Cannabis: Rapid nondestructive detection of mixed pesticides residues on fruit surface using SERS combined with self-modeling mixture analysis method

2.1. Materials

Gold chloride trihydrate (HAuCl₄·3H₂O, Mw: 393.83, CAS: 16961-25-4), hexadecyl trimethyl ammonium bromide (CTAB, Mw: 364.45, CAS: 57-09-0), rhodamine 6G (R6G, Mw: 479.01, CAS: 989-38-8), tetramethylthiuram disulphide (Thiram, Mw: 240.43, CAS: 137-26-8), thiabendazole (TBZ, Mw: 201.25, CAS: 148-79-8), and methanol (HPLC reagent, CAS: 67-56-1) were purchased from Aladdin Industrial Co., Ltd. (Shanghai, China). Silver nitrate (AgNO₃, Mw: 169.87, CAS: 7761-88-8), ascorbic acid (AA, Mw: 176.12, CAS: 50-81-7), and acetone (Mw: 58.08, CAS: 67-64-1) were obtained from Sinopharm Chemical Regent Co., Ltd. (Shanghai, China). Sodium borohydride (NaBH₄, Mw: 37.83, CAS: 16940-66-2) was supplied by Tianjin Fuchen Chemical Reagent Factory (Tianjin, China). Sodium oleate (NaOL, Mw: 304.44, CAS: 143-19-1), nitric acid (HNO₃) and hydrochloric acid (HCl) were provided by Shanghai Macklin Biochemical Technology Co., Ltd. (Shanghai, China). Cyclohexane (Mw: 84.16, CAS: 110-82-7) and ethanol (CAS: 64-17-5) were bought from Tianjin Fuyu Fine Chemical Co., Ltd. (Tianjin, China). Silicon wafer was supplied by Topvendor Technology Co. Ltd. (Beijing, China). Red Fuji apple (Malus pumila Mill, Shandong, China), crown pear (Pyrus bretschneideri Rehd., Hebei, China), cherry tomatoes (Lycopersicon esculentum Mill., Yunnan, China) and the swab were bought from China Resources Vanguard Supermarket (Guangzhou, China). Ultrapure water (resistivity ˃ 18.2 MΩ cm⁻¹) purified by a Milli-Q system (Millipore Co., Billerica, MA, USA) was used for all the solutions. All glassware was treated with freshly prepared aqua regia (HCl: HNO₃ = 3:1, v:v) followed by rinsing with copious amounts of water.

2.2. Preparation of Au NRs

Highly consistent Au NRs were prepared using a modified seed-mediated method with a binary surfactant mixture [[35], [47]]. In order to synthesize high-quality seed particles for Au NRs growth, the amount of NaBH₄ to be added for HAuCl₄ reduction was very important. In details, the seed solution was synthesized as follows: 5 mL of CTAB (0.1 M) solution was mixed with 125 μL HAuCl₄ (10 mM), and 0.3 mL of freshly prepared ice-cold NaBH₄ (10 mM) was injected into the mixture under vigorous stirring (1200 rpm, 2 min) by a magnetic stirrer (MS7-H550-Pro, DLAB Scientific Co., Ltd., Beijing, China), and change in colour from pale to brownish was observed with the addition of NaBH_4, and the resulting seed solution was then incubated in an incubator (DZF-6050, Shanghai Qixin Scientific Co., Ltd., Shanghai, China) at 30 °C for 2 h.

To prepare growth solution, 0.7 g CTAB and 0.1234 g NaOL were dissolved in 25 mL warm water (50 °C) in a 100 mL conical flask. After cooling to 30 °C, an amount of 1 mL AgNO₃ (10 mM) was added in the flask with continuously stirring for 5 min at 600 rpm, then kept undisturbed in the incubator at 30 °C for 15 min. Next, 25 mL of HAuCl₄ (1 mM) was mixed with the solution followed by 90 min of stirring (600 rpm), and the yellow solution became colourless. Then an amount of 2.5 mL HCl (1.0 M) was added to adjust the pH, followed by gentle stirring at 400 rpm for 15 min. After that, 80 μL of AA (0.1 M) was added into the solution under vigorous stirring at 1200 rpm for 30 s. Finally, 100 μL of seed solution was injected into the growth solution followed by stirring at 700 rpm for 1 min. The resultant mixture was kept undisturbed in the incubator (30 °C, 12 h) to complete the growth. The prepared Au NRs colloid stock solution was purified and concentrated by centrifugation (8000 rpm, 25 min) (JW-3024HR, Anhui Jia Wen Instrument and Equipment Industry Co., Ltd., Hefei, China) and washed with ultrapure water for three times to remove the residual CTAB and NaOL before further uses.

