Towards a validation of surface-enhanced Raman scattering (SERS) for use in forensic science: repeatability and...

In order for a new analytical technique such as surface-enhanced Raman scattering (SERS) to be used in a routine manner, data and studies on the validation of the method are required. In that context, we performed a systematic study of the variability observed at different levels of the analytical procedure (i.e. respectively measurement, sampling, colloids aliquots, colloids batches, laboratories). Our goal is to provide data towards a qualitative validation of the technique for identification purposes. Three molecules of forensic interest were used as probes, respectively crystal violet, methamphetamine and 2,4,6-trinitrotoluene (TNT). We demonstrate that the method is repeatable with RSD and multivariate techniques (PCA). The % RSD at the different analytical stages vary between the molecules and the peaks considered. The repeatability is on the order of 2-6% for crystal violet, and 5-16% for TNT. Methamphetamine binds very weakly to the silver colloids giving much greater variability in the measurements (5-29%). We show that spectra measured in the same conditions (e.g. same laboratory and instrument), even a few days apart, are comparable and stable. The largest source of variation has been identified to be the measurement conditions and the associated random fluctuations in intensity (i.e. Brownian motion of the particles, solvent evaporation and concentration). The influence of the substrate is confirmed to be negligible. However, the reproducibility between different laboratories and different instruments introduced the largest source of variability (∼ 10-70%). Despite these factors, we demonstrate that qualitative identification of the species under analysis by measurement and comparison of peaks position is always successful even though quantitative analysis is, at present, difficult. Regardless of the amount of variability determined, the molecules could always be successfully identified, even on different instruments from different laboratories by utilizing the criterion proposed in the literature (i.e. 3:1 signal-to-noise ratio).