Surface treatment optimization of silicon wafer
In this experiment, the sample was dropped onto a silicon wafer for detection. Silicon wafers need to be properly processed before use because the new silicon wafer has a natural oxide layer and other impurities on the surface, while the surface hydrophilicity and hydrophobicity of the used silicon wafer often change, resulting in reuse failure. Commonly used silicon wafer cleaning fluids were investigated, and the CBZ standard solution was used to evaluate the SERS performance of the silicon wafer after surface cleaning (Table 1). After processing with cleaning fluids ① and ②, the edge of the sample droplet on the silicon wafer shrank rapidly under laser irradiation, and it was difficult to collect SERS signals. When the silicon wafer was treated with cleaning fluids ③, ④, and ⑤, the edge of the sample infiltrated the silicon and shrunk slowly with laser irradiation. The SERS spectra were obtained under the processing conditions of solutions ④ and ⑤.
Table 1
SERS enhanced performance of silicon wafers with different hydrophilic and hydrophobic treatments.
Cleaning fluid number | Cleaning fluid composition | Volume ratio | Incubation time/temperature | SERS detection performance |
① | H2SO4/H2O2 | 3:1 | 10min/75℃ | The edge of the droplet shrinks quickly, unable to successfully detect the signal |
② | H2O2 /H2O | 1:1 | 10min/75℃ | The edge of the droplet shrinks quickly, unable to successfully detect the signal |
③ | NH4OH/H2O2/ H2O | 1:1:5 | 10min/75℃ | Edge infiltration, unable to successfully detect the signal |
④ | HF/H2O | 1:50 | 15s /25℃ | Edge infiltration, Good SERS signal, noise interference |
⑤ | HCL /H2O | 1:6 | 10min/75℃ | Edge infiltration, Good SERS signal |
Surface wetting occurs when a liquid sample comes into contact with a solid silicon wafer [19]. The contact angle was used to assess the degree of surface wetting; the smaller the contact angle, the more hydrophilic the solid surface and the better the wetting performance, and vice versa. In this study, we found that the difference in the hydrophilicity and hydrophobicity of the silicon wafer surface affected the peak intensity and peak shape of the SERS spectrum of CBZ. If the contact angle was too small, the edge shrank too quickly; thus, the SERS spectrum could not be obtained. When the contact angle on the surface of the silicon wafer was 75.5° (treated by cleaning fluid ④), the laser could not be focused well, and the droplet edge was easily boiled by the laser (Fig. 1a), resulting in a very large signal-to-noise ratio on the SERS spectrum. Moreover, some impurity signals were present. These problems could be resolved when the silicon wafer was treated with cleaning fluid ⑤; the contact angle was lowered to 61.5°, and the SERS spectrum could be collected easily. Thus, cleaning fluid ⑤ was used in subsequent experiments. The silicon wafer was cut into small pieces of 1 × 1 cm for the sample to be tested for each piece detection to avoid mutual interference between the sample droplets (Figure S1).
Optimization of plasma pretreatment conditions
Protein precipitation method
For biological samples rich in protein, the large amount of protein that interferes with the determination should be removed during purification. The protein precipitation method is commonly used for pre-processing biological samples. During the analysis process, the protein precipitation agent can be selected according to the different characteristics of the pre-separated sample.
Methanol and acetonitrile are the most commonly used protein precipitation agents. These two agents were used to treat the simulated plasma samples. The protocol was as follows: A protein precipitator (300 µL) was added to the simulated sample, followed by vortexing (5 min) and centrifugation at 10,000 rpm for 20 min. The upper fluid was collected and gently evaporated to dryness using a stream of nitrogen. The supernatant was extracted after dissolving the product in methanol. The SERS spectrum of the supernatant obtained using the protein precipitator methanol is shown in Fig. 2b. By comparison with the SERS spectrum of the blank control samples (Fig. 2a), it was found that the detected signals were derived from blank samples. If the protein precipitator was changed to acetonitrile and the same protocol was followed, another simulated sample was obtained. However, interference still existed in the SERS spectrum, indicating that protein precipitation was unsuitable for SERS detection. Although most of the proteins were removed by the protein precipitation method, there were still many blank SERS signals owing to poor selectivity and serious matrix effects, which could not meet the requirements of spectral detection.
Liquid–liquid extraction method
The liquid–liquid extraction method exploits the difference in solubility or partition coefficient between the target substance and the water–immiscible organic solvent to transfer the target substance from the biological matrix to the organic solvent. Compared to the protein precipitation method, this method showed a lower matrix effect and better repeatability. Two commonly used reagents, ethyl acetate and chloroform, were used in this study. Chloroform is an easily soluble solvent for the target analyte, CBZ. The experimental procedure for liquid–liquid extraction was as follows. Fifty microliters of 5% ammonia water were added to the simulated sample by vortexing for 30 s. Extraction was performed by adding 2 mL of extracting agent, followed by vortexing for 3 min and centrifugation at 8000 rpm for 5 min. The upper layer was blow-dried with nitrogen and reconstituted with methanol. SERS spectra of the extracted products from the simulated sample and blank plasma were collected (Fig. 3). The SERS signals from the blank group (Fig. 3a and c), especially those extracted with ethyl acetate, were much lower than those from the protein precipitation method, indicating that protein interference was greatly reduced by the liquid–liquid extraction method. However, SERS signals originating from the ethyl acetate group could not be found in the spectrum of the simulated sample; this may be because of residual protein signal interference. It was clear that SERS peaks, such as 697, 719, 1021, 1038, 1561, and 1619 cm− 1, originating from CBZ appeared in the spectrum of the chloroform group; thus, chloroform was applied for liquid–liquid extraction in the following experiment.
