Article pubs.acs.org/est
Energy Recovery in Membrane Capacitive Deionization Piotr Długołęcki* and Albert van der Wal Voltea B.V., Wasbeekerlaan 24, 2171 AE Sassenheim, The Netherlands ABSTRACT: Membrane capacitive deionization (MCDI) is a water desalination technology based on applying a cell voltage between two oppositely placed porous carbon electrodes. In front of each electrode, an ion-exchange membrane is positioned, and between them, a spacer is situated, which transports the water to be desalinated. In this work, we demonstrate for the first time that up to 83% of the energy used for charging the electrodes during desalination can be recovered in the regeneration step. This can be achieved by charging and discharging the electrodes in a controlled manner by using constant current conditions. By implementing energy recovery as an integral part of the MCDI operation, the overall energy consumption can be as low as 0.26 (kW·h)/m3 of produced water to reduce the salinity by 10 mM, which means that MCDI is more energy efficient for treatment of brackish water than reverse osmosis. Nevertheless, the measured energy consumption is much higher than the thermodynamically calculated values for desalinating the water, and therefore, a further improvement in thermodynamic efficiency will be needed in the future.
■
INTRODUCTION Sufficient access to clean, safe, and inexpensive water for human consumption, agricultural use, and industrial processes will be one of the main challenges facing humanity in the coming decades. In many parts of the world, fresh water aquifers are being depleted, and in order to meet the increasing demand for fresh water, there will be a growing need to use poorer quality water. Currently, reverse osmosis (RO) is the state-of-the-art technology for water desalination, especially for seawater, but also for the treatment of brackish water.1−5 Separation of a feedwater stream into a diluted (desalinated) stream and a concentrated (waste or brine) stream always requires a certain amount of work or energy.6 In that respect, the energy consumption in a RO installation is relatively high as the water is taken out of the salt solution by pressing it through a semipermeable membrane instead of taking only the ions out from the solution.2,3 Another proven desalination technology is multistage flash evaporation, which requires even more energy, and therefore, it seems not to be a sustainable technical option for future energy-efficient desalination.1,4,5 Membrane capacitive deionization (MCDI) is a promising electrochemical desalination technique, which is suitable for the removal of ions from brackish water up to concentrations of about 100 mM (ca. 5000 ppm).7−11 In MCDI, ions are removed from feedwater by applying a cell voltage difference between two electrodes, in front of which ion-exchange membranes are placed, separated from each other by a spacer.12,13 Water can flow through the spacer compartment in between the membranes that are placed against the oppositely charged electrodes. Ions are removed from the feedwater and are temporarily stored in the electrical double layers (EDLs) that are formed at the electrode−water interface within the carbon electrode structure.14−16 The electrodes need © 2013 American Chemical Society
to be regenerated once they are saturated with ions, which can be done by reducing or even reversing the applied cell voltage. After the ions have been released from the electrodes, they can be flushed out of the spacer compartment, resulting in a waste stream concentrated in ions. After regeneration of the electrodes, the system is again ready to remove ions from the feedwater. An important advantage of MCDI compared to the more conventional capacitive deionization (CDI) technique is that, in MCDI, the co-ions that are expelled from the EDLs during the desalination step give rise to additional adsorption of salt inside the macropores of the carbon electrodes.7−13,17,18 This means that, in MCDI, the transport of, say, 100 electrons from the anode to the cathode during the desalination process results in the removal of, for instance, 95 NaCl molecules from the water, while in CDI, this number is significantly lower.11 In this context, it is useful to define charge efficiency as the ratio between the ionic charge removed from the water and the electrical charge, which for MCDI, is close to unity and for conventional CDI this number is below unity, depending on salt concentration and applied voltage.11 The minimum energy to separate a feedwater stream into diluted stream and concentrated streams can be quantified by using a thermodynamic approach.6 All desalination technologies will consume a greater amount of energy than the thermodynamic minimum because of additional energy losses, for example, due to the movement of ions and/or water. An important advantage of MCDI is that the energy needed to Received: Revised: Accepted: Published: 4904
December 30, 2012 March 1, 2013 March 11, 2013 March 11, 2013 dx.doi.org/10.1021/es3053202 | Environ. Sci. Technol. 2013, 47, 4904−4910
Environmental Science & Technology
Article
Figure 1. Scheme of the MCDI stack used in this study.
