exactory
Sign inGet started
openMathematics, Number TheoryPosted 15 Nov 2026 by Thijs Missiaen

Close the global threshold gap and unconditional critical-boundary finiteness for Pólya's sinc multiplier on the Davenport–Heilbronn function

1score · Sign in to vote

Problem statement

For the odd primitive Dirichlet character χ5(mod5)\chi_5 \pmod 5 (χ5(2)=i\chi_5(2) = i) and κ:=(10−25−2)/(5−1)≈0.28407904\kappa := (\sqrt{10 - 2\sqrt{5}} - 2)/(\sqrt{5} - 1) \approx 0.28407904, let DDH(s):=1−iκ2L(s,χ5)+1+iκ2L(s,χ5‾)=∑n=1∞aDH(n) n−s\mathcal{D}_{\mathrm{DH}}(s) := \frac{1-i\kappa}{2}L(s,\chi_5) + \frac{1+i\kappa}{2}L(s,\overline{\chi_5}) = \sum_{n=1}^{\infty} a_{\mathrm{DH}}(n)\,n^{-s} be the Davenport–Heilbronn Dirichlet series, and let ξDH(s):=(5π)(s+1)/2Γ ⁣(s+12)DDH(s),ΞDH(z):=ξDH ⁣(12+iz2)=2∫0∞ΦDH(u)cos⁡(zu) du,\xi_{\mathrm{DH}}(s) := \left(\frac{5}{\pi}\right)^{(s+1)/2}\Gamma\!\left(\frac{s+1}{2}\right)\mathcal{D}_{\mathrm{DH}}(s), \qquad \Xi_{\mathrm{DH}}(z) := \xi_{\mathrm{DH}}\!\left(\frac{1}{2} + \frac{iz}{2}\right) = 2\int_{0}^{\infty} \Phi_{\mathrm{DH}}(u)\cos(zu)\,du, where ΦDH(u):=4e3u∑n=1∞n aDH(n) e−πn2e4u/5\Phi_{\mathrm{DH}}(u) := 4e^{3u}\sum_{n=1}^{\infty} n\,a_{\mathrm{DH}}(n)\,e^{-\pi n^2 e^{4u}/5}. Let σ∗:=sup⁡{Re⁡ρ:DDH(ρ)=0}=1.120362498183325…\sigma^* := \sup\{\operatorname{Re}\rho : \mathcal{D}_{\mathrm{DH}}(\rho) = 0\} = 1.120362498183325\dots, h∞∗:=σ∗−12=0.620362498183325…h_\infty^* := \sigma^* - \frac{1}{2} = 0.620362498183325\dots, and let χ5,quad\chi_{5,\mathrm{quad}} denote the real primitive quadratic Dirichlet character modulo 55. Consider the class G0\mathcal{G}_0 of real entire Pólya–Schur differential multipliers G(w)=eλw∏j=1∞(1+wwj)(λ≥0,  0<w1≤w2≤⋯ ,  ∑j=1∞wj−1<∞),G(w) = e^{\lambda w}\prod_{j=1}^{\infty} \left(1 + \frac{w}{w_j}\right) \qquad \left(\lambda \ge 0,\; 0 < w_1 \le w_2 \le \cdots,\; \sum_{j=1}^{\infty} w_j^{-1} < \infty\right), acting via G(−∂z2)ΞDH(z)=2∫0∞G(u2)ΦDH(u)cos⁡(zu) duG(-\partial_z^2)\Xi_{\mathrm{DH}}(z) = 2\int_{0}^{\infty} G(u^2)\Phi_{\mathrm{DH}}(u)\cos(zu)\,du, with Stieltjes logarithmic derivative SG(u):=ddulog⁡G(u2)=4λu+∑j=1∞2uwj+u2S_G(u) := \frac{d}{du}\log G(u^2) = 4\lambda u + \sum_{j=1}^{\infty} \frac{2u}{w_j + u^2} (u>0u > 0). Let Noff(F;∞)N_{\mathrm{off}}(F; \infty) count the non-real zeros of F(z)F(z) in C∖R\mathbb{C} \setminus \mathbb{R} with multiplicity. Compare the two canonical exponential-type-2h2h multipliers Gcosh,h(w):=cosh⁡(2hw)G_{\mathrm{cosh},h}(w) := \cosh(2h\sqrt{w}) and Pólya's hyperbolic sinc multiplier Gsinc,h(w):=sinh⁡(2hw)2hw=∏j=1∞(1+4h2wπ2j2)G_{\mathrm{sinc},h}(w) := \frac{\sinh(2h\sqrt{w})}{2h\sqrt{w}} = \prod_{j=1}^{\infty} (1 + \frac{4h^2 w}{\pi^2 j^2}), both having identical asymptotic shift lim⁡u→∞SG(u)=2h\lim_{u\to\infty} S_G(u) = 2h and identical infinite-height threshold h∞∗=σ∗−12h_\infty^* = \sigma^* - \frac{1}{2}. Because {Gsinc,h}h>0\{G_{\mathrm{sinc},h}\}_{h>0} is **not** a one-parameter semigroup in hh, define both the first global real-rootedness threshold and the ultimate real-rootedness threshold: hglobal∗,inf⁡(G∙;ΞDH):=inf⁡{h>0:G∙,h(−∂z2)ΞDH has only real zeros},h_{\mathrm{global}}^{*,\inf}(G_\bullet; \Xi_{\mathrm{DH}}) := \inf\{h > 0 : G_{\bullet,h}(-\partial_z^2)\Xi_{\mathrm{DH}} \text{ has only real zeros}\},hglobal∗,sup⁡(G∙;ΞDH):=inf⁡{h0>0:∀h≥h0,  G∙,h(−∂z2)ΞDH has only real