Untangling Particles with Artificial Intelligence

In 2022, after a series of upgrades, CERN's Large Hadron Collider (LHC) is expected to turn back on for a final run and, once again, furiously smash particles together in search of new clues about the fundamental structure of our universe. When the LHC turns on, it will be operating at its highest energies yet: the collider will crash together protons every 25 nanoseconds, leading to the production of hundreds of particles passing through the detector. To help with the deluge of data, scientists are turning to new artificial intelligence (AI) technologies.

We met with Jennifer Ngadiuba, a Robert A. Millikan Postdoctoral Scholar Research Associate in Physics at Caltech, over Zoom to learn more about these developments. Ngadiuba is also supported by the U.S. Department of Energy through Fermilab's Machine Intelligence group. She is Italian-Nigerian and resides in Geneva, Switzerland, near the LHC, where scientists, including team members from Caltech, famously discovered the Higgs boson, a particle that endows all matter with mass, in 2012. Ngadiuba, who has been working at home during the coronavirus pandemic, also recently took part in an online CERN event on diversity in physics.

How did you come to work at CERN?

I knew I wanted to study physics as long ago as high school, and then I went to University of Milano. That is when I discovered particle physics and got excited about CERN. I actually started early at CERN because they have this nice initiative where they take young students from universities and give them a chance to work with senior scientists. After that, I did my PhD at the University of Zurich, and then I worked as a CERN fellow for three years. That was when I started learning about and exploring AI for physics. Now, I'm still at CERN through a postdoctoral program at Caltech. I actually have not been to Caltech yet, due to the coronavirus.

Can you tell us more about your research at CERN?

I got very interested recently in applying modern artificial intelligence methods to enable new physics discoveries. CERN was supposed to start back up next year, but this has been delayed until 2022, due to the coronavirus. When we turn back on, we'll have more energy, going from 13 TeV (terra-electron volts) to 14 TeV. This is not as much of an improvement as we will have later, in the High-Luminosity LHC (HL-LHC) program, scheduled to begin in 2027, in which we will see even more particles produced for every collision. We will have to disentangle these very complex collision events, so we are borrowing tools from the computer science industry, which has enormous potential. What we learn from the 2022 run will teach us how to handle data for HL-LHC.

How will the AI technologies help?

There will be so much data coming in, even in 2022, that we can't store it all. It's just too much. So, what we want to do is discriminate between new rare signals we are interested in and the Standard Model background ones we are more familiar with, and then we want to store the rare signals with high efficiency and purity. For instance, we want to find the Higgs boson signals, but those only show up from time to time. So, one of the things I'm working on is to apply AI technologies in real-time, such that our system can discriminate between what we need and what we don't need. This system has to be ultrafast. Our detector must acquire and send data every 25 nanoseconds, because that is the frequency at which the proton bunches will collide. We have only a few microseconds to decide which events we want to store. And that's tricky because if we decide to throw away something, we will never get those data back, and there could be a signal hidden in the data. The big challenge we are working on now is to replace the simple algorithms used previously with new AI technologies capable of handling the high data rates.

What kinds of signals will you be looking for in this data?

We have a huge range of new physics scenarios we are looking at. We are looking for new particles, hidden dimensions, for evidence of supersymmetry [which predicts that every particle has a partner particle], for hints of dark matter particles [which make up an invisible substance that accounts for 85 percent of matter in the universe], and more. One thing my colleagues and I are working on is a more general search strategy, as opposed to looking for a specific signal. This is how we will find the unexpected. AI can help with this general search approach too.

How have you been spending your time at home during the pandemic?

I mainly work all the time. My husband, who is also a particle physicist, plays guitar as a hobby, so we used to go concerts. We can't do that anymore, but we listen to post rock, avant-garde jazz, and experimental electronic music, and more. I used to play piano and hope to get back to that one day. I also plan to visit my parents soon in Italy.