Skip to content
Tutorials Center
My learning

Topics

Lesson 25 of 36 · Cauchy Theorems

Derivatives from Integrals and Liouville's Theorem

See why analytic functions have derivatives of every order, use Cauchy's formula for f^(n)(a), the Cauchy estimates, and prove Liouville and the algebra theorem.

Cauchy's integral formula recovers f(a) from boundary values. Differentiate it with respect to a, and you get formulas for f'(a), f''(a), and every higher derivative. The consequences are striking: an analytic function is automatically infinitely differentiable, and its derivatives are controlled by its size. This leads to Liouville's theorem and to a short proof of the fundamental theorem of algebra.

The derivative formula

Let f be analytic on and inside a simple closed counterclockwise contour C, and a inside C. Then for every n = 0, 1, 2, ...

Text
f^(n)(a) = n! / (2πi) · ∮_C f(z) / (z - a)^(n+1) dz

For n = 0 this is Cauchy's integral formula. For n = 1:

Text
f'(a) = (1 / 2πi) ∮_C f(z) / (z - a)² dz

Where it comes from. In f(a) = (1/2πi) ∮ f(z)/(z - a) dz only the factor 1/(z - a) depends on a. Differentiating under the integral sign gives 1/(z - a)², then 2/(z - a)³, then 6/(z - a)⁴, and in general n!/(z - a)^(n+1). Differentiation under the integral is legitimate because the integrand stays continuous in (z, a) when a is away from C.

Consequence. Since the right-hand side exists for every n, an analytic function has derivatives of all orders, and every derivative is again analytic. Compare with real analysis, where a function can be once differentiable and not twice.

Example 1

∮_(|z|=2) e^z / z³ dz. Here n + 1 = 3 so n = 2, f(z) = e^z, a = 0. The integral equals 2πi · f''(0)/2! = 2πi · 1/2 = πi.

Example 2

∮_(|z|=1) sin z / z² dz. Here n = 1, f = sin, a = 0. The integral is 2πi · f'(0) = 2πi · cos 0 = 2πi.

Cauchy estimates

Take C to be the circle |z - a| = R inside which f is analytic, and suppose |f(z)| ≤ M on the circle. Apply the ML inequality to the derivative formula: the integrand has modulus at most M / R^(n+1) and the length is 2πR, so

Text
|f^(n)(a)| ≤ n! · M / R^n

These are the Cauchy estimates. They say that a function that stays small on a big circle has small derivatives at the center, and the bound for higher derivatives gets stronger as R grows (the denominator R^n).

Liouville's theorem

Theorem (Liouville). A bounded entire function is constant.

Proof. Let f be entire with |f(z)| ≤ M for all z. Use the Cauchy estimate with n = 1 at any point a, on a circle of any radius R:

Text
|f'(a)| ≤ M / R

Let R → ∞. The right side tends to 0, so f'(a) = 0 for every a. A function with zero derivative on the connected plane is constant. ∎

So sin z and cos z, which are bounded on the real line, must be unbounded in the complex plane; indeed |sin(iy)| = sinh |y| grows like e^|y|/2.

The fundamental theorem of algebra

Theorem. Every non-constant polynomial p(z) has a root in C.

Proof. Suppose p has no root. Then g(z) = 1/p(z) is entire. For large |z| we have |p(z)| ≥ c|z|^k for some c > 0 and k ≥ 1 (the leading term dominates), so |g(z)| → 0 as |z| → ∞. A continuous function on the plane that tends to 0 at infinity is bounded (it is bounded on a large closed disc by continuity, and small outside it). By Liouville, g is constant, so p is constant, a contradiction. ∎

A one-line corollary: counting multiplicities, a polynomial of degree n has exactly n complex roots.

flowchart TD
    A["Cauchy integral formula"] --> B["Differentiate in a:<br/>f^(n)(a) = n!/(2πi) ∮ f/(z-a)^(n+1)"]
    B --> C["Cauchy estimates<br/>|f^(n)(a)| ≤ n! M / R^n"]
    C --> D["Liouville: bounded entire is constant"]
    D --> E["Fundamental theorem of algebra"]

From one integral formula to the fundamental theorem of algebra.

Try it: derivatives by integration

We compute f^(n)(a) for f(z) = sin z at a = 0.5 + 0.3i using only the contour integral (a circle of radius 1 around a), for n = 0..5, and compare with the exact derivatives sin, cos, -sin, -cos, sin, cos evaluated at a.

Python Derivatives from a contour integral

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Try it: how good are the Cauchy estimates?

For f(z) = e^z at a = 0, every derivative f^(n)(0) = 1. On |z| = R the maximum of |e^z| is e^R, so the Cauchy estimate says 1 ≤ n! e^R / R^n for every R > 0. The best choice of R is R = n (the minimum of e^R / R^n), giving n! e^n / n^n. The table shows how close this is to 1: surprisingly close, only off by a factor of about √(2πn).

Python Optimizing the radius in the Cauchy estimate

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. What is ∮_(|z|=1) e^z / z² dz?

    n + 1 = 2 so n = 1: the integral is 2πi · f'(0) = 2πi · e^0 = 2πi.
  2. 2. What follows from the derivative formula about an analytic function?

    The formula for f^(n)(a) exists for every n, so analytic functions are infinitely differentiable.
  3. 3. Liouville's theorem says that an entire function that is bounded must be

    The Cauchy estimate |f'(a)| ≤ M/R tends to 0 as R grows, so f' ≡ 0.
  4. 4. Which entire function must be unbounded on the complex plane?

    sin z is entire and non-constant, so by Liouville it must be unbounded; for example |sin(iy)| = sinh|y|.

Summary

  • f^(n)(a) = n!/(2πi) ∮ f(z)/(z - a)^(n+1) dz: analytic functions have derivatives of all orders.
  • Cauchy estimates: |f^(n)(a)| ≤ n! M / R^n on a circle of radius R where |f| ≤ M.
  • Liouville: a bounded entire function is constant.
  • Fundamental theorem of algebra: otherwise 1/p would be a non-constant bounded entire function.

Next: Maximum modulus and mean value