Chapter 3 · Series and Residues

3.1 Taylor and Laurent Series; Singularities

An analytic function is locally the sum of its Taylor series; around an isolated singularity it is the sum of a Laurent series in positive and negative powers of $z-a$ — and the negative powers decide whether the singularity is removable, a pole or essential.

In §2.3 we saw that a function analytic on a disc is the sum of its Taylor series there. This section asks two natural follow-up questions. First, how far does the Taylor series of $f$ about $a$ converge? The answer is geometric and surprisingly clean: up to the nearest point where $f$ genuinely breaks down. It explains a puzzle from real analysis — why the Maclaurin series of the perfectly smooth function $1/(1+x^2)$ stops converging at $|x|=1$ — and it lets us read off radii of convergence without computing a single coefficient.

Second, what happens near a point where $f$ is not analytic? If $f$ is analytic on an annulus — in particular on a punctured disc $0\lt|z-a|\lt r$ — it can be expanded in a Laurent series $\sum_{n=-\infty}^{\infty}c_n(z-a)^n$, a power series that is allowed negative powers. The negative-power part, the principal part, classifies isolated singularities into three kinds: removable ones (no negative powers: $\frac{\sin z}{z}$), poles (finitely many: $\frac1{z^3}$) and essential singularities (infinitely many: $e^{1/z}$), each with its own striking behaviour — culminating in the Casorati–Weierstrass and Picard theorems. The coefficient $c_{-1}$ will be the residue of §3.2. In CSIR NET this section supplies a steady stream of Part B questions ("the radius of convergence of the Taylor series of … about $z=\dots$ is", "the point $z=0$ is a … of …", "the coefficient of $z^{-3}$ in the Laurent series valid in $1\lt|z|\lt2$ is") and Part C questions on what an isolated singularity can or cannot look like.

Notation. As in Chapter 2, $D(a,r)=\{|z-a|\lt r\}$, $\overline{D}(a,r)$ is its closure, $C(a,\rho)$ is the positively oriented circle $|z-a|=\rho$, $\mathbb{D}=D(0,1)$, and $D'(a,r)=\{0\lt|z-a|\lt r\}$ is the punctured disc.

Taylor's Theorem and the Radius of Convergence

Theorem 2.25 already gave the expansion on a disc. We now record it in its final form, with an integral formula for the coefficients, an explicit error bound and uniqueness — the three things one actually uses.

📐 Theorem 3.1 — Taylor's theorem
Let $f$ be analytic on an open set $U$, let $a\in U$, and let $R=\sup\{r\gt0: D(a,r)\subseteq U\}$ (the distance from $a$ to $\mathbb{C}\setminus U$; $R=\infty$ if $U=\mathbb{C}$).
  1. Expansion. For every $z\in D(a,R)$, $$f(z)=\sum_{n=0}^{\infty}c_n(z-a)^n,$$ $$c_n=\frac{f^{(n)}(a)}{n!}=\frac1{2\pi i}\oint_{C(a,\rho)}\frac{f(\zeta)}{(\zeta-a)^{n+1}}\,d\zeta\quad(0\lt\rho\lt R).$$ In particular the radius of convergence of the Taylor series is at least $R$.
  2. Remainder. If $|z-a|\lt\rho\lt R$ and $N\ge0$, then $$f(z)-\sum_{n=0}^{N-1}c_n(z-a)^n=\frac{(z-a)^N}{2\pi i}\oint_{C(a,\rho)}\frac{f(\zeta)\,d\zeta}{(\zeta-a)^N(\zeta-z)},$$ and its modulus is at most $\dfrac{M(\rho)\,q^N}{1-q}$, where $q=\dfrac{|z-a|}{\rho}$ and $M(\rho)=\max_{|\zeta-a|=\rho}|f(\zeta)|$.
  3. Uniqueness. If $f(z)=\sum_{n\ge0}b_n(z-a)^n$ on some disc $D(a,r)$, $r\gt0$, then $b_n=c_n$ for every $n$.

Proof. (1) The expansion is Theorem 2.25 applied on $D(a,R)\subseteq U$, and the integral formula for $f^{(n)}(a)$ is Cauchy's integral formula for derivatives (Theorem 2.18) on the circle $C(a,\rho)$.

(2) Put $w=\frac{z-a}{\zeta-a}$ for $\zeta$ on the circle; then $|w|=q\lt1$. The finite geometric sum $\frac1{1-w}=\sum_{n=0}^{N-1}w^n+\frac{w^N}{1-w}$ gives, after dividing by $\zeta-a$, $$\frac1{\zeta-z}=\sum_{n=0}^{N-1}\frac{(z-a)^n}{(\zeta-a)^{n+1}}+\frac{(z-a)^N}{(\zeta-a)^N(\zeta-z)}.$$ Multiply by $\frac{f(\zeta)}{2\pi i}$ and integrate over $C(a,\rho)$. The left side gives $f(z)$ (Cauchy's integral formula) and the $n$-th term of the sum gives $c_n(z-a)^n$, which is the identity. For the bound, $|\zeta-z|\ge|\zeta-a|-|z-a|=\rho(1-q)$ on the circle, so the ML-inequality gives at most $\frac{1}{2\pi}\cdot2\pi\rho\cdot\frac{|z-a|^NM(\rho)}{\rho^N\cdot\rho(1-q)}=\frac{M(\rho)q^N}{1-q}$.

(3) The sum $g$ of $\sum b_n(z-a)^n$ may be differentiated term by term (Theorem 1.23), so $g^{(n)}(a)=n!\,b_n$. Since $g=f$ on $D(a,r)$, $b_n=\frac{f^{(n)}(a)}{n!}=c_n$. ∎

Uniqueness is what makes series computations legitimate: any method that produces a convergent power series for $f$ near $a$ — substitution, multiplication, division, partial fractions — produces the Taylor series.

How large is the radius exactly? Theorem 3.1 says "at least the distance to the boundary of $U$". The next theorem says the radius is exactly the distance to the first point where $f$ really cannot be continued.

📐 Theorem 3.2 — The radius is the distance to the nearest singularity
Let $f$ be analytic near $a$ and let $\rho\in(0,\infty]$ be the radius of convergence of its Taylor series at $a$.
  1. $\rho=\sup\{r\gt0:$ there is $g$ analytic on $D(a,r)$ with $g=f$ near $a\}$: the series converges on the largest disc about $a$ to which $f$ extends analytically.
  2. Suppose $f$ is analytic on $D(a,R)$, $0\lt R\lt\infty$, and there is a point $b$ with $|b-a|=R$ such that $f$ is unbounded on $D(a,R)\cap D(b,\delta)$ for every $\delta\gt0$ — for example $|f(z)|\to\infty$ as $z\to b$, as happens at a pole. Then $\rho=R$ exactly.

Proof. (1) If $g$ is analytic on $D(a,r)$ and $g=f$ near $a$, then $g$ and $f$ have the same Taylor coefficients at $a$ (they are derivatives at $a$), and by Theorem 3.1 applied to $g$ the series converges on $D(a,r)$; so $\rho\ge r$. Conversely, the sum of the series is analytic on $D(a,\rho)$ (Theorem 1.23) and equals $f$ near $a$; so the supremum is at least $\rho$.

(2) By Theorem 3.1, $\rho\ge R$. Suppose $\rho\gt R$ and let $S$ be the sum of the series on $D(a,\rho)$. $S$ is continuous, hence bounded on the compact disc $\overline D(b,\delta)$ for any $\delta\lt\rho-R$ (this disc lies inside $D(a,\rho)$). But $S=f$ on $D(a,R)$ (Theorem 3.1), so $f$ would be bounded on $D(a,R)\cap D(b,\delta)$ — a contradiction. ∎

$\frac{1}{1+z^2}$ at $0$Poles $\pm i$ at distance $1$: radius $1$. So $\sum(-1)^nx^{2n}$ diverges for $|x|\ge1$, although $\frac1{1+x^2}$ is smooth on all of $\mathbb{R}$ — the obstruction is invisible on the real line.
$\tan z$ at $0$Nearest zeros of $\cos z$: $\pm\frac{\pi}{2}$, where $|\tan z|\to\infty$. Radius $\frac\pi2$.
$\frac{z}{e^z-1}$ at $0$The zero of $e^z-1$ at $0$ is cancelled (the singularity is removable); the next zeros $\pm2\pi i$ give radius $2\pi$, not $0$.
$\operatorname{Log}z$ at $-1+i$The branch cut is at distance $1$, but the radius is $|{-1+i}|=\sqrt2$, the distance to the branch point $0$ (Example 2(d)).
🧭 Computing Taylor coefficients in practice
  1. Known series and substitution: $e^w$, $\sin w$, $\cos w$, $\frac1{1-w}$, $\operatorname{Log}(1+w)$ with $w=z^2$, $w=-z$, $w=\frac{z-a}{b-a}$, ….
  2. Partial fractions + geometric series for rational functions: $\dfrac{1}{z-b}=-\dfrac1b\sum_{n\ge0}\dfrac{z^n}{b^n}$ for $|z|\lt|b|$.
  3. Products by the Cauchy product $\left(\sum a_nz^n\right)\left(\sum b_nz^n\right)=\sum\left(\sum_{k=0}^na_kb_{n-k}\right)z^n$, valid inside both discs.
  4. Quotients $g/h$ with $h(a)\ne0$: write $g=h\cdot\sum c_n(z-a)^n$ and solve for $c_0,c_1,\dots$ in turn ("undetermined coefficients"); use parity to skip zero coefficients.
  5. Differentiate or integrate a known series term by term (e.g. $\frac{1}{(1-z)^2}=\sum(n+1)z^n$).
  6. Radius without coefficients: the distance from $a$ to the nearest point where $f$ cannot be continued analytically (Theorem 3.2) — ignore removable singularities, and measure to branch points, not cuts.
✏️ Example 1 — Taylor series and their radii
(a) Find the Taylor series of $f(z)=\dfrac{1}{z^2-3z+2}$ about $0$ and its radius of convergence.
(b) Find the radius of convergence of the Taylor series of $g(z)=\dfrac{1}{z^2+2z+2}$ about $z=1$.
(c) Find the Taylor series of $\tan z$ about $0$ up to $z^5$, and its radius.
(d) Show that $\dfrac{e^z}{1-z}=\sum_{n\ge0}s_nz^n$ with $s_n=\sum_{k=0}^{n}\frac1{k!}$, and decide where on $|z|=1$ the series converges.
(a) Partial fractions $z^2-3z+2=(z-1)(z-2)$ and $f=\frac{1}{z-2}-\frac1{z-1}$. For $|z|\lt1$: $-\frac{1}{z-1}=\sum z^n$ and $\frac1{z-2}=-\frac12\cdot\frac{1}{1-z/2}=-\sum\frac{z^n}{2^{n+1}}$. Hence $$f(z)=\sum_{n=0}^\infty\left(1-\frac{1}{2^{n+1}}\right)z^n,$$ radius $1$ (the pole $z=1$). Check: $c_0=\frac12=f(0)$.
(b) Distance The zeros of $z^2+2z+2$ are $-1\pm i$. Their distance from $1$ is $|2\mp i|=\sqrt5$, and $|g|\to\infty$ there, so the radius is $\sqrt5$ (Theorem 3.2).
(c) Undetermined coefficients $\tan z$ is odd, so $\tan z=a_1z+a_3z^3+a_5z^5+\cdots$, and $\sin z=\tan z\cdot\cos z$: $$\left(a_1z+a_3z^3+a_5z^5\right)\left(1-\tfrac{z^2}2+\tfrac{z^4}{24}\right)=z-\tfrac{z^3}{6}+\tfrac{z^5}{120}+\cdots$$ Comparing: $a_1=1$; $a_3-\frac{a_1}2=-\frac16$, so $a_3=\frac13$; $a_5-\frac{a_3}2+\frac{a_1}{24}=\frac1{120}$, so $a_5=\frac1{120}+\frac16-\frac1{24}=\frac{2}{15}$. Thus $\tan z=z+\frac{z^3}{3}+\frac{2z^5}{15}+\cdots$, radius $\frac\pi2$.
(d) Cauchy product For $|z|\lt1$, $e^z\cdot\frac{1}{1-z}=\left(\sum\frac{z^k}{k!}\right)\left(\sum z^j\right)=\sum_n\left(\sum_{k\le n}\frac1{k!}\right)z^n$. The radius is $1$ (pole at $1$). On $|z|=1$ the terms satisfy $|s_nz^n|=s_n\to e\ne0$, so the series diverges at every point of the circle — even at $z=-1$, where $f$ itself is analytic.
✏️ Example 2 — Prove or disprove: statements about radii
(a) If $f$ is analytic on $\mathbb{C}\setminus\{1,-2\}$, then its Taylor series at $0$ has radius of convergence exactly $1$.
(b) The Maclaurin series of $\frac1{1+x^2}$ converges for every real $x$.
(c) The Taylor series of $\frac1{1+x^2}$ about $x_0=1$ converges for $|x-1|\lt\sqrt2$ and diverges for $|x-1|\gt\sqrt2$.
(d) The Taylor series of the principal logarithm $\operatorname{Log}z$ about $a=-1+i$ has radius $1$, the distance from $a$ to the branch cut, and its sum is $\operatorname{Log}z$ on its whole disc of convergence.
(a) False The hypothesis allows $f(z)=e^z$ (radius $\infty$) or $f(z)=\frac{1}{z+2}$ (radius $2$). What is true: the radius is at least $1$ (Theorem 3.1), and it is exactly $1$ if $f$ is unbounded near $1$ (Theorem 3.2).
(b) False As a function of $z$, $\frac1{1+z^2}$ has poles at $\pm i$, at distance $1$ from $0$; the radius is $1$. At $x=1$ the series $\sum(-1)^n$ diverges.
(c) True $\frac{1}{1+z^2}$ is analytic on $D(1,\sqrt2)$ and unbounded near $\pm i$, which are at distance $|1\mp i|=\sqrt2$ from $1$. By Theorem 3.2 the radius is exactly $\sqrt2$; a power series converges inside and diverges outside its radius.
(d) False, twice The disc $\Delta=D(a,\sqrt2)$ does not contain $0$ and is simply connected, so $\frac1z$ has a primitive $L$ on $\Delta$ with $L(a)=\operatorname{Log}a$ (Theorem 2.15); $L$ is a branch of $\log z$ on $\Delta$, so the radius is at least $\sqrt2$. Since $L'(z)=\frac1z$ is unbounded near $0\in\partial\Delta$, the derived series has radius exactly $\sqrt2$ (Theorem 3.2), and so does the series of $L$. The disc $\Delta$ meets the negative real axis in the segment $(-2,0)$. Above it $L=\operatorname{Log}z$ (both are branches, equal near $a$); below it $L=\operatorname{Log}z+2\pi i$, because $L$ is continuous across the segment while $\operatorname{Log}$ jumps by $-2\pi i$. The series "does not see" the branch cut — only the branch point.

