Machine Learning-Assisted Colorimetric Assay Based on Au@Ag Nanorods for Chromium Speciation

• Published in: Analytical chemistry (Washington) 2023-07, Vol.95 (26), p.10110-10118
• Main Authors: Orouji, Afsaneh, Ghasemi, Forough, Hormozi-Nezhad, Mohammad Reza
• Format: Article
• Language: English
• Subjects: Chemical analysis; Chemical speciation; Chemistry; Chromium; Color; Colorimetry; Contamination; Detection limits; Discriminant analysis; Drinking water; Gold; Groundwater; Groundwater pollution; Ions; Learning algorithms; Least squares method; Machine learning; Metallizing; Mineral water; Nanorods; Oxidation; Pattern analysis; Pattern recognition; Rainbows; Rivers; Seawater; Silver; Speciation; Species; Trivalent chromium; Valence; Water analysis; Water sampling
• ISSN: 0003-2700; 1520-6882
• DOI: 10.1021/acs.analchem.3c01904

The oxidation state of an element significantly controls its toxicological impacts on biological ecosystems. Therefore, design of robust sensing strategies for multiplex detection of species with respect to their oxidation states or bonding conditions, i.e., chemical speciation, is quite consequential. Chromium (Cr) species are known as the most abundant inorganic groundwater pollutants and can be quite harmful to human health depending on their oxidation states. In the present study, a multicolorimetric probe based on silver-deposition-induced color variations of gold nanorods (AuNRs) was designed for identification and quantification of Cr species including Cr (III) and Cr (VI) (i.e., CrO4 2– and Cr2O7 2–) in water samples. In fact, the presence of Cr species leads to inhibition of the silver metallization of AuNRs to various degrees depending on the concentration and identity of the analyte. This process is accompanied by the blue shift of the longitudinal peak which results in sharp-contrast rainbow-like color variations, thereby providing great opportunity for highly accurate visual detection. The gathered dataset was then statistically analyzed using two pattern recognition and regression machine learning techniques. In particular, linear discriminant analysis was used as a classification method to discriminate the unicomponent and mixture samples of Cr species with 100% accuracy. Then, a well-known multivariate calibration technique called partial least-squares regression was employed for quantitative analysis of Cr species. Responses were linearly related to Cr species concentrations over a wide range of 10.0–1000.0, 1.0–200.0, and 1.0–200.0 μmol L–1 with detection limits of 37.7, 8.7, and 2.9 μmol L–1 for Cr3+, CrO4 2–, and Cr2O7 2–, respectively. The practical applicability of the multicolorimetric probe was successfully evaluated by analyzing Cr species in several water specimens comprising tap water, mineral water, river water, and seawater. Above all, the vivid rainbow color tonality of the proposed assay further improves the accuracy of the naked eye detection, making it a practical platform for on-site monitoring of Cr contamination.