Quantitative optimization and identification of Nitroaniline isomers: An application case of MOF-based fluorescent...

Detecting pollutants such as electrically neutral aromatic amines in environmental water bodies is a challenging endeavor. Sensing array technique, especially MOF-based fluorescence sensor array, has become the preferred approach due to the unique advantages of MOF sensing elements in providing multiple recognition sites and the multi parameter advantages of fluorescence properties. The multidimensional advantages provided by these features should be further explored and utilized. However, the current optimization efforts are primarily focused on the selection of sensor types, and the methods of signal acquisition and data analysis have not yet received much attention. In this study, we used 3D full range fluorescence scanning to obtain three-dimensional sensing signals and made the quantitative analysis method work best for this kind of information. Three distinct phases of the optimization were conducted taking three non-fluorescent isomers, o-nitroaniline, p-nitroaniline, and m-nitroaniline as detection targets. Initially, based on the internal filtering effect (IFE) sensing mechanism, Zn-MOF-74 was selected and synthesized as the sensor. The optimal dosage was determined and the stability was further investigated. Secondly, signal acquisition was achieved with excitation-emission matrix (EEM) to construct a virtual sensor array (VSA). Thirdly, the qualitative PCA-SVM (principal component analysis-support vector machine) model and quantitative PCA-LUT (lookup table) model, which combine two-stage nonlinear fitting with interpolation, were formulated using the first three principal components as features. The research results suggest that the classification accuracy of a single sample is 97.7 %, binary mixed samples is 76.2 %. And this method also shows good detection results for tap and river water samples. Especially, the PCA-LUT has a stable performance in predicting all three compounds with significantly better MAE and MSE and high analysis efficiency. Compared with neural network method, the advantage of PCA-LUT is that it can be used for limited calibration samples and the analysis process is intuitive with good interpretability. This study provides a construction and optimization case for the fluorescence sensing array method, covering the entire process from sensor selection, signal acquisition to data processing method.