riemann

The Lean code claims a proof of the Riemann Hypothesis. This site has not checked the claim yet. See what is checked.

Complex analysis

Holomorphic and meromorphic functions, zeros and poles, analytic continuation, the logarithmic derivative, the zeros of the cosine, the Gamma function, Poisson summation, theta functions and the Mellin transform: what the last step of the proof and the continuation of Eisenstein series need.

Prerequisites: Analysis.

Complex analysis enters through the zeta function and the Eisenstein series. The last step RhOfRight of the overview moves a zero of ζ\zeta across the functional equation, which needs that Γ\Gamma has no zeros and where the cosine vanishes. The theta branch continues Eisenstein series in a complex variable ss.

Complex numbers

Definition. The field of complex numbers is ℂ={a+bi:a,b∈ℝ}\mathbb{C} = \{a + bi : a, b \in \mathbb{R}\} with i2=−1i^2 = -1. For s=σ+its = \sigma + it with σ,t∈ℝ\sigma, t \in \mathbb{R}, the real part is Re⁡(s)=σ\operatorname{Re}(s) = \sigma and the imaginary part is Im⁡(s)=t\operatorname{Im}(s) = t. The complex conjugate is s‾=σ−it\overline{s} = \sigma - it, and ∣s∣2=ss‾|s|^2 = s \overline{s}.

Conjugation is a field automorphism of order 22 with fixed field ℝ\mathbb{R}, and Re⁡(s)=(s+s‾)/2\operatorname{Re}(s) = (s + \overline{s})/2. For x>0x > 0, ∣xs∣=xRe⁡(s)|x^{s}| = x^{\operatorname{Re}(s)}.

RiemannHypothesis is about the real parts of the zeros of ζ\zeta, and the Rallis integrand of block determinants has a conjugated matrix coefficient.

Holomorphic functions

Definition. Let U⊆ℂU \subseteq \mathbb{C} be open. A function f:U→ℂf : U \to \mathbb{C} is holomorphic if the limit f′(z0)=lim⁡z→z0(f(z)−f(z0))/(z−z0)f'(z_0) = \lim_{z \to z_0} (f(z) - f(z_0))/(z - z_0) exists for each z0∈Uz_0 \in U. A function holomorphic on all of ℂ\mathbb{C} is entire.

Theorem. A holomorphic function ff on UU is analytic: for each disc D(z0,r)⊆UD(z_0, r) \subseteq U, f(z)=∑n≥0an(z−z0)nf(z) = \sum_{n \ge 0} a_n (z - z_0)^n on D(z0,r)D(z_0, r). So f′f' is again holomorphic.

Example. Polynomials, eze^z, cos⁡z\cos z and xs=eslog⁡xx^{s} = e^{s \log x} for fixed x>0x > 0 are entire.

Proposition (holomorphic family). Let (X,μ)(X, \mu) be a measure space, U⊆ℂU \subseteq \mathbb{C} open, and F:X×U→ℂF : X \times U \to \mathbb{C} measurable with F(x,⋅)F(x, \cdot) holomorphic for each xx. Suppose that each compact K⊆UK \subseteq U has an integrable gKg_K with ∣F(x,s)∣≤gK(x)|F(x, s)| \le g_K(x) for s∈Ks \in K. Then s↦∫XF(x,s) dμ(x)s \mapsto \int_X F(x, s) \, d\mu(x) is holomorphic on UU.

The proof uses dominated convergence, Fubini and Morera's theorem. Such an FF is a holomorphic family, for example F(g,s)=a(g)sϕ(g)F(g, s) = a(g)^{s} \phi(g) with a>0a > 0. Likewise a series of holomorphic functions that converges uniformly on compact sets has a holomorphic sum.

In Step 4 of block determinants the Rallis integral with the factor a(g,1)z−s0a(g,1)^{z - s_0} is holomorphic in zz, and the section fs=as−s0ff_s = a^{s - s_0} f of the Siegel-Weil chapter is a holomorphic family.