Detection of simulated samples
In this study, silver colloids prepared by the classic Lee–Meisel method were applied as SERS substrates because of their simple synthesis in batches at a relatively low cost. When the original concentration of the silver colloids was used, the intensity of the CBZ SERS spectrum was weak. The SERS intensity increased when the concentration of the silver colloids was doubled; however, the stability was not ideal. During the optimization process, we found that the group of 2× concentration silver colloids in KI solution presented a stable SERS signal with high resolution (Figure S2). This detection condition was also applied to the two metabolites of CBZ, CBZE and CBZD. Figure 4 shows the structures and their SERS spectra.
The three analytes showed remarkable similarities in most regions of the SERS spectra because of their extremely similar structures. As shown in Fig. 4B, the SERS spectra of CBZ, CBZE, and CBZD presented many signals in the spectral range of 300–1700 cm− 1. The SERS spectrum of CBZ showed SERS features at 580, 719, 807, 1220, and 1620 cm− 1 while neither of its metabolites peaked at these positions. The characteristic peak at 719 cm− 1 presented the strongest intensity. Thus, the peak at 719 cm− 1 was regarded as a unique qualitative and quantitative detection peak of CBZ. In the spectrum of CBZE, two peaks at 697 and 758 cm− 1 appeared simultaneously, which can be used as the characteristic signals of metabolites together. The SERS peak at 697 cm− 1 and 758 cm− 1 were used as the characteristic peaks of CBZE and CBZD, respectively, to evaluate the lowest detection concentration.
Under the optimized parameters, the simulated positive samples of different concentrations were detected with a deposition of 1µL. The limit of detection (LOD) was calculated using a signal-to-noise ratio (S/N) of 3: the specific LOD values were 0.01 µg·mL− 1 for CBZ, 1 µg·mL− 1 for CBZE, and 1 µg·mL− 1 for CBZD (Figure S3). The established method exhibited good specificity and acceptable sensitivity for CBZ and its metabolites. Thus, it can be applied to the detection of real plasma samples.
Quantitative analysis of CBZ in rat plasma
It is well known that quantitative analysis using the SERS method is challenging [20]. However, significant efforts have been made [21, 22]to obtain convincing results [21, 22]. The design and optimization of SERS substrates with high sensitivity and reproducibility are currently prevalent [16, 14, 23]. Different SERS substrates cause significant variations in peak intensities, which may dramatically affect the accuracy of the quantitative results. Methanol [24], sodium thiocyanate [25, 26], and melamine [27] have often been used as internal standards for SERS quantitative analyses. In this study, sample detection was conducted on silicon wafers, which exhibited low effects on the characteristic SERS peaks of the sample. The stable and unique Raman scattering at 521 cm− 1 derived from silicon was designated as a quantitative internal standard peak, which eliminated the tedious operation of adding internal standards. The inherent internal standard method simplifies the preparation process based on the embedded internal standard method and retains stability and quantitative accuracy. The design of purposefully selecting a special substance to modify the substrate has strong controllability, operability, and practicality.
The SERS intensity ratios of peaks 719 and 521 cm− 1 from CBZ and internal standard silicon, respectively (I719/I521), were used for the quantitative analysis. The relationship between the standing time of the mixed solutions and I1040/I2120 revealed that the intensity ratio was fairly stable during a standing time of 60 min. This observation indicates that mixed treatment with the silver solutions used in this study was feasible. A series of gradient concentrations of CBZ in plasma, ranging from 2.5 to 40 µg·mL− 1, were measured using the established method. SERS spectra were recorded with a laser irradiating the edge of the analyte and the silver solid mixture. Figure 5 depicts the measured spectra for a representative selection of CBZ concentrations (2.5, 5, 10, 20, 30, and 40 µg·mL− 1). The relative intensity of the peak at 719 cm− 1 basically showed a concentration-dependent manner. The corresponding linear relationship is shown in Fig. 5b. The calibration curve presented a good linearity with a correlation coefficient (R2) of 0.9913 and small error with an average relative standard deviation (RSD) of 3.19% in the range of 2.5–40 µg·mL− 1. Moreover, methodological evaluation using three plasma samples spiked with CBZ showed that the recoveries were in the range of 85.26–112.73%, with an RSD of 8.50% (Table S1).
Four quality control samples of high, medium, low, and lower limits of quantification were prepared according to the method described in Section 2.4, each of which had six samples in parallel. The concentration of CBZ was calculated according to the established standard curve, and accuracy and precision were obtained (Table S2). The accuracy (RE) was in the range of 95.45–112.49%, and the precision RSD was less than 17%, which is in line with the 2020 edition of the Pharmacopoeia for biological samples. The quality control samples with two concentrations of 5 and 30 µg·mL− 1, representing low and high concentrations, respectively, were processed under the following experimental conditions for the stability investigation experiment (Table S3), each of which had three samples in parallel. The results showed that the stability test met the requirements under all treatment conditions. The results of the methodological investigation indicated that the calibration curve had the potential for use in the quantitative analysis of CBZ in plasma samples while efficiently predicting the concentration of the unknown sample.
The established method was used to analyze a real sample of CBZ in rat plasma. Plasma samples collected at different time of 0.5, 1, 2, 3, 5, 7, and 9 h after administration were analyzed. The sample concentrations were calculated according to the calibration curve. The average plasma concentration–time curve of CBZ was plotted using the blood sampling time as the abscissa and the corresponding concentration values as the ordinate (Fig. 6). The results showed that CBZ could be quickly absorbed into the blood circulation, and the blood concentration reached a peak of approximately 17.96 µg·mL− 1 at 2 h after of administration. The steady-state plasma concentration of CBZ was approximately 14.97 µg·mL− 1 after the last administration.