Self-Powered Photoelectrochemical Assay for Hg 2+ Detection Based on g-C 3 N 4 -CdS-CuO Composites and Redox Cycle Signal Ampliﬁcation Strategy

: A highly sensitive self-powered photoelectrochemical (spPEC) sensing platform was constructed for Hg 2+ determination based on the g-C 3 N 4 -CdS-CuO co-sensitized photoelectrode and a visible light-induced redox cycle for signal ampliﬁcation. Through successively coating the single-layer g-C 3 N 4 , CdS, and CuO onto the surface of an electrode, the modiﬁed electrode exhibited signiﬁcantly enhanced PEC activity. The microstructure of the material was characterized by scanning electron microscopy (SEM), X-ray diffraction (XRD) and Fourier transform infrared spectroscopy (FT-IR). However, the boost in photocurrent could be noticeably suppressed due to the consumption of hole-scavenging agents (reduced glutathione) by the added Hg 2+ . Under optimal conditions, we discovered that the photocurrent was linearly related to the Hg 2+ concentration in the range of 5 pM–100 nM. The detection limit for Hg 2+ was 0.84 pM. Moreover, the spPEC sensor demonstrated good performance for the detection of mercury ions in human urine and artiﬁcial saliva.


Introduction
It is well known that heavy metals are very harmful to the human body. Among such metals, mercury is the most widely used heavy metal, and it is used in the manufacture of compact fluorescent-lamp light bulbs, jewelry, thermometers, blood pressure monitors, and silver dental fillings [1]. Mercury ions (Hg 2+ ) are the general form of mercury and these can cause various diseases, such as hypertension, enteritis, bronchitis, and pulmonary edema [2]. At present, several methods have been utilized for the detection of mercury ions, including fluorescent and colorimetric sensors [3][4][5], surface-enhanced Raman scattering [6], and electrochemistry [7,8]. Although these methods have the advantages of high sensitivity and accuracy [9], they are limited by complex sample pretreatment and high operation cost [10]. Therefore, developing a mercury-ion detection method with a facile pretreatment process, and low cost, is imperative.
Photoelectrochemical (PEC) sensors demonstrate significant potential for analysis owing to advantages such as weak background noise, simple operation, high sensitivity, and quick response [11]. The self-powered photoelectrochemical (spPEC) sensor is an emerging photoelectrochemical detection method owing to its portability, lack of external power supply, and sustainability [12]. The spPEC sensor is primarily based on a three-electrode system, including the type of anode or cathode. The three-electrode scheme adopts a single electrode system, including the type of anode or cathode. The three-electrode sch adopts a single working electrode as the signal source and a specific recognition platf [13]. However, the interaction between the optical electrode and the reducing agent o target molecule inevitably leads to deterioration of the performance of the PEC se [14]. Cao et al. [15] constructed an ingenious visible light-induced membraneless powered PEC biosensing platform by integrating a signal amplification strategy for bioanalysis. Owing to the simple enzyme-induced chemical redox cycle process, the si could be effectively and repeatedly regenerated through coupling reduction and ox tion reactions, which were used for ultrasensitive PEC analysis.
