Concentration of crotonic acid using capacitive deionization technology

Concentration of crotonic acid using capacitive deionization technology

Separation and Purification Technology 209 (2019) 658–665 Contents lists available at ScienceDirect Separation and Purification Technology journal ho...

1MB Sizes 0 Downloads 22 Views

Separation and Purification Technology 209 (2019) 658–665

Contents lists available at ScienceDirect

Separation and Purification Technology journal homepage:

Concentration of crotonic acid using capacitive deionization technology a




Ellen Hack , Dominik Hümmer , Matthias Franzreb a b

Karlsruher Institute of Technology (KIT), Institute of Functional Interfaces, Hermann v. Helmholtz, Platz 1, 76344 Eggenstein-Leopoldshafen, Germany ETH Zürich, Bioanalytics Group, Mattenstr. 26, 4058 Basel, Switzerland


Capacitive deionization (CDI) was developed for desalination of brackish water. In contrast to ion exchange, CDI is based on the accumulation of ions within the electrical double layer formed at the surface of conductive materials when an electrical potential is applied. Ion adsorption and desorption can therefore be controlled by switching between a potential of approx. 1.2 V and zero. In this work we demonstrate that the principle of CDI is also suitable for concentrating dilute solutions of crotonic acid resulting from processes such as biotechnological production via fermentation. For this we built a CDI cell using rapid prototyping and fabricated capacitive electrodes based on activated carbon. Applying voltages of 1.2 V and adjusting the feed solution to pH = 7 caused the electrodes to reach a salt adsorption capacity of approx. 0.1 mmol crotonic acid (in the form of sodium crotonate) per g of electrode material. If an aqueous feed solution of pH = 5, which is in the range of the pKs of crotonic acid is used, the electrodes only captured around 0.02 mmol∙g−1. This shows that pH dependent dissociation of weak electrolytes has a strong impact onto potential applications of CDI, and must be taken into account when designing a process combining production and separation steps. Despite the relatively low adsorption capacities, a process is suggested which accumulates sodium crotonate in a CDI system followed by a release of the accumulated ions with a concentration factor up to two. In the discussion we present suggestions how to improve these numbers and a first estimate of the energy consumption of such a system, which can be assumed to be around 1 kWh for a feed of 1 m3·h−1 and 20 mM crotonic acid. In the case of dilute ionic