*3.4. Detection of Individual Pesticide on AgNPs-PDMS Substrate*

The optimized substrate was employed to test standard solutions of thiram, malathion, and phoxim. Figure 4 presents the SERS spectra of different concentrations of thiram (a), malathion (c), and phoxim (e). The linear calibration plots of all three pesticides are also illustrated in Figure 4b,d,f, respectively. SERS detection of thiram has been extensively studied in previous reports [38,40,43,44]. In this study, thiram was chosen first for pesticide detection. As illustrated in Figure 4a, characteristic peaks at 558, 925, 1147, 1384, and 1513 cm−<sup>1</sup> are identified and assigned in Table 1 according to Figure S4a and the literature [55]. In detail, the SERS band at 558 cm−<sup>1</sup> corresponds to the stretching of the <sup>S</sup>−S bond. The broad peak at 925 cm−<sup>1</sup> is the combination of the stretching mode of the

C=S double bond and CH3N. The major signal at 1384 cm−<sup>1</sup> is assigned to symmetric CH3 deformation. The Raman shift of all these peaks matches well with previous literature reports. Among them, the peak at 1384 cm−<sup>1</sup> provides the highest SERS intensity, thus it was chosen for semi-quantitative analysis. As presented in Figure 4b, for thiram between the concentrations of 10 and 1000 μg L<sup>−</sup>1, the plotting of an integrated peak area at 1384 cm−<sup>1</sup> against the logarithm of thiram concentrations presents a linear relationship, with a coefficient of determination (R2) of 0.9665. This result demonstrates that the proposed substrate can be applied for qualitative and semi-quantitative detection of thiram at low levels. The limit of detection (LOD) is calculated as 3.62 μg L−<sup>1</sup> using the widely adopted formula LOD = 3σ/S, in which σ represents the standard deviation of background signal and S stands for the slope of the linear calibration curve. For practical application, the maximum residue limit (MRL) for thiram in apple is 5 mg kg<sup>−</sup>1, according to the national food safety standard (GB 2763-2019, China). Therefore, the sensitivity of this method is satisfied for thiram detection.

**Figure 4.** SERS spectra of different concentrations of thiram (**a**), malathion (**c**), and phoxim (**e**). Linear calibration plots between the integrated peak area and the logarithm of concentration for thiram (**b**), malathion (**d**), and phoxim (**f**). Inset of (**b**,**d**,**f**): molecular structure of thiram, malathion, and phoxim.


**Table 1.** Assignments of some characteristic bands in Raman and SERS spectra of three tested pesticides.

b—broad; s—strong; m—middle; w—weak; *ν*—stretching; *δ*—bending; *ρ*—rocking; *γ*—out-of-plane bending; *ν*s—symmetric stretching; *ν*as—asymmetric stretching; *δ*s—symmetric bending; *δ*as—asymmetric bending.

For organophosphates insecticides, malathion is tested first. The resulted SERS spectra are presented in Figure 4c. By comparing with the Raman spectrum of the reference material, it is confirmed that the recorded SERS signal is produced by SERS enhancement of malathion molecules (see Supplementary Materials, Figure S4b). The characteristic Raman band at 815 cm−<sup>1</sup> is attributed to the out-of-plane bending of the C−H bond. The strong signal at 862 cm−<sup>1</sup> is caused by the symmetric stretching mode of the C−O−C bond. The broad peak at 1144 cm−<sup>1</sup> is generated from the stretching of the P−S bond. Finally, the two bands at 1444 and 1726 cm−<sup>1</sup> are correlated with the stretching mode of the C−O and C=O bonds, respectively. The pattern of these characteristic peaks can be used for qualitative identification of malathion. For semi-quantitative detection, Figure 4d shows the linear fitting curve of data plotted between the logarithm of malathion concentration and integrated peak area at 862 cm−1. As summarized in Table 2, a linear range of 100 to 5000 μg L−<sup>1</sup> is achieved using this method, with a calculated LOD of 41.46 μg L−1. Even though the sensitivity is not as high as thiram, probably due to the weak affinity between malathion molecules and the Ag substrate, it can still be useful for on-site inspection, considering the relatively high MRL of 4 mg kg−<sup>1</sup> in oranges.

Phoxim is another target insecticide in this study due to its known toxicity, which is an acetylcholinesterase inhibitor that affects the human nervous system. According to GB 2763-2019, the MRLs of phoxim are set as 0.05 mg kg−<sup>1</sup> in most fruits. Unlike thiram, SERS detection of phoxim has rarely been reported. Using an optimized AgNPs-PDMS substrate, SERS measurement of different concentrations of phoxim standard solutions were performed. Characteristic peaks can be observed on the acquired SERS spectra in Figure 4e. By comparing with the Raman spectrum of phoxim in Figure S4c, characteristic SERS bands can be identified. Among them, the broad peak at 758 cm−<sup>1</sup> should be recognized as the stretching vibration of the P=S double bond. Peaks at 995, 1092, and 1178 cm−<sup>1</sup> are all attributed to the in-plane bending of the C−H bond. Different from thiram and malathion, SERS spectra of phoxim showed characteristic stretching vibration signals of the phenyl ring at 1437, 1502, and 1591 cm−1, which provide important information for distinguishing its signal in a mixed sample. For semi-quantitative detection, the strongest peak at 1502 cm−<sup>1</sup> is used as a reference. A linear relationship is established between

the logarithm of concentrations and signal strength. The linear range is between 50 and 5000 μg L<sup>−</sup>1, with LOD reaching 15.69 μg L<sup>−</sup>1.


**Table 2.** Summary of validation parameters of the proposed method.

\* Recovery rates were calculated according to SERS spectra of the mixed samples.

As presented in Table 2, the R<sup>2</sup> of the calibration curves are in the range of 0.9665~0.9891 for the tested pesticides. Comparing with other quantitative analysis techniques such as GC-MS and LC-MS, the linearity of this method is relatively poorer. It might be caused by surface defects on the AgNPs-PDMS substrate, which give rise to uneven enhancement of the Raman signal. Nonetheless, the coefficients of determination achieved in this study are at the same level with previously reported SERS-based methods [41,42,56]. Considering that the priority of this study is to establish a method for the rapid screening of pesticide residues, the linearity of the regression equations should be acceptable for semi-quantitative SERS analysis.

To investigate the precision of this method for pesticide detection, a series of experiments was conducted by measuring the SERS spectra of each pesticide using substrates from the same and different batches. The results are summarized in Table 2. For the tested pesticides, the RSDs are between 3.18% and 4.65% for the SERS signals obtained from the same batch of AgNPs-PDMS substrates. In the case of inter-batch precision, SERS spectra from six different batches of substrates were collected, and the RSDs varies from 5.96% to 8.27%. These results indicate that consistent SERS measurements can be performed on different substrates, and the precision of this method is acceptable for rapid and on-site detection.
