Spectral Geometry of the Primes

1. Problem Statement

The paper addresses the fundamental question of whether the sequence of prime numbers possesses an intrinsic "geometry" that can be probed using spectral analysis, independent of their embedding in Euclidean space. While connections between prime distribution and spectral statistics (e.g., the Riemann Hypothesis and Random Matrix Theory) are well-known, this work seeks to construct a self-contained arithmetic geometry directly from the primes. The core problem is to define a metric-like structure based on "arithmetic dissimilarity" rather than spatial distance and to determine if this structure yields a coherent spectral dimension and thermodynamic behavior.

2. Methodology

The author constructs a framework called Coherence Geometry, treating primes as elements of a discrete, unordered set where structure arises solely from pairwise divergences.

A. Construction of the Coherence Hamiltonian

1. Divergence Matrix ($\delta_{ij}$): A symmetric matrix is defined over the first $N$ primes ($P_N$) representing arithmetic separation. Several divergence functions are tested:
   • Logarithmic scale: $\delta_{ij} = \log(p_i p_j)$
   • Relative asymmetry: $\delta_{ij} = (\log(p_i/p_j))^2$
   • Information-theoretic (Jensen-Shannon type): $\delta_{ij} = \log\left(\frac{p_i + p_j}{2\sqrt{p_i p_j}}\right)$
   • Index-based fractal: $\delta_{ij} = |\log i - \log j|^\gamma$
2. Coherence Kernel ($K_{ij}$): An unnormalized kernel is formed by exponentiating the negative divergence: $K_{ij} = \exp(-\delta_{ij}/\delta_0)$, where $\delta_0$ is a coherence length scale.
3. Normalization and Laplacian:
   • A degree matrix $D$ is defined where $D_{ii} = \sum_j K_{ij}$.
   • A symmetric normalized kernel is formed: $S = D^{-1/2} K D^{-1/2}$.
   • Two Laplacian-like generators are considered:
     • Normalized: $L = I - S$ (bounded spectrum).
     • Unnormalized (Combinatorial): $L_c = D - K$ (unbounded as $N \to \infty$).
4. Hamiltonian Definition: The primary operator of study is the squared Hamiltonian $H = L_c^2$ (an order-4, bi-Laplacian-type operator). This choice is crucial for capturing short-time asymptotic behavior.

B. Spectral Observables

The paper analyzes the spectrum $\{\lambda_k\}$ of $H$ using three primary probes:

1. Heat Trace ($\Theta(t)$): $\Theta(t) = \text{Tr}(e^{-tH}) = \sum e^{-t\lambda_k}$.
2. Spectral Dimension ($d_s(t)$): Defined via the short-time scaling of the heat trace:
   $$d_s(t) = -2 \frac{d \log \Theta(t)}{d \log t}$$
   (Note: For an order-$m$ operator, the standard relation is $\Theta(t) \sim t^{-d_s/m}$).
3. Thermal Entropy ($S(t)$): The von Neumann entropy of the Gibbs state $\rho(t) = e^{-tH}/\Theta(t)$.
4. Eigenvalue Growth: Fitting the scaling law $\lambda_k \sim k^\alpha$.

3. Key Contributions

A. Analytic Derivation of $d_s = 1/2$

The paper's most significant theoretical contribution is a first-principles proof (Theorem 7.3) that does not rely on the Riemann Hypothesis.

• Continuum Limit: Under mild regularity assumptions on the divergence kernel, the unnormalized combinatorial Laplacian $L_c$ converges to a one-dimensional differential operator $c_2(-\partial_u^2)$ in the log-index coordinate $u = \log i$.
• Bi-Laplacian Limit: Consequently, the squared Hamiltonian $H = L_c^2$ converges to the one-dimensional bi-Laplacian $c_4(-\partial_u^2)^2$.
• Scaling Result: By Weyl's law for a 1D bi-Laplacian, the eigenvalue counting function scales as $N(\lambda) \sim \lambda^{1/4}$. Applying the Laplace-Tauberian theorem, the heat trace scales as $\Theta(t) \sim t^{-1/4}$.
• Dimension Calculation: Using the convention $d_s = -2 \frac{d \log \Theta}{d \log t}$, the result is:
  $$d_s = 2 \times (1/4) = 1/2$$
  This establishes an effective spectral dimension of 0.5 for the primes.

B. Spectral Rigidity and Compression

The paper demonstrates that this spectral compression is rigid and universal:

• It persists across logarithmic, entropic, and fractal-type divergence functions.
• It is distinct from Random Matrix Theory (GUE) ensembles. While GUE spectra show slower decay and higher effective dimensions, the prime-induced spectra exhibit "maximal spectral compression" with $d_s \approx 0.5$.
• The system behaves as a coherence-limited medium rather than an extended geometric space, indicating that arithmetic sparsity enforces a non-Euclidean geometry.

C. The Prime Coherence Profile (PCP)

The author identifies a universal shape for the spectral dimension function $d_s(t)$, termed the Prime Coherence Profile:

1. Activation: $d_s(t)$ rises from near zero at small scales.
2. Peak: It reaches a single broad maximum at intermediate scales.
3. Decay: It decays rapidly toward zero at large scales, never stabilizing at a constant Euclidean dimension.
   This morphology signifies that long-range coherence is inhibited by the multiplicative structure of the primes.

4. Key Results

• Spectral Dimension: The effective spectral dimension of the prime coherence ensemble is analytically proven to be $d_s = 1/2$ in the thermodynamic limit ($N \to \infty$) for the unnormalized generator.
• Eigenvalue Growth: Numerical fits confirm $\lambda_k \sim k^4$ (implying $\alpha \approx 4$), consistent with the analytic $d_s = 2/\alpha = 0.5$.
• Entropy Scaling: The thermal entropy scales as $S(t) \sim \log(1/t)^\beta$ with $\beta \approx 0.25$, which is consistent with the derived spectral dimension.
• Comparison with Zeta Zeros: The local statistics of the rescaled eigenvalues of the prime Hamiltonian approximate the Gaussian Unitary Ensemble (GUE) statistics of the Riemann zeta zeros, suggesting a shared universality class, though the global spectral dimension ($d_s=0.5$) is a distinct feature of the prime coherence model.
• Normalization Sensitivity: The normalized generator ($L = I-S$) yields a bounded spectrum where $d_s(t) \to 0$ as $t \to 0$, highlighting that the non-trivial $d_s=1/2$ result is specific to the unnormalized (combinatorial) formulation which captures the true thermodynamic limit.

5. Significance

• New Arithmetic Geometry: The paper proposes a novel framework where primes are viewed not as points in space, but as nodes in an emergent "arithmetic medium" with well-defined spectral properties.
• Independence from RH: The derivation of $d_s = 1/2$ is unconditional and does not rely on the Riemann Hypothesis, offering a new perspective on the structural constraints of the primes.
• Bridge Between Fields: It unifies tools from spectral geometry (heat kernels, Tauberian theorems), information geometry (divergences, entropy), and analytic number theory.
• Physical Interpretation: The result $d_s = 1/2$ implies "maximal spectral compression" and the absence of classical diffusion. This suggests that the multiplicative independence and sparsity of primes create a geometry where information propagation is severely restricted, fundamentally different from Euclidean or standard graph-based geometries.
• Future Directions: The framework opens avenues for classifying other arithmetic sets (e.g., square-free numbers) by their spectral dimensions and exploring the relationship between arithmetic rigidity and spectral organization in broader contexts.

In summary, Watson demonstrates that the primes possess an intrinsic, non-Euclidean geometry characterized by a spectral dimension of 1/2, driven by the combinatorial sparsity of the prime sequence and revealed through the spectral analysis of coherence Hamiltonians.