Searcharxiv⌕ Search

arXiv subjects

Atsushi J. Nishizawa

Publications and source records attributed to Atsushi J. Nishizawa.

At least 19 recordsLinked to original sources

Limited Impact of CMB B-mode Polarization on Constraints on Cosmic Topology

The global spatial structure of the universe is a fundamental question in cosmology. Although the standard model usually assumes a trivial topology, a multiply connected universe can break statistical isotropy and generate off-diagonal covariance between the spherical-harmonic coefficients of cosmic microwave background fluctuations. In this work, we investigate whether CMB B-mode polarization improves constraints on the cubic three-torus, or $E_1$ topology. We construct the full harmonic-space covariance matrices of the temperature, E-mode, and primordial B-mode anisotropies, including correlations between different $(\ell,m)$ modes, and evaluate Gaussian likelihoods using random realizations generated from a statistically isotropic $Λ$CDM model with trivial topology. To isolate the intrinsic topological information carried by each CMB observable, we consider the ideal cosmic-variance-limited case, neglecting instrumental noise and gravitational lensing. We find that adding primordial B-mode polarization yields only a marginal improvement in the ability to distinguish the $E_1$ topology from the trivial topology, even under these ideal conditions. The different constraining powers of the temperature, E-mode, and B-mode anisotropies can be understood through the combined effects of their transfer-function responses and the topology-induced off-diagonal correlations in harmonic space. }

astro-ph.CO↗

The CAMELS-CROCODILE Simulation Suite: A New Cosmology--Astrophysics Playground for Machine Learning

We present CAMELS-CROCODILE, a suite of cosmological hydrodynamic simulations that extends the CAMELS framework with the \gadget{} smoothed particle hydrodynamics code and the Osaka feedback model. As a companion to the IllustrisTNG-based second-generation CAMELS suite, it provides an independent implementation of the baryonic physics that shapes galaxies, the intergalactic medium, and large-scale structure, broadening the range of subgrid models over which machine-learning (ML) inference can be marginalized and tested. The suite comprises $25$ and $50Mpc/h$ boxes organized into one-parameter, cosmic-variance, and Sobol-sequence sets that vary $26$ cosmological and astrophysical parameters around a fiducial Osaka model. We describe the simulation design and validate the suite against benchmarks and other CAMELS suites. Compared with IllustrisTNG, the fiducial model suppresses the matter power spectrum by only a few per cent, against up to ${\sim}27\%$ at $z=0$, and forms stars with similar efficiency below $10^{12.2} M_\odot/h$ but up to $2.5$ times more efficiently in group-scale halos, both consistent with weaker AGN feedback. We also present the stellar and black hole bias and the Ly$α$ absorption around galaxies. As a first ML application, we apply a graph neural network trained on IllustrisTNG galaxy catalogs, with frozen weights, to CAMELS-CROCODILE. It fails to recover $Ω_{\rm m}$ and $σ_8$ and returns overconfident posteriors, mainly for simulations containing more galaxies than any in its training set; within that range $Ω_{\rm m}$ is recovered with an error only $1.5$ times the in-distribution value. These results underline the need to train and test ML inference across physically distinct galaxy formation models.

astro-ph.GA↗

Future of Artificial Intelligence for Science in Japan 2024 Community Report

This white paper summarizes scientific challenges and AI/ML research opportunities identified through the FAIRS Japan 2024 unconference process. The discussion focuses on three major physics domains: accelerator physics, cosmology and astrophysics, and neutrino physics. Although each domain has distinct scientific goals and experimental constraints, several common technical themes emerge: high-dimensional reconstruction, fast and accurate simulation, uncertainty propagation, simulation-to-data mismatch, anomaly detection, real-time decision-making, and shared infrastructure.

hep-ph↗

Weak-lensing Shear-Selected Galaxy Clusters from the Hyper Suprime-Cam Subaru Strategic Program: III. A precision cosmological sample enabled by optical confirmation

