Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Testing modified gravity with 3x2pt analyses in galaxy mocks

Marc Alemany-Gotor, Cristian Viglione, Pablo Fosalba, Isaac Tutusaus

TL;DR

The paper benchmarks the impact of assuming General Relativity when analyzing 3x2pt statistics on high-fidelity full-sky mocks built from Hu–Sawicki $f(R)$ gravity, using twin GR and $f(R)$ simulations with identical initial conditions and SDSS-calibrated galaxy mocks. Through a CosmoSIS-based 3x2pt pipeline and conservative scale cuts, the authors show that fitting $f(R)$-driven data with a GR model yields substantial biases, with FoB and S8 shifting by up to ~12σ, persisting even when marginalising galaxy bias and including baryonic feedback. The results are robust across variations in parameter sets and priors, indicating that MG can masquerade as different ΛCDM parameters and that nuisance parameters cannot trivially absorb this mis-specification. The study concludes that ignoring MG can lead to severe, detectable biases for Stage-IV surveys, motivatingMG-aware analyses in upcoming cosmological inferences.

Abstract

Stage-IV surveys will enable unprecedented tests of gravity on cosmological scales. However, assuming General Relativity in the analysis of large-scale structure could introduce systematic biases if gravity deviates from GR at these scales. Modified gravity theories, such as the Hu-Sawicki formulation of $f(R)$ gravity, offer an alternative explanation for cosmic acceleration without invoking a cosmological constant, while remaining consistent with Solar System tests through screening mechanisms. In this work, we quantify the cosmological parameter biases that arise when using a combination of galaxy clustering and weak-lensing data-vectors, the so-called 3x2pt analysis, from an $f(R)$ galaxy mock under the incorrect assumption of GR, using for the first time high-fidelity full-sky galaxy mock catalogues. We employ a pair of twin simulations: one with GR and one with Hu--Sawicki $f(R)$ gravity with $|f_{R0}| = 10^{-5}$. The mocks are built using an HOD method to populate the dark matter haloes with galaxies, calibrated against SDSS observations at low redshift. Using conservative scale cuts to minimise modelling uncertainties, we perform 3x2pt analyses and infer cosmological parameters through nested sampling, validating our pipeline with the GR mock. Our results show that when analysing the $f(R)$ galaxy mock assuming GR, the recovered cosmological parameters are very significantly biased, even when considering conservative scale cuts: the Figure of Bias reaches $\sim12σ$ for both $\{Ω_{\rm m}, σ_8\}$ and $S_8$. These biases persist even when marginalising over the galaxy bias and baryonic feedback, demonstrating that nuisance parameters cannot absorb the effects of modified gravity. We conclude that incorrectly assuming GR in a universe governed by $f(R)$ gravity leads to severe and detectable biases in cosmological inference for Stage-IV surveys.

Testing modified gravity with 3x2pt analyses in galaxy mocks

TL;DR

The paper benchmarks the impact of assuming General Relativity when analyzing 3x2pt statistics on high-fidelity full-sky mocks built from Hu–Sawicki gravity, using twin GR and simulations with identical initial conditions and SDSS-calibrated galaxy mocks. Through a CosmoSIS-based 3x2pt pipeline and conservative scale cuts, the authors show that fitting -driven data with a GR model yields substantial biases, with FoB and S8 shifting by up to ~12σ, persisting even when marginalising galaxy bias and including baryonic feedback. The results are robust across variations in parameter sets and priors, indicating that MG can masquerade as different ΛCDM parameters and that nuisance parameters cannot trivially absorb this mis-specification. The study concludes that ignoring MG can lead to severe, detectable biases for Stage-IV surveys, motivatingMG-aware analyses in upcoming cosmological inferences.

Abstract

Stage-IV surveys will enable unprecedented tests of gravity on cosmological scales. However, assuming General Relativity in the analysis of large-scale structure could introduce systematic biases if gravity deviates from GR at these scales. Modified gravity theories, such as the Hu-Sawicki formulation of gravity, offer an alternative explanation for cosmic acceleration without invoking a cosmological constant, while remaining consistent with Solar System tests through screening mechanisms. In this work, we quantify the cosmological parameter biases that arise when using a combination of galaxy clustering and weak-lensing data-vectors, the so-called 3x2pt analysis, from an galaxy mock under the incorrect assumption of GR, using for the first time high-fidelity full-sky galaxy mock catalogues. We employ a pair of twin simulations: one with GR and one with Hu--Sawicki gravity with . The mocks are built using an HOD method to populate the dark matter haloes with galaxies, calibrated against SDSS observations at low redshift. Using conservative scale cuts to minimise modelling uncertainties, we perform 3x2pt analyses and infer cosmological parameters through nested sampling, validating our pipeline with the GR mock. Our results show that when analysing the galaxy mock assuming GR, the recovered cosmological parameters are very significantly biased, even when considering conservative scale cuts: the Figure of Bias reaches for both and . These biases persist even when marginalising over the galaxy bias and baryonic feedback, demonstrating that nuisance parameters cannot absorb the effects of modified gravity. We conclude that incorrectly assuming GR in a universe governed by gravity leads to severe and detectable biases in cosmological inference for Stage-IV surveys.

Paper Structure

This paper contains 19 sections, 23 equations, 12 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Halo catalogues
  4. Galaxy catalogues
  5. Framework and pipeline of analysis
  6. Tomographic bin selection
  7. Measured data vectors
  8. Predicted data vectors
  9. Scale cuts
  10. Covariance matrix estimation
  11. Inference methods
  12. Metrics
  13. Results
  14. Data vector comparison
  15. Parameter inference
  16. ...and 4 more sections

Figures (12)

  • Figure 1: Normalised true redshift distribution, $n(z)$, for the 6 tomographic bins. The left plot shows the distribution for the GR mock while the right plot shows the distribution for the F5 mock.
  • Figure 2: Analytic correlation matrices for the 3×2pt data vectors, computed with OneCovariance. (a) Gaussian-only terms. (b) Full covariance including Gaussian, super-sample covariance (SSC), and non-Gaussian (nG) contributions. The matrix is block-structured into GC, GGL, and WL observables, from left to right.
  • Figure 3: Ratio between the measured data vectors (superscript m) in the GR and F5 mocks and the analytic data vectors (superscript p) of GC generated with our model and the fiducial cosmology of the mock. The label $\text{z}_\text{l}$ refers to the redshift of the centre of the tomographic bin.
  • Figure 4: Ratio between the measured data vectors (superscript m) in the GR and F5 mocks and the analytic data vectors (superscript p) of GGL generated with our model and the fiducial cosmology of the mock. The labels $\text{z}_\text{s}$ and $\text{z}_\text{l}$ refer to the redshift of the centre of the tomographic bin for the source and lens galaxies, respectively.
  • Figure 5: Ratio between the measured data vectors (superscript m) in the GR and F5 mocks and the analytic data vectors (superscript p) of WL generated with our model and the fiducial cosmology of the mock. The labels $\text{z}_\text{s}$ and $\text{z}_\text{l}$ refer to the redshift of the centre of the tomographic bin for the source and lens galaxies, respectively.
  • ...and 7 more figures