Laurent Series in an Annulus

A function such as $\frac1{(z-1)(z-2)}$ has no Taylor series valid in $1\lt|z|\lt2$, but it can still be expanded there — if negative powers are allowed. The natural domains for such expansions are annuli.

📘 Definition 3.1 — Annuli and Laurent series

For $a\in\mathbb{C}$ and $0\le r\lt R\le\infty$ the annulus is $A(a;r,R)=\{z:r\lt|z-a|\lt R\}$; for $r=0$ it is the punctured disc $D'(a,R)$, and for $R=\infty$ the exterior of a circle.

A Laurent series about $a$ is a doubly infinite series $\sum_{n=-\infty}^{\infty}c_n(z-a)^n$. It converges at $z\ne a$ if both of its halves converge: the regular part $\sum_{n\ge0}c_n(z-a)^n$ and the principal part $\sum_{n\ge1}c_{-n}(z-a)^{-n}$; its sum is the sum of the two.

The regular part is a power series in $z-a$: it converges for $|z-a|\lt R_2$, where $\frac1{R_2}=\limsup|c_n|^{1/n}$. The principal part is a power series in $\frac{1}{z-a}$: it converges for $|z-a|\gt r_2$, where $r_2=\limsup|c_{-n}|^{1/n}$ (Cauchy–Hadamard). So a Laurent series converges on the annulus $r_2\lt|z-a|\lt R_2$ (if $r_2\lt R_2$), diverges where $|z-a|\lt r_2$ or $|z-a|\gt R_2$, and its sum is analytic on the annulus.

📐 Theorem 3.3 — Laurent's theorem
Let $f$ be analytic on the annulus $A=A(a;r,R)$, $0\le r\lt R\le\infty$. Then for every $z\in A$ $$f(z)=\sum_{n=-\infty}^{\infty}c_n(z-a)^n,$$ $$c_n=\frac1{2\pi i}\oint_{C(a,\rho)}\frac{f(\zeta)}{(\zeta-a)^{n+1}}\,d\zeta\qquad(n\in\mathbb{Z}),$$ where $\rho$ is any radius with $r\lt\rho\lt R$ (the integral does not depend on $\rho$). The series converges absolutely on $A$ and uniformly on every closed sub-annulus $\rho_1\le|z-a|\le\rho_2$ with $r\lt\rho_1\lt\rho_2\lt R$. The expansion is unique: if $f(z)=\sum b_n(z-a)^n$ for all $z\in A$, then $b_n=c_n$ for all $n$.

Proof. Independence of $\rho$. For $r\lt\rho\lt\rho'\lt R$, the function $\zeta\mapsto f(\zeta)(\zeta-a)^{-n-1}$ is analytic on a neighbourhood of the closed region between $C(a,\rho)$ and $C(a,\rho')$, so the two integrals agree by the deformation theorem (Theorem 2.16(1)).

A Cauchy formula for the annulus. Fix $z\in A$ and choose $\rho_1,\rho_2$ with $r\lt\rho_1\lt|z-a|\lt\rho_2\lt R$; let $C_1=C(a,\rho_1)$, $C_2=C(a,\rho_2)$ and let $\gamma$ be a circle about $z$ so small that it lies between $C_1$ and $C_2$. The function $\zeta\mapsto\frac{f(\zeta)}{\zeta-z}$ is analytic on $A\setminus\{z\}$, so Theorem 2.16(2) (with $\gamma_0=C_2$ and the two inner contours $C_1$, $\gamma$) gives $\oint_{C_2}=\oint_{C_1}+\oint_\gamma$. By Cauchy's integral formula on a small disc about $z$ inside $A$, $\oint_\gamma\frac{f(\zeta)}{\zeta-z}d\zeta=2\pi if(z)$. Therefore $$f(z)=\frac1{2\pi i}\oint_{C_2}\frac{f(\zeta)\,d\zeta}{\zeta-z}-\frac1{2\pi i}\oint_{C_1}\frac{f(\zeta)\,d\zeta}{\zeta-z}.$$

Expanding the two integrals. On $C_2$, $|z-a|\lt|\zeta-a|$, and as in Theorem 3.1 $$\frac{1}{\zeta-z}=\sum_{n\ge0}\frac{(z-a)^n}{(\zeta-a)^{n+1}},$$ uniformly in $\zeta\in C_2$ (M-test with ratio $|z-a|/\rho_2\lt1$). Integrating term by term, the first integral is $\sum_{n\ge0}c_n(z-a)^n$. On $C_1$, $|\zeta-a|\lt|z-a|$, so now we expand in the other direction: $$-\frac{1}{\zeta-z}=\frac{1}{(z-a)-(\zeta-a)}=\sum_{n\ge1}\frac{(\zeta-a)^{n-1}}{(z-a)^n},$$ uniformly on $C_1$ (ratio $\rho_1/|z-a|\lt1$). Integrating term by term, the second term becomes $$\sum_{n\ge1}(z-a)^{-n}\cdot\frac1{2\pi i}\oint_{C_1}f(\zeta)(\zeta-a)^{n-1}d\zeta=\sum_{n\ge1}c_{-n}(z-a)^{-n},$$ because $(\zeta-a)^{n-1}=(\zeta-a)^{-(-n)-1}$. Adding gives the expansion.

Convergence. The ML-inequality on $C(a,\rho)$ gives $|c_n|\le M(\rho)\rho^{-n}$ for every $\rho\in(r,R)$ and every $n\in\mathbb{Z}$, where $M(\rho)=\max_{C(a,\rho)}|f|$. On $\rho_1\le|z-a|\le\rho_2$ pick $\rho_1'\in(r,\rho_1)$ and $\rho_2'\in(\rho_2,R)$; then $|c_n(z-a)^n|\le M(\rho_2')(\rho_2/\rho_2')^n$ for $n\ge0$ and $|c_{-n}(z-a)^{-n}|\le M(\rho_1')(\rho_1'/\rho_1)^n$ for $n\ge1$ — convergent geometric majorants, so the M-test gives absolute and uniform convergence.

Uniqueness. If $\sum b_n(z-a)^n$ converges to $f$ on $A$, its two halves converge on $A$, hence (by the previous paragraph's argument applied to power series in $z-a$ and in $\frac1{z-a}$) uniformly on the circle $C(a,\rho)$. Multiply by $(z-a)^{-k-1}$ and integrate term by term: since $\oint_{C(a,\rho)}(z-a)^m\,dz$ is $2\pi i$ for $m=-1$ and $0$ for every other integer $m$ (§2.2, Example 1), only $n=k$ survives and $\frac{1}{2\pi i}\oint\frac{f(z)}{(z-a)^{k+1}}dz=b_k$. So $b_k=c_k$. ∎

Three remarks. (i) If $f$ is analytic on the whole disc $D(a,R)$, then $c_{-n}=\frac1{2\pi i}\oint f(\zeta)(\zeta-a)^{n-1}d\zeta=0$ for $n\ge1$ by Cauchy's theorem, and Laurent's theorem reduces to Taylor's. (ii) The coefficient $c_{-1}=\frac1{2\pi i}\oint_{C(a,\rho)}f$ is the only one that the integral of $f$ itself sees; it will be the residue of §3.2. (iii) A Laurent series belongs to an annulus, not to a point: the same $f$ has different Laurent series about the same $a$ in different annuli (Example 3). Nobody computes $c_n$ from the integral formula; uniqueness lets us use geometric series instead.

