2, 5-dimethoxyterephthalaldehyde (DMTP) was purchased from Alfa Aesar Chemical Co., Ltd. (Shanghai, China). Graphene (GR), 1, 3, 5-tris(4-aminophenyl) benzene (TAPB), sodium phosphate dibasic dodecahydrate (Na₂HPO₄·12H₂O), formic acid and n-butanol were bought from Shanghai Macklin Biochemical Co., Ltd. (China). 1, 4-dioxane and anhydrous sodium dihydrogen phosphate (NaH₂PO₄) were purchased from Aladdin Industrial Co., Ltd. (Shanghai, China). Acetone and lactic acid were supplied by Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Methanol, acetic acid and tetrahydrofuran (THF) were provided from Tianjin Fuchen Chemical Reagent Co., Ltd. (China). Ethanol was obtained from Mairuida Technology Co., Ltd. (Beijing, China). Ultrapure water (18.25 MΩ·cm) was prepared by Jingjiang Hengxin Environmental Protection Equipment Co., Ltd. (Jiangsu, China). Phosphate buffers (PBS, 0.1 M) at different pH were achieved by mixing different proportions of Na₂HPO₄ and Na₂HPO₄·12H₂O. All chemicals were of analytical grade quality and were used without further purification.

Fourier-transform infrared spectroscopy (FTIR) measurements were recorded on a Fourier-transform infrared spectrometer (Nicolet iS10, United States). The X-ray diffraction (XRD) spectra were recorded on an X'Pert3 Powder multifunctional X-ray diffractometer (PANalytical, Holland) in the range of 2θ = 0.5°–80° with Cu-Kα radiation (λ = 1.5418 Å). Scanning electron microscope (SEM) images were obtained on a Zeiss Supra 55 field emission scanning electron microscope. A JEM-2100 transmission electron microscope (Japan) was used to obtain the transmission electron microscopy (TEM) images. High-resolution transmission electron microscopy (HRTEM) and energy dispersive X-ray (EDX) spectrum studies of COF_DPTB were carried out on a Tecnai-G2 F30 S-TWIN microscope (Philips, Netherlands). A V-Sorb 2800 (Gold APP Instruments Corporation China) analyzer was used to collect the Brunauer-Emmett-Teller (BET) surface area and pore volume of COF_DPTB. The all electrochemical experiments were performed on a CHI760 electrochemical workstation (Shanghai Chenhua Instrument, China) with a three-electrode system containing a calomel reference electrode in the saturated KCl solution at room temperature, a platinum wire counter electrode and a GR/COF_DPTB/GCE (3 mm diameter) as the working electrode.

COF_DPTB was prepared by the aldehyde-amine Schiff-base condensation reaction of TAPB and DMTP as our previous work (Yu et al., 2023). Specially, 85.0 mg DMTP and 105.0 mg TAPB were firstly dissolved in 45.0 mL mixed solution of 1, 4-dioxane, n-butanol and methanol (volume ratio of 4:4:1) for 30 min by ultrasound. Secondly, 0.5 mL of 3.0 M acetic acid was added by drops and the obtained solution was reacted at room temperature for 2 h. Then, another 4.5 mL of 3.0 M acetic acid was added drop by drop and the mixed solution was placed in an oven at 70°C for 24 h. Next, the solution was cooled naturally to room temperature and the solid product was collected by centrifugation and washed three times with acetone and THF, respectively. The yellow powder of COF_DPTB was obtained by vacuum drying at 50°C for 24 h. Before each modification, the GCE was ground on the polishing pad with 0.05 μm Al₂O₃ powder until the electrode surface was as smooth as a mirror. Then it was washed ultrasonically with ethanol and water for 3 min in turn and dried. 1.0 mg COF_DPTB and 1.0 mg GR were separately dispersed in 1.0 mL H₂O and treated by ultrasound for about 1 h. The COF_DPTB/GCE and GR/GCE was obtained by casting 10 μL of COF_DPTB suspension and 10 μL of GR suspension on the surface of the pretreated GCE and then dried under the infrared light, respectively. The GR/COF_DPTB/GCE was prepared layer by layer by casting 10 μL of GR suspension on the surface of COF_DPTB/GCE and then dried under the infrared light. Lastly, 5.0 μL of 0.5% Nafion was dropped on the modified electrode surfaces and dried for use.

