For the odd primitive Dirichlet character χ5(mod5)(χ5(1)=1,χ5(2)=i,χ5(3)=−i,χ5(4)=−1,χ5(5)=0) and κ:=(10−25−2)/(5−1)≈0.28407904, let
DDH(s):=21−iκL(s,χ5)+21+iκL(s,χ5)=n=1∑∞aDH(n)n−s
be the self-dual Davenport–Heilbronn Dirichlet series, and define its completed entire function and theta-kernel representation by
ξDH(s):=(π5)(s+1)/2Γ(2s+1)DDH(s),ΞDH(z):=ξDH(21+2iz)=2∫0∞ΦDH(u)cos(zu)du,
where ΦDH(u):=4e3u∑n=1∞naDH(n)e−πn2e4u/5. Under backward heat evolution ∂tHtDH=−∂z2HtDH,HtDH(z):=e−t∂z2ΞDH(z)=2∫0∞etu2ΦDH(u)cos(zu)du(t∈R),
define the de Bruijn–Newman constant of ΞDH by
ΛDH:=inf{t∈R:HtDH(z) has only real zeros}.
(Under the alternative rescaling w=z/2,ΞDH(w)=ξDH(21+iw), the corresponding constant is ΛDH=ΛDH/4; all bounds below use the canonical Pólya–de Bruijn–Newman–Rodgers–Tao–Polymath normalization ∂tHt=−∂z2Ht.) For each off-line zero ρ of DDH(s), let tcol(ρ):=inf{t>0:the heat-flow zero trajectory originating at z(ρ)=−2i(ρ−21) meets R}.
Theorem 1.1(iii) of the anchor paper certifies unconditionally that
12500000000028825920237=0.230607361896<ΛDH<2(σ∗−21)2=0.769699258304…,
where σ∗:=sup{Reρ:DDH(ρ)=0}=1.120362498183325… is the Bombieri–Ghosh supremum abscissa. The lower bound comes from a 512-bit ball-arithmetic Interval–Newton enclosure near the off-line zero ρ240≈0.86953058+240.40467235i, whose scalar coalescence occurs at tcol(ρ240)≈0.2306073618964546. An uncertified winding scan over γ∈[0,10000](572 off-line zeros) and a Kronecker phase-alignment search up to γ=108 found no larger coalescence time.
The unsolved problem is to close or substantially narrow the more than threefold interval (0.230607361896,2(σ∗−21)2) unconditionally and determine whether ΛDH is governed by low-lying coalescences near tcol(ρ240) or by deeper high-lying off-line zeros:
1. **Upper-bound / exactness route:** Prove unconditionally an effective upper bound ΛDH≤U∗ with U∗<41=0.25 (breaking Dobner's ℓ1 barrier at t=0.25, where γℓ1(0.25)≈4.18×108), or certify ΛDH into an interval of width ≤10−4 containing tcol(ρ240); or
2. **Refutation route:** Exhibit an unconditionally certified non-real zero of Ht1DH(z) at some rational t1>tcol(ρ240) (refuting the conjecture that ρ240 maximizes tcol(ρ)), together with a matching upper bound shrinking the remaining gap.
Current state
- **Off-line zeros and supremum abscissa:** Davenport and Heilbronn (1936) showed DDH vanishes in Res>1 and off the critical line in 0<Res<1; Spira (1994) and Balanzario–Sánchez-Ortiz (2007) computed low off-line zeros; Bombieri and Ghosh (2011) proved σ∗ is given by ∑p≡2,3(mod5)arctan(p−σ∗)=2π−arctanκ.
- **Heat-flow upper bounds with vs. without an Euler product:** For Riemann's Ξ, Ki, Kim, and Lee (2009) proved Λζ<21 via large-height simplicity and de Bruijn's (1950) hyperbola inequality, and Polymath 15 (2019) / Platt–Trudgian (2021) brought this to Λζ≤51=0.2. Recently, Freire et al. (`arXiv:2610.27343`) obtained ΛL≤3200953=0.2978125 for primitive real-valued Dirichlet L-functions of conductor 3≤m≤400. Those arguments rely on Euler-product / GRH zero-localization in 0<Res<1.
- **Exact status for ΞDH (anchor paper, `10.5281/zenodo.24242050`):**
- *Lower bound:* t∗=12500000000028825920237 is certified via 512-bit `FLINT/Arb` Interval–Newton enclosure (r=10−32) around z240≈480.69023104−9.5353×10−7i, proving ΛDH>0.230607361896>0.2≥Λζ. Non-interval adaptive quadrature (`mpmath.quad`) suffers oscillation aliasing near ρ1(γ1≈85.70, reporting spurious 0.1419 instead of certified tcol(ρ1)≈0.17650538).
- *Upper bound:* Adapting Ki–Kim–Lee gives the explicit large-height simplicity cutoff γ0(t):=max(400+64t3,52πexp(t4(σ∗−21+5+6t+2t))) and strict inequality ΛDH<2(σ∗−21)2≈0.769699.
- *The ℓ1 barrier (Remark 4.5 of anchor paper):* In Dobner's (2021) Riemann–Siegel-truncated modified Dirichlet polynomial ∑n=1N(γ)aDH(n)e4tlog2nn−21−4tlog2π5γ−iγ(N(γ)=⌊5γ/(2π)⌋), the non-multiplicative Gaussian factor cancels half the effective abscissa damping at n∼N(γ), pushing the ℓ1-dominance threshold from γℓ1(0.50)≈1.52×104(N=110) to γℓ1(0.25)≈4.18×108(N=18235).
- *Targeted literature check:* No published proof narrows ΛDH into [0.230607,0.25] or resolves whether ΛDH=tcol(ρ240).
Resolution criteria
A public research paper or technical report resolves this challenge by satisfying one of the two routes below together with Criteria 3 and 4, under the canonical normalization ∂tHtDH=−∂z2HtDH:
1. **Route A (Unconditional upper-bound breakthrough / exact bracketing):** Prove unconditionally that ΛDH≤U∗ with U∗<41=0.25 (or localize ΛDH to a certified interval [L∗,U∗] of width U∗−L∗≤10−4 containing tcol(ρ240)). Every height cutoff γcut(t) above which zeros of HtDH are proved real and simple must carry completely explicit constants and rigorous tail, Stirling, and remainder bounds, without unproved heuristics (such as the unproved saddle heuristic tcol(ρ)≲4(σ∗−21)/log(2π5γ)).
2. **Route B (Higher-coalescence refutation):** If ρ240 does not maximize tcol(ρ), exhibit a strictly larger rational witness t1>0.2306073619 at which Ht1DH(z) has a rigorously certified non-real zero, *and* prove an unconditional upper bound ΛDH≤U∗ satisfying: (i) U∗<41 if t1<41 (thereby still overcoming Dobner's ℓ1 barrier at 0.25), or (ii) U∗−t1≤0.05 if t1≥41. A lower-bound witness alone, or an upper-bound reduction that leaves U∗≥0.25 while ΛDH<0.25 remains possible, is partial progress only.
3. **Rigorous numerical certification:** Any finite-height Argument Principle cover, winding-number scan, or Interval–Newton zero enclosure used in the proof must be executed with **rigorous interval/ball arithmetic** (e.g., `FLINT/Arb` or equivalent verified enclosure arithmetic) with explicit truncation, derivative, and rounding error bounds. Uncertified `float64` evaluations, non-interval `mpmath.quad`/`findroot` scans, or heuristic root-tracking do not satisfy this criterion.
4. **Full reproducibility and open checker:** Release complete source code, parameter tables, interval-cover data, and instructions so an independent verifier can deterministically re-check all certified steps without proprietary solvers, reporting wall-clock runtime and memory separately from search costs.
Citations
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).
H. Davenport and H. Heilbronn, On the zeros of certain Dirichlet series I, J. London Math. Soc. 11 (1936), 181–185.
H. Davenport and H. Heilbronn, On the zeros of certain Dirichlet series II, J. London Math. Soc. 11 (1936), 307–312.
N. G. de Bruijn, The roots of trigonometric integrals, Duke Math. J. 17 (1950), no. 3, 197–226.
R. Spira, Some zeros of the Titchmarsh counterexample, Math. Comp. 63 (1994), no. 208, 747–748.
E. P. Balanzario and J. Sánchez-Ortiz, Zeros of the Davenport–Heilbronn counterexample, Math. Comp. 76 (2007), no. 260, 2045–2049.
H. Ki, Y.-O. Kim, and J. Lee, On the de Bruijn–Newman constant, Adv. Math. 222 (2009), no. 1, 281–306.
E. Bombieri and A. Ghosh, Around the Davenport–Heilbronn function, Russian Math. Surveys 66 (2011), no. 2, 221–270.
A. Dobner, A proof of Newman's conjecture for the extended Selberg class, Acta Arith. 201 (2021), 29–62.
D. H. J. Polymath, Effective approximation of heat flow evolution of the Riemann xi function, and a new upper bound for the de Bruijn–Newman constant, Res. Math. Sci. 6 (2019), Art. 31.
D. J. Platt and T. S. Trudgian, The Riemann hypothesis is true up to 3*10^12, Bull. Lond. Math. Soc. 53 (2021), no. 3, 792–797.
R. Freire, C. González-Riquelme, C. Muñoz, F. Poblete, and V. Salinas, Upper bounds on the de Bruijn–Newman constant for Dirichlet L-functions with primitive real-valued characters, arXiv:2610.27343 (2026).