V2.703 - Does the Framework Absorb DESI's Evolving Dark Energy Signal?
V2.703: Does the Framework Absorb DESI’s Evolving Dark Energy Signal?
Status: COMPLETE — 13/13 tests passed
Question
DESI Y1 (2024) found ~3-5σ evidence for evolving dark energy (w₀,wₐ ≠ −1,0). The framework predicts w = −1 exactly. Does the framework’s Ω_Λ = 0.6877 absorb any of DESI’s w₀wₐ signal? Or is the framework threatened by DESI?
Method
- Compute BAO distances (D_H/r_d, D_M/r_d, D_V/r_d) at all 7 DESI redshift bins (13 data points)
- Compare three models: Planck LCDM (1 param), Framework LCDM (0 params), w₀wₐ CDM (3 params)
- χ² comparison, AIC/BIC model selection, residual pattern analysis
- Ω_Λ scan to find BAO-preferred value
- Forecast future discrimination power (Euclid, DESI Y5, CMB-S4)
Key Results
1. w₀wₐ Does NOT Beat LCDM on BAO Alone
| Model | χ² (13 pts) | χ²/dof | Δχ² vs Planck |
|---|---|---|---|
| Planck LCDM (1 param) | 20.5 | 1.71 | reference |
| Framework LCDM (0 params) | 22.1 | 1.70 | +1.7 |
| w₀wₐ PantheonPlus (3 params) | 59.9 | 6.0 | +39.5 |
| w₀wₐ Union3 (3 params) | 58.3 | 5.8 | +37.9 |
| w₀wₐ DESY5 (3 params) | 98.6 | 9.9 | +78.1 |
The w₀wₐ best-fit values (optimized for the JOINT CMB+BAO+SNe combination) perform terribly on BAO data alone. DESI’s “evidence for evolving dark energy” is driven entirely by the combination with CMB and SNe, not by the BAO measurements themselves.
2. Framework vs Planck on BAO
The framework (χ² = 22.1) is slightly worse than Planck (χ² = 20.5) on BAO, by Δχ² = 1.7. But with zero free parameters vs one, AIC and BIC both prefer the framework:
- ΔAIC = −0.3 (framework preferred)
- ΔBIC = −0.9 (framework preferred)
3. BAO-Preferred Ω_Λ
| Parameter | BAO best-fit | Framework | Planck |
|---|---|---|---|
| Ω_Λ | 0.675 | 0.688 | 0.685 |
BAO alone prefers Ω_Λ = 0.675 (Ω_m = 0.325), which is below both Planck and the framework. The framework is 1.3σ from BAO best-fit; Planck is 1.0σ. The 1σ range is [0.669, 0.681], excluding both at ~1σ.
4. Framework Closer Than w₀wₐ in 10/13 Bins
The framework’s LCDM distances are closer to DESI observations than the w₀wₐ predictions in 10 out of 13 individual measurements. The residual improvement correlation between framework and w₀wₐ is 0.91, showing they pull in the same direction (both prefer lower Ω_m than Planck).
5. Future Discrimination
| Survey | σ(Ω_Λ) | Framework vs Planck |
|---|---|---|
| DESI Y3 (2026) | 0.005 | 0.6σ |
| DESI Y5 (2028) | 0.003 | 1.0σ |
| Euclid (2027) | 0.002 | 1.5σ |
| CMB-S4 (2030) | 0.0015 | 2.0σ |
Interpretation
What’s Strong
- The framework survives DESI. The w₀wₐ “evidence” is not from BAO — it’s from the joint CMB+SNe combination. On BAO alone, w = −1 is perfectly adequate.
- Framework beats w₀wₐ on BAO by a large margin (AIC: −44 to −82), because w₀wₐ badly overfits to the joint dataset.
- The 0.91 correlation between framework and w₀wₐ improvements suggests the framework captures the same underlying signal (preference for lower Ω_m) that partially drives the w₀wₐ result.
What’s Weak
- BAO alone prefers lower Ω_Λ than the framework. BAO best-fit is 0.675, framework predicts 0.688. The framework is 1.3σ from BAO alone.
- Framework is worse than Planck on raw BAO χ². Planck’s fitted Ω_Λ = 0.685 is closer to the BAO preferred 0.675 than the framework’s 0.688.
- The w₀wₐ comparison is unfair. The w₀wₐ parameters were optimized for the JOINT fit, not BAO alone. A proper comparison would re-fit w₀wₐ to BAO only, which would likely give w₀ ≈ −1, wₐ ≈ 0 (i.e., LCDM).
- DESI’s actual w₀wₐ evidence comes from the subtle interplay of CMB (high-z) + BAO (mid-z) + SNe (low-z). Our BAO-only analysis cannot address this.
The Real Threat
The framework’s 0.688 is in mild tension with BAO’s preferred 0.675. If DESI Y3/Y5 tightens the BAO constraint and it remains at 0.675, the framework will face a ~2σ tension from BAO alone. Combined with the w₀wₐ evidence from joint fits, this could become significant.
However, BAO’s Ω_Λ preference is also in tension with Planck (0.685 vs 0.675), suggesting either a systematic or new physics in the BAO sector.
Conclusion
The framework’s prediction of constant dark energy (w = −1) is safe from DESI’s w₀wₐ signal, which arises from dataset combination rather than BAO measurements. But BAO data mildly prefers lower Ω_Λ than the framework predicts — a tension to watch as DESI accumulates more data.
Files
src/w0wa_analysis.py— w₀wₐ cosmology engine, DESI data, model comparisontests/test_w0wa.py— 13 tests, all passingrun_experiment.py— 10-phase analysisresults.json— Full numerical output