We develop fCAMIRA (forced-mode CAMIRA), a tool for optical cluster confirmation, and apply it to a sample of 129 weak-lensing (WL) shear-selected galaxy clusters identified in aperture-mass maps obtained from the Hyper Suprime-Cam Subaru Strategic Program Three-Year (HSC-SSP Y3) weak-lensing data. fCAMIRA is built upon the CAMIRA cluster-finding algorithm and relies on a red-sequence (RS) galaxy model that is calibrated in a data-driven way. The RS model adopts the metallicity-luminosity relation measured in this work using X-ray-selected clusters up to redshift $z\approx1.3$, followed by the calibration of color offsets using large spectroscopic samples. With the RS model, we build two types of galaxy richness maps, one obtained with a spatial filter matched to a typical cluster size of $R=0.8\,h^{-1}\,\mathrm{Mpc}$ and the other obtained with a fixed angular-size filter identical to that used in constructing the WL aperture-mass maps. The fCAMIRA algorithm utilizes these two richness maps, identifies all optical counterpart candidates along the line of sight of each shear-selected cluster, and measures the cluster photometric redshift from the highest-ranked counterpart. The ranking is determined by the fractional lensing contribution $f_{\mathrm{lens}}$ of each candidate. Using available spectroscopic cluster redshifts, we quantify the mean bias and scatter in the cluster photometric redshifts at levels of approximately 0.005 and 0.008, respectively, demonstrating excellent photometric-redshift performance. We compare the fCAMIRA photometric redshifts with estimates from direct positional cross-matching and find that approximately 8% of the total sample exhibits redshift discrepancies greater than 0.15. This outlier fraction is primarily attributed to projection effects, leading to the misidentification of the optical counterparts. (abridged)

astro-ph.CO↗

Standard Reconstruction Shifts the Optimal Input Scale for CNN-Based Density-Field Reconstruction

We investigate convolutional neural network (CNN) methods for reconstructing the high-redshift density field from late-time large-scale structure, focusing on how the physical scale of the CNN input changes when standard first-order reconstruction is applied beforehand. Using dark-matter-only $N$-body simulations, we compare three approaches: a single-input CNN, a dual-input CNN combining two physical scales, and a single-input CNN applied to the density field after standard reconstruction. We vary the physical side length of the input sub-box over $L_\mathrm{sub}\sim38$-$380~h^{-1}\mathrm{Mpc}$ while keeping its numerical size fixed at $39^3$ voxels, allowing us to examine the trade-off between spatial context and resolution. For the CNN applied directly to the evolved density field, the reconstruction performs best at $L_\mathrm{sub}\sim150$-$200~h^{-1}\mathrm{Mpc}$. After standard reconstruction, however, the preferred scale shifts to $L_\mathrm{sub}\sim38$-$114~h^{-1}\mathrm{Mpc}$. The single-input CNN after standard reconstruction consistently outperforms both the single- and dual-input CNNs without standard reconstruction according to the normalized loss, density probability distribution, Kullback-Leibler divergence, residual field, and Fourier-space correlation. These results indicate that coherent large-scale displacements are more efficiently recovered by perturbative reconstruction, while the CNN is better suited to modelling the remaining quasi-linear and non-linear evolution on smaller scales. The preferred post-reconstruction input range includes the effective receptive scale of approximately $60~h^{-1}\mathrm{Mpc}$ adopted in previous hybrid reconstruction studies. Our findings therefore support a physically motivated separation of scales between analytic and data-driven reconstruction and demonstrate the advantage of combining the two approaches.

astro-ph.CO↗

Searching optimal scales for reconstructing cosmological initial conditions using convolutional neural networks

Reconstructing the initial density field of the Universe from the late-time matter distribution is a nontrivial task with implications for understanding structure formation in cosmology, offering insights into early Universe conditions. Convolutional neural networks (CNNs) have shown promise in tackling this problem by learning the complex mapping from nonlinear evolved fields back to initial conditions. Here we investigate the effect of varying input sub-box size in single-input CNNs. We find that intermediate scales ($L_\mathrm{sub} \sim 152\,h^{-1}\,\mathrm{Mpc}$) strike the best balance between capturing local detail and global context, yielding the lowest validation loss and most accurate recovery across multiple statistical metrics. We then propose a dual-input model that combines two sub-boxes of different sizes from the same simulation volume. This model significantly improves reconstruction performance, especially on small scales over the best single-input case, despite utilizing the same parent simulation box. This demonstrates the advantage of explicitly incorporating multi-scale context into the network. Our results highlight the importance of input scale and network design in reconstruction tasks. The dual-input approach represents a simple yet powerful enhancement that leverages fixed input information more efficiently, paving the way for more accurate cosmological inference from large-scale structure surveys.

astro-ph.CO↗

Exploring the Cosmic Dawn through the 21 cm Forest and High-redshift Radio Sources with the SKA

