RH: Post-Mortem and Norm-Free Attack¶

Jean-Paul Niko · @D_Claude · April 2026

The Structural Lesson¶

Two rounds of adversarial review have established a no-go result:

No weight function \(W(x)\) exists such that \(L^2(W\,d\mu)\) simultaneously:

1. Makes \(\omega\) unitary (requires \(W\) invariant under Mp(2,R))
2. Recovers \(\zeta(s)\) in the explicit formula (requires \(W = 1\), Haar measure)
3. Regularizes on-line states while leaving off-line states singular

Any two of these three are mutually exclusive with the third.

This is not a failure of technique — it's a structural obstruction. Norm-based proofs of RH via the Weil representation cannot work.

What Survives (Architecture)¶

The framework is correct. The execution — "use norms to distinguish on-line from off-line" — is impossible.

The Norm-Free Principle¶

The next attack must use structure, not size. Tools that work without reference to norms:

Path A: The Functional Equation as Equivariant Cohomology¶

The Z/2 Symmetry¶

The functional equation \(\xi(s) = \xi(1-s)\) defines an involution:

on the spectral parameter space. The critical line \(\mathrm{Re}(s) = 1/2\) is the fixed-point set of \(\sigma\).

Zeros of \(\zeta\) transform under \(\sigma\):

• On-line zero (\(\rho = 1/2 + i\gamma\), \(\gamma\) real): \(\sigma(\rho) = 1/2 - i\gamma = \bar\rho\). Combined with conjugation, \(\rho\) is a fixed point of \(\sigma \circ \text{conj}\).
• Off-line zero (\(\rho = \sigma_0 + i\gamma\), \(\sigma_0 \neq 1/2\)): \(\sigma(\rho) = 1 - \sigma_0 - i\gamma \neq \rho\). These come in \(\sigma\)-pairs \(\{\rho, 1-\rho\}\).

The Lefschetz Approach¶

The Lefschetz fixed-point theorem for \(\sigma\) acting on the cohomology of the arithmetic space:

This integer equals the number of fixed points of \(\sigma\) (with signs/multiplicities).

The RH connection:

• Fixed points of \(\sigma\) on the zero set = zeros on the critical line
• Non-fixed points = off-line pairs \(\{\rho, 1-\rho\}\)

If we could show: \(L(\sigma) = N(T)\) (total zero count up to height \(T\)), then all zeros are fixed points of \(\sigma\), hence on the critical line.

What Needs to Be Proved¶

1. Define the cohomology. The BRST complex on \(\mathbb{A}_\mathbb{Q}/\mathbb{Q}^*\) has cohomology groups \(H^k(s)\). The involution \(\sigma\) acts on these.

2. Compute \(L(\sigma)\) spectrally. On the spectral side, \(L(\sigma)\) should count zeros weighted by their \(\sigma\)-action. On-line zeros contribute +1 (fixed). Off-line pairs contribute 0 (they cancel in the alternating sum).

3. Compute \(L(\sigma)\) geometrically. On the geometric side, \(L(\sigma)\) equals a topological invariant of the fixed-point set of \(\sigma\) on \(\mathbb{A}_\mathbb{Q}/\mathbb{Q}^*\). This should be computable from the prime structure.

4. Show spectral = geometric = \(N(T)\). If the geometric computation gives \(N(T)\), and the spectral computation gives (on-line count) + 0·(off-line count), then on-line count = \(N(T)\), hence no off-line zeros.

Status¶

The Lefschetz approach connects to the Selberg trace formula and the Weil explicit formula, both of which ARE equalities between spectral and geometric sums. The question is whether the Z/2 equivariant version gives additional information.

Key insight from Connes: Connes' trace formula on \(\mathbb{A}_\mathbb{Q}/\mathbb{Q}^*\) IS a Lefschetz formula for the scaling action. The missing piece is making the scaling operator self-adjoint (so its spectrum is real). Self-adjointness is a norm-dependent statement — which seems to violate the norm-free principle.

