Over the project’s duration, the team made significant advances in dark-matter signal modelling, data-analysis methodology, and astrophysical searches.
A major focus lay on axion searches using neutron stars. The team developed the first end-to-end modelling framework for axion–photon conversion in realistic magnetospheres, capturing anisotropic emission, line broadening, and time-dependent signatures. Guided by these models, they performed targeted searches in archival Green Bank Telescope data and obtained the strongest constraints to date on axion dark matter in parts of the micro-electron-volt range.
A key breakthrough was the discovery that axions produced near neutron stars can naturally form long-lived gravitationally bound “axion clouds”, which may later reconvert into brief radio bursts as the star spins down. This unexpected prediction opened an entirely new search channel and attracted broad interest.
In parallel, the project advanced small-scale probes of dark matter using strong gravitational lensing and stellar streams. It pioneered neural-network–based methods to infer the warm-dark-matter halo-mass cutoff from ensembles of lensing images, and developed a dynamical inference framework for stellar streams in the Milky Way. These tools strengthen our ability to test the cold-dark-matter paradigm on subgalactic scales.
The project also delivered substantial progress in statistical methodology, most notably the development of Truncated Marginal Neural Ratio Estimation (TMNRE) and the open-source analysis package Swyft. These simulation-based inference methods enabled rigorous parameter estimation in high-dimensional problems previously out of reach, with applications ranging from radio searches and gravitational lensing to 21-cm cosmology, cosmological parameter inference, and stochastic gravitational-wave background analyses. Together, these developments significantly expand the scientific toolkit for post-WIMP dark-matter searches and establish new directions for future discovery.
