Summary of Super Calculus

01 Gamma Function & Digamma Function

Although the factorialn! and the harmonic number Hn ( = 1+1/2++1/n ) are usuallydefined for a natural
number, if a gamma function and a digamma function are used, these can be defined for the real number p . That is,
p! = (1+ p) , Hp =  (1+ p)+ ( =0.57721)
The former is convenient to express the coefficient of the higher order primitive or derivative of a power function, and the
latter is indispensable to non-integer order calculus of the logarithmic function.
Although some formulas about these functions are described here, the following two formulas proved in Section 3 are
especially important. That is, when m , n =0, 1, 2, 3,  ,
(-n) m!  (-n)
= (-1)m-n , = (-1)n+1n!
(-m) n! (-n)
These show that the ratios of the singular points of (z) or (z) reduce to integers or its reciprocals. The former is
necessary to express the higher order derivative of fractional functions, and the latter is required for the super calculus
of a logarithmic function.

02 Multifactorial
The relational expression of multifactorial and the gamma function is shown here.
For instance, in case of double factorial,

     / 
1 1 1
(2n -1)!! = 2n n + / = 2n n +
2 2 2

    = 2 (n +1) = 2 n !
2 2
(2n)!! = 2n n + / n n
2 2
These are used to express the half-integration of a integer-power function later.
Moreover, we obtain the following Maclaurin expansions as by-products.

f (0) 0 f '(0) 1 f " (0) 2 f 3 (0) 3

( )

 2 =
x
f x + x + x + x +
0!! 2!! 4!! 6!!
f (0) 0 f '(0) 1 f " (0) 2 f 3 (0) 3
( )
f  =
x
x + x + x + x +
3 0!!! 3!!! 6!!! 9!!!

03 Generalized Multinomial Theorem

First, the binomial theorem and a generalized binomial theorem are mentioned. The Leibniz rule and the Leibniz rule
about super-differentiation are expressed just like these later.

Next, multinomial thorem and generalized multinomial thorem are shown as follows.

Theorem 3.3.1
For real numbers x1 , x2 ,  , xm and a natural number n , the following expressions hold.

 r  r  r 
n r1 rm -2 n r1 rm-2 rm-2-rm-1 rm-1
x1+ x2++xm = Σ Σ  Σ  x2  xm-1
n n -r1 r1-r2
x1 xm
r1=0 r2=0 rm -1=0 1 2 m-1
(1.1)
r1+ r2+ + rm-1
 r + r + + r     
n n n n rm-2 + rm-1
= Σ Σ Σ 
r1=0 r2=0 rm -1=0 1 2 m-1 r2+ + rm-1 rm-1
n -r1- - rm-1 r
 x1 x21 x32  xmm-1
r r

Theorem 3.4.1
The following expressions hold for real numbers  and x1 , x2 ,  , xm s.t.x1 x2+ x3++ xm .

-1-
 
 r  r  r 
  r1 r m -2 r1 rm-2  -r1 r1-r2 r -rm-1 rm-1
x1+ x2++xm = Σ Σ  Σ   xm-1
m-2
. x1 x2 xm
r1=0 r2=0 rm -1=0 1 2 m-1
(1.1)
 r1+ r2+ + rm-1
 r + r + + r     
   rm-2 + rm-1
= Σ Σ Σ 
r1=0 r2=0 rm -1=0 1 2 m-1 r2+ + rm-1 rm-1
 -r1- - rm-1 r1 r2 r
 x1 x2 x3  xmm-1
(1.2)
Where, x 1 x 2+ x 3++ x m is allowed at  >0 .

Higher order calculus of the product of many functions and super order calculus of the product of many functions are
expressed just like these later.

What should be paid attention here are the following property of generalized binomial coefficients.

r 
n  n
Σ
r=0
nＣ r = Σ
r=0

That is, once generalized binomial coefficients was used, the upper limit of  should be . Therefore, when n is
a natural number and p is a real number, the following holds in most cases.

 r  f(p,x)
n  p
n p 　 Σ nＣ r f(n,x)  Σ
r=0 r=0
When the original coefficient is not binomial coefficient e.g. 1 ,

 r  f(p,x)
n  p
n p 　 Σ nＣ r f(n,x)  Σ
r=0 r=0

 r  f(p,x)
p  p
Although Σ
r=0
1f(p,x) is difficult, Σ
r=0
is satisfactory. What enables super calculus in this text

is just this property of the generalized binomial coefficient. Newton is great !

04 Higher Integral

(1) Definitions and Notations

The 1st order primitive function of f (x ) is usually denoted F(x ). However, such a notation is unsuitable for the
description of the 2nd or more order primitive functions. Then, f
<1>
(x), f <2>(x),  , f <n>(x) denote the each order
<1>
f (x ) in this text. Here, for example, when f (x )= sin x , f
primitive functions of (x) might mean -cos x or
might mean -cos x + c . Which it means follows the definition at that time.
Furthermore, each order integrals of f (x ) are denoted as follows.

    f(x)dx
x x x x x
f(x)dx 1 , f(x)dx 2 ,  ,  n
a1 a2 a1 an a1
And these are called higher integral with variable lower limits . On the other hand,

 f(x)dx ,  f(x)dx  f(x)dx

x x x x x
1 2
, , n
a a a a a
are called higher integral with a fixed lower limit .

(2) Fundamental Theorem of Higher Integral

There is Fundamental Theorem of Calculus for the 1st order integral. The same theorem holds for the higher order
integral.

-2-
Theorem 4.1.3
r =0, 1,  , n
<r> <r+1>
Let f be continuous functions defined on a closed interval I and f be the arbitrary

ar , x  I .
<r>
primitive function of f . Then the following expression holds for

 
x x n -1 x x
 f(x)dx n = f (x) - Σ f -r a n-r 
<n > <n >
dx r
an a1 r=0 an a n- r+ 1

Especially, when ar = a for r =1, 2,  ,n ,

 f(x)dx
x x
<n > n -1
<n -r> (x -a)r
n
=f (x) - Σf (a)
a a r=0 r!

(3) Lineal and Collateral

We call Constant-of-integration Polynomial the 2nd term of the right sides of these. And when Constant-of-integration
Polynomial is 0, we call the left side Lineal Higher Integral and we call f <n>(x) Lineal Higher Primitive Function.
Oppositely, when Constant-of-integration Polynomial is not 0, we call the left side Collateral Higher Integral and we
call the right side Collateral Higher Primitive Function.
For example,

 
x x x Left: Lineal 3rd order integral
3
sin xdx = cosx
3 2  1 Right: Lineal 3rd order primitive
2 2 2


x x x
3x2 Left: Collateral 3rd order integral
sin xdx = cosx + -1 Right: Collateral 3rd order primitive
0 0 0 2
for r =1, 2, , n
<r>
Furthermore, from Theorem 4.1.3, we see that ar must be all zeros of f in order for the
higher integral of f (x ) to be lineal.

(4) Higher Integral and Reimann-Liouville Integral

The higher integral with a fixed lower limit reduce to the 1st order integral which called Reimann-Liouville Integral.

Theorem 4.2.1 ( Cauchy formula for repeated integration )

When f (x ) is continuously integrable function and (z) denotes a gamma function ,

  (x -t)
x x 1 x
 f(x)dx n =
n-1
f(t)dt
a a (n) a

Reimann-Liouville Integral of the right side is important. However, in the higher integral, since the left side itself has
an operation functions, the right side is not indispensable.
By the way, replacing the left side in Theorem 4.1.3 for Reimann-Liouville Integral and shifting the index by -n ,
we obtain the following.

 (x -t)
(x -a)r
n -1
(r) 1 x
n-1 (n )
f
<0>
(x) = Σf (a) + f (t)dt
r=0 r! (n) a
(n)
This is the Taylor expansion of f (x ) around a . And Reimann-Liouville Integral of f is the remainder term called
Bernoulli form

(5) Higher Integrals of Elementary Functions.

When m is a natural number, the 2nd order integral of xm becomes as follows.

 x dx
x x 1 m!
m 2 m+2 m+2
= x = x
0 0 (m +1)(m +2) (m +2)!

-3-
Then, when  is a positive number, it is as follows.
!
 x dx
x x
 2 
= x +2
0 0 ( +2)!
Here, ! can be expressed by gamma function (1+) . Thus
(1+ ) +2
 x dx
x x
 2
= x
0 0 (1+ +2)
By such an easy calculation, we obtain the following expressions for elementary functions. Where, ,  denote the
ceiling function and the floor function respectively.

Higher Integrals of Power Function etc.

(1+ )
  x dx = 1+ + n  x
x x
 +n
 n
(  0)
0 0
(- - n)
  x dx = (-1) (-) x
x x
 +n
n n
( <-n)
 

  e dx = (1) e
x x
x n n x
 

 log x dx = n ! log x -Σ k 

x x n
x 1 n
n
x>0
0 0 k =1

n
    2
x x x
 sin xdx = sin x - n
n 2 1
2 2 2
n
 
x

 
x x
 cosxdx n = cos x -
(n -1) 1 0 2
2 2 2

  
e x - (-1)ne
x x x -x
 sinh xdx n =
n i 2 i 1 i 2
2 2 2

   e x + (-1)ne -x
x x x
 cosh xdx = n
(n -1) i 1 i 0 i 2
2 2 2

Higher Integrals of Inverse Trigonometric Functions


x x tan -1x n /2
tan x dx = -1
Σ (-1)k nＣ n-2k x n-2k
n
0 0 n ! k=0
log1+ x 2 n /2
+ Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k=1

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+n)- (2r)x n+1-2r
n! r=1


x x xn tan -1x n /2
 (-1)k nＣ n-2k x n-2k
n! Σ
-1 n -1
cot x dx = cot x -
0 0 n! k =1

logx +1 n /2

2
- Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k=1
1 n /2
+ Σ (-1)r nＣ n+1-2r  (1+n)- (2r)x n+1-2r
n! r=1

-4-
 n-2r


x x n/2 x
 sin x dx = Σ
-1 n
2 sin -1x
an a1 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
+ Σ Σ (-1) n -2r+1Ｃ s
s
2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
Where, a1 = i 1.508879 , a2 = 0 , a3 = -i 0.475883 , a4 = 0 , 
n-2r


x x xn n/2 x
 cos x dx = cos x - Σ
-1 -1 n
2
-1
sin x
an a1 n! r=1 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1
s (s -1)!!2 x
- Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
Where, a1 = 1 , a2 = 0 , a3 = ? , a4 = 0 , 
n-2r-1


x x xn (n -1)/2 (2r -1)!! x
-1
sec xdx = n
sec x - Σ
-1
logx + x 2-1
1 1 n ! r=0 (2r)!!( 2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 x 2-1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
n-2r-1


x x xn (n -1)/2 (2r -1)!! x
 csc x dx = csc x + Σ
-1
-1 n
logx + x 2-1
an a1 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
n /2 n -2r Ｃs
n -2r (s -1)!!2 x
-Σ Σ (-1)s x 2-1
r=1 s=0 2r +s (s +2r -1)!!2 (n -2r)!
Where, a1 , a2 ,  , an are all complex numbers.

Higher Integrals of Inverse Hyperbolic Functions


x x tanh -1x n /2 n-2k log1- x 2 n /2 n+1-2k
n
Σ nＣ n-2k x Σ nＣ n+1-2k x
-1
tanh x dx = +
0 0 n ! k =0 2n ! k =1

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
n! Σ
-
r=1


x x xn n tanh -1x n /2 n-2k
Σ nＣ n-2k x
-1 -1
coth x dx = coth x +
0 0 n! n ! k=1
log1-x 2 n /2 n+1-2k
+
2n ! k=1 Σ nＣ n+1-2k x

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
n! Σ
- |x|<1
r=1

(-1)r x n-2r

x x n/2
 sinh x dx = Σ
-1 n
2 sinh -1x
an a1 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1
r+s (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1+x 2
r=1 s=0 (s +2r -1)!!
Where, a1 = 1.508879 , a2 = 0 , a3 = -0.475883 , a4 = 0 , 
n-2r


x x n/2 x
 cosh -1x dx n = Σ 2 cosh -1x
1 1 r=0 (2r)!! (n -2r)!

-5-
 n-2r+1
n /2 n -2r+1
s (s -1)!!2 x
-Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)! x
2
-1
r=1 s=0 (s +2r -1)!!


n-2r-1
x x xn (n -1)/2 (2r -1)!! x
 sech x dx = sech x + Σ
-1
-1 n
sin -1x
an a1 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
n /2 n -2r n -2r Ｃs
(s -1)!!2 x
+Σ
r=1
Σ (-1)s
s=0 2r +s (s +2r -1)!!2 (n -2r)!
1- x 2

Where, a1 = a3 = a5 =  = 0 , a2 , a 4 , a 6 ,  are complex numbers.

n-2r-1


x x xn (n -1)/2 (-1)r(2r -1)!! x
 csch -1x dx n = csch -1x + Σ sinh -1x
an a1 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
r+s n -2rＣ s (s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 x 2+1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
Where, a1 = 0 , a2 = 0.6079 , a3 = 0 , a4 = 1.5539 , 

(6) Termwise Higher Integral and Taylor series of higher primitive function

Theorem 4.6.1
r =0, 1,  , n
<r> <r +1>
Let f be continuous functions defined on [a , b ] and f be the arbitrary primitive function
<r>
off . At this time, if f (x ) can be expanded to the Taylor series around a, the following expressions hold for
x [a , b] .
<n > n -1
<n -r> (x -a)r  (r) (x -a)n+ r 
<n -r> (x -a)r
f (x) =Σf (a) +Σ f (a) =Σf (a)
r=0 r! r=0 (n + r)! r=0 r!
<n>
This expression shows that the Taylor series of f consists of the constant-of-integration polynomial and the
termwise higher integral of f (x ). The following can be said from this.

(1) A termwise higher integral with a fixed lower limit is collateral generally.
(2) It is the following case that a termwise higher integral with a fixed lower limit is lineal.
(a)=0 for r =1, 2, ,n (a) 0 ,  for at least one s  0
<r> (s)
f & f
For example,


x x (x -a)n+ r 
 e dx x n
a -
= Σ ea is collateral higher integral.
a a r=0 (n + r)!


x x  (2r)!
 tan -1x dx n = Σ(-1)r x
n+2r+ 1
is lineal higher integral.
0 0 r=0 ( n +2r +1)!

Next, the following is obtained from the Taylor series of the higher integral of log x .
n-1
 1 (-1) n -1

H 
n 1
Σ
r=1 r(r +1)(r + n )
=
n! Σ
r=0
(-1)rnＣ r Hn-r n =Σ
k =1 k

 (-1)r-1 2n 1 n -1
Σ
r=1 r(r +1)(r + n)
=
n! 
log 2 -H n +
n! Σ
r=0
nＣ r Hn-r

n -1 (-1)n-1
Σ
r=0
(-1) nＣ r Hn-r r
=
n

-6-
 
(1- x)p

1  p
dx = 2plog 2-  (1+p )-+Σ  (1+r)+
0 1+ x r=1 r

Example
1 1 1 1 1
+ + + + =
1234 2345 3456 4567 18
1 1 1 1 4 8
- + - +- = log 2 -
1234 2345 3456 4567 3 9

05 Termwise Higher Integral (Trigonometric, Hyperbolic)

In this chapter, for the function which second or more order integral cannot be expressed with the elementary functions
among trigonometric functions and hyperbolic functions, we integrate the series expansion of these function termwise
and obtain the following expressions. Where, ,  denote the ceiling function and the floor function respectively. And
Bernoulli Numbers and Euler Numbers are as follows.
1 1 1 1 5
B0 =1, B2= , B4=- , B 6= , B8=- , B10= ,
6 30 42 30 66
E0 =1, E2 =-1, E4 =5, E6 =-61, E8 =1385, E10=-50521, 

(1) Taylor Series

22k 22k -1B2k 2k+ n-1 
  tan x dx
x x 
n
=Σ x |x|<
0 0 k=1 2k (2k + n -1)! 2
22k 22k -1B2k 2k+ n-1 
 
x x 
 tanh x dx = Σ n
x |x|<
0 0 k=1 2k (2k + n -1)! 2

 
x x  E2k
 secx dx n =Σ x 2k+
n
|x|<
0 0 k =0 (2k + n )! 2

(2) Fourier Series

 (-1)k-1  (-1)k
When (x)=Σ ,  (n)=Σ ,
k =1 kx k=0 (2k +1)n
n
 
x x (-1)k-1
 
1 
 tan x dx =n
n-1 Σ sin 2k x -
0 0 2 k=1 kn 2
n/2 (2k -1) x
n+1-2k

+ Σ (-1)k-1 |x|<
k=1 22k-2 (n +1-2k)! 2

 
x x (-1)n-1  e -2k xn
x
 tanh x dx n = Σ(-1)
k-1
+
0 0 2n-1 k =1 kn n!
n -1 (k +1) x n-1- k
- Σ(-1)k x >0
k=0 2k (n -1- k)!
n
 
(-1)k
 
x x 
sec x dx = 2Σ
n
n cos (2k +1)x -
0 0 k=0 (2k +1) 2
n /2 2 (2k) n-2k 
+ Σ (-1)k-1 x |x|<
k=1 (n -2k)! 2

-7-
  
)x
x x  e -(2k +1
 sech x dx n = (-1)n2Σ(-1)k x >0
  k=0 (2k +1)n
(3) Riemann Odd Zeta & Dirichlet Odd Eta
Comparing Taylor series and Fourier series, we obtain Riemann Odd Zeta and Dirichlet Odd Eta. For example,

 2n+1
Σ
2 -1B2k

2k
(-1)n 22n 
 (2n +1) =  2k+2n+1 - log 2
 22n-1 k=1 2k (2k +2n +1) ! (2n +1) !
(-1)n 22n n -1  2n+1-2k 22k -1
- Σ(-1)k (2k +1)
 22n-1 k =1 (2n +1-2k) ! 22k
 2n+1
 (2n)!  
 B2k

1 2n +1 1
(2n +1) = (-1)n log - Σ -Σ  2k+2n+1
 k=1 k k=1 2k (2k +2n +1) !
1 n -1  2n-2k+1
(2k +1)
Σ
+ (-1)n (-1)k-1
k=1 (2n -2k +1) !
2n-1 n-1- j

 
e -k n -2 
j-1 2 -1 1
 (n) = n-1 Σ (-1) n
+ Σ(-1) n-1- j
k-1
 (n -j)
2 -1 k=1 k j=1 2 j!
(-2)n-1  2 -1B2k
 
2k
1 1 log 2
- n-1 Σ - +
2 -1 k=1 2k (2k + n -1)! 2 n! (n -1)!
(-1)n-1  E2k  2k+2n
 (2n -2k)  2k

 2  2
n
 (2n) = Σ +Σ(-1) k-1
2 k=0 (2k +2n ) ! k=1 (2k) !

06 Termwise Higher Integral (Inv-Trigonometric, Inv-Hyperbolic)

Here, we integrate the series of an inverse trigonometric function or an inverse hyperbolic function term by term.
Then, we obtain formulas simpler than what were obtained in "04 HIgher Integral". Moreover, both are compared and
we obtain various by-products.

(1) Taylor Series


x x  (2k)!
 tan -1x dx n = Σ(-1)k x 2k+
n+1
|x|< 1
0 0 k=0 (2k + n +1)!
 xn

x x  (2k)!
 cot -1x dx n = - Σ(-1)k x 2k+
n+1
0< x 1
0 0 2 n! k=0 (2k + n +1)!

 sin
x x 2
 (2k -1)!! n+1
-1
x dx = Σ
n
x 2k+ |x|< 1 collateral
0 0 k=0 (2k + n +1)!
 xn

x x 2
 (2k -1)!!
 cos -1x dx n = -Σ x 2k+
n+1
|x|< 1 collateral
0 0 2 n! k=0 (2k + n +1)!

 tanh
x x  (2k)! n+1
-1
x dx n = Σ x 2k+ |x|< 1
0 0 k=0 (2k + n +1)!

 sinh
x x 2

k (2k -1)!! n+1
-1
x dx = Σ(-1)
n
x 2k+ |x|< 1 collateral
0 0 k =0 (2k + n +1)!


xn 2

 
x x 2 n 1  (2k -1)!!
 -1
sech x dx = n
log +Σ -Σ x 2k+n
0 0 n! x j=1 j k=1 2k (2k +n)!
0< x <1 collateral

-8-
 
xn 2

 
x x 2 n 1  (2k -1)!!
 -1
csch x dx = n
log +Σ -Σ(-1)k x
2k+n
0 0 n! x j=1 j k=1 2k (2k +n)!
0< x <1 collateral

(2) By-products


1 (1- x)n n /2
nＣ n+1-2r  (1+ n )- (2r)
n-1
0 1- x 2
dx = 2 log 2 - Σr=1


1 (1- x)n  n /2 log 2 n /2
0 1+ x 2 dx =
4 Σk=0
(-1)k
nＣ n-2k +
2 Σk=1
(-1)k nＣ n+1-2k
n /2
- Σ (-1)r nＣ n+1-2r  (1+n)- (2r)
r=1
n-1
 2
(2k)! 1 n /2
Σ nＣ n+1-2r  (1+ n )- (2r)
n! Σ
log 2 - =
k=0 (2k + n +1)! n! r=1
 (2k)!  n /2 log 2 n /2
4 n! Σ 2n! Σ
k k
Σ (-1) = ( -1 ) nＣ n-2k + (-1)k nＣ n+1-2k
k=0 (2k +n +1)! k=0 k =1

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)
!
n r=1
2n -1Ｃ n-1
2
 (2k -1)!! n /2 1
Σ =  -Σ 2
k=0 (2k + n +1)! (2n)!! k=1 (2k -1)!! (n -2k +1)!

 
 (2k -1)!! 1 n 1 n /2 1 1
Σ 2k (2k +n)!
=
n!
log 2+Σ
j=1 j
+Σ 2 (n -2r)!
k=1 2r(2r -1)!!
r=1

 (n -1)/2 (2r -1)!! 1

- Σ
2 r=0 (2r)!!(2r +1)! (n -2r -1)!
Example
1 1 1 1 2 5
+ + + + = log 2 -
1234 3456 5678 78910 3 12
1 1 1 1 5  log 2
- + - +- = - -
1234 3456 5678 78910 12 12 6
(-1)!!2 1!!2 3!!2 5!!2 
+ + + + = -1
2! 4! 6! 8! 2
1!!2 3!!2 5!!2 7!!2 1 
+ + + + = (log 2+1)-
23! 45! 67! 89! 1! 2

07 Super Integral (Non-integer order Integral)

p m
It is Σ
j=0
a = ap which extended the domain of index j of Σ
j=1
a to the real number interval [0 , p ] from the natural
q n
number interval [1 , m ]. And it is Πb = b q which extended the domain of
k=0
index k of Π
k =1
b to the real number
interval [0 , q ] from the natural number interval [1 , n ]. It is called analytic continuation to extend the domain
generaly. Although usually analytic continuation is used for extending the domain of a function, it can be used also
for extending the domain of the index of a operator. Here, extending the domain of the index of the integration operator,
we obtain Super Integral (Non-integer order Integral ).

-9-
(1) Definitions and Notations
<p>
f (x) denotes the non-integer order primitive function of f (x ). And we call this Super Primitive Function of
f(x). Since there is a super primitive function innumerably, which f p (x) means follows the definition at that time.
< >

We call it Super Integral to integrate a function f with respect to an independent variable x from a (0) to a (p )
continuously. And it is described as follows.

    
x x


x x
 p
f(x)dx =  f(x)dx  dx
a(p ) a(0) a(p ) a(0)
And
when a(k)= a for all k [0 , p ] , we call it super integral with a fixed lower limit ,
when a(k) a for some k [0 , p ] , we call it super integral with variable lower limits .

(2) Fundamental Theorem of Super Integral

r [0 , p ] be an continuous function on the closed interval I and be arbitrary the r -th order primitive
<r>
Let f
function of f . And let a (r ) be a continuous function on the closed interval [0 , p ].
Then the following expression holds for a(r ), x  I .

   
x x p -1 x x
 <p > <p >
f(x)dx p = f (x) - Σ f -r a(p -r)  dx r
a(p ) a(0) r=0 a(p ) a(p -r+1)
Especially, when a(r)= a for all k [0 , p ] ,


x x p -1 (x -a)r
 f(x)dx n = f
<p >
(x) - Σf <p -r>
(a)
a a r=0 (1+ r)

(3) Lineal and Collateral

We call Constant-of-integration Function the 2nd term of the right sides of these.

And when Constant-of-integration Function is 0, we call the left side Lineal Super Integral . and we call f <p>(x)
Lineal Super Primitive Function.
Oppositely, when Constant-of-integration Function is not 0, we call the left side Collateral Super Integral and we call
the right side Collateral Super Primitive Function.
For example,
p
 
x

 
x Left: Lineal p th order integral
p
sin xdx = sin x -
p 0 2 Right: Lineal p th order primitive
2 2


x x  (-1)r Left: Collateral p th order integral
 sin xdx p =Σ x 2r+1+
p
0 0 r=0 (2r +2+ p) Right: Collateral p th order primitive

(4) Super Integral and Reimann-Liouville Integral

The super integral with a fixed lower limit reduce to the 1st order integral which called Reimann-Liouville Integral.

 f(x)dx  (x -t)

x x 1 x
p p-1
= f(t)dt
a a (p) a
This is what extended the parameter n to the real number in Cauchy formula for repeated integration. Since the left
side has lost the operating function, Reimann-Liouville Integral of the right side is very important. All the super integral
with a fixed lower limit can be verified numerically by this. On the other hand, since the super integral with variable
lower limits cannot apply Reimann-Liouville Integral, the verification is vary difficult.

(5) Fractional Integral & Super Integral

In traditional Fractional Integral , the super primitive function is drawn from Riemann-Liouville Integral.

For example, in the case of f(x )= x , it is as follws.

- 10 -
Let

 (x -t)p-1
f(t) = t , g(x -t) =
(p)
Then

  f(t)g(x -t)dt
<p > 1 x x
x  t (x -t)
p-1
= dt =
(p) 0 0
(f g )(x) . Then we take the Laplace transform of (f g)(x)
We find out that this is a convolution

(f g)(x)  F(s)G(s)
( +1)
f(x) = x   
= F(s)
s +1
x
p-1
1 (p) 1
g(x) =  = = G(s)
(p) (p) s p sp
( +1) 1 ( +1) ( +p +1)
 F(s)G(s) = =

s +1 s
p ( +p +1) s +p+1
( +1) a+p
Finally, taking the inverse Laplace transform, we obtain x .
( +p +1)
Because the technique of Fractional Integral is difficult like this, it is more difficult to obtain the super primitive
function of log x by this technique.
Above all, the problem is that Fractional Integral cannot treat the lineal non-integer order integral such as sin x.
Because Riemann-Liouville Integral cannot be applied to the integral with variable lower limets.
On the other hand, in Super Integral that I advocates, first of all, we obtain the following higher integral.
(1+ ) +n
 x
x x

 dx n = x (  0)
0 0 1+ + n 
And replacing the index of the integration operator n with a real number p , we obtain the following very easily.
(1+ ) +p
 x
x x

 dx p = x (  0)
0 0 1+ + p 
Furthermore, performing the higher integral with variable lower limits and replacing the index of the integration operator
with a real number, we can obtain the super integral such as sin x easily.

(6) Super Integrals of Elementary Functions.

In this way, the following super integrals obtined from " 04 Higher Integral " ,
(1+ ) +p
  x dx =  (1+ + p ) x
x x
 p
(  0)
0 0
(- - p)
 
x x
 +p
 x dx = (-1) p
x p
( <-p)
  (- )

  e dx = (1) e
x x
x p p x
 
log x - (1+ p) - 
  log xdx = (1+p)
x x
p
xp
0 0

p
 
x

 2
x
 sin xdx = sin x - p
p 0
2 2
p
 
x

 
x
 cosxdx p = cos x -
(p -1) -1 2
2 2

- 11 -
  
x x e x - (-1)pe -x
 sinh xdx p
=
p i 0 i 2
2 2

  e x + (-1)pe -x
x x
 cosh xdx p
=
(p -1) i -1 i 2
2 2


tan -1x 
 
x x p
 tan -1x dx p =
Σ (-1)k x
p-2k
0 0  (1+ p ) k =0 p -2 k
log1+ x 2 
 
p p+1-2k
+ Σ (-1)k x
2(1+ p) k=1 p +1-2k

 
1  p
 (1+ p )- (2r)x
p+1-2r
-
(1+ p) r=1 Σ (-1)r
p +1-2r


p -1

 
x x x tan x  p
 cot x dx =
-1 p -1
cot x - Σ (-1)k x
p-2k
0 0 (1+ p) (1+ p) k=1 p -2k
log1+ x 2 
 
p p+1-2k
-
2(1+ p) k=1 Σ (-1 )k
p +1-2k
x

 
1  p
 (1+ p )- (2r)x
p+1-2r
+ Σ
(1+ p) r=1
(-1)r
p +1-2r


tanh -1x 
 
x x p
 tanh -1x dx p = Σ x
p-2k
0 0  (1+ p ) k=0 -2
p k
log1- x 2 
 
p p+1-2k
+ Σ
2(1+ p) k=1 p +1-2k
x

 
1  p
 (1+ p )- (2r)x
p+1-2r
- Σ
(1+ p) r=1 p +1-2r

(7) By-products
(-1)k (-1)k (1+ n , 1/m)
k  k 
n n (2n )!! n n
Σ
k=0 2k +1
=
(2n +1)!!
, Σ
k=0 mk +1
=
m
q+p
(-1)r r

r r 
 q -1 q <p > x  (-1) q
(q , p) = Σ x 
(p) Σ
, =
r=0 p+r r=0 p + r

Where, B (x,y) is the beta function.

08 Termwise Super Integral

The following termwise super integrals are obtained from " 05 Termwise Higher Integral (Trigonometric,Hyperbolic) "
and " 06 Termwise Higher Integral (Inv-Trigonometric, Inv-Hyperbolic) ".
Where, ,  denote the ceiling function and the floor function respectively. And Bernoulli Numbers and Euler Numbers
are as follows.
1 1 1 1 5
B0 =1, B2= , B4=- , B 6= , B8=- , B10= ,
6 30 42 30 66
E0 =1, E2 =-1, E4 =5, E6 =-61, E8 =1385, E10=-50521, 

- 12 -
 22k 22k -1B2k 
 tan xdx
x x 
p-1
p
=Σ x 2k+ 0< x <
0 0 k =1 2k (2k + p) 2
22k 22k -1B2k 
  tanh xdx
x x 
 p
=Σ x 2k+
p-1
0< x <
0 0 k =1 2k (2k + p) 2

 secxdx
x x  E2k
p
=Σ x 2k+p 0< x < collateral
0 0 k =0 (2k +p +1) 2

 sech xdx
x x
p
 E2k
=Σ x 2k+p 0< x < collateral
0 0 k =0 (2k +p +1) 2
22k 22k -1B2k  
 cotxdx
x x 2k+ p-1

 

p
= -Σ x- < x< 
  k=1 2k (2k + p) 2 2
2 2

 

E2k
x 2k+p

 
x 
 cscxdx p =Σ x- < x<  collateral
  k=0 (2k +p +1) 2 2
2 2


)x
x x e -(2k +1 
 csch x dx = (-1) 2Σ p
p
p
x>0
  k=0 (2k +1)


)x
x x  e -(2k +1
 sech x dx p = (-1)p2Σ(-1)k x>0
  k=0 (2k +1)p


x x  (2k)!
 tan -1xdx p = Σ(-1)k x 2k+p+1 0< x <1
0 0 k =0  (2 k + p +2 )

  xp
x x  (2k)!
 cot -1xdx p = -Σ(-1)k x 2k+p+1 0< x <1
0 0 2 (1+ p) k=0 (2k + p +2)

 sin xdx

x x 2
 (2k -1)!!
p
-1
=Σ x 2k+p+1 0< x <1 collateral
0 0 k =0 (2k +p +2)

 cos xdx
xp
x x 2
 (2k -1)!!
p
-1
= -Σ x 2k+p+1 0< x <1 collateral
0 0 2 (1+ p) k=0 (2k + p +2)

 tanh xdx

x x  (2k)!
p p+1
-1
=Σ x 2k+ 0< x <1
0 0 k =0 (2 +
k p +2)

 sinh xdx

x x 2
 (2k -1)!!
p p+1
-1
= Σ(-1)k x 2k+ 0< x <1 collateral
0 0 k =0  (2 +
k p +2 )

 sech xdx

x x xp 2

 
2  (2k -1)!!
-1 p
= log + (1+ p) + -Σ x 2k+p
0 0 (1+ p) x k=1 2k (2 k + p +1)
0< x <1 collateral


xp
x x 2

 
2 
k (2k -1)!!
-1
csch xdx = log + (1+ p)+ -Σ(-1)
p
x 2k+p
0 0 (1+ p) x k=1 2k (2k + p +1)
0< x <1 collateral

09 Higher Derivative

(1) Definitions and Notations

When f (x) denotes the derivative function of f (n-1)(x) for n =1, 2, 3,  ,
(n)
we call f
(n)
(x) Higher Derivative

- 13 -
of f(x)..
Moreover, we call it Higher Differentiation to differentiate a function f with respect to an independent variable x
repeatedly. And it is described as follows.
n

      
d d d d d d
f(x) =  f(x)  : n pieces
dx n dx dx dx dx dx

(2) Fundamental Theorem of Higher Differentiation

The following theorem holds from Theorem 4.1.3 .

r =0, 1,  , n
(r)
When f are continuous functions on a closed interval I and are the r th derivative functions of f,
the following expression holds for x I .
dn
( ) = f n (x)
( )
nf x
dx
This theorem guarantees that only the lineal exists in the higher differentiation.

(3) Higher Derivative of Elementary Functions

When m is a natural number, the 2nd order derivative of x m becomes as follows.
m (2) m!
x  = m(m -1)x m-2 = x
m-2
(m -2)!
Then, when  is a positive number, it is as follows.
! 
x 
(2)
= x -2
( -2)!
Here, ! can be expressed by gamma function (1+) . Thus
(1+ ) +2
x 
(2)
= x
(1+ -2)
By such an easy calculation, we obtain the following expressions for elementary functions.
(n ) (1+ ) -n
x   = x (  0)
1+ - n
(- + n) -n
= (-1)-n x ( < 0)
(- )
 x (n )
e  = (1)-ne  x
(log x) n = (-1)n-1(n -1)! x -n
( )

n
(sin x) n  
()
= sin x +
2
n
(cosx) n  
( )
= cos x +
2
(n ) e x - (-1)-ne -x
(sinh x) =
2
e x + (-1)-ne -x
(cosh x) n =
( )
2
etc.

(4) Higher Derivative of Inverse Trigonometric Functions

When ,  denote the ceiling function and the floor function and n is a natural number,

- 14 -
 (n ) (n -1)! n /2
n+1-2r
tan x = (-1)n nΣ
(-1)rnＣ n+1-2r x
-1

x +1
2 r=1

(n ) (n -1)! n /2
cot x
-1
= (-1)n-1 nΣ
(-1)rnＣ n+1-2r x n+1-2r

x +1
2 r=1

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n ) n/2 n -1
sin x Σ
-1
= 1
r=0 n -1-2r n-r-
2
1- x 
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n ) n/2 n -1
cos x = -Σ
-1
1
r=0 n -1-2r n-r-
2
1- x 
2

etc.

(5) Higher Derivative of Inverse Hyperbolic Functions

When ,  denote the ceiling function and the floor function and n is a natural number,
(n ) (n ) (n -1)! n /2
n+1-2r
tanh
-1
x = coth -1x = (-1)n Σ Ｃ n+1-2r x
n r=1 n
x -1
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n ) n/2 n -1
sinh
-1
x = (-1)n-1 Σ (-1)r 1
r=0 n -1-2r n-r-
2
x +1
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n ) n/2 n -1
cosh
-1
x = (-1)n-1 Σ (-1)r 1
r=0 n -1-2r n-r-
2
x -1
2

etc.

(6) By-Products
n /2 n
Σ (-1)k nＣ n+1-2k = -2n/2 sin
k=1 4
 k +(n -1)! (n -1)!
Σ
k=0
(-1)k
k!
=
2n

Example
-5Ｃ 4 + 5Ｃ 2 - 5Ｃ 0 =4
1
12 - 23 + 34 - 45 +-  =
4

10 Termwise Higher Derivative (Trigonometric, Hyperbolic)

In this chapter, for the function which second or more order integral cannot be expressed with the elementary functions
among trigonometric functions and hyperbolic functions, we differentiate the series expansion of these function termwise
and obtain the following expressions. Where, ,  denote the ceiling function and the floor function respectively. And
Bernoulli Numbers and Euler Numbers are as follows.
1 1 1 1 5
B0 =1, B2= , B4=- , B 6= , B8=- , B10= ,
6 30 42 30 66
E0 =1, E2 =-1, E4 =5, E6 =-61, E8 =1385, E10=-50521, 

- 15 -
(1) Taylor Series

(n )  22k 22k -1B2k 2k- n-1 

(tan x) = Σ x |x|<
k=
n +1

2k(2k - n -1)! 2
2

(n )
22k 22k -1B2k 2k- n-1
 
(tanh x) = Σ x |x|<
n +1 2k(2k - n -1)! 2

k=
2

n!  22k B2k
(cot x)
(n )
= (-1) n
n+1
-Σ x 2k-n-1 0< x < 
x n +1

2 2 - -1
k( k n ) !
k=
2

n!  22k B2k
(coth x)
(n )
= (-1) n
n+1
-Σ x 2k-n-1 0< x < 
x n +1 2k(2k -n -1)!
k= 
2

n!  22k -2B2k 2k- n-1

0< x < 
(n ) n
(csc x) = (-1) n+1
+ Σ x
x n +1 2k (2k - n -1)!
k= 
2

n! 22k -2B2k
(csch x)
(n )
= (-1) n
n+1
+ Σ x 2k-
n-1
0< x < 
x n +1

2 2 -
k( k n )-1 !
k=
2

 E2k 
(sec x) n =
( )
Σ
n +1 (2k -n)!
x 2k-n |x|<
2
k= 
2

 E2k 
(sech x) n =
( )
Σ
n +1 (2k -n)!
x 2k-n |x|<
2
k= 
2

(2) Fourier Series


(tanh x) n = (-1)n-12n+1Σ(-1)k-1k ne -2kx
()
x>0
k=1

(coth x) n = (-1)n2n+1Σk ne -2kx
( )
x>0
k =1

(csch x) n = (-1)n2Σ(2k +1)n e -(2k +1)x
( )
x>0
k=0

(sech x) n = (-1)n2Σ(-1)k (2k +1)n e -(2k +1)x
( )
x>0
k=0

(3) Dirichlet Odd Eta (minus) & Even Beta (minus)

Comparing Taylor series and Fourier series, we obtain Dirichlet Odd Eta (minus) and Even Beta (minus) .

(-1)n-1 22k 22k -1B2k  2k-2n

 4

(1-2n) = Σ
24n-1 k =n 2k(2k -2n)!

(2n -1)! + Σ 
(-1)n-1 2n 22k B2k  2k

  4
4 
(1-2n) =
24n-1 k=n 2k(2k -2n)!
22n -1B2n
(1-2n) =
2n

- 16 -
 (-1)n 22k 22k -1B2k  2k-2n-1

 4

(-2n) = Σ
22n+1 k=n +1 2k(2k -2n -1)!

 
(-1)n 2n 22k B2k  2k

  4
4 
(-2n) = (2n)! - Σ
22n+1 k=n +1 2k(2k -2n -1)!
E2n
(-2n) =
2
(4) Other by-products
1n 2n 3n 4n  B2k
+ + + + = n! + (-1)n Σ
e1 e2 e3 e4 k=
n +1

2k(2k -n -1)!
2

2 -1B2k
n n n n 2k
1 2 3 4 
n-1
- + - +- = (-1) Σ
e1 e2 e3 e4 k=
n +1 2k (2k -n -1)!

2

1 n
3 n
5 n
7 n
n! (-1)n  22k -2B2k
2 k =Σ
+ + + + = -
e1 e3 e5 e7 2 n +1 2k(2k - n -1)!

2
n n n n n  E2k
1 3 5 7 (-1)
- + - +- = Σ
e1 e3 e5 e7 2 k=
n +1 (2k -n )!

2
p p p p
1 2 3 4
1
+ 2
+ 3
+ +  (1+ p ) p >0
e e e e4
1p 3p 5p 7p (1+ p)
1
+ 3
+ 5
+ 7
+  p >0
e e e e 2
22k -2B2k  2k
 

(2n -1)! = Σ
k=n 2k(2k -2n )! 2
2k
2 B2k  2k

 

(2n)! = Σ
k= n +1 2 2
k k n
( -2 -1)! 2

11 Termwise Higher Derivative (Inv-Trigonometric, Inv-Hyperbolic)

Among trigonometric functions and hyperbolic functions, there are functions that it is difficult to obtain the general
form of the second or more order derivative. In this chapter, we differentiate the series expansion of these functions
termwise and obtain the following expressions. Where, ,  denote the ceiling function and the floor function
respectively.

(1) Taylor Series

(n )  (2k)! n
tan x Σ (-1)k
-1
= x 2k+1- |x| <1
k=
n -1

(2k +1- n)!
2

(n )  (2k)! n
cot x Σ (-1)k
-1
=- x 2k+1- |x| <1
k=
n -1

(2k +1- n)!
2
2
(n )  (2k -1)!!
sin x Σ
-1
= x 2k+1-n |x| <1
k=
n -1

(2k +1- n)!
2

- 17 -
 2
(n )  (2k -1)!!
cos x Σ
-1
=- x 2k+1-n |x| <1
k=
n -1

(2k +1- n)!
2
(2k -1)!! (2k +n)! -2k-n-1
(n ) 
csc x
-1
= (-1)nΣ x |x| > 1
k=0 (2k)!! (2k +1)!
-1 (n ) n-1  (2k -1)!! (2k +n )! -2k-n-1
sec x = (-1) Σ x |x| > 1
k =0 (2k)!! (2k +1)!
(n )  (2k)! n
tanh x Σ
-1
= x 2k+1- |x| <1
n -1
k= 
(2k +1- n)!
2
(n )  (2k + n)! -2k-1- n
coth x = (-1)(n)Σ
-1
x |x| > 1
k=0 (2k +1)!
2
(n )  (2k -1)!!
sinh x Σ (-1)k
-1
= x 2k+1-n |x| <1
n -1
k= 
(2k +1- n)!
2
(n ) (n -1)! n-1  (2k + n -1)! -2k-n
cosh
-1
x = (-1)n-1 + (-1) Σ 2 x x >1
xn k=1 (2k)!!

2
(n -1)! 
(n ) k (2k -1)!!
csch x - Σ (-1)
-1 n
= (-1) n x 2k-n 0< x <1
x n 2k (2k -n)!
k= 
2
2
(n ) (n -1)!  (2k -1)!!
sech x -Σ
-1 n
= (-1) n x 2k-n 0< x <1
x n 2k (2k - n )!
k= 
2

(4) By-Products
 (2k)! (n -1)! n /2
Σ (-1)k = (-1)n Σ (-1)k nＣ n+1-2k
k=
n -1

(2k +1- n)! 2n k =1
2

Example
1
12 - 34 + 56 - 78 +-  =-
2

12 Super Derivative (Non-integer times Derivative)

Here, extending the domain of the index function of a differenciation oprator, we obtain Super Derivative (non-integer
order derivative).

(1) Definitions and Notations

f <p>(x ) denotes the non-integer order derivative function of f (x ). And we call this Super Derivaive of f(x).
Since there is a super derivative function innumerably, which f <p >(x ) means follows the definition at that time.
We call it Super Differentiation to differentiate a function f with respect to an independent variable continuously.
And it is described as follows.
p

 
d d d d
f(x) =  f(x) : p pieces
dx p dx dx dx

(2) Fundamental Theorem of Super Differentiation

r [0 , p ] be an continuous function on the closed interval I
(r)
Let f and be arbitrary the r-th order derivative

- 18 -
function of f . And let a (r ) be a continuous function on the closed interval [0 , p ].
Then the following expression holds for a(r ), x  I .

 
dp d p p -1 (r) x x
= (p )
f(x) f (x) + Σ f a(p -r)  dx r
dx p dx p r=0 a(p ) a(p -r)
Especially, when a(r)= a for all k [0 , p ] ,
dp (p ) d p p -1 (r) (x -a)r
f(x) = f (x) + Σ f (a )
(1+ r)
dx p dx p r=0

(3) Lineal and Collateral

We call the 2nd term of the right sides of these Constant-of-differentiation Function.
(p )
And when Constant-of-differentiation Function is 0, we call the left side Lineal Super Differentiation and we call f (x )
Lineal Super Derivaive Function .
Oppositely, when Constant-of-differentiation Function is not 0, we call the left side Collateral Super Differentiation and
we call the right side Collateral Super Derivaive Function.
For example,
when a  - ,
dp x d p p -1 a (x -a)r Left: Collateral p -th order differenciation
x
e =e + Σ e (1+ r)
dx p dx p r=0 Right: Collateral p -th order derivative

when a = -
dp x Left: Lineal p -th order differenciation
pe
= ex
dx Right: Lineal p -th order derivative
As seen from this example, the lineal and the collateral exist in the super differentiation unlike the higher differenciation

(4) Riemann-Liouville Differintegral

When the super integral of f (x ) is the one with a fixed lower limit, the super derivative of
f (x ) is obtained by
the following formula. In this formula, the n th order differenciation is subtracted from the n -p th order integration
and the p th order derivative is obtained. Then, this formula is called Riemann-Liouville Differintegral .

 (x -t)
dp 1 dn x
n-p-1
f(x) = f(t)dt n = ceil(p)
dx p (n -p) dx n a
Since the left side has lost the operating function, Reimann-Liouville Differintegral of the right side is very important.

(5) Fractional Derivative & Super Derivative

In traditional Fractional Derivative , the super derivative is drawn from Riemann-Liouville Differintegral.
For example, in the case of f (x )= log x , it is as follws.

 =
1 1

 (x -t)
1 x
2 1 d 1-
2
-1
(log x) log t dt
(1-1/2) dx 1 0
1

 (x -t)
1 d x -
2
= log t dt
 dx 0
x

  -2 
1 d x -t
= -4 x tanh -1 x -t (log t -2)
 dx x 0
Long calculation continues.


- 19 -
 1
1 d log x +2 log 2 - 2
= 4 x (log 2-1) +2 x log x = x
 dx 
Because the technique of Fractional Derivative is difficult like this, it is unknown in whether the case of p =1/3 is
calculable in this way.
Also, it is the same as the case of super integral that Riemann-Liouville Differintegral cannot be applied to the non-
integer order derivative such as sin x .
On the other hand, in Super Derivative that I advocates, first of all, we obtain the following higher derivative.
log x - (1+p)-  p
  log xdx
x x
p
= x
0 0 (1+ p)
Since differentiation is the reverse operation of integration, reversing the sign of the indexof the integration operator,
we obtain the following immediately.
log x - (1-p)-  -p
(log x) p =
( )
x
(1- p)
Substituting p = 1/2 for this,

  log x - 1/2-  x - 2
1 1
(log x) 2 =
 1/2  

log x - (- -2 log 2) - 

1 1
-
2 log x + 2 log 2 -
2
= x = x
 
Thus, we obtain the desired super derivative very easily.
By the way, when p = 2, according to the formulas
 (-n)
= (-1)n+1n! , (-n)= 
(-n)
it is as follows.
log x - (-1)-  -2 1
(log x)(2) = x = -(-1)1+11! x -2 = - 2
(-1) x
Furthermore, in a similar way, we can obtain the super derivative such as sin x easily.

(6) Super Derivatives of Elementary Functions

In this way, the following super derivatives obtined from " 09 Higher Derivative " .
(p ) (1+ ) -p
x   = x (  0)
1+ - p
(- + p) -p
= (-1)-p x ( < 0)
(- )
 x (p )
e  = (1)-pe  x
log x - (1-p)-  -p
(log x) p =
( )
x
(1- p)
p
(sin x) p  
()
= sin x +
2
p
= cos x +
2 
(cosx) p
( )

e x - (-1)-pe -x
(sinh x) p =
( )
2

- 20 -
 e x + (-1)-pe -x
(p )
(cosh x) =
2
tan -1x 
 
-1 (p )
-p
tan x =
(1-p) Σ
(-1)k x -p-2k
k=0 -p -2k
log 1+ x  
2

 
-p
+ Σ (-1)k x -p+1-2k
2(1-p) k=1 -p +1-2k

 
1  -p
Σ (-1)r  (1- p )- (2r)x
-p+1-2r
-
(1-p) r=1 -p +1-2r
tan -1x 
 
-1 (p )
-p
cot x = - Σ (-1)k x -p-2k
(1-p) k=0 -p -2k
log1+ x  
2

 
-p
- Σ (-1)k x
-p+1-2k
2(1-p) k=1 -p +1-2k

 
1  -p
Σ (-1)r  (1- p )- (2r)x
-p+1-2r
+
(1-p) r=1 -p +1-2r

(7) By-Products
(-1)k (1+ n , -1/m)
k  = -
n n
Σ
k=0 mk +(m -1) m(mn +m -1)
B() is Beta function

q +1- p  (-1)r q
q (p )
x  = Σ
(1-p) r=0 r +1-p r
 x q-p

q -p  (-1)r q -1
(q , -p) = -
p Σr=0 r +1- p r   B() is Beta function

13 Termwise Super Derivative

The following termwise super derivatives are obtained from "10 Termwise Higher Derivative (Trigonometric, Hyperbolic)"
and "11 Termwise Higher Derivative (Inv-Trigonometric, Inv-Hyperbolic)" .
Where, ,  denote the ceiling function and the floor function respectively. And Bernoulli Numbers and Euler
Numbers are as follows.
1 1 1 1 5
B0 =1, B2= , B4=- , B 6= , B8=- , B10= ,
6 30 42 30 66
E0 =1, E2 =-1, E4 =5, E6 =-61, E8 =1385, E10=-50521, 

(p )  22k 22k -1B2k p-1 

(tan x) = Σ
p +1 2k (2k - p)
x 2k- 0< x <
2
k= 
2

(p )  22k 22k -1B2k p-1 

(tanh x) = Σ
p +1 2k (2k - p)
x 2k- 0< x <
2
k= 
2

 E2k 
(sec x) p
( )
Σ
2k-p
= x 0< x < collateral
p +1 (2k -p +1) 2
k= 
2

 E2k 
(sech x) p =
( )
Σ p +1 (2k -p +1)
x 2k-p 0< x <
2
collateral
k= 
2

- 21 -
 22k 22k -1B2k  2k- p-1

 

(cot x)
(p )
=- Σ x- < x< 
k=
p +1

2k (2k - p) 2 2
2

E2k  2k-p

p +1 (2k -p +1) 
x- 

(csc x)
(p )
= Σ < x< 

2 2
k=
2

 (2k +1)p
(csch x) (p )
= (-1) 2Σ -p
(2k +1)x x >0
k=0 e
 (2k +1)p
(sech x)(p) = (-1)-p2Σ(-1)k ( )x x >0
e 2k +1 k=0

 (2k)!
-1 (p ) p
tan x = Σ (-1)
k
x 2k+1- 0< x < 1
k=
p -1

(2 k +2- p )
2

(p ) x -p  (2k)! 2k+1- p
cot x
-1
= - Σ (-1)k x 0< x < 1
2 (1-p) k=
p -1

(2k +2- p)
2
2
(p )  (2k -1)!!
sin x Σ
-1
= x 2k+1-p 0< x < 1 collateral
p -1 (2k +2- p )
k= 
2

(p ) x -p  (2k -1)!!
2
cos x Σ
-1
= - x 2k+1-p 0< x < 1 collateral
2 (1-p) p -1 (2k +2- p )
k= 
2
(p ) 
(2k)! p
tanh x = Σ
-1
x 2k+1- 0< x < 1
p -1 (2k +2- p )
k= 
2
2

(p ) k (2k -1)!! p
sinh x Σ
-1
= (-1) x 2k+1- 0< x < 1 collateral
k=
p -1

(2k +2- p)
2

x -p 2

  - Σ 2k (2k -p+1) x
(p ) 2  (2k -1)!!
sech
-1
x = log + (1- p)+ 2k-p
(1- p) x k=
p

2

0< x <1 collateral

-p 2

 
(p ) x 2  (2k -1)!!
csch
-1
x = log + (1- p)+ - Σ (-1)k x 2k-p
(1- p) x k=
p

2k (2k - p+1)
2

0< x <1 collateral

14 Higher and Super Calculus of Logarithmic Integral etc.

Here, the higher integrals and the super calculus of the double logarithm function and the following four functions are
shown.

 et
  
x x cos t x sin t x 1
Ei(x) = dt , Ci(x)= dt , Si(x)= dt , li(x)= dt
- t  t 0 t 0 log t

(1) Higher Integrals of Logarithmic Integral etc.

 
x x 1 n -1

 
n-1- r
Ei(x)dx n = x nEi(x)- e xΣr! x
- - n ! r=0

- 22 -
 
x x 1

(n-1)/2
 Ci(x)dx n = Ci(x)x n - sin x Σ (-1)r(2r)! x
n-1-2r
  n ! r=0


(n-2)/2
+ cosx Σ (-1)r(2r +1)! x n-2-2r n 2
r=0


x x 1

(n-1)/2
 Si(x)dx n = Si(x)x n + cosx Σ (-1)r(2r)! x
n-1-2r
0 0 n! r=0


(n-2)/2
+ sin x Σ (-1)r(2r +1)! x n-2-2r
r=0
n-1-2r
(n-1)/2
rx
- Σ
r=0
(-1)
(2r +1)(n -1-2r)!

  li(x)dx = n ! Σ(-1) Ｃ x Ei (r +1)log x

x x 1 n
 n r
n r
n- r
 
0 0 r=0

 log|log x|dx = n ! x log|logx| + Σ(-1) Ｃ x

x x 1 n


n n r n- r
n r Ei(rlog x)
0 0 r=1

(2) Super Calculus of Logarithmic Integral etc.

 
x x 1  p
 li(x)dx p =
Σ (-1)r p- r
x Ei(r +1)log x x 0
0 0 (1+ p) r=0 r

 
1  -p
x 0
r
Σ
(p )
li(x) = (-1)r x -p- Ei(r +1)log x
(1- p) r=0 r

   
x x 1  p
 log|log x|dx p = x plog|logx| +Σ(-1)r
p-r
x Ei(rlog x)
0 0 (1+ p) r=1 r

   
1  -p Ei(rlog x)
(log|log x|)(p) = p log|logx| + Σ(-1)r x 0
x (1-p) r=1 r xr

15 Higher and Super Calculus of Elliptic Integral

When |x|  1 , |k|  1 , |c|  1 , Elliptic Integrals of the 1st 3rd kind are expressed as follows.

 
x dx x 1- k 2x 2
F(x, k) = , E(x, k) = dx
0 1- x 1- k x 
2 2 2 0 1- x 2


x dx
 (x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

In this chapter, we expand these to a double series or a triple series and calculate the arc length of an ellipse and
a lemniscate using thiese. Next, we calculate these term by term.

(1) Double (triple) Series Expansion

(-1)r
 r-s   s  k x
 r -1/2 -1/2
F(x, k) = ΣΣ 2s 2r+1
r=0 s=0 2r +1
r

2r +1  r - s   s 
(-1) -1/2 1/2 r
E(x, k) = ΣΣ k x 2s 2r+1
r=0 s=0
r
c 
s-t   t 
(-1)  -1/2 -1/2
r s
r-s
(x,c,k) = ΣΣΣ k 2t
x 2r+1
2r +1 r=0 s=0 t=0

- 23 -
(2) Termwise Higher Calculus


(-1)r(2r)!
 r-s   s  k x
x x  r -1/2 -1/2
 F(x, k)dx =ΣΣ n 2s 2r+n+1
0 0 r=0 s=0 (2r +n +1)!


r

(2r +n +1)!  r - s   s 
x x (-1) (2r)! -1/2 1/2
 r
 E (x, k)dx = Σ Σ n
k x 2s 2r+n+1
0 0 r=0 s=0


r
ΣΣΣ (2r +n +1)!  s - t   t  k
x x (-1) (2r)! -1/2 -1/2
 r s
  ( x,c,k)dx = c n r-s 2t
x 2r+n+1
0 0 r=0 s=0 t=0

dn r (-1)r(2r)!

 r-s   k
 -1/2 -1/2 2s 2r- n+1
F(x, k) = Σ Σ x
dx n r=
n -1 s=0 (2r - n +1)!

s
2
n r

 r-s   s  k
d 
r (-1) (2r)! -1/2 1/2
E(x, k) = Σ Σ 2s 2r-n+1
x
dx n r=
n -1 s=0 (2r - n +1)!

2
n s (-1)r(2r)!

 
-1/2
k
d  r -1/2
r-s
 (x,c,k) = Σ Σ Σ c 2t
x 2r-n+1
dx n r=
n -1 s=0 t=0 (2r - n +1)!

s-t t
2

(3) Termwise Super Calculus

(-1)r(2r +1)
  r-s   s  k x
x x  r -1/2 -1/2
 F(x, k)dx p =ΣΣ 2s 2r+p+1
0 0 r=0 s=0 (2r +p +2)
(-1)r(2r +1)
  E(x, k)dx =ΣΣ  r-s   s  k x
x x -1/2p
1/2  r
2s 2r+p+1
0 0 r=0 s=0 (2r +p +2)
(-1) (2r +1)
   (x,c,k)dx =ΣΣΣ (2r +p +2) c  s - t   t  k
x x -1/2 
r
p
-1/2 r s
r-s 2t
x 2r+p+1
0 0 r=0 s=0 t=0

dp r (-1)r(2r +1)

 r-s   k
 -1/2 -1/2
Σ Σ (2r -p +2)
2s 2r-p+1
F x, k
( ) = x
dx p r=
p -1 s=0

s
2
p r (-1)r(2r +1)

 r-s   s  k
d  -1/2 1/2
E(x, k) = Σ Σ 2s 2r-p+1
x
dx p r=
p -1 s=0

(2r -p +2)
2
p s (-1)r(2r +1)

 
-1/2
k
d  r -1/2
r-s
 (x,c,k) = Σ Σ Σ c 2t
x 2r-p+1
dx p r=
p -1 s=0 t=0

(2r - n +2) s -t t
2

16 Higher Integral of the Product of Two Functions

We obtain the following thorem for the product of two functions.

(1) Theorem 16.1.2

<r> <r>
Let f be the arbitrary r th order primitive function of f (x )and g be the r th order derivative function of g (x )

r =1, 2,  , m +n -1 . Let f a , g a be the function values of f ak for k =1, 2,  , n

<r> (r) <r> (r)
for ,g on .
k k
And let B(n, m ) be the beta function. Then, the following formulas hold.

a a r
x x m -1 -n
 f g dx n = Σ
<0> (0)
f
<n + r> (r)
g
r=0
n 1

 s f a a
-n + r
n -1 m -1 x x
- ΣΣ <n -r+ s>
an-r
(s)
ga  dx r
r=0 s=0 n-r
n n- r+1

- 24 -
 a a
n -1 r -1 r -1 x x
t s  m +n-1-r + tＣ m-1 f a 
m+ n -r+s (m+ s)
+ (-1)m ΣΣΣ Ｃ ga dx r
r =1 s =0 t=s n- r n- r
n n- r+1

a a f
m
(-1) n -1 n -1Ｃ k x x
 <m+ k > (m+ k )
dx n
(n , m) Σ
+ g
k=0 m + k 1
n

Especially, when ar = a for r =1, 2,  ,n and

when f
<r>
(a) = 0 (r =1, 2,  ,m+n -1) or g (s)(a ) = 0 (s =0, 1,  ,m+n -2) ,

 (-1)m n -1 n -1Ｃ k x x <m +k > (m +k ) n

  
x x m -1 -n
 f g dx = Σ  f
<0> (0) n <n+r> (r)
 (n , m) Σ
f g + g dx
a a r=0 r k=0 m + k a a

(2) Higher Integral of the Product of Two Functions

We obtain the following expressions using this theorem.

 
x x
p q n
(ax +b) (cx +d) dx
b b
- -
a a

(1/a)n+ r (1+p)(1+q) (ax +b)p+n+ r

r
 -n
=Σ
r=0 (1/c)r (1+p +n + r)(1+q -r) (cx +d)r-q

 
x x
 (ax +b)p (cx +d)mdx n
b b
- -
a a

(1/a)n+ r (1+p)(1+m) (ax +b)p+n+ r

 
m -n
=Σ
r=0 r (1/c)r (1+p +n + r)(1+m -r) (cx +d)r-m
-n log x - (1+ n+r)-  (1+ ) +n
  
x x 
 x  log xdx n = Σ x
0 0 r=0 r (1+n+ r) (1+ -r)
-n log x - (1+ n+ r)-  (1+ m) m+ n
  
x x m
 x mlog xdx n = Σ x
0 0 r=0 r (1+ n+r) (1+ m - r)
(1+ m) m- r (n + r)
  
x -n
 
x m
 x m sin x dx n = Σ x sin x -
an a1 r=0 r  ( 1+ m - r ) 2
(1+ m) (n + r)
a a x cosx dx = Σ  
-n
 
x x m
 m n
x
m- r
cos x -
n 1 r  ( 1+ m - r )
r=0 2
(1+ m) m- r e x -(-1)n+ re -x
aa x r
x x m -n
m
sinh x dx = Σn
x
n 1
r=0  ( 1+ m - r ) 2
(1+ m) m- r e x +(-1)n+ re -x
aa x r
x x m -n
m
cosh x dx = Σn
x
n 1
r=0  ( 1+ m - r ) 2
n


x x

 
n 1 n x
log xdx = log x log x -Σ
2
0 0 k=1 k n!
(r)
  (1+ n +r) 
-n

 n+r 1
+ Σ(-1)r-1 x
n
log x -Σ
r=1 r k =1 k
x  n -1 Ｃ r (n -r + , -x)x r

x x
 n -1
 x
e x dx = n
Σ +R
- - (-x) r=0 (n -1)!

- 25 -
 x 0


0
R= n -1 n -1 Ｃ r (n -r + )x r
2i sin Σ x >0
r=0 (n -1)!
(1+ m) m- r
 r
x x m -n
 e x x mdx n = e x Σ x
- - r=0 (1+ m - r)

 
s! x r xr
= e log |x| +Σ Σ 
x x n -2 n -2-r n -1
 e x log x dx n x
- Ei(x)Σ
- - r=0 s=0 (r + s +1)! r=0 r!
 n

n

 4  4 
x x
 e sin xdx = sin e x
sin x - n x
- -

 n
 e cosxdx = sin 4  e cos x - 4 
x x n
x n x
- -


x x (-1)n n -1 
 e x dx = -x
Σ (-1)rn -1Ｃ r (n -r + , x)x r
n
  (n -1 ) ! r=0

(1+ m) m- r

n

 
x x (-1) m -n
 e -x x mdx n = Σ ( -1) r
x
  e x r=0 r (1+ m - r)


s!(-x)r
 
x x n -2 n -2-r
 e log x dx = (-1)n e -x log|x| + Σ
-x n
Σ
  r=0 s=0 (r + s +1)!
n -1 (-x)r
- (-1)n Ei(-x)Σ
r=0 r!
 n
 e
n

 
x x
sin xdx n = (-1)n sin  
-x
e -xsin x +
  4 4
 n
 e
n

   
x x
-x
cosxdx = (-1)n sin
n
e -xcos x +
  4 4

(3) By-Products
n /2  (-1)k  -n-1
(n +1)
 
1
Σ
k=0 2k +1 n
Ｃ 2k =
n +1
sin
4
sin
4
 (n +1)
 
n /2  (-1)k -n-1


1
Σ
k=1 2k
nＣ 2k-1 =
n +1
sin
4
cos
4
-1

Example
Ｃ0
8 Ｃ2
8 Ｃ4
8 Ｃ6
8 Ｃ8
8 16
- + - + =
1 3 5 7 9 9
Ｃ1
8 Ｃ3
8 Ｃ5
8 Ｃ7
8 5
- + - + =
2 4 6 8 3

17 Super Integral of the Product of Two Functions

We obtain the following thorems for the product of two functions.

(1) Theorem 17.1.2

f r be an arbitrary r th order primitive function of f (x ) , g r be the r th order
<> ()
Let r, p are positive numbers,
derivative function of g (x ),  (n , m) , (p ) are the beta function and the gamma function respectively. At this

- 26 -
time, if there is a number a such that
f (a) = 0 r [0 , m + p] g (a ) = 0 s [0 , m +p -1] ,
<r> s()
or
then the following expression holds.

  r f
x x m -1 -p
 f g dx p = Σ
<p + r> (r)
g + Rmp
a a r=0

 k   f
(-1)m  1 p -1 x x
 <m+ k > (m+ k )
(p , m) Σ
Rmp = g dx p
k=0 m + k a a

(2) Super Integral of the Product of Two Functions

We obtain the following expressions using this theorem.

 
x x
p q s
(ax +b ) (cx +d) dx
b b
- -
a a
(1/a)s+ r (1+p)(1+q) (ax +b)p+s+ r
r
 -s
=Σ
r=0 (1/c)r (1+p +s + r)(1+q -r) (cx +d)r-q

 
x x
 (ax +b )p (cx +d)mdx s
b b
- -
a a
(1/a)s+ r (1+p)(1+m) (ax +b)p+s+ r
r
m -s
=Σ
r=0 (1/c)r (1+p +s + r)(1+m -r) (cx +d)r-m
log x - (1+ p+r)-  (1+ ) +p
 r
x x  -p
 x  log xdx p = Σ x
0 0 r=0 (1+p+ r) (1+ -r)
-p log x - (1+p+ r)-  (1+ m)
 x log xdx  r  (1+p+r) (1+ m -r) x
x x m
m+ p
=Σ m p
0 0 r=0

(1+ m) (p + r)
 Σ  r  (1+ m - r) x sin x - 2 
x x-p m
 x sin x dx = m p m- r
ap a0 r=0

(1+ m) (p + r)
 Σ  r  (1+ m - r) x cosx - 2 
x x-p m
 x cosx dx = m p m- r
ap a0 r=0

(1+ m) m- r e x -(-1)p+ re -x
a a r
x x m -p
 x m sinh x dx p = Σ x
p 0
r=0  ( 1+ m - r ) 2
(1+ m) m- r e x +(-1)p+ re -x
aa x r
x x
m p
m -p
cosh x dx = Σ x
p 0
r=0  ( 1+ m - r ) 2

x plog x - (1+ p)- log x


x x
 log xdx = 2 p
0 0 (1+ p)
-p log x - (1+ p +r)-(r)
 

+ x p Σ(-1)r-1
r=1 r (1+ p +r)
(-1)  p -1
  
x x
x 
 e x dx =p
Σ (p -r + , -x)x r
- - (p) r=0 r
(1+ m)
 e x dx = e Σ  r  (1+ m - r) x
x -p
x
x m p x
m
m- r
- - r=0

- 27 -
  p

p

 
x x

-

-
e xsin xdx p = sin  4  e xsin x -
4
 p

p

 
x x

-

-
x
e cosxdx = sin p
 4  e xcos x -
4


(-1)p 
 
x x

p -1
 -x
e x dx = Σ p
(-1)r (p -r + , x)x r
  ( p r=0
) r
(1+ m) m- r

p

 
x x (-1) m -p
 e -x x mdx p = Σ (-1) r
x
  e x r=0 r (1+ m - r)
 p

p

 
x x
 e sin xdx = (-1)p sin
p
 
-x
e -xsin x +
  4 4
 p
 e
p

   
x x
-x p p
cosxdx = (-1) sin e -xcos x +
  4 4

(3) By-Products
(-1)k  -p-1
(p +1)
 
p
 
 1
Σ
k=0 2k +1 2k
=
p +1
sin
4
sin
4
 (p +1)
  = p +1  sin 4  
 (-1)k p
1 -p-1
Σ
k=1 2k 2k -1
cos
4
-1

18 Higher Derivative of the Product of Two Functions

The following Leibniz rule is drawn from the Theorem 16.1.2.

Theorem 18.1.1 (Leibniz)

When functions f (x ) and g (x ) are n times differentiable, the following expression holds.

r f
(n ) n n (n -r)
(x)g r (x)
()
f(x)g (x) =Σ
r=0

(2) Higher Derivative of the Product of Two Functions

We obtain the following expressions using this theorem. Where, p >0 ,  are real numbers and m is a non-negative
integer. Furthermore, if  = -1, -2, -3, , it shall read as follows.
(1+ ) (- + r)
 (-1)-r
 ( 1+ - r ) (- )

(1/a)-n+ r (1+p)(1+q) (ax +b)p-n+r

r 
p q (n ) n n
(ax +b ) (cx +d)  =Σ
r=0 (1/c)r (1+p -n + r)(1+q -r) (cx +d)r-q
(1/a)-n+ r (1+p)(1+ m) (ax +b)p-n+ r
r 
m (n )
m n
(ax +b ) (cx +d)  =Σ
p
r=0 (1/c)r (1+p -n +r)(1+m -r) (cx +d)r-m
(n -r)(1+ ) -n (1+ ) -n
r 
(n ) n -1 n
x  log x = -Σ(-1)n-r x + x log x
r=0 (1+ -r) (1+ -n)
n (n -r)(1+m)
 r  (1+m -r) x
n -1
(n ) -n
x log x
m
= -Σ(-1)n-r n > m =0, 1, 2, 3, 
r=0

- 28 -
 (1+ ) -r (n -r)
r 
n
 
{n } n
x  sin x =Σ x sin x +
r=0 (1+ -r) 2
(1+ ) -r (n -r)
r 
n
 
{n } n

x cosx =Σ x cos x +
r=0 (1+ -r) 2
(1+m) m-r (n -r)
r 
n
 
{n } m
x sin x =Σ
m
x sin x +
r=0 (1+m -r) 2
(1+m) m-r (n -r)
r 
n
 
{n } m
x cosx =Σ
m
x cos x +
r=0 (1+m -r) 2
(1+ ) x -(n -r) -x

 r  (1+ -r)

n (n ) e -(-1)
n e - r
x sinh x = Σ x
r=0 2
(1+ ) x -(n -r) -x

 r  (1+ -r)

n (n ) e +(-1)
n e - r
x cosh x = Σ x
r=0 2
(1+ m) x -(n -r) -x

 r  (1+m -r)
m
n (n ) e -(-1)
m e m- r
x sinh x = Σ x
r=0 2
(1+m) x -(n -r) -x

 r  (1+m -r)
m
n (n ) e +(-1)
m e m- r
x cosh x = Σ x
r=0 2
n-1

  r (n - r)(r)
(-1)(n ) n n -1
log x =
2
2(n)log x -Σ n
x r=1

(1+ )
 r  (1+ -r) x
x  (n )
n x
n
- r
e x  = e Σ
r=0

(1+m)
 r  (1+m -r) x
x m (n )
n x
m
m- r
e x  = e Σ
r=0

n (r)
r  x
(n ) n
x x x r-1
e log x = e log x + e Σ(-1) r
r=1

 n
-n

   
x (n )
e sin x = sin e xsin x +
4 4
 n -n
= sin  e cos x +
4 
x (n ) x
e cosx
4
e x - (-1)-re -x
r 
(n ) n n
e sinh x = e xΣ
x
r=0 2
x -r -x

r  2
n e + (-1) e
(n ) n
e cosh x = e Σ
x x
r=0

(1+ )
 r  (1+ -r) x
 (n )
n n
-(n -r) - r
e x  = e Σ(-1)
-x -x
r=0

(1+ m)
 r  (1+m -r) x
(n ) n m
e x  = e Σ(-1) -(n -r)
-x m -x m- r
r=0

n (r) -n

 log x - Σ  
x 
(-1)(n ) n
e log x =
-x
x r
e r r=1

- 29 -
  -n
n
   
(n )
e sin x
-x
= -sin e -xsin x -
4 4
 -n
n
   
(n )
e cosx
-x
= -sin e -xcos x -
4 4
e x - (-1)-re -x
r 
(n ) n n
e sinh x
-x
=e -x
Σ (-1) -n+ r
r=0 2
x -r -x

r  2
(n ) n n e + (-1) e
e cosh x
-x
= e -xΣ(-1)-n+ r
r=0

n
 
(n )
sin x
2
= -2n-1cos 2x +
2
n
2 (n )
 
n-1
cos x = 2 cos 2x +
2
n 3n n
   
(n ) 3
sin x
3
= sin x + - sin 3x +
4 2 4 2
n n n
= cos x +
2   2
(n ) 3 3
cos x
3
+ cos 3x +
4 4
etc.

(3) By-Products
n n
n n
n /2 (n -1)/2
r r
Σ (-1) nＣ 2r Σ (-1) nＣ 2r+1 = 2 sin
2 2
= 2 cos ,
r=0 4 r=0 4
n /2 (n -1)/2
n-1
Σ Ｃ 2r = 2
n , Σ nＣ 2r+1 = 2n-1
r=0 r=0

n /2 3n + (-1)n (n -1)/2 3n - (-1)n

Σ Ｃ 2r Σ 2 nＣ 2r+1
2r-1 2r
2 n = , =
r=0 4 r=0 4

19 Super Derivative of the Product of Two Functions

The following Leibniz rule for Super Derivative is drawn from the Theorem 16.1.2.

Theorem 19.1.1
r =0, 1, 2,  , let f -p+ r be arbitrary
< >
Let B(x,y ) be the beta function and p be a positive number. And for
(r)
primitive function of f (x )and g be the r th order derivative function of g (x ). Then the following expressions hold

r 
(p ) m -1 p (p -r)
(x)g r (x) + Rmp
()
f(x)g (x) =Σ f
r=0

(-1)m  1
 
-p -1 <m+ k > (p )
f (x)g (m+ k )(x)
(-p , m) Σ
Rmp =
k=0 m + k k
Especially, when n =0, 1, 2, 

r f
(n ) n n (n -r)
(x)g r (x)
()
f(x)g (x) =Σ ( Leibniz )
r=0

(2) Super Derivative of the Product of Two Functions

We obtain the following expressions using this theorem. Where, p >0 ,  are real numbers and m is a non-negative

- 30 -
integer. Furthermore, if  = -1, -2, -3, , it shall read as follows.
(1+ ) (- + r)
 (-1)-r
1+ - r  (- )
(1/a)-s+ r (1+p)(1+q) (ax +b)p-s+r
r 
(s)  s
(ax +b ) (cx +d)  =Σ
p q
r=0 (1/c)r (1+p -s + r)(1+q -r) (cx +d)r-q
-s+ r
(1+p)(1+ m) (ax +b)p-s+ r
r  (1/c)
p m (s)
s (1/a)
m
(ax +b ) (cx +d)  =Σ
r=0
r (1+p -s + r)(1+ m -r) (cx +d)r-m
log x - (1-p + r)-  (1+  ) - p
r 
(p )  p
x  log x =Σ x
r=0 (1- p +r) (1+  -r)
log x - (1-p+ r)-  (1+ m) m- p
log x = Σ  
m (p ) p m
x x
r r=0 (1- p+r) (1+ m -r)
(1+ m) (p -r)
sin x =Σ  
p
x sin x + 
(p ) m
m m- r
x
r (1+ m -r)
r=0 2
(1+ m) (p -r)
cosx =Σ  
p
x cos x + 
(p ) m
m m- r
x
r (1+ m -r)
r=0 2
(1+ m) x -p+ r -x
sinh x = Σ  
p
{p } m e -(-1) e m- r
x
m
x
r (1+ m - r)
r=0 2
(1+ m) x -p+ r -x
cosh x = Σ  
m {p }p m e +(-1) e m- r
x x
r (1+ m - r)
r=0 2
(p ) log x log x - 1-p - 
x -p
log x
2
=
1- p
p log x - (1- p +r)-(r)


- x -p Σ(-1)r
r=1 r 1- p + r
(1+ ) - r

(p ) m -1 p
e x  = e Σ
x x
x + Rmp
r=0 r  1+  - r
(-1)m  1 (1+ )
 
-p -1 x - m-k
(p )
 
(-p , m) Σ
Rmp = e x
k=0 m + k k 1+ - m -k
(1+m) m- r

x m (p ) x
m p
e x  = e Σ x
r=0 r (1+ m - r)
 -p x p
   
(p )
e sin x = sin
x
e sin x +
4 4
 -p
p
   
x (p )
e cosx = sin e xcos x +
4 4
x
p e - (-1) e -x -r


(p ) 
x x
e sinh x = e Σ
r=0 r 2
p e x + (-1)-re -x

(p ) 
x x
e cosh x = eΣ
r=0 r 2
(-1)-p m -1 (1+ ) - r
-x  (p )
e x  =
e x Σ
r=0
( -1 ) r p

r 1+ - r
x + Rmp

- 31 -
 x
(p)
 (-1)k (1+) - m-k

   
1 -p -1
Rmp =
 (-p , m) Σ
k =0 m+ k k  (1+ - m -k ) ex
(-1)-p (1+ m) m- r
r 
m (p )
m p
e x 
r
Σ(-1)
-x
= x
e x r=0 1+ m - r
 -p
p
 4  
(p )
e sin x
-x
= (-1)-p sin -x
e sin x -
4
 -p
p
cos x -
 4 4 
(p )
e cosx
-x
= (-1)-p sin e
-x

e x - (-1)-re -x
r 
(p ) -x  -p+ r
p
e sinh x Σ
-x
=e (-1)
r=0 2
x -r -x

r  2
(p ) -x  -p+ r
p e + (-1) e
e cosh x
-x
=e Σ (-1)
r=0

p
 
(p )
sin x
2
= -2p-1cos 2x +
2
p
2 (p )
 
p-1
cos x = 2 cos 2x +
2
p 3p p
   
3 (p ) 3
sin x = sin x + - sin 3x +
4 2 4 2
p p
p
   
3 (p ) 3 3
cos x = cos x + + cos 3x +
4 2 4 2
etc.

(3) By-Products

p p
p p

   

k
p 
k
p
Σ Σ
2 2
(-1) = 2 cos , (-1) = 2 sin
k=0 4
2k k =0 2k +1 4

    r  = 2
 p  p  p
Σ = 2p-1 , Σ = 2p-1 , Σ p
p >-1
r=0 2r r=0 2r +1 r=0

20 Higher Calculus of the product of many functions

20.1 Higher Derivative of the product of many functions

The following theorems are drawn from the Theorem 18.1.1.

Theorem 20.1.1
When fk
(r)
denotes the r th order derivative function of fk (x) (k =1, 2, ,) ,

r  r  r 
n r1 r-2 n r1 r-2
f1
n - r1 r1-r2
 f -1
r
f1 f2  f =Σ Σ  Σ 
(n )
f2
r1=0 r2=0 r-1=0 1 2 -1

Theorem 20.1.2
When f
(r)
denotes the r th order derivativefunction of f (x ) and  is a natural number,

r  r  r 
n r1 r-2 n r1 r-2
f
n - r1 r1 -r2

 f  -1
(n ) r
f (x) =Σ Σ  Σ  f
r1=0 r2=0 r-1=0 1 2 -1

- 32 -
Example
(1+ ) s
 r  s 
n r
 
n r

x e sin x
x (n ) r
=Σ Σ x -n+ e x sin x +
r =0 s =0 1+ - n + r 2
n-1 (n ) 3(n -1)! n 3log x
n-1 n!
log x = (-1) Σ
3 2
log x + (-1 )
xn x n r =1 (n -r)r
(-1)n-1 n-1 r -1 n!
+ n Σ Σ
r =2 s =1 (n -r)(r -s)s
x

Higher Derivatives of cos m x , sin m x

n
 
m (n ) 1 m /2
cos x = m-1 Σ m r m
Ｃ ( -2r)n cos (m -2r)x +
2 r=0 2
 n
   
m (n ) 1 m /2
sin x = m-1 Σ m r m
Ｃ ( -2r)n cos (m -2r) x - +
2 r=0 2 2
1   n
   
(n )
cos x = Σ r
( -2r)n cos ( -2r)x +  >0
2 r=0 2
1    n
     
(n )
sin x =    >0
n
Σ
2 r=0 r
( -2r) cos ( -2r) x -
2
+
2

20.2 Higher Integral of the product of many functions

The following theorems are drawn from the Theorem 16.1.2 and Theorem 20.1.1.

Theorem 20.2.1
be the r th order derivative function of fk (x ) k =1, 2, , ,
(r )
fk r
<>
Let fk be the arbitrary r th order primitive

function of fk (x ) , m, n are natural numbers and B(n , m )be the beta function. If there is a number a such that

f1r (a ) =0 (r =1, 2,  ,m +n -1) or fk s (a ) =0 (s =0, 1,  ,m +n -2)

<> ()
for at least one k > 1, then the
following expression holds.

  r  r  r 
x x m -1 r1 r-2 -n r1 r-2
f1
n + r1 r1 -r2
 f -1
r
 f1 f2  f dx n =Σ Σ  Σ  f2
a a r1=0 r2=0 r-1=0 1 2 -1
+ Rmn

 k  k 
(-1)m n-1 m + k1 k2 n -1Ｃ k1 k-2

k -2 m + k1 k2
Rmn = Σ Σ Σ  Σ 
(n , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 k2 3 -1

 f
x x
m+ k 1 m+k 1 -k 2 k 2-k 3
 f -1dx n
k
  1 f2 f3
a a

Theorem 20.2.2
(r )
f (x ), fk r
<>
Let f be the r th order derivative function of be the arbitrary r th order primitive function of f (x ),
m, n are natural numbers and B(n , m)be the beta function. At this time , if there is a number a such that

f (a) =0 (r =1, 2,  ,m +n -1) f (a) =0 (s =0, 1,  ,m +n -2)

<r> s ()
or
then the following expression holds for  =2, 3, 4,  .

  r  r  r  f
x x m -1 r1 r-2 -n r1 r-2 n + r1 r1 -r2

 f  -1 + Rmn
r
 f dx =Σ Σ  Σ
n
 f
a a r1=0 r2=0 r-1=0 1 2 -1

- 33 -
  k  k 
(-1)m n-1 m + k1 k2 n -1Ｃ k1 k-2

k -2 m + k1 k2
Rmn = Σ Σ Σ Σ 
(n , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 k2 3 -1

 f
x x
m+ k 1 m+k 1-k 2 k 2-k 3
 f  -1dx n
k
  f f
a a

Example

  r   s   (1+ + n + r ) x
(1+) s
x x m -1 r -n r
x  e x sin x dx n = Σ Σ   +R
+n+ r
 e x sin x + n
m
- - r =0 s =0 2
(1+)
 
s
Ｃ m+ r
m

m+ r  s 
x x

 
(-1) n -1 m + r n -1 r +m+ r x
Rmn = Σ Σ x e sin x + dx n
 (n , m)  (1+ + m + r )
r =0 s =0 - - 2

log x - (1+n ) -  n
  log x dx
x x
 3 n
= x (log x)
2
0 0  (1+n )
-n log x - (1+n + r ) - 
 

- 2x n log xΣ(-1)r  (r )
r=1 r  (1+n + r )
r log x - (1+n + r ) - 
 r   s   (1+n + r )  (r -s ) (s)
n  -n
r-1
+ x ΣΣ(-1) r
r=2 s=1

Higher Integrls of cos m x , sin m x

  n
x x x Ｃr
cos (2m -2r +1) x - 
1 m 2m +1
 cos
2m+1
x dx n = Σ
(n-1) 1 0 22m r=0 (2m -2r +1) n 2
2 2 2

   (-1)m-r2m +1Ｃ r n
x x x

 
n 1 m
Σ
2m+1
sin x dx = 2m n
sin (2m -2r +1) x -
n 2 1 2 r=0 (2m -2r +1) 2
2 2 2

 n

x x Ｃm Ｃm x x

 
1 m -1 2m 2m
 cos
2m
x dx n = Σ n
cos (2m -2r)x - +  dx n
an a1 22m-1 r=0 (2m -2r ) 2 22m an a1

 (-1)m-r2mＣ m n
 dx
x x Ｃm x x

 
1 m -1 2m
 sin
2m
x dx n = Σ n
cos (2m -2r)x - + n

an a1 22m-1 r=0 (2m -2r) 2 2 2m

an a1

Where, a1 , a2 ,  ,an are the solutions of the following transcendental equations.

Ｃm n
cos (2m-2r)x - 
m -1 2m
Σ n
=0 k =1, 2,  ,n
r=0 (2m -2r ) 2

(-1)m-r2mＣ m n
 = 0
m -1
Σ n
cos (2m -2r)x - k =1, 2,  ,n
r=0 (2m -2r) 2


 cos x dx = 2 Σ  r   -2r
x 1 1 

sin ( -2r) x   
0 r=0


 sin x dx = 2 Σ  r   -2r sin ( -2r)x - 2 
x 1 1 

 r=0

2


x 1 m -1 2m Ｃr 2m Ｃm
2m-1 Σ
cos 2m x dx = sin {(2m -2r)x } + x
0 2 r=0 2m -2r 22m

- 34 -
 
 sin
Ｃr Ｃm
 
x 1 m -1 2m 2m
m-r
2m-1 Σ
2m
x dx = (-1) sin(2m -2r)x + x-
 2 r=0 2m -2r 22m 2
2

21 Super Calculus of the product of many functions

21.1 Super Integrals of the product of many functions

The following theorems are drawn from the Theorem 20.2.1.

Theorem 21.1.1
p, r are positive numbers, m be a natural number, fk r be the r th order derivative function of
()
Let

fk (x) k =1, 2, , , fk r be arbitrary r th order primitive function of fk (x) and B(p , m ) be
<>

the beta function. At this time, if there is a number a such that

f1r (a) = 0 r [0 , m + p ] or fk s (a ) = 0 s [0 , m +p -1]

<> ()
for at least one k > 1,
the following expression holds

  r  r  r 
x x m -1 r1 r -2 -p r1 r-2
f1
p + r1 r1-r2
 f -1
r
 f1 f2  f dx =Σ Σ  Σp
 f2 + Rmp
a a r1=0 r2=0 r-1=0 1 2 -1

 k  k 
m k-2
 k 
(-1)  m + k1 k2 k -2
1 p -1 m + k1 k2
Rmp = Σ Σ Σ  Σ 
(p , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 1 k2 3 -1

f
x x
m+ k 1 m+k 1-k 2 k 2-k 3
 f -1dx p
k
  1 f2 f3
a a

Theorem 21.1.2
(r)
Let p, r are positive numbers, m be a natural number, f be the r th order derivative function of f (x ),
<r>
f be arbitrary r th order primitive function of f (x ) and B(p , m ) be the beta function. At this time,
if there is a number a such that
(a ) = 0 r [0 , m + p] or f (s)(a ) = 0 s [0 , m +p -1]
f
<r>

the following expression holds for  = 2, 3, 4,  .

  r  r  r 
x x m -1 r1 r-2 -p r1 r-2
f
p + r1 r1 -r2
f  dx p =Σ Σ  Σ  f  -1 + Rmp
r
  f
a a r1=0 r2=0 r-1=0 1 2 -1

m + k  k   k  k  k 
(-1) 1m p -1 m +k k
 m + k1 k2 k k -2 1 2 -2
Rmp = Σ Σ Σ Σ 
(p , m) k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1

f
x x
m+ k 1 m+k 1-k 2 k 1-k 2
 f  -1dx p
k
  f f
a a

Example

 x  r   s   (1+ + p + r ) x
(1+) s
x x m -1 r -p r
  +R
  +p+ r
e x sin x dx p = Σ Σ e x sin x + p
m
- - r =0 s =0 2
(1+)
 
s
m
Ｃ
 
p -1  m+r x x

 
p (-1) m +r s  +m+ r x
R = Σ Σ x e sin x + dx p
m
 (p , m) r m+ r  (1+ + m + r )
r =0 s =0 - - 2
log x - (1+p ) - 
 log x dx =  (1+p
x x
3 p p 2
x (log x)
0 0 )

- 35 -
 log x - (1+p + r ) - 
 r   (1+p + r )  (r )
 -p
- 2x p log xΣ(-1)r
r=1

r log x - (1+p + r ) - 
 r   s   (1+p + r )  (r -s ) (s)
p  -p
r-1
+ x ΣΣ(-1) r
r=2 s=1

Super Integrls of cos m x , sin m x

p

2m +1Ｃ r
 
x x 1 m
 cos 2m+1x dx p = 2m Σ cos (2m -2r +1)x -
a(p ) a(0) 2 r=0 (2m -2r +1)p 2

p
 
(-1)m-r2m +1Ｃ r
 
x x 1 m
p
Σ
2m+1
sin x dx = p sin
(2m -2r +1)x -
a(p ) a(0) 22m r=0 (2m -2r +1) 2
a(s) s [0 , p ] are zeros of lineal super primitives of cos
2m+1 2m+1
Where, x or sin x .

21.2 Super Derivatives of the product of many functions

The following theorems are drawn from the Theorem 21.1.1.

Theorem 21.2.1
p, r are positive numbers, m be a natural number, fk r be the r th order derivative function of
()
Let
fk (x) k =1, 2, , , fk r be arbitrary r th order primitive function of fk (x) and B(p , m ) be
<>

the beta function. At this time, if there is a number a such that

f1r (a) = 0 r [0 , m - p ] or fk s (a ) = 0 s [0 , m -p -1]
<> ()
for at least one k > 1,
the following expression holds

r  r  r 
m -1 r1 r-2 p r1 r-2
f1
p - r1 r1-r2
 f -1
r
f1 f2  f =Σ Σ  Σ 
(p ) p
+R f2 m
r1=0 r2=0 r-1=0 1 2 -1

(-p , m) Σ Σ Σ Σ m + k  k   k  k  k 
(-1) m
1 -p -1 m +k k
 m + k1 k2 k k -2 1 2 -2
Rmp =  
k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1
(p )
 f1
m+ k 1 m+k 1-k 2 k 2 -k 3 k -1
f2 f3  f 
Theorem 21.2.2
(r)
Let p, r are positive numbers, m be a natural number, f be the r th order derivative function of f (x ),
<r>
f be arbitrary
r th order primitive function of f (x ) and B(p , m ) be the beta function. At this time,
if there is a number a such that

(a ) = 0 r [0 , m - p] or f (s)(a) = 0 s [0 , m -p -1]

f
<r>

the following expression holds for  = 2, 3, 4,  .

r  r  r 
m -1 r1 r-2 p r1 r-2
f
p + r1 r1 -r2
 (p )
 f  -1 + Rmp
r
f  = ΣΣ Σ  f
r1=0 r2=0 r-1=0 1 2 -1

m + k  k   k  k  k 
(-1) m
1 -p -1 m +k
 m + k1 k2 k k k -2 1 2 -2
Rmp = Σ Σ Σ Σ 
(-p , m) k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1
(p )
f   f  -1
m+ k 1 m+k 1-k 2 k 2-k 3 k
 f f

- 36 -
 cos m x , sin m x
Super Derivatives of
p
 
m (p ) 1 m /2
cos x = m-1 Σ mＣ r (m -2r) cos (m -2r)x +
p
2 r=0 2
 p
   
m (p ) 1 m /2
sin x = m-1 Σ mＣ r (m -2r) cos (m -2r) x -
p
+
2 r=0 2 2

22 Higher Derivative of Composition

22.1 Formulas of Higher Derivative of Composition

About the formula of the higher derivative of composition, the following formula was shown by Faà di Bruno about 150
years ago.

Formula 22.1.1 ( Faà di Bruno )

Let j1 , j2 ,  Bn,rf1 , f2 , 
(n)
, jn are non-negative integers. Let g , fn are derivative functions and are

the 2nd kind of Bell polynomials such that

(n =1, 2, 3, )
(n ) (n ) (n )
g = g ( f) , fn = f (x)
j1 j2 jn

     
n! f1 f2 fn
Bn,rf1 , f2 ,  , fn =Σ
j1! j2!  jn! 1! 2! n!

j1 + j2 +  + jn = r & j1 +2j2 +  + n jn = n 

Then, the higher derivative function with respect to x of the composition g {f (x )} is expressed as follows.
n
(n ) (r)
gf(x) = Σ
r=1
g Bn,rf1 , f2 ,  , fn

Next, the algorithm that generates Bn,rf1 , f2 ,  , fn without an omission is shown. And the derivatives up to
the 8 th order of z = g {f (x )} are calculated using this.
Even if the above algorithm is used, it is not so easy to obtain Bn,rf1 , f2 ,  , fn .
However, when the core function f (x )is the 1st degree, it becomes remarkably easy.

Formula 22.1.4
When g
(n)
=g
(n)
( f ) , fn = f
(n)
(x) (n =1, 2, 3, ), if f (x ) is the 1st degree, the higher derivative function
with respect to x of the composition g {f (x )} is expressed as follows.
(n ) (n ) n
gf(x) = g f1
And, this can be easily enhanced to the super-differentiation in this case. That is, the following expression holds
for p > 0 . This is the grounds for which we have used "Linear form" since "12 Super Deribative" as a fait accompli .
(p ) (p ) p
gf(x) = g f1

Next, trying the higher differentiation of some compositions, we obtain the following formulas.
n
f (x) (n )
e  Σ
f (x)
=e Bn,rf1 , f2 ,  , fn
r=1

(x) (n ) (x) n
e -f  = e -f Σ (-1)rBn,rf1 , f2 ,  , fn
r=1


(n ) n 1 r r
ge 
x
= ΣS(n,r)g (r)e rx , S(n,r) = r Σ(-1)s (r -s)
n
r=1 ! s=0 s

- 37 -
 (n ) n
ge -x = (-1)nΣS(n,r)g (r)e -rx
r=1
n
(n ) r-1
log f(x) = Σ(-1) (r -1)!Bn,rf1 , f2 ,  , fnf -r
r=1

 
(n ) n
(r) 0! 1! n-1 (n -1)!
g (log x) = Σ g Bn,r , - 2 ,  , (-1)
r=1 x x xn

f (x) (n) -f (x) (n)

In e  , e  , especially when f(z)= log(z) , we obtain the following formula. In addition, this
formula was discovered by Masayuki Ui living in Yokohama City in early December 2016.

Formula 22.3.1 ( Masayuki Ui )

When (z) is the gamma function, n(z) is the polygamma function and Bn,rf1 , f2 ,  are
Bell polynomials , the following expressions hold.
n
d n
(z) = (z)ΣBn,k  0(z) ,  1(z) ,  ,  n-1(z)
dz n k=1

dn 1 1 n
n (z)
= Σ (-1)k Bn,k  0(z) ,  1(z) ,  ,  n-1(z)
dz ( z) k=1

23 Higher Integral of Composition

Formula 23.1.1
<n > (k )
Let f = f(x), g be the lineal higher primitive function of g (f ) and h , h are the functions of f such that

dx 1 d kh
k = 1, 2, 3, 
(k )
h= = (1) , h =
df f df k
Let Sk , Mk , rk are the polynomials such taht

   r  r 
 
m k -1 -1 rk1 rk1 rk2 rk 2 rk k-1 rk k -1
Sk = Σ r Σ Σ  Σ k =1, 2, ,n
rk1=0 k1 r =0k2 k2 r =0k3 k3 r =0
kk rkk

 Σ Σ  Σ 

mk mk rk2 rk 2 rk3 rk 3 rk k-1 rk k -1
k =2, 3, ,n
mk
Mk = (-1) Σ
rk2=0 rk 2 rk3=0 rk 3 rk4=0 rk 4 rkk =0 rkk
j
Rj k =Σrik j,k = 1, 2, ,n
i=k
And let a , fa are the zeros of the lineal higher primitive functions of g {f (x )} , g h respectively.
Then, the lineal higher integral with respect to x of the composition g {f (x )} is expressed as follows.


x x R
 h
n+ Rn1 Rn1 -Rn2 -Rnn Rnn
 gf(x )dx
n
= S1S2Sn g  h n n-1
h + Rmn 1
a a

       g
f f f

  hdf  hdf  hdf

f
m1 m1 m1
Rmn 1 = (-1) h df (n-fold nest)
fa fa fa fa

     g
f f f

  hdf  hdf
1+ R11+m2 R11-R22+m2 R22
+ S1M2 h h df
fa fa fa

+S S M   g
f f

 
2+ R21 + m3 R21 -R32+m3 R32-R33 R33
1 2 3 h h h df  hdf
fa fa

- 38 -
 g
f
n-1+ Rn-1 1+ mn Rn-1 1-Rn2+mn Rn2-Rn3 Rnn
+ S1S2Sn-1Mn h h  h df
fa

If it writes down up to the 3rd order without using Sk , Mk , rk , it is s follows.

 r  g
x m 1-1 -1
1+ r11 r11
gf(x)dx =Σ h + Rm11
a r11=0 11

= (-1)  g
f
m1 m1 m1
Rm11 h df
fa

  r  r 
r21
r 
x x m 1-1 -1 m 2-1 r21 -1
g
2+ r11 +r21  r11 +r21 -r22 r22
gf(x)dx = Σ ΣΣ
2
h h
a a r11=0 11 r =0 r =0
21 22 21 22
+ Rm21

r  
m2
r 
m 1-1 -1 m2 f 1+ r11+ m2 r11-r22+ m2 r22
g
m2
Rm21 =Σ (-1) Σ h h df
r11=0 11 r =022 22 fa

 g h df hdf

f f
m1 m1 m1
+ (-1)
fa fa

  r  r   r  r  r 
r r r
r 
x x x -1
m 1-1 -1 -1 m 2-1 r21 21 m 3-1 r31 r32 31 32
gf(x)dx
3
=Σ Σ Σ Σ Σ Σ
a a a r11=0 11 r21=0 r22=0 21 22 r31=0 r32=0 r33=0 31 32 33

 g
3+ r11+ r21+ r31 r11+ r21+ r31-r22-r32 r22+ r32-r33 r33
h h h + Rm31

r  r 
m3
 r  Σ Σ  r  r 
m 1-1 -1 m 2-1 r21 -1 r21 m3
m3 r32 r32
Rm31 = Σ (-1) ΣΣ
r11=0 11 r21=0 r22=0 21 22 r32=0 r33=0 32 33

 g
f
2+ r11+ r21+m3 r11+r21-r22-r32+m3 r22 +r32-r33 r33
h h h df
fa

r     g
m2
r 
m 1-1 -1 m2 f f


m2 1+ r11+ m2 r11-r22+ m2 r22
+Σ (-1) Σ h h df hdf
r11=0 11 r =0
22 22 fa fa

+ (-1)    g
f f f

   
m1 m1 m1
h df hdf hdf
fa fa fa

Next, trying the higher integration of some compositions, we obtain the following formulas
n+ R
(1+)
  xn
n1
x x  
 
  1
x - c dx = x - c S S S n
1-

c 
c n
1 2 n
 1+ + n + Rn1 x
(1/ ) (1/ ) (1/ )
 
 1/ - Rn1+Rn2  1/  - Rn n-1 +Rnn  1/ - Rn n
n n
Σ (1+) + n+ Σ rk
 (log x) dx
x x    rk n

= x nΣΣΣ(-1)
rk
n k =1
Π k n (log x) k =1


 1+ + n +Σrk 
0 0 r1=0r2=0 rn=0 k=1
k=1

(1+ ) n x
n-k
+ (-1) k-1
Σ(-1) nＣ k
n! k=1 k
n
n

 n +Σrk
 (log x) n


x x    Σ rk n r n-1+Σ rk
1 k =1
dx n = -x nΣ Σ Σ(-1) k =1 Πk k n
k =1

 
log x
0 0 r1=0 r2=0 rn=0 k=1
 n +Σr k
k =1

- 39 -
 1 n -1

 
n-1 r
+ (x -1) log|log x| -Σ(-1)n-rn -1Ｃ r x log(n -r)
(n -1)! r=0
n

  (log x)n Σr n  1+n +Σrk n

 log|log x|dx

x x    Σ rk n rk k =1
+ k
n
= -x Σ Σ Σ(-1) Πk
n k 1=
n
k =1


 1+n +Σr

0 0 r1=0 r2=0 rn=0 k=1
k
k =1
1 n

n!  
log|log x| - (x -1) +Σ(-1) Ｃ ( + logr ) x
n r n-r
+ n r
r=1

Formula 23.1.1 is so complicated and is not applicable to any composition. The 1st, the inverse function of the core
function must be known. The 2nd, the higher primitive function of the enclosing function g (f ) must have the property

such as lim g <n>( f )= c . Considering these, there are not many compositions that Formula 23.1.1 is applicable.
n 
However, when the core function f (x ) is the 1st degree, it becomes remarkably easy.

Formula 23.1.2
When f (x )= cx +d ,

   
x x n f f
1
 n
gf(x)dx =  g( f )df n
a a c fa fa

And, this can be easily enhanced to the super integral in this case. That is, the following expression holds for p > 0.
This is the grounds for which we have used "Linear form" since " 07 Super Integral " as a fait accompli.

  c   g(f )df

x x
p 1 p f f p
gf(x)dx = f(x)= cx +d
a a fa fa

24 Sugioka's Theorem on the Series of Higher (Repeated) Integrals

Mikio Sugioka showed the following in his work " 数学の研究 ".

1. The series of the higher integral of a function f (x ) results in one integral of e x , sin x, cos x, sinh x, cosh x ,
2. The series of the higher integral of a function f (x ) is expanded into the series of the higher integrals of e x , etc.

  f(x)dx
x x
These are introduced and proved in this chapter. In this summary, we assume lim
m  a
 m
= 0. Also,
a
fa n  f (x)x=a
() (n)
we describe the n th order differential coefficient at a with

(1) Sum of the Series of Higher Integrals

   f(x)e dx
 x x x
Σ  f(x)dx r = ex -x
r=1 a a a

 = e  f(x)e dx
 x x x
Σ(-1)
r-1
 f(x)dx r -x x
r=1 a a a

a f(x)dx 2 
f(x)e dx + e  f(x)e dx
x


 x 1 x x
Σ x x
2r-1 -x -x
= e
r=1 a a a

  f(x)dx = cosx f(x)cosxdx + sin x f(x)sin xdx

 x x x x
Σ (-1) r-1
 2r-1
r=1 a a a a

a f(x)dx 2 
f(x)e dx - e  f(x)e dx
x


 x 1 x x
Σ x x
2r -x -x
= e
r=1 a a a

- 40 -
    f(x)sin xdx
 x x x x
Σ(-1) 
r-1
f(x)dx 2r = sin x f(x)cosxdx - cosx
r=1 a a a a

Example

 e dx + e dx + e dx  e dx e e

x x x x x x x x x
x x x x
+  = ex x -x
dx = e x(x -a)
2 3 4
a a a a a a a a a

    e dx +-  
x x x x x x x x
e xdx  e xdx +  e xdx   x
3 5 7
a a a a a a a a

= cos x e cos x dx + sin x e sin x dx =

x x x a
x e e x
- cos(x -a ) - sin(x -a )
a 2 2 a

   x dx + x dx  x dx + 
x x x x x x x x
x dx  
2 2 2 4 2 6 2 8
0 0 0 0 0 0 0 0

= sin x x cos x dx - cos x x sin x dx = x -2 + 2cos x

x x
2 2 2
0 0

(2) Integrals Series Expansion

a f(x)dx   e dx
 s*r

()  
x x (x -a) x x
Σ
r=0 a
r
=Σ
r=0
fa r Σ
s=0 (s + r) !
= Σ fa r
r=0
()

a a
x-a r

(-1)r f(x)dx = Σ f (a)Σ(-1)s (a ) e

 s+ r
x x   (x -a)  x x
Σ
r=0 a a
r
r=0
(r)
s=0 (s + r ) !
= Σf r
r=0
()
a a
-(x-a)
dx r

2s+1+ r

 a   sinh(x -a)dx
 
() (x -a) 
x x x x
Σ
r=1 a
 f(x)dx 2r-1 = Σ fa r Σ
r=0 s=0 (2s +1+ r) !
= Σfa r 
()
r=0 a a
r

   sin(x -a)dx
2s+1+
  () (x -a)  () x
x x x
Σ(-1)r-1 
r=1 a a
f(x)dx
2r-1
= Σfa r Σ(-1)s
r=0 s=0 (2s +1+ r) !
= Σ fa r 
r=0 a a
r

a   cosh(x -a)dx

2s+
 
( ) (x -a) 
x x x x
Σ
r=0 a
f(x)dx
2r
=Σ
r=0
fa r Σ
s=0 (2s + r) !
= Σfa r 
()
r=0 a a
2r

2s+ r
(-1)r f(x)dx = Σfa Σ(-1)s   cos(x -a)dx
 x x   (x -a)  x x
Σ
r=0 a a
2r
r=0
(r)
s=0 (2s + r) !
= Σfa r 
r=0
()
a a
r

Example
s+r

  
x x x x x x   (x -1)
log x + log x dx + log x dx 2 + log x dx 3 + = Σ(-1)r-1(r -1)!Σ
1 1 1 1 1 1 r=1 s=0 (s + r) !

= 0! e dx -1! e dx +2! e e

x x x x x x x x
x-1 x-1 x-1
dx -3!  x-1
2 3 4
dx +-
1 1 1 1 1 1 1 1

 secx dx  secx dx + secx dx - secx dx

x x x x x x x x
- 
3 5 7
0 0 0 0 0 0 0 0
2s+1+ 2r
 x 
= ΣE2rΣ(-1) s
E2r : Euler Number
r=0 s=0 (2s +1+ 2r) !

    sin x dx
x x x x x x
sin x dx  5  sin x dx  61 
2 4 6
= sin x + +
0 0 0 0 0 0

 tan   tan   tan

x x x x x x
x dx   x dx +  x dx + 
-1 -1 2 -1 4 -1 6
tan x +
1 1 1 1 1 1

- 41 -
 3r
2s 2s+ r
 
(x -1)  (r -1)!  (x -1)
= Σ +Σ
4 s=0 (2s) ! r=1 2 r/2
sin
4  Σ s=0 (2s + r) !

  cosh(x -1)dx
1 x 1 x x
2
= cosh(x -1) + cosh(x -1) dx - +
4 2 1 2 1 1

(3) Series of Higher Integrals with coefficients

   f(x)e dx
 x x x x x
Σr  f(x)dx r = ex f(x)e -xdx + e x -x 2
r=1 a a a a a

  f(x)dx  f(x)e dx - e  f(x)e dx

 x x x x x
r 
r-1 r x x
Σ = e -x -x 2
(-1)
r=1 a a a a a

Example

      e
x x x x x x x x x x x
dx 4  dx +  = e x e dx + e x
2 3 4 -x -x 2
1 dx +2 dx +3 dx
a a a a a a a a a a a

= e x-a(x -a)

     log x dx +-

x x x x x x x x
log x dx 4 
2 3 4
1 log x dx -2 log x dx +3
0 0 0 0 0 0 0 0

= e  log x e dx - e  log x e dx
x x x
-x x -x x 2
0 a a

=e
-x
x Ei(x)-   + e -x
-1

25 Series of Higher Integral with Geometric Coefficients

This chapter is a generalization of " 24 Sugioka's Theorem on the Series of Higher Integral "

  f(x)dx
x x
In this summary, we introduce the formulas in the case of lim c m  m
=0 (c > 0) .
m  a a

(1) Sum of the Series of Higher Integrals with Geometric Coefficient

  f(x)e
 x x x
Σc r  f(x)dx r = ce cx -cx
dx
r=1 a a a

  f(x)e
m x x x
Σ(-1)r-1c r 
r=1 a a
f(x)dx r = ce -cx
a
cx
dx

    f(x)e dx 
 x x c x x
Σc 2r-1 
r=1 a a
f(x)dx 2r-1 =
2
e cx
a
f(x)e -cxdx + e -cx
a
cx

 = c cos cx f(x)cos cx dx + c sin cx f(x)sin cx dx

 x x x x
Σ(-1) r-1c
r=1
2r-1
a

a
f(x)dx
2r-1
a a

c  f(x)dx
2 
f(x)e dx - e  f(x)e dx
 x x x x


c
Σ
r=1
2r
a a
2r
= e cx
a
-cx -cx
a
cx

(-1) r-1c  f(x)dx = c sin cx f(x)cos cx dx - c cos cx f(x)sin cx dx
 x x x x
Σ
r=1
2r
a a
2r
a a

Example

     e dx
x x x x x x x x
c
1
e x dx + c
2
e x dx + c
2 3
e x dx  c
3 4
 x 4
+
a a a a a a a a

ee ce cx e
(1-c)x (1-c)a
cx
x
x -cx -e 
= ce dx =
a 1-c

- 42 -
      e dx
x x x x x x x x
c
1
e xdx  c
3
e xdx + c
3 5
 e xdx  c
5 7
 x 7
+- 
a a a a a a a a

  e sin cx dx
x x
= c cos cx e xcos cx dx + c sin cx x
a a
x a
ce ce
= 2
- 2
[cos{c(x -a )}- csin{c(x -a )}]
1+ c 1+ c

  
  x dx + 
x x x x x x x x
x dx  c   x dx  c 
2 2 2 4 2 4 6 2 6 8 2 8
c x dx + c
0 0 0 0 0 0 0 0

= c sin cx x cos cx dx - c cos cx x sin cx dx =

x x 2 2
2 c x -2 + 2cos cx 2
2
0 c 0

(2) Series of Higher Integrals with coefficients

rc r f(x)dx r = ce cx f(x)e dx + c e cx f(x)e dx

 x x x x x
Σ
r=1 a a a
-cx 2
a a
-cx 2

(-1)r-1 rc  f(x)dx = ce  f(x)e dx - c e  f(x)e dx

 x x x x x
Σ
r=1
r
a a
r -cx
a
cx 2 -cx
a a
cx 2

Example

     dx + 
x x x x x x x x
dx  4c 
1 2 2 3 3 4 4
1c dx + 2c dx + 3c
a a a a a a a a

 e dx + ce  e
x x x
= ce cx cx
dx = ce c(x-a (x - a)
-cx -cx 2 )
a a a

(3) Calculation by Double Series


s+r
 x x   c s(x -a)
Σc r 
r ()
f(x)dx r = Σ f (a)Σ
r=0 a a r=0 s=0 (s + r)!

(-1) c  f(x)dx

s+ r

r r
x x
r
 (r) 
s c s(x -a)
Σ
r=0 a a
= Σ f (a)Σ(-1)
r=0 s=0 (s + r)!
2s+1+ r

 f(x)dx
2s+1
 x x   c (x -a)
Σ
r=1
c 2r-1
a a
2r-1
= Σ f (a)Σ
r=0
(r)
s=0 (2s +1+ r) !
2s+1+ r

 f(x)dx
2s+1
 x x   c (x -a)
Σ
r=1
(-1) r-1 2r-1
c
a a
2r-1
= Σ f (a)Σ(-1)
r=0
(r)
s=0
s
(2s +1+ r) !

 f(x)dx = Σf (a)Σ (2s + r)!

x x 2s 2s+ r
  (r)  c (x -a )
Σ 2r 2r
c
r=0 a a r=0 s=0

Σ(-1) c  f(x)dx =Σf (a)Σ(-1) (2s + r)!

x x 2s 2s+ r
 c (x -a)  (r) 
r 2r 2r s
r=0 a a r=0 s=0

Example

   log x dx
x x x x x x
1 2 2 3 3
log x + c log x dx + c log x dx + c +
1 1 1 1 1 1
s+r

r-1  c s(x -1)
= Σ(-1) (r -1)!Σ
r=1 s=0 (s + r)!

- 43 -
  secx dx  c  secx dx  secx dx  secx dx
x x x x x x x x
- 
1 3 3 5 5 7 7
c +c -c
0 0 0 0 0 0 0 0
2s+1 2s+1+ 2r
  c x
= ΣE2rΣ(-1)s E2r : Euler Number
r=0 s=0 (2s +1+ 2r) !

    tan
x x x x x x
tan x dx  c  tan x dx  c  x dx + 
-1 2 -1 2 4 -1 4 6 -1 6
tan x + c
1 1 1 1 1 1

3r
2s 2s 2s 2s+ r
  c (x -1)  (r -1)!  c (x -1)
=
4Σ
s=0 (2s) !
+Σ
r=1 2r/2
sin  4 Σ
s=0 (2s + r) !

26 Higher and Super Calculus of Zeta Function etc

In this chapter, n is a natural number, p is a complex number, (p) is gamma function, (p) is digamma
function, Hn is a harmonic number and r is a Stieltjes constant defined by the following equation.

 
n (log k)r (log n)r+1
r = nlim

Σ -
k=1 k r +1
Then, the following higher calculus and super calculus are obtain. These are all lineal.

26.1 Higher and Super Calculus of Riemann Zeta Function

n-1 r+n
(z -1)  (z -1)

<n >
(z) = n-1 Σ
r

(n -1) ! 
log(z -1) - H + (-1) r
r=0 (r + n )!
log(z -1)- (p) - 0  (z -1)r+p
 (z -1)p-1 + Σ(-1)r r
<p >
(z) =
(p) r=0 (1+ r + p)
(-1)-n n !  (z -1)r-n
 (z) ( ) r
(n ) r
= + Σ -1
1+ r - n
(z -1)n+1 r=0

log(z -1) -(-p) - 0  (z -1)

r-p

(p )
+ Σ(-1) r
-p-1 r
(z)= (z -1)
(-p) r=0 (1+ r - p)

26.2 Higher and Super Calculus of Dirichlet Lambda Function

n-1
(z -1)
 <n>(z) =
2(n -1) ! 
log(z -1) -Hn-1

log r+12 r-1 r r+n

  
1  (z -1)
+ Σ(-1)r r + -Σ s (log 2) r-s
2 r=0 r +1 s=0 s (r+n )!
log(z -1) -(p) - 0
 <p>(z) = (z -1)
p-1
2(p )
log r+12 r-1 r r+p

  
1  (z -1)
+ Σ(-1)r r + -Σ s (log 2) r-s
2 r=0 r +1 s=0 s  (1+ r + p )
-n
(-1) n !
 (n)(z) =
2(z -1) n+1
log r+12 r-1 r r-n

  
1  (z -1)
+ Σ(-1) r +
r
-Σ s (log 2) r-s
2 r=0 r +1 s=0 s (1+ r - n)
log(z -1) -(-p) - 0
 p (z) =
( )
(z -1)-p-1
2(-p)
log r+12 r-1 r (z -1)r-p
   (1+ r - p)
1 
+ Σ(-1) r +
r
-Σ s (log 2) r-s
2 r=0 r +1 s=0 s

- 44 -
26.3 Higher and Super Calculus of Dirichlet Eta Function

Formula 26.3.1h ( Higher Integral )

zn 
-n
log r
 n (z) =
< >
+ (-1)nΣ(-1)r-1 Re(z) > 0
n! r=2 rz
r-1
zn log -nr
r 
 k (-1) k

<n >
(z) = + (-1) ΣΣ
n
n! k =2 r=2 2k+1 rz

Formula 26.3.1s ( Super Integral )

zp 
-p
log r
 (z) = + e p iΣ(-1)r-1
p < >
Re(z) > 0
(1+p) r=2 rz
r-1
zp log -p r
r 
p i
 k (-1) k

<p >
(z) = + e ΣΣ
(1+p) k=2 r=2 2k+1 rz

Formula 26.3.2h ( Higher Derivative )

z -n 
n
r-1 log r

(n )
(z) = + (-1) Σ(-1)
-n
Re(z) > 0
(1-n) r=1 rz
r-1
z -n log nr

-n 
k (-1) k

(n )
(z) = +(-1) ΣΣ
(1-n) k=2 r=2 2k+1 r rz

Formula 26.3.2s ( Super Derivative )

z -p  log p r
 + e -p iΣ(-1)r-1
(p )
(z) = Re(z) > 0
(1-p) r=1 rz
r-1
z -p log p r

-p  i 
k (-1) k

(p )
(z) =
(1-p)
+e Σ Σ
k=2 r=2 2k+1 r rz

26.4 Higher and Super Calculus of Dirichlet Beta Function

Formula 26.4.1h ( Higher Integral )

zn 
-n
log 2r -1

<n >
(z) = + (-1)nΣ(-1)r-1 Re(z) > 0
n! r=2 2r -1
z

-n
zn  k (-1)r-1 k


log (2r -1)
 (z) =
<n>
+ (-1) ΣΣ
-n
n! k=2 r=2 2k+1 r (2r -1)
z

Formula 26.4.1s ( Super Integral )

zp  log -p 2r -1
 <p>(z) = + e p iΣ(-1)r-1 Re(z) > 0
(1+p) r=2 2r -1
z

zp  k (-1)r-1 k -p


log (2r -1)
 <p>(z) = + e p iΣΣ
(1+p) k=2 r=2 2k+1 r (2r -1)
z

- 45 -
Formula 26.4.2h ( Higher Derivative )
-n n

z r-1 log (2r -1)
 (n)(z) = + (-1) Σ(-1)
-n
Re(z) > 0
(1-n) r=2 (2r -1)
z

-n  k (-1)r-1 k log n (2r -1)


z
 (n)
(z) = + (-1)-nΣΣ k+1
(1-n) k=2 r=2 2 r (2r -1)
z

Formula 26.4.2s ( Super Derivative )

-p p
z -p  i  r-1 log (2r -1)
 (z) =
(p )
(1-p)
+e Σ
r=2
(-1)
(2r -1)z
Re(z) > 0

r-1
z -p k log p (2r -1)
-p  i 

k (-1)
 (z) =
(p )
(1-p)
+e Σ Σ k+1 r (2 -1)z
k =2 r=2 2 r

2011.04.16
2012.10.28 Renewal
2016.01.05 Updated
2018.04.04 Added Chap.25
2018.06.11 Added Chap.26
Kano Kono
Alien's Mathematics

- 46 -
1 Gamma Function & Digamma Function

1.1 Gamma Function

The gamma function is defined to be an extension of the factorial to real number arguments. By this,
for example, a definition of (1/2) ! and the calculation is enabled.

1.1.1 Gauss expression

n! nz
(z) = lim (1.1)
n  z(z +1)(z +2)(z + n)

Calculation.
Deform (z -1)! as follws

z! 1 123 z (z +1)(z +2)(z +n)

(z -1)! = = 
z z (z +1)(z +2)(z +n)
(n +1) (n +2) (n +z)
n !n z 
n !(n +1)(n +2)(n +z) n n n
= =
z(z +1)(z +2)(z +n) z(z +1)(z +2)(z +n )
Because natural number n may be arbitrary, when n  , as follows:
n +1 n +2 n +z
 1,  1 , , 1
n n n
Hence
(n +1) (n +2) (n +z)
n !n z 
n n n n !n z
lim = lim
n  z(z +1)(z +2)(z + n) n  z(z +1)(z +2)(z + n )

Since z is already good in anything besides 0, -1, -2,  , replacing function z -1! with z ,
we obtain the desired expression

1.1.2 Euler expression

  
z -1


1  1 z
(z)= Π 1+ 1+ (2.1)
z n =1 n n

Calculation.
Further deform (1.1) as follows
z
n !n
lim
n  z(z +1)(z +2)(z + n )

 
1 123 n 2 3 n
= lim    
n  z (z +1)(z +2)(z + n) 1 2 n -1
z

 
1 123 n 2 3 n +1 n +1
= lim      1
n  z (z +1)(z +2)(z + n) 1 2 n n

-1-
 z

    
1 1 1
1+ 1+  1+
1 1 2 n
= lim 
    
n  z z z z
1+ 1+  1+
1 2 n

  
z -1


1  1 z
= Π
z n =1
1+
n
1+
n

1.1.3 Integral expression

e

(z) = -t z-1
t dt Re(z)>0 (3.1)
0

Proof.


n

t
n t
n(z) = 1- z-1
dt (3.2)
0 n
Here
n

 
t
lim 1- = e -t
n  n
so

 e
n 

t
n t z-1 -t z-1
lim 1- dt = t dt (3.3)
n  0 n 0

Further, in (3.2) let t = ns . Then because t :0n  s :01 , dt = nds

   (1-s)
n


n t 1 1
1- t z-1 dt = (1-s)n n z-1 s z-1 n ds = n z n
s z-1 ds
0 n 0 0
1
n zn
   (1-s)
sz 1
n-1
=n z
(1-s)n + s z ds
z 0 z 0
1
n zn n zn (n -1)
   (1-s)
s z+1 1
= (1-s)n-1 + n-2
s z+1 ds
z z +1 0 z(z +1) 0


n zn(n -1)321 n z n !
s
1
z+n-1
= ds =
z(z +1)(z +2)(z +n -1) 0 z(z +1)(z +2)(z +n)
Hence


n


n t n! nz
lim 1- t z-1 dt = lim
n  0 n n  z(z +1)(z +2)(z +n)
Thus, from (1.1),(3.3),(3.4) we obtain

-2-
 e

(z) = -t z-1
t dt Re(z)>0 (3.1)
0

1.1.4 Weierstrass expression

z

 e
1  z -
= e  z z Π 1+ n
(4.1)
(z) n =1 n

Calculation.
We employ a reciprocal of Gauss expression (1.1). Then

z(z +1)(z +2)(z +n)

    
1 z z z
=  z 1+ 1+  1+
n zn ! nz 1 2 n

 n
1 1 1
1+ + ++ z z z
2 3

     e
e z z - z -
=  z 1+ e -z 1+ e 2  1+ n
elog n z 1 2 n

1+ 2 + 3 ++ n - log n z z

1 1 1 z z

     e
z z - z -
=e 1+ e -z
1+ e 2  1+ n
1 2 n

 z
1 1 1 z z

     e
1+ + ++ - log n z z z - z -
 lim e 2 3 n
1+ e -z 1+ e 2  1+ n
n  1 2 n
z

 


 
z z - 1 1 1
=e zΠ 1+ e n
where  = lim 1+
n 
+ ++ - log n
n =1 n 2 3 n
z

 e
1  z -
 = e  z z Π 1+ n
(z) n =1 n

1.1.5 Properties of the Gamma Function (The 1)

(z +1) = z z  (5.1)

(1-z) = -z -z  (5.1')

(1) = (2) = 1 (5.2)

(n +1) = n ! n is a nonnegative integer (5.3)

(z + n)
= z(z +1)(z +2)(z + n -1) n is a natural number (5.4)
(z)
(z)
= (z -1)(z -2)(z - n) n is a natural number (5.4')
(z - n)
(-z) (1+z +n)
= (-1)-n n is a nonnegative integer (5.5)
(-z -n) (1+ z)
(z +n)
lim =1 (5.6)
n  (n )n z

-3-
Proof.
From Gauss expression (1.1)

(z +1) n ! n z+1z(z +1)(z +2)(z +n)

 
n
= lim = lim z  =z
(z) n  (z +1)(z +2)(z +n +1)n ! n z n  z +n +1
Then (5.1) follows. And reversing the sign of (5.1), we obtain (5.1').

Next,
1 1 n +1

k =1  e     
n 1 - 1 1 1 - -
Π 1+ k
= 1+ 1+  1+ e -1e 2  e n
k 1 2 n
n +1 -1+ 2 + + n  (n +1) Log n -1+ 2 + + n 
1 1 1 1
2 3
=   e = e e
1 2 n n
(n +1) -1+ 2 + + n - Log n 
1 1

= e
n
1 1

n =1    
 1 - n 1 -
 Π 1+ e n
= lim Π 1+ e k = e -
n n  k =1 k
Substitute this for Weierstrass expression (4.1) as follows.
1

 e
1  1 -
= e  1 1 Π 1+ n
= e   e - = 1
(1) n =1 n
Hence (1)=1 , and from (5.1) (2)=1(1)=1 i.e. (5.2) follows.
Substitute positive integer n for (5.1) one by one as follows.

(n +1) = n(n) = n(n -1)(n -1) =  = n !(1) = n !

i.e. (5.2) follows. And because (1)=0! (5.3) holds also for n = 0 .
From (5.1)
(z +1) (z +2) (z +n)
=z, = z +1 ,  = z +n -1
(z) (z +1) (z +n -1)
Multiply these on each other as follows.
(z +1) (z +2) (z +n)
  = z(z +1)(z +2) (z +n -1)
(z) (z +1) (z +n -1)
Hence (5.4) follows. and in (5.4) by replacing z with z -n , (5.4') follows.
Let replace z with -z in (5.4') , then

(-z)
= (-z -1)(-z -2)(-z -n )
(-z -n)
(1+z +n)
= (-1)-n(z +1)(z +2)(z +n) = (-1)-n
(1+z)
This equation holds also for n = 0. Moreover, this equation holds obviously for the positive integer z .
Then,this quation includs a part of Singular Point Formulas ( Later 1.3 ).
From (5.4)

(z +n) = z(z +1)(z +2) (z +n -1) ( z )

-4-
 1
= z(z +1)(z +2) (z +n -1)(z +n) ( z ) 
(z +n)
(z +n) z(z +1)(z +2) (z +n -1)(z +n) ( z )
 lim = lim
n  (n -1)! n z n  (n -1)! n z(z +n)
z(z +1)(z +2) (z +n -1)(z +n) n
= lim  (z)
n  n !n z z +n
1
= 1(z) = 1
(z)
By substituting n -1! = n for this, (5.6) follows.

1.1.6 Properties of the Gamma Function (The 2)


 
1
 n+ = (2n -1)!!
2 2n
2n 
 
1
 -n = (-1) n
2 (2n -1)!!

n +  =  
1 147(3n -2) 1
3 3 n
3

   
2 258(3n -1) 2
 n+ = 
3 3n 3

   
1 159(4n -3) 1
 n+ = 
4 4n 4

   
3 3711(4n -1) 3
 n+ = n

4 4 4
 
(z)(-z) = - , (z)(1-z) =
z sin  z sin  z
 z
  
1 1
 +z  -z = , (1+z)(1-z) =
2 2 cos  z sin  z
22z-1
 
1
(2z) = (z) z+
 2
33z-1/2
  
1 2
(3z) = (z) z+  z+
2 3 3

Proof.: Omitted.

1.1.7 Logarithmic Function & Gamma Function

 '(z)

x+1
dz = log x (7.1)
x (z)

-5-
  '(z)

n+1
dz = log(n !) (7.2)
1 (z)

Calculation.
Just do logarithm integral calculus obediently, as follows:

 '(z)

x+1
dz = [log(z)]x+1 = log (x +1) - log(x)
x (z) x

= logx(x) - log(x) = log x + log(x)- log(x)

= log x
Also

 '(z)

n+1
dz = [log(z)]n+1 = log (n +1) - log(1)
1 (z) 1

= log(n !) - log 1
= log(n !)

1.1.8 2nd order differential calculus of the Gamma Function

' 2 2
 "(z)  '(z)  '(z)  '(z)
  +  = Σ (n + z) +  
 1
= (8.1)
(z) (z) (z) n =0 2 (z)

Calculation.
'
 '(z)  "(z)(z) -  '(z)2  "(z)  '(z)
 
2

(z)
=
(z)2
=
(z)
-
 (z) 
From this
' 2
 "(z)  '(z)  '(z)
(z)
=  (z)  + (z) 
On the other hand, from next section (2.8 Trigamma Function)

 '(z) '

 
d  1
=  (z) = Σ
(z) dz n =0 (z + n )2

Hence (8.1) follows.

1.1.9 Spcial Values of the Gamma Function

From properties of the gamma function ( 1.1.5 , 1.1.6 ), the following special values are obtained.
Because these are used frequently, we write here.

(1) The 1
(0) =
(1) =1 , (2) =1

-6-
   2   2
1 -1!! 3 1!! 1
 =  =  ,  =  = 
2 0 2 1 2

  =   =
5 3!! 3 7 5!! 15
 =  ,  = 
2 2 4 2 2 2 8 3

  = 
2 
9 7!! 105 11 9!! 945
 =  , =  = 
2 2 16 4
2 325

  = 2.678938   = 1.354118
1 2
,
3 3

  = 3.625600   = 1.225417
1 3
,
4 4

(2) The 2
22
   
1 21 3 4
 - = -  = -2  ,  - =  = 
2 1!! 2 3!! 3
24
   
5 23 8 7 16
 - =-  =-  ,  - =  = 
2 5!! 15 2 7!! 105
26
   
9 25 32 11
 - =-  =-  ,  - = 
2 9!! 945 2 11!!

-7-
1.2 Digamma Function

1.2.1 The definition of the Digamma Function

1st order derivative of the logarithm of the gamma function (z) is called Digamma Function, and is defined
in the following expression.
d  '(z)
 (z)= log(z) = (1.0)
dz (z)

 
 e -t e -zt
 (z)= - dt (integral expression) (1.0')
0 t 1-e -t
In addition, the derivative more than 2nd order are callecd Trigamma ,Tetragamma, Pentagamma, etc.,
and generaly, n th order derivative is called Polygamma Function.
These are dnoted like  z  , z  ,  ,
(0) (1) (2) (n)
z  , z  including the digamma, and named
generally with Psi Function.

1.2.2 Properties of the Digamma Function

   
1  1 1 n 1
 (z) = - - +Σ - = lim log n -Σ (2.1)
z n =1 n n + z n  k =0 k + z

 (1)= - ,  (2) = 1-  (2.2)

 
1  1 1
 (z +1) =  (z) + = - + Σ - (2.3)
z n =1 n n+ z
n -1 1 n 1
 (z +n) =  (z) + Σ ,  (z -n) =  (z) - Σ (2.4)
k =0 z + k k =1 z - k

n 1
 (1+n) = - + Σ (2.5)
k =1 k

 
1
 = - - 2 log 2 (2.6)
2

 
1 n -1 1
  n = - - 2 log 2 + 2Σ (2.7)
2 k =0 2k+ 1

d  1
 ( 1 ) (z) =  (z) = Σ (Trigamma Function) (2.8)
dz n =0 (z + n )2

where  = 0.577215664901532860606512090082402431042

Proof
Let invers Weierstrass expression in the previous section (1.1.4 ) as follows.

 
z
 n
- z
(z) = e z Π
-1
e n
n =1 n +z
And differentiate the logarithm of this with respect to z, then

-8-
  '(z)  e -rz z -2  e z/n / n  n /(n + z)
2
= - - z - -1 + Σ z/n - Σ
(z) e z n =1 e n =1 n /(n + z)

 
1  1 1
= - - +Σ -
z n =1 n n+ z
Hence the first half of (2.1) follows.
Next,

 
1  1 1  1  1
- - +Σ - = - + Σ -Σ
z n =1 n n+ z n =1 n n =0 n + z

 
n 1  1  1
= - lim Σ - log n + Σ -Σ
n  k =1 k n =1 n n =0 n + z

 
n 1
= lim log n - Σ
n  k =0 k + z

Hence the latter half of (2.1) follows.

Substitute z =1, z =2 for the first half of (2.1), then

1 1 1 1 1 1 1
 (1) = - - + - + - + - + = -
1 1 2 2 3 3 4
1 1 1 1 1 1 1 1 1
 (2) = - - + - + - + - + - +
2 1 3 2 4 3 5 4 6

    = 1 -
1 1 1 1 1 1
= 1 - - + + + + + + +
2 3 4 2 3 4
Thus (2.2) follows.
Differentiate both side of (z +1) = z z(the previous (5.1)) with respect to z , then
 '(z +1) = z  ' z+ z
Divide both side of this by (z +1) = z  z , then

 '(z +1)  '(z) 1 1

= + , i.e.  (z +1) =  (z) +
(z +1) (z) z z
Hence the first half of (2.3) follows, and by substituting (2.1) for this, the latter half of (2.3)
follows.

z +1, z +2,  , z +n -1 and z, z -1,  , z -n for (2.3) sequentially; then

Next, substitute

1 1
 (z +1) -  (z) = ,  (z) -  (z -1) =
z z -1
1 1
 (z +2) -  (z +1) = ,  (z -1) -  (z -2) =
z +1 z -2

1 1
 (z +n) -  (z +n -1) = ,  (z -n +1) -  (z -n) =
z +n -1 z -n
Add these on each other, then

-9-
 n -1 1 n 1
 (z +n) -  (z) = Σ ,  (z) -  (z -n) = Σ
k =0 z + k k =1 z - k

Hence (2.4) follows.

Substitute z =1 for the first half of (2.4); then
n -1 1 n 1
 (1+n) =  (1) + Σ = - + Σ
k =0 1+ k k =1 k

Hence (2.5) follows.

Substitute z = 1/2 for (2.1) ; then

 
1  1  2  1  1
 = - + Σ -Σ = - + 2Σ - 2Σ
2 k =1 k k =0 2k +1 k =1 2k k =0 2k +1

 (-1)k-1
= - - 2Σ = - - 2 log 2
k =1 k
Hence (2.6) follows.
Substitute z = 1/2  n for (2.1) ; then

 
1  1  1
 + n = - + Σ - Σ
2 k =1 k k =0 1
k+ +n
2
1   1 n -1 1 n -1 1
= - + 2Σ - 2Σ - 2Σ + 2Σ
k =1 2k k =0 2k +1+2n k =0 2k +1 k =0 2k +1

 1  1 n -1 1
= - + 2Σ - 2Σ + 2Σ
k =1 2k k =0 2k +1 k =0 2k +1

n 1
= - - 2 log 2 + 2Σ
k =1 2k -1

 
1  1  1  1  1
 - n = - +Σ -Σ = - +2Σ -2Σ
2 k =1 k k =0 1 k =1 2k k =0 2k +1-2n
k+ -n
2
 1 n -1 1  1
= - + 2Σ - 2Σ - 2Σ
k =1 2k k =0 2k +1-2n k =n 2k +1-2n

 1 n 1  1
= - + 2Σ + 2Σ - 2Σ
k =1 2k k =1 2k -1 k =1 2k +1

n 1
= - - 2 log 2 + 2Σ
k =1 2k -1

Hence (2.7) follows.

Finally,

  
d d 1  1 1
 (z) = - - + Σ -
dz dz z n =1 n n + z
1  1  1
= 2 +Σ = Σ
z n =1 (n +z)2 n =0 (n + z)2

Hence (2.8) follows.

- 10 -
1.2.3 Spcial values of the Digamma Function
From properties of the digamma function (1.2.2 ), the following special values are obtained.

(1) The 1
 (0)= -
 (1)= - ,  (2) = - + 1
3 11
 (3)= - + ,  (4) = - +
2 6
5 29
 (5)= - + ,  (6) = - +
4 20
97 739
 (7)= - + ,  (8) = - +
60 420

(2) The 2

 
1
 = - - 2 log 2
2

   
3 1
= - = - - 2 log 2 + 2
2 2

   
5 3 4
= - = - - 2 log 2 + 2
2 2 3

   
7 5 23
= - = - - 2 log 2 + 2
2 2 15

   
9 7 176
= - = - - 2 log 2 + 2
2 2 105

   
11 9 563
= - = - - 2 log 2 + 2
2 2 315

   
13 11 6508
= - = - - 2 log 2 + 2
2 2 3465

- 11 -
1.3 Singular Point Formulas

Formula 1.3.1
When (z),  (z),  m(z) denote gamma function, digamma function and polygamma function
respectively, following expressions hold for n =0, 1, 2, 3,  .
(0)
= (-1)n(1+n) = (-1)n n! (1.1)
(-n)
(-n) m!
= (-1)m-n m =0, 1, 2, 3,  (1.2)
(-m) n!
 (-n)
= (-1)n+1n! (1.3)
(-n)
 (-n)
=1 (1.4)
 -(n +1)
 m(-n)
 
dm
=1  m(z)= m  (z) (1.5)
 m-(n +1) dz

Proof
From Gauss expression ( 1.1.1 ), the following expression holds.

(z +1) n ! n z+1 z (z +1)(z +2)(z +n)

= lim
(z) n  n ! n z (z +1)(z +2)(z +n +1)

When z = -1, -2 , this expression becomes the indeterminate form, and the value is not decided.
Now assume z  -1 ,  -2 ,  , then z +1 , z +2 ,  , z + n become nonzero all.

Hence, we can divide the numerator and the denominator by z +1 , z +2 ,  , z +n as follows:

(z +1) nz z
= lim = lim =z (1.0)
(z) n  z +n +1 n  z 1
+1+
n n
Here substitute z =0 for this; then

(z +1) (1)

= =0
(z) (0)
Because we have supposed it to be z -1 , we cannot substitute z =-1 for (1.0).
Therefore let z  -1 . Then
(z +1) (0)
lim = = -1
z-1 (z) (-1)
Similarly let z  -2, -3,  , -k,  , then
(-1) (-2) (-k +1)
= -2, = -3,  , = -k, 
(-2) (-3) (-k)
Multiply these from 1 to k ; then
(0) (0) (-1) (-k +1)
=  = (-1)k k ! = (-1)k (1+ k)
(-k) (-1) (-2) (-k)

- 12 -
Replacing k with n gives (1.1). And using this

(-n) (-n) (0) (-1)m m! m-n m !

= = = (-1)
(-m) (0) (-m) (-1)n n! n!
i.e. we obtain (1.2). This expression also holds for m =0, n =0 .
Next
n! n z
(z) = lim
n  z(z +1)(z +2)(z +n)
d log(z)
  
1 1 1 1
 (z) = = lim log n - + + ++
dz n  z z +1 z +2 z+n
From these two expressions

 
1 1 1 1
log n - + + ++
 (z) z z +1 z +2 z+n
= lim
(z) n  n! n z
z(z +1)(z +2)(z +n)
Assum z 0 and multiply z + k k =0, 1, 2,  by numerator and denominator; then

 
z+k z+k z+k z+k z+k
(z + k)log n - + ++ +1+ ++
 (z) z z +1 z + k -1 z + k +1 z+n
= lim
(z) n  n! n z

z(z +1)(z + k -1)1(z + k +1)(z +n)

Here let z  -k k =0, 1, 2,  , then as follows.

 
0 0 0 0 0
0log n - + ++ +1+ ++
 (-k) -k - k +1 -1 1 n
= lim
(-k) n  n! n -k
(-k)(-2)(-1)112(n -k)
(-k)(-2)(-1)112(n -k)
= lim -
n 
n! n -k
Calculate this in detail ;
When z  -k k =0
 (-k) n 0n!
= lim - =1
(-k) n  n!
When z  -k k =1, 2, 3, 
 (-k) k+1 nk
= lim (-1) k !
(-k) n  (n -k +1)(n -k +2)(n -k +k)
1
= lim (-1)k+1k !
    
n  k 1 k 2 k k
1- + 1- +  1- +
n n n n n n

- 13 -
 = (-1)k+1k !
Hence Replacing k with n gives (1.3).
Next using (1.2),(1.3)
 (-n)  (-n) (-n)  -(n +1)
=
 -(n +1) (-n) -(n +1)  -(n +1)
1
= (-1)n+1n!-(n +1) n+2
=1
(-1) (n +1)!
Thus we obtain (1.4).
Finally,

z 
1 1 1
 m(z) = (-1)m+1m! m+1
+ + +
(z +1)m+1 (z +2)m+1
From this
1 1 1 1
+ + + +
 m-(z +1) ( z )
- -1 m+1
( z)
- m+1
( z )
- +1 m+1
( z )
- +2 m+1
=
 m(-z) 1 1 1
+ + +
(-z)m+1 (-z +1)m+1 (-z +2)m+1
Assuming z 0 , denominator is also non-zero and positive. Then
1
 m-(z +1) (-z -1)m+1
=1+
 m(-z) 1 1 1
+ +  + +
(-z)m+1 (-z +1)m+1 (-z + n)m+1
1 1
Let z n then
m+1
 . Therefor the summation of terms after
m+1
-z +n -z +n +1
converges to a finite value.  m >0 . Consequently, second term in the right hand side converges to 0.
Thus
 m-(n +1)  m-(z +1)
= lim =1
 m(-n) zn  m(-z)
i.e. (1.5) was poroved.

Examples
The results actually calculated with computational software are as follows.

- 14 -
Significance of these formulas
z = 0, -1, -2 are singular points (pole of order 1). The value of these functions is  . Nevertheless,
the arbitrary ratio of z,z on these points is removable singular point. esides, all those ratios reduce
to the integer or the reciprocal number of the integer. formulas (1.1)  (1.4) mentioned above insists on this.

For exsample
(-3) (-3) (0) -7!
= = = 840
(-7) (0) (-7) -3!
(-3) (-3) (-7) -7! 1 1
= = =
 (-7) (-7)  (-7) -3! (-1)87! 6
 (-8)  (-8) (-8) -5!
= = (-1)98! = 120
(-5) (-8) (-5) 8!
etc.
These are phenomena which is peculiar to the gamma function and the digamma function and do not occur
in the polygamma functon more than the trigamma function.
For example, if we adopt the ratio 1z/0z of the trigamma functkion and the digamma function,

z =0, -1, -3,  are all poles of order 1 of this function (ratio) , and generaly, if we adopt the ratio

mz /nz m >n , z =0, -1, -2, -3,  become poles of order m -n . Functions more than

the trigamma are right different from Functions less than the digamma in a dimension.
What I can say in polygamma functions more than the trigamma is only that the ratio of polygamma functions

of the same dimension becomes 1 in z =0, -1, -2, -3,  entirely.

It is indispensable for Super Calculus ( non-integer order calculus ) of the power function and the logarithmic
function that the ratio between singular points of the gamma function or the digamma function is a rational
number.

2003.12.19
K. Kono
Alien's Mathematics

- 15 -
2 Multifactorial

2.1 Double Factorial

2.1.1 The definition of a Double Factorial

Definition 2.1.1
When m, n denote a natural number,
m !!  n (n -2)(n -4)  531 m =2n -1 (11)

 n (n -2)(n -4)  642 m =2n (10)

1 m =0, -1 (1- )

Example.
2!! = 2, 4!! = 24, 6!! = 246, 
1!! = 1, 3!! = 13, 5!! = 135, 
0!! = (-1)!! = 1

2.1.2 The basic formulas of a double factorial

Formula 2.1.2
When n, and (z) denote a natural number, a circular constant and a gamma function respectively,
the following expressions hold.
n! = n!!(n -1)!! (2)

     / 
1 1 1
(2n -1)!! = 2n n + / = 2n n + (31)
2 2 2

    = 2 (n +1) = 2 n !
2 2
(2n)!! = 2n n + / n n
(30)
2 2
(-2n)!! =, 0!! = 1 (4)
2
-(2n +1)!! = (-1)-n , (-1)!! = 1 (5)
(2n -1)!!

Proof
First, calculate (11)  (10), as follows.
(2n)!!(2n -1)!! = (2n -1)(2n -3)(2n -5)  531
 2n (2n -2)(2n -4)  642 = (2n)! (2)
Replacing 2n with n, we obtain (2).

Next ,since (1+ z) = z z,

   
1 1 1
 1+ = 
2 2 2

   =    
1 3 3 3 3 1 1
 2+ =  1+  = 
2 2 2 2 2 2 2

   =    
1 5 5 5 5 3 1 1
 3+ =  1+  = 
2 2 2 2 2 2 2 2

-1-
 

     
135(2n -1)
1 1 (2n -1)!! 1
 n+ n
=  = n

2 2 2 2 2

     
1 1 1
 (2n -1)!! = 2  n + / = 2n n + / 
n
(31)
2 2 2
In a similar way, we obtain (30).
Substitute n =0 for (30) ; then

0!! = 20(1) = 1
Replace n with -n in (30) ; then

(-2n)!! = (1-n)2-n =  (approach from +) (4)

0
Next, from (2) and (3 )
(2n)! (1+2n)
(2n -1)!! = = n (w)
(2n)!! 2 (1+n)
Substitute n =0 for (w) , then
(1) 0!
(-1)!! = = =1
(1)2 0 0!1
This secures the justification of the definition (1- ) conjointly with 0!!=1 .
Finally, replace n with -n in (w) , then

(1-2n) 2n-(2n 1)

-(2n +1)!! = =
2-n(1-n) -(n -1)
According to the Formula 1.3.1 in 1.3 (Singular Point Formulas) ,
 -(2n 1) (n -1)! (n -1)!
= (-1)(n -1)-(2n -1) = (-1)-n
-(n -1) (2n -1)! (2n -1)!
Then
-n (n -1)!2n 2 n!2n
-(2n +1)!! = (-1) = (-1)-n
(2n -1)! (2n)!

 
2 (2n)!
= (-1)-n  = (2n -1)!! (5)
(2n -1)!! 2nn !
Example 1

     
1 1 15
13!! = (27-1)!! = 27 7+ / = 27 / 
2 2 2
= 1281871.25430543/1.77245385 = 135135.0000
Example 2
2 2 2
(-3)!! = (-1)1 = -2 , (-5)!! = (-1)2 =
1!! 3!! 3

2.1.3 Expression by the double factorial of Maclaurin expansion

Formula 2.1.3
f (0) 0 f '(0) 1 f "(0) 2 f (3)(0) 3
 
x
f = x + x + x + x + (6)
2 0!! 2!! 4!! 6!!

-2-
Calculation
f '(0) 1 f " (0) 2 f (3)(0) 3
f(x) = f(0)+ x + x + x +
1! 2! 3!
' " (3)
f (0) 0 f (0) 1 1 f (0) 2 2 f (0) 3 3
= 0 2x + 1 2x + 2 2x + 3 2x +
2 0! 2 1! 2 2! 2 3!
f (0) f '(0) f "(0) f (3)(0)
= (2x)0+ (2x)1+ (2x)2+ (2x)3+ 
0!! 2!! 4!! 6!!
Replacing x with x/2 , we obtain (6).

2.1.4 Expansion by the double factorial of an elementary function

Applying Formula 2.1.3 to an elementary function, we obtain the following expressions.

(1) Expansion by the double factorial of the exponental function

x
 x0 x1 x2 x3 x4
e 2
=  +  + 
0!! 2!! 4!! 6!! 8!!

(2) Expansions by the double factorial of the trigonometric function

x x0 x2 x4 x6 x8
cos = - + - + -
2 0!! 4!! 8!! 12!! 16!!
x x1 x3 x5 x7 x9
sin = - + - + -
2 2!! 6!! 10!! 14!! 18!!

(3) Expansions by the double factorial of the hyperbolic function

x x0 x2 x4 x6 x8
cosh = + + + + +
2 0!! 4!! 8!! 12!! 16!!
x x1 x3 x5 x7 x9
sinh = + + + + +
2 2!! 6!! 10!! 14!! 18!!

(4) Expansions by the double factorial of the logarithmic function

 
x 0! 1! 2 2! 3 3! 4
log 1 = x- x  x - x 
2 2!! 4!! 6!! 8!!
2+ x 0! 2! 3 4! 5 6! 7
log = x+ x + x + x +
2- x 2!! 6!! 10!! 14!!
1 2+ xi 0! 2! 3 4! 5 6! 7
log = x- x + x - x +
2i 2- xi 2!! 6!! 10!! 14!!

(5) Expansions by the double factorial of the inverse trigonometric function

(-1)!! x1 1!! x 3 3!! x 5 5!! x 7
sin x = -1
+ + + +
0!! 1 2!! 3 4!! 5 6!! 7
x 0! 2! 3 4! 5 6! 7
tan -1 = x- x + x - x +
2 2!! 6!! 10!! 14!!

-3-
 x 1 2+ xi
 tan -1 = log
2 2i 2- xi

(6) Expansions by the double factorial of the inverse hyperbolic function

(-1)!! x 1 1!! x 3 3!! x 5 5!! x 7
sinh x = -1
- + - +
0!! 1 2!! 3 4!! 5 6!! 7
x 0! 2! 3 4! 5 6! 7
tanh -1 = x+ x + x + x +
2 2!! 6!! 10!! 14!!
-1 x 2+ x
 tanh = log |x|<2
2 2- x

(7) Expansions by the double factorial of the irrational function

1
(-1)!! 1!! 2 3!! 3 5!! 4
(1 x)
2
=1 x- x  x - x 
2!! 4!! 6!! 8!!
1
- 1!! 3!! 2 5!! 3 7!! 4
(1 x)
2
=1 x+ x  x + x 
2!! 4!! 6!! 8!!


x (-1)!! 1 3 1!! 1 5 3!! 1 7
1 x 2dx = x  x - x  x - 
0 2!! 3 4!! 5 6!! 7


x dx 1!! 1 3 3!! 1 5 5!! 1 7
=x  x + x  x +
0 1 x 2 2!! 3 4!! 5 6!! 7

2.1.5 The double factorial series of the elementary function

The following series are obtained as the special values of 2.1.4 .
1
1 1 1 1 1
e 2
= + + + + +
0!! 2!! 4!! 6!! 8!!
1
- 1 1 1 1 1
e 2
= - + - + -
0!! 2!! 4!! 6!! 8!!
1 1 1 1 1 1
cos = - + - + -
2 0!! 4!! 8!! 12!! 16!!
1 1 1 1 1 1
sin = - + - + -
2 2!! 6!! 10!! 14!! 18!!
1 1 1 1 1 1
cosh = + + + + +
2 0!! 4!! 8!! 12!! 16!!
1 1 1 1 1 1
sinh = + + + + +
2 2!! 6!! 10!! 14!! 18!!
3 0! 1! 2! 3!
log = - + - +
2 2!! 4!! 6!! 8!!
0! 1! 2! 3!
log 2 = + + + +
2!! 4!! 6!! 8!!
0! 2! 4! 6!
log 3 = + + + +
2!! 6!! 10!! 14!!

-4-
 1 0! 2! 4! 6!
tan -1 = - + - +
2 2!! 6!! 10!! 14!!
1 0! 2! 4! 6!
tanh -1 = + + + +
2 2!! 6!! 10!! 14!!
(-1)!! 1!! 3!! 5!!
2 =1+ - + - +
2!! 4!! 6!! 8!!
1 1!! 3!! 5!! 7!!
=1- + - + -
2 2!! 4!! 6!! 8!!
(-1)!! 1!! 3!! 5!!
1 = + + + +
2!! 4!! 6!! 8!!

