140 INTEGRALS CHAP.

7. Apply inequality (1), Sec. 47, to show that for all values of x in the interval −1 ≤ x ≤ 1,
the functions∗
 
1 π
Pn (x) = (x + i 1 − x 2 cos θ)n dθ (n = 0, 1, 2, . . .)
π 0
satisfy the inequality |Pn (x)| ≤ 1.
8. Let C N denote the boundary of the square formed by the lines
   
1 1
x =± N+ π and y=± N+ π,
2 2
where N is a positive integer and the orientation of C N is counterclockwise.
(a) With the aid of the inequalities
|sin z| ≥ |sin x| and |sin z| ≥ |sinh y|,
obtained in Exercises 8(a) and 9(a) of Sec. 38, show that | sin z| ≥ 1 on the vertical
sides of the square and that |sin z| > sinh(π/2) on the horizontal sides. Thus show
that there is a positive constant A, independent of N, such that |sin z| ≥ A for all
points z lying on the contour C N .
(b) Using the final result in part (a), show that



dz

16

≤

z sin z
(2N + 1)π A
2
CN

and hence that the value of this integral tends to zero as N tends to infinity.

48. ANTIDERIVATIVES
Although the value of a contour integral of a function f (z) from a fixed point z 1 to
another fixed point z 2 depends, in general, on the path that is taken, there are certain
functions whose integrals from z 1 to z 2 have values that are independent of path.
Recall statements (a) and (b) at the end of Sec. 45. Those statements also remind
us of the fact that the values of integrals around closed paths are sometimes, but not
always, zero. Our next theorem is useful in determining when integration is independent
of path and, moreover, when an integral around a closed path has value zero.
The theorem contains an extension of the fundamental theorem of calculus that
simplifies the evaluation of many contour integrals. The extension involves the concept
of an antiderivative of a continuous function f (z) on a domain D, or a function F(z)
such that F  (z) = f (z) for all z in D. Note that an antiderivative is, of necessity,
an analytic function. Note, too, that an antiderivative of a given function f (z) is
unique except for an additive constant. This is because the derivative of the difference
F(z) − G(z) of any two such antiderivatives is zero; and, according to the theorem

∗
These functions are actually polynomials in x. The are known as Legendre polynomials and are
important in applied mathematics. See, for example, the authors’ book (2012) that is listed in Appendix
1. The expression for Pn (x) used in Exercise 7 is sometimes called Laplace’s first integral form.
SEC. 48 ANTIDERIVATIVES 141

in Sec. 25, an analytic function is constant in a domain D when its derivative is zero
throughout D.

Theorem. Suppose that a function f (z) is continuous in a domain D. If any one

of the following statements is true, then so are the others:
(a) f (z) has an antiderivative F(z) throughout D;
(b) the integrals of f (z) along contours lying entirely in D and extending from any
fixed point z 1 to any fixed point z 2 all have the same value, namely
 z2 z 2
f (z) dz = F(z) = F(z 2 ) − F(z 1 )
z1 z1

where F(z) is the antiderivative in statement (a);

(c) the integrals of f (z) around closed contours lying entirely in D all have value
zero.

It should be emphasized that the theorem does not claim that any of these state-
ments is true for a given function f (z). It says only that all of them are true or that
none of them is true. The next section is devoted to the proof of the theorem and can
be easily skipped by a reader who wishes to get on with other important aspects of
integration theory. But we include here a number of examples illustrating how the
theorem can be used.

EXAMPLE 1. The continuous function f (z) = eπ z evidently has an antideriva-

tive F(z) = eπ z /π throughout the finite plane. Hence
 i/2
i/2
πz eπ z 1 iπ/2  1 1
e dz = = e − eiπ = (i + 1) = (1 + i).
i π i π π π

EXAMPLE 2. The function f (z) = 1/z 2 , which is continuous everywhere ex-

cept at the origin, has an antiderivative F(z) = − 1/z in the domain |z| > 0, consisting
of the entire plane with the origin deleted. Consequently,

dz
2
=0
C z

when C is the positively oriented unit circle z = eiθ (−π ≤ θ ≤ π) about the origin.
Note that the integral of the function f (z) = 1/z around the same circle cannot
be evaluated in a similar way. For, although the derivative of any branch F(z) of log z
is 1/z (Sec. 33), F(z) is not differentiable, or even defined, along its branch cut. In
particular, if a ray θ = α from the origin is used to form the branch cut, F  (z) fails to
exist at the point where that ray intersects the circle C (see Fig. 51). So C does not lie
in any domain throughout which F  (z) = 1/z, and one cannot make direct use of an
142 INTEGRALS CHAP. 4

i
C

O x

–i

FIGURE 51

antiderivative. But Example 3, just below, illustrates how a combination of two different
antiderivatives can be used to evaluate f (z) = 1/z around C.

EXAMPLE 3. Let C1 denote the right half

 
π π
z = eiθ − ≤θ ≤
2 2
of the circle C in Fig. 51. The principal branch
Log z = ln r + i (r > 0, −π < < π)
of the logarithmic function serves as an antiderivative of the function 1/z in the eval-
uation of the integral of 1/z along C1 (Fig. 52):
  i
dz dz i
= = Log z −i = Log i − Log (−i)
z z
C1
−i   
π π
= ln 1 + i − ln 1 − i = πi.
2 2

y
i
C1

O x

–i
FIGURE 52

Next let C2 denote the left half

 
π 3π
z=e iθ
≤θ ≤
2 2
of the same circle C and consider the branch
log z = ln r + iθ (r > 0, 0 < θ < 2π)
SEC. 48 ANTIDERIVATIVES 143

of the logarithmic function (Fig. 53). One can write

  −i −i
dz dz
= = log z = log(−i) − log i
C2 z i z i
   
3π π
= ln 1 + i − ln 1 + i = πi.
2 2

y
i
C2

O x

–i
FIGURE 53

The value of the integral of 1/z around the entire circle C = C1 + C2 is thus
obtained:
  
dz dz dz
= + = πi + πi = 2πi.
C z C1 z C2 z

EXAMPLE 4. Let us use an antiderivative to evaluate the integral


(1) z 1/2 dz,
C1

where the integrand is the branch

 
1 √
(2) f (z) = z = exp
1/2
log z = r eiθ/2 (r > 0, 0 < θ < 2π)
2
of the square root function and where C1 is any contour from z = −3 to z = 3 that,
except for its end points, lies above the x axis (Fig. 54). Although the integrand is
piecewise continuous on C1 , and the integral therefore exists, the branch (2) of z 1/2 is

C1

–3 O 3 x

C2
FIGURE 54
148 INTEGRALS CHAP. 4

50. CAUCHY–GOURSAT THEOREM

In Sec. 48, we saw that when a continuous function f has an antiderivative on a
domain D, the integral of f (z) around any given closed contour C lying entirely in
D has value zero. In this section, we present a theorem giving other conditions on a
function f which ensure that the value of the integral of f (z) around a simple closed
contour (Sec. 43) is zero. The theorem is central to the theory of functions of a complex
variable; and some modifications of it, involving certain special types of domains, will
be given in Secs. 52 and 53.
We let C denote a simple closed contour z = z(t) (a ≤ t ≤ b), described in
the positive sense (counterclockwise), and we assume that f is analytic at each point
interior to and on C. According to Sec. 44,
  b
(1) f (z) dz = f [z(t)]z  (t) dt;
C a
and if
f (z) = u(x, y) + iv(x, y) and z(t) = x(t) + i y(t),
the integrand f [z(t)]z  (t) in expression (1) is the product of the functions
u[x(t), y(t)] + iv[x(t), y(t)], x  (t) + i y  (t)
of the real variable t. Thus
  b  b
(2) f (z) dz = (ux  − vy  ) dt + i (vx  + uy  ) dt.
C a a
In terms of line integrals of real-valued functions of two real variables, then,
  
(3) f (z) dz = u d x − v dy + i v d x + u dy.
C C C
Observe that expression (3) can be obtained formally by replacing f (z) and dz on the
left with the binomials
u + iv and d x + i dy,
respectively, and expanding their product. Expression (3) is, of course, also valid
when C is any contour, not necessarily a simple closed one, and when f [z(t)] is only
piecewise continuous on it.
We next recall a result from calculus that enables us to express the line integrals
on the right in equation (3) as double integrals. Suppose that two real-valued functions
P(x, y) and Q(x, y), together with their first-order partial derivatives, are continuous
throughout the closed region R consisting of all points interior to and on the simple
closed contour C. Green’s theorem states that
  
Pd x + Q dy = (Q x − Py ) d A.
C R
Now f is continuous on R, since it is analytic there. Hence the functions u and
v are also continuous on R. Likewise, if the derivative f  of f is continuous on R, so
SEC. 50 CAUCHY–GOURSAT THEOREM 149

are the first-order partial derivatives of u and v. Green’s theorem then enables us to
rewrite equation (3) as
    
(4) f (z) dz = (−vx − u y ) d A + i (u x − v y ) d A.
C R R
But, in view of the Cauchy–Riemann equations
u x = vy , u y = −vx ,
the integrands of these two double integrals are zero throughout R. So when f is
analytic in R and f  is continuous there,

(5) f (z) dz = 0.
C
This result was obtained by Cauchy in the early part of the nineteenth century.
Note that once it has been established that the value of this integral is zero, the
orientation of C is immaterial. That is, statement (5) is also true if C is taken in the
clockwise direction, since then
 
f (z) dz = − f (z) dz = 0.
C −C

EXAMPLE. If C is any simple closed contour, in either direction, then


sin(z 2 ) dz = 0.
C

This is because the composite function f (z) = sin(z 2 ) is analytic everywhere and its
derivative f  (z) = 2z cos(z 2 ) is continuous everywhere.

Goursat∗ was the first to prove that the condition of continuity on f  can be omitted.
Its removal is important and will allow us to show, for example, that the derivative f 
of an analytic function f is analytic without having to assume the continuity of f  ,
which follows as a consequence. We now state the revised form of Cauchy’s result,
which is known as the Cauchy-Goursat theorem.

Theorem. If a function f is analytic at all points interior to and on a simple

closed contour C, then

f (z) dz = 0.
C

The proof is presented in the next section, where, to be specific, we assume that
C is positively oriented. The reader who wishes to accept this theorem without proof
may pass directly to Sec. 52.

∗
E. Goursat (1858–1936), pronounced gour-sah .
154 INTEGRALS CHAP. 4

Conclusion
We now use the theorem in Sec. 47 to find an upper bound for each modulus on the
right in inequality (8). To do this, we first recall that each C j coincides either entirely
or partially with the boundary of a square. In either case, we let s j denote the length
of a side of the square. Since, in the jth integral, both the variable z and the point z j
lie in that square,
√
|z − z j | ≤ 2s j .
In view of inequality (5), then, we know that each integrand on the right in inequality
(8) satisfies the condition
√
(9) |(z − z j )δ j (z)| = |z − z j | |δ j (z)| < 2s j ε.
As for the length of the path C j , it is 4s j if C j is the boundary of a square. In that case,
we let A j denote the area of the square and observe that



√ √

(10)
(z − z j )δ j (z) dz
< 2s j ε4s j = 4 2A j ε.

Cj

If C j is the boundary of a partial square, its length does not exceed 4s j + L j , where
L j is the length of that part of C j which is also a part of C. Again letting A j denote
the area of the full square, we find that



√ √ √

(11)
(z − z j )δ j (z) dz
< 2s j ε(4s j + L j ) < 4 2A j ε + 2S L j ε,

Cj

where S is the length of a side of some square that encloses the entire contour C as
well as all of the squares originally used in covering R (Fig. 59). Note that the sum of
all the A j ’s does not exceed S 2 .
If L denotes the length of C, it now follows from inequalities (8), (10), and (11)
that



√ √

f (z) dz
< (4 2S 2 + 2S L)ε.

Since the value of the positive number ε is arbitrary, we can choose it so that the right-
hand side of this last inequality is as small as we please. The left-hand side, which
is independent of ε, must therefore be equal to zero ; and statement (3) follows. This
completes the proof of the Cauchy–Goursat theorem.

52. SIMPLY CONNECTED DOMAINS

A simply connected domain D is a domain such that every simple closed contour within
it encloses only points of D. The set of points interior to a simple closed contour is an
example. The annular domain between two concentric circles is, however, not simply
connected. Domains that are not simply connected are discussed in the next section.
SEC. 52 SIMPLY CONNECTED DOMAINS 155

The closed contour in the Cauchy–Goursat theorem (Sec. 50) need not be simple
when the theorem is adapted to simply connected domains. More precisely, the contour
can actually cross itself. The following theorem allows for this possibility.
Theorem. If a function f is analytic throughout a simply connected domain D,
then

(1) f (z) dz = 0
C
for every closed contour C lying in D.
The proof is easy if C is a simple closed contour or if it is a closed contour that
intersects itself a finite number of times. For if C is simple and lies in D, the function
f is analytic at each point interior to and on C; and the Cauchy–Goursat theorem
ensures that equation (1) holds. Furthermore, if C is closed but intersects itself a finite
number of times, it consists of a finite number of simple closed contours, and the
Cauchy-Goursat theorem can again be applied. This is illustrated in Fig. 60, where
two simple closed contours C1 and C2 make up C. Since the values of the integrals
around C1 and C2 are zero, regardless of their orientations,
  
f (z) dz = f (z) dz + f (z) dz = 0.
C C1 C2

C1

C2

O x FIGURE 60
Subtleties arise if the closed contour has an infinite number of self-intersection
points. One method that can sometimes be used to show that the theorem still applies
is illustrated in Exercise 5, Sec. 53.∗

EXAMPLE. If C denotes any closed contour lying in the open disk |z| < 2
(Fig. 61), then

sin z
dz = 0.
C (z + 9)
2 5

∗
For a proof of the theorem involving more general paths of finite length, see, for example, Secs. 63–65
in Vol. I of the book by Markushevich that is cited in Appendix 1.
156 INTEGRALS CHAP. 4

O 2 x

FIGURE 61

This is because the disk is a simply connected domain and the two singularities z = ±3i
of the integrand are exterior to the disk.

Corollary 1. A function f that is analytic throughout a simply connected domain

D must have an antiderivative everywhere in D.

We begin the proof of this corollary with the observation that a function f is
continuous on a domain D when it is analytic there. Consequently, since equation (1)
holds for the function in the hypothesis of this corollary and for each closed contour
C in D, f has an antiderivative throughout D, according to the theorem in Sec. 48.

Corollary 2. Entire functions always possess antiderivatives.

This corollary is an immediate consequence of Corollary 1 and the fact that the
finite plane is simply connected.

53. MULTIPLY CONNECTED DOMAINS

A domain that is not simply connected (Sec. 52) is said to be multiply connected.
The following theorem is an adaptation of the Cauchy–Goursat theorem to multiply
connected domains. While the statement of the theorem involves n contours labeled
Ck (k = 1, 2, . . . , n), we shall be guided in the proof by Fig. 62, where n = 2.

Theorem. Suppose that

(a) C is a simple closed contour, described in the counterclockwise direction;
(b) Ck (k = 1, 2, . . . , n) are simple closed contours interior to C, all described in
the clockwise direction, that are disjoint and whose interiors have no points in
common (Fig. 62).
SEC. 53 MULTIPLY CONNECTED DOMAINS 157

If a function f is analytic on all of these contours and throughout the multiply connected
domain consisting of the points inside C and exterior to each Ck , then
 n 
(1) f (z) dz + f (z) dz = 0.
C k=1 Ck

y
C
L2 L3
L1 C1
C2

O x FIGURE 62

Note that in equation (1), the direction of each path of integration is such that the
multiply connected domain lies to the left of that path.
To prove the theorem, we introduce a polygonal path L 1 , consisting of a finite
number of line segments joined end to end, to connect the outer contour C to the inner
contour C1 . We introduce another polygonal path L 2 which connects C1 to C2 ; and we
continue in this manner, with L n+1 connecting Cn to C. As indicated by the single-
barbed arrows in Fig. 62, two simple closed contours 1 and 2 can be formed, each
consisting of polygonal paths L k or −L k and pieces of C and Ck and each described in
such a direction that the points enclosed by them lie to the left. The Cauchy–Goursat
theorem can now be applied to f on 1 and 2 , and the sum of the values of the integrals
over those contours is found to be zero. Since the integrals in opposite directions along
each path L k cancel, only the integrals along C and the Ck remain; and we arrive at
statement (1).

