Bismuth nitrate (Bi(NO₃)₃·5H₂O), ethylene glycol (EG), and sodium molybdate (Na₂MoO₄·2H₂O) were purchased from Macklin Biochemical Co., Ltd. (Shanghai, China). l-Serine (L-Ser), glycine (Gly), and l-tyrosine (L-Tyr) were purchased from Sinopharm Chemical Reagent Co., Ltd. (China). L-Cys and glutathione (GSH) were obtained from Aladdin Reagent Inc. (Shanghai, China). Ascorbic acid (AA), sodium sulfate (Na₂SO₄), and sodium sulfite (Na₂SO₃) were purchased from Sinopharm Chemical Reagent Co., Ltd. (China). Phosphate buffer solution of 0.01 M (PBS, pH 7.4) was prepared with NaH₂PO₄.2H₂O, K₂HPO₄.3H₂O, and KCl. All chemical reagents were of analytical grade, and all aqueous solutions were prepared with ultrapure water (18.2 MΩ cm).

The PEC system consists of a CHI660E electrochemical workstation (Shanghai Chenhua Apparatus Corporation, China) and a PEAC 200A PEC reaction instrument (Tianjin Aidahengsheng Science-Technology Development Co., Ltd., China). PEC experiments and linear sweep voltammetry (LSV) curves were conducted on the PEC system using a three-electrode system: an ITO electrode with a geometric area of 0.25 cm² as the working electrode, a saturated Ag/AgCl electrode as the reference electrode, and a Pt wire as the counter electrode. The electrochemical impedance spectra (EIS) were implemented on a CHI660E electrochemical workstation in 5.0 mM K₃ [Fe(CN)₆]/K₄ [Fe(CN)₆] solution containing 0.1 M KCl. The scanning electron microscope (SEM) images were acquired from the Hitachi S-4800 SEM (Tokyo, Japan). UV-visible diffuse reflection spectra were recorded using a PerkinElmer Lambda 950 UV-visible spectrophotometer (United States). X-ray photoelectron spectroscopy (XPS) images were recorded on a K-Alpha X-ray photoelectron spectrometer (Thermo Fisher Scientific Co., Waltham, MA, United States). Fourier transform infrared (FT-IR) spectra were acquired from the Bruker TENZOR 27 spectrophotometer (Bruker Optics, Germany).

Bi₂MoO₆ was synthesized by a hydrothermal method (Dai et al., 2018). First, 0.4210 g of Na₂MoO₄·2H₂O was dissolved in 5 ml of EG under stirring for 0.5 h, and 1.6866 g of Bi(NO₃)₃·5H₂O solution was prepared in the same way. After mixing them together, 20 ml of ethanol was added dropwise under stirring. Second, the resulted solution was transferred into the Teflon-lined stainless steel autoclave, heated to 160°C for 12 h, and cooled to room temperature. Finally, the resultant product collected by centrifugation was washed three times with ethanol as well as water, dried overnight at 80°C, and then annealed at 400°C for 3 h to obtain Bi₂MoO₆ nanoparticles.

Bi₂MoO₆ suspension of 20 microliters with a concentration of 3 mg ml⁻¹ was evenly dropped onto the cleaned ITO electrode and dried at 60°C for 20 min. Afterward, 20 µL of L-Cys solution was cast onto the surface of Bi₂MoO₆/ITO gently. After the reaction at 37°C for 0.5 h, the electrode was washed with water and then immersed in 0.01 M PBS (pH 7.4) containing 0.1 M AA for PEC measurement.

The morphology of Bi₂MoO₆ was characterized using the SEM. Figures 1A,B depicted that Bi₂MoO₆ possessed a nanosheet-assembled spherical structure, and the diameters of the microsphere were less than 3 µm. The stacked sheet structure makes the material have a large specific surface area, which benefits for the subsequent ion exchange reaction and the PEC detection. After incubated with L-Cys, parts of nanosheets granulated on the microsphere of Bi₂MoO₆ (Figure 1C), indicating the interaction between Bi₂MoO₆ and L-Cys. Additionally, the elemental mapping images in Supplementary Figure S1 suggested that Bi, Mo, O, and S elements existed in the material, indicating the reaction between Bi₂MoO₆ and L-Cys.

