4 Integral calculus

As indicated at the start of the previous chapter, the differential calculus and its
complement, the integral calculus, together form the most widely used tool for the
analysis of physical systems. The link that connects the two is that they both deal with
the effects of vanishingly small changes in related quantities; one seeks to obtain the
ratio of two such changes, the other uses such a ratio to calculate the variation in one of
the quantities resulting from a change in the other.
Any change in the value of any one property (or variable) of a physical system almost
always results in the values of some or all of its other properties being altered; in
general, the size of each consequential change depends upon the current values of all of
the variables. As a result, during a finite change in any one of the values, that of x say,
those associated with all of the other variables are continuously changing, making
computation of the final situation difficult, if not impossible. The solution to this
difficulty is provided by the integral calculus, which allows only vanishingly small
changes, and, after any such change in one variable, brings all the other values ‘up to
date’ (by infinitesimal amounts) before allowing any further change. By notionally
carrying through an infinitely large number of these infinitesimally small changes, the
effect of one or more finite changes can be calculated and the correct final values for all
variables determined. This procedure has its basic mathematical representation in the
formal definition of an integral, as given in Section 4.1.1.

4.1 Integration
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

The notion of an integral as the area under a curve will be familiar to the reader, and has
already been used in Appendix A in connection with the definition of a natural logarithm.
In Figure 4.1, in which the solid line is the plot of a function f (x), the shaded area
represents the quantity denoted by
 b
I= f (x) dx. (4.1)
a

This expression is known as the definite integral of f (x) between the lower limit x = a
and the upper limit x = b, and f (x) is called the integrand. In this context, it should be
noted that the definite integral I does not depend upon x, which is known as the variable
of integration or dummy variable. If x were replaced throughout by a new variable of
integration, u say, then the value of I would not change in any way; I does depend on a, b
141
 142 Integral calculus

f (x )

a b x
Figure 4.1 An integral as the area under a curve.

and the form of f (x), but not upon x. However, despite having just emphasised this point,
we must draw the reader’s attention to the remarks immediately following Equation (4.11).

4.1.1 Integration from first principles

The definition of an integral as the area under a curve is not a formal definition, but one
that can be readily visualised. The formal definition1 of I involves subdividing the finite
interval a ≤ x ≤ b into a large number of subintervals, by defining intermediate points ξi
such that a = ξ0 < ξ1 < ξ2 < · · · < ξn = b, and then forming the sum
n
S= f (xi )(ξi − ξi−1 ), (4.2)
i=1
where xi is an arbitrary point that lies in the range ξi−1 ≤ xi ≤ ξi (see Figure 4.2). If now
n is allowed to tend to infinity in any way whatsoever, subject only to the restriction that
the length of every subinterval ξi−1 to ξi tends to zero, then S might, or might not, tend to
a unique limit, I . If it does, then the definite integral of f (x) between a and b is defined
as having the value I . If no unique limit exists the integral is undefined. For continuous
functions and a finite interval a ≤ x ≤ b the existence of a unique limit is assured and the
integral is guaranteed to exist.

Example  b
Evaluate from first principles the integral I = x 2 dx.
0

We first approximate the area under the curve y = x 2 between 0 and b by n rectangles of equal
width h. If we take the value at the lower end of each subinterval (in the limit of an infinite number
of subintervals we could equally well have chosen the value at the upper end) to give the height
of the corresponding rectangle, then the area of the kth rectangle will be (kh)2 h = k 2 h3 . The total
area is thus

n−1
A= k 2 h3 = (h3 ) 16 n(n − 1)(2n − 1),
k=0

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

1 This definition defines the Riemann integral. Other, more abstract, procedures for integration are possible and
enable the integration of more esoteric functions; but for integrations arising from physical situations, Riemann
integration is adequate.
143 4.1 Integration

f (x )

a x1 ξ1 x2 ξ2 x3 ξ3 x4 ξ4 x 5 b x
Figure 4.2 The evaluation of a definite integral by subdividing the interval
a ≤ x ≤ b into subintervals.

where we have used the expression for the sum of the squares of the natural numbers as given by
Equation (2.40). Now h = b/n and so
b3 n b3 1 1
A= (n − 1)(2n − 1) = 1− 2− .
n3 6 6 n n
As n → ∞, A → b3 /3, which is thus the value I of the integral.

Some straightforward properties of definite integrals that are almost self-evident are as
follows:
 b  a
0 dx = 0, f (x) dx = 0, (4.3)
a a
 c  b  c
f (x) dx = f (x) dx + f (x) dx, (4.4)
a a b
 b  b  b
[f (x) + g(x)] dx = f (x) dx + g(x) dx. (4.5)
a a a

Combining (4.3) and (4.4) with c set equal to a shows that

 b  a
f (x) dx = − f (x) dx. (4.6)
a b

4.1.2 Integration as the inverse of differentiation

The definite integral has been defined as the area under a curve between two fixed limits.
Let us now consider the integral
 x
F (x) = f (u) du (4.7)
a
in which the lower limit a remains fixed but the upper limit x is now variable.
 144 Integral calculus

It will be noticed that this is essentially a restatement of (4.1), but that the variable x
in the integrand has been replaced by a new variable u; as emphasised in the introduction
to this section, this makes no difference to the value of the integral. It is conventional to
rename the dummy variable in this way so that the same variable name does not appear in
both the integrand and the integration limits.2
It is apparent from (4.7) that F (x) is a continuous function of x, but at first glance the
definition of an integral as the area under a curve does not connect with our assertion that
integration is the inverse process to differentiation. However, by considering the integral
(4.7) and using the elementary property (4.4), we obtain
 x+x
F (x + x) = f (u) du
a
 x  x+x
= f (u) du + f (u) du
a x
 x+x
= F (x) + f (u) du.
x

Rearranging and dividing through by x yields


F (x + x) − F (x) 1 x+x
= f (u) du.
x x x

Letting x → 0 and using (3.1) we find that, by definition, the LHS becomes dF /dx,
whereas the RHS becomes f (x). The latter conclusion follows because when the integral
range x is small the value of the integral on the RHS is approximately f (x)x [since
f (u) is essentially constant throughout the range and has value f (x)], and in the limit
x → 0 no approximation is involved. Thus
dF (x)
= f (x), (4.8)
dx
or, substituting for F (x) from (4.7),
 x 
d
f (u) du = f (x). (4.9)
dx a
Pictorially, we can interpret Equation (4.9) as saying that, if we were to steadily increase
the value of b in Figure 4.1, the rate at which the shaded area in the figure would be
increasing at any point would vary as the value of f (b) at that point; for an increase of db
in b the shaded area would increase by f (b) db.
From the last two equations it is clear that integration can be considered as the inverse of
differentiation. However, we see from the above analysis that the lower limit a is arbitrary
and so differentiation does not have a unique inverse. Any function F (x) obeying (4.8)

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

2 When we come to study the integration of functions of more than one variable in Chapter 7, we will see that it is
possible to have the same variable appearing in both the integrand and the integration limits – however, even here,
the variable will be a fixed parameter so far as the integration is concerned, and it will be distinct from the dummy
variable of integration.
145 4.1 Integration

is called an indefinite integral of f (x), though any two such functions can only differ
by at most an arbitrary additive constant. Since the lower limit is arbitrary, it is usual to
write
 x
F (x) = f (u) du (4.10)

and explicitly include the arbitrary constant only when evaluating F (x). The evaluation is
conventionally written in the form

f (x) dx = F (x) + c (4.11)

where c is called the constant of integration. It will be noticed that, in the absence of any
integration limits, we use the same symbol for the arguments of both f and F . This can
be confusing, but is sufficiently common practice that the reader needs to become familiar
with it.
We also note that the definite integral of f (x) between the fixed limits x = a and x = b
can be written in terms of F (x). From (4.7) we have
 b  b  a
f (x) dx = f (x) dx − f (x) dx
a x0 x0

= F (b) − F (a), (4.12)

where x0 is any third fixed point. Using the notation F (x) = dF /dx, we may rewrite
(4.8) as F (x) = f (x), and so express (4.12) as
 b
F (x) dx = F (b) − F (a) ≡ [F ]ba .
a
The final identity symbol ≡ defines the meaning of the square bracket notation. It is a
generally accepted convention that [g(x)]ba is a shorthand way of writing g(b) − g(a), i.e.
the difference between the value of whatever function is contained between the brackets
when it is evaluated at x = b, and the value of the same function evaluated at x = a. For
example [x 3 ]32 = 27 − 8 = 19.
In contrast to differentiation, where repeated applications of the product rule and/or the
chain rule will always give the required derivative, it is not always possible to find the
integral of an arbitrary function. Indeed, in most real physical problems exact integration
cannot be performed and we have to revert to numerical approximations. Despite this
cautionary note, it is in fact possible to integrate many simple functions, and the following
sections introduce the most common types.