2.3. Interfacial self-assembly of Au NRs array

As shown in Fig. 1, Fig. 5, 5 mL of Au NRs colloid stock solution was concentrated to 3 mL and then put in a 5 mL beaker, which was cleaned ultrasonically in acetone for 30 min. The organic-inorganic interphase was formed by adding 1 mL of cyclohexane on the Au NRs colloid surface. By injecting 1 mL of ethanol rapidly into the system using a 1.5 mL range syringe, interfacial self-assembly reaction was induced and a gold film was formed on the cyclohexane-aqueous interphase. After cyclohexane evaporation, the Au NRs array film floating on water was transferred onto the silicon wafer (6 × 6 mm²) and allowed for drying at room temperature of 25 °C, which was used as the substrate and fixed on the Si wafer. Finally, the self-assembly Au NRs array substrate was immersed in ethanol for 30 min to remove the CTAB layer from the surface of Au NRs followed by air-dried before uses [[35], [48]].

1. Download : Download high-res image (2MB)
2. Download : Download full-size image

Fig. 1. Characterization of the prepared Au NRs colloid and Au NRs array substrates. (A) FE-SEM images of as-prepared Au NRs. (B–C) TEM images of the prepared Au NRs with 15 KX (B) and 500 KX (C) magnifications. (D) Extinction spectra of Au NRs colloid before and after interfacial self-assembly. (E) Photographs of interfacial self-assembly of Au NRs array using concentrated Au NRs colloid: (E₁) 3 mL Au NRs colloid in a beaker condensed using 5 mL Au NRs colloid stock solution; (E₂) the formation of organic-inorganic interface after 1 mL cyclohexane added on the concentrated Au NRs colloid surface; (E₃) interfacial self-assembly reaction was induced by adding 1 mL ethanol; (E₄) Interfacial self-assembled Au NRs array on the water-cyclohexane interface; (E₅) Au NRs array membrane floating on water after cyclohexane volatilization; (E₆) uplift Au NRs array membrane on a silicon wafer; (E₇) dried Au NRs array substrate. (F–G) FE-SEM images of Au NRs array with 50 KX (F) and 100 KX (G) magnifications. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

2.4. Characterization

The UV–Vis extinction spectra of Au NRs colloids were acquired from a UV spectrophotometer (UV-1800, Shimadzu Co., Kyoto, Japan). The transmission electron microscopy (TEM) images were collected from a JEM-2100 Plus high-resolution transmission electron microscope (JEOL Ltd., Tokyo, Japan) operating at an acceleration voltage of 200 kV. The field-emitting scanning electron microscope (FE-SEM) images were acquired from a scanning electron microscope (Zeiss Merlin FESEM, Carl Zeiss NTS GmbH, Oberkochen, Germany) at an acceleration voltage of 10.0 kV. The aspect ratio of Au NRs was calculated by measuring the length and width of the Au NRs from the SEM images using a particle size distribution software (Nano Measurer 1.2.5, Department of Chemistry, Fudan University, Shanghai, China, https://nano-measurer.software.informer.com/).

2.5. SERS measurements of probe molecule R6G and pesticides

In order to evaluate the SERS sensitivity, uniformity and feasibility of the proposed Au NRs array substrates developed in the current study, a series of R6G and thiram in water, and TBZ in methanol were prepared. In details, an amount of 0.5 mL solution sample was put in a 1 mL centrifugation tube and then an Au NRs array substrate was put into the tube. After incubation for 30 min, the Raman spectra was recorded using a laser confocal Raman microspectroscopy system (LabRAM HR, Horiba France SAS, Villeneuve d’ Ascq, France) equipped with a 50× long working distance (LWD) visible objective (LMPlanFL N 50×/0.50) (Olympus Co., Center Valley, PA, USA) and a 633 nm He–Ne laser (laser rated power: 17 mW) as excitation source. The laser power was attenuated to 25% for SERS measurements (4.25 mW) and the exposure time was set for 10 s for each scanning. Each spectrum was the average of 10 scans.

