Residues and the Residue Theorem

A function that is analytic everywhere in a small disc except at one central point still carries a complete record of what goes wrong there, written in its Laurent series. Of the infinitely many coefficients in that series, exactly one survives integration around a loop: the coefficient of $(z-z_0)^{-1}$ , called the residue. Every other power, positive or negative, integrates to zero around a closed contour, because $\oint (z-z_0)^n\,\mathrm{d}z$ vanishes unless $n=-1$ , where it equals $2\pi i$ . The residue is therefore not an arbitrary label but the single number a contour integral is physically able to detect — the one fingerprint of a singularity that flux around it can measure. The Residue Theorem promotes this local fact to a global accounting rule. If $f$ is analytic on and inside a simple closed contour $C$ except at finitely many isolated singularities, then $\oint_C f\,\mathrm{d}z$ equals $2\pi i$ times the sum of the enclosed residues, each weighted by how many times $C$ winds around it. A global integral collapses into a finite, local, algebraic sum. This is not a new axiom: it is the exact meeting point of Cauchy's Integral Theorem (deform freely where $f$ is analytic), Cauchy's Integral Formula (the kernel $1/(z-z_0)$ supplies one residue), and the Laurent series (which supplies all the others). For a physicist the payoff is enormous. Real integrals that defeat elementary calculus — Fourier and Laplace inversions, dispersion relations, momentum-space Green's functions, infinite sums summed by pairing with $\pi\cot\pi z$ — all reduce to residue arithmetic the moment one closes a contour in the complex plane. Learning to read off a residue is, in practice, learning to do a large fraction of theoretical physics.

Key takeaways

The residue of $f$ at an isolated singularity $z_0$ is the Laurent coefficient $a_{-1}$ of $(z-z_0)^{-1}$ — equivalently $\tfrac{1}{2\pi i}\oint f\,\mathrm{d}z$ around a small loop about $z_0$ .
The Residue Theorem: $\oint_C f\,\mathrm{d}z = 2\pi i\sum_k n(C,z_k)\operatorname{Res}_{z_k}f$ , converting a global contour integral into a finite sum of local residues.
For a simple pole $\operatorname{Res}=\lim_{z\to z_0}(z-z_0)f(z)$ ; for an order- $m$ pole differentiate $m-1$ times; for a quotient $g/h$ with a simple zero of $h$ , just $g(z_0)/h'(z_0)$ .
It unifies Cauchy's Integral Theorem (no residues) and Cauchy's Integral Formula (one residue from the kernel) as the two simplest special cases.
It is the master tool for real definite integrals ( $\int_{-\infty}^\infty \mathrm{d}x/(1+x^2)=\pi$ ), infinite series ( $\sum 1/n^2=\pi^2/6$ ), transform inversion, and Green's functions.
Key idea: around a closed loop every Laurent power $(z-z_0)^n$ integrates to zero except $n=-1$ , which gives $2\pi i$ — so a contour integral sees only the residues, the $a_{-1}$ coefficients, of whatever singularities it encloses.

The core insight

Around a closed loop every Laurent power $(z-z_0)^n$ integrates to zero except $n=-1$ , which gives $2\pi i$ ; hence a contour integral sees only the residues — the $a_{-1}$ coefficients — of whatever singularities it encloses.

Where this lives today

Residue calculus is not a historical curiosity — it is the everyday arithmetic of theoretical physics. In quantum field theory every Feynman propagator is a pole in the complex energy plane, and the loop integrals of perturbation theory are evaluated by contour rotation and residues; the choice of which pole to enclose (the $i\varepsilon$ prescription) is what encodes causality and distinguishes the Feynman, retarded, and advanced Green's functions. The one-line Yukawa integral derived above generalizes directly to the massive scalar propagator. In condensed matter, Matsubara sums over discrete imaginary frequencies are performed by inserting the Bose or Fermi distribution — which supplies exactly the unit-residue poles of a $\cot$ or $\tanh$ kernel — and closing a contour, the finite-temperature analogue of the Basel trick. Signal processing inverts Laplace transforms through the Bromwich integral, reading each system mode off a pole's residue, and the Kramers-Kronig relations of optical response come from pushing a contour past the poles of a causal susceptibility. Numerical analysts even use residue-based contour integrals (rational eigensolvers like FEAST) to compute eigenvalues to high accuracy — a two-hundred-year-old idea running inside modern linear-algebra libraries.

