V2.300: Heat Kernel Double-Counting Identity

Motivation

The coupling matrix M_l for each angular channel l determines three spectral invariants:

• α_l: entanglement area coefficient (from symplectic eigenvalues of reduced state)
• δ_l: entanglement log coefficient (same)
• ρ_l: vacuum energy density = (1/2) tr(M_l^{1/2})

The heat kernel K_l(t) = tr(exp(−t M_l)) encodes the full spectrum through its small-t expansion in Seeley-DeWitt (SD) coefficients {a_k}. If α, δ, and ρ are all expressible as functions of {a_k}, the double-counting identity is manifest: they share the same spectral origin.

This is a novel approach to the B.3 sub-experiment (heat kernel analysis) from the RESEARCH_GUIDE.

Method

1. For each angular channel l, compute eigenvalues {λ_k} of M_l (tridiagonal, N=300)
2. Compute heat kernel K_l(t) = Σ exp(−t λ_k) for t ∈ [10⁻³, 10⁻⁰·⁵]
3. Fit K_l(t) = a₀ t⁻¹/² + a₁ + a₂ t¹/² + a₃ t + … (Seeley-DeWitt expansion)
4. Compute ρ_l = (1/2) Σ √λ_k exactly
5. Attempt Mellin reconstruction: ρ from {a_k} via spectral zeta function
6. Attempt linear regression: ρ and α as functions of {a_k}
7. Test per-channel α_l/ρ_l relationship

Key Results

1. SD Expansion Fits Perfectly

l	a₀	a₁	a₂	ρ_l	R²
0	0.024	298.69	23.84	191.26	0.9999995
5	0.017	299.06	18.45	196.76	0.9999998
20	0.028	299.14	−12.34	224.41	0.9999995
40	0.023	299.23	−48.41	269.37	0.9999997

Six SD terms capture K_l(t) to R² > 0.99999 for all channels. The expansion converges well at small t.

Key observation: a₁ ≈ 299 ≈ N−1 for all l (this is the “bulk” contribution, insensitive to angular momentum). a₂ varies strongly with l and tracks ρ_l (corr = −0.97).

2. Mellin Reconstruction of ρ_vac FAILS

Using the formal relation tr(M^{−s}) ~ Σ a_k Γ(s−(k−1)/2) / Γ(s) at s = −1/2:

l	ρ_exact	ρ_Mellin (6 terms)	Error
0	191.3	808.3	+323%
5	196.8	809.6	+311%
20	224.4	790.6	+252%

The Mellin route diverges. The SD expansion is asymptotic (not convergent), and analytic continuation to s = −1/2 amplifies errors catastrophically. Adding more SD terms makes it WORSE (4 terms: +129%, 6 terms: +323%, 8 terms: +457%).

This is a genuine mathematical obstruction: the spectral zeta function cannot be reliably computed from a finite number of SD coefficients at the values needed for ρ_vac. The formal identity exists but is computationally useless.

3. Linear Regression: ρ from SD Coefficients

Despite Mellin failure, empirical regression works:

ρ_l = c₀·a₀ + c₁·a₁ + c₂·a₂ + c₃·a₃ + c₄·a₄ + c₅·a₅
R² = 0.997, max error = 3.2%

The dominant predictor is a₂ (correlation −0.97 with ρ). This makes physical sense: a₂ captures the “mass” or angular momentum contribution to the heat kernel, which also determines the mode frequencies and hence ρ_vac.

4. Per-Channel α Extraction is Unreliable

Per-channel α_l values from d²S_l/dn² at n=12−28 are:

• Order 10⁻⁶ (extremely small)
• Some negative (unphysical — sign instability)
• Noisy across l (no clean scaling)

The area law α is a collective property of ALL channels, not meaningfully decomposable per-channel at these lattice sizes. Individual S_l(n) grows too slowly for the n² coefficient to be reliably extracted.

5. No Clean α-ρ Relationship Per Channel

α_l ∝ ρ_l^{5.35}, R² = 0.37
ρ_l = 3.7×10⁶ · α_l + 240, R² = 0.59

Both fits are poor. The per-channel relationship between α and ρ is not a simple power law or linear function. This is partly because α_l extraction is unreliable (Finding 4), and partly because the relationship may genuinely be nonlocal (involving correlations between channels).

6. Correlation Structure

Quantity	Most correlated SD coeff	Correlation
ρ_l	a₂	−0.97
δ_l	a₂	−0.84
α_l	a₀	+0.82

ρ and δ share the same dominant SD coefficient (a₂), while α correlates most with a₀. This separation — α linked to a₀ (volume/area), δ and ρ linked to a₂ (curvature/mass) — is physically meaningful but doesn’t constitute an algebraic identity.

Physical Interpretation

What Works

• The SD expansion perfectly captures the spectral content of M_l
• ρ_vac is well-predicted by SD coefficients (R² = 0.997) through regression
• a₂ is the spectral invariant linking ρ and δ (both sensitive to angular momentum)

What Fails

• Formal Mellin reconstruction diverges: The SD expansion is asymptotic, making the spectral zeta approach to double-counting computationally inaccessible
• Per-channel α is not reliably extractable: The area coefficient is a collective (multi-channel) quantity
• No simple α-ρ identity per channel: The double-counting relationship, if it exists, is nonlocal across channels

Implication for Λ_bare = 0

The heat kernel approach does NOT provide the algebraic identity the RESEARCH_GUIDE calls for. The SD expansion encodes the spectrum but the map from SD coefficients to {α, δ, ρ} is:

• ρ → well-determined by a₂ (linear regression R² = 0.997)
• δ → correlated with a₂ (r = −0.84)
• α → poorly determined per-channel

The double-counting identity must be formulated at the total level (summed over channels), not per-channel. This is consistent with V2.295 (matrix locality) which showed α requires the full Williamson decomposition, not boundary-site approximation.

Honest Assessment

Achieved

• First systematic extraction of Seeley-DeWitt coefficients on the Srednicki lattice
• Demonstrated Mellin reconstruction failure (important negative result)
• Identified a₂ as the key spectral invariant linking ρ and δ
• Confirmed that ρ_l is well-predicted by SD coefficients (R² = 0.997)

Not Achieved

• No algebraic double-counting identity
• Per-channel α extraction too noisy for meaningful comparison
• The heat kernel approach does not straightforwardly prove Λ_bare = 0

Key Insight

The double-counting identity is a global (channel-summed) statement, not a per-channel one. The SD coefficients per channel encode ρ and δ cleanly, but α requires the full multimode entanglement structure. This suggests the proof must use the collective entanglement spectrum, not individual mode properties.