On-chip trace detection of Cd²⁺ and Pb²⁺ of deep seawater using CMOS-integrated low-noise transimpedance amplifiers

Abstract

Electrochemical techniques are commonly employed for heavy metal detection. However, due to parasitic capacitance and noise issues arising from their structural design, conventional workstations face limitations in detection performance and system scalability when used for trace analysis in complex environments. To address these limitations, here we developed a custom-designed, low-noise, multi-channel complementary metal-oxide-semiconductor transimpedance amplifier integrated circuit with vertically integrated on-chip electrodes. This system achieves a low noise level of 273.9 fA_RMS and reduces the electrochemical reaction area to just 1 mm², enabling sensitive and specific detection of Cd²⁺ and Pb²⁺ in the wide range of 0.05–500 μg/L. We validated the system’s performance by detecting Cd²⁺ and Pb²⁺ in real seawater samples collected from a depth of 8,448 meters in the Mariana Trench, achieving concentrations of 0.859 μg/L for Cd²⁺ and 0.921 μg/L for Pb²⁺. Compared to inductively coupled plasma-mass spectrometry, our system demonstrated excellent agreement for Cd²⁺ (0.10% deviation) and reasonable consistency for Pb²⁺ (28.0% deviation), reflecting its selectivity for free ions. Our work provides a robust, portable, and miniaturized solution for off-line trace Cd²⁺ and Pb²⁺ detection with seawater background for advanced in situ oceanic monitoring technologies.

To overcome noise limits in heavy metal detection, Yiming Yu and colleagues developed a miniaturized CMOS transimpedance amplifier with integrated electrodes. The device successfully detected trace Cd²⁺ and Pb²⁺ in seawater samples collected from the Mariana Trench at a depth of 8,448 m.

Introduction

Heavy metals, such as cadmium (Cd) and lead (Pb), are persistent environmental pollutants with notable adverse effects on human health and ecosystems¹. These metals interact strongly with biological molecules, leading to high toxicity and irreversible damage^2–4. In remote and poorly characterized environments like the deep sea, heavy metal concentrations are exceedingly low, making their detection highly sensitive to even minor disturbances. While techniques such as inductively coupled plasma-mass spectrometry (ICP-MS) offer unparalleled accuracy for trace-level analysis, their high cost, complex sample preparation, and stringent operational requirements limit their applicability in situ, particularly in remote or extreme conditions^5,6. A compact, scalable solution tailored for such environments is therefore highly desirable.

Electrochemical methods provide a cost-effective alternative for heavy metal detection, leveraging redox reactions to quantify metal ion concentrations^7–9. These methods are particularly attractive due to their simplicity, portability, and ability to operate in real-time, making them suitable for field applications such as environmental monitoring and industrial quality control^10,11. However, traditional electrochemical workstations face notable challenges in achieving the sensitivity and detection limits required for ultra-trace analysis. The core limitation lies in the noise performance of transimpedance amplifiers (TIAs), which directly determines the system’s ability to resolve weak current signals in the picoampere (pA) to nanoampere (nA) range. In low-concentration regimes, the current generated by redox reactions can be as low as a few pA, making it critically important to minimize noise contributions from the TIA. Furthermore, the dynamic range of the TIA must be carefully optimized to accommodate both high and low current levels without introducing distortion or saturation, further complicating the design process.

Additionally, parasitic capacitance introduced by external wiring and discrete components further degrades signal fidelity, hindering accurate measurements in low-concentration regimes. Parasitic capacitance arises from the physical separation between the working electrode, reference electrode, and counter electrode, as well as from the connections between these components and the TIA. This capacitance creates a low-impedance path for high-frequency noise to couple into the system, reducing the effective bandwidth and increasing the likelihood of signal distortion. Moreover, the use of discrete components in traditional systems introduces additional noise and variability, as each component contributes its own inherent noise characteristics and tolerances. These issues are exacerbated in portable or miniaturized systems, where space constraints limit the ability to mitigate parasitic effects through shielding or filtering. Addressing these limitations requires innovative approaches, such as integrating electrodes and amplifiers onto a single chip to minimize parasitic effects and enhance signal integrity.

To address these limitations, we proposed a fully integrated electrochemical detection system based on complementary metal-oxide-semiconductor (CMOS) technology. By vertically integrating on-chip electrodes with low-noise TIAs, parasitic effects were minimized, and good noise cancellation was achieved. The performance of this design was compared with that of the commercial benchtop system CHI760E. For the purpose of benchmarking fundamental detection limits under controlled lab conditions, the CHI760E is a representative and pragmatic choice. Furthermore, we demonstrated the system’s capability to detect Cd²⁺ and Pb²⁺ in real seawater samples collected from a depth of 8448 m in the Mariana Trench. Our work provides a robust solution for in situ monitoring of other heavy metals (Hg²⁺, Cu²⁺, Zn²⁺, et al.) in demanding environments, while also highlighting the potential of our integrated multi-channel chip for the long-term monitoring of trace heavy metals in challenging marine settings.

Results

Design considerations

Figure 1a illustrates the concept of a miniaturized low-noise electrochemical chip for trace Cd²⁺ and Pb²⁺ detection in extreme deep sea, leveraging CMOS technology and vertical integration. The system employs a three-electrode configuration (Titanium working electrode (WE), reference electrode (RE), and counter electrode (CE)) immersed in a solution containing lead and cadmium ions in Fig. 1b. A potential difference is applied between the WE and RE, and the generated current between the WE and CE is detected. Lead and cadmium ions are reduced into amalgams during deposition and re-released during dissolution, enabling quantitative analysis via differential pulse stripping voltammetry (DPSV). Figure 1c shows the potentiostat principle and the circuit system framework for lead and cadmium detection. The potentiostat, which connects the WE, RE, and CE, is the core component of the three-electrode system, typically composed of two main operational amplifiers and other components. The amplifier at the input allows a specific external signal to be directly transmitted to the RE, forming a fixed potential difference between the RE and WE. The potential of the WE to ground is constant, so controlling the signal applied to the RE enables precise control of the signal applied to the WE without being affected by current-induced voltage drops. This independent performance markedly enhances the precision of experiments, particularly when studying the kinetics and mechanisms of electrochemical reactions, providing more reliable data. Low-noise electrochemical chip encapsulated in epoxy resin on a customized base (Fig.1d). The base is installed on a customized PCB and assembled into a lead and cadmium detection system. The test PCB adopts a main-sub board structure. The sub-board is bonded with the TIA, while the main board consists of the power supply module, controlling module and low-noise interfaces.

Fig. 1. System overview and detection principle.

a Schematic illustration of deep-sea operations using a manned submersible for lead and cadmium detection at depths exceeding 8000 m in the Mariana Trench. b Detailed view of the electrochemical cell and detection principle. Multiple Ti electrodes are integrated on a multi-channel transimpedance amplifier (TIA) chip using complementary metal-oxide-semiconductor (CMOS) technology as the working electrode (WE). Lead and cadmium ions form amalgams during deposition and are released during dissolution, enabling reversible detection. c Framework of the lead and cadmium detection system. The electrochemical cell facilitates redox reactions of lead and cadmium ions. The three-electrode system consists of the WE, reference electrode (RE), and counter electrode (CE), integrated with the TIA. d Photograph of the TIA and PCB system. TIA is encapsulated in epoxy resin and bonded to a customized base. The PCB is reserved with interfaces for connecting to TIA. The base is installed on a customized PCB and assembled into the lead and cadmium detection system. The test PCB adopts a main-sub board structure. The sub-board is bonded with the TIA, while the main board consists of the power supply module, controlling module and low-noise interfaces (all image copyrights belong to Wei Cai.)

The core of the system is the TIA (Supplementary Methods. 1), which converts the weak current signals from the WE into measurable voltages. To achieve high sensitivity, we adopted a fully differential architecture with a resistive continuous-time feedback design. The feedback resistor (R_F = 257 MΩ) was optimized to balance gain, bandwidth, and noise performance. At the transistor level, PMOS inputs were selected for their lower flicker noise¹², with critical dimensions of W/L = 784 μm/240 nm. Source degeneration resistors further reduced effective transconductance, enhancing noise optimization¹³ (Supplementary Table 1).

