The goal of scientific computing has always been to recreate the physical world, but the physical world is extraordinarily complex. Whether you’re simulating a nuclear reactor or something as seemingly simple as the flow of air over a wing, there are more factors—or variables—to consider than even the most powerful systems can handle if you need all the details.
Even physical experiments fall victim to chance.
“Nothing is deterministic,” ORNL mathematician Clayton Webster explained. “The air flowing over a wing, normally we would say it flows at some rate, so we fix that. But that’s definitely not true. If you’ve been on an airplane and experienced turbulence, it’s simply because of the way that air flows over a wing. It’s definitely not a constant rate.”
Webster is the winner of an Early Career Research Program award from DOE’s Office of Science. He is working to develop more sophisticated means to identify the mathematical underpinnings for complex systems, whether they are associated with a computational simulation or a physical experiment.
His job, finding the important information in mountains of data, is akin to the task of compressing digital images. While the original image has three values for each pixel—corresponding with its red, green and blue levels—developers have been able to identify the data most important to re-creating the image and, possibly more importantly, the information that can be left behind with minimal loss to the image’s quality.
“The things that we do in some sense mimic that procedure,” Webster explained. “The challenge is that now I don’t have an image, but I have something very complicated under the hood, and I want to play the same game. I want to look at some complicated problem—it can be air flow over a wing or a reactor—and I want to understand some behavior of that complex object, but I want to do it in some reduced form where I can still get back all the features I want, but with minimal information.”
If it’s done right, the process will allow even the most powerful supercomputers to handle problems potentially thousands of times more complex than they otherwise could.
Webster, who leads the lab’s Computational and Applied Mathematics Group, is a math person to the core and always has been.
“I was in gifted math classes starting in the second grade,” he said. “That and baseball were all I ever wanted to do. I was lucky that I found what I wanted to do very early, because I pretty much am not good at anything else.”
Webster grew up in Branford, Ontario—“the home of Wayne Gretzky and Alexander Graham Bell”—earning bachelor’s and master’s degrees from McMaster University before moving on to Florida State University for a Ph.D. Before coming to ORNL, he was a John von Neumann Fellow at Sandia National Laboratories and also worked in commodities trading, another math-intensive activity.
To the apparent surprise of many people he has encountered, Webster has not found his obsession with math to be limiting.
“Outside of my parents, I was told by everyone else, ‘I don’t know what you will do with a degree in math.’ I pointed out to them that you can do anything with a degree in math,” Webster said.
“It’s been very good to me, but it’s also a curse, because a mathematician never stops thinking about their problems. Otherwise, it’s my hobby. It’s what I do.”
See also: