1 2

PART I
GENERATING FUNCTIONS FOR PRODUCTS OF POWERS

OF FIBONACCI NUMBERS*
H. W. GOULD, WEST VIRGINIA UNIVERSITY, MORGANTOWN, W. VA.
1.
INTRODUCTION
We may define the Fibonacci n u m b e r s , F . by F n = 0, F< = 1,

'
n' J
' 1
+ F . A well-known generating function for these n u m b e r s is
d.i)
1i _ xx _ xx2
Zn -=JQ
F n = F
n+2
n+1
F nxn m
Intimately associated with the numbers of Fibonacci a r e the n u m b e r s of Lucas,

L , which we may define by L01 =, 2,
= L
+ L . The n u m b e r s
5 1L t = 1, L
'
n+29
n+1
n
n
F and L may be c o n s i d e r e d as special c a s e s of general functions first studied
in g r e a t detail by Lucas [8 J, though as Bell [ 1 ] has observed many expansions for
the Lucas functions appeared in p a p e r s of Cauchy and others p r i o r to Lucas.
Dickson [4 | devotes all of one chapter (17) to r e c u r r i n g s e r i e s and m o r e p a r t i c u larly Lucas functions.
Here one may find further r e f e r e n c e s to the many p a p e r s
on the subject which have appeared since Leonardo P i s a n o , or Fibonacci,
first
introduced the famous n u m b e r s in 1202. It would be difficult to estimate how many

p a p e r s r e l a t e d to Fibonacci n u m b e r s have appeared since Dickson's monumental
History was written, however it may be of i n t e r e s t to point out that a project h a s
been initiated under the direction of P r o f e s s o r
Vern Hoggatt,
San J o s e
State
College, San J o s e , California, to collect formulas, maintain a bibliography and c o ordinate work on Fibonacci n u m b e r s .
As p a r t of the w r i t e r ' s activity with this
Fibonacci Bibliographical P r o j e c t the subject of generating functions for powers

of the F i b o n a c c i ' n u m b e r s has come in for some study, and the object of this p r e s ent paper is to develop some very general generating functions for the
Lucas
functions.
* R e s e a r c h supported by a June R e s e a r c h Grant from the College of A r t s

Sciences of West Virginia University.
1
and
[April
Riordan [10 J has recently made a very i n t e r e s t i n g study of a r i t h m e t i c p r o p e r t i e s of c e r t a i n c l a s s e s of coefficients which a r o s e in his analysis of the generating
function defined by the p - t h powers of Fibonacci n u m b e r s .
fp W
d.2)
fP x n
n=0
where f
= F
Golomb [ 5 J had found essentially that for s q u a r e s of Fibonacci
n u m b e r s we have
(1 - 2x - 2x 2 + x 3 ) f2 (x) = 1 - x
and it was this which led Riordan to seek the general form of f (x).
However, t h e r e a r e other simple generating functions for the n u m b e r s of
Fibonacci. F i r s t of all, let us o b s e r v e that we may define the Fibonacci and Lucas
n u m b e r s by
(13)
KXm
'
=
n
un
-
a - b
, L = a11 + b n
'
n
where
(1.4)
a = | ( 1 +N/5) ,
b = | ( 1 - N/5) .
The very general functions studied by L u c a s , and generalized by Bell [ 1 , 2 ],

a r e essentially the F
and L
defined by (1.3) with a, b being the roots of the
quadratic equation x 2 = P x - Q so that a + b = P and ab = Q.
In view of this
formulation it is easy to show that we also have the generating function
ax
d-5>
bx
<~n
^T^F- = Z
^T
F o r many p u r p o s e s this expansion is e a s i e r to consider than (1.1), and one i s natur a l l y led to ask what form of generating function holds if we put p-th powers of the
1963]
OF FIBONACCI AND LUCAS NUMBERS
Fibonacci n u m b e r s
in (1.5).
Similar questions a r i s e for
3
LM.
We shall also
powe
consider negative powers
of. F , L , and suggest an analogy with the polynomials
of Bernoulli and Euler.
2.
GENERATING FUNCTIONS FOR LUCAS FUNCTIONS
Suppose we a r e given for any initial generating function,
(2-1)
F(x), say
Anxn
F(x) = Y
n=0
with no p a r t i c u l a r r e s t r i c t i o n s on A . It follows at once from this that
(2.2)
A n xn(an + bn) = ^ T
F(ax) + F(bx) = ^
n=0
An xn Ln
n=0
and
FI
(2S)
2: b
F(bx
OO
'
=Z\*nrn=0
This incidentally is r a t h e r like the method used by Riordan [10 ] to begin his study
of r e c u r r e n c e r e l a t i o n s for the generation function (1.2), except that we could study
(1.5) as well as (1.1) in the general expansion of (2.3).
Now of c o u r s e we may i t e r a t e upon the formulas (2.2) and (2.3) by making s u c c e s s i v e substitutions, replacing x by ax,
or bx,
and adding or subtracting, and
by this iteration build up generating functions involving F ^ and L

from (2.2)
F(a 2 x) + F(abx) = 2 A X n a11 L

'
'
n
n
F(abx) + F(b 2 x) = 2 A
x b
Thus we have
GENERATING FUNCTIONS AND PRODUCTS OF POWERS
[April
so that
F(a 2 x) + 2F(abx) + F(b 2 x) =
(2.4)
A n x11 L^
n=0
and in s i m i l a r fashion
(2 5)
F(a%) - 2F(abx) + Ffofrl

(a - b ) 2
V
^
xn
p2
Clearly we may proceed inductively to obtain a general r e s u l t .
We find the
general r e l a t i o n s
(2.6)
>
(J
F(ap-kbkx)
>
nÔ
An*nLP
and
P
(2.7)
oo
F aP kbkx
(a-b)^ Y> t - ^ f )
k=0
( "
) =E
n xIlF n '
n =0
In fact we may readily combine the r e l a t i o n s to obtain
(2. 8)
F(a
p+q-k-jbk+jx)=
AnxnFPL^
n=6
for any non-negative i n t e g e r s p , q.
Thus in principle we may set down generating
functions for products of p o w e r s of the Fibonacci and Lucas n u m b e r s , though the

r e s u l t may not usually be in the s i m p l e s t form;
We obtain (1.1) when A
= 1 identically;
The expansion analogous to (1.5) for
is
(1.5) when A
= 1/n!
identically.
1963 ]

OO
a X
/o
(2.9)
<ri
kX
Z_) n;
n=0
+ e
Proceding in the s a m e manner as above, it is c l e a r that we also have
F(amx) + F(bmx) = T
A xn L
n=0
(2.10)
and
F(a
-
(2 11)
v^11/
x] - F(b
l
r
a-b
x)
L
\
/
A x
n
ZLJ
n
F
mn
n=0
which include other well-known generating functions.
(2.12)
Consequently we have
with a corresponding result lor F ,
3.
RECIPROCALS OF FIBONACCI NUMBERS
Landau [7 ] showed that a certain s e r i e s of r e c i p r o c a l s of Fibonacci n u m b e r s

could be e x p r e s s e d in t e r m s of a Lambert s e r i e s . In fact he showed that if we write
n
(3.1)
L(x)
n=l
then we have
X
l
The device used to obtain this i s to expand (a - b)/(a
- b ) = (a - b ) / a
[April
1/(1 - z),
where z = (b/a) , by a power s e r i e s , and then invert the o r d e r of summation in

Landau f s r e s u l t i s noted i n B r o m w i c h [3, p. 194,
the s e r i e s , this being justifiable.
example 32], in Knopp [6, p. 279, ex. 144; p. 468, ex. 9], and in Dickson
p. 404].
Dickson also notes that 2 l / F
[4,
was put in finite form by A. A r i s t a .
Let us now define in general
R(x) ^
(3.3)
n=l
n f
n
Then by the s a m e technique we have used e a r l i e r to obtain (2.2) and (2.3) we see at
once that R(x) satisfies a functional equation
(3.4)
R(ax) - R(bx) = (a - b)
>
Thus if we have
(3.5)
R(x)
_
n=l
|_
then
R(ax) - R(bx) = ( a - b ) - j - ^
- \[E for ordinary Fibonacci

* "" x
numbers.
For R(x) as defined by (3.3) we also easily verify that
(3.6)
R(a 2 x) - R(b 2 x) = (a - b) ^
n=l
A^
n
L^
1963 J
To use this when A n = 1 we need to note the generating

&
& function for L n which is
a companion to (1.1). Let us obtain this from (2.2). Inasmuch as we have
(3.7)
a + b = 1,
a - b = N/5J and
we find
1
,
1
1 - ax ' 1 - bx
F(ax) + F(bx)
2 - (a + b)x
1 - (a + b)x + abx
2 - x
1 - x - x2
and so for the ordinary Lucas n u m b e r s we find
<3- 8 )
1 - x -
X2
^_J
n=0
xn
Sometimes this i s stated in the equivalent form (since L0 = 2)
x 2x +
(3. 9)
<
L xn .
*\ =Y
1 - x - x
Z_J
n=l
Thus if we have
(3
m\
10)
<c-i
TD/X _ \
R(X) = )
n=
Y~
l
then
R a2x
<
) - R * 2 x ) = (a-b) .
n
4.
x(2x
^ 1>q
1 - x - xBILINEAR GENERATING FUNCTIONS
We wish to turn next to some simple r e s u l t s for what a r e called bilinear gener a t i n g functions for Fibonacci and Lucas n u m b e r s .
To discuss this we first intro-
duce what we shall call a general Turaii operator defined by
(4. i )
Tf = T
f (x) = f (x + u) f (x + v) - f(x) f (x + u + v) .
[April
F o r the Fibonacci n u m b e r s it i s a c l a s s i c formula first discovered apparently by

Tagiuri (Cf. Dickson [4. p. 404]) and l a t e r given as a problem in the A m e r i c a n
Mathematical Monthly (Problem E 1396) that
<4-2>
n =
We may determine
n+u
n+v "
n+u+v = ^
a general bilinear generating function for the
2A x F
F
if we can first determine a r e s u l t for
n
n+u n+v
this let us consider 2A x F F
n
n n+j
Again, let us set as in (2.1)
2A x F F
n
n n+u+v
CO
Anxn
F(x) = 2
11=0
so that
*J. F(x) = Y \ *n+j

n=0
Then we find
J xJ F(ax) - b j J F(bx) = J
An xn+j(an+j - bn+j)
n=0
and hence ultimately
CO
( 4 .3)
v
;
aJF(ax) - b'F(bx)
a - b
Next we introduce F
y
/ ,
n=0
xn
n+j
by the s a m e device and we find

n
a j F(a 2 x), - b J F(abx) _ a 3 F(abx) - b 3 F(b 2 x)

a - b
a - b
2 A
(a
n+ j
_ bn}
series
To
do
1963]
and consequently we have

( 4 4^
K
' '
' a J F(a^x) - (a3 + b 3 ) F(abx) + b j F(b*x) _ ^

.
n
(a - b)2
" 2_* n x
n n+j '
n=0
Moreover, since we have the T u r a n expression (4.2), we have
00
(4.5)
K
' '
00
> A x
n
ZJ
n=0
n+
n+
=
v
> A xnF F
+ F F F(-x)
( X)
ZJ n
n n+u+v
u v
'
n=0
w h e r e the relation (4.4) i s used to simplify the right-hand side.
In principle at
l e a s t we have a way to write down explicit bilinear generating functions provided

m e r e l yj that we have F(x)
= 2A x
\
/
n
given.
&
It should be r e m a r k e d that since we have made no special assumptions about

the coefficients
in any of the work so far, we could apply our work to finite
s e r i e s just as well by supposing that A
= 0 identically for n ^ some value n 0 ,
As far as relation (4.4) is concerned, t h e r e is an alternative method.
Call the
s e r i e s M.(x), i . e . , let
M.(x) J '
ZJ
An x n F
.
n+j
n=0
If it Is possible to evaluate M0(x) = 2 A x F 2 ,
then one may note that M
(x)
= M.(x) + M._ 1 (x) and so a simple formula could be written down giving M. .
5. BERNOULLI AND EULER POLYNOMIALS

Returning to the p r o b l e m s p r e s e n t e d by r e c i p r o c a l s of Fibonacci and Lucas
n u m b e r s , it would appear to be of value to introduce some new polynomials b a s e d
upon the polynomials of Bernoulli and Euler.
Using a standard notation [9] we d e -
fine Euler and Bernoulli polynomials by
(5.D
>
k=0
k
EV<X) | r
2 e
e
tx
1
10
and
,k
\ - Bk. ww yk!
(5.2)
t e
"= t
tx
k=0
n
1
Now we have for general Lucas functions

1
L; ~t~r}j
t
.,, ~
u/
with c = b / a
T 3
a + b
a C + 1
so that
(5-3)
ctX
1 \
= AClL
where
b/a
C1 + 1
Similarly we find that when F, = (a - b )/(a - b) we have
(5.4)
ta C
*
- a>Ft
where
b/a
Cl - 1
The s i m i l a r i t y of (5.3) with (5.1) and (5.4) with (5.2) motivates what follows.
We define generalized Bernoulli and Euler polynomials by
tx
(5.5)
= ^
k<x>c>!r
and
(5 6)
tx
^T
E C)
2 ^ fe
[April
1963J
11
Now in fact
(5.7)
JLC^l
ct _ x
tlogC- eX(tl0gC)
e t log C _ 2
JL_
log C
OO
1
log C
zeXZ
ez _x
= 15FC 2 V ) = E Bk(x, (log C)1-1

k=0
k=0
so that
B k ( x ; C ) = (log C) k
(5,8)
B k (x)
Similarly one easily finds that
(5.9)
Ek(x;C)
= (log C ) k E k ( x )
Putting these observations together we ultimately have the expansions
(5.10)
L
t
^ ^
2 a (b/a)
E ( x )
(1ogb/a)
and
(5.11)
Jj- = _ k ^
t
t a (b/a)
V B k (x,
^
(log b / a ) 1 - 1
Thus we also have an amusing analogy between four n a m e s :
'
2)
}
BERNOULLI
FIBONACCI
EULER
LUCAS
12
[April
We may extend the analogy by considering the m o r e general Bernoulli and Euler
polynomials of higher o r d e r as d i s c u s s e d i n [9] and find expansions for the r e c i p r o cals of powers of the n u m b e r s of Fibonacci and Lucas.
We have
2^
(5.13)
k*
(X)
k!
^
,n xt
t e
,(et
, ,n
- 1)
and
*
eXt
I *!> *r " V7TF '
(5.14)
and we ultimately find as before the r e c i p r o c a l expansions (with C = b / a )
and
(5.16)
^
k ,
k
Qn n x t
1
) ^ t (log C)
- 2__C
^
== \ >
gE( n '(x)
Z_J
,t - +, 1n )nn"
Z_i
k w
k!
(C
k~0
nt ^tx
0n
2 a C
TLnt
Consequently we have
00
( 5 - 17)
t=m
00
k=0
""
00
fei
"
\a"C
m > 1
1963]
13
and
oo
î8>
ô
u-
l!s
oo
2*.h-""2
t?-2v
k=0 f
Pm
L
t=m
\'t
\a C J
m > 0.
In t h e s e , log C will be r e a l provided, e. g. that both a and b a r e positive,
or
both negative.
C.
In c a s e C i s negative we may take principal'values for the log

6.
SOME MISCELLANEOUS FIBONACCI FORMULAE
We shall conclude our r e m a r k s h e r e by d e r i v i n g a f e w miscellaneous r e l a t i o n s .

In relation (2.1) let A
(6-D
= J
2-> U
, z being any r e a l number. Then we find by (2.2)
LT
=_ / -(1! +, ax
>
^ z
<! +
bx
> ,
n=0
and by (2.3)
(6.2)
(A
xn p
6b W
(1 + ax) Z - (1 + bx) 2
Let us examine one special instance.
;b
Let z = r be any non-negative integer,
and take the o r d i n a r y F i b o n a c c i - L u c a s n u m b e r s when a = ^(1 + ^ 5 ) , b = (1 - N/S).

Then we find 1 + a = a 2 , 1 + b = b 2 , whence (6.1) and (6.2) become
(6-3)
n = d
+ b r
> =
a2r +
^ = L2r
n=0
and
(6.4)
to'.
n=0
2r
- Kb
a - b
2r
F
2r "
14
[April
It is c l e a r that by using the general r e l a t i o n s previously developed h e r e , we

could go on to derive many i n t e r e s t i n g F i b o n a c c i - L u c a s number r e l a t i o n s .
As
another example using the s a m e value for A , we find from (2.10) that
(6.5)
x n L m n = (1
amx)r
(1
In t h i s , let x = - 1 , m = 2, and a and b as above.
(6-6)
3 <-')
>
(| (-l)"L2n
bmx)r
Then we find
= (-1)1" L r
Also we have
J Q (-l)n FQ2n = v(-l)' r F r
(6.7)
n=0
By (4,3) we have
V
?
A xn F
n
n+
a JFT ?( a/ OxV) \
_ VJT
" hFQ>rt
a-b
n=0
and s i m i l a r l y we have
(6.8)
\
x\nxnLn+.
n=0
= a : i F(ax) + b^ F(bx)
OO
where as before in (2.1) we have F(x) = n =2 A n x .

,\
0
Let A = ( M and take x = 1. We then have for the ordinary Fibonacci-Luc as
numbers
19631
..>
Sn=0( D F - n+j
" = F2r+j
15
'
and
(6.10)
(r)
. = LQ
which a r e well-known r e c u r s i o n s .
A v e r y elegant s y m m e t r i c a l relation may be gotten from (2.6). In that relation
we choose A
= (
J and set x = a
formula then b e c o m e s , since F(a
,TX
A
(6,11)
Z-JA W
n=0
Similarly in (2.7) if we set A
wu>
i ,
The r e a d e r may easily verify that the
b x) = (1 + b a
Lp
aP
/pN
L;
ZJ
= L a
rn
n=0
= (
j and take x = - a
/r\
FP
P .
we find easily that
/P\ F r
i F,
.-
<.-^2<- ' (j>--'" 2:<- ' (j^

-p-Q
And s i m i l a r l y we find from (2.8) that (here we take x = a
,,18)
2(>-nM^M-^2-kQ20^T.-
With other choices of x we could give s i m i l a r r e s u l t s .
In fact with x = - a -p-q.
we have
- tr'0%^-Hrt)W%
n=0
E=0
jô
16

[April
1963J
It may be of i n t e r e s t to note that Kelisky [12] developed some curious r e s u l t s

involving Bernoulli, E u l e r , Fibonacci, and Lucas n u m b e r s . The r e l a t i o n s he gives
should be compared with those developed in the p r e s e n t p a p e r .
In p a r t i c u l a r ,
Kelisky has since written the p r e s e n t author that the unpublished proofs of the l a s t
collection of relations he found a r e somewhat s i m i l a r to the methods of the p r e s e n t
note.
REFERENCES
1.
E. T. Bell, A revision of the algebra of Lucas functions, Annals of M a t h . ,

(2)36(1935), 733-742.
2.
E. T. Bell, Arithmetical t h e o r e m s on Lucas functions and Tchebycheff polyn o m i a l s , A m e r . J. Math. , 57 (1935), 781-788.
3.
T. J. P a . B r o m w i c h , An Introduction to the Theory of Infinite S e r i e s ,
second
edition r e v i s e d , London, 1942.

