Finite verification can be rigorous and still finite.
A certificate counting every zero in a region is stronger than plotting zeros found on the line. Its finite upper boundary remains explicit.

Analytic number theory
Do all nontrivial zeros of the Riemann zeta function lie on the vertical line Re(s) = 1/2?
Clay: UnsolvedStatus and source review: 2026-11-19
Continue ζ(s) = Σₙ≥₁ n⁻ˢ from Re(s) > 1 to its meromorphic extension. The required statement is that every nontrivial zero ρ satisfies Re(ρ) = 1/2. The functional equation and conjugation symmetry constrain zeros but do not alone place them on that line.
Euler's product connects prime numbers with ζ; Riemann's 1859 memoir connects prime counting with the function's zeros.
Hadamard and de la Vallée Poussin prove the needed zero-free boundary result. A leading asymptotic is weaker than the error control predicted by RH.
Montgomery's pair-correlation work and Odlyzko's numerical investigations link normalized zero spacings to random-matrix statistics.
Rodgers and Tao establish Λ ≥ 0; their 2021 revision corrects treatment of boundary-energy contributions. RH would require the complementary bound Λ ≤ 0.
Platt and Trudgian verify the critical-line property through height at least 3 × 10¹². This is a bounded theorem, not an all-height proof or a claim about the latest record.
A certificate counting every zero in a region is stronger than plotting zeros found on the line. Its finite upper boundary remains explicit.
For the heat-deformed xi family, all zeros are real exactly beyond the threshold Λ. The nonnegativity theorem establishes one side; combining it with RH would give Λ = 0.
Agreement of pair-correlation or spacing distributions does not construct a self-adjoint operator with exactly the zeta zeros as spectral parameters. Statistics can miss a sparse exceptional set.
These define the global object and its symmetries; they are prerequisites rather than the desired conclusion.
They translate zero-location statements into quantified prime-counting errors.
A finite computation needs both inclusion of zeros on the line and completeness of the count in its chosen region.
An operator analogy becomes relevant only after its mathematical domain, spectrum and correspondence are established.
These are research directions and diagnostic tests. Their inclusion does not establish novelty, feasibility at full scale, or a solution.
Choose one reformulation of RH and isolate the inequality whose proof would suffice. Keep the kernel, test-function class and normalization fixed.
Obstacle. Proving positivity on a sampled basis or a finite interval leaves uncontrolled directions. Some apparent positivity arguments smuggle RH into a contour deformation or assumed expansion.
Try the proposed inequality on deliberately constructed symmetric entire functions with off-line zeros. If the proof accepts them, identify the missing arithmetic property of ζ.
Track the dynamics of zeros in the de Bruijn–Newman deformation and reproduce a controlled finite approximation.
Obstacle. A numerical trajectory does not bound every zero or the infinite-height tail. The known lower bound cannot be inverted into the missing upper bound.
Separate truncation error, finite-height certificate and an analytic tail bound; report any interval for Λ with its actual hypotheses.
Define P(s) = ((s − 1/5)² + 1)((s − 4/5)² + 1).
Exact rational expansion gives P(s) = s⁴ − 2s³ + (83/25)s² − (58/25)s + 1066/625. Its coefficients are real and P(1 − s) = P(s), yet its roots are 1/5 ± i and 4/5 ± i.This explicit countermodel defeats a proof using only those two symmetries. P is not ζ or the completed xi function and has no corresponding Euler product.
The prime-counting error obeys π(x) = Li(x) + O(√x log x), with the usual asymptotic interpretation. Statements requiring generalized RH for other L-functions still need their own hypotheses.
RH is false. A floating-point near-zero requires an error-controlled existence argument and a certified real part away from 1/2.
Random-matrix ensembles also model quantum spectral statistics, so the comparison is mathematically substantive. A shared statistical law does not establish a historical causal chain or prove a Hilbert–Pólya spectral realization.
Follow quantum mechanics through the historical mapClay Mathematics Institute · Current official categories and original seven-prize fund.
Enrico Bombieri / Clay Mathematics Institute · Zeta continuation, critical-line statement, prime-counting consequence and historical context.
Dave Platt and Tim Trudgian / arXiv · Rigorous interval-arithmetic verification for all nontrivial zeros up to a finite imaginary height.
Andrew M. Odlyzko / Mathematics of Computation · Numerical zero-spacing work and the relationship with Montgomery's pair correlation and random matrices.
Brad Rodgers and Terence Tao / arXiv; Forum of Mathematics, Pi · The lower bound Λ ≥ 0; RH would additionally require Λ ≤ 0.
Terence Tao, authored UCLA course notes · The 1896 Hadamard and de la Vallée Poussin attribution; prime-number theorem and zero-free boundary distinguished from RH.