The physical picture

Think of each isolated singularity as a point source dropped into the complex plane and of the contour integral $\oint_C f\,\mathrm{d}z$ as a flux measurement taken along the loop $C$ . Cauchy's Theorem says the loop can be stretched, dented, or slid at will without changing the reading, as long as it never crosses a source. So $C$ can be reeled in until it becomes a bunch of tiny circles, one hugging each enclosed singularity. On each tiny circle the Laurent series is a clean inventory of the local behaviour, and integrating it term by term annihilates everything except the $(z-z_0)^{-1}$ term. What is left is a fixed quantum $2\pi i\,a_{-1}$ per source. The total flux is just the bookkeeping of these quanta — nothing about the far-away shape of $C$ matters, only which sources sit inside it and with what residue.

Figure

The loop C picks up a fixed quantum 2πi·Res at each enclosed singularity: deforming C onto small circles γₖ hugging the poles leaves ∮꜀ f dz = 2πi Σₖ Res f — a global integral reduced to a local algebraic sum

The theorem is only useful because the residue is easy to get without ever doing the integral: for a simple pole you cancel one factor and take a limit; for a higher-order pole you differentiate. The second picture shows the workhorse application — turning a real integral over the whole line into a single residue by closing the contour with a large arc that contributes nothing. The pole at $z=+i$ is captured, its twin at $z=-i$ is left outside, and the arithmetic $2\pi i\cdot \tfrac{1}{2i}=\pi$ delivers a definite integral that would otherwise need an antiderivative.

Figure

Closing the real axis with a semicircle of radius R in the upper half plane captures only the pole z = i: the arc contribution falls as πR/(R²−1) → 0, and Res at i of 1/(1+z²) = 1/(2i) gives ∫dx/(1+x²) = 2πi·1/(2i) = π

The core equations

The residue is defined two equivalent ways at once: as a specific Laurent coefficient, and as a normalized contour integral. In practice one reads the coefficient off algebraically and uses it to evaluate the integral — the definition runs left to right, the application runs right to left.

\operatorname{Res}_{z=z_0} f \equiv a_{-1} = \frac{1}{2\pi i}\oint_{C} f(z)\,\mathrm{d}z \qquad (1)

where: $a_{-1}$ — Laurent coefficient of $(z-z_0)^{-1}$ , the residue; $C$ — small positively oriented circle enclosing only $z_0$ ; $f$ — analytic in a punctured neighbourhood of $z_0$ . Reading: the residue is both a number in a series and the value of a loop integral, normalized by $2\pi i$ . Every isolated singularity carries a full Laurent expansion, and the residue is one entry in the table of coefficients. The integral formula for $a_n$ makes the definition above the special case $n=-1$ .

f(z) = \sum_{n=-\infty}^{\infty} a_n (z-z_0)^n,\qquad a_n = \frac{1}{2\pi i}\oint_C \frac{f(z)}{(z-z_0)^{n+1}}\,\mathrm{d}z \qquad (2)

where: $a_n$ — the $n$ -th Laurent coefficient; $n$ — any integer; $z_0$ — centre of expansion and location of the singularity. Reading: setting $n=-1$ makes the kernel $1/(z-z_0)^0=1$ , recovering equation (1). The engine that makes only $a_{-1}$ matter is a single orthogonality integral. It is the reason the whole theory works, and it depends on nothing but the geometry of a circle.

\oint_{C}(z-z_0)^{n}\,\mathrm{d}z = 2\pi i\,\delta_{n,-1} \qquad (3)

where: $\delta_{n,-1}$ — Kronecker delta, equal to $1$ when $n=-1$ and $0$ otherwise; $C$ — a closed contour winding once around $z_0$ . Reading: integrating any integer power over a full loop kills it, except the $-1$ power, which returns the universal constant $2\pi i$ . Assembled globally, these give the Residue Theorem in the form that keeps track of multiply-wound contours through the winding number.

\oint_{C} f(z)\,\mathrm{d}z = 2\pi i \sum_{k} n(C,z_k)\,\operatorname{Res}_{z=z_k} f \qquad (4)

where: $z_k$ — the isolated singularities enclosed by $C$ ; $n(C,z_k)$ — winding number of $C$ about $z_k$ , usually $+1$ ; $\operatorname{Res}_{z=z_k}f$ — the residue there. Reading: the integral counts each singularity with a multiplicity equal to how many times the loop circles it. To use the theorem you need residues, and for a pole of finite order there is a closed formula that never requires the full series. Multiply away the pole, then differentiate to pick out $a_{-1}$ .

