The Maximum Modulus Principle

Suppose a function is holomorphic on a disk. Suppose you know its values only on the boundary circle. Can the function's absolute value suddenly spike to a new record somewhere in the interior? Could it go higher than anything it achieves on the boundary? The Maximum Modulus Principle answers this with a resounding no. The reason is not a numerical coincidence. Instead, it reflects a deep rigidity built into holomorphy itself. A holomorphic function's local behavior is entirely dictated by an averaging property. This property is inherited from Cauchy's Integral Formula. This rigidity is one of the sharpest ways complex analysis diverges from real analysis. A smooth real function of two variables can easily have an interior local maximum of its magnitude. Think of a bump function. A holomorphic function cannot do this, unless it is constant. The principle transforms a seemingly delicate analytic question into a soft topological statement. The question is: where is ∣f∣|f| largest? The answer: nowhere in the interior, unless ff is trivial. It does so via the mean value property. Every holomorphic function satisfies this property on circles. The importance of this result radiates outward. It gives one of the cleanest proofs that a polynomial's roots control its growth. This works via the fundamental theorem of algebra's analytic proof. It underlies estimates on how fast holomorphic maps can distort distances. This is the Schwarz Lemma. It is also the analytic engine behind rigidity theorems for conformal maps of the disk. Consequences include the Schwarz–Pick lemma and uniqueness theorems for automorphisms of the disk. They also include a rich family of Phragmén–Lindelöf type extensions to unbounded domains. This article states and proves the Maximum Modulus Principle in full generality on domains (open connected sets). It derives the Schwarz Lemma as its centerpiece application. It works through concrete examples showing how sharp the bound is. We close by tracing the principle's fingerprints across Conformal Mappings, Mobius Transformations, and further afield.

Key takeaways
  • A non-constant holomorphic function on a domain has no interior local maximum of ∣f∣|f|.
  • The proof rests on the mean value property: f(z0)f(z_0) equals the average of ff over any circle centered at z0z_0 inside the domain.
  • On a bounded domain with continuous boundary values, ∣f∣|f| attains its maximum only on the boundary, never strictly inside.
  • The Schwarz Lemma sharpens this for self-maps of the disk fixing the origin, forcing ∣f(z)∣≤∣z∣|f(z)| \le |z| and ∣f′(0)∣≤1|f'(0)| \le 1.
  • Equality anywhere in Schwarz's Lemma forces ff to be a rotation — a rigidity phenomenon typical of holomorphic maps.
  • Key idea: holomorphy's averaging property makes ∣f∣|f| behave like a subharmonic function with no room for interior peaks.

Where this lives today

The Maximum Modulus Principle and Schwarz Lemma are load-bearing walls of geometric function theory. In hyperbolic geometry, the Schwarz–Pick Lemma shows holomorphic self-maps of the disk are exactly the maps that do not increase the hyperbolic (Poincaré) metric, giving an analytic route into non-Euclidean geometry and laying groundwork for the study of Mobius Transformations as isometries. In the classification of Conformal Mappings, the rigidity clause of Schwarz's Lemma is precisely what proves that automorphisms of the disk are Möbius maps, a key step toward the Riemann Mapping Theorem's uniqueness statement. In operator theory and control engineering, contraction properties analogous to Schwarz's Lemma appear in the study of Schur functions (bounded analytic functions on the disk), used in signal processing and stability analysis of linear systems. In several complex variables and functional analysis, maximum modulus type arguments extend to Banach-algebra-valued holomorphic functions, underpinning parts of spectral theory. The principle also gives the standard analytic proof that non-constant entire functions bounded on the whole plane cannot exist except constants — closely related to, though logically distinct from, Liouville's Theorem and Applications. In numerical analysis, maximum modulus estimates control error propagation for holomorphic approximants, and in complexity theory and cryptography, bounds on holomorphic generating functions sometimes invoke exactly this style of boundary-controls-interior reasoning.

Ask the tutor: How does the Schwarz–Pick Lemma generalize this to maps that don't fix the origin, using the hyperbolic metric?

DefinitionMaximum Modulus Principle — setup

Let D⊆CD \subseteq \mathbb{C} be a domain, meaning a nonempty, open, connected subset of C\mathbb{C}. Let f:D→Cf : D \to \mathbb{C} be holomorphic. We say ∣f∣|f| has a local maximum at z0∈Dz_0 \in D if there exists r>0r > 0 with B(z0,r)‾⊂D\overline{B(z_0,r)} \subset D such that

∣f(z)∣≤∣f(z0)∣for all z∈B(z0,r).|f(z)| \le |f(z_0)| \quad \text{for all } z \in B(z_0, r).

The local maximum is called strict in a punctured sense if ∣f(z)∣=∣f(z0)∣|f(z)| = |f(z_0)| forces z=z0z = z_0 within that ball — though as the theorem shows, this case cannot occur unless ff is constant.

Intuition

The heart of the matter is the mean value property. If ff is holomorphic on a disk containing B(z0,r)‾\overline{B(z_0,r)}, Cauchy's Integral Formula gives

f(z0)=12πi∮∣z−z0∣=rf(z)z−z0 dz=12π∫02πf(z0+reiθ) dθ.f(z_0) = \frac{1}{2\pi i} \oint_{|z - z_0| = r} \frac{f(z)}{z - z_0}\, dz = \frac{1}{2\pi} \int_0^{2\pi} f(z_0 + re^{i\theta})\, d\theta.

So f(z0)f(z_0) is literally the average of ff over the circle. Averages can never exceed the maximum of the quantity being averaged, and — crucially — an average can only equal the maximum value if the quantity is constant on the circle. This is the entire mechanism: a value that is an average of its neighbors cannot be a strict peak unless everything nearby is frozen at that same value. Picture ∣f∣|f| as a landscape over the domain. Holomorphy forces this landscape to have no true "summits" in the interior — only saddle-like or flat behavior, with all genuine peaks pushed to the boundary. This is utterly different from generic smooth surfaces, where bumps are common. The rigidity comes entirely from the complex structure: ff is not just smooth, it is complex-differentiable, and that extra structure is what forbids interior maxima. The Schwarz Lemma specializes this picture to the unit disk: if a holomorphic self-map of the disk fixes the center, the maximum modulus principle — applied cleverly to f(z)/zf(z)/z — shows the map cannot expand distances from the origin. Geometrically, holomorphic self-maps of the disk fixing 00 can only contract or rotate, never stretch.

Figure
Mean value property: f(z0)f(z_0) is the average over the circle, so ∣f(z0)∣|f(z_0)| cannot exceed the max on the circle
z₀ f-value large here θ=0 smaller |f| here f(z₀) = (1/2π) ∫₀^{2π} f(z₀ + re^{iθ}) dθ Average over circle ⇒ |f(z₀)| ≤ max on circle, equality only if f constant on it
TheoremMaximum Modulus Principle

Let D⊆CD \subseteq \mathbb{C} be a domain and let f:D→Cf: D \to \mathbb{C} be holomorphic. Suppose there exists z0∈Dz_0 \in D such that

∣f(z)∣≤∣f(z0)∣for all z in some open ball B(z0,r)⊆D.|f(z)| \le |f(z_0)| \quad \text{for all } z \text{ in some open ball } B(z_0, r) \subseteq D.

Then ff is constant on DD.

Corollary (boundary maximum form). Let DD be a bounded domain and let ff be continuous on D‾\overline{D} and holomorphic on DD. Then

max⁡z∈D‾∣f(z)∣=max⁡z∈∂D∣f(z)∣,\max_{z \in \overline{D}} |f(z)| = \max_{z \in \partial D} |f(z)|,

and if ∣f∣|f| attains this maximum at any interior point of DD, then ff is constant on DD.

LemmaMean Value Property

Let ff be holomorphic on an open set containing B(z0,r)‾\overline{B(z_0, r)}. Then

f(z0)=12π∫02πf(z0+reiθ) dθ.f(z_0) = \frac{1}{2\pi} \int_0^{2\pi} f(z_0 + re^{i\theta})\, d\theta.
Proof

This is a direct consequence of Cauchy's Integral Formula. Since ff is holomorphic on an open neighborhood of the closed disk B(z0,r)‾\overline{B(z_0,r)}, Cauchy's Integral Formula applied to the circle γ(θ)=z0+reiθ\gamma(\theta) = z_0 + re^{i\theta}, θ∈[0,2π]\theta \in [0, 2\pi], gives

f(z0)=12πi∮γf(z)z−z0 dz.f(z_0) = \frac{1}{2\pi i} \oint_{\gamma} \frac{f(z)}{z - z_0}\, dz.

