The Residue Theorem

How can we compute the integral of eix/(1+x2)e^{ix}/(1+x^2) over the entire real line? We want to do this without ever finding an antiderivative. The answer lies in a single remarkable fact. A holomorphic function's behavior near an isolated singularity is captured entirely by one complex number. This number is its residue. The integral of the function around any closed loop equals 2πi2\pi i times the sum of these numbers, provided the loop encloses such singularities. This is the Residue Theorem, the computational engine of complex analysis. The theorem's power comes from a dramatic reduction of information. A holomorphic function on a punctured disk around a singularity has infinitely many Laurent coefficients. Yet only one of them survives integration around a small loop. That coefficient is the one for (z−z0)−1(z-z_0)^{-1}. Every other term integrates to zero. This single surviving number, the residue, packages all the information needed to compute contour integrals. By extension, it also handles a vast class of real integrals. These are integrals that resist elementary calculus. The big picture is this: the Residue Theorem is the natural culmination of a philosophy. That philosophy runs through complex analysis, moving from local to global. Cauchy's Integral Formula already tells us that values of a holomorphic function are determined by a boundary integral. The Residue Theorem generalizes this to functions with isolated singularities. It replaces "the function itself" with "the singular part of its Laurent expansion." It sits between Singularities and Classification, which sets up the local Laurent theory, and later tools like The Argument Principle and Rouche's Theorem. Those tools specialize the residue calculus to counting zeros and poles. This article gives the precise definition of a residue. It proves the Residue Theorem in full rigor. The proof covers a general finite union of simple closed contours. It develops the standard techniques for computing residues at poles of any order. It also works through asymmetric examples showing the theorem's reach. These include a first glimpse of how it evaluates genuinely difficult real integrals.

Key takeaways
  • A residue Res⁡(f,z0)\operatorname{Res}(f, z_0) is the coefficient of (z−z0)−1(z-z_0)^{-1} in the Laurent expansion of ff around an isolated singularity z0z_0.
  • The Residue Theorem says ∮γf(z) dz=2πi∑Res⁡(f,zk)\oint_\gamma f(z)\,dz = 2\pi i \sum \operatorname{Res}(f, z_k), summed over singularities enclosed by γ\gamma.
  • It generalizes Cauchy's Integral Formula, which is the special case of a removable-looking pole handled by direct expansion.
  • For a pole of order mm, the residue is computed by a explicit derivative formula, with the simple-pole case Res⁡(f,z0)=lim⁡z→z0(z−z0)f(z)\operatorname{Res}(f,z_0) = \lim_{z\to z_0}(z-z_0)f(z) being the workhorse.
  • The theorem converts hard real integrals (rational functions, trigonometric integrals, Fourier-type integrals) into algebraic residue computations.
  • Key idea: all the analytic complexity of a singularity collapses into one number that a closed contour integral can detect.

Where this lives today

The Residue Theorem is the standard machine for evaluating definite real integrals that lack elementary antiderivatives — improper integrals of rational functions, Fourier-type integrals ∫f(x)eiωx dx\int f(x) e^{i\omega x}\,dx arising in signal processing, and integrals over trigonometric functions reduced to unit-circle contour integrals — a systematic technique developed further in Evaluation of Real Integrals. In physics, residue calculus computes dispersion relations, Green's functions for differential operators, and propagators in quantum field theory via contour deformation in the complex energy or momentum plane. In engineering, the residue theorem underlies the inverse Laplace and Z-transforms, recovering time-domain signals from their transform via residues at poles of a transfer function — central to control theory and digital filter design. In analytic number theory, residues at poles of zeta and L-functions (e.g., the simple pole of ζ(s)\zeta(s) at s=1s=1) drive the proof of the Prime Number Theorem through Perron's formula, a contour-integral representation of summatory functions. The closely related The Argument Principle and Rouche's Theorem specializes residue calculus to counting zeros and poles of meromorphic functions, with applications to stability analysis (Nyquist criterion) in control systems.

Ask the tutor: How do we choose the right contour shape (semicircle, rectangle, keyhole) to evaluate a specific real integral using the Residue Theorem?

