Search

In this study, we investigate the impact of covariance within uncertainties on the inference of cosmological and astrophysical parameters, specifically focusing on galaxy stellar mass functions derived from the CAMELS simulation suite. Utilizing both Fisher analysis and Implicit Likelihood Inference (ILI), we explore how different covariance structures, including simple toy models and physics-motivated uncertainties, affect posterior distributions and parameter variances. Our methodology utilizes forward modeling via emulators that are trained on CAMELS simulations to produce stellar mass functions based on input parameters, subsequently incorporating Gaussian noise as defined by covariance matrices. We examine both toy model covariance matrices and physically motivated covariance matrices derived from observational factors like the stellar Initial Mass Function (IMF) and photometric aperture size. Our results demonstrate that covariance terms significantly influence parameter inference, often leading to tighter constraints or revealing complex, multimodal posterior distributions. These findings underscore the necessity of accounting for covariance when interpreting astrophysical observations, especially in fields where accurate parameter estimation is critical for model validation and hypothesis testing., Comment: 19 pages, 6 figures, submitted to ApJ

The baryonic physics shaping galaxy formation and evolution are complex, spanning a vast range of scales and making them challenging to model. Cosmological simulations rely on subgrid models that produce significantly different predictions. Understanding how models of stellar and active galactic nuclei (AGN) feedback affect baryon behavior across different halo masses and redshifts is essential. Using the SIMBA and IllustrisTNG suites from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project, we explore the effect of parameters governing the subgrid implementation of stellar and AGN feedback. We find that while IllustrisTNG shows higher cumulative feedback energy across all halos, SIMBA demonstrates a greater spread of baryons, quantified by the closure radius and circumgalactic medium (CGM) gas fraction. This suggests that feedback in SIMBA couples more effectively to baryons and drives them more efficiently within the host halo. There is evidence that different feedback modes are highly interrelated in these subgrid models. Parameters controlling stellar feedback efficiency significantly impact AGN feedback, as seen in the suppression of black hole mass growth and delayed activation of AGN feedback to higher mass halos with increasing stellar feedback efficiency in both simulations. Additionally, AGN feedback efficiency parameters affect the CGM gas fraction at low halo masses in SIMBA, hinting at complex, non-linear interactions between AGN and SNe feedback modes. Overall, we demonstrate that stellar and AGN feedback are intimately interwoven, especially at low redshift, due to subgrid implementation, resulting in halo property effects that might initially seem counterintuitive., Comment: 28 pages, 8 figures, submitted to ApJ

We present a study on the inference of cosmological and astrophysical parameters using stacked galaxy cluster profiles. Utilizing the CAMELS-zoomGZ simulations, we explore how various cluster properties--such as X-ray surface brightness, gas density, temperature, metallicity, and Compton-y profiles--can be used to predict parameters within the 28-dimensional parameter space of the IllustrisTNG model. Through neural networks, we achieve a high correlation coefficient of 0.97 or above for all cosmological parameters, including $\Omega_{\rm m}$, $H_0$, and $\sigma_8$, and over 0.90 for the remaining astrophysical parameters, showcasing the effectiveness of these profiles for parameter inference. We investigate the impact of different radial cuts, with bins ranging from $0.1R_{200c}$ to $0.7R_{200c}$, to simulate current observational constraints. Additionally, we perform a noise sensitivity analysis, adding up to 40\% Gaussian noise (corresponding to signal-to-noise ratios as low as 2.5), revealing that key parameters such as $\Omega_{\rm m}$, $H_0$, and the IMF slope remain robust even under extreme noise conditions. We also compare the performance of full radial profiles against integrated quantities, finding that profiles generally lead to more accurate parameter inferences. Our results demonstrate that stacked galaxy cluster profiles contain crucial information on both astrophysical processes within groups and clusters and the underlying cosmology of the universe. This underscores their significance for interpreting the complex data expected from next-generation surveys and reveals, for the first time, their potential as a powerful tool for parameter inference., Comment: Submitted to ApJ

Traditional star formation subgrid models implemented in cosmological galaxy formation simulations, such as that of Springel & Hernquist (2003, hereafter SH03), employ adjustable parameters to satisfy constraints measured in the local Universe. In recent years, however, theory and spatially-resolved simulations of the turbulent, multiphase, star-forming ISM have begun to produce new first-principles models, which when fully developed can replace traditional subgrid prescriptions. This approach has advantages of being physically motivated and predictive rather than empirically tuned, and allowing for varying environmental conditions rather than being tied to local Universe conditions. As a prototype of this new approach, by combining calibrations from the TIGRESS numerical framework with the Pressure-Regulated Feedback-Modulated (PRFM) theory, simple formulae can be obtained for both the gas depletion time and an effective equation of state. Considering galaxies in TNG50, we compare the "native" simulation outputs with post-processed predictions from PRFM. At TNG50 resolution, the total midplane pressure is nearly equal to the total ISM weight, indicating that galaxies in TNG50 are close to satisfying vertical equilibrium. The measured gas scale height is also close to theoretical equilibrium predictions. The slopes of the effective equations of states are similar, but with effective velocity dispersion normalization from SH03 slightly larger than that from current TIGRESS simulations. Because of this and the decrease in PRFM feedback yield at high pressure, the PRFM model predicts shorter gas depletion times than the SH03 model at high densities and redshift. Our results represent a first step towards implementing new, numerically calibrated subgrid algorithms in cosmological galaxy formation simulations., Comment: 33 pages, 12 figures, Accepted for publication in ApJ. This is a Learning the Universe Publication. All codes and data used to produce this work can be found at the following $\href{https://github.com/sultan-hassan/tng50-post-processing-prfm}{GitHub \,Link.}$

We use the cosmological magneto-hydrodynamical simulation TNG50 to study the relationship between black hole feedback, the presence of stellar bars, and star formation quenching in Milky Way-like disc galaxies. Of our sample of 198 discs, about 63 per cent develop stellar bars that last until z=0. After the formation of their bars, the majority of these galaxies develop persistent 3-15 kpc wide holes in the centres of their gas discs. Tracking their evolution from z=4 to 0, we demonstrate that barred galaxies tend to form within dark matter haloes that become centrally disc dominated early on (and are thus unstable to bar formation) whereas unbarred galaxies do not; barred galaxies also host central black holes that grow more rapidly than those of unbarred galaxies. As a result, most barred galaxies eventually experience kinetic wind feedback that operates when the mass of the central supermassive black hole exceeds $M_{BH} > 10^8 M_{\odot}$. This feedback ejects gas from the central disc into the circumgalactic medium and rapidly quenches barred galaxies of their central star formation. If kinetic black hole feedback occurs in an unbarred disc it suppresses subsequent star formation and inhibits its growth, stabilising the disc against future bar formation. Consequently, most barred galaxies develop black hole-driven gas holes, though a gas hole alone does not guarantee the presence of a stellar bar. This subtle relationship between black hole feedback, cold gas disc morphology, and stellar bars may provide constraints on subgrid physics models for supermassive black hole feedback., Comment: 9 pages, 7 figures, submitted to MNRAS. Comments welcome

We introduce the DREAMS project, an innovative approach to understanding the astrophysical implications of alternative dark matter models and their effects on galaxy formation and evolution. The DREAMS project will ultimately comprise thousands of cosmological hydrodynamic simulations that simultaneously vary over dark matter physics, astrophysics, and cosmology in modeling a range of systems -- from galaxy clusters to ultra-faint satellites. Such extensive simulation suites can provide adequate training sets for machine-learning-based analyses. This paper introduces two new cosmological hydrodynamical suites of Warm Dark Matter, each comprised of 1024 simulations generated using the Arepo code. One suite consists of uniform-box simulations covering a $(25~h^{-1}~{\rm M}_\odot)^3$ volume, while the other consists of Milky Way zoom-ins with sufficient resolution to capture the properties of classical satellites. For each simulation, the Warm Dark Matter particle mass is varied along with the initial density field and several parameters controlling the strength of baryonic feedback within the IllustrisTNG model. We provide two examples, separately utilizing emulators and Convolutional Neural Networks, to demonstrate how such simulation suites can be used to disentangle the effects of dark matter and baryonic physics on galactic properties. The DREAMS project can be extended further to include different dark matter models, galaxy formation physics, and astrophysical targets. In this way, it will provide an unparalleled opportunity to characterize uncertainties on predictions for small-scale observables, leading to robust predictions for testing the particle physics nature of dark matter on these scales., Comment: 28 pages, 8 figures, DREAMS website: https://www.dreams-project.org

