betaFeedback

WorksMathematics

Many odd-power reciprocal sums of Lucas sequences are irrational

Abstract

Let a≥3a\ge3, let U0=0U_0=0, U1=1U_1=1, Un+1=aUn−Un−1U_{n+1}=aU_n-U_{n-1}, and let δ=±1\delta=\pm1. We prove that 11 and the numbers ∑n≥1δn/Unl\sum_{n\ge1}\delta^n/U_n^l, l=1,3,…,A−1l=1,3,\dots,A-1, span a Q\mathbb{Q}-vector space of dimension at least (π/12+π2+o(1))A(\pi/\sqrt{12+\pi^2}+o(1))\sqrt A. So infinitely many of these sums with ll odd are irrational, for instance infinitely many of ∑n1/F2nl\sum_n1/F_{2n}^l. With the known linear independence of 11, ∑nδn/Un\sum_n\delta^n/U_n and ∑nδn/Un3\sum_n\delta^n/U_n^3, one of ∑nδn/Unl\sum_n\delta^n/U_n^l, l=5,7,…,23l=5,7,\dots,23, lies outside the span of these three numbers. The proof adapts the author's construction for odd qq-zeta values at real quadratic units, with a weighted bilateral sum. No single sum such as ∑n1/Un5\sum_n1/U_n^5 is decided, and sequences of norm −1-1, such as the Fibonacci numbers, are not covered.

Introduction

Fix an integer a≥3a \ge3 and let U0=0U_0=0, U1=1U_1=1, Un+1=aUn−Un−1U_{n+1}=aU_n-U_{n-1}. Then Un=(ηn−η−n)/(η−η−1)U_n=(\eta^n-\eta^{-n})/(\eta-\eta^{-1}) with η=(a+a2−4)/2>1\eta=(a+\sqrt{a^2-4})/2>1, a unit of norm +1+1 of K=Q(a2−4)K=\mathbb{Q}(\sqrt{a^2-4}), and every unit η>1\eta>1 of norm +1+1 of a real quadratic field arises in this way, with a=η+η−1a=\eta+\eta^{-1}. Examples are Un=F2nU_n=F_{2n} (a=3a=3), Un=Fkn/FkU_n=F_{kn}/F_k for even kk (a=Lka=L_k) and Un=P2n/2U_n=P_{2n}/2, where PnP_n are the Pell numbers (a=6a=6). Let δ∈{1,−1}\delta\in\{1,-1\} and, for even A≥4A\ge4 and 1≤r<A/21\le r<A/2,

τ(A,r)=4π2r(A−2r)−12A12A+π2(A+4r2).\tau(A,r)=\frac{4\pi^2r(A-2r)-12A}{12A+\pi^2(A+4r^2)}.

Theorem 1.1. For every even A≥4A\ge4 and every integer rr with 1≤r<A/21\le r<A/2,

dim⁡Q(Q+∑1≤l≤A−1l oddQ∑n≥1δnUnl)≥1+τ(A,r).\dim_{\mathbb{Q}}\left(\mathbb{Q}+\sum_{\substack{1\le l\le A-1\\ l\ \mathrm{odd}}}\mathbb{Q}\sum_{n\ge1}\frac{\delta^n}{U_n^l}\right)\ge1+\tau(A,r).

As A→∞A\to\infty through even integers, this dimension is at least (π/12+π2+o(1))A=(0.67178…+o(1))A(\pi/\sqrt{12+\pi^2}+o(1))\sqrt{A}=(0.67178\ldots+o(1))\sqrt{A}. In particular, at least ⌈1+τ(A,r)⌉−1\lceil1+\tau(A,r)\rceil-1 of the numbers ∑nδn/Unl\sum_n\delta^n/U_n^l with odd l≤A−1l\le A-1 are irrational, and infinitely many of the numbers ∑nδn/Unl\sum_n\delta^n/U_n^l, ll odd, are irrational.

2020 Mathematics Subject Classification. Primary 11J72; Secondary 11B39, 11R11, 33D15.

Key words and phrases. Lucas sequences, Fibonacci numbers, reciprocal sums, Lambert series, irrationality, linear independence, Nesterenko’s criterion.

The function τ\tau is that of the author’s dimension bound for odd qq-zeta values at real quadratic units [12 Theorem 1.1]. The bound depends neither on aa nor on δ\delta.

Corollary 1.2. Let AkA_k be the smallest even AA with max⁡r(1+τ(A,r))>k−1\max_r(1+\tau(A,r))>k-1, so that Theorem 1.1 gives dim⁡≥k\dim\geq k for the sums with odd l≤Ak−1l\leq A_k-1. For 2≤k≤112\leq k\leq11 the values are those of Table 1, attained at r=k−1r=k-1; at every even A<AkA<A_k and every rr, 1+τ(A,r)≤k−11+\tau(A,r)\leq k-1.

kkAkA_krr1+τ(Ak,r)1+\tau(A_k,r)kkAkA_krr1+τ(Ak,r)1+\tau(A_k,r)
2411.243878866.0650
31222.1601811877.0518
42433.0948915288.0296
54044.03581019099.0012
66255.0635112341010.0097

Table 1. Thresholds of Corollary 1.2.

For the first new cases we use the following result, proved with a Lean formalisation in [13]: the numbers 11, ∑nδn/Un\sum_n \delta^n/U_n and ∑nδn/Un3\sum_n \delta^n/U_n^3 are linearly independent over Q\mathbb{Q}. Let V3V_3 be their Q\mathbb{Q}-span, so dim⁡QV3=3\dim_{\mathbb{Q}}V_3=3.

Corollary 1.3. (1) At least one of the numbers ∑nδn/Unl\sum_n\delta^n/U_n^l with l∈{5,7,…,23}l\in\{5,7,\ldots,23\} lies outside V3V_3; in particular it is irrational. (2) At least two of the numbers ∑nδn/Unl\sum_n\delta^n/U_n^l, l∈{5,7,…,39}l\in\{5,7,\ldots,39\}, are linearly independent modulo V3V_3. (3) In general, the ∑nδn/Unl\sum_n\delta^n/U_n^l with ll odd, 5≤l≤Ak−15\leq l\leq A_k-1, span a space of dimension at least k−3k-3 modulo V3V_3.

At A=12A=12 Theorem 1.1 gives dim⁡≥3\dim\geq3, which is contained in [13]; part (1), from A=24A=24, is the first finite statement that goes beyond it.

Corollary 1.4. Theorem 1.1 and Corollaries 1.2 and 1.3 hold for ∑n1/F2n\sum_n1/F_{2n} and ∑n(−1)n/F2n\sum_n(-1)^n/F_{2n}, for ∑nδn/Fkn\sum_n\delta^n/F_{kn} with k≥2k\geq2 even, and for ∑nδn/P2n\sum_n\delta^n/P_{2n}. For instance, one of ∑n1/F2n\sum_n1/F_{2n}, l∈{5,7,…,23}l\in\{5,7,\ldots,23\}, lies outside Q+Q∑n1/F2n+Q∑n1/F3n\mathbb{Q}+\mathbb{Q}\sum_n1/F_{2n}+\mathbb{Q}\sum_n1/F_{3n}.

The Lambert form

For 0<∣ν∣<10<|\nu|<1 and odd j≥1j\geq1 put

Wj(ν)=∑m≥1m oddmj−1νm1−νm=ζν(j)−2j−1ζν2(j),ζq(j)=∑m≥1mj−1qm1−qm.W_j(\nu)=\sum_{\substack{m\geq1\\m\ \mathrm{odd}}}\frac{m^{j-1}\nu^m}{1-\nu^m}=\zeta_\nu(j)-2^{j-1}\zeta_{\nu^2}(j),\qquad\zeta_q(j)=\sum_{m\geq1}\frac{m^{j-1}q^m}{1-q^m}.

