Visual walkthrough — Laurent series — principal part, annulus of convergence

Before we start, one promise: every symbol gets a plain-word meaning and a picture the first time it appears. Nothing is assumed.

Step 1 — What is $z$, and what does "$\oint$ around $a$" even mean?

WHAT. A complex number $z$ is just a point on a flat map (a plane). Its horizontal position is the "real part," its vertical position is the "imaginary part." We fix a special centre point and call it $a$. The little arrow from $a$ to $z$ is written $z-a$.

WHY. Everything in this whole theory is about how a function behaves as you circle around the point $a$. So the very first thing we need is a way to walk in a loop around $a$. The symbol ${\oint }_{\gamma }$ means "add up (integrate) as you travel once, counter-clockwise, all the way around a closed path $\gamma$ (a loop)."

PICTURE. Look at the loop $\gamma$ in the figure. The centre dot is $a$. The moving dot is $z$. The green arrow $z-a$ spins around as $z$ walks the loop.

Step 2 — The one integral that powers everything

WHAT. We compute a single, deceptively simple loop integral for every whole-number power $k$: ${\oint }_{\gamma }(z-a{)}^{k}dz.$ Here $k$ is any integer ($\dots ,-2,-1,0,1,2,\dots$), and $(z-a{)}^k$ is the green arrow raised to the $k$-th power.

WHY. This is the sieve we'll use later. If we can show this integral is zero for almost every $k$ and nonzero for exactly one special $k$, then integrating a whole series will kill every term except one — which is precisely how we'll isolate a single coefficient.

PICTURE. Watch what $(z-a{)}^k$ traces as $z$ goes once around. Because $z-a=\rho e^{i\theta }$, we get $(z-a{)}^k={\rho }^k{e}^{ik\theta }$: an arrow of fixed length ${\rho }^k$ that spins around $(k)$ times as $\theta$ goes once around. The figure shows the traced circle for a few values of $k$.

Substitute $z=a+\rho e^{i\theta }$, so $dz=i\rho e^{i\theta }d\theta$ (a tiny step along the loop): ${\oint }_{\gamma }(z-a{)}^{k}dz={\int }_0^{2\pi }\underset{(z-a{)}^k}{\underbrace{{\rho }^k{e}^{ik\theta }}}\cdot \underset{dz}{\underbrace{i\rho e^{i\theta }}}d\theta =i{\rho }^{k+1}{\int }_0^{2\pi }{e}^{i(k+1)\theta }d\theta .$

• ${\rho }^k{e}^{ik\theta }$ — the power itself: length ${\rho }^k$, angle $k\theta$.
• $i\rho e^{i\theta }d\theta$ — the direction and length of the tiny push $dz$ along the loop.
• ${\rho }^{k+1}$ — the two radii ${\rho }^k$ and $\rho$ combined; a harmless constant.
• $e^{i(k+1)\theta }$ — the only part that still spins; its spinning is what decides everything.

Step 3 — Why the sieve is zero (almost always)

WHAT. Evaluate ${\int }_0^{2\pi }{e}^{i(k+1)\theta }d\theta$.

WHY. This one number decides the whole outcome, so we look at it carefully and split into two cases.

PICTURE. The quantity $e^{i(k+1)\theta }$ is a unit arrow that makes $(k+1)$ complete turns as $\theta$ goes $0\to 2\pi$. Adding up an arrow that spins full circles means adding up arrows pointing in every direction equally — they perfectly cancel, giving zero. But if the arrow never turns (points the same way the whole time), all the pieces stack up and give a big nonzero total. The figure contrasts these two: cancelling petals vs. a solid stack.

Step 4 — Assume the series exists, then multiply by a "shifter"

WHAT. Suppose (the parent note's Annulus of convergence guarantees this) that on the ring $r<|z-a|<R$ our function equals a two-sided series $f(z)={\sum }_{m=-\infty }^{\infty }{c}_m(z-a{)}^m.$ We want just one number, say $c_n$. Multiply both sides by $(z-a{)}^{-n-1}$: $\frac{f(z)}{(z-a{)}^{n+1}}={\sum }_{m=-\infty }^{\infty }{c}_m(z-a{)}^{m-n-1}.$

WHY. We are deliberately shifting every power. The term we care about (the $m=n$ term) becomes $(z-a{)}^{n-n-1}=(z-a{)}^{-1}$ — exactly the special power the sieve keeps! Every other term becomes some other power, which the sieve will destroy. This is the plan made precise.

