Cal

ulus for mathemati ians

D. J. Bernstein, University of Illinois at Chi ago, djbpobox. om

19970403, draft 5

1. Introdu tion
This booklet presents the main on epts, theorems, and te hniques of single-variable
al ulus. It diers from a typi al undergraduate real analysis text in that (1) it fo uses
purely on al ulus, not on developing topology and analysis for their own sake; (2) it's
short.

Notation and terminology. The reader must be omfortable with fun tions, not just

numbers, as obje ts of study. I use the notation x 7! x2 for the fun tion that takes x to
x2 ; thus (x 7! x2 )(3) = 9. In general f = (t 7! f (t)) for any fun tion f .

An open ball around means an interval Ball( ; h) = fx : jx j < hg for some positive
real number h. The interse tion of two open balls around is another open ball around
.
If S is a set, and f (x) is dened for all x 2 S , then f (S ) is dened as ff (x) : x 2 S g.

Part I. Continuity

2. Continuous fun tions

Denition 2.1. Let f be a fun tion dened at . Then f is ontinuous at if, for
any open ball F around f ( ), there is an open ball B around su h that f (B )  F .
In other words, if f is ontinuous at , and F is an open ball around f ( ), then there is
some h > 0 su h that f (x) 2 F for all x with jx j < h.
Example: The fun tion x 7! 3x is ontinuous|i.e., ontinuous at for every . Indeed,
Ball(3 ; ) ontains (x 7! 3x)(Ball( ; =3)), be ause jx j < =3 implies j3x 3 j < .
Another example: If f (x) = 3 for x < 2 and f (x) = 5 for x  2, then f is not ontinuous
at 2. Indeed, onsider the open ball F = Ball(5; 1). If B is any open ball around 2, then
B ontains numbers smaller than 2, so 3 2 f (B ); thus f (B ) is not ontained in F .
Theorem 2.2. Let f and g be fun tions ontinuous at . Assume that f (x) = g(x) for
all x 6= su h that f (x) and g(x) are both dened. Then f ( ) = g( ).

Proof. I will show that jf ( )

g ( ) < 2 for any  > 0. Write F = Ball(f ( ); ) and

G = Ball(g ( ); ). By ontinuity of f and g , there are balls A and B around su h that
f (A)
F and g (B )
G. Find a point x = ontained in both A and B . By onstru tion
f (x)
F and f (x) = g (x)
G, so f ( )
g ( )
f (x)
f ( ) + f (x)
g ( ) < 2 as


2

laimed.

jj

j j

3. Continuity of sums, produ ts, and ompositions

Theorem 3.1. Let f and g be fun tions ontinuous at . Dene h = f + g. Then h is
ontinuous at .

Proof. Given a ball H = Ball(h( ); ), onsider the balls F = Ball(f ( ); =2) and G =
Ball(g ( ); =2). By ontinuity of f and g , there are open balls A and B around su h
that f (A)  F and g (B )  G. Dene D = A \ B ; D is an open ball around . If x 2 D
then f (x) 2 F and g (x) 2 G so h(x) = f (x) + g (x) 2 H . Thus h(D)  H .

Theorem 3.2. Let

ontinuous at .

Proof. Dene
H

L
Ball(LM; ),

and

be fun tions ontinuous at . Dene

f g.

Then

is

= f ( ) and M = g ( ), so that LM = h( ). Given an open ball

I will nd an open ball D around so that h(D)  H .

If L = M = 0, take the interse tion of open balls where jf (x)j <  and jg (x)j < 1. Then
jh(x)j < .

6 0, take the interse tion of open balls where jf (x)j < =(2 jM j) and
=
j
j j j. Then jg(x)j < 2 jM j so jh(x)j < . Similarly if L 6= 0 and M = 0.
If L 6= 0 and M =
6 0, take the interse tion of open balls where jf (x) Lj < =(4 jM j),
jg(x) M j < =(2 jLj), and jg(x) M j < jM j. Then jg(x)j < 2 jM j so jh(x) LM j =
jg(x)(f (x) L) + L(g(x) M )j < 2 jM j (=(4 jM j)) + jLj (=(2 jLj)) = .