Parametrize: z=z0+reiθz = z_0 + re^{i\theta}, so dz=ireiθ dθdz = ire^{i\theta}\, d\theta and z−z0=reiθz - z_0 = re^{i\theta}. Substituting,

f(z0)=12πi∫02πf(z0+reiθ)reiθ⋅ireiθ dθ=12π∫02πf(z0+reiθ) dθ,f(z_0) = \frac{1}{2\pi i} \int_0^{2\pi} \frac{f(z_0 + re^{i\theta})}{re^{i\theta}} \cdot ire^{i\theta}\, d\theta = \frac{1}{2\pi} \int_0^{2\pi} f(z_0 + re^{i\theta})\, d\theta,

as claimed. ■\blacksquare

∎
Proof

(Main theorem, via the Mean Value Property and the identity theorem.)

Step 1: Local constancy near z0z_0. Let M=∣f(z0)∣M = |f(z_0)|. By hypothesis, ∣f(z)∣≤M|f(z)| \le M for all z∈B(z0,r)z \in B(z_0, r). Fix any ρ\rho with 0<ρ<r0 < \rho < r. By the Mean Value Property lemma above,

f(z0)=12π∫02πf(z0+ρeiθ) dθ.f(z_0) = \frac{1}{2\pi} \int_0^{2\pi} f(z_0 + \rho e^{i\theta})\, d\theta.

Taking absolute values and using the triangle inequality for integrals,

M=∣f(z0)∣≤12π∫02π∣f(z0+ρeiθ)∣ dθ≤12π∫02πM dθ=M.M = |f(z_0)| \le \frac{1}{2\pi} \int_0^{2\pi} |f(z_0 + \rho e^{i\theta})|\, d\theta \le \frac{1}{2\pi} \int_0^{2\pi} M\, d\theta = M.

So every inequality in this chain is actually an equality. In particular,

12π∫02π(M−∣f(z0+ρeiθ)∣) dθ=0,\frac{1}{2\pi} \int_0^{2\pi} \big(M - |f(z_0 + \rho e^{i\theta})|\big)\, d\theta = 0,

and since the integrand M−∣f(z0+ρeiθ)∣M - |f(z_0 + \rho e^{i\theta})| is continuous and non-negative (because ∣f(z)∣≤M|f(z)|\le M on the ball), an integral of a non-negative continuous function that vanishes must have the function identically zero. Hence

∣f(z0+ρeiθ)∣=Mfor all θ∈[0,2π].|f(z_0 + \rho e^{i\theta})| = M \quad \text{for all } \theta \in [0, 2\pi].

This holds for every ρ∈(0,r)\rho \in (0, r), so ∣f(z)∣=M|f(z)| = M for all z∈B(z0,r)z \in B(z_0, r).

Step 2: Constant modulus forces constant function. We now show that if ∣f∣|f| is constant on the open ball B(z0,r)B(z_0, r), then ff itself is constant there. If M=0M = 0, then f≡0f \equiv 0 on the ball and we are done. Assume M>0M > 0, so ff is nowhere zero on B(z0,r)B(z_0, r). Write f=u+ivf = u + iv in real and imaginary parts (harmonic real-valued functions since ff is holomorphic). Since ∣f∣2=u2+v2=M2|f|^2 = u^2 + v^2 = M^2 is constant, differentiate with respect to xx and yy:

uux+vvx=0,uuy+vvy=0.u u_x + v v_x = 0, \qquad u u_y + v v_y = 0.

By the Cauchy–Riemann equations, ux=vyu_x = v_y and uy=−vxu_y = -v_x. Substitute into the second equation:

u(−vx)+vux=0  ⟹  vux−uvx=0.u(-v_x) + v u_x = 0 \implies v u_x - u v_x = 0.

Combined with uux+vvx=0u u_x + v v_x = 0, we have a linear system in (ux,vx)(u_x, v_x):

(uvv−u)(uxvx)=(00).\begin{pmatrix} u & v \\ v & -u \end{pmatrix} \begin{pmatrix} u_x \\ v_x \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix}.

The determinant of the coefficient matrix is −u2−v2=−M2≠0-u^2 - v^2 = -M^2 \ne 0 since M>0M > 0. Hence the only solution is ux=vx=0u_x = v_x = 0. By the Cauchy–Riemann equations again, uy=−vx=0u_y = -v_x = 0 and vy=ux=0v_y = u_x = 0 as well. So all partial derivatives of uu and vv vanish identically on B(z0,r)B(z_0, r), meaning uu and vv are constant there, so ff is constant on B(z0,r)B(z_0, r).

Step 3: Globalizing via the Identity Theorem. We have shown ff agrees with the constant function c:=f(z0)c := f(z_0) on the open ball B(z0,r)⊆DB(z_0, r) \subseteq D. Since DD is connected and f−cf - c is holomorphic on DD vanishing on a set with an accumulation point in DD (indeed, on an entire open ball), the Identity Theorem for holomorphic functions — itself a consequence of the fact that zero sets of non-identically-zero holomorphic functions are discrete, provable from power series expansions guaranteed by Cauchy's Integral Formula — implies f≡cf \equiv c on all of DD. This completes the proof that ff is constant on DD. ■\blacksquare

Proof of the Corollary. Since D‾\overline{D} is compact (as DD is bounded) and ff is continuous on D‾\overline{D}, ∣f∣|f| attains its maximum MM at some point p∈D‾p \in \overline{D}. If p∈∂Dp \in \partial D, we are done: the maximum over D‾\overline D is attained on the boundary, hence

max⁡D‾∣f∣=∣f(p)∣≤max⁡∂D∣f∣≤max⁡D‾∣f∣,\max_{\overline D}|f| = |f(p)| \le \max_{\partial D}|f| \le \max_{\overline D}|f|,

forcing equality throughout. If instead p∈Dp \in D (interior), then pp is an interior point at which ∣f(z)∣≤∣f(p)∣|f(z)| \le |f(p)| for all zz in a ball around pp inside DD (since DD is open) — exactly the hypothesis of the main theorem. Hence ff is constant on DD, and by continuity constant on D‾\overline D, so trivially max⁡D‾∣f∣=max⁡∂D∣f∣\max_{\overline D}|f| = \max_{\partial D}|f| as well. ■\blacksquare

∎
CorollaryMinimum Modulus Principle

Let DD be a domain and ff holomorphic and zero-free on DD. If ∣f∣|f| attains a local minimum at an interior point of DD, then ff is constant on DD.

Proof

Since ff is zero-free on DD, 1/f1/f is holomorphic on DD. A local minimum of ∣f∣|f| at z0z_0 is exactly a local maximum of ∣1/f∣|1/f| at z0z_0 (because ∣1/f(z)∣=1/∣f(z)∣|1/f(z)| = 1/|f(z)|, and minimizing ∣f∣|f| is the same as maximizing 1/∣f∣1/|f|). Applying the Maximum Modulus Principle to 1/f1/f shows 1/f1/f is constant on DD, hence so is ff. ■\blacksquare

∎

Worked examples

Example 1: A polynomial on an annular region, not a disk. Let f(z)=z3−2z+3if(z) = z^3 - 2z + 3i and let DD be the open annulus {1<∣z∣<2}\{1 < |z| < 2\}, which is a domain (open and connected, though not simply connected). Note that the maximum modulus principle as proved applies to any domain — connectivity is all that was used, not simple-connectivity. We estimate max⁡∣z∣=2∣f(z)∣\max_{|z|=2}|f(z)| using the triangle inequality:

∣f(z)∣≤∣z∣3+2∣z∣+3=8+4+3=15when ∣z∣=2.|f(z)| \le |z|^3 + 2|z| + 3 = 8 + 4 + 3 = 15 \quad \text{when } |z| = 2.