Theorem 1.5. Let d≥2d\geq2 be squarefree, K=Q(d)K=\mathbb{Q}(\sqrt{d}), and let ν∈OK×\nu\in\mathcal{O}_K^\times have norm +1+1 and 0<∣ν∣<10<|\nu|<1. For every even A≥4A\geq4 and every integer rr with 1≤r<A/21\leq r<A/2,

dim⁡Q(Q+∑1≤j≤A−1j oddQd Wj(ν))≥1+τ(A,r).\dim_{\mathbb{Q}}\left(\mathbb{Q}+\sum_{\substack{1\leq j\leq A-1\\j\ \mathrm{odd}}}\mathbb{Q}\sqrt{d}\,W_j(\nu)\right)\geq1+\tau(A,r).

Theorems 1.1 and 1.5 are equivalent by an elementary change of basis (Lemma 2.1, Remark 2.2). Theorem 1.5 does not follow from [12 Theorem 1.1], which concerns 1,d ζq(3),…,d ζq(A−1)1,\sqrt{d}\,\zeta_q(3),\ldots,\sqrt{d}\,\zeta_q(A-1) at one unit qq: the Wj(ν)W_j(\nu) mix the units ν\nu and ν2\nu^2, and j=1j=1 occurs.

The method

In [12] the rational function Rn(T;q)R_n(T;q) of Krattenthaler, Rivoal and Zudilin [11 §3] is summed over qkq^k and q−n−kq^{-n-k}, k≥1k\geq1, with opposite signs. This gives linear forms in the odd ζq(l)\zeta_q(l) whose coefficients are invariant or anti-invariant under q↦1/qq\mapsto1/q, up to a power of qq. At a unit of norm +1+1, where Galois conjugation is q↦1/qq\mapsto1/q, integer forms in the numbers d ζq(l)\sqrt{d}\,\zeta_q(l) are then small at both real places, and Nesterenko’s criterion applies. We keep the rational function and its partial-fraction coefficients, at q=ν2q=\nu^2, and change only the sum:

Sn(ν)=∑k≥1νkR(ν2k)−∑k≥1ν−n−kR(ν−2n−2k).S_n(\nu)=\sum_{k\geq1}\nu^kR(\nu^{2k})-\sum_{k\geq1}\nu^{-n-k}R(\nu^{-2n-2k}).

The weights turn the qq-zeta values into the Wj(ν)W_j(\nu), the minus sign removes the even jj, and j=1j=1 survives (Lemma 4.1); the reciprocity under ν↦1/ν\nu\mapsto1/\nu holds for all odd jj, j=1j=1 included (Lemma 4.2). Coefficients, decay and denominators change only by factors eO(n)e^{O(n)}, so τ\tau is unchanged.

We cite [12] for the properties of RnR_n and prove what is new. Nothing is formalised. Besides [12] the proofs use Nesterenko’s criterion [16, 8, 11] and the asymptotic ∑ℓ≤nϕE(ℓ)=3n2/π2+O(nlog⁡n)\sum_{\ell\leq n}\phi_E(\ell)=3n^2/\pi^2+O(n\log n) [9 Theorem 330]; Corollary 1.3 also uses [13].

The change of basis

Lemma 2.1. Let η>1\eta>1 be a unit of norm +1+1 of a real quadratic field, with trace aa and Lucas sequence (Un)(U_n), let δ∈{1,−1}\delta\in\{1,-1\}, ν=δ/η\nu=\delta/\eta, and let dd be the squarefree part of a2−4a^2-4. For odd l≥1l\geq1,

∑n≥1δnUnl=(η−η−1)lVl,Vl:=∑n≥1νnl(1−ν2n)l=∑m≥1m oddPl(m)νm1−νm.(1)\sum_{n\geq1}\frac{\delta^n}{U_n^l}=(\eta-\eta^{-1})^lV_l,\qquad V_l:=\sum_{n\geq1}\frac{\nu^{nl}}{(1-\nu^{2n})^l}=\sum_{\substack{m\geq1\\m\ \mathrm{odd}}}P_l(m)\frac{\nu^m}{1-\nu^m}. \tag*{(1)}

where Pl(m)=∏i(m2−i2)/(2l−1(l−1)!)P_l(m)=\prod_i(m^2-i^2)/(2^{l-1}(l-1)!), the product over odd ii with 1≤i≤l−21\leq i\leq l-2. Hence Vl=∑j≤lj oddal,jWj(ν)V_l=\sum_{\substack{j\leq l\\j\ \mathrm{odd}}}a_{l,j}W_j(\nu), over odd j≤lj\leq l, with al,j=[mj−1]Pl∈Qa_{l,j}=[m^{j-1}]P_l\in\mathbb{Q} and al,l≠0a_{l,l}\ne0; and for every odd L≥1L\geq1,

span⁡Q(1,∑nδnUn,…,∑nδnUnL)=span⁡Q(1,d W1(ν),…,d WL(ν)).\operatorname{span}_{\mathbb{Q}}\left(1,\sum_n\frac{\delta^n}{U_n},\ldots,\sum_n\frac{\delta^n}{U_n^L}\right)=\operatorname{span}_{\mathbb{Q}}\left(1,\sqrt{d}\,W_1(\nu),\ldots,\sqrt{d}\,W_L(\nu)\right).

Proof. Since 1/Unl=(η−η−1)lη−nl(1−η−2n)−l1/U_n^l=(\eta-\eta^{-1})^l\eta^{-nl}(1-\eta^{-2n})^{-l} and ll is odd, δnη−nl=νnl\delta^n\eta^{-nl}=\nu^{nl} and η−2n=ν2n\eta^{-2n}=\nu^{2n}. Expanding (1−ν2n)−l(1-\nu^{2n})^{-l} and summing over nn (absolutely convergent), Vl=∑i≥0(i+l−1l−1)νl+2i/(1−νl+2i)V_l=\sum_{i\geq0}\binom{i+l-1}{l-1}\nu^{l+2i}/(1-\nu^{l+2i}). With m=l+2im=l+2i, ((m−l)/2+l−1l−1)=∏t=1l−1(m−l+2t)/(2l−1(l−1)!)=Pl(m)\binom{(m-l)/2+l-1}{l-1}=\prod_{t=1}^{l-1}(m-l+2t)/(2^{l-1}(l-1)!)=P_l(m), since the factors m−l+2tm-l+2t are the numbers m±im\pm i, ii odd, i≤l−2i\leq l-2. As PlP_l vanishes at m=1,3,…,l−2m=1,3,\ldots,l-2, the sum may run over all odd mm. Pl\mathcal{P}_l is even of degree l−1l-1 with leading coefficient 1/(2l−1(l−1)!)1/(2^{l-1}(l-1)!), which gives the triangular expansion. Finally (η−η−1)2=a2−4(\eta-\eta^{-1})^2=a^2-4, so (η−η−1)l=(a2−4)(l−1)/2fd(\eta-\eta^{-1})^l=(a^2-4)^{(l-1)/2}f\sqrt{d} with f∈Q×f\in\mathbb{Q}^{\times}. □

For example V3=(W3−W1)/8V_3=(W_3-W_1)/8 and V5=(9W1−10W3+W5)/384V_5=(9W_1-10W_3+W_5)/384.

Remark 2.2. By Lemma 2.1, Theorem 1.1 for (a,δ)(a,\delta) is Theorem 1.5 at ν=δ/η\nu=\delta/\eta. Conversely, for ν\nu as in Theorem 1.5, η=1/∣ν∣>1\eta=1/|\nu|>1 is a unit of norm +1+1, its trace a=η+η−1a=\eta+\eta^{-1} is an integer >2>2, and ν=δ/η\nu=\delta/\eta with δ\delta the sign of ν\nu. Since η∈K∖Q\eta\in K\setminus\mathbb{Q}, Q(a2−4)=K\mathbb{Q}(\sqrt{a^2-4})=K, so the squarefree dd of Lemma 2.1 is that of Theorem 1.5.

