Midnight exp_08: Cascade signal panel — two z-trend-immune signals (2/4)

After exp_07 showed z-trends confound single-quantity statistics, this
panel tests channels smooth z-trends CANNOT mimic.

Panel C PASS: Absorber pairs along the SAME sightline that straddle a
cascade transition have larger EW differences (0.485 vs 0.425, p~0).
Z-trend immune: both absorbers share the same line of sight.

Panel D PASS: In narrow z-windows (dz=0.06) around transitions vs
troughs, FWHM discrepancy is 0.128 vs 0.136 (p=0.0001). Doublet
lines lock tighter at transitions. Narrow windows are immune to
smooth z-evolution.

Panel A FAIL: No N-space periodicity at cascade frequency.
Panel B FAIL: Transitions don't stand out above random z-windows.

Surviving after full z-trend controls: sightline-straddling pairs
and narrow-window doublet coherence. Both are relationship metrics
that smooth evolution cannot produce.

Author: lornecodes

foundational/experiments/midnight/README.md

@@ -2,7 +2,7 @@
 
 *Night removes the noise. Midnight sees the structure.*
 
-**Status**: Active | **Score**: 20/28
+**Status**: Active | **Score**: 22/32
 **Created**: 2026-06-03
 **Confidence**: 0.3 (exploratory)
 **Journals**:
@@ -186,6 +186,7 @@ Timeline and priorities TBD.
 | 02 | PAC Tree Correlation Profile | Phase-Rate | 3/4 | T1 PASS (conservation exact to 10^{-17}). T2 FAIL: correlation decay base = 1.13, not phi (30% off). The level-mean correlation doesn't decay as phi^(-delta) — the averaging across all nodes at each level washes out the phi structure. **Key finding:** the PAC tree is deterministic, so correlations come from the noise, not from the conservation structure directly. T3 PASS: Fibonacci-index separations DO show enhanced correlation above trend (mean residual +0.003 vs -0.001). T4 PASS: phi-split is closest to phi-decay of all tested splits (29.6% vs 33-36% for others), and all others are >10% from phi. Phi is privileged but the effect is subtle. |
 | 03 | Photon Archaeology | Photon Arch. | 4/4 | T1 PASS: alpha = 5.7 ppm from CODATA, 0.1% perturbation of any component breaks it by 8-181x. F7 is least sensitive (8x) because it only appears in the near-unity correction term. phi is a unique fixed point (phi^2-phi-1=0 to machine precision). T2 PASS: A_8 line ratios match hydrogen to 2.5%, identical at all z by construction (discrete graph topology). T3 PASS: **line widths vary 220% across z=[0.1,3]**, perfectly correlated with cascade disequilibrium (Spearman rho=1.000). Widths peak near cascade transitions (z=0.1: N≈2, z=0.3: N≈4) and collapse between (z=3: N=6.46, diseq=0.07). T4 PASS: clean PAC/SEC separation — ratio spread = 0, width spread = 220%. |
 | 04 | Cascade Width Oscillation | Photon Arch. | 4/4 | T1 PASS: five cascade transition redshifts mapped — N=2 at z=0.101, N=3 at z=0.171, N=4 at z=0.302, N=5 at z=0.579, N=6 at z=1.416. Troughs at half-integers correctly interleaved. T2 PASS: **5 oscillation cycles** in 150-point W(z) curve, all 5 peaks aligned with transition redshifts. Width range 0.001 to 0.336. T3 PASS: oscillatory model fits **24x better** than monotonic power law (chi2 ratio 24.1). Standard broadening cannot produce this. T4 PASS: width contrast 398x at every transition. Observable in CaII K (z<1), MgII (z=0.3-2.5), CIV (z=1-3.5). **Prediction: quasar absorption lines at z=0.101, 0.171, 0.302, 0.579, 1.416 should be hundreds of times broader than at adjacent half-integer-N redshifts.** |
+| 08 | Cascade Signal Panel | Photon Arch. | 2/4 | **Z-trend-immune tests.** A FAIL: no periodicity at cascade frequency in N-space (peak at period 0.44, not 1.0). B FAIL: transition windows have higher EW spread than troughs but not above random controls. **C PASS: multi-absorber sightlines — pairs straddling cascade transitions show larger EW differences (0.485 vs 0.425, p≈0).** This is immune to z-trends: both absorbers are on the SAME sightline. **D PASS: narrow-window doublet shape — FWHM discrepancy 0.128 at transitions vs 0.136 at troughs (p=0.0001), doublet ratio also differs (p≈0).** Coherence signal survives narrow-window test. |
 | 07 | Deep Cascade Probe | Photon Arch. | 1/4 | **HONEST FAILURE.** T1 FAIL: FeII subset (53K) same direction (rho=0.237) but p=0.12 — not significant with smaller sample. T2 PASS: CIV Doppler b spread correlates at **p=0.004** and KS p=10^{-157}, but EW spread does NOT correlate (rho=-0.113). T3 FAIL: **Z-DETRENDING KILLS THE SIGNAL.** MgII EW std vs diseq goes from rho=0.288 (p=0.026) to rho=-0.021 (p=0.87) after removing quadratic z-trend. The exp_05 result was substantially confounded with smooth z-evolution. T4 FAIL: no multi-ion confirmation. **Key finding: exp_05's EW spread signal (p=0.007) was partially a z-trend artifact. The doublet coherence (exp_06, p=0.018) and CIV b-param (p=0.004) remain — these are harder to explain as z-trend.** |
 | 06 | Doublet Coherence (SDSS) | Photon Arch. | 3/4 | **REAL DATA: doublet coupling oscillates with cascade.** T1 PASS: FWHM discrepancy anticorrelates with disequilibrium (rho=-0.334, **p=0.018**) — doublet lines LOCK together at cascade transitions, diverge between. T2 FAIL: cascade rho at 97.2nd percentile of shuffled (needs 0.05 on median). T3 PASS: inter-line FWHM correlation tighter at transitions (r=0.540) than troughs (r=0.453). T4 PASS: **KS separation p=10^{-11} (FWHM discrepancy), p=10^{-7} (doublet ratio), p=10^{-4} (FWHM ratio).** Transition absorbers: discrepancy 0.123 vs trough 0.134 — doublet 9% more coherent at transitions. |
 | 05 | Bifractal Mesh Signal (SDSS) | Photon Arch. | 3/4 | **REAL DATA: SDSS DR16, 89,851 absorbers, zero tuning.** T1 PASS: EW spread correlates with cascade disequilibrium — EW IQR rho=0.343 **p=0.0073**, EW std rho=0.290 p=0.025. Wider distributions at cascade transitions. T2 FAIL: FWHM kurtosis correlation suggestive (rho=-0.234, p=0.07) but doesn't reach significance. T3 PASS: cascade correlation above 99.5th percentile of shuffled controls. Plain z-correlation is stronger (rho=0.813) — cascade signal is a modulation ON TOP of z-trend. T4 PASS: **KS test: transition vs trough populations distinguishable at p=10^{-42} (EW) and p=10^{-30} (FWHM).** Transition absorbers: EW median 1.478 vs trough 1.357; FWHM median 178.7 vs 168.8. |

