riemann

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

Zeta and L-functions

The Riemann zeta function, Dirichlet characters and L-functions, the Riemann Hypothesis and its generalization in the Mathlib forms, the explicit formula and theorems on primes, and the zeta functions of number fields.

Prerequisites: Class field theory, Complex analysis.

The Riemann zeta function and the Dirichlet L-functions are Dirichlet series with Euler products over the primes. The final theorem, riemannHypothesis, is about the zeros of ζ\zeta, and the overview reaches it from EveryCharacter, about the zeros of all Dirichlet L-functions. With the zeta functions of number fields they are also inputs, to the explicit formula of ZeroFreeAll and the class number formula of the CM side. Chapter 3 defines analytic continuation and Γ\Gamma, and Chapter 5 defines Hecke L-functions.

The Riemann zeta function

Definition. For Re⁡(s)>1\operatorname{Re}(s) > 1, the Riemann zeta function is the absolutely convergent series ζ(s)=∑n≥1n−s\zeta(s) = \sum_{n \geq 1} n^{-s}.

Theorem (Euler product; Euler, 1737). For Re⁡(s)>1\operatorname{Re}(s) > 1, ζ(s)=∏p(1−p−s)−1\zeta(s) = \prod_p (1 - p^{-s})^{-1} over the primes, and the product converges absolutely. So ζ(s)≠0\zeta(s) \neq 0 for Re⁡(s)>1\operatorname{Re}(s) > 1.

Theorem (Riemann, 1859). The function ζ\zeta extends to a meromorphic function on ℂ\mathbb{C} whose only pole is a simple pole at s=1s = 1 with residue 11: ζ(s)=1s−1+γ+O(s−1),\zeta(s) = \frac{1}{s - 1} + \gamma + O(s - 1), where γ=lim⁡N→∞(∑n≤N1/n−log⁡N)=0.5772…\gamma = \lim_{N \to \infty} \bigl(\sum_{n \leq N} 1/n - \log N\bigr) = 0.5772\ldots is the Euler-Mascheroni constant. Also ζ(0)=−1/2\zeta(0) = -1/2, and ζ(−2n)=0\zeta(-2n) = 0 for each integer n≥1n \geq 1.

Remark (the value at the pole in Mathlib). Mathlib defines riemannZeta on all of ℂ\mathbb{C}, and riemannZeta_one says that its value at s=1s = 1 is (γ−log⁡4π)/2≠0(\gamma - \log 4\pi)/2 \neq 0. A statement about the zeros of riemannZeta must exclude s=1s = 1 by hand or use that this value is not zero.

Theorem (Hadamard, de la Vallée Poussin, 1896). ζ(s)≠0\zeta(s) \neq 0 for Re⁡(s)=1\operatorname{Re}(s) = 1, hence for Re⁡(s)≥1\operatorname{Re}(s) \geq 1. This gives the prime number theorem #{p≤x}∼x/log⁡x\#\{p \leq x\} \sim x / \log x.

The functional equation and the trivial zeros

Definition (completed zeta function). Put Γℝ(s)=π−s/2Γ(s/2)\Gamma_\mathbb{R}(s) = \pi^{-s/2}\Gamma(s/2) and ξ(s)=Γℝ(s)ζ(s)\xi(s) = \Gamma_\mathbb{R}(s)\zeta(s).

Theorem (functional equation of the zeta function; Riemann, 1859). The function ξ\xi is meromorphic on ℂ\mathbb{C} with simple poles at s=0s = 0 and s=1s = 1 only, and ξ(s)=ξ(1−s)\xi(s) = \xi(1 - s). An equivalent form, riemannZeta_one_sub in Mathlib, is ζ(1−u)=2(2π)−u Γ(u)cos⁡(πu/2) ζ(u)for u∉{0,−1,−2,… }, u≠1.\zeta(1 - u) = 2(2\pi)^{-u}\,\Gamma(u)\cos(\pi u/2)\,\zeta(u) \qquad \text{for } u \notin \{0, -1, -2, \dots\},\ u \neq 1.