E X E R C I S E S 4.1
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Evaluate
1 3  π/2  2 π/2
(a) 4
x4 1
, (b) sin2 θ 0 , (c) x 2 −2 , (d) [cos θ]−π/2 .
 x  e
1 1
2. Given that du = ln x, evaluate dv.
1 u e −1 v
 146 Integral calculus

4.2 Integration methods

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Many of the techniques discussed in this section will probably be familiar to the reader
and so are summarised largely by example.

4.2.1 Integration by inspection

The simplest method of integrating a function is by inspection. Some of the more ele-
mentary functions have well-known integrals that should be remembered. The reader will
notice that these integrals are precisely the inverses of the derivatives found near the end
of Section 3.1.1. A few are presented below, using the form given in (4.11):
 
ax n+1
a dx = ax + c, ax n dx = + c,
n+1
 
eax a
eax dx = + c, dx = a ln x + c,
a x
 
a sin bx −a cos bx
a cos bx dx = + c, a sin bx dx = + c,
b b
 
−a ln(cos bx) a sinn+1 bx
a tan bx dx = + c, a cos bx sinn bx dx = + c,
b b(n + 1)

 
a −1 x −a cosn+1 bx
dx = tan + c, a sin bx cos n
bx dx = + c,
a2 + x 2 a b(n + 1)

x  
x 
−1 1
√ dx = cos−1 + c, √ dx = sin−1 + c,
a2 − x 2 a a2 − x 2 a
where the integrals that depend on n are valid for all n = −1 and where a and b are
constants. In the two final results |x| ≤ a.3

4.2.2 Integration of sinusoidal functions

 
Integrals of the type sinn x dx and cosn x dx may be found by using trigono-
metric expansions. Two methods are applicable, one for odd n and the other for
even n.
In the first of these, when n is odd, use is made of the fact that, to within a possible minus
sign, the pair of functions cos x and sin x are each the derivative of the other. This means
that if one factor is separated off, the remaining ones form an even power of the original
sinusoid, which can then be converted, using cos2 x + sin2 x = 1, to a sum of terms that
contain only even powers of the other sinusoid, together with constants. When in this
form, each term in the integrand is in the form of a power of a sinusoid multiplied by the
derivative of that same sinusoid, and so can be integrated immediately. In the illustrative
example that follows, one factor of sin x is used as (minus) the derivative of a sum of even
powers of cos x.

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

3 Reconcile the stated form of the first of these results with what might have been expected from the second, namely
that the indefinite integral of −(a 2 − x 2 )−1/2 is − sin−1 (x/a) + c.
147 4.2 Integration methods


Example
Evaluate the integral I = sin5 x dx.

Following the approach outlined above, we rewrite the integral as a product of sin x and an even
power of sin x, and then use the relation sin2 x = 1 − cos2 x:

I = sin4 x sin x dx

= (1 − cos2 x)2 sin x dx

= (1 − 2 cos2 x + cos4 x) sin x dx

= (sin x − 2 sin x cos2 x + sin x cos4 x) dx
= − cos x + 23 cos3 x − 15 cos5 x + c,
where the integration has been carried out using the results of Section 4.2.1. If the integrand had
been of the form cos2m+1 x, we would have separated off a single cos x and then expressed cos2m x
in terms of even powers of sin x, again giving a sum of terms each of which could be integrated by
inspection.

The second method, used for integrating even powers of sinusoids, depends on rewriting
the square of that sinusoid in terms of cos 2x and thus halving the power to which
any sinusoidal function is raised. If the integrand still contains squares (or higher even
powers) of sinusoids, the process is repeated. This reduction procedure comes at the
‘price’ of introducing multiples of x as the arguments of the sinusoids, but, for subsequent
integration, this presents no added difficulty.


Example
Evaluate the integral I = cos4 x dx.

Rewriting the integral as a power of cos2 x and then using the double-angle formula cos2 x =
1
2
(1 + cos 2x) yields
 
1 + cos 2x 2
I = (cos2 x)2 dx = dx
2

= 1
4
(1 + 2 cos 2x + cos2 2x) dx.

Using the double-angle formula again, we may write cos2 2x = 12 (1 + cos 4x), and hence

1 1 
I = 4
+ 2 cos 2x + 18 (1 + cos 4x) dx

= 14 x + 14 sin 2x + 18 x + 1
32
sin 4x + c
= 3
8
x + sin 2x +
1
4
1
32
sin 4x + c.
 148 Integral calculus

If the original integrand had been of the form sin2m x, we would have used the relationship sin2 x =
1
2
(1 − cos 2x), and so reduced it to one containing up to the mth power of cos 2x. This could then
be handled either by a repeat of the current procedure or by the method described in the previous
worked example (depending upon whether m is even or odd).4

4.2.3 Logarithmic integration

Integrals for which the integrand may be written as a fraction in which the numerator is
proportional to the derivative of the denominator may be evaluated using

Af (x)
dx = A ln f (x) + c. (4.13)
f (x)
This follows directly from the differentiation of a logarithm as a function of a function
(see Section 3.1.3).

Example Evaluate the integral


6x 2 + 2 cos x
I= dx.
x 3 + sin x

We note first that the numerator can be factorised to give 2(3x 2 + cos x), and then that the quantity
in parentheses is the derivative of the denominator. Hence

3x 2 + cos x
I =2 dx = 2 ln(x 3 + sin x) + c,
x 3 + sin x
where we have used (4.13) with f (x) = x 3 + sin x.

Sometimes the rearrangement needed to express the integrand in a form suitable for
logarithmic integration is a bit more subtle, as is illustrated by the following two examples.5
 
cos ax 1
cot ax dx = dx = ln(sin ax) + c,
sin ax a
 
1 ekx 1
−kx
dx = dx = ln(ekx + 1) + c.
1+e e +1
kx k

4.2.4 Integration using partial fractions

The method of partial fractions was discussed at some length in Section 2.3, but in essence
consists of the manipulation of a fraction (here the integrand) in such a way that it can be
written as the sum of two or more simpler fractions. In that discussion it was shown that

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

4 Note that the constant term in the integrand in its final form gives the average value of the power of the sinusoid
around a complete cycle. Thus cos4 x has an average value of 38 , whilst sin5 x has zero average value. An important
result, worth memorising, is that both cos2 x and sin2 x have an average value of 12 .
5 Use a similar method to that of the first to show that the indefinite integral of tan ax is a −1 ln(sec ax) + c.
149 4.2 Integration methods

each term in the partial fraction expansion of a rational fraction is of one of three types;
each of these types can be integrated directly in a standard way. More specifically:
1
integrates to ln(x − a) + c,
x−a
1 −1 1
integrates to + c, n = 1,
(x − a)n n − 1 (x − a)n−1
Ax + B A B x
integrates to ln(x 2 + a 2 ) + tan−1 + c.
x 2 + a2 2 a a
We illustrate the method with a simple example.

Example Evaluate the integral


1
I= dx.
x2 + x

The denominator factorises as x(x + 1), and so we separate the integrand into two partial fractions
and integrate each directly:

1 1 x
I= − dx = ln x − ln(x + 1) + c = ln + c.
x x+1 x+1
In this case, both terms were of the first type listed above and so each gave rise to a logarithm.6

4.2.5 Integration by substitution

Sometimes it is possible to make a substitution of variables that turns a complicated
integral into a simpler one, which can then be integrated by a standard method. There are
many useful substitutions, but knowing which to use is a matter of experience. We now
present a few examples of particularly useful ones.

Example Evaluate the integral


1
I= √ dx.
1 − x2

Making the substitution x = sin u, we note that dx = cos u du, and hence
  
1 1
I= √ cos u du = √ cos u du = du = u + c.
1 − sin2 u cos2 u
Now substituting back for u,
I = sin−1 x + c.
This corresponds to one of the results given in Section 4.2.1.

•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

6 Factorise f (x) = x 4 + 2x 3 + 5x 2 + 8x + 4 and so write down the general form of the integral of [f (x)]−1 , leaving
multiplicative constants unevaluated.
 150 Integral calculus
√
As a general guide, if an integrand in the form of a fraction contains a term a 2 − x 2
in its denominator, then some progress can usually be made by making the substitution
x = a sin u. The reason for this is that dx then becomes a cos u and this cos u in the
numerator cancels with the square root in the denominator, since the latter has also
become a cos u. This assumes that |x| ≤ a throughout the integration region (as it must
be if the integrand is to remain √ real).
If a square root is of the form x 2 − a 2 , with |x| ≥ a, then the appropriate substitution is
x = a cosh u, where cosh u is a hyperbolic cosine. The hyperbolic functions are introduced
and discussed in Chapter 5, where it is shown that, with this substitution, both dx and
the square root become a sinh u.7 If the square root is in the denominator, the two terms
cancel, but, even if it is in the numerator, the explicit square root has been removed.
Correspondingly,
√ and based on the same relationship (see footnote), square roots of the
form x + a are dealt with by substituting x = a sinh u, with both dx and the square
2 2

root becoming a cosh u.

Another particular example of integration by substitution is afforded by integrals of the
form
 
1 1
I= dx or I= dx. (4.14)
a + b cos x a + b sin x
In these cases, making the substitution t = tan(x/2) yields integrals that can be solved
more easily than the originals. Formulae expressing sin x and cos x in terms of t were
derived in Equations (1.73) and (1.74) (see p. 31), but before we can use them we must
relate dx to dt as follows.
Since
dt 1 x 1
x  1 + t2
= sec2 = 1 + tan2 = ,
dx 2 2 2 2 2
the required relationship is
2
dx = dt. (4.15)
1 + t2
We now have all the elements needed to change integrals of the form (4.14) into integrals
of rational polynomials, as is illustrated by the following example.

Example Evaluate the integral


2
I= dx.
1 + 3 cos x

Rewriting cos x in terms of t, as in Equation (1.74), and using (4.15) yields


2 2
I =   dt
1 + 3 (1 − t )(1 + t )
2 2 −1 1 + t2

2(1 + t 2 ) 2
= dt
1 + t 2 + 3(1 − t 2 ) 1 + t 2

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

7 The analogue for hyperbolic functions of cos2 x + sin2 x = 1 (for sinusoids) is cosh2 x − sinh2 x = 1.
151 4.2 Integration methods

 
2 2
= dt = √ √ dt
2 − t2 ( 2 − t)( 2 + t)

1 1 1
= √ √ +√ dt
2 2−t 2+t
1 √ 1 √
= − √ ln( 2 − t) + √ ln( 2 + t) + c
2 2
√
1 2 + tan (x/2)
= √ ln √ + c.
2 2 − tan (x/2)
In the final line we resubstituted for t in terms of x, the original variable.8

Integrals of a similar form to (4.14), but involving sin 2x, cos 2x, tan 2x, sin2 x, cos2 x or
tan2 x, rather than cos x and sin x, should be evaluated by using the substitution t = tan x.9
In this case
t 1 dt
sin x = √ , cos x = √ and dx = . (4.16)
1+ t2 1 + t2 1 + t2

A final example of the evaluation of integrals using substitution is the method of

completing the square (cf. Section 2.1). This method can be used where a quadratic
expression in the variable of integration x appears in the integrand; a change to a new
variable y that is a linear function of x, i.e. y = ax + b, reduces the quadratic expression
to one containing no linear term in y, but without introducing any complication into that
for dx, which simply becomes a dy. The following illustrates this procedure.

Example Evaluate the integral


1
I= dx.
x2 + 4x + 7

We can write the integral in the form


1
I= dx.
(x + 2)2 + 3
Substituting y = x + 2, we have that dy = dx and hence that

1
I= dy.
y +3
2

•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

8 Show that the value of x at which this final integral → ∞ is the same as that at which the original integrand → ∞.
9 Demonstrate, using the substitutions given by (4.16), that the integral of sin 2x is given, within an integration
constant, by cos2 x, which, again within a constant, is equal to 12 cos 2x.
152 Integral calculus

Comparison with the table of standard integrals (see Section 4.2.1) then shows that
1 y 1 x+2
I = √ tan−1 √ + c = √ tan−1 √ + c.
3 3 3 3
If, after completing the square, the denominator had been of the form y 2 − b2 , rather than y 2 + b2 ,
then the integral would have been of the general form (2b)−1 ln[(y − b)/(y + b)], rather than an
inverse tangent.

E X E R C I S E S 4.2
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Find the indefinite integrals of the following integrands, first rearranging them into
standard form where necessary:
a x 3 2
(a) , (b) , (c) , (d) , (e) cos 2x sin x.
a−x a−x 4 + (x + 1)2 x2 + 4x + 8
 π/4  π/4
2. Evaluate the definite integrals cos5 x dx and sin4 x dx.
0 0

3. Find the indefinite integrals of

x x(x + 3) 1
(a) , (b) , (c) .
a2 − x2 x3 + 3x 2 + x + 3 x3 + 3x 2 +x+3

4. Find the derivative of ln(tan x) and hence evaluate cosec 2x dx.

5. Evaluate
  
1
(a) x a2 − x2 dx, (b) √ dx,
x2 a2 − x 2
  π/2
1 1
(c) dx, (d) dx.
2 − sin x 0 1 + sin 2x

4.3 Integration by parts

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Integration by parts is the integration analogy of product differentiation. The principle is

to break down a complicated function into two functions, at least one of which can be
integrated by inspection. In fact, the method relies on the result for the derivative of a
product. Recalling from (3.7) that
d dv du
(uv) = u + v,
dx dx dx
where u and v are functions of x, we now integrate to find
 
dv du
uv = u dx + v dx.
dx dx
153 4.3 Integration by parts

Rearranging into the standard form for integration by parts gives

 
dv du
u dx = uv − v dx. (4.17)
dx dx
Integration by parts is often remembered for practical purposes in the form the integral
of a product of two functions is equal to {the first times the integral of the second} minus
the integral of {the derivative of the first times the integral of the second}. Here, u is ‘the
first’ and dv/dx is ‘the second’; clearly the integral v of ‘the second’ must be determinable
by inspection.


Example
Evaluate the integral I = x sin x dx.

In the notation given above, we identify x with u and sin x with dv/dx. Hence v = −cos x and
du/dx = 1 and so using (4.17)

I = x(−cos x) − (1)(−cos x) dx = −x cos x + sin x + c.

Since both x and sin x can be both integrated and differentiated by inspection, there was a decision
to be made about which would be u and which would be dv/dx in the general formula. The actual
choice of x as u was dictated by the fact that x ‘gets simpler’ when it is differentiated and more
complicated when integrated, whereas sin x ‘stays about the same’ in terms of complexity under
either operation.

The separation of the functions is not always so apparent, as is illustrated by the

following example.


Example
x 3 e−x dx.
2
Evaluate the integral I =

Firstly we rewrite the integral as



x 2 xe−x dx.
2
I=

Now, using the notation given above, we identify x 2 with u and xe−x with dv/dx. Hence v =
2

− 12 e−x and du/dx = 2x, and so

2


I = − 12 x 2 e−x − (−x)e−x dx = − 12 x 2 e−x − 12 e−x + c.
2 2 2 2

Here, there was very little real choice for u. It had to be whatever was left over from x 3 after
something proportional to the derivative (−2x) of the exponent (−x 2 ) of the exponential had been
taken from it as a factor; without the derivative of the exponent being included in what was taken
as dv/dx it would not be possible to carry out even one stage of integration.

