A new $U(1)$ dark gauge group coupled to the Standard Model (SM) via the kinetic mixing portal provides a natural dark matter candidate in the form of the Higgs field, $h_d$, responsible for generating the mass of the dark photon, $\gamma_d$. We show that the condition $m_{h_d}\leq m_{\gamma_d}$, together with smallness of the kinetic mixing parameter, $\epsilon$, and/or dark gauge coupling, $g_d$, leads the dark Higgs to be sufficiently metastable to constitute dark matter. We analyze the Universe's thermal history and show that both freeze-in, ${\rm SM}\to \{\gamma_d, h_d\}$, and freeze-out, $ \{\gamma_d, h_d\} \to {\rm SM}$, processes can lead to viable dark Higgs dark matter with a sub-GeV mass and a kinetic mixing parameter in the range $10^{-13}\lesssim\epsilon\lesssim10^{-6}$. Observable signals in astrophysics and cosmology include modifications to primordial elemental abundances, altered energetics of supernovae explosions, dark Higgs decays in the late Universe, and dark matter self-interactions.

Low-mass structures of dark matter (DM) are expected to be entirely devoid of light-emitting regions and baryons. Precisely because of this lack of baryonic feedback, small-scale substructures of the Milky Way are a relatively pristine testing ground for discovering aspects of DM microphysics and primordial fluctuations on subgalactic scales. In this work, we report results from the first search for Galactic DM subhalos with time-domain astrometric weak gravitational lensing. The analysis is based on a matched-filter template of local lensing corrections to the proper motion of stars in the Magellanic Clouds. We describe a data analysis pipeline detailing sample selection, background subtraction, and handling outliers and other systematics. For tentative candidate lenses, we identify a signature based on an anomalous parallax template that can unequivocally confirm the presence of a DM lens, opening up prospects for robust discovery potential with full time-series data. We present our constraints on substructure fraction $f_l \lesssim 5$ at 90% CL (and $f_l \lesssim 2$ at 50% CL) for compact lenses with radii $r_l < 1\,\mathrm{pc}$, with best sensitivity reached for lens masses $M_l$ around $10^7$-$10^8\,M_\odot$. Parametric improvements are expected with future astrometric data sets; by end of mission, $Gaia$ could reach $f_l \lesssim 10^{-3}$ for these massive point-like objects, and be sensitive to lighter and/or more extended subhalos for $\mathcal{O}(1)$ substructure fractions.