The presence of dense, neutral hydrogen clouds in the hot, diffuse intra-group and intra-cluster medium is an important clue to the physical processes controlling the survival of cold gas and sheds light on cosmological baryon flows in massive halos. Advances in numerical modeling and observational surveys means that theory and observational comparisons are now possible. In this paper, we use the high-resolution TNG50 cosmological simulation to study the HI distribution in seven halos with masses similar to the Fornax galaxy cluster. Adopting observational sensitivities similar to the MeerKAT Fornax Survey (MFS), an ongoing HI survey that will probe to column densities of $10^{18}$ cm$^{-2}$, we find that Fornax-like TNG50 halos have an extended distribution of neutral hydrogen clouds. Within one virial radius, we predict the MFS will observe a total HI covering fraction around $\sim$ 12\% (mean value) for 10 kpc pixels and 6\% for 2 kpc pixels. If we restrict this to gas more than 10 half-mass radii from galaxies, the mean values only decrease mildly, to 10\% (4\%) for 10 (2) kpc pixels (albeit with significant halo-to-halo spread). Although there are large amounts of HI outside of galaxies, the gas seems to be associated with satellites, judging both by the visual inspection of projections and by comparison of the line of sight velocities of galaxies and intracluster HI., Comment: 20 pages, 13 figures, Accepted for publication in ApJ

Galaxy formation models within cosmological hydrodynamical simulations contain numerous parameters with non-trivial influences over the resulting properties of simulated cosmic structures and galaxy populations. It is computationally challenging to sample these high dimensional parameter spaces with simulations, particularly for halos in the high-mass end of the mass function. In this work, we develop a novel sampling and reduced variance regression method, CARPoolGP, which leverages built-in correlations between samples in different locations of high dimensional parameter spaces to provide an efficient way to explore parameter space and generate low variance emulations of summary statistics. We use this method to extend the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) to include a set of 768 zoom-in simulations of halos in the mass range of $10^{13} - 10^{14.5} M_\odot\,h^{-1}$ that span a 28-dimensional parameter space in the IllustrisTNG model. With these simulations and the CARPoolGP emulation method, we explore parameter trends in the Compton $Y-M$, black hole mass-halo mass, and metallicity-mass relations, as well as thermodynamic profiles and quenched fractions of satellite galaxies. We use these emulations to provide a physical picture of the complex interplay between supernova and active galactic nuclei feedback. We then use emulations of the $Y-M$ relation of massive halos to perform Fisher forecasts on astrophysical parameters for future Sunyaev-Zeldovich observations and find a significant improvement in forecasted constraints. We publicly release both the simulation suite and CARPoolGP software package., Comment: The manuscript was submitted to arxiv after receiving and responding to comments from the first referee report

We analyze measurements of the thermal Sunyaev-Zeldovich (tSZ) effect arising in the circumgalactic medium (CGM) of $L^*$ galaxies, reported by Bregman et al. 2022 and Das et al. 2023. In our analysis we use the Faerman et al. 2017 and Faerman et al. 2020 CGM models, a new power-law model (PLM), and the TNG100 simulation. For a given $M_{\rm vir}$, our PLM has four parameters; the fraction, $f_{\rm hCGM}$, of the halo baryon mass in hot CGM gas, the ratio, $\phi_T$, of the actual gas temperature at the virial radius to the virial temperature, and the power-law indicies, $a_{P,{\rm th}}$ and $a_n$ for the thermal electron pressure and the hydrogen nucleon density. The B+22 Compton-$y$ profile implies steep electron pressure slopes ($a_{P,{\rm th}}\simeq 2$). For isothermal conditions the temperature is at least $1.1\times 10^6$ K, with a hot CGM gas mass of up to $3.5\times 10^{11}$ M$_\odot$ for a virial mass of $2.75\times 10^{12}$ M$_\odot$. However, if isothermal the gas must be expanding out of the halos. An isentropic equation of state is favored for which hydrostatic equilibrium is possible. The B+22 and D+23 results are consistent with each other and with recent (0.5-2 keV) CGM X-ray observations by Zhang et al. 2024 of Milky Way mass systems. For $M_{\rm vir}\simeq 3\times 10^{12}$ M$_\odot$, the scaled Compton pressure integrals, $E(z)^{-2/3}Y_{500}/M_{\rm vir,12}^{5/3}$, lie in the narrow range, $2.5\times 10^{-4}$ to $5.0\times 10^{-4}$ kpc$^2$, for all three sets of observations. TNG100 underpredicts the tSZ parameters by factors $\sim 0.5$ dex for the $L^*$ galaxies, suggesting that the feedback strengths and CGM gas losses are overestimated in the simulated halos at these mass scales., Comment: Accepted for publication in the Astrophysical Journal. 29 pages, 17 figures

In recent years, cosmological hydrodynamical simulations have proven their utility as key interpretative tools in the study of galaxy formation and evolution. In this work, we present a like-for-like comparison between the baryon cycle in three publicly available, leading cosmological simulation suites: EAGLE, IllustrisTNG, and SIMBA. While these simulations broadly agree in terms of their predictions for the stellar mass content and star formation rates of galaxies at $z\approx0$, they achieve this result for markedly different reasons. In EAGLE and SIMBA, we demonstrate that at low halo masses ($M_{\rm 200c}\lesssim 10^{11.5}\, M_{\odot}$), stellar feedback (SF)-driven outflows can reach far beyond the scale of the halo, extending up to $2-3\times R_{\rm 200c}$. In contrast, in TNG, SF-driven outflows, while stronger at the scale of the ISM, recycle within the CGM (within $R_{\rm 200c}$). We find that AGN-driven outflows in SIMBA are notably potent, reaching several times $R_{\rm 200c}$ even at halo masses up to $M_{\rm 200c}\approx10^{13.5}\, M_{\odot}$. In both TNG and EAGLE, AGN feedback can eject gas beyond $R_{\rm 200c}$ at this mass scale, but seldom beyond $2-3\times R_{\rm 200c}$. We find that the scale of feedback-driven outflows can be directly linked with the prevention of cosmological inflow, as well as the total baryon fraction of haloes within $R_{\rm 200c}$. This work lays the foundation to develop targeted observational tests that can discriminate between feedback scenarios, and inform sub-grid feedback models in the next generation of simulations., Comment: 20 pages, 8 figures. Accepted for publication in MNRAS 03/07/24

Galaxies play a key role in our endeavor to understand how structure formation proceeds in the Universe. For any precision study of cosmology or galaxy formation, there is a strong demand for huge sets of realistic mock galaxy catalogs, spanning cosmologically significant volumes. For such a daunting task, methods that can produce a direct mapping between dark matter halos from dark matter-only simulations and galaxies are strongly preferred, as producing mocks from full-fledged hydrodynamical simulations or semi-analytical models is too expensive. Here we present a Graph Neural Network-based model that is able to accurately predict key properties of galaxies such as stellar mass, $g-r$ color, star formation rate, gas mass, stellar metallicity, and gas metallicity, purely from dark matter properties extracted from halos along the full assembly history of the galaxies. Tests based on the TNG300 simulation of the IllustrisTNG project show that our model can recover the baryonic properties of galaxies to high accuracy, over a wide redshift range ($z = 0-5$), for all galaxies with stellar masses more massive than $10^9\,M_\odot$ and their progenitors, with strong improvements over the state-of-the-art methods. We further show that our method makes substantial strides toward providing an understanding of the implications of the IllustrisTNG galaxy formation model., Comment: 18 pages, 7 figures, 3 tables, 4 pages of appendices. Submitted to ApJ