the MCDI stack is 1.18 m2. Both anion- and cation-exchange membranes have a resistance of ∼2 Ω·cm 2 and a permselectivity higher than 90%.18 The water flow through the MCDI stack was set at 1 L/min. The NaCl (analytical grade, BDH Prolabo, VWR International) used in this study was dissolved in 60 L of water. All of the stack components were compressed at 0.9 bar by 14 stainless steel bolts situated in the end plates. Electrochemical Characterization. MCDI is a cyclic desalination technique, where the energy used for charging the electrodes with ions (desalination step) can be recovered during the electrode discharge step (ion desorption). During the charging process, ions are adsorbed into the EDLs present at the carbon−water interface of the electrodes.14−16 The discharge step starts once the electrodes are saturated with ions. Discharging can be done in a controlled manner at constant current conditions via an external load, while the ions are moving from the electrode through the ion exchange membrane into the spacer compartment. At the same time, an equivalent number of electrons is transported over the external load. The external load (B&K Precision, model 8500, USA) absorbs the released (recovered) energy. Note that this load has a minimum resistance of 35 mΩ, which means that the constant current condition cannot be maintained at lower voltages (e.g., 0.42 V at 12 A). The experimental MCDI stack was operated under constant current conditions at currents in the range 2−15 A, where the electrodes were charged up to a total charge of 10 000 (±2000) A·s (i.e., 10 kC) by a custommade programmable power supply. Exact duration of the cycles and applied currents are presented in the Thermodynamic Efficiency section in Table 1. Prior to starting the experiment, the MCDI stack was equilibrated at room temperature and a flow of 1 L/min for 10 min with the short-circuit electrodes. Ohmic resistance presented in Figure 5a was calculated from the initial cell voltage increase immediately after the start of the desalination step (and dividing by current). This means that we used the first data point of the cell voltage after the desalination current was applied, see Figure 4. Measured resistance was multiplied by the total cell area of 0.59 m2. This initial cell voltage increase was measured twice for each current value, and the average from these two points is presented in Figure 5a. The initial cell voltage increase is graphically presented in Figures 3 and 4.
remove ions from water during desalination is to a certain extent stored in the EDLs that form at the carbon−water interface.14−16 This stored energy is released again during the electrode regeneration step, when the ions move back from the electrode into the spacer compartment. The amount of recoverable energy during the desalination step in the MCDI stack has up to now not been determined. In recent literature, Bijmans et al. and Sales et al. have shown that energy can be extracted using a Capmix technique in a capacitive setup by mixing seawater and fresh water.19−21 However, up to now, no papers have been published that provide experimental data on the amount of energy that can be recovered during electrode regeneration in MCDI. In addition, our study provides insight in how to reduce energy consumption in a MCDI system. In our paper, we demonstrate that, by choosing the right operational settings, the energy invested during desalination can be recovered to a large extent, which significantly reduces the overall energy consumption. Objective. The objective of this work is to identify energy recovery values under different constant current conditions and at two salt concentrations for MCDI. Furthermore, charge− discharge characteristics of the MCDI system are presented. Thermodynamic calculations are compared with energy consumption values measured under practical conditions, where ion removal is up to 91%. By comparing the data for energy consumption with thermodynamic calculations, we obtain a better insight into the thermodynamic efficiency for different ion removals. Thus, we can define optimized operational settings for the use of MCDI as an energy efficient desalination technique suitable for treatment of brackish water.