zeros}.h_{\mathrm{global}}^{*,\sup}(G_\bullet; \Xi_{\mathrm{DH}}) := \inf\{h_0 > 0 : \forall h \ge h_0,\; G_{\bullet,h}(-\partial_z^2)\Xi_{\mathrm{DH}} \text{ has only real zeros}\}. Theorems 1.1(ii) and 3.2 of the anchor paper prove unconditionally that hglobal∗,inf⁡(Gcosh;ΞDH)=hglobal∗,sup⁡(Gcosh;ΞDH)=σ∗−12<58<hglobal∗,inf⁡(Gsinc;ΞDH)≤hglobal∗,sup⁡(Gsinc;ΞDH)≤2σ∗−1,h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{cosh}}; \Xi_{\mathrm{DH}}) = h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{cosh}}; \Xi_{\mathrm{DH}}) = \sigma^* - \frac{1}{2} < \frac{5}{8} < h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) \le h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) \le 2\sigma^* - 1, with a certified non-real zero of Gsinc,h(−∂z2)ΞDHG_{\mathrm{sinc},h}(-\partial_z^2)\Xi_{\mathrm{DH}} across all h∈[0,58]h \in [0, \frac{5}{8}] (and an isolated point witness at h=625121106=0.625121h = \frac{625121}{10^6} = 0.625121, where (0.625,0.625121)(0.625, 0.625121) remains uncovered). Moreover, at the critical asymptotic boundary h∗:=σ∗−12h_* := \sigma^* - \frac{1}{2} (lim⁡u→∞SGsinc,h∗(u)=2σ∗−1\lim_{u\to\infty} S_{G_{\mathrm{sinc},h_*}}(u) = 2\sigma^* - 1), Proposition 3.3 proves unconditionally that Noff(Gsinc,h∗(−∂z2)ΞDH;∞)≥8N_{\mathrm{off}}(G_{\mathrm{sinc},h_*}(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) \ge 8, and **conditionally on RH for ζ(s)\zeta(s) and GRH for L(s,χ5,quad)L(s, \chi_{5,\mathrm{quad}})** that Noff(Gsinc,h∗(−∂z2)ΞDH;∞)<∞N_{\mathrm{off}}(G_{\mathrm{sinc},h_*}(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) < \infty. Three sharp open problems arise from Sections 2, 3, and 6 of the anchor paper: 1. **Global GsincG_{\mathrm{sinc}} threshold gap:** Determine hglobal∗,inf⁡(Gsinc;ΞDH)h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) and hglobal∗,sup⁡(Gsinc;ΞDH)h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) inside the factor-two interval (58, 2σ∗−1](\frac{5}{8},\, 2\sigma^* - 1] (2σ∗−1=1.240724996…2\sigma^* - 1 = 1.240724996\dots), and determine whether hglobal∗,inf⁡=hglobal∗,sup⁡h_{\mathrm{global}}^{*,\inf} = h_{\mathrm{global}}^{*,\sup} holds for Gsinc,hG_{\mathrm{sinc},h} on ΞDH\Xi_{\mathrm{DH}} (i.e., whether real-rootedness is upward-closed in hh despite the lack of a semigroup law). 2. **Unconditional critical-boundary finiteness:** Prove or refute unconditionally (removing RH for ζ(s)\zeta(s) and GRH for L(s,χ5,quad)L(s, \chi_{5,\mathrm{quad}})) that Noff(Gsinc,σ∗−1/2(−∂z2)ΞDH;∞)<∞N_{\mathrm{off}}(G_{\mathrm{sinc},\sigma^*-1/2}(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) < \infty. 3. **Crossing-oscillation boundary regime (Remark 2.8(a)):** Proposition 2.7 already exhibits tower-lacunary products GlacG_{\mathrm{lac}} with lim inf⁡u→∞SGlac(u)=0<2σ∗−1<lim sup⁡u→∞SGlac(u)=+∞\liminf_{u\to\infty} S_{G_{\mathrm{lac}}}(u) = 0 < 2\sigma^* - 1 < \limsup_{u\to\infty} S_{G_{\mathrm{lac}}}(u) = +\infty and Noff=∞N_{\mathrm{off}} = \infty. For λ=0\lambda = 0 multipliers in G0\mathcal{G}_0 oscillating across 2σ∗−12\sigma^* - 1 (lim inf⁡SG<2σ∗−1<lim sup⁡SG\liminf S_G < 2\sigma^* - 1 < \limsup S_G), determine whether Noff<∞N_{\mathrm{off}} < \infty can occur despite lim inf⁡SG<2σ∗−1\liminf S_G < 2\sigma^* - 1, and whether Noff=∞N_{\mathrm{off}} = \infty can occur in the bounded-oscillation regime 0<lim inf⁡SG<2σ∗−1<lim sup⁡SG<∞0 < \liminf S_G < 2\sigma^* - 1 < \limsup S_G < \infty.

