Electrochemical Deposition of Lead for Water Quality Sensing

To detect lead presence in drinking water, a 300/1000 Å Ti/Pt four-electrode sensor was designed and fabricated as shown in Fig. 1A. The four-electrode system consists of an anode (+), anode's adjacent floating electrode (F2), cathode's adjacent floating electrode (F3), and a cathode (−). Here, F2 and F3 serve as a reference point to determine if conductive materials are present. The gap between the two adjacent electrodes (e.g., anode and F2) was designed to be 3, 5, 8, 10 and 12 μm. Photolithography, electron beam evaporation, and lift-off were employed to fabricate the electrodes on borofloat wafers (Precision Glass & Optics) following published procedures. ¹² The borofloat wafers were diced into 2.0 mm × 3.6 mm pieces. Each wafer produced approximately 500 sensors. Fabricated sensors were epoxied (Loctite) to custom printed circuit boards (PCBs) and were electrically connected by Gold ball wirebonding (Model 4124, Kulicke & Soffa) and protected by epoxy encapsulation (Loctite). Wires connecting the sensors to the LabVIEW test setup were soldered to the other end of the PCBs.

Zoom In Zoom Out Reset image size

Lead (II) chloride (PbCl₂) and sodium chloride (NaCl, >98%) were purchased from Sigma-Aldrich (St. Louis, MO, USA). All experiments were conducted in 0.01 M NaCl with a conductivity of 1000 μS cm⁻¹, the upper limit in drinking water. ¹⁶ The lead test solutions were adjusted to 15 ppb PbCl₂ using deionized water. Control solutions consisted of 0.01 M NaCl and deionized water.

The experimental setup uses LabVIEW Data Acquisition software to operate at three different modes: deposition mode (anode-cathode), lead measurement mode (anode-F2), and heavy metal measurement mode (F3-cathode). The sensor was connected in series with a 100 kΩ load resistor and a 3.2 V DC power supply (Fig. 1B). The sensor was submerged in 100 ml of test solution and ran in deposition mode for one hour. Next, the LabVIEW Data Acquisition software recorded potential measurements across the load resistor every hour for 3 min in deposition mode (∆V), lead measurement (∆V₁) mode and heavy metal measurement mode (∆V₂). A MATLAB program determined the average potential for each operating configuration by averaging five data points across the span of the 3-min recording. The sensor triggered after the load resistor in lead measurement mode (∆V₁) reached a 1 V threshold potential. Tests have been conducted in stagnant and agitated solution. Agitated solutions were mixed by placing a 15.9 mm × 8.0 mm stir bar in solution on a stir plate.

All SEM images were taken with a Field Emission SEM (Hitachi SU8000 In-Line) in the secondary electron detection mode. Samples were mounted on a 4'' wafer mount and loaded into the machine. A 2 KV acceleration voltage and a 7 μA emission current were used. It should be noted that the SEM images were taken following the electrical measurements, rinsing the sensor with deionized water, and preparing the sensor for SEM imaging. The presence of lead on the sensor was confirmed by Energy Dispersive X-ray Spectroscopy (EDS, Bruker Quantax 200) with a 5 KV acceleration voltage and a 7 μA emission current. The EDS spectra of the electrodeposits had oxygen and lead composition suggesting the formation of lead oxide (Figs. S1 and S2 in Supplementary Material (available online at stacks.iop.org/JES/169/017505/mmedia)).

Fractal box-counting method was employed to calculate fractal dimension (${f}_{D}$) of the electrodeposited structure between the anode and F2. The ${f}_{D}$ analysis was performed on the SEM images of six different lead sensors after applying potential for 6, 12, 18, 24, 48, and 72 h. Individual SEM images were analyzed with ImageJ^® software and ImageJ^® plugin FracLac V 2.5 software developed by Karperien. ¹⁷ Briefly, using ImageJ^® software, a threshold was applied to the SEM images to distinctly separate pixels of the electrodeposited structures from those in the background. Next, the area of interest was selected by manually removing the anode and F2 from the SEM images. Using plugin FracLac V 2.5, the box-counting method was executed. SEM images were automatically converted into binary images and covered with grids of boxes. The software counts the number of boxes ($N$) that covers the structures and contains the pixels of interest. The pixels of interest are considered to contain meaningful information while other pixels are background or noise. Finally, the ${f}_{D}$ is calculated as follows:

$$\begin{eqnarray}&&{f}_{D\,}=\mathop{\mathrm{lim}}\limits_{\varepsilon \to 0}\displaystyle \frac{\mathrm{log}\,N}{\mathrm{log}\,\varepsilon }\end{eqnarray} \tag{ 1 }$$

where $\varepsilon$ is the length scale and is equal to the box size divided by the image size. ${f}_{D\,}\,\,$is a unit-less measure and can be calculated by measuring the negative slope of $\mathrm{log}\,N\,\,$vs $\mathrm{log}\,\varepsilon .$ When the ${f}_{D\,}\,$value nears 2, the growth is more of a compact cluster, while an ${f}_{D\,}\,$value of 1 indicates the growth is closer to a one-dimensional structure. ¹⁸

Lead ion concentration gradient and flux were numerically modeled with COMSOL Multiphysics^® 5.5. The lead ion transportation was adapted from Lupo et al. ^19,20 Due to the symmetry of the four-electrode system, the geometry was modeled and contained the electrolyte, anode, and F2. "Transport of Diluted Species" and "Level Set" were used to simulate the concentration gradient and ion flux. The transport module solved the diffusive term of the Nernst-Planck equation. ²¹ The initial concentration of the lead ion was defined as 0 mol m⁻³ at the anode, and 15 ppb (7.24 × 10⁻⁵ mol m⁻³) a distance of 15 μm away from the sensor in the z direction. The F2 and electrolyte regions between electrodes were assumed to have a no flux boundary condition. To save computation time, the mesh was defined to be coarser at a distance of 5 to 15 μm away from the electrode interface with a maximum element size and growth rate of 1.34 and 1.25μm, respectively. At the anode boundary, the mesh contained a maximum element size and growth rate of 0.43 and 1.1μm, respectively.

We designed and fabricated an interdigitated four-electrode sensor capable of detecting lead in drinking water with no false positive results (Fig. 1). As previously described, ¹² the four-electrode lead sensor operates by depositing lead oxide onto the anode. The sensor returns a positive signal for lead when the lead oxide bridges the gap between the anode and the adjacent floating electrode, F2. The presence of lead oxide in this gap lowers the resistance between the anode and F2 thus increasing the potential across the load resistor. Lin et al demonstrated the four-electrode sensor can also detect other heavy metals (i.e., Zn²⁺, Cu²⁺, and Fe²⁺) between the cathode and F3 and found that other heavy metals had minimal effect on the lead detection capability. ¹² Figure 2A shows the electrical potential signal at three different operational modes: deposition mode, heavy metal measurement mode, and lead measurement mode. Lead solutions did not trigger a response during the deposition and heavy metal measurement modes. In the lead measurement mode, the initial potential (${\rm{\Delta }}{V}_{1}$) was below the 1 V threshold potential before 18 h, oscillating below and above 1 V between 18 and 32 h, and above 1 V after 32 h. The rise in potential served as the detection signal for the presence of lead oxide on the sensor and, therefore, lead in the contacting solution. Control lead-free test solutions did not trigger a detection signal in the lead or heavy metal measurement modes (Fig. S3 in Supplementary Material).

