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

Inverting Transforms: The Bromwich Contour

The Laplace transform turns a hard differential equation into easy algebra — but then you have to climb back. This guide shows how a single vertical line in the complex plane, closed into a loop and read by residues, inverts the transform and hands you the answer in time.

Where the integral lives: a vertical line

You meet the Laplace transform long before complex analysis: it sends a function f(t) of time, defined for t >= 0, to a function F(s) = integral from 0 to infinity of e^(-s t) f(t) dt of a new variable s. The point of it is leverage. Differentiating in time becomes multiplying by s; an awkward differential equation in f(t) collapses into ordinary algebra in F(s), which you solve in a line. The catch is the return journey. Once you hold F(s), how do you recover the f(t) it came from? That inverse step is where the complex plane finally has to be invoked, and it is the whole subject of this guide.

Here is the surprise. The transform F(s) is not just a real formula — fed a complex s, the integral converges and defines a genuine meromorphic function of s, holomorphic wherever the integral makes sense and decorated with isolated singularities elsewhere. The integral converges in a half-plane Re s > c, where c, the abscissa of convergence, sits just to the right of f's worst growth. So F(s) is an analytic object you may move around the plane, and the inversion formula reflects exactly that: it integrates F(s) e^(s t) up a vertical line Re s = gamma, where gamma is any real number larger than c — a line drawn safely to the right of every singularity.

F(s) = integral from 0 to infinity of e^(-s t) f(t) dt        (the transform)

            1            gamma + i*infinity
f(t)  =  -------  integral                  F(s) e^(s t) ds      (the inverse)
          2 pi i        gamma - i*infinity

   the path is the vertical line  Re s = gamma,  with  gamma > c
The Bromwich inversion: integrate F(s) e^(s t) straight up a vertical line lying to the right of every singularity of F.

The Bromwich contour: close the line into a loop

A vertical line stretching from gamma - i*infinity up to gamma + i*infinity is not a closed loop, so the residue theorem cannot touch it directly — and yet a bare infinite line integral is exactly the kind of object that machinery devours. The fix is the move that names the whole method. Close the line into a loop by adding a large arc — for t > 0, a big semicircle bulging out to the LEFT, into the half-plane Re s < gamma — so the straight piece and the arc together bound a region. That closed path is the Bromwich contour, and the inversion problem becomes a contour integral you can finally read.

Why bulge to the left, and only for t > 0? Look at the factor e^(s t) you are dragging along. Write s = gamma + R e^(i theta) on the arc; then e^(s t) carries a magnitude e^(t * Re s), and on the left half (Re s < gamma) that exponential is being driven DOWN as Re s falls. For positive t, pushing the arc leftward makes e^(s t) decay, and the contribution of the far arc can be made to vanish. For t < 0 the sign flips: you must instead close to the RIGHT, where F is analytic, every residue you collect is zero, and the answer is f(t) = 0 — which is exactly right, since f was only defined for t >= 0. The geometry is enforcing causality.

Making the arc vanish: Jordan's lemma

It is tempting to wave the arc away by saying 1/s shrinks, but that alone is too weak: the arc's length grows like pi R, and a crude bound pairs a 1/R decay against an R-sized path, giving a constant that refuses to die. The honest tool is Jordan's lemma, which exploits the oscillation in e^(s t) rather than just its size. On the left-closing arc the magnitude e^(t * Re s) is not merely small somewhere — it decays so steeply away from the real axis that, once you integrate over the arc, the answer is genuinely squeezed to zero. The lemma needs only that F(s) -> 0 uniformly as |s| -> infinity along the arc, a mild condition most transforms of decent functions satisfy.

There is a fine print worth saying out loud, because it trips people. Jordan's lemma in this form wants t > 0 strictly. At t = 0 the e^(s t) factor is just 1, it provides no decay, and the clean argument breaks; the inversion there returns the average of the left and right limits of f at a jump, not a single value — the same midpoint behaviour you may remember from Fourier series at a discontinuity. And if F decays only like 1/s without any extra room, you are right at the boundary of what the lemma allows: you keep the e^(s t) oscillation but cannot afford to also throw away a 1/s, so the bookkeeping must be done carefully rather than by reflex.

Reading the answer off the poles

Once the arc is gone, the picture is gorgeously simple. The line integral equals 2 pi i times the sum of residues, the 1/(2 pi i) out front cancels that factor exactly, and you are left with f(t) = sum of the residues of F(s) e^(s t) at every singularity to the left of the line. For the common case where F is a ratio of polynomials, those singularities are just poles, and each one contributes a clean piece of the time signal. A simple pole at s = a contributes a term proportional to e^(a t); the real part of a sets the growth or decay rate, the imaginary part sets the oscillation frequency. The poles of F literally are the building blocks of f(t).

  1. Locate every singularity of F(s) e^(s t) — for a rational F, these are the zeros of the denominator, i.e. the poles.
  2. Check that all of them lie to the LEFT of the line Re s = gamma, so the leftward Bromwich contour encloses them.
  3. Confirm F(s) -> 0 on the big arc so Jordan's lemma kills the arc for t > 0.
  4. Compute the residue of F(s) e^(s t) at each enclosed pole and add them; the sum is f(t).

Watch it run on the gentlest example. Take F(s) = 1/(s - a), the transform of a single exponential. Its only singularity is a simple pole at s = a, and for any gamma > Re a the Bromwich contour encloses it. The residue of F(s) e^(s t) = e^(s t)/(s - a) at s = a is, by the simple-pole rule, just e^(a t). Summing that one residue gives f(t) = e^(a t) — and indeed the Laplace transform of e^(a t) is 1/(s - a). The machine recovered exactly the function we started from, with the pole's location a appearing directly as the exponential rate.

When the picture is not all poles

The tidy pole story is the friendly half of the subject. Plenty of honest transforms have a branch point instead — F(s) = 1/sqrt(s), say, whose inverse is 1/sqrt(pi t), or transforms involving sqrt(s) that arise from the heat equation. At a branch point you cannot just sum residues, because F is not even single-valued there; you must run a branch cut out from the point and deform the Bromwich contour around it, turning the inversion into a branch-cut integral that wraps the cut like a keyhole. The same closing-and-reading philosophy holds, but the bookkeeping is richer and the answer often comes out as a real integral rather than a finite sum of exponentials.

Two honest warnings before you reach for residues automatically. First, if F has infinitely many poles — as transfer functions of distributed systems do — you collect an infinite series of exponential terms, and you owe a convergence argument, not just a hopeful sum. Second, an essential singularity is genuinely beyond the simple residue recipes; there the local behaviour is wild and you must expand by hand. None of this breaks the Bromwich method — it is still close the contour, justify the arc, sum the local contributions — it merely reminds you that the residue theorem is a precise instrument, not a slogan, and each of its hypotheses has to be paid for.