We derive constraints on a possible velocity-dependent DM-nucleon scattering cross section, for Dark Matter in the 10 MeV -- 100 GeV mass range, using the XQC, DAMIC, and CRESST 2017 Surface Run experiments. We report the limits on cross sections of the form $\sigma=\sigma_0\,v^n$, for a range of velocity dependencies with $n\in\{-4,-2,-1,0,1,2\}$. We point out the need to measure the efficiency with which nuclear recoil energy in the sub-keV range thermalizes, rather than being stored as Frenkel pairs in the semi-conductor lattice. The possibility of a significant inefficiency leaves open a considerable `hole' in the limits for mass in the $\sim$ 0.2 -- 2 GeV range, which XQC and CRESST can potentially fill when the thermalization efficiency is measured. We call attention to the asymmetry between a conventional lower limit cross section and the `upper-reach cross section' imposed by attenuation in an overburden -- an upper boundary being extremely sharp but quite insensitive to the statistics of the experiment. Considering the recent interest to use dark matter-baryon interaction with velocity dependence $n=-4$ to explain the EDGES 21 cm anomaly, we also derive the limits on milli-charged DM that scatters off protons and electrons under a Coulomb-like interaction. We find that much but not all of the region of interest for the EDGES anomaly can be excluded.

We improve limits on the spin-independent scattering cross section of Dark Matter on nucleons, for DM in the 300 MeV -- 100 GeV mass range, based on the DAMIC and XQC experiments. Our results close the window which previously existed in this mass range, for a DM-nucleon cross section of order $\approx \mu$b, assuming the standard velocity distribution.

As experimental constraints on DM interactions become ever more sensitive and push into new regimes of DM mass, it becomes more and more challenging to accurately model the process by which Dark Matter particles lose energy through scattering in the Earth's surface or other overburdens. We show that a commonly-used approximation due to Starkman, Gould, Esmailzadeh and Dimopoulos (SGED) can fail badly in computing the attenuation, even while being useful for an order-of-magnitude estimate of the maximum cross section reach. We introduce a method of importance sampling which makes Monte-Carlo simulation of energy loss feasible, in spite of factor-$10^7$ or greater attenuation. We demonstrate the validity of our new method and expose multiple problems with the SGED approximation, this reveals interesting features of the energy loss process. We spot-check the recent Emken, Kouvaris and Shoemaker "$5\,\Delta v$" prescription to place limits on cross sections based on a limited-statistics analysis, and find that an accurate simulation yields a factor of $\rm{4.4\times10^6}$ and $\rm{2.4\times10^4}$ larger number of events, for 50 MeV and 1 GeV DM mass respectively, than if the EKS $5\,\Delta v$ prescription were valid.