What is used from [Kri26c]

From now on A≥4A\geq4 is even, 1≤r<A/21\leq r<A/2, n≥1n\geq1 and e=(A−2r)n/2∈Ze=(A-2r)n/2\in\mathbb{Z}. Write (x;q)m=∏i=0m−1(1−xqi)(x;q)_m=\prod_{i=0}^{m-1}(1-xq^i), Φℓ\Phi_\ell for the ℓ\ell-th cyclotomic polynomial, dn(q)=∏ℓ≤nΦℓ(q)d_n(q)=\prod_{\ell\leq n}\Phi_\ell(q), ϕE\phi_E for Euler’s function and T(n)=∑ℓ≤nϕE(ℓ)\mathcal{T}(n)=\sum_{\ell\leq n}\phi_E(\ell). Let

R(T)=Rn(T;q)=(q;q)nA−2r(q−rnT;q)rn(qn+1T;q)rnTe(T;q)n+1AR(T)=R_n(T;q)=\frac{(q;q)_n^{A-2r}(q^{-rn}T;q)_{rn}(q^{n+1}T;q)_{rn}T^e}{(T;q)_{n+1}^{A}}

be the rational function [12 (2.1)], with partial fractions [12 (3.1)]

R(T)=∑j=0n∑s=1Acs,j(q)(1−qjT)s,cs,j∈Q(q).R(T)=\sum_{j=0}^{n}\sum_{s=1}^{A}\frac{c_{s,j}(q)}{(1-q^jT)^s},\qquad c_{s,j}\in\mathbb{Q}(q).

We use the following statements of [12] verbatim. Items (a)–(c) are identities for an indeterminate qq; (d) and (e) hold for every complex qq with 0<∣q∣<10<|q|<1, where κ=∣q∣\kappa=|q|; (f) holds for every unit qq of norm +1+1 of a real quadratic field with 0<∣q∣<10<|q|<1.

(a) [12 Lemma 2.1]: R(q−n/T)=qAn/2TAR(T)R(q^{-n}/T)=q^{An/2}T^AR(T).

(b) [12 Corollary 3.1]: cs,j(1/q)=q−ecs,n−j(q)c_{s,j}(1/q)=q^{-e}c_{s,n-j}(q).

(c) [12 Lemma 3.2]: dn(q)Acs,j(q)∈Z[q±1]d_n(q)^Ac_{s,j}(q)\in\mathbb{Z}[q^{\pm1}]; and 1−q±m1-q^{\pm m} divides dn(q)d_n(q) in Z[q±1]\mathbb{Z}[q^{\pm1}] for 1≤m≤n1\leq m\leq n [12 §3].

(d) The proof of [12 Lemma 6.1]: with c±=∏i≥1(1±κi)c_{\pm}=\prod_{i\geq1}(1\pm\kappa^i) and C=(c+/c−)2AC=(c_+/c_-)^{2A}, R(qk)=0R(q^k)=0 for 1≤k≤rn1\leq k\leq rn, and C−1κke≤∣R(qk)∣≤CκkeC^{-1}\kappa^{ke}\leq|R(q^k)|\leq C\kappa^{ke} for k≥rn+1k\geq rn+1.

(e) [12 Lemma 9.2]: with ρ=(1−κ)/2\rho=(1-\kappa)/2 and constants c±,γ±>0c_{\pm},\gamma_{\pm}>0 depending only on κ\kappa,

∣cs,j(q)∣≤ρs−A(j+1)c+A−2r+1γ+γ−−A(1+ρ)rn+eκ−E(j),|c_{s,j}(q)|\leq\rho^{s-A(j+1)}c_+^{A-2r+1}\gamma_+\gamma_-^{-A}(1+\rho)^{rn+e}\kappa^{-\mathcal{E}(j)},

where E(j)≤(A+4r2)n2/8+rn/2\mathcal{E}(j)\leq(A+4r^2)n^2/8+rn/2.

(f) [12 Lemma 7.1, Remark 7.2, and §9 before (9.2)]: σ(dn(q))=dn(1/q)=−q−T(n)dn(q)\sigma(d_n(q))=d_n(1/q)=-q^{-\mathcal{T}(n)}d_n(q), log⁡∣dn(q)∣=O(n)\log|d_n(q)|=O(n), and T(n)=3n2/π2+O(nlog⁡n)\mathcal{T}(n)=3n^2/\pi^2+O(n\log n) [9 Theorem 330].

We also use the trace argument of [12 Lemma 9.3(a)]: Tr⁡K/Q(x/d)∈Z\operatorname{Tr}_{K/\mathbb{Q}}(x/\sqrt{d}) \in\mathbb{Z} for x∈OKx \in\mathcal{O}_K, dd squarefree; and Nesterenko’s criterion as stated in [12 §9, proof of Theorem 1.1], from [16], [8 Theorem A], [11 Proposition 2]: if ϑ1,…,ϑN\vartheta_1,\ldots,\vartheta_N are real and there are integers pj,np_{j,n} with 1n2log⁡∣∑jpj,nϑj∣→α<0\frac{1}{n^2}\log\left|\sum_j p_{j,n}\vartheta_j\right|\to\alpha<0 and lim sup⁡n1n2log⁡∣pj,n∣≤β\limsup_{n}\frac{1}{n^2}\log|p_{j,n}|\le\beta for every jj, then dim⁡Q∑jQϑj≥1−α/β\dim_{\mathbb{Q}}\sum_j\mathbb{Q}\vartheta_j\ge1-\alpha/\beta.

We apply all of this at q=ν2q=\nu^2 with ν\nu a unit of norm +1+1, 0<∣ν∣<10<|\nu|<1; then qq is a unit of norm +1+1 with 0<q<10<q<1 and Q(q)=K\mathbb{Q}(q)=K. The sums SnS_n and coefficients PlP_l, P0P_0 of [12] are not used; ours are defined below. (In [12 §9] the letter ν\nu denotes log⁡(1/∣q∣)\log(1/|q|); here we write λ=log⁡(1/κ)\lambda=\log(1/\kappa).)

Let Rn=Q[ν,ν−1,dn(ν2)−1]R_n=\mathbb{Q}[\nu,\nu^{-1},d_n(\nu^2)^{-1}] for an indeterminate ν\nu. By (c), cs,j(ν2)∈Rnc_{s,j}(\nu^2)\in R_n. Since the identity dn(1/q)=−q−T(n)dn(q)d_n(1/q)=-q^{-T(n)}d_n(q) of (f) holds in Q(q)\mathbb{Q}(q) (see also [12 §4]), RnR_n is stable under ν↦1/ν\nu\mapsto1/\nu, and evaluation at ν∈C\nu\in\mathbb{C} with 0<∣ν∣≠10<|\nu|\ne1 is a ring homomorphism Rn→CR_n\to\mathbb{C}, since the roots of dn(ν2)d_n(\nu^2) are roots of unity.

The weighted bilateral forms

Let ν∈C\nu\in\mathbb{C}, 0<∣ν∣<10<|\nu|<1, q=ν2q=\nu^2 and κ=∣q∣\kappa=|q|, and let Sn=Sn(ν)S_n=S_n(\nu) be as in the introduction. No argument is a pole, and by (a), ν−n−kR(q−n−k)=ν−n−kqA(n/2+k)R(qk)=νkν(A−1)(n+2k)R(qk)\nu^{-n-k}R(q^{-n-k})=\nu^{-n-k}q^{A(n/2+k)}R(q^k)=\nu^k\nu^{(A-1)(n+2k)}R(q^k), so by (d) both series converge absolutely. For s≥1s\ge1 let ps(μ)=∏t=1s−1(μ+2t−1)/(2s−1(s−1)!)p_s(\mu)=\prod_{t=1}^{s-1}(\mu+2t-1)/(2^{s-1}(s-1)!), so that ps(2i+1)=(i+s−1s−1)p_s(2i+1)=\binom{i+s-1}{s-1}; psp_s has degree s−1s-1, nonnegative coefficients, and ps(1)=1p_s(1)=1. For odd ll with 1≤l≤A−11\le l\le A-1 put, with cs,j=cs,j(ν2)c_{s,j}=c_{s,j}(\nu^2) and sums over 1≤s≤A1\le s\le A, 0≤j≤n0\le j\le n,

Pl=2∑s,jcs,jν−j[μl−1]ps(μ),P0=∑s,jcs,jν−jBs,j,P_l=2\sum_{s,j}c_{s,j}\nu^{-j}[\mu^{l-1}]p_s(\mu),\qquad P_0=\sum_{s,j}c_{s,j}\nu^{-j}B_{s,j},
Bs,j=−∑t=1jνt(1−qt)−s+∑m=1n−jν−m(1−q−m)−s.B_{s,j}=-\sum_{t=1}^{j}\nu^t(1-q^t)^{-s}+\sum_{m=1}^{n-j}\nu^{-m}(1-q^{-m})^{-s}.