\operatorname{Res}_{z=z_0} f = \frac{1}{(m-1)!}\lim_{z\to z_0}\frac{\mathrm{d}^{m-1}}{\mathrm{d}z^{m-1}}\Big[(z-z_0)^m f(z)\Big] \qquad (5)

where: $m$ — the order of the pole at $z_0$ ; $(z-z_0)^m f(z)$ — the regularized function, analytic at $z_0$ . Reading: for a simple pole ( $m=1$ ) this is just $\lim_{z\to z_0}(z-z_0)f(z)$ ; each extra order of pole costs one derivative. The single most common case — a ratio whose denominator has a simple zero — has an even shorter shortcut, obtained by L'Hôpital on the simple-pole limit.

\operatorname{Res}_{z=z_0}\frac{g(z)}{h(z)} = \frac{g(z_0)}{h'(z_0)} \qquad (6)

where: $g(z_0)\neq 0$ — numerator value; $h(z_0)=0,\ h'(z_0)\neq 0$ — denominator with a simple zero. Reading: differentiate only the denominator and evaluate — no factoring, no limits. On the Riemann sphere the accounting closes exactly: the residue at infinity balances all the finite ones, giving a powerful consistency check and an alternative route when the finite poles are messy.

\operatorname{Res}_{z=\infty} f = -\frac{1}{2\pi i}\oint_{C} f\,\mathrm{d}z = -\operatorname{Res}_{w=0}\!\Big[\tfrac{1}{w^2}f\!\big(\tfrac1w\big)\Big],\qquad \sum_{\text{all incl. }\infty}\operatorname{Res} f = 0 \qquad (7)

where: $C$ — large circle enclosing all finite singularities; $w=1/z$ — the map sending infinity to the origin. Reading: for any rational or globally meromorphic $f$ , the finite residues plus the residue at infinity sum to zero.

Derivation AProof of the Residue Theorem

The goal is to show that the contour integral of a function analytic except at finitely many isolated singularities equals $2\pi i$ times the sum of its residues there, using only two ingredients already in hand: Cauchy's Theorem, which lets a contour be deformed freely through any region where $f$ is analytic, and the orthogonality integral $\oint(z-z_0)^n\,\mathrm{d}z=2\pi i\,\delta_{n,-1}$ , which does the actual selecting. We isolate each singularity on its own small circle, expand there, and integrate term by term.

Step 1. Let $f$ be analytic on and inside a simple closed contour $C$ except at the isolated points $z_1,\dots,z_N$ , each encircled once positively. Draw a tiny circle $\gamma_k$ around each $z_k$ and join every $\gamma_k$ to $C$ by a pair of infinitesimally close cross-cuts. The resulting region — $C$ on the outside, the $\gamma_k$ punched out on the inside — contains no singularities, so Cauchy's Theorem gives $\oint_{C} f\,\mathrm{d}z = \sum_{k=1}^{N}\oint_{\gamma_k} f\,\mathrm{d}z$ . Why: the two passes along each cross-cut run in opposite directions and cancel exactly, so the boundary integral reduces to $C$ minus the small circles.

Step 2. On the annulus surrounding $z_k$ the Laurent series $f(z)=\sum_{n}a_n^{(k)}(z-z_k)^n$ converges uniformly, so it may be integrated term by term over $\gamma_k$ : $\oint_{\gamma_k} f\,\mathrm{d}z=\sum_n a_n^{(k)}\oint_{\gamma_k}(z-z_k)^n\,\mathrm{d}z$ . Why: uniform convergence on the compact circle is exactly the license needed to exchange the infinite sum and the integral without changing the value.

Step 3. Apply the orthogonality integral (equation 3): every term $\oint_{\gamma_k}(z-z_k)^n\,\mathrm{d}z$ vanishes except $n=-1$ , which contributes $2\pi i$ . Hence $\oint_{\gamma_k} f\,\mathrm{d}z = 2\pi i\,a_{-1}^{(k)} = 2\pi i\,\operatorname{Res}_{z=z_k} f$ . Why: this single fact is the whole selection mechanism — of the entire Laurent inventory only the residue is visible to a loop integral.