If

L = 0 and M
g (x)
M < M

Theorem 3.3. Let g be a fun tion ontinuous at . Let f be a fun tion ontinuous at
Dene h = (x 7! f (g(x))). Then h is ontinuous at .
For example, x 7! os 2x is ontinuous, sin e x 7! 2x and y 7! os y are ontinuous,

g ( ).

Proof. Let F be an open ball around h( ) = f (g( )). By ontinuity of f , there is some
open ball G around g ( ) with f (G)  F . By ontinuity of g , there is some open ball B
around with g (B )  G. Finally h(B ) = f (g (B ))  f (G)  F .


4. Continuity of simple fun tions

Theorem 4.1. x 7! b is ontinuous at , for any b and .
Proof. Ball(b; h) ontains (x 7! b)(D) for any open ball D.

Theorem 4.2. x 7! x is ontinuous at , for any .

Proof. Ball( ; h) ontains (x 7! x)(Ball( ; h)).

By Theorems 3.2 and 4.2, x 7! x2 is ontinuous; x 7! x3 is ontinuous; in general

x 7! xn is ontinuous for any positive integer n. Thus, by Theorems 3.1, 3.2, and 4.1,
any polynomial fun tion x 7! 0 + 1 x +


+ n xn is ontinuous.
2

The fun tion x 7! 1=x is ontinuous at for 6= 0. (It's not even dened at 0, so it
an't be ontinuous there.) By Theorem 3.3, x 7! 1=f (x) is ontinuous whenever f is
ontinuous and nonzero. For example, x 7! xn is ontinuous ex ept at 0 when n is a
negative integer.
Part II. Derivatives

5. Dierentiable fun tions

Denition 5.1. Let f be a fun tion dened at . Then f is dierentiable at if there

is a fun tion f1 , ontinuous at , su h that f = (x 7! f ( ) + (x

)f1 (x)).

Denition 5.2. Let f be a fun tion dened at . Then f has derivative d at if there

is a fun tion f1 , ontinuous at , su h that f = (x 7! f ( ) + (x )f1 (x)) and f1 ( ) = d.

By Theorem 2.2, there is at most one ontinuous fun tion f1 satisfying f1 (x) = (f (x)
f ( ))=(x
) for all x 6= , so f has at most one derivative at , alled the derivative
of f at . The derivative of f at is written f 0 ( ). The derivative of f , written f 0 , is
the fun tion 7! f 0 ( ).
For example, onsider the fun tion f = (x 7! x2 ). Here f (x) = f (3) + (x 3)f1 (x) with
f1 = (x 7! x + 3). The fun tion f1 is ontinuous at 3, so f is dierentiable at 3; its
derivative at 3 is f1 (3) = 6. In general f 0 ( ) = 2 .

Theorem 5.3. Let f be a fun tion. If f is dierentiable at then f is ontinuous at .

Proof. By denition of dierentiability, there is a fun tion
f

= (x 7! f ( ) + (x

)f1 (x)).

f1 , ontinuous at , with
Apply Theorems 3.1, 3.2, 4.1, and 4.2.


6. Derivatives of sums, produ ts, and ompositions

Theorem 6.1. Let f and g be fun tions. Dene h = f + g. If f and g are dierentiable
at then h is dierentiable at . Furthermore h0 ( ) = f 0 ( ) + g0 ( ).
In short (f + g )0 = f 0 + g 0 if the right side is dened. This is the sum rule.

Proof. Say f (x) = f ( ) + (x

)f1 (x) and g (x) = g ( ) + (x

)g1 (x) with f1 and
ontinuous at . Dene h1 = f1 + g1 ; then h1 is ontinuous at by Theorem 3.1,
and h(x) = h( ) + (x )h1 (x), so h is dierentiable at . Finally h0 ( ) = h1 ( ) =
f1 ( ) + g1 ( ) = f 0 ( ) + g 0 ( ).

g1

Theorem 6.2. Let f and g be fun tions. Dene h = f g. If f and g are dierentiable

at then h is dierentiable at . Furthermore h0 ( ) = f 0 ( )g( ) + f ( )g0 ( ).
In short (f g )0 = f 0 g + f g 0 if the right side is dened. This is the produ t rule.
3

