Generating Dark Matter Subhalo Populations Using Normalizing Flows

Strong gravitational lensing is a powerful tool for probing the nature of dark matter, as lensing signals are sensitive to the dark matter substructure within the lensing galaxy. We present a comparative analysis of strong gravitational lensing signatures generated by dark matter subhalo populations using two different approaches. The first approach models subhalos using an empirical model, while the second employs the Galacticus semi-analytic model of subhalo evolution. To date, only empirical approaches have been practical in the analysis of lensing systems, as incorporating fully physical models was computationally infeasible. To circumvent this, we utilize a generative machine learning algorithm, known as a normalizing flow, to learn and reproduce the subhalo populations generated by Galacticus. We demonstrate that the normalizing flow algorithm accurately reproduces the Galacticus subhalo distribution while significantly reducing computation time compared to direct simulation. Moreover, we find that subhalo populations from Galacticus produce comparable results to the empirical model in replicating observed lensing signals under the fiducial dark matter model. This work highlights the potential of machine learning techniques in accelerating astrophysical simulations and improving model comparisons of dark matter properties.