-5-
2.2 Triple factorial

2.2.1 The definition of a Triple Factorial

Definition 2.2.1
When m, n denote a natural number,
m !!!  m (m -3)(m -6)  741 m =3n -2 (12)

 m (m -3)(m -6)  852 m =3n -1 (11)

 m (m -3)(m -6)  963 m =3n (10)

1 m =0, -1, -2 (1- )

Example.
1!!! = 1, 4!!! = 14, 7!!! = 147, 
2!!! = 2, 5!!! = 25, 8!!! = 258, 
3!!! = 3, 6!!! = 36, 9!!! = 369, 
0!!! = (-1)!!! = (-2)!!! =1

2.2.2 The basic formulas of a triple factorial

Formula 2.2.2
When n and (z) denote a natural number and a gamma function respectively,
n! = n!!!(n -1)!!!(n -2)!!! (2)

   
1 1
(3n -2)!!! = 3n n + / (32)
3 3

   
2 2
(3n -1)!!! = 3n n + / (31)
3 3

   
3 3
(3n)!!! = 3n n + / = 3n(n +1) = 3nn ! (30)
3 3
(-3n)!!! = , 0!!! = 1 (4)

(-1)!!!(-2)!!! = 1 (5)
3
-(3n +1)!!!-(3n +2)!!! = (6)
(3n -1)!!!(3n -2)!!!

Proof
First, calculate (12)  (11)  (10), as follows.
(3n -2)!!!(3n -1)!!!(3n)!!!
= (3n -2)(3n -5)(3n -8)  741
(3n -1)(3n -4)(3n -7)  852
 3n (3n -3)(3n -6)  963 = (3n)!
Replacing 2n with n, we obtain (2).

-6-
 Next, since (1+ z) = z z ,

   
1 1 1
 1+  =
3 3 3

       
1 4 4 4 4 1 1
 2+ =  1+ =  = 
3 3 3 3 3 3 3

       
1 7 7 7 7 4 1 1
 3+ =  1+ =  = 
3 3 3 3 3 3 3 3


     
1 147(3n -2) 1 (3n -2)!!! 1
 n+ = n
 = n

3 3 3 3 3

   
1 1
 (3n -2)!!! = 3n n + / (32)
3 3
In a similar way, we obtain (31),(30).
Substitute n =0 for (30) ; then

0!!! = 3 (1) = 1
0

Replace n with -n in (30) ; then

(-3n)!!! = 3-n(1- n) =  (approach from +) (4)

0
Next, from (2) and (3 )
(3n)! (1+3n)
(3n -1)!!!(3n -2)!!! = = n (w)
(3n)!!! 3 (1+n)
Substitute n =0 for (w) ; then
(1) 0!
(-1)!!!(-2)!!! = = =1 (5)
(1)30 0!1
This secures the justification of a definition (1- ) conjointly with 0!!!=1 .
Finally, replace n with -n in (w) , then

(1-3n) 3n-(3n -1)

-(3n +1)!!!-(3n +2)!!! = =
3-n(1-n) -(n -1)
According to Formulas1.1 in 1 .3 (Singular Point Formulas) ,
-(3n -1) (n -1)! (n -1)!
= (-1)(n -1)-(3n -1) =
-(n -1) (3n -1)! (3n -1)!
Then

(n -1)!3n 3n(n -1)!3n

-(3n +1)!!!-(3n +2)!!! = =
(3n -1)! (3n)!
3 n!3n
= { from (2) }
(3n)!!!(3n -1)!!!(3n -2)!!!
3 n!3n
= n { from (30) }
3 n!(3n -1)!!!(3n -2)!!!
3
= (6)
(3n -1)!!!(3n -2)!!!

-7-
Example 1

       
2 2 20 2
17!!! = (36-1)!!! = 36 6+ / = 36 /
3 3 3 3
= 729389.03492617/1.35411794 = 209439.9998

Example 2
3 3
(-4)!!!(-5)!!! = =
1!!!2!!! 2

2.2.3 Expression by the triple factorial of Maclaurin expansion

Formula 2.2.3
f (0) 0 f '(0) 1 f " (0) 2 f 3 (0) 3
()

 
x
f = x + x + x + x + (7)
3 0!!! 3!!! 6!!! 9!!!

Calculation
f '(0) 1 f " (0) 2 f (3)(0) 3
f(x) = f(0)+ x + x + x +
1! 2! 3!
' " 3 ()
f (0) 0 f (0) 1 1 f (0) 2 2 f (0) 3 3
= 0 3x + 1 3 x + 2 3 x + 3 3 x +
3 0! 3 1! 3 2! 3 3!
' " (3)
f (0) 0 f (0) 1 f (0) 2 f (0)
= (3x) + (3x) + (3x) + (3x)3+ 
0!!! 3!!! 6!!! 9!!!
Replacing x with x/3 , we obtain (7).

2.2.4 Expansion by the triple factorial of an elementary function

Applying Formula 2.2.3 to an elementary function, we obtain the following expressions.

(1) Expansion by the triple factorial of the exponental function

x
 x0 x1 x2 x3 x4
e 3
=  +  + 
0!!! 3!!! 6!!! 9!!! 12!!!

(2) Expansions by the triple factorial of the trigonometric function

x x0 x2 x4 x6 x8
cos = - + - + -
3 0!!! 6!!! 12!!! 18!!! 24!!!
x x1 x3 x5 x7 x9
sin = - + - + -
3 3!!! 9!!! 15!!! 21!!! 27!!!

(3) Expansions by the triple factorial of the hyperbolic function

x x0 x2 x4 x6 x8
cosh = + + + + +
3 0!!! 6!!! 12!!! 18!!! 24!!!
x x1 x3 x5 x7 x9
sinh = + + + + +
3 3!!! 9!!! 15!!! 21!!! 27!!!

-8-
(4) Expansions by the triple factorial of the logarithmic function

 
x 0! 1! 2 2! 3 3! 4
log 1 = x- x  x - x 
3 3!!! 6!!! 9!!! 12!!!
3+ x 0! 2! 3 4! 5 6! 7
log = x+ x + x + x +
3- x 3!!! 9!!! 15!!! 21!!!
1 3+ xi 0! 2! 3 4! 5 6! 7
log = x- x + x - x +
2i 3- xi 3!!! 9!!! 15!!! 21!!!

(5) Expansions by the triple factorial of the inverse trigonometric function

x 0! 2! 3 4! 5 6! 7
tan -1 = x- x + x - x +
3 3!!! 9!!! 15!!! 21!!!
x 1 3+ xi
 tan -1 = log
3 2i 3- xi

(6) Expansions by the triple factorial of the inverse hyperbolic function

x 0! 2! 3 4! 5 6! 7
tanh -1 = x+ x + x + x +
3 3!!! 9!!! 15!!! 21!!!
x 3+ x
 tanh -1 = log |x|<3
3 3- x

(7) Expansions by the triple factorial of the irrational function

1
(-1)!!! 2!!! 2 5!!! 3 8!!! 4
(1 x)
3
=1 x- x  x - x 
3!!! 6!!! 9!!! 12!!!
1
- 1!!! 4!!! 2 7!!! 3 10!!! 4
(1 x)
3
=1 x+ x  x + x 
3!!! 6!!! 9!!! 12!!!


x3 (-1)!!! 1 4 2!!! 1 7 5!!! 1 10
1 x 3dx = x  x - x  x - 
0 3!!! 4 6!!! 7 9!!! 10


x dx 1!!! 1 4 4!!! 1 7 7!!! 1 10
3
=x  x + x  x +
0 1 x 3 3!!! 4 6!!! 7 9!!! 10

2.2.5 The triple factorial series of the elementary function

The following series are obtained as the special values of 2.2.4 .

1
1 1 1 1 1
e 3
= + + + + +
0!!! 3!!! 6!!! 9!!! 12!!!
1
- 1 1 1 1 1
e 3
= - + - + -
0!!! 3!!! 6!!! 9!!! 12!!!
1 1 1 1 1 1
cos = - + - + -
3 0!!! 6!!! 12!!! 18!!! 24!!!
1 1 1 1 1 1
sin = - + - + -
3 3!!! 9!!! 15!!! 21!!! 27!!!

-9-
 1 1 1 1 1 1
cosh = + + + + +
3 0!!! 6!!! 12!!! 18!!! 24!!!
1 1 1 1 1 1
sinh = + + + + +
3 3!!! 9!!! 15!!! 21!!! 27!!!
4 0! 1! 2! 3!
log = - + - +
3 3!!! 6!!! 9!!! 12!!!
2 0! 1! 2! 3!
-log = + + + +
3 3!!! 6!!! 9!!! 12!!!
0! 2! 4! 6!
log 2 = + + + +
3!!! 9!!! 15!!! 21!!!
1 0! 2! 4! 6!
tan -1 = - + - +
3 3!!! 9!!! 15!!! 21!!!
1 0! 2! 4! 6!
tanh -1 = + + + +
3 3!!! 9!!! 15!!! 21!!!
(-1)!!! 2!!! 5!!! 8!!!
3
2 = 1+ - + - +
3!!! 6!!! 9!!! 12!!!
1 1!!! 4!!! 7!!! 10!!!
3
= 1- + - + -
2 3!!! 6!!! 9!!! 12!!!
(-1)!!! 2!!! 5!!! 8!!!
1 = + + + +
3!!! 6!!! 9!!! 12!!!

- 10 -
2.3 Multi factorial

2.3.1 The definition of a Multi factorial

Definition 2.3.1
When m, n denote natural number and !!! (k pieces) !k ,

m!k  n (n -k)(n -2k)  (1+ k)1 m =kn -(k -1) (1k-1)


 n (n -k)(n -2k)  (k -1)+k(k -1) m =kn -1 (11)

 n (n -k)(n -2k)  (k + k)k m =kn (10)

1 m =0, -1, , -(k -1) (1- )

Example
1!4 = 1, 5!4 = 15, 9!4 = 15 9, 13!4 = 15 913, 
2!4 = 2, 6!4 = 26, 10!4 = 2610, 14!4 = 261014, 
3!4 = 3, 7!4 = 37, 11!4 = 3711, 15!4 = 371115, 
4!4 = 4, 8!4 = 48, 12!4 = 4812, 16!4 = 4812, 16, 
0!4 = (-1)!4 = (-2)!4 = (-3)!4 = 1

2.3.2 The basic formulas of a multifactorial

Formula 2.3.2
When k , n and (z) denote a natural number and a gamma function respectively,
n! = n!k (n-1)!k (n-2)!k   n-(k -1)!k (2)

   
k -s k -s
(kn -s)!k = k n n + / , s =1, 2, k -1 (3s )
k k

   
k k
(kn)!k = k n n + / = k n(n +1) = k nn! (30)
k k
(-kn)!k = , 0!k = 1 (4)

(-1)!k (-2)!k   -(k -1)!k = 1 (5)

-(kn +1)!k -(kn +2)!k  -(kn + k -1)!k

k
= (6)
(kn -1)!k (kn -2)!k  kn -(k -1)!k

Proof
First, calculate (1k-1)  (11)  (10), as follows.

kn-(k -1)!k   (kn-1)!k (kn)!k

= kn -(k -1)kn -(2k -1)kn -(3k -1)  (k +1) 1
 kn -(k -2)kn -(2k -2)kn -(3k -2)  (k +2) 2

- 11 -
 
 (kn -1) kn -(k +1) kn -(2k +1)  k +(k -1) (k -1)
 kn (kn -k) (kn -2k)  (k + k)  k
= (kn )!
Replacing 2n with n, we obtain (2).
Next, since (1+ z) = z z ,

   
k-s k-s k-s
 1+ = 
k k k

   =    
k-s 2k - s 2k - s 2k - s 2k - s k - s k-s
 2+ =  1+  = 
k k k k k k k

   =  
k-s 3k - s 3k - s 2k - s k - s k-s
 3+ =  1+ 
k k k k k k

(nk -s)(2k - s)(k - s) (kn -s)!k
     
k-s k-s k-s
 n+ =  = 
k kn k kn k

   
k-s k-s
 (kn -s)!k = k n n + / (3s )
k k
Especially substituting s =0 for (3s ) , we obtain (30)
Substitute n =0 for (30) ; then

0!k = k 0(1) = 1
Replace n with -n in (30) ; then

(-kn)!k = k -n(1- n) =  (approach from +) (4)

0
Next, from (2) and (3 )
(kn)! (1+kn)
(kn -1)!k (kn -2)!k   kn -(k -1)!k = = n (w)
(kn)!k k (1+ n)
Substitute n =0 for (w) ; then
(1) 0!
(-1)!k (-2)!k   -(k -1)!k = = =1
(1)k 0 0!1
This secures the justification of a definition (1- ) conjointly with 0!k =1 .
Finally, replace n with -n n (w) , then

-(kn +1)!k -(kn +2)!k  -(kn + k -1)!k

(1- kn) k n -(kn -1)
= -n =
k (1- n)  -(n -1)
According to Frmulas1.1 in 1 .3 (Singular Point Formulas) ,
 -(kn -1) (n -1)! (n -1)!
= (-1)(n -1)-(k n -1) = (-1)-(k -1)n
 -(n -1) (kn -1)! (kn -1)!
Tthen

-(kn +1)!k -(kn +2)!k  -(kn + k -1)!k

- 12 -
 -(k -1)n (n -1)! k n -(k -1)n kn(n -1)! k
n
= (-1) = (-1)
(kn -1)! (kn)!
kn! k n
= (-1)-(k -1)n { from (2) }
(kn)!k (kn-1)!k (kn-2)!k   kn-(k -1)!k
-(k -1)n kn! k n
= (-1) { from (30) }
k nn!(kn-1)!k (kn-2)!k   kn-(k -1)!k
k
= (-1)-(k -1)n (6)
(kn-1)!k (kn-2)!k   kn-(k -1)!k

Example 1

       
5-3 5-3 32 2
27!5 = (56-3)!5 = 56 6+ / = 56 /
5 5 5 5
= 15625240.83377994/2.21815954 = 1696464.0025

Example 2
(-1)-(5-1)5 5
(-6)!5(-7)!5(-8)!5(-9)!5 = =
1!52!53!54!5 24

2.3.3 Expression by the k-fold factorial of Maclaurin expansion

Formula 2.3.3
f (0) 0 f (1)(0) 1 f (2)(0) 2 f (3)(0) 3
 
x
f = x + x + x + x + (7)
k 0!k k !k (2k)!k (3k)!k

Calculation
f (1)(0) 1 f (2)(0) 2 f (3)(0) 3
f(x) = f(0)+ x + x + x +
1! 2! 3!
f (0) 0 f (1)(0) 1 1 f (2)(0) 2 2 f (3)(0) 3 3
= 0 kx + 1 k x + 2 k x + 3 k x +
k 0! k 1! k 2! k 3!
(1) (2) (3)
f (0) 0 f (0) 1 f (0) 2 f (0)
= (kx) + (kx) + (kx) + (kx)3+ 
0!k k !k (2k)!k (3k)!k
Replacing x with x /k , we obtain (7).

2.3.4 Expansion by the multifactorial of an elementary function

Applying Formula 2.3.3 to an elementary function, we obtain the following expressions.

(1) Expansion by the multifactorial of the exponental function

x
 x0 x1 x2 x4 x3
e n
=  +  + 
0!n n !n (2n)!n (3n)!n (4n)!n

- 13 -
(2) Expansions by the multifactorial of the trigonometric function
x x0 x2 x4 x6 x8
cos = - + - + -
n 0!n (2n)!n (4n)!n (6n)!n (8n)!n
x x1 x3 x5 x7 x9
sin = - + - + -
n n !n (3n)!n (5n)!n (7n)!n (9n)!n

(3) Expansions by the multifactorial of the hyperbolic function

x x0 x2 x4 x6 x8
cosh = + + + + +
n 0!n (2n)!n (4n)!n (6n)!n (8n)!n
x x1 x3 x5 x7 x9
sinh = + + + + +
n n !n (3n)!n (5n)!n (7n)!n (9n)!n

(4) Expansions by the multifactorial of the logarithmic function

 
x 0! 1! 2! 3!
log 1 = x- x2  x3 - x4  
n n !n (2n)!n (3n)!n (4n)!n
n +x 0! 2! 4! 6!
log = x+ x3 + x5 + x7 + 
n -x n !n (3n)!n (5n)!n (7n)!n
1 n + xi 0! 2! 4! 6!
log = x- x3 + x5 - x7 + 
2i n - xi n !n (3n)!n (5n)!n (7n)!n

(5) Expansions by the multifactorial of the inverse trigonometric function

x 0! 2! 4! 6!
tan -1 = x- x3 + x5 - x7 + 
n n !n (3n)!n (5n)!n (7n)!n
x 1 n + xi
 tan -1 = log
n 2i n - xi

(6) Expansions by the multifactorial of the inverse hyperbolic function

x 0! 2! 4! 6!
tanh -1 = x+ x3 + x5 + x7 + 
n n !n (3n)!n (5n)!n (7n)!n
x n +x
 tanh -1 = log |x|< n
n n -x

(7) Expansions by the multifactorial of the irrational function

1
(-1)!n (n -1)!n 2 (2n -1)!n 3 (3n -1)!n 4
x  x 
n
(1 x) = 1 x- x -
n !n (2n)!n (3n)!n (4n)!n
1
- 1!n (n +1)!n 2 (2n +1)!n 3 (3n +1)!n 4
(1 x) = 1 n
x + x  (3 )! x + x 
n !n (2n)!n n n (4n)!n
(-1)!n x n+1 (n -1)!n x 2n+1 (2n -1)!n x 3n+1

xn
1 x dx = x 
n
-  -
0 n !n n +1 (2n)!n 2n +1 (3n)!n 3n +1

- 14 -
 1!n x n+1 (n +1)!n x 2n+1 (2n +1)!n x 3n+1

x dx
= x +  (3n)! 3 +1 + 
0
n
1 x n n !n n +1 (2n )!n 2n +1 n n

2.3.5 The multifactorial series of the elementary function

The following series are obtained as the special values of 2.3.4 .
1
 1 1 1 1 1
e n
=  +  + 
0!n n !n (2n)!n (3n)!n (4n)!n
1 1 1 1 1 1
cos = - + - + -
n 0!n (2n)!n (4n)!n (6n)!n (8n)!n
1 1 1 1 1 1
sin = - + - + -
n n !n (3n)!n (5n)!n (7n)!n (9n)!n
1 1 1 1 1 1
cosh = + + + + +
n 0!n (2n)!n (4n)!n (6n)!n (8n)!n
1 1 1 1 1 1
sinh = + + + + +
n n !n (3n)!n (5n)!n (7n)!n (9n)!n

 
1 0! 1! 2! 3!
log 1 = -  - 
n n !n (2n)!n (3n)!n (4n)!n
n +1 0! 2! 4! 6!
log = + + + + (n >1)
n -1 n !n (3n)!n (5n)!n (7n)!n
1 0! 2! 4! 6!
tan -1 = - + - +
n n !n (3n)!n (5n)!n (7n)!n
1 0! 2! 4! 6!
tanh -1 = + + + +
n n !n (3n)!n (5n)!n (7n)!n
(-1)!n (n -1)!n (2n -1)!n (3n -1)!n
n
2 =1+ - + - +  (n >1)
n !n (2n)!n (3n)!n (4n)!n
1 1!n (n +1)!n (2n +1)!n (3n +1)!n
n
= 1 - + - + -
2 n !n (2n)!n (3n)!n (4n)!n
(-1)!n (n -1)!n (2n -1)!n (3n -1)!n
1 = + + + + (n >1)
n !n (2n)!n (3n)!n (4n)!n

Especially when n=1

1 1 1 1
e =1+ + + + +
1! 2! 3! 4!
1 1 1 1 1
=1- + - + -
e 1! 2! 3! 4!
1 1 1 1 1
cos 1 = - + - + -
0! 2! 4! 6! 8!

- 15 -
 1 1 1 1 1
sin 1 =
- + - + -
1! 3! 5! 7! 9!
1 1 1 1 1
cosh 1 = + + + + +
0! 2! 4! 6! 8!
1 1 1 1 1
sinh 1 = + + + + +
1! 3! 5! 7! 9!
1 1 1
log 2 =1- + - +
2 3 4
1 1 1
tan -11 = 1 - + - +
3 5 7
1 1 1
tanh -11 = 1 + + + + =
3 5 7
1
=1-1+1-1+1-
2

2.3.6 Expansions by the multifactorial of the Pochhammer Symbol

Pochhammer Symbol of a rational number smaller than 1 can be expressed by the multi-factorial.

Formula 2.3.4
When n , k are two or more natural numbers and a is an integer such that |a|< n ,
a +(k -1)n!n
 
a
= (when a >0) (8)
n k nk
aa +(k -1)n!n
= (when a <0) (8')
nk

Calculation
When a >0 ,
a(a +1n)a +(k -1)n
     
a a a a
= +1  + k -1 =
n k n n n nk
a +(k -1)n!n
=
nk
When a < 0 , since a !n = 1, a +1n >0,  , a +(k -1) n > 0 , then

a = a a !n (k =1)
a(a +1n) = a (a +1n)!n (k =2)
a(a +1n)(a +2n) = a (a +2n)!n (k =3)
hence by inducton

a(a +1n )a +(k -1)n a a +(k -1)n!n

=
nk nk

- 16 -
Example

 
1 1471+3(k -1) 1+3(k -1)!!!
= k
=
3 k 3 3k

 
-1 -125-1+3(k -1) -1-1+3(k -1)!!!
= k
= (-1)!!! = 1
3 k 3 3k

 
-2 -214-2+3(k -1) -2-2+3(k -1)!!!
= k
= (-2)!!! = 1
3 k 3 3k

2.3.7 Expansions by the multifactorial of the Hypergeometric Function

The hypergeometric function of which parameters are rational number smaller than 1 can be exprssed by
the multifactorial.

Formula 2.3.5
When n is two or more natural number and a is an integer such that |a|< n ,
 a +n(k -1)!n (b )k xk
 
a
2F 1 , b, c ; x = 1 +Σ (a >0) (9)
n k =1 (c)k (nk)!n
 a +n(k -1)!n (b )k xk
= 1 + aΣ (a <0) (9')
k =1 (c)k (nk)!n
When n is two or more natural number and a,b,c are integers such that |a |,|b |,|c|< n ,

 
a b c
2F 1 , , ;x
n n n
 a +n(k -1)!n b +n(k -1)!n xk
= 1 +Σ (a,b,c >0) (10)
k =1 c +n(k -1)!n (nk)!n
 a +n(k -1)!n b +n(k -1)!n xk
= 1 + aΣ (only a <0) (10')
k =1 c +n(k -1)!n (nk)!n
1  a +n(k -1)!nb +n(k -1)!n xk
=1+ Σ (only c <0) (10")
c k =1 c +n(k -1)!n (nk)!n

Calculation
From (8),(8') and n k k! = (nk)!n

 
a
(b)k k
n k
 
a  x
2F 1 , b, c ; x = Σ
n k =0 (c)k k!
 a +(k -1)n!n (b )k xk
= 1 +Σ (a >0) (9)
k =1 (c)k (nk)!n
 a +(k -1)n!n (b )k xk
= 1 + aΣ (a <0) (9')
k =1 (c)k (nk)!n

- 17 -
   
a b
n n xk
  Σ
a b c  k k
2F 1 , , ;x =
n
n n n c k =0 k!
k
a +(k -1)n!n b +(k -1)n!n

 nk nk xk
= 1 +Σ (a,b,c > 0)
k =1 c +(k -1)n!n k!
nk
 a +n(k -1)!n b +n(k -1)!n xk
= 1 +Σ (a,b,c > 0) (10)
k =1 c +n(k -1)!n (nk)!n
Next, because a symbol whose sign is negative among a,b,c serves as a coefficient of Σ,
(10') and (10") are obvious.

Example 1

 
1
2F 1 3
, c, c ; x

 1+3(k -1)!3(c)k xk  (3k -2)!!! k

= 1 +Σ = 1 +Σ x
k =1 (c)k (3k)!3 k =1 (3k)!!!
1
1!!! 4!!! 2 7!!! 3 10!!! 4 -
=1+ x+ x + x + x +  = (1- x) 3
3!!! 6!!! 9!!! 12!!!

Example 2

 
-1
2F 1 3
, c, c ; x

 -1+3(k -1)!3(c)k xk  (3k -4)!!! k

= 1 +Σ = 1 + (-1)Σ x
k =1 (c)k (3k)!3 k =1 (3k)!!!
1
(-1)!!! 2!!! 2 5!!! 3 8!!! 4
=1- x- x - x - x -  = (1- x)3
3!!! 6!!! 9!!! 12!!!

Example 3

 
1 1 3
2F 1 , , ; x2
2 2 2
 1+2(k -1)!21+2(k -1)!2 x 2k
= 1 +Σ
k =1 3+2(k -1)!2 (2k)!2
2k
 (2k -1)!!(2k -1)!! x
= 1 +Σ
k =1 (2k +1)!! (2k)!!
2 4
1!! x 3!! x 5!! x 6 sin -1 x
=1+ + + + =
2!! 3 4!! 5 6!! 7 x

- 18 -
Example 4

 
1 n -1 n +1
2F 1 , , ; xn
n n n
 1+n(k -1)!n n -1+n(k -1)!n x nk
= 1 +Σ
k =1 n +1+n(k -1)!n (nk)!n
 n(k -1)+1!n (nk -1)!n x nk
= 1 +Σ
k =1 (nk +1)!n (nk)!n
(n -1)!n x n (2n -1)!n x 2n (3n -1)!n x 3n
=1+ + + +
n !n n +1 (2n)!n 2n +1 (3n)!n 3n +1

2004.05.08
K. Kono
Alien's Mathematics

- 19 -
3 Generalized Multinomial Theorem

3.1 Binomial Theorem

Theorem 3.1.1
If x1 , x2 are real numbers and n is a positive integer, then
n
x1+ x2 = ΣnＣ r x1 x2
n n-r r
(1.1)
r=0

Binomial Coefficients
Binomial Coefficient in (1.1) is a positive number and is described as nＣ r . Here, n and r are both

non-negative integer. nＣ r is the number of ways of picking r unordered outcomes from n possibilities
and is calculated as follows.
n!
n Ｃr =
(n -r)! r!

Pascal's Triangle
About nＣ r , what arranged n in the row and r in the column is called Pascal's Triangle .

Ｃ0
0 1
1Ｃ0 Ｃ1
1 1 1
Ｃ0
2 Ｃ1
2 Ｃ2
2 1 2 1
Ｃ0
3 Ｃ1
3 3 Ｃ2 3Ｃ3 1 3 3 1
4 Ｃ0 Ｃ1
4 Ｃ2
4 Ｃ3
4 Ｃ4
4 1 4 6 4 1
 

Propertis of the Binomial Coefficient

Although a lot of properties of the binomial coefficient are known, fundamental ( understood from Pascal's
Triangle immediately ) some are as follows. Among these, ii is used for step-by-step calculation of nＣ r .

i nＣ r = nＣ n-r
ii nＣ r = n -1Ｃ r-1 + n -1Ｃ r
n -1
iii Ｃ r = Σ kＣ r-1
n
k = r-1
n
iv Σ
r=0
nＣ r = 2
n

n
v Σ
r=0
(-1)rnＣ r = 0

-1-
3.2 Generalized Binomial Theorem

3.2.1 Newton's Generalized Binomial Theorem

Theorem 3.2.1
The following formulas hold for a real number  .


 r

Σ x r
|x| 1 ( |x| = 1 is allowed at  >0 ) ( 1.1)
r=0

(1+ x) =

Σ   -r  x

-r
|x|>1 ( 1.2)
r=0

Proof
When n is a natural number, the following expression holds from the binomial theorem.
n
(1+ x)n = ΣnＣ r x r
r=0

Since nＣ r = 0 for r > n , this can be written as follows.


(1+ x)n = ΣnＣ r x r
r=0
Extending n to real number  ,
 ( +1)
r
 
(1+ x) = Σ xr = Σ xr (1.1)
r=0 r=0 ( -r +1) r!
Here, let
( +1)
x r  ar .
( -r +1) r!
Then
(p +1) r+1
x
ar+1 (p -r) (r +1)! (p -r)x
= = .
ar (p +1) r +1
xr
(p -r +1) r!
From this,

 
ar+1
      = |x|
(p -r)x (p /r -1)x -x
lim = lim = lim =
r ar r r +1 r 1+1/r 1
According to d'Alembert's ratio test, (1.1) converges absolutely if |x|< 1 .
then x < 1 . Therefore, from (1.1) ,
-1
Next, if |x|> 1

r
 
-1 r
1+ x  = Σ x 
-1
r=0

Multiplying by x both sides,

 
r  
 

(x +1) = x Σ x -r = Σ x -r (1.2)
r=0 r=0  -r
The proof in case of |x|= 1 is accomplished in the following sub section..

-2-
3.2.2 General Binomial Coefficient

The coefficient
r  in Theorem 3.2.1 is called General Binomial Coefficient and is as follows.

 ( +1)  ( -1)( -2)( -r +1)

r =
( -r +1)(r +1)
=
r!
(2.0)

The first few are as follows.

    ( -1)  ( -1) ( -2)
0 = 1,
1 =
1!
,
 2 =
2!
,
3 =
3!
,

Although properties similar to binomial coefficient also about general binomial coefficient are known, especially
an important thing is sum of the general binomial coefficient. We prepare some Lemma, in order to obtain this.

Lemma 3.2.2
 -1
r

When  is not positive integer, binomial series Σ ar converges or diverges simultaneously
r=0

 ar
with Dirichlet series Σ (-1)r .
r=1 r

Source 「岩波数学公式Ⅱ」 p132

Lemma 3.2.3

r

Σ converges absolutely for non-integer  >0 .
r=0

Proof
   -1
r =
 -r r
Then
   -1
r r
 
Σ
r=0
=Σ
r=0  -r

Let ar = . Then
 -r
 -1 
r r
 
Σ
r=0
ar =Σ
r=0
(s1)

 ar  
Σ (-1)r 
= Σ(-1)r 
(s2)
r=1 r r=1 r ( -r)
Here, let

(-1)r  br
r ( -r)
Then

-3-
 
(-1)r+1 
br+1 (r +1) ( -r -1) r ( -r)
= =-
br r  (r +1)( -r -1)
(-1) 
r ( -r)
From this

r ( -r)  r -r +1

  = lim   = lim  (r +1) -(r +1) 
br+1
lim
r br r
(r +1)( -r -1) r  +1

 r

  



+1 +1
-1 +1 -1


r r 1 r
= lim = lim =1
r
(r +1)+1
 (r +1) r 1 
-1 1+ -1
(r +1)+1 r r +1
Since the judgment is impossible, we try Raabe's test for convergence.

(r +1)( -r -1)
     
br
lim r -1 = lim r -1
r br+1 r
r ( -r)
 

 r r ( -r)  - r
r(r +1) r(r +1)
= lim -  
r

 

       - r
1 1 r
= lim r 1+ - 1+
r r r  -r

= lim  1+  r +
 1- /r  - r
1 1
r r
Since 1- /r > 0 for sufficiently large r ,
 

     
br
 
1 1 1
lim r -1 = lim 1+ r + 1+ -r
r br+1 r r r 1- /r
 
= lim r 1+    
1 1 1
-1 + lim 1+
r r r r 1- /r
Here,
   1 ( -1) 1 ( -1)( -2) 1
s  r


 
1 1
1+ =Σ s
=1+ 1
+ 2
+ 3
+
r s=0 1! r 2! r 3! r
Then,
   
  s  r s  r s  r
  


1 1 1 1
r 1+ -1 = rΣ s
=Σ s-1
= +Σ s-1
r s=1 s=1 s=2
Therfore,
 
   s r
1  1
lim r 1+ - 1 =  + lim Σ s-1
=
r r r s=2

Moreover,


 
1 1
lim 1+ =1
r r 1- /r
After all,

-4-
   
br
lim r -1 =  +1
r br+1
Thus, if  >0 , (s2) converges absolutely. Then, (s1) also converges absolutely according to Lemma 3.2.2 .

Theorem 3.2.4
The following expressions hold for arbitrary real number  >0 .

r = 2


Σ
r=0
(2.1)


Σ(-1)  r  = 0

r
(2.2)
r=0

Proof

r

According to Lemma 3.2.3 , Σ converges absolutely for non-integer  >0 .
r=0
Therefore, from Theorem 3.2.1 (1.1) ,

r

Σ 1r = (1+1) = 2 (2.1)
r=0


Σ  r  (-1) = (1-1)

r 
=0 (2.2)
r=0

Note
In fact, it is known that (2.1) holds if  > -1 . ( Where, it is conditional convergence. )
For example, when  = -0.9 , the right side is 2-0.9 = 0.53588673 and the left side seems to
converge to this value. However, the confirmation is difficult as the convergence is very slow. So, we apply
Knopp Transformation to this and accelerate the convergence. It is as follows.

We can see that (2.1) converges to 2-0.9 .

3.2.3 Generalized Binomial Theorem

Theorem 3.2.1 can be further generalized.

Theorem 3.2.5
When  is a real number, the following expression holds for x1 , x2 s.t. x 1 x 2 .


r

 
x1+ x2 = Σ x1 -rx2r ( x 1 = x 2 is allowed at  >0 ) (3.1)
r=0

Proof
x2
 
x2
If x 1 x 2 then 1> = . Therefore, using Theorem 3.2.1 (1.1) ,
x1 x1

-5-
  

x   = x  

x2 
x2
x1+ x2 = 1 1+
x1 1 1+
x1
 x  
  1  x  2   x   3   x  + 
2 3

x 2 x 2 2
=x 1 +
1 + +
1 1 1
   
=
 0  1
x + 
1x x +
 2 x x +
1  3 x x + 
-1 1
2
-2 2
1 2
-3 3
1 2


r x x

-r r
=Σ 1 2
r=0

Note
As is clear from the process of the proof, if x 1 x 2 then (3.1) converges absolutely.

Where, x 1 x2 is allowed at  >0 .

This becomes important in Generalized Multinomial Theorem.

-6-
3.3 Multinomial Theorem

Theorem 3.3.0
For real numbers x1 , x2 ,  , xm and non negative integers n , r1 , r2 ,  , rm , the followings hold.

n!
x11 x22  xmm
n r r r
x1+ x2++ xm = Σ (0.1)
r1! r2! rm!
where  denotes the sum of all combinations of r1 , r2 ,  , rm s.t. r1+ r2++ rm= n .
n! n-r  -r r
x1 1 m-1 x21  xmm-1
n r
x1+ x2++ xm = Σ (0.2)
n -r1 -rm-1! r1! rm-1!
where  denotes the sum of all possible combinations of n , r1 , r2 ,  , rm-1 .

Since (0.1) is well known, the proof is omitted. In addition, it is also clear (0.2) and (0.1) are synonymous.
These are near a definition rather than a theorem.

How to generate multinomial coefficients

Theorem 3.3.0 is not difficult in theory. Difficulty is its proviso. This is to actually generate combinations
( m choose n ) with repetition. But this is not easy when it becomes more than 3 terms. Since I found out
the formulas which generates these without leak, I present it here as a theorem. (1.2) realizes the provis by
an iterated series (multiple series) and (1.1) realizes it by a diagonal series (half-multiple series).

Theorem 3.3.1
For real numbers x1 , x2 ,  , xm and a natural number n , the following expressions hold.

 r  r  r 
n r1 rm -2 n r1 rm-2 rm-2-rm-1 rm-1
x1+ x2++ xm = Σ Σ  Σ  x2  xm-1
n n -r1 r1-r2
x1 xm
r1=0 r2=0 rm -1=0 1 2 m-1
(1.1)
r1+ r2+ + rm-1
 r + r + + r     
n n n n rm-2 + rm-1
= Σ Σ Σ 
r1=0 r2=0 rm -1=0 1 2 m-1 r2+ + rm-1 rm-1
n -r1- - rm-1 r
 x1 x21 x32  xmm-1
r r

(1.2)

Proof
According to Theorem 3.1.1 , the following expressions hold.
n n n-r1 r1
x1+ x2+ x3+ x4++ xm = Σ nＣ r1 x1 x2+ x3+ x4++ xm (1)
r1=0

r r1 r
r
x2+ x3+ x4++ xm
1
= Σ r1Ｃ r2 x21-r2 x3+x4++xm 2 (2)
r2=0

r2 r2 r
r
x3+ x4++ xm = Σ r2Ｃ r3 x32-r3 x4++xm 3 (3)
r3=0


rm -3
r r -rm-2 rm-2
xm-2 + xm-1 + xm
m-3
= Σ r Ｃr
r =0
m-3 m-2
m-3
xm-2 xm-1 + xm (m -2)
m -2

-7-
 rm -2
r r -rm-1 r
xm-1 + xm
m-2
= Σ r Ｃr
r =0
m-2 m-1
m-2
xm-1 xmm-1 (m -1)
m -1

Substituting (2), (3),  ,(m -2) ,(m -1) for (1) one by one, we obtain (1.1) .
Next, according to Theorem 3.4.1 ( later ) , when x1  x2+ x3++ xm ,
 r1+ r2+ + rm-1
 r + r + + r     
    rm-2 + rm-1
x1+ x2++ xm = Σ Σ Σ r2+ + rm-1

rm-1
r1=0 r2=0 rm -1=0 1 2 m-1

 -r1-- rm -1 r1 r2 r
 x1 x2 x3  xmm-1
Replacing the real number  with non-negative integer n ,
 r1+ r2+ + rm-1
 r + r + + r     
   rm-2 + rm-1
x1+ x2++ xm = Σ Σ Σ
n

r1=0 r2=0 rm -1=0 1 2 m-1 r2+ + rm-1 rm-1
n -r1-- rm -1 r
 x1 x21 x32  xmm-1
r r

 r  = 0 for r > n, r =1, 2, 3,  , this is a definite multiple series. Therefore, the condition
n
Since

x1 x 2+ x3++ xm is unnecessary. Although this is not bad as it is, replacing  on the  with
n , we obtain (1.2) .

cf.
(1.2) results in (0.2) . Because,
r1+ r2+ + rm-1
 r + r + + r     
n rm-2 + rm-1 n!
 =
1 2 m-1 r2+ + rm-1 rm-1 n -r1 - - rm-1! r1!+ + rm-1!

4
Example 1: The expasion of x1+ x2+ x3
Using (1.1) ,

 r  s 
4 r 4 r r-s
x1+ x2+ x3 = ΣΣ x14-rx2 x3s
4
r=0 s=0

0 Σ  s  x
4 0 4
0
0-s s
= x 1 2 x3
s=0

1 Σ  s  x 2  s  x
4 1 1 4 2 2
+ x 3
1
1-s s
2 x3 + x12Σ 2-s s
2 x3
s=0 s=0

3  s  x 4  s  x
4 3 3 4 4 4
Σ Σ
1 3-s s 0 4-s s
+ x 1 2 x3 + x 1 2 x3
s=0 s=0

= x14
+ 4x13 x2 + x3
+ 6x12 x22 + 2x2 x3 + x32 
+ 4x1x23 + 3x22 x3 + 3x2 x32 + x33
+ x24 + 4x23 x3 + 6x22 x32 + 4x2 x33 + x34

-8-
 Using (1.2) ,

r + s   s 
4 4 4 r+s
x1+ x2+ x3 = ΣΣ x14-r-s x2r x3s
4
r=0 s=0

 0+ s   s  x
44 0+ s
= Σ 4-s 0
1 x2 x3s
s=0

1+ s   s  x 2+ s   s  x
44 1+ s 4 4 2+ s
+Σ 3-s 1 s
1 x2 x3 +Σ 2-s 2 s
1 x2 x3
s=0 s=0

3+ s   s  x 4+ s   s  x
44 3+ s 4 4 4+ s
+Σ 1-s 3 s
1 x2 x3 +Σ 0-s 4 s
1 x2 x3
s=0 s=0

= x14 + 4x13 x3 + 6x12 x32 + 4x1x33 + x34

+ 4x13 x2 + 12x12 x2 x3 + 12x1 x2 x32 + 4x2x33 + 0
+ 6x12 x22 + 12x1 x22 x3 + 6x22 x32 + 0 + 0
+ 4x1 x23 + 4x23 x3 + 0 + 0 + 0
+ x24 + 0 + 0 + 0 + 0
we can see that what totaled this along the diagonal line is equal to the above.

3
Example 2: The expasion of x1+ x2+ x3+ x4
Now, the formulas of the theorem are expanded using mathematical software. (1.1) and (1.2) are expanded
and are verified respectively.

-9-
Sum of multinomial coefficients
n r1 r m -2
Σ Σ  Σ nＣ r1 r1Ｃ r2  rm-2Ｃ rm-1 = m n
r1=0 r 2=0 rm -1=0
(1.1")

Proof
n n
n-r r
Σ
r=0
nＣ r =ΣnＣ r1
r=0
1 = (1+1)n = 2n
n r n r n

 s=0  = Σ
n-r r
Σ ΣnＣ r rＣ s =Σ
r=0 s=0 r=0
nＣ r Σ rＣ s
r=0
nＣ r1 2 = (1+2)n = 3n
n r s n r s n
Σ Σ Σ
r=0 s=0 t=0 
nＣ r rＣ s sＣ t =ΣnＣ r ΣΣrＣ s sＣ t
r=0 s=0 t=0 r=0 
=ΣnＣ r1n-r3r = (1+3)n = 4n
Hereafter, by induction we obtain the desired expression.

4
Example: Sum of multinomial coefficients of x1+x2+x3
Let's calculate sum of multinomial coefficients in Example 1 . Then it is as follows.

1+(4+4)+(6+12+6)+(4+12+12+4)+(1+4+6+4+1) = 81 = 34

- 10 -
3.4 Generalized Multinomial Theorem
Although I do not know whether the theorem like generalized multinomial theorem exists or not , since this
is essential for Higher Calculus of Function Product, I present this here.

Theorem 3.4.1
The following expressions hold for real numbers  and x1 , x2 ,  , xm s.t.x 1 x2+ x3++ x m .

 r  r  r 
  r1 r m -2 r1 rm-2  -r1 r1-r2 r -rm -1 rm -1
x1+ x2++xm = Σ Σ  Σ   xm-1
m -2
. x1 x2 xm
r1=0 r 2=0 rm -1=0 1 2 m-1
(1.1)
 r1+ r2+ + rm-1
 r + r + + r     
   rm-2 + rm-1
= Σ Σ Σ 
r 1=0 r 2=0 rm -1=0 1 2 m-1 r2+ + rm-1 rm-1
 -r1- - rm-1 r1 r2 r
 x1 x2 x3  xmm-1
(1.2)
Where, x 1 x 2+ x 3++ x m is allowed at  >0 .

Proof
From Theorem 3.2.5 , when x 1 x 2+ x 3++ x m , the following expression holds.



  -r1 r1
x1+ x2+ x3+ x4++ xm = Σ x1 x2+ x3+ x4++xm
r1=0 r1

Here, the right side converges absolutely.

On the other hand, from Theorem 3.3.1 (1.1) , the following expression holds.

    
r1
r1 r2 r m -2 r1 r2 rm-2 r -rm -1 rm -1
x1+ x2++ xm = Σ Σ  Σ   xm-1
r -r2 r -r3 m -2
x21 x32 xm
r2=0 r 3=0 r m -1=0 r2 r3 rm-1
Substituting the latter for the former ,

 r  r   r  x
  r1 rm -2 r1 rm-2 -r1 r1-r2 r -rm -1 rm -1
x1+ x2++ xm = Σ Σ  Σ  xm-1
m -2
1 x2 xm
r 1=0 r 2=0 r m -1=0 1 2 m-1
(1.1)
Naturally, the right side also converges absolutely.
Next, let us describe a multiple series and its iterated series as follows respectively.
   
Σ ar ,r ,,r
r , r ,  ,r =0
1 2 m
, Σ Σ Σ ar ,r ,, r
r =0 r =0 r =0 1 2 m -1 ,rm
1 2 m 1 2 m

In order to convert the iterated series to its diagonal series, we should just perform the following operations.
( See " 02 Multiple Series & Exponential Function " ).
Replace rm-1 with rm-1- rm , and replace the 1st  with rm-1 from the right.

Replace rm-2 with rm-2- rm-1 , and replace the 2nd  with rm-2 from the right.


Replace r1 with r1- r2 , and replace the (m-1)th  with r1 from the right.
If so, in order to return the diagonal series to the original iterated series, we should just perform this opposite
operation. That is,
Replace r1 with r1+ r2 , and replace r1 on the 2nd  with  from the left.

Replace r2 with r2+ r3 , and replace r2 on the 3rd  with  from the left.

- 11 -
 Replace rm-1 with rm-1+ rm , and replace rm-1 on the (m -1)t h  with  from the left.
For example,

  r  r  x x x
 r1 r2 r1 r2  -r1 r1-r2 r2-r3 r3

x1+ x2+ x3+ x4 = Σ Σ Σ r x1 2 3 4
r1=0 r 2=0 r 3=0 1 2 3


 r +r   r   r  x x x
r1  r +r r   r2 1 2 2  -r1-r2 r1 r2-r3 r3
r  r + r ,ΣΣ; = Σ Σ Σ
1 1 2 1 2 3 x4
r1=0 r2=0 r 3=0 1 2 2 3


Σ Σ Σ  r +r +r   r +r   r  x
r2  r +r +r r +r
   1 2 3 2 3  -r1-r2-r3 r1 r2 r3
2 2 
r r + r ,Σ Σ;
3 = 1 x2 x3 x4
r1=0 r 2=0 r3=0 1 2 3 2 3 3
Thus, performing this operation to (1.1) , we obtain the following.
 r1+ r2+ + rm-1
 r + r + + r     
    rm-2 + rm-1
x1+ x2++ xm = Σ Σ Σ r2+ + rm-1

rm-1
r 1=0 r 2=0 rm -1=0 1 2 m-1

 - r + + rm-1 r1 r2
 x1  1
r
x2 x3  xmm-1 (1.2)
Since (1.1) converges absolutely, this rearrangement is allowed.

3.9
Example 1: The expasion of x1+ x2+ x3
Using (1.1) ,

 r  s x x x
 r 3.9 r 3.9-r r-s s
x1+ x2+ x3 = ΣΣ
3.9
1 2 3
r=0 s=0

 0  Σs   1  Σs  x x
3.9 0 3.9
3.9
0 1 0-s s 2.9
1
1-s s
= x x 1x + x 2 3 1 2 3
s=0 s=0

 2  s  3  s x x
3.9 2 3.9
2 3 3
+ x Σ x x + 1.9
1 x Σ 2-s s
2 3
0.9
1
3-s s
2 3
s=0 s=0

 4  x Σs  5  x Σs  x
3.9 1 4 4.1 41 5 5
+ x x0.1
+ 4-s s
2 3 1.1
5-s s
2 x3 +
s=0 s=0
1 1

= x13.9
+ 3.9 x12.9 x2 + x3
+ 5.655 x11.9 x22 + 2x2 x3 + x32 
+ 3.5815 x10.9 x23 + 3x22 x3 + 3x2 x32 + x33
0.805838 4
x2 + 4x2 x3 + 6x2 x3 + 4x2 x3 + x3
3 2 2 3 4
+ 0.1
x1
0.0161168 5
x2 +5x2 x3 +10x2 x3 +10x2 x3 +5x2 x3 + x3
4 3 2 3 2 4 5
- 1.1
x1


Using (1.2) ,

r+s  s  x
  3.9 r+s
x1+ x2+ x3
3.9
= ΣΣ 3.9-r-s r
1 x2 x3s
r=0 s=0

- 12 -
 s=0  0+ s    1+ s   s  x
3.9 3.9
s 
 0+ s  1+ s
= Σ x13.9-s x20 x3s +Σ 2.9-s 1 s
1 x2 x3
s=0

 2+ s   s  x  3+ s   s  x
3.9
 2+ s  3.9 3+ s
+Σ 1.9-s 2 s
1 x2 x3 +Σ 0.9-s 3 s
1 x2 x3
s=0 s=0

 4+ s   s  x
3.9
 4+ s
+Σ -0.1-s 4 s
1 x2 x3 +
s=0
4
0.805838 x3
= x13.9 +
2.9
3.9 x1 x3 +
1.9 2
5.655 x1 x3 +
0.9
3.5815 x1 x 33 + +
x10.1
3
2.9 1.9 0.9 2 3.22335 x 2x3 0.0805838
+ 3.9 x1 x 2 + 11.31 x1 x2 x3 + 10.7445 x1 x2 x3 + -
x10.1 x11.1
2 2 2 3
4.83503 x2 x3 0.161168 x2 x3
+
1.9 2 0.9 2
+ 5.655 x1 x2 + 10.7445 x1 x2 x3 + -
x10.1 x11.1
3 3 2 3 3
3.22335 x 2 x3 0.161168 x2 x3 0.0590948 x2 x3
+
0.9 3
+ 3.5815 x1 x2 + - +
x10.1 x11.1 x12.1
4 4 4 2 4 3
0.805838 x2 0.0805838 x2 x 3 0.0443211 x2 x3 0.0310247 x2 x3
+ - + - +
x10.1 x11.1 x12.1 x13.1


we can see that what totaled this along the diagonal line is equal to the above.

Example 2: The expasion of (a + b + c + d)2.9

Although a = b +c+d in this numerical example, (1.1) and (1.2) are consistent.

Sum of General Multinomial Coefficients


 r  r  r 
 r1 rm -2 r1 rm-2
ΣΣ Σ   m (1.1")
r 1=0 r2=0 rm -1=0 1 2 m-1

- 13 -
 As expected, the following expression does not hold.

 r  r  r  = m
 r1 rm -2 r1 rm-2

ΣΣ Σ
r 1=0 r2=0 rm -1=0 1

2 m-1

It is because x 1= x2== xm = 1 does not satisfy the condition x 1 x 2+ x 3++ x m .

Let its partial sum be

 r  r  r 
n r1 rm -2 r1 rm-2
Sn = Σ Σ  Σ 
r 1=0 r2=0 r m -1=0 1 2 m-1

Then, when n  , Sn oscillates and diverges. And m

is the median of this oscillating divergent series.

In fact, applying Knopp Transformation to S n , we can obtain the approximate value of m with high precision.
However, it does not become a series but becomes an asymptotic expansion.

2007.07.06
2016.02.13 updated
2016.02.22 renewed
Kano. Kono
Alien's Mathematics

- 14 -
4 Higher Integral

4.1 Higher Primitive Function and Higher Integral

4.1.1 Higher Primitive Function

Definition 4.1.1
(x) denotes the primitive function of f <n -1>(x) for n =1, 2, 3,  ,
<n>
When f
<n>
we call f (x) Higher Order Primitive Function of f(x) ,
or for short , Higher Primitive of f(x) .

Where, f
<n>
(x) might mean indefinite integral f <n -1>
(x)dx + cn ,

f
x
<n -1>
or might mean integral function (x)dx .
an

The primitive function of f(x) is denoted by a capital letter F(x) in many cases. Then with what symbols
should the primitive function of F(x ) and its further primitive functions be denoted ?

For example, G(x), H(x),  etc. can be considered. However, by these notation, it is difficult to denote
the primitive function such as the 27 th order. So, applying the notation of the differentiation, I had denoted

F(x), G(x), H(x),  (x), f (-2)(x), f (-3)(x),  . Soon, the minus sign became obst-
(-1)
by f
ructive. Then I thought to use [n ] instead of (-n) , but this notation was as confusing as floor function.
So I came to adopt the notation f (x), f <2>(x), f <3>(x), 
<1>
using <> . The power of this notation
will become clear gradually .

Example
(sin x)<1> = -cosx + c1 c1 is an arbitrary constant.

(sin x)<2> = -sin x + c1 x + c2 c1 , c2 are arbitrary constants.

(sin x)<3> = cosx + c1 x 2 + c2 x + c3 c1 , c2, c3 are arbitrary constants.

4.1.2 Higher Integral

Definition 4.1.2
We call it Higher Integral to integrate a function f with respect to an independent variable x repeatedly.
And it is described as follows.

a a       f(x)dx dx dx   dx 
x x x x x x
 n
f(x)dx = 
n 1 an a3 a2 a1
And
when ak = a for k =1,  , n , we call it higher integral with a fixed lower limit ,

when ak  a for some k , we call it higher integral with variable lower limits .

Difference between Multiple Integral and Higher Integral

Multiple integral is what integrated f(x) with respect to x1 , x 2 ,  , xn . On the other hand, Higher integral

-1-
is what integrated f(x) with respect to x repeatedly. Then it is also called repeated integration.
By familiar expression, a triple integral is what integrates f(x,y,z) with respect to length, width and height.
On the other hand, the 3 rd order integral is what integrates f(x) 3 times with respect to only length.
As a notation of the higher integral, it is denoted in many cases as follows using a notation of multiple integral.

  ft dt dt  dt
x tn t2
1 1 2 n
a a a
However, this notation is complicated and inconvenient. Hence in this paper, for these problem solutions and
saving of variables, we replaced t1 , t2 , , tn with x and replaced dt1 dt2 dtn with dx n , as seen in
the definition.

Example

     e -1 dx = e - x -1
x x x x x
e xdx 2 = e xdx dx = x x
0 0 0 0 0

  e dx =   e dx dx =  e dx = e
x x x x x
x 2 x x x
- - - - -

   sin xdx =   (1-cosx)dx =  (x -sin x)dx = cosx + x2 -1

x x x x x x 2
3 2
0 0 0 0 0 0

   sin xdx =   (-cosx)dx =  (-sin x)dx = cosx

x x x x x x
3 2
3 2 1  3 2 3
2 2 2 2 2 2

4.1.3 Fundamental Theorem of Higher Integral

In the case of the first order integral, the relation between a primitive function and the integral is shown by
the following theorem.

Fundamental Theorem of Calculus

Let f be a continuous function defined on a closed interval [a, b] . Let F be the arbitrary primitive
function of f . Then the following equation holds for x [a, b] .

 f(x)dx = F(x) - F(a)

x

a
Where, F(a) is called Constant-of-integration .

The following theorem holds also about the Higher Integrl.

Theorem 4.1.3
r =0, 1,  , n
<r> r+1>
Let f be continuous functions defined on a closed interval I and f< be

ar , x  I .
<r>
the arbitrary primitive function of f . Then the following expression holds for

 f(x)dx 

x x n -1 x x
<n> <n -r>
n
=f (x) - Σ f an-r dx r (1.1)
an a1 r=0 an a n- r+ 1

Especially, when ar = a for r =1, 2,  ,n ,

 f(x)dx
x x
<n> n -1
<n -r> (x -a)r
n
=f (x) - Σf (a) (1.2)
a a r=0 r!

-2-
Proof
First, the following equations hold according to Fundamental Theorem of Calculus.


x
f <0>dx = f <1> - f <a11> (01)
a1

f
x
<1>
dx = f <2> - f <a22> (12)
a2

f
x
<2>
dx = f <3> - f <a33> (23)
a3

f
x
<3>
dx = f <4> - f <a44> (34)
a4


Integrating both sides of (01) with respect to x from a 2 to x ,

   dx
x x x x
f <0>dx 2 = f <1>dx - f <a11>
a2 a1 a2 a2
Substituting (12) for this,

 f  dx
x x x
<0> 2
dx = f <2> - f <a22> - f <a11>
a2 a1 a2
Integrating both sides of this with respect to x from a3 to x ,

     dx
x x x x x x x
f(x)dx 3 = f <2>dx - f <a22> dx - f <a11> 2
a3 a2 a1 a3 a3 a3 a2
Substituting (23) for this,

    dx
x x x x x x
f(x)dx 3 = f <3> - f <a33 > - f <a22> dx - f <a11 > 2
a3 a2 a1 a3 a3 a2
Integrating both sides of this with respect to x from a4 to x ,

  f(x)dx =  f  dx - f   dx    dx
x x x x x x x x x
4 <3> <3> <2> 2 <1> 3
dx - f a3 a2 -f a1
a4 a1 a4 a4 a4 a3 a4 a3 a2
Substituting (34) for this,

      dx
x x x x x x x x
 f(x)dx 4 = f <4> - f <a44> - f <a33> dx - f <a22> dx 2 - f <a11 > 3
a4 a1 a4 a4 a3 a4 a3 a2
Hereafter by induction, we obtain the following expression.

 
x x n -1 x x
 n <n>
f(x)dx = f
<n> <n -r>
-fa -Σfa  dx r
an a1 n r=1 n-r an a n- r+ 1
<n>
And pushing f an into Σ, we obtain (1.1) .

Especially, when a r = a for r =1, 2,  ,n ,


x x (x -a)r
dx = r
r =1, 2,  ,n -1
a a r!
Then, using this, we obtain

 f(x)dx
x x
<n> n -1
<n -r> (x -a)r
n
=f (x) - Σf (a) (1.2)
a a r=0 r!

-3-
Constant-of-integration Polynomial
(x -a)r

n -1 x x n -1
<n-r>
Σf a n-r  Σ
<n-r>
We call dx r and f (a) in the theorem
r=0 an an-r +1 r=0 r!
Constant-of-integration Polynomial.

In the case of (1.1) , it is very difficult to express the higher integral of 1 by a polynomial.
For reference, if this is calculated by force and the 4th order integral of f(x) is expressed by a polynomial,
it is as follows. The 5th order or more are unmanageable...
1

 x -a 4 x -a 4x -2a 3+ a 4
x x
 f(x)dx = f 4 <4>
-fa <4>
-fa <3>
- f <a2>
a4 a1 4 3 1! 2 2!
x -a 4x + a 4 x -3a2 x + 6a 2 a3 -3a 2 a4 -3a 3 + a 4 
2 2 2
<1>
-fa
1 3!
Since the remainder term of (1.2) is already expressed by the polynomial, we can find out easily that it is
the (n-1)-th order. We can also express it explicitly as follows, though there is no merit not much.
n
f < -s>

x x n -1 n -1
s-r a
 f(x)dx = f (x) -
n <n> r
cr+1 x , cr+1 = Σ (-1) sＣ s-r a
s-r
Σ
a a r =0 s =r s!

Example1
1

a a f(x)dx a dx x -a2
x x x
2 <2> <2> <1> <2> <2> <1>
=f -fa -fa =f -fa -fa
2 1
2 1
2
2 1 1!
<1> <2>
Let f = e x , e x = e x + c1 , e x = e x + c1 x + c2
Substituting these for the above experession, we obtain as follows.
1

a a a 1 x -a 2
x x
x 2 x a2
Left: e dx = e - e - e
2 1
1!
1
x <2> <2> <1> x -a 2
Right: e  - e xa2 - e xa1
1!
1
x -a2
= e + c1 x + c2 - e - c1 a2- c2 - e + c1
x a2 a1
1!
1
a a1 x -a2
= ex - e 2 - e
1!

Example2

a a a a a a dx
x x x x x x
f(x)dx 3 = f <3> - f <a33> - f <a22> dx - f <a11> 2

3 2 1 3 3 2
1
x -a 3 x -a 3x -2a 2+ a 3
= f <3> - f <a3> - f <a2> - f <a1>
3 2 1! 1 2!
<1> <2> <3>
Let f = sin x , (sin x) = -cosx , (sin x) = -sin x , (sin x) = cosx
Substituting these for the above experession, we obtain as follows.

-4-
 a a a sin x dx
x x x
3
Left:
3 2 1
x -a3cosa1
2 2
= cosx -cosa3 + + x -a3sin a2 - a2x -a3cosa1
2!
1
x -a 3 x -a 3x -2a2+ a3
Right: cosx - cosa3 + sin a2 + cosa1
1! 2!
x -a32cosa1
2
= cosx -cosa3 + x -a3sin a2 + - a2x -a3cosa1
2!

4.1.4 Lineal and Collateral

Definition 4.1.4

 
x x n -1 x x
 <n>
f(x)dx = f (x) - Σ f
n <n -r>
an-r  dx r (1.1)
an a1 r=0 an a n- r+ 1
In this expression,
when Constant-of-integration Polynomial is 0,


x x
we call  f(x)dx n Lineal Higher Integral and
an a1
we call the function equal to this Lineal Higher Primitive Function .

(or for short , Lineal Higher Primitive. )

when Constant-of-integration Polynomial is not 0,


x x
we call  f(x)dx n Collateral Higher Integral and
an a1
we call the function equal to this Collateral Higher Primitive Function .
(or for short , Collateral Higher Primitive. )
These are the same also in (1.2).

In short , Lineal Higher Primitive Function is what integrated f(x) with respect to x repeatedly
without considering the constant of the integration.

Example
Left: Collateral Higher Integral Right: Collateral Higher Primitive

   sin xdx
x x x 1
3 (x -3) (x -3)(x -22+3)
= cosx - cos 3+ sin 2 + cos 1
3 2 1 1! 2!

  e dx
x x
x 2
= e x-1- x
0 0

Left: Lineal Higher Integral Right: Lineal Higher Primitive

 
x x x
sin xdx 3 = cosx
3 2 1 
2 2 2

  e dx
x x
x 2
= ex
- -

-5-
4.1.5 The Conditions for the Higher Integral being Lineal
In Theorem 4.1.3 (1.1), since the higher integral of 1 can be arbitrary value, in order for Constant-of-integration

ar = 0 for r = 1, 2,  , n . Since this becomes

<r>
Polynomial to be 0, we find out that it must be f
important later, we state here as a theorem.

Theorem 4.1.5

 f(x)dx 

x x n -1 x x
<n> <n -r>
n
=f (x) - Σ f an-r dx r (1.1)
an a1 r=0 an a n- r+ 1

 f(x)dx
x x
<n> n -1
<n -r> (x -a)r
a a
= f (x) -
r=0
n
f Σ (a)
r!
(1.2)

In these expressions, the necessary conditions for higher integrals being lineal are as follows respectively

r = 1, 2,  , n
<r>
f ar = 0
r = 1, 2,  , n
<r>
f (a) = 0

That is, The necessary condition for the higher integral being lineal is that a r or a is a zero of the higher
<r>
primitive function f for each r from 1 to n.

Note
These are necessary conditions and are not sufficient conditions. For example, let

 f(x)dx
x x
r = 1, 2,  , n
<r> r
f (x) =
ar a1

ar = 0 r =1, 2,  ,n
<r>
Then, whether or not this higher integral is lineal, certainly f
In the case of f(x)=e x , let

  e dx
x x
<r>
f (x) =  x r
r = 1, 2
0 0

Then, f
<1>
(x)= e x -1 , f <2>(x)= e x -1- x and the lower limit 0 is a zero of these primitive functions.
Nevertheless, substituting these for (1.2), we obtain


x x x0 x1
e dx = e -1- x - e -1-0
x 2 x
- e -1
0
= e x -1- x
0
0 0 0! 1!
This is not lineal higher integral.

-6-
4.2 Reimann-Liouville Integral and Higher Integral

  f(x)dx
x x
When ak = a , k =1, 2,  , n the higher ingegral  n
can result in the single integral
an a1
called Reimann-Liouville Integral .

Theorem 4.2.1 ( Cauchy formula for repeated integration )

When f(x) is continuously integrable function and (z) denotes the gamma function, the following
equation holds.

  (x -t)
x x 1 x
 f(x)dx n =
n-1
f(t)dt (2.1)
a a (n) a

Proof
First, Reimann-Liouville Integral can be expanded as follows.

 (x -t)
1 x
n-1 <n> n -1 (x -a)r <n -r>
f(t)dt = f (x) - Σ f (a) (2.2)
(n) a r=0 r!
Because,

  f(t)dt = f
1 x x x
0
(x -t) f(t)dt =
<1>
(t)a
(1) a a
<1>
= f (x) - f <1>(a)

 f
x x

 
1 1 1 <1> x <1>
(x -t) f (t)a +
1
(x -t) f(t)dt = (t)dt
(2) a 1! a
1 <1> <2> <2>
= -(x -a) f (a)+ f (x)- f (a)

 
x x

 
1 1 2 <1> x 1 <1>
(x -t) f (t)a + 2 (x -t) f (t)dt
2
(x -t) f(t)dt =
(3) a 2! a


1 x x
2 <1> 1 <2> <>
=- (x -a) f (a )+ (x -t) f (t)a + f 2 (t)dt
2! a
2 1
(x -a ) <1> (x -a ) <2> <> < >
=- f (a)- f (a)+ f 3 (x )- f 3 (a)
2! 1!


 (x -t)
1 x
n-1 <n> n -1 (x -a)r <n -r>
f(t)dt = f (x) - Σ f (a) (2.2)
(n) a r=0 r!
On the other hand, Theorem 4.1.3 (1.2) was as follows.

 f(x)dx
x x
<n> n -1
<n -r> (x -a)r
n
=f (x) - Σf (a) (1.2)
a a r=0 r!
From these, we obtain the desired expressin.

Note
This theorem shows that the higher integral with fixed lower limit is equivalent to Riemann-Liouville integral.
Therefore, we can examine the higher integral in left side by Riemann-Liouville integral in right side. Moreover,
this theorem shows that we can not apply Riemann-Liouville Integral to the higher integral with variable lower
limits. So, we can not apply Riemann-Liouville integral to the higher integral of sin x etc.. Strictly speaking,
if we apply it by force, then we obtain the collateral higher integral inevitably.

-7-
Reimann-Liouville Integral and Taylor's Theorem
The right side of (2.2) is rewritten as follows.

 (x -t)
<n> n -1
<n -r> (x -a)r 1 x
n-1
f (x) = Σf (a) + f(t)dt (2.2)
r=0 r! (n) a
<n>
Then we notice that this is Taylor expansion of f (x) around a and Reimann-Liouville Integral is
the remainder term.
Actually, from this, we can derive usual Taylor's theorem. Let us shift by -n the index of the integration
operator of f(x) . Then

 (x -t)
n -1
(r) (x -a)r 1 x
n-1 (n)
f <0>
(x) = Σ f (a) + f (t)dt (2.20)
r=0 r! (n) a
(n)
Just, this is the Taylor expansion of f(x) around a . And Reimann-Liouville Integral of f is the remainder
term called Bernoulli form .
And according to the first mean value theorem for integration, when g(t) is a continuous real valued function
on the [a,b] and (t) is an integrable function that does not change the sign on the a,b , there exists
a  (a,b ) such that

   (t)dt
b b
g(t) (t)dt = g( )
a a
Applying this to (2.20) , we obtain the following.

f ( )
(n)

  (x -t)
1 x x
n-1 (n) n-1
(x -t) f (t)dt = dt
(n) a (n -1)! a

f
(n)
( ) n x (x -a)n (n)
= -(x -t) a = f ( )
n! n!
This is the remainder term called Lagrange form .

-8-
4.3 Higher Integrals of Power Function etc.

Formula 4.3.1 : Higher Integral of Power Function

When (z) denotes a gamma function, the following expressions hold.
(1) Basic form
(1+ ) +n
 x
x x

 dx n = x (  0)
0 0 1+ + n 
(- - n) +n
  x
x x

dx n = (-1)n x ( <-n)
  (- )
(2) Linear form
(1+ )
 
x n

 
x 1
 (ax +b) dx n = (ax +b)+n (  0)
b b a (1+ +n)
- -
a a

(- - n)
 
x n

 
x 1

(ax +b) dx = - n
(ax +b)+n ( <-n)
  a (- )

Proof
When   0 , the zero of the higher primitive functions of (ax +b)  is -b /a . Hence


x 1 1
(ax +b) dx = (ax +b)+1
b a  +1
-
a


x x 2

 
1 1
(ax +b) dx  2
= (ax +b)+2
b b a ( +1)(  +2)
- -
a a


 
x n

 
x 1 1
 (ax +b) dx =  n
(ax +b)+n
b
-
b a ( +1)( +2)( +n)
-
a a
1 (1+)
Substituting = for this, we obtain (2).
( +1) ( +2) ( + n )  ( 1+ + n )
Furthermore, substituting a =1, b =0 for this, we obtain (1).

 < -n  - (ax +b)  = (ax +b) 

-
When , if we replace with then . And the zero of higher
-
primitive functions of (ax +b) is . Therfore,


x 1 1
(ax +b)-dx = (ax +b)-+1
 a - +1

  (ax +b)
2

 
x x 1 1
-
dx =2
(ax +b)-+2
  a (- +1)(- +2)


 (ax +b)

n

 
x x 1 1
-
dx = n
(ax +b)-+n
  a (- +1)(- +2)(- + n)

-9-
Since

1 (-1)n
=
(- +1)(- +2)(- + n) ( -1)( -2)( - n)
 [1+ -(n +1)] ( -n)
= (-1)n = (-1)n
1+( -1) ()
we can substitute this for the above as follows.

( - n)

n

 
x x 1 -
 (ax +b) dx = - n
(ax +b)-+n
  a ( )
Then replacing  with - we botain (2), furthermore, substituting a =1, b =0 for this, we obtain (1).

Note
When -n   < 0 , we do not dare define the higher integral of the power function. It is because the lower
limit of the integral changes irregularly in this case. For example as follows.
5

 x  x
x x x x x x  8
2 3 2 3
dx = -x(log x -1) , dx = x
0 1  0   3

Examples

Formula 4.3.2 : Higher Integral of Exponential Function

(1) Basic form

 e
x x
x
 dx n = (1)ne  x
 

(2) Linear form

 
n

 a <0 : +
x a >0 : -
 
x 1
 e
ax+b
dx n
= e
ax+b
  a

Proof
When a >0 the zero of higher primitive functions of e ax+b is - , and when a <0 the zero of same one
is + . Hence we obtain (2). Furthermore, substituting a = 1 , b =0 for this, we obtain (1).

- 10 -
Formula 4.3.3 : Higher Integral of Logarithmic Function

(1) Basic form

  log x dx
x xn
 
x n 1
n
= log x -Σ
0 0 n! k =1 k

(2) Linear form

 
x n

  
x 1 b n 1
 log(ax +b)dx n = x+ log(ax +b) -Σ
b b n! a k =1 k
- -
a a

Proof
m+1 m+1


x x x
m
x log x dx = log x -
0 m +1 (m +1)2
Therfore,


x1 x1 x1
   
x 1 1
log x dx = x(log x - 1) = log x - = log x -
0 1! 1! 1 1! 1

 log x dx =  x(log x - 1)dx =  x log x dx -  x dx

x x x x x
2
0 0 0 0 0

 
2 2 2 2
1 x x x x x2 x2
= log x - 2 - = log x - -
1! 2 2 2 2! 2!2 2!
x2 x2 x2
    
1 1
= log x - 1+ = log x - 1+
2! 2! 2 2! 2

   
x x x 1 x 1 1 x
log x dx 3 = x 2 log x dx - 1+ x 2dx
0 0 0 2! 0 2! 2 0

 
x3 x3
 
1 1 1 3
= log x - 2 - 1+ x
2! 3 3 3! 2
x3 x3
 
1 1 3
= log x - - 1+ x
3! 3!3 3! 2
x3 x3 x3
    
1 1 1 1
= log x - 1+ + = log x - 1+ +
3! 3! 2 3 3! 2 3
Hereafter, we obtain (1) by induction. And replacing x with ax +b , we obtain (2) by similar way.

Example: 3rd order integral of log x

- 11 -
Formula 4.3.4 : Higher Integrals of sin x , cos x

(1) Basic form

n
   
x x x
 sin xdx n = sin x -
n 2 1 2
2 2 2
n
   
x x x
 cosxdx n = cos x -
(n -1) 1 0 2
2 2 2
(2) Linear form
n
 
n

   
x x 1
 sin(ax +b)dx n
= sin ax +b -
n b 1 b a 2
- -
2a a 2a a

n
 
n

   
x x 1
 cos(ax +b)dx = n
cos ax +b -
(n -1) b 0 b a 2
- -
2a a 2a a

Proof
Lineal higher primitive function of f(x)= sin x and its zeros are as follows.
1 1
f<1>(x) = -cosx 
= sin x -
2  , x=
2
2 2
f<2>(x) = -sin x 
= sin x -
2  , x=
2
3 3
f<3>(x) = cosx 
= sin x -
2  , x=
2
4 4
f<4>(x) = sin x 
= sin x -
2  , x=
2

Thus we obtain (1). And replacing x with ax +b cos x
, we obtain (2) by similar way. For , it is obtained by
deducting  /2 from the lower limit of the higher integral in the left side and replacing sin with cos in the
right hand.

Termwise higher integrals of sinx , cosx

If common value 0 is adopted as lower limits of higher integrals of sin x, cos x, we obtain the following
termwise higher integral. These are collateral higher integrals as understood from the constant-of-integration
polynomial in the right side

 sinxdx
x x  (-1)k 2k+1+ n
n
=Σ x
0 0 k =0 (2k +1+ n)!
n k
n-k

  +Σ
n x
= sin x - sin
2 k =1 (n -k)! 2


x x 
k
(-1)
 cosxdx n = Σ x 2k+
n
0 0 k =0 (2k + n)!

- 12 -
 n k
n-k

  -Σ
n x
= cos x - cos
2 k=1 (n -k)! 2

If these are written without using Σ , it is as follows.

  cosxdx
x x
sinxdx = -cosx + 1 , = sinx
0 0

 sinxdx x1
 cosxdx
x x x x
2 2
= -sinx + , = -cosx + 1
0 0 1! 0 0

 sinxdx x2
 cosxdx x1
x x x x x x
3 3
= cosx + -1 , = -sinx +
0 0 0 2! 0 0 0 1!

 sinxdx  cosxdx

1 3
x x
4 x x x x
4 x2
= sinx - + , = cosx + -1
0 0 1! 3! 0 0 2!
 

Formula 4.3.5 : Higher Integral of sinhx , coshx

(1) Basic form
n i
  
e x - (-1)ne -x
 
x x x
n n
sinh xdx = i sinh x - =
n i 2 i 1 i 2 2
2 2 2

n i
    
e x + (-1)ne -x
x x x
 n n
cosh xdx = i cosh x - =
(n -1) i 1 i 0 i 2 2
2 2 2
(2) Linear form
n i
 
n

 
x

 
x i
 sinh(ax +b)dx = n
sinh ax +b -
n i b 1 i b a 2
- -
2a a 2a a
n

  e
1 1
= ax+b
- (-1)ne -(ax+b)
2 a
n i
 
n

 
x

 
x i
 cosh(ax +b)dx = n
cosh ax +b -
(n -1) i b 0 i b a 2
- -
2a a 2a a
n

  e
1 1
= ax+b
+ (-1)ne -(ax+b)
2 a

Proof
Lineal higher primitive function of f(x)= sinh x and its zeros are as follows.
1 i 1 i
f<1>(x) = cosh x 
= i sinh x -
2
,  x=
2
2 i 2 i
f<2>(x) = sinh x 
= i 2sinh x -
2  , x=
2

- 13 -
 3 i 3 i
f<3>(x) = cosh x = i 3sinh x - 2  , x=
2
4 i 4 i
f<4>(x) = sinh x = i 4sinh x - 2  , x=
2

Thus we obtain the hyperbolic form of (1). And replacing x with ax +b , we obtain the hyperbolic form of (2)
by similar way. For cosh x , it is obtained by deducting  i/2 from the lower limit of the higher integral in
the left side and replacing sinh with cosh in the right side.
Next,
n i n i
n i

sinh ax +b -
2  =
2
e 
1 ax+b- 2
-e
 - ax+b -
2 

n i i -n

=
e
-
2
e
ax+b
-e -(ax+b -n  i)
=
e 2  e ax+b - e  i e -(ax+b)
n
2 2
-n
i
e - (-1)ne -(ax+b)
ax+b
=
2
n i n i
n i

cosh ax +b -
2
=
2
e 
1 ax+b- 2
+e 
-ax+b -
2 
-n
i
e + (-1)ne -(ax+b)
ax+b
=
2
Therefore, using these we botain exponential forms of (1), (2).

Termwise higher integrals of sinhx , coshx

If common value 0 is adopted as lower limits of higher integrals of sinhx, coshx , we obtain the following
termwise higher integral. These are collateral higher integrals as understood from the constant-of-integration
polynomial in the right side

 sinhxdx
x x  1 2k+1+ n
n
=Σ x
0 0 (2k +1+ n)! k =0

n i k i
n-k

 
n x
n
= i sinh x - -Σ i -k sinh
2 k =1 (n -k)! 2


x x  1
 coshxdx n = Σ x 2k+
n
0 0 k =0 (2k n
+ )!
n i k i
n-k

 
n x
n
= i cosh x - -Σ i -k cosh
2 k =1 (n -k)! 2

- 14 -
4.4 Higher Integrals of Inverse Trigonometric Functions

-1 -1
Formula 4.4.1 : Higher Integrals of tan x , cot x
When  , ,  (x) denote floor function, ceiling function and digamma function respectively, the following
expressions hold for a natural number n2.


x x tan -1x n /2
-1
tan x dx = Σ
n
(-1)k nＣ n-2k x n-2k
0 0 n ! k =0
log1+ x 2 n /2
+ Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k =1

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)x n+1-2r
n! r=1


x x xn tan -1x n /2
(-1)k nＣ n-2k x n-2k
n! Σ
-1 -1n
cot x dx = cot x -
0 0 n! k =1

log1+ x 2 n /2
- Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k =1

1 n /2
+ Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)x n+1-2r
n r=1
!

Proof


x 1
tan -1x dx = xtan -1x - log1+ x 2
0 2
1 1
1Ｃ 1 xtan 1Ｃ 0 x log 1+ x 
-1 0 2
= x-
1! 21!

 tan  
x x x2 1 x x
-1
x dx = 2
- tan -1x - log1+ x 2 +
0 0 2 2 2 2

 
1 1 1 1 x
2Ｃ 2 x -2Ｃ 0 tan x - 2Ｃ 1 x log 1+ x  +
2 -1 1 2
=
2! 22! 1! 2 1!

 tan    
x x x x3 x x2 1 5x 2
-1
x dx =3
- tan -1x - - log1+ x  +
2
0 0 0 6 2 4 12 12
1
3Ｃ 3 x -3Ｃ 1 x tan x
3 -1
=
3!
x2
 
1 1 1 1
3Ｃ 2 x -3Ｃ 0 x  log 1+ x  +
2 0 2
- +
23! 1! 2 3 2!

  
x4 x2 1 3

 
x x 3
x x 13x x
-1
tan x dx = 4
- + - log 1+ x 2 +
tan -1x - -
0 0 24 4 24
12 12 72 24
1 1
4Ｃ 4 x -4Ｃ 2 x +4Ｃ 0 tan x - 4Ｃ 3 x -4Ｃ 1 x  log1+ x 
4 2 -1 3 1 2
=
4! 24!
x3
    1!
1 1 1 1 1 1 x
+ + + -
1! 2 3 4 3! 3! 4

- 15 -

x
   
5 3 4 2
x x x x x x 1 77x 4 3x 2
 tan -1x dx 5 = - + -1
tan x - - + 2
log 1+ x  + -
0 0 120 12 24 48 24 240 1440 80
1 1
5Ｃ 5 x -5Ｃ 3 x +5Ｃ 1 x tan x - 5Ｃ 4 x -5Ｃ 2 x +5Ｃ 0 x  log1+ x 
5 3 -1 4 2 0 2
=
5! 25!
x4 x2
   
1 1 1 1 1 1 1 1
+ + + + - +
1! 2 3 4 5 4! 3! 4 5 2!
Hereafter, by induction, we obtain the desired expression.

The desired expression for cot -1x is also obtained in a similar way.

Example: 4th order integral of tan -1x

-1 -1
Formula 4.4.2 : Higher Integrals of sin x , cos x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n .
n-2r

 sin
x x n/2 x
-1
x dx = Σ
n
2 sin -1x
an a1 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s
s
2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
Where a1 = i 1.5088 , a 2 = 0 , a3 = -i 0.4758 , a4 = 0 , 
n-2r


x x xn n/2 x
 cos x dx = -1
cos -1x - Σ
n
2 sin -1x
an a1 n! r=1 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
- Σ Σ (-1) n -2r+1Ｃ ss
2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
Where a1 = 1 , a2 = 0 , a3 = a complex number , a4 = 0 , 

- 16 -
Proof

 sin
x
-1
x dx = xsin -1x + 1-x 2
a1

x1 -1 1(-1)!!2 x 0
= sin x + 1-x 2
0!! 1!
2
1!!2 0!

 sin  
x x
-1 2 x2 1 3x
x dx = + sin -1x + 1-x 2
a2 a1 2 4 4

   
x2 x0 1(-1)!!2 10!!2 x1
= + sin -1x + - 1-x 2
0!! 2! 2!! 0!
2 2
1!!2 2!!2 1!

 sin    
x x x
-1 3 x3 x 11x 2 1
x dx = + sin -1x + + 1-x 2
a3 a2 a1 6 4 36 9

 
x3 x1
= + sin -1x
0!! 3! 2!! 1!
2 2

 1!! 2!! 3!!  2! 3!! 0!

2 2 2 2 2 0
1(-1)!! 20!! 11!! x 1(-1)!! x
+ - +
2
+ 2 2 2
1-x 2

  sin    
x x x4 x2 1 25x
3
55 x
 -1
x dx = 4
+ + sin -1x + + 1-x
2
a4 a1 24 8 64 288 576

 
x4 x2 x0
= + + sin -1x
0!! 4! 2!! 2! 4!! 0!
2 2 2

  
x3 x1
 
1(-1)!!2 30!!2 31!!2 12!!2 1(-1)!!2 10!!2
+ - + - + -
1!!2 2!!2 3!!2 4!!2 3! 3!!2 4!!2 1!


2
1-x
Hereafter, by induction, we obtain the desired expression.
The desired expression for cos -1x is also obtained in a similar way.

Note
Integration lower limits a1 , a2 ,  , an are the solutions of the transcendental equations which made
-1
the formal right-hand sides equal to 0. For example, in the case of sin x , a1 , a2 are the solutions
of the following transcendental equation respectively.

2 4
-1 2 x2 1 -1 3x
xsin x + 1-x = 0 , + sin x + 1-x 2 = 0
4

On the other hand, if the integration lower limit is replaced by a fixed lower limit 0 in this formula,
the following collateral high integrals are obtained. The difference among both is only the existence
of a constant-of-integration polynomial .

-1 -1
Formula 4.4.2' : Collateral Higher Integrals of sin x , cos x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n .

- 17 -
 n-2r


x x n/2 x
sin x dx = Σ
-1 n
2 sin -1x
0 0 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
+ Σ Σ (-1) n -2r+1Ｃ s
s
2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
n-2r+1
n /2 x
+Σ
r=1 (2r -1)!!2(n -2r +1)!
n-2r


x x xn n/2 x
-1
cos x dx = n
cos -1x - Σ 2 sin -1x
0 0 n! r=1 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
-Σ Σ s
(-1) n -2r+1Ｃ s 2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
n-2r+1
n /2 x
+Σ
r=1 (2r -1)!!2(n -2r +1)!
-1
Example: Collaterral the 3rd order integral of cos x

-1 -1
Formula 4.4.3 : Higher Integrals of sec x , csc x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n2.


x x
-1 xn n
sec xdx = sec -1x
1 1 n!
n-2r-1
(n -1)/2 (2r -1)!! x
- Σ logx + x 2-1
r=0 (2r)!!(2r +1)! (n -2r -1)!

- 18 -
 n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 x 2-1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!


x x xn
 csc x dx = -1
csc -1x n
an a1 n !
n-2r-1
(n -1)/2 (2r -1)!! x
+ Σ logx + x 2-1
r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
-Σ Σ (-1) 2 x 2-1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
Where a1 , a2 ,  , an are all complex numbers.

Proof

 sec
x
-1
xdx = x sec -1x - logx + x 2-1
1

x1 (-1)!! x 0
= -1
sec x - logx + x 2-1
1! 0!!1! 0!


x x x2 1
sec xdx =-1
sec -1x - x log x + x 2-1 +
2
x 2-1
1 1 2 2
x2 (-1)!! x 1 x 0 (-1)!!2
= -1
sec x - logx + x 2
-1 0!
+ x 2-1
2! 0!!1! 1! 21!!2

  
x x x x3 x2 1 5x
-1
sec xdx = 3
sec -1x - + logx + x 2-1 + x 2-1
1 1 1 6 2 12 12

x-
0!!1! 2! 2!!3! 0! 
x3 (-1)!! x 1!! x 2 0
= sec -1 + logx + x 2-1
3!

 
x1 (-1)!!2 0!!2
+ - x 2-1
1! 21!!2 32!!2

    
x 4 3 2
x x x x 13x 1
 sec -1x dx 4 = sec -1x - + logx + x 2-1 + + x 2-1
1 1 24 6 12 72 36

 
x4 (-1)!! x 3 1!! x 1
= sec -1x - + logx + x 2-1
4! 0!!1! 3! 2!!3! 1!

   
x2 (-1)!!2 20!!2 1!!2 x 0 (-1)!!2
+ - + + x 2-1
2! 21!!2 32!!2 43!!2 0! 43!!2
Hereafter, by induction, we obtain the desired expression.
The desired expression for csc -1x is also obtained in a similar way.

-1
Example: The 5th order integral of sec x

- 19 -
Since both sides overlapped exactly, the blue (Riemann-Liouville Integral) can not be seen.

On the other hand, if the integration lower limit is replaced by a fixed lower limit 1 in this formula, the following
collateral high integral is obtained. The difference among both is only the existence of a constant-of-integration
polynomial.

-1
Formula 4.4.3' : Collateral Higher Integral of csc x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n2.


x x
-1 n xn
csc x dx = csc -1x
1 1 n!
n-2r-1
(n -1)/2 (2r -1)!! x
+ Σ logx + x 2-1
r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
-Σ Σ (-1) 2 (n -2r)! x
2
-1
r=1 s=0 2r +s (s +2r -1)!!
 n -1
- Σ
2n! r=0
nＣ r(x -1)
r

- 20 -
4.5 Higher Integrals of Inverse Hyperbolic Functions
-1 -1
Formula 4.5.1 : Higher Integrals of tanh x , coth x
When  , ,  (x) denote floor function, ceiling function and digamma function respectively, the following
expressions hold for a natural number n2.


x x tanh -1x n /2 n-2k
n! Σ
nＣ n-2k x
-1 n
tanh x dx =
0 0 k =0

log1- x 2 n /2 n+1-2k

+
2n ! Σk =1
nＣ n+1-2k x

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
n! Σ
-
r=1


x x xn tanh -1x n /2 n-2k
n
Σ nＣ n-2k x
-1 -1
coth x dx = coth x +
0 0 n! n ! k =1
log1-x 2 n /2 n+1-2k
+
2n ! k =1Σ nＣ n+1-2k x

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
- Σ
n! r=1
|x|<1

Proof
The integral up to the 4th order of tanh -1x are as follows.


x 1
tanh -1x dx = xtanh -1x + log1- x 2
0 2

  2 2
x x 2
x 1 x x
 
-1 2 -1 2
tanh x dx = + tanh x + log 1- x -
0 0 2 2

  6 2  4 12 
x x x 3 2 2
x x x 1 5x
 
-1 3 -1 2
tanh x dx = + tanh x + + log 1- x -
0 0 0 12

 tanh x dx = 24 + 4 + 24  tanh x+ 12

4 2 3


x x 3
x x 1 x x 13x x
-1 4
+ log 1- x  - -1 2
-
0 0 12 72 24
The absolute values of these coefficients are all consistent with the absolute values in Formula 4.4.1 .
Therefore, these coefficients can be easily obtained by changing the signs of the coefficients in Formula 4.4.1.
This is the same also for coth -1x .

-1
Example: The 3rd order integral of coth x

- 21 -
 -1 -1
Formula 4.5.2 : Higher Integrals of sinh x , cosh x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n .

(-1)r x n-2r
 sinh
x x n/2
-1
x dx = Σ n
2 sinh -1x
an a1 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1
r+s (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1+ x 2
r=1 s=0 (s +2r -1)!!
Where a1 = 1.5088 , a2 = 0 , a 3 = -0.4758 , a4 = 0 , 
n-2r

 cosh
x x n/2 x
-1
x dx = Σ n
2 cosh -1x
1 1 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
-Σ Σ s
(-1) n -2r+1Ｃ s 2 (n -2r +1)! x
2
-1
r=1 s=0 (s +2r -1)!!

Proof
The integral up to the 4th order of sinh -1x are as follows.

 sinh
x
-1
x dx = xsinh -1x - 1+ x 2
a1

 sinh  
x x
-1 2 x2 1 3x
x dx = - sinh -1x - 1+ x 2
a2 a1 2 4 4

 sinh     1+ x
x x x
-1 3 x3 x 11x 2 1
x dx = - sinh -1x - - 2
a3 a2 a1 6 4 36 9

 sinh =  sinh x - 

288 576 
x x 4 2 3
-1 4 x x 1 25x 55 x -1
x dx - + - 1+ x 2
a4 a1 24 8 64
The absolute values of these coefficients are all consistent with the absolute values in Formula 4.4.2 .
Therefore, these coefficients can be easily obtained by changing the signs of the coefficients in Formula 4.4.2.
This is the same also for cosh -1x .

-1
Example: The 4th order integral of cosh x

- 22 -
 On the other hand, if the integration lower limit is replaced by a fixed lower limit 0 in this formula, the following
collateral high integral is obtained. The difference among both is only the existence of a constant-of-integration
polynomial.

-1
Formula 4.5.2' : Collateral Higher Integral of sinh x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n .

(-1)r x n-2r
 sinh
x x n/2
-1
x dx = Σ
n
2 sinh -1x
0 0 r=0 (2r)!! (n -2r)!
n-2r+1
n /2 n -2r+1
r+s (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1+ x 2
r=1 s=0 (s +2r -1)!!
n /2 (-1)r-1 x n-2r+1
+Σ
r=1 (2r -1)!!2(n -2r +1)!

-1 -1
Formula 4.5.3 : Higher Integrals of sech x , csch x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n2.


n-2r-1
x x xn (n -1)/2 (2r -1)!! x
 sech x dx = sech x + Σ
-1-1 n
sin -1x
an a1 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
+ Σ Σ (-1) 2 (n -2r)!
1- x 2
r=1 s=0 2 r +s (s +2r -1)!!
Where a1 = a3 = a5 =  = 0 , a2 , a4 , a6 ,  are complex numbers
n-2r-1


x x xn (n -1)/2 (-1)r(2r -1)!! x
 csch x dx = csch x + Σ
-1
-1 n
sinh -1x
an a1 n ! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
r+s n -2rＣ s(s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 x 2+1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
Where a1 = 0 , a 2 = 0.6079 , a 3 = 0 , a4 = 1.5539 , 

- 23 -
Proof
The integral up to the 4th order of sech -1x are as follows.

 sech
x
-1
x dx = x sech -1x + sin -1x
a1

 sech
x x
-1 2 x2 1 1
x dx = sech -1x + xsin -1x + 1- x 2 -
a2 a1 2 2 2

  
x x x x3
-1 3 x2 1 5x
sech x dx = sech -1x + + sin -1x + 1- x 2
a3 a2 a1 6 2 12 12

 sech 
x3 x
  
2
x x
-1 x4 13 x 1
x dx 4 = sech -1x + + sin -1x + + 1- x
2
a4 a1 24 6 12 72 36
The absolute values of these coefficients are all consistent with the absolute values in Formula 4.4.3.
Therefore, these coefficients can be easily obtained by changing the signs of the coefficients in Formula 4.4.3.
This is the same also for csch -1x .

On the other hand, if the integration lower limit is replaced by fixed lower limit 0 in this formula,
the following collateral high integrals are obtained. The difference among both is only the existence
of a constant-of-integration polynomial.

-1 -1
Formula 4.5.3' : Collateral Higher Integrals of sech x , csch x
When  ,  denote floor function and ceiling function respectively, the following expressions hold for
a natural number n2.


n-2r-1
x x xn (n -1)/2 (2r -1)!! x
-1
sech x dx = sech x + Σ
n-1
sin -1x
0 0 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2
1- x 2
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
n-2r
n /2 1 x
-Σ
r=1 2r(2r -1)!!2 (n -2r)!


n-2r-1
x x xn (n -1)/2 (-1)r(2r -1)!! x
-1
csch x dx = csch x + Σ
n-1
sinh -1x
0 0 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
r+s n -2rＣ s (s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 x 2+1
r=1 s=0 2r +s (s +2r -1)!! (n -2r)!
n-2r
n /2 (-1)r x
-Σ
r=1 2r(2r -1)!!2 (n -2r)!

-1
Example: Collateral the 4th order integral of sech x

- 24 -
Since both sides overlapped exactly, blue (Riemann-Liouville Integral) can not be seen.
This result also shows that Formula 4.5.3 is right.

- 25 -
4.6 Termwise Higher Integral and Taylor series of higher primitive function

4.6.1 Theorem of the Termwise Higher Integral with a fixed lower limit

Theorem 4.6.1
r =0, 1,  , n
<r> <r +1>
Let f be continuous functions defined on [a , b] and f be the arbi-trary
<r>
primitive function of f . At this time, iff(x) can be expanded to the Taylor series around a ,
the following expressions hold for x [a , b] .
<n> n -1 (x -a)r
<n -r> (r) (x -a)n+ r
f (x) = Σf (a) + Σ f (a) (1.1)
r=0 r! r=0 (n + r)!
 (x -a)r
= Σf <n -r>(a) (1.2)
r=0 r!

Proof
The following expression holds according to Theorem 4.1.3 .

  f(x)dx
n -1 (x -a)r x x
f
<n>
(x) = Σf <n -r>
(a) +  n
r=0 r! a a
On the other hand, if f(x ) can be expanded to the Taylor series around a ,
(r)  (x -a)r
f(x) = Σf (a)
r=0 r!
This can be integrated termwise as follows.

 f(x)dx
x x  (r) (x -a)n+ r
n
=Σf (a)
a a r=0 (n + r)!
Substituting this for the above,

<n> n -1
<n -r> (x -a)r  (r) (x -a)n+ r
f (x) = Σf (a) + Σ f (a) (1.1)
r=0 r! r=0 (n + r)!
Here,

 (r) (x -a)n+ r 
(r-n ) (x -a) r 
<n -r> (x -a) r
Σ
r=0
f (a)
(n + r)!
=Σf
r=n
(a)
r!
=Σf
r=n
(a)
r!
Using this for (1.1),

<n> n -1 (x -a)r
<n -r> 
<n -r> (x -a) r
f (x) = Σf (a) +Σf (a)
r=0 r! r=n r!

<n -r> (x -a)r
= Σf (a) (1.2)
r=0 r!

Remark
(1.1) shows that
<n>
f (x) consists of the termwise integral of f(x) and the constant-of-integration-polynomial.
<n>
(1.2) shows that these constitute the Taylor series of f (x ) .

Conclusion
(1) A termwise higher integral with a fixed lower limit is collateral generally.

- 26 -
(2) It is the following case that a termwise higher integral with a fixed lower limit is lineal.

(a)=0 for r =1, 2, ,n (a) 0 ,  for at least one s  0

<r> (s)
f & f

Example 1: Termwise higher integral of f(x)= e x

<n>
Let f (x) be the lineal hegher primitive function of e x . Then,

for a -
<n>
(x)= e x , f <n -r>(a)= e a , f
(n)
f (a)= e a
Substituting these for the theorem,
n+ r
x <n> (x -a)r
n -1 
a (x -a )

a (x -a )
r
e  = Σe +Σe
a
= Σe
r=0 r! r=0 (n + r)! r=0 r!
The 1st term (constant-of-integration-polynomial) is not 0 for a - .
On the other hand, f(x)= e x is expanded to the Taylor series around a as follows.

(x -a)r 
e = Σe
x a
r=0 r!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain


x x (x -a)n+ r

e dx = Σ e
x n a
a a r=0 (n + r)!
x <n>
This termwise higher integral is a part of the Taylor series of e  . Therefore, this is a collateral haigher
integral.

-1
Example 2: Termwise higher integral of f(x)= tan x
<n>
Let f (x) be the lineal higher primitive function of tan -1x . Then, a =0 is a zero of these functions

(0)= 0 for r =0, 2, ,n -1 .

<n-r>
according to Formula 4.4.1 . That is, f
Therefore, the constant-of-integration-polynomial becomes as follows.
n -1
<n -r> (x -0)r
Σ
r=0
f (0)
r!
=0 (w1)

On the other hand, from Formula 9.2.7 (later 9.2 ),

(n) (n -1)! n /2
n+1-2k
tan x = (-1)n Σ (-1)k nＣ n+1-2k x
-1
n k =1
x +1
2

From this
n /2
(n)
f (0) = (-1)n(n -1)!Σ (-1)k nＣ n+1-2k 0n+1-2k
k =1

n +1-2k  0 k =1,  ,n /2 , f n (0) = 0 .

()
When n is even, since
When n is odd,
r+1
f (2r+1)(0) = (-1)2r+1(2r)!Σ(-1)k 2r+1Ｃ 2r+2-2k 02r+2-2k
k =1
r
= (-1)2r+1(2r)!
Σ (-1)k 2r+1Ｃ 2r+2-2k 0 + (-1)r+12r+1Ｃ 000

2r+2-2k
k =1

= (-1)r(2r)!
-1
Thus, the Taylor series around 0 of f(x)= tan x becoms as follows.

- 27 -
  (2r)! 2r+1
tan -1x = Σ(-1)r x
r=0 (2r +1)!
And we obtain the following termwise higher integral.

 tan
x x  (2r)! n+2r+ 1
-1 n
x dx = Σ(-1)
r
x (w2)
0 0 r=0 (n +2r +1)!
Substituting (w1) and (w2) for the thorem,
<n>  (2r)! n+2r+ 1
tan
-1
x = Σ(-1)r x
r=0 (n +2r +1)!
<n>
The termwise higher integral (w2) is all of Taylor series of tan x
-1
. Therefore, this is the lineal higher
integral.

4.6.2 Taylor expansion of the higher integral of log x

The higher primitive function of log x has already shown in the previous section. If this and the Taylor series
around 1 of this is compared, interesting results are obtained.

Formula 4.6.2
k 1
When a harmonic number is denoted by Hk = Σ =  (1+ k)+ , the following expressions hold.
s=1 s

 log xdx

x x  (-1)r-1(x -1)n+ r 1 n -1
n
=Σ - Σ nＣ r H n-r(x -1)
r
(2.1)
0 0 r=1 r(r +1)(r + n) n ! r=0
 1 (-1)n-1 n -1
Σ
r=1 r(r +1)(r + n)
=
n! Σ r=0
(-1)rnＣ r Hn-r (2.2)

 (-1)r-1 2n 1 n -1
Σ
r=1 r(r +1)(r + n)
=
n! 
log 2 -H n +
n! Σ
r=0
nＣ r Hn-r (2.3)

Proof
From Formula 4.3.3 , the lineal higher primitive function of f(x )=log x as follows.
r r 1

 
x
(log x)
<r>
= log x -Σ r =1, 2,  ,n
r! s =1 s
From Formula 9.2.3 (later 9.2 ), the higher derivative function of this as follows.
(r -1)!
(log x) r = (-1)r-1 r =1, 2, 3, 
()
xr
Substituting these for Thorem 4.6.1 (1.1) ,

xn
 
n 1
log x -Σ
n! s=1 s

a n-r (x -a) r  r-1 (r -1)! (x -a )

n+ r

 
n -1 n -r 1
=Σ log a -Σ +Σ(-1)
r=0 (n -r)! s=1 s r! r=0 ar (n + r) !
k 1
Substitute a =1 , H k = Σ for the right side, then
s=1 s
xn (-1)r-1(x -1)n+ r
  =Σ
n 1  1 n -1
log x -Σ - Σ nＣ r H n-r(x -1)
r
(2.1)
n! s=1 s r=1 r(r +1)(r + n) n ! r=0

- 28 -
Substitute x =0 , x =2 for this respectively, then

0n (-1)r-1(-1)n+ r
 
n 1  1 n -1
log 0 -Σ =Σ - Σ nＣ r H n-r(-1)
r
n! s=1 s r=1 r(r +1)(r + n) n r=0
!
2n (-1)r-11n+ r
  =Σ
n 1  1 n -1
log 2 -Σ - Σ nＣ r H n-r1
r
n! s=1 s r=1 r(r +1)(r + n) n ! r=0
From these, we obtain (2.2), (2.3) .

Example: n =3
1 1 1 1
+ + + +
1234 2345 3456 4567
(-1)2
        2 
3 1 1 3 1 3 1
= 1+ + - 1+ + =
3! 0 2 3 1 2 18
1 1 1 1
- + - +-
1234 2345 3456 4567

  0       
23 3 3 3
  
1 1 1 1 1 1
= log 2- 1+ + + 1+ + + 1+ +
3! 2 3 3! 2 3 1 2 2
4 8
= log 2 -
3 9

By-product
The following equations are known about (2.2) and (2.3).
 1 1
Σ
r=1 r(r +1)(r + n )
=
n n !
(2.2')

r-1


 (-1) 1 1 (1-t)n
Σ
r=1 r(r +1)(r + n)
=
n ! 0 1+ t
dt (2.3')

Therefore , the following expressions are derived from (2.2),(2.2') and (2.3),(2.3')
n-1
n -1 (-1)
Σ
r=0
(-1)rnＣ r Hn-r =
n Hr =(1+ r)+   (2.4)


1 (1- x)n n
dx = 2nlog 2- Hn+ΣnＣ r Hr Hr =(1+ r)+   (2.5)
0 1+ x r=1

(2.5) can be generalized further, and the following expression holds for p >0 .


p


1 (1- x)  p
dx = 2plog 2-  (1+p)-+Σ  (1+r)+ (2.5')
0 1+ x r=1 r

Example1

- 29 -
Example2

2012.05.01 Renewal
K. Kono
Alien's Mathematics

- 30 -
05 Termwise Higher Integral (Trigonometric, Hyperbolic)
In this chapter, for the function which second or more order integral cannot be expressed with the elementary
functions among trigonometric functions and hyperbolic functions, we integrate the series expansion of these
function termwise. Therefore, sin x, cos x, sinhx, cosh x mentioned in " 4.3 Higher Integral of Elementary
Functions " are not treated here.

5.1 Termwise Higher Integral of tan x

Since both zeros of tan x and the primitive function -log cos x are x =0 , the latter seems to be a lineal
primitive function with a fixed lower limit. Then, we assume zeros of the second or more order primitive function
are also x =0 .

5.1.0 Higher Integral of tan x

 tan x dx
x
= -log cos x
0

 tan x dx
x x
2
= non-elementary function
0 0

5.1.1 Termwise Higher Integral of Taylor Series of tan x

Formula 5.1.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers, the following expressions
hold for |x|<  /2 .
22k22k -1B2k 2k
 tan x dx
x 
=Σ x
0 k =1 2k (2k)!
22k22k -1B2k 2k+1
 tan x dx
x x 
2
=Σ x
0 0 k =1 2k (2k +1)!
22k22k -1B2k 2k+2
 tan x dx
x x x 
3
=Σ x
0 0 0 k=1 2k (2k +2)!

22k22k -1B2k 2k+ n-1
  tan x dx
x x 
n
=Σ x
0 0 k =1 2k (2k + n -1)!

Proof
tan x can be expanded to Taylor series as follows.

 22k22k -1B2k 2k-1 

tan x = Σ x |x| <
k=1 2k (2k -1)! 2
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

(1) Termwise Higher Definite Integral of Taylor Series of tan x

Substituting x = /2 for Formula 5.1.1 , we obtain the following expressions.

22k22k -1B2k 
  tan x dx
2k+1


 2
2 x
2
=Σ (1.2')
0 0 k=1 2k (2k +1)!

-1-
 
22k22k -1B2k 
  tan x dx
2 x x 2k+2

 2

3
=Σ (1.3')
0 0 0 k=1 2k (2k +2)!

22k22k -1B2k 
  tan x dx
2 2k+3

 2
x x 
4
=Σ (1.4')
0 0 0 k =1 2k (2k +3)!


22k22k -1B2k 
  tan x dx
2 2k+ n-1


 2
x x
n
=Σ (1.n')
0 0 0 k =1 2k (2k + n -1)!

(2) Taylor Series of the 1st order Integral of tan x

The1st order Integral of tan x is as follows.

 2 2 -1B2k
2k 2k


x
tan x dx = Σ x 2k = -log cos x |x|<
0 k=1 2k (2k)! 2
From this, the following expression follows immediately.

 2 2 -1B 2k
2k 2k

Σ
2k
x = -log cos x |x|< (1.t)
k =1 2k (2k)! 2
Substituting x =1 , 1/2 ,  /4 for this one by one, we obtain the following special values.

 22k22k -1B2k
Σ
k =1 2k (2k)!
= -log cos 1

 22k -1B2k 1
Σ
k =1 2k (2k)!
= -log cos
2
2 -1B2k
2k
 2k

 2

Σ
k =1 2k (2k)!
= -log cos
4
This is used in 5.3.1 later.

5.1.2 Termwise Higher Integral of Fourier Series of tan x

Formula 5.1.2
 (-1)k-1
Let (x) =Σ be a Dirichlet Eta Function and  be a ceiling function, then the following
k =1 kx
expressions hold for |x|<  /2 .

1 (1) x 0

1  (-1)k-1
 +
x
tan x dx = 0Σ sin 2k x -
0 2 k=1 k 1 2 20 0!
2 (1)

1  (-1)k-1 x1
 
x x
tan x dx 2
= 1Σ sin 2k x - + 0
0 0 2 k=1 k 2 2 2 1!
3 (1) x 2 (3) x 0

1  (-1)k-1
 
x x x
tan x dx = 2 Σ
3
3 sin 2k x - + 0 - 2
0 0 0 2 k=1 k 2 2 2! 2 0!
4 (1) x 3 (3) x 1
 
1  (-1)k-1
 +
x x
tan x dx = 3 Σ
4
4 sin 2k x - -
0 0 2 k=1 k 2 20 3! 22 1!

-2-
 
n
  tan x dx (-1)k-1
 
x x 1 
n
= Σ sin 2k x -
0 0 2n-1 k=1 k n 2
n/2
k-1 (2k -1) x
n+1-2k
+ Σ (-1)
k =1 22k-2 (n +1-2k)!

Proof
tan x can be expanded to Fourier series in a broad meaning.as follows.

1 e ix - e -ix 1- e 2ix
tan x = =i = i1- e 2ix1 - e 2ix + e 4ix - e 6ix +-
i ix -ix
e +e 2ix
1+ e
= i - i2e 2ix - 2e 4ix + 2e 6ix - 2e 8ix +-
= 2(sin 2x - sin 4x + sin 6x - sin 8x +-)
-2i(cos 2x - cos 4x + cos 6x - cos 8x +-) + i (2.0)

y
40

30

20

10

-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-10 x

-20

-30

-40
tan(x)
2*sin(2*x) - 2*sin(4*x) + 2*sin(6*x) - 2*sin(8*x)

tan x is the locus of the median (central line) of this real number part. (See the figure) However calculus of the
right side of (2.0) may be carried out, the real number does not turn into an imaginary number and the contrary
does not exist either. Therefore, limiting the range of tan x and the higher order integral to the real number, we
calculate only the real number part of the right side of (2.0) . Although the Riemann zeta (2n) is obtained
from calculation of the imaginary number part, since it is long, it is omitted in this section.
Now, integrate the real number part of both sides of (2.0) with respect to x from 0 to x , then

  2(sin 2x - sin 4x + sin 6x - sin 8x +-)dx

x x
tan x dx =
0 0
x

 
cos 2x cos 4x cos 6x cos 8x
=- - + - +-
1 2 3 4 0

 
cos 2x cos 4x cos 6x cos 8x 1 1 1 1
=- - + - +- + - + - +-
1 2 3 4 1 2 3 4
i.e.
k-1


x 1  k-1 cos 2k x  (-1)
tan x dx = - 0 Σ(-1) +Σ
0 2 k=1 k1 k=1 k1

-3-
  (-1)k-1
 

Here, let -cos x = sin x - , (x)=Σ , then we obtain
2 k=1 kx
 (1) x 0

1  (-1)k-1
 +
x
tan x dx = 0 Σ 1 sin 2k x -
0 2 k=1 k 2 20 0!
This is consistent with the real number part of the Fourier series of -log cos x .
Next, integrate both sides of this with respect to x from 0 to x , then

 (1) x 0
 tan x dx =    dx
1  (-1)k-1
 -
x x x
0Σ
2
1 sin 2kx -
0 0 0 2 k =1 k 2 20 0!
x
2 (1) x
 1! 
1  (-1)k-1 1
sin 2kx -
Σ 2 
= -
21 k=1 k1 2 0
0
k-1
2 (1) x 1

 
1  (-1)
= Σ
21 k =1
sin 2kx - +
k1 2 20 1!
Next, integrate both sides of this with respect to x from 0 to x , then

 tan x dx =  2 Σ
x
3 (1) x 2

x x x  (-1)k-1
 +
3 1
2 1
sin 2k x -
0 0 0 k=1 k 2 20 2!
0
 k-1
3 (1) x 2 

 
1 (-1) 1 cos 2k0
= Σ
22 k=1 k
1
sin 2k x -
2
+
20 2!
- 2Σ
2 k=1
(-1)k-1
k
3

 (-1)k-1 3 (1) x 2 (3) x 0

 
1
=
22
Σ
k=1 k
1
sin 2k x -
2
+
20 2!
-
22 0!
Hereafter, in a similar way we obtain the desired expressions.

(1) Termwise Higher Definite Integral of Fourier Series of tan x

Substituting x = /2 for Formula 5.1.2 , we obtain the following expressions.

(1) 
  tan x dx
1

 2
2 x
2
= (2.2')
0 0 201!

(1)  (3)   (3)
 
2 x x 2 0

 2  2
3
tan x dx = - - (2.3')
0 0 0 202! 220! 22

(1)  (3) 
  tan x dx
2 3 1

 2  2
x x
4
= - (2.4')
0 0 0 203! 221!

(1)  (3)  (5)  (5)

2 4 2

 2  2
x x
 tan x dx = 5
- + + (2.5')
0 0 0 204! 222! 240! 24


(-1)k-1 (2k -1)   (n) n

2 n+1-2k

 2
x x n /2
 tan x dx = Σ 6
+ sin (2.n')
0 0 0 k =1 22k-2(n +1-2k)! 2n-1 2

-4-
(2) Fourier Series of the1st order Integral of tan x
The1st order Integral of tan x is as follows.


x 1  k-1 cos 2k x log 2 0
tan x dx = - 0 (-1) Σ + x = -log cos x
0 2 k =1 k1 0!
From this, the following expression follows immediately.

k-1 cos 2k x 
Σ (-1) = log(2cos x) |x|< (2.f)
k =1 k1 2
Substituting x =1 , 1/2 for this one by one, we obtain the following special values.
 cos 2k
k-1
Σ (-1) = log(2cos 1)
k =1 k1
 cosk
k-1
Σ (-1) = log2cos(1/2)
k =1 k1

5.1.3 Dirichlet Odd Eta

Comparing Taylor series of tan x and Fourier series of tan x, we obtain Dirichlet Odd Eta

Formula 5.1.3
When B2k , (x),  (x) denote Bernoulli Numbers, Dirichlet Eta Function and Riemann Zeta Function
respectively, the following expressions hold.

 2 -1B 2k
2k
2
(3) = -Σ  2k+2
+ (1) -  (3)
k=1 2k (2k +2)! 2!
22k -1B2k 2k+3
 3

1 
(3) = -  (1)
 Σ
-
k=1 2k (2k +3)! 3!
22k -1B2k 2k+4  4
 2
(5) = Σ  - (1) + (3) -  (5)
k=1 2k (2k +4)! 4! 2!
22k -1B2k 2k+ 5
 5 3

1 
(5) =  (1)+ (3)
 Σ
-
k=1 2k (2k +5)! 5! 3!
2 -1B2k 2k+ 6  6
2k
 4 2
(7) = -Σ  + (1)- (3)+ (5)-  (7)
k=1 2k (2k +6)! 6! 4! 2!
22k -1B2k 2k+7  7
 5 3

1 
(7) = -  (1)+ (3)- (5)
 Σ
-
k=1 2k (2k +7)! 7! 5! 3!

 2 -1B 2k 2k+2n
2k
(2n +1) = (-1) Σ
n

k=1 2k (2k +2n )!

n -1  2n-2k
- (-1) n
Σ (-1) k
(2k +1) -  (2n +1)
k =0 (2n -2k)!
(-1)n  22k -1B2k 2k+2n +1
(2n +1) = Σ 
 k=1 2k (2k +2n +1)!

-5-
 (-1)n n -1  2n+1-2k
- Σ (-1) k
(2k +1)
 k=0 (2n +1-2k)!

Proof

(1)  (3)   (3)
   tan x dx
2 x 2 0

 2  2
x
3
= - - (2.3')
0 0 0 202! 220! 22

22k22k -1B2k 
   tan x dx
2 x 2k+2

 2
x 
3
=Σ (1.3')
0 0 0 k=1 2k (2k +2)!
from these

(1)  2
(3)  0
 (3) 22k22k -1B2k  2k+2

 2  - 2 0!  2  -  2

=Σ
202! 2
22 k =1 2k (2k +2)!

(1)  (3) 22k -1B2k 2k+2


 -(3) -
2
=Σ 
2! 22 k =1 2k (2k +2)!

22k -1B2k 2k+2  2


i.e. (3) = -Σ  + (1) -  (3)
k=1 2k (2k +2)! 2!
Next

(1)  (3) 
  tan x dx
2 3 1

 2  2
x x
4
= - (2.4')
0 0 0 203! 221!

22k22k -1B2k 

2 2k+3

 2
x x 
 tan x dx = Σ 4
(1.4')
0 0 0 k =1 2k (2k +3)!
from these

(1)  3
(3)  1 22k22k -1B2k  2k+3

 2  2


203!
-
221!
=Σ
k=1 2k (2k +3)!  2
(1) (3) 22k -1B2k 2k+3

 - 3
 =Σ 1

3! 1! k =1 2k (2k +3)!

22k -1B2k 2k+3

 3

1 
(3) = -  (1)
 Σ
i.e. -
k=1 2k (2k +3)! 3!
Hereafter, in a similar way we obtain the desired expressions.

5.1.4 Riemann Odd Zeta

2n-1
Applying (n) = (n) for Formula 5.1.3 , we obtain Riemann Zeta Function.
2n-1 - 1
Formula 5.1.4
22k -1B2k 2k+2  2
22
 

 (3) = - 3 Σ  - log 2
2 -1 k =1 2k (2k +2)! 2!

-6-
 22k -1B2k 2k+3
22 1
 3


 (3) = - 2 Σ  - log 2
2 -1  k=1 2k (2k +3)! 3!

 
24  22k -1B2k 2k+4  4  2 22-1
 (5) = 5 Σ  - log 2 +  (3)
2 -1 k =1 2k (2k +4)! 4! 2! 22

 
2 -1B2k 2k+ 5  5
2k
24 1   3 22-1
 (5) = 4 Σ  - log 2 +  (3)
2 -1  k=1 2k (2k +5)! 5! 3! 22
22k -1B2k 2k+ 6  6
26
 

 (7) = - 7 Σ  + log 2
2 -1 k =1 2k (2k +6)! 6!
 4 22-1  2 24-1
 
26
+ 7 -  (3)+  (5)
2 -1 4! 22 2! 24
22k -1B2k 2k+7
 
log 2 
7
26 1 
 (7) = - 6 Σ  -
2 -1  k=1 2k (2k +7)! 7!
 5 22-1  3 24-1
 
26 1
+ 6 -  (3) +  (5)
2 -1  5! 22 3! 24

 2 -1B 2k 2k+2n

 
2k
22n  2n
 (2n +1) = (-1) 2n+1
n
Σ  - log 2
2 -1 k=1 2k (2k +2n)! (2n)!
22n n -1  2n-2k 22k-1-1
- (-1) 2n+1 Σ(-1)
n k
 (2k +1)
2 -1 k=1 (2n -2k)! 22k-1
 2n+1
Σ
2 -1B2k

2k
(-1)n 22n 
 (2n +1) =  2k+2n+1 - log 2
 22n-1 k=1 2k (2k +2n +1) ! (2n +1) !
(-1)n 22n n -1  2n+1-2k
22k -1
-
 Σ(-1) (2n +1-2k) ! 22k (2k +1)
22n-1 k=1
k

-7-
5.2 Termwise Higher Integral of tanh x
Since both zeros of tanh x and the primitive function log cosh x are x =0 , the latter seems to be a lineal
primitive function with a fixed lower limit. Then, we assume zeros of the second or more order primitive function
are also x =0 .

5.2.0 Higher Integral of tanh x

 tanh x dx
x
= log cosh x
0

 tanh x dx
x x
2
= non-elementary function
0 0

5.2.1 Termwise Higher Integral of Taylor Series of tanh x

Formula 5.2.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers, the following expressions
hold for |x|<  /2 .
22k22k -1B2k 2k
 tanh x dx
x 
=Σ x
0 k=1 2k (2k)!
22k22k -1B2k 2k+1
 tanh x dx
x x 
2
=Σ x
0 0 k=1 2k (2k +1)!
22k22k -1B2k 2k+2
 tanh x dx
x x x 
3
=Σ x
0 0 0 k=1 2k (2k +2)!

22k22k -1B2k 2k+ n-1
  tanh x dx
x x 
n
=Σ x
0 0 k=1 2k (2k + n -1)!

Proof
tanh x can be expanded to Taylor series as follows.

 22k22k -1B2k 2k-1 

tanh x = Σ x |x| <
k =1 2k (2k -1)! 2
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

(1) Termwise Higher Definite Integral of Taylor Series of tanh x

Substituting x = 1/2 for Formula 5.2.1 , we obtain the following expressions.
1
22k22k -1B2k

2k

 
2  1
tanh x dx =Σ (1.1')
0 k=1 2k (2k)! 2
1
22k22k -1B2k
  tanh x dx
2k+1

 
2 x  1
2
=Σ (1.2')
0 0 k=1 2k (2k +1)! 2

-8-
 1
22k22k -1B2k
   tanh x dx
2 x 2k+2

 
x  1
3
=Σ (1.3')
0 0 0 k =1 2k (2k +2)! 2

1
22k22k -1B2k
  tanh x dx
2 2k+ n-1

 
x x  1
n
=Σ (1.n')
0 0 0 k=1 2k (2k + n -1)! 2

(2) Taylor Series of the 1st order Integral of tanh x

The 1st order Integral of tanh x is as follows.

 2 2 -1 B 2k
2k 2k


x
tanh xdx = Σ x 2k = log cosh x |x|<
0 k=1 2k (2k)! 2
From this, the following expression follows immediately.

 2 2 -1 B 2k
2k 2k

Σ
2k
x = log cosh x |x|< (1.t)
k =1 2k (2k)! 2
Substituting x =1 , 1/2 ,  /4 for this one by one, we obtain the following special values.

 22k22k -1B2k
Σ
k =1 2k (2k)!
= log cosh 1

 22k -1B2k 1
Σ
k =1 2k (2k)!
= log cosh
2
2 -1B2k
2k
 2k

 2

Σ
k =1 2k (2k)!
= log cosh
4
This is used in 5.4.1 later.

5.2.2 Termwise Higher Integral of Fourier Series of tanh x

Formula 5.2.2
Let (x) be a Dirichlet Eta Function, then the following expressions hold for x >0 .
-2k x
(1) x 0
 tanh x dx
x 1  k-1 e x1
= Σ(-1)
20 k=1
+ -
0 k1 1! 20 0!
-2k x
(1) x 1 (2) x 0

x x 1  k-1 e x2
tanh x dx = - 1 Σ(-1)
2
+ - + 1
0 0 2 k=1 k2 2! 20 1! 2 0!


-2k x
x x x 1  k-1 e x3
tanh x dx = 2 Σ(-1) 3
+
0 0 0 2 k=1 k3 3!
(1) x 2 (2) x 1 (3) x 0
- + 1 -
20 2! 2 1! 220! 0!


  (-1)n-1  -2k x
x x
k-1 e xn
 tanh x dx n = Σ(-1) +
0 0 2n-1 k=1 kn n!

-9-
 n -1 (k +1) x
n-1- k
- Σ(-1) k
k =0 2k (n -1- k)!

Proof
tanh x can be expanded to a Fourier series as follows. (See the following figure).

e x - e -x 1- e -2x
tanh x = x -x
= -2x
= 1- e -2x1 - e -2x + e -4x - e -6x +-
e +e 1+ e
= 1 - 2e -2x - 2e -4x + 2e -6x - 2e -8x +- (2.0)

= 1 - 2(cos 2ix - cos 4ix + cos 6ix - cos 8ix +- )

- 2i(sin 2ix - sin 4ix + sin 6ix - sin 8ix +- )

y
1.0

0.8

0.6

0.4

0.2

0.0
0 1 2 3 4 5
x
tanh(x)
1 - 2*(exp(-2*x) - exp(-20002*x))/(exp(-2*x) + 1)

Integrate both sides of (2.0) with respect to x from 0 to x , then

  1 - 2e
x x
tanh xdx = -2x
- e -4x + e -6x - e -8x +- dx
0 0
x

  
e -2x e -4x e -6x e -8x
= x+ - + - +-
1 2 3 4 0

=x +  
-2x -4x -6x -8x


e e e e 1 1 1 1
- + - +- - - + - +-
1 2 3 4 1 2 3 4
i.e.
x
(-1)k-1

x
k-1  e -2k 
tanh x dx = Σ(-1) +x -Σ
0 k =1 k1 k=1 k1
 (-1)k-1
Here, let (x)=Σ , then we obtain
k=1 kx
x

 tanh x dx = Σ(-1)
x  e -2k
k-1
+ x - (1)
0 k =1 k1
Of course, this is also the Fourier series expansion of log cosh x .
Next, integrate both sides of this with respect to x from 0 to x , then

- 10 -
  
-2k x

 tanh x dx
x x x 1  k-1 e
2
= Σ(-1) + x - (1) dx
0 0 0 20 k =1 k1
x

 
-2k x
1  k-1 e x2 x1
= - 1 Σ(-1) + - (1)
2 k=1 k2 2! 1! 0
1  k-1 e
-2k x
x 2
(1) x 1 (2) x 0
=- Σ
21 k =1
(-1) + -
20 1!
+ 1
k2 2! 2 0!
Hereafter, in a similar way we obtein the desired expressions.

(1) Termwise Higher Definite Integral of Fourier Series of tan x

Substituting x = 1/2 > 0 for Formula 5.2.2 , we obtain the following expressions.
1
(1)

e -k 1 1 0

 
2  1
k-1
tanh x dx = Σ(-1) 1
+ - 0 (2.1')
0 k =1 k 1! 2 2 1! 2
1


-k
1 2
 
2 x 1  k-1 e 1
tanh x dx = - 1 Σ(-1)
2
+
0 0 2 k=1 k 2 2! 2
(1) 1 1 (2) 0

   
1
- 0 + 1 (2.2')
2 1! 2 2 0! 2
1

 
2 x -k 3
1 
 
x 1 1
k-1 e
tanh x dx = 2 Σ(-1)
3
3
+
0 0 0 2 k=1 k 3! 2
(1) 1 2 (2) 1 1
(3) 0

    - 2 0!  
1
- 0 + 1 2 (2.3')
2 2! 2 2 1! 2 2

1

 (-1)n-1 n
2 e -k
 
x x  1 1
 tanh x dx = n
Σ(-1)
+k-1
0 0 0 2n-1 k=1 kn n! 2
n -1 (-1)k (k +1) 1 n-1- k
-Σ
k =0 2k (n -1- k)! 2   (2.n')

(2) Taylor Series of the 1st order Integral of tanh x

The 1st order Integral of tanh x is as follows.
-2k x


x 
k-1 e
tanh x dx = (-1) Σ + x - log 2 = log cosh x x >0
0 k =1 k1
From this, the following expression follows immediately.
-2k x

k-1 e
Σ (-1) 1
= log(2cosh x) - x x >0 (2.f)
k =1 k
Substituting x =1 , 1/2 ,  /2 for this one by one, we obtain the following special values.


k-1 e -2k
Σ (-1) = log(2cosh 1) - 1
k =1 k1

- 11 -
 e -k
 
 1 1
k-1
Σ (-1) = log 2cosh -
k =1 k1 2 2

e -k  
 

k-1
Σ(-1) = log 2cosh -
k =1 k1 2 2

5.2.3 Dirichlet Eta Function

Comparing Taylor series of tanh x and Fourier series of tanh x, we obtain Dirichlet Eta Function.

Formula 5.2.3
When B2k , (x) denote Bernoulli Numbers and Dirichlet Eta Function, the following expressions hold.


k-1 -k 22k -1B2k 1 1
0 = Σ(-1) e +Σ -
k=1 k=1 2k (2k -1)! 2 0!
 2 -1 B 2k
2k
 e -k 1 1
(1) = Σ(-1) k-1
1
-Σ +
k=1 k k=1 2k (2k)! 2 1!
 2 -1 B 2k
2k
 e -k 1 1 1
(2) = Σ(-1) k-1
2
+Σ - + (1)
k=1 k k =1 2k (2k +1)! 2 2! 1!
 2 -1 B 2k
2k
 e -k 1 1 1 1
(3) = Σ(-1) k-1
-Σ + - (1)+ (2)
k=1 k3 k =1 2k (2k +2)! 2 3! 2! 1!
 2 -1 B 2k
2k
 e -k 1 1
(4) = Σ(-1)k-1 4
+Σ -
k=1 k k =1 2k (2k +3)! 2 4!
1 1 1
+ (1) - (2) + (3)
3! 2! 1!

 e -k   22k
-1B2k (-1)n 1
(n) = Σ(-1) k-1
+ (-1) Σ
n
-
k=1 kn k=1 2k (2k + n -1)! 2 n!
n -1 (-1)k
-Σ (n -k)
k=1 k!

Proof
1
tanh = 1 - 2e -1 - e -2 + e -3 - e -4 +-  (2.0')
2
22k22k -1B2k 2k-1

 
1  1
tanh =Σ (1.0')
2 k=1 2k (2k -1)! 2
 22k -1B2k
 1
 0 = Σ(-1) k-1 -k
e +Σ -
k =1 k =1 2k (2k -1)! 2
1


2  e -k 1
tanh x dx = Σ(-1) k-1
+ - (1) (2.1')
0 k =1 k1 2

- 12 -
 1
22k22k -1B2k

2k

 
2  1
tanh x dx =Σ (1.1')
0 k =1 2k (2k)! 2
 2 -1 B 2k
2k
 e -k 1
 (1) = Σ(-1) k-1
-Σ +
k=1 k1 k=1 2k (2k)! 2
1


-k
1 2
 
2 x 1  k-1 e 1
tanh x dx = - 1 Σ(-1)
2
+
0 0 2 k=1 k 2 2! 2
(1) 1 1 (2) 0

   
1
- 0 + 1 (2.2')
2 1! 2 2 0! 2
1
22k22k -1B2k
  tanh x dx
2k+1


 
2 x 1
2
=Σ (1.2')
0 0 k=1 2k (2k +1)! 2
 2 -1 B 2k
2k
 e -k 1 1 1
 (2) = Σ(-1) k-1
- + (1)+Σ
k=1 k2 2 2! 1! k =1 2k (2k +1)!

Hereafter, in a similar way we obtain the desired expressions..

5.2.4 Riemann Zeta Function

2n-1
Applying (n) = (n) for Formula 5.2.3 , we obtain Riemann Zeta Function.
2n-1 - 1

Formula 5.2.4

Σ(-1)
 2 -1 B 2k


2k
21  e -k 1 log 2
 (2) = 1 k-1
+ Σ - +
2 -1 k=1 k2 k=1 2k (2k +1)! 22! 1!
 2 -1 B 2k
2k


22  e -k 1 log 2
 (3) = 2 Σ (-1) k-1
- Σ + -
2 -1 k=1 k3 k=1 2k (2k +2)! 23! 2!
2 -1  (2)
1
+
21 1! 
 2 -1 B 2k
2k


23  e -k 1 log 2
 (4) = 3 Σ (-1) k-1
4
+Σ - +
2 -1 k=1 k k=1 2k (2k +3)! 24! 3!
22-1  (3) 21-1  (2)
+
22 1!
-
21 2! 

-1  (n -j)
Σ(-1) 
2n-1  e -k n -2
j-1 2
n-1- j
 (n) = k-1
+ Σ (-1)
2n-1-1 k=1 kn j=1 2n-1- j j!

2 -1  
(-2) n-1
 22k -1B2k 1 log 2
- Σn-1 k =1
-
2k (2k + n -1)! 2n !
+
(n -1)!

- 13 -
5.3 Termwise Higher Integral of cot x
Since both zeros of cot x and the primitive function log sin x are x = /2 , the latter seems to be a lineal
primitive function with a fixed lower limit. Then, we assume zeros of the second or more order primitive function
are also x = /2 .

5.3.0 Higher Integral of cot x

 cot x dx
x
= log sin x

2

 cotx dx
x x
2
= non-elementary function
 
2 2

5.3.1 Termwise Higher Integral of Taylor Series of cot x

Formula 5.3.1
(x) on 0< x < 
<n>
When B2k denote Bernoulli Numbers and functions f are as follows.

xn 22kB2k 2k+n
  -Σ
n 1 
n =1, 2, 3, 
<n>
f (x)= log x -Σ x
n! k =1 k k=1 2k (2k +n)!
the following expressions hold .

22kB2k 2k
 cot x dx
x 
= f<0>(x) = log x -Σ x
 k=1 2k (2k)!
2

 
  cot x dx
0

   2
x x 1
2 <1>
=f (x) - x- f<1>
  0! 2
2 2

    
x x x 0 1

   2    2
1 1
cot x dx 3 = f <2>(x) - x- f
<2>
- x- f
<1>
   0! 2 1! 2
2 2 2

     
x x 0 1

   2   f  2
1 1
 4
cot x dx = f <3>(x) - 0! x - 2 f
<3>
- x-
<2>
  1! 2
2 2
 2

   2
1 <1>
- x- f
2! 2

 
  cot x dx
x x k

   2  n =2, 3, 4, 
<n -1>
n -2 1 n -1-k >
n
=f (x) -Σ x- f<
  k=0 k! 2
2 2
  
where f   = 0 , f   = - log 2
<0> <1>
2 2 2

Proof
22k B2k 2k 
2k
 2 B 2k
x cot x = 1+Σ(-1) x = 1+Σ k
x 2k
k =1 (2k)! k=1 (2k)!

- 14 -
From this
2k
1  2 B2k 2k-1
cot x = -Σ x
x k =1 2k (2k -1)!
Integrate the both sides with respect to x from  /2 to x, then

 
22kB2k 2k 
2k

 cot x dx = log x -Σ
 2 B 2k
2k

 2
x 

2 Σ
x - log -
 k=1 2k (2k)! k=1 2k (2k)!
2
Here

 22kB2k  2k

 2

log -Σ =0
2 k =1 2k (2k)!
Because, the following equation is known ( 岩波数学公式Ⅱ p151)
2k
 2 B 2k
log x -Σ 2k
x = log sin x 0< x <
k =1 2k (2k)!
Then, substituting x = /2 for this, we obtain the above.
Thus,

22kB2k 2k
 cot x dx = log x -Σ
x 
x = f<0>(x)
 k=1 2k (2k)!
2
Next, integrate the both sides with respect to x from  /2 to x, then

 f  2
x x x
cot x dx 2 = <0>
(x)dx = f<1>(x) - f<1>
  
2 2 2
Furthermore, integrate the both sides with respect to x from  /2 to x , then

  f
x

 2  dx
x x x
cot x dx 3 = <1>
(x) - f<1>
   
2 2 2 2
  1

 2   f  2
<2> <2> 1 <1>
=f (x) - f - x-
1! 2
Hereafter, in a similar way we obtein the desired expressions.

Although Formula 5.3.1 is a general expression anyway, if it is expand actually, it is very complicated.
Then, this was devised and was made the same easy expression as tan x. It is as follows.

Formula 5.3.1'
When B2k denote Bernoulli Numbers, the following expressions hold for 0< x <  .
 2 2 -1B 2k
2k 2k
 2k+ n-1
 
x

 
x
 cotx dx = -Σn
x-
  k =1 2k (2k + n -1)! 2
2 2

Proof

cot x = -tan x -  2 
- 15 -
  22k22k -1B2k 2k-1 
tan x = Σ x |x|<
k =1 2k (2k -1)! 2
From these
22k22k -1B2k  2k-1

 

cotx = -Σ x- 0< x < 
k =1 2k (2k -1)! 2
Integratingboth sides of this with respect to x from  /2 to x repeatedly, we obtain the desired expressions.

(1) Termwise Higher Definite Integral of Taylor Series of cot x

When
xn 22kB2k 2k+n
  -Σ
n 1 
n =1, 2, 3, 
<n>
f (x)= log x -Σ x
n! k =1 k k=1 2k (2k +n)!
substituting x =0 for Formula 5.3.1 , we obtain the following expressions.

 

0

 2  2
0 x 1
2
cot x dx = - f<1> (1.2')
  0!
2 2
   
 
0 x 0 1

 2  2  2  2
x 1 1
3 <2>
cot x dx = - f + f<1> (1.3')
   0! 1!
2 2 2
 0
  1

   2  2  2  2
0 x x 1 1

<3>
cot x dx = -
4
f + f<2>
   0! 1!
 
2 2 2 2

 2 f  2
1 <1>
- (1.4')
2!

 
   cot x dx (-1)k+1
0 x k

 2 f  2
x n -2
<n -1-k >
n
=Σ (1.n')
   k=0 k!
2 2 2

(2) Taylor Series of the 1st order Integral of cot x

The 1st order Integral of cot x is as follows.
2k

 cot x dx
x  2 B 2k
= log x -Σ x 2k = log sin x 0< x < 
 k =1 2k (2k)!
2
From this, the following expression follows immediately.
2k
 2 B 2k
Σ x 2k
= -log sin x + log x 0< x <  (1.t)
k =1 2k (2k)!
Substituting x =1 , 1/2 ,  /4 for this one by one, we obtain the following special values.
2k
 2 B2k
Σ
k =1 2k (2k)!
= -log sin 1
 B2k 1
Σ
k =1 2k (2k)!
= -log sin
2
- log 2

B2k  2k
 
 2

Σ
k =1 2k (2k)!
= -log sin
4
+ log
4

- 16 -
 Moreover, from the last expressions of this and 5.1.1 (2) , the next expression follows

 2 -2B 2k
2k
 2k 
Σ
k =1 2k (2k)!  2
= -log This is used in 5.5.1 later.
4

5.3.2 Termwise Higher Integral of Fourier Series of cot x

Formula 5.3.2
Let (x) be Dirichlet Eta Function and  be the ceiling function, then the following expressions hold

for 0< x <  .

1 (1) 
 cot x dx
0

  - 2 0! x - 2 
x 1  1
= 0 Σ 1 sin 2kx - 0
 2 k=1 k 2
2
2 (1) 
  cot x dx
1

  - 2 1! x - 2 
x x 1  1
2
= 1 Σ 2 sin 2kx - 0
  2 k=1 k 2
2 2

 cot x dx 3 (1)  (3) 

x x x  2 0

 - 202! x - 2  + 220! x - 2 
1 1
Σ 
3
= sin 2kx -
   22 k =1 k 3 2
2 2 2

 4 (1)  (3) 
x x  3 1

   
1 1
 Σ  
4
cot x dx = sin 2kx - - x- + x-
  23 k =1 k
4 2 203! 2 221! 2
2 2


n
  cot x dx  
x x 1  1
n-1 Σ
n
= n sin
2kx -
  2 k =1 k 2
2 2

(-1)k (2k -1)  n+1-2k

 
n/2
+Σ 2k-2 (n +1-2k)!
x-
k =1 2 2

Proof
cot x can be expanded to Fourier series in a broad meaning.as follows.

e ix + e -ix 1+ e 2ix
cotx = i ix -ix
= -i 2ix
= -i1 + e 2ix1 + e 2ix + e 4ix + e 6ix +
e -e 1- e
= -i - i2e 2ix + 2e 4ix + 2e 6ix + 2e 8ix +
= 2(sin 2x + sin 4x + sin 6x + sin 8x +)
-2i(cos 2x + cos 4x + cos 6x + cos 8x +) - i (2.0)

cot x is the locus of the median (central line) of this real number part. (See the figure of the following page.)
However calculus of the right side of (2.0) may be carried out, the real number does not turn into an imaginary
number and the contrary does not exist either. Therefore, limiting the range of cot x and the higher order
integral to the real number, we calculate only the real number part of the right side of (2.0).
Although the Riemann zeta (2n) is obtained from the calculation of the imaginary number part, since it is
long, it is omitted in this section.

- 17 -
 y
40

30

20

10

0
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
-10 x

-20

-30

-40
cot(x)
2*sin(2*x) + 2*sin(4*x) + 2*sin(6*x) + 2*sin(8*x)

Now, integrate the real number part of both sides of (2.0) with respect to x from  /2 to x , then

 cot x dx =  2(sin 2x + sin 4x + sin 6x * sin 8x +)dx

x x

 
2 2
x

 
cos 2x cos 4x cos 6x cos 8x
=- + + + +
1 2 3 4 
2

   
cos 2x cos 4x cos 6x cos 8x 1 1 1 1
=- + + + + - - + - +-
1 2 3 4 1 2 3 4
i.e.
k-1


x 1  cos 2k x  (-1)
cot x dx = log sin x = - 0 Σ -Σ
 2 k=1 k 1 k=1 k1
2

 (-1)k-1
 

Here, let -cos x = sin x - , (x)=Σ , then we obtain
2 k=1 kx
1 (1) 

0

  - 2 0! x - 2 
x 1  1
cot x dx = 0 Σ 1 sin 2kx - 0
 2 k =1 k 2
2
This is consistent with the real number part of the Fourier series of log sin x .
Next, integrate both sides of this with respect to x from  /2 to x , then
1 (1) 
  cot x dx =    - 2 0! x - 2  dx
0


x x x 1  1
0 Σ 1 sin
2
2kx - 0
   2 k=1 k 2
2 2 2
x
2 (1) 
  - 2 1! x - 2  
1


1  1
1 Σ 2 sin
= 2kx - 0
2 k=1 k 2

2
2 (1)  1

  - 2 1! x - 2 
1  1
= 1 Σ 2 sin 2kx - 0
2 k =1 k 2

- 18 -
 Next, integrate both sides of this with respectto x from  /2 to x , then

 cot x dx =   2 2 (1) 

dx
x x  1

 -
x x
x - 2 
1 1
Σ
3
1 2
sin 2kx -
    k =1 k 2 201!
2 2 2 2
x
3 (1) 
2 Σ 
 2

 -
1
x - 2 
1
= 2 3
sin 2kx -
k =1 k 2 202! 
2
 3 (1)  2  cos 2k
 
1 1
 
1
=
22 Σ
k=1 k
3
sin 2kx -
2
-
2 2!0
x-
2
-
22 Σ
k=1 k
3

 3 (1)  2
(3)  0

 
1
   
1
=
22
Σ
k=1 k
3
sin 2kx -
2
-
202!
x-
2
+
222!
x-
2
Hereafter, in a similar way we obtain the desired expressions.

(1) Termwise Higher Definite Integral of Fourier Series of cot x

Substituting x =0 for Formula 5.3.2 , we obtain the following expressions.

(1) 
  cot x dx
1

 2
0 x
2
= (2.2')
  201!
2 2

(1)  (3)   (3)

   cot x dx
0 2 0

 2  2
x x
3
=- 0
+ 2
+ (2.3')
   2 2! 2 0! 22
2 2 2

(1)  (3) 
   cot x dx
0 x 3 1

 2  2
x
4
= - (2.4')
   203! 221!
2 2 2

(1)  (3)  (5)  (5)

   cot x dx
0 x 4 2

 2  2
x
5
=- 0
+ 2
- 4
- (2.5')
   2 4! 2 2! 2 0! 24
2 2 2

(-1)k-1 (2k -1)   (n) n
   cot x dx
n+1-2k

(n +1-2k) !  2 
0 x x n /2
n
  
= (-1) n
Σ
k=1 22k-2
-
2 n-1
sin
2
2 2 2
(2.n')

(2) Fourier Series of the 1st order Integral of cot x

The 1st order Integral of cot x is as follows.

 cotx dx = - 2
x 1  cos 2k x
0 Σ - log 2 = log sin x 0< x < 
0 k =1 k1
From this, the following expression follows immediately.
 cos 2k x
Σ = -log(2sin x) 0< x <  (2.f)
k =1 k1
Substituting x =1 , 1/2 for this one by one, we obtain the following special values.
 cos 2k
Σ = -log(2sin 1)
k =1 k1

- 19 -
  cosk
Σ = -log2sin(1/2)
k =1 k1

5.3.3 Dirichlet Odd Eta & Riemann Odd Zeta

Comparing Taylor series of cot x and Fourier series of cot x, we obtain Dirichlet Odd Eta and Riemann
Odd Zeta.

Formula 5.3.3
When B2k , (x),  (x) denote Bernoulli Numbers, Dirichlet Eta Function and Riemann Zeta Function
respectively, the following expressions hold.
 B2k
log = 1 +Σ  2k
2k (2k +1)!
k=1

2 B2k
 
2 1 
(3) = - log -Σ +Σ  2k+2 -  (3)
2! k =1 k k=1 2k (2k +2)!

1 
 
B2k

3


3 1 
 (3) = - log -Σ -Σ  2k+3
 3! k=1 k k =1 2k (2k +3 )!
 4
B2k 2
 
4 1 
(5) = log -Σ -Σ  2k+4
+  (3) -  (5)
4! k =1 k k=1 2k (2k +4)! 2!
1 5 3
 
B2k
 
5 1 
 (5) = log -Σ -Σ  2k+5
+  (3)
 5! k=1 k k =1 2k (2k +5)! 3!
6  B2k 4 2
log -Σ  +Σ 2k (2k +6)! 
6 1
(7)= - 2k+6
- (3)+ (5) - (7)
6! k=1 k k=1 4! 2!
7 5 3
 7! log -Σ  -Σ 2k (2k +7)!  
1 7 1  B2k
(7)= - 2k+7
+ (3)- (5)
 k =1 k k =1 5! 3!

 2n
 (2n)!  
 B2k

2n 1
(2n +1) = (-1) n
log -Σ -Σ  2k+2n
k =1 k k =1 2k (2k +2n ) !
n -1  2n-2k
+ (-1)nΣ(-1)k-1 (2k +1) - (2n +1)
k=1 (2n -2k) !
 2n+1
  (2n +1) !  
 B2k

1 2n +1 1
(2n +1) = (-1) n
log  - Σ -Σ  2k+2n+1
k=1 k k=1 2k (2k +2n +1) !
1 n -1  2n-2k+1
+ (-1) n
Σ (-1) k-1
(2k +1)
 k=1 (2n -2k +1) !

Proof
From Fourier series (2.2') and Taylor series (1.2')

log 2  1  0 <1> 
     2
1
0
= - f (2.2")
2 1! 2 0! 2
 1
 22kB2k  2k+1

 2  log 2 -1 -Σ  2
1 
=-
1! k =1 2k (2k +1)!

- 20 -
Hence we obtain
 B2k
log = 1 + Σ  2k
k =1 2k (2k +1)!
Next, substitute (2.2") for Taylor series (1.3') , then

  
   cot x dx
0 1

 2  2  2
x x 1
3 <2>
= -f + f<1>
   1!
2 2 2
  2

 2  2
<2> log 2
= -f -
1!
From this and Fourier series (2.3') , we obtain
2
 1 (3)   2

 2    2
1 log 2 log 2
 3 -
2 ( )
+ 2 = -f<2> -
2 2! 2 0! 2 1!
1 (3) <2>   2

   2
1 log 2
i.e.  (3) + = -f - (a)
22 22 0! 2 2!
From (a), it is as follows.

 2
(3) = -2 f  2 - log 2 -  (3)
2 <2>
2!

  2  log 2 -Σ  -Σ 
 2
 22kB2k  2k+2

 2
1 2 1 
2
= -2
2! k=1 k k=1 2k (2k +2)!
 2
- log 2 -  (3)
2!
2 B2k
 
2 1 
=- log -Σ +Σ  2k+2 -  (3)
2! k =1 k k=1 2k k
(2 +2)!
Next, substitute (2.2") and (a) for Taylor series (1.4') one by one, then

   cot x dx <3>   1 <2>   

0 x x 2

 2 + 1!  2 
f    2 f  2
4 1 1 <1>
= -f -
   2 2!
2 2 2
  1
1 (3)
 2  2  
1
= -f <3>
-  (3) +
22 22 0!
From this and Fourier series (2.4') , we obtain

 3
1 (3)  1
  1
1 (3)
 2 - 2  2 
log 2
 2  2
1
- 2 = -f
<3>
2
(3)+
3! 2 1! 22 0!
 1
  3

 2  2  2
1 log 2
i.e.  (3) = -f<3> - (b)
22 3!

Substitute f<3>  2 for (b), then

 1
  3

 2   (3) = -f  2   2
1 <3> log 2
-
22 3!
 3
  22kB2k  2k+3
log 2  3

 2    2  2
1 3 1
=- log -Σ +Σ -
3! 2 k =1 k k =1 2k (2k +3) ! 3!

- 21 -
  3 22kB2k  2k+3

 2  log -Σ  +Σ  2
1 3 1 
=-
3! k =1 k k=1 2k (2k +3)!
From this, we obtain

  2  
 -1
 3 22kB2k  2k+3

 2   2
1 1 
3
 (3) = -2 2
log -Σ +Σ
3! k=1 k k=1 2k (2k +3)!

3
 3! 
B2k
 
1 3 1 
=- log -Σ -Σ  2k+3
 k =1 k k=1 2k k(2 +3)!
Hereafter, in a similar way we obtein the desired expressions.

5.3.4 Another set of Riemann Odd Zeta

22n - 1
Substituting (2n +1) = 2n  (2n +1) for Formula 5.3.3 , we obtain another set of Riemann
2
Odd Zeta.

Formula 5.3.4
2
 2!   2k (2k +2)!  
22 2 1  B2k
( )
3 = - log -Σ -Σ 2k+2
23-1 k =1 k k=1

 
 4! 
B 
 (3)
24 4 2

k
1 4 
 (5) =  
2k 2k+4
25-1
log - Σ - Σ 2k (2k +4)!
k =1
+
2! k =1


 6!  
6
26 B 
k
1 6 
 (7)= - 7  
2k 2k+6
2 -1
log -Σ -Σ 2k (2k +6)!
k=1 k =1

 
 4!  (5)
4 2
26
- 7  (3) -
2 -1 2!

 2n
-1  (2n) !  
22n  B2k

2n 1
(2n +1) = (-1) n
2n+1
log -Σ -Σ  2k+2n
2 k=1 k k=1 2k (2k +2n ) !
22n n -1  2n-2k
+ (-1)n Σ(-1)k-1 (2k +1) n =2, 3, 4, 
22n+1 -1 k=1 (2n -2k) !

- 22 -
5.4 Termwise Higher Integral of coth x
Since the zero of the first order integral coth x is x = sinh -11 = 0.8813 ,
log sinh x of
we assume zeros of the second or more order integral of coth x are also x =0.8813 .
Where, since the zero of coth x is not 0.8813 , these seem to be a collateral higher integral.

5.4.0 Collateral Higher Integral of coth x

When  = 2sinh -11 = 1.762747174

 coth x dx
x
= log sinh x

2

 coth x dx
x x
2
= non-elementary function
 
2 2

5.4.1 Termwise Higher Integral of Taylor Series of coth x

Formula 5.4.1
Let  = 2sinh -11 (= 1.762747174) and B0=1, B2=1/6, B4=-1/30, B6=1/42, 
<n>
are Bernoulli Numbers and functions (x) on x >0 be as follows.
f
xn 22k B2k
 
n 1 
n =1, 2, 3, 
<n>
f (x)= log x -Σ +Σ x 2k+n
n! k =1 k k=1 2k (2 k +n ) !
Then the following expressions hold for x >0 .

22k B2k 2k
 coth x dx
x 
=f <0>
(x) = log x +Σ x
 k=1 2k (2k)!
2
 
  coth x dx
0

   2
x x 1
2 <1>
=f (x) - x- f<1>
  0! 2
2 2

    
x x x 0 1

   2    2
3 1 1
coth x dx = f <2>(x) - x- f
<2>
- x- f
<1>
   0! 2 1! 2
2 2 2

  coth x dx    
x x 0 1

  f  2   f  2
4 <3> 1 <3> 1 <2>
=f (x) - x- - x-
  0! 2 1! 2
2 2
 2

   2
1 <1>
- x- f
2! 2

 
  coth x dx
x x k

 f  2
<n -1>
n -2 1 <n -1-k >
n
=f (x) -Σ x-
  k=0 k! 2
2 2

Proof
 22k B2k 2k
x coth x = 1+Σ x
k =1 (2k)!

- 23 -
From this

1  22k B2k
coth x = +Σ x 2k-1
x k=1 2k (2k -1)!
Integrate the both sides with respect to x from  /2 to x , then

 
22k B2k 2k  22k B2k 
 coth x dx = log x +Σ
2k

 2
x  

2 Σ
x - log +
 k=1 2k (2k)! k=1 2k (2k)!
2
Here

 22k B2k  2k

 2

log +Σ =0
2 k =1 2k (2k)!
Because, the following equation is known ( 岩波数学公式Ⅱ p150)
2k-1
 2 B2k 2k
log x +Σ x = log sinh x x >0
k=1 k (2k)!
Then, substituting x = /2 for this, we obtain the above.
Thus,

22k B2k 2k
 coth x dx = log x +Σ
x 
x = f<0>(x)
 k=1 2k (2k)!
2
Next, integrate the both sides with respect to x from  /2 to x , then

   2
x x x
coth x dx 2 = f<0>(x)dx = f<1>(x) - f<1>
  
2 2 2
Furthermore, integrate the both sides with respect to x from  /2 to x , then


  f
x

 2  dx
x x x
coth x dx 3 = <1>
(x) - f<1>
   
2 2 2 2

  1

 2   f  2
<2> <2> 1 <1>
=f (x) - f - x-
1! 2
Hereafter, in a similar way we obtein the desired expressions.

