Year: 2026
Type: preprint
Abstract: Cosmological hydrodynamical simulations connect observed galaxies to cosmology and the physics of galaxy formation. Each simulation is expensive, and different codes model galaxy formation in different ways. We present Celia (Catalog Embeddings Learned from cosmologIcal simulAtions), a self-supervised foundation model that maps catalogs of subhalos and their galaxy properties to an embedding vector for the purposes of compression and organization of astrophysical and large-scale structure information. Celia is pretrained without parameter labels on the IllustrisTNG, SIMBA and Astrid suites of CAMELS and their N-body counterparts, by reconstructing masked subhalo properties and predicting the representations of a teacher network. Simple linear models fitted to the embeddings recover $Ω_{\mathrm m}$ with $R^2=0.999$, and $σ_8$ and both supernova-feedback amplitudes with $R^2\geq0.97$, in all three suites. The embeddings translate well to out-of-distribution simulations such as Swift-EAGLE, a code that Celia never saw. A linear model fitted on the three training codes predicts its $Ω_{\mathrm m}$ with a correlation coefficient $r=0.985$, and linear models fitted on 20 labeled simulations reach $r\geq0.94$ for $Ω_{\mathrm m}$, $σ_8$ and a supernova-feedback amplitude. We decompose the variation among the simulations of each suite into principal components of the embedding, and simple power laws of the simulation parameters describe the leading components. Celia gives simulated galaxy catalogs a shared coordinate system for simulation-based inference, emulation, the comparison of galaxy-formation models, the detection of physical degeneracies and the discovery of new astrophysical probes.
Source: arXiv (Cornell University)
Institutions: Carnegie Mellon University, Institute for Advanced Study
Language: English
Subfield: Astronomy and Astrophysics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
Awards: G2689075568, INTENSITY MAPPING (IM) IS THE USE OF ONE OR MORE EMISSION LINES TO TRACE OUT THE STRUCTURE OF THE UNIVERSE WITHOUT NEEDING TO RESOLVE INDIVIDUAL OBJECTS (SUCH AS GALAXIES OR GAS CLOUDS). IT IS ONE OF THE MOST PROMISING WAYS TO RADICALLY EXTEND THE SKY SURVEY REVOLUTION IN COSMOLOGY. BY MAKING SPECTRA OF THE ENTIRE SKY RATHER THAN THE ONE PART IN ONE MILLION CAPTURED BY CURRENT FIBER SPECTROGRAPHS ONE WOULD BE SENSITIVE TO ALL STRUCTURE. THERE ARE POTENTIALLY HUGE DISCOVERIES TO BE MADE IN THE VAST MAJORITY OF THE SKY THAT IS CURRENTLY SPECTRALLY UNMAPPED AND ALSO GREAT GAINS IN SIGNAL TO NOISE OF COSMOLOGICAL CLUSTERING MEASUREMENTS. INTENSITY MAPPING WITH THE 21CM RADIO LINE HAS BEEN EXPLORED THEORETICALLY BY MANY AND INSTRUMENTS ARE BEING BUILT PARTICULARLY TARGETING THE EPOCH OF REIONIZATION. IN THE UV VISIBLE AND INFRARED HOWEVER OTHER LINES HAVE ENORMOUS PROMISE AND WILL BE EXPLOITED BY A RANGE OF FUTURE NASA MISSIONS INCLUDING WFIRST EUCLID AND THE PROPOSED SPHEREX INSTRUMENT A DEDICATED INTENSITY MAPPING SATELLITE.THE FIRST MEASUREMENT OF LARGE-SCALE STRUCTURE OUTSIDE THE RADIO (USING LYMAN-ALPHA EMISSION) WAS RECENTLY MADE BY THE PI AND COLLABORATORS. THE LY-A ABSORPTION LINE ALSO TRACES A CONTINUOUS COSMOLOGICAL FIELD THE LYMAN-ALPHA FOREST AND THE ENORMOUS RECENT INCREASE IN THE NUMBER OF OBSERVED QUASAR SPECTRA HAVE MADE IT POSSIBLE TO INTERPOLATE BETWEEN QUASAR SIGHTLINES TO CREATE THREE DIMENSIONAL MAPS. BEING ABLE TO TRACE THE SAME COSMIC STRUCTURE IN EMISSION AND ABSORPTION OFFERS HUGE ADVANTAGES WHEN WE SEEK TOUNDERSTAND THE PROCESSES INVOLVED. IT WILL HELP US MAKE COMPREHENSIVE MAPS OF THE UNIVERSE'S CONTENTS AND OFFER US THE OPPORTUNITY TO CREATE NEW POWERFUL COSMOLOGICAL TESTS.IN OUR PROPOSED WORK WE WILL EXPLORE THE POSSIBILITIES AFFORDED BY TAKING GRISM AND INTEGRAL FIELD SPECTRA OF LARGE VOLUMES OF THE UNIVERSE USING STATE-OF-THE-ART COSMOLOGICAL HYDRODYNAMIC SIMULATIONS. WE WILL MAKE USE OF ANALYSIS TECHNIQUES DEVELOPED FOR THE LYMAN-ALPHA FOREST AS WELL AS FOREST DATA ITSELF TO TEST THEM. OUR AIM IS TO DEVELOP INTENSITY MAPPING AS A COSMOLOGICAL TOOL AND SHOWHOW IT CAN BE USED TO ANSWER QUESTIONS ABOUT THE CONTENTS OF THE UNIVERSE AND THE FORMATION OF STRUCTURE THAT ARE NOT ACCESSIBLE TO TRADITIONAL TECHNIQUES. THE PROJECT WILL INVOLVE BOTH DIRECT SAMPLING OF COSMIC STRUCTURE AND CROSS-CORRELATIONS OF LINE INTENSITY AND OBJECTS (INCLUDING GALAXIES QUASARS AND ABSORPTION LINES). EMISSION (E.G. H-ALPHA EMISSION) AND ABSORPTION (LY ALPHA FOREST) WILLBE VIEWED AS CONTINUOUS FIELDS. USING LARGE VOLUME COSMOLOGICAL SIMULATIONS COMBINED WITH POPULATION SYNTHESIS TECHNIQUES WE WILL MAKE SIMULATED SPECTRAL DATA SETS. THE TECHNIQUES TO ANALYZE THESE COSMOLOGICAL DATA CUBES WILL BE DEVELOPED. THE EXPECTED OUTCOMESARE THE FOLLOWING:(A) PREDICTIONS FOR THE LARGE-SCALE STRUCTURE OF STRONG EMISSION LINES (INCLUDING HA HB LYA OII OIII) IN THE UNIVERSE USING HYDRODYNAMIC SIMULATIONS INCLUDING THE CONTRIBUTION FROM ALL COMPONENTS FROM QUASARS TO DIFFUSE EMSSION.(B) SIMULATIONS OF REALISTIC EXAMPLES OF THE USE OF IM AS A COSMOLOGICAL PROBE} INCLUDING BARYON OSCILLATIONS AND WEAK GRAVITATIONAL LENSING.(C) TESTS OF TECHNIQUES TO DETECTION AND QUANTIFY THE LOW SURFACE BRIGHTNESS UNIVERSE LEADING TO A COMPLETE CENSUS OF THE COSMIC INTENSITY IN SPECIFIC LINES SUCH AS OII AND HA.(D) DEVELOPMENT OF TECHNIQUES TO EXTRACT REDSHIFTS FOR INDIVIDUAL GALAXIES FROM LOW ANGULAR RESOLUTION IM SPECTROSCOPY.(E) MOCK CATALOGS FOR SPHEREX EUCLID AND WFIRST SPECTROSCOPY OF DIFFUSE EMISSION AS WELL AS FOR THE GALEX GRISM SURVEY AND TESTS OF ANALYSIS TECHNIQUES ON DATA FROM THE LATTER., G824294325, AI Institute: Planning: Physics of the Future
OpenAlex ID: W7224738713