Suppose you are trying to understand the behavior of the complex function $f(z) = \frac{1}{z(z-1)}$ near the point $z = 0$. Taylor series cannot help you: the function is not even defined at $z = 0$, let alone holomorphic there. Yet the function is perfectly well-behaved on the punctured disk $0 < |z| < 1$, and you can feel intuitively that there should be some kind of series expansion that captures its behavior in that annular region. The ordinary power series machinery, which demands holomorphicity at the center, is simply the wrong tool. The same difficulty arises whenever you want to extract quantitative information about a function at an isolated singularity — a pole, an essential singularity, or a removable singularity in disguise. Residue calculus, contour integration, and the classification of singularities all depend on having a way to expand $f$ as a series in a punctured neighborhood. That series is the Laurent series. The key insight, due to Pierre Alphonse Laurent (1843), is that if you relax the requirement of a full power series — allowing not just non-negative powers $z^n$ but also negative powers $z^{-n}$ — then any function holomorphic on an annulus $r < |z - z_0| < R$ admits a unique bi-infinite expansion. The non-negative powers reconstruct the "regular" part of $f$, while the negative powers — the *principal part* — encode the singularity structure at $z_0$. [example: A Function That Resists Taylor Expansion] Consider $f: \mathbb{C} \setminus \{0, 1\} \to \mathbb{C}$ defined by \begin{align*} f(z) &= \frac{1}{z(z-1)}. \end{align*} Using partial fractions, we can write \begin{align*} \frac{1}{z(z-1)} &= \frac{-1}{z} + \frac{1}{z-1}. \end{align*} On the annulus $0 < |z| < 1$, the factor $\frac{1}{z-1}$ can be expanded as a geometric series. Since $|z| < 1$, we have $\frac{1}{z-1} = \frac{-1}{1-z} = -\sum_{n=0}^{\infty} z^n$, and this series converges absolutely and uniformly on compact subsets of $|z| < 1$. Combining with the $-\frac{1}{z}$ term: \begin{align*} f(z) &= \frac{-1}{z} - \sum_{n=0}^{\infty} z^n \\ &= \frac{-1}{z} - 1 - z - z^2 - z^3 - \cdots \end{align*} This is a perfectly convergent series for $0 < |z| < 1$, but it contains the term $z^{-1}$, which no Taylor series can produce. On the annulus $1 < |z| < \infty$, the expansion looks entirely different: now $|1/z| < 1$, so \begin{align*} \frac{1}{z-1} &= \frac{1}{z} \cdot \frac{1}{1 - 1/z} = \frac{1}{z} \sum_{n=0}^{\infty} z^{-n} = \sum_{n=0}^{\infty} z^{-(n+1)}, \end{align*} giving \begin{align*} f(z) &= \frac{-1}{z} + \sum_{n=0}^{\infty} z^{-(n+1)} = \sum_{n=2}^{\infty} z^{-n}. \end{align*} The same function has two completely different Laurent expansions on two different annuli. This is not a contradiction — it is a feature: the Laurent series is local to the annulus, and the coefficients adapt to the geometry of the domain. [/example] ## Definition The right framework is the annulus. An annulus centered at $z_0$ with radii $0 \le r < R \le \infty$ is the open set $A(z_0; r, R) := \{ z \in \mathbb{C} : r < |z - z_0| < R \}$. When $r = 0$ this is a punctured disk; when $r = 0$ and $R = \infty$ it is the entire punctured plane. This is the natural domain of convergence for a bi-infinite power series. The key representation theorem is that every function holomorphic on an annulus admits a unique bi-infinite series expansion. This is the content of Laurent's theorem, and it is the central definition around which the rest of the theory is built. [definition: Laurent Series] Let $\Omega \subset \mathbb{C}$ be an open set, $z_0 \in \mathbb{C}$, and suppose $f: \Omega \to \mathbb{C}$ is holomorphic on the annulus $A(z_0; r, R) \subset \Omega$. The **Laurent series** of $f$ centered at $z_0$ on $A(z_0; r, R)$ is the bi-infinite series \begin{align*} \sum_{n=-\infty}^{\infty} a_n (z - z_0)^n, \end{align*} where the coefficients $a_n \in \mathbb{C}$ are given by the formula \begin{align*} a_n &= \frac{1}{2\pi i} \oint_{\gamma} \frac{f(w)}{(w - z_0)^{n+1}}\, dw \end{align*} for any positively oriented simple closed curve $\gamma$ in $A(z_0; r, R)$ with $z_0$ in its interior. The series converges absolutely and uniformly on compact subsets of $A(z_0; r, R)$. [/definition] [remark: Coefficient Formula Independence] The formula $a_n = \frac{1}{2\pi i} \oint_{\gamma} \frac{f(w)}{(w-z_0)^{n+1}} dw$ is independent of the choice of $\gamma$, as long as $\gamma$ is a positively oriented simple closed curve in the annulus encircling $z_0$. This follows from Cauchy's theorem: any two such curves are homologous in the annulus, so the integrals agree. In practice one almost always takes $\gamma$ to be the circle $|w - z_0| = \rho$ for some $r < \rho < R$. [/remark] The series splits naturally into two parts. The part with $n \ge 0$ is called the **analytic part** (or regular part) of the Laurent series, and the part with $n \le -1$ is called the **principal part**. The principal part is not merely a bookkeeping device. A natural question arises once you have the Laurent expansion: which terms carry the information about the singularity at $z_0$, and which terms reflect the regular behavior of $f$ nearby? The non-negative powers $(z - z_0)^n$ for $n \ge 0$ extend holomorphically to $z_0$ and contribute nothing pathological — they are the Taylor part. It is the negative powers that encode everything singular. Whether the principal part vanishes, has finitely many terms, or has infinitely many terms determines the type of singularity at $z_0$ — and the single coefficient $a_{-1}$ of $(z - z_0)^{-1}$ controls the value of every contour integral around $z_0$. Isolating the negative-power terms makes both the singularity classification and the residue extraction entirely explicit. [definition: Principal Part] Let $\sum_{n=-\infty}^{\infty} a_n (z - z_0)^n$ be the Laurent series of $f$ on $A(z_0; r, R)$. The **principal part** of $f$ at $z_0$ is the finite or infinite series \begin{align*} P(z) &:= \sum_{n=1}^{\infty} \frac{a_{-n}}{(z - z_0)^n} = \frac{a_{-1}}{z - z_0} + \frac{a_{-2}}{(z-z_0)^2} + \cdots \end{align*} [/definition] The principal part $P(z)$ is holomorphic on the exterior region where the negative-power series converges, namely $|z-z_0|>r$, while the analytic part $\sum_{n=0}^\infty a_n(z-z_0)^n$ is holomorphic on $|z - z_0| < R$. On the overlap annulus $A(z_0;r,R)$, the Laurent decomposition $f = P + (\text{analytic part})$ separates the singular behavior from the regular behavior. The coefficient formula above is useful only if it really recovers every holomorphic function on an annulus. The next theorem supplies that guarantee: it says that the annulus, the coefficient integrals, and the resulting bi-infinite series are not optional features but the canonical representation of the function there. [quotetheorem:350] Uniqueness is the crucial point: if you find *any* bi-infinite series representation $\sum_{n=-\infty}^{\infty} c_n (z-z_0)^n$ that converges to $f(z)$ on $A(z_0; r, R)$, then necessarily $c_n = a_n$ for all $n$. This means that whenever you manipulate geometric series, partial fractions, or known Taylor expansions to produce a bi-infinite series for $f$, you have automatically found the Laurent coefficients, without ever evaluating the contour integral directly. [illustration:annulus-coefficient-contour] ## Convergence and the Annulus of Convergence The comparison with power series illuminates the global convergence structure. A Taylor series $\sum_{n=0}^\infty a_n z^n$ converges on a disk $|z| < R$ and diverges outside it: the disk of