Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Zeta5: rational functionals and Hankel determinant polynomials

Definition
Zeta5_SourceConstruction

by tomasz · Oct 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

formalizationirrationalitynumber-theoryzeta5-mo271

For every natural number mmm, define Dm(t)=∏j=1m(t+j2)D_m(t)=\prod_{j=1}^m(t+j^2)Dm​(t)=∏j=1m​(t+j2). The rational linear functional on monomials is

μ(te)=(−1)eB2e+2(2e+3)(2e+4)(2e+5)24.\mu(t^e)=\frac{(-1)^e B_{2e+2}(2e+3)(2e+4)(2e+5)}{24}.μ(te)=24(−1)eB2e+2​(2e+3)(2e+4)(2e+5)​.

Let Hj(5)=∑v=1jv−5H_j^{(5)}=\sum_{v=1}^j v^{-5}Hj(5)​=∑v=1j​v−5. Its polynomial-valued pole functional is

μX((t+j2)−1)=j4(X−Hj(5))−14+12j.\mu_X\bigl((t+j^2)^{-1}\bigr)=j^4(X-H_j^{(5)})-\frac14+\frac1{2j}.μX​((t+j2)−1)=j4(X−Hj(5)​)−41​+2j1​.

Polynomial division by DKD_KDK​, together with the residues P(−j2)/DK′(−j2)P(-j^2)/D_K'(-j^2)P(−j2)/DK′​(−j2), defines μX(P/DK)\mu_X(P/D_K)μX​(P/DK​). For n∈Nn\in\mathbb Nn∈N, set K=40nK=40nK=40n, N=3nN=3nN=3n, h=37nh=37nh=37n and define

Gn(X)ij=μX(DN(t)6ti+j/DK(t)),Δn(X)=det⁡Gn(X),G_n(X)_{ij}=\mu_X\bigl(D_N(t)^6t^{i+j}/D_K(t)\bigr),\qquad \Delta_n(X)=\det G_n(X),Gn​(X)ij​=μX​(DN​(t)6ti+j/DK​(t)),Δn​(X)=detGn​(X),

where 0≤i,j<h0\le i,j<h0≤i,j<h. Finally put

Sn=K!2h4h−1N!12h∏i=1h−1(2i)!2,Fn(X)=SnΔn(X).S_n=\frac{K!^{2h}4^{h-1}}{N!^{12h}\prod_{i=1}^{h-1}(2i)!^2},\qquad F_n(X)=S_n\Delta_n(X).Sn​=N!12h∏i=1h−1​(2i)!2K!2h4h−1​,Fn​(X)=Sn​Δn​(X).

These are the explicit rational polynomials used throughout the project. The source module is retained verbatim, including its namespace and its total definitions at n=0n=0n=0; the exponent and endpoint written h−1h-1h−1 use natural-number subtraction.

Definition code
import Mathlib

/-!
# The Hankel determinant construction of the paper "ζ(5) is irrational"

This file *defines* (in Lean) the objects of Section 2.1 of the paper
"ζ(5) is irrational" (A. Fauzan, 17 September 2026):

* the polynomials `D_m(t) = ∏_{j=1}^m (t + j²)`;
* the `ℚ`-linear functional `μ_X`, given on monomials by (2.2) and on the simple poles
  `1/(t + j²)` by (2.3), extended to `P / D_K` by polynomial division and partial fractions;
* the Hankel matrix `G_K(X)` of (2.4), its determinant `Δ_K(X)`, the scalar `S_K` of (2.5)
  and `F_K = S_K Δ_K`.

Throughout, `K = 40 n`, `N = 3 n`, `h = 37 n` as in (2.1).

**Convention.** Two polynomial rings `ℚ[X]` occur: the paper's variable `t` (the argument of the
rational functions the functional is applied to) and the paper's indeterminate `X` (the
value of the functional is affine in `X`). Both are represented by `Polynomial ℚ`; the
docstrings say which is which.

