Cosmology and Gravitation

Multiprobe techniques are at the forefront of modern cosmological analysis. In this talk, I will demonstrate the advantages of the multiprobe approach for assessing internal consistency, mitigating systematic effects, and breaking parameter degeneracies across datasets spanning different scales and redshifts. I will showcase these benefits using a comprehensive multiprobe pipeline, developed over a series of analyses, that combines cosmic microwave background (CMB) and large-scale structure (LSS) observations, including all possible cross-correlations between probes. In the second half of the talk, I will present recent results from the pipeline demonstrating how multiprobe analysis can address fundamental questions such as the nature of dark energy and the sum of neutrino masses. Finally, I will look forward to applications with upcoming CMB and LSS surveys. I will highlight how the multiprobe pipeline, combined with modular tools I have developed for effective field theory of large-scale structure (EFTofLSS) modelling and beyond-Limber angular power spectrum computation, positions this work for natural extension to next-generation (Stage-IV) datasets.

The Hubble tension, recently reaching as high as 7σ, increasingly presents a challenge to the ΛCDM model. A key question is whether this tension stems from our use of the sound horizon as a standard ruler. In this talk, I will present a new, sub-2% measurement of the Hubble constant (H₀) that is independent of the sound horizon scale, using data from the first data release of the Dark Energy Spectroscopic Instrument (DESI). The analysis employs a power spectrum rescaling technique that marginalizes over the sound horizon information, drawing instead on the matter-radiation equality scale as a standard ruler. Combining DESI’s full-shape galaxy clustering with uncalibrated post-reconstruction BAO and the CMB acoustic scale θ*, along with various external sound horizon-free datasets, results in highly robust constraints that are the most precise to date from LSS. Looking ahead, this measurement offers a new window on beyond-ΛCDM physics and the physical origin of the Hubble tension.

Maximizing our chances of discovering new physics from next-generation cosmological surveys requires making robust theoretical predictions of the nonlinear processes governing both the early Universe and the late-time large-scale structure. In this talk, I will present advances in two complementary methods for constraining the physics of the early Universe with cosmological survey data: forward simulation from inflation to late-time observables, and machine-learning-assisted Bayesian reconstruction of primordial initial conditions. In the first approach, we study the early Universe by directly simulating its dynamics. I have developed the first code for simulating multifield inflationary theories on a discrete lattice with enough precision to robustly predict primordial non-Gaussianity in higher-order N-point correlation functions. Focusing on axion-U(1) inflation, we accurately characterize the rich phenomenology predicted by this model, including modifications to the power spectrum, bispectrum, and higher-order correlation functions that violate parity. Using the output of these inflation simulations as initial conditions for N-body simulations, we effectively simulate the entire history of the Universe from inflation to late-time large-scale structure, finally revealing what inflation actually predicts about the origin of structures in the Universe. In the second approach, we reconstruct the initial conditions of the Universe using Bayesian inference with a fast, accurate, differentiable model of large-scale structure formation. I developed the first field-level emulator for large-scale structure, training a convolutional neural network to map linear initial conditions directly to the late-time matter field as modelled by computationally expensive N-body simulations. The model accelerates predictions by a factor of 1000 over full N-body simulations and achieves percent-level accuracy at deeply nonlinear scales (1 Mpc), outperforming fast particle-mesh simulations. This field-level emulator opens the possibility of reconstructing the initial conditions of the Universe using Bayesian inference techniques that include information from late-time nonlinear scales, as I will demonstrate with an application to simulated data. This highly nontrivial task involved exploring a million-dimensional parameter space using Hamiltonian Monte Carlo sampling and was only possible due to the combined speed, accuracy, and differentiability of the emulator.

Primordial non-Gaussianities (PNGs) are imprints in the density field caused by particle interactions during the inflationary epoch. A subclass of these signatures ("cosmological colliders") encode information about the interactions and properties of any additional particle fields present during inflation. I introduce a novel method for producing cosmological simulations with arbitrary collider signals, and present signatures from over thirty collider models in the deeply non-linear regime of the density field. I show how weak lensing surveys can offer competitive constraints that are complementary to other probes of PNGs in the late Universe. This simulation framework expands the probes of nonlinear structure (beyond lensing) that are usable in constraining primordial particle physics, which in turn expands the space of testable models and enhances the precision with which we constrain them.