🧭 Finding a Laurent series
  1. Identify the annulus: centre $a$, and the singular points that lie inside the inner circle and outside the outer one.
  2. Split $f$ (partial fractions) into pieces, each with a single singular point $b$.
  3. Expand each piece by the geometric series in the variable that is small on the annulus: $$\frac{1}{z-b}=-\sum_{n\ge0}\frac{(z-a)^n}{(b-a)^{n+1}}\ \ (|z-a|\lt|b-a|),$$ $$\frac{1}{z-b}=\sum_{n\ge1}\frac{(b-a)^{n-1}}{(z-a)^n}\ \ (|z-a|\gt|b-a|).$$
  4. For $e^{g}$, $\sin g$, $\cos g$ with $g=\frac{1}{z-a}$, substitute into the Taylor series (valid for all $z\ne a$), then multiply by the remaining powers of $z-a$.
  5. Check one coefficient, e.g. the leading behaviour of $f$ as $z\to a$ or as $z\to\infty$.
✏️ Example 3 — One function, four Laurent series
Expand $f(z)=\dfrac{1}{(z-1)(z-2)}$ in Laurent series valid in (a) $|z|\lt1$; (b) $1\lt|z|\lt2$; (c) $|z|\gt2$; (d) $0\lt|z-1|\lt1$.
Split $f=\dfrac{1}{z-2}-\dfrac{1}{z-1}$; the singular points are $1$ and $2$.
(a) $|z|\lt1$ Both points outside: this is the Taylor series of Example 1(a), $f=\sum_{n\ge0}\left(1-2^{-n-1}\right)z^n$.
(b) $1\lt|z|\lt2$ Now $|z|\gt1$, so $-\frac1{z-1}=-\frac1z\cdot\frac1{1-1/z}=-\sum_{n\ge1}z^{-n}$, while $|z|\lt2$ still gives $\frac1{z-2}=-\sum_{n\ge0}\frac{z^n}{2^{n+1}}$: $$f(z)=-\sum_{n=1}^{\infty}\frac{1}{z^{n}}-\sum_{n=0}^{\infty}\frac{z^n}{2^{n+1}}.$$
(c) $|z|\gt2$ Both points inside: $\frac1{z-2}=\sum_{n\ge1}\frac{2^{n-1}}{z^n}$, so $$f(z)=\sum_{n\ge1}\frac{2^{n-1}-1}{z^n}=\frac1{z^2}+\frac3{z^3}+\frac7{z^4}+\cdots$$ Check: $f(z)\sim\frac{1}{z^2}$ as $z\to\infty$.
(d) About $1$ Put $w=z-1$: $f=\frac{1}{w(w-1)}=-\frac1w\cdot\frac{1}{1-w}=-\sum_{n\ge0}w^{n-1}$, i.e. $$f(z)=-\frac{1}{z-1}-1-(z-1)-(z-1)^2-\cdots$$ valid for $0\lt|z-1|\lt1$ (the next singular point, $2$, is at distance $1$). The coefficient of $\frac1{z-1}$ is $-1$.
Lesson In (b) there are infinitely many negative powers, yet $0$ is a perfectly regular point of $f$. Negative powers in an annulus $r\lt|z|\lt R$ with $r\gt0$ record the singular points inside the inner circle (here $z=1$), not the behaviour at the centre.
✏️ Example 4 — Laurent series by substitution
Find the Laurent series about $0$, valid for $0\lt|z|\lt\infty$, of (a) $z^3e^{1/z}$; (b) $\dfrac{\sin z}{z^4}$; and (c) show that $e^{z+1/z}=\sum_{n\in\mathbb{Z}}c_nz^n$ with $$c_n=c_{-n}=\sum_{k=0}^{\infty}\frac{1}{k!\,(n+k)!}\quad(n\ge0).$$ Deduce the value of $\int_0^{2\pi}e^{2\cos\theta}\cos\theta\,d\theta$ as a series.
(a) $e^{1/z}=\sum_{k\ge0}\frac{1}{k!\,z^k}$ for $z\ne0$, so $$z^3e^{1/z}=z^3+z^2+\frac z2+\frac16+\frac1{24z}+\frac{1}{120z^2}+\cdots$$ — infinitely many negative powers; $c_{-1}=\frac1{4!}=\frac1{24}$.
(b) $\frac{\sin z}{z^4}=\frac{1}{z^4}\left(z-\frac{z^3}{6}+\frac{z^5}{120}-\cdots\right)=\frac{1}{z^3}-\frac{1}{6z}+\frac{z}{120}-\cdots$ — the principal part has just two terms; $c_{-1}=-\frac16$.
(c) Product For $z\ne0$ both $e^z=\sum_j\frac{z^j}{j!}$ and $e^{1/z}=\sum_k\frac{z^{-k}}{k!}$ converge absolutely, so their product may be rearranged: the coefficient of $z^n$ collects the pairs with $j-k=n$, i.e. $j=n+k$: $c_n=\sum_{k\ge\max(0,-n)}\frac{1}{(n+k)!\,k!}$. The substitution $z\mapsto\frac1z$ leaves $e^{z+1/z}$ unchanged, so by uniqueness $c_{-n}=c_n$.
(c) Integral By Theorem 3.3 with $\rho=1$, $z=e^{i\theta}$, $dz=iz\,d\theta$: $$c_n=\frac{1}{2\pi}\int_0^{2\pi}e^{2\cos\theta}e^{-in\theta}\,d\theta.$$ The imaginary part $-\frac1{2\pi}\int e^{2\cos\theta}\sin n\theta\,d\theta$ vanishes (the integrand is odd about $\theta=\pi$). With $n=1$: $\int_0^{2\pi}e^{2\cos\theta}\cos\theta\,d\theta=2\pi\sum_{k\ge0}\frac{1}{k!(k+1)!}\approx2\pi\times1.590637$. (These coefficients are the modified Bessel numbers $I_n(2)$.)
Interactive 1 — Laurent Series Live in Annuli

Choose a function and an annulus about $0$ bounded by its singular points (×). The truncated Laurent series $S_N(z)=\sum_{n=-N}^{N}c_nz^n$ of that annulus is compared with $f$ on the circle $|z|=\rho$. Right: $\log_{10}$ of the largest error on the circle against $N$. Press Play to add terms. Inside the shaded annulus the error falls geometrically; move $\rho$ outside it and the same series diverges.

annulus—
circle |z| = ρ—
max |f − S_N| on circle—
c₋₁—
c₀—
c₁—

Isolated Singularities: Removable Singularities, Poles and Essential Singularities

📘 Definition 3.2 — Isolated singularities and their types

A point $a$ is an isolated singularity of $f$ if $f$ is analytic on some punctured disc $D'(a,r)$, $r\gt0$ (whether or not $f$ is defined at $a$). Let $f(z)=\sum_{n\in\mathbb{Z}}c_n(z-a)^n$ be its Laurent series on $D'(a,r)$ (Theorem 3.3), with principal part $\sum_{n\ge1}c_{-n}(z-a)^{-n}$. The singularity is

  • removable if $c_{-n}=0$ for all $n\ge1$ (no principal part);
  • a pole of order $m$ ($m\ge1$) if $c_{-m}\ne0$ and $c_{-n}=0$ for all $n\gt m$ (a principal part with finitely many terms); a pole of order $1$ is a simple pole, of order $2$ a double pole;
  • an essential singularity if $c_{-n}\ne0$ for infinitely many $n\ge1$.

Not every singular point is isolated: $0$ is a non-isolated singularity of $\frac{1}{\sin(1/z)}$ (the poles $\frac1{k\pi}$ accumulate at $0$), and $0$ is a branch point of $\operatorname{Log}z$ and $\sqrt z$ (no punctured disc about $0$ carries an analytic branch). Definition 3.2 says nothing about such points.

The Laurent coefficients are seldom the quickest way to decide the type. The next three theorems characterise each type by the behaviour of $f$ near $a$.

📐 Theorem 3.4 — Riemann's removable singularity theorem
Let $f$ be analytic on $D'(a,r)$. The following are equivalent:
  1. $a$ is a removable singularity;
  2. $f$ extends to an analytic function on the whole disc $D(a,r)$;
  3. $\lim_{z\to a}f(z)$ exists in $\mathbb{C}$;
  4. $f$ is bounded on $D'(a,\delta)$ for some $\delta\gt0$;
  5. $\lim_{z\to a}(z-a)f(z)=0$.

Proof. (1)⇒(2): the Laurent series is then a power series $\sum_{n\ge0}c_n(z-a)^n$ that converges at every $z$ with $0\lt|z-a|\lt r$; hence its radius of convergence is at least $r$ and its sum $g$ is analytic on $D(a,r)$ (Theorem 1.23). Since $g=f$ on $D'(a,r)$, setting $f(a)=c_0$ gives the extension. (2)⇒(3)⇒(4) because analytic functions are continuous. (4)⇒(5): $|(z-a)f(z)|\le M|z-a|\to0$.

(5)⇒(1): let $\varepsilon(\rho)=\max_{|\zeta-a|=\rho}|(\zeta-a)f(\zeta)|$, so $\varepsilon(\rho)\to0$ as $\rho\to0$. For $n\ge1$ and $0\lt\rho\lt\min(r,1)$, the ML-inequality gives $$|c_{-n}|=\left|\frac{1}{2\pi i}\oint_{C(a,\rho)}f(\zeta)(\zeta-a)^{n-1}d\zeta\right|\le\frac{2\pi\rho}{2\pi}\cdot\varepsilon(\rho)\rho^{n-2}=\varepsilon(\rho)\rho^{n-1}\le\varepsilon(\rho).$$ Letting $\rho\to0$ gives $c_{-n}=0$. ∎

So $\frac{\sin z}{z}$, $\frac{e^z-1}{z}$, $\frac{z}{e^z-1}$ and $\frac{z^2-1}{z-1}$ all have removable singularities at the obvious points: each is bounded nearby. Condition (5) is remarkably weak — $|f(z)|\le\frac{C}{|z-a|^{0.99}}$ already forces removability.

📐 Theorem 3.5 — Characterisation of poles
Let $f$ be analytic on $D'(a,r)$ and $m\ge1$. The following are equivalent:
  1. $a$ is a pole of order $m$;
  2. $f(z)=\dfrac{g(z)}{(z-a)^m}$ on $D'(a,r)$ for some $g$ analytic on $D(a,r)$ with $g(a)\ne0$;
  3. $\lim_{z\to a}(z-a)^mf(z)$ exists and is nonzero;
  4. $f\ne0$ near $a$, and $\frac1f$ has a removable singularity at $a$ which, once removed, is a zero of order $m$.
Moreover, $a$ is a pole (of some order) if and only if $|f(z)|\to\infty$ as $z\to a$.

Proof. (1)⇒(2): $g(z)=(z-a)^mf(z)=\sum_{n\ge0}c_{n-m}(z-a)^n$ is a power series converging on $D'(a,r)$, hence analytic on $D(a,r)$ as in Theorem 3.4, with $g(a)=c_{-m}\ne0$. (2)⇒(1): expand $g=\sum b_k(z-a)^k$ and divide by $(z-a)^m$; by uniqueness this is the Laurent series, and its lowest term is $b_0(z-a)^{-m}$ with $b_0=g(a)\ne0$. (2)⇒(3) is clear. (3)⇒(2): $(z-a)^mf(z)$ has a limit, so its singularity is removable (Theorem 3.4); call the extension $g$; $g(a)$ is the nonzero limit. (2)⇒(4): $g\ne0$ on some $D(a,\delta)$, and there $\frac1f=\frac{(z-a)^m}{g}$ with $\frac1g$ analytic and nonzero: a zero of order exactly $m$ (Theorem 2.26). (4)⇒(2): $\frac1f=(z-a)^mh$ near $a$ with $h$ analytic and $h(a)\ne0$; so $f=\frac{1/h}{(z-a)^m}$ near $a$, and $g=(z-a)^mf$ is analytic on $D'(a,r)$ and equals $\frac1h$ near $a$, hence analytic on $D(a,r)$ with $g(a)\ne0$.