Therein, the selection of photoactive materials is the key to improving the detec of photoelectrochemical sensors. Currently, g-C3N4 is considered as the photoactive terial with the highest potential owing to its advantages such as suitable band gap, ex lent stability, and low toxicity [16]. Nevertheless, the photocatalytic efficiency of sin phase g-C3N4 is low owing to the lack of surface redox active sites and the high recom nation rate of electron-hole (e − -h + ) pairs [17]. To effectively improve the photocata activity of g-C3N4, it is generally combined with other photoactive materials with lo band gaps. As n-type narrow-band gap semiconductors, CdS quantum dots (QDs) often used for the study of visible light-active materials [18]. However, CdS QDs are only photocorrosive [19] but also exhibit a high electron-hole recombination rate which can be resolved by combining them with other p-type photoactive materials. C nanoparticles (NPs) are p-type semiconductors [21] that have been widely employe photocatalysts for different reactions owing to their high conductivity and low band [22]. Therefore, combining n-type CdS and p-type CuO to form composites [23] may s as an effective strategy for enhancing stability and inhibiting electron-hole recombina Herein, we propose a spPEC assay for Hg 2+ detection based on g-C3N4-CdS-C composites and a redox cycle signal amplification strategy. As presented in Scheme the g-C3N4-CdS-CuO material is prepared via layer-by-layer self-assembly as a self-p ered system. A fluorine-doped tin oxide (FTO) electrode is modified using the compo material as a photoanode. As illustrated in Scheme 1B, under irradiation, glutath (GSH) is oxidized by CuO to form oxidized glutathione (GSSG) for generating a str photocurrent owing to the GSH-oxidase and peroxidase-like activities of CuO NPs [24 In the presence of GSH reductase (GR), GSSG is reduced to GSH using a reductive c zyme II (NADPH) as a substrate, resulting in a redox cycle system [26], signal ampli tion is achieved through redox cycle. However, in the presence of Hg 2+ , GSH can bin Hg 2+ based on the hard-soft acid-base (HSAB) theory [27], which influences the ab redox cycle system and significantly reduces the photocurrent. Based on these results can conclude that the g-C3N4-CdS-CuO composite was successfully used to develo spPEC senor for the detection of Hg 2+ in human urine and artificial saliva. Scheme 1. Schematic diagram of electrode preparation process (A) and schematic diagram of r cycle on photoanode of self-powered device and test in actual sample (B).

Photoelectrochemical Measurement
Prior to the photoelectrochemical measurement, g-C 3 N 4 and CdS were prepared. The detailed preparation processes of these composites are illustrated in Support Information. Firstly, the bared FTO was cleaned ultrasonically in water, ethanol and acetone in sequence and dried at 60 • C. Then, 20 µL g-C 3 N 4 suspension (1.0 mg mL −1 ) was dropped on the FTO electrode with an active area of 0.25 cm 2 and dried at room temperature. After that, 20 µL CdS (5 mg mL −1 ) and 20 µL CuO (1.0 mg mL −1 ) were dropped on the FTO in the same operation, successively, to form layer upon layer of electrode materials. A three-electrode system was used to measure the photocurrent in CHI660E Electrochemical Workstation. The modified FTO was used as the working electrode, Pt wire as the counter electrode, and the saturated Ag/AgCl electrode as the reference electrode. GSH (15 mM, 1 mL), GR (170 u mg −1 , 2 µL) and NADPH (10 mg mL −1 , 250 µL) were added into the 0.1 M phosphate buffer solution (PBS) (5 mL, pH 7.4) as a redox cycle to realize signal amplification. The excitation light was turned on every 10 s in the photoelectric measurement process, and the voltage was constant at 0.0 V. The change in photocurrent was observed by adding different concentrations of mercury ions into the electrolyte.

Real Sample Processing
Urine samples [28]: 1 mL 50% hydrochloric acid and 0.8 mL 0.1 mol L −1 KBrO 3 / 0.084 mol L −1 KBr were added to 10 mL urine samples; after reaction for 15 min, the appropriate amount of hydroxylamine (120 g L −1 ) hydrochloride/sodium chloride (120 g L −1 ) solution was added into the above mixture until the yellow color disappeared; samples were diluted to 50 mL with deionized water and diluted 10 times with PBS buffer before use.
Artificial saliva [29]: 0.6 mg mL −1 disodium hydrogen phosphate, 0.6 mg mL −1 anhydrous calcium chloride, 0.4 mg mL −1 potassium chloride, 0.4 mg mL −1 sodium chloride, 4 mg mL −1 mucin and 4 mg mL −1 urea was dissolved in 1 mL of the deionized water. This was set to pH 7.2, stored in the refrigerator and diluted 10 times with PBS buffer before use.

Characterizations of Composite Material
Scanning electron microscopy (SEM), X-ray diffraction (XRD), and Fourier transform infrared spectroscopy (FT-IR) were used to characterize the prepared composites and to analyze the electrode preparation process.
SEM images reveal that the bare FTO electrode is a uniform spongy layer ( Figure 1A) [30]. After g-C 3 N 4 is loaded onto FTO, the surface of the g-C 3 N 4 /FTO electrode presents a typical plate-like structure ( Figure 1B), which results from aggregation of the g-C 3 N 4 sheets [31]. A large number of CdS QD nanoparticles are distributed on the surface of g-C 3 N 4 ( Figure 1C), which is consistent with the literature [32]. The large surface area of g-C 3 N 4 serves as an effective anchor for loading CdS NPs. When the CdS QD-g-C 3 N 4 /FTO electrode is covered with CuO nanoparticles ( Figure 1D), the CuO nanoparticles appear to be nearly spherical [33]. SEM ( Figure 1E,F) and elemental mapping by energy dispersive X-ray spectroscopy ( Figure 1I) reveal that C, N, Cd, S, Cu, and O exist in the g-C 3 N 4 -CdS-CuO electrode, indicating that the composite material is successfully loaded onto the FTO electrode.
The XRD analysis results of the crystal phase of the synthesized material are presented in Figure 1G. The XRD spectrum of g-C 3 N 4 exhibits two evident diffraction peaks at 24.3 • and 27.0 • . These peaks represent the (100) and (002) planes of the graphite material, respectively, corresponding to the in-plane structure of the triazine ring and layered superposition of the conjugated aromatic groups [34]. The diffraction peaks of CdS at 26.5 • , 37.6 • , and 43.7 • correspond to the (002), (102), and (103) crystal planes, respectively, which correlate with the wurtzite phase standard cards (JCPDS 80-0006) [35]. The diffraction peaks of the CuO nanoparticles at 35.6 • , 38.7 • , and 48.8 • , in accordance with their monoclinic and crystalline nature, correspond to the (110), (111), and (200) crystal planes, respectively [33].
In the FT-IR spectra ( Figure 1H) with the direct detection (Resolution: 4, Number of scans: 8, Detector: DTGS), g-C 3 N 4 (curve a) exhibits strong absorption peaks in the range of 1200-1600 cm −1 , which correspond to C-N heterocycles. The absorption band at 806 cm −1 corresponds to the typical breathing mode of the tri-s-triazine ring, and the intensity of the absorption peak near 2900-3400 cm −1 corresponds to the N-H stretching vibration [36]. The Cd-S band of CdS (curve b) demonstrates stretching and bending vibration absorption peaks at 631 cm −1 , whereas the peak at 1630 cm −1 corresponds to the O-H stretching vibrations [37]. For the CuO NPs (curve c), the peaks appearing at 701 cm −1 belong to the CuO NPs and confirm the formation of NPs [38]. Moreover, peaks corresponding to OH − and C-O stretching vibrations can be observed from 2800 to 3600 cm −1 [39]. These results confirm that the composite and modified electrodes were successfully synthesized. superposition of the conjugated aromatic groups [34]. The diffraction peaks of CdS at 26.5°, 37.6°, and 43.7° correspond to the (002), (102), and (103) crystal planes, respectively, which correlate with the wurtzite phase standard cards (JCPDS 80-0006) [35]. The diffraction peaks of the CuO nanoparticles at 35.6°, 38.7°, and 48.8°, in accordance with their monoclinic and crystalline nature, correspond to the (110), (111), and (200) crystal planes, respectively [33]. In the FT-IR spectra ( Figure 1H) with the direct detection (Resolution: 4, Number of scans: 8, Detector: DTGS), g-C3N4 (curve a) exhibits strong absorption peaks in the range of 1200-1600 cm −1 , which correspond to C-N heterocycles. The absorption band at 806 cm −1 corresponds to the typical breathing mode of the tri-s-triazine ring, and the intensity of the absorption peak near 2900-3400 cm −1 corresponds to the N-H stretching vibration [36]. The Cd-S band of CdS (curve b) demonstrates stretching and bending vibration absorption peaks at 631 cm −1 , whereas the peak at 1630 cm −1 corresponds to the O-H stretching vibrations [37]. For the CuO NPs (curve c), the peaks appearing at 701 cm −1 belong to the CuO NPs and confirm the formation of NPs [38]. Moreover, peaks corresponding to OH − and C-O stretching vibrations can be observed from 2800 to 3600 cm −1 [39]. These results confirm that the composite and modified electrodes were successfully synthesized.

