15–19 Jun 2026
Dipartimento di Fisica G. Occhialini, Università Degli Studi di Milano-Bicocca
Europe/Zurich timezone

Numerical polology: next-generation model-building for precision cosmology

16 Jun 2026, 15:20
20m
U3-01

U3-01

Oral contribution Topical Session 1 - Astro

Speaker

Dr Will Barker (Central European Institute for Cosmology and Fundamental Physics, Institute of Physics of the Czech Academy of Sciences)

Description

Particle dark matter, along with many ultraviolet scenarios, suggest that additional low-energy degrees of freedom remain to be discovered. Theories of new physics may be understood as statistical models, for which the Lagrangian couplings are model parameters. The net worth of a theory is determined by its Bayesian evidence: the likelihood of precision cosmology data is multiplied by the prior probability of the couplings, and integrated over the coupling-space. Precision cosmology has made great advances both in the collection of data and the efficient computation of likelihoods. But whilst the complementary programme of manufacturing candidate models is very active, it is far less systematic, and priors are seldom specified.

We present a framework for massively automating the construction of new physics models, designed to scale with the ever-increasing volume of data. Numerical polology uses nested sampling to identify unitary and technically natural regions in coupling-space. Such models form self-consistent effective field theories, which is essential since the predictivity of a model (the ability to compute a likelihood) is endowed by the systematics of QFT alone. The framework is adapted to bosonic theories of the dark sector: the phenomenological implications of arbitrary field content (field number, rank and index-symmetry) can be systematically explored with recourse to tools such as GetDist and Cobaya. The framework is inspired by the SOFTSUSY/SARAH tools for supersymmetric model-building, and builds directly on the PSALTer software for modified gravity. The latter is computer algebra software, which scales badly with complexity: numerical polology overcomes this technical hurdle and facilitates data-driven model-building.

We illustrate numerical polology with a Stueckelberg extension of massive gravity, and derive simple posterior reweightings from black hole superradiance, large scale structure, pulsar timing data and gravitational wave dispersion. We also discuss realistic prospects for more sophisticated likelihood plugins. We then perform a high-resolution survey of the coupling-space of symmetric rank-two fields, using high-performance computing. We discuss numerical challenges, and the benefits of migrating the present Julia implementation to JAX.

Parallel session Astrophysical Probes of Dark Matter and Dark Energy

Author

Dr Will Barker (Central European Institute for Cosmology and Fundamental Physics, Institute of Physics of the Czech Academy of Sciences)

Co-authors

Presentation materials

There are no materials yet.