For Raman mapping in uniformity evaluation, Au NRs array substrates that incubated with 10⁻⁶ M R6G was laid on the x-y mapping stage. Raman spectra of the substrates with an area of 40 × 40 μm² and a step of 0.5 μm were collected in the range of 543–1683 cm⁻¹ under 50 × microscopic objective with the 633 nm laser.

For nondestructive and rapid detection of thiram, TBZ, and their mixture on the surfaces of fruits (apple, tomato, and pear), a modified surface swab method [3] was performed. In details, 50 μL of pesticide solution was spiked on the surface of the fruit (1 × 1 cm²) by a pipette followed by air-drying. Then the fruit surface was swabbed using a swab pre-soaked with methanol-water (1:1, v/v) solution, after that, the head of the swab was removed and immersed in 500 μL of the methanol-water (1:1, v/v) solution in a 1 mL centrifugation tube and vortexed for 30 s to extract the pesticides, i.e., the pesticide sample on the fruit surface was diluted in 10 times during extraction procedure. Finally, an Au NRs array substrate was put into the tube and incubated for 30 min for Raman measurements, or the 500 μL extracted pesticides solution was filtrated using a filter before HPLC analysis.

2.6. HPLC analysis of thiram and TBZ

Chromatographic analysis of thiram and TBZ was performed according to the Chinese National Food Safety Standard “Determination of thiabendazole residue in fruits and vegetables: Liquid chromatography” (GB 23200.17–2016) for TBZ, and the Chinese Entry-Exit Inspection and Quarantine Industry Standard “Determination of thiram residue in fruit and vegetable for export” (SN/T 0525–2012) for thiram, respectively. In details, the standard solution or samples of thiram and TBZ were analysed using HPLC instrument (ACQuity® Arc™, Waters Co., MA, USA) with a reverse-phase C-18 column (XBridge® BEH, 100 mm × 2.1 mm, particle size 23.5 μm, Waters Chromatography Ireland Ltd., Dublin, Ireland) equipped with a quaternary solvent manager-R (ACQuity® Arc™, Waters Pacific Pte Ltd., Singapore), a sample manager FTN-R (ACQuity® Arc™, Waters Co., MA, USA) and a dual absorbance ultraviolet detector (2998 PDA detector, Waters Pacific Pte Ltd., Singapore). The samples were eluted using a linear isocratic elution procedure of methanol and water (50/50, v/v) used as mobile phases at a flow rate of 0.417 mL/min at 25 °C. The detection wavelength was set at 270 nm for thiram and 300 nm for TBZ, respectively.

2.7. Processing of SERS spectra of mixed pesticides

All the collected Raman spectra were processed using LabSpec 6 spectroscopy software suite (Horiba France SAS, Villeneuve d’Ascq, France): de-spiking, de-noising and baseline correction.

The SERS enhancement factor (EF) was calculated using the following equation [29]:(1)Enhancementfactor=ISERS×NRamanIRaman×NSERSwhere N_SERS and N_Raman are the corresponding numbers of molecules of R6G for SERS and Raman measurements, I_SERS and I_Raman are the corresponding SERS and Raman intensities, respectively. The conventional Raman spectrum of R6G molecule (Fig. S1) and details of the calculation are available in the supporting information.

Linear calibration curves were determined by monitoring SERS intensities of characteristic peaks as a function of R6G or pesticides concentration. The LOD was determined using the following formula:(2)LOD=3Skwhere S represents the standard deviation of Raman intensity of blank samples, k is the slope of the calibration curve [49].

In this study, an alternating least squares (ALS) algorithm, which decomposes mixture data into single component spectra and contributions, was used in SMA. Initial estimates were determined using a pure variable detection method based on the SIMPLISMA approach [50]. Purity function in PLS-TOOLBOX (Eigenvector Research, Wenatchee, WA, USA) was employed to conduct a self-modelling analysis [51] of SERS spectra of pesticides mixture. The spectral data analysis procedures were executed using in-house software developed by MATLAB (MathWorks, Natick, MA, USA). The program of the SIMPLISMA used refers to the code available in literature [52]. The basic formulas for SMA are as follows:(3)D=CST+Ewhere D is an (m × n) matrix of mixed spectral data, C is the contribution matrix (m × k), S^T (k × n) is the transpose of S (n × k), which represents the matrix of pure component spectra, k is the number of components, E is the residual error. The contribution C can be estimated using the spectral intensities. As a result, S can be determined by the following equation:(4)S=DTC(CTC)−1where S represents the estimated pure component spectra. Next, the estimated contribution C* can be calculated by the following equation:(5)C*=DSSTS−1

More details for SMA in this study are available in supporting information.

3.1. Characterization of Au NRs and Au NRs arrays substrates

The prepared Au NRs colloids were characterized by FE-SEM, TEM, and UV-NIR spectroscopy and the results are displayed in Fig. 1A–D. As shown in Fig. 1A, the synthesized Au NRs showed uniform morphology, and their average aspect ratio was 3.4 with an average width diameter of 23.3 ± 1.9 nm and a length diameter of 79.8 ± 6.4 nm. Fig. 1B and C shows the TEM images of fabricated Au NRs in 15 KX (Fig. 1B) and 500 KX (Fig. 1C) magnifications with the lattice fringes of 0.234 nm (high-resolution Fig. S2 is available in supporting information), and the Au NRs were single-crystalline growing along the [001] direction [53]. The extinction spectrum of Au NRs colloids stock solution in Fig. 1D exhibited two distinct absorption characteristic bands: the transverse localized surface plasmon resonance (LSPR) at 515 nm and a longitudinal LSPR at 745 nm, due to the surface plasmonic resonance. The UV-NIR spectrum of the residual Au NRs colloids after the interfacial self-assembly showed the same LSPRs with the Au NRs colloids stock solution, with dramatically decreased absorbance intensity, revealing that the Au NRs particles in colloid were moved to organic-inorganic interphase to complete the self-assembly.

In aqueous solution, the surface plasmon resonance (SPR) absorption maximum (λ_max) of Au NRs was linearly proportional to the aspect ratio (AR) by the following relationship [54]:(6)λmax=95AR+420

Based on Equation (1) and the AR (3.4) of the Au NRs, λ_max was calculated as 743 nm, which was in line with the longitudinal LSPR (745 nm), presented in the UV- NIR spectra (Fig. 1D). Correspondingly, the theoretical AR can also be calculated as 3.4 by using Equation (6) and the longitudinal LSPR (745 nm).

Moreover, Fig. 1E presents the preparation procedure for Au NRs array by interfacial self-assembly using as-prepared Au NRs colloids. In details, the condensed Au NRs colloid (Fig. 1E₁) as the inorganic phase and the cyclohexane as the organic phase formed distinct interphase (Fig. 1E₂). Then, the inducer ethanol was rapidly squirted into the system (Fig. 1E₃) to decrease the surface charge of Au NRs, facilitating the Au NRs-water surface tension and driving the Au NRs rapidly to the cyclohexane-aqueous interphase, which eventually formed a compact Au NRs array (Fig. 1E₄) [29]. With the evaporation of cyclohexane spontaneously at the room temperature, the Au NRs array film that floating on aqueous phase (Fig. 1E₅) was transferred onto the silicon wafer (6 × 6 mm²) by a tweezer (Fig. 1E₆). After air-drying, the Au NRs array was fixed on the silicon wafer with a golden appearance (Fig. 1E₇). The FE-SEM images of the Au NRs array displayed in Fig. 1F and G and Fig 2B₃-D₃ further demonstrated that Au NRs were closely packed on the silicon wafer with a smooth, dense and crack-free morphology, which would provide abundant hot-spots for enhancing the Raman scattering signals of analytes on the array.

1. Download : Download high-res image (1MB)
2. Download : Download full-size image

Fig. 2. SERS enhancement effect, morphology and uniformity of Au NRs array substrates affected by the volume of Au NRs colloid stock solution used in interfacial assembly. (A1-A3) SERS spectra of R6G (10⁻⁶ M) collected on the Au NRs array substrates with an increase in Au NRs colloid stock solution (3–5 mL). (B1–B3, C1–C3, D1-D3) corresponding FE-SEM image of the Au NRs array substrates with different magnifications (20–100 KX); (E1-E3) SERS intensity distribution of R6G at 1510 cm⁻¹ collected from 100 randomly spots on the corresponding Au NRs array substrates.

