New SERS method turns signal noise into concentration fingerprints

Researchers at Capital Normal University and the Henan Academy of Sciences say they can use full surface-enhanced Raman scattering intensity distributions, not just peak averages, to identify trace concentrations more reliably. The approach improved classification across dye, drug, pesticide, and serum tests and could strengthen ultrasensitive analysis in food safety, environmental monitoring, and public security.

Why it matters: - Ultrasensitive SERS measurements often vary too much for standard peak-based calibration to deliver reliable concentration readings. - The new method turns those fluctuations into usable information, which could improve trace analysis in chemical testing, biosensing, food safety, environmental monitoring, drug analysis, and public security. - The approach also reduces dependence on single-molecule binary counting, which usually requires stricter experimental conditions.

What happened: - Researchers from Capital Normal University and the Henan Academy of Sciences developed a statistical SERS strategy based on concentration-encoded intensity distributions. - The team paired histogram-based SERS patterns with a compact one-dimensional residual neural network called ResNet4. - The study was published online Aug. 27, 2026, in Opto-Electronic Advances. - The paper is titled "Statistical SERS spectroscopy based on concentration-encoded intensity distributions for deep-learning-assisted quantitation." - The paper DOI is https://doi.org/10.29026/oea.2026.260163.

The details: - The researchers used relatively uniform silver nanoparticle arrays made by the Langmuir-Blodgett method. - Rhodamine 6G served as the model analyte. - SERS mapping covered 30 μm × 30 μm areas. - Conventional peak-intensity analysis showed large point-to-point variation even when average signal rose with concentration. - At 1 × 10⁻⁹ M, the relative standard deviation of the 612 cm⁻¹ band was 37.9%. - Calibration slopes and intercepts varied across substrate batches. - When the same data were treated as intensity distributions, the histograms changed predictably as concentration rose from 1 × 10⁻¹¹ to 1 × 10⁻⁷ M. - The distribution shape shifted from a highly skewed long tail to an almost symmetric Gaussian-like profile. - Skewness fell from 3.515 to 0.086 across that concentration range. - Physical modeling used Poisson statistics, generalized Mie theory, and Monte Carlo simulations. - The simulations modeled molecules randomly adsorbed in a plasmonic hot spot formed by two 60 nm silver nanoparticles. - The model showed that more contributing molecules shifted the distribution from long-tailed to nearly Gaussian. - A smaller 1 nm gap produced stronger fields and greater low-concentration asymmetry, but the same overall transition remained. - The team says hot-spot strength shapes the distribution, while molecule-number fluctuations drive its concentration-dependent evolution. - ResNet4 was trained and tested on independent SERS substrate batches and mapping measurements. - For five R6G concentrations from 1 × 10⁻¹¹ to 1 × 10⁻⁷ M, the model reached 100% identification accuracy. - The model outperformed KNN and SVM classifiers on the R6G task. - Methamphetamine in methanol was classified from 1 ppb to 10 ppm with 99.6% average accuracy. - Thiram in bean-sprout extract reached 99% accuracy from 1 × 10⁻⁸ to 1 × 10⁻⁶ M. - Rosiglitazone maleate in methanol-treated artificial serum reached about 96% accuracy from 1 × 10⁻⁷ to 1 × 10⁻⁵ M.

Between the lines: - The core shift is from asking how tall a SERS peak is to reading the full statistical shape of many measurements. - That matters because rare bright events and substrate variation can distort single-point calibration, while distributions capture more of the underlying measurement behavior. - The result suggests random-looking SERS variability can function like a concentration fingerprint rather than a nuisance.

What's next: - The approach could be extended to more analytes and real-world matrices if the distribution patterns hold up beyond the demonstrated tests. - The method may support broader deployment in trace chemical analysis and biosensing where conventional quantitation remains unstable. - The authors frame the strategy as a route toward more reliable ultrasensitive quantitation across applied settings.