Semi-analytic models (SAMs) systematically predict higher stellar-mass scatter at a given halo mass than hydrodynamical simulations and most empirical models. Our goal is to investigate the physical origin of this scatter by exploring modifications to the physics in the SAM Dark Sage. We design two black hole formation models that approximate results from the IllustrisTNG 300-1 hydrodynamical simulation. In the first model, we assign a fixed black hole mass of $10^{6}\, \mathrm{M}_{\odot}$ to every halo that reaches $10^{10.5}\, \mathrm{M}_{\odot}$. In the second model, we disregard any black hole growth as implemented in the standard Dark Sage model. Instead, we force all black hole masses to follow the median black hole mass-halo mass relation in IllustrisTNG 300-1 with a fixed scatter. We find that each model on its own does not significantly reduce the scatter in stellar mass. To do this, we replace the native Dark Sage AGN feedback model with a simple model where we turn off cooling for galaxies with black hole masses above $10^{8}\, \mathrm{M}_{\odot}$. With this additional modification, the SMBH seeding and fixed conditional distribution models find a significant reduction in the scatter in stellar mass at halo masses between $10^{11-14}\, \mathrm{M}_{\odot}$. These results suggest that AGN feedback in SAMs acts in a qualitatively different way than feedback implemented in cosmological simulations. Either or both may require substantial modification to match the empirically inferred scatter in the Stellar Mass Halo Mass Relation (SMHMR)., Comment: 21 pages, 16 figures

We quantify the cosmological spread of baryons relative to their initial neighboring dark matter distribution using thousands of state-of-the-art simulations from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project. We show that dark matter particles spread relative to their initial neighboring distribution owing to chaotic gravitational dynamics on spatial scales comparable to their host dark matter halo. In contrast, gas in hydrodynamic simulations spreads much further from the initial neighboring dark matter owing to feedback from supernovae (SNe) and Active Galactic Nuclei (AGN). We show that large-scale baryon spread is very sensitive to model implementation details, with the fiducial \textsc{SIMBA} model spreading $\sim$40\% of baryons $>$1\,Mpc away compared to $\sim$10\% for the IllustrisTNG and \textsc{ASTRID} models. Increasing the efficiency of AGN-driven outflows greatly increases baryon spread while increasing the strength of SNe-driven winds can decrease spreading due to non-linear coupling of stellar and AGN feedback. We compare total matter power spectra between hydrodynamic and paired $N$-body simulations and demonstrate that the baryonic spread metric broadly captures the global impact of feedback on matter clustering over variations of cosmological and astrophysical parameters, initial conditions, and galaxy formation models. Using symbolic regression, we find a function that reproduces the suppression of power by feedback as a function of wave number ($k$) and baryonic spread up to $k \sim 10\,h$\,Mpc$^{-1}$ while highlighting the challenge of developing models robust to variations in galaxy formation physics implementation., Comment: 17 pages, 15 figures

Using a large sample of galaxies taken from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project, a suite of hydrodynamic simulations varying both cosmological and astrophysical parameters, we train a normalizing flow (NF) to map the probability of various galaxy and halo properties conditioned on astrophysical and cosmological parameters. By leveraging the learnt conditional relationships we can explore a wide range of interesting questions, whilst enabling simple marginalisation over nuisance parameters. We demonstrate how the model can be used as a generative model for arbitrary values of our conditional parameters; we generate halo masses and matched galaxy properties, and produce realisations of the halo mass function as well as a number of galaxy scaling relations and distribution functions. The model represents a unique and flexible approach to modelling the galaxy-halo relationship., Comment: 8 pages, 2 figures, accepted for ICML 2023 Workshop on Machine Learning for Astrophysics

We present CAMELS-ASTRID, the third suite of hydrodynamical simulations in the Cosmology and Astrophysics with MachinE Learning (CAMELS) project, along with new simulation sets that extend the model parameter space based on the previous frameworks of CAMELS-TNG and CAMELS-SIMBA, to provide broader training sets and testing grounds for machine-learning algorithms designed for cosmological studies. CAMELS-ASTRID employs the galaxy formation model following the ASTRID simulation and contains 2,124 hydrodynamic simulation runs that vary 3 cosmological parameters ($\Omega_m$, $\sigma_8$, $\Omega_b$) and 4 parameters controlling stellar and AGN feedback. Compared to the existing TNG and SIMBA simulation suites in CAMELS, the fiducial model of ASTRID features the mildest AGN feedback and predicts the least baryonic effect on the matter power spectrum. The training set of ASTRID covers a broader variation in the galaxy populations and the baryonic impact on the matter power spectrum compared to its TNG and SIMBA counterparts, which can make machine-learning models trained on the ASTRID suite exhibit better extrapolation performance when tested on other hydrodynamic simulation sets. We also introduce extension simulation sets in CAMELS that widely explore 28 parameters in the TNG and SIMBA models, demonstrating the enormity of the overall galaxy formation model parameter space and the complex non-linear interplay between cosmology and astrophysical processes. With the new simulation suites, we show that building robust machine-learning models favors training and testing on the largest possible diversity of galaxy formation models. We also demonstrate that it is possible to train accurate neural networks to infer cosmological parameters using the high-dimensional TNG-SB28 simulation set.

We discover analytic equations that can infer the value of $\Omega_{\rm m}$ from the positions and velocity moduli of halo and galaxy catalogues. The equations are derived by combining a tailored graph neural network (GNN) architecture with symbolic regression. We first train the GNN on dark matter halos from Gadget N-body simulations to perform field-level likelihood-free inference, and show that our model can infer $\Omega_{\rm m}$ with $\sim6\%$ accuracy from halo catalogues of thousands of N-body simulations run with six different codes: Abacus, CUBEP$^3$M, Gadget, Enzo, PKDGrav3, and Ramses. By applying symbolic regression to the different parts comprising the GNN, we derive equations that can predict $\Omega_{\rm m}$ from halo catalogues of simulations run with all of the above codes with accuracies similar to those of the GNN. We show that by tuning a single free parameter, our equations can also infer the value of $\Omega_{\rm m}$ from galaxy catalogues of thousands of state-of-the-art hydrodynamic simulations of the CAMELS project, each with a different astrophysics model, run with five distinct codes that employ different subgrid physics: IllustrisTNG, SIMBA, Astrid, Magneticum, SWIFT-EAGLE. Furthermore, the equations also perform well when tested on galaxy catalogues from simulations covering a vast region in parameter space that samples variations in 5 cosmological and 23 astrophysical parameters. We speculate that the equations may reflect the existence of a fundamental physics relation between the phase-space distribution of generic tracers and $\Omega_{\rm m}$, one that is not affected by galaxy formation physics down to scales as small as $10~h^{-1}{\rm kpc}$., Comment: 32 pages, 13 figures, summary video: https://youtu.be/STZHvDHkVgo

We train graph neural networks to perform field-level likelihood-free inference using galaxy catalogs from state-of-the-art hydrodynamic simulations of the CAMELS project. Our models are rotational, translational, and permutation invariant and do not impose any cut on scale. From galaxy catalogs that only contain $3$D positions and radial velocities of $\sim 1, 000$ galaxies in tiny $(25~h^{-1}{\rm Mpc})^3$ volumes our models can infer the value of $\Omega_{\rm m}$ with approximately $12$ % precision. More importantly, by testing the models on galaxy catalogs from thousands of hydrodynamic simulations, each having a different efficiency of supernova and AGN feedback, run with five different codes and subgrid models - IllustrisTNG, SIMBA, Astrid, Magneticum, SWIFT-EAGLE -, we find that our models are robust to changes in astrophysics, subgrid physics, and subhalo/galaxy finder. Furthermore, we test our models on $1,024$ simulations that cover a vast region in parameter space - variations in $5$ cosmological and $23$ astrophysical parameters - finding that the model extrapolates really well. Our results indicate that the key to building a robust model is the use of both galaxy positions and velocities, suggesting that the network have likely learned an underlying physical relation that does not depend on galaxy formation and is valid on scales larger than $\sim10~h^{-1}{\rm kpc}$., Comment: 34 pages, 12 figures. For a video summarizing the results, see https://youtu.be/b59ep7cyPOs