convergence is determined by a single radius. A Laurent series has *two* radii determining its domain of convergence, and the series converges on the annular region between them. To understand why, note that the Laurent series $\sum_{n=-\infty}^{\infty} a_n (z-z_0)^n$ splits into two separately convergent series: - The **analytic part** $\sum_{n=0}^{\infty} a_n (z-z_0)^n$ is an ordinary power series. By the theory of power series, it has a radius of convergence $R$ defined by $1/R = \limsup_{n \to \infty} |a_n|^{1/n}$, and it converges absolutely on $|z - z_0| < R$. - The **principal part** $\sum_{n=1}^{\infty} a_{-n} (z-z_0)^{-n}$ is a power series in $1/(z-z_0)$. Setting $w = 1/(z-z_0)$, it becomes $\sum_{n=1}^\infty a_{-n} w^n$. If $r := \limsup_{n \to \infty} |a_{-n}|^{1/n}$, then this power series has radius of convergence $1/r$ in the $w$-variable. Converting back: it converges when $|w| < 1/r$, i.e., when $|z - z_0| > r$. [quotetheorem:3364] This also explains why the same formula can have different Laurent series in annular regions separated by singularities. If two annuli lie in the same connected region of holomorphy and one is obtained by shrinking the other, uniqueness forces the coefficients to agree on the overlap. Different expansions appear when a singularity, such as $z=2$ below, separates one annular domain from another. [example: Geometric Series Yielding Different Laurent Expansions] Let $f: \mathbb{C} \setminus \{0\} \to \mathbb{C}$ be defined by \begin{align*} f(z) &= \frac{1}{z - 2}. \end{align*} On the annulus $A(0; 0, 2) = \{0 < |z| < 2\}$, we have $|z/2| < 1$, so \begin{align*} \frac{1}{z - 2} &= \frac{-1}{2} \cdot \frac{1}{1 - z/2} = \frac{-1}{2} \sum_{n=0}^{\infty} \left(\frac{z}{2}\right)^n = -\sum_{n=0}^{\infty} \frac{z^n}{2^{n+1}}. \end{align*} This is actually a Taylor series — it has no principal part, because $f$ has no singularity at $z = 0$; the only singularity is at $z = 2$, which lies outside the annulus. The inner radius $r = 0$ here is not forced by a singularity of $f$ but by our choice of domain. On the annulus $A(0; 2, \infty) = \{|z| > 2\}$, we have $|2/z| < 1$, so \begin{align*} \frac{1}{z - 2} &= \frac{1}{z} \cdot \frac{1}{1 - 2/z} = \frac{1}{z} \sum_{n=0}^{\infty} \left(\frac{2}{z}\right)^n = \sum_{n=0}^{\infty} \frac{2^n}{z^{n+1}}. \end{align*} In this region, only negative powers of $z$ appear: $\frac{1}{z} + \frac{2}{z^2} + \frac{4}{z^3} + \cdots$. Both expansions converge on their respective domains, and neither contradicts the other. [/example] ## Classification of Isolated Singularities The greatest application of Laurent series in basic complex analysis is the classification of isolated singularities. If $f$ is holomorphic on a punctured disk $A(z_0; 0, R)$ and fails to extend holomorphically to $z_0$, then $z_0$ is called an *isolated singularity*. The structure of the principal part in the Laurent expansion at $z_0$ determines the type of the singularity, and this classification has sharp consequences for the behavior of $f$. An isolated singularity can be one of exactly three types. The principal part may be identically zero, may have finitely many nonzero terms, or may have infinitely many nonzero terms. Each case leads to qualitatively different behavior. The first case is the most surprising: a function can appear to have a singularity at $z_0$ simply because it was never defined there, even though it extends perfectly well. The principal part being zero is the precise way to detect this. [illustration:singularity-classification] [definition: Removable Singularity] Let $f: A(z_0; 0, R) \to \mathbb{C}$ be holomorphic. The point $z_0$ is a **removable singularity** of $f$ if all coefficients $a_n = 0$ for $n \le -1$ in the Laurent expansion of $f$ at $z_0$. [/definition] Equivalently, the principal part of $f$ at $z_0$ is identically zero. If $z_0$ is a removable singularity, then $f(z) = \sum_{n=0}^\infty a_n (z-z_0)^n$ on $A(z_0; 0, R)$, and defining $f(z_0) := a_0$ extends $f$ to a holomorphic function on the full disk $|z - z_0| < R$. The "singularity" was entirely illusory — the function was already holomorphic, just not yet defined at $z_0$. In practice, checking every Laurent coefficient is rarely the easiest way to decide whether a puncture can be filled. A more usable test asks only whether the function stays bounded near the missing point. Here $\mathcal{O}(U)$ denotes the holomorphic functions on an open set $U$, and $B(z_0,r)=\{z\in\mathbb{C}: |z-z_0|<r\}$ denotes the open disk used to state that local boundedness criterion. [quotetheorem:3356] The theorem separates genuine singularities from missing values. Once the bounded case has been removed, the next structured possibility is that the function diverges to infinity in a controlled, polynomial way. The number of negative-power terms in the Laurent expansion, called the *order* of the pole, measures exactly how rapidly the divergence occurs. [definition: Pole] Let $f: A(z_0; 0, R) \to \mathbb{C}$ be holomorphic. The point $z_0$ is a **pole of order $m$** (for some $m \in \mathbb{N}$, $m \ge 1$) if $a_{-m} \ne 0$ but $a_n = 0$ for all $n < -m$ in the Laurent expansion of $f$ at $z_0$. A pole of order $1$ is called a **simple pole**. [/definition] At a pole of order $m$, the function $f$ blows up like $1/(z-z_0)^m$ as $z \to z_0$: we have $|f(z)| \to \infty$ as $z \to z_0$. Moreover, the function $(z - z_0)^m f(z)$ has a removable singularity at $z_0$, and extends holomorphically there with nonzero value $a_{-m}$. Poles still have only finitely much singular data, so multiplying by a high enough power of $z-z_0$ restores holomorphy. The remaining case is different: infinitely many negative Laurent coefficients persist, and no finite multiplication removes the singular behavior. [definition: Essential Singularity] Let $f: A(z_0; 0, R) \to \mathbb{C}$ be holomorphic. The point $z_0$ is an **essential singularity** of $f$ if the principal part of $f$ at $z_0$ has infinitely many nonzero terms, i.e., $a_{-n} \ne 0$ for infinitely many $n \ge 1$. [/definition] An infinite principal part should produce more than just the failure of a limit; it should force genuinely unpredictable local behavior. The central question is how large the image of every punctured neighborhood must be when no finite pole order controls the singularity. [quotetheorem:3355] [example: The Essential Singularity of $e^{1/z}$] The function $f: \mathbb{C} \setminus \{0\} \to \mathbb{C}$ defined by $f(z) = e^{1/z}$ has an essential singularity at $z_0 = 0$. To see this, compute the Laurent expansion by substituting $1/z$ into the exponential series: \begin{align*} e^{1/z} &= \sum_{n=0}^{\infty} \frac{1}{n!} \left(\frac{1}{z}\right)^n = \sum_{n=0}^{\infty} \frac{1}{n!\, z^n} = 1 + \frac{1}{z} + \frac{1}{2! z^2} + \frac{1}{3! z^3} + \cdots \end{align*} This series converges for all $z \ne 0$ (it is valid on $A(0; 0, \infty)$), and the principal part $\sum_{n=1}^\infty \frac{1}{n!} z^{-n}$ has infinitely many nonzero terms. So $z = 0$ is an essential singularity. To see the wild behavior concretely, consider just the target value $w = 1$. Along the imaginary axis, $e^{1/(iy)} = e^{-i/y}$, which has modulus $1$ for all $y \ne 0$. As $y \to 0$, the argument $-1/y \to \pm\infty$, and $e^{-i/y}$ winds around the unit circle infinitely often. In particular, $e^{-i/y} = 1$ whenever $1/y = 2\pi k$ for $k \in \mathbb{Z}$, so $e^{1/z}$ takes the value $1$ at points $z = i/(2\pi k)$ accumulating at $0$. This single calculation illustrates the behavior