Building Up the Foundation of Science: an Interview with Professor Shanhui Fan
One of the many highlights of the WPI-SKCM² Summer School 2024 was having Professor Shanhui Fan, the Joseph and Hon Mai Goodman Professor of Stanford University, come to Hokkaido to give a seminar to our Summer School participants. He traveled all the way from California with his daughter, Elsa Fan, who spent much of her flight watching Chiikawa, a Japanese animated series about cute animals. It was her first time in Japan, so on the first night they decided to try real Hokkaido style miso ramen! They said the kind they tried at home in California could not compare to the real thing.
During her time with us, we had many chances to hear Elsa’s thoughts about what it’s like to have such a prominent scientist as a dad. When asked how she would describe it to other children, she explained, “I’d tell them that they can learn stuff, but it’s sort of like just having any other parent, honestly.” This observation is certainly hard to disagree with! Scientists are all people, just like non-scientists. Although, A few moments later when asked if her dad ever helps with her homework, she said “Well, he doesn’t really talk about homework, he sometimes talks about learning and I sometimes ask him questions [about] quantum mechanics and stuff.” Well, we certainly don’t remember our parents tucking us into bed with stories of quantum mechanics!
Nevertheless, there may be many ways in which Professor Fan is an ordinary dad, but of course there are also many ways in which he is extraordinary. And much like Elsa Fan, we were able to learn some things not only from his seminar entitled “Synthetic Frequency Dimensions in Modulated Ring Resonators”, covering many different concepts including nonreciprocity in optics, but also afterwards when we were able to follow up with him on some topics we were curious about. During this interview, we were able to get more insight into his research and his success in incorporating his findings into two successful startup companies: Skycool and Flexcompute.
Interviewer: I saw on your website [that there are tons of different topics you research]. Are they a natural extension of each other?
Prof. Fan: No, but the fundamentals are quite related . For example, most of them come from solving Maxwell’s equations so there’s a connection that way.
The underlying argument is all [about] controlling electromagnetic fields and electromagnetic waves. And it actually permeates into all these things. It is remarkable that a single equation can have that implications in so many different fields. But that’s what we explore.
Interviewer: That’s amazing. And so, is it your students and your postdocs that come to you with these new ideas and then you kind of branch out …?
Prof. Fan: Well, we certainly generate new ideas through interactions among the group members. We have a few major directions that we have consistently pursued over the years. One of them is understanding thermal radiation. Which I didn’t talk about here but I will talk about it at the conference tomorrow [NanoRad 2024]. We have been interested in thermal radiation for a long time in part because I have always been intrigued by the thermodynamic consequences of light. This is perhaps something that’s commonly thought about. But if you look at some of the most important energy technology for example solar cells, the basic theoretical arguments are all about the thermodynamics of them, so I’ve been quite intrigued by that aspect. And my group has been trying to take those ideas into different directions. Our efforts in radiative cooling came out of this line of thinking.
The other area that I always find fascinating is the issue of nonreciprocity, which I talked about at my talk here. In optics, nonreciprocity is rare. And most materials that we encounter are reciprocal materials. So most theory of optical phenomena are developed with reciprocity in mind. Removing the reciprocity constraint gives qualitatively different physics, which is why we are interested in this area. There are practical reasons for understanding reciprocity, like the study of isolators. And these practical reasons give [an] additional guide and stimulus as to what’s important to work on.
So, in general we’re interested in new physics phenomena that have practical use – or at least have practical arguments as to why such phenoman are potentially useful. We usually do the basic work on these phenomena. So that probably characterizes most of the stuffs that we do.
Interviewer: Basic work, but you said in your talk and, also, I saw on your website that you started two startups so… How did you decide which ideas were good for startups and how did those start?
Prof. Fan: One of them is on radiative cooling. The company, Skycool systems, aims to do passive air conditioning for buildings. We published the initial theoretical work on radiative cooling in 2013. After a few years of working on this idea, it became clear that you can make an economic argument about under what circumstances radiative cooling actually makes sense commercially. So, together with two members of my group, Aaswath Raman, and Eli Goldstein, we started a company trying to put radiative cooling on buildings. The company is now deploying radiative cooling technology in a number of places, including supermarkets and data centers in California. On the company’s website you can find pictures of Skycool’s radiative cooling system deployed on the roofs of supermarkets. I am quite proud of how far this technology has gone. In ten years, it went from our initial simulation, to the deployment on somebody’s roof. That, I’m very proud of.
The other company, Flexcompute, works on GPU Accelerated Scientific Simulation. In my group, as I mentioned, the underlying thing is to solve Maxwell’s equations, so we’re actually a group that works a lot on computations and simulations. Through these works we recognize a few exciting opportunities. One of them is that if you look at machine learning, the GPU has a huge impact on machine learning. And many of those impacts can be realized for scientific simulations as well.
