CAREER: Deep Learning-Enabled SERS Sensors for Next-Generation Drinking Water Monitoring

Maintaining safe drinking water requires detecting trace-level amounts of harmful contaminants quickly. However, many water systems use laboratory tests that are slow and expensive. This CAREER project will create a faster, low-cost way to check drinking water for harmful chemicals by combining a chemical sensor with artificial intelligence (AI). The project will use a method that measures how light is scattered from chemicals attached to a surface. Each chemical generates its own pattern, like a fingerprint, that can be read. Drinking water may contain a collection of chemicals. The project will use AI to help read and understand the test results. The project will study how real-world conditions such as acidity and salts in the water affect how well the method works. The detection system will quickly flag possible contamination, so water utilities can do follow-up testing. The project will also use computer models to better understand how chemicals interact with the sensor, which will help improve the sensor design. Project outcomes will help protect public health and lower the cost of water testing. The project will also provide research opportunities to college students and high school teachers, create teaching materials about water sensors and AI, and offer hands-on outreach activities about drinking water quality problems. This CAREER project will develop a deep learning-enabled surface-enhanced Raman spectroscopy (SERS) platform for rapid, low-cost, and quantitative detection of organic contaminants regulated under the U.S. Environmental Protection Agency National Primary Drinking Water Regulations. The project will integrate plasmon-enabled spectroscopy, aquatic chemistry, and AI to improve sensitivity, selectivity, and matrix robustness in drinking water analysis. Objective 1 will establish predictive relationships among analyte physicochemical properties, adsorption thermodynamics, aquatic chemistry variables, and SERS sensitivity by combining adsorption isotherm measurements, controlled matrix experiments, and density functional theory modeling of analyte–sensor interactions. Objective 2 will develop multivariate statistical and computational methods for contaminant identification and quantification by leveraging orientation-dependent vibrational signatures as reproducible analytical features across plasmonic substrates. Objective 3 will build matrix-transferable deep learning models for spectral deconvolution, contaminant quantification, and interference mitigation in complex drinking water matrices, including mixed-contaminant conditions and variable water chemistries, with the goal of achieving sub-part-per-billion detection for most regulated organic contaminants and establishing a pathway toward part-per-trillion performance for priority analytes. The project will contribute new mechanistic insight into SERS signal generation in environmental systems, generalizable AI-enabled methods for quantitative spectral interpretation, and a scalable framework for intelligent water quality sensing. It will provide training for undergraduate and graduate students and high school teachers, curriculum modules in sensing and data analytics, and outreach activities that connect water quality challenges with hands-on learning in spectroscopy, AI, and molecular modeling. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2543927 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Haoran Wei | Institution: University of Wisconsin-Madison, MADISON, WI | Award Amount: $550,000 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543927 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543927.html