Researcher of the Month
April 2024
Daniel Julian
Majors: Physics, Applied Mathematics; Class of 2025
Research Mentor: Dr. Jesús Pérez-Ríos, Physics & Astronomy
“So even after 10 weeks, it is good to have a poster and to be able to explain it
to people. That's why I liked the summer symposium so much. You have something to
show: Here's some proof. Here's my poster. This is what I did. This is what I learned
from doing this experience.”
– Daniel Julian, Class of ‘25
Daniel Julian is an Honors College junior double majoring in physics and applied mathematics. Undergraduate research and science communication have both played a large part in his undergraduate education. Since January 2023, Daniel has been working under the mentorship of Dr. Jesús Pérez-Ríos on using machine learning for atomic and molecular (AMO) physics, and was listed as a co-first author for his contributions to a Journal of Molecular Spectroscopy article titled: “The database of spectroscopic constants of diatomic molecules (DSCDM): a dynamic and user-friendly interface for molecular physics and spectroscopy.” Daniel’s full time research in summer 2023 was supported through the Explorations in STEM program, a URECA-Career Center undergraduate research program (PI: Dr. Monica Bugallo). In the previous summer, Daniel was introduced to machine learning tools as an intern in the SULI internship program at BNL where he worked with Dr. Garth Williams at the National Synchrotron Light Source II on using neural networks to predict the number of grains contributing to the diffraction pattern of a polycrystal. Daniel’s long-term goals are to pursue graduate studies in the field of theoretical physics.
In summer 2022, Daniel presented a poster at the annual Summer Symposium on “Learning chemical reactions: Neural Networks for predicting final states of an atom-diatom collision.” In addition, he has presented at the spring Society of Physics Students inaugural poster day event (March 2023); at the URECA spring 2023 Undergraduate Research Celebration (May 2023); and at the 54th Annual Meeting of the APS Division of Atomic, Molecular and Optical Physics (DAMOP) in Spokane, Washington (June 2023). He was among a group of physics students who volunteered to practice science communication skills through fun, hands on physics exhibits at the CommUniversity event (October 2023); and recently presented a poster at the Physics Undergraduate Symposium (March 2024). Be sure to look for his poster at the upcoming URECA Celebration on April 30th!
On campus, Daniel has been involved as a tutor for the EOP program and has been recognized with the Outstanding Academic Achievement Award. He is also an accomplished jazz saxophonist who has played at multiple venues on and off campus. Daniel is a graduate of Shoreham-Wading River HS and during high school participated in BNL science outreach programs. Below are excerpts of his interview with Karen Kernan, URECA Director.
The Interview:
Karen: Tell me about your current research, and how you got started.
Daniel: I work under the mentorship of Dr. Jesús Pérez-Ríos in the Department of Physics and Astronomy and our overall goal is to develop a database using machine learning in atomic and molecular physics, more specifically for the prediction of the spectroscopic constants of diatomic molecules. My individual project is focused on predicting the outcome state space of reactive atom-diatom collisions-- more specifically, the final ro-vibrational state distributions and the total cross sections of the collisions of a calcium atom and a diatomic hydrogen molecule. I first started working with Dr. Pérez-Ríos and his groupon the database for the spectroscopic constants of diatomic molecules in the beginning of 2023, a project that is about to be published; and over summer 2023, I worked with him as part of Explorations in STEM summer research program.
When you first started this work, did you have a background in machine learning?
Yes, in the previous summer I did an internship through the SULI program at Brookhaven National Lab, where I first learned about machine learning and worked on a project with Dr. Garth Williams, one of the synchrotron physicists at the NSLS-II. That's really where I started to learn how to do it and over the few last projects, I've really cemented my skills. I am more of a self-taught machine learning user, having never taken a class on the subject. But I'm glad I've been able to learn hands-on tools that are relevant and important for both of my majors, and for doing research.
What I worked on in the previous SULI internship would be considered more condensed matter physics … and now I'm doing atomic and molecular AMO physics, but both use machine learning for predictive purposes. And that's really what I like about machine learning in particular, and theoretical physics as a whole: the emphasis on using mathematical models to make predictions. … Now F= ma is a model that's much more beautiful model than, say, a very complicated neural network. But at the end of the day, what I like about my major area of study is trying to describe the physical world using mathematics.
What do you enjoy most about working in Dr. Perez-Rios’s research group?
