Monitoring air contaminant exposure is vital in ongoing efforts to detect the presence of dangerous chemicals and understand how urban pollution affects human health.
And a team of researchers at the University of Wisconsin-Madison and Cornell University is developing liquid crystal-based sensors that can quickly and accurately detect trace amounts of air contaminants—everything from carbon monoxide to chemical weapons such as sarin. A $1.9 million grant from the National Science Foundation is funding their efforts.
Liquid crystals, like those used in electronics displays, are states of matter where the molecules can either move and flow like a liquid or adopt an organized arrangement. That switch from the free-flowing to organized phase often alters how liquid crystals interact with light. It’s what produces the vibrant images on LCD screens, which use electricity to switch liquid crystals between phases pixel-by-pixel.
“The versatility of liquid crystal sensors is quite remarkable,” says Victor Zavala, the Baldovin-DaPra Associate Professor in chemical and biological engineering at UW-Madison. “We want people to have more information about what they’re exposed to in their daily lives, especially within cities. Right now, we measure few contaminants and at a few locations, but that doesn’t tell the whole story.”
Liquid crystal technology is already under exploration for low cost, portable, personal sensors. However, such existing liquid-crystal-based chemical sensors are limited in response times; detectable changes become visible to human eyes only after several minutes.
That’s too long, given that most of our exposures to airborne contaminants occur briefly—for example, if we walk through a poorly ventilated tunnel or catch a blast of exhaust from a construction site.
The researchers already have developed a sensor that can detect specific sarin concentrations within a few seconds. And to speed up their sensors, the scientists turned to artificial intelligence: Deep neural networks are computer programs that learn to recognize patterns without input from humans.
When Zavala and colleagues allowed deep neural networks to take a look at snapshots of the new sensors, those algorithms spotted changes in patterns inside the liquid crystals after only three seconds—correctly identifying exposure to the sarin-like chemical with 99-percent accuracy.
And though the deep neural networks excelled at identification, the researchers weren’t satisfied to blindly trust the machines.
So drawing on the expertise of colleagues Reid Van Lehn and Manos Mavrikakis, the team is starting to run molecular simulations and computationally modeled interactions between contaminants and the liquid crystal—linking the abstract “thoughts” of deep neural networks to actual physical mechanisms occurring within the liquid crystal sensors.
“We’re scientists; we want to know what is actually happening,” says Zavala. “Deep neural networks tell us that there is something hiding in there, and our job is to try to find it and understand it.”
The UW-Madison team will collaborate with Nicholas Abbott of Cornell University. He will run extensive experiments to explore the sensitivity of different sensor designs under diverse chemical environments and to collect data to train artificial intelligence algorithms. Armed with data, physical models, and artificial intelligence techniques, the team plans to develop new and powerful liquid-crystal-based sensors for several additional chemicals, which they’ll select with input from leading air quality expert James Schauer, a professor of civil and environmental engineering at UW-Madison.
In the Department of Chemical and Biological Engineering at UW-Madison, Reid Van Lehn is an assistant professor and the Jay and Cynthia Ihlenfeld Faculty Scholar and Manos Mavrikakis is the Vilas Distinguished Achievement Professor and Paul A. Elfers Professor. James Schauer also directs the Wisconsin State Laboratory of Hygiene. The research is supported by the National Science Foundation Big Data Science & Engineering Initiative, award number 1837812.
Author: Sam Million-Weaver