Nothing is proved about these objects here; see `Apery.MainEstimate` for the main estimate.
-/

open Polynomial Finset

namespace Apery

/-- `D_m(t) = ∏_{j=1}^m (t + j²)`, a monic polynomial in `t`. -/
noncomputable def D (m : ℕ) : ℚ[X] := ∏ j ∈ Icc 1 m, (X + C ((j : ℚ) ^ 2))

/-- The polynomial moments (2.2): `μ(t^e) = (-1)^e B_{2e+2} (2e+3)(2e+4)(2e+5) / 24`.
Mathlib's `bernoulli` uses the convention `B₁ = -1/2`, as does the paper. -/
def μmono (e : ℕ) : ℚ :=
  (-1) ^ e * _root_.bernoulli (2 * e + 2) * ((2 * e + 3) * (2 * e + 4) * (2 * e + 5)) / 24

/-- `μ` on polynomials in `t`, extended `ℚ`-linearly from the monomials. -/
noncomputable def μpoly (P : ℚ[X]) : ℚ := P.sum fun e c => c * μmono e

/-- The generalized harmonic number `H_j^{(5)} = ∑_{v=1}^j v⁻⁵`. -/
def H5 (j : ℕ) : ℚ := ∑ v ∈ Icc 1 j, 1 / (v : ℚ) ^ 5

/-- The pole values (2.3), as polynomials in the indeterminate `X`:
`μ_X(1/(t + j²)) = j⁴ (X - H_j^{(5)}) - 1/4 + 1/(2j)`. -/
noncomputable def poleValue (j : ℕ) : ℚ[X] :=
  C ((j : ℚ) ^ 4) * (X - C (H5 j)) - C (1 / 4) + C (1 / (2 * (j : ℚ)))

/-- `μ_X(P(t) / D_K(t))` for a polynomial `P` in `t`, as a polynomial in `X`.
Since `D_K` is monic with the simple roots `t = -j²` (`1 ≤ j ≤ K`),
`P / D_K = (P /ₘ D_K) + ∑_j (P(-j²) / D_K'(-j²)) / (t + j²)`, and `μ_X` is applied termwise. -/
noncomputable def μX (K : ℕ) (P : ℚ[X]) : ℚ[X] :=
  C (μpoly (P /ₘ D K)) +
    ∑ j ∈ Icc 1 K,
      C (P.eval (-(j : ℚ) ^ 2) / (derivative (D K)).eval (-(j : ℚ) ^ 2)) * poleValue j

/-- The `h × h` Hankel matrix `G_K(X)` of (2.4), whose `(i, j)` entry is
`μ_X (D_N(t)^6 t^(i+j) / D_K(t))`, with `K = 40 n`, `N = 3 n`, `h = 37 n`. -/
noncomputable def G (n : ℕ) : Matrix (Fin (37 * n)) (Fin (37 * n)) ℚ[X] :=
  fun i j => μX (40 * n) (D (3 * n) ^ 6 * X ^ ((i : ℕ) + (j : ℕ)))

/-- `Δ_K(X) = det G_K(X)`, a polynomial in `X` (of degree `h`, by (2.9)). -/
noncomputable def Δ (n : ℕ) : ℚ[X] := (G n).det

/-- The scalar `S_K` of (2.5). -/
noncomputable def S (n : ℕ) : ℚ :=
  ((40 * n).factorial : ℚ) ^ (2 * (37 * n)) * 4 ^ (37 * n - 1) /
    (((3 * n).factorial : ℚ) ^ (12 * (37 * n)) *
      ∏ i ∈ Icc 1 (37 * n - 1), ((2 * i).factorial : ℚ) ^ 2)

/-- `F_K = S_K Δ_K` of (2.5). -/
noncomputable def F (n : ℕ) : ℚ[X] := C (S n) * Δ n

end Apery
Source
https://github.com/mo271/Zeta5/blob/7fe736760f4b96bfdb4334b68e3b3124ecbe10b0/Apery/Construction.lean#L1-L73
Read-back

What the Lean code literally says, in plain math · GPT-6 (Codex independent auditor)

D. For each natural number mmm, including m=0m=0m=0, this defines the rational polynomial Dm(t)=∏j=1m(t+j2)D_m(t)=\prod_{j=1}^{m}(t+j^2)Dm​(t)=∏j=1m​(t+j2), where each natural number jjj is regarded as a rational number inside the polynomial. The product at m=0m=0m=0 is empty, so D0(t)=1D_0(t)=1D0​(t)=1.

μmono. For every natural number eee, this defines the rational number βe=(−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\beta_e=(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24βe​=(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24, where BkB_kBk​ is the rational Bernoulli number with convention B1=−1/2B_1=-1/2B1​=−1/2 and the natural-number factors are interpreted in Q\mathbb QQ. The index e=0e=0e=0 is included, giving β0=5/12\beta_0=5/12β0​=5/12.

μpoly. For every rational polynomial P(t)P(t)P(t), this defines the rational number ∑e∈supp⁡(P)[te]P (−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\sum_{e\in\operatorname{supp}(P)}[t^e]P\,(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24∑e∈supp(P)​[te]P(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24, where supp⁡(P)\operatorname{supp}(P)supp(P) is the finite set of exponents having nonzero coefficient, [te]P[t^e]P[te]P is the coefficient of tet^ete, and BkB_kBk​ are the Bernoulli numbers with B1=−1/2B_1=-1/2B1​=−1/2. For the zero polynomial the sum is empty and the value is 000.