By (c), Pl,P0∈RnP_l,P_0\in R_n.

Lemma 4.1 (expansion). For 0<∣ν∣<10<|\nu|<1, Sn=P0+∑lPlWl(ν)S_n=P_0+\sum_lP_lW_l(\nu), the sum over odd ll with 1≤l≤A−11\le l\le A-1.

Proof. Positive side. By the partial fractions and with m=j+km=j+k,

∑k≥1νk(1−qj+k)s=ν−j[∑m≥1−∑m=1j]νm(1−ν2m)s,\sum_{k\ge1}\frac{\nu^k}{(1-q^{j+k})^s} =\nu^{-j}\left[\sum_{m\ge1}-\sum_{m=1}^{j}\right]\frac{\nu^m}{(1-\nu^{2m})^s},
∑m≥1νm(1−ν2m)s=∑μ oddps(μ)νμ1−νμ,\sum_{m\ge1}\frac{\nu^m}{(1-\nu^{2m})^s} =\sum_{\mu\ {\rm odd}}\frac{p_s(\mu)\nu^\mu}{1-\nu^\mu},

by expanding (1−ν2m)−s=∑i(i+s−1s−1)ν2mi(1-\nu^{2m})^{-s}=\sum_i\binom{i+s-1}{s-1}\nu^{2mi} and putting μ=2i+1\mu=2i+1.

Negative side. With m=n+k−jm=n+k-j, so that ν−n−k=ν−jν−m\nu^{-n-k}=\nu^{-j}\nu^{-m},

∑k≥1ν−n−k(1−qj−n−k)−s=ν−j[∑m≥1−∑m=1n−j]ν−m(1−q−m)−s.\sum_{k\geq1}\nu^{-n-k}(1-q^{j-n-k})^{-s}=\nu^{-j}\left[\sum_{m\geq1}-\sum_{m=1}^{n-j}\right]\nu^{-m}(1-q^{-m})^{-s}.

Here

ν−m(1−ν−2m)s=(−1)sνm(2s−1)(1−ν2m)s=(−1)s∑i≥0(i+s−1s−1)νm(2i+2s−1).\frac{\nu^{-m}}{(1-\nu^{-2m})^s}=\frac{(-1)^s\nu^{m(2s-1)}}{(1-\nu^{2m})^s}=(-1)^s\sum_{i\geq0}\binom{i+s-1}{s-1}\nu^{m(2i+2s-1)}.

For μ=2i+2s−1\mu=2i+2s-1, ps(−μ)=(−1)s−1(i+s−1s−1)p_s(-\mu)=(-1)^{s-1}\binom{i+s-1}{s-1}, and ps(−μ)=0p_s(-\mu)=0 for odd μ≤2s−3\mu\leq2s-3. Hence ∑m≥1ν−m(1−q−m)−s=−∑μ oddps(−μ)νμ/(1−νμ)\sum_{m\geq1}\nu^{-m}(1-q^{-m})^{-s}=-\sum_{\mu\ \mathrm{odd}}p_s(-\mu)\nu^\mu/(1-\nu^\mu).

Difference. Altogether

Sn=∑s,jcs,jν−j[∑μ odd(ps(μ)+ps(−μ))νμ1−νμ+Bs,j].S_n=\sum_{s,j}c_{s,j}\nu^{-j}\left[\sum_{\mu\ \mathrm{odd}}\bigl(p_s(\mu)+p_s(-\mu)\bigr)\frac{\nu^\mu}{1-\nu^\mu}+B_{s,j}\right].

and ps(μ)+ps(−μ)=2∑λ even[μλ]ps⋅μλp_s(\mu)+p_s(-\mu)=2\sum_{\lambda\ \mathrm{even}}[\mu^\lambda]p_s\cdot\mu^\lambda with λ≤s−1≤A−1\lambda\leq s-1\leq A-1, so λ≤A−2\lambda\leq A-2. Summing over μ\mu gives ∑lPlWl\sum_l P_lW_l with l=λ+1l=\lambda+1. All rearrangements are of absolutely convergent double series. □

No relation such as ∑s,jcs,j=0\sum_{s,j}c_{s,j}=0 is used, since the weights make every series converge. The term λ=0\lambda=0 gives W1W_1; a one-sided weighted sum would keep the even WW’s as well.

Lemma 4.2 (reciprocity). In RnR_n: Pl(1/ν)=νn−2ePl(ν)P_l(1/\nu)=\nu^{n-2e}P_l(\nu) for every odd ll, l=1l=1 included, and P0(1/ν)=−νn−2eP0(ν)P_0(1/\nu)=-\nu^{n-2e}P_0(\nu).

Proof. The substitution ν↦1/ν\nu\mapsto1/\nu is q↦1/qq\mapsto1/q, so cs,j(ν−2)=ν−2ecs,n−j(ν2)c_{s,j}(\nu^{-2})=\nu^{-2e}c_{s,n-j}(\nu^2) by (b). With j′=n−jj'=n-j,

Pl(1/ν)=2∑s,jν−2ecs,n−jνj[μl−1]ps=νn−2e⋅2∑s,j′cs,j′ν−j′[μl−1]ps.P_l(1/\nu)=2\sum_{s,j}\nu^{-2e}c_{s,n-j}\nu^{j[\mu^{l-1}]p_s=\nu^{n-2e}\cdot2\sum_{s,j'}c_{s,j'}\nu^{-j'}[\mu^{l-1}]p_s.}

For P0P_0, term by term,

Bs,j(1/ν)=−∑t=1jν−t(1−q−t)−s+∑m=1n−jνm(1−qm)−s=−Bs,n−j(ν),B_{s,j}(1/\nu)=-\sum_{t=1}^{j}\nu^{-t}(1-q^{-t})^{-s}+\sum_{m=1}^{n-j}\nu^m(1-q^m)^{-s}=-B_{s,n-j}(\nu),

so

P0(1/ν)=∑s,jν−2ecs,n−jνj(−Bs,n−j(ν))=−νn−2eP0(ν).P_0(1/\nu)=\sum_{s,j}\nu^{-2e}c_{s,n-j}\nu^j(-B_{s,n-j}(\nu))=-\nu^{n-2e}P_0(\nu).

If ν\nu is a unit of norm +1+1 in KK, then σν=1/ν\sigma\nu=1/\nu, so σ(f(ν))=f(1/ν)\sigma(f(\nu))=f(1/\nu) for f∈Rnf\in R_n, and

σ(Pl(ν))=νn−2ePl(ν),σ(P0(ν))=−νn−2eP0(ν).(2)\sigma(P_l(\nu))=\nu^{n-2e}P_l(\nu),\qquad\sigma(P_0(\nu))=-\nu^{n-2e}P_0(\nu). \tag*{(2)}

Heuristically, Wj(1/ν)=−Wj(ν)−(1−2j−1)ζ(1−j)W_j(1/\nu)=-W_j(\nu)-(1-2^{j-1})\zeta(1-j) formally, and the constant vanishes for every odd jj, including j=1j=1 (where $1-2^0=0): all WjW_j behave as anti-invariant, which is why dWj\sqrt{d}W_j is the right twist.