Step 4. Sum over the singularities: $\oint_{C} f\,\mathrm{d}z = 2\pi i\sum_{k=1}^{N}\operatorname{Res}_{z=z_k} f$ , which is the theorem. Why: with each local circle reduced to its residue, the global integral is just their total, and inserting winding numbers for contours that loop more than once gives the general form (4). *Two classical results emerge as immediate special cases: with $N=0$ (no singularities) the right side is empty and $\oint_C f\,\mathrm{d}z=0$ — Cauchy's Integral Theorem. With $f(z)=g(z)/(z-z_0)$ , $g$ analytic, the sole residue is $g(z_0)$ , so $\oint_C f\,\mathrm{d}z=2\pi i\,g(z_0)$ — Cauchy's Integral Formula.*

Go deeper · Derivation B · 1 minThe order-m residue formula from the Laurent series

Derivation BThe order-m residue formula from the Laurent series

The aim is to derive the practical formula (5) that extracts a residue by differentiation, so one never has to construct the whole Laurent series to read off a single coefficient. The idea is that multiplying $f$ by $(z-z_0)^m$ turns its principal part into an ordinary Taylor series, after which the residue is buried as one particular Taylor coefficient that repeated differentiation exposes. We track exactly which derivative lands on $a_{-1}$ .

Step 1. Suppose $z_0$ is a pole of order $m$ , so the Laurent series starts at $n=-m$ : $f(z)=\sum_{n=-m}^{\infty}a_n(z-z_0)^n$ . Multiply through by $(z-z_0)^m$ to get $\phi(z)\equiv(z-z_0)^m f(z)=\sum_{n=-m}^{\infty}a_n(z-z_0)^{n+m}$ . Why: the smallest power is now $(z-z_0)^0$ , so $\phi$ is analytic at $z_0$ — the pole has been removed and what remains is an honest Taylor series.

Step 2. In that Taylor series the residue $a_{-1}$ sits on the power $(z-z_0)^{n+m}$ with $n=-1$ , i.e. the power $(z-z_0)^{m-1}$ . The Taylor coefficient of $(z-z_0)^{m-1}$ is $\phi^{(m-1)}(z_0)/(m-1)!$ . Why: differentiating a power series $m-1$ times and setting $z=z_0$ kills every lower power (they vanish) and every higher power (they still carry a factor $z-z_0$ ), leaving only the $(m-1)$ -th term.

Step 3. Equate the two expressions for that coefficient: $a_{-1}=\dfrac{1}{(m-1)!}\phi^{(m-1)}(z_0)=\dfrac{1}{(m-1)!}\lim_{z\to z_0}\dfrac{\mathrm{d}^{m-1}}{\mathrm{d}z^{m-1}}\big[(z-z_0)^m f(z)\big]$ , which is equation (5). Why: the limit is written instead of a plain evaluation so the formula also applies when $\phi$ is only defined by a removable singularity and must be reached as a limit.

Step 4. Specialize. For $m=1$ the derivative is the zeroth one, so $\operatorname{Res}=\lim_{z\to z_0}(z-z_0)f(z)$ . For a quotient $g/h$ with $h(z_0)=0$ , $h'(z_0)\neq0$ , write $\lim(z-z_0)\,g/h=\lim g\big/\big[(h(z)-h(z_0))/(z-z_0)\big]=g(z_0)/h'(z_0)$ , recovering (6). Why: the difference quotient of $h$ tends to $h'(z_0)$ , so the simple-pole limit becomes the numerator over the derivative of the denominator. *As a concrete check: for $f=e^z/(z-1)^2$ at $z=1$ ( $m=2$ ), formula (5) gives $\tfrac{1}{1!}\tfrac{\mathrm{d}}{\mathrm{d}z}e^z\big|_{1}=e$ . Applying the wrong simple-pole limit $\lim_{z\to1}(z-1)f=\lim e^z/(z-1)$ diverges, correctly flagging that $m=2$ .*

Go deeper · Derivation C · 1 minThe Yukawa Green's function by Jordan's lemma

Derivation CThe Yukawa Green's function by Jordan's lemma

The purpose here is to evaluate the one-dimensional Fourier integral $\int_{-\infty}^{\infty} e^{ikx}/(k^2+m^2)\,\mathrm{d}k$ , which is the momentum-space form of the screened (Yukawa or Klein-Gordon) Green's function and a template for essentially every propagator integral in physics. The technique is to close the real axis with a large semicircle in whichever half-plane makes the exponential decay, use Jordan's lemma to discard the arc, and collect the single enclosed residue.

