Cauchy's Theorem

Suppose a function is complex-differentiable at every point of a region. Does that alone force integrals of the function around closed loops to vanish? This holds no matter how wildly the loop bends and winds. This question sits at the heart of complex analysis. The surprising answer — yes, provided the domain has no holes — is Cauchy's theorem. It is surprising because the analogous statement in real analysis is false. A real differentiable function need not have any special integral-vanishing property over closed curves. Even in the plane, smooth vector fields routinely have nonzero circulation. Holomorphy is a far stronger constraint than real differentiability. Cauchy's theorem is the first place this strength becomes visible. It appears as a global, geometric fact rather than a local, algebraic one. The result matters because it is the hinge on which the rest of classical complex analysis turns. From it flows Cauchy's integral formula. This formula recovers the values of a holomorphic function inside a curve from its values on the curve. From that flows the fact that holomorphic functions are automatically analytic. They are infinitely differentiable and rigid in ways real functions never are. Path independence of contour integrals, the existence of primitives, and the residue calculus all trace back to this. Liouville's theorem and the maximum modulus principle also trace back to this single vanishing statement. The big picture is topological as well as analytic. Cauchy's theorem is really a statement about an interaction. It concerns the local condition of holomorphy — captured by the Cauchy-Riemann equations — and the global shape of the domain, captured by simple connectedness. A holomorphic function on an annulus can have nonzero integral around a loop encircling the hole. The same function, restricted to a simply connected piece of that annulus, cannot. This dichotomy is what makes complex analysis feel like geometry and analysis in the same breath. It is the reason winding numbers, homotopy, and residues all enter the subject through the same door. This article gives the precise definition of the objects involved. It states and completely proves Cauchy's theorem for continuously differentiable holomorphic functions via Green's theorem. It builds the geometric intuition with figures, works two nontrivial examples, and shows exactly what breaks when simple connectedness or holomorphy fails. It closes by connecting the theorem to Cauchy's Integral Formula and to Antiderivatives and Path Independence.

Key takeaways

Cauchy's theorem says: if $f$ is holomorphic on a simply connected domain $D$ , then $\oint_\gamma f(z)\,dz = 0$ for every closed contour $\gamma$ in $D$ .
The proof (in the smooth case) reduces the complex line integral to two real line integrals and applies Green's theorem, using the Cauchy–Riemann equations to kill both resulting double integrals.
Simple connectedness is essential: on domains with holes, holomorphic functions like $1/z$ can have nonzero integral around a loop encircling the hole.
The theorem is purely topological in its hypothesis on the domain but purely analytic in its hypothesis on the function — both are needed.
It immediately yields path-independence of contour integrals and the existence of holomorphic antiderivatives on simply connected domains.
It is the direct ancestor of Cauchy's integral formula, the residue theorem, and the entire machinery of contour integration.
Key idea: holomorphy plus no holes forces every closed-loop integral to vanish, because Cauchy–Riemann makes the relevant circulation and flux integrands identically zero.

Where this lives today

Cauchy's theorem is the foundation of the residue theorem, which turns contour integration into a bookkeeping device for evaluating real definite integrals (improper integrals, Fourier transforms, inverse Laplace transforms) by summing residues at enclosed poles — a technique used constantly in engineering signal processing and control theory. In physics, contour integration via Cauchy's theorem underlies dispersion relations in quantum field theory (Kramers–Kronig relations), the evaluation of Green's functions and propagators, and stability analysis of dynamical systems through the Nyquist criterion, which literally counts encirclements of a point by a contour image — a direct descendant of the winding-number ideas that Cauchy's theorem makes rigorous. In pure mathematics, it underlies the proof that holomorphic functions are analytic (via Cauchy's Integral Formula, obtained by applying the theorem to a specially chosen auxiliary contour), the theory of conformal mapping in geometry, and analytic number theory, where contour shifting arguments used to prove the prime number theorem rely on Cauchy's theorem to justify moving integration paths across the complex plane without changing certain integral's values, as long as no singularities (poles of the Riemann zeta function, for instance) are crossed. It also underlies Antiderivatives and Path Independence: on a simply connected domain, Cauchy's theorem is exactly what guarantees that $\int_{z_0}^{z} f(w)\,dw$ is well-defined independent of path, allowing construction of a holomorphic antiderivative $F$ with $F'=f$ .

Ask the tutor: How does Goursat's triangle-bisection argument prove the theorem without assuming $f'$ is continuous?

