The ΛCDM model accounts for a remarkable range of cosmological observations, yet the physical nature of its two dominant components (dark matter and dark energy) remains unknown, and the rate at which cosmic structure grows offers one of the most direct tests of General Relativity on the largest observable scales. Progress on these questions rests on our ability to measure, with ever greater precision, how matter is distributed throughout the Universe. Spectroscopic galaxy surveys, from BOSS to the ongoing DESI and the incoming Euclid mission, have mapped the three-dimensional positions of millions of galaxies, turning the large-scale structure of the cosmic web into one of the sharpest probes available of both the initial conditions of the Universe and their subsequent gravitational evolution.
Two-point statistics (the correlation function in configuration space, or the power spectrum in Fourier space) fully describe a Gaussian random field, and much of the last two decades of galaxy clustering has been built on them: baryon acoustic oscillations as a standard ruler, redshift-space distortions as a probe of the growth rate. But the observed galaxy field is not Gaussian. Nonlinear gravitational evolution, the nonlinear and non-local relation between galaxies and the underlying matter field, redshift-space distortions, and possibly primordial physics from inflation all transfer cosmological information from the two-point function to higher order, chiefly to the three-point correlation function (3PCF) in configuration space and its Fourier-space counterpart, the bispectrum. Exploiting them jointly with the 2PCF has repeatedly been shown to break parameter degeneracies and tighten cosmological constraints beyond what two-point statistics alone can deliver.
Most full-shape analyses of three-point statistics to date have worked in Fourier space, where efficient estimators and mature perturbation-theory frameworks (from standard and Lagrangian perturbation theory to the effective field theory of large-scale structure) are already well established. Configuration space offers a complementary advantage: its estimators handle the complex geometry of a spectroscopic survey directly, without the need to convolve theoretical predictions with a survey window function. What has long held the 3PCF back is computational cost (direct triplet counting scales as the cube of the number of galaxies) together with the limited availability of accurate, efficient models suitable for a genuine full-shape likelihood analysis.
My research has been built around closing that gap. I developed an analytic next-to-leading-order model for the real-space matter 3PCF, validated it against N-body simulations, and used it for the first full-shape joint analysis of the 2PCF and 3PCF on synthetic catalogues built to match the Euclid spectroscopic sample. More recently, I led the first joint full-shape analysis of the 2PCF and 3PCF multipoles in redshift space on real data, using BOSS DR12, building on a non-perturbative treatment of line-of-sight velocity statistics combined with a one-loop description of the density field. This step had remained missing from configuration-space clustering analyses until now. To make this kind of joint analysis computationally tractable, I also built machine-learning emulators for fast 2PCF+3PCF likelihood exploration. Details and the full list of publications are on the Publications page.