Corollary. Let C1 and C2 denote positively oriented simple closed contours,

where C1 is interior to C2 (Fig. 63). If a function f is analytic in the closed region
consisting of those contours and all points between them, then
 
(2) f (z) dz = f (z) dz.
C1 C2

This corollary is known as the principle of deformation of paths since it tells us

that if C1 is continuously deformed into C2 , always passing through points at which
f is analytic, then the value of f over C1 never changes. To verify this corollary. we
need only observe how it follows from the theorem that
 
f (z) dz+ f (z) dz = 0.
C2 −C1
But this the same as equation (2).
158 INTEGRALS CHAP. 4

C2

C1

O x FIGURE 63

EXAMPLE. When C is any positively oriented simple closed contour surround-

ing the origin, the corollary can be used to show that

dz
= 2πi.
C z

This is done by constructing a positively oriented circle C0 with center at the origin
and radius so small that C0 lies entirely inside C (Fig. 64). Since (see Exercise 13,
Sec. 46)

dz
= 2πi
C0 z

and since 1/z is analytic everywhere except at z = 0, the desired result follows.
Note that the radius of C0 could equally well have been so large that C lies entirely
inside C0 .

C0

O x

FIGURE 64
SEC. 53 MULTIPLY CONNECTED DOMAINS 159

EXERCISES
1. Apply the Cauchy–Goursat theorem to show that

f (z) dz = 0
C
when the contour C is the unit circle |z| = 1, in either direction, and when
z2 1
(a) f (z) = ; (b) f (z) = z e−z ; (c) f (z) = 2 ;
z+3 z + 2z +2
(d) f (z) = sech z; (e) f (z) = tan z; (f) f (z) = Log(z + 2).
2. Let C1 denote the positively oriented boundary of the square whose sides lie along the
lines x = ±1, y = ±1 and let C2 be the positively oriented circle |z| = 4 (Fig. 65). With
the aid of the corollary in Sec. 53, point out why
 
f (z) dz = f (z) dz
C1 C2

when
1 z+2 (c) f (z) =
z
(a) f (z) = ; (b) f (z) = ; 1 − ez
.
3z 2 + 1 sin(z/2)

C2

C1

1 4 x

FIGURE 65

3. If C0 denotes a positively oriented circle |z − z 0 | = R, then

 
0 when n = ±1, ±2, . . . ,
(z − z 0 )n−1 dz =
C0 2πi when n = 0,
according to Exercise 13, Sec. 46. Use that result and the corollary in Sec. 53 to show
that if C is the boundary of the rectangle 0 ≤ x ≤ 3, 0 ≤ y ≤ 2, described in the positive
sense, then
 
0 when n = ±1, ±2, . . . ,
(z − 2 − i)n−1 dz =
C 2πi when n = 0.
4. Use the following method to derive the integration formula
 ∞ √
−x 2 π −b2
e cos 2bx d x = e (b > 0).
0 2
160 INTEGRALS CHAP. 4

(a) Show that the sum of the integrals of e−z along the lower and upper horizontal legs
2

of the rectangular path in Fig. 66 can be written

 a  a
e−x d x − 2eb e−x cos 2bx d x
2 2 2
2
0 0
and that the sum of the integrals along the vertical legs on the right and left can be
written
 b  b
ie−a e y e−i2ay dy − ie−a
2 2 2 2
e y ei2ay dy.
0 0
Thus, with the aid of the Cauchy–Goursat theorem, show that
 a  a  b
e−x cos 2bx d x = e−b e−x d x + e−(a +b2 )
2 2 2 2 2
e y sin 2ay dy.
0 0 0

y
– a + bi a + bi

–a O a x
FIGURE 66
∗
(b) By accepting the fact that
 ∞
√
π
e−x d x =
2

0 2
and observing that




b
b

e y sin 2ay dy

≤
2
e y dy,
2

0 0
obtain the desired integration formula by letting a tend to infinity in the equation at
the end of part (a).
5. According to Exercise 6, Sec. 43, the path C1 from the origin to the point z = 1 along
the graph of the function defined by means of the equations

x 3 sin (π/x) when 0 < x ≤ 1,
y(x) =
0 when x = 0
is a smooth arc that intersects the real axis an infinite number of times. Let C2 denote
the line segment along the real axis from z = 1 back to the origin, and let C3 denote
any smooth arc from the origin to z = 1 that does not intersect itself and has only its
end points in common with the arcs C1 and C2 (Fig. 67). Apply the Cauchy–Goursat

∗
The usual way to evaluate this integral is by writing its square as
 ∞  ∞  ∞  ∞
e−x d x e−y dy = e−(x +y 2 )
2 2 2
d xd y
0 0 0 0
and then evaluating this iterated integral by changing to polar coordinates. Details are given in, for
example, A. E. Taylor and W. R. Mann, “Advanced Calculus,” 3d ed., pp. 680–681, 1983.