Finally, at a pole $|f|=\frac{|g|}{|z-a|^m}\to\infty$. Conversely, if $|f(z)|\to\infty$ then $f\ne0$ near $a$ and $\frac1f\to0$; by Theorem 3.4, $\frac1f$ extends analytically with value $0$ at $a$, and it is not identically zero, so $a$ is a zero of some finite order $m$ (Definition 2.10). By (4), $a$ is a pole of order $m$. ∎

Order arithmetic. Write $\operatorname{ord}_a f=m$ if $f$ has a zero of order $m$ at $a$, $\operatorname{ord}_af=-m$ at a pole of order $m$, and $0$ if $f$ is analytic and nonzero at $a$. Theorems 2.26 and 3.5 show that $f=(z-a)^{k}h$ with $h(a)\ne0$ and $k=\operatorname{ord}_af$, so $$\operatorname{ord}_a(fg)=\operatorname{ord}_af+\operatorname{ord}_ag,\qquad\operatorname{ord}_a\frac fg=\operatorname{ord}_af-\operatorname{ord}_ag.$$ For instance $\operatorname{ord}_0\frac{\sin z}{z^4}=1-4=-3$: a pole of order $3$, not $4$.

🧭 Classifying an isolated singularity
  1. Is it isolated? Poles or zeros of a denominator that accumulate at $a$ (as for $\frac1{\sin(1/z)}$ at $0$), and branch points, are not isolated singularities.
  2. Quotients $g/h$ of functions analytic at $a$: compute $\operatorname{ord}_ag-\operatorname{ord}_ah$ from the first nonzero derivative (or the first term of the Taylor series): $\ge0$ removable, $=-m$ pole of order $m$.
  3. A function of $\frac{1}{z-a}$, such as $e^{1/z}$ or $\cos\frac1z$ (an entire non-polynomial function of $\frac1{z-a}$): substitute — infinitely many negative powers, essential.
  4. Limits: bounded near $a$ ⇒ removable; $|f|\to\infty$ ⇒ pole, order $m$ = the $m$ with $(z-a)^mf\to$ a finite nonzero limit; two paths giving different limits of $|f|$ in $[0,\infty]$ ⇒ essential.
✏️ Example 5 — Classify the singularities
Find and classify all singularities in $\mathbb{C}$ of: (a) $\dfrac{1-\cos z}{z^5}$; (b) $\dfrac{z}{e^z-1}$; (c) $\dfrac{1}{z(e^z-1)}$; (d) $z\cos\dfrac1z$; (e) $\dfrac{1}{\sin(\pi/z)}$; (f) $\dfrac{z^2-\pi^2}{\sin^2z}$.
(a) Only $z=0$. $1-\cos z=\frac{z^2}{2}-\frac{z^4}{24}+\cdots$ has a zero of order $2$; order $2-5=-3$: a pole of order $3$. Indeed $\frac{1-\cos z}{z^5}=\frac{1}{2z^3}-\frac{1}{24z}+\frac{z}{720}-\cdots$.
(b) $e^z=1$ exactly at $z=2\pi ik$, $k\in\mathbb{Z}$, and these zeros are simple ($(e^z)'=e^z\ne0$). At $0$ the numerator also has a simple zero: order $1-1=0$, removable (value $\lim\frac{z}{e^z-1}=1$). At $2\pi ik$, $k\ne0$, the numerator is nonzero: simple poles.
(c) At $0$: order $-1-1=-2$, a double pole ($\frac{1}{z(e^z-1)}=\frac1{z^2}-\frac1{2z}+\frac1{12}-\cdots$). At $2\pi ik$, $k\ne0$: simple poles.
(d) Only $0$: $z\cos\frac1z=z-\frac{1}{2z}+\frac{1}{24z^3}-\cdots$ has infinitely many negative powers: essential.
(e) $\sin\frac\pi z=0\iff\frac\pi z=k\pi\iff z=\frac1k$, $k\in\mathbb{Z}\setminus\{0\}$. The derivative $-\frac{\pi}{z^2}\cos\frac\pi z$ is nonzero there, so each $\frac1k$ is a simple pole. The point $0$ is the limit of the poles $\frac1k$: a non-isolated singularity — no punctured disc about $0$ is free of poles.
(f) $\sin^2z$ has double zeros at $z=k\pi$. At $z=\pm\pi$ the numerator has a simple zero: order $1-2=-1$, simple poles. At $z=k\pi$ with $k\ne\pm1$ (including $0$, where the numerator is $-\pi^2\ne0$): double poles.

Essential singularities behave in a way that has no analogue for real functions: near such a point, $f$ comes arbitrarily close to every complex value.

📐 Theorem 3.6 — Casorati–Weierstrass theorem
Let $a$ be an essential singularity of $f$, analytic on $D'(a,r)$. Then for every $\delta\in(0,r]$ the image $f(D'(a,\delta))$ is dense in $\mathbb{C}$: for every $w\in\mathbb{C}$ there are points $z_k\to a$ with $f(z_k)\to w$. Consequently, an isolated singularity $a$ is essential if and only if $\lim_{z\to a}|f(z)|$ does not exist in $[0,\infty]$.

Proof. Suppose that for some $\delta$ the image is not dense: there are $w\in\mathbb{C}$ and $\varepsilon\gt0$ with $|f(z)-w|\ge\varepsilon$ for all $z\in D'(a,\delta)$. Then $g=\frac{1}{f-w}$ is analytic on $D'(a,\delta)$ and bounded by $\frac1\varepsilon$, so $a$ is a removable singularity of $g$ (Theorem 3.4); extend $g$ to $D(a,\delta)$. Since $g$ has no zeros on $D'(a,\delta)$, $g\not\equiv0$. If $g(a)\ne0$, then $f=w+\frac1g$ is analytic near $a$: removable. If $g(a)=0$, it is a zero of some order $k$, and by Theorem 3.5(4) $f-w=\frac1g$ has a pole of order $k$, hence so does $f$ (adding a constant does not change the principal part). Either way $a$ is not essential — a contradiction.

For the last statement: at a removable singularity $|f|\to|f(a)|$, at a pole $|f|\to\infty$. At an essential singularity, density gives points $z_k\to a$ with $f(z_k)\to0$ and points $z_k'\to a$ with $f(z_k')\to1$, so $\lim|f|$ does not exist. ∎

📐 Theorem 3.7 — Picard's theorems (statements)
  1. Great Picard theorem. If $a$ is an essential singularity of $f$, then in every punctured disc $D'(a,\delta)$ the function $f$ takes every complex value, with at most one exception, infinitely many times.
  2. Little Picard theorem. A nonconstant entire function takes every complex value with at most one exception.

The proofs need tools beyond this course (the elliptic modular function, or Bloch's and Schottky's theorems). Both are sharp: $e^{1/z}$ omits only $0$ near $0$, and $e^z$ omits only $0$ on $\mathbb{C}$. For Part C the key consequences are: an entire function whose image misses two distinct values is constant, and Casorati–Weierstrass gives density, while Picard gives "onto, with at most one exception".

✏️ Example 6 — The essential singularity of $e^{1/z}$
Show that $0$ is an essential singularity of $f(z)=e^{1/z}$; that $f(x)\to\infty$ as $x\to0^+$ and $f(x)\to0$ as $x\to0^-$ along the real axis, while $|f|=1$ on the imaginary axis; and that every $w\ne0$ is taken infinitely often in every $D'(0,\delta)$, while $0$ is never taken. Find explicitly the points where $e^{1/z}=i$.
Type $e^{1/z}=\sum_{k\ge0}\frac{1}{k!\,z^k}$ has infinitely many negative powers: essential.
Three directions For real $x\to0^+$, $\frac1x\to+\infty$ and $e^{1/x}\to\infty$; for $x\to0^-$, $e^{1/x}\to0$. For $z=iy$, $y\ne0$, $e^{1/(iy)}=e^{-i/y}$ has modulus $1$. Three different limits of $|f|$: no limit, as Theorem 3.6 predicts.
Every value but one Since $e^\zeta=w$ has the solutions $\zeta=\operatorname{Log}w+2\pi ik$ ($k\in\mathbb{Z}$) for $w\ne0$ and none for $w=0$: $$e^{1/z}=w\iff z_k=\frac{1}{\operatorname{Log}w+2\pi ik},\quad k\in\mathbb{Z}.$$ As $|k|\to\infty$, $|z_k|\to0$, so infinitely many $z_k$ lie in any $D'(0,\delta)$ — the great Picard theorem with exactly one exceptional value, $0$.
$w=i$ $\operatorname{Log}i=\frac{i\pi}2$, so $z_k=\frac{1}{i\pi(2k+\frac12)}=\frac{-2i}{(4k+1)\pi}$: points on the imaginary axis accumulating at $0$. Check: $\frac1{z_k}=\frac{i\pi}{2}+2\pi ik$ and $e^{i\pi/2}=i$.
Interactive 2 — Zoom In on a Singularity

Left: a circle $|z|=r$ around the singular point $0$. Right: its image under $f$, drawn in a compressed $w$-plane (distance $|w|$ squeezed so that the rim is $\infty$; grey rings at $|w|=1,10,100,10^4$). Press Play to shrink the circle. Watch three behaviours: the image collapses to a point (removable), runs off to $\infty$ while winding $m$ times clockwise round $0$ (pole of order $m$), or sweeps out ever larger regions (essential).

radius r—
min |f| on circle—
max |f| on circle—
winding of f(circle) about 0—
type—

The Point at Infinity; Entire and Meromorphic Functions

The extended plane $\mathbb{C}_\infty$ of §1.1 lets us treat $\infty$ like any other point: a neighbourhood of $\infty$ is a set $\{|z|\gt R\}$, and $z\mapsto\frac1z$ carries it to a punctured neighbourhood of $0$.

📘 Definition 3.3 — Behaviour at $\infty$

Let $f$ be analytic on $\{|z|\gt R\}$; we say $\infty$ is an isolated singularity of $f$. Put $F(w)=f\!\left(\frac1w\right)$ for $0\lt|w|\lt\frac1R$. Then $\infty$ is a removable singularity, a pole of order $m$ or an essential singularity of $f$ according as $0$ is one for $F$; similarly $f$ has a zero of order $m$ at $\infty$ if $F$ (with its removable singularity removed) has a zero of order $m$ at $0$.

In terms of the Laurent series $f(z)=\sum_{n\in\mathbb{Z}}c_nz^n$ on $|z|\gt R$ we have $F(w)=\sum c_nw^{-n}$, so the roles of positive and negative powers are exchanged: $\infty$ is removable iff $c_n=0$ for all $n\ge1$; a pole of order $m$ iff $c_m\ne0$ and $c_n=0$ for $n\gt m$; essential iff $c_n\ne0$ for infinitely many $n\ge1$.

