JOVANA
Explore Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Liouville and the Fundamental Theorem of Algebra

One short chain of inequalities settles two of the most famous facts in mathematics: a bounded entire function can only be constant, and every non-constant polynomial has a root. We follow that chain from the Cauchy estimates to Liouville's theorem and watch it crack open the fundamental theorem of algebra.

From the derivative formula to a size bound

In the previous guide you earned something extraordinary: the formula for every derivative of a holomorphic function as an integral over a surrounding loop. Write it for a circle of radius R centred at z_0. The n-th derivative f^(n)(z_0) equals (n! / (2 pi i)) times the integral over that circle of f(w) / (w - z_0)^(n+1) dw. The left side is a derivative — a delicate local thing — yet the right side only ever looks at the values of f out on the circle. That mismatch is the whole engine of this guide: control f on a far-away circle, and you control its derivatives at the centre.

Now reach for the only estimating tool you need, the ML inequality from the integration rung. The path is the circle, so its length is L = 2 pi R. The integrand is f(w) / (w - z_0)^(n+1); on the circle |w - z_0| = R, so the denominator has modulus R^(n+1). If we know |f(w)| is at most some bound M on that circle, then the integrand's size is at most M / R^(n+1). Multiply size by length, then fold in the constant n! / (2 pi) sitting out front, and the 2 pi R cancels almost everything.

| f^(n)(z_0) |  =  ( n! / (2 pi) ) * | integral over |w - z_0| = R  of  f(w) / (w - z_0)^(n+1) dw |

              <=  ( n! / (2 pi) ) * ( M / R^(n+1) ) * ( 2 pi R )      [ ML inequality ]

              =   n! * M / R^n                                       [ Cauchy estimate ]

   where   M = max of |f(w)|  on the circle  |w - z_0| = R
The Cauchy estimate: each derivative at the centre is controlled by the size of f on the circle, divided by R to the power n.

What the Cauchy estimates actually say

The result, |f^(n)(z_0)| at most n! M / R^n, is the family of Cauchy estimates, one for each n. Read them as a trade between three quantities. M measures how big f is allowed to get on the circle. R is how far out that circle sits. And the left side is how steep the n-th derivative can be at the centre. The estimates say steepness is bought with size and discounted by distance: a function that stays small far away simply cannot have wild high derivatives at the middle.

Pause on how strange this is from a real-variable point of view. On the real line a function can be bounded — pinned between -1 and 1, say — and still wiggle so violently that its derivatives are astronomically large; think of sin(1000 x), bounded by 1 yet with slope up to 1000. Nothing like that survives in the complex world. For a holomorphic function the single complex derivative you assumed at the start has already forced infinite differentiability and this rigid bookkeeping that ties high derivatives to mere boundedness. The Cauchy estimates are where that rigidity becomes a number you can write down.

Liouville's theorem: bounded plus entire means constant

Here is the payoff. Suppose f is entire — holomorphic on the whole plane, an entire function — and suppose it is bounded, meaning there is a single number M with |f(z)| at most M for every z, no matter how far out. Then Liouville's theorem says f must be constant. Not approximately constant, not eventually constant — exactly, identically constant. A function defined and differentiable everywhere that never escapes a fixed bound has no choice but to be a flat horizontal value.

The proof is the section above, used once. Fix any centre z_0 and apply the Cauchy estimate with n = 1: |f'(z_0)| is at most 1! times M over R, that is M / R. Because f is entire we may take the circle as large as we like — R is free to run to infinity. Since the global bound M never changes, M / R goes to zero. So |f'(z_0)| is at most something that goes to zero, forcing f'(z_0) = 0. But z_0 was arbitrary, so f' is zero everywhere. A function with vanishing derivative on a connected domain is constant. Done.

  1. Fix an arbitrary point z_0 in the plane; we will show f'(z_0) = 0.
  2. Apply the Cauchy estimate with n = 1 on the circle |z - z_0| = R: |f'(z_0)| is at most M / R.
  3. Let R run to infinity. The bound M is global and fixed, so M / R goes to zero.
  4. Conclude f'(z_0) = 0. Since z_0 was arbitrary, f' vanishes everywhere, so f is constant on the connected plane.

The fundamental theorem of algebra, in three lines

Centuries of algebra struggled to prove that every polynomial has a root; complex analysis disposes of it almost casually. The fundamental theorem of algebra states that any polynomial p(z) of degree at least 1, with complex coefficients, has at least one root in the complex plane. The strategy is a beautiful piece of judo: assume p has no root, manufacture a bounded entire function out of it, and let Liouville crush the assumption.

  1. Suppose, for contradiction, that p(z) is never zero anywhere in the plane.
  2. Then g(z) = 1 / p(z) is defined everywhere and is entire — a quotient of holomorphic functions whose denominator never vanishes.
  3. As |z| goes to infinity, |p(z)| goes to infinity (the leading term z^n dominates), so |g(z)| goes to zero. Far out, g is tiny.
  4. Hence g is bounded: it is small outside some large disk, and continuous on the closed disk, so bounded there too. A bounded entire function is constant by Liouville.
  5. If g is constant then p is constant — contradicting degree at least 1. So our assumption was false: p must have a root.

Step 3 deserves a second look, because it is the only place real estimation enters. Write p(z) = a_n z^n + (lower terms) with a_n not zero. Factor out z^n: p(z) = z^n times (a_n + a_(n-1)/z + ... + a_0/z^n). The bracket tends to a_n as |z| grows, so for large |z| it stays bounded below by, say, |a_n|/2, while |z^n| blows up. Thus |p(z)| is eventually at least (|a_n|/2)|z|^n, which races to infinity. That single growth fact — a polynomial of positive degree cannot stay bounded — is the one honest inequality the whole proof rests on.

Why one root gives all of them, and what to carry away

We only proved one root exists, but the full theorem — n roots counted with multiplicity — follows immediately. If p has degree n and root r_1, polynomial division gives p(z) = (z - r_1) q(z) with q of degree n - 1. Apply the same theorem to q to peel off r_2, and repeat. After n steps p has split into n linear factors, so a degree-n polynomial has exactly n roots in the plane, counted with multiplicity. This is why C is called algebraically closed, and why the complex numbers, not the reals, are the natural home of polynomial root-finding.

Step back and admire the shape of the argument, because it recurs all over complex analysis. A purely algebraic statement about polynomials was settled by an analytic fact about growth and boundedness. The bridge was the Cauchy integral formula, whose derivative version gave the Cauchy estimates, which gave Liouville, which gave the root. Every link is one you forged in this rung. Nothing here is a trick reserved for special functions — it is the ordinary power of holomorphy, that a function knowing its values on a circle knows everything inside.