H5. For every natural number jjj, this defines the rational number Hj=∑v=1j1/v5H_j=\sum_{v=1}^{j}1/v^5Hj​=∑v=1j​1/v5. The indices are positive natural numbers interpreted in Q\mathbb QQ; when j=0j=0j=0 the sum is empty and H0=0H_0=0H0​=0.

poleValue. For each natural number jjj, this defines the rational polynomial j4(x−∑v=1j1/v5)−1/4+1/(2j)j^4\bigl(x-\sum_{v=1}^{j}1/v^5\bigr)-1/4+1/(2j)j4(x−∑v=1j​1/v5)−1/4+1/(2j) in the output variable xxx. There is no hypothesis j>0j>0j>0: at j=0j=0j=0, the sum is empty and division by zero in Q\mathbb QQ gives 1/(2j)=01/(2j)=01/(2j)=0, so the defined polynomial is the constant −1/4-1/4−1/4.

μX. For every natural number KKK and every rational polynomial P(t)P(t)P(t), this defines the rational polynomial MK(P;x)M_K(P;x)MK​(P;x) by the following formula. Here Dm(t)=∏r=1m(t+r2)D_m(t)=\prod_{r=1}^{m}(t+r^2)Dm​(t)=∏r=1m​(t+r2) for m∈Nm\in\mathbb Nm∈N, Hj=∑v=1jv−5H_j=\sum_{v=1}^{j}v^{-5}Hj​=∑v=1j​v−5 for j∈Nj\in\mathbb Nj∈N, and βe=(−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\beta_e=(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24βe​=(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24 for e∈Ne\in\mathbb Ne∈N, where BkB_kBk​ are the rational Bernoulli numbers with B1=−1/2B_1=-1/2B1​=−1/2. For K∈NK\in\mathbb NK∈N and P∈Q[t]P\in\mathbb Q[t]P∈Q[t], let qK,Pq_{K,P}qK,P​ be the polynomial quotient on division of PPP by the monic polynomial DKD_KDK​, and put

MK(P;x)=∑e[te]qK,P βe+∑j=1KP(−j2)DK′(−j2)(j4(x−Hj)−14+12j).\begin{aligned} M_K(P;x)&=\sum_e [t^e]q_{K,P}\,\beta_e\\ &\quad+\sum_{j=1}^{K}\frac{P(-j^2)}{D_K'(-j^2)} \bigl(j^4(x-H_j)-\frac14+\frac{1}{2j}\bigr). \end{aligned}MK​(P;x)​=e∑​[te]qK,P​βe​+j=1∑K​DK′​(−j2)P(−j2)​(j4(x−Hj​)−41​+2j1​).​

The first sum is over the finitely many nonzero coefficients of qK,Pq_{K,P}qK,P​; [te][t^e][te] denotes a coefficient and the prime denotes formal differentiation. The variable ttt is the input polynomial variable and xxx is the output polynomial variable. Empty products are 111, empty sums are 000, and rational division by zero has value 000. In particular, D0=1D_0=1D0​=1, H0=0H_0=0H0​=0, and M0(P;x)=∑e[te]P βeM_0(P;x)=\sum_e[t^e]P\,\beta_eM0​(P;x)=∑e​[te]Pβe​. For P=0P=0P=0 this gives the zero polynomial. This declaration assigns the displayed value to the pair (K,P)(K,P)(K,P).

G. For every natural number nnn, this defines a square matrix over Q[x]\mathbb Q[x]Q[x] with row and column indices i,j∈{0,…,37n−1}i,j\in\{0,\ldots,37n-1\}i,j∈{0,…,37n−1} and entries Gn(x)ij=M40n(D3n(t)6ti+j;x)G_n(x)_{ij}=M_{40n}(D_{3n}(t)^6t^{i+j};x)Gn​(x)ij​=M40n​(D3n​(t)6ti+j;x). Here Dm(t)=∏r=1m(t+r2)D_m(t)=\prod_{r=1}^{m}(t+r^2)Dm​(t)=∏r=1m​(t+r2) for m∈Nm\in\mathbb Nm∈N, Hj=∑v=1jv−5H_j=\sum_{v=1}^{j}v^{-5}Hj​=∑v=1j​v−5 for j∈Nj\in\mathbb Nj∈N, and βe=(−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\beta_e=(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24βe​=(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24 for e∈Ne\in\mathbb Ne∈N, where BkB_kBk​ are the rational Bernoulli numbers with B1=−1/2B_1=-1/2B1​=−1/2. For K∈NK\in\mathbb NK∈N and P∈Q[t]P\in\mathbb Q[t]P∈Q[t], let qK,Pq_{K,P}qK,P​ be the polynomial quotient on division of PPP by the monic polynomial DKD_KDK​, and put