Lemma 4.3 (denominators). Put FA=2A−1(A−1)!F_A=2^{A-1}(A-1)!. Then, in Rn\mathcal{R}_n, FAdn(ν2)A−lPl∈Z[ν±1]F_A d_n(\nu^2)^{A-l}P_l\in\mathbb{Z}[\nu^{\pm1}] for odd 1≤l≤A−11\le l\le A-1, and dn(ν2)AP0∈Z[ν±1]d_n(\nu^2)^A P_0\in\mathbb{Z}[\nu^{\pm1}]. Hence Dn:=FAdn(ν2)AD_n:=F_A d_n(\nu^2)^A satisfies DnPl,DnP0∈Z[ν±1]D_nP_l,D_nP_0\in\mathbb{Z}[\nu^{\pm1}].

Proof. Write dn=dn(ν2)d_n=d_n(\nu^2). Since 2s−1(s−1)! ps∈Z[μ]2^{s-1}(s-1)!\,p_s\in\mathbb{Z}[\mu] and 2s−1(s−1)!2^{s-1}(s-1)! divides FAF_A for s≤As\le A, we have FA[μl−1]ps∈ZF_A[\mu^{l-1}]p_s\in\mathbb{Z}, and [μl−1]ps=0[\mu^{l-1}]p_s=0 unless λ≤s−1\lambda\le s-1. So only s≥ls\ge l contributes to PlP_l, and then FAdnA−ldn,cs,jν−j[μl−1]psF_A d_n^{A-l}d_{n,c_{s,j}}\nu^{-j}[\mu^{l-1}]p_s lies in Z[ν±1]\mathbb{Z}[\nu^{\pm1}] by (c). For P0P_0: 1−qt1-q^t divides dnd_n for 1≤t≤n1\le t\le n by (c), and (1−q−t)−s=(−qt)s(1−qt)−s(1-q^{-t})^{-s}=(-q^t)^s(1-q^t)^{-s}, so dnsBs,j∈Z[ν±1]d_n^sB_{s,j}\in\mathbb{Z}[\nu^{\pm1}], and dnAcs,jν−jBs,j=(dnA−scs,j)(dnsBs,j)ν−jd_n^A c_{s,j}\nu^{-j}B_{s,j}=(d_n^{A-s}c_{s,j})(d_n^sB_{s,j})\nu^{-j}.

Coefficients and decay

Lemma 5.1 (coefficients). Let ν∈C\nu\in\mathbb{C}, 0<∣ν∣<10<|\nu|<1, κ=∣ν∣2\kappa=|\nu|^2, ρ=(1−κ)/2\rho=(1-\kappa)/2 and Γn=∑j=0n∑s=1A∣ν∣−j∣cs,j(ν2)∣\Gamma_n=\sum_{j=0}^{n}\sum_{s=1}^{A}|\nu|^{-j}|c_{s,j}(\nu^2)|. For every odd ll with 1≤l≤A−11\le l\le A-1, ∣Pl(ν)∣≤2Γn|P_l(\nu)|\le2\Gamma_n and ∣P0(ν)∣≤n∣ν∣(1−κ)−AΓn|P_0(\nu)|\le n|\nu|(1-\kappa)^{-A}\Gamma_n, and

Γn≤(n+1)A∣ν∣−nρ−A(n+1)c+A−2r+1γ+γ−−A(1+ρ)rn+eκ−(A+4r2)n2/8−rn/2.\Gamma_n\le(n+1)A|\nu|^{-n}\rho^{-A(n+1)}c_+^{A-2r+1}\gamma_+ \gamma_-^{-A}(1+\rho)^{rn+e}\kappa^{-(A+4r^2)n^2/8-rn/2}.

Consequently, with an implied constant depending only on AA, rr and ∣ν∣|\nu|,

log⁡max⁡{∣P0(ν)∣,max⁡l∣Pl(ν)∣}≤A+4r28n2λ+O(n).(3)\log\max\left\{|P_0(\nu)|,\max_l|P_l(\nu)|\right\}\le\frac{A+4r^2}{8}n^2\lambda+O(n). \tag*{(3)}

Proof. Apply (e) at q=ν2q=\nu^2. Since ρ,κ<1\rho,\kappa<1 and s≥1s\ge1, ρ−A(j+1)≤ρ−A(n+1)\rho^{-A(j+1)}\le\rho^{-A(n+1)} and κ−E(j)≤κ−(A+4r2)n2/8−rn/2\kappa^{-E(j)}\le\kappa^{-(A+4r^2)n^2/8-rn/2}; with ∣ν∣−j≤∣ν∣−n|\nu|^{-j}\le|\nu|^{-n} and (n+1)A(n+1)A terms this gives the bound for Γn\Gamma_n. The coefficients of psp_s are nonnegative with sum ps(1)=1p_s(1)=1, so ∣[μl−1]ps∣≤1|[\mu^{l-1}]p_s|\le1 and ∣Pl∣≤2Γn|P_l|\le2\Gamma_n. For 1≤t≤n1\le t\le n, ∣νt(1−qt)−s∣≤∣ν∣(1−κ)−A|\nu^t(1-q^t)^{-s}|\le|\nu|(1-\kappa)^{-A} and ∣ν−t(1−q−t)−s∣=∣ν∣t(2s−1)∣1−qt∣−s≤∣ν∣(1−κ)−A|\nu^{-t}(1-q^{-t})^{-s}|=|\nu|^{t(2s-1)}|1-q^t|^{-s}\le|\nu|(1-\kappa)^{-A}; as Bs,jB_{s,j} has nn terms, this gives the bound for P0P_0. All factors except κ−(A+4r2)n2/8\kappa^{-(A+4r^2)n^2/8} are eO(n)e^{O(n)}, since ∣ν∣−n=κ−n/2|\nu|^{-n}=\kappa^{-n/2}.

Lemma 5.2 (decay). Fix ν∈C\nu\in\mathbb{C} with 0<∣ν∣<10<|\nu|<1. For all large nn, Sn(ν)≠0S_n(\nu)\ne0 and

log⁡∣Sn(ν)∣=−(rn+1)(e+12)λ+O(1)=−r(A−2r)2n2λ+O(n).\log|S_n(\nu)|=-(rn+1)\left(e+\frac{1}{2}\right)\lambda+O(1)=-\frac{r(A-2r)}{2}n^2\lambda+O(n).

Proof. By (d) and the computation at the start of § 4, Sn=∑k≥rn+1νk(1−ν(A−1)(n+2k))R(qk)S_n=\sum_{k\ge rn+1}\nu^k(1-\nu^{(A-1)(n+2k)})R(q^k). For large nn, ∣ν∣(A−1)(n+2rn+2)≤12|\nu|^{(A-1)(n+2rn+2)}\le\frac12, so each factor 1−ν(A−1)(n+2k)1-\nu^{(A-1)(n+2k)} has modulus in [12,32]\left[\frac12,\frac32\right], and ∣νkR(qk)∣∈[C−1,C]κk(e+1/2)|\nu^kR(q^k)|\in[C^{-1},C]\kappa^{k(e+1/2)} by (d). So the term k=rn+1k=rn+1 has modulus at least (2C)−1κ(rn+1)(e+1/2)(2C)^{-1}\kappa^{(rn+1)(e+1/2)}, and the others have total modulus at most 32Cκ(rn+2)(e+1/2)/(1−κe+1/2)\frac32C\kappa^{(rn+2)(e+1/2)}/(1-\kappa^{e+1/2}). The ratio of these bounds tends to 00 as e→∞e\to\infty, so ∣Sn∣∈[(4C)−1,4C]κ(rn+1)(e+1/2)|S_n|\in[(4C)^{-1},4C]\kappa^{(rn+1)(e+1/2)} for large nn; and (rn+1)(e+12)=12r(A−2r)n2+O(n)(rn+1)(e+\frac12)=\frac12r(A-2r)n^2+O(n). □