PICTURE. Think of a row of numbered dials, one per term $m$. Multiplying by $(z-a{)}^{-n-1}$ slides the whole row so that dial $m=n$ lands on the magic "$-1$" slot. The figure shows the slide.

• $c_m$ — the unknown coefficient of the $m$-th power (what we're hunting).
• $(z-a{)}^{m-n-1}$ — the shifted power; its exponent is $m-n-1$.
• The special case $m=n$ gives exponent $-1$.

Step 5 — Integrate: the sieve keeps exactly one term

WHAT. Integrate both sides around a loop $\gamma$ inside the annulus (it may pass through the sum term by term because the series converges there): ${\oint }_{\gamma }\frac{f(z)}{(z-a{)}^{n+1}}dz={\sum }_{m=-\infty }^{\infty }{c}_m{\oint }_{\gamma }(z-a{)}^{m-n-1}dz.$

WHY. Now apply the Step 3 sieve to each inner integral. Every one is zero unless the exponent $m-n-1=-1$, i.e. unless $m=n$. So the entire infinite sum collapses to a single surviving term.

PICTURE. All the dials go dark except the one sitting on the "$-1$" slot, which lights up with value $2\pi i$. Everything else contributes nothing. The figure shows the survivor.

${\oint }_{\gamma }\frac{f(z)}{(z-a{)}^{n+1}}dz=c_n\cdot \underset{=2\pi i}{\underbrace{{\oint }_{\gamma }(z-a{)}^{-1}dz}}=c_n\cdot 2\pi i.$

Divide by $2\pi i$:

Step 6 — Free bonus: uniqueness, and the residue

WHAT. The formula gives each $c_n$ directly from $f$ — no choices were made. So the coefficients are forced.

WHY it matters. This proves the parent note's claim that the Laurent series is unique per annulus: if two series both equal $f$ on the ring, the formula pins their coefficients to the same numbers. (Different annuli $\Rightarrow$ different loop $\gamma$ available $\Rightarrow$ possibly different coefficients — no contradiction.)

PICTURE. Set $n=-1$. Then $(z-a{)}^{n+1}=(z-a{)}^0=1$, and the formula becomes the plain loop integral of $f$ itself: $c_{-1}=\frac{1}{2\pi i}{\oint }_{\gamma }f(z)dz={\mathrm{Res}}_{z=a}f.$ That single coefficient is the residue — the only thing that survives when you integrate $f$ around a singularity — which is exactly why the Residue theorem and the Cauchy integral formula both fall out of this one calculation. The figure spotlights the $n=-1$ row.

• $n=-1$ — the choice that removes the denominator.
• ${\oint }_{\gamma }fdz$ — literally the loop integral of the function.
• Result $=2\pi i\cdot c_{-1}$ — one loop measures one number: the residue.

Step 7 — Degenerate & edge cases (never leave the reader stranded)

WHAT & WHY. We test the formula at its boundaries so no scenario surprises you.

Case A — $f$ is analytic (no singularity). Then $f$ is smooth inside, the annulus can swell into a full disk ($r\to 0$), and for every negative $n$ the loop can be shrunk to a point with nothing singular inside, so ${\oint }_{\gamma }\frac{f}{(z-a{)}^{n+1}}dz=0$. All negative coefficients vanish: the Laurent series degenerates into a Taylor series. Principal part $=0$ $\Rightarrow$ removable.

Case B — a pole of order $m$. Multiply $f$ by $(z-a{)}^m$ to get an analytic function; its lowest surviving negative coefficient is $c_{-m}\ne 0$ and there is nothing below it. The principal part is a finite stack of the "scream" powers $\frac{1}{z-a},\dots ,\frac{1}{(z-a{)}^m}$.

Case C — essential singularity. The negative half never stops; infinitely many $c_{-n}\ne 0$. The formula still holds for each $n$ — it just never produces a last term.

Case D — choice of loop $\gamma$. Which loop? Any loop inside the annulus that circles $a$ once counter-clockwise gives the same answer, because between two such loops $f$ is analytic and the integral doesn't change (deformation). If the loop circled $a$ twice, you'd get $2\times$; if clockwise, a minus sign. The figure shows all three.

The one-picture summary

The whole derivation in a single sweep: multiply $f$ by the shifter $(z-a{)}^{-n-1}$ so that the term you want lands on the magic "$-1$" slot → integrate around the loop → the sieve zeroes every other term → one survivor emerges, scaled by $2\pi i$ → divide it out to read off $c_n$.

Active recall