TheoremComplete

Rouche's Theorem

Rouche's theorem is a fundamental result for locating zeros of holomorphic functions. It states that a "small" perturbation of a holomorphic function does not change the number of zeros inside a contour.


Statement

Theorem6.8Rouche's theorem (restated)

Let ff and gg be holomorphic inside and on a simple closed contour γ\gamma, with g(z)<f(z)|g(z)| < |f(z)| on γ\gamma. Then ff and f+gf + g have the same number of zeros inside γ\gamma, counted with multiplicity.

RemarkSymmetric form

An equivalent symmetric formulation: if f(z)g(z)<f(z)+g(z)|f(z) - g(z)| < |f(z)| + |g(z)| on γ\gamma, then ff and gg have the same number of zeros inside γ\gamma. This form is sometimes easier to verify.


Proof via the Argument Principle

Proof

Consider the homotopy ht(z)=f(z)+tg(z)h_t(z) = f(z) + tg(z) for t[0,1]t \in [0,1]. On γ\gamma, ht(z)f(z)tg(z)>f(z)f(z)=0|h_t(z)| \geq |f(z)| - t|g(z)| > |f(z)| - |f(z)| = 0, so hth_t has no zeros on γ\gamma for any t[0,1]t \in [0,1].

By the argument principle, the number of zeros of hth_t inside γ\gamma is

N(t)=12πiγht(z)ht(z)dz.N(t) = \frac{1}{2\pi i}\oint_\gamma \frac{h_t'(z)}{h_t(z)}\,dz.

Since hth_t depends continuously on tt and is nonvanishing on γ\gamma, the integrand depends continuously on tt. Therefore N(t)N(t) is a continuous integer-valued function, hence constant.

In particular, N(1)=N(0)N(1) = N(0), i.e., f+gf + g and ff have the same number of zeros. \blacksquare


Applications

ExampleZeros of a polynomial

Determine the number of zeros of p(z)=z75z3+12p(z) = z^7 - 5z^3 + 12 in the annulus 1<z<21 < |z| < 2.

On z=1|z| = 1: 12=12>z75z31+5=6|12| = 12 > |z^7 - 5z^3| \leq 1 + 5 = 6. By Rouche, pp and 1212 have the same number of zeros in z<1|z| < 1: zero zeros.

On z=2|z| = 2: z7=128>5z3+1240+12=52|z^7| = 128 > |-5z^3 + 12| \leq 40 + 12 = 52. By Rouche, pp and z7z^7 have the same number of zeros in z<2|z| < 2: seven zeros.

Therefore pp has 70=77 - 0 = 7 zeros in 1<z<21 < |z| < 2.

ExamplePerturbation of zeros

For small ε>0\varepsilon > 0, the polynomial pε(z)=zn+εq(z)p_\varepsilon(z) = z^n + \varepsilon q(z) (where degq<n\deg q < n) has all nn zeros near the origin. On z=r|z| = r for any fixed r>0r > 0, zn=rn|z^n| = r^n while εq(z)εCrn1|\varepsilon q(z)| \leq \varepsilon C r^{n-1}. For ε<r/C\varepsilon < r/C, Rouche applies, showing pεp_\varepsilon has nn zeros in z<r|z| < r. As ε0\varepsilon \to 0, all zeros converge to the origin.


Hurwitz's Theorem

Theorem6.9Hurwitz's theorem

Let {fn}\{f_n\} be a sequence of holomorphic functions on a domain DD converging uniformly on compact subsets to ff, where ff is not identically zero. If f(z0)=0f(z_0) = 0 with multiplicity mm, then for sufficiently large nn, fnf_n has exactly mm zeros (counted with multiplicity) in any sufficiently small disk around z0z_0.

In particular, if each fnf_n is nonvanishing on DD and fnff_n \to f uniformly on compacts, then either f0f \equiv 0 or ff is also nonvanishing on DD.

RemarkSignificance of Hurwitz's theorem

Hurwitz's theorem is crucial in the proof of the Riemann mapping theorem: it ensures that the limit of injective holomorphic functions is either injective or constant. Combined with normal families theory, it provides the key tool for extracting convergent subsequences with the desired properties.