DefinitionResidue and the Residue Theorem

Let ff be holomorphic on a punctured disk D∗(z0,r)={z:0<∣z−z0∣<r}D^*(z_0, r) = \{z : 0 < |z - z_0| < r\}, so that z0z_0 is an isolated singularity of ff. By the Laurent expansion theorem, ff has a unique representation

f(z)=∑n=−∞∞an(z−z0)n,0<∣z−z0∣<r,f(z) = \sum_{n=-\infty}^{\infty} a_n (z - z_0)^n, \qquad 0 < |z-z_0| < r,

convergent on the punctured disk. The residue of ff at z0z_0 is the coefficient

Res⁡(f,z0):=a−1.\operatorname{Res}(f, z_0) := a_{-1}.

A pole of order m≥1m \geq 1 is an isolated singularity at which a−n=0a_{-n} = 0 for n>mn > m and a−m≠0a_{-m} \neq 0; a pole of order 11 is called simple. A domain Ω⊆C\Omega \subseteq \mathbb{C} is called a region with positively oriented piecewise-C1C^1 boundary ∂Ω\partial\Omega if ∂Ω\partial\Omega consists of finitely many piecewise-C1C^1 simple closed curves, pairwise disjoint, oriented so that Ω\Omega lies to the left as the boundary is traversed (the standard counterclockwise convention for the outer boundary, clockwise for inner boundary components/holes).

Intuition

Think of a holomorphic function as an incompressible fluid flow in the plane, and think of an isolated singularity as a point source or sink. Cauchy's theorem tells us that if there is no source inside a loop, the total "circulation" integral around it vanishes. If there is a source, the circulation picks up a fixed contribution depending only on the strength of that source — not on the loop's exact shape, only on which sources it encloses. The residue is precisely this strength. Algebraically, the reason only the a−1a_{-1} term survives is a direct computation: for any integer n≠−1n \neq -1, the function (z−z0)n(z-z_0)^n has an explicit antiderivative (z−z0)n+1n+1\frac{(z-z_0)^{n+1}}{n+1} single-valued on the punctured plane, so its integral around a closed loop is zero. Only n=−1n = -1 produces the multivalued logarithm, whose loop integral is the nonzero quantity 2πi2\pi i. The Residue Theorem is Cauchy's theorem applied term-by-term to the Laurent series, term by term, with exactly one surviving term per singularity.

Figure
A loop enclosing three poles, each contributing its residue
z₁ z₂ z₃ z₄ (outside) ∮_γ f(z) dz = 2πi [Res(f,z₁)+Res(f,z₂)+Res(f,z₃)] γ (counterclockwise)

The central theorem

TheoremResidue Theorem

Let Ω⊆C\Omega \subseteq \mathbb{C} be a bounded region whose boundary ∂Ω\partial \Omega is a finite union of positively oriented, piecewise-C1C^1, pairwise disjoint simple closed curves (so Ω‾=Ω∪∂Ω\overline{\Omega} = \Omega \cup \partial\Omega is compact). Let z1,…,znz_1, \dots, z_n be finitely many distinct points in Ω\Omega, and suppose ff is holomorphic on an open set containing

Ω‾∖{z1,…,zn}.\overline{\Omega} \setminus \{z_1, \dots, z_n\}.

Then

∮∂Ωf(z) dz=2πi∑k=1nRes⁡(f,zk).\oint_{\partial\Omega} f(z)\, dz = 2\pi i \sum_{k=1}^n \operatorname{Res}(f, z_k).
Proof

We use the strategy of excision: remove small disjoint disks around each singularity, apply Cauchy's theorem to the resulting multiply-connected region where ff is holomorphic throughout, and then shrink the disks to recover residues via Laurent expansions.

