Available student projects

This page is a non-exhaustive list of possible projects for graduate students and research assistants to work on.  Some of these can lead to multiple theses and all are expected to lead to publishable work.

First tests of a Large P-type Point Contact (LPPC) detector

Our group will soon (Spring 2019) be receiving a very large prototype point contact detector, the largest of its kind in the world. Increasing the size of these detectors can allow backgrounds in rare-event searches to be decreased substantially as one requires less cables and electronics to read out fewer detectors. We are particularly interested in using this detector as a test bench to understand:

  • The limitations in how large these detectors can be made.
  • How charges drift inside a germanium detector.
  • How temperature affects the drift and trapping of charges
  • How precisely one can determine the position of interactions in the detector.
  • How well one can model the signals in the detector.
  • Whether the ability for these detectors to reject backgrounds is affected by their size.

Understanding surface events in point contact detectors

In previous work, we showed that the region near the outer electrode of point contact detectors forms a "transition layer" that can lead to the measured energy of events being significantly degraded, to the point of mimicking the signal that one expects from light WIMP dark matter interacting in the detector. We are interested in furthering our understanding of the transition layer, in particular, we wish to:

  • Develop efficient methods to measure the size of the transition layer.
  • Develop a model of signals that arise in the transition layer.
  • Develop a model for the expected energy spectrum of events from the transition layer.

Participating in data analysis from rare-event searches

Our group is involved in several experiments at SNOLAB and elsewhere in the world that are searching for dark matter and furthering our understanding of neutrinos. Participating in the analysis of the physics data from these experiments is usually done in conjunction with other technical work which is usually considered "service" for the experiments (the other projects listed here can, for example, be considered as service). Usually, the physics analysis topic is left as a choice to the student based on their interest and their expertise. For example, it would make sense for a student that is working on Machine Learning tools to use those tools to extract a physics signal, or for a student that is studying the transition layer in point contact detectors to search for low-energy signals such as WIMPs in the Majorana Demonstrator experiment. We can provide supervision on topics in the following experiments:

Applying machine learning techniques to NEWS-G

Data from the NEWS-G experiment consist primarily in "pulses" (charge as a function of time) that contain information about the energy deposited in a spherical proportional counter. Those pulses require substantial processing in order to recover the original physics that led to their formation. By using machine learning, we plan to improve our treatment of these pulses and extract a maximum of information. We are particularly interested to develop tools that:

  • Can remove electronic noise, ideally using unsupervised machine learning.
  • Determine the energy deposited in the detector.
  • Identify events that are non-physical.
  • Identify the radial position of the events.

Applying machine learning techniques to point contact detectors (Majorana Demonstrator/LEGEND)

Data from the Majorana Demonstrator experiment consist primarily in "pulses" (charge as a function of time) that contain information about the energy deposited in a one of the germanium detectors. Broadly, the issue of treating these pulses is very similar to treating those from the NEWS-G experiment (described above), even if the physics is very different. Thus, we want to develop a similar set of tools (if not a set of tools to work for both experiments), that

  • Can remove electronic noise, ideally using unsupervised machine learning.
  • Determine the energy deposited in the detector.
  • Identify events that are non-physical.
  • Identify the position of the events.

Applying machine learning techniques to SNO+

Data from the SNO+ detector consist primarily in a collection of the time and charge of the "hits" on the 10,000 photomultiplier tubes (PMTs) that instrument the detector. In any one "event", a large fraction of the PMTs are typically not triggered, creating data sets that are very sparse and for which current machine learning algorithms are not well equipped to handle. However, some promising work from our group shows that one can indeed construct algorithms (e.g. neural networks) to deal with the sparsity of the data.  We are also interested in using machine learning tools (e.g GAN) to simulate events in SNO+ that will supplement the official Monte Carlo data. We are thus interested to develop a common set of tools for SNO+ to:

  • Use PMT hits to determine position, direction  and energy of event.
  • Use PMT hits to identify signal and background events.
  • Use PMT hits to reject instrumental backgrounds (non-physical events).
  • Develop a GAN type of network to create "fake" Monte Carlo data.