The TIA layout fabricated using a 0.18 μm CMOS process, measures 1.255  × 0.8 mm and integrates four channels with on-chip electrodes (Fig. 2a, b). Aluminum-coated electrode surfaces were used for durability, and mercury film modification enhanced sensitivity for lead and cadmium detection. Compared to the state-of-the-art commercial amplifiers like the AD8139, our TIA achieves a noise level of 273.9 fA_RMS at 1 kHz low-pass filtering, which is 50× lower than that of AD8139 (Fig. 2e). This ultra-low noise performance, which is comparable to commercial instruments and previous integration solutions (Table 1), ensures high-fidelity detection of ultra-weak electrochemical signals. Electrical characterization confirmed the system’s accuracy and linearity, with an I-V slope of 1.9868 pA/mV (R² = 0.9999) (Fig. 2c, d).

Fig. 2. TIA photograph, layout and performance comparison.

a Photograph of the TIA, showing aluminum-coated electrode surfaces. b Layout implementation of the TIA, measuring 1.255  × 0.8 mm, with four individual channels and integrated on-chip electrodes. c Current-time curve and d Current-voltage curve of the TIA, converting pA-level currents to voltage using a 500 MΩ input resistor. e Power spectral density of our TIA and the state-of-the-art (AD8139). Both are connected to an identical feedback resistance (R_F = 257 MΩ), shielded in a Faraday cage.

Table 1.

Comparison of TIA performance between this work and other works

Performance	AD8139	Reference31	Reference32	Reference33	This work
Technology	XFCB	Separate components	0.35 µm	0.18 µm	0.18 µm
Power rail (V)	±5	/	±1.5	1.8	1.8
Gain (MΩ)	/	1000	498	7.5 / 45	534
Equivalent input current noise (pARMS)	14.55 (<1 kHz)	1.10 (<5 kHz)	3.50 (<10 kHz)	0.10 (<1 kHz)	0.2739 (<1 kHz)
Feedback capacitor (fF)	/	100 − 500	/	50 − 1000	7.14
Power consumption (mW)	200	/	0.2	5.1	1.5

On-chip electrochemical system outperforms standard potentiostats

To validate the performance of our on-chip electrochemical system, we compared it with a commercial potentiostat (CHI760E) under identical cyclic voltammetry (CV) conditions. Figure 3a, b shows the recorded voltammograms for both systems using 0.02 μM K₃[Fe(CN)₆]. Our system exhibited a distinct reduction peak near +0.15 V, indicative of excellent sensitivity, while the CHI760E failed to resolve any peaks across all scan rates.

Fig. 3. Performance validation using cyclic voltammetry.

a CV measurements of the state-of-the-art commercial instrument CHI760E in 0.02 μM K₃[Fe(CN)₆] experiments at varying scan rates. Parameters: initial potential = +0.5 V, high potential = +0.5 V, low potential = −0.1 V, sampling interval =  1 mV. b CV measurements of our system under identical conditions. c Linear relationship between peak current and scan rate for our system, demonstrating excellent linearity (the meanings represented by different colors are the same as (a) and (b)). d Detection limits for K₃[Fe(CN)₆] at a scan rate of 10 mV/s.

Figure 3c highlights the linear relationship between the square root of the scan rate and the peak current in our system, confirming its superior detection capability. The detection limit for K₃[Fe(CN)₆] was 1 nM, one order of magnitude lower than the CHI760E (0.02 μM) (Fig. 3d). This demonstrates the ability of our system to capture redox characteristics at ultra-low concentrations.

For the lead and cadmium detection, the electrode surface was modified with a mercury film to enhance sensitivity (Fig. 4(a–e)). Using DPSV, we tested artificial seawater samples containing Cd²⁺ and Pb²⁺ (Supplementary Fig. 4). Figure 4f, g show excellent linearity (R² > 0.999) across two concentration ranges: 500 μg/L to 5 μg/L and 5 μg/L to 0.05 μg/L. These results confirm the robustness of our system for trace-level detection. Meanwhile, we performed 20 days of independent detections on samples containing 5 μg/L of both Pb²⁺ and Cd²⁺ (Supplementary Fig. 5). The results showed good consistency in the peak current responses, with standard deviations of less than 0.5% for both metals.