Step 1 (Setup and choice of radii). Since z1,…,zn∈Ωz_1,\dots,z_n \in \Omega are finitely many distinct points and Ω\Omega is open, choose ρ>0\rho > 0 small enough that: (a) the closed disks D‾(zk,ρ)\overline{D}(z_k, \rho), k=1,…,nk=1,\dots,n, are pairwise disjoint; (b) each D‾(zk,ρ)⊂Ω\overline{D}(z_k,\rho) \subset \Omega. This is possible: let d=min⁡(min⁡j≠k∣zj−zk∣, dist⁡({zk},∂Ω))>0d = \min\big(\min_{j\neq k}|z_j - z_k|,\ \operatorname{dist}(\{z_k\}, \partial\Omega)\big) > 0 (a minimum of finitely many strictly positive numbers, since Ω\Omega is open and the zkz_k are distinct interior points), and take ρ<d/2\rho < d/2.

Step 2 (The excised region). Define

Ωρ:=Ω∖⋃k=1nD‾(zk,ρ).\Omega_\rho := \Omega \setminus \bigcup_{k=1}^n \overline{D}(z_k, \rho).

This is a bounded region whose boundary is

∂Ωρ=∂Ω ∪ ⋃k=1nCk−,\partial \Omega_\rho = \partial \Omega \ \cup\ \bigcup_{k=1}^n C_k^{-},

where Ck−C_k^- denotes the circle ∣z−zk∣=ρ|z - z_k| = \rho traversed clockwise (i.e., with the standard orientation reversed), since Ωρ\Omega_\rho lies outside these disks and the boundary-orientation convention requires the region to lie to the left of the traversal direction. On Ωρ‾\overline{\Omega_\rho}, the function ff is holomorphic (we removed exactly the singular points), and Ωρ‾\overline{\Omega_\rho} is compact.

Step 3 (Apply Cauchy's Theorem for multiply-connected regions). Since ff is holomorphic on an open neighborhood of the compact set Ωρ‾\overline{\Omega_\rho}, and ∂Ωρ\partial\Omega_\rho is a finite union of piecewise-C1C^1 closed curves bounding Ωρ\Omega_\rho with the standard orientation, Cauchy's theorem (in its general form for regions with holes, itself a consequence of Green's theorem applied to the real and imaginary parts, valid since ff has no singularities in Ωρ‾\overline{\Omega_\rho}) gives

∮∂Ωρf(z) dz=0.\oint_{\partial \Omega_\rho} f(z)\, dz = 0.

Explicitly,

∮∂Ωf(z) dz+∑k=1n∮Ck−f(z) dz=0,\oint_{\partial\Omega} f(z)\,dz + \sum_{k=1}^n \oint_{C_k^-} f(z)\, dz = 0,

that is,

∮∂Ωf(z) dz=∑k=1n∮Ckf(z) dz,\oint_{\partial\Omega} f(z)\, dz = \sum_{k=1}^n \oint_{C_k} f(z)\, dz,

where each CkC_k now denotes the circle ∣z−zk∣=ρ|z-z_k| = \rho traversed counterclockwise (we flipped the sign by reversing orientation on each term). This step reduces the problem to computing ∮Ckf dz\oint_{C_k} f\, dz for a small circle around a single isolated singularity — precisely the situation Laurent's theorem addresses.

Step 4 (Laurent expansion and term-by-term integration). Fix kk. Since zkz_k is an isolated singularity of ff, there is rk>0r_k > 0 such that ff is holomorphic on the punctured disk D∗(zk,rk)D^*(z_k, r_k), and (choosing ρ<rk\rho < r_k, which we may do by shrinking ρ\rho further if necessary — finitely many further constraints, so a common ρ\rho working for all kk exists) ff has a Laurent expansion

f(z)=∑m=−∞∞am(k)(z−zk)mf(z) = \sum_{m=-\infty}^{\infty} a_m^{(k)} (z-z_k)^m

converging uniformly on the circle Ck={∣z−zk∣=ρ}C_k = \{|z - z_k| = \rho\} (uniform convergence on compact subsets of the annulus of convergence is part of the Laurent theorem). Because the convergence is uniform on the compact curve CkC_k, we may integrate term by term:

∮Ckf(z) dz=∑m=−∞∞am(k)∮Ck(z−zk)m dz.\oint_{C_k} f(z)\, dz = \sum_{m=-\infty}^{\infty} a_m^{(k)} \oint_{C_k} (z-z_k)^m\, dz.