MK(P;x)=∑e[te]qK,P βe+∑j=1KP(−j2)DK′(−j2)(j4(x−Hj)−14+12j).\begin{aligned} M_K(P;x)&=\sum_e [t^e]q_{K,P}\,\beta_e\\ &\quad+\sum_{j=1}^{K}\frac{P(-j^2)}{D_K'(-j^2)} \bigl(j^4(x-H_j)-\frac14+\frac{1}{2j}\bigr). \end{aligned}MK​(P;x)​=e∑​[te]qK,P​βe​+j=1∑K​DK′​(−j2)P(−j2)​(j4(x−Hj​)−41​+2j1​).​

The first sum is over the finitely many nonzero coefficients of qK,Pq_{K,P}qK,P​; [te][t^e][te] denotes a coefficient and the prime denotes formal differentiation. The variable ttt is the input polynomial variable and xxx is the output polynomial variable. Empty products are 111, empty sums are 000, and rational division by zero has value 000. In particular, D0=1D_0=1D0​=1, H0=0H_0=0H0​=0, and M0(P;x)=∑e[te]P βeM_0(P;x)=\sum_e[t^e]P\,\beta_eM0​(P;x)=∑e​[te]Pβe​. At n=0n=0n=0 both index sets are empty, so the matrix has no entries.

Δ. For every natural number nnn, this defines the rational polynomial Δn(x)=det⁡[M40n(D3n(t)6ti+j;x)]0≤i,j<37n\Delta_n(x)=\det\bigl[M_{40n}(D_{3n}(t)^6t^{i+j};x)\bigr]_{0\le i,j<37n}Δn​(x)=det[M40n​(D3n​(t)6ti+j;x)]0≤i,j<37n​. Here Dm(t)=∏r=1m(t+r2)D_m(t)=\prod_{r=1}^{m}(t+r^2)Dm​(t)=∏r=1m​(t+r2) for m∈Nm\in\mathbb Nm∈N, Hj=∑v=1jv−5H_j=\sum_{v=1}^{j}v^{-5}Hj​=∑v=1j​v−5 for j∈Nj\in\mathbb Nj∈N, and βe=(−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\beta_e=(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24βe​=(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24 for e∈Ne\in\mathbb Ne∈N, where BkB_kBk​ are the rational Bernoulli numbers with B1=−1/2B_1=-1/2B1​=−1/2. For K∈NK\in\mathbb NK∈N and P∈Q[t]P\in\mathbb Q[t]P∈Q[t], let qK,Pq_{K,P}qK,P​ be the polynomial quotient on division of PPP by the monic polynomial DKD_KDK​, and put

MK(P;x)=∑e[te]qK,P βe+∑j=1KP(−j2)DK′(−j2)(j4(x−Hj)−14+12j).\begin{aligned} M_K(P;x)&=\sum_e [t^e]q_{K,P}\,\beta_e\\ &\quad+\sum_{j=1}^{K}\frac{P(-j^2)}{D_K'(-j^2)} \bigl(j^4(x-H_j)-\frac14+\frac{1}{2j}\bigr). \end{aligned}MK​(P;x)​=e∑​[te]qK,P​βe​+j=1∑K​DK′​(−j2)P(−j2)​(j4(x−Hj​)−41​+2j1​).​

The first sum is over the finitely many nonzero coefficients of qK,Pq_{K,P}qK,P​; [te][t^e][te] denotes a coefficient and the prime denotes formal differentiation. The variable ttt is the input polynomial variable and xxx is the output polynomial variable. Empty products are 111, empty sums are 000, and rational division by zero has value 000. In particular, D0=1D_0=1D0​=1, H0=0H_0=0H0​=0, and M0(P;x)=∑e[te]P βeM_0(P;x)=\sum_e[t^e]P\,\beta_eM0​(P;x)=∑e​[te]Pβe​. When n=0n=0n=0 the matrix is empty and its determinant is 111, so Δ0(x)=1\Delta_0(x)=1Δ0​(x)=1.

