摘要
DNPH-modified PS@Ag serves as a measurable pre-exposure reference state. • Paired ΔSNV resolves aldehyde-induced signed spectral redistribution. • Full-spectrum ΔSNV classifies four aldehydes with balanced accuracy 0.977. • Endpoint and ΔSNV provide complementary views of aldehyde identity. • Accuracy remained 0.90 in preparation holdout and 0.80 in concentration holdout. Reactive surface-enhanced Raman spectroscopy (SERS) interfaces are challenging to interpret because endpoint spectra combine reagent bands, analyte-induced changes, and hotspot-dependent variation. Here, a PS@Ag surface modified with 2,4-dinitrophenylhydrazine (DNPH) served as a structured pre-exposure reference, and aldehyde exposure was represented as a preparation-linked state transition derived from repeated spatial measurements on the same substrate. Four aldehydes were examined across broad nominal concentration ranges using independently executed preparations. Across three formaldehyde preparations, global transition descriptors showed more consistent concentration directionality than endpoint total intensity, whereas local coordinates were more preparation sensitive. At the matched 10 mg g −1 benchmark, endpoint spectra achieved a preparation-grouped balanced accuracy of 1.000. Full-spectrum standard-normal-variate difference spectra (ΔSNV) achieved balanced accuracy of 0.977 and macro-F1 of 0.971 while resolving signed redistribution from the DNPH prelayer to the endpoint. Preparation-replicate and leave-one-concentration-out validation yielded balanced accuracies of 0.90 ± 0.05 and 0.80 ± 0.05, respectively, for aldehyde-identity classification. A fixed full-spectrum ΔSNV model was subsequently evaluated on 12 newly fabricated substrate-preparation blocks comprising 96 endpoint-derived target observations, yielding balanced accuracy and macro-F1 of 0.979. Complementary XPS of a representative 10 mg g −1 formaldehyde endpoint supported DNPH deposition and aldehyde-exposure-associated surface-composition redistribution, while HPLC/PDA resolved DNPH derivatives of formaldehyde, acetaldehyde, and acrolein. The results establish a qualitative, state-resolved framework for aldehyde discrimination. Download: Download high-res image (190KB) Download: Download full-size image
危险品化学品经营许可证(带存储)