In conversation with Richard Qiu
Hi Richard. Good to have you here. So can you tell us a little bit about yourself ?
Sure. My name is Richard Qiu. I just finished my second year as an undergraduate at Harvard. I’m studying physics and computer science, as well as minoring in maths. That’s probably gonna change to a minor in computer science but that’s a different story.
Great, so how many research projects have you been involved in so far?
Quite a few actually. So in High school I did this program where they match high-schoolers with research mentors at NASA. That was my first real research experience. There I was working on computational material science trying to deduce properties of both metals and composite materials using a combination of data and computing algorithms.
From there my interest diverged into biology. I spent about a year and a half working in a biology lab, studying diabetes, particularly focused on blood vessel development. I noticed that I don’t really care about the hands-on aspect of the biology lab. What I really liked was the computational part of it, analysing data etc.
So in college I switched to statistics. So in my freshmen year I did research in epidemiology, because I was still kind of interested in biology. That research went no-where.
And then finally I switched to where I currently am – Physics and computer science. I did a project under a professor in Harvard, in the beginning of my second year.
What can you tell us about the most exciting project you’ve been part of?
Well amongst the ones mentioned, I’d say the High Energy Physics was the most exciting. I did it under Prof. Matthew Schwartz at Harvard. So my project involved using machine learning to study the collision of quarks and to compare observational data and match it with theoretical data.
So high energy physics is basically particle physics. The idea behind the field as a whole is to break down the universe at the fundamental particle level. In school we learned about molecules being made of atoms, and atoms made of protons and neutrons. High energy physics breaks it down even further, and studies quarks, which makes up the protons and neutrons.
So we now know that quarks and gluons are the most fundamental particles, meaning we can’t break them down any further. The way we probe these interactions between quarks and gluons and the way we try to find these new particles is through particle accelerators.
So how exactly do you do that?
We accelerate particles at speeds faster than 99.99% of the speed of light and smack them into each other. That’s why it’s called high energy. The reason this works is because when particles have a lot of energy, the fundamental interactions change, and we look for these kinds of interactions. So we put two particles in a magnetic field, spin them really fast and line them up with each other forcing them to collide.
At the point of collision, we have detectors all around, which picks up the properties of the quarks. It’s like picking up the pieces after a really bad car accident, and trying to figure out what kind of car it was. The outcome of these collisions are called jets. Recently, there has been a lot of work to understand the behaviour of these jets, predicted by derivatives of quantum field theory. So when we smash two quarks together, we notice a bunch of 30 -100 particles flying outwards. We want to know how two colliding particles can produce so many particles.
These particles, how do you detect them?
To put it simply, let’s say we have a bubble chamber, which is a huge vat of liquid hydrogen. When particles move through this bubble chamber, they leave a trail of bubbles through the hydrogen. The detectors used are extremely sophisticated versions of the bubble chambers.
The specific thing I was working with is called a CMS – Compact Muon Solenoid. What’s important is the Solenoid. The way it detects particles is that it creates a superconducting magnetic field. The particles that come out are charged, and hence it looks for perturbations in the magnetic field. From there, we can back-calculate the properties of each particle that comes out.
How exactly were you involved in the research project?
I was more involved with designing and training the machine learning algorithm. So my day-to-day consisted of more hands on coding. The specific algorithm that we’re learning from using ML algorithms isn’t the quantum field theory per say, it’s the decomposition theory of these particle collisions, the jet decomposition.
It processes this jet data, which can come from the simulation as well as the experimental data. We use ML to understand the landscape of these jets. We have quarks which collide and produce more particles. These collisions are fairly well understood. The simulations are also based on fairly well understood quantum field theory. Of course there are certain discrepancies, but they aren’t in question in the algorithm we are concerned with.
So we can break down these jets into splitting trees, where one particle disintegrates into a bunch of other particles. So our machine learning algorithm uses something called a Recurrent Neural Network to analyze the jet data and, to put it very naively, calculated the probability of these splittings.
What exactly are the skills/technical requirements to pursue such a project?
I don’t think a theoretical background is necessary to enjoy research. If you enter with a mindset that you need to know all about the theory behind a project before doing it, you’ll never really do the project. The most important thing, honestly, is just computational skills.
These types of projects mostly involve simulations and calculations. Of course theoretical knowledge would be helpful, but the specific theoretical requirements for such a project tend to be fairly narrow. I definitely didn’t know as much as I do now before entering the project.
The machine learning part of this project happened in python, and the pre-processing of the data was done in C++. My knowledge of machine learning has also greatly increased since then.
I’m sure this was enlightening for our readers as well. Thank you for talking to us.
Thank you for having me :)
Further Resources:
How Artificial Intelligence can supercharge the search for new particles
.svg)
.png)