4.
L. E. Dickson, History of the Theory of N u m b e r s , vol. 1, Washington, D. C. ,

1919.
5.
S. W. Golomb, P r o b l e m 4720, A m e r . Math Monthly, 64(1957), 49.
6.
Konrad Knopp, Theorie und Anwendung d e r unendlichen Reihen, fourth edition,

B e r l i n , 1947.
7.
8.
9.
10.
Edmund Landau, Sur la s e r i e des i n v e r s e s des n o m b r e s de Fibonacci, Bull.

Soc. M a t h . , F r a n c e , 27(1899), 298-300.
*
/
/
Edouard Lucas, T h e o r i e des fonctions n u m e r i q u e s simplement periodiques,
A m e r . J. Math. , 1(1878), 184-240; 289-321.
N. E. Nbrlund, Vorlesungen- \iber Differenzenrechnung, B e r l i n , 1924.
John Riordan, Generating functions for powers of Fibonacci n u m b e r s , Duke
Math. J. , 29(1962), 5-12.
11.
P r o b l e m E 1396, A m e r i c a n Math. Monthly, 67(1960), 81-82; 694.
12.
R. P . Kelisky, On formulas involving both the Bernoulli and Fibonacci n u m b e r s ,

Scripta Mathematica, 23(1957), 27-35.
A FIBONACCI ARRAY*
L.
C A R L I T Z , DUKE U N I V E R S I T Y ,
DURHAM, N.
C.
We take u0 = 0, u1 = 1,
u ,., = u + u .,
n+1
n
n-1
(n > 1) ,
;
'
and define
(1)
'
un
= u
0}n
n
as the 0-th row of the a r r a y

(2)
uljn
the first row of F.
F.
(n = 0, 1, 2,
'
We next put
= un+2
(n = 0, 1, 2,
F o r r ^ 2 we define u
by means of
J
r, n
(3)
= u
r,n
Thus u
+ u
r-l,n
(n = 0, 1, 2,
r-2,n
i s defined for all r , n ^ 0. It follows from the definition that

r,n
(4)
v ;
u
r,n
= u
i + u
rt
r,n-l
r,n-2
(n > 21
~ ' '
Indeed, assuming the truth of (4), we get
u , _,
r+l,n
= u
+ u .,
r,n
r-l,n
u
, + u
+ u 1
, + u . ,
r5n-i
r,n-2
r-l,n-l
r-l,n-2
r+l,n-l
14
r+l,n-2
Supported in p a r t by National Science Foundation Grant G16485.

17
A FIBONACCI ARRAY
[April
The following table is easily computed
r\n
f
13
211
13
21
34
55 1
11
18
29
47
76
12
19
31
50
81
131
11
19
30
49
79
128
207
15
13
18
31
49
80
129
209
338
21
29
50
79
129
208
337
545
13
34
47
81
128
209
337
546
883
21
55
76
131
207
338
545
883
1428
143
15
1
161 8
7
[j3
The s y m m e t r y p r o p e r t y
(5)
r,n
n,r
i s easily proved by making use of (3) and (4).

We now put
f r (x)
(6)
V
>
Z_i
>n
(r = 0, 1, 2, . . . )
n=0
In p a r t i c u l a r , it follows from (1) and (2) that
(?)
o(x) = i -
- x2
'
fl(x)
1 + X
*2
and by (3) we have also
(8)
Vx>
= f
r-l ( x )
WX)
(r
" 2)
>
A FIBONACCI ARRAY
1963]
19
Using (7) and (8), we prove readily that
+ u
Thus (6) yields

(10)
u
r,n
= u u , _. + u , . u
r n+1
r+1 n
which again implies the truth of (5).

If we put
f(X y) =
'
CO
CO
Z,
r=0
n=0
r n
r,n
X y
then by (9)
f(x,y)
^
- ^
r=0
r + ur+ly
l-y-y2
"
1
f
x
,
y_
l-y-y2[l-x-x2+l-x-x2
so that
(ii)
f( X , y) =
[
>Y)
y
(l-y-y2)
(l-x-X2)
We r e m a r k that (10) is equivalent to

(I 2 )
u
r,n
= u u + u
r n
r+n
as is easily proved,
It a p p e a r s from the table that
(13)
v
u ^
n - u
r+l,r-l
r3r
= (-l)r
v
(r > 1)
;
20
A FIBONACCI ARRAY
[April
Indeed (13) holds for r = 1. Then
V2,r - V l . r u = (Ur+l,r
+ U
rr) " ( V l . r " V l , r - l )
u. . - u _ r_x
r,r V l , r -
. .r+1
= (-1)
Also the relation
(14)
V2,r-2-Vr = ^r+1
<rs2>
is suggested; the proof of (14) is s i m i l a r to the proof of (13).

In the next place we have
(15)
V3,r-3-
r,r
= ^
<r ^ 3)
The general formula of which (13), (14), and (15) a r e special c a s e s is
(16)
uM
- u
= (-l)r
r+s,r-s
r,r
S+1
u2
s
(r > s)
Indeed it follows from (12) that
r+s,r-s
~r,..r ..
.. r+s " r - s
and (16) is an easy consequence.

F o r a l a t e r purpose we shall r e q u i r e the formula
(17)
"1
fo *>
V"1
hsu> u
=<
Zj
r,r
|2un+1un_1
( n even)
(n odd) .
A FIBONACCI ARRAY
1963]
This is equivalent to
n-l,n-l
2(u2 - u u 0 ); == 2u u -.
n
n n-2
n n-1
(n even)
2(u ^ u 1 - uL _.) = 2u u 1
n+1 n-1
n-1
n n-1
(n odd)
which i s in agreement with (10).

In connection with (17) we note that
(18)
ur , r x
= (1 + x)(l - 3x + x-)
r=0
F o r m u l a s of this kind a r e perhaps most easily proved by using the
representation
n
a
_n
- 13
a - (3
where
i + \H
1 - *sf5
To i l l u s t r a t e we shall evaluate
<-,
\
u
L>
r
n+r?r
r=0
Since by (12)
*[>
~ , n+2r+l
_,n+2r+l\
, , r n
,n
u ,
= u , u + u , n = - l o2 b
+ 8
- (-1) [a + (3
v
J
n+r,r
n+r r
n+2r 5 ' >
'
/
M
>
[April
A FIBONACCI ARRAY
22
we get
_, , 0 n+1
11 2a
u n+r,
, r x = 5\ l-a2x
nnn+l
n
n
2/3
_ a + 0(3
2
l-/3 x
1+x
r=0
n
1
5
/ 2 (v , - - v .. x)
v
n+1
n -2 1 '
n
1 - 3x + x
"l + x
where
(19)
vn = a
^n
+ (3
It follows that
T 22 !
\
Z^
r=0
( 2 )
r
U
n+r, r X
i (v _L1 + v J C l - x2) + 5v x
- i
n+1
n-l;
'
n
" 5
(1 + x)(l - 3x + x 2 )
When n = 0. (20) r e d u c e s to (18). When n = 1, 2 we get
\~"
2,
r=0
(21)
(22
'v"1
Xi
r=0
>
r?r+l
r,r+2X
1 + x - x2
(l + x)(l - 3 x + x * )
1 + 3x - x
(1 + x)(l - 3 x + x 2 )
>
'
respectively.
Returning to (11), we r e p l a c e x, y by xt, yt,
(23)
"sT* ji \*
r n-r
2 , t ZVn-rXy
n=0
r=0
r e s p e c t i v e l y , so that
(x + y ) t
(1-xt-x2t2)(l-yt-y2t2)
A FIBONACCI ARRAY
1963]
23
Since the right m e m b e r of (23) is equal to
x +y
(x-y)(x"+3xy + y)
xy + x-(x-ry)t _ xy-t-y*(x + y)t

I-Xt-X2^"
l-yt-y2t2
OO
x +y
(x-y)(x2+3xy+y2
xy + y 2 (x + y)t
xy + x" (x + y)t
"X-1
Vi
>
U , -, X
n ,n i
it follows that
V"1
n r
xy(x + y)(x -v )u
-(x-y)-(x
-y
)u
r=0
The polynomials
= D (x,y)
n
n 'J'
r-0
r5n-r
r n-r
x y
J
c o r r e s p o n d to the secondary diagonals in the Fibonacci a r r a y .
For example, we
have
D0 = 0 3
Dj + x + y ,
D2 =- ( x - y r r
D3 -
2(x-. y) 3 - 3 x y ( x - y ) ,
D4 -
3(x -r y) 4 - 7xy(x ~ y) 2 .
Since
n-rl
x
- y
x - y
n-1
^-D
1 r l (x y)
(x - y)
n-2r
24
[April
A FIBONACCI ARRAY
we find, after a little manipulation, that (24) implies
(25)
( X
'
y )
- - - E [ t > n - ( r > n
l ] ^
-2r+2
(x + y ) 2 r - (~l) r (xy) r
(x + y) 2 + xy
In p a r t i c u l a r , if we take
x = a =
1 + -s/5
i - 4s
Y = 0
(25) r e d u c e s to
D (
n ^> = H O *
(26)
n+1
However, it is s i m p l e r to make use of (11).
/ "n
It is easily verified that
OO
CO
n=0
n=0
so that
(27)
Dn(*,|3) - ]T ( " 1 ) r
( r + 1)U
2n-2r
r=0
It is not obvious that (26) and (27) a r e identical.
As an instance of (27), we have
D 4 (a, (3) = u 8 - 2u6 + 3u4 - 4u2 + 5u0 = 21 - 16 + 9 - 4 = 10.
1963]
A FIBONACCI ARRAY
25
In the next place we evaluate the determinant

u
u
r,m
A (r, s; m , n)
r?n
u
s,m
ssn
Using (10) we get

A (r, s; m , n ) = (u u , . - u , n u )(u , ., u
'
' 7
r s+1
r+1 s 7 m+1 n
- u u n) .
m n+1 7
Since, for n m,
, -.U
m+1 n
- U
, - = -(U
m n+1
-, - U
m n-1
-U
m - 1 n7
(-1)
- UU
(UHU
'
n-m
-, )
n-m+1'
it follows that
(28)
A(r,s;m,n)
= (_i)m+r+1u
7
In p a r t i c u l a r , when m = r , n = s,
(29)
'
'
(28) b e c o m e s
( r 5 s ; r , s ) = - u2g_r
(s => r) .
Consider the s y m m e t r i c m a t r i x of o r d e r
(30)
(n>m,s>r).
n-m s-r
M = (u
)
n
r,s7
n:
( r , s = 0, 1, . . . , n-1)
Clearly the rank of M
< 2 and indeed is equal to 2 for n > 2.
istic polynomial of M
is given by
P (x) = x
n
n-1
V
- > u
r=0
n-1
The character-
n-2
:..(r, s; r , s ) x '
x
o<r<s<n
26
A FIBONACCI ARRAY
[April
n-1
The coefficient of x
can be found by means of (1.7).
n-2
x
, it follows from (29) that
As for the coefficient of
n-2
V
*r,B;r.*)
<r<s<n
= -
Y]
o<r<s<n
-r
n-1
= -
r=0 s = r + l
n-2 n - r - 1
n-1
s-r
n-r-1
= - v v ^ = - y y .
r=0
s=l
r=0
s=0
But
n-1
n-1
zs=0 u s = z,
r
s=0
? " ( } J" ^^~ rf*""
= 1 - v 0 + v - 2e
2n-2
2n
n
w h e r e as above v
=a
+ (3
,
'
and
(3D
p
={
1
(n even)
(n odd) .
Then
n-1 n - r - 1
5
E
r=0
n-1
U
2
s=0
E
r=0
'
+ V
2n-2r-2
2n"2
+ V
2n-2r "
[|(n
+l|]
2
( n + 1 )
= V
2 n -
so that
< 32 )
/>
/
A ( r , s ; r , s ) - - | (v 0
i
o<r<s<n
2n
Vr
- 2 -e )
n7
- V
1963]
A FIBONACCI ARRAY
T h e r e f o r e , using (17) and (32), we find that the c h a r a c t e r i s t i c polynomial of

i s given by
x 11 - 2 U 2 x 1 1 " 1 (33)
27
U2
xn~2
(n even)
P n (x) = +
l** ~ 2 u n+l u n-l x n _ 1 " K - i ) ^ " 2 <n od d, n >1).
F o r example, we have
p 2 (x) = x 2 - 2x - 1
p 3 (x) = x 3 - 6x 2 - 3x ,
as can be verified directly.

By m e a n s of (33) we can compute the c h a r a c t e r i s t i c values of M . In addition
to n - 2 z e r o s we have
u2 u J u 2 + 1
n
nv n
(n even)
(34)
u , - u 1 / u 2 , - u 2 1 + u2 - 1
I n+1 n - 1
v n+1 n - 1
n
(n odd)
NOTICE TO ALL SUBSCRIBERS!!!

P l e a s e notify the Managing Editor AT ONCE of any a d d r e s s change. The P o s t Office
Department,
r a t h e r than forwarding magazines mailed third c l a s s ,
directly to the d e a d - l e t t e r office.
sends them
Unless the a d d r e s s e e specifically r e q u e s t s the
Fibonacci Quarterly be forwarded at first c l a s s r a t e s to the new a d d r e s s , he will not

r e c e i v e it. (This will usually cost about 30 cents for f i r s t c l a s s postage.) If p o s s i b l e ,
p l e a s e notify us AT LEAST THREE WEEKS PRIOR to publication dates: F e b r u a r y 15,
April 15, October 15, and December 15.
[April 1963]
FIBONACCI PAPERS READ
28
S0 L B a s i n , 'Fibonacci N u m b e r s / p r e s e n t e d to the Cupertino High School Mathem a t i c s Club, Cupertino, Calif,, F e b r u a r y 1 1 , 1963.
B r o t h e r U, Alfred, 'Fibonacci D i s c o v e r y , ' p r e s e n t e d to the California Mathematics
Council, Northern Section, at St. M a r y ' s College, Calif., March 30, 1963.
Verner E. Hoggatt, J r .
'Fibonacci N u m b e r s , ' p r e s e n t e d to the California Mathe-
m a t i c s Council, Northern Section, at St. M a r y ' s College, Calif. , March 30, 1963.
D. W. Robinson, 'The Fibonacci Matrix Modulo m , ' p r e s e n t e d to the Mathematical
Association of A m e r i c a , March 9, 1963.
H. W. Gould and T. A. Chapman, 'Solution of Functional Equations Involving Turan
E x p r e s s i o n s , ' p r e s e n t e d to the West Virginia Academy of Science, April 26, 1963.
H0 W. Gould,
A b - p a r a m e t e r S e r i e s T r a n s f o r m with Novel Applications to B e s s e l
and Legendre p o l y n o m i a l s , ' p r e s e n t e d to the A m e r i c a n Mathematical Association,

May 4, 1963.
N. J . F i n e , 'An Elementary Arithmetic M e a s u r e , ' p r e s e n t e d to Allegheny Mountain
Section of the A m e r i c a n Mathematical Association, May 4, 1963.
FIBONACCI RELATED MASTER'S THESES
1. John E. Vinson, 'Modulo m p r o p e r t i e s of the Fibonacci S e q u e n c e , ' Oregon State
University, 1961, Advisor: Prof. Robert Stalley.
2.
Charles H. King, 'Some P r o p e r t i e s of the Fibonacci S e q u e n c e , ' San J o s e
State
College, 1960, Advisor: Prof. V e r n e r E. Hoggatt, J r .

3.
R i c h a r d A. H a y e s , 'Fibonacci and Lucas P o l y n o m i a l s , ' San J o s e State
College,
Advisor: Prof. V e r n e r E. Hoggatt, J r . (Not yet completed.)

4.
S i s t e r Mary de Sales McNabb, 'Fibonacci N u m b e r s : Some P r o p e r t i e s and

r a l i z a t i o n s ' Catholic University of A m e r i c a ,
Advisor:
Prof.
Gene-
Raymond
W.
5. L. Basin, An Application of Continuants as a Link Between Chebyshev
and
Moller. (Not yet c o m p l e t e d . )

FIBONACCI ARTICLES SOON TO A P P E A R
Fibonacci, Mathematics Magazine.

D. Zeitlin,
On Identities for Fibonacci N u m b e r s ,
C l a s s r o o m Notes,
American
Mathematical Monthly.
A. F , Horadam, On Khazanov's F o r m u l a e , Mathematics Magazine.
D. E. Thoro, Regula F a l s i and the Fibonacci N u m b e r s ,
matical Monthly.
The A m e r i c a n
Mathe-
THE FIBOiACCi MATRIX MODULO 111"

D.
W.
ROBINSON, C A L I F O R N I A
INSTITUTE
OF TECHNOLOGY,
PASADENA,
CALIF,
In this paper we investigate some of the arithmetical p r o p e r t i e s of the famous

Fibonacci sequence by use of e l e m e n t a r y m a t r i x algebra.
to be conceptual ancls at l e a s t in p a r t , novel.
We believe the approach
Thus, it is our purpose to explore
the pedagogical advantages of m a t r i x methods for p r o b l e m s of this kind as well as

to provide a refreshing appreciation of the a r i t h m e t i c a l p r o p e r t i e s t h e m s e l v e s .
At
the conclusion of the paper we also indicate how the methods may be applied to other
l i n e a r r e c u r r e n t sequences.
We begin by considering the following example.
Suppose that the Fibonacci
sequence
0, 1. 1, 2. 3, 5, 8, 13. 2 1 , 34 ; 55. 89, 144.. . . .
i s reduced modulo 8:
, 0, 1 , 1 ,
2. 3. 5, 0, 5. 5. 2, 7. 1, 0. F
We observe that the reduced sequence is periodic.
1,
Indeed, the 12 t e r m s of the
period form two s e t s of 6 t e r m s each, the t e r m s of the second half being 5 t i m e s

the corresponding t e r m s of the first half.
We say that the Fibonacci sequence r e -
duced modulo 8 is of period 12 and r e s t r i c t e d period 6 with multiplier 5,
Also, we
o b s e r v e that the multiplier is of exponent 2 modulo 8.

