Residues and poles: Visual walkthrough — Maths deep dive

A visual, from-zero walkthrough of the parent result Residues and poles. We are going to prove, with a picture at every step, why the messy-looking integral $\oint_{C} f (z) d z = 2 π i Res (f, z_{0})$ collapses into one single number. No prior fluency needed — we build every symbol as it appears.

What we are even looking at (setup in plain words)

Before any algebra, let us agree on the picture.

$z$ is a complex number. Think of it as a point on a flat sheet (the "complex plane"): the horizontal distance is its real part, the vertical distance is its imaginary part. So $z$ is just an arrow-tip location on graph paper.
$f (z)$ is a machine: feed it a point $z$ , it spits out another complex number. Nothing mystical.
A contour $C$ is a closed loop drawn on that sheet — you start somewhere, wander around, and come back. We will use a circle because circles are the easiest loops to describe with an angle.
$\oint_{C} f (z) d z$ means: walk once around the loop, adding up $f (z)$ times each tiny step $d z$ you take. The little circle on the integral sign $\oint$ just says "the path is a closed loop".
A singularity $z_{0}$ is a point where $f$ misbehaves (blows up to infinity). Our whole story is about what a loop feels as it circles such a point.

Step 1 — Zoom in on the singularity and write $f$ as a sum of simple pieces

WHAT. Very close to the trouble-spot $z_{0}$ , we replace the complicated $f$ by an infinite sum of pure powers of $(z - z_{0})$ . This is the Laurent series:

$f (z) = add up all powers n = - \infty \sum \infty a fixed number a_{n} how far we are from z_{0}, raised to power n (z - z_{0})^{n}$

Reading the symbols where they sit:

$(z - z_{0})$ is the displacement arrow from the bad point to where we currently stand.
$n$ runs over all whole numbers — positive and negative. Negative $n$ means $(z - z_{0})^{- 1} = \frac{1}{z - z _{0}}$ , which grows huge as we approach $z_{0}$ . Those negative-power terms are the "blow-up" pieces.
$a_{n}$ is just the fixed weight attached to each power. The star of the show will be $a_{- 1}$ .

WHY. Powers of $(z - z_{0})$ are the only thing we know how to integrate around a circle cleanly. So we trade the hard $f$ for a pile of easy pieces, and integrate each piece separately.

PICTURE. Each power $(z - z_{0})^{n}$ is a little "flow field" on the sheet. Step 1's figure shows three sample terms stacked into one function.

Step 2 — Turn the loop into an angle: parametrize the circle

WHAT. We choose $C$ to be a circle of radius $ρ$ centred on $z_{0}$ . Every point on that circle is

$z = z_{0} + ρ e^{i θ}, θ \in [0, 2 π]$

Term by term:

$ρ$ (Greek "rho") is the fixed radius — how far the loop sits from the centre.
$e^{i θ}$ is a unit arrow pointing at angle $θ$ . As $θ$ sweeps $0 \to 2 π$ , this arrow spins once fully around. (That is the whole meaning of $e^{i θ}$ : a direction dial.)
So $z_{0} + ρ e^{i θ}$ = "start at centre, step out by $ρ$ in direction $θ$ " — exactly the rim of the circle.

The tiny step $d z$ becomes, by differentiating, $d z = i ρ e^{i θ} d θ$ where $d θ$ is the sliver of angle we advance, and the factor $i ρ e^{i θ}$ says the step points tangent to the circle (the $i$ is a $9 0^{\circ}$ turn — perpendicular to the radius, i.e. along the rim).

WHY. An integral over a wandering path is awkward. Once everything is written in the single dial-variable $θ$ , the loop integral becomes an ordinary integral from $0$ to $2 π$ — school-level calculus.

PICTURE. Step 2's figure shows the radius arrow $ρ e^{i θ}$ and the tangent step $d z$ at a right angle to it.

Step 3 — Integrate ONE power term and watch the exponentials spin

WHAT. Take a single term $(z - z_{0})^{n}$ and push the parametrization through. Since $(z - z_{0}) = ρ e^{i θ}$ :

= i\,\rho^{\,n+1}\int_0^{2\pi} e^{\,i(n+1)\theta}\,d\theta$$ What each piece did: - $\rho^n e^{in\theta}$ is the term itself on the rim. - $i\rho e^{i\theta}\,d\theta$ is the step $dz$. - Multiplying, the two exponentials merge: $e^{in\theta}\cdot e^{i\theta}=e^{i(n+1)\theta}$. **The powers of $\rho$ combine into $\rho^{n+1}$**, and everything constant slides outside the integral. **WHY.** We are left with the cleanest possible object: $\int_0^{2\pi} e^{i k\theta}\,d\theta$ with $k=n+1$. This one integral decides the whole theorem. **PICTURE.** Step 3's figure shows the merged spinning arrow $e^{i(n+1)\theta}$ tracing a circle as $\theta$ advances. ![[deepdives/dd-maths-4.10.05-d2-s04.png]] --- ## Step 4 — The magic cancellation: a full spin averages to zero (unless it doesn't spin) **WHAT.** Evaluate $\displaystyle\int_0^{2\pi} e^{ik\theta}\,d\theta$ for an integer $k$. $$\int_0^{2\pi} e^{ik\theta}\,d\theta= \begin{cases} 2\pi & k=0\\[4pt] 0 & k\neq 0 \end{cases}$$ Why each case, *geometrically*: - If $k\neq 0$: the arrow $e^{ik\theta}$ spins around the origin $|k|$ complete times as $\theta$ goes $0\to 2\pi$. Adding up arrows that point in **every direction equally** gives a perfect cancellation — the sum is the centre of the circle, i.e. $0$. Every push has an equal-and-opposite push half a turn later. - If $k=0$: then $e^{i\cdot 0\cdot\theta}=e^0=1$. The arrow *never moves* — it just sits pointing right the whole time. Adding a constant $1$ over an interval of length $2\pi$ gives $2\pi$. Nothing cancels because nothing spun. **WHY.** This is the heart of the whole subject: **spinning things cancel, non-spinning things accumulate.** The integral is a lie detector that returns $0$ for every genuinely-spinning term. **PICTURE.** Step 4's figure puts the two cases side by side: a spinning arrow whose tips cancel to the centre (left) versus a frozen arrow that piles up (right). ![[deepdives/dd-maths-4.10.05-d2-s05.png]] > [!formula] The single decisive result > $$\oint_C (z-z_0)^n\,dz=\begin{cases}2\pi i & n=-1\\[3pt]0 & n\neq -1\end{cases}$$ > The exponent is $k=n+1$. It is zero — the only surviving case — **exactly when $n=-1$**. (The leftover $i$ comes from the $i$ inside $dz$.) --- ## Step 5 — Which term survives? Only $n=-1$. Meet the residue. **WHAT.** We now sum the individual results back up. Feed the whole Laurent series into the loop: $$\oint_C f(z)\,dz=\sum_{n=-\infty}^{\infty} a_n \underbrace{\oint_C (z-z_0)^n\,dz}_{\;=\,0\text{ for every }n\ne -1}=a_{-1}\cdot(2\pi i)$$ Reading it: - Every single term with $n\neq -1$ contributed $a_n\cdot 0=0$. Gone. - The lone term $n=-1$ contributed $a_{-1}\cdot 2\pi i$. So the entire, possibly-infinite function is squeezed by the loop down to **one number times $2\pi i$**: $$\boxed{\;\oint_C f(z)\,dz=2\pi i\,a_{-1}=2\pi i\,\operatorname{Res}(f,z_0)\;}$$ **WHY.** This is *why* $a_{-1}$ was singled out and named. Physically: it is the strength of the one flow pattern, $\tfrac{1}{z-z_0}$, that circulates *exactly once per loop* and therefore never cancels. **PICTURE.** Step 5's figure shows the tower of terms, all struck out except the glowing $n=-1$ survivor. ![[deepdives/dd-maths-4.10.05-d2-s06.png]] > [!mistake] "But which power is the swirling one?" > It is tempting to think the biggest blow-up, $(z-z_0)^{-m}$, dominates. **Wrong:** even a strong term like $(z-z_0)^{-2}$ *spins* around ($k=-1\ne 0$) and fully cancels. Only $(z-z_0)^{-1}$ has $k=0$. Strength of blow-up is irrelevant; **only the once-around swirl survives.** --- ## Step 6 — Many singularities: shrink the loop onto each one **WHAT.** Suppose $C$ encloses several bad points $z_1,\dots,z_k$. By [[Cauchy's Integral Theorem]], we may deform the big loop — without changing the integral, as long as we don't cross a singularity — into a bunch of tiny circles, one hugging each singularity, joined by thin corridors. The corridors are traced **twice in opposite directions**, so their contributions cancel. What remains is one small loop per singularity, each giving its own $2\pi i\operatorname{Res}$: $$\oint_C f(z)\,dz=2\pi i\sum_{j=1}^{k}\operatorname{Res}(f,z_j)$$ - The $\sum_j$ adds up the swirl strengths of **only the singularities inside $C$**. - A singularity *outside* $C$ contributes nothing — the loop never circles it. **WHY.