Zeros and poles

Definition. Let ff be holomorphic on a connected neighbourhood of z0z_0 and not identically 00 there. The order of a zero z0z_0 is the least m≥0m \ge 0 with f(m)(z0)≠0f^{(m)}(z_0) \ne 0. Then f(z)=(z−z0)mg(z)f(z) = (z - z_0)^m g(z) with gg holomorphic and g(z0)≠0g(z_0) \ne 0. The number mm is also the multiplicity of the zero, and the zero is simple if m=1m = 1.

Definition. Let ff be holomorphic on a punctured disc 0<∣z−z0∣<r0 < |z - z_0| < r. The point z0z_0 is a pole of order m≥1m \ge 1 if f(z)=(z−z0)−mg(z)f(z) = (z - z_0)^{-m} g(z) with gg holomorphic near z0z_0 and g(z0)≠0g(z_0) \ne 0. The residue of ff at z0z_0 is the coefficient a−1a_{-1} of its Laurent series ∑n≥−man(z−z0)n\sum_{n \ge -m} a_n (z - z_0)^n.

Definition. Let UU be open. A function ff is meromorphic on UU if there is a set P⊆UP \subseteq U with no limit point in UU such that ff is holomorphic on U∖PU \setminus P and has a pole at each point of PP. On a connected UU the meromorphic functions form a field.

Example. 1/(s−1)1/(s - 1) is meromorphic on ℂ\mathbb{C} with a simple pole at 11 of residue 11. The Riemann zeta function ζ(s)=∑n≥1n−s\zeta(s) = \sum_{n \ge 1} n^{-s}, defined for Re⁡(s)>1\operatorname{Re}(s) > 1, continues to a meromorphic function on ℂ\mathbb{C} whose only pole is a simple pole at s=1s = 1 of residue 11.

The Riemann hypothesis of the overview excludes the pole s=1s = 1. EisValueAt, in the Siegel-Weil chapter, asks for a meromorphic function equal to the Eisenstein series for large Re⁡(s)\operatorname{Re}(s), and HolLead of Step 6 says that it has no pole at s0s_0.

Analytic continuation and the identity theorem

Theorem (identity theorem). Let U⊆ℂU \subseteq \mathbb{C} be open and connected, and let f,gf, g be holomorphic on UU. If f=gf = g on a set with a limit point in UU, for example a nonempty open subset, then f=gf = g on UU.

Definition. Let U⊆VU \subseteq V be open, UU nonempty, VV connected, and ff holomorphic on UU. An analytic continuation of ff to VV is a holomorphic (or meromorphic) FF on VV with F=fF = f on UU. By the identity theorem there is at most one, so an identity between continued functions holds on all of VV once it holds for large Re⁡(s)\operatorname{Re}(s).

The L-function L(s,χ)L(s, \chi) of EveryCharacter is the continuation of a Dirichlet series to ℂ\mathbb{C}, and the lemma zero_of_half_pk of Step 4 of block determinants extends an identity from large Re⁡(z)\operatorname{Re}(z) to z=s0z = s_0 in this way.

The logarithmic derivative

Definition. The logarithmic derivative of a meromorphic f≢0f \not\equiv 0 is f′/ff'/f.

Proposition. If ff has a zero of order mm at z0z_0, then f′/ff'/f has a simple pole at z0z_0 with residue mm; at a pole of order mm the residue is −m-m. Away from zeros and poles f′/ff'/f is holomorphic, and (fg)′/(fg)=f′/f+g′/g(fg)'/(fg) = f'/f + g'/g.

Example. For Re⁡(s)>1\operatorname{Re}(s) > 1, the Euler product gives −ζ′(s)/ζ(s)=∑n≥1Λ(n)n−s-\zeta'(s)/\zeta(s) = \sum_{n \ge 1} \Lambda(n) n^{-s}, with the von Mangoldt function Λ\Lambda of Chapter 6.