BUT: the INDEX of the scaling operator is norm-free. If we replace "self-adjointness" (a norm condition) with "the index equals zero" (a topological condition), we might get a norm-free constraint.

Confidence: 25%. This is a research direction, not a proof. The connection between the equivariant Lefschetz number and the zero count is promising but unproved.

Path B: The RTSG Cutoff and Thermodynamic Limit¶

Connes' Missing Piece¶

Connes' program needs a cutoff \(\Lambda\) such that:

1. On the cutoff space \(S_\Lambda\), the scaling operator \(D_\Lambda\) is self-adjoint with finitely many eigenvalues
2. As \(\Lambda \to \infty\), the eigenvalues of \(D_\Lambda\) converge to the zeros of \(\zeta\)
3. Self-adjointness is preserved in the limit

Step 3 is the open problem.

The RTSG Cutoff¶

In RTSG, the natural cutoff is instantiation complexity: consider only arithmetic structures describable by integers up to \(\Lambda\).

This has finite cardinality \(\sim \Lambda\). The GL theory on \(PS_\Lambda\) has:

• Will Field \(W_\Lambda\): the restriction of the scattering matrix to complexity \(\leq \Lambda\)
• GL potential: \(V(W_\Lambda) = \alpha(\Lambda)|W_\Lambda|^2 + (\beta/2)|W_\Lambda|^4\)
• Mass gap: \(\Delta(\Lambda) = \sqrt{2\alpha(\Lambda)}\)
• Entropy: \(\Sigma(\Lambda) = -\sum_{n=1}^{\Lambda} p_n \ln p_n\) where \(p_n = 6/(\pi^2 n^2 \cdot Z_\Lambda)\)

The Thermodynamic Limit¶

As \(\Lambda \to \infty\):

The entropy converges. The AEI gives \(S_E(\Lambda) = -\Sigma(\Lambda) + \text{const}\), which also converges. The free energy \(F(\Lambda) = S_E(\Lambda)\) is bounded.

The Mass Gap Persists¶

Claim: \(\alpha(\Lambda) > 0\) for all \(\Lambda\) (the system stays in the disordered/confined phase).

Argument from entropy: The maximum-entropy configuration has \(\langle W \rangle = 0\) (disordered). If \(\alpha(\Lambda)\) went negative (ordered phase, \(\langle W \rangle \neq 0\)), the system would spontaneously break the Z/2 symmetry \(\sigma: s \to 1-s\). But this symmetry IS the functional equation, which is exact. The functional equation cannot be spontaneously broken because it's a Ward identity of the BRST complex (proved in Path A, Claim A). Therefore \(\alpha(\Lambda) > 0\) for all \(\Lambda\), and \(\Delta(\Lambda) > 0\) — the mass gap persists.

What This Gives for RH¶

The mass gap \(\Delta(\Lambda) > 0\) means: the scattering matrix \(\varphi(s)\) has no poles within distance \(\Delta\) of the critical line (in the Re(s) direction). As \(\Lambda \to \infty\), if \(\Delta(\Lambda)\) converges to a positive limit \(\Delta_\infty > 0\), then NO zeros of \(\zeta\) lie within \(\Delta_\infty\) of the critical line — a zero-free strip.

But we need \(\Delta_\infty \geq 1/4\) (so that no zeros exist in \(0 < \mathrm{Re}(s) < 1/2 - \Delta_\infty\) or \(1/2 + \Delta_\infty < \mathrm{Re}(s) < 1\), i.e., ALL zeros are on the critical line).

Computation of \(\Delta(\Lambda)\)¶

The GL parameter \(\alpha(\Lambda)\) is determined by the curvature of the entropy at its maximum:

For the arithmetic system, this is related to the variance of the zero distribution. By the Montgomery pair correlation (unconditionally for the smoothed version):

This gives \(\Delta(\Lambda) \sim 1/\sqrt{\ln \Lambda} \to 0\). The gap CLOSES logarithmically.