Current state

- **Subcritical vs. supercritical regimes (Theorem 1.1(i) & Proposition 2.7 of anchor paper):** - If λ=0\lambda = 0 and lim⁡u→∞SG(u)∈[0,2σ∗−1)\lim_{u\to\infty} S_G(u) \in [0, 2\sigma^* - 1) (including all polynomials and all ρ(G)<1/2\rho(G) < 1/2), then Noff(G(−∂z2)ΞDH;T)≫GTN_{\mathrm{off}}(G(-\partial_z^2)\Xi_{\mathrm{DH}}; T) \gg_G T via conjugate-diagram Borel–Laplace asymptotics generalizing Dobner (2021) and single-zero syndetic Diophantine translation (avoiding Righetti's 2017 non-density phenomenon on (1,σ∗)(1, \sigma^*)). - If λ>0\lambda > 0 or lim inf⁡u→∞SG(u)>2σ∗−1\liminf_{u\to\infty} S_G(u) > 2\sigma^* - 1 (including all multipliers of lower order ρ‾(G)>1/2\underline{\rho}(G) > 1/2, with no root-gap regularity), then Noff(G(−∂z2)ΞDH;∞)<∞N_{\mathrm{off}}(G(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) < \infty via a dynamic uu-plane horizontal-contour shift Im⁡u=π8−SG(v0)+38γ\operatorname{Im} u = \frac{\pi}{8} - \frac{S_G(v_0)+3}{8\gamma}. - Conversely, for every upper order ρ∈(1/2,1)\rho \in (1/2, 1), the tower-lacunary canonical product Glac(w)=∏k=1∞(1+w/Rk2)⌊Rk2ρ⌋G_{\mathrm{lac}}(w) = \prod_{k=1}^{\infty} (1 + w/R_k^2)^{\lfloor R_k^{2\rho}\rfloor} (R1=16R_1 = 16, Rk=22Rk−1R_k = 2^{2^{R_{k-1}}}) has ρ‾(Glac)=0\underline{\rho}(G_{\mathrm{lac}}) = 0, lim inf⁡u→∞SGlac(u)=0\liminf_{u\to\infty} S_{G_{\mathrm{lac}}}(u) = 0, lim sup⁡u→∞SGlac(u)=+∞\limsup_{u\to\infty} S_{G_{\mathrm{lac}}}(u) = +\infty, and leaves infinitely many non-real zeros in ∣Im⁡z∣>1|\operatorname{Im} z| > 1. - **Sinc vs. cosh separation:** Gcosh,hG_{\mathrm{cosh},h} satisfies hglobal∗,inf⁡(Gcosh;ΞDH)=hglobal∗,sup⁡(Gcosh;ΞDH)=σ∗−12h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{cosh}}; \Xi_{\mathrm{DH}}) = h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{cosh}}; \Xi_{\mathrm{DH}}) = \sigma^* - \frac{1}{2} by Hermite–Biehler and Rouché, whereas for Gsinc,hG_{\mathrm{sinc},h} a singularity-free 384-bit `FLINT/Arb` 360-ball Interval–Newton continuation from ρ44838\rho_{44838} across the full compact interval [0,58][0, \frac{5}{8}] certifies hglobal∗,inf⁡(Gsinc;ΞDH)>58h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) > \frac{5}{8} (with an isolated point witness at h=0.625121h = 0.625121), and the computer-free interval-splitting identity Gsinc,2h1(−∂z2)ΞDH=Gcosh,h1(−∂z2)Gsinc,h1(−∂z2)ΞDHG_{\mathrm{sinc},2h_1}(-\partial_z^2)\Xi_{\mathrm{DH}} = G_{\mathrm{cosh},h_1}(-\partial_z^2)G_{\mathrm{sinc},h_1}(-\partial_z^2)\Xi_{\mathrm{DH}} gives hglobal∗,sup⁡(Gsinc;ΞDH)≤2σ∗−1h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) \le 2\sigma^* - 1. - **Why removing RH/GRH at h∗=σ∗−12h_* = \sigma^* - \frac{1}{2} is blocked by known barriers (Remark 3.4 of anchor paper):** Under RH + GRH, a smoothed Mellin prime-angle sum yields ∣DDH(σ+iγ)∣≫(log⁡γ)−1/3|\mathcal{D}_{\mathrm{DH}}(\sigma + i\gamma)| \gg (\log\gamma)^{-1/3} on σ≥σ∗−(log⁡γ)−1/3\sigma \ge \sigma^* - (\log\gamma)^{-1/3}, beating the Borel–Laplace cap error O((log⁡γ)−1)\mathcal{O}((\log\gamma)^{-1}). Unconditionally, the Vinogradov–Korobov zero-free region (Korobov 1958) requires log⁡X≍(log⁡γ)2/3(log⁡log⁡γ)4/3\log X \asymp (\log\gamma)^{2/3}(\log\log\gamma)^{4/3} and yields a lower bound ≫exp⁡(−c′(log⁡γ)2/3(log⁡log⁡γ)4/3)\gg \exp(-c'(\log\gamma)^{2/3}(\log\log\gamma)^{4/3}), while direct Baker–Wüstholz (1993) linear forms in π(P)≍(log⁡γ)8.308⋯−o(1)\pi(P) \asymp (\log\gamma)^{8.308\dots - o(1)} logarithms give a doubly exponential lower bound—both lower bounds decaying faster than the (log⁡γ)−1(\log\gamma)^{-1} Borel cap remainder.

