Fig. 1: Plot of a point cloud and the associated log-likelihood at each point. The log-likelihood represents a score that indicates how likely it is for an event
In this project, the XENONnT detector was calibrated using a Y-Be neutron source. This calibration uses -emission from Y to induce Be to emit neutrons. neutrons look similar to B solar neutrinos, hence allowing us to calibrate the detector response for such a measurement. This calibration is challenging as the energy of these events is extremely low, and many of the gamma rays actually escape the source instead of generating neutrons, presenting a high rate of events that are not useful for our calibration. The solution is thus to use a dense material to construct a gamma shield. Neutrons scattering off of such a heavy nucleus, such as tungsten or lead, loses very little energy, making such materials ideal for a selective shield that would block gamma rays but leave neutrons relatively unaffected.
Due to the physical constraints of upgrading XENON1T in-place to XENONnT, this calibration source, including shielding, is confined to a cube. Thus, the calibration system needs to be designed to make optimal use of the compact space. Thus, the geometry of the source assembly containing the gamma source and the beryllium is optimised using detailed simulations of the neutron source and associated gamma event rate in the XENONnT detector. The ratio of useful neutron events to gamma events is used as a figure-of-merit to determine the performance of various possible designs, allowing for a final optimised design to be selected.
The Windchime project aims to use mechanical accelerometers to detect dark matter via the gravitational coupling alone. This is very exciting for two reasons. First, the gravitational interaction is the only interaction that Dark Matter is known to have, as dark matter is currently only observed indirectly via its gravitational effects. Second, the mass region around and above the Planck mass is a natural mass region to look for Dark Matter, as new physics is expected around the Planck mass scale, and there are many theoretical models for heavy dark matter around this mass scale. However, as the gravitational force is the weakest of the forces, this is expected to be extremely hard. Thus, the Windchime project is a project with a long time horizon, where we expect to have multiple generations of the experiment before one can gain sensitivity to dark matter via the gravitational interaction.
As part of my PhD, I worked on analysis methods for uncovering a track formed by long-range interactions in a large array of sensors, where the signal is below the noise floor of the array if one looks at the total signal. Despite this, I showed that it is possible to recover the signal using both template matching and Bayesian fitting using nested sampling with ultranest. A demonstration of this using template matching is shown in Fig 2.
I am also working on the estimation the look-elsewhere effect correction for such a computationally difficult problem. The look-elsewhere effect refers to the fact that the more independent experiments one does, the more often one would expect to get a statistically "surprising" result by chance. While it is easy to correct for this when one is doing multiple independent experiments, searching for a track that can come from any direction over a long period of time also results in many experiments that are effectively independent! For example, for tracks that are a few microseconds long, one would expect tracks from today and data from tomorrow to be essentially truly independent. However, the actual number of effectively independent trials, known as the trial factor, is not always easy to estimate – while the result from a search of dark matter tracks days apart is obviously completely independent, what about microseconds apart, where the times of the tracks overlap? While this seems like a technical detail, an estimate of the trial factor is necessary to determine whether any signal is a statistical fluctuation, or is actually dark matter. Finally, I am working on a framework to estimate the sensitivity of Windchime accelerometer arrays with different quantum-sensing methods and array sizes. This would allow us to determine the sensitivity of large sensor arrays efficiently, and inform both the design of a first-generation Windchime experiment, and the R&D direction for the longer-term project.
Fig. 2: 2D slices of a significance map obtained with simulated data. Even though the signal is buried below the noise for single sensors in this simulated dataset, a clear peak around the expected parameter values of a simulated dark matter track can be seen.