Examples: $z^3$ has a pole of order $3$ at $\infty$; $\frac1{z^2}$ has a zero of order $2$ there; $e^{1/z}$ is analytic at $\infty$ with value $1$; $\frac{z^3}{z^2+1}=z-\frac1z+\frac1{z^3}-\cdots$ ($|z|\gt1$) has a simple pole at $\infty$; $e^z$, $\sin z$, $\cos z$ have essential singularities at $\infty$; and for $\tan z$ the point $\infty$ is not an isolated singularity at all (the poles $\frac\pi2+k\pi$ tend to $\infty$).

📐 Theorem 3.8 — Entire functions at $\infty$
Let $f$ be entire. Then:
  1. $\infty$ is a removable singularity of $f$ if and only if $f$ is constant;
  2. $\infty$ is a pole of order $m$ if and only if $f$ is a polynomial of degree $m$;
  3. otherwise — when $f$ is a transcendental entire function, i.e. not a polynomial — $\infty$ is an essential singularity.
In particular, an entire function with $|f(z)|\to\infty$ as $z\to\infty$ is a nonconstant polynomial.

Proof. By Theorem 3.1 with $U=\mathbb{C}$, $f(z)=\sum_{n\ge0}c_nz^n$ for all $z$; by uniqueness this is also its Laurent series on $|z|\gt0$, the punctured neighbourhood of $\infty$. Read Definition 3.3: (1) no positive powers means $f=c_0$; (2) $c_m\ne0$ and $c_n=0$ for $n\gt m$ means $f$ is a polynomial of degree $m$; (3) is what remains. If $|f(z)|\to\infty$, then $|F(w)|\to\infty$ as $w\to0$, so $0$ is a pole of $F$ (Theorem 3.5) and case (2) applies. ∎

Applying Casorati–Weierstrass at $\infty$: for a transcendental entire function $f$ and every $R$, the set $f(\{|z|\gt R\})$ is dense in $\mathbb{C}$. Together with the open mapping theorem this pins down the injective entire functions.

📐 Theorem 3.9 — Injective entire functions are linear
An entire function $f$ is injective on $\mathbb{C}$ if and only if $f(z)=\alpha z+\beta$ with $\alpha\ne0$.

Proof. Linear maps with $\alpha\ne0$ are injective. Conversely, let $f$ be entire and injective; it is not constant. If $f$ were transcendental, $f(\mathbb{D})$ would be a nonempty open set (open mapping theorem, Theorem 2.35) and $f(\{|z|\gt1\})$ would be dense, so the two sets would meet: $f(z_1)=f(z_2)$ with $|z_1|\lt1\lt|z_2|$, contradicting injectivity. So $f$ is a polynomial of degree $n\ge1$. If $n\ge2$, $f'$ has a root $z_0$ (fundamental theorem of algebra), so $f-f(z_0)$ has a zero of order at least $2$ at $z_0$ and, by the local mapping theorem (Theorem 2.36), $f$ is at least $2$-to-$1$ near $z_0$ — not injective. Hence $n=1$. ∎

📘 Definition 3.4 — Meromorphic functions
A function $f$ is meromorphic on an open set $U$ if there is a set $P\subseteq U$ with no limit point in $U$ such that $f$ is analytic on $U\setminus P$ and has a pole at every point of $P$. It is meromorphic on the extended plane $\mathbb{C}_\infty$ if it is meromorphic on $\mathbb{C}$ and $\infty$ is a removable singularity or a pole.

Rational functions are meromorphic on $\mathbb{C}_\infty$; $\tan z$, $\sec z$, $\frac1{\sin z}$ and $\frac{z}{e^z-1}$ are meromorphic on $\mathbb{C}$ but not on $\mathbb{C}_\infty$; $e^{1/z}$ is not meromorphic on $\mathbb{C}$ (essential singularity at $0$); $\frac1{\sin(1/z)}$ is meromorphic on $\mathbb{C}\setminus\{0\}$ (its poles $\frac1k$ have no limit point in that set) but not on $\mathbb{C}$. If $g,h$ are analytic on a domain $D$ and $h\not\equiv0$, then $g/h$ is meromorphic on $D$, because the zeros of $h$ have no limit point in $D$ (identity theorem). (The converse — every meromorphic function on $\mathbb{C}$ is a quotient of two entire functions — is a theorem of Weierstrass.)

📐 Theorem 3.10 — Meromorphic functions on $\mathbb{C}_\infty$ are rational
If $f$ is meromorphic on $\mathbb{C}$ and $\infty$ is a removable singularity or a pole of $f$, then $f$ is a rational function. More precisely, $f$ has finitely many poles $a_1,\dots,a_k$ in $\mathbb{C}$, and $$f=p+P_1+\cdots+P_k,$$ where $P_j$ is the principal part of $f$ at $a_j$ (a polynomial in $\frac1{z-a_j}$ without constant term) and $p$ is a polynomial — the partial fraction decomposition of $f$.

Proof. Since $\infty$ is an isolated singularity, $f$ is analytic on $\{|z|\gt R\}$ for some $R$, so all poles lie in the compact disc $\overline D(0,R)$. The poles have no limit point in $\mathbb{C}$, so only finitely many lie in this disc (Bolzano–Weierstrass): $a_1,\dots,a_k$. Each principal part $P_j$ is analytic on $\mathbb{C}\setminus\{a_j\}$ and tends to $0$ as $z\to\infty$. The function $h=f-\sum_jP_j$ is analytic on $\mathbb{C}$ minus the $a_j$, and at $a_j$ the difference $f-P_j$ has no principal part while the other $P_i$ are analytic, so every $a_j$ is a removable singularity of $h$: after removal $h$ is entire. As $z\to\infty$, $h$ differs from $f$ by a function tending to $0$, so $\infty$ is removable or a pole of $h$; by Theorem 3.8, $h$ is a polynomial $p$. ∎

✏️ Example 7 — Prove or disprove: singularities, entire and meromorphic functions
(a) If $f$ and $g$ have poles of order $m$ and $n$ at $a$, then $f+g$ has a pole of order $\max(m,n)$ at $a$.
(b) If $f$ has a pole at $a$, then $e^{f}$ has an essential singularity at $a$.
(c) If $f$ is analytic on $D'(0,1)$ and $|f(z)|\le|z|^{-1/2}$ there, then $0$ is a removable singularity.
(d) An entire function $f$ with $f(z)\to\infty$ as $z\to\infty$ is a polynomial.
(e) A meromorphic function on $\mathbb{C}$ with only finitely many poles is rational.
(a) False True when $m\ne n$ (the term $(z-a)^{-\max(m,n)}$ cannot cancel), false when $m=n$: $f=\frac1z$, $g=1-\frac1z$ give $f+g=1$, a removable singularity. With $f=\frac1{z^2}+\frac1z$, $g=-\frac1{z^2}$ the order drops to $1$.
(b) True Write $f(z)=\frac{g(z)}{(z-a)^m}$ with $g(a)=\beta\ne0$. Along the ray $z=a+te^{i\varphi}$, $t\to0^+$: $\operatorname{Re}f=t^{-m}\operatorname{Re}\!\left(g(a+te^{i\varphi})e^{-im\varphi}\right)$. Choose $\varphi$ with $\beta e^{-im\varphi}=|\beta|$; then the real part of $g(\cdot)e^{-im\varphi}$ tends to $|\beta|\gt0$, so $\operatorname{Re}f\to+\infty$ and $|e^f|=e^{\operatorname{Re}f}\to\infty$. Choosing $\varphi$ with $\beta e^{-im\varphi}=-|\beta|$ gives $|e^f|\to0$. So $\lim|e^f|$ does not exist, and $a$ is essential (Theorem 3.6). Example: $e^{1/z}$.
(c) True $|zf(z)|\le|z|^{1/2}\to0$, which is condition (5) of Theorem 3.4.
(d) True This is the last part of Theorem 3.8: $\infty$ is then a pole, so $f$ is a (nonconstant) polynomial.
(e) False $e^z$ has no poles at all, and $\frac{e^z}{z-1}$ has one; neither is rational. Theorem 3.10 needs the extra hypothesis that $\infty$ is at worst a pole (here it is essential).
⚠️ Counterexamples and pitfalls
  1. A Laurent series belongs to an annulus. $\frac1{(z-1)(z-2)}$ has three different Laurent series about $0$ ($|z|\lt1$, $1\lt|z|\lt2$, $|z|\gt2$). "The coefficient of $z^{-3}$" means nothing until the annulus is named.
  2. Negative powers in an annulus are not a singularity at the centre. In $1\lt|z|\lt2$ the series above has infinitely many negative powers, but $0$ is a regular point. Types of singularities are read only from the series on a punctured disc $0\lt|z-a|\lt r$.
  3. Not every singular point is isolated. $0$ for $\frac1{\sin(1/z)}$ and $\tan\frac1z$ (limits of poles), and the branch points of $\operatorname{Log}z$, $\sqrt z$, $z^{1/3}$: neither removable, pole nor essential.
  4. Removable is not "undefined". $\frac{\sin z}{z}$, $\frac{e^z-1}{z}$, $\frac{z}{e^z-1}$ at $0$ and $\frac{z-\pi}{\sin z}$ at $\pi$ are analytic once the value is filled in; they play no role in radii of convergence or residues.
  5. Orders of quotients need the numerator too. $\frac{\sin z}{z^4}$ has a pole of order $3$; $\frac{1-\cos z}{z^5}$ of order $3$; $\frac{z}{\sin^2z}$ a simple pole at $0$.
  6. $|f|\to\infty$ along one path is not a pole. $e^{1/x}\to\infty$ as $x\to0^+$, but $e^{1/z}$ has an essential singularity; at a pole $|f|\to\infty$ along every path.
  7. Dense is not onto. Casorati–Weierstrass gives a dense image; $e^{1/z}$ never takes the value $0$. Picard allows exactly one omitted value, not two.
  8. Cancellation. Two poles of the same order can add up to a removable singularity ($\frac1z+(1-\frac1z)$); two essential singularities can multiply to a removable one ($e^{1/z}\cdot e^{-1/z}=1$) or add to one ($e^{1/z}+(z-e^{1/z})$).
  9. Radius = distance to the nearest genuine singularity. Ignore removable ones ($\frac{z}{e^z-1}$ at $0$: radius $2\pi$), and measure to branch points, not branch cuts ($\operatorname{Log}z$ at $-1+i$: radius $\sqrt2$, not $1$).
  10. Smooth on $\mathbb{R}$ is not radius $\infty$. $\frac{1}{1+x^2}$ has Maclaurin radius $1$ because of $\pm i$; $e^{-1/x^2}$ (extended by $0$) is $C^\infty$ on $\mathbb{R}$ with Maclaurin series $0$, because $e^{-1/z^2}$ has an essential singularity at $0$.
  11. At $\infty$ the roles of the powers swap. $e^{1/z}$ is removable at $\infty$, $z$ has a simple pole there, $\frac{1}{z^2}$ a double zero, $\sin z$ an essential singularity, and for $\tan z$ the point $\infty$ is not isolated.
  12. Finitely many poles is not enough for rationality. $e^z$, $\frac{e^z}{z-1}$ are meromorphic on $\mathbb{C}$ with finitely many poles; Theorem 3.10 also needs $\infty$ to be at worst a pole.