Step 1. The integrand $e^{ikx}/(k^2+m^2)=e^{ikx}/[(k-im)(k+im)]$ has simple poles at $k=\pm im$ on the imaginary axis. Take $x>0$ first and close the contour with a semicircle of radius $R$ in the upper half-plane, where $\operatorname{Im}k>0$ makes $|e^{ikx}|=e^{-x\,\operatorname{Im}k}$ decay. This encloses only $k=+im$ . Why: the sign of $x$ dictates the half-plane — you must close where the oscillating exponential turns into a decaying one, or the arc will not vanish.

Step 2. Jordan's lemma bounds the arc: since $1/(k^2+m^2)\to0$ uniformly as $R\to\infty$ and the factor $e^{ikx}$ decays in the upper half-plane for $x>0$ , the semicircular contribution vanishes in the limit. The closed-contour integral therefore equals the original real-line integral. Why: Jordan's lemma is the sharpened arc estimate that works even though $1/(k^2+m^2)$ only decays like $1/R^2$ — the exponential does the heavy lifting.

Step 3. Compute the residue at $k=im$ using the quotient rule (6) with $h=k^2+m^2$ , $h'=2k$ : $\operatorname{Res}_{k=im}=\dfrac{e^{ikx}}{2k}\Big|_{k=im}=\dfrac{e^{-mx}}{2im}$ . By the Residue Theorem the integral is $2\pi i$ times this: $\int_{-\infty}^{\infty}\dfrac{e^{ikx}}{k^2+m^2}\,\mathrm{d}k=2\pi i\cdot\dfrac{e^{-mx}}{2im}=\dfrac{\pi}{m}e^{-mx}$ . Why: $e^{i(im)x}=e^{-mx}$ is real and decaying, so the imaginary factors cancel and the answer is a clean real exponential.

Step 4. For $x<0$ close in the lower half-plane instead, picking up $k=-im$ with a sign flip from the clockwise orientation; the result is $\dfrac{\pi}{m}e^{+mx}$ . The two cases combine into $\dfrac{\pi}{m}e^{-m|x|}$ . Why: the absolute value is forced by symmetry — the Green's function must decay away from the source in both directions, which is exactly the screened Coulomb falloff of the Yukawa potential. *As a numerical check, at $m=1.5,\ x=0.7$ the integral evaluates to $0.7329079$ , matching $(\pi/m)e^{-m|x|}=0.7329079$ to seven digits; the result is real and even in $x$ as required.*

Limiting cases

The general apparatus specializes to a small family of everyday tools, selected entirely by the type of singularity enclosed.

Case	Result	Meaning
No singularity enclosed	$\oint_C f(z)\,\mathrm{d}z = 0$	All residues absent — this is Cauchy's Integral Theorem, the $N=0$ base case.
Simple pole (order 1)	$\operatorname{Res}_{z_0} f = \lim_{z\to z_0}(z-z_0)\,f(z)$	The overwhelmingly common case; cancels the single $(z-z_0)^{-1}$ factor.
Pole of order $m$	$\operatorname{Res}_{z_0} f = \dfrac{1}{(m-1)!}\lim_{z\to z_0}\dfrac{\mathrm{d}^{m-1}}{\mathrm{d}z^{m-1}}\big[(z-z_0)^m f\big]$	Differentiate $m-1$ times to extract $a_{-1}$ from the regularized function.
Quotient $g/h$ , simple pole	$\operatorname{Res}_{z_0}\dfrac{g}{h} = \dfrac{g(z_0)}{h'(z_0)}$	Valid when $g(z_0)\neq0$ , $h(z_0)=0$ , $h'(z_0)\neq0$ ; skips all algebra.
Kernel $g/(z-z_0)$ , $g$ analytic	$\oint_C f\,\mathrm{d}z = 2\pi i\,g(z_0)$	Cauchy's Integral Formula — the single-residue special case.

Common trapThe simple-pole limit is not a universal residue formula

The residue is by definition the coefficient of $(z-z_0)^{-1}$ in the Laurent series — it is not the value of $f$ , nor of any single derivative of $f$ , taken directly. Two traps follow. First, the shortcut $\operatorname{Res}=\lim_{z\to z_0}(z-z_0)f(z)$ works only for a simple pole; applied to a double pole it cancels just one of the two offending factors and diverges, so you must confirm the pole order before using it and switch to the order- $m$ formula (differentiate $m-1$ times) otherwise. Second, branch points and essential singularities are not poles at all: $f(z)=e^{1/z}$ has an essential singularity at $0$ with residue $1$ read straight from its Laurent series ( $\sum z^{-n}/n!$ ), a value no pole formula will ever give, and functions with branch cuts such as $\sqrt{z}$ or $\log z$ demand keyhole contours that respect the cut rather than blind residue counting.

