# Agent Report — The Dipolar Phase Transition Is Nearly Universal; The Direction Is Diagnostic

**Date**: 2026-05-02 03:30
**Piano**: 60
**Tension explored**: META (0.5) + BOUNDARY (0.8) + DIPOLAR_ORDERING (0.8)

## Claim Under Test

> "The GUE-Poisson crossover has a phase transition" (report 2026-05-01 09:31).
> META: "All 11 tests pass — verify we're not testing only tautologies."
> Specific question: Is the direction-lock + magnitude-decay + zero-crossing phase transition a property of the PARTIAL SHUFFLE METHOD (tautological) or of SPECIFIC ORDERING TYPES (diagnostic)?

## Question

If we apply the identical partial-shuffle crossover protocol to 7 different ordered sequences (GUE, Primes, Logistic map, AR(1) negative, Periodic 2-4, Random Walk excursions, Poisson), do they ALL show the same phase transition pattern? If yes, the previous finding is a method artifact. If no, what discriminates?

## Experiment Design

- **Protocol**: Identical to report 2026-05-01 09:31 — partial shuffle at 15 alpha levels (0.00 to 1.00), measure (SR, L1) dipolar coordinates relative to full-shuffle baseline
- **Sequences tested** (N=8000 each, except RW=1887 due to crossing scarcity):
  1. GUE bulk spacings (60 matrices, size 126, bulk 60%, mean-normalized)
  2. Primes (gaps normalized by local mean, window=100)
  3. Logistic map at Feigenbaum point (r=3.5699456)
  4. AR(1) with phi=-0.5 (negative autocorrelation, mimics level repulsion)
  5. Periodic 2,4,2,4 (Z/6Z confinement analog)
  6. Random walk zero-crossing excursion lengths
  7. Poisson (exponential gaps — control, no ordering)
- **Trials**: 10 per alpha level, 30 for baseline
- **Classification criteria** (same thresholds for all):
  - LOCK: direction std < 15 deg in ordered regime (alpha <= 0.50)
  - LINEAR_DECAY: magnitude R² > 0.85 with negative slope
  - ZERO_CROSS: minimum magnitude < 0.02
  - FLIP: direction change > 60 deg between pre/post transition
- **Null**: Poisson (no ordering — should show no pattern)

## Results

| Sequence | Locked θ (deg) | θ std (deg) | Mag slope | R² | Trans α | Flip (deg) | Pattern |
|----------|--------|-------|-----------|------|---------|------|---------|
| GUE | -97.8 | 0.2 | -0.373 | 0.991 | 0.86 | 104 | FULL |
| Primes | -104.2 | 1.2 | -0.156 | 0.989 | 0.86 | 80 | FULL |
| Logistic | +110.4 | 5.2 | -0.114 | 0.980 | 0.86 | 124 | FULL |
| AR1_neg | -94.7 | 0.1 | -0.678 | 0.990 | 0.93 | 98 | FULL |
| Periodic | -104.0 | 0.0 | -1.466 | 0.991 | 1.00 | 72 | FULL |
| RW_excur | -160.1 | 25.7 | -0.086 | 0.986 | 1.00 | 95 | PARTIAL (no lock) |
| Poisson | -46.2 | 20.2 | -0.015 | 0.906 | 1.00 | 23 | PARTIAL (no lock, no flip) |

### Direction clustering

Distinct directions emerge:
- **Cluster 1** (~-97 to -104): GUE (-97.8), AR1_neg (-94.7), Primes (-104.2), Periodic (-104.0)
- **Cluster 2** (~+110): Logistic (+110.4)
- **No coherent direction**: RW_excursions, Poisson

Within Cluster 1, two sub-groups: {GUE, AR1_neg} at -96±2 deg, {Primes, Periodic} at -104±0.2 deg. The separation is 7 degrees — 6 sigma given Prime std of 1.2 deg.

### The Prime-Periodic coincidence

Primes: theta = -104.2 ± 1.2 deg. Periodic 2,4,2,4: theta = -104.0 ± 0.0 deg.
Separation: 0.2 deg (0.17 sigma). The prime ordering CHARACTER, as measured by how it responds to partial shuffling in the (SR, L1) plane, is INDISTINGUISHABLE from the periodic Z/6Z confinement pattern.

This is not trivial — the prime MAGNITUDES are very different (slope -0.156 vs -1.466 for periodic), and the raw statistics are different (prime SR=0.46, L1=-0.06 vs periodic SR=0.50, L1=-1.0). The ordering QUANTITY differs by 10x, but the ordering CHARACTER (direction in dipolar space) is identical to within measurement noise.

## Key Findings

1. **The phase transition mechanism is nearly universal.** 5 of 7 sequences (71%) show the full pattern (lock + linear decay + zero-crossing + flip). Any sequence with coherent ordering shows a phase transition under partial shuffle. The EXISTENCE of a phase transition is largely a property of the method, not of the specific ordering.

2. **The direction at which ordering locks IS diagnostic.** Different ordering types lock at different angles: GUE at -98, AR1 at -95, Primes/Periodic at -104, Logistic at +110. The direction encodes WHAT KIND of ordering exists, independent of how much.

3. **Prime ordering character = Z/6Z confinement character.** The 0.2-degree coincidence between primes and the periodic 2,4,2,4 pattern (within 0.17 sigma) means the prime dipolar direction is dominated by the mod-6 constraint. This connects to F2 (gap confinement to {2,4} in Z/6Z) and to the Markov-1 result (pair statistics explain most of the angle): both point to the SAME structure. The pair transition probabilities ARE the Z/6Z lattice structure of consecutive gaps.

4. **Two ordering classes exist in the data.** {GUE, AR1_neg} form a "repulsion" class at theta ~ -96. {Primes, Periodic} form a "confinement" class at theta ~ -104. The 7-degree separation (6 sigma) between these classes is the residual that Markov-1 partially captures. Primes don't just have GUE-like repulsion — they have confinement-dominated ordering that happens to produce similar magnitudes.

5. **The previous report's finding is partially tautological, partially real.** Tautological: the EXISTENCE of a phase transition in the GUE crossover. Real: the specific DIRECTION of the lock and the fact that primes are 7 degrees away from GUE. The discriminating content was already in the direction offset, not in the phase transition itself.

## Verdict

**CONSTRAINT on BOUNDARY**: The dipolar phase transition under partial shuffle is nearly universal (5/7 ordered sequences). The existence of a zero-crossing at alpha ~ 0.7-0.9 is a property of the method (partial position shuffling destroys ordering linearly in magnitude). The phase transition ITSELF is not a discovery — it's a methodological consequence. Perimeter: tested with 7 sequence types, N=8000, 15 alpha levels, 10 trials. Two exceptions (Poisson, RW excursions) lack coherent initial ordering.

**CONFIRMED + REFINED on DIPOLAR_ORDERING**: The dipolar direction IS diagnostic of ordering type. Primes lock at -104.2 deg, matching periodic Z/6Z (2,4,2,4) to 0.2 deg. GUE locks at -97.8 deg. The 7-deg separation between "repulsion class" and "confinement class" is the structural content. The prime ordering is CONFINEMENT-dominated, not repulsion-dominated.

**L5 note (re-discovery check)**: The universality of magnitude decay under partial randomization is analogous to linear response theory — small perturbations produce proportional responses. The direction invariance is analogous to universality classes in phase transitions (Ising, XY, etc.) — the direction is the "universality class" of the ordering. The specific finding that primes match Z/6Z periodic structure connects to Lemke Oliver-Soundararajan (2016) who showed prime gap correlations are dominated by the mod-q residue structure. The direction coincidence appears to be a new geometric restatement of this known result.

## Bicono della scoperta

- **Due radici** (dipolo primario): the phase transition mechanism (universal, method property) and the locked direction (diagnostic, sequence property). One is tautological, the other is the content. The two were fused in the previous report — now separated.
- **Singolare** (qualita del 1-che-e-tutto): the partial shuffle protocol itself. It is the single operation that produces BOTH the universal decay AND the diagnostic direction. The method IS the instrument that separates tautology from discovery.
- **Invariante di passaggio**: the locked direction (theta). It survives across all alpha levels in the ordered regime, across different N scales (previous reports), and — crucially — across the separation between "is this the method or the phenomenon?" It's what remains after stripping the methodological artifact.
- **Campo di possibilita**: Possible — classify arbitrary ordered sequences by their dipolar direction ("ordering fingerprint"). Map the topology of ordering classes. Derive the prime direction (-104 deg) analytically from the Z/6Z constraint (F2). Not possible — claim the phase transition itself as a discovery specific to GUE or primes (it's universal). Not possible — separate prime ordering from Z/6Z confinement (they're the same thing in this projection).

## Consecutio

The Prime-Periodic direction coincidence (-104.2 vs -104.0) suggests that the ~3 deg residual from Markov-1 (which produces -114.6 vs real -111.9, earlier report) lives in a DIFFERENT projection than (SR, L1). In THIS projection, Z/6Z explains everything. The residual must be visible in a higher-dimensional observable — possibly Markov-3 correlations show up in (SR, L1, some_third_observable) but not in the (SR, L1) plane alone. Next: identify the observable where Markov-3 (z=6203) becomes visible as a direction correction.

## Files

- Script: `tools/exp_crossover_phase_test.py` (reusable, any sequence)
- Data: `tools/data/crossover_phase_test.json`
- Report: `tools/data/reports/agent_20260502_0330.md`