Lambda Cold Dark Matter (LCDM) is the most successful theory for the formation of structure in the Universe. Although its predictions have been verified on large scales, they are still contested on the scale of dwarf galaxies, whose dynamical properties are often cited as evidence for the need to revise some of LCDM’s basic tenets. In this context, I will discuss the recent discovery of the faintest galaxies known to date, and how their properties may be used to place constraints on the clustering of dark matter on the smallest galactic and sub-galactic scales, as well as on the viability of some of the proposed alternatives to LCDM.

Modern telescopes such as Hubble and JWST deliver exquisitely detailed images that encode rich information about galaxies, stars, and dark matter. Extracting this information requires solving high-dimensional inverse problems in which each pixel represents a parameter of the sky. In this talk, I will present a Bayesian framework that uses diffusion-based generative models to perform inference directly in these high-dimensional spaces. By learning expressive priors over galaxy morphologies and likelihoods over complex, non-Gaussian noise distributions, this approach enables the reconstruction of astronomical images at unprecedented fidelity. This talk will cover a wide range of applications where we have applied these techniques, from the recovery of detailed images of protoplanetary discs using ALMA data and high redshift galaxies using Hubble data and strong gravitational lensing to a discussion about the misscalibration of the generative model. The generative model often needs to be trained on low redshift galaxies or using simulations, all of which could introduce biases in our inference. To address this, I will present a framework which can update the parameters of the generative model using corrupted data only. Together, these results illustrate how generative modeling can be used in a principled statistical framework for astronomy.

Galaxy science and cosmology are rapidly extending to ever higher redshifts, where new observations have motivated a wide range of galaxy formation models. Because these models are all tuned to reproduce observed number densities, clustering provides the critical observable for breaking their degeneracies. Pure-parallel surveys, often underexploited, enable clustering to be measured through field-to-field variance across a broad redshift range. I will show measurements from the JWST PANORAMIC survey, spanning quiescent galaxies at intermediate redshift to luminous Lyman-break galaxies above z∼10. Linear-bias results from COSMOS-Web up to z∼8 reveal a consistent picture: galaxies in the early universe are highly clustered, placing strong constraints on models of early formation. Remarkably, even carefully tuned models struggle to reproduce the observed amplitudes. Cosmic variance also emerges as both a systematic and a tool in number-count statistics, with particular relevance for explaining the presence of ultra-massive quiescent galaxies already in the young universe. I will also provide a brief overview of complementary low-redshift work, including the use of small-scale clustering to reconstruct halo merger trees with graph neural networks, and other efforts to maximize the scientific return of the next generation of galaxy surveys.

Galaxy formation in the first billion years marks a time of great upheaval in our cosmic history: the first sources of light in the Universe, these galaxies ended the 'cosmic dark ages' and produced the first photons that could break apart the hydrogen atoms suffusing all of space starting the process of 'cosmic reionization'. The past few years have seen cutting-edge instruments such as the James Webb Space Telescope (JWST) provide tantalising glimpses of such galaxies assembling in an infant Universe. Puzzlingly, these observations are also yielding a sample of unexpectedly numerous and large black holes (up to a 100 million solar masses) within the first 600 million years, posing an enormous challenge for galaxy formation models. I will show how this data is providing an unprecedented opportunity to pin down the reionization state of the Universe in addition to providing an unrivalled resource for understanding the reionization topology in the forthcoming era of 21cm cosmology. I will also show how these early systems provide a powerful testbed for Dark Matter models beyond "Cold Dark Matter". Finally, I will try to give a flavour of the gravitational wave event rates expected from such early black holes in the Laser Interferometer Space Antenna Array (LISA) era.

  Supermassive black holes (SMBHs) exist in many galaxies. Their growth is accompanied by strong energy output, which is capable of regulating host galaxy evolution. Understanding SMBH growth is thus critical for studying galaxy formation and evolution. However, it has been difficult to quantify SMBH growth in different galaxies and cosmic epochs. In this talk, I will present Trinity, an empirical technique to determine the typical SMBH mass and growth rate in different galaxies and dark matter halos from z=0-10. I will discuss how the galaxy—SMBH connection from Trinity will help observational astronomers extract more information from data, as well as theoretical astronomers create better simulations of galaxy evolution. In addition, I will give an overview of the Trinity predictions that match latest JWST data, as well as those that do not match observations. Finally, I will talk about how observations with next-generation telescopes will enable a better understanding of the galaxy—SMBH connection, and how Trinity will be helpful for quantifying the constraining power of future observations.