More generally, let u 0 , Uj, , u , *** be the Fibonacci sequence of i n t e g e r s
satisfying u
= u n + 1 + u n for n > 0 with
(U()5 at) = (0, 1).
Given any integer
m ^ 1 we provide below an elementary proof of the fact that t h e r e is a positive

integer n such that (u , u
. ) = (0. 1) (mod m).
The least such integer
6(m)
is
called the period of the Fibonacci sequence modulo m. The least positive integer n
sLieh that (u . u
ir
..) = s(0. 1) (mod m), where s
n+1'
'
r e s t r i c t e d period aim)
is some integer, is called the
of the sequence modulo m. If (u
(mod m). 0 < s(m) < m,

quence modulo m.
'
. u
) - s(m) (0.1)
then s(m) is called the multiplier of the Fibonacci s e -
Obviously
s(m) = u
exponent modulo m of the multiplier
,-.
(mod m),
Finally, we denote the
s(m) by /j(m).
^ P r e s e n t e d to the Mathematical Association of A m e r i c a , Southern California Section, March 9, 1963.

f P r e s e n t l y on sabbatical leave from Brigham Young University and at California
Institute of Technology under an N. S. F. Faculty Fellowship P r o g r a m .
29
THE FIBONACCI MATRIX-MODULO m
30
[ April
By d i r e c t calculation we obtain the following table:
10
11
12
13
14
15
16
17
18
19
Q,(m)
12
12
15
10
12
24
20
12
12
18
(3{my
6(m)
24
16
12
24
60
1
6
4
20
1
10
4
24
2
28
2
48
2
40
24
4
36
2
24
1
18
The r e s u l t s of this table i l l u s t r a t e s e v e r a l i n t e r e s t i n g a r i t h m e t i c a l p r o p e r t i e s .

In fact, if (a,b) a n d [ a , b ] denote the g r e a t e s t common divisor and the l e a s t c o m mon multiple, r e s p e c t i v e l y , of the integers a and b ,
then we propose to e s t a b -
lish the following:

(i) m ju
if and only if a(m) in, and m | u , m|(u
(ii) 6(m) = a(m)P(m)
. - 1) if and only if <5(m) i n;
= (2,/3(m)) [y(m),ar(m)] , where 7(2) = 1 and 7(m) = 2
for m > 2;
(iii) a([m1,m2]\=
[fffm]), a(m2)] , and 6([m 1 , m 2 ]) = [&(mt). 6(m 2 )] ;
(iv) for every odd p r i m e p t h e r e i s a positive integer e(p) such that

, . max(0.
e-e(p))
, s. e.
, max(0. e-e(p))
v
= a(p)p
'
^" and <5(p
) = w<5(p)p
;
(v) #(p)|(p - (5/p)), where (5/p) is the usual Legendre symbol;
if p ^ 5,
then
6(p)|(p-l) or
a(p )
furthermore,
6(p)|2(p + 1).
With the possible exception of the last equation of (ii). which i s due to Morgan
Ward, these p r o p e r t i e s a r e all well known. Indeed, the fact that reduced sequences
of this type a r e periodic was observed by J. L. Lagrange in the eighteenth century.
A century l a t e r E. Lucas engaged in an extensive study of the a r i t h m e t i c d i v i s o r s
of such sequences.
These early r e s u l t s together with some of the l a t e r develop-
ments in the subject a r e reviewed in Chapter 17 of Dickson's History [6].
How-
e v e r , it is suggested that this general background be supplemented with at l e a s t

the p a p e r s of Carmichael [ 3 ] , L e h m e r [ l l ] , and Ward [ 1 9 ] ,
(See also [4, 7,
8,
9, 10, 17, 20, 2 1 ] . )

F u r t h e r m o r e , since the main purpose of this p r e s e n t p a p e r is to indicate the
use of m a t r i x algebra for the study of linear r e c u r r e n c e r e l a t i o n s , we also r e m a r k
that such techniques a r e certainly not new.
(See for example [ 1, 13, 16, 18]. ) In
fact, some of the a r i t h m e t i c a l p r o p e r t i e s of linear r e c u r r e n t sequences have been

studied by means of m a t r i c e s .
(See for example [ 2 , 12, 1 5 ] . ) It is our aim to now
indicate some of the further possibilities of this method.
THE FIBONACCI MATRIX MODULO m
1963]
31
We begin by introducing the main tool of our discussion.
Specifically,
view the linear r e c u r r e n c e above as defining a mapping of the o r d e r e d p a i r (u
we
,u )
onto the o r d e r e d p a i r (u u ,..)

Since u . = u ., + u 9. it i s c l e a r that this
^
n' n + l ;
n+1
n-1
n
(u J, u , _,),
wher^
mapping i s r e p r e s e n t e d by the m a t r i x product (u _1 , u )U
n n+l;?
^
0
F u r t h e r m o r e , by induction on n," we observe that ( u n > u n + i )
(>
)u
and
. u
n-1 n
u
u. ^
n
n+1
B e c a u s e of these r e s u l t s we call U the Fibonacci x matrix a

F r o m the foregoing it is evident that
if U
(u ,u
..) =
(0, 1) (mod m) if and only
i s congruent (elementwise) modulo m to the identity m a t r i x . Thus, the study
of the period of the Fibonacci sequence modulo m is equivalent to the study of the
p e r i o d of the sequence I, U, U 2 ,
reduced modulo m.
since t h e r e a r e only a finite number of distinct m a t r i c e s in this reduced sequence,

it follows that t h e r e a r e i n t e g e r s k and n such that U
with k + n > k > 0.
But since the determinant of U is the unit - 1 , this
that for some positive integer n, U

such positive integer n,
= 1 (mod m).
which is in fact
means
Thus, t h e r e exists a l e a s t
<5(m) as defined above. Also, it is c l e a r
that every such n i s an integral multiple of

and only if
i s congruent to U
<5(m).
That i s ,
= I (mod m)
if
<5(m)|n, which is equivalent to the second statement of (i).
By a s i m i l a r argument we have (u , u _.) = s(0, 1) (mod m) if and only if

n
n
U = si (mod m). Indeed, U i s congruent to a s c a l a r m a t r i x modulo m if and
only if a(m)\n, where a(m) i s the r e s t r i c t e d period defined above. This r e s u l t is
equivalent to the first p a r t of (i).
F u r t h e r m o r e , we have
ifW
w h e r e the multiplier
(mod m)
= s(m) I
s(m) i s of exponent /3(mj modulo m.
Since
= s ( m ) ^ m ^ 1 = 1 (mod m ) , 6(m)|cKm)/3(m). On the other hand, since it is evident that

a(m)\ <5(m),
we have by a s i m i l a r argument that
/3(m)| 6(m)/a(m).
= a(m)/3(m), which e s t a b l i s h e s the first equation of (ii).
Thus,
<5(m)
32
[April
Also, since the determinant of U is - 1 , we have from the m a t r i x congruence

above that
( _ 1 ) ^(m)
^ (S(m))2
(modm)
Hence, these congruent i n t e g e r s have the s a m e exponent modulo m.
Specifically,
7(m)
_
g(m)
(/(m), or(m))
(2,0(m))
where
7(m) i s the exponent of -1 modulo m.
Tnat is
6(m) = (mW?<m) = (2,/3(m)) ^ J ^ y
which is c l e a r l y equivalent to (ii). In p a r t i c u l a r , we observe that
<5(m) i s even for
m > 2 and that /3(m)|4.

We now demonstrate the second equation of (iii).
T
m |m,
then U
= I (modm ) and
divide
[m1.m2],
<5(m2).
On the other hand, suppose d(mt)
it follows that
We first observe that if
6(m )| <5(m). Thus, since m 1 and m 2 both
6([m 1 ,m 2 ]) i s a common multiple of

and
congruent to the identity m a t r i x modulo both m t and m 2 ,

valid modulo [m 1 ? m 2 ] ,
That i s ,
l e a s t common multiple of
5(m1) and <5(m2).
d(rn.t) and
<5(m2) both divide <5. Since U
^([m^mg])
the congruence is also
divides d and i s therefore
We obtain s i m i l a r l y the f i r s t equation of (iii).
is
the
Thus, we observe that both a
and 6 a r e factorable (1. c. m. multiplicative) functions of the argument m, which

suggests next the consideration of property (iv).
T h e r e f o r e , let p be any odd p r i m e and let e be any positive integer.
U
6(P
p6(p
G+1
Since
* = I + p B for some m a t r i x B, U
= (I + p B ) = I (mod p
).
That
e+1 t
e
e [
e+1
is,
6(p
) | p <5(p ). But obviously 6(p ) | 6(p
). We conclude, since p i s a
e+1
e
e
e
p r i m e , that dip
) is either <5(p ) or p dip ). In p a r t i c u l a r , dip )/ dip) is some
non-negative power of p. Similarly, a(p )/ctip) is some non-negative power of p.
Recalling that for any given modulus the ratio of the period to the r e s t r i c t e d period
divides 4 and that p is odd, it is immediate from the identity
1963]

g(p e )
<MR!)
^(P) ' a(p^
that a(pe)/a(p)
= S(p e )/5(p) and
Moreover,
B 0 (mod p).
suppose that
6{pe)/a(pe)
6(p
) ^
_6(p)
<5(p) ' a(p)
51R!)
= 6(p)/or(p).
<5(pe).
Then
U
e+1
;)
p+?
' = I + peB
with
p+1
).
Consequently, if
e(p)
<5(pe) = 6(p) for 1 < e _< e(p)
g _ Q /p \
6(p ) = p
<5(p) for e > e(p).
Finally, the existence of e(p) i s a s s u r e d
from a consideration of the alternative: if U ^

U
,
, pe+2.)
= TI +, p e + 1B^ ^/ i T (mod
That i s , if 6(p
) = p6(p ), then 5(p
) = pS(p
i s the l a r g e s t positive e such that 6(p e ) = 6(p), then
g
U ^
Hence s
T T P<5(p
and
33
= I (mod p e ) , e = 1,2, , then
- I, which is impossible. This completes the proof of (iv).

It is of i n t e r e s t to r e m a r k that a t e s t [17J with a digital computer has shown
that e(p) = 1 for all p r i m e s p l e s s than 10,000. However, the problem of
identi-
fying the exceptional p r i m e s p with e(p) > 1 r e m a i n s unsolved.

Finally, we prove p r o p e r t y (v). F o r every p r i m e p we define the r e s t r i c t e d
graph R(p) of U modulo p to consist of the p + 1 points P 0 = (0, l ) , P i = (1,1),
, P
= (p - 1, 1), P = (1, 0) together with the collection of all directed edges

P _L
P. P . T , where P. T
is the unique point which is linearly dependent upon the
m a t r i x product P.U.
(Contrast this with, for example, [ 5 ] . ) B y w a y of i l l u s t r a -
tion,
R(5) c o n s i s t s of the 1-cycle P 2 * P2 and the 5-cycle
-- P
- P 0 . In g e n e r a l , since this graph i s determined by a one-to-one c o r r e s p o n d -
ence, it follows that
R(p)
P 0 - * Pj - * P 3 - * P 4
consists of a collection of disjoint cycles.
(See for
example [ 1 4 J pp. 25-27.) F u r t h e r m o r e , it is c l e a r that P. belongs to a 1-cycle

(or in other words is a fixed point under the correspondence) if and only if P. is a
c h a r a c t e r i s t i c vector of U modulo p.
Moreover, suppose that P. belongs to an
a - c y c l e with a > 1. Since { P . , P . U} is a linearly independent set, it follows that

P. Ua = s P . (mod p) implies
since obviously a\a(p)}
= si ( m o d p ) ,
which means that a(p)\a.
Thus,
a = a(p). That i s , R(p) c o n s i s t s of a collection o f l - c y c l e s
and a ( p ) - c y c l e s . Consequently,
#(p)|(p + 1 - t),where t is the number o f l - c y c l e s
of R(p). But t is also the number of linearly independent characteristic vectors

of U modulo p ,
or equivalently the number of distinct roots modulo p of the m i n i -
mum polynomial X2 - X - 1 of U.
Since the discriminant of this quadratic is 5,
it
follows that t i s 0, 1, or 2 according as the Legendre symbol (5/p) is - 1 , 0, or 1.
34
[April
That i s , <*(p)|(p-(5/p)), which means that u p " ( 5 / / p ) = si and U p = s U ( 5 / / p ) ( m o d p),

for some integer s depending upon p.
m a t r i c e s in this l a s t congruence, t r U
But. since U
I implies U
and t r . t r = 1 (mod p).
Therefore
UP"<5/P)S
Now, considering the t r a c e of each of the

= 2s (mod 5) and t r U
= (5/p)s ( m o d p ^ 5 ) .
= tr Up - 2
= U - I (mod p), we have - t r U

U 5 =. 31 (mod 5). and
(mod p 5) ,
(5/p)I
which e s t a b l i s h e s p r o p e r t y (v). As a corollary we obtain the well-known congruence

u
= (5/p) (mod p).
law, we have
Also, it is of i n t e r e s t to add that, by the quadratic r e c i p r o c i t y
(5/p) = l i f p
= 5kl
and (5/p) = - 1 if p = 5 k 2 .
Thus, by u s e of the Fibonacci m a t r i x , we have established some of the p r i n c i pal a r i t h m e t i c a l p r o p e r t i e s of the sequence 0, 1, 1, 2,
. Although we may u s e
this m a t r i x to establish many other interesting p r o p e r t i e s and identities of the

Fibonacci n u m b e r s , we feel that the foregoing is sufficient to i l l u s t r a t e the application of this tool (at l e a s t as far as the a r i t h m e t i c a l p r o p e r t i e s a r e concerned).
However, we indicate in conclusion how the idea m a y b e readily adapted to the study
of m o r e general linear r e c u r r e n t sequences.
Specifically, let x 0 , x l s . . . , x , be the sequence of i n t e g e r s
satisfying
the linear r e c u r r e n c e
x , = a.]X
,
+ + a x
,
l
n+r
n + r - 1n
r n
for n > 0 where x 0 , , x __1 and a l 9 . . . , a
a r e given i n t e g e r s .
A study of
this linear r e c u r r e n t sequence may be made by m e a n s of the equation X

where X
= X0A ,
= (x , , x ,
') and
n'
' n+r-1;
0 a
0a
A =
r-1
0 l a v
Indeed, the a r i t h m e t i c a l p r o p e r t i e s of this sequence m a y b e investigated by a g e n e r alization of the methods suggested by this p r e s e n t paper.
positive integer such that {X 0 , , X
In p a r t i c u l a r , if m i s a
-} i s linearly independent modulo m , then
1963]
35
X ^ = sX 0 (mod m) if and only if A n = si (mod m). F u r t h e r m o r e , if (m, a ) = 1, then

the determinant of A is the unit (-1) ~ a
modulo m and the sequence of powers
of this m a t r i x reduced modulo m i s periodic.
That i s , under these assumptions,
the periodic p r o p e r t i e s of the sequence of integers reduced modulo m m a y b e identified with those of the sequence I, A, , A , reduced modulo m. For example,
we have that (ii) above is a special case of the equation
6(m) = a(m)/3(m) = (r, (3(m)) [ y(m)9a(m)
w h e r e r i s the o r d e r of the r e c u r r e n c e ,
of A modulo m , and a(m),
],
7(m) i s the exponent of the determinant
/3(m), and <5(m) a r e obvious extensions of the defini-
tions above.
REFERENCES
1.
E. T. Bell, Notes on r e c u r r i n g s e r i e s of the third o r d e r , Tohoku Math. J. 24

(1924) 168-184.
2.
J. L. B r e n n e r , Linear r e c u r r e n c e r e l a t i o n s , A m e r . Math. Monthly 61 (1954)

171-173.
3.
R. D. Carmichael, On sequences of integers defined by r e c u r r e n c e r e l a t i o n s ,

Quart. J. Math. 48 (1920) 343-372.
4.
, A simple principal of unification in the elementary theory

of n u m b e r s , A m e r . Math. Monthly 36 (1929) 132-143.
5.
R. H. Crowell,
Graphs of linear transformations over finite fields, J. Soc.
Indust. Appl. Math. 10 (1962) 103-112.

6.
L. E. Dickson, History of the theory of n u m b e r s , vol. I, Chelsea, New York,

1952.
7.
E. B. Dynkin and W. A. Uspenski, Mathematische Unterhaltungen II, Kleine

Erganzungsreihe XIV, Deutscher Verlag d e r Wissenschaften, Berlin, 1956.
8.
H. T. E n g s t r o m , On sequences defined by linear r e c u r r e n c e r e l a t i o n s , T r a n s .

A m e r . Math. Soc. 33 (1931)210-218.
9.
M. Hall, An i s o m o r p h i s m between linear r e c u r r i n g sequences and algebraic

r i n g s , T r a n s . A m e r . Math. Soc. 44 (1938) 196-218.
36
10.
[April 1963]
G. H. Hardy and E. M. Wright, An introduction to the theory of n u m b e r s ,

4th ed. , Oxford University P r e s s , London, I960, pp. 148-150.
11.
D. H. L e h m e r , An extended theory of L u c a s ' f u n c t i o n s , Ann. of Math. (2) 31

(1930) 419-448.
12.
N. S. Mendelsohn. Congruence relationships for integral r e c u r r e n c e s ,
Can.
Math. Bull. 5 (1962) 281-284.

13.
E. P . Miles, J r . . Generalized Fibonacci n u m b e r s and associated
matrices,
A m e r . Math. Monthly 67 (1960) 745-752.

14.
O. O r e , Theory of g r a p h s ,
Amer.
Math.
Soc.
Colloq.
Publ. , v o l .
38,
P r o v i d e n c e , 1962.
15.
D. W. Robinson, A note on l i n e a r r e c u r r e n t sequences modulo m,
submitted
for publication in the A m e r . Math. Monthly.

16.
R. A. Rosenbaum, An application of m a t r i c e s to l i n e a r r e c u r s i o n
relations,
A m e r . Math Monthly 66 (1959) 792-793.

17.
D. D. Wall, Fibonacci s e r i e s modulo m , A m e r .
Math, Monthly 67 (1960)
525-532.
18.
M. Ward, The algebra of r e c u r r i n g s e r i e s , Ann. of Math. (2) 32 (1931) 1-9.
19.
, The c h a r a c t e r i s t i c number of a sequence of i n t e g e r s satisfying a

l i n e a r r e c u r s i o n r e l a t i o n , T r a n s . A m e r . Math. Soc. 33 (1931) 153-165.
20.
, The a r i t h m e t i c a l theory of linear r e c u r r i n g s e r i e s , T r a n s .
Amer.
Math. Soc. 35 (1933) 600-628.

21.
N. Z i e r l e r ,
Linear r e c u r r i n g sequences,
J.
Soc.
Indust.
Appl.
Math. 7
(1959) 31-48.
Brigham Young University

and
California Institute of Technology
/AVMVVYXW^^
REQUEST
The Fibonacci Bibliographical R e s e a r c h Center d e s i r e s that any r e a d e r finding

a Fibonacci r e f e r e n c e send a c a r d giving the r e f e r e n c e and a b r i e f description of the
contents.
P l e a s e forward all such information to:

Fibonacci Bibliographical R e s e a r c h Center,
Mathematics Department,
San J o s e State College,
San J o s e , Calif.
THE RELATION OF THE PERIOD MODULO

TO THE RANK OF APPARITION OF m IN THE FIBONACCI SEQUENCE
JOHN VINSON, AEROJET-GENERAL CORPORATION, SACRAMENTO, CALIF.
The Fibonacci sequence i s defined by the r e c u r r e n c e relation,

(1)
n = 0, 1, 2,
and the initial values u 0 = 0 and VLt = 1. Lucas [ 2 , pp. 297-301] h a s shown that
every i n t e g e r , m 5 divides some m e m b e r of the sequence, and also that the sequence
i s periodic modulo m for every m. By this we mean t h e r e i s an integer, k, such
that
(2)
uk+
Definition.
tive integer,
= u (mod m) ,
n = 0, 1, 2, *** .
The period modulo m, denoted by s ( m ) , i s the s m a l l e s t p o s i -
k, for which the s y s t e m (2) i s satisfied.
Definition.
The rank of apparation of m , denoted by f(m), i s the s m a l l e s t
positive integer, k, for which u, = 0 (mod m).

Wall [3] has shown that
(3)
u n = 0 ( m o d m ) iff
f(m)|n.
In 1p a r t i c u l a r , since u , . = uun = 0 (mod m); we have

'
s(m)
(4)
f(m) |s(m) .
Definition.
We define a function t(m) by the equation f(m)t(m) = s(m).
We note that t(m) i s an integer for all m. The purpose of this paper i s to
give c r i t e r i a for the evaluation of t(m).
Now we give some r e s u l t s which will be needed l a t e r .
(5)
v
'
2
n
u
= u u
+ (-1)
n-1
n n-2 v ;
This can be proved by induction, using the r e c u r r e n c e relation (1).

This paper was p a r t of a t h e s i s submitted in 1961 to Oregon State University in p a r t i a l
fulfillment of the r e q u i r e m e n t s for the degree of m a s t e r of A r t s .
37
THE RELATION OF THE PERIOD MODULO m TO THE RANK
38
/^
(6)
<?n - (3n
,
1 + N/5
, .
= -f1
- , w
here # = +
5 V5 and, >
w n o T O ru
n
a - j8
[April
1 - \/5
This is the well-known "Binet f o r m u l a . " It gives a natural extension of the Fibonacci
sequence to negative values of n.
By using the relation a (3 = (-1) , we find
(7)
_n
1
(* -)
n+1
u.S i
F r o m this we see that the r e c u r r e n c e (1) holds for the extended sequence.
By solving the s y s t e m
ak - /3 k = (a - (3) u k
a ak - /3 . /3 k = (or - 0) u k + 1
k
for a
k
and ft , we obtain
or
= u.k+1
/3u k = (1 - ) u k + u k _ x = , u
k +
uk_x
and
/3 k =, u k + 1 - u k = (1 - a) u k + u k _ x = u k + u k - 1
Then
{a
- 0 , ^ = nk+r - ^
= (.u,+
V l
, V - (ûk +
By expanding and recombining we get (for n > 0)
%k+r = X I .UkVlVj
Now if we set k = f(m). we find
(8)
v ;
u r, x, = Up. . , u (mod m)
nf(m)+r
f(m)-.l r
V l )V
1963 ]
OF APPARITION OF m IN THE FIBONACCI SEQUENCE
39
We note that this is valid for negative as well as non-negative i n t e g e r s ,
= u
r.
L e m m a 1. t(m) is the exponent to which u f , . - belongs (mod m).

n
Proof: Suppose u f / . ., = 1 x(mod m).
Then from v(8)
we have u -, x,
;
;
-^
f(m)-l
nf(m)+r
(mod m) for all r . It follows from the definition of s(m) that s(m) _< nf(m)
and thus u f ,
,_-. = 1 (mod m) implies t(m) = s(m)/f(m) <_ ne
Now we s e t r = 1 and n = t(m) in (8) to obtain

t(m)
uf
'
f(m)-l
Thus
uf,
-, ,
-, x
= 1 1 , / .-, v, -, = u , ,, .. = u t = 1 (mod m) .
x
v
;
t(m)f(m)+l
s(m)+l
t(m) i s the s m a l l e s t positive n for which u nf , , = 1 (mod m),
that i s ,
._ 1 belongs to t(m) (mod m).

T h e o r e m 1.
F o r m > 2 we have
i) t(m) = 1 or 2 if f(m) i s even, and

ii) t(m) = 4 if f(m) is odd.
A l s o , t(l) = t(2) = 1.
Conversely,
t(m) = 4 implies f(m) is odd, t(m) = 2 i m -
plies f(m) i s even, and t(m) = 1 implies f(m) is even or m = 1 or 2.

Proof.
The c a s e s
m = 1 and m = 2 a r e easily verified.
Now suppose
m > 2 and s e t n = f(m) in (5) to get

U
f(m)-1 -
f(m)uf(m)-2
("1)f(m) -
( " 1 ) f ( m ) (modm)
If f(m)
i s even we have u?, x _. = 1 (mod m ) , and i) follows from L e m m a 1.
v
'
t(m)-l
'
If f(m)
i s odd we have uS, x -. = - 1 (mod m )/ ?, and since m > 2 , u | , . 1
v
'
f(m)-l
' l(m)-l
^ 1 (mod m). This implies u f , ,_-. ^ 1 (mod m) and then
3
Up, x *- = u_p. . _. u. , _, = -Up, , ., r 1 v (mod m).

f(m)-l
f(m)-l f(m)-l
f(m)-l
'
Finally,
u4
= (u2
)2 = (-1) 2 = 1 (mod m) and, by L e m m a 1,
t(m) = 4.
The c o n v e r s e follows from the fact that the c a s e s in the direct statement of
the t h e o r e m a r e all inclusive.
T h e o r e m 2. Let p be an odd p r i m e and let e be any positive integer. Then
i)
ii)
t(p e ) = 4 if 2 | f(p),
t(p e ) = 1 if 2 | f(p) but 4 1 f(p),
iii)
t(p e ) = 2 if 4 | f(p), and
iv)
t(2 e ) = 2 for e 2 3 and t(2) = t(2 2 ) = 1 .
40
[April
Conversely, if q r e p r e s e n t s any p r i m e , then t(q ) = 4 implies f(q) is odd,

e
a
e
t(q ) = 2 i m p l i e s 4|f(q) or q = 2 and e _>. 3, and t(q ) = 1 implies 2jf(q) but
4 | f(q) or q e = 2 or 4.
Proof. Wall [3, p. 527] has shown that if p n + 1 \ u
then f(p n + 1 )= pf(p n ).
It follows by induction that f(p ) = p f(p), where k is some non-negative integer.

e
We emphasize that f(p ) and f(p) a r e divisible by the s a m e power of 2, since this
fact i s used s e v e r a l t i m e s in the sequel without further explicit r e f e r e n c e .
e
e
In c a s e i), f(p ) i s odd and the r e s u l t is given by setting m = p in T h e o r e m 1.
e
e
In Gc a s e s ii) a n d i i i ) , f(p ) i s even and we may set m = p , n = 1, and
r - | f ( p ) in (8) to get
UI-P/ ep\
|f(p )
2
4T/ P\e 1
f(p )-l
H o,
-|f(pe) x
v (modP
pe)
'
which, in view of (7), i s the s a m e as
f(pe)-1
if(pe)
2
Now |f(p e ). = p-p^(p),
where
l f ( P e/ ), i+u1
. (-D2iVF
/Hr^
<mod
if(pe)
k i s s o m e non-negative integer, and we see that
f(p) I
if(pe). Then from (3) we have p | u l f , e so that we may divide the above
2
2XvP )
congruence by u l f , y We get
i,
uA . ^ ^
f(p6)Now in c a s e ii),
.= ( - l ) 2 f ( P e ) + 1
(modpe)
~f(p) i s odd and so is H{pe)f

2
<*
and the l a s t congruence gives

e
u f / ^x - = 1 (mod pe) and t h u s , by L e m m a 1, t(p ) = 1 .

ilp^J-l
In c a s e iii) the congruence b e c o m e s
u
f(pe)-i
-1
( m o d P e ) '.
since
Hip)
a r e both even.
and
if(pe>
Then
V)-!
and by L e m m a 1 again,
t(p e ) = 2.
l (modpe)
1963 J

In c a s e iv) we can easily verify t(2) = t(2 2 ) = 1.
41
That t(2 e ) = 2 for e ^ 3
follows from r e s u l t s given by Carmichael [1, p. 42] and Wall [ 3 , p. 527].
These
results are, respectively:

A.
Let q be any p r i m e and let r be any positive integer such that (q, r) = 1.
, X+l I
,-,
X+a
If q u and q
/ u , then q
u
_ and q X+a+1 / u
_ except when q = 2
n
H
L
I n
A
I nrqa
/] n r q a
and X = 1.
B. Let q be any p r i m e and let X be the l a r g e s t integer such that s(q )
e-X
= s(q). Then s(q e ) = q
s(q) for e > \ .
xl
The hypotheses of A. a r e satisfied by q = 2, X = 3, and n = f(2 3 ), and

->3+a| u
.
a|
T4-.cn
e
/OV4.U 4. .c/o3+ax
we find that 2
I v.f/o3\ ^ff 20 T|k- It follows from (3) that f(2
) m u s t be a m u l t i '\f(23)
pie of f(2 3 ),
for e ^ 3.
hence f ( 2 3 + a ) = 2 a f(2 3 ).
Since f(23) = 2f(2) we have f(2 e ) = 2 e " 2 f(2)
Now set q = 2 and X = 1 in B.
We get s(2 e ) = 2 e _ 1 s ( 2 ) . Thus for
e ^ 3 we have
t(2e} =
3(2!) = f^m
f(2 e )
2e
= 2
f(2)
The converse follows from the fact that the c a s e s in the d i r e c t statement
the t h e o r e m a r e all inclusive.
of
This completes the proof,
Now we give a l e m m a which is needed in the proof of the next theorem.

L e m m a 2.
If m has the p r i m e factorization
m = qt
i) s(m) = 1. c. m.
lî^
ii)
f(m) = 1. c. m.
l<i^n
. . . qr
n
then
( s ( q . 1 ) } , and
(ffq1)}
1
Wall has given i).

relatively p r i m e ,
q2
m|u,
The proof of ii) is as follows: Since the q. 1 a r e p a i r w i s e

i s equivalent to q. i u, (i = 1, 2, 0* , n), which, by (3),
i s equivalent to f (q. i ) k (i = 1, 2, , n).
The s m a l l e s t .positive k which s a t i s -
fies these conditions i s

k = 1. c. m.
l<i<n
which, according to the definition of f(m),
(ffa?1)}
1
gives the d e s i r e d r e s u l t .
42
[April
T h e o r e m 3. We have
i)
ii)
t(m) = 4 if m > 2 and f(m) i s odd.

t(m) = 1 if 8/f m and 2 | f (p) but 4 f f (p) for every odd p r i m e ,
divides m ,
iii)
p,
which
and
t(m) = 2 for all other
m.
Proof: F r o m what has already been given in T h e o r e m 1, we see that it suffices to show that the conditions given h e r e in ii) a r e both n e c e s s a r y and sufficient
for t(m) = 1.
Let m have the p r i m e factorization
f(q^) = 2 V
w h e r e the K. a r e odd i n t e g e r s .
6
at
t(q
Then
s^1)
(i = 1, 2, -
m = q11q22
, n)
q n
m
and set
By T h e o r e m 1, we may set
) = 2
(i = 1, 2, , n)
= f (q?i )t (q?i) = 2
where 6. = 0, 1, or 2.
K. (i = 1, 2, , n).
F r o m L e m m a 2 we
have, where K i s an odd integer,

, x
1. c. m. r / # i \ i
0 max(7i + &\ )xr
w
s(m) = - , ' .
isfa-Mr = 2
i
i 'K,
L
J
'
l<i<n
l
.
1. c. m.
p,^.
,
/
0maxîT/
f(m) = , .
ffa- 1 )
= 2
iK, and
x ;
VH ;
5
li<n
i
4., ^
i X/-P/ x
0max(7-+<5 ) - m a x 7 x
t(m) = s(m)/f(m) = 2
i i'
i
Now suppose t(m) = 1.
Then max (7. + 5.) = m a x / . .
7, <L 7i
+ 5,
Let y, = m a x / . . We have
< m a x (7 + 6.) = max 7, = 7u ,
6
and thus <5k = 0 and t ( q Q'
Kkv) = 2~ k = 1. It follows from T h e o r e m 2 that 4 ( f ( q k k ) ,
that i s , that y^ < 1.
Then for all i,

6. < max (7. + 6.) = max 7. = 7 ,
Furthermore,
<1 .
6. = 1 is i m p o s s i b l e , for 5. = 1 is the s a m e as t(q. i ) = 2 which
i m p l i e s , by T h e o r e m 2, that
2 f(q. *)
and thus 7. > 1.
Then we would have
1963]
/. 1 + t
> 2j3 which

is c o n t r a r y to max (:>'
. +.) < 1.
z.
" I I 7
^1
id t CX'
and t ( q . 1 ) = 2 = l s
in ii).
43
Thus for all
i,
61
'
which, by T h e o r e m 2, is equivalent to the conditions given
Now suppose, conversely, that the conditions given in ii) a r e satisfied, which,
as we have just seen, is equivalent to the condition t ( q . x ) = 1 for all i.
s(q?i) = f(qj'i) t ( q ^ )
Then
for all i.
Then L e m m a 2 gives
. ,
S(m
1. c. m.
. a\ . -,
1. c . m .
' = l<i<_n { S <V>} = l<i<n
rp/îxi
^ t(q i
'' =
P.
f(m)
and thus t(m) = s(m)/f(m) = 1.

Our l a s t t h e o r e m is of r a t h e r different c h a r a c t e r . Once again, we need a p r e liminary l e m m a .
L e m m a 3.
Let p be an odd p r i m e .
i)
f(P) | (p - 1)
if
Then
p = . 1 (mod 10) ,
ii)
f(p) | (p + 1)
if
p = 3 (mod 10) ,
iii)
s(p) | (p - 1)
if
p = i l (mod 10) , and
iv)
s ( p ) | ' ( p + 1) but s(p) | 2(p+ 1) if p SE+ 3 (mod 10).
Lucas [2, p. 297] gave the following r e s u l t :

P
| U p - l if
" ~
^mod
10
and
P I UD+1
if
P = - 3 (mod 10).
We get i) and ii) by applying (3) to this result, Wall [3. p. 528] has given iii) and iv).
T h e o r e m 4.
i)
ii)
Let p be an odd p r i m e and let e be any positive integer. Then
t(p e ) = 1 if p = 11 or 19 (mod 20),

t(p e ) = 2 if p = 3 or 7 (mod 20),
iii)
t(p e ) = 4 if p = 13 or 17 (mod 20), and
iv)
t(p e ) 4 2 if p = 21 or 29 (mod 40).
Proof: T h e o r e m 2 shows that t(p ) is independent of the value of e,

i s sufficient to consider e = 1 throughout the proof.
hence
44
[April
so that by
Fermatrs
belongs to t(p) (mod p), it follows that t(p) j(p-l).
Now if
If follows from the definition of ff(p) that p j( u f , .

theorem,
u
Then, since u f . .
f(p)-l
X (mod
p)
'
p = 3 (mod 4) we have 4 \ (p - 1) and thus t(p) ^ 4.

i)
H e r e p = 3 (mod 4)
4 | f(p).
so t(p) ^ 4.
Suppose t(p) = 2.
Now p = 1 (mod 10) and, by L e m m a 3 i ) ,
4 | (p - 1).
Then, by T h e o r e m 2,
f(p)
(p - 1)
But this is impossible when p = 3 (mod 4), hence
and
thus
t(p) ^ 2 and we
m u s t have t(p) = 1.
ii) Again p = 3 (mod 4) and t(p) ^ 4.
Also p = . 3 (mod 10) and it follows from
L e m m a 3 that s(p) ^ f(p) and t(p) = s(p)/f(p) ^ 1.
Hence t(p) = 2.
iii) We have just seen that t(p) ^ 1 when p = 3 (mod 10), which i s h e r e the c a s e .
Also,
f(p) J (p + 1).
Now p = 1 (mod 4) so that
and it follows from T h e o r e m 2 that t(p) ^ 2.

iv) Suppose t(p) = 2.
= t(p)f(p) = 2f(p)).
t(p) = 2 i m p l i e s
Then by T h e o r e m 2,
Furthermore,
8( (p - 1).
4 j (p + 1)
and thus
f(p),
Hence t(p) = 4.
4 | f(p) and thus
s(p) | (p - 1)
since
8 js(p)
(since
s(p)
p = 1 (mod 10). Then
But we have p - 1 = 20 or 28 (mod 40)
p - 1 = 4 (mod 8), so that 81 (p - 1) i s impossible.
which
gives
Hence t(p) ^ 2.
We naturally ask if a ^ t h i n g m o r e can be said about t(p ) for

29 (mod 40).
4 |
p = 1, 9, 2 1 ,
The following examples show that the t h e o r e m is "complete' 1 :
p =
1 (mod 40)
p =
9 (mod 40)
p = 21 (mod 40)
p = 29 (mod 40)
t(521) = 1,
t(809) = 1,
t(101) = 1,
t(29) = 1,
t(41) = 2,
t(761) = 4
t(409) = 2,
t(89) = 4.
t(61) = 4.
t(109) = 4.
Now we might ask whether t h e r e is a number, m, for which t(p ) is

d e t e r m i n e d by the modulo m r e s i d u e c l a s s to which
this question i s not known.
p belongs.
always
The answer to
We note that the p r i n c i p l e s upon which the proof of
T h e o r e m 4 is b a s e d a r e not applicable to other moduli.
1963]
45
REFERENCES
1.
n
n
R. D. C a r m i c h a e l , On the n u m e r i c a l factors of the arithmetic forms a (3 ,
Ann. of Math. 15 (1913-1914), 30-70.
2.
E. Lucas, Theorie des fonctions numeriques simplement periodiques,
3.
J. Math. 1 (1878), 184-240, 289-321.