A trick that is useful when the integral of the given integrand is not known, but its
derivative is, is to take ‘1’ and the integrand as the two factors in a product, which is then
integrated by parts. This is illustrated by the following example.
 154 Integral calculus


Example
Evaluate the integral I = ln x dx.

Firstly we rewrite the integral as


I= (ln x) 1 dx.

Now, using the notation above, we identify ln x with u and 1 with dv/dx. Hence we have v = x
and du/dx = 1/x, and so

1
I = (ln x)(x) − x dx = x ln x − x + c.
x
When this method is used, the ‘1’ must always be identified with dv/dx, and the original integrand
with u.10

It is sometimes necessary to integrate by parts more than once. In doing so, we may
occasionally encounter a multiple of the original integral I . In such cases we can obtain a
linear algebraic equation for I that can be solved to obtain its value.


Example
Evaluate the integral I = eax cos bx dx.

Integrating by parts, taking eax as the first function,11 we find


sin bx sin bx
I = eax − aeax dx,
b b
where, for convenience, we have omitted the constant of integration. Integrating by parts a second
time,

sin bx − cos bx − cos bx
I =e ax
− ae ax
+ a 2 eax dx.
b b2 b2
Notice that the integral on the RHS is just −a 2 /b2 times the original integral I . Thus
1 a a2
I = eax sin bx + 2 cos bx − 2 I.
b b b
Rearranging this expression to obtain I explicitly and including the constant of integration we find
eax
I= (b sin bx + a cos bx) + c. (4.18)
a 2 + b2
Another method of evaluating this integral, using the exponential of a complex number, is given in
Section 5.6.

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

10 Use the same technique to show that the integral of sin−1 x between 0 and 1 has the value 12 π − 1.
11 For this particular integral it does not matter whether cos bx or eax is taken as u, since both generate multiples
of themselves after two integrations or differentiations. The reader should check this by taking cos bx as u, the
opposite choice to that made in the text.
155 4.4 Reduction formulae

E X E R C I S E S 4.3
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Evaluate the following indefinite integrals:

  
(a) (3x 2 + 2x)e−x dx, x 5 e−x dx, tan−1 x dx.
3
(b) (c)


2. Evaluate sin x sin 2x dx by (a) rewriting the integrand and (b) using repeated inte-
gration by parts, showing that the two methods produce the same result.

4.4 Reduction formulae

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Integration using reduction formulae is a process that involves first evaluating a simple
integral and then, in stages, using it to find a more complicated one. In general structure,
the procedure follows the same lines as proof by induction (see Section 2.4.1), in that the
only direct calculations are for simple cases, with more complicated ones being tackled by
indirect methods. Reduction formulae also share with induction the feature that a positive
integer parameter characterises the various stages.
In practice, calculations using reduction formulae usually start with a form that is
expressed in terms of a general integer parameter n and then aim to relate that to one with
a lower value of n; that relationship is the reduction formula. The formula is then applied
repeatedly until the original integral is related to one in which n has a very low value,
usually 0 or 1, but occasionally 2 or 3. Finally, that low-n integral is evaluated directly.
The method can be illustrated by the following worked example.

Example Using integration by parts, find a relationship between In and In−1 where
 1
In = (1 − x 3 )n dx
0
1
and n is any positive integer. Hence evaluate I2 = 0 (1 − x 3 )2 dx.

Writing the integrand as a product and separating the integral into two we find
 1
In = (1 − x 3 )(1 − x 3 )n−1 dx
0
 1  1
= (1 − x )
3 n−1
dx − x 3 (1 − x 3 )n−1 dx.
0 0

The first term on the RHS is clearly In−1 and so, writing the integrand in the second term on the
RHS as a product,
 1
In = In−1 − (x)x 2 (1 − x 3 )n−1 dx.
0
156 Integral calculus

Integrating by parts we find

!1  1
x 1
In = In−1 + (1 − x 3 )n − (1 − x 3 )n dx
3n 0 0 3n
1
= In−1 + 0 − In ,
3n
which on rearranging gives
3n
In = In−1 .
3n + 1
We now have a reduction formula connecting successive integrals. Hence, if we can evaluate I0 , we
can find I1 , I2 etc. Evaluating I0 is trivial:
 1  1
I0 = (1 − x 3 )0 dx = dx = [x]10 = 1.
0 0

Hence
(3 × 1) 3 (3 × 2) 3 9
I1 = ×1= , I2 = × = .
(3 × 1) + 1 4 (3 × 2) + 1 4 14
Although the first few In could be evaluated by direct expansion of the integrand, using the
binomial theorem, followed by direct integration of terms like x r , this becomes tedious for integrals
containing higher values of n; these are therefore best evaluated using the reduction formula, which
gives
3n n!
In = n
r=0 (3r + 1)

as the general result.

E X E R C I S E 4.4
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •


1. Using the identity sec2 x = 1 + tan2 x, find a reduction formula for In = tann x dx,
where n is a non-negative integer. Hence write down a general expression for In ,
distinguishing between n even and n odd. Evaluate the definite integrals
 π/4  π/4
tan4 x dx and tan5 x dx.
0 0

4.5 Infinite and improper integrals

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

The definition of an integral given previously does not allow for cases in which either
of the limits of integration is infinite (an infinite integral) or for cases in which f (x) is
infinite in some part of the range (an improper integral), e.g. f (x) = (2 − x)−1/4 near the
point x = 2. Nevertheless, modification of the definition of an integral gives infinite and
improper integrals each a meaning.
157 4.5 Infinite and improper integrals
b
In the case of an integral I = a f (x) dx, the infinite integral, in which b tends to ∞,
is defined by
 ∞  b
I= f (x) dx = lim f (x) dx = lim F (b) − F (a).
a b→∞ a b→∞

As previously, F (x) is the indefinite integral of f (x) and limb→∞ F (b) means the limit
(or value) that F (b) approaches as b → ∞; it is evaluated after calculating the integral.
The formal concept of a limit will be introduced in Chapter 6, but for the present purposes
an intuitive interpretation will be sufficient.
Of course it may happen that as b → ∞ the indefinite integral F (b) does not tend to a
limit; in that case, the infinite integral is not defined. To take an example that has already
been discussed in Appendix A, the integral of x −1 between a and b is given by
 b
1
dx = ln b − ln a.
a x

As b → ∞ so does ln b, albeit very slowly. Consequently, ln b does not approach any

limit and as a result the infinite integral of x −1 is undefined. Our later, more precise,
treatment of limits will not change this conclusion. It will be seen that the existence or
otherwise of an infinite integral has to be tested against the definition on a case-by-case
basis.
Integrals for which the two limits are ±∞ are first evaluated between limits of ±b, and
then b is allowed to approach ∞. This can result in either outcome. For example:
 ∞  4 b
x
x dx = lim
3
= lim [ 1 (b4 − b4 )] = lim 0 = 0
−∞ b→∞ 4 −b b→∞ 4 b→∞

has a well-defined limit and the integral is defined, but

 ∞
 b
ex dx = lim ex −b = lim (eb − e−b ) = ∞ − 0 = ∞
−∞ b→∞ b→∞

does not and the integral is undefined.

As a non-trivial example of a defined infinite integral, consider the following.

Example Evaluate the integral

 ∞
x
I= dx.
0 (x 2 + a 2 )2
Integrating, we find F (x) = − 12 (x 2 + a 2 )−1 + c and so
 
−1 −1 1
I = lim − = .
b→∞ 2(b2 + a 2 ) 2a 2 2a 2
The value of the constant of integration c is immaterial, as it always is for integrals between
definite limits, here 0 and b.

