🧮 Brain Teaser

A Polynomial That Lives on the Unit Circle

Let pp be a complex polynomial that maps the unit circle into itself:

∣p(z)∣=1for all ∣z∣=1.|p(z)| = 1 \quad \text{for all } |z| = 1.

Prove that p(z)=c znp(z) = c\,z^n for some n≥0n \ge 0 and some constant cc with ∣c∣=1|c| = 1.

(Bonus: what is the analogous statement for rational functions RR with ∣R(z)∣=1|R(z)| = 1 on ∣z∣=1|z| = 1?)


Field: Complex Analysis

Why it's interesting: The hypothesis only concerns the values of pp on a circle. It says nothing about the coefficients, and nothing directly about the zeros. Yet it forces pp to be a pure monomial. The result illustrates a recurring theme in complex analysis: an equation that holds on a curve becomes an identity everywhere. Here the bridge is the observation that on the unit circle, complex conjugation coincides with inversion, since zˉ=1/z\bar z = 1/z. The bonus generalizes the result to finite Blaschke products, which are the building blocks of inner functions and of disk automorphisms.

Estimated difficulty: 2 / 5

Hint (if you want one): On ∣z∣=1|z|=1, rewrite ∣p(z)∣2=p(z)p(z)‾|p(z)|^2 = p(z)\overline{p(z)} using zˉ=1/z\bar z = 1/z.

identity theoremreflectionBlaschke productmaximum moduluspolynomials

Answer: A Polynomial That Lives on the Unit Circle

Key Idea / Intuition

On the unit circle, conjugation is inversion: zˉ=1/z\bar z = 1/z. So the condition p pˉ=1p\,\bar p = 1 is really the algebraic identity p(z) p∗(1/z)=1p(z)\,p^*(1/z) = 1, where p∗p^* is pp with conjugated coefficients. An identity between holomorphic functions on a circle holds everywhere on C∖{0}\mathbb{C}\setminus\{0\}. A zero z0≠0z_0 \neq 0 of pp would make the left side vanish, which is impossible. So every zero of pp sits at the origin, and pp is a monomial.

Equivalently, one can view it through reflection symmetry. A function of modulus 11 on the circle must have its zeros at z0z_0 "mirrored" by poles at 1/zˉ01/\bar z_0. A polynomial has no poles to spare, so its zeros cannot be anywhere except 00, whose mirror point is ∞\infty.


Formal Proof / Solution

Setup. Write p(z)=∑k=0nakzkp(z) = \sum_{k=0}^{n} a_k z^k with an≠0a_n \neq 0. The polynomial is not identically zero, since ∣p∣=1|p|=1 on the circle. Define the "conjugate polynomial"

p∗(w)=∑k=0nak‾ wk,so thatp(z)‾=p∗(zˉ).p^*(w) = \sum_{k=0}^{n} \overline{a_k}\, w^k, \qquad\text{so that}\qquad \overline{p(z)} = p^*(\bar z).

Step 1: Turn the modulus condition into an analytic identity. For ∣z∣=1|z| = 1 we have zˉ=1/z\bar z = 1/z. Hence

1=∣p(z)∣2=p(z) p(z)‾=p(z) p∗(zˉ)=p(z) p∗(1/z).1 = |p(z)|^2 = p(z)\,\overline{p(z)} = p(z)\,p^*(\bar z) = p(z)\,p^*(1/z).

Step 2: Extend off the circle. The function F(z)=p(z) p∗(1/z)−1F(z) = p(z)\,p^*(1/z) - 1 is holomorphic on the connected domain C∖{0}\mathbb{C}\setminus\{0\}. It vanishes on the unit circle, a set with accumulation points. By the identity theorem, F≡0F \equiv 0:

p(z) p∗(1/z)=1for all z≠0.p(z)\,p^*(1/z) = 1 \qquad \text{for all } z \neq 0.

Step 3: Locate the zeros. Suppose p(z0)=0p(z_0) = 0 for some z0≠0z_0 \neq 0. Plugging z0z_0 into the identity gives 0⋅p∗(1/z0)=10 \cdot p^*(1/z_0) = 1, where p∗(1/z0)p^*(1/z_0) is a finite number. This is a contradiction. Therefore every zero of pp is at 00.

Step 4: Conclude. By the fundamental theorem of algebra, p(z)=an∏i=1n(z−zi)p(z) = a_n \prod_{i=1}^{n}(z - z_i). Step 3 shows all zi=0z_i = 0, so p(z)=anznp(z) = a_n z^n. Evaluating at z=1z=1 gives ∣an∣=1|a_n| = 1. ■\blacksquare


Alternative proof (maximum modulus + Blaschke factors)

pp has no zeros on ∣z∣=1|z|=1, so its zeros z1,…,zmz_1,\dots,z_m (with multiplicity) in the closed disk all lie in the open disk. Form the Blaschke product

B(z)=∏i=1mz−zi1−zi‾ z.B(z) = \prod_{i=1}^{m} \frac{z - z_i}{1 - \overline{z_i}\, z}.

Each factor has modulus 11 on ∣z∣=1|z|=1. The quotient g=p/Bg = p/B is holomorphic and zero-free on a neighborhood of the closed disk, with ∣g∣=1|g| = 1 on the boundary.

  • The maximum modulus principle applied to gg gives ∣g∣≤1|g| \le 1 inside.
  • The maximum modulus principle applied to 1/g1/g gives ∣g∣≥1|g| \ge 1 inside.

So ∣g∣≡1|g| \equiv 1, and an open mapping argument shows g≡cg \equiv c is constant. Hence p=cBp = cB.

BB has a pole at 1/zi‾1/\overline{z_i} for every zi≠0z_i \neq 0, while pp has no poles. Therefore all zi=0z_i = 0, giving B(z)=zmB(z) = z^m and p=czmp = c z^m.


Bonus: rational functions

The same reasoning shows that a rational function RR with ∣R∣=1|R| = 1 on ∣z∣=1|z| = 1 satisfies R(z) R(1/zˉ)‾=1R(z)\,\overline{R(1/\bar z)} = 1. Zeros at aa must therefore be paired with poles at 1/aˉ1/\bar a. Consequently

R(z)=c zk∏iz−ai1−ai‾ z,∣c∣=1, k∈Z, ai≠0.R(z) = c\, z^{k} \prod_{i} \frac{z - a_i}{1 - \overline{a_i}\, z}, \qquad |c| = 1,\ k \in \mathbb{Z},\ a_i \neq 0.

So RR is a unimodular constant times a power of zz times a finite Blaschke product. Taking ∣ai∣<1|a_i| < 1 for all ii is exactly the condition that RR map the unit disk into itself. These are precisely the proper holomorphic self-maps of the disk.

Type: Complex AnalysisEdit on GitHub ↗