D. D. Wall, Fibonacci s e r i e s modulo m, A m e r . Math. Monthly 67 (I960), 525-532.
^wywyYwvwvvvv^^
Amer.
SPECIAL NOTICE
The Fibonacci Association has on hand 14 copies of Dov J a r d e n ,
Recurrent
Sequences, Riveon L e m a t e m a t i k a , J e r u s a l e m , I s r a e l . This i s a collection of p a p e r s

on Fibonacci and Lucas n u m b e r s with extensive tables of factors extending to the
385th Fibonacci and Lucas n u m b e r s .
The volume s e l l s for $5.00 and is an e x c e l -
lent investment. Check or money o r d e r should be sent to V e r n e r Hoggatt at San J o s e

State College, San J o s e , Calif.
REFERENCES TO THE QUARTERLY

Martin G a r d n e r , Editor, Mathematics G a m e s , Scientific A m e r i c a n ,
J u n e , 1963
(Column devoted this i s s u e to the helix. )

A Review of The Fibonacci Quarterly will appear in the Feb.
1963 i s s u e of the
Recreational Mathematics Magazine.

FIBONACCI NEWS
B r o t h e r U. Alfred r e p o r t s that he is c u r r e n t l y offering a one unit c o u r s e
on
Fibonacci N u m b e r s at St. M a r y ' s College,

Murray B e r g ,
Oakland, Calif. , r e p o r t s that he has computed phi to
2300 decimals by dividing F l l o 0 4 by F 1 1 0 0 3 on a computer.
some
Any inquiries should
be a d d r e s s e d to the editor.
Charles R. Wall, Ft. Worth, T e x a s , r e p o r t s that he is working on his m a s t e r ' s
t h e s i s in the a r e a of Fibonacci r e l a t e d topics.
SORTING ON THE B-5000 - - Technical Bulletin 5000-21004P S e p t . , 1961,
Burroughs Corporation, Detroit 32, Michigan.
This contains in Section 3 the use of Fibonacci n u m b e r s in the merging of information using t h r e e tape units instead of the usual four thus effecting considerable
efficiency.
(This was brought to our attention by Luanne Angle my e r and the pamphlet
was sent to us by Ed Olson of the San J o s e office.)
HAVE YOU SEEN?
46
[April
S. L. Basin, Generalized Fibonacci Numbers and Squared Rectangles, A m e r i c a n

Mathematical Monthly, pp. 372-379, April, 1963.
J . A. H. Hunter and J. S. Madachy s Mathematical D i v e r s i o n s , D. Van Nostrand
Company, I n c . , P r i n c e t o n , New J e r s e y , 1963.
S. K. Stein, The Intersection of Fibonacci Sequences,
Michigan Math.
Journal,
9 (1962), Dec. , No, 4 a pp. 399-402. (Correction)

L. Carlitz, Generating Functions for P o w e r s of Certain Sequences of N u m b e r s ,
Duke Math. J o u r n a l , Vol. 29(1962), D e c , No. 4, pp. 521-538. (Correction)
V. E. Hoggatt, J r . and S. L. Basin, The F i r s t 571 Fibonacci N u m b e r s ,
Recrea-
tional Mathematics Magazine, Oct. , 1962, pp. 1 9 - 3 1 .

A. F. Horadam, Complex Fibonacci Numbers and Fibonacci Quaternions,
The
A m e r i c a n Mathematical Monthly, M a r . , 1963, pp. 289-291.

S. L. Basin, The Appearance of Fibonacci Numbers and the Q - m a t r i x in E l e c t r i cal Network Theory, Math. Magazine, Vol. 36, No. 2, March 1963, pp. 84-97.
J. Browkin and A. Schinzel, On the Equation 2
- D = y 2 , Bulletin de l'Academie
Polonaise des Sciences, Serie des sci. math., a s t r , et phys. Vol. VIII. No. 5,
I960, pp. 311-318.
Georges Browkin and Andre Schnizel, Sur les n o m b r e s
triangulares,
de Mersenne qui
Comptes rendues des s e a n c e s de l'Academie
des
sont
sciences,
t. 242, pp. 1780-1781, seance du 4 Avril 1956.

C. D. Olds,
Continued F r a c t i o n s ,
Random House (New Mathematical
Library
S e r i e s - - p a r t of the Monograph P r o j e c t of SMSG) 1963.

This is an excellent understandable t r e a t m e n t of the subject at a reasonable
level with many i n t e r e s t i n g topics for those devoted to the study of i n t e g e r s
with special p r o p e r t i e s .
L. Zippin, Uses of Infinity7, Random House (New Mathematical L i b r a r y S e r i e s , 1962.)
This has no index which, makes the Fibonacci topics h a r d e r to find but t h e r e
a r e s e v e r a l i n t e r e s t i n g comments t h e r e .
Mannis Charosh, P r o b l e m Department, Mathematics Student J o u r n a l , May, 1963.
In the editorial comment following the solution of P r o b l e m 187, t h e r e is a
little generalized r e s u l t s i m i l a r to problem B-2 of the E l e m e n t a r y P r o b l e m s
and Solutions section of the Fibonacci Q u a r t e r l y , Feb. , 1963.
A. Rotkiewicz, On Lucas Numbers with Two Intrinsic P r i m e D i v i s o r s ,
Bulletin
de l'Academie Polonaise des Sciences, Serie des sci. math. , a s t r . et phys.

Vol. X. , No. 5, 1962.
FIBONACCI MATRICES AND LAMBDA FUNCTIONS

MARJORIE BICKNELL AND VERNER E.
HOGGATT, J R . ,
SAN JOSE STATE C O L L E G E ,
SAN J O S E ,
When we speak of a Fibonacci m a t r i x , we shall have in mind m a t r i c e s
CALIF.
which
contain m e m b e r s of the Fibonacci sequence a s e l e m e n t s . An example of a Fibonacci

m a t r i x i s the Q m a t r i x a s defined by King in [ l ] , pp. 11-27, w h e r e
-I 1
I 1 /
The determinant of Q i s - 1 , written det Q = - 1 . F r o m a t h e o r e m in m a t r i x theory 3

det Q n = (det Q) n = ( - l ) n
By mathematical induction, it can be shown that
/ n+1
Qn =
\F
^. n
n\
!
F ,/
n-1
so that we have the familiar Fibonacci identity
- F 2 = (-1) 11
1 F
v ;
n+1 n - 1
n
The lambda function of a m a t r i x was studied extensively in [2] by Fenton

S. Stancliff, who was a professional musician.
tion
Stancliff defined the lambda func-
A(M) of a m a t r i x M a s the change in the value of the determinant of M when
the number one i s added to each element of M. If we define

m a t r i x formed from
have the identity
(1
>
(M + k) to be that
M by adding any given number k to each element of M5 we
det (M + k) = det M + k X_(M).

47
48
F o r an example, the determinant
\{Q ) is given by
+ 1
n+1
MQ )
(F
v
[April
- detQ"
+ 1
F 2 ); + v(F , + F : 1 - 2 F ); - det Q n
n
n-1
n+1
n
F
n+1
, + 1
n-1
n-1
n-3
which follows by use of Fibonacci identities.
Now if we add
k to each element
of Q , the r e s u l t i n g determinant is
F
F
- + k
n+1
n
+ k
+ k
det Q
+ k F
n-3
- + k
n-1
However, t h e r e a r e m o r e convenient ways to evaluate the lambda function.

F o r simplicity, we consider only 3 x 3 m a t r i c e s .
THEOREM.
F o r the given general 3 x 3
of the e x p r e s s i o n s (2) or (3).
For
a
d
kg
(2)
b
e
h
c^
f
J I
a + e - ( b + d )
b + f - ( c
+ e)
d + h - (g + e)
e + j - (h + f)
X(M) =
1
(3)
m a t r i x M, X(M) i s e x p r e s s e d by either
MM) =
Proof: This is made by d i r e c t evaluation and a simple e x e r c i s e in algebra.
1963 ]
FIBONACCI" MATRICES AND LAMBDA FUNCTIONS

An application of the lambda function i s
49
in the evaluation of d e t e r m i n a n t s .
Whenever t h e r e is an obvious value of k such that
det (M + k) is easy to evaluate.
we can use equation (1) advantageously. To i l l u s t r a t e this fact, consider the m a t r i x
000
998
554
990
988
554
675
553
554
We notice that, if we add k = -554 to each element of M, then det (M.+ k) = 0

since every element in the t h i r d column will be z e r o .
10
-120
435
F r o m (2) we compute
1200.
X(M)
and from (1) we find that

0 - det M + (-554) (1200) ,
so that det M = (554) (1200).

R e a d e r s who enjoy m a t h e m a t i c a l c u r i o s i t i e s can c r e a t e d e t e r m i n a n t s which
a r e not changed in value when any given number k is added to each element, by
writing any m a t r i x D such that X (D) = 0.
LEMMA: If two rows (or columns) of a m a t r i x D have a constant difference b e tween corresponding e l e m e n t s , then X(D) = 0.
Proof: Evaluate
X(D) directly, by (2) or (3).
F o r example, we w r i t e the m a t r i x D, where corresponding elements in the first

and second rows differ by 4, such that
det D
1 + k
2+k
3 + k
5 + k
6+k
7 + k
4 + k
9+k
8 + k
Now, we consider other Fibonacci m a t r i c e s .

a Fibonacci m a t r i x U such that det U
Now
24
Suppose that we want to write
50
[ April

a
0I
d|
.
= abd.
We can write F
= F ^ F
= F ^ F
= F 2 F 2 F for any n, and for some n we
n
n
n
n
will also have other Fibonacci factorizations. Hence, F = det U for
p.\
where F 0 = 0.
If we choose m = k = 3 and p = 2, we find that
X(U) = 0. If we
choose m = 1 or 2, k = 1 or 2. and let p be an a r b i t r a r y integer, then
\(U) = F .
A m o r e elegant way to write such a m a t r i x was suggested by Ginsburg in [3 ],

who showed that if
then det B = n,
a =
2p'
c = b = F
and
f-
2p+3 J
where
B =
Letting n = F
F2p+2,
2p+l'
, we can write F
= det U, where
2p
'2p+l
F
2p+l
F
2p+2
F
2p+2
m
F
2p+3
Using equation (3) we have

m
X(U)
= 0 + 0 + 1/F
(det U) = 1
1963 J
51
If we let k F ^ ^, from (1) we see that

det (U + F
J
m-r
- F
m
+ (F
J (1) = F
v
m-l; v ;
m+1 *
Notice the possibilities for finding Fibonacci identities using the lambda function and evaluation of d e t e r m i n a n t s . As a brief example, we let k = F and conn
n
s i d e r det (Q + F ), which gives us
n+1
F
+ F
+ F
= det Q n + F n X ( Q n )
, + F
n-1
n
or
2 F
n+2
2 F
-1);
n
+ F n F n - 03
n+1
so that
4 F 2 = F . F _ - F F _ + ( - l;) n + 1
n
n+2 n+1
n n-3
As a final example of a Fibonacci m a t r i x , we take the m a t r i x R,

/0
R =
\l
which has been c o n s i d e r e d by Brennan [ 4 ] .

It can be shown by mathematical induction that
/
R11 =
2F
, F
n-1 n
F \
- F
.F
n+1
n-1 n
F
, F
n-1 n
,i
n+1
2F
n+1
n
n+1
given by
52
[April 1963
The r e a d e r may verify by equation (2) and by Fibonacci identities that
\(R
F2
+ F2
- 4F
F
n-1
n+1
n-1 n
1J
3F
2F
_, F - F 2 - F 2
+ F F ,
n-1 n
n
n+1
n n+1
r
2n-3
n-2
-F
H e r e we see that the value of
n-20
n - n1
, F + 2 F F
n-1
n n+1
_ F
- F
n+1
2F2
- 3 F F in - F F ,
n+1
n n+1
n n-1
2n-2
+ v( - l ); n
= <-1>X-l-Fn-3Fn-2>-
(R ) i s the center element of R
multiplied by
(-D .
REFERENCES
1.
Charles H. King, Some P r o p e r t i e s of the Fibonacci N u m b e r s , ( M a s t e r ' s Thesis)

San J o s e State College, J u n e , 1960.
2.
F r o m the unpublished notes of Fenton S. Stancliff.
3.
Jukethiel Ginsburg, " D e t e r m i n a n t s of a Given V a l u e , " Scripta Mathematica,

Vol, 18, i s s u e s 3-4, S e p t . - D e c . , 1952, p. 219.
4.
F r o m the unpublished notes of T e r r y Brennan.
NOTICE TO ALL SUBSCRIBERS!!!

P l e a s e notify the Managing Editor AT ONCE of any a d d r e s s change. The P o s t Office
Department,
r a t h e r than forwarding magazines mailed t h i r d c l a s s ,
directly to the d e a d - l e t t e r office.
sends them
Unless the a d d r e s s e e specifically r e q u e s t s the
| Fibonacci Quarterly be forwarded at first c l a s s r a t e s to the new a d d r e s s , he will not

j r e c e i v e it. (This will usually cost about 30 cents for first c l a s s postage.) If possible,
; please notify us AT LEAST THREE WEEKS PRIOR to publication dates: F e b r u a r y 15,
i
! April 15, October 15, and December 15.
ADVANCED PROBLEMS AND SOLUTIONS

EDITED BY VERNER E. HOGGATT, JR., SAN JOSE STATE COLLEGE
Send all communications concerning Advanced P r o b l e m s
and Solutions to
V e r n e r E, Hoggatt, J r . , Mathematics Department, San J o s e State College,

J o s e , Calif.
San
This d e p a r t m e n t especially welcomes p r o b l e m s believed to be new or
extending old r e s u l t s .
P r o p o s e r s should submit solutions or other information that
will a s s i s t the editor. To facilitate t h e i r consideration, solutions should be submitted

on s e p a r a t e signed sheets within two months after publication of the p r o b l e m s .
H-9
Proposed by Olga Taussky, California
Institute
of Technology, Pasadena,
California
Find the n u m b e r s a n, r 5, where n > 0 and

to r a r e i n t e g e r s>, f o r
relations
which
a.
+ a
- + a
= a .,
n, r
n.r-1
n, r - 2
n+l.r
and
aA
0.r
= <5
. o.r
r ^ 0
,1
r =0
hold.
H-10
Proposed by R. L. Graham, Bell Telephone Laboratories,

OO
Show that
-i
Z_I F
n=l
H-ll
Murray Hill,
x-1
"
(_i\n+l
Z JZ _ Fi F F , F ^
d +
n=l
n + 1
n + 2
Proposed by John L. Brown, Jr., Ordnance Research Laboratory,

State University, University Park, Penna.
Find the function whose formal F o u r i e r s e r i e s is

OO
-5>
sin nx
f(x) =
where F
Sew Jersey
30
n=l
is the nth Fibonacci number.
53
The Pennsylvania
the
54
H-12
ADVANCED P R O B L E M S AND SOLUTIONS

Proposed by D. E. Thoro, San Jose State College, San Jose,
[April
California
Find a formula for the nth t e r m in the sequence:

1. 3 , 4 , 6 , 8, 9 , 1 1 r 1 2 , 1 4 , 1 6 , 1 7 , 1 9 , 2 1 , 2 2 , 2 4 , 2 5 ,
H-13
Proposed by H. W. Gould, West Virginia University, Morgantown, W. Va., and

Verner E. Hoggatt, Jr., San Jose State College, San Jose, Calif.
Show that
F
F -J *JF rF
TT = \ 7 \ T?
n =j = 0S(0 î i n+j-rk
n
r
H - 1 4 Proposed by David Zeitlin,

College, Alaska.
Minneapolis, Minnesota, and F. D. Parker, University
of Alaska,
P r o v e the Fibonacci identity

F 3 , , - 3 F 3 _ - 6 F^
+ 3 F3
+ F3 i>:-4
n+3
n+2
n+1
n
H-15
Proposed by Malcolm H. Tallman, Brooklyn, New York

Do t h e r e e x i s t i n t e g e r s
N l5 N 2 , and N3 for which the following e x p r e s s i o n s
cannot equal other Fibonacci n u m b e r s ?

(i)
F3 - F2 F
- F3
n
n m
m
m, n > N j
(ii)
F3 + F 2 F
+ F F2
n
n m
n m
m, n -
N2
m, n >
N3
F2 - 3 F 3
n
m
Proposed by H. W. Could, West Virginia University,
(iii)
H-16
Morgantown, West Virginia
Define the o r d i n a r y H e r m i t e polynomials by H
(i)
H (X/2)
Y n
= (-1)
n x
n _x
e D (e
).
n=0
Show that:
(")
Y n<
x/2)
n! Fr,n
x
n=0
OO
(iii)
where
and
V"1
xn
H
( x / 2 )
2i n
-HT
n=0
-x 2
L
2 G
'
a r e the nth Fibonacci and nth Lucas n u m b e r s ,
respectively.
1963 ]
ADVANCED P R O B L E M S AND SOLUTIONS
H - 1 7 Proposed by Brother U. Alfred, St. Mary's College,
55
Calif.
n
Sum:
k3Fk
k=l
H-18 Proposed by R. G. Buschman, Oregon State University,
Gorvallis,
Ore.
"Symbolic r e l a t i o n s " a r e s o m e t i m e s used to e x p r e s s identities.

if F
and L
For example,
denote, r e s p e c t i v e l y 2 Fibonacci and Lucas numbers^ then

(1
L)n i
L2n,
(1
F)n # F 2 n
a r e known i d e n t i t i e s , where denotes that the exponents on the symbols a r e to be

lowered to s u b s c r i p t s after the expansion i s m a d e .
(a) P r o v e
(L + F ) n = (2F) n .
(b) Evaluate
(L + L ) n .
(c) Evaluate
(F + F ) n .
(d) How can this be suitably generalized?