 158 Integral calculus

To define improper integrals, we adopt the approach of excluding the unbounded range
from the integral, next performing the integration, and then letting the length of the
excluded range tend to zero. If the value of the integral tends to a definite limit as the
excluded length tends to zero, then that limit is defined as the value of the improper
integral. Expressed in more mathematical terms, if the integrand f (x) is infinite at x = c
(say), with a ≤ c ≤ b, then
 b  c−δ  b
f (x) dx = lim f (x) dx + lim f (x) dx,
a δ→0 a →0 c+
12
provided both limits exist. The following example illustrates the procedure.

Example  2
Evaluate the integral I = (2 − x)−1/4 dx.
0

Since the integrand becomes infinite at x = 2 and this is in the integration range, we initially cut
out the interval from x = 2 −  to x = 2. Then, integrating directly gives
 2−
I = − 43 (2 − x)3/4 0 .
We now test whether this tends to a finite limit as  is allowed to tend to zero.
   
I = lim − 43  3/4 + 43 23/4 = 43 23/4 .
→0

It clearly does, and that limit is therefore the value of I . Notice that the result does not, and must
not, depend upon any particular value of .

E X E R C I S E S 4.5
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Determine whether the following infinite integrals exist and, where they do, evaluate
them:
 ∞  ∞
x
xe−λx dx,
2
(a) (b) dx,
0 0 a + x2
2
 ∞  ∞
x
(c) dx, (d) sin x dx.
−∞ a + x
2 2
0

2. Determine whether the following improper integrals exist and, where they do, evaluate
them:
 π/2  π/2
(a) tan θ dθ , (b) tan θ dθ ,
0 −π/2
 1  1
x2 1
(c) dx, (d) dx.
0 (1 − x 3 )3/4 0 (3x − 1)2

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

12 If a common quantity, h say, is used instead of both δ and , and the limit exists, then the limit is called the
principal value of the integral.
159 4.6 Integration in plane polar coordinates

y
C

ρ dφ
ρ(φ+ dφ)
dA
ρ(φ) B

O
x
Figure 4.3 Finding the area of a sector OBC defined by the curve ρ(φ) and the radii
OB, OC, at angles to the x-axis φ1 , φ2 respectively.

4.6 Integration in plane polar coordinates

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

As described in Section 2.2.4, a curve is defined in plane polar coordinates ρ, φ by its

distance ρ from the origin as a function of the angle φ between the line joining a point on
the curve to the origin and the x-axis, i.e. ρ = ρ(φ). The size of an element of area is given
in the same coordinate system by dA = 12 ρ 2 dφ, as is illustrated in Figure 4.3. The total
area enclosed by the curve in the sector defined by angles φ1 and φ2 is therefore given by
 φ2
A= 1 2
2
ρ dφ. (4.19)
φ1

One immediate, but hardly novel, deduction from this is that the area of a circle of
radius a, is given by
 2π
 2π
A= a dφ = 12 a 2 φ 0 = πa 2 .
1 2
2
0
A more substantial calculation is provided by the following example.

Example The equation in polar coordinates of an ellipse with semi-axes a and b is

1 cos2 φ sin2 φ
2
= 2
+ .
ρ a b2
Find the area A of the ellipse.

Using (4.19) and symmetry, we have

  π/2
1 2π a 2 b2 1
A= dφ = 2a 2 2
b dφ.
2 0 b2 cos2 φ + a 2 sin2 φ 0 b 2 cos2 φ + a 2 sin2 φ
 160 Integral calculus

To evaluate this integral we divide both numerator and denominator by cos2 φ and then write
tan φ = t:13
 π/2
sec2 φ
A = 2a 2 b2 dφ
0 b + a 2 tan2 φ
2
 ∞  ∞
1 1
= 2a 2 b2 2 + a2t 2
dt = 2b 2
2 + t2
dt.
0 b 0 (b/a)
Finally, using the list of standard integrals (see Section 4.2.1),
 ∞
π 
1 t
A = 2b2 tan−1 = 2ab − 0 = πab.
(b/a) (b/a) 0 2
Of course, if we let a = b, the familiar result for a circle is recovered.

E X E R C I S E 4.6
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Using symmetry to avoid any ambiguity concerned with ‘negative ρ-values’, find the
total areas of (a) the lemniscate of Bernoulli, ρ 2 = a 2 cos 2φ, and (b) the particular
limaçon ρ = 12 a(3 + 2 cos φ).

4.7 Integral inequalities

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Consider the functions f (x), φ1 (x) and φ2 (x) such that φ1 (x) ≤ f (x) ≤ φ2 (x) for all x in
the range a ≤ x ≤ b. It immediately follows that
 b  b  b
φ1 (x) dx ≤ f (x) dx ≤ φ2 (x) dx, (4.20)
a a a
which gives us a way of putting bounds on the value of an integral that is difficult to
evaluate explicitly.

Example Show that the value of the integral

 1
1
I= dx
0 (1 + x2 + x 3 )1/2
lies between 0.810 and 0.882.

What makes this integral difficult to evaluate is the x 3 term in the denominator. If it were absent,
we would have an integrand of the form (a 2 + x 2 )−1/2 ; this could be handled in closed form, as
we indicated in Section 4.2.5. We therefore need to put bounds on x 3 , with the bounds expressed
in terms of functions that can be managed. We note that for x in the range 0 ≤ x ≤ 1, the double
inequality 0 ≤ x 3 ≤ x 2 holds. Hence
(1 + x 2 )1/2 ≤ (1 + x 2 + x 3 )1/2 ≤ (1 + 2x 2 )1/2 ,

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

13 Verify that the direct substitution of the relationships given in (4.16) gives the same t-integral.
161 4.8 Applications of integration

and so
1 1 1
≥ ≥ .
(1 + x )
2 1/2 (1 + x + x )
2 3 1/2 (1 + 2x 2 )1/2
Consequently,
 1  1  1
1 1 1
dx ≥ dx ≥
dx.
0 (1 + x 2 )1/2
0 (1 + x 2 + x 3 )1/2
0 (1 + 2x 2 )1/2
√
We have not yet found the integral of (a 2 + x 2 )−1/2 ; it can be expressed as ln(x + a 2 + x 2 ) or,
in terms of inverse hyperbolic functions (see Chapter 5), as sinh−1 (x/a). The first of these can be
verified by direct differentiation. Taking this result for granted at this stage, we have
 !1
 !1
ln(x + 1 + x 2 ) ≥ I ≥ √12 ln x + 12 + x 2
0 0

0.8814 ≥ I ≥ 0.8105
0.882 ≥ I ≥ 0.810.
In the last line the calculated values have been rounded to three significant figures, one rounded
up and the other rounded down so that the proved inequality cannot be unknowingly made
invalid.

E X E R C I S E 4.7
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Noting that, for 0 ≤ x ≤ π/2, the double inequality 2x/π ≤ sin x ≤ x holds, find to 3
 π/2
1
s.f. limits for the value of I = dx. Using an appropriate substitution,
0 1 + sin2 x
evaluate I exactly and so verify the validity of the limits.

4.8 Applications of integration

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

In this section we give brief outlines of some standard procedures that involve the use of
integration. They typically form only a part of a larger calculation, and each would not
normally be specified in any more detail than that given by the corresponding subsection
heading.

4.8.1 Mean value of a function

The mean value m of a function between two limits a and b is defined by
 b
1
m= f (x) dx. (4.21)
b−a a
The mean value may be thought of as the height of the rectangle that has the same area
(over the same interval) as the area under the curve f (x). This is illustrated in Figure 4.4.
 162 Integral calculus

f (x )

a b x
Figure 4.4 The mean value m of a function.

Example Find the mean value m of the function f (x) = x 2 between the limits x = 2 and x = 4.

Using (4.21),
 4  4
1 1 x3 1 43 23 28
m= x 2 dx = = − = .
4−2 2 2 3 2 2 3 3 3
As expected, because x 2 increases more rapidly than x, this result is (slightly) more than the square
of the mid-value of x over the given range; that would give 32 = 9.