Definition (trivial zeros, critical strip). Since Γ(s/2)\Gamma(s/2) has poles at s=0,−2,−4,…s = 0, -2, -4, \dots and ξ\xi has no zeros with Re⁡(s)<0\operatorname{Re}(s) < 0, ζ\zeta vanishes at s=−2,−4,−6,…s = -2, -4, -6, \dots: these are the trivial zeros. The set 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 is the critical strip, and the zeros of ζ\zeta in it are the nontrivial zeros.

Proposition. The zeros of ζ\zeta are the trivial zeros and the nontrivial zeros. The nontrivial zeros are the zeros of ξ\xi; they are infinite in number and symmetric under s↦1−ss \mapsto 1 - s and s↦sˉs \mapsto \bar s.

It follows from ζ≠0\zeta \neq 0 for Re⁡(s)≥1\operatorname{Re}(s) \geq 1 and the functional equation, which leaves only the trivial zeros −2(n+1)-2(n + 1), n∈ℕn \in \mathbb{N}, in Re⁡(s)≤0\operatorname{Re}(s) \leq 0 (and ζ(0)=−1/2≠0\zeta(0) = -1/2 \neq 0). The step rh-of-right uses riemannZeta_one_sub and ζ(0)=−1/2\zeta(0) = -1/2 to move a zero ss with Re⁡(s)<1/2\operatorname{Re}(s) < 1/2 to the zero 1−s1 - s with Re⁡(1−s)>1/2\operatorname{Re}(1 - s) > 1/2.

The Riemann Hypothesis

Definition (strip form). The Riemann Hypothesis (Riemann, 1859) is the statement: each zero ss of ζ\zeta with 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 has Re⁡(s)=1/2\operatorname{Re}(s) = 1/2.

Definition (Mathlib form). Mathlib states the hypothesis as

def RiemannHypothesis : Prop :=
  ∀ (s : ℂ) (_ : riemannZeta s = 0) (_ : ¬∃ n : ℕ, s = -2 * (n + 1)) (_ : s ≠ 1),
    s.re = 1 / 2

It says that each zero ss of ζ\zeta other than the trivial zeros −2(n+1)-2(n+1) and the point 11 has Re⁡(s)=1/2\operatorname{Re}(s) = 1/2. Since riemannZeta 1 is not zero, the hypothesis s≠1s \neq 1 does no work.

Proposition. RiemannHypothesis is equivalent to the strip form: by the previous section, a zero that is neither trivial nor 11 lies in the critical strip, and each zero in the strip is neither.

The theorem riemannHypothesis proves RiemannHypothesis. It is the theorem rhOfAll, restated as riemannZeta_eq_zero_re in Main.lean.

Dirichlet characters

Fix an integer N≥1N \geq 1.

Definition. A Dirichlet character modulo NN is a group homomorphism χ:(ℤ/Nℤ)×→ℂ×\chi : (\mathbb{Z}/N\mathbb{Z})^\times \to \mathbb{C}^\times, extended to ℤ\mathbb{Z} by χ(n)=χ(n mod N)\chi(n) = \chi(n \bmod N) when gcd⁡(n,N)=1\gcd(n, N) = 1 and χ(n)=0\chi(n) = 0 otherwise. The number NN is the modulus. The trivial character modulo NN is 11 at each nn prime to NN. The order of χ\chi is its order in the group of the φ(N)\varphi(N) characters modulo NN. Since χ(−1)2=1\chi(-1)^2 = 1, the parity χ(−1)\chi(-1) is ±1\pm 1: χ\chi is even if χ(−1)=1\chi(-1) = 1 and odd if χ(−1)=−1\chi(-1) = -1.

Example. Modulo 44 there are two characters. The nontrivial one, χ−4\chi_{-4}, has χ−4(1)=1\chi_{-4}(1) = 1 and χ−4(3)=−1\chi_{-4}(3) = -1; it is odd, of order 22.

Definition (conductor). Let M∣NM \mid N. The character χ\chi modulo NN is induced from χ′\chi' modulo MM if χ(n)=χ′(n)\chi(n) = \chi'(n) for each nn prime to NN. The conductor of χ\chi is the least such MM, and χ\chi is primitive if its conductor is NN.

Each χ\chi is induced from exactly one primitive χ∗\chi^*, of modulus the conductor of χ\chi; the trivial character modulo 11 induces each trivial character. In Mathlib, χ.IsPrimitive means conductor χ = N. By Chapter 5, primitive characters of conductor NN are the finite order Hecke characters of ℚ\mathbb{Q} of conductor NN, and characters modulo NN are characters of Gal(ℚ(ζN)/ℚ)\mathrm{Gal}(\mathbb{Q}(\zeta_N)/\mathbb{Q}).

EveryCharacter is about each character of each modulus, primitive or not, while VWin and the branch ZeroFreeAll (step 7 of the theta matching chapter) are about primitive characters PChar.

Dirichlet L-functions

Definition. Let χ\chi be a Dirichlet character modulo NN. For Re⁡(s)>1\operatorname{Re}(s) > 1, the Dirichlet L-function is L(s,χ)=∑n≥1χ(n)n−s=∏p(1−χ(p)p−s)−1.L(s, \chi) = \sum_{n \geq 1} \chi(n) n^{-s} = \prod_p (1 - \chi(p) p^{-s})^{-1}.

Theorem. L(s,χ)L(s, \chi) extends to a meromorphic function on ℂ\mathbb{C}. It is entire if χ\chi is nontrivial; if χ\chi is trivial, its only pole is a simple pole at s=1s = 1. For N=1N = 1, L(s,χ)=ζ(s)L(s, \chi) = \zeta(s) (Davenport, ch. 9).

Mathlib defines DirichletCharacter.LFunction χ s on all of ℂ\mathbb{C} as N−s∑j=0N−1χ(j) ζ(s,j/N)N^{-s}\sum_{j=0}^{N-1} \chi(j)\,\zeta(s, j/N), with the Hurwitz zeta function ζ(s,a)\zeta(s, a). It equals the series for Re⁡(s)>1\operatorname{Re}(s) > 1 and is entire for χ≠1\chi \neq 1, and LFunction_modOne_eq says it is riemannZeta for N=1N = 1.

Proposition (imprimitive characters). Let χ\chi modulo NN be induced from the primitive χ∗\chi^*. Then L(s,χ)=L(s,χ∗)∏p∣N(1−χ∗(p)p−s).L(s, \chi) = L(s, \chi^*) \prod_{p \mid N} \bigl(1 - \chi^*(p) p^{-s}\bigr). For p∣Np \mid N with p∤cond⁡(χ)p \nmid \operatorname{cond}(\chi), the factor vanishes where ps=χ∗(p)p^s = \chi^*(p); since ∣χ∗(p)∣=1|\chi^*(p)| = 1, these are infinitely many points on Re⁡(s)=0\operatorname{Re}(s) = 0. For Re⁡(s)>0\operatorname{Re}(s) > 0, ∣χ∗(p)p−s∣<1|\chi^*(p)p^{-s}| < 1, so the factor has no zeros there.

Example. For the trivial character χ\chi modulo 22, L(s,χ)=(1−2−s)ζ(s)L(s, \chi) = (1 - 2^{-s})\zeta(s), which vanishes at s=2πik/log⁡2s = 2\pi i k/\log 2 for each k∈ℤk \in \mathbb{Z}.

Gauss sums and the functional equation

Definition. The Gauss sum of χ\chi modulo NN is τ(χ)=∑a=1Nχ(a)e2πia/N\tau(\chi) = \sum_{a = 1}^{N} \chi(a)e^{2\pi i a/N}. If χ\chi is primitive, ∣τ(χ)∣=N|\tau(\chi)| = \sqrt N.

Theorem (functional equation). Let χ\chi be primitive modulo NN with χ(−1)=(−1)κ\chi(-1) = (-1)^\kappa, κ∈{0,1}\kappa \in \{0, 1\}, and put Λ(s,χ)=(N/π)(s+κ)/2 Γ((s+κ)/2) L(s,χ)\Lambda(s, \chi) = (N/\pi)^{(s + \kappa)/2}\,\Gamma\bigl((s + \kappa)/2\bigr)\,L(s, \chi). Then Λ(s,χ)=τ(χ)iκN Λ(1−s,χˉ),\Lambda(s, \chi) = \frac{\tau(\chi)}{i^\kappa\sqrt N}\,\Lambda(1 - s, \bar\chi), and the constant has absolute value 11 (Davenport, ch. 9). In Mathlib this is IsPrimitive.completedLFunction_one_sub.

Since L(sˉ,χ)‾=L(s,χˉ)\overline{L(\bar s, \chi)} = L(s, \bar\chi), the zeros of L(s,χ)L(s, \chi) in the strip 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 are symmetric under s↦1−sˉs \mapsto 1 - \bar s. For primitive χ≠1\chi \neq 1, L(s,χ)≠0L(s, \chi) \neq 0 for Re⁡(s)≥1\operatorname{Re}(s) \geq 1, and the zeros outside the strip are the trivial zeros s=−κ,−κ−2,−κ−4,…s = -\kappa, -\kappa - 2, -\kappa - 4, \dots.

The step imprimitive proves the relation with LFunction_changeLevel and passes from primitive to all characters through the bound ∣χ∗(p)p−s∣<1|\chi^*(p)p^{-s}| < 1. The epsilon factors of ThetaIntegralsAgree contain local Gauss sums, the analogue of τ(χ)\tau(\chi) at a place.

Zeros and the Generalized Riemann Hypothesis

Definition (order of a zero). Let ff be holomorphic near ρ\rho and not identically 00. The order of a zero, or multiplicity mρm_\rho, of ff at ρ\rho is the least m≥0m \geq 0 with f(m)(ρ)≠0f^{(m)}(\rho) \neq 0. In sums over the zeros of an L-function, each zero counts mρm_\rho times.

Definition. The Generalized Riemann Hypothesis for Dirichlet L-functions is the statement: for each Dirichlet character χ\chi, each zero ss of L(s,χ)L(s, \chi) with 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 has Re⁡(s)=1/2\operatorname{Re}(s) = 1/2.

The proof states GRH in three forms with the Mathlib LFunction, for every N≥1N \geq 1 ([NeZero N]) and every χ\chi : DirichletCharacter ℂ N:

  • DirichletLNonvanishingRightOfHalf, the statement EveryCharacter: if 1/2<Re⁡(s)1/2 < \operatorname{Re}(s) and (χ≠1\chi \neq 1 or s≠1s \neq 1), then L(s,χ)≠0L(s, \chi) \neq 0. The condition removes only the pole of the trivial character.
  • DirichletLZerosOnLine: if 0<Re⁡(s)<10 < \operatorname{Re}(s) < 1 and L(s,χ)=0L(s, \chi) = 0, then Re⁡(s)=1/2\operatorname{Re}(s) = 1/2. The strict bound 0<Re⁡(s)0 < \operatorname{Re}(s) is needed because imprimitive characters have zeros on Re⁡(s)=0\operatorname{Re}(s) = 0.
  • PrimitiveDirichletLZerosOnLine: the same, for primitive χ\chi.

By the symmetry s↦1−sˉs \mapsto 1 - \bar s, no zeros with Re⁡(s)>1/2\operatorname{Re}(s) > 1/2 means no zeros off the line in the strip. Main.lean derives the three forms as LFunction_ne_zero_of_half_lt_re, LFunction_eq_zero_re and LFunction_eq_zero_re_of_isPrimitive.

The step all-chi proves EveryCharacter, and rh-of-every applies it at N=1N = 1, where L=ζL = \zeta. The field logDerOne of ThetaLocalInputs counts zeros with their orders: ∣L′/L(1,χ)−∑ρmρ/(1−ρ)∣≤Clog⁡(3q)\bigl|L'/L(1, \chi) - \sum_\rho m_\rho/(1 - \rho)\bigr| \leq C\log(3q) for primitive χ\chi modulo q>1q > 1, the sum over the zeros with 0<Re⁡(ρ)<10 < \operatorname{Re}(\rho) < 1 and ∣Im⁡(ρ)∣≤1|\operatorname{Im}(\rho)| \leq 1.

Primes and the explicit formula

Definition. The von Mangoldt function is Λ(n)=log⁡p\Lambda(n) = \log p if n=pkn = p^k with pp prime and k≥1k \geq 1, and Λ(n)=0\Lambda(n) = 0 otherwise. For Re⁡(s)>1\operatorname{Re}(s) > 1, −ζ′ζ(s)=∑n≥1Λ(n)n−s,−L′L(s,χ)=∑n≥1Λ(n)χ(n)n−s.-\frac{\zeta'}{\zeta}(s) = \sum_{n \geq 1} \Lambda(n) n^{-s}, \qquad -\frac{L'}{L}(s, \chi) = \sum_{n \geq 1} \Lambda(n)\chi(n) n^{-s}.

Theorem (explicit formula; von Mangoldt, 1895). Let ψ(x)=∑n≤xΛ(n)\psi(x) = \sum_{n \leq x}\Lambda(n). For x>1x > 1 not a prime power, ψ(x)=x−∑ρxρρ−log⁡2π−12log⁡(1−x−2),\psi(x) = x - \sum_\rho \frac{x^\rho}{\rho} - \log 2\pi - \tfrac12\log(1 - x^{-2}), where ρ\rho runs over the nontrivial zeros with multiplicity and the sum is lim⁡T→∞∑∣Im⁡ρ∣<T\lim_{T \to \infty}\sum_{|\operatorname{Im}\rho| < T}. In the form of Guinand (1948) and Weil (1952): let gg be even, smooth and compactly supported on ℝ\mathbb{R}, let h(r)=∫g(u)eiru duh(r) = \int g(u)e^{iru}\,du, and write ρ=1/2+itρ\rho = 1/2 + i t_\rho. Then ∑ρh(tρ)=h(i2)+h(−i2)−g(0)log⁡π+12π∫ℝh(r) Γ′Γ(14+ir2)dr−2∑n≥1Λ(n)n g(log⁡n).\sum_\rho h(t_\rho) = h\bigl(\tfrac{i}{2}\bigr) + h\bigl(-\tfrac{i}{2}\bigr) - g(0)\log\pi + \frac{1}{2\pi}\int_{\mathbb{R}} h(r)\,\frac{\Gamma'}{\Gamma}\bigl(\tfrac14 + \tfrac{ir}{2}\bigr)dr - 2\sum_{n \geq 1}\frac{\Lambda(n)}{\sqrt n}\,g(\log n).

So sums over primes control sums over zeros: a zero with Re⁡(ρ)>θ\operatorname{Re}(\rho) > \theta makes a prime sum grow like xRe⁡ρx^{\operatorname{Re}\rho}, and a bound O(xθ)O(x^\theta) on a suitable prime sum excludes zeros with Re⁡(s)>θ\operatorname{Re}(s) > \theta.

The branch ZeroFreeAll (step 7 of the theta matching chapter) uses Gz(y)=∑n≤yΛ(n)(y/n−1)−ylog⁡y+(1+γ)yG_z(y) = \sum_{n \leq y}\Lambda(n)(y/n - 1) - y\log y + (1 + \gamma)y, with Euler's constant γ\gamma, and its character twists. Its lemma FormB turns a bound Gm(x)≤xθG_m(x) \leq x^\theta on long windows into L(s,ψ′)≠0L(s, \psi') \neq 0 for Re⁡(s)>θ\operatorname{Re}(s) > \theta, and ZeroFreeAll from Green, VWin and HypsAlt assembles it.

Primes in progressions and in short intervals

Theorem (Dirichlet's theorem, 1837). Let q≥1q \geq 1 and gcd⁡(a,q)=1\gcd(a, q) = 1. There are infinitely many primes p≡a(modq)p \equiv a \pmod q, and ∑p≡a, p≤x1/p∼1φ(q)log⁡log⁡x\sum_{p \equiv a,\ p \leq x} 1/p \sim \frac{1}{\varphi(q)}\log\log x. The proof uses L(1,χ)≠0L(1, \chi) \neq 0 for each nontrivial χ\chi modulo qq.

Theorem (Linnik, 1944). There are absolute constants c,L>0c, L > 0 such that for each q≥2q \geq 2 and each aa prime to qq, the least prime p≡a(modq)p \equiv a \pmod q is at most c qLc\,q^L. One can take L=5L = 5 (Xylouris, 2011).

Theorem (primes in short intervals). For each θ>7/12\theta > 7/12, #{p:x<p≤x+xθ}∼xθ/log⁡x\#\{p : x < p \leq x + x^\theta\} \sim x^\theta/\log x as x→∞x \to \infty (Huxley, 1972). For large xx, [x−x0.525,x][x - x^{0.525}, x] contains a prime (Baker, Harman and Pintz, 2001). Under RH, for some absolute CC, each interval [x,x+Cxlog⁡x][x, x + C\sqrt x\log x] with x≥2x \geq 2 contains a prime (Cramér, 1920). See prime gaps.

The Chebotarev density theorem

Let L/KL/K be a finite Galois extension of number fields with group GG. Chapter 4 defines the Frobenius class Frob𝔭\mathrm{Frob}_{\mathfrak p} of a prime 𝔭\mathfrak p of KK unramified in LL.

Theorem (Chebotarev, 1926). For a conjugacy class C⊆GC \subseteq G, the primes 𝔭\mathfrak p of KK unramified in LL with Frob𝔭=C\mathrm{Frob}_{\mathfrak p} = C have natural density ∣C∣/∣G∣|C|/|G|: lim⁡x→∞#{𝔭:N(𝔭)≤x, Frob𝔭=C}#{𝔭:N(𝔭)≤x}=∣C∣∣G∣.\lim_{x \to \infty}\frac{\#\{\mathfrak p : N(\mathfrak p) \leq x,\ \mathrm{Frob}_{\mathfrak p} = C\}}{\#\{\mathfrak p : N(\mathfrak p) \leq x\}} = \frac{|C|}{|G|}. The same holds for the primes of degree one (those with N(𝔭)N(\mathfrak p) prime), since the others have density 00.

Example. For K=ℚK = \mathbb{Q} and L=ℚ(ζq)L = \mathbb{Q}(\zeta_q), Frobp=σp\mathrm{Frob}_p = \sigma_p, and the theorem gives Dirichlet's density 1/φ(q)1/\varphi(q).

ThetaLocalInputs states Linnik's theorem as linnik, the input hLit : LitThin of HypsAlt asks for three statements ShortIntervalPrimes, and chebotarevDegreeOne of UnitaryGroupThirdPlaceHypotheses is Chebotarev for primes of degree one.

Zeta functions of number fields

Let KK be a number field with r1r_1 real and r2r_2 complex places, discriminant dKd_K, class number hKh_K, regulator RKR_K, and wKw_K roots of unity (Chapter 4).

Definition. The Dedekind zeta function of KK is ζK(s)=∑𝔞N(𝔞)−s=∏𝔭(1−N(𝔭)−s)−1\zeta_K(s) = \sum_{\mathfrak a} N(\mathfrak a)^{-s} = \prod_{\mathfrak p}(1 - N(\mathfrak p)^{-s})^{-1} for Re⁡(s)>1\operatorname{Re}(s) > 1, the sum over the nonzero ideals of 𝒪K\mathcal{O}_K. It is the Hecke L-function of the trivial character.

Definition (partial zeta function). Let 𝔣\mathfrak f be a modulus and 𝔠\mathfrak c a class in the ray class group modulo 𝔣\mathfrak f. The partial zeta function is ζ(s,𝔠)=∑𝔞∈𝔠N(𝔞)−s\zeta(s, \mathfrak c) = \sum_{\mathfrak a \in \mathfrak c} N(\mathfrak a)^{-s}, over the integral ideals in 𝔠\mathfrak c. For a character χ\chi of the ray class group, L(s,χ)=∑𝔠χ(𝔠)ζ(s,𝔠)L(s, \chi) = \sum_{\mathfrak c}\chi(\mathfrak c)\zeta(s, \mathfrak c), and by orthogonality ζ(s,𝔠)\zeta(s, \mathfrak c) is an average of these L(s,χ)L(s, \chi). So it is meromorphic on ℂ\mathbb{C}, with a simple pole at s=1s = 1, and has a functional equation.

Theorem (class number formula; Dirichlet, 1839; Dedekind, 1877). ζK\zeta_K has a simple pole at s=1s = 1 with residue Res⁡s=1ζK(s)=2r1(2π)r2hKRKwK∣dK∣.\operatorname{Res}_{s = 1}\zeta_K(s) = \frac{2^{r_1}(2\pi)^{r_2}h_K R_K}{w_K\sqrt{|d_K|}}. For an imaginary quadratic field of discriminant −d-d, ζK(s)=ζ(s)L(s,χ−d)\zeta_K(s) = \zeta(s)L(s, \chi_{-d}) with the quadratic character χ−d\chi_{-d}, and for d>4d > 4 this gives L(1,χ−d)=πhK/dL(1, \chi_{-d}) = \pi h_K/\sqrt d.

Theorem (Kronecker limit formula; Kronecker, 1863). For τ=x+iy\tau = x + iy in the upper half plane and Re⁡(s)>1\operatorname{Re}(s) > 1, let E(τ,s)=∑(m,n)≠(0,0)ys/∣mτ+n∣2sE(\tau, s) = \sum_{(m, n) \neq (0, 0)} y^s/|m\tau + n|^{2s}. Then, with Dedekind's eta function η\eta, E(τ,s)=πs−1+2π(γ−log⁡2−log⁡(y ∣η(τ)∣2))+O(s−1).E(\tau, s) = \frac{\pi}{s - 1} + 2\pi\bigl(\gamma - \log 2 - \log(\sqrt y\,|\eta(\tau)|^2)\bigr) + O(s - 1). Summed over the ideal classes of an imaginary quadratic field, it computes L′/L(1,χ−d)L'/L(1, \chi_{-d}), which leads to the Chowla-Selberg formula.

Theorem (Siegel-Klingen; Klingen, 1962; Siegel, 1970). Let FF be totally real and 𝔠\mathfrak c a ray class modulo 𝔣\mathfrak f. For each integer k≥1k \geq 1, ζ(1−k,𝔠)\zeta(1 - k, \mathfrak c) is rational.

Theorem (Deligne-Ribet, 1980). These rational values satisfy ℓ\ell-adic congruences, which give the ℓ\ell-adic L-functions of FF. In one form: for an integral ideal 𝔠\mathfrak c prime to ℓ𝔣\ell\mathfrak f, each ray class 𝔟\mathfrak b modulo 𝔣\mathfrak f and each k≥1k \geq 1, ζ(1−k,𝔟𝔠)−N(𝔠)k ζ(1−k,𝔟)∈ℤℓ.\zeta(1 - k, \mathfrak b\mathfrak c) - N(\mathfrak c)^k\,\zeta(1 - k, \mathfrak b) \in \mathbb{Z}_\ell.

The CM divisibility chapter states the Chowla-Selberg mean 12log⁡2π−6log⁡d+6γ−6L′(1,χ)/L(1,χ)12\log 2\pi - 6\log d + 6\gamma - 6L'(1, \chi)/L(1, \chi), proved with the Kronecker limit formula (tendsto_realEisenstein_sub_pole) and the class number formula (LFunction_one_eq_classNumber). The record of CM inputs holds SiegelKlingenDeligneRibet and PartialZetaFunctionalEquation; only the second does work. The unitary hypotheses chapter uses a class number formula at each datum (StandardLattice.ClassNumberFormula₂), and the error term of the Weil bound chapter contains L′/L(1,χd)L'/L(1, \chi_d).

Further reading

  • H. Davenport, Multiplicative Number Theory, 3rd ed., chapters 1, 4 to 9, 12 and 17.
  • E. C. Titchmarsh, The Theory of the Riemann Zeta-Function, 2nd ed., chapters 2 and 3.
  • H. Iwaniec and E. Kowalski, Analytic Number Theory, chapters 5 (5.5, the explicit formula) and 18 (Linnik).
  • J. Neukirch, Algebraic Number Theory, chapter VII (Dedekind zeta, Hecke L-functions, Chebotarev, class number formula).
  • S. Lang, Elliptic Functions, chapter 20 (the Kronecker limit formula).

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