To characterize the chemical composition and chemical state of Bi₂MoO₆ before and after reacting with L-Cys, XPS analysis was performed. As shown in Figure 2A, the elements of Bi, Mo, and O exist in Bi₂MoO₆ samples, whereas a new element of sulfur appeared after the reaction between Bi₂MoO₆ and L-Cys. Peaks in Bi 4f spectra in Figure 2B showed that two main peaks at 159.0 and 164.3 eV belong to Bi 4f_5/2 and Bi 4f_7/2 in Bi₂MoO₆ (Jia et al., 2018), shifted to 159.3 and 164.6 eV after the chemical reaction. This chemical shift originated from the formation of new bonds between bismuth and sulfur which changed the original chemical environment of bismuth atoms. The high-resolution XPS spectra of Mo 3d, S 2p, and O 1s of Bi₂MoO₆ after reacting with L-Cys were also conducted. The binding energy at 232.3, 235.4, 159.2, 164.4, and 531.1 eV pictured in Figures 2C–E were ascribed to Mo 3d_5/2, Mo 3d_3/2, S 2p_3/2, S 2p_1/2, and O 1 s, respectively. The result further witnessed the in situ formation of Bi₂S₃ on Bi₂MoO₆ (Li et al., 2020).

The optical property of Bi₂MoO₆ before and after reacting with L-Cys was studied by FT-IR spectroscopy and UV-vis DRS. As can be seen from Figure 3A, the characteristic peak at 712 cm⁻¹ existed both in the FT-IR spectrum of Bi₂MoO₆ and that after reacting with L-Cys, attributing to the symmetrical tensile vibration of the top oxygen atom of MoO₆ ⁶⁻ (Zhang et al., 2010; Li et al., 2014a; Tian et al., 2015). Compared with the FT-IR spectrum of Bi₂MoO₆, a new peak at 842 cm⁻¹ appeared in the chart of Bi₂MoO₆ after the reaction with L-Cys. This new peak corresponds to the stretching vibration of Bi–S, indicative of the formation of Bi₂S₃ through the reaction between Bi₂MoO₆ and L-Cys (Zhao et al., 2017). The UV-vis DRS in Figure 3B suggested that the formation of Bi₂MoO₆–Bi₂S₃ widened the absorption range of the light irradiation and thus is benefit for the subsequent PEC analysis.

As a photoactive material to construct the photoelectrode, the concentration of Bi₂MoO₆ plays a crucial effect on the PEC performance of the sensor. The photocurrent signal of the Bi₂MoO₆/ITO electrode constructed with varied concentration of Bi₂MoO₆ was recorded, and the photocurrent response reached a maximum value when the concentration of Bi₂MoO₆ was 3 mg ml⁻¹ (Supplementary Figure S2). So, 3 mg ml⁻¹ Bi₂MoO₆ was used for the subsequent experiments. In addition, the reaction time of Bi₂MoO₆ with L-Cys was optimized. According to Supplementary Figure S3, the photocurrent response gradually enhanced with the increase of reaction time, but the signal tended to stabilize when the reaction time reached 30 min. Therefore, 30 min was used as the reaction time.