Worked examples

1 · A real integral by closing a contour (fully worked)

Evaluate $\displaystyle\int_{-\infty}^{\infty}\frac{\mathrm{d}x}{1+x^2}$ using residues, and check it against the elementary antiderivative.

Step 1. Extend to the complex plane: $f(z)=1/(1+z^2)=1/[(z-i)(z+i)]$ has simple poles at $z=\pm i$ . Close the real segment $[-R,R]$ with a semicircle of radius $R$ in the upper half-plane, forming a contour $C$ that encloses only $z=+i$ .

Step 2. Bound the arc. On the semicircle $|1+z^2|\ge R^2-1$ , so $\big|\int_{\text{arc}}f\,\mathrm{d}z\big|\le \dfrac{\pi R}{R^2-1}\xrightarrow{R\to\infty}0$ . The arc drops out and the closed-contour integral equals the real-line integral.

Step 3. Take the residue with the quotient rule (6), $h=1+z^2,\ h'=2z$ : $\operatorname{Res}_{z=i}f=\dfrac{1}{2z}\Big|_{z=i}=\dfrac{1}{2i}$ .

Step 4. Apply the theorem: $\displaystyle\int_{-\infty}^{\infty}\frac{\mathrm{d}x}{1+x^2}=2\pi i\cdot\frac{1}{2i}=\boldsymbol{\pi}$ . Cross-check with $\arctan x\big|_{-\infty}^{\infty}=\tfrac{\pi}{2}-(-\tfrac{\pi}{2})=\pi$ — the two agree exactly, and the residue method needed no antiderivative.

2 · An infinite sum from a cotangent kernel (fully worked)

Sum the Basel series $\displaystyle\sum_{n=1}^{\infty}\frac{1}{n^2}$ by pairing $1/z^2$ with $\pi\cot\pi z$ .

Step 1. The function $\pi\cot\pi z$ has a simple pole with residue $1$ at every integer, because near $z=n$ it behaves like $1/(z-n)$ . So $g(z)=\pi\cot(\pi z)/z^2$ has, at each nonzero integer $n$ , a simple pole with $\operatorname{Res}_{z=n}g=1/n^2$ .

Step 2. At $z=0$ , $g$ has a triple pole. Using the standard expansion $\pi\cot\pi z=\dfrac{1}{z}-\dfrac{\pi^2 z}{3}-\dfrac{\pi^4 z^3}{45}-\cdots$ , divide by $z^2$ : $g(z)=\dfrac{1}{z^3}-\dfrac{\pi^2}{3z}-\cdots$ . The coefficient of $1/z$ is $-\pi^2/3$ , so $\operatorname{Res}_{z=0}g=\boldsymbol{-\pi^2/3}$ .

Step 3. On a large square contour $C_N$ with half-integer sides, $\cot\pi z$ stays bounded while $1/z^2$ forces $\oint_{C_N} g\,\mathrm{d}z\to0$ as $N\to\infty$ . By the Residue Theorem the total of all enclosed residues therefore vanishes: $\displaystyle\sum_{n\neq0}\frac{1}{n^2}-\frac{\pi^2}{3}=0$ .

Step 4. The double-sided sum is even, $\sum_{n\neq0}1/n^2=2\sum_{n\ge1}1/n^2$ , so $2\sum_{n=1}^{\infty}1/n^2=\pi^2/3$ and $\displaystyle\sum_{n=1}^{\infty}\frac{1}{n^2}=\boldsymbol{\frac{\pi^2}{6}}\approx1.6449$ — Euler's value $\zeta(2)=\pi^2/6$ , obtained purely from residue bookkeeping.

3 · A double pole (quick)

Find $\operatorname{Res}_{z=1}\dfrac{e^z}{(z-1)^2}$ . The pole is order $m=2$ , so formula (5) gives $\operatorname{Res}=\dfrac{1}{1!}\dfrac{\mathrm{d}}{\mathrm{d}z}\big[(z-1)^2 f\big]_{z=1}=\dfrac{\mathrm{d}}{\mathrm{d}z}e^z\Big|_{z=1}=\boldsymbol{e}$ . The naive simple-pole limit $\lim_{z\to1}(z-1)f=\lim_{z\to1}e^z/(z-1)$ diverges — a live reminder to check the order first.

