Cosmological constraints from galaxy surveys
A major part of my research has consisted of extracting cosmological information from large-scale structure and weak gravitational lensing measurements. In particular I have used observations from the Dark Energy Survey (DES), a photometric galaxy survey that covers ~5000 sq. deg. of the southern sky and has measured the positions and shapes of over one hundred million galaxies. Specifically, I have been part of the core team that has performed a so-called 3×2pt analysis, which stands for the combination of three two-point correlation functions: one from position-position correlations (galaxy clustering), one from shear-shear correlations (cosmic shear), and one from position-shear correlations (galaxy-galaxy lensing). As Co-Convener of the DES Weak Lensing Working Group since 2021, I was a core member of the team for the flagship analysis of the full six-year dataset, whose results were released in January 2026.
For the first time, DES combined four dark energy probes from a single experiment—3×2pt (weak lensing + galaxy clustering), Type Ia supernovae, BAO, and galaxy clusters—delivering constraints more than twice as constraining as the previous analysis. In the standard ΛCDM model, the combined DES result is consistent with the CMB (Planck+ACT+SPT) at the 2.8σ level. In the wCDM extension, where the dark energy equation of state is allowed to differ from −1, it is found to be consistent with a cosmological constant (w ≈ −1), with a 2.5σ consistency with the CMB.
In May 2026, we extended this analysis to the w0waCDM model, in which the dark energy equation of state is allowed to evolve over cosmic time. Combining all DES probes—3×2pt, Type Ia supernovae, and BAO—we find w0 = −0.84 ± 0.10 and wa = −0.44+0.60−0.55, a 2.2σ deviation from a cosmological constant using data from a single experiment for the first time. Combining further with DESI BAO and CMB data tightens this to w0 = −0.82 ± 0.05 and wa = −0.63+0.21−0.18, a 3.0σ preference for evolving dark energy, adding to the growing evidence that dark energy may not be constant after all. This result was featured in a press release by the Niels Bohr Institute.
Cosmological constraints from the DES Year 6 dataset in ΛCDM (left) and wCDM (right). Shown are results from 3×2pt (pink), SNe Ia + BAO (black/gray), clusters + 3×2pt (brown), all DES probes combined (orange), and CMB from Planck+ACT+SPT (blue). For the first time, DES combined four independent dark energy probes from a single experiment, delivering constraints more than twice as constraining as the previous analysis. The DES combination is consistent with the CMB at the 2.8σ level in ΛCDM and 2.5σ in wCDM.
Constraints on the w0waCDM model from the full six-year DES dataset, showing 68% and 95% credible regions from DES 3×2pt (light blue), 3×2pt + DESI BAO (dotted), all DES probes combined (3×2pt + SNe + DES BAO; red), all DES probes + DESI BAO (blue), and all datasets combined including the CMB (black). The dashed crosshair marks the cosmological constant (w0 = −1, wa = 0), which is disfavored at 2.2σ by DES data alone and at 3.0σ when further combining with DESI BAO and the CMB.
Review article: I also wrote a comprehensive review of weak lensing cosmology, published as a chapter in the Encyclopedia of Astrophysics (Elsevier). It covers the theoretical framework, observational techniques, key systematic effects, and the path from measurements to cosmological inference—written as an accessible entry point for graduate students and researchers new to the field.
Developing new probes and methods
Gravitational lensing ratios: a geometrical probe of dark energy
A big focus of my research is working on gravitational lensing ratios as probes for cosmology. The idea is the following: by taking ratios of different lensing measurements (like galaxy-CMB lensing compared to galaxy-galaxy lensing), we get something that depends purely on geometry—specifically, angular diameter distances. This means we don’t have to worry as much about all the messy astrophysics (like galaxy bias and the matter power spectrum), and we can even use data from smaller scales that we’d normally have to throw away because the astrophysics is too uncertain.
Together with close collaborators, we were among the first to make these measurements work with photometric data. My work includes both CMB lensing ratios (Prat et al. 2019) and galaxy-galaxy lensing ratios (Prat et al. 2018; Sánchez, Prat et al. 2022). It’s been really cool to see other collaborations like KiDS, HSC, and DESI adopt these techniques.
What makes lensing ratios especially interesting is that they are particularly sensitive to spatial curvature and dark energy evolution, and they constrain different parameter combinations than standard probes like baryon acoustic oscillations. With next-generation data from LSST and Simons Observatory coming soon, we’ll be able to make much more precise measurements and really test whether dark energy is evolving over time.
Higher-order statistics and machine learning
The early Universe was very Gaussian, so two-point statistics (like the correlations we measure in 3×2pt) capture most of the information. But the late-time Universe is non-Gaussian, which means there is a lot more information hiding in higher-order statistics. The problem is that these are computationally expensive and hard to model with traditional methods, so most analyses still focus on two-point statistics.
During my time as a Schmidt AI in Science Fellow, I worked on using machine learning to get at this extra information. This led to publishing a persistent homology analysis (a technique from topological data analysis) with DES Y3 weak lensing mass maps, using simulation-based inference. This methodology yields constraints that are 70% tighter than those obtained through traditional cosmic shear two-point analysis.
Combining galaxy surveys and gravitational wave observatories
On very large scales, optical surveys face some challenges: systematic errors from selection effects, and limited statistical power because of finite sky coverage. Interestingly, gravitational wave (GW) sources seem to be less affected by these issues. We have been exploring how much we could gain by combining LSST 3×2pt with large-scale measurements from next-generation GW experiments.
Building for the future: Rubin Observatory and LSST
TXPipe: an end-to-end analysis pipeline
I am a member of the Dark Energy Science Collaboration (DESC) for the Vera Rubin Observatory’s Legacy Survey of Space and Time (LSST), where I have helped develop and validate the pipelines that will be used when the data arrives. I led the development of TXPipe, a modular, end-to-end analysis pipeline designed to produce robust data vectors for 3×2pt cosmology analyses with LSST. TXPipe takes raw catalog data and handles every step through to final data vectors, incorporating systematic tests and map-based diagnostics at each stage. The goal is to make the science reproducible and to set a community standard for how these analyses should be run at Rubin scale.
Galaxy-galaxy lensing at small scales
Extracting useful cosmological information from small-scale two-point measurements is tricky because of non-linearities and baryonic effects. Cosmological analyses often discard scales below a certain threshold—in the DES Y3 3×2pt analysis, this meant throwing out roughly 50% of the available signal-to-noise! Figuring out how to reliably push to smaller scales is one of the most impactful improvements we can make for future analyses.
I have been working on this from several angles: understanding the galaxy-halo connection from galaxy-galaxy lensing, measuring the stellar-to-halo mass relation (SHMR) using a new stellar mass sample for DES Y3 (available publicly here), and publishing a comparison of mitigation methods for small-scale systematics. I also led the galaxy-galaxy lensing measurements for DES Y1 and Y3 (Prat et al. 2018, 2022). Most recently, together with undergraduate student Nathalie Chicoine (now a graduate student at the University of Pittsburgh), we made the first detection of lensing signals around low surface brightness galaxies (Chicoine, Prat et al. 2024).
List of publications
To follow my latest papers, check the following ADS libraries:
