The Identity Theorem and Zeros

Suppose two holomorphic functions agree on a tiny arc of points. These points converge to some interior point of a domain. Must the functions agree everywhere on that domain? For general smooth functions this is absurdly false. A bump function can vanish on a whole interval. Yet it can still be nonzero elsewhere. Holomorphic functions are astonishingly more rigid. Agreement on any set with an accumulation point forces agreement everywhere. This phenomenon, the Identity Theorem, is one of the deepest structural facts. It distinguishes complex-differentiable functions from their real counterparts. The reason lies in the power of Taylor Series of Holomorphic Functions. A holomorphic function is locally equal to its Taylor series. It is not merely approximated by it. This single fact — analyticity — means something important. A holomorphic function is entirely encoded by the infinite sequence of derivatives at one point. If those derivatives vanish, the function is trapped into being identically zero. The same happens if the function's zero set clusters anywhere inside the domain. Zeros of holomorphic functions are therefore isolated, unlike zeros of general smooth functions. This holds unless the function itself is trivial. This rigidity underlies much of complex analysis. It explains why complex functions cannot "turn off" on a sub-region. They cannot turn back on elsewhere either. It explains why the notion of order of a zero is well-defined. This order is computable from a Laurent-type expansion. It also explains why analytic formulas propagate uniquely across connected domains. It is the theoretical seed from which Analytic Continuation grows. If you know a function on a small piece of a domain, you know it everywhere it can be defined analytically. The extension — if it exists — is unique. This article develops the theory of isolated zeros and their orders. It proves the Identity Theorem in full rigor using connectedness and power series. It shows how these ideas set the stage for counting zeros. This counting happens via The Argument Principle and Rouche's Theorem. We include complete proofs and concrete worked computations of zero orders. We also include a counterexample showing where the theorem fails without connectedness. Finally, we discuss the theorem's role across mathematics.

Key takeaways
  • A nonconstant holomorphic function's zeros are isolated: no zero has other zeros arbitrarily close to it, unless the function vanishes identically nearby.
  • The order of a zero aa is the unique integer m≥1m \ge 1 such that f(z)=(z−a)mg(z)f(z) = (z-a)^m g(z) with g(a)≠0g(a) \ne 0 and gg holomorphic.
  • The Identity Theorem: if two holomorphic functions on a connected open set agree on a set with an accumulation point in that set, they agree everywhere on it.
  • Equivalently: if ff's zero set has an accumulation point inside a domain of holomorphy, then f≡0f \equiv 0 there.
  • Connectedness of the domain is essential — the theorem fails on disconnected domains, as the counterexample shows.
  • Key idea: holomorphy forces global rigidity from purely local Taylor data, because power series determine a function completely on its disk of convergence and analytic continuation forces consistency across a connected domain.

Where this lives today

The Identity Theorem is the conceptual bedrock for analytic continuation: if a function is prescribed on a small piece of a domain (a real interval, an arc, or a small disk), the theorem guarantees any holomorphic extension to a larger connected domain is unique, which is exactly what lets mathematicians unambiguously define functions like the Riemann zeta function, the Gamma function, and hypergeometric functions across their full domains via continuation from a region of convergence. It also underlies the rigidity used in the theory of several complex variables, in the uniqueness of solutions to certain PDEs with analytic data (Cauchy–Kovalevskaya-type uniqueness), and in complex-analytic proofs throughout algebraic geometry where holomorphic (and eventually algebraic) functions are shown to be globally determined by local data. In applied contexts, the theorem justifies why engineers and physicists can compute a complex potential or a transfer function on a small test region and trust the formula elsewhere in a connected domain of analyticity — a principle exploited in analytic signal processing, control theory (poles and zeros of transfer functions, where the order-of-zero concept directly measures multiplicity of a system's zero), and in quantum field theory's use of analytic continuation between Minkowski and Euclidean signatures. In cryptography and coding theory, finite-field analogues of "polynomials determined by few values" (interpolation, Reed–Solomon codes) echo the same rigidity philosophy: an algebraic/analytic object of bounded complexity is determined by comparatively little data. The order-of-zero machinery developed here is also the direct precursor to The Argument Principle and Rouche's Theorem, where the sum of zero orders inside a contour is computed via a contour integral, turning the local notion of multiplicity into a global counting tool.

Ask the tutor: Why exactly does connectedness let "agreement near one point" propagate to "agreement everywhere," and what breaks in the open–closed argument if the domain has two components?

DefinitionZeros, order of a zero, and the identity principle

Let Ω⊆C\Omega \subseteq \mathbb{C} be open and connected (a domain), and let f:Ω→Cf : \Omega \to \mathbb{C} be holomorphic.

Zero. A point a∈Ωa \in \Omega is a zero of ff if f(a)=0f(a) = 0.

Isolated zero. A zero aa of ff is isolated if there exists r>0r > 0 such that f(z)≠0f(z) \ne 0 for all zz with 0<∣z−a∣<r0 < |z - a| < r.

Order of a zero. Suppose f(a)=0f(a) = 0 and f≢0f \not\equiv 0 on any neighborhood of aa. Since ff is holomorphic at aa, it has a Taylor expansion

f(z)=∑n=0∞cn(z−a)n,cn=f(n)(a)n!,f(z) = \sum_{n=0}^{\infty} c_n (z-a)^n, \qquad c_n = \frac{f^{(n)}(a)}{n!},

convergent on some disk D(a,r)⊆ΩD(a,r) \subseteq \Omega. Because ff is not identically zero near aa, not all cnc_n vanish; let

m=min⁡{n≥0:cn≠0}.m = \min\{ n \ge 0 : c_n \ne 0 \}.

This mm is called the order (or multiplicity) of the zero at aa. Since f(a)=0f(a)=0 we have c0=0c_0 = 0, so m≥1m \ge 1. We say aa is a zero of order mm, and equivalently ff has an mm-th order zero at aa.

Accumulation point. A point a∈Ωa \in \Omega is an accumulation point of a set S⊆ΩS \subseteq \Omega if every neighborhood of aa contains a point of SS other than aa itself.

Identity principle. Two holomorphic functions f,g:Ω→Cf, g: \Omega \to \mathbb{C} on a domain Ω\Omega coincide if f(z)=g(z)f(z) = g(z) for all z∈Ωz \in \Omega.

Intuition

Think of a holomorphic function as a rigid sheet of information: knowing it on an arbitrarily small curled-up patch near a point pins down every Taylor coefficient there, and the Taylor series itself reconstructs the function on a whole disk. If the function also agrees with another holomorphic function along a sequence approaching that point, the coefficients must match to all orders — the two functions are Taylor-identical at that point, hence literally identical near it. The connectedness of the domain then acts like a chain of overlapping disks: agreement propagates disk by disk across the whole domain, the way a rumor "the two functions agree here" spreads to every corner reachable by a path of overlapping neighborhoods. Contrast this with the real line: f(x)=e−1/x2sin⁡(1/x)f(x) = e^{-1/x^2}\sin(1/x) for x≠0x\neq0 (extended by 00 at x=0x=0) is smooth, and vanishes at the sequence xn=1/(nπ)→0x_n = 1/(n\pi) \to 0, yet is not identically zero anywhere near 00. Real-differentiability, even infinite differentiability, allows a function's derivatives to encode only "local" information that need not force global vanishing. Complex differentiability is stronger: it bundles in the full power series representation, so the zero set of a nontrivial holomorphic function cannot accumulate anywhere inside the domain. The order of a zero measures how flat ff is there: order 11 means ff crosses zero "transversally" (nonzero derivative), while higher order mm means the first m−1m-1 derivatives vanish too, and locally ff behaves like cm(z−a)mc_m(z-a)^m — an mm-fold covering map near aa, which is the geometric seed of the argument principle's zero-counting formula.

Figure
Order-3 zero: local behavior of f(z)=(z−a)3g(z)f(z) = (z-a)^3 g(z) near aa, mapping a small circle around aa to a curve winding three times around 00
z-plane near a a |z-a|=r w-plane: f(z), winds 3× 0 image winds around 0 three times as z winds once
TheoremIsolated Zeros and the Identity Theorem

Let Ω⊆C\Omega \subseteq \mathbb{C} be a domain (open and connected), and let f:Ω→Cf : \Omega \to \mathbb{C} be holomorphic.

(A) Isolated Zeros. If f≢0f \not\equiv 0 on Ω\Omega, then every zero of ff in Ω\Omega is isolated. Equivalently, the zero set Z(f)={z∈Ω:f(z)=0}Z(f) = \{z \in \Omega : f(z) = 0\} has no accumulation point in Ω\Omega.

(B) Identity Theorem. Let g:Ω→Cg : \Omega \to \mathbb{C} also be holomorphic. If the set

S={z∈Ω:f(z)=g(z)}S = \{ z \in \Omega : f(z) = g(z) \}

has an accumulation point a∈Ωa \in \Omega, then f≡gf \equiv g on all of Ω\Omega. (Part (A) is the special case g≡0g \equiv 0.)

LemmaLocal rigidity at an accumulation point

Let Ω\Omega be open, h:Ω→Ch : \Omega \to \mathbb{C} holomorphic, and suppose hh has a zero at a∈Ωa \in \Omega that is not isolated, meaning every punctured neighborhood of aa contains another zero of hh. Then h≡0h \equiv 0 on some disk D(a,r)⊆ΩD(a, r) \subseteq \Omega.

Proof

(Lemma, by contradiction using the Taylor expansion.) Since hh is holomorphic at aa, by Taylor Series of Holomorphic Functions there is ρ>0\rho > 0 with D(a,ρ)⊆ΩD(a,\rho) \subseteq \Omega and

h(z)=∑n=0∞cn(z−a)nfor all z∈D(a,ρ),h(z) = \sum_{n=0}^{\infty} c_n (z - a)^n \qquad \text{for all } z \in D(a,\rho),

where cn=h(n)(a)/n!c_n = h^{(n)}(a)/n!. Suppose, for contradiction, that hh is not identically zero on D(a,ρ)D(a,\rho). Then not all cnc_n are zero, so we may set

m=min⁡{n≥0:cn≠0}<∞.m = \min\{ n \ge 0 : c_n \ne 0 \} < \infty.

Since h(a)=0h(a) = 0, we have c0=0c_0 = 0, forcing m≥1m \ge 1. Factor out (z−a)m(z-a)^m:

h(z)=(z−a)m∑k=0∞cm+k(z−a)k=:(z−a)m ϕ(z),h(z) = (z-a)^m \sum_{k=0}^{\infty} c_{m+k} (z-a)^k =: (z-a)^m \, \phi(z),

where ϕ(z)=∑k≥0cm+k(z−a)k\phi(z) = \sum_{k \ge 0} c_{m+k}(z-a)^k is itself a convergent power series on D(a,ρ)D(a,\rho) (same radius of convergence, since we've only shifted indices and the original series converges there), hence ϕ\phi is holomorphic on D(a,ρ)D(a,\rho). By construction ϕ(a)=cm≠0\phi(a) = c_m \ne 0. Since ϕ\phi is continuous at aa and ϕ(a)≠0\phi(a) \ne 0, there exists 0<r≤ρ0 < r \le \rho such that ϕ(z)≠0\phi(z) \ne 0 for all z∈D(a,r)z \in D(a,r) (continuity: choose rr so that ∣ϕ(z)−ϕ(a)∣<∣ϕ(a)∣|\phi(z) - \phi(a)| < |\phi(a)| on D(a,r)D(a,r), which forces ϕ(z)≠0\phi(z) \neq 0 there by the triangle inequality). Then for z∈D(a,r)z \in D(a,r) with z≠az \ne a,

h(z)=(z−a)mϕ(z)≠0,h(z) = (z-a)^m \phi(z) \ne 0,

since neither factor vanishes. Thus aa is an isolated zero of hh — contradicting the hypothesis that every punctured neighborhood of aa contains another zero. This contradiction shows h≡0h \equiv 0 on D(a,ρ)D(a,\rho), proving the lemma. ■\blacksquare

∎
Proof

(Theorem, using the lemma and a connectedness/open-closed argument.) Part (A). Suppose f≢0f \not\equiv 0 on Ω\Omega but some zero a∈Z(f)a \in Z(f) is not isolated. By the contrapositive reading of the Lemma (applied with h=fh = f), non-isolation of the zero at aa forces f≡0f \equiv 0 on a disk around aa. We now show this propagates to force f≡0f\equiv 0 on all of Ω\Omega, contradicting f≢0f \not\equiv 0. Define

U={z∈Ω:f≡0 on some open disk around z}.U = \{ z \in \Omega : f \equiv 0 \text{ on some open disk around } z \}.

We claim UU is nonempty, open, and closed in Ω\Omega; since Ω\Omega is connected this forces U=ΩU = \Omega, i.e. f≡0f \equiv 0 on Ω\Omega.