S. For every natural number nnn, this defines the rational scalar sns_nsn​ by the following formula. Put rn=max⁡(37n−1,0)r_n=\max(37n-1,0)rn​=max(37n−1,0), the truncated natural-number subtraction used in the code, and sn=((40n)!)2(37n)4rn((3n)!)12(37n)∏i=1rn((2i)!)2∈Qs_n=\frac{((40n)!)^{2(37n)}4^{r_n}}{((3n)!)^{12(37n)}\prod_{i=1}^{r_n}((2i)!)^2}\in\mathbb Qsn​=((3n)!)12(37n)∏i=1rn​​((2i)!)2((40n)!)2(37n)4rn​​∈Q. All factorials have natural-number arguments and are then embedded in Q\mathbb QQ; the finite product is over natural numbers 1≤i≤rn1\le i\le r_n1≤i≤rn​. In particular r0=0r_0=0r0​=0, the product at n=0n=0n=0 is empty, all zero exponents give 111, and s0=1s_0=1s0​=1.

F. For every natural number nnn, this defines the rational polynomial FnF_nFn​ as the product of the constant polynomial with value sns_nsn​ and the determinant polynomial Δn\Delta_nΔn​. The rational polynomial in this assertion is Fn(x)=sndet⁡[M40n(D3n(t)6ti+j;x)]0≤i,j<37nF_n(x)=s_n\det\bigl[M_{40n}(D_{3n}(t)^6t^{i+j};x)\bigr]_{0\le i,j<37n}Fn​(x)=sn​det[M40n​(D3n​(t)6ti+j;x)]0≤i,j<37n​. Put rn=max⁡(37n−1,0)r_n=\max(37n-1,0)rn​=max(37n−1,0), the truncated natural-number subtraction used in the code, and sn=((40n)!)2(37n)4rn((3n)!)12(37n)∏i=1rn((2i)!)2∈Qs_n=\frac{((40n)!)^{2(37n)}4^{r_n}}{((3n)!)^{12(37n)}\prod_{i=1}^{r_n}((2i)!)^2}\in\mathbb Qsn​=((3n)!)12(37n)∏i=1rn​​((2i)!)2((40n)!)2(37n)4rn​​∈Q. Here Dm(t)=∏r=1m(t+r2)D_m(t)=\prod_{r=1}^{m}(t+r^2)Dm​(t)=∏r=1m​(t+r2) for m∈Nm\in\mathbb Nm∈N, Hj=∑v=1jv−5H_j=\sum_{v=1}^{j}v^{-5}Hj​=∑v=1j​v−5 for j∈Nj\in\mathbb Nj∈N, and βe=(−1)eB2e+2(2e+3)(2e+4)(2e+5)/24\beta_e=(-1)^eB_{2e+2}(2e+3)(2e+4)(2e+5)/24βe​=(−1)eB2e+2​(2e+3)(2e+4)(2e+5)/24 for e∈Ne\in\mathbb Ne∈N, where BkB_kBk​ are the rational Bernoulli numbers with B1=−1/2B_1=-1/2B1​=−1/2. For K∈NK\in\mathbb NK∈N and P∈Q[t]P\in\mathbb Q[t]P∈Q[t], let qK,Pq_{K,P}qK,P​ be the polynomial quotient on division of PPP by the monic polynomial DKD_KDK​, and put

MK(P;x)=∑e[te]qK,P βe+∑j=1KP(−j2)DK′(−j2)(j4(x−Hj)−14+12j).\begin{aligned} M_K(P;x)&=\sum_e [t^e]q_{K,P}\,\beta_e\\ &\quad+\sum_{j=1}^{K}\frac{P(-j^2)}{D_K'(-j^2)} \bigl(j^4(x-H_j)-\frac14+\frac{1}{2j}\bigr). \end{aligned}MK​(P;x)​=e∑​[te]qK,P​βe​+j=1∑K​DK′​(−j2)P(−j2)​(j4(x−Hj​)−41​+2j1​).​

The first sum is over the finitely many nonzero coefficients of qK,Pq_{K,P}qK,P​; [te][t^e][te] denotes a coefficient and the prime denotes formal differentiation. The variable ttt is the input polynomial variable and xxx is the output polynomial variable. Empty products are 111, empty sums are 000, and rational division by zero has value 000. In particular, D0=1D_0=1D0​=1, H0=0H_0=0H0​=0, and M0(P;x)=∑e[te]P βeM_0(P;x)=\sum_e[t^e]P\,\beta_eM0​(P;x)=∑e​[te]Pβe​. At n=0n=0n=0 the determinant is that of the empty matrix and equals 111, and s0=F0=1s_0=F_0=1s0​=F0​=1.

Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by tomasz · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me