Liouville's Theorem: Why Bounded Entire Functions Are Constant

The precise statement

Call a function f: ℂ → ℂ entire if it is holomorphic (complex-differentiable) at every point of the plane. Equivalently, f has a power series ∑ₙ aₙ zⁿ with infinite radius of convergence.

Liouville's Theorem. If f is entire and there exists M ≥ 0 with |f(z)| ≤ M for all z ∈ ℂ, then f is constant.

Two hypotheses do all the work, and both are indispensable:

• Entire — holomorphic on all of ℂ, with no singularities anywhere. A single pole (as in 1/z) breaks it.
• Bounded on ℂ — one global constant M caps |f| everywhere. Boundedness on a proper subset is not enough.

The conclusion is maximally strong: not merely that f is nice, but that it takes a single value. There is no intermediate 'slowly varying' case.

The intuition: rigidity of holomorphic maps

Holomorphic functions are far more rigid than merely smooth ones. Being complex-differentiable forces the real and imaginary parts to satisfy the Cauchy–Riemann equations, which chain the function's behavior together across the whole plane — the value on any small disk secretly encodes the value everywhere via the power series.

The geometric picture: an entire function is an open map (unless constant), so its image is an open subset of ℂ. If |f| ≤ M, the image lives inside the closed disk of radius M. A non-constant entire function would have to spread its image out infinitely — Picard's theorems later make this brutally precise, showing it must hit almost every complex value. Squeezing that sprawling image into a bounded disk is impossible.

Another slogan: information cannot be created. The derivative f′ is controlled by how much room f has to grow. Give f no room (a fixed ceiling), and its derivative is forced to vanish identically — so f can't change.

The key idea of the proof: Cauchy's estimate, then R → ∞

The mechanism is a single inequality plus a limit. Start from Cauchy's integral formula for the derivative: for any center z₀ and radius R,

f′(z₀) = (1/2πi) ∮_|z−z₀|=R f(z)/(z−z₀)² dz.

Bound the integral crudely (the ML-inequality): the integrand has modulus ≤ M/R², and the circle has length 2πR. This yields Cauchy's estimate:

|f′(z₀)| ≤ (1/2π) · (M/R²) · 2πR = M/R.

Now the punchline: R was arbitrary because f is entire — the circle of any radius stays in the domain. Let R → ∞. The right side M/R → 0, forcing |f′(z₀)| = 0. Since z₀ was arbitrary, f′ ≡ 0 on ℂ, and a function with zero derivative on a connected domain is constant. That is the whole argument.

Worked example and the power-series view

A concrete check. Take f(z) = sin z. Is it entire? Yes — power series everywhere. Is it bounded? Only on the real line, where |sin x| ≤ 1. But sin(iy) = i sinh y, whose modulus sinh|y| → ∞. So sin z is entire but unbounded, and Liouville imposes nothing. This is the classic trap: real boundedness is irrelevant; you need boundedness on all of ℂ.

Coefficient proof (the cleanest special case). Write f(z) = ∑ₙ aₙ zⁿ. Cauchy's estimates give |aₙ| ≤ M/Rⁿ for every radius R. Fix any n ≥ 1 and let R → ∞: M/Rⁿ → 0, so aₙ = 0. Every coefficient beyond a₀ dies, leaving f(z) = a₀ — a constant. This packages the whole theorem into one line of coefficient bookkeeping.

Why the hypotheses matter — and counterexamples

Each hypothesis is sharp; drop it and non-constant examples appear immediately.

• Drop boundedness: f(z) = z, or e^z, are entire and non-constant. So 'entire' alone never forces constancy — the theorem genuinely needs the ceiling M.
• Drop 'entire' (allow a domain ≠ ℂ): on the unit disk 𝔻, f(z) = z is holomorphic and bounded (|f| < 1) yet non-constant. The proof breaks because you cannot take R → ∞ — the circle would leave the domain.
• Drop 'holomorphic': a bounded real-smooth function like u(x,y) = sin x is fine. Rigidity comes from Cauchy–Riemann, not smoothness.

The theorem sits in a family of sharpenings. Extended Liouville: if |f(z)| ≤ C(1 + |z|)ᵏ, then f is a polynomial of degree ≤ k (same Cauchy-estimate proof, killing coefficients past degree k). Picard's little theorem is the deepest cousin: a non-constant entire function omits at most one value of ℂ.

Applications and significance

Liouville's Theorem is small to state but load-bearing across analysis and algebra.

• Fundamental Theorem of Algebra. The most famous corollary. If a non-constant polynomial p had no root, then 1/p would be entire; since |p(z)| → ∞, the reciprocal is bounded, so by Liouville 1/p is constant — contradiction. Hence p has a root, and by induction ℂ is algebraically closed.
• Uniqueness in PDE and potential theory. Bounded harmonic functions on ℝ² and Liouville-type theorems for harmonic and analytic functions give uniqueness of solutions with growth constraints.
• Removable behavior and normal families. It anchors the theory of how entire functions grow, feeding Hadamard factorization, the study of order and type, and Montel/Picard theory.

The deeper lesson: global holomorphy is enormously constraining. On ℂ there is no such thing as a bounded, genuinely varying analytic function — a principle that recurs whenever complex methods tame a real problem.

How the conclusion depends on which hypotheses hold — dropping any one lets non-constant functions survive.

Function / setting	Entire?	Bounded on ℂ?	Forced constant?
f(z) = e^z	Yes	No (grows on ℝ⁺)	No — escapes the theorem
f(z) = sin z	Yes	No (unbounded on iℝ)	No
f(z) = 1/(1+z²)	No (poles at ±i)	No (blows up near ±i)	No — not entire
f(z) = c (constant)	Yes	Yes	Yes (trivially)
f bounded, holomorphic on unit disk	No (domain ≠ ℂ)	Yes	No — e.g. f(z)=z