Projet CRAB : détection et simulation précises des reculs nucléaires pour la calibration des détecteurs cryogéniques à l'échelle de 100eV

Cryogenic detectors have reached energy thresholds of a few tens of electronvolts, for detector masses on the order of a gram. These characteristics make them valuable technologies for detecting low-energy nuclear recoils induced by coherent elastic scattering of reactor neutrinos or for the direct detection of light dark matter. A detailed understanding of the detector response to sub-keV nuclear recoils will however be crucial for interpreting future experimental data. The CRAB (Calibrated Recoils for Accurate Bolometry) project proposes using radiative thermal neutron capture to induce mono-energetic nuclear recoils in the 100 eV to 1keV range, which can be leveraged for calibration purposes.Following thermal neutron capture, the formed compound nucleus may de-excite by emitting a single gamma-ray. The resulting nuclear recoil is perfectly determined given the nuclear mass and the energy of the emitted gamma-ray, which is characteristic of the isotope. This work presents a first experiment with a portable thermal neutron source which led to the observation of a nuclear recoil peak at 112 eV, corresponding to neutron capture on ¹⁸²W in a CaWO₄ detector. The statistical significance of the signal is found to be 3σ. The statistical analysis proves also that the measured excess of events above the external background is compatible with the simulated CRAB energy spectrum, with an associated significance greater than 6σ. These results experimentally validate the principle of CRAB as an in-situ non-invasive calibration method. A high-precision experimental campaign is currently being prepared at a thermal neutron beamline of the Atominstitut in Vienna, with the goal of unlocking the full scientific potential of the method. In particular, some solid-state physics effects should be accounted for when predicting the expected energy spectra. In general, nuclear de-excitation tends to occur rather through the sequential emission of several gamma-rays. The corresponding recoil energy therefore depends on the energy and direction of the emitted gamma-rays, as well as on the time evolution of the gamma cascade relative to the stopping time of the recoil in matter. We have simulated this effect using Iradina to model, within the binary collision approximation, the atomic collisions induced by the initial recoil in the surrounding matter. This code has been modified to intertwine the displacement cascade with the gamma cascade that triggered the intial recoil, which is predicted by Fifrelin. For the four most common materials used in low-threshold cryogenic detectors (Si, Ge, Al₂O₃, CaWO₄), this approach shows that additional structures are expected in the CRAB energy spectra, providing new calibration anchors. In CaWO₄ , CRAB recoils are below 200eV, thus falling within the region of displacement energy thresholds. Whenever crystal defects subsist at the end of the cascade, a fraction of the initial recoil energy is retained in these defects and escapes detection. To quantify this effect in CaWO₄, we have simulated a large number of displacement cascades using molecular dynamics, for various initial recoil energies and directions of a tungsten atom. These simulations required the prior development of a Machine Learning interatomic potential, using MiLaDy, which was trained on well-chosen DFT calculations performed with VASP. A preliminary assessment of the impact of the energy stored in defects on the CRAB energy spectrum suggests that the latter could serve as an original experimental probe of this effect. Similarly, the energy spectra of dark matter and coherent elastic scattering of neutrinos will be affected, with expected shape distortions on the order of 10%. Given the potential degeneracy with new physics signatures, the energy stored in defects will need to be accounted for in the precise interpretation of future experimental data.