I'm broadly interested in cosmology, astrophysics, and aspects of theoretical physics. My current research focuses on the late universe and how to better understand and predict the formation of large-scale structure in the universe.
Dark matter is typically modelled as a cold collisionless fluid, which is hugely successful at large scales. However, to understand bound structures, the perfect fluid model must break down, as different streams of dark matter can intersect, causing the distribution to no longer have single valued velocity. Better understanding the dynamics of dark matter within this multi-stream region is critical to understanding the smaller scale effects of large scale structure. My work focuses on using techniques inspired by the quantum-classical correspondence to model cold dark matter beyond the simple perfect fluid approximation.
Time evolution of a simple 1D model for structure formation. Top shows the full 2+1D phase space evolution, middle shows collisionless particle/fluid trajectories, and bottom shows a wavefunction analogue system (also shown separately). The multi-streaming in the classical models is imprinted in wave interference.
Because of the nonlinearities of gravitational interaction, even Gaussian initial conditions evolve into highly non-Gaussian distributions over time. Finding a set of statistics that are easy to measure, theoretically under control, and concisely capture interesting physics is not trivial. My work focuses on accurately predicting simple choices of statistic, for example the one point function of the density field, and trying to wrestle down their dependence on fundamental physics.
Covariances for 1-point PDFs
The statistics of the cosmic web are highly non-Gaussian, unlike the distribution of matter fluctuations in the early universe (e.g. CMB). Accurately predicting non-Gaussian statistics is crucial for extracting cosmological information from the large scale structure. The PDF of densities in spheres is one particularly simple statistic which already captures much of the non-Gaussian information and can be predicted accurately by theory.