To explore the interfacial electrochemical behavior of the biosensor, EIS analysis was conducted. As seen from Figure 4A, the bared ITO electrode displayed a small electron-transfer resistance (R _et), whereas the Bi₂MoO₆/ITO electrode gave an increased R _et because the coating of the semiconductor impedes the electron transfer. After Bi₂MoO₆/ITO was incubated with L-Cys, the R _et declined. This result may be because the in situ formation of Bi₂S₃ on the interface of Bi₂MoO₆/ITO improved the electrical conductivity of the electrode. The photocurrent responses of the sensor at different modification stages were also investigated. As illustrated in Figure 4B, almost no PEC response was shown on the bare ITO electrode, while an evident photocurrent response was observed when Bi₂MoO₆ was immobilized on the electrode. After reacting with L-Cys (10 μmol L⁻¹), the Bi₂MoO₆/ITO electrode gave a much stronger photocurrent response. This is because the compact heterostructure formed between Bi₂S₃ and Bi₂MoO₆ by in situ formation of Bi₂S₃ on Bi₂MoO₆ and the matchable band-edge levels of Bi₂MoO₆ and Bi₂S₃ could effectively accelerate the transfer of the photo-excited charge carriers. The valence band (VB) and conduction band (CB) energy levels of Bi₂MoO₆ and Bi₂S₃ were determined by the electrochemical method (Supplementary Figure S4), and the charge transfer in Bi₂MoO₆–Bi₂S₃ heterostructure is illustrated in Scheme 2. Under the light irradiation, the photo-generated electrons in the CB of Bi₂S₃ (−0.36 eV) easily transferred to the CB of Bi₂MoO₆ (−0.17 eV), whereas the holes in the VB of Bi₂MoO₆ (2.69 eV) moved to the VB of Bi₂S₃ (1.33 eV).

The PEC response of the Bi₂MoO₆/ITO electrode toward L-Cys was explored. As depicted in Figure 5A, the photocurrent intensity enhanced along with the increase in L-Cys concentration. The reason of this variation trend may be that more L-Cys increased the amount of Bi₂S₃ in situ formed on the Bi₂MoO₆/ITO electrode, thus facilitating the charge transfer and boosting the photocurrent enhancement. As demonstrated in Figure 5B, the photocurrent intensity of the sensor showed a linear relationship with the logarithm of L-Cys concentrations when the concentrations varied in the range of 5.0 × 10⁻¹¹–1.0 × 10⁻⁴ mol L⁻¹. The linear equation is I = 128.7 + 8.1 log C _L-Cys (R ² = 0.997). The limit of detection is 5.0 × 10⁻¹² mol L⁻¹. Compared with some reported methods, this method demonstrates high detection sensitivity and a wide linear range for L-Cys (Table 1). The excellent performance of the sensor can be attributed to the in situ formation of Bi₂MoO₆–Bi₂S₃ heterostructure, which possesses an excellent photoelectric response under light irradiation.

The selectivity of the sensor was evaluated by testing the PEC response of Bi₂MoO₆/ITO toward Gly, L-Tyr, L-Lys, GSH, L-Ser, SO₃ ^2-, and SO₄ ^2- and the mixture of the aforementioned substances with L-Cys (all the aforementioned solutions have a concentration of 5 μmol L⁻¹). As pictured in Figure 6A, the PEC responses of Bi₂MoO₆/ITO to Gly, L-Tyr, L-Lys, GSH, and L-Ser showed no obvious change compared with the blank solution, whereas the response of L-Cys as well as the mixture of the aforementioned interferents with L-Cys exhibited an obvious enhancement, thus demonstrating good selectivity. The reproducibility of the sensor was studied by intra-assay and inter-assay of 10 μmol L⁻¹ L-Cys. The relative standard deviations (RSDs) of intra-assay by using five Bi₂MoO₆/ITO electrodes in the same batch and inter-assay of the electrodes in different batches were 3.0 and 4.2%, respectively, indicating good reproducibility of the sensor. In addition, the photocurrent response of Bi₂MoO₆/ITO for 100 nmol L⁻¹ L-Cys within 4 weeks of storage was investigated to study the stability of the sensor. As shown in Figure 6B, the photocurrents show negligible change with RSDs less than 5.1%. The signal of this system for 15 cycles was monitored. In Supplementary Figure S5, the photocurrent was stable with a RSD of 3.2%. The data indicate the good stability of the sensor.

To explore the practical application of the sensor, seven undiluted human serum samples from Xinyang Central Hospital were measured. As listed in Supplementary Table S1, compared with the reference method (enzymatic cycling) used by the hospital, the relative errors between the reference method and this method are less than 6.1%, and the RSDs are no more than 6.2%. In addition, the standard addition test results suggest that the recoveries of L-Cys are in the range of 90.0–110.0% with RSDs less than 6.8%, as shown in Supplementary Table S2. The aforementioned results show that this method has good accuracy and feasibility.