3.2. SERS performance, uniformity and reproducibility of Au NRs arrays

Under a specified volume of the container (5 mL beaker) and reactants, the structure, morphology and SERS performance of the Au NRs array depended on the concentration of the Au NRs colloid. As shown in Fig. 2, the morphology of the Au NRs array was tuned by adjusting the volume of the Au NRs colloid stock solution from 3 to 5 mL, and the SERS performance and uniformity were evaluated using Raman probe molecule R6G. For the interfacial self-assembly reaction using 3 mL prefabricated Au NRs colloid stock solution, the SERS intensity of R6G at 1510 cm⁻¹ was 7985 ± 1081 (Fig. 2A₁) with the relative standard deviations (RSD) of 13.53%, collected from 100 randomly selected spots on the Au NRs array substrates (Fig. 2E₁), the Au NRs were not packed closely, and there were many cavities in the array (Fig. 2B₁-D₁). When 4 mL Au NRs colloid stock solution (re-dispersed in 3 mL ultra-pure water after centrifugation in wash procedure) was used, the SERS intensity of R6G at 1510 cm⁻¹ was 19272 ± 4121 (Fig. 2B₁), which was increased by more than two-folds, however, the morphology of the substrates was rough (Fig. 2B₂-D₂) with RSD of 21.38% (Fig. 2E₂), and such an RSD was too high for practical applications as quantification analysis required RSD to be less than 20% [29]. When the Au NRs colloid stock solution was increased to 5 mL, no obvious cavities in the array were observed, showing that Au NRs were closely assembled with a compact and crack-free appearance (Fig. 2B₃-D₃), and the SERS intensity of R6G at 1510 cm⁻¹ was 43185 ± 4009 (Fig. 2A₃) with RSD of 9.28% (Fig. 2E₃), demonstrating that this could be the ideal substrate with excellent SERS activity and uniformity. Therefore, 5 mL Au NRs colloid stock solution was selected for the interfacial self-assembly fabrication of Au NRs array in the following investigation.

For further verification of the uniformity of the fabricated Au NRs array substrates, Raman mapping was carried out on a selected area of 40 × 40 μm² on the substrate with a resolution of 0.5 μm. Fig. S3 displays the recorded 6561 SERS spectra (81 × 81 spots) of R6G (Fig. S3A) incubated with the Au NRs array substrate and their Raman mapping images (Fig. S3C). The colour distribution displayed good uniformity and without obvious fluctuations. The RSDs by targeting Raman intensities at 612, 772, 1182, 1311, 1362, 1510 and 1650 cm⁻¹ were calculated to be 15.02%, 13.81%, 13.40%, 14.90%, 13.45%, 13.57%, 14.77%, respectively. These values were higher than those from 100 randomly selected spots on the Au NRs array substrate (9.29% at 1510 cm⁻¹) as the area illuminated by the laser (1.86 μm²) was larger than the step (0.5 μm) for Raman mapping. This phenomenon could be ascribed to the possible destruction of the analyte on the substrate caused by continuous laser irradiation during Raman mapping, leading to the fluctuation of Raman intensities between the centre and the edge of the mapping area. Nevertheless, these results revealed that the Au NRs array showed excellent uniformity and could serve as reliable SERS substrates.

Furthermore, for the batch-to-batch reproducibility of the Au NRs array substrates evaluation, the SERS spectra of R6G at 90 random spots (3*3*10) from three batches of Au NRs colloids and nine batches of Au NRs array were collected, and their SERS intensity distribution at 1510 cm⁻¹ are presented in Fig. S3B with the average of 41524 ± 5928 and the RSD of 14.28%. Additionally, the SEM image of the Au NRs array substrate after incubation with the sample and SERS analysis is shown in Fig. S4. The Au NRs array had no significant difference from that before detection, indicating no occurrence of lixiviation of the Au NRs from the substrate. Therefore, the ideal SERS performance with good uniformity and reproducibility confirmed that the Au NRs array could serve as excellent SERS substrates for practical applications.