NOTE: On occasion t h e r e will be p r o b l e m s listed at the ends of the a r t i c l e s in the
advanced and e l e m e n t a r y sections of the magazine. These p r o b l e m s a r e to be cons i d e r e d as logical extensions of the corresponding problem sections and solutions for
t h e s e p r o b l e m s will be d i s c u s s e d in these sections a s they a r e received.
See, for example, "Expansion of Analytic Functions In Polynomials Associated
with Fibonacci N u m b e r s , " by Paul F. Byrd, San Jose State College, in the first i s s u e
of the Q u a r t e r l y , and " L i n e a r R e c u r r e n c e Relations - P a r t I, " by J a m e s J e s k e , San
J o s e State College, in this i s s u e .
Solutions for p r o b l e m s in ISSUE ONE will appear in ISSUE THREE.
56
GENERATING FIBONACCI SERIES ON A DESK CALCULATOR
[April
A convenient method of generating Fibonacci n u m b e r s i s the alternating s u b total exchange and add p r o c e d u r e which i s easily p e r f o r m e d on the Olivetti T e t r a c t y s
desk calculator.
The Olivetti T e t r a c t y s has two separate accumulating r e g i s t e r s so a r r a n g e d that
the contents of either can be added to the other by manual commands and without r e writing n u m b e r s into the input keyboard. The detailed p r o c e d u r e for generating any
Fibonacci sequence on the T e t r a c t y s follows:
Set automatic total l e v e r (14) to left and set automatic accumulation lever (29) to
up position.
a) Clear the r e g i s t e r s by d e p r e s s i n g the green and black total keys.
b) Write the value of the initial t e r m of the sequence in the 10-key keyboard.
c) D e p r e s s the green add key. The initial value now printed on the tape in blue
ink is the sequence title but is not the f i r s t t e r m .
d) Advance the paper tape 3 l i n e s .
e) D e p r e s s simultaneously the g r e e n sub-total key and the black add key.
initial t e r m of the sequence is now printed on the tape in r e d ink.
The
f) D e p r e s s simultaneously the black sub-total key and the g r e e n add key.

second t e r m of the sequence is now printed on the tape in r e d ink.
The
g) Repeat step (e) for the third t e r m .

h) Repeat step (f) for the fourth t e r m .
i) Continue alternating steps (e) and (f) for as many t e r m s as d e s i r e d up to the
13 decimal digit capacity of the a r i t h m e t i c r e g i s t e r s .
j) The sequence can be continued beyond the 13 digit limit by clearing the r e g i s t e r s 9 step (a), and rewriting the r e q u i r e d m o s t significant digits of the l a s t
t e r m obtained as a new initial value, r e s t a r t i n g at step (b).
It takes a little p r a c t i c e to develop the manual knack of simultaneously d e p r e s sing the adjacent sub-total and add keys,
A firm push i s n e c e s s a r y but it m u s t not
l a s t too long or the operation will be done twice, producing an e r r o r ,

In generating long sequences it is a r e a d - o u t convenience to d e p r e s s the nonadd key after every fifth t e r m .
This provides a blank line and makes it easy to
count t e r m s ,
It is difficult? but not impossible to make e r r o r s .
The usual e r r o r s consist in
skipping a step or doing a step twice, which amount to the same thing.
The printed tape can be checked for e r r o r s in two ways.
1) Each t e r m approximates 1.6 t i m e s the preceding t e r m .
2) The + symbols on the far right side of the printed tape should alternate b e tween the two symbol columns. If two successive + signs fall in the s a m e
vertical symbol column, an e r r o r wras made at that point.
(Continued on p. 88)
PART H
BEGINNERS' CORNER
EDITED BY DMITRI THORO, SAN JOSE STATE COLLEGE
DIVISIBILITY II
We shall continue our investigation of some "background m a t e r i a l " for the b e ginning Fibonacci e x p l o r e r . Whenever n e c e s s a r y , we may a s s u m e that the i n t e g e r s
involved a r e not negative (or zero).
1.
DEFINITIONS
Two i n t e g e r s a and b a r e relatively p r i m e if their g r e a t e s t common divisor

(g. c. d. ) i s 1.
When convenient we will use the c u s t o m a r y abbreviation (a,b) to
designate the g. c. d. of a and b. Finally, as previously implied, we shall say that

n
* s composite if n has m o r e than two divisors,
2.
El.
ILLUSTRATIONS
If d = a and d| b , then d ( a b ) ; i. e. , a common divisor of two n u m b e r s is a
divisor of their sum or difference.

PROOF: d | a m e a n s t h e r e e x i s t s an integer a such that a = a'd. Similarly.
we may write b = b T d. Thus a b = d (a? b ? ) which p r o v e s that d[ (a ^ b). [ This
follows from the definition of divisibility. ]
E2.
If
d = (a,b).
then j -7 , -7 J = 1; i. e. , if two n u m b e r s a r e divided by their
g. c. d. , then the quotients a r e relatively p r i m e .
[This r e s u l t is a wide!}- used
"tool. "]
a
b
PROOF: Since d i s the g. c. d. of a and b. , and -7 a r e certainly i n t e g e r s :
to
let us call them a'

show that
and b , respectively.
(a'. b ) = 1.
a' = a"d'
Thus a = a'd,
"
and b = b"d'.
Then (there exist i n t e g e r s
b = b'd. and we a r e to
a" and b " such that)
This implies a = a'd = a"d d and
i . e . . d d is a common divisor of a and b. But we a s s u m e d d'

T
dd
E3.
to
The trick is to use an indirect argument.
Suppose (a ? , bT) = df > 1.

T
b = b'd - b"d'd;
1. which means
> d c o n t r a r y to the fact that d is the g r e a t e s t common divisor of
a and b.
If a and b a r e relatively p r i m e , what can you say about the g. c. d. of a -r b
and a - b?
For example, (13 -f 8, 13 - 8) - 1. but (5 - 3. 5 - 3) - 2.
out, however, that these a r e the only possibilities.'

57
It t u r n s
58
BEGINNERS' CORNER
If (a,b) = 1, then
PROOF:
(a + b , a - b) = 1 or 2.
(a + b, a - b) = d.
Then by E l , dj [(a + b) (a - b)] ; i. e. ,
(I) If d i s even, set d = 2 K. Then from
2a = ' 2KaT, 2b = 2Kb' we have a = Ka r ,
d|2a
Let
[April
and d | 2 b .
b = Kb\
Therefore
K m u s t be 1 (why?), and hence d = 2.
(ii) Similarly if d is odd, then d would have to divide both a and b ,
whence
d = 1.
Can you see what objection a pedantic r e a d e r might have to this proof?
E4y
The following is a special c a s e of a r e s u l t due to Sophie Germain, a F r e n c h
mathematician (1776-1831).
n 4 + 4 is composite for n > 1 .
PROOF: Unlike the preceding i l l u s t r a t i o n s , h e r e one needs to stumble onto
a factorization,
n 4 + 4 = (n2 + 2)2 - (2n)2 = (n2 + 2 + 2n) (n2 + 2 - 2n) does the t r i c k ,
for this shows that N = n 4 + 4 is divisible by a number between 1 and N and

hence must be composite,
E5.
We now consider one of the s i m p l e s t p r o p e r t i e s of Fibonacci n u m b e r s .

Two consecutive Fibonacci n u m b e r s are always relatively p r i m e .
PROOF: Certainly this i s obvious for the first few n u m b e r s : 1. 1, 2 f 3, 5,
8,
Let us use an i n d i r e c t argument.
Suppose
to
II-
and
_, i s the f i r s t
n +
p1 a i r for which (F . F , ) = d > 1. Now examine the p a i r F _, and F . Since

ii' n+1
n-1
n
F ., - F = F 1 ?J d is a divisor of F 1 (for d| F , 1 , d! F and hence, byJ E l .
n+1
n
n-l
n-1
< n+V
n
'
their difference). This m e a n s that F - and F a r e not relatively p r i m e a con'
n-1
n
tradiction to our assumption that F and F
i s the first such p a i r .
3.
Tl.
SOME USEFUL THEOREMS
Any composite integer n has at least one p r i m e factor.

PROOF:
(i) Since n i s composite, it must have at least one divisor g r e a t e r
than 1 and l e s s than n.

(ii) Let d be the s m a l l e s t divisor of n such that 1 < d < ii.
(iii) Suppose d is composite; let d r jd, 1 < d' < d.
(iv) Thus we have n = n t d = li^l'd"; i. e. , d' j n but d < d'
to the definition of d.
T2. Given n > 1.
than a.
Therefore
a contradiction
d must be a p r i m e .
Suppose that the quotient q.
If n is not divisible by 2. 3, 4.
in the division of n by a, is l e s s
. (a - 1). a, then n is a p r i m e .
1963]
BEGINNERS' CORNER
59
PROOF: (i) If q (q < a ) is the quotient and r the r e m a i n d e r in the division

of n by a,
we may w r i t e
n
r
- = q + -,
0 <
<
a.
(ii) A s s u m e that n i s not divisible by 2, 3 , 9 9 ' , a but has a divisor d, 1 < d < n.
We shall show that this leads to a contradiction.
(iii) Since djn,
n = dd T ,
where 1 < dT < n.
(iv) By (ii), df > a and by (i) a > q; hence d? >q o r d? > q + 1.
Also
d>a;
multiplying d - q + 1 by d > a. we a r r i v e at
(v) n = ddf > aq + a.
But by (i), n = aq + r < a q + a
since r < a.
This i s the
d e s i r e d contradiction which p r o v e s that n cannot be divisible by d, 1 < d < n9

and hence m u s t be a p r i m e .
T3. If n > 1 i s not divisible by pt = 2, p 2 = 3, p 3 = 5, p 4 = 7, , p k ? where
p^. i s the l a r g e s t p r i m e whose s q u a r e does not exceed n,
then n i s a p r i m e .
PROOF: A s s u m e a2 n < (a + l ) 2 and that n i s not divisible by 2, 3, ,'

(a - 1), a. Then n = dd? i m p l i e s d ^ a + 1, df a + 1,
a contradiction.
Thus n m u s t be p r i m e .
whence n = dd? ^
(a+1) 2
The r e a d e r should convince himself
that h e r e (as well as in T2) it suffices to consider only p r i m e divisors - Vn.
4.
PROBLEMS
1.1 Suppose that

(i) p i s the s m a l l e s t p r i m e factor of n and
V .
(ii) p > vn
What i n t e r e s t i n g conclusion can you draw?
1.2 P r o v e that two consecutive Fibonacci n u m b e r s a r e relatively p r i m e by
using one of the identities on p. 66 (Fibonacci Q u a r t e r l y , F e b r u a r y , 1963).
1.3
by 6.
P r o v e that if p and p + 2 a r e (twin) p r i m e s , then p + 1 i s divisible
(Assume p > 3. ) [This p r o b l e m was suggested by J a m e s Smart. ]

n
1.4 P r o v e that if n i s divisible by k, 1 < k < n, then 2 - 1 i s divisible
by 2 k - 1.
F o r example, 2 35 - 1 = 34 359 738 367 is divisible by 2 5 - 1 =31 and
2 7 - 1 = 127.
BEGINNERS' CORNER
60
[April 1963 ]
1.5 P r o v e that t h e r e a r e infinitely many p r i m e s .
Hint: Assuming that
i s the l a r g e s t p r i m e , Euclid considered the expression N = 1 + 2 - 3 - 5 - 7 - - - p .

Now either N i s p r i m e or N i s composite.
Complete h i s proof by investigating
the consequences of each alternative.

Additional hints may be found on p. 80.
FIBONACCI FORMULAS
Maxey Brooke 3 Sweeny, Texas
If you have a favorite Fibonacci formula, send it to us and we will try to publish
it.
Some historically i n t e r e s t i n g ones a r e shown below.
1.
P e r h a p s the f i r s t Fibonacci formula was developed by Simpson in 1753.

F
2.
- F 2 = v( - l ) n
1 F
n+1
n - 1n
n
'
A very i m p o r t a n t formula was developed in 1879 by an o b s c u r e F r e n c h m a t h e matician, Aurifeuille.
In fact, it is his one claim to fame,
L_ = L (L Q - 5F + 3 ) ( L 0 + 5F + 3)
5n
n 2n
n
' 2n
n
'
3.
The only formula involving cubes of Fibonacci n u m b e r s given in Dickson T s

"History of the Theory of N u m b e r s " i s due to L u c a s .
F3
n+1
F3
F3
n-1
3n
F* ,' - 3F + F 3 n = 3 F 0
.
n+2
n
n-1
3n
The r e c u r s i o n formula for sub-factorials i s s i m i l a r to the one for Fibonacci
The late Jekuthiel Ginsburg offers
te
4.
numbers:
5.
P , , = n(P + P J ; P 0 = 1 ,
v
n+1
n
n-17' u
*
Vx = 0 .
i
Fibonacci n u m b e r s have been r e l a t e d to a l m o s t every other kind of n u m b e r .

H e r e i s He S. Vandiver's relation with Bernoulli n u m b e r s .
2kF2k
2k
hmod^
^ P = 5a+ 1
k=0
^ 3
)
2(k-l) ~
X (modp
i f P = 5a 2
k=0
p - 3 denotes the g r e a t e s t integer not exceeding (p - 3 ) / 2 .
I think that this is a good idea.
Ed.
A PRIMER ON THE FIBONACCI SEQUENCE - PART II

S.. L. BASIN A M I VERNER E. HOGGATT, J R . , SAN JOSE STATE COLLEGE
1.
INTRODUCTION
The proofs of existing Fibonacci identities and the discovery of new identities
can be greatly simplified if matrix algebra, and a particular 2 x 2 matrix are introduced. The matrix approach to the study of recurring sequences has been used for
some time [1J and the Q matrix appeared in a thesis by C. H. King [2] . We first
present the basic tools of matrix algebra.
THE ALGEBRA OF (TWO-BY-TWO) MATRICES
The two-by-two matrix A is an array of four elements a, b. c, d:
A =
The zero matrix. Z, is defined as.

0
The identity matrix, I, is
The matrix C. which is the matrix sum of two matrices A and B. is
C = A + B=
/a
b\
/e
f \
\c
d/
\g
h/
Ut
/a^
b+f
\c-g
d-h
U/
The matrix P, which is the matrix product of two matrices A and B5 is
P = AB = / a
\ c
Ve
V/"*1*
d /\g
h/
61
\ce+dg
^cf+dh
62
[April
The determinant D(A) of m a t r i x A i s

a
ad - be.
D(A)
Two m a t r i c e s a r e equal if and only if the corresponding elements a r e equal; t h a t i s ,
A =
if and only if,
b\
/e
d/
\ g
a = e, b = f, c = g, d = h.
A SIMPLE THEOREM
The determinant,
D(P),
of the product,
P = AB.
of two m a t r i c e s A and B i s
the product of the d e t e r m i n a n t s D(A) and D(B)

D(P) = D(AB) = D(A) D(B)
The proof is left as a simple e x e r c i s e in algebra.

THE Q MATRIX
The Q m a t r i x and the determinant of Q, D(Q)5
Q = I
If we designate Q = (
(T
,0
1/
) = l?
J,
are:
D(Q) = - 1 .
then
Q = Q 1 = Qu Q = IQ = QI = QQU .
DEFINITION: Q n + 1 = Q n Q 1 ,
law of exponents for m a t r i c e s .
an inductive definition where Q1 = Q.
This i s the
1963 J
It is easily proved by mathematical induction that

F
Q1J
^
n+1
F
n
F
n
where F
n-1
is the nth Fibonacci n u m b e r , and the d e t e r m i n a n t of Q

'
is
D(Q n ) = D n (Q) = ( - l ) n
MORE PROOFS
We may now prove s e v e r a l of the identities very nicely.
Let us prove identity III
(given in P a r t I), that i s ,

F . -. F
- F2
n+1 n - 1
n
Proof:
If
Q" =
/F ^
n+1
= (-l)n
v
F
n
F
n
and D(Q n ) = ( - i ) n
n-1
then
F
n
D(Q ) =
x 1
n+1
n
n-1
- F2 n+1 n - 1
n
F
Let us prove identity VI
2n+l
j
then
n + 1
F*
+ F2
n+1
n
Qn = Q 2 n + 1
(-n 1 1
;
64
[April
Qn Q n + 1
F
+ F F
n+1 n+2
n n 4 -!
F*
+ F"
n+1
n
F F , 0 + F
,
n F
n n+2
n - 1 n+1
But this is also
,2ml
F F ^ + F
, F
n n+1
n-1 n
2n+2
211+1
F
2n+l
2n
Since these two m a t r i c e s a r e equal we may equate corresponding elements so that

F0 0 = F
F 0 + F F
,
2n+2
n+1 n+2
n n+1
(Upper
Left);
v
*^
F . ., = F 2
t F2
2n+l
n+1
n
(Upper Right)
1K
&
'
F0 , = F F
+F
- F ^,
2n+l
n n+j
n - 1 n+1
Fn
2n
(Lower Right)
to
'
= F F
+ F , F
n n+1
n-1 n
(Lower Left)
;
F u(F
+ F
)
^n n+1
n-l;
If we accept
1 identity
J V: L
= F
, - F 1 , then
n+1
n-1
F0
= F L
2n
n n
which gives identity VIII.
Return again to
F0
= F v(F
+ F
J
2n
n n+1
n-l;
From
F, . = F, , + F, . for
k- -2
k-f 1
k
thus also
k = n - 1.
'
one can write
= F
n
. - F
n-1
n-1
1963]
65
F 0 . = v(F _L1 - F 1 ) v(F , n + F n7) = F 2 , - F 2

2n
n+1
n - 1 ' n+1
n-1
n+1
n
which i s identity VII.
It is a simple task to verify
Q2 - Q + I
and
Qn+2 = Qn+1
Qn ,
and
Qn = Q F n
where F
I Fn_x ,
i s the nth Fibonacci number and the multiplication of m a t r i x A, by a
number q,
i s defined by
b \
/aq
bq
d/
\ cq
dq
qA = q
GENERATION OF FIBONACCI NUMBERS BY LONG DIVISION
2
= Fil + FL9 x + Fox
+ + F xn
6
1 - x - x'
In the p r o c e s s of long division below
1 - x - x2 I
t h e r e i s no ending.
As far as you c a r e to go the p r o c e s s will yield Fibonacci Num-
b e r s as the coefficients.
6Q

F
[April
AS A FUNCTION OF ITS SUBSCRIPT
It is not difficult to show by mathematical induction that

n
P(n):
Fr
, j (l ^y. ^ j - j
This can be derived in many ways,
P ( l ) and P(2) a r e clearly t r u e .
From
= F, _.. + F, _ 9 and the inductive assumption that P(k-2) and P ( k - l ) a r e t r u e ,
(a)
F,
then
fM^T-Ml
Fk_2
1 + NO
"!
/ i - ^ -
Adding, after a simple algebra step, we get
"1
+F
"2
-L U LJ*fZ (l*
N/5
IV
\ ./i^^f2/!^
Observing that
1 + 45
2
1 -
N/5 + x
3 + \fe
"
/ 1 + N/5 \ 2
2
3-^5
it follows simply that if (a) and (b) a r e t r u e (P(k - 2)
and
P(k - 1) a r e
then for n = k,
P(k):
Fk = Fk_x
Fk_2
1 U 1 + ^5 \ k
^
/ l
- ^5
X k
true),
19631
The proof i s complete by m a t h e m a t i c a l induction,,

L
67
Similarly it may be shown that
= F j_1 + F ,
n+1
n-1
and
, - (H^ 5 )" ( H * ) "

Let us now prove identity VIII
FQ
= F L
2n
n n
Proof:
'i
2n
%r 5 \ n
^r
N/5
Now factoring:
* -75{{~v~>
-(-7-)nvr~);{-r~)
F0
= F L
2n
n n
MORE IDENTITIES
.-ih*;-^)")
- w-^H
XIV.
1 + ^5\n
/l - ^5\n
Fc
XVI.
Bf
XVII.
l-F
F|
<
= -
+ (-1)
^
+ 2F? + 3 F o + - - . + n F
z
n+1
6F
+ 5
5ll
= (n+1) F n - F- , + 2
v
' n+2
n+4
68
XVIH.
XIX.
[April 1963]
F2 + F4 +
+ F = F 0 _ - 1
2n
2n+l
F 11 F2 2 + F l2 Fo
, F = \ (F - F^ F
) - 1
6 + Fo
6 F4 4 + . . . + F
n-1 n
2 n+2
n n-1'
1
JoGH-,
XX.
where
XXI
**A'
XXII.
(
1
) =
J
n
,
and m ! =
( n - i)I i !
1 . 2 . 3 . . . - m .
F
= F3 + F3
- F3
*3n+3
n+1
n+2
n
F F - F , F M = (-l) 1 1 " 1 ^^ F
.
v
n m
n - k m+k
'
k m+k-n
REFERENCES
1.
J. S. F r a m e ,
"Continued fractions and m a t r i c e s , " A m e r .
Math.
Monthly,
'Some p r o p e r t i e s of the Fibonacci n u m b e r s /
Master's
Feb. 1949, p. 38.