Fig. 4. Electrode preparation and lead and cadmium detection in artificial seawater.

a Photograph of the TIA prepared for electrochemical experiments. TIA is encapsulated with epoxy resin, exposing the aluminum layer on the electrode surface. b TIA has damaged the aluminum layer, exposing the titanium layer and modifying it with a mercury film. The electrodes turned silver-white. c, d, e Scanning electron microscope image of exposed Ti electrodes electroplated with a mercury film. f, g Differential pulse stripping voltammetry (DPSV) detection of Cd²⁺ and Pb²⁺ at different concentration gradients in artificial seawater. Parameters: initial potential = −0.9 V, final potential = −0.3 V, increment = 4 mV, amplitude = 0.05 V, pulse width = 0.05 s, sample width = 0.02 s, pulse period = 0.1 s, deposition potential = −1.4 V, deposition time = 400 s (high concentrations) or 600 s (low concentrations), quiet time = 10 s. Calibration curves for Cd²⁺ and Pb²⁺. In the high concentration range (5–500 μg/L), R² values are 0.999 for Cd²⁺ and 0.997 for Pb²⁺. In the low concentration range (0.05–5 μg/L), R² values are 0.999 for both Cd²⁺ and Pb²⁺.

In order to adapt to the low-temperature environment of the deep seawater, we validated our system under low-temperature conditions in Supplementary Fig. 7. The TIA system demonstrates reliable reproducibility and robust performance at temperatures as low as 3 °C, with peak current variations remaining within 13% across all replicate measurements. This confirms its suitability for accurate and continuous monitoring of heavy metals in cold seawater environments.

Finally, we validated the system’s performance using real seawater samples from the Mariana Trench at a depth of 8448 m. The pH of these samples was subsequently measured in the shipboard laboratory, yielding an approximate value of 7.6. Our system resolved clear stripping peaks for Cd²⁺ and Pb²⁺, achieving concentrations of 0.859 μg/L and 0.921 μg/L, respectively, with standard deviations of 0.10% and 28.0% (Fig. 5). While cadmium detection closely matched ICP-MS results (0.858 μg/L), lead values were consistently lower due to the selective nature of our electrochemical system^14–16.

Fig. 5. Offline demonstration of the Mariana Trench seawater samples.

a Schematic diagram of sample collection by the manned submersible Fendouzhe. (All image copyrights belong to Wei Cai). b Seawater samples collected at 8448 m in the Mariana Trench. c, d DPSV detection of real seawater samples collected from a depth of 8448 m in the Mariana Trench using our TIA system. Results of five consecutive measurements of Cd²⁺ and Pb²⁺ concentrations. The dashed lines indicate, inductively, coupled plasma-mass spectrometry (ICP-MS) results of 0.858 μg/L for Cd²⁺ (orange) and 1.28 μg/L for Pb²⁺ (green).

Discussion

The development of a miniaturized, low-noise electrochemical detection system represents a useful advancement in the field of trace Cd²⁺ and Pb²⁺ analysis, particularly for applications in extreme environments such as the deep sea. By integrating CMOS technology with on-chip titanium electrodes, our system achieves unparalleled sensitivity and noise performance, surpassing traditional benchtop electrochemical workstations. Our system addresses the detection limits through several key innovations: (1) the use of source degeneration techniques and non-unity gain buffers in the TIA design, which reduces noise levels to 273.9 fA_RMS; and (2) the vertical integration of on-chip electrodes, which minimizes parasitic capacitance and confines the ECR area to just 1 mm². These advancements enable a detection limit of 0.05 μg/L, two orders of magnitude lower than conventional systems (~5 μg/L) under the same conditions.

