CM221A

ANALYSIS I

NOTES ON WEEK 8

DIFFERENTIATION
Let f be a function defined on an interval I. We say that f 0 (x) = k if


 f (x) f (y)


> 0 x, > 0 : 0 < |x y| < x, 
k  < .
xy
Equivalently, putting y = x + ,
> 0 x, > 0 : 0 < || < x,
or



 f (x + ) f (x)


k  < ,

f (x) f (y)
= k.
yx
xy
lim

The number k is called the derivative of f at the point x. If f 0 (x) exists for every
point x from the interval, we can consider f 0 as a function of the variable x. This
d
function is called the derivative of f . Another notation for the derivative is dx
f (x).
It may well happen that the limit does not exist, in which case we say that f is
not differentiable at x.
The right and left limits
f (x) f (y)
yx+0
xy
lim

and

f (x) f (y)
yx0
xy
lim

(if they exist) are said to be the right and left derivatives of f at the point x. The
function f is differentiable if both the right and left derivative exist and have the
same value. If x is an end point of the interval I then one can speak only about
one of these derivatives (the other limit does not make sense).
Theorem. If the derivative f 0 (x) exists then f is continuous at x.
Proof.
If f is differentiable at x then there exists a number x,1 > 0 such that

 f (x)f (y)

 xy k  < 1 whenever 0 < |x y| < x,1 (because we can take = 1 in the




 
 f (x)f (y)
 f (x)f (y) 
  f (x)f (y) 
definition). Since  xy k  >  xy  |k|, this implies that  xy  <
|k| + 1. Multiplying both parts of this inequality by |x y|, we see that
|f (x) f (y)| < (|k| + 1)|x y| whenever 0 < |x y| < x,1 .
Let us fix an arbitrary > 0 and choose > 0 so small that < x,1 and <
((|k| + 1))1 . Then, by the above, we have |f (x) f (y)| < whenever |x y| < .
This means that f is continuous at x.
1

The converse is not true; a continuous function may not be differentiable.

Example. The function f (x) = |x| is continuous at the point x = 0 because
|f (0) f (y)| = |y| 0 as y 0. However,
 


  ||

 f (0 + ) f (0)



k  =  k  .


If < 0 then the right hand side is equal to |k + 1|. If > 0 then the right hand
side is equal to |1 k|. Clearly, for all sufficiently small at least one of these
numbers is greater than . Therefore the derivative at x = 0 does not exist.
Theorem. Let f and g be differentiable functions. Then
+ g(x)) = f 0 (x) + g 0 (x);

(a)

d
(f (x)
dx

(b)

d
(f (x)g(x))
dx

(c)

d
dx

f (x)
g(x)

= f 0 (x)g(x) + f (x)g 0 (x);

f 0 (x)g(x)f (x)g 0 (x)

g 2 (x)

provided that g(x) 6= 0.

Proofs.


(y)+g(y))
(y)
(y)
(a) limyx f (x)+g(x)(f
= limyx f (x)f
+ f (x)f
xy
xy
xy
(y)
= limyx f (x)f
+ limyx g(x)g(y)
= f 0 (x) + g 0 (x)
xy
xy

because the limit of the sum is equal to the sum of limits.



f (x)g(x)f (y)g(y)
(f (x)f (y))g(x)
f (y)(g(x)g(y)
(b) limyx
= limyx
+
xy
xy
xy
(y))g(x)
limyx (f (x)f
+ limyx f (y)(g(x)g(y))
xy
xy 




f (x)f (y)
+ limyx f (y) limyx g(x)g(y)
= limyx g(x) limyx xy
xy
= f 0 (x)g(x) + f (x)g 0 (x)
because the limits of the products are equal to the products of limits and
limyx f (y) = f (x) as f is continuous.
f (x)g 1 (x)f (y)g 1 (y)
(y)g(x)
= limyx f (x)g(y)f
xy
(xy)g(x)g(y)
(x)(g(x)g(y))
= limyx (f (x)f (y))g(x)f

 (xy)g(x)g(y)

(y))g(x)
f (x)(g(x)g(y))
1
limyx g(x)g(y)
limyx (f (x)f

lim
yx
(xy)
(xy)
f 0 (x)g(x)f (x)g 0 (x)
=
g 2 (x)

(c) limyx

because limyx

1
g(x)g(y)

1
limyx g(x)g(y)

1
g 2 (x)

as g is continuous.