Zoom In Zoom Out Reset image size

The detection signal is based on depositing conductive lead oxide and bridging between the anode and F2 to trigger the sensor. To visually capture this bridging event, SEM images were studied over time as illustrated in Fig. 2B. Prior to 18 h, SEM images show fast linear growth of lead oxide deposition and correspond to the observed constant ${\rm{\Delta }}{V}_{1}$ in Fig. 2A. Following 18 h of lead oxide deposition, the deposited material starts to bridge the gap between the electrodes which coincides with the oscillating potential response. At this stage, lead oxide structures are similar to very thin wires and can easily break with the passage of current during the lead measurement mode. The interplay between connection and breakage appears to slow down the overall rate of lead oxide growth as observed in Fig. 2C. By 72 h, the electrical potential signal stabilizes as a thick layer of deposited lead oxide fills the electrode gap. The presence of sufficient material between the electrodes to carry the current can eliminate the stress caused by the measurement process. Note that the lead oxide deposition has a growth rate between 0.05 and 0.15 μm h⁻¹ as indicated in Fig. 2C.

The relatively slow process of oxide growth to bridge the electrodes is due to the limitation caused by ion diffusion and is supported by the observed morphology. As shown in Fig. 2B, the morphology agrees well with an open fractal pattern suggesting the ions are performing a random walk and aggregating one-by-one (i.e., a diffusion mechanism). ^22,23 The open fractal patterns of lead oxide deposits are self-similar with a non-uniform growing front with respect to electric field lines and continuous tip splitting facilitating growth into a tree-like structure. The open fractal structures that emerge from this process can be described by the fractal dimension, ${f}_{D},$ which is an index for quantifying open fractal pattern complexity as a ratio of the change in detail ($\mathrm{log}\,N$) with respect to the change in scale ($\mathrm{log}\,\varepsilon$). As shown in Fig. 3, the ${f}_{D}$ of the lead oxide formation varied between 1.65 to 1.74, which is approximately consistent with the diffusion limited value of 1.71 reported in the literature. ^18,24

Zoom In Zoom Out Reset image size

The diffusion-limited growth was modeled with COMSOL Multiphysics® simulations and agreed with the spatial observation of lead oxide deposition across the anode. Generally, in dilute systems, the ion flux, ${{\boldsymbol{N}}}_{i},$ is governed by diffusion, migration, and convection:

$$\begin{eqnarray}\begin{array}{cccc}{{\boldsymbol{N}}}_{i}= & -{D}_{i}{\rm{\nabla }}{C}_{i} & -F{z}_{i}{C}_{i}{u}_{i}{\rm{\nabla }}{\rm{\Phi }} & +{\bf{v}}{C}_{i}\\ {\rm{flux}} & {\rm{diffusion}} & {\rm{migration}} & {\rm{convection}}\end{array}\end{eqnarray} \tag{ 2 }$$

where ${D}_{i}$ is ion diffusivity, ${C}_{i}$ is the bulk ion concentration, $F$ is Faraday's constant, ${z}_{i}$ is the ion charge, ${u}_{i}$ is the ion mobility, ${\rm{\Phi }}$ is the electric potential, and ${\bf{v}}$ is the bulk velocity. In stagnant solutions that contain highly conductive supporting electrolytes, convection and migration can be assumed to be negligible. ²⁵ Thus, Eq. 2 simplifies to:

$$\begin{eqnarray}&&{{\boldsymbol{N}}}_{i}=-{D}_{i}{\rm{\nabla }}{C}_{i}\end{eqnarray} \tag{ 3 }$$

For a diffusion limited mechanism, the ion flux is primarily driven by the concentration gradient. Figure 4A shows the 2D cross-sectional plot of the concentration gradient in the x–z plane at y = 0. The anode tip (E₁) and near-anode tip (E₂) are identified as the area of interest across the 1st anode digit. After 12 h, a drastic drop in concentration occurs at the anode tip (E₁) because this area receives a greater supply of ions giving rise to a larger concentration gradient. The large gradient at the anode tip is beneficial as the detection mechanism of the sensor relies on lead bridging across the anode and F2.

Zoom In Zoom Out Reset image size

To correlate ion flux with the spatial observation of lead oxide deposition across the 1st anode digit, two additional area of interest were identified at the anode corner (C) and anode palm (D). Figure 4B shows the ion flux across the x direction at z = 0.2 μm. The ion flux is largest at the anode tip (E₁) and corner (C), resulting in a greater possibility of forming electrodeposits at these locations. The simulated results agree with the increased deposition experimentally observed in SEM images, as shown in Figs. 4C and 4E. At the center of the anode (between D and E₁), the ion flux decreases significantly resulting in less deposition displayed in Figs. 4D and 4E.