Proof. Say f (x) = f ( ) + (x

)f1 (x) and g (x) = g ( ) + (x

)g1 (x) with f1 and g1
ontinuous at . Then h(x) = h( ) + (x )h1 (x) where h1 (x) = f1 (x)g (x) + f ( )g1 (x).
This fun tion h1 is ontinuous at by Theorems 3.1, 3.2, 4.1, and 5.3, so h is dierentiable
at , with derivative h1 ( ) = f1 ( )g ( ) + f ( )g1 ( ) = f 0 ( )g ( ) + f ( )g 0 ( ).

Theorem 6.3. Let f and g be fun tions. Dene h = (x 7! f (g(x))). If g is dierentiable

at , and f is dierentiable at g( ), then h is dierentiable at . Furthermore h0 ( ) =

f 0 (g ( ))g 0 ( ).
In short (f g )0 = (f 0 g )g 0 if the right side is dened. This is the hain rule.

Proof. Write b = g( ). Say f (x) = f (b)+(x b)f1 (x) and g(x) = b +(x )g1 (x) with f1

ontinuous at b and g1 ontinuous at . Now h(x) = f (g (x)) = f (b)+(g (x) b)f1 (g (x)) =
f (b)+(x )g1 (x)f1 (g (x)). Thus h(x) = h( )+(x )h1 (x) where h1 (x) = g1 (x)f1 (g (x)).
Finally h1 is ontinuous at by Theorems 3.3, 3.2, and 5.3, so h is dierentiable at ,
with derivative h1 ( ) = g1 ( )f1 (g ( )) = g 0 ( )f 0 (g ( )).


7. Derivatives of simple fun tions

A onstant fun tion, su h as x 7! 17, has derivative 7! 0, sin e 17 = 17 + (x )0.
The identity fun tion x 7! x has derivative 7! 1, sin e x = + (x )1.
In general, for any positive integer n, the fun tion x 7! xn has derivative 7! n n 1 ,
sin e xn = n + (x )(xn 1 + xn 2 +


+ n 1 ).
The fun tion x 7! 1=x, dened for nonzero inputs, has derivative 7! 1= 2 . Indeed,
1=x = 1= + (x )( 1= x), and x 7! 1= x is ontinuous at with value 1= 2 .
Now the hain rule, with f = (x 7! 1=x), states that 1=g has derivative g 0 =g 2 at any
point where g ( ) =
6 0. In parti ular, for any negative integer n, x 7! xn has derivative
n
1
7! n
.
Finally, the produ t rule implies that h=g has derivative (gh0
where g ( ) 6= 0; this is the quotient rule.

hg 0 )=g 2

at any point

Part III. Completeness and its onsequen es

8. Completeness of the real numbers

Denition 8.1. Let S be a set of real numbers. A real number is an upper bound
for S if x  for all x 2 S .

For example, any number   is an upper bound for the set f3; 3:1; 3:14; 3:141; : : :g.
The smallest upper bound is  .
The real numbers are omplete: if S is a nonempty set, and there is an upper bound
for S , then there is a smallest upper bound for S . The smallest upper bound is unique;
it is alled the supremum of S , written sup S .
4

9. The intermediate-value theorem

Theorem 9.1. Let f be a ontinuous real-valued fun tion. Let y be a real number. Let

 be real numbers with f (b)  y  f ( ). Then f (x) = y for some x 2 [b; .

Here [b; means fx : b  x  g. For example, if f (3) = 5 and f (4) = 7, and f

ontinuous, then f must have a root between 3 and 4.

Proof. Dene S = fx 2 [b; : f (x)  yg.

is

is nonempty, be ause it ontains b, and it

has an upper bound, namely , so it has a smallest upper bound, say u.
S

Suppose f (u) > y . By ontinuity, there is an open ball D around u su h that f (x) > y
for x 2 D. Pi k any t 2 D with t < u. If x 2 [t; u then x 2 D so f (x) > y so x 2= S .
Thus t is an upper bound for S |but u is the smallest upper bound. Contradi tion.

