Leading Thoughts: "At the Center of Technology"

Dean Alexis Abramson talks with Professor Geoffroy Hautier. (Photo by Rob Strong '04)

What drew you to Dartmouth? 

HAUTIER: Dartmouth is in a very unique spot with world-class education and research in a beautiful environment. One thing that I didn't realize when I first visited is the sense of community. I was not even here yet and I was already on two proposals with Jifeng Liu and Ian Baker. That sense of "Let's bring everyone together as much as possible" struck me very quickly. Our community is very interdisciplinary and because there are no walls, it's very easy to interact with others. This applies to students too. I co-advise some students doing experimental work and have a student doing both computational and experimental work who's advised by a colleague who's an experimentalist. 

Can you explain your research? 

HAUTIER: Materials science is at the center of technology. It has always been the case, from the Bronze Age and the Steel Age, all these materials are changing our lives. But usually, the way these materials have been developed is empirically—you mix chemicals together and see what happens. There is a lot of trial and error, but for a few thousand years we developed a lot of good materials that are important in our everyday life. Now, with the advent of better theories, including quantum mechanics, and stronger, bigger computers, we can predict properties of materials. 

Of all the possible combinations, what percentage do we understand?

HAUTIER: For known inorganic compounds—we know the composition, we know the crystal structure—then it is around 100,000 in organic compounds. But there are probably millions and millions more out there. We know they exist, but we have no sense of their electrical, mechanical, thermal properties. That's the ambitious part of my research: There's hundreds of thousands of things that are known—let's compute their properties. I'm involved in the Materials Project, which is putting all this data on a website where people can browse and ask questions. 

Can you talk about your methods? 

HAUTIER: This is really a crucial moment for the field. When I started 10 years ago, we could run computation, but we couldn't do much machine learning because the data wasn't there. But now we are building a lot of computed data. We can bypass the modeling and use machine learning. We have become really good at predicting properties of materials. But there's still a bottleneck in the way you synthesize materials—that's still a bit of a black art. There are plenty of synthesis paths; some of them are expensive, some are less expensive. In the future, we might reach that level where we are going to have many different materials and maybe we'll pick them not only based on their properties but also on how easy, for instance, they are to be made. 

Is that something you can predict? 

HAUTIER: My intuition is that it might be a combination of modeling of certain things, diffusion, nucleation events, maybe machine learning. We've been synthesizing materials for a long time now, and there are rules. So, combining this with a machine learning and an AI approach with modeling could bring us to the point we can predict synthesis. 

You and Jifeng Liu recently discovered a new material for solar absorbers … 

HAUTIER: This one is very exciting because the impact is huge. Solar power is going to grow, and we need alternatives or materials you could use in combination with silicon as the current solar cell material. We've been taking candidate materials and computing their defects. We are working now on barium cadmium phosphide. The material has a known composition, but nobody had suggested it would be a good solar absorber. But when you do a deep dive and use computation, then you start to see that it has very promising properties.