Electrochemical Activity of g-C 3 N 4 -CdS-CuO Electrode
The electrochemical properties of the bare and material electrodes were studied, including the diffusion coefficient and heterogeneous electron-transfer rate constant. First, the electrochemical performances of the bare and modified electrodes were determined using cyclic voltammetry, in a 5 mM [Fe(CN) 6 ] 3−/4− solution containing 0.1 M KCl, by employing platinum as the counter electrode and saturated Ag/AgCl as the reference electrode (See supporting literature for details). As illustrated in Figure 2A,D, the current response increases gradually with an increase in the scanning rate and demonstrates a good linear relationship with the square root of the potential scanning rate, indicating that the redox reaction on the surface of the bare and modified electrodes is a diffusion control process. The electrically active area of the bare and g-C 3 N 4 -CdS-CuO electrodes was calculated according to the Randles-Sevcik equation [40]: where I p denotes the maximum current (A), D denotes the diffusion coefficient (cm 2 /s), C = [Fe (CN) 6 ] 3−/4− (mol/cm 3 ), N represents the number of transferred electrons, υ denotes the scanning rate (V/s), and A indicates the effective working area (cm 2 ) of the modified electrode. The effective working area (A) can be calculated by a linear fitting of I p to υ 1/2 . The active area of the bare electrode was 53.72 cm 2 ( Figure 2B) and that of the g-C 3 N 4 -CdS-CuO electrode was 72.40 cm 2 ( Figure 2E). Thus, the larger the specific surface area of the g-C 3 N 4 -CdS-CuO electrode, the more numerous the active sites are, and this is conducive to the amplification of electrical signals. In addition, the electron transfer rate constants of the bare and g-C 3 N 4 -CdS-CuO electrodes can be calculated according to Nicholson's equation [41]: ϕ = k 0 [πDnFV⁄(RT)] −1⁄2 ; where ϕ denotes the peak-to-peak separation (δ Ep), k 0 indicates the heterogeneous electron transfer rate constant, D represents the electroactive material diffusion coefficient, F denotes the Faraday constant, R represents the molar gas constant, and T denotes the temperature. The slope of the ϕ pair [πDnF/(RT)] −1/2 υ −1/2 curve corresponds to the heterogeneous electron transfer rate constant k 0 , which is 3.35 × 10 −4 cm s −1 ( Figure 2C) for the bare electrode and 2.72 × 10 −4 cm s −1 ( Figure 2F) for the g-C 3 N 4 -CdS-CuO electrode. This indicates that the electrochemical equilibrium time required by the g-C 3 N 4 -CdS-CuO-modified electrode was not significantly shortened. material diffusion coefficient, F denotes the Faraday constant, R represents the molar gas constant, and T denotes the temperature. The slope of the φ pair [πDnF/(RT)] −1/2 υ −1/2 curve corresponds to the heterogeneous electron transfer rate constant 0, which is 3.35 × 10 −4 cm s −1 ( Figure 2C) for the bare electrode and 2.72 × 10 −4 cm s −1 ( Figure 2F) for the g-C3N4-CdS-CuO electrode. This indicates that the electrochemical equilibrium time required by the g-C3N4-CdS-CuO-modified electrode was not significantly shortened.

Feasibility of the Designed Strategy
To evaluate the spPEC properties of the fabricated CuO-CdS-g-C 3 N 4 /FTO electrode, photocurrent responses of bare FTO, g-C 3 N 4 /FTO, CdS-g-C 3 N 4 /FTO, and CuO-CdS-g-C 3 N 4 /FTO were monitored ( Figure 3A). The photocurrent of CuO-CdS-g-C 3 N 4 /FTO (column c) is better than that of g-C 3 N 4 /FTO (column a) and CuO-CdS-g-C 3 N 4 /FTO (column b), which is further verified by electronic impedance spectroscopy (EIS), recorded by the EIS different changes of preparing the electrodes in ferricyanide/ferrocyanide mixed solution ([Fe(CN) 6 ] 3−/4− ) (See supporting literature for details). As illustrated in Figure 3B, the resistance of g-C 3 N 4 /FTO (curve b) is larger than that of FTO (curve a), which may be attributed to the poor conductivity of g-C 3 N 4 /FTO. The resistance of CdS-g-C 3 N 4 /FTO (curve c) is lower than that of FTO, indicating that CdS promotes g-C 3 N 4 electron transfer. Moreover, the resistance of CuO-CdS-g-C 3 N 4 /FTO (curve d) is less than that of CdS-g-C 3 N 4 /FTO, which indicates that CuO increases the transfer of electrode current and further increases the photocurrent. The feasibility analysis further verifies this principle.
To verify the amplification caused by the redox cycle strategy, the developed spPEC sensor was studied using the photocurrent output. For comparison, we also carried out the comparison experiment of redox single signal amplification and double signal amplification; that is, under the same conditions, different concentrations of Hg 2+ (0.5 nM, 10 nM) were added to GSH ( Figure 3C) and GSH, NADPH, and GR ( Figure 3D) solutions respectively. As presented in Figure 3C, GSH is oxidized by CuO to GSSG to generate a strong photocurrent (column b vs. a) and single signal amplification was achieved. When Hg 2+ was added from 0.5 nM to 10 nM, the photocurrent I increased from 0.97 µA to 2.27 µA (column c and d). These results demonstrate that the GSH could specifically recognize Hg 2+ , and the photocurrent was inversely proportional to the concentration of Hg 2+ . Evidently, when NADPH and GR are added ( Figure 3D), the photocurrent reaches a maximum (column b vs. a) and double signal amplification. In the presence of GR, GSSG is reduced to GSH using NADPH as a substrate, thereby resulting in a self-powered redox cycle system. When Hg 2+ was added from 0.5 nM to 10 nM, the photocurrent I increased from 2.49 µA to 3.59 µA (columns c and d), which was the same as the photocurrent trend in Figure 3C, indicating that both methods were beneficial to PEC detection, and the detection effect of double signal amplification was better. Therefore, the GSH-NADPH-GR redox cycle can be used as a signal amplification strategy to improve the detection sensitivity for Hg 2+ . b), which is further verified by electronic impedance spectroscopy (EIS), recorded by the EIS different changes of preparing the electrodes in ferricyanide/ferrocyanide mixed solution ([Fe(CN)6] 3−/4− ) (See supporting literature for details). As illustrated in Figure 3B, the resistance of g-C3N4/FTO (curve b) is larger than that of FTO (curve a), which may be attributed to the poor conductivity of g-C3N4/FTO. The resistance of CdS-g-C3N4/FTO (curve c) is lower than that of FTO, indicating that CdS promotes g-C3N4 electron transfer. Moreover, the resistance of CuO-CdS-g-C3N4/FTO (curve d) is less than that of CdS-g-C3N4/FTO, which indicates that CuO increases the transfer of electrode current and further increases the photocurrent. The feasibility analysis further verifies this principle. To verify the amplification caused by the redox cycle strategy, the developed spPEC sensor was studied using the photocurrent output. For comparison, we also carried out the comparison experiment of redox single signal amplification and double signal amplification; that is, under the same conditions, different concentrations of Hg 2+ (0.5 nM, 10 nM) were added to GSH ( Figure 3C) and GSH, NADPH, and GR ( Figure 3D) solutions respectively. As presented in Figure 3C, GSH is oxidized by CuO to GSSG to generate a strong photocurrent (column b vs. a) and single signal amplification was achieved. When Hg 2+ was added from 0.5 nM to 10 nM, the photocurrent △I increased from 0.97 μA to 2.27 μA (column c and d). These results demonstrate that the GSH could specifically recognize Ultraviolet-visible diffuse reflection spectroscopy (DRS) was used to study the electronic orientation of the composite materials. The band gap of the three semiconductors was calculated according to the equation (αhν) 2 = C (hν-Eg) (where α, hν, C, and Eg are the absorption coefficient, photoenergy, the constant, and band gap energy, respectively) [42] and extrapolated the intercept of the linear part of (αhν) 2 vs. hν plot to the x-axis shown in Figure S1A. The band gaps (E g ) of g-C 3 N 4 , CdS, and CuO are 2.84 eV, 2.10 eV, and 1.38 eV, respectively. The conduction bands (ECB) and valence bands (E VB ) were estimated using the following formula: E CB = X − E c − E g /2 (E c = 4.5 eV) and E VB = E CB + E g , where X and E g denote the geometric average value of the absolute electronegativity of each atom in the semiconductor and the band gap width of the semiconductor, respectively. The E CB values of g-C 3 N 4 , CdS, and CuO were 0.92 eV, −0.36 eV, and 0.62 eV, respectively, and the E VB values of g-C 3 N 4 , CdS, and CuO were 3.76 eV, 1.74 eV, and 2.0 eV, respectively. The band gaps of g-C 3 N 4 , CdS, and CuO reported in the literature are presented in Scheme 2A [43]. When g-C 3 N 4 , CdS, and CuO are in contact, the energy level differences between CdS and g-C 3 N 4 and between CdS and CuO (Fermi-Dirac statistics) cause electrons to flow from CdS (high energy level) to g-C 3 N 4 and CuO (low energy level). If the material is described in terms of a Fermi-Dirac distribution, this electron transfer is referred to as a Fermi level alignment. The redistribution of electrons between CdS and g-C 3 N 4 and between CdS and CuO is assumed to trigger the downward and upward movement of CdS, g-C 3 N 4 and CuO band edges, respectively. Therefore, we can infer that the electron orientation of the g-C 3 N 4 -CdS-CuO electrode demonstrates a stepped structure, as illustrated in Scheme 2B.
