JOVANA
Explore Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Deforming Contours and the Winding Number

Cauchy's theorem says a holomorphic loop integral is zero in a hole-free region. This guide turns that into a moving picture: you may bend and slide a contour freely as long as you never cross a singularity, and the integer that counts how many times a loop wraps a point becomes the bookkeeping that runs the rest of complex analysis.

From "the integral is zero" to "the integral does not move"

The previous guide left you holding a sharp tool with one frustrating limit: Cauchy's integral theorem gives zero only when the loop encloses no bad point. The moment a singularity is trapped inside — think 1/z around the origin, with its stubborn value 2 pi i — the zero answer fails. This guide salvages everything. The key is a small but powerful shift in viewpoint: instead of asking whether an integral is zero, ask whether it CHANGES when you move the contour.

Here is the clean statement of deformation of contours. Suppose f is holomorphic everywhere in some region, and you have two closed contours, C_1 and C_2, that can be continuously slid into each other while staying inside that region — never sweeping across a point where f fails to be holomorphic. Then the two integrals are equal: integral over C_1 of f dz = integral over C_2 of f dz. The shape of the curve no longer matters; only the holes it does or does not pass over do.

Why deformation is just Cauchy's theorem in disguise

Deformation invariance is not a new theorem; it is the old one looked at sideways. To see why two loops C_1 and C_2 give the same integral, build a single closed path that goes out along C_1, runs over a thin bridge to C_2, comes back around C_2 in the OPPOSITE direction, and returns along the bridge. This compound loop encloses no singularity — the bad point sits in the gap between the two curves, outside the region swept — so by Cauchy's theorem its total integral is zero.

Now read off the pieces. The two trips along the thin bridge run in opposite directions over the same path, so they cancel exactly. What remains is the integral over C_1 plus the integral over C_2 reversed. Setting the sum to zero gives integral over C_1 = integral over C_2. That is the whole proof — a bridge, a cancellation, and one application of the theorem you already trust. This same picture, with two cuts instead of one, handles a region with a hole punched in the middle.

In symbols the compound loop is C_1, then the bridge out, then C_2 traversed backwards, then the bridge back: a single closed curve enclosing no singularity, so its total integral is zero. The bridge legs cancel, leaving integral over C_1 minus integral over C_2 equal to zero — that is, the two integrals agree. It is worth savouring how little this costs: one auxiliary curve and the theorem you already trust, and deformation falls out for free.

Homotopy: the precise idea of "slid into each other"

The phrase "continuously slid into each other" deserves a real definition, and it is homotopy. Two loops in a region are homotopic if there is a continuous family of loops, parametrised by a time s running from 0 to 1, that starts at the first loop, ends at the second, and stays inside the region the whole time. Picture a film clip: frame 0 shows C_1, frame 1 shows C_2, and every frame in between is a legal loop avoiding all the holes. Homotopy invariance of the integral says every frame of that clip carries the same value.

This reframes the last two guides in one sentence. In a simply connected domain every loop is homotopic to a point — you can shrink it away entirely — and a constant point-loop has integral zero, so every loop integrates to zero. That is exactly Cauchy's theorem. In a multiply connected domain a loop encircling a hole CANNOT be shrunk to a point, because the hole blocks the way; such a loop sits in a different homotopy class, and its integral need not vanish. The topology of the region, not the algebra of the curve, decides everything.

The winding number: counting the wraps

Homotopy classes near a single hole are labelled by one integer, and that integer is the winding number. The winding number of a closed contour C about a point z_0 counts how many net times C travels counterclockwise around z_0 — a circle once around gives +1, twice around gives +2, the same circle traversed clockwise gives -1, and a loop that never encircles z_0 gives 0. The miracle is that this purely geometric count is computed by a specific integral.

n(C, z_0)  =  (1 / (2 pi i))  *  integral over C of  dz / (z - z_0)

   one CCW circle about z_0 :  n = +1
   k times CCW            :  n = +k
   one CW circle          :  n = -1
   z_0 outside the loop   :  n =  0
The winding number is an integral that always returns an integer — the net number of counterclockwise wraps.

Where does this formula come from? It is the 1/z integral wearing a disguise. Recall that 1/(z - z_0) has the multivalued primitive log(z - z_0); as you travel once counterclockwise around z_0, the argument increases by 2 pi while the modulus returns to its start, so the integral picks up exactly 2 pi i per loop. Dividing by 2 pi i strips that down to the bare count of laps. That the answer is always a whole number is itself a small marvel: a continuous integral is forced, by topology, to land on an integer.

Putting it together: the residue-counting machine

Deformation and the winding number combine into a complete accounting rule. Take any loop and a function holomorphic except at a few isolated bad points. Deform the loop, without crossing those points, into a bundle of tiny circles, one hugging each singularity it encloses. By deformation the original integral equals the sum over those small circles, and each small circle's contribution is its winding number times a number attached to that singularity. That attached number, you will soon learn, is the residue.

  1. List the singular points of f, and check that the contour C passes through none of them.
  2. For each singular point z_k, compute the winding number n(C, z_k) — usually just +1 for the points inside a simple counterclockwise loop, and 0 for the points outside.
  3. Deform C, without touching any z_k, into separate tiny circles, one around each enclosed singularity; the total integral is unchanged.
  4. Read the integral as 2 pi i times the sum, over all singular points, of (winding number) x (residue at that point). That single line is the residue theorem.

This is the shape of the residue theorem, the engine of the next rung, and you can see now that it is nothing but deformation plus winding numbers carried to their conclusion. The same machinery, applied to f(z)/(z - z_0), collapses to the Cauchy integral formula, and the more abstract homology form of Cauchy's theorem simply states this winding-number bookkeeping for any contour at once. One honest idea — you can bend a contour freely until it hits a hole, and integers count the holes — has quietly become the spine of complex integration.