We now compute ∮Ck(z−zk)m dz\oint_{C_k} (z - z_k)^m \, dz directly. Parametrize CkC_k by z=zk+ρeiθz = z_k + \rho e^{i\theta}, θ∈[0,2π]\theta \in [0, 2\pi], so dz=iρeiθdθdz = i\rho e^{i\theta} d\theta:

∮Ck(z−zk)m dz=∫02πρmeimθ⋅iρeiθ dθ=iρm+1∫02πei(m+1)θ dθ.\oint_{C_k} (z-z_k)^m\, dz = \int_0^{2\pi} \rho^m e^{im\theta} \cdot i \rho e^{i\theta}\, d\theta = i \rho^{m+1} \int_0^{2\pi} e^{i(m+1)\theta}\, d\theta.

If m≠−1m \neq -1, then m+1≠0m + 1 \neq 0 and

∫02πei(m+1)θ dθ=[ei(m+1)θi(m+1)]02π=1−1i(m+1)=0,\int_0^{2\pi} e^{i(m+1)\theta}\, d\theta = \left[\frac{e^{i(m+1)\theta}}{i(m+1)}\right]_0^{2\pi} = \frac{1 - 1}{i(m+1)} = 0,

since ei(m+1)⋅2π=1e^{i(m+1)\cdot 2\pi} = 1. If m=−1m = -1, then ρm+1=ρ0=1\rho^{m+1} = \rho^0 = 1 and

∫02πei⋅0⋅θ dθ=∫02π1 dθ=2π.\int_0^{2\pi} e^{i \cdot 0 \cdot \theta}\, d\theta = \int_0^{2\pi} 1\, d\theta = 2\pi.

Hence

∮Ck(z−zk)m dz={2πim=−10m≠−1.\oint_{C_k} (z-z_k)^m\, dz = \begin{cases} 2\pi i & m = -1 \\ 0 & m \neq -1. \end{cases}

Substituting back into the term-by-term sum, every term vanishes except m=−1m = -1, giving

∮Ckf(z) dz=a−1(k)⋅2πi=2πiRes⁡(f,zk).\oint_{C_k} f(z)\, dz = a_{-1}^{(k)} \cdot 2\pi i = 2\pi i \operatorname{Res}(f, z_k).

Step 5 (Conclusion). Combining Steps 3 and 4:

∮∂Ωf(z) dz=∑k=1n∮Ckf(z) dz=∑k=1n2πiRes⁡(f,zk)=2πi∑k=1nRes⁡(f,zk).\oint_{\partial \Omega} f(z)\, dz = \sum_{k=1}^n \oint_{C_k} f(z)\, dz = \sum_{k=1}^n 2\pi i \operatorname{Res}(f, z_k) = 2\pi i \sum_{k=1}^n \operatorname{Res}(f, z_k).

This is exactly the claimed formula. ■\blacksquare

∎
LemmaResidue at a pole of order mm

Let ff have a pole of order mm at z0z_0, meaning f(z)=(z−z0)−mg(z)f(z) = (z-z_0)^{-m} g(z) on a punctured disk around z0z_0, where gg is holomorphic at z0z_0 and g(z0)≠0g(z_0) \neq 0. Then

Res⁡(f,z0)=g(m−1)(z0)(m−1)!=1(m−1)!lim⁡z→z0dm−1dzm−1[(z−z0)mf(z)].\operatorname{Res}(f, z_0) = \frac{g^{(m-1)}(z_0)}{(m-1)!} = \frac{1}{(m-1)!}\lim_{z \to z_0} \frac{d^{m-1}}{dz^{m-1}}\Big[(z-z_0)^m f(z)\Big].

In particular, for a simple pole (m=1m=1),

Res⁡(f,z0)=lim⁡z→z0(z−z0)f(z).\operatorname{Res}(f, z_0) = \lim_{z\to z_0} (z - z_0) f(z).
Proof

Since gg is holomorphic at z0z_0, it has a Taylor expansion

g(z)=∑n=0∞g(n)(z0)n!(z−z0)ng(z) = \sum_{n=0}^\infty \frac{g^{(n)}(z_0)}{n!}(z-z_0)^n

convergent on some disk D(z0,r)D(z_0, r). Multiplying by (z−z0)−m(z-z_0)^{-m} gives the Laurent expansion of ff on the punctured disk:

f(z)=∑n=0∞g(n)(z0)n!(z−z0)n−m.f(z) = \sum_{n=0}^\infty \frac{g^{(n)}(z_0)}{n!} (z-z_0)^{n-m}.

