Capacitive Deionization of NaCl Solutions at Non-Steady-State

Jul 19, 2011 - In this Article, the consequence of continuous electrochemical oxidation at the positive electrode in an initially symmetrical capaciti...
0 downloads 8 Views 4MB Size
ARTICLE pubs.acs.org/JPCC

Capacitive Deionization of NaCl Solutions at Non-Steady-State Conditions: Inversion Functionality of the Carbon Electrodes Yaniv Bouhadana,* Eran Avraham, Malachi Noked, Moshe Ben-Tzion, Abraham Soffer, and Doron Aurbach Department of Chemistry, Bar Ilan University, Ramat Gan 52900 Israel

bS Supporting Information ABSTRACT: In this Article, the consequence of continuous electrochemical oxidation at the positive electrode in an initially symmetrical capacitive deionization (CDI) cell, comprising two identical activated carbon electrodes, is examined and discussed. Extensive and intensive parameters of the CDI cell are defined, and the deviations occurring among them as a result of continuous electrochemical oxidation processes at the positive electrode during prolonged charge discharge cycling are discussed. A special flowthrough CDI cell containing activated carbon fiber (ACF) electrodes was developed for this purpose. Ex situ XPS measurements were conducted to prove the presence of oxidized surface groups on the positive electrode of these cells due to cycling. A surprising phenomenon that looks like an inversion functionality of the carbon electrodes occurring after numerous charge discharge cycles is observed and explained.

’ INTRODUCTION Capacitive deionization (CDI) may serve as an energy efficient technique for salt removal from water.1 5 The charged species are electroadsorbed onto high surface carbon electrodes, whereas a potential difference is applied between the carbon electrodes. Fortunately, most of the charges injected into high surface carbon electrodes are utilized for electrostatic interactions, and the charged species participate in the construction of the electrical double layer (EDL). The charged species can be easily released back into the flow solution when the potential difference applied is altered or canceled. The CDI method was already described in detail in previous publications.1 11 One important advantage that CDI is expected to offer over other competitive water desalination methods like reverse osmosis (RO)12 and electrodialysis (ED)13 is long-term stability. For example, problems like biofouling on the membrane surface in the RO method14 or irreversible hydroxide precipitation due to extreme pH changes nearby the ion exchange membrane in the ED method13 require special maintenance. The commonly accepted model of the CDI cell as described in the literature is that of a simple electrical circuit composed of two capacitors in series (Figure 1). Because it is assumed that only electrostatic interactions take place, the desalination process (charge discharge cycle) could be expected to last for numerous cycles without deterioration. Another important aspect of the behavior of CDI cells relates to the electrodes’ working potential ranges and the ratio between adsorption of counterions and desorption of coions. The ideal situation is presented in Figure 2, which shows a schematic dependence of the amounts of counterions adsorbed and co-ions desorbed as a function of the electrodes’ potential. r 2011 American Chemical Society

Figure 1. Basic model of a capacitive deionization cell as a simple electrical circuit analog composed of two capacitors in series.

The behavior of both the negative and the positive electrodes is demonstrated. The charts of Figure 2 reflect a symmetric CDI cell in which the electrodes are identical and possess large enough pores, so there is no difference in the specific capacitance of the electrodes for anions and cations adsorption. However, because not only electrostatic interactions but also Faradaic interactions may be involved in the desalination process, the whole desalination system may exit the steady-state, and as a consequence, the Received: May 22, 2011 Revised: July 11, 2011 Published: July 19, 2011 16567

dx.doi.org/10.1021/jp2047486 | J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

Figure 2. Schematic description of adsorption/desorption response of symmetric activated carbon electrodes. Their pores are wide enough to allow a similar adsorption of anions and cations. The amount of counterions adsorbed and co-ions desorbed is plotted as a function of potential (arbitrary units).

