fairfolio
Saved Projects

Accelerating LHC Exotic Physics Discovery via Optimal Transport Flow Matching Calorimeter Simulation

CSEF · 2026 Physics & Astronomy (Senior Division)

Raunav Mendiratta

Overview

The Large Hadron Collider (LHC) at CERN smashes protons together at nearly the speed of light to search for particles never seen before, potential keys to understanding dark matter, extra dimensions, and why the universe exists at all. When these collisions produce exotic particles, they leave unique fingerprints as cascading "showers" of energy inside the detector's calorimeter. To recognize these rare signals hidden among trillions of ordinary collisions, physicists must simulate billions of showers using Geant4, a program that painstakingly models every particle interaction from first principles. The problem is that Geant4 is extraordinarily slow, and the upcoming High-Luminosity LHC upgrade will increase collision rates tenfold, creating a simulation crisis that could delay the next major discovery in fundamental physics. This project presents a deep learning solution that generates realistic calorimeter showers approximately 3,500 times faster than Geant4. The model uses Optimal Transport Conditional Flow Matching (OT-CFM), a technique that learns to transform random noise into physically accurate energy patterns by computing the most efficient transformation paths through high-dimensional space. A two-stage pipeline first predicts how energy spreads across 45 detector layers using a Transformer network, then fills in the complete 6,480-voxel shower shape using a Vision Transformer tailored to the detector's cylindrical geometry. The Sinkhorn optimal transport algorithm pairs noise samples with real showers to produce straighter generation trajectories, cutting the required computation steps in half. Tested on the international CaloChallenge benchmark for electrons spanning three orders of magnitude in energy (1 GeV to 1 TeV), the model produces showers that classifier networks cannot reliably distinguish from genuine Geant4 simulations (AUC = 0.519, where 0.500 represents perfect indistinguishability). At 17 milliseconds per shower and twice the speed of the previous leading method, this work demonstrates that AI-driven surrogate simulation can meet the computational demands of the High-Luminosity LHC era, removing a critical bottleneck on the path to discovering new physics beyond the Standard Model.

Competition history

  • CSEF 2026 Physics & Astronomy (Senior Division) · Entry S-17-20

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

Source: California Science & Engineering Fair public projects

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google