The edges of the conduction and valence bands of these three materials increased in the following order: g-C 3 N 4 < CdS < CuO; that is, the CdS layer was inserted between g-C 3 N 4 and CuO to increase the conduction band edge of CuO, providing a higher driving force for the injection of excited-state electrons from outside the CuO layer. When g-C 3 N 4 -CdS-CuO was photoexcited under irradiation, the stepped structure at the edge of the band proved not only beneficial for electron injection in g-C 3 N 4 -CdS-CuO, but also for hole recovery. Under irradiation, the g-C 3 N 4 -CdS-CuO material generated a photocurrent, and the GSH-NADPH-GR redox cycle acted as a self-powered system to further increase the photocurrent and achieve signal amplification ( Figure S1B).
CdS and g-C3N4 and between CdS and CuO is assumed to trigger the downwa upward movement of CdS, g-C3N4 and CuO band edges, respectively. Therefore, infer that the electron orientation of the g-C3N4-CdS-CuO electrode demonst stepped structure, as illustrated in Scheme 2B. The edges of the conduction and v bands of these three materials increased in the following order: g-C3N4 < CdS < Cu is, the CdS layer was inserted between g-C3N4 and CuO to increase the conductio edge of CuO, providing a higher driving force for the injection of excited-state el from outside the CuO layer. When g-C3N4-CdS-CuO was photoexcited under irrad the stepped structure at the edge of the band proved not only beneficial for electro tion in g-C3N4-CdS-CuO, but also for hole recovery. Under irradiation, the g-C3N CuO material generated a photocurrent, and the GSH-NADPH-GR redox cycle act self-powered system to further increase the photocurrent and achieve signal amplif ( Figure S1B).

Optimization of Experimental Conditions
To improve the sensitivity of the sensor, we optimized the experimental parameters by using the photocurrent of PEC. We primarily optimized the concentration of NADPH ( Figure S2A), and a maximum change was noted when the amount of NADPH was 0.40 µg/µL (250 µL). The amount of GR was also optimized, which is one of the most important factors in the signal detection process. As illustrated in Figure S2B, when different concentrations of GR are added to the bath solution, a maximum photocurrent signal can be obtained at 2.08 ng/µL (2 µL). Next, we determined the optimal GSH concentration ( Figure S2C). The maximum signal was obtained at a concentration of 15 mM, after which the signal slowly diminished. Thus, 15 mM was selected as the optimal concentration for GSH. Finally, we optimized the mass ratio of g-C 3 N 4 to CuO. As presented in Figure S2D, the photocurrent signal of the material reaches a maximum when the ratio of g-C 3 N 4 and CuO is 1:1. As shown in Figure S2E, when using a Na 2 CO 3 -NaHCO 3 (0.1 M, pH 7.4) and tris-HNO 3 (0.1 M, pH 7.4) buffer, the signal response of the photocurrent decreases. In contrast, no corresponding interference can be observed in PBS (0.1 M, pH 7.4). Therefore, we think that using PBS (0.1 M, pH 7.4) to prepare the bath solution will result in a better signal response.