Proof of the theorems

Let dd, KK, ν\nu be as in Theorem 1.5, q=κ=ν2q=\kappa=\nu^2, λ=log⁡(1/κ)\lambda=\log(1/\kappa), and Dn=FAdn(q)A∈Z[ν±1]⊂OKD_n=F_A d_n(q)^A\in\mathbb{Z}[\nu^{\pm1}]\subset\mathcal{O}_K as in Lemma 4.3. By (f), ∣σDn∣=κ−AT(n)∣Dn∣|\sigma D_n|=\kappa^{-A T(n)}|D_n| and log⁡∣Dn∣=O(n)\log|D_n|=O(n), so

log⁡∣NDn∣=log⁡∣Dn σDn∣=3Aπ2n2λ+O(nlog⁡n).(4)\log|\mathop{\rm N}D_n|=\log|D_n\,\sigma D_n|=\frac{3A}{\pi^2}n^2\lambda+O(n\log n). \tag*{(4)}

Balancing (as in [12 (9.3)]). Put ϱn=∣ν∣n−2e∣σDn∣/∣Dn∣\varrho_n=|\nu|^{n-2e}|\sigma D_n|/|D_n|, and let m=mn∈Zm=m_n\in\mathbb{Z} be the unique integer with

2≤∣ν∣2mϱn<2∣ν∣−2.(5)2\leq|\nu|^{2m}\varrho_n<2|\nu|^{-2}. \tag*{(5)}

Put Dn′=ν−mDn∈OKD'_n=\nu^{-m}D_n\in\mathcal{O}_K and Mn=∣ν∣n−2e∣σDn′∣M_n=|\nu|^{n-2e}|\sigma D'_n|. Since σν=1/ν\sigma\nu=1/\nu, Mn/∣Dn′∣=∣ν∣2mϱn∈[2,2∣ν∣−2)M_n/|D'_n|=|\nu|^{2m}\varrho_n\in[2,2|\nu|^{-2}); so ∣Dn′∣≤Mn/2|D'_n|\leq M_n/2 and Mn2=(Mn/∣Dn′∣)∣ν∣n−2e∣NDn∣M_n^2=(M_n/|D'_n|)|\nu|^{n-2e}|\mathop{\rm N}D_n|, whence by (4)

log⁡Mn=3A2π2n2λ+O(nlog⁡n).(6)\log M_n=\frac{3A}{2\pi^2}n^2\lambda+O(n\log n). \tag*{(6)}

For odd ll with 1≤l≤A−11\leq l\leq A-1 let ℓ0,n=Tr⁡K/Q(Dn′P0(ν))\ell_{0,n}=\operatorname{Tr}_{K/\mathbb{Q}}(D'_nP_0(\nu)), ℓl,n=Tr⁡K/Q(Dn′Pl(ν)/d)\ell_{l,n}=\operatorname{Tr}_{K/\mathbb{Q}}(D'_nP_l(\nu)/\sqrt d) and Ln=ℓ0,n+∑lℓl,ndWl(ν)L_n=\ell_{0,n}+\sum_l\ell_{l,n}\sqrt d W_l(\nu).

Lemma 6.1. (a) ℓ0,n,ℓl,n∈Z\ell_{0,n},\ell_{l,n}\in\mathbb{Z}. (b) Ln=(Dn′−νn−2eσDn′)Sn(ν)L_n=(D'_n-\nu^{n-2e}\sigma D'_n)S_n(\nu). (c) For large nn, Ln≠0L_n\ne0 and

log⁡∣Ln∣=(3A2π2−r(A−2r)2)n2λ+O(nlog⁡n).\log|L_n|=\left(\frac{3A}{2\pi^2}-\frac{r(A-2r)}{2}\right)n^2\lambda+O(n\log n).

(d)

log⁡max⁡(∣ℓ0,n∣,max⁡l∣ℓl,n∣)≤(3A2π2+A+4r28)n2λ+O(nlog⁡n).\log\max\left(|\ell_{0,n}|,\max_l|\ell_{l,n}|\right)\leq\left(\frac{3A}{2\pi^2}+\frac{A+4r^2}{8}\right)n^2\lambda+O(n\log n).

Proof. (a) By Lemma 4.3, Dn′P0(ν),Dn′Pl(ν)∈OKD'_nP_0(\nu),D'_nP_l(\nu)\in\mathcal{O}_K; use Tr⁡(OK)⊂Z\operatorname{Tr}(\mathcal{O}_K)\subset\mathbb{Z} and the trace argument of Section 3. (b) Write Tr⁡(z)=z+σz\operatorname{Tr}(z)=z+\sigma z. The terms zz add up to Dn′SnD'_nS_n by Lemma 4.1. By (2) and σd=−d\sigma\sqrt d=-\sqrt d, the conjugate terms are σ(Dn′P0)=−νn−2eσ(Dn′)P0\sigma(D'_nP_0)=-\nu^{n-2e}\sigma(D'_n)P_0 and σ(Dn′Pl/d)dWl=−νn−2eσ(Dn′)PlWl\sigma(D'_nP_l/\sqrt d)\sqrt d W_l=-\nu^{n-2e}\sigma(D'_n)P_lW_l, which add up to −νn−2eσ(Dn′)Sn-\nu^{n-2e}\sigma(D'_n)S_n. (c) ∣Dn′−νn−2eσDn′∣∈[Mn−∣Dn′∣,Mn+∣Dn′∣]⊂[12Mn,32Mn]|D'_n-\nu^{n-2e}\sigma D'_n|\in[M_n-|D'_n|,M_n+|D'_n|]\subset[\frac12M_n,\frac32M_n]; use Lemma 5.2 and (6). (d) ∣ℓ0,n∣≤(∣Dn′∣+Mn)∣P0∣≤32Mn∣P0∣|\ell_{0,n}|\leq(|D'_n|+M_n)|P_0|\leq\frac32M_n|P_0|, and likewise ∣ℓl,n∣≤32Mn∣Pl∣/d|\ell_{l,n}|\leq\frac32M_n|P_l|/\sqrt d. Use (3) and (6). □

Proof of Theorem 1.5. Apply Nesterenko’s criterion to the A/2+1A/2+1 numbers 11 and dWj(ν)\sqrt d W_j(\nu), j=1,3,…,A−1j=1,3,\ldots,A-1, and to the forms LnL_n, which are defined for every n≥1n\geq1. By Lemma 6.1(c), (d), whose first estimate is two-sided because Lemma 5.2 and the asymptotic in (f) are,

α=(3A2π2−r(A−2r)2)λ,β=(3A2π2+A+4r28)λ>0.\alpha=\left(\frac{3A}{2\pi^2}-\frac{r(A-2r)}{2}\right)\lambda,\qquad \beta=\left(\frac{3A}{2\pi^2}+\frac{A+4r^2}{8}\right)\lambda>0.

If α≥0\alpha\geq0 then τ(A,r)≤0\tau(A,r)\leq0 and there is nothing to prove. Otherwise the dimension is at least

1−αβ=1+r(A−2r)/2−3A/(2π2)3A/(2π2)+(A+4r2)/8=1+τ(A,r).1-\frac{\alpha}{\beta} =1+\frac{r(A-2r)/2-3A/(2\pi^2)}{3A/(2\pi^2)+(A+4r^2)/8} =1+\tau(A,r).

Proof of Theorem 1.1. By Lemma 2.1 at ν=δ/η\nu= \delta/\eta, the span in Theorem 1.1 is that of Theorem 1.5, with dd the squarefree part of a2−4a^{2}-4. The asymptotic is [12 proof of Corollary 1.2]: with t=12+π2/(2π)t=\sqrt{12+\pi^{2}}/(2\pi) and r=⌈tA⌉r=\lceil t\sqrt{A}\rceil (admissible for A≥6A\geq6), τ(A,r)/A→4π2t/(12+π2+4π2t2)=π/12+π2\tau(A,r)/\sqrt{A}\to4\pi^{2}t/(12+\pi^{2}+4\pi^{2}t^{2})=\pi/\sqrt{12+\pi^{2}}. If at most uu of the sums with odd l≤A−1l\leq A-1 were irrational, the span would have dimension at most u+1u+1; this gives the count, and max⁡rτ(A,r)→∞\max_r\tau(A,r)\to\infty.

