New city, new project, and a fresh collection of things I do not understand yet. Excellent conditions for a research degree.
I started my MPhil by research in Physics at the University of Cambridge in October 2026, working with Professor Oleg Brandt on the ATLAS experiment. My project explores machine-learning approaches to reconstructing displaced vertices in the ATLAS muon spectrometer, supporting searches for long-lived particles.
That is quite a sentence. Let’s unpack it.
What am I working on?
Some proposed new particles could travel a measurable distance from a proton–proton collision before decaying. Their decay products would then appear to originate somewhere away from the original collision point—a displaced vertex.
Reconstructing that point means working backwards from the traces left in the detector. In the ATLAS muon spectrometer, this can involve identifying and grouping short reconstructed track segments, called tracklets, and using their geometry to estimate where the decay happened.
The question I’m investigating is whether machine learning could help associate the right tracklets with the right vertex. First comes understanding the existing reconstruction and its difficulties; then comes testing whether a learned approach actually helps.
Adding a neural network is easy to suggest. Establishing that it improves anything requires rather more work.
For a broader introduction to the physics, this ATLAS explainer is a useful starting point.
What will you find here?
Research diary
Progress updates, meetings, practical obstacles, and the reasoning behind the choices I make along the way.
The physics
Long-lived particles, displaced decays, detector signatures, and why the muon spectrometer is an interesting place to look.
Reconstruction and machine learning
How detector information becomes vertex candidates, where the existing methods struggle, and the ML ideas I’m investigating.
Papers, tools, and things I had to learn
Reading notes and explanations of the software and concepts behind the project, starting with the basics where necessary.
My UCL project focused on boosted Higgs production at a future collider. Here, the focus shifts to reconstructing unusual decays inside an existing experiment.
This section will follow that learning process as it happens: the useful results, the revisions, and the occasions when understanding the problem turns out to be most of the work.
The plan is to document it before my browser tabs become its only surviving record.