2.
Charles H.
King,
T h e s i s , San J o s e State College, J u n e . 1960.
REQUEST
The Fibonacci Bibliographical R e s e a r c h Center d e s i r e s that any r e a d e r finding
a Fibonacci r e f e r e n c e send a c a r d giving the r e f e r e n c e and a b r i e f description of the
contents.
P l e a s e forward all such information to:

Fibonacci Bibliographical R e s e a r c h Center,
Mathematics Department,
San Jose State College,
San J o s e , Calif.
SPECIAL NOTICE
The Fibonacci Association h a s on hand 14 copies of Dov J a r d e n ,
Recurrent
Sequences, Riveon Lematematika, J e r u s a l e m , I s r a e l . This is a collection of p a p e r s

on Fibonacci and Lucas n u m b e r s with extensive tables of factors extending to the
385th Fibonacci and Lucas n u m b e r s .
The volume sells for $5.00 and i s an e x c e l -
lent investment. Check or money o r d e r should be sent to Verner Hoggatt at San J o s e

State College, San J o s e , California.
LINEAR RECURRENCE RELATIONS - PART I

J/W1ES A.
JESKE,
SAN J O S E STATE
COLLEGE
1.
INTRODUCTION
Most of t h e special s e q u e n c e s , which appear in The Fibonacci Q u a r t e r l y ,

satisfy a type of equation called a r e c u r r e n c e relation i . e . , a difference equation
whose independent variable i s r e s t r i c t e d to integral values.
Although t h e r e a r c
s e v e r a l good textbooks (e. g. , s e e [1] , [2] , [3] o r [ 4 ] ) which p r e s e n t v a r i o u s

methods of solution for many such equations, the beginner may not b e acquainted
with any of t h e m , and in fact i s likely to have m o r e knowledge of the theory of differential equations than of that of r e c u r r e n c e r e l a t i o n s .
The purpose of this s e r i e s of a r t i c l e s i s to introduce the beginner to the s u b ject, and to derive explicit e x p r e s s i o n s for the solution of c e r t a i n g e n e r a l ,
r e c u r r e n c e r e l a t i o n s by applying a generating function transformation.
linear
The p a r -
t i c u l a r generating function chosen i s seldom used in the t r e a t m e n t of r e c u r r e n c e

r e l a t i o n s , but for the purpose of developing general formulas it h a s the advantage
of immediately t r a n s f o r m i n g t h e problem to a m o r e familiar one involving differential equations, for which t h e r e i s already available a g r e a t wealth of special
formulas and techniques.
2.
DEFINITIONS
A l i n e a r r e c u r r e n c e relation of o r d e r k i s an equation of the form

k
I2-1)
Saj,nVj
= b
n >
j=0
and b
w h e r e a^ , a, . . . . a,
a r e given functions
of the independent v a r i to
0,n' l , n '
k, n
n
^
able n over the s e t of consecutive non-negative i n t e g e r s S, and a 0 a,
^ 0
on S . If b
= 0 , the relation i s called homogeneous, otherwise it i s said to be
non-homogeneous.
x(2.2)
' '
We may introduce the translation o p e r a t o r

E m Jy = Jy
n
n+m
v(m
= 0, 1, , k ) ,
' '
' ' '
and thus we can w r i t e (2.1) a s

(2.3)
Lk(E)yn
= bn
69
E ,
defined by
70
where the linear operator
[April
L, (E) is
(2.4)
Lk(E)
j=0
A sequence whose t e r m s a r e v
the set S if the substitution y
i s a solution to the r e c u r r e n c e relation on
= v
reduces relation (2.3) to an identity on S.
If a set of k s u c c e s s i v e initial values y 0s y 1} --, y, _-, i s given a r b i t r a r i l y ,

equation (2.1) or (2.3) enables us to extend this s e t to k + 1. successive values.
Using mathematical induction, it can be easily established that the r e c u r r e n c e r e lation (2.3) over the set S of consecutive non-negative integers has one and only
one solution for which the k values a r e p r e s c r i b e d .
3.
A SERIES TRANSFORM
F o r the sequence {y }, n = 0, 1, 2, # -v
we introduce the exponential gen-
erating function defined by tne infinite s e r i e s
(3.1)
Y(t)
V y n.
jLjyn
n=0
which we suppose is convergent for some positive value of t.
F r o m (3.1) we find
the derived s e r i e s
(3.2)
dJY __ v ( j ) , , x
ax
V*
0 = 0 , 1 , - . - , k).
n=0
These s e r i e s of c o u r s e have the same radius of convergence as (3.1), and a r e seen

as the generating functions of the sequences {y
. } , j = 0,l,--*,k.
we have the i n v e r s e t r a n s f o r m
(3.3)
yn =
^\o)
Y(t)
n
dt
t=0
(n=0,l,2,---,k) .
Now from (3.1),
1963]
71
The relations (3.2) and (3.3) follow from known p r o p e r t i e s of Taylor s e r i e s .

4.
EXPLICIT SOLUTION OF A LINEAR RECURRENCE RELATION
We shall now derive the formula for the g e n e r a l solution to the l i n e a r h o m o geneous r e c u r r e n c e relation
k
Eajyn+j
3=0
(4.1)
with constant coefficients.
k(E^n
= 0
(Discussion of the non-homogeneous c a s e will appear in
the next i s s u e of this j o u r n a l . ) The derivation i s based on the application of the e x ponential generating function (3.1) which t r a n s f o r m s the r e c u r r e n c e relation into a
m o r e familiar differential equation.
Multiplying both sides of (4.1) by t /n.T
and summing over n from
0 to o}
we thus obtain the t r a n s f o r m e d equation

k
(4'2>
a Yd)(t) = L
k(D)Y
'
(D
dt ) '
J=0
which i s an ordinary l i n e a r differential equation of o r d e r k. Now it is well known 1
that, if r l 5 r 2 ; . r,
(4.3)
a r e k distinct r o o t s of the c h a r a c t e r i s t i c equation

L k ( r ) = 0,
then the general solution of (4.2) is given by

k
(4.4)
Y(t) = 2^
c.e
i=l
w h e r e c. a r e k a r b i t r a r y constants.
Application of the i n v e r s e t r a n s f o r m (3,
then yields immediately the explicit formula
^ e e for example, almost any textbook on ordinary differential equations.
72
[April
y n -E c i4
(4.5)
i=l
for the general solution of the r e c u r r e n c e relation (4.1).
In the c a s e where the c h a r a c t e r i s t i c equations LjJr) = 0 p o s s e s s e s m d i s tinct r o o t s rl9 r 2 , , r
and each root r . being of multiplicity m. ( i = l , % m ) ,
with
(4.6).
y^m, = k ,
1=1
the differential equation (4. 2) is known to have the general solution
<4-7>
Y(t) = e
1=1
where b . . a r e k a r b i t r a r y c o n s t a n t s ,
m.-l
I
b tJ
ij '
j=0
Applying the i n v e r s e t r a n s f o r m (3.3), we
then obtain the general solution
m.-l
l
i=l
3=0
to the r e c u r r e n c e relation (4.1).

In P a r t II of this a r t i c l e , we shall not only derive an explicit formula for the
general solution of the non-homogeneous l i n e a r r e c u r r e n c e relation with constant
coefficients, but shall also show how the method employing the exponential generating function may solve c e r t a i n r e c u r r e n c e r e l a t i o n s having variable coefficients.
5
EXAMPLE
The Fibonacci n u m b e r s satisfy the s e c o n d - o r d e r r e c u r r e n c e relation

( 5 a )
n+2 "
n+1 "
n = >
= >
i = 1
L963]
(5.2)
L2(E)F^ = 0
73
where
(5.3)
L 2 (E) = E ' - E - 1.
The c h a r a c t e r i s t i c equation L 2 (r) = 0 has the distinct roots

(5.4)
i-i
(1 + V5)/2,
R2
(1 -
N/5)/2
so that the formula (4.5) immediately yields

^
(5.5)
Now since F 0 = 0, Fj = 1,
iri
2r2
we obtain ct - - c 2 = l / \ / 5 ; hence the general solu-
tion for the Fibonacci sequence is e x p r e s s e d by
(5.6)
i + V" 5 \ n
^5\
^5
We note from (4.2) that the t r a n s f o r m e d equation for (5,1) i s the s e c o n d - o r d e r

differential equation
(5.7)
y n - Y' - Y = 0,
Y(0) = 0, Y f (0) = 1
Hence the exponential generating function for the Fibonacci sequence is
OO
(5.8)
r2t
Y(t) = [e^ - e J/V5 = ^ F n g

n=0
while the well-known o r d i n a r y generating function for this sequence is
74
(5.9)
W(t) =
1 - t - t
g =^2
~
n=0
ntn
[April 1963]
'
The two generating functions W(t) and Y(t) a r e r e l a t e d by the e x p r e s s i o n

OO
(5.10)
W(t) =
I e" Z Y(tz)dz
o.
REFERENCES
1.
L. M. Milne-Thompson, The Calculus of Finite Differences, London, 1933.
2.
C. J o r d a n , Calculus of Finite Differences, New York, 2nd Ed. , 1947.
3.
S. Goldberg, Introduction to Difference Equations, New York, 1958.
4.
G. Boole, Calculus of Finite Differences, New York, 4th Ed. , 1926.
PROBLEM DEPARTMENT
P-l.
The r e c u r r e n c e relation for the sequence of Lucas n u m b e r s is

L
l0
n+2
- L ,., - L = 0
n+1
n
with LH
x == 1,
'
LL2 = 3 .
Find the t r a n s f o r m e d equation, the exponential generating function,
and the
general solution.
P-2.
Find the general solution and the exponential generating function for the r e c u r r e n c e relation
yn+3 with
5y
n+2
V l
"
4y
= .
y 0 = 0, y t = 0, y 2 = - 1 .
REQUEST
Maxey Brooke would like any r e f e r e n c e s suitable for a Lucas bibliography,
His a d d r e s s i s 912 Old Ocean A v e . , Sweeny, Tex.
SOME FIBONACCI RESULTS USING FIBONACCI-TYPE SEQUENCES

I.
DALE RUGGLES, SAN JOSE STATE
COLLEGE
The elements of the Fibonacci sequence satisfy the r e c u r s i o n formula,
n+1
= F + F 1 5 where Fu0 = 0 and F11
1. Let us define an F-sequence a s one for
n
n-1
which the r e c u r s i o n formula u
u + u -, holds for the elements u of the
n
n-1
n
n+1
sequence.
Suppose {u } and {v } a r e two F - s e q u e n c e s .
|cu
+ dv 1, i s also an F - s e q u e n c e .
Then a linear combination,
If the determinant
"I
Vl
v2
^0
then by an application of a t h e o r e m from algebra, every F-sequence
can be e x -
p r e s s e d a s a unique l i n e a r combination of the F - s e q u e n c e s {u } and {v } .

Consider the sequence 1 , 7 , 7 2 , 7 3 ,
7
=7
This will be an
F - s e q u e n c e if
for all i n t e g e r s n; that i s , for 7 such that 7

7+1.
i + 'sfh
i
equation h a s solutions which we will denote by [3 = ~
and a =
T h u s , t h e ^ - s e q u e n c e 1, a, a2, and the (3-sequence
1, /3, [32,
This
\T5
.
are
F - s e q u e n c e s . These can be extended to include negative integer exponents a s well.

Since
2|'
P = p - a = ^5
every F - s e q u e n c e can be written a s a unique linear combination of the cy-sequence

and the >-sequence.
(Note that a + (3 = 1 and a[3 - - 1 . )
In p a r t i c u l a r this applies to the Fibonacci sequence.

' F j = 1 = ca + d{3
F 2 = 1 = ca2 + d/32
one finds that c = - 1 / V 5 and d = l/4~5.

nn
Thus,
75
n
_ (3
[3(3 -- aa
n
- a
\fe
F r o m the equations
SOME FIBONACCI RESULTS
76
[April
The F - s e q u e n c e with L^L = 1 and L 2 = 3 is known a s the Lucas sequence.

In the c a s e of the Lucas sequence.
T
L
The a-
on
= B + a
n
*
and ^ - s e q u e n c e s can be used to prove many well-known r e l a t i o n s
involving Fibonacci n u m b e r s , Lucas n u m b e r s , and general

1.
Since
0n
and L
f3 - a
n
n
(3
+ a then it follows immediatelv that
= (3
r
Since (3U+1 + /3 n
follows that / 3
Also,
'
g^JL.
2.
n+1
n+1
+ L
n 2n
= p
n
2n
-a
(3 -
= F 2n
= f3n ((3 + (3 X) = (3U ((3 - a) and aU+1 + ^ n
n+1
- a
n X
+ p~
11 1
- a11 {a - 0),
it
- a '
= ((3 - a){f3 + a \ thus L n = F n + 1 + F n _ r
, = 5F
can be similarlyJ shown.
n-1
n
n
3.
F-sequences:
Let {u } be an F - s e q u e n c e , such that u
= ca
n
+ d(3 .
Then the determinant
n+1
-*
n-1
i
can be simplified a s follows:
ca
n+1
A.
-.
n-1
ca
n+1
, n+l
+ dp0
11
n11
+ d/3
n-1
,0n-l
ca
+ d/3
+ d/3
n+1
0 n+l
cd
= cd
.n-1
, - n+1 _ ,
(-1)x
5 cd .
ca
n-1
1^
1963]
USING FIBONACCI-TYPE SEQUENCES
77
"n+1
F
4.
= (-D
,
n-1
F2
- F2
= F 0 F 0 for all p and n.
n+p
n-p
2n
2p
^
n+p
n-p
n+p
-j
p - a
Consider F
/Dn~P
n+p
+ F
n+p
n-p
Then,
'
~P
P- a
0 P + /3-P)[/3 n + (-D P + Vj
/3 -
[/?"
(-l)P
since a
= (-1) P /3 P
V ] [fP
(-ifgP]
P - a
T h e r e f o r e , if p i s even, F , + F
= F . L and if p i s odd. F
+ F
n+p
n-p
n
p
^
'
n+p
n-p
= L F .
n
p
Also, F
- F
= L . F for p even and F
- F
= F L for
'
n+p
n -2 p
n
p
n+p
n-p
n
p
2
p odd. T h u s , F
- F
= F 0 F 0 for all p and n.
^
'
n+p
n-p
2n
2p
5. Let us simplify F 3 + F 6 + + F .
Since the ^ - s e q u e n c e and t h e
psequence a r e also g e o m e t r i c sequences it follows that
0 3 / 0 3n
P(P
p + ... + p 3n
/3 3 -
~ 1)
1
and
a
Thus,
F3 + F6 +
+ ..
+ a
3, 3n
a (a
- 1X
1)
3n
^3n
3n
Q
3 + a
+ p
+ F
F
3n+3
3n "
3n+3
0 3n+3
__
+ a
a)
3
78
6.
[April
SOME FIBONACCI RESULTS

As another example consider
F 1 + 2 F 2 + + n F , n positive.
+ n/3
P + 2/3'
= /3
^-l
0 n+2
n/3
- gP
n+3 ,
j8
Now
+ jS
since
n-1
1 + 2x + + nx
Also, a + 2a2 + + na
= nF
7.
n+20
- F
0
n+3
+ Fo.
6
= no?
-a
dx
+ ad.
x(x - 1)
(x - 1)
Therefore,
Fx + 2 F 2 + . . . + n F
Note that a s i m i l a r r e s u l t holds for a &general
Let us consider some r e s u l t s that utilize the binomial t h e o r e m .
/3n = (1-af
vU
(-1)^
j=0
and
*n = a-/3)n = E ( ) ( - D V .
it follows that
j=0\ j ,
hence,
r
Also,
=y
(-D 3 + 1 F,
j=0
ÊOA
j=0
F-sequence.
Since
1963]
8.
USING FIBONACCI-TYPE SEQUENCES
Again using the binomial t h e o r e m ,
2n
>
= a+^ = v r w
and
^ = (i + ^ n = i :
3=0 O
Therefore
j=o
VJ
also
If
{u } i s a general F - s e q u e n c e , it also follows that
u0
2n
SO-
9. As a final example to i l l u s t r a t e the usefulness of the aestablishing Fibonacci r e l a t i o n s we will derive the r e s u l t
F = F
^ F + F
n
n-p+1 p
n-p
for all n and p .
F i r s t , from
F ,
p-1
and /3-sequences in
79
80
SOME FIBONACCI RESULTS USING FIBONACCI-TYPE SEQUENCES

,p , n - p - l
[3^ [3
nP-1 ,n-p
+ [^
j3
,n-l
,n-l
= j3
nn.
- fS
= (3 (j3 -
a)
and
,-# n-p+1
.p-1 n-p
we obtain
n-p+1
n-p
Similarly, one can show that

a
It then follows that F
n
- a r
_, -i- a
n-p-^l
= F F
-, + F J
p n-p+1
p-1
1
n-p
and if
n-p
{u } i s an F - s e q u e n c e .
L
nJ
then
u
= u F
, + u , F
p n-p+1
p - 1 n-p
Note that if q = n - p + 1. then u

1 = u F - r u ^ 1 F
1
^
p+q-1
p q
p - 1 q-1
^
n
n
n
n
Since p -a
= V5 F and [3 + a ~ L , it follows that
'
n
m
L + \/5 F
.n
n
n
/3
and
L
n
/
- \/5 F
n
_
a
=
VMVWWW\V\VYVW^
HINTS TO BEGINNERS' CORNER PROBLEMS

(See page 59)
..1 Examine
P
1.2
Use identity III .
1.3 Notice that p. p -r- 1. p + 2 a r e three consecutive i n t e g e r s .

an odd p r i m e ,
1.4
p - 1 is even.
Since p > 3 is
Why must p * 1 be a multiple of 3?
2 5 * 7 - 1 = ( 2 5 ) ' - ( 1 ) T = ( 2 s - 1 ) [ ( 2 5 ) 6 + ( 2 5 ) 5 4- . . . + ( 2 5 ) t- 1 ] .
1.5 If N is composite, then by T l it must have a p r i m e factor p.
This factor
must be one of the following: 2. 3, 5, 7, , p . Thus p;N and p|(2-3-5---p ).
EXPLORING RECURRENT SEQUENCES

EDITED BY BROTHER U. ALFRED, S T . M A R Y ' S COLLEGE,
CALIF.
The following a r t i c l e constitutes the Elementary R e s e a r c h Department of the

p r e s e n t i s s u e of the Fibonacci Quarterly.
R e a d e r s a r e r e q u e s t e d to send their d i s -
c o v e r i e s , q u e r i e s , and suggestions dealing with this portion of the Quarterly
to
B r o t h e r U. Alfred, St. M a r y ' s College, Calif.

