Contour Integration
On the real line, the definite
integral adds up a function along an interval. In the complex plane there is no
single "interval" between two points — there are infinitely many paths. So a
complex integral is taken along a chosen curve \gamma, called a
contour, and is written
\int_\gamma f(z)\,dz.
Everything in complex analysis — Cauchy's theorem, the integral formula, residues — is built on
this one idea. Let us define it precisely and then compute the single integral the whole subject
turns on.
Defining the contour integral
Describe the contour by a parametrisation z(t) for
t \in [a, b]: as t runs from
a to b, the point
z(t) traces out the curve. Then
\int_\gamma f(z)\,dz = \int_a^b f\bigl(z(t)\bigr)\,z'(t)\,dt.
The factor z'(t)\,dt is just dz expressed
in the parameter — the chain rule, made into a recipe. The right-hand side is an ordinary
integral of a complex-valued function of the real variable t, which
we evaluate by integrating its real and imaginary parts separately.
A worked example: integrating over a circle
Let \gamma be the circle of radius R
centred at the origin, traced once anticlockwise. The natural parametrisation is
z(t) = R\,e^{it}, \qquad t \in [0, 2\pi].
Step 1 — find dz. Differentiate the
parametrisation:
z'(t) = iR\,e^{it}, \qquad\text{so}\qquad dz = iR\,e^{it}\,dt.
Step 2 — substitute. For a concrete f we feed
z(t) into f and multiply by
dz. This circle, with its tidy
e^{it} form, is exactly the contour we now use to crack the most
important integral in the theory.
The cornerstone: \oint z^n\,dz around the unit circle
Take the unit circle, z(t) = e^{it} on
[0, 2\pi], so dz = i\,e^{it}\,dt, and
integrate the power f(z) = z^n for an integer
n. (The closed-loop symbol \oint just
marks that \gamma is a closed contour.)
Step 1 — substitute. Since
z^n = e^{int}:
\oint_{|z|=1} z^n\,dz = \int_0^{2\pi} e^{int}\,\bigl(i\,e^{it}\bigr)\,dt = i\int_0^{2\pi} e^{i(n+1)t}\,dt.
Step 2 — the case n \ne -1. Then
n + 1 \ne 0, and we are integrating a complex exponential over a full
period. Its antiderivative is \frac{e^{i(n+1)t}}{i(n+1)}, and
e^{i(n+1)t} returns to its start after one full turn:
i\int_0^{2\pi} e^{i(n+1)t}\,dt = i\left[\frac{e^{i(n+1)t}}{i(n+1)}\right]_0^{2\pi} = \frac{e^{i(n+1)2\pi} - 1}{n+1} = \frac{1 - 1}{n+1} = 0.
Step 3 — the case n = -1. Now
n + 1 = 0, so the exponent vanishes and the integrand is just the
constant i:
i\int_0^{2\pi} e^{0}\,dt = i\int_0^{2\pi} 1\,dt = i\,(2\pi) = 2\pi i.
So a single power behaves in two utterly different ways: every power except
z^{-1} integrates to zero around the loop, while
1/z alone leaves behind 2\pi i.
Integrating z^n once anticlockwise around the unit circle gives:
-
\displaystyle\oint_{|z|=1} z^n\,dz = 0 for every integer
n \ne -1 — a full period of
e^{i(n+1)t} averages to nothing;
-
\displaystyle\oint_{|z|=1} \frac{dz}{z} = 2\pi i — the lone
exception, because the integrand collapses to the constant i.
The number 2\pi i is not an accident of this particular circle. As
we will see, deforming the contour does not change a contour integral of a holomorphic
function, and 1/z is holomorphic everywhere except at the origin.
What survives is a record of how many times — and which way — the path winds around that one
bad point. The single term z^{-1} is the detector for it, and
2\pi i is the signal it returns. The entire theory of
residues grows out of this one computation.
Watch a point trace the contour
Turn the t slider and the radius R: the
point z(t) = R\,e^{it} rides around the circle, and the arrow shows
dz = z'(t)\,dt = iR\,e^{it}\,dt, the tangent direction of travel —
always a quarter-turn ahead of the radius. That tangent is what
dz measures.