desalination performance may decrease in the long term, as explained below. Also, changes in the electrodes’ structure and surface due to side reactions may drive the CDI cell away from symmetric behavior. The work presented herein is critically important for the application of CDI in water desalination processes. It is also relevant to other topics related to activated carbon electrodes, such as super (EDL) capacitors, in which the stability of the electrodes is also critically important. Desalination System at Steady-State. The important parameters in an electrochemical desalination process are the salt removal rate (Δn), accumulation of charge (Δq), and charge efficiency (η), where charge efficiency is defined as the ratio between the charge used for salt removal (derived fromΔn) and the charge consumed during the charge discharge cycle (Δq).15 Let us define the following parameters in the desalination process: a The electric capacity of each electrode in the desalination cell (C1 and C2). b The ion transport in the carbon electrode pores (adsorption/ desorption) in relation to the potential difference applied: Γ+ (E) for cation adsorption and Γ (E) for anion adsorption. In previous work,16 we demonstrated that the behavior of the cation and anion is quite symmetric with respect to the electrodes’ PZC (as illustrated in Figure 2) if the electrodes have open pore structure. c The carbon electrode PZC,17 which is derived from the measured potential of the carbon electrodes versus Ag/ AgCl electrode at shortcut circuit potential (E0). d The potential difference distribution between each electrode versus Ag/AgCl/Cl reference electrode in the desalination cell (E1 and E2, respectively). e The overall potential applied to the CDI cell E (= E1 + E2). The first three parameters are intrinsic features of the system. The last two are external features, which depend on the operational conditions of the entire CDI cell. Note that the only controllable parameter is the potential applied to the desalination cell (E). When one describes a CDI desalination system at steady-state, no deviation of the parameters mentioned above should be observed. Determining whether our system is at steady-state entails verifying that the measurable parameters, Δn, Δq, and η, remain constant throughout the charge discharge cycles.

Figure 3. Schematic representation of the unit cell in a “flow through” operation mode.

The surface interactions are not necessarily electrostatic (nonFaradaic), as reported in the literature.7,8,16,18 There are various possible Faradaic reactions that may occur during CDI processes, which may involve pH changes and chemical changes in the carbon surface.7 In this Article, the consequence of continuous changes of the carbon electrodes in CDI cells due to electrochemical surface oxidation processes resulting in some interesting effects on the desalination parameters and subsequently on the desalination performance is examined and discussed. Some interesting effects on the desalination parameters and subsequently on the desalination performance are also described.

’ EXPERIMENTAL SECTION System Setup. The system is composed of a desalination reactor, a PC-controlled power supply, a conductometer, a pH meter, a bidirectional gear pump, solenoids, and a storage bulb, all of which are illustrated and well-described in a previous publication.7 The entire system is connected to a PC and controlled by a homemade customized program created using LabVIEW software, which controls the applied voltage and the solution flow rate and reads and plots all the pertinent system data. Sodium chloride solution (1000 ppm >99% Frutarom, Israel) was circulated by a gear pump through the conductivity sensor and the electrochemical cell using a minimal volume of tubing. 16568

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

Figure 4. Cumulative charge for each charge discharge cycle versus time during periodic experiments in a “flow by” mode of operation at various applied potentials with and without nitrogen purging, as indicated.

Preliminary experimental work was carried out to choose the optimized flow rate conditions. The chosen solution flow rate was 50 mL/minute. Both slower flow rate and faster flow rates demonstrated nonuniform results in the conductivity measurements because of diffusion effects (too slow flow rates) or turbulence (too fast flow rates) inside the cell. When an inert atmosphere was needed (e.g., to eliminate any possible influence of oxygen reduction), the solutions were deaerated by nitrogen purging (99.999%). We chose to work with solutions containing 1000 ppm of NaCl to explore methodology and conditions that are relevant to brackish water concentration. As already demonstrated by Oren,21 CDI is more effective than RO for brackish water, containing only a few thousands of ppm NaCl. Note that there are many sources of brackish water throughout the world. Thereby, CDI is important because it rivals RO for desalination of brackish water. Flow-through CDI cell. Unit Cell. The cell is defined as a “flow-through” capacitor in which the solution is forced to flow through the activated carbon electrodes. (The diameter of the ACF was 54 mm.) A sheet of a porous glassy paper cloth served as a separator, and a sealing gasket was formed by soaking silicon glue into the separators’ rims. A poly tetrafluoroethylene (Teflon) ring spacer formed a 1 mm gap filled by two activated carbon cloth sheets. A flexible impermeable graphite paper ring served as the current collector. This was a ring-shaped, electrically conductive carbon sheet whose internal diameter was 4 mm smaller than the outer diameter of the carbon cloth. Thus, there was a 2 mm radial overlap between the current collector and the activated carbon electrode to secure the electrical contact between them. Covers, with distributors, ensure the homogeneous flow of the solution throughout the ACFs. A schematic representation of a unit cell is illustrated in Figure 3. The CDI reactor consists of a 100 unit cell (50 positive, 50 negative). The ACF used was a wide-pore size commercial carbon cloth (ACC-5092-15) from Nippon Kynol Japan with high surface area of 1500 m2 g 1 (B.E.T). Surface analysis of the carbon electrodes by photoelectron spectroscopy (XPS) was performed with a Kratos AXIS-HS spectrometer using a monochromatic Al Ka source (1486.68 eV). All binding energies were calibrated to a carbon 1S graphitic peak position at 285 eV.