  • Nonempty: a∈Ua \in U by the Lemma applied above.
  • Open: if z∈Uz \in U, then by definition there is a disk D(z,δ)D(z,\delta) on which f≡0f \equiv 0; every point w∈D(z,δ)w \in D(z,\delta) then also has a disk (a smaller one contained in D(z,δ)D(z,\delta)) on which f≡0f \equiv 0, so w∈Uw \in U. Hence D(z,δ)⊆UD(z,\delta) \subseteq U and UU is open.
  • Closed in Ω\Omega: let z0∈Ωz_0 \in \Omega be a limit of points zk∈Uz_k \in U. Since ff is continuous (holomorphic functions are continuous) and f≡0f \equiv 0 on a neighborhood of each zkz_k, we get f(zk)=0f(z_k) = 0 for all kk, hence f(z0)=lim⁡f(zk)=0f(z_0) = \lim f(z_k) = 0 by continuity. Moreover, since zk→z0z_k \to z_0 with zk≠z0z_k \ne z_0 eventually (or z0z_0 itself already in UU), z0z_0 is a zero of ff that is not isolated: every neighborhood of z0z_0 contains some zk∈Uz_k \in U, hence contains a whole disk of zeros, hence contains zeros of ff other than z0z_0. By the Lemma applied at z0z_0, f≡0f \equiv 0 on a disk around z0z_0, so z0∈Uz_0 \in U. Thus UU is closed in Ω\Omega.

Since Ω\Omega is connected and UU is a nonempty subset that is both open and closed in Ω\Omega, we conclude U=ΩU = \Omega. Hence f≡0f \equiv 0 on Ω\Omega, contradicting the hypothesis f≢0f \not\equiv 0. This contradiction proves that every zero of ff (when f≢0f \not\equiv 0) is isolated, and correspondingly Z(f)Z(f) has no accumulation point in Ω\Omega: if it did, that accumulation point would itself be a non-isolated zero of ff, which we have just ruled out. This proves Part (A). Part (B). Let h=f−gh = f - g, which is holomorphic on Ω\Omega (differences of holomorphic functions are holomorphic). The set S={z:f(z)=g(z)}=Z(h)S = \{z : f(z) = g(z)\} = Z(h), and by hypothesis SS has an accumulation point a∈Ωa \in \Omega. If h≢0h \not\equiv 0 on Ω\Omega, then by Part (A) every zero of hh is isolated — but aa is a zero of hh (since hh is continuous and zk→az_k \to a with h(zk)=0h(z_k) = 0 gives h(a)=0h(a) = 0) that is not isolated, since aa is an accumulation point of Z(h)=SZ(h) = S. This contradicts Part (A). Hence h≡0h \equiv 0 on Ω\Omega, i.e. f≡gf \equiv g on Ω\Omega. ■\blacksquare

∎
CorollaryWell-definedness of the order

If ff is holomorphic and not identically zero near aa with f(a)=0f(a) = 0, the order mm from the Definition is a well-defined positive integer, independent of any choices, and ff factors uniquely as

f(z)=(z−a)mϕ(z),ϕ holomorphic near a,ϕ(a)≠0.f(z) = (z-a)^m \phi(z), \qquad \phi \text{ holomorphic near } a, \quad \phi(a) \ne 0.
Proof

This is exactly the factorization constructed inside the proof of the Lemma above (the case where the contradiction was not reached, i.e. genuinely f≢0f \not\equiv 0 near aa): mm is uniquely determined as the index of the first nonvanishing Taylor coefficient, which is intrinsic to ff (Taylor coefficients are determined by f(n)(a)/n!f^{(n)}(a)/n!, not by any choice), and the factorization f(z)=(z−a)mϕ(z)f(z) = (z-a)^m \phi(z) with ϕ(a)=cm≠0\phi(a) = c_m \ne 0 was shown to hold on a neighborhood of aa. Uniqueness: if f(z)=(z−a)mϕ(z)=(z−a)m′ϕ′(z)f(z) = (z-a)^m\phi(z) = (z-a)^{m'}\phi'(z) with ϕ(a),ϕ′(a)≠0\phi(a),\phi'(a) \ne 0 and m<m′m < m' say, dividing by (z−a)m(z-a)^m and taking z→az \to a gives ϕ(a)=0\phi(a) = 0, a contradiction; so m=m′m = m', and then ϕ=ϕ′\phi = \phi' off aa, hence everywhere by continuity. ■\blacksquare