3.3. Detection of thiram and TBZ in standard solutions

To examine the performances of the prepared Au NRs array substrates in practical uses, thiram and TBZ were quantitatively detected by SERS method using R6G as the Raman probe molecule. Fig. 3A shows the SERS spectra of R6G solution with R6G concentrations from 1 × 10⁻⁹ to 1 × 10⁻⁶ M (1 × 10⁻³ μM–1 μM). From the spectra of the blank control sample, weak background of Au NRs array was observed, and the slight bulge at 1444 cm⁻¹ was attributed to C–N binding mode of CTAB (Fig. S5) on the surface of Au NRs [55], which almost disappeared when detecting the analytes. With an increase in the concentrations of R6G, characteristic peaks at 612, 772, 1182, 1311, 1362, 1510, 1578 and 1650 cm⁻¹ were clearly observed. The assignments of these peaks are listed in Table S1 [56]. Considering the intensity and stability, the peak at 1510 cm⁻¹ was selected to establish the relationship of the SERS intensity as a function of R6G concentrations. The SERS intensity is linearly dependent on the R6G concentration in a narrow range but showed a logarithmic relationship in a wide concentration range. It is attributed to the saturation of the analyte molecules on the substrate at high concentrations. A similar phenomenon can also be found for thiram (Fig. 3E and F), TBZ (Fig. 3H and I) and in the literature [38]. Therefore, piecewise functions were adopted to express the relationship of the SERS intensity and analyte concentrations. Fig. 3 B–C demonstrate that the SERS intensity at 1510 cm⁻¹ showed a linear correlation with the logarithmic R6G concentrations ranging of from 1 × 10⁻³ to 1 × 10⁻¹ μM and the R6G concentrations from 0.1 to 1 μM. The calibration curves were Y₁₅₁₀ = 3369.15 Lg(X) + 10632.22 and Y₁₅₁₀ = 41447.5X + 3192.6 with coefficients of determination (R²) of 0.9883 and 0.9981 (Table S2), respectively. The LOD was calculated as 8.3 × 10⁻⁴ μM. Furthermore, the EF of the Au NRs array substrates was calculated as 1.64 × 10⁶ using the Raman probe molecule R6G, and details of the calculation are available in the supporting information.

1. Download : Download high-res image (1MB)
2. Download : Download full-size image

Fig. 3. SERS spectra of R6G (A) in water, thiram (D) in methanol, and TBZ (G) in water with different concentrations collected on the Au NRs array substrates. (B–C, E-F, H–I) Corresponding SERS intensity of R6G at (B–C) 1510 cm⁻¹, thiram at (E–F) 1378 cm⁻¹, TBZ at (H) 778 cm⁻¹ and (I) 1007 cm⁻¹ as a function of the concentrations of R6G, thiram, or TBZ, respectively.

Thiram and TBZ were further detected on the Au NRs array substrates. Fig. 3D, G displays the SERS spectra of thiram (Fig. 3D) with concentrations from 1 × 10⁻¹⁰ to 1 × 10⁻⁶ M (1 × 10⁻⁴ to 1 μM) and TBZ (Fig. 3G) with concentrations from 1 × 10⁻⁹ to 1 × 10⁻⁵ M (1 × 10⁻³ to 10 μM), respectively. The SERS intensities of the corresponding characteristic peaks at 562, 932, 1144, 1378, 1512 cm⁻¹ for thiram, and 778, 880, 1007, 1278, 1321, 1578, 1601 and 1622 cm⁻¹ for TBZ increased with increases in the concentrations of thiram and TBZ increase. The assignments of the peaks of thiram and TBZ are also listed in Table S1 [1,[57], [58], [59]]. For the detection of thiram, the most intense peak at 1378 cm⁻¹ was attributed to C–N stretching and CH₃ deformation, which was selected to build the calibration curves of SERS intensities versus thiram concentrations. Fig. 3 E-F exhibit that the SERS intensity at 1378 cm⁻¹ showed linear correlations (Table S2) with the logarithmic concentrations of thiram from 1 × 10⁻⁴ to 1 μM and with the concentrations of thiram from 1 to 10 μM. The LOD of thiram was calculated to be 3.6 × 10⁻⁵ μM. For the detection of TBZ, the peak at 778 cm⁻¹ was assigned to in-plane CN bending and C–S stretching, and the peak at 1007 cm⁻¹ was attributed to C–C stretching and C–N stretching, and both were selected to establish the relationships of SERS intensities with concentrations of TBZ. Fig. 3 H–I present that the SERS intensities at 778 and 1007 cm⁻¹ displayed linear correlation (Table S2) with the logarithmic concentrations of TBZ from 1 × 10⁻³ to 10 μM, and the LOD of TBZ was calculated to be 1.03 × 10⁻³ μM and 9.5 × 10⁻⁴ μM, respectively. Additionally, it was noticed that the sensitivity for thiram detection was significantly higher than that of TBZ, indicating that more Au–S bond formed on the substrate during thiram detection. These results indicated the accuracy of SERS detection based on the Au NRs array, and its high sensitivity made the high-density Au NRs array a promising substrate for practical food safety detection.