Analytical Performance and Selectivity and Stability
To verify the analytical performance of the proposed spPEC sensing system, different concentrations of Hg 2+ were added and measured under the optimal conditions. As illustrated in Figure 4A, the photocurrent signal gradually weakens with increasing Hg 2+ concentration. Hg 2+ is linearly and logarithmically correlated in the range between 5 pM and 100 nM. The regression equation was Y = 0.90LogC Hg2+ − 2.40 (R 2 = 0.993), and the calculation formula for detection limit is S/N = 3 [44]. The detection limit was 0.84 pM ( Figure 4B). Compared with other Hg 2+ detection methods, this sensor demonstrated a picomolar detection sensitivity for Hg 2+ without requiring aptamer fixing and pretreatment processes (Table S1). Therefore, we can conclude that the detection efficiency of the sensor is significantly improved. trated in Figure 4A, the photocurrent signal gradually weakens with increasing Hg 2+ concentration. Hg 2+ is linearly and logarithmically correlated in the range between 5 pM and 100 nM. The regression equation was Y = 0.90LogCHg2+ − 2.40 (R 2 = 0.993), and the calculation formula for detection limit is S/N = 3 [44]. The detection limit was 0.84 pM ( Figure  4B). Compared with other Hg 2+ detection methods, this sensor demonstrated a picomolar detection sensitivity for Hg 2+ without requiring aptamer fixing and pretreatment processes (Table S1). Therefore, we can conclude that the detection efficiency of the sensor is significantly improved.  The experimental results indicated that the PEC response did not change significantly within 200 s ( Figure 4C), and the relative standard deviation (RSD) of the PEC response change was less than 0.77%. To assess the selectivity of the constructed spPEC sensor, we investigated its specificity for Hg 2+ (0.5 nM). Among several analogs, including Mg 2+ , Mn 2+ , Cd 2+ , Fe 3+ , Zn 2+ , K + , and Na + (50 nM), a much lower current value was recorded with Hg 2+ than with other interfering species ( Figure 4D). This result demonstrates the excellent selectivity of the spPEC method for Hg 2+ . The stability of the sensor was also evaluated for the detection of 0.5 nM Hg 2+ . Moreover, to evaluate the repeatability of the proposed sensor, six replicate measurements were performed with the sensor using 0.5 nM Hg 2+ ( Figure S3). The results revealed an RSD of 0.53%, indicating reasonable repeatability of the spPEC sensor for Hg 2+ detection.

Analysis of Real Samples
To investigate whether the proposed method could be used to detect Hg 2+ in complex samples, the experiment was conducted for two different substances (the urine of healthy people and artificial saliva). The standard addition method was used to evaluate the feasibility of the method for actual samples, as presented in Table 3. The concentration of Hg 2+ was in the range 0.1-10 nM, and the recovery rate was between 93.56% and 103.02%, indicating possible applicability of the spPEC sensor for detecting Hg 2+ in real-sample application. "-" Indicates that it is below detection limit.

Conclusions
A new spPEC sensor for Hg 2+ detection based on g-C 3 N 4 -CdS-CuO composites, as well as a redox-cycle signal-amplification strategy were developed. The novel g-C 3 N 4 -CdS-CuO co-sensitized modified electrode, which exhibits large surface area and improved photoelectric response due to a cascade structure, was successfully used to detect target ions in complex samples. In addition, the self-powered sensor demonstrated satisfactory sensitivity owing to double signal amplification (GSH and GSH-NADPH-GR) and a limit of detection of 0.84 pM for Hg 2+ . More importantly, in vitro application of the GSH-NADPH-GR redox cycle demonstrated its potential as a photoelectrochemical sensor and, therefore, presented a new method for constructing novel sensors. The sensor also demonstrated significant potential for detecting Hg 2+ in human urine and artificial saliva samples. More significantly, the developed approach can also be used to detect other heavy metal ions according to the HSAB theory and may serve as a new analytical tool for specific, sensitive, and reliable analysis of heavy metal ions in clinical toxicology, food, and the environment.  Table S1: Comparison of other analysis strategies for Hg 2+ detection. References [10,25,34,[45][46][47][48][49][50] are cited in the supplementary materials.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.