The following chemicals were obtained from Sigma-Aldrich: sodium silicate solution, ascorbic acid, sulfuric acid, polytetrafluoroethylene (PTFE) solution. Ammonium molybdate tetrahydrate was purchased from Fluka. Carbon black (Vulcan^® XC72R) was purchased from Cabot. Powdered activated carbon (CEP21) was purchased from Power Carbon Technology. Ethanol was purchased from Daejung Chemicals & Metal. Anion exchange membrane (ASE) and cation exchange membrane (CSE) were purchased from Astom.

Carbon electrodes were prepared using powdered activated carbon, carbon black, and PTFE in an 86:7:7 weight ratio, respectively. A mixture containing these components was kneaded with a few milliliters of ethanol until it solidified, after which sheet-type electrodes were made by pressing with a roll-pressing machine. The desired thickness was then obtained by adjusting the gap between the two rollers; the final thickness was approximately 250 μm. The electrodes were dried in a vacuum oven at 120°C overnight to remove the residual ethanol. After drying, the electrodes were cut into 50 mm diameter circles with a center hole having a 4 mm diameter.

Three kinds of CDI configurations were employed for silicate removal using a custom-built electrochemical cell with built-in graphite current collectors. For the typical-CDI configuration, a pair of round-shaped carbon electrodes with a center hole was placed onto the current collectors. Two nylon spacers (thickness 100 μm) were positioned between the two electrodes to allow a feed solution to pass from the outside to the center hole. For MCDI, ion exchange membranes were incorporated into conventional CDI: an anion exchange membrane covered the anode, and a cation exchange membrane covered the cathode, where each was sized to match its corresponding electrode. For HCDI, the graphite current collector was used as the cathode while the assembly of the anion exchange membrane and carbon electrode was used as the anode. In particular, this HCDI was a modified form of MCDI where the cathode assembly (cation exchange membrane + carbon electrode) was removed.

The silicate removal performance of three configurations (typical-CDI, MCDI, and HCDI configurations) was evaluated in a flow-mode cell with a flow rate of 10 mL/min. A feed solution containing 1.0 ppm silicate was used. The silicate removal step was performed by applying a constant voltage ranging from open circuit voltage (OCV) to 18 V using a potentiostat (VSP); each voltage step was maintained for 6 min. The silicate concentration was determined from the effluent at 2 and 4 min during each step, and the pH was measured from the effluent at 4 min of each step. The molybdenum blue method [12] was used to determine the silicate concentration, and the pH of the solution was measured using a bench-top pH meter (Thermo Fisher). After passing through the cell, the effluent was not re-circulated into the influent (single-pass mode).

The silicate recovery test was conducted via adsorption and desorption stages. For the adsorption stage, a 1.0 ppm silicate solution was passed through the cell for 3 h at a flow rate of 10 mL/min under 18 V. Then, for the desorption stage, −20 mA was applied with a flow rate of 0.3 mL/min for 33 min. During the desorption stage, 10 mL of solution was collected in total, and this was used to determine the silicate concentration.

This study investigated the removal of silicate in typical-CDI and MCDI configurations under varying voltage conditions (Fig. 1). The experiments were conducted with different voltage steps, starting with an OCV to stabilize the fluid flow and confirm the physicochemical adsorption of each configuration. The voltages increased from 1 to 18 V and each voltage step was maintained for 6 min.

No physicochemical adsorption of silicate was observed in the typical-CDI configuration. As the voltage increased, the current intensity increased, but no reduction in silicate concentration was detected in the effluent. Similarly, the MCDI configuration showed no physicochemical adsorption of silicate at the OCV step. No silicate removal was observed up to 7 V, but silicate removal began to be observed from 9 V. When evaluating the performance at each voltage step based on the average observed effluent concentration, the proportions of silicate removal at 9, 12, 15, and 18 V were 1.6, 3.8, 4.6, and 6.7%, respectively. Although the MCDI configuration showed silicate removal, the performance was insignificant.