4.8.2 Finding the length of a curve

Finding the area between a curve and certain straight lines provides one example of the use
of integration, as we saw in Section 4.6. Another is that of finding the length of a curve.
If a curve is defined by y = f (x) then the distance along the curve, s, that corresponds
to small changes x and y in x and y is given by

s ≈ (x)2 + (y)2 ; (4.22)
this follows directly from Pythagoras’ theorem (see Figure 4.5). Dividing (4.22) through
by x and letting x → 0 we obtain14
2
ds dy
= 1+ . (4.23)
dx dx
Clearly the total length s of the curve between the points x = a and x = b is then given
by integrating both sides of the equation:
 b
dy 2
s= 1+ dx. (4.24)
a dx
The following provides a simple example of the use of this method.

••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

14 Instead of considering small changes x and y and letting these tend to zero, we could have derived (4.23)
by considering infinitesimal changes dx and dy from the start. After writing (ds)2 = (dx)2 + (dy)2 , (4.23) may
be deduced by using the formal device of dividing through by dx. Although not mathematically rigorous, this
method is often used and generally leads to the correct result.
163 4.8 Applications of integration

f (x )
y = f (x )

∆s ∆y
∆x

x
Figure 4.5 The distance moved along a curve, s, corresponding to the small
changes x and y.

Example Find the length of the curve y = x 3/2 from x = 0 to x = 2.

√
Using (4.24) and noting that dy/dx = 32 x, the length s of the curve is given by
 2
s= 1 + 94 x dx
0
4  3/2 !2  3/2 !2
= 2
3 9
1 + 94 x = 8
27
1 + 94 x
0 0
 11 3/2 !
= 8
27 2
−1 .

For a more general power curve y = x n (n > 0), the integration would not be so straightforward;
n = 3/2 gives a linear function of x under the square root sign and so makes the integration
elementary.

Although less often done, it is equally valid to divide (4.22) through by y and let
y → 0 and so obtain
2  d 2
ds dx dx
= 1+ leading to s= 1+ dy, (4.25)
dy dy c dy
where c and d are the y-values marking the beginning and end of the curve. If the extremes
of the curve are given in this form, then this can be the best way to proceed. The hyperbolic
functions cosh x and sinh x are not introduced formally until the next chapter, but we can
use them expressed in exponential form to provide a worked example.

Example Find the length of the curve given by y(x) = 1 (ex + e−x ) between the points at which y = 1 and
2
y = Y , where Y > 1.

The y = 1 end of the curve clearly corresponds to x = 0, but solving Y = 12 (ex + e−x ) for x
is somewhat more complicated (though it can be done; see Section 5.7.5) and so we make y the
variable of integration. We need (dx/dy)2 but, given the equation of the curve, it is easier to calculate
 164 Integral calculus

(dy/dx)2 as
dy 2
1 2  2
= 2
(ex − e−x ) = 12 (ex + e−x ) − 1 = y 2 − 1.
dx
Inserting the reciprocal of this result into the alternative expression for s and using the limits for y
(not x) gives
 Y  Y  !Y 
1 y
s= 1+ dy =  dy = y2 − 1 = Y2 − 1
1 y −1
2
1 y2 −1 1

as the length of the curve between y = 1 and y = Y .

In the other two-dimensional coordinate system we have met so far, namely plane polar
coordinates, the corresponding expression for the length of a curve is

ds = (dρ)2 + (ρ dφ)2 ,
leading to
 ρ2 2  φ2 2
dφ dρ
s= 1+ ρ2 dρ or s= ρ2 + dφ (4.26)
ρ1 dρ φ1 dφ
For the simple spiral given by ρ = bφ, the two equivalent expressions for the length of
the spiral up to the point where it has completed one ‘orbit’ are
 2πb   2π 
ρ2
s= 1 + 2 dρ or s = b φ 2 + 1 dφ.
0 b 0

These integrals could be evaluated in terms of the hyperbolic functions that are studied in
the next chapter, but doing so would add little to the main point of this subsection.15

4.8.3 Surfaces of revolution

Whilst it is easy to give an expression for the curved surface area of a uniform circular
cylinder – 2πrh for a cylinder of constant radius r and length h – it is less straightforward
if the radius of the cylinder varies along its length. In the former case, we could imagine
the surface cut and rolled out into a flat rectangle with sides of 2πr and h, but, for the
latter, the unrolled surface would not have a shape for which a readily available expression
gives its area.
To make it into a set of calculable areas, we could cut the cylinder surface, perpen-
dicularly to the axis of symmetry, into narrow strips; the length of any particular strip
would be 2π times the local radius r of the cylinder. If the strip were the result of two
cuts made a distance x apart, measured along the axis of the cylinder, then the width of
the strip when laid out flat would be s, with s as given by Equation (4.22).16 The area
of the strip would therefore be 2πrs and the area of the surface would be the sum of all
••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

15 Extend this formalism to three dimensions and show that the length of thread on a uniform machine screw that
has radius a and pitch h is (h2 + 4π 2 a 2 )1/2 per turn.
16 Note that, in general, s will be larger than x, and that it will never be smaller.
165 4.8 Applications of integration

y f (x)
ds

V
dx
a b x

S
Figure 4.6 The surface and volume of revolution for the curve y = f (x).

such quantities. This approach, in the limit that x becomes infinitesimal, is the basis of
finding the area using integration. The following derivation is a much terser mathematical
description of this procedure.
Consider the surface S formed by rotating the curve y = f (x) about the x-axis (see
Figure 4.6). The surface area of the ‘collar’ formed by rotating an element of the curve,
ds, about the x-axis is 2πy ds, and hence the total surface area is
 b
S= 2πy ds.
a

Since, from (4.23), ds = [1 + (dy/dx) ] dx , the total surface area between the planes
2 1/2

x = a and x = b is
 b
dy 2
S= 2πy 1 + dx. (4.27)
a dx
We now illustrate this result with a simple example.

Example Find the curved surface area of the cone formed by rotating about the x-axis the line y = 2x
between x = 0 and x = h.

Using (4.27), the surface area is given by

 h  2
d
S= (2π)2x 1 + (2x) dx
0 dx
 h  h √
 1/2
= 4πx 1 + 22 dx = 4 5πx dx
0 0
√ !h √ √
= 2 5πx 2 = 2 5π(h2 − 0) = 2 5πh2 .
0

As it must be, this result is in agreement with the standard formula for the area of the curved surface
of a cone, namely S = πr, where r is the√radius of its base (here r = 2h) and is its slope length,
given in this case by = h2 + (2h)2 = 5h.

166 Integral calculus

We note that a surface of revolution may also be formed by rotating a line about the
y-axis. In this case the surface area between y = a and y = b is

 b 2
dx
S= 2πx 1 + dy. (4.28)
a dy

As an example of this kind of calculation, consider the following problem.

Example Find the curved surface area of a parabolic bowl that has the form x 2 = 4ay, a base that is 4a in
diameter, and height h.

Most of the calculation consists of algebraic manipulation, but we do need to find dx/dy. This is
easily done by differentiating x 2 = 4ay and obtaining 2x(dx/dy) = 4a. As the bowl has a base
of radius 2a, its profile corresponds to the section of the parabola between y = (2a)2 /4a = a and
y = a + h. Substitution into (4.28) gives
 a+h 2
2a
S= 2πx 1 + dy
a x
 a+h 
= 2π x 2 + 4a 2 dy
a
 a+h 
= 2π 4ay + 4a 2 dy
a
 a+h √ √
= 4π a y + a dy
a
√  a+h
= 4π a 23 (y + a)3/2 a
√  
= 83 π a (2a + h)3/2 − (2a)3/2 .
It should be noted that to obtain the integrand in the third line entirely in terms of y we used
the curve-defining equation to replace x 2 , and that it is the y-limits, a and a + h, that are
appropriate.

