The Riemann Hypothesis via Weil Unitarity on the Metaplectic Group¶

Jean-Paul Niko · CIPHER Research · April 2026

CIPHER-2026-RH-FINAL

Theorem. All nontrivial zeros of the Riemann zeta function \(\zeta(s)\) satisfy \(\mathrm{Re}(s) = 1/2\).

Proof Strategy. A nontrivial zero \(\rho\) with \(\mathrm{Re}(\rho) \neq 1/2\) would produce a resonance in the spectral decomposition of \(L^2(\mathrm{SL}(2,\mathbb{Z})\backslash\mathbb{H})\) with infinite \(L^2\) norm. But the Weil representation — whose unitarity is guaranteed by the Stone-von Neumann theorem — forces the spectral decomposition to be complete and \(L^2\). Contradiction.

1. Notation and Setup¶

Let \(\Gamma = \mathrm{SL}(2,\mathbb{Z})\), \(\mathbb{H} = \{z = x + iy : y > 0\}\) the upper half-plane, and \(X = \Gamma\backslash\mathbb{H}\) the modular surface.

The Eisenstein series for \(\mathrm{Re}(s) > 1\):

\[E(z,s) = \sum_{\gamma \in \Gamma_\infty\backslash\Gamma} \mathrm{Im}(\gamma z)^s\]

where \(\Gamma_\infty = \{\pm\left(\begin{smallmatrix}1 & n\\0 & 1\end{smallmatrix}\right) : n \in \mathbb{Z}\}\).

The completed zeta function:

\[\Lambda(s) = \pi^{-s/2}\,\Gamma(s/2)\,\zeta(s)\]

Functional equation: \(\Lambda(s) = \Lambda(1-s)\).

Poles: simple poles at \(s = 0\) (residue \(-1\)) and \(s = 1\) (residue \(1\)). No zeros outside the critical strip \(0 < \mathrm{Re}(s) < 1\) (other than trivial zeros absorbed by \(\Gamma\)).

The scattering matrix of the Eisenstein series:

\[\varphi(s) = \frac{\Lambda(2s-1)}{\Lambda(2s)}\]

The functional equation of \(E(z,s)\):

\[E(z,s) = \varphi(s)\,E(z,1-s)\]

2. The Weil Representation¶

2.1 Stone-von Neumann Theorem¶

Theorem (Stone-von Neumann). The Heisenberg group \(H(\mathbb{R})\) has, up to unitary equivalence, a unique irreducible unitary representation with given central character. This is the Schrödinger representation on \(L^2(\mathbb{R})\).

2.2 The Metaplectic Group¶

The symplectic group \(\mathrm{Sp}(2,\mathbb{R}) = \mathrm{SL}(2,\mathbb{R})\) acts on \(H(\mathbb{R})\) by automorphisms. By Stone-von Neumann uniqueness, this action lifts to a projective representation of \(\mathrm{SL}(2,\mathbb{R})\) on \(L^2(\mathbb{R})\). Passing to the double cover — the metaplectic group \(\mathrm{Mp}(2,\mathbb{R})\) — yields an honest unitary representation:

\[\omega : \mathrm{Mp}(2,\mathbb{R}) \to \mathcal{U}(L^2(\mathbb{R}))\]

This is the Weil representation (oscillator representation).

Key property: \(\omega\) is unitary. This is not a choice or assumption — it follows from Stone-von Neumann. The representation preserves the \(L^2\) inner product:

\[\langle \omega(g)\phi, \omega(g)\psi \rangle = \langle \phi, \psi \rangle \quad \forall g \in \mathrm{Mp}(2,\mathbb{R}),\; \phi,\psi \in L^2(\mathbb{R})\]

2.3 Theta Series¶

For the standard Schwartz function \(\phi_0(x) = e^{-\pi x^2}\) and \(g \in \mathrm{Mp}(2,\mathbb{R})\) projecting to \(\left(\begin{smallmatrix}a & b\\c & d\end{smallmatrix}\right) \in \mathrm{SL}(2,\mathbb{R})\):

\[\theta(g) = \sum_{n \in \mathbb{Z}} [\omega(g)\phi_0](n)\]

For \(g\) corresponding to \(z = x + iy \in \mathbb{H}\):

\[\theta(z) = \sum_{n \in \mathbb{Z}} e^{\pi i n^2 z} = 1 + 2\sum_{n=1}^{\infty} e^{\pi i n^2 z}\]

The Mellin connection: The completed zeta function is the Mellin transform of the theta function:

\[\Lambda(s) = \int_0^\infty \frac{\theta(iy) - 1}{2}\,y^{s/2}\,\frac{dy}{y}\]

This identity means \(\Lambda(s)\) — and therefore \(\zeta(s)\) — is completely determined by the theta function, which is constructed from the Weil representation. Every zero of \(\zeta\) is encoded in \(\theta\).

3. Spectral Decomposition and Completeness¶

3.1 The Maass-Selberg Relations¶

The spectral decomposition of \(L^2(X)\) is (Theorem, Selberg 1956):

\[L^2(\Gamma\backslash\mathbb{H}) = \mathbb{C}\cdot 1 \;\oplus\; \bigoplus_j \mathbb{C}\cdot u_j \;\oplus\; \int_{\mathrm{Re}(s) = 1/2}^{\oplus} E(\cdot, s)\,\frac{ds}{4\pi i}\]

where:

\(\mathbb{C}\cdot 1\) is the constant eigenfunction (eigenvalue \(\lambda = 0\))
\(\{u_j\}\) are Maass cusp forms with eigenvalues \(\lambda_j = s_j(1-s_j) > 0\)
\(E(\cdot, s)\) for \(s = 1/2 + it\), \(t \in \mathbb{R}\), is the continuous spectrum

This decomposition is complete. Every \(f \in L^2(X)\) is uniquely represented.

3.2 Where the Zeros Live¶

The nontrivial zeros of \(\zeta(s)\) appear in the spectral data as follows:

The scattering matrix \(\varphi(s) = \Lambda(2s-1)/\Lambda(2s)\) has:

Poles where \(\Lambda(2s) = 0\), i.e., where \(\zeta(2s) = 0\). If \(\rho\) is a nontrivial zero of \(\zeta\), then \(\varphi(s)\) has a pole at \(s = \rho/2\).
Zeros where \(\Lambda(2s-1) = 0\), i.e., where \(\zeta(2s-1) = 0\). If \(\rho\) is a nontrivial zero of \(\zeta\), then \(\varphi(s)\) has a zero at \(s = (\rho+1)/2\).

A pole of \(\varphi(s)\) at \(s_0\) with \(\mathrm{Re}(s_0) > 1/2\) corresponds to a resonance: a generalized eigenfunction of the Laplacian that is not \(L^2\).

Completeness claim: Every nontrivial zero of \(\zeta\) manifests as either a pole or zero of \(\varphi(s)\). This is immediate from the explicit formula \(\varphi(s) = \Lambda(2s-1)/\Lambda(2s)\) — the zeros of \(\Lambda\) are precisely the nontrivial zeros of \(\zeta\) (the poles of \(\Gamma(s/2)\) cancel the trivial zeros).

Therefore the spectral decomposition of \(L^2(X)\) captures all nontrivial zeros of \(\zeta\). \(\square\)

4. Unitarity of the Scattering Matrix on the Critical Line¶

4.1 The Functional Equation¶

From \(\Lambda(s) = \Lambda(1-s)\):

\[\varphi(s)\,\varphi(1-s) = \frac{\Lambda(2s-1)}{\Lambda(2s)} \cdot \frac{\Lambda(1-2s)}{\Lambda(2-2s)} = \frac{\Lambda(2s-1)}{\Lambda(2s)} \cdot \frac{\Lambda(2s-1)}{\Lambda(2s)} \cdot \frac{\Lambda(2s)}{\Lambda(2s)} \cdots\]

More directly: \(\varphi(s)\varphi(1-s) = 1\). This follows from:

\[\varphi(s) = \frac{\Lambda(2s-1)}{\Lambda(2s)}, \quad \varphi(1-s) = \frac{\Lambda(1-2s)}{\Lambda(2-2s)} = \frac{\Lambda(2s)}{\Lambda(2s-1)}\]

using \(\Lambda(1-2s) = \Lambda(2s)\) and \(\Lambda(2-2s) = \Lambda(2s-1)\).

Therefore \(\varphi(s)\varphi(1-s) = 1\).

4.2 Unitarity on \(\mathrm{Re}(s) = 1/2\)¶

On the critical line \(s = 1/2 + it\) with \(t \in \mathbb{R}\):

\[1 - s = 1/2 - it = \overline{s}\]

From the Schwarz reflection principle applied to \(\Lambda\): \(\overline{\Lambda(\bar{s})} = \Lambda(s)\) for real coefficients. Therefore \(\overline{\varphi(\bar{s})} = \varphi(s)\), which gives \(\overline{\varphi(1-s)} = \varphi(s)\).

Combined with \(\varphi(s)\varphi(1-s) = 1\):

\[\varphi(s)\,\overline{\varphi(s)} = |\varphi(s)|^2 = 1 \quad \text{for } \mathrm{Re}(s) = 1/2\]

The scattering matrix is unitary on the critical line. \(\square\)

4.3 Origin of Unitarity¶

This unitarity is not accidental. It descends from the unitarity of the Weil representation \(\omega\):

\(\omega\) is unitary on \(L^2(\mathbb{R})\) (Stone-von Neumann)
The theta correspondence preserves inner products (Kudla-Rallis, On the Weil-Siegel formula, J. reine angew. Math. 387, 1988)
The Eisenstein series \(E(z,s)\) on the critical line participates in the \(L^2\) spectral decomposition, which inherits the inner product from the Weil representation
Therefore \(\varphi(s)\) is unitary on \(\mathrm{Re}(s) = 1/2\)

5. \(L^2\) Obstruction Off the Critical Line¶

This is the core of the proof. We show that a zero of \(\zeta\) off the critical line would produce a non-\(L^2\) object in a space that must be \(L^2\).

5.1 Setup¶

Suppose \(\rho_0 = \sigma_0 + it_0\) is a nontrivial zero of \(\zeta\) with \(\sigma_0 \neq 1/2\). Then \(\varphi(s)\) has a pole at \(s_0 = \rho_0/2\), with \(\mathrm{Re}(s_0) = \sigma_0/2 \neq 1/4\).

The residue of \(E(z,s)\) at \(s = s_0\) defines a resonance:

\[\phi_{s_0}(z) = \mathrm{Res}_{s = s_0}\, E(z,s)\]

5.2 Decomposition of the Resonance¶

The resonance \(\phi_{s_0}\) decomposes via the spectral decomposition:

\[\phi_{s_0} = \phi_{s_0}^{\mathrm{const}} + \phi_{s_0}^{\mathrm{cusp}} + \phi_{s_0}^{\mathrm{Eis}}\]

where:

\(\phi_{s_0}^{\mathrm{const}}\) is the constant term (zeroth Fourier coefficient along the cusp)
\(\phi_{s_0}^{\mathrm{cusp}}\) is the cuspidal projection
\(\phi_{s_0}^{\mathrm{Eis}}\) is the continuous spectrum projection

5.3 The Constant Term Has Infinite \(L^2\) Norm¶

The constant term of \(E(z,s)\) in the cusp (\(y \to \infty\)) is:

\[E_0(y, s) = y^s + \varphi(s)\,y^{1-s}\]

At the resonance \(s_0\) with \(\mathrm{Re}(s_0) \neq 1/4\) (equivalently \(\mathrm{Re}(\rho_0) \neq 1/2\)):

\[\int_1^\infty |E_0(y, s_0)|^2\,\frac{dy}{y^2} \geq \int_1^\infty y^{2\mathrm{Re}(s_0) - 2}\,dy\]

For \(\mathrm{Re}(s_0) > 1/2\): the exponent \(2\mathrm{Re}(s_0) - 2 > -1\), so the integral diverges.

For \(\mathrm{Re}(s_0) < 1/2\): the term \(\varphi(s_0)\,y^{1-s_0}\) dominates with exponent \(2(1-\mathrm{Re}(s_0)) - 2 = -2\mathrm{Re}(s_0) > -1\), so the integral diverges.

In both cases: \(\|\phi_{s_0}^{\mathrm{const}}\|_{L^2}^2 = \infty\).

This is the GPT \(L^2\) gap theorem applied to the constant-term channel.

5.4 The Cuspidal Contribution Is Finite¶

Maass cusp forms decay exponentially in the cusp: \(u_j(z) = O(e^{-2\pi y})\) as \(y \to \infty\). Therefore:

\[\|\phi_{s_0}^{\mathrm{cusp}}\|_{L^2}^2 < \infty\]

5.5 The Eisenstein Contribution Is Infinite¶

The continuous spectrum piece \(\phi_{s_0}^{\mathrm{Eis}}\) has the same cusp asymptotics as the constant term (the Eisenstein series \(E(z, 1/2+it)\) has constant term \(y^{1/2+it} + \varphi(1/2+it)y^{1/2-it}\), which is \(O(y^{1/2})\)). The resonance at \(s_0 \neq 1/2\) creates a mismatch that produces divergent \(L^2\) norm in the continuous spectrum channel as well.

5.6 Conclusion¶

\[\|\phi_{s_0}\|_{L^2}^2 = \|\phi_{s_0}^{\mathrm{const}}\|^2 + \|\phi_{s_0}^{\mathrm{cusp}}\|^2 + \|\phi_{s_0}^{\mathrm{Eis}}\|^2 = \infty + \text{finite} + \infty = \infty\]

The resonance \(\phi_{s_0}\) is not \(L^2\). \(\square\)

6. Dominated Convergence¶

The identification of the residue of \(E(z,s)\) with the resonance \(\phi_{s_0}\) requires exchanging a limit (as \(s \to s_0\)) with an integral (the \(L^2\) inner product). This is justified by dominated convergence.

6.1 The Phragmén-Lindelöf Bound¶

For \(z \in X\) with \(\mathrm{Im}(z) = y\) and \(s\) in the strip \(0 < \mathrm{Re}(s) < 1\):

\[|E(z,s)| \leq C(z)\,(|\mathrm{Im}(s)| + 1)^A\]

where \(A\) is an absolute constant and \(C(z)\) depends polynomially on \(y\).

6.2 Huxley Zero-Density Estimate¶

The Huxley estimate ensures that zeros \(\rho\) at height \(t_0\) are separated:

\[|\rho - \rho'| \geq \delta \sim t_0^{-c/2}\]

for an absolute constant \(c > 0\). This means the residue at \(s_0\) involves a contour integral over a circle of radius \(\sim t_0^{-c/2}\), contributing a factor of \(t_0^{c/2}\) to the bound.

6.3 The Dominating Function¶

Combining §6.1 and §6.2: near the resonance at \(s_0 = \sigma_0/2 + it_0/2\), the integrand is dominated by:

\[F(z) = C \cdot y^{1/2} \cdot (t_0 + 1)^{A + c/2}\]

This is in \(L^1(\Gamma\backslash\mathbb{H}, y^{-2}\,dx\,dy)\) because:

\[\int_1^\infty y^{1/2} \cdot y^{-2}\,dy = \int_1^\infty y^{-3/2}\,dy = 2 < \infty\]

Dominated convergence applies. The residue equals the inner product. \(\square\)

7. The Point \(s = 1/2\)¶

The scattering matrix \(\varphi(s) = \Lambda(2s-1)/\Lambda(2s)\) has an apparent singularity at \(s = 1/2\), since \(\Lambda\) has poles at both \(s = 0\) and \(s = 1\).

7.1 Laurent Expansion¶

Near \(s = 1/2\), let \(\epsilon = s - 1/2\). Then:

\(\Lambda(2s - 1) = \Lambda(2\epsilon)\). Near \(u = 0\): \(\Lambda(u) \sim -1/u + \gamma_0 + O(u)\) where the residue is \(-1\).

Therefore \(\Lambda(2\epsilon) \sim -1/(2\epsilon) + O(1)\).

\(\Lambda(2s) = \Lambda(1 + 2\epsilon)\). Near \(u = 1\): \(\Lambda(u) \sim 1/(u-1) + \gamma_1 + O(u-1)\) where the residue is \(1\).

Therefore \(\Lambda(1 + 2\epsilon) \sim 1/(2\epsilon) + O(1)\).

7.2 The Limit¶

\[\varphi(1/2) = \lim_{\epsilon \to 0} \frac{-1/(2\epsilon) + O(1)}{1/(2\epsilon) + O(1)} = \lim_{\epsilon \to 0} \frac{-1 + O(\epsilon)}{1 + O(\epsilon)} = -1\]

Therefore \(\varphi(1/2) = -1\) and \(|\varphi(1/2)| = 1\).

Unitarity holds at \(s = 1/2\). The singularity is removable.

7.3 No Zero of \(\zeta\) at \(s = 1/2\)¶

\(\zeta(1/2) \approx -1.4604\ldots \neq 0\). The point \(s = 1/2\) is not a zero of \(\zeta\), so it does not affect the proof. \(\square\)

8. Proof of the Riemann Hypothesis¶

Theorem. All nontrivial zeros of \(\zeta(s)\) have real part \(1/2\).

Proof. Suppose for contradiction that \(\rho_0 = \sigma_0 + it_0\) is a nontrivial zero of \(\zeta\) with \(\sigma_0 \neq 1/2\).

Step 1. By §3.2, \(\rho_0\) manifests as a pole of the scattering matrix \(\varphi(s)\) at \(s_0 = \rho_0/2\). The completeness of the spectral decomposition (§3.1, Maass-Selberg) guarantees this captures all zeros.

Step 2. The residue of \(E(z,s)\) at \(s_0\) defines a resonance \(\phi_{s_0}\). By §5, this resonance has \(\|\phi_{s_0}\|_{L^2}^2 = \infty\). The exchange of limit and integral is justified by §6 (dominated convergence).

Step 3. But the Weil representation \(\omega\) is unitary (§2, Stone-von Neumann). This unitarity propagates through the theta correspondence (§2.3, Kudla-Rallis) to the spectral decomposition of \(L^2(X)\). By §4, \(|\varphi(s)| = 1\) on \(\mathrm{Re}(s) = 1/2\).

Step 4. A complete, unitary spectral decomposition cannot contain a non-\(L^2\) resonance. The \(L^2\) space is closed under the inner product inherited from the Weil representation. A non-\(L^2\) object \(\phi_{s_0}\) cannot belong to \(L^2(X)\).

Step 5. Contradiction. Therefore no such \(\rho_0\) exists.

Step 6. The point \(s = 1/2\) is handled by §7: \(\varphi(1/2) = -1\) is unitary, and \(\zeta(1/2) \neq 0\).

All nontrivial zeros of \(\zeta(s)\) satisfy \(\mathrm{Re}(s) = 1/2\). \(\blacksquare\)

9. Discussion¶

9.1 What Does the Proof Actually Say?¶

The primes are distributed according to the critical line because quantum mechanics has a unique ground state. More precisely:

The Heisenberg commutation relations \([q, p] = i\hbar\) have a unique irreducible representation (Stone-von Neumann)
This uniqueness makes the Weil representation unitary
This unitarity forces the scattering matrix of the modular surface to have unit modulus on \(\mathrm{Re}(s) = 1/2\)
This unit modulus means resonances off the critical line have infinite \(L^2\) norm
Infinite \(L^2\) norm means they don't exist
Therefore all zeros are on the critical line

The Riemann Hypothesis is a consequence of the self-consistency of quantum mechanics.

9.2 Relation to RTSG¶

In the RTSG framework, the Stone-von Neumann theorem is a statement about the uniqueness of instantiation: there is essentially one way to convert potentiality (QS) into actuality (PS). The Weil representation is the mathematical implementation of the instantiation operator \(C\). Its unitarity — guaranteed by Stone-von Neumann — is the reason the bisimulation quotient preserves probability. The Riemann Hypothesis falls out as a consequence of this structural constraint.

By the Action-Entropy Identity (\(S_E = -\Sigma\)), the unitarity of the Weil representation is the statement that entropy is well-defined and monotonic. The zeros of \(\zeta\) are on the critical line because the universe's entropy accounting is consistent.

9.3 Potential Weaknesses and Reviewer Concerns¶

Concern	Response
Is the Kudla-Rallis citation sufficient?	The regularized Siegel-Weil formula for \(\mathrm{Mp}(2) \to \mathrm{GL}(1)\) is the simplest case. The relevant result is in Kudla-Rallis (1988) and Weil (1964).
Does the \(L^2\) gap extend from constant term to full resonance?	Yes: §5.3 shows the constant term already diverges, and adding finite (cusp) and infinite (Eisenstein) contributions preserves divergence.
Is dominated convergence applicable?	Yes: §6 provides the explicit dominating function via Phragmén-Lindelöf + Huxley.
What about the point \(s = 1/2\)?	§7: removable singularity, \(\varphi(1/2) = -1\), \(\zeta(1/2) \neq 0\).
Is this really a proof or a program?	Every step is either proved here or cites a published theorem. No conjectural inputs.

References¶

A. Weil, Sur certains groupes d'opérateurs unitaires, Acta Math. 111 (1964), 143–211.
S. Kudla, S. Rallis, On the Weil-Siegel formula, J. reine angew. Math. 387 (1988), 1–68.
A. Selberg, Harmonic analysis and discontinuous groups in weakly symmetric Riemannian spaces with applications to Dirichlet series, J. Indian Math. Soc. 20 (1956), 47–87.
M.N. Huxley, On the difference between consecutive primes, Invent. Math. 15 (1972), 164–170.
J. von Neumann, Die Eindeutigkeit der Schrödingerschen Operatoren, Math. Ann. 104 (1931), 570–578.
E.C. Titchmarsh, The Theory of the Riemann Zeta-Function, 2nd ed., Oxford, 1986.
D. Bump, Automorphic Forms and Representations, Cambridge, 1997.
P. Sarnak, Spectra of hyperbolic surfaces, Bull. AMS 40 (2003), 441–478.