To validate the robustness of our system, we conducted experiments using real seawater samples collected from a depth of 8448 meters in the Mariana Trench. The system successfully detected Cd²⁺ and Pb²⁺ concentrations of 0.859 μg/L and 0.921 μg/L, respectively, without the need for added electrolytes. Compared to ICP-MS, our results showed excellent agreement for Cd²⁺ (0.1% deviation) and reasonable consistency for Pb²⁺ (28.0% deviation). The good agreement observed for Cd²⁺ can be attributed to its speciation in deep-sea environments. Under high-pressure and high-salinity conditions, cadmium predominantly exists as free Cd²⁺ ions or weakly coordinated chloro-complexes (e.g., CdCl⁺), which remain electrochemically labile and can be efficiently deposited onto the electrode surface¹⁷. As a result, the electrochemical measurements show excellent consistency with ICP-MS, with a deviation of only 0.10%. In contrast, the consistency for Pb²⁺ is comparatively lower. ICP-MS determines the total Pb concentration, including all chemical forms, whereas the electrochemical method primarily responds to free Pb²⁺ ions. Lead is known to form relatively stable chloro-complexes (e.g., PbCl₃⁻ and PbCl₄²⁻) in seawater and exhibits a strong affinity for natural organic ligands and suspended particulate matter¹⁸, which materially reduces its electrochemical lability. Consequently, the different distributions of Pb species in seawater are considered the main reason for the larger discrepancy between the electrochemical results and ICP-MS measurements. This distinction highlights the complementary role of our system alongside analytical techniques like ICP-MS, offering a portable and cost-effective alternative for in situ monitoring.

The successful detection of our system for samples from one of the most extreme environments on Earth - the Mariana Trench - demonstrates its potential for broader applications in environmental monitoring¹⁹, resource exploration²⁰, and ecological research^21,22. For instance, the ability to detect trace heavy metals in situ could provide critical insights into deep-sea ecosystems and anthropogenic pollution pathways^23,24, to name a few. Additionally, the system’s miniaturization and low-power requirements make it suitable for integration into autonomous underwater vehicles or remote sensing platforms²⁵.

Future work will focus on further optimizing the system’s sensitivity and expanding its applicability to other analytes, such as other heavy metals, organic pollutants or biomarkers. Although matrix components such as high salinity and dissolved organic matter may influence the electrochemical response, further optimization of electrode modification and sample pretreatment is expected to improve interference tolerance and expand the detectable metal speciation. Meanwhile, it is invaluable to explore the in-situ application capability of the system and its long-term stability in complex deep-sea environments. The TIA system successfully passed pressure tests equivalent to a depth of 5200 m in the ocean. We therefore anticipate that it can be extended to even deeper environments, such as the Mariana Trench (~8448 m), for trace heavy metal analysis. In addition, the exploration of advanced surface modification techniques is underway to enhance electrode selectivity and stability. Furthermore, machine learning algorithms are being developed for real-time data analysis and interpretation. These efforts will further solidify the role of integrated electrochemical systems in advancing scientific discovery and environmental stewardship.

Conclusions

We have developed a CMOS amplifier with on-chip electrodes based on vertical integration technology to push the detection limits of trace Cd²⁺ and Pb²⁺ in deep-sea electrochemical detection. Leveraging the miniaturization advantages of modern CMOS processes, the ECR system was reduced to an area of approximately 1 mm², while the distance between the ECR system and the TIA was shortened to a few hundred micrometers. This design noticeably reduces the noise level for multi-channel electrochemical experiments to 273.9 fA_RMS. In terms of low-concentration signal measurements, our chip extends the detection limit by at least two orders of magnitude compared to traditional electrochemical workstations, achieving a sensitivity of 0.05 μg/L versus the typical 5 μg/L. Furthermore, we validated the ability of the TIA to detect trace Cd²⁺ and Pb²⁺ in real seawater samples collected from a depth of 8448 m in the Mariana Trench, achieving accurate detection without the need for added electrolytes. These results demonstrate the potential of our system for applications in other fields requiring ultra-low signal long-term detection.

Methods

Per-unit cost

Based on a wafer cost of $2500, a compact die area of 1 mm², and an average yield of 90%, the fabrication cost per CMOS chip is estimated to be approximately $0.88. When low-cost packaging, basic testing, minimal peripheral components, system assembly, and amortized overhead under scalable production conditions are taken into account, the total cost of the single-chip TIA system is estimated to be on the order of $7.08 (Table 2).