Exercises

I
Taylor series and radii of convergence.
  • 1 Expand $\frac{1}{z+2}$ in powers of $z-1$ and state the radius of convergence.
  • 2 Find the Taylor series of $\dfrac{z+1}{z^2-4z+3}$ about $0$, with its general coefficient and radius.
  • 3 Find the first three nonzero terms of the Taylor series of $\sec z$ about $0$ and its radius of convergence.
  • 4 Without computing coefficients, find the radius of convergence of the Taylor series of (a) $\frac{z}{\sin z}$ about $0$; (b) $\frac1{z^2+4}$ about $1$; (c) $\cot z-\frac1z$ about $0$; (d) $\operatorname{Log}z$ about $2i$.
  • 5 Find the Taylor series of $\dfrac{1}{1+z+z^2}$ about $0$ (hint: multiply numerator and denominator by $1-z$), and its radius.
  • 6 Define $B_n$ by $\dfrac{z}{e^z-1}=\sum_{n\ge0}\dfrac{B_n}{n!}z^n$. Find $B_0,\dots,B_4$, show that $B_{2k+1}=0$ for $k\ge1$ (hint: $\frac{z}{e^z-1}+\frac z2$ is even), and give the radius.
  • 7 Let $f=\sum c_nz^n$ be entire. Use Theorem 3.1(2) with $\rho=2|z|$ to show $\left|f(z)-\sum_{n\lt N}c_nz^n\right|\le\dfrac{2M(2|z|)}{2^N}$. Evaluate the bound for $e^z$ at $z=1$, $N=10$, and compare with the true error.
  • 8 Let $f$ be analytic on $D(0,2)$ except for a simple pole at $z=1$, and $f=\sum c_nz^n$ on $\mathbb{D}$. Prove that $c_n\to-b$, where $b=\lim_{z\to1}(z-1)f(z)$. Check this against Example 1(d).
1 $\frac1{z+2}=\frac{1}{3+(z-1)}=\frac13\cdot\frac{1}{1+\frac{z-1}{3}}=\sum_{n\ge0}\frac{(-1)^n(z-1)^n}{3^{n+1}}$, for $|z-1|\lt3$; radius $3$, the distance from $1$ to the pole $-2$.
2 $z^2-4z+3=(z-1)(z-3)$ and $\frac{z+1}{(z-1)(z-3)}=\frac{-1}{z-1}+\frac{2}{z-3}=\frac1{1-z}-\frac23\cdot\frac{1}{1-z/3}$. So $c_n=1-\frac{2}{3^{n+1}}$ ($c_0=\frac13=f(0)$), radius $1$.
3 $\sec z$ is even: $\sec z=1+a_2z^2+a_4z^4+\cdots$ with $(1+a_2z^2+a_4z^4)(1-\frac{z^2}2+\frac{z^4}{24})=1$. Then $a_2=\frac12$ and $a_4-\frac{a_2}{2}+\frac1{24}=0$, $a_4=\frac5{24}$: $\sec z=1+\frac{z^2}2+\frac{5z^4}{24}+\cdots$. Radius $\frac\pi2$ (zeros of $\cos z$).
4 (a) $\pi$: the singularity at $0$ is removable; the nearest poles are $\pm\pi$. (b) $\sqrt5$: the poles $\pm2i$ are at distance $|1\mp2i|=\sqrt5$. (c) $\pi$: $\cot z-\frac1z$ is removable at $0$ (it equals $-\frac z3-\cdots$), with poles at $\pm\pi$. (d) $2$: the distance from $2i$ to the branch point $0$ (as in Example 2(d)).
5 $\frac{1}{1+z+z^2}=\frac{1-z}{1-z^3}=(1-z)\sum_k z^{3k}=\sum_k\left(z^{3k}-z^{3k+1}\right)$: coefficients $1,-1,0,1,-1,0,\dots$ Radius $1$ (poles at the primitive cube roots of unity $\omega,\omega^2$).
6 Solve $\left(\sum\frac{B_n}{n!}z^n\right)\left(1+\frac z2+\frac{z^2}6+\frac{z^3}{24}+\frac{z^4}{120}+\cdots\right)=1$ (dividing $e^z-1$ by $z$): $B_0=1$, $B_1=-\frac12$, $B_2=\frac16$, $B_3=0$, $B_4=-\frac1{30}$. Since $\frac{z}{e^z-1}+\frac z2=\frac z2\cdot\frac{e^z+1}{e^z-1}=\frac z2\coth\frac z2$ is even, all odd coefficients except $B_1$ vanish. Radius $2\pi$ (poles $\pm2\pi i$).
7 With $q=\frac{|z|}{\rho}=\frac12$: $\frac{M(\rho)q^N}{1-q}=\frac{2M(2|z|)}{2^N}$. For $e^z$, $z=1$: $M(2)=e^2$, so the bound is $\frac{2e^2}{1024}\approx0.0144$. The true error $e-\sum_{n\lt10}\frac1{n!}\approx3.03\times10^{-7}$: the bound is crude (it ignores the factorials), but it holds for every entire function and every $z$.
8 $f(z)=\frac{b}{z-1}+g(z)$ with $g$ analytic on $D(0,2)$ (Theorem 3.5: subtract the principal part). For $1\lt\rho\lt2$, $|g_n|\le M_g(\rho)\rho^{-n}\to0$ (Cauchy's estimates), while $\frac{b}{z-1}=-b\sum z^n$. So $c_n=-b+g_n\to-b$. For $\frac{e^z}{1-z}$: $b=\lim(z-1)\frac{e^z}{1-z}=-e$, so $c_n\to e$, matching $s_n\to e$.
II
Laurent series.
  • 1 Find the Laurent series of $\dfrac{1}{z(1-z)}$ valid in (a) $0\lt|z|\lt1$; (b) $|z|\gt1$; (c) $0\lt|z-1|\lt1$.
  • 2 Expand $\dfrac{1}{z^2+1}$ (a) for $|z|\gt1$; (b) for $0\lt|z-i|\lt2$, giving the general term.
  • 3 Find the coefficient of $\frac1z$ in the Laurent series about $0$ ($0\lt|z|\lt\infty$) of (a) $z^2\sin\frac1z$; (b) $\frac{e^z}{z^3}$; (c) $\frac{\sin z}{z^6}$.
  • 4 Find the first four nonzero terms of the Laurent series of $\dfrac{1}{z^2\sin z}$ in $0\lt|z|\lt\pi$. (You may use $\frac{1}{\sin z}=\frac1z+\frac z6+\frac{7z^3}{360}+\frac{31z^5}{15120}+\cdots$, and check it by multiplying.)
  • 5 Show that the Laurent coefficients of $\exp\!\left(\frac12\left(z-\frac1z\right)\right)$ on $0\lt|z|\lt\infty$ are $c_n=\frac1{2\pi}\int_0^{2\pi}\cos(n\theta-\sin\theta)\,d\theta$, that $c_{-n}=(-1)^nc_n$, and find $c_0$ as a series.
  • 6 Let $c_n=2^{-n}$ for $n\ge0$ and $c_n=3^{n}$ for $n\lt0$. Find the largest annulus about $0$ on which $\sum_{n\in\mathbb{Z}}c_nz^n$ converges, and its sum.
  • 7 Let $f$ be analytic on $1\lt|z|\lt3$ with $|f|\le5$ there. Show that $|c_n|\le5\cdot3^{-n}$ for $n\ge0$ and $|c_n|\le5$ for $n\lt0$. What can you conclude if instead $f$ is analytic and bounded by $5$ on $0\lt|z|\lt3$?
  • 8 Prove that if $f$ is analytic on $A(0;r,R)$ and extends analytically to $D(0,R)$, its Laurent series in the annulus has no negative powers. Show by an example that $c_{-1}=0$ alone does not imply this.
Check: a Laurent coefficient is an average over a circle
numerical c_n (mean of the curve)—
claimed c_n—
difference—
1 (a) $\frac1{z(1-z)}=\frac1z\sum_{n\ge0}z^n=\frac1z+1+z+z^2+\cdots$. (b) $\frac{1}{z(1-z)}=-\frac{1}{z^2}\cdot\frac{1}{1-1/z}=-\sum_{n\ge2}z^{-n}$. (c) With $w=z-1$: $\frac{1}{(1+w)(-w)}=-\frac1w\sum(-1)^nw^n=\sum_{n\ge-1}(-1)^nw^n=-\frac1{z-1}+1-(z-1)+(z-1)^2-\cdots$
2 (a) $\frac{1}{z^2}\cdot\frac{1}{1+z^{-2}}=\sum_{k\ge1}(-1)^{k-1}z^{-2k}=\frac1{z^2}-\frac1{z^4}+\cdots$. (b) $w=z-i$, $z+i=w+2i$: $\frac{1}{w(w+2i)}=\frac{1}{2iw}\cdot\frac{1}{1+\frac{w}{2i}}=\sum_{n\ge0}\frac{(-1)^nw^{n-1}}{(2i)^{n+1}}=\frac{-i/2}{z-i}+\frac14+\frac{i}{8}(z-i)-\cdots$, valid for $|w|\lt2=|2i|$.
3 (a) $z^2\left(\frac1z-\frac{1}{6z^3}+\cdots\right)=z-\frac{1}{6z}+\cdots$: $-\frac16$. (b) $\frac{1}{z^3}\left(1+z+\frac{z^2}2+\cdots\right)$: $\frac12$. (c) $\frac{1}{z^6}\left(z-\frac{z^3}6+\frac{z^5}{120}-\cdots\right)$: $\frac1{120}$.
4 Check: $\left(z-\frac{z^3}6+\frac{z^5}{120}-\frac{z^7}{5040}\right)\left(\frac1z+\frac z6+\frac{7z^3}{360}+\frac{31z^5}{15120}\right)=1+0\cdot z^2+0\cdot z^4+0\cdot z^6+\cdots$ (e.g. the $z^4$ coefficient is $\frac7{360}-\frac1{36}+\frac{1}{120}=0$). Dividing by $z^2$: $$\frac1{z^2\sin z}=\frac1{z^3}+\frac{1}{6z}+\frac{7z}{360}+\frac{31z^3}{15120}+\cdots$$
5 On $\rho=1$, $z=e^{i\theta}$: $\frac12(z-\frac1z)=i\sin\theta$, so $c_n=\frac{1}{2\pi}\int_0^{2\pi}e^{i(\sin\theta-n\theta)}d\theta$; the sine part is odd about $\theta=\pi$ and integrates to $0$, leaving $\frac1{2\pi}\int\cos(n\theta-\sin\theta)\,d\theta$. The map $z\mapsto-\frac1z$ leaves $z-\frac1z$ fixed, so $\sum c_n(-1)^nz^{-n}=\sum c_nz^n$ and $c_{-n}=(-1)^nc_n$. From $e^{z/2}e^{-1/(2z)}$: $c_0=\sum_{k\ge0}\frac{(-1)^k}{4^k(k!)^2}\approx0.765198$ (the Bessel value $J_0(1)$).
6 Regular part $\sum(z/2)^n$: converges iff $|z|\lt2$, sum $\frac{2}{2-z}$. Principal part $\sum_{n\ge1}3^{-n}z^{-n}=\sum(3z)^{-n}$: converges iff $|z|\gt\frac13$, sum $\frac{1}{3z-1}$. Annulus $\frac13\lt|z|\lt2$, sum $\frac{2}{2-z}+\frac1{3z-1}$.
7 $|c_n|\le M\rho^{-n}=5\rho^{-n}$ for every $\rho\in(1,3)$. For $n\ge0$ let $\rho\to3$; for $n=-k\lt0$, $|c_{-k}|\le5\rho^{k}$ and let $\rho\to1$. If $|f|\le5$ on $0\lt|z|\lt3$, then $|c_{-k}|\le5\rho^k$ for all small $\rho$, so $c_{-k}=0$: $0$ is removable — Riemann's theorem again.