The bound LogDerOne of the literature compares L′(1,χ)/L(1,χ)L'(1, \chi)/L(1, \chi) with a sum Z(χ)Z(\chi) of 1/(1−ρ)1/(1 - \rho) over the zeros ρ\rho, counted with multiplicity.

The zeros of the cosine

Definition. For z∈ℂz \in \mathbb{C}, cos⁡z=(eiz+e−iz)/2\cos z = (e^{iz} + e^{-iz})/2 and sin⁡z=(eiz−e−iz)/(2i)\sin z = (e^{iz} - e^{-iz})/(2i). Both are entire, and cos⁡2z+sin⁡2z=1\cos^2 z + \sin^2 z = 1.

Proposition (zeros of the cosine). cos⁡z=0\cos z = 0 if and only if z=(2k+1)π/2z = (2k + 1)\pi/2 for some k∈ℤk \in \mathbb{Z}, and each of these zeros is simple. The zeros of sin⁡z\sin z are the kπk\pi, k∈ℤk \in \mathbb{Z}.

The idea: cos⁡z=0\cos z = 0 means e2iz=−1e^{2iz} = -1, and ew=−1e^w = -1 exactly for w∈πi+2πiℤw \in \pi i + 2\pi i \mathbb{Z}. In RhOfRight, cos⁡(πu/2)=0\cos(\pi u/2) = 0 forces u=2k+1u = 2k + 1, so s=1−u=−2ks = 1 - u = -2k is 00, a trivial zero, or has Re⁡(s)≥2\operatorname{Re}(s) \ge 2.

The Gamma function

Definition. For Re⁡(s)>0\operatorname{Re}(s) > 0, the Gamma function is Γ(s)=∫0∞ts−1e−t dt.\Gamma(s) = \int_0^\infty t^{s - 1} e^{-t} \, dt. It is holomorphic there, Γ(1)=1\Gamma(1) = 1, and Γ(s+1)=sΓ(s)\Gamma(s + 1) = s \Gamma(s). So Γ(n+1)=n!\Gamma(n + 1) = n!.

Theorem. Γ\Gamma continues to a meromorphic function on ℂ\mathbb{C}. Its poles are simple poles at s=0,−1,−2,…s = 0, -1, -2, \dots, and the residue at −n-n is (−1)n/n!(-1)^n / n!.

Theorem (Euler reflection formula). For s∈ℂ∖ℤs \in \mathbb{C} \setminus \mathbb{Z}, Γ(s)Γ(1−s)=π/sin⁡(πs)\Gamma(s) \Gamma(1 - s) = \pi / \sin(\pi s).

Theorem (Gamma has no zeros). The Gamma function has no zeros: Γ(s)≠0\Gamma(s) \ne 0 for each s∈ℂ∖{0,−1,−2,… }s \in \mathbb{C} \setminus \{0, -1, -2, \dots\}. So 1/Γ1/\Gamma is entire.

For s∉ℤs \notin \mathbb{Z} this follows from the reflection formula, since π/sin⁡(πs)≠0\pi/\sin(\pi s) \ne 0 and Γ(1−s)\Gamma(1 - s) is finite; for s=n≥1s = n \ge 1, Γ(n)=(n−1)!≠0\Gamma(n) = (n - 1)! \ne 0. For Re⁡(s)>0\operatorname{Re}(s) > 0 it is Mathlib's Complex.Gamma_ne_zero_of_re_pos.

Theorem (Legendre duplication formula). Γ(s)Γ(s+12)=21−2sπ Γ(2s)\Gamma(s) \Gamma(s + \tfrac12) = 2^{1 - 2s} \sqrt{\pi} \, \Gamma(2s). With s=12s = \tfrac12 it gives Γ(12)=π\Gamma(\tfrac12) = \sqrt{\pi}.

The functional equation of Mathlib used in RhOfRight is ζ(1−u)=2(2π)−uΓ(u)cos⁡(πu2)ζ(u),u∉{1,0,−1,−2,… },\zeta(1 - u) = 2 (2\pi)^{-u} \Gamma(u) \cos\Big(\frac{\pi u}{2}\Big) \zeta(u), \qquad u \notin \{1, 0, -1, -2, \dots\}, and the step uses Γ(u)≠0\Gamma(u) \ne 0 for Re⁡(u)>12\operatorname{Re}(u) > \tfrac12. By the reflection and duplication formulas the equation is equivalent to Λ(s)=Λ(1−s)\Lambda(s) = \Lambda(1 - s) for Λ(s)=π−s/2Γ(s/2)ζ(s)\Lambda(s) = \pi^{-s/2} \Gamma(s/2) \zeta(s).

Poisson summation, theta and Mellin

Theorem (Poisson summation formula). Let ff be a Schwartz function on ℝn\mathbb{R}^n with Fourier transform f^\hat f (see the Fourier transform). Then ∑m∈ℤnf(m)=∑k∈ℤnf^(k),\sum_{m \in \mathbb{Z}^n} f(m) = \sum_{k \in \mathbb{Z}^n} \hat f(k), and both series converge absolutely. For a lattice LL with dual lattice L∗L^*, ∑m∈Lf(m)=vol⁡(ℝn/L)−1∑k∈L∗f^(k)\sum_{m \in L} f(m) = \operatorname{vol}(\mathbb{R}^n/L)^{-1} \sum_{k \in L^*} \hat f(k).

Definition. For t>0t > 0 the theta function is θ(t)=∑n∈ℤe−πn2t\theta(t) = \sum_{n \in \mathbb{Z}} e^{-\pi n^2 t}. Poisson summation applied to the Gaussian e−πtx2e^{-\pi t x^2} gives θ(1/t)=t θ(t).\theta(1/t) = \sqrt{t} \, \theta(t).

Definition. Let f:(0,∞)→ℂf : (0, \infty) \to \mathbb{C} and s∈ℂs \in \mathbb{C} with xs−1f(x)x^{s - 1} f(x) integrable. The Mellin transform of ff at ss is ℳf(s)=∫0∞f(x)xs dx/x\mathcal{M}f(s) = \int_0^\infty f(x) x^{s} \, dx/x. For example, Γ\Gamma is the Mellin transform of e−xe^{-x}.

Theorem (Riemann, 1859). For Re⁡(s)>1\operatorname{Re}(s) > 1, Λ(s)=π−s/2Γ(s/2)ζ(s)=∫0∞θ(t)−12 ts/2 dtt.\Lambda(s) = \pi^{-s/2} \Gamma(s/2) \zeta(s) = \int_0^\infty \frac{\theta(t) - 1}{2} \, t^{s/2} \, \frac{dt}{t}. With θ(1/t)=t θ(t)\theta(1/t) = \sqrt t \, \theta(t), this integral continues Λ\Lambda to ℂ\mathbb{C} with simple poles at s=0,1s = 0, 1 only, and Λ(s)=Λ(1−s)\Lambda(s) = \Lambda(1 - s).

The theta function θ(g,h;Φ)=∑ξ(ω(g,h)Φ)(ξ)\theta(g, h; \Phi) = \sum_\xi (\omega(g, h)\Phi)(\xi) of the theta matching is, like θ(t)\theta(t), a sum of a Schwartz function over rational points. The proof siegelWeilStd_pre of the Siegel-Weil chapter gets ∑XTΦ(X)=θ(Φ)\sum_X T\Phi(X) = \theta(\Phi) in the Schrödinger model from a Poisson summation theorem.

Further reading

  • L. Ahlfors, Complex Analysis, Chapters 4 and 5 (Cauchy's theorem, zeros and poles, residues, the Gamma function).
  • E. Stein and R. Shakarchi, Complex Analysis, Chapter 6 (the Gamma and zeta functions).
  • E. C. Titchmarsh, The Theory of the Riemann Zeta-Function, Chapter II (the functional equation).
  • E. T. Whittaker and G. N. Watson, A Course of Modern Analysis, Chapter XII (the Gamma function).

The source of this page. Markdown. Something is wrong? Report it.