DPASV was employed for the high sensitivity detection of Cd²⁺, Pb²⁺, and Cu²⁺ in 10 mL of PBS solution (0.1 M, pH 4.0) under nitrogen atmosphere. DPASV was performed at the potential range from −1.2–0 V (vs. SCE) with the pulse amplitude of 50 mV, the pulse width of 0.2 s, the potential increment of 5 mV, and the quiet time of 10 s. Before every DPASV measurement, nitrogen was firstly injected for 30 min, and then an electrochemical deposition step was carried out at −1.2 V for 260 s and left for 10 s. After each test, a 300 s cleaning step was followed under agitation conditions to remove the residual HMIs at 0.2 V. The same experimental condition was applied for the individual detection as well as the simultaneous one of HMIs.

Three kinds of Chinese baijiu were purchased from the local supermarket, which were used directly for sample testing without pretreatment. For the sample analysis, 10 µL of Baijiu was added to 10 mL of PBS solution (0.1 M, pH 4.0) and the recovery rate was studied by using the standard addition method. Each experiment was repeated three times under the same conditions to obtain the relative standard deviation (RSD).

The functional groups of DMTP (curve a), TAPB (curve b) and COF_DPTB (curve c) were investigated by FTIR in Figure 2A. In curve a, the characteristic peak at 1,681 cm⁻¹ belonged to the C=O stretching vibration of DMTP. As for curve b, the FTIR peaks appeared at 3,430, 3,347, and 3,206 cm⁻¹ were originated from the N-H stretching vibration of TAPB. Compared to the two monomers of DMTP and TAPB, the appearance of a new characteristic peak at 1,614 cm⁻¹ corresponded to the stretching vibration of C=N (Zhang et al., 2018a), suggesting that the aldehyde group of DMTP formed imine bonds with the amino group of TAPB. Meanwhile, the intensity of the characteristic peaks belonging to DMTP and TAPB decreased significantly, indicating that COF_DPTB was generated by the reaction of DMTP and TAPB. XRD was later applied for the phase and structure characterization of the synthesized COF_DPTB. As shown in Figure 2B, the characteristic diffraction peaks appeared at 2.81°, 4.81°, 5.48°, 7.32°, 9.73°, and 25.39° can be indexed to the plane (100), (110), (200), (210), (220), and (001) of the crystalline COF_DPTB, respectively, which was consistent with the literature (Zhang et al., 2018b).

Next, SEM was used to characterize the surface morphologies of the prepared sensor. In Figure 3A, the SEM images of COF_DPTB/GCE showed the spherical structure and rough surface, and that of GR/GCE exhibited a lamellar structure in Figure 3B. As shown in Figure 3C, the bottom dark gray rough surface indicated the presence of COF_DPTB and the layer structure of GR was clearly, which proved that GR/COF_DPTB/GCE was successfully obtained. The TEM image of COF_DPTB was shown in Figure 3D, which was consistent with its SEM image, proving that it was successful prepared.

Figure 4 further describes (A) the HRTEM image, (B–E) elemental mapping images, (F) EDX spectrum and (G) N₂ adsorption-desorption isotherm (inset: shows the pore size distribution) of COF_DPTB. As Figure 4A shown, the morphology of COF_DPTB in HRTEM image was consistent with that of SEM and TEM. In the elemental mapping images (Figure 4B–E), it was observed that the elements of C, O, and N were uniformly distributed. As can be seen from the EDX spectrum in Figure 4F, the content distribution of each element was as follows: C was the highest, N was less, and O was the least. Based on the above investigation, it can be concluded that COF_DPTB was successfully synthesized. Figure 4G characterized the porosity of COF_DPTB by the N₂ adsorption-desorption isotherm and the pore size distribution curves (inset), which presented the type IV isotherm with a distinct H3 hysteresis loop, indicating the well-defined mesoporous structure possessed by COF_DPTB. Its surface area, pore volume and average pore size were measured to be 810.760 m² g⁻¹, 0.518 cm³ g⁻¹, and 2.781 nm, respectively, which greatly improved the effective active sites on the electrode surface.