The results indicated that both typical-CDI and MCDI were ineffective for silicate removal. To understand the cause of the ineffectiveness of silicate removal using the CDI technique, the behavior of silicate in the solution was investigated. Silicate either gains or loses charge depending on the pH of the solution (Fig. 2). For example, silicate has a −1 charge at pH 8 and above, while it has a −2 charge at pH 12 and above. Consequently, at pH 9.9 and higher, total charged silicate ([H₃SiO₄ ⁻] + [H₂SiO₄ ²⁻]) becomes a major species. However, at pH below 8, it exists as silicic acid (Si(OH)₄) with no charge. The ineffective removal of silicate in typical-CDI and MCDI is attributed to this pH-dependent charge property of silicate. In fact, the pH of 1.0 ppm silicate stock solution is 6.7. At this pH, Si(OH)₄ is the dominant species that has no charge. A similar phenomenon was reported by Mossand and Zou [13]. In their system with a feed solution pH of 7.7–8.0, they observed no dissolved silica removal in CDI. Conversely, Ma et al. reported 25% of dissolved silica removal in a flow-through CDI configuration, which was mainly attributed to local pH changes at the electrode surface [11]. This local pH change was used to keep silicate in a stable ionic state, thus allowing silicate to be removed from the solution by electrochemical sorption. Considering the pH speciation of silicate and previous research, an increase in pH is needed for effective silicate removal.

Section 3.1 has highlighted the limitations of silicate removal using conventional CDI technologies. Effective silicate removal via CDI techniques requires pH conditions above 8. While the pH of the solution could be increased by the addition of NaOH or borate buffer, the addition of these chemicals necessitates further post-treatment.

A key advantage of electrochemical technologies is the ability to introduce useful reactions through applied voltage. In electrochemical systems, pH can be manipulated through water reduction, oxidation, or dissociation, thus producing H⁺ or OH⁻ from water (Eq. (1)–(3)) [14, 15]. Among these, the water reduction reaction can increase the pH of the solution.

The pH-increasing water reduction occurs at the cathode. As silicate maintains a stable negative charge at high pH, it would be expected to be removed via electrochemical sorption at the anode. This separation of pH elevation and silicate removal functions allows for specific reactions to be assigned to each electrode.

In our electrochemical set-up, the graphite current collector can be utilized as the cathode instead of the fabricated carbon electrode designed for capacitive performance. For the anode, either typical-CDI or MCDI configuration is possible for silicate removal. However, based on the experimental results presented in Fig. 1, the MCDI configuration is expected to give a higher silicate removal efficiency than the typical-CDI configuration. This arrangement allows for the construction of an HCDI configuration, where a Faradaic reaction (i.e., water reduction reaction) increases pH, while a non-Faradaic (capacitive) reaction (electrochemical sorption) removes silicate from the solution. Fig. 3 illustrates the expected behavior of silicate in the HCDI configuration compared to typical-CDI and MCDI configurations. The proposed HCDI configuration is expected to achieve higher silicate removal efficiency than that achieved by the previous study [11] for two reasons: i) the HCDI configuration can overcome the limitations of conventional CDI operating conditions that avoid water electrolysis by intentionally inducing a pH-increasing reaction at one electrode, and ii) the HCDI configuration induces pH changes across the entire channel between electrodes, rather than inducing local pH changes at the electrode surface, potentially resulting in a larger volume of high pH solution.

The removal efficiency of silicate in the HCDI was evaluated (Fig. 4a). In the HCDI, silicate removal was observed from 2 V. A drastic increase in removal efficiency was observed from 3 V, with the removal rate saturating at 73% at 12 V. The silicate removal efficiencies of typical-CDI, MCDI, and HCDI were compared at various operating voltages (Fig. 4b). While no silicate removal was observed in typical-CDI at any voltage condition, MCDI showed slight removal from 9 V, although this removal was insignificant (with a maximum removal rate of 6.7% at 18 V). HCDI exhibited silicate removal from 2 V, achieving a maximum removal rate of 75% at 18 V. This removal rate is 11 times higher than the maximum removal rate of MCDI.

The high removal efficiency of HCDI is considered to be attributable to the pH increase caused by the water reduction reaction at the cathode. To verify this, the pH of the effluent (pH_eff.) was investigated in the HCDI system (Fig. 5a). Compared to the initial pH of 6.7 for the 1.0 ppm solution, both typical-CDI and MCDI were found to lower the pH at all voltage conditions. It is believed that these results occur because the ions removed from the solution by electrochemical sorption are primarily Na⁺, which is the counter ion for silicate in sodium silicate solution and OH⁻ anions, finally leaving H⁺ in the solution.

