science
Testing gravity and the dark sector with Euclid, and making sure we can trust the answer.
I want to understand what dark energy, dark matter and gravity really are, using the large-scale structure of the Universe. Euclid and the other Stage IV surveys will measure it with unprecedented precision, and that precision only pays off if our models, our software and our statistics are up to the task. My research sits at the intersection of these three things.
Beyond ΛCDM: model extensions
ΛCDM describes our Universe remarkably well, yet we still do not know what dark energy is, whether general relativity holds on cosmological scales, or what set up the initial conditions. I coordinate the model extensions analysis for Euclid's first cosmology data release (DR1), which tests dark energy, modified gravity and other extensions of the standard model with Euclid's first cosmological data. Before Euclid, I worked on inflation and on new observables such as the cross-correlation of gravitational waves with galaxy clustering.
- Euclid preparation. Review of forecast constraints on dark energy and modified gravity
- CLOE 5. Extensions beyond the standard modelling of theoretical probes and systematic effects
- Euclid preparation. Testing multi-field inflation with galaxy power spectrum and bispectrum
- Illuminating the dark sector: modified gravity signatures with cross-correlations of gravitational waves and large-scale structure
- Learning how to surf: reconstructing the propagation and origin of gravitational waves with Gaussian processes
- Cross-correlation of the astrophysical gravitational-wave background with galaxy clustering
- Current and future constraints on single-field α-attractor models
- Bayesian reconstruction of the inflaton's speed of sound using CMB data
Verification and validation of Euclid's performance
Before Euclid can tell us anything new about the Universe, we need to know that the full chain, from the theory predictions to the final constraints, is correct and that it delivers the performance the mission was designed for. I have spent much of the last decade on exactly that: developing, verifying and validating the Cosmology Likelihood for Observables in Euclid (CLOE), and forecasting what Euclid will achieve by the end of the mission. I led the CLOE inference and forecasts paper, and this work is what led us to build the cloe-org software ecosystem.
- Euclid. I. Overview of the Euclid mission
- CLOE 3. Inference and forecasts lead author
- CLOE 1. Theoretical recipe
- CLOE 2. Code implementation
- CLOE 4. Validation and performance
- CLOE 6. Impact of systematic uncertainties on the cosmological analysis
- cloelib: a flexible Python library for computing cosmological observables in the Euclid era
- cloelike: a Python library for cosmological likelihood inference in the Euclid era
Statistics in cosmology
Cosmology is, at its heart, a statistical science: we only get one Universe to observe. I am interested in how we make robust inferences from it, from Bayesian model comparison and reconstructions to simulation-based inference and interpretable machine learning, and in how we quantify when a result is biased or in tension. My PhD thesis, Playing dice with the universe, was about exactly this, and it is still the thread that runs through my work.
- Playing dice with the universe: Bayesian statistical analyses of cosmological models and new observables
- Merlin: fast and flexible 3x2pt cosmology with simulation-based inference
- Fast likelihood-free inference in the LSS Stage IV era
- Neural networks for cosmological model selection and feature importance using CMB data
- Explaining neural networks on the sky: machine learning interpretability for CMB maps
- Bayesian inference methodology for primordial power spectrum reconstructions from large-scale structure
The full list is on the publications page.