Abundance matching studies have shown that the average relationship between galaxy radius and dark matter halo virial radius remains nearly constant over many orders of magnitude in halo mass, and over cosmic time since about $z=3$. In this work, we investigate the predicted relationship between galaxy radius $r_{e}$ and halo virial radius $R_{\rm h}$ in the numerical hydrodynamical simulations Illustris and IllustrisTNG from $z\sim 0$--3, and compare with the results from the abundance matching studies. We find that Illustris predicts much higher $r_e/R_{\rm h}$ values than the constraints obtained by abundance matching, at all redshifts, as well as a stronger dependence on halo mass. In contrast, IllustrisTNG shows very good agreement with the abundance matching constraints. In addition, at high redshift it predicts a strong dependence of $r_e/R_{\rm h}$ on halo mass on mass scales below those that are probed by existing observations. We present the predicted $r_e/R_{\rm h}$ relations from Illustris and IllustrisTNG for galaxies divided into star-forming and quiescent samples, and quantify the scatter in $r_e/R_{\rm h}$ for both simulations. Further, we investigate whether this scatter arises from the dispersion in halo spin parameter and find no significant correlation between $r_e/R_{\rm h}$ and halo spin. We investigate the paths in $r_e/R_{\rm h}$ traced by individual haloes over cosmic time, and find that most haloes oscillate around the median $r_e/R_{\rm h}$ relation over their formation history., Comment: 13 pages, 10 figures, accepted at MNRAS

Extracting information from the total matter power spectrum with the precision needed for upcoming cosmological surveys requires unraveling the complex effects of galaxy formation processes on the distribution of matter. We investigate the impact of baryonic physics on matter clustering at $z=0$ using a library of power spectra from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project, containing thousands of $(25\,h^{-1}{\rm Mpc})^3$ volume realizations with varying cosmology, initial random field, stellar and AGN feedback strength and sub-grid model implementation methods. We show that baryonic physics affects matter clustering on scales $k \gtrsim 0.4\,h\,\mathrm{Mpc}^{-1}$ and the magnitude of this effect is dependent on the details of the galaxy formation implementation and variations of cosmological and astrophysical parameters. Increasing AGN feedback strength decreases halo baryon fractions and yields stronger suppression of power relative to N-body simulations, while stronger stellar feedback often results in weaker effects by suppressing black hole growth and therefore the impact of AGN feedback. We find a broad correlation between mean baryon fraction of massive halos ($M_{\rm 200c} > 10^{13.5}$\,\Msun) and suppression of matter clustering but with significant scatter compared to previous work owing to wider exploration of feedback parameters and cosmic variance effects. We show that a random forest regressor trained on the baryon content and abundance of halos across the full mass range $10^{10} \leq M_\mathrm{halo}/$\Msun$< 10^{15}$ can predict the effect of galaxy formation on the matter power spectrum on scales $k = 1.0$--20.0\,$h\,\mathrm{Mpc}^{-1}$.

Feedback from active galactic nuclei and stellar processes changes the matter distribution on small scales, leading to significant systematic uncertainty in weak lensing constraints on cosmology. We investigate how the observable properties of group-scale halos can constrain feedback's impact on the matter distribution using Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS). Extending the results of previous work to smaller halo masses and higher wavenumber, $k$, we find that the baryon fraction in halos contains significant information about the impact of feedback on the matter power spectrum. We explore how the thermal Sunyaev Zel'dovich (tSZ) signal from group-scale halos contains similar information. Using recent Dark Energy Survey (DES) weak lensing and Atacama Cosmology Telescope (ACT) tSZ cross-correlation measurements and models trained on CAMELS, we obtain $10\%$ constraints on feedback effects on the power spectrum at $k \sim 5\, h/{\rm Mpc}$. We show that with future surveys, it will be possible to constrain baryonic effects on the power spectrum to $\mathcal{O}(<1\%)$ at $k = 1\, h/{\rm Mpc}$ and $\mathcal{O}(3\%)$ at $k = 5\, h/{\rm Mpc}$ using the methods that we introduce here. Finally, we investigate the impact of feedback on the matter bispectrum, finding that tSZ observables are highly informative in this case., Comment: 18 pages, 15 figures. Comments are welcome

In a novel approach employing implicit likelihood inference (ILI), also known as likelihood-free inference, we calibrate the parameters of cosmological hydrodynamic simulations against observations, which has previously been unfeasible due to the high computational cost of these simulations. For computational efficiency, we train neural networks as emulators on ~1000 cosmological simulations from the CAMELS project to estimate simulated observables, taking as input the cosmological and astrophysical parameters, and use these emulators as surrogates to the cosmological simulations. Using the cosmic star formation rate density (SFRD) and, separately, stellar mass functions (SMFs) at different redshifts, we perform ILI on selected cosmological and astrophysical parameters (Omega_m, sigma_8, stellar wind feedback, and kinetic black hole feedback) and obtain full 6-dimensional posterior distributions. In the performance test, the ILI from the emulated SFRD (SMFs) can recover the target observables with a relative error of 0.17% (0.4%). We find that degeneracies exist between the parameters inferred from the emulated SFRD, confirmed with new full cosmological simulations. We also find that the SMFs can break the degeneracy in the SFRD, which indicates that the SMFs provide complementary constraints for the parameters. Further, we find that the parameter combination inferred from an observationally-inferred SFRD reproduces the target observed SFRD very well, whereas, in the case of the SMFs, the inferred and observed SMFs show significant discrepancies that indicate potential limitations of the current galaxy formation modeling and calibration framework, and/or systematic differences and inconsistencies between observations of the stellar mass function., Comment: This is the revised version from the reviewer's report (submitted to ApJ)

Recent studies illustrate the correlation between the angular momenta of cosmic structures and their Lagrangian properties. However, only baryons are observable and it is unclear whether they reliably trace the cosmic angular momenta. We study the Lagrangian mass distribution, spin correlation, and predictability of dark matter, gas, and stellar components of galaxy-halo systems using IllustrisTNG, and show that the primordial segregations between components are typically small. Their protoshapes are also similar in terms of the statistics of moment of inertia tensors. Under the common gravitational potential they are expected to exert the same tidal torque and the strong spin correlations are not destroyed by the nonlinear evolution and complicated baryonic effects, as confirmed by the high-resolution hydrodynamic simulations. We further show that their late-time angular momenta traced by total gas, stars, or the central galaxies, can be reliably reconstructed by the initial perturbations. These results suggest that baryonic angular momenta can potentially be used in reconstructing the parameters and models related to the initial perturbations., Comment: 8 pages, 5 figures, 1 table. Matches the accepted version in ApJ

Active galactic nuclei (AGN) feedback models are generally calibrated to reproduce galaxy observables such as the stellar mass function and the bimodality in galaxy colors. We use variations of the AGN feedback implementations in the IllustrisTNG (TNG) and Simba cosmological hydrodynamic simulations to show that the low redshift Lyman-$\alpha$ forest can provide constraints on the impact of AGN feedback. We show that TNG over-predicts the number density of absorbers at column densities $N_{\rm HI} < 10^{14}$ cm$^{-2}$ compared to data from the Cosmic Origins Spectrograph (in agreement with previous work), and we demonstrate explicitly that its kinetic feedback mode, which is primarily responsible for galaxy quenching, has a negligible impact on the column density distribution (CDD) of absorbers. In contrast, we show that the fiducial Simba model which includes AGN jet feedback is the preferred fit to the observed CDD of the $z = 0.1$ Lyman-$\alpha$ forest across five orders of magnitude in column density. We show that the Simba results with jets produce a quantitatively better fit to the observational data than the Simba results without jets, even when the UVB is left as a free parameter. AGN jets in Simba are high speed, collimated, weakly-interacting with the interstellar medium (via brief hydrodynamic decoupling) and heated to the halo virial temperature. Collectively these properties result in stronger long-range impacts on the IGM when compared to TNG's kinetic feedback mode, which drives isotropic winds with lower velocities at the galactic radius. Our results suggest that the low redshift Lyman-$\alpha$ forest provides plausible evidence for long-range AGN jet feedback., Comment: 15 pages, 5 figures, 1 table, accepted to ApJL