4.8.4 Volumes of revolution

The volume V enclosed by rotating the portion of the curve y = f (x) between x = a and
x = b about the x-axis (see Figure 4.6) can also be found using integration.
We consider the complete volume as made up of a very large number (formally, an
infinitely large number) of thin discs, each of thickness dx. The volume of the single disc
that lies between between x and x + dx, and has radius y(x), is given by dV = πy 2 dx;
since the disc is vanishingly thin, we can ignore any variation of its radius within the
disc – more formally, the contribution to the volume of this variation is second order in
167 4.8 Applications of integration

dx. To obtain the total volume enclosed by the rotating curve, we integrate this infinitesimal
volume between x = a and x = b:
 b
V = πy 2 dx. (4.29)
a

Our final worked example uses the same cone as the first example in the previous
subsection.

Example Find the volume of the cone enclosed by the surface that is formed when the portion of the line
y = 2x between x = 0 and x = h is rotated about the x-axis.

Using (4.29), the volume is given by

 h  h
V = π(2x)2 dx = 4πx 2 dx
0 0
4 h
= 3
πx 3 0
= 43 π(h3 − 0) = 43 πh3 .

Again agreement is obtained with a standard formula: for a cone, the volume V = 13 πr 2 h, with
r = 2h in the current case.

As before, it is also possible to form a volume of revolution by rotating a curve about

the y-axis. In this case,
 b
V = πx 2 dy. (4.30)
a

gives the volume enclosed between y = a and y = b.17

E X E R C I S E S 4.8
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

1. Find the mean values of the following functions over the ranges indicated:
(a) x 3 in [0, 2], (b) x 3 in [−2, 2], (c) sin θ in [0, π],
(d) tan2 θ in [0, π/4], (e) x 3 e−x in [0, ∞].
2. Show that the length L(x) of the curve y = ln(cos x), with 0 ≤ x ≤ π/2, as measured
x
from the origin x = 0, y = 0, is given by L(x) = sec u du. Using the substitution
0
t = tan u/2, evaluate L(x) as
1 + tan(x/2)
L(x) = ln .
1 − tan(x/2)

•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••

17 Show that the capacity of the parabolic bowl discussed in the second worked example of the previous subsection
is 2π ah(2a + h).
168 Integral calculus

3. Show that the (outside) surface area of a flat-bottomed, straight-sided tumbler, 4a high,
that has a base diameter of 2a and a diameter of 3a at its widest part, is
 √ 
5 65
πa 1 +
2
.
4

4. By considering them as a surface and volume of revolution generated by the semi-

circular arc x 2 + y 2 = a 2 , establish the well-known formulae for the surface area and
volume of a sphere.
5. Find the volume of the solid obtained by rotating the curve y = x(1 − x) for 0 ≤ x ≤ 1
around the x-axis.

SUMMARY
1. Elementary properties of integrals
 b  a  b  a
0 dx = 0, f (x) dx = 0, f (x) dx = − f (x) dx,
a a a b
 c  b  c
f (x) dx = f (x) dx + f (x) dx,
a a b
 b  b  b
[ f (x) + g(x)] dx = f (x) dx + g(x) dx,
a a a
 x 
d d
F (x) ≡ f (u) du = f (x).
dx dx x0

2. Standard integrals
r For the integrals of elementary functions, including exponentials and sinusoids, see
p. 146.
r Some particularly important cases for physical science:
 
sin x dx = −cos x + c, cos x dx = sin x + c,


1 1 −1 x
dx = tan + c,
a2 + x 2 a a
 nπ/2  nπ/2
nπ
cos x dx =
2
= sin2 x dx,
0 4 0
 x0 +(nπ/α)  x0 +(nπ/α)
nπ
cos2 (αx) dx = = sin2 (αx) dx.
x0 2α x0
169 Summary

3. Common substitutions
With t = tan θ/2,
2t 1 − t2 2
sin θ = , cos θ = , dθ = dt.
1 + t2 1 + t2 1 + t2

Integrand
contains Substitution Differential
√
a2 − x 2 x = a sin u dx = a cos u du
√
a2 + x 2 x = a sinh u dx = a cosh u du,
√
x 2 − a2 x = a cosh u dx = a sinh u du

4. Integration by parts
 
dv du
u dx = uv − v dx
dx dx
  x   x
du
or uw dx = u w dx − w dx dx.
dx
It is sometimes helpful to use the second form with w as (a hidden) unity.

5. Infinite
 ∞ and improper integrals
 b
r f (x) dx = lim f (x) dx = lim F (b) − F (a).
a b→∞ a b→∞
r If limx→c f (x) = ∞ with a ≤ c ≤ b, then
 b  c−δ  b
f (x) dx = lim f (x) dx + lim f (x) dx,
a δ→0 a →0 c+

provided both limits exist.

6. Curve lengths, and areas and volumes of revolution

r Curve length
 b 2  d 2
dy dx
s= 1+ dx or s= 1+ dy.
a dx c dy
r Area of solid of revolution
 b 
dy 2 d
dx 2
S = 2π y 1+ dx or S = 2π x 1+ dy.
a dx c dy
r Volume of solid of revolution
 b  d
V =π y 2 dx or V =π x 2 dy.
a c
170 Integral calculus

PROBLEMS
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

4.1. Find, by inspection, the indefinite integrals of

(a) 7x 6 ; (b) e3x + e−3x ; (c) cot 3x; (d) sin x sin 2x; (e) cos x sin 2x;
(f) (a − 2x)−1 ; (g) (4 + x 2 )−1 ; (h) (4 − x 2 )−1/2 ; (i) x(4 + x 2 )−1 .

4.2. Find the following indefinite

 integrals:
(a) (4 + x 2 )−1 dx; (b) (8 + 2x − x 2 )−1/2 dx for 2 ≤ x ≤ 4;
  √
(c) (1 + sin θ)−1 dθ; (d) (x 1 − x)−1 dx for 0 < x ≤ 1.

4.3. Find the indefinite integrals J of the following ratios of polynomials:

(a) (x + 3)/(x 2 + x − 2);
(b) (x 3 + 5x 2 + 8x + 12)/(2x 2 + 10x + 12);
(c) (3x 2 + 20x + 28)/(x 2 + 6x + 9);
(d) x 3 /(a 8 + x 8 ).

4.4. Express x 2 (ax + b)−1 as the sum of powers of x and another integrable term, and
hence evaluate
 b/a
x2
dx.
0 ax + b

4.5. Find the integral J of (ax 2 + bx + c)−1 , with a = 0, distinguishing between the
cases (i) b2 > 4ac, (ii) b2 < 4ac and (iii) b2 = 4ac.

4.6. Use logarithmic integration to find the indefinite integrals J of the following:
(a) sin 2x/(1 + 4 sin2 x);
(b) ex /(ex − e−x );
(c) (1 + x ln x)/(x ln x);
(d) [x(x n + a n )]−1 .

4.7. Find the derivative of f (x) = (1 + sin x)/ cos x and hence determine the
indefinite integral J of sec x.

4.8. Find the indefinite integrals, J , of the following functions involving sinusoids:
(a) cos5 x − cos3 x;
(b) (1 − cos x)/(1 + cos x);
(c) cos x sin x/(1 + cos x);
(d) sec2 x/(1 − tan2 x).

4.9. By making the substitution x = a cos2 θ + b sin2 θ, evaluate the definite integrals
J between limits a and b (> a) of the following functions:
(a) [(x − a)(b − x)]−1/2 ;
(b) [(x − a)(b − x)]1/2 ;
(c) [(x − a)/(b − x)]1/2 .
171 Problems

4.10. Determine whether the following integrals exist and, where they do, evaluate
 ∞
them:  ∞
x
(a) exp(−λx) dx; (b) 2 + a 2 )2
dx;
0 −∞ (x
 ∞  1
1 1
(c) dx; (d) dx;
1 x+1 0 x
2
 π/2  1
x
(e) cot θ dθ ; (f) dx.
0 0 (1 − x 2 )1/2

4.11. Useintegration by parts to evaluate the following:

y y
(a) x 2 sin x dx; (b) x ln x dx;
0 y 1 y
(c) sin−1 x dx; (d) ln(a 2 + x 2 )/x 2 dx.
0 1

4.12. Show, using the following methods, that the indefinite integral of x 3 /(x + 1)1/2 is
J = 2
35
(5x 3 − 6x 2 + 8x − 16)(x + 1)1/2 + c.
(a) Repeated integration by parts.
(b) Setting x + 1 = u2 and determining dJ /du as (dJ /dx)(dx/du).