Table 2.

Cost breakdown of the single-chip TIA system (per chip)

Item	Parameter	Calculation	Cost
Wafer fabrication	0.18 µm CMOS process	$2500	-
Area	1 mm2	3140 dies/8-inch wafer	-
Yield rate	90%	3140×0.9 = 2826 dies	-
Chip fabrication	-	2500/2826	$0.88
Packaging and testing	Low-cost packaging	-	$2.0
PCB and passive components	Minimal peripherals	Limited discrete components	$1.5
Electrode interface and connectors	-	Interface and connectors	$1.2
System assembly and losses	-	Assembly, yield loss and labor	$1.5
Total cost per chip	-	Sum	$7.08

Amplifier chip fabrication and packaging

The custom-designed integrated circuit was fabricated using a 0.18 μm CMOS process. The chip was directly connected to a custom printed circuit board via epoxy resin, serving as a test daughterboard. The daughterboard was interfaced with the main PCB through pin headers. The main PCB included the front-end amplifier, analog-to-digital converter, and interface circuitry required for the electrochemical detection system. Gold wire bonding under a microscope was used to connect the chip to the PCB, and epoxy resin encapsulation protected the chip and gold wires from corrosion by the solution.

A custom-designed DC power supply was implemented to meet the system’s requirements, primarily using low-dropout regulators and DC-DC converters to generate ±1.8 V, ±0.9 V, and ±5 V. For example, the TIA was powered by -0.9/ + 0.9 V during noise testing, while also powered by -0.9/ + 0.9 V during electrochemical experiments. Data acquisition was supported by the NI USB-6003 module, which includes four analog input channels and two analog output channels, enabling simultaneous detection across the four TIA channels. Data were transferred to a PC via a USB 2.0 interface. A user-friendly graphical interface was developed in MATLAB²⁶, offering scalability and cross-platform operation.

Electrode preparation

Four on-chip pads were directly integrated as electrodes using a standard 0.18-µm CMOS process. The top aluminum layer was selectively removed by post-CMOS etching, exposing the underlying titanium layer, which was used as the working electrode.

The electrode surface of the fabricated CMOS chip consists of titanium and aluminum. To expose the titanium working electrode, the aluminum layer was etched using Type A etching solution (Sigma-Aldrich, China) at room temperature (~25 °C) for 30 min. The etching process was monitored under a magnifier and stopped when the electrode surface transitioned from silver to black²⁷. After etching, the electrodes were rinsed three times with deionized water to ensure no residual etchant remained²⁸.

For CV experiments, no mercury plating was required. However, for DPSV, the electrode surface was modified with a mercury film. A 0.02 M mercury nitrate solution (Beifangweiye, China) was used for plating. The working electrode was the on-chip titanium electrode, the reference electrode was Ag/AgCl (Gaoss Union, China), and the counter electrode was Pt (Gaoss Union, China). A constant negative potential of −0.3 V was applied for 15 min to form a dense mercury film. Optical microscopy confirmed the silver-gray appearance of the electrode surface, and scanning electron microscopy (Nova NanoSem450, USA) verified the formation of a uniform mercury layer.

Noise measurement of the CMOS TIA Chip

Noise measurements were conducted under zero-current input conditions. The system was shielded from external interference using a Faraday cage. The ADC recorded the TIA’s voltage output signal over a 10-s interval. The root-mean-square noise of the current signal was calculated, and the power spectral density was derived from the recorded data. (Supplementary Methods 7)

CV measurements

The CV experiments used a mixture of potassium ferricyanide (Macklin, China) and 1 M KCl (Macklin, China). The working electrode was a 0.5 mm-diameter glassy carbon electrode (Gaoss Union, China), the reference electrode was Ag/AgCl (Gaoss Union, China), and the counter electrode was Pt (Gaoss Union, China). A triangular wave input signal was generated using MATLAB and supplied to the PCB. The starting potential was set to −0.1 V, and the ending potential was +0.5 V, with the slope representing the scan rate. The redox potential of potassium ferricyanide lies between −0.1 V and +0.5 V, making this range suitable for CV analysis.