3.4. SERS and HPLC detection of thiram and TBZ on the fruit surface

Fig. 4 shows the residues of thiram and TBZ on the surface of the apple, pear and tomato as detected using SERS technology combined with the modified surface swab method. The characteristic peaks of thiram at 1378 cm⁻¹ and of TBZ at 1007 cm⁻¹ were chosen to build the calibration curves (Table S2) for SERS intensities versus the concentrations of thiram (Fig. 4B_1-3) or TBZ (Fig. 4D_1-3), respectively. Two intense and clear peaks could still be identified when the concentrations of the thiram (Fig. 4A_1-3) and TBZ (Fig. 4C_1-3) were decreased to ppb level. The LODs of pesticides on the surface of apple, tomato and pear were 0.81, 0.58, 0.93 ppb for thiram, and 15.89, 15.24, 15.38 ppb for TBZ, respectively. The difference in LODs might be due to different absorptivity on pericarp structures of apple, tomato and pear. The LODs of pesticides on the fruit surface can also be expressed as mass-to-area ratio (0.041, 0.029 and 0.047 ng/cm² for thiram, and 0.79, 0.76 and 0.80 ng/cm² for TBZ on the surface of apple, tomato and pear, respectively), or mass of pesticide to mass of fruit (0.039, 0.073 and 0.045 ng/g for thiram, and 0.76, 1.91 and 0.78 ng/g for TBZ on the surface of apple, tomato and pear, respectively), which are listed in Table S3. Details for calculating the concentration conversion are available in the supporting information.

1. Download : Download high-res image (2MB)
2. Download : Download full-size image

Fig. 4. SERS spectra of thiram (A1-A3) and TBZ (C1–C3) extract from (A1, C1) apple, (A2, C2) tomato, and (A3, C3) pear surface with different concentration collected on the Au NRs array substrates, respectively. (B1–B3, D1-D3) Corresponding SERS intensities of thiram at (B1–B3) 1378 cm⁻¹ and TBZ at (D1-D3) 1007 cm⁻¹ as a function of the logarithmic thiram or TBZ concentrations on (B1, D1) apple, (B2, D2) tomato, and (B3, D3) pear surface, respectively.

1. Download : Download high-res image (2MB)
2. Download : Download full-size image

Fig. 5. Schematic diagram of process and outcome of self-modelling mixture analysis of SERS spectra of pesticides mixture.

Nevertheless, the LODs were lower than both the maximum residue limit (MRLs) for thiram (5 ppm) and TBZ (12 ppm) in fruits prescribed by the USA Environmental Protection Agency (EPA, 2020) [60], and for thiram (5 mg/kg) and TBZ (3 mg/kg) in fruits prescribed by GB 2763–2016 “China National Food Safety Standard: Maximum Residue Limits for Pesticides in Food”, respectively. The comparison of the current results with those reported in the literature is listed in Table S4.

In addition, HPLC analysis was adopted to validate the SERS method. Fig. S6 shows the retention time (Fig S6 A, D) and standard curve of thiram (Fig S6 B-C) and TBZ (Fig S6 E-F) in methanol, respectively. Considering that the swab cannot release all the pesticides in the extract procedure, the extraction efficiency was calculated to calibrate the recovery on the fruit surface by using the HPLC standard curve of thiram and TBZ in methanol. To calculate the extraction factor, 50 μL of 100 ppm thiram or TBZ solution was dropped on a clean glass slide and dried. The same procedure with the swab test on fruit surface was carried out. The final pesticides concentration was calculated using the standard curve. The average extraction factor was calculated to be 72.1% for thiram and 65.8% for TBZ using the following equation:(7)Extractionfactor=CD×VDCS×VS×100%and the calibrated recovery on the fruit surface was calculated by the equation below:(8)Calibratedrecovery=CD×VDExtractionfactor×VSwhere C_D is the detected concentration of pesticide on glass or fruit surface, C_S is the spiked concentration of pesticide on glass or fruit surface, V_D is the detected volume of pesticide solution extract from swab, V_S is the spiked volume of pesticide on glass or fruit surface, respectively.