D \subseteq \mathbb{C}

Intuition

Think of $f\,dz$ as encoding two real planar vector fields at once. Writing $f = u+iv$ and $dz = dx + i\,dy$ ,

f(z)\,dz = (u+iv)(dx+i\,dy) = (u\,dx - v\,dy) + i(v\,dx + u\,dy).

The real part is the line integral of the vector field $(u,-v)$ , and the imaginary part is the line integral of $(v,u)$ . Green's theorem converts each of these closed-loop line integrals into a double integral, over the region enclosed, of a specific combination of partial derivatives. For the real part that combination is $-v_x - u_y$ ; for the imaginary part it is $u_x - v_y$ . The Cauchy–Riemann equations say precisely that both of these combinations vanish identically. So both double integrals are integrals of the zero function, hence zero, and the whole contour integral is zero. This is the entire content of the theorem in the smooth case: holomorphy is exactly the condition that makes these two circulation integrands cancel. The role of simple connectedness is to guarantee that Green's theorem applies — that the closed curve actually bounds a region lying inside $D$ , where $f$ is holomorphic and the partials behave. If $D$ has a hole and the curve loops around it, the "enclosed region" is not entirely in $D$ ; $f$ may fail to be holomorphic (or even defined) at the missing point, and the cancellation argument breaks down. This is exactly the case of $f(z) = 1/z$ around a hole at the origin, worked out below.

Figure

Left: a simply connected domain with a contractible loop, integral forced to zero. Right: an annulus with a hole; the same shape of loop cannot be shrunk to a point without leaving the domain.

D \subseteq \mathbb{C}

R \subset \mathbb{R}^2

C^1

D

Worked examples

Example 1: A polynomial times an exponential, over a non-circular contour. Let $f(z) = (z^2 - 3z + 1)e^{z}$ , which is entire (holomorphic on all of $\mathbb{C}$ , since it is built from polynomials and $e^z$ by products, both of which preserve holomorphy). Let $\gamma$ be the boundary of the triangle with vertices $0$ , $4$ , and $2+3i$ , positively oriented — an asymmetric, non-circular closed contour. Since $\mathbb{C}$ is simply connected (any loop in the plane can be contracted to a point through the plane itself) and $f$ is holomorphic on all of $\mathbb{C} \supset D$ for $D=\mathbb{C}$ , Cauchy's Theorem (Cauchy–Goursat, smooth case) applies directly with $D = \mathbb{C}$ . Hence

\oint_\gamma (z^2-3z+1)e^z\,dz = 0,

regardless of the specific triangle chosen, its side lengths, or its orientation details beyond sign. No computation of the antiderivative or parametrization of the three edges is needed — the vanishing is immediate from holomorphy on a simply connected domain containing $\gamma$ and its interior. To see the mechanism explicitly (as a check), note $u_x - v_y$ and $-v_x-u_y$ vanish pointwise by Cauchy–Riemann as proved in Step 4 above; this is what makes the triangle's specific shape irrelevant. Contrast this with a real vector field like $(-y,x)$ , whose circulation around any triangle containing the origin is twice its area — never zero. The vanishing here is special to holomorphic integrands, not a generic feature of smooth planar integrands.

Example 2: Where the theorem does not directly apply, and how to fix the domain. Let $g(z) = \dfrac{e^{iz}}{z^2+4}$ and let $\gamma$ be the rectangle with vertices $-3-i$ , $3-i$ , $3+i$ , $-3+i$ (positively oriented), a rectangle centered at the origin. The function $g$ is holomorphic wherever $z^2+4 \neq 0$ , i.e. on $\mathbb{C}\setminus\{2i,-2i\}$ . The point $z=2i$ lies outside the rectangle (since the rectangle only reaches height $\pm 1$ ), but $z=-2i$ also lies outside. So in fact both singularities are outside $\gamma$ 's interior. Let $D$ be the open rectangle $\{-4 < \mathrm{Re}(z) < 4,\ -2 < \mathrm{Im}(z) < 2\}$ minus nothing — since neither $\pm 2i$ lies in $\overline{D}$ (as $|{\pm}2i| = 2$ is exactly the boundary height of $D$ ; choose $D$ slightly smaller in imaginary extent, say height strictly less than $2$ , e.g. $|\mathrm{Im}(z)|<1.9$ , to keep $\pm 2i \notin \overline D$ safely). This $D$ is simply connected (a rectangle is convex, hence simply connected), $g$ is holomorphic on $D$ , and $\gamma \subset D$ with its interior in $D$ . Cauchy's theorem gives

\oint_\gamma \frac{e^{iz}}{z^2+4}\,dz = 0.