4.13. The gamma function (n) is defined for all n > −1 by

 ∞
(n + 1) = x n e−x dx.
0
Find a recurrence relation connecting (n + 1) and (n).
(a) Deduce (i) the value of (n +
 1) when n is a non-negative integer and (ii) the
√
value of  72 , given that  12 = π.
(b) Now, taking factorial m for any m to be defined by m! = (m + 1), evaluate
− 32 !.

4.14. Define J (m, n), for non-negative integers m and n, by the integral
 π/2
J (m, n) = cosm θ sinn θ dθ.
0
(a) Evaluate J (0, 0), J (0, 1), J (1, 0), J (1, 1), J (m, 1), J (1, n).
(b) Using integration by parts, prove that, for m and n both > 1,
m−1 n−1
J (m, n) = J (m − 2, n) and J (m, n) = J (m, n − 2).
m+n m+n
(c) Evaluate (i) J (5, 3), (ii) J (6, 5) and (iii) J (4, 8).

4.15. By integrating by parts twice, prove that In , as defined in the first equality below
for positive integers n, has the value given in the second equality:
 π/2
n − sin(nπ/2)
In = sin nθ cos θ dθ = .
0 n2 − 1
172 Integral calculus

4.16. Evaluate
 ∞ the following definite
 1  3 integrals: 
−x
(a) 0 xe dx; (b) 0 (x + 1)/(x 4 + 4x + 1) dx;
 π/2 ∞
(c) 0 [a + (a − 1) cos θ]−1 dθ with a > 12 ; (d) −∞ (x 2 + 6x + 18)−1 dx.

4.17. If Jr is the integral

 ∞
x r exp(−x 2 ) dx
0
show that
(a) J2r+1 = (r!)/2,
(b) J2r = 2−r (2r − 1)(2r − 3) · · · (5)(3)(1) J0 .

4.18. Find positive constants a, b such that ax ≤ sin x ≤ bx for 0 ≤ x ≤ π/2. Use this
inequality to find (to two significant figures) upper and lower bounds for the
integral
 π/2
I= (1 + sin x)1/2 dx.
0
Use the substitution t = tan(x/2) to evaluate I exactly.

4.19. By noting that for 0 ≤ η ≤ 1, η1/2 ≥ η3/4 ≥ η, prove that

 a
2 1 π
≤ 5/2 (a 2 − x 2 )3/4 dx ≤ .
3 a 0 4

4.20. The official specifications for a rugby ball allow one that has a length of 300 mm
and a smallest circumference of 600 mm. By treating it as an ellipsoid of
revolution, find its volume.

4.21. A vase has curved sides that are generated by rotating the part of the curve
x = 12 a(ey/a + e−y/a ) that lies between y = 0 and y = ha around the y-axis.
Show that the area of the curved surface is πa 2 [ 14 (e2h − e−2h ) + h].

4.22. Show that the total length of the astroid x 2/3 + y 2/3 = a 2/3 , which can be
parameterised as x = a cos3 θ, y = a sin3 θ, is 6a.

4.23. By noting that sinh x < 12 ex < cosh x, and that 1 + z2 < (1 + z)2 for z > 0, show
that, for x > 0, the length L of the curve y = 12 ex measured from the origin
satisfies the inequalities sinh x < L < x + sinh x.

4.24. The equation of a cardioid in plane polar coordinates is

ρ = a(1 − sin φ).
Sketch the curve and find (i) its area, (ii) its total length, (iii) the surface area of
the solid formed by rotating the cardioid about its axis of symmetry and (iv) the
volume of the same solid.
173 Hints and answers

HINTS AND ANSWERS

4.1. (a) x 7 ; (b) 13 (ex − e−x ); (c) 13 ln sin 3x; (d) consider sin 2x as 2 sin x cos x, 23 sin3 x;
(e) − 23 cos3 x; (f) − 12 ln(a − x); (g) 12 tan−1 (x/2) (h) sin−1 (x/2); (i) 12 ln(4 + x 2 ).
4.3. (a) Express in partial fractions; J = 13 ln[(x − 1)4 /(x + 2)] + c.
(b) Divide the numerator by the denominator and express the remainder in partial
fractions; J = x 2 /4 + 4 ln(x + 2) − 3 ln(x + 3) + c.
(c) After division of the numerator by the denominator the remainder can be
expressed as 2(x + 3)−1 − 5(x + 3)−2 ; J = 3x + 2 ln(x + 3) + 5(x + 3)−1
+c.
(d) Set x 4 = u; J = (4a 4 )−1 tan−1 (x 4 /a 4 ) + c.
4.5. Writing b2 − 4ac as 2 > 0, or 4ac − b2 as  2 > 0:
(i) −1 ln[(2ax + b − )/(2ax + b + )] + k;
(ii) 2 −1 tan−1 [(2ax + b)/ ] + k;
(iii) −2(2ax + b)−1 + k.
4.7. f (x) = (1 + sin x)/ cos2 x = f (x) sec x; J = ln(f (x)) + c = ln(sec x + tan x)
+c.
4.9. Note that dx = 2(b − a) cos θ sin θ dθ.
(a) π; (b) π(b − a)2 /8; (c) π(b − a)/2.
4.11. (a) (2 − y 2 ) cos y + 2y sin y − 2; (b) [(y 2 ln y)/2] + [(1 − y 2 )/4];
(c) y sin−1 y + (1 − y 2 )1/2 − 1;
(d) ln(a 2 + 1) − (1/y) ln(a 2 + y 2 ) + (2/a)[tan−1 (y/a) − tan−1 (1/a)].
√ √
4.13. (n + 1) = n(n); (a) (i) n!, (ii) 15 π /8; (b) −2 π .
4.15. By integrating twice recover a multiple of In .
4.17. J2r+1 = rJ2r−1 and 2J2r = (2r − 1)J2r−2 .
4.19. Set η = 1 − (x/a)2 throughout and x = a sin θ in one of the bounds.
4.21. Note that (u + u−1 )2 = (u − u−1 )2 + 4.
x  1/2
4.23. L = 0 1 + 14 exp 2x dx.
5 Complex numbers and
hyperbolic functions

This chapter is concerned with the representation and manipulation of complex

numbers. Complex numbers pervade this book, underscoring their wide application in
the mathematics of the physical sciences. Some elementary applications of complex
numbers to the description of physical systems appear in later chapters, but only the
basic tools are presented here.

5.1 The need for complex numbers

• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •

Although complex numbers occur in many branches of mathematics, they arise most
directly out of solving polynomial equations. We examine a specific quadratic equation
as an example.
Consider the quadratic equation
z2 − 4z + 5 = 0. (5.1)
Equation (5.1) has two solutions, z1 and z2 , such that
(z − z1 )(z − z2 ) = 0. (5.2)
Using the familiar formula for the roots of a quadratic equation, (2.4), the solutions z1 and
z2 , written in brief as z1,2 , are

4 ± (−4)2 − 4(1 × 5)
z1,2 =
2
√
−4
=2± . (5.3)
2
Both solutions contain the square root of a negative number. However, it would not be
true to say that there are no solutions to the quadratic equation. The fundamental theorem
of algebra states that a quadratic equation will always have two solutions and these are in
fact given by (5.3). The second term on the RHS of (5.3) is called an imaginary term since
it contains the square root of a negative number; the first term is called a real term. The full
solution is the sum of a real term and an imaginary term and is called a complex number.
A plot of the function f (z) = z2 − 4z + 5 is shown in Figure 5.1. It will be seen that the
plot does not intersect the z-axis, corresponding to the fact that the equation f (z) = 0 has
no purely real solutions.
The choice of the symbol z for the quadratic variable was not arbitrary; the conventional
representation of a complex number is z, where z is the sum of a real part x and i times
174