Suppose f (u) < y . Then u = so u < . By ontinuity, there is an open ball D around
u su h that f (x) < y for x
D . Pi k any x
D with u < x < ; then f (x) < y . But
x = S sin e u is an upper bound for S ; so f (x) > y . Contradi tion.

10. The maximum-value theorem

Theorem 10.1. Let f be a ontinuous real-valued fun tion. Let b  be real numbers.

Then there is an upper bound for f ([b; ).

Proof. Let S be the set of x 2 [b; su h that f ([b; x) is bounded|i.e., has an upper

bound.

is nonempty, be ause it ontains b. Dene u = sup S .

By ontinuity, there is an open ball D around u su h that f (D)  Ball(f (u); 1). Sele t
2 with t < u; then t is not an upper bound for S , so there is some x 2 S with
 u. Now f ([b; x) and f ([x; u)  f (D) are bounded, so f ([b; u) is bounded.

t
D
t<x

Suppose u < . Sele t v 2 D with u < v < . Then f ([u; v ) is bounded, so

Contradi tion. Hen e u = , and f ([b; ) = f ([b; u) is bounded.

2 S.


Theorem 10.2. Let f be a ontinuous real-valued fun tion. Let b  be real numbers.
Then there is some u 2 [b; su h that, for all z 2 [b; , f (u)  f (z ).
This is the maximum-value theorem: a ontinuous fun tion on a losed interval
a hieves a maximum. The same is not true for open intervals: onsider 1=x for 0 < x < 1.
Proof. By Theorem 10.1, there is an upper bound for f ([b; ). Dene M = sup f ([b; ).

Let S be the set of x 2 [b; su h that sup f ([x; ) = M . Then b 2 S . Dene u = sup S .

Suppose f (u) < M . By ontinuity there is an open ball D around u su h that f (D) 
Ball(f (u); (M f (u))=2); then sup f (D) < M . Sele t t 2 D with t < u; then t is not an
upper bound for S , so there is some x 2 S with t < x  u. Then sup f ([x; ) = M , but
sup f ([x; u) < M , so u < . Sele t v 2 D with u < v < . Then sup f ([x; v ) < M , so
sup f ([v; ) = M , so v 2 S . Contradi tion. Hen e f (u) = M = sup f ([b; ).

5

Theorem 10.3. Let f be a ontinuous real-valued fun tion. Let b  be real numbers.

Then there is some u 2 [b; su h that, for all x 2 [b; , f (u)  f (x).

Proof. Apply Theorem 10.2 to

f.

Part IV. The mean-value theorem

11. Fermat's prin iple

Theorem 11.1. Let f be a real-valued fun tion dierentiable at t. Assume that f (t) 

f (x)

for all x in an open ball B around t. Then f 0 (t) = 0.

Proof. By assumption f (x) = f (t) + (x

t)f1 (x)

where f1 is ontinuous at t. Suppose

f1 (t) > 0. Then f1 (x) > 0 for all x in an open ball D around t. Pi k x > t in both B
and D; then f (t)  f (x) = f (t) + (x t)f1 (x) > f (t). Contradi tion. Thus f1 (t)  0.
Similarly f1 (t)  0. Hen e f 0 (t) = f1 (t) = 0.


Theorem 11.2. Let f be a real-valued fun tion dierentiable at t. Assume that f (t) 

f (x)

for all x in an open ball B around t. Then f 0 (t) = 0.

Proof. Apply Theorem 11.1 to

f.

12. Rolle's theorem

Theorem 12.1. Let f be a dierentiable real-valued fun tion. Let b < be real numbers.
If f (b) = f ( ) then there is some x with b < x < su h that f 0 (x) = 0.

Proof. By Theorem 10.2, there is some t 2 [b; su h that f 's maximum value on [b;

is a hieved at t. If f (t) > f (b) then t 6= b and t 6= , so there is an open ball B around t
su h that B  [b; . By Theorem 11.1, f 0 (t) = 0.

Similarly, by Theorem 10.3, there is some u 2 [b; su h that f a hieves its minimum at
If f (u) < f (b) then f 0 (u) = 0 as above.

u.