By uniqueness of Laurent expansions, the coefficient of (z−z0)−1(z-z_0)^{-1} occurs when n−m=−1n - m = -1, i.e., n=m−1n = m-1, giving

Res⁡(f,z0)=g(m−1)(z0)(m−1)!.\operatorname{Res}(f, z_0) = \frac{g^{(m-1)}(z_0)}{(m-1)!}.

Since g(z)=(z−z0)mf(z)g(z) = (z-z_0)^m f(z) on the punctured disk, and this product extends holomorphically to z0z_0 (as gg is holomorphic there), the value g(m−1)(z0)g^{(m-1)}(z_0) is computed as lim⁡z→z0dm−1dzm−1[(z−z0)mf(z)]\lim_{z\to z_0}\frac{d^{m-1}}{dz^{m-1}}[(z-z_0)^m f(z)], using continuity of the (m−1)(m-1)-th derivative of the holomorphic function gg. This proves the general formula; setting m=1m=1 recovers Res⁡(f,z0)=lim⁡z→z0(z−z0)f(z)\operatorname{Res}(f,z_0) = \lim_{z\to z_0}(z-z_0)f(z) (the "0-th derivative" being the function itself). ■\blacksquare

∎

Worked examples

Example 1: A rational function with a double pole and a simple pole, off-center. Compute

∮∣z∣=5z2−2z+3(z−1)2(z+3) dz.\oint_{|z| = 5} \frac{z^2 - 2z + 3}{(z-1)^2(z+3)}\, dz.

Setting up. The integrand has singularities where the denominator vanishes: z=1z = 1 (order 2) and z=−3z = -3 (order 1). Both lie inside ∣z∣=5|z|=5 since ∣1∣=1<5|1| = 1 < 5 and ∣−3∣=3<5|{-3}| = 3 < 5. Write f(z)=z2−2z+3(z−1)2(z+3)f(z) = \dfrac{z^2-2z+3}{(z-1)^2(z+3)}. Residue at z=−3z=-3 (simple pole). By the simple-pole formula,

Res⁡(f,−3)=lim⁡z→−3(z+3)f(z)=(−3)2−2(−3)+3(−3−1)2=9+6+316=1816=98.\operatorname{Res}(f,-3) = \lim_{z\to -3}(z+3)f(z) = \frac{(-3)^2 - 2(-3) + 3}{(-3-1)^2} = \frac{9+6+3}{16} = \frac{18}{16} = \frac{9}{8}.

Residue at z=1z=1 (double pole). Here g(z)=(z−1)2f(z)=z2−2z+3z+3g(z) = (z-1)^2 f(z) = \dfrac{z^2-2z+3}{z+3}, and by the order-2 formula,

Res⁡(f,1)=g′(1).\operatorname{Res}(f,1) = g'(1).

Compute g′(z)g'(z) by the quotient rule:

g′(z)=(2z−2)(z+3)−(z2−2z+3)(1)(z+3)2.g'(z) = \frac{(2z-2)(z+3) - (z^2-2z+3)(1)}{(z+3)^2}.

At z=1z=1: numerator =(2⋅1−2)(1+3)−(1−2+3)=(0)(4)−2=−2= (2\cdot 1 - 2)(1+3) - (1 - 2 + 3) = (0)(4) - 2 = -2, and denominator =(1+3)2=16= (1+3)^2 = 16. So

Res⁡(f,1)=−216=−18.\operatorname{Res}(f,1) = \frac{-2}{16} = -\frac{1}{8}.

Assemble via the theorem.

∮∣z∣=5f(z) dz=2πi(Res⁡(f,−3)+Res⁡(f,1))=2πi(98−18)=2πi⋅1=2πi.\oint_{|z|=5} f(z)\, dz = 2\pi i\left(\operatorname{Res}(f,-3) + \operatorname{Res}(f,1)\right) = 2\pi i\left(\frac{9}{8} - \frac{1}{8}\right) = 2\pi i \cdot 1 = 2\pi i.

