F Appendix F Notes on Complex Variables

These notes are meant to accompany a graduate level physics course, to provide a
basic introduction to the necessary concepts in complex analysis. They are not complete,
nor are any of the proofs considered rigorous. The immediate goal is to carry through
enough of the work needed to explain the Cauchy residue theorem.

F.1 Complex Numbers and Complex Functions

A complex number z can be written as

z = x + iy or z = reiφ with r ≥ 0
√
where i = −1, and x, y, r, and φ are real numbers. Clearly, x = r cos φ and y = r sin φ,
leading to a description in terms of the “complex plane.” The complex conjugate of z is

z∗ = x − iy or z∗ = re−iφ .
√ 
The “modulus” of z is |z| ≡ z∗ z = r = x2 + y2 , and φ is often called the “phase” of z.
A complex function f (z) typically returns a complex number. Generically, we write

f (z) = u(x, y) + iv(x, y) (F.1)

for purposes of proofs or illustrations. The behavior of the (real) functions u(x, y) and
v(x, y) are critical for classifying complex functions, as seen when we consider taking
derivatives.

F.2 Differentiation and Analyticity

We define the derivative f  (z) = df/dz of a complex function f (z) in the same way as we do
for the derivatives of real functions. That is, for z0 ≡ x0 + iy0 ,
df f (z) − f (z0 )
f  (z0 ) = = lim .
dz z=z0
z→z0 z − z0

535
536 Appendix F

However, there is clearly an ambiguity, depending on whether we approach z0 along the

line y = y0 or along x = x0 . (Of course, we could also say the ambiguity is along any line
of constant φ = φ0 , but it is sufficient to consider just two orthogonal directions.) That is,
u(x, y0 ) − u(x0 , y0 ) v(x, y0 ) − v(x0 , y0 ) ∂ u ∂v
f  (z0 ) = lim + i lim = +i
x→x0 x − x0 x→x0 x − x0 ∂x ∂x
or
u(x0 , y) − u(x0 , y0 ) v(x0 , y) − v(x0 , y0 ) ∂u ∂v
f  (z0 ) = lim + i lim = −i + .
y→y0 iy − iy0 y→y0 iy − iy0 ∂y ∂y
Therefore, in order to remove the ambiguity and have a consistent definition of the
derivative,
∂u ∂v ∂u ∂v
= and =− . (F.2)
∂x ∂y ∂y ∂x
These are called the Cauchy–Riemann conditions. A function f (z) which satisfies these
rather restrictive conditions is called analytic. Indeed, analytic functions have very many
applications in physics, and we will merely scratch the surface here.
For example, the function f (z) = ez = ex (cos y + i sin y) is analytic. This is easy to prove.
Putting u(x, y) = ex cos y and v(x, y) = ex sin y,
∂u ∂v ∂u ∂v
= ex cos y = and = −ex sin y = −
∂x ∂y ∂y ∂x
so the Cauchy–Riemann conditions (F.2) are satisfied.
It is simple to show that f (z) = az is analytic, where a is a complex constant. It is also
not hard to show that the product of two analytic functions is analytic, so any function of
the form f (z) = an zn , where n is a nonnegative integer, is also analytic. Of course, any sum
of analytic functions is analytic, so we see that any polynomial in z is analytic in the entire
complex plane.
These examples beg the question: If a function f (z) can be written explicitly in terms of
z, is it analytic? The answer is “Yes.” To see this, realize that instead of x and y, we could
always write a complex function in terms of z and z∗ using x = (z+z∗ )/2 and y = (z−z∗ )/2i.
Now consider
     
∂f ∂f ∂x ∂f ∂y ∂u ∂v 1 ∂u ∂v 1
= + = + i + + i −
∂ z∗ ∂ x ∂ z∗ ∂ y ∂ z∗ ∂x ∂x 2 ∂y ∂y 2i
   
1 ∂u ∂v i ∂u ∂v
= − + + =0
2 ∂x ∂y 2 ∂y ∂x
so long as the Cauchy–Riemann conditions (F.2) are satisfied. That is, if the expression for
f (z) contains only z (and not z∗ ) then the function is analytic.
There is some common terminology. A function f (z) need not be analytic in the entire
complex plane. (If it is, we called the function “entire.”) If it is analytic at a point z0 then
we call that a “regular point.” Otherwise, z0 is called a “singular point.” Much of our
discussion of complex integration will focus on the notion of singular points.
537 Notes on Complex Variables

F.3 Integration and Series Expansion

Similarly to differentiation, we approach integration of complex functions the same way as

with real functions, but we need to be aware that there is now an arbitrariness of the “path”
of integration. With dz = dx + idy and using using (F.1), we have
 z2  z2  z2  z2  z2
f (z)dz = (u dx − v dy) + i (v dx + u dy) = A · dx + i B · dx (F.3)
z1 z1 z1 z1 z1

where A = ux̂ − vŷ and B = vx̂ + uŷ. So, we can now think of the two integrals on the right
as real integrals of vector functions over curves in the xy plane. However, if we invoke
Stokes’s theorem, these become integrals of the curls, and using (F.2), we find
   
