[Paper Review] The Self-Consistency of DESI Analysis and Comment on "Does DESI 2024 Confirm $Λ$CDM?"

Deng Wang|arXiv (Cornell University)|Apr 22, 2024

Manufacturing Process and Optimization4 citations

TL;DR

This paper reanalyzes DESI's first-year BAO data using broader prior ranges for dark energy parameters, revealing a >2σ preference for quintessence-like dark energy and evidence of evolving dark energy beyond 2σ. When combined with CMB and SN data, the analysis shows a ~4σ evidence for dynamical dark energy, challenging the DESI collaboration's original constraints and highlighting limitations in their binning and prior choices.

ABSTRACT

We demonstrate that the constraints on the evolution of dark energy implemented by the DESI collaboration may be insufficient or incomplete using their own BAO data. Using large enough prior ranges for the present-day equation of state of dark energy $ω_0$ and amplitude of dark energy evolution $ω_a$, we obtain the complete $1\,σ$ and $2\,σ$ constraints $ω_0=1.04^{+0.91+2.00}_{-1.00-1.90}$ and $ω_a=-7.4^{+3.8+6.8}_{-3.2-7.3}$ indicating a beyond $2\,σ$ preference of quintessence-like dark energy today and an evidence of evolving dark energy at beyond $2\,σ$ CL, respectively. Our results are different from $ω_0=-0.55^{+0.39}_{-0.21}$ and the $2\,σ$ upper limit $ω_a<-1.32$ reported by the DESI collaboration \cite{DESI:2024mwx}. Employing a data combination of cosmic microwave background, DESI BAO and type Ia supernova, we obtain the $1\,σ$, $2\,σ$ and $3\,σ$ constraints $ω_0=-0.707^{+0.089+0.18+0.24}_{-0.089-0.17-0.22}$ and $ω_a=-1.09^{+0.38+0.67+0.82}_{-0.31-0.72-1.00}$, which reveals a $\sim4\,σ$ evidence of dynamical dark energy when the redshift $z\lesssim0.1$. We verify that the BAO data point from luminous red galaxies at the effective redshift $z_{ m eff}=0.51$ hardly affects the joint constraint from the data combination of cosmic microwave background, DESI BAO and type Ia supernova. We also point out the shortcomings and advantages of the binning method widely used in cosmological analyses.

Motivation & Objective

• To assess the self-consistency of the DESI collaboration's cosmological analysis using their own BAO data.
• To investigate whether the DESI-reported constraints on dark energy evolution are complete or potentially biased due to narrow prior ranges.
• To evaluate the impact of individual BAO data points—particularly the LRG1 point at z_eff=0.51—on joint constraints with CMB and SN data.
• To examine the reliability and limitations of the binning method commonly used in cosmological data analysis.
• To provide a more robust constraint on the CPL model parameters ω₀ and ωₐ using combined CMB, DESI BAO, and SN datasets.

Proposed method

• Re-analyzes DESI's BAO data using large prior ranges for ω₀ (present-day dark energy equation of state) and ωₐ (evolution amplitude), avoiding the narrow priors used in the original DESI analysis.
• Applies the Chevallier-Polarski-Linder (CPL) parameterization to model time-evolving dark energy in the Friedmann-Lemaître-Robertson-Walker metric.
• Combines DESI BAO data with Planck CMB and Pantheon+ Type Ia supernova data to constrain ω₀ and ωₐ in a joint likelihood analysis.
• Tests the robustness of results by splitting DESI BAO data into different redshift bins (e.g., BGS+LRG1+LRG2+LRG3+ELG1+ELG2 vs. QSO+Lyα) to assess binning effects.
• Compares constraints from the full data combination (CMB+DESI BAO+SN) with those excluding the LRG1 data point to isolate its influence.
• Evaluates model dependence by analyzing the behavior of dark energy parameters across redshift, particularly near z≈4 where phantom crossing is observed.

Experimental results

Research questions

• RQ1 Are the DESI collaboration's reported constraints on dark energy evolution self-consistent when using their own BAO data with broader prior ranges?
• RQ2 Does the inclusion of CMB and SN data significantly alter the evidence for dynamical dark energy compared to DESI's standalone BAO analysis?
• RQ3 What is the impact of the LRG1 BAO data point at z_eff=0.51 on the joint constraints of ω₀ and ωₐ?
• RQ4 How do different binning strategies affect the cosmological constraints, and what are the trade-offs between resolution and bias?
• RQ5 Is there a significant preference for quintessence-like dark energy (ω₀ > -1) or phantom-like behavior (ω₀ < -1) in the combined dataset?

Key findings

• Using large prior ranges, the analysis yields ω₀ = 1.04⁺⁰.⁹¹⁺².⁰⁰₋₁.⁰⁰₋₁.⁹⁰, indicating a >2σ preference for quintessence-like dark energy at late times.
• The constraint ωₐ = -7.4⁺³.⁸⁺⁶.⁸₋³.²₋⁷.³ shows evidence of evolving dark energy at >2σ CL, contradicting the DESI collaboration’s reported 2σ upper limit of ωₐ < -1.32.
• In the combined CMB+DESI BAO+Pantheon+ SN analysis, the 1σ, 2σ, and 3σ constraints are ω₀ = -0.707⁺⁰.⁰⁸⁹⁺⁰.¹⁸⁺⁰.²⁴₋₀.⁰⁸⁹₋₀.¹⁷₋₀.²² and ωₐ = -1.09⁺⁰.³⁸⁺⁰.⁶⁷⁺⁰.⁸²₋₀.³¹₋₀.⁷²₋₁.⁰⁰, showing a ~4σ evidence for dynamical dark energy.
• The LRG1 BAO data point at z_eff=0.51 has minimal impact on the joint CMB+BAO+SN constraints, as its exclusion does not significantly alter the results.
• The analysis reveals a significant phantom-crossing behavior, with dark energy becoming phantom at the 2σ CL around z≈4, indicating strong model dependence in the constraints.
• The binning method is shown to introduce bias when applied to small subsets; larger bins better approximate the full data distribution and reduce systematic errors.