■
EXPERIMENTAL SECTION MCDI Stack. Figure 1 gives a schematic overview of the experimental MCDI stack that was used in the present study. The MCDI stack contains 24 repeating cells that are sandwiched between two end plates made from PVC. Each cell consist of a graphite current collector (thickness δ = 250 μm), porous carbon electrodes (Materials Methods, PACMM 203, Irvine, CA, USA, δe = 362 μm), anion- and cationexchange membranes (Neosepta AM-1 and Neosepta CM-1, Tokuyama Co., Japan, δ ≈ 130 μm), and a woven spacer (δ = 115 μm). The electrodes are chemically the same and can act as an anode and as a cathode. The resistance of the porous carbon electrode equals 1 (±0.2) Ω·cm2. The total electrode area in 4905
dx.doi.org/10.1021/es3053202 | Environ. Sci. Technol. 2013, 47, 4904−4910
Environmental Science & Technology
Article
The so-called “non-ohmic resistance” is the resistance that derives from the initial cell voltage increase until the voltage becomes linear with time. A visual representation of this calculation is presented in Figure 4a. Energy recovery values plotted in Figure 6 have been calculated from the ratio of the energy (joules) obtained during discharge of the electrodes and the amount of energy that was required during the desalination step (charging of the electrodes).
■
RESULTS AND DISCUSSION Thermodynamic Energy. Separating an inflowing water stream into diluted and concentrated streams leads to a decrease of entropy for the two product streams and therefore requires the input of a certain amount of work, the minimum of which can be calculated from the difference in Gibbs energy between inflow and outflow, see eq 1:6 ΔGdesalination = Gd + Gc − Gfeed
(1)
where, ΔGdesalination is the Gibbs energy (J) that is needed to separate an ionic solution into two product solutions, one enriched in ions and another depleted of ions, Gd is the Gibbs energy of the diluted solution (J), Gc is the Gibbs energy of the concentrated solution (J), and Gfeed is the Gibbs energy of the feed salt solution (J). Ion removal is defined as the ratio of the ion concentration in the dilute product water and the ion concentration in the feedwater. Figure 2 shows the thermodynamic energy needed to produce 1 m3 of product water (in kJ/m3) as a function of ion removal, inflow ion concentration, and water recovery. Figure 2 shows that, to reach higher ion removal at higher salt concentrations in the feedwater, more energy is needed to produce 1 m3 of purified water. For example, a 10 times increase in inflow ion concentration (from 8.6 to 86 mM NaCl) leads to a 10 times increase of required thermodynamic energy. Water recovery, which is the ratio of the volume of produced water versus the volume of feedwater, is another important parameter for all desalination techniques. Figure 2b shows that a high water recovery of 90% requires almost double the energy to produce 1 m3 of purified water compared to the situation when the water recovery is 50%. This thermodynamic energy that is needed to separate the feedwater stream into diluted and concentrated streams cannot be recovered. Therefore, this energy can be considered as the minimal energy that is required to desalinate a water stream. However, any desalination technique will in practice require more energy as there are always additional energy loss terms, for example, due to ion transport and hydrodynamic flow. Charge−Discharge Characteristics of the MCDI Stack. The energy that is required to charge the electrodes during desalination and the energy that can be recovered from the electrodes during regeneration are determined by the charging and discharging conditions of the MCDI stack. In this study, we applied constant current conditions both for the charging and for the discharging of the electrodes, and the cell voltage was measured, being the voltage between the two electrodes. In Figure 3, the MCDI stack was charged at two different currents, 2 and 8 A, although in both cases, the final, total, charge was the same (thus the duration of charging was 4 times shorter in the second case). Figure 3 shows the charge−discharge behavior of the MCDI stack when operated at these two constant current conditions in a low salt concentration (8.6 mM NaCl solution).
Figure 2. Required thermodynamic energy needed to produce 1 m3 of product water at different ion removal and at different ion concentrations in the feedwater (a) and different water recoveries (b).