And on ∣z∣=1|z| = 1:

∣f(z)∣≤1+2+3=6.|f(z)| \le 1 + 2 + 3 = 6.

Since ff is holomorphic on the closed annulus D‾={1≤∣z∣≤2}\overline D = \{1 \le |z| \le 2\} (indeed entire), the corollary tells us the maximum of ∣f∣|f| on D‾\overline D occurs on the boundary ∂D={∣z∣=1}∪{∣z∣=2}\partial D = \{|z|=1\} \cup \{|z|=2\}, not in the interior. Our bound 1515 (attained in the worst case estimate on the outer circle) dominates the inner circle's bound of 66, so we conclude

max⁡D‾∣f(z)∣≤15,\max_{\overline D} |f(z)| \le 15,

with the true maximum guaranteed to occur somewhere on ∣z∣=2|z| = 2 specifically if that bound is close to sharp (a finer argument using the actual angular dependence would pin down the exact maximizing angle, but the qualitative conclusion — no interior spike — is guaranteed regardless of the precise constant). This example underscores that the theorem needs no simple-connectivity, no disk shape — only that DD be a domain (connected open set) and ff holomorphic there.

Example 2: The Schwarz Lemma, proved via the Maximum Modulus Principle.

TheoremSchwarz Lemma

Let f:D→Df: \mathbb{D} \to \mathbb{D} be holomorphic on the open unit disk D={∣z∣<1}\mathbb{D} = \{|z| < 1\}, with f(0)=0f(0) = 0. Then: 1. ∣f(z)∣≤∣z∣|f(z)| \le |z| for all z∈Dz \in \mathbb{D}; 2. ∣f′(0)∣≤1|f'(0)| \le 1; 3. If equality holds in (1) for some z≠0z \ne 0, or if equality holds in (2), then f(z)=λzf(z) = \lambda z for some constant λ\lambda with ∣λ∣=1|\lambda| = 1.

Proof

Since ff is holomorphic on D\mathbb{D} and f(0)=0f(0) = 0, the power series of ff about 00 has no constant term:

f(z)=a1z+a2z2+⋯(∣z∣<1).f(z) = a_1 z + a_2 z^2 + \cdots \qquad (|z| < 1).

Define

g(z)={f(z)/zz≠0f′(0)=a1z=0.g(z) = \begin{cases} f(z)/z & z \ne 0 \\ f'(0) = a_1 & z = 0. \end{cases}

The function gg is holomorphic on all of D\mathbb{D}: away from 00 it is a quotient of holomorphic functions with nonvanishing denominator, and at 00 the removable singularity is resolved because f(z)/z=a1+a2z+⋯f(z)/z = a_1 + a_2 z + \cdots converges to a1a_1 as z→0z \to 0, matching the defined value; formally, gg's power series a1+a2z+a3z2+⋯a_1 + a_2 z + a_3 z^2 + \cdots has the same radius of convergence as ff's, namely at least 11.

Bounding gg on circles ∣z∣=ρ<1|z| = \rho < 1. For 0<ρ<10 < \rho < 1 and ∣z∣=ρ|z| = \rho,

∣g(z)∣=∣f(z)∣∣z∣=∣f(z)∣ρ≤1ρ,|g(z)| = \frac{|f(z)|}{|z|} = \frac{|f(z)|}{\rho} \le \frac{1}{\rho},

since ff maps into D\mathbb{D}, so ∣f(z)∣<1|f(z)| < 1 for all z∈Dz \in \mathbb D (in fact ∣f(z)∣≤1|f(z)|\le 1 suffices, we even have strict inequality since ff maps into the open disk). Since gg is holomorphic on the closed disk B(0,ρ)‾⊂D\overline{B(0,\rho)} \subset \mathbb D (for ρ<1\rho<1) and continuous there, the Maximum Modulus Principle (corollary form) gives

∣g(z)∣≤max⁡∣w∣=ρ∣g(w)∣≤1ρfor all ∣z∣≤ρ.|g(z)| \le \max_{|w|=\rho} |g(w)| \le \frac{1}{\rho} \quad \text{for all } |z| \le \rho.