3.5. Spectral analysis of pesticides mixture using SMA method

For simultaneous measurements of mixed thiram and TBZ, the SMA method was adopted to analyse the obtained Raman spectra from the mixture according to the following equations:(9)IpureA∗A+IpureB∗B+…=Mixture1IpureA∗A+IpureB∗B+…=Mixture2⋯IpureA∗A+IpureB∗B+…=Mixturenwhere A, B, …, are pure spectra of each single component, I_{pure A}, I_{pure B}, …, are contributions of A, B, …, respectively.

Fig. 5 shows the self-modelling mixture analysis of the SERS spectra for the mixed pesticides. In the current study, singular value decomposition (SVD) was used to find the number of components [41]. In order to resolve a spectral dataset with SMA, pure variables should be present [43]. For the presence of overlapped peaks in spectra, to avoid selecting poor pure variables that would result in inferior resolution results, second-order derivatives were introduced [44] and to decrease/suppress any ambiguity in the SMA solution, constraints were introduced [61]. Details of the analysis are available in the supporting information. Finally, the original SERS spectra of the mixture and its corresponding decomposed pure spectra (variables multiplied by scores) of thiram and TBZ are displayed in Fig. 5, in which the blue, pink, and purple rectangles represent the characteristic peaks of thiram, TBZ, and their overlapped peaks, respectively. The characteristic spectra and fingerprint peaks of each individual pesticide were clearly resolved and presented. Results showed that the acquired SERS spectra of each pesticide in the mixture had no obvious variation compared with the spectra of the corresponding pure pesticide, indicating the feasibility of SMA method for extracting the pure spectra from spectral mixture data in detecting complex matrixes.

By taking the advantages of the fingerprint-based SERS technique and the powerful pure spectra extracting method of SMA, mixed pesticides spiked on the surface of the apple, pear and tomato were simultaneously detected and decomposed (Fig. 6A–C), and the resolved individual pure spectra are displayed in Fig. 6A₁-C₁ for TBZ and in Fig. 6A₂-C₂ for thiram, respectively. The characteristic peaks of thiram at 1378 cm⁻¹ and of TBZ at 1007 cm⁻¹ were selected to calculate the recovery of the two pesticides on the surface of three fruits. Compared with the mixture spectra, satisfactory recoveries of 66.0–142.54% for detecting thiram and 83.06–136.76% for detecting TBZ at an appropriate proportion of the pesticides (Table 2) were obtained, revealing that SERS technique combined with SMA method had the potential in practical applications for multiple-analytes sensing in complex systems.

1. Download : Download high-res image (2MB)
2. Download : Download full-size image

Fig. 6. SERS spectra of thiram and TBZ mixture (A, B, C) extracted from (A) apple, (B) tomato, and (C) pear surface. The decomposed pure spectra of individual TBZ (A1, B1, C1) and thiram (A2, B2, C2). B is for TBZ and H is for thiram.

Table 2. Recoveries for detecting the mixture of thiram and TBZ on surface of apple, tomato, and pear using SERS technique combined with SMA method.

Samples	Spiked pesticides (ppm)	Detected TBZ (ppm)	Detected thiram (ppm)	Recovery of TBZ (%)	Recovery of thiram (%)
Apple	B200 H0.3	198.18	0.3404	99.09	113.47
	B30 H0.01	41.03	0.0143	136.76	142.54
	B3 H0.02	3.38	0.0204	112.76	101.96
	B2 H0.2	2.47	0.2122	123.55	106.09
	B1 H0.1	1.14	0.0750	114.19	75.00
	B0.1 H0.01	0.13	0.0132	129.33	132.18
Tomato	B200 H0.3	219.10	0.3739	109.55	124.62
	B100 H0.1	88.91	0.0832	88.91	83.25
	B5 H0.05	5.27	0.0330	105.33	66.00
	B5 H0.02	5.29	0.0174	105.75	87.01
	B1 H0.5	1.29	0.3950	129.21	79.00
	B0.1 H0.005	0.12	0.0038	123.89	75.78
Pear	B200 H0.1	220.07	0.0989	110.04	98.89
	B100 H5	83.06	4.0147	83.06	80.29
	B30 H0.5	39.21	0.5318	130.70	106.36
	B5 H0.05	5.24	0.0480	104.76	95.92
	B3 H0.001	2.93	–	97.73	–
	B1 H0.02	1.30	0.0263	129.74	131.48

Note: B is for TBZ and H is for thiram.