Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite 2026 Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite – F

Adam Andrews, Jens Jasche, Guilhem Lavaux, William Coulton, Francisco Villaescusa-Navarro, Marco Baldi, Drew Jamieson, Gabriel Jung, Dionysios Karagiannis, Florent Leclercq, Michele Liguori, Marco Marinucci, Benjamin D. Wandelt · 2026

Read Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite 2026 Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite – F by Adam Andrews, Jens Jasche, Guilhem Lavaux, William Coulton, Francisco Villaescusa-Navarro, Marco Baldi, Drew Jamieson, Gabriel Jung, Dionysios Karagiannis, Florent Leclercq, Michele Liguori, Marco Marinucci, Benjamin D. Wandelt (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-20855.txt.

Introduction

Andrews and co-authors show a field level method to measure local primordial non-Gaussianity, a clue to early universe physics. They infer this value, nuisance terms, and the starting conditions jointly from halo catalogs. The catalogs come from full dark matter simulations in the Quijote suite. Their Bayesian forward model uses the full data field, not just summary statistics. They report about a 1.3 times tighter error than power spectrum and bispectrum methods. Higher resolution improves the result further. The work prepares for surveys such as Euclid.

Catalog record

CREATORAdam Andrews, Jens Jasche, Guilhem Lavaux, William Coulton, Francisco Villaescusa-Navarro, Marco Baldi, Drew Jamieson, Gabriel Jung, Dionysios Karagiannis, Florent Leclercq, Michele Liguori, Marco Marinucci, Benjamin D. Wandelt
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73714
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.20855

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite 2026 Field-Level Inference of Primordial Non-Gaussianity with the Quijote Simulation Suite – F Adam Andrews, Jens Jasche, Guilhem Lavaux, William Coulton, Francisco Villaescusa-Navarro, Marco Baldi, Drew Jamieson, Gabriel Jung, Dionysios Karagiannis, Florent Leclercq, Michele Liguori, Marco Marinucci, Benjamin D. Wandelt Abstract Local primordial non-Gaussianity, parameterised as f NL local {f^{\rm{local}}_{\mathrm{NL}}} , will be stringently constrained using state-of-the-art methods applied to next-generation galaxy redshift survey data. In this paper, in preparation for the upcoming data sets, we demonstrate for the first time the joint field-level inference of f NL local {f^{\rm{local}}_{\mathrm{NL}}} , nuisance parameters, and the initial conditions in realistic halo catalogues, ones which are generated through full dark-matter-only N N -body simulations. The field-level inference algorithm optimally constrains f NL local {f^{\rm{local}}_{\mathrm{NL}}} through a Bayesian forward-modelling approach at the field level, which outperforms traditional methods by leveraging the full statistical power of the data at the scales considered. In addition, we assess its performance under various design choices in the forward model, including tests of the structure formation model and resolution. We demonstrate the robustness of our approach by applying it to a subset of the Quijote simulation suite, performing the inference at scales down to k max ≈ 0.1 ​ h ​ Mpc − 1 k_{\rm max}\approx 0.1h\;\text{Mpc}^{-1} . Compared with a power spectrum and bispectrum estimator, we find a ∼ 1.3 \sim 1.3 improvement in σ ⁡ ( f NL local ) \sigma({f^{\rm{local}}_{\mathrm{NL}}}) when applying BORG , while marginalising over the initial conditions and bias parameters. From the small-scale information sensitivity tests, we show that the constraints on f NL local {f^{\rm{local}}_{\mathrm{NL}}} improve as we increase the resolution of the inference. These findings underscore the transformative potential of field-level inference to leverage the information available in ongoing surveys such as Euclid , providing accurate insights into the physics of cosmic inflation and the number of fields driving it. 1 Introduction One of the major questions in modern cosmology is the physical mechanism that generates the initial conditions of the Universe ( 21 ; 30 ; 2 ) . These initial conditions set the foundation for the formation of the large-scale structure we observe today. However, the actual process, or even the dominant particle(s), remains unclear. The canonical mechanism for generating these initial conditions is cosmic inflation, an era of the early universe where an accelerated expansion occurred ( 164 ; 83 ) . In particular, vanilla models, such as the single-field slow-roll model, predict nearly Gaussian initial conditions ( 152 ; 75 ; 3 ; 119 ; 53 ; 33 ; 52 ; 13 ) . Consequently, detecting local Primordial Non-Gaussianity (PNG) would provide strong evidence for the involvement of multiple fields during cosmic inflation and would serve as compelling evidence for the era of cosmic inflation ( 45 ; 105 ; 6 ; 70 ; 43 ; 126 ; 69 ; 75 ; 104 ; 119 ; 21 ; 46 ; 30 ; 126 ; 137 ; 82 ) . Figure 1 : Comparison of constraints on f_{\mathrm{NL}}^{\mathrm{local}} from two different techniques in the Quijote simulation suite. The grey posteriors represent the Fisher constraints using a combination of power spectrum and bispectrum estimators. The blue posterior shows the constraints on f_{\mathrm{NL}} from a BORG analysis, with an uncertainty of \sigma(f_{\mathrm{NL}})=23 . The mean and the uncertainty of the posterior is estimated from the weighted mean and the total standard deviation of the runs performed. The vertical dashed line indicates the fiducial value f_{\mathrm{NL}}=0 . The results showcase the possible constraints on {f_{\mathrm{NL}}} at the given resolution of k_{\rm max}\approx 0.1h\;\text{Mpc}^{-1} when using a field-level inference method to the Quijote -PNG suite over a traditional estimator. The description of the computation of the forecasted constraints is provided in Section 2.4 . The most stringent constraint on local PNG, parameterised as {f^{\rm{local}}_{\mathrm{NL}}} (henceforth denoted as {f_{\mathrm{NL}}} ) currently comes from Cosmic Microwave Background (CMB) measurements by the Planck satellite, reporting {f_{\mathrm{NL}}}=-0.1\pm 5.0 at a confidence level of 68.3% ( 96 ; 141 ) . While the CMB remains a valuable cosmological probe of the early universe, its potential for new information on PNG is limited; large-scale temperature fluctuations are expected to have reached the cosmic-variance limit. Although small-scale polarisation data may yield marginal improvements ( 23 ; 64 ; 1 ) , it is unlikely to achieve the target of \sigma(f_{\mathrm{NL}})=1 ( 30 ; 126 ; 2 ) . In contrast, next-generation 3D galaxy surveys, which cover vast cosmic volumes and capture the largest scales of matter distribution, hold promise for providing new insights into {f_{\mathrm{NL}}} ( 30 ; 125 ; 2 ) . The next generation of galaxy redshift surveys are underway to significantly enhance our understanding of PNG with the aim of reducing the uncertainty in \sigma(f_{\mathrm{NL}}) to the order of unity ( 116 ; 62 ; 66 ; 67 ) . In the large-scale structure, PNG leaves characteristic imprints in various cosmological observables, with notable effects including scale-dependent bias ( 28 ; 175 ; 129 ; 101 ; 27 ; 77 ; 151 ; 32 ; 140 ; 181 ; 100 ) , small-scale signatures ( 13 ; 102 ; 74 ; 79 ; 80 ) , and higher-order statistical moments ( 42 ; 154 ; 107 ; 13 ; 118 ; 171 ; 30 ; 22 ) . These effects provide crucial information about the early universe and the processes that shaped the Universe in which we exist and observe. Current methods for constraining {f_{\mathrm{NL}}} rely on summary statistics to extract primordial information from galaxy clustering data ( 102 ; 41 ; 103 ; 130 ; 65 ; 34 ; 38 ) . These approaches have yielded encouraging results, with studies measuring {f_{\mathrm{NL}}} down to the level of \sigma(f_{\mathrm{NL}})=7.4 ( 44 ; 48 , DESI, p=1.6 1 1 1 See 72 for an analysis of DESI quasars with other values for p . for quasars,) , -6<{f_{\mathrm{NL}}}<20 ( 38 ; 37 , SDSS4/eBOSS,) and f_{\rm NL}=-3\pm 14 ( 168 ; 24 ; 149 ; 68 , Quaia,) using either a power spectrum estimator, or in conjunction with a bispectrum estimator. However, they reduce the vast amount of information available in galaxy fields to a limited set of statistical moments. This compression is not sensitive to higher-order correlations or non-Gaussian features, which are relevant to constrain {f_{\mathrm{NL}}} at the required precision. As a result, they ignore a substantial amount of information, especially in the era of high-precision data expected from upcoming surveys. Overcoming these limitations requires methods that exploit the full statistical content of the data; a wide variety of statistical methods have been developed to access information beyond 2-pt statistics, e.g., higher-order statistics or ( 102 ; 32 ; 128 ; 179 ; 121 ; 100 ) , persistent homology ( 27 ; 181 ; 39 ) , including machine learning approaches ( 77 ; 78 ; 106 ; 71 ; 140 ; 131 ; 47 ; 136 ; 36 , e.g.,) . A prominent example is field-level inference, which preserves all available information and allows for a more direct connection between theory and observation. Figure 2 : Illustration of the forward model used in

Read the complete text