Fix any z0∈Dz_0 \in \mathbb D. For every ρ\rho with ∣z0∣<ρ<1|z_0| < \rho < 1, the inequality ∣g(z0)∣≤1/ρ|g(z_0)| \le 1/\rho holds. Letting ρ→1−\rho \to 1^-, we obtain

∣g(z0)∣≤1.|g(z_0)| \le 1.

Since z0∈Dz_0 \in \mathbb D was arbitrary, ∣g(z)∣≤1|g(z)| \le 1 for all z∈Dz \in \mathbb D. Translating back:

  • For z≠0z \ne 0: ∣f(z)∣/∣z∣≤1|f(z)|/|z| \le 1, i.e. ∣f(z)∣≤∣z∣|f(z)| \le |z|. This proves (1) (and trivially holds at z=0z=0 too, since both sides are 00).
  • At z=0z = 0: ∣g(0)∣=∣f′(0)∣≤1|g(0)| = |f'(0)| \le 1, proving (2).

Equality case. Suppose either ∣f(z1)∣=∣z1∣|f(z_1)| = |z_1| for some z1≠0z_1 \ne 0, or ∣f′(0)∣=1|f'(0)| = 1. In either case, ∣g∣|g| attains the value 11 at some point of D\mathbb D (z1z_1 in the first case, 00 in the second), and we've shown ∣g∣≤1|g| \le 1 everywhere on D\mathbb D. So ∣g∣|g| attains an interior maximum (value 11, which is the global sup) at that point. By the Maximum Modulus Principle (main theorem: a local, hence global, interior maximum of ∣g∣|g| forces gg constant), gg is constant on D\mathbb D. Write g≡λg \equiv \lambda where ∣λ∣=1|\lambda| = 1 (since ∣g∣≡1|g|\equiv 1 there by continuity of the constant value). Then f(z)=zg(z)=λzf(z) = z g(z) = \lambda z for all z∈Dz \in \mathbb D, proving (3). ■\blacksquare

∎
CorollaryRigidity of disk automorphisms fixing 00

If f:D→Df: \mathbb D \to \mathbb D is a holomorphic bijection (automorphism) with f(0)=0f(0) = 0, then f(z)=λzf(z) = \lambda z for some ∣λ∣=1|\lambda| = 1; that is, ff is a rotation.

Proof

Apply the Schwarz Lemma to ff: since f(D)⊆Df(\mathbb D) \subseteq \mathbb D and f(0)=0f(0)=0, we get ∣f(z)∣≤∣z∣|f(z)| \le |z| for all zz. Apply the Schwarz Lemma to f−1f^{-1}, which is also holomorphic (a standard fact for bijective holomorphic maps with holomorphic inverse, provable via the inverse function theorem for holomorphic maps once one knows f′≠0f' \ne 0, which follows since ff is injective) and satisfies f−1(0)=0f^{-1}(0) = 0, f−1:D→Df^{-1}: \mathbb D \to \mathbb D. This gives ∣f−1(w)∣≤∣w∣|f^{-1}(w)| \le |w| for all w∈Dw \in \mathbb D. Substituting w=f(z)w = f(z): ∣z∣=∣f−1(f(z))∣≤∣f(z)∣|z| = |f^{-1}(f(z))| \le |f(z)|. Combining ∣f(z)∣≤∣z∣|f(z)| \le |z| and ∣z∣≤∣f(z)∣|z| \le |f(z)| gives ∣f(z)∣=∣z∣|f(z)| = |z| for all z∈Dz \in \mathbb D. In particular equality holds in the Schwarz Lemma at every nonzero zz, so by the equality clause, f(z)=λzf(z) = \lambda z for a unimodular constant λ\lambda. ■\blacksquare

∎
Figure
Schwarz Lemma: the region ∣f(z)∣≤∣z∣|f(z)|\le|z| traps the graph of ∣f∣|f| against the diagonal
|z| |f(z)| y = |z| (bound line) graph of |f(z)| stays below diagonal 0 |f(z₀)| < |z₀| (strict, generic case)
CounterexampleWhy connectivity/domain hypothesis and boundedness cannot be dropped