The simplest way to decrease the detection time of the sensor is to reduce the gap distance between the anode and F2. As previously shown in Fig. 2, lead detection in the sensor geometry is a function of the time it takes for the lead oxide to form and bridge the gap between electrodes. To study the effect of gap distance on sensor detection time, sensors with different gap (3, 5, 8, 10, and 12 μm) were fabricated. Table S1 in the Supplementary Material summarizes the average actual gap distance, electrode width, and detection time of each studied device geometry. Figure 5A shows a representative curve for each gap distance. The oscillation period and detection time are significantly shorter with smaller gap distances of 3 and 5 μm, as smaller amounts of lead oxide deposits are needed to bridge the gap between electrodes.

Zoom In Zoom Out Reset image size

Figure 5B illustrates the average detection time vs gap distance. The dispersion of average detection time for 3 and 5 μm gap sensors is relatively low with the values of 19.8 ± 3.04 and 33.2 ± 5.44 h, respectively. The rate of lead oxide growth was found to be between 0.05 and 0.15 μm h⁻¹, and agrees with the growth rate range calculated in Fig. 2C. Note that, for short times and distances in Figs. 2C and 5B, the growth rate appears to be linear with a constant value of 0.15 μm h⁻¹. While this phenomenon is associated with the morphology of the growing lead oxide, further investigation is needed to determine the exact mechanism and to validate that the region truly involves linear growth. The growth rate, as determined by the relationship between gap distance and detection time, can also be used to envision an improved detection time with a smaller gap sensor (e.g., sub-micron made via nanopatterning) below what was possible in this work (i.e., 3 μm gap).

Electrolyte agitation was employed to facilitate transport of lead ions to the electrode surface and resulted in more stable electrical potential signals. The electrical response of three lead sensors are plotted for stagnant and 360 rpm agitated lead solutions in Figs. 6A and 6C, respectively. The electrical response of a stagnant lead solution exhibits noisy signals oscillating around the threshold potential, while electrical response of the agitated lead solution produces a smooth curve that reaches a steady potential. Additionally, two sensors (i.e., lead sensor 1 and 2 in Fig. 6C) show stronger signal response during agitation with a ${\rm{\Delta }}{V}_{1}$ larger than 2 V when triggered. Note that the agitated solution created a laminar flow regime, which demonstrates that the sensor can perform well in conditions consistent with the real world (e.g., laminar faucet flow). Further studies are needed to understand the effect of water flow in a turbulent regime on sensor performance.

Zoom In Zoom Out Reset image size

The improvement of signal stability could be due to the morphology of the electrodeposited material, which has been shown to be influenced by agitation. ^26–28 Compared to the stagnant solutions (Fig. 6B), the SEM images of agitated solutions (Fig. 6D) were more densely packed, had a growing front that varied in direction, and were more uniformly distributed across the anode. In agitated solutions, the dense nature of the lead oxide structures could contribute to the stable signal observed during experiments. When agitating a solution, the diffusion layer thickness decreases ²⁹ causing the rate of ion diffusion to increase. As more ions are closer to the surface of the electrode in agitated solutions, denser structures are more likely to form.

To study the lead oxide growth mechanism during agitation, the SEM images were horizontally divided at the center of the image into the top and the bottom regions, and the resulting segmented images were used to calculate ${f}_{D}.$ Despite increasing the ion transport to the surface of the electrode, ${f}_{D}$ of the top region is similar to that of stagnant solutions. However, the ${f}_{D}\,\,$of the bottom region was found to be 1.82 ± 0.04 displaying deviation from the fractal patterns by forming denser structures. As stated previously, the formation of these dense structures greatly improves the sensor's potential signal stability.