Controlled Tension Forecasting: Quantifying Cross-Probe Biases in $\omega_0\omega_a$CDM

Problem Statement

Recent analyses combining DESI DR2 Baryon Acoustic Oscillations (BAO), Planck Cosmic Microwave Background (CMB) compressed distance priors, and Pantheon+ Supernovae (SNe) have reported mild deviations from the standard $\Lambda$CDM model, often interpreted as evidence for dynamical dark energy (DDE) within the Chevallier–Polarski–Linder (CPL) parametrization ($\omega_0, \omega_a$). However, a central challenge remains: determining whether these deviations reflect genuine dark energy evolution or arise from cross-probe inconsistencies, prior choices, or mismatches in likelihood construction. Previous work has shown that biased priors on $\Omega_{m0}$ or differing degeneracy geometries among probes can artificially displace inferred parameters from the $\Lambda$CDM expectation. This paper addresses the need for a transparent framework to quantify how specific cross-probe inconsistencies propagate into shifts in dark energy parameters ($\omega_0, \omega_a, \omega_p$) within a unified inference pipeline.

Methodology

The authors develop a controlled tension-injection framework using self-consistent mock datasets generated from a fiducial flat $\Lambda$CDM cosmology ($\Omega_{m0}=0.30, H_0=70, \omega_0=-1, \omega_a=0$). The methodology relies on the following components:

1. Mock Generation:

   • BAO: Distance ratios $D_M(z)/r_d$ and $D_H(z)/r_d$ at DESI effective redshifts, using DESI DR2 covariance matrices.
   • CMB: Compressed distance priors $\{R, \ell_A, \omega_b\}$ calibrated to Planck-2018 constraints.
   • SNe: Pantheon+-like Hubble diagram data with full statistical and systematic covariances.
   • Mocks are drawn from multivariate normal distributions to reproduce survey statistical properties.

2. Tension Injection Scheme:
   The framework introduces parameterized mismatches (tensions) into otherwise consistent mocks. These are implemented via three modes:

   • Data-level: Modifying theoretical predictions (e.g., shifting SNe zero-point $\Delta M$, BAO sound horizon $\Delta r_d$, or distance ratios).
   • Prior-level: Applying biased external priors (e.g., shifting $\Omega_{m0}$).
   • Hybrid: Combining data and prior offsets to simulate complex multi-probe conflicts.
     The primary tension vector considered is $\Delta t = (\Delta \Omega_{m0}, \Delta H_0, \Delta M, \Delta r_d)$.

3. Analysis Pipeline:

   • A unified Markov Chain Monte Carlo (MCMC) pipeline (using emcee) analyzes BAO, CMB, SNe, and their combinations.
   • The parameter space includes cosmological parameters $(\Omega_{m0}, H_0, \omega_0, \omega_a, r_d)$ and nuisance parameters ($M, \Delta M(z)$).
   • Likelihoods are assumed Gaussian in the compressed data vectors.
   • Fisher Matrix Analysis: Used to derive linear-response bias estimates ($\Delta \theta \approx F^{-1} \nabla \chi^2$) to complement MCMC results and map tension-to-bias relations.

4. Experimental Runs:
   Seven specific runs are conducted:

   • Run 1: Tension-free baseline (validates the pipeline).
   • Run 2: SNe $H_0$ tension via zero-point shift ($\Delta M \neq 0$).
   • Run 3: BAO $\Omega_{m0}$ tension via distance-ratio distortion.
   • Run 4: CMB acoustic-scale tension via low-$H_0$ distance prior.
   • Run 5: Opposite-directed BAO–CMB tensions.
   • Run 6: SNe anchor miscalibration (subtle zero-point shift).
   • Run 7: Combined multi-probe systematic tension (simultaneous BAO and CMB offsets).

Key Results

1. Baseline Consistency (Run 1):
   In the absence of injected tensions, the joint BAO+CMB+SNe analysis recovers the fiducial $\Lambda$CDM parameters within $1\sigma$, confirming the internal consistency of the mock datasets and the validity of the pipeline.

2. Probe-Specific Bias Signatures:

   • SNe Tension (Run 2): A SNe zero-point shift favoring higher $H_0$ ($\Delta M < 0$) forces the joint posterior to shift along the BAO+CMB degeneracy direction (lower $H_0$, higher $\Omega_{m0}$, $\omega_0 > -1, \omega_a > 0$). The BAO and CMB constraints dominate over the SNe preference, producing an apparent quintessence-like shift despite the tension originating in SNe.
   • BAO Tension (Run 3): A BAO preference for lower $\Omega_{m0}$ drives the joint posterior toward $\omega_0 < -1$ and $\omega_a > 0$ (phantom-like), illustrating how BAO-specific distortions can mimic DDE.
   • CMB Tension (Run 4): A CMB preference for lower $H_0$ induces a strong phantom-like shift ($\omega_0 < -1, \omega_a > 0$) and a large pivot redshift ($z_p \approx 0$), as the fit compensates for the acoustic-scale inconsistency via rapid late-time evolution.
   • Opposing Tensions (Run 5): When BAO and CMB tensions pull in nearly orthogonal directions, the joint posterior represents a compromise, resulting in a phantom-like shift that is milder than in Run 4 due to partial cancellation.
   • Combined Tensions (Run 7): Simultaneous BAO ($\Omega_{m0}$) and CMB ($H_0$) tensions produce the largest deviation from $\Lambda$CDM. The CPL parameters move deeply into the phantom regime ($\omega_0 \approx -1.2, \omega_a \approx 1.1$), and the pivot redshift shifts anomalously to $z_p \approx 1.6$, indicating a loss of geometric coherence where the pivot lies outside the constraining power of any single probe.

3. Tension–Bias Transfer Functions:

   • 1D Relations: The authors derive empirical linear mappings between injected tensions ($\Delta \Omega_{m0}, \Delta H_0, \Delta M$) and posterior shifts. These work well for small, isolated tensions.
   • 2D Maps: When multiple tensions act simultaneously, the response becomes highly nonlinear and anisotropic. The 2D maps reveal that $\Delta \Omega_{m0}$ and $\Delta H_0$ mix non-separably, creating curved ridges in the $(\omega_0, \omega_a)$ plane.
   • Limitations: The 1D linear transfer functions fail to predict the magnitude of shifts for large, realistic tensions (e.g., DESI DR2 + CMB) because they do not account for the strong curvature of the CMB acoustic-scale degeneracy and the nonlinear mixing of tension vectors.

Significance and Claims

The paper explicitly positions itself as a diagnostic and illustrative tool rather than a precision prediction engine for real data. Its primary contributions and claims are:

• Demonstration of Spurious DDE: The framework demonstrates that apparent dynamical dark energy signatures (shifts in $\omega_0, \omega_a$) can emerge purely from cross-probe inconsistencies in a mock universe generated from exact $\Lambda$CDM.
• Geometric Origin of Bias: The results highlight that the direction and magnitude of inferred biases are determined by the degeneracy geometry of the probes. The joint posterior often follows the "steepest" or most constraining degeneracy ridge (typically BAO+CMB) rather than the direction of the injected tension.
• Role of Parametrization: The authors emphasize that these results are specific to the CPL parametrization and the use of compressed CMB priors. The apparent DDE behavior is a projection effect within this specific basis; different parametrizations would yield different bias directions.
• Diagnostic Utility: The framework provides a reproducible method to stress-test inference pipelines. It allows researchers to identify which probe combinations and degeneracy directions are most vulnerable to tension-induced biases before drawing physical conclusions from observational anomalies.
• Modesty on Real Data: The authors caution that the specific transfer functions derived here are not parametrization-independent and should not be used as a direct substitute for full-likelihood analyses of current data. The work serves to clarify how inconsistencies propagate geometrically, rather than quantifying the exact magnitude of effects in specific observational datasets.

In summary, the paper argues that interpreting deviations from $\Lambda$CDM requires careful synchronization of absolute calibrations across probes, as even modest cross-probe mismatches can be absorbed by flexible late-time parametrizations, mimicking the signatures of dynamical dark energy.