(1) Termwise Higher Definite Integral of Taylor Series of coth x

When
xn 22k B2k
 
n 1 
n =1, 2, 3, 
<n>
f (x)= log x -Σ +Σ x 2k+n
n! k =1 k k=1 2k (2k +n )!

substituting x =0 for Formula 5.4.1 , we obtain the following expressions.

 

0

 2  2
0 x 1
2
coth x dx =- f<1> (1.2')
  0!
2 2

   
 
0 0 1

 2  2  2  2
x x 1 1
3 <2>
coth x dx = - f + f<1> (1.3')
   0! 1!
2 2 2

- 24 -
    
 
0 1

 2 f  2  2 f  2
0 x x 1 1
 coth x dx = - 4 <3> <2>
+
   0! 1!
2 2 2

 2

 2  2
1
- f<1> (1.4')
2!

 
  (-1)k
0 x k

 2  2
x n -2
 coth x dx = -Σ
n n -1-k >
f< (1.n')
   k=0 k!
2 2 2

(2) Taylor Series of the 1st order Integral of coth x

The 1st order Integral of coth x is as follows.
22k B2k
 coth x dx = log x +Σ 2k (2k)! x
x 
2k
= log sinh x 0< x < 
 k=1
2
From this, the following expression follows immediately.

 22k B2k 2k
Σ x = log sinh x - log x 0< x <  (1.t)
k =1 2k (2k)!

Substituting x =1 , 1/2 ,  /4 for this one by one, we obtain the following special values.

 22k B2k
Σ
k =1 2k (2k)!
= log sinh 1

 B2k 1
Σ
k =1 2k (2k)!
= log sinh
2
+ log 2

B2k  2k
 
 2

Σ
k =1 2k (2k)!
= log sinh
4
- log
4
Moreover, from the last expressions of this and 5.2.1 (2) , the next expression follows

 2 -2 B 2k
2k
 2k  
Σ
k =1 2k (2k)!  2 = log coth
4
+ log
4

5.4.2 Termwise Higher Integral of Fourier Series of coth x

Formula 5.4.2
Let  = 2sinh -11 = 1.762747174, then the following expressions hold for x >0 .

1  e -2k x  
  +
x
coth x dx = log sinh x = - 0 Σ 1 + x - - log 2
 2 k=1 k 2 2
2
1  e -2k x   /2-log 2 
  coth x dx
2

   
x x 1
2
= 1Σ 2 + x- + x-
  2 k=1 k 2! 2 1! 2
2 2
1  e -k 
- 1Σ 2
2 k=1 k

- 25 -
   /2-log 2 

x 1  e -2k x 1 3 2

 + x - 2 
x x
coth x dx = - 2 Σ 3 +
3
x-
   2 k =1 k 3! 2 2!
2 2 2

1  e -k   1  e -k 
- 1 Σ 2 x-
2 k=1 k 2   + 2Σ 3
2 k=1 k
  /2-log 2 
 
x 1  e -2k x 1 4 3

   
x
 coth x dx = 3 Σ 4 + 4
x- + x-
  2 k=1 k 4! 2 3! 2
2 2

1  e -k  1  2
1  e -k   1  e -k 
- Σ
21 k=1 k 2 2!
x -
2   + Σ
22 k=1 k 3
x -
2   - Σ
23 k=1 k 4

  /2-log 2 
   coth x dx
x n-1
x (-1)n e -2k n

 + x - 2 
x  1
n
= Σ + x -
  2n-1 k =1 kn n! 2 (n -1)!
2 2

(-1)n-j 1  j 
e -k
 Σ k
n -2
-Σ x -
j=0 2n-1- j j ! 2 k=1
n-j

Proof
coth x can be expanded to a Fourier series as follows. (See the following figure).

e x + e -x 1+ e -2x
coth x = x -x
= -2x
= 1+ e -2x1 + e -2x + e -4x + e -6x +
e -e 1- e
= 1 + 2e -2x + 2e -4x + 2e -6x + 2e -8x +- (2.0)

= 1 + 2(cos 2ix + cos 4ix + cos 6ix + cos 8ix + )

+ 2i(sin 2ix + sin 4ix + sin 6ix + sin 8ix + )

y
60

50

40

30

20

10

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7
x
coth(x)
2*exp(-2*x) + 2*exp(-4*x) + 2*exp(-6*x) + 2*exp(-8

Now, let  /2 =sinh -11 (= 0.881373587) . Integrate the both sides of (2.0) with respect to x
from  /2 to x , then

- 26 -
   2e
x x
coth x dx = -2x
+ e -4x+ e -6x+ e -8x+ + 1dx
 
2 2
x

 
e -2x e -4x e -6x e -8x
=-
1
+
2
+
3
+
4
+ - x -
2  
2

 
-2x -4x -6x -8x


e e e e
=- + + + + - x -
1 2 3 4 2
   
e- e -2 e -3 e -4
+ + + + +
1 2 3 4
i.e.

1  e -2k x 

e -k
  +Σ
x 
coth x dx = - 0 Σ 1 + x -
 2 k=1 k 2 k=1 k1
2
Here, the following equation holds,
-k 
 e 
Σ = - log 2 (w)
k =1 k1 2
Because,
 xk
log(1-x) = -Σ -1 x < 1
k =1 k
From this,

 e -2kx
Σ = -log1-e -2x x >0
k =1 k
On the other hand,

e x- e -x e x1- e -2x
1 = sech x = =
2 2
From this,

1 ex 1 ex
=  log = log  -log1-e -2x = x - log 2
1- e -2x 2 1- e -2x 2
Then,

 e -2kx
Σ = -log1-e -2x = x - log 2
k =1 k

Substituting x = sech -11 = for this, we obtain (w).
2
Thus, using (w), we obtain

1  e -2k x  
  
x
coth x dx = log sinh x = - 0 Σ 1 + x - + - log 2
 2 k=1 k 2 2
2
This is consistent with the real number part of the Fourier series of log sinh x .

Next, integrate both sides of this with respect to x from  /2 to x , then

- 27 -
  
 -Σ  dx
x

  coth x dx
e -2k
  2
x x x 
2
= 1
+ x- + - log 2
   k=1 k 2
2 2 2
x
  /2-log 2 
 x - 2 
1  e -2k x 2

 
1
= 1 Σ
2 k =1 k 2
+
2!
x-
2
+
1!

2

1  e -2k x  2
 /2-log 2  1  e -k 
 + x - 2 
1
= 1Σ 2 + x- - 1Σ 2
2 k =1 k 2! 2 1! 2 k=1 k
Hereafter, in a similar way we obtein the desired expressions.

(1) Termwise Higher Definite Integral of Fourier Series of coth x

Substituting x =0 for Formula 5.4.2 , we obtain the following expressions.

 (2)   1  e -s
  coth x dx
2

 2  2
0 x 1 log 2
2
= - + - 1 Σ 2
  21 2! 1! 2 0! s=1 s
2 2
(2.2')
 (3)  3 log 2  2
 
0

   
x x 2
3
coth x dx = - 2 + -
   2 3! 2 2! 2
  e -s 1  e -s
2 2 2

 
1
+ 1 Σ
2 1! 2 s=1 s 2
+ 2 Σ 3
2 0! s=1 s
(2.3')

 (4)  log 2  3
 
0 x 4

   
x 3
 coth x dx 4
= - +
   23 4! 2 3! 2
2 2 2
  
 2
e -s  e -s 1  e -s
 2 Σ  2 Σ
1  1 
- 1 2
- 2 - Σ (2.4')
2 2! s=1 s 2 1! s=1 s 3 230! s=1 s 4


 (n)  
   coth x dx 2 
0 x n n-1

 2  2
x n -1 log 2
n n
= (-1) n-1
- +
   n! (n -1)!
2 2 2

 r 
e -s
 2 Σ s
n -2 1
-(-1)n Σ (2.n')
j=0 2n-1- r r! s=1
n-r

(2) Fourier Series of the 1st order Integral of coth x

The 1st order Integral of coth x is as follows.
-2k x
 
  2
x 1  e

coth x dx = - Σ
20 k=1 k 1
+ x- +
2
- log 2 = log sinh x x >0
2
From this, the following expression follows immediately.
-2k x
 e
Σ = x - log(2sinh x) x >0 (2.f)
k =1 k1

- 28 -
Substituting x =1 , 1/2 ,  /2 for this one by one, we obtain the following special values.

 e -2k
Σ = 1 - log(2 sinh 1)
k =1 k1
e -k
 
 1 1
Σ = - log 2 sinh
k =1 k1 2 2

e -k  
 

Σ = - log 2 sinh
k =1 k1 2 2

5.4.3 Riemann Zeta Function

Comparing Taylor series of coth x and Fourier series of coth x, we obtain Riemann Zeta Function.

Formula 5.4.3
When  = 2sinh -11 = 1.762747174 and B2k are Bernoulli Numbers, the following
expressions hold.

B2k 
 1! log -Σ  +Σ 
 2k+1
 e -k  1 1  1 2
 (2) = Σ - -
k =1 k2 k =1 k k=1 2k (2k +1)! 2 2!
 B2k 
2k+2
e -k 2 1 3
  +Σ
 2 1 
 (3) = Σ + log  -Σ -
k =1 k3 2! k=1 k k=1 2k (2k +2)! 2 3!

+  (2)
1!
B2k 
 3! log -Σ  +Σ 
 2k+3
 e -k 3 3 1  1 4
 (4) = Σ - -
k =1 k4 k =1 k k =1 2k (2k +3)! 2 4!
 2
+  (3) -  (2)
1! 2!


j-1 
j
 e -k n -2
 (n) = Σ + Σ(-1)  (n - j)
k =1 kn j=1 j!
B2k 
 (n -1)! log -Σ  +Σ 
2k+ n-1
n-1  n-1 n -1 1  1 n
+ (-1) -
k=1 k k=1 2k (2k + n -1)! 2 n!
Proof
From Fourier series (2.2') and Taylor series (1.2')

  2
 1  e -k 
 2  2  2
1 1 log 2
f <1>
=-  (2) + - + Σ (a)
21 2! 1! 21 k =1 k 2
From this

  2
 log 2 e -k
 2  2 -
2 
 (2) = -2f <1>
+ +Σ
2! 1! k=1 k2

- 29 -
   2  log 2 -Σ  +Σ 
 1
 22k B2k  2k+1

 2
1 1 1 
= -2
1! k=1 k k=1 2k (2k +1)!

 2
 log 2 e -k
 2
2 
+ - +Σ
2! 1! k =1 k2
B2k 
 1! log -Σ   1  2  e -k 
2k+1
 1 1 
+Σ
2 2! Σ
=- + + 2
k =1 k k=1 2k (2k +1)! k=1 k

Next, substitute (a) for Taylor series (1.3') , then

<2> 
   coth x dx
0 x x

 2
3
= -f (1.3')
  
2 2 2

  3
 2
 e -k
   2  2  2 Σ
1 1 log 2 1 
- 1  (2)+ - + 1
2 2 2! 1! 2 k=1 k2
From this and Fourier series (2.3') , we obtain

   3
 2

 2    2  2
1 1 log 2
f<2>
=- 1  (2) + -
2 2 3! 2!
1 1  e -k 
+ 2  (3)- 2 Σ 3 (b)
2 2 k=1 k
From this,

 22  3
 2log 2 e -k
 2    2 +

 (3) = 2 f 2 <2>
+2  (2) - +Σ
2 3! 2! k=1 k3

  2  log 2 
 2
 22k B2k  2k+2

  2
1 2 1 
=2 2
-Σ +Σ
2! k =1 k k=1 2k (2k +2)!

22  3
 2log 2 e -k
 2 +

+  (2)- +Σ
3! 2! k =1 k3
2  B2k 1 3
2k+2 -k 

log -Σ 
2 1   e
+Σ +  (2)
2 3! Σ
= - + 3
2! k =1 k k=1 2k (2k +2)! k=1 k

Next, substitute (a), (b) for Taylor series (1.4') , then

  
   coth x dx = - f  2 
0 x 1

 2  2
x 1
4 <3>
+ f<2>
   1!
2 2 2

 2

 2  2
1
- f<1> (1.4')'
2!

  e -k  2  e -k 
 2    
1  1
2 Σ 22! 2 Σ
<3>
= -f - 2 3
- 2
2 k=1 k k=1 k

 2 1   4

     
1 1
-  (2) + 2  (3) -
22! 2 2 2 23! 2
From this and Fourier series (2.4') , we obtain

- 30 -
   2

 2  2  2   (3)
1 1
f<3>
=-  (2) +
22! 22
 4
 3
1  e -k 
 2  2
1 1 log 2
- 3  (4) + - + 3Σ 4 (c)
2 4! 3! 2 k=1 k
From this,

  2  log 2 -Σ  +Σ 
 3
 22k B2k  2k+3

 2
1 3 1 
 (4) = -2 3
3! k=1 k k=1 2k (2k +3)!
 3
23  4

 2  2
log 2
- +
3! 4!
23  e -k  23  2
23 
+ 3Σ 4 -
2 k =1 k 22!  2  (2) +
22  2   (3)
B2k 
 3! log -Σ 
2k+3
3 4
 +Σ
3 1 
=- -
k =1 k k =1 2k (2k +3)! 24!

 e -k  2
+Σ 4
-  (2) +  (3)
k =1 k 2!
Hereafter, by induction we obtain (n) for natural number n >2 .

- 31 -
5.5 Termwise Higher Integral of csc x
log tan (x /2) of cscx is x =  /2 , we assume zeros of
Since the zero of the first order integral
the second or more order integral are also x =  /2 . Where, since the zero of cscx is i , these seem
to be a collateral higher integral.

5.5.0 Collateral Higher Integral of csc x

  
x x 1 1- cosx
csc x dx = log tan = log
 2 2 1+ cosx
2

  csc x dx
x x
2
= non-elementary function
 
2 2

5.5.1 Termwise Higher Integral of Taylor Series of csc x

Formula 5.5.1
(x) on 0< x < 
<n>
When B2k denote Bernoulli Numbers and functions f are as follows

xn 22k -2B2k 2k+n

 
n 1 
n =1, 2, 3, 
<n>
f (x) = log x -Σ +Σ x
n! k =1 k k=1 2k (2k +n )!

the following expressions hold for x >0 .

 

0

   2
x 1
<0>
csc x dx =f (x) - x- f<0>
 0! 2
2
22k -2B2k 2k

= log x +Σ x - log 2
k =1 2k (2k)!

  0 <1>   
x x 1

   2    2
1 1
csc x dx 2 = f <1>(x) - x- f
0!
-
2 1!
x-
2
f
<0>
 
2 2

    
x x x 0 1

  f  2   f  2
3 1 1
csc x dx = f <2>(x) - x-
<2>
- x-
<1>
   0! 2 1! 2
2 2 2
 2

   2
1 <0>
- x- f
2! 2

 
 
x x k

   2
n -1 1
 csc x dx = f n <n -1>
(x) -Σ x- f<
n -1-k >
  k=0 k! 2
2 2

Where f<0>  2  = log 2
Proof
x


22k -2B2k 2k
 csc x dx = f

 2 
x 
<0>
(x) - f<0> = log x +Σ x
 k=1 2k (2k)! 
2 2

- 32 -
 22k -2B2k 2k
 
 2 -2B2k
2k
   2k
= log x +Σ
k=1 2k (2k)!
x - log 2 +Σ
k =1 2k (2k ) !  2
22k -2B2k 2k  
 

= log x +Σ x - log - log
k=1 2k (2k)! 2 4

 
2 -2B2k
2k
  2k

 Σ
k=1 2k (2k ) !  2 = -log
4
5.3.1 (2)

22k -2B2k 2k

= log x +Σ x - log 2
k=1 2k (2k)!

   2  dx
x x x
2
csc x dx = f<0>(x) - f<0>
  
2 2 2
x
 
  2 
1

 
<1> 1 <0>
= f (x) - x- f
1! 2 
2
  1

 2    2
1
= f<1>(x) - f<1> - x- f<0>
1! 2
Hereafter, in a similar way we obtain the desired expressions..

Although Formula 5.5.1 is a general expression anyway, if it is expand actually, it is very complicated.
Then, this was devised and was made the same easy expression as sec x. It is as follows.

Formula 5.5.1'
When E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are Euler Numbers,the following expressions
hold for 0< x < .

  cscx dx
2k+ n
x E2k
x - 2 
x 
n
=Σ
  k =0 (2k + n)!
2 2

Proof

 
1 1
csc x = = = sec x -

 
sin x 2
cos x -
2
 E2k 
secx =Σ x 2k |x|<
k =0 (2k)! 2
From these

E2k  2k

x - 2 

cscx =Σ 0< x < 
k =0 (2k)!
Integrating both sides of this with respect to x from  /2 to x repeatedly, we obtain the desired expressions.

(1) Termwise Higher Definite Integral of Taylor Series of csc x

Substituting x =0 for Formula 5.5.1 , we obtain the following expressions.

- 33 -
    
  csc x dx
0 1

 2 f  2  2 f  2
0 x 1 1
2 <1> <0>
=- + (1.2')
  0! 1!
2 2
   
 
0 x 0 1

 2  2  2  2
x 1 1
3 <2>
csc x dx = - f + f<1>
   0! 1!
2 2 2

 2

 2  2
1
- f<0> (1.3')
2!

 
   csc x dx
0 x (-1)k k

 2  2
x n -1
n -1-k >
n
= -Σ f< (1.n')
   k=0 k!
2 2 2

Where f<0>  2  = log 2
(2) Taylor Series of the 1st order Integral of csc x
The 1st order Integral of csc x is as follows.
 2 -2B 2k
2k


x x
csc x dx = log x +Σ x 2k - log 2 = log tan 0< x < 
 k=1 2k (2k)! 2
2
From this, the following expression follows immediately.

 2 -2B 2k
2k
x x
Σ
2k
x = log tan - log 0< x <  (1.t)
k =1 2k (2k)! 2 2
Substituting x =1 for this , we obtain the following special value.
 22k -2B2k 1
Σ
k =1 2k (2k)!
= log tan + log 2
2

5.5.2 Termwise Higher Integral of Fourier Series of csc x

Formula 5.5.2
(-1)k 
Let  (n ) =Σ be Dirichlet Beta Function and  be floor function. Then the following
k =0 (2k +1)
n

expressions hold for 0< x <  .

1
  
x  1
csc x dx = 2Σ sin (2k +1)x -
 k =0 2k +1 2
2
2 2(2) 
  csc x dx
0

   
x x  1
2
= 2Σ 2 +1)x -
2 sin ( k
+ x-
  k =0 (2k +1) 2 0! 2
2 2

3 2(2) 
   csc x dx
1

 
x

 
x x  1
3
= 2Σ sin (2k +1)x - + x-
   k =0 (2k +1)3 2 1! 2
2 2 2

- 34 -
 4
  csc x dx  
x x  1
 4
= 2Σ 2 +1)x -
4 sin ( k
  k =0 (2k +1) 2
2 2

2(2)  2 2(4)  0
+
2! 
x-
2
- 
0!
x-
2  
5
   
x x  1
 csc x dx 5 = 2Σ 2 +1)x -
5 sin ( k
  k =0 (2k +1) 2
2 2

2(2)  3
2(4)  1
+
3!
x-
2   -
1! 
x-
2 

n
  csc x dx  
x x  1
n
= 2Σ sin (2k +1)x -
  k=0 (2k +1)n 2
2 2
2 (2k)  n-2k

 
n /2
k-1
+ Σ (-1) x-
k=1 (n -2k)! 2

Proof
csc x can be expanded to Fourier series in a broad meaning.as follows.

2i 2ie ix
cscx = ix -ix = - 2ix
= -2ie ix1 + e 2ix + e 4ix + e 6ix +
e -e 1- e
= -2 ie ix + e 3ix + e 5ix + e 7ix +
= 2(sin x + sin 3x + sin 5x ++) -2i(cosx + cos 3x + cos 5x + ) (2.0)

csc x is the locus of the median (central line) of this real number part. (See the figure)

However, calculus of the right side of (2.0) may be carried out, the real number does not turn into an imagi-
nary number and the contrary does not exist, either. Therefore, limiting the range of csc x and the higher order
integral to the real number, we calculate only the real number part of the right side of (2.0). Although Dirichlet
Odd Beta is obtained from calculation of the imaginary number part, since it is long, it is omitted in this section
Now, integrate the real number part of both sides of (2.0) with respect to x from  /2 to x . Then

- 35 -
   2(sin x + sin 3x + sin 5x + sin 7x +)dx
x x
csc x dx =
 
2 2
x

 
cos 1x cos 3x cos 5x cos 7x
= -2 + + + +
1 3 5 7 
2

 
cos 1x cos 3x cos 5x cos 8x
= -2 + + + +
1 3 5 7
i.e.


x  cos(2k +1)x
csc x dx = -2Σ
 k=0 2k +1
2

Here, let -cos x = sin x -  2  . Then using this we obtain

1
  
x  1
csc x dx = 2Σ sin (2k +1)x -
 k =0 2k +1 2
2
This is consistent with the real number part of the Fourier series of log tan(x /2) .
Next, integrate both sides of this with respect to x from  /2 to x , then
1
  2Σ 2k +1 sin (2k +1)x - 2  dx
x x x  1
csc x dx 2 =
   k=0
2 2 2
2 x

  
 1
= 2Σ sin (2k +1)x -
k=0 (2k +1)2 2

2

2 (-1)k
 
 1 
= 2Σ sin (2k +1)x - + 2Σ
k =0 (2k +1)2 2 k=0 (2k +1)2
 (-1)k
Here, let  (n) =Σ , then we obtain
k =0 (2k +1)n
2
   + 2(2)
x x  1
csc x dx 2 = 2Σ sin (2k +1)x -
  k =0 (2k +1)2 2
2 2
Next, integrate both sides of this with respect to x from  /2 to x , then
2
  2Σ (2k +1) sin (2k +1)x - 2  + 2(2) dx
x x x x  1
csc x dx 3 = 2
    k =0
2 2 2 2

3  x

   + 2(2)x - 2 
 1
= 2Σ 2 +1)x -
3 sin ( k
k =0 (2k +1) 2

3 
2

  + 2(2)x - 2 
 1
= 2Σ sin (2k +1)x -
k =0 (2k +1)3 2
Hereafter, in a similar way we obtain the desired expression..

- 36 -
(1) Termwise Higher Definite Integral of Fourier Series of csc x
Substituting x =0 for Formula 5.5.2 , we obtain the following expressions.

2(2) 

0

 2
0 x
2
csc x dx = (2.2')
  0!
2 2
2 (2) 
   csc x dx
1
0 x 23-1
 2
x
3
=- +  (3) (2.3')
   1! 22
2 2 2
2 (2)  2(4) 
   csc x dx
0 x 2 0

 2  2
x
4
= - (2.4')
   2! 0!
2 2 2
2 (2)  2(4) 
 
3 1
0 x x 25-1
 csc x dx = -  2  2  (5)
5
+ - (2.5')
   3! 1! 24
2 2 2


2(2k) 
   csc x dx
0 x n-2k

 2
x n /2
k-1
  
n
= (-1) n
Σ
k =1
(-1)
(n -2k)!
2 2 2
n 2n-1
- sin   (n) (2.n')
2 2n-1

Proof
Since
n

sin (2k +1)0 -
2 =0 for n =2, 4, 6, 
n

sin (2k +1)0 -
2  = 1 for n =3, 5, 7,  ,

giving x =0 to Formula 5.5.2 and substituting these for them,

2
   + 2(2) = 2(2)
0 x  1
csc x dx 2 = 2Σ 2 +1)0-
2 sin ( k
  k =0 (2k +1) 2
2 2

3 2 (2) 
   csc x dx
1

 
0

 
x x  1
3
= 2Σ sin (2k +1)0- + 0-
   k =0 (2k +1)3 2 1! 2
2 2 2
2(2)  1
23-1 2 (2)  1

 2  2
 1
= 2Σ 3
- = 2
 (3) -
k =0 (2k +1) 1! 2 1!
Hereafter, in a similar way we obtain the desired expressions..

(2) Fourier Series of the 1st order Integral of csc x

The 1st order Integral of csc x is as follows.


x  cos(2k +1)x x
csc x dx = -2Σ = log tan 0< x < 
 k =0 2k +1 2
2
From this, the following expression follows immediately.

- 37 -
  cos(2k +1)x 1 x
Σ = - log tan 0< x <  (2.f)
k =0 2k +1 2 2
Substituting x =1 for this, we obtain the following special values.
 cos(2k +1) 1 1
Σk =0 2k +1
= - log tan
2 2

5.5.3 Dirichlet Even Beta

Comparing Taylor series of csc x and Fourier series of csc x, we obtain Dirichlet Even Beta. Although the
general expression by the Euler Number is known about Dirichlet Odd Beta, the general expression of Dirichlet
Even Beta is not known. The following formula shows this by a little complicated Bernouilli series.

Formula 5.5.3
When B2k , (x),  (x) denote Bernoulli Numbers, Dirichlet Beta Function, Riemann Zeta Function
(x) on 0< x <  are as follows
<n>
respectively, and functions f

 2 -2B 2k
2k
xn
 
n 1
n =1, 2, 3,  ,
<n>
f (x)= log x -Σ +Σ x 2k+n
n ! k =1 k k=1 2 2
k k n
( + )!
the following expressions hold.
1 1 (-1)k  k
  log 2 
 2  2  2  2 
1
 (2) = - Σ f <1-k > = - f<1>
2 k=0 k ! 2
1 2 (-1)k  k-1
 2 -1  (3) 3
 (2) =
2Σ  2  2 2 
<2-k >
f + 2
k=0 k!
1 3 (-1)k  k
  (2)  2
 (4) =
2Σ  2 f  2  + 2!  2 
<3-k >
k=0 k!
1 4 (-1)k  k-1
 (2)  2
25-1  (5)
 (4) = - Σ  2  2  3!  2  +
<4-k >
f +
2 k=0 k ! 24 
1 5 (-1)k  k
  (4)  2
 (2)  4
 (6) = - Σ  2  f  2  + 2!  2  -  2
<5-k >
2 k =0 k ! 4!
1 6 (-1)k  
k-1
 (4)  2
 (2)  4
 (6) =
2Σ  2  2  3!  2  5!  2 
<6-k >
f + -
k=0 k!
27-1 (7)
+
26 

(-1)n 2n -1 (-1)k  k
  (2n -2k)  2k

 2 f  2  -Σ(-1)  2
n -1
 (2n) =
2 Σ
<2n-1-k > k
k =0 k! k=1 (2k) !
(-1)n 2n (-1)k  k-1
  (2n -2k)  2k

 2  2  -Σ(-1)  
n -1
 (2n) = -
2 Σ
<2n-k > k
f
k =0 k! k =1 (2k +1) ! 2
22n+1 -1 (2n +1)
+
22n 

Proof
   

0 1

 2  2  2  2
0 x 1 1
2 <1>
csc x dx = - f + f<0> (1.2')
  0! 1!
2 2

- 38 -
 (-1)k  k

 2 f  2
1
= -Σ <1-k >
k =0 k!
2(2)

0 x
csc x dx 2 = (2.2')
  0!
2 2

1 1 (-1)k  k
  log 2 
 2  2  2  2 
1
  (2) = - Σ f <1-k > = - f<1>
2 k=0 k ! 2
 
   csc x dx (-1)k
0 k

 2 f  2
x x 3-1
3
= -Σ <3-1-k >
(1.3')
   k =0 k!
2 2 2

2 (2) 
 
0 23-1 1

 2
x x
csc x dx =3
2
 (3) - (2.3')
   2 1!
2 2 2

1 3-1 (-1)k  k-1

 23-1  (3)
  (2) = Σ  2  2
<3-1-k >
f +
2 k=0 k ! 22 
Hereafter, in a similar way we obtain the desired expressions.

5.5.4 Riemann Odd Zeta

Riemann Odd Zeta can be conversely obtained from the half of Formula 5.5.3 . The following formula shows
the relation between Riemann Odd Zeta and Dirichlet Even Beta by a little complicated expression.

Formula 5.5.4
When B2k , (x),  (x) denote Bernoulli Numbers, Dirichlet Beta Function, Riemann Zeta Function
(x) on 0< x <  are as follows
<n>
respectively, and functions f

 2 -2B 2k
2k
xn
 
n 1
n =1, 2, 3,  ,
<n>
f (x)= log x -Σ +Σ x 2k+n
n! k =1 k k=1 2k (2k +n )!
the following expressions hold.

 (2)3
 1
22 2 (-1)k  k

 2  2 f  2
2
 (3) = 3 - 3 Σ <2-k >
2 -1 1! 2 -1 k=0 k !
 (4)   (2) 
  2 
25 1 3
 (5) = 5
2 -1 1!  2 -
3!
24 4 (-1)k  k

+ 5 Σ
2 -1 k =0 k !  2 f <4-k >  2
(6)   (4)   (2) 
  2 
27 1 3 5
 (7) = 7
2 -1 1!  2 -
3!  2 +
5!
26 6 (-1)k  k

- 7 Σ
2 -1 k=0 k !  2 f <6-k >  2


- 39 -
 22n+1 (2n -2k)  2k+1

 2
n -1
 (2n +1) = Σ (-1) k
22n+1 -1 k =0 (2k +1)!
22n (-1)k  k

 2  2
2n
+ (-1) n
Σ k!
f <2n -k >
22n+1 -1 k=0

- 40 -
5.6 Termwise Higher Integral of csch x
Since both zeros of csch x and the primitive function log tanhx /2
x = , the latter seems to be
are
a lineal primitive function with a fixed lower limit. On the other hand, the Fourier series of csch x is known for
x >0 . Then, we assume zeros of the second or more order primitive function are also x = .

5.6.0 Higher Integral of csch x


x x
csch x dx = log tanh
2

 csch x dx
x x
2
= non-elementary function

5.6.1 Termwise Higher Integral of Taylor Series of csch x

The Taylor series of csch x is known as follows.

 2 -2 B 2k
2k
1
csch x = -Σ x 2k-1 0< x < 
x k =1 2k (2k -1)!
However, this can not be integrated with the lower limit x = .

5.6.2 Termwise Higher Integral of Fourier Series of csch x

Formula 5.6.2
The following expressions hold for x >0 .
x

 csch x dx = -2Σ
x  e -(2k +1)
 k =0 2k +1

  csch x dx = 2Σ (e2k +1)

x x  -(2k +1)x
2
2
  k =0

   csch x dx = -2Σ (e2k +1)

x x x  -(2k +1)x
3
3
   k =0

  csch x dx
x
x x  e -(2k +1)
n
= (-1) 2Σ n
  k=0 (2k +1)n

Proof
csch x can be expanded to a Fourier series as follows. (See the figure of the following page.)

2 2e -x
csch x = x -x
= -2x
= 2e -x1 + e -2x + e -4x + e -6x +
e -e 1- e
= 2e -1x + e -3x + e -5x + e -7x + (2.0)

= 2(cosh 1x +cosh 3x +cosh 5x +) - 2(sinh 1x +sinh 3x +sinh 5x +)

= 2(cosh ix +cosh 3ix +cosh 5ix +)- 2i(sinh ix +sinh 3ix +sinh 5ix +)
  cosh x = cosix , sinh x = -isin ix 

- 41 -
 Integrate the both sides of (2.0) with respect to x from  to x , then

  2e
x x
csch x dx = -1x
+ e -3x + e -5x + e -7x +dx
 
x

 
e -x e -3x e -5x e -7x
= -2 + + + +
1 3 5 7 
i.e.
x


x  e -(2k +1)
csch x dx = -2Σ
 k=1 2k +1
x
This is consistent with the real number part of the Fourier series of log tanh .
2
Next, integrate both sides of this with respect to x from  to x , then
x

  csch x dx =  -2Σ  dx = 2Σ 

x x
x x
2
x  e -(2k +1)  e -(2k +1)
   k=0 2k +1 k =0 (2k +1)2 
i.e.
x

  csch x dx
x x  e -(2k +1)
2
= 2Σ
  k=0 (2k +1)2
Hereafter, in a similar way we obtein the desired expression.

5.6.3 Dirichlet Lambda Function

Substituting x =0 for Formula 5.6.2 , we obtain Dirichlet Lambda Function immediately.

Formula 5.6.3
 1
When (n) =Σ , the following expressions hold.
k=0 (2k +1)n

  csch x dx = 2(2)
0 x
2
 

   csch x dx = -2(3)
0 x x
3
  

- 42 -
 

   csch x dx
0 x x
 n
= (-1)n2 (n)
  

5.6.4 Riemann Zeta Function

2n
Applying (n) = (n) for Formula 5.6.3 , we obtain Riemann Zeta Function.
2n - 1

Formula 5.6.4

   csch x dx
2n-1 0 x x
 (n) = (-1) n n n
2 -1   

- 43 -
5.7 Termwise Higher Integral of sec x
1 1+ sin x
Since the zero of the first order integral log of secx is x = 0 , we assume the zeros of
2 1- sin x
the second or more order integral are also x = 0 . Where, since the zero of secx is i , these seem
to be a collateral higher integral.

5.7.0 Collateral Higher Integral of sec x


x 1 1+ sin x
sec x dx = log
0 2 1- sin x

 secx dx
x x
2
= non-elementary function
0 0

5.7.1 Termwise Higher Integral of Taylor Series of sec x

Formula 5.7.1
When E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are Euler Numbers, the following expressions

hold for |x|<  /2 .


x  E2k
sec x dx =Σ x 2k+1
0 k =0(2k +1)!

 sec x dx
x x  E2k
2
=Σ x 2k+2
0 0 k =0 (2k +2)!

 sec x dx
x x x  E2k
3
=Σ x 2k+3
0 0 0 k =0 (2k +3)!

 
x x  E2k
 secx dx n =Σ x 2k+
n
0 0 k =0 (2k + n)!

Proof
secx can be expanded to Taylor series as follows.

 E2k 
secx =Σ x 2k |x| <
k=0 (2k)! 2
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

(1) Termwise Higher Definite Integral of Taylor Series of sec x

Substituting x =  /2 for Formula 5.7.1 , we obtain the following expressions.


  sec x dx
E2k
2k+2

 2
2 x 
2
=Σ (1.2')
0 0 k=0 (2k +2)!


  sec x dx E2k
2 x x 2k+3

 2

3
=Σ (1.3')
0 0 0 k =0 (2k +3)!


- 44 -
 

  secx dx
2k+ n
2
E2k
 2
x x 
n
=Σ (1.n')
0 0 0 k =0 (2k + n)!

(2) Taylor Series of the 1st order Integral of sec x

The 1st order Integral of sec x is as follows.


x  E2k 1 1+ sin x
sec x dx =Σ 2k+1
x = log |x|<
0 k=0 (2k +1)! 2 1- sin x 2
From this, the following expression follows immediately.
 E2k 1 1+ sin x 
Σ
2k+1
x = log |x|< (1,t)
k=0 (2k +1)! 2 1- sin x 2
Substituting x =1 for this , we obtain the following special value.
 E2k 1 1+ sin 1
Σ
k=0 (2k +1)!
= log
2 1- sin 1

Reference : Euler Number , Tangent number, Bernoulli Number

Euler Number is defined as follows.
Ek k  
sech x =Σ x |x| <
k=0 k ! 2
As a result, E2n-1 = 0 n =1, 2, 3,  .
When Euler numbers , Tangent numbers and Bernoulli Numbers (all non-zero) are listed, it is as follows.
E0=1, E2=-1, E4=5, E6=-61, E8=1385, E10=-50521, 
T1=1, T3=2, T5=16, T7=272, T9=7936, T11= 353792, 
1 1 1 1 5
B0=1, B2= , B4=- , B 6= , B8=- , B10= ,
6 30 42 30 66
According to Mr. Sugimoto , the following equations holds between Euler Numbers and Tangent Numbers.

 
n 2n
E2n =Σ(-1)k T +1 n =1, 2, 3,  (1)
k =1 2k -1 2k-1
On the other hand, between Tangent numbers and Bernoulli Numbers, the following equations hold.

n-1 2 2 -1
2n 2n
T2n-1 ( )= -1 B2n n =1, 2, 3,  (2)
2n
Therefore, the following equations have to hold between Euler Numbers and Bernoulli Numbers.
2k 2k
2 2 -1
 
n 2n
E2n = 1 -Σ B2k n =1, 2, 3,  (3)
k =1 2k 2k -1
In fact, for example,


2222-1
 
2424-1
 
2626-1
 
23 23 23
E6 = 1 - B2 + B4 + B6
2 1 4 3 6 5

 2 6  1 4 30  3 6 42  5  = -61
43 1 6 1615 1 6 6463 1 6
=1- - +
On the contrary, the following equations between Bernoulli Numbers and Euler Numbers are shown by
Mr. Nishimura .

- 45 -
  
2n n -1 2n -1
B2n = Σ E2k n =1, 2, 3,  (4)
22n22n-1 k =0 2k
The following example shows the formula is right.

   2   E 
23 23-1 23-1 23-1
B6 = E0 + E2 + 4
223223-1 0 4

      4  42
23 5 5 5 1
= 1- 1+ 5 =
6463 0 2
In addition, Mr. Nishimura expressed Even Zeta with the Euler numbers using this equation.

5.7.2 Termwise Higher Integral of Fourier Series of sec x

Formula 5.7.2
When E2k ,  (n) ,  denote Euler Numbers, Dirichlet Beta Function, floor function respectively,
the following expressions hold for |x |<  /2 .

1

(-1)k
 
x 
sec x dx = 2Σ cos (2k +1)x -
0 k=0 2k +1 2
2 2 (2) 0

(-1)k
 
x x 
secx dx 2
= 2Σ 2 cos (2k +1)x - + x
0 0 k =0 (2k +1) 2 0!
3 2 (2) 1

k

 
x x x  (-1)
sec x dx 3 = 2Σ 2 +1)x -
3 cos ( k
+ x
0 0 0 k=0 (2k +1) 2 1!
4
 
(-1)k
 
x x 
 sec x dx 4 = 2Σ 2 +1)x -
4 cos ( k
0 0 k =0 (2k +1) 2
2 (2) 2 2 (4) 0
+ x - x
2! 0!
5
  sec x dx (-1)k
 
x x 
5
= 2Σ 2 +1)x -
5 cos ( k
0 0 k=0 (2k +1) 2
2 (2) 3 2 (4) 1
+ x - x
3! 1!

n
  sec x dx
(-1)k
 
x x 
n
= 2Σ cos (2k +1)x -
0 0 k =0 (2k +1)n 2
n /2 2 (2k) n-2k
+ Σ (-1)k-1 x
k=1 (n -2k)!

Proof
sec x can be expanded to Fourier series in a broad meaning.as follows.
2 2e ix
secx = ix -ix = 2ix
= 2e ix - e 3ix + e 5ix - e 7ix +-
e +e 1+ e
= 2(cosx - cos 3x + cos 5x - cos 7x +-)
+ 2i(sin x - sin 3x + sin 5x - sin 7x +-) (2.0)

- 46 -
sec x is the locus of the median (central line) of this real number part. (See the figure)

y
40

30

20

10

-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
x
1/cos(x)
2*cos(x) - 2*cos(3*x) + 2*cos(5*x) - 2*cos(7*x) +

However calculus of the right side of (2.0) may be carried out, the real number does not turn into an imaginary
number and the contrary does not exist, either. Therefore, limiting the range of sec x and the higher order
integral to the real number, we calculate only the real number part of the right side of (2.0) .
Although the Dirichlet Odd Beta is obtained from calculation of the imaginary number part, since it is long, it is
omitted in this section.
Now, integrate the real number part of both sides of (2.0) with respect to x from 0 to x, then

 sec x dx =  2(cosx - cos 3x + cos 5x - cos 7x +-)dx

x x

0 0
x

 
sin 1x sin 3x sin 5x sin 8x
=2 - + - +-
1 2 3 4 0

 
sin 1x sin 3x sin 5x sin 8x
=2 - + - +-
1 2 3 4
i.e.

 sec x dx = 2Σ
x  (-1)k
sin (2k +1)x
0 k =0 2k +1 

Here, since sin x = cos x -  2  , we obtain

1
 sec x dx = 2Σ
(-1)k
 
x 
cos (2k +1)x -
0 k =0 2k +1 2
1 1+ sin x
This is consistent with the real number part of the Fourier series of log .
2 1- sin x
Next, integrate both sides of this with respect to x from 0 to x , then

1
   dx
(-1)k

x x x 
secx dx = 2
2Σ cos (2k +1)x -
0 0 0 k =0 2k +1 2
x
2
 
(-1)k


= 2Σ 2 cos (2k +1)x -
k =0 (2k +1) 2
0

- 47 -
 (-1)k 2 (-1)k
  - 2Σ
 
= 2Σ cos (2k +1)x - cos(-)
k =0 (2k +1)2 2 k =0 (2k +1)2
 (-1)k
Here, let (n) =Σ , then we obtain
k =0 (2k +1)n
2 2 (2) 0
 secx dx
(-1)k
 
x x 
2
= 2Σ 2 +1)x -
2 cos ( k
+ x
0 0 k =0 (2k +1) 2 0!
Next, integrate both sides of this with respect to x from 0 to x , then

2 2 (2) 0
  
(-1)k
 
x x x x 
sec x dx = 3
2Σ 2 cos (2k +1)x - + x dx
0 0 0 0 k=0 (2k +1) 2 0!
x
3 2 (2) 1
 
(-1)k
 

= 2Σ 3 cos ( k
2 +1)x - + x
k =0 (2k +1) 2 1!
0
k
3 2 (2) 1
 
 (-1)
= 2Σ 3 cos ( k
2 +1)x - + x
k =0 (2k +1) 2 1!
Hereafter, in a similar way , we obtain the desired expressions.

(1) Termwise Higher Definite Integral of Fourier Series of sec x

Substituting x = /2 for Formula 5.7.2 , we obtain the following expressions.

2 (2) 
  secx dx
0

 2
2 x
2
= (2.2')
0 0 0!

2 (2) 
 
1
2 x x 23-1
 2  (3)
3
sec x dx = - (2.3')
0 0 0 1! 22

2 (2)  2 (4) 

2 x 2 0

 2  2
x
4
sec x dx = - (2.4')
0 0 0 2! 0!

2 (2)  2 (4) 

3 1
2 x 25-1
 2  2
x
sec x dx = 5
- +  (5) (2.5')
0 0 0 3! 1! 24


2 (2k)  n 2n-1
 
2 x n-2k

 2
x n /2
 sec x dx = Σ (-1)
n k-1
+ sin   (n)
0 0 0 k =1 (n -2k)! 2 2n-1
(2.n')

Proof
Since
n 

cos (2k +1)
2 2
- =0 for n =2, 4, 6, 
 n n =1, 3, 5, 

cos (2k +1) -
2 2  = (-1) k+1
for
k =0, 1, 2, 
,

- 48 -
giving x = /2 to Formula 5.7.2 and substituting these for them,

 2
  secx dx
(-1)k
  + 2(2) = 2(2)
2 x 
2
= 2Σ cos (2k +1) -
0 0 k=0 (2k +1)2 2 2

 3 2 (2) 
  sec x dx
(-1)k 1

 
2 x x

 2

3
= 2Σ 2 +1) -
3 cos ( k
+
0 0 0 k=0 (2k +1) 2 2 1!

(-1)k (-1)k+1 2 (2)  1

 

= 2Σ +
k=0 (2k +1)3 1! 2
23-1 2 (2)  1
= - 2  (3) +
2 1! 2  
Hereafter, in a similar way we obtein the desired expressions.

(2) Fourier Series of the 1st order Integral of sec x

The 1st order Integral of sec x is as follows.

 sec x dx = 2Σ(-1)
x 
k sin(2k +1)x 1 1+ sin x
= log |x|<
0 k =0 2k +1 2 1- sin x 2
From this, the following expression follows immediately.
 sin(2k +1)x 1 1+ sin x 
Σ
k=0
(-1)k
2k +1
= log
4 1- sin x
|x|<
2
(2.f)

Substituting x =1 for this, we obtain the following special value.

 sin(2k +1) 1 1+ sin 1
Σ
k=0
(-1)k
2k +1
= log
4 1- sin 1

5.7.3 Dirichlet Even Beta

Comparing Taylor series of sec x and Fourier series of sec x, we obtain Dirichlet Even Beta. Although the
general expression by the Euler Number is known about Dirichlet Odd Beta, the general expression of
Dirichlet Even Beta is not known. The following formula shows this by the easy Euler series.

Formula 5.7.3
When E2k ,  (x),  (x) denote Euler Numbers, Dirichlet Beta Function, Riemann Zeta Function
respectively, the following expressions hold.
1  E2k  2k+2
(2) =
2Σk =0 (2k +2)!  2
1  E2k  2k+2
23-1  (3)
(2) =
2Σk =0 (2k +3)!  2 +
22 
1  E2k  2k+4
(2)  2
(4) = - Σ
2 k=0 (2k +4)!  2 +
2!  2
1  E2k  2k+4
(2)  2
25-1  (5)
(4) = - Σ
2 k=0 (2k +5)!  2 +
3!  2 +
24 
1  E2k  2k+6
(4)  2
(2)  4
(6) =
2Σk =0 (2k +6)!  2 +
2!  2 -
4!  2
- 49 -
 1  E2k  2k+6
(4)  2
(2)  4
(6) =
2Σk =0 (2k +7)!  2 +
3!  2 -
5! 2  
2 -1  (7)
7
+
26 

(-1)n-1  E2k  2k+2n
 (2n -2k)  2k

 2  2
n
 (2n) = Σ +Σ(-1)k-1
2 k=0 (2k +2n ) ! k=1 (2k) !
(-1)n-1  E2k   (2n -2k ) 
2k+2n+1 2k

 2  2
n
 (2n) = Σ +Σ(-1)k-1
2 k =0 (2 k +2 n +1) ! k =1 (2k +1)!
22n+1-1 (2n +1)
+
22n 

Proof


  sec x dx
E2k
2k+2

 2
2 x 
2
=Σ (1.2')
0 0 k=0 (2k +2)!

2 (2)
  secx dx
2 x
2
= (2.2')
0 0 0!
1  E2k  2k+2
  (2) = Σ
2 k =0 (2k +2)!  2


  sec x dx E2k
2 x x 2k+3

 2

3
=Σ (1.3')
0 0 0 k =0 (2k +3)!

2 (2)  1
 
2 x x 23-1
sec x dx = - 2  (3) +  
3
(2.3')
0 0 0 2 1! 2
1  E2k  2k+2 23-1  (3)
  (2) = Σ
2 k=0 (2k +3)! 2
+
22   
Hereafter, in a similar way we obtein the desired expressions.

c.f.
Mr.Sugioka expressed Dirichlet Even Beta by definite Reamann Odd Zeta and infinite Reiamnn Even Zeta
using his " Taylor System ".  (4),  (6) which he calculated are as follows.
22-1  1

 2
1
 (4) = (3)
22 1!
3
 2 3! 
log 2 22-1 1! 24-1 3! 26-1 5!
- - (2) - (4) - (6) -
22 1
22 22 5! 24 24 7! 26 26 9!
24-1  1
22-1  3

 2  2
1 1
 (6) = (5) - (3)
24 1! 22 3!
5
 2 5! 
log 2 22-1 1! 24-1 3! 26-1 5!
+ 4 1
- 2
(2) 2
- 4
(4) 4
- 6
(6) 6
-
2 2 2 7! 2 2 9! 2 2 11!

- 50 -
5.7.4 Riemann Odd Zeta
Riemann Odd Zeta can be conversely obtained from the half of Formula 5.7.3 . The following formula shows
the relation between Riemann Odd Zeta and Dirichlet Even Beta by the easy expression.

Formula 5.7.4
When E2k , (x),  (x) denote Euler Numbers, Dirichlet Beta Function, Riemann Zeta Function
respectively, the following expressions hold.

23  (2)  1
22  E2k  2k+3
 (3) = 3
2 -1 1!  2 - 3 Σ
2 -1 k=0 (2k +3)!  2
(4)  (2) 
  2 
25 1 3
 (5) = 5
2 -1 1!  2 -
3!
24  E2k  2k+5
+ 5 Σ
2 -1 k =0 (2k +5)!  2
(6)  (4)  (2) 
  2 
27 1 3 5
 (7) = 7
2 -1 1!  2 -
3!  2 +
5!
26  E2k  2k+7
- 7 Σ
2 -1 k =0 (2k +7)!  2

22n+1 (2n -2k)  2k+1

 2
n -1
 (2n +1) = Σ (-1) k
22n+1 -1 k=0 (2k +1)!
22n E2k  2k+2n+1

k=0 (2k +2n +1)!  2 


+ (-1) n
Σ
22n+1 -1

- 51 -
5.8 Termwise Higher Integral of sech x
2tan e x-
-1
Since both zeros of sech x and the primitive function are x = , the latter seems to be
a lineal primitive function with a fixed lower limit. Then, we assume zeros of the second or more order primitive
function are also x = .

5.8.0 Higher Integral of sech x

 sech x dx = 2tan e  - 
x
-1 x


 sech x dx = non-elementary function

x x
2
 

5.8.1 Termwise Higher Integral of Taylor Series of sech x

When E2k denote Euler Numbers, the Taylor series of sech x is known as follows.
 E2k 2k 
sech x =Σ x |x|<
k=0 (2k)! 2
However, this can not be integrated with the lower limit x = .

5.8.2 Termwise Higher Integral of Fourier Series of sech x

Formula 5.8.2
The following expressions hold for x >0 .
x

 sech x dx
x  e -(2k +1)
= -2Σ(-1)k
 k =0 2k +1

  sech x dx = 2Σ(-1) (e2k +1)

x x -(2k +1)x

2 k
2
  k=0


x x x -(2k +1)x
e 
3 k
sech
  
x dx = -2Σ(-1)
(2k +1) k=0
3

 
x
x x e -(2k +1) 
sech x dx = (-1) 2Σ(-1)
n n k
  k=0 (2k +1)n

Proof
sech x can be expanded to a Fourier series as follows. (See the figure of the following page.)

2 2e -x
sech x = x -x
= -2x
= 2e -x 1 - e -2x + e -4x - e -6x +-
e +e 1 +e
= 2e -1x - e -3x + e -5x - e -7x +- (2.0)

= 2(cos 1ix - cos 3ix + cos 5ix - cos 7ix +- )

+ 2i(sin 1ix - sin 3ix + sin 5ix - sin 7ix +- )

- 52 -
 y
1.0

0.8

0.6

0.4

0.2

0.0
0 1 2 3 4 5
x
1/cosh(x)
2*(exp(-x) - exp(-201*x))/(exp(-2*x) + 1)

Integrate the both sides of (2.0) with respect to x from  to x , then

  2e
x x
sech x dx = -x
- e -3x + e -5x - e -7x +-dx
 
x

 
e -x e -3x e -5x e -7x
= -2 - + - +-
1 3 5 7 0
i.e.
x


x e -(2k +1)
sech x dx = -2Σ(-1) k
 k =0 2k +1
2tan e x-
-1
This is consistent with the real number part of the Fourier series of .
Next, integrate both sides of this with respect to x from  to x , then
x

 
x x


-(2k +1)
x x
k e
  e -(2k +1)
sech x dx = 2Σ(-1)
2
= 2Σ(-1) k
  k=0 (2k +1)2 
k=0 (2k +1)2
Hereafter, in a similar way we obtein the desired expressions.

5.8.3 Dirichlet Beta Function

Substituting x =0 for Formula 5.8.2 , we obtain Dirichlet Beta Function immediately.

Formula 5.8.3
 (-1)k
When (n) =Σ is the Dirichlet Beta Function, the following expressions hold.
k=0 (2k +1)n

 sech x dx = -2(1)
0

  sech x dx = 2(2)
0 x
2
 

   sech x dx = -2(3)
0 x x
3
  

- 53 -
 

   sech x dx
0 x x
 n
= (-1)n2 (n)
  

Note
1 1 1 1 
(1) = 1
- 1 + 1 - 1 +- = : Madhava series
1 3 5 7 4
1 1 1 1
(2) = 2 - 2 + 2 - 2 +- = 0.9159655941 : Catalan's constant
1 3 5 7

2006.10.05
2012.05.29 Updated
K. Kono
Alien's Mathematics

- 54 -
06 Termwise Higher Integral (Inv-Trigonometric, Inv-Hyperbolic)
The 2nd or more order integrals of Inverse Trigonometric Functions and Inverse Hyperbolic Functions were
all shown in 4.4 and 4.5 . However, as for these, the integration lower limits and the function forms were
complicated. On the contrary, integrating term by term the Taylor series of these functions, we can obtain the
simple general forms. Naturally, many of these are the collateral higher integrals. However, it cannot be said
that the lineal higher integral is useful and the collateral one is useless. Simple is the best ! A simple infinite
series is superior to the complicated definite terms function.

6.1 Termwise Higher Integral of arctan x

Formula 6.1.1
The following expression holds for |x|< 1 .


x x  (2k)!
 tan -1x dx n = Σ(-1)k x 2k+
n+1
0 0 k=0 (2k + n +1)!

Proof
 1  (2k)! 2k+1
tan -1x = Σ(-1)k x 2k+1 = Σ(-1)k x
k =0 2k +1 k=0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The lineal higher integral of this was as follows as shown in Formula 4.4.1 ( 4.4 ).

 tan
x x tan -1x n /2 n-2k
n
Σ (-1) nＣ n-2k x
k
-1
x dx =
0 0 n! k=0

log1+ x 2 n /2
+ Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k=1

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)x n+1-2r
!
n r=1
Therefore, the termwise higher integral above is a lineal higher integral. And matching both, we obtain the
following formula.

Formula 6.1.1'
The following expression holds for |x| 1 .
 (2k)! tan -1x n /2
Σ(-1)
k
x 2k+n
= Σ (-1)k nＣ n-2k x n-2k
k =0 (2k +n +1)! n ! k =0

log1+ x 2 n /2
+ Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k=1

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)x n+1-2r
n! r=1

Substituting x =1 for this, we obtain the following special values.

 (2k)! 1 1 1 1  log 2
Σ
k =0
(-1)k
(2k +2)!
= - + -
12 34 56 78
+- =
4
-
2
 (2k)! 1 1 1 1 1 log2
Σ
k =0
(-1)k
(2k +3)!
= - + -
123 345 567 789
+- =
2
-
2

-1-
  (2k)! 1 1 5  log 2
Σ
k =0
(-1)k
(2k +4)!
= -
1234 3456
+- = -
12 12
-
6
 (2k)! 1 1 5 
Σ
k =0
(-1)k
(2k +5)!
= -
12345 34567
+- =
36
-
24

 (2k) !  n /2 log 2 n /2
Σ
k =0
(-1)k
(2k +n +1) !
=
4 n! Σ
k=0
(-1)k nＣ n-2k + Σ
2n! k=1
(-1)k nＣ n+1-2k

1 n /2
- Σ (-1)r nＣ n+1-2r  (1+ n)- (2r) (1.n')
!
n r=1

Note
The following formula is known about these series.

(1- t)m-1

 (-1)k 1 1
Σ
k =0 (2k +1)(2k +2)(2k +m)
=
(m -1)! 0 1+ t 2
dt
This is easily drawn from Formula 6.1.1 using Riemann-Liouville Integral.
And the following expression is also drawn easily..


n
1 (1- x)  n /2 log 2 n /2
2 dx
= Σ (-1)k nＣ n-2k + Σ (-1)k nＣ n+1-2k
0 1+ x 4 k =0 2 k =1
n /2
- Σ (-1)r nＣ n+1-2r  (1+ n )- (2r)
r=1

Higher order Gregory Series

In the special values, a noteworthy series is the 4th. i.e.
 (2k)! 1 1 5 
Σ
k =0
(-1)k
(2k +5)!
= -
12345 34567
+- =
36
-
24
Remembering that the 0th series was Gregory Series
 (2k)! 1 1 1 1 
Σ
k =0
(-1)k
(2k +1)!
= - + - +- =
1 3 5 7 4
we can find out such a series occurs at four cycles. Then, let us derive such a series.
First, from (1.n')
 n /2 1 n /2
Σ (-1)k nＣ n-2k - Σ (-1)r nＣ n+1-2r  (1+ n)- (2r)
4 n ! k =0 n ! r=1
log 2 n /2  (2k)!
+ Σ
2n ! k=1
(-1)k nＣ n+1-2k = Σ(-1)k
k=0 (2k + n +1)!
n /2
k-1
Especially, when n is the multiple of 4, since Σ
k=1
(-1) Ｃ n+1-2k = 0
n

 n /2 1 n /2
(-1)r nＣ n+1-2r  (1+ n)- (2r)
4 n! Σ (-1)knＣ n-2k -
k =0 n! Σ
r=1
 (2k)!
=Σ(-1)k
k =0 (2k + n +1)!
Replacing n with 4n, we obtain the following equation.

-2-
  2n 1 2n
Σ(-1)k 4nＣ 4n-2k - (-1)r4nＣ 4n+1-2r  (1+4 n)- (2r)
4(4n)! k =0 (4n )! Σ
r=1
 (2k)!
=Σ(-1)k
k =0 (2k +4n +1)!

This is the one having to call Higher order Gregory Series . When n=2,3, it is as followes.
 109 1 1
- = - +-
10080 325800 123456789 34567891011
87217  1 1
- = - +-
829870272000 29937600 12  1213 34 1415

-3-
6.2 Termwise Higher Integral of arccot x

Formula 6.2.1
The following expression holds for 0< x 1 .
 xn

x x  (2k)!
 -1
cot x dx =n
- Σ(-1)k x 2k+
n+1
0 0 2 n! k =0 (2k + n +1)!

Proof
  1   (2k)! 2k+1
cot -1x = -Σ(-1)k x 2k+1 = -Σ(-1)k x
2 k=0 2k +1 2 k=0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The lineal higher integral of this was as follows as shown in Formula 4.4.1 ( 4.4 ).

 cot
x x xn tan -1x n /2 n-2k
n
Σ k
(-1) nＣ n-2k x
-1 -1
x dx = cot x -
0 0 n! n! k=1

log1+ x 2 n /2
- Σ (-1)k nＣ n+1-2k x n+1-2k
2n ! k =1

1 n /2
(-1)r nＣ n+1-2r  (1+ n)- (2r)x n+1-2r
n! Σ
+
r=1
Therefore, the termwise higher integral above is a lineal higher integral.

-4-
6.3 Termwise Higher Integral of arcsin x

Formula 6.3.1
|x|< 1 .
The following expression holds for


x x 2
 (2k -1)!!
 sin x dx = Σ
-1 n
x 2k+
n+1
Collateral
0 0 k=0 (2k + n +1)!

Proof
 (2k -1)!!  (2k -1)!!(2k)!
sin -1x = Σ x 2k+1 = Σ x 2k+1
k =0 (2k )!! (2k +1 ) k=0 (2k)!! (2k +1)!
2
 (2k -1)!!
=Σ x 2k+1  (2k)! = (2k)!!(2k -1)!!
k=0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The collateral higher integral of this was as follows as shown in Formula 4.4.2' ( 4.4 ).
n-2r


x x n/2 x
 sin x dx =-1 n
Σ 2 sin -1x
0 0 r=0 (2r)!! (n -2r)!

n-2r+1
n /2 n -2r+1 (s -1)!!2 x
+Σ Σ (-1)sn -2r+1Ｃ s 2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
n-2r+1
n /2 x
+Σ
r=1 (2r -1)!!2(n -2r +1)!
Therefore, the termwise higher integral above is a collateral higher integral.
And matching both, we obtain the following formula.

Formula 6.3.1'
The following expression holds for |x| 1 .
2 n-2r
 (2k -1)!! n/2 x
2k+ n+1
Σ
k=0 (2k + n +1)!
x =Σ
r=0
2
(2r)!! (n -2r)!
sin -1x
n-2r+1
n /2 n -2r+1 (s -1)!!2 x
+Σ Σ s
(-1) n -2r+1Ｃ s 2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
n-2r+1
n /2 x
+Σ
r=1 (2r -1)!!2(n -2r +1)!

x =1 for this, we obtain the following special values.

Substituting

 (2k -1)!!
2

Σ
k=0 (2k +2)!
=
2
-1

 (2k -1)!!
2
3
Σ
k=0 (2k +3)!
=
8
-1

 (2k -1)!!
2
5 11
Σ
k=0 (2k +4)!
=
24
-
18

-5-
 
 (2k -1)!!
2
Ｃ n-1 n /2 1
 -Σ
2n -1
Σ
k=0 (2k + n +1)!
=
(2n)!! k=1 (2k -1)!!2(n -2k +1)!
(1.n')

Calculation of the circular constant by the double factorial series

We can calculate a circular constant using (1.n') . That is

 
2
 (2j -1)!! n /2 1 (2n)!!
= Σ +Σ
j=0 (2j + n +1)!
2
k=1 (2k -1)!! (n -2k +1)! Ｃ n-1
2n -1

Calculating formula
Whencn , c =0.851.00 denotes the number of significant figures,

  
1 n /2 (2k -1)!!2 1 (2n)!!
(cn)= +Σ +
(n +1)! k=1 (2k + n +1)! (2k -1)!!2(n -2k +1)! 2n -1Ｃ n-1
In this formula, the number of terms required for the number of significant figures cn is about n /2 . And
we can assume c=1 when large n is taken.
In the following example, supposing n =1,700 and c=0.9 , we obtained the significant figure of 1,530
digits. And the number of terms required for this was 851 .

Example (1530)
 n:=1700: DIGITS:=floor(0.9*n): MAXDEPTH:=1000:
 f := sum((2*k-1)!!^2/(2*k+n+1)!+1/((2*k-1)!!^2*(n-2*k+1)!)
, k=1..ceil(n/2)):
 pi := (1/(n+1)!+f)*(2*n)!!/binomial(2*n-1,n-1):
 float(pi)

-6-
6.4 Termwise Higher Integral of arccos x

Formula 6.4.1
The following expression holds for |x|< 1 .
 xn
 cos
x x 2
 (2k -1)!! n+1
-1 n
x dx = -Σ x 2k+ Collateral
0 0 2 n! k =0 (2k + n +1)!

Proof
 (2k -1)!!   (2k -1)!!(2k)!
cos -1x = -Σ x 2k+1 = -Σ x 2k+1
2 k=0 (2k)!! (2k +1) 2 k=0 (2k)!! (2k +1)!
  (2k -1)!!
2
= -Σ x 2k+1  (2k)! = (2k)!!(2k -1)!!
2 k =0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The collateral higher integral of this was as follows as shown in Formula 4.4.2' ( 4.4 ).
n-2r


x x xn n/2 x
 cos x dx = -1 n -1
cos x - Σ 2 sin -1x
0 0 n ! r=1 (2r)!! (n -2r)!

n-2r+1
n /2 n -2r+1 (s -1)!!2 x
-Σ Σ (-1) n -2r+1Ｃ s
s
2 (n -2r +1)!
1-x 2
r=1 s=0 (s +2r -1)!!
n-2r+1
n /2 x
+Σ
r=1 (2r -1)!!2(n -2r +1)!
Therefore, the termwise higher integral above is a collateral higher integral.

-7-
6.5 Termwise Higher Integral of arctanh x

Formula 6.5.1
|x|< 1 .
The following expression holds for


x x  (2k)!
 tanh -1x dx n = Σ x 2k+
n+1
0 0 k =0 (2k + n +1)!

Proof
 1  (2k)! 2k+1
tanh -1x = Σ x 2k+1 = Σ x
k=0 2k +1 k=0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The lineal higher integral of this was as follows as shown in Formula 4.5.1 ( 4.5 ).

 tanh
x x tanh -1x n /2 n-2k
n
Σ nＣ n-2k x
-1
x dx =
0 0 n! k=0

log1- x 2 n /2 n+1-2k

+
2n ! Σk=1
nＣ n+1-2k x

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
- Σ
n! r=1
Therefore, the termwise higher integral above is a lineal higher integral.
And matching both, we obtain the following formula.

Formula 6.5.1'
The following expression holds for |x| 1 .
 (2k)! 2k+ n+1 tanh -1x n /2 n-2k
Σ
k=0 (2k + n +1)!
x = Σ
n ! k =0
nＣ n-2k x

log1- x 2 n /2 n+1-2k

+
2n ! Σ
k=1
nＣ n+1-2k x

1 n /2
nＣ n+1-2r  (1+ n )- (2r)x
n+1-2r
-
! Σ
n r=1

Substituting x =1 for this, we obtain the following special values.

 (2k)! 1 1 1 1
Σ
k=0 (2k +2)!
= + +
12 34 56 78
+ + = log 2
 (2k)! 1 1 1 1 1
Σ
k=0 (2k +3)!
= + +
123 345 567 789
+ + = log 2 -
2
 (2k)! 1 1 1 2 5
Σ
k=0 (2k +4)!
= +
1234 3456 5678
+ + = log 2 -
3 12

 (2k)! 2n-1 1 n /2
Σ nＣ n+1-2r  (1+ n )- (2r)
n! Σ
= log 2 -
k=0 (2k + n +1)! n! r=1

-8-
Note
The following formula is known about these series.

(1- t)m-1

 1 1 1
Σ
k=0 (2k +1)(2k +2)(2k +m)
=
(m -1)! 0 1- t 2
dt
This is easily drawn from Formula 6.5.1 using Riemann-Liouville Integral.
And the following expression is also drawn easily..


1 (1- x)n n /2
nＣ n+1-2r  (1+ n )- (2r)
n-1
0 1- x 2
dx = 2 log 2 - Σ
r=1

-9-
6.6 Termwise Higher Integral of arcsinh x

Formula 6.6.1
The following expression holds for |x|< 1 .

 sinh
2
x x 
k (2k -1)!! n+1
-1
x dx = Σ(-1)
n
x 2k+ Collateral
0 0 k=0 (2k + n +1)!

Proof
 (2k -1)!!  (2k -1)!!(2k)! 2k+1
sinh -1x =Σ(-1)k x 2k+1 =Σ(-1)k x
k=0 (2k)!! (2k +1) k=0 (2k)!! (2k +1)!
2

k (2k -1)!!
= Σ(-1) x 2k+1  (2k)! = (2k)!!(2k -1)!!
k =0 (2k +1)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The collateral higher integral of this was as follows as shown in Formula 4.5.2' ( 4.5 ).

(-1)r x n-2r

x x n/2
 sinh x dx = -1 n
Σ 2 sinh -1x
0 0 r=0 (2r)!! (n -2r)!

n-2r+1
n /2 n -2r+1
r+s (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1+ x 2
r=1 s=0 (s +2r -1)!!
n /2 (-1)r-1 x n-2r+1
+Σ
r=1 (2r -1)!!2(n -2r +1)!
Therefore, the termwise higher integral above is a collateral higher integral.
And matching both, we obtain the following formula.

Formula 6.6.1'
The following expression holds for |x| 1 .

k (2k -1)!!
2
2k+ n+1
n/2 (-1)r x n-2r
Σ
k=0
(-1)
(2k + n +1)!
x =Σ
r=0
2
(2r)!! (n -2r)!
sinh -1x
n-2r+1
n /2 n -2r+1
r+s (s -1)!!2 x
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
1+ x 2
r=1 s=0 (s +2r -1)!!
n /2 (-1)r-1 x n-2r+1
+Σ
r=1 (2r -1)!!2(n -2r +1)!

Substituting x =1 for this, we obtain the following special values.

2
 (2k -1)!!
Σ(-1)k
k=0 2 +2
( k ) !
= sinh -11 - 2 + 1
2

k (2k -1)!! 1 3 2
Σ
k=0
(-1)
(2k +3)!
= sinh -11 -
4 4
+1
2

k (2k -1)!! 1 7 2 7
Σ
k=0
(-1)
(2k +4)!
=-
12
sinh -11 -
36
+
18

- 10 -
 

k (2k -1)!!
2 n/2 (-1)r
Σ
k=0
(-1)
(2k + n +1)!
=Σ
r=0
2
(2r)!! (n -2r)!
sinh -11

n /2 n -2r+1
r+s (s -1)!!2 2
+Σ Σ (-1) n -2r+1Ｃ s 2 (n -2r +1)!
r=1 s=0 (s +2r -1)!!
n /2 (-1)r-1
+Σ
r=1 (2r -1)!!2(n -2r +1)!

- 11 -
6.7 Termwise Higher Integral of arcsech x

Formula 6.7.1
The following expression holds for 0< x <1 .


xn 2

  -Σ 2k (2k +n)! x
x x 2 n 1  (2k -1)!!
-1
sech x dx = n
log +Σ 2k+n
Collateral
0 0 n! x j=1 j k=1

Proof
2  (2k -1)!! 2  (2k -1)!!(2k)! 2k
sech -1x = log -Σ x 2k = log -Σ x
x k=1 (2k)!! 2k x k=1 (2k)!! 2k (2k)!
2
2  (2k -1)!!
= log -Σ x 2k  (2k)! = (2k)!!(2k -1)!!
x k=1 2k (2k)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The collateral higher integral of this was as follows as shown in Formula 4.5.3' ( 4.5 ).

 sech
n-2r-1
x x xn (n -1)/2 (2r -1)!! x
n
Σ
-1 -1
x dx = sech x + sin -1x
0 0 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
+Σ Σ (-1) 2 (n -2r)!
1- x 2
r=1 s=0 2r +s (s +2r -1)!!
n-2r
n /2 1 x
-Σ
r=1 2r(2r -1)!!2 (n -2r)!
Therefore, the termwise higher integral above is a collateral higher integral.
And matching both, we obtain the following formula.

Formula 6.7.1'
The following expression holds for 0< x 1 .
n-2r
2
xn
  + Σ 2r(2r -1)!!
 (2k -1)!! 2 n 1 n /2 1 x
Σ log +Σ
2k+n
x =
k =1 2k (2k +n)! n! x j=1 j r=1
2 (n -2r)!
n-2r-1
xn (n -1)/2 (2r -1)!! x
- sech x - Σ
-1
sin -1x
n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
s n -2rＣ s
(s -1)!!2 x
n /2 n -2r
-Σ Σ (-1) 2 (n -2r)!
1- x 2
r=1 s=0 2r +s (s +2r -1)!!

Substitutingx =1 for this, we obtain the following special values.

 (2k -1)!!
2
1 
Σ
k=1 2k (2k +1)!
=
1!
(log 2+1) -
2
2

 
 (2k -1)!! 1 3 1
Σ
k=1 2k (2k +2)!
=
2!
log 2+
2
+
2
-
2
2

   
 (2k -1)!! 1 11 1 1 1
Σ
k=1 2k (2k +3)!
=
3!
log 2+
6
+
2
-
2
+
2 12

- 12 -
 
2

 
 (2k -1)!! 1 n 1 n /2 1 1
Σ 2k (2k +n)!
=
n!
log 2+Σ
j=1 j
+Σ
(n -2r)!
k=1 r=12r(2r -1)!!2
 (n -1)/2 (2r -1)!! 1
- Σ
2 r=0 (2r)!!(2r +1)! (n -2r -1)!

- 13 -
6.8 Termwise Higher Integral of arccsch x

Formula 6.8.1
The following expression holds for 0< x <1 .


xn 2

  -Σ(-1)
x x 2 n 1 
k (2k -1)!!
-1 n
csch x dx = log +Σ x 2k+n Collateral
0 0 n! x j=1 j k =1 2k (2k +n)!

Proof
2  (2k -1)!! 2k
csch -1x = log +Σ(-1)k-1 x
x k=1 (2k)!! 2k
2  (2k -1)!!(2k)! 2k
= log -Σ(-1)k x
x k=1 (2k)!! 2k (2k)!
2
2 
k (2k -1)!!
= log -Σ(-1) x 2k  (2k)! =(2k)!!(2k -1)!!
x k=1 2k (2k)!
Integrating both sides of this with respect to x from 0 to x repeatedly, we obtain the desired expression.

The collateral higher integral of this was as follows as shown in Formula 4.5.3' ( 4.5 ).

 csch
n-2r-1
x x xn (n -1)/2 (-1)r(2r -1)!! x
n
Σ
-1 -1
x dx = csch x + sinh -1x
0 0 n! r=0 (2r)!!(2r +1)! (n -2r -1)!
n-2r
n /2 n -2r
r+s n -2r Ｃs
(s -1)!!2 x
+Σ Σ (-1) x 2+1
r=1 s=0 2r +s (s +2r -1)!!2 (n -2r)!
n-2r
n /2 (-1)r x
-Σ
r=1 2r(2r -1)!!2 (n -2r)!
Therefore, the termwise higher integral above is a collateral higher integral.

2011.02.15
K. Kono
Alien's Mathematics

- 14 -
7 Super Integral (Non-integer order Integral)

7.1 Super Primitive Function and Super Integral

  Σ , Π ,  , etc . are operators,

m n x x
In Σaj , Πb k ,  f(x)dx n , etc . ,
j=1 k=1 a(n) a(1)
and j , k , etc . are indexes, and [1 , m], [1 , n], etc . are the domains of the index.
As for these index and its domain, a natural number and the set are usually used.
However, the domain may sometimes be extended. For example,
p m
It is Σ
j=0
a = ap which extended the domain of index j of Σ
j=1
a to the real number

interval [0 , p] from the natural number interval [1 , m].

q n
It is Π
k=0
b = b q which extended the domain of index k of Π
k =1
b to the real number

interval [0 , q] from the natural number interval [1 , n].

A fractional and an irrational number are obtained by extending the domain of the index of the operator so
that these two examples may show. It is called analytic continuation to extend the domain generaly.
Although usually analytic continuation is used for extending the domain of a function, it can be used also for
extending the domain of the index of a operator as mentioned above.

7.1.1 Super Primitive Function

Definition 7.1.1
<p>
f (x) obtained by continuing analytically the index of the integration operator of a higher primitive function
<n>
f (x) to a complex plane [0 , p] from a natural number interval [1 , n ] is called Super Primitive Function
f(x) . f p (x)
<>
of may mean the Super Indefinite Integral, and may mean a Super Integration Function.

Example
p
(sin x) p = sin x -   + c (x)
< >
p cp(x) is an arbitrary function.
2

7.1.2 Super Integral

Definition 7.1.2
We call it Super Integral to integrate a function f with respect to an independent variable x from a(0)
to a(p) continuously. And it is described as follows.

     
x x x x
 f(x)dx p =  f(x)dx  dx
a(p) a(0) a(p) a(0)
And
when a(k)= a for all k [0 , p] , we call it super integral with a fixed lower limit ,
when a(k) a for some k [0 , p] , we call it super integral with variable lower limits .

Example

  sin xdx
x x  (-1)r p
p
=Σ x 2r+1+
0 0 r=0 (2r +2+ p)

-1-
 p
  
x x
 sin xdx p = sin x -
p 0 2
2 2

7.1.3 Fundamental Theorem of Super Integral

Continuing analytically the index of the integration operator in Theorem 4.1.3 to a complex plane [0 , p ]
from a natural number interval [1 , n ] , we obtain the following theorem.

Theorem 7.1.3
r [0 , p ] be an continuous function on the closed interval I
<r>
Let f and be arbitrary the r-th order
primitive function of f . And let a(r) be a continuous function on the closed interval [0 , p] .
Then the following expression holds for a(r), x  I .

   
x x p -1 x x
 <p>
f(x)dx p = f (x) - Σ f <p -r>
a(p -r)  dx r (1.1)
a(p) a(0) r=0 a(p) a(p -r+1)
Especially, when a(r)= a for all k [0 , p ] ,

 f(x)dx
x x
<p> p -1
<p -r> (x -a)r
p
=f (x) - Σf (a) (1.2)
a a r=0 (1+ r)

Constant-of-integration Function
p -1
We call Σ
r=0
etc. Constant-of-integration Function of f(x) . Since p is a real number, in order to obtain
p -1
Σ
r=0
etc. generally, the calculus about the calculus operator <r>,(s) is required. This is very difficult.

However, it becomes easy exceptionally at the time of f(x)= e x , and we can obtain the following expression
from (1.2) .

  
p -1 (x -a)r x x  (x -a)r (x -a)r+ p
Σ e a
= e x
-  e xdx p = e aΣ -
r=0 (1+ r) a a r=0 (1+ r) (1+ r + p)

  
 (x -a)r x x  (x -a)r+ p
 e =e Σ
a x
,  e dx = e aΣ
x p
r=0 (1+ r) a a r=0 (1+ r + p)

7.1.4 Lineal and Collateral

Definition 7.1.4

   
x x p -1 x x
<p> <p -r>
f(x)dx = f p
(x) - Σ f a(p -r) dx r (1.1)
a(p) a(0) r=0 a(p) a(p -r+1)

In this expression,
when Constant-of-integration Functin is 0,

 
x x
we call  f(x)dx p Lineal Super Integral and
a(p) a(0)
we call the function equal to this Lineal Super Primitive Function.

when Constant-of-integration Functin is not 0,

-2-
  
x x
we call  f(x)dx p Collateral Super Integral and
a(p) a(0)
we call the function equal to this Collateral Super Primitive Function.

These are the same also in (1.2).

In short , Lineal Super Primitive Function is what integrated f(x) with respect to x continuously without
considering the constant-of-integration function.

Example
Left : collateral integral Right : collateral primitive function

p p
     
x x p -1 x x
sin xdx = sin x - p
- Σ sin (p -r)- dx r
p 0 2 r=0 2 p p-r+1

  e dx   =Σ 
r+ p r+ p
x x  xr x  x
x p
= e -Σ
x
-
0 0 r=0 (1+ r) 1+ r + p  r=0 1+ r + p 
Left : lineal integral Right : lineal primitive function

p
  
x x
 sin xdx p = sin x -
p 0 2
2 2

 e dx
x x
x p
= ex
- -

7.1.5 The Necessary Conditions for the Super Integral being Lineal
In Theorem 7.1.3 (1.1), since the higher integral of 1 can be arbitrary value, in order for Constant-of-integration

= 0 for all r  [0 , p ] .
<r>
Function to be 0, we find out that it must be f a(r)
Since this is important, it is stated here as a theorem.

Theorem 7.1.5

   
x x p -1 x x
 <p>
f(x)dx p = f (x) - Σ f <p -r>
a(p -r)  dx r (1.1)
a(p) a(0) r=0 a(p) a(p -r+1)

 f(x)dx
x x
<p> p -1
<p -r> (x -a)r
p
=f (x) - Σf (a) (1.2)
a a r=0 (1+ r)
The necessary condition for the super integral being lineal is as follows respectively.

for all r  [0 , p]
<r>
f a(r) =0
for all r  [0 , p]
<r>
f (a) =0

That is, the necessary condition for the super integral being lineal is that a(r) or a are zeros of the super
<r>
primitive function f for all continuous r from 0 to p.
In addition, this is not sufficient condition as well as the case of Higher Integral.

Note
Although I was taught to Mr. Sugimoto and knew in 2005, seemingly such integration is the field which is
called "Fractinal Integral" and studied in Europe in recent years. I thought I follow this term in this text.
However, although the number of times of integration is extensible even to a complex plane, wrapping cloth of
"Fractional" is too small. So, in this text, I decided to use "Super Integral" that I was using from before.

-3-
7.2 Fractional Integral

7.2.1 Fractional Integral & Riemann-Liouville Integral

There seems to be Fractional Integral for about 200 years, and the most general form is as follows.


1 x
D -p
f(x) = (x -t)p-1 f(t)dt (a is a zero of the left side ) (2.0)
a x
(p) a
-p
The left side a Dx f(x) is Non-integer order Primitive Function , and this is the same as Super Primitive
<p>
Function f (x) defined in the previous section. Notation aDx-p is called Riemann-Liouville operator.
On the other hand, the integration of the right side is called Riemann-Liouville Integral , and this is equivalent
to Super Integral defined in the previous section.
Although many of super primitive function (non-integer order primitive function) cannot be expressed with an
elementary function, the some can be expressed with elementary functions. In traditional Fractional Integral ,
the super primitive function expressed with these elementary functions are drawn from Riemann-Liouville
integral. However, the method is very difficult. Two examples are shown below.

(1) super primitive function of f(x)= x

Let

 (x -t)p-1
f(t) = t , g(x -t) =
(p)
then

  f(t)g(x -t)dt
<p> 1 x x
x  t (x -t)
p-1
= dt =
(p) 0 0
(f g)(x) . Then we take the Laplace transform of (f g)(x)
We find out that this is a convolution ;

(f g)(x)  F(s)G(s)
( +1)
f(x) = x   
= F(s)
s +1
x
p-1
1 (p) 1
g(x) =  = = G(s)
(p) (p) s p sp
( +1) 1 ( +1) ( +p +1)
 F(s)G(s) = =

s +1 s
p ( +p +1) s +p+1
( +1) a+p
Finally, taking the inverse Laplace transform, we obtain x .
( +p +1)
x
(2) super primitive function of f(x)= e


<p> 1 x
 e x = (x -t)p-1 e tdt
(p) -


 p-1
-
The indefinite integral of the right-hand side is expressed as follows using (p, x)= e d .
x

(x -t) p-1

e tdt = e x(p, x -t)
Then,

-4-
  (x -t)
x <p> 1 x ex
(p, x -t)-
p-1 x
e  = t
e dt =
(p) - (p)
x
e ex
= (p, x - x) - (p, x +) = (p)
(p) (p)
= ex

7.2.2 The demerit and the strong point of Fractional Integral

As seen in two upper examples, Fractional Integral is difficult like this. Calculation of the power function
of the first example is a masterful performance. I do not know whether it will solve by this method also in the
<p>
case of (ax +b ) . Calculationof the exponential function of the 2nd example is by force. In fact,
I borrowed the power of mathematical software for this. Although I also challenged Fractional Integral of a
logarithmic function, in spite of the help of mathematical software, it was too difficult for me. When p is non-
integer, the calculation which obtains the super-primitive function expressed with the elementary function
by Riemann-Liouville integration looks just like the trial which makes infinite decimals with unknown rational
number or irrational number a fraction.
Next, how to take the lower limit of Riemann-Liouville integral in Fractional Integral cannot understand well
to me. Probably, this originates in the fact that (2.0) holds for any lower limit a . However, the greatest cause
will be that there is no concept of Lineal and Collateral in Fractional Integral. Even if it is the usual integral and
is Fractional Integral, I think that original asks for a lineal primitive function.
Finally, since Fractional Integral is dependent on Riemann-Liouville integral, it cannot treat super integral with
a variable lower limit. Specifically, it cannot perform lineal super integral of trigonometric functions or hyperbolic
functions. In order to make these possible, as for us, it is unescapable to stop use of Riemann-Liouville integral.
Although only the demerit of Fractional Integral was mentioned above, there is nothing beyond this for the
super integral with a fixed lower limit. The super integral which is the continued type of the higher order integral
can be expressed with single integral. Numerical integral is possible. These merits and powers are greatest.
If the concept of super integral is taken in and suitable usage is carried out, Riemann-Liouville integral will serve
as a very powerful tool.

7.2.3 Fractional Integral and Super Integral

Theorem 7.2.3
When f(x) denotes a continuously differentiable function and (z) denotes Gamma Function,
the following expression holds for any p >0 .

 
x x 1 x
 f(x)dx p = (x -t)p-1 f(t)dt (2.1)
a a  (p) a

Proof
Analytically continuing the index of the integration operator in Theorem 4.2.3 to [0 , p ] from [1 , n ],
we obtain the desired expression.

Remark
This theorem means that Super Integral with a fixed lower limit is equivalent to Fractional Integral.
However, in Super Integral, instead of Riemann-Liouville integral, the Higher Integral is used to draw the super
primitive function expressed with elementary functions. That is, we ask for the function form of the higher order
primitive function by the higher integral first, and extend the order to the real number from an integer. In Super
Integral, by this way, we can easily obtain the super primitive function which is expressed with elementary

-5-
functions, without the difficult calculation such as previous examples. And in order to verify the result
numerically, Riemann-Liouville integral is used.

Proof of the pudding is in the eating.

In order to show the justification of Theorem 7.2.3, we describe the formulas in the following 3 sections here
in advance.
(1+ ) +p
  x dx  (x -t)
x x 1 x
 p p-1 
= x = t dt
0 0 1+ + p  (p) 0

  e dx (p) 
x x 1 x
x p
= (1)pe  x = (x -t) p-1 t
e dt
  
log x - (1+ p) -  p
  log xdx (p) 
x x 1 x
p p-1
= x = (x -t) log t dt
0 0 (1+ p) 0
Middle is the Super Integral obtained by analytically continuing the index of the operator of the Higher Integral
to real number from natural number, and the right side is Riemann-Liouville Integral. These formulas are inputted
into mathematical software, arbitrary one point is chosen suitably, and the function values are calculatied. The
value of bothe sides are in agreement and it is shown numerically that Super Integral and Riemann-Liouville
Integral are equal.

-6-
7.3 Super Integral of Power Function

7.3.1 Super Integral of Power Function

Analytically continuing the index of the integration operator in Formula 4.3.1 to [0 , p ] from[1 , n ],
we obtain the following formula. In addition, Rieamnn-Liouville integrals are also expressed together

Formula 7.3.1
(1) Basic form
(1+ ) +p
  (x -t)
x x 1 x
x  dx p =
p-1 
 x = t dt (  0)
0 0 1+ + p  (p) 0
(- - p) +p
  (x -t)
x x 1 x
 x  dx p = (-1)p x = p-1 
t dt ( <-p)
  (- ) (p) 
(2) Linear form
(1+ )
 
x p

 
x 1
 (ax +b) dx p = (ax +b)+p (  0)
b b a (1+ +p)
- -
a a


1 x
= (x -t)p-1 (at +b) dt
(p) b
-
a

(- - p)
 
x p

 
x 1

(ax +b) dx = - p
(ax +b)+p ( <-p )
  a (- )

 (x -t)
1 x
= p-1
(at +b) dt
(p) 

Note
When -p   < 0 , we do not define the super-integral of a power function. It is because a lower limit of
the super integral changes irregularly at this time.

Example 1 : When  0
 10 
1
(1+1)
1 11 11
1+ 100
x 
1 10 10 10
= x = x = 0.955579 x
   
1 1
 1+1+ 11
10 10
 10 
9
(1+1)
9 19 19
1+ 100
x 
1 10 10 10
= x = x = 0.547239 x
   
9 9
 1+1+ 171
10 10
2 <e> (1+2) 2+e 2
x  = x = x 2+e = 0.026755 x 4.718281828
1+2+e 3+ e
1 <i> (1+1) 1+ i x 1+
i
x  = x = = (1.200176 - 0.630568 i) x 1+ i
1+1+ i  1+ i !

<1/10> <9/10>
When x 1 , x 1 , x 1 , x 2/2 are drawn on a figure side by side, it is as follows.

-7-
 y
50

40

30

20

10

0
0 1 2 3 4 5 6 7 8 9 10
x
x
100/11*x^(11/10)/gamma(1/10)
100/171*x^(19/10)/gamma(9/10)
1/2*x^2

Example 2 : When  <-p

 2
3 3
5/2 -3/2
2 5 3

   
1 1 - +
= - (5x +4) 2 2
(5x +4)5/2 5 5/2
4 5i 0.0672835
=- =- i
75  (5x +4) 5x +4
-2 <1- i > 2- 1- i  -2+1- i
x  = (-1)1- i x
2
1 i! -11.524427 + 3.585646 i
=- =
(-1) i x 1+ i x 1+
i

 2
1
(1-1/2) -1+ 2
1 1 1
-
x 
-1
= (-1) x 2
= i x 2
1

Example 3 : Outside of a definition (improper)

 2
1

x 
1 1 1

  x
- x x -
2
= 2
dx 2
(-p =  < 0)
0 

=  x
<2> x x
x -1  -1
dx 2 = x(log x -1) (-p <  < 0)
0 1

7.3.2 Half Integral of a power function

Especially, Super Integral of order 1/2 is called Half Integral.

Formula 7.3.2
Let n be a non-negative integer, -1!!  1, (2n -1)!!  135(2n -1) ,
0!!  1, 2n !!  2462n ,
then following expressions hold.

-8-
(1) Basic form

 2
1 1
2(2n)!! n+
x 
n 2
= x
(2n +1)!! 
 2
1

(2n +1)!! 
x 
1
n+
2 n+1
= x
2(2n)!!(n +1)
(2) Linear form
1
 
1 1
2

 
2 1 2(2n)!! n+
(ax +b)  n 2
= (ax +b)
a (2n +1)!! 
 2
1
1
(2n +1)!! 
(ax +b) 
1 2

 
n+ 1
2
= (ax +b)n+1
a 2(2n)!!(n +1)

Proof
From Formula 7.3.1 (2) Linear form,
1
 
1
(1+n)
2 1

 
2 1 n+
(ax +b)  n 2
= (ax +b)
 
a 1
 1+n +
2
 
1
1+ n + 
2
1 1

(ax +b)  =  1a 
1 2
n+
2 2 n+1
(ax +b)
1+n +  + 
1 1
2 2
n +  =
3 (2n +1)!!
Where, since  n+1
,
2 2
(1+n) (1+n) 2n+1 n ! 2(2n)!!
= = =
 (2n +1)!! 
    (2n +1)!!
1 3
 1+n +  n+
2 2

   n + 
1 3
 1+ n +
2 2
=
(n +2)
1+n +  + 
1 1
2 2
(2n +1)!!  n+1 (2n +1)!!  n+1
= x = x
2n+1 (n +1)! 2(2n)!!(n +1)
Substituting these for above expressions, we obtain the linear form. And giving a =1 , b =0 ,
we obtain the basic form.

Example 1

 2
1 1 1
20!! 2
x 
0 2 2
= x = x
1!!  

-9-
  2
1 3 3
22!! 4
x 
1 2 2
= x = x
3!!  3 
 2
1 5 5
24!! 16
x 
2 2 2
= x = x
5!!  15 
 2
1 7 7
26!! 32
x 
3 2 2
= x = x
7!!  35 

Example 2

 2
1

1!!  1 
x 
1
2
= x = x1
20!! 2
 2
1

3!!  2 3  2
x 
3
2
= x = x
22!!2 8
 2
1

5!!  3 15  3
x 
5
2
= x = x
24!!3 48
 2
1

7!!  4 35  4
x 
7
2
= x = x
26!!4 128

7.3.3 Half Integral of an integer power function

Next, using Riemann-Liouville integral, we obtain the Half Integral of an integer power function.

Formula 7.3.3
When n denots a natural number, the following expression holds.

 
1 1
2 n (-1)k n
k   x
2 n+
x 
n
Σ
2
=
 k=0 2k +1

Proof
Let n be a natural number,  = n , p =1/2 . Then since  >-p , Applying the Formula 7.3.1 to these,
we obtain the following expression.

 
1 1

 (x -t) 
2 1 x - 1 x tn
x 
n 2 n
= t dt = dt
(1/2) 0  0 x -t
Where, the following equation is known. ( 岩波数学公式Ⅰ p96 ).

(x -t)n-r

tn

2 x -t n n
dt = Σ (-x)r
x -t (-1)n+1 r=0 r 2n -2r +1
Then

- 10 -
 x

 
(x -t)n-r

tn

x 2 x -t n n
dt = Σ (-x)r
0 x -t (-1)n+1 r=0 r 2n -2r +1
0
1
xn
r 
2 n n
Σ(-1)r
2
= x
(-1)n r=0 2n -2r +1
1
(-1)r-n
r  x
n n n+
= 2Σ 2
r=0 2n -2r +1
 2
1 1
(-1)r-n
r  x
2 n n n+
 x  n
= Σ
2
 r=0 2n -2r +1
Where, we devise further.
r-n
(-1)n-r (-1)k
r    k 
n (-1) n n n n n
Σ
r=0 2n -2r +1
=Σ
r=0 2(n -r)+1 n -r
=Σ
k=0 2k +1
Using this, we obtain

 
1 1
(-1)k
k   x
2 2 n n n+
x 
n
Σ
2
=
 k=0 2k +1

By-product
According to Formula 7.3.2,

 
1 1
2 2(2n)!! n+
x 
n 2
= x
(2n +1)!! 
Since this must be consistent with Formula 7.3.3, then
k

k 
2 n n
(-1) 2(2n)!!
Σ =
 k=0 2k +1 (2n +1)!! 
From this,

(-1)k
k 
n n (2n)!!
Σ
k =0 2k +1
=
(2n +1)!!
(3.2)

If expanded, it is as follows.

0
1 0 0!!
=
1 1!!

1 0 3 1
1 1 1 1 2!!
- =
3!!

1 0 3 1 5 2

1 2 1 2 1 2 4!!
- + =
5!!

Regrettably this was already known.
Although it is a digression, the following equations hold.
k

k 
n (-1) n (1n)! 1
Σ
k =0 1k +1
=
(1n +1)!
=
n +1
(3.1)

- 11 -
 (-1)k
k 
n n (3n)!!!
Σ
k =0 3k +1
=
(3n +1)!!!
!!! means triple factorial. (3.3)

k
Σ mk +1  k 
(-1) n
n (mn)!m
= !m means multi factorial. (3.n)
k =0 (mn +1)!m
Seemingly, it is not known m= 3 or more.

7.3.4 Fractional Integral of an integer power function

Generalizing Formula 7.3.3, we calculate a fractional integral of an integer power function.
First, we prepare the following lemma.

Lemma
When m,n are natural numbers, the following expression holds.
1
m-1
(x -t)n-r

m

r 
- m(x -t) n n
(x -t) m t n dt = n+1 r=0
(- x) r
Σ m(n -r)+1
(-1)

Proof
1
m
(x -t)n-r
r 
m(x -t) n n
F(t) = Σ (-x) r
(-1)n+1 r=0 m(n -r)+1
Differentiate this with respect to t. Then
1
m
(x -t)n-r
r 
d d m(x -t) n n
dt
F(t) =
dt (-1)n+1 Σ
r=0
(-x) r
m(n -r)+1
1
m
(x -t)n-r

m(x -t) d n n
+ Σ (-x)r
(-1)n+1 dt r=0 r m(n -r)+1
1
-1
m
(x -t)n-r
r 
(x -t) n n
=- Σ (-x) r
(-1)n+1 r=0 m(n -r)+1
1
m
(x -t)n-r-1
r 
m(x -t) n n
- Σ(n -r)(-x)r
(-1)n+1 r=0 m(n -r)+1
1
-1
m
(x -t)n-r
r 
(x -t) n n
=- Σ (-x) r
(-1)n+1 r=0 m(n -r)+1
1
-1
m
(x -t)n-r
r 
(x -t) n n
- Σ m(n -r)(-x) r
(-1)n+1 r=0 m(n -r)+1
1
-1
m
(x -t)n-r
r 
(x -t) n n
=
(-1)n
Σ
r=0
1+m(n -r)(-x)
r
m(n -r)+1

- 12 -
 1 1
-1 -1
m m

r 
(x -t) n n (x -t)
= Σ (-x)r(x -t)n-r = (-t)n
(-1)n r=0 (-1)n
m-1
-
m n
= (x -t) t Q.E.D
Using this lemma, we obtain the following formula.

Formula 7.3.4
When m,n are natural numbers, the following expressions hold

 
1 1
n (-1)k

  x
m m n n+

(1/m) Σ
x 
n m
= (4.1)
k=0 mk +1 k
(-1)k (1+ n , 1/m)
k  =
n n
Σ ( ) denots beta function. (4.2)
k =0 mk +1 m

Proof
(1+ ) +p
 (x -t)
<p> 1 x
x  
p-1 
= x = t dt (  0)
1+ +p  (p) 0
Giving  = n , p =1/m ,
 m
1 m-1

 (x -t)
1 x -
x n  = m n
t dt
(1/m) 0
Using above Lemma, we calculate as follows.
x

 
1
m-1 n-r


m

r 
x - m(x -t) n n (x -t)
(x -t) m n
t dt = Σ(-x)r
0 (-1)n+1 r=0 m(n -r)+1
0
1 n

r 
m n n x
=
(-1)n
x m
Σ(-1)r
r=0 m(n -r)+1
1
(-1)r-n
r  x
n n n+
= mΣ m
r=0 m(n -r)+1
Furthermore, since r,n are integer,
r-n
(-1)n-r (-1)k
r    k 
n (-1) n n n n n
Σ
r=0 m(n -r)+1
=Σ
r=0 m(n -r)+1 n -r
=Σ
k=0 mk +1
Using this,
m-1 1

 (x -t)
(-1)k
k   x
x - n n n+
m n
t dt = mΣ m
0 k =0 mk +1
Thus, we obtain

 
1 1
n (-1)k

k   x
m m n n+

(1/m) Σ
x 
n m
= (4.1)
k=0 mk +1
Next,

- 13 -
  m
1
(1+ n) (1+ n)  1/m 
1 1
n+ n+
x 
n m
= x = x m
1+ n +1/m   1/m  1+ n +1/m 
Since this must be equal to (4.1), we obtain
k
(1+ n) 1/m  (1+ n , 1/m)
k 
n (-1)n
Σ
k =0 mk +1
=
m 1+ n +1/m 
=
m
(4.2)

Remark
(4.2) shows that (4.1) can be expressed with a beta function and n can be the real number. Actuality,
(1+ ) 1 (1+ )(p)  1+ , p 
= =
1+ +p  (p) 1+ +p  (p)
Then,
<p> (1+ )   1+ , p  +p
x   = x +p = x ( 0) (4.3)
1+ +p  (p)
Therefore, also,

 (x -t)
x
p-1  
t dt =  1+ , p  x +p ( 0) (4.4)
0

Example 1

 3
1
(1+2)
1 7 7
2+ 2
x  2
= x 3
= x3 = 0.7199 x 3
1+2+1/3 10/3
 3
1 7
2 (-1)k

k   x
3 2
x 
(1/3) Σ
2 3
=
k =0 3k +1
7 7

(1/3)  1 0 4 1 7 2  

3 1 2 1 2 1 2
= - + x 3
= 0.7199 x 3

 4
1 13
3 (-1)k

k   x
4 3
x 
(1/4) Σ
3 4
=
k =0 4k +1
13 13

     3   x
4 1 3 1 3 1 3 1 3 4 4
= - + - = 0.7241 x
(1/4) 1 0 5 1 9 2 13

Example 2

0 1  2
1 2 1 2 1 2 B(1+2 , 1/3) 9
- + = =
1 4 7 3 14

   
1 9 1 9 1 9 1 9 B(1+9, 1/3) 1, 594, 323
- + -+  - = =
1 0 4 1 7 2 28 9 3 3, 803, 800

   
3 1 3 1 3 1 3 B(1+ 3, 1/4) 128
- + - = =
0 5 1 9 2 13 3 4 195

7.3.5 Super Integral of an integer power function

Replacing 1/m with p , we obtain the following formula.

- 14 -
Formula 7.3.5
When n is a natural number, the following expressions hold for p >0 .
n (-1)k


n <p> 1 n
x  = Σ  x n+ p (5.1)
(p) k=0 p + k k
n -1 (-1)k

 
n -1
(n , p) = Σ ( )denots beta function. (5.2)
k =0 p + k k
Example 1
<e> (1+2) 2
x 2 
e e
= x 2+ = x 2+ = 0.026755 x 2+ e
1+2+ e 3+ e 
1 2 (-1)k 2
2 <e>
x  = Σ
(e) k=0 e + k k
 x 2+ e

   2  
1 1 2 1 2 1 2 e e
= - + x 2+ = 0.026755 x 2+
(e) e 0 e +1 1 e +2

Example 2

 
1 1 1 11 1
(2 , e) = -- = 0.098938
=
0 e 1 e e +1
e +1

  
1 2 1 2 1 2
(3 , ) = - + = 0.029896
 0  +1 1  +2 2

7.3.6 Super Integral of an positive power function

Since Formula 7.3.5 are binomial forms, the further generalization is possible.

Formula 7.3.6
The following expressions hold for positive numbers p ,q .
q+p r


q <p> x  (-1) q
(p) Σ
x  = (6.1)
r=0 p + r r
r

 
 (-1) q -1
(q , p) = Σ ( )is Beta Function (6.2)
r=0 p + r r

7.3.7 Super Integral of a polynomial

m
m-k
In the case of a polynomial f(x) =Σcm-k x , as for the zero of the super primitive function, it is good
k=0
to perform it as follows. It is based on experience of a writer.

(1) When f(x) is factored by primary formula ax +b , let -b /a be the zero.

(2) When f(x) is not factored by primary formula ax +b , let 0 be the zero.

f(x) = x 2-2 x + , the following calculation is

2
For example, in the case of the 1/2 times integral of
right in many case.

 2  2
1 1 5
16
x 2-2 x +  = (x -)  (x -)
2 2 2
=
15 

- 15 -
 If this is calculated termwise as follows, the result is different from the former.

 
1
     
1 1 1
2 2 2 2
x -2 x +  - 2x  +  x 
2 2
2
= x  2 1 0

8 2 4
1

 
2
x +
2 2
= x x -
 15 3
Needless to say, this cause is the difference between
1 1

   
x x x x
 x -2 x + dx and
2 2 2
 x -2 x + dx 2 .
2 2
  0 0
That is, it is because the latter regarded it as 0 although the former regarded the zero of the super primitive
function as pi.
The latter is right when there is a special reason why the zero of the super primitive function should be 0.
However, such a case is rare, and in almost all cases the former is right.

- 16 -
7.4 Super Integral of Exponential Function

7.4.1 Super Integral of Exponential Function

Analytically continuing the index of the integration operator in Formula 4.3.2 to [0 , p ] from[1 , n ],
we obtain the following formula. In addition, Rieamnn-Liouville integrals are also expressed together

Formula 7.4.1
(1) Basic form

   (x -t)
x x 1 x
x
 e dx p = (1)pe  x = p-1
e dt
t
  (p) 
(2) Linear form

   (x -t)
x p

 
x 1 1 x
 e
ax+b
dx p
= e
ax+b
= p-1
e
at+b
dt
  a (p) 
a >0 : - , a <0 : +
(3) General form

   (x -t)
x p

 
x 1 1 x
  ax+bdx p =  ax+b = p-1
 at+bdt
  a log (p) 
a >0 : - , a <0 : +

Proof of the general form

Let c =a log , d = b log , then

= e xa log+b log = e axloge blog = e log e log =  ax b = 

cx+d ax b ax+b
e
Applying this to (2) Linear form, we obtain (3) General form immediately.

Example

 2
1 1
e  -x
= (-1) e 2 -x
= i e -x
2
3x-4  2
 
1
e  = e 3x-4 = 0.211469 e 3x-4
3
 2
1
1/2

 
1
3 
x
= 3x = 0.954064  3x
log 3
1
 
1
2

 
2 1
(-3)x = (-3)x
log(-3)
= (0.447018 - 0.317241 i)  (-3)x
1 i x
2  x <i>
=
 log 2 
2 = (0.933582 + 0.358362 i)  2x

<1/2> <1>
When 3x , 3x , 3x are drawn on a figure side by side, it is as follows.

- 17 -
 y

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
x
3^x/ln(3)
3^x*(1/ln(3))^(1/2)
3^x

7.4.2 Super Integral of Exponential Function (hyperbolic function form)

Also as follows, the super integral of an exponential function is expressed using a hyperbolic function.

Formula 7.4.2
(1) Basic form
p i p i
      
x x
x
 e dx p = i p cosh x -  sinh x -
  2 2
= (cosh x)<p>  (sinh x)<p>
(2) Linear form
p i p i
  e  
p p

 
x

 
x i
ax+b
dx p
= cosh ax +b -  sinh ax +b -
  a 2 2
= cosh(ax +b)<p>  cosh(ax +b)<p>

Proof
From the Formula 7.7.1 mentioned later, the following expressions hold.
p i
 
1 x
(cosh x) p = i p cosh x - e + (-1)pe -x
< >
=
2 2
p i
 
1
(sinh x) p
<>
= i p sinh x - = e x - (-1)pe -x
2 2
Adding and subtracting these two formulas, we obtain the following expressions.
p i p i

i p cosh x -
2   + sinh x -
2  =e x

p i p i
ip  cosh x -
2   - sinh x -
2    = (-1) e p -x

Substituting these for the Formula 7.4.1, we obtain (1) basic form.
In a similar way, (2) linea form is obtained

Note
Therefore, in the concept, the following expression holds.

- 18 -
      
x x x x x x
x
 e dx p =  cosh xdx p   sinh xdx p
  (p -1) i -1 i p i 0 i
2a 2a 2 2
Actually, when p is a natural number n, both sides recover the function of operation immediately and the result
in the higher integral as follows..

  e    
x x x x x x
x
dx =n
 cosh xdx 
n n
sinh xdx
  (n -1) i 0 i n i 1 i
2a 2a 2 2

- 19 -
7.5 Super Integral of Logarithmic Function
Analytically continuing the index of the integration operator in Formula 4.3.3 to [0 , p ] from[1 , n ],
we obtain the following formula. In addition, Rieamnn-Liouville integrals are also expressed together

Formula 7.5.1
(1) Basic form
log x - (1+p) -  p
   (x -t)
x x 1 x
 log xdx p = x =
p-1
log t dt
0 0 (1+p) (p) 0
(2) Linear form
log(ax +b)- (1+ p) - 
 
x p

 
x b
p
log(ax +b)dx = x+
b b (1+ p) a
- -
a a


1 x
= (x -t)p-1 log(at +b)dt
(p) b
-
a

Proof
Formula 4.3.3 in "04 Higher Integral" was as follows.

  log x dx
x xn
 
x n 1
 n
= log x -Σ
0 0 n! k=1 k

  log(ax +b)dx
x n

  
x 1 b n 1
n
= x+ log(ax +b) -Σ
b b n! a k=1 k
- -
a a
On the other hand, in "01 Gamma Function & Digamma Function" there were the following formulas.
n 1
n ! = (1+ n), Σ =  (1+n) + 
k =1 k
Then, subsutituting these for the upper formulas and analytically continuing the index of the integration operator
to [0 , p ] from[1 , n ], we obtain the desired expressions.

Example

log x - (1+1) -  1
(log x)<1> = x = x(log x -1)
(1+1)

 
1 1
log(3x +4) - 1+ -
 2
1
4 2
 
2
log(3x +4) = x+
 
1 3
 1+
2
= 1.1283(log x - 0.6137) x +1.3333

 
1
log x - 1+ -
 10 
1 1
10
(log x) = x 10
= 1.0511 10 x (log x -0.1534)
 
1
 1+
10

- 20 -
  
9
log x - 1+ -
 =
9 9 9
10 10 10 10
(log x) x = 1.0397 x (log x -0.9333)
 
9
 1+
10

When log x , (log x)<1/10> , (log x)<9/10> , (log x)<1> are drawn on a figure side by side,
<1/10>
it is as follows. We can see slightly that the zero of (log x) (red) is also x =0 .

y
6

0
1 2 3 4 5 6 7
x
-2

ln(x)
-10*x^(1/10)/gamma(1/10)*(EULER - ln(x) + psi(1/10
-10/9*x^(9/10)/gamma(9/10)*(EULER - ln(x) + psi(9/
x*(ln(x) - 1)

- 21 -
7.6 Super Integral of Trigonometric Function

7.6.1 Super Integrals of sin x , cos x

Analytically continuing the index of the integration operator in Formula 4.3.4 to [0 , p ] from[1 , n ],
we obtain the following formula.

Formula 7.6.1
(1) Basic form
p
   
x x
 sin xdx p = sin x -
p 0 2
2 2
p
   
x x
 cosxdx p = cos x -
(p -1) -1 2
2 2
(2) Linear form
p
 
p

   
x x 1
 sin(ax +b)dx p = sin ax +b -
p b 0 b a 2
- -
2a a 2a a

p
 
p

   
x x 1
 cos(ax +b)dx p = cos ax +b -
(p -1) b 1 b a 2
- - -
2a a 2a a

Example

 
1
1  
(sin x ) 2 
= sin x -
2 2  = sin x -  4 
      
1 1 1
((sin x ) 2 ) 2 = (sin (x - )) 2
1
4
2
 1  
    = sin x - 2  = -cos x
1
= sin x - -
1 4 2 2
 <1> 1
 1
       = sin x
1
sin x + = sin x + -
2 1 2 2
 
1
1  
(cos x) 2 
= cos x -
2 2  = cos x -  4 
  2 
2
(cos x)  
= cos x -
 2  = cos(x -1)
When cos x , (cos x)1/2 , sin x are drawn on a figure side by side, it is as follows. Red shows 1/2
order super integral. It is clear also in the figure that super integral which is the easiest to understand is super
integral of trigonometric functions.

- 22 -
7.6.2 Termwise Super Integrals of sin x , cos x
If common lower limit 0 is employed as the lower limits of the super integral, we obtain the following termwise
super integral. These are collateral super integrals as understood from the constant-of-integration function in
the right side

p
 sinxdx
(-1)k
  + C(p, x)
x x 
2k+1+ p
p
=Σ x = sin x -
0 0 k=0 (2k +2+ p) 2
k
p-k
p x
C(p, x) = Σ sin
k =1 (1+p -k) 2
p
 cosxdx (-1)k
  + C(p, x)
x x 
p
p
=Σ x 2k+ = cos x -
0 0 k =0 (2k +1+ p) 2
k
p-k
p x
C(p, x) = Σ cos
k =1 (1+p -k) 2

When the 1/2th order collateral super integral of cos x is drawn as cos x and sin x side by side, it is as
follows. Red shows the 1/2th order collateral super integral.

- 23 -
7.7 Super Integral of Hyperbolic Function

7.7.1 Super Integrals of sinh x , cosh x

Analytically continuing the index of the integration operator in Formula 4.3.5 to [0 , p ] from [1 , n ],
we obtain the following formula.

Formula 7.7.1
(1) Basic form
p i
 
e x - (-1)pe -x
 
x x
p p
sinh xdx = i sinh x - =
p i 0 i 2 2
2 2

p i
   
e x + (-1)pe -x
x x
 cosh xdx p p
= i cosh x - =
(p -1) i -1 i 2 2
2 2
(2) Linear form
p i
 
p

 
x

 
x i
 sinh(ax +b)dx p
= sinh ax +b -
p i b 0 i b a 2
- -
2a a 2a a
p

 
1 1
e - (-1)pe -(ax+b)
ax+b
=
2 a
p i
 
p

 
x

 
x i
 cosh(ax +b)dx p = cosh ax +b -
(p -1) i b -1 i b a 2
- -
2a a 2a a
p

  e
1 1
= ax+b
+ (-1)pe -(ax+b)
2 a

Example 1

  1 i i
1 1 1
(sinh x ) 2 2
= i sinh x -
2 2   2
= i sinh x -  4 
    i  2
1 1 1 1
((sinh x ) 2 ) 2 = (i 2 sinh (x - ))
4
1
i 1 i
1 2

   4 
2 i
=i sinh x - -
1 2 2
i
= i sinh x -
2
= cosh x

i <1> 1
i 1 i
       = i cosh x
i
cosh x + = cosh x + -
2 1 2 2
 
i i
 
(cosh x) 10   = 0.854635  cosh x +  
10
=i cosh x +
20 20

- 25 -
  
9i 9i
9 9
(cosh x) 10 =i 10
cosh x +  20  = 0.243237  cosh x +  20 
Super integral of hyperbolic function is the most incomprehensible in super integrals. The reason is that the
super primitive function turns into a complex function except order p is an integer or a purely imaginary number.
i 9i

Then, when
   
cosh x , (cosh x) 10 , (cosh x) 10 , sinh x which can be displayed on a real number
domain are drawn on a figure side by side, it is as follows.

i
 10 
All of four curves have overlapped in the positive area. It is natural that (cosh x) is near cosh x in
9i
 10  is far apart from sinh x
a negative area. But (cosh x) in why.

Example 2
1

 
1 x
2 e - (-1) e 2 -x
e x - i e -x
(sinh x ) = =
2 2
 2
1

 2
1
   
1 1

 
e x - i e -x ex i
((sinh x ) 2 ) 2 = = -  e -x
2 2 2
ex i e x + e -x
= - i e =
-x
= cosh x
2 2 2
i i
i
  
<1>
1-1 x+
  
- x+
2 1 2
cosh x + = e + (-1) e
2 2
i i
=
1
2
 -
e 2 e x - e 2 e -x =
1
2
1
i e x - e -x
i
  
i
=  e x + e -x = i cosh x
2
i i
i  -x
  e x - (-1)  e -x
i
e x - e  e e x - e -1-x
(sinh x)  = = =
2 2 2

- 26 -
Note
Super integral of a hyperbolic function is the integral with variable lower limits, and super integral of an
exponential function is the integral with a fixed lower limit. So, for example, the p-th order super integral
of sinh x is as follows in the concept.

   
<p>
x x
p
x x e x - e -x p
(sinh x) = sinh xdx = dx
p i 0 i p i 0 i 2
2 2 2 2

But the example forgets such a thing and perform super integral of e x , e -x with lower limits - ,  .
The result is all right. Truly, this is convenience and interesting.

7.7.2 Termwise Super Integrals of sinh x , cosh x

If common lower limit 0 is employed as the lower limits of the super integral, we obtain the following termwise
super integral. These are collateral super integrals as understood from the constant-of-integration function in the
right side

p i
 sinhxdx
p
x 2k+1+
  + C(p, x)
x x 
p
=Σ p
= i sinh x -
0 0 k=0 (2k +2+ p) 2
k i
p-k
p x
C(p, x) = Σ sinh
k =1 (1+p -k) 2
p i
 coshxdx
p
x 2k+
  + C(p, x)
x x 
p
=Σ = i p cohs x -
0 0 k=0 (2k +1+ p) 2
k i
p-k
p x
C(p, x) = Σ cosh
k =1 (1+p -k) 2

When the values of the lineal and the collateral the 1.2th order integral of cosh x on x=1, x=8 are calculated
respectively, it is as follows.

Though the difference of both is large where x is small, both are almost corresponding ing where x is large.
Since this is similar also in sinh x, it is thought that the termwise super integrals of sinh x and cosh x are
asymptotic expansions of the lineal super integrals.

- 27 -
7.8 Super Integral of Inv-Trigonometric Function etc.

7.8.1 Super Integrals of arctan x , arccot x

Analytically continuing the index of the integration operator in Formula 4.3.6 to [0 , p ] from[1 , n ],
we obtain the following formula.

Formula 7.8.1
When (x),(x) denote gamma function and digamma function respectively, the following expressions
hold for x 1 .


tan -1x 
 
x x p
 tan x dx =-1 p
Σ ( -1) k
x
p-2k
0 0 (1+ p) k =0 p -2k
log1+ x 2 
 
p p+1-2k
+ Σ
2(1+ p) k =1
(-1)k
p +1-2k
x

 
1  p
 (1+ p )- (2r)x
p+1-2r
- Σ
(1+ p) r=1
(-1)r
p +1-2r


xp tan -1x 
 
x x p
 cot x dx =
-1 p -1
cot x - Σ (-1)k x
p-2k
0 0 ( p)
1+ ( p) k =1
1+ p -2k
log1+ x  
2

 
p p+1-2k
- Σ
2(1+ p) k =1
(-1)k
p +1-2k
x

 
1  p
 (1+ p )- (2r)x
p+1-2r
+ Σ
(1+ p) r=1
(-1) r
p +1-2r

-1
Example: The 5/2 th order integral of cot x

7.8.2 Super Integrals of arctanh x , arccoth x

Analytically continuing the index of the integration operator in Formula 4.3.7 to [0 , p ] from[1 , n ],
we obtain the following formula.

- 28 -
Formula 7.8.2
When (x),(x) denote gamma function and digamma function respectively, the following expression
holds for x 1 .


tanh -1x 
 
x x p p-2k
p
Σ
-1
tanh x dx = x
0 0 (1+ p) k =0 p -2k
log1- x 2 
 
p p+1-2k
+ Σ
2(1+ p) k =1 p +1-2k
x

 
1  p
 (1+ p )- (2r)x
p+1-2r
(1+ p) Σ
-
r=1 p +1-2 r

-1
Example: The 3/2 th order integral of tanh x

2010.07.07
K. Kono
Alien's Mathematics

- 29 -
Examples
Below, the examples of the termwise super integral in Formula 8.1.1 are shown. One arbitrary point is chosen
suitably. fl is the the function value on the point by the formula and fr is the function value on the same point
by Riemann-Liouville Integral. All digits are corresponding to both, and this shows the justification of the above
termwise super integrals numerically. In the figure, blue shows the function to be integrated, redshows the
termwise super integral, green shows the 1st order integral.

-2-
Formula 8.1.2
When B2k , E2k , (x) denote Bernoulli Numbers, Euler Numbers, Gamma Function respectively,

the following expressions hold for any p  0 and any  /2< x <  .

-3-
9 Higher Derivative

9.1 Higher Derivative and Higher Differentiation

9.1.1 Higher Derivative

Definition 9.1.1
(x) denotes the derivative function of f (n -1)(x) for n =1, 2, 3,  ,
(n)
When f
(n)
we call f (x) ,
or for short , .

9.1.2 Higher Differentiation

Definition 9.1.2
We call it to differentiate a function f with respect to an independent variable x
repeatedly. And it is described as follows.
n

      
d d d d d d
f(x) =  f(x)  : n pieces
dx n dx dx dx dx dx

9.1.3 Fundamental Theorem of Higher Differentiation

The following theorem holds from Theorem 4.1.3 in 4.1 .

Theorem 9.1.3
r =0, 1,  , n
(r )
When f are continuous functions on a closed interval I and are the r th derivative
functions of f, the following expression holds for x I.
dn
) = f n (x)
()
n f(x
(1.1)
dx

Proof
Theorem 4.1.3 in 4.1 can be rewritten as follows.
x x n -1 x x
<n>
f (x) =  n
f(x)dx + <n -r>
f an-r  dx r
an a1 r=0 an a n- r+ 1
Differentiating both sides with respect to x n times, we obtain

d n <n> d n n -1 <n -r> x x

nf
<0>
(x) = f (x) + n r=0 f an-r  dx r
dx dx an a n- r+ 1

Here, since the constant-of-integration polynomial of the right side is degree n -1 , if this is differentiated n
times, it must become 0. Then we obtain

d n <n>
nf
(x) = f<0>(x)
dx
Shifting by -n the index in the integration operator <> and replacing <> by differentiation operator ( ) ,
we obtain the desired expression.

-1-
Remark
Since differentiation is an inverse operation of integration, it cannot be unrelated to the constant-of-integration.
In fact, the lineal and the collateral exist in Super Differentiation (non-integer times differentiation). (See later
12.1.2.) But this thorem guarantees that only the lineal exists in Higher Differentiation.

9.1.4 The basic formulas of Higher Differentiation

The following formulas hold like the 1st order differentiation.
(n) (n)
c f(x) =cf (x) c 0 : constant multiple rule
(n)
 f(x)+ g(x) = f(n)(x) + g(n)(x) : sum rule
n
(n) r
() n
f(x)g (x) = n r f (x) g ( -r)(x) : product rule (Leibniz rule)
r=0

-2-
9.2 Higher Derivative of Elementary Functions

Formula 9.2.1 Higher Derivative of a power function

When  (z) denots the zeta function, the following expressions hold.
(1) Basic form
(n)  (1+) -n
x  = x   0
 1+ - n 
 (- + n) -n
= (-1)-n x ( < 0)
 (-)
(2) Linear form
 (1+)
-n

 
1
(ax +b)   (n)
= (ax +b)-n   0
a  (1+ -n)
1 -n  (- + n)
= -  
a  (-)
(ax +b)-n ( < 0)

Proof
-1

 
1
(ax +b) (ax +b)-1
(1)
=
a
-2

 
1
(ax +b)   (2)
= ( -1)(ax +b)-2
a

-n

 
 (n) 1
(ax +b)  = ( -1) -(n -1)x -n
a
 (1+)
Substituting ( -1) -(n -1) = for this, we obtain (2).
 (1+ -n)
Furthermore, substituting a =1, b =0 for this, we obtain (1).

However, when  is a negative integer, a denominator and a numerator become the infinite form of infinity,
and as for this right side, the value can not be calculated.

Therfore, when  <0 let  =-. . Then, since (ax +b) = (ax +b)- ,
-1

 
 1
-(ax +b)
(1)
(ax +b)-  = -1
a
-2

 
- (2) 1
(ax +b)  = +( +1)(ax +b) -2
a

-n

 
- (n) 1
(ax +b)  = (-1)n( +1)( +n -1)(ax +b) -n
a
Here
 ( + n)
( +1)( +n -1) =
 ()
we can substitute this for the above as follows.

-3-
 -n
 ( + n)
 
- (n) 1
(ax +b)  = (-1)n (ax +b) -n
a  ()
Then replacing  with - we botain (2), furthermore, substituting a =1, b =0 for this, we obtain (1).

Formula 9.2.2 : Higher Derivative of Exponential Functions

(1) Basic form
 x (n)
e  = ( 1)-ne  x
(2) Linear form
-n

 
ax+b (n) 1
e 
ax+b
= e
a

Proof
According to Formula 4.3.2 in 4.3 , it was as follows.

  e
x x
x n x
n
dx = ( 1) e

 

n

  e  a <0 : +
x a >0 : -
 
x 1
ax+b n ax+b
dx = e

 
 a
Since the differentiation is a reverse-operation of integration, replacing the index n of the integration operator
with -n , we obtain the desired expression.

Formula 9.2.3 : Higher Derivative of Logarithmic Functions

(1) Basic form
(log x) n = (-1)n-1(n -1)! x -n
()

(2) Linear form

-n

 
(n) n-1 b
log (ax +b ) = (-1) (n -1)! x +
a

Proof
Differentiating log x with respect to x one by one, we obtain the desired expression. As for (2), it is similar.

Formula 9.2.4 : Higher Derivatives of sin x, cos x

(1) Basic form
n
(sin x) n  2
()
= sin x +

n
= cos  x +
2 
(cosx) n
()

(2) Linear form

-n
n
   
(n) 1
sin(ax +b ) = sin ax +b +
a 2
-n
n
   
(n) 1
cos(ax +b ) = cos ax +b +
a 2

-4-
Proof
Replacing the index n of the integration operator with -n in Formula 4.3.4 (4.3 ) , we obtain the desired
expressions.

Formula 9.2.5 : Higher Derivatives of sinh x, cosh x

(1) Basic form
n i e x - (-1)-ne -x
 
(n) -n
(sinh x) =i sinh x + =
2 2
n i e x + (-1)-ne -x
 
(n) -n
(cosh x) =i cosh x + =
2 2
(2) Linear form
-n
n i
   
(n) i
sinh (ax +b ) = sinh ax +b +
a 2
-n

 
1 1
e - (-1)-ne -(ax+b)
ax+b
=
2 a
-n
n i
   
(n) i
cosh(ax +b ) = cosh ax +b +
a 2
-n

 
1 1
e + (-1)-ne -(ax+b)
ax+b
=
2 a

Proof
Replacing the index n of the integration operator with -n in Formula 4.3.5 (4.3 ) , we obtain the desired
expressions.

Formula 9.2.6 : Higher Derivatives of tan x, cot x

When  denots ceiling function, the following expressions hold for a natural number n.
n /2
(tan x) n = n+1-2r
()
n T r(tan x)
r=0
n /2
(cotx) n = (-1)n n+1-2r
()
n T r(cotx)
r=0
where nTr are coefficients as follows.

0T 0 1
1T 0 1T 1 1 1
2T 0 2T 1 2 2
3T 0 3T 1 3T 2 = 6 8 2
4T 0 4T 1 4T 2 24 40 16
5T 0 5T 1 5T 2 5T 3 120 240 136 16
6T 0 6T 1 6T 2 6T 3 720 1680 1232 272
 

-5-
And these coefficients are obtained by the following sequential computation

0! Calculating formula
1
1! 1 1=0! 1
2 0
2! 2 2=1! 2+1 0
3 1
3! 8 2 8=2! 3+2 1 , 2=2 1
4 2 0
4! 16 40 40=3! 4+8 2 , 16=8 2+2 0
5 1 3
5! 240 136 16 240=4! 5+40 3 , 136=40 3+16 1 , 16=16 1
6 4 2 0
6! 1680 1232 272 1680=5! 6+240 4 , 1232=240 4+136 2 ,
272=136 2+16 0
 

Note
When T1=1 , T3=2 , T5=16 , T7=272 , T9=7936 ,  , T2n-1 ,  are tangent numbers,
there is the following relation between these and the above coefficients.

T2n-1 = 2n-1T n n =1, 2, 3, 

Formula 9.2.7 : Higher Derivatives of tanh x, coth x

When  denots ceiling function, the following expressions hold for a natural number n.
n /2
(tanh x) n = (-1)n (-1)r nTr(tanh x)n+1-2r
( )
r=0
n /2
(coth x) n = (-1)n (-1)r nTr(coth x)n+1-2r
()
r=0

where nTr are same as ones in Formula 9.2.6 .

Formula 9.2.8 : Higher Derivatives of sec x, csc x

When  denots floor function, the following expressions hold for a natural number n.
1 n/2
(secx) n = n-2r
()
n E r(tan x)
cosx r=0
(-1)n n/2
(cscx) n n-2r
()
= n E r(cotx)
sin x r=0
where nEr are coefficients as follows.

-6-
 0E 0 1
1E 0 1
2E 0 2E 1 2 1
3E 0 3E 1 = 6 5
4E 0 4E 1 4E 2 24 28 5
5E 0 5E 1 5E 2 120 180 61
6E 0 6E 1 6E 2 6E 3 720 1320 662 61
7E 0 7E 1 7E 2 7E 3 5040 10920 7266 1385
 

And these coefficients are obtained by the following sequential computation

1!
1
2! 1 1=1! 1
2 1
3! 5 5=2! 2+1 1
3 2 1
4! 28 5 28=3! 3+5 2 , 5=5 1
4 3 2 1
5! 180 61 180=4! 4+28 3, 61=28 2+5 1
5 4 3 2 1
6! 1320 662 61 1320=5! 5+180 4 , 662=180 3+61 2 , 61=61 1
6 5 4 3 2 1
7! 10920 7266 1385 10920=6! 6+1320 5 , 7266=1320 4+662 3 ,
  1385=662 2+61 1

Note
When E0 =1 , E2 =-1 , E4 =5 , E6 =-61 , E8 =1385 ,  , E2n ,  are Euler numbers,
there is the following relation between these and the above coefficients.

E2n = (-1)n2nEn n =0, 1, 2, 

Formula 9.2.9 : Higher Derivatives of sech x, csch x

When  denots floor function, the following expressions hold for a natural number n.
(-1)n n/2
(sech x) n = (-1)r nE r(tanh x)n-2r
( )
cosh x r=0
(-1)n n/2
(csch x) n (-1)r nE r(coth x)n-2r
( )
=
sinh x r=0
where nEr are same as ones in Formula 9.2.8 .

-7-
9.3 Higher Derivative of Inverse Trigonometric Functions

Formula 9.3.1 : Higher Derivatives of arctan x , arccot x

When  denots ceiling function, the following expressions hold for a natural number n.
(n) (n -1)! n /2
tan
-1
x = (-1)n n r=1
(-1)rn>An+1-2r x n+1-2r
x +1
2

(n) (n -1)! n /2

cot x
-1
= (-1)n-1 n r=1
(-1)rn>An+1-2r x n+1-2r
x +1
2

Proof
Differentiating arctan x with respect to x one by one, it is as follows.
(2) 1! 2x 1!2>A1 x
tan x
-1
=- 2
=- 2
x +1 x +1
2 2

(3) 2!3x 2-1 2!3>A2 x 2 - 3>A0

tan x
-1
= 3
= 3
x +1 x +1
2 2

(4) 3!4x 3-4x 3!4>A3 x 3- 4>A1 x

tan x
-1
=- 4
=- 4
x +1 x +1
2 2

(5) 4!5x 4-10x 2+1 4! 6>A4 x 4 - 5>A2 x 2 + 5>A0

tan x
-1
= 5
= 5
x +1 x +1
2 2

(6) 5!6x 5-20x 3+6x 5! 6>A5 x 5 - 6>A3 x 3 + 6>A1 x

tan x
-1
=- 6
=- 6
x +1 x +1
2 2


Hereafter by induction, we obtain the desired expression.

c.f.
According to < X ¶ ’ Ł p39,40, the following expressions hold for a natural number n.

 
(n )
tan x = (n -1)!cos n(tan -1 x )sin n tan -1 x +
-1
2
(n )
cot x
-1
= (-1)n(n -1)! sin n(cot -1 x )sinn cot -1 x
Therefore, combining with Formula 9.3.1, these formulas lead to the following equations.
n /2 n
(-1)k n>An+1-2k x n+1-2k = -x 2+1 sin n(cot -1 x )sinn cot -1 x
k =1
n
n /2 n
(-1) n>An+1-2k
k
= -2 sin 2
k =1 4

Example
5
-5>A4 x 4+ 5>A2 x 2- 5>A0 x 0 = -x 2+1 sin 5(cot -1 x )sin5 cot -1 x
5
5
-5>A4 + 5>A2 - 5>A0 = -2 sin 2
=4
4

-8-
Formula 9.3.2 : Higher Derivatives of arcsin x, arccos x
When  denots floor function, the following expressions hold for a natural number n.
(2r -1)!!(2n -3-2r)!!x n-1-2r
 
(n ) n/2 n -1
sin x
-1
= 1
r=0 n -1-2r n-r-
2
1- x 
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n ) n/2 n -1
cos x
-1
=- 1
r=0 n -1-2r n-r-
2
1- x 
2

Proof
Differentiating arcsin x with respect to x one by one, it is as follows.
(1) 1
sin x
-1
= 1
2
1- x 
2

(2) x
sin x
-1
= 3
2
1- x 
2

(3) 3x 2 1
sin x
-1
= 5
+ 3
2 2
1- x  1- x 
2 2

(4) 15x 3 9x
sin x
-1
= 7
+ 5
2 2
1- x  1- x 
2 2

4
(5) 105x 90x 2 9
sin x
-1
= 9
+ 7
+ 5
2 2 2
x +1 x +1 x +1
2 2 2

(6) 945x 5 1050x 3 225x

sin x
-1
= 11
+ 9
+ 7
2 2 2
x +1 x +1 x +1
2 2 2


These coefficients are expressed as follows using binomial coefficients and double factorials.

1 0>A0(-1)!!(-1)!!
2 1>A1 (-1)!!1!!

3 2>A2 (-1)!!3!! 2>A0 1!!1!!

4 3>A3 (-1)!!5!! 3>A1 1!!3!!

5 4>A4 (-1)!!7!! 4>A2 1!!5!! 4>A03!!3!!

6 5>A5 (-1)!!9!! 5>A3 1!!7!! 5>A1 3!!5!!


Substituting these for the above,

-9-
 (1) >A0(-1)!!(-1)!!x 0
0
sin x
-1
= 1
1-
2
x +1
2

(2) >A1(-1)!!1!!x 1
1
sin x
-1
= 1
2-
2
x +1
2

(3) >A2(-1)!!3!!x 2
2 >A01!!1!!x 0
2
sin x
-1
= 1
+ 1
3- 2-
2 2
x +1 x +1
2 2

(4) >A3(-1)!!5!!x 3
3 >A11!!3!!x 1
3
sin x
-1
= 1
+ 1
4- 3-
2 2
x +1 x +1
2 2

(5) >A4(-1)!!7!!x 4
4 >A21!!5!!x 2
4 >A03!!3!!x 0
4
sin x
-1
= 1
+ 1
+ 1
5- 4- 3-
2 2 2
x +1 x +1 x +1
2 2 2

(6) >A5(-1)!!9!!x
5
5
>A31!!7!!x
5
3
>A13!!5!!x 1
5
sin x
-1
= 1
+ 1
+ 1
6- 5- 4-
2 2 2
x +1 x +1 x +1
2 2 2


(2r -1)!!(2n -3-2r)!!x n-1-2r
 
(n ) n/2 n -1
sin x
-1
= 1
r=0 n -1-2r n-r-
2
1- x 
2
(n)
cos x
-1
And reversing this sighn , we obtain .

-1
Example: the 9th order derivative of sin x
Re pe ate d diffe re ntiation
 g := diff(arcsin(x),x,x,x,x,x,x,x,x,x)
2 4 6 8
 2182950  x  
396900  x   3783780  x  
2027025  x  
11025
 
11  
13  
15  
17  9
2 2 2 2 2 2 2 2 2 2
1 x 1 x 1 x 1 x 1 x

Formula
 f := n-> sum(binomial(n-1,n-1-2*r)
*(2*r-1)!!*(2*n-3-2*r)!!*x^(n-1-2*r)
/(1-x^2)^(n-r-1/2), r=0..floor(n/2))

 f(9)
2 4 6 8
 2182950  x  
396900  x   3783780  x  
2027025  x  
11025
 
11  
13  
15  
17  9
2 2 2 2 2 2 2 2 2 2
1 x 1 x 1 x 1 x 1 x

- 10 -
 Formula 9.3.2 can also be expressed as follows, if partial fraction decomposition is applied to the right side.

Formula 9.3.2'

r
(n ) 1 n -1 n -1 (2r -1)!!(2n -2r -3)!!
sin x (-1)r
-1
=
2n-1 1-x 2 r=0 (1+x)r(1-x)n-1-r

 r  (1+x) (1-x)
(n ) 1 n -1 n -1 (2r -1)!!(2n -2r -3)!!
cos x
-1 r
=- (-1) r n-1-r
n-1
2 1-x 2 r=0

For example,
(3) 3x 2 1 1 2x 2+1
sin x
-1
= 5
+ 3
=
2 2 1-x 2 (1+ x)2(1- x)2
1- x  1- x 
2 2

Let us decompose this to the partial fractions. then,

2x 2+1 B0 B1 B2
2 2
= 2
+ 1 1
+
(1+ x) (1- x) (1-x) (1+x) (1-x) (1+x)2
Using Heaviside cover-up method,
2
2x +1 B1(1- x)1 B2(1- x)2
= B0 +
+ (1)
(1+ x)2 (1+x)1 (1+x)2
2x 2+1 B0(1+x)1 B2(1-x)1
= + B1 + (2)
(1+ x)1(1- x)1 (1-x)1 (1+x)1
2x 2+1 B0(1+x)2 B1(1+x)1
= + + B2 (3)
(1- x)2 (1-x)2 (1-x)1
Giving x =1, 0 , -1 to (1),(2),(3) respectively,
3 3
= B0 , 1 = B0 + B1 + B2 , = B2
22 22
2
From these B1 = - . Then,
22

 
(3) 1 3 2 3
sin x
-1
= 2
- 1 1
+
22 1-x 2 (1-x) (1+x) (1-x) (1+x)2

 
1 (-1)!!3!! 2 1!!1!! 3!!(-1)!!
= 0 2
- +
22 1-x 2 (1+x) (1-x) (1+x)1(1-x)1 (1+x)2(1-x)0

r 
1 3-1 2 (2r -1)!!(2 3-2r -3)!!
= (-1)r
22 1-x 2 r=0 (1+x)r(1-x)3-1-r

Formula 9.3.3 : Higher Derivatives of arcsec x, arccsc x

The following expressions hold for a natural number n.

(n) n n Ar
sec x
-1
= (-1)n-1 1
r=1 r-
n-1+2r 2
1-x 
-2
x

- 11 -
 (n) n n Ar
csc x = (-1)n
-1
1
r=1 r-
n-1+2r 2
1-x 
-2
x
where n Ar are coefficients as follows.

1A1 1
2A1 2A2 2 1
3A1 3A2 3A3 6 7 3
4A1 4A2 4A3 4A4 = 24 48 45 15
5A1 5A2 5A3 5A4 5A5 120 360 549 390 105
6A1 6A2 6A3 6A4 6A5 6A6 720 3000 6570 7425 4200 945
 
And these coefficients are obtained by the following sequential computation

2! 1!!
1, 5
3! 7 3!! 7=2! 1+1!! 5
1, 6 3, 8
4! 48 45 5!! 48=3! 1+7 6, 45=7 3+3!! 8
1, 7 3, 9 5, 11
5! 360 549 390 7!! 360=4! 1+48 7, 549=48 3+45 9,
1, 8 3, 10 5, 12 7, 14 390=45 5+5!! 11
6! 3000 6570 7425 4200 9!! 3000=5! 1+360 8, 6570=360 3+549 10,
7425=549 5+390 12, 4200=390 7+7!! 14
 

Proof
Differentiating arcsec x with respect to x one by one, it is as follows.
(1) 1
sec x
-1
= 1
2
x 21-x -2
(2) 2 1
sec x
-1
=- 1
- 3
2 2
x 1-x 
3 -2
x 1-x 
5 -2

(3) 6 7 3
sec x
-1
= 1
+ 3
+ 5
2 2 2
x 41-x -2 x 61-x -2 x 81-x -2
(4) 24 48 45 15
sec x
-1
=- 1
- 3
- 5
- 7
2 2 2 2
x 1-x  x 1-x  x 1-x  x 1-x 
5 -2 7 -2 9 -2 11 -2

- 12 -
These coefficients are calculable with the algorithm in the formula. Then, by induction, we obtain the desired
expressions.

Note
These coefficients and the direct calculation method is not known. However, these have the following
character.
1 = 0!
2 - 1 = 1!
6 - 7 + 3 = 2!
24 - 48 + 45 - 15 = 3!
120 - 360 + 549 - 390 + 105 = 4!
720 - 3000 + 6570-7425+4200-945 = 5!
 

- 13 -
9.4 Higher Derivative of Inverse Hyperbolic Functions

Formula 9.4.1 : Higher Derivatives of arctanh x , arccoth x

When  denots ceiling function, the following expressions hold for a natural number n.
(n) (n) (n -1)! n /2
tanh
-1
x = coth -1x = (-1)n >A
n r=1 n n+1-2r x
n+1-2r

x +1
2

Proof
If we differentiate arctanh x with respect to x one by one, these coefficients become the same as the ones
(n)
tan x
-1
of ( Formula 9.3.1 ). Then we obtain the desired expression.

Formula 9.4.2 : Higher Derivatives of arcsinh x, arccosh x

When  denots floor function, the following expressions hold for a natural number n.

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n) n-1
n/2 n -1
sinh x
-1 r
= (-1) (-1) 1
r=0 n -1-2r n-r-
2
x +1
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n) n-1
n/2 n -1
cosh x
-1 r
= (-1) (-1) 1
r=0 n -1-2r n-r-
2
x -1
2

Proof
If we differentiate arcsinh x with respect to x one by one, these coefficients become the same as the ones
(n)
sin x
-1
of ( Formula 9.3.2 ). Then we obtain the desired expression.

Formula 9.4.2 can also be expressed as follows, if partial fraction decomposition is applied to the right side.

Formula 9.4.2'
n-1 n -1 1 1

 r (2r -1)!!(2n -2r -3)!!(1+ix)

n -1 - -r + r-n
 
(n) i
sinh x
-1 2 2
= (-1) r (1-ix)
2 r=0

n-1 n -1 1 1

 r (2r -1)!!(2n -2r -3)!!(x +1)

n -1 - -r + r-n
 
(n) 1
cosh x
-1 2 2
= - (x -1)
2 r=0

-1
Example: the 5th order derivative of sinh x
Re pe ate d diffe re ntiation
 g := diff(arcsinh(x),x,x,x,x,x):simplify(%)
4 2
24  x  72  x  9

 9
2 2
x  1
Formula
 asinh := n-> (I/2)^(n-1)*
sum((-1)^r*binomial(n-1,r)*(2*r-1)!!*(2*n-2*r-3)!!
*(1+I*x)^(-1/2-r)*(1-I*x)^(1/2+r-n), r=0..n-1)

- 14 -
 asinh(5):simplify(%)
4 2
24  x  72  x  9 

9 9
2 2
(1  i  x)  (i  x  1)

Formula 9.4.3 : Higher Derivatives of arcsech x , arccsah x

The following expressions hold for a natural number n.

(n) n-1
n (-1)r n Ar
sech x
-1
= (-1) 1
r=1 r-
2
x -2 -1
n-1+2r
x
(n) n (-1)r n Ar
csch
-1
x = (-1)n-1 1
r=1 r-
2
x -2 +1
n-1+2r
x
where n Ar are same as ones in Formula 9.3.3 .

- 15 -
9.5 Termwise Higher Derivative of a Logarithmic Function
Since the higher derivative of a logarithmic function has already shown in the previous section, there is no
necessity of differentiating this termwise. However, interesting results are obtained if the termwise higher
derivative is compared with the higher integral.

Formula 9.5.1
 k +(n -1)! k (n -1)!
(-1)k x = -1< x <1 (1.1)
k=0 k! (1+ x)n

k
 -n
x k = (1+ x)-n -1< x <1 (1.1')
k=0

Proof
x -1 (x -1)2 (x -1)3 (x -1)4
- + - +- = log x
1 2 3 4
Differentiating both sides of this with respect to x,
0!
1 - (x -1) + (x -1)2 - (x -1)3 +- = 0< x <2
x
Further differentiating both sides of this with respect to x,
1!
1 - 2(x -1) + 3(x -1)2 - 4(x -1)3 +- = 0< x <2
x2
Furthermore differentiating both sides of this with respect to x,
2!
1 2 - 2 3(x -1) + 3 4(x -1)2 - 4 5(x -1)3 +- = 0< x <2
x3
Thus generally,
 k +(n -1)! (n -1)!
(-1)k (x -1)k = 0< x <2
k=0 k! xn
Replacing x with 1+ x , we obtain
 k +(n -1)! k (n -1)!
(-1)k x = -1< x <1 (1.1)
k=0 k! (1+ x)n
Next, if we divide both sides by (n -1)! , the left side becomes as follows.

  k
 k +(n -1)! k  n -1+ k  -n
(-1)k x = (-1)k xk = xk
k=0 k !(n -1)! k=0 k k=0
Therefore we obtain

 k x
 -n k
= (1+ x)-n -1< x <1 (1.1')
k=0

c.f.
In (1.1'), n may be the real number already. Here, let  = -n , then

k 

x k = (1+ x) -1< x <1
k=0
This is Newton's generalized binomial theorem.

- 16 -
Special Values
In fact, Formula 9.5.1 holds also on x=1. Therfore, giving x=1 to (1.1) without considering convergence
conditions, we obtain the following special values.
0!
1 - 1 + 1 - 1 +-  =
21
1!
1 - 2 + 3 - 4 +-  = 2
2
2!
1 2 - 2 3 + 3 4 - 4 5 +-  = 3
2
3!
1 2 3 - 2 3 4 + 3 4 5 - 4 5 6 +-  = 4
2

 k +(n -1)! (n -1)!
(-1)k =
k =0 k! 2n

2012.01.06 Renewal
K. Kono
Alien's Mathematics

- 17 -
10 Termwise Higher Derivative (Trigonometric, Hyperbolic)
In this chapter, for the function which the second or more order derivative is difficult to express with an easy
unification notation among trigonometric functions and hyperbolic function, we differentiate the series expansion
of these functions term by term. Therefore, sin x, cosx, sinhx, cosh x mentioned in " 9.2 Higher Derivative of
Elementary Functions " are not treated here.

10.1 Termwise Higher Derivative of tan x

10.1.0 Higher Derivative of tan x

Accordong to Formula 9.2.6 ( 9.2 ), Higher Derivative of tan x is expressed as follows.
n /2
(tan x) n = n+1-2r
()
Σ
r=0
n T r(tan x) (0.n)

Where,  is celling function and nTr are coefficients as follows.

0T 0 1
1T 0 1T 1 1 1
2T 0 2T 1 2 2
3T 0 3T 1 3T 2 = 6 8 2
4T 0 4T 1 4T 2 24 40 16
5T 0 5T 1 5T 2 5T 3 120 240 136 16
6T 0 6T 1 6T 2 6T 3 720 1680 1232 272
 

10.1.1 Termwise Higher Derivative of Taylor Series of tan x

Formula 10.1.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,

the following expressions hold for |x |<  /2 .

 22k 22k -1B2k 2k-2

(tan x) (1)
=Σ x (1.1)
k =1 2k (2k -2)!
 22k 22k -1B2k 2k-3
(tan x) (2)
=Σ x (1.2)
k =2 2k (2k -3)!
 22k 22k -1B2k 2k-4
(tan x) (3)
=Σ x (1.3)
k =2 2k (2k -4)!
 22k 22k -1B2k 2k-5
(tan x)(4) = Σ x (1.4)
k =3 2k (2k -5)!


(n)  22k 22k -1B2k 2k- n-1

(tan x) = Σ x (1.n)
k=
n +1

2 k (2k - n -1)!
2

-1-
Proof
tan x is expanded to Tylor series as follows.

 2 2 -1B 2k
2k 2k

tan x = Σ x 2k-1 |x| <
k =1 2k (2k -1)! 2
Differentiating both sides of this with respect to x repeatedly, we obtain (1.1) ~ (1.4).

Here, considering the relation between the derivative order n and the first term k0 of Σ, it is as follows.

n 0 1 2 3 4 5 
k0 1 1 2 2 3 3 

n +1
Such a relation can be expressed by k0 =  using a ceiling function x = x. Then we obtain
2
the desired expression.

Sum of Taylor series of the higher derivative of tan x

From (0.n) and (1.n) , we obtain the following expressions for |x|<  /2 .
 22k 22k -1B2k 2k- n-1 n /2
Σ x = Σ nTr(tan x)n+1-2r (1.t)
k=
n +1

2k(2k - n -1)! r=0
2

And giving x = /4 to this, we obtain the following special values.

22k 22k -1B2k  2k-2

 4
 1/2
Σ
k =1 2k (2k -2)!
= Σ 1T r = 2
r=0

22k 22k -1B2k  2k-3

 4
 2/2
Σ
k =2 2k (2k -3)!
= Σ 2T r = 4
r=0

2 2 -1B2k
2k 2k
 2k-4

 4
 3/2
Σ
k =2 2k (2k -4)!
= Σ 3Tr = 16
r=0

22k 22k -1B2k  2k-5

 4
 4/2
Σ
k =3 2k (2k -5)!
= Σ 4Tr = 80
r=0


10.1.2 Termwise Higher Derivative of Fourier Series of tan x

As seen in 5.1.2 , tan x is expanded to Fourier series in a broad meaning.as follows.
tan x = 2(sin 2x - sin 4x + sin 6x - sin 8x +-)
-2i(cos 2x - cos 4x + cos 6x - cos 8x +-)+ i

Although this does not hold as an equation, differentiating the real part of the both sides repeatedly,
we obtain the following perfunctory expressions.
(1) 2
(tan x) = 2 (1cos 2x -2cos 4x +3cos 6x -4cos 8x +-)
(tan x)(2) = -2312sin 2x -22sin 4x +32sin 6x -42sin 8x +-
(tan x)(3) = -2413cos 2x -23cos 4x +33cos 6x -43cos 8x +-

-2-
 (tan x)(4) = 2514sin 2x -24sin 4x +34sin 6x -44sin 8x +-


(tan x)(2n -1) = (-1)n-122nΣ(-1)k-1 k 2n-1cos 2kx (2.2n-1)
k =1

(tan x)(2n) = (-1)n22n+1Σ(-1)k-1 k 2nsin 2kx (2.2n)
k =1
Of course, these also do not hold as an equation. However, if the calculation is advanced assuming that
thiese hold, interesting results are obtained as follows.

10.1.3 Dirichlet Odd Eta & Even Beta

Formula 10.1.3
 (-1)k-1  (-1)k
When (x) =Σ ,  (n) =Σ , B2k are Bernoulli Numbers and nTr are the
k =1 kx k =0 (2k +1) n
coefficients mentioned in 10.1.0 , the following expressions hold.

(-1)n-1  2 2 -1B2k
2k 2k
 2k-2n
(-1)n-1
 4
n
(1-2n) = Σ = Σ 2n-1T r
4n-1 k =n
2 2k(2k -2n)! 24n-1 r=0

(-1)n 22k 22k -1B2k  2k-2n-1

(-1)n
 4
 n
(-2n) = Σ = Σ 2n T r
22n+1 k =n +1 2k(2k -2n -1)! 22n+1 r=0

Proof
From (0.n) , (1.n) , (2.2n-1) and (2.2n) , we obtain the following perfunctory expressions.

(-1)n-1  2 2 -1B2k 2k-2n

2k 2k

k-1 2n-1
Σ (-1) k cos 2kx = 2n Σ2k(2k -2n)!
x
k =1 2 k =n

(-1)n-1 n
Σ
2n-2r
= 2n 2n-1T r(tan x)
2 r=0


k-1 (-1)n  22k 22k -1B2k 2k-2n-1
Σ Σ
2n
(-1) k sin 2kx = x
k =1 22n+1 k =n +1 2k(2k -2n -1)!
(-1)n n
Σ
2n+1-2r
= 2n+1 2n T r(tan x)
2 r=0

If x = /4 is substituted for these, the left sides are as follows respectively.

 k 
k-1 2n-1
Σ(-1) k cos = 22n-1Σ(-1)k-1 k 2n-1 = 22n-1(1-2n)
k =1 2 k =1
 k 
Σ (-1)k-1 k 2nsin k = Σ(-1)k-1 (2k -1)2n = (-2n)
k =1 2 k =1
Therefore, we obtain the desired expressions.

Example1
1  2 -1B2k
2k 2k
1 1 1
(-1) = Σ 2k (2k -2)! = Σ 1T r =
23 k =1 23 r=0 4

-3-
 1  2 -1B2k
2k 2k
1 2 1
(-3) = - 7 Σ = - 7 Σ 3T r = -
2 k =2 2k (2k -4)! 2 r=0 8

1  2 -1B2k
2k 2k
1 1 1
(-2) = - 3 Σ = - 3 Σ 2T r = -
2 k =2 2k(2k -3)! 2 r=0 2

1  2 -1B2k
2k 2k
1 n 5
(-4) = Σ 2k(2k -5)! = Σ 4T r =
25 k =3 25 r=0 2

Example2 (-5),  (-6)

Note
Using a relation  (-2n ) = E2n / 2 between Dirichlet Even Beta and Euler number, .we obtain the
following expression.