Figure 3 shows an initial jump in the cell voltage when charging the electrodes, which is caused by internal electrical resistances at different locations in the MCDI stack. For example, these resistances can occur in the graphite current collectors, electrodes, membranes, and spacer compartments. The initial cell voltage jump is higher when the MCDI stack is charged at the higher current of 8 A. We expect the cell voltage jump to be linear with the current, which would imply that the initial resistance is ohmic. Indeed, the measured initial cell voltage increase at the higher current (8 A) is 4 times larger compared to a current of 2 A. Obviously, the higher the initial cell voltage jump, the more energy will be required to charge the electrodes during desalination, and therefore, charging the electrodes at 8 A is less energy efficient than charging them at 2 A. The integral of power (current multiplied by voltage) over time represents the energy used for charging of the electrodes. Thus, the surface area under the charging curve must be proportional to the total energy invested during the desalination process. Part of this energy is stored in the EDLs at the electrodes and is potentially recoverable, which is indicated by the surface area under the dashed line in Figure 3. On the other hand, part of the applied energy is needed to overcome the internal resistances in the MCDI stack, indicated by the surface area between the charging curve and the dashed line in Figure 3, and this energy cannot be recovered. In the discharge step, the surface area below the discharging curve represents the amount of recoverable energy, and the surface area between the dashed line and the discharging curve represents energy losses to overcome the internal resistances 4906
dx.doi.org/10.1021/es3053202 | Environ. Sci. Technol. 2013, 47, 4904−4910
Environmental Science & Technology
Article
maintained at lower voltages. This feature does not affect our conclusions. In the MCDI stack, not only ohmic resistance can play a role, but also, additional non-ohmic resistance may appear during operation. This additional non-ohmic resistance is likely related to the formation of diffusion boundary layers, which can develop at the membrane−spacer interface as well as at the membrane−electrode interface.24,25 In order to elucidate the non-ohmic resistance further, the MCDI stack was charged and discharged at 12 A, where the contribution of the non-ohmic resistance to the total resistance is expected to be more pronounced.25 In Figure 4a, it can be seen that, at a charging current of 12 A, not only the initial cell voltage jump is higher, but also, there
Figure 3. Charge−discharge curve of the MCDI stack for 2 A charge− discharge current (a) and 8 A charge−discharge current (b).
during the discharge process. Figure 3 shows that, by charging the electrodes at a lower current, not only less energy is required during the charging of the electrodes, but also, more energy can potentially be recovered during the discharge process. The charging profile of the electrodes does not follow a straight line but is rather curved (Figure 3), which is most likely related to a redistribution of the (ionic) charges in the porous electrodes. Initially, the ions are only filling up the larger pores and/or only the pore openings. Nevertheless, during a prolonged charging cycle, the ions are redistributed inside the electrode and also start to fill up the micro- or even nanopores in the carbon particles. This redistribution of ions leads to a lowering of the cell voltage.22,28 In addition, at higher voltages (>1.5 V), also, additional energy losses may occur caused by faradaic reactions at the electrode and the graphite current collector surface.33 We measured the decrease of cell voltage in an open circuit, where the MCDI stack was charged at 6 A for 1000 s up to a cell voltage of 1.53 V, after which the power supply was disconnected and the electrodes were left in open circuit. During a period of 200 s, the cell voltage dropped to 1.47 V, which implies a loss of stored energy. Obviously, during longer charging cycles, more energy losses will occur due to ion distribution. At the end of the discharge curve (Figure 3b), a “tail” effect can be observed. This is caused by a technical feature of the electronic load, which has a minimum internal resistance of 35 mΩ, and therefore, constant current can no longer be
Figure 4. Charge−discharge curve of the MCDI stack under constant current of 12 A at 8.6 mM salt concentration (a) and at 273 mM salt concentration (b).
is a significant cell voltage increase due to non-ohmic resistance, compared to charging at 8 A. Figure 4b shows the charge−discharge profile of the MCDI stack operated at a high current of 12 A and at a high salt concentration of 273 mM NaCl. In this case, ohmic resistance measured at the beginning of the charging cycle is significantly reduced compared to that at the lower salt concentration of 8.6 mM NaCl (Figure 4a). At the higher salt concentration, the flow channel resistance is approximately 30 times lower, which has a direct impact on reduction of the overall system resistance. Obviously, the lower overall system resistance improves the charging efficiency of the electrodes due to lower energy losses over the resistive part of the system. This implies that, at higher salt concentration, the recoverable energy should also be higher. 4907
dx.doi.org/10.1021/es3053202 | Environ. Sci. Technol. 2013, 47, 4904−4910
Environmental Science & Technology
Article
concentration is low. The non-ohmic resistance at 4 A equals approximately 50 Ω·cm2, and by increasing the current to 12 A, the non-ohmic resistance doubles, which shows the large contribution of the non-ohmic resistance to the overall MCDI stack resistance. The difference in ion transport between the bulk solution and the membrane results in the formation of diffusion boundary layers at the membrane−solution interface as well as at the membrane−electrode interface. It has been shown in a previous paper that these diffusion boundary layers lead to an increase of the resistance, especially at low salt concentrations, which can be up to 1 order of magnitude higher than the membrane resistance.23−25 On the other hand, at high salt concentrations, the non-ohmic resistance becomes negligible.24 In order to maximize the energy recovery in MCDI, it is important to minimize both the ohmic and nonohmic energy losses during charging and discharging of the electrodes. Energy Recovery. Energy recovery is defined as the ratio of the energy recovered during ion release and the invested energy during the charging of the electrodes. It is important to note that a certain amount of energy cannot be recovered, because of the thermodynamic work that is needed to divide a feedwater stream into diluted and concentrated streams. Figure 6a shows how much energy can be recovered from the electrodes when they are discharged at the same constant current level as at which they are charged, whereas in Figure 6b, the electrodes were charged at 2 A and discharged at currents in the range 2−15 A.
To further understand the potential for energy recovery at the different operating conditions, it is crucial to get to a quantitative description of both ohmic and non-ohmic resistance. Figure 5 shows the ohmic and non-ohmic resistance contributions to the overall resistance as a function of applied current at two salt concentrations.
Figure 5. Ohmic (a) and non-ohmic (b) resistance as a function of applied current.
Figure 5a shows that the ohmic resistance at the low salt concentration (8.6 mM NaCl) is about a factor of 3 larger than the value measured at the higher salt concentration (273 mM NaCl). The absolute difference between both values is 140 Ω·cm2, whereas the calculated resistance difference between these two solutions taking into account the distance between the two membranes (115 μm) equals 11.1 Ω·cm2. The theoretical value for the resistance based on the ion concentration in the spacer compartment is about 1 order of magnitude lower than the value calculated from the voltage profiles. Therefore, this difference cannot be attributed to only the ionic resistance in the flow channel. There are clearly additional resistances that may come from the resistances of the electrode, the membrane as well as contact resistances at the membrane−electrode interface, the electrode−graphite interface, and the connectors to the graphite current collectors. In addition to this, the so-called spacer shadow effect, which occurs because the spacer material is ionically nonconductive thus reducing the effective area available for ionic transport, may also play a role.25,26 Currently, we have not been able to quantify all the different contributions to the overall resistance, which will be part of further research. Figure 5b shows a nonlinear increase of the non-ohmic resistance as a function of the applied current when the salt
Figure 6. Recovered energy for the MCDI stack charged and discharged with the same magnitude of the current (a) and when the electrodes are charged at 2 A and discharged at different currents (b). 4908
dx.doi.org/10.1021/es3053202 | Environ. Sci. Technol. 2013, 47, 4904−4910
Environmental Science & Technology
Article
Table 1. Measured and Calculated Energy Consumption and Thermodynamic Efficiency at Different Charge−Discharge Currents at 8.6 mM Salt Concentration charge−discharge
thermodynamic energy
current (A)
time (s)
ion removal (%)
energy recovery (%)
(kJ/mol)
2 4 6 8 10 12
4800 2400 1350 1500 1050 900
17.9 32.1 50.7 61.7 76.1 90.6
79.1 73.2 50.1 49.0 37.8 26.8
0.36 0.65 1.06 1.33 1.72 2.22
experimental energy consumption
thermodynamic efficiency
(kJ/m3 of prod)
no recovery (kJ/m3 of prod)
recovery (kJ/m3 of prod)
no recovery (%)
recovery (%)
0.6 1.8 4.6 7.0 11.3 17.3
114 258 347 594 674 974
23.9 69.1 173 303 420 712
0.49 0.70 1.33 1.19 1.67 1.78
2.32 2.61 2.67 2.33 2.68 2.43
Table 1 shows the applied charge−discharge current, time of the charge−discharge cycle, ion removal, energy recovery, thermodynamic energy, experimental energy consumption and thermodynamic efficiency with and without energy recovery. In general, Table 1 shows values of thermodynamic efficiency, which from an engineering point of view are very low (
