Predicting the Subhalo Mass Functions in Simulations from Galaxy Images

TL;DR

This work develops a pipeline that predicts galaxy-specific SHMFs conditioned on an assumed WDM mass by linking galaxy images to SHMF parameters via a CNN-NSF architecture trained on DREAMS simulations and image synthesis. It shows that including galaxy morphology reduces uncertainties in the SHMF parameters $(A, m_{\rm wdm})$ compared to mass-only predictions, enabling per-galaxy tests against strong lensing constraints and providing a scalable framework to compare DM models with observations. The approach emphasizes the need for diverse high-resolution simulations to capture environmental variation and paves the way for applying per-galaxy SHMF predictions to lensing data, though current training is limited to Milky Way–mass isolated halos with simplified observational realism. Future work will broaden the training set, incorporate more realistic imaging effects, and extend applicability to observed lens systems.

Abstract

Strong gravitational lensing provides a powerful tool to directly infer the dark matter (DM) subhalo mass function (SHMF) in lens galaxies. However, comparing observationally inferred SHMFs to theoretical predictions remains challenging, as the predicted SHMF can vary significantly between galaxies - even within the same cosmological model - due to differences in the properties and environment of individual galaxies. We present a machine learning framework to infer the galaxy-specific predicted SHMF from galaxy images, conditioned on the assumed inverse warm DM particle mass $M^{-1}_{\rm DM}$. To train the model, we use 1024 high-resolution hydrodynamical zoom-in simulations from the DREAMS suite. Mock observations are generated using Synthesizer, excluding gas particle contributions, and SHMFs are computed with the Rockstar halo finder. Our neural network takes as input both the galaxy images and the inverse DM mass. This method enables scalable, image-based predictions for the theoretical DM SHMFs of individual galaxies, facilitating direct comparisons with observational measurements.

Figures (3)

• Figure 1: A flowchart showing the training procedure, label generation, and inference.
• Figure 2: The galaxy-specific halo mass functions under different cosmologies. The figure shows the galaxy-specific halo mass functions for different WDM masses, with the corresponding galaxy image in the plot. The displayed galaxies come from three different WDM mass regions. The black dots are the counts of subhalos from the DREAMS simulation with Poisson uncertainties. The blue contours show the samples of the NF conditioned only on $M_{\rm DM}$, and the red contours the samples of the NF conditioned on both $M_{\rm DM}$ and the galaxy image. The NF conditioned only on $M_{\rm DM}$ shows a broader posterior range and bigger uncertainties on the fit.
• Figure 3: Similar to Figure \ref{['fig:NF_visual_comp']}. The shown galaxies are the same as in Figure \ref{['fig:hmf_fits']}, but the orientation of the galaxy images is different. The blue contours show the samples of the MCMC fit, and the red contours the samples of the NF conditioned on $M_{\rm DM}$ and the galaxy image. Overall, the sampled parameters of the NF represent the fits obtained directly through the likelihood and cover the full range of MCMC sampled parameters.