Electrochemical impedance spectroscopy (EIS) was applied to further characterize the behavior of the preparation and assembly process of the prepared sensor. The charge transfer resistances (Rct) of each modification step were estimated according to the semicircle diameters of Nyquist plots in 5.0 mM [Fe(CN)₆]³⁻/⁴⁻ probe containing 0.1 M KCl solution. Figure 5 displays the Nyquist plots of the bare GCE (curve a), COF_DPTB/GCE (curve b), and GR/COF_DPTB/GCE (curve c). As is shown, the Rct value of the GCE electrode (curve a) was the largest, and it gradually decreased with the deposition of COF_DPTB (curve b), indicating that the modifier would promote the surface charge transfer. For GR/COF_DPTB/GCE (curve c), a significantly reduced semicircle diameter was further observed because the electron transfer rate and mass exchange of the electroactive indicators on the electrode surface were effectively facilitated owing to the synergistically increased specific surface area and high electrical conductivity by COF_DPTB combined GR. The results revealed the successful preparation of GR/COF_DPTB/GCE.

The electroactive surface areas of GCE and GR/COF_DPTB/GCE were determined by cyclic voltammetry (CV) in 5.0 mM [Fe(CN)₆]³⁻/⁴⁻ solution containing 0.1 M KCl (Figure 6). Obviously, both the anodic and cathodic peak currents (I _p ) were linearly enhanced with the increase of scan rate, and showed a linear dependence on its square root (v ^1/2). The electrochemically active surface areas were calculated by using Randles-Sevcik formula (Dashtian et al., 2019): I p = 2.69 × 10 5 n 2 / 3 A D 1 / 2 C ν 1 / 2 where I _p , n, A, D, C and ν correspond to the peak current (A), the number of electrons transferred, the electrode’s electrical activity surface areas (cm²), the scan rate (V s^–1), the diffusion coefficient (cm² s^–1) and the bulk concentration of electrochemically active molecules in this solution (M), respectively. The electroactive areas of GCE and GR/COF_DPTB/GCE were calculated from the slope of plot of the I _p against ν ^1/2 and found to be 0.046 and 0.068 cm², respectively. It indicated that GR/COF_DPTB/GCE in [Fe(CN)₆]³⁻/⁴⁻ solution showed a fast rate of electron transfer from the modified electrode to the redox probe. The effective surface area of the GR/COF_DPTB/GCE was 1.48 times higher than that of the GCE, confirming that GR/COF_DPTB/GCE not only improved the conductivity but also enlarged the active surface area of the modified electrode. It proved that the constructed GR/COF_DPTB/GCE was a promising candidate for high-performance electrochemical sensor.

To evaluate the detection performance of the proposed electrochemical sensor, the DPASV responses of GCE (curve a), COF_DPTB/GCE (curve b), GR/GCE (curve c) and GR/COF_DPTB/GCE (curve d) for 1.0 μM each of Cd²⁺, Pb²⁺, and Cu²⁺ in the N₂-saturated PBS (0.1 M, pH 4.0) was studied and the results were recorded in Figure 7. As shown, the stripping peak currents of GCE to Cd²⁺, Pb²⁺, and Cu²⁺ were very low (0.090, 0.300, 0.080 μA, respectively). The COF_DPTB modified GCE was almost unresponsive to Cd²⁺ and had inconspicuous response to Pb²⁺ and Cu²⁺. The response of GR/GCE to Cd²⁺ and Cu²⁺ was weak (0.146 and 0.442 μA, respectively). It indicated that the GCE, COF_DPTB/GCE and GR/GCE did not have the ability to detect Cd²⁺, Pb²⁺, and Cu²⁺ at the same time. While for GR/COF_DPTB/GCE, three anodic stripping peaks of Cd²⁺, Pb²⁺, and Cu²⁺ presented respectively at −0.855 V, −0.590 V and −0.135 V with high response current and large peak-to-peak separation, indicating that there was no mutual interference between the three HMIs. On the one hand, GR can make up for the shortcomings of poor electrical conductivity of COF_DPTB. On the other hand, the layered graphene and the porous COF_DPTB synergistically increased the specific surface area. Moreover, the periodic porous network of COF_DPTB enabled the HMIs to be identified and captured selectively, because the N and O atoms on COF_DPTB had good selective coordination ability with Cd²⁺, Pb²⁺, and Cu²⁺, which provided a large number of effective binding sites. And the difference in coordination ability with Cd²⁺, Pb²⁺, and Cu²⁺ produced different anodic stripping peaks on the GR/COF_DPTB/GCE. The above results showed that GR/COF_DPTB/GCE had the electrochemical analysis ability for the simultaneous detection of Cd²⁺, Pb²⁺, and Cu²⁺. In addition, there was a small shoulder peak near the Pb²⁺ response peak at about −0.48 V, which was attributed to the possible formation of a Pb-Cu alloy (Wang Z. H. et al., 2020).