captured by Casorati-Weierstrass; the theorem asserts the full density statement for every target value. [/example] [remark: The Classification Is Exhaustive] Every isolated singularity is exactly one of: removable, pole, or essential. There are no other cases. This is because the Laurent expansion is unique, and the principal part is either zero, has finite many nonzero terms, or has infinitely many. The trichotomy is sharp. [/remark] ## Residues and the Residue Theorem The Laurent expansion at an isolated singularity $z_0$ produces a single coefficient that plays a privileged role in integration theory: the coefficient $a_{-1}$ of $(z - z_0)^{-1}$. Why is $a_{-1}$ special? Consider integrating the Laurent series term by term around a small positively oriented circle $\gamma_\rho: |z - z_0| = \rho$. For $n \ne -1$, the antiderivative of $(z-z_0)^n$ is $\frac{(z-z_0)^{n+1}}{n+1}$, which is single-valued and returns to its starting value after one circuit: the integral is zero. But $(z-z_0)^{-1}$ has antiderivative $\log(z - z_0)$, which is multi-valued: traversing $\gamma_\rho$ once increases the logarithm by $2\pi i$. The upshot is: \begin{align*} \oint_{\gamma_\rho} (z - z_0)^n\, dz &= \begin{cases} 2\pi i & \text{if } n = -1, \\ 0 & \text{if } n \ne -1. \end{cases} \end{align*} Therefore, integrating the Laurent series term by term, all terms vanish except $n = -1$, and \begin{align*} \oint_{\gamma_\rho} f(z)\, dz &= 2\pi i \cdot a_{-1}. \end{align*} This is the content of the definition: [definition: Residue] Let $z_0$ be an isolated singularity of $f$, and let $\sum_{n=-\infty}^{\infty} a_n (z-z_0)^n$ be the Laurent expansion of $f$ on $A(z_0; 0, R)$. The **residue** of $f$ at $z_0$ is \begin{align*} \operatorname{Res}(f, z_0) &:= a_{-1}. \end{align*} [/definition] The local computation around one singularity becomes powerful only when several singularities lie inside the same contour. The residue theorem is the global form of the calculation above: it replaces a contour integral by a weighted sum of the residues enclosed by the curve. [quotetheorem:352] After the integral has been reduced to residues, the practical problem is extracting the coefficient $a_{-1}$ efficiently. Poles have finite principal parts, so their residues can be recovered from limits and derivatives instead of from a full Laurent expansion. [quotetheorem:353] In practice, many functions arising in contour integration take the form of a quotient $g/h$, and the quotient formula in the theorem avoids expanding the Laurent series entirely. The diagnostic is simple: if the denominator has a simple zero at the pole while the numerator is nonzero there, the residue is read off from the numerator value and the derivative of the denominator. [example: Computing a Contour Integral via Residues] Compute $I = \oint_{|z| = 2} \frac{e^z}{z^2(z-1)}\, dz$. The integrand $f(z) = e^z / (z^2(z-1))$ has isolated singularities at $z = 0$ (a pole of order $2$) and $z = 1$ (a simple pole). Both lie inside the circle $|z| = 2$. For the simple pole at $z = 1$, the residue formula gives \begin{align*} \operatorname{Res}(f, 1) &= \lim_{z \to 1} (z - 1) \cdot \frac{e^z}{z^2(z-1)} = \lim_{z \to 1} \frac{e^z}{z^2} = \frac{e^1}{1^2} = e. \end{align*} For the pole of order $2$ at $z = 0$, the higher-order formula requires one differentiation. We have $z^2 f(z) = \frac{e^z}{z-1}$, so \begin{align*} \frac{d}{dz}\left[\frac{e^z}{z-1}\right] &= \frac{e^z(z-1) - e^z \cdot 1}{(z-1)^2} = \frac{e^z(z-2)}{(z-1)^2}. \end{align*} Evaluating at $z = 0$: \begin{align*} \frac{d}{dz}\left[\frac{e^z}{z-1}\right]\bigg|_{z=0} &= \frac{e^0(0-2)}{(0-1)^2} = \frac{1 \cdot (-2)}{1} = -2. \end{align*} By the formula, $\operatorname{Res}(f, 0) = \frac{1}{(2-1)!