In the HCDI configuration, pH_eff. was observed to exceed the initial pH from 2 V. The pH_eff. increased with applied voltage, reaching a maximum of 9.1 at 9 V. Subsequently, pH_eff. decreased from 12 V, ultimately dropping to 7.1 at 18 V. While pH_eff. elevation was confirmed in HCDI, pH_eff. alone cannot fully explain the performance of HCDI. This is because the expected distribution of total ionic silicate is only 14% at pH 9.1 (Fig. 2). It must be noted that pH_eff. is a combined result of pH changes occurring at both the anode and cathode. Even if pH increases at the cathode, pH may decrease at the anode due to OH⁻ adsorption. It is therefore necessary to investigate the pH of the channel solution between the anode and the cathode, which is influenced by the cathode reactions (i.e., water reduction reaction).

Assuming that only OH⁻ generation occurs at the cathode in the HCDI configuration, the pH of the channel solution (pH_exp.) can be estimated using Eq. (4) and (5) as follows:

where I_avg is the time-averaged current for each voltage step, HRT is the hydraulic retention time of the solution in the channel (2.36 s), n is the number of moles of electrons (1 for Eq. (1)), F is the Faraday constant (96485.3321 C/mol) [14], V is the channel volume (0.39 mL), and K_w is the dissociation constant of water (10⁻¹⁴ M²) [16].

pH_exp. was calculated from the current data of each voltage step (Fig. 5b). As the voltage increases, pH_exp. increases, consequently increasing the proportion of total charged silicate. The silicate removal rate also increased with the proportion of total charged silicate. Notably, from 2 V, the total charged silicate reached 59%; as charged silicate species became dominant, the removal of silicate began. At 12 V, where the removal rate began to saturate, the proportion of total charged silicate also converged to 94%. The results from pH_exp. provide a more intuitive and accurate interpretation than those from pH_eff. Consequently, the increase in pH_exp. by the HCDI configuration maintained the charge of silicate and ultimately allowed for effective silicate removal.

Energy efficiency is a crucial consideration in processes using electrical energy, as it directly impacts operational costs. Additionally, electrical energy consumption serves as a representative figure-of-merit for system performance, facilitating comparisons with other technologies. For these reasons, the electrical energy consumption for silicate removal by HCDI was investigated (Fig. 6). Energy consumption was the lowest at 3 V (148.8 Wh/g), and it increased with voltage increase. While 3 V is the most energy efficient stage, when considering both energy efficiency and removal rate, 7 V or 9 V could emerge as the optimal operating voltage. This is because, when compared to 3 V, the energy consumption at 7 V and 9 V increased by 1.6 and 2.3 times, respectively, but the removal rate increased from 12.4% to 50.6% and 63.9%, respectively representing 4- and 5-fold increases. The optimal operating condition identified below 9 V is likely due to the fact that, at voltages above 9 V, oxygen evolution reaction (OER) begins to hinder system performance. As the voltage increases below 9 V, energy consumption rises moderately; however, above 9 V, it increases sharply. This sharp increase is likely due to the intensified OER, which parasitically consumes electrical energy without effectively contributing to silicate removal. By investigating energy consumption, it was possible to identify operating conditions that allow effective system performance while avoiding the detrimental effects of OER.

Since HCDI removes silicate utilizing a part of the MCDI configuration, silicate can be recovered by desorption of adsorbed ions. In our experiments, a 1.0 ppm feed was passed through the system at 18 V with a flow rate of 10 mL/min for 3 h. The subsequent desorption process was conducted at −20 mA with a flow rate of 0.3 mL/min for 33 min. As a result, a total of 10 mL of concentrated solution was obtained. This solution had a concentration of 10.3 ppm, which was approximately 10-fold higher than that of the feed solution. When comparing the amount of silicate passing through the system with the amount collected, the silicate recovery rate was 5.8%, likely due to the low desorption efficiency of silicate. The desorption process also requires a high pH environment for effective ion expulsion from the electrode to the solution via an electric field. In the HCDI configuration, pH cannot be increased during the desorption process due to the reverse polarity. Since complete desorption of silicate was not achieved electrochemically, there were concerns regarding the recyclability of the system. However, after disassembling the cell, rinsing the carbon electrode, and reassembling it for reuse, no significant decline in silicate removal efficiency or recovery was observed in the HCDI configuration. In conclusion, while the HCDI configuration is effective for silicate removal, it faces certain limitations in terms of silicate recovery. To overcome these limitations, further research is needed to improve the desorption of silicate.