Proof of Corollaries 1.3 and 1.4. Let VV be the span of Theorem 1.1 at A=AkA=A_k, so dim⁡V≥k\dim V\geq k by Corollary 1.2. As VV is V3V_3 plus the span of the sums with 5≤l≤Ak−15\leq l\leq A_k-1, and dim⁡V3=3\dim V_3=3 by [13 Theorem 1.1] (its case (P,Q)=(a,1)(P,Q)=(a,1), for both δ\delta; in its Lean statement (a,E)=(a,−1)(a,E)=(a,-1)), dim⁡V/V3≥k−3\dim V/V_3\geq k-3; (1) and (2) are k=4k=4 and k=5k=5. For Corollary 1.4: F2nF_{2n} is the sequence with a=3a=3; for even kk, Fkn=FkUnF_{kn}=F_kU_n with η=ϕk\eta=\phi^k and a=Lka=L_k, so ∑nδn/Fknl=Fk−l∑nδn/Unl\sum_n\delta^n/F_{kn}^l=F_k^{-l}\sum_n\delta^n/U_n^l; and P2n=2UnP_{2n}=2U_n with a=6a=6.

Thresholds and computations

Proof of Corollary 1.2. Since 12A+π2(A+4r2)>012A+\pi^2(A+4r^2)>0, the inequality 1+τ(A,r)>k−11+\tau(A,r)>k-1 is equivalent to

π2[4r(A−2r)−(k−2)(A+4r2)]>12A(k−1).\pi^2\left[4r(A-2r)-(k-2)(A+4r^2)\right]>12A(k-1).

For 2≤k≤112\leq k\leq11, every even A≤AkA\leq A_k and every 1≤r<A/21\leq r<A/2, this inequality was decided in exact rational arithmetic from 3.14159265358979<π<3.141592653589803.14159265358979<\pi<3.14159265358980 (no case was left undecided), with the results stated; for example max⁡r(1+τ(22,r))=2.9499<3\max_r(1+\tau(22,r))=2.9499<3.

From k=7k=7 (A7=88A_7=88) on, two values of rr reach the threshold at AkA_k (at A=234A=234, r=11r=11 gives 10.020810.0208), and A12=282A_{12}=282. The ratio max⁡r(1+τ(A,r))/A\max_r(1+\tau(A,r))/\sqrt{A} is 0.64750.6475, 0.66350.6635, 0.66910.6691, 0.67090.6709 at A=102,…,105A=10^2,\ldots,10^5.

Computations

Computations. The ancillary files contain scripts (python-flint, mpmath) and logs in two independent sets, the author’s (anc/) and those of the AI referee pass described under Provenance, written without the author’s code (anc/referee/). Only the threshold certification (manyu_opt.py, Arb balls; referee/ref_opt.py, exact rationals) enters a proof. The other checks confirm Lemma 2.1 (a≤7a\leq7, l≤23l\leq23), Lemmas 4.2 and 4.3 exactly in Q(ν)\mathbb{Q}(\nu) (35 triples (A,r,n)(A,r,n), A≤24A\leq24), and, at five units such as ±φ−2\pm\varphi^{-2} and 2−32-\sqrt{3} with (A,r)(A,r) up to (24,3)(24,3) (120 rows, 8000 bits), the integrality of the ℓ\ell’s, Lemmas 4.1, 5.1, 5.2 and 6.1(b); the referee repeated the last group in 44 rows and computed the denominator profiles of Remark 8.1(c).

Remarks and limits

Remark 8.1. (a) Norm −1-1. For FnF_n, PnP_n, and FknF_{kn} with kk odd, the unit has norm −1-1, Galois conjugation acts as ν↦−1/ν\nu\mapsto-1/\nu, and Lemma 4.2 gives no relation between σ(Pl(ν))\sigma(P_l(\nu)) and Pl(ν)P_l(\nu). These sequences are not covered; for them the results we know concern l≤3l\leq3 [1, 15, 14, 19, 13].

(b) Single values. The theorems decide no individual sum such as ∑n1/Un5\sum_n 1/U_n^5, and do not say which ll give irrational values. We see no way to eliminate W1W_1 and W3W_3 from the forms LnL_n without multiplying their coefficients by factors ecn2e^{cn^2} with c>0c>0.

(c) Sharper thresholds. At q=1/pq=1/p the bounds of [11] were improved by saving one power of dnd_n [10] and by a refinement of Nesterenko’s criterion [8]; we have not adapted either. A full power of dn(ν2)d_n(\nu^2) cannot be saved in Lemma 4.3: in every case we computed exactly (in Q(ν)\mathbb{Q}(\nu), A≤24A\leq24) the denominator of P0P_0 contains Φ2n(ν)A\Phi_{2n}(\nu)^A, so dn(ν2)A−1P0∉Z[ν±1]d_n(\nu^2)^{A-1}P_0\notin\mathbb{Z}[\nu^{\pm1}] there. In these cases the exponent AA occurs only at Φ2ℓ(ν)\Phi_{2\ell}(\nu) with ℓ\ell in roughly (2n/3,n](2n/3,n] and at a few small indices, so a partial saving is not excluded.

(d) Odd AA. We have not proved a version of the construction for odd AA (then ee is a half-integer for odd nn). Such a version, with the same function τ\tau, would give nothing new. For fixed r≥1r\geq1,

τ(A,r)=(4π2r−12)A−8π2r2(12+π2)A+4π2r2\tau(A,r)=\frac{(4\pi^2r-12)A-8\pi^2r^2}{(12+\pi^2)A+4\pi^2r^2}

is a Möbius function of AA whose denominator is positive for A>0A>0 and whose determinant (4π2r−12)⋅4π2r2+8π2r2(12+π2)(4\pi^2r-12)\cdot4\pi^2r^2+8\pi^2r^2(12+\pi^2) is positive, so it is strictly increasing in AA. Moreover r<A/2r<A/2 implies r<(A+1)/2r<(A+1)/2. Hence τ(A,r)<τ(A+1,r)\tau(A,r)<\tau(A+1,r), and Theorem 1.5 at the even number A+1A+1 concerns the same numbers d W1,d W3,…,d WA\sqrt{d}\,W_1,\sqrt{d}\,W_3,\ldots,\sqrt{d}\,W_A.

Related work

André-Jeannin proved that ∑nxn/wn\sum_n x^n/w_n is irrational for a class of binary recurrences that includes every Un(P,±1)U_n(P,\pm1) with x=±1x=\pm1, in particular ∑nqn/Un∉Q\sum_n q^n/U_n\notin\mathbb{Q} [1]. Tachiya [18] proved irrationality results for weight-one Lambert series at algebraic integers of Pisot type; Bundschuh and Väätänen [2] proved linear independence results for such series, with applications to reciprocal sums of Fibonacci and Lucas numbers, and [3] treats mixed weights one and two at rational bases; Duverney, Suzuki and Tachiya [5] proved linear independence results for reciprocal sums of Fibonacci and Lucas numbers along certain sequences of indices. For both signs of the norm the even powers ∑nUn−2s\sum_n U_n^{-2s} are transcendental [4], and their algebraic relations are known [6, 7]; these are values of quasi-modular forms, to which Nesterenko’s theorem [17] applies. Wang [24] expresses ∑nnpUn−s\sum_n n^pU_n^{-s} through derivatives of Eisenstein series when p≡s(mod2)p\equiv s\pmod2 (norm +1+1) and when p,sp,s are even (norm −1-1); the sums ∑nUn−l\sum_n U_n^{-l}, ll odd, of this note are outside these parity classes, and we know of no such expression for them or for their alternating versions.