Now modify the contour: let $\gamma'$ be the rectangle with vertices $-3-3i, 3-3i, 3+3i, -3+3i$ , which does enclose $z=2i$ (and $-2i$ ). Here no simply connected domain containing $\gamma'$ and its interior avoids both singularities, so this instance of Cauchy's theorem as stated does not apply to $g$ on such a domain — the integral is in fact generally nonzero and is computed instead by the residue theorem (a direct descendant of Cauchy's theorem, covered under Cauchy's Integral Formula and its generalizations), giving

\oint_{\gamma'} \frac{e^{iz}}{z^2+4}\,dz = 2\pi i \cdot \operatorname{Res}_{z=2i}\frac{e^{iz}}{z^2+4} = 2\pi i \cdot \frac{e^{i(2i)}}{2(2i)} = 2\pi i \cdot \frac{e^{-2}}{4i} = \frac{\pi e^{-2}}{2}.

This second computation is not itself a proof of Cauchy's theorem; it illustrates precisely the boundary of the theorem's applicability — when a singularity is enclosed, one must either shrink the domain to exclude it (making the integral zero on the smaller region, vacuously not applying to $\gamma'$ ) or leave holomorphy's simply-connected hypothesis behind and compute via residues.

D = \mathbb{C}\setminus\{0\}

Scope, edge cases & common traps

Situation	What happens	Why
$D$ simply connected, $f$ holomorphic on $D$	$\oint_\gamma f\,dz = 0$ for all closed $\gamma\subset D$	Full theorem applies
$D$ has a hole, loop encircles it	integral generally $\neq 0$	Green's theorem region exits $D$ ; see counterexample
$D$ simply connected, $f$ holomorphic except at finitely many points enclosed by $\gamma$	integral generally $\neq 0$ , computed via residues	Domain of holomorphy is $D$ minus those points, no longer simply connected around $\gamma$
$f$ only real-differentiable (not holomorphic) on simply connected $D$	integral generally $\neq 0$	Cauchy–Riemann equations fail, double integrals in Step 4 do not vanish
$f'$ merely exists (not assumed continuous)	theorem still holds (Goursat's refinement) but the proof above (via Green's theorem) does not directly apply	Goursat's proof avoids Green's theorem via a triangle-bisection argument; see below
$\gamma$ passes through a singularity of $f$	contour integral may be improper/undefined	$f$ not defined at points of $\gamma$ itself

A remark on rigor: the proof given uses Green's theorem, which additionally requires $u,v \in C^1$ , i.e. $f'$ continuous. Goursat proved in 1900 that continuity of $f'$ is not needed — complex differentiability alone suffices — via a clever bisection argument that shrinks any triangle on which the integral is nonzero to a point, deriving a contradiction with the definition of the derivative at the limiting point. That argument is independent of Green's theorem and is the reason the result is often called the Cauchy–Goursat theorem. It is not reproduced here in full, but readers should know the $C^1$ hypothesis in the theorem above is a genuine (removable) restriction of this particular proof, not of the theorem's true scope.

The mathematician behind it

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

Augustin-Louis Cauchy

1789–1857 · Post-revolutionary France, École Polytechnique and the founding of rigorous analysis

Cauchy was among the principal architects of modern analysis, rigorizing calculus with epsilon-delta style limit arguments and founding complex analysis as an independent discipline with a systematic theory of contour integration. Trained as an engineer before turning to mathematics, he brought an insistence on rigor that was unusual for his era, reacting against the looser 18th-century style of Euler and Lagrange. He was extraordinarily prolific, publishing over 800 papers, and politically inflexible — his royalist sympathies led to two periods of exile from French academic positions. Cauchy's theorem on contour integrals, first announced around 1814 and developed over subsequent decades, became the cornerstone on which he and later mathematicians built the residue calculus, the theory of power series representations of holomorphic functions, and the modern treatment of complex integration.

History

- 1814: Cauchy submits his first memoir on definite integrals taken between imaginary limits, containing early versions of the vanishing-integral idea.

- 1825: Cauchy publishes a more complete treatment of integrals of complex functions over paths, sharpening the contour integral framework.

- 1831: Cauchy, in exile in Turin, further develops the integral formula and residue ideas building directly on the vanishing theorem.