The only remaining ase is that f (t)  f (b) and f (u)  f (b). Then f (b) is both the
maximum and the minimum value of f on [b; ; i.e., f is onstant on [b; . Hen e
f 0 (x) = 0 for any x between b and .


13. The mean-value theorem

Theorem 13.1. Let f be a dierentiable real-valued fun tion. Let b < be real numbers.
Then there is some x with b < x < su h that f ( )
6

f (b)

= f 0 (x)(

b).

This is the mean-value theorem. The terminology \mean value" omes from the fundamental theorem of al ulus, whi h an be interpreted as saying that (f ( ) f (b))=( b)
is the average (\mean") value of f 0 (x) for x 2 [b; . See Theorem 16.1.

Proof. Dene g(x) = (

b)f (x)
(x b)(f ( ) f (b)). Then g is dierentiable, and
g (b) = ( b)f (b) = ( b)f ( ) ( b)(f ( ) f (b)) = g ( ). By Theorem 12.1, g 0 (x) = 0
for some x between b and . But g 0 (x) = ( b)f 0 (x) (f ( ) f (b)).

Theorem 13.2. Let

be a dierentiable real-valued fun tion. If f 0 (x) = 0 for all

then f is onstant.
More generally, two fun tions with the same derivative must dier by a onstant.

Proof. Pi k any real numbers

f ( )
f (b) = f 0 (x)(

b)

= 0,

b < . By Theorem
so f ( ) = f (b).

13.1, there is some

su h that

Part V. Integration

14. Tagged divisions and gauges

Denition 14.1. Let b  be real numbers. Let x0 ; x1 ; : : : ; xn and t1 ; : : : ; tn be real
numbers. Then x0 ; t1 ; x1 ; : : : ; tn ; xn is a tagged division of [b; if b = x0  t1  x1 

t2




 xn 1  tn  xn = .

The idea is that [b; is divided into the intervals [x0 ; x1 ; [x1 ; x2 ; : : : ; [xn 1 ; xn ; in
ea h interval [xk 1 ; xk there is a tag tk . For example, onsider the tagged division
0; 1; 4; 5; 6; 6; 7 of [0; 7; here the intervals are [0; 4; [4; 6; [6; 7, with tags 1; 5; 6 respe tively.

Denition 14.2. Let b  be real numbers. A gauge on [b; is a fun tion assigning

to ea h point t 2 [b; an open interval ontaining t.

For example, given  > 0, the fun tion t 7! Ball(t; ) is a gauge on any interval.

Denition 14.3. Let b  be real numbers. Let G be a gauge on [b; . A tagged division
of [b; is inside G if [xk 1 ; xk  G(tk ) for every k.

x0 ; t1 ; x1 ; : : : ; tn ; xn

Theorem 14.4. Let b  be real numbers. Let G be a gauge on [b; . Then there is a
tagged division of [b; inside G.

Proof. Let S be the set of x 2 [b; su h that there is a tagged division of [b; x inside

is nonempty: b; b; b is a tagged division of [b; b inside G, so b 2 S . Also,

upper bound for S . Thus there is a smallest upper bound for S , say y .
G. S

is an

Sele t v 2 G(y ) su h that v < y . Then v is not an upper bound for S , so there is some
x > v with x 2 S . Let x0 ; t1 ; x1 ; : : : ; tn ; xn be a tagged division of [b; x inside G.
7

Suppose y

< . Pi k z
G(y ) with y < z
. Then [xn ; z = [x; z
[v; z
x0 ; t1 ; x1 ; : : : ; tn ; xn ; y; z is a tagged division of [b; z inside G. Thus z
S;
and y is an upper bound for S . Contradi tion.

 G(y), so
but

z > y,

Thus y = . Finally x0 ; t1 ; x1 ; : : : ; tn ; xn ; y; y is a tagged division of [b; inside G.