8 For $n\ge1$, $c_{-n}=\frac{1}{2\pi i}\oint_{C(0,\rho)}f(\zeta)\zeta^{n-1}d\zeta=0$ by Cauchy's theorem, since the integrand is analytic on the disc $D(0,R)$. Example: $\frac1{z^2}$ on $0\lt|z|\lt\infty$ has $c_{-1}=0$ but a double pole.
III
Isolated singularities.
  • 1 Classify the singularity at $0$ of (a) $\frac{e^z-1-z}{z^2}$; (b) $\frac{\sin z}{z^3}$; (c) $\frac{\operatorname{Log}(1+z)}{z^2}$; (d) $\frac1z\sin\frac1z$; (e) $\frac{z-\sin z}{z^5}$.
  • 2 Find and classify all singularities in $\mathbb{C}$ of (a) $\frac{1}{z^3-z^2}$; (b) $\tan z$; (c) $\frac{z-1}{z^2\sin\pi z}$; (d) $\frac{e^{1/(z-1)}}{z-2}$.
  • 3 Find the order of the pole of $\dfrac{1}{\left(2\cos z-2+z^2\right)^3}$ at $0$.
  • 4 Find all $z$ with $e^{1/z}=-1$, and check that infinitely many of them lie in every punctured disc $D'(0,\delta)$.
  • 5 Prove: if $f$ is analytic on $D'(a,r)$ and $\operatorname{Re}f(z)\gt0$ there, then $a$ is a removable singularity.
  • 6 Prove: if $f$ is analytic on $D'(a,r)$ and $|f(z)|\le A|z-a|^{-3/2}$ there, then $a$ is removable or a simple pole. Give an example of each.
  • 7 Give an example, or prove that none exists: (a) $f$ with an essential singularity at $0$ such that $f^2$ has a pole at $0$; (b) $f$ with a pole at $0$ such that $f^2$ has a removable singularity; (c) $f,g$ with essential singularities at $0$ such that $fg$ is removable; (d) $f$ analytic and unbounded on $D'(0,1)$ with $zf(z)\to0$ as $z\to0$.
  • 8 Let $f$ have a pole of order $m$ at $a$. Prove that $f'$ has a pole of order $m+1$ and that $\frac{f'}{f}$ has a simple pole at $a$ with coefficient $-m$ of $\frac{1}{z-a}$.
  • 9 Show that $\cos z$ takes every complex value infinitely often (so it has no Picard exceptional value), whereas $e^z$ omits exactly one value.
1 (a) $\frac{z^2/2+z^3/6+\cdots}{z^2}$: removable, value $\frac12$. (b) $1-3=-2$: pole of order $2$. (c) $\operatorname{Log}(1+z)=z-\frac{z^2}2+\cdots$: order $1-2=-1$, simple pole. (d) $\frac1z\sin\frac1z=\frac1{z^2}-\frac{1}{6z^4}+\cdots$: essential. (e) $z-\sin z=\frac{z^3}{6}-\cdots$: order $3-5=-2$, pole of order $2$.
2 (a) $z^2(z-1)$: double pole at $0$, simple pole at $1$. (b) Simple poles at $\frac\pi2+k\pi$ ($\cos$ has simple zeros there, $\sin\ne0$). (c) $\sin\pi z$ has simple zeros at all integers. At $0$: order $0-(2+1)=-3$, pole of order $3$. At $1$: numerator order $1$, denominator order $1$ ($z^2=1\ne0$): removable. At $k\ne0,1$: simple poles. (d) Essential singularity at $1$ (an $e^{1/w}$ with $w=z-1$, times a function analytic and nonzero near $1$), simple pole at $2$.
3 $2\cos z-2+z^2=2\left(1-\frac{z^2}2+\frac{z^4}{24}-\cdots\right)-2+z^2=\frac{z^4}{12}-\cdots$ has a zero of order $4$; its cube has order $12$: a pole of order $12$.
4 $e^{1/z}=-1\iff\frac1z=(2k+1)\pi i\iff z=\frac{-i}{(2k+1)\pi}$, $k\in\mathbb{Z}$. These tend to $0$ as $|k|\to\infty$.
5 Not essential: $f(D'(a,r))$ lies in the right half-plane, which is not dense (Theorem 3.6). Not a pole: then $\frac1f$ would have a zero at $a$ and, by the open mapping theorem, $\frac1f$ would map a small disc about $a$ onto a neighbourhood of $0$, which contains points $w$ with $\operatorname{Re}w\lt0$, i.e. points where $\operatorname{Re}f=\operatorname{Re}\frac1w\lt0$. So $a$ is removable. (Alternatively: $\frac{1}{1+f}$ is bounded by $1$, hence removable, and so on.)
6 Put $g=(z-a)f$. Then $|(z-a)g(z)|=|z-a|^2|f(z)|\le A|z-a|^{1/2}\to0$, so $g$ has a removable singularity (Theorem 3.4(5)). Hence $f=\frac{g}{z-a}$ is removable (if $g(a)=0$) or a simple pole (if $g(a)\ne0$). Examples: $f=1$ and $f=\frac{1}{z-a}$.
7 (a) None: if $f^2$ has a pole, $|f|^2\to\infty$, so $|f|\to\infty$ and $f$ has a pole (Theorem 3.5). (b) None: $|f^2|=|f|^2\to\infty$ is not bounded. (c) $f=e^{1/z}$, $g=e^{-1/z}$, $fg=1$. (d) None: $zf\to0$ makes $0$ removable (Theorem 3.4), so $f$ is bounded near $0$.
8 $f=\frac{g}{(z-a)^m}$, $g(a)\ne0$: $f'=\frac{(z-a)g'-mg}{(z-a)^{m+1}}$ and the numerator is $-mg(a)\ne0$ at $a$, so $f'$ has a pole of order $m+1$. Also $\frac{f'}{f}=\frac{g'}{g}-\frac{m}{z-a}$ with $\frac{g'}{g}$ analytic near $a$. (This is the key computation of the argument principle, §3.2.)
9 $\cos z=w\iff u+u^{-1}=2w$ with $u=e^{iz}\iff u^2-2wu+1=0$. This quadratic has roots $u_1,u_2$ with $u_1u_2=1$, so both are nonzero, and each $e^{iz}=u_j$ has infinitely many solutions $z=-i\operatorname{Log}u_j+2\pi k$. So every $w$ is taken infinitely often. $e^z$ omits $0$ and takes every other value infinitely often.
IV
The point at infinity; entire and meromorphic functions.
  • 1 Classify the behaviour at $\infty$ of (a) $\frac{z^3}{z^2+1}$; (b) $\frac{z^2+1}{z^3}$; (c) $e^{1/z}$; (d) $\sin z$; (e) $z^2e^{1/z}$; (f) $\frac1{\sin z}$.
  • 2 Find all entire $f$ such that $f(1/z)$ has a pole of order $3$ at $0$, $f(0)=f(1)=f(-1)=0$ and $f(2)=6$.
  • 3 Prove that if $f$ is entire and $f(\{|z|\gt1\})$ is not dense in $\mathbb{C}$, then $f$ is a polynomial.
  • 4 Find the partial fraction decomposition (Theorem 3.10) of $\dfrac{z^4+1}{z^2(z-1)}$, naming the principal part at each pole, including $\infty$.
  • 5 Show that $\tan z$ is meromorphic on $\mathbb{C}$, but that $\infty$ is not an isolated singularity of it.
  • 6 Which of the entire functions $3z-i$, $z^3$, $z+e^z$, $e^z$ are injective on $\mathbb{C}$? Justify.
  • 7 Find the meromorphic function $f$ on $\mathbb{C}_\infty$ whose only pole is a simple pole at $z=1$, with $f(0)=1$ and $f(\infty)=3$.
1 (a) $z-\frac1z+\cdots$: simple pole. (b) $\frac1z+\frac{1}{z^3}$: removable, a simple zero at $\infty$. (c) Removable, value $1$. (d) Essential ($\sin z$ is transcendental entire). (e) $F(w)=\frac{e^w}{w^2}$: double pole at $\infty$. (f) The poles $k\pi$ tend to $\infty$: $\infty$ is not isolated.
2 By Theorem 3.8, $f$ is a polynomial of degree $3$ with roots $0,\pm1$: $f=cz(z-1)(z+1)$. Then $f(2)=6c=6$, so $f(z)=z^3-z$.
3 If $f$ were transcendental, $\infty$ would be an essential singularity and Casorati–Weierstrass, applied to $F(w)=f(1/w)$ on $D'(0,1)$, would make $f(\{|z|\gt1\})=F(D'(0,1))$ dense.
4 Division: $z^4+1=(z^3-z^2)(z+1)+(z^2+1)$. Then $\frac{z^2+1}{z^2(z-1)}=\frac{A}{z}+\frac{B}{z^2}+\frac{C}{z-1}$ with $C=\frac{1+1}{1}=2$, $B=\lim_{z\to0}\frac{z^2+1}{z-1}=-1$, $A=\frac{d}{dz}\frac{z^2+1}{z-1}\Big|_0=\frac{2z(z-1)-(z^2+1)}{(z-1)^2}\Big|_0=-1$. So $$\frac{z^4+1}{z^2(z-1)}=z+1-\frac1z-\frac1{z^2}+\frac2{z-1}.$$ Principal parts: $-\frac1z-\frac1{z^2}$ at $0$, $\frac2{z-1}$ at $1$, and $z$ at $\infty$ (a simple pole). Check: $A+C=1$, the coefficient of $\frac1z$ at $\infty$ of $\frac{z^2+1}{z^2(z-1)}\sim\frac1z$.
5 $\tan z=\frac{\sin z}{\cos z}$ is a quotient of entire functions with $\cos\not\equiv0$, so it is meromorphic, with simple poles at $\frac\pi2+k\pi$. These poles tend to $\infty$, so no set $\{|z|\gt R\}$ is free of them.
6 Only $3z-i$ (Theorem 3.9). $z^3$: $1^3=\omega^3$. $e^z$: $e^0=e^{2\pi i}$. $z+e^z$ is transcendental, hence not injective by Theorem 3.9.
7 By Theorem 3.10, $f=p+\frac{b}{z-1}$ with $p$ a polynomial; $f(\infty)=3$ finite forces $p\equiv3$. Then $f(0)=3-b=1$, $b=2$: $f(z)=3+\frac{2}{z-1}=\frac{3z-1}{z-1}$.
🎯 Exam corner CSIR Part BCSIR Part CGATE MA