} \cdot (-2) = -2$. Applying the Residue Theorem to sum the contributions from both singularities: \begin{align*} I &= 2\pi i \left( \operatorname{Res}(f, 0) + \operatorname{Res}(f, 1) \right) = 2\pi i(e - 2). \end{align*} [/example] ## Laurent Series at Infinity So far, we have worked at a finite center $z_0 \in \mathbb{C}$. But in many applications — especially in studying the behavior of meromorphic functions on the Riemann sphere, or evaluating integrals over the real line by closing contours in the upper or lower half-plane — we need to understand the behavior of $f(z)$ as $|z| \to \infty$. The Laurent expansion at a finite point $z_0$ tells us about the singularity structure there, but many problems require understanding what happens as $|z| \to \infty$. For instance, when closing a contour in the upper half-plane to evaluate a real integral, you need to know whether $f$ decays at infinity, has a pole there, or behaves erratically — and the finite Laurent expansions tell you nothing about this. The point at infinity on the Riemann sphere is just as much a potential singularity as any finite point, and it needs its own Laurent analysis. The standard device is the substitution $w = 1/z$, which maps a punctured neighborhood of $\infty$ to a punctured neighborhood of $0$, reducing the problem to one we have already solved. [definition: Laurent Expansion at Infinity] Let $f$ be holomorphic on $\{z \in \mathbb{C} : |z| > R\}$ for some $R \ge 0$. Define $g: \{w : 0 < |w| < 1/R\} \to \mathbb{C}$ by \begin{align*} g(w) &:= f(1/w). \end{align*} The **Laurent expansion of $f$ at infinity** is the series obtained by expressing $g(w) = \sum_{n=-\infty}^\infty b_n w^n$ (the Laurent expansion of $g$ at $w = 0$) and rewriting in terms of $z = 1/w$: \begin{align*} f(z) &= \sum_{n=-\infty}^{\infty} b_n z^{-n} = \sum_{n=-\infty}^{\infty} c_n z^n, \end{align*} where $c_n := b_{-n}$. [/definition] The point $z = \infty$ is a **removable singularity** of $f$ if the expansion at infinity has no positive powers of $z$ (so $c_n = 0$ for $n \ge 1$), a **pole of order $m$** if $c_m \ne 0$ and $c_n = 0$ for $n > m$, and an **essential singularity** if infinitely many positive powers of $z$ appear. The **residue at infinity** deserves special mention. It is defined as \begin{align*} \operatorname{Res}(f, \infty) &:= -\frac{1}{2\pi i} \oint_{|z| = R+1} f(z)\, dz = -b_1 = -c_{-1}. \end{align*} The sign convention is designed to make infinity behave like an ordinary point on the compact Riemann sphere. With that convention, residues satisfy a global balance law: all finite contributions are exactly offset by the contribution at infinity. [quotetheorem:3366] [example: Using the Residue at Infinity] Consider $f(z) = \frac{1}{z^2 + 1} = \frac{1}{(z+i)(z-i)}$. The finite singularities are at $z = i$ and $z = -i$. Computing directly: $\operatorname{Res}(f, i) = \lim_{z \to i} (z-i) \cdot \frac{1}{(z+i)(z-i)} = \frac{1}{2i}$ and $\operatorname{Res}(f, -i) = \frac{1}{-2i} = \frac{-1}{2i}$. The sum of finite residues is $\frac{1}{2i} - \frac{1}{2i} = 0$. By the global residue theorem, $\operatorname{Res}(f, \infty) = 0$ as well. We can verify this directly. As $|z| \to \infty$: \begin{align*} f(z) &= \frac{1}{z^2(1 + 1/z^2)} = \frac{1}{z^2} \sum_{n=0}^\infty \frac{(-1)^n}{z^{2n}} = \frac{1}{z^2} - \frac{1}{z^4} + \frac{1}{z^6} - \cdots \end{align*} The coefficient of $z^{-1}$ is zero, confirming $\operatorname{Res}(f, \infty) = 0$. The expansion at infinity is a series in $z^{-2}$: all even negative powers, and no $z^{-1}$ term. [/example] ## Partial Fractions and Mittag-Leffler Decomposition Every rational function — a ratio of polynomials — is meromorphic on $\hat{\mathbb{C}}$, and the theory of Laurent series gives a complete structural decomposition. Partial