∎

Worked examples

Example 1: Computing the order of a zero from a non-obvious factorization. Let f(z)=ez−1−z−z22f(z) = e^{z} - 1 - z - \tfrac{z^2}{2}, holomorphic on all of C\mathbb{C}. We determine the order of its zero at a=0a = 0. The Taylor expansion of eze^z about 00 is ∑n=0∞zn/n!\sum_{n=0}^\infty z^n/n!, so

f(z)=∑n=0∞znn!−1−z−z22=∑n=3∞znn!=z36+z424+z5120+⋯ .f(z) = \sum_{n=0}^{\infty} \frac{z^n}{n!} - 1 - z - \frac{z^2}{2} = \sum_{n=3}^{\infty} \frac{z^n}{n!} = \frac{z^3}{6} + \frac{z^4}{24} + \frac{z^5}{120} + \cdots.

The coefficients c0=c1=c2=0c_0 = c_1 = c_2 = 0 and c3=1/6≠0c_3 = 1/6 \ne 0. By definition the order of the zero at 00 is m=3m = 3. Explicitly,

f(z)=z3ϕ(z),ϕ(z)=16+z24+z2120+⋯=∑k=0∞zk(k+3)!,f(z) = z^3 \phi(z), \qquad \phi(z) = \frac{1}{6} + \frac{z}{24} + \frac{z^2}{120} + \cdots = \sum_{k=0}^\infty \frac{z^k}{(k+3)!},

and ϕ(0)=1/6≠0\phi(0) = 1/6 \ne 0, matching the Corollary's factorization. Since f≢0f \not\equiv 0 (it's a nonzero entire function — its Taylor series has nonzero coefficients), Part (A) guarantees this zero at 00 is isolated: indeed one can check directly that f(z)=0  ⟺  ez=1+z+z2/2f(z) = 0 \iff e^z = 1 + z + z^2/2, and by uniqueness-of-power-series arguments (or just direct estimate) no other zero lies in a small punctured disk around 00, consistent with the theorem though we do not need to verify this by hand — the theorem guarantees it structurally once we know f≢0f\not\equiv0.

Example 2: Applying the Identity Theorem to prove a global functional identity from local data. Claim: if ff is entire (holomorphic on all of C\mathbb{C}) and satisfies

f(x)=sin⁡(x)cosh⁡(x) for every x in the real interval (−0.01, 0.01),f(x) = \sin(x)\cosh(x) \quad \text{ for every } x \text{ in the real interval } (-0.01,\, 0.01),

then in fact f(z)=sin⁡(z)cosh⁡(z)f(z) = \sin(z)\cosh(z) for every z∈Cz \in \mathbb{C} — not just near 00, and not just on the reals. Proof of claim. Let g(z)=sin⁡(z)cosh⁡(z)g(z) = \sin(z)\cosh(z), an entire function (product of entire functions). Let S={z∈C:f(z)=g(z)}S = \{z \in \mathbb{C} : f(z) = g(z)\}. By hypothesis, S⊇(−0.01,0.01)⊂R⊂CS \supseteq (-0.01, 0.01) \subset \mathbb{R} \subset \mathbb{C}, an interval of real numbers. Every point of this interval, e.g. a=0a = 0, is an accumulation point of SS (indeed the whole interval accumulates at every one of its points). Since Ω=C\Omega = \mathbb{C} is a domain (open and connected) and f,gf, g are both holomorphic on Ω\Omega with SS having an accumulation point in Ω\Omega, the Identity Theorem gives f≡gf \equiv g on all of C\mathbb{C}. ■\blacksquare This example demonstrates the theorem's real power: an equality verified on a minuscule real sub-interval — data of "measure zero" in the plane, and utterly negligible compared to C\mathbb{C} — completely determines the entire function everywhere. This is precisely the mechanism used to define complex extensions of real-analytic functions (like sin⁡,cos⁡,exp⁡\sin, \cos, \exp) uniquely, and it is the starting mechanism of Analytic Continuation.