Interviewer: What kind of impact does it have?
Prof. Fan: It’s a lot faster when compared with traditional CPU based technology,. In Flexcompute’s demonstration, we were able to take a standard algorithm for simulating Maxwell’s equations, something called the finite difference time domain method, and show that if you run it on a GPU, it will be 100 times faster.
Interviewer: Oh, I see, that’s a big, big difference.
Prof. Fan: Huge difference. The second realization, is that the business model of doing simulation has to change. if you look at how typically in academia and in industry the simulations are done, Typically you start by buying your computer cluster. Then you go buy a license for a software. Then you install the license and the software on the cluster, and then you run it. This process alone may take weeks or months before anything gets done. And also, then you have to maintain the cluster. For most engineers this is a complete waste of time. Because why would you care about…
Interviewer: You just want the simulation!
Prof. Fan: So, Flexcompute is an entirely cloud based setup where you just have a web interface. Then you can run our code on our cluster, the moment you sign up. We built this up gradually, improving things like user interfaces, features, until these aspects become competitive with existing commercial softwares, and that took many years. We build codes that are native GPU and also natively for cloud. It took us a while to build it up but now it has very great market adoption. The company now has 80 people. And we just closed another round of 50 million dollars. So next year we’ll probably have a couple hundred people. It’s been very widely used by industry companies. Almost all of our revenue comes from industry.
Interview: I wanted to ask, what is your role…?
Prof. Fan: In both companies I’m an advisor, and I’m quite involved in the technological side of Flexcompute, Many of my former students ended up there, because as I mentioned my group does this kind of stuff as the bread and butter of the group. Therefore, a lot of students have the background that fits very well what our company needs. The company is led by Professor Zongfu Yu, who is a former student of mine.
Interviewer: I just want to go back to the application side, the practical side, and the sustainability. You have a lot of sustainability applications. But you work at a very fundamental level and I think for some of them it’s very natural to see potential applications. So, for you, how do you see those things in the future?
Prof. Fan: Even though some of our works are indeed quite useful practically, I don’t necessarily insist that every project we do is practically meaningful, at least in the short term. To me, if it is fundamentally interesting, it is good enough for me. Occasionally, we’ve done some things that actually are useful, and I’m very happy with it.
We have spent a lot of time, for example, developing ideas and fundamental theories about fundamental limits for solar energy harvesting. We have a recent paper of a new design to reach the fundamental limit. Now these things are nowhere near practical. Not even close. But that is how things develop. If you go and read one of the pioneering papers on silicon solar cell, Shockley and Queisser‘s analysis of silicon solar cells, it’s a paper that predicts the silicon cell’s going to get 30% efficiency. This is at the time when the practical cell efficiency is a few percent. So, it’s not surprising that everybody has great respect for it – Shockley – got a Nobel Prize, his name is on the paper – first. [There] must be a lot of respect for the paper. But for many years, the paper was not well cited. For many, many decades, nobody cared much about it. We only had a few percent efficient solar cell. You told me about a theory of a 30 percent efficient cell, okay? Sounds wonderful! Doesn’t do me any good. If you look at the citation record, the impact of the work took off only in the past fifteen years. Because now silicon cells are at the limit. Now, understanding solar cell performance at the fundamental limit is important.
For me, I certainly want to go after the practical side. But I am also inspired by this kind of fundamental work. In energy technologies, there are many questions that can be phrased in a fundamental way (the same way Shockley and Queisser described silicon cells): These fundamental work may not be useful in the next five years, but ultimately these work points to where the field should be. The Shockley-Queisser paper was written in 1960, so it is indeed old. But that is a paper that I ask every one of my students to read, because if you read it, you understand how good fundamental work really should look like. That’s the inspiration.
—–
In any case, one cannot argue with the fact that Professor Fan’s dedication to research has begun to make an impact on the world. Whether it’s working on fundamental theories to further our understanding on the subject for future researchers, or applying the science to create a more environmentally friendly cooling system or providing high quality and quick simulations to fellow scientists, much of his work has made a positive impact. We hope that the members of our institute can also remember that any work they do will make an impact, no matter how small. Even if there is no immediate application for the science you want to work on, it has the potential to provide essential knowledge for other scientists now, or even many, many years into the future. Your work can help “make your field what it should be” and be a part of building the foundation for future scientists.
And finally, we hope that we may once again have the privilege of hosting and learning from Professor Shanhui Fan in the near future!
If you would like to learn more about Professor Shanhui Fan’s work, please visit his profile on the Stanford website here: Professor Shanhui Fan (stanford.edu)