In order to achieve the best electrochemical performance of the developed sensor, DPASV method was used to optimize the relevant experimental parameters, including the dosage ratio of COF_DPTB to GR, the pH of PBS solution, the deposition potential and the deposition time.

Firstly, the dosage ratio of COF_DPTB to GR was optimized. Figure 8A exhibits the influence of the modified dosage ratios of COF_DPTB to GR at 1:0.6, 1:0.8, 1:1, 1:1.2, and 1:1.4 on the stripping peak current of the three HMIs. The results showed that the peak currents of the three ions were the highest when the dosage ratio was 1:1. When the ratio was lower than 1:1, the current response was small. It was possibly because of the small content of GR and the low conductivity. And when the ratio was higher than 1:1, the current decreased gradually. It may be because the modified material became thicker, which hindered the mass transfer process of the electrode surface. Therefore, 1:1 was selected as the optimal material modification dosage ratio for the following experiments.

Next, the pH influence of 0.1 M PBS buffer in the range of 2.0–6.0 was studied, as shown in Figure 8B. It can be seen that in this pH range, the peak currents of Pb²⁺ and Cu²⁺ increased first and then decreased, while that of Cd²⁺ increased with the increase of pH. At pH 4.0, the response currents of Pb²⁺ and Cu²⁺ were the highest and Cd²⁺ also had high response current. Under the comprehensive consideration, pH 4.0 was used for the follow-up studies.

The influence of the deposition potential from −0.8 to −1.6 V on the response currents of Cd²⁺, Pb²⁺, and Cu²⁺ was further studied and the results were shown in Figure 8C. From −0.8 to −1.2 V, the stripping signals of Cd²⁺, Pb²⁺, and Cu²⁺ increased gradually because the more negative the deposition potential was, the easier the three ions would be reduced. However, the response currents decreased when it was more negative than −1.2 V, which was due to the increased effect of hydrogen evolution hindering the deposition of the metal alloy being deposited onto the electrode surface (Yu et al., 2018). Thus, −1.2 V was set as the optimal deposition potential for the electrochemical analysis.

Figure 8D displays the DPASV response signals to Cd²⁺, Pb²⁺, and Cu²⁺ in the deposition time range of 160–360 s. As it shown, the current signals increased with the increasing deposition time when it increased from 160 to 260 s. But when it was increased up to 260 s, the DPASV response currents decreased gradually as the deposition time increased, which may be due to the competitive adsorption between Cd²⁺, Pb²⁺, and Cu²⁺, or to the saturated accumulation of the three HMIs mentioned above on the electrode surface that lead to the decreasing of the electron transfer rate at the electrode/solution interface (Isa et al., 2017; Roushani et al., 2017). So, 260 s was chosen as the optimal deposition time.

Under optimal conditions, DPASV responses of Cd²⁺, Pb²⁺, and Cu²⁺ detected by GR/COF_DPTB/GCE alone and simultaneously were investigated. The individual determination was carried out by increasing the concentration of one target ion and keeping the other two unchanged to study the mutual interference between the three ions. As shown in Figure 9A, the response signals of Cd²⁺ increased accordingly with increasing its amount in the present of 1.0 μM each of Pb²⁺ and Cu²⁺, and the linear regression equation of I (μA) = 0.467 [Cd²⁺ (μM)] + 0.538 (R ² = 0.998) was obtained in the range of 0.5–100 μM with the limit of detection (LOD) calculated to be 9.384 nM (S/N = 3). Similarly, as shown in Figure 9B, the analytical curve for Pb²⁺ detection demonstrated a good linearity in the range of 0.01–10 μM and the correlation equation of I (μA) = 1.877 [Pb²⁺ (μM)] − 0.244 (R ² = 0.996) was achieved with the LOD calculated to be 4.508 nM (S/N = 3). As for the separate detection of Cu²⁺ shown in Figure 9C, a good linearity in the range of 0.5–10 μM and the correlation equation of I (μA) = 1.171 [Cu²⁺ (μM)] − 0.906 (R ² = 0.991) was achieved with the LOD calculated to be 4.987 nM (S/N = 3). The above results verified that no mutual interference existed among the three HMIs.