’ RESULTS AND DISCUSSION In a previous work, we explored CDI processes in cells that had “flow-by” architecture, under various potentials and gaseous

Figure 5. Variation of solution pH measured at the outlet of the CDI cell (pH vs time) for each charge discharge cycle at various applied potentials with and without nitrogen purging, as indicated.

environments. These studies provided us with the appropriate starting point for the present study. Figure 4 presents important results related to experiments with the “flow-by” cells in which different voltages were applied to them (500, 700, 900, and 1100 mV) and each charge discharge cycle lasted 60 min. The activated carbon electrodes of these cells were used without any pretreatment. Figure 4 exhibits the amplitude of the cumulative charge versus time for each charge discharge cycle measured as a function of the different voltages applied, with and without nitrogen purging.7 A gradual increase in the cumulative charge/ time profile is seen when a voltage of 900 mV is applied to the cell without nitrogen purging. This increase in the cumulative charge has to be connected to Faradaic currents (red-ox reactions), which are more pronounced as the voltage applied to the cell is higher. It is important to note that the Faradaic processes were clearly indicated in these preliminary experiment,7 although the voltages applied to the cell were much below the electrolysis potential of water (1.23 V). The relevant Faradaic processes can belong to the reduction of oxygen dissolved in the water and to possible red-ox reactions of water, the carbon surfaces, or both. 16569

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

Figure 6. Conductivity/time profile of the solution measured at the outlet of the cell in a CDI experiment during 5 days. Each charging or discharging step lasted 60 min, whereas the voltage applied to the cell fluctuated between 900 and 0 mV.

Such Faradaic processes are expected to affect the pH of the solution in the cell. We therefore also measured changes in the pH of the solution at the outlet of the CDI cell. The pH response of the periodic charge discharge experiments is depicted in Figure 5 (pH vs time measured during cycling). Periodic changes in pH during cycling are indeed observed in these measurements and, as exhibited in Figure 5, are greater as the voltage applied is higher and decrease when the solution is purged with nitrogen. Additionally, to evaluate what happens to the electrodes and their surface chemistry upon cycling, we measured electrodes before and after cycling by XPS with an emphasis on oxygen spectra. Representative O1s XPS spectra of these carbon electrodes are presented in the Supporting Information, Figure 1S. These studies prove that the cycled positive electrode underwent oxidation and contain oxygenated surface groups. To examine the effects of these Faradaic reactions on the desalination performance over an extended time scale, a new deionization cell in an “flow-through” architecture was assembled, as described in Figure 3. The new CDI cell was tested in repeated 60 min charge discharge cycles over five days periods, whereas a periodic voltage of 900 and 0 V (short) was applied to the cell. The solution concentration used for the rest of this work was 1000 PPM NaCl (∼1.8 mS) for the reasons explained above. Figure 6 shows the conductivity/time profile of this experiment. Intuitively, we can see that the system did not reach steadystate at any point. Moreover, a gradual decrease in the salt removal amplitude was observed. At minute 5800 during this experiment, the system, which was under air, was purged with nitrogen. As a consequence, a gradual increase in the conductivity amplitude versus time was obtained (i.e., after dissolved oxygen was removed, the cell had partially recovered in terms of desalination ability). In addition, a unique phenomenon was observed as the desalination process progressed; there was a slight increase in the conductivity at the moment the potential difference was applied (i.e., at the beginning of the “charge” stage of the charge discharge cycle). This increase, shown as a peak in the expanded scale section in Figure 6, decayed very rapidly. This peak, however, became more pronounced as the number of

Figure 7. Typical steady-state cyclic voltammogram of the activated carbon electrodes used herein in 1000 ppm NaCl aqueous solution.

charge discharge cycles increased. Intuitively, this would mean that the charge consumed at the beginning of the charging step is used for desorption rather than adsorption of salt, indicating that the electrode stores a significant amount of salt at the end of each discharge cycle. Further demonstration of the phenomena discussed above can be seen upon long-term measurements, in which the electrode’s stability was examined during prolonged periodic CDI processes. Typical results of these experiments are presented in the Supporting Information (Figure S2). Figure 7 exhibits a representative, typical steady-state cyclic voltammetry of the activated carbon electrodes that were used in the CDI cells in the experiments described herein. The CV curves of these carbon electrodes are symmetrical, showing the typical expected butterfly shape. The potential at which the I/E curve has a minimum can be assigned as the PZC of these electrodes. This symmetrical shape of the CV curves means that the adsorption of cations (Na+) and anions (Cl ) to these electrodes is similar (no molecular sieving or specific adsorption effects effect19). Therefore, a CDI cell comprising two electrodes such as those related to Figure 7 is initially fully symmetric: C1(E) = C2(E), which results in an a symmetric distribution of the potential (E) over the two activated carbon electrodes in the desalination cell (E1 = E2). The results of the periodic CDI experiments with these cells show that these initial situations change. The distribution of potential between the electrodes ceases to be symmetrical. 16570

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

Figure 8. Potential distribution on the carbon electrodes of the CDI cell used herein versus a reference electrode (Ag/AgCl/Cl ) during periodic CDI experiments (900 0 mV, no nitrogen purging).

E1 and E2 become different, and the PZC of the electrodes changes as well. E0, the short-circuit potential, is shifted to a more positive value, as explained in previous work.16 Because the surface oxidation is a continuous process, so too are the changes at E1 and E2. To measure directly the change that each electrode in the CDI undergoes, we introduced a reference electrode (Ag/ AgCl) to the cell and thus were able to measure each electrode separately. The difference in the potential distribution between the electrodes was recorded during the periodic cycling experiment and is shown in Figure 8. At first, the potential distribution is quite symmetric, but as the charge discharge cycles progress, E1 > E2 (where E is constant). We believe that this indicates that there is a leakage current due to a Faradaic side reaction at the negative electrode (e.g., reduction of trace oxygen that is unavoidably present in the solution). Such a reaction needs relatively small polarization potential negative to the PZC of the electrode. Hence, only a small portion of the 900 mV applied to the cell falls on the negative electrode, which means that most of the potential applied to the cell falls on the positive electrode. Consequently, the potential that the positive electrode sees may exceed its potential limit of stability. The potential that the positive electrode can reach in the asymmetric situations that develop in the CDI cell may be high enough for both water oxidation and oxidation of the carbon surface as well. In previous work,19 we were able to calculate the amount of sodium and chloride adsorbed onto/desorbed from activated carbon fiber electrodes (similar to those presently used) in CDI experiments (denoted herein as Γ+ (E) and Γ (E)). It was shown there that in aqueous NaCl solutions the charge consumption at potentials close to the PZC of the electrodes ((∼150 mV) is divided almost equally between the desorption of the co-ions and adsorption of the counterions. As a result of the asymmetric situation developed in the CDI experiments, as described above, E0, E1, and E2 deviate from their original values, so the functions Γ+ (E) and Γ (E) are also no longer symmetric with respect to E0. Moreover, reduction of trace oxygen and the consequent push of the positive electrode to water oxidation potentials may also contribute to the “inversion effect” that is seen during prolonged repeated measurements. When a potential is applied to the cell, three processes that affect the ions concentration in solution can occur simultaneously: 1 The expected adsorption of counterions (i.e., Cl ions to the positive electrodes and Na+ ions to the negative electrode).

2 Desorption of co-ions (i.e., Cl ions from the negative electrode and Na+ ions from the positive electrode). It is noteworthy that, unlike the present study, the great majority of works published on CDI do not mention the case of coion desorption. 3 Reduction of trace oxygen on the negative electrode that produces HO ions and oxidation of water on the positive electrode that produces H+. (Obviously, neutralization of these two ions occurs because of the continuous mixing of the contents of the two-electrode CDI cell.) The conductivity response of the CDI experiments (measured at the outlet of the CDI cells) always reflects a balance among these three processes. It is interesting that the steady state is not achieved at all during prolonged cycling. We are measuring a dynamic situation rather than any kind of steady state. When the solution was purged with nitrogen during the CDI experiment, the conductivity/time profile was found to gradually increase (denoted in the right side of Figure 6). This increase should not be attributed to any release of oxidized surface groups but rather to more efficient utilization of the induced charges at the EDL of the electrodes due to the removal of oxygen. The lack of oxygen in the system (due to purging with nitrogen) avoids a situation in which a Faradaic reaction takes place at the negative electrode (oxygen reduction), which means that mostly electrostatic interactions are relevant to the cell (i.e., more effective desalination). In such a case, the potential applied to the cell is distributed more symmetrically between the electrodes, and hence the charge consumed by the electrodes upon their polarization is counterbalanced at the solution side by electroadsorption of the ions. Surface parasitic reactions with H2O can still take place at the positive electrode in the absence of oxygen, but their kinetics is dramatically decreased. This is evident in Figure 9, which represents the cumulative charge (a good indication for Faradaic reactions), the magnitude of salt removal, and the salt/charge efficiency15 as a function the charge discharge cycles before and after nitrogen introduction. The presented results relate to solutions containing 1000 ppm NaCl with a flow rate of 50 mL/g. These parameters may have an impact on the efficiency of the CDI processes. Starting from low concentrations up to 2000 ppm, an increase in the concentration leads to higher charge efficiency in the desalination process.20The charge efficiency may reach a maximum at a certain salt concentration (leaving all other parameters such as voltage applied or flow rate invariant). Then, beyond this 16571

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

Figure 9. Cumulative charge, the magnitude of salt removal, and the salt/charge efficiency as a function of the charge discharge cycles in a continuous CDI process over 5 days. Each cycle step lasted 60 min. 900 0 mV.

Figure 10. Comparison of the cell behavior under normal and inverted polarization.

maximum, an increase in the salt concentration should result in a lower charge efficiency due to the adverse effect of salt concentration on the ions transport dynamics in solutions and the negative impact of desorption of co-ions on the charge efficiency.21 The effect of the flow rate on the charge efficiency may also behave as a function with a maximum. At too-low flow rate, the removal of ions upon discharge may be too slow. However, at too-high flow rates, we may have turbulence in the CDI cell, which avoids good separation among feed, product, and waste solutions. Also, at too-high flow rates, the feed solution cannot reach the necessary retention time required for effective electroadsorption of the ions.20 It is clear that the system can partially recover by introducing nitrogen due to the inhibiting of surface reactions and consequently better utilization of the charge for electrostatic reactions. However, when the potential applied to the cell was inverted (i.e., a potential difference of 900 mV was applied to the cell), we were able to bring the CDI cell back almost to its original state (Figure 10). This means that the electrochemical oxidation of the carbon surface is, at least in part, a reversible process: cathodic

polarization of oxidized carbons probably removes some of the oxygenated surface groups. A possible mechanism for the decrease in productivity of prolonged CDI processes in which cells comprising activated carbon electrodes was previously suggested: The inversion in the functionality of the CDI cell results from continuous changes in the PZC of the positive carbon electrode due to its surface oxidation. Consequently, after the surface oxidation reaches a certain level, the change in the ions’ behavior at the positive electrode with respect to the applied potential (Γ+ (E) and Γ (E)) results in a shift of the potential distribution in the cell even after the removal of oxygen by purging with nitrogen. The oxidation of the positive electrode may change its surface morphology in a way that reduces considerably its capacity. Because the specific capacity of the negative electrode remains nearly unchanged whereas that of the positive electrode deteriorates due to oxidation, the potential is distributed unevenly, lower at the negative electrode and higher at the positive electrode (i.e., E1 > E2). This asymmetric behavior of the electrodes obviously results in a decrease in salt removal per charge discharge cycle. Surprisingly, after inversion of the electrodes’ functionality by inverting the potential applied to the cell, the system may reach a steady state that can be attributed to the following reasons: 1 The oxidation process is limited and stops after the carbon electrode reaches a certain level of coverage by surface groups. 2 The oxygenated surface groups can be reduced in the aqueous solutions upon cathodic polarization of the electrode.

’ CONCLUSIONS We studied and addressed important problems that can adversely affect the effectiveness of electrochemical water desalination (CDI processes). CDI cells comprising identical activated carbon electrodes, in which initial behavior is fully symmetric (similar specific capacity for adsorptions of anions and cations at potentials positive and negative to the PZC, respectively), may lose their symmetry. The most probable reason is the presence of trace oxygen in the aqueous solution, which can be easily reduced on the negative electrodes. This situation automatically pushes the positive electrodes to potentials much higher than E/2 (E being the potential applied to the cell). Hence, even if E is much lower than the electrochemical window of the solution, the asymmetric situations developed in the CDI cell when there is a parasitic cathodic reaction in the cell 16572

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573

The Journal of Physical Chemistry C

ARTICLE

may push the positive electrode to dangerous surface oxidation. Such a change in the surface properties of the positive electrode can further increase the difference between the positive and negative electrodes. Such situations may bring CDI processes to demonstrate an inversion behavior, that is, increase in the ions concentration in solution when the CDI cell is charged. This means that the desired desalination process by adsorption of counterions is fully masked by two opposing processes: 1 Desorption of co-ions. 2 Formation of protons and hydroxide ions due to the parasitic processes that may exist in these systems (mainly trace oxygen reduction). After the inversion occurs, the CDI process may achieve a steady state. This inverted behavior of CDI processes may be reversible. It is possible to reduce the surface group formed on the positive electrode by cathodic polarization (inverted potential). To make CDI practical, it is clear that an effort must be made to avoid this electrochemical oxidation of the carbon electrodes. This might be accomplished by the selection of appropriate carbons that are resistive to oxidation or can be made so by appropriate pretreatments.

’ ASSOCIATED CONTENT

bS

Supporting Information. O1s spectra obtained from XPS measurements of pristine and cycled activated carbon electrodes (after periodic CDI experiments, 900 0 mV). Conductivity/time profile over a 6 week period employing 60 min charge discharge cycles with periodic application of 900 and 0 V (short) to the cell. This material is available free of charge via the Internet at http://pubs.acs.org.

’ REFERENCES (1) Johnson, A. M.; Newman, J. J. Electrochemic. Soc. 1971, 118, 510. (2) Oren, Y.; Soffer, A. J. Appl. Electrochem. 1983, 13, 473. (3) Oren, Y.; Soffer, A. J. Appl. Electrochem. 1983, 13, 489. (4) Farmer, J. C.; Fix, D. V.; Mack, G. V.; Pekala, R. W.; Poco, J. F. J. Electrochem. Soc. 1996, 143, 159. (5) Farmer, J. C.; Fix, D. V.; Mack, G. V.; Pekala, R. W.; Poco, J. F. J. Appl. Electrochem. 1996, 26, 1007. (6) Anderson, M. A.; Cudero, A. L.; Palma, J. Electrochim. Acta 2010, 55, 3845. (7) Bouhadana, Y.; Ben-Tzion, M.; Soffer, A.; Aurbach, D. Desalination 2011, 268, 253. (8) Lee, J.-H.; Bae, W.-S.; Choi, J.-H. Desalination 2010, 258, 159. (9) Zhao, R.; Biesheuvel, P. M.; Miedema, H.; Bruning, H.; van der Wal, A. J. Phys. Chem. Lett. 2009, 1, 205. (10) Li, H.; Zou, L.; Pan, L.; Sun, Z. Sep. Purif. Technol. 2010, 75, 8. (11) Villar. Energy Fuels 2010, 24, 3329. (12) Radjenovic, J.; Petrovic, M.; Ventura, F.; Barcelo, D. Water Res. 2008, 42, 3601. (13) Xu, T.; Huang, C. AIChE J. 2008, 54, 3147. (14) Flemming, H. C. Exp. Therm. Fluid Sci. 1997, 14, 382. (15) Avraham, E.; Bouhadana, Y.; Soffer, A.; Aurbach, D. J. Electrochem. Soc. 2009, 156, P95. (16) Avraham, E.; Noked, M.; Bouhadana, Y.; Soffer, A.; Aurbach, D. Electrochim. Acta 2010, 56, 441. (17) Tobias, H.; Soffer, A. J. Electroanal. Chem. 1983, 148, 221. (18) Muller, M.; Kastening, B. J. Electroanal. Chem. 1994, 374, 149. (19) Avraham, E.; Yaniv, B.; Soffer, A.; Aurbach, D. J. Phys. Chem. C 2008, 112, 7385. (20) Xu, P.; Drewes, J. E.; Heil, D.; Wang, G. Water Res. 2008, 42, 2605. (21) Oren, Y. Desalination 2008, 228, 10. 16573

dx.doi.org/10.1021/jp2047486 |J. Phys. Chem. C 2011, 115, 16567–16573