∂v ∂u ∂u ∂v
∇×A = − − =0 and ∇×B = − =0 (F.4)
∂x ∂y ∂x ∂y
and each of the two integrals on the right in (F.3) is path independent. Hence, the integral of
an analytic complex function f (z) is path independent and can be unambiguously defined.
From here on, we assume all functions to be analytic unless explicitly noted
otherwise. It is obvious from (F.3) that, when integrating around a closed path C,
,
f (z)dz = 0
C

which is known as the Cauchy–Goursat theorem. We will be exploring circumstances

where the integrand is explicitly singular at one or more points.
For the first example, we prove the Cauchy integral formula, namely
,
1 f (z)
f (z0 ) = dz (F.5)
2πi C − z0
z
where C is a closed contour in the complex plane that contains the point z0 and traversed
in the counterclockwise direction. We can break up a contour C into something that looks
like Figure F.1. Notice that C0 is a tiny circular contour around the singular point, but in
the clockwise direction. That is, we replace C with limC0 →0 (C + C0 ). However, for C0 = 0,
the new contour C does not include the singular point, so by (F.4) we write (F.5) as
,
1 f (z)
f (z0 ) = − dz. (F.6)
2πi C0 z − z0

The shrinking contour C0 is parameterized as z − z0 = reiφ for r → 0 and φ = 2π → 0, so

,  0
1 f (z) 1 1 1
− dz = − f (z0 ) ireiφ dφ = − f (z0 )i(−2π) = f (z0 )
2πi C0 z − z0 2πi 2π reiφ 2πi
proving the Cauchy integral formula (F.5). A trivial, but suggestive, rewriting of (F.5) gives
,
1 f (ξ)
f (z) = dξ (F.7)
2πi C ξ−z
 538 Appendix F

Fig. F.1 A contour in the complex plane for proving the Cauchy integral formula.

which leads to a convenient way to write the derivatives of a complex function, namely
,
dn f n! f (ξ)
f(n) (z) = = dξ. (F.8)
C (ξ − z)
dz n 2πi n+1

Now consider the series expansion of an analytic function f (z). We would naturally write
∞
f (z) = f (z0 ) + f  (z0 )(z − z0 ) + · · · = ∑ an (z − z0 )n (F.9)
n=0

where
,
1 (n) 1 f (ξ)
an ≡ f (z0 ) = dξ. (F.10)
C (ξ − z0 )
n! 2πi n+1

Such a Taylor series expansion works out as expected, but the curve C specifies regions in
which the series converges.
This idea can be expanded to include −∞ ≤ n ≤ ∞, still using the right side of (F.10) to
define an , and with modified regions of convergence. Such an expansion is called a Laurent
series. It clearly is not, in general, an analytic function because of poles that appear for
n < 0. These, however, lead us to one of the most important theorems of complex analysis,
so far as mathematical physics is concerned.

F.4 The Cauchy Residue Theorem

Let g(z) have an isolated singularity at z = z0 . If the Laurent expansion can be written as
∞ ∞
b1
g(z) = ∑ an (z − z0 )n = + ∑ an (z − z0 )n
z − z0 n=0
(F.11)
n=−∞
539 Notes on Complex Variables

then we say that g(z) has a “simple pole” at z = z0 . Higher-order poles are possible, but we
are not going to consider them here.
Consider a contour C within the radius of convergence of g(z). Separate the integral of
g(z) around this contour into two terms, one for each of the two terms on the right in (F.11).
The second term is a polynomial in z; therefore it is analytic and the integral is zero. Recall
that we reduced the contour to a small circle around the pole in order to prove the Cauchy
integral formula. We can do the same thing here, and
, ,
1
g(z)dz = b1 = 2πi b1 . (F.12)
C C z − z0

We refer to b1 as the “residue” of g(z0 ), sometimes written as Res[g(z0 )]. We have

Res[g(z0 )] = lim (z − z0 )g(z)

z→z0

for a simple pole at z0 .

If there is more than one simple pole within the contour C, this result is easy to
generalize. Instead of redrawing the contour with a small loop about the single pole, do
it for all N poles within the contour. The result is clearly
, N
g(z)dz = 2πi ∑ Res[g(zk )]. (F.13)
C k=1

We refer to this as the Cauchy residue theorem. It is widely used in mathematical physics.
The usefulness of the residue theorem can be illustrated in many ways, but here is one
important example. It is a warm-up to evaluating the integral in (6.44). The exercise is to
evaluate the integral
 ∞  ∞
eika eika
I= k dk = lim k dk (F.14)
−∞ q − k q2 − k2 + iε
2 2 ε→0 −∞

where k, a, q, and ε > 0 are all real variables. We use the second version above because
this moves the singularities at k = ±q off the real axis. To be sure, we could have moved
off the real axis by using −iε instead of +iε, and in fact, this would give us a different
answer. A physical rationale is needed to justify one sign or the other. Leave that for a
physics course.
We can evaluate (F.14) using contour integration by first allowing k to be complex and
then noting that eikx → 0 as Im(k) → +∞. Therefore (F.14) can be rewritten as an integral
over a semicircular contour C that runs (counterclockwise) along the Re(k) axis and closes
as a semicircle in the Im(k) > 0 plane. Then for ε → 0, the integrand in (F.14) has poles at

  
ε
k=± q2 − iε = ±q 1 − i 2 ⇒ ±q ∓ iε
q

where we redefine ε (with q > 0) so that it is still small and has the same sign.
540 Appendix F

The pole at k = k0 ≡ +q − iε does not matter to us, since it is outside the integration
contour. However, the pole at k = −k0 = −q + iε is inside, so we use the residue theorem
to write
,
eika
I = lim k dk
ε→0 C (k − k0 )(k + k0 )

eika
= lim 2πi k = πi lim e−ik0 a = πie−iqa .
ε→0 k − k0 k=−k0 ε→0