Field-Level Inference and New Observables for the Next Era of Galaxy Surveys

Upcoming galaxy surveys will probe larger volumes and higher redshifts than ever before, bringing within reach physics that has so far eluded cosmological tests — among them late-time dynamical dark energy and primordial non-Gaussianity from inflation. Realizing this potential will require analysis frameworks that go beyond traditional summary statistics.

I will present field-level inference as a principled alternative: a forward-modeling framework in which theory and data are compared directly at the level of the three-dimensional galaxy field, with cosmological parameters, bias and nuisance parameters, and the realization of the initial conditions treated on the same footing. Within this framework, I will then discuss two natural extensions — galaxy shapes and sizes as additional observables, and the multi-tracer technique for suppressing cosmic variance — and how they sharpen constraints on the physics targets above. I will close with an outlook on bringing these methods to real data.

‍