This is a clean check: the answer is a simple multiple of 2πi2\pi i despite the asymmetric numerator and mixed pole orders, exactly as the theorem guarantees since only the residues (not the full function) matter.

Example 2: A pole of order 3 combined with an essential-singularity check, and a semicircular contour foreshadowing real-integral applications. Let f(z)=eiz(z−2i)3f(z) = \dfrac{e^{iz}}{(z-2i)^3}, and compute

∮γRf(z) dz\oint_{\gamma_R} f(z)\, dz

where γR\gamma_R is the boundary of the upper half-disk {∣z∣≤R, Im⁡z≥0}\{|z| \le R,\ \operatorname{Im} z \ge 0\} for R=5R = 5 (a diameter along the real axis from −5-5 to 55, plus the semicircular arc), traversed counterclockwise. Locating singularities. The only singularity of ff is at z=2iz = 2i, a pole of order 33 (since eize^{iz} is entire and nonvanishing, it contributes no zero/pole behavior). We check 2i2i lies inside γ5\gamma_5: ∣2i∣=2<5|2i| = 2 < 5 and Im⁡(2i)=2>0\operatorname{Im}(2i) = 2 > 0, so yes, it's enclosed. Computing the residue at a pole of order 3. Here m=3m=3 and g(z)=eizg(z) = e^{iz} (already the holomorphic factor after stripping (z−2i)−3(z-2i)^{-3}), so

Res⁡(f,2i)=g′′(2i)2!.\operatorname{Res}(f, 2i) = \frac{g''(2i)}{2!}.

We compute derivatives of g(z)=eizg(z) = e^{iz}:

g′(z)=ieiz,g′′(z)=i2eiz=−eiz.g'(z) = i e^{iz}, \qquad g''(z) = i^2 e^{iz} = -e^{iz}.

At z=2iz = 2i:

g′′(2i)=−ei(2i)=−e−2.g''(2i) = -e^{i(2i)} = -e^{-2}.

(Here we used i⋅2i=2i2=−2i \cdot 2i = 2i^2 = -2, so ei(2i)=e−2e^{i(2i)} = e^{-2}, a real number since 2i2i is purely imaginary.) Hence

Res⁡(f,2i)=−e−22=−12e2.\operatorname{Res}(f, 2i) = \frac{-e^{-2}}{2} = -\frac{1}{2e^2}.

Applying the theorem.

∮γ5eiz(z−2i)3 dz=2πi⋅(−12e2)=−πie2.\oint_{\gamma_5} \frac{e^{iz}}{(z-2i)^3}\, dz = 2\pi i \cdot \left(-\frac{1}{2e^2}\right) = -\frac{\pi i}{e^2}.

This example is the prototype computation used later (see Evaluation of Real Integrals) to evaluate integrals like ∫−∞∞cos⁡x(x2+4)n dx\int_{-\infty}^\infty \frac{\cos x}{(x^2+4)^n}\,dx by letting R→∞R\to\infty and showing the arc contribution vanishes (Jordan's lemma) — here we only needed the finite-RR residue computation itself, which is exact for every R>2R > 2, not merely in a limit.

CounterexampleDropping "isolated" singularity / finiteness — a non-isolated singularity breaks the theorem

The Residue Theorem requires the singularities enclosed by γ\gamma to be isolated and finite in number, so that a Laurent expansion (hence a well-defined residue) exists at each and Step 1 of the proof (choosing disjoint disks) is possible. Drop this and the conclusion fails outright — there is no residue to sum, and no version of the formula survives. Consider