(-1)n 22k 22k -1B2k  2k-2n-1

(-1)n
 4
 n
E2n = Σ 2k (2k -2n -1)!
= Σ 2n T r (3.E)
22n k =n +1 22n r=0

-4-
10.2 Termwise Higher Derivative of tanh x

10.2.0 Higher Derivative of tanh x

Accordong to Formula 9.2.7 ( 9.2 ), Higher Derivative of tanh x is expressed as follows.
n /2
(tanh x) n = (-1)n Σ (-1)r nTr(tanh x)n+1-2r
( )
(0.n)
r=0

Where,  is celling function and nTr are the same as the coefficients in 10.1.0 .

10.2.1 Termwise Higher Derivative of Taylor Series of tanh x

Formula 10.2.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,

the following expressions hold for |x |<  /2 .

 22k22k -1B2k 2k-2

(tanh x) (1)
=Σ x
k=1 2k (2k -2)!
 22k22k -1B2k 2k-3
(tanh x) (2)
=Σ x
k=2 2k (2k -3)!

(n)  22k22k -1B2k 2k- n-1
(tanh x) = Σ x (1.n)
n +1 2k(2k - n -1)!
=k 
2

Proof
tanh x is expanded to Tylor series as follows.

 2 2 -1 B 2k
2k 2k

tanh x = Σ 2k-1
x |x| <
k =1 2k (2k -1)! 2
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

10.2.2 Termwise Higher Derivative of Fourier Series of tanh x

Formula 10.2.2
The following expressions hold for x >0 .
(tanh x)(1) = 2211e -2x -21e -4x +31e -6x -41e -8x +- 
(tanh x)(2) = -2312e -2x -22e -4x +32e -6x -42e -8x +- 


(tanh x) n = (-1)n-12n+1Σ(-1)k-1k ne -2kx
( )
(2.n)
k =1

Proof
tanh x is epanded to Fourier series for x > 0 as follows.

e x - e -x 1- e -2x
tanh x = x -x
= -2x
= 1- e -2x1 - e -2x + e -4x - e -6x +-
e +e 1+ e

-5-
 = 1 - 2e -2x - 2e -4x + 2e -6x - 2e -8x +- (2.0)

= 1 - 2(cos 2ix - cos 4ix + cos 6ix - cos 8ix +- )

- 2i(sin 2ix - sin 4ix + sin 6ix - sin 8ix +- )
Differentiating both sides of (2.0) with respect to x repeatedly, we obtain the desired expression .

10.2.3 Exponential series and Bernouilli series

Replacing x with x /2 in (0.n) , (1.n) and (2.n) , we obtain the following formula.

Formula 10.2.3
When  B2n are Bernoulli Numbers and nTr
is ceilling function, are the coefficients mentioned in 10.1.0 ,
the following expressions hold. for 0 < x <  .


k-1 kn n-1  22k -1B2k 2k- n-1
Σ (-1) = (-1) Σ x
k =1 e kx k=
n +1 2k(2k - n -1)!

2
n+1-2r

 
1 n /2x
= - n+1 Σ (-1) nTr tanh r
2 r=0 2

Example
2 -1B2k 2k-2
2k
11 21 31 41 

 
1 2 x
e
1x -
e
2x
+
e
3x
-
e
4x
+- = Σ
k=1 2k (2k -2) !
x = - 2 tanh
2 2
-1

2 -1B2k 2k-3
2k
12 22 32 42 

 
1 3 x x
1x - 2x
+ 3x
- 4x
+- = -Σ x = - 3 2 tanh -2 tanh
e e e e k=2 2k (2k -3) ! 2 2 2

Dirichlet Odd Eta (minus)

Disregarding the convergence condition and substituting x =0 for Formula 10.2.3 , we obtain Dirichlet Odd
Eta (minus)

Formula 10.2.3'
 (-1)k-1
Let (x) =Σ , and let Bernoulli number B2n and tangent number Tr are as follows.
k =1 kx
B2=1/6, B4=-1/30, B6=1/42, B8=-1/30, B10=5/66, 
T1=1, T3=2, T5=16, T7=272, T9=7936, 
Then the following expressions hold.
2n
2 -1B2n (-1)n
(-2n +1) = = - 2n T2n-1 (3.2n-1')
2n 2
(-2n) =0 (3.2n')

Proof
Sbstitute x =0 for Formula 10.2.3 . Then, the Fourier series is

 kn  (-1)k-1
Σ (-1) k-1
k0
=Σ = (-n)
k =1 e k=1 k -n

-6-
The Taylor series and the polynomial are
when n =2m-1

n-1  22k -1B2k

(-1) Σ n +1 2 2 -
k( k n )-1 !
02k-n-1
k= 
2

 22k -1B2k
Σ
2m-1-1
= (-1) 02k-(2m-1)-1
2m -1+1

2 2
k k -(2 -1
m ) -1 !
k=
2

2 -1B2k 2k-2m 2 -1B2m 0 2 -1B2m

2k 2m 2m

=Σ 0 = 0 =
k=m 2k (2k -2m)! 2m 0! 2m
n+1-2r

 
1 n /2 0
- n+1 Σ (-1) nTr tanh r
2 r=0 2
1 m 1
Σ(-1)r 2m-1Tr0 (-1)m 2m-1Tm 00
2m-2r
=- =-
22m r=0 2 2m

(-1)m
=- T2m-1  2m-1T m = T2m-1
22m
when n =2m

n-1  22k -1B2k

(-1) Σ n +1 2k (2k -n -1)!
02k-n-1
k= 
2

 22k -1B2k
Σ
2m-1
= (-1) 02k-2m-1 = 0
2m +1

2 2
k k m
( -2 -1 )!
k=
2
n+1-2r

 
1 n /2 0
- n+1 Σ (-1) nTr tanh r
2 r=0 2
1 m

2m+1 Σ
=- (-1)r 2mTr0 2m+1-2r
=0
2 r=0

Replacing m with n , we obtain (3.2n-1') and (3.2n') .

-7-
10.3 Termwise Higher Derivative of cot x

10.3.0 Higher Derivative of cot x

Accordong to Formula 9.2.6 ( 9.2 ), Higher Derivative of cot x is expressed as follows.
n /2
(cotx) n = (-1)n Σ nTr(cotx)n+1-2r
( )
(0,n)
r=0

Where,  is celling function and nTr are the same as the coefficients in 10.1.0 .

10.3.1 Termwise Higher Derivative of Taylor Series of cot x

Formula 10.3.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,
the following expressions hold for 0 < x <  .

1!  22kB2k 2k-2
(cot x) (1)
=- -Σ x
x2 k=1 2k(2k -2)!

2!  22kB2k 2k-3
(cot x)(2) = -Σ x
x3 k=2 2k(2k -3)!


(n) n!  22kB2k
(cot x) = (-1) n
n+1
-Σ x 2k-n-1 (1.n)
x n +1

2k(2k -n -1)!
=k
2

Proof
 22kB2k 2k
x cot x = 1 +Σ x 0< x < 
k =1 (2k)!
From this

0!  22kB2k 2k-1
cot x = -Σ x
x1 k=1 2k(2k -1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

Formula 10.3.1 can also be expressed as follows.

Formula 10.3.1'
When B2k denote Bernoulli Numbers and  denotes ceilling function, the following expressions hold for

 /2 < x <  .

22k22k -1B2k  2k- n-1

 
(n) 
(cot x) =-Σ x-
n +1

2k(2k - n -1)! 2
= k
2

Proof

cot x = -tan x -  2 
-8-
  22k22k -1B2k 2k-1 
tan x = Σ x |x|<
k=1 2k (2k -1)! 2
From these

22k22k -1B2k  2k-1

 

cotx = -Σ x- 0< x < 
k =1 2k (2k -1)! 2
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

Sum of Taylor series of the higher derivative of cot x

From (0,n) and (1.n) , we obtain the following expressions for 0<x < .

 22kB2k
 
n /2
n+1 n+1-2r
Σ n
Σ
2k
x = (-1) n ! - x n Tr(cotx) (1.t)
n +1 2k(2k -n -1)! r=0
k= 
2

And giving x = /4 to this, we obtain the following special values.

22kB2k  2k
 2

 4  4

Σ
k =1 2k(2k -2)!
= -1! + 2

22kB2k  2k
 3

 4  4

Σ
k =2 2k(2k -3)!
= 2! - 4

22kB2k  2k
 4

 4  4

Σ
k =2 2k(2k -4)!
= -3! + 16

22kB2k  2k
 5

 4  4

Σ
k =3 2k(2k -5)!
= 4! - 80


10.3.2 Termwise Higher Derivative of Fourier Series of tan x

As seen in 5.3.2 , cot x is expanded to Fourier series in a broad meaning as follows.
cotx = 2(sin 2x + sin 4x + sin 6x + sin 8x +)
-2i(cos 2x + cos 4x + cos 6x + cos 8x +) - i
Although this does not hold as an equation, differentiating the real part of the both sides repeatedly,
we obtain the following perfunctory expressions..
(1) 2
(cotx) = 2 (1cos 2x +2cos 4x +3cos 6x +4cos 8x +)
(cotx)(2) = -2312sin 2x +22sin 4x +32sin 6x +42sin 8x +
(cotx)(3) = -2413cos 2x +23cos 4x +33cos 6x +43cos 8x +
(cotx)(4) = 2514sin 2x +24sin 4x +34sin 6x +44sin 8x +


(cotx)(2n -1) = (-1)n-122nΣk 2n-1cos 2kx (2.2n-1)
k=1

(cotx)(2n) = (-1)n22n+1Σk 2nsin 2kx (2.2n)
k =1

-9-
 Of course, these also do not hold as an equation. However, if the calculation is advanced assuming that
thiese hold, interesting results are obtained as follows.

10.3.3 Dirichlet Odd Eta & Even Beta

Formula 10.3.3
 (-1)k-1  (-1)k
When (x) =Σ ,  (n ) =Σ , B2k are Bernoulli Numbers and nTr are the
k =1 kx k =0 (2k +1)
n

coefficients mentioned in 10.1.0 , the following expressions hold.

(2n -1)! + Σ 
(-1)n-1 2n 22kB2k  2k

  4
4 
(1-2n) =
24n-1 k=n 2k(2k -2n)!
(-1)n-1 n
= Σ 2n-1T r
24n-1 r=0

(2n)! - Σ 
(-1)n 2n 22kB2k  2k

  4
4 
(-2n) =
22n+1 k =n +1 2k(2k -2n -1)!
n
= (-1)nΣ 2nTr
r=0

Proof
From (0.n) , (1.n) , (2.2n-1) and (2.2n) , we obtain the following perfunctory expressions.

 
2k
B2k

n  2
(-1 )
Σ k 2n-1cos 2kx = 2n 2n (2n -1)! +
2 k( 2 k -2 n Σ
)!
x 2k
k =1 2 x k=n

(-1)n n
Σ
2n-2r
= 2n 2n-1T r(cotx)
2 r=0

(2n)! - Σ 
 (-1)n  22kB2k
Σ
2n
k sin 2kx = x 2k
k =1 22n+1x 2n+1 k=n +1 2k(2k -2n -1)!
n
= (-1)nΣ 2nTr(cotx)2n+1-2r
r=0
If x = /4 is substituted for these, the left sides are as follows respectively.
 k 
Σk cos
2n-1
= -22n-1Σ(-1)k-1 k 2n-1 = -22n-1(1-2n)
k =1 2 k=1
 k 
Σ k 2n
sin k = Σ (-1)k-1 (2k -1)2n = (-2n)
k =1 2 k =1
Therefore, we obtain the desired expressions.

Example

  
2 22kB2k  2k

 4
1 4  1 1 1
(-1) = 1!+Σ = Σ 1T r =
23 k =1 2k(2k -2)! 23 r=0 4

(-3) = -   3!+Σ
2k(2k -4)!  4  
2k
1 4 2 B   4
1  2k
2k 2 1
2  7
= -
2
Σ k =2
7
r=0
3T r =-
8

- 10 -
   
2
 22kB2k  2k

 4
1 4 1 1 1
(-2) = - 3 2!-Σ =- Σ T
2 r = -
2 k =2 2k(2k -3)! 23 r=0 2

 2k(2k -5)!  4  
2k
2 B 4
 2k

2 
1 4  1 n 5
(-4) =
2k
5
4!-Σ =
k =3
Σ 4T r =
25 r=0 2

10.3.4 Factorial and Bernouilli series

Formula 10.3.4
When B2k denotes a Bernoulli Number, the following expressions hold.
22k -2B2k  2k

(2n -1)! = Σ
k=n 2k(2k -2n )! 2   (4.1)

2k
2 B2k  2k

 

(2n)! = Σ (4.2)
k= n +1 2k(2k -2n -1)! 2
Proof
Replacing n with 2n -1 in Formula 10.3.1' , we obtain

 2 2 -1B 2k
2k 2k
 2k-2n
= -Σ x - 2 
(2n -1)
(cot x)
k =n 2k(2k -2n)!
Substituting x =  /4 for this,
 22k22k -1B2k  2k-2n
|x=/4 = -Σ  4
(2n -1)
(cotx)
k=n 2k(2k -2n)!
On the othe hand, substituting x =  /4 for (2.2n-1) ,
2n  22kB2k  2k-2n

  4
(2n -1) 4
(cotx) |x=/4 = -(2n -1) ! -Σ
k=n 2k(2k -2n) !
Since these have to be equal, we obtain (4.1).
Next, replacing n with 2n in Formula 10.3.1' ,

2 2 -1B2k
2k 2k
 2k-2n-1

 

(cot x)(2n) = - Σ x-
k=n +1 2k(2k -2n -1)! 2
x =  /4 for this, we obtain
Substituting

2 2k
22k
-1B2k  2k-2n-1

 4

(cotx)(2n)|x=/4 = Σ
k=n +1 2k(2k -2n -1)!
On the othe hand, substituting x =  /4 for (2.2n) ,
2n+1  22kB2k  2k-2n-1

  4
(2n ) 4
(cotx) |x=/4 = (2n) ! -Σ
k=n +1 2k(2k -2n -1) !
Since these have to be equal, we obtain (4.2).

Proof of the pudding is in the eating.

- 11 -
- 12 -
10.4 Termwise Higher Derivative of coth x

10.4.0 Higher Derivative of coth x

Accordong to Formula 9.2.7 ( 9.2 ), Higher Derivative of coth x is expressed as follows.
n /2
(coth x) n = (-1)n Σ (-1)r nTr(coth x)n+1-2r
()
(0.n)
r=0

Where,  is celling function and nTr are the same as the coefficients in 10.1.0 .

10.4.1 Termwise Higher Derivative of Taylor Series of coth x

Formula 10.4.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,

the following expressions hold for 0<x < .

1!  22k B2k
(coth x) (1)
=- +Σ x 2k-2
x2 k=1 2k(2k -2)!
2!  22k B2k
(coth x)(2) = +Σ x 2k-3
x3 k=2 2k(2k -3)!


(n) n!  22k B2k

(coth x) = (-1) n
n+1
+ Σ x 2k-n-1 (1.n)
x n +1

2k(2k -n -1)!
k=
2

Proof

 22k B2k 2k
x coth x = 1 + Σ x 0< x < 
k =1 (2k)!
From this

0!  22k B2k
coth x = +Σ x 2k-1
x k=1 2 2
k( k -1 )!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

10.4.2 Termwise Higher Derivative of Fourier Series of coth x

Formula 10.4.2
The following expressions hold for x >0 .
(coth x)(1) = -2211e -2x +21e -4x +31e -6x +41e -8x + 
(coth x)(2) = 2312e -2x +22e -4x +32e -6x +42e -8x + 


(coth x) n = (-1)n2n+1Σk ne -2kx
()
(2.n)
k=1

- 13 -
Proof
coth x is expanded to Fourier series for x >0 as follows.

e x + e -x 1+ e -2x
coth x = x -x
= -2x
= 1+ e -2x1 + e -2x + e -4x + e -6x +
e -e 1- e
= 1 + 2e -2x + 2e -4x + 2e -6x + 2e -8x +- (2.0)

= 1 + 2(cos 2ix + cos 4ix + cos 6ix + cos 8ix + )

+ 2i(sin 2ix + sin 4ix + sin 6ix + sin 8ix + )
Differentiating both sides of (2.0) with respect to x repeatedly, we obtain the desired expression.

10.4.3 Exponential series and Bernouilli series

Replacing x with x /2 in (0.n) , (1.n) and (2.n) , we obtain the following formula.

Formula 10.4.3
When  is ceilling function, B2n are Bernoulli Numbers and nTr are the coefficients mentioned in 10.1.0 ,
the following expressions hold for 0 < x <  .

 
 kn 1 B2k x 2k

Σ = n ! +(-1) Σ n
k=1 e kx x
n+1
k=
n +1 2k(2k -n -1)!

2
n+1-2r

 
1 x
n /2
= n+1 Σ (-1) nTr coth r
2 r=0 2

Example

  2
11 21 31 41  B2k x 2k

1 1 x
1x + 2x
+ 3x
+ 4x
+ = 2
1! -Σ = 2
coth 2 -1
e e e e x k=1 2k(2k -2) ! 2

x  2k(2k -3) ! 
2k
12 22 32 42 B x 

2  
1 1 2k x x
2! +Σ
3
1x + 2x
+ 3x
+ 4x
+ = 3
= 2 coth 3
-2 coth
e e e e k=2 2 2

Exponential series and Factorial

Giving x =1 to Formula 10.4.3, we obtain the following special values.

Formula 10.4.3'
When  is ceilling function, B2n are Bernoulli Numbers and nTr are the coefficients mentioned in 10.1.0 ,
the following expressions hold.

n n+1-2r
k  B2k
 
1 n /2 1
Σ k
= n ! + (-1)n
Σ n +1 2k(2k -n -1) !
= n+1 Σ (-1)r nTr coth
2
k=1 e k= 
2 r=0
2

Note
Mr. Sugimoto was foreseeing this formula. ( http://homepage3.nifty.com/y_sugi/sp/sp56.htm )

Example
11 21 31 41  B2k
 
1 2 1
1+ 2
+ 3
+ 4
+ = 1! -Σ = 2 coth -1
e e e e k =1 2k(2k -2) ! 2 2

- 14 -
 12 22 32 42  B2k
 
1 3 1 1
1 + 2
+ 3
+ 4
+ = 2! +Σ = 3 2 coth -2 coth
e e e e k =2 2k(2k -3) ! 2 2 2
13 23 33 43  B2k
 
1 4 1 2 1
1+ 2
+ 3
+ 4
+ = 3! -Σ = 4 6 coth -8 coth +2
e e e e k =2 2k(2k -4) ! 2 2 2

Therefore also, the following approximation formula is obtained.

Formula 10.4.3"
1p 2p 3p 4p
+ + + +  (1+ p) p >0 (3.p)
e1 e2 e3 e4

Example (1+2.5)

- 15 -
10.5 Termwise Higher Derivative of csc x

10.5.0 Higher Derivative of csc x

Accordong to Formula 9.2.8 ( 9.2 ), Higher Derivative of csc x is expressed as follows.

(n) (-1)n n/2 n-2r

(cscx) =
sin x Σ
n E r(cotx)
r=0
(0.n)

Where,  is floor function and n Er are coefficients as follows.

0E 0 1
1E 0 1
2E 0 2E 1 2 1
3E 0 3E 1 = 6 5
4E 0 4E 1 4E 2 24 28 5
5E 0 5E 1 5E 2 120 180 61
6E 0 6E 1 6E 2 6E 3 720 1320 662 61
7E 0 7E 1 7E 2 7E 3 5040 10920 7266 1385
 

10.5.1 Termwise Higher Derivative of Taylor Series of csc x

Formula 10.5.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,

the following expressions hold for 0<x < .

1! 22k -2B2k 2k-2


(csc x) (1)
=- +Σ x
x2 k=1 2k (2k -2)!

2! 22k -2B2k 2k-3


(csc x) (2)
= +Σ x
x3 k=2 2k (2k -3)!

(n) n!  22k -2B2k 2k- n-1

(csc x) = (-1) n
n+1
+ Σ x (1.n)
x n +1 2k (2k - n -1)!
k= 
2

Proof
22k -2B2k 2k

xcscx = 1 +Σ x 0< x < 
k=1 (2k)!
From this

 2 -2B 2k
2k
0!
cscx = +Σ x 2k-1
x k=1 2k (2k -1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

Formula 10.5.1 can also be expressed as follows.

- 16 -
Formula 10.5.1'
When E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are Euler Numbers and  is floor function,

the following expressions hold for  /2 < x <  .

E2k  2k-n

 
(n) 
(csc x) =Σ x-
n +1 (2k -n )! 2
k= 
2

proof

csc x = sec x -  2 
 E2k 
secx =Σ x 2k |x|<
k=0 (2k)! 2
From these

E2k  2k

x - 2 

cscx =Σ 0< x < 
k=0 (2k)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of sec x. )

Sum of Taylor series of the higher derivative of csc x

From (0,n) and (1.n) , we obtainthe following expressions for 0<x < .

22k -2B2k 2k
 
n+1 n/2
 x
n-1 n-2r
Σ n +1 2k (2k - n -1)!
x = (-1) n! -
sin x Σ
r=0
n E r(cotx) (1.t)
k= 
2

And giving x = /2 to this, we obtain the following special values.

2 -2B2k
2k
 2k

 2

Σ
k=1 2k (2k -2)!
= 1!

22k -2B2k  2k
 3

 2  2

Σ
k=2 2k (2k -3)!
= -2! + 1

22k -2B2k  2k

 2

Σ
k=2 2k (2k -4)!
= 3!

22k -2B2k  2k
 5

 2  2

Σ
k=3 2k (2k -5)!
= -4! + 5


10.5.2 Termwise Higher Derivative of Fourier Series of csc x

As seen in 5.5.2 , csc x is expanded to Fourier series in a broad meaning.as follows.
cscx = 2(sin x + sin 3x + sin 5x + sin 7x +)
-2i(cosx + cos 3x + cos 5x + cos 7x +)
Although this does not hold as an equation, differentiating the real part of the both sides repeatedly,
we obtain the following perfunctory expressions..

- 17 -
 (csc x)(1) = 2(1cosx +3cos 3x +5cos 5x +7cos 7x +-)
(csc x)(2) = -212sin x +32sin 3x +52sin 5x +42sin 7x +-
(csc x)(3) = -213cosx +33cos 3x +53cos 5x +43cos 7x +-
(csc x)(4) = 214sin x +34sin 3x +54sin 5x +44sin 7x +-


(csc x)(2n -1) = (-1)n-12Σ(2k +1)2n-1cos(2k +1)x (2.2n-1)
k=0

(csc x)(2n) = (-1)n2Σ(2k +1)2nsin(2k +1)x (2.2n)
k=0

Of course, these also do not hold as an equation. However, if the calculation is advanced assuming that
thiese hold, interesting results are obtained as follows.

10.5.3 Dirichlet Even Beta (minus)

Formula 10.5.3
 (-1)k
Let  (n ) =Σ n
, and let Bernoulli number B2n and Euler number Er are as follows.
k=0 (2k +1)
B0=1, B2=1/6, B4=-1/30, B6=1/42, B8=-1/30, 
E0=1, E2=-1, E4=5, E6=-61, E8=1385, 
Then the following expressions hold.

22k -2B2k
(-1)n
 

2n+1  2k E2n
  2
2
 (-2n) = (2n)!+ Σ =
2 k =n+1 2k(2k -2n -1)! 2

Proof
From (0.n) , (1.n) and (2.2n) , we obtain the following perfunctory expression.


22k -2B2k 2k
(-1)n

 
Σ (2n)! + Σ
2n
(2k +1) sin(2k +1)x = x
k=0 2x 2n+1 k=n +1 2k (2k -2n -1)!
(-1)n n
= Σ
2sin x r=0 2n E r(cotx)2n-2r
If x = /2 is substituted for this, the left side is as follows.
 (2k +1) 
Σ(2k +1) sin
2n
= Σ(-1)k (2k +1)2n = (-2n)
k=0 2 k=0
Then,

22k -2B2k
(-1)n
(2n)! + Σ 

2n+1  2k

  2
2
 (-2n) =
2 k =n+1 2k(2k -2n -1)!
(-1)n n (-1)n
Σ
2n-2r
= 2n Er 0 = 2n E n00
2 r=0 2
E2n
 (-1) 2nE n = E2n
n
=
2

- 18 -
Example
22k -2B2k
  
3   2k E2
 2
1 2 1
 (-2) = - 2! + Σ = =-
2 k =2 2k(2k -3)! 2 2
2 -2B  
2  Σ 2k(2k -5)!  2   2
5 2k 2k
1 2  E 5
 (-4) =
2k 4
4! + = =
k =3 2

- 19 -
10.6 Termwise Higher Derivative of csch x

10.6.0 Higher Derivative of csch x

Accordong to Formula 9.2.9 ( 9.2 ), Higher Derivative of csch x is expressed as follows.

(n) (-1)n n/2

(csch x) = (-1)r n E r(coth x)n-2r
Σ (0.n)
sinh x r=0

Where,  is floor function and n Er are the same as the coefficients in 10.5.0.

10.6.1 Termwise Higher Derivative of Taylor Series of csch x

Formula 10.6.1
When B0=1, B2=1/6, B4=-1/30, B6=1/42,  are Bernoulli Numbers and  is ceilling function,

the following expressions hold for 0<x < .

1! 22k -2B2k 2k-2

(csch x) (1)
=- -Σ x
x2 k =1 2k (2k -2)!

2!  22k -2B2k 2k-3

(csch x) (2)
= -Σ x
x3 k =2 2k (2k -3)!


(n) n!  22k -2B2k n-1
(csch x) = (-1) n
n+1
-Σ x 2k- (1.n)
x n +1 2k(2k - n -1)!
k= 
2

Proof
2 -2B2k 2k
2k

xcsch x = 1 -Σ x 0< x < 
k=1 (2k)!
From this
 2 -2 B 2k
2k
0!
csch x = -Σ x 2k-1
x k=1 2k ( 2k -1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of tan x. )

10.6.2 Termwise Higher Derivative of Fourier Series of csch x

Formula 10.6.2
The following expressions hold for x >0 .
(csch x)(1) = -21e -1x +3e -3x +5e -5x +7e -7x +
(csch x)(2) = 212e -1x +32e -3x +52e -5x +42e -7x +


(csch x) n = (-1)n2Σ(2k +1)n e -(2k +1)x
( )
(2.n)
k=0

Proof
csch x is expanded to Fourier series for x >0 as follows.

- 20 -
 2 2e -x
csch x = = = 2e -x1 + e -2x + e -4x + e -6x +
e x- e -x 1- e -2x

= 2e -1x + e -3x + e -5x + e -7x + (2.0)

= 2(cosh ix +cosh 3ix +cosh 5ix +)- 2i(sinh ix +sinh 3ix +sinh 5ix +)
Differentiating both sides of (2.0) with respect to x repeatedly, we obtain the desired expression.

10.6.3 Exponential series and Bernouilli series

From (0.n) , (1.n) and (2.n) , we obtain the following formula.

Formula 10.6.3
When ,  B2n
are Bernoulli Numbers and n Er are
are ceilling function and floor function respectivly,
the coefficients mentioned in 10.5.0 , the following expressions hold for 0 < x <  .


22k -2B2k

 (2k +1)
n 
1
Σ (2k +1)x
= n+1 n!-(-1) n
n +1 2k(2k - n -1)!
x 2kΣ
k =0 e 2x k= 
2
1 n/2
= Σ (-1)r n E r(csch x)n-2r
2sinh x r=0
Example

 
2 -2B2k 2k
2k
11 31 51 71 1  coth x
1

1x + 3x
+ 5x
+ 7x
+ = 2
1!+Σ x =
e e e e 2x k =1 2k (2k -2) ! 2sinh x

 
2 -2B
2k
12 32 52 72 1  2coth x - coth x
2k
2 0
2!-Σ
2k
1x + 3x
+ 5x
+ 7x
+ = 3
x =
e e e e 2x 2k (2k -3) !
k =1 2sinh x

Exponential series and Factorial

Giving x =1 to Formula 10.6.3, we obtain the following special values.

Formula 10.6.3'
When ,  B2n
are Bernoulli Numbers and n Er are
are ceilling function and floor function respectivly,
the coefficients mentioned in 10.5.0 , the following expressions hold.

 
2 -2B2k
2k
 2k +1n 1  1 n/2
Σ
k =0 e
(2k +1)
=
2
n!-(-1) n
Σ
n +1 2k(2k-n -1)!
= Σ
2sinh 1 r=0
(-1)rnE r(coth 1) n-2r
k= 
2

Example


 2 -2B2k

2k
11 31 1 51 71 coth 1
1

1 + 3 + 5 + 7 + = 2 1! +Σ 2k(2k-2)! =
2sinh 1
e e e e k=1

2
2 -2B
2k(2k-3)! 
2 2 2 2 2k 2 0
1 3 5 7 1 2coth 1 - coth 1  2k
+ 1+ + + =
3
2! - Σ
5 7
=
e e e e 2sinh 1 k =2

Therefore also, this leads to the following expression conjointly with Formula 10.4.3'.

2n 4n 6n 8n n! (-1)n  22k B2k

2
+ 4 + 6 + 8 + =
2
+
2 2k(2k - n -1)! Σ (3.e')
e e e e n +1
k= 
2

- 21 -
10.7 Termwise Higher Derivative of sec x

10.7.0 Higher Derivative of sec x

Accordong to Formula 9.2.8 ( 9.2 ), Higher Derivative of sec x is expressed as follows.
1 n/2
(secx) n = n-2r
()
Σ
cosx r=0 n E r(tan x) (0.n)

Where,  is floor function and n Er are the same as the coefficients in 10.5.0.

10.7.1 Termwise Higher Derivative of Taylor Series of sec x

Formula 10.7.1
When E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are Euler Numbers and  is floor function,

the following expressions hold for |x |<  /2.

 E2k
(sec x)(1) =Σ x 2k-1 (1.1)
k =1 (2k -1)!
 E2k
(sec x)(2) =Σ x 2k-2 (1.2)
k =1 (2k -2)!
 E2k
(sec x)(3) =Σ x 2k-3 (1.3)
k =2 (2k -3)!
 E2k
(sec x)(4) =Σ x 2k-4 (1.4)
k =2 (2k -4)!

 E2k
(sec x) n =
()
Σ
n +1 (2k -n)!
x 2k-n (1.n)
k= 
2

Proof
sec x is expanded to Tylor series as follows.
 E2k 
secx =Σ x 2k |x| <
k=0 (2k)! 2
Differentiating both sides of this with respect to x repeatedly, we obtain (1.1) ~ (1.4).
Here, considering the relation between the derivative order n and the first term n of Σ, it is as follows.

n 0 1 2 3 4 
k0 0 1 1 2 2 

n +1
Such a relation can be expressed by k0 =  using a floor function x  = x . Then we obtain
2
the desired expression.

Sum of Taylor series of the higher derivative of sec x

From (0.n) , (1.n) , we obtain the following expressions for |x|<  /2 .
 E2k 1 n/2 n-2r
Σ
n +1 (2k -n)!
x 2k-n = Σ
cosx r=0 n E r(tan x) (1.t)
k= 
2

- 22 -
And giving x = /4 to this, we obtain the following special values.
E2k  2k-1
 
 0
Σ = 2 Σ n E r = 2 1
k=1 (2k -1)! 4 r=0

E2k  2k-1
 
 1
Σ = 2 Σ n E r = 2 3
k=1 (2k -1)! 4 r=0

E2k  2k-3

 
 1
Σ = 2 Σ n E r = 2 11
k=2 (2k -3)! 4 r=0

E2k  2k-4

 
 2
Σ = 2 Σ n E r = 2 57
k=2 (2k -4)! 4 r=0


- 23 -
10.8 Termwise Higher Derivative of sech x

10.8.0 Higher Derivative of sech x

Accordong to Formula 9.2.9 ( 9.2 ), Higher Derivative of sech x is expressed as follows.

(n) (-1)n n/2

(sech x) = (-1)r n E r(tanh x)n-2r
Σ (0.n)
cosh x r=0

Where,  is floor function and n Er are the same as the coefficients in 10.5.0.

10.8.1 Termwise Higher Derivative of Taylor Series of sech x

Formula 10.8.1
When E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are Euler Numbers and  is floor function,

the following expressions hold for |x |<  /2.

 E2k
(sech x)(1) =Σ x 2k-1
k=1 2 -1
( k ) !
 E2k
(sech x)(2) =Σ x 2k-2
k=1 (2k -2)!

 E2k
(sech x) n =
( )
Σ
n +1 (2k -n)!
x 2k-n (1.n)
k= 
2

Proof
sech x is expanded to Tylor series as follows.
 E2k 2k 
sech x =Σ x |x| <
k=0 (2k)! 2
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of sec x. )

10.8.2 Termwise Higher Derivative of Fourier Series of sech x

Formula 10.8.2
The following expressions hold for x >0 .
(sech x)(1) = -21e -1x -3e -3x +5e -5x -7e -7x +-
(sech x)(2) = 212e -1x -32e -3x +52e -5x -42e -7x +


(sech x) n = (-1)n2Σ(-1)k (2k +1)n e -(2k +1)x
( )
(2.n)
k=0

Proof
sech x is expanded to Fourier series for x >0 as follows.
-x
2 2e
sech x = = = 2e -x 1 - e -2x + e -4x - e -6x +-
e x + e -x 1 +e -2x

- 24 -
 = 2e -1x - e -3x + e -5x - e -7x +- (2.0)

= 2(cos 1ix - cos 3ix + cos 5ix - cos 7ix +- )

+ 2i(sin 1ix - sin 3ix + sin 5ix - sin 7ix +- )
Differentiating both sides of (2.0) with respect to x repeatedly, we obtain the desired expression.

10.8.3 Exponential series and Euler series

From (0.n) , (1.n) and (2.n) , we obtain the following formula.

Formula 10.8.3
When  is floor function, E2n are Euler Numbers and n Er are the coefficients mentioned in 10.5.0 ,

the following expressions hold for 0 < x <  /2

 (2k +1)n (-1)n  E2k
Σ(-1)k x
=
2 Σ 2 -
( k n) !
x 2k-n
k=0 e(2k +1) n +1
k= 
2

1 n/2
= Σ (-1)r n E r(tanh x)n-2r
2cosh x r=0

Example1
11 31 51 1  E2k71 tanh 1x
- + - +- = - Σ x 2k-1
=
e 1x e 3x e 5x e 7x 2 k =1 (2k -1)! 2cosh x
12 32 52 72 1  E2k 2 0
2tanh x - tanh x
2Σ
2k-2
- + - +- = x =
e 1x e 3x e 5x e 7x k=1 (2k -2)! 2cosh x

Example2 n=3, x=1.4

Dirichlet Even Beta (minus)

Giving x =0 to Formula 10.8.3 , we obtain Dirichlet Even Beta (minus).

Formula 10.8.3'
 (-1)k
When  (n) =Σ and E0=1, E2=-1, E4=5, E6=-61, E8=1385,  are
k =0 (2k +1) n
Euler Numbers, the following expressions hold.
 (-2n +1) = 0 (3.2n-1')

- 25 -
 E2n
 (-2n) = (3.2n')
2

Proof
Sbstitute x =0 for Formula 10.8.3 . Then, the Fourier series is

 (2k +1)n  (-1)k

Σ (-1) k
=Σ =  (-n)
k =0 e(2k +1)0 k =0 (2k +1)-n
The Taylor series is
when n =2m -1
(-1)n  E2k
2 Σ
n +1 (2k -n)!
02k-n
k= 
2

(-1)2m-1  E2k
=
2 Σ
2m -1+1 2k -(2m -1)!
02k-(2m-1)
k= 
2

1  E2k
2Σ
=- 02k-2m+1 = 0
k=m (2k -2m +1)!
when n =2m
(-1)n  E2k (-1)2m  E2k
Σ Σ
2k-n
0 = 02k-2m
2 n +1

(2k -n )! 2 2m +1

(2k -2m )!
k= k=
2 2

1  E2k 1 E2m 0
2Σ
2k-2m
= 0 = 0
k=m (2k -2m)! 2 0!
E2m
=
2
Replacing m with n , we obtain (3.2n-1') and (3.2n') .

Note
These results are obtained also from the polynomial in Formula 10.8.3 ..

2006.11.11
2012.07.04 Renewal
K. Kono
Alien's Mathematics

- 26 -
11 Termwise Higher Derivative (Inv-Trigonometric, Inv-Hyperbolic)

11.1 Termwise Higher Derivative of Inverse Trigonometric Functions

11.1.1 Termwise Higher Derivative of arctan x , arccot x

Formula 11.1.1
When  is ceilling function, the following expressions hold for |x |< 1 .
(n)  (2k)! n
tan x Σ (-1)k
-1
= x 2k+1- (1.t)
n
k=
-1

(2k +1- n)!
2

(n)  (2k)! 2k+1- n

cot x Σ (-1)k
-1
=- x (1.c)
k=
n -1 (2k +1- n)!

2

Proof
arctan x is expanded to Tylor series as follows.
 (2k)! 2k+1
tan -1x = Σ(-1)k x |x| < 1
k =0 (2k +1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the following.
(1)  (2k)! 2k
tan
-1
x = Σ(-1)k x
k=0 (2k)!
(2)  (2k)! 2k-1
tan
-1
x = Σ(-1)k x
k=1 (2k -1)!
(3)  (2k)! 2k-2
tan
-1
x = Σ(-1)k x
k=1 (2k -2)!
(4)  (2k)! 2k-3
tan
-1
x = Σ(-1)k x
k=2 (2k -3)!

Here, considering the relation between the derivative order n and the first term k0 of Σ, it is as follows.

n 0 1 2 3 4 5 
k0 0 0 1 1 2 2 
n -1
Such a relation can be expressed by k0 =  using a ceiling function x = x. Then we obtain
2
(1.t). And,

 x 0 : +
cot -1x = 
2
- tan -1x
 x <0 : - 
Therefore, we obtain (1.c) immediately.

Sum of Taylor series of the higher derivative of arctan x

Formula 9.2.7 in "9 Higher Derivative" was as follows.
(n) (n -1)! n /2
n+1-2r
tan x = (-1)n nΣ
(-1)rnＣ n+1-2r x
-1

x +1
2 r=1

-1-
From (1.t) and this , the following equation follows for |x|< 1 .
 (2k)! (n -1)! n /2
n
Σ (-1)k x 2k+1- = (-1)n n Σ (-1)rnＣ n+1-2r x n+1-2r
k=
n -1

(2k +1- n)! x +1 r=1
2
2

And giving x =1 to this without considering the convergence condition, we obtain the following special value.
 (2k)! (n -1)! n /2
Σ
n -1
(-1)k
(2k +1- n)!
= (-1)n
2n Σr=1
(-1)rnＣ n+1-2r (1.t')
k= 
2

Example
0! 1
1 - 1 + 1 - 1 +-  1 1Ｃ 0
= =
2 2
1! 1
2 - 4 + 6 - 8 +-  = 2Ｃ 1 =
22 2
2! 1
12 - 34 + 56 - 78 +-  = - 3 3Ｃ 2-3Ｃ 0 = -
2 2
3!
234 - 456 + 678 - 8910 +-  = - 4 4Ｃ 3-4Ｃ 1 = 0
2

11.1.2 Termwise Higher Derivative of arcsin x , arccos x

Formula 11.1.2
When  is ceilling function, the following expressions hold for |x |< 1.
2
(n)  (2k -1)!!
sin x Σ
-1
= x 2k+1-n (2.s)
k=
n -1

(2k +1- n)!
2
2
(n)  (2k -1)!!
cos x Σ
-1
=- x 2k+1-n (2.c)
n -1
k= 
(2k +1- n)!
2

Proof
 (2k -1)!!  (2k -1)!!(2k)!
sin -1x = Σ x 2k+1 = Σ x 2k+1
k =0 (2 k)!! (2k +1) k=0 (2k)!! (2k +1)!
2
 (2k -1)!!
=Σ x 2k+1  (2k)! = (2k)!!(2k -1)!!
k =0 (2k +1)!
Differentiating both sides of this with respect to x repeatedly, we obtain (2.s). ( The number of the first term
of Σ is the same as it of arctan x. )

cos x =  /2 - sin x
-1 -1
(2.c) is obtained immediately from .

Formula 11.1.2'

  (2r -1)!!(4n -2r -3)!! = 0

2n -1 2n -1
Σ (-1)r
r=0 r
(2.so)

-2-
 r
2n 2n
Σ
r=0
(-1)r (2r -1)!!(4n -2r -1)!! = 22n(2n -1)!!2 (2.se)

Proof
Formula 9.3.2' in "9 Higher Derivative" was as follows.

r
(n ) 1 n -1 n -1 (2r -1)!!(2n -2r -3)!!
sin x Σ(-1)r
-1
=
2n-1 1-x 2 r=0 (1+x)r(1-x)n-1-r
From this,

 
1 2n -1
2n -1
x=0 =
(2n )
sin x Σ r
-1
(-1) (2r -1)!!(4n -2r -3)!!
22n-1 r=0 r

 
1 2n 2n
x=0
(2n +1)
sin x = 2n Σ(-1)r
-1
(2r -1)!!(4n -2r -1)!!
2 r=0 r
From (2.s) ,
2
 (2k -1)!!
x=0 = Σ (2 +1-2 )! 02k+1-2n = 0
(2n )
sin x
-1
k =n k n
2 2
 (2k -1)!! (2n -1)!!
x=0 = Σ (2 +1-2 -1)! 0
(2n +1)
sin x
-1 2k-2n
= 00
k =n k n 0!
Thus, we obtain the desired expressions.

Example
3Ｃ 0(-1)!!5!!- 3Ｃ 11!! 3!!+ 3Ｃ 23!!1!!- 3Ｃ 35!!(-1)!! = 0
4Ｃ 0 (-1)!!7!!- 4Ｃ 1 1!!5!!+ 4Ｃ 2 3!!3!!- 4Ｃ 3 5!!1!!+ 4Ｃ 4 7!!(-1)!! = 144

11.1.3 Termwise Higher Derivative of arccsc x , arcsec x

Formula 11.1.3
The following expressions hold for |x |> 1 .
(n) 
(2k -1)!! (2k +n)! -2k-n-1
csc x
-1
= (-1)nΣ x (3.c)
k=0 (2k)!! (2k +1)!
(n)  (2k -1)!! (2k +n )!
sec x
-1
= (-1)n-1Σ x -2k-n-1 (3.s)
k=0 (2k)!! (2k +1)!

Proof
1  (2k -1)!!
csc -1x = sin -1 =Σ x -2k-1 |x|>1
x k=0 (2k)!!(2k +1)
Differentiating both sides of this with respect to x repeatedly, we obtain (3.c).

sec x =  /2 - csc x
-1 -1
(3.s) is obtained immediately from .

-3-
11.2 Termwise Higher Derivative of Inverse Hyperbolic Functions

11.2.1 Termwise Higher Derivative of arctanh x , arccoth x

Formula 11.2.1t
When  is ceilling function, the following expressions hold for |x |< 1 .
(n)  (2k)! n
tanh x Σ
-1
= x 2k+1- (1.t)
n -1
k= 
(2k +1- n)!
2

Proof
arctanh x is expanded to Tylor series as follows.
 (2k)! 2k+1
tanh -1x = Σ(-1)k x |x| < 1
k=0 (2k +1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of arctan x. )

Formula 11.2.1c
The following expression holds for |x |> 1 .
(n)  (2k + n)! -2k-1- n
coth
-1
x = (-1)(n)Σ x (1.c)
k =0 (2k +1)!

Proof
arccoth x is expanded to Tylor series as follows.
 (2k)! -2k-1
coth -1x = Σ x |x| > 1
k =0 (2k +1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression.

11.2.2 Termwise Higher Derivative of arcsinh x , arccosh x

Formula 11.2.2s
When  is ceilling function, the following expression holds for |x |< 1 .
2

(n) k (2k -1)!!
sinh x Σ
-1
= (-1) x 2k+1-n (2.s)
n -1
k= 
(2k +1- n)!
2

Proof
arcsinh x is expanded to Tylor series for |x |< 1 as follows.
 (2k -1)!!  (2k -1)!!(2k)! 2k+1
sinh -1x =Σ(-1)k x 2k+1 =Σ(-1)k x
k =0 (2k)!! (2k +1) k =0 (2k)!! (2k +1)!
2

k (2k -1)!!
=Σ(-1) x 2k+1  (2k)! = (2k)!!(2k -1)!!
k =0 (2k +1)!
Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression. ( The number
of the first term of Σ is the same as it of arctan x. )

-4-
Sum of Taylor series of the higher derivative of arcsin x
Formula 9.4.2 in "9 Higher Derivative" was as follows.

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
(n) n-1
n/2 n -1
sinh x Σ r
-1
= (-1) (-1) 1
r=0 n -1-2r n-r-
2
x +1
2

From (2.s) and this , the following equation follows for |x|< 1 .
2
 (2k -1)!!
Σ
n -1
(-1)k
(2k +1- n)!
x 2k+1-n
k= 
2

(2r -1)!!(2n -3-2r)!!x n-1-2r

 
n-1
n/2 n -1
= (-1) Σ
r=0
(-1) r
n -1-2r n-r-
1
2
x +1
2

And giving x =1 to this without considering the convergence condition, we obtain the following special value.
2 n/2

 n -1-2r 

k (2k -1) !!
n -1 (2r -1)!!(2n -3-2r)!!
Σ
n -1
(-1)
(2k +1-n ) !
= (-1)n-1 Σ (-1)r
r=0 n-r-
1
k=  2
2 2

Example
(-1)!!2 3!!2 5!!2 7!!2 2
- + - +- =
0! 2! 4! 6! 2
1!!2 3!!2 5!!2 7!!2 2
- + - +- =
1! 3! 5! 7! 4
1!!2 3!!2 5!!2 7!!2 2
- + - +- =-
0! 2! 4! 6! 8
3!!2 5!!2 7!!2 9!!2 3 2
- + - +- =
1! 3! 5! 7! 16

Formula 11.2.2c
The following expressions hold for |x |> 1 .
(n) (n -1)! n-1  (2k + n -1)! -2k-n
cosh
-1
x = (-1)n-1 + (-1) Σ 2 x (2.c)
xn k =1 (2k)!!

Proof
arccosh x is expanded to Tylor series for |x | 1 as follows.

(2k -1)!! -2k  (2k -1)!!(2k -1)!
cosh -1x = log 2x -Σ x = log 2x -Σ x -2k
k=1 (2k) !!2k k =1 ( 2k)!!(2k) !
 (2k -1)!
= log 2x -Σ 2x
-2k
 (2k )! = (2k )!!(2k -1)!!
k =1 (2k)!!

Differentiating both sides of this with respect to x repeatedly, we obtain the desired expression.

-5-
11.2.3 Termwise Higher Derivative of arccsch x , arcsech x

Formula 11.2.3
When  is ceilling function, the following expressions hold for 0<x <1 .
2
(n -1)! 
(n) k (2k -1)!!
csch
-1
x = (-1) n
- Σ (-1)
n
x 2k-n (3.c)
x n 2k (2k - n)!
k= 
2
2
(n) (n -1)!  (2k -1)!!
sech
-1
x = (-1) n
-Σ
n
x 2k-n (3.s)
x n 2k (2k - n )!
k= 
2

Proof
2  (2k -1)!! 2k
csch -1x = log +Σ(-1)k-1 x
x k=1 (2k)!! 2k
2  (2k -1)!!(2k)! 2k
= log -Σ(-1)k x
x k =1 (2k)!! 2k (2k)!
2

k (2k -1)!!
= log 2 - log x -Σ(-1) x 2k  (2k )!=(2k )!!(2k -1)!!
k=1 2k k
(2 )!
Differentiating both sides of this with respect to x repeatedly, we obtain the following.
2
0! 
-1 (1) k (2k -1)!!
csch x = - 1
- ( -1 ) Σ 2k (2k -1)!
x 2k-1
x k =1

2
1! 
(2) k (2k -1)!!
csch
-1
x = 2
- Σ(-1) x 2k-2
x k=1 2k (2k -2)!
2
2! 
(3) k (2k -1)!!
csch x - Σ(-1)
-1
=- 3 x 2k-3
x k =2 2k (2k -3)!
2
(4) 3!  (2k -1)!!
csch
-1
x = 4
- Σ(-1)k x 2k-4
x k=2 2k (2k -4)!


Here, considering the relation between the derivative order n and the first term k0 of Σ, it is as follows.

n 0 1 2 3 4 
k0 1 1 1 2 2 
n
Such a relation can be expressed by k0 =  using a ceiling function x = x. Then we obtain
2
(3.c) . Next,
2  (2k -1)!! 2  (2k -1)!!(2k)! 2k
sech -1x = log +Σ x 2k = log -Σ x
x k=1 (2k)!! 2k x k =1 (2k)!! 2k (2k)!
2
 (2k -1)!!
= log 2 - log x -Σ x 2k  (2k )!=(2k )!!(2k -1)!!
k =1 2k (2 k)!
Differentiating both sides of this with respect to x repeatedly, we obtain (3.s) . ( The number of the first term
of Σ is the same as it of arctan x. )

-6-
Sum of Taylor series of the higher derivative of arccsch x
Formula 9.4.2 in "9 Higher Derivative" was as follows.

(n) n-1
n (-1)r n Ar
csch x Σ
-1
= (-1) 1
r=1 r-
2
x +1
n-1+2r -2
x
where n Ar are coefficients as follows.

1A1 1
2A1 2A2 2 1
3A1 3A2 3A3 6 7 3
4A1 4A2 4A3 4A4 = 24 48 45 15
5A1 5A2 5A3 5A4 5A5 120 360 549 390 105
6A1 6A2 6A3 6A4 6A5 6A6 720 3000 6570 7425 4200 945
 
From (3.s) and this , the following equation follows for 0<x <1 .

 

k (2k -1)!!
2
(-1)n n (-1)r n Ar
Σn (-1) (n -1)! +Σ
2k-n
x =
k= 
2k (2k - n)! xn r=1 r-
1
2
x 2r-1 x -2 +1
2

And giving x =1 to this without considering the convergence condition, we obtain the following special value.

 

k (2k -1)!!
2 n (-1)r n Ar
Σ (-1)
n 2k (2k - n)!
= (-1) (n -1)! + 2 Σ
n
r=1 2r
(3.c')
k= 
2

Example
1!!2 3!!2 5!!2 7!!2 2
- + - +- = 0! -
21! 43! 65! 87! 2
1!!2 3!!2 5!!2 7!!2 3 2
- + - +- = -1! +
20! 42! 64! 86! 4
3!!2 5!!2 7!!2 9!!2 13 2
- - - +- = -2! +
41! 63! 85! 107! 8
3!!2 5!!2 7!!2 9!!2 75 2
- - - +- = 3! -
40! 62! 84! 106! 16

2006.11.20
2012.07.08 Renewal
K. Kono
Alien's Mathematics

-7-
12 Super Derivative (Non-integer times Derivative)

12.1 Super Derivative and Super Differentiation

Defintion 12.1.1
(p)
f (x) obtained by continuing analytically the index of the differentiation operator of Higher Derivative of
a function f(x ) to a complex plane [0 , p ] from a natural number interval [1 , n ] is called Super Derivative
of f(x) .

Example
p
(sin x) p = sin x +   + c (x)
()
p cp(x) is an arbitrary function.
2

12.1.2 Super Differentiation

Definition 12.1.2
We call it Super Differentiation to differentiate a function f with respect to an independent variable x
non-integer times continuously. And it is described as follows.
p

 
d d d d
f(x) =  f(x) : p pieces
dx p dx dx dx

Example
dp p
dx p
cosx = cos x +
2  
12.1.3 Fundamental Theorem of Super Differentiation
The following theorem holds from Theorem 7.1.3 in 7.1.

Theorem 12.1.3
r [0 , p] be an continuous function on the closed interval I and be arbitrary the r-th order
(r)
Let f
derivative function of f . And let a(r ) be a continuous function on the closed interval [0 , p ] .
Then the following expression holds for a(r), x  I .

 
dp d p p -1 (r) x x
(p)
f(x) = f (x) + Σ f a(p -r)  dx r (1.1)
dx p dx p r=0 a(p) a(p -r)
Especially, when a(r)= a for all k [0 , p] ,
dp (p) d p p -1 (r) (x -a)r
f(x) = f (x) + Σf (a) (1+ r) (1.2)
dx p dx p r=0

Proof
Theorem 7.1.3 in 7.1 can be rewritten as follows.

   
x x p -1 x x
<p>
f (x) =  f(x)dx + Σ f
p <p -r>
a(p -r)  dx r
a(p) a(0) r=0 a(p) a(p -r+1)

-1-
Especially, when a(r)= a for all k [0 , p] ,

 f(x)dx
<p>
x x p -1
<p -r> (x -a)r
f (x) = n
+ Σf (a)
a a r=0 (1+ r)
Differentiating these both sides with respect to x p times,

 
d p <p> d p p -1 <p -r> x x
Σ dx r
<0>
pf
(x) = f (x) + p r=0 f a(p -r)
dx dx a(p) a(p -r)

d p <p> d p p -1 <p -r> (x -a)r

Σf
<0>
f (x) = f (x) + (a)
dx p dx p r=0 (1+ r)
Shifting by -p the index in the integration operator <> and replacing <> by differentiation operator ( ) ,
we obtain the desired expression.

Constant-of-differentiation Function
d p p -1
We call Σ etc. Constant-of-differentiation Function of f(x) . Since p is a real number,
dx p r=0
generally it is difficult to obtain this. However, it becomes easy exceptionally at the time of f(x)= e x .
That is, Constant-of-integration Function in 7.1.3 was as follows.

 
p -1
a (x -a )
r
 (x -a)r (x -a)r+ p
Σ
r=0
e
(1+ r)
=e a
Σ
r=0
-
(1+ r) (1+ r + p)
Differentiating both sides with respect to x p times,

 
dp (x -a)r
p -1  dp (x -a)r (x -a)r+ p
Σe a
dx p r=0 (1+ r)
=e Σ p
a
r=0 dx
-
(1+ r) (1+ r + p)
From the 12.3.1 mentioned later, the following expressions hold.

dp (1+r) r-p dp (1+r +p)

p x a
( - )r
= -
(x a) , (x -a)r+p = (x -a)r
dx (1+r -p) p
dx (1+r)
Substituting this for the above, we obtain the following expression.

 
d p p -1 a (x -a)r  (x -a)r-p (x -a)r
Σ
p r=0 e (1+ r)
= e a
Σ (1+r -p)
-
(1+r)
dx r=0

12.1.4 Lineal and Collateral

In the case of the higher differentiation, since the constant-of-integration polynomial was degree n -1 ,
the constant-of-differentiation function which differentiated this n times became 0. However, in the case of
the super differentiation, since the constant-of-integration function is expressed by a series in general, the
constant-of-differentiation function which differentiated this p times does not become 0. This shows that
there are lineal and collateral in the super differentiation.

Definition 12.1.4

 
dp d p p -1 (r) x x
(p)
f(x) = f (x) + Σ f a(p -r)  dx r (1.1)
dx p dx p r=0 a(p) a(p -r)

In this expression,
when Constant-of-differentiation Functin is 0,

-2-
 dp
we call f(x) Lineal Super Differentiation and
dx p
we call the function equal to this Lineal Super Derivaive Function.

when Constant-of-differentiation Functin is not 0,

dp
we call f(x) Collateral Super Differentiation and
dx p
we call the function equal to this Collateral Super Derivaive Function.
These are the same also in (1.2) .

In short , Lineal Super Derivaive Function is what differentiated f(x) with respect to x continuously
without considering the constant-of-differentiation function.

Example: lineal derivative and collateral derivative of ex

In the case of easier fixed lower limit, from (1.2) in the theorem

dp x d p p -1 a (x -a)r
pe
= e + x
Σ
p r=0 e (1+ r)
dx dx
Here, using the former expression i.e.

 
d p p -1 a (x -a)r  (x -a)r-p (x -a)r
Σ e = e a
Σ -
dx p r=0 (1+ r) r=0 (1+r -p) (1+r)
we obtain

  = e Σ 
dp x  (x -a)r-p (x -a)r  (x -a)r-p
e = e +e Σ
x a
- a
dx p r=0 (1+r -p) (1+r) r=0 (1+r -p)
When a - , since the constant-of-differentiation functin can not be 0, this is a collateral differentiation.

When a =-, since e a =0 , we obtain the following lineal differentiation.

dp x
pe
= ex
dx

When p =1/2 , a =0 , if the differential quotients on x =0.3 are compared with the calculation result
by Riemann-Liouville differintegral (later 12.2.1), it is as follows.

-3-
 And if the lineal super derivative and the collateral super derivative are illustrated side by side, it is as follows.

Remark
It is thought that this collateral super derivative is an asymptotic expansion. And this collateral super derivative
is corresponding with the termwise super differentiation. In general, a termwise super differentiation seems to
become a collateral super differentiation.

12.1.5 The basic formulas of the Super Differentiation

The following formulas hold like the higher differentiation.

c 0
(p) (p)
c f(x) =cf (x) : constant multiple rule
(p) (p) (p)
 f(x)+ g(x) = f (x) + g (x) : sum rule

-4-
12.2 Fractional Derivative

12.2.1 Riemann-Liouville differintegral

Among the super integrals of function f(x), the super integral whose lower limit function a(k ) is a constant
a was calculable by Riemann-Liouville integral. The super derivativeof such a function f(x) is calculable by
Riemann-Liouville integral and integer times differentiation. It is as follows.
Let n = p = ceil(p ) . First, integrate with f (x) n -p times. Next, differentiate it n times. And,
since the result is n- (n-p), it means that f(x) was differentiated p times.


(p) 1 dn x
f (x) = (x -t)n-p-1f(t)dt n = p (2.0)
(n -p) dx a n

This expression is called Riemann-Liouville differintegral. "differintegral" is a coined word which combined
"differential" and "integral".
Although the numerical integration and the numerical differentiation are possible for (2.0) with this, the
accuracy of numerical differentiation is bad and the desired result may not be obtained. In this case, the
following formula which replaced the calculation order of integration and differentiation is effective.

  
(p) 1 x
n-p-1 dn
f (x) = (x -t) f(t) dt n = p (2.0')
(n -p) a dt n
Although this formula has a possibility of cutting off a constant of integration as a result of differentiating
previously, in many cases, it is correctly calculable. This (2.0') is often used in the following chapters.

12.2.2 Riemann-Liouville differintegral expressions of super derivatives of elementary functions

Riemann-Liouville differintegral expressions of super derivatives of some elementary functions are as follows.
In the right side, super derivatives obtained by super differentiation are shown in advance. Needless to say,
Riemann-Liouville differintegral holds only if the lower limit function a(k) is a constant a . In addition, p is
a positive non-integer and n = p = ceil(p ) in all the expressions.

(1+) -p
dn
 (x -t)
1 x
x 
(p) n-p-1 
= t dt = x (  0)
(n -p) dx n 0  1+ -p 
(- + p) -p
 dn
 dt 
1 x
= (x -t)
n-p-1
t  dt = (-1)-p x ( < 0)
(n -p) 
n (-)

 (x -t)
 x (p) 1 dn x
t -p  x
e  e dt = (1) e
n-p-1
=
(n -p) dx n 

log x - (1- p)-  -p

 (x -t)
(p) 1 dn x
n-p-1
(log x) = log t dt = x
(n -p) dx n 0 (1- p)

Note
When n =p , -(n -p) <  <0 , the following expression does not hold.


 (p) 1 dn x
x  = (x -t)n-p-1 t  dt (2.1)
(n -p) dx n

n -p times integral of x  ( <0) is carried out, and super primitive function x  n-p
+
In this case,

 + n -p > 0 is obtained. According to this, the zero of the super primitive function changes from 
to 0. The integral with a fixed lower limit is inapplicable to such an integral.
In this case, the following formula which replaced the calculation order of integration and differentiation
is effective.

-5-
  (x -t)  
 (p) 1 x
n-p-1 dn 
x  = t dt n =[p] (2.1')
(n -p)  dt n
x ( -n <0) is carried out, and super primitive function
-n
If this formula is used, n -p times integral of
 -p
x ( -p < 0) is obtained. Since the zero of the super primitive function does not change, the integral
with a fixed lower limit is applicable. (See Example 5 in the following chapter).

Super Derivative & Fractional Derivative

In Super Derivative which I developed, first, we obtain the higher derivative, next, extending the index of
the operator to real number, we obtain the super derivative.
On the contrary, in traditional Fractional Derivative , the super derivative is directly drawn from Riemann-
Liouville differintegral. However, the calculation is very difficult.
Three examples are shown below. In each example, the 1st is Super Derivative , and the 2nd is Fractional
Derivative .

Example 1

 2
1
(1+1) (2) 2
1 1 1
1- 2 2
x 
1 2
= x = x = x
1+1-1/2 3/2 
 2
1 1

 (x -t)
1 d1 x 1- -1
x 
1 2 1
= t dt
(1-1/2) dx 1 0
1

 (x -t)
x

 
1 d x - 1 d 2
2 1
= t dt = - x -t (2x +t)
(1/2) dx 0  dx 3 0

 
3 1
1 d 4 2 2 2
= x = x
 dx 3 
Example 2

 2
1 1
- 1 -x
e -x
= (-1) 2 -x
e = e
i
 2
1 1

 (x -t)
1 d1 x 1- -1
e 
-x 2
= e -tdt
(1-1/2) dx 1 +
1


1 d x 1 d - x
-e  erfi x -t 
x
2 -t
= (x -t) e dt =
 dx +  dx
1 d 1 d
= e -x  erfi x - = e -xerfi x -
 dx i dx
1 d 1 d 1
= e -xerf-  = -e -x = e -x
i dx i dx i

Example 3

  log x - 1-1/2-  x - 2 = log x - 1/2- x - 2

1 1 1
(log x) 2 =
1-1/2 1/2

-6-
 log x - (- -2 log 2) - 
1 1
-
2 log x + 2 log 2 -
2
= x = x
 
 
1 1

 (x -t)
1 d1 x 1- -1
(log x) 2 = 2
log t dt
(1-1/2) dx 1 0
1

 (x -t)
1 d x -
2
= log t dt
 dx 0
x

  -2 
1 d x -t
= -4 x tanh -1 x -t (log t -2)
 dx x 0
Here

x n +x 1
tanh -1 = log = log(n + x) - log(n - x) |x|< n
n n -x 2
Then

 
x -t 1
tanh -1 = log x + x -t  - log x - x -t 
x 2
From this

 
x -t
4 x tanh -1 =2 x log x + x -t  - log x - x -t 
x
= 4 x log x + x -t  - 2 x log x + x -t  + log x - x -t 
= 4 x log x + x -t  - 2 x logx -(x -t)
= 4 x log x + x -t  - 2 x log t
Therefore

 
x -t
4 x tanh -1 + 2 x -t (log t -2)
x
= 4 x log x + x -t  - 2 x log t + 2 x -t log t - 4 x -t
= 4 x log x + x -t  - 2 x - x -t  log t - 4 x -t
Using this
x
 
1

-4  -2 
1 d x -t
(log x) 2 = x tanh -1
x -t (log t -2)
 dx x 0
1 d x
= -4 x log x + x -t  +2 x - x -t  log t +4 x -t 0
 dx
1 d
= -4 x log x + 2 x log x
 dx
1 d
+ 4 x log x + x -2 x - x  log 0-4 x
 dx
1 d
= 4 x log2 x  - 4 x  2 x log x = x log x
 dx

-7-
Furthermore

4 x log2 x - 4 x = 4 x log 2 + 4 x log x - 4 x

= 4 x log 2 + 2 x log x - 4 x
= 4 x (log 2-1) + 2 x log x
Thus

 2
1 1
1 d log x +2 log 2 - 2
(log x) = 4 x (log 2-1) +2 x log x = x
dx  

Reference
1

(x -t)
-
2 2
t dt = - x -t (2x +t)
3
1

 
- 2 z 2
e dt = -e  erf x -t 
2 t x
(x -t) erf(z)= e -t dt
 0
1

(x -t)
-
2
e -t dt = -e -x  erf i x -t  erf i(z)= erf(iz)/i
1

(x -t)  -2

-
2 x -t
log t dt = -4 x tanh -1 x -t (log t -2)
x

12.2.3 The demerit and the strong point of Fractional Derivative

As seen in three upper examples, Fractional Derivative based on Riemann-Liouville differintegral is difficult
like this. Although Example 3 was the 1/2 times derivative of a logarithmic function, for this calculation, the
delicate technique was used abundantly, and one day was required. When p is a real number, the p times
derivative of this function is hopelessly difficult. Thus, in Fractional Derivative , it is very difficult to obtain the
derivative from Riemann-Liouville differintegral. Moreover, how to take a lower limit is not clear, and it cannot
treat trigonometric functions etc. These are the same as that of Fractional Integral .
A peculiar problem to Riemann-Liouville differintegral is not to be able to use it when p is an integer. It is
because of becoming (0)= of n=p at this time. In this case, it will rely on the Higher Derivative.
However, (2.0) and (2.0') are the powerful tools of numerical computation and can obtain super derivative of
the arbitrary points of arbitrary functions easily. And super derivatrive can be verified numerically with these.

-8-
12.3 Super Derivative of Power Function

12.3.1 Formula of Super Derivative of Power Function

Analytically continuing the index of the differentiation operator in Formula 9.2.1 ( 9.2 ) to [0 , p ] from [1,n]
we obtain the following formula. In addition, Rieamnn-Liouville differintegrals are also expressed together

Formula 12.3.1
When (z) is gamma function and n = p is ceiling function, the following expressions hold

(1) Basic form

(1+) -p dn
 (x -t)
1 x
x 
(p) n-p-1 
= x = t dt (  0)
(1+ -p) (n -p) dx n 0

(- + p) -p
 (x -t)  dt 
dn
1 x
= (-1) -p
x = n-p-1
t  dt ( < 0)
(-) (n -p) 
n

(2) Linear form

-p
(1+ )
 
1
(ax +b)   (p)
= (ax +b)-p (  0)
a (1+ -p)


1 dn x
= (x -t)n-p-1(at +b) dt
(n -p) dx n b
-
a
-p
(- + p)
 
1
(ax +b)   (p)
= - (ax +b)-p ( < 0)
a (- )

 (x -t)  
1 x dn
= n-p-1
( + ) dt
n at b
(n -p)  dt

Caution !
-p

 
1 q
Do not describe to be a p in the upper formula. Because, since the law of exponents a p = a pq
a
does not hold for the arbitrary real numbers p, q at the time a < 0,
-p -p

   
1 1
= (a ) -1 -p
a (-1)(-p )
=a p
i .e .  ap
a a
-p

 
1
If it is described as = ap, the rotation direction on the complex plane becomes reverse and the
a
mistaken result is caused.
For example, when a = -3 , p =1/2 ,
1 1
- - 1 1
-p 2 2

     
1 1 1 -
= = (-1) = (-1) 2
3 2
= -i 3
a -3 3

1 1 1 1
= (-1)3
p 2 2 2 2
a = (-3) = (-1) 3 = i 3

-9-
Example 1  > p

 10 
1

(1+2)
1 19 19

 
x2 2-
10 100
= x = x 10 =0.547239 x 10
2 21+2-1/10 1719/10
 10 
9

(1+2)
9 11 11

 
2
x 2-
10 100
= x = x 10 = 0.955579 x 10
2 21+2-9/10 111/10
<1/10> <9/10>
When x 2/2 , x 2/2 , x 2/2 , x1 are drawn on a figure side by side, it is as follows.

Example 2  = p

1 (1) (1+1) 1-1 (2) 0 1 0

x  = x = x = x =1
1+1-1 1 1
 2
1

   
1 3
 1+ 

x 
1 1 1
2 2 -
2 2 2
= x = x0 =
1
 
1 1 2
 1+ -
2 2
Example 3 0   < p

 2
1
(1+0) (1) - 2
1 1 1
0- 1 -2
x 
0 2
= x = x = x
1+0-1/2 (1/2) 
1 (2) (1+1) 1-2 (2) -1 1
x  = x = x = x -1 = 0
1+1-2 (0) 

Example 4  < 0

 2
1
-(-1)+1/2 -1- 2  -2
1 1 3
-
x 
-1 2
= (-1) x = -i x
-(-1) 2
-1 (1) -(-1)+1 -1-1 (2) -2
x  = (-1)-1 x =- x = -x -2
-(-1) (1)
1 (2)
-(-1/2)+2 - 2 -2 (5/2) - 2
x 
1 5 5
-
2 -2 3 -
= (-1) x = x = x 2
-(-1/2) (1/2) 4

- 10 -
Example 5 -p - p <  < 0

 2
1

   x
1 1
 - - +
x 
1 1 1 5
(5/6) - 6
1
- - 3 2 - -
= -i
3 2 3 2
= (-1) x
(1/3)
  
1
 - -
3
5 5
1.1287 - 6 -
6
=- ix = -0.42135 i x
2.6789
 2
1

x 
1 1 1 4

 (x -t)   dt = - 3   (x -t)
1
-
3 1 x -
2 d1 - 3 1 x -
2
-
3
= t t dt
 
 1-
1  dt 1 
2
This last integral cannot be expressed with an elementary function. Then, if the value on x=1 is numerically
integrated by mathematical software, it is as follows. This result is consistent with the previous value
-0.42135i exactly.
Rie mann-Liouville diffe rinte gral
 a := -1/3: p := 1/2:
 df := diff(t^a,t)
1
 4
3t
3

 fl := x-> 1/gamma(ceil(p)-p)
*int((x-t)^(ceil(p)-p-1)*df, t=infinity..x)
x
1  p  p  1
x    (x  t)  df d t
( p  p)


 float(fl(1))
 0.4213560763  i

12.3.2 Half Derivative of a power function

Especially, Super Derivative of order 1/2 is called Half Derivative.

Formula 12.3.2
Let n be a non-negative integer, -1!!  1, (2n -1)!!  135(2n -1) ,
0!!  1, 2n !!  2462n ,
then following expressions hold.

(1) Basic form

 2
1 1
(2n)!! n-
x n
= x 2
(2n -1)!! 
 2
1

(n +1)(2n +1)!!  n
x 
1
n+
2
= x
2(n +1)!!

- 11 -
(2) Linear form

 2
1 1
(2n)!! a n-
(ax +b)  n 2
= (ax +b)
(2n -1)!! 
 2
1

(ax +b) 
1
n+ (n +1)(2n +1)!!
2
= a (ax +b)n
2(n +1)!!

Proof
Upper rows   0 of the linear form of Formula 12.3.1 were as follows.
1
 2
1 -
(1+n)
2 1

 
1 n-
(ax +b)n
2
= (ax +b)
 
a 1
 1+n -
2
 
1

  (ax +b)
1 1
2 -  1+n +
(ax +b) 
1 2

 
n+
2 1 2 n
=
a (1+n)
Here, when n is a non-negative integer,

 
1 (2n -1)!!
 n+ = n
 , (2n)!! = 2n n !
2 2
Then
(1+n) (1+n) 2n n ! (2n)!!
= = =
    (2n -1)!!  (2n -1)!! 
1 1
 1+n -  n+
2 2

  (1+n)+ 
1 1
 1+n +
2 2 1 2(1+n)-1 !!   
= =
(1+n) (1+n) 2 n! 1+n

(n +1)(2n +1)!!  (n +1)(2n +1)!! 

= =
2n+1 (n +1)! 2(n +1)!!

Substituting these for the previous formula, we obtain the linear form, andgiving a =1, b =0 to them,
we obtain the basic form.

Example 1

 2
1 1 1 1
0!! - 1 - -
x 
0
= x 2
= x 2
(  a 0x 2
)
(-1)!!  
 2
1 1 1 1
2!! 2
x 
1
= x 2
= x 2
(  a 1x ) 2
1!!  
 2
1 3 3 3
4!! 8
x 
2
= x 2
= x 2
(  a 2x ) 2
3!!  3 
 2
1 5 5 5
6!! 16
x 
3
= x 2
= x 2
(  a 3x ) 2
5!!  5 

- 12 -
Example 2

 2
1

11!!  0 
x 
1
2 1 0
= x = x0 (= x )
2!! 2 a1
 
1
2
23!!  1 3  1
x 
3
2 2 1
= x = x (= x )
4!! 4 a2
 2
1

35!!  2 15  2
x 
5
2 3 2
= x = x (= x )
6!! 16 a3
 2
1

47!!  3 35  3
x 
7
2 4 3
= x = x (= x )
8!! 32 a4

12.3.3 Half Derivative of an integer power function (Fractional Derivative)

Next, using Riemann-Liouville differintegral, we obtain the Half Derivative of an integer power function.

Formula 12.3.3
When n denots a natural number, the following expression holds.

 
1 1
2n +1 n (-1)k n
k   x
2 n-
x 
n
Σ
2
=
 k =0 2k +1

Proof
Let n be a natural number,  = n , p =1/2 . Then, since  > p , from Formula 12.3.1 we obtain
the following expression.

 
1 1

 (x -t) 
2 1 d1 x - 1 d x tn
x 
n 2 n
= t dt = dt
(1/2) dx 1 0  dx 0 x -t
Here, according to「岩波数学公式Ⅰ」p96, the following expressions hold

(x -t)n-r

tn
r 
2 x -t n n
dt = n+1 r=0
(-x)r
Σ
2n -2r +1
x -t (-1)
Then
x

 
(x -t)n-r

tn

x 2 x -t n n
dt = Σ (-x)r
0 x -t (-1)n+1 r=0 r 2n -2r +1
0
1
xn
r 
2 n n
Σ(-1)r
2
= x
(-1)n r=0 2n -2r +1
Differentiate both sides of this with respect to x as follows.

 
1


tn

d x d 2 n n 1 n+

dx
dt = n
dx (-1) r=0Σ (-1) r
r 2n r
-2 +1
x 2
0 x -t
1


n
 
2 1 n 1 n-
n n 2 Σ
r
= + (-1) x 2
(-1) r=0 r 2n -2r +1

- 13 -
 1
(-1)r-n
r  x
n n n-
= (2n +1)Σ 2
r=0 2n -2r +1
 2
1 1
(-1)r-n
r  x
2n +1 n n n-
 x 
n
= Σ
2
 r=0 2n -2r +1
Here, we devise further,
r-n
(-1)n-r (-1)k
r    k 
n (-1) n n n n n
Σ
r=0 2n -2r +1
=Σ
r=0 2(n -r)+1 n -r
=Σ
k =0 2k +1
Using this, we obtain

 
1 1
(-1)k
k   x
2 2n +1 n n n-
x 
n
Σ
2
=
 k =0 2k +1

By-product
Comparing Formula 12.3.2 and Formula 12.3.3 , we obtain the following formula.
k

k 
n (-1)n (2n)!!
Σ
k =0 2k +1
=
(2n +1)!!
This is the same as the by-product in 7.3.3 .

12.3.4 Fractional Derivative of an integer power function

Generalizing Formula 12.3.3 , we calculate a fractional derivative of an integer power function. First, we
prepare the following lemma.

Lemma
When m,n are natural numbers, the following expression holds.
m-1
1
(x -t)n-r

m(x -t) m n
r 
- n
m n
(x -t) t dt = n+1
(- x) r
Σ m(n -r)+(m -1)
(4.0)
(-1)
r=0

Proof
Let m-1
m
(x -t)n-r
r 
m(x -t) n n
F(t) = Σ (-x) r
(-1)n+1 r=0 m(n -r)+(m -1)
Differentiate this with respect to t. Then
m-1
m
(x -t)n-r
r 
d d m(x -t) n n
dt
F(t) =
dt (-1)n+1 Σ
r=0
(-x) r
m(n -r)+(m -1)
m-1
m
(x -t)n-r
r 
m(x -t) d n n
+ Σ (-x) r
(-1)n+1 dt r=0 m(n -r)+(m -1)
1
-
m
(x -t)n-r
r 
(m -1)(x -t) n n
=- Σ (-x) r
(-1)n+1 r=0 m(n -r)+(m -1)

- 14 -
 m-1
m
(x -t)n-r-1
r 
m(x -t) n n
- Σ (n -r)(-x) r
(-1)n+1 r=0 m(n -r)+(m -1)
1
-
m
(x -t)n-r
r 
(x -t) n n
=
(-1)n
Σ
r=0
(m -1)+m(n -r)(-x)
r
m(n -r)+(m -1)
1 1
- -
m m

r 
(x -t) n n (x -t)
= Σ (-x)r(x -t)n-r = (-t)n
(-1)n r=0 (-1) n

1
-
m n
= (x -t) t

Using this Lemma, we obtain the following formula.

Formula 12.3.4
When n, m are natural numbers such that n 1, m 2 , the following expressions hold.

 
1 1
(-1)k
k   x
m m(mn +m -1) n n n-
x 
(-1/m) Σ
n m
=- (4.1)
k =0 mk +(m -1)
k 1

(1-1/m) Σ mk +(m -1)  k 

mn +(m -1) (-1) n n n-
= x m
(4.1')
k =0
k
(1+ n , -1/m)
Σ mk +(m -1)  k  = - m(mn +m -1) ( )
n (-1) n
: beta function (4.2)
k =0

Proof
(1+ ) -p
 (x -t)
 (p) 1 dk x
k-p-1 
x  = x = t dt (  p)
1+ -p  (k -p) dx k 0

Give  = n , p =1/m to this. Then since k = 1/m=1 ,

 m
1 1

 (x -t)
1 d1 x -
x  n
= m n
t dt
(1-1/m) dx 1 0
1

 (x -t)
m d x -
m n
=- t dt
(-1/m) dx 0
Using the above Lemma, we calculate as follows.
x

 
m-1
1
(x -t)n-r
 (x -t)
m

r 
x - m(x -t) n n
m n
t dt = Σ (-x) r
0 (-1)n+1 r=0 m(n -r)+(m -1)
0
m-1
m n-r

r 
mx n n x
=
(-1)n
Σ
r=0
(-x) r
m(n -r)+(m -1)

- 15 -
 1
(-1)r-n
r  x
n n n- +1
= mΣ m
r=0 m(n -r)+(m -1)
Furthermore, since r,n are integers,
r-n
(-1)n-r
r   
n (-1) n n n
Σ
r=0 m(n -r)+(m -1)
=Σ
r=0 m(n -r)+(m -1) n -r
(-1)k

n n
=Σ
k =0 mk +(m -1) k
Using this,
1 1

 (x -t)
(-1)k
k   x
x - n n n- +1
m n
t dt = mΣ m
0 k =0 mk +(m -1)
Differentiating this with respect to x,
1 1

 (x -t)
(-1)k
k   x
d x - n n n-
m n
t dt = (mn +m -1)Σ m
dx 0 k =0 mk +(m -1)
Thus

 m
1 1
(-1)k
k   x
m(mn +m -1) n n n-
x 
(-1/m) Σ
n m
=- (4.1)
k =0 mk +(m -1)
Moreover, (4.1') follows immediately from this.
Next,

 
1
(1+ n) (1+ n) (-1/m)
1 1
m n- n-
x 
n
= m
x = x m
(1+ n -1/m)  (-1/m) (1+ n -1/m)
Since this have to be equal to (4.1) ,

(-1)k (1+ n) (-1/m)

k 
n n
-m(mn +m -1)Σ =
k =0 mk +(m -1) 1+ n -1/m 
From this

(-1)k (1+ n , -1/m)

k 
n n
Σ
k =0 mk +(m -1)
=-
m(mn +m -1)
(4.2)

Remark
(4.2) suggests that (4.1) can be expressed with a beta function and n,m can be real numbers. Actuality,
(1+ ) (1+ ) -p  (1+ , -p)
= =
(1+ - p) (-p) (1+ -p) (-p)
Then,
(p) (1+ )  (1+ , -p) -p
x   = x -p = x (  0)
(1+ -p) (-p)

12.3.5 Super Derivative of an integer power function

Replacing 1/m with p in Formula 12.3.4 , we obtain the following formula.

Formula 12.3.5
When n is a natural number, the following expressions hold for 0< p  n .

- 16 -
 n +1- p n (-1)k
k 
n (p)
n
x   x n-p
(1-p) Σ
= (5.1)
k =0 k +1- p
k

k +1- p  k 
n -p (-1) n -1
n -1
(n , -p) = - Σ ( )denots beta function. (5.2)
p k =0

Example 1
2  3 (1+2) 2
x  x 2- 3
= = x 2- 3 = 2.2151 x 2- 3
1+2- 3   3- 3 

    
2  3 3- 3 1 2 1 2 1 2
x  = - + x 2- 3
1- 3 1- 3 0 2- 3 1 3- 3 2
= 2.2151 x 2- 3

Example 2

   
2- 3 1 1 1 1
2, - 3 = - - = 0.788675
3 1- 3 0 2- 3 1

    
3- e 1 2 1 2 1 2
(3, -e) = - - + = -0.596137
e 1- e 0 2- e 1 3- e 2

12.3.6 Super Derivative of a positive power function

Since Formula 12.3.5 are binomial forms, the further generalization is possible.

Formula 12.3.6
The following expressions hold for p ,q such that 0< p  q .
r

r  x
(p) q +1- p  (-1) q q-p
x q
(1-p) Σ
= (6.1)
r=0 r +1- p
r

p Σ r +1- p  r 
q -p (-1) q -1 
(q , -p) = - ( )denots beta function. (6.2)
r=0

12.3.7 Super Derivative of a polynomial

m
m-k
In the case of a polynomial f(x) =Σcm-k x , as for the zero of the super derivative, it is good to
k =0
perform it as follows. It is based on experience of a writer.

(1) When f(x) is factored by primary formula ax +b , let -b /a be the zero.

(2) When f(x) is not factored by primary formula ax +b , let 0 be a the zero.

For example, in the case of the 1/2 times derivative of f(x)= x 2-2 x + 2, the following calculation is
right in many case.

 2  2
1 1 3
8
x -2 x +  = (x -)  (x -)
2 2 2 2
=
3 
If this is calculated termwise as follows, the result is different from the former.

- 17 -
  2
1
 2  2  2
1 1 1

x -2 x +  - 2x 1 +  2x 0

2
2
= x 2

 
3 1 1
1 8 2 2 -
= x - 4 x 2 +  x 2
 3
Needless to say, this cause is the difference between
1 1 1

       dx
x x - x x - x x -
 x -2 x + dx and
2 2 2
 x -2 xdx 2
2
+ 2 2
  0 0  
That is, it is because the latter regarded it as 0 and  although the former regarded the zero of the super
derivative as pi.
The latter is right when there is a special reason why the zero of the super derivative should be 0. However,
such a case is rare, and in almost all cases the former is right.

- 18 -
12.4 Super Derivative of Exponential Function
Analytically continuing the index of the differentiation operator in Formula 9.2.2 ( 9.2 ) to [0 , p ] from [1,n ]
we obtain the following formula. In addition, Rieamnn-Liouville diffeintegrals are also expressed together

Formula 12.4.1
When n = p is ceiling function, the following expressions hold

(1) Basic form

 (x -t)
 x (p) 1 dn x
e  = (1)-pe  x = n-p-1
e dt
t
(n -p) dx n 
(2) Linear form

 (x -t)
-p
dn
 
1 1 x
ax+b (p)
e 
ax+b n-p-1 at+b
= e = e dt
a (n -p) dx n 
a >0 : - , a <0 : +
(3) General form

 (x -t)
-p
dn
 
1 1 x
ax+b (p)
  =  ax+b = n-p-1
 a t+bdt
a log (n -p) dx n 
a >0 : - , a <0 : +

Proof of the general form

Let c =a log , d = b log , then

= e xa log+b log = e axloge blog = e log e log =  ax b = 

cx+d ax b ax+b
e
Applying this to (2) Linear form, we obtain (3) General form immediately.

Example

 2
1 1
- 1 -x
e  -x
= (-1) 2 -x
e = e = -ie -x
i
- 2
3x-4  2
 
1
e  = e 3x-4 = 4.728804 e 3x-4
3
-i

 
x (i) 1
2  = 2x = (0.933582 - 0.358362 i)  2x
log 2
1
 
1 -
2

 
2 1
(-3)x = (-3)x
log(-3)
= (1.487742 + 1.05582 i)  (-3)x
 2
1
-1/2

 
1
3 
x
= 3x = 1.048147  3x
log 3

x (1) (1/2)
When 3  , 3x , 3x are drawn on a figure side by side, it is as follows.

- 19 -
y
10

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
x
3^x*ln(3)
3^x/(1/ln(3))^(1/2)
3^x

- 20 -
12.5 Super Derivative of Logarithmic Function
Reversing the sign of the index of the integration operator in Formula 7.5.1 ( 7.5 ), we obtain the following
formula. In addition, Rieamnn-Liouville differintegrals are also expressed together

Formula 12.5.1
When (z), (z), n = p denote zeta function, psi function, ceiling function respectively, the following
expressions hold.

(1) Basic form

log x - (1-p)-  -p
 (x -t)
(p) 1 dn x
n-p-1
(log x) = x = log t dt
(1-p) (n -p) dx n 0
(2) Linear form
log(ax +b)- (1-p) -  -p

 
(p) b
log(ax +b ) = x+
(1-p) a


1 dn x
= (x -t)n-p-1log(at +b)dt
(n -p) dx n b
-
a
 (1- p)
Where, = (-1)p(p -1)! for p =1, 2, 3, 
(1- p)

Proof
Formula 7.5.1 ( 7.5 ) was as follows.
- (1+p)-  p
 
x x log x
 p
log xdx = x
0 0 (1+ p)
log(ax +b)- (1+ p) - 
 
x p

 
x b
p
log(ax +b)dx = x+
b b (1+ p) a
- -
a a
Since differentiation is the reverse operation of integration, replacing the index p of the integration operator
with -p , we obtain the desired expressions. And Formula 1.3.1 ( 1.3 ) was
 (-n)
= (-1)n+1n! , n =0, 1, 2, 3, 
(-n)
Then, replacing n with p -1 , we obtain
 (1-p)
= (-1)p(p -1)! , p =1, 2, 3, 
(1-p)
Example
log x - 1-1-  -1
(log x)(1) = x = -(-1)10! x -1 = x -1
1-1

 
1 1
log(3x +4) - 1- - -
 2
1 2

 
2 4
log(3x +4) = x+
 
1 3
 1-
2
= 0.5641(log x + 1.3862) / x +1.3333

- 21 -
  
1
log x - 1- -
 =
1 1 1
10 10 -
10
-
10
(log x) x = 0.9357x (log x +0.1777)
 
1
 1-
10

 
9
log x - 1- -
 =
9 9 9
10 10 -
10
-
10
(log x) x = 0.1051 x (log x +9.8465)
 
9
 1-
10

(1/10) (9/10) (1)

When log x , (log x) , (log x) , (log x) are drawn on a figure side by side,
it is as follows.

y
2

0
1 2 3 4 5 6 7
x
-2

-4

-6

ln(x)
-1/x^(1/10)/gamma(9/10)*(EULER - ln(x) + psi(9/10)
-1/x^(9/10)/gamma(1/10)*(EULER - ln(x) + psi(1/10)
1/x

- 22 -
12.6 Super Derivative of Trigonometric Function

12.6.1 Super Derivatives of sin x , cos x

Analytically continuing the index of the differentiation operator in Formula 9.2.4 ( 9.2 ) to [0 , p ] from [1,n ]
fwe obtain the following formula.

Formula 12.6.1
(1) Basic form
p
(sin x) p  
()
= sin x +
2
p
(cosx) p  
()
= cos x +
2
(2) Linear form
-p
p
   
(p) 1
sin(ax +b ) = sin ax +b +
a 2
-p
p
   
(p) 1
cos(ax +b ) = cos ax +b +
a 2

Example

 
1
1  
(sin x ) 2 
= sin x +
2 2  = sin x + 4 
      
1 1 1
((sin x ) 2 ) 2 = (sin (x + )) 2
4
1
-
2
 1  
      = cos x
1
= sin x + + = sin x +
1 4 2 2 2
 (1) -1
 1
       = sin x
1
sin x - = sin x - +
2 1 2 2
 
1
1  
(cos x) 2 
= cos x +
2 2  = cos x +  4 
  2 
2
(cos x)  
= cos x +
 2  = cos(x +1)
When cos x , (cos x)(1/2) , -sin x are drawn on a figure side by side, it is as follows. Red shows
1/2 order super derivative. It is clear also in the figure that super derivative which is the easiest to understand
is super derivative of trigonometric functions.

- 23 -
12.6.2 Termwise Super Derivative of sin x , cos x
Reversing the sign of the index of the operator of the collateral super integrals of sin x , cos x in 7.6.2 ( 7.6 ) ,
we obtain the following termwise super derivatives. These are collateral super derivatives as understood from
the constant-of-differentiation function in the right side.

(-1)k p
  + C(p, x)
(p) 
2k+1- p
(sin x) =Σ x = sin x +
k =0 (2k +2- p) 2
p x -p-k k
C(p, x) =Σ sin
k =1 (1- p -k) 2
k
p
  - C(p, x)
 (-1)
(cosx) p =Σ p
()
x 2k- = cos x +
k =0 (2k +1- p) 2
p x -p-k k
C(p, x) =Σ cos
k =1 (1- p -k) 2

When the 1/2th order collateral super derivative of cos x is drawn as cos x and -sin x side by side, it is as
follows. Red shows the 1/2th order collateral super derivative.

- 24 -
 Compared with the upper figure, the collateral super derivative is curving unnaturally near the coordinate origin
in this figure. Since this is similar also in sin x, it is thought that the termwie super derivatives of sin x and cos x
are asymptotic expansions of the lineal super derivatives.

- 25 -
12.7 Super Derivative of Hyperbolic Function

12.7.1 Super Derivatives of sinh x , cosh x

Analytically continuing the index of the differentiation operator in Formula 9.2.5 ( 9.2 ) to [0 , p ] from [1,n]
we obtain the following formula.

Formula 12.7.1
(1) Basic form
p i e x - (-1)-pe -x
 
(p) -p
(sinh x) =i sinh x + =
2 2
p i e x + (-1)-pe -x
 
(p) -p
(cosh x) =i cosh x + =
2 2
(2) Linear form
-p
p i
   
(p) i
sinh(ax +b ) = sinh ax +b +
a 2
-n

 
1 1
e - (-1)-pe -(ax+b)
ax+b
=
2 a
-p
p i
   
(p) i
cosh(ax +b ) = cosh ax +b +
a 2
-n

 
1 1
e + (-1)-pe -(ax+b)
ax+b
=
2 a

Example 1

  1 i i
1 1 1

   
- -
(sinh x ) 2 =i 2
sinh x + =i 2
sinh x +
2 2 4
    i  2
1 1 1 1
-
((sinh x ) 2 ) 2 = (i 2 sinh (x + ))
4
i 1 i
1 1

 
- -
2 2
=i i sinh x + +
4 2 2
i
= i -1 sinh x +  2  = cosh x
i (1) -1
i 1 i
      
i
cosh x + = cosh x + +
2 1 2 2
= -i cosh(x + i) = i cosh x
 
i i
 
   
-
(cosh x) 10 =i 10
cosh x - = 1.170088  cosh x -
20 20
 
9i 9i
9 9
   
-
(cosh x) 10 =i 10
cosh x - = 4.111207  cosh x -
20 20

- 26 -
 Super derivative of hyperbolic function is the most incomprehensible in super derivatives. The reason is that
the super derivative turns into a complex function except the order p is an integer or a purely imaginary number.
(i/10) (9 i/10)
Then, when cosh x , (cosh x) , (cosh x) , sinh x which can be displayed on a real number
domain are drawn on a figure side by side, it is as follows.

400

300

200

100

-4 -3 -2 -1 1 2 3 4 5 6
x
cosh(x)
1/(I)^(1/10*I)*cosh(x - 1/20*PI)
1/(I)^(9/10*I)*cosh(x - 9/20*PI)
sinh(x)

(i/10)
All of four curves have overlapped in the positive area. It is natural that (cosh x) is near cosh x in
(9 i/10)
a negative area. But (cosh x) is far apart from sinh x in why.

Example 2
1
-
 
1 x
2 e - (-1) 2 -x
e e x + i e -x
(sinh x ) = =
2 2
 2
1

 2
1
 )  =
1 1

 
x x
e + i e -x e i
((sinh x ) 2 2
= +  e -x
2 2 2
ex i e x + e -x
= + -i e -x = = cosh x
2 2 2
i i

  e
i  
(1) -1

cosh x + 2 
1 1 x+
2 -1
- x+
2
= + (-1) e
2 1
i i
=
1
2

-
e 2 e x - e 2 e -x =
1
2
1
i e x - e -x
i
  
i
=  e x + e -x = i cosh x
2
i i
-
-
 -x i  -x
 
i
e x - (-1) e e x - e  e e x - e 1-x
(sinh x)  = = =
2 2 2

- 27 -
12.7.2 Termwise Super Derivative of sinh x , cosh x
Reversing the sign of the index of the operator of the collateral super integrals of sinh x , cosh x in 7.7.2
we obtain the following termwise super derivatives. These are collateral super derivatives as understood from
the constant-of-integration function in the right side.

p i
p
x 2k+1-
  - C(p, x)
(p) 
(sinh x) =Σ =i -p
sinh x +
k=0 (2k +2- p) 2
p x -p-k k i
C(p, x) Σ i -k sinh
k=1 (1- p -k) 2
p i
p
x 2k-
  - C(p, x)
(p) 
(cosh x) =Σ = i -p cohs x +
k=0 (2k +1- p) 2
p x -p-k k i
C(p, x) Σ i -k cosh
k=1 (1- p -k) 2

When the values of the lineal and the collatera l .7th order derivatives of cosh x on x=1, x=6 are calculated
respectively, it is as follows.

Though the difference of both is large where x is small, both are almost corresponding ing where x is large.
Since this is similar also in sinh x, it is thought that the termwise super derivatives of sinh x and cosh x are
asymptotic expansions of the lineal super derivatives.

- 28 -
12.8 Super Derivative of Inverse Trigonometric Function

-1 -1
12.8.1 Super Derivatives of tan x , cot x
Reversing the sign of the index of the integration operator in Formula 7.8.1 ( 7.8 ), we obtain the following
formula.

Formula 12.8.1
When (x),(x) denote gamma function and digamma function respectively, the following expressions
hold for x 1 .

tan -1x 
 
(p) -p
tan x Σ (-1)k
-1
= x -p-2k
(1-p) k =0 -p -2k
log1+ x 2 
 
-p
+
2(1-p) k=1 Σ (-1)k
-p +1-2k
x -p+1-2k

 
1  -p
Σ (-1)r  (1- p)- (2r)x
-p+1-2r
-
(1-p) r=1 -p +1-2r
x -p tan -1x 
 
(p) -p
cot x Σ (-1)k
-1 -1
= cot x - x -p-2k
(1- p) (1-p) k =1 -p -2k
log1+ x 2 
 
-p
-
2(1-p) k=1 Σ (-1)k
-p +1-2k
x -p+1-2k

 
1  -p
 (1- p)- (2r)x
(1-p) Σ
r -p+1-2r
+ (-1)
r=1 - p +1-2 r
-1
Example: 1/2th order derivative of cot x

Rie mann-Liouville diffe rinte gral

 g := x-> 1/gamma(1-p)*int((x-t)^(1-p-1)*arccot(t), t=0..x)

- 29 -
 x
1  1 p1
x   (x  t)  arccot(t) d t
(1  p)
0

2010.07.07
K. Kono
Alien's Mathematics

- 30 -
13 Termwise Super Derivative
In this chapter, for the function whose super derivatives are difficult to be expressed with easy formulas, we
differentiate the series expansion of these functions non integer times termwise. Therefore, e x, log x , sin x ,
cos x , sinh x , cosh x mentioned in " 12 Super Derivative " are not treated here.

13.1 Termwise Super Derivative of Trigonometric Functions & Hyperbolic Functions

Formula 13.1.1
Let (x) be gamma function, ,  are ceilling function and floor function and let Bernoulli number B2n
and Euler number E2k are as follows.

B0=1, B2=1/6, B4=-1/30, B6=1/42, B8=-1/30, 

E0=1, E2=-1, E4=5, E6=-61, E8=1385, 
Then the following expressions hold for p  0 and 0 < x <  /2 .

(p)  22k 22k -1B2k 2k- p-1

(tan x) = Σ
p +1 2k (2k - p)
x
k= 
2

(p)  22k 22k -1B2k p-1

(tanh x) = Σ
p +1 2k (2k - p)
x 2k-
k= 
2

 E2k
(sec x) p
()
Σ
2k-p
= x : collateral
p +1 (2k -p +1)
k= 
2

 E2k
(sech x) p =
( )
Σ p +1 (2k p
- +1)
x 2k-p : collateral
k= 
2

Proof
There were the following formulas in " 10 Termwise Higher Derivative "

(n)  22k 22k -1B2k 2k- n-1

Formula 10.1.1 (tan x) = Σ
n +1
x
2k (2k - n -1)!
k= 
2

(n) 22k 22k -1B2k 2k- n-1

Formula 10.2.1 (tanh x) = Σ x
n +1 2k (2k - n -1)!
k= 
2

 E2k
(sec x) n
()
Formula 10.7.1 = Σ
n +1 (2k -n)!
x 2k-n
k= 
2

 E2k
(sech x) n =
()
Formula 10.8.1 Σ n +1 (2k n
- )!
x 2k-n
k= 
2

In these formulas, replacing m! with gamma function (1+ m) and analytically continuing the index of
the differentiation operator to [0 , p ] from [1 , n ], we obtain the desired expressions.
Since the termwise super integrals of arcsin x and arccos x were collateral, the super derivatives have to be
collateral too. (This is the same in the following chapters.)

-1-
Example 1: 3/4 th order derivative of tan x
We calculated the the super differential coefficients on arbitrary one point x =0.4 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding. Moreover, in the figure,
blue shows tan x, red shows the 3/4 th order derivative and green shows the 1 st order derivative.
Te rmwise supe r de rivative of tan x
 tanp := (p,x)-> sum(2^(2*k)*(2^(2*k)-1)*abs(bernoulli(2*k))/((2*k
*gamma(2*k-p))*x^(2*k-1-p),k=ceil((p+1)/2)..200
   
2k 2k

200
2  2  1  bernoulli(2  k) 2  k  1  p

( p, x)  x

p1
 (2  k)  (2  k  p)
k  
2

Rie mann-Liouville diffe rinte gral

 p:=3/4: h:=10^-10:
 f := x-> 1/gamma(1-p)*int((x-t)^(1-p-1)*tan(t), t=0..x)
x
1  1 p1
x   (x  t)  tan(t) d t
(1  p)
0

Example 2: 9/10 th order derivative of tanh x

Only the figure is shown. Blue shows tanh x, red shows the 9/10 th order derivative and green shows the
1 st order derivative.

-2-
Example 3: 1/2 th order derivative of sec x
We calculated the the super differential coefficients on arbitrary one point x =0.3 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding. Moreover, in the figure,
blue shows sec x, red shows the 1/2 th order derivative and green shows the 1 st order derivative.

Formula 13.1.2
Let (x) be gamma function, ,  are ceilling function and floor function and let Bernoulli number B2n
and Euler number E2k are as follows.

B0=1, B2=1/6, B4=-1/30, B6=1/42, B8=-1/30, 

E0=1, E2=-1, E4=5, E6=-61, E8=1385, 
Then the following expressions hold for p  0 and  /2 < x < .
22k 22k -1B2k  2k- p-1

x - 2 
(p) 
(cot x) =- Σ
p +1 2k (2k - p)
k= 
2

E2k  2k-p

p +1 (2k -p +1) 
x- 
(p) 
(csc x) = Σ 2
: collateral
k= 
2

-3-
Proof
There were the following formulas in " 10 Termwise Higher Derivative "

22k 22k -1B2k  2k- n-1

x - 2 
(n) 
Formula 10.3.1' (cot x) = - Σ
n +1 2k(2k - n -1)!
k= 
2

E2k  2k-n

x - 2 
(n) 
Formula 10.5.1' (csc x) = Σ
n +1 (2k -n)!
k= 
2

In these formulas, replacing m! with gamma function (1+ m) and analytically continuing the index o
the differentiation operator to [0 , p ] from [1 , n ], we obtain the desired expressions.

Example 1: 3/4 th order derivative of cot x

We calculated the the super differential coefficients on arbitrary one point x =1.7 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding. Moreover, in the figure,
blue shows cot x, red shows the 3/4 th order derivative and green shows the 1 st order derivative.
Te rmwise supe r de rivative of cot x
 cotp := (p,x)-> -sum(2^(2*k)*(2^(2*k)-1)*abs(bernoulli(2*k))/((2*
*gamma(2*k-p))*(x-PI/2)^(2*k-1-
p),k=ceil((p+1)/2)..200)
⎛    ⎞
 2k 2k   
2  2  1  bernoulli(2  k)
200
 x  
 2k 1 p
( p, x)   ⎜
⎝  
⎟
⎠
p1
(2  k)  (2  k  p) 2
k  
2

Rie mann-Liouville diffe rinte gral

 p:=3/4: h := 10^-11:
 f := x-> 1/gamma(1-p)*int((x-t)^(1-p-1)*cot(t), t=PI/2..x)
x
1  1 p1
x   (x  t)  cot(t) d t
(1  p)

2

-4-
Example 2: 14/15 th order derivative of csc x
Only the figure is shown. Blue shows csc x, red shows the 14/15 th order derivative and green shows the
1 st order derivative.

The following lineal termwise super derivatives exist for csch x and sech x.

Formula 13.1.3
The following expressions hold for p 0 , x >0 .
 (2k +1)p
(csch x) (p)
= (-1) 2Σ
-p
x
e(2k +1) k =0


p
k (2k +1)
(sech x) = (-1) 2Σ(-1) (2k +1)x
(p) -p
k =0 e

Proof
Formula 8.1.3 ( 8.1 ) was as follows.

 csch x dx
x
x x  e -(2k +1)
p
= (-1) 2Σ p
  k =0 (2k +1)p


x
x x e -(2k +1)

sech x dx = (-1) 2Σ(-1)
p p k
  k =0 (2k +1)p
Since differentiation is the reverse operation of integration, replacing the index p of the integration operator
with -p , we obtain the desired expressions.

Example: 7/9 th order derivative of csch x .

We calculated the the super differential coefficients on arbitrary one point x =3.8 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding.

-5-
-6-
13.2 Termwise Super Derivative of Inverse Trigonometric Functions

Formula 13.2.1
When (x) is gamma function and  is ceilling function, the following expressons hold for p0 and
0 < x <1 .
(p)  (2k)! 2k+1- p
tan x Σ (-1)k
-1
= x
k=
p -1

(2k +2- p )
2

(p) x
-p  (2k)! 2k+1- p
cot x Σ (-1)k
-1
= - x
2 (1-p)
k=
p -1

(2k +2- p)
2
2
(p)  (2k -1)!!
sin x Σ
-1
= x 2k+1-p : collateral
p -1 (2k +2- p )
k= 
2

(p) x
-p  (2k -1)!!
2
cos x Σ
-1
= - x 2k+1-p : collateral
2 (1-p) p -1 (2k +2- p )
k= 
2

Proof
There were the following formulas in " 11 Termwise Higher Derivative " .
(n)  (2k)! 2k+1- n
tan x Σ (-1)k
-1
Formula 11.1.1 = x
n -1
k=
(2k +1- n)!

2
2
(n)  (2k -1)!!
sin x Σ
-1
Formula 11.1.2 = x 2k+1-n
n -1
k= 
(2k +1- n)!
2
In these formulas, replacing m! with gamma function (1+ m) and analytically continuing the index of
-1 (p ) (p )
the differentiation operator to [0 , p ] from [1 , n ], we obtain tan x , sin -1x  . Next,

cot -1x = x 0 - tan -1x
2
From this,
(p)  0 (p) (p)
cot x x  - tan -1x
-1
=
2
Substituting
-p
x  (2k)!
0 (p) (p) 2k+1- p
x  tan x Σ (-1)k
-1
= , = x
(1-p)
k=
p -1

(2k +2- p)
2
-1 (p ) -1 (p )
for this, we obtain cot x . cos x is also obtained in a similar way.

Note
When p =1, 2, 3, , (1-p) =(0),(-1),(-2),  i.e. (1-p) =  . Then,
-p
x
= 0 for p =1, 2, 3, 
(1-p)
(p ) (p )
Therefore, If we replace p with n , cot -1x  , cos -1x  results in the following formulas in 11.1 .

-7-
 (n)  (2k)! 2k+1- n
cot x Σ (-1)k
-1
Formula 11.1.1 =- x
k=
n -1

(2k +1- n)!
2
2
(n)  (2k -1)!!
cos x Σ
-1
Formula 11.1.4 = - x 2k+1-n
k=
n -1

(2k +1- n)!
2

Example 1: 9/10 th order derivative of arctan x

We calculated the the super differential coefficients on arbitrary one point x =0.1 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding. Moreover, in the figure,
blue shows arctan x, red shows the 9/10 th order derivative and green shows the 1 st order derivative.

Example 2: 1/2 th order derivative of arccot x

We calculated the the super differential coefficients on arbitrary one point x =0.05 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding.

-8-
Example 3: 4/5 th order derivative of arcsin x
Only the figure is shown. Blue shows arcsin x, red shows the 4/5 th order derivative and green shows the 1 st
order derivative.

Example 4: 1st order derivative of arccos x

-p
x
When p =1 , (1-p) =(0)= . Then = 0. Therefore, from the formula,
(1-p)
2
(1)  (2k -1)!! 1
cos x = -Σ
-1
x 2k = -
k =0 (2k +1) 1-x 2
In fact, this series converges to the right-hand side on |x|< 1 .

-9-
13.3 Termwise Super Derivative of Inverse Hyperbolic Functions

Formula 13.3.1
When (x), (x),  ,  are gamma function, digamma function, ceilling function and Euler-Mascheroni
constant (= 0.57721566...) , the following expressons hold for p0 and 0 < x <1 .
(p)  (2k)! 2k+1- p
tanh x Σ
-1
= x
p -1 (2k +2- p )
k= 
2
2

(p) k (2k -1)!! p
sinh x Σ
-1
= (-1) x 2k+1- : collateral
k =
p -1

(2k +2- p)
2

x -p 2

  - Σ 2k (2k -p+1) x
(p) 2  (2k -1)!!
sech
-1
x = log +  (1- p)+  2k-p
(1- p) x p
k= 
2
: collateral
-p 2
x  (2k -1) !!
 
(p) 2
csch x log + (1- p) + 
-1
= - Σ (-1)k x 2k-p
(1- p) x p 2k (2k -p+1)
k= 
2
: collateral

Proof
There were the following formulas in " 11 Termwise Higher Derivative "
(n)  (2k)! n
tanh x Σ
-1
Formula 11.2.1t = x 2k+1-
n -1
k= 
(2k +1- n )!
2
2

(n) k (2k -1)!!
sinh x Σ
-1
Formula 11.2.2s = (-1) x 2k+1-n
n -1
k= 
(2k +1- n)!
2
In these formulas, replacing m! with gamma function (1+ m) and analytically continuing the index of
-1 (p ) (p )
the differentiation operator to [0 , p ] from [1 , n ], we obtain tanh x , sinh -1x  .
Next, arcsech x is expanded to Tylor series for 0< x <1 as follows.
2
 (2k -1)!!
sech x = log 2x - log x -Σ
-1 0
x 2k
k =1 2k (2k)!
Differentiating both sides of this with respect to x n times,
2
(n) (n)  (2k -1)!!
sech
-1
x = log 2x 0 - (log x)(n) - Σ x 2k-n
n
k= 
2k (2k - n)!
2
Substituting

0 (n) x -n log x - (1-n)-  -n

(log x) n =
( )
x  = , x
(1-n) (1-n)
for this,

x -n 2

 -Σ
(n) 2  (2k -1)!!
sech
-1
x = log +  (1- n)+  x 2k-n
(1- n) x n
k= 
2k (2k - n)!
2
-1 (p ) -1 (p )
Replacing n with p , we obtain sech x . csch x is also obtained in a similar way.

- 10 -
Note
According to Formula 1.3.1 ( 1.3 ) ,
 (1-n)
= (-1)n(n -1)! n =1, 2, 3, 
(1-n)
-1 (n )
If we substitute this for sech x in the proof, it results in Formula 11.2.3 ( 11.2 ) as follows.
2
(n) (n -1)!  (2k -1)!!
sech x n
-Σ
-1
= (-1) n x 2k-n
x n 2k (2k - n )!
k= 
2

Example 1: 3/4 th order derivative of arctanh x

We calculated the the super differential coefficients on arbitrary one point x =0.2 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding. Moreover, in the figure,
blue shows arctanh x, red shows the 3/4 th order derivative and green shows the 1 st order derivative.

Example 2: 3/2 th order derivative of arcsinh x

Replacing the calculation order of differentiation and integration in Riemann-Liouville differintegral, we obtain

  
(p) 1 x
n-p-1 dn
f (x) = (x -t) f(t) dt n = p
(n -p) a dt n
We calculated the the super differential coefficients on arbitrary one point x =0.3 according to the formula
and this expression. As the result, two values were corresponding.

- 11 -
Example 3: 6/7 th order derivative of arcsech x
Only the figure is shown. Blue shows arcsech x, red shows the 6/7 th order derivative and green shows the
1 st order derivative.

0.2 0.4 0.6 0.8 1.0

-5

2006.11.22
2012.07.22 Renewal
K. Kono
Alien's Mathematics

- 12 -
14 Higher and Super Calculus of Logarithmic Integral etc.

14.1 Higher Integral of Exponential Integral

Exponential Integral is defined as follows.


x t e
Ei(x) = dt (1.0)
- t

Integrating both sides of (1.0) with respect to x repeatedly by ONLINE INTEGRATOR (Wolfram Mathematica)
and arranging the results, we obtain the following higher indefinite integrals.

Ei(x)dx = 1! xEi(x)- e 0!

1 x

Ei(x)dx = 2! x Ei(x)-e (0!x +1!)

1 2 2 x

Ei(x)dx = 3! x Ei(x)-e 0!x +1!x +2!

1 3 3 x 2


1 n -1
 Ei(x)dx n =
 
n-1- r
x nEi(x)-e xΣr! x
n! r=0

Although these right sides are the lineal primitive functions of Ei(x), since both zeros of Ei(x) and ex are - ,
zeros of the right sides are all - . Therfore, the lineal higher primitive function of Ei(x) can be expressed by
the higher integral with a fixed lower limit - .

Formula 14.1.1


x et
When Exponential Integral is Ei(x)= dt , the following expressions hold.
- t

 
x x 1 n -1

 
n-1- r
Ei(x)dx n = x nEi(x)-e xΣr! x (1.n)
- - n! r=0

Example : 3rd order integral of Ei (x)

-1-
-2-
14.2 Higher Integral of Cosine Integral
Cosine Integral is defined as follows.


x cos t
Ci(x)= dt
 t

Integrating both sides of this with respect to x repeatedly by ONLINE INTEGRATOR and arranging
the results, we obtain the following higher indefinite integrals.

Ci(x)dx
1
x Ci(x) - 0!sin x
1
=
1!

Ci(x)dx
1
x Ci(x) - 0!xsin x + 1!cos x
2 2
=
2!

Ci(x)dx
1
x Ci(x) - 0!x -2!sin x + 1!x cos x
3 3 2
=
3!

Ci(x)dx
1
x Ci(x) - 0!x -2!xsin x + 1!x -3!cos x 
4 4 3 2
=
4!


 
1 (n -1)/2
 Ci(x)dx n = Ci(x)x n - sin x Σ (-1)r(2r)! x n-1-2r
n! r=0


(n -2)/2
+ cos x Σ (-1)r(2r +1)! x n-2-2r
r=0

Although these right sides are the lineal primitive functions of Ci(x), these zeros are all  . (See the above
figure.) Therfore, the lineal higher primitive function of Ci(x) can be expressed by the higher integral with a
fixed lower limit .

Formula 14.2.1


x cost
When  is floor function and Ci(x)= dt ,is Cosine Integral, the following expressions hold.
 t

 Ci(x)dx
x 1
x Ci(x) - 0!sin x
1
=
 1!

 Ci(x)dx
x x 1

(n-1)/2
n-1-2r
n
= Ci(x x
) n
- sin x Σ (-1)r(2r)! x
  n ! r=0

-3-
 
(n-2)/2
+ cosx Σ (-1)r(2r +1)! x n-2-2r n 2
r=0

Example : 4th order integral of Ci (x)

If both sides are illustrated, it is as follows. Since both sides overlap exactly, the left side (blue) is not visible.

-4-
14.3 Collateral Higher Integral of Sine Integral
Sine Integral is defined as follows.


x sin t
Si(x)= dt
0 t
Integrating both sides of this with respect to x repeatedly by ONLINE INTEGRATOR and arranging
the results, we obtain the following higher indefinite integrals.

Si(x)dx
1
x Si(x) + 0!cos x
1
=
1!

Si(x)dx
1
x Si(x) + 0!x cos x + 1!sin x
2 2
=
2!

Si(x)dx
1
x Si(x) + 0!x -2!cos x +1! x sin x
3 3 2
=
3!

Si(x)dx
1
x Si(x) + 0!x -2!xcos x + 1!x -3!sin x
4 4 3 2
=
4!


 
1 (n -1)/2
 Si(x)dx n = Si(x)x n + cos x Σ (-1)r(2r)! x n-1-2r
n! r=0


(n -2)/2
n-2-2r
+ sin x Σ
r=0
(-1)r(2r +1)! x

Although these right sides are the lineal primitive functions of Si(x), those zeros are all 0 at the time of even
order and are not 0 at the time of odd order. That is, the lineal higher primitive function of Si(x) can not be
expressed by the higher integral with a fixed lower limit.(See the above figure.) Therfore, the higher integra
of Si(x) with a fixed lower limit 0 is not lineal but collateral. However, the idea which makes 0 a common lower
limit is natural. It is because the Si(x) itself is defined by the integral with a lower limit 0.

Collateral Higher Integral of Sine Integral

Collateral Higher Integrals of Si(x) are obtained by compensating the above lineal higher primitive functions
with Constant-of-integration Polynomials .

Formula 14.3.1


x sin t
When  is floor function and Si(x)= dt ,is Sine Integral, the following expressions hold.
0 t

-5-
 
x 0
1 x
x Si(x) + 0!cos x -
1
Si(x)dx =
0 1! 10!

Si(x)dx = 2! x Si(x) + 0!x cos x + 1!sin x - 11!

x x 1
1 2 2 x
0 0

 Si(x)dx = 3! x Si(x) + 0!x -2!cos x +1! x sin x - 12!

x x x 2 0
1 3 3 2 x x
+
0 0 0 30!


x x 1
 Si(x )dx = x Si(x) + 0!x -2!xcos x + 1!x -3!sin x
4 4 3 2
0 0 4! 3 1
x x
- +
13! 31!



x x 1

(n-1)/2
 Si(x)dx n = Si(x)x n + cosx Σ (-1)r(2r)! x
n-1-2r
0 0 n ! r=0


(n-2)/2
+ sin x Σ (-1)r(2r +1)! x n-2-2r
r=0
n-1-2r
(n-1)/2 x
- Σ
r=0
(-1) r
(2r +1)(n -1-2r)!

Example : Collateral the 4th order integral of Si (x)

If both sides are illustrated, it is as follows. Since both sides overlap exactly, the left side (blue) is not visible.

-6-
14.4 Higher Integral of Logarithmic Integral
Logarithmic Integral is defined as follows.


x 1
li(x)= dt (1.0)
0 log t
6

2 4 6 8 10

-2

-4

-6

First, we prepare two Lemmas.

Lemma 14.4.1


x et
When Exponential Integral is Ei(x)= dt , the following expressions hold.
- t

Ei(2log x)dx = xEi(2log x) - Ei(3log x)

Ei(3log x)dx = xEi(3log x) - Ei(4log x)


Ei(n log x)dx = xEi(n log x) - Ei (n +1)log x   (1.n)

Proof
t t
2 x 1
Let 2log x = t . Then x = e , dx = dt = e 2 dt . Hence
2 2
t

 Ei(t)e
1 2
Ei(2log x)dx = dt
2
Calculating the integral of the right side by ONLINE INTEGRATOR, we obtain
t t

Ei(t)e  2 t
2 2 3
dt = 2Ei(t)e - 2Ei
Using this,
t

   = xEi(2log x) - Ei(3log x)
2 3
Ei(2log x)dx = Ei(t)e - Ei t
2
Next, let 3log x = t . Then we obtain the following expresssion by the same calculation.
t t t

Ei(3log x)dx =  Ei(t)e  

1 1 4
Ei(t) e 3 dt = 3 3
dt = Ei(t)e - Ei t
3 3 3

-7-
 = xEi(3log x) - Ei(4log x)
Hereafter, by induction, we obtain the desired expresson.

Note
Since log x - at the time x +0 , x =0 is clearly a zero of these functions. Then, (1.n) can be
written as follows.

 Ei(n log x)dx = xEi(n log x) - Ei (n +1)log x

x
  (1.n')
0

Lemma 14.4.2


et x
When Exponential Integral is Ei(x)= dt , the following expressions hold.
- t


1 n+1 1
x nEi(log x)dx = x Ei(log x)- Ei(n +2)log x (2.n)
n +1 n +1

Calculation
Calculating by ONLINE INTEGRATOR , we obtain (2.n) immediatly.

Note
Since log x - at the time x +0 , x =0 is crealy a zero of these functions. Then, (2.n) can be
written as follows.


x 1 n+1 1
n
x Ei(log x)dx = x Ei(log x)- Ei(n +2)log x (2.n')
0 n +1 n +1

Formula 14.4.3
When

 
x 1 x et
li(x)= dt , Ei(x)= dt
0 log t - t
the following expression holds for x 0 .


x x 1 n
 li(x)dx n = Σ (-1)r nＣ r x n- r Ei(r +1)log x (3.n)
0 0 n ! r=0

Proof
Let t = log x . Then [0 , x][- , t] , dx = e tdt . Hence

 
x 1 t et
dx = dt = [Ei (t )]t- = Ei (log x ) = li(x)
0 log x - t
Next, let

  li(x)dx
x x 1 x
dx 2 =
0 0 log x 0
Calculating the integral of the right side by ONLINE INTEGRATOR, we obtain

li(x)dx = xli (x )- Ei (2log x )

-8-
Since the zero of this right side is x=0 obviously,

 li(x)dx = xEi(log x) - Ei(2log x)

x

0
Next, integrating both sides of this with respect to x and applying Lemma 14.4.1 , 14.4.2
to the result, we obtain the following

 li(x)dx =  xEi(log x)dx -  Ei(2log x)dx

x x x x
2
0 0 0 0
1 2 1
= x Ei(log x)- Ei(3log x)- xEi(2log x)- Ei(3log x)
2 2
1
= x 2Ei(log x)-2xEi(2log x)+Ei(3log x)
2
Next, integrating both sides of this with respect to x and applying Lemma 14.4.1 , 14.4.2 to the result,
we obtain the following

    Ei(3log x)dx

x x x 1 x x 1 x
3 2
li(x)dx = x Ei(log x) dx - x Ei(2log x) dx +
0 0 0 2 0 0 2 0

1 3 1 1 1
= x Ei(log x)- Ei(4log x)- x 2Ei(2log x)+ Ei(4log x)
3! 3! 2 2
1 1
+ xEi(3log x) - Ei(4log x)
2 2
1
x Ei(log x)-3x Ei(2log x)+3xEi(3log x)-Ei(4log x)
3 2
=
3!
Hereafter, by induction, we obtain the desired expresson .

Example : 2nd order integral of li(x)

-9-
14.5 Higher Integral of Double Logarithmic Function
Double Logarithmic Function is defined as follows.
f(x) = log|log x| (1.0)

Integrating both sides of this with respect to x repeatedly and arranging the results, we obtain the following
higher indefinite integrals. Where, li(x)=Ei(log x) is Logarithmic Integral mentioned in the previous.

log|log x|dx = 1! xlog|logx| - li(x)

1
 

log|log x|dx = 2! x log|logx| -2xli(x)+ Ei(2log x)

1 2 2


1
x log|logx| -3x li(x)+3xEi(2log x)- Ei(3log x)
3 3 2
log |log x|dx =
3!



1 n
 log|log x|dx n =
 
n-r
x nlog|logx| + Σ(-1)rnＣ r x Ei(rlog x)
n! r=1

Although these right sides are the lineal primitive functions of log|logx | , since both zeros of x nlog|logx |
and Ei(nlog x ) are 0 , zeros of the right sides are all 0 . Therfore, the lineal higher primitive function of
log|logx | can be expressed by the higher integral with a fixed lower limit 0 .

Formula 14.5.1


x et
When Ei(x)= dt , the following expressions hold for x 0 .
- t


x x 1 n

 
n-r
 log|log x|dx n = x n log|logx| + Σ(-1)rnＣ r x Ei(rlog x) (1.n)
0 0 n! r=1

Example : 2nd order integral of log|logx|

When the one arbitrary point x =1.6 is given) , the values of the both sides are as follows.

- 10 -
14.6 Super Calculus of Logarithmic Integral
Among the higher integrals mentioned in previous sections, Higher Integral of Logarithmic Integral is
extensible even to Super Calculus. It is because this higher integral is expressed with binomial coefficients.

14.6.1 Super Integral of Logarithmic Integral

Formula 14.6.1
When

 
x 1 x et
li(x)= dt , Ei(x)= dt
0 log t - t
p 0 and x 0 .
the following expression holds for

 
x x 1  p
 li(x)dx p = Σ (-1)r p- r
x Ei(r +1)log x
0 0 (1+ p) r=0 r

Proof

r 
n
First, replace n!, nＣ r with (1+n) , respectively in Formula 14.4.3 . Next, analytically

continuing the index of the integration operator to [0 , p ] from[1 , n ], we obtain the desired formula.

Example : 3/2th order integral of li(x)

We calculated the function values on arbitrary one point x =4 according to the formula and Riemann-
Liouville integral. As the result, two values were almost corresponding.
Compared with the figure of the 2nd order integral in 14.4 , we can find that this curvature is loose.

14.6.2 Super Derivative of Logarithmic Integral

Formula 14.6.2
When

- 11 -
  
x 1 x et
li(x)= dt , Ei(x)= dt
0 log t - t
p >0 , p  1, 2, 3,  and x 0 .
the following expression holds for

 
1  -p r
Σ
(p) r
li(x) = (-1) x -p- Ei(r +1)log x
(1- p) r=0 r

Proof
In fact, Formula 14.6.1 holds for p  -1, -2, -3,  , x 0 . Therefore, in Formula 14.6.1 , replacing the
integration operator <p > with the differentiation operator (p ) = <-p > , we obtain the desired expression.

Example : 1/2th order derivative of li(x)

We calculated the the super differential coefficients on arbitrary one point x =2 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding.

- 12 -
14.7 Super Calculus of Double Logarithmic Function
Among the higher integrals mentioned in previous sections, Higher Integral of Double Logarithmic Function is
extensible even to Super Calculus. It is because this higher integral is expressed with binomial coefficients.

14.7.1 Super Integral of Double Logarithmic Function

Formula 14.7.1


x et
When Ei(x)= dt , the following expressions hold for p >0 , x 0.
- t

   
x x 1  p

p-r
log|log x|dx p = x plog|logx| + Σ(-1)r x Ei(rlog x)
0 0 (1+ p) r=1 r

Proof

r 
n
First, replace n!, nＣ r with (1+n) , respectively in Formula 14.5.1 . Next, analytically

continuing the index of the integration operator to [0 , p ] from[1 , n ], we obtain the desired formula.

Example : 5/3th order integral of log|logx|

We calculated the function values on arbitrary one point x =1.5 according to the formula and Riemann-
Liouville integral. As the result, two values were almost corresponding.

14.7.2 Super Derivative of Double Logarithmic Function

Formula 14.7.2


x et
When Ei(x)= dt , the following expression holds for p >0 , p  1, 2, 3,  , x 0 ,
- t

- 13 -
the following expression holds.

   
1  -p Ei(rlog x)
log|log x|
(p)
= log|logx| + Σ (-1)r
x p(1-p) r=1 r xr

Proof
In fact, Formula 14.7.1 holds for p  -1, -2, -3,  , x 0 . Therefore, in Formula 14.7.1 , replacing the
integration operator <p> with the differentiation operator (p) = <-p > , we obtain the desired expressions.

Example : 0.3th order derivative of log|logx|

We calculated the the super differential coefficients on arbitrary one point x =0.5 according to the formula
and Riemann-Liouville differintegral. As the result, two values were almost corresponding.
Since the number of the order of the differentiation is small, it resembles the figure of log|logx | well.

2007.10.05
K. Kono
Alien's Mathematics

- 14 -
15 Higher and Super Calculus of Elliptic Integral

15.1 Double series expansion of Elliptic Integral

15.1.1 Double series expansion of Elliptic Integral of the 1st kind

Formula 15.1.1
The following expressions hold for |k|  1 , |x|  1 .


x dx
F(x, k) =
1- x 1- k x 
0 2 2 2

(-1)r -1/2
  
 r -1/2 2s 2r+1
= ΣΣ k x (1.1)
r=0 s=0 2r +1 r-s s
 r 1 (2r -2s -1)!! (2s -1)!! 2s 2r+1
= ΣΣ k x (1.1')
r=0 s=0 2r +1 (2r -2s)!! (2s)!!

Proof
Since |k|  1 , |x|  1 , from Generalized Binomial Theorem (See 3.2 ) .
1

 x
-
2  -1/2
1- x  = Σ(-1)r
2 2r
r=0 r
1

 
-
2  -1/2
1- k x  = Σ(-1)r
2 2
k 2rx 2r
r=0 r
Multiplying each other, it is as follows.
1

       x
- -1/2 -1/2 -1/2 -1/2
2 2
1- x  = x0 - x2 + x4 - 6
+-
0 1 2 3
1

        k x +-
- -1/2 -1/2 -1/2 -1/2
2
 1- k x 
2 2
= k 0x 0- k 2x 2+ k 4x 4- 6 6
0 1 2 3

 0   0 k x
-1/2 -1/2 0 0
=

 0   1   1   0  k x
-1/2 -1/2 -1/2 -1/2 0 2 2
- k +

 0   2   1   1   2   0  k x
-1/2 -1/2 -1/2 -1/2 -1/2 -1/2 0 2 4 4
+ k + k +

 r-s   
 r -1/2 -1/2
= ΣΣ(-1)r k 2sx 2r
r=0 s=0 s
That is

 r-s   
1  r -1/2 -1/2
= ΣΣ(-1)r k 2sx 2r
1- x 1- k x 
2 2 2 r=0 s=0 s
Then, integrating both sides of this with respect to x from 0 to x, we obtain (1.1).
Next, from the definition of General Binomial Coefficient (See 3.2 ) ,

-1-
 (-1/2+1) (1/2)
 
-1/2
= =
s (-1/2-s +1)(s +1) (1/2-s)(s +1)
On the other hand, from the properties of the Gamma Function (See 1.1.6 ) ,

 2n 
 =  
1 1
 n+ (2n -1)!! ,  -n = (-1) n
2 2n 2 (2n -1)!!
Then, for non-negative integer s,

2s 
 =  
1 1
  ,  -s = (-1) s
2 2 (2s -1)!!
Using these,

 (2s -1)!!
 
-1/2 s (2s -1)!!
= (-1)s = (-1)
s 2ss!  2ss!
i.e.

   r -s 
-1/2 (2s -1)!! -1/2 (2r -2s -1)!!
= (-1)s , = (-1)r-s
s (2s)!! (2r -2s)!!
Hence, substituting these for (1.1) , we obtain (1.1') .

 
1
Example 1 : Double series expansion of F x ,
2
When arbitrary point x= 0.7 is given to this elliptic integral and its double series, both are compared and the
right side is illustrated, it is as follows.

-2-
15.1.2 Double series expansion of Elliptic Integral of the 2nd kind

Formula 15.1.2
The following expressions hold for |k|  1 , |x|  1 .


x 1- k 2x 2
E(x, k) = dx
0 1- x 2
(-1)r -1/2
  
 r 1/2 2s 2r+1
= ΣΣ k x (1.2)
r=0 s=0 2r +1 r-s s
 r 1 (2r -2s -1)!! (2s -1)!! 2s 2r+1
= ΣΣ k x (1.2')
r=0 s=0 (2r +1)(1-2s) (2r -2s)!! (2s)!!

Proof
Since |k|  1 , |x|  1 , from Generalized Binomial Theorem .
1

 r x
-
2  -1/2
1- x  = Σ(-1)r
2 2r
r=0
1

 r k x
2 1/2
1- k x  = Σ(-1) r
2 2 2r 2r
r=0
Multiplying each other and integrating both sides of the result with respect to x from 0 to x, we obtain (1.2).
This is only what reversed the sign of 1/2 in the 2nd binomial coefficient in (1.1).
Next, from the definition of General Binomial Coefficient ,
(3/2) (3/2)
s
1/2
= =
(3/2-s)(s +1) (1/2-s)(1/2-s)(s +1)
On the other hand, from the properties of the Gamma Function (See 1.1.6 ) , the following expressions hold
for non-negative integer s,
 2s 
 2 =  2 -s = (-1)
3 1
 ,  s
2 (2s -1)!!
Using these ,


(-1)s (2s -1)!!
s
1/2 (2s -1)!!
s
= (-1) =
(1-2s)s! 2s  (1-2s) (2s)!!

 r -s 
-1/2 (2r -2s -1)!!
Substituting this and previous =(-1)r-s for (1.2) , we obtain (1.2') .
(2r -2s)!!

 
2
Example 2 : Double series expansion of E x ,
3
When arbitrary point x= 0.8 is given to this elliptic integral and its double series and both are compared,
it is as follows.

-3-
15.1.3 Triple series expansion of Elliptic Integral of the 3rd kind

Formula 15.1.3
The following expressions hold for |c|  1 , |k|  1 , |x|  1 .


x dx
(x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

(-1)r r-s -1/2

  k
 r s -1/2
= ΣΣΣ c 2t
x 2r+1 (1.3)
r=0 s=0 t=0 2r +1 s-t t
 r s (-c)r-s (2s -2t -1)!! (2t -1)!! 2t 2r+1
= ΣΣΣ k x (1.3')
r=0 s=0 t=0 2r +1 (2s -2t)!! (2t)!!

Proof
Since |c|  1 , |k|  1 , |x|  1 , from Generalized Binomial Theorem .
1 

1+ cx 2
= Σr=0
(-1)rc rx 2r
1

 x
-
2  -1/2
1- x  = Σ(-1)r
2 2r
r=0 r
1

 k
-
2  -1/2
1- k x  = Σ(-1)r
2 2 2r 2r
x
r=0 r
Multiplying each other and integrating both sides of the result with respect to x from 0 to x, we obtain (1.3) .
And replacing the binomial coefficients with the double factorial in (1.3) , we obtain (1.3') .

 
1 1
Example 3 : Triple series expansion of  x, ,
3 2
When arbitrary point x= 0.9 is given to this elliptic integral and its triple series and both are compared,
it is as follows.

-4-
15.2 Arc length of an ellipse
The ellipse which has foci on x-axis is shown by the following formula.
2 2
x y
2
+ 2 =1 (0< b < a) (1)
a b
Comparing this with cos  + sin  = 1 ,
2 2

x2 y2
2
= cos 2 , 2
= sin 2
a b
i.e.
x = a cos , y = b sin , 0   2
This is illustrated as shown in Fig.A. By this expression, since the arc length is calculated by the counter-
rclockwise rotation from x-axis, the calculation is complicated.
Therefore we think about only the 1st quadrant, and let us replace  with  /2- .
i.e.
x = a sin , y = b cos , 0    /2 (1')
Then, we can calculate the arc length clockwise from y-axis as shown in Fig.B. Because the calculation is
easy, we adopt expression (1').

Fig.A Fig.B

When the parametric equation of a curve on a plane is x = f() , y = g (), the length l is given by the
following expression.



2 2

  + 
dx dy
l=   
 d d
From (1') ,
dx dy
= a cos , = -b sin
d d
Then
2 2

   
dx dy
+ = a 2cos 2 + b 2sin 2 = a 2 - a 2-b 2sin 2
d d
= a 21- k 2sin 2  (k  a -b /a
2 2 2
: Eccentricity )

-5-
Therefore, the length l of bQ is as follows.

l =a 0
1- k 2sin 2 d (2.0)

Furthermore, let t = sin (= x /a) . Then,

x


1- k 2t 2
   
a x a 2-b 2
l =a dt = aE ,k k= (2.1)
0 1- t 2 a a2
Since the right side is a Elliptic Integral of the 2nd kind, using (1.2) in previous section , we obtain

(-1)r 2r+1

 r-s   s     
 r -1/2 1/2 x a 2-b 2
l = a ΣΣ k 2s
k= (2.1')
r=0 s=0 2r +1 a a2

Example : The length of bQ in the following figure

From a =3 , b =2 , xq =2 ,
32-22 5 xq 2
k= = , =
32 3 a 3
Then the length l of bQ is as follows from (2.1') .
2s

 r-s   s     
 r (-1)r -1/2 1/2 5 2 2r+1
l = 3ΣΣ
r=0 s=0 2r +1 3 3
When this is calculated, it is as follows.

-6-
15.3 Arc length of a lemniscate

15.3.0 Definition of a lemniscate

The locus of the point for which the product of the distance from some fixed points on a plane is a constant
is called Lemniscate of a wide sense. Especially, when the number of the fixed points is 2, and the distance
is FF' = 2c , and the constant product is c2 , it is called Lemniscate in a narrow sense.

The equation is drawn from FP 2F'P 2 = c 4 and is as follows.

2
x + y  - 2c x - y  = 0
2 2 2 2 2
Orthogonal coordinates :

Polar coordinates : r 2 -2c 2cos 2 = 0

2 2
Although these equations are used with this, the following formula replaced by 2c = a is often used.
In this chapter, these are used hereafter.
2
x + y  - a x - y  = 0
2 2 2 2 2
Orthogonal coordinates : (3.0o)

Polar coordinates : r 2 - a 2cos 2 = 0 (a >0) (3.0p)

Explicit function (Orthogonal coordinates)

From (3.0o),

2 -2x 2+ a 2  a 8x 2+ a 2
y =
2
Since a>0 , 8x 2+ a 2 > 0 , Employing +, we obtain

-2x 2+ a 2  a 8x 2+ a 2
y= (3.0h)
2
These zeros are obtained from x 2 - a 2x 2 =0 which is drawn from (3.0h) = 0. They are as follows.

x0 = a , =0
That is, the x-coordinate of A points is a in Fig.1. The expression of + of (3.h) is an upper half (blue) and
the expression of - is a lower half (red) in Fig.1.

Relation between orthogonal coordinates and polar coordinates

Substituting r = a cos 2 for x = r cos , y = r sin  , we obtain the following equations.

x = a cos 2 cos , y = a cos 2 sin (3.xy)

-7-
Next, substituting the following expressions for (3.xy),

cos = 1-sin 2 , cos 2 = 1-2sin 2

we can extract sin as follows.
3a 2  a a 2+8x 2 a 2  a a 2-8y 2
sin = = (3.0')
2a 2a
Furthermore, from this, we obtain .
3a 2  a a 2+8x 2 a 2  a a 2-8y 2
 = sin -1
= sin -1 (3.0")
2a 2a

Arc length from the right end point (polar coordinates)

In Fig.1, the figure of each quadrant is symmetrical with a point, and also symmetrical with a line.
So, we pay our attention only to the 1st quadrant.

Then, the length l of AP in Fig.2 is obtained as follows.

Differentiating (3.xy) with respect to ,

dx a cos sin 2
= -a sin cos 2 -
d cos 2
dy a sin sin 2
= a cos cos 2 -
d cos 2
From these,
2 2
a2
   
dx dy
+ =
d d cos 2
Hence,
 d  d 
l =a 
0 cos 2
=a  0 1-2sin  2  0  
4  (3.0)

15.3.1 Elliptic integral expression of arc length (Part 1)

Let t= 2 sin in (3.0) , then

t t2
dt = 2 cos d , sin = , cos = 1-
2 2

-8-
From these
d 1 dt 1 dt
= =
1-2 sin 2 2 cos
t2
 
2 1 2
1- t  1-
2
1-2 t
2 2
Then, the length l of AP in Fig.2 is obtained as follows.

  
a u dt a 1
l= = F u, (3.1)

   
2 0 2 2 2
1
1- t  1-
2
t2
2
3 a 2+8x 2
u= 2 sin = - (3.1u)
2 2a
Where, (3.1u) is obtained from t = 2 sin and (3.0') .
Since the right side of (3.1) is the elliptic Integral of the 1st kind, using (1.1) in previous section, we obtain

a  r (-1)r 2s

 r-s    
-1/2 -1/2 1
l= ΣΣ
2 r=0 s=0 2r +1 s 2
u 2r+1 (3.1')

Example1: Arc length from x= a to x= 0.8 at the time of a= 1 (Part 1)

Note

When K(k) =F(1, k) is a complete elliptic integral of the 1st kind, the length L/4 of AO in Fig.2
is given by the following expression.

 
L a 1 a
= K = 1.854074677301372 (3.1q)
4 2 2 2
Furthermore, the length l of OP in Fig.2 is drawn from (3.1) and (3.1q) as follows.

   
a 1 1
l = K - F t, (0 t  1) (3.1)
2 2 2

15.3.2 Elliptic integral expression of arc length (Part 2)

du
Let t = tan in (3.0) , then d = from  = tan -1t .
1+ t 2
On the other hand,

-9-
 1 cos 2 + sin 2 1+ tan 2 1+ t 2
= = =
1-2 sin 2 cos 2 - sin 2 1- tan 2 1- t 2
Then, the length l of AP in Fig.2 is

 
u 1+ t 2 dt u dt
l =a =a .
0 1- t 2 1+ t 2 0 1- t 2
1+ t 2

That is,


u dt
l =a = aF(t, i) (3.2)
0 2 2 2
1- t 1- i t

u = tan sin  3a 2  a a 2+8x 2

-1
2a  (3.2u)

Where, (3.2u) is obtained from u = tan  and (3.0") .

Although the right side of (3.2) is slightly irregular ( k = -1 ), since it is the elliptic Integral of the 1st kind,
using (1.1) in previous section, we obtain the following expression.
r (-1)r

 r-s   i
 -1/2 -1/2 2s 2r+1
l = a ΣΣ u (3.2')
r=0 s=0 2r +1 s

Example2: Arc length from x= a to x= 0.8 at the time of a= 1 (Part 2)

15.3.3 Single series expression of arc length

In the previous two examples, elliptic integral was calculated by the double series. On the other hand, (3.2)
is calculable by single series. According to the Generalized Binomial Theorem,

 x
1  -1/2
= Σ(-1)r 4r

1- x 4 r=0 r
From this,


(-1)r
 u
u dt  -1/2
l =a = aΣ 4r+1
(3.3)
0 1- t 4 r=0 4r +1 r
or


u dt  (2r -1)!! u 4r+1
l =a = aΣ (3.3')
0 1- t 4 r=0 (2r)!! 4r +1
The upper limit of the integration is the same as (3.2u), as follows.

u = tan sin  -1 3a 2  a a 2+8x 2

2a 
- 10 -
Example3: Arc length from x= a to x= 0.8 at the time of a= 1 (Part 3)
Using (3.3'), we try to calculate Example 2 by hand. The significant figure is to three numbers below a decimal
point. The upper limit of (3.3') at the time of a= 1 and x= 0.8 is same as Example2. Then,

u = tan sin -1 312  1 12+8(0.8)2

21  = 0.389
Substituting this for (3.3') and calculating the first 5 terms, we obtain the following.

(-1)!! 0.3891 0!! 0.3895 3!! 0.3899 5!! 0.38913 7!! 0.38917
s= + + + +
0!! 1 1!! 5 4!! 9 6!! 13 8!! 17
= 0.390

Note
After all, there are the following relations between the three integrals.
 d
  
1 t dt u du
= =
1-2sin 2
 
0 2 0 1 2 0 1-u 4
1- t  1- t
2
2
2u
t= 2 sin = 2 sintan -1u =
1+ u 2

 =
t t
u = tan = tan sin -1
2 2- t 2

- 11 -
15.4 Termwise Higher Calculus of Elliptic Integral

15.4.1 Termwise Higher Integral of Elliptic Integral

Formula 15.4.1
When |x|  1 , |k|  1 , |c|  1

 
x dx x 1- k 2x 2
F(x, k) = , E(x, k) = dx
0 1- x 1- k x 
2 2 2 0 1- x 2


x dx
 (x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

the following expressions hold for natural number n.


r (-1)r(2r)!

(2r + n +1)!  r - s   s 
x x  -1/2 -1/2
 F(x, k)dx =ΣΣ
n
k x 2s 2r+n+1
(4.1)
0 0 r=0 s=0

 
r

(2r + n +1)!  r - s   s 
x x (-1) (2r)! -1/2 1/2  r
 E( x, k)dx =Σ Σ n
k x 2s 2r+n+1
(4.2)
0 0 r=0 s=0

  x,c,k dx ΣΣΣ (2r +n +1)! c  s - t  t  k

x x -1/2 -1/2 
r
(-1) (2r)! r s
  ( ) = n r-s 2t
x 2r+n+1
0 0 r=0 s=0 t=0
(4.3)

Proof
From Formula 15.1.1 ,

(-1)r
 r-s   s  k x
 r -1/2 -1/2
F(x, k) = ΣΣ 2s 2r+1
r=0 s=0 2r +1
r

(2r +1)!  r - s   s 
(-1) (2r)! -1/2
 r -1/2
= ΣΣ k 2s 2r+1
x
r=0 s=0
Integrating both sides of this with respect to x from 0 to x n times, we obtain (4.1) . (4.2) and (4.3) are also
obtained from the Formula 15.1.2 and 15.1.3 in a similar way.

Note
The following expression using the double factorial is also possible. However, it is complicated and does not
have a merit.


x x  r (2r)! (2r -2s -1) !! (2s -1) !! 2s 2r+n+1
 F(x , k) dx n =ΣΣ k x
0 0 r=0 s=0 (2r +n +1)! (2r -2s) !! (2s) !!

   x , 
x x 2
Example :  dx 3
0 0 3
In the left side, the integrand of elliptic integral of the 2nd kind is integrated 3+1 times by Riemann-Liouville
integral. In the right side, this higher integral is calculated by the double series (4.2). Moreover, one arbitrary
point x= 0.8 is given to the both sides, and the values of the both sides are compared. Both are corresponding.

- 12 -
15.4.2 Termwise Higher Derivative of Elliptic Integral

Formula 15.4.2
When |x|  1 , |k|  1 , |c|  1 , x = x

 
x dx x 1- k 2x 2
F(x, k) = , E(x, k) = dx
0 1- x 1- k x 
2 2 2 0 1- x 2


x dx
 (x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

the following expressions hold for natural number n.

n r

 r-s   k
d  (-1) (2r)!
r -1/2 -1/2 2s 2r- n+1
F(x, k) = Σ Σ x (4.4)
dx n r=
n -1 s=0 (2r - n +1)!

s
2

dn r (-1)r(2r)!

 r-s   s  k
 -1/2 1/2
Σ Σ
2s 2r-n+1
E(x, k) = x (4.5)
dx n r=
n -1 s=0 (2r - n +1)!

2

dn s (-1)r(2r)!

 
-1/2
k
 r -1/2
r-s
 (x,c,k) = Σ Σ Σ c 2t
x 2r-n+1
dx n r=
n -1 s=0 t=0 (2r - n +1)!

s-t t
2
(4.6)

Proof
From Formula 15.1.1 ,

(-1)r(2r)!
 r-s   k
 r -1/2 -1/2
F(x, k) = ΣΣ 2s 2r+1
x
r=0 s=0 (2r +1)! s
Differentiating both sides of this with respect to x n times, we obtain (4.4). The 1st term r0 of outside Σ is
decided as the power 2r -n +1 of x is nonnegative. And there is a following relation between the 1st term
r0 and the order n of derivative.

n 0 1 2 3 4 5 
r0 0 0 1 1 2 2 

Such a relation can be expressed as follows using a ceiling function x = x.

n -1
r0 =  n 0
2
(4.5) and (4.6) are also obtained from the Formula 15.1.2 and 15.1.3 in a similar way.

- 13 -
 d3
 
1 1
Example :  x , ,
dx 3 3 2
In the left side, the integrand of the elliptic integral of the 3rd kind is differentiated 3- 1 times directly. In the
right side, this higher derivative is calculated by the double series (4.6). Moreover, one arbitrary point x= -0.9
is given to the both sides, and the values of the both sides are compared. Both are corresponding.

- 14 -
15.5 Termwise Super Calculus of Elliptic Integral

15.5.1 Termwise Super Integral of Elliptic Integral

Formula 15.5.1
When |x|  1 , |k|  1 , |c|  1

 
x dx x 1- k 2x 2
F(x, k) = , E(x, k) = dx
0 1- x 1- k x 
2 2 2 0 1- x 2


x dx
 (x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

p0.
the following expressions hold for


(-1)r(2r)!
 r-s   s  k x
x x  r -1/2 -1/2
 F(x, k)dx =ΣΣ
p 2s 2r+p+1
(5.1)
0 0 r=0 s=0 (2r + p +2)


r
ΣΣ (2r +p +2)  r - s   s  k x
x x (-1) (2r)! -1/2 1/2  r
 E( x, k)dx = p 2s 2r+p+1
(5.2)
0 0 r=0 s=0


r

 s-t   t  k
x x (-1) (2r)! -1/2 -1/2  r s
  (x,c,k )dx =Σ Σ Σ c p r-s 2t
x 2r+p+1
0 0 (2r + p +2) r=0 s=0 t=0
(5.3)

Proof
Analytically continuing the index of the integration operator in Formula 15.4.1 to [0 , p ] from[1 , n ],
we obtain the desired expressions.

  
x


x 1
Example : F x, dx 2
0 0 2
In the left side, the integrand of elliptic integral of the 1st kind is integrated 3/2+1 times by Riemann-Liouville
integral. In the right side, this higher integral is calculated by the double series (5.1). Moreover, one arbitrary
point x= 0.7 is given to the both sides, and the values of the both sides are compared. Both are corresponding,

15.5.2 Termwise Super Derivative of Elliptic Integral

Formula 15.5.2
When |x|  1 , |k|  1 , |c|  1 , x = x

- 15 -
  
x dx x 1- k 2x 2
F(x, k) = , E(x, k) = dx
0 1- x 1- k x 
2 2 2 0 1- x 2


x dx
 (x,c,k) =
1+ cx  1- x 1- k x 
0 2 2 2 2

p0.
the following expressions hold for

dp (-1)r(2r)!
 r-s   
 r -1/2 -1/2
F(x, k) = Σ Σ k 2s x 2r-p+1 (5.4)
dx p p -1 s=0 (2r - p +2) s
r= 
2
p
(-1)r(2r)!
 r-s   s 
d  r -1/2 1/2
E(x, k) = Σ Σ k 2s x 2r-p+1 (5.5)
dx p p -1 s=0 (2r - p +2)
r= 
2
p s (-1)r(2r)!

 
-1/2

d  r -1/2
r-s
 (x,c,k) = Σ ΣΣ c k 2t x 2r-p+1
dx p p -1 s=0 t=0 (2r - n +2) s-t t
r= 
2
(5.6)

Proof
Analytically continuing the index of the differentiation operator in Formula 15.4.2 to [0 , p ] from [1 , n],
we obtain the desired expressions.

 
d 3/2 2
Example : E x,
dx 3/2 3
In the left side, the 2nd order derivative of elliptic integral of the 2nd kind is integrated 1/2 times by Riemann-
Liouville integral. In the right side, this super derivative is calculated by the double series (5.5). Moreover, one
arbitrary point x= 0.6 is given to the both sides, and the values of the both sides are compared. Both sides are
corresponding,

2007.11.07
K. Kono
Alien's Mathematics

- 16 -
16 Higher Integral of the Product of Two Functions

16.1 Higher Integral of f (x) g (x)

16.1.1 Higher Intagration by parts

Formula 16.1.1
<r> <r>
Let f be the arbitrary r th order primitive function of f(x) and g be the r th order derivative function

g(x) for r =1, 2,  , n

<r> (r ) <r> (r)
of . And let f a , ga be the function values of f ,g on ak k =1,  , n
k k
Then, the following expressions hold.

 
x x n -1 r x x
 <0> (0) n
f
<n> (0)
g dx = f g - ΣΣrＣ s f a<n -r+s> g(as)  dx r
an a1 r=0 s=0 n-r n-r an a n- r+ 1

 f
n x x
- ΣnＣ r  <r> (r)
g dx n (1.1)
r=1 an a1

Especially, when ar = a for r =1, 2,  ,n ,

 f
x x
<n> (0) n -1 r
<n -r+s> (x -a)r
(s)
<0> (0)
g dx = f n
g - ΣΣrＣ s f a ga
a a r=0 s=0 r!

 f
n x x
- ΣnＣ r  <r> (r)
g dx n (1.2)
r=1 a a

f (a) = 0 (r =1, 2,  ,n) g (a) = 0 (s =0, 1,  ,n -1),

r
<> s ()
Especially, when or

  f
x x n x x
 f <0>g(0)dx n = f g(0) - ΣnＣ r 
<n> <r> (r)
g dx n (1.3)
a a r=1 a a

Proof
The following equations hold according to the formula of integration by parts.

 
x x
f <0>g(0)dx = f <1>g(0) - f <a11 >g(a01) - f <1>g(1)dx (001)
a1 a1

f - f
x x
<1> (0)
g dx = f <2>g(0) - f <a22>g(a02) <2> (1)
g dx (102)
a2 a2

f - f
x x
<1> (1)
g dx = f <2>g(1) - f <a21 >g(a11) <2> (2)
g dx (111)
a1 a1

f - f
x x
<2> (0)
g dx = f <3>g(0) - f <a33>g(a03) <3> (1)
g dx (203)
a3 a3

 f
x x
f <2>g(1)dx = f <3>g(1) - f <a32>g(a12) - <3> (2)
g dx (212)
a2 a2

f f
x x
<2> (2)
g dx = f <3>g(2) - f <a31 >g(a21) - <3> (3)
g dx (221)
a1 a1

Integrating both sides of (001) with respect to x from a2 to x ,

-1-
     f
x x x x x x
f <0>g(0)dx 2 = f <1>g(0)dx - f <a11 >g(a01) dx - <1> (1)
g dx 2
a2 a1 a2 a2 a2 a1
Substituting (102) for this,

 f
x x x
<0> (0) 2
f g dx = f <2>g(0) - f <a22>g(a02) - <2> (1)
g dx
a2 a1 a2

 dx -  f
x x x
<1> (0) <1> (1)
-f a1 ga1 g dx 2
a2 a2 a1

Here, integrating both sides of (111) with respect to x from a2 to x ,

 f f  dx -  f
x x x x x x
<1> (1) 2 <2> (1) <2> (1) <2> (2)
g dx = g dx - f a1 ga1 g dx 2
a2 a1 a2 a2 a2 a1
From this,

    f
x x x x x x
f <2>g(1)dx = f <a21 >g(a11) dx + f <1>g(1)dx 2 + <2> (2)
g dx 2
a2 a2 a2 a1 a2 a1
Substituting this for the above,

  dx
x x x
<0> (0) 2
f g dx = f <2>g(0) - f <a22>g(a02) - f <a11 >g(a01 ) + f <a21>g(a11)
a2 a1 a2

 f  f
x x x x
<1> (1)
-2 g dx 2 - <2> (2)
g dx 2 (2)
a2 a1 a2 a1

Next, integrating both sides of (2) with respect to x from a 3 to x ,

 f f
x x x x
<0> (0) 3 <2> (0)
g dx = g dx
a3 a2 a1 a3

   dx
x x x
- f <a22>g(a02) dx - f <a11>g(a01 ) + f <a21>g(a11 )
a3 a3 a2

- 2   f  f
x x x x x x
<1> (1)
g dx 3 - <2> (2)
g dx 3
a3 a2 a1 a3 a2 a1
Substituting (203) for this,

 f
x x x x
<0> (0)
f g dx 3 = f <3>g(0) - f <a33>g(a03) - <3> (1)
g dx
a3 a2 a1 a3

   dx
x x x
- f <a22>g(a02 ) dx - f <a11>g(a01 ) + f <a21>g(a11 )
a3 a3 a2

- 2   f  f
x x x x x x
<1> (1)
g dx 3 - <2> (2)
g dx 3
a3 a2 a1 a3 a2 a1

Here, integrating both sides of (212) with respect to x from a3 to x ,

    f
x x x x x x
f <3>g(1)dx = f <a32>g(a12) dx + f <2>g(1)dx 2 + <3> (2)
g dx 2
a3 a3 a3 a2 a3 a2
Substituting this for the above,

-2-
   dx
x x x x
f <0>g(0)dx 3 = f <3>g(0) - f <a33>g(a03) - f <a22>g(a02) + f <a32>g(a12)
a3 a2 a1 a3

  dx
x x
- f <a11 >g(a01 ) + f <a21 >g(a11)
a3 a2

 f g -  f g
x x x x
<2> (1) 2 <3> (2)
- dx dx 2
a3 a2 a3 a2

- 2   f  f
x x x x x x
<1> (1)
g dx 3 - <2> (2)
g dx 3
a3 a2 a1 a3 a2 a1

Here, integrating both sides of (111) , (221) with respect to x from a 2 to x respectivly,

    f
x x x x x x
f <1>g(1)dx 2 = f <2>g(1)dx - f <a21 >g(a11) dx - <2> (2)
g dx 2
a2 a1 a2 a2 a2 a1

 f   dx -  f
x x x x x x
<2> (2)
g dx 2 = f <3>g(2)dx - f <a31 >g(a21) <3> (3)
g dx 2
a2 a1 a2 a2 a2 a1

Furthermore, integrating both sides of these with respect to x from a3 to x respectivly,

    f
x x x x x x x x x x
f <1>g(1)dx 3 = f <2>g(1)dx 2- f <a21 >g(a11) dx 2- <2> (2)
g dx 3
a3 a2 a1 a3 a2 a3 a2 a3 a2 a1

 f =  f   dx -   f
x x x x x x x x x x
<2> (2)
g dx 3 <3> (2)
g dx 2- f <a31 >g(a21) 2 <3> (3)
g dx 3
a3 a2 a1 a3 a2 a3 a2 a3 a2 a1
i.e.

    f
x x x x x x x x x x
f <2>g(1)dx 2 = f <1>g(1)dx 3+ f <a21 >g(a11) dx 2+ <2> (2)
g dx 3
a3 a2 a3 a2 a1 a3 a2 a3 a2 a1

 f =   f   dx +   f
x x x x x x x x x x
<3> (2)
g dx 2 <2> (2)
g dx 3+ f <a31 >g(a21) 2 <3> (3)
g dx 3
a3 a2 a3 a2 a1 a3 a2 a3 a2 a1
Substituting this for the above,

 f
x x x
<0> (0) 3
g dx = f <3>g(0) - f <a33>g(a03)
a3 a2 a1

 dx
x
- f <a22>g(a02) + f <a32>g(a12)
a3

  dx
x x
- f <a11>g(a01 ) + 2f <a21 >ga(11) + f <a31 >g(a21) 2
a3 a2

   f

x x x x x x x x x
-3 f <1>g(1)dx 3-3 f <2>g(2)dx 3- <3> (3)
g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

= f <3>g(0) - f <a33>g(a03)

 
1 x x 2 x x
-Σ1Ｃ s f a<22+ s>g(as2)  dx 1 -Σ2Ｃ s f a<11+ >ga1
s (s)
 dx 2
s=0 a3 a 3-1 + 1 s=0 a3 a 3-2 + 1

 f
3 x x
- Σ3Ｃ r  <r> (r)
g dx 3
r=1 a3 a1

-3-
 
2 r x x
= f <3>g(0) - f <a33>g(a03) - ΣΣrＣ s f a<3-r+s>g(as)  dx r
r=1 s=0 3 -r 3 -r a a 3- r+ 1
3

 f
3 x x
- Σ3Ｃ r  <r> (r)
g dx 3
r=1 a3 a1
Thus,

 f 
x x x 2 r x x
<0> (0)
g dx = f <3>g(0) - f <a33>g(a03) - ΣΣrＣ s f a<33-r+s
3
-r
ga3-r 
> (s)
dx r
a3 a2 a1 r=1 s=0 a3 a 3- r+ 1

 f
3 x x
- Σ3Ｃ r  <r> (r)
g dx 3 (3)
r=1 a3 a1
Hereafter by induction, we obtain the following expression.

 f 
x x x x
<n> (0) <n> (0) n -1 r <n -r+s> (s)
<0> (0) n
g dx = f g - f a g a -ΣΣrＣ s f a ga dx r
an a1 n n r=1 s=0 n-r n-r an a n- r+ 1

 f
n x x
- ΣnＣ r  <r> (r)
g dx n
r=1 an a1
<n>
And pushing f a g(a0) into ∑∑, we obtain (1.1) .
n n

Next, when ar = a for r =1, 2,  ,n ,

 
x x x x (x -a)r
dx =r r
dx = for r =0, 1,  , n -1
an a n- r+ 1 a a r!
Then, substituting this for (1.1) , we obtain (1.2) .
Last, when f (a ) = 0 (r =1, 2,  ,n ) or (a ) = 0 (s =0, 1,  ,n -1) ,
<r> (s)
g (1.3) is clear.

Note
Different from the integration by parts of the 1st order, These formulas are seldom used directly.

16.1.2 Higher Integral of the Product of Two Functions

Theorem 16.1.2
<r> <r>
Let f be the arbitrary r th order primitive function of f(x) and g be the r th order derivative function

g(x) for r =1, 2,  , m +n -1 . Let f ar , g ar be the function values of f r , g r

<> () <> ()
of on a k for
k k
k =1, 2,  , n . And let  (n , m) be the beta function. Then, the following formulas hold.

a a r
x x m -1 -n
 f <0>g(0)dx n = Σ f <n + r> (r)
g
r=0
n 1

 s f a a
-n + r
n -1 m -1 x x
- ΣΣ <n -r+ s>
a n-r ga
(s)
 dx r
r=0 s=0 n-r
n n- r+1

a a
n -1 r -1 r -1 x x
t s  m +n-1 -r + tＣ m-1 f a
m+ n -r+s (m+ s)
+ (-1) m
Σ ΣΣ Ｃ
r =1 s =0 t=s n- r
ga
n- r
dx r
n n- r+1

a a f
(-1)m n -1 n -1Ｃ k x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx n (2.1)
k=0 m + k 1
n

-4-
 ar = a for r =1, 2,  ,n ,
Especially, when

  
x x m -1 -n
 f <0>g(0)dx n = Σ n r (r)
f < + >g
a a r=0 r
r

 
n -1 m -1 -n + r
<n -r+ s> (s) (x -a )
- ΣΣ fa ga
r=0 s=0 s r!
n -1 r -1 r -1 (x -a)r
t s  m +n-1 -r + tＣ m-1 f a
m + n -r+s (m+ s)
+ (-1) m
Σ ΣΣ Ｃ
r =1 s =0 t=s
ga
r!

 f
(-1)m n -1 n -1Ｃ k x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx n (2.2)
k=0 m + k a a

Especially, when f (a) = 0 (r =1, 2,  ,m+n -1) or g (s)(a ) = 0 (s =0, 1,  ,m+n -2),
<r >

  r f
x x m -1 -n
 f <0>g(0)dx n = Σ <n + r> (r)
g
a a r=0

 f
(-1)m n -1 n -1Ｃ k x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx n (2.3)
k=0 m + k a a

Proof
k =0, 1, 2, 
<k > (k )
According to Formula 16.1.1 , the 3rd order integral of f g are as follows

 f
x x x
<0> (0)
g dx 3 = f <3>g(0)
a3 a2 a1

 dx
x
- f <a3>g(a0) - f <a2>g(a0) + f <a3>g(a1)
3 3 2 2 2 2 a3

  dx
x x
- f <a11>g(a01 ) + 2f <a21 >ga(11) + f <a31 >g(a21) 2
a3 a2

   f

x x x x x x x x x
-3 f <1>g(1)dx 3-3 f <2>g(2)dx 3- <3> (3)
g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

 f
x x x
<1> (1)
g dx 3 = f <4>g(1)
a3 a2 a1

 dx
x
- f <a4>g(a1) - f <a3>g(a1) + f <a4>g(a2)
3 3 2 2 2 2 a3

g + f g   dx
x x
- f <a21>g(a11 ) + 2f <a31 > (2)
a1
<4> (3)
a1 a1
2
a3 a2

 f dx -3   f g dx -   f
x x x x x x x x x
<2> (2) 3 <3> (3) 3 <4> (4)
-3 g g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

 f
x x x
<2> (2)
g dx 3 = f <5>g(2)
a3 a2 a1

 dx
x
- f <a5>g(a2) - f <a4>g(a2) + f <a5>g(a3)
3 3 2 2 2 2 a3

-5-
   dx
x x
- f <a31>g(a21 ) + 2f <a41 >ga(31) + f <a51 >g(a41) 2
a3 a2

   f

x x x x x x x x x
-3 f <3>g(3)dx 3-3 f <4>g(4)dx 3- <5> (5)
g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

Substituting the 2nd less than expressions for the 1st expression one by one,

 f
x x x
<0> (0) 3 <3> (0) <4> (1)
g dx = f g -3 f g
a3 a2 a1

- f <a3>g(a0) - 3f <a4>g(a1)
3 3 3 3

 dx
x
- f a ga - 2f a ga - 3f a ga 
<2> (0) <3> (1) <4> (2)
2 2 2 2 2 2 a3

  dx
x x
- f <a11>g(a01 ) - f <a21 >g(a11 ) - 5f <a31 >ga(21) - 3f <a41>g(a31 ) 2
a3 a2

  +3   f

x x x x x x x x x
+6 f <2>g(2)dx 3 + 8 f <3>g(3)dx 3 <4> (4)
g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

= f <3>g(0) -3 f <4>g(1) + 6f <5>g(2)

- f <a3>g(a0) - 3f <a4>g(a1) + 6f <a5>g(a2)
3 3 3 3 3 3

 dx
x
- f <a2>g(a0) - 2f <a3>g(a1) + 3f <a4>g(a2) + 6f <a5>g(a3)
2 2 2 2 2 2 2 2 a3

  dx
x x
- f <a11>g(a01 ) - f <a21 >g(a11 ) + f <a31 >ga(21) + 9f <a41>ga(31) + 6f <a51>g(a41 ) 2
a3 a2

 f  f dx -6   f g

x x x x x x x x x
<3> (3) 3 <4> (4) 3 <5> (5)
-10 g dx -15 g dx 3
a3 a2 a1 a3 a2 a1 a3 a2 a1

This formula can be expressed as follows.

  0  1 f g
x x x -3 -3
f <0>g(0)dx 3 = f <3>g(0) + <4> (1)
a3 a2 a1

 0  f  1f g 
-3 <3> (0)
-3 <4> (1)
- a3 ga3 + a3 a3

 0  f  1  f g - 3f   dx
-2 <2> (0)
-2 <3> (1) <4> (2)
x
- a2 ga2 + a2 a2 a2 ga2
a3

 0  f  1  f g - 5f    dx
-1 <1> (0)
-1 <2> (1) <3> (2)
x x
- a1 ga1 + a1 a1 a1 ga1 - 3f <a41 >g(a31 ) 2
a3 a2

 f
2 Ｃ k 2(2+1)(2+2)
2 x x x
<2+ k > (2+ k )
+(-1)2Σ g dx 3
k =0 2! 2+ k a3 a2 a1

-6-
  0  1  2
-3 -3 -3
= f <3>g(0) + f <4>g(1) + f <5>g(2)

 0  f g  1  f g  2  f 
-3 -3 -3<3> (0) <4> (1) <5> (2)
- + + a3 a3 a3 a3 a3 ga3

 0   1  2 f   dx
-2 -2 -2<2> (0) <3> (1) <4> (2)
x
- f g + f g + a2 a2 a2 a2 a2 ga2
a3

 0  f g  1  f g  2  f    dx
-1 -1 -1<1> (0) <2> (1) <3> (2)
x x
2
- + + a1 a1 a1 a1 a1 ga1
a3 a2

 dx
x
- 6f <a5>g(a3)
2 2 a3

  dx
x x
- 9f <4> (3)
a1 ga1 + 6f <5> (4)
a1 ga1 
2
a3 a2

 f
2 2 Ｃk 3(3+1)(3+2) x x x
<3+ k > (3+ k )
Σ
3
+(-1) g dx 3
k =0 2! 3+ k a3 a2 a1

Here, red and magenta coefficients are the elements of the following Pascal type triangles. Bule coefficients are
3m
given, magenta coefficients are equal to c00 , and a red coefficient is obtained as the sum of two coefficients
of the upper row.
32
c00 3 Ｃ1
3

32
c10 32
c11 = 2 3 = 2Ｃ1 Ｃ1
3

32
c20 32
c21 32
c22 1 5 3 1Ｃ1 5 Ｃ1
3

33
c00 6 Ｃ2
4

33
c10 33
c11 = 3 6 = 3Ｃ2 Ｃ2
4

33
c20 33
c21 33
c22 1 9 6 2Ｃ2 9 Ｃ2
4

If bule coefficients are given, the other coefficients can be calculated by the following expression. ( See
Lemma 16.8.1 )
r -1
crs3m = ΣＣ t s-1 cr-1-t 0
3m
r, s  1
t=s -1

When
32
m =2 , c00 = 3Ｃ 1 , c10
32
= 2Ｃ 1 , c20
32
= 1Ｃ 1 i.e.
32
ct0 = 3-tＣ 1 t =0, 1, 2
0 = 3- r -1-tＣ 1 = 4 -r +tＣ 1
32
From this, cr-1-t
r -1
 crs32 = ΣＣ
t s-1 4 -r +tＣ 1 r, s  1
t=s -1

When
33
m =3 , c00 = 4Ｃ 2 , c10
33
= 3Ｃ 2 , c20
33
= 2Ｃ 2 i.e.
33
ct0 = 4-tＣ 2 t =0, 1, 2
0 = 4- r -1-tＣ 2 = 5-r +tＣ 2
33
From this, cr-1-t

-7-
 r -1
 crs33 = ΣＣ t s-1 5-r+ tＣ 2 r, s  1
t=s -1
r -1
Thus, crs3m = t s-1  m +2-r+ Ｃ
ΣＣ t m-1
t=s -1
If these are used, the above formula is expressed as follows.

 r  s f g
x x x 1 -3 <3+ r> (r)
1 -3 <3+ s> (s)
f <0>g(0)dx 3 = Σ f g -Σ a3 a3
a3 a2 a1 r=0 s=0

 s f  dx Σ  0  f   dx
-2 1
<2+ s> (s)
x -1 1
<1+ s> (s)
x x
-Σ a2 ga2 - a1 ga1 2
s=0 a3 s=0 a3 a2

   dx
x x x
32 <4> (2)
+ c11 f a ga 2 2
dx + c21
32 <3> (2) 32 <4> (3)
f a1 ga1 + c22 f a1 ga1 2
a3 a3 a2

 f
2 Ｃ k 2(2+1)(2+2)
2 x x x
<2+ k > (2+ k )
+(-1)2Σ g dx 3
k =0 2! 2+ k a3 a2 a1

r  s f
2 -3 <3+ r> (r)
2 -3 <3+ s> (s)
=Σ f g -Σ a3 ga3
r=0 s=0

s   0   dx
2 -2 s (s)
x 2 -1 s (s)
x x
-Σ f a<2+ >
ga2 dx - Σ f a<11+ >ga1 2
s=0 2 a3 s=0 a3 a2

   dx
x x x
- c11 f a2 ga2 
33 <5> (3)
dx - c21
33 <4> (3) 33 <5> (4)
f a1 ga1 + c22 f a1 ga1 2
a3 a3 a2

 f
2 Ｃ k 3(3+1)(3+2)
2 x x x
<3+ k > (3+ k )
+(-1)3Σ g dx 3
k =0 2! 3+ k a3 a2 a1

That is,

 f  r f g
x x x 1 -3 <3+ r> (r)
<0> (0)
g dx = Σ
3
a3 a2 a1 r=0

 s f a a
-3+ r
2 1 x x
- ΣΣ <3-r+ s> (s)
a 3- r ga  dx r
r=0 s=0 3- r
3 3- r+1

a a
2 r-1 x x
Σ Σ dx r
2
+ (-1) c 32
r 1+s fa <5-r+s> (2+s)
ga
r=1 s=0 3- r 3- r
3 3- r+1

 f
2 Ｃ k 2(2+1)(2+2)
2 x x x
<2+ k > (2+ k )
+(-1)2Σ g dx 3
k =0 2! 2+ k a3 a2 a1

 r f
2 -3 <3+ r> (r)
=Σ g
r=0

  a a
2 2 -3+ r x x
- ΣΣ f a<3-r+ >ga
s (s)
 dx r
r=0 s=0 s 3- r 3- r
3 3- r+1

a a
2 r-1 x x
+ (-1)3ΣΣc 33 <6-r+s> (3+s)
r 1+s f a ga  dx r
r=1 s=0 3- r 3- r
3 3- r+1

-8-
  f
2 Ｃ k 3(3+1)(3+2)
2 x x x
<3+ k > (3+ k )
+(-1)3Σ g dx 3
k =0 2! 3+ k a3 a2 a1

r -1
Here, if cr3m t s m +2- r +tＣ m-1 is used, the expansion of 0m -1 terms is as follows.
s+1 =Σ Ｃ
t=s

  r f
x x x m -1 -3 <3+ r> (r)
f <0>g(0)dx 3 = Σ g
a3 a2 a1 r=0

 s f a a
-3+ r
2 m -1 x x
- ΣΣ <3-r+ s> (s)
a 3- r ga  dx r
r=0 s=0 3- r
3 3- r+1

a a
2 r-1 r-1 x x
t sm +2-r +tＣ m-1 f a 
<m+3-r+s> (m+s)
+ (-1)mΣΣΣ Ｃ ga dx r
r=1 s =0 t=s 3- r 3- r
3 3- r+1

 f
2 2 Ｃ k m(m +1)(m +2) x x x
<m+ k > (m+ k )
+(-1)mΣ g dx 3 (3)
k =0 2! m +k a3 a2 a1
In a similar way, calculating for the 4th integral,

 f
x x
<0> (0) 4 <4> (0) <5> (1)
g dx = f g - 4f g
a4 a1

- f <a4>g(a0) - 4f <a4>g(a1)
4 4 4 4

 dx
x
- f <a3>g(a0) - 3f <a4>g(a1)
3 3 3 3 a4

- f a ga - 2f a ga   dx
x x
<2> (0) <3> (1) 2
2 2 2 2 a4 a3

- f g - f g    dx
x x x
<1> (0) <2> (1) 3
a1 a1 a1 a1
a4 a3 a2

+ 4f a ga  dx
x
<5> (2)
3 3 a4

+ 7f a ga + 4f a ga   dx
x x
<4> (2) <5> (3) 2
2 2 3 3 a4 a3

   dx
x x x
+ 9f <a31 >ga(21) + 11f <a41>ga(31) + 4f <a51>g(a41 ) 3
a4 a3 a2

   f g dx
x x x x
+ 10  f <2> g(2)dx 4 + 20  <3> (3) 4
a4 a1 a4 a1

+ 15 f + 4 f g dx
x x x x
<4> (4)
g dx 4 <5> (5) 4
a4 a1 a4 a1

= f <4>g(0) - 4f <5>g(1) + 10f <6>g(2)

- f <a4>g(a0) - 4f <a5>g(a1) + 10f <a6>g(a2)
4 4 4 4 4 4

 dx
x
- f <a3>g(a0) - 3f <a4>g(a1) + 6f <a5>g(a2)
3 3 3 3 3 3 a4

-9-
   dx
x x
- f <a2>g(a0) - 2f <a3>g(a1) + 3f <a4>g(a2) 2
2 2 2 2 2 2 a4 a3

   dx
x x x
- f <1> (0)
a1 ga1 - f <2> (1)
a1 ga1 + f <3> (2)
a1 ga1
3
a4 a3 a2

 dx
x
- 10f <a6>g(a3)
3 3 a4

  dx
x x
- 16f <a5>g(a3) + 10f <a6>g(a4) 2
2 2 2 2 a4 a3

   dx
x x x
- 19f <a41 >ga(31) + 26f <a51 >ga(41) + 10f <a61>g(a51) 3
a4 a3 a2

 f  f
x x x x
- 20  <3> (3)
g dx 4 - 45  <4> (4)
g dx 4
a4 a1 a4 a1

- 36 f - 10 f
x x x x
<5> (5)
g dx 4 <6> (6)
g dx 4
a4 a1 a4 a1

Here, red and magenta coefficients are the elements of the following Pascal type triangles. Bule coefficients are
4m
given, magenta coefficients are equal to c00 , and a red coefficient is obtained as the sum of two coefficients
of the upper row.
42
c00 4 Ｃ1
4
42 42
c10 c11 = 3 4 = Ｃ1
3 Ｃ1
4
42
c20
42
c21
42
c22 2 7 4 2 Ｃ1 7 Ｃ1
4
42
c30 c31
42
c32
42
c33
42
1 9 11 4 Ｃ1
1 9 11 Ｃ1
4

43
c00 10 Ｃ2
5

c10
43
c11
43
= 6 10 = Ｃ2
4 5 Ｃ2
43
c20
43
c21
43
c22 3 16 10 3 Ｃ2 16 Ｃ2
5
43 43 43 43
c30 c31 c32 c33 1 19 26 10 Ｃ2
2 19 26 Ｃ2
5

As well as before, if bule coefficients are given, the other coefficients can be calculated by the following
equation.
r -1
crs4m = Σt s-1  m +3-r+ Ｃ
Ｃ t m-1
t=s -1
If these are used, the above formula is expressed as follows.

 f  r f
x x 1 -4 <4+ r> (r)
<0> (0) 4
g dx = Σ g
a4 a1 r=0

 s f a a
-4+ r
3 1 x x
- ΣΣ <4-r+ s> (s)
a 4- r ga4-  dx r
r=0 s=0 r
4 4- r+1

a a
3 r-1 x x
+ ΣΣc 42 <6-r+s> (2+s)
r 1+ s f a ga dx r
r=1 s=0 4- r 4- r
4 4- r+1

- 10 -
  f
3 Ｃ k 2(2+1)(2+2)(2+3) x x

3 <2+ k > (2+ k )
+ (-1)2Σ g dx 4
k =0 3! 2+ k a4 a1

 r f g
2 -4 <4+ r> (r)
=Σ
r=0

 s f a a dx
3 -4+ r
2 x x
- ΣΣ <4-r+ s> (s)
a 4- r ga4- r  r
r=0 s=0
4 4- r+1

a a dx
3 r-1 x x
7-r+s (3+s)
- ΣΣc 43
r 1+ s f a ga r
r=1 s=0 4- r 4- r
4 4- r+1

 f
3 Ｃ k 3(3+1)(3+2)(3+3) x x

3 <3+ k > (3+ k )
+ (-1)3Σ g dx 4
k =0 3! 3+ k a4 a1

r -1
Here, if t s  m +3-r+ Ｃ
s+1 =Σ Ｃ
cr4m t m-1 is used, the expansion of 0m -1 terms is as follows.
t=s

 r
x x m -1 -4
 f <0>g(0)dx 4 = Σ f <4+ >g
r (r)
a4 a1 r=0

 s f a a
-4+ r
3 m -1
<4-r+ s> (s)
x x
- ΣΣ a 4- r ga4- dx r
r=0 s=0 r
4 4- r+1

a a
3 r-1 r -1 x x
t s m +3-r +tＣ m-1 f a 
m +4 -r+s (m+s)
+ (-1)mΣΣΣ Ｃ ga dx r
r=1 s=0 t=s 4- r 4- r
4 4- r+1

 f
3 Ｃ k m(m +1)(m +2)(m +3) x x

3 <m+ k > (m+ k )
+ (-1)mΣ g dx 4
k =0 3! m +k a4 a1

Hereafter, in a similar way, the expansion of 0 m -1 terms of n th order integral is obtained as follows.

a a r
x x m -1 -n
 f <0>g(0)dx n = Σ f<
n + r> (r)
g
r=0
n 1

 s f a a
-n + r
n -1 m -1 x x
- ΣΣ <n -r+ s>
a n-r ga
(s)
 dx r
r=0 s=0 n-r
n n- r+1

a a
n -1 r -1 r -1 x x
t s m + n -1- r + tＣ m-1 f a 
m +n -r+s m +s
+ (-1)m ΣΣΣ Ｃ ga dx r
r =1 s =0 t=s n- r n- r
n n- r+1

m(m +1)(m + n -1)

a a f
n -1 n -1 Ｃk x x
<m+ k > (m+ k )
+ (-1)m Σ g dx 4
k=0 (n -1)! m +k n 1
Here,
m(m +1)(m + n -1) (m + n -1)! (m + n) 1
= = =
(n -1)! (n -1)!(m -1)! (n)(m) (n , m)
Then, substituting this for the above,

a a  r f
x x m -1 -n
 f <0>g(0)dx n = Σ <n + r> (r)
g
r=0
n 1

- 11 -
   a a
n -1 m -1 -n + r <n -r+ s> (s)
x x
- ΣΣ fa ga dx r
r=0 s=0 s n-r n-r
n n- r+1

a a
n -1 r -1 r -1 x x
t s m + n -1- r + tＣ m-1 f a
m +n -r+s m +s
+ (-1) m
Σ ΣΣ Ｃ
r =1 s =0 t=s n- r
ga
n- r
dx r
n n- r+1

a a f
(-1) n -1 n -1Ｃ k
m x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx 4 (2.1)
k=0 m + k 1
n

ar = a for r =1, 2,  ,n ,
Especially, when

  
x x m -1 -n
 f <0>g(0)dx n = Σ n r (r)
f < + >g
a a r=0 r

   dx
n -1 m -1 -n + r x x
n s (s)
- ΣΣ f a< -r+ > ga r
r=0 s=0 s a a

  dx
n -1 r -1 r -1 x x
t s m + n -1- r + tＣ m-1 f a 
m +n -r+s m +s
+ (-1)m ΣΣΣ Ｃ ga r
r =1 s =0 t=s a a

 f
(-1) n -1 n -1Ｃ k
m x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx 4
k=0 m + k a a
Here, since


x x (x -a)r
dx =r
r = 0, 1, 2,  ,n -1
a a r!
substituting this for the above, we obtain (2.2) .

Last, when f (a ) = 0 (r =1, 2,  ,m+n -1) or g (a) = 0 (s =0, 1,  ,m+n -2) ,

<r> s ()

(2.3) is clear. Q.E.D.

Note
As mentioned in 4.1.3, the polynomial expression for the higher integral of 1 is difficult.

  e sin x dx
x x
x 2
Example 1
2 1

When n =2, m =3 , from (2.1) ,

a a  r f
x x 3-1 -2 <2+ r> (r)
f <0>g(0)dx 2 = Σ g
r=0
2 1

 s f a a
-2+ r
2-1 3-1
<2-r+ s> (s)
x x
- ΣΣ a 2-r ga dx r
r=0 s=0 2-r
2 2- r+1

a a
2-1 r -1 r -1 x x
t s  3+2-1- r + tＣ 3-1 f a
3+2-r+s 3+s
Σ ΣΣ Ｃ dx r
3
+ (-1) ga
r =1 s =0 t=s 2- r 2- r
2 2- r+1

a a f
(-1)3 2-1 2-1Ｃ k x x
<3+ k > (3+ k )
(2, 3) Σ
+ g dx 2
k=0 3+ k 2 1

- 12 -
 r
2 -2 <2+ r> (r)
=Σ f g
r=0

 s f  s f a dx
-2 2
<2+ s> (s) 2 -1 <1+ s> (s)
x
-Σ a2 ga -Σ a1 ga 1
s=0 2 s=0 1
2

a dx
x
- 3 f <a41> g(a31)
2

  f
x x x x
-4 f <3>g(3)dx 2 - 3 <4> (4)
g dx 2
a2 a1 a2 a1

Let f = e x , g = sin x , a1 = 1 , a2 = 2 , then

a a f   e sin x dx
x x x x
<0> (0) 2 x 2
Left: g dx =
1 2 1
2

e xcos x - e 2cos 2 e 1x(sin 1- cos 1)

=- - + e 1(sin 1- cos 1)
2 2
Next,
r
 
<r> <2+ r> (r)
f =f = ex , g = sin x +
2
s s
= e , ga = sin 2+
2   
s> s> (s) (s)
f a<2+ = e 2 , f a<1+ 1
, ga = sin 1+
2 1 2 1 2
3
a a f =   e sin  x +
2 
x x x x
<3> (3) 2 x 2
g dx dx
1 2 1
2

e xsin x - e 2sin 2 e 1x(sin 1+ cos 1)

=- + - e 1(sin 1+ cos 1)
2 2
4
a    
x x x x
f <4>g(4)dx 2 = e x sin x + dx 2
2 a1 2 1 2
e xcos x - e 2cos 2 e 1x(sin 1- cos 1)
=- - + e 1(sin 1- cos 1)
2 2
Substituting these for the right,

r r
r   r  
2 -2 2 -2
Right: = e xΣ sin x + - e 2Σ sin 2+
r=0 2 r=0 2
s x -2 3 x -2
 s   2  1!
-1
 2  1!
2
-e Σ 1
sin 1+ - 3 e sin 1+ 1
s=0

- 4 - - e (sin 1+ cos 1)

x 2 1
e sin x - e sin 2 e x(sin 1+ cos 1) 1
+
2 2

- 3 - + e (sin 1- cos 1)

x 2 1
e cos x - e cos 2 e x(sin 1- cos 1) 1
-
2 2
= -2e xcos x -2e xsin x + 2e 2cos 2 + 2e 2sin 2
+ 4x e 1cos 1 - 8e 1cos 1

- 13 -
 + 2e xsin x -2e 2sin 2-2e 1x sin 1-2e 1x cos 1+4e 1sin 1+4e 1cos 1
3 3 3 3
+ e xcos x - e 2cos 2+ e 1x sin 1- e 1x cos 1-3e 1sin 1+3e 1cos 1
2 2 2 2
1 1 1 1
= - e xcos x + e 2cos 2 - e 1x sin 1 + e 1x cos 1 + e 1sin 1 - e 1cos 1
2 2 2 2
This is consistent with the left. And, since this result contains the constant-of-integration polynomial with
degree 1, we can see that this 2nd order integral is collateral.


x x
Example 2 e x sin x dx 2
2 1
4 4
When n =2, m =3 , from (2.1) ,

a a f  r f
x x 2 -2 <2+ r> (r)
<0> (0)
g dx = Σ 2
g
r=0
2 1

 s f s a dx
-2 2
<2+ s> (s) 2 -1 s> (s)
x
-Σ a2 ga - Σ f a<1+
1
ga1
s=0 2 s=0
2

a dx
x
- 3 f <a41> g(a31)
2

 f  f
x x x x
<3> (3) 2 <4> (4)
-4 g dx - 3 g dx 2
a2 a1 a2 a1

Let f = e x , g = sin x , a1 = 1 /4 , a2 = 2 /4 , then

a a f 
x x
<0> (0) 2
x x
x e xcos x 2
Left: g dx = e sin x dx = -
2 1 2 1 2
4 4
Next,
r
 
<r> <2+ r> (r)
f =f = ex , g = sin x +
2
2 1
2 s 1 s
f a2<2+ s>
=e 4
, f a1 <1+ s>
=e 4
, g(as2) = sin  4
+
2  , g(as1) = sin  4
+
2 
3
a a   
x x x x
f <3>g(3)dx 2 = e x sin x + dx 2
2 1 2 1 2
4 4
1 1
2
2  2
 
4 4
1 x 4 e x 2 e
=- e sin x - e sin + -
2 4 2 4
4
a a f =   e sin  x +
x

2 
x x x x e cos x
<4> (4) 2 x 2
g dx dx = -
2 1 2 1 2
4 4
Substituting these for the right side,
2
s 2 s
s  s 
-2 -2
 
2 2
Right: e x
Σ
s=0
sin x +
2
-e 4
Σ
s=0
sin
4
+
2

- 14 -
 1
1 s 2
s 
-1
 
2
-e 4
Σ
s=0
sin
4
+
2
x-
4
1
1 3 2
- 3e 4
sin  4
+
2  x-
4  1 1

 
2
2  2
 +
1 x 4 e 4
x 2 e 4
3e xcos x
-4 - e sin x - e sin - +
2 4 2 4 2
= -2e xcos x -2e xsin x
2 1
4 4
+ 2e + 2e x 2
1
- e 4
 2
2 1 1
3e xcos x
x
+ 2e sin x - 2e 4
- 2e 4
x 2 +e 4
 2 +
2
e xcos x
=-
2
This is consistent with the left. And, since this result does not contains the constant-of-integration polynomial,
we can see that this 2nd order integral is lineal.

  e sin x dx
x x
x 2
Example 3
0 0
When n =2, m =3 , from (2.2) ,

 f  r f
x x 2 -2 <2+ r> (r)
<0> (0)
g dx = Σ 2
g
a a r=0

 s f  
-2 2
<2+ s> (s) (x -a)0 2 -1 <1+ s> (s) (x -a)1
-Σ a ga -Σ fa ga
s=0 0! s=0 s 1!
1
(x -a)
- 3 f <a4> g(a3)
1!

 f  f
x x x x
<3> (3) 2 <4> (4)
-4 g dx - 3 g dx 2
a a a a
x
Let f = e , g = sin x , a = 0 , then

 f 
x x
<0> (0) 2
x x
x e xcos x2 x +1
Left: g dx = e sin x dx = - +
a a 0 0 2 2
Next
r
 
<r> <2+ r> (r)
f =f = ex , g = sin x +
2
s> s
f a<1+ = f <a4> = e 0 , g(as) = sin
2
3
  
e xsin x

x x x x x
<3> (3) 2 x 2
f g dx = e sin x + dx = - +
a a 0 0 2 2 2
4
 f =  e sin  x +
e xcos x
2 
x x x x x +1
<4> (4) 2 x 2
g dx dx =- +
a a 0 0 2 2

- 15 -
Substituting these for the right side,

r
r  
2 -2
Right: = e xΣ sin x +
r=0 2
s x 3 x r

 s  2 r!
1 2 -2+ r 1
- e ΣΣ
0
sin - 3 e sin 0
r=0 s=0 2 1!

- 4 - +  - 3 -
2 
x x
e sin x x e cos x x +1
+
2 2 2
= -2e xcos x - 2e xsin x + 2 + x + 3x
x 3e xcos x 3x +3
+ 2e sin x - 2x + -
2 2
x
e cos x x 1
=- + +
2 2 2
This is consistent with the left. And, since this result contains the constant-of-integration polynomial with
degree 1, we can see that this 2nd order integral is collateral.

  e sin x dx
x x
x 2
Example 4
- -
When n =2, m =3 , from (2.3) ,

  f g
(-1)3 1 1Ｃ k
r
x x 2 -2 <n + r> (r)
x x
<3+ k > (3+ k )
dx =Σ
(2, 3) Σ
<0> (0) 2
f g f g + dx 2
a a r=0 k =0 3+ k a a

 s f  dx - 3 f
-22
<n + s> (s)
x x x x
=Σ g -4 f<3>g(3) 2 <3> (3)
g dx 2
s=0 a a a a

Let f = e x , g = sin x , a = - , then

 f 
x x
<0> (0) 2
x x
x e xcos x
2
Left: g dx = e sin x dx = -
a a - - 2
Next,
r
 
<r> <2+ r> (r)
f =f = ex , g = sin x +
2
3
 f  
e xsin x

x x x x
<3> (3) 2 x 2
g dx = e sin x + dx = -
a a - - 2 2
4
 f =   e sin  x +
e xcos x
2 
x x x x
<4> (4) 2 x 2
g dx dx =-
a a - - 2
Substituting these for the right side,

r
   
e xsin x e xcos x
r 
-2

2
Right: =e x
Σ
r=0
sin x +
2
-4 -
2
-3 -
2
x 3e xcos x x
= e (sin x - 2cos x - 3sin x) + 2e sin x +
2
e xcos x
=-
2

- 16 -
This is consistent with the left. And, since this result does not contains the constant-of-integration polynomial,
we can see that this 2nd order integral is lineal.

Remark 1
The following thing can be seen from Example 2 and Example 4 which are lineal together.

   e sin x dx
<2> e xcos x x x x x
e sin x
x x x
2 2
=- = e sin x dx =
2 2 1 - -
4 4
That is, the higher integral which gives the higher primitive function of the product of two functions is not
necessarily unique.

Remark 2
As understood from the above proof process and Example 1,2 , (2.1) holds unconditionally and is perfect.
However, it is too complicated and the practical use is difficult. (2.2) is not simplified considering the
conditions.
(2.3) is the most practicable compared with these. The condition looks severe apparently. But, let

  f(x)dx (r =1, 2,  ,n)

x x
f
<r>
=  r
a a

f (a)= 0 (r =1, 2,  ,n) holds without trouble.

<r>
Then

<r> <r> r-1 xk

Therefore, in Example 3, if not f =e x
but f =e -Σ x
was adopted, not (2.2) but (2.3) was
k =0 k!
<r >
applicable. But if it does so, since f will become a collateral higher primitive function, the calculation may
be complicated. Like Example4, In the case that the zero of the lineal higher primitive function is consistent
with the lower limit a of the integral of f g , the calculation is the easiest. Therefore, below, we adopt this case
as much as possible, and will mainly use (2.3).

16.1.3 Riemann-Liouville Integral Expression

According to Formula 4.2.1 in 4.2 , Higher Integral is expressed by Riemann-Liouville
Integral as follows.

 f(x)dx  (x -t)

x x 1 x
n n-1
= f(t)dt
a a (n) a
If (2.2), (2.3) in Theorem 16.1.2 are rewritten using this, it is as follows.

Theorem 16.1.3
<r> <r>
Let f be the arbitrary r th order primitive function of f(x) and g be the r th order derivative function

g(x) for r =1, 2,  , m +n -1 . Let f a , g a be the function values of f r , g r on a k for

<r > (r ) <> ()
of
k k
k =1, 2,  , n . And let  n be the gamma function and  (n , m) be the beta function. Then, the
( )
following formulas hold.

 r
1 x m -1 -n
(x -t)n-1 f g dt = Σ f <n + r> (r)
g
(n) a r=0

(x -a)r
 s f
-n + r
n -1 m -1
<n -r+ s> (s)
- ΣΣ a ga
r=0 s=0 r!
n -1 r -1 r -1 (x -a)r
t s  m + n -1- r + tＣ m-1 f a
m +s + n -r m +s
+ (-1) m
Σ ΣΣ Ｃ
r =1 s =0 t=s
ga
r!

- 17 -
  (x -t)
(-1)m n -1 n -1Ｃ k x
n-1 <m+ k > (m+ k )
+ Σ
(n , m)(n) k=0 m + k a
f g dt (3.1)

Especially, when f (a) = 0 (r =1, 2,  ,m+n -1) or g (s)(a ) = 0 (s =0, 1,  ,m+n -2),
<r >

 r
1 x m -1 -n
(x -t)n-1 f g dt = Σ f <n + r> (r)
g
(n) a r=0

 (x -t)
(-1)m n -1 n -1Ｃ k x
n-1 <m+ k > (m+ k )
+ Σ
(n , m)(n) k=0 m + k a
f g dt (3.2)

- 18 -
16.2 Higher Integral of x^a f (x) (general)
There are two features in the higher integral of x  f(x) . That is
1 When  is a positive integer, the remainder and a part of the constant-of-integration polynomial disappear .
2 When the commonn lower limit of Higher Integral is 0, the constant-of-integration polynomial disappear.

Formula 16.2.0
Let (z) be the gamma function,  (n , m) be the beta function, f <r> be the arbitrary r th primitive
<r > <r>
function of f(x) and fa be the function values of f on a k . Then the following expressions hold for
k
a natural number n.
(1)
(1+ ) - r
a a r
x x m -1 -n
 f <0>x  dx n = Σ f <n + r>
x
n 1
r=0 1+ - r 
(1+ ) - s x
 s f  a
-n + r
n -1 m -1 x
- ΣΣ <n -r+ s>
a  dx r
r=0 s=0
a n-r 1+ - s  n-r an n- r+1
n -1 r -1 r -1
t s  m + n -1- r + tＣ m-1
+ (-1)m ΣΣΣ Ｃ
r =1 s =0 t=s
(1+ )
 a
x x
- m- s
 f am + n -r+ s an-r  dx r
n- r 1+ - m - s  an n- r+1

(1+ )
a a f
(-1) n -1 n -1Ｃ k
m x x
 m- k
 <m+ k >
(n , m) Σ
+ x - dx n (0.1)
k=0 m + k
n 1
1+ - m - k 
Especially, when  = m =0, 1, 2, 
(1+m) m- r
a a r
x x m -n
 f <0>x m dx n = Σ f <n + r>
x
n 1
r=0 1+m - r 
(1+m) m- s x
 f  a
n -1 m -n + r x
- ΣΣ <n -r+ s>
a  dx r (0.2)
r=0 s=0 s a n-r 1+m - s n-r an n- r+1

f (a) = 0 (r =1, 2,  ,m+ n -1) or a = 0

<r>
Especially, when

(1+ ) - r
  
x x m -1 -n
 f <0>x  dx n = Σ n r
f<+> x
a a r=0 r 1+ - r 
(1+ )

(-1)m n -1 n -1Ｃ k x x <m+ k >  m- k
+ Σ  f x - dx n (0.3)
(n , m) k=0 m k a a
+ 1+ - m - k 
Where, if  = -1, -2, -3,  , it shall read as follows.
(1+ ) (- + r)
 (-1)-r r = r, s, m+ s , m+ k
1+ - r  (- )
(2) When   -1, -2, -3,  &  + n  -1, -2, -3, 
(1+ )
  
x x m -1 -n

 x f(0)dx n = Σ x +n+ f
r (r)
an a1 r=0 r (1+ + n + r)
(1+ )
 s  (1+ + n -r + s) a a a
-n + r
n -1 m -1
+n-r+ s (s)
x x
- ΣΣ n-r fa  dx r
r=0 s=0 n-r
n n- r+1

- 19 -
 n -1 r -1 r -1
t s  m + n -1- r + tＣ m-1
+ (-1)m ΣΣΣ Ｃ
r =1 s =0 t=s
(1+ )
 a
x x
+ m + n-r+ s (m+ s)
 an-r fa  dx r
1+ + m + n - r + s n- r an n- r+1

(1+ )
a a
(-1) n -1 n -1Ｃ k
m x x
+m
 m+ k
(n , m) Σ
+ x +k f ( )dx n (0.4)
k=0 m + k
n 1
(1+ + m + k)
a = 0 or f (a) = 0 (s =0, 1,  ,m+ n -2)
(s)
Especially, when

(1+ )
  
x x m -1 -n

 x f(0)dx n = Σ r (r)
x +n+ f
a a r=0 r (1+ + n + r)
(1+ )

(-1)m n -1 n -1Ｃ k x x +m
+ Σ  m+ k
x +k f ( )dx n (0.5)
(n , m) k=0 m + k a a (1+ + m + k)

Proof (1)
When   -1, -2, -3, 
(r) (1+ ) - r (m+ k ) (1+ )  m- k
x  = x , x  = x -
1+ - r 1+ - m - k 
Then, substituting these for (2.1) in Theorem 16.1.2,
(1+ ) - r
a a f  
x x
<0> 
m -1 -n
 x dx = Σ n
f <n + r>
x
n 1
r=0 r 1+ - r
(1+ ) - s x x
   
n -1 m -1 -n + r
- ΣΣ fa<n -r+ s>
an-r  dx r
r=0 s=0 s n-r 1+ - s a n a n- r+1
(1+ )
 a
x x
- m- s
 f am +s + n -r an-r  dx r
n- r 1+ - m - s  an n- r+1
n -1 r -1 r -1
t s  m + n -1- r + tＣ m-1
+ (-1)m ΣΣΣ Ｃ
r =1 s =0 t=s

(1+ )
a a f
(-1) n -1 n -1Ｃ k
m x x
 m- k
 <m+ k >
(n , m) Σ
+ x - dx n (0.1)
k=0 m + k
n 1
1+ - m - k 
When  = m -1 m =0, 1, 2, 
(1+ - m - s)=  s =0, 1, 2, 3, 
(1+ - m - k)=  k =0, 1, 2, 3, 
Hence, ΣΣΣ and the remainder term in (0.1) disappear, and is as follows.
(m) m-1- r
a a  
x x m -1 -n
 f <0>x
m-1
dx n = Σ f <n + r>
x
n 1
r=0 r m - r 
(m) m-1- s x
   a
n -1 m -1 -n + r x
- ΣΣ n s
f a< -r+ > a  dx r
r=0 s=0 s n-r m - s n-r an n- r+1
Then, replacing m-1 with m, we obtain (0.2) .

f r(a ) = 0 (r =1, 2,  ,m+ n -1) or a = 0 , the condition of

<>
Next, when (2.3) in Thorem 16.1.2 is
satisfied. Then we obtain (0.3) .

- 20 -
 When  = -1, -2, -3,  , from (5.5) in 1.1.5
(-z) (1+z + n)
= (-1)-n ( n is a nonnegative integer )
(-z - n) (1+ z)
Then replaceng -z = 1+ , n = r , we obtain the proviso.

Proof (2)
When   -1, -2, -3,  &  + n  -1, -2, -3, 
n r (1+ )  n+ r m+ k (1+ )  m+ k
f<+> = x + , f< > = x +
1+ + n + r 1+ + m + k
(1+ ) + n-r+ s
f a -r+  =
n s
an-r
n- r 1+ + n - r + s 
(1+ ) + m + n-r+ s
f a -r+  =
m+n s
an-r
n- r 1+ + m + n - r + s 
Then, substituting these for (2.1) in Theorem 16.1.2 ,
(1+ )
a a r
x x m -1 -n  n+ r (r)
 x  g(0)dx n = Σ x + g
n 1
r=0 1+ + n + r
(1+ )
 s   a
-n + r
n -1 m -1
+ n-r+ s (s)
x x
- ΣΣ an-r ga  dx r
r=0 s=0 1+ + n - r + s  n-r a
n n- r+1
n -1 r -1 r -1
+ (-1)m ΣΣΣ Ｃ t s  m +n-1 -r + tＣ m-1
r =1 s =0 t=s
(1+ )
 
x x
+ m + n-r+ s m +s
 an-r ga  dx r
1+ + m + n - r + s  n- r a n a n- r+1
(1+ )

m n -1
n -1Ｃ k
(-1) x x
 m+ k m+ k
+ Σ  x + g ( )dx n
(n , m) k=0 m + k an a1 1+ + m + k
Here, replacing g with f , we obtain (0.4) .

Last, when a = 0 or f (a ) = 0 (s =0, 1, 

(s)
,m+ n -2), since the condition of the Theorem
16.1.2 (2.3) is satisfied, we obtain (0.5) .


x x
Example 1 x log x dx 2
2 1
Let n =2, m =3 , f = log x ,  =1/2 , a1 =1 , a2 =2 . And integrating the left directly

a a 
x x x x
Left: f <0>x  dx 2 = x log x dx 2
1 2 1
2
5 5

   
4 log x 64 4 log 2 64 4
= - x2 - - 2 2 + (x -2)
15 225 15 225 9
Next ,
log x - (1+ r)-  r
a  dx = x -2
x x
<r>
f = (log x)<r> = x , dx 1 =
(1+ r) 2
2

r log x - (1+2+ r)-  2+ r log 1 - (1+4)-  4

f <2+ > = x , f <a4> = 1
(1+2+ r) 1 (1+4)

- 21 -
 s log 2 - (1+2+ s)-  2+ s s log 1 - (1+1+ s)-  1+ s
f a<2+ > = 2 , f a<1+ > = 1
2 (1+2+ s) 1 (1+1+ s)
Substituting these for the right side of (0.1) ,

log x - (1+2+ r)-  2+ r (1+1/2)

1

r
2 -2 -r
Right: =Σ x x2
r=0 (1+2+ r) 1+1/2- r 
log 2 - (1+2+ s)-  2+ s (1+1/2)
1

s
2 -2 -s
-Σ 2 22
s=0 (1+2+ s) 1+1/2- s 
-1 log 1 - (1+1+ s)  (1+1/2) 2 - s
1

 s  (1+1+ s) 1
2 - 1+ s
-Σ 1 (x -2)
s=0 1+1/2- s 
log 1 - (1+4)-  4 (1+1/2)
1
-3
-3 1 1 2 (x -2)
(1+4) 1+1/2-3
log x - (1+3)-  3 (1+1/2)
1


x x -3
-4 x x 2 dx 2
2 1 (1+3) 1+1/2-3
log x - (1+4)-  3 (1+1/2)
1


x x -4
-3 x x 2 dx 2
2 1 (1+4) 1+1/2-4
5

       x
1 3 1 11 1 25 2
= log x - - log x - - log x -
2 2 6 6 32 12
5

       2
1 3 1 11 1 25 2
- log 2- - log 2- - log 2-
2 2 6 6 32 12
79 25
+ (x -2) + (x -2)
144 256
5 5

   
1 29 1 29 5
- log x - x2+ log 2- 22 - (x -2)
15 10 15 10 12
5 5

   
1 63 1 63 55
+ log x - x2- log 2- 22 + (x -2)
32 20 32 20 256
5 5

   
4 log x 64 4 log 2 64 4
= - x2 - - 2 2 + (x -2)
15 225 15 225 9
This is consistent with the left side.

  x log x dx
x x
3 2
Example 2
2 1
Let n =2, m =3 , f = log x ,  =1/2 , a1 =1 , a2 =2 . And integrating the left directly

a a   x log x dx
x x x x
Left: f <0>x  dx 2 = 3 2

1 2 1
2

   
log x 9 log 2 9 1
= - x5 - - 25 + (x -2)
20 400 20 400 16
Next ,

- 22 -
 log x - (1+ r)-  r
 
x x
<r>
f = (log x)<r> = x , dx 1 = dx = x -2
(1+ r) a2 2

r log x - (1+2+ r)-  2+ r

f <2+ > = x
(1+2+ r)
s log 2 - (1+2+ s)-  2+ s s log 1 - (1+1+ s)-  1+ s
f a<2+ > = 2 , f a<1+ > = 1
2 (1+2+ s) 1 (1+1+ s)
Substituting these for the right side of (0.2) ,

log x - (1+2+ r)-  2+ r (1+3) 3- r

 
3 -2
Right: =Σ x x
r=0 r (1+2+ r) 1+3- r
-2 log 2 - (1+2+ s)-  2+ s (1+3) 3- s
 
3
-Σ 2 2
s=0 s (1+2+ s) 1+3- s
-1 log 1 - (1+1+ s)-  1+ s (1+3) 3- s
 
3
-Σ 1 1 (x -2)
s=0 s (1+1+ s) 1+3- s

         x
1 3 1 11 3 25 1 137 5
= log x - - log x - + log x - - log x -
2 2 1 6 4 12 5 60

 
log 2 9 1
- - 25 + (x -2)
20 400 16

=
400 
x - 
log x 9 5 log 2 9 1
- - 25 + (x -2)
20 20 400 16


x x
Example 3 x log x dx 2
0 0
Let n =2, m =3 , f = log x ,  =1/2 . And integrating the left side directly
5

 f 
x x

 
x x 4 log x 64
<0>  2 2
Left: x dx = x log x dx = - x2
0 0 0 0 15 225
Next,
<r> log x - (1+ r)-  r
f = (log x)<r> = x
(1+ r)
Substituting these for the right side of (0.3) ,

log x - (1+2+ r)-  2+ r (1+1/2)

1

r
2 -2 -r
Right: =Σ x x2
r=0 (1+2+ r) 1+1/2- r 
log x - (1+3)-  3 (1+1/2)
1


x x -3
-4 x x 2 dx 2
0 0 (1+3) 1+1/2-3
log x - (1+4)-  3 (1+1/2)
1


x x -4
-3 x x 2 dx 2
0 0 (1+4) 1+1/2-4
5

       x
1 3 1 11 1 25 2
= log x - - log x - - log x -
2 2 6 6 32 12
5 5

   
1 29 1 63
- log x - x2+ log x - x2
15 10 32 20

- 23 -
 5

 
4 log x 64
= - x2
15 225


x x
Example 4 x sin x dx 2
2 1
Let f = sin x , then

r (m + k)
   
(r) m+ k )
f = sin x + , f( = sin x +
2 2
r (m + s)
(s)
fa = sin a +  n-r ,
m+ s
f (a ) = sin an-r+  
n-r 2 n- r 2
Substitute these for (0.4) ,
(1+ ) r
a a r
-n
 
x x m -1

 x  sin x dx n = Σ
r
x +n+ sin x +
n 1
r=0 (1+ + n + r) 2
(1+ ) r
 s  (1+ + n -r + s) a a a
-n + r

n -1 m -1 x x
+n-r+ s
-Σ Σ n-r sin an-r+  dx r
r=0 s=0 2
n n- r+1
n -1 r -1 r -1
t s  m + n -1- r + tＣ m-1
+ (-1)m ΣΣΣ Ｃ
r =1 s =0 t=s
(1+ ) (m + s)
 a a
x x
+ m + n-r+ s
 an-r sin an-r+  dx r
1+ + m + n - r + s 2
n n- r+1

(1+) (m + k) 

(-1) n -1 n -1Ｃ k x
m x
Σ  x dx n  
+m+k
+ sin x +
 (n , m) k=0 m + k an a1 (1+ + m + k) 2

This right side is very complicated. However, fortunately, in the case of this function f = sin x , if m 
ΣΣΣ and the remainder term converge to 0. And the following expression holds.

(1+ ) r
a a r
-n
 
x x 

 x  sin x dx n = Σ
r
x +n+ sin x +
n 1
r=0 (1+ + n + r) 2
(1+ ) r
 s  (1+ + n -r + s) a a a
-n + r

n -1  x x
+n-r+ s
-Σ Σ n-r sin an-r+  dx r
r=0 s=0 2
n n- r+1
Substituting n =2,  =1/2, a 1 =1, a 2 =2 for this,

(1+1/2) r
1

  r  (1+1/2+2+ r)
-2
 
x x  +2+ r
x sin x dx = Σ 2
x 2
sin x +
2 1 r=0 2
0
1
(1+1/2)
 s  (1+1/2+2+ s)  
 -2 +2+ s
-Σ 2 2
sin 2+
s=0 2
1
1
(1+1/2)
 s  (1+1/2+1+ s)  
 -1 +2+ s
-Σ 1 2
sin 1+ (x -2)
s=0 2
When the both sides are illustrated by mathematical software, it is as follows. Both overlap exactly and
blue (left) can not be seen.

- 24 -
- 25 -
16.3 Higher Integral of x^a f (x) (particulars)
In this section, substituting various functions f for the Formula 16.2.0 in previous section, we obtain a various
formula. There are (1) and (2) in Formula 16.2.0, and we may also choose whichever. However, what we want

is the expression or approximation of higher integral of x f(x ) by the series. So, in the selection (1) or (2),
we choose the way where such a well-behaved series is obtained.
Moreover, also in which formula, if  = -1, -2, -3,  , it shall read as follows.
(1+ ) (- + r)
 (-1)-r r = r, s, m+ s , m+ k
1+ - r  (- )

16.3.1 Higher Integral of (ax+b ) p(cx+d) q

Formula 16.3.1
The following expressions hold for p > 0 and a natural number n.

 
x x
p q n
(ax +b) (cx +d) dx
b b
- -
a a

(1/a)n+ r (1+p)(1+q) (ax +b)p+n+ r

r
m -1 -n
=Σ + Rmn (1.1)
r=0 (1/c) (1+p +n + r)(1+q -r) (cx +d)
r r-q

(-1)m n -1 n -1Ｃ k m+ k
(1+p)(1+q)
 
c
(n , m) Σ
Rmn =
k=0 m + k a (1+p +m + k)(1+q -m - k)

 
x x
  (ax +b)p+m+ k (cx +d)q-m- k dx n (1.1r)
b b
- -
a a
lim Rmn = 0
m 

Especially, when m =0, 1, 2, 

 
x x
 (ax +b)p (cx +d)mdx n
b b
- -
a a

(1/a)n+ r (1+p)(1+m) (ax +b)p+n+ r

r
m -n
=Σ (1.1')
r=0 (1/c)r (1+p +n + r)(1+m -r) (cx +d)r-m

3
Example 1st order integral of x-2 3x+4
The zeros of this primitive function are x =-4/3 , 2 . If -4/3 is adopted, since x -2 is a complex
number, this higher integral becomes a complex function. It is inconvenient. Then if we adopt x =2 as
the lower limit of the integral, since a =1, b =-2, p =1/2 , c=3, d =4, q =1/3 , n =1 , substituting
these for (1.1), (1.1r) , we obtain


x
3
x -2 3x +4 dx
2
(3/2)(4/3)
3 1

r
m -1 -1 +r -r
=Σ 3r
(x -2) 2 (3x +4) 3 + Rm1
r=0 (5/2+ r)(4/3-r)

- 26 -
 (3/2)(4/3)
1 1

 (x -2)
x +m -m
m m
Rm1 = (-1) 3 2
(3x +4) 3
dx
(3/2+ m)(4/3-m) 2
1
This remainder becomes lim R =0 . And the convergence is quick. When m =5 , if both sides are
m  m
illustrated, it is as follows. Both overlap exactly and blue (left) can not be seen. In addition, this integral is
not an elementary function.

16.3.2 Higher Integral of x log x

Lineal higher primitive functions of x  log x are as follows.
+1
 <1> x ( +1)log x - 1
x log x =
( +1)
2

+2
x ( +1)( +2)log x -2 -3
x  log x
<2>
=
( +1) ( +2)
2 2


x +3( +1)( +2)( +3)log x -3 -12 -11
2
x  log x
<3>
=
( +1) ( +2) ( +3)
2 2 2


And when  > -n , the zeros of these are all x =0 .

Formula 16.3.2
When (z), (z) are the gamma function and the psi function respectively, the following expressions hold
for  , n such that  + n > 0 .
log x -(1+n+r) -  (1+) +n
  
x x m -1 -n
 x  log x dx n = Σ x + Rmn (2.1)
0 0 r=0 r ( 1+n+r )  (1+  -r )
log x -(1+m+k) -  (1+) x 
m+ k  
(-1) m
Ｃ n -1 n -1 k x x
Rmn = Σ  dx n (2.1r)
 (n , m) k=0 0 0 (1+m+k) (1+ - m- k)
lim Rmn = 0
m 

Especially, when m =0, 1, 2, 

- 27 -
 log x -(1+n+r) -  (1+ m) m+ n
 r
x x m -n
 x mlog x dx n = Σ x (2.1')
0 0 r=0 (1+n+r) (1+ m -r)

Complete Automorphism
Observing well the Formula 16.3.2, we notice that the integral of the completely same type as the left side
is included in the remainder. In such a case, we can take out the integral of the purpose by transposition.

Formula 16.3.2'
The following expression holds for ,n such that  +n > 0.

 x
x x

 log xdx n
0 0
log x - (1+ n)-  (1+ ) n -1 n -1Ｃ k  (2+ k)+ 
(1+ n)
+n Σ
1+ + n  k=0 (1+ k)2 (1+ k,  - k) +n
= x
n -1 n -1Ｃ k 1
1 + nΣ 2 (1+ k,  - k)
k=0 (1+ k)
(2.2)

Calculation
Let m =1 in (2.1), (2.1r), then
log x - (1+n+0)-  (1+ ) +n
  
x x 1-1 -n
 x  log xdx n = Σ x + R1n
0 0 r=0 0 (1+n+0) (1+ -0)
log x - (1+1+k)-  (1+ )x 

(-1)1 n -1 n -1Ｃ k x x

(n , 1) Σ
R1n = dx n
k =0 1+ k 0 0 (1+1+ k) (1+ -1- k)
(1+ )

Ｃk
n -1 n -1 x x
= -nΣ  x  log xdx n
k =0 1+ k (2+ k)( - k) 0 0
n -1 n -1Ｃ k (1+ ) (2+ k)+ 


x x
+ nΣ  x dx n
k=0 1+ k (2+ k)( - k) 0 0
Here,
(1+ ) +n
  x dx
x x

(2+k) = (1+ k)(1+k) ,  n
= x
0 0 1+ + n 
Substituting these for the remainder,


n -1 Ｃk 1 x x
 x  log xdx n
n -1
Rmn = -nΣ 2 (1+ k,  - k)
k=0 (1+ k) 0 0
n -1 n -1Ｃ k  (2+k)+  (1+ ) +n
+ nΣ 2 (1+ k,  - k)  1+ + n x
k=0 (1+ k)  
Thus
log x - (1+ n)-  +n

x x
 x  log xdx n = x
0 0 (1+n)


n -1 n -1Ｃ k 1 x x
- nΣ 2 (1+ k,  - k)
 x  log xdx n
k=0 (1+ k) 0 0
n -1 n -1Ｃ k  (2+k)+  (1+ ) +n
+ nΣ 2 (1+ k,  - k)  1+ + n x
k=0 (1+ k)  
From this, we obtain (2.2).

- 28 -
Example When n =1 ,
log x - (2)-  (1+ ) 0Ｃ 0  (2)+ 
+
(2) 2+  12 (1 ,  ) +1

x

x log xdx = x
0 0Ｃ 0 1
1+
12 (1 ,  )

log x -1 + +1
1+
 
+1 x 1
= x = log x -
1 +  +1  +1
But in this case, direct calculation is far easier !

16.3.3 Higher Integral of x sinx , x cosx

Lineal higher primitive functions of x 3sin x are as follows.
<1>
x 3 sin x = 3x 2 -6sin x - x 3-6xcosx
<2>
x sin x = -6x 2 -24cosx - x 3-18xsin x
3

<3>
x sin x = -9x 2 -60sin x + x 3-36xcosx
3

<4>
x 3 sin x = 12x 2 -120cosx + x 3-60xsin x

And the zeros of these are x =0 , 4.9762 , 0 , 3.1224 , 0 ,  respectively. Then the lower
3
limit of the higher integral of x sin x is variable. Similarly, the lower limit of the lineal higher integral of

x  sin x for any  is variable.

Formula 16.3.3
(1) When m =0, 1, 2,  .
(1+ m) m- r (n + r)
a a r
-n
 
x x m
 x m sin x dx n = Σ x sin x -
n 1
r=0 1+ m - r 2
(1+ m) m- s (n -r + s)
  a a
-n + r

n -1 m x x
- ΣΣ an-r sin an-r -  dx r
r=0 s =0 s  ( 1+ m - s ) 2
n n- r+1
(3.1)

Especially, when a 1 , a2 ,  , a n are the zeros of the lineal heigher primitive of x sin x m

(1+ m) m- r (n + r)
a a r
-n
 
x x m
 x m sin x dx n = Σ x sin x - (3.1')
n 1
r=0 1+ m - r 2
(2) When 
 -1, -2, -3,  &  + n  -1, -2, -3,  .
(1+ ) r
  
m -1 -n

 
x x

 x  sin x dx n = Σ r
x +n+ sin x +
an a1 r=0 r (1+ + n + r) 2
(1+ ) s
  a a
-n + r

n -1 m -1 x x
+n-r+ s
-Σ Σ an-r sin an-r+  dx r
r=0 s=0 s (1+ + n -r + s) 2
n n- r+1
+ Rmn (3.2)

- 29 -
 n -1 r -1 r -1
t s  m + n -1- r + Ｃ
Rmn = (-1)m ΣΣΣ Ｃ t m-1
r =1 s =0 t=s

(m + s)
a a
(1+)

x x
+ m + n-r+ s
 an-r sin an-r+  dx r
 1+ + m + n -r + s  2
n n- r+1

(1+) (m + k) 

(-1) n -1 n -1Ｃ k x
m x
Σ  x dx n  
+m+k
+ sin x +
 (n , m) k=0 m + k a a 1 (1+ + m + k) 2
n

lim Rmn = 0
m 

In addition, the formula of x  cos x is obtained only by replacing sin x by cos x .

Example 1
(2+ r)
(1+3) 3- r
   
-2

x x 3
x 3 sin x dx 2 = Σ sin x - x
2 1 r=0 r 1+3- r 
2
(2+ s) (1+3) 3- s
  
-2

3
-Σ sin 2- 2
s=0 s 2 1+3- s
(1+ s) (1+3) 3- s
  
-1

3
-Σ sin 1- 1 (x -2)
s=0 s 2 1+3- s
= -x 3 -18 xsin x - 6x 2 -24cosx
+ (3x -6)sin 1 - (5x -10)cos 1 - 28 sin 2
Especially, when a 1=0 , a 2 =4.9762 ,

(2+ r) (1+3) 3- r

   
-2

x x 3
x 3 sin x dx 2 = Σ sin x - x
a2 a1 r=0 r 2 1+3- r
= -x 3 -18 xsin x - 6x 2 -24cosx

Example 2
See Example 4 in 16.2 .

16.3.4 Higher Integrals of x sinhx , x coshx

Lineal higher primitive functions of x 3sinh x are as follows.

   
<1> x 3 3x 2 x 3 3x 2
x sinh x
3
= e -x
+ +3x +3 +e x
- +3x -3
2 2 2 2

 2 +3x +9x +12 - e - -9x +12

3 3
<2> x x
x sinh x
3
= -e -x 2 x
+3x 2
2


And the zeros of these are x =0 , 2.7085 , 0 , 3.1224 i  , 0 ,  respectively. Then the lower
3
limit of the higher integral of x sinh x is variable. Similarly, the lower limit of the lineal higher integral of

x sinh x for any  is variable.

- 30 -
Formula 16.3.4
(1) When m =0, 1, 2,  .
(1+ m) m- r e x -(-1)n+ re -x
aa x r
x x m -n
m
sinh x dx = Σ n
x
n 1
r=0 1+ m - r 2
a n-r -a n-r
(1+ m) m- s e
a a
-(-1)n-r+ s e
 
n -1 m -n + r x x
- ΣΣ a  dx r
r=0 s =0 s 1+ m - s  n-r 2
n n- r+1
(4.1)
Especially, when a1 , a2 ,  , an are the zeros of the lineal heigher primitive of
m
x sinh x
(1+ m) m- r e x -(-1)n+ re -x
aa x r
x x m -n
m
sinh x dx = Σ n
x (4.1')
n 1
r=0 1+ m - r 2
(2) When 
 -1, -2, -3,  &  + n  -1, -2, -3,  .
(1+ )

x -r -x

 
x x m -1 -n
 +n+ r e -(-1) e
 x sinh x dx = Σ
n
x
an a1 r=0 r (1+ + n + r) 2
a n-r -a n-r
(1+)
a a
-(-1)-se
 
-n + r x x
n -1 m -1
+n-r+ s e
-Σ Σ an-r  dx r
r=0 s=0 s (1+ + n -r + s) 2
n n- r+1

+ Rmn (4.2)
n -1 r -1 r -1
t s  m + n -1- r + Ｃ
Rmn = (-1)m ΣΣΣ Ｃ t m-1
r =1 s =0 t=s
a n-r -a n-r
(1+)
a a
+ m + n-r+ s e -(-1)-m-se x x
　 an-r  dx r
 ( 1+ + m + n -r + s ) 2
n n- r+1

a a
(-1) m
Ｃk n -1 n -1 x x (1+) +m+k e -(-1)
x -m-k -x
e
+ Σ  x dx n
(n , m) k =0 m + k n 1
(1+ + m + k) 2
lim Rmn = 0
m 

In addition, the formula of x  cosh x is obtained by only replacing -(-1) by +(-1) .


x x
Example x sinh x dx 2
2 1
From (4.2) ,

(1+1/2)
1


e x -(-1)-re -x
 r  (1+1/2+2+ r) x
x x  -2 +2+ r
x sinh x dx = Σ 2 2
2 1 r=0 2
1 2
(1+1/2) e -(-1)-se -2
 s  (1+1/2+2+ s) 2
 -2 +2+ s
-Σ 2
s=0 2
1 1
(1+1/2) e -(-1)-se -1
 s  (1+1/2+1+ s) 1
 -1 +2+ s
-Σ 2
(x -2)
s=0 2
When the both sides are illustrated by mathematical software, it is as follows. Both overlap exactly and
blue (left) can not be seen.

- 31 -
Left:
 a:=1/2:
 Fl := int(int(t^a*sinh(t),t=1..u),u=2..x)
x u
t  sinh(t) d t d u
2 1

Right:
 fx := r-> (E^x-(-1)^-r*E^-x)/2:
 gx := (r,s)-> gamma(1+a)/gamma(1+a+2-r+s)*x^(a+2-r+s):
 m:=15:
 f1 := sum(binomial(-2,s)*gx(0,s)*fx(s), s=0..m):
 f2 := -sum(binomial(-2,s)*subs(gx(0,s)*fx(s),x=2), s=0..m):
 f3 := -sum(binomial(-1,s)*subs(gx(1,s)*fx(s),x=1), s=0..m)*(x-2):
 Fr := f1+f2+f3:

- 32 -
16.4 Higher Integral of log x f (x)

16.4.1 Higher Integral of (log x)2

2
Lineal higher primitive functions of (log x) are as follows.
2 <1>
(log x)  = x(log x (log x -2)+2)
2 <2> 1
(log x)  = x 2(2log x (log x -3)+7)
4
2 <3> 1 3
(log x)  = x (6 log x (3 log x -11)+85)
108

And the zeros of these are all x =0 .

Formula 16.4.1
n 1
When Hn = Σ =  (1+n)+ is a harmonic number, the following expression holds.
j=1 j
n


x x x
2 n
log xdx = log xlog x -Hn
0 0 n!
(r)
 
m -1 -n
+ Σ (-1)r-1 n
x log x -Hn+r+ Rmn (1.1)
r=1 r (1+ n +r)
n
1 x n -1 n -1Ｃ k
Rmn Σ (-1)k-1
(m + k) 
= log x - Hn - Hm+ k (1.1r)
(n , m) n ! k=0 2

lim
m
R1 = 0
 m

Calculation
f(x)= g(x)= log x , f (0) = 0 (r =1, 2,  ,m+ n -1).
<r>
Since Then (2.3) in Theorem 16.1.2
is applicable.
log x - (1+ n + r)-  n+ r r-1 (r)
(log x)<n + r> =
(r)
x , (log x) = (-1)
(1+ n + r) xr
Substituting these for (2.3) in Theorem 16.1.2 ,
log x - (1+ n)- log x
  log xdx
x x
 2 n
= x
n
0 0 (1+ n)
log x - (1+ n +r)-(r)
r
m -1 -n
+ Σ (-1)r-1 n
x + Rmn
r=1 (1+ n +r)
log x -Hn+r(r)
n

 
x m -1 -n
= log x log x -Hn + Σ (-1)r-1 n
x + Rmn
n! r=1 r (1+ n + r)
log x -(1+ m + k) -  m+ k (m+ k) n
(-1)m n -1 n -1Ｃ k

x x
Rmn = 
 (n , m) Σ
x (-1)m+ k-1 dx
k=0 m + k 0 0 (1+ m + k) x m+ k
n -1Ｃ k
  log x - (1+ m + k)  dx
1 n -1 x x
= Σ (-1)k-1  -
n
(n , m) k =0 (m + k)
2
0 0

- 33 -
 n -1Ｃ k log x -(1+ n) -  (1+ m + k) +  n
 
1 n -1
= Σ
 (n , m) k =0
(-1)k-1
(m + k)
2
(1+ n)
-
(1+ n)
x

n
1 x n -1 n -1Ｃ k
Σ (-1)k-1
(m + k) 
= log x - Hn - Hm+ k
(n , m) n ! k=0 2

2
Example 3rd order integral of log x
Let n =3, m =1000 in (1.1). And if both sides are illustrated, it is s follows. Since the convergence of
logarithm family is usually slow, both sides are not overlapping completely.
Left:
 g := int(int(int(ln(x)^2,x),x),x)
x  ln(x) 11  x  ln(x) 85  x3
3 2 3
    
6 18 108

Right: Series
 n:=3: m:=1000: MAXDEPTH:=1000:
 f := ln(x)*(ln(x)-psi(1+n)-EULER)*x^n/gamma(1+n) +
sum((-1)^(r-1)*binomial(-n,r)*(ln(x)-psi(1+n+r)-EULER)
*gamma(r)/gamma(1+n+r)*x^n, r=1..m-1):

- 34 -
16.5 Higher Integral of e^x f (x)

16.5.1 Higher Integral of e x x

When (a , x) denots the incomplete gamma function, the lineal higher primitive functions of e x x  can be
expressed as follows.

 <1> x  (1+ , -x)x 0 +1

e x  = -x E-(-x)
x
=
(-x) 0!

 <2> x  (2+ , -x)x 0 + (1+ , -x)x 1

e x 
x
=
(-x) 1!

 <3> x  (3+ , -x)x 0 + 2(2+ , -x)x 1 + (1+ , -x)x 2

e x 
x
=
(-x) 2!

 <n> x  n -1 Ｃ r (n -r + , -x)x r
n -1
e x 
x
= Σ
(-x) r=0 (n -1)!
Since these zeros are all -
and these do not include constant-of-integration polynomial, surely these are
x 
the lineal higher primitive functions of e x . And these are complex functions generally so that clearly from
(-x)  .
x 
Although it seems that the lineal higher primitive functions of e x can be immediately obtained from these,
it does not go so. When these are calculated numerically, the sighn of the imaginary number part is contrary
to the result of Riemann-Liouville integration at the time of x >0 . That is, these are right at the time of x 0 ,
and are not right at the time of x >0 . So, I found the following formula that holds also at the time of x >0 .

Formula 16.5.1

t

When (a , x) = a-1 -t
e dt denots the incomplete gamma function, the following expressions hold
x
for   -1,-2, -3, .
(1) When x  0
x  n -1 Ｃ r (n -r + , -x)x r

x x
 x
 x e dx =
n -1
Σ
(-x) r=0
n
(n -1)!
(1.n- )
- -
(2) When x >0

x  n -1 Ｃ r (n -r + , -x)x r

x x
 x
 x e dx =
n -1
Σ
(-x) r=0
n
(n -1)!
- -

Ｃ r (n -r + )x r
n -1 n -1
+ 2i sin Σ (1.n+ )
r=0 (n -1)!

Proof
As mentioned above, the lineal higher primitive function of exx was as follows.

 <n> x  n -1 Ｃ r (n -r + , -x)x r

n -1
e x 
x
= Σ
(-x) r=0 (n -1)!
From this, when x <0 , (1.n- ) holds immediately.

- 35 -
And when   -1, -2, -3, ,
x (-1) x 0
=
(-x) (-1)- x >0
Then, (1.n- ) holds as the limit of x -0 also at the time of x =0 .
When x >0 , we calculate separately for [-, 0] and [0 , x] . If the lineal higher integral is displayed
separately by Riemann-Liouville Integral, it is as follows..

 (x -t )
1 x
n-1  t
t e dt
(n) -

  (x -t )
1 0 1 x
= (x -t )n-1t e t dt + n-1  t
t e dt (w)
(n) - (n) 0

Calculation of the 1st term

Since x 0 , then x  /(-x) = (-1) . So, the 1st,

 (x -t )
0
n-1  t
t e dt
-


<1> 0 0
= (x -t )n-1t e t  -+ (n -1) (x -t )n-2t e t  dt
<1>

-
0
t t 0 (1+ , -t)
   (x -t )
0
t e 
t <1>
= (x -t ) n-1
+ (n -1) n-2
 dt
(-t)  0!
-
-

(1+ )
 (x -t )
0
= (-1) x n-1 t e t  dt
n-2 <1>
+ (n -1)
0! -
the 2nd,

 (x -t )
0
t e t  dt
n-2 <1>

-


<2> 0 0
= (x -t )n-2t e t  -+ (n -2) (x -t )n-3t e t  dt
<2>

-
(2+ )
+ (n -2) (x -t )
0
= (-1) x n-2 t e t  dt
n-3 <2>
1! -
the 3rd,,

 (x -t )
0
t e t  dt
n-3 <2>

-
(3+ )
 (x -t )
0
= (-1) x n-3 t e t  dt
n-4 <3>
+ (n -3)
2! -

the last (n-1)th,


0
(x -t )1t e t  dt
<n -1>

-
(n -1+ )
 (x -t ) t e 
0
= (-1) x 1
(n -2)!
+ 1 0  t <n -1>
dt
-

- 36 -
Substituting these for the former expression one by one, we obtain the following expression.

 (x -t )
0 n -1 (n -1)!
n-1  t 
t e dt = (-1) Σ (n -r + )x
r
- r=1 r !(n -1-r)!

 (x -t ) t e 
0
+ (n -1)! 0  t <n -1>
dt
-
Thus,
(n -r + )x r
 (x -t )
1 0 n -1
n-1  t 
(n)
t e dt = (-1) Σ
r=1
n -1Ｃ r
(n -1)!
-


0 <n -1>
+ t e t  dt (w1)
-
Calculation of the 2nd term
Since x > 0 , then x  /(-x) = 1/(-1) . So, the 1st,

 (x -t )
x
n-1  t
t e dt
0


<1> x x
= (x -t )n-1t e t  0+ (n -1) (x -t )n-2t e t  dt
<1>

0
x
t t (1+ , -t)
   (x -t )
0 x
t e t
<1>
= (x -t ) n-1
 0!
+ (n -1) n-2
 dt
(-t) 0
0

(1+ )
n-1

 (x -t )
x x
t e t  dt
n-2 <1>
=- 
+ (n -1)
(-1) 0! 0
2nd,

 (x -t )
x
t e t  dt
n-2 <1>


<2> x x
= (x -t )n-2t e t  0+ (n -2) (x -t )n-3t e t  dt
<2>

0
n-2
(2+ )
 (x -t )
x x
t e t  dt
n-3 <2>
=- + (n -2)
(-1) 1! 0

the last (n-1)th,


x
(x -t )1t e t  dt
<n -1>

x 1 (n -1+ )

x
(x -t )0t e t  dt
<n -1>
=- + 1
(-1) (n -2)! 0
Substituting these for the former expression one by one, we obtain the following expression.

 (x -t )
x 1 n -1 (n -1)!
n-1  t
t e dt = - Σ (n -r + )x
r
0 (-1) r=1 r !(n -1-r)!

 (x -t ) t e 
x
+ (n -1)! 0  t <n -1>
dt
0
Thus,

- 37 -
 (n -r + )x r
 (x -t )
1 x 1 n -1
n-1  t
(n) 0
t e dt = -
(-1)
Σ
r=1
n -1Ｃ r
(n -1)!

 t e 
x
 t <n -1>
+ dt (w2)
0
Total
Substituting (w1) , (w2) for (w) ,

(n -r + )x r
  
1 x 1 n -1
(x -t )n-1t e t dt = (-1) - Σ Ｃr
(n) (-1)
n -1
- r=1 (n -1)!

  t e 
0 x
t e t 
<n -1>  t <n -1>
+ dt + dt
- 0
Here,
1 2 i
(-1) -
(-1)
= 2i sin 
= ()(1-) 
   t e 
0 x x <n>
t e t 
<n -1>
t e t 
<n -1>  t <n -1>
dt + dt = dt = x e x 
- 0 -

x  n -1 Ｃ r (n -r + , -x)x r
n -1
= Σ
(-x) r=0 (n -1)!
Substituting these for the above, we obtain (1.n+ ) .

Example 2nd order integral of ex x

Let us substitute  =1/2 , n =2 for (1.n- ), (1.n+ ) and calculate the value on the arbitrary two points
x =1.7 . Then, it is as follows.
Le ft: Rie mann-Liouville integral
 a:=1/2: n:=2:
 G := x-> 1/gamma(n)*int((x-t)^(n-1)*t^a*E^t, t=-infinity..x)
x
1 n1
x   (x  t) t E dt
a t
(n)


 float(G(-1.7)); float(G(1.7))
0.3457294964  i

2.255551158  2.835926162  i

Right: Higher int by incomple ate gamma

x <= 0
 Fm := x-> x^a/(-x)^a*(1/(n-1)!)*sum(binomial(n-1,r)
*igamma(n-r+a,-x)*x^r, r=0..n-1)
 
n1 
 
n  1  (n  r  a,  x)  xr
a
x   
x   1 
(  x) (n  1) 
a
r
r0

- 38 -
x>0
 Fp := x-> Fm(x)+2*I*sin(a*PI)/(n-1)!
*sum(binomial(n-1,r)*gamma(n-r+a)*x^r, r=0..n-1)
 
n1  
2  i  sin(  a) n  1  (n  r  a)  xr
x  Fm(x)   
(n  1)  r0
r

 float(Fp(1.7))
2.255551158  2.835926161  i

When  is a natural number, we can differentiate x completly. And the following formula holds.

Formula 16.5.1'
When m =0, 1, 2, ,
(1+ m) m- r
 r
x x m -n
 e x x mdx n = e x Σ x (1.n')
- - r=0 (1+ m - r)

Proof
Let f(x)= e x in the Theorem16.2.0 (0.2) . Then, since a1 = a2 == an = - ,
(1+m) m- r
   
x x m -n
 e xx m dx n = e xΣ x
- - r=0 r 1+m - r 
(1+m)
    dx
n -1 m -n + r m- s x
x
- e -Σ Σ (-)  r
r=0 s=0 s 1+m - s  - -
Since this 2nd line is 0 clearly, we obtain (1.n') immediately.

4
Example 5th order integral of ex x
Substituting m =4 , n =5 for (1.n'),

(5) 4- r
 r  
x x 4 -5 4 4+ r 4!
 e x x 4dx 5 = e xΣ x = e xΣ(-1)r 4 r
x -
- - r=0 5- r  r=0 4 (4- r)!
When the both sides are calculated by mathematical software, it is as follows.
Le ft: Rie mann-Liouville integral
 m:=4: n:=5:
 g := 1/gamma(n)*int((x-t)^(n-1)*t^m*E^t, t=-infinity..x)
 
e  x  20  x  180  x  840  x  1680
x 4 3 2

Right: Formula
 f := E^x*sum(binomial(-n,r)*gamma(1+m)/gamma(1+m-r)
*x^(m-r), r=0..m)
 
e  x  20  x  180  x  840  x  1680
x 4 3 2

Collateral higher integral of e x x

About the collateral higher integral whose zero is 0, the following formula holds from the Formula 16.2.0 (0.5) .

- 39 -
Formula 16.5.1"
The following expressions hold for  , n such that
  -1, -2, -3,  &  + n >0.
(1+ )
  
x x m -1 -n

 e x x dx n = e xΣ r
x +n+ + Rmn (1.n")
0 0 r=0 r  (1+  + n r
+ )
(1+ )
n -1Ｃ k

m n -1 x x
(-1)  m+k
Rmn = Σ  e x x + dx n
(n , m) k =0 m + k (1+ + m +k) 0 0
lim Rmn = 0
m 

3
Example Collateral the 2nd order integral of ex x
This function can be illustrated in a positive domain. Blue (Riemann-Liouville integral) hidden from red (Series)
cannot be seen. Although this turns into complex function in a negative domain, of course, both sides are
corresponding.
Le ft: Rie mann-Liouville integral
 a:=1/3: n:=2:
 g := x-> 1/gamma(n)*int((x-t)^(n-1)*t^a*E^t, t=0..x):

Right: Se rie s
 m:=20:
 f := x-> E^x*sum(binomial(-n,r)*gamma(1+a)/gamma(1+a+n+r)
*x^(a+n+r), r=0..m-1):

16.5.2 Higher Integral of e x log x

Formula 16.5.2


x et
When Ei(x) = dt is the Exponential Integral, the following expressions hold.
- t

 e log x dx = e log x - Ei(x)

x
x x
-

  e log x dx   
x0 x1

x x 0! 0
x 2 x
=e log x + x - + Ei(x)
- - 1! 0! 1!

- 40 -
    e log x dx    
x x x 0! 1 0! 1! 0
x 3
= e x log x + x + + x
- - - 2! 1! 2!

 
x0 x1 x2
- + + Ei(x)
0! 1! 2!

 
x x
     
0! 0! 1! 0! 1! 2!
 e x log x dx 4 = e x log x + 3! x 2+ 2! + 3! x 1+ 1! + 2! + 3! x 0
- -

 
0 1 2 3
x x x x
- + + + Ei(x)
0! 1! 2! 3!


  e log x dx log x + Σ Σ  - Ei(x)Σ

x x
x n x
n -2 n -2-r s! x r n -1 xr
=e (2.n)
- - r=0 s=0 (r + s +1)! r=0 r!

Proof


x et
When Ei(x) = dt , lineal higher primitive functions of e x log x are as follows.
- t
<1>
e x log x = e xlog x - Ei(x)
<2>
e x log x = e x(log x +1) -(x +1)Ei(x)
x 2+2x +2
 
<3> x +3
e log x
x
=e x
log x + - Ei(x)
2 2

 
<4> x 2+4x +11 x 3+3x 2+6x +6
e log x
x
=e x
log x + - Ei(x)
3! 6

And the zeros of these are all x =- . Therefore, 1st ~ 4th order integrals can be written as mentioned above.
And hereafter, by induction, we obtain (2.n) .

Example 3rd order integral of e xlog x

F n_ : x Log Absx     ExpIntegralEix 
n 2 n 2r n 1
s  xr xr
r  s  1  r
Plot F3, x, 5, 4
r0 s0 r0

-4 -2 2 4

-1

-2

- 41 -
Note
All polynomials obtained by applying Theorem 16.1.2 to e xlog x become the asymptotic expansions, and
they are hardly useful.

16.5.3 Higher Integrals of e x sinx , e x cosx

Formula 16.5.3
 n

n

   
x x
 x
e sin xdx = sin n
e xsin x - (3.0s)
- - 4 4
 n
 e cosxdx = sin 4  e cos x - 4 
x x n
x n x
(3.0c)
- -

Proof
About the higher derivative of e xsin x , the following formula is known.(See 共立 数学公式 p187).

 -n
n
   
(n)
e sin x
x
= sin e xsin x +
4 4
Replacing n with -n ,

 n
n
   
(-n )
e sin x
x
= sin e xsin x -
4 4
Since a differentiation operator (-n) is equal to an integration operator <n > ,
 n
n
   
<n>
e xsin x = sin e xsin x -
4 4
And since x = - is zero of this clearly, rewriting the left side, we obtain (3.0s).
The following formula holds also about e xcos x .

 -n
n
   
(n)
e cosx
x
= sin e xcos x +
4 4
From this, we obtain (3.0c) in a similar way.

Example
 
 e sin xdx = sin 4  e sin x - 4  =
x 2 x sin x - cosx
x x
e
- 2 2
1 x
= e (sin x - cosx)
2
 2

2

   
x x 1 x
x
e sin xdx = sin 2
e xsin x - =- e cosx
-- 4 4 2
 3 
 e cosxdx = sin 4  e cos x - 4  =
3

 
x x 2 x
x 3 x
e sin x -
- - 4 4
2 x sin x - cosx 1
= e = e x(sin x - cosx)
4 2 4
Higher Integral of e x sin x , e x cos x ends now. There is no necessity for Theorem 16.1.2.
However, daring use Theorem 16.1.2, we obtain an interesting result.

- 42 -
Formula 16.5.3'
(1)
r
 r 
-n
+R
x x m -1
 e xsin xdx n = e xΣ sin x + n
m (3.1)
- - r=0 2
(m + k)
  e sin x +
(-1)m n -1 n -1Ｃ k

x x

(n , m) Σ
Rmn = x
dx n (3.1r)
k =0 m + k - - 2
(2)

(n + r)
 r 
-n
+ R
x x m -1
 e xsin xdx n = e xΣ sin x - n
m (3.2)
- - r=0 2
(m + k)

(-1)m n -1 n -1Ｃ k
 
x x

(n , m) Σ
Rmn = e xsin x - dx n (3.2r)
k =0 m + k - - 2
(3)

r
 r 
-n
+R
x x m -1
 e xcosxdx n = e xΣ cos x + n
m (3.3)
- - r=0 2
(m + k)

(-1)m n -1 n -1Ｃ k
 
x x

(n , m) Σ
Rmn = e xcos x + dx n (3.3r)
k =0 m + k - - 2
(4)
(n + r)
 r 
-n
+ R
x x m -1
 e xcosxdx n = e xΣ cos x - n
m (3.4)
- - r=0 2
(m + k)
  e cosx -
(-1)m n -1 n -1Ｃ k

x x

(n , m) Σ
Rmn = x
dx n (3.4r)
k =0 m + k - - 2

Calculation
Let f(x)= e x , g(x)= sin x . Then

r
 
()
(sin x) r = sin x +
2
Substituting this for (2.3) in Theorem 16.1.2 , we obtain (3.1), (3.1r).

Next, let f(x)= sin x , g(x)= e x . Then

(n +r)
(sin x)<n + r> = sin x -  2 
Substituting this for (2.3) in Theorem 16.1.2, we obtain (3.2), (3.2r).

Also about e xcos x , in a similar way, we obtain (3.3) ~ (3.4r).

Example 2nd order integral of e xsinx

Let n =2, m =10 in (3.1), (3.1r) . And calculating the function value on the arbitrary point x =3
by mathematical software, it is as follows.

- 43 -
Se rie s
 m:=100:
 f := E^x*sum(binomial(-n,r)*sin(x+r*PI/2), r=0..m-1):
 float(subs(f, x=3))
 1135.950099
Re mainde r
 r0 := -1/(m+0)*binomial(n-1,0)*E^x/2*cos(x+(m+0)*PI/2):
 r1 := -1/(m+1)*binomial(n-1,1)*E^x/2*cos(x+(m+1)*PI/2):
 R := (-1)^m/beta(n,m)*(r0+r1):
 float(subs(R, x=3))
1145.892364

Se rie s + Re mainde r
 float(subs(f+R, x=3))
9.942265422

Complete Automorphism
As understood from this example, although (3.1), (3.1r) hold as an equation, it is not helpful at all. The same
is said for (3.2), (3.2r). However, if these are used in combination, the complete automorphism of the higher

integral of e xsin x is drawn as follows. This is the same for e xcos x

Formula 16.5.3"
n n

e xsin x -  cos

x x 4 4
 e xsin xdx n = (3.5)
- - n -1 n -1Ｃ k (1+ k)
1+ nΣ cos
k=0 1+ k 2
n n
e xcos x -  
sin

x x 4 4
 e xcosx dx n = (3.6)
- - n -1 n -1Ｃ k (1+ k)
nΣ sin
k =0 1+ k 2
n n
e x
cos 
x - 
cos

x x 4 4
 e xcosxdx n = (3.7)
- - n -1 n -1Ｃ k (1+ k)
1+ nΣ cos
k =0 1+ k 2
 
 
n n
e xsin x - sin

x x 4 4
 e xsin x dx n = (3.8)
- - n -1 n -1Ｃ k (1+ k)
nΣ sin
k =0 1+ k 2

Calculation
Let m =1 in (3.1) ~ (3.2r) . Since (n , 1) =1/n ,
(1+ k)
 1+ k  
n -1 n -1Ｃ k
e sin x + 
x x x x
 e xsin xdx n = e xsin x - nΣ  x
dx n
- - k=0 - - 2

- 44 -
 n (1+k)
 
n -1 n -1Ｃ k

   
x x x x
 e sin xdx = e xsin x -
x n
- nΣ  e xsin x - dx n
- - 2 k =0 1+ k - - 2
n -1 Ｃk (1+ k)
Here, let Ck = , Bk = . Then
1+ k 2

   e sinx +B dx
x x n -1 x x
 e xsin xdx n = e xsin x - nΣCk  x
k
n
- - k=0 - -
n
 e sin xdx = e sin x -
2 
- nΣC  e sinx -B dx
x x n -1 x x
x n x x n
k k
- - k =0 - -
Using the sum and difference formulas,

   e sin x cosB + cosx sinB dx

x x n -1 x x
 x
e sin xdx = e sin x - nΣCk
n x
 x
k k
n
(a)
- - k=0 - -

 n
  e sin x cosB - cos x sinB dx
x x x x

 
n -1
 e xsin x dx n = e xsin x - -n ΣCk  x
k k
n
(b)
- - 2 k =0 - -
Adding (a) to (b),
n
      e sin x cosB dx
x x n -1 x x
2  e xsin xdx n = e x sin x + sin x - -2nΣCk  x
k
n
- - 2 k=0 - -
n n
  - 2nΣC cosB  e sin x dx
n -1 x x
= 2e xsin x - cos k k
x n
4 4 k =0 - -
From this,
n n

e xsin x -  cos

x x 4 4
 e xsin xdx n = n -1
- - 1 + nΣCk cosBk
k=0
Returning Ck , B k before,
n n

e xsin x -  cos

x x 4 4
 e xsin xdx n = (3.5)
- - n -1 n -1Ｃ k (1+ k)
1+ nΣ cos
k =0 1+ k 2
Next, subtract (b) from (a) . Then
n
     e cosx sinB dx
n -1 x x
0 = e x sin xdx n - sin x - -2nΣCk  x
k
n
2 k =0 - -
n n
  -2nΣC sinB  e cosx dx
n -1 x x
x n
= 2cos x - sin k k
4 4 k =0 - -
From this,
n n
e xcos x -   sin
 e cosx dx
x x 4 4
x n
= n -1
- - nΣCk sinBk
k =0
Returning Ck , B k before,
n n
e xcos x -   sin

x x 4 4
 e xcosx dx n = (3.6)
- - n -1 n -1Ｃ k (1+ k)
nΣ sin
k =0 1+ k 2

- 45 -
 Also about e xcosx , in a similar way, we obtain (3.7), (3.8).

Example When n=2

   
e xsin x -   cos e xsin x -  cos

x x 2 2 2 2
e xsin xdx 2 = =
1Ｃ k (1+ k)  1 2
 
- - 1 1
1+2Σ cos 1+2 cos + cos
k =0 k +1 2 1 2 2 2
  
=
 2 2 =
e xsin x - cos 
e xsin x -
2  =-
e xcosx
 1
1+2  cos - 
1 2 2
1 2 2
   
e cos x -  sin
x
e xcos x -  sin

x x 2 2 2 2
e xcosx dx 2 = =
Ｃk (1+ k)  1 2
 
- - 1 1 1
2Σ sin 2 sin + sin
k =0 1+ k 2 1 2 2 2
  
=
e xcos x -  2  sin
2
=

e xcos x -
2  =
e xsin x
 2 2
2sin
2

16.5.4 Trigonometric Polynomial and Binomial Polynomial

(1) Trigonometric Polynomial

Although Formla 16.5.3" is helpful, it is fearfully complicated and cannot be compared to Formula 16.5.3.
However, if we dare compare this with Formula16.5.3 , the following Trigonometric Polynomial is obtained.

Formula 16.5.4
Ｃ k-1 k  -n
n
 
n n -1 1
Σ
k =1 k
sin
2
=
n
sin
4
sin
4
(4.1)

n -1Ｃ k-1 k  -n
n
 
n 1 1
Σ
k =1 k
cos
2
=
n
sin
4
cos
4
-
n
(4.2)

Proof
 n
 e cosxdx = sin 4  e cos x - 4 
x x n
x n x
(3.0c)
- -
n n
e cosx -
4 
x
sin

x x 4
 e cosx dx = x n
(3.6)
- - Ｃ (1+ k) n -1 n -1 k
nΣ sin
k =0 1+ k 2
From these,
 n
n n -1 n -1Ｃ k (1+ k)
 sin
4  = sin
4
/ nΣ 
k =0 1+ k
sin
2 
Replacing k with k -1 , we obtain (4.1).

- 46 -
 Next,

 n
 e sin xdx = sin 4  e sin x - 4 
x x n
x n x
(3.0s)
- -
n n
e sin x -  cos x

 e sin xdx =

x x 4 4
x n
(3.5)
- - Ｃ (1+ k) n -1 n -1 k
1+ nΣ cos
k =0 1+ k 2
From these,

 n
n n -1 n -1Ｃ k k
 sin
4  = cos
4 
/ 1+ nΣ
k =0 1+ k
cos
2 
Replacing k with k -1 , we obtain (4.2).

(2) Binomial Polynomial

However, Formula 16.5.4 is tedious for a natural number n and is not interesting. Then removing
k k
sin , cos from the formula, we obtain the following interesting polynomials.
2 2

Formula 16.5.4'
When  , denote a floor function and a ceiling function respectivly,

n /2  (-1)k  -n-1
(n +1)
 
1
Σ
k =0 2k +1 n
Ｃ 2k =
n +1
sin
4
sin
4
(4.3)

 (n +1)
 
n /2  (-1)k -n-1


1
Σ
k =1 2k
nＣ 2k-1 =
n +1
sin
4
cos
4
-1 (4.4)

Proof
Replacing n,k with n +1, k +1 in (4.1) ,

Ｃk (k +1)  -n-1
(n +1)
 
n n 1
Σ
k =0 k +1
sin
2
=
n +1
sin
4
sin
4
Since the even-numbered terms of the left side are all 0,
n Ｃk (k +1) nＣ 0 nＣ 2 nＣ 4 (-1)k
-+  + 2k  n
n
Σ
k =0 k +1
sin
2
=
1
-
3
+
5 2k +1
nＣ 2k

n /2 (-1)k
=Σ nＣ 2k
k =0 2k +1

(-1)k
n /2   -n-1
(n +1)
 
1
 Σ nＣ 2k = sin sin (4.3)
k =0 2k +1 n +1 4 4
Replacing n , k with n +1 , k +1 in (4.2) ,

(k +1)  (n +1)
 
-n-1
Ｃk

n n 1
Σ
k =0 k +1
cos
2
=
n +1
sin
4
cos
4
-1
Since the odd-numbered terms of the left side are all 0,
n Ｃk (k +1) nＣ 1 nＣ 3 nＣ 5 (-1)k
+-  +
n
Σ
k =0 k +1
cos
2
=-
2
+
4
-
6 2k
nＣ 2k-1 2k -1 n

- 47 -
 n /2  (-1)k
=Σ nＣ 2k-1
k=1 2k
 (n +1)
 
n /2  (-1)k -n-1


1
 Σ nＣ 2k-1 = sin cos -1 (4.4)
k =1 2k n +1 4 4

Example
Ｃ0 Ｃ2 Ｃ4 Ｃ6 Ｃ8  -9-1
(9+1)
 
9 9 9 9 9 1 16
- + - + = sin sin =
1 3 5 7 9 9+1 4 4 5
Ｃ0 Ｃ2 Ｃ4 Ｃ6 Ｃ8  -8-1
(8+1)
 
8 8 8 8 8 1 16
- + - + = sin sin =
1 3 5 7 9 8+1 4 4 9
 (7+1)
 4  
Ｃ1
7 Ｃ3
7 Ｃ5
7 Ｃ7
7 1 -7-1
15
- + - + = sin cos -1 =
2 4 6 8 7+1 4 8
 (8+1)
8+1  
- 1 =
Ｃ1 Ｃ3 Ｃ5 Ｃ7 -8-1

4
8 8 8 8 1 5
- + - + = sin cos
2 4 6 8 4 3

- 48 -
16.6 Higher Integral of f (x) / e^x

e x
-x
16.6.1 Higher Integral of

Formula 16.6.1

t

When (a , x) = a-1 -t
e dt is the incomplete gamma function,
x


x 1
e -x x  dx (1+ , x) =-
 0!


x x 1
e -x x  dx 2 = {(2+ , x)- (1+ , x)x }
  1!


x x x 1
e x dx = - {(3+ , x)- 2(2+ , x)x + (1+ , x)x 2}
-x 3
   2!

  e
x x 1
x dx = {(4+ , x) -3(3+ , x)x +3(2+ , x)x -(1+ , x)x }
-x 4 2 3
  3!



x x
(-1)n n -1
(-1)rn -1Ｃ r (n -r + , x)x r
(n -1)! Σ
-x n
e x dx = (1.n)
  r=0

Proof
If the integration is obediently repeated, they are led naturally. And these zeros are all x = . Then these
are lineal higher integrals.

-x
Example 2nd order integral of e x
Let substitute  =1/2 , n =2 for (1.n), and calculate the value on thearbitrary point x = 2.3 .
Then, it is as follows. Both sides are corresponding exactry.

Left: Riemann-Liouville inte gral

 a := 1/2: n:=2:
 g := x-> 1/gamma(n)*int((x-t)^(n-1)*t^a/E^t,t=infinity..x):
 float(g(-2.3))
3.367662317  6.637309509  i

Right: Inte gral by incomple te gamma function

 f := x-> (-1)^n/(n-1)! *
sum((-1)^r*binomial(n-1,r)*igamma(n-r+a,x)*x^r, r=0..n-1)
 
n1
  
(  1)
n
x    (  1)  n  1  (n  r  a, x)  x
r r
(n  1)  r0
r

 float(f(-2.3))
3.367662317  6.637309509  i


When  is a natural number, we can differentiate x completly. And the following formula holds.

- 49 -
Formula 16.6.1'
When m =0, 1, 2, ,
(1+ m) m- r

(-1)n
r
x x m -n
 -x m
e x dx = n
Σ (-1) r
x (1.n')
  ex r=0 (1+ m - r)

-x 4
Example 5th order integral of e x
Substituting m =4 , n =5 for (1.n') ,

(1+4) 4- r

(-1)5
r
x x 4 -5
 -x 4
e x dx = 5
Σ(-1)
r
x
  ex r=0 (1+4- r)
When the both sides are calculated by mathematical software, it is as follows. Naturally, both sides are
corresponding.
Left: Riemann-Liouville inte gral
 m:=4: n:=5:
 g := 1/gamma(n)*int((x-t)^(n-1)*t^m/E^t,t=infinity..x):
 expand(g)
840  x  
180  x  
20  x  
2 3 4
  x  
1680
x x x x x
e e e e e

Right: Formula
 f := (-1)^n/E^x*sum((-1)^r*binomial(-n,r)*gamma(1+m)
/gamma(1+m-r)*x^(m-r), r=0..m):
 expand(f)
180  x  
20  x   840  x
2 3 4
  x  
1680  
x x x x x
e e e e e

e x
-x
Collateral higher integral of
The following formula holds about the collateral higher integral whose zero is 0.

Formula 16.6.1"
The following expressions hold for  , n such that   -1, -2, -3,  &  + n >0.
(1+ )
  
x x 1 m -1 -n 
 e -x x dx n = -r r
Σ (-1)  
x +n+ + Rmn
0 0 e x r=0 r ( 1+ + +
n r)
+ m+k
(1+ )
n -1Ｃ k

1 n -1 x x x
Rmn = Σ
(n , m) k =0
(-1)-k
m + k (1+ + m +k) x dx n
0 0 e
lim R n
m  m
=0

-x
Example Collateral the 2nd order integral of e x
This function can be illustrated in a positive domain. Blue ( Riemann-Liouville integral ) hidden from
red (Series) cannot be seen. Although this turns into complex function in a negative domain, of course,
both sides are corresponding.

- 50 -
Left: Riemann-Liouville inte gral
 a := 1/2: n:=2:
 g := x->1/gamma(n)*int((x-t)^(n-1)*t^a/E^t,t=0..x):

Right: Serie s
 m:=20:
 f := x->sum((-1)^-r*binomial(-n,r)*gamma(1+a)/gamma(1+a+n+r)
*x^(a+n+r)/E^x, r=0..m-1):

-x
16.6.2 Higher Integral of e log x
Formula 16.6.2


x et
When Ei(x) = dt is the Exponential Integral, the following expressions hold.
- t

 e log x dx = -e log x + Ei(-x)

x
-x -x


 x  -
0! 1! 
0 1


x x 0! x x
-x 2 -x 0
e log x dx = e log x + - Ei(-x)
  1!

   1! + 2!  x 
x x x 0! 0! 1!
-x 3 -x 1 0
e log x dx = -e log x - x +
   2!

+
0! 1! 2! 
0 1 2
x x x
- + Ei(-x)

 e log xdx = e log x + 0!

x x
x -
2! 3!   1! + 2! + 3!  x 
-x 4 0! 1! -x 0! 1! 2!
2 1 0
+ x +
  3!

-
0! 1! 2! 3! 
0 1 2 3
x x x x
- + - Ei(-x)


  
x x n -2 n -2-r s!
 e -x log x dx n = (-1)n e -x log x + Σ (-x)r Σ
  r=0 s=0 (r + s +1)!

- 51 -
 n -1 (-x)r
- (-1) Ei(-x)Σ
n
(2.n)
r=0 r!
Proof
Lineal higher primitive functions of e -x log x are as follows.
<1>
e -x log x = -e -x log x + Ei(-x)
<2>
e log x
-x
= e -x(log x +1) + (x -1)Ei(-x)
x 2-2x +2
 
<3> x 3
e -x log x = -e -x
log x - + + Ei(-x)
2 2 2!

 
<4> x 2-4x 11 x 3-3x 2+6x -6
e log x
-x
= e -x log x + + + Ei(-x)
3! 6 3!


And the zeros of these are all x = . Therefore, 1st ~ 4th order integrals can be written as mentioned above.
And hereafter, by induction, we obtain (2.n).

-x
Example 3rd order integral of e log x

Note
All polynomials obtained by applying Theorem 16.1.2 to e -xlog x become the asymptotic expansions,
and they are hardly useful.

-x -x
16.6.3 Higher Integrals of e sinx , e cosx

Formula 16.6.3
 n

n

   
x x
 e sin xdx = (-1)n sin
-x n
e -xsin x + (3.0s)
  4 4

- 52 -
  n

n

   
x x
 -x
e cosxdx = (-1) n n
sin e -xcos x + (3.0c)
  4 4

Proof
About the higher derivative of e xsin x , the following formula is known.(See 共立 数学公式 p187).

 -n
n
   
(n)
e sin x
x
= sin e xsin x +
4 4
Replacing x with -x ,

 -n
n
   
(n)
(-1) e sin(-x)
n -x
= sin e -xsin -x +
4 4
Using sin(-x) = -sin x
 -n
n
   
(n)
e sin x
-x
= (-1)n sin e -xsin x -
4 4
Replacing n with -n ,

 n
n
   
(-n )
e sin x
-x
= (-1)n sin e -xsin x +
4 4
Replacing (-n ) with <n > ,
 n
n
   
<n>
e sin x
-x
= (-1)n sin e -xsin x +
4 4
And since x =  is zero of this clearly, rewriting the left side, we obtain (3.0s).
The following formula holds also about e xcos x .

 -n
n
   
(n)
e cosx
x
= sin e xcos x +
4 4
From this, we obtain (3.0c) in a similar way.

Example
 
e    
x 2 -x sin x + cosx
-x
sin xdx = (-1)1 sin e -xsin x + =- e
 4 4 2 2
1 -x
=- e (sin x + cosx)
2
 2

2

   
x x 1 -x
-x 2 2
e sin xdx = (-1) sin e -xsin x + = e cosx
  4 4 2
 3 
 e cosxdx
3

     
x x 2 -x
x 3 3
= (-1) sin e -xcos x + = e sin x +
  4 4 4 4
2 x cosx + sin x 1
= e = e -x(sin x + cosx)
4 2 4

- 53 -
16.7 Series Expansion of Higher Integral
By adopting easy function such as 1 or ex as one side of the functions f and g in Thorem 16.1.2, various
function series is obtained.

16.7.1 Series Expansion of Higher Integral with fixed lower limit

First, adopting 1 as the function f in Theorem 16.1.2, we obtain the following theorem.

Theorem 16.7.1
r =0, 1,  ,m +n
(r)
Let m,n be natural numbers, f be the r th derivative of f, a be an arbitrary

constant on the domain of f ,  (x , y) be the beta function. Then the following expressions hold.

 f(x)dx
(x -a)n+ r (r)
r
x x m -1 -n
n
=Σ f (x) + Rmn (1.1)
a a r=0 (n + r)!


(-1)m n -1 n -1Ｃ k x t t (t -a)m+ k (m+ k )
(n , m) Σ
Rmn = f (t)dt n (1.1r)
k=0 m + k a a a (m + k)!

  f(x)dx
x x m -1 (x -a )
n+ r
 n
=Σ (r)
f (a) + Rmn (1.2)
a a r=0 (n + r)!
n -1 r -1 r -1 (x -a)m + n+s (m+s)
t s  m +n -1- r + Ｃ
t m -1  m +n+sＣ r
n-r+s
Rmn = ΣΣ(-1) Σ Ｃ f (a)
r =1 s =0 t=s (m + n + s)!


(-1)m n -1 n -1Ｃ k x t t (t -x)m+ k (m+ k )
(n , m) Σ
+ f (t)dt n (1.2r)
k=0 m + k a a a (m + k)!
Where, there shall be no term of ΣΣ at the time of n <2 .

Proof
Exchanging f and g in Theorem 16.1.2 ,

 r
x x m -1 -n
 g <0>f(0)dx n = Σ g <n + r> (r)
f
a a r=0

(x -a)r
 s g
-n + r
n -1 m -1
<n -r+ s> (s)
- ΣΣ a fa
r=0 s=0 r!
n -1 r -1 r -1 (x -a)r
t s  m + n -1- r + tＣ m-1 g a
m + n -r+s m +s
+ (-1)m ΣΣΣ Ｃ fa
r =1 s =0 t=s r!

 g
m
(-1) n -1 n -1Ｃ k x x
 <m+ k > (m+ k )
(n , m) Σ
+ f dx n
k=0 m + k a a
r
(x -c)
( a  c) for this,
<r>
Substituting g =
r!

 1f
(x -c)n+ r (r)
r
x x m -1 -n
(0)
dx = Σ
n
f
a a r=0 (n + r)!
n-r+ s r

 s  (n -r + s)! r ! f
-n + r (a -c)
n -1 m -1 (x -a) (s)
- ΣΣ a
r=0 s=0

- 54 -
 n -1 r -1 r -1 (a -c)m + n-r+s (x -a)r m +s
Σ ΣΣ Ｃ t s  m + n -1- r + tＣ m-1
m
+ (-1) fa
r =1 s =0 t=s (m + n - r + s)! r !


(-1)m n -1 n -1Ｃ k x x (x -c)m+ k (m+ k ) n
(n , m) Σ
+ f dx
k=0 m + k a a (m + k)!
Change this to the definite higher integral from a to b ( c) , and rewrite the variable in the remainder term
from x to t . Then,

  f
(b -c)n+ r (r)
r
b x x m -1 -n
 (0)
dx = Σn
f
a a a r=0 (n + r)! b
n-r+ s r

 s  (n -r + s)! r !
-n + r (a -c)
n -1 m -1 (b -a) (s)
- ΣΣ fa
r=0 s=0

n -1 r -1 r -1 (a -c)m + n-r+s (b -a)r m +s

Σ ΣΣ Ｃ t s  m + n -1- r + tＣ m-1
m
+ (-1) fa
r =1 s =0 t=s (m + n - r + s)! r!


(-1)m n -1 n -1Ｃ k b t t (t -c)m+ k (m+ k ) n
(n , m) Σ
+ f dt (1.0)
k=0 m + k a a a (m + k)!
Here, put c=a , then

 f
(b -a)n+ r (r)
r
b x x m -1 -n
(0)
dx = Σn
f
a a a r=0 (n + r)! b
n-r+ s
(b -a)r (s)
 s  (n -r + s)!
-n + r (a -a)
n -1 m -1
- ΣΣ fa
r=0 s=0 r!
n -1 r -1 r -1 (a -a)m + n-r+s (b -a)r m +s
Σ ΣΣ Ｃ t s  m + n -1- r + tＣ m-1
m
+ (-1) fa
r =1 s =0 t=s (m + n - r + s)! r!

 
(-1)m n -1 n -1Ｃ k b t t (t -a)m+ k (m+ k ) n

(n , m) Σ
+ f dt
k=0 m + k a a a (m + k)!
(b -a)n+ r (r)
 
m -1 -n
=Σ f
r=0 r (n + r)! b


(-1)m n -1 n -1Ｃ k b t t (t -a)m+ k (m+ k ) n
(n , m) Σ
+ f dt
k=0 m + k a a a (m + k)!
Returning b to x , we obtain (1.1), (1.1r) .
Next, let c=b in (1.0) . Then,


(b -b)n+ r (r)
 
b x x m -1 -n
f dx = Σ
(0) n
f
a a a r=0 r (n + r)! b
(a -b)n-r+ s (b -a)r (s)
 
n -1 m -1 -n + r
- ΣΣ fa
r=0 s=0 s (n - r + s)! r!
n -1 r -1 r -1 (a -b)m + n-r+s (b -a)r m +s
+ (-1) ΣΣΣ Ｃ t s  m + n -1- r + tＣ m-1
m
fa
r =1 s =0 t=s (m + n - r + s)! r!

 
(-1)m n -1 n -1Ｃ k b t t (t -b)m+ k (m+ k ) n

(n , m) Σ
+ f dt
k=0 m + k a a a (m + k)!

- 55 -
 (b -a)n+ s
 
n -1 m -1
n-r+ s
-n + r (s)
= -ΣΣ (-1) fa
r=0 s=0 s r!(n - r + s)!
n -1 r -1 r -1 (b -a)m + n+s
t s  m + n -1- r + tＣ m-1 f a +s
n-r+s m
+ ΣΣΣ(-1) Ｃ
r =1 s =0 t=s r!(m + n - r + s)!

 
m n -1 m+ k
n -1Ｃ k
(-1) b t t (t -b )
+ Σ  m+ k
f ( )dt n
(n , m) k=0 m + k a a a (m + k)!
Here,
n+sＣ r
 
-n + r 1
(-1) s = n-r -1+sＣ s , =
s r!(n - r + s)! (n + s)!
Substituting these for the above,


b x x m -1 n -1 (b -a)n+ s (s)

f dx = -Σ Σ(-1) n-1-r +sＣ s  n+sＣ r 
(0) n n-r
f
a a a s=0 r=0 (n + s)! a
n -1 r -1 r -1 (b -a)m + n+s m +s
t s  m + n -1- r + tＣ m-1  m +n+sＣ r
n-r+s
+ ΣΣ(-1) Σ Ｃ f
r =1 s =0 t=s (m + n + s)! a

 
(-1)m n -1 n -1Ｃ k b t t (t -b)m+ k (m+ k ) n

(n , m) Σ
+ f dt
k=0 m + k a a a (m + k)!
Here, the following equation holds for a natural number n and a non-negative integer s .
n -1
Σ (-1) n-r
n-1 -r +s Ｃ s  n+sＣ r = -1
r=0
Because,
n -1 n -1 (n -1-r + s)! (n + s)!
Σ(-1) n-1-r +sＣ s  n+sＣ r = Σ(-1)
n-r n-r
r=0 r=0 (n -1-r)! s ! (n -r + s)! r !
(n + s)! n -1 (n -1)! 1
= Σ (-1)n-r
(n -1)! s ! r=0 (n -1- r)! r ! n -r + s
(n + s)! n -1 n-1Ｃ n-1- r
=- Σ (-1)n-1-r
(n -1)! s ! r=0 s +1+ n -1-r
r
(n + s)! n -1 (-1)
(n -1)! s ! Σ n-1Ｃ r
=-
r=0 s +1+ r
According to 岩波数学公式Ⅱ p12,
m !(a)
r
m (-1)
Σ mＣ r = a  0, -1, -2, 
r=0 a+ r (m +a +1)
Then,
n -1 (-1) r (n -1)! (s +1) (n -1)! s !
Σ
r=0 s +1+ r
n-1 Ｃr =
(n -1+ s +1+1)
=
(n + s)!
Therefore

 f
b x x m -1 (b -a)n+ s (s)
(0)
dx = Σ
n
f
a a a s=0 (n + s)! a
n -1 r -1 r -1 (b -a)m + n+s m +s
t s  m + n -1- r + Ｃ
t m-1  m +n+sＣ r
n-r+s
+ ΣΣ(-1) Σ Ｃ f
r =1 s =0 t=s (m + n + s)! a

 
(-1)m n -1 n -1Ｃ k b t t (t -b)m+ k (m+ k ) n

(n , m) Σ
+ f dt
k =0 m + k a a a (m + k)!

- 56 -
ΣΣ of the 2nd line cannot be simplified any more. Moreover, since this is a disturbance term, we combine this
with the 3rd line and make it a remainder. Needless to say, when this 2nd line doesn't exist when n is smaller
than 2. Thus, rewriting s to r in the 1st line and returning b to x , we obtain (1.2), (1.2r) .

Series Expansion of Riemann-Liouville Integral

Higher Integral with fixed lower limit is Riemann-Liouville Integral itself. Therefore, the above thorem is also
a series expansion of Riemann-Liouville Integral. If the above thorem is rewritten by Riemann-Liouville Integral,
it is as follows.

Thorem 16.7.1'
n and an arbitrary number a on the domain of f .
The following formulas hold for a natural number
(x -a)n+ r (r)
  
1 x m -1 -n
n-1
(x -t) f(t)dt = Σ f (x) + Rmn (1.1')
(n) a r=0 r (n + r)!
x (x -t)n-1(t -a )m+ k

(-1)m n -1 n -1Ｃ k
m+ k
n
Rm = Σ
(n , m)(n) k =0 m k a
+ ( m k+ ) !
f ( )(t)dt n (1.1'r)

(x -a)n+ r (r)
 (x -t)
1 x m -1
n-1
f(t)dt = Σ f (a) + Rmn (1.2')
(n) a r=0 (n + r)!
n -1 r -1 r -1 (x -a)m + n+s (m+s)
t s  m +n -1- r + Ｃ
t m -1  m +n+sＣ r
n-r+s
Rmn = ΣΣ(-1) Σ Ｃ f (a)
r =1 s =0 t=s (m + n + s)!
(-1)m+ n x (t -x)m+n+ k

n -1 n -1Ｃ k
m+ k
- Σ
(n , m)(n) k =0 m + k a (m + k)!
f ( )(t)dt n (1.2'r)

Where, there shall be no term of ΣΣ at the time of n <2 .

Example 1 Series expansion of the 2nd order collateral integral of ex

Let f(x)= e x , then

r =1, 2, 3, 
r () (r)
f(x)= e x , f (x)= e x , f (0)= 1
Substituting these for Theorem 16.7.1 ,

 e dx
(x -a)n+ r
r
x x m -1 -n
a a
x n
=e x
Σ
r=0 (n + r)!
+ Rmn


(-1)m n -1 n -1Ｃ k x t t (t -a)m+ k t n
(n , m) Σ
Rmn = e dt
k=0 m + k a a a (m + k)!


x x m -1 (x -a)n+ r a
 e dx = Σ
x n
e + Rmn
a a r=0 +
(n r) !
n -1 r -1 r -1 (x -a)m + n+s a
t s  m + n -1- r + Ｃ
t m-1  m +n+sＣ r
n-r+s
Rmn = ΣΣ(-1) Σ Ｃ e
r =1 s =0 t=s (m + n + s)!

 
(-1)m n -1 n -1Ｃ k x t t (t -x)m+ k t n

(n , m) Σ
+ e dt
k=0 m + k a a a (m + k)!
Although the zero of the lineal higher integral of e x is a = - , these expressons can not have this zero.
That is, by the above formulas, the lineal higher integral of e x can not be expanded to the series. Then, we

- 57 -
put a =0 . and we expand the collateral higher integral of e x to the series. When n =2 ,

 e dx
(x -0)2+ r
r
x x m -1 -2
x x
Σ
2
=e + Rm2
0 0 r=0 (2+ r)!
x t(t -0)m+ k
 (m + k)! e dt
(-1)m 2-1 2-1Ｃ k
(2, m) Σ
t
Rm2 = 2
k =0 m + k 0 0


x x 2+ r
m -1 (x -0)
e xdx 2 = Σ e 0 + Rm2
0 0 r=0 (2+ r)!
(x -0)m +2 0
Rm2 = - mＣ m-1 m +2Ｃ 1 e
(m +2)!
x t(t -x)m+ k
 (m + k)! e dt
(-1)m 2-1 2-1Ｃ k
(2, m) Σ
t 2
+
k=0 m + k 0 0

Example 2 Series expansion of the 3rd order integral of log x

Let f(x)= log x , then
(r -1)!
r =1, 2, 3, 
(r)
f(0) = log x , f = (-1)r-1
xr
Substituting these for Theorem 16.7.1 ,


(x -a)n -n (x -a)n+ r (r -1)!
 
x x m -1
 log x dx = log x -Σ (-1)
n r
r
+ Rmn
a a n! r=1 r (n + r)! x


n+ r
x x (x -a)n m -1
r (x -a ) (r -1)!
 log x dx =
n
log a - Σ (-1) r
+ Rmn
a a n ! r=1 +
(n r) ! a
If a =0 then the first formula is lineal higher integral, else it is collateral higher integral. On the oter hand,
the second formula can not be lineal higher integral.
By the first formula, we calculate the lineal 3rd order integral and illustrate it as follows. Although the
convergence of this series was slow, as a result of taking m very greatly, both sids overlapped somehow.
Left: lineal highe r integral
 g := n-> x^n/n!*(ln(x) - sum(1/k,k=1..n)):
Right: Series
 m:=1000: a:=0:
 f := n-> (x-a)^n/n!*ln(x)-sum((-1)^r*binomial(-n,r)
*((x-a)^(n+r)/(n+r)!)
*((r-1)!/x^r),r=1..m-1):

- 58 -
16.7.2 Series Expansion of a Function
Sliding the index of the integration operator in Theorem 16.7.1, we obtain the series expansion of a function.
Of course in these, Taylor expansion is also included.

Theorem 16.7.2
r =0, 1,  ,m +n
(r)
Let m,n are natural numbers, f be the r th derivative of f , a be the arbitrary

constant on the domain of f ,  (x , y) be the beta function. Then the following expressions hold.
(x -a)r (r) (x -a)n+ r (n + r)
 
n -1 m -1 -n
f(x) = Σ f (a) + Σ f (x) + Rmn (2.1)
r=0 r! r=0 r (n + r)!


(-1)m n -1 n -1Ｃ k x t t (t -a)m+ k (n + m+ k )
(n , m) Σ
Rmn = f (t)dt n (2.1r)
k =0 m + k a a a (m + k)!
m + n -1 (x -a)r (r)
f(x) = Σ
r=0 r!
f (a) + Rmn (2.2)

n -1 r -1 r -1 (x -a)m + n+s (m+ n + s)

t s  m +n-1 -r + Ｃ
t m -1  m +n+sＣ r
n-r+s
Rmn = ΣΣ(-1) Σ Ｃ f (a)
r =1 s =0 t=s (m + n + s)!


(-1)m n -1 n -1Ｃ k x t t (t -x)m+ k (n + m+ k )
(n , m) Σ
+ f (t)dt n (2.2r)
k=0 m + k a a a (m + k)!
Where, there shall be no term of ΣΣ at the time of n <2 .

Proof
(0)
Adding n to the index of the differentiation operator of the function f in Theorem16.7.1 ,


n+ r

r
x x m -1 -n (x -a)
 f
(n)
(x)dx n = Σ n r
f ( + )(x) + Rmn
a a r=0 (n + r)!


(-1)m n -1 n -1Ｃ k x t t (t -a)m+ k (n + m+ k )
(n , m) Σ
Rmn = f (t)dt n
k=0 m + k a a a (m + k)!


x x n+ r
m -1 (x -a)
 (n)
f (x)dx n = Σ n r
f ( + )(a) + Rmn
a a r=0 (n + r)!
n -1 r -1 r -1 (x -a)m + n+s (n + m+s)
t s  m +n -1- r + Ｃ
t m -1  m +n+sＣ r
n-r+s
Rmn = ΣΣ(-1) Σ Ｃ f (a)
r =1 s =0 t=s (m + n + s)!


(-1)m n -1 n -1Ｃ k x t t (t -x)m+ k (n + m+ k )
(n , m) Σ
+ f (t)dt n
k=0 m + k a a a (m + k)!
Here,

 f   f 
x x x x (x -a)0 (n -1)
dx = 
(n) n (n -1) n-1
- fa dx
a a a a 0!

= f 
0
x
(x -a) x
(n -2) (n -2) (x -a)1 (n -1) n-2
- fa - fa dx
a 0! a 1!

= f 
0
x
(x -a) x
(n -3) (n -3) (x -a)1 (n -2) (x -a)2 (n -1) n-3
- fa - fa - fa dx
a 0! a 1! 2!

- 59 -
 
n -1 (x -a)r (r)
=f (0)
-Σ fa
r=0 r!
And,
n+ r
n -1 (x -a)r (r) m -1 (x -a )
(n + r)
m + n -1 (x -a )r
(r)
Σ
r=0 r !
f a ( ) + Σ
r=0 ( +
n r )!
f (a ) = Σr=0 r !
f (a)
Substituting these for the above, we obtain the desired expressions.

Remark
When n =1 in (2.2), (2.2r) ,


m (x -a)r (r) x(t -x)m
m
f(x) = Σ f (a) + (-1)m f (1+ )(t)dt
r=0 r! a m!
Needless to say, this is the usual Taylor expansion. However, in fact, (2.2), (2.2r) is alredy the usual Taylor
expansion. Because, from Theorem 4.1.3 (1.2), this fearful remainder must be as follows.

 
x x 1 x
Rmn =  f ( )(x)dx
m+ n m+ n
= (x -t)m+ n-1 f(t)dt
a a  ( +
m n a )
(2.2), (2.2r) is a trivial expression.

After all, the clarification from this theorem is the following two.
(1) A series of functions of a function f(x) about a point a exists innumerably.
(2) The Taylor series of a function f(x) about a point a is unique.

Example 1 Series Expansion of ex about 0

r =1, 2, 3, 
r
() (r)
f(x)= e x , f (x)= e x , f (0)= 1
Substituting these for (2.1) in the theorem,
n+ r
xr
 
n -1 m -1 -n x
e =Σx
+ e xΣ + Rmn
r=0 r! r=0 r (n + r)!
n
However, this is an automorphism and is not interesting. Then, using lim R = 0,
m  m

 
n+ r
xr
r
 -n x n -1
e x
1-Σ =Σ
r=0 (n + r)! r=0 r!
From this,

  /Σ
(-x)n+ r (-x)r
r
 -n n -1
e = 1 -Σ
x
r=0 (n + r)! r=0 r!
And giving n =1, 2, 3,  to this, we obtain the following series.

   
 -1 (-x)1+ r  x
r
e = 1 -Σ
x
/1 =Σ
r=0 r (1+ r)! r=0 r !

    
-2 (-x)2+ r

 x1
= 1 -Σ / 1-
r=0 r (2+ r)! 1!

  r  (3+ r)! 
3+ r
/ 1 -
2! 
-3 (-x) 1 2
 x x
= 1 -Σ +
r=0 1!

- 60 -
 

Example 2 Series Expansion of log x about 1

(r -1)!
r =1, 2, 3, 
r
() r-1
f(x)= log x , f(1)= 0 , f (x)= (-1)
xr
Substituting these for (2.1) in the theorem,
n+ r
(1-x)r m -1 1
 
-n

n -1 1 1
log x = -Σ -Σ -1 + Rmn <x
r=1 r r=0 n + r r x 2
And giving n =1, 2, 3,  to this, we obtain the following series.
1+ r

  
 1 -1 1 1
log x = -Σ -1 <x
r=0 1+ r r x 2
2+ r
(1-x)1
  
 1 -2 1
= -Σ -1 -
r=0 2+ r r x 1
3+ r
(1-x)1 (1-x)2
3+ r  r   x 
1 -3 1
= -Σ -1 - -
r=0 1 2


- 61 -
16.8 Pascal Type Triangle
When there is a number triangle, such that the numbers on two sides are given and the other numbers are
given by sum of the two numbers on the upper line, we will call it "Pascal Type Triangle".

Lemma 16.8.1
In the following number triangle, cr0 (r =0, 1, 2, ) are given, and let crr = c00 ,
crs = cr-1 s + cr-1 s-1 r, s  1 .
c00
c10 c11
c20 c21 c22
c30 c31 c32 c33
c40 c41 c42 c43 c44

Then, the following expressions hold.
r -s r -1
crs = Σ r -1-tＣ s-1 ct 0 = ΣＣ
t s-1 cr-1 -t 0
t=0 t=s -1

Proof
If we calculate one byone according to the condition crs = cr-1 s + cr-1 s-1 , it is as follows.
1 c10 c11
2 c20 c10 + c11 c22
3 c30 c10 + c11 + c20 c10 + c11 + c22 c33
4 c40 c10 + c11 + c20 + c30 2c10 + 2c11 + c20 + c22 c10 + c11 + c22 + c33 c44


When crr = c00 r =1, 2, 3, 

0 c00 1 2 3 4
1 c10 c00
2 c20 c00 + c10 1c00
3 c30 c00 + c10 + c20 2c00 + 1c10 1c00
4 c40 c00 + c10 + c20 + c30 3c00 + 2c10 + 1c20 3c00 + c10 1c00


Using binomial coefficient rＣ s ,

0 c00 1 2 3
1 c10 0Ｃ 0 c00
2 c20 1Ｃ 0 c00 + 0Ｃ 0 c10 Ｃ 1 c00
1

3 c30 2 Ｃ 0 c00 + 1Ｃ 0 c10 + 0Ｃ 0 c20 Ｃ 1 c00 + 1Ｃ 1 c10

2 Ｃ 2 c00
2

- 62 -
That is,
0 c00 1 2 3 4
0
1 c10 Σ
t=0
0- tＣ 0 ct0

1 0
2 c20 Σ1-Ｃ
t 0 ct0 Σ1- Ｃ
t=0
t 1 ct0
t=0
2 1 0
3 c30 Σ
t=0
2- tＣ 0 ct0 Σ
t=0
2- tＣ 1 ct0 Σ2- tＣ 2 ct0
t=0
3 2 1 0
4 c40 Σ
t=0
Ｃ 0 ct0
3- t Σ
t=0
Ｃ 1 ct0 Σ3-tＣ 2 ct0 Σ3-tＣ 3 ct0
3- t
t=0 t=0

Thus, we obtain the following expression.
r -s
crs = Σr -1-tＣ s-1 ct 0 r, s  1
t=0
And,
r -s
crs = Σr -1-tＣ s-1 ct 0
t=0

= r -1-(r -s)Ｃ s-1 cr-s 0 + r -1-(r -s -1)Ｃ s-1 cr-s-10 +  + r -1-1Ｃ s-1 c10 + r -1-0Ｃ s-1 c00
= s -1Ｃ s-1 cr-s 0 + sＣ s-1 cr-s-10 +  + r -2Ｃ s-1 c10 + r -1Ｃ s-1 c00
r -1
= ΣＣ
t s-1 cr-1 -t 0
t=s -1

Example When c00 = 4 , c10 = 3 , c20 = 2 , c30 = 1

Since c00 = 4Ｃ 1 , c10 = 3Ｃ 1 , c20 = 2Ｃ 1 , c30= 1Ｃ 1 , ct0= 4-tＣ 1 .
Then,
 C2 := (r,s)-> sum(binomial(r-1-t,s-1)*binomial(4-t,1), t=0..r-s)
rs 
   
(r, s)  r1t  4t
s1 1
t0

 C2(3,0);C2(3,1);C2(3,2);C2(3,3)

Note
When ct0 = 1 t =0, 1, 2,  , this number triangle reduces to Pascal's Triangle.

2009.02.18
K. Kono
Alien's Mathematics

- 63 -
17 Super Integral of the Product of Two Functions

17.1 Super Integral of f (x) g (x)

17.1.1 Super Intagration by parts

Formula 17.1.1
r , p are positive numbers, f r be an arbitrary r th order primitive function of f(x) and g
<> (r )
Let
be the r th order derivative function of g(x ) . At this time, if there is a certain constant a and

f
<r>
(a ) =0 or g (r)(a) =0 for r [0, p], the following expression holds.

  r   f
x x  p x x
 <p>
f <0>g(0)dx p = f g(0) - Σ  <r> (r)
g dx p (1.1)
a a r=1 a a

Proof
In Formula 16.1.1 (1.3) in 16.1.1 , let us analytically continue the index of the integration operator to
[0 , p] [1 , n] . Since Σ have to become an infinite series according to this, the upper limit n
from
is expanded to  .
1


x x
3 6
Example Super Intagration by parts of x x dx
0 0

Let f (0) = x  , g (0) = x  in (1.1) . Then

(1+ +) ++p

x x
+
Left:  x dx p = x
0 0 (1+ + + p)
Higher Integral of right side is
(1+ ) +r (1+) -r p
 
x x x x
 <r>  (r)
 x  x  dx =
p
 x x dx
0 0 0 0 (1+ + r) (1+ - r)
(1+ ) (1+)
=
(1+ +r) (1+ - r)
 x + dx p 
(1+ +) ++p (1+ ) (1+)
= x
(1+ + + p) (1+ + r) (1+ - r)
Then, from (1.1) ,
(1+) +p  (1+ + ) ++p  p (1+) (1+ )
Right:
(1+ +p)
x x -
(1+ + +p)
x Σr=1 r   (1+ +r) (1+ -r)
Substitute  =1/2,  =1/3, p =1/6 for both sides. Then
1 1 1
(11/6)
x
x x
 2 3
x dx
6
= x
0 0 2
(3/2) (11/6)  1/6 (3/2) (4/3)
=
 5/3
x-
2
xΣ
r=1 r    3/2+r    4/3-r 
When rhese are calculated, it is as follows. Both sides are corresponding completely

-1-
Note1
The above example is the one for the numerical proof of this formula. Actually, this formula is rarely used
directly.

Note2
The more general formulas of Theorem 16.1.1 in 16.1.1 are as follows.

   s  f a ga  
x x p -1 r r x x
 f <0> (0)
g
<p>
dx = f g(0) - ΣΣ
p <p -r+s> (s)
 dx r
a(p) a(0) r=0 s=0 p-r p-r ap a p- r

 r    f g dx
p  x x
<r> (r)
-Σ p
r=1 a(p) a(0)

 f
r

s a
x x p -1 r r (x -a)
 <0> (0)
g p <p>
dx = f g(0) - ΣΣ f g a
<p -r+s> (s)
a a r=0 s=0 r!

 r   f g dx
p x x
<r> (r)
-Σ p
a a r=1
These are notionally right. However, it is almost impossible to obtain the constant-of-integration function

r=0 s=0  s 
p -1 r r
Σ Σ  contained in these. So, regrettably below, we build a theory based on the easiest (1.1) .

17.1.2 Super Integral of the Product of Two Functions

Theorem 17.1.2
<r> (r)
Let r, p are positive numbers, f be an arbitrary r th order primitive function of f(x) , g be the r th
order derivative function of g(x) ,  (n , m),(p ) are the beta function and the gamma function respectively
At this time, if there is a number a such that
f (a) = 0 r [0 , m + p] g (a) = 0 s [0 , m + p -1,]
<r> s ()
or
then the following expression holds.

 r
x x m -1 -p
 f g dx p = Σ f<
p + r> (r)
g + Rmp (2.1)
a a r=0

 k   f
(-1)m  1 p -1 x x
<m+ k > (m+ k )
(p , m) Σ
Rmp = g dx p (2.1r)
k =0 m + k a a

Proof
In Formula 16.1.2 (2.3) in 16.1.2 , let us analytically continue the index of the integration operator to
[0 , p] [1 , n] .
from Since Σ have to become an infinite series according to this, the upper limit n
is expanded to  .

-2-
Riemann-Liouville Integral Expression
According to Formula 7.2.3 in 7.2.3 , Super Integral is expressed by Riemann-Liouville Integral as follows.

 f(x)dx  (x -t)

x x 1 x
p p-1
= f(t)dt
a a (p) a
If Theorem 17.1.2 is rewritten using this, it is as follows. Since the left side does not have operation functions,
this right side is indispensable for the calculation.

Theorem 17.1.2'
<r> (r)
Let r, p are positive numbers, f be an arbitrary r th order primitive function of f(x) , g be the r th
order derivative function of g(x) ,  (n , m),(p ) are the beta function and the gamma function respectively
At this time, if there is a number a such that
f (a) = 0 r [0 , m + p] g (a) = 0 s [0 , m + p -1,]
<r> s()
or
then the following expression holds.

 r
1 x m -1 -p
(x -t)p-1 f g dt = Σ f<
p + r> (r)
g + Rmp (2.1')
(p) a r=0

m+k  k  
m p -1  x
(-1) 1 p-1 <m+ k > (m+ k )
Rmp =
(p)(p , m) Σ k=0
(x -t)
a
f g dt (2.1'r)

-3-
17.2 Super Integral of x^a f (x) (general)

Formula 17.2.0
Let (z) be the gamma function,  (n , m) be the beta function, f <r> be an arbitrary r th primitive
f ar
<> <r >
function of f(x) and be the function values of f on a . Then the following expressions hold

for a positive number p .

r [0 , m + p] or a = 0
<r>
(1) When f (a) = 0
(1+ ) - r
  
x x m -1 -p
<0> 
 f x dx p = Σ p r
f<+> x
a a r=0 r (1+ - r)
(1+ )
  
(-1)m  1 p -1 x x
 m- k
+ Σ  f <m+ k >
x - dx p (0.1)
(p , m) k =0 m + k k a a (1+  - -
m k )
Especially, when  = m =0, 1, 2, 

(1+m) m- r
  
x x m -p
 f <0>x m dx p = Σ p r
f<+> x (0.1')
a a r=0 r (1+ m - r)
where, if  = -1, -2, -3,  , it shall read as follows.
(1+ ) (- + r)
 (-1)-r r = r, s, m+ k
(1+ - r) (- )
f (a) = 0 s [0 , m + p -1]
s ()
(2) When a =0 or
(1+ )
  
x x m -1 -p

 x f(0)dx p = Σ r (r)
x +p+ f
a a r=0 r ( 1+ + +
p r )
(1+ )
  
m  x x
(-1) 1 p -1 +m
+ Σ  m+ k
x +k f ( )dx p (0.2)
(p , m) k =0 m + k k a a  (1+  + +
m k )
where   -1, -2, -3,  &  + p  -1, -2, -3, .

Proof
Analytically continuing the index of the integration operator in (0.3) and (0.5) in Formula 16.2.0 ( 16.2 ) to
[0 , p] [1 , n] , we obtain (0.1) and (0.2) in this theorem.
from
Since (1+ - m - k) =  k =0, 1, 2, 3,  when  = m-1 m =0, 1, 2,  , the remainder
term in (0.1) disappears, and is as follows.
(1+m -1) m-1- r
  
x x m -1 -p
 f <0>x
m-1
dx p = Σ f <p + r>
x
a a r=0 r 1+m -1- r 
Then, replacing m-1 with m, we obtain (0.1').

-4-
17.3 Super Integral of x^a f (x) (particulars)
In this section, substituting various functions f for Formula 17.2.0 in previous section, we obtain various
formulas. There are (1) and (2) in Formula 17.2.0, and we may also choose whichever. However, what we wont

is the expression or approximation of super integral of x f(x ) by the series. So, in the selection of (1) or (2),
we choose the way where such a well-behaved series is obtained.
Moreover, also in which formula, if  = -1, -2, -3,  , it shall read as follows.
(1+ ) (- + r)
 (-1)-r r = r, s, m+ k
1+ - r  (- )

17.3.1 Super Integral of (ax+b)p(cx+d)q

Formula 17.3.1
The following expressions hold for p >0 and s >0 .

 
x x
(ax +b)p(cx +d)qdx s
b b
- -
a a

(1/a)s+ r (1+p)(1+q) (ax +b)p+s+ r

 
m -1 -s
=Σ + Rms (1.1)
r=0 r (1/c) (1+p +s + r)(1+q -r) (cx +d)
r r-q

(-1)m  1 m+ k
(1+p)(1+q)
  
s -1 c
s
Rm = Σ
(s, m) k =0 m + k k a (1+p +m + k)(1+q -m - k)

 
x x
  (ax +b)p+m+ k (cx +d)q-m- k dx s (1.1r)
b b
- -
a a
lim Rms = 0
m 

Especially, when m =0, 1, 2, 

 
x x
 (ax +b)p (cx +d)mdx s
b b
- -
a a

(1/a)s+ r (1+p)(1+m) (ax +b)p+s+ r

r
m -s
=Σ (1.1')
r=0 (1/c)r (1+p +s + r)(1+m -r) (cx +d)r-m

3
Example1 The 1/2th order integral of x-2 3x+4
The zeros of this primitive function are x =-4/3 , 2 . If -4 / 3 is adopted, since x -2 is a complex
number, this higher integral becomes a complex function. It is inconvenient. Then if we assume x =2 as
the lower limit of the integral, since a =1, b =-2, p =1/2 , c=3, d =4, q =1/3 , n =1 , substituting
these for (1.1), (1.1r) , we obtain
1


x x
3 2
x -2 3x +4 dx
2 2
(3/2)(4/3)
1 1

 
m -1 -1/2 -r
=Σ 3r
(x -2)1+ r(3x +4) 3 + Rm2
r=0 r (2+ r)(4/3-r)

-5-
 (-1)m (3/2)(4/3)
 
 1 -1/2
Rms = Σ 3m+ k
(1/2, m) k=0 m + k k (3/2+m + k)(4/3-m - k)
1 4 1

 (x -2)
x x +m+ k -m- k
 2
(3x +4) 3
dx 2
2 2
When the left side is replaced with Riemann-Liouville Integral and given one arbitrary point x =3.3 ,
it is as follows.
 a:=1: b:=-2: p:=1/2: c:=3: d:=4: q:=1/3: s:=1/2: m:=22:
Rie mann-Liouville Integral
 g := x->1/gamma(s)*int((x-t)^(s-1)*(a*t+b)^p*(c*t+d)^q,t=-b/a..x)
x
1 s1 p q
x   (x  t)  (a  t  b)  (c  t  d) d t
(s)
 b
a

Series
 f := x-> sum(binomial(-s,r)*((1/a)^(s+r)/(1/c)^r)
*(gamma(1+p)*gamma(1+q)/(gamma(1+p+s+r)*gamma(1+q-r))
*((a*x+b)^(p+s+r)/(c*x+d)^(r-q)), r=0..m-1)
m1
  (1 sr psr
 s   (1  p)  (1  q)
a )    (a  x  b)

x 1 r  (1  p  s  r)  (1  q  r)  rq
r0
r (c) (c  x  d)

 float(g(3.3)),float(f(3.3))
2.701947993, 2.701947994

Although this integral cannot be expressed with an elementary function, it can be expressed with a series.
Here is the meaning of this formula.

3 2
Example1' The 5 / 2th order integral of 3x+4 (x-2)
Since this is the example of application of (1.1') , we assume the noninteger root as the lower limit of the
integral. Since a =3, b =4, p =1/3 , c=1, d =-2, s =5/2 , substituting these for (1.1') , we obtain
5 17
5 +r +r

  (4/3)(3)
x x 2 6

  3
3 2
2 -5/2 1 (3x +4)
3x +4 x -22dx =Σ
4 4 r =0 r 23/6+ r(3-r) (x -2)r-2
- -
3 3

(4/3)(3)
 
3 2 4 5
3x +4 x -2 5 3 x +4 x -2 35 3 x +4
= - +
5 1 23/6 2! 23 29/6 1! 832 35/6 0!
2 6
3 3 x +4

When the left side is replaced with Riemann-Liouville Integral and given one arbitrary point x =2.1 ,
it is as follows.
 a:=3: b:=4: p:=1/3: c:=1: d:=-2: m:=2: s:=5/2:

-6-
Series
 f := x-> sum(binomial(-s,r)*((1/a)^(s+r)/(1/c)^r)*(gamma(1+p)
*gamma(1+m)/(gamma(1+p+s+r)*gamma(1+m-r)))
*((a*x+b)^(p+s+r)/(c*x+d)^(r-m)), r=0..m)
m 
  (1 sr psr
(1  p)  (1  m)
a )  
 s   (a  x  b)

x 1 r  (1  p  s  r)  (1  m  r)  (c  x  d)r  m
r0
r (c)

 float(g(2.1)),float(f(2.1))
44.54679172, 44.54679172

17.3.2 Super Integral of x log x

Since the zero of the lineal higher primitive function of x  log x for  , n such that  +n > 0 was 0,
naturally, the zero of the lineal super primitive function for  , p such that  +p > 0 is also 0.

Formula 17.3.2
When (z),  (z) are the gamma function and the psi function respectively, the following expressions
hold for  , p such that  + p >0 .
log x -(1+p+r) -  (1+) +p
 x r
x x

m -1 -p
 log x dx p = Σ x + Rmp (2.1)
0 0 r=0 (1+p+r) (1+ -r)
log x -(1+m+k) -  (1+) x 
 (p , m) Σ m + k  k   
m p -1 x x
(-1) 1
Rmp =  dx p
k=0 0 0 (1+m+k) (1+ - m - k)
lim Rmp = 0
m 

Especially, when m =0, 1, 2, 

log x -(1+p+r) -  (1+ m) m+ p
 x r
x x m -p
 m
log x dx p = Σ x (2.1')
0 0 r=0 (1+p+r) (1+ m -r)

Example1 The 1.9th order integral of x log x

Substituting  =1/2 , p =1.9 for (2.1), we obtain

log x -(2.9+r)-  (3/2) 2.4

  
x x m -1 -1.9
 x log x dx
1.9
=Σ x
1.9
+ Rm
0 0 r=0 r (2.9+r) (3/2-r)
When m=20000, if the values of the both sides on arbitrary point x=2 are calculated, it is as follows.
Since the convergence is very slow, even if calculated so far, both sides are corresponding only to 5 digits
below a decimal point.

-7-
 3
Example1' The 2.1th order integral of x log x
Substituting m =3 , p =2.1 for (2.1') , we obtain

log x - (3.1+r)-  (4) 5.1

  
x x 3 -2.1
 x 3 log xdx 2.1 = Σ x
0 0 r=0 r (3.1+ r) (4-r)
If the values of the both sides on arbitrary point x=3 are calculated, it is as follows.
All digits in both sides are corresponding this time.

Complete Automorphism
Observing well Formula 17.3.2 , we notice that the integral of the completely same type as the left side is
included in the remainder. In such a case, we can take out the integral of the purpose by transposition.

Formula 17.3.2'
The following expression holds for  , p such that  + p >0 .

 x
x x

 log xdx p
0 0
log x -(1+p) -  (1+)  p -1 (2+k) + 

=
(1+p)
+p Σ
 1+ + p  k=0 k  
(1+ k)  (1+ k,  - k) +p
2
x

 
 p -1 1
1 + pΣ
(1+ k)  (1+ k,  - k)
2
k =0 k
(2.2)

Proof
Analytically continuing the index of the integration operator in Formula 16.3.2' ( 16.3 ) to [0 , p ] from
[1 , n] , we obtain the desired expression.