In this work, we analyze a sample of $\sim$4000 massive ($M_*\geq 10^{11} M_\odot$ at $z=0$) galaxies in TNG300, the $(300 \mathrm{Mpc})^3$ box of the IllustrisTNG simulation suite. We characterize the shape and kinematics of these galaxies with a focus on the kinematic misalignment ($\Psi_\mathrm{int}$) between the angular momentum (AM) and morphological major axis. We find that the traditional purely shape- or kinematics-based classifications are insufficient to characterize the diversity of our sample and define a new set of classes based on the rates of change of the galaxies' morphological and kinematic axes. We show that these classes are mostly stable over time and correspond to six distinct populations of galaxies: the rapid AM reorienters (58% of our sample), unsettled galaxies (10%), spinning disks (10%), twirling cigars (16%), misaligned slow reorienters (3%), and regular prolate rotators (galaxies that display major axis rotation; 2%). We demonstrate that the most-recent significant (mass-ratio $\mu>1/10$) mergers of these galaxies are the primary cause for their present-day properties and find that these mergers are best characterized at the point of the satellite's final infall -- that is, much closer to the final coalescence than has been previously thought. We show that regular prolate rotators evolve from spinning disk progenitors that experience a radial merger along their internal AM direction. Finally, we argue that these regular prolate rotators are distinct from the similarly-sized population of rapid AM reorienters with large $\Psi_\mathrm{int}$, implying that a large $\Psi_\mathrm{int}$ is not a sufficient condition for major axis rotation., Comment: 26 pages, 17 figures; accepted for publication in ApJ

According to the standard model of cosmology, the arrangement of matter in the cosmos on scales much larger than galaxies is entirely specified by the initial conditions laid down during inflation. But zooming in by dozens of orders of magnitude to microscopic (and human?) scales, quantum randomness reigns, independent of the initial conditions. Where is the boundary of determinism, and how does that interplay with chaos? Here, we make a first attempt at answering this question in an astronomical context, including currently understood processes. The boundary is a function, at least, of length scale, position, and matter type (dark matter being more simply predictable). In intergalactic voids, the primordial pattern of density fluctuations is largely preserved. But we argue that within galaxies, the conditions are at minimum chaotic, and may even be influenced by non-primordial information, or randomness independent of the initial conditions. Randomness could be supplied by events such as supernovae and jets from active galactic nuclei (AGN) and other accretion disks, which, with the help of chaotic dynamics, could broadcast any possible microscopic randomness to larger scales, eventually throughout a galaxy. This may be generated or amplified by a recently investigated process called spontaneous stochasticity, or effective randomness in turbulent systems arising from arbitrarily small perturbations., Comment: Submitted to Philosophy of Astrophysics volume, ed. Boyd et al., Springer. 17pp, 6 figs

The driving of turbulence in galaxies is deeply connected with the physics of feedback, star formation, outflows, accretion, and radial transport in disks. The velocity dispersion of gas in galaxies therefore offers a promising observational window into these processes. However, the relative importance of each of these mechanisms remains controversial. In this work we revisit the possibility that turbulence on galactic scales is driven by the direct impact of accreting gaseous material on the disk. We measure this effect in a disk-like star-forming galaxy in IllustrisTNG, using the high-resolution cosmological magnetohydrodynamical simulation TNG50. We employ Lagrangian tracer particles with a high time cadence of only a few Myr to identify accretion and other events, such as star formation, outflows, and movement within the disk. The energies of particles as they arrive in the disk are measured by stacking the events in bins of time before and after the event. The average effect of each event is measured on the galaxy by fitting explicit models for the kinetic and turbulent energies as a function of time in the disk. These measurements are corroborated by measuring the cross-correlation of the turbulent energy in the different annuli of the disk with other time series, and searching for signals of causality, i.e. asymmetries in the cross-correlation across zero time lag. We find that accretion contributes to the large-scale turbulent kinetic energy even if it is not the dominant driver of turbulence in this $\sim 5 \times 10^{9} M_\odot$ stellar mass galaxy. Extrapolating this finding to a range of galaxy masses, we find that there are regimes where energy from direct accretion may dominate the turbulent energy budget, particularly in disk outskirts, galaxies less massive than the Milky Way, and at redshift $\sim 2$., Comment: Submitted to AAS Journals, comments welcome

As the next generation of large galaxy surveys come online, it is becoming increasingly important to develop and understand the machine learning tools that analyze big astronomical data. Neural networks are powerful and capable of probing deep patterns in data, but must be trained carefully on large and representative data sets. We developed and generated a new `hump' of the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project: CAMELS-SAM, encompassing one thousand dark-matter only simulations of (100 $h^{-1}$ cMpc)$^3$ with different cosmological parameters ($\Omega_m$ and $\sigma_8$) and run through the Santa Cruz semi-analytic model for galaxy formation over a broad range of astrophysical parameters. As a proof-of-concept for the power of this vast suite of simulated galaxies in a large volume and broad parameter space, we probe the power of simple clustering summary statistics to marginalize over astrophysics and constrain cosmology using neural networks. We use the two-point correlation function, count-in-cells, and the Void Probability Function, and probe non-linear and linear scales across $0.68<$ R $<27\ h^{-1}$ cMpc. Our cosmological constraints cluster around 3-8$\%$ error on $\Omega_{\text{M}}$ and $\sigma_8$, and we explore the effect of various galaxy selections, galaxy sampling, and choice of clustering statistics on these constraints. We additionally explore how these clustering statistics constrain and inform key stellar and galactic feedback parameters in the Santa Cruz SAM. CAMELS-SAM has been publicly released alongside the rest of CAMELS, and offers great potential to many applications of machine learning in astrophysics: https://camels-sam.readthedocs.io., Comment: 28 pages main text, additional 12 in appendices, 23 figures, 9 tables

We use the TNG100 simulation of the IllustrisTNG project to investigate the stellar specific angular momenta ($j_{\ast}$) of $\sim$12,000 central galaxies at $z=0$ in a full cosmological context, with stellar masses ($M_{\ast}$) ranging from $10^{9}$ to $10^{12} \, {\rm M}_{\odot}$. We find that the $j_{\ast}$-$M_{\ast}$ relations for early-type and late-type galaxies in IllustrisTNG are in good overall agreement with observations, and that these galaxy types typically `retain' $\sim$10-20 and $\sim$50-60 per cent of their host haloes' specific angular momenta, respectively, with some dependence on the methodology used to measure galaxy morphology. We present results for kinematic as well as visual-like morphological measurements of the simulated galaxies. Next, we explore the scatter in the $j_{\ast}$-$M_{\ast}$ relation with respect to the spin of the dark matter halo and the mass of the supermassive black hole (BH) at the galactic centre. We find that galaxies residing in faster spinning haloes, as well as those hosting less massive BHs, tend to have a higher specific angular momentum. We also find that, at fixed galaxy or halo mass, halo spin and BH mass are anticorrelated with each other, probably as a consequence of more efficient gas flow toward the galactic centre in slowly rotating systems. Finally, we show that halo spin plays an important role in determining galaxy sizes - larger discs form at the centres of faster-rotating haloes - although the trend breaks down for massive galaxies with $M_{\ast} \gtrsim 10^{11} \, {\rm M}_{\odot}$, roughly the mass scale at which a galaxy's stellar mass becomes dominated by accreted stars., Comment: 17 pages, 11 figures. Published in MNRAS

We analyze the velocity anisotropy of stars in real and energy space for a sample of Milky Way-like galaxies in the TNG50 simulation. We employ different selection criteria, including spatial, kinematic and metallicity cuts, and make three halo classes ($\mathcal{A}$-$\mathcal{C}$) which show mild-to-strong sensitivity to different selections. The above classes cover 48%, 16% and 36% of halos, respectively. We analyze the $\beta$ radial profiles and divide them into either monotonically increasing radial profiles or ones with peaks and troughs. We demonstrate that halos with monotonically increasing $\beta$ profiles are mostly from class $\mathcal{A}$, whilst those with peaks/troughs are part of classes $\mathcal{B}$-$\mathcal{C}$. This means that care must be taken as the observationally reported peaks/troughs might be a consequence of different selection criteria. We infer the anisotropy parameter $\beta$ energy space and compare that against the $\beta$ radial profile. It is seen that 65% of halos with very mild sensitivity to different selections in real space, are those for which the $\beta$ radial and energy profiles are closely related. Consequently, we propose that comparing the $\beta$ radial and energy profiles might be a novel way to examine the sensitivity to different selection criteria and thus examining the robustness of the anisotropy parameter in tracing stellar kinematics. We compare simulated $\beta$ radial profiles against various observations and demonstrate that, in most cases, the model diversity is comparable with the error bars from different observations, meaning that the TNG50 models are in good overall agreement with observations., Comment: 21 pages, 16 figures, Accepted for publication in ApJ

Uncertain feedback processes in galaxies affect the distribution of matter, currently limiting the power of weak lensing surveys. If we can identify cosmological statistics that are robust against these uncertainties, or constrain these effects by other means, then we can enhance the power of current and upcoming observations from weak lensing surveys such as DES, Euclid, the Rubin Observatory, and the Roman Space Telescope. In this work, we investigate the potential of the electron density auto-power spectrum as a robust probe of cosmology and baryonic feedback. We use a suite of (magneto-)hydrodynamic simulations from the CAMELS project and perform an idealized analysis to forecast statistical uncertainties on a limited set of cosmological and physically-motivated astrophysical parameters. We find that the electron number density auto-correlation, measurable through either kinematic Sunyaev-Zel'dovich observations or through Fast Radio Burst dispersion measures, provides tight constraints on $\Omega_{m}$ and the mean baryon fraction in intermediate-mass halos, $\bar{f}_{\mathrm{bar}}$. By obtaining an empirical measure for the associated systematic uncertainties, we find these constraints to be largely robust to differences in baryonic feedback models implemented in hydrodynamic simulations. We further discuss the main caveats associated with our analysis, and point out possible directions for future work., Comment: 31 pages, 10 figures, to be submitted to JCAP

Galaxies can be characterized by many internal properties such as stellar mass, gas metallicity, and star-formation rate. We quantify the amount of cosmological and astrophysical information that the internal properties of individual galaxies and their host dark matter halos contain. We train neural networks using hundreds of thousands of galaxies from 2,000 state-of-the-art hydrodynamic simulations with different cosmologies and astrophysical models of the CAMELS project to perform likelihood-free inference on the value of the cosmological and astrophysical parameters. We find that knowing the internal properties of a single galaxy allow our models to infer the value of $\Omega_{\rm m}$, at fixed $\Omega_{\rm b}$, with a $\sim10\%$ precision, while no constraint can be placed on $\sigma_8$. Our results hold for any type of galaxy, central or satellite, massive or dwarf, at all considered redshifts, $z\leq3$, and they incorporate uncertainties in astrophysics as modeled in CAMELS. However, our models are not robust to changes in subgrid physics due to the large intrinsic differences the two considered models imprint on galaxy properties. We find that the stellar mass, stellar metallicity, and maximum circular velocity are among the most important galaxy properties to determine the value of $\Omega_{\rm m}$. We believe that our results can be explained taking into account that changes in the value of $\Omega_{\rm m}$, or potentially $\Omega_{\rm b}/\Omega_{\rm m}$, affect the dark matter content of galaxies. That effect leaves a distinct signature in galaxy properties to the one induced by galactic processes. Our results suggest that the low-dimensional manifold hosting galaxy properties provides a tight direct link between cosmology and astrophysics., Comment: 20+6 pages, 15 figures, data and codes to reproduce the results publicly available at https://github.com/franciscovillaescusa/Cosmo1gal

The Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project was developed to combine cosmology with astrophysics through thousands of cosmological hydrodynamic simulations and machine learning. CAMELS contains 4,233 cosmological simulations, 2,049 N-body and 2,184 state-of-the-art hydrodynamic simulations that sample a vast volume in parameter space. In this paper we present the CAMELS public data release, describing the characteristics of the CAMELS simulations and a variety of data products generated from them, including halo, subhalo, galaxy, and void catalogues, power spectra, bispectra, Lyman-$\alpha$ spectra, probability distribution functions, halo radial profiles, and X-rays photon lists. We also release over one thousand catalogues that contain billions of galaxies from CAMELS-SAM: a large collection of N-body simulations that have been combined with the Santa Cruz Semi-Analytic Model. We release all the data, comprising more than 350 terabytes and containing 143,922 snapshots, millions of halos, galaxies and summary statistics. We provide further technical details on how to access, download, read, and process the data at \url{https://camels.readthedocs.io}., Comment: 18 pages, 3 figures. More than 350 Tb of data from thousands of simulations publicly available at https://www.camel-simulations.org

We present new constraints on the masses of the halos hosting the Milky Way and Andromeda galaxies derived using graph neural networks. Our models, trained on thousands of state-of-the-art hydrodynamic simulations of the CAMELS project, only make use of the positions, velocities and stellar masses of the galaxies belonging to the halos, and are able to perform likelihood-free inference on halo masses while accounting for both cosmological and astrophysical uncertainties. Our constraints are in agreement with estimates from other traditional methods., Comment: 2 figures, 2 tables, 7 pages. Code publicly available at https://github.com/PabloVD/HaloGraphNet

Understanding the halo-galaxy connection is fundamental in order to improve our knowledge on the nature and properties of dark matter. In this work we build a model that infers the mass of a halo given the positions, velocities, stellar masses, and radii of the galaxies it hosts. In order to capture information from correlations among galaxy properties and their phase-space, we use Graph Neural Networks (GNNs), that are designed to work with irregular and sparse data. We train our models on galaxies from more than 2,000 state-of-the-art simulations from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project. Our model, that accounts for cosmological and astrophysical uncertainties, is able to constrain the masses of the halos with a $\sim$0.2 dex accuracy. Furthermore, a GNN trained on a suite of simulations is able to preserve part of its accuracy when tested on simulations run with a different code that utilizes a distinct subgrid physics model, showing the robustness of our method. The PyTorch Geometric implementation of the GNN is publicly available on Github at https://github.com/PabloVD/HaloGraphNet, Comment: 20 pages, 8 figures, code publicly available at https://github.com/PabloVD/HaloGraphNet

We present the first results from applying the Santa Cruz semi-analytic model (SAM) for galaxy formation on merger trees extracted from a dark matter only version of the IllustrisTNG (TNG) simulations. We carry out a statistical comparison between the predictions of the Santa Cruz SAM and TNG for a subset of central galaxy properties at z = 0, with a focus on stellar mass, cold and hot gas mass, star formation rate (SFR), and black hole (BH) mass. We find fairly good agreement between the mean predictions of the two methods for stellar mass functions and the stellar mass vs. halo mass (SMHM) relation, and qualitatively good agreement between the SFR or cold gas mass vs. stellar mass relation and quenched fraction as a function of stellar mass. There are greater differences between the predictions for hot (circumgalactic) gas mass and BH mass as a function of halo mass. Going beyond the mean relations, we also compare the dispersion in the predicted scaling relations, and the correlation in residuals on a halo-by-halo basis between halo mass and galaxy property scaling relations. Intriguingly, we find similar correlations between residuals in SMHM in the SAM and in TNG, suggesting that these relations may be shaped by similar physical processes. Other scaling relations do not show significant correlations in the residuals, indicating that the physics implementations in the SAM and TNG are significantly different., Comment: 22 pages, 15 figures, accepted for publication in MNRAS

In large-scale hydrodynamical cosmological simulations, the fate of massive galaxies is mainly dictated by the modeling of feedback from active galactic nuclei (AGN). The amount of energy released by AGN feedback is proportional to the mass that has been accreted onto the BHs, but the exact sub-grid modeling of AGN feedback differs in all simulations. Whilst modern simulations reliably produce populations of quiescent massive galaxies at z<2, it is also crucial to assess the similarities and differences of the responsible AGN populations. Here, we compare the AGN population of the Illustris, TNG100, TNG300, Horizon-AGN, EAGLE, and SIMBA simulations. The AGN luminosity function (LF) varies significantly between simulations. Although in agreement with current observational constraints at z=0, at higher redshift the agreement of the LFs deteriorates with most simulations producing too many AGN of L_{x, 2-10 keV}~10^43-10^44 erg/s. AGN feedback in some simulations prevents the existence of any bright AGN with L_{x, 2-10 keV}>=10^45 erg/s (although this is sensitive to AGN variability), and leads to smaller fractions of AGN in massive galaxies than in the observations at z<=2. We find that all the simulations fail at producing a number density of AGN in good agreement with observational constraints for both luminous (L_{x, 2-10 keV}~10^43-10^45 erg/s) and fainter (L_{x, 2-10 keV}~10^42-10^43 erg/s) AGN, and at both low and high redshift. These differences can aid us in improving future BH and galaxy subgrid modeling in simulations. Upcoming X-ray missions (e.g., Athena, AXIS, and LynX) will bring faint AGN to light and new powerful constraints. After accounting for AGN obscuration, we find that the predicted number density of detectable AGN in future surveys spans at least one order of magnitude across the simulations, at any redshift., Comment: Published in MNRAS

A wealth of cosmological and astrophysical information is expected from many ongoing and upcoming large-scale surveys. It is crucial to prepare for these surveys now and develop tools that can efficiently extract most information. We present HIFlow: a fast generative model of the neutral hydrogen (HI) maps that is conditioned only on cosmology ($\Omega_{m}$ and $\sigma_{8}$) and designed using a class of normalizing flow models, the Masked Autoregressive Flow (MAF). HIFlow is trained on the state-of-the-art simulations from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project. HIFlow has the ability to generate realistic diverse maps without explicitly incorporating the expected 2D maps structure into the flow as an inductive bias. We find that HIFlow is able to reproduce the CAMELS average and standard deviation HI power spectrum (Pk) within a factor of $\lesssim$ 2, scoring a very high $R^{2} > 90\%$. By inverting the flow, HIFlow provides a tractable high-dimensional likelihood for efficient parameter inference. We show that the conditional HIFlow on cosmology is successfully able to marginalize over astrophysics at the field level, regardless of the stellar and AGN feedback strengths. This new tool represents a first step toward a more powerful parameter inference, maximizing the scientific return of future HI surveys, and opening a new avenue to minimize the loss of complex information due to data compression down to summary statistics., Comment: updated: 14 pages, 10 figures, a new section on inference has been added during revision. Accepted for publication in ApJ

We present the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) Multifield Dataset, CMD, a collection of hundreds of thousands of 2D maps and 3D grids containing many different properties of cosmic gas, dark matter, and stars from 2,000 distinct simulated universes at several cosmic times. The 2D maps and 3D grids represent cosmic regions that span $\sim$100 million light years and have been generated from thousands of state-of-the-art hydrodynamic and gravity-only N-body simulations from the CAMELS project. Designed to train machine learning models, CMD is the largest dataset of its kind containing more than 70 Terabytes of data. In this paper we describe CMD in detail and outline a few of its applications. We focus our attention on one such task, parameter inference, formulating the problems we face as a challenge to the community. We release all data and provide further technical details at https://camels-multifield-dataset.readthedocs.io., Comment: 17 pages, 1 figure. Third paper of a series of four. Hundreds of thousands of labeled 2D maps and 3D grids from thousands of simulated universes publicly available at https://camels-multifield-dataset.readthedocs.io

We train neural networks to perform likelihood-free inference from $(25\,h^{-1}{\rm Mpc})^2$ 2D maps containing the total mass surface density from thousands of hydrodynamic simulations of the CAMELS project. We show that the networks can extract information beyond one-point functions and power spectra from all resolved scales ($\gtrsim 100\,h^{-1}{\rm kpc}$) while performing a robust marginalization over baryonic physics at the field level: the model can infer the value of $\Omega_{\rm m} (\pm 4\%)$ and $\sigma_8 (\pm 2.5\%)$ from simulations completely different to the ones used to train it., Comment: 7 pages, 4 figures. Second paper of a series of four. The 2D maps, codes, and network weights used in this paper are publicly available at https://camels-multifield-dataset.readthedocs.io

Astrophysical processes such as feedback from supernovae and active galactic nuclei modify the properties and spatial distribution of dark matter, gas, and galaxies in a poorly understood way. This uncertainty is one of the main theoretical obstacles to extract information from cosmological surveys. We use 2,000 state-of-the-art hydrodynamic simulations from the CAMELS project spanning a wide variety of cosmological and astrophysical models and generate hundreds of thousands of 2-dimensional maps for 13 different fields: from dark matter to gas and stellar properties. We use these maps to train convolutional neural networks to extract the maximum amount of cosmological information while marginalizing over astrophysical effects at the field level. Although our maps only cover a small area of $(25~h^{-1}{\rm Mpc})^2$, and the different fields are contaminated by astrophysical effects in very different ways, our networks can infer the values of $\Omega_{\rm m}$ and $\sigma_8$ with a few percent level precision for most of the fields. We find that the marginalization performed by the network retains a wealth of cosmological information compared to a model trained on maps from gravity-only N-body simulations that are not contaminated by astrophysical effects. Finally, we train our networks on multifields -- 2D maps that contain several fields as different colors or channels -- and find that not only they can infer the value of all parameters with higher accuracy than networks trained on individual fields, but they can constrain the value of $\Omega_{\rm m}$ with higher accuracy than the maps from the N-body simulations., Comment: 11 pages, 7 figures. First paper of a series of four. All 2D maps, codes, and networks weights publicly available at https://camels-multifield-dataset.readthedocs.io

From 1,000 hydrodynamic simulations of the CAMELS project, each with a different value of the cosmological and astrophysical parameters, we generate 15,000 gas temperature maps. We use a state-of-the-art deep convolutional neural network to recover missing data from those maps. We mimic the missing data by applying regular and irregular binary masks that cover either $15\%$ or $30\%$ of the area of each map. We quantify the reliability of our results using two summary statistics: 1) the distance between the probability density functions (pdf), estimated using the Kolmogorov-Smirnov (KS) test, and 2) the 2D power spectrum. We find an excellent agreement between the model prediction and the unmasked maps when using the power spectrum: better than $1\%$ for $k<20 h/$Mpc for any irregular mask. For regular masks, we observe a systematic offset of $\sim5\%$ when covering $15\%$ of the maps while the results become unreliable when $30\%$ of the data is missing. The observed KS-test p-values favor the null hypothesis that the reconstructed and the ground-truth maps are drawn from the same underlying distribution when irregular masks are used. For regular-shaped masks on the other hand, we find a strong evidence that the two distributions do not match each other. Finally, we use the model, trained on gas temperature maps, to perform inpainting on maps from completely different fields such as gas mass, gas pressure, and electron density and also for gas temperature maps from simulations run with other codes. We find that visually, our model is able to reconstruct the missing pixels from the maps of those fields with great accuracy, although its performance using summary statistics depends strongly on the considered field., Comment: 14 pages, 6 figures, Submitted to APJ

We use a generic formalism designed to search for relations in high-dimensional spaces to determine if the total mass of a subhalo can be predicted from other internal properties such as velocity dispersion, radius, or star-formation rate. We train neural networks using data from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project and show that the model can predict the total mass of a subhalo with high accuracy: more than 99% of the subhalos have a predicted mass within 0.2 dex of their true value. The networks exhibit surprising extrapolation properties, being able to accurately predict the total mass of any type of subhalo containing any kind of galaxy at any redshift from simulations with different cosmologies, astrophysics models, subgrid physics, volumes, and resolutions, indicating that the network may have found a universal relation. We then use different methods to find equations that approximate the relation found by the networks and derive new analytic expressions that predict the total mass of a subhalo from its radius, velocity dispersion, and maximum circular velocity. We show that in some regimes, the analytic expressions are more accurate than the neural networks. We interpret the relation found by the neural network and approximated by the analytic equation as being connected to the virial theorem., Comment: 23 pages, 15 figures

The circumgalactic medium (CGM) contains information on gas flows around galaxies, such as accretion and supernova-driven winds, which are difficult to constrain from observations alone. Here, we use the high-resolution TNG50 cosmological magnetohydrodynamical simulation to study the properties and kinematics of the CGM around star-forming galaxies in $10^{11.5}-10^{12}\;M_{\odot}$ halos at $z\simeq$ 1 using mock MgII absorption lines, which we generate by postprocessing halos to account for photoionization in the presence of a UV background. We find that the MgII gas is a very good tracer of the cold CGM, which is accreting inward at inflow velocities of up to 50 km s$^{-1}$. For sight lines aligned with the galaxy's major axis, we find that MgII absorption lines are kinematically shifted due to the cold CGM's significant corotation at speeds up to 50% of the virial velocity for impact parameters up to 60 kpc. We compare mock MgII spectra to observations from the MusE GAs FLow and Wind (MEGAFLOW) survey of strong MgII absorbers ($\rm{EW}^{2796\r{A}}_{0}>0.5 \; \r{A}$). After matching the equivalent-width (EW) selection, we find that the mock MgII spectra reflect the diversity of observed kinematics and EWs from MEGAFLOW, even though the sight lines probe a very small fraction of the CGM. MgII absorption in higher-mass halos is stronger and broader than in lower-mass halos but has qualitatively similar kinematics. The median-specific angular momentum of the MgII CGM gas in TNG50 is very similar to that of the entire CGM and only differs from non-CGM components of the halo by normalization factors of $\lesssim$ 1 dex., Comment: Accepted to ApJ. 17 pages, 11 figures. The TNG50 data is now publicly available at https://www.tng-project.org

