Nathaniel Starkman

Cold dark matter predicts triaxial halos; self-interacting dark matter rounds them toward spheres.

Cold dark matter predicts triaxial halos; self-interacting dark matter rounds them toward spheres. A stellar stream in a spherical halo can never curve away from its host galaxy’s center in projection, but one in a triaxial halo can. The rate of these convexities tracks halo shape across a population, so catalogs on the scale of Euclid, Rubin and Roman can tell CDM from SIDM at up to 5σ.

Euclid is turning up stellar streams around galaxies across a cosmological volume.

Euclid is turning up stellar streams around galaxies across a cosmological volume. A stream’s shape on the sky constrains the shape and center of its host’s dark matter halo, complementing weak lensing. Across 13 galaxies in Euclid’s first quick data release, halos are consistent with spherical, albeit with a mild preference for flattening, q = 0.95 (+0.05, −0.10), in line with ΛCDM. Thousands more streams are expected over the mission.

A physics-informed neural network learns only the corrections to an analytic galactic potential, so it stays interpretable while reaching sub-percent acceleration errors and more faithful orbits than analytic models alone.

A physics-informed neural network learns only the corrections to an analytic galactic potential, so it stays interpretable while reaching sub-percent acceleration errors and more faithful orbits than analytic models alone. Bayesian and time-dependent extensions add uncertainties and follow the potential as it evolves. Led by undergraduate Charlotte Myers; code in galactoPINNs, first shown at NeurIPS 2025.

JAX gives scientific Python automatic differentiation, compilation and GPUs, but no physical units.

JAX gives scientific Python automatic differentiation, compilation and GPUs, but no physical units. unxt adds them: quantities that carry their units through jitted, differentiated and vectorized code, with astropy.units under the hood and an interface astropy users will recognize. It underpins the GalacticDynamics stack, including coordinax and galax.

Using a stellar stream to weigh the Milky Way’s dark matter starts with knowing which stars belong to it.

Using a stellar stream to weigh the Milky Way’s dark matter starts with knowing which stars belong to it. Mixture density networks model a stream’s track, width, density and kinematics from all the available astrometry and photometry, with normalizing flows for the background, even where phase-space data are incomplete. Applied to GD-1 and Palomar 5, the result is public catalogs of stellar membership probabilities.