15. The denite integral

Denition 15.1. Let b  be real numbers. Let

x0 ; t1 ; x1 ; : : : ; tn ; xn
fun tion dened on [b; . The
x0 )f (t1 ) +
+ (xn xn 1 )f (tn ).

be a tagged di-

be a
Riemann sum for f on
x0 ; t1 ; x1 ; : : : ; tn ; xn is (x1




For example, the Riemann sum for f on 0; 1; 4; 5; 6; 6; 7 is (4 0)f (1) + (6 4)f (5) + (7
6)f (6). This may be visualized as the sum of areas of three re tangles: one stret hing
from 0 to 4 horizontally with height f (1), another from 4 to 6 with height f (5), and
another from 6 to 7 with height f (6).
vision of [b; . Let

Denition 15.2. Let b  be real numbers. Let f be a fun tion dened on [b; . Let I
be a number. Then f has integral I on [b; if, for every open ball E around I , there is

a gauge G on [b; su h that E ontains the Riemann sum for f on any tagged division
of [b; inside G.

Theorem 15.3. Let b  be real numbers. Let f be a fun tion. If f has integral I on
[b; and

has integral J on [b; then I = J .

Thus there is at most one number I su h that f has integral
R I on [b; . If this number
exists, it is alled the integral of f from b to , written b f .
f

Proof. I will show that jI

J < 2

for any  > 0.

By denition of integral, there is a gauge G on [b; su h that Ball(I; ) ontains the

Riemann sum for f on any tagged division of [b; inside G.
Similarly, there is a gauge H on [b; su h that Ball(J; ) ontains the Riemann sum for
f on any tagged division of [b; inside G.
Dene F (t) as the interse tion of G(t) and H (t). Then F is a gauge on [b; . By Theorem
14.4, there is a tagged division x0 ; : : : ; xn of [b; inside F .
Let R be the Riemann sum for f on x0 ; : : : ; xn . Observe that x0 ; : : : ; xn is inside both G
and H , so R 2 Ball(I; ) and R 2 Ball(J; ). Hen e jI J j  jI Rj + jR J j < 2. 

16. The fundamental theorem of al ulus

Theorem 16.1.
R Let

f ( )

f (b)

0
b f .

be a dierentiable fun tion. Let

8

 be real numbers.

Then

Proof. Pi k  > 0. I will onstru t a gauge G su h that Ball(f ( )

f (b); (

ontains the Riemann sum for f 0 on any tagged division of [b; inside G.

b + 1))

Fix t 2 [b; . Sin e f is dierentiable at t, there is a fun tion f1 , ontinuous at t,

su h that f (x) = f (t) + (x t)f1 (x). By denition of ontinuity, f1 (x) is within  of
f1 (t) = f 0 (t) for all x in some open ball around t. Dene G(t) as the union of all su h
balls. Then G is a gauge on [b; .

Observe that if x; y 2 G(t), with x  t  y , then (y x)f 0 (t) is within (y x) of

f (y ) f (x). Indeed, jf1 (x)
f 0 (t)j <  by denition of G, and f (x) f (t) = (x t)f1 (x),
so
jf (x) f (t) (x t)f 0 (t)j = j(x t)(f1 (x) f 0 (t))j   jx tj :
Similarly jf (y ) f (t) (y t)f 0 (t)j   jy tj. Thus jf (y ) f (x) (y x)f 0 (t)j 
(jy
tj + jx
tj); and jy
tj + jx tj = y
x.

Finally, say x0 ; t1 ; x1 ; : : : ; tn ; xn is a tagged division of [b; inside G. Then xk 1 ; xk 2

G(tk ), with xk 1  tk  xk , so (xk xk 1 )f 0 (tk ) is within (xk xk 1 ) of f (xk ) f (xk 1 )
as above. Thus the Riemann sum for f 0 on x0 ; t1 ; x1 ; : : : ; tn ; xn is within

1kn

(xk

of

1kn

xk 1 )

(f (xk )

= (xn

f (xk 1 ))

x0 )

= (

= f (xn )

b) < (

f (x0 )

= f ( )

b + 1)

f (b)

as laimed.

17. Integration rules

Theorem 17.1. Let

f
[b;

has integral aI on
R
R
In short b af = a b f

be a fun tion. Let b  be real numbers. If

for any real number a.
if the right side is dened.

b f

then

af

0. Sin e b f = I , there is a gauge G on [b; su h that Ball(I; )

ontains the Riemann sum for f on any tagged division of [b; inside G. The Riemann
sum for af is exa tly a times the Riemann sum for f , so it is inside Ball(aI; jaj ) for
a 6= 0 or Ball(0; ) for a = 0.


Proof. Pi k

 >

Theorem
17.2. Let f and g be fun tions. Let b  be real numbers. If
R

b g

= J then

In short

b (f

+ g has integral I + J on [b; .

+ g) =

b f

b g

b f

= I and

if the right side is dened.

Proof. Pi k  > 0. There is a gauge F on [b; su h that Ball(I; ) ontains the Riemann

sum for f on any tagged division of [b; inside F ; and there is a gauge G on [b; su h
that Ball(J; ) ontains the Riemann sum for g on any tagged division of [b; inside G.
9

Dene H (t) = F (t) \ G(t). Then H is a gauge on [b; . If x0 ; : : : ; xn is a tagged division

of [b; inside H , then x0 ; : : : ; xn is also inside both F and G, so the Riemann sums for
f and g on x0 ; : : : ; xn are within  of I and J respe tively; thus the Riemann sum for
f + g on x0 ; : : : ; xn is within 2 of I + J .


RTheorem 17.3. Let f be a fun tion. Let a  b  be real numbers. If

b f

= J then

In short

af

has integral I + J on [a; .

Rb

af

b f

Rb

af

and

if the right side is dened.

Proof. Pi k  > 0. There is a gauge G on [a; b su h that Ball(I; ) ontains the Riemann

sum for f on any tagged division of [a; b inside G; there is a gauge H on [b; su h that
Ball(J; ) ontains the Riemann sum for f on any tagged division of [b; inside H .
I dene a new gauge as follows. For t < b dene F (t) = fx 2 G(t) : x < bg. For t =
dene F (t) = G(t) \ H (t). For t > b dene F (t) = fx 2 H (t) : x > bg.

Say x0 ; : : : ; xn is a tagged division of [a; inside F . Then b 2 [xk 1 ; xk  F (tk ) for

some k; by onstru tion of F , tk must equal b. Now x0 ; t1 ; x1 ; : : : ; xk 1 ; tk ; b is a tagged
division of [a; b inside F , hen e inside G. Thus the Riemann sum (x1 x0 )f (t0 ) +



+ (b xk 1 )f (tk ) is within  of I . Similarly the Riemann sum (xk b)f (tk ) +


+
(xn xn 1 )f (tn ) is within  of J . Add: the Riemann sum (x1 x0 )f (t0 ) +


+ (xk
xk 1 f (tk ) +


+ (xn
xn 1 )f (tn ) is within 2 of I + J .


TheoremR 17.4. Let f be a fun tion. Let b  be real numbers. If f is nonnegative on

[b; and

b f

Proof. Pi k

= I then

is nonnegative.

0. Sele t an appropriate gauge G. By Theorem 14.4, there is an

appropriate tagged division of [b; . The orresponding Riemann sum is nonnegative, so
I  .

 >

Part VI. Limits

18. Convergen e and limits

Denition 18.1. Let f be a fun tion. Then f onverges to L at if the fun tion
x

7!

L
f (x)

if x =
if x 6=

is ontinuous at .
Equivalent terminology: f (x) onverges to L as x approa hes .
By Theorem 2.2, there is at most one number L su h that f onverges to L at . If this
number exists, it is alled the limit of f at , or the limit of f (x) as x approa hes
, written limx! f (x). Note that f is ontinuous if and only if limx! f (x) = f ( ).
10

Example: os(1=x) does not onverge to 0 as x approa hes 0.

19. Limits of sums, produ ts, and ompositions

Theorem 19.1. Let

and g be fun tions. If limx! f (x) = L and limx! g(x) = M

then f (x) + g(x) onverges to L + M as x approa hes .
In short limx! (f (x) + g (x)) = limx! f (x) + limx! g (x) if the right side is dened.
f

Proof. Repla e f ( ) by L and g( ) by M to obtain new fun tions a and b. Then a and

are ontinuous, so a + b is ontinuous by Theorem 3.1.

Theorem 19.2. Let f and g be fun tions. If limx! f (x) =

then f (x)g(x) onverges to LM as x approa hes .

and limx! g(x) =

Proof. Theorem 3.2.