In the practical applications, multiple HMIs usually coexist. Therefore, we also studied the linear relationship of the three HMIs detected at the same time, as shown in Figure 10. The results showed that the response current of each ion increased with the increase of its concentration. For Cd²⁺, the linearization equation was I (μA) = 0.350 [Cd²⁺ (μM)] + 0.054 (R ² = 0.997) and the corresponding linear range was 0.1–25 μM. For Pb²⁺, the linearization equation was I (μA) = 1.445 [Pb²⁺ (μM)] − 0.859 (R ² = 0.986) and the corresponding linear range was 0.1–11 μM. For Cu²⁺, the linearization equation was I (μA) = 1.232 [Cu²⁺ (μM)] − 0.159 (R ² = 0.991) and the corresponding linear range was 0.1–11 μM. The LODs of Cd²⁺, Pb²⁺, and Cu²⁺ were 0.011 μM, 8.747 nM and 6.373 nM, respectively (S/N = 3). Compared with the single ion detection, the linear range became narrower, which may be the result of competitive adsorption among the different HMIs (Han et al., 2020). Table 1 summarized the performance comparison of the developed sensor with some other reported electrochemical sensors for the simultaneous detection of Cd²⁺, Pb²⁺, and Cu²⁺. As it shown, the analytical performance of the sensor designed in this study was competitive owing to the widest linear range and the lowest LOD for each HMI. It can be attributed to the large effective surface area and plentiful binding sites provided by the GR/COF_DPTB film.

It is necessary to study the selectivity of the electrochemical platform because a variety of ions are typically present in Baijiu samples. To examine the selectivity of the proposed sensor, the interference measurements were performed by DPASV under the selected experiment conditions. The applied tolerance limit for the interfering species was the maximum concentration that gave a relative error of ±5% in the presence of 1.0 μM each of Cd²⁺, Pb²⁺, and Cu²⁺, and the results were listed in Table 2. It revealed that 5-fold of inorganic ions (including Ag⁺, Fe³⁺, Zn²⁺, Cr³⁺, Mn²⁺, K⁺, Na⁺, Ca²⁺, Mg²⁺), 1000-fold of small organic molecules (such as methanol, formic acid and lactic acid), and 10⁷-fold of ethanol had no effect on the detection of Cd²⁺, Pb²⁺, and Cu²⁺. It suggested the good specificity of the constructed sensor.

Next, the reproducibility and stability of the fabricated sensor were examined by measuring the DPASV response currents of 0.1 M PBS solution containing 1.0 μM each of Cd²⁺, Pb²⁺, and Cu²⁺ under the optimized working conditions. The repeatability was evaluated by comparing the stripping currents from ten successive measurements on a new prepared GR/COF_DPTB/GCE and the RSDs were found to be 2.01% for Cd²⁺, 2.56% for Pb²⁺, and 5.29% for Cu²⁺, respectively. The reproducibility was further studied with seven independent GR/COF_DPTB/GCE sensors, and the RSDs of 4.80% for Cd²⁺, 4.90% for Pb²⁺, and 5.60% for Cu²⁺, respectively were obtained. These results revealed that the developed sensor had good repeatability and reproducibility. When not in use, the prepared sensor was stored at room temperature, and it could respectively retain 91.20%, 96.08%, and 98.18% of Cd²⁺, Pb²⁺, and Cu²⁺ of the original current response after 1 week deposit, indicating good storage stability.

In order to evaluate the practicability of GR/COF_DPTB/GCE, three different brands of Baijiu purchased from local supermarkets were taken as the samples and the recovery tests were carried out by the standard addition method. The obtained results were listed in Table 3, which showed that the recovery ranges of Cd²⁺, Pb²⁺, and Cu²⁺ were 92.8%–94.2%, 107.8%–109.9% and 92.7%–98.1%, respectively, and RSDs were less than 1.86%, 2.42%, and 4.99%, respectively. The results indicated that the novel sensor had good accuracy and feasibility in the simultaneous assay of Cd²⁺, Pb²⁺, and Cu²⁺ in real samples.