The 21~cm forest, seen as absorption features in the spectra of distant radio sources, is produced by intervening neutral hydrogen and offers a direct probe of the neutral intergalactic medium during the epoch of reionization (EoR). Because it is sensitive to small-scale structure and gas temperature, it can constrain the thermal history of the early Universe and physics that affects structure formation. Detecting individual absorption lines is challenging, mainly because of their weakness and the scarcity of high-redshift radio-bright sources. Recent progress, however, has made 21~cm forest studies increasingly feasible: new statistical observables can improve sensitivity within realistic observing times, updated radio-source counts have revised expectations for suitable background quasars, and deep-learning methods can extract physical information more efficiently. In addition, new approaches have been developed to separate astrophysical effects from early galaxies from fundamental-physics effects on small-scale structure. With the Square Kilometre Array (SKA), the 21~cm forest will therefore provide a promising route to study early heating, possible exotic energy injection, dark matter properties, neutrino mass, the running spectral index, and baryon--dark-matter relative velocity. This chapter reviews recent developments in 21~cm forest research and discusses observational strategies and prospects for constraining the first galaxies and fundamental physics with SKA-Low.

astro-ph.CO↗

Sensitivity toward dark matter annihilation imprints on 21-cm signal with SKA-Low: A convolutional neural network approach

This study investigates the sensitivity of the radio interferometers to identify imprints of spatially inhomogeneous dark matter annihilation signatures in the 21-cm signal during the pre-reionization era. We focus on the upcoming low-mode survey of the Square Kilometre Array (SKA-Low) telescope. Using CNNs, we analyze simulated 3D 21-cm differential brightness temperature maps generated via the DM21cm code, which is based on 21cmFAST and DarkHistory, to distinguish between spatially homogeneous and inhomogeneous energy injection/deposition scenarios arising from dark matter annihilation. The inhomogeneous case accounts for local dark matter density contrasts and gas properties, such as thermal and ionization states, while the homogeneous model assumes uniform energy deposition. Our study focuses on two primary annihilation channels to electron-positron pairs ($e^+e^-$) and photons ($γγ$), exploring dark matter masses from 1 MeV to 100 MeV and a range of annihilation cross-sections. For $γγ$ channel, the distinction across dark matter models is less pronounced due to the larger mean free path of the emitted photons, resulting in a more uniform energy deposition. For $e^+e^-$ channel, the results indicate that the CNNs can effectively differentiate between the inhomogeneous and homogeneous cases. Despite observational challenges, the results demonstrate that these effects remain detectable even after incorporating noise from next-generation radio interferometers, such as the SKA. We find that the inhomogeneous dark matter annihilation models can leave measurable imprints on the 21-cm signal maps distinguishable from the homogeneous scenarios for the dark matter masses $m_{\rm DM}=1$ MeV and the annihilation cross-sections of $\geq 5 \times 10^{-30}~{\rm cm^3/sec}$ ($\geq 5 \times 10^{-29}~{\rm cm^3/sec}$ for $m_{\rm DM}=100$ MeV) for moderate SKA-Low noise.

astro-ph.CO↗

Catalogs of optically-selected clusters and photometric luminous red galaxies from the Hyper Suprime-Cam Subaru Strategic Program final year dataset

We construct samples of optically-selected clusters and photometric luminous red galaxies (LRGs) from the Hyper Suprime-Cam Subaru Strategic Program final year dataset covering $\sim 1200$~deg$^2$. The cluster catalogs extend out to the redshift of $1.38$ and contain more than 10000 clusters with richness larger than $15$, where the richness is defined to be a membership probability weighted number of galaxies above the stellar masses of approximately $10^{10.2}M_\odot$. The total number of probable red cluster member galaxies in these clusters are more than $6\times 10^5$. Photometric redshifts of the clusters are shown to be precise with the scatter better than $\sim 0.01$ for a wide redshift range. We detect stacked weak lensing signals of clusters out to the redshift of $1$, and use them to update constraints on the mass-richness relation. Our catalog of about 6 million photometric LRGs extend out to the redshift of $1.25$, and have the scatter of the photometric redshift better than $\sim 0.02$ for the redshift range between $0.4$ and $1.0$.

astro-ph.CO↗

Insights into the dependence of galaxy properties on the environment with explainable machine learning models