Everyone who buys i n s u r a n c e i s urged to r e a d the fine print because it u s u a l ly contains qualifications of an important n a t u r e .
In a s i m i l a r vein the r e a d e r s of
the newly c r e a t e d Fibonacci Quarterly should turn to the inside cover and examine
the sub-title: "A journal devoted to the study of i n t e g e r s with special p r o p e r t i e s . M
This in no way indicates that the editors could not fill the pages of t h e i r magazine
with m a t e r i a l dealing exclusively with Fibonacci sequences.
It does, however, p r o -
vide for a m e a s u r e of latitude and a c e r t a i n variety in the contents while adhering to

the main theme indicated by the title of the magazine.
In this s p i r i t , the
Tt
Fibonacci
e x p l o r e r s ' ' a r e invited to look into a somewhat b r o a d e r topic: R e c u r r e n t Sequences.

The word " r e c u r r e n t " need not frighten anyone.
repetition.
R e c u r r e n c e simply means
A sequence is a set of mathematical quantities that a r e o r d e r e d in the
s a m e way as the i n t e g e r s : 1, 2, 3,"9
Put the two ideas together- and the
result
i s a " r e c u r r e n t sequence. "

P e r h a p s the s i m p l e s t example of such a sequence i s the i n t e g e r s t h e m s e l v e s .
Let us denote the t e r m s of our sequence by T l 5 T 2 , T 3 , e o e , T .
In the case of the
i n t e g e r s , the relation involved i s

T
i
- T
n+l
+ 1
n
that i s , every integer is one m o r e than the integer preceding it. This idea is readily
extended to even i n t e g e r s and odd i n t e g e r s .
If, for example,
the next even integer i s

T
Likewise, if T
, = T + 2
n+1
n
i s an odd i n t e g e r , the next odd integer i s
, = T + 2
n+1
n
81
T n is an even integer
82
Now look at the last two laws of r e c u r r e n c e .
[April
They a r e the same! This fact is a
s o u r c e of confusion to students of elementary algebra wTho think that if x and x - 2

r e p r e s e n t consecutive even i n t e g e r s , something else would r e p r e s e n t
odd i n t e g e r s .
consecutive
The answer l i e s , of c o u r s e , in the "if" portion of the proposition. If
x is an odd integer, then x - 2 is also the next odd integer.

The natural extension of such relations which we have been considering is the
arithmetic p r o g r e s s i o n in which each t e r m differs from the preceding t e r m by a
fixed quantity,
a,
called the common difference.
Thus for this type of sequence
we have
- = .T
n+T
n
+ a
,
*
where a can be any r e a l or complex quantity.

Another, well-known type of r e c u r r e n c e sequence is the geometric
sion in which each t e r m is a fixed multiple, r , of the previous t e r m .
progres-
The relation
in this case is
T
= rT
n+1
A simple example i s : 2. 6, 18, 54, 162, , where r = 3, Tj = 2,

We now come to the c o m m e r c i a l .
Recurrent sequences in which each t e r m is
the sum of the two preceding t e r m s a r e known as Fibonacci sequences.
The law of
r e c u r r e n c e for all such sequences is

T .- = T + T - .
n+1
n
n-1
Starting with the values of T t and To, it is possible to build up such a s e q u e n c e .
Thus, if Tj = 3 and T 2 = 1 1 , it follows that T 3 = 14, T 4 = 25,
One can go on to variations of this idea.
T
n+11
= 2T
For example:
+ 3T
,
n-1
or
T
n-1n
- T
+ T
. + T 0
n-1
n-2
T 5 = 395 .
1963J
83
Any one such sequence can be the subject of a g r e a t deal of investigation and r e s e a r c h which can lead to many i n t e r e s t i n g mathematical r e s u l t s .
At this juncture it may be well to point out that in some i n s t a n c e s ,
the law
of r e c u r r e n c e i s such that it i s possible to work out an explicit m a t h e m a t i c a l e x p r e s sion f o r
the
nth
term.
In o t h e r s ,
example; if Tx = 1, T 2 = 1,
this i s not convenient o r p o s s i b l e . F o r
and every t e r m i s the sum of all the t e r m s p r e c e d i n g
it, we find directly that T 3 = 2 , T 4 = 4, T 5 = 8, T 6 = 16,
T 7 = 32, . . .
, so that
a p e r s o n not endowed with mathematical genius can see that the nth t e r m i s given
by
?
On the other hand, if Tt = 1, T 2 = 1 and
n+1
n-1
'
we have the well-known Fibonacci sequence: 1, 1, 2, 3, 5, 8, 13, 2 1 , 34, whose

t e r m s a r e such that they a r e not as readily e x p r e s s i b l e by a s i m p l e formula. Hence,
we establish them as a s t a n d a r d sequence which can s e r v e to e x p r e s s r e s u l t s found
in other sequences.
EXPLORATION
A few sequences worthy of exploration have already been indicated.
Other
suggestions follow, and beginning r e a d e r s a r e urged to c r e a t e additional sequences

of t h e i r own.
Interesting mathematical r e s u l t s derived from such work should be
communicated to the Editor of this department of the Fibonacci Quarterly. H e r e a r e

a few suggestions to s t a r t you exploring:
1.
Let Tt = a,
T2 = b ,
w h e r e a and b a r e any positive n u m b e r s , and let the
law of formation in the sequence be that each t e r m i s the quotient of the two p r e ceding t e r m s .
2.
Starting with the s a m e initial t e r m s , let each t e r m be the product of the two
previous t e r m s .
3.
Another law: Let each odd-numbered t e r m be the sum of the two previous t e r m s
and each e v e n - n u m b e r e d t e r m be the difference of the two p r e v i o u s t e r m s .
4.
Let each odd-numbered t e r m be the product of the two preceding t e r m s and each
even-numbered t e r m be the quotient of the two preceding t e r m s .
5.
Starting with Tt = a, T 2 = b , T 3 = c,
T
let the law of formation b e :
., = T + T
- - T 0
n+1
n
n-1
n-2
LETTERS TO THE EDITOR
84
[April
A Computer Investigation of a Property of the Fibonacci Sequence

Stephen P. Geller
Mathematics Department, University of Alaska
February 18, 1963
Publication of a table of the first 571 Fibonacci numbers in Recreational Mathematics Magazine (Oct. 1962) brought out the fact that the last (units) digit of the sequence is periodic with period 60, i.e. , the 1,1,2,3,4, sequence repeats on the
last digit every 60 entires of the sequence. It also appeared that the last two are
similarly periodic with a period of 300. Noting that the table had been calculated by
an IBM 7090 digital computer, I resolved to set up our IBM 1620 to check out the
above observations and extend to more digits. The size of our memory (20K) prohibited calculation of the terms of the sequence in their entirety, but this was no1
necessary since it was quite easy on this machine to truncate off all the digits oJ
the running sums beyond those under consideration. The machine verified that the
last two digits repeat every 300 times, the last three every 1500, the last four ever}
15000, the last five every 150,000, and finally after the computer ran for nearly
three hours a repetition of the last six digits appeared at the 1, 500, 000th Fibonacc:
number. These may be written in the form:
, (n+60)
^rm " F n = 0 v(mod 10);
(n+300)-Fn"(mod100)
Vl500) ' V
0(-odl000)
F,(n+15000)
, , , " F n = 0 (mod 10000)'
F
V150000) - Fn
F
(n + 1500000) " n
<m0d
S
mod
100000
>
1000000
>
There does not yet seem to be any way of guessing the next period, but perhaps
a new program for the machine which will permit initialization at any point in the sequence for a test will cut down computer time enough so that more data can be'
gathered for conjecturing some rule for these repetition periods.
ELEMENTARY PROBLEMS AND SOLUTIONS

EDITED BY S.
L.
BASIN,
SAN JOSE STATE
COLLEGE
Send all communications r e g a r d i n g E l e m e n t a r y P r o b l e m s and Solutions to

S. L. B a s i n , 946 Rose Ave. , Redwood City, California. We welcome any p r o b l e m s
believed to be new in the a r e a of r e c u r r e n t sequences a s well a s new approaches to
existing p r o b l e m s .
The p r o p o s e r must submit h i s p r o b l e m with solution in legible
f o r m , preferably typed in double spacing, with the name(s) and a d d r e s s of the p r o p o s e r c l e a r l y indicated.
Solutions should be submitted within two months of the
appearance of the p r o b l e m s .
B-9
Proposed by R. L. Graham, Bell Telephone Laboratories,
Prove
î
n=2
n-1 V l "
Murray Hill,
New Jersey
= 1
OO
r-i
= 2
ZJ F - F , "
n - 1 n+1
n
n=2
where F
B-10
and
i s the nth Fibonacci number.
Proposed by Stephen Fisk, San Francisco,

tf
P r o v e the de M o i v r e - t y p e
California
identity,
/ L + ^5 F A P
L + N/5 F
/ __n
n 1 _ np
np
\
2
/
=
2
where L denotes the nth Lucas number and F denotes the nth Fibonacci number.
n
n
B-ll
Proposed by S. L. Basin, Sylvania Electronic
Defense Laboratory, Mt. View,
Show that the h y p e r g e o m e t r i c function
~ 2 2, k ,
G(x,n) = V
>
k)-' (* - 1)
k=0
ZJ
generates the sequence G ( ^ , n ) = F 0
(f' n )
(n - k - 1 ) : (2k + i ) :
85
California

B-12
Proposed
by Paul F. Byrd, San Jose State
College,
San Jose,
[April
Calif.
Show that
n+1
where
.0
is the nth Lucas number given by Lj = 1, L 2 = 3, L
n+i
and i = ^^-l .
B-13
Proposed
by S. L. Basin,
Sylvama
Electronic
Defense
Laboratory,
[At. View,
Determinants of o r d e r n which a r e of the form,
K (b.c.a) =
nv ' '
c
b
a
c
0
a
0
0
0
0
a r e known as CONTINUANTS,
P r o v e that,
- (c - Vc2 - 4~ab)n+1
(c + Vc2 - 4~ab)
K (b,c,a)
nv ' ' '
n+1
N/c2 _ 4 a b
and show, for special values of a, b, and c, that K ( b , c , a ) =
B-14
Proposed
by Maxey Brooke,
Sweeny,
Texas,
and C R. Wail,
Ft.
tiortn,
Show that
oo
\
io
n=l
10
n
89
V
^
n=l
Y\+
io
fn_
_10
"
109
^n+l
Texas
Calif.
n'
1963]
B-15

Proposed by R. B. Wallace, Beverly Hills,
College, Alaska.
Calif, and Stephen Celler,
87
University
of Alaska,
If p, i s the s m a l l e s t positive integer such that

F ^
= F
mod(10k)
for all positive n, then p, i s called the p e r i o d of the Fibonacci sequence r e l a t i v e

to 10 . Show that p,
e x i s t s for each k, and find a specific formula for p,
as a
function of k.
B - 1 6 Proposed by Marjorie Bickneil, San Jose State College,
& Space Co., Sunnyvale, Calif.
and Terry Brennan, Lockheed
Missiles
Show that if
R =
then
/F2 ,
/ n-1
R
2 F
" =
\
F
F
n-l n
F2
\ n
, F
n-1 n
F2
- F
F
n+1
n-1 n
F F _
n n+1
F2
n
2F
\
n
,
n+1
F
n+1
NOTE: On occasion t h e r e will be p r o b l e m s listed at the ends of the a r t i c l e s in the

advanced and elementary sections of the magazine, These p r o b l e m s a r e to be cons i d e r e d as logical extensions of the corresponding p r o b l e m sections and solutions for
t h e s e p r o b l e m s will be d i s c u s s e d in t h e s e sections as they a r e received.
See, for example, "Expansion of Analytic Functions In Polynomials Associated
with Fibonacci N u m b e r s , " by Paul F. Byrd ? San J o s e State College, in the first i s s u e
of the Q u a r t e r l y , and " L i n e a r R e c u r r e n c e Relations Part I, " by J a m e s J e s k e , San
J o s e State College, in this i s s u e .
Solutions for p r o b l e m s in ISSUE ONE will appear in ISSUE THREE.
88
GENERATING FIBONACCI SERIES (Cont'd, from p . 56) [April 1963]

When an e r r o r i s found, c l e a r the r e g i s t e r s and s t a r t at Step (b) with the l a s t
accurate values before the e r r o r . It i s not n e c e s s a r y to s t a r t fresh and do the whole

sequence over 0
It may be noted for any simple Fibonacci sequence starting with an odd digit s
that the odd probability of any m o s t significant digit i s 0 a 6 s and the odd probability
of any l e a s t significant digit i s 0W66. F o r a sequence starting with an even digit in
the LSD position, the odd probability of any MSD i s still 0 9 6 s but the odd probability
of any LSD is 0o0I
1<+
<<
^<
<<
<<
<<
<<
1+S
10+
2+S
30+
5+S
<<
80+
13+S
210+
34+S
550+
<<
89+S
1440+
233+S
3770+
610+S
<<
9870+
1597+S
2 5 840+
4181+S
67 6 50+
<<
10946+S
17 7 H O +
28657+S
4 6 3 6 SO +
7 5 025 +S
<<
1213 930+
196418+S
3 1 7 81lO +
514229+S
FIBONACCI SEQUENCE
8320400+
<<
1346269+S
2 17 830 90+
3524578+S
57028870+
9227465+S
<<
149 3 03 520+
24157817+S
3 90 8 816 90+
63245986+S
1023341550+
<<
165580141+S
2679142960+
433494437+S
7014087330+
1134903170+S
<<
18363119030+
2971215073+S
48075269760+
7778742049+S
125 86 2 6 902 50+
<<
20365011074+S
3 2 9512 8 00 990+
53316291173+S
862 67 57 127 20+
139583862445+S
<<
2258514337170 +
365435296162+S
5912867298790 +
9 5 6 7 2 2 0 2 6 0 41 + S
1 5 4 8 0 0 8 7 5 5 9200+
<<
2504730781961+S
4 0 5 2 7 3 9 5 3 7 8 8 10 +
6557470319842+S

1 2

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

1 2

Hochgeladen von

Copyright:

Verfügbare Formate

PART I

GENERATING FUNCTIONS FOR PRODUCTS OF POWERS

We may define the Fibonacci n u m b e r s , F . by F n = 0, F< = 1,

Intimately associated with the numbers of Fibonacci a r e the n u m b e r s of Lucas,

Here one may find further r e f e r e n c e s to the many p a p e r s

on the subject which have appeared since Leonardo P i s a n o , or Fibonacci,

introduced the famous n u m b e r s in 1202. It would be difficult to estimate how many

As p a r t of the w r i t e r ' s activity with this

Fibonacci Bibliographical P r o j e c t the subject of generating functions for powers

* R e s e a r c h supported by a June R e s e a r c h Grant from the College of A r t s

GENERATING FUNCTIONS FOR PRODUCTS OF POWERS

Golomb [ 5 J had found essentially that for s q u a r e s of Fibonacci

The very general functions studied by L u c a s , and generalized by Bell [ 1 , 2 ],

defined by (1.3) with a, b being the roots of the

quadratic equation x 2 = P x - Q so that a + b = P and ab = Q.

formulation it is easy to show that we also have the generating function

OF FIBONACCI AND LUCAS NUMBERS

Similar questions a r i s e for

GENERATING FUNCTIONS FOR LUCAS FUNCTIONS

Suppose we a r e given for any initial generating function,

with no p a r t i c u l a r r e s t r i c t i o n s on A . It follows at once from this that

and adding or subtracting, and

by this iteration build up generating functions involving F ^ and L

F(a 2 x) + F(abx) = 2 A X n a11 L

GENERATING FUNCTIONS AND PRODUCTS OF POWERS

F(a 2 x) + 2F(abx) + F(b 2 x) =

F(a%) - 2F(abx) + Ffofrl

Clearly we may proceed inductively to obtain a general r e s u l t .

In fact we may readily combine the r e l a t i o n s to obtain

Thus in principle we may set down generating

functions for products of p o w e r s of the Fibonacci and Lucas n u m b e r s , though the

The expansion analogous to (1.5) for

OF FIBONACCI AND LUCAS NUMBERS

Proceding in the s a m e manner as above, it is c l e a r that we also have

with a corresponding result lor F ,

RECIPROCALS OF FIBONACCI NUMBERS

Landau [7 ] showed that a certain s e r i e s of r e c i p r o c a l s of Fibonacci n u m b e r s

GENERATING FUNCTIONS FOR PRODUCTS OF POWERS

The device used to obtain this i s to expand (a - b)/(a

where z = (b/a) , by a power s e r i e s , and then invert the o r d e r of summation in

the s e r i e s , this being justifiable.

Dickson also notes that 2 l / F

was put in finite form by A. A r i s t a .

Let us now define in general

- \[E for ordinary Fibonacci

For R(x) as defined by (3.3) we also easily verify that

OF FIBONACCI AND LUCAS NUMBERS

To use this when A n = 1 we need to note the generating

and so for the ordinary Lucas n u m b e r s we find

Sometimes this i s stated in the equivalent form (since L0 = 2)

1 - x - xBILINEAR GENERATING FUNCTIONS

To discuss this we first intro-

duce what we shall call a general Turaii operator defined by

GENERATING FUNCTIONS FOR PRODUCTS OF POWERS

F o r the Fibonacci n u m b e r s it i s a c l a s s i c formula first discovered apparently by

a general bilinear generating function for the

*J. F(x) = Y \ *n+j

by the s a m e device and we find

a j F(a 2 x), - b J F(abx) _ a 3 F(abx) - b 3 F(b 2 x)

OF FIBONACCI AND LUCAS NUMBERS

and consequently we have

' a J F(a^x) - (a3 + b 3 ) F(abx) + b j F(b*x) _ ^

Moreover, since we have the T u r a n expression (4.2), we have

J. F(x) = Y \ n+j

I !> r " V7TF '