17.3.3 Super Integral of x sinx , x cosx

When   0, -1, -2, , the lineal higher integral of x  sin x had variable lower limits. Therefore,
the lineal super integral of this has also variable lower limits. Since the value of the lower limit to given 
becomes the function of p , we denote this ak , k [0 , p] .

Formula 17.3.3
When a0  ap are the zeros of the lineal super primitive of x m sin x , the follwing expressions hold.

(1+ m) m- r (p + r)
a a r
-p
 
x x m
 x m sin x dx p = Σ x sin x - (3.1s)
p 0
r=0 (1+ m - r) 2
(1+ m) (p + r)
  r  (1+ m - r) x
-p
 
x x m
 x cosx dx = Σ m p m- r
cos x - (3.1c)
ap a0 r=0 2

-8-
Proof
Analytically continuing the index of the integration operator in Formula 16.3.3 (3.1') in 16.3.3 to [0 , p]
from[1 , n ] , we obtain (3.1s). Also, (3.1c) is obtained in a similar way.

3
Example The 7 / 2th order integral of x sinx
Substituting m =3 , p =7/2 for (3.1s) , we obtain

(4) 3- r (7/2+ r)

7

  
-7/2
 
x x 3
 x sin xdx = Σ
3 2
x sin x -
ap a0 r=0 r (4- r) 2
This super integral can not be examined by Riemann-Liouville Integral. Then we draw this with the 3rd and
the 4th order integral on a figure side by side. It is as follows.

Formula 17.3.3' (Collateral Super Integral)

 , p such that   -1, -2, -3,  &  + p >0 .
The following expressions hold for

(1+ ) r
  
m -1 -p

 
x x

 x  sin xdx p = Σ r
x +p+ sin x + + Rmp (3.2s)
0 0 r=0 r 1+ + p + r  2
(1+) (m + k) 
 k  
(-1)m  1 p -1 x x
Rmp =  x   dx
+m+ k p
 (p , m) Σ
sin x +
k=0 m+ k 0 0  1+ +m +k  2
(1+ ) r
 r
-p
 +R
x x m -1

 x  cosxdx p = Σ
r p
x +p+ cos x + (3.2c)
0 0 r=0 1+ + p + r  2 m

 1+ (m + k) 
(-1)m  1
 
 k   1+ +m +k
p -1 x x

  dx
( )
Rmp =  x
+m+ k p
 (p , m) Σ
cos x +
k=0 m+ k 0 0   2
lim Rmp = 0
m 

Proof
Substituting f(x)= sin x , a =0 , etc . for Formula 17.2.0 (2) , we obtain the desired expression.

3/2
Example Collateral the 1 / 2th order integral of x sinx
Substituting  =3/2 , p =1/2 for (3.2s), we obtain

-9-
 (5/2) 2+ r r
3 1

  
-1/2
 +R
x x m -1
 2
x sin xdx 2
=Σ x sin x + 1/2
0 0 r=0 r (3+ r) 2 m

When the left side is replaced with Riemann-Liouville Integral and given one arbitrary point x =3 ,
it is as follows.

17.3.4 Super Integrals of x sinhx , x coshx

Formula 17.3.4
When a0  ap are the zeros of the lineal super primitive of x m sinh x , the follwing expression holds.

(1+ m) m- r e x -(-1)p+ re -x
aa x r
x x m -p
m
sinh x dx = Σ
p
x (4.1s)
p 0
r=0 1+ m - r 2
(1+ m) m- r e x +(-1)p+ re -x
aa x r
x x m -p
m
cosh x dx = Σ
p
x (4.1c)
p 0
r=0 1+ m - r 2

Proof
Analytically continuing the index of the integration operator in Formula 16.3.4 (4.1') in 16.3.4 to [0 , p]
from[1 , n ] , we obtain (4.1s). Also, (4.1c) is obtained in a similar way.

3
Example The 0.999th order integral of x sinhx
Substituting m =3 , p =0.999 for (4.1s), we obtain

(1+3) 3- r e x -(-1)0.999+ re -x
aa x  
x x m -0.999
3
sinh x dx = Σ
p
x
p 1
r=0 r 1+3- r 2
This super integral can not be examined by Riemann-Liouville Integral. Then we calculate on the arbitrary point
x =1 the value of the 0.999th order integral and the value of the 1st order integral. They are as follows.
Although the former becomes a complex number, both real number parts are very near.

- 10 -
Formula 17.3.4' (Collateral Super Integral)
 , p such that   -1, -2, -3,  &  + p >0 .
The following expressions hold for

(1+ )

x -r -x

 
x x m -1 -p
 +p+ r e -(-1) e
 x sinh x dx = Σ
p
x (4.2s)
0 0 r=0 r (1+ + p + r) 2
(1+)
x

 k  
-m-k -x
(-1)m  1 p -1 x x
 +m+ k e -(-1) e
Rmp dx p
 (p , m) Σ
= x
k=0 m+ k 0 0  1+ +m +k  2
 1+
00 x cosh x dx = Σ  r  (1+( + p)+ r) x
e x +(-1)-re
-x
x x
 -p p
m -1
 +p+ r
+ Rmp (4.2c)
r=0 2
(1+) e x +(-1)-m-ke

m -x

m+ k  k 
(-1) 1 p -1  x x
p
x
+m+ k
R = Σ dx p
m
 (p , m) k=0  1+ +m +k 0 0   2
lim Rmp = 0
m 

Proof
Substituting f(x)= sinh x , a =0 , etc . for Formula 17.2.0 (2) , we obtain the desired expression.

Example Collateral the 3 / 2th order integral of x sinhx

Substituting  =1/2 , p =3/2 for (4.2s), we obtain

(3/2) 2+ r e x-(-1)-re -x
3

  
x x m -1 -3/2
x sinh xdx =Σ 2
x + Rm3/2
0 0 r=0 r (3+ r) 2
When the left side is replaced with Riemann-Liouville Integral and given one arbitrary point x =2 ,
it is as follows.

- 11 -
17.4 Super Integral of log x f (x)

17.4.1 Super Integral of (log x)2

2
Since the zero of the lineal higher primitive of (log x) was x =0 , naturally, the zero of the lineal super
primitive is also x =0 .

Formula 17.4.1
x p log x log x - (1+ p)-

x x
2 p
log xdx =
0 0 (1+ p)
-p log x - (1+ p +r)-(r)
 
m -1
- x p Σ (-1)r + Rmp (1.1)
r=1 r (1+ p +r)
xp
Rmp = 
(p , m)(1+ p)
k-1

 k  log x - (1+ p)- (1+ m + k)-2

 (-1) p -1
Σ (1.1r)
k =0 ( m + k)2
lim Rmp = 0
m 

Proof
Analytically continuing the index of the integration operator in Formula 16.4.1 in 16.4.1 to [0 , p ] from
[1 , n ] , we obtain the desired expression.

The remainder in the super integral of product of two functions becomes a series of super integrals. So it is
2
very difficult to calculate this. However, the calculation of the remainder in the super integral of (log x) is
exceptionally easy. So we show it as follows.

2
Example The 3/2th order integral of (log x)
We substitute p =3/2 , m=3 for (1.1), (1.1r) and calculate the remainder 300 terms. The function values
on the arbitrary point x =0.3 are as follows. Although the difference of the left side and the series is large,
if the remainder is added, all digits in both sides are corresponding.

- 12 -
 Incidentally, even if this series is calculated to m=300,000 , both are corresponding only to 4 digits below
the decimal point. From this, we know that the convergence of the series is extraordinarily slower than the
remainder.

- 13 -
17.5 Super Integral of e^x f (x)

17.5.1 Super Integral of e x x

Since the zero of the lineal higher primitive of exx was - , naturally, the zero of the lineal super primitive
is also - .

Formula 17.5.1

t

When (a , x ) = x0
a-1 -t
e dt is the incomplete gamma function, the following expression holds for
x
(-1)
  r  (p -r + , -x)x
x x
 x
 p -1
 x e dx = p
Σ r
(1.n)
- - (p) r=0

Proof
Analytically continuing the index of the integration operator in Formula 16.5.1 in 16.5.1 to [0 , p ] from
[1 , n] , we obtain the desired expression.

3
Example1 The 2.1th order integral of e x x
When  =1/3 , p =2.1 , if the left side is replaced by Riemann-Liouville Integral and an arbitrary point
x = 1.7 is given to both sides, it is as follows.

When  is a natural number, x can be differentiated completely and the following formula holds.

Formula 17.5.1'
When m =0, 1, 2,  ,
(1+ m) m- r
 r
x x m -p
 e x x mdx p = e xΣ x (1.n')
- - r=0 (1+ m - r)

4
Example1' The 5.1th order integral of ex x
Substituting m=4 , n =5.1 for (1.n'), we obtain

(1+4) 4- r
  
x x 4 -5.1
 e x x 4dx 5.1 = e xΣ x
- - r=0 r 1+4- r
If the values of the both sides on arbitrary point x =2 are calculated, it is as follows.

- 14 -
Collateral Super Integral of e x x
For the collateral super integral with the zero x=0, the following formula holds.

Formula 17.5.1"
The following expressions hold for  , p such that
  -1, -2, -3,  &  + p >0.
(1+ )
  
x x m -1 -p

 e x x dx p = e xΣ r
x +p+ + Rmp (1.n")
0 0 r=0 r (1+ + p + r)
(1+ )

(-1)m  1
 
p -1 x x
p
Rm = Σ  e x x + m+k dx p
(p , m) k =0 m + k k (1+ + m +k) 0 0
lim Rmp = 0
m 

3
Example1" Collateral the 2.1th order integral of e x x
Assume  =1/3 , p =2.1 and an arbitrary point x =1.7 are the same as Example1 ( Lineal 2.1 th
order integral ) . The values of the both sides are completely corresponding as follows. However, these are
considerably different from the value of Example1.

17.5.2 Super integral of e x log x

All the polynomials obtained by applying Theorem 17.1.2 to e xlog x become asymptotic expansions, and
they are hardly helpful.

17.5.3 Super integral of e x sinx , e x cosx

Formula 17.5.3
 p
 e sin xdx = sin 4  e sin x - 4 
x x p
x p x
(3.0s)
- -

 p

p

 4  4 
x x
 e cosxdx x
= sin p
e cos x - x
(3.0c)
- -

- 15 -
Proof
Analytically continuing the index of the integration operator in Formula 16.5.3 in 16.5.3 to [0 , p] from
[1 , n] , we obtain the desired expressions.

Example The 1/2th order integral of e x sinx

If the values of the both sides on arbitrary point x =2.3 are calculated, it is as follows.
 p:=1/2:
Le ft; Rie mann-Liouville Inte gral
 g := x-> 1/gamma(p)*int((x-t)^(p-1)*E^t*sin(t), t=-infinity..x)
x
1  p1
x   (x  t)  E  sin(t) d t
t
( p)


Right: Formula
 s := x-> sin(PI/4)^p*E^x*sin(x-p*PI/4)
 p  


   p

x  sin  E  sin x 
x
4 4

Le ft , Right
 float(g(2.3)), float(s(2.3))
7.916851572, 7.916851572

Trigonometric Series
If Formula 16.5.4 in 16.5.4 is extended to the real number, the following trigonometric series can be obtained

Formula 17.5.4
k  -p
p
 
p -1
 
 1 1
Σ
k =1 k k -1
sin
2
=
p
sin
4
sin
4
(4.1)

k  -p
p
Σk  
p -1
 
1 1 1
cos = sin cos - (4.2)
k =1 k -1 2 p 4 4 p

Alternating Binomial Series

If Formula 16.5.4' in 16.5.4 is extended to the real number, the following alternating binomial series can be
obtained.

Formula 17.5.4'
(-1)k  -p-1
(p +1)
 
p
 
 1
Σ
k =0 2k +1 2k
=
p +1
sin
4
sin
4
(4.3)

 (p +1)
  p +1   4  
 (-1)k p 1 -p-1
Σ
k =1 2k
=
2k -1
sin cos
4
-1 (4.4)

If a horizontal axis is set as p and these are illustrated , it is as follows. Although the left side is blue and
right side is red, since both sides overlap exactly, the left side (blue) is not visible.

- 16 -
- 17 -
17.6 Super Integral of f (x) / e^x

e x
-x
17.6.1 Super Integral of
Since the zero of the lineal higher primitive of e -x x  was , naturally, the zero of the lineal super
primitive is also .

Formula 17.6.1

t

When (a , x ) =
a-1 -t
e dt is the incomplete gamma function, the following expression holds.
x


(-1)p 
 
x x

p -1
-x
e x dx = Σp
(-1)r (p -r + , x)x r (1.1)
  (p) r=0 r

Proof
Analytically continuing the index of the integration operator in Formula 16.6.1 in 16.6.1 to [0 , p ] from
[1 , n] , we obtain the desired expression.
- 3
Example1 The 2.1th order integral of e x x
When  =1/3 , p =2.1 , if the left side is replaced by Riemann-Liouville Integral and an arbitrary point
x = 1.7 is given to both sides, it is as follows.

When  is a natural number, x  can be differentiated completely. And the following formula holds.

Formula 17.6.1'
When m =0, 1, 2,  ,
(1+ m) m- r

(-1)p
r
x x m -p
 -x
e x dx = m p
Σ (-1) r
x (1.1')
  ex r=0 (1+ m - r)

- 4
Example1' The 5 .1 th order integral of e x x
Substituting m =4 , p =5.1 for (1.1'), we obtain

(1+4) 4- r

(-1)5.1
 
x x 4 -5.1
 -x
e x dx 4 5.1
= Σ(-1)
r
x
  e x r=0 r (1+4- r)
If the values of the both sides on arbitrary point x =1.3 are calculated, it is as follows.

- 18 -
Collateral Super Integral of e x x 
-

For the collateral super integral with the zero x=0, the following formula holds.

Formula 17.6.1"
The following expressions hold for  , p such that   -1, -2, -3,  &  + p >0.
(1+ )
  
x x
 1 m -1 -p 
 -x
e x dx = p
Σ (-1)
-r r
x +p+ + Rmp (1.1")
0 0 e x r=0 r (1+ + p + r)
+ m+k
(1+ )
 k  (1+ + m +k) 
-k p -1 x x
1 n -1 (-1) x
(p , m) Σ
Rmp = x dx p
k =0 m + k 0 0 e
lim R n = 0
m  m

- 3
Example1" Collateral the 2.1th order integral of e x x
Assume  =1/3 , p =2.1 and an arbitrary point x =1.7 are the same as Example1 ( Lineal 2.1 th order
integral ). The values of the both sides are completely corresponding as follows. However, these are different
from the value of Example1 considerably.

-x
17.6.2 Super integral of e log x
All the polynomials obtained by applying Theorem 17.1.2 to e -xlog x become asymptotic expansions, and
they are hardly helpful.

-x -x
17.6.3 Super integral of e sinx , e cosx

Formula 17.6.3
 p

p

   
x x
 e sin xdx = (-1)p sin
-x p
e -xsin x + (3.0s)
  4 4
 p
 e
p

   
x x
-x
cosxdx p = (-1)p sin e -xcos x + (3.0c)
  4 4

- 19 -
Proof
Analytically continuing the index of the integration operator in Formula 16.6.3 in 16.6.3 to [0 , p] from
[1 , n] , we obtain the desired expressions.

-
Example The 4 /3th order integral of e x cos x
If the values of the both sides on arbitrary point x =1.2 are calculated, it is as follows.

- 20 -
17.7 Series Expansion of Super Integral

17.7.1 Series Expansion of Super Integral with a fixed lower limit

Theorem 17.7.1
m be a natural number, f r =0, 1, 2,  be the r th derivative of f , a be an arbitrary constant
(r)
Let

on the domain of f , (x, y) be the beta function. Then the following expressions hold for p >0 .


(x -a)p+ r (r)
 
x x m -1 -p
 f(x)dx = Σ p
f (x) + Rmp (1.1)
a a r=0 r  (1+ p + r)

  
(-1)m  1 p -1 x t (t -a )m+ k
 m+ k
(p , m) Σ
p
Rm = f ( )(t)dt p (1.1r)
k =0 m + k k a a ( m + k) !

 f(x)dx
x x m -1 (x -a)p+ r (r)
p
=Σ f (a) + Rmp (1.2)
a a r=0 (1+p + r)
 r -1 (x -a ) m+p+s
  Σ Ｃ
m +p +s r -1 m + p -1-r + t
Rmp = ΣΣ(-1) p-r+s
f m +s(a)
r =1 s =0 r t=s m -1
t s
 1+m +p +s

  
(-1)m  1 p -1 x (t -x)m+ k (m+ k )
t

(p , m) Σ
+ f (t)dt p (1.2r)
k =0 m + k k a a (m + k)!
Where, there shall be no term of ΣΣ at the time of p <2 .

Proof
Analytically continuing the index of the integration operator in Formula 16.7.1 in 16.7.1 to [0 , p] from
[1 , n] , we obtain the desired expressions.

Series Expansion of Riemann-Liouville Integral

Super Integral with a fixed lower limit is Riemann-Liouville Integral itself. Therefore, the above thorem is also
a theorem concerning the series expansion of Riemann-Liouville Integral. If the above thorem is rewritten by
Riemann-Liouville Integral, it is as follows.

Thorem 17.7.1'
p and an arbitrary number a on the domain of f .
The following formulas hold for a positive number

(x -a)p+ r (r)
  
1 x m -1 -p
p-1
(x -t) f(t)dt = Σ f (x) + Rmp (1.1')
(p) a r=0 r (1+p + r)
x (x -t)p-1(t -a )m+ k

 
(-1)m  1 p -1 m+ k
p
Rm = Σ
(p , m)(p) k =0 m + k k a (1+m + k)
f ( )(t)dt p

(x -a)p+ r (r)
 (x -t)
1 x m -1
p-1
f(t)dt = Σ f (a) + Rmp (1.2')
(p) a r=0 (1+p + r)
(x -a) m+p+s
t=s  s  
 r -1 t
   1+m +p +s
r -1 m + p -1-r + t m +p +s m+ s)
Rmp = ΣΣ(-1) p-r+s
Σ f( (a )
r =1 s =0 m -1 r  

(-1)m+ p (t -x)m+p+ k (m+ k )

 k 
 1 p -1 x
- Σ
(p , m)(p) k =0 m + k a (1+ m + k)
f (t)dt p (1.2'r)

Where, there shall be no term of ΣΣ at the time of p <2 .

- 21 -
Example1 Series expansion of collateral the 2.1th order integral of ex
Let f(x)= e x , then

r =1, 2, 3, 
r () (r)
f(x)= e x , f (x)= e x , f (0)= 1
Substituting these for Theorem 17.7.1 ,

 e dx
(x -a)p+ r
 
x x m -1 -p
a a
x p
=e x
Σ
r=0 r  (1+ p r
+ )
+ Rmp

  
(-1)m  1 p -1 x t (t -a )m+ k

(p , m) Σ
Rmp = e tdt p
k =0 m k
+ k a a ( m k
+ )!


x x m -1 (x -a)p+ r
 e dx = e aΣ
x p
+ Rmp
a a r=0 (1+p + r)
(x -a) m+p+s
t=s  s  
 r -1 t
   1+m +p +s
r -1 m + p -1-r + t m +p +s
Rmp = ΣΣ(-1) Σ
p-r+s
ea
r =1 s =0 m -1 r  

 k 
(-1)m  1 p -1 x t (t -x)m+ k t p
(p , m) Σ
+ e dt
k =0 m + k a a (m + k)!
x
Although the zero of the lineal super integral of e is a = - , these expressions can not have this zero.
That is, by the above formulas, the lineal super integral of e x can not be expanded to the series. Then, we put

a =0 and expand the collateral super integral of e x to the series. When p =2.1 , if these are illustrated,
it is as follows. The left side is calculated by Riemann-Liouville Integral. Both sides overlap exactly and blue (left)
can not be seen.

Example2 Series expansion of the 2.9th order integral of log x

Let f(x)= log x , then
(r -1)!
r =1, 2, 3, 
(r)
f(0) = log x , f = (-1)r-1
xr

- 22 -
Substituting these for Theorem 17.7.1 ,


(x -a)p (x -a)p+ r (r -1)!
 
x x m -1 -p
 log x dx = p
log x -Σ (-1)r
+ Rmp
a a (1+ p) r=1 r (1+ p + r) x r


p+ r
x x (x -a)p m -1
r (x -a ) (r -1)!
 log x dx =
p
log a - Σ (-1) + Rmp
a a ( p)
1+ r=1 ( p r) a
1+ + r

If a =0 then the first formula is lineal super integral, else it is collateral super integral. On the other hand,
the second formula can not be lineal super integral.
Let us calculate the left side by the direct integration and compare it with the series by the first formula. Then
it is as follows. Although the convergence of this series is slow, as a result of taking very large m, both sids
overlapped somehow.

2010.07.14
K. Kono
Alien's Mathematics

- 23 -
18 Higher Derivative of the Product of Two Functions

18.1 Leibniz Rule about the Higher Order Differentiation

Theorem 18.1.1 (Leibniz)

When functions f(x) and g(x)are n times differentiable, the following expression holds.

 r
(n) n n (n -r)
(x)g r (x)
()
f(x)g (x) =Σ f (1.1)
r=0

Proof
Theorem 16.1.2 (2.1) in 16.1.2 was as follows.

a a r
x x m -1 -n
 f <0>g(0)dx n = Σ f <n + r> (r)
g
r=0
n 1

 s f a a
-n + r
n -1 m -1 x x
- ΣΣ <n -r+ s>
a n-r ga
(s)
 dx r
r=0 s=0 n-r
n n- r+1

a a
n -1 r -1 r -1 x x
t s  m +n-1 -r + tＣ m-1 f a 
m+ n -r+s (m+ s)
+ (-1)m ΣΣΣ Ｃ ga dx r
r =1 s =0 t=s n- r n- r
n n- r+1

a a f
m
(-1) Ｃk n -1 n -1 x x
<m+ k > (m+ k )
+ Σ
(n , m) k=0 m + k
g dx n
n 1

Should be noted here is the next two.

i When n =1 ∑∑∑ of the 3rd line does not exist, when n =0 ∑∑ of the 2nd line does not exist also.
ii When the binomial coefficient of the 4th line is generalized, the upper limit n -1 of ∑ can be replaced
by .
Since 1 > 0  -n at the time n =0, 1, 2,  , if the index n of the integration operator is substituted for
-n in consideration of these, it becomes as follows.

a a  r
x x m -1 n
 f <0>g(0)dx -n = Σ f <-n + r> (r)
g
r=0
n 1

  f
(-1)m  1

-n -1 x x
<m+ k > (m+ k )
(-n , m) Σ
+ g dx -n
k=0 m + k k an a1
Since m may be arbitrary integer, when m = n +1 , it is as follows.

a a  r f
x x n n
 f <0>g(0)dx -n = Σ <-n + r> (r)
g
r=0
n 1

 a a f
(-1)n+1

 1 -n -1 x x
<n +1+ k > (n +1+ k )
+ Σ
(-n , n +1) k =0 n +1+ k
k
g dx -n
n 1

However, since (-n , n +1)=  for n =0, 1, 2,  , the 2nd line disappears. That is,

 
x x n n <-n + r> (r)
 f g dx -n = Σ f g n =0, 1, 2, 
an a1 r=0 r
-n
Then, replacing the integration operators dx , <-n + r> with the differentiation operators (n ),(n -r)
respectivly, we obtain the desired expression.

-1-
18.2 Higher Derivative of x^a f (x)

Formula 18.2.0
When (z) denotes the gamma function and f(x) is n times differentiable continuous function,
the following expressions hold for a natural number n.
(1)
(1+ ) - r (n - r)
 r
(n) n n
x  f(x) =Σ x f (x) (0.1)
r=0 (1+ -r)
Where, if  = -1, -2, -3, , it shall read as follows.
(1+ ) (- + r)
 (-1)-r
1+ - r  (- )
(2) Especially, when  = m =0, 1, 2, 
(1+ m) m- r (n - r)

(n) m n
x m f(x) = Σ x f (x) (0.1')
r=0 r (1+ m -r)
(3) When   -1, -2, -3,  &  -n  -1, -2, -3, 
(1+ )

(n) n n 
x  f(x) = Σ
r (r)
x -n+ f (x) (0.2)
r=0 r (1+ - n + r)

Proof
When g(x)= x  in Theorem 18.1.1 , since
(r) (1+ ) - r
x  = x
(1+ -r)
we obtain the following expression immediately.
(1+ ) - r (n - r)
 r
n n
x  f(x)
(n)
=Σ x f (x) (0.1)
r=0 (1+ -r)
Especially, when  = m =0, 1, 2,  , (0.1) is as follows.

(1+m) (1+ m)
r 
(n) n n
m
x f(x) x m- r f n- r (x) = 0 for m < r  n
( )
=Σ
r=0 (1+m -r) (1+ m -r)
(1+m)
 r  (1+m -r) x r  =0
nm n
=Σ m- r (n- r )
f (x) for n < r  m
r=0
We adopt the convenient latter for mathematical software.
When  = -1, -2, -3,  , from 1.1.5 ( Properties of the Gamma Function ) (5.5),
(-z) (1+z + n)
= (-1)-n (n is a non-negative integer )
(-z - n) (1+ z)
Then substituting -z = 1+ , n = r for this, we obtain the proviso.
Last, replacing r with n -r in (0.1), we obtain (0.2).

Below, substituting various functions f for Formula 18.2.0 , we obtain various formulas.
Although there are (1) and (2) in Formula 18.2.0, since (2) is almost meaningless in the case of higher
differentiation, we adopt (1) in principle.

-2-
18.2.1 Higher Derivative of (ax+b)p(cx+d)q

Formula 18.2.1
The following expressions hold for p >0 and n =1, 2, 3,  .
q (n)
(ax +b)p(cx +d) 
(1/a)-n+ r (1+p)(1+q) (ax +b)p-n+r

n n
=Σ (1.1)
r=0 r (1/c)r (1+p -n + r)(1+q -r) (cx +d)r-q
Especially, when m =0, 1, 2, 
(n)
(ax +b) (cx +d) 
p m

(1/a)-n+ r (1+p)(1+ m) (ax +b)p-n+ r

 r
m n
=Σ (1.1')
r=0 (1/c)r (1+p -n + r)(1+ m -r) (cx +d)r-m

Proof
Let f(x)=(ax +b)p , g(x)=(cx +d)q , then
-n+r
(1+p)
 
n - r) (n -r) 1
f( = (ax +b)p = (ax +b)p-n+r
a (1+p -n + r)
-r
(1+q)
 
(r) 1
q  1, 2, 3, 
(r)
g = (cx +d)q = (cx +d)q-r
c (1+q - r)
Substituting these for Theorem 18.1.1 , we obtain (1.1) .
And especially, when q = m = 0, 1, 2,  , from (1.1)
(n)
(ax +b)p(cx +d)m
(1/a)-n+ r (1+p)(1+m) (ax +b)p-n+r
 r
n n
=Σ
r=0 (1/c)r (1+p -n + r)(1+m -r) (cx +d)r-m
(1/a)-n+ r (1+p)(1+m) (ax +b)p-n+r
 r
m n
=Σ
r=0 (1/c)r (1+p -n + r)(1+m -r) (cx +d)r-m
We adopt the latter expression as (1.1').

3
Example1 The 2nd order derivative of x-2 3x+4
a =1, b =-2, p =1/2 , c =3, d =4, q =1/3 , n =2 for (1.1) ,
Substituting

2 r (3/2)(4/3) 3 1


(2) 2 r- -r
 x -2 3x +4 =Σ
3
3 (x -2) (3x +4) 3
2
r=0 r (r -1/2)(4/3-r)

 
1 1 2
= - x -2 3 3x +4 - +
4(x -2)2 (x -2)(3x +4) (3x +4)2

2
Example1' The 3rd order derivative of x-2 (3x+4)
Substituting a =1, b =-2, p =1/2 , c =3, d =4, m =2 , n =3 for (1.1') ,

-3-
 5
- +r
(3/2)(3) 2

r
(3) 2 3 (x -2)
 x -2 (3x +4) 
2
=Σ 3r
r=0 (-3/2+r)(3-r) (3x +4)r-2

 8(x -2) 
1
3 2 32 33
= x -2 (3x +4) 3
- 2
+ 2
2(3x +4) (x -2) (3x +4) (x -2)

Example2 The 3rd order derivative of x-2 / (3x+4)

When q = -1, -2, -3,  , (1.1) can be read as follows.
q (n)
(ax +b)p(cx +d) 
(1/a)-n+ r (1+p)(-q + r) (ax +b)p-n+ r

n n
=Σ
r=0 r (-1/c)r (1+p -n + r)(-q) (cx +d)r-q
Substituting a =1, b =-2, p =1/2 , c =3, d =4, q =-1 , n =3 for this,
5
(3) - +r
(3/2)(1+ r) (x -2)
 
2

r
x -2 3 3
=Σ (-3)r
3x +4 r=0 (-3/2+r)(1) (3x +4)r+1
x -2
 8(x -2) 
3 9 27 162
= 3
+ 2
+ 2
- 3
3x +4 4(x -2) (3x +4) (x -2) (3x +4) (3x +4)

18.2.2 Higher Derivative of x  log x

Formula 18.2.2
(n -r) (1+) -n (1+) -n
r 
 (n) n -1 n
x log x = -Σ(-1)n-r x + x log x
r=0 (1+ -r) (1+ -n) (2.1)
Especially, when m = 0, 1, 2, 
(n -r) (1+m) m-n (1+m) m-n
r 
m (n) n -1 n
x log x = -Σ(-1)n-r x + x log x
r=0 (1+m -r) (1+m -n)
(2.1')
Where, there shall be no 2nd term of the right side at the time of m< n .

Proof
Let f(x)= log x . Then

(log x) (n -r)
= -(-1)n-r(n -r) x -n+r r = 0, 1, , n -1
= log x r =n
Substituting these for (0.1) in Theorem 18.2.0 , we obtain (2.1).
When m = 0, 1, 2,  , applying (0.1') , we obtain the following.

(n -r) (1+m) (1+m)

 r  (1+m -r) x + (1+m -n) x
m (n) n -1 n m-n m-n
x log x = -Σ(-1)n-r log x
r=0

(1+m)
 r  (1+m -r) x (log x)
nm
m-n (n -r)
+Σ
r=n +1
Where, since r does not reach n at the time of m< n , the 2nd term does not exist.

-4-
Example1 The 3rd order derivaive of x log x
Substituting  =1/2 , n =3 for (2.1) ,
5 5
(2-r) (3/2) - 2 (3/2) - 2
r
(3) 2 3
 x log x = -Σ(-1) 3-r
x + x log x
r=0 (3/2-r) (-3/2)
5
(3/2) (3/2) (3/2)
   
3 3 3 -
2
=- - 2! + 1! - 0! x
0 (3/2) 1 (1/2) 2 (-1/2)
5
(3/2) -
2
+ x log x
(-3/2)
5 5 5
(3/2) - 2
 
1 3 -
 
3 3 -
= 2- - x +
2
x log x = - + log x x 2
2 4 (-3/2) 4 8
3
Example1' The 2nd order derivaive of x log x
Substituting m =3 , n =2 for (2.1') ,

(2-r)(4) 1 (4) 1
r
(2) 1 2
x log x
3
= -Σ(-1)2-r x + x log x
r=0 (4-r) (2)

  0 3!  1 2!  x
2 1! 3! 2
2 0! 3! 1 1 3! 1
= - (-1) + (-1) + x log x
1!
= (5 + 6 log x)x 1
2
Example1" The 3rd order derivaive of x log x
Substituting m =2 , n =3 for (2.1') ,

(3-r)(3) -1
r
(3) 2 3
x log x = -Σ
2
(-1)3-r x
r=0 (3-r)

  0  1  2 (-1) 2! x
3 3 3
3 2 1 -1
=- (-1) 2! + (-1) 2! +

= 2 x -1
Example2 The 3rd order derivaive of log x /x
When  = -1, -2, -3,  , (2.1) can be read as follows.
(n -r) (- + r) -n (- + n) -n
r 
 (n) n -1 n
x log x = -(-1)nΣ x + (-1)n x log x
r=0 (-) (-)
Substituting  =-1 , n =3 for this ,
(3)
(3-r)(1+ r) -4 (4) -4
r
3
 
log x 2
= -(-1)-3Σ x + (-1)-3 x log x
x r=0 (1) (1)

  0  1  2 (1)(3)- 6 log x 
3 3 3
= x -4 (3)(1)+ (2)(2)+
1
= (11 - 6 log x)
x4

-5-
18.2.3 Higher Derivatives of x sinx , x cosx

Formula 18.2.3
(1+ ) -r (n -r)
 r
n
 
{n} n
x  sin x =Σ x sin x + (3.1s)
r=0 (1+ -r) 2
(1+ ) -r (n -r)
 r
n
 
{n} n
x  cosx =Σ x cos x + (3.1c)
r=0 (1+ -r) 2
Especially, when m = 0, 1, 2, 
(1+m) m-r (n -r)

n
 
{n} m
x m sin x =Σ x sin x + (3.1's)
r=0 r (1+m -r) 2
(1+m) m-r (n -r)

n
 
{n} m
x m
cosx  =Σr=0 r (1+m -r)
x cos x +
2
(3.1'c)

3
Example1 The 2nd order derivative of x sinx
Substituting  =1/3 , n =2 for (3.1s) ,

(4/3) 3 -r (2-r)
1

r
2
 
(2) 2
 x sin x
3
=Σ x sin x +
r=0 4/3-r  2
1 2 5
4/3 3 4/3 - 3   4/3
 0   2  -2/3
2 2 2   -
= x sinx ++ x sin x +  + x 3
sin x
4/3 1 1/3 2  
1 2 5
2 - 2 -
= - x sin x + x 3 cosx - x 3 sin x
3
3 9
2
Example1' The 3rd order derivative of x sinx
m =2 , n =3 for (3.1's) ,
Substituting

(3) 2-r (3-r)


3
 
{3} 2
x 2 sin x =Σ x sin x +
r=0 r 3- r  2
(3) 2 3 (3) 1 2
 0   1
3 3
=
3
x sin x +
2  +
2 
x sin x +
2 
(3) 0 
 2
3
+
1 
x sin x +
2 
= -x 2cosx - 6 x sin x + 6 cosx

18.2.4 Higher Derivatives of x sinhx , x coshx

Formula 18.2.4
(1+ ) - r e x -(-1)-(n -r)e -x
 r
 (n) n n
x sinh x =Σ x (4.1s)
r=0 (1+ -r) 2
(1+ ) - r e +(-1)-(n -r)e -x
x

 r
(n) n n
x  cosh x =Σ x (4.1c)
r=0 (1+ -r) 2

-6-
 Especially, when m = 0, 1, 2, 
(1+ m) m- r e x -(-1)-(n -r)e -x

(n) m n
x sinh x = Σ
m
x (4.1's)
r=0 r (1+ m -r) 2
(1+ m) n -r) -x
e x +(-1)-(
  (1+m -r) x
(n) m n m- r e
x m cosh x = Σ (4.1'c)
r=0 r 2

3
Example1 The 2nd order derivaive of x sinhx
Substituting  =1/3 , n =2 for (4.1s) ,

(4/3) 3 -r e x -(-1)-(2-r)e -x
1

r
(2) 2 2
 x sinh x =Σ
3
x
r=0 4/3- r  2
1 2 5
(4/3) 3 (4/3) - 3 (4/3) - 3
 0  
2 2 2
= x sinh x + x cosh x + x sinh x
4/3 1 1/3 2 -2/3
1 2 5
3 2 -3 2 -
= x sinh x + x cosh x - x 3 sinh x
3 9

2
Example1' The 3rd order derivative of x sinhx
m =2 , n =3 for (4.1's) ,
Substituting

(3) 2-r e x -(-1)-(3-r)e -x


{3} 2 3
x sinh x
2
=Σ x
r=0 r 3- r  2
(3) 2 (3) 1 (3)
 0  1  2  1
3 3 3 0
= x cosh x + x sinh x + x cosh x
3 2  

= x 2cosh x + 6 x sinh x + 6 cosh x

-7-
18.3 Higher Derivative of log x f (x)

18.3.1 Higher Derivative of (log x)2

Formula 18.3.1
(-1)n-1
  r (n - r)(r)
(n) n -1 n
log x
2
= 2(n)log x -Σ (1.1)
xn r=1

Proof
Let f(x)= g(x)= log x . Then
(n - r) (r)
n =1, 2, 
(n- r) (r)
(log x) = (-1)n- r-1 , (log x) = (-1)r-1
x n- r xr
Substituting these for Theorem 18.1.1 ,

 r
(n) n n
log x
2
=Σ (log x)(n - r)(log x)( r)
r=0

 0 (log x)  r  (log x) (log x)

n (n) n -1 n (n - r) ( r)
= (log x)(0) + Σ
r=1

 n  (log x) (log x)
n (n - n ) (n)
+
n n -1
Σ  r (n - r)(r)
2 (-1) n
= (-1)n-1 n
(n)log x + n
x x r=1

(-1)n-1
  r (n - r)(r)
n n -1
= 2(n)log x -Σ
xn r=1

2
Example The 3rd order derivative of (log x)
(-1)3-1
  r (3- r)(r)
(3) 2 3
log x
2
= 2(3)log x -Σ
x3 r=1

x   1  2(1)(2)
1 3 3
= 22log x -
3
(2)(1)-
1
= (4log x -6)
x3

18.3.2 Higher Derivatives of log x sinx , log x cosx

Formula 18.3.2
n
 
(n)
(log xsin x) = log x  sin x +
2
(r) (n -r)
 r
n
 
n
+Σ(-1)r-1 sin x + (2.0s)
r=1 xr 2

-8-
 n
 
(n)
(log xcos x) = log xcos x +
2
(r) (n -r)
 r
n
 
n
+Σ(-1)r-1 cos x + (2.0c)
r=1 xr 2

Example The 3rd order derivative of log xsin x

3 (r) (3-r)
r x 
3
  
3
(log x sin x) = log x  sin x +
(3)
+Σ(-1)r-1 r sin x +
2 r=1 2
3 (1) 2 3 (2) 1
= -log x  cosx +
 1 x  2   2 x  2 
sin x + - 1 sin x + 2

3 (3) 0
+
 3 x  2 
sin x + 3

3 3 2
= -log x  cosx - 1 sin x
- 2 cosx
+ sin x
x x x3

18.3.3 Higher Derivatives of log x sinhx , log x coshx

Formula 18.3.3
(n) e x -(-1)-ne -x
log xsinh x = log x 
2
(r) e x -(-1)r-ne -x
 r
n
r-1 n
+Σ(-1) (3.0s)
r=1 xr 2
(n) e x +(-1)-ne -x
log xcosh x = log x 
2
(r) e x +(-1)r-ne -x
 r
n
r-1 n
+Σ(-1) (3.0c)
r=1 xr 2

Example The 4th order derivative of log xcosh x

e x+(-1)-4e
-x x r-4 -x
(r) e +(-1) e
r
4 4
(log x cosh x)
(4)
= log x  +Σ(-1) r-1
2 r=1 xr 2
(1) (2)
 1 x  2 x cosh x
4 4
= log x  cosh x + 1 sinh x - 2

4 (3) 4 (4)
+
 3 x sinh x -
 4 x cosh x
3 4

4 6 8 6
= log x  cosx + 1 sinh x
- 2 cosh x
+ 3 sinh x
- cosh x
x x x x4

-9-
18.4 Higher Derivative of e^x f (x)

18.4.1 Higher Derivative of e x x

Formula 18.4.1
(1+ ) - r

(n) n n
e x x  = e xΣ x for   -1, -2, -3,  (1.1)
r=0 r (1+ -r)
n (- + r) - r

n
= e xΣ(-1)r x for  = -1, -2, -3,  (1.2)
r=0 r (- )
Especially, when m = 0, 1, 2, 
(1+ m) m- r

(n) m n
e x x m = e xΣ x (1.1')
r=0 r (1+ m -r)

Proof
(n-r )
Substite f(x)= e x for Theorem 18.2.0 . Then since e x = ex, we obtain the desired expression
immediately.

Example1 The 2nd order derivative of e x x

(3/2) 2 - r
1

r
(2) 2 2
e x
x =e x
Σ
r=0 3/2- r 
x

  0 
1 1 3
(3/2) 2 (3/2) - 2 (3/2) - 2
 1  2
x
2 2 2
=e x + x + x
3/2 1/2 -1/2

= e x + x  = e x 1+ x - 4 x 
1 1 3
x 2
-
2 1 - x 1 1
- x 2 2
4

Example2 The 2nd order derivative of e x/x

(2)
(1+ r) 1- r
 
ex

2 2
= e xΣ(-1)r x
x r=0 r (1)
(1) -1 2 (2) -2 (3) -3
  0   2 
2 2
= ex x - x + x
(1) 1 (1) (1)
x

x  x 
e 2 2
= 1- + 2
x

18.4.2 Higher Derivative of e x log x

Formula 18.4.2
(r)
 r
(n) n n
e x log x = e x log x + e xΣ(-1)r-1 (2.1)
r=1 xr
Proof
Let f(x)= e x , g(x)= log x . Then

- 10 -
 (r)
(log x) r = (-1)r-1 r = 1, 2, 3, 
()
xr
Substituting this for Theorem 18.1.1 ,

 r 
(n) n n (r) n n
(r)
e x log x =Σ e x (log x) = e x (log x)(0) +Σe x (log x)
r=0 0 r=1
n (r)

n
= e x log x + e xΣ(-1)r-1
r=1 r xr

Example The 4th order derivative of e xlog x

(r)
r x
(4) 4 4
e x log x = e x log x + e xΣ(-1)r-1 r
r=1

4 (1) 4 (2) 4 (3) 4 (4)

x
= e log x + e
  1 x -  2 x +  3 x -  4 x 
x
1 2 3 4

x x x x 
x 4 6x 8 6
= e log x + e - + - 1 2 3 4

18.4.3 Higher Derivatives of e x sinx , e x cosx

Formula 18.4.3
 -n
n
   
(n)
e sin x
x
= sin e xsin x + (3.0s)
4 4
 -n
n
   
(n)
e cosx
x
= sin e xcos x + (3.0c)
4 4

Proof
"共立 数学公式" p187 was posted as it was.

Example
 -2
2
    = 2e cosx
(2)
e sin x
x
= sin e xsin x + x
4 4
 3-3
=  sin  e cos  x +
4 
(3)
e cosx
x x
= -2e (sin x + cosx) x
4

Higher Derivatives of e xsin x, e xcos x end now. There is no necessity for Theorem 18.1.1.
However, daring use Theorem 18.1.1, we obtain an interesting result.

Trigonometric Polynomial

Formula 18.4.3'
r  -n
n
 r 
n
    
n
Σ
r=0
sin x +
2
= sin
4
sin x +
4
(3.1s)

- 11 -
 r  -n n
 
n
    
n
Σr=0 r
cos x +
2
= sin
4
cos x +
4
(3.1c)

Especially, when x =0

r  -n n

n
 
n
Σr=0 r
sin
2
= sin
4
sin
4
(3.1's)

r  -n
n
Σ  r  cos 2 = sin 4 
n
n
cos (3.1'c)
r=0 4

Proof
Substitutingf(x)= e x , g(x)= sin x, cosx for Theorem 18.1.1 ,
r
 
n

(n) n
e x sin x = e xΣ sin x +
r=0 r 2
r
 
n

(n) n
e x cos x = e xΣ cos x +
r=0 r 2
And comparing these with Formula 18.4.3 , we obtain the desired expressions.

When n =5 , if both sides of (3.1s) are illustrated, it is as follows. Both overlap exactly and blue (left) can
not be seen.

Alternative Binomial Polynomial

r r
Removing sin , cos from (3.1's), (3.1'c), we obtain the following interesting polynomial.
2 2

Formula 18.4.3"
When 
denotes the floor function, the following expressions hold.

n
n

 2r +1 = 2 sin 4
(n -1)/2 n
Σ r 2
(-1) (3.2s)
r=0

n
n

Σ (-1)  2r 
n /2 n r 2
= 2 cos (3.2c)
r=0 4

- 12 -
Proof
Since the odd-numbered terms of the left side in (3.1's) are all 0,

r 0 1 2 n
 r    n 
n n n n n n
Σ
r=0
sin
2
=
0
sin
2
+
1
sin
2
+
2
sin
2
++ sin
2
2 
n-1

 
1 3 5 n
  
n n n
= sin + sin + sin + n-1 sin
2 
1 2 3 2 5 2 2

 
n
    
n n n (n -1)/2 n
= - + - n-1 = Σ (-1)r
1 3 5 2  r=0 2r +1
Also, since sin  /4 = 2n/2 in the right side in
-n
(3.1's) ,

n
n

 
(n -1)/2 n
 Σ (-1)r = 2 2 sin (3.2s)
r=0 2r +1 4
Next, since the even-numbered terms of the left side in (3.1'c) are all 0,

r 0 1 2 n
 r    n 
n n n n n n
Σ
r=0
cos
2
=
0
cos
2
+
1
cos
2
+
2
cos
2
++ cos
2
0 2 4 n /2
 0    n /2
n n n n
= cos + cos + cos + cos
2 2 2 4 2 2

      
n n n n n /2 n
= - + - = Σ (-1)r
0 2 4 n /2 r=0 2r
Also, since sin  /4 = 2n/2
-n
in the right side in (3.1'c) ,

n
n

 
n /2 n
 Σ (-1) r
= 2 cos 2
(3.2c)
r=0 2r 4
In addition, this formula is known. (See "岩波 数学公式Ⅱ" p11)

Note
When n = 4k -3 , k =1, 2, 3, 

   
(n -1)/2 n n /2 n
Σ
r=0
(-1)r
2r +1
= Σ
r=0
(-1)r
2r
Example
5
5

 1  3  5
5 5 5 2
- + = 5 - 10 + 1 = 2 sin = -4
4
5
5

 0  2  4 = 1 - 10 + 5 = 2
5 5 5 2
- + cos = -4
4

18.4.4 Higher Derivatives of e x sinhx , e x coshx

Formula 18.4.4
e x - (-1)-re -x
 r
(n) n n
e sinh x
x
=e x
Σ
r=0 2
(4.0s)

- 13 -
 e x + (-1)-re -x
 r
(n) n n
e cosh x
x
=e x
Σ
r=0 2
(4.0c)

Example
e x - (-1)-re -x
r
(0) 0 0
e sinh x
x
=e x
Σ
r=0 2
= e xsinh x
x -r -x

r
(3) 3 e + (-1) e
3
e cosh x
x
=e Σ x
r=0 2

  0  1  2  3 cosh x
x
3 3 3 3
=e sinh x + cosh x + sinh x +

= 4e x(sinh x + cosh x)

Note
The following formula is known for a natural number n .
(n) (n) n-1 x
e sinh x = e cosh x = 2
x x
e (sinh x + cosh x)
However, this formula does not hold for n =0 . That is, in this formula, the natural number n is inextensible
to the real number p . So, this is insufficient as a general formula.

- 14 -
18.5 Higher Derivative of f (x) / e^x

e x
-x
18.5.1 Higher Derivative of

Formula 18.5.1
(1+)
r 
-x 
n n
x  r for   -1, -2, -3, 
(n)
e x  Σ
-x -
=e (-1)-(n-r )
(1.1)
r=0 (1+ -r)
n (- + r)
Σ  r  (-) x
n
- r
for  = -1, -2, -3, 
-x
= (-1)-ne (1.2)
r=0

Especially, when m = 0, 1, 2, 
(1+ m) m- r
r 
(n) m n
e -x x m = e -xΣ(-1)-(n -r) x (1.1')
r=0 (1+m -r)

Proof
-x (n-r )
e 
-x -x
Substite f(x)= e for Theorem 18.2.0 . Then since = (-1)-(n-r)e , we obtain the
desired expression immediately.

-
Example1 The 3rd order derivative of e x/x
(3)
(1+r) -1-r
 
e -x
r 
3 3
= (-1)-3e -xΣ x
x r=0 (1)
(1) -1 3 (2) -2 3 (3) -3 3 (4) -4
  0    
-x
3
= -e x + x + x + x
(1) 1 (1) 2 (1) 3 (1)
e -x
 
3 6 6
=- 1+ + 2 + 3
x x x x
- 7
Example1' The 3rd order derivative of e x x
(8) 7- r
r 
(3) 7 3
e -x x 7 = e -xΣ(-1)-(3-r) x
r=0 (8-r)
(8) 7 3 (8) 6 3 (8) 5 3 (8) 4
  0    
3
= e -x - x + x - x + x
8 1 7 2 6 3 5
= e -xx 4-x 3 + 21 x 2 - 126 x 1 + 210

-x
18.5.2 Higher Derivative of e log x
Formula 18.5.2
(-1)- n (r)
 r  
(n) n n
e log x log x - Σ
-x
= (2.1)
ex r=1 xr

Proof
Let f(x)= e -x , g(x)= log x . Then

- 15 -
 -x (n -r)
e  = (-1)- n+re -x
(r)
(log x) r = (-1)r-1 r r = 1, 2, 3, 
()
x
Substituting these for Theorem 18.1.1 ,

r 
(n) n n
e -x log x (-1)- n+re -x(log x) r
()
=Σ
r=0

 0  r  (-1)
n n - n -x
n
- n+r -x r()
= (-1) e (log x) +Σ (0)
e (log x)
r=1

n (r)
-n
log x - Σ  
 x 
(-1) n
= x r
e r r=1

-
Example The 4th order derivative of e xlog x
(-1)- 4 (r)
 r  
(4) 4 4
e log x
x
= x log x - Σ
e r=1 xr
(1) (2) (3) (4)
  1 x  2 x  3 r 
log x 1 4 4 4 4
= - + + +
ex ex
1 2 3 4
x x

x x x x 
log x 1 4 6 8 6
= - + + +
ex ex 1 2 3 4

-x -x
18.5.3 Higher Derivatives of e sinx , e cosx
Formula 18.5.3
 -n
n
   
(n)
e sin x
-x
= -sin e -xsin x - (3.0s)
4 4
 -n
n
cos  x -
  4 
(n)
e cosx
-x
= -sin e -x (3.0c)
4

Proof
Replacing x with -x in Formula 18.4.3 , we obtain the desired expressions.

Example
2
 -2

    = -2e
(2)
e sin x
-x
= -sin e -xsin x - -x
cosx
4 4
 -3
3
    = -2e
(3)
e cosx
-x
= -sin e -xcos x - -x
(sin x - cosx)
4 4
-x -x
Higher Derivatives of e sinh x , e cosh x end now. There is no necessity for Theorem 18.1.1.
Daring use Theorem 18.1.1, we obtain the following expression first.
r
r  
n

(n) n
e -x sin x = e -xΣ(-1)n-r sin x +
r=0 2

- 16 -
And from this and (3.0s) , we obtain

r  -n
n
r  
n
    
n
Σ(-1)
-r
sin x + = sin sin x -
r=0 2 4 4
r r
A similar expression is obtained about e -xcosx too. Then removing sin
2
, cos from these, we obtain
2
the completely same results as Formula 18.4.3" .

-x -x
18.5.4 Higher Derivatives of e sinhx , e coshx

Formula 18.5.4
e x - (-1)-re -x
r 
(n) n n
e sinh x
-n+ r
Σ
-x -x
=e (-1) (4.0s)
r=0 2
x -r -x

r  2
(n) n n e + (-1) e
e cosh x
-n+ r
Σ
-x -x
=e (-1) (4.0c)
r=0

Proof
-x
Substituting f(x)= e , g(x)= sinh x , cosh x for Theorem 18.1.1 , we obtain the dsired expressions.

Example
e x - (-1)-re -x
r 
(0) 0 0
e sinh x
-0+ r
Σ
-x -x
=e (-1) = e -xsinh x
r=0 2
x -r -x

r  2
(3) 3 3 e + (-1) e
e cosh x
-x
=e -x
Σ (-1 ) -3+ r
r=0

 0 1 2 3 sinh x 

3 3 3 3
= e -x - cosh x + sinh x - cosh x +

= -4e -x(cosh x - sinh x)

Note
The following formula is known for a natural number n .
(n) (n)
e sinh x = e cosh x = (-2)n-1e x(cosh x - sinh x)
-x -x

However, this formula does not hold for n =0 . That is, in this formula, the natural number n is inextensible
to the real number p . So, this is insufficient as a general formula.

- 17 -
18.6 Higher Derivatives of sin x f (x), cos x f (x)

2 2
18.6.1 Higher Derivatives of sin x , cos x

Formula 18.6.1
n
 
(n)
sin x
2
= -2n-1cos 2x + (1.0s)
2
n
 
(n)
cos x
2
= 2n-1cos 2x + (1.0c)
2

Proof
From Formula 18.6.1' mentioned next ,
(n - r) r
 r 
n
  
(n) n
cos x =Σ
2
cos x + cos x +
r=0 2 2
Here
1
cosA cosB = cos(A +B )+cos(A -B)
21
Using this,

n n
  
1 n n
  
(n)
cos x
2
= Σ cos 2 x + + cos -r
21 r=0 r 2 2
n n n
  
n n

1 n 1
= 1 cos 2x +
2
Σ
2 r=0 r
+ 1 cos
2
Σ
2 r=0
(-1)r
r
And since

 r  r = 0
n n n n
Σ
r=0
= 2n , Σ
r=0
(-1)r
substituting these for the above, we obtain (1.0c). (1.0s) is also obtained in a similar way.

Example
2
 
(2)
sin x = 2 cos 2x = 2cos 2x - sin 2x
2
= -22-1cos 2x +
2
3
  = 4 sin 2x = 8 sin x cosx
(3)
cos x
2
= 23-1cos 2x +
2

Formula 18.6.1'
(n - r) r
 r 
n
  
(n) n
sin x =Σ
2
sin x + sin x + (1.1s)
r=0 2 2
(n - r) r
 r  2   2 
(n) n
n
cos x
2
=Σ cos x + cos x + (1.1c)
r=0

Proof
Substituting f(x)= g(x)= sin x for Theorem 18.1.1 , we obtain (1.1s). (1.1c) is also obtained in a similar
way.

- 18 -
Formula 18.6.1"
When  denotes the floor function, the following expressions hold.

 
n /2 n
Σ = 2n-1 (1.2e)
r=0 2r

Σ  2r +1 = 2
n
(n -1)/2
n-1
(1.2o)
r=0

Proof
(1.1s) is transformed as follows.
(n - r) r
 r 
n
  
(n) n
sin x
2
=Σ sin x + sin x +
r=0 2 2
(n -2r) 2r
 2r   2   2 
n
n /2
=Σ sin x + sin x +
r=0

(n -2r -1) (2r +1)

 2r +1 
n
  
(n -1)/2
+ Σ sin x + sin x +
r=0 2 2
(n -2r)
 2r   2  sin x
n /2 n
= Σ (-1) sin r
x +
r=0

(n -2r -1)
 2r +1 
n
 cosx
(n -1)/2
+ Σ (-1) sin x + r
r=0 2
i.e.

n n
   
n n
   
(n) n /2 (n -1)/2
sin x
2
= sin x + sin xΣ - cos x + cosx Σ
2 r=0 2r 2 r=0 2r +1
On the other hand, (1.0s) is transformed as follows too.
n n
   
(n)
sin x
2
= 2n-1sin x + n-1
sin x - 2 cos x + cosx
2 2
From these, the following expression follows .
n n
Σ      2r +1 - 2 
n /2 n (n -1)/2 n

sin x +
2  sin x
r=0 2r
- 2n-1 = cos x +  2  cos x Σ
r=0
n-1

In order to hold this equation for arbitrary x , the followings are necessary.

   2r +1 - 2
n /2 n (n -1)/2 n
Σ
r=0 2r
- 2n-1 = 0 , Σ
r=0
n-1
=0

In addition, this formula is known. (See "岩波 数学公式Ⅱ" p11)

3 3
18.6.2 Higher Derivatives of sin x , cos x
Formula 18.6.2
n 3n n
   
(n) 3
sin x
3
= sin x + - sin 3x + (2.0s)
4 2 4 2
n n n
= cosx +
2   2
(n) 3 3
cos x
3
+ cos 3x + (2.0c)
4 4

- 19 -
Proof
From Formula 18.6.2' mentioned next, it is obtained in a similar way in the case of the 2nd degree.
However, it is not so easy as the case of the 2nd degree. ( See 20.1.3 )

Example
2 32 2
   
(2) 3 3 9
sin x
3
= sin x + - sin 3x + =- sin x + sin 3x
4 2 4 2 4 4
3 3 3
= cosx +
2   2=
(3) 3 3 3 27
cos x
3
+ cos 3x + sin x + sin 3x
4 4 4 4

Formula 18.6.2'
n
 
(n) (0)
sin x = sin 2x sin x +
3
2
r (n -r)
 r
n
2 cos 2x + sin x + 
n
r-1
-Σ (2.1s)
r=1 2 2
n
= cos x cos x +
2 
(n) (0)
cos x
3 2

r (n -r)
 r
n
2 cos 2x + cosx + 
n
r-1
+Σ (2.1c)
r=1 2 2

Proof
2
Substituting f(x)= sin x , g(x)= sin x for Theorem 18.1.1, we obtain (2.1s). (2.1c) is also obtained in
a similar way.

Formula 18.6.2"
3n + (-1)n
 
n /2 n
Σ
2r-1
2 = (2.2e)
r=0 2r 4
n n

 2r +1
(n -1)/2 n 3 - (-1)
Σ
2r
2 = (2.2o)
r=0 4

Proof
From (2.1s)
n r (n -r) 
r 
n
     
(n) (0) n
sin x = sin x -Σ
3 2
sin x + 2r-1cos 2x + sin x +
2 r=1 2 2
n
 
(0)
= sin 2x sin x +
2
(2r +1) (n -2r -1)
 
n
   
(n -1)/2
- Σ
r=0 2r +1
22rcos 2x +
2
sin x +
2
2r (n -2r)
 
n
   
n /2
- Σ 22r-1cos 2x + sin x +
r=1 2r 2 2

- 20 -
 n
   
1
=- 2-1cos 2x sin x +
2 2
(n -2r -1)
 
n
 
(n -1)/2
+ Σ (-1)r 22rsin 2x sin x +
r=0 2r +1 2
(n -2r)
 
n
 
n /2
- Σ (-1)r 22r-1cos 2x sin x +
r=1 2r 2
n n
 
n
   
1 (n -1)/2
= sin x + - Σ 22rsin 2x cos x +
2 2 r=0 2r +1 2
n n
  
n n
   
n /2
- 2-1cos 2x sin x + -Σ 22r-1cos 2x sin x +
0 2 r=1 2r 2
n n
 
n
   
1 (n -1)/2
= sin x + - Σ 22rsin 2x cos x +
2 2 r=0 2r +1 2
n
 
n
 
n /2
-Σ 22r-1cos 2x sin x +
r=0 2r 2
When n =1 ,
(1)
sin x
3

1 1 1
  0   1
1 1
   
1
= sin x + - cos 2x sin x + 2-1 - sin 2x cos x + 20
2 2 2 2
1
 
1 1
= sin x+ - cos 2x cosx + 1 sin 2x sin x
2 2 2
1
 
1 1 1
= sin x+ + sin 2x sin x - (cos 2x cosx - sin 2x sin x)
2 2 2 2
1
 
1 1 1
= sin x+ + sin 2x sin x - cos 3x
2 2 2 2
1
 
1 1 1 1
= sin x+ - (cos 3x - cosx) - cos 3x
2 2 2 2 2
1 1 1
= sin x +
2   2
3 3
- sin 3x +
4 4
31-1 31+(-1)1
 2r 
1/2 1 1
 Σ 2 2r-1
= = =
r=0 4 2 4
1 1 1 1

 2r +1 4 2 4
(1-1)/2 1 3 -1 1 3 +1 3 -(-1)
 Σ 2 2r
= + = =1=
r=0 4
When n =2 ,
2 2
   0  2 
2 2
  
(2) 1
sin x
3
= sin x + - cos 2x sin x + 2-1 + 21
2 2 2
2
2   1
2
- sin 2x cosx + 2 0

- 21 -
 2
 
1 5
= sin x + + cos 2x sin x + 2 sin 2x cosx
2 2 2
2
 
1 1 1
= sin x + + (sin 3x - sin x) + 2sin 3x
2 2 2 2
2
 
3 9
= sin x + + sin 3x
4 2 4
2 32 2
= sin x +
2   
3
- sin 3x +
4 4 2
32-1 1 32+1 32+(-1)2
 2r 
2/2 2 5
 Σ 2 2r-1
= + = = =
r=0 4 2 2 4 4
2 2 2

 2r +1 4
(2-1)/2 2 3 -1 3 -(-1)
 Σ 2 2r
= =2=
r=0 4
Hereafter, by induction, we obtain the desired expressions.

If both sides of Formula18.6.2" are illustrated, it is as follows. The left side is blue line and the right side is
red point.

- 22 -
18.6.3 Higher Derivatives of the product of trigonometric and hyperbolic functions

Formula 18.6.3
(n - r) e x -(-1)-re -x
 
n

(n) n
(sinx  sinhx) =Σ sin x + (3.1)
r=0 r 2 2
(n - r) e x +(-1)-re -x
 
n

(n) n
(sinx  coshx) =Σ sin x + (3.2)
r=0 r 2 2
(n - r) e -(-1) e x -r -x

 r  2  2
(n) n n
(cosx  sinhx) =Σ cos x + (3.3)
r=0

(n - r) e +(-1) e x -r -x

 r  2  2
(n) n n
(cosx  coshx) =Σ cos x + (3.4)
r=0

Proof
Substituting f(x)= sin x , g(x)= sinh x for Theorem 18.1.1, we obtain (3.1).
The others are also obtained in a similar way.

Example
(0- r) e x +(-1)-0e -x
 
0

0
(sinx  coshx) = Σ = sinx  coshx
(0)
sin x +
r=0 r 2 2
(2- r) e x +(-1)-re -x
 
2

2
(cosx  coshx) = Σ
(2)
cos x +
r=0 r 2 2
2 1 0
= cos x +
2  
coshx +2cos x +
2 
sinhx + cos x + 
2
coshx  
= -cosx coshx - 2sin x sinhx + cosx coshx = -2sin x sinhx

- 23 -
18.7 Higher Derivatives of sinh x f (x), cosh x f (x)

2 2
18.7.1 Higher Derivatives of sinh x , cosh x

Formula 18.7.1
e x -(-1)-n+re -x e x -(-1)-re -x
 r
(n) n n
sinh x =Σ
2
(1.1s)
r=0 2 2
x -n+r -x
e x +(-1)-re -x
 r
(n) n n e +(-1) e
cosh x =Σ
2
(1.1c)
r=0 2 2

Proof
Substituting f(x)= g(x)= sinh x for Theorem 18.1.1, we obtain (1.1s). (1.1c) is also obtained in
a similar way.

Example
e x -(-1)-0+re -x e x -(-1)-re -x
r
(0) 0 0
sinh x =Σ
2
= sinh 2x
r=0 2 2
x -3+r -x
e x +(-1)-re -x
r
(3) 33 e +(-1) e
cosh x
2
=Σ
r=0 2 2

 0  1  2  3 coshxsinhx
3 3 3 3
= sinhxcoshx + coshxsinhx + sinhxcoshx +

= 8sinhxcoshx = 4sinh(2x)

2007.05.06
K. Kono
Alien's Mathematics

- 24 -
19 Super Derivative of the Product of Two Functions

19.1 Super Leibniz Rule

Theorem 19.1.1
Let  (x,y) be the beta function and p be a positive number. And for r =0, 1, 2,  , let f
<-p+ r >
be
(r )
arbitrary primitive function of f(x) and g be the r th order derivative function of g(x) .
Then the following expressions hold.

 r
(p) m -1 p (p -r)
(x)g r (x) + Rmp
()
f(x)g(x) =Σ f (1.1)
r=0

(-1)m  1
 
-p -1 (p)
 <m+ k > (m+ k )

(-p , m) Σ
Rmp
= f (x) g (x) (1.1r)
k =0 m + k k
Especially, when n =0, 1, 2, 


(n) n n n (r)
f(x)g (x) =Σ f ( -r)(x)g (x) ( Leibniz ) (1.1')
r=0 r

Proof
Theorem 16.1.2 (2.1) in 16.1.2 was as follows.

a a r
x x m -1 -n
 f <0>g(0)dx n = Σ f<
n + r> (r)
g
r=0
n 1

  a a
n -1 m -1 -n + r x x
- ΣΣ f a<
n -r+ s>
ga
(s)
 dx r
r=0 s=0 s n-r n-r
n n- r+1

a a
n -1 r -1 r -1 x x
t s  m +n-1 -r + tＣ m-1 f a 
m+ n -r+s (m+ s)
+ (-1)m ΣΣΣ Ｃ ga dx r
r =1 s =0 t=s n- r n- r
n n- r+1

a a f
(-1) n -1 n -1Ｃ k
m x x
 <m+ k > (m+ k )
(n , m) Σ
+ g dx n
k=0 m + k 1
n
When n =1 ∑∑∑ of the 3rd line does not exist, when
n =0 ∑∑ of the 2nd line does not exist also. And the
upper limit n -1 of ∑ of the 4th line can be replaced by  . Therefore, if the index n of the integration
operator is substituted for -n in consideration of these, it is as follows.

a a  r
x x m -1 n
 f <0>g(0)dx -n = Σ f <-n + r> (r)
g
r=0
n 1

  f
(-1)m  1

-n -1 x x
<m+ k > (m+ k )
(-n , m) Σ
+ g dx -n
k=0 m + k k an a1
Analytically continuing the index of the integration operator to [0 , p ] from[1 , n ] ,

a(p) a(0)  r
x x m -1 p
 f <0>g(0)dx -p = Σ g f <-p + r> (r)
r=0


(-p , m) Σ m + k  k  
m -p -1  x x
(-1) 1 m+ k > (m+ k )
+ f< g dx -p
k=0 a (p ) a (0)

-1-
 -p
Then, replacing the integration operators dx , <-p + r> with the differentiation operators (p ),(p -r)
respectivly, we obtain (1.1) and (1.1r) .
Especially, when p = m -1 , m =1, 2, 3,  , since  (-p , m) =  (1-m , m) =  , Rmp = 0 .
Then

 
(m-1)
m -1 m -1 m-1-r) (r)
f(x)g(x) = Σ
r=0 r
f( g
Furthermore, replacing m -1 with n , we obtain (1.1') .
Q.E.D

-2-
19.2 Super Derivative of x^a f (x)

Formula 19.2.0
Let (z) be the gamma function,  (n , m) be the beta function, f <r> be arbitrary the r th primitive
<r> <r>
function of f(x) and f a be the function values of f on a . Then the following expressions hold for
a positive number p .
(1)
(1+ ) - r (p -r)
 r
(p) m -1 p
x  f(x) =Σ x f
r=0 (1+ - r)
(1+ )
(p)
(-1)m  1
  
-p -1 - m- k <m+ k >
+ Σ
(-p , m) k =0 m + k k (1+ - m - k)
x f (0.1)

Especially, when m =0, 1, 2, 

(1+m) m- r (p -r)

(p) m p
x m f(x) = Σ x f (0.1')
r=0 r (1+ m - r)
Where, if  = -1, -2, -3,  , it shall read as follows.
(1+ ) (- + r)
 (-1)-r
1+ - r  (- )
(2) When   -1, -2, -3,  &  - p  -1, -2, -3, 

(1+ )

m -1 p
(p)  r (r)
x m f(x) = Σ x -p+ f
r=0 r (1+ - p + r)
(-1)m  1 (1+ ) (p)

  
-p -1 +m m+ k
(-p , m) Σ
+ x +k f ( ) (0.2)
k =0 m + k k (1+ + m + k)

Proof
Since differentiation is an inverse operation of integration, replacing the index p of the integration operator
with -p in Formula17.2.0 in 17.2 , we obtain the desired expressions.

19.2.1 Super Derivative of (ax+b)p(cx+d)q

Formula 19.2.1
The following expressions hold for p >0 , s >0 such that p - s  -1, -2, -3, 
q (s)
(ax +b)p(cx +d) 
(1/a)-s+ r (1+p)(1+q) (ax +b)p-s+r

m -1 s
=Σ + Rms (1.1)
r=0 r (1/c) (1+p -s + r)(1+q -r) (cx +d)
r r-q

(-1)m  1 m+ k
(1+p)(1+q)
  
-s -1 c
s
Rm = Σ
(-s, m) k=0 m + k k a (1+p +m + k)(1+q -m - k)
( s)
 (ax +b)p+m+ k (cx +d)q-m- k (1.1r)
lim R s = 0
m  m

-3-
Especially, when m =0, 1, 2, 
(s)
(ax +b)p (cx +d)m
(1/a)-s+ r (1+p)(1+ m) (ax +b)p-s+ r
r
m s
=Σ (1.1')
r=0 (1/c)r (1+p - s + r)(1+ m -r) (cx +d)r-m

Proof
Although it is an original way to substitute f(x)=(ax +b ) p , g(x)=(cx +d ) q for Theorem 19.1.1 ,
here, we reverse the sign of the index of the integration operator <s > in Formula 17.3.1 ( 17.3 ) and replace
it with the differentiation operator (s ) .

3
Example1 The 1/2th order derivative of x-2 3x+4
Substituting a =1, b =-2, p =1/2 , c =3, d =4, q =1/3 , s =1/2 for (1.1) ,

 
1
(3/2)(4/3)
1

r
2  1/2 -r
 x -2 3x +4
3
=Σ 3r
(x -2)r(3x +4) 3
r=0 (1+ r)(4/3-r)
The left side is calculated by the expression which replaced the order of integration and differentiation in
Riemann-Liouville differintegral. The integration lower limit is taken as -b /a = 2 according to Example1
in 17.3 . When m=17, the values of the both sides on arbitrary point x =5 are as follows.

3
Example1' The 1/2th order derivative of 3x+4(x-2)2
Substituting a =3, b =4, p =1/3 , c =1, d =-2, m =2 , s =1/2 for (1.1') ,
1
 
1 - +r
(4/3)(3)
1 6

r
2 2 1/2 -r (3x +4)
 3x +4(x -2)2
3
=Σ 3 2
r=0 (5/6+ r)(3-r) (x -2)r-2
The integration lower limit of Riemann-Liouville differintegral in the left side is taken as-b /a = -4/3
according to Example1' in 17.3.1 . The values of the both sides on arbitrary point x =3 are as follows.

Example2 The 1/2th order derivative of x-2 /(3x+4)

When q = -1, -2, -3, 

-4-
 (1+q) (-q + r)
= (-1)-r
1+q - r  (-q)
Then (1.1) can be read as follows.
q (s)
(ax +b) (cx +d) 
p

(1/a)-s+ r (1+p)(-q + r) (ax +b)p-s+ r


m -1 s
=Σ r (1+p -s + r)(-q ) r-q
+ Rms
r=0 r (-1/c) (cx +d)
Substituting a =1, b =-2, p =1/2 , c =3, d =4, q =-1 , s =1/2 for this ,

 2
1

(3/2)(1+ r) (x -2) r
  r
x -2  1/2
=Σ (-3) r
3x +4 r=0 (1+ r)(1) (3x +4)r+1
The integration lower limit of Riemann-Liouville differintegral in the left side is -b /a = 2 like Example1 .
When m=10, the values of the both sides on arbitrary point x =4 are as follows.

19.2.2 Super Derivative of x log x

Formula 19.2.2
When (z),  (z) denotes the gamma function and the digamma function respectively, the following
expressions hold.
(1)
log x -(1-p+r) -  (1+) -p
r 
 (p) m -1 p
x log x =Σ x + Rmp (2.1)
r=0 (1-p+r) (1+ -r)
log x -(1+m+k) -  (1+) x 
(p)

 
(-1)m

 1 -p -1
Rmp = Σ
 (-p , m) k =0 m + k k (1+m+k) (1+ - m - k)
(2.1r)
lim R p = 0
m  m

Especially, when m =0, 1, 2, 

p log x - (1- p+ r)-  (1+ m) m- p

(p) m
x mlog x = Σ x (2.1')
r=0 r (1- p+ r) (1+ m - r)
(2) When   -1, -2, -3,  &  - p  -1, -2, -3, 
(p) (1+ ) -p
x  log x = x log x
1+ - p 
(1+ )(r) -p

m -1 p
+ Σ (-1)r-1 x + Rmp (2.2)
r=1 r 1+ - p + r

-5-
 
x -p (1+ )  (1+ , m + k)
 
k-1 -p -1
(-p , m) 1+ - p  Σ
Rmp = (-1) (2.2r)
k =0 k m+k
lim R p = 0
m  m

Proof
Substituting f(x)= log x for Formula 19.2.0 , we obtain the desired expressions.

3/2
Example1 The 4/5 th order derivative of x log x
Substituting  =3/2 , p =4/5 for (2.1) ,
(4/5)

 
3 7 4
log x -(1/5 +r) -  (5/2)
 
m -1 4/5
=Σ
2 10 5
x log x x + Rm
r=0 r (1/5 +r) (5/2-r)
The integration lower limit of Riemann-Liouville differintegral in the left side is x =0 .
When m=120, the values of both sides on arbitrary point x =3 are as follows.

3
Example1' The 1.3 th order derivative of x log x
Substituting m =3 , p =1.3 for (2.1') ,

log x - (-0.3+ r)-  (4) 1.7

r
(1.3) 3 1.3
x log x
3
=Σ x
r=0 (-0.3+ r) (4- r)
The values of the both sides on arbitrary point x=2.1 are as follows.

The remainder in the super derivative of the product of two functions becomes a series of super derivatives.
So it is very difficult to calculate this. However, the calculation of the above (2.2r) is exceptionally easy.
So we show it as follows.

-6-
 3
Example2 The 1/2th order derivative of x log x
Substituting  =1/3 , p =1/2 for the above (2) ,

(4/3) - 6
1
 (p)
x log x = x log x
5/6
(4/3)(r) - 6
1 1

 
m -1
r-1
-1/2
+ Σ (-1) x + Rm2
r=1 r 5/6+ r 
1
-
(4/3)  (4/3 , m + k)
1 6

 
x k-1
-3/2
(-1/2, m) 5/6 Σ
Rm2 = (-1)
k =0 k m+k
When m=10, the values of both sides on arbitrary point x =1 are as follows.

Complete Automorphism
Reversing the sign of the index of the integration operator <p> in Formula 17.3.2 in 17.3 , we obtain the
following expression without the remainder term. However, it is complicated and the convergence is also very
slow.

Formula 19.2.2'
(p)
x  log x
log x -(1-p) -  (1+)  -p -1 (2+k) + 

=
(1-p)
-p Σ
 1+ - p  k =0 k 
(1+ k)  (1+ k,  - k) -p
2
x

 
 -p -1 1
1 - pΣ
(1+ k)  (1+ k,  - k)
2
k=0 k

19.2.3 Super Derivatives of x sinx , x cosx

Formula 19.2.3
When m =0, 1, 2,  , the following expressions hold for p >0 .

-7-
 (1+ m) m- r (p -r)
 r
p
 
(p) m
x m sin x =Σ x sin x + (3.1s)
r=0 (1+ m -r) 2
(1+ m) m- r (p -r)
 r
p
 
(p) m
x m cosx =Σ x cos x + (3.1c)
r=0 (1+ m -r) 2

Proof
Reversing the sign of the index of the integration operator <p> in Formula 17.3.3 in 17.3 and replacing it
with the differentiation operator (p ), we obtain the desired expressions.

3
Example The 3/2th order derivative of x sinx
Substituting m =2 , p =3/2 for (3.1s) ,

 2
3
(4) 3-r (3/2- r)
r
3/2
 
3
x 3 sin x =Σ x sin x +
r=0 (4- r) 2
This super derivative can not be examined by Riemann-Liouville differntegral. Then we draw this with the 1st and
the 2nd order derivative on a figure side by side. It is as follows.

Formula 19.2.3' (Collateral Super Derivative)

The following expressions hold for , p such that  -1, -2, -3,  &  - p -1, -2, -3,  .
(1+ ) r
 r
p
  +R
m -1
{p} 
x  sin x
r
=Σ x -p+ sin x + p
(3.2s)
r=0 (1+ -p + r) 2 m

(p)
(-1)m (1+ ) m + k  
 
-p -1
 
 1
Rmp = Σ x +m+ k sin x +
(-p , m) k =0 m + k k 1+ + m +k  2
(1+ ) r
 r
p
  +R
m -1
{p} 
x  cosx
r
=Σ x -p+ cos x + p
(3.2c)
r=0 (1+ -p + r) 2 m

(p)
(-1)m (1+ ) m + k  
 
-p -1
 
 1
Rmp = Σ x +m+ k cos x +
(-p , m) k =0 m + k k 1+ + m +k  2
lim Rmp = 0
m 

-8-
Proof
Reversing the sign of the index of the integration operator <p> in Formula 17.3.3' in 17.3 and replacing it
with the differentiation operator (p ), we obtain the desired expressions.

3/2
Example Collateral the 1 / 2th order derivative of x sinx
Substituting  =3/2 , p =1/2 for (3.2s), we obtain

 2
1

(5/2) 1-r r
1
x sin x
3

r
1/2
 +
m -1
2
=Σ x sin x + Rm2
r=0 (2- r) 2
The integration lower limit of Riemann-Liouville differintegral in the left side is x =0 .
When m=15, the values of the both sides on arbitrary point x =5 are as follows.

19.2.4 Super Derivatives of x sinhx , x coshx

Formula 19.2.4
When m =0, 1, 2,  , the following expressions hold for p >0 .
(1+ m) m- r e x -(-1)-p+ re -x
 r
{p} m p
x sinh x
m
=Σ x (4.1s)
r=0 (1+ m - r) 2
(1+ m) e x +(-1)-p+ e -x
r

 r  (1+ m - r) x
{p} m p m- r
x cosh x
m
=Σ (4.1c)
r=0 2

3
Example The 0.999th order derivaive of x coshx
Substituting m =3 , p =0.999 for (4.1c), we obtain

(4) 3- r e x +(-1)-0.999+ re -x
 
{0.999} 3 0.999
x cosh x
3
=Σ x
r=0 r (4- r) 2
This super derivative can not be examined by Riemann-Liouville differntegral. Then we calculate the values on
arbitrary point x=2 for the 0.999th order derivative and the 1st order derivative respectively. Although the former
is a complex number, the real part is naturally near to the coefficient of the later.

-9-
Formula 19.2.4' (Collateral Super Derivative)
The following expressions hold for , p such that  -1, -2, -3,  &  - p -1, -2, -3,  .
(1+ ) x -r -x

 r
 {p} m -1 p -p+ r e -(-1) e
x sinh x =Σ x + Rmp (4.2s)
r=0 1+ -p + r  2
x (p)

 
-x
(1+ )
-m-k
(-1)m

 1 -p -1 +m+ k e -(-1) e
Rmp = Σ x
(-p , m) k =0 m + k k 1+ + m + k  2
(1+ ) x -r -x

 r
 {p} m -1 p -p+ r e +(-1) e
x cosh x =Σ x + Rmp (4.2c)
r=0 1+ -p + r 2
x (p)

 
-x
(1+ )
-m-k
(-1)m

 1 -p -1 +m+ k e +(-1) e
Rmp = Σ x
 p , m k =0 m + k
(- ) k 1+ + m + k  2
lim Rmn = 0
m 

3
Example Collateral the 3 / 2th order integral of x sinhx
Substituting  =1/3 , p =3/2 for (4.2s), we obtain

 
3
(4/3)
1
e x -(-1)-re -x
 r   -1/6+ r
2 m -1 3/2 - +r
 x sinh x
3
=Σ x 6
+ Rm3/2
r=0   2
The integration lower limit of Riemann-Liouville differintegral in the left side is x =0 .
When m=50, the values of the both sides on arbitrary point x =17 are as follows.

- 10 -
19.3 Super Derivative of log x f (x)

19.3.1 Super Derivative of (log x)2

Formula 19.3.1
(p) x -p log x log x - 1- p - 
log x
2
=
1-p
p log x - (1- p +r)-(r)

m -1
- x -p Σ (-1)r + Rmp (1.1)
r=1 r 1-p + r
-p
x
Rmp = 
 (-p , m)  1-p 
(-1) k-1
k =0 (m+ k )  
 -p -1
Σ 2 log x - 1-p  -(1+ m+ k ) -2 (1.1r)
k
lim Rmp = 0
m 

Proof
Reversing the sign of the index of the integration operator <p> in Formula 17.4.1 in 17.4 and replacing it
with the differentiation operator (p ), we obtain the desired expressions.

Example The 1/2th order derivative of (log x)2

When m=4000, the values of both sides on arbitrary point x =2.7 are as follows. Since the convergence
is slow, even if it calculates so far, both sides does not match only up to 3 digits after the decimal point.

- 11 -
19.4 Super Derivative of e^x f (x)

19.4.1 Super Derivative of e x x

Formula 19.4.1
(1+ ) - r

(p) m -1 p
e x x  = e xΣ x + Rmp (1.1)
r=0 r  1+ - r 
(-1)m  1 (1+ )
 
-p -1  m-k (p)
p
Rm = Σ e x x -  (1.1r)
(-p , m) k =0 m + k k  1+ - m -k
Especially, when m =0, 1, 2, 
(1+m) m- r

(p) m p
e x x m = e xΣ x (1.1')
r=0 r (1+ m - r)

Proof
Let f(x)= e x , g(x)= x  . Then
(r) (1+ ) - r
x   = x ( -1, -2, -3, )
1+ - r 
Substituting these for Theorem 19.1.1 , we obtain (1.1) and (1.1r).
Especially, when  = m-1 m =1, 2, 3,  , (1+ - m - k) =  k =0, 1, 2, 3,  .
Then, the remainder term disappears and (1.1) is as follows.
(1+m -1) m-1- r
r 
m-1 (p)
m -1 p
e x x  = e xΣ x
r=0 1+m -1- r 
Thus, replacing m -1 with m , we obtain (1.1').

3/4
Example1 The 1/2.1th order integral of ex x
Substituting  =3/4 , p =1/2 for (1.1) ,

 2
1
(1+3/4) 4 - r
3

r
m -1 1/2
e x
x
 x
Σ
3/4
=e x + Rm1/2
r=0  1+3/4- r 
The left side is calculated by the expression which replaced an order of integration and differentiation in
Riemann-Liouville differintegral. When m=10, the values of the both sides on arbitrary point x =10 are
as follows.

7
Example1' The 5/2th order derivative of ex x
Substituting m =7 , p =5/2 for (1.1') ,

- 12 -
  2
5
(8) 7- r
r
7 5/2
e x x 
7
= e xΣ x
r=0 8- r 
The values of the both sides on arbitrary point x = .5 are as follows.

Formula 19.4.1" ( Collateral Super Derivative)

The following expression holds for , p such that  -1, -2, -3,  &  - p -1, -2, -3,  .
(1+ )

(p) m -1 p 
e x x 
r
= ex Σ x -p+ + Rmp (1.1")
r=0 r 1+ - p + r 
(-1)m  1 (1+ )
 
-p -1 x + m+k p
()
 
(-p , m) Σ
p
Rm = e x
k =0 m + k k  1+ + m +k
lim Rmp = 0
m 

3/4
Example1" Collateral the 1 / 2th order derivative of ex x
Substituting  =3/4 , p =1/2 for (1.1") , we obtain

 2
1
(3/2) 4 + r
3

r
m -1 1/2
e x
x
 x
Σ
3/4
=e x + Rm1/2
r=0 3/4+ r 
The values of both sides on the same point x =-10 as Example1 are as follows. Though the both sides are
corresponding, they are considerably different from the values of Example1 (Lineal the 1/2th order derivative.)

19.4.2 Super derivative of e x log x

All the polynomials obtained by applying Theorem 19.1.1 to e xlog x become asymptotic expansions, and
they are hardly helpful.

19.4.3 Super derivatives of e x sinx , e x cosx

- 13 -
Formula 19.4.3
 -p
p
   
(p)
e sin x
x
= sin e xsin x + (3.0s)
4 4
 -p
p
   
(p)
e cosx
x
= sin e xcos x + (3.0c)
4 4

Proof
Analytically continuing the index of the differentiation operator in Formula 18.4.3 in 18.4 to [0 , p ] from
[1 , n] , we obtain the desired expressions.

Example The 1/3 th order derivative of e x sinx

When p =1/3 , the values of the both sides of (3.0s) on arbitrary point x =1.2 are as follows.

Trigonometric Series
If Formula 18.4.3' in 18.4 is extended to the real number, the following trigonometric series are obtained.

Formula 19.4.3'
r  -p p
 
p
    

Σr=0 r
sin x +
2
= sin
4
sin x +
4
(3.1s)

r  -p p
 
p
    

Σr=0 r
cos x +
2
= sin
4
cos x +
4
(3.1c)

Especially, when x =0 ,

r  -p p

p
 

Σr=0 r
sin
2
= sin
4
sin
4
(3.1's)

r  -p
p
Σ  r  cos 2 = sin 4 
p
n
cos (3.1'c)
r=0 4

Alternating Binomial Series

r r
Removing sin , cos from (3.1's), (3.1'c), we obtain the following interesting series.
2 2

Formula 19.4.3"
p
p

 
 p
Σ k 2
(-1) = 2 sin (3.2s)
k=0 2k +1 4

- 14 -
 p
p

 
 p
Σ k 2
(-1) = 2 cos (3.2c)
k=0 2k 4

Proof
Since the odd-numbered terms of the left side in (3.1's) are all 0,

r
r       
 p p p p p  p
Σ
r=0
sin
2
=
1
-
3
+
5
-
7
+- = Σ(-1)k
k=0 2k +1
p p
p
 -p

  = sin 4 
 p
 Σ (-1) k
sin = 2 2 sin
k=0 2k +1 4 4
Next, since the even-numbered terms of the left side in (3.1'c) are all 0,

r
r       
n p p p p p  p
Σ
r=0
cos
2
=
0
-
2
+
4
-
6
+- = Σ(-1)k
k =0 2k
p p
p
 -p

  = sin 4 
 p
 Σ (-1) k
cos = 2 2 cos
k=0 2k 4 4

If a horizontal axis is set as p and these are illustrated , it is as follows. Although the left side is blue and
right side is red, since both sides overlap exactly, the left side (blue) is not visible.

19.4.4 Super Derivatives of e x sinhx , e x coshx

Formula 19.4.4
e x - (-1)-re -x
r 
(p)  p
e sinh x
x
=e x
Σ (4.0s)
r=0 2
x -r -x

r  2
(p) p e + (-1)
 e
e cosh x
x
=e Σ x
(4.0c)
r=0

- 15 -
Example The 3/2th order derivative of e x coshx
When p =3/2 , the values of the both sides of (4.0c) on arbitrary point x =1.3 are as follows.

- 16 -
19.5 Super Derivative of f (x) / e^x

e x
-x
19.5.1 Super Derivative of

Formula 19.5.1
(-1)-p (1+ ) - r
 r
 (p)
m -1 p
e x 
r
Σ + Rmp
-x
= (-1) x (1.1)
ex r=0 1+ - r
(p)
(1+) x
 
 (-1)k - m-k

 
1 -p -1
Rmp
 (-p , m) Σ
= (1.1r)
k=0 m+ k k  1+ - m -k  ex
Especially, when m =0, 1, 2, 
(-1)-p (1+ m) m- r
 r
m (p)
m p
e x 
r
Σ(-1)
-x
= x (1.1')
e x r=0 1+ m - r 
-x
Example1 The 1/2th order derivative of e x
If we substitute  =1/2 , p =1/2 for (1.1) and calculate the values of the both sides on arbitrary point
x =10 , it is as follows. In addition, this is an asymptotic expansion.

-x 7
Example1' The 5/2th order derivative of e x
If we substitute m =7 , p =5/2 for (1.1') and calculate the values of the both sides on arbitrary point
x =6 , it is as follows.

Formula 19.5.1" ( Collateral Super Derivative)

The following expression holds for , p such that  -1, -2, -3,  &  - p -1, -2, -3,  .
(1+ )
 r
(p) 1 m -1 p 
e x  (-1)- r r
xΣ x -p+ + Rmp
-x
= (1.1")
e r=0 1+ - p + r 

- 17 -
 (p)
(1+) x
 
 (-1)-k - m-k

 
1 -p -1
Rmp
 (-p , m) Σ
=
k=0 m + k k  1+ + m +k  ex
lim Rmp = 0
m 

-x
Example1" Collateral the 1 / 2th order derivative of e x
If we substitute  =1/2, p =1/2, m =30 for (1.1") and calculate the values of the both sides on arbitrary
point x =10 , it is as follows. Though the both sides are corresponding, they are completely different from
the values of Example1 (Lineal the 1/2th order derivative.)

-x
19.5.2 Super Derivative of e log x
All the polynomials obtained by applying Theorem 19.1.1 to e -xlog x become asymptotic expansions, and
they are hardly helpful.

-x -x
19.5.3 Super derivatives of e sinx , e cosx

Formula 19.5.3
 -p
p
   
(p)
e sin x
-x -p
= (-1) sin e -xsin x - (3.0s)
4 4
 -p
p
   
(p)
e cosx
-x
= (-1)-p sin e -xcos x - (3.0c)
4 4

-x
Example The 3/2th order derivative of e sinx
The values of the both sides of (3.0s) on arbitrary point x =1.7 are as follows.

-x -x
19.5.4 Super derivatives of e sinhx , e coshx

- 18 -
Formula 19.5.4
e x - (-1)-re -x
 r
(p) -x 
p
e sinh x
-x
=e Σ (-1) -p+ r
(4.0s)
r=0 2
x -r -x

 r
(p) -x 
p e + (-1) e
e cosh x
-p+ r
Σ
-x
=e (-1) (4.0c)
r=0 2

-x
Example The 3/2th order derivative of e coshx
The values of the both sides of (4.0c) on arbitrary point x =1.3 are as follows.

- 19 -
19.6 Super Derivatives of sin x f (x), cos x f (x)

2 2
19.6.1 Super Derivatives of sin x , cos x

Formula 19.6.1
p
 
(p)
sin x
2
= -2p-1cos 2x + (1.0s)
2
p
 
(p)
cos x
2
= 2p-1cos 2x + (1.0c)
2

Proof
Analytically continuing the index of the differentiation operator in Formula 18.6.1 in 18.6.1 to [0 , p ] from
[1 , n] , we obtain the desired expressions.

Example

 2
1 1

 
-1 1
sin x
2 2
= -2 cos 2x + = (sin 2x + cos 2x)
4 2
 2
3
3
3

  = -(sin 2x + cos 2x)

-1
cos x
2 2
= 2 cos 2x +
4

Analytically continuing the index of the differentiation operator in Formula 18.6.1' and Formula 18.6.1" in 18.6
to [0 ,p ] from[1 ,n ] respectively, we obtain the following two formualas.

Formula 19.6.1'
(p - r) r
 r 
p
  
(p) 
sin x =Σ
2
sin x + sin x + (1.1s)
r=0 2 2
(p - r) r
 r  2   2 
(p)  p
cos x =Σ
2
cos x + cos x + (1.1c)
r=0

Formula 19.6.1"

  =2
 p p-1
Σ
r=0 2r
p >0 (1.2e)

Σ  =2
 p p-1
p >0 (1.2o)
r=0 2r +1

If a horizontal axis is set as p and these are illustrated , it is as follows. (1.2e) , (1.2o) and 2p-1 are blue,
p-1
red and green respectively. Since three curves overlap exactly, only red ( 2 ) is visible.

- 20 -
 The following formula follows from Formula 19.6.1 immediately.

Formula 19.6.1"'

 r = 2
 p
Σ
r=0
p
p >0 (1.3)

Note
In fact, Formula 19.6.1" and Formula 19.6.1"' hold for p > -1 .

3 3
19.6.2 Super Derivatives of sin x , cos x
Formula 19.6.2
p 3p p
   
(p) 3
sin x
3
= sin x + - sin 3x + (2.0s)
4 2 4 2
p p p
= cosx +
2   2
(p) 3 3
cos x
3
+ cos 3x + (2.0c)
4 4

Example

 2
1
 
   
3 3
sin x
3
= sin x + - sin 3x +
4 4 4 4
 2
3
3 3
   
3 3 3
cos x
3
= cos x + + cos 3x +
4 4 4 4
If the later is drawn with the 1st and the 2nd order derivatives on a figure side by side, it is as follows.

- 21 -
19.6.3 Super Derivatives of the product of trigonometric and hyperbolic functions

Formula 19.6.3
(p - r) e x -(-1)-re -x
 
p

(p) 
(sinx  sinhx) =Σ sin x + (3.1)
r=0 r 2 2
(p - r) e x +(-1)-re -x
 
p

(p) 
(sinx  coshx) =Σ sin x + (3.2)
r=0 r 2 2
(p - r) e -(-1) e x -r -x

 r  2  2
(p) p
(cosx  sinhx) =Σ cos x + (3.3)
r=0

(p - r) e +(-1) e x -r -x

 r  2  2
(p) p
(cosx  coshx) =Σ cos x + (3.4)
r=0

Example The 3/2th order derivative of cosx sinhx

(3/2- r) e x -(-1)-re -x
  
3/2


(cosx  sinhx)
(3/2)
=Σ cos x +
r=0 r 2 2

If this is drawn with the 1st and the 2nd order derivatives on a figure side by side, it is as follows.

10

1 2 3 4

-5

-10

- 22 -
19.7 Super Derivatives of sinh x f (x), cosh x f (x)

2 2
19.7.1 Super Derivatives of sinh x , cosh x

Formula 19.7.1
e x -(-1)-p+re -x e x -(-1)-re -x
r 
(p)  p
sinh x
2
=Σ (1.1s)
r=0 2 2
x -p+r -x
e x +(-1)-re -x
r 
(p) p e +(-1) e
cosh x =Σ
2
(1.1c)
r=0 2 2

Proof
Analytically continuing the index of the differentiation operator in Formula 18.7.1 in 18.7 to [0 ,p ] from
[1 , n], we obtain the desired expressions.

2
Example: The 0.01th and the 0.99th order derivatives of sinh x
According to (1.1s), if each differential coefficient of the 0th order, the 0.01th order, the 0.99th order and
the 1st order on arbitrary point x =2 are calculated , they are as follows. Naturally, the super differential
coefficients turn into complex numbers.

2010.11.11
K. Kono
Alien's Mathematics

- 23 -
20 Higher Calculus of the product of many functions

20.1 Higher Derivative of the product of many functions

(1) Binomial Theorem and Leibniz Rule

According to the binomial theorem in 3.1 , the following expression holds for real numbers x1 , x2 and natural

number n .

 r x
n n n-r r
x1+ x2 = Σ
n
1 x2
r=0

On the other hand, according to the Leibniz Rule, the following expression holds for functions f1 , f2 of x and

natural number n .

 r
(n) n n n -r) (r)
f1 f2 =Σ f1( f2
r=0

(2) Multinomial Theorem and Higher Derivative of the product of many functions
According to the multiomial theorem in 3.3 , the following expression holds for real numbers x 1 , x 2 ,  , x
and natural number n .

 r  r  r 
n r1 r-2 n r1 r-2 n-r1 r -r2 r
x1+ x2++ x =Σ Σ  Σ   x-1
n
x1 x21
r1=0 r 2=0 r -1=0 1 2 -1

Therefore, the following expression must hold for functions f1 , f2 ,  , f of x and natural number n .

 r  r  r 
n r1 r -2 n r1 r-2
f1
n - r1 r1 -r2
 f -1
r
f1 f2  f = ΣΣ Σ 
(n)
f2
r1=0 r 2=0 r -1=0 1 2 -1

20.1.1 Higher Derivative of the product of many functions

Theorem 20.1.1
When fk
(r)
r th order derivative function of fk (x) (k =1, 2, , ) ,
denotes the

    
n r1 r-2 n r1 r-2
f1 1f2 1 2 f -1
n- r r -r r
  
(n)
 1 2 
f f f =Σ Σ Σ
r 1=0 r2=0 r-1=0 r1 r2 r-1

Proof
According to Theorem 18.1.1 (Leibniz) in 18.1 , the following expressions hold.

r 
n n r1
f1
(n) n -r1
f1 f2 f3 f4  f =Σ
r1=0
f2 f3 f4  f (1)
1

r 
r1
r1 r1 r2
f2 1
r -r2
f2 f3 f4  f =Σ
r2=0
f3 f4  f (2)
2

r  f
r2
r r2 2 r3
r2-r3
f3 f4  f =Σ
r3=0
3 f4  f (3)
3


-1-
 r 
r-3 r-3
 -3
f-2-3
r r - r-1 r-2
f-2 f-1 f = Σ f-1 f ( -2)
r =0
-2 -2

r  f
r r-2 -2
 -2 r r-2- r-1 r-1
f-1 f = Σ -1 f ( -1)
r-1=0 -1
Substituting (2), (3),  , ( -2), ( -1) for (1) one by one, we obtain the desired expression.

Example

 r  s f f
n r n r (n - r) (r-s) (s)
f1 f2 f3
(n)
= ΣΣ 1 2 f3
r =0 s =0

 r  s  t  f
n n r r s s
(n - r) (r-s) (s -t) (t)
f1 f2 f3 f4
(n)
= ΣΣΣ 1 f2 f3 f4
r =0 s =0 t=0

Example computation of the product of three functions

Since the combination of the product of many functions are numerous, we cannot calculate these one by one.
Then we pick up some and calculate them.

Example1 Higher Derivative of x e x sinx

Let f1 = x  , f2 = e x , f3 = sin x . Then

(n -r) (1+)
x  = x -n+ r
 1+ - n + r 
x (r-s)
e  = ex
s
 
(s)
(sin x) = sin x +
2
Substituting these for Theorem 20.1.1 , we obtain
(1+) s
 r  s 
n r n r
 
 x (n)
x e sin x =Σ Σ x -n+ r e x sin x + (1.1)
r =0 s =0  1+ - n + r  2

Although this formula is troublesome in manual calculation, it is easy in mathematical software. When the 4th
order derivative of x e xsin x is calculated using the mathematical software MuPad, it is as follows.

Diffe re ntiation by dire ct calculation

 fl := diff(x^a*E^x*sin(x) ,x,x,x,x):
 simplify(fl)

Diffe re ntiation by the formula

 n:=4:
 delete r: fr:=0:
 for r from 0 to n do
fs:=0:
for s from 0 to r do

-2-
 fs := fs + binomial(n,r)*binomial(r,s)
*gamma(1+a)/gamma(1+a-n+r)*x^(a-n+r)
*E^x*sin(x+s*PI/2)
end_for:
fr:=fr+fs
end_for:
 fr := expand(fr):
 simplify(fr)

Verification of the e quivalence

 testeq(fl,fr)
TRUE

Example2 Higher Derivative of x e x log x

Let f1 = x  , f2 = e x , f3 = log x . Then

(n -r) (1+ )  (r-s)

x 
r
= x -n+ , e x = ex
1+ - n + r 
(log x) s = (-1)s-1(s -1)! x -s (s =1, 2, 3, )
()
(log x)(0) = log x ,
(0)
Separating the terms containing f3 from Theorem20.1.1 , we obtain

 r  r  s f
n n n - r) (r-s) (0)
n r n r (n - r) (r-s) (s)
f1 f2 f3 = Σ Σ Σ
(n)
f1( f2 f3 + 1 f2 f3
r =0 r =1 s =1
Substituting the abave expressions for this, we obtain
(1+ )

n n 
x  e x log x
(n) r
= Σ
r =0 r  1+  - n + r 
x -n+ e x log x
(1+ )(s) -n+ r-s x
  
n r n r
+ ΣΣ (-1)s-1 x e (1.2)
r =1 s =1 r s 1+ - n + r 
When n =1
(1+ )

1 1 
x  e x log x
(1) r
= Σ
r =0 r 1+ -1+ r 
x -1+ e x log x

(1+ )(s) -1+ r-s x

  
1 r 1 r
+Σ Σ (-1)s-1 x e
r =1 s =1 r s 1+ -1+ r 
(1+ ) -1 1 (1+ )  x
   
1
= x + x e log x
0 1+ -1 1 1+ 
(1+ )(1) -1 x
  
1 1 1 1
+Σ Σ (-1)1-1 x e
r =1 s =1 1 1 1+ 
=  x -1 + x  e x log x + x -1 e x
When n =2

-3-
 (1+ )

2 2 
x  e x log x
(2) r
= Σ r 1+ -2+ r 
r =0
x -2+ e x log x

(1+ )(s) -2+ r-s x

  
2 r 2 r
+Σ Σ (-1)s-1 x e
r =1 s =1 r s 1+ -2+ r 
(1+) -2 2 (1+) -1 2 (1+)  x
    
2
= x + x + x e log x
0  1+ -2 1  1+ -1 2  1+ 
1 (1+) (1) -2 x

2
+ x e
1 1  1+ -1
2 (1+) (1) -1 x 2 (1+) (2) -2 x
     
2
+ x e - x e
2 1  1+  2  1+ 
=  ( -1)x -2 + 2 x -1 + x e x log x
+ 2 x -2 + 2x -1 - x -2e x

20.1.2 Higher Derivative of the power of a function

Especially, when f1 = f2 =  = f in Theorem 20.1.1 , the following theorem follows immediately.

Theorem 20.1.2
When f
(r)
r th order derivativefunction of f(x) and  is a natural number,
denotes the

    
n r1 r-2 n r1 r-2
f (x) =Σ Σ  Σ f  1 f  1 2 f  -1
(n) n- r r -r r

r 1=0 r2=0 r-1=0 r1 r2 r-1

Example1 Higher Derivative of e x
x (r)
Since e  = ex, immediately from the theorem,

  r  r 
r1 r -2 n r1 r-2
x  x 
(n) n
e   = e  Σ Σ  Σ 
r1=0 r 2=0 r-1=0 r1 2 -1
Here, from (1.1") in 3.3 ,
n r3 r-2

Σ Σ  Σ nＣ r1 r1Ｃ r2 r -2Ｃ r -1 =  n  

r =0 r =0 r
1 2=0 -1

Substituting this for the above, we obtain

x 
(n)
n x
e   =  e (2.1)

3
Example2 Higher Derivative of log x
Let f = log x . Then

(r =1, 2, 3, )
(r)
f(0) = log x , f = (-1)r-1(r -1)! x -r
r =1, 2, 
(0) (r ) (0)
Since f is different from f in function type, we have to separate the terms containing f
from

 r  s 
(n) n r n r
f 
3 n -r) (r-s) (s)
= ΣΣ f( f f
r =0 s =0
It is as follows.

-4-
  0
n
r =0  r  s =0  s  s =0  s 
n n n -r)
r r r-s) (s) (n)
0 0 (0-s) (s)
Σ f( Σ f( f = f Σ f f

n 
n
r  s  f f s =0  s 
n n-1 r (n -r)
r
(r-s) (s) n -n )
n n (n -s) (s)
+Σ f Σ + f( Σ f f
r =1 s =0

f + Σ f s  f s  f
n r
n-1 r n n
= f(n)f(0) (0) Σ
(n -r) (r-s) (s)
f + f(0)Σ (n -s) (s)
f (w)
r r =1 s =0 s =0
Here,

r =1  r  s =0  s 
n-1 n (n -r)
r r (r-s) (s)
Σ f Σ f f

 1
n
s =0  s  r  s =0  s 
n -1)
1 1 (1-s) (s)
n-1 n n -r)
r r (r-s) (s)
= f( Σ f f +Σ
r =2
f( Σ f f

=  f
  0  1 f f 
n (n -1) 1 1
f(1-0)f(0)+ (1-1) (1)
1

 r    0 +  f

r
s 
n n-1 r (n -r) r (r-0) (0)
r -1
(r-s) (s) (r-r) (r)
+Σ f f f +Σ f f f
r =2 r s =1

= 2f   f
n
f + 2f Σ   f f +Σ Σ     f
(0) (n -1) (1)n n
(0) r
n-1
(n -r) (r)
n-1 r -1
(n -r) (r-s) (s)
f f
1 r r s
r =2 r =2 s =1

r   r  s  f f f
n n-1 n r
(n -r) (r)
n-1 r -1
(n -r) (r-s) (s)
= 2f Σ (0)
f f + ΣΣ
r =1 r =2 s =1

  0 Σs  n  f 
n n
Σs 
n n (n -s) (s) n (n -0) (0)
n-1
(n -s) (s) (n -n ) (n)
f(0) f f = f f f + (0)
f f + f
s =0 s =1

= f Σ f
n (0)
n-1
(n -s) (s) (0) (0) (n)
f + 2f f f
s s =1
Substitute these for (w) ,

r =0 s =0  r   s 
n r n r (n -r) (r-s) (s)
Σ Σ f f f

s  f
n-1 n
= f(n)f(0)f(0) + f(0)Σ (n -s) (s)
f + 2f(0)f(0)f
(n)
s =1

Σr  f  r  s 
n-1 n (n -r) (r)
n-1 r -1 n r n -r) (r-s) (s)
+ 2f(0) f + ΣΣ f( f f
r =1 r =2 s =1
i.e.

r  f
(n) n-1 n
f (x)
3
= 3f(0)f(0) f(n)+ 3f(0)Σ (n -r) (r)
f
r =1

 r  s  f
n r
n-1 r -1
(n -r) (r-s) (s)
+ ΣΣ f f
r =2 s =1

When f(x)= log x

(log x)(n -r) = (-1)n-r-1(n -r -1)! x -n+ r (r =0, 1, 2,  , n -1)
(log x)(r-s) = (-1)r-s-1(r -s -1)! x -r+s (s =0, 1, 2,  , r -1)
(s =1, 2, 3, )
(s) s-1
(log x) = (-1) (s -1)! x -s
Substituting these for the above,

-5-
 
3(n -1)! n-1 n
(n) n 3log x
log x
3
= (-1)n-1 log 2
x +( -1) Σ (n -r -1)!(r -1)!
xn x n r =1 r
(-1)n-1 n-1 r -1 n
  
r
+ Σ Σ (n -r -1)!(r -s -1)!(s -1)!
xn r =2 s =1 r s
Furthermore,

r 
n n! n!
(n -r -1)!(r -1)! = (n -r -1)!(r -1)! =
(n -r)!r ! (n -r)r

r   
n r n!
(n -r -1)!(r -s -1)!(s -1)! =
s (n -r)(r -s)s
Using these, we obtain
(n) 3(n -1)! n 3log x
n-1 n!
log x
3
= (-1)n-1 n log 2
x + (-1) n Σ
x x r =1 (n -r)r

(-1)n-1 n-1 r -1 n!
+ n Σ Σ
r =2 s =1 (n -r)(r -s)s
(2.2)
x
When n =1
(1) 3(1-1)! 3log 2x
log x
3 1-1 2
= (-1) log x =
x1 x1
When n =2
(2) 3(2-1)! 32(2-2)!
log x
3
= (-1)2-1 log 2x + (-1)2 log x
x2 x2
3log 2x 6log x
=- +
x2 x2
When n =3
(3) 3(3-1)! 3log x 3-1 3!
log x = (-1) Σ
3 3-1
log 2x + (-1)3
x3 x 3 r =1 (3-r)r
(-1)3-1 3-1 r -1 3!
+ 3 Σ Σ
r =2 s =1 (3-r)(r -s)s
x
6log 2x 18log x 6
= 3
- 3
+ 3
x x x

20.1.3 Higher Derivatives of cos m x , sin m x

Although this formula can also be derived from the Theorem20.1.2 , the proof is long and complicated.
Therefore, we will use the following lemma.

Lemma1
1 m
cos m x = ΣmＣ r cos(m -2r)x m =1, 2, 3,  (3.m)
2m r=0

Proof
The following formulas are known about the power of cos x . (See "岩波 数学公式Ⅱ " p190. )

-6-
 1 m -1 2m Ｃm
2m-1 Σ 2mＣ r cos{(2m -2r)x } +
cos 2mx =
2 r=0 22m
1 m
cos 2m+1x = Σ 2m +1Ｃ r cos{(2m +1-2r)x }
22m r=0

When the pwer is even,

m -1 2m
Σ
r=0
2mＣ r cos{(2m -2r)x } = Σ 2mＣ r cos{(2m -2r)x }
r=m +1

2m Ｃ m = 2mＣ m cos{(2m -2m)x }

Using these,
1 m -1 2mＣm
2m-1 Σ 2mＣ r cos{(2m -2r)x } +
cos 2mx =
2 r=0 22m
1 m -1 m -1 Ｃm
2m Σ 
2m
= 2mＣ r cos{(2m -2r)x } + Σ 2mＣ r cos{(2m -2r)x } +
2 r=0 r=0 22m
1 m -1

2m Σ 
2m
= 2mＣ r cos{(2m -2r)x } + Σ 2mＣ r cos{(2m -2r)x }
2 r=0 r=m +1

1
+ Ｃ m cos{(2m -2m)x }
2m
22m
1 2m

2m Σ 2mＣ r cos{(2m -2r)x }

=
2 r=0

When the power is odd,

m 2m +1
Σ
r=0
2m +1 Ｃ r cos{(2m +1-2r)x } = Σ 2m +1Ｃ r cos(2m +1-2r)x
r=m +1
Using this,
1 m

2m Σ 2m +1Ｃ r cos{(2m +1-2r)x }

cos 2m+1x =
2 r=0

1 m m
=
22m+1 Σ r=0
2m +1Ｃ r cos{(2m +1-2r)x }+Σ2m +1Ｃ r cos{(2m +1-2r)x }
r=0 
1 m

2m+1 Σ 
2m +1
= 2m +1Ｃ r cos{(2m +1-2r)x }+ Σ 2m +1Ｃ r cos(2m +1-2r)x
2 r=0 r=m +1

1 2m +1
= Σ 2m +1Ｃ r cos(2m +1-2r)x
22m+1 r=0

Formula 20.1.3
When  denotes the floor function, the following expressions hold for natural number m .
n
Σ Ｃ r (m -2r) cos(m -2r)x + 
(n) 1 m /2
cos mx = m-1 m
n
(3.mc)
2 r=0 2
 n
Σ Ｃ r (m -2r) cos(m -2r)x - 2  + 
(n) 1 m /2
sin mx = m-1 m
n
(3.ms)
2 r=0 2

-7-
Proof
Differentiating the both sides of (3.m) in Lemma1 with respect to x n times,
n
 
(n) 1 m
cos mx = mΣ
(m -2r)n mＣ r cos (m -2r)x +
2 r=0 2
When m is even,

n
n

  m -2 2  x + 2  = 0
m m
m-2 Ｃ m cos
m
2 2

n n
Ｃ cos (m-2r) x +
2   
m /2-1 m
Σ (m-2r) n m r = Σ (m -2r) n
m Ｃ r cos (m-2r) x +
r=0 r=m /2+1 2
Using these,
n
 
(n) 1 m
cos mx = Σ (m -2r)n
mＣ r cos (m -2r)x +
2m r=0 2
n
 
2 m /2-1
= m Σ (m -2r)n mＣ r cos (m -2r)x +
2 r=0 2
2 m n n
    
m
+ m m -2 mＣ m cos m -2 x+
2 2 2 2 2
n
 
1 m /2
= m-1 Σ (m -2r)n mＣ r cos (m -2r)x +
2 r=0 2
When m is odd,
n n
   
m /2 m
n n
Σ
r=0
(m-2r) m Ｃ r cos (m -2r) x +
2
= Σ (m -2r)
r=m /2+1
Ｃ r cos (m -2r) x +
m
2
Using this,
n
 
(n) 1 m
cos mx = mΣ
(m -2r)n mＣ r cos (m -2r)x +
2 r=0 2
n
 
2 m /2
= m Σ (m -2r)n mＣ r cos (m -2r)x +
2 r=0 2
Thus, we obtain (3.mc) . (3.ms) follows by the replacing x with x - /2 .

6
Example The 7th order derivative of cos x

-8-
20.1.4 Higher Derivatives of cos  x , sin  x

Formula 20.1.4
The following expressions hold for a positive number .
1   n
   
(n)
cos x Σ r
= ( -2r)n cos ( -2r)x + (4.c)
2 r=0 2
1    n
     
(n)
sin x =  Σ ( -2r)n cos ( -2r) x - + (4.s)
2 r=0 r 2 2

Proof
Differentiating the both sides of (3.m) in Lemma1 with respect to x ,
n
 
(n) 1 m
cos mx = Σ (m -2r)n
mＣ r cos (m -2r)x +
2m r=0 2
Analytically continuing the power from the natural number m to the positive number  , we obtain (4.c) .
(4.s) follows by the replacing x with x - /2 .

5.3
Example The 3rd order derivative of sin x
The differential quotient on arbitrary point x= 0.9 was calculated by the direct calculation and the formula.
Both are corresponding very well.

-9-
20.2 Higher Integral of the product of many functions

(1) Generalized binomial theorem and Higher integral of the product of 2 functions
According to generalized binomial theorem in 3.2 , the following expression holds for real numbers x 1, x2
such that x 1 >x 2 and natural number n .

 r x
 -n
x1+ x2 = Σ
-n -n-r r
1 x2
r=0

On the other hand, according to Formula 16.1.2 in 16.1 , the following expression holds for functions f1, f2

of x and natural number n .

  r f
x x m -1 -n
 f1 f2 dx n= Σ <n +r> (r)
1 f2 + Rmn
a a r=0

 f
(-1)m n -1 n -1Ｃ k x x
 <m+ k > (m+ k ) n
(n , m) Σ
Rmn = 1 f2 dx
k=0 m + k a a
Reversing the sign of the index of the differentiation operator (n ) in the Leibniz Rule ,

r
 -n
f1 f2
(-n )
=Σ f1(-n -r)f2(r)
r=0

Next, Replacing (-n) with the intagration operator <n > , and dividing the series into a polynomial and
a remainder, we obtain the above formula.

(2) Generalized multinomial theorem and Higher Integral of the product of many functions
According to the generalized multinomial theorem in 3.4 , the following expression holds for real numbers
x1 , x2 ,  , x such that x 1 >x 2 + x 3 ++ x  and natural number n .

 r r  r 
 r1 r -2 -n r1 r-2 -n-r1 r -r2 r
x1+ x2++x =Σ Σ  Σ   x-1
-n
x1 x21
r1=0 r2=0 r-1=0 1 2 -1

Therefore, the following expression must hold for functions f1 , f2 ,  , f of x and natural number n .

  r r  r 
x x m -1 r1 r-2 -n r1 r-2
f1
n - r1 r1-r2
 f -1
r
 f1 f2  fdx = Σ
n
Σ Σ  f2
a a r =0 r =0 r =0 1 2 -1 1 2 -1

+ Rmn

20.2.1 Higher Integral of the product of many functions

Theorem 20.2.1
Let fk
(r)
be the r th order derivative function of fk (x) k =1, 2, , , fk<r> be the arbitrary r th order

primitive function of fk (x ) , m,n are natural numbers and  (n , m) be the beta function. If there is a

f1r (a) =0 (r =1, 2,  ,m +n -1) fk s (a) =0 (s =0, 1,  ,m +n -2)

<> ()
number a such that or
for at least one k > 1 , then the following expression holds.

  r  r  r 
x x m -1 r1 r-2 -n r1 r-2
f1
n + r1 r1-r2
 f -1
r
 f1 f2  f dx =Σ Σ  Σ
n
 f2
a a r1=0 r2=0 r-1=0 1 2 -1
+ Rmn

- 10 -
  k  k 
n -1Ｃ k1 k-2

(-1)m n-1 m + k1 k2 k -2 m + k1 k2
Rmn = Σ Σ Σ  Σ 
(n , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 k2 3 -1

 f
x x
m+ k 1 m+k 1 -k 2 k 2-k 3
 f -1dx n
k
  1 f2 f3
a a

Proof
Let us integrate f1 f2 f3  f with respect to x from a to x n times. Then according to Theorem 16.1.2
in 16.1 , we obtain

 r 
x x m -1 -n r1
 f1 f2 f3  f dx n = Σ f1 f2 f3  f
n + r1
+ Rmn
a a r1=0 1

(-1)m n -1 n -1Ｃ k1
 f
x x
m+ k 1
f2 f3  f
m+ k 1
(n , m) Σ
Rmn = 1 dx n
k=0 m + k1 a a
According to Theorem20.1.1 ,

r  r  r 
r1 r2 r-2 r1 r2 r-2
f2 1
r -r2 r2- r3
 f -1
r1 r
f2 f3  f = ΣΣ Σ  f3
r2=0 r3=0 r-1=0 2 3 -1

 k  k  k  f
m + k1 k2 m +k k k k -2 -2
k 1 -k 2 k 2- k 3
 f -1
m+ k 1 1 2 k
f2 f3  f =Σ Σ  Σ  2 f3
k2=0 k 3=0 k -1=0 2 3 -1
Substituting these for the above, we obtain the disired expression.

Example

  r  s f f f + R
x x m -1 r -n r
 f1 f2 f3 dx =Σ Σ
n <n + r> (r-s) (s)
1 2 3
n
m
a a r =0 s =0

m + r  s  s  
(-1) Ｃ m +r r m n -1 m + r n -1 x x
 f
r <m+ r> (m+r-s) (s) n
R = n
(n , m)
m Σ Σ r =0 s =0 a
1 f2 f3 dx
a

  f f f f dx ΣΣΣ  r   s   t  f f
x x -n r s m -1 r s
 = 1 2 3 4
n <n + r> (r-s) (s-t) (t)
1 2 f3 f4 + Rmn
a a r =0 s =0 t=0

Ｃ m +r m

m + r  s  s  t 
(-1) r s n -1 m + r s n -1 r
R = n
(n , m)
m Σ Σ Σ r =0 s =0 t=0

 f
x x
  1
<m+ r> (m+r-s) (s-t) (t) n
f2 f3 f4 dx
a a

Example computation of the product of three functions

Since the combination of the product of many functions are numerous, we cannot calculate these one by one.
Then we pick up some and calculate them.

Example1 Higher Integral of x e x sinx

The common zero of the higher order integral of x  e x sin x is x =- .
Then let f1 = x  , f2 = e x , f3 = sin x ,

- 11 -
 <n + r> (1+ )  <m+ r> (1+ ) 
x  x +n+ , x 
r r
= = x +m+
1+ + n + r 1+ + m + r 
(r-s) (m+ r-s)
e x = e x = ex
s
(sin x) s = sin x +  
()
2
Substituting these for Theorem 20.2.1 , we obtain

 x
x x
 x n
e sin xdx
- -

(1+ ) s
 r  s
-n r
  +R
m -1 r
 r
= ΣΣ x +n+ e x sin x + n
(1.1)
r =0 s =0 1+ + n + r 2 m

(-1)m n -1 m + r 1
 r  
n -1 m+r
(n , m) Σ Σ
Rmn =
r =0 s =0 m + r s
(1+ ) s
  
x x

  r
x +m+ e xsin x + dx n (1.1r)
- -
1+ + m + r 2

And when n =2
( +1) +2+ r x
  r  s 
s
x x m -1 r -2 r
x  e x sin x dx = Σ Σ  +R
2 2
x e sin x +
- - r =0 s =0   +3+ r  2 m

m+ r  r   s 
(-1) 1 1
1 m+rm+ r
Σ Σ
2
Rm =
 (2, m) r =0 s =0

(1+)

x s
 
x
x +m+ re xsin x + dx 2
 1+ + m+ r - -   2
This remainder term is an automorphism. And the calculation is difficult when m is large. So let m =1 ,
then the above expression is simplified as follows.
( +1) +2 x
 x
x x

e x sin x dx 2 = x e sin x + R12
- -   +3
 1+
(-1)1 1 1
 0  s    s
1 1 x x

 
( )  +1 x
 (2, 1) Σ
R12 = x e sin x + dx 2
s =0 1   +2 - -   2
(1+)
    s
1 2
 
1 1 x x

 
(-1) 2
 +2 x
Σ
2
+ x e sin x + dx
 (2, 1) 2 1 s
s =0   +3 - -   2
2( +1)
  +2  
x e sin xdx +  x
x x x


x
 +1 x 2  +1 x 2
= - e cos xdx
  - - - -
2( +1)
  +3  
x x
 +2 x 2
- x e cos xdx
  - -

When  =3 , if the values of the both sides on arbitrary point x =4 are calculated, it is as follows.
2
Although the remainder R1 is large because of m=1 , both sides are corresponding exactly.

- 12 -
When m 
R = 0 at the time of m  is asked, unfortunately, it is not so.
2
On the contrary, if whether to be
Surprisingly, when  =3 , it becomes R = -3 regardless of the value of x . Then
2

(3+1) 3+2+ r x s
  -3
x x

 r  s 
 r -2 r
x 3e xsin xdx 2 =Σ Σ x e sin x +
- - r =0 s =0 3+3+ r 2
As the result of much calculation, it turned out that the following formula holds.

Formula 20.2.1'
When  = 0, 1, 2, 
(1+ ) s
  r  s
 -n r
 
x x r

 x e xsin xdx n =Σ Σ
r
x +n+ e xsin x +
- - r =0 s =0 1+ + n + r 2
1 n -1
r
Ｃ r c +r x n-1-r
(n -1)! Σ
+ (-1) n -1 (1.1' )
r=0
Where
k-1
c4k+0 = -(-1)k 2 k !(2k -1)!!(4k -1)!!
k-1
c4k+1 = (-1)k 2 k !(2k -1)!!(4k +1)!!
k-1
c4k+2 = -(-1)k 2 k !(2k +1)!!(4k +1)!!
c4k+3 = 0

According to this formula, R2 at the time of  =3 is obtained as follows.

1 2-1
R2 = r
Ｃ r c3+r x 2-1-r = 1Ｃ 0 c3+0 x 1 - 1Ｃ 1 c3+1 x 0
(2-1)! Σ
(-1) 2-1
r=0
Here, from the proviso,

- 13 -
 c3+0 = c40+3 = 0
c3+1 = c41+0 = -(-1)121-1 1!(21-1)!!(41-1)!! = 3
Substituting these for the above, we obtain R2 = -3 .

By reference, when Rn is shown for n =1, 2, 3 ,  =1, 2, , 10 , it is as followins.

 R1 = c R2 R3

1 1/2 x /2 + 1/2 x 2/4 + x /2
2
2 -1/2 -x /2 -x /4 + 3/2
3 0 -3 -3 x - 15/2
2
4 3 3 x + 15 3x /2 + 15 x + 45/2
2
5 -15 -15 x - 45 -15 x /2 - 45 x
2
6 45 45 x 45 x /2 - 630
7 0 1260 1260 x + 5670
2
8 -1260 -1260 x - 11340 -630 x - 11340 x - 28350
2
9 11340 11340 x + 56700 5670 x + 56700 x
2
10 -56700 -56700 x -28350 x + 1871100

Example2 Higher Integral of x e x log x

A common zero of the higher order integral of x  e x log x is x =- .
 x
Then let f 1 = x , f 2 = e , f 3 = log x
(1+)
x   x r ,
<n+ r> x (r -s)
e  = ex
+n+
=
 1+ + n + r 
(1+)
x   x
<m + r> +m+ r x (m + r -s)
= , e  = ex
 1+ + m + r 
(s 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
(0)
First, we need to separate the terms containing f3 from Theorem 20.2.1 . Then

  f f f dx r=0  r   r  s  f
x x m -1 -n m -1 r -n r
 1 2 3
n
= f3
(0)
Σ f1n+ r f2 r + Σ Σ
< > () <n+ r > (r -s) (s)
1 f2 f3 + Rmn
a a r=1 s=1

 0   f f
(-1) n-1 n -1Ｃ r
m m+ r x x
<m + r> (m +r) (0) n
Rmn = 
 (n , m) Σ 1 2 f3 dx
r =0 m + r a a

 (n , m) ΣΣ m+ r  s   
(-1) Ｃ m
m+r n-1 m + r n -1 r x x
<m + r > (m +r -s) (s) n
+  f 1 f2 f3 dx
r =0 s =1 a a
Substituting the above expressins for this,
(1+)
 r
x x m -1 -n
 x  e x log x dx n = e x log xΣ x +n+ r
- - r=0  1+ + n + r 
(1+) (s)
 r   s   1+ + n + r x
m -1 r -n r  +n+ r-s x
+ Σ Σ(-1)s-1 e + Rmn (1.2)
r=1 s=1  

- 14 -
 (1+)
(-1)m n-1 n -1Ｃ r

x x
Rmn =  x +m+ re x log x dx n
 (n , m) Σ
r =0 m + r - -  1+ + m + r 
(1+) (s) +m+ r-s x n
 s  
n -1Ｃ r
m m+ r x x
(-1) n-1 m + r
+ Σ Σ (-1)s-1  x e dx
 (n , m) r =0 s =1 m+ r - -  1+ + m + r 
(1.2r)

When  =3/2 , n =3, m =80 , the values of the both sides on arbitrary point x=25 are as follows.
The right side of (1.2) seems to be asymptotic expansion.

20.2.2 Higher Integral of the power of a function

Especially, when f1 = f2 =  = f in Theorem 20.2.1 , the following theorem follows immediately.

Theorem 20.2.2
(r)
Let f be the r th order derivative function of f(x) , fk<r> be the arbitrary r th order primitive function of
f(x) , m , n are natural numbers and  (n , m) be the beta function. At this time , if there is a number a
such that

f (a) =0 (r =1, 2,  ,m +n -1) f (a) =0 (s =0, 1,  ,m +n -2)

<r> s
()
or
then the following expression holds for  =2, 3, 4,  .

  r  r  r 
x x m -1 r1 r-2 -n r1 r-2
f
n + r1 r1-r2

 f  -1 + Rmn
r
 f dx =Σ Σ  Σ
n
 f
a a r1=0 r2=0 r-1=0 1 2 -1

(n , m) Σ Σ Σ Σ m + k  k  k  k 
(-1) m
Ｃ m +k k
n-1 m + k1 k2k k -2
n -1 k1 1 2 -2
Rmn =  
k 1=0 k 2=0 k 3=0 k -1=0 1 2 3 -1

 f
x x
m+ k 1 m+k 1-k 2 k 2 -k 3
 f  -1dx n
k
  f f
a a

3
Example Higher Integral of log x
Let f = log x . Then
log x - 1+n + r -  n+ r
<n+ r >
(log x) = x
 1+n + r 
(log x)
(r -s )
= log x (r =s) , (log x)
(r -s )
= (-1)r-s-1(r -s -1) ! x
-r+s
(r s)
(r 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
r =1, 2, 
(0) (r ) (0)
Since f is different from f in function type, we have to separate the terms containing f

- 15 -
from

 f  r  s  f f f + R
x x m -1 r -n r
 dx n = Σ Σ
3 <n+ r> (r -s ) (s ) n
m
a a r=0 s=0

 (n , m) ΣΣ m + r  r   s   
m n -1 m+ r
n-1 m + r x x
(-1) 1 <m + r> (m +r -s) (s)
Rmn =  f f f dx n
r =0 s =0 a a
It is as follows.

r=0 
r  s=0  s 
m -1 -n <n+ r >
r r (r -s) (s)
Σ f f Σ f

 0   0  1    0  1 f f 
-n 0 -n
<n+0>
1 1
(0) (0) <n+1> (1) (0) (0) (1)
= f f f + f f f +

 r    0 s  r  f f 
-n m -1 r r <n+ r> r (r) (0)
r-1
(r -s) (s) (0) (r)
+Σ f f f +Σ f f +
r=2 s=1

r r s  f f
-n
<n> (0) (0) -n
m -1 r <n+ r> (r) (0)
m -1
<n+ r >
r-1
(r -s) (s)
= f f f + 2Σ f f f +Σ f Σ
r=1 r=2 s=1

Σ s f
m+ r
Σ r  f
n -1
n-1
<m + r>
m +r (m +r -s) (s)
f
r =0 s =0

r   s f f +f f 
n -1 n-1 m+ r
<m + r> (m +r) (0) m + r-1 (m +r -s) (s ) (0) (m + r)
= Σ f f f + Σ
r =0 s =1

 s f f
m+ r
r r
n -1 n-1 n -1 <m + r> (m +r) (0)
n-1
<m + r>
m + r-1 (m +r -s) (s)
= 2Σ f f f +Σ f Σ
r =0 r =0 s =1
Substituting thes for the above ,

 f  r f f f
x x m -1 -n
 dx n = f
<n> (0) (0) <n+ r> (r) (0)
+ 2Σ
3
f f
a a r=1

 r  s  f f f
-n r m -1 r-1
<n+ r> (r -s) (s)
+ ΣΣ
r=2 s=1

m+ r  r   
mn -1 n-1 x x
(-1) 2 <m + r> (m +r ) (0)
Rmn = Σ  f f f dx n
 (n , m) r =0 a a

m+ r  r   s   
m n -1 m+ r
n-1 m + r-1 x x
(-1) 1 <m + r > (m +r -s) (s)
+ Σ Σ  f f f dx n
 (n , m) r =0 s =1 a a
Substituting the above expressions and the following expressions for this

<m + r> log x - 1+m+ r -  m+ r

(log x) = x
 1+m + r 
(r =0 , , n -1)
(m + r) -m -r
(log x) = (-1)m+ r-1(m + r-1) ! x
(s =1 , , m+r -1)
(m + r -s) -m-r+s
(log x) = (-1)m+ r-s-1(m + r -s-1) ! x
we obtain
log x - 1+n -  n

x x
 log x dx n =
3
x (log x)
2
0 0  1+n 
-n log x - ( 1+n + r ) - 
 
m -1
- 2 x n log xΣ (-1)r  (r )
r=1 r  ( 1+n + r )
r log x - ( 1+n + r ) - 
  
m -1 r-1 -n
+ x n Σ Σ(-1)r  ( r -s ) (s)
r=2 s=1 r s  ( 1+n + r )

- 16 -
 + Rmn (2.1)

 r    log x - (1+m + r ) -  log x dx

2 n-1 (-1)r n -1 x x
Rmn  n
 (n , m) Σ
=- 2
r =0 (m + r ) 0 0

n-1 m + r-1 (-1)r

 r  
1 n -1 m+ r
 (n , m) Σ Σ
+
r =0 s =1 m+ r s
log x - 1+m + r - 

x x
  (m+ r -s) (s) dx n (2.1r)
0 0  1+m+ r 
Although this formula is complicated, fortunately lim Rmn = 0 holds. However, the proof is difficult and the
m 
convergence speed is very slow.
When n =2 , m=275 , the values of the both sides of (2.1) on arbitrary point x =0.3 are as follows.
The both sides are corresponding up to only 1 digit below the decimal point. If larger m is given, the accuracy
can be raised. However, it takes a lot of time.

20.2.3 Higher Integrls of cos m x , sin m x

Formula 20.2.3c
(1) Oddth power

  n
x x x Ｃr
cos (2m-2r +1) x - 
1 m 2m +1
 cos x dx n =
2m+1
2m Σ n
(n-1) 1 0 2 r=0 (2m-2r +1) 2
2 2 2
(3.co)
(2) Eventh power
n
 
Ｃr
 
x x x 1 m -1

2m
cos 2m x dx n = Σ (2 n cos (2m -2r)x -
2
an a2 a1 22m-1 r=0 m -2r)

   dx
Ｃm x x x

2m n
+ 2m
(3.ce)
2 an a2 a1

Where, a1 , a2 ,  ,an are the solutions of the following transcendental equation.

Ｃr k
 =0
m -1
k =1, 2,  ,n
2m
Σ cos (2m -2r)x -
r=0 (2 m -2r)k 2

- 17 -
 Especially when m=1 ,
n
     
x x x 1 1 x x x
 2 n
cos x dx = n+1 cos 2x - +  dx n
(n -1) 1 0 2 2 2 (n -1) 1 0
4 4 4 4 4 4
(3.c2)

Proof
The following expression holds from Lemma1 .
1 m
cos 2m+1x = Σ 2m +1Ｃ r cos{(2m -2r +1)x }
22m r=0

Integrating the both sides of this with respect to x n times,

Ｃr n
 
<n> 1 m 2m +1
cos 2m+1x = 2m Σ n cos (2m -2r +1)x -
2 r=0 (2 m -2r +1) 2
Here, let us consider the solutions of the following transcendental equations.
Ｃr k
 =0
m
k =1, 2,  ,n
2m +1
Σ cos (2m -2r +1)x -
r=0 (2 m -2r +1)k 2
Then
(k -1)
x = ak = k =1, 2,  ,n
2
are the solutions of the transcendental equations. If it is why,
(k -1) k 

cos (2m -2r +1)
2
-
2  
= cos (m -r)(k -1) -
2 

= (-1)(m-r)(k -1)cos =0
2
These solutions are not dependent on m. Thus we obtain (3.co) .
Next, the following expression holds from Lemma1 .
1 m -1 Ｃm
2m
cos 2mx = 2m Σ Ｃ r cos{(2m -2r)x } +
22m-1 r=0 22m
Integrating the both sides of this with respect to x n times,
Ｃr n Ｃm
 
<n> 1 m -1 2m 2m n -1
k
cos x Σ Σ
2m
= 2m-1 n
cos (2m-2r) x - + 2m
ck+1 x
2 r=0 (2m -2r ) 2 2 k=0

Here, let a k k =1, 2,  ,n are the solutions of the following transcendental equations.

Ｃr k
 =0
m -1
k =1, 2,  ,n
2m
Σ cos (2m -2r)x -
r=0 (2 m -2r)k 2
Then, (3.ce) follows immediately.
Especially when m =1 , these transcendental equations are as follows.
2Ｃ 0 k
2k
cos 2 x -
2  =0 
k =1, 2,  ,n
Then
(k -1)
x = ak = k =1, 2,  ,n
4

- 18 -
are solutions of the transcendental equations. If it is why,
Ｃ0 (k -1) k 
 
1
k =1, 2,  ,n
2
cos 2 - = k cos 2
=0
2k 4 2 2
These solutions are unrelated to m. Thus, (3.c2) holds.

Note1
According to Formula20.1.3 , the n th order derivative of cos 2m+1x is as follows.
n
 
(n) 1 m
cos 2m+1x = Σ 2m +1Ｃ r (2m +1-2r)
n
cos (2m +1-2r)x +
2 2m
r=0 2
Since an integration is the inverse operation of a differentiation, replacing n with -n ,
Ｃr n
 
(-n ) 1 m 2m +1
cos 2m+1x = Σ n cos (2m -2r +1)x -
2 2m
r=0 (2m -2r +1) 2
Replacing the differentiation operator with the integration operator, we can obtain (3.co) .

Note2

  dx
x x
The general notation of  n
is difficult. By reference, if these are shown for n =1, 2, 3 , it is as
an a1
follws. The 4th order or more are unmanageable.

 x -a 1

x -a 2x -2a 1+ a2
x x x
1 2
dx = , dx =
a1 1! a2 a1 2!
x -a3 x + a3 x -3a 1 x + 6a 1 a2 -3a1 a3 -3a 2 + a3 
2 2 2

  dx
x x x
3
=
a3 a2 a1 3!
These resemble a part of constant-of-integration in 4.1.3 . However, these are completely another. That is,
these are undoubted parts of the lineal higher primitive function that originates in cos 2x = (cos 2x - 1)/2
etc. .

5
Example1 The 3rd order integral of cos x

2
Example2 The 3rd order integral of cos x

- 19 -
Formula 20.2.3s
(1) Oddth power

  (-1)m-r2m +1Ｃ r n
x x x

 
1 m
 sin
2m+1
x dx n = 2m Σ n
sin (2m -2r +1) x -
n 2 1 2 r=0 (2m -2r +1) 2
2 2 2
(3.so)
(2) Eventh power
n
   sin
(-1)m-r2mＣ r
 
x x x 1 m -1
n
Σ
2m
x dx = n cos
(2m -2r)x -
an a2 a1 22m-1 r=0 (2m -2r) 2

   dx
Ｃm x x x

2m n
+ 2m
(3.se)
2 an a2 a1

Where, a1 , a2 ,  ,an are the solutions of the following transcendental equation.

(-1)m-r2mＣ r k
 =0
m -1
Σ k cos (2m -2r)x - k =1, 2,  ,n
r=0 (2m -2r) 2
Especially when m =1 ,

  n
 
x x x x x x

 
1 1
 sin x dx n = -
2
n+1
cos 2x - +  dx n
(n+1) 3 2 2 2 2 (n+1) 3 2
4 4 4 4 4 4
(3.s2)

Proof
Replacing x with x - /2 in (3.co) in Formula 20.2.3c ,

   n
x x x Ｃr
cos (2m -2r +1) x -  - 
1 m 2m +1
 sin
2m+1
x dx n = Σ
n 2 1 22m r=0 (2m-2r +1) n 2 2
2 2 2
Here
 n (2m-2r +1)  n

cos (2m-2r +1) x -  2  -
2  
= cos (2m-2r +1) x -
2
-
2 
n

= sin (2m -2r +1)x -
2
-(m -r)

n

= (-1)m-rsin (2m -2r +1)x -
2 
Substituting this for the above, we obtain (3.so) . (3.se) and (3.s2) are obtained from Formul20.2.3c in a
similar way.

7
Example1 The 3rd order integral of sin x

- 20 -
 2
Example2 The 4th order integral of sin x

20.2.4 Higher Integrals of cos  x , sin  x

Lemma1 is a polynomial including binomial coefficients. Such a formula can be easily extended to a real
number region.

Lemma2
1   
cos  x = Σ
2 r=0 r 
cos( -2r)x  >0 , |x| 
2
(4. )

Formula 20.2.4

 cos x dx = 2 Σ  r   -2r
x 11 

 sin ( -2r) x   (4.c)
0 r=0


  

x 1 1 
sin x dx = 
Σ sin ( -2r) x - 2  (4.s)
 2 r  -2r r=0
2
Especially when  = 2m ,


x 1 m -1 Ｃr
2m 2mＣm
2m-1 Σ
cos 2m x dx = sin {(2m -2r)x } + x (4.ce)
0 2 r=0 2m -2r 22m

- 21 -
  (-1)m-r2mＣ r 
x Ｃm
 
1 m -1 2m


sin 2m x dx =
2 2m-1 Σ
r=0 2m-2r
sin(2m-2r) x +
2 2m
x-
2
2 (4.se)

Proof
Integrating both sides of (4. ) in Lemma2 with respect to x from 0 to x, we obtain

 cos r=0  r 
x 1  1

x dx = Σ sin( -2r) x (4.c)
0 2  -2r
Especially when  = 2m ,
m -1 2m Ｃr 2m 2m Ｃr
Σ
r=0 2m -2r
sin (2m -2r)x = Σ
r=m +1 2m -2r
sin (2m -2r)x
Using this,


x 1 2m 2mＣr
2m Σ
cos 2m x dx = sin (2m -2r)x
0 2 r=0 2m -2r

Ｃr Ｃr
2m Σ
sin (2m -2r)x
1 m -1 2m 2m 2m
= sin (2m -2r)x + Σ
2 r=0 2m -2r r=m +1 2m -2r

Ｃm
2m sin (2m -2r)x
+ lim
22m rm 2m -2r
2 m -1 2m Ｃr Ｃm
2m
= Σ 2m -2r sin (2m -2r)x + x (4.ce)
22m r=0 22m
(4.s), (4.se) are obtained by the replacing x with x - /2 .

Note
Even if the number of  is non-even number, the following formula does not hold generally.

1   n
    ( -2r)  
x x x 1
 cos  x dx n = Σ cos ( -2r)x -
an a2 a1 2 r=0 r n 2
It is because that the right side is always a real valued function although the left side is generally a complex
valued function for the real number  . Although the left side can be a real valued function for the real number
, when all a 1 , a 2 ,  ,an is below  /2 , it is difficult to find out such a case. This is the same also

about sin x .

3
Example1 The 1st order integral of sin x

- 22 -
Since both sides have overlapped, the left side (blue) can not be seen.

8
Example2 The 1st order integral of cos x

Since both sides have overlapped, the left side (blue) can not be seen.

2010.12.21
K. Kono
Alien's Mathematics

- 23 -
21 Super Calculus of the product of many functions

21.1 Super Integrals of the product of many functions

(1) Generalized binomial theorem and Super integral of the product of 2 functions
According to the generalized binomial theorem in 3.2 , the following expression holds for the real numbers
x1 , x2 such that x 1 >x 2 and a positive number p .

r
 -p
x1+ x2 = Σ x1-p-rx2r
-p
r=0

On the other hand, according to Formula 17.1.2 in 17.1 , the following expression holds for the functions f1 , f2
of x and a positive number p .

  r f
x x m -1 -p
 f1 f2
<p>
dx 2= Σ <p +r> (r)
1 f2 + Rmp
a a r=0

 k   f
(-1)m  1 p -1 x x
 <m+ k > (m+ k ) p
(p , m) Σ
Rmp = 1 f2 dx
k =0 m + k a a
Reversing the sign of the index of the differentiation operator (p ) in the Leibniz Rule about the Super
Differentiation in 19.1 ,

r
m -1 -p
f1 f2
(-p )
=Σ f1(-p -r)f2(r) + Rm-p
r=0

(-1)m  1
k
p -1 <m+ k > (m+ k ) (-p )
Rm-p =
(p , m) Σ
k =0 m + k
f1 f2 
And, replacing (-p) with the intagration operator <p > , we obtain the above formula.

(2) Generalized multinomial theorem and Super Integral of the product of many functions
According to the generalized multinomial theorem in 3.4 , the following expression holds for real numbers
x1 , x2 , , x such that x 1 >x 2 + x 3 ++ x  and a positive number p .

 r  r  r 
 r1 r-2 -p r1 r-2 -p-r1 r -r2 r
x1+ x2++ x =Σ Σ  Σ   x-1
-p
x1 x21
r1=0 r2=0 r-1=0 1 2 -1

Therefore, the following expression must hold for functions f1, f2 ,  , f of x and a positive number p .

  r  r  r 
x x m -1 r1 r-2 -p r1 r-2
f1
p + r1 r1 -r2
 f -1
r
 f1 f2  f dx =Σ Σ  Σ
p
 f2
a a r1=0 r2=0 r-1=0 1 2 -1
+ Rmp

21.1.1 Super Integral of the product of many functions

Theorem 21.1.1
m be a natural number, fk r
()
Let p,r are positive numbers, be the r th order derivative function of
fk (x) k =1, 2, , , fk r r th order primitive function of fk (x) and  (p ,m) be
<>
be arbitrary
the beta function. At this time, if there is a number a such that
f1r (a ) = 0 r [0 , m+ p] or fk s (a ) = 0 s [0 ,m +p -1] for at least one k > 1 ,
<> ()

the following expression holds

-1-
   r  r  r 
x x m -1 r1 r-2 -p r1 r-2
f1
p + r1 r1 -r2
 f -1
r
 f1 f2  f dx =Σ Σ  Σ
p
 f2
a a r1=0 r2=0 r-1=0 1 2 -1
+ Rmp

 k   k 
k-2
 k 
(-1)m  m + k1 k2 k -2
1 p -1 m + k1 k2
Rmp = Σ Σ Σ Σ
(p , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 1 k2 3 -1


x x
f1
m+ k 1 m+k 1 -k 2 k 2 -k 3
 f -1dx p
k
  f2 f3
a a

Proof
Analytically continuing the index of the integration operator in Formula 20.2.1 in 20.2 to [0 ,p ] from[1 , n ]
we obtain the desired expression.

Example computation of the product of three functions

Example1 Super Integral of x e x sinx

Since the zero of the higher integral of x  e x sin x is x =- , the zero of the super integral is also
 x
the same. Then let f 1 = x , f 2 = e , f 3 = sin x
<p + r> (1+ )  <m+ r> (1+ ) 
x  x +p+ , x 
r r
= = x +m+
1+ + p + r 1+ + m + r 
s
 
(r-s) x (m+ r-s) (s)
e x = e  x
=e , (sin x) = sin x +
2
Substituting these for Theorem 21.1.1 , we obtain

 x
x x
 x p
e sin xdx
- -

(1+ ) s
 r  s
-p r
  +R
m -1 r
 r
= ΣΣ x +p+ e x sin x + p
(1.1)
r =0 s =0 1+ + p + r 2 m

(-1)m  m + r 1
 r  
p -1 m+r
(p , m) Σ Σ
Rmp =
r =0 s =0 m + r s
(1+ ) s
  
x x

  r
x +m+ e xsin x + dx p (1.1r)
-

- 1+  + m r
+ 2
As seen in Theorem 20.2.1 in 20.2 , this Rmp is not converged on 0 at the time of m  . That is, this
polynomial is an asymptotic expansion. When  =3 , p =5/2 , the values of the both sides on arbitrary
point x = 20 are as follows.

-2-
Example2 Super Integral of x e x log x
Since the zero of the higher integral of x  e x log x is x =- , the zero of the super integral is also
 x
the same. Then let f 1 = x , f 2 = e , f 3 = log x
(1+) (1+)
x  x r x  x
<p+ r> +p+ <m + r> +m+ r
= , =
 1+ + p + r   1+ + m + r 
(r -s) (m + r -s)
e x = ex , e x = ex
(s 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
(0)
Separating the terms containing f3 from Theorem 21.1.1 and substituting these fot it,

(1+)
x r
x x

m -1 -p
 e x log x dx p = e x log xΣ x r
+p+
- - r=0  1+ + p + r 
(1+) (s)
 r   s   1+ + p + r x e + R
-pm -1 r r
 +p+ r-s x p
- Σ Σ(-1) s
m (1.2)
r=1 s=1  

(1+)

m 

m+ r  r   0 
(-1) 1 p -1 m+ r x x
 +m+ r x
Rmp = Σ x e log x dx p
 (p , m) r =0  1+ + m + r - -  

(1+) (s)
 
m  m +r s-1

m+ r  r   s 
(-1) (-1) p -1 m+r x x
+m+ r-s x p
+
 (p , m) Σ Σ r =0 s =1  1+ + m+ r
x e dx
- -  
(1.2r)
When  =3/2 , p =2.7 , the values of the both sides on arbitrary point x = 17 are as follows. This also
seems to be asymptotic expansion.

21.1.2 Super Integral of the power of a function

Especially, if f1 = f2 =  = f in Theorem 21.1.1 , the following theorem holds immediately.

Theorem 21.1.2
m be a natural number, f r be the r th order derivative function of f(x) ,
()
Let p,r are positive numbers,

f r be arbitrary r th order primitive function of f(x) and  (p , m) be the beta function. At this time,
<>

if there is a number a such that

f r (a ) = 0 r [0 , m+ p ] or f s (a) = 0 s [0 ,m +p -1]
<> ()


-3-
the following expression holds for  = 2, 3, 4,  .

  r  r  r 
x x m -1 r1 r-2 -p r1 r-2
f
p + r1 r1 -r2
f  dx p =Σ Σ  Σ  f  -1 + Rmp
r
  f
a a r1=0 r2=0 r-1=0 1 2 -1

m + k  k   k  k  k 
(-1) m p -1
 m + k1 k2
1 m +k k kk -2 1 2 -2
Rmp = Σ Σ Σ Σ 
(p , m) k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1

 f
x x
m+ k 1 m+k 1 -k 2 k 1 -k 2
 f  -1dx p
k
  f f
a a

3
Example Super Integral of log x
3
Since the zero of the higher integral of log x is x =0 , the zero of the super integral is also the same.
Then let f = log x
<p+ r> log x - 1+p + r  -  p+ r log x - 1+m + r  -  m+ r
, (log x) m r =
< + >
(log x) = x x
 1+p + r   1+m + r 
(log x)
(r -s)
= log x (r =s) , (log x)
(r -s)
= (-1)r-s-1(r -s -1) ! x
-r+s
(r s)
(r 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
(0)
f from Theorem 21.1.2 and substituting these fot it,
Separating the terms containing
log x - 1+p  -  p

x x
 (log x) 3dx p = x (log x)
2
0 0   1+p 
-p log x - 1+p + r  - 
 
m -1
- 2x p log xΣ (-1)r r
r=1 r  1+p + r 
r log x - 1+p + r  - 
  
m -1 r-1 -p
+ x p Σ Σ(-1)r  r -s  (s)
r=2 s=1 r s  1+p + r 
+ Rmp (2.1)

 m + r (-1) r-s

 r  s 
1 p -1 m+ r
Rmp = -
 (p , m) Σ Σ
r =0 s =0 m + r
log x - 1+m + r -
 
x x 
(m+r -s) x sdx p (2.1r)
0 0 1+m+ r  
n
And lim R =0 holds although the proof is difficult. Then
m  m
log x - 1+p  -  p
  (log x) dx
x x
 3 p
= x (log x)
2
0 0  1+p 
-p log x - 1+p + r  - 
 

- 2x p log xΣ(-1)r r
r=1 r  1+p + r 
r log x - 1+p + r  - 
  
 r-1 -p
+ x pΣΣ(-1)r  r -s  (s)
r=2 s=1 r s  1+p + r 
(2.1')
However, the convergence speed is very slow.
When p =3/2 , m =400 , the values of the both sides on arbitrary point x = 0.2 are as follows. Even if
calculated so far, both sides are corresponding only to 1 digit below the decimal point. Although (2.1') is not
suitable for the calculation, it is meaningful that the integral can be expressed by the double series.

-4-
21.1.3 Super Integrls of cos m x , sin m x

Formula 21.1.3
Let m be a natural number, p be a positive number and a(s) s [0 , p ] be the zero of the lineal super
2m+1
primitive function of cos x or sin 2m+1x . Then the following expressions hold.

 p
2m +1Ｃ r
x x

 
1 m
 cos
2m+1
x dx p = 2m Σ cos ( 2m -2r +1 ) x - (3.c)
a(p) a(0) 2 r=0 (2m -2r +1) p 2

  (-1)m-r2m +1Ｃ r p
x x

 
1 m
x dx p = Σ
2m+1
sin 2m p
sin (2m-2r +1) x - (3.s)
a(p) a(0) 2 r=0 (2m -2r +1) 2

Proof
Analytically continuing the index of the integration operator in Formula 20.2.3c (1) and Formula 20.2.3s (1)
in 20.2 to [0 , p ] from[1 , n] , we obtain the desired expressions.