Galaxies reside within dark matter halos, but their properties are influenced not only by their halo properties but also by the surrounding environment. We construct an interpretable neural network framework to characterize the surrounding environment of galaxies and investigate the extent to which their properties are affected by neighboring galaxies in IllustrisTNG300 data ($z=0$). Our models predict galaxy properties (stellar mass and star formation rate) given dark matter subhalo properties of both host subhalo and of surrounding galaxies, which serve as an explainable, flexible galaxy-halo connection model. We find that prediction accuracy peaks when incorporating only the nearest neighboring galaxy for stellar mass prediction, while star formation rate prediction benefits from information from up to the third-nearest neighbor. We determine that environmental influence follows a clear hierarchical pattern, with the nearest neighbor providing the dominant contribution that diminishes rapidly with additional neighbors. We confirm that central and satellite galaxies, as well as different galaxy categories based on mass and star-forming activity, exhibit distinct environmental dependencies. Environmental dependence for low-mass galaxies ($\log(M_*/M_\odot) < 10$) shows 35-50% environmental contribution compared to just 8-30% for massive centrals, while satellite galaxies experience consistently stronger environmental effects than centrals across all populations. Furthermore, we find that the most significant attribute from neighboring subhalos for predicting target galaxy properties is its distance to the nearest neighboring galaxy. These quantitative results offer guidance for constructing more sophisticated empirical and semi-analytic models of galaxy formation that explicitly include environmental dependence as a function of galaxy type and mass.

astro-ph.GA↗

Modelling Galaxy Clustering and Tomographic Galaxy-Galaxy Lensing with HSC Y3 and SDSS using the Point-Mass Correction Model and Redshift Self-Calibration

The combination of galaxy-galaxy weak lensing and galaxy clustering is a powerful probe of the cosmological model, and exploration of how to best model and extract this information from the signals is essential. We present the measurement of the galaxy-galaxy weak lensing signals using the SDSS DR11 spectroscopic galaxies as lens galaxies, and the HSC Y3 shear catalog as source galaxies, binned into four tomographic bins by their photometric redshift. The SDSS DR11 galaxies, with a redshift range $0.15<z<0.7$, are binned into three redshift bins, each as a probe for measuring the projected correlation function, $w_p(R_p)$. We measure the galaxy-galaxy lensing signal $ΔΣ(R_p)$ in 12 lens-source bin pairs and show that there is no evidence for significant systematic biases in the measurement with null testing. We combine our $w_p(R_p)$ and $ΔΣ(R_p)$ ($2\times2$pt) data vectors and perform likelihood inference with a flat $Λ$CDM model. For $ΔΣ(R_p)$, we extend the lower limit of the scale cut compared to previous HSC Y3 analyses to $2 h^{-1}$Mpc by including a point-mass correction term in addition to the minimal bias model. We present various tests to validate our model and provide extended consistency tests. In the $Λ$CDM context, our fiducial model yields $S_8 = 0.804^{+0.051}_{-0.051}$. The $2\times2$pt data vector provides redshift parameter constraints for the third and fourth redshift bins $Δz_3 = -0.079^{+0.074}_{-0.084}$, and $Δz_4 = -0.203^{+0.167}_{-0.206}$, which is consistent with results from the previous tomographic cosmic shear studies, and serves as the foundation for a future $3\times 2$pt analysis.

astro-ph.CO↗

Cosmology and Source Redshift Constraints from Galaxy Clustering and Tomographic Weak Lensing with HSC Y3 and SDSS using the Point-Mass Correction Model

The combination of galaxy clustering and weak lensing is a powerful probe of the cosmology model. We present a joint analysis of galaxy clustering and weak lensing cosmology using SDSS data as the tracer of dark matter (lens sample) and the HSC Y3 dataset as source galaxies. The analysis divides HSC Y3 galaxies into four tomographic bins for both galaxy-galaxy lensing and cosmic shear measurements, and employs a point-mass correction model to utilize galaxy-galaxy lensing signals down to 2$h^{-1}$Mpc, extending up to 70$h^{-1}$Mpc. These strategies enhance the signal-to-noise ratio of the galaxy-galaxy lensing data vector. Using a flat $Λ$CDM model, we find $S_8 = 0.780^{+0.029}_{-0.030}$, and using a $w$CDM model, we obtain $S_8 = 0.756^{+0.038}_{-0.036}$ with $w = -1.176^{+0.310}_{-0.346}$. We apply uninformative priors on the redshift mean-shift parameters for the third and fourth tomographic bins. Leveraging the self-calibration power of tomographic weak lensing, we measure $Δz_3 = -0.112^{+0.046}_{-0.049}$ and $Δz_4 = -0.185^{+0.071}_{-0.081}$, in agreement with previous HSC Y3 results. This demonstrates that weak lensing self-calibration can achieve redshift constraints comparable to other methods such as photometric and clustering redshift calibration.