Resolution criteria

A public research paper or technical report resolves this challenge by completely solving **any one** of the three standalone parts below (or all three together), together with Criterion 4: 1. **Part A — Closing the GsincG_{\mathrm{sinc}} global threshold gap:** Either - (i) prove an unconditional strict upper bound hglobal∗,sup⁡(Gsinc;ΞDH)≤huph_{\mathrm{global}}^{*,\sup}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) \le h_{\mathrm{up}} with hup≤1.0h_{\mathrm{up}} \le 1.0 (or localize both hglobal∗,inf⁡h_{\mathrm{global}}^{*,\inf} and hglobal∗,sup⁡h_{\mathrm{global}}^{*,\sup} to certified intervals of width ≤10−2\le 10^{-2}), or - (ii) prove that hglobal∗,inf⁡(Gsinc;ΞDH)=2σ∗−1h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) = 2\sigma^* - 1 (which requires exhibiting non-real zeros for **every** h∈(58,2σ∗−1)h \in (\frac{5}{8}, 2\sigma^* - 1), not merely along a sequence hn→2σ∗−1h_n \to 2\sigma^* - 1), or - (iii) settle whether hglobal∗,inf⁡(Gsinc;ΞDH)=hglobal∗,sup⁡(Gsinc;ΞDH)h_{\mathrm{global}}^{*,\inf}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) = h_{\mathrm{global}}^{*,\sup}(G_{\mathrm{sinc}}; \Xi_{\mathrm{DH}}) and prove an unconditional upper bound strictly below 2σ∗−12\sigma^* - 1 (e.g., ≤1.20\le 1.20). Merely extending numerical Interval–Newton continuation slightly past 58\frac{5}{8} (e.g., covering [58,0.63][\frac{5}{8}, 0.63]) without closing the macroscopic gap toward 2σ∗−12\sigma^* - 1 is partial progress only. 2. **Part B — Unconditional critical-boundary resolution (h∗=σ∗−12h_* = \sigma^* - \frac{1}{2}):** Unconditionally prove or refute Noff(Gsinc,σ∗−1/2(−∂z2)ΞDH;∞)<∞N_{\mathrm{off}}(G_{\mathrm{sinc},\sigma^*-1/2}(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) < \infty without assuming RH, GRH, or unproved zero-density or Diophantine conjectures—either by establishing an unconditional pointwise lower bound on ∣DDH(σ∗+iγ)∣|\mathcal{D}_{\mathrm{DH}}(\sigma^* + i\gamma)| (or an unconditional contour/Rouché argument bridging every close encounter to σ∗\sigma^*) that dominates the Borel cap remainder, or by constructing an infinite sequence of non-real zeros at h=σ∗−12h = \sigma^* - \frac{1}{2}. (An "almost-all" or metric lower bound on ∣DDH(σ∗+iγ)∣|\mathcal{D}_{\mathrm{DH}}(\sigma^* + i\gamma)| alone does not bridge exceptional γ\gamma-intervals wider than 4π/log⁡γ4\pi/\log\gamma and is insufficient without an explicit bridging proof.) 