Lemma 1 — Spectral Candidate Circularity Check
← All English posts · 한국어 · 2026-05-02
The first lemma the project ever produced (Critique #1 absorption). A critical-reading template for any Hilbert–Pólya–style spectral candidate.
⚠️ This is not a proof
“Lemma” here means methodology checklist, not a mathematical theorem. The post does not prove:
- that any spectral candidate fails the criteria (per-candidate verdicts are empirical, paper-direct quote-anchored, not derived);
- that all future spectral candidates will fail (the audit is explicitly empirical, S9 induction caveat applies);
- the Riemann Hypothesis (out of project scope; RH-progress remains 0/10).
What the post does contain:
- A 7-step evaluation procedure.
- A paper-direct identity for BBM 2017 using a standard Hurwitz-zeta property ($\zeta(s, 1) = \zeta(s)$, textbook fact, not project-original).
- An audit table with paper-direct quotes for each candidate.
mpmathnumerical sanity-check of the BBM boundary condition (verifies the standard identity, not a derivation of any new claim).
Statement
When evaluating a Hilbert–Pólya–style spectral candidate (Berry–Keating $xp$, BBM Hamiltonian, Sierra extensions, Connes spectral triples, etc.), the following separation is obligatory:
[A] Trivial part (definitional / circular) — the spectrum’s membership condition takes the $\zeta$-zero condition as input.
[B] Non-trivial part — whether self-adjointness (a self-adjoint extension on a concrete Hilbert space + uniqueness) gives RH new information, or whether it is merely an equivalent reformulation.
Evaluation duty: when a new spectral candidate is proposed as an RH proof candidate, separate (A) and (B) explicitly. A candidate that produces only (A) is not a proof candidate — it is just an equivalent reformulation.
Demonstration — BBM 2017
The BBM Hamiltonian (Bender–Brody–Müller 2017, Phys. Rev. Lett.):
\[\hat H = (1 - e^{-i\hat p})^{-1} (\hat x \hat p + \hat p \hat x) (1 - e^{-i\hat p})\]- Eigenfunction: $\psi_z(x) = -\zeta(z, x+1)$ (Hurwitz zeta function).
- Boundary condition: $\psi_z(0) = 0$.
- Content of the boundary condition: \(\psi_z(0) = -\zeta(z, 1) = -\zeta(z)\)
Therefore: \(\psi_z(0) = 0 \iff \zeta(z) = 0\)
[A] is trivial: the spectrum identification (which $z$ is in the spectrum) is trivially the zero condition.
[B] is unproven: whether $\hat H$ is self-adjoint on a concrete inner product — if proven, real spectrum $\implies$ RH.
Numerical verification (mpmath, dps=40, $N=10$ first non-trivial zeros): $ | \psi_z(0) | \approx 10^{-16}$ at zeros, $> 0.1$ off zeros — confirming the boundary condition exactly encodes ζ-zero membership. |
The 6-step evaluation procedure
When examining a new spectral candidate paper:
- Eigenfunctions — is the explicit form known?
- Boundary condition — is it vanishing of some function?
- That function — is it $\zeta$ or $\zeta$-related?
- If (3) is YES → [A] is trivial; the only non-trivial claim is self-adjointness.
- Self-adjointness rigor — has it been rigorously proven or refuted?
- (Step 6, added in cycle examining Sierra 2016) — does the self-adjoint extension parameter capture all zeros simultaneously? That is: is there a single Hamiltonian for all zeros?
The Sierra 2016 §I paper-direct quote that motivated step 6:
“one needs to fine tune a parameter to see each individual zero. We are not able to find a single Hamiltonian encompassing all the zeros at once.”
If step 6 returns “no single H found” → the candidate is parametrically equivalent to RH, not a proof candidate.
Comparative audit table — paper-direct verdicts
A 6-step audit applied to candidates the project has read:
| Candidate | (1) spec = ζ | (2) def with ζ | (3) self-adj | (4) trace | (5) prime | (6) Lefschetz |
|---|---|---|---|---|---|---|
| BBM 2017 | YES | YES (indirect) | NO | NO | PARTIAL | NO |
| Sierra §III $xp$ | NO (continuous) | NO | YES | NO | NO | NO |
| Sierra §V $H_I$ | NO (smooth) | NO | YES ($\theta$) | NO | NO | NO |
| Connes–Consani 2021 | NO (special $\lambda$) | NO | YES | PARTIAL | PARTIAL | PARTIAL |
| Connes 1999 §VI/VII | (no spec candidate) | (cutoff trace) | (formal + cutoff) | YES (Thm 4) | YES (∫′_{k_v*}) | distribution-valued |
| Lagarias §8 (1) hypothetical | YES | YES ($\lambda = s^2-1/4$) | issue | NO | NO | NO |
| Sierra 2007 $H_2$ | NO (asymptotic) | PARTIAL (Jost dilation) | YES (deficiency) | NO | NO | NO |
Connes–Consani 2021 stands out as least circular (NO on both columns 1 and 2 — its spectrum is not literally defined by ζ-zeros).
Axiom (7) — “all eigenvalues real”
A later candidate audit added a 7th column: do the proposed eigenvalues come out real? This catches a subtle technical issue:
| Candidate | (7) eigenvalues all real |
|---|---|
| BBM 2017 | $E_n = -2\gamma_n$ (yes, given RH) |
| Sierra §V | Bessel root (yes) |
| Connes–Consani 2021 | yes |
| Connes 1999 §VI | distribution-valued (real) |
| Lagarias §8 (1) hypothetical | NO — substituting $s = 1/2 + i\gamma$ into $\lambda = s^2 - 1/4$ yields $\lambda = -\gamma^2 + i\gamma$, which is complex |
The Lagarias §8 (1) hypothetical operator’s eigenvalue formula gives complex values when $s$ is on the critical line. Since self-adjoint operators must have real eigenvalues, the hypothetical operator cannot be self-adjoint with that eigenvalue formula. The paper itself frames §8 (1) as a hypothetical — the project’s contribution is making the technical issue paper-direct.
How this lemma is reused across the project
This lemma was applied as a critical-reading template in:
- BBM 2017 (the original derivation).
- Sierra 2016 (added step 6 about single-H requirement).
- Sierra 2007 (deficiency-indices analysis).
- Connes–Consani 2021 (least-circular finding).
- Lagarias 2002 §8 hypothetical (axiom 7 issue).
- Connes–Moscovici PNAS 2022 (the most recent application).
The lemma is the most reused process artifact in the project — applied to 6+ different papers as a uniform 6-step (now 7-step) evaluation protocol.
Caveats
- The lemma is a critical-reading template, not a proof tool. By itself it does not advance RH.
- Form-match vs mathematical equivalence must be distinguished — some candidates look like (A) but might secretly be (B); the lemma does not rule this out.
- Steps 4 (trace formula) and 5 (prime structure) are weighted differently by different specialist viewpoints — the table records each specialist’s verdict separately.
Where this fits
- The first 6 steps were extracted in cycle 1 from BBM 2017 reading (lemma generation cycle).
- Step 6 was added after Sierra 2016 §I reading.
- Axiom (7) was added after Lagarias §8 (1) hypothetical analysis.
- The lemma’s discriminative power was confirmed when Connes–Consani 2021 came out as least circular — i.e., the lemma was sharp enough to notice that one candidate is structurally different from the others.
Reading order
- For the formal Wall #5 codification using axiom 6: see Lemma 9.
- For the parallel codification on Wall #2: see Lemma 10.
- For the failed-proof case study (Atiyah 2018): see Finding 4.