Note
For example, in the case of cos 2m+1x , a(s) is a solution of the following transcendental equation.
Ｃr s
s cos ( m =0
m
s [0 , p]
2m +1
Σ
r=0 (2m -2r +1)
2 -2r +1)x -
2
Although it is difficult to calculate this solution, the necessity does not exist in this formula.

3
Example The 1.5 th order intagral of cos x

-5-
-6-
21.2 Super Derivatives of the product of many functions

(1) Generalized binomial theorem and Leibniz Rule about Super Differentiation
According to the generalized binomial theorem in 3.2 , the following expression holds for the real numbers
x1 , x2 such that x 1 >x 2 and a positive number p .

 r
 p p-r
x1+ x2 = Σ
p
x1 x2r
r=0

On the other hand, according to Formula 19.1.1 in 19.1 , the following expression holds for the functions f1 , f2

of x and a positive number p .

 rf
(p) m -1 p (p -r) (r)
f1 f2 =Σ 1 f2 + Rmp
r=0

(-1)m  1
 
-p -1 <m+ k > (m+ k ) (p)
Rmp =
(-p , m) Σ
k =0 m + k k
f1 f2 

(2) Generalized multinomial theorem and Super Derivative of the product of many functions
According to the generalized multinomial theorem in 3.4 , the following expression holds for real numbers
x1 , x2 ,  , x such that x 1 >x 2 + x3 ++ x  and a positive number p .

  r  r 

r1 r-2 p r1 r-2 p-r1 r -r2 r
   x-1
p
1 2
x + x ++ x =Σ Σ Σ
r1=0 r2 =0 r-1 =0 r1
x1 x21
2 -1
Therefore, the following expression must hold for functions f1 , f2 ,  , f of x and a positive number p .

 r  r  r 
m -1 r1 r-2 p r1 r-2
f1
p - r1 r1 -r2
 f -1 +
r
f1 f2  f = ΣΣ Σ 
(p)
f2 Rmp
r1=0 r2=0 r-1=0 1 2 -1

21.2.1 Super Derivative of the product of many functions

Theorem 21.2.1
m be a natural number, fk r
()
Let p,r are positive numbers, be the r th order derivative function of fk (x)
k =1, 2, , , fk r r th order primitive function of fk (x) and  (p , m) be the beta function.
<>
be arbitrary
At this time, if there is a number a such that
f1r (a) = 0 r [0 , m - p] or fk s (a) = 0 s [0 , m-p -1]
<> ()
for at least one k >1 ,
the following expression holds

 r  r  r 
m -1 r1 r-2 p r1 r-2
f1
p - r1 r1 -r2
 f -1
r
f1 f2  f =Σ Σ  Σ 
(p) p
+R f2 m
r1=0 r2=0 r-1=0 1 2 -1

(-p , m) Σ Σ Σ Σ m + k  k   k  k  k 
(-1) m -p -1 m +k
 m + k1 k2
1 k kk -2 1 2 -2
Rmp =  
k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1
(p)
 m+ k 1 m+k 1-k 2 k 2-k 3
f1 f2 f3 k -1
 f 
Proof
Reversing the sign of the index of the integration operator <p> in Formula 21.1.1 ,

-7-
   r  r  r 
x x m -1 r1 r-2 p r1 r-2
f1
-p + r1 r1 -r2
 f -1
r
 f1 f2  f dx -p
=Σ Σ  Σ  f2
a a r1=0 r2=0 r-1=0 1 2 -1
+ Rm-p

 k  k 
k-2

(-1)m  m + k1 k2 m + k1 k2

k -2
1 -p -1
Rm-p = Σ Σ Σ  Σ 
(-p , m) k 1=0 k 2=0 k 3=0 k -1=0 m + k1 k1 k2 3 -1

 f
x x
m+ k 1 m+k 1-k 2 k 2-k 3
 f -1dx -p
k
  1 f2 f3
a a
Then, replacing the integration operator <-p> with the differentiation operator (p) , we obtain the desired
expression.

Example computation of the product of three functions

Example1 Super Derivative of x e x sinx


Let f 1 = x , f 2 = e x , f 3 = sin x , then
(p - r) (1+) <m+ r> (1+)
x  x r , x  x
-p+ +m+ r
= =
 1+ - p + r   1+ + m + r 
s
 
x (r-s) (m + r -s) (s)
e  = e x = ex , (sin x) = sin x +
2
Substituting these for Theorem 21.2.1 , we obtain
(1+) s
 r  s 
m -1 r p r
  +R
 x (p)
x e sin x = ΣΣ x  r e x sin x + p
-p+
r =0 s =0  1+ - p + r  2 m

(1.1)
m

 
m+ r

(-1)  m+r 1 -p -1
Rmp = Σ Σ
 (-p , m) r =0 s =0 m + r r s
(1+) s
(p)

 x  
+m+ r x
e sin x + (1.1r)
 1+ + m + r  2

When  =3/2 , p =1/2 , the values of the both sides on arbitrary point x = 1.4 are asfollows.

Example2 Super Derivative of x e x log x


Let f 1 = x , f 2 = e x , f 3 = log x , then
(1+) (1+)
x  x r x  x
(p- r) -p+ <m + r> +m+ r
= , =
 1+ - p + r   1+ + m + r 
x (r -s) x (m + r -s)
e  = ex , e  = ex

-8-
 (s 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
(0)
Separating the terms containing f3 from Theorem21.2.1 and substituting these for it,

(1+)
r 
 x
m -1 p
x r
(p)
x e log x = e xlog xΣ
-p+
r =0  ( 1+ - p + r )
(1+) (s)
 r   s   (1+ - p + r ) x
m -1 rp r  -p+ r-s
- e Σ Σ (-1)
x s
+ Rmp (1.2)
r =1 s =1

(-1)m (1+) (p)

 
 -p -1

1
x
+m+ r x
Rmp = Σ e log x
 (-p , m) r =0 m + r r  ( 1+ + m + r )
(-1)m  m + r (-1)s (1+) (s) +m+ r-s x (p)

  s  
-p -1 m+r
-
 (-p , m) Σ Σ
r =0 s =1 m + r r  ( 1+ + m+ r )
x e

(1.2r)
Polynomials in (1.2) seem to be asymptotic expansion. when  =3/2 , p =1/2 , the values of the both sides
on arbitrary point x = 16 are as followas.

21.2.2 Super Derivative of the power of a function

Especially, if f1 = f2 =  = f in Theorem 21.2.1 , the following theorem holds immediately.

Theorem 21.2.2
m be a natural number, f r be the r th order derivative function of f(x) ,
()
Let p,r are positive numbers,

f r be arbitrary r th order primitive function of f(x) and  (p , m) be the beta function. At this time,
<>

if there is a number a such that

f r (a) = 0 r [0 , m - p] or f s (a) = 0 s [0 , m -p -1]
<> ()

the following expression holds for  = 2, 3,4,  .

 r  r  r 
m -1 r1 r-2 p r1 r-2
f
p + r1 r1 -r2
 (p)
 f  -1 + Rmp
r
f  = ΣΣ Σ  f
r1=0 r2=0 r-1=0 1 2 -1

(-p , m) Σ Σ Σ Σ m + k  k   k  k  k 
(-1) m -p -1 m +k
 m + k1 k2
1 k k k -2 1 2 -2
Rmp =  
k 1=0 k 2=0 k 3=0 k -1=0 1 1 2 3 -1
(p)
f   f  -1
m+ k 1 m+k 1 -k 2 k 2 -k 3 k
 f f

3
Example Super Derivative of log x
Let f = log x , then

-9-
 (p- r) log x - 1-p + r  -  -p+ r log x - 1+m + r  -  m+ r
, (log x) m r =
< + >
(log x) = x x
 1-p + r   1+m + r 
(log x)
(r -s)
= log x (r =s) , (log x)
(r -s)
= (-1)r-s-1(r -s -1) ! x
-r+s
(r s)
(r 0)
(0) (s) -s
(log x) = log x (s =0) , (log x) = (-1)s-1(s -1) ! x
(0)
Separating the terms containing f3 from Theorem21.2.2 and substituting these for it , we obtain
(p) log x - 1-p  -  -p
(log x) 
3 2
= x (log x)
 1-p 
p log x - ( 1-p + r ) - 

m -1
- 2x log xΣ (-1)r
-p
(r)
r=1 r  ( 1-p + r )
r log x - ( 1-p + r ) - 
  
m -1 r-1 p
+ x Σ Σ(-1)r
-p
(r -s) (s)
r=2 s=1 r s  ( 1-p + r )
+ Rmp (2.1)

 (-1) r 2

 
1 -p -1
log x - 1+m + r  -
(p)
Rmp = -
 (-p , m) Σ 2
log x
r =0 (m + r ) r
 m + r-1 (-1)r

 
m+ r

1 -p -1
 (-p , m) Σ Σ
+
r =0 s =1 m + r r s
log x - 1+m + r  - 
(p)

  1+m + r 
(m + r -s)  (s)
 (2.1r)

n
And lim R =0 holds although the proof is difficult. Then
m  m

(p) log x - ( 1-p ) -  -p

(log x) 
3 2
= x (log x)
 ( 1-p )
p log x - ( 1-p + r ) - 


- 2x log xΣ(-1)r
-p
(r)
r=1 r  ( 1-p + r )
r log x - ( 1-p + r ) - 
  
-p 
r-1 p
+ x ΣΣ(-1)r (r -s) (s)
r=2 s=1 r s  ( 1-p + r )
(2.1')
However, the convergence speed is very slow.
When p =1/2 , m =120 , the values of the both sides on arbitrary point x =4 are as follows. Even if
calculated so far, both sides are corresponding only to 1 digit below the decimal point.

- 10 -
21.2.3 Super Derivatives of cos m x , sin m x

Formula 21.2.3
When m is a natural number, p is a positive number and  is the floor function,
p
Σ Ｃ r (m -2r) cos(m -2r)x + 
(p) 1 m /2
cos mx = m-1 m
p
(3.c)
2 r=0 2
 p
Σ Ｃ r (m -2r) cos(m -2r)x - 2  + 
(p) 1 m /2
sin x
m p
= m-1 m (3.s)
2 r=0 2

Proof
Analytically continuing the index of the differentiation operator in Formula 20.1.3 in 20.1 to [0 , p ] from
[1 , n ] , we obtain the desired expressions.

4
Example The 2.5 th order derivative of cos x

Collateral Super Derivatives of cos m x , sin m x

Unlike the higher derivative, in the super derivative, the lineal super derivative and the collateral one exist.
Although this was described in 12.1.4 , I describe it once now.

- 11 -
 3
For example, if we differenciate cos x with respect to x p times according to Theorem 21.2.2 , it is as
follows.
(p -r)  (r -s)  s
 r  s  
p r
   cos x + 2 
(p) m -1 r
cos x =Σ Σ
3
cos x + cos x +
r=0 s=0 2 2
+ Rmp (4.c')

(-1)m
 s 
 m +r m+r

1 -p -1
Rmp = Σ Σ
 (-p , m) r =0 s =0 m + r r
(m +r)  (m +r -s)  s
p

 
 cos x -
2  
cos x +
2  
cos x +
2  (4.cr')

Then, this is a collateral super derivative. If it is why, this theorem was drawn out as a reverse-operation of the
super integral of the product of many functions as seen during the proof of Theorem 21.2.1 . And the base was
the super integral with a fixed lower limit.
3
However, the lineal super integral of cos x is one with a variable lower limit. Therefore, (4.c') and (4.cr') derived
based on the fixed lower limit cannot be the lineal super derivative. Althogh this holds as a equation, the
polynomial of (4.c') is not well behaved.
By reference, let us compare (4.c') with the following lineal super derivative (4.c) derived from Formula 21.2.3 .

p 3p p
   
3 (p) 3
cos x = cos x + + cos 3x + (4.c)
4 2 4 2
Althogh (4.c') fits (4.c) most at the time of m=3 , a big difference is still seen by both.

However, since 1/ (-p , m) = 0 at

p
p = m -1 , m =1, 2, 3,  , Rm = 0 . Therefore
(m -1-r) (r -s) s
  s  cosx +
m -1 r m -1 r
 cosx +  cos x + 2 
(m -1)
cos x
3
=Σ Σ
r=0 s=0 r 2 2
Furthermore, replacing m -1 with n ,

n -r (r -s)  s
 r  s  
n r
   
n r


(n)
cos x =ΣΣ
3
cos x + cos x + cos x +
r=0 s=0 2 2 2
And this results in the following lineal super derivative. ( The proof is long, then omitted. )

n 3n n
 +  
3 (n) 3
cos x = cos x + cos 3x +
4 2 4 2

2010.12.25
K. Kono
Alien's Mathematics

- 12 -
22 Higher Derivative of Composition

22.1 Formulas of Higher Derivative of Composition

22.1.1 Faà di Bruno's Formula

About the formula of the higher derivative of composition, the one by a mathematician Faà di Bruno in Italy of
about 150 years ago seems to be the beginning. And it is called Faà di Bruno's Formula .

Formula 22.1.1 ( Faà di Bruno )

Let j1 , j2 ,  , jn are non-negative integers. Let g Bn,rf1 , f2 , 
(n)
, fn are derivative functions and
are Bell polynomials such that

(n =1, 2, 3, )
(n) (n) (n)
g = g ( f) , fn = f (x)
j1 j2 jn

     
n! f1 f2 fn
Bn,rf1 , f2 ,  , fn =Σ
j1! j2!  jn! 1! 2! n!

j1 + j2 +  + jn = r & j1 +2j2 +  + n jn = n 

Then, the higher derivative function with respect to x of the composition gf(x) is expressed as follows.
n
(n) (r)
gf(x) = Σ
r=1
g Bn,rf1 , f2 ,  , fn (1.1)

(4)
Example gf(x)
First, j1 , j2 , j3 , j4 such that j1+j2+j3+j4 = 1 , j1+2j2+3j3+4j4 = 4 are

j1 , j2 , j3 , j4 = (0 , 0 , 0 , 1)
Then,
0 0 0 1

    
4! f1 f2 f3 f4
B4,1f1 f4 = = f41
0!0!0!1! 1! 2! 3! 4!
2nd, j1 , j2 , j3 , j4 such that j1+j2+j3+j4 = 2 , j1+2j2+3j3+4j4 = 4 are

j1 , j2 , j3 , j4 = (1 , 0 , 1 , 0)
j1 , j2 , j3 , j4 = (0 , 2 , 0 , 0)
Then,
1 0 1 0

    
4! f1 f2 f3 f4
B4,2f1 f4 =
1!0!1!0! 1! 2! 3! 4!
0 2 0 0

0!2!0!0!  1!   2!   3!   4! 
4! f f f 1 f 2 3 4 1 1 2
+ = 4f 1 f3 +3f2

3rd, j1 , j2 , j3 , j4 such that j1+j2+j3+j4 = 3 , j1+2j2+3j3+4j4 = 4 are

j1 , j2 , j3 , j4 = (2 , 1 , 0 , 0)
Then,
2 1 0 0

    
4! f1 f2 f3 f4
B4,3f1 f4 = = 6f12f21
2!1!0!0! 1! 2! 3! 4!
Last, j1 , j2 , j3 , j4 such that j1+j2+j3+j4 = 4 , j1+2j2+3j3+4j4 = 4 are

j1 , j2 , j3 , j4 = (4 , 0 , 0 , 0)

-1-
Then,
4 0 0 0

    
4! f1 f2 f3 f4
B4,4f1 f4 = = f14
4!0!0!0! 1! 2! 3! 4!
Thus,
(4)
gf(x) = g(1) B4,1f1 , f2 , f3 , f4 + g(2) B4,2f1 , f2 , f3 , f4
+ g(3)B4,3f1 , f2 , f3 , f4 + g(4) B4,4f1 , f2 , f3 , f4
= g(1)f41 + g(2)4f11 f31 + 3f22 + 6 g(3)f12 f21 + g(4)f14

As understood from this example, obtaining j1 , j2 , j3 , j4 such that

j1+ j2+ + jn = k , j1 +2j2 +3j3 +  + njn = n , jk  0 k =1, 2,  ,n

is equivalent with solving the following indeterminate equation. It is not easy.
1j2 +2j3 +  + (n -1)jn = n -k , jk  0 k =2, 3,  ,n

Althogh the condition is jk  0 k =1, 2,  ,n in Formula 22.1.1 , if jk > 0 k =1, 2,  is

adopted, Formula 22.1.1 can be expressed as follows.

Formula 22.1.1'
Let , , , and n are natural numbers, g
(n)
, fn are g
(n)
=g
(n)
(f ) , fn = f
(n)
(x)
Then, the higher derivative function with respect to x of the composition gf(x) is expressed as follows.
  
( + ++)
n! fp fq  ft
p +q ++t = n
(n)
gf(x) = Σg
p! q!  t! !!!
(1.1')

Algorithm of calculation of the Formula 22.1.1'

If the formula of a composition is expressed in this way, (1.1') can be calculated with a easier algorithm. This
algorithm of author design is suitable for the computer. However, since it is easy, manual calculation can also
be performed.

  
(1) Obtain all the products fp fq  ft such that p +q ++t = n as follows.

Step1 Put n on the 1st row 1st column, and arrange 0 from the 2nd column to the
n th column .
Step2 Look for the first number that is smaller than the number of the 1st column as for 2 or more
sequentially from the 2 nd column.
I When such a number exists in the k th column, make a new sequence ( new row ) as follows.
i From the 2 nd column to the k th column, arrange the number that is 1 greater than
the number in the k th column of the old sequence ( the row just above ) .
ii From the k +1 th column to the n th column, copy the numbers of old sequence ( row just above )
iii To the 1 st column, write the complement that the total of each column becomes n .
II When such a number does not exist, finish calculating and go to Step4 .
Step3 Repeat Step2 for the new sequence ( new row ) .
Step4 Let the number in the each sequence be the order p and the repetition frequency
  
be the degree  . And generate the product fp fq  ft .

-2-
 For example, the combination of the 7 th order is generated as follows.

Seq. 1 2 3 4 5 6 7
7 0 0 0 0 0 0 f71 Red is smaller than the 1st column by 2 or more.

6 1 0 0 0 0 0 f61 f11 Blue is a complement.

5 2 0 0 0 0 0 f51 f21
4 3 0 0 0 0 0 f41 f31
5 1 1 0 0 0 0 f51 f12
4 2 1 0 0 0 0 f41 f21 f11
3 3 1 0 0 0 0 f32f11
3 2 2 0 0 0 0 f31 f22
4 1 1 1 0 0 0 f41 f13
3 2 1 1 0 0 0 f31 f21 f12
2 2 2 1 0 0 0 f23f11
3 1 1 1 1 0 0 f31 f14
2 2 1 1 1 0 0 f22f13
2 1 1 1 1 1 0 f21 f15
1 1 1 1 1 1 1 f17
 
(2) The coefficient of each term fp fq  ft  is calculated as follows, for example.

7!
Coefficient of f31 f21 f12 : = 210
3! 2! 1!21!1!2!
1 1

3 1 7!
Coefficient of f2 f1 : 3 1
= 105
2! 1! 3!1!

fp fq  ft 
(s)
(3) The multiplier g of each term is calculated as follows, for example..
(s)
Multiplier of f31 f21 f12 : g = g(1+1+2) = g(4)

(4) Therefore, the 7th order derivative z7 is as follows, for example..

 6! 1! 1!1! 
7! f71 (1)
f61 f11 f51 f21 f41 f31
z7 = 1 g + 7! 1 1
+ 1 1
+ 1 1 g(2)
7! 1! 5! 2! 1!1! 4! 3! 1!1!

 5! 1! 1!2! + 4! 2! 1! 1!1!1! + 3! 1! 2!1! + 3! 2! 1!2!  g

f51 f12 f41 f21 f11 f32f11 f31 f22 (3)
+ 7! 1 2 1 1 1 2 1 1 2

 4! 1! 1!3! 3! 2! 1! 1!1!2! 2! 1! 3!1!  g

1 3 1 1 2 3 1
f f 4 1 f f f f f 3 2 1 2 1 (4)
+ 7! + 1 1
+ 1 1 2 3 1

-3-
  3! 1! 1!4! 
f31 f14 f22f13
+ 7! 1 1
+ 2 3 g(5)
2! 1! 2!3!
7! f21 f15 (6)
7! f17
+ 1 5 g + 1 g(7)
2! 1! 1!5! 1! 7!
= g(1)f7 + g(2)7 f6 f1 + 21 f5 f2 + 35 f4 f3
+ g(3)21 f5 f12 + 105 f4 f2 f1 + 70 f32f1 + 105 f3 f22
+ g(4)35 f4 f13 + 210 f3 f2 f12 + 105 f23f1
+ g(5)35 f3 f14 + 105 f22f13 + 21g(6)f2 f15 + g(7)f17

Higher differentiation up to the 8th order

If the higher derivative zn up to the 8th order of composition z = gf(x) is calculatedby such a way,
it becomes the following.
z1 = g(1)f1
z2 = g(1)f2 + g(2)f12
z3 = g(1)f3 + 3g(2)f2 f1 + g(3)f13
 
z4 = g(1)f4 + g(2) 4f3 f1 +3f22 + 6g(3)f2 f12 + g(4)f14
z5 = g(1)f5 + g(2)5f4 f1 +10f3 f2+ g(3) 10f3 f12+15f22f1  +10g f2 f1 + g f1
(4) 3 (5) 5

z6 = g(1)f6 + g(2)6f5 f1 +15f4 f2 +10f32 + g(3)15f4 f12 +60f3 f2 f1 +15f23

+ g(4)20f3 f13 + 45f22f12 + 15g(5)f2 f14 + g(6)f16
z7 = g(1)f7 + g(2)7 f6 f1 + 21 f5 f2 + 35 f4 f3
+ g(3)21 f5 f12 + 105 f4 f2 f1 + 70 f32f1 + 105 f3 f22
+ g(4)35 f4 f13 + 210 f3 f2 f12 + 105 f23f1
+ g(5)35 f3 f14 + 105 f22f13 + 21g(6)f2 f15 + g(7)f17
z8 = g(1)f8 + g(2)8f7 f1 +28f6 f2 +56f5 f3 +35f42
+ g(3)28f6 f12 +168f5 f2 f1 +280f4 f3 f1 +210f4 f22 +280f32f2
+ g(4)56f5 f13 +420f4 f2 f12 +280f32f12 +840f3 f22f1 +105f24
+ g(5)70f4 f14 +560f3 f2 f13 +420f23f12
+ g(6)56f3 f15 +210f22f14 + 28g(7)f2 f16 + g(8)f18

3
Example When z = gf(x) = sinx 

g(1) = cos f , g(2) = -sin f , g(3) = -cos f

f1 = 3x 2 , f2 = 6 x , f3 = 6

-4-
Then,

z1 = cosx 33x 2
2
z2 = cosx 36x - sinx 33x 2
3
z3 = cosx 36 - 3 sinx 33x 26x - cosx 33x 2

22.1.2 Hoppe's Formula

Afterwards, it was discovered that Formula 22.1.1 is expressed with a double series of g(r)(f ) and
f(s)(x) . The finished type is Hoppe's formula mentioned as follows.

Formula 22.1.2 (Reinhold Hoppe)

(n) (n) (n) (n)
Let n be a natural number, g , fn are g =g (f ) , fn = f (x). Then, the higher derivative
function with respect to x of the composition gf(x) is expressed as follows.
(r)

s 
(n) n g r r (n)
gf(x) =Σ Σ (-f)r-sf s (1.2)
r=0 r! s=0

Although this formula is beautiful, the calculation is not easy. When the differentiation is actually performed
according to (1.2), it is as follows.
(r)

s  (-f)
2 g r r r-s s (2)
gf(x)
(2)
=Σ Σ f 
r=0 r! s=0
() ()
g 0 g 1
 0 s=0  s 
0 0 (2)
1 1 s (2)
(-f) f  Σ f 
0 1-s
= + (-f)
0! 1!
( )
g 2
s=0  s 
2 2 s (2)
Σ f 
2-s
+ (-f)
2!
()
g 1
  0  1 
1 (2) 1 (2)
(-f) f  (-f) f 
1 0 0 1
= +
1!
(2)

2!   0  1  2 (-f) f  
g 2 2 0 (2)
2 1 (2) 2 (2)
(-f) f  + (-f) f  +
2 1 0
+
(1) (2)

1!  1 2!   1  2 2f f + f f 
g 1 g 2
1
2 1 1
= f + 2 (-f) f + 2 1 1 2

g(1)f21 g(2)
2! 
= + -2f f2 + 2f1 f1 + f f2 = g(1)f2 + g(2)f12
1!

22.1.3 Higher order differentiation using Mathematica

A function BellY [n , r, {x1 , x 2 ,  , xn-r+1}] is implemented in a formula manipulation software
Mathematica since 2010, By using this, we can easily generate the Bell polynomials Bn,rf1 , f2 ,  , fn .
For example, B8,3f1 , f2 ,  , f8 is as follows.

-5-
 Further, replacing the table {x1 , x2 ,  , xn-r+1} with a function, we can generate Formula 22.1.1 itself.
For example, let

Then, the above z8 is easily generated as follows.

Of course, if a specific function is given toor fk , a desired higher order derivative function can be obtained.
gr
The examples are shown in the following chapters.

22.1.4 Higher Derivative in case the core function is the 1st degree
Although the higher differentiation of a general composition is complicated in this way, If the core function
f(x ) is the 1st degree, it becomes remarkably easy.

Formula 22.1.4
When g
(n)
=g
(n)
( f ) , fn = f (x) (n =1, 2, 3, ),
(n)
if f(x) is the 1st degree, the higher derivative

function with respect to x of the composition gf(x) is expressed as follows.

(n) (n)
gf(x) = g f1n (1.4)

Proof
In Formula 22.1.1 , when f(x) is the 1st degree, f2 = f3 = f4 =  = 0 . Then, the Bell polynomials
Bn,rf1 , f2 ,  are as follows.
j1 j2 jn

      =0
n! f1 f2 fn
Bn,rf1 , f2 ,  , fn =Σ
j1! j2!  jn! 1! 2! n!

j1 + j2 +  + jn = r < n & j1 +2j2 +  + n jn = n 

n 0 0

     
n! f1 f2 fn
Bn,nf1 , f2 ,  , fn =Σ = f1n
n! 0!  0! 1! 2! n!
Therefore,
n
(n)
gf(x) = Σg(r) Bn,rf1 , f2 ,  , fn
r=1
n -1
= Σg(r) Bn,rf1 , f2 ,  , fn + g(n) Bn,nf1 , f2 ,  , fn
r=1

= g(n)f1n

-6-
Example When z = sin(ax+b )
1
z1 = cos(ax +b)a 1 = a 1 sin ax +b +
2  
2
z2 = -sin(ax +b)a 2 = a 2 sin ax +b +
2  

n
zn = 
a n sin ax +b +
2 
The last expression is consistent with linear form in Formula 9.2.1 in " 9.2 Higher Derivative ".

-7-
22.2 Higher Derivative of Some Composition
There are quite a lot of combinations even if the composition is limited to the elementary function. We cannot
calculate these one by one. So, in this section, we calculate only an easy and interesting thing.

e f
(x)
22.2.1 Higher Derivative of
g = e  f(x
)
Placing with in Formula 22.1.1 , we obtain the following formulas immediately.
n
f(x) (n)
e  = e f(x)ΣBn,kf1 , f2 ,  , fn (2.1+)
k =1
n
-f(x) (n)
e  = e -f(x)Σ(-1)rBn,kf1 , f2 ,  , fn (2.1-)
k =1
If (2.1+) is written down to the 5th order, it is as follows.
( ) (1)
e f x  = e f(x)f1
( ) (2)
e f x  = e f(x)f2 + f12
( ) (3)
e f x  = e f(x)f3 + 3f2 f1 + f13
( ) (4)
e f x  = e f(x)f4 + 4f3 f1 +3f22 + 6f2 f12 + f14 
( ) (5)
e f x  = e f(x)f5 + 5f4 f1 +10f3 f2 + 10f3 f12+15f22f1 + 10f2 f13 + f15 

2 (3)
Example1 e -x 
f = -x 2 , f1 = -2x , f2 = -2 , f3 =0
Then,
2 (3) 2 2
e -x  = e -x 0 + 3(-2)(-2x) + (-2x)3 = 4e -x x3 - 2x 2

(4)
Example2 e sin x
f = sin x , f1 = cosx , f2 = -sin x , f3 = -cosx , f4 = sin x
Then,
sin x (4)
e  = e sin xsin x + -4cos 2x +3sin 2x - 6 sin xcos 2x + cos 4x

-cos x (8)
Example3 e 
We calculate using a formula manipulation software Mathematica according to (2.1-) .
n
f = cosx , fn = cos x +  2  n =1, 2, 3, 
So,

-8-
22.2.2 Higher Derivative of ge  x
Placing with f = e x in Formula 22.1.1 , we obtain the following formulas.
(n) n
ge x = Σg(r) Bn,re x ,  , e x
r=1
(n) n
ge  = (-1)nΣg(r) Bn,re -x ,  , e -x
-x
r=1
If the first expression is written down to the 5th order, it is as follows.
x (1)
ge 
(1) x
=g e
(2)
ge x = g(1)e x + g(2)e 2x
(3)
ge x = g(1)e x + 3g(2)e 2x + g(3)e 3x
(4)
ge x = g(1)e x + 7g(2)e 2x + 6g(3)e 3x + g(4)e 4x
(5)
ge x = g(1)e x + 15g(2)e 2x +25g(3)e 3x +10g(4)e 4x + g(5)e 5x

Coefficients (1), (1, 1) , (1, 3, 1) , (1, 7, 6, 1) , (1, 15, 25, 10, 1) ,  of these right sides are called
Stirling numbers of the 2nd kind and is given by the following expression.


1 r r
S(n,r) = r Σ(-1)s (r -s) n (2.s)
! s=0 s

Using this notation, the above formulas are more briefly expressed as follows
n
x (n) (r) rx
ge  = ΣS(n,r)g e (2.1+)
r=1
(n) n
ge  = (-1)nΣS(n,r)g(r)e -rx
-x
(2.1-)
r=1

(3)
Example1 tane x

g(1) = tan 2e x +1 , g(2) = 2tan 3e x +2tan e x , g(3) = 6tan 4e x +8tan 2e x +2

Then,
x (3)
tan e  = tan e x+1e x + 32tan e x+2tan e xe + 6tan e x+8tan e x+2e
2 3 2x 4 2 3x

= e xtan e x+1 + 6e tan e x+ tan e x + 2e 3tan e x+4tan e x+1

2 2x 3 3x 4 2

- x (5)
Example2 e e 
-x
(n =1, 2, 3, )
(n)
g = ef , g = ef = ee
So,

-9-
 (5) -x
e e 
-x
= -e e e -x + 15e -2x +25e -3x +10e -4x + e -5x

Note
The horizontal total of Stirling numbers of the 2nd kind is called Bell Number . That is
n
Bn = ΣS(n,r)
r=1

The first few Bell numbers for n =1, 2, 3,  are 1, 2, 5, 15, 52, 203, 877,  .
This Bell number is given by the following expression that is called Dobinski's formula , too.

n n n -r s
1 r n r (-1)
Bn = Σ
e r=0 r !
=Σ Σ
r=1 r ! s=0 s!

22.2.3 Higher Derivative of log f(x)

(r)
g = (-1)r-1(r -1)! f -r from g = log f . Substituting this for Formula 22.1.1 ,
n
(n)
log f(x) = Σ(-1)r-1(r -1)!Bn,rf1 , f2 ,  , fnf -r (2.3)
r=1
If these are written down to the 5th order, it is as follows.
(1)
log f(x) = 0!f1 f -1
log f(x)(2) = 0!f2 f -1 -1!f12f -2
log f(x)(3) = 0!f3 f -1 -1!3f2 f1 f -2 +2!f13f -3
log f(x)(4) = 0!f4 f -1 -1!4f3 f1 +3f22f -2+2!6f2 f12f -3-3!f14 f -4
log f(x)(5) = 0!f5 f -1 -1!5f4 f1 +10f3 f2f -2+2!10f3 f12+15f22f1f -3
-3!10f2 f13 f -4 + 4!f15 f -5

(3)
Example (log sinx)
f = sin x , f1 = cosx , f2 = -sin x , f3 = -cosx ,
So,
3
cos x
(3) sin x cos x (cos x) 3
(log sin x) = -0! + 1!3 2
+ 2! 2
= 2cot x + 2(cot x)
sin x (sin x ) (sin x)

22.2.4 Higher Derivative of g log x

(r)
f = (-1)r-1(r -1)! x -r from f = log x . Substituting this for Formula 22.1.1 ,

 
0! 1! n-1 (n -1)!
n
(n) (r)
g(log x) = Σ g Bn,r , - 2 ,  , (-1) (2.4)
r=1 x x xn
If these are written down to the 4th order, it is as follows.

(1) 1 (1)
g(log x) = g 0!
x
1 (1)
g(log x) = - 2 g 1!- g (0!) 
(2) (2) 2
x

- 10 -
 1
g 2!- 3g(2)1!0!+ g(3)(0!)3
(3) (1)
g(log x) = 3
x
(4) 1
g (0!)4
(1) (2) (3) (4)
42! 0!+3(1!)  + 6g
2
g(log x) = - 4
3! - g 1!(0!)2 - g
x

(4)
Example1 (cos log x)

Sinceg(1) = -sin f , g(2) = -cosf , g(3) = sin f , g(4) = cosf ,

1
(cos log x)(4) = - 4 6g(1) - 11g(2) + 6g(3) - g(4)
x
1 10 cos log x
= - 4 (-6sin f + 11cosf + 6sin f - cosf) = -
x x4
(3)
Example2 (1/log x)
1! 2! 3!
Since g(1) = - , g(2) = , g(3) = - ,
f2 f3 f4

 
1 1 21! 32! 3!
(1/log x)(3) = 2g - 3g(2) + g(3) =
(1)
3 3
- 2
- 3
-
x x f f f4

 (log x) 
1 2 6 6
=- + +
x3 2
(log x)3 (log x)4
(3)
Example3 (log log x)
0! 1! 2!
Since g(1) = , g(2) = - 2 , g(3) = 3 ,
f f f

 
1 1 20! 31! 2!
(log log x)(3) = 3 2g(1) - 3g(2) + g(3) = 3 + + 3
x x f f2 f

 
1 2 3 2
= 3 + +
x log x (log x)2 (log x)3

- 11 -
22.3 Higher Derivative of Gamma Function & the Reciprocal

Formula 22.3.1 ( Masayuki Ui )

When (z) is the gamma function, n(z) is the polygamma function and Bn,rf1 , f2 ,  are
Bell polynomials , the following expressions hold.
n
d n
(z) = (z)ΣBn,k  0(z) ,  1(z) ,  ,  n-1(z) (3.1+)
dz n k =1

dn 1 1 n
n (z)
= Σ (-1)k Bn,k  0(z) ,  1(z) ,  ,  n-1(z) (3.1-)
dz (z) k=1

Proof
When f(z)= log(z) ,
d
f1 = log(z) =  0(z)
dz
d
f2 =  (z) =  1(z)
dz 0

d
fn =  (z) =  n-1(z)
dz n-2
Substituting these for (2.1+) , (2.1-) in Formula 22.2.1 , we obtain
n
( ) (n)
e log z  = e log(z)ΣBn,r 0(z) ,  1(z) ,  ,  n-1(z) (3.1+)
r=1
n
(z) (n)
e -log  = e -log(z)Σ(-1)rBn,r 0(z) ,  1(z) ,  ,  n-1(z) (3.1-)
r=1

Example1  (z)
(4)

We calculate using a formula manipulation software Mathematica according to (3.1+) . When this table
definition is adopted, the result is slightly hard to see, but the differential coefficient is calculable.

Example2 1/(z)
(4)

We calculate using a formula manipulation software Mathematica according to (3.1-) . When this table
definition is adopted, the result is easy to see, but the differential coefficient is incalculable.

- 12 -
Note
On December 9, 2016, I received a mail from Mr. Ui living in Yokohama city. In the mail, it was written that
the coefficients of the Bell polynomials appear in the higher order derivative of the gamma function. I was vary
surprised. Because, it means that the gamma function is a composite function. In fact, it was a too simple
composite function. ( Mr. Ui seems to have noticed it soon, but I needed 3 days to notice it. )
As far as I get to know, the discoverer of Formula 22.3.1 is Mr. Ui. This section is what I added a simple proof
with the consent of Mr. Ui.

- 13 -
22.4 Possibility to super differentiation of composition
Faà di Bruno's Formula was as follows.
n
(n) (r)
gf(x) = Σ
r=1
g Bn,rf1 , f2 ,  , fn (1.1)

Hoppe's Formula was as follows.

(r)

s 
(n) n g r r (n)
gf(x) =Σ Σ (-f)r-sf s (1.2)
r=0 r! s=0

We cannot make the upper limit n of ∑  in both formulas. It is because the number of terms in the right
side cannot become larger than the number of orders in the left side. So, it is hopeless to extend the domain
of (1.1) or (1.2) from the natural number n to the real number p . That is, the super differentiation of the
general composition gf(x ) is impossible now.
However, when the core function f(x) is the 1st degree, according to Formula 22.1.4 ,
(n) (n) n
gf(x) = g f1 (1.4)

So, there is no obstacle in extending this domain from the natural number n to real number p .
Thus, the following expression holds for a real number p >0 .
(p) (p)
gf(x) = g f1p (4.4)
This is the grounds for which we have used "Linear form" since " 12 Super Derivative " as a fait accompli .

2011.01.06
2016.12.26 updated.
K. Kono
Alien's Mathematics

- 14 -
23 Higher Integral of Composition

23.1 Formula of Higher Integral of Composition

Formula 23.1.1
<n> (k)
Let f = f(x), g be the lineal higher primitive function of g(f ) and h , h are the functions of f
such that

dx 1 d kh
k = 1, 2, 3, 
(k)
h= = (1) , h =
df f df k
Let Sk , Mk , rk are the polynomials such that

r  Σ  r  Σ r 
 
m k -1 -1 rk1 rk1 rk2 rk 2 rk k-1 rk k -1
Sk = Σ  Σ k =1, 2, ,n
rk1=0 k1 rk2=0 k2 rk3=0 k3 r =0
kk rkk

 Σ Σ  Σ 

mk mk rk2 rk 2 rk3 rk 3 rk k-1 rk k -1
k =2, 3, ,n
mk
Mk = (-1) Σ
rk2=0 rk 2 rk3=0 rk 3 rk4=0 rk 4 rkk =0 rkk
j
Rj k =Σrik j,k = 1, 2, ,n
i=k

And let a , fa are the zeros of the lineal higher primitive functions of gf(x) , g h respectively. Then,

the lineal higher integral with respect to x of the composition gf(x) is expressed as follows.


x x R
 h
-Rnn
n+ Rn1 Rn1 -Rn2 Rnn
 gf(x )dx
n
= S1S2Sn g h n n-1
h + Rmn 1 (n.1)
a a

    g
f f f

  hdf  hdf  hdf

f
m1 m1 m1
Rmn 1 = (-1)  h df (n-fold nest)
fa fa fa fa

+S M    g
f f f

   
1+ R11 +m2 R11 -R22 +m2 R22
1 2 h h df  hdf hdf
fa fa fa

+S S M   g
f f

 
2+ R21 + m3 R21 -R32+m3 R32 -R33 R33
1 2 3 h h h df  hdf
fa fa


f n-1+ Rn-1 1 + mn Rn-1 1 -Rn2 +mn Rn2 -Rn3 Rnn
+ S1S2Sn-1Mn g h h  h df (n.r)
fa

Proof
df
dx = from g = g ( f ) , f = f(x) .
f(1)
Then, considering the transformation of variabl such that x  f , [a , x] fa , f ,

 
x f g(f)
gf(x)dx = df
a fa f(1)
Here, let
 
f    df 
r r
1 dx
r = 0, 1, 2, 
(r)
h = =
f r (1)
f r
Then,

-1-
   g( f)h( f)df
x f
gf(x)dx =
a fa
According to Formula 16.1.2 in 16.1 ,

 r  g
f m 1-1 -1
1+ r11 r11
g h df = Σ h + Rm11
fa r11=0 11

= (-1)  g
f
m1 m1 m1
Rm11 h df
fa
Using this,

 r 
x m 1-1 -1
g
1+ r11 r11
gf(x)dx =Σ h + Rm11 (1)
a r11=0 11

= (-1)  g
f
m1 m1 m1
Rm11 h df (1r)
fa
Next, integrating both sides of (1) and (1r) with respect to x from a to x ,

 r   g R
x x m 1-1 -1 x x
1+ r11 r11
gf(x)dx = Σ
2 1
h dx + m1dx
a a r11=0 11 a a

 R dx = (-1)   g 
x x f
1 m1 m1 m1
m1 h df dx
a a fa

Substituting dx = h df for this and rewriting the 2nd term of the right side with Rm21 ,

 r   g
x x m 1-1 -1 f
1+ r11 r11
gf(x)dx = Σ
2
h hdf + Rm21
a a r11=0 11 fa

= (-1)   g
 
f f
m1 m1 m1
Rm21 h df hdf
fa fa
Here, according to Formula 20.2.1 in 20.2 ,

 f f f dx = ΣΣ  r   s  f
x m -1 r -1 r <1+ r> (r-s) (s)
1 2 3 1 f2 f3 + Rm1
a r=0 s=0

R = (-1) Σ    f
m m x
1 m <m> (m-s) (s)
m 1 f2 f3 dx
s s=0 a
Then, replacing m with m2 and using this,

  r  r 
f m 2-1 r21 -1 r21
g
2+ r11 +r21  r11 +r21 -r22 r22
g
1+ r11 r11
h hdf = Σ Σ h h + Rm12
fa r21=0 r22=0 21 22

Σ r   g
m
m2 f
m2 2
1+ r11 + m2 r11 -r22 + m2 r22
Rm12 = (-1) h h df
r22=0 22 fa
Substituting this for the above,


x x
2
gf(x)dx
a a

 r  r 
r21
r 
m 1-1 -1 m 2-1 r21 -1
g
2+ r11 +r21  r11 +r21 -r22 r22
=Σ ΣΣ h h
r11=0 11 r =0 r =0
21 22 21 22

r  R + R
-1
m 0-1
+ Σ 1
m1
2
m0
r11=0 11

-2-
 r   g
m2 m2 f
m2 1+ r11 + m2 r11 -r22 + m2 r22
Rm12 = (-1) Σ
r =0
h
fa
h df
22 22

 g h df hdf

f f
m1 m1 m1
Rm21 = (-1)
fa fa
2
Redefining R m ,
1


x x
2
gf(x)dx
a a

 r  r 
r21
r 
m 1-1 -1 m 2-1 r21 -1
g
2+ r11 +r21  r11 +r21 -r22 r22
=Σ ΣΣ h h + Rm21 (2)
r11=0 11 r =0 r =0
21 22 21 22

r   g
m
r 
-1
m 1-1
m2
m2 2 f
1+ r11 + m2 r11 -r22 + m2 r22
Rm21 =Σ (-1) Σ h h df
r11=0 11 r22=0 22 fa

+ (-1)   g
 
f f
m1 m1 m1
h df hdf (2r)
fa fa
Here, let

 r  Σ r  r 
r21 m2
r 
m 1-1 -1 m 2-1 -1 r21
m2
m2
S1 =Σ , S2 =Σ , M2 =(-1) Σ
r11=0 11 r21=0 21 r22=0 22 r22=0 22

R11 = r11 , R21 = r11+ r21 , R22 = r22

Then, using these, (2) and (2r) can be expressed as follows.


x x
2 2+ R21  R21 -R22 R22
gf(x)dx = S 1 S 2 g h h + Rm21
a a


f 1+ R + m
Rm21 = S1 M2 g 11 2 h  11 22 2h 22df
R -R + m R

fa

 g 
f f
m1 m1 m1
+ (-1) h df hdf
fa fa
Next, integrating both sides of (2) and (2r) with respect to x from a to x ,

  r  Σ Σ  r  r 
x x x m 1-1 -1 m 2-1 r21 -1 r21
=Σ
3
gf(x )dx
a a a r11=0 11 r21=0 r22=0 21 22

 R
x x
g
2+ r11 +r21  r11 +r21 -r22 r22

2
h h dx + m1dx
a a

 r     g
m2
r 
x m 1-1 -1 m2 x f


m2 1+ r11 + m2 r11 -r22 + m2 r22
Rm21dx = Σ (-1) Σ h h df dx
a r11=0 11 r =022 22 a fa

  g
f

 
x f
m1 m 1 m 1
+ (-1) h df hdf dx
a fa fa

Substituting dx = h df for this and rewriting the 2nd term of the right side with Rm31 ,


x x x
3
gf(x )dx
a a a

 r  r   g
r21
r 
m 1-1 -1 m 2-1 r21 -1 f
2+ r11 +r21  r11 + r21 -r22 r22
=Σ ΣΣ h h hdf + Rm31
r11=0 11 r =0 r =0
21 22 21 22 fa

-3-
 r   
m2
r 
m 1-1 -1 m2 f f
g

m2 1+ r11 + m2 r11 -r22 + m2 r22
Rm31 = Σ (-1) Σ h h df hdf
r11=0 11 r =022 22 fa fa

  g
f

 
f f
m1 m1 m1
+ (-1) h df hdf hdf
fa fa fa
Here again, according to Formula 20.2.1 in 20.2 ,

  r  s   t  f
x m -1 r s -1 r s <1+ r> (r-s) (s-t) (t)
f1 f2 f3 f4dx = Σ ΣΣ 1 f2 f3 f4 + Rm1
a r=0 s=0 t=0

s  t 
= (-1) ΣΣ  
m s m s x
m <m> (m-s) (s-t) (t)
Rm1 f 1 f2 f3 f4 dx
s=0 t=0 a
Then, replacing m with m3 and using this,

  r  r  r 
f m 3-1 r31 r32 -1 r31 r32
g
2+ r11+ r21  r11+r21-r22 r22
h h hdf = Σ Σ Σ
fa r31=0 r32=0 r33=0 31 32 33
3+ r11+ r21+ r31 r11+ r21+ r31-r22-r32 r22+ r32-r33 r33
g h h h + Rm13

r  r 
m3
m3 r32 m3 r32
Rm13 = (-1) ΣΣ
r32=0 r33=0 32 33


f
g
2+ r11+ r21+m3 r11+r21-r22-r32+m3 r22+r32-r33 r33
 h h h df
fa
Substituting this for the above,


x x x
3
gf(x)dx
a a a

 r  Σ Σ  r  r  Σ Σ Σ  r  r   r 
m 1-1 -1 m 2-1 r21 -1 r21 m 3-1 r31 r32 -1 r31 r32
=Σ
r11=0 11 r21=0 r22=0 21 22 r31=0 r32=0 r33=0 31 32 33

 g
3+ r11+ r21+ r31 r11+ r21+ r31- r22-r32 r22+ r32-r33 r33
h h h

 r  Σ Σ  r  r  R
m 1-1 -1 m 2-1 r21 -1 r21
+Σ 1
m3 + Rm31
r11=0 11 r21=0 r22=0 21 22

r  r 
m3
m3 r32 m2 r32
Rm13 = (-1) ΣΣ
r =0 r =0
32 33 32 33


f
g
2+ r11+ r21+m3 r11+r21-r22-r32+m3 r22+r32-r33 r33
 h h h df
fa

r    g
m2
r 
m 1-1 -1 m2 f f


m2 1+ r11 + m2 r11 - r22 + m2 r22
Rm31 = Σ (-1) Σ h h df hdf
r11=0 11 r =0
22 22 fa fa

  g
f

 
f f
m1 m 1 m 1
+ (-1) h df hdf hdf
fa fa fa

Redefining Rm31 ,


x x x
3
gf(x)dx
a a a

-4-
  r  r   r  r  r 
r21 r31 r32
r 
m 1-1 -1 m 2-1 r21 -1 m 3-1 r31 r32 -1
=Σ ΣΣ ΣΣΣ
r11=0 11 r =0 r =0
21 22 21 22 r =0 r =0 r =0
31 32 33 31 32 33
3+ r11+ r21+ r31 r11+ r21+ r31-r22-r32 r22+ r32-r33 r33
g h h h + Rm31 (3)

r  r 
m3
 r  Σ Σ  r  r 
m 1-1 -1 m 2-1 r21 -1 r21 m3
m3 r32 r32
Rm31 = Σ (-1) ΣΣ
r11=0 11 r21=0 r22=0 21 22 r32=0 r33=0 32 33

 g
f
2+ r11+ r21+m3 r11+r21-r22-r32+m3 r22+r32-r33 r33
h h h df
fa

r   
m2
r 
m 1-1 -1 m2 f f


1+ r11 + m2 r11 -r22 + m2 r 22
g
m2
+Σ (-1) Σ h h df hdf
r11=0 11 r =022 22 fa fa

+ (-1)    g h df hdf hdf

   
f f f
m1 m1 m1
(3r)
fa fa fa
Further, (3) and (3r) are expressed as follows using Sk , Mk , Rjk .


x x x
= S 1S 2S 3 g 
3 3+ R31 R31-R32 R32-R33 R33
gf(x)dx h h h + Rm31
a a a


f
Rm31 = S1S2M3 g 
2+ R21+m3 R21-R32+m3 R32-R33 R33
h h h df
fa

+ S M  g
 
f f
1  + R11+m2 R11-R22+m2 R22
1 2 h h df hdf
fa fa

+ (-1)    g
   
f f f
m1 m1 m1
h df hdf hdf
fa fa fa
Calculating the 4th order integral in a similar way,


x x 4+ R41 R41 -R42 R42 -R43 R43 -R44 R44
 = S1S2S3S4 g
4 4
gf(x )dx h h h h + Rm1
a a


f
= S 1S 2S 3M4 g 
3+ R31+m4 R31-R42+m4 R42-R43 R43-R44 R44
Rm41 h h h h df
fa

+ S S M  g
  hdf
f f
2  + R21+m3 R21-R32+m3 R32-R33 R33
1 2 3 h h h df
fa fa

+ S M   g
   
f f f
1  + R11+m2 R11-R22+m2 R22
1 2 h h df hdf hdf
fa fa fa

+ (-1)     g
     
f f f f
m1 m1 m1
h df hdf hdf hdf
fa fa fa fa
Hereafter, by induction, we obtained the desired expression.

Higher Integral in case the core function is the 1st degree

Although the higher integration of a general composition is complicated in this way, If the core function f(x)
is the 1st degree, it becomes remarkably easy.

Formula 23.1.2
When f(x)=cx +d ,

-5-
    
x x n f f
1
 gf(x)dx
n
=  g( f )df n
a a c fa fa

Proof
When f(x)= cx +d in Formula 23.1.1 ,
dx 1 1 d kh
k = 1, 2, 3, 
(k)
h= = (1) = , h = =0
df f c df k
Then, the remainder term

    g
f f f

  hdf  hdf  hdf

f
m1 m1 m1
Rmn 1 = (-1)  h df (n-fold nest)
fa fa fa fa

+S M    g
f f f

   
1+ R11 +m2 R11 -R22 +m2 R22
1 2 h h df  hdf hdf
fa fa fa

+S S M   g
f f

 
2+ R21 + m3 R21 -R32+m3 R32 -R33 R33
1 2 3 h h h df  hdf
fa fa


f n-1+ Rn-1 1 + mn Rn-1 1 -Rn2 +mn Rn2 -Rn3 Rnn
+ S1S2Sn-1Mn g h h  h df (nr)
fa
becomes as follows.

m1 > 0 , h 
m1
1st line: Since = 0. Then the 1st line = 0.

R22 > 0 , h
R22
2nd line: When = 0. Then, the 2nd line = 0.

R22 = 0 , h
R11 - R22 + m2
When = 0. Then, the 2nd line = 0.

R33 > 0 , h
R33
3rd line: When = 0. Then, the 3rd line = 0.

R33 = 0 , if R32 > 0 , h

R32 - R33
When = 0. Then, the 3rd line = 0.
R21 - R32 + m3
When R33 = R32 = 0 , h = 0. Then, the 3rd line = 0.


h
Rnn
n th line: When Rnn > 0 , = 0. Then, the n th line = 0.

When Rnn = 0 , if at least one is positive in Rn n-1 Rn2 , Then, the n th line = 0.

Rnn = Rn n-1 =  = Rn2 = 0 , h 

Rn-11 - Rn2 + mn
When = 0. Then, the n th line = 0.

Thus, it is sure to become Rmn1 = 0 under the assumption. Therefore,


x x
= S 1S 2 S n g 
n + Rn1 Rn1 -Rn2
 h
Rn n-1 -Rnn Rnn
 gf(x)dx
n
h h (n)
a a
Next, observing (n), we find out Rnn = Rn n-1 =  = Rn1 = 0 immediately. If it is why, in order for (n)

not to be 0, Rnn = 0 is required first. The 2nd, Rnn-1 = 0 is required. The 3rd,  . Hereafter, it advances
one after another, finaly Rn1 = 0 is required. Then,
n

 
Rn1 -Rn2 Rn n-1 -Rnn Rnn (0) n 1
h h h = h  =
c
S1 S2  Sn = 1

-6-
Thus, we obtain

   
n n

 
x x 1 1 f f
 gf(x)dx =
n
g
<n>
=  g( f )df n
a a c c fa fa
Q.E.D

Example When g = log(ax+b )

 
n
x (ax +b)n
   
x 1 n 1
log(ax +b)dx = n
log(ax +b) -Σ
b b a n! k =1 k
- -
a a
This is consistent with linear form in Formula 4.3.3 in " 4 HigherIntegral " .

Compositions that Formula 23.1.1 is applicable

Even if restricted only to the combination of an elementary function, there are many kinds of composition.
However, Formula 23.1.1 is not applicable all of these. The 1st, the inverse function of the core function have to
be known. If it is not so, h cannot be expressed by the function of f . The 2nd, the higher primitive function

g n ( f )= c . If it is not so,
<>
of the enclosing function g( f ) must have the property such as nlim

not the

series but the remainder term becomes the prime part of the integration.
Considering these, there are not many compositions that Formula 23.1.1 is applicable. This decreases the
usefulness of Formula 23.1.1. However, about some compositions that Formula 23.1.1 was applicable, a quite
interesting result was obtained. I describe it in the following sections.

-7-
23.2 Higher Integral of (x^ - c)^

Formula 23.2.1
Let (z) be the gamma function. Let Sk , rk and gf(x) are as follows respectively.

gf(x) = x  - c ,  >0 &  1/k (k =1, 2, ) ,  , , c > 0

r  Σ  r  Σ r   
 -1 rk1 rk1 rk2 rk 2 rk k-1 rk k -1
Sk =Σ  Σ k =1, 2, ,n
rk1=0 k1 rk2=0 k2 rk3=0 k3 r =0
kk rkk
j
Rj k =Σrik j,k = 1, 2, ,n
i=k

Then, the following expressions hold for x> c .

(1+) (1/ )

1+ r

r=0  
x    -1
 
  x c

x - c dx = x - c
 Σ r  1+ +1+ r   1/ - r 
1- 
x
c
(1.1)

 Σ  r  ΣΣ  s
(1+ )
  t   1+ + 2+ r +s
2 -1 -1
x x
 
  x   s s
x - c dx = x - c
2

c 
c  2
r =0 s =0 t =0  

(1/ ) (1/ )
2+ r+s

1- x 
c
 (1.2)
1/ - r - s + t  1/ - t  

 
3
x x
 
  x

3
x - c dx = x - c

c 
c 3

 r  r   r  r  r 
r21 r31 r32
r 
 -1  r21 -1  r31 r32 -1
Σ ΣΣ ΣΣΣ
r11=0 11 r =0 r =0
21 22 21 22 r =0 r =0 r =0
31 32 33 31 32 33
(1+) (1/ )

 1+ + 3+ r11+ r21+ r31  1/ - r11- r21- r31+ r22+ r32
(1/ ) (1/ ) 3+ r11+ r21+ r31

 
c
 1-  (1.3)
 1/ - r22- r32+ r33  1/ - r33 x

n + Rn1
(1+)
  xn
x x  
 
  1
x - c dx = x - c
n
S S Sn 1- 

c 
c  n 1 2  1+ + n + Rn1 x
(1/ ) (1/ ) (1/ )
  (1.n)
 1/ - Rn1+Rn2  1/ - Rn n-1+Rnn  1/ - Rn n

Proof
1
  dx 1 -1
gf(x ) = x - c , h= = (f + c) 
df 
From these,
n+ Rn1 n+ Rn1 (1+)  + n+ Rn1
g = f  = f
 1+ +n + Rn1

-8-
 Rn1 -Rn2

(f + c)  
1
1 -1
Rn1 -Rn2
h =

1
1(1/ ) -1-Rn1 + Rn2
= (f + c) 
  1/ - Rn1+Rn2

1
Rn n-1-Rnn 1(1/ ) -1-R +R
(f + c) 
n n-1 nn
h =
  1/ - Rn n-1+Rnn
1
(1/ )
1 -1-Rnn
h m 1
= (f + c) 
  1/ - Rnn
Substituting these for Formula 23.1.1 (n.1) ,
(1+)
 
x x
   + n+ Rn1
 x - c dx = S1S2Sn
n
f

c 
c  1+ +n + Rn1
1
1(1/ ) -1-Rn1 + Rn2
 (f + c) 
  1/ - Rn1+Rn2

1
1(1/ ) -1-R +R
(f + c) 
n n-1 nn

  1/ - Rn n-1+Rnn
1
(1/ )
1 -1-Rnn
 (f + c)  + Rmn 1
  1/ - Rnn
n
1 (1+) 
 + n+ Rn1
 
-n-Rn1
= n S1S2S n  x - c  
x
  1+ +n + Rn1
(1/ ) (1/ ) (1/ )
  + Rmn 1
 1/ - Rn1+Rn2  1/  - Rn n-1 +R nn  1/ - Rnn
i.e.
n+ R
(1+)
  xn n1
x x  
 
  c
x - c dx = x - c S n
S S 1-

c 
c n
1 2 n
 1+ +n + Rn1 x
(1/ ) (1/ ) (1/ )
  + Rmn 1 (n)
 1/ - Rn1+Rn2  1/ - Rn n-1+Rnn  1/ - Rnn
Since Rmn 1 is too long, we omit it. If (n) is written about the 1st order  the 3rd order, it is as follows.


x
 
x - c dx

c

(1+) (1/ ) 1+ r

r=0  r 
 m 1-1 -1
 
 x c
= x - c Σ
1
1-  + Rm1 (1)
  1+ +1+ r   1/ - r  x
m1 1
(1+) (1/ )

(-1) f  +m -1-m1
Rm11 = f 1
(f + c)  df (1r)
  1+ + m1  1/ - m1 0


x x
 
x - c dx
2
 
c c

-9-
 2 m -1
(1+)
 2 r=0  Σ
s=0 t=0   t 
 x -1 m 2-1 s -1 s
= x  - c
1

Σ r
Σ s  1+ + 2+ r +s 
(1/ ) (1/ ) 2+ r+s

 
c

2
1-  + Rm1 (2)
 1/ - r - s +t   1/ - t  x
(1+) (1/ )
m2 m -1

t
m2
r=0  Σ
(-1) 1 -1 m2
Σ
2
Rm1 =
 2
r t=0  1+ + 1+ r + m2  1/ - r + t - m2
2
(1/ )
f
f  +1+ r+ m -2-r- m2
 2
(f + c)  df
 1/ - t  0

(1+) (1/ )
m1
(-1)
+
 2  1+ + m1  1/ - m1

  
1 1
f f -1-m1 -1
 +m 1 
 f (f + c) df (f + c)  df (2r)
0 0

 
x x
   x3

3
x - c dx = x - c

c 
c 3

 r  Σ Σ  r  r  Σ Σ Σ  r  r   r 
m 1-1 -1 m 2-1 r21 -1 r21 m 3-1 r31 r32 -1 r31 r32
Σ
r11=0 11 r21=0 r22=0 21 22 r31=0 r32=0 r33=0 31 32 33
(1+) (1/ )

 1+ + 3+ r11+ r21+ r31  1/ - r11- r21- r31+r22+r32
(1/ ) (1/ ) 3+ r11+ r21+ r31

 
c
 1-  + Rm31 (3)
 1/ - r22- r32+r33  1/ - r33 x
m3 m -1

r  r 
m3
 r  r 
r21 r32
r 
(-1) 1 -1 m 2-1 r21 -1 m3 r32
Rm31 =
3
Σ
r =0 11
ΣΣ
r =0 r =0 21
ΣΣ
r =0 r =0
11 21 22 22 32 33 32 33
(1+) (1/ )

 1+ +2+ r11+ r21+m3  1/ - r11-r21+r22+r32-m3
(1/ ) (1/ )

 1/ - r22- r32+r33  1/ - r33
3


f  + 2+ r + r +m -3- r11 -r21-m3
 f 11 21 3
(f + c)  df
0
m2 m -1

 r  Σ r 
(-1) 1 -1 m2 m2
+
3
Σ
r11=0 11 r22=0 22
(1+) (1/ ) (1/ )

 1+ +1+ r11+m2  1/ - r11+r22- m2  1/ - r22

  
2 1
f f  + 1+ r +m -2-r11 - m2 -1

 f 11 2
(f + c) df (f + c)  df
0 0
m1
(-1) (1+) (1/ )
+
  1+ +m1  1/ - m1

- 10 -
      
1 1 1
f f f -1-m1 -1 -1
 + m1  
 f (f + c) df (f + c) df (f + c)  df (3r)
0 0 0

Rmn 1 also roughly looks like (3r). The integration included in these is more difficult than the original one (the left
side), and solving these is unthinkable. However, If  >0,  >0 , Rmn 1 0 at the time mi i =1, 2,  .
That is shown as follows.
Now, cnsider the function S(m) such that
1
S(m) =
(1+ m)(1/b - m)
Giving the values to b of this function parametrically and illustrating this, we find out the following. (Refer to the
figure.)
b <0 : S(m) diverges to  .
b =0 : S(m) oscillates around 0.
b >0 : S(m) converges to  .

Although the description was omitted previously, the last term of the remainder Rmn 1 is as follows.
m1
(-1) (1+) (1/ )
  1+ +m1  1/ - m1
n n n

     
f f f  + n-1+ R -n- Rn-11 - mn -1 -1
n-11 m n
+  
  f (f + c) df (f + c) df (f + c)  df
0 0 0
And when  >0 ,  >0 ,
1 1 1 1
lim < lim =0
m  1  1+ +m1  1/ - m1 m 1  1+m1  1/ - m1
Moreover, the integrand is as follows.
n
 + n-1+ Rn-11 + mn -n- Rn-11 - m n
f (f + c) 
n
 + n-1+ Rn-11 + mn 
-n- Rn-11 - mn
= x  - c 
x 

- 11 -
 n mn
 + n-1+ Rn-11 
-n- Rn-11

1- x 
  c
= x - c x  


By assumption, x> c , c >0 . Then,
n mn
 + n-1+ Rn-11 
-n- Rn-11

1- x 
  c
lim x - c
m n
x  
=0

Together with these, if mi i =1, 2, 3,  , the last term of the remainder Rmn 1 have to converge to 0 .
. Next, terms except the last term of the remainder Rmn 1 contain the products as follows.
(1+) (1/ )
 1+ +k -1+ Rk-1 1+ mk  1/ - Rk-11+Rk2-mk
(1/ ) (1/ )
 
 1/ - Rk2+Rk3  1/ - Rkn
k k k

    
f f f  + k-1+ R -k- Rk-11 - mk -1 -1
k-11 m k
+  
  f (f + c) df (f + c) df (f + c)  df
0 0 0
The term that does not include mk among these can be regarded as a constant. Since k -10 ,

The term that include mk is as follows.

1 1
lim
m 
k  1+ +k -1+ Rk-1 1+ mk  1/ - Rk-11+Rk2-mk
1 1
< mlim =0
k  1+ Rk-1 1+ mk 1/ +Rk2 -  Rk-11+mk
And the integrand is as follows.
k
 + k-1+ Rk-11 + mk -k- Rk-11 - mk
lim f (f + c)  =0
m 
k

Together with these, if mi i =1, 2, 3,  , the non-last term of the remainder Rmn 1 have to also converge
to 0 . Therefore, mlim R n = 0 i =1, 2, 3,  .
 m 1
Thus, (1),(2),(3),  ,(n) reduce to (1.1),(1.2),(1.3),  ,(1.n)
i
respectively.

Example1 Area between the third hyperbola and x-axis

Let  =1/3 ,  =3 , c =1 in (1,1) . Then,

(1+1/3) (1/3)

1+ r

   1+1/3+1+ r
x3 x  -1
 
3 1
3Σ
x 3-1 dx = x 3-1 1-
1 r=0 r    1/3- r  x3
3 3 1 1
The function y= x -1 is a part of the third hyperbola 3 - = 1 in the 1st quadran.
x y3
Therefore, this integration means to ask for the function expressing the area between the hyperbola and x-axis in
the 1st quadrant. Although even the elliptic integral is impossible to express this area, this example can express
it with a series. In fact, when both sides are drawn in figure, it is as follows. As the result of the calculation until
the 100th term in the right side, both sides almost overlapp.

- 12 -
 log2
x -
e
Example2 The 2nd integral of
Let  =log 2 ,  =e , c = in (1,2) . Then,


x2
e r=0  Σ
s=0 t=0   t 
x x log2 log2  -1  s -1 s
x -   dx 2 = x e-  
e
2Σ Σ
e

e
 r s
(1+log 2) 1/e 1/e  2+ r+s

 1+log 2+ 2+ r +s   1/e - r - s + t  
 1/e - t 
1- e
x 
The values of the both sides on arbitrary point x = 3.5 are as follows.

- 13 -
23.3 Higher Integral of g (log x)
f = log x , the inverse function is x = e f . Then h = e f r =0, 1,  .
(r )
When the core function is
Thus, the n th order integral of g (log x ) is expressed with n-fold series.

Formula 23.3.1
<n>
When g is the lineal higher primitive function of g(f ) and a , fa are zeros of the lineal higher primitive
f
functions of g(log x) , g ( f)e respectively, the lineal higher integral of g(log x)is expressed as follows.
n n
Σ rk  n +Σ rk
 + Rn

x x m 1-1 m 2-1 m n -1 n
= x n Σ Σ  Σ (-1)k=1
rk
a a
g(log x)dx
n
r1=0 r2=0 rn =0
Π
k=1
k g k=1
m1 (1.n)

    g e df  e df e df  e df
f f f f
m1 f

m1
Rmn1 = (-1)m11 f f f
(n-fold nest)
fa fa fa fa

1 2    g
   
m 1-1 f f f
1
r1+ m2 m1 m2  + r1+m2 2f
+ Σ (-1) e df  e df e df f f
r1=0 fa fa fa

1 2 3   g
 
m 1-1 m 2-1 f f
2 r1+ r2+m3 m1 m2 m3  + r1+ r2+m3 3f
+ Σ Σ (-1) e df  e df f
r1=0 r2=0 fa fa

n -1 n -1

 g e n fdf
m 1-1 m 2-1 m n-1-1 Σ rk+ mn n mk f n -1 + Σ rk+ mn
+ Σ Σ Σ (-1)k =1
Π k k=1
(1.nr)
r1=0 r2=0 r =0n-1 k=1 fa

Proof
Since f = log x , x = e f
dx
(r =1, 2, 3, ,n)
(r)
h= = ef = x , h = ef = x
df
Substituting these for (1), (1,r) in Formula 23.1.1 ,

 r 
x m 1-1 -1
g
1+ r11
g(log x)dx = xΣ + Rm11
a r11=0 11

= (-1)  g
f
m1 m1 f
Rm11 e df
fa

r 
-1 r r m1
Replacing with (-1) 11 and adding 1 11 , 1 to this, we obtain
11


x m 1-1
= xΣ (-1) 11 1 11 g1+
r r r11
g(log x)dx + Rm11 (1.1)
a r11=0

g
f
m m1 m1 f
Rm11 = (-1) 1 1 e df (1.1r)
fa

= e f = x (r =0, 1, 2, ,n) for (2), (2,r)

(r)
Next, substituting h in Formula 23.1.1 ,

  r  r 
r21
r 
x x m 1-1 -1 m 2-1 r21 -1
g
2+ r11+r21
Σ ΣΣ
2 2
g (log x)dx =x + Rm21
a a r =0
11 11 r =0 r =0
21 22 21 22

- 14 -
 r   g
m2
r 
m 1-1 -1 m2
m2 f
1+ r11 + m2 2f
Rm21 = Σ (-1) Σ e df
r11=0 11 r =0
22 22 fa

+ (-1)   g
 
f f
m1 m1 f
e df e fdf
fa fa
Here,

r  r 
-1 r
-1 r
= (-1) 11 , = (-1) 21
11 21

r 
m2
r 
r21 r21 r21
m2
m2
Σ
r =0
=2 , Σ
r =0
=2
22 22 22 22
r m1
Substituting these for the above, and adding 1 11 , 1 to this, we obtain


x x m 1-1 m 2-1
r +r21 r11 r21 2+ r11+r21
g (log x)dx
2
= x 2Σ Σ (-1) 11 1 2 g + Rm21 (1.2)
a a r11 =0 r =0 21

g
m 1-1 f
r + m2 m1 m2 1+ r11 + m2 2f
Rm21 = Σ (-1) 11 1 2 e df
r11=0 fa

1  g
 
f f
m1 m1 m1 f
+ (-1) e df e fdf (1.2r)
fa fa

= e f = x (r =0, 1, 2, ,n) for

(r)
Next, substituting h (3), (3,r) in Formula 23.1.1 ,


x x x
3
g(log x)dx
a a a

 r  r   r  r  r 
r21 r31 r32
r 
m 1-1 -1 m 2-1 r21 -1 m 3-1 r31 r32 -1
= x 3Σ ΣΣ ΣΣΣ
r11=0 11 r =0 r =0
21 22 21 22 r =0 r =0 r =0
31 32 33 31 32 33
3+ r11+ r21+ r31
g + Rm31

r  r 
m3
 r  r 
r21 r32
r 
m 1-1 -1 m 2-1 r21 -1 m3
m3 r32
Rm31 =Σ ΣΣ (-1) ΣΣ
r11=0 11 r =0 r =0
21 22 21 22 r =0 r =0
32 33 32 33

g
f
2+ r11+ r21+m3 3f
 e df
fa

r    g
m2
r 
-1

m 1-1 m2 f f
m2 1+ r11 + m2 2f
+Σ (-1) Σ e df e fdf
r11=0 11 r =0
22 22 fa fa

+ (-1)    g
   
f f f
m1 m1 f f f
e df e df e df
fa fa fa
Here,

r  r  r 
-1 r
-1 r
-1 r
= (-1) 11 , = (-1) 21 , = (-1) 31
11 21 31

r  r  r 
r21 r21 r
r31 r32 r31 r32 r
Σ
r =0
= 2 21 , ΣΣ
r =0 r =0
= 3 31
22 22 32 33 32 33

Σ r  = 2 Σ Σ r  r  = 3
m2m 2 m2
m
m3 r r32 3 32 m3
,
r22=0 22 r32=0 r33=0 32 33

- 15 -
 r m1
Substituting these for the above, and adding 1 11 , 1 to this, we obtain


x x x
3
g (log x)dx
a a a
m 1-1 m 2-1 m 3-1
1 2 21 3 31 g 
r + r21+ r31 r11 r r 3+ r11+ r21+ r31
Σ Σ Σ
3
=x (-1) 11 + Rm31 (1.3)
r =0 r =0 r =0
11 21 31

g
m 1-1 m 2-1 f
r + r21+m3 m1 m2 m3 2+ r11+ r21+m3 3f
Rm31 = Σ Σ (-1) 11 1 2 3 e df
r11=0 r21=0 fa

1 2  g
 
m 1-1 f f
r11+ m2 m1 m2 1+ r11 + m2 2f
+ Σ (-1) e df e fdf
r11=0 fa fa

+ (-1) 1    g
   
f f f
m1 m1 m1 f
e df e fdf e fdf (1.3r)
fa fa fa
Hereafter, by induction, we obtain
n n
Σ rk1 n +Σ  + Rn

x x m 1-1 m 2-1 m n -1 n rk1
rk 1
g (log x)dx
n
= x n Σ Σ  Σ (-1)k=1 Π k g k=1
m1 (1.n)
a a r11=0 r21=0 rn 1=0 k=1

    g e df  e df e df  e df
f f f f
m1 f
Rmn1 = (-1) 1 m1 m1
 f f f
(n-fold nest)
fa fa fa fa

1 2    g
   
m 1-1 f f f
1
r11+ m2 m1 m2  + r11+m2 2f
+ Σ (-1) e df  e df e df f f
r11=0 fa fa fa

1 2 3   g
 
m 1-1 m 2-1 f f
2r11 + r21+m3 m1 m2 m3  + r11+ r21+m3 3f
+ Σ Σ (-1) e df  e df f
r11=0 r21=0 fa fa

n -1 n -1

g e n fdf
m 1-1 m 2-1 m n-1-1 Σ rk1+ mn n mk f n -1 + Σ rk1+ mn
+ Σ Σ Σ (-1)k=1 Π k k=1
(1.nr)
r11=0 r21=0 r =0
n-1 1 k=1 fa
And replacing rk1 with rk , we obtain the desired expression.
Q.E.D.

Although fa is determined by the function g(f ) in Formula 23.3.1 , in many cases, fa = - that is a zero
f
of e . The calculation at the time of applying a concrete function to g(f ) should be a part of this section.
However, since these are very long, these are described in the following independent sections.

- 16 -
23.4 Higher Integrals of (log x) ^ , (log x)^m

23.4.1 Higher Integral of (log x) 

Lemma1

t

When (a , x ) =  >0
a-1 -t
e dt is the incomplete gamma function, the following expressions hold for
x
and   -1, -2, -3,  .
(1) When x  0

x  n -1 Ｃ r (n -r + , - x)( x)r


x x
 x
 x e dx =
n -1
Σ
(-x) r=0
n
- -  +n(n -1)!
(2) When x >0

x  n -1 Ｃ r (n -r + , - x)( x)r


x x
 x
 x e dx =
n -1
Σ
(-x) r=0
n
- -  +n(n -1)!
n -1 n -1Ｃ r (n -r + )( x)
r
+ 2i sin Σ
r=0  +n(n -1)!

Proof
Let g(f ) = f e f , f = x in Formula 23.1.2 . Then,

    f
x x n f f
 x 1  
 ( x) e dx = n
 e df n
- - - -
Here, substitute Formula 16.5.1 in 16.5 for this right side. Then, when x >0 ,
f  n -1 Ｃ r (n -r + , -f)f r

n


x x 1
 ( x) e x dx n =
n -1
Σ
(-f) r=0 (n -1)!
- -
n
Ｃ r (n -r + )f r
   2i sin Σ
1 n -1 n -1
+
r=0 (n -1)!
n
( x) Ｃ r (n -r + , - x)(x)r

1 n -1 n -1
= Σ
(- x) r=0 (n -1)!
n
Ｃ r (n -r + )(x)r
   2i sin Σ
1 n -1 n -1
+
r=0 (n -1)!
i.e.

x  n -1 Ｃ r (n -r + , - x)(x)r

x x
  x
  x e dx =
n -1
Σ
(-x) r=0
n
- -  n(n -1)!
n -1 n -1Ｃ r (n -r + )( x)
r
+ 2i sin Σ
r=0  n(n -1)!
From this, we obtain the desired expressions.

- 17 -
Formula 23.4.1
When (z) is the gamma function, the following expressions hold for a real number   -1, -2, -3,  .
(1+ )

x 
(log x) dx = x 1Σ(-1)r1r (log x)+1+ r
0 r=0 (1+ +1+ r)
 (1+ ) x0
+ (-1) 1Ｃ 1 (1.1)
 1! 1
(1+ )

x x  
(log x) dx 2 = x 2ΣΣ(-1)r+ s 1r2s (log x)+2+ r+s
0 0 r=0 s=0 1+ +2+ r +s 
(1+ )
 
 x1 x0
+ (-1) 2Ｃ 1 - 2Ｃ 2  (1.2)
2! 1 2


x x x   
(log x) dx 3 = x 3ΣΣΣ(-1)r+ s+ t 1r2s3t
0 0 0 r=0 s=0 t=0

(1+ )
 (log x)+3+ r+ s+ t
1+ +3+ r + s + t 
(1+ )
 
 x2 x1 x0
+ (-1) 3Ｃ 1 - 3Ｃ 2  + 3Ｃ 3  (1.3)
3! 1 2 3

n


x x

   Σ rk n rk
 n
(log x) dx = x n
Σ Σ Σ (-1) k =1
Π k
0 0 r =0 r =0
1 2 r =0 n k=1

(1+ )
n
+ n+ Σ rk
 n (log x) k =1


 1+ + n + Σrk
k=1 
(1+ ) n x
n-k
+ (-1) k-1
Σ(-1) nＣ k (1.n)
n! k=1 k

Proof

When the 1st order, substituting g =f , f = log x for (1.1), (1.1r) in Formula 23.3.1 ,
(1+ )

x m -1
(log x) dx = xΣ (-1)r1r (log x)+1+r + Rm1 (1)
0 r=0 1+ +1+r
(1+ )

f

Rm1 = (-1)m1m f +me fdf (1r)
1+ + m  -
If f 0 , from Lemma1,


f
f
+m f f +m
e df = (1+ +m , -f)
- (-f)+m
Substituting this for (1r),

(1+ ) f +m
Rm1 = (-1) m1m (1+ +m , -f)
1+ + m  (-f)+m

- 18 -
 (1+ ) 

 
f
= (-1)m(-1)+m1m (1+ + m , -f)  = (-1)
1+ + m  (-f ) 

i.e.
(1+ +m , -f)
Rm1 = (-1)(1+ )
1+ + m 
Let m  . Then,
(1+ + m , -f)
lim =1
m  1+ + m 
Therefore,

R1 = (-1)(1+ )
If f >0 , from Lemma1,


f
f
+m f f +m
e df = (1+ +m , -f)+ 2i sin( +m)(1+ + m)
- (-f)+m
Substituting this for (1r) ,
(1+ )
f
f
+m f
Rm1 = (-1)m1m e df
1+ + m  -

(1+) f + m
= (-1) m
 1+ + m   (-f) + m
(1+ + m, -f) + 2i sin( +m  )(1+ + m)

(1+)
= (-1)m (-1)--m(1+ + m, -f) + 2i sin( +m  )(1+ + m)
 1+ + m 
(1+ ) (1+ + m, -f)
= + 2i (1+ )(-1)msin( + m )
(-1)  1+ + m
i.e.
(1+ ) (1+ + m, -f)
Rm1 = + 2i (1+ )sin
(-1) 1+ + m 
Let m  . Then,

 
1
R1 = (1+ ) + 2i sin
(-1)
Here,
1

+ 2i sin  = (-1)
(-1)
Therefore R1 = (-1)(1+) . This is the same result as the time of f 0.
Thus, (1) becomes as follows.
(1+ )
 (log x) dx = x Σ(-1) 1 (1+ +1+ r) (log x)
x
 1  r r +1+ r
0 r=0

 (1+ ) x0
+ (-1) 1Ｃ 1 (1.1)
1! 1

When the 2nd order, substituting g =f , f = log x for (1.2), (1.2r) in Formula 23.3.1 ,

(1+)

x x m 1-1 m 2-1
(log x)  dx = x Σ Σ (-1)r+s 1r 2s (log x) 
2 2 +2+ r+s 2
+ Rm1 (2)
0 0 r=0 s=0  1+ +2+r + s 

- 19 -
 (1+)
  f
f f


m m1  +m1 f
Rmn 1 = (-1) 11 e df e fdf
 1+ + m1 - -
(1+)
f
m 1-1 r+ m 2 m 1 m 2 f  +1+ r+ m
+ Σ (-1) 1 2 2 2f
e df (2r)
r=0  1+ +1+ r + m2 -

If f 0 , from Lemma1,
+m1

f
f f
+m1 f
e df = 1+ + m1 , -f
+m1
- (-f)
+1+ r+ m2
1+ +1+ r + m2, -2f
x
f +1+ r+ m2  f f
e df = +1+ r+ m2 +1+ r+ m2+1
- (-f) 2
Substituting these for (2r),
+m1

  (-f) 
(1+) f f
Rm21 = (-1) 1
m1 m1
1+ + m1 , -f e fdf
 1+ + m1 - +m1

 +1+ r+ m
m 1-1 (1+)
r+ m2 m1 m2 f 2
1+ +1+ r + m2, -2f
+ Σ (-1) 1 2
r=0  1+ +1+ r + m2 (-f)  +1+ r+ m 2
2
 +1+ r+ m2 +1

(1+) 1+ +m1 , -2f

 
 +m
e f 1+ +m1 , -f -
m
= (-1) 1(-1) 1
 1+ + m1  +m +1
2 1
m 1-1 r+ m 2  +1+ r+ m2 (1+) 1+ +1+ r + m2, -2f
+ Σ (-1) (-1)
r=0  1+ +1+ r + m2 2+1+ r+1
(1+) 1+ +m1 , -2f
= (-1)
 1+ + m1
e f

1+ +m1 , -f - 
2 +m1+1 
m 1-1
+1 (1+) 1+ +1+ r + m2, -2f
+ Σ (-1)
r=0  1+ +1+ r + m2 2+1+ r+1
i.e.

e f 1+ + m1 , -f 1+ +m1 , -2f

2
Rm1 = 
(-1) (1+)
  1+ + m1
- +m +1
2 1  1+ + m1 
m 1-1 1+ +1+ r + m2, -2f
- (-1)(1+) Σ
r=0 2+2+ r 1+ +1+ r + m2
Let m1 , m2   . Then,

 1
R2 = (-1)(1+) e f - (-1)(1+) Σ +2+ r
r=0 2

 
1
= (-1)(1+) e log x -
2+1
(1+ )
 
 2x 1 x0
= (-1) - 
2! 1 2
If f >0 , from Lemma1,

- 20 -
  +m1

f
f x
 +m 1 f
e df =  +m 1
1+ +m1 , -f + 2i sin  +m1 1+ +m1
- (-f)
+1+ r+ m2
1+ +1+ r + m2, -2f

f +1+ r+ m2  f f
x e df = +1+ r+ m2 +1+ r+ m2+1
- (-f) 2
1+ +1+ r + m2
+ 2i sin +1+ r + m2 +1+ r+ m2+1
2
Substituting these for (2r) ,
+m1

  (-f) 
(1+) f f
Rm21 = (-1) 1
m1 m1
1+ + m1 , -f e fdf
 1+ + m1 - +m1

(1+)
 2i sin +m 1+ +m e df
f
m m1 f
+ (-1) 11
 1+ + m1 -
1 1

 +1+ r+ m
m 1-1 (1+)
r+ m 2 m 1 m 2 f 2
1+ +1+ r + m2, -2f
+ Σ (-1) 1 2
r=0  1+ +1+ r + m2 (-f) +1+ r+ m2 2
 +1+ r+ m2+1

m 1-1 r+ m m m (1+) 1+ +1+ r + m2

+ Σ (-1) 21 12 2 2i sin +1+ r + m2
r=0  1+ +1+ r + m2 2
 +1+ r+ m2+1

(1+) 1+ + m1 , -2f

 
m1
(-1)
= e f
1+ + m1 , -f -
+m1  1+ + m1 
2 +m1+1
(-1)
(1+)
m1
2i sin + m11+ + m1e f
 1+ + m1 
+ (-1)

1+ +1+ r + m2, -2f

r+ m 2
m 1-1 (-1) (1+)
+Σ +1+ r+ m2
r=0 (-1)  1+ +1+ r + m2 2+1+ r+1
m 1-1 (1+) 1+ +1+ r + m2
2i sin +1+ r + m2
r+ m 2
+ Σ (-1)
r=0  1+ +1+ r + m2 2+1+ r+1
i.e.

e f 1+ +m1 , -f 1+ +m1 , -2f

 
(1+)
Rm21 = - +m +1 + 2i sin  (1+) e f
(-1)   1+ + m1 2 1  1+ + m1
1 (1+) m 1-1 1+ +1+ r + m2, -2f
+ Σ
(-1)+1 2+2 r=0 2 r 1+ +1+ r + m2
(1+) m 1-1 1+ +1+ r + m2
+ 2i sin( +1) Σ 
2+2 r=0 2 r 1+ +1+ r + m2
Let m1 , m2   . Then,

(1+) e f
+ 2i sin  (1+) e f
2
R = 
(-1)
1 (1+)  1 (1+)  1
+ Σ + 2i sin( +1) Σ 
(-1)+1 2+2 r=0 2r 2+2 r=0 2r

- 21 -
  (-1) 
1
= 
+ 2i sin  (1+) e log x

(1+)
 (-1) 
1
+ +1
+ 2i sin( +1)
2+1
Here,
1

+ 2i sin  = (-1)
(-1)
Therefore,
(1+)
 
+1 (1+) 2x 1 x0
R2 = (-1)(1+) x + (-1) = (-1) -
2+1 2 1 2
This is the same result as the time of f 0.
Thus, (2) becomes as follows.
(1+ )
 (log x) dx
x x  
 2
= x 2ΣΣ(-1)r+ s 1r2s (log x)+2+ r+s
0 0 r=0 s=0 1+ +2+ r +s 
(1+ )
 
 x1 x0
+ (-1) 2Ｃ 1 - 2Ｃ 2  (1.2)
2! 1 2
Hereafter, by induction, we obtain the general expression.

Example The 3rd integral of (log x)5/2

When the left side of (1.3) is replaced with Riemann-Liouville Integral and the one arbitrary point x =2 is
given, the values of the both sides are as follows.

23.4.2 Higher Integrals of (log x) m

Lemma2
Let  be a positive number, (z) be the gamma function and m,n are non-negative integers.

e x (1+ m)
 r
x x
x m
m -n
 (x)
m- r
m+n Σ
n
e x dx =
- -  r=0 (1+ m - r)

Proof
Let g(f ) = f me f , f = x in Formula 23.1.2 . Then,

    f
x x n f f
m x 1
 ( x) e dx = n
 m f
e df n
- - - -

- 22 -
Here, substitute Formula 16.5.1' in 16.5 for this right side. Then,

(1+ m) m- r
 (x) e
1 n f m -n
 r
x x
m x n
dx = e Σ
r=0 (1+ m - r)
f
- -
i.e.

e x (1+ m)
 r
x x
x m
m -n
 m
 e x dx = n
Σ ( x)
m- r

- -  n r=0 (1+ m - r)
From this, we obtain the desired expression.

Formula 23.4.2
When m =1, 2, 3,  ,


m- r
x m! 1 m (-1)r (log x)
(log x) dx = m
xΣ (2.1)
0 1 r=0 1r m - r  !
r+s
(log x)m+1- r- s

x x m! 2 m +1 m +1- r (-1)
(log x) dx = - x Σ Σ
m 2
(2.2)
0 0 2 r=1 s=0 1r 2s m +1- r - s  !
r+ s+ t
(log x)m+2- r- s- t

x x x m! 3 m +2 m +2- r m +2- r- s (-1)
(log x) dx = m
x Σ Σ Σ 3
0 0 0 3 r=1 s=1 t=0 1r 2s 3t m +2- r - s - t !
(2.3)

n -2 n -1
m + n -1- Σ rk m + n -1- Σ rk


x x m! n m + n -1 m + n -1- r1 k =1 k =1
 (log x)m dx n = (-1)n x Σ Σ  Σ Σ
0 0 n r =1
1 r =1
2 r =1
n-1 r =0n
n n
Σ rk m+ n-1-Σ rk
(-1) k =1 (log x) k =1

 n n (2.n)

m + n -1-Σrk  !
rk
Π
k=1
k
k=1

Proof

When the 1st order, substituting g =f , f = log x for (1.1), (1.1r) in Formula 23.3.1 ,
(1+m)

x m 1-1
(log x)m dx = xΣ (-1)r1r (log x)m+1+r + Rm11
0 r=0 1+m +1+r
(1+m)
1+m + m  
f
m1 m+m1 f
Rm11 = (-1)m11 f e df
1 -
On the other hand, from Lemma2 ,
1+ m +m1 m+m1- r
ex r
f
f m+m1
m +m 1 -1
dx = e f
n
Σ
r=0 1+ m +m1- r
f
-

Substituting this for Rm11 ,

(1+m) 1+ m +m1 m+m1- r

r
m1
m +m 1 -1
Rm11 = (-1)m11
1+m + m1
e f
Σ
r=0 1+ m +m1- r
f

(1+ m)
 r  1+ m +m - r (log x)
m1 m1 log x
m +m 1 -1 m+m1- r
= (-1) 1 e Σ
r=0 1

- 23 -
Therefore,
(1+m)

x m 1-1
(log x) dx = xΣ (-1)r1r
m
(log x)m+1+r
0 r=0 1+m +1+r
m
m +m (1+ m) 1
m+m - r
+ (-1)m11 1 x Σ (-1)r (log x) 1
r=0 1+ m +m1- r
Here, let m1 = 1 . Then,

(1+m)
 (log x)
x
m
dx = x (log x)m+1
0 1+m +1
m +1 (1+ m)
- xΣ (-1)r (log x)m+1- r
r=0 (1+ m +1- r)
(1+m) (1+ m)
=x (log x)m+1 - x (log x)m+1
1+m +1 (1+ m +1)
m +1 (1+ m)
- xΣ (-1)r (log x)m+1- r
r=1 (1+ m +1- r)
m (1+ m)
= xΣ(-1)r (log x)m- r
r=0 (1+ m - r)
i.e.


m- r
x m! 1 m (-1)r (log x)
m
(log x) dx = xΣ (2.1)
0 1 r=0 1r m - r  !

When the 2nd order, substituting g =f , f = log x for (1.2), (1.2r) in Formula 23.3.1 ,
(1+m)

x x m 1-1 m 2-1
m m+2+ r+s
(log x) dx = x Σ Σ (-1)r+s 1r 2s
2 2 2
(log x) + Rm1
0 0 r=0 s=0  1+m +2+r + s 
(1+m)
 1+m + m   
f f m+m


m m1 1 f
Rm21 = (-1) 11 f e df e fdf
1 - -
(1+m)
 1+m +1+ r + m  
m 1-1 r+ m2 m1 m2 f m+1+ r+ m
+ Σ (-1) 1 2 f 2 2f
e df
r=0 2 -
On the other hand, from Lemma2 ,
1+ m +m1 m+m1- r
 r
f
f m+m1
m +m 1 -1
ef dx = e n f
Σ
r=0 1+ m +m1- r
f
-

e
f
2f m+1+ r+ m2 e 2f
f dx = m+1+ r+ m2 +1
- 2
1+ m +1+ r + m2
 
m +1+ r+ m 2 -1
m+1+ r+ m2- s
Σ  (2f)
s=0 s 1+ m +1+ r + m2- s
1+ m +m1- r
  
f
2f m+m - r e 2f m +m - r -1 1
m+m - r- s
e f df = m+m - r+1 Σ
1
(2f) 1
- 2 1 s=0 s 1+ m +m1- r - s
Substituting these for Rm21 ,
(1+m) 1+ m+m1 m+m1- r f
 r=0   
m m1
f m +m 1 -1
Rm21 = (-1) 11 ef Σ f e df
 1+m + m1 - r 1+ m +m1- r

- 24 -
 m 1-1 r+ m 2 m1 m2 (1+m) e 2f
+ Σ (-1) 1 2
r=0  1+m +1+ r + m2 2m+1+ r+ m2+1
1+ m +1+ r + m2
s
m +1+ r+ m 2 -1 m+1+ r+ m2- s
 Σ (2f)
s=0 1+ m +1+ r + m2- s
(1+ m)
r=0  r  1+ m +m - r 
m m1
m +m 1 -1 f m+m1- r
Σ
2f
= (-1) 11 e f df
1 -
m+1+ r+ m2 - s
e 2f m +1+ r + m2
(1+ m)
 s   1+ m +1+ r + m - s
m 1-1 r+ m 2 m1 m2 -1 f
+ Σ (-1) 1 2 Σ
r=0 21 s=0  2  2s
x
(1+ m) e 2log
r=0   1+ m +m - r 2
m1 m1
m +m 1 -1
= (-1) 1 Σ r 1
m+m1- r+1

1+ m+m1- r
s
m +m 1- r -1

m+m 1- r- s
Σ
s=0 1+ m +m1- r - s
(2f)

2log x m +1+ r + m2 m+1+ r+ m2- s

(1+ m)
 s   1+ m +1+ r + m - s
m 1-1 r+ m2 m1 m2 e -1 f
+ Σ (-1) 1 2 Σ
r=0 2 s=0  2  2s
m+m 1- r- s
m1 m1 x 2 m +m 1 m +m 1- r (1+ m) (log x)
= (-1) 1 Σ
2 r=0
(-1)r Σ (-1)s
s=0 1+ m+m1- r - s 2s
m+1+ r+ m 2- s
(1+ m)
m +1+ r+ m 2
m 1-1 r+ m2 m1 m2 x2 (log x )
+ Σ (-1) 1 2 Σ (-1) s

r=0 2 s=0 1+ m +1+ r + m2- s 2s

Therefore,
(1+m)

x x m 1-1 m 2-1
m m+2+ r+s
Σ Σ
2 2
(log x) dx = x (-1)r+s 1r 2s (log x)
0 0 r=0 s=0  1+m +2+r + s 
m+m1- r- s
m1 m1 x 2 m +m 1m +m 1- r (1+ m) (log x)
+ (-1) 1 Σ
2 r=0 s=0 Σ (-1)r+s
1+ m+m1- r - s 2s
m+1+ r+ m2- s
(1+ m)
m +1+ r+ m 2
m 1-1 r+ m 2 m 1 m 2 x2 (log x )
+ Σ (-1) 1 2 Σ (-1) s
r=0 2 s=0 1+ m +1+ r + m2- s 2s
Here, let m1 = m2 = 1 , Then,
(1+m)
 (log x) dx
x x
m 2 2 m+2
= x (log x)
0 0  1+m +2
x 2 m +1 m +1- r (1+ m) (log x) m+1- r- s
2Σ Σ (-1) (1+ m +1- r - s)
r+s
-
r=0 s=0 2s
m +2 (1+ m) (log x) m+2- s
Σ (-1)s+1
2
+x
s=0 (1+ m +2- s) 2s
2 (1+m) m+2 2 (1+ m) (log x) m+2
= x (log x) -x
 1+m+2 (1+ m +2) 20
m +1 (1+ m) (log x) m+1- s
Σ (-1)s
2
-x
s=0 (1+ m +1- s) 2s+1

- 25 -
 x 2 m +1 m +1- r (1+ m) (log x) m+1- r- s
2Σ Σ (-1) (1+ m +1- r - s)
r+s
-
r=1 s=0 2s
m +2 (1+ m) (log x) m+2- s
Σ
2 s+1
+x (-1)
s=1 (1+ m +2- s) 2s
x 2 m +1 m +1- r (1+ m) (log x) m+1- r- s
2 Σ Σ (-1) (1+ m+1- r - s)
r+s
=-
r=1 s=0 2s
i.e.
r+s
(log x)m+1- r- s

x x m! 2 m +1 m +1- r (-1)
m 2
(log x) dx = - x Σ Σ (2.2)
0 0 2 r=1 s=0 1r 2s m +1- r - s  !
Hereafter, by induction, we obtain the general expression.

Example The 3rd integral of (log x)4

When both the sides are calculated in (2.3) at m=4 , it is as follows.

- 26 -
23.5 Higher Integral of 1 / log x

Formula 23.5.1
When (z), (z ),  denote the gamma function, the digamma function, Euler-Mascheroni-Constant
(= 0.57721566...) respectively, the following expressions hold for x 0 .

r r log|log x|- (1+ r)-   + log 1

 log x
x 1 
dx = xΣ(-1) 1 (log x) r + (1.1)
0 r=0 (1+ r) 0!
log|log x|- (2+ r +s)-
 log x dx = x ΣΣ(-1)
x x
1 2  
r+s r s
2
12 (log x) 1+ r+s
0 0 r=0 s=0 (2+ r +s)
1
( + log 1)x - ( + log 2)
1! 
+ (1.2)


x x x 1   
r+ s+ t r s t
dx 3 = x 3ΣΣΣ(-1) 12 3
0 0 0 log x r=0 s=0 t=0

log|log x|- (3+ r + s + t)-

 (log x) 2+ r+ s+ t
(3+ r + s + t)
1
+ ( + log 1)x 2 -2( + log 2)x +( + log 3) (1.3)
2!

n


x x 1    Σ rk n r
dx x Σ Σ Σ
n
= n
 (-1) k =1
Π kk
0 0 log x r1=0 r2=0 rn=0 k=1
n


log|log x|- n +Σrk -   k =1  (log x)
n-1+Σ rk
k =1
n


 n +Σrk
k =1 
1 n -1
+ Σ (-1)rn -1Ｃ r + log(r +1)x n-1-r
(1.n)
(n -1)! r=0

Calculation
Let g = 1/f . Then, for f > 0,
log f - (k) -  k-1
k =1, 2, 3, 
<k>
g = (log f)<k -1> = f
(k)
Therefore,
n

g
n + Σ
k=1 
rk
n

=

log f - n + Σrk - 
k =1  f
n
n+ Σ rk-1
k=1
n


 n + Σrk
k=1 
log f - m1 -  m1-1
g
m1
= f
m1
log f - 1+ r1+ m2 -  1+ r1+ m2-1
g
1+ r1+ m2
= f
1+ r1+ m2


- 27 -
 n -1

g

n -1
n -1 + Σ rk+ mn
k =1 = 
log f - n -1+ Σrk+ mn - 
k =1  f
n -1
n-1 + Σ rk+ mn -1
k=1
n -1


 n -1+ Σrk+ mn
k=1 
Substitute these for Formula 23.3.1 and returning f to x in the non-remainder terms,
since [- , f ]  [0, x ] ,
n
Σ rk n r

x x 1 m 1-1 m 2-1 m n -1
dx = x Σ Σ  Σ (-1)k=1 Πk k
n n
0 0 log x r1=0 r2=0 rn =0 k=1
n


log|log x| - n + Σrk -   k=1  (log x) k=1
n
n+ Σ rk-1
+ Rmn1 (n)
n

  n + Σrk
k =1 
log f - m1 - 
       
f f f f
m m m 1-1 f
Rmn 1 = (-1) 11 1 f e df  e fdf e fdf e fdf
- - - - m1
m 1-1 r + m2 m1 m2
+ Σ (-1) 1 1 2
r1=0
log f -1+ r1+ m2 - 
     e df  e df
f f f r 1 + m2 2f
  f e df f f
- - - 1+ r1+ m2

n -1
m 1-1 m 2-1 m n-1-1 Σ r k+ m n n mk
+ Σ Σ Σ (-1)k =1 Π k
r1=0 r2=0 r =0
n-1 k =1
n -1


log f - n -1 + Σrk+ mn - 

n -1
Σ rk+ mn -1

f n-1 +
k=1
 n -1
f k=1
e nf df (n,r)


 n -1 + Σrk+ mn

-
k=1
When the 1st order,
log|log x|- 1+ r  - 

x 1 m 1-1
dx = x Σ (-1)r 1r (log x) r + Rm11
0 log x r=0 1+ r 
f log f - m1 -  m -1

m m
Rm11 = (-1) 1 1 1 f 1 e fdf
-
m1
Here, surprisingly, when m1  ,
log f - m1 -  m1-1 f

f
e df = 
1 m m1
lim R = mlim (-1) 1 1 f (1r)
m 
1
m1  1
-
m1
Therefore,
log|log x|- (1+ r)- 

x 1 
dx = xΣ(-1)r 1r (log x) r +  (1.1)
0 log x r=0 (1+ r)
When the 2nd order,

 log|log x| - 2+ r +s  - 
x x 1 m 1-1 m 2-1
dx 2 = x 2 Σ Σ (-1)r+s1r2s (log x) r+s-1
2+
0 0 log x r=0 s=0  2+ r +s 
2
+ Rm1

- 28 -
 log f - 1+ r + m2 -  1+ r+ m2-1 2f

m 1-1
r+ m2 m1 m2 f
Rm21 = Σ (-1) 1 2 f e df
r=0 - 1+ r + m2
f log f - m1 - 
  
f
m m1 m -1
+ (-1) 11 f 1 e fdf e fdf
- - m1
Here, surprisingly again, when m1 , m2  ,
lim Rm21 =  x - ( + log 2)
m i
i =1, 2 (2r)

Therefore,
log|log x|- 2+ r +s  -

x x 1  
r+s
dx 2= x 2Σ Σ(-1) 1r2s (log x) 1+ r+s
0 0 log x r=0 s=0 2+ r + s
+ ( + log 1) x - ( + log 2) (1.2)
Also, when the 3rd order,


x x x 1 m 1-1 m 2-1 m 3-1 r + r2+ r3 r1 r r
dx 3 = x 3 Σ Σ Σ (-1) 1 1 2 23 3
0 0 0 log x r1=0 r2=0 r3=0

log f - 3+ r1+ r2+ r3 -  2+ r1 + r2 + r3

 f + Rm31
 3+ r1+ r2+ r3
log f - 2+ r1+ r2+m3 - 

m 1-1 m 2-1 r + r 2 +m3 m 1 m2 m3 f 1+ r1 + r2 +m3 3f
Rm31 = Σ Σ (-1) 1 1 2 3 e df f
r1=0 r2=0 -  2+ r1+ r2+m3
log f - 1+ r + m  - 
1 2  
 
m 1-1 r1+ m2 m1 m2 f f r1 + m2 2f
1 2 f
+ Σ (-1) f e df e df
r1=0  1+ r + m 
- - 1 2

log f - m  - 
+ (-1) 1   
   
f f f 1 m1-1 f
m1 m1 f f
f e df e df e df
 m 
- - - 1
And
  + log 2  + log 3
lim R 3 =
m  m1
x2 - x+ i =1, 2, 3 (3r)
i 2! 1! 2!
Therefore,


x x x 1   
r+ s+ t r s t
dx 3 = x 3ΣΣΣ(-1) 12 3
0 0 0 log x r=0 s=0 t=0

log|log x| - (3+ r + s + t)-

 (log x) 2+ r+ s+ t
(3+ r + s + t)
1
+ ( + log 1)x 2 -2( + log 2)x +( + log 3) (1.3)
2!
Hereafter, by induction, we obtain the general expression.

Although the problems are (1r), (2r), (3r),  , these proofs are difficult. However, surely they hold on
the numerical computation

Example The 2nd integral of 1/log x

When both sides of (1.2) are illustrated, it is as follows. Since both sides overlapp exactly, the left side (blue)
is not visible.

- 29 -
Note


x


1
Therefore, the higher integral of the logarithmic integral li(x) = dt can be expressed
0 log t
as follows.
n

 li(x)dx
x x
n+1    Σ rk n +1 r

0 0
n
=x Σ Σ Σ
r =0 r =0 r =0
(-1) k =1
Π k

k =1
k
1 2 n

n +1


log|log x|- 1+ n +Σrk -   k =1  n+Σ rk
(log x)
n +1

k =1
n +1


 1+ n +Σrk
k =1 
1 n
+ Σ (-1)rnＣ r + log(r +1)x n-r
(1.n)
n! r=0

Formula 14.4.3 ( 14.4 ) using exponential integral Ei(x) is easier as the formula of the higher integral of
li(x) , In contrast, Formula 23.5.1 looks truly complicated and not useful. However, both have a decisive
difference. That is well understood if (1.n) is rewritten using a harmonic number Hn and a factorial ! as follows.
n log|log x| -H n +1 n +1
n+Σ rk

 li(x)dx
x x    Σ rk n +1 n+Σ rk
rk
Σ Σ(-1)
n n+1
Σ Π
=
k 1 k =1 k =1
=x k n +1
(log x)
n +Σr !
0 0 r =0 r =0 r =0
1 2 n k =1
k
k =1
1 n
+ Σ (-1)rnＣ r + log(r +1)x n-r
(1.n)
n! r=0
That is, (1.n) is expressed by the elementary functions while Formula 14.4.3 is expressed by non-elementary
functions. Here is the meaning of Formula 23.5.1 .

Now, Formula 23.5.1 can be simplified a little more. For the purpose, the following two lemmas are necessary.

Lemma3
 1r r 1 1
Σ
r=0
(-1)
(1+ r)
x =r
0! e x

- 30 -
 1r2s
e 
  1 1 1
r+s r+s
Σ Σ(-1) (2+ r +s)
x 1+ =
1! x
-
r=0 s=0 e 2x
1r2s3t
e 
   1 1 2 1
r+s+t r+s+t
Σ ΣΣ(-1) (3+ r + s +t)
x 2+ =
2! x
- 2x
+
r=0 s=0 t=0 e e 3x

n
rk
  
n
Σ rk Π k n
n-1+Σ rk
1 n -1 n -1Ｃ r
Σ (-1)k =1
k=1
Σ Σ n x k=1
= Σ (-1)r 1+ r
 
(n -1)! r=0
r =0 r =0
1 r =0
2 n  n +Σrk e
k=1

Proof
By induction. Details are discussed in one chapter of "À la carte".

Replacing x with log x in Lemma3 , we obtain the following Lemma immediately.

Lemma3'

1  1r r x0
x Σ(-1) (log x) = r
r=0 (1+ r) 0!
2  r+s  1r2s (x -1)1
x ΣΣ(-1) (log x)1+ r+s =
r=0 s=0 (2+ r +s) 1!
3   
r+s+t 1r2s3t 2+ r+s+t (x -1)2
x Σ ΣΣ(-1)
r=0 s=0 t=0 (3+ r +s + t)
(log x) =
2!

n
rk
  
n
Σ rk Π k n-1+Σ rk
n
(x -1)n-1
Σ Σ
k=1
x n
Σ (-1) k=1
n (log x) k =1
= (3'.n)

 
(n -1)!
r =0 r =0
1 r =0
2 n  n +Σrk
k=1

Using this Lemma3' , Formula 23.5.1 can be simplified as follows.

Formula 23.5.1'
When (z), (z ) denote the gamma function and the digamma function respectively, the following
expressions hold for x 0 .
 (1+ r)

x 1 
dx = -xΣ(-1)r1r (log x) r
0 log x r=0 (1+ r)
1
(x -1) log|log x|+ x log 1
0 0
+ (1.1')
0!
 (2+ r + s)

x x 1  
r+s
dx 2 = -x 2ΣΣ(-1) 1r2s (log x) 1+ r+s
0 0 log x r=0 s=0 ( 2+ r s
+ )
1
+ (x -1)1 log|log x|+ x 1log 1 - x 0log 2 (1.2')
1!

- 31 -
 
x x x 1   
r+ s+ t r s t
dx 3 = -x 3ΣΣΣ(-1) 1 23
0 0 0 log x r=0 s=0 t=0
 (3+ r + s + t)
 (log x)2+ r+ s+ t
(3+ r + s + t)
1
+ (x -1)2 log|log x|+ x 2log 1-2x 1log 2+ x 0log 3 (1.3)
2!

n
n

 n +Σrk
 (log x) n


x x    Σ rk n r n-1+Σ rk
1 k =1
dx n = -x nΣ Σ Σ(-1) k =1 Πk k n
k =1

 
log x
0 0 r1=0 r2=0 rn=0 k=1
 n +Σr k
k =1
1 n -1
+
(n -1)! 
(x -1) n-1 log|log x| -Σ(-1)n-rn -1Ｃ r x r log(n -r)
r=0 
(1.n')

Proof
Substituting Lemma3' (3'.n) for Formula 23.5.1 (1.n) ,
n
n

 n +Σrk
 (log x) n


x x    Σ rk n r n-1+Σ rk
1 k =1
dx n = -x nΣ Σ Σ(-1) k =1 Πk k n
k =1

 
log x
0 0 r1=0 r2=0 rn=0 k=1
 n +Σr k
k =1
n rk
  
n
Σ rk Π k n
n-1+Σ rk
+ log|log x| - x nΣ Σ Σ(-1) k =1
k=1 k =1
n
(log x)

 n +Σrk

r1=0 r2=0 rn=0
k =1
1 n -1
+ Σ (-1)rn -1Ｃ r + log(r +1)x n-1-r
(n -1)! r=0
n


= -x nΣ Σ Σ(-1)
  Σ
n
rk n rk 
 n +Σrk
 (log x)
k =1
n
n-1+Σ rk
Π
=
k 1 k =1
k n


 n +Σr

r1=0 r2=0 rn=0 k=1
k
k =1
log|log x| 
+ (x -1) n-1 - (x -1) n-1
(n -1) ! (n -1) !
 n -1
n-1-r 1 n -1
+ Σ (-1)r
n Ｃ
-1 rx + Σ (-1)rn -1Ｃ r x n-1-rlog(r +1)
(n -1)! r=0 (n -1)! r=0
n

n    Σ
n
rk n rk 
 n +Σrk
 (log x)
k =1
n
n-1+Σ rk
Σ Σ Σ(-1) Π
=
k 1 k =1
= -x k n


 n +Σr

r =0 r =0 r =0
1 2 n k=1
k
k =1
1 n -1
+
(n -1)! 
(x -1) n-1 log log x - Σ(-1)n-rn -1Ｃ r x rlog(n -r)
r=0 
n -1 n -1

 Σ(-1)rn -1Ｃ r x n-1-rlog(r +1)= -Σ(-1)n- rn -1Ｃ r x rlog(n -r)

r =0 r =0

- 32 -
Example The 3rd order integral of 1/log x (The 2nd order integral of li(x) )
When both sides of (1.3') are illustrated, it is as follows. Since both sides overlapp exactly, the left side (blue)
is not visible.

- 33 -
23.6 Higher Integral of log|log x|

Formula 23.6.1
When (z), (z ),  denote the gamma function, the digamma function, Euler-Mascheroni Constant
(= 0.57721566...) respectively, the following expressions hold for x 0 .
log|log x| - 2+ r - 
 log|log x|dx = xΣ(-1) 1
x 
(log x) r - 
r r 1+
(1.1)
0 r  2+ r 
r+s r s log|log x | - 3+ r +s  - 
 log|log x|dx = x ΣΣ(-1)
x x  
(log x) r+s
2 2 2+
12
0 0 r=0 s=0  3+ r +s 
1
- 2( + log 1) x - ( + log 2) (1.2)
2!

 log|log x|dx

x x x
3
0 0 0

3    log|log x| - 4+ r +s +t  - 
Σ Σ Σ (log x ) r+s+t
3+
=x (-1)r+ s+ t1r 2s 3t
r=0 s=0 t=0  4+ r +s +t 
1
3  + log 1x 2 -3 + log 2x +  + log 3
3!  
- (1.3)

 log|log x|dx

x x
n
0 0
n


= x nΣ Σ Σ(-1)
  Σ
n
rk n rk
log|log x| - 1+ n +Σrk - 
 k =1  n
n+Σ r k
Π
=
k 1 k =1
k n
(log x)

 1+ n +Σrk

r1=0 r2=0 rn=0 k =1
k =1
1 n
(-1)rnＣ r( + logr) x n-r
n! Σ
+ (1.n)
r=1

Calculation
Let g = log f , f = log x . Then,
n

g

n
n +Σ rk
k =1 = log f - 1+ n +Σrk -  k=1  f
n
n + Σ rk
k=1
n
 1+n +Σrk
k =1  
log f -   1+ m 1 - 
g 1 =
m m
f 1
1+m1
log f - 1+1+ r1+ m2 -  1+ r1+m2
g  1 2 =
1+ r +m
f
1+1+ r1+m2
log f - 1+2+ r1+ r2+m3 -  2+ r1+ r2+m3
g  1 2 3 =
2+ r + r +m
f
1+2+ r1+ r2+ m3
n -1

g

n -1
n -1 + Σ rk+ mn
k =1 = 
logf - 1+ n -1+Σrk+ mn - 
k=1  f
n -1
n -1 + Σ rk+ mn
k =1
n -1


 1+n -1+Σrk+ mn
k =1 
- 34 -
Substitute these for Formula 23.3.1 and returning f to x in the non-remainder terms,
since [- , f ]  [0, x ] ,
n
Σ rk

x x m 1-1 m 2-1 m n -1 n
 log|log x|dx n = x n Σ Σ  Σ (-1)k=1
rk
0 0 r1=0 r2=0 rn =0
Π
k=1
k
n
log|log x|- 1+ n +Σrk -   k=1  (log x)
n +Σ rk
k =1
n

+ Rmn1 (n)
n


 1+n +Σrk
k =1 
Rmn1 =
log f - 1+m1- 
       
f f f f m1 f
m m1
(-1) 11  f e df  e fdf e fdf e fdf
- - - -  1+m1
m 1-1
r + m2 m1 m2
+ Σ (-1) 1 1 2
r1=0

log f - 2+ r1+m2 - 

     
f f f

1+ r1+m2 2f
 f e df  e fdf e fdf
- - -  2+ r1+m2
m 1-1 m 2-1
r + r2+m3 m1 m2 m3
+ Σ Σ (-1) 1 1 2 3
r1=0 r2=0

log f - 3+ r1+ r2+m3 - 

  
f f
 e df  e fdf
2+ r1+ r2+m3 3f
 f
- -  3+ r1+ r2+m3

n -1
m 1-1 m 2-1 m n-1-1 Σ rk+ mn n mk
+ Σ Σ Σ (-1)k =1
Π k
r1=0 r2=0 r =0
n-1 k=1
n-1

 
n -1
logf - n + Σrk + mn -  n -1 + Σ rk+ mn

f

k =1
n-1
f k=1
e nfdf (nr)


 n + Σrk + mn 
-
k =1
When the 1st order,
log|log x| - 2+ r - 

x m 1-1
log|log x| dx = x Σ (-1)r 1r (log x) r + Rm1
1+ 1
0 r  2+ r 
log|f|- 1+m1-  m1 f

1 m m1 f
Rm1 = (-1) 1 1 f e df
-  1+m1
Here, surprisingly, when m1  ,
log|f|- 1+ m1 -  m1 f

f
f e df = -
1 m m1
lim R = mlim (-1) 1 1 (1r)
m 
1
m1  1
-
1+m1
Therefore,
log|log x| - 2+ r - 
 log|log x|dx = xΣ(-1) 1
x 
(log x) r - 
r r 1+
(1.1)
0 r  2+ r 
When the 2nd order,

- 35 -
  log|log x|dx
x x
2
0 0
m 1-1 m 2-1 log|log x| - 3+ r +s  - 
Σ Σ (log x) r+s + Rm1
2 2+ 2
=x (-1)r+s1r2s
r=0 s=0  3+ r +s 
f log |f|- 2+ r + m2 -  1+ r+ m

m 1-1 r+ m m m
Rm21 = Σ (-1) 21 12 2 f 2 2f
e df
r=0 -  2+ r + m2
f log |f|- 1+m1- 
  
m m f m
+ (-1) 11 1 f 1e fdf e fdf
- -  1+m1
Here, surprisingly again, when m1 , m2  ,
1
lim R 2 = - 2( + log 1)x - ( + log 2)
2! 
m  m1
(i =1, 2) (2r)
i

Therefore,
log|log x| - 3+ r +s  - 
 log|log x|dx 2
x x 
= x Σ Σ(-1)r+s1r2s (log x) r+s
2 2+
0 0 r=0 s=0  3+ r +s 
1
- 2( + log 1) x - ( + log 2) (1.2)
2!
Also, when the 3rd order,

 log|log x|dx

x x x
3
0 0 0
m 1-1 m 2-1 m 3-1 log|log x| - 4+ r +s +t  - 
Σ Σ Σ (log x) r+s+t
3 3+
=x (-1)r+ s+ t1r 2s 3t
r=0 s=0 t=0  4+ r +s +t 
3
+ Rm1
log|f|- 3+ r + s +m3 -  2+ r+ s+m3 3f

m 1-1 m 2-1 r+ s+m3 m1 m2 m3 f
Rm31 = Σ Σ (-1) 1 2 3 f e df
r=0 s=0 -  3+ r + s +m3
f log |f|- 2+ r + m2 -  1+ r+ m
1 2  
 
m 1-1 r+ m2 m1 m2 f
+ Σ (-1) f e df e fdf
2 2f

r=0 - -  2+ r + m2

log|f|- 1+m1-  m1 f
+ (-1) 1   
   
f f f
m1 m1
f e df e fdf e fdf
- - -  1+m1
And
1
lim R 3 = -
m  m1
3 + log 1x 2 -3 + log 2x +  + log 3 (i =1, 2, 3) (3r)
i 3!
Therefore,

 log|log x|dx

x x x
3
0 0 0

3    log|log x | - 4+ r +s +t  - 
Σ Σ Σ (log x) r+s+t
3+
=x (-1)r+ s+ t1r 2s 3t
r=0 s=0 t=0  4+ +s +t 
r
1
- 3 + log 1x 2 -3 + log 2x +  + log 3 (1.3)
3!
Hereafter, by induction, we obtain the general expression.

Although the problems are (1r), (2r), (3r),  , these proofs are difficult. However, surely they hold on
the numerical computation.

- 36 -
Example The 2nd integral of loglog x
When the one arbitrary point x =4.7 is given in (1.2) , the values of the both sides are as follows.

Note
Formula 14.5.1 (14.5 ) using Ei(x) is easier as the formula of the higher integral of log|log(x)| . In contrast,
Formula 23.6.1 looks truly complicated and not useful. However, both have a decisive difference. For example,
if (1.2) is rewritten using a harmonic number Hn and a factorial ! , it is as follows.


x x   log|log x| -H2+ r+s
log|log x| dx 2 = x 2Σ Σ(-1)r+s1r2s (log x) r+s
2+
0 0 r=0 s=0 2+ r +s  !
1
- 2( + log 1) x - ( + log 2) (1.2)
2!
That is, (1.2) is expressed by the elementary functions while Formula 14.5.1 is expressed by non-elementary
functions. Here is the meaning of Formula 23.6.1.

Now, Formula 23.6.1 can be simplified a little more. For the purpose, the following two lemmas are necessary.

Lemma4
1r
 
 1 1
r
Σ (-1) r
(2+ r)
x 1+ =
1!
1- x
r=0 e
r s

 
  12 1 2 1
Σ Σ (-1)r+s r+s
x 2+ = 1- x + 2x
r=0 s=0 (3+ r +s) 2! e e
r s t

 
   123 1 3 3 1
Σ Σ Σ (-1)r+s+t x 3+
r+s+t
= 1- x + 2x - 3x
r=0 s=0 t=0 (4+ r + s +t) 3! e e e

n
rk
  
n
Σ rk Π k n
n+Σ rk
1 n nＣ r
Σ Σ Σ (-1)k =1
k =1
n x k=1
= Σ (-1)r r
 
r1=0 r2=0 rn =0
 1+ n +Σrk n ! r=0 e
k =1

Proof
By induction. Details are discussed in one chapter of "À la carte". In addition, this can be drawn also from
Lemma3.

Replacing x with log x in Lemma4 , we obtain the following Lemma immediately.

- 37 -
Lemma4'
1r
 
 1 1
Σ (-1) (log x)1+ r =
r
1-
r=0 (2+ r) 1! x
1r2s 1 2
 
  1
r+s 2+ r+s
Σ Σ(-1) (3+ r +s) (log x)
r=0 s=0
=
2!
1-
x
r s t 3

 
   123 1 1
Σ Σ Σ (-1)r+s+t (log x)3+ r+s+t = 1-
r=0 s=0 t=0 (4+ r + s +t) 3! x

n
rk
   Σ
n
rk Π k n+Σ rk
n
n

 
1 1
Σ Σ Σ (-1)k =1 k =1 k=1
n (log x) = 1- (4'.n)

 
r1=0 r2=0 rn =0
 1+ n +Σrk n! x
k =1

Using this Lemma4' , Formula 23.6.1 can be simplified as follows.

Formula 23.6.1'
When (z), (z ),  denote the gamma function, the digamma function, Euler-Mascheroni-Constant
(= 0.57721566...) respectively, the following expressions hold for x 0 .
 2+ r 
 log|log x|dx = -xΣ(-1) 1
x 
(log x ) r
r r 1+
0 r  2+ r 
1
log|log x| -  (x -1) -( + log1) x 
1 0
+ (1.1')
1!
 3+ r +s 

x x  
log|log x | dx 2 = -x 2ΣΣ(-1)r+s1r2s (log x) r+s
2+
0 0 r=0 s=0  3+ r +s 
1
 
2
log|log x | -  (x -1) +Σ(-1) 2Ｃ r( + logr) x
2 r 2-r
+ (1.2')
2! r=1
 4+ r +s +t 

x x x   
log|log x| dx 3 = -x 3ΣΣΣ(-1)r+ s+ t1r 2s 3t (log x) r+s+t
3+
0 0 0 r=0 s=0 t=0  4+ r +s +t 
1
 
3
log|log x | -  (x -1) +Σ(-1) 3Ｃ r( + logr) x
3 r 3-r
+ (1.3')
3! r=1

n

  (log x)n Σ r n  1+n +Σrk n


x x    Σ rk n rk k =1
+ k
 log|log x | dx = -x ΣΣ Σ  ( n
-1 ) Π k n k 1=
n
k =1


 1+n +Σr

0 0 r1=0 r2=0 rn=0 k=1
k
k =1
1 n

n!   (1.n')
log|log x | -  (x -1) +Σ(-1) Ｃ ( + logr) x
n r n-r
+ n r
r=1

Proof
Substituting Lemma4' (4'.n) for Formula 23.6.1 (1.n) ,

  log|log x|dx
x x
 n
0 0

- 38 -
 n

n    Σ
n
rk n rk 
 1+ n +Σrk
 (log x)
k =1
n
n+Σ rk
Σ Σ Σ(-1) Π
=
k 1 k =1
= -x k n


 1+ n +Σr

r =0 r =0 r =0
1 2 n k =1
k
k =1
n rk
   Σ
n
rk Π k n
n+Σ rk
+ log|log x| -  x nΣ Σ Σ(-1) =
k 1 k=1 k =1

n
(log x)

 1+ n +Σrk

r1=0 r2=0 rn=0
k =1
1 n
(-1)rnＣ r( + logr) x n-r
n! Σ
+
r=1
n

   Σ
n
rk n rk 
 1+ n +Σrk  (log x)n+Σ r n
k

Σ Σ
k =1
= -x n
Σ (-1) k =1
Π k n
k =1

r =0 r =0
1 2 r =0 n k =1
 1+ n +Σrk
k =1
n n

 
x 1 1 n
+ log|log x|-  1- + Σ (-1)rnＣ r( + logr)x n-r
n! x n! r=1
n


= -x nΣ Σ Σ(-1)
  Σ
n
rk n rk 
 1+ n +Σrk  (log x)n+Σ r
k =1
n
k
k =1
Πk n
k =1

r1=0 r2=0 rn=0 k =1

 1+ n +Σrk
k =1

1 n


log |log x|-  (x -1) +Σ(-1) nＣ r( + logr)x

n r n-r
+
n! r=1

Example The 3rd order integral of log|log x|

When the one arbitrary point x =4.9 is given in (1.3') , the values of the both sides are as follows.

- 39 -
23.7 Possibility to super integral of composition
As understood from Formula23.1.1 , the n th order integral of composision g{f(x )} becomes n(n+1)/2 - fold
series in general. Therefore, for example, the 1.5 th order integral of composition g{f(x )} will be expressed by
1.875-fold series. However, such a multiple series does not exist. i.e. it is impossible to make Formula23.1.1
Super Integral.
However, when the core function f(x) is the 1st degree, according to Formula 23.1.2 ,

   
x x n f f
1
 gf(x)dx
n
=  g(f)df n f(x)= cx +d
a a c fa fa
So, there is no obstacle in extending this domain from the natural number n to real number p .
Thus, the following expression holds for a real number p >0 .

   
x x p f f
1
 gf(x)dx
p
=  g(f)df p f(x)= cx +d
a a c fa fa
(1.3p)
This is the grounds for which we have used "Linear form" since " 07 Super Integral " as a fait accompli .

2011.01.15
K. Kono
Alien's Mathematics

- 40 -
24 Sugioka's Theorem on the Series of Higher (Repeated) Integrals
Mikio Sugioka discovered that the series of the higher integral of arbitrary functions results in one integral in
March, 2003. ( http://www5b.biglobe.ne.jp/~sugi_m/page030.htm ). I was introducing these theorems in the
one section in my work " 16 Higher Integral of the product of two functions ". Since Mr. Sugioka discovered
many theorems after that, they can no longer be housed in one section. Then, I decided to prepare a chapter
independent here and to introduce his wonderful theorems.
Although Mr. Sugioka named the thorem "An operator's theorem", in this paper, I named like the title so that
the contents might be understood well.

24.1 Series of the n-th order Integrals

In this section, we ask for the sum of the following series of higher integrals.

 f(x)dx  f(x)dx + f(x)dx  f(x)dx

x x x x x x x x
+ 
2 3 4
a a a a a a a a

Theorem 24.1.1 ( Sugioka's Theorem 3 )

When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

 f(x)dx  f(x)e   f(x)dx  e

r=1 
m x x x x x x
Σ r x -x
=e x
dx - e m -x
dx (1.1)
a a a a a a

 f(x)dx r = e  f(x)e
r=1 
 x x x
Σ x -x
dx (1.1')
a a a

Proof
Let

 f(x)dx
x x
r =1, 2,  ,m
<r> r
f (x) = (1.r)
a a

f (a) = 0 (r =1, 2,  ,m).

r <>
Then So,

 f f
x x x x
f(x)e -xdx = f <1>(x)e -xa + <1>
(x)e -x = f <1>(x)e -x + <1>
(x)e -x
a a a

f
x x
= f <1>(x)e -x + f <2>(x)e -xa + <2>
(x)e -x
a

f
x
= f <1>(x)e -x + f <2>(x)e -x + <2>
(x)e -x
a


f
m x
= e -x Σf < r>
(x) +
<m> -x
e dx
r=1 a
Substituting (1.r) for this,

     f(x)dx  e
x m x x x x x
f(x)e dx = e -x Σ 
-x
f(x)dx r +  m -x
dx
a r=1 a a a a a

Multiplying by ex the both sides, we obtain (1.1) .

Next, according to Theorem 4.1.3 ( 4.1 ), the following expression holds.

  f(x)dx
m -1 (x -a)r x x
f
<m>
(x) = Σ f <m-r>
(a) +  m
(1.m)
r=0 r! a a

-1-
On the other hand, since f(x)is analytic function on I , this can be expanded to Taylor series as follows.
 (r) (x -a)r  (x -a)r
f(x) = Σf (a) = Σf <-r>
(a)
r=0 r! r=0 r!
Shifting by m the index of the integration operator of f(x)
<m> 
<m-r> (x -a)r
f (x) = Σf (a) (1.t)
r=0 r!
Comparing the right-hand sides of (1.m) and (1.t) , we obtain

 f(x)dx
x x
m
lim
m 
=0
a a
Thus, (1.1') holds.

Relation with a differential equation ( Sugioka's Theorem 2 )

The right side of the theorem is transformed as follows.

 f(x)e f(x)e
x x
ex -x x
dx = e-x
dx
a a

y1(x) e f(x)e
x
= e xe -xy1(x)a x
dx
-x

= y1(x)- y1(a)e x-a

And y1(x) is transformed as follows.

(-1)dx (-1)dxdx
  
-
y1(x)= e x f(x)e -xdx = e f(x) e
Then, we notice that this is a particular solution of the 1st order differential equation y'(x)- y(x)= f(x).

   dx
x x x x x x
Example 1 dx + dx 2 + 3
+
a a a a a a

 dx = m !
x x m
(x -a) m
f(x)=1 ,
a a
Substituting these for (1.1) and using the incomplete gamma function (x, y ) ,

 dx e 
r=1 
m x x x x (x -a)m -x
Σ r
=e x -x
dx - e x
m!
e dx
a a a a
x
x e x
= e x-e -xa - -e -a(m +1, x -a)a
m!
That is

 dx
x-a

r=1 
m x x e
Σ (m +1, x -a) - (m +1, 0)
r x-a
=e -1 + (1.2)
a a m!
x-a
e
And, since lim (m +1, x -a ) - (m +1, 0) = 0 ,
m  m !

 dx r = e x-a -1
r=1 
 x x
Σ (1.2')
a a
The higher order integral of the left side of (1.2) is as follows.

-2-
 a dx
m x x m (x - a)r
Σ
r=1 a
r
=Σ
r=1 r!
Using this as the left side of (1.2) and (1.2') ,
r x-a
m (x - a ) e
(m +1)-(m +1 , x -a)
x-a
Σ
r=1 r!
=e -1 -
m!
(1.3)

(x - a)r

Σ = e x-a - 1 (1.3')
r=1 r!
When x =2, a =1, m =4 , Both sides of (1.3) are calculated as follows.

Also, it is clear from the following that (1.3') is correct.


r 
r 0
(x - a) (x - a) (x - a)
Σ =Σ - = e x-a -1
r=1 r! r=0 r! 0!

   e dx + 
x x x x x x
Example 2 e x dx + e x dx 2 + x 3
a a a a a a

 e dx = e - e Σ r !
x x r
x (x -a)
x m x a
m -1
f(x)= e ,
a a r=0

Substituting these for (1.1) and using the incomplete gamma function (x, y ) ,

r=1 
     e dx  e dx
m x x x x x x
Σ  e xdx r = e x e xe -xdx - e x  x m -x
a a a a a a

= e  dx - e  e - e Σ
r! 
x x r
x (x -a) x x a
m -1
-x
e dx
a a r=0

= e (x -a) - e  e e dx + e Σ
r! 
1 x m -1 x
x x x -x x+ a r -x
(x -a) e dx
a r=0 a

= e (x -a) - e  e e dx - e Σ
x m -1 -a
e
(r +1, x -a)- (r +1, 0)
x x x -x x+ a
r! a r=0
m -1 (r +1, x -a)- (r +1)
= e x(x -a) - e x(x -a) - e x Σ
r=0 r!
i.e.
(r +1, x -a)
 e xdx r = e x(x -a)- e x(x -a)- e x Σ 
r=1  
m x x m -1
Σ r=0 r!
-1 (1.4)
a a

Next,

-3-
 n -1
xs
(n , x) = (n -1)! e Σ -x
s=0 s!
From this,
(r +1, x -a) r (x -a)s
r!
=e -(x-a )
Σ
s=0 s!
Using this,
(r +1, x -a)
 e
r=1  r=0  
m x x m -1
Σ x r
dx = -e x
Σ r!
-1
a a

Σ 
m -1 r (x -a)s
= - e xe -(x-a)Σ - e (x-a)
r=0 s=0 s!
Here,

   
m -1 r (x -a)s (m -r)(x -a)r
m -1
Σ
r=0
Σ
s=0 s !
- e (x-a)
r=0
=Σ
r !
- e (x-a)
m -1 r(x -a )r m -1 m(x -a )r
= -Σ +Σ - m e (x-a)
r=0 r! r=0 r!
m -1 (x -a )r-1 m -1 (x -a )r
= -(x -a)Σ + mΣ - m e (x-a)
r=1 (r -1)! r=0 r!
Substituting this for the above,
(r +1, x -a)
 e
r=1   
m x x m -1
Σ x r
dx = -e x Σ
r=0 r!
-1
a a

 
m -1 (x -a )
r-1 m -1 (x -a )r
= e xe -(x-a) (x -a)Σ - mΣ + m e (x-a)
r=1 (r -1) ! r=0 r !

When m  ,

 e xdx r = e xe -(x-a)(x -a)e (x-a) - me (x-a) + m e (x-a)

 x x
Σ
r=1 a a
That is,

 e xdx r = e x(x -a)

 x x
Σ
r=1
(1.4')
a a

Note

 e dx = e r=1 

x x  x x
If we adopt the lineal higher integral
x r x
then Σ e x dx r =  .
- - - -
Therefore, it is for a collateral higher integral that this theorem is significant.

A by-product

Σ  = lim Σ  
m -1 r (x -a)s m -1 (m -r)(x -a)r
lim Σ
m  r=0
- e (x-a)
m  r=0
- e (x-a)
s=0 s! r!
= -e (x-a)(x -a)

-4-
    log xdx
x x x x x x
Example 3 log xdx + log xdx 2 + 3
+
0 0 0 0 0 0
Substituting f(x)= log x , a =0 for (1.1') ,

 log xdx r = e x log xe -xdx

 x x x
Σ
r=1 0 0
0
The integral of the right side is as follows according to Formula 16.6.2 ( 16.6 )


x x
log xe -xdx = -e -x log|x|+ Ei(-x)0
0
-x
= -e log|x|+ Ei(-x) - 


x et
Where, Ei(x) = dt ,  = 0.5772 (Euler-Mascheroni Constant ).
- t
This integral is not lineal as understood from this constant. Multiplying by e x the both sides, we obtain

r=1 
 log xdx
 x x
Σ  r
= -log|x|+ e xEi(-x) -  (1.6')
0 0
The higher integral of the left side is the following lineal integral.

 log xdx

x xn
log|x| -Σ s 
x n 1
n
=
0 0 n! s=1
Then, (1.6') becomes
 r

  = -log|x|+ e Ei(-x)- 
x r 1
Σ
r=1 r!
log|x| -Σ
s=1 s
x

When the first 10 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

-5-
Theorem 24.1.2 ( Sugioka's Theorem 4 )
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

(-1)r-1 f(x)dx = e  f(x)e dx -(-1)me   f(x)dx e dx

m x x x x x x

a  a 
Σ
r=1 a a
r -x
a
x -x
a
m x

(2.1)

(-1)r-1 f(x)dx r = e -x f(x)e xdx

 x x x
Σ
r=1
(2.1')
a a a

Proof
In a similar way to the proof of Theorem 24.1.1 , we obtain the desired expressions.

     sin xdx

x x x x x x x x
Example 4 sin xdx - sin xdx 2 + sin xdx 3 -  4
+- 
a a a a a a a a
Substituting f(x)= sin x for (2.1') ,

 sin xdx  sin xe dx

 x x x
Σ
r=1
(-1) r-1 r
=e -x x
a a a
The right side is as follows according to Formula 16.5.3 ( 16.5 ).

 x  x
 sin x e dx =  sin 4  e sin x - 4  
x
x
a a
 
    
2
= e xsin x - - e asin a -
2 4 4

Here, 2 sin x - 4  = sin x -cosx . Then

 sin x e dx = sin x 2cosx e

x - sin a -cosa a
x x
- e
a 2
-x
Multiplying this by e , we obtain

(-1)r-1 sin xdx r =

 x x 1
Σ
r=1 2
sin x -cosx -(sin a -cosa)e 
a-x
(2.2')
a a
The higher integral of the left side is as follows according to Theorem 4.1.3 ( 4.1 )

n  (n -s)
 sin x dx
n -1 (x -a )s

  -Σ  
x x
n
= sin x - sin a -
a a 2 s=0 s! 2
Then, the left side is
r  (r -s)
sin x - 2  -Σ 
 (x -a)s

r-1
Σ
r=1
(-1) r-1
s=0 s!
sin a -
2
When a =3/2 , the first 25 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

-6-
-7-
24.2 Series of the odd-th order Integrals
In this section, we ask for the sum of the following series of higher integrals.

     f(x)dx
x x x x x x x x
f(x)dx  f(x)dx +  f(x)dx   + 
3 5 7
a a a a a a a a

Theorem 24.2.1
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

a f(x)dx 2   f(x)e dx

m x x 1 x x x
Σ x
2r-1
= -x
e f(x)e dx + e -x
r=1 a a a

  
e x +(-1)-2m+1e
x x x -x

  2m-1
- (-1)2m-1cosh x f(x)dx dx
a a a 2

sinh x  f(x)dx

x -x
x x x -(-1)-2m+1e
 
e 2m-1
+ (-1)2m-1 dx
a a a 2
(1.1)

     f(x)e dx
 x x 1 x x x
Σ  f(x)dx 2r-1 = e f(x)e -xdx + e -x x
(1.1')
r=1 a a 2 a a

Proof
Formula of repeated integration by parts was as follows. (" 01 Generalized Taylor's Theorem " ( A la Carte ) )

 
x m x x
f(x)g(x)dx = (-1) f <r>(x)g ( )(x)a + (-1)m f <m>(x)g (x)dx
r-1 r-1 (m)
a r=1
Σ a
When g(x)= cosh x , sinh x ,
(r-1) e x +(-1)-r+1e -x (m) e x +(-1)-me -x
(cosh x) = , (cosh x) =
2 2
(r-1) e x -(-1)-r+1e -x (m) e x -(-1)-me -x
(sinh x) = , (sinh x) =
2 2
Substituting these for the above,
x

 f(x)cosh xdx = Σ(-1)  

x m
r-1 <r> e x +(-1)-r+1e -x
f (x) ,
a r=1 2 a

f
m
x
<m> e x +(-1)-me -x
+ (-1) (x) dx ,
a 2
x

 f(x)sinh xdx = Σ(-1)  

x m
r-1 e x -(-1)-r+1e -x
<r>
f (x) ,
a r=1 2 a

f
m
x
<m> e x -(-1)-me -x
+ (-1) (x) dx ,
a 2
Here, let

  f(x)dx
x x
<r>
f (x) =  r
r =1, 2,  ,m
a a

f (a) = 0 (r =1, 2,  ,m) ,

<>r
Since

-8-
    f(x)dx
x m e x +(-1)-r+1e -x x x
f(x)cosh xdx = Σ(-1)
r-1
 r
a r=1 2 a a

  f(x)dx 
m
x x x
m e x +(-1)-me -x
+ (-1) dx
a a a 2

  f(x)dx
x x -r+1 -x x x
r-1
m e +(-1) e
f(x)cosh xdx = Σ(-1) r
a r=1 2 a a

  f(x)dx 
m
x x x
m e x -(-1)-me -x
+ (-1) dx
a a a 2
Expanding a part of the 1st term of the right side,
x -r+1 -x
r-1 e +(-1) e
m
Σ
r=1
(-1)
2
= cosh x - sinh x + cosh x - sinh x +-

m
r-1 e x-(-1)-r+1e -x
Σ
r=1
(-1)
2
= sinh x - cosh x + sinh x - cosh x +-
Multiplying both sides by cosh x, sinh x respectively,

e x+(-1)-r+1e
m
-x
cosh xΣ(-1) r-1
2 2
= cosh x - cosh x sinh x + cosh x - cosh x sinh x +-
r=1 2
e x-(-1)-r+1e
m
-x
sinh xΣ(-1) r-1 2 2
= sinh x - sinh x cosh x + sinh x - sinh x cosh x +-
r=1 2
The 2nd term of the right side is as follows.

 
e x +(-1)-me
x x x -x
m
(-1) cosh x
 a a a
f(x)dx m
2
dx

(-1) sinh x  f(x)dx

x x x x -m -x

 
m e -(-1) m e
dx
a 2 a a
2 2
Since cosh x - sinh x = 1 ,

    f(x)dx
x x m x x
cosh x f(x)cosh xdx - sinh x f(x)sinh xdx = Σ  2r-1
a a r=1 a a

 
e x +(-1)-2m+1e
x x x -x


2m-1
+ (-1)2m-1cosh x f(x)dx dx
a a a 2

sinh x  f(x)dx

x -x
x x x -(-1)-2m+1e
 
2m-1 e 2m-1
- (-1) dx
a a a 2
Here, m in the 2nd term and the 3rd term on the right side has to be an odd number corresponding to the 1st
term. Furthermore,

  f(x)sinh xdx
x x
cosh x f(x)cosh xdx - sinh x
a a

 
e x+e -x x e x+e -x e x-e -x x e x-e -x
= f(x) dx - f(x) dx
2 a 2 2 a 2

  f(x)e  f(x)e dx

1 x x x
-x x
= e dx + e -x
2 a a

-9-
Using this, we obtain

a f(x)dx    f(x)e dx

m x x 1 x x x
Σ x
2r-1
= e f(x)e -xdx + e -x
r=1 a 2 a a

  
e x +(-1)-2m+1e
x x x -x

  2m-1
- (-1)2m-1cosh x f(x)dx dx
a a a 2

sinh x  f(x)dx

x -x
x x x -(-1)-2m+1e
 
2m-1 e 2m-1
+ (-1) dx
a a a 2

 f(x)dx
x x
m
And since lim =0 holds by assumption ( See proof of Theorem 24.1.1 ), we obtain (1.1').
m  a a

     dx
x x x x x x x x
dx  dx +  dx   
3 5 7
Example1
a a a a a a a a


x x
m (x -a)m
f(x)=1 , dx =
a a m!
Substituting these for (1.1) ,

a dx  e  e dx
m x x 1 x x x
Σ x
2r-1 -x -x
= e dx + e
r=1 a 2 a a


2m-1
x (x -a)2m-1 e x +(-1)-2m+1e -x
- (-1) cosh x dx
a (2m -1)! 2


2m-1
x (x -a)2m-1 e x -(-1)-2m+1e -x
+ (-1) sinh x dx
a (2m -1)! 2
Here,

 e  e dx
1 x x x ex -x x ex x x
e -x
dx + e -x x
= -e a + e a
2 a a 2 2
x-a
e - e -(x-a)
= = sinh(x -a)
2
Then,

  dx
m x x
Σ  2r-1
= sinh(x -a)
r=1 a a


2m-1
x (x -a)2m-1 e x +(-1)-2m+1e -x
- (-1) cosh x dx
a (2m -1)! 2


2m-1
x (x -a)2m-1 e x -(-1)-2m+1e -x
+ (-1) sinh x dx
a (2m -1)! 2
(1.2)

  dx
 x x
Σ  2r-1
= sinh(x -a) (1.2')
r=1 a a
The left side is

a dx
m x x m (x - a)2r-1
Σ =Σ
2r-1
r=1 a r=1 (2r -1)!

- 10 -
Using this as the left side of (1.2) and (1.2') ,
2r-1
m (x -a)
Σ
r=1 (2r -1)!
= sinh(x -a)


2m-1
x (x -a)2m-1 e x +(-1)-2m+1e -x
- (-1) cosh x dx
a (2m -1)! 2


x (x -a)2m-1 e x -(-1)-2m+1e -x
+ (-1)2m-1sinh x dx
a (2m -1)! 2
(1.3)
2r-1
 (x -a)
Σ
r=1 (2r -1)!
= sinh(x -a) (1.3')

When x =3, a =1, m =5 , Both sides of (1.3) are calculated as follows.

Also, (1.3') is correct. Because the left side is Taylor expansion of the right side.

Theorem 24.2.2 ( Sugioka's Theorem 5 )

When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

   f(x)sin xdx
m x x x x
Σ(-1)
r-1
 2r-1
f(x)dx = cosx f(x)cosxdx + sin x
r=1 a a a a
2m -1 
  f(x)dx  cos x + ( 2 )  dx
x x x
2m-1
- (-1)2m-1cos x
a a a
(2m -1)
sin x  f(x)dx
x x x
sin x +  dx
 
2m-1
- (-1)2m-1
a a a 2
(2.1)

   f(x)sin xdx
 x x x x
Σ (-1) r-1
 f(x)dx 2r-1 = cosx f(x)cosxdx + sin x (2.1')
r=1 a a a a

Proof
Formula of repeated integration by parts was as follows (" 1 Generalized Taylor's Theorem ") .

 
x m x x
f(x)g(x)dx = (-1) f <r>(x)g ( )(x)a + (-1)m f <m>(x)g (x)dx
r-1 r-1 (m)
a r=1
Σ a
When g(x)= cosx , sin x ,

- 11 -
 (r -1) m
(cosx)(r-1) = cos x +   (cosx) m = cos x +  
( )
,
2 2
(r -1) m
(sin x)(r-1)   (sin x) m = sin x +  
( )
= sin x + ,
2 2
Substituting these for the above,

(r -1) x

 f(x)cosxdx = Σ(-1)   
x m
r-1 <r>
f (x)cos x +
a r=1 2 a
m
  
x
+ (-1)m f <m>(x)cos x + dx
a 2
(r -1) x

 f(x)sin xdx = Σ(-1)   2 

x m
r-1 <r>
f (x)sin x +
a r=1 a
m
+ (-1)  f (x)sin  x +
2 
x
m <m>
dx
a
Here, let

  f(x)dx
x x
<r>
f (x) =  r
r =1, 2,  ,m
a a

f (a) = 0 (r =1, 2,  ,m) ,

<>r
Since

(r -1)
    f(x)dx
x m x x
r-1
f(x)cosxdx = Σ(-1) cos x +  r
a r=1 2 a a
m
+ (-1)   f(x)dx cos  x +
  2 
x x x
m m
dx
a a a
(r -1)
 f(x)sin xdx = Σ(-1) 2  
sin x +
x m x x
r-1
 f(x)dx r
a r=1 a a
m
+ (-1)   f(x)dx sin  x +
  2 
x x x
m m
dx
a a a
Expanding a part of the 1st term of the right side,
(r -1)
  = cosx + sin x - cosx - sin x ++--
m
r-1
Σ
r=1
(-1) cos x +
2
(r -1)
  = sin x - cosx - sin x + cosx ++--
m
Σ (-1) r-1 sin x +
r=1 2
Multiplying both sides by
cos x, sin x respectively,
(r -1) 
 
m
cos xΣ(-1) r-1 cos x +
2 2
= cos x + sin x cos x - cos x - sin x cos x ++--
r=1 2
(r -1) 
sin x +  = sin x - sin x cos x - sin x + sin x cos x ++--
m
sin xΣ(-1) r-1
2 2
r=1 2
The 2nd term of the right side is as follows.

   f(x)dx  cos x + m2 dx

x x x
(-1)mcos x  m
a a a

(-1) sin x  f(x)dx sin x +

x m x x

  2 
m m
dx
a a a

- 12 -
 2 2
Since cos x + sin x = 1 ,

    f(x)dx
x x m x x
cosx f(x)cosxdx + sin x f(x)sin xdx = Σ(-1)
r-1
 2r-1
a a r=1 a a
2m -1 
  f(x)dx  cos x + ( 2 )  dx
x x x
2m-1
+ (-1)2m-1cos x
a a a
(2m -1)
sin x  f(x)dx
x x x

  sin x + 2  dx
2m-1
+ (-1)2m-1
a a a
Here, m in the 2nd term and the 3rd term on the right side has to be an odd number corresponding to the 1st
term. Then, transposing the remainder terms , we obtain (2.1) .

  f(x)dx
x x
And since lim
m  a
 m
=0 holds by assumption ( See proof of Theorem 24.1.1 ), we obtain (2.1').
a

     e dx
x x x x x x x x
e xdx  e xdx +  e xdx   x
+- 
3 5 7
Example2
a a a a a a a a

Substituting f(x)= e x for (2.1') ,

   e sin xdx
 x x x x
Σ(-1) 
r-1
e xdx 2r-1 = cosx e xcosxdx + sin x x
r=1 a a a a
The right side is as follow.


x 1 x
x
e cosxdx = e (cosx + sin x)-e a (cosa + sin a)
a 2


x 1
e xsin xdx = - e x(cosx - sin x)-e a(cosa - sin a)
a 2
From these,

 
x
x
x
x ex ea
cosx e cosxdx + sin x e sin xdx = - cos(x -a)- sin(x -a)
a a 2 2
Therefore,

 e dx

r-1
x x ex ea
Σ x 2r-1
(-1) = - cos(x -a)- sin(x -a)
2
(2.2')
r=1 a a 2
The higher integral of the left side is as follows according to Theorem 4.1.3 ( 4.1 )


x x (x -a)s n -1
e dx = e - Σe
x n x a
a a s=0 s!
Then, (2.2') becomes
s

e 
 2r-2 (x -a ) ex ea
Σ
r=1
(-1) r-1 x
-e a
Σ
s=0 s!
=
2
-
2
cos(x -a ) - sin(x -a ) (2.3')

When a =0.3 , the first 20 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

- 13 -
- 14 -
24.3 Series of the even-th order Integrals
In this section, we ask for the sum of the following series of higher integrals.

     f(x)dx
x x x x x x x x
f(x)dx   f(x)dx +  f(x)dx   + 
2 4 6 8
a a a a a a a a

Theorem 24.3.1
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

a f(x)dx e  f(x)e  f(x)e dx

m x x 1 x x x
Σ x
2r -x
= dx - e -x
r=1 a 2 a a

- sinh x  f(x)dx

  cosh x dx
x x x
2m
a a a

+ cosh x  f(x)dx

  sinh x dx
x x x
2m
a a a
(1.1)

a f(x)dx    f(x)e dx

 x x 1 x x x
Σ x
2r
= e f(x)e -xdx - e -x (1.1')
r=1 a 2 a a

Proof
The following expressions were obtained during the proof of Theorem 24.2.1.

 f(x)cosh xdx = Σ(-1)  f(x)dx

x x -r+1 -x x x
r-1 e +(-1) e
m
r
a r=1 2 a a

  f(x)dx 
m
x x x
m e x +(-1)-me -x
+ (-1) dx
a a a 2

  f(x)dx
x x -r+1 -x x x
r-1
m e -(-1) e
f(x)sinh xdx = Σ(-1) r
a r=1 2 a a

   f(x)dx 
x x x e x -(-1)-me -x
+ (-1)m  m
dx
a a a 2
Expanding a part of the 1st term of the right side,
x -r+1 -x
r-1 e +(-1) e
m
Σ
r=1
(-1)
2
= cosh x - sinh x + cosh x - sinh x +-

m
r-1 e x-(-1)-r+1e -x
Σ
r=1
(-1)
2
= sinh x - cosh x + sinh x - cosh x +-
Multiplying both sides by sinh x, cosh x respectively,

e x+(-1)-r+1e
m
-x
sinh xΣ(-1) r-1
2 2
= sinh x cosh x - sinh x + sinh x cosh x - sinh x +-
r=1 2
e x-(-1)-r+1e
m
-x
cosh xΣ(-1) r-1 2 2
= cosh x sinh x - cosh x + cosh x sinh x - cosh x +-
r=1 2
The 2nd term of the right side is as follows.

  f(x)dx 
e x +(-1)-me
x x x -x
m m
(-1) sinh x dx
a a a 2

- 15 -
   
e x -(-1)-me
x x x -x
(-1)mcosh x
a a

a
f(x)dx m
 2
dx
2 2
Since cosh cx - sinh cx = 1

    f(x)dx
x x m x x
sinh x f(x)cosh xdx - cosh x f(x)sinh xdx = Σ  2r
a a r=1 a a

  f(x)dx 
e x +(-1)-2me
x x x -x
2m 2m
+ (-1) sinh x dx
a a a 2

cosh x  f(x)dx

x x x x -2m -x

 
e -(-1) 2m e
- (-1)2m dx
a 2 a a
Here, m in the 2nd term and the 3rd term on the right side has to be an even number corresponding to the 1st
term. Furthermore,

  f(x)sinh xdx
x x
sinh x f(x)cosh xdx - cosh x
a a

 
e x-e -x x e x+e -x e x+e -x x e x-e -x
= f(x) dx - f(x) dx
2 a 2 2 a 2

   
1 x x x
= e f(x)e -xdx - e -x f(x)e xdx
2 a a
Using this, we obtain

     f(x)e dx
m x x 1 x x x
Σ  f(x)dx 2r = e f(x)e -xdx - e -x x
r=1 a a 2 a a

- sinh x  f(x)dx

  cosh x dx
x x x
2m
a a a

+ cosh x  f(x)dx

  sinh x dx
x x x
2m
a a a

 f(x)dx
x x
m
And since lim =0 holds by assumption ( See proof of Theorem 24.1.1 ), we obtain (1.1').
m  a a
Q.E.D.

     dx
x x x x x x x x
dx +  dx +  dx + 
2 4 6 8
Example1 +
a a a a a a a a
Substituting f(x)= 1 for (1.1') ,

    
x


 x 1 x x x 1 x
e -e -xa - e -xe xa
x x
Σ  dx =2r
e e dx - e -x
-x
e xdx =
r=1 a a 2 a a 2
x-a a-x
1 e +e
= -1+ e x-a
-1-e  =
a-x
-1
2 2
i.e.

  dx
 x x
Σ  2r
= cosh(x -a) - 1
r=1 a a

- 16 -
Theorem 24.3.2
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .

   f(x)sin xdx
m x x x x
Σ (-1) r-1 2r
 f(x)dx = sin x f(x)cosxdx - cosx
r=1 a a a a

- sin x  f(x)dx

  cos (x +m )dx
x x x
2m
a a a

+ cosx  f(x)dx

  sin (x +m )dx
x x x
2m
a a a
(2.1)

 f(x)dx  f(x)cosxdx - cosx f(x)sin xdx

 x x x x
r-1
Σ
2r
(-1) = sin x (2.1')
r=1 a a a a

Proof
The following expressions were obtained during the proof of Theorem 24.2.2.
(r -1)
    f(x)dx
x m x x
r-1
f(x)cosxdx = Σ(-1) cos x +  r
a r=1 2 a a
m
+ (-1)   f(x)dx cos x +
  2 
x x x
m m
dx
a a a
(r -1)
 2  
sin x +
x m x x
f(x)sin xdx = Σ(-1)
r-1
 f(x)dx r
a r=1 a a
m
+ (-1)   f(x)dx sin x +
  2 
x x x
m m
dx
a a a
Expanding a part of the 1st term of the right side,
(r -1)
  = cosx + sin x - cosx - sin x ++--
m
r-1
Σ
r=1
(-1)
2
cos x +
(r -1)
  = sin x - cosx - sin x + cosx ++--
m
Σ (-1) r-1 sin x +
r=1 2
Multiplying both sides by
sin x, cos x respectively,
(r -1) 
 
m
sin xΣ(-1) r-1 cos x +
2 2
= sin x cos x + sin x - sin x cos x - sin x ++--
r=1 2
(r -1) 
 
m
cos xΣ(-1) r-1 sin x +
2 2
= cos x sin x - cos x - cos x sin x + cos x ++--
r=1 2
The 2nd term of the right side is as follows.

   f(x)dx  cos x + m2 dx

x x x
(-1)msin x  m
a a a

(-1) cos x  f(x)dx sin x +

x m x x

  2 
m m
dx
a a a
2 2
Since cos cx + sin cx = 1 ,

    f(x)dx
x x m x x
sin x f(x)cosxdx - cosx f(x)sin xdx = Σ(-1)
r-1
 2r
a a r=1 a a

- 17 -
    f(x)dx  cos (x +m )dx
x x x
+ (-1)2msin x  2m
a a a

cos x  f(x)dx

x x x

  sin (x +m )dx

2m
- (-1)2m
a a a
Here, m in the 2nd term and the 3rd term on the right side has to be an even number corresponding to the 1st
term. Then, transposing the remainder terms , we obtain (2.1) .

  f(x)dx
x x
And since lim  m
=0 holds by assumption ( See proof of Theorem 24.1.1 ), we obtain (2.1').
m  a a

     x dx
x x x x x x x x
x dx   x dx +  x dx   + 
2 2 2 4 2 6 2 8
Example2
0 0 0 0 0 0 0 0

Substituting f(x)= x 2 for (2.1) ,

   x sin xdx
m x x x x
Σ (-1) r-1
 x 2dx 2r = sin x x 2cosxdx - cosx 2
r=1 0 0 0 0

- sin x  x dx
  cos (x +m )dx
x x x
2 2m
0 0 0

+ cosx  x dx
  sin (x +m )dx
x x x
2 2m
0 0 0
Here,

  x sin xdx = x -2 + 2cosx

x x
sin x x 2cosxdx - cosx 2 2
0 0

 x
x x 2!
2
dx 2m = x 2+2m
0 0 (2+2m )!
Substituting these for the above,

 x dx
m x x
r-1
Σ
2 2r
(-1) = x 2-2 + 2cosx
r=1 0 0


x 2!
- sin x x 2+2mcos (x +m )dx
0 (2+2m )!
+ cosx
x 2!
x 2+2msin (x +m )dx (2.2)
0 (2+2m )!

  x dx
 x x
Σ(-1)
r-1
 2 2r
= x 2-2 + 2cosx (2.2')
r=1 0 0
The higher integral of the left side is

 x
x x 2!
 2
dx n = x
2+n
0 0 ( 2+ n )!
Using this as the left side of (2.2) and (2.2') ,
m
r-1 2!
Σ
r=1
(-1)
(2+2r )!
x 2+2r = x 2-2 + 2cosx


x 2!
- sin x x 2+2mcos (x +m )dx
0 (2+2m )!

- 18 -
 
x 2!
+ cosx x 2+2msin (x +m )dx (2.3)
0 (2+2m )!
 2!
r-1
Σ (-1) x
2+2r
= x 2-2 + 2cosx (2.3')
r=1 ( 2+ 2r )!
When x =5, m =6 , Both sides of (2.3) are calculated as follows.

- 19 -
24.4 Integrals Series Expansion (exp x)
Mikio Sugioka discovered that the series of the higher integral of arbitrary functions is expanded into
the series of the higher integrals of exponential functions or trigonometric functions in March, 2012.
( http://www5b.biglobe.ne.jp/~sugi_m/page216.htm ). The expressions by the higher order integral series are
beautiful. However, more importantly, the expressions by a double series are useful for numerical calculations.
Then I introduce them in the following three sections.

Theorem 24.4.1
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .

 f(x)dx
s+r


 x x   (x -a)
Σ  r
= Σ f (r)(a)Σ (1.1)
r=0 a a r=0 s=0 (s + r)!

 e
 x x
= Σ f (a) (r) x-a
dx r (1.1')
r=0 a a

(-1) 
 x x (x -a)s+ r
 (r) 
Σ
r=0 a a
r
f(x)dx = Σ f (a)Σ(-1)
r=0
r
s=0 (s + r)!
s
(1.2)

 e
 x x
= Σ f (r)(a)  -(x-a )
dx r (1.2')
r=0 a a

Proof
r
()
Let f (a)= far and expand f(x) into Taylor series around a. Then

(x -a)0 1 (x -a )
1
2 (x -a )
2
3 (x -a )
3
f(x)= fa0 + fa  + fa  + fa  + (0)
0! 1! 2! 3!
Integrating both sides of this with respect to x from a to x one by one,


x 1 2 3 4
0 (x -a ) 1 (x -a ) 2 (x -a ) 3 (x -a )
f(x)dx = fa  + fa  + fa  + fa  +
a 1! 2! 3! 4!
2 3 4 5

 0 (x -a ) 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
f(x)dx = fa  + fa  + fa  + fa 
2
+
a a 2! 3! 4! 5!
(x -a)3 4 5 6

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x x
= fa  + fa  + fa  + fa 
3 0
+
a a a 3! 4! 5! 6!

Adding these also including (0) perpendicularly.

   f(x)dx
x x x x x x
+
2 3
f(x)+ f(x)dx + f(x)dx +
a a a a a a

 
(x -a)0 (x -a)1 (x -a)2 (x -a)3
= fa0 + + + +
0! 1! 2! 3!

+f  +
1
(x -a)
1 (x -a)2 (x -a)3 (x -a)4
a + + +
1! 2! 3! 4!

+f  +
2
(x -a)
2 (x -a)3 (x -a)4 (x -a)5
a + + +
2! 3! 4! 5!


- 20 -
   (x -a)s+ r
=Σ far Σ
r=0 s=0 (s + r)!
i.e.

a f(x)dx
 x x  (r)  (x -a)s+ r
Σ
r=0 a
r
= Σ f (a)Σ
r=0 s=0 (s + r)!
(1.1)

Furthermore,

 e
 (x -a)s+ r x x
x-a
Σ
s=0 (s + r)!
=
a a
dx r
Then

   e
 x x  x x
Σ  f(x)dx = Σ f (a) 
(r) r x-a
dx r (1.1')
r=0 a a r=0 a a
Next,

(x -a)0 1 (x -a )
1
2 (x -a )
2
3 (x -a )
3
f(x)= fa0 + fa  + fa  + fa  +
0! 1! 2! 3!


x (x -a)1 (x -a)2 (x -a)3 (x -a)4
- f(x)dx = -fa0 - fa1 - fa2 - fa3 -
a 1! 2! 3! 4!
(x -a)2 (x -a)3 (x -a)4 (x -a)5

x x
f(x)dx = fa  + fa1 + fa2 + fa3
2 0
+
a a 2! 3! 4! 5!
(x -a)3 4 5 6

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x x
-fa  - fa  - fa  - fa 
3 0
- = -
a a a 3! 4! 5! 6!
Adding these perpendicularly.

   f(x)dx
x x x x x x
+- 
2 3
f(x)- f(x)dx + f(x)dx -
a a a a a a

 
(x -a)0 (x -a)1 (x -a)2 (x -a)3
= fa0 - + - +-
0! 1! 2! 3!

+f  +-
1
(x -a)
1 (x -a)2 (x -a)3 (x -a)4
a - + -
1! 2! 3! 4!

+f  +-
2
(x -a)
2 (x -a)3 (x -a)4 (x -a)5
a - + -
2! 3! 4! 5!

  (x -a)s+ r
=Σ (-1)rfar Σ (-1) s
r=0 s=0 (s + r)!
i.e.


 x x (x -a)s+ r  (r) 
Σ
r=0
(-1) r
a a
f(x)dx = Σ f (a)Σ(-1)
r=0 s=0
r
(s + r)!
s
(1.2)

Furthermore,

 e
 (x -a)s+ r x x
Σ (-1) =s
 -(x-a )
dx r
s=0 (s + r)! a a
Then

- 21 -
   e
 x x  x x
Σ(-1)r  f(x)dx r = Σ f (a) 
(r) -(x-a )
dx r (1.2')
r=0 a a r=0 a a

   log xdx

x x x x x x
Example1 log x + log xdx + log xdx 2 + 3
+
1 1 1 1 1 1

n =1, 2, 3, 
(n ) n-1
f(x) = log x , (log x) =(-1) (n -1)! x -n
Substituting these for (1.1), (1.1') ,

1 log xdx

 x x 
r-1  (x -1)s+r
Σ
r=0 1
r
= Σ(-1)
r=1
(r -1)!Σ
s=0 (s + r)!

 e
 x x
= Σ(-1)r-1(r -1)!  x-1
dx r
r=1 1 1

   e
x x x x x x x x
e x-1dx -1! e x-1dx +2! e x-1dx -3!  x-1
2 3 4
= 0! dx +-
1 1 1 1 1 1 1 1
When higher integrals are replaced with Riemann-Liouville Integral and one arbitrary point x =0.8 is given,
it is as follows.

  
x x x x x x
Example2 x - x dx + x dx 2 - x dx 3 +- 
1 1 1 1 1 1

f(x) = x
(1+1/2)
1
(n ) -n (2n -3)!!
f (1) = 1 2
= (-1)n-1
1+1/2- n  2n
Substituting these for (1.2), (1.2') ,

 (x -1) s+r
 x x  (2r -3) !! 
Σ(-1)r
r=0

1 1
x dx r = e
2r
-(x-1)
s=0
+ Σ (-1)r-1
r=1
Σ(-1)s (s + r) !


 (2r -3) !! x x -(x-1) r
= e x-1 + Σ (-1)r-1
-( )
 e dx
r=1 2r 1 1

  
x 1!! x x x-1 2 3!! x x x x-1 3
-( ) (-1)!!
= e x-1 + e x-1
dx - e dx + 3 e dx -+
21 1 22 1 1 2 1 1 1

During the proof of the previous theorem, if the calculation is done without including (0) ,
we obtain the following theorem.

- 22 -
Theorem 24.4.1'
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .

a f(x)dx
x s+ r
 x  ()  (x -a )
r
Σ
r=1 a
r
= Σ f (a)Σ
r=0 s=1 (s + r)!

 f(x)dx

r-1
x x  (r) 
s-1 (x -a)s+ r
Σ
r=1
(-1)
a a
r
= Σ f (a)Σ(-1)
r=0 s=1 (s + r)!

Combining Theorem 24.4.1' and 24.1 , we obtain the following formula which gives the collateral integral of
the product of an exponential function and arbitrary functions.

Formula 24.4.2
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .


x s+ r
(r)   (x -a)
s-1
e x f(x)dx = e xΣ f (a)Σ(-1) (2.1)
a r=0 s=1 (s + r)!

e
x
-x  (r)  (x -a)s+ r
Σ (a)Σ
-x
f(x)dx = e f (2.2)
a r=0 s=1 (s + r)!

e
x
-x
Example cosxdx
a

(cosx) n = cos x +n /2 n =1, 2, 3, 

( )
Since ,

r  (x -a)

s+1

 
x
-x 
e cosxdx = e Σ cos a +
2 Σ
-x
a r=0 s=r (s +1)!

When a =2 and the first 50 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

- 23 -
24.5 Integrals Series Expansion (sin x)

Theorem 24.5.1
When f(x )is analytic function defined on a closed interval I, the following expressions hold for a ,x  I .


2s+1+ r


 x x   (x -a)
Σ  (r)
f(x)dx 2r-1 = Σ f (a)Σ (1.1)
r=1 a a r=0 s=0 (2s +1+ r)!

  sinh(x -a)dx
 x x
= Σ f (r)(a)  r
(1.1')
r=0 a a

  f(x)dx
 x x   (x -a)2s+1+ r
Σ(-1) 
r-1 2r-1
= Σ f (r)(a)Σ(-1)s (1.2)
r=1 a a r=0 s=0 (2s +1+ r)!

  sin(x -a)dx
 x x
= Σ f (a)  (r) r
(1.2')
r=0 a a

Proof
r
()
Let f (a)= far and expand f(x) into Taylor series around a. Then

(x -a)0 1 (x -a )
1
2 (x -a )
2
3 (x -a )
3
f(x)= fa0 + fa  + fa  + fa  +
0! 1! 2! 3!
Integrating both sides of this with respect to x from a to x alternately,


x 1 2 3 4
0 (x -a ) 1 (x -a ) 2 (x -a ) 3 (x -a )
f(x)dx fa=  + fa  + fa  + fa  +
a 1! 2! 3! 4!
3
(x -a)4 (x -a)5 (x -a)6
 0 (x -a )
x x x
f(x)dx 3 = fa  + fa1 + fa2 + fa3 +
a a a 3! 4! 5! 6!
(x -a)5 6 7 8

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
= fa  + fa  + fa  + fa 
5 0
+
a a 5! 6! 7! 8!

Adding these perpendicularly.

    f(x)dx
x x x x x x
f(x)dx + f(x)dx 3 +  5
+
a a a a a a

 
(x -a)1 (x -a)3 (x -a)5 (x -a)7
= fa0 + + + +
1! 3! 5! 7!

+f  +
2
(x -a)
1 (x -a)4 (x -a)6 (x -a)8
a + + +
2! 4! 6! 8!

+f  +
3
(x -a)
2 (x -a)5 (x -a)7 (x -a)9
a + + +
3! 5! 7! 9!

  (x -a)2s+1+ r
=Σ far Σ
r=0 s=0 (2s +1+ r)!
i.e.

- 24 -
 (x -a)2s+1+ r
a f(x)dx
 x x  
Σ = Σ f (a)Σ
2r-1 (r)
(1.1)
r=1 a r=0 s=0 (2s +1+ r)!
Furthermore,

  sinh(x -a)dx
 (x -a)2s+1+ r x x
Σ =  r
s=0 (2s +1+ r)! a a
Then,

 a   sinh(x -a)dx
 x x  x x
Σ  f(x)dx 2r-1 = Σ f r(a)  r
()
(1.1')
r=1 a r=0 a a
Next,

 f(x)dx
x (x -a)1 1 (x -a )
2
2 (x -a )
3
3 (x -a )
4
= fa0 + fa  + fa  + fa  +
a 1! 2! 3! 4!
(x -a)3 4 5 6

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x x
-fa  - fa  - fa  - fa 
3 0
- = -
a a a 3! 4! 5! 6!
(x -a)5 6 7 8

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
5
= fa0 + fa  + fa  + fa  +
a a 5! 6! 7! 8!
(x -a)7 (x -a)8 (x -a)9 (x -a)10

x x
-  f(x )dx = -fa  - fa1 - fa2 - fa3
7 0
-
a a 7! 8! 9! 10!

Adding these perpendicularly.

     f(x)dx
x x x x x x x x
f(x)dx - f(x)dx 3 +  f(x)dx 5 -  7
+- 
a a a a a a a a

 
(x -a)1 (x -a)3 (x -a)5 (x -a)7
= fa0 - + - +-
1! 3! 5! 7!

+f  +-
2
(x -a)
1 (x -a)4 (x -a)6 (x -a)8
a - + -
2! 4! 6! 8!

+f  +-
3
(x -a)
2 (x -a)5 (x -a)7 (x -a)9
a - + -
3! 5! 7! 9!

+f  +-
4
(x -a)
3 (x -a)6 (x -a)8 (x -a)10
a - + -
4! 6! 8! 10!

  (x -a)2s+1+ r
=Σ far Σ (-1) s
r=0 s=0 (2s +1+ r)!
i.e.

 f(x)dx (x -a)2s+1+ r

 x x  
r-1 (r)
Σ = Σ f (a)Σ(-1) s
2r-1
(-1) (1.2)
r=1 a a r=0 s=0 (2s +1+ r)!
Furthermore,

 sin(x -a)dx

 (x -a)2s+1+ r x x
Σ
s=0
(-1)
(2s +1+ r)!
= s
a a
r

Then,

- 25 -
    sin(x -a)dx
 x x  x x
Σ(-1)
r-1
 f(x)dx 2r-1 = Σ f (a) 
(r) r
(1.2')
r=1 a a r=0 a a

     tan x dx
x x x x x x x x
Example1 tan x dx  tan x dx 3 +  tan x dx 5 +  7

0 0 0 0 0 0 0 0
The higher differential quotient of tan x on x =0 is as follows according to Theorem 9.2.6 ( 9.2 )

(2n -1)
22n22n-1B2n
(tan x) |x=0 = T2n-1 = , (tan x)(2n)|x=0 = 0
2n
Where, T 2n-1 is the tangent number and B2n is the Bernoulli number. Thus, from (1.1), (1.1') ,

 0   sinh xdx
 2s+2r
x x   x  x x
Σ
r=1 0
 tan x dx 2r-1
= ΣT2r-1Σ
r=1 s=0 (2s +2r) !
= ΣT2r-1 
r=1 0 0
2r-1

= 1 sinh x dx  2 sinh x dx +16 sinhx dx

x x x x x x
3 5
+
0 0 0 0 0 0
When higher integrals are replaced with Riemann-Liouville Integral and one arbitrary point x =0.9 is given,
it is as follows.

     secx dx
x x x x x x x x
Example2 secx dx  secx dx 3 +  secx dx 5 -  7
- 
0 0 0 0 0 0 0 0
The higher differential quotient of secx on x =0 is as follows according to Theorem 9.2.8 ( 9.2 )

(secx)(2n)|x=0 = E2n , (secx)(2n +1)|x=0 = 0

Here, E2n is an Euler number. Thus, from (1.2), (1.2') ,

 secxdx
 x x  x 2s+1+ 2r

Σ
r=1
(-1) r-1
0 0
2r-1
= ΣE2rΣ(-1)
r=0 s=0 (2s +1+ 2r) !
s

  sin xdx
 x x
= ΣE2r 
2r
r=0 0 0

  5 sin x dx   sin x dx

x x x x x x
sin x dx 2  61 
4 6
= sin x + +
0 0 0 0 0 0

- 26 -
24.6 Integrals Series Expansion (cos x)

Theorem 24.6.1
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .

 f(x)dx
2s+ r


 x x   (x -a)
Σ  2r
= Σf (r)(a)Σ (1.1)
r=0 a a r=0 s=0 (2s + r)!

  cosh(x -a)dx
 x x
= Σf (a)  (r) 2r
(1.1')
r=0 a a

(-1) 
 x x (x -a)2s+ r
 (r) 
Σ r
f(x)dx =Σf (a)Σ(-1) s
2r
(1.2)
r=0 a a r=0 s=0 (2s + r)!

  cos(x -a)dx
 x x
= Σf (r)(a)  r
(1.2')
r=0 a a

Proof
r
()
Let f (a)= far and expand f(x) into Taylor series around a. Then

(x -a)0 1 (x -a )
1
2 (x -a )
2
3 (x -a )
3
f(x)= fa0 + fa  + fa  + fa  + (0)
0! 1! 2! 3!
Integrating both sides of this with respect to x from a to x alternately,
2 3 4
(x -a)5
 0 (x -a ) 1 (x -a ) 2 (x -a )
x x
) 2
f(x dx = fa  + fa  + fa  + fa3 +
a a 2! 3! 4! 5!
(x -a)4 5 6 7

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
= fa  + fa  + fa  + fa 
4 0
+
a a 4! 5! 6! 7!
(x -a)6 7 8 9

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
= fa  + fa  + fa  + fa 
6 0
+
a a 6! 7! 8! 9!

Adding these also including (0) perpendicularly.

 f(x)dx + f(x)dx + f(x)dx

x x x x x x
2 4 6
f(x)+
a a a a a a

 
0 2
(x -a) (x -a) (x -a)4 (x -a)6
= fa0 + + + +
0! 2! 4! 6!

+f  +
1
(x -a)
1 (x -a)3 (x -a)5 (x -a)7
a + + +
1! 3! 5! 7!

+f  +
2
(x -a)
2 (x -a)4 (x -a)6 (x -a)8
a + + +
2! 4! 6! 8!

  (x -a)2s+ r
=Σ far Σ
r=0 s=0 (2s + r)!
i.e.

a f(x)dx
 x x  (r)  (x -a)2s+ r
Σ = Σf (a)Σ
2r
(1.1)
r=0 a r=0 s=0 (2s + r)!

- 27 -
Furthermore,

  cosh(x -a)dx
 (x -a)2s+r x x
Σ =  r
s=0 (2s + r)! a a
Then,

  f(x)dx   cosh(x -a)dx

 x x  x x
Σ  2r
= Σf (r)(a)  r
(1.1')
r=0 a a r=0 a a
Next,
(x -a)0 (x -a)1 (x -a)2 (x -a)3
f(x) = fa0 + fa1 + fa2 + fa3 +
0! 1! 2! 3!
(x -a)2 3 4 5

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
-fa  - fa  - fa  - fa 
2 0
- = -
a a 2! 3! 4! 5!
(x -a)4 5 6 7

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
4
= fa0 + fa  + fa  + fa  +
a a 4! 5! 6! 7!
(x -a)6 7 8 9

  f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x
-  -fa  - fa  - fa  - fa 
6 0
= -
a a 6! 7! 8! 9!

Adding these perpendicularly and doing calculation similar to Theorem 24.5.1, we obtain (1.2) and (1.2') .

    tan
x x x x x x
tan x dx   tan x dx   x dx + 
-1 -1 2 -1 4 -1 6
Example1 tan x+
1 1 1 1 1 1
According to " 岩波数学公式Ⅰ" p39, the following expression holds for a natural number n .

 
(n )
tan x = (n -1)!cos n(tan -1 x )sin n tan -1 x +
-1
2
From this,
3n
 
(n ) (n -1)!
tan x
-1
|x=1 = sin
2n/2 4
Substituting this for (1.1), (1.1') ,
3r  (x -1)
r


2s 2s+
 x x (x -1) 
 (r -1)! 
Σ  tan x dx =
4Σ
+Σ
4 Σ  
-1 2r
sin
r=0 1 1 s=0 (2s ) ! r=1 2r/2 s=0 (2s + r) !

3r
 
  (n -1)! x x
=
4
cosh(x -1) +Σ
r=1 2r/2
sin
4

1 1
cosh(x -1) dx r 

 
1 x 1 x x
2
= cosh(x -1) + cosh(x -1) dx - cosh(x -1) dx +
4 2 1 2 1 1
When higher integrals are replaced with Riemann-Liouville Integral and one arbitrary point x =1.8 is given,
it is as follows.

- 28 -
     sin
x x x x x x
sin x dx   sin x dx -  x dx + 
-1 -1 2 -1 4 -1 6
Example2 sin x-
0 0 0 0 0 0
-1
The higher differential quotient of sin x on x =0 is as follows according to Theorem 9.3.2 ( 9.3 )
( 2n +1 )
sin x |x=0 = 2nＣ 0(2n -1)!!(2n -1)!!00 = (2n -1)!!2
-1

Then, from (1.2), (1.2'),

(-1)  sin

 2s+2r+1
x x  x 
Σ
r=0 0 0
r -1
x dx
2r
= Σ(2r -1)!! Σ(-1)
r=0 s=0 (2s +2r +1) !
2 s

  cos xdx
 x x
= Σ(2r -1)!!2 
2r+1
r=0 0 0

= (-1)!!  cos x dx  1!!  cos x dx +3!!  cos x dx

x x x x x x

2 2 3 2 5
0 0 0 0 0 0

During the proof of the previous theorem, if the calculation is done without including (0) ,
we obtain the following theorem.

Theorem 24.6.1'
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .


2s+ r


 x x   (x -a)
Σ  (r)
f(x)dx 2r = Σf (a)Σ (1.3)
r=1 a a r=0 s=1 (2s + r)!

 f(x)dx

r-1
x x  (r) 
s-1 (x -a)2s+ r
Σ = Σf (a)Σ(-1)
2r
(-1) (1.4)
r=1 a a r=0 s=1 (2s + r)!

Note
This theorem is not so beautiful. However, this is useful for the calculation.

     sin
x x x x x x x x
sin x dx   sin x dx +  sin x dx   x dx + 
-1 2 -1 4 -1 6 -1 8
Example
0 0 0 0 0 0 0 0
-1
Since the higher differential quotient of sin x on x =0 is same as Example2, from (1.4),

 sin

r-1
x x 
2  s-1 x 2s+2r+1
Σ = Σ(2r -1)!! Σ
-1 2r
(-1) xdx (-1)
r=1 0 0 r=0 s=1 (2s +2r +1)!
If the left side is replaced with Riemann-Liouville Integral and one arbitrary point x =0.6 is given, then

- 29 -
24.7 Series of Higher Integrals with coefficients
Theorem 24.1.1 is extensible to the higher integral series with coefficients n -1+ rＣ n-1 . Although these
coefficients are numbers on the hypotenuse of Pascal's triangle, coefficients such as n 3 is not so
interesting. Then, in this section, we ask for the sum of the higher integral series at n=2 i.e.

     f(x)dx
x x x x x x x x
f(x)dx  2 f(x)dx  4  + 
2 3 4
1 f(x)dx + 3
a a a a a a a a

Theorem 24.7.1
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .

r f(x)dx = e  f(x)e dx  f(x)e

m x x x x x
Σ r x x
-x -x 2
dx + e
r=1 a a a a a

- (m +1)e   f(x)dx

  e dx
x x x x
x m-1 -x 2
a a a a

+ m e   f(x)dx e dx
 
x x x x
x m -x 2
(1.1)
a a a a

Σ r f(x)dx = e  f(x)e dx + e  f(x)e dx

 x x x x x
r x -x x -x 2
(1.1')
r=1 a a a a a

Proof
f (a) = 0 (r =1, 2,  ,m+ n -1), Theorem 16.1.2 ( 16.1 ) was as follows.
r
<>
When

  
x x m -1 -n
 f <0>g (0)dx n = Σ n r (r)
f < + >g
a a r=0 r


(-1)m n -1 n -1Ｃ k x x <m+ k > (m+ k ) n
 f
(n , m) Σ
+ g dx
k=0 m + k a a
(r)
Here, let g = e -x , n =2 . Then, since g =(-1)re -x ,

  
x x m -1 -2
r
f <0>e -xdx 2 = Σ f <2+ >(-1)re -x
a a r=0 r


(-1)m 1Ｃ 0 x x <m+0>
+ f (-1)m+0e -xdx 2
(2, m) m +0 a a


(-1)m 1Ｃ 1 x x <m+1>
+ f (-1)m+1e -xdx 2
(2, m) m +1 a a
i.e.

   f
x x m -1 x x x x
<0> -x <2+ r > -x <m > -x <m +1> -x
e dx = Σ (1+ r) f
2 2 2
f e + (m +1) f e dx - m e dx
a a r=0 a a a a

  r  = 1+ r , 
-2 1
 (-1) r
= m(m +1)
 (2, m)
Here, subtracting 1 from the index <r> of the integration operator of the function f,

 f
x x m -1
<1+ r> -x
<-1> -x 2
e dx = Σ (1+ r)f e
a a r=0

 f  f
x x x x
<m-1> -x 2 <m> -x
+ (m +1) e dx - m e dx 2
a a a a

- 30 -
Left side becomes as follows.

  f
x x


x x
<0> -x x
<-1> -x
f2
e dx = f e a + <0> -x
e dx dx
a a a a

= f dx +  f
x x x
<0> -x <0> -x
e e dx 2
a a a
Substituting this for the above,

 
x x x m -1
r
f <0> -x
e dx + f <0>e -xdx 2 = Σ (1+ r)f <1+ >e -x
a a a r=0

  f
x x x x
m-1> -x <m> -x
+ (m +1) f< e dx 2 - m e dx 2
a a a a
From this,

  f(x)e dx
m x x x
r <>
e -xΣr f (x)= f(x)e -xdx + -x 2
r=1 a a a

- (m +1) f  f
x x x x
<m-1> -x 2 <m> -x
e dx +m e dx 2
a a a a
Here, let

    f(x)dx
x x x x x x
f
<r>
=  f(x)dx r , f <
m-1>
=  f(x)dx
m-1
, f
<m>
=  m
a a a a a a

(r =1, 2,  ,m +1) , substituting these for the above

<r>
Since these satisfy the condition f (a )=0
x
and multiplying by e the both sides,

   f(x)e dx
m x x x x x
Σr  r x
f(x)e -xdx + e x -x 2
f(x)dx = e
r=1 a a a a a

- (m +1)e   f(x)dx

 e
x x x x
x m-1 -x
dx 2
a a a a

+ m e   f(x)dx e
 
x x x x
x m -x
dx 2 (1.1)
a a a a

  f(x)dx
x x
And since lim  m
=0 holds by assumption ( See proof of Theorem 24.1.1 ), we obtain (1.1').
m  a a

   dx
x x x x x x
Example 1 1 dx + 2 dx 2 + 3 3
+
a a a a a a

 dx 
m-1
x x (x -a) x x (x -a)m
f(x)=1 ,
m-1
= ,  dx m =
a a (m -1)! a a m!
Substituting these for (1.1) ,

r dx  e  (m -1)! e
x x (x -a )m-1
e
m x x x x x
Σ r x x x
-x -x 2 -x
=e dx + e dx -(m +1)e dx 2
r=1 a a a a a a a

+ m e 
x x m
(x -a) x -x 2
e dx
m! a a
The 1st term and the 2nd term of the right side are as follows.

- 31 -
    e - e dx
x x x x
ex e -xdx + e x e -xdx 2 = e xe a- e x + e x a x
a a a a
x-a
=e (x - a)
After the long calculation, the 3rd term and the 4th term on the right side become as follows.

 
x x (x -a )m-1 x x (x -a )m
x -x 2 x -x 2
-(m +1)e e dx + me e dx
a a (m -1)! a a m!
x-a
= -e (x -a)
m +1
e (x -a)(m , x -a)- (1+ m , x -a)
x-a
+
(m -1)!
1
e (x -a)(1+ m , x -a)- (2+ m , x -a)
x-a
-
(m -1)!
Substituting these for the above,

r dx
m x x
x-a
Σ r
=e (x -a) - e x-a(x -a)
r=1 a a
m +1
e (x -a)(m , x -a)- (1+ m , x -a)
x-a
+
(m -1)!
1
e (x -a)(1+ m , x -a)- (2+ m , x -a)
x-a
-
(m -1)!
(1.2)

r dx r = e
 x x
x-a
Σ
r=1 a a
(x -a) (1.2')

The left side is

r dx
m x x m r(x -a)r
Σ
r=1 a a
r
=Σ
r=1 r!
Using this as the left side of (1.2) and (1.2') ,
m r(x -a)r x-a x-a
Σ
r=1 r!
=e (x -a) - e (x -a)
m +1
e (x -a)(m , x -a)- (1+ m , x -a)
x-a
+
(m -1)!
1
e (x -a)(1+ m , x -a)- (2+ m , x -a)
x-a
-
(m -1)!
(1.3)
 r(x -a)r
Σ = e x-a(x -a) (1.3')
r=1 r!
When x =4, a =3, m =7 , Both sides of (1.3) are calculated as follows.

- 32 -
 Further, when a =3 , the first 50 terms of ∑ are calculated and both sides of (1.3') are illustrated, it is as
follows. Both sides overlap exactly and blue (left) can not be seen.

   cos x dx
x x x x x x
Example 2 1 cos x dx + 2 2
cos x dx + 3 3
+
a a a a a a
Substituting f(x)= cosx for (1.1') ,

   cosx e
 x x x x x
Σr  cosx dx = e x
r
cosx e dx + e x
-x -x
dx 2
r=1 a a a a a
Integrating the right side with respect to x from a to x by force,

r cosx dx
x x-a x-a
 x e (x -a)(cos a - sin a) e cos a - cos x
Σ
r=1 a a
r
=
2
+
2
The higher integral of the left side is as follows according to Theorem 4.1.3 ( 4.1 )

n  (n -s)
 cosx dx
n -1 (x -a )s

 - Σ  
x x
n
= cos x - cos a -
a a 2 s=0 s! 2
Then, the left side is
r  (r -s)
  
(x -a)s
 
 r-1
Σr cos x -
r=1 2
-Σ
s=0 s!
cos a -
2
When a =3/2 , the first 18 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

- 33 -
Theorem 24.7.2
When f(x )is analytic function defined on a closed interval I , the following expressions hold for a ,x  I .

  f(x)dx  f(x)e dx - e  f(x)e dx

m x x x x x
r 
r-1
Σ r x x
-x -x 2
(-1) =e
r=1 a a a a a

- (-1) (m +1)e   f(x)dx

  e dx
x x x x
m x m-1 x 2
a a a a

+ (-1) m e   f(x)dx e dx

 
x x x x
m+1 x m x 2
(2.1)
a a a a

r f(x)dx = e  f(x)e dx - e  f(x)e dx

 x x x x x
r-1
Σ r x x
-x -x 2
(-1) (2.1')
r=1 a a a a a

Proof
In a way similar to Theorem 24.7.1, we obtain the desired expressioons

     log x dx
x x x x x x x x
log x dx 4 
2 3 4
Example3 1 log x dx -2 log x dx +3 +-
0 0 0 0 0 0 0 0

Substituting f(x)= log x , a =0 for (2.1) ,

  log x dx  log x e dx - e  log x e dx

m x x x x x
r 
r-1
Σ r x x
-x -x 2
(-1) =e
r=1 a a a a a

- (-1) (m +1)e   log x dx

  e dx
x x x x
m x m-1 x 2
a a a a

+ (-1) m e   log x dx e dx

 
x x x x
m+1 x m x 2
a a a a
The integrasl of the right side are as follows.

- 34 -
  log xe dx = e log|x| - Ei(x) = e log|x|- Ei(x)+ 
x x
x x x
0
0

 log xe dx =  e log|x|- Ei(x)+ dx

x x x
x x
0 0 0
x
= e xlog|x|+1-(x +1)Ei(x) +  x0
= e xlog|x|+1-(x +1)Ei(x)+  x - (1-)


x et
Where, Ei(x) = dt ,  = 0.5772
(Euler-Mascheroni Constant ).
- t
Therefore, the 1st term and the 2nd term on the right side are

 log xe dx - e  log xe dx
x x x
-x
e x -x x 2 -x
= e xEi(x) -   + e
-x
-1
0 a a
Next,


x xm
 
x m 1
 log x dx = m
log|x|-Σ
0 0 m! s=1 s
Using thes for the 3rd term and the 4th term on the right side, the right side becomes

e xEi(x)-   + e
-x -x
-1


x x m-1

 
x m -1 1
m
- (-1) (m +1)e x
log|x|-Σ e xdx 2
a a (m -1)! s=1 s

 
x x x m


m 1
+ (-1)m+1 m e x log|x|-Σ e xdx 2
a a m! s=1 s
i.e.

e -x xEi(x)-   + e -x - 1

 
x x


m +1 m-1
m -1 1
- (-1)m ex x log|x|-Σ e xdx 2
(m -1)! a a s=1 s

 
x


1 x m 1
+ (-1)m+1 ex x m log|x|-Σ e xdx 2
(m -1)! a a s=1 s
Thus, we obtain

  log x dx
m x x
Σ(-1) r 
r-1 r
= e -x xEi(x) -   + e -x - 1
r=1 a a

 
x x


m +1 m-1
m -1 1
- (-1)m ex x log|x|-Σ e xdx 2
(m -1)! a a s=1 s

 
x x


1 m 1
+ (-1)m+1 ex x m log|x|-Σ e xdx 2 (2.2)
(m -1)! a a s=1 s

  log x dx
 x x
Σ (-1) r-1
r  r
= e -x xEi(x) -   + e -x - 1 (2.2')
r=1 a a
The left side is

 xn
x x

 
n 1
n
log x dx = log|x|-Σ
0 0 n! s=1 s
Using this,

- 35 -
 m rx r
  =e
r 1
xEi(x)-   + e -x - 1
r-1
Σ
-x
(-1) log x -Σ
r=1 r! s=1 s

 
x x


m +1 m-1
m -1 1
- (-1)m ex x log|x|-Σ e xdx 2
(m -1)! a a s=1 s

 
x x


1 m 1
+ (-1)m+1 ex x m log|x|-Σ e xdx 2 (2.3)
(m -1)! a a s=1 s
 rx r
 
r 1
Σ(-1)
r-1
log|x|-Σ = e -x xEi(x) -   + e -x - 1 (2.3')
r=1 r! s=1 s
When x =-0.2, m =4 , Both sides of (2.3) are calculated as follows.

Further, when the first 30 terms of ∑ are calculated and both sides of (2.3') are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

c.f.
Mr. Sugioka showed the following theorem ( Sugioka's Theorem 3-1) just like Theorem24.7.1 .

    f(x)e
x x x x x x x x x
f(x)dx  3  f(x)dx +  = e x
2 3 4 -x 2
1 f(x)dx + 2 dx
a a a a a a a a a
Adding this to Theorem 24.1.1 (1.1') , we obtain Theorem 24.7.1 (1.1') .

- 36 -
2012.09.01
2018.06.05 Updated
Kano Kono
Alien's Mathematics

- 37 -
25 Series of Higher Integral with Geometric Coefficients
This chapter is a generalization of " 24 Sugioka's Theorem on the Series of Higher Integral ". The origin of
theses paper is " e ^x に関する公式の発見 " ( Mikio Sugioka 2003 ) .

25.1 Series of the n-th order Integrals

In this section, we ask for the sum of series of higher integral with geometric coefficients, such as

     f(x)dx
x x x x x x x x
c
1
f(x)dx  c
2
f(x)dx + c
2 3
f(x)dx  c
3 4
 4
+ 
a a a a a a a a

Theorem 25.1.1
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

c  f(x)dx    f(x)dx  e
r=1  
m x x x x x x
Σ r r
= ce cx m+1
f(x)e -cxdx - c e cx  m -cx
dx
a a a a a a
(1.1)

  f(x)dx = 0
x x
Especially, when lim c m
 m
,
m  a
a

Σc  f(x)dx = ce  f(x)e

 x x x
r r cx -cx
dx (1.1')
r=1 a a a

Proof
Let

  f(x)dx
x x
<r>
f (x) =  r
r =1, 2,  ,m (1.r)
a a

f (a) = 0 (r =1, 2,  ,m).

r <>
Then So,

 f
x x x
f(x)e -cxdx = f <1>(x)e -cxa + c <1>
(x)e -cxdx
a a

f
x
= f <1>(x)e -cx + c <1>
(x)e -cxdx
a

f
x x
= f <1>(x)e -cx + c f <2>(x)e -cxa + c 2 <2>
(x)e -cxdx
a

f
x
= f <1>(x)e -cx + c 1f <2>(x)e -cx + c 2 <2>
(x)e -cxdx
a


f
m x
= e -cx Σc r-1f < r>
(x) + c m
<m> -cx
e dx
r=1 a
Substituting (1.r) for this,

     f(x)dx  e
x m x x x x x
-cx
f(x)e dx = e -cxΣc
r-1
 f(x)dx r + c m  m -cx
dx
a r=1 a a a a a
Multiplying by e cx the both sides,

     f(x)dx  e
x m x x x x x
e cx
f(x)e -cxdx = Σc
r-1
 f(x)dx r + c me cx  m -cx
dx
a r=1 a a a a a
Multiplying by c the both sides and transposing them, we obtain (1.1) .

-1-
    dx
x x x x x x
Example 1 c 1
dx + c 2 dx + c 3
2 3
+
a a a a a a

  dx
x x m
(x -a)
f(x)=1 ,  m
=
a a m!
Substituting these for (1.1) ,

c  dx
r=1  e 
m x x x x (x -a)m -cx
Σ r r
= ce cx -cx
dx - c
m+1 cx
e
m!
e dx
a a a a
The 1st term of the right sideis
x

e  
x e -cx
ce cx -cx
dx = ce cx - = e c(x-a)-1
a c a
The 2ndt term of the right sideis is expressed as follows using the incomplete gamma function (x, y ) .

e -ac(x -a)mc(x -a)-m(m +1)-m +1, c(x -a)


x (x -a)m -cx
e dx =
a m! c m!
e -ac (x -a)m
= (m +1)-m +1, c(x -a)
c m ! c(x -a)m
When c >0 & x -a > 0 ,


x (x -a)m -cx e -ac
e dx = m+1 (m +1)-m +1 , c(x -a )
a m! c m!
Thus,

c  dx
x x c(x-a )
m e
(m +1)-m +1 , c(x -a)
c(x-a )
Σ
r=1 a
r
a
r
=e -1 -
m!
(1.2)

c(x-a )
e
And, since lim (m +1)-m +1 ,c(x -a) = 0,
m  m!

c  dx r = e
 x x
c(x-a )
Σ
r=1 a
r
a
-1 (1.2')

The higher order integral of the left side of (1.2) is as follows.

 dx
x x
r (x - a)r
=
a a r!
Then, the left side is

c 
m x x (x - a)r
m
Σ
r=1 a
r
a
dx = Σc
r
r=1 r!
r

Using this as the left side of (1.2) and (1.2') ,

m (x -a ) r e c(x-a
)
(m +1) -m +1 , c(x -a )
r c(x-a)
Σ
r=1
c
r!
=e -1 -
m!
(1.3)

(x - a)r
Σ c r
= e c(x-a)-1 (1.3')
r=1 r!
When x =2, a =0, c =3.3, m =4 , both sides of (1.3) are calculated as follows.

-2-
Also, it is clear from the following that (1.3') is correct.

r  c(x - a )
r 0
r (x - a ) c(x - a )
Σ c =Σ - = e c(x-a)-1
r=1 r! r=0 r! 0!

   e dx
x x x x x x
Example 2 c 1 x
e dx + c 2 x
e dx + c 3 2 x 3
+
a a a a a a

 e dx
x x m -1 (x -a)r
f(x)= e x , x m
= e x- e a Σ
a a r=0 r!
Substituting these for (1.1) and using the incomplete gamma function (x, y ) ,

c  ee    e dx  e
m x x x x x x
Σ r
 e dx = ce cx
x r x -cx m+1
dx - c e cx  x m -cx
dx
r=1 a a a a a a

= ce  e  
x
(1-c)x m+1
x m -1 (x -a)r -cx
cx
dx - c e cx e - ea Σ
x
e dx
a a r=0 r!

= ce  e e 
x
(1-c)x m+1 cx
x
(1-c)x m+1 cx+a
m -1 x (x -a)r -cx
cx
a
dx - c e
a
dx + c e Σ
r=0 a r!
e dx

ce cx e (1-c) - e (1-c)  c e e (1-c) - e (1-c) 

x a m+1 cx x a
= -
1-c 1-c
 
e - (x -a)rc(x -a)-r(1+r)-1+r, c(x -a)
m -1
ac
m+1 cx+a
+c e Σ
r=0 c r!
ce cx e (1-c) - e (1-c)  c e e (1-c) - e (1-c) 
x a x a
m+1 cx
= -
1-c 1-c
1+r, c(x -a)
 
m -1 1
- c me cx+(1-c)a Σ r -1
r=0 c r!
i.e.

 ce cx e(1-c) - e (1-c)  c e cx e(1-c) - e (1-c) 

x a m+1 x a
c
m x x
Σ r
 x
e dx = r
-
r=1 a a 1-c 1-c
1+ r, c(x -a)
 
m -1 1
- c me cx+(1-c)a Σ r -1 (1.4)
r=0 c r!
Especially, when a =- ,

-3-
  e dx
x x
x r
= ex r =1, 2, 3,  : Lineal Higher Integral
- -

a 1+r, c(x -a)

e(1-c) = 0 , =0
cr r!
So, (1.4) becoms as follows.
m 1- c m
e Σc = ce
x r x
r=1 1-c
And from this, the following well-known equation is obtained.
m -1 1- c m
Σ
r=0
cr
=
1- c
When 0 < c < 1 , lim
m 
c m =0 . Then,

c 
ce cx e(1-c) - e (1-c) 
x x a
 x
Σ
r=1 a a
r x
e dx = r
1- c
(1.4')

Especially, when a =- ,

 ce cx
e Σc =
cx r
r=1 1- c
And from this, the following well-known equation is obtained.
 1
Σ
r=0
cr =
1- c

The higher order integral of the left side of (1.4) is as follows.

 e dx
x x r-1 (x -a )s

a a
x r x a
= e -e Σ
s=0 s!
Then, the left side is

c  dx e - e Σ 
m x x m r-1 (x -a)s
Σ
r=1 a a
r r
= Σc
r=1
r x a
s=0 s!
Using this as the left side of (1.4) and (1.4') ,

 
m r-1 (x -a )s 1- c m
ce cx e (1-c) - e(1-c) 
x a
Σ
r=1
c r x a
e -e Σ
s=0 s!
=
1- c
1+r, c(x -a)
 
m -1 1
- c me cx+(1-c)a Σ r
-1 (1.5)
r=0 c r!
ce cx e (1-c) - e (1-c) 
x a

e - e Σ 
 r-1 (x -a)s
Σ
r=1
c r x a
s=0 s!
=
1- c
(1.5')

When x =4, a =1, c =2, m =3 , both sides of (1.5) are calculated as follows. Both sides coincide exactly.
So, (1.4) is confirmed to be correct.

-4-
    log x dx
x x x x x x
Example 3 c1 log x dx + c 2 log x dx 2 + c 3 3
+
0 0 0 0 0 0
Substituting f(x)= log x , a =0 for (1.1') ,

  log x e
 x x x
Σc r 
r=1
log x dx r = ce cx -cx
dx
a a a
The integral of the right side is as follows.
x

 log x e  
x
-cx -e -cx log|x|+ Ei(-cx)
dx =
0 c 0
-e -cx
log|x|+ Ei(-cx) log c + 
= -
c c


x et
Where, Ei(x) = dt ,  = 0.5772 (Euler-Mascheroni Constant ).
- t
Multiplying by e cx the both sides,

 log x e
x
ce cx -cx
dx = -log|x|+ e cxEi(-cx) -  - log c
0
Substituting this for the right side of the above,

c  log x dx
 x x
Σ r r
= -log|x|+ e cxEi(-cx) -  - log c (1.6)
r=1 a a
The higher order integral of the left side becomes lineal higher integral as follows.

 log x dx
x xn
log|x|-Σ s 
x n 1
n
=
0 0 n! s=1
Then, (1.6) is
r r

  = -log|x|+ e
 cx r 1
Σ log|x|-Σ Ei(-cx) - 
cx
- log c
r=1 r! s=1 s

When c =1.7 , the first 10 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

-5-
Theorem 25.1.2
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

  f(x)e
m x x x
Σ (-1) r-1 r
c  f(x)dx = ce -cx r cx
dx
r=1 a a a

   f(x)dx  e
x x x
- (-1)mc m+1e -cx  m cx
dx (2.1)
a a a

  f(x)dx = 0
x x
Especially, when lim c m 
m 
m
,
a a

c  f(x)dx = ce  f(x)e

m x x x
Σ r r cx
-cx
(-1)r-1 dx (2.1')
r=1 a a a

Proof
In a similar way to the proof of Theorem 25.1.1 , we obtain the desired expressions.

Example 4

     sin xdx

x x x x x x x x
c1 sin xdx - c 2 sin xdx 2 + c 3 sin xdx 3 - c 4  4
+- 
a a a a a a a a

Substituting f(x)= sin x for (2.1') ,

  sin xe
 x x x
Σ(-1)r-1c r 
r=1
r
sin xdx = ce -cx cx
dx
a a a
The right side is as follows.

 sin x e
cx
x
cx csin x -cosx e - csin a -cosa e ca
dx =
a 1+ c 2
-cx
Multiplying by ce the both sides,

-6-
  sin x e
x c
2 csin x cosx
ce -cx cx
dx = - - csin a -cosa e c(a -x)
a 1+ c
Thus, we obtain


 x x c
Σ(-1) c r  2 csin x cosx
- csin a -cosa e c(a -x)
r-1
sin xdx r = -
r=1 a a 1+ c
(2.2')
The higher integral of the left side is as follows according to Theorem 4.1.3 ( 4.1 )

n  (n -s)
 sin x dx
n -1 (x -a )s

  -Σ  
x x
n
= sin x - sin a -
a a 2 s=0 s! 2
Then, (2.2') is
r  (r -s)
sin x - 2  -Σ 
 (x -a)s

r-1
Σ
r=1
(-1) r-1 r
c
s=0 s!
sin a -
2
c
2 csin x cosx
= - - csin a -cosa e c(a -x)
1+ c

When a =3/2 , c=1.3 , the first 30 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

-7-
25.2 Series of the odd-th order Integrals
In this section, we ask for the sum of series of higher integral with geometric coefficients, such as

     f(x)dx
x x x x x x x x
c 1 f(x)dx  c 3   + 
5 7
f(x)dx  c
3 5 7
f(x)dx + c
a a a a a a a a

Theorem 25.2.1
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

  f(x)dx 2r-1 =
   f(x)e dx 
m x x c x x
Σ
r=1
c
2r-1
a a 2
e cx
a
f(x)e
-cx
dx + e
-cx
a
cx

cosh cx  f(x)dx

x x x cx -2m+1 -cx

 
2m-1 2m e +(-1) 2m-1 e
- (-1) c dx
a a 2 a

sinh cx  f(x)dx

x x x cx -2m+1 -cx

 
2m e -(-1) 2m-1 e
+ (-1)2m-1c dx
a a 2 a
(1.1)

  f(x)dx
x x
Especially, when lim c 2m 
m 
2m-1
= 0,
a a

  f(x)dx 2r-1 =
   f(x)e dx 
 x x c x x
Σ
r=1
c
2r-1
a a 2
e cx
a
f(x)e
-cx
dx + e
-cx
a
cx
(1.1')

Proof
Formula of repeated integration by parts was as follows. (" 01 Generalized Taylor's Theorem " ( A la Carte ) )

 
x m x x
f(x)g(x)dx = (-1) f <r>(x)g ( )(x)a + (-1)m f <m>(x)g (x)dx
r-1 r-1 (m)
a r=1
Σ a
When g(x)=cosh cx , sinh cx ,
(r-1) r-1 e cx +(-1)-r+1e -cx (m)
cx -m -cx
m e +(-1) e
(cosh cx) =c 　, (cosh cx) = c
2 2
(r-1) r-1 e cx -(-1)-r+1e -cx (m)
cx -m -cx
m e -(-1) e
(sinh cx) =c , (sinh cx) = c
2 2
Substituting these for the above,
x

 f(x)cosh cxdx = Σ(-1) f 

x m
r-1 <r> r-1 e cx +(-1)-r+1e -cx
(x)c ,
a r=1 2 a

f
m
x
<m> e cx +(-1)-me -cx
m
+ (-1) (x)c dx ,
a 2
x

 f(x)sinh cxdx = Σ(-1) f 

x m
r-1 <r> r-1 e cx -(-1)-r+1e -cx
(x)c ,
a r=1 2 a

f
m
x
<m> e cx -(-1)-me -cx
m
+ (-1) (x)c dx ,
a 2
Here, let

  f(x)dx
x x
<r>
f (x) =  r
r =1, 2,  ,m
a a

-8-
 f (a) = 0 (r =1, 2,  ,m) ,
<>r
Since

   f(x)dx
x m e cx +(-1)-r+1e -cx x x
r-1 r-1
f(x)cosh cxdx = Σ(-1) c  r
a r=1 2 a a

  f(x)dx 
m
x x
+ (-1)
a a
x
m me cx +(-1)-me -cx
c dx
a 2

   f(x)dx
x m e cx -(-1)-r+1e -cx x x
f(x)sinh xdx = Σ(-1)
r-1
c
r-1
 r
a r=1 2 a a

  f(x)dx 
m
x x x
m me cx -(-1)-me -cx
+ (-1) c dx
a a a 2
Expanding a part of the 1st term of the right side,

e cx +(-1)-r+1e
m -cx
Σ
r=1
(-1) r-1 r-1
c
2
0 1
= c cosh cx - c sinh cx + c cosh cx - c sinh cx +-
2 3

e cx -(-1)-r+1e
m -cx
Σ(-1) r-1 c r-1
r=1 2
0 1
= c sinh cx - c cosh cx + c sinh cx - c cosh cx +-
2 3

Multiplying both sides by cosh cx, sinh cx respectively,

cx -r+1 -cx
m e +(-1) e
cosh cxΣ(-1) r-1 c r-1
r=1 2
0 2 1 2 2 3
= c cosh cx - c cosh cxsinhcx + c cosh cx - c cosh cxsinh cx +-
e cx -(-1)-r+1e
m
-cx
sinh cxΣ(-1) r-1 c r-1
r=1 2
0 2 1 2 2 3
= c sinh cx - c sinh cxcosh cx + c sinh cx - c sinh cxcosh cx +-

The 2nd term of the right side is as follows.

  f(x)dx 
e cx +(-1)-me
x x x -cx
m m m
(-1) cosh cx c dx
a a a 2

(-1) sinh cx  f(x)dx c

e cx -(-1)-me
x x x -cx

 
m m m
dx
a a a 2
2 2
Since cosh cx - sinh cx = 1 ,

    f(x)dx
x x m x x
cosh cx f(x)cosh cx dx - sinh cx f(x)sinh cx dx = Σc
2r-2
 2r-1
a a r=1 a a

cosh cx  f(x)dx

x x x cx -2m+1 -cx

 
2m-1 e +(-1) 2m-1 2m-1 e
+ (-1) c dx
a 2
a a

sinh cx  f(x)dx

x x x cx -2m+1 -cx

 
2m-1 e -(-1) 2m-1 2m-1 e
- (-1) c dx
a a 2 a
Here, m in the 2nd term and the 3rd term on the right side has to be an odd number corresponding to the 1st
term. Furthermore,

  f(x)sinh cx dx
x x
cosh cx f(x)cosh cx dx - sinh cx
a a

-9-
 e cx+e e cx+e e cx-e e cx-e
 
-cx x -cx -cx x -cx
= f(x) dx - f(x) dx
2 a 2 2 a 2

   f(x)e dx
1 cx x x
cx
= e f(x)e -cxdx + e -cx
2 a a
Using this, we obtain

 f(x)dx   f(x)e  f(x)e dx

m x x 1 cx x x
Σ cx
2r-2 2r-1 -cx -cx
c = e dx + e
r=1 a a 2 a a

  f(x)dx 
e cx +(-1)-2m+1e
x x x -cx
2m-1 2m-1 2m-1
- (-1) c cosh cx dx
a a a 2

sinh cx  f(x)dx

cx -cx
x x x -(-1)-2m+1e
 
e
+ (-1)2m-1c 2m-1 2m-1
dx
a a a 2
Multiplying by c the both sides, we obtain (1.1) .

     dx
x x x x x x x x
c 1 dx  c 3  
5 7

3 5 7
Example 1 dx + c dx + c
a a a a a a a a


x x
m (x -a)m
f(x)=1 , dx =
a a m!
Substituting these for (1.1) ,

    e dx 
m x x c x x
Σc
r=1
2r-1
a

a
dx
2r-1
=
2
e cx
a
e
-cx
dx + e
-cx
a
cx

cosh cx
2m-1 cx -cx
2m-1 2m (x -a) e x +(-1)-2m+1e
- (-1) c dx
(2m -1) ! a 2

sinh cx
2m-1 cx -cx
2m-1 2m (x -a ) e x -(-1)-2m+1e
+ (-1) c dx
(2m -1) ! a 2
Here,
x x

 e  e dx    
c x x ce cx e -cx ce cx e cx
e cx -cx
dx + e -cx cx
= - +
2 a a 2 c a 2 c a
c(x-a ) -c(x-a )
e -e
= = sinhc(x -a)
2
Then,

  dx
m x x
Σc
2r-1
 2r-1
= sinhc(x -a)
r=1 a a


2m-1
e cx +(-1)-2m+1e
-cx
2m-1 2m
x (x -a)
- (-1) c cosh cx dx
a (2m -1) ! 2

sinh cx
2m-1
e cx -(-1)-2m+1e
-cx
2m-1 2m (x -a ) x
+ (-1) c dx
(2m -1) ! a 2
(1.2)

aa dx
 x x
Σ 2r-1 2r-1
c = sinhc(x -a) (1.2')
r=1

- 10 -
The left side is

 dx
m x x m (x - a)2r-1
Σ = Σc
2r-1 2r-1 2r-1
c
r=1 a a r=1 (2r -1)!
Using this as the left side of (1.2) and (1.2') ,
2r-1
m (x - a)
Σ
r=1
c
2r-1
(2r -1)!
= sinhc(x -a)


2m-1
e cx +(-1)-2m+1e
-cx
2m-1 2m
x (x -a)
- (-1) c cosh cx dx
a (2m -1) ! 2

sinh cx
2m-1
e cx -(-1)-2m+1e
-cx
2m-1 2m (x -a ) x
+ (-1) c dx
(2m -1) ! a 2
(1.3)
 (x - a)2r-1
Σc
r=1
2r-1
(2r -1)!
= sinhc(x -a) (1.3')

When x =3, a =1, c =2.1, m =5 , Both sides of (1.3) are calculated as follows.

Also, (1.3') is correct. Because the left side is Taylor expansion of the right side.

Theorem 25.2.2
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

   f(x)sin cx dx
m x x x x
Σ(-1) r-1c
r=1
2r-1

a a
f(x)dx
2r-1
= c cos cx
a
f(x)cos cx dx + c sin cx
a
(2m -1)
cos cx  f(x)dx
x x x
cos cx +  dx
-(-1)2m-1c 2m
 
2m-1
a a 2 a
(2m -1)
sin cx  f(x)dx
x x x
sin cx +  dx
-(-1)2m-1c 2m
 
2m-1
a a 2 a
(2.1)

  f(x)dx
x x
Especially, when lim c 2m 
m 
2m-1
= 0,
a a

   f(x)sin cx dx
 x x x x
Σ(-1) r-1c
r=1
2r-1

a a
f(x)dx
2r-1
= c cos cx
a
f(x)cos cx dx + c sin cx
a
(2.1')

- 11 -
Proof
Formula of repeated integration by parts was as follows (" 1 Generalized Taylor's Theorem ") .

 
x m x x
f(x)g(x)dx = (-1) f <r>(x)g ( )(x)a + (-1)m f <m>(x)g (x)dx
r-1 r-1 (m)
a r=1
Σ a
When g(x)= sin cx , coscx ,
(r -1) m
(coscx)(r-1) = c r-1cos cx +    
(m)
, (coscx) = c mcos cx +
2 2
(r -1) m
(sin cx)(r-1) = c r-1sin cx +    
(m)
, (sin cx) = c msin cx +
2 2
Substituting these for the above,

(r -1) x

 f(x)coscxdx = Σ(-1) f  
x m
r-1 <r> r-1
(x)c cos cx +
a r=1 2 a
m
+ (-1)  f  
x
m <m>
(x)c mcos cx + dx
a 2
(r -1) x

 f(x)sin cxdx = Σ(-1) f  

x m
r-1 <r> r-1
(x)c sin cx +
a r=1 2 a
m
f  
x
+ (-1)m <m>
(x)c msin cx + dx
a 2
Here, let

 f(x)dx
x x
r =1, 2,  ,m
<r> r
f (x) =
a a

f (a) = 0 (r =1, 2,  ,m) ,

r
<>
Since

(r -1)
    f(x)dx
x m x x
r-1 r-1
f(x)coscxdx = Σ(-1) c cos cx +  r
a r=1 2 a a
m
+ (-1)   f(x)dx c cos cx +
  2 
x x x
m m m
dx
a a a
(r -1)
 2  
c sin cx +
x m x x
f(x)sin xdx = Σ(-1)
r-1 r-1
 f(x)dx r
a r=1 a a
m
+ (-1)   f(x)dx c sin cx +
  2 
x x x
m m m
dx
a a a
Expanding a part of the 1st term of the right side,
(r -1)
(-1) r-1c r-1cos cx + 
m
Σ 0 1 2 3
= c cos cx + c sincx - c cos cx - c sin cx ++--
r=1 2
(r -1)
 
m
Σr=1
r-1
(-1) r-1c sin cx +
2
0 1 2 3
= c sin cx - c cos cx - c sin cx + c cos cx ++--

Multiplying both sides by coscx, sincx respectively,

(r -1) 
 
m
cos cxΣ(-1) r-1c r-1cos cx +
r=1 2
0 2 1 2 2 3
= c cos cx + c sincx cos cx - c cos cx - c sin cx cos cx ++--

- 12 -
 (r -1) 
 
m
sin cxΣ(-1) r-1c r-1sin cx +
r=1 2
0 2 1 2 2 3
= c sin cx - c sincx cos cx - c sin cx + c sin cx cos cx ++--

The 2nd term of the right side is as follows.

   f(x)dx  c cos cx + m2 dx

x x x
(-1)mcos cx  m m
a a a

(-1) sin cx  f(x)dx c sin cx +

m
x x x

  2 
m m m
dx
a a a

2 2
Since cos cx + sin cx = 1 ,

    f(x)dx
x x m x x
cos cx f(x)cos cx dx + sin cx f(x)sin cx dx = Σ(-1) r-1c
2r-2
 2r-1
a a r=1 a a
2m -1 
  f(x)dx  cos cx + ( 2 ) dx
x x x
+ (-1)2m-1c 2m-1cos cx 2m-1
a a a
(2m -1)
sin cx  f(x)dx
x x x
sin cx + dx
+ (-1)2m-1c 2m-1
 
2m-1
a a 2 a
Here, m in the 2nd term and the 3rd term on the right side has to be an odd number corresponding to the 1st
term. Then, multiplying by c the both sides and transposing them, we obtain (2.1) .

     e dx
x x x x x x x x
c 1 e xdx  c 3   +- 
5 7
e xdx + c e xdx  c x
3 5 7
Example 2
a a a a a a a a

Substituting f(x)= e x for (2.1') ,

   e sin cx dx
 x x x x
Σ(-1) r-1c
r=1
2r-1
a

a
e xdx
2r-1
= c cos cx
a
e xcos cx dx + c sin cx
a
x

The right side is as follow.


x 1
2 e (csin cx coscx) e (csin ca cosca )
x x
e coscxdx = + - a +
a 1+ c


x 1
e xsin cxdx = 2 e (sin cx ccoscx) e (sin ca ccosca )
x
- - a -
a 1+ c
From these,

 
x x ex ea
cos x e xcos x dx + sin x e xsin x dx = 2
- 2
[cos{c(x -a )}- c sin{c(x -a )}]
a a 1+ c 1+ c
Therefore,

 e dx
 x x ce x ce a
Σ
r=1
(-1) r-1 2r-1
c
a a
x 2r-1
=
1+ c
2
-
1+ c
2
[cos{c(x -a )}- csin{c(x -a )}]

(2.2')
The higher integral of the left side is as follows according to Theorem 4.1.3 ( 4.1 )


x x (x -a)s n -1
e dx = e - Σe
x n x a
a a s=0 s!
Then, (2.2') becomes

- 13 -
 s
c ex ce a
  = 1+ c
 (x -a )
2r-2
Σ
r=1
(-1) r-1 2r-1
c x a
e -e Σ
s=0 s! 2
-
1+ c
[cos{c(x -a )}- csin{c(x -a )}]
2

(2.3')
When a =2 , c=0.8 , the first 10 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

- 14 -
25.3 Series of the even-th order Integrals
In this section, we ask for the sum of series of higher integral with geometric coefficients, such as

     f(x)dx
x x x x x x x x
c
2
f(x)dx  c
2 4
 f(x)dx + c
4 6
 f(x)dx  c
6 8
 8
+ 
a a a a a a a a

Theorem 25.3.1
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

c  f(x)dx =
   f(x)e dx 
m x x c x x
Σ
r=1
2r
a a
2r
2
e cx
a
f(x)e
-cx
dx - e
-cx
a
cx

sinh cx  f(x)dx

x x x

  cosh cx dx
2m
- c 2m+1
a a a

cosh cx  f(x)dx

x x x

  sinh cx dx
2m
+ c 2m+1 (1.1)
a a a

  f(x)dx = 0
x x
Especially, when lim c 2m+1  2m
,
m  a a

 f(x)dx = c2 e  f(x)e  f(x)e dx 

 x x x x
Σc r=1
2r
a a
2r cx
a
-cx
dx - e
-cx
a
cx
(1.1')

Proof
The following expressions were obtained during the proof of Theorem 25.2.1 .

   f(x)dx
x cx -r+1 -cx x x
r-1 r-1 e +(-1) e
m
f(x)cosh cxdx = Σ(-1) c  r
a r=1 2 a a

  f(x)dx 
m
x x x
m e cx +(-1)-me -cx
m
+ (-1) c dx
a a a 2

   f(x)dx
x m e cx -(-1)-r+1e -cx x x
f(x)sinh cxdx = Σ(-1)
r-1
c
r-1
 r
a r=1 2 a a

  f(x)dx 
cx
m
x x x
m e -(-1)-me -cx
m
+ (-1) c dx
a a a 2
Expanding a part of the 1st term of the right side,

e cx +(-1)-r+1e
m -cx
Σ(-1) r-1 c r-1
r=1 2
0 1
= c cosh cx - c sinhcx + c cosh cx - c sinh cx +-
2 3

e cx -(-1)-r+1e
m -cx
Σ
r=1
(-1) r-1 r-1
c
2
0 1
= c sinh cx - c cosh cx + c sinh cx - c cosh cx +-
2 3

Multiplying both sides by sinh cx, cosh cx respectively,

cx -cx
m
r-1 r-1 e +(-1)-r+1e
sinh cxΣ(-1) c
r=1 2
0 1 2 2 3 2
= c sinh cx cosh cx - c sinh cx + c sinh cx cosh cx - c sinh cx +-
e cx -(-1)-r+1e
m -cx
cosh cxΣ(-1) r-1 r-1
c
r=1 2
0 1 2 2 3 2
= c cosh cx sinh cx - c cosh cx + c cosh cx sinh cx - c cosh cx +-

- 15 -
The 2nd term of the right side is as follows.

  f(x)dx 
e cx +(-1)-me
x x x -cx
m m m
(-1) sinh cx c dx
a a a 2

(-1) cosh cx  f(x)dx c

e cx -(-1)-me
x x x -cx

 
m m m
dx
a a a 2
2 2
Since cosh cx - sinh cx = 1

    f(x)dx
x x m x x
sinh cx f(x)coshcx dx - cosh cx f(x)sinh cx dx = Σc
2r-1
 2r
a a r=1 a a

  f(x)dx  c
e cx +(-1)-2me
x x x -cx
2m 2m 2m
+(-1) sinh cx dx
a a a 2

cosh cx  f(x)dx

e cx -(-1)-2me
x x x -cx

 c
2m 2m
-(-1)2m dx
a a a 2
Here, m in the 2nd term and the 3rd term on the right side has to be an even number corresponding to the 1st
term. Furthermore,

  f(x)sinh cx dx
x x
sinh cx f(x)cosh cx dx - cosh cx
a a
cx cx cx cx

 
-cx -cx -cx -cx
e -e e +e e +e e
x x -e
= f(x ) dx - f(x ) dx
2 2 2 a a 2

2 
f(x)e dx - e  f(x)e dx
x x


1 cx -cx -cx cx
= e
a a
Using this, we obtain

    f(x)e dx 
m x x 1 x x
Σ c
2r-1
 f(x)dx
2r
= e cx f(x)e
-cx
dx - e
-cx cx
r=1 a a 2 a a

sinh cx  f(x)dx

x x x
- c 2m
  cosh cx dx
2m
a a a

cosh cx  f(x)dx

x x x
+ c 2m
  sinh cx dx
2m
a a a
Multiplying by c the both sides, we obtain (1.1) .

     dx
x x x x x x x x
c2   
2 4 4 6 6 8 8
Example 1 dx + c dx + c dx + c +
a a a a a a a a


x x (x -a)m
m
f(x)=1 , dx =
a a m!
Substituting these for (1.1) ,

    e dx
m x x c x x
Σ
2r
c 2r
dx = e cx e -cxdx - e -cx cx
r=1 a a 2 a a

sinh cx
2m+1
x (x -a)2m
-c cosh cx dx
a (2m)!


2m+1
x (x -a)2m
+c cosh cx sinh cx dx
a (2m)!

- 16 -
Here,
x x

 e  e dx    
c x x ce cx e -cx ce cx e cx
e cx -cx
dx - e -cx cx
= - -
2 a a 2 c a 2 c a
c(x-a ) -c(x-a )
e -1 -e +1
= -
2 2
= coshc(x -a) - 1
Then,

 dx 
m x x x (x -a)2m
Σ
2r 2r 2m+1
c = coshc(x -a) - 1 - c sinh cx cosh cx dx
r=1 a a a (2m)!


2m+1
x (x -a)2m
+c cosh cx sinh cx dx
a (2m)!
(1.2)

  dx
 x x
Σc 
2r 2r
= coshc(x -a) - 1 (1.2')
r=1 a a
The left side is

  dx
m x x m (x - a)2r
Σ c
2r
 2r
= Σc 2r
r=1 a a r=1 (2r)!
Using this as the left side of (1.2) and (1.2') ,


2r
m (x - a) x (x -a)2m
Σ
r=1
c
2r
(2r)!
= coshc(x -a) - 1 - c 2m+1sinh cx
a (2m)!
cosh cx dx


2m+1
x (x -a)2m
+c cosh cx sinh cx dx
a (2m)!
(1.3)
 (x - a)2r
Σ
r=1
c
2r
(2r)!
= coshc(x -a) - 1 (1.3')

When x =3, a =1, c =2.3, m =4 , Both sides of (1.3) are calculated as follows.

Also, it is clear that (1.3') is correct if coshc(x -a) is expanded to Taylor series.

- 17 -
Theorem 25.3.2
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

   f(x)sin cx dx
m x x x x
Σ(-1) r-1c
r=1
2r

a a
f(x)dx
2r
= c sin cx
a
f(x)cos cx dx - c cos cx
a

sin cx  f(x)dx

x x x

  cos (cx +m )dx

2m
- c 2m+1
a a a

cos cx  f(x)dx

x x x

  sin (cx +m )dx

2m
+ c 2m+1
a a a
(2.1)

  f(x)dx = 0
x x
Especially, when lim c 2m+1  2m
,
m  a a

 f(x)dx = c sin cx f(x)cos cx dx - c cos cx f(x)sin cx dx

 x x x x
Σ
r=1
(-1) r-1c
2r
a a
2r
a a
(2.1')

Proof
The following expressions were obtained during the proof of Theorem 25.2.2.
(r -1)
    f(x)dx
x m x x
r-1 r-1
f(x)coscxdx = Σ(-1) c cos cx +  r
a r=1 2 a a
m
+ (-1)   f(x)dx c cos cx +
  2 
x x x
m m m
dx
a a a
(r -1)
  2  
x m x x
f(x)sin cxdx = Σ (-1) c sin cx +  f(x)dx r-1 r-1 r
a r=1 a a
m
+ (-1)   f(x)dx c sin cx +
  2 
x x x
m m m
dx
a a a
Expanding a part of the 1st term of the right side,
(r -1)
(-1) r-1c r-1cos cx + 
m
Σ 0 1 2 3
= c cos cx + c sincx - c cos cx - c sin cx ++--
r=1 2
(r -1)
  = c sin cx - c cos cx - c sin cx + c cos cx ++--
m
Σ
r=1
(-1) r-1c r-1sin cx +
2
0 1 2 3

Multiplying both sides by

sin cx, cos cx respectively,
(r -1) 
 
m
sin cxΣ(-1) r-1 cos x +
r=1 2
0 1 2 2 3 2
= c sin cx cos cx + c sin cx - c sin cx cos cx - c sin cx ++--
(r -1) 
 
m
cos cxΣ(-1) r-1 sin x +
r=1 2
0 1 2 2 3 2
= c cos cx sin cx - c cos cx - c cos cx sin cx + c cos cx ++--
The 2nd term of the right side is as follows.

   f(x)dx  c cos cx + m2 dx

x x x
(-1)msin cx  m m
a a a

(-1) cos cx  f(x)dx c sin cx +

x xm x

  2 
m m m
dx
a a a

- 18 -
 2 2
Since cos cx + sin cx = 1 ,

    f(x)dx
x x m x x
sin cx f(x)cos cx dx - cos cx f(x)sin cx dx = Σ(-1) r-1c
2r-1
 2r
a a r=1 a a

sin cx  f(x)dx

x x x

  cos (cx +m )dx

2m
+ c 2m
a a a

cos cx  f(x)dx

x x x

  sin (cx +m )dx

2m
- c 2m
a a a
Here, m in the 2nd term and the 3rd term on the right side has to be an even number corresponding to the 1st
term. Then, multiplying by c the both sides and transposing them, we obtain (2.1) .

     x dx
x x x x x x x x
Example 2 c
2
x dx  c
2 2 4
 2
x dx + c
4 6
 x dx  c
2 6 8
 2 8
+ 
0 0 0 0 0 0 0 0

Substituting f(x)= x 2 for (2.1') ,

     x dx
x x x x x x x x
c
2
x dx  c
2 2 4
 2
x dx + c
4 6
 x dx  c
2 6 8
 2 8
+ 
0 0 0 0 0 0 0 0
The right side is,

 x cos cx dx =
2cx cos cx + c x -2sin cx
x 2 2
2
3
0 c

 x sin cx dx =
2cx sin cx - c x -2cos cx - 2
x 2 2
2
3
0 c
So,

 x cos cx dx - c cos cx x sincx dx =

x x 2 2
2 2 c x -2 + 2cos cx
c sin cx 2
0 0 c
Therefore,

 x dx
 x x 2 2
c x -2 + 2cos cx
Σ
r=1
(-1) r-1 2r
c
0 0
2 2r
=
c
2
(2.2')

The higher integral of the left side is


x x 2!
 x dx n =
2
x
2+n
0 0 ( 2+ n )!
Then, (2.2') becoms
 2 2
2! c x -2 + 2cos cx
Σ
r=1
(-1) r-1 2r
c
( 2+2r ) !
x
2+2r
=
c
2
(2.3')

When c=3.1 , the first 25 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

- 19 -
25.4 Series of Higher Integrals with coefficients
In this section, we ask for the sum of series of higher integral with arithmetic coefficients and geometric
coefficients, such as

     f(x)dx
x x x x x x x x
1c
1
f(x)dx  2c
2
f(x)dx + 3c
2 3
f(x)dx  4c
3 4
 4
+ 
a a a a a a a a

Theorem 25.4.1
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

rc r f(x)dx r = ce cx f(x)e dx + c e cx f(x)e dx

m x x x x x
Σ
r=1 a a a
-cx 2
a a
-cx 2

   f(x)dx  e
x x x x
- (m +1) c m+1 e cx  m-1 -cx
dx
2
a a a a

e   f(x)dx e dx
x x x x
+ m c m+2
 
cx m -cx 2
(1.1)
a a a a

  f(x)dx = 0
x x
Especially, when lim m c m+2  m
,
m  a a

Σrc  f(x)dx = ce  f(x)e dx + c e  f(x)e

 x x x x x
r r cx -cx 2 cx -cx 2
dx (1.1')
r=1 a a a a a

Proof
f (a) = 0 (r =1, 2,  ,m+ n -1), Theorem 16.1.2 ( 16.1 ) was as follows.
r
<>
When

  
x x m -1 -n
 f <0>g (0)dx n = Σ n r (r)
f < + >g
a a r=0 r


(-1)m n -1 n -1Ｃ k x x <m+ k > (m+ k ) n
 f
(n , m) Σ
+ g dx
k=0 m + k a a
r ()
Here, let g = e -cx , n =2 . Then, since g =(-1)rc re -cx ,

  
x x m -1 -2
r
f <0>e -cxdx 2 = Σ f <2+ >(-1)rc re -cx
a a r=0 r


(-1)m 1Ｃ 0 x x <m+0>
+ f (-1)m+0c m+0e -cxdx 2
(2, m) m +0 a a

(2, m) m +1 
(-1) Ｃ m
1 1 x x
<m+1>
+ f (-1)m+1c m+1e -cxdx 2
a a
i.e.


x x m -1
r
f <0> -cx
e dx 2 = Σ (1+ r)f <2+ >c re -cx
a a r=0

  f
x x x x
<m> -cx m+1 <m+1> -cx
+ (m +1)c m f e dx 2 - m c e dx 2
a a a a

 r 
-2 1
 (-1)r = 1+ r , = m(m +1)
 (2, m)
Here, subtracting 1 from the index <r> of the integration operator of the function f,

 f
x x m -1
<1+ r> r -cx
<-1> -cx 2
e dx = Σ (1+ r)f ce
a a r=0

- 20 -
   f
x x x x
m-1> -cx m+1 <m> -cx
+ (m +1)c m f< e dx 2 - m c e dx 2
a a a a
Left side becomes as follows.

  f
x x


x x
<0> -cx x
<-1> -cx
f 2
e dx = f e a +c <0> -cx
e dx dx
a a a a

= f dx + c f
x x x
<0> -cx <0> -cx
e e dx 2
a a a
Substituting this for the above,

 
x x x m -1
<1+ r> r -cx
f <0>e -cxdx + c f <0>e -cxdx 2 = Σ (1+ r)f ce
a a a r=0

 f  f
x x x x
m <m-1> -cx 2 m+1 <m> -cx
+ (m +1)c e dx - m c e dx 2
a a a a
From this,

f  f
m x x x
r-1 <r>
Σ
-cx <0> -cx <0> -cx
e rc f (x)= e dx + c e dx 2
r=1 a a a

 f  f
x x x x
m <m-1> -cx 2 m+1 <m> -cx
- (m +1)c e dx + m c e dx 2
a a a a
Here, let

  f(x)dx   f(x)dx   f(x)dx

x x x x x x
f
<r>
=  r
, f <m-1>
=  m-1
, f
<m>
=  m
a a a a a a

(r =1, 2,  ,m +1) , substituting these for the above

<r>
Since these satisfy the condition f (a )=0
cx
and multiplying by ce the both sides,

   f(x)e dx
m x x x x x
Σrc r
r=1 a

a
f(x)dx r = ce cx
a
f(x)e
-cx
dx + c e cx
2
a a
-cx 2

e   f(x)dx
x x x x
- (m +1) c m+1
  e dx
cx m-1 -cx 2
a a a a

e   f(x)dx e dx
x x x x
+ m c m+2
 
cx m -cx 2
(1.1)
a a a a

   dx
x x x x x x
+
1 2 2 3 3
Example 1 1c dx + 2c dx + 3c
a a a a a a

 dx  dx
x x m-1 x x m
m-1 (x -a ) m (x -a )
f(x) = 1 , = , =
a a (m -1) ! a a m!
Substituting these for (1.1) ,

rc r dx r = ce cx e dx + c e cx e dx

m x x x x x
Σ
r=1 a a a
-cx 2
a a
-cx 2


x x m-1
m+1 cx (x -a ) -cx 2
- (m +1) c e e dx
a a (m -1) !

 (x -a) m -cx 2
x x
m+2 cx
+ mc e e dx
a a m!
The 1st term and the 2nd term of the right side are as follows.

- 21 -
 e  e 
cx
x
cx
x x
cx e ca- e cx x e ca- e cx
+ ce cx
-cx -cx 2
ce dx + ce dx = ce dx
a a a c a c
c(x-a)
　　 = ce (x - a )
After the long calculation, the 3rd term and the 4th term on the right side become as follows.

 
x x m-1 x x m
(x -a ) (x -a) -cx 2
-(m +1) c m+1 e cx e dx + mc m+2 e cx
-cx 2
e dx
a a (m -1) ! a a m!
c(x-a)
= -ce (x -a)
m +1
e c(x-a c(x -a) m , c(x -a) - 1+ m , c(x -a )
)
+
(m -1) !
1
e c(x-a c(x -a) 1+ m, c(x -a) - 2+ m , c(x -a )
)
-
(m -1) !
Substituting these for the above,

  dx
m x x
Σrc r  r
= ce c(x-a (x -a ) - ce c(x-a (x -a )
) )
r=1 a a
m +1
e c(x-a c(x -a) m , c(x -a) - 1+ m , c(x -a )
)
+
(m -1) !
1
e c(x-a c(x -a) 1+ m, c(x -a ) - 2+ m , c(x -a )
)
-
(m -1) !
(1.2)

rc r dx r = c e
 x x
c(x-a )
Σ
r=1 a a
(x -a) (1.2')

The left side is

rc   dx
m x x m rc r(x -a ) r
Σ
r=1 a a
r
 r
=Σ
r=1 r!
Using this as the left side of (1.2) and (1.2') ,
m rc r(x -a) r
Σ = ce c(x-a (x -a) - ce c(x-a (x -a)
) )
r=1 r!
m +1
e c(x-a c(x -a) m , c(x -a) - 1+ m , c(x -a )
)
+
(m -1) !
1
e c(x-a c(x -a) 1+ m, c(x -a ) - 2+ m , c(x -a )
)
-
(m -1) !
(1.3)
 r r
rc (x -a)
Σ = c e c(x-a)(x -a) (1.3')
r=1 r!
When x =4, a =3, c=2, m =7 , both sides of (1.3) are calculated as follows.

- 22 -
 Further, when a =3, c=2 , the first 50 terms of ∑ are calculated and both sides of (1.3') are illustrated,
it is as follows. Both sides overlap exactly and blue (left) can not be seen.

Theorem 25.4.2
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

   f(x)e dx
m x x x x x
r-1
Σ(-1) rc r 
r=1 a a
f(x)dx r = ce
-cx
a
f(x)e cxdx - c e
2 -cx
a a
cx 2

e   f(x)dx
x x x x
- (-1)m(m +1) c m+1
 e
-cx m-1 cx 2
dx
a a a a

  f(x)dx  e dx
x x x x
+ (-1)m+1 m c m+2 e m cx
-cx 2
(2.1)
a a a a

  f(x)dx
x x
Especially, when lim m c m+2  m
= 0,
m  a a

   f(x)e
 x x x x x
Σ(-1)r-1 rc r 
r=1 a a
f(x)dx r = ce
-cx
a
f(x)e cxdx - c e
2 -cx
a a
cx
dx
2
(2.1')

Proof
In a way similar to Theorem 25.4.1, we obtain the desired expressioons

Example 2

     log x dx
x x x x x x x x
1c log x dx - 2c
2
log x dx + 3c
2 3
log x dx  4c
3 4
 4
+-
0 0 0 0 0 0 0 0

Substituting f(x)= log x , a =0 for (2.1) ,

- 23 -
    log x e dx
m x x x x x
r-1
Σ(-1) rc r 
r=1 0 0
log x dx r = ce
-cx
0
log x e cxdx - c e
2 -cx
0 0
cx 2

e   log x dx
x x x x

 e
m
- (-1) (m +1) c m+1 m-1 cx
-cx 2
dx
0 0 0 0

  log x dx  e dx
x x x x
m+1
m c m+2 e m cx
-cx 2
+ (-1)
0 0 0 0

The 1st term and the 2nd term of the right side are as follows.


x 1 cx x 1
e log|x| - Ei(cx)0 = e cx log|x|- Ei(cx)+  + log c
log xe cxdx =
0 c c

 
x x 1 x cx
log xe cxdx = e log|x|- Ei(cx)+  + log cdx
0 0 c 0
1 x
= 2 e cxlog|x|+1-(cx +1)Ei(cx)+ ( + log c)cx0
c
1 1
= 2 e cxlog|x|+1-(cx +1)Ei(cx)+( + log c)cx - 2(1- - log c)
c c
1 cx
= 2 e log|x|+1-(cx +1)Ei(cx)+ cx( + log c)+  + log c -1
c


x et
Where, Ei(x) = dt ,  = 0.5772 (Euler-Mascheroni Constant ).
- t
From this,

 log x e 
x x x -cx
ce
e log|x|- Ei(cx)+  + log c
-cx cx 2 -cx cx 2 cx
ce dx - c e log x e dx =
0 0 0 c
2 -cx
c e
e log|x|+1-(cx +1)Ei(cx)+ cx( + log c) +  + log c -1
cx
- 2
c
-  cx e
-cx -cx -cx -cx
= -1 + e + cx e Ei(cx)- cx e log c
Ei(cx)- log c - 
-cx -cx
= cx e +e -1
i.e.

  log x e
x x x
cx
ce
-cx
log x e cxdx - c e
2 -cx
dx 2 = cx e -cxEi(cx)- logc -  + e -cx - 1
0 0 0
Next,


x xm
 
x m 1
log x dx = m
log|x|-Σ
0 0 m! s=1 s

Using thes for the 3rd term and the 4th term on the right side, the right side becomes

Ei(cx )- logc -  + e
-cx -cx
cx e -1

 x m-1
x x

 
m -1 1
m
- (-1) (m+1) c m+1e log|x| -Σ e cxdx
-cx 2
0 0 (m -1) ! s=1 s

 mx !log|x| -Σ s  e
x x m
m+1 1 m
m c m+2e cx
-cx 2
+ (-1) dx
0 0 s=1
i.e.

- 24 -
 Ei(cx )- logc -  + e
-cx -cx
cx e -1

 
m +1 x x


m m -1 1
c m+1e x m-1 log|x| -Σ e cxdx
-cx 2
- (-1)
(m -1) ! 0 0 s=1 s

 
x x


m+1 1 m 1
c m+2e x m log|x| -Σ e cxdx
-cx 2
+ (-1)
(m-1) ! 0 0 s=1 s
Thus, we obtain

  log x dx
m x x
r-1
Σ (-1) rc r  r
= cx e -cxEi(cx)- logc -  + e -cx - 1
r=1 0 0

 
m +1 x x


m m -1 1
c m+1e x m-1 log|x| -Σ e cxdx
-cx 2
- (-1)
(m -1) ! 0 0 s=1 s

 
x x


m+1 1 m 1
c m+2e x m log|x| -Σ e cxdx
-cx 2
+ (-1) (2.2)
(m -1) ! 0 0 s=1 s

  log x dx
 r-1
x x
Σ (-1) rc r  r
= cx e -cxEi(cx)- logc -  + e -cx - 1 (2.2')
r=1 0 0

The left side is

 xn
x x

 
n 1
n
log x dx = log|x|-Σ
0 0 n! s=1 s

Using this,

r c rx r
 
m r 1
r-1
Σ(-1) log|x|-Σ = cx e -cxEi(cx)- logc -  + e -cx - 1
r=1 r! s=1 s

 
m +1 x x


m -1 1
m
c m+1e x m-1 log|x| -Σ e cxdx
-cx 2
- (-1)
(m -1) ! 0 0 s=1 s

 
x x


m+1 1 m 1
c m+2e x m log|x| -Σ e cxdx
-cx 2
+ (-1) (2.3)
(m -1) ! 0 0 s=1 s

 rx r
  =e
r 1
xEi(x) -   + e -x - 1
r-1
Σ(-1)
-x
log|x|-Σ (2.3')
r=1 r! s=1 s
When x =-0.2, c =1.7, m =3 , Both sides of (2.3) are calculated as follows.

Further, when c=1.7 , the first 30 terms of ∑ are calculated and both sides of (2.3') are illustrated, it is as

- 25 -
follows. Both sides overlap exactly and blue (left) can not be seen.

- 26 -
25.5 Calculation by Double Series
The sum of series of higher integral with geometric coefficients can also be calculated by a double series.
That way is almost the same as 24.4 ~ 24.6 .

Theorem 25.5.1
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

c  f(x)dx
s+r
 x x  (r)  c s(x -a )
Σ
r=0 a a
r r
= Σ f (a)Σ
r=0 s=0 (s + r)!
(1.1)

 e
 x x
= Σ f (r)(a)  c(x-a )
dx r (1.1')
r=0 a a

(-1) c  f(x)dx

s+ r
 x x  (r)  c s(x -a )
Σ
r=0 a a
r r r
= Σ f (a)Σ(-1)
r=0 s=0
s
(s + r)!
(1.2)

 e
 x x
= Σ f (a)  (r) -c(x-a )
dx r (1.2')
r=0 a a

Proof
r
()
Let f (a)= far and expand f(x) into Taylor series around a. Then

(x -a)0 1 (x -a )
1
2 (x -a )
2
3 (x -a )
3
f(x)= fa0 + fa  + fa  + fa  + (0)
0! 1! 2! 3!
Integrating both sides of this with respect to x from a to x one by one,


x 1 2 3 4
0 (x -a ) 1 (x -a ) 2 (x -a ) 3 (x -a )
f(x dx fa
) =  + fa  + fa  + fa  +
a 1! 2! 3! 4!
2
(x -a)3 (x -a)4 (x -a)5
 0 (x -a )
x x
f(x)dx = fa  + fa1 + fa2 + fa3
2
+
a a 2! 3! 4! 5!
(x -a)3 4 5 6

 f(x)dx 1 (x -a ) 2 (x -a ) 3 (x -a )
x x x
= fa  + fa  + fa  + fa 
3 0
+
a a a 3! 4! 5! 6!

c , c , c ,
0 1 2
Multiplying these also including (0) by , respectively,
0 0 0 1 0 2 0 3
c (x -a ) 1 c (x -a ) 2 c (x -a ) 3 c (x -a )
fa  + fa  + fa  + fa 
0 0
c f(x)= +
0! 1! 2! 3!

 f(x)dx =
1 1 1 2 1 3 1 4
x c (x -a) 1 c (x -a ) 2 c (x -a ) 3 c (x -a )
fa  + fa  + fa  + fa 
1 0
c +
a 1! 2! 3! 4!

c  f(x)dx
2 2 2 3 2 4 2 5
x x c (x -a) 1 c (x -a ) 2 c (x -a ) 3 c (x -a )
fa  + fa  + fa  + fa 
2 2 0
= +
a a 2! 3! 4! 5!

 f(x)dx
3 3 3 4 3 5 3 6
x x x c (x -a) 1 c (x -a ) 2 c (x -a ) 3 c (x -a )
fa  + fa  + fa  + fa 
3 3 0
c = +
a a a 3! 4! 5! 6!

Adding these perpendicularly.

- 27 -
    f(x)dx
x x x x x x
0
c f(x)+ c
1
f(x)dx + c
2
f(x)dx + c
2 3 3

a a a a a a

 
0 0 1 1 2 2 3 3
0 c (x -a ) c (x -a) c (x -a ) c (x -a)
= fa + + + +
0! 1! 2! 3!

 1! +
0 1 1 2 2 3 3 4
1 c (x -a ) c (x -a) c (x -a ) c (x -a)
+ fa + + +
2! 3! 4!

 2! +
0 2 1 3 2 4 3 5
2 c (x -a ) c (x -a) c (x -a ) c (x -a)
+ fa + + +
3! 4! 5!

s+ r
  c s(x -a)
=Σ far Σ
r=0 s=0 (s + r)!
i.e.

c  f(x)dx
s+r
 x x  (r)  c s(x -a )
Σ
r=0 a a
r r
= Σ f (a)Σ
r=0 s=0 (s + r)!
(1.1)

Furthermore,

 e
s+ r
 c s(x -a) x x
c(x-a )
Σ
s=0 (s + r)!
=
a a
dx r
Then

  e
 x x  x x
Σc  f(x)dx = Σ f (a) 
(r) r c(x-a )
r
dx r (1.1')
r=0 a a r=0 a a
(1.2) and (1.2') are also proved in a similar way.

Note
If (1.1') and (1.2') are written down, they are as follows.

   f(x)dx
x x x x x x
c f(x )  c f(x)dx  c + 
0 1 2 2 3 3
f(x)dx + c
a a a a a a

   e
x x x x x x
= f (0()a) f (1()a) e c(x-a)dx + f (2()a) e c x-a dx  f (3(
( ) ) 2
a) c(x-a)
dx + 
3
a a a a a a
These are similar to the Taylor series and are beautiful, so these are very good for viewing.
However, these are not so useful.

   log x dx
x x x x x x
1 2 2 3 3
Example 1 log x + c log x dx + c log x dx + c +
1 1 1 1 1 1

n =1, 2, 3, 
(n ) n-1
f(x) = log x , (log x) =(-1) (n -1)! x -n
Substituting these for (1.1), (1.1') ,

c  log xdx

s+r
 x x 
r-1  c s(x -1)
Σ
r=0 1 1
r r
= Σ(-1)
r=1
(r -1)!Σ
s=0 (s + r)!

 e
 x x
= Σ(-1)r-1(r -1)!  c(x-1)
dx r
r=1 1 1
When higher integrals are replaced with Riemann-Liouville Integral and x =0.8 , c=1.7 , m =10 are given,
each is calculated as follows.

- 28 -
   
x x x x x x
Example 2 x - c1 x dx + c 2 x dx 2 - c 3 x dx 3 +- 
1 1 1 1 1 1

f(x) = x
(1+1/2)
1
(n ) -n (2n -3)!!
f (1) = 1 2
= (-1)n-1
1+1/2- n  2n
Substituting these for (1.2),

(-1) c 
s+ r
 x x  (2r -3) !!  c s(x -a)
Σ
r=0 a a
r r r
x dx = e
-c(x-1)
2r s=0
+Σ(-1)
(s + r)!
r=0
r-1
Σ(-1)
s

When higher integrals are replaced with Riemann-Liouville Integral and x =0.9 , c=1.3 , m =15 are given,
both sides are calculated as follows.

During the proof of the previous theorem , if the calculation is done without including (0) ,
we obtain the following theorem.

Theorem 25.5.1'
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold


s+r
 x x   c s(x -a )
Σc r 
(r)
f(x)dx r = Σ f (a)Σ (1.3)
r=1 a a r=0 s=1 (s + r)!

 f(x)dx
s+ r

r-1 r
x x  (r) 
s-1 c s(x -a )
Σ
r=1
(-1) c
a a
r
= Σ f (a)Σ(-1)
r=0 s=1 (s + r)!
(1.3')

- 29 -
 Combining this theorem and 25.1 , we obtain the following theorem which gives the collateral integral
of the product of an exponential function and arbitrary functions.

Theorem 25.5.1"
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold
s+ r


x
(r) s-1   c s-1(x -a)
e f(x)dx = e cxΣ f (a)Σ(-1)
cx
(1.4)
a r=0 s=1 (s + r)!

e
s+r
x
(r)   c s-1(x -a )
-cx
f(x)dx = e -cxΣ f (a)Σ (1.4')
a r=0 s=1 (s + r)!

e
x
cx
Example 2" sin x dx
a

(sin x) n = sinx +n /2 n =1, 2, 3, 

( )
Since , from (1.4),

r

s+ r
c s-1(x -a )
 Σ(-1)
x  
s-1
e sin xdx = e Σ sin a +
cx cx
a r=0 2 s=1 (s + r)!
When a =2 , c=1.3 and the first 50 terms of ∑ are calculated and both sides are illustrated, it is as follows.
Both sides overlap exactly and blue (left) can not be seen.

Theorem 25.5.2
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold
2s+1+ r

 f(x)dx
2s+1
 x x   c (x -a)
Σ
(r )
= Σ f (a )Σ
2r-1 2r-1
c (2.1)
r=1 a a r=0 s=0 (2s +1+ r) !

  sinh c(x -a) dx

 x x
= Σ f r(a)  r
()
  (2.1')
r=0 a a
2s+1+ r

 f(x)dx
2s+1
 x x   c (x -a )
Σ
(r)
r-1 2r-1
= Σ f (a )Σ(-1)
2r-1 s
(-1) c (2.2)
r=1 a a r=0 s=0 (2s +1+ r) !

  sin c(x -a) dx

 x x
= Σ f r(a)  r
()
  (2.2')
r=0 a a

- 30 -
Proof
In a way similar to Theorem 25.5.1, we obtain the desired expressioons

     tan x dx
x x x x x x x x
Example 3 c
1
tan x dx  c
3
tan x dx + c
3 5
 5
tan x dx + c
7
 7

0 0 0 0 0 0 0 0
The higher differential quotient of tan x on x =0 is as follows according to Theorem 9.2.6 ( 9.2 )

(2n -1)
22n22n-1B2n
(tan x) |x=0 = T2n-1 = , (tan x)(2n)|x=0 = 0
2n
Where, T 2n-1 is the tangent number and B 2n is the Bernoulli number. Thus, from (2.1) ,


 2s+1 2s+2r
x x   c x
Σc
r=1
2r-1
0

0
tan x dx
2r-1
= ΣT2r-1Σ
r=1 s=0 (2s +2r) !
When higher integrals are replaced with Riemann-Liouville Integral and x =0.9 , c=1.2 , m =12 are given,
both sides are calculated as follows.

     secx dx
x x x x x x x x
Example 4 c
1
secx dx  c
3
secx dx + c
3 5
 secx dx - c
5 7
 7
- 
0 0 0 0 0 0 0 0
The higher differential quotient of secx on x =0 is as follows according to Theorem 9.2.8 ( 9.2 )

(secx)(2n)|x=0 = E2n , (secx)(2n +1)|x=0 = 0

Here, E2n is an Euler number. Thus, from (2.2) ,


 2s+1 2s+1+ 2r
x x   c x
Σ(-1)r-1c
r=1
2r-1

0 0
secxdx
2r-1
= ΣE2rΣ(-1)s
r=0 s=0 (2s +1+ 2r) !
When higher integrals are replaced with Riemann-Liouville Integral and x =0.3 , c=1.1 , m =12 are given,
both sides are calculated as follows.

- 31 -
Theorem 25.5.3
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

 f(x)dx
x 2s 2s+ r
 x  (r)  c (x -a)
Σ = Σf (a)Σ
2r 2r
c (3.1)
r=0 a a r=0 s=0 (2s + r)!

  cosh c(x -a) dx

 x x
= Σf (r)(a)   
2r
(3.1')
r=0 a a


x x 2s 2s+ r
   c (x -a)
Σ(-1)rc 2r 
(r)
f(x)dx =Σf (a)Σ(-1)s
2r
(3.2)
r=0 a a r=0 s=0 (2s + r)!

  cos c(x -a) dx

 x x
= Σf (a)  (r)
 
r
(3.2')
r=0 a a

Proof
In a way similar to Theorem 25.5.1, we obtain the desired expressioons

    tan
x x x x x x
Example 5 tan
-1
x +c
2
tan x dx  c
-1 2 4
 tan x dx  c
-1 4 6
 -1
x dx + 
6
1 1 1 1 1 1
According to " 岩波数学公式Ⅰ" p39, the following expression holds for a natural number n .

 
(n )
tan x = (n -1)!cos n(tan -1 x )sin n tan -1 x +
-1
2
From this,
3n
 
(n ) (n -1)!
tan x
-1
|x=1 = sin
2n/2 4
Substituting this for (3.1) ,

 tan
2s 2s
 x x   c (x -1)
Σ 4Σ
2r -1 2r
c x dx =
r=0 1 1 s=0 (2s) !
3r
2s 2s+ r
 (r -1)!  c (x -1)
+Σ
r=1 2r/2
sin  4 Σ s=0 (2s + r) !
When higher integrals are replaced with Riemann-Liouville Integral and x =1.8 , c=2.1 , m =17 are given,
both sides are calculated as follows.

During the proof of the previous theorem , if the calculation is done without including (0) ,
we obtain the following theorem.

- 32 -
Theorem 25.5.3'
When c is a positive number and a is a real number on the domain of analytic function f(x) ,
the following expressions hold

 f(x)dx = Σf (a)Σ (2s + r)!

x 2s 2s+ r
 x  (r)  c (x -a)
Σ 2r 2r
c (4.1)
r=1 a a r=0 s=1

Σ(-1) c  f(x)dx =Σf (a)Σ(-1)

x x 2s 2s+ r
   c (x -a )
r-1 2r 2r (r) s-1
(4.2)
r=1 a a r=0 s=1 (2s + r)!

Example 6

     sin
x x x x x x x x
c
2
sin x dx  c
-1 2 4
 -1
sin x dx + c
4 6
 sin x dx  c
-1 6 8
 x dx + 
-1 8
0 0 0 0 0 0 0 0
-1
The higher differential quotient of sin x on x =0 is as follows according to Theorem 9.3.2 ( 9.3 )
( 2n +1 )
sin x |x=0 = 2nＣ 0(2n -1)!!(2n -1)!!00 = (2n -1)!!2
-1

Then, from (4.2) ,

 sin
 2s 2s+2r+1
x x   c x
Σ
r=1
(-1) r-1 2r
c
0 0
-1
x dx
2r
= Σ(2r -1)!!
r=0
2
Σ
s=1
(-1) s-1
(2s +2r +1) !
When higher integrals are replaced with Riemann-Liouville Integral and x =0.6 , c=2.3 , m =10 are given,
both sides are calculated as follows.

2018.06.09
Kano Kono
Alien's Mathematics

- 33 -
26 Higher and Super Calculus of Zeta Function etc

26.1 Higher and Super Calculus of Riemann Zeta Function

26.1.1 Higher and Super Integral of Riemann Zeta Function

Formula 26.1.1h ( Higher Integral )

n
When z is Riemann zeta function,  <n>z is the lineal n -th order primitive, Hn ( =Σ1/k ) is
k =1

a harmonic number ( where H0 = 0 ) , the following expression holds on whole complex plane.
n-1 r+n
z -1  z -1
 n(z) = logz -1 - Hn-1 Σ
+ (-1) r
 r n =1, 2, 3, 
n -1! r=0 r + n!
(1.1h)
Where, r is Stieltjes constant defined by the following expression.

 
n (log k)r (log n)r+1
r = lim Σ -
n  k =1 k r +1

Proof
It is known that the Riemann zeta function z is expanded to Laurent series around 1 as follows.

1  (z -1)r
 (z) = + Σ(-1) r r
z -1 r=0 r!
Then, integrating the both sides with respect to z without considering the constant of the integration,
r+1
 z -1
 <1>
(z) = logz -1 + Σ(-1) r r
r=0 r +1!
Integrating this once more without considering the constant of the integration,
1 r+2
z -1  z -1
 <2>(z) = logz -1-1 + Σ(-1) r
r
1! r=0 r +2!
Integrating this once more without considering the constant of the integration,
2 r+3

  
z -1 1  z -1
 <3>(z) = logz -1- 1+ +Σ(-1) r r
2! 2 r=0 r +3!
Hereafter, by induction,
n-1 r+n

    +Σ(-1) 
z -1 1 1  z -1
 n
(z) = logz -1- 1+ ++ r
r
n -1! 2 n -1 r=0 r +n!
Rewriting the harmonic number as Hn-1 , we obtain the desired expression.

Example The 1st order integral

When the real part and the imaginary part of  <1>x +i y are illustrated, it is as follows. The left figure is
the real part and the right figure is the imaginary part.

- 1-
Note
If the left side of (1.1h) is represented by integral symbols, it is as follows.


z n-1 r+n
z z -1  z -1
  (z)dz =    n-1 Σ
log z -1 - H + (-1) r
 r
an a1 n -1! r=0 r + n!
Here, the lower limits of the integral are as follows.

a1 = 1.669008 , a2 = 2.641300 , a3 = 3.610288 , 

That is, this is a higher integral with variable lower limits. And this is a lineal higher integral .

cf.
The higher integral with variable lower limits a1 = a2 =  = an = 0 is as folows.


z n-1 r+n
z
z -1  z -1
  (z)dz n =    n-1 Σ
log z -1 -H + (-1) r
 r
0 0 n -1! r=0 r + n!

z -1
n-1 n -1 (-1) r Hr z n-1- r  n (-1) sr z n-s
- i +Σ -ΣΣ
n -1! r=1 r!n -1- r! r=0 s=1 r + s!n - s!
As seen from the existence of constant-of-integration polynomials, this is a collateral higher integral.

Formula 26.1.1s ( Super Integral )

When p is a complex number, z is Riemann zeta function,  <p>z is the lineal p -th order primitive,
p is gamma function, p is digamma function and r is Stieltjes constant, the following expression
holds on whole complex plane.

log(z -1)- (p) - 0  (z -1)r+p

 p
(z) = (z -1) + Σ(-1) r
p-1 r
(1.1s)
(p) r=0 (1+ r + p)

Proof
From (1.1h) ,

- 2-
 n-1 r+n
z -1  z -1
 n(z) = logz -1 - Hn-1 + Σ(-1) r
r
n =1, 2, 3, 
n -1! r=0 r + n!
At first,

(n -1)! = (n) , Hn-1 =  (n) + 0 , (r + n)! = (1+ r + n)

Using these,
n-1 r+n
z -1  z -1
 n(z) = logz -1 - n + 0 + Σ(-1) r
r
n r=0 1+r + n
Then, this expression also holds at n =0 . Because,

logz -1 - 0 logz -1 - 0 -(0)

= =0 , =1
(0)  (0)
So, replacing the natural number n with a complex number p , we obtain the desired expression.

Example The 0.1th order integral

z = x + i y , z and 
<0.1>
When z are illustrated as follows. The left figure is a real part and the right

figure is an imaginary part. In both figures, the orange is z and the blue is  <0.1>z . Since p is neer
0 , both curved surfaces look double.

26.1.2 Higher and Super Derivative of Riemann Zeta Function

Formula 26.1.2h ( Higher Derivative )

When z is Riemann zeta function,  (n)z is the lineal n -th order derivative and r is Stieltjes
constant, the following expression holds on whole complex plane.

(-1)-n n !  (z -1)r-n
 n
(z) = + Σ(-1)  r r
(1.2h)
(z -1)n+1 r=0 1+ r - n

Proof
It is known that the Riemann zeta function z is expanded to Laurent series around 1 as follows.

- 3-
 1  (z -1)r
 (z) = + Σ(-1) r
r
z -1 r=0 r!
Differentiating the both sides n times with respect to z,
n
r
 
1 
 n (n )
(z) = + Σ(-1)r (z -1)r
z -1 r=0 r!
According to Formula 9.2.1 in " 09 Higher Derivative ",
n 1+ -n
x   = x   0
1+ - n
- + n -n
= (-1)-n x  < 0
-
Applying this,
(n ) (1+ r) r!
(z -1)r = (z -1)r-n = (z -1)r-n
1+ r - n 1+ r - n
n
(1+ n)
 
1 n!
= (-1)-n (z -1)-1-n = (-1)-n
z -1 (1) (z -1)n+1
Substituting these for the above,

(-1)-n n !  (z -1)r-n
 n
(z) = + Σ(-1)  r r
(1.2h)
(z -1)n+1 r=0 1+ r - n

Example  (0.3) ,  -1.1 + 2.3i

(1) 2 ( )

When these are calculated by formula manipulation soft Mathematica , it is as follows. We can see that this
formula is numerically right.

Formula 26.1.2s ( Super Derivative )

When p is a complex number, z is Riemann zeta function,  (p)z is the lineal p -th order derivative,
p is gamma function, p is digamma function and r is Stieltjes constant, the following expression
holds on whole complex plane.
logz -1--p - 0  z -1
r-p
 p(z)= z -1
-p-1
+ Σ(-1) r r (1.2s)
-p r=0 1+ r - p

- 4-
Proof
From Formula 26.1.2h ,

(-1)-n n !  (z -1)r-n
 n(z) = + Σ(-1)r  r (1.2h)
(z -1)n+1 r=0 r - n !
According to Formula 1.3.1 in " 01 Gamma Function & Digamma Function " ,
 (-n)
(-1)-n n ! = - , n =0, 1, 2, 3, 
(-n)
Using this, (1.2h) is rewritten as follows.
logz -1--n - 0  z -1
r-n
 n(z)= z -1
-n-1
+ Σ(-1) r r
-n r=0 1+ r - p
Because,
logz -1 - 0 logz -1 - 0
= =0 for n =0, 1, 2, 3, 
-n 
So, replacing the natural number n with a complex number p , we obtain the desired expression.

logz -1--p - 0  z -1

r-p
 p(z)= z -1
-p-1
+ Σ(-1) r r (1.2s)
-p r=0 1+ r - p

Note
If the sign of p is inverted in (1.2s), it becomes as follows.

logz -1-p - 0  z -1

r+p
 -p
(z) = z -1
p-1
+ Σ(-1) r r
p r=0 1+ r + p
This results in Formula 26.1.1s . That is, the lineal super calculus of Riemann zeta function is seamless.

Example The 1.9th order derivative

z = x +i y ,   (1.9)z
(2)
When z and are illustrated as follows. The left figure is a real part and the

right figure is an imaginary part. In both figures, the orange is  (2)z and the blue is  (1.9)z . Since
p is near 2 , both curved surfaces look double.

- 5-
26.2 Higher and Super Calculus of Dirichlet Lambda Function

26.2.1 Higher and Super Integral of Dirichlet Lambda Function

Formula 26.2.1h ( Higher Integral )

n
When z is Dirichlet lambda function,  <n>z is the lineal n -th order primitive, Hn ( =Σ1/k )
k =1
is a harmonic number ( where H0 = 0 ) , the following expression holds on whole complex plane.
n-1
z -1
 <n>z = logz -1-Hn-1
2n -1!

 
log r+12 r-1 r

r+n
1  r-s z -1
+ Σ (-1) r
 r + -Σ  slog 2 n =1, 2, 3, 
2 r=0 r +1 s=0 s r + n!
(2.1h)
Where, r is Stieltjes constant defined by the following expression.

 
n (log k)r (log n)r+1
r = lim Σ -
n  k =1 k r +1

Proof
According to Formula 3.1.3 in " 03 Complementary Series of Dirichlet Series " ( Dirichlet Series ), Dirichlet
lambda function is expanded to Laurent series around 1 as follows.

 
log r+12 r-1 r

r
1 1  z -1
(z) = + Σ(-1) r r + -Σ s log 2r-s
2z -1 2 r=0 r +1 s=0 s r!
Where,  in { } is absent for r =0 .
Hereafter, in a way similar to the proof of Formula 26.1.1h , the desired expression is obtained.

Example The 2nd order integral

When the real part and the imaginary part of  <1>x +i y are illustrated, it is as follows. The left figure is
the real part and the right figure is the imaginary part.

Note
This is also lineal higher integral with variable lower limits. The first few of the integral lower limits are as follows.

- 6-
 a1 = 1.509052 , a2 = 2.203125 , a3 = 2.891846 , 

Formula 26.2.1s ( Super Integral )

When p is a complex number, z is Dirichlet lambda function,  <p>z is the lineal p -th order primitive,
p is gamma function, p is digamma function and r is Stieltjes constant, the following expression
holds on whole complex plane.
logz -1-p - 0
 <p>z = z -1
p-1
2p

   1+ r + p
log r+12 r-1 r

r+p
1  z -1
+ Σ (-1) r
 r + -Σ s log 2r-s (2.1s)
2 r=0 r +1 s=0 s  

Proof
In a way similar to the proof of Formula 26.1.1s , the desired expression is obtained.

Example The 0.05th order integral

z and  <0.05>z are illustrated as follows.

26.2.2 Higher and Super Derivative of Dirichlet Lambda Function

Formula 26.2.2h ( Higher Derivative )

When z is Dirichlet lambda function,  (n)z is the lineal n -th order derivative and r is Stieltjes
constant, the following expression holds on whole complex plane.

(-1) -n n !
 (n)z =
2z -1n+1

 
log r+12 r-1 r

r-n
1  z -1
+ Σ(-1) r +
r
-Σ s log 2 r-s
n =0, 1, 2, 
2 r=0 r +1 s=0 s 1+ r - n
(2.2h)

Proof
As seen in the proof of Formula 26.2.1h , the Dirichlet lambda function z is expanded to Laurent series

- 7-
around 1 as follows.

 
log r+12 r-1 r

r
1 1  r-s z -1
(z) = + Σ(-1) r + r
-Σ s log 2
2z -1 2 r=0 r +1 s=0 s r!
Where,  in { } is absent for r =0 .
Hereafter, in a way similar to the proof of Formula 26.1.2h , the desired expression is obtained.

Example  (-2.1) ,  -0.8 - 1.9i

(2) 3 ( )

When these are calculated by formula manipulation soft Mathematica , it is as follows. We can see that this
formula is numerically right.

Formula 26.2.2s ( Super Derivative )

When p is a complex number, z is Dirichlet lambda function,  (p)z is the lineal p -th order derivative,
p is gamma function, p is digamma function and r is Stieltjes constant, the following expression
holds on whole complex plane.
logz -1--p - 0
p(z) = (z -1)-p-1
2-p

   1+ r - p
log r+12 r-1 r (z -1)r-p

1 
+ Σ(-1) r r + -Σ s log 2r-s (2.2s)
2 r=0 r +1 s=0 s  

Proof
In a way similar to the proof of Formula 26.1.2s , the desired expression is obtained.

Note
If the sign of p is inverted in (2.2s), it becomes as follows.

(-p) logz -1-p - 0

 z = z -1
p-1
2p

   1+ r + p
log r+12 r-1 r

r+p
1  z -1
+ Σ(-1) r +
r
-Σ s log 2r-s
2 r=0 r +1 s=0 s  
This results in Formula 26.2.1s . That is, the lineal super calculus of Dirichlet lambda function is seamless.

- 8-
Example The 0.9th order derivative
z = x +i y ,   (0.9)z
(1)
When z and are illustrated as follows. The left figure is a real part and the

right figure is an imaginary part. In both figures, the orange is  (1)z and the blue is  (0.9)z . Since
p is near 1 , both curved surfaces look double.

- 9-
26.3 Higher and Super Calculus of Dirichlet Eta Function

26.3.1 Higher and Super Integral of Dirichlet Eta Function

Formula 26.3.1h ( Higher Integral )

When z is Dirichlet eta function,  <n>z is the lineal n -th order primitive,

(1) The following expression holds for z s.t. Rez > 0 .

zn  -n
r-1 log r
 n
(z) = + (-1) Σ(-1)
n
n =0, 1, 2,  (3.1h)
n! r=2 rz
(2) The following expression holds on whole complex plane.

r
r-1
zn  k (-1) k log -nr
 n
(z) = + (-1) ΣΣ
n
n =0, 1, 2,  (3.1h')
n! k =2 r=2 2k+1 rz

Proof
At Rez > 0 , Dirichlet eta function z is expressed with the following series which is called DIrichlet
eta series.
 1 1 1
(z) = 1 + Σ(-1)r-1e -z log r = 1 - + - +- 
r=2 2z 3z 4z
So, integrating the both sides n times with respect to z without considering the constant of the integration,

zn  log -n r
n(z) = + (-1) nΣ(-1) r-1 n =0, 1, 2,  (3.1h)
n! r=2 rz
Applying Euler transformation to this second term, we obtain (3.1h') . By this transformation, (3.1h) is
analytically continued from Rez > 0 to the whole complex plane. In addition, about Euler transformation,
see " 10 Convergence Acceleration & Summation Method by Double Series of Functions " ( A la carte ).

Example The 1st order integral

The imaginary part of  <1>x +i y are illustrated as follows. The left is (3.1h) and the right is (3.1h') .
In the left figure, the line of convergence is visible at x =0 .

- 10 -
Note
If the left side of (3.1h) is represented by integral symbols, it is as follows.


z z
zn  -n
r-1 log r
 (z)dz = + (-1) Σ(-1)
n
an a1 n! r=2 rz
Here, the lower limits of the integral are as follows.

a1 = -1.809613 + i 1.766080 , a 2 = 1.216967 ,

a3 = 1.337211 + i 1.289222 , a4 = 2.163768 , 
That is, this is a higher integral with variable lower limits. And this is a lineal higher integral .

Formula 26.3.1s ( Super Integral )

p is a complex number, z is Dirichlet eta function, 
<p>
When z is the lineal p -th order primitive
and p is gamma function,
(1) The following expression holds for z s.t. Rez > 0 .

zp p i

r-1 log
-p
r
 p
(z) = + e Σ(-1) (3.1s)
(1+p) r=2 rz
(2) The following expression holds on whole complex plane.

r
r-1
zp p i
 k (-1) k log -p r
 p
(z) = + e ΣΣ (3.1s')
(1+p) k =2 r=2 2k+1 rz

Proof
In Formula 26.3.1h , replacing n! with 1+n , replacing (-1) -n with e -n i and replacing the natural

number n with a complex number p , we obtain the desired expressions.

Example The 0.2th order integral

When z = x +i y , the real parts of z and  (0.2)z are illustrated as follows. The left figure is (3.1s)
and the right figure is (3.1s'). In both figures, the orange is z and the blue is  (0.2)z . Since p is near
0, both curved surfaces look double. In the left figure, the line of convergence is visible at x =0 .

- 11 -
26.3.2 Higher and Super Derivative of Dirichlet Eta Function

Formula 26.3.2h ( Higher Derivative )

When z is Dirichlet eta function,  (n)z is the lineal n -th order drivative and n is gamma
function,
(1) The following expression holds for z s.t. Rez > 0 .

z -n  n
r-1 log r
 n
(z) = + (-1) Σ(-1)
-n
n =0, 1, 2,  (3.2h)
(1-n) r=1 rz
(2) The following expression holds on whole complex plane.


r-1
z -n  k (-1) k log nr
 n
(z) = +(-1) ΣΣ
-n
n =0, 1, 2,  (3.2h')
(1-n) k =2 r=2 2k+1 r rz

Proof
As seen in the proof of Formula 26.3.1h ,
 1 1 1
(z) = 1 + Σ(-1)r-1e -z log r = 1 - + - +- 
r=2 2z 3z 4z
0
So, assuming 1 = z /0! and differentiating the both sides n times with respect to z,
z -n  n
r-1 log r
 n
(z) = + (-1) Σ(-1)
-n
n =0, 1, 2, 
(-n)! r=1 rz
Replacing -n! with 1-n , we obtain (3.2h) . In addition,

0
z -n 1 for n =0
=
(1-n) for n =1, 2, 3, 
And applying Euler transformation to this second term, we obtain (3.2h') . By this transformation, (3.2h) is
analytically continued from Rez > 0 to the whole complex plane.

Example  (0.2) ,  0.5 + 14.1i

(1) 3 ( )

If these are calculated according to (3,2h') by formula manipulation soft Mathematica , it is as follows. We can
see that this formula is numerically right.

Formula 26.3.2s ( Super Derivative )

When p is a complex number, z is Dirichlet eta function,  (p)z is the lineal p -th order derivative

- 12 -
and p is gamma function,
(1) The following expression holds for z s.t. Rez > 0 .

z -p  log p r
 p
(z) = + e -p iΣ(-1) r-1 (3.2s)
(1-p) r=1 rz
(2) The following expression holds on whole complex plane.


r-1
z -p -p  i
 k (-1) k log p r
 p
(z) = +e Σ Σ (3.2s')
(1-p) k =2 r=2 2k+1 r rz

Proof
In Formula 26.3.2h , replacing (-1) -n with e -n i and replacing the natural number n with a complex

number p , we obtain the desired expressions.

Note
If the sign of p is inverted in (3.2s), it becomes as follows.

zp  log -p r
-p(z) = + e p iΣ(-1) r-1
(1+p) r=1 rz
This results in Formula 26.3.1s . That is, the lineal super calculus of Dirichlet eta function is seamless.

Example The 2.2th order derivative

When z = x +i y , the real parts of  (2)z and  (2.2)z are illustrated as follows. The left figure is
In both figures, the orange is  z and the blue is 
(2) (2.2)
(3.2s) and the right figure is (3.2s'). z .
In the left figure, the line of convergence is visible at x =0 .

- 13 -
26.4 Higher and Super Calculus of Dirichlet Beta Function

26.4.1 Higher and Super Integral of Dirichlet Beta Function

Formula 26.4.1h ( Higher Integral )

When z is Dirichlet beta function,  <n>z is the lineal n -th order primitive,

(1) The following expression holds for z s.t. Rez > 0 .

zn  -n
r-1 log 2r -1
 n
(z) = + (-1) Σ(-1)
n
n =0, 1, 2,  (4.1h)
n! r=2 2r -1
z

(2) The following expression holds on whole complex plane.


zn  k (-1) r-1 k log -n2r -1
 <n>
(z) = + (-1) ΣΣ
-n
n =0, 1, 2,  (4.1h')
n! k =2 r=2 2k+1 r 2r -1
z

Proof
At Rez > 0 , Dirichlet beta function z is expressed with the following series which is called DIrichlet
beta series.
 1 1 1
(z) = 1 + Σ(-1)r-1e -z log2r-1 = 1 - + - +- 
r=2 3z 5z 7z
So, integrating the both sides n times with respect to z without considering the constant of the integration,

zn  -n
r-1 log 2r -1
 n
(z) = + (-1) Σ(-1)
n
n =0, 1, 2,  (4.1h)
n! r=2 2r -1
z

Applying Euler transformation to this second term, we obtain (4.1h') . By this transformation, (4.1h) is
analytically continued from Rez > 0 to the whole complex plane.

Example The 3rd order integral

The imaginary part of  <3>x +i y are illustrated as follows. The left is (4.1h) and the right is (4.1h') .
In the left figure, the line of convergence is visible at x =0 .

- 14 -
Note
If the left side of (4.1h) is represented by integral symbols, it is as follows.


z z
zn  log -n2r -1
 (z)dz = + (-1) nΣ(-1) r-1
an a1 n! r=2 2r -1
z

Here, the lower limits of the integral are as follows.

a1 = -1.027077 + i 0.978760 , a 2 = 0.754520 ,

a3 = 0.831711 + i 0.807718 , a 4 = 1.357089 , 
That is, this is a higher integral with variable lower limits. And this is a lineal higher integral .

Formula 26.4.1s ( Super Integral )

When p is a complex number, z is Dirichlet beta function,  <p>z is the lineal p -th order primitive
and p is gamma function,
(1) The following expression holds for z s.t. Rez > 0 .

zp  log -p 2r -1

 <p>(z) = + e p iΣ(-1) r-1 (4.1s)
(1+p) r=2 2r -1
z

(2) The following expression holds on whole complex plane.


zp  k (-1) r-1 k log -p2r -1
 (z) = + e p iΣΣ
<p>
(4.1s')
(1+p) k =2 r=2 2k+1 r 2r -1
z

Proof
n! with 1+n , replacing (-1) -n with e -n i
In Formula 26.4.1h , replacing and replacing the natural
number n with a complex number p , we obtain the desired expressions.

Example The 0.3th order integral

When z = x +i y , the real parts of z and  (0.3)z are illustrated as follows. The left figure is
(4.1s) and the right figure is (4.1s'). In both figures, the orange is z and the blue is  (0.3)z .
In the left figure, the line of convergence is visible at x =0 .

- 15 -
26.4.2 Higher and Super Derivative of Dirichlet Beta Function

Formula 26.4.2h ( Higher Derivative )

When z is Dirichlet beta function,  (n)z is the lineal n -th order drivative and n is gamma
function,

(1) The following expression holds for z s.t. Rez > 0 .

n
z -n 
r-1 log 2r -1
 (n)
z = + (-1) Σ(-1)
-n
n =0, 1, 2,  (4.2h)
1-n r=2 2r -1
z

(2) The following expression holds on whole complex plane.

log n 2r -1

z -n  k (-1) r-1 k
 (n)
z = + (-1) ΣΣ
-n
n =0, 1, 2, 
1-n k =2 r=2 2k+1 r 2r -1
z

(4.2h')

Proof
As seen in the proof of Formula 26.4.1h ,
 1 1 1
(z) = 1 + Σ(-1)r-1e -z log2r-1 = 1 - + - +- 
r=2 3z 5z 7z
Hereafter, in a way similar to the proof of Formula 26.3.2h , (4.2h) is obtained.
And applying Euler transformation to this second term, we obtain (4.2h') . By this transformation, (4.2h) is
analytically continued from Rez > 0 to the whole complex plane.

Example  0.9 + 8.7i , 

(2) (3)
(1.3)
If these are calculated according to (4,2h) by formula manipulation soft Mathematica , it is as follows. Though
the convergence is slow, we can see that this formula is numerically right.

Formula 26.4.2s ( Super Derivative )

p is a complex number, z is Dirichlet beta function, 
(p)
When z is the lineal p -th order derivative
and p is gamma function,
(1) The following expression holds for z s.t. Rez > 0 .
p
z -p -p  i

r-1 log (2r -1)
 p
(z) = +e Σ (-1) (4.2s)
(1-p) r=2 (2r -1)z

- 16 -
(2) The following expression holds on whole complex plane.


r-1
z -p  k (-1) k log p (2r -1)
 p
(z) = + e -p iΣΣ k+1 (4.2s')
(1-p) k =2 r=2 2 r (2r -1)z

Proof
In Formula 26.4.2h , replacing (-1) -n with e -n i and replacing the natural number n with a complex

number p , we obtain the desired expressions.

Note
If the sign of p is inverted in (4.2s), it becomes as follows.

zp p i

r-1 log
-p
(2r -1)
 -p
(z) = + e Σ(-1)
(1+p) r=2 (2r -1)z
This results in Formula 26.4.1s . That is, the lineal super calculus of Dirichlet beta function is seamless.

Example The 1.1th order derivative

When z = x +i y , the imaginary parts of  (1)z and  (1.1)z are illustrated as follows. The left figure
In both figures, the orange is  and the blue is 
(1) (1.1)
is (4.2s) and the right figure is (4.2s'). z . In the
left figure, the line of convergence is visible at x =0 .

2018.01.13
Kano Kono
Alien's Mathematics

- 17 -
27 Zeros of Super Derivative of Riemann Zeta

27.1 Zeros of  'z

27.1.0 Laurent Expansion of  ' z

Accordint to " 26 Higher and Super Calculus of Zeta Function etc " Formula 26.1.2h , when r is a Stieltjes

constant, the1st order derivative  (1)z of Rieamnn zeta fanction is expanded to Laurent series around 1
as follows

1 (z -1)r-1

 (1)
(z) = - + Σ(-1) r r
(1.0)
(z -1)2 r=0 (1+r -1)

As is clear from this formula and the figure,  (1)z has a pole of order 2 at z =1 .

27.1.1 Non-Trivial Zeros of  'z

According to " New zero-free regions for the derivatives of the Riemann Zeta Function " ( T. Binder etc.) ,

the followings are known regarding non-trivial zeros of  (1)z

(i) The Riemann hypothesis is equivalent to that the zeros of 
(1)
x +i y do not exist in 0 < x < 1/2 .
(ii) The zeros of 
(1)
x +i y exist in x < 2.93938 .
(iii) In x 1/2 , the number of zeros of   z .
(1)
z is less than the number of zeros of

In this section, we actually ask for the zeros of  (1)z bearing these things in mind.

At the zeros of  (1)z , both the real part Re (1)z and the imaginary part Im (1)z have
to be 0. That is, solutions of the following simultaneous equations must be obtained.

Re (1)(x +iy) = 0


(1.1r)

Im (1)(x +iy) = 0 (1.1i)

Since it is impossible to solve this, we are obliged to use the numerical solution method like the Newton-Raphson
method. Since the function FindRoot [] for it is implemented in formula manipulation software Mathematica,
we use this. However, for that purpose, approximate position of the zeros must be known. In order to know this,

- 1-
we draw contour plots of (1.1r) and (1.1i) and find the intersection. Since the function ContourPlot [] for it
is implemented in formula manipulation software Mathematica, we use this. At this time, x can be up to 3
at most from the above (ii) .

(1) 0  y  15
The contour plot in this interval is drawn as follows. Although (1 ,0) looks like the intersection of the real part

and the imaginary part, this is not a zero but a singular point (pole) . Therefore, there is no zero of  (1)z
in this interval.

(2) 15  y  25
Zero is found near (2.5 ,23) . The result of the numerical calculation is written on the right. This is the first

zero of  (1)z .

- 2-
(3) 25  y  35
Zero is found near (1 ,32) . However, an exact figure cannot be drawn by the series of (1.0) and the upper
one becomes a ghost. Of course, accurate numerical calculation is also impossible. So, it was calculated by
the function Zeta'[] of Mathematica . It is written on the right.

(4) 35  y  50
For y  35 , we draw and calculate by the left side ( Zeta'[] of Mathematica ) rather than the right side
of (1.0) . Then, the contour plot is drawn as follows. 3 zeros are seen. The results of the numerical calculation
are written on the right.

As mentioned above, 5 zeros were obtained in the Interval 0  y  50 . If these are recalculated with
significant 16 digits, it is as follows.

- 3-
For trial, if the first zero is substituted for Zeta'[z ] , it is as follows.

If these are plotted on the complex plane with the zeros of z , it is as follows. Blue is a zero of z
and Red is a zero of  (1)z . The latter number is reduced by half from the former. Further, each of the latter
exists on the right side ( y > 1/2 ) of the former.

27.1.2 Trivial Zeros of  'z

I have not seen trivial zeoros of  (1)z . So, first, let us draw the 2D diagram and the contour plot. Then,

- 4-
 Observing these diagrams, we can see that trivial zeros of  (1)z are distributed in the negative region of
the x -axis. If the first few are calculated with significant 16 digits, it is as follows.

Observing these, we can see that

(1) There is no zero near -2 .
(2) There is no zero near -4n n =1, 2, 3,  .
(3) There are zeros near -4n -1 n =1, 2, 3,  .
(4) The number of contours of the real part and the number of contours of the imaginary part are the same.

- 5-
27.2 Zeros of 
(1/2)
z

27.2.0 Laurent Expansion of Super Derivative of  z

Accordint to " 26 Higher and Super Calculus of Zeta Function etc " Formula 26.1.2s , when p is a positive
number, p is the Gamma Function, p is the Digamma Function and r is a Stieltjes constant,

the p  (p)z of Rieamnn zeta fanction is expanded to Laurent series around 1 as follows
th order derivative

logz -1--p - 0  z -1

r-p
 (z)=
p
z -1
-p-1
+ Σ(-1) r r
(2.0)
-p r=0 1+ r - p
-p
Where, = (-1)1+p p! for p =0, 1, 2, 
-p
p =1/2 , this is illustrated as follows. As is clear from this formula and the figure, 
(p)
When z has a pole
of order 1+p at z =1 .

Though the super differentiation for arbitrary p >0 is possible using (2.0) , we deal with the super derivative

between  (0)z and  (1)z in this chapter. And, in order to overlook the zeros of the super derivative

in this interval, it seems good to observe the zeros of  (1/2)(z) which is in the middle.
From (2.0) , the function used for this is as follows.
logz -1--1/2 - 0  z -1
r-1/2
 (1/2)
(z)= z -1
-3/2
+Σ(-1) r r
(2.1)
-1/2 r=0 r +1/2

27.2.1 Non-Trivial Zeros of 

(1/2)
z
The method of finding the zeros follows the method of the previous section. However, the interval of the real
part x of z should be slightly wider.

(1) 0  y  15
The contour plot in this interval is drawn as follows. The intersection of the real part and the imaginary part
( zero ) is found in two places. The results of the numerical calculation are written on the right. The bottom
intersection is not a zero but a singularity point.

- 6-
(2) 15  y  25
2 zeros are seen also in this interval. The results of the numerical calculation are written on the right. However,
The first zero is an estimate.

(3) 25  y  35
The contour plot is drawn as follows. Most are ghosts and the location of the intersection can not be seen.
This figure shows that it is impossible to obtain zeros in wider interval, by the above formula (2.1) and my
personal computer & the software.

- 7-
 As mentioned above, 4 zeros were obtained in the Interval 0  y  25 . If the first 3 are recalculated with
significant 16 digits, it is as follows.

For trial, if the first zero is substituted for (2.1) , it is as follows.

27.2.2 Trivial Zeros of 

(1/2)
z
The contour plots in -24  x  12 and -12  x  0 are respectively drawn as follows. The left is the
former and the right is the latter. The blue line is the real part and the orange line is the imaginary part.

- 8-
 If these are calculated with significant 16 digits, it is as follows.

Observing these, we can see that

(1) There is no zero near x = -2 .
(2) There is no zero near x = -4n n =1, 2, 3,  .
(3) There are zeros near x = -4n +2 n =1, 2, 3,  .
(4) The number of contours of the real part is half the number of contours of the imaginary part.

- 9-
27.3 Transition of Non-Trivial Zeros associated with the Super Dderivative.
In this section, we investigate the transition of non-trivial zeros associated with the Super Derivative using
formula (2.0) .

27.3.1 Non-Trivial Zeros of  z    y  2

(0) (1)
z 0
 z   0 y  2.
(2/10) (7/10)
Here, we investigate the transition of the zeros of z in The interval
of x is set to 0 x  2. Drawing these contour lines and looking for intersections, we obtain the followings.

The zero of  (p)z appears near p =3/10 , moves to the right at p =4/10 , 5/10 , 6/10 , and
disappears at p =7/10 . This is a zero peculiar to super derivative that appears only on non-integer order.

27.3.2 Non-Trivial Zeros of  z    y  16

(0) (1)
z 8
 z   z 8  y  16 .
(0) (1)
Next, we investigate the transition of the zeros of in The interval of x
is set to 0 x  7. if these contour lines and intersections are drawn on 1 figure, it is as follow.

- 10 -
 The blue line is the real part and the orange line is the imaginary part. The red dot is the intersection of these

namely the zero. The red dot on the left is the first zero 1/2+ i 14.1347 of (z) . This zero moves to
the right at p =1/10 9/10 and exists as long as p <1 . And the zero disappears at the moment of p =1 .
Because, at this time, the real part and the imaginary part become parallel and do not cross.

By reference, the zero of  (0.99999)z is z = 20.59398 + i 9.68836 .

27.3.3 Non-Trivial Zeros of  z    y  22.5

(0) (1)
z 18

Next, we investigate the transition of the zeros of  (0)z   (1)z in 18  y  22.5 . The interval
of x is set to 0  x  7.5 . if these contour lines and intersections are drawn on 1 figure, it is as follow.

The blue line is the real part and the orange line is the imaginary part. The red dot is the intersection of these

namely the zero. The red dot on the left is the 2nd zero 1/2+ i 21.0220 of (z) . This zero moves to
the right at p =1/10 9/10 and exists as long as p <1 . And the zero disappears at the moment of p =1 .
Because, at this time, the real part and the imaginary part become parallel and do not cross.

By reference, zero of  (0.999)z is about z = 13.984 + i 19.522 .

27.3.4 Non-Trivial Zeros of  z    y  26

(0) (1)
z 22.5

Last, we investigate the transition of the zeros of  (0)z   (1)z in 22.5  y  26 . The interval
of x is set to 0 x  3. if these contour lines and intersections are drawn on 1 figure, it is as follow.
The blue line is the real part and the orange line is the imaginary part. The red dot is the intersection of these
namely zero.

- 11 -
 The red dot on the left is the 3rd zero 1/2+ i 25.0108 of (z) . This zero moves to the right at

p =1/10 9/10 and finally reaches the zero 2.4631+ i 23.2983 of  (1)(z) .

Summary
Investigating the transition of the non-triviall zeros of  (p)(z) p = 0 1 in interval 0  y  26 ,
we obtained the following results.
(1) There exist non-trivial zeros peculiar to super derivative that appear only on non-integer order 0 < p < 1 .
(2) The first two zeros 1/2+ i 14.1347 and 1/2+ i 21.0220 of (z) move to the lower right with

increasing p , and disappear suddenly at  (1)(z) . This is consistent with the fact that the number of

zeros of  (1)z is less than the number of zeros of  z ( 27.1.1 (iii) ) .

(3) The 3rd zero 1/2+ i 25.0108 of (z) moves to the lower right with increasing p , and becomes

the 1st zero 2.4631+ i 23.2983 of  (1)(z) . This assists the assertion that the zeros of

 (1)(z) do not exist in 0 < x < 1/2 ( 27.1.1 (i) ) .

(4) It is impossible to obtain zeros in y  26 by the formula (2.0) and my personal computer & the software.
We need a faster convergent formula, a faster personal computer, a dedicated program, or all of them.

- 12 -
27.4 Transition of Trivial Zeros associated with the Super Derivative.
In this section, we investigate the transition of trivial zeros associated with the Super Derivative using formula (2.0) .

27.4.1 Transition of the trivial zero z = -2 at  z  

(0) (1)
z
 z  
(0) (1)
Here, we investigate the transition of the trivial zero z = -2 at z . The interval of x
is set to -3  x  4.2 . When the zero is traced with nine contour plots, it is as follows. The blue line is
the real part, the orange line is the imaginary part, and the red dot is a zero.

 (z) 's zero -2 draws a ballistic trajectory on a complex plane with increasing p , and falls near 1.77 .
0 0.67 show the transition. In 0.68 , this zero is separated into two zeros on the x -axis. In 0.78 , two

zeros move to the right and the left respectively. In 0.82 , the left zero is swallowed in the singular point 1 and

disappears. In 1, the right zero disappears. Simultaneously,  (z)'s zero -2.71726

(1)
appears on the left

side. This looks like the transmigration from x =1 . Additionally, the zeros existing on the half line x >1, y = 0
at 0.670597  p < 1 are trivial zeros peculiar to the non-integer order derivative.

- 13 -
27.4.2 Transition of the trivial zeros z = -4 , -6 at  z  
(0) (1)
z
 z  
(0) (1)
Here, we investigate the transition of the trivial zeros z = -4 , -6 at z . The interval of x
is set to -7.5  x  -3 . When the zeros are traced with nine contour plots, it is as follows.
The blue line is the real part, the orange line is the imaginary part, and the red dot is the zero.

In 0 , there are two zeros -4 , -6 of (z) . Zero -4 disappears at the moment p increases. Zero -6 draws

a lofted trajectory on a complex plane with increasing p , and falls to  (1)(z) ' s zero -4.93676 .
Simultaneously, another  (1)(z) ' s zero -7.07459 appears on the left side. This looks like the trans-

migration from -6 .

27.4.3 Transition of the trivial zeros z = -8 , -10 at  z  

(0) (1)
z

Here, we investigate the transition of the trivial zeros z = -8 , -10 at  (0)z   (1)z . The interval of x
is set to -11.5  x  -7.5 . When the zeros are traced with nine contour plots, it is as follows.
The blue line is the real part, the orange line is the imaginary part, and the red dot is a zero .

- 14 -
 In 0 , there are two zeros -8 , -10 of (z) . Zero -8 disappears at the moment p increases. Zero -10
draws a lofted trajectory on a complex plane with increasing p , and falls to  (z) ' s zero -9.17049 .
(1)

Simultaneously, another  (z) ' s zero -11.24121

(1)
appears on the left side. This looks like the trans-

migration from -10 .

Summary
Investigating the transitions of  (z) 's trivial zeros -2 , -4 , , - 10 associated  (p)(z) p = 0 1 ,
we obtained the following results.

(1) There exist trivial zeros peculiar to super derivative that appear only on non-integer order 0.670597  p < 1 .
And, these are real numbers greater than 1.
(2) Half of  (z) 's zero -2 transmigrates from the singular point 1 to  (1)(z) 's zero -2.71726 , and.
the other half disappears at the moment of p =1 . The above (1) occurs in this process.
(3)  (z) 's zero -6  (1)(z) 's zero -4.93676 , and transmigrates to  (1)(z) 's zero
moves to

-7.07459 simultaneously.  (z) 's zero -4 disappears at the moment of p > 0 .

- 15 -
(4)  (z) 's zero -10  (1)(z) 's zero -9.17049 , and transmigrates to  (1)(z) 's zero
moves to

-11.24121 simultaneously.  (z) 's zero -8 disappears at the moment of p > 0 .

2018.12.07
2018.12.20 Renewed
Kano Kono
Alien's Mathematics

- 16 -
28 Zeros of Super Integral of Riemann Zeta

28.1 Zeros of 
<1>
z

27.1.0 Series Expansion of 

<1>
z
Accordint to " 26 Higher and Super Calculus of Zeta Function etc " Formula 26.1.1h , when r is a Stieltjes

constant, the1st order lineal primitive  <1>z of Rieamnn zeta fanction is expessed as follows
r+1
 z -1
 <1>(z) = logz -1 + Σ(-1) r r (1.0)
r=0 r +1!

The 3D figures of the real part and the imaginary part are as follows. The left is the real part and the right is
the imaginary part. As is clear from both figures, there is no zero on the half line x <0 , y =0 .

28.1.1 Zeros of 
<1>
z

At the zeros of  <1>z , both the real part Re <1>z and the imaginary part Im <1>z have
to be 0. That is, solutions of the following simultaneous equations must be obtained.


Re (x +iy) = 0
<1>
(1.1r)

Im (x +iy) = 0
<1> (1.1i)

Since it is impossible to solve this, we are obliged to use the numerical solution method like the Newton-Raphson
method. Since the function FindRoot [] for it is implemented in formula manipulation software Mathematica,
we use this. However, for that purpose, approximate position of the zeros must be known. In order to know this,

we draw contour plots of (1.1r) and (1.1i) and find the intersection. Since the function ContourPlot [] for it
is implemented in Mathematica, we use this.

(1) Wide area figures

In order to explore the location of the zeros, let us draw 3 contour plots. The interval of y is set to 0  x  32
Then, it is as follows. The blue line is the real part, the orange line is the imaginary part, and the intersection is
a zero.

- 1-
 Fig.1 is -43  x  -20 . There is no zero here. Fig.2 is -20  x  3 . There are nine zeros here.
Fig.3 is 3  x  -26 . There is no zero here.
When these upper ( y  32 ) contour plots were drawn, the three plots became ghosts. It is difficult to draw
any more by the formula (1.0) and my personal computer & the software. However, the followings may be said.

(i) There seems to be no zero in the upper part of Fig.1 and Fig.3.
(iI) There seems to be innumerous zeros in the upper part of Fig.2.

(2) Enlarged figure

When Fig.2 is enlarged and the zeros are calculated with significant 16 digits using the FindRoot [] of
Mathematica , it is as follows.

- 2-
- 3-
28.2 Transition of Non-Trivial Zeros associated with the Super Integral.
In this section, we investigate the transition of non-trivial zeros associated with the Super Integral.

28.2.0 Laurent Expansion of Super Integral of  z

Accordint to " 26 Higher and Super Calculus of Zeta Function etc " Formula 26.1.1s , when p is a positive
number, p is the Gamma Function, p is the Digamma Function and r is a Stieltjes constant,

the p th order lineal primitive  <p>

z of Rieamnn zeta fanction is expessed as follows

logz -1-p - 0  z -1

r+p
 p(z) = z -1
p-1
+ Σ(-1) r r (2.0)
p r=0 1+ r + p
0 < p < 1, 
<p>
As is clear from this formula, when z has a pole of order 1- p at z =1 .

<0> <1>
28.2.1 Non-Trivial Zeros of  z   z 12.5  y  32
z  
<0> <1>
Here, we investigate the transition of the zeros of  z in 12.5  y  32 . The interval
of x is set to -3  x  2 . When the zeros are traced with 5 contour plots, it is as follows. The blue line is
the real part, the orange line is the imaginary part, and the red dot is a zero.

In Fig.1, the non-trivial zeros of  <0>z are drawn. In Fig.2, the zeros of  <0.2>z are drawn. Looking at
these figures, we can see that the zeros are moving to the left with increasing p . And finally they reach Fig.5.
If Fig.5 and Fig.1 are shown numerically, it is as follows.

- 4-
 -2.3007 + i 16.0854  1/2 + i 14.1347
-2.2574 + i 22.0090  1/2 + i 21.0220
-1.9080 + i 26.4561  1/2 + i 25.0108
-2.0862 + i 30.8578  1/2 + i 30.4248

Summary
The non-trivial zeros of z move to the upper left on the complex plane with increasing p .

- 5-
28.3 Transition of Trivial Zeros associated with the Super Integral.
In this section, we investigate the transition of trivial zeros associated with the Super Integral using formula (2.0) .

28.3.1 Transition of the trivial zero z = -2 at  z  

<0> <1>
z

z  
<0> <1>
Here, we investigate the transition of the trivial zero z = -2 at  z . The interval of x is set
to -2.25  x  2.25 . When the zero is traced with nine contour plots, it is as follows. The blue line is the real
part, the orange line is the imaginary part, and the red dot is a zero.

In 0 ,  (z) 's trivial zero -2 is drawn. Another intersection

1 is a singular point. This zero point transmigrates
to the immediate right of singular point 1 at the moment of p > 0 . 0.07 show the transition.
Although this transmigration also occurs at a very small order p 0 , the transmigration at p < 0.07 is not able
to be detected by the formula (2.0) and my personal computer & the software. Incidentally, an enlarged figure of
0.07 near 1 is as follows.

-6-
 After this, the zero point moves to the right with increasing p , and reaches 1.6690082124782 at 1.

28.3.2 Transition of the trivial zeros z = -4 , -6 ,  at  z  

<0> <1>
z
z = -4 , -6 ,   z  
<0> <0>
Here, we investigate the transition of the trivial zeros at z . The interval
of x is set to -20  x  0 . When the zeros are traced with nine contour plots, it is as follows. The blue line is
the real part, and the orange line is the imaginary part. The zeros are drawn with a red dot and a purple dot.

-7-
 In 0 ,  (z) 's trivial 8 zeros are drawn. -4 , -8 , -12 , -16 are drawn with a red dot and -6 , -10 , -14 , -18
are drawn with a purple dot.
Although all zeros disappear at the moment of p > 0 , among them, -4 , -8 , -12 , -16 revive in the order near
the origin with increasing p , and rises on the complex plane. On the contrary, -6 , -10 , -14 , -18 do not revive.
0.01 , 0.2 , 0.445 show the transitions. After
0.5 , these zeros rise further in the complex plane.
Although partner changes of the contour plots are also observed In 0.445  0.972 , the four zeros finally
reach 1 . The results of these transitions are numerically shown as follows.

-4  -5.2076 + i 6.7222
-8  -10.4916 + i 4.6111
-12  -14.1483 + i 2.6351
-16  -17.1410 + i 0.7711

Summary
 (z) 's trivial zeros -2 , -4 , -6 ,  p(z) p = 0 1 ,
< >
Investigating the transitions of associated
we obtained the following results.

(1)  (z) 's zero -2 transmigrates to the immediate right of singular point 1 at the moment of p >0 ,
and moves to the right on the x axis with increasing p
(2) Although  (z) 's trivial zeros -4 , -6 , -8 ,  disappear at the moment of p >0 , among them, zeros
-4 , -8 , -12 , -16 revive in the order near the origin with increasing p , and move to the upper left on the
complex plane. In addition, zeros -20 , -24 , -28 ,  can not revive at 0<p  1 .

2018.12.28
Kano Kono
Alien's Mathematics

-8-