3. **Part C — Crossing-oscillation dichotomy beyond GlacG_{\mathrm{lac}}:** For λ=0\lambda = 0 multipliers in G0\mathcal{G}_0 satisfying lim inf⁡u→∞SG(u)<2σ∗−1<lim sup⁡u→∞SG(u)\liminf_{u\to\infty} S_G(u) < 2\sigma^* - 1 < \limsup_{u\to\infty} S_G(u), establish **both**: (i) whether Noff(G(−∂z2)ΞDH;∞)<∞N_{\mathrm{off}}(G(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) < \infty can occur despite lim inf⁡u→∞SG(u)<2σ∗−1\liminf_{u\to\infty} S_G(u) < 2\sigma^* - 1, and (ii) whether Noff(G(−∂z2)ΞDH;∞)=∞N_{\mathrm{off}}(G(-\partial_z^2)\Xi_{\mathrm{DH}}; \infty) = \infty occurs in the bounded crossing regime 0<lim inf⁡u→∞SG(u)<2σ∗−1<lim sup⁡u→∞SG(u)<∞0 < \liminf_{u\to\infty} S_G(u) < 2\sigma^* - 1 < \limsup_{u\to\infty} S_G(u) < \infty, together with sharp structural criteria on SG(u)S_G(u) or the root-counting function nG(t)n_G(t) separating the two behaviors. Re-exhibiting an example with lim inf⁡SG=0\liminf S_G = 0 and lim sup⁡SG=+∞\limsup S_G = +\infty (already given by Proposition 2.7) does not satisfy this criterion. 4. **Verification and reproducibility:** Any computational witness, contour bound, or Interval–Newton continuation must use rigorous ball/interval arithmetic (such as `FLINT/Arb`) with full analytical truncation, derivative, and rounding bounds, and release reproducible source code and certificates.

Citations

  1. T. Missiaen, Asymptotic Shift Thresholds for Pólya–Schur Multipliers and a Certified Lower Bound on the de Bruijn–Newman Constant of the Davenport–Heilbronn Function, Zenodo v15.3.0 (2026).
  2. G. Pólya, Bemerkung über die Integraldarstellung der Riemannschen xi-Funktion, Acta Math. 48 (1926), 305–317.
  3. G. Pólya, Über die algebraisch-funktionentheoretischen Untersuchungen von J. L. W. V. Jensen, Kgl. Danske Vidensk. Selsk. Math.-Fys. Medd. 7 (1927), no. 17, 3–33 (reprinted in Collected Papers, Vol. 2: Location of Zeros, MIT Press, 1974).
  4. E. Benz, Über lineare, verschiebungstreue Funktionaloperationen und die Nullstellen ganzer Funktionen, Comment. Math. Helv. 7 (1934), 243–289.
  5. N. G. de Bruijn, The roots of trigonometric integrals, Duke Math. J. 17 (1950), no. 3, 197–226.
  6. E. Bombieri and A. Ghosh, Around the Davenport–Heilbronn function, Russian Math. Surveys 66 (2011), no. 2, 221–270.
  7. M. Righetti, On the density of zeros of linear combinations of Euler products for sigma > 1, Algebra Number Theory 11 (2017), no. 9, 2131–2163.
  8. A. Dobner, A proof of Newman's conjecture for the extended Selberg class, Acta Arith. 201 (2021), 29–62.
  9. N. M. Korobov, Estimates of trigonometric sums and their applications, Uspekhi Mat. Nauk 13 (1958), no. 4(82), 185–192.
  10. A. Baker and G. Wüstholz, Logarithmic forms and group varieties, J. Reine Angew. Math. 442 (1993), 19–62.
  11. F. Johansson, Arb: efficient arbitrary-precision midpoint-radius interval arithmetic, IEEE Trans. Comput. 66 (2017), no. 8, 1281–1292.

Linked papers