DPSV measurements

The DPSV experiments were conducted using artificial seawater prepared with a 3.5% NaCl solution (Macklin, China). Stock solutions of 100  μmol/mL cadmium nitrate and 1000 μmol/mL lead nitrate (Boer, China) were used to prepare Cd²⁺ and Pb²⁺ solutions. The working electrode was an on-chip Ti electrode (75 × 75 μm), with Ag/AgCl as the reference electrode and Pt as the counter electrode. The input signal, a staircase-shaped square wave, was generated using MATLAB and supplied to the PCB.

The experiment consisted of three phases, including deposition, resting, and stripping. During the deposition phase, a potential of −1.4 V was applied for 400 s (high concentrations) or 600 s (low concentrations). In the resting phase, a potential of −0.9 V was applied for 10 s. During the stripping phase, a staircase rising square wave was applied from −0.9 V to −0.3 V. The data acquisition system collected current values during the high and low levels of the stripping phase, generating two I–V curves. The voltammetry curve was obtained by calculating the difference between these two current values. Finally, the raw data undergoes denoising, filtering, and limit calibration. (Supplementary Fig. 6)

To avoid contamination, measurements were performed in ascending order of concentration. Peak heights were quantified relative to the baseline, which was obtained from DPSV measurements of 3.5% NaCl solution under identical conditions. Each curve was measured repeatedly until two consecutive results were consistent, ensuring accuracy.

Seawater sample collection and analysis

The seawater sample was collected from a depth of 8448 m at coordinates 142.15717^∘E, 11.5698^∘N in the Mariana Trench during the 62^nd dive of the manned submersible Fendouzhe, which is employed for deep-sea exploration. The sample was stored at 3 °C. During analysis, five consecutive measurements were performed, and the average current was calculated.

Inductively coupled plasma-mass spectrometry (Agilent 7700X ICP-MS, USA) was used to validate the results. The seawater sample was diluted twofold, and three measurements were taken to calculate the original concentration.

Calibration and standardization

A standardized analytical workflow was established to ensure measurement accuracy, including on-chip electrode cleaning and activation, Hg-film modification, and calibration based on Pb²⁺ and Cd²⁺ concentration gradients. To maintain consistent performance, the system was recalibrated every three weeks using DPSV measurements over two concentration ranges, conducted from low to high concentrations to prevent contamination, followed by the construction of calibration curves within the corresponding linear response ranges.

Author contributions

Y.M.Y., W.F., C.Y.M., and T.D. designed the TIA and device. Y.M.Y., X.Y., and Y.F.W. prepared the TIA and PCB. Y.L., C.Y.M., Y.M.Y., and W.F. completed the architecture of the software. Y.M.Y., C.H.C., X.Z., Y.F.W., W.F., and T.D. prepared the figures. X.Y., T.Y.Z., S.Q.D., D.L., and S.Q.L. designed the experiment. Y.M.Y., W.F., Y.L., Y.G. and W.C. drafted the manuscript. Y.L., Y.G., and W.C. supervised the study. All authors read, revised, and approved the final manuscript.

Peer review

Peer review information

Communications Engineering thanks Babankumar S Bansod, Shiwei Wang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: [Philip Coatsworth]. A peer review file is available.

Data availability

The datasets generated and/or analyzed during the current study are available in the Zenodo repository under accession code²⁹ (10.5281/zenodo.19060448). Other specific datasets generated and analyzed during this study may be provided upon reasonable request by the corresponding author.

Code availability

The custom code supporting the findings of this study is openly available in the Zenodo repository under the DOI 10.5281/zenodo. 19060708 (10.5281/zenodo.19060708). The version of the code described in this paper is the version deposited in Zenodo³⁰. All analyses were performed using MATLAB (R2021b). Any specific parameters used for data processing are detailed in the “Methods” section of the paper. Other specific codes generated and analyzed during this study may be provided upon reasonable request by the corresponding author.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s44172-026-00671-y.