Theorem 19.3. Let f and g be fun tions. If limx! g(x) = L, and f is ontinuous at

L,

then f (g(x)) onverges to f (L) as x approa hes .

In short limx! f (g (x)) = f (limx! g (x)) if the right side is dened, provided that f is
ontinuous.

Proof. Theorem 3.3.

20. L'H^opital's rule

Theorem 20.1. Let f and g be real-valued fun tions dierentiable at . If f ( ) = g( ) =

0, and g 0 ( ) 6= 0, then f (x)=g (x) onverges to f 0 ( )=g 0 ( ) as

approa hes .
For example, limx!0 (x= sin x) = 1=1 = 1, sin e sin0 = os and os 0 = 1 6= 0.
x

Proof. By assumption f (x) = f ( ) + (x

)f1 (x) = (x )f1 (x) where f1 is ontinuous

at . Similarly g (x) = (x )g1 (x) where g1 is ontinuous at . By assumption g1 ( ) =
g 0 ( ) = 0, so the fun tion x
f1 (x)=g1 (x) is ontinuous at , with value f1 ( )=g1 ( ).
Finally f (x)=g (x) = f1 (x)=g1 (x) for x = .

7!

Theorem 20.2. Let f and g be dierentiable real-valued fun tions. If f ( ) = g( ) = 0,

and limx! (f 0 (x)=g0 (x)) = L, then f (x)=g(x) onverges to L as x approa hes .

Proof. Fix a ball E around L. There is a ball D around su h that f 0 (x)=g0 (x) 2 E
for all x 2 D with x 6= . In parti ular, g 0 (x) is nonzero for x 2 D. By Theorem 12.1,
g (y ) is nonzero for y 2 D .
I will show that f (y )=g (y ) 2 E for all y 2 D with y 6= . Thus f (y )=g (y ) onverges to L
as y approa hes .
11

Given y 2 D, y 6= , onsider the fun tion h = (x 7! f (x)g (y )

h is dierentiable, with h0 (x) = f 0 (x)g (y )
f (y )g 0 (x).

f (y )g (x)).

Noti e that

Now h( ) = f ( )g (y ) f (y )g ( ) = 0, and h(y ) = f (y )g (y ) f (y )g (y ) = 0, so there is

some x between and y with h0 (x) = 0 by Theorem 12.1. Thus f 0 (x)g (y ) = f (y )g 0 (x).
Both g 0 (x) and g (y ) are nonzero, so f (y )=g (y ) = f 0 (x)=g 0 (x) 2 E .

Theorem 20.2 may be used repeatedly. For example:
lim

x!0

os x

x2

sin x
os x 1
= lim
= :
x!0 2x
x!0 2
2

= lim

99. Expository notes

Common pra ti e in al ulus books is to dene ontinuity using limits. I dene limits
using ontinuity; ontinuity is a simpler on ept.
\An open ball around " is substantially easier to read than \for some h > 0, the set of
su h that jx j < h."

I use Caratheodory's denition of the derivative of f . The point is to give a name to

the fun tion x 7! (f (x) f ( ))=(x ). I learned about this from an arti le by Stephen
Kuhn in the Monthly. It's also used in the se ond edition of Apostol's text.
My proof of Theorem 6.2 uses the formula h1 (x) = f1 (x)g (x)+ f ( )g1 (x), whi h is shorter
than the (more obvious) formula h1 (x) = f1 (x)g ( ) + f ( )g1 (x) + (x )f1 (x)g1 (x). I
was reminded of this simpli ation by a letter in the Monthly from Gunter Pi kert.
The Heine-Borel theorem follows immediately from Theorem 14.4. See Botsko's 1987
Monthly arti le for this approa h to all the basi ompleteness theorems. Thanks to Joe
Buhler for the referen e.
I follow the Kurzweil-Hensto k approa h to integration. The resulting integral is more
general than the Lebesgue integral; it is equivalent to the integrals onstru ted by Denjoy
and Perron. There is no need for any te hni al onditions in the fundamental theorem
of al ulus, Theorem 16.1; every derivative is integrable. I learned about this from
advertisements by Robert G. Bartle in the Bulletin and the Monthly.

12