Dropping connectivity. Let D=B(0,1)∪B(5,1)D = B(0,1) \cup B(5,1), a disjoint union of two disks — open but not connected, so not a domain. Define ff to equal 11 (constant) on B(0,1)B(0,1) and equal 12\frac{1}{2} (a different constant) on B(5,1)B(5,1). This ff is holomorphic on DD (holomorphic on each connected component separately is sufficient since holomorphy is a local condition). At the point z0=5∈B(5,1)z_0 = 5 \in B(5,1), we have ∣f(z)∣=1/2|f(z)| = 1/2 for all zz near z0z_0, an interior "local maximum" equal to the local constant value — this is consistent, not a counterexample to the local statement. But the global conclusion that ff is constant on all of DD fails: ff takes two different values (11 and 1/21/2) on the two components. The proof's Step 3 (globalizing via the Identity Theorem) crucially used connectedness; without it, local constancy near one point says nothing about far-away components.

Dropping boundedness of the domain (for the corollary). Consider D=CD = \mathbb{C} (unbounded) and f(z)=ezf(z) = e^z. There is no boundary to speak of, and indeed ∣f(z)∣=eRe⁡(z)|f(z)| = e^{\operatorname{Re}(z)} is unbounded on C\mathbb{C}: taking z=nz = n real and letting n→∞n \to \infty shows ∣f∣|f| has no global maximum at all. The corollary's compactness argument (extracting a maximum of a continuous function on a compact set D‾\overline D) completely breaks down without boundedness — there may be no maximum to locate on the boundary because there may be no finite maximum whatsoever. This is precisely why extensions to unbounded domains (Phragmén–Lindelöf theorems) require extra growth hypotheses at infinity to compensate for the lost compactness.

Scope, edge cases & common traps

SituationWhat holdsWhat fails / caution
DD a domain (open, connected), ff holomorphicNo interior local max of ∣f∣|f| unless ff constantRequires connectedness for the global conclusion
DD bounded, ff continuous on D‾\overline D, holomorphic on DDmax⁡D‾∣f∣=max⁡∂D∣f∣\max_{\overline D}|f| = \max_{\partial D}|f|Needs DD bounded for compactness of D‾\overline D
ff merely harmonic (real-valued)Maximum principle for harmonic functions holds separatelyThe complex proof here (Cauchy–Riemann step) is specific to holomorphic f=u+ivf=u+iv; real harmonic functions need their own mean value argument
ff real-valued and holomorphicForces ff constant (Cauchy–Riemann collapse)A common trap: a real-valued non-constant "holomorphic" function does not exist on a domain
Domain not simply connected (e.g. annulus)Principle still holdsSimple-connectivity is not a hypothesis of this theorem, unlike some other complex-analytic tools
Schwarz Lemma without f(0)=0f(0)=0Conclusion ∣f(z)∣≤∣z∣|f(z)|\le |z| can failMust first compose with a Möbius transformation sending the relevant point to 00; this is the content of the Schwarz–Pick lemma

The mathematician behind it

Origins · 1 minHermann Amandus Schwarz and how this was discovered
HA
Hermann Amandus Schwarz
1843–1921 · German mathematician, University of Berlin, Göttingen

Schwarz was a student and successor of Karl Weierstrass, inheriting Weierstrass's insistence on rigor in complex and real analysis. He worked on minimal surfaces, conformal mapping (the Schwarz–Christoffel formula for mapping polygons), and the calculus of variations. His lemma, published in the context of studying conformal self-maps of the disk, became a cornerstone technique reused across geometric function theory, hyperbolic geometry, and later in the Schwarz–Pick metric formulation of hyperbolic distance.

History
- 1851: Riemann's dissertation introduces holomorphic mapping ideas underlying later maximum modulus arguments.
- 1868: Schwarz develops key inequality techniques while working on conformal mapping and minimal surfaces.
- 1890s: The maximum modulus principle is formalized in complex analysis textbooks following Weierstrass's rigorous function theory program.
- 1916: Georg Pick extends Schwarz's Lemma to the Schwarz–Pick inequality, describing hyperbolic-metric-contracting maps.
- 20th century: Phragmén–Lindelöf extends the maximum modulus principle to unbounded strips and sectors with growth conditions.
Does the Maximum Modulus Principle say ∣f∣|f| has no maximum at all?