astro-ph.CO↗

CMB lensing tomography with clustering estimation of lens redshift distributions

We develop a clustering-based redshift estimation approach for CMB lensing tomography, focusing on the kernel function of the lensing galaxies. Within a linear galaxy bias framework, we derive estimators for this kernel from two-point cross-correlations between lens mass and reference samples. The reconstructed kernel then enables a theoretical prediction for the angular cross-power spectrum \(C_{gκ}\) between CMB lensing convergence and lens galaxies. As a proof of concept, we measure \(C_{gκ}\) by correlating the \emph{Planck} PR4 convergence map with NVSS+SUMSS radio galaxies (\(0\lesssim z\lesssim 3\)). We estimate the radio-galaxy kernel by collectively cross-correlating their distribution with spectroscopic and photometric surveys (2MPZ, LOWZ-CMASS, eBOSS DR16 LRGs, and Gaia-unWISE QSOs). From the measured \(C_{gκ}\), we obtain \(σ_8 = 0.86^{+0.12}_{-0.09}\) when the density parameter is set to the {\it Planck} value of $Ω_m = 0.315$; this is in good agreement with the \emph{Planck} normalisation of $σ_8 = 0.812$.

astro-ph.CO↗

Lyman-$α$ forest power spectrum and its cross-correlation with dark matter halos in different astrophysical models

The Ly$α$ forest, a series of HI absorption lines in the quasar spectra, is a powerful tool for probing the large-scale structure of the intergalactic medium. Its three-dimensional (3D) correlation and cross-correlations with quasars allow precise measurements of the baryon acoustic oscillation feature and redshift space distortions at redshifts $z>2$. Understanding small-scale astrophysical phenomena, such as star formation and feedback, is crucial for full-shape analyses. In this study, we measure the 3D auto-power spectrum of the Ly$α$ forest and its cross-power spectrum with halos using hydrodynamic simulations from the GADGET3-OSAKA code, which includes models for star formation and supernova feedback. Across five astrophysical models, we find significant deviations from the Fiducial model, with $5-10\,\%$ differences for wavenumbers $k>2\,h\mathrm{Mpc}^{-1}$ in the Ly$α$ auto-power spectrum. The Ly$α\,\times\,$halo cross-power spectra show even larger deviations, exceeding $10\,\%$ in some cases. Using the fitting models of Arinyo-i-Prats et al. (2015) and Givans et al. (2022), we jointly fit the Ly$α$ auto- and Ly$α$ $\times$ halo cross-power spectra, and assess the accuracy of the estimated $fσ_8$ parameter by comparing it with the ground truth from the simulations, while varying the maximum wavenumber $k_\mathrm{max}$ and minimum halo mass $M_h$. Our results demonstrate that the extended model of Givans et al. (2022) is highly effective in reproducing $fσ_8$ at $k_\mathrm{max}\leq3.0\,h\mathrm{Mpc}^{-1}$ for $M_h>10^{10.5} M_\odot$, and remains robust against astrophysical uncertainties.

astro-ph.CO↗

Machine Learning Constraints on Dark Matter Annihilation during the Epoch of Reionization: A CNN Analysis of the 21-cm Signal

We explore the impact of dark matter annihilation on the 21-cm signal during the cosmic dawn and epoch of reionization (EoR). Using modified 21cmFAST simulations and convolutional neural networks (CNNs), we investigate how energy injected into the intergalactic medium (IGM) through dark matter annihilation affects the evolution of the 21-cm differential brightness temperature. Focusing on two annihilation channels, photon-photon ($γγ$) and electron-positron ($e^+e^-$), we examine a broad range of dark matter masses and annihilation cross-sections. Our results show that CNNs outperform traditional power spectrum analysis by effectively distinguishing between subtle differences in simulated 21-cm maps produced by annihilation and non-annihilation scenarios. We also demonstrate that the structure formation boost, driven by dark matter clumping into halos and subhalos, significantly enhances the annihilation signal and alters the thermal and ionization history of the IGM. This enhancement leads to a noticeable effect on the 21-cm signal, including a shift from absorption to emission as dark matter annihilation heats the IGM at lower redshifts. By incorporating observational noise from upcoming radio interferometers, particularly the Square Kilometer Array (SKA), we show that these effects remain detectable despite observational challenges. We find that the dark matter annihilation models can leave measurable imprints on the 21-cm signal distinguishable from the non-annihilation scenarios for the dark matter masses $m_{\rm DM}=100$ MeV and the annihilation cross-sections of $\langle σv\rangle \simeq 10^{-31}~{\rm cm}^3/{\rm s}$ ($\langle σv\rangle \simeq 10^{-32}~{\rm cm}^3/{\rm s}$ for $m_{\rm DM}=1$ MeV and $\langle σv\rangle \simeq 10^{-24}~{\rm cm}^3/{\rm s}$ for $m_{\rm DM}=1$ TeV).