Example. The derivative of a constant function is identically equal to zero because

in this case f (x) f (y) = 0 for all x and y. This fact and the part (b) imply that
d
(c g(x)) = c g 0 (x) for any constant c.
dx
2

y
d n
Example. If n is a positive integer then dx
x = limxy xxy
. Substituting the
n
n
n1
n2
n2
n1
expansion x y = (x y)(x
+ x y + + xy
+ y ) (see Sheet 1) and
passing to the limit, we obtain

d n
x = lim (xn1 + xn2 y + + xy n2 + y n1 )
yx
dx
= (xn1 + xn2 x + + xxn2 + xn1 ) = nxn1 .

Example.

d x
e
dx

Proof. We have

= ex .
d x
e
dx

= limyx

ex ey
xy

= ex limyx

eyx 1
yx
z

ez 1
. Recall
z
z2
z3
+ 3! + . . . It
2!

= ex limz0

that the exponential function is given by the series e = 1 + z +

follows that
z
z2
ez 1
= 1+ +
+ = 1 + z F (z)
z
2! 3!
2
where F (z) = 2!1 + 3!z + z4! + . . . The ratio test shows that the series in the right
hand side is absolutely convergent for all z. If |z| < 1 then |F (z)| 6 2!1 + 3!1 + 4!1 +. . .
because the modulus of each partial sum of the series defining F is estimated by
the corresponding partial sum of the series on the right hand side. Thus F is
1
1
1
implies
estimated
 ez 1 by the
 constant C = 2! + 3! + 4! + . . . whenever |z| < 1. eThis

z 1



that z 1 < C |z| for all z with |z| < 1. It follows that limz0 z 1 = 0
z
and, consequently, limz0 e z1 = 1.
Example.

d
dx

sin x = cos x and

d
dx

cos x = sin x.

Proof. The trigonometric formulae






x+y
xy
sin x sin y = 2 cos
sin
,
2
2




x+y
xy
cos x cos y = 2 sin
sin
2
2
imply that
x+y
2

d
cos x = sin x lim
z0
dx

sin x sin y
d
sin x = lim
= lim cos
yx
yx
dx
xy


lim

yx


xy
2


sin z
= cos x lim
z0
z

2
sin
xy

and, similarly,

Since sin z = z

z3
3!

z5
5!

z7
7!

sin z
z


.

+ . . . , we have sin z = z + z 2 F (z) , where

F (z) =

z
z3 z5
+

+ ...
3! 5!
7!
3

By the ratio test, the series in the right hand side is absolutely convergent for all
z and y. If |z| < 1 then
|F (z)| 6 C =

1
1
1
+ + + ...
3! 5! 5!
2

It follows that |sin z z| = |z 2 F (z)| 6 C |z|2 . Therefore



 sin z

lim 
1 = lim z 1 |sin z z| = 0
z0
z0
z
and, consequently, limz0

sin z
z

= 1.

Chain rule (differentiating a function of a function). If f (x) is differentiable

at x = c and g(y) is differentiable at y = f (c) then g(f (x)) is differentiable at x = c
and
d
g(f (x))|x=c = g 0 (f (c))f 0 (c).
dx

Proof. By definition,


g(f (x)) g(f (c)) f (x) f (c)
g(f (x)) g(f (c))
= lim
lim
xc
xc
xc
f (x) f (c)
xc





g(f (x)) g(f (c))
f (x) f (c)
g(f (x)) g(f (c))
lim
lim
= lim
f 0 (c) .
xc
xc
xc
f (x) f (c)
xc
f (x) f (c)
Since f is continuous at c, we have f (x) f (c) as x c. Therefore
g(f (x)) g(f (c))
=
xc
f (x) f (c)
lim

g(f (x)) g(f (c))

= g 0 (f (c)) .
f (x)f (c)
f (x) f (c)
lim

Example. Let f (x) = ln x where x > 0. By definition, g(f (x)) = x for all x
where g(y) = ey . Since (ey )0 = ey , differentiation this identity and applying the
d ln x
previous theorem, we obtain dx
e
= (ln x)0 eln x = (ln x)0 x = 1. Thus we have
(ln x)0 = x1 .