E1. CSIR Part B GATE MA The radius of convergence of the Taylor series of $f(z)=\dfrac{z-\pi}{\sin z}$ about $z=3$ is (A) $\pi-3$; (B) $3$; (C) $2\pi-3$; (D) $\pi$.

Idea Radius = distance to the nearest non-removable singularity (Theorem 3.2).
Singular points $\sin z=0$ at $z=k\pi$. At $z=\pi$ the numerator also vanishes (simple zeros of both): removable. At $0$ and $2\pi$, $|f|\to\infty$. Distances from $3$: $|3-0|=3$ and $2\pi-3\approx3.28$; option (A) is the distance to the removable point.
Answer (B) $3$.

E2. CSIR Part B Let $\sum_{n\in\mathbb{Z}}c_nz^n$ be the Laurent series of $f(z)=\dfrac{1}{(z-1)(z-3)}$ valid in $1\lt|z|\lt3$. Then $c_{-3}$ equals (A) $0$; (B) $-\frac12$; (C) $4$; (D) $\frac12$.

Split $f=\frac12\left(\frac{1}{z-3}-\frac1{z-1}\right)$. In $1\lt|z|\lt3$: $\frac{1}{z-3}=-\sum_{n\ge0}\frac{z^n}{3^{n+1}}$ (no negative powers) and $-\frac1{z-1}=-\sum_{n\ge1}z^{-n}$.
Read off $c_{-n}=-\frac12$ for every $n\ge1$. (In $|z|\lt1$ one gets $0$, option (A); in $|z|\gt3$, $c_{-3}=\frac12(3^2-1)=4$, option (C).)
Answer (B) $-\frac12$.

E3. CSIR Part B For $f(z)=\dfrac{1-\cos z}{z^3\,(e^z-1)}$, the point $z=0$ is (A) a removable singularity; (B) a simple pole; (C) a pole of order $2$; (D) a pole of order $4$.

Orders $\operatorname{ord}_0(1-\cos z)=2$, $\operatorname{ord}_0z^3=3$, $\operatorname{ord}_0(e^z-1)=1$; so $\operatorname{ord}_0f=2-3-1=-2$. Indeed $f=\frac{1}{2z^2}-\frac{1}{4z}+\cdots$
Answer (C). (Option (D) forgets the zero of the numerator.)

E4. CSIR Part C Let $f$ be analytic on $D'(0,1)$. Which of the following statements are true? (One or more options may be correct.) (A) If $zf(z)$ is bounded near $0$, then $0$ is a removable singularity or a simple pole of $f$. (B) If $f\left(\frac1n\right)=0$ for all $n\ge2$ and $f\not\equiv0$, then $0$ is an essential singularity of $f$. (C) If $|f(x)|\to\infty$ as $x\to0^+$ along the real axis, then $0$ is a pole of $f$. (D) If $f^3$ has a removable singularity at $0$, so does $f$.

(A) True $g=zf$ is bounded, so removable (Theorem 3.4); then $f=\frac gz$ is removable ($g(0)=0$) or has a simple pole ($g(0)\ne0$).
(B) True Not removable: the extension would be analytic on $\mathbb{D}$ with zeros $\frac1n$ accumulating at $0\in\mathbb{D}$, so $f\equiv0$ by the identity theorem. Not a pole: $|f|\to\infty$ is contradicted by $f(\frac1n)=0$. Hence essential (e.g. $\sin\frac{\pi}{z}$).
(C) False $e^{1/z}$: $e^{1/x}\to\infty$ as $x\to0^+$, but the singularity is essential.
(D) True $|f|=|f^3|^{1/3}$ is bounded near $0$; apply Theorem 3.4.
Answer (A), (B), (D).

E5. CSIR Part C GATE MA Which of the following statements are true? (A) An entire function $f$ with $|f(z)|\to\infty$ as $|z|\to\infty$ is a polynomial. (B) An entire function whose range omits two distinct complex numbers is constant. (C) Every function meromorphic on $\mathbb{C}$ with finitely many poles is rational. (D) If $f$ is entire and nonconstant, then $f(1/z)$ has a pole at $0$.

(A) True $\infty$ is then a pole (Theorem 3.5 for $f(1/w)$), so $f$ is a polynomial (Theorem 3.8).
(B) True Little Picard theorem (Theorem 3.7). (For an elementary special case — omitting a whole disc — Liouville suffices, §2.3.)
(C) False $e^z$ is meromorphic with no poles and is not rational.
(D) False For transcendental $f$ such as $e^z$, $f(1/z)=e^{1/z}$ has an essential singularity at $0$. Only polynomials give poles.
Answer (A), (B).

E6. CSIR Part C Let $\sum_{n\in\mathbb{Z}}c_nz^n$ be the Laurent series of $f(z)=\dfrac{1}{z(z-1)(z-2)}$ in the annulus $1\lt|z|\lt2$. Which of the following are true? (A) $c_{-1}=-\frac12$. (B) $c_{-5}=-1$. (C) $c_0=\frac14$. (D) The series converges at $z=2.5$.

Split $f=\frac{1/2}{z}-\frac{1}{z-1}+\frac{1/2}{z-2}$ (residues $\frac{1}{(-1)(-2)}$, $\frac{1}{1\cdot(-1)}$, $\frac{1}{2\cdot1}$). In $1\lt|z|\lt2$: $-\frac1{z-1}=-\sum_{n\ge1}z^{-n}$ and $\frac{1/2}{z-2}=-\sum_{n\ge0}\frac{z^n}{2^{n+2}}$.
(A) True $c_{-1}=\frac12-1=-\frac12$.
(B) True $c_{-n}=-1$ for all $n\ge2$.
(C) False $c_0=-\frac14$.
(D) False The regular part $-\sum\frac{z^n}{2^{n+2}}$ diverges for $|z|\gt2$; at $z=2.5$ one needs the different Laurent series of the region $|z|\gt2$.
Answer (A), (B).
🧑‍🏫 For the teacher

Suggested plan — 3 lectures × 60 min.

  1. Taylor series and radii. Theorem 3.1 (remainder proof on the board — it is just a finite geometric sum), Theorem 3.2 with the four rule cards, the first strategy box, Examples 1–2. Spend ten minutes on Example 2(d): draw the disc $D(-1+i,\sqrt2)$ crossing the cut, and ask where the sum of the series equals $\operatorname{Log}z$.
  2. Laurent series. Definition 3.1, Theorem 3.3 with the full proof (the annulus Cauchy formula is the only new idea), the second strategy box, Examples 3–4. Project Interactive 1 with $\frac1{(z-1)(z-2)}$: show the three annuli in turn and drag $\rho$ out of the annulus to watch the error explode. End with $e^{1/z}$ and its super-geometric convergence.
  3. Singularities. Definition 3.2, Theorems 3.4–3.6 with proofs, Theorem 3.7 as statements; Interactive 2 (removable → simple pole → order 3 → $e^{1/z}$ → $\sin\frac1z$); Example 5 on the board, Example 6, then Definition 3.3 and Theorems 3.8–3.10 more briefly; Example 7 and Exam-corner E4–E5 live.

Discussion questions. (1) Why can a real function be $C^\infty$ on $\mathbb{R}$ and still have a Maclaurin series with finite radius — or radius $\infty$ but the wrong sum ($e^{-1/x^2}$)? What does complex analysis see that real analysis does not? (2) In Interactive 2, the winding number of the image circle stabilises at $-m$ for a pole of order $m$, but for $\sin\frac1z$ it grows without bound. Explain this using the zeros $\frac1{k\pi}$. (3) Is there a function analytic on $D'(0,1)$ with $|f(z)|=e^{1/|z|}$? (Hint: what would $\frac{1}{f}$ be?)

Misconceptions to watch for. Reading singularity types from a Laurent series in an annulus that is not a punctured disc; counting orders of quotients without the numerator; believing $|f|\to\infty$ along one ray means a pole; confusing "dense image" with "onto"; measuring radii to branch cuts or to removable singularities; forgetting that positive powers are the principal part at $\infty$.

Suggested homework. I(2, 4, 6, 8), II(1, 2, 4, 6, 7), III(1, 2, 5, 7, 8), IV(1, 3, 4, 6); Exam corner E1, E3, E6.

🔬 Research–Teaching NexusResearch-tutored · optional, about 60 minutes

Research question. Can you "see" the nearest singularity of a generating function from its Taylor coefficients alone — and predict how fast those coefficients grow?

Why it matters. In the analysis of algorithms and in combinatorics (counting tree structures in compilers, data-base indices, random permutations in card shuffling or secret-gift draws), the numbers of interest are Taylor coefficients of a generating function, and their growth is governed by the singularities closest to $0$. This "singularity analysis" (developed by Flajolet, Odlyzko and Sedgewick) turns Theorem 3.2 into a quantitative tool used daily in computer science.

What to do.

  1. For $f(z)=\frac{z}{e^z-1}=\sum c_nz^n$ (coefficients $c_n=\frac{B_n}{n!}$, computable exactly with Python's fractions or sympy.bernoulli), compute $\sqrt{|c_{2k}/c_{2k+2}|}$ for $k=1,\dots,20$. What number does it approach, and why (Theorem 3.2)?
  2. Derangements: the number $D_n$ of permutations of $n$ objects with no fixed point has $\sum\frac{D_n}{n!}z^n=\frac{e^{-z}}{1-z}$. Use Exercise I.8 to predict $\lim\frac{D_n}{n!}$, then compute $\frac{D_n}{n!}-\frac1e$ for $n\le15$ and compare it with $\frac{1}{(n+1)!}$. Why is the convergence so fast?
  3. For $\frac{1}{1-z-z^2}=\sum F_{n+1}z^n$ (Fibonacci numbers), locate the poles, find the residue at the nearest one, $b$, and conjecture a formula $c_n\approx\frac{-\operatorname{Res}_bf}{b^{n+1}}$. Test it numerically.
  4. Formulate the general rule for a meromorphic $f$ whose nearest pole $b$ is simple, and prove it by the argument of Exercise I.8.

Tools: Python (sympy, fractions) or any computer algebra system; a spreadsheet for the tables.

Hand in: A one-page note with the three tables, your general rule and its proof.

Going further. Read about the "transfer theorems" of singularity analysis, which handle algebraic singularities such as $\sqrt{1-4z}$ (Catalan numbers $\sim\frac{4^n}{\sqrt\pi\,n^{3/2}}$).

Instructor note: the ratios in step 1 are $6.2855$ ($k=5$), $6.283188$ ($k=10$) and $6.2831853072$ ($k=19$), converging to $2\pi$ (poles at $\pm2\pi i$). In step 2, $D_n/n!\to1/e$; for $n=10$, $D_{10}=1334961$ and $\frac{D_{10}}{10!}-\frac1e\approx2.31\times10^{-8}$, close to $\frac1{11!}\approx2.51\times10^{-8}$ — the remainder is a tail of the entire function $e^{-z}$ at $z=1$. In step 3 the poles are $b=\frac{\sqrt5-1}{2}$ and $-\frac{\sqrt5+1}{2}$, $\operatorname{Res}_bf=-\frac1{\sqrt5}$, and the rule gives Binet's $F_{n+1}\approx\frac{\varphi^{n+1}}{\sqrt5}$ with $\varphi=\frac1b$.

🧪 Self-check quiz

0/5 correct

Q1. The radius of convergence of the Taylor series of $\dfrac{1}{1+z^2}$ about $z=1$ is

Q2. The point $z=0$ is, for $\dfrac{\sin z}{z^3}$,

Q3. The Laurent series of $\dfrac{1}{z-1}$ valid for $|z|\gt1$ is

Q4. If $a$ is an essential singularity of $f$, the Casorati–Weierstrass theorem guarantees that

Q5. An entire function with a pole of order $3$ at $\infty$