Physics of the portrait

The residue calculus is the creation of one mathematician who, more than any other, turned the vague eighteenth-century manipulation of complex integrals into a rigorous machine — and who literally coined the word.

Origins · 1 minAugustin-Louis Cauchy and how this was discovered

Augustin-Louis Cauchy

1789-1857 · founder of rigorous complex analysis and the calculus of residues

Augustin-Louis Cauchy was the most prolific mathematician of the nineteenth century and the architect of nearly the entire theory in this article. Beginning with his 1814 memoir on definite integrals, he spent two decades building the rigorous complex analysis that his predecessors — Euler, Laplace, Poisson — had only groped toward with formal symbol-shuffling. In an 1825 memoir on integrals taken between imaginary limits he proved what is now Cauchy's Integral Theorem: the integral of an analytic function around a closed path is zero. Then, in 1826, in his self-published Exercices de mathématiques, he isolated the one quantity that a contour integral does detect at a singularity and named it the résidu — the residue — launching the "calculus of residues" as a systematic method. By the early 1830s he had the general Residue Theorem and the integral formula that bears his name, and he applied residues relentlessly to real integrals, series, and differential equations. Cauchy insisted on convergence and rigor in an era that tolerated neither; the modern definitions of limit, continuity, and the Cauchy sequence are all his, across nearly 800 papers. The Laurent series that supplies the coefficients he integrates was contributed in 1843 by Pierre Alphonse Laurent, extending Cauchy's Taylor expansion to the annulus around a singularity — the final piece that lets every residue be read off as a coefficient.

History

1814

— Cauchy's memoir on definite integrals seeds the theory of complex integration

1825

— Cauchy proves the integral theorem for integrals between imaginary limits

1826

— Cauchy coins the term residue and founds the calculus of residues in the Exercices

1831

— Cauchy states the general residue theorem and the integral formula

1843

— Laurent extends the expansion to an annulus, giving the Laurent series

Going further

Texts. Stone & Goldbart, Mathematics for Physics, ch. on Complex Analysis · Arfken, Weber & Harris, Mathematical Methods for Physicists, 7th ed., ch. on Calculus of Residues · Hassani, Mathematical Physics, ch. on Calculus of Residues · Byron & Fuller, Mathematics of Classical and Quantum Physics, ch. on Analytic Functions · Riley, Hobson & Bence, Mathematical Methods for Physics and Engineering · Ahlfors, Complex Analysis · Whittaker & Watson, A Course of Modern Analysis. Key results in this article: (1) $\oint(z-z_0)^n\,\mathrm{d}z=2\pi i\,\delta_{n,-1}$ , verified by direct parametrization $z-z_0=re^{i\theta}$ ; (2) the order- $m$ residue formula derived from the Laurent series; (3) $\int_{-\infty}^{\infty}\mathrm{d}x/(1+x^2)=\pi$ , consistent with $\arctan$ ; (4) the Basel sum $\sum1/n^2=\pi^2/6\approx1.6449$ using the expansion $\pi\cot\pi z=1/z-\pi^2z/3-\pi^4z^3/45-\cdots$ ; (5) the Yukawa integral $\int e^{ikx}/(k^2+m^2)\,\mathrm{d}k=(\pi/m)e^{-m|x|}$ , checked numerically to seven digits at $m=1.5,\ x=0.7$ ; (6) the residue $e$ of $e^z/(z-1)^2$ at $z=1$ .

Prerequisites. Laurent Series · Zeros, Poles, and Isolated Singularities · Cauchy's Integral Formula and Derivatives · Complex Integration and Contour Integrals · Cauchy's Integral Theorem. Leads to. Consequences: Liouville, Maximum Modulus, Morera · Elementary Complex Functions.

Ask the tutor: Bring me any integral, series, or transform you are stuck on and I will show you where to put the contour and which residues to collect.

What exactly is a residue, in one sentence

It is the coefficient of $(z-z_0)^{-1}$ in the Laurent expansion of the function about the singularity $z_0$ — equivalently, $\tfrac{1}{2\pi i}$ times the integral of the function around a small loop enclosing only $z_0$ . Those two descriptions are the same number viewed two ways: the series picture tells you how to compute it algebraically, while the integral picture tells you what it does. Of all the infinitely many Laurent coefficients, this is the single one that a contour integral can register, which is why it deserves its own name and its own battery of shortcut formulas.