f(z)=1sin⁡(1/z)f(z) = \frac{1}{\sin(1/z)}

on the punctured disk D∗(0,1)D^*(0,1). The point z=0z=0 is a singularity of ff, but it is not isolated: ff has poles at every point z=1nπz = \frac{1}{n\pi} for n=±1,±2,…n = \pm 1, \pm 2, \dots (since sin⁡(1/z)=0\sin(1/z) = 0 exactly when 1/z=nπ1/z = n\pi), and these poles accumulate at z=0z=0 as n→∞n \to \infty. Every punctured disk around 00, however small, contains infinitely many of these poles. Consequently, ff has no Laurent expansion valid on any punctured disk 0<∣z∣<r0 < |z| < r around 00 — Laurent's theorem itself requires holomorphy on an annulus, and no such annulus around 00 exists free of poles. There is no number "Res⁡(f,0)\operatorname{Res}(f,0)" to speak of, and the proof's Step 1 (choosing ρ\rho small enough to isolate the singularity in a disk disjoint from the rest of the domain) is impossible: any disk around 00 still contains infinitely many other poles. If one naively tries to integrate ff around a small circle ∣z∣=ϵ|z| = \epsilon, the value depends delicately on exactly which poles the circle happens to avoid or graze as ϵ\epsilon varies, and no clean closed formula "2πi×(one number)2\pi i \times (\text{one number})" exists. This shows why the hypothesis "isolated singularities, finitely many enclosed" cannot be weakened to merely "singularities in Ω\Omega": the entire mechanism (Laurent expansion at a single accumulation-free point) collapses.

Scope, edge cases & common traps

SituationWhat to checkCommon error
Pole order mmUse 1(m−1)!lim⁡(z−z0)mf(z)\frac{1}{(m-1)!}\lim (z-z_0)^m f(z) differentiated (m−1)(m-1) timesUsing the simple-pole shortcut for a higher-order pole
Singularity on the contourTheorem does not apply; contour must avoid all singularitiesSilently including a boundary pole in the sum
Removable singularityResidue is 00 (Laurent series has no negative terms)Assuming every "singularity" contributes a nonzero residue
Essential singularityResidue still well-defined as a−1a_{-1}, but no polynomial-derivative shortcut exists — must extract the Laurent coefficient directlyTrying to force the pole-order formula on an essential singularity
OrientationClockwise traversal reverses the sign: ∮=−2πi∑Res⁡\oint = -2\pi i \sum \operatorname{Res}Forgetting to flip sign for clockwise contours or inner boundary components
Singularity outside Ω\OmegaExcluded from the sum entirelyIncluding all "nearby-looking" poles regardless of whether they're enclosed

The mathematician behind it

Origins · 1 minAugustin-Louis Cauchy and how this was discovered
AC
Augustin-Louis Cauchy
1789–1857 · French mathematician, École Polytechnique and Sorbonne

Cauchy essentially founded rigorous complex analysis, proving the integral theorem that bears his name and laying the groundwork from which the calculus of residues emerged. Though the term "residue" and its systematic calculus were developed and named by Cauchy himself in the 1820s–1840s, the theorem synthesizes his integral theorem, his integral formula, and the Laurent expansion (due to Pierre Alphonse Laurent, building on Cauchy's methods) into the unified tool used today. Cauchy's prolific output — over 800 papers — reshaped analysis with the standards of rigor still used in defining limits, continuity, and convergence.

History
- 1814/1825 — Cauchy proves foundational integral theorems for contour integrals of holomorphic functions.
- 1831 — Cauchy develops the integral formula expressing function values via boundary integrals.
- 1843 — Pierre Alphonse Laurent proves the Laurent series expansion theorem for annuli, submitted to the Paris Academy.
- 1846 — Cauchy formalizes the calculus of residues, naming and systematizing the residue as a computational tool.
- Late 1800s — Residue methods become standard for evaluating real integrals, cementing their place in analysis curricula worldwide.
Why does only the a−1a_{-1} Laurent coefficient matter, and not a−2,a−3,…a_{-2}, a_{-3}, \dots?

Every power (z−z0)n(z-z_0)^n with n≠−1n \neq -1 has an explicit single-valued antiderivative (z−z0)n+1n+1\frac{(z-z_0)^{n+1}}{n+1} on the punctured plane, so its integral around any closed loop is zero by the fundamental theorem of calculus for contour integrals. Only n=−1n=-1 fails to have such an antiderivative on the punctured plane (its "antiderivative" would be a logarithm, which is multivalued), and its direct parametrized integral evaluates to 2πi2\pi i, not zero. So when we integrate a Laurent series term-by-term around a small loop, every term but the (z−z0)−1(z-z_0)^{-1} term vanishes, leaving only a−1⋅2πia_{-1}\cdot 2\pi i.

