Complex Analysis
Entire and Meromorphic Functions
1799, Göttingen. The 22-year-old Carl Friedrich Gauss proved the Fundamental Theorem of Algebra geometrically in his doctoral thesis. By 1816 he had three more proofs in hand; a fourth came in 1849. But the shortest and most elegant proof appeared with Joseph Liouville in 1844: a bounded entire function must be constant, so $1/p(z)$ for a polynomial without zeros leads to a contradiction. Today the same rigidity of entire functions underlies the Nyquist-Shannon sampling theorem: a band-limited signal is an entire function of finite exponential type, hence uniquely recoverable from discrete samples. One class of functions, from Gauss to 5G.
- **Nyquist-Shannon sampling theorem:** a band-limited signal $f(t)$ is entire of exponential type $\sigma$; by Paley-Wiener it is uniquely reconstructed from samples at rate $2\sigma$ - the foundation of all digital communication
- **Riemann zeta function:** $\zeta(s)$ is meromorphic on $\mathbb{C}$ with a single simple pole at $s = 1$; the location of its zeros is the content of the 1859 Riemann hypothesis
- **Euler gamma function:** $\Gamma(s)$ is meromorphic with poles at $s = 0, -1, -2, \ldots$, and the Weierstrass factorization gives $1/\Gamma(s) = s e^{\gamma s} \prod (1 + s/n) e^{-s/n}$
- **Euler product for $\sin$:** $\sin(\pi z) = \pi z \prod_{n \geq 1} (1 - z^2/n^2)$ - the first historical instance of Weierstrass factorization (Euler, 1734)
Предварительные знания
Entire Functions and Liouville's Theorem
An entire function is analytic on all of C. Liouville's theorem reveals the extreme rigidity of complex analysis: a bounded entire function must be constant - something with no real-analysis analogue.
**Liouville's Theorem:** a bounded entire function is constant. **Proof sketch:** Cauchy's inequality gives |f'(z)| ≤ M/R for any R. Let R→∞ → f'(z)=0 → f=const. **Polynomial Liouville:** if |f(z)| ≤ C(1+|z|)ⁿ for all z, then f is a polynomial of degree ≤ n. **Growth order:** f has order ρ if |f(z)| ≤ e^{|z|^ρ} for large |z|. - eᶻ: order 1 - e^{z²}: order 2 - Polynomials: order 0
f(z) = sin(z) is entire and satisfies |sin(z)| ≤ e^{|z|} for all z. This means:
The Fundamental Theorem of Algebra
The Fundamental Theorem of Algebra - every nonconstant polynomial has a complex root - follows almost instantly from Liouville. The complex-analysis proof is one of the shortest known.
**Fundamental Theorem of Algebra:** every polynomial p(z) of degree n ≥ 1 has at least one root in C. **Proof via Liouville:** 1. Suppose p(z) ≠ 0 for all z. 2. Then 1/p(z) is entire. 3. Since |p(z)| → ∞ as |z|→∞, |1/p(z)| → 0 → 1/p is bounded. 4. Liouville: 1/p = const → contradiction. **Corollary:** p(z) factors completely over C: p(z) = aₙ(z−r₁)···(z−rₙ)
In the Liouville proof of the FTA, why is 1/p(z) bounded?
Weierstrass Factorization Theorem
The Weierstrass Factorization Theorem generalizes the polynomial factorization p(z) = a·Π(z−rₖ) to entire functions. Any entire function can be expressed as a product over its zeros, times an exponential factor.
**Weierstrass Factorization:** any entire function with zeros {aₙ} (repeated by multiplicity) is: f(z) = zᵐ · e^{g(z)} · Πₙ Eₚₙ(z/aₙ) where Eₚ(z) = (1−z)·exp(z + z²/2 + ... + zᵖ/p) are elementary factors. **Hadamard's theorem:** if |f(z)| ≤ C·e^{|z|^ρ}, then: - g is a polynomial of degree ≤ ρ - The product converges if Σ|aₙ|^{−ρ−1} < ∞ **Classic example:** sin(πz) = πz · Πₙ₌₁^∞ (1−z²/n²)
Why are the elementary factors Eₚ(z) = (1−z)exp(z+z²/2+...+zᵖ/p) used instead of just (1−z)?
Picard's Theorems
Picard's theorems describe how "densely" an entire or meromorphic function covers C. The results are striking: a non-polynomial entire function misses at most one value.
**Little Picard Theorem:** a nonconstant entire function takes every value in C, with at most one exception. Example: eᶻ ≠ 0 but takes every other value. **Great Picard Theorem:** in any punctured neighborhood of an essential singularity, f takes every value in C, with at most one exception. Example: e^{1/z} near z=0: takes every value except 0. **Contrast with poles:** near a pole, |f(z)| → ∞ (misses all small values). Essential singularities are far more pathological.
From Gauss to Picard: 180 years of zeros
Carl Friedrich Gauss gave the first (geometric) proof of the Fundamental Theorem of Algebra in his 1799 doctoral thesis, and published four more proofs in his lifetime. Joseph Liouville in 1844 stated the theorem that a bounded entire function is constant, leading to the shortest known proof of the FTA. Karl Weierstrass in 1876 factored arbitrary entire functions into elementary factors, the transcendental counterpart of factoring polynomials. Émile Picard in 1879 proved that a nonconstant entire function misses at most one value, sharpening the older Casorati-Weierstrass theorem on essential singularities. Four theorems across 180 years form the modern theory of entire functions.
The function eᶻ does not take the value 0. Does this contradict the Little Picard Theorem?
Key Ideas
- **Liouville:** bounded entire = constant; proof via Cauchy inequality as R→∞
- **FTA:** every polynomial has a root in C - proved in two lines via Liouville
- **Weierstrass:** f = zᵐ·e^{g}·Πₙ Eₚₙ(z/aₙ) - factors entire functions over their zeros
- **Picard:** entire function misses at most one value; essential singularity misses at most one near z₀
Connected Topics
Entire functions connect algebra, analysis, and number theory:
- Analytic Continuation — The Riemann zeta function - a meromorphic continuation - uses these results as foundation
- Harmonic Functions — Real and imaginary parts of entire functions are harmonic - the bridge to PDE theory
- Complex Analysis in CS — Entire functions of finite order appear as generating functions in algorithm analysis
Вопросы для размышления
- Why does Liouville's theorem fail for real-analytic functions? Give a counterexample.
- How does Weierstrass factorization explain why sin has Euler's infinite product formula?
- The Great Picard Theorem says e^{1/z} takes all values except 0 near z=0. How can you visualize this geometrically using the Riemann sphere?