Why is 2πi the magic constant that appears everywhere

It comes straight from the geometry of a circle. Parametrize a small loop as $z-z_0=re^{i\theta}$ with $\theta$ running from $0$ to $2\pi$ ; then $\mathrm{d}z=ire^{i\theta}\,\mathrm{d}\theta$ , and integrating $(z-z_0)^{-1}=e^{-i\theta}/r$ against it gives $\int_0^{2\pi} i\,\mathrm{d}\theta=2\pi i$ . Every other integer power carries a factor $e^{i(n+1)\theta}$ that completes a whole number of oscillations and averages to zero. So $2\pi i$ is nothing more exotic than "once around the circle" measured with the complex logarithm, and it is universal because it does not depend on the radius $r$ or on the particular function — only on winding once.

How do I choose which half-plane to close the contour in

Let the exponential decide. For an integrand carrying $e^{i k x}$ you need the added semicircle to make that factor decay, not blow up. Writing $k=\operatorname{Re}k+i\operatorname{Im}k$ , the magnitude is $e^{-x\operatorname{Im}k}$ : if $x>0$ this decays when $\operatorname{Im}k>0$ , so close in the upper half-plane; if $x<0$ , close below. For a rational integrand with no exponential you can close in either half-plane, provided the function falls off faster than $1/|z|$ so the arc vanishes — pick whichever side has the simpler poles. Jordan's lemma is the precise statement that legitimizes discarding the arc even when the rational factor only decays like $1/|z|$ .

When does the simple-pole formula fail, and what do I use instead

The formula $\operatorname{Res}=\lim_{z\to z_0}(z-z_0)f(z)$ is valid only for a pole of order one. If the pole has order $m>1$ , that single factor of $(z-z_0)$ cannot cancel all $m$ offending factors, and the limit diverges. The fix is equation (5): multiply by $(z-z_0)^m$ to regularize, differentiate $m-1$ times, evaluate at $z_0$ , and divide by $(m-1)!$ . Always determine the pole order first — usually by inspecting the lowest power in the denominator — and never apply the order-one shortcut on faith. For essential singularities and branch points, neither formula applies and you must read the residue from the Laurent series directly.

How can a contour integral possibly compute an infinite sum

The trick is to manufacture a function whose residues are the terms you want to add. The kernels $\pi\cot\pi z$ and $\pi/\sin\pi z$ both have simple poles with residue $\pm1$ at every integer, so multiplying one of them by $f(z)$ produces a function whose residue at $z=n$ is (up to sign) $f(n)$ . Integrating that product around a huge contour that swells to infinity gives zero when $f$ decays fast enough, and the Residue Theorem then says the sum of all residues — the sum $\sum f(n)$ plus the residues at $f$ 's own poles — vanishes. Rearranging isolates the series. That is exactly how the Basel sum $\sum1/n^2=\pi^2/6$ falls out, and it generalizes to every even zeta value and to finite-temperature Matsubara sums in physics.

Is the Residue Theorem really just Cauchy's theorem and formula combined

Yes, and seeing that is the whole point. Cauchy's Integral Theorem is the case with no singularities enclosed: the integral is zero. Cauchy's Integral Formula is the case of a single simple pole coming from a kernel $1/(z-z_0)$ : the integral is $2\pi i$ times the value of the analytic numerator, which is precisely its residue. The Residue Theorem is the general statement that contains both — deform the contour onto small circles (that is Cauchy's Theorem at work), expand each singularity's Laurent series, and keep only the $a_{-1}$ terms. No genuinely new principle is introduced; the theorem is the organizing statement that makes the earlier two look like the two simplest instances of one accounting rule.

What about branch points, essential singularities, and the residue at infinity

Poles are the easy case. Essential singularities, like $e^{1/z}$ at the origin, still have a residue — you read it as the $a_{-1}$ coefficient of the Laurent series ( $1$ , here) — but no pole formula produces it. Branch points, such as those of $\sqrt{z}$ or $\log z$ , are not isolated singularities at all: the function is multivalued around them, and you must use a keyhole or dogbone contour that runs along both sides of a branch cut rather than counting a residue. The residue at infinity, $\operatorname{Res}_\infty f=-\operatorname{Res}_{w=0}[w^{-2}f(1/w)]$ , closes the accounting on the Riemann sphere: for any rational function the finite residues plus the one at infinity sum to zero, which is often the fastest way to get a stubborn residue and always a good consistency check.