Does the Residue Theorem apply if the singularity is an essential singularity, not just a pole?

Yes — the theorem itself makes no distinction between poles and essential singularities; it only requires the singularity to be isolated so that a Laurent expansion exists on some punctured disk. The residue is still defined as the a−1a_{-1} coefficient. What differs is how you compute it: for poles there is a closed-form derivative formula (via the lemma in this article), but for an essential singularity (e.g. e1/ze^{1/z} at z=0z=0) you generally must extract the Laurent coefficient directly from the known series expansion, since no finite-order pole formula applies.

What happens if a singularity lies exactly on the contour γ\gamma, not strictly inside?

The Residue Theorem does not apply in that case at all — its hypotheses require ff to be holomorphic on a neighborhood of Ω‾\overline{\Omega} except at finitely many interior points. If a singularity sits on ∂Ω\partial\Omega, the integral itself may not even converge (for example, the integral of 1/z1/z along a path passing through z=0z=0 diverges). In practice, one perturbs the contour slightly to route around such boundary singularities, often via a small semicircular indentation, and separately analyzes the contribution of that indentation using local behavior near the pole (a technique used heavily in real-integral applications).

How is the Residue Theorem related to Cauchy's Integral Formula?

Cauchy's Integral Formula is a special case in disguise. Given ff holomorphic on Ω\Omega and z0∈Ωz_0 \in \Omega, apply the Residue Theorem to h(z)=f(z)z−z0h(z) = \frac{f(z)}{z-z_0}, which has a single simple pole at z0z_0 with residue Res⁡(h,z0)=f(z0)\operatorname{Res}(h,z_0) = f(z_0) (by the simple-pole limit formula, since ff is holomorphic and continuous at z0z_0). The Residue Theorem then gives ∮∂Ωf(z)z−z0dz=2πif(z0)\oint_{\partial\Omega} \frac{f(z)}{z-z_0}dz = 2\pi i f(z_0), exactly Cauchy's formula. So Cauchy's Integral Formula is the Residue Theorem applied to one very specific meromorphic function.

Can the Residue Theorem be used when Ω\Omega has holes, i.e., is not simply connected?

Yes, and this is built directly into the theorem's statement: ∂Ω\partial\Omega is allowed to be a union of several disjoint simple closed curves, with inner boundary components (around holes) oriented clockwise so that Ω\Omega still lies to the left of the traversal. The proof's excision argument works identically — the "holes" are simply additional boundary pieces alongside the small circles cut out around each singularity, and Cauchy's theorem for multiply-connected regions handles them without modification.

Why must the radius ρ\rho in the proof be chosen small enough for disjoint disks, and is this always possible?

The disks must be disjoint so that each contains exactly one singularity, isolating it enough to apply the Laurent expansion theorem uninterrupted by nearby singularities. Since there are only finitely many singularities z1,…,znz_1,\dots,z_n, all pairwise distances ∣zj−zk∣|z_j - z_k| are strictly positive, and their minimum (over finitely many pairs) is a fixed positive number dd. Choosing ρ<d/2\rho < d/2 (and also less than the distance from each zkz_k to ∂Ω\partial\Omega) guarantees disjointness and containment in Ω\Omega simultaneously — this is a completely constructive, finite argument, not an existence claim requiring compactness tricks.

How does the Residue Theorem foreshadow the Argument Principle?

The Argument Principle applies the Residue Theorem to the specific function f′(z)/f(z)f'(z)/f(z) for a meromorphic ff. At a zero of order kk of ff, this logarithmic derivative has a simple pole with residue kk; at a pole of order kk of ff, it has a simple pole with residue −k-k. Summing residues via the Residue Theorem then counts zeros minus poles (with multiplicity) enclosed by a contour — turning residue calculus into a zero/pole-counting tool, which is exactly the subject of The Argument Principle and Rouche's Theorem.

Before this

Where this leads