Tracing the Cosmic Origins: Machine Learning Reconstruction of the Primordial Density Field from EoR Observations
6 affiliations
- 1Max-Planck-Insitut für Radioastronomie , Auf dem Hügel 69, D-53121 Bonn ,
- 2Van Swinderen Institute, University of Groningen , Nijenborgh 4, 9747 AG Groningen ,
- 3Northwestern University CIERA and Department of Physics and Astronomy, , 1800 Sherman Ave, Evanston, IL 60201 ,
- 4Jet Propulsion Laboratory, California Institute of Technology , 4800 Oak Grove Drive, Pasadena, CA 91109 ,
- 5California Institute of Technology , 1200 E. California Blvd., Pasadena, CA 91125 ,
- 6University of California Department of Astronomy and Astrophysics, , Santa Cruz, 1156 High Stpreet, Santa Cruz, CA 95064 ,
Abstract
Reconstructing the initial conditions of the Universe from late-time tracers would unlock cosmological information buried by non-linear structure formation and astrophysics. We reconstruct the initial density field corresponding to z = 300, with its linear amplitude extrapolated to z = 0, from simulated 21-cm and CO(1–0) line-intensity maps at z = 8 generated with the semi-numerical code, LIMFAST. Using a three-dimensional U-Net, we reconstruct the initial conditions and evaluate their impact on cosmological parameter constraints. The two tracers probe complementary environments: 21-cm emission traces neutral, low-density regions of the intergalactic medium, while CO traces overdense, star-forming regions. We additionally train and evaluate the reconstruction on maps incorporating SKA1-Low– and COMAP-ERA–like instrumental response and thermal noise. In the noiseless case, combining both tracers delivers the most accurate reconstruction across ionisation states, with the cross-correlation coefficient between the reconstructed and true initial density fields, |C(k)| ≳ 0.90 for k ≲ 0.75 Mpc−1. With observational effects, small-scale information is degraded, but the joint reconstruction retains substantial large-scale coherence, with |C(k)| ≳ 0.70 for k ≲ 0.3 Mpc−1. To quantify the cosmological information gain, we perform simulation-based inference using power-spectrum summaries pre- and post-reconstruction. Across an ensemble of realisations, reconstruction yields the largest improvements for σ8 and ns, with a median aggregate marginal compression of 1.17 bits in the noiseless case, corresponding to a 56% reduction in the factorised product of prior-normalised 68% marginal widths. With observational effects, the corresponding median compression is 1.29 bits. These results demonstrate that joint 21-cm–CO reconstruction can recover cosmological information beyond that captured by the pre-reconstruction summaries.
Versions
Related
- Probing Inflationary Origins of Primordial Black Holes with LIGO--Virgo--KAGRA O1--O4a data
- Tracing M22's origins: Spatial and chemical constraints on its formation history
- Tracing the Cosmic web across Cosmic time through SKA observations of radio galaxies
- Tracing M22’s Origins: Spatial and Chemical Constraints on Its Formation History
- Tracing the main belt origins of S-complex near-Earth objects: From space mission targets to meteorite sources
BibTeX
@article{saxena2026tracing,
title = {{Tracing the Cosmic Origins: Machine Learning Reconstruction of the Primordial Density Field from EoR Observations}},
author = {Anchal Saxena and P Daniel Meerburg and Guochao Sun and Tzu-Ching Chang and Lluís Mas-Ribas},
journal = {Monthly Notices of the Royal Astronomical Society},
year = {2026},
month = oct,
doi = {10.1093/mnras/stag1972},
url = {https://academic.oup.com/mnras/advance-article/doi/10.1093/mnras/stag1972/8931748},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Tracing the Cosmic Origins: Machine Learning Reconstruction of the Primordial Density Field from EoR Observations
AU - Anchal Saxena
AU - P Daniel Meerburg
AU - Guochao Sun
AU - Tzu-Ching Chang
AU - Lluís Mas-Ribas
PY - 2026
DA - 2026/10/16
JO - Monthly Notices of the Royal Astronomical Society
DO - 10.1093/mnras/stag1972
UR - https://academic.oup.com/mnras/advance-article/doi/10.1093/mnras/stag1972/8931748
AB - Reconstructing the initial conditions of the Universe from late-time tracers would unlock cosmological information buried by non-linear structure formation and astrophysics. We reconstruct the initial density field corresponding to z = 300, with its linear amplitude extrapolated to z = 0, from simulated 21-cm and CO(1–0) line-intensity maps at z = 8 generated with the semi-numerical code, LIMFAST. Using a three-dimensional U-Net, we reconstruct the initial conditions and evaluate their impact on cosmological parameter constraints. The two tracers probe complementary environments: 21-cm emission traces neutral, low-density regions of the intergalactic medium, while CO traces overdense, star-forming regions. We additionally train and evaluate the reconstruction on maps incorporating SKA1-Low– and COMAP-ERA–like instrumental response and thermal noise. In the noiseless case, combining both tracers delivers the most accurate reconstruction across ionisation states, with the cross-correlation coefficient between the reconstructed and true initial density fields, |C(k)| ≳ 0.90 for k ≲ 0.75 Mpc−1. With observational effects, small-scale information is degraded, but the joint reconstruction retains substantial large-scale coherence, with |C(k)| ≳ 0.70 for k ≲ 0.3 Mpc−1. To quantify the cosmological information gain, we perform simulation-based inference using power-spectrum summaries pre- and post-reconstruction. Across an ensemble of realisations, reconstruction yields the largest improvements for σ8 and ns, with a median aggregate marginal compression of 1.17 bits in the noiseless case, corresponding to a 56% reduction in the factorised product of prior-normalised 68% marginal widths. With observational effects, the corresponding median compression is 1.29 bits. These results demonstrate that joint 21-cm–CO reconstruction can recover cosmological information beyond that captured by the pre-reconstruction summaries.
SN - 0035-8711
SN - 1365-2966
ER -