At q=1/pq=1/p with p∈Z∖{0,±1}p\in\mathbb{Z}\setminus\{0,\pm1\}, Krattenthaler, Rivoal and Zudilin proved a lower bound of order A\sqrt{A} for the dimension of the span of 1,ζq(3),ζq(5),…,ζq(A−1)1,\zeta_q(3),\zeta_q(5),\ldots,\zeta_q(A-1) [11], refined in [10, 8]. The analogue at real quadratic units for d ζq(l)\sqrt{d}\,\zeta_q(l) is [12], of which this note is a variant.

For weight three, the author proved ∑n1/Fn3∉Q\sum_n 1/F_n^3 \notin\mathbb{Q} and ∑n1/Un3∉Q\sum_n 1/U_n^3 \notin\mathbb{Q} for all real quadratic units [15, 14], and the linear independence of 1,∑nδn/Un,∑n(δε)n/Un31,\sum_n \delta^n/U_n,\sum_n(\delta\varepsilon)^n/U_n^3, with ε\varepsilon the norm of the unit [13], all with Lean formalisations. The Hexagon contributor twinheterodox posted neighbouring results of weight at most three: the alternating sums ∑n(−1)n−1/Un3\sum_n(-1)^{n-1}/U_n^3 [19], the weight-three Lambert sum ∑nn2/(bn−1)\sum_n n^2/(b^n-1) at Pisot, Salem and other algebraic bases [22, 20], a weighted reciprocal Lucas sum [23] and the reciprocal Mersenne cubes [21]; these preprints are described by their author as primarily AI-generated, have no formal proofs, and were not checked by us. None of these works treats an odd power l≥5l \ge5, and in our searches of Hexagon, arXiv, GitHub and Zenodo (November 2026) we found no earlier result on ∑nδn/Unl\sum_n \delta^n/U_n^l for an odd l≥5l \ge5.

Provenance

The construction was found by varying the summation in [12 §9], choosing weights so that the resulting values are those of Lemma 2.1. The proof was checked line by line in a separate whole-chain referee pass, which rebuilt the exact computations of § 7 with independent code and found no error affecting a statement. Nothing in this note is formalised. Its external inputs are those of [12 Theorem 1.1] and the lemmas listed in § 3; the Lean-verified result of [13] enters only Corollary 1.3. The referee pass was itself carried out by an AI model (a separate Claude session, without access to the author’s code); it is not a peer review.

Use of AI tools

This work was carried out with substantial assistance from AI systems (Anthropic’s Claude models), which were used to search the literature, to develop and check the proofs, to run the computations and to draft this text. The author directed the project and is responsible for its content.

References

  1. [1]R. André-Jeannin, Irrationalité de la somme des inverses de certaines suites récurrentes, C. R. Acad. Sci. Paris Sér. I Math. 308 (1989), no. 19, 539–541.
  2. [2]P. Bundschuh and K. Väanänen, Linear independence of certain Lambert and allied series, Acta Arith. 120 (2005), no. 2, 197–209.DOI
  3. [3]———, Linear independence of q-analogues of certain classical constants, Results Math. 47 (2005), no. 1-2, 33–44.DOI
  4. [4]D. Duverney, Ke. Nishioka, Ku. Nishioka, and I. Shiokawa, Transcendence of Rogers–Ramanujan continued fraction and reciprocal sums of Fibonacci numbers, Proc. Japan Acad. Ser. A Math. Sci. 73 (1997), no. 7, 140–142.DOI
  5. [5]D. Duverney, Y. Suzuki, and Y. Tachi, Linear independence results for sums of reciprocals of Fibonacci and Lucas numbers, Acta Math. Hungar. 162 (2020), no. 1, 375–392, arXiv:1908.07290.arxiv.org/abs/1908.07290
  6. [6]C. Elsner, S. Shimomura, and I. Shiokawa, Algebraic relations for reciprocal sums of Fibonacci numbers, Acta Arith. 130 (2007), 37–60.DOI
  7. [7]———, Algebraic relations for reciprocal sums of even terms in Fibonacci numbers, J. Math. Sci. (N.Y.) 180 (2012), no. 5, 650–671, translated from Fundam. Prikl. Mat. 16 (2010), no. 5, 173–200.
  8. [8]S. Fischler and W. Zudilin, A refinement of Nesterenko’s linear independence criterion with applications to zeta values, Math. Ann. 347 (2010), 739–763.
  9. [9]G. H. Hardy and E. M. Wright, An introduction to the theory of numbers, sixth ed., Oxford University Press, Oxford, 2008, Revised by D. R. Heath-Brown and J. H. Silverman.
  10. [10]F. Jouhet and E. Mosaki, Irrationalité aux entiers impairs positifs d’un q-analogue de la fonction zêta de Riemann, Int. J. Number Theory 6 (2010), 959–988.arxiv.org/abs/0712.1762
  11. [11]C. Krattenthaler, T. Rivoal, and W. Zudilin, Séries hypergéométriques basiques, q-analogues des valeurs de la fonction zêta et séries d’Eisenstein, J. Inst. Math. Jussieu 5 (2006), no. 1, 53–79, arXiv:math/0311033.DOI
  12. [12]S. Krisman, Irrationality of odd q-zeta values at real quadratic units, Hexagon 2611.00693v1, https://hexagonmath.org/abs/2611.00693v1, 2026.
  13. [13]———, Linear independence of reciprocal and reciprocal-cube sums of Lucas sequences, Hexagon preprint; Lean formalisation at https://github.com/telaviv/lucas-reciprocal-independence, 2026.hub.valency.io/works/W_dyu427fe
  14. [14]———, The reciprocal cube sums of Lucas sequences of quadratic units are irrational, Hexagon 2611.00491v1, https://hexagonmath.org/abs/2611.00491v1; Lean formalisation at https://github.com/telaviv/lucas-zeta3-irrationality, 2026.
  15. [15]———, The sum of the reciprocal cubes of the Fibonacci numbers is irrational, Hexagon 2611.00473v1, https://hexagonmath.org/abs/2611.00473v1; Lean formalisation at https://github.com/telaviv/fibonacci-zeta3-irrationality, 2026.
  16. [16]Yu. V. Nesterenko, On the linear independence of numbers, Moscow Univ. Math. Bull. 40 (1985), no. 1, 69–74.
  17. [17]———, Modular functions and transcendence questions, Sb. Math. 187 (1996), no. 9, 1319–1348.DOI
  18. [18]Y. Tachiya, Irrationality of certain Lambert series, Tokyo J. Math. 27 (2004), 75–85.DOI
  19. [19]twinheterodox, The alternating sum of reciprocal Fibonacci cubes is irrational, Hexagon 2611.00613v1, https://hexagonmath.org/abs/2611.00613v1, 2026, Described by its author as primarily AI-generated; no formal proof.
  20. [20]———, A height criterion for irrational weight-three Lambert sums, Hexagon 2611.00638v1, https://hexagonmath.org/abs/2611.00638v1, 2026, Described by its author as primarily AI-generated; no formal proof.
  21. [21]———, The sum of the reciprocal Mersenne cubes is irrational, Hexagon 2611.00643v1, https://hexagonmath.org/abs/2611.00643v1, 2026, Described by its author as primarily AI-generated; no formal proof.
  22. [22]———, The weight-three Lambert sum at the golden ratio is irrational, Hexagon 2611.00623v1, https://hexagonmath.org/abs/2611.00623v1, 2026, Described by its author as primarily AI-generated; no formal proof.
  23. [23]———, A weighted reciprocal Lucas sum lies outside ℚ(√5), Hexagon 2611.00639v1, https://hexagonmath.org/abs/2611.00639v1, 2026, Described by its author as primarily AI-generated; no formal proof.
  24. [24]W. Wang, Evaluation of reciprocal sums of hyperbolic functions using quasimodular forms, J. Math. Anal. Appl. 549 (2025), no. 2, Paper No. 129526, arXiv:2312.16929.DOI

Paper details

vlnc:W_rxfgc7sh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 12

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Anthropic Claude Opus 5.5

View statement
Looking for related papers…
Contents