I feel that his group is special. There's a huge emphasis on critical thinking and scientific communication. In addition, what I also like about it is that not we're being told what to do. We are given project ideas, and we're allowed to explore what we think is a good approach. As we work on the problems, Dr. Perez Rios will mentor us, and lead us in the right path. But there's always a level of individual freedom in the group, which I really like.
Do you enjoy presenting your work?
Yes, what I like about presenting is trying to explain and break it down from something that seems very abstract. For instance, I could say: “using neural networks for the prediction of the final ro-vibrational state distributions of an atom-diatom collision of calcium and diatomic hydrogen to form calcium monohydride;” Or I could just say: “I'm trying to teach a computer how to predict the outcome of a chemical reaction.” There is a very technical way of saying things, but then there's a much more human way of explaining it that is important for getting people interested in your work.
It was interesting being at CommUniversity day recently, trying to explain what chaos is to kids who even haven't attended high school yet or taken physics. We had to learn to present concepts in a way that gets people with very different backgrounds interested in the topic. At the same time, we also had somebody come talk to us at CommUniversity who was quite knowledgeable about it.
You're also been involved in tutoring. Has that also helped you to practice those same skills of presenting and communicating science?
Yes, but I feel like with tutoring, you get to be a little more hands-on with the learning. You get to show exactly why things happen the way they do. Even with introductory physics, I think it’s interesting to teach it to see how you can help other people learn it, and to try to give a new path forward, because sometimes the first way things are taught or are given to students is not necessarily the most intuitive way for them to learn. They may need to see other ways of doing it. So when I’m teaching, I try to focus a lot on showing students the way from the beginning to the end of a problem. And I feel like that path, that story of problem solving, is important and interesting. That's why I teach not to just recite facts. It's more so to show a process, a process of thinking and problem solving.
What for you has been valuable about participating in summer research programs, like SULI and Explorations in STEM?
What I like the most about those programs is that you get support to do your research, and you have a lot more uninterrupted time to focus on your work. You also get the professional development workshops. Those can be really useful – such as the graduate school panel. One of the things I also particularly like about the programs is that there's an emphasis on having deliverables. With Explorations in STEM, for instance, you had to give a poster presentation. And I don't view it as a bad thing. It's actually a really good thing because it forces you to take what you've done, even if you feel like you're not done yet done with the project, and you have to be able to succinctly explain what you're doing. That's why I liked the summer symposium so much. You have something to show: Here's some proof. Here's my poster. This is what I did. This is what I learned from doing this experience.
What are the most challenging aspects of the kind of research you do?
Machine learning takes a lot of experimenting. It's not one size fits all when it comes to determining what is exactly the right model, or what is the right model architecture. It does take playing around. But as you gain more experience, you gain intuition. And the more you read, the more intuition you develop as to what will work-- so that can help to cut down the amount of time that you are spending trying to get things to work. Sometimes you will have to change trajectories, because maybe it is not the correct model type. Sometimes, you change one thing-- and all of a sudden it works. So those kind of breakthrough moments can be important to fueling the forward motion of the project.
I’ve had times where I make changes to make it better and it actually ends up making it worse … so there comes an element that you have to be realistic with what works or what doesn’t work. You do have to try a lot, and you do have to repeat a lot of steps. But having the intuition and working on trying to think it out mathematically, can help lessen the feeling of repetition. You have to always be thinking: what other useful things can be done with the machine learning tools that you can pivot to? You don't want your project to come to a screeching halt because one thing doesn't work. Maybe sometimes you’re asking too much of the model… Instead of trying to ask it to predict an entirely new chemical reaction, say, tritium and calcium, maybe I don’t ask the model to extrapolate to such an extreme. Maybe I try to get something that is a little more similar to the training data—something that gives the machine learning algorithm a better shot of getting it right. Because there's no guarantee it will work, and that's very much the case with research. There’s no guarantee It will work when you set out, especially if you're trying something new.
What advice do you have for students interested in physics research?
Well, just starting a project is good. You don't want to wait until you've learned everything, because you're not going to learn everything. So there has to be an element of being open to learn, but also being open to not knowing, even while you are in the research process. No matter how many classes you take, there will be a point to where there is a part of the research that you do not know. It's not going to be covered in class. It will never be covered in class, and you do have to learn it on your own. But that's not a bad thing, and it should not deter you from continuing into research. It's okay to not know everything. You can still do a lot with the subset of knowledge that you do have. And it's important to keep on building from what you have, instead of focusing on what you don't.