Per- and polyfluoroalkyl substances (PFAS) are oil- and water-resistant chemicals that are used in a range of consumer and industrial products, including nonstick pots and pans, fast-food packaging, firefighting foam, raincoats and stain-resistant carpeting. Sometimes called "forever chemicals," they are long-lasting and do not naturally degrade but instead accumulate over time.
But researchers at the Department of Energy's (DOE) Argonne National Laboratory and the University of Chicago Pritzker School of Molecular Engineering (UChicago PME) have developed a novel method to detect miniscule levels of PFAS in water. The method, which they plan to share via a portable, handheld device, uses unique probes to quantify levels of PFAS, some of which are toxic to humans.
"Existing methods to measure levels of these contaminants can take weeks and require state-of-the-art equipment and expertise," said Junhong Chen, lead water strategist at Argonne and the Crown Family Professor at UChicago PME. "Our new field test can measure these contaminants in just minutes."
The technology, published in the journal Nature Water, can detect PFAS present at 250 parts per quadrillion (ppq) - like one grain of sand in an Olympic-sized swimming pool. This makes the sensor 16 times more sensitive than U.S. Environmental Protection Agency (EPA) requirements, giving the test utility in monitoring drinking water for two of the most toxic PFAS - perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) - for which the EPA recently proposed limits of 4 parts per trillion.
"PFAS detection and elimination is a pressing environmental and public health challenge," said co-senior author Andrew Ferguson, professor of molecular engineering at UChicago PME. "Computer simulations and machine learning have proven to be an incredibly powerful tool to understand how these molecules bind to molecular sensors and can guide experimental efforts to engineer more sensitive and selective molecular probes."
"Even though they are typically present at miniscule concentrations, PFAS do have certain molecular characteristics that differentiate them from other things dissolved in water, and our probes are designed to recognize those features," said co-senior author Seth Darling, senior scientist at Argonne and UChicago and director of the Advanced Materials for Energy-Water Systems (AMEWS), an Argonne-led DOE Energy Frontier Research Center.
A Detection Challenge
In recent years, studies have linked PFAS to health concerns, including cancers, thyroid problems and weakened immune systems. In light of some of these findings, the EPA proposed new limits for PFOS and PFOA.
"The problem with enforcing these limits is that it's very challenging and time-consuming to detect PFAS," said Chen. "You currently can't just take a sample of water and test it at home."
The gold standard for measuring PFAS levels is an expensive laboratory test known as liquid chromatography/tandem mass spectrometry, which separates chemical compounds and provides information on each one.
Researchers attempting to make their own faster and cheaper PFAS tests face a few challenges: for one, PFAS are often present in water at much lower concentrations than dozens of other more common chemical contaminants. In addition, there are thousands of different PFAS with only slight variations between their chemical structures - but important differences in their health effects and regulations.
But Chen's team has been developing highly sensitive, portable sensors on computer chips for the last 15 years. Chen is already using the technology in a lead sensor for tap water, and his lab group suspected that the same method could be used in PFAS sensing. Their proposal to adapt the technology for PFAS became part of Great Lakes RENEW, the National Science Foundation (NSF) Water Innovation Engine in the Great Lakes.
Designed by AI
The gist of Chen's sensor is that if a PFAS molecule attaches to his device, it changes the electrical conductivity that flows across the surface of the silicon chip. But he and his colleagues had to figure out how to make each sensor highly specific for just one PFAS chemical - such as PFOS.
To do this, Chen, Ferguson, Darling and their teams turned to machine learning to help select unique probes that could sit on the sensing device and ideally bind only to the PFAS of interest. In 2021, they won a Discovery Challenge Award from the UChicago Center for Data and Computing (CDAC) to support their use of artificial intelligence (AI) in designing PFAS probes.
"In this context, machine learning is a tool that can quickly sort through countless chemical probes and predict which ones are the top candidates for binding to each PFAS," said Chen.
In the new paper, the team showed that one of these computationally predicted probes does indeed selectively bind to PFOS - even when other chemicals common in tap water are present at much higher levels. When water containing PFOS flows through their device, the chemical binds to the new probe and changes the electrical conductivity of the chip. How much the conductivity changes depends on the level of PFOS present.
To ensure that the readings from the new device were correct, the team collaborated with the EPA and used EPA-approved liquid chromatography/tandem mass spectrometry methods to confirm concentrations and verified that the levels were in line with what the new device detected. The team further showed that the sensor could maintain its accuracy even after cycles of detection and rinsing, suggesting the potential for real-time monitoring.
"Our next step is to predict and synthesize new probes for other, different PFAS chemicals and show how this can be scaled up," says Chen. "From there, there are many possibilities about what else we can sense with this same approach - everything from chemicals in drinking water to antibiotics and viruses in wastewater."
The end result may eventually be that consumers can test their own water and make better choices about their environment and what they consume.
Other authors on the paper are Yuqin Wang and Yining Liu of Argonne, UChicago PME and AMEWS; Hyun-June Jang, Wen Zhuang, Haihui Pu and Xiaoyu Sui of Argonne and UChicago PME; Max Topel, Siva Dasetty and Ellie Ouyang of UChicago PME; Mohamed Ateia of the EPA and Rice University; Aaron Tam of the Oak Ridge Institute for Science and Education; Vepa Rozyyev of AMEWS; and Sang Soo Lee and Jeffrey Elam of Argonne.
The research was funded by CDAC, the DOE Office of Science, AMEWS, the NSF and Great Lakes RENEW.