foundational/experiments/midnight/results/exp_08_cascade_panel_20260608_091251.json

@@ -0,0 +1,51 @@
+{
+  "experiment": "exp_08_cascade_panel",
+  "initiative": "midnight",
+  "thread": "photon_archaeology",
+  "test_results": {
+    "A": {
+      "test": "panel_A_periodicity",
+      "peak_period": 0.4426505810343298,
+      "peak_power": 0.14227826346147662,
+      "near_cascade": false,
+      "shuffle_percentile": 91.4,
+      "PASS": false
+    },
+    "B": {
+      "test": "panel_B_sharp_excess",
+      "mean_trans": 0.6782623413644698,
+      "mean_trough": 0.6413584018985674,
+      "mean_control": 0.7012589275269846,
+      "trans_above_trough": true,
+      "trans_above_control": false,
+      "PASS": false
+    },
+    "C": {
+      "test": "panel_C_sightlines",
+      "n_straddling": 6437,
+      "n_non_straddling": 9000,
+      "median_straddle": 0.4848470382430742,
+      "median_non_straddle": 0.4252800164676564,
+      "mannwhitney_p": 1.8446982308274112e-09,
+      "ks_p": 1.6537785114010026e-10,
+      "PASS": true
+    },
+    "D": {
+      "test": "panel_D_doublet_shape",
+      "n_transition": 14020,
+      "n_trough": 10235,
+      "ks_doublet_ratio": 0.031723192563906544,
+      "p_doublet_ratio": 1.3171073421006342e-05,
+      "ks_fwhm_disc": 0.02932460223269573,
+      "p_fwhm_disc": 7.465079747313287e-05,
+      "dr_median_trans": 1.2199944422919509,
+      "dr_median_trough": 1.2265253682488946,
+      "disc_median_trans": 0.1277383854173625,
+      "disc_median_trough": 0.1360633146107863,
+      "PASS": true
+    }
+  },
+  "score": "2/4",
+  "n_pass": 2,
+  "n_total": 4
+}