CounterexampleWhy connectedness of the domain cannot be dropped

Consider the domain

Ω=D(0,1)∪D(3,1)⊂C,\Omega = D(0,1) \cup D(3,1) \subset \mathbb{C},

the union of two disjoint open unit disks centered at 00 and at 33 — an open set, but not connected (it splits into two components with no path between them). Define f:Ω→Cf : \Omega \to \mathbb{C} by

f(z)={0z∈D(0,1)1z∈D(3,1).f(z) = \begin{cases} 0 & z \in D(0,1) \\ 1 & z \in D(3,1). \end{cases}

This ff is holomorphic on Ω\Omega: on each component it is a constant function, and constants are holomorphic; holomorphy is a local condition, checked in a neighborhood of each point, and every point of Ω\Omega lies entirely inside one component or the other (the components being open and disjoint), so ff is complex-differentiable (with derivative 00) at every point of Ω\Omega. Now compare ff to g≡0g \equiv 0 on Ω\Omega. The set S={z∈Ω:f(z)=g(z)}=D(0,1)S = \{z \in \Omega : f(z) = g(z)\} = D(0,1), the entire left disk. This set certainly has accumulation points in Ω\Omega — every point of D(0,1)D(0,1) is one, e.g. a=0a = 0. Yet f≢gf \not\equiv g on Ω\Omega: they disagree everywhere on D(3,1)D(3,1), where f≡1≠0=gf \equiv 1 \ne 0 = g. This shows every hypothesis of the Identity Theorem except connectedness can hold — Ω\Omega open, f,gf,g holomorphic on Ω\Omega, SS has an accumulation point in Ω\Omega — and still the conclusion f≡gf \equiv g on Ω\Omega fails. The proof's key step (the open-closed argument forcing U=ΩU = \Omega) uses connectedness essentially: with two components, UU can equal one component and stop there, being simultaneously open and closed in Ω\Omega without covering everything. This pinpoints exactly why domains (connected opens) rather than merely open sets are the natural habitat for the theorem.

Scope, edge cases & common traps

SituationWhat holdsWhat can fail
Ω\Omega connected, f,gf,g holomorphic, S={f=g}S=\{f=g\} has accumulation point in Ω\Omegaf≡gf \equiv g on Ω\Omega— (this is exactly the theorem)
Ω\Omega disconnected—Conclusion can fail entirely; see counterexample
Accumulation point outside Ω\Omega (e.g. on the boundary)Nothing guaranteedf,gf, g may still differ inside Ω\Omega
SS merely infinite, but with no accumulation point in Ω\Omega (e.g. accumulates only at ∂Ω\partial\Omega or at ∞\infty)Nothing guaranteedClassic trap: infinitude of agreement points is not enough
ff real-analytic instead of holomorphicTheorem failsReal-analytic case is subtler; some real-analytic functions still satisfy an analogous identity theorem if defined by convergent power series, but generic real C∞C^\infty or merely real-analytic-looking constructions can fail without the complex structure
Order of zero mmWell-defined positive integer, intrinsic to ff and aaDo not confuse with "vanishing to order mm" in real calculus, which permits non-integer or non-isolated behavior

The mathematician behind it

Origins · 1 minKarl Weierstrass and how this was discovered
KW
Karl Weierstrass
1815–1897 · Rigorization of complex analysis and power-series foundations

Karl Weierstrass was a German mathematician who, alongside Cauchy and Riemann, built one of the three classical foundations of complex analysis — the power-series approach, where a holomorphic function is defined as one locally representable by a convergent power series, and global behavior is understood via analytic continuation along chains of overlapping disks. This viewpoint makes the Identity Theorem almost definitional: it is the natural rigidity of a theory built from power series glued together, and it directly foreshadows Weierstrass's broader program of analytic continuation, elements, and the monodromy phenomena that later authors formalized.

History
- 1815 — Karl Weierstrass born in Ostenfelde, Westphalia.
- 1841–1842 — Weierstrass, while an unknown schoolteacher, develops the power-series ("Weierstrassian") approach to complex function theory in unpublished manuscripts.
- 1850s–1860s — Weierstrass at Berlin systematizes rigorous foundations of analysis (uniform convergence, epsilon-delta rigor) and the theory of analytic continuation via power series chains.
- 1886 — Weierstrass's lectures at Berlin, widely circulated among students, cement the "elements of an analytic function" viewpoint underlying the Identity Theorem.
- 1897 — Karl Weierstrass dies in Berlin; his approach becomes one of the standard modern treatments of complex analysis alongside Cauchy's integral methods and Riemann's geometric methods.
Why does the Identity Theorem fail for real-differentiable or even real-C∞C^\infty functions?

Real differentiability, even to infinite order, only encodes local Taylor data that need not reconstruct the function globally — the Taylor series may fail to converge to the function at all (as with e−1/x2e^{-1/x^2}), or the function may not even be represented by any convergent power series near the point. Complex differentiability is far stronger: a function differentiable once in a complex-analytic sense on an open set is automatically representable by a convergent power series there (a consequence of Cauchy's integral formula machinery, established in Taylor Series of Holomorphic Functions). This "power series = the function itself" property is precisely what makes the Identity Theorem's rigidity possible; without it, local vanishing data simply doesn't propagate.

Does the Identity Theorem mean holomorphic functions can never vanish on an infinite set without being identically zero?

Not quite — it means the zero set of a nonzero holomorphic function on a domain cannot have an accumulation point inside that domain. It can certainly be infinite: for example sin⁡(z)\sin(z) has infinitely many zeros at z=nπz = n\pi for every integer nn, but these zeros march off to infinity and never accumulate at any finite point of C\mathbb{C}, so there's no contradiction. The theorem constrains where infinite zero sets can cluster, not whether they can be infinite in cardinality.

How is the order of a zero used in practice?

The order of a zero is the local multiplicity that later feeds into global zero-counting: The Argument Principle and Rouche's Theorem shows that a contour integral of f′/ff'/f around a closed curve computes exactly the sum of orders of zeros enclosed (minus orders of poles, with sign). It also governs local mapping behavior: near a zero of order mm, ff behaves like a degree-mm map w↦wmw \mapsto w^m up to a change of coordinates, so small circles around the zero get wrapped mm times around 00 in the image, which is the geometric content illustrated in the intuition figure.

Is the Identity Theorem the same as analytic continuation?

They are closely related but distinct. The Identity Theorem is a uniqueness statement: if an extension exists, it is unique on a connected domain. Analytic Continuation is the existence-and-construction process: given a function defined on a small region, one builds a holomorphic extension to a larger region by chaining overlapping power series disks (or other methods). The Identity Theorem is what guarantees that however you continue the function — along whichever path or method — you cannot get two different answers on the same simply-laid-out domain, though on multiply-connected domains continuation along different paths can produce genuinely different branches (monodromy), which is a separate and subtler phenomenon.

Can the Identity Theorem be applied when Ω\Omega is all of C\mathbb{C} (entire functions)?

Yes, and this is one of its most common uses. Since C\mathbb{C} is connected, any two entire functions that agree on a set with an accumulation point anywhere in C\mathbb{C} — a convergent sequence, an interval, an arc, a disk, anything with a cluster point — must be identical as entire functions. This is exactly how one proves that the complex extensions of sin⁡\sin, cos⁡\cos, exp⁡\exp, defined via power series, are the unique entire extensions of their real-variable counterparts, since they agree with the real functions on the real axis, which has plenty of accumulation points.

What's the difference between "order of a zero" and "order of vanishing" used loosely in calculus?

In elementary real calculus, people sometimes say a function "vanishes to order mm" if its first m−1m-1 derivatives vanish at a point, without asserting anything about whether higher derivatives are nonzero or whether the function is nonzero nearby — this can apply to functions like xmsin⁡(1/x)x^m \sin(1/x)-type constructions that don't fit a clean power series picture. In complex analysis, "order of a zero" is much more rigid: it is the unique positive integer mm from the first nonvanishing Taylor coefficient, and it comes bundled with the guarantee that f=(z−a)mϕ(z)f = (z-a)^m \phi(z) with ϕ\phi holomorphic and ϕ(a)≠0\phi(a) \ne 0 — a clean, checkable factorization with no pathological exceptions, precisely because of analyticity.

Does the theorem require ff and gg to be defined on exactly the same domain?

Yes — the theorem as stated requires both ff and gg to be holomorphic on the same connected open set Ω\Omega, and the accumulation point of their agreement set must lie inside Ω\Omega. If they are defined on different domains that merely overlap, one typically applies the theorem to the intersection (if it is itself connected and open) or invokes analytic continuation arguments more carefully — the naive same-domain statement does not automatically extend to functions defined on different or disconnected supersets without extra care about connectedness, exactly as the counterexample demonstrates.

Before this

Where this leads