- 1851: Riemann's dissertation formalizes the geometric, mapping-theoretic viewpoint on holomorphic functions, reframing Cauchy's analytic results geometrically.

- 1900: Édouard Goursat proves the theorem assuming only that

f'

exists, removing Cauchy's implicit continuity-of-derivative assumption via a triangle-subdivision argument.

Why does Cauchy's theorem fail for merely real-differentiable functions?

The proof crucially uses the Cauchy–Riemann equations to force the two double integrals produced by Green's theorem to vanish identically. A function that is differentiable as a map $\mathbb{R}^2 \to \mathbb{R}^2$ but does not satisfy $u_x=v_y$ , $u_y=-v_x$ has no reason for those integrands to cancel. For example, $f(x+iy) = x - iy$ (complex conjugation) is real-differentiable everywhere with continuous partials, but $u=x,v=-y$ gives $u_x=1\neq v_y=-1$ , violating Cauchy–Riemann. Indeed $\oint_{|z|=1}\bar z\,dz = 2\pi i \neq 0$ , confirming the vanishing genuinely depends on holomorphy, not mere smoothness.

Does the theorem require the contour to be simple (non-self-intersecting)?

No. The statement holds for any closed piecewise- $C^1$ contour, simple or not, as long as it lies in the simply connected domain. The proof handles the general case by decomposing an arbitrary closed contour into finitely many simple closed sub-loops (using simple connectedness to guarantee each sub-loop bounds a region inside the domain), proving the vanishing for each simple piece via Green's theorem, and summing; shared edges traversed in opposite directions cancel algebraically, reconstructing the original integral as a sum of zeros.

Is simple connectedness of the domain the same as the contour being contractible?

They are related but distinct. Simple connectedness is a property of the domain $D$ : every loop in $D$ is contractible within $D$ . Given that, any particular closed contour $\gamma$ in $D$ is automatically contractible, because $D$ itself has this property. One does not need to check contractibility contour-by-contour; simple connectedness of the ambient domain guarantees it for all contours at once, which is why the theorem's hypothesis is stated on $D$ rather than on $\gamma$ .

Can Cauchy's theorem be used even when

f

has singularities, as long as they're outside the contour?

Yes, and this is the typical situation in applications. One restricts attention to a simply connected sub-domain $D$ that contains $\gamma$ and its interior but excludes the singularities; $f$ is holomorphic on this $D$ , and the theorem applies directly, giving a vanishing integral. The key technical point, illustrated in Example 2, is verifying that a suitable simply connected $D$ can be found — typically by choosing a disk or rectangle small enough to avoid the singular points while still containing the contour and its interior.

How does Cauchy's theorem lead to path-independence of contour integrals?

If $D$ is simply connected and $f$ is holomorphic on $D$ , take any two paths $\gamma_1,\gamma_2$ in $D$ from $z_0$ to $z_1$ . The concatenation of $\gamma_1$ with the reverse of $\gamma_2$ is a closed contour in $D$ , so by Cauchy's theorem its integral is zero. Splitting that integral into the two pieces (the second with a sign flip from reversal) shows $\int_{\gamma_1} f\,dz = \int_{\gamma_2} f\,dz$ . Hence the integral from $z_0$ to $z_1$ depends only on the endpoints, not the path, which is exactly Antiderivatives and Path Independence.

Why is Goursat's version of the theorem considered stronger, and is it still "the same theorem"?

Cauchy's original argument (and the proof given above) assumes $f'$ is continuous, which is a real extra hypothesis needed to invoke Green's theorem. Goursat showed in 1900 that this hypothesis is unnecessary: complex differentiability alone, with no continuity assumption on the derivative, still forces the integral to vanish, via a self-contained triangle-bisection argument that never invokes Green's theorem. It is the same conclusion under a weaker hypothesis, so it is a genuine strengthening; the combined statement is often named the Cauchy–Goursat theorem to credit both contributions.

Does Cauchy's theorem generalize to domains that are not simply connected?

Yes, via the notion of winding number. On a general domain $D$ (possibly with holes), if $\gamma$ is a closed contour that is null-homologous in $D$ (meaning its winding number around every point outside $D$ is zero), then $\oint_\gamma f\,dz = 0$ still holds for $f$ holomorphic on $D$ . This homological form of Cauchy's theorem includes the simply connected case as a special instance, since in a simply connected domain every closed contour automatically has winding number zero around every excluded point.