Integrated SERS and machine learning platform for rapid acrylamide detection in water

Overview

This solution addresses the critical challenge of detecting acrylamide, a probable human carcinogen, in drinking water at regulatory-relevant concentrations. Current analytical methods rely on laboratory-based techniques that are slow, expensive, and labor-intensive. The proposed platform integrates surface-enhanced Raman spectroscopy (SERS) with machine learning algorithms to deliver rapid, low-cost, and highly accurate acrylamide quantitation in water matrices.

The technology targets two performance benchmarks: detection down to 5 ppb in deionized water using simple single-band analysis, and detection down to 0.1 ppb in real drinking water using machine learning-enhanced spectral interpretation. With a per-test time of under 10 seconds and a projected cost of $0.01, this approach makes large-scale water quality monitoring economically feasible for utilities, regulatory agencies, and food safety laboratories.

Technical specifications

Core technology components:

• SERS substrates: Low-cost colloidal gold nanoparticles (50 nm) that amplify Raman signals from acrylamide molecules, enabling trace-level detection
• Raman spectroscopy: Targets the acrylamide C=C stretching band at approximately 1700 cm⁻¹, normalized against the Au-Cl reference band at approximately 250 cm⁻¹ for quantitative analysis
• Automated spectral processing pipeline: Includes smoothing, baseline correction, normalization, feature extraction, and multivariate analysis
• Machine learning integration: Support vector machine (SVM) models trained to distinguish acrylamide signals from background interference caused by organic coatings on nanoparticles and dissolved organic matter in drinking water matrices
• Sample preparation: Simple mixing of gold colloid, water sample, and 10 mM K₂SO₄ aggregation agent

Key performance targets:

• Quantification limit: 0.1 ppb in drinking water (with ML); 5 ppb in deionized water (single-band)
• Accuracy: ±0.05 ppb (ML-enhanced) or ±1 ppb (single-band)
• False positive rate: less than 0.5%
• Test time: under 10 seconds per sample
• Cost: approximately $0.01 per test

Technology readiness level

This technology is currently at TRL 3–4 (analytical and experimental proof-of-concept). The principal investigator has previously demonstrated SERS-based quantification of amine-containing pollutants such as atrazine and imidacloprid at low-ppb levels, establishing the feasibility of the approach. Acrylamide is expected to yield even stronger signals due to its C=C moiety, which is an excellent Raman scatterer.

The six-month validation plan involves collecting and analyzing Raman spectra across acrylamide concentrations from 0 to 1000 ppb in treated Lake Mendota water and Madison groundwater. Hypothesis 1 will be validated through single-band intensity ratio analysis in deionized water, while Hypothesis 2 will be validated by training and testing the SVM model on spectra from complex drinking water matrices. Successful completion will advance the technology toward prototype development and field testing in real-world water distribution systems.