We investigate the specific angular momentum (sAM) $ j(

We investigate the morphology of the stellar distribution in a sample of Milky Way (MW) like galaxies in the TNG50 simulation. Using a local in shell iterative method (LSIM) as the main approach, we explicitly show evidence of twisting (in about 52% of halos) and stretching (in 48% of them) in the real space. This is matched with the re-orientation observed in the eigenvectors of the inertia tensor and gives us a clear picture of having a re-oriented stellar distribution. We make a comparison between the shape profile of dark matter (DM) halo and stellar distribution and quite remarkably see that their radial profiles are fairly close, especially at small galactocentric radii where the stellar disk is located. This implies that the DM halo is somewhat aligned with stars in response to the baryonic potential. The level of alignment mostly decreases away from the center. We study the impact of substructures in the orbital circularity parameter. It is demonstrated that in some cases, far away substructures are counter-rotating compared with the central stars and may flip the sign of total angular momentum and thus the orbital circularity parameter. Truncating them above 150 kpc, however, retains the disky structure of the galaxy as per initial selection. Including the impact of substructures in the shape of stars, we explicitly show that their contribution is subdominant. Overlaying our theoretical results to the observational constraints from previous literature, we establish fair agreement., Comment: 29 pages, 21 figures, Accepted for publication in ApJ

Using the TNG100 (100 Mpc)^3 simulation of the IllustrisTNG project, we demonstrate a strong connection between the onset of star formation quenching and the stellar size of galaxies. We do so by tracking the evolutionary history of extended and normal-size galaxies selected at z=2 with log(M_star) = 10.2 - 11 and stellar-half-mass-radii above and within 1-sigma of the stellar size--stellar mass relation, respectively. We match the stellar mass and star formation rate distributions of the two populations. By z=1, only 36% of the extended massive galaxies have quenched, in contrast to a quenched fraction of 69% for the normal-size massive galaxies. We find that normal-size massive galaxies build up their central stellar mass without a significant increase in their stellar size between z=2-4, whereas the stellar size of the extended massive galaxies almost doubles in the same time. In IllustrisTNG, lower black hole masses and weaker kinetic-mode feedback appears to be responsible for the delayed quenching of star formation in the extended massive galaxies. We show that relatively gas-poor mergers may be responsible for the lower central stellar density and weaker supermassive black hole feedback in the extended massive galaxies., Comment: 15 pages, 9 figures, Accepted for publication in the Astrophysical Journal

Many different studies have shown that a wealth of cosmological information resides on small, non-linear scales. Unfortunately, there are two challenges to overcome to utilize that information. First, we do not know the optimal estimator that will allow us to retrieve the maximum information. Second, baryonic effects impact that regime significantly and in a poorly understood manner. Ideally, we would like to use an estimator that extracts the maximum cosmological information while marginalizing over baryonic effects. In this work we show that neural networks can achieve that. We made use of data where the maximum amount of cosmological information is known: power spectra and 2D Gaussian density fields. We also contaminate the data with simplified baryonic effects and train neural networks to predict the value of the cosmological parameters. For this data, we show that neural networks can 1) extract the maximum available cosmological information, 2) marginalize over baryonic effects, and 3) extract cosmological information that is buried in the regime dominated by baryonic physics. We also show that neural networks learn the priors of the data they are trained on. We conclude that a promising strategy to maximize the scientific return of cosmological experiments is to train neural networks on state-of-the-art numerical simulations with different strengths and implementations of baryonic effects., Comment: 16 pages, 12 figures. Codes publicly available at https://github.com/franciscovillaescusa/baryons_marginalization

Galactic outflows have density, temperature, and velocity variations at least as large as that of the multiphase, turbulent interstellar medium (ISM) from which they originate. We have conducted a suite of parsec-resolution numerical simulations using the TIGRESS framework, in which outflows emerge as a consequence of interaction between supernovae (SNe) and the star-forming ISM. The outflowing gas is characterized by two distinct thermal phases, cool (T<10^4 K) and hot (T>10^6 K), with most mass carried by the cool phase and most energy and newly-injected metals carried by the hot phase. Both components have a broad distribution of outflow velocity, and especially for cool gas this implies a varying fraction of escaping material depending on the halo potential. Informed by the TIGRESS results, we develop straightforward analytic formulae for the joint probability density functions (PDFs) of mass, momentum, energy, and metal loading as distributions in outflow velocity and sound speed. The model PDFs have only two parameters, SFR surface density \Sigma_SFR and the metallicity of the ISM, and fully capture the behavior of the original TIGRESS simulation PDFs over \Sigma_SFR~(10^{-4},1)M_sun/kpc^2/yr. Employing PDFs from resolved simulations will enable galaxy formation subgrid model implementations with wind velocity and temperature (as well as total loading factors) that are based on theoretical predictions rather than empirical tuning. This is a critical step to incorporate advances from TIGRESS and other high-resolution simulations in future cosmological hydrodynamics and semi-analytic galaxy formation models. We release a python package to prototype our model and to ease its implementation., Comment: ApJL accepted. For associated python package, see https://twind.readthedocs.io/en/latest/

We compare three major large-scale hydrodynamical galaxy simulations (EAGLE, Illustris-TNG, and SIMBA) by forward modeling simulated galaxies into observational space and computing the fraction of isolated and quiescent low mass galaxies as a function of stellar mass. Using SDSS as our observational template, we create mock surveys and synthetic spectroscopic and photometric observations of each simulation, adding realistic noise and observational limits. All three simulations show a decrease in the number of quiescent, isolated galaxies in the mass range $\mathrm{M}_* = 10^{9-10} \ \mathrm{M}_\odot$, in broad agreement with observations. However, even after accounting for observational and selection biases, none of the simulations reproduce the observed absence of quiescent field galaxies below $\mathrm{M}_*=10^{9} \ \mathrm{M}_\odot$. We find that the low mass quiescent populations selected via synthetic observations have consistent quenching timescales, despite apparent variation in the late time star formation histories. The effect of increased numerical resolution is not uniform across simulations and cannot fully mitigate the differences between the simulations and the observations. The framework presented here demonstrates a path towards more robust and accurate comparisons between theoretical simulations and galaxy survey observations, while the quenching threshold serves as a sensitive probe of feedback implementations., Comment: 19 pages, 8 figures. Figure 4 presents the main result. Code used in this work may be accessed at github.com/IQcollaboratory/orchard. Submitted to ApJ

We present the Cosmology and Astrophysics with MachinE Learning Simulations --CAMELS-- project. CAMELS is a suite of 4,233 cosmological simulations of $(25~h^{-1}{\rm Mpc})^3$ volume each: 2,184 state-of-the-art (magneto-)hydrodynamic simulations run with the AREPO and GIZMO codes, employing the same baryonic subgrid physics as the IllustrisTNG and SIMBA simulations, and 2,049 N-body simulations. The goal of the CAMELS project is to provide theory predictions for different observables as a function of cosmology and astrophysics, and it is the largest suite of cosmological (magneto-)hydrodynamic simulations designed to train machine learning algorithms. CAMELS contains thousands of different cosmological and astrophysical models by way of varying $\Omega_m$, $\sigma_8$, and four parameters controlling stellar and AGN feedback, following the evolution of more than 100 billion particles and fluid elements over a combined volume of $(400~h^{-1}{\rm Mpc})^3$. We describe the simulations in detail and characterize the large range of conditions represented in terms of the matter power spectrum, cosmic star formation rate density, galaxy stellar mass function, halo baryon fractions, and several galaxy scaling relations. We show that the IllustrisTNG and SIMBA suites produce roughly similar distributions of galaxy properties over the full parameter space but significantly different halo baryon fractions and baryonic effects on the matter power spectrum. This emphasizes the need for marginalizing over baryonic effects to extract the maximum amount of information from cosmological surveys. We illustrate the unique potential of CAMELS using several machine learning applications, including non-linear interpolation, parameter estimation, symbolic regression, data generation with Generative Adversarial Networks (GANs), dimensionality reduction, and anomaly detection., Comment: 34 pages, 19 figures, Matches published version. CAMELS webpage at https://www.camel-simulations.org