** The plane between the loops is singularity-free, so integrals there vanish; only the swirls trapped inside register. **PICTURE.** Step 6's figure: one big loop deformed into three little loops connected by cancelling corridors. ![[deepdives/dd-maths-4.10.05-d2-s07.png]] > [!mistake] Counting a pole that lies outside the contour > A perfectly correct residue is *useless* if that singularity is not enclosed. **Fix:** first draw $C$, mark every singularity, keep only the ones inside the loop, then sum. --- ## Step 7 — The degenerate & edge cases (so nothing surprises you) Every scenario the reader could meet: > [!example] No singularity inside ($k=0$) > The sum is empty, so $\oint_C f\,dz=0$. This is just Cauchy's theorem — the loop feels calm water everywhere. **Picture:** a loop drawn where $f$ is perfectly nice returns nothing. > [!example] Removable singularity ($a_{-1}=0$ and no negative terms) > Example $\tfrac{\sin z}{z}$ at $0$: the Laurent series has *no* negative powers, so $a_{-1}=0$ and the loop integral is $0$. The point looked bad but the swirl strength is zero — a "fake" whirlpool. > [!example] Higher-order pole, e.g. $\tfrac{1}{(z-z_0)^2}$ > Here the strongest term has $n=-2$, giving $k=-1\ne 0$, so it **cancels**. If $a_{-1}=0$ as well, the whole loop integral is $0$ — a double pole can still have zero residue. The order of the pole and the residue are **independent**. > [!example] Essential singularity, e.g. $e^{1/z}$ at $0$ > Infinitely many negative powers, yet still only $n=-1$ survives. From $e^{1/z}=\sum_{n\ge 0}\tfrac{1}{n!}z^{-n}$, the $z^{-1}$ coefficient is $\tfrac{1}{1!}=1$, so $\operatorname{Res}=1$ and $\oint = 2\pi i$. **No pole formula applies** — you read $a_{-1}$ straight off the series. > [!mistake] Radius-dependence anxiety > Notice the final answer had **no $\rho$ in it** — the $\rho^{n+1}$ vanished for the surviving term ($n=-1\Rightarrow \rho^0=1$). So the loop's size does not matter, only that it encloses $z_0$. If your answer still contains $\rho$, you kept a term that should have cancelled. --- ## The one-picture summary Everything above in a single frame: the Laurent tower feeds into the loop; the loop is a filter; every spinning term drains to zero; the lone $\tfrac{1}{z-z_0}$ swirl passes through, scaled by $2\pi i$. ![[deepdives/dd-maths-4.10.05-d2-s08.png]] > [!recall]- Feynman retelling — explain the whole walkthrough to a friend > Picture a pond with a whirlpool at one spot. You want to know how strong the whirlpool is, but you can't reach the centre — it's a mess there. So you paddle a big circle around it and measure the net spin you pick up. > > Here is the trick: the messy water near the whirlpool can be split into many simple flow patterns, each swirling around the centre a different number of times per lap. When you paddle one full lap, any pattern that swirls a *whole* number of times ($1,2,-2,\dots$) pushes you equally in every direction — those pushes perfectly cancel and you gain nothing from them. Only **one** special pattern, the one that circulates exactly once per lap ($\tfrac{1}{z-z_0}$), never cancels. Its strength is the number we call the **residue**. > > So one lap around returns exactly $2\pi i$ times that residue — the size of your circle doesn't matter, only that you went around the whirlpool once. And if there are several whirlpools, you just add up all their strengths. That is the entire Residue Theorem. > [!mnemonic] > **"Only the once-around swirl survives."** Everything else spins to zero; $a_{-1}$ is the survivor; a lap pays you $2\pi i\times$ it. --- ## Connections - [[Laurent series]] — supplied the tower of terms we filtered. - [[Cauchy's Integral Theorem]] — the reason spinning terms and empty regions give zero (Steps 4 & 6). - [[Cauchy's Integral Formula]] — the special case $\oint \tfrac{f(z)}{z-z_0}dz = 2\pi i\,f(z_0)$. - [[Contour integration]] — where this machine is put to work on real integrals. - [[Argument principle]] / [[Rouché's theorem]] — count zeros and poles using these residues. - [[Singularities]] — the classification behind Step 7's cases.