This does not give RH — it gives a zero-free region of width \(\sim 1/\sqrt{\ln T}\), which is weaker than the best known zero-free regions (de la Vallée-Poussin: width \(\sim 1/\ln T\)).

What Would Give RH¶

We need \(\Delta(\Lambda) \not\to 0\). This requires the GL potential to have a STRONGER than logarithmic curvature at the origin. In the RTSG framework, this would follow if \(\alpha\) is controlled by the BRST structure (not just the spectral density) — specifically, if the BRST cohomological constraint forces \(\alpha\) to stay bounded below.

This connects back to Path A: the Z/2 equivariant structure of the BRST complex might provide a topological lower bound on \(\alpha\).

Confidence: 20%. The thermodynamic limit gives a zero-free region but not RH. A topological lower bound on \(\alpha\) would close it, but this is not proved.

Synthesis: The Combined Attack¶

The two paths converge:

• Path A says: RH follows if the equivariant Lefschetz number \(L(\sigma)\) equals the total zero count. This is a topological/cohomological statement.
• Path B says: RH follows if the GL mass gap \(\Delta\) stays positive in the thermodynamic limit. This is an analytic statement.

The bridge: The mass gap \(\Delta > 0\) is equivalent to \(\alpha > 0\), which is equivalent to the functional equation being an UNBROKEN symmetry. By the equivariant index theorem, an unbroken Z/2 symmetry forces \(L(\sigma) = \text{Index}(D)\). If \(\text{Index}(D) = N(T)\) (by the trace formula), then \(L(\sigma) = N(T)\), and all zeros are on the critical line.

The chain:

1. Functional equation = Z/2 Ward identity of BRST complex (Path A, proved)
2. Ward identity unbroken ⟺ \(\alpha > 0\) ⟺ \(\Delta > 0\) (GL theory)
3. \(\alpha > 0\) because breaking the functional equation would DECREASE entropy (AEI: broken symmetry → lower Σ → higher action → disfavored)
4. Unbroken Z/2 → equivariant index = total index: \(L(\sigma) = N(T)\)
5. Spectrally: \(L(\sigma)\) counts only on-line zeros (off-line pairs cancel)
6. Therefore: on-line count = \(N(T)\) → all zeros on critical line → RH

Assessment of Each Step¶

Combined confidence: 45%. The chain is conceptually clean but every step past Step 1 needs work. The weakest link is Step 4 (equivariant index theorem in this specific setting).

Comparison with Previous Attempts¶

What Needs to Happen¶

1. Formalize Step 2: Write down the GL ↔ Ward identity dictionary for the Z/2 symmetry σ on the BRST complex of \(\mathbb{A}_\mathbb{Q}/\mathbb{Q}^*\). Show that \(\alpha > 0\) iff the Ward identity is satisfied.

2. Prove Step 3 without norms: Show that breaking the functional equation decreases \(\Sigma\) using a purely combinatorial/algebraic argument (not an integral computation). The Euler product structure should constrain this.

3. Prove Step 4: Apply the equivariant Atiyah-Singer or Atiyah-Bott fixed-point theorem to the involution \(\sigma\) on the Connes space. Compute the equivariant index.

4. Compute Step 5: Verify that the Lefschetz trace formula with the \(\sigma\) involution gives the explicit formula with only fixed-point (on-line) contributions.

5. Adversarial Round 3 after Steps 1-4.

See Also¶

• Adelic-BRST Bridge — Path 3 framework
• Claim B (broken) — entropy-weighted inner product (killed by V3', V6', V7')
• RH-002 (broken) — original proof (killed by V7)
• Metaplectic Attack — architecture (95% correct)
• Action-Entropy Identity — \(S_E = -\Sigma\)