fractions, familiar from calculus, is a special case of a much deeper theorem. The partial fraction theorem says: if $f(z) = P(z)/Q(z)$ is a rational function with $\deg P < \deg Q$ and $Q$ has simple zeros at $z_1, \ldots, z_k$, then \begin{align*} f(z) &= \sum_{j=1}^{k} \frac{\operatorname{Res}(f, z_j)}{z - z_j}. \end{align*} This is exactly the statement that $f$ equals the sum of its principal parts at each finite singularity (with a simple pole, the principal part is just $\frac{a_{-1}}{z - z_j}$). For poles of higher order, the principal part at $z_j$ is $\sum_{n=1}^{m_j} \frac{a_{-n}^{(j)}}{(z-z_j)^n}$. The partial fraction decomposition then states that $f$ equals the sum of its principal parts at all finite poles, plus a polynomial (which is the principal part at $\infty$). The Laurent expansion at each singularity makes this decomposition completely explicit. The Mittag-Leffler theorem generalizes this to infinite collections of poles: [quotetheorem:3367] This theorem is the additive analogue of Weierstrass's factorization theorem (which prescribes zeros multiplicatively). Together they give a complete picture of the possible singularity structures of meromorphic functions. [remark: Laurent Series and Partial Fractions] For a rational function with simple poles at $z_1, \ldots, z_k$, the Laurent expansion at $z_j$ immediately reads off the residue: $\operatorname{Res}(f, z_j) = a_{-1}^{(j)}$, the $-1$ coefficient. The partial fraction decomposition is therefore a direct corollary of the Laurent expansion theorem and uniqueness. No separate argument is needed once the Laurent theory is in place. [/remark] ## Connection to Fourier Series Laurent series on the unit circle are precisely Fourier series. If $f$ is holomorphic on an annulus containing the unit circle $|z| = 1$, the Laurent expansion \begin{align*} f(z) &= \sum_{n=-\infty}^{\infty} a_n z^n \end{align*} restricts to $z = e^{i\theta}$ to give \begin{align*} f(e^{i\theta}) &= \sum_{n=-\infty}^{\infty} a_n e^{in\theta}, \end{align*} which is the Fourier series of the function $\theta \mapsto f(e^{i\theta})$ on the circle $[0, 2\pi)$. The Laurent coefficient formula $a_n = \frac{1}{2\pi i} \oint_{|z|=1} \frac{f(z)}{z^{n+1}} dz$ becomes, after substituting $z = e^{i\theta}$, \begin{align*} a_n &= \frac{1}{2\pi} \int_0^{2\pi} f(e^{i\theta}) e^{-in\theta}\, d\theta, \end{align*} which is exactly the standard Fourier coefficient formula. This connection runs deep: the $L^2$ theory of Fourier series on the circle corresponds to the $H^2$ Hardy space of holomorphic functions on the disk, and the analytic and principal parts of the Laurent expansion correspond to the non-negative and negative frequency components of the Fourier series. The decay rate of $|a_n|$ as $|n| \to \infty$ reflects the smoothness of the boundary function — a fact exploited throughout harmonic analysis. ## References Lars Ahlfors, *Complex Analysis* (3rd ed., 1979). The standard rigorous treatment of Laurent series, convergence, and residues. Henri Cartan, *Elementary Theory of Analytic Functions of One or Several Complex Variables* (1963). Emphasizes the sheaf-theoretic and Cauchy-theory perspective on Laurent expansions. Elias M. Stein and Rami Shakarchi, *Complex Analysis* (Princeton Lectures in Analysis, Vol. II, 2003). Excellent motivated treatment of Laurent series, singularity classification, and the residue theorem, with computational examples. Reinhold Remmert, *Theory of Complex Functions* (1991). Encyclopedic reference with historical notes on Laurent's original contributions and the development of the principal part decomposition. Walter Rudin, *Real and Complex Analysis* (3rd ed., 1987). Brief but rigorous treatment within the broader framework of harmonic analysis and measures.