Improving High-Energy Particle Detectors with Machine Learning

Microseconds after the big bang, the universe existed in a state called the quark-gluon plasma (QGP). To experimentally study its properties, the QGP is recreated in high-energy nuclear collisions at the LHC, and the particles produced from the QGP are reconstructed from their energy deposition in the ATLAS calorimeter. This requires both classifying the particles and calibrating their deposited energy. The objective of this project is to improve the reconstruction by using machine learning techniques, where the energy depositions of clusters of cells, formed by ATLAS topo-clustering methods, are treated as three-dimensional images when inputted to neural networks. This approach significantly improves the calibration of deposited energies when cross-validating while training, and models trained on idealized data predict the calibrated energies of particles in more complex data sets well. Additionally, implementation of a data generator using uproot allows the program to load input data into memory as needed while training or predicting, significantly reducing the amount of memory used. The data generator also allows for use of multiprocessing to speed up training and evaluating. This work illustrates that using machine learning methods for both classification and calibration has the potential to significantly improve particle reconstruction.

I worked on this research project at Lawrence Livermore National Laboratory (LLNL) under Aaron Angerami from June 2020 to August 2020. At the end of my internship, I presented the results of my research at the Summer SLAM! (abstract and presentation). I also wrote a research report to culminate my internship.

New resummation techniques of divergent series: the Painlevé equation PII

I worked on this research project at The Ohio State University under Professor Ovidiu Costin from February 2018 to March 2019 and from August 2019 to May 2020. The research was on advanced methods for resummation of divergent series to convergent solutions for differential equations. When resumming a divergent series numerically, one generically has to deal with limited information due to having computed only a finite number of terms. To deal with this, we used novel approximation methods that are more precise than the standard Padé approximation. Additionally, Professor Costin developed a new resummation method which leads to rapidly convergent uniform rational expansions for solutions to differential equations. This method relies on resurgence theory, a new area in analysis with many applications to different areas of physics, such as quantum field theory. I applied this method to the tritronquée solutions of Painlevé equation PII. Our results are now being written up for publication, and I presented the ongoing results at the 2018 Ohio State Autumn Undergraduate Research Festival, the 2019 Ohio State Denman Undergraduate Research Forum (Denman poster), and the 2019 Young Mathematicians Conference at Ohio State (YMC abstract and YMC poster). Finally, I successfully completed and defended my undergraduate thesis on this research in April 2020 (thesis and defense slides).

Smoothed Particle Hydrodynamics for hydrodynamic fluctuations

I worked on this research project at Wayne State University under Professor Chun Shen from May 2019 to July 2019. During my 10 weeks at WSU, I wrote an open source code package in C++ with C++ 11 standard that used a Smoothed Particle Hydrodynamics (SPH) algorithm from astrophysics to solve the hydrodynamic evolution equations (a set of non-linear partial differential equations (PDEs)) for relativistic fluids created in high-energy collisions between atomic nuclei. I also wrote a report on my findings.

I worked on this research project at The Ohio State University under Professor Sergei Chmutov from June 2017 to December 2017. In this research project, I worked with two other undergraduates to try to develop a novel knot invariant that would be able to distinguish more knots than current invariants on virtual knots. Together, we developed a program to output different knot invariants for any inputted virtual knot to assist us in testing out our potential knot invariants in comparison to the existing ones. A description of the project and program can be found here.

I worked on this project at The Ohio State University under Professor Robert Perry from June 2016 to December 2016. I independently studied various problems in quantum mechanics and discussed my findings with Professor Perry. I also attended the meetings of the Low-Energy Nuclear Theory group at Ohio State, allowing me to experience many different people’s perspectives on various research projects and how a large research group functioned.