No. It says a non-constant holomorphic function on a domain has no interior local maximum of ∣f∣|f|. If the domain is bounded and ff extends continuously to the closure, a maximum absolutely exists (by compactness) — it is simply guaranteed to occur on the boundary, never strictly inside, unless ff is constant. The principle relocates where maxima can live; it does not deny their existence. This is why boundary value problems for holomorphic and harmonic functions are so central: the boundary data controls everything, including the interior maximum size.

Is there an analogous minimum modulus principle?

Yes, but with a crucial caveat: it requires ff to be zero-free on the domain. Without that hypothesis, ∣f∣|f| can certainly have interior local minima — the simplest example is f(z)=zf(z) = z on any disk containing 00, where ∣f∣|f| has an interior minimum value 00 at z=0z=0, and ff is manifestly non-constant. The minimum modulus principle sidesteps this by excluding zeros: apply the Maximum Modulus Principle to 1/f1/f, which is only holomorphic where f≠0f\ne 0.

Why does the proof work with circles specifically, and not other closed curves?

Circles are used because Cauchy's Integral Formula in its simplest form is stated for circular contours, giving the exact averaging identity f(z0)=12π∫02πf(z0+reiθ) dθf(z_0) = \frac{1}{2\pi}\int_0^{2\pi} f(z_0+re^{i\theta})\,d\theta. More general versions of Cauchy's formula work for other simple closed curves, and one could in principle redo the argument with any contour surrounding z0z_0, but circles make the averaging interpretation transparent and the estimate immediate. The key structural fact — an average cannot strictly exceed the values being averaged — is curve-independent in spirit.

How does the Schwarz Lemma relate to hyperbolic geometry?

The Schwarz–Pick refinement of the lemma shows that every holomorphic self-map of the disk is a contraction (weak or strict) for the Poincaré hyperbolic metric on the disk. This means holomorphic maps of the disk to itself never increase hyperbolic distances, exactly mirroring how the plain Schwarz Lemma shows they never increase the ratio ∣f(z)∣/∣z∣|f(z)|/|z| near 00. Automorphisms of the disk, which are exactly the hyperbolic isometries, correspond to the equality case — a beautiful bridge between complex analysis and non-Euclidean geometry.

Can the Maximum Modulus Principle be used to prove the Fundamental Theorem of Algebra?

Indirectly, yes, though the more standard proof uses Liouville's Theorem. One route: suppose a non-constant polynomial p(z)p(z) has no root. Then 1/p(z)1/p(z) is entire and, because ∣p(z)∣→∞|p(z)| \to \infty as ∣z∣→∞|z|\to\infty, 1/p1/p is bounded, hence constant by Liouville's Theorem — contradiction. The Maximum Modulus Principle enters more naturally in proving Liouville's Theorem itself is sharp, and in estimating growth via Cauchy estimates, which share the same mean-value-property DNA as the maximum modulus argument.

Does the theorem require ff to be injective or one-to-one?

No, injectivity plays no role in the Maximum Modulus Principle or its proof — the argument only uses that ff is holomorphic and that ∣f∣|f| has a local max somewhere. Injectivity becomes relevant later, in the corollary about automorphisms of the disk, where we needed ff to be a bijection in order to apply the Schwarz Lemma to both ff and its inverse f−1f^{-1}. That specific corollary needs injectivity; the core principle does not.

What goes wrong if ff is only assumed continuous, not holomorphic?

Everything breaks. The entire proof rests on the mean value property, which is a consequence of Cauchy's Integral Formula and hence requires holomorphy (or at least harmonicity of Re⁡f\operatorname{Re} f separately). A merely continuous function, like f(z)=1−∣z∣2f(z) = 1 - |z|^2 (real-valued, so trivially "not holomorphic" unless constant, but illustrative), can have a genuine interior maximum with no boundary control at all. Complex-differentiability is precisely the extra rigidity that forbids interior peaks; mere continuity or even real-differentiability is far too weak.

Before this

Where this leads