astro-ph.CO↗

A Systematic Search of Distant Superclusters with the Subaru Hyper Suprime-Cam Survey

Superclusters, encompassing environments across a wide range of overdensities, can be regarded as unique laboratories for studying galaxy evolution. Although numerous supercluster catalogs have been published, none of them goes beyond redshift $z=0.7$. In this work, we adopt a physically motivated supercluster definition, requiring that superclusters should eventually collapse even in the presence of dark energy. Applying a friends-of-friends (FoF) algorithm to the CAMIRA cluster sample constructed using the Subaru Hyper Suprime-Cam survey data, we have conducted the first systematic search for superclusters at $z=0.5-1.0$ and identified 673 supercluster candidates over an area of 1027 deg$^2$. The FoF algorithm is calibrated by evolving $N$-body simulations to the far future to ensure high purity. We found that these high-$z$ superclusters are mainly composed of $2-4$ clusters, suggesting the limit of gravitationally bound structures in the younger Universe. In addition, we studied the properties of the clusters and brightest cluster galaxies (BCGs) residing in different large-scale environments. We found that clusters associated with superclusters are typically richer, but no apparent dependence of the BCG properties on large-scale structures is found. We also compared the abundance of observed superclusters with mock superclusters extracted from halo light cones, finding that photometric redshift uncertainty is a limiting factor in the performance of superclusters detection.

astro-ph.CO↗

Non-Linearity-Free prediction of the growth-rate $fσ_8$ using Convolutional Neural Networks

The growth-rate $fσ_8(z)$ of the large-scale structure of the Universe is an important dynamic probe of gravity that can be used to test for deviations from General Relativity. However, for galaxy surveys to extract this key quantity from cosmological observations, two important assumptions have to be made: i) a fiducial cosmological model, typically taken to be the cosmological constant and cold dark matter ($Λ$CDM) model and ii) the modeling of the observed power spectrum, especially at non-linear scales, which is particularly dangerous as most models used in the literature are phenomenological at best. In this work, we propose a novel approach involving convolutional neural networks (CNNs), trained on the Quijote N-body simulations, to predict $fσ_8(z)$ directly and without assuming a model for the non-linear part of the power spectrum, thus avoiding the second of the assumptions above. This could serve as an initial step towards the future development of a method for parameter inference in Stage IV surveys. We find that the predictions for the value of $fσ_8$ from the CNN are in excellent agreement with the fiducial values since they outperform a maximum likelihood analysis and the CNN trained on the power spectrum. Therefore, we find that the CNN reconstructions provide a viable alternative to avoid the theoretical modeling of the non-linearities at small scales when extracting the growth rate.

astro-ph.CO↗

Differentiating Warm Dark Matter Models through 21cm Line Intensity Mapping: A Convolutional Neural Network Approach

We apply the convolutional neural networks (CNNs) to the mock 21cm maps from the post-reionization epoch to show that the $Λ$ cold dark matter and warm dark matter (WDM) model can be distinguished for WDM particle masses $m_{FD}<3$\,keV, under the assumption of thermal production of WDM following the Fermi-Dirac (FD) distribution. We demonstrate that the CNN is a potent tool in distinguishing the dark matter masses, highlighting its sensitivity to the subtle differences in the 21cm maps produced by varying dark matter masses. Furthermore, we extend our analysis to encompass different WDM production mechanisms, recognizing that the dark matter production mechanism in the early Universe is among the sources of the most significant uncertainty for the dark matter model building. In this work, given the mass of the dark matter, we discuss the feasibility of discriminating four different WDM models: Fermi-Dirac (FD) distribution model, neutrino minimal Standard Model ($ν$MSM), Dodelson-Widrow (DW), and Shi-Fuller (SF) model. For instance, when the WDM mass is 2\,keV, we show that one can differentiate between CDM, FD, $ν$MSM, and DW models while discerning between the DW and SF models turns out to be challenging. Our results reinforce the viability of the CNN as a robust analysis for 21cm maps and shed light on its potential to unravel the features associated with different dark matter production mechanisms.

astro-ph.CO↗