ADE Spectral Ladder

The spectral determinant of the scalar Laplacian on each ALE space in the ADE series, compared against the E₇ target derived from the fine structure constant.

The ADE Classification of ALE Spaces

The finite subgroups of SU(2) are classified by ADE:

For each ALE space, the smooth crepant resolution (Kronheimer’s construction) carries a hyper-Kähler metric. The scalar Laplacian on this metric has a zeta-function-regulated spectral determinant. We compute the relative determinant: the ALE space versus (1/|Γ|) times flat C².

The E₇ Target OBSERVED

The E₇ theory derives the fractional part of 1/α from the self-consistent quadratic:

The fractional part is:

The conjecture is that this number appears as the E₇ ALE spectral determinant:

If this holds, the geometry of the binary octahedral resolution would directly encode α without any free parameters. The computation is the test.

The A₁ Result: Eguchi–Hanson VERIFIED

The A₁ ALE space is the Eguchi–Hanson manifold: the minimal resolution of C²/Z₂. The relative spectral determinant of the scalar Laplacian uses the Barnes double-zeta function:

The ingredients:

Evaluated at a = 1/2 and a = 1 using the Barnes G-function and Glaisher constant:

The target is −0.035999177. The gap:

A₁ is already within 2% of the conjectured E₇ value. The question is whether the remaining 0.000748 is closed by ascending the ADE ladder to E₇.

The A_n Series LADDER KILLED

For general A_n (Z_n+1), the relative spectral determinant is:

The heat-kernel coefficient zeta_rel(0) = n/12 grows linearly with n (each exceptional divisor contributes 1/12). The derivative Δ_An computed numerically:

Why A_n and E₇ Are Different Structures

The divergence is not surprising once you see the reason. The A_n groups are abelian (cyclic). The D and E groups are non-abelian. The spectral determinant formula changes structurally:

• A_n (abelian): contributions from each non-trivial element add independently. As n grows, more elements contribute, and the sum grows.
• D_n, E₆, E₇, E₈ (non-abelian): elements come in conjugacy classes; the character theory mixes contributions. The cancellations between representations can suppress the total.

For the binary octahedral group 2O (|2O| = 48), the elements fall into 8 conjugacy classes. The spectral determinant receives contributions weighted by the irreducible representations, not by a simple angular-sector sum. The A_n series is not the right path to E₇.

What Is Actually Open

D₄ computation (next achievable step)

The D₄ ALE space resolves C²/BD₈ where BD₈ is the binary dihedral group of order 8 (the quaternion group Q₈). It has 5 conjugacy classes: {e}, {−e}, {±i}, {±j}, {±k}. The spectral determinant requires:

1. Decompose the scalar field on C²/BD₈ into representations of BD₈.
2. Compute the angular zeta function on S³/BD₈ using characters of BD₈.
3. Evaluate the double zeta integral for each conjugacy class.

The group BD₈ acts on C² with two non-trivial types of elements (order 4 and order 2), requiring two distinct angular integrals versus the single integral in the A₁ case. Doable in principle; not yet computed.

E₇ computation (the target)

The binary octahedral group 2O of order 48 has 8 conjugacy classes. The Kronheimer hyper-Kähler metric on the E₇ ALE space is not explicitly known in closed form (it exists by construction, not by formula). Two methods are available:

• Scattering approach: compute the spectral determinant from the asymptotic boundary data and the Jost function of the resolved metric, using the McKay correspondence to organize the angular channels.
• Character sum approach: use the known character table of 2O to compute the double zeta sum with weights from each conjugacy class. Requires the action of 2O on C² written explicitly.

Either method would give a definite number. If that number equals −0.035999177 = −(1/α − 137), the conjecture survives. If not, it is killed.

The Striking Feature of A₁

Despite the A_n ladder being killed, A₁ itself is notable. The Eguchi–Hanson spectral determinant −0.036747 lands within 2% of the target −0.035999 without any tuning. The exact A₁ value is:

This is a closed-form expression in standard transcendental constants. It is not 1/137, but it is close enough to suggest the E₇ target is in the right neighborhood. Whether the additional group structure of 2O versus Z₂ produces exactly the 0.000748 correction is the question the computation would answer.

2O Computation — Cycle 9 (May 2026) IN PROGRESS

A dedicated computational project (2O_Spectral_Determinant_Project) is running the E₇/2O case directly through 9 cycles. Full attack history below.

Step 1: Naive Molien + F(l) diverges

The direct approach — compute n_l(2O) via the Molien series, subtract the free-field baseline, weight by F(l) = −2ζ_H'(0, l+1) — was run to high cutoff:

Growth consistent with δ_l ~ −l²/48 leading to an L³ log L divergence. The naive method fails — higher-order polynomial tails must be removed analytically before the finite oscillatory part is accessible.

Step 2: Three-term decomposition (Cycle 9 architecture)

Cycle 9 replaced the earlier monolithic solver with a clean three-term decomposition:

• Resolution — first-principles G/h = 48/18 = 2.667 multiplier. Old normalized version KILLED (collapsed to tautology hiding the true mass). New: −0.097992.
• Oscillatory — per-class Dirichlet eta regularization of the −I class. Net: ~7 × 10⁻⁶ (negligible after separation). Old proxy: KILLED.
• Orbifold — full 8-irrep McKay-weighted character projections. Old sin² proxy: KILLED. New: +0.007046.
• Cross-term — intrinsic McKay-adjacency inner product (osc-independent). Old osc-slaved version: KILLED. New: −0.01331.

Step 3: Cycle 9 results

Best 3-front combination (old resolution + new orbifold + new cross + new oscillatory):

Sign is now correct. Gap reduced by factor >2 versus previous best (+0.028). The new first-principles resolution (G/h mass) was the hidden dominant lever — the old normalization bug masked it across 8 prior cycles.

4-front (all new pieces, default coefficients): −0.10423 — expected overshoot from the large new resolution term. Joint retuning of orbifold coefficient and intrinsic coupling against the G/h resolution is the live math.

Step 4: Cycle 9.5 mass-consistent result (May 30 2026)

If the same G/h = 48/18 mass scaling is applied uniformly to all three geometric terms (resolution, orbifold, cross), the gap collapses by another order of magnitude:

Best result to date. The G/h uniform rescaling is the principled move — when you change the divisor-basis normalization from probability weights (divide by G) to affine Kac weights (divide by h), every linear-in-charge term must scale by the same factor for the relative physics to be preserved.

Step 5: Derivation attack on the surviving free parameters (May 30 2026)

Even with the mass-consistent gap of −0.000681, two O(1) coefficients still sat inside the v11 character-projection modules: orb_coeff = 0.75 and cross_intrinsic = 0.00018. Rounds 1–6 of a dedicated attack tried to derive these from the same McKay graph / exact E₇~ Cartan / representation-ring data using four natural conditions (determinant-style, variational, A₁ normalization, lowest-mode). None reproduced the empirical ratio; closest was off by ~2.9×.

Step 6: ADE subgroup consistency — the missing principle (May 30 2026)

Round 7–8 of the attack imposed a new global constraint: the relative strengths chosen for E₇/2O must reduce exactly to the known closed form on any A₁ sub-system (where no extra boost is required). Combined with mass-rescaling stationarity, this gives a hard fixed point:

Off the (uncorrected) empirical 0.19175 by 1.21×. The closest pure first-principles number across all 8 rounds.

Step 7: Character-table correction closes the gap (May 30 2026)

Independent audit then found that the v11 character table was lumping two distinct 2O conjugacy classes at θ = π/2 (face vs edge π/2 rotations are different conjugacy classes; v11 treated them as one). The character-weighted orbifold kernel:

The "empirical 0.19175" was computed with the lumped table, so it inherits the same systematic bias. Correcting:

The first-principles derivation and the corrected empirical agree to 7%. The "two surviving free coefficients" framing from Step 5 is superseded at the orbifold-coefficient level: ADE subgroup consistency is the missing principle there.

Step 8: Full mass-consistent three-term re-run with proper 8-class table (May 30 2026)

Plugging the proper 8-class orbifold kernel into the full mass-consistent three-term solver:

The kernel ratio alone (1/0.7716 ≈ 1.30×) is far short of what's actually needed. Recovering anything close to the previous gap requires 3.5–4× the old effective orbifold coefficient. This means the cross and orbifold pieces in the 7-class pipeline were not just individually slightly biased — they were structurally compensating each other's lumping errors. Correcting one without the other exposes a multiplicative artifact.

Honest takeaway: the orbifold-coefficient principle (ADE consistency → λ ≈ 0.23) survives in isolation but cannot rescue the total three-term match alone.

Step 9: The cross sector also broke (May 30 2026)

Audit of the cross-term's character-weighted kernel under the proper 8-class table:

The cross-orb relative geometric strength shifted by ~5× when the π/2 class is split — the 7-class lumping was distorting the cross's χ₂·sin² inner product even more than the orbifold sin² kernel. This rules out any single-knob retune.

Three consequences:

• Cross ansatz needs an 8-class rebuild. The "intrinsic McKay·χ₂·sin²" form inherited the same lumping bias; no boost on orb alone can compensate.
• The uniform G/h mass-consistent boost is wrong for the cross sector. The rescaling argument that justified a uniform G/h on all three terms used 7-class characters; with the proper table, the cross's response to basis rescaling is sector-specific.
• The three-term decomposition itself is partially an artifact. Orb and cross kernels both weaken under 8-class but at very different rates; their ratio is no longer what the v11 architecture assumed. The decomposition that worked on the lumped table may not be the natural one on the proper table.

Net position (Step 9): the 2O numerical pipeline is structurally open. Closing it requires rebuilding the cross sector on the proper 8-class table, deriving sector-specific scaling (not uniform G/h), and re-checking whether the res/orb/cross decomposition is even the right organizing structure under the corrected characters.

Step 10: Cross sector closes on the proper table with a sector-specific multiplier (May 30 2026)

Sweeping the cross multiplier on the proper 8-class raw cross geometric (orbifold and resolution held first-principles):

The proper 8-class three-term solver does reach the target with a sector-specific cross multiplier ~0.03 — almost two orders of magnitude smaller than the uniform G/h boost that the 7-class pipeline used. The mass-consistent rule was never uniform; it was masquerading as uniform under the lumped characters.

Higher-precision fitting gives a measured cross multiplier ~0.0308. Three structural candidates were tested:

None of the three candidates hit the measured value at precision-fit accuracy. 1/32 and 1/(2π²φ) both overshoot by ~1.5%; 1/33 undershoots by similar. The precision fit has not selected a structural form — it has actually pushed all three candidates out of the "clean match" zone.

Three honest possibilities going forward:

• The structural form is none of the three. 0.0308 wants a different first-principles expression — possibly involving α directly (4α = 0.0292, off by ~5%), Casimir factors specific to E₇, or a small correction to one of the candidates above.
• Residual systematic in the foundation. The orb and resolution sectors are first-principles in isolation, but their interaction under the proper 8-class table may carry a small residual that biases the cross multiplier ~1.5% low. Higher-precision foundation check needed.
• The structural form is 1/32 or 1/(2π²φ) plus a small genuine correction (one-loop renormalization, finite-volume effect). Both might still be the right leading order with a known sub-leading correction.

Honest takeaway: the cross-sector scaling rule is sector-specific and small (~0.03), confirmed across multiple fits. The full proper-table three-term solver does reach +0.0004 of the spectral-determinant target. But the structural identification of the cross multiplier is not yet pinned down at precision-fit accuracy — ~1.5% remains between the measurement and the nearest framework-natural candidates.

Honest Status

• A₁ value: VERIFIED. −0.036747 computed from Barnes double zeta, 50-digit precision.
• A_n series: COMPUTED. Diverges from target after n=1. Abelian ladder does not reach E₇.
• D₄: OPEN. Method known; computation not done.
• E₇ / 2O naive: COMPUTED. Diverges as L³ log L (confirmed).
• E₇ / 2O 3-front (Cycle 9): COMPUTED. −0.04372 (gap −0.00772).
• E₇ / 2O mass-consistent (Cycle 9.5): COMPUTED. −0.036680 (gap −0.000681). This number used the 7-class lumped character table and was materially helped by that approximation (see Step 8).
• ADE consistency derivation (Round 8): RUN. Hard A₁ reduction + mass stationarity gives λ = 0.2327 from first principles. PRINCIPLE STANDS
• Character-table audit (Step 7): RUN. v11 7-class table lumps two distinct π/2 conjugacy classes. Proper 8-class kernel is 0.772× the lumped value (23% systematic).
• Full re-run with proper 8-class table (Step 8): RUN. At the same coefficients that gave the −0.000681 result, the proper-table gap is −0.083 — two orders of magnitude worse. The previous best gap was substantially driven by the lumping.
• Conjecture Δ_E7 = −(1/α−137): sign, order, magnitude correct to within a factor of ~2 with the proper table. The ADE-consistency principle for the orbifold coefficient still holds. The full three-term proper-table fit is open.

The clean claim: not “E₇ gives α.”
The clean claim is: “A₁ gives −0.036747. Target is −0.035999. With the lumped 7-class table the 2O mass-consistent result reached −0.036680, but that match was a compound artifact of the lumping — matching it on the proper 8-class table requires ~3.5–4× the old orbifold coefficient. ADE consistency derives a coefficient near the corrected kernel ratio (~1.21−1.30×) but that's not enough to close the full gap. The 2O case is genuinely open; the principle direction is identified, the numerical match is not.”

Related pages: E₇ Uniqueness Theorem · Electroweak Scale · The α Fixed Point · Why Three Generations · Theory Index