Experiments / V2.358
V2.358
Dynamical Selection COMPLETE

V2.358 - DESI Year 1 w0-wa Confrontation

V2.358: DESI Year 1 w0-wa Confrontation

Question

DESI Year 1 (2024) reports ~3.9 sigma preference for dynamical dark energy (w0 = -0.45, wa = -1.79) over LCDM. This framework predicts w = -1 exactly. Is the framework already dead?

The Threat

DESI’s w0-wa fit suggests the dark energy equation of state evolves with time. The framework derives Omega_Lambda from entanglement entropy — a topological invariant that cannot evolve. If w != -1, the framework is falsified.

Results

Framework vs DESI BAO Data (13 measurements across 7 redshift bins)

ModelFree paramschi2chi2/NBIC
Framework (R=0.688)013.021.0013.02
Planck LCDM114.961.1517.52
LCDM best-fit18.510.6511.07
w0-wa (DESI fit)374.425.7282.11

The framework fits DESI BAO data better than Planck LCDM, with zero free parameters.

Key Finding: w0-wa Fails on BAO Alone

The w0-wa model with DESI’s best-fit parameters (w0 = -0.45, wa = -1.79) gets chi2 = 74 against DESI’s own BAO data. This is terrible. The 3.9 sigma preference for w0-wa comes from the combined BAO + CMB + SN fit, not from BAO data alone.

This means the w0-wa signal is driven by tension between datasets, not by a clean signal in any single dataset. This is a hallmark of either:

  • A real but subtle effect requiring multiple probes
  • Systematic tension between datasets mimicking dynamical DE

Individual BAO Measurements

Tracerz_effTypeFrameworkObservedPull
BGS0.295DV/rd7.907.93 +/- 0.15-0.2 sigma
LRG10.510DM/rd13.2513.62 +/- 0.25-1.5 sigma
LRG10.510DH/rd22.3322.33 +/- 0.58+0.0 sigma
LRG20.706DM/rd17.3717.86 +/- 0.33-1.5 sigma
LRG20.706DH/rd19.8219.33 +/- 0.53+0.9 sigma
LRG3+ELG10.930DM/rd21.5221.71 +/- 0.28-0.7 sigma
LRG3+ELG10.930DH/rd17.3217.88 +/- 0.35-1.6 sigma
ELG21.317DM/rd27.5327.79 +/- 0.69-0.4 sigma
ELG21.317DH/rd13.8713.82 +/- 0.42+0.1 sigma
QSO1.491DM/rd29.8330.69 +/- 0.80-1.1 sigma
QSO1.491DH/rd12.6313.23 +/- 0.47-1.3 sigma
Lya2.330DM/rd38.5039.71 +/- 0.94-1.3 sigma
Lya2.330DH/rd8.488.52 +/- 0.17-0.2 sigma

All pulls < 1.6 sigma. No single measurement is in significant tension. The largest pulls are in D_M measurements (comoving distance), which all pull in the same direction (framework predicts slightly smaller distances). This is consistent with the Eisenstein-Hu r_d approximation being ~2% too large.

Sound Horizon Sensitivity

The comparison is sensitive to r_d. Using Planck’s r_d = 147.09 Mpc instead of EH’s 150.0 Mpc:

r_d (Mpc)chi2chi2/N
147.097.870.61
148.007.090.55
149.008.790.68
150.00 (EH)13.021.00
151.0019.781.52

With proper r_d ~ 147-148 Mpc, the framework chi2 drops to ~7-8 (excellent for 13 data points, 0 free parameters).

DESI Best-Fit Omega_Lambda

DESI BAO data alone (within LCDM) prefers Omega_Lambda = 0.703 +/- 0.007. The framework’s R = 0.688 is 2.2 sigma from this — mild tension but not excluded. Planck’s 0.685 is at 2.6 sigma.

DESI Year 3/5 Forecast

ScenarioYear 3Year 5Framework
Statistical fluctuation~3.9 sigma~3.5 sigmaSURVIVES
Real dynamical DE~6.8 sigma~8.7 sigmaKILLED
Systematic error~3.9 sigma~3.9 sigmaSURVIVES

DESI Year 3 (expected ~2026) is the decisive test.

The Honest Assessment

What the framework gets right

  • chi2/N = 1.00 against DESI BAO data with zero free parameters
  • Beats Planck LCDM (chi2/N = 1.15) despite having one fewer parameter
  • BIC strongly favors framework over w0-wa (Delta BIC = -69)
  • All individual pulls < 1.6 sigma

What threatens the framework

  • DESI’s 3.9 sigma preference for w0-wa over LCDM (combined fit)
  • DESI best-fit Omega_Lambda = 0.703 is 2.2 sigma from framework’s 0.688
  • If w0-wa signal grows with more data, framework is falsified

Why the framework survives (for now)

  1. The w0-wa signal comes from inter-dataset tension, not BAO alone
  2. w0-wa gets chi2 = 74 against DESI’s own BAO data (terrible fit)
  3. Different SN datasets give different w0-wa significance (2-4 sigma)
  4. The EH r_d approximation inflates our chi2 by ~5 units
  5. DESI Year 1 is 1/3 of planned data — too early to conclude

Limitations

  1. Eisenstein-Hu approximation: r_d = 150 Mpc vs true ~147 Mpc adds ~2% systematic to all D/r_d predictions. This is the dominant source of chi2.
  2. No covariance matrix: DM and DH at the same redshift are correlated. We ignore this.
  3. H0 is an input: We use Planck’s H0 = 67.4. DESI’s own fit may prefer a different value.
  4. DESI data values: We use published central values. The full likelihood analysis with covariances would give a more accurate comparison.
  5. w0-wa is not the only alternative: More general DE models (e.g., w(z) bins) might fit better.

Files

  • src/desi_confrontation.py: BAO predictions, model comparison, w0-wa analysis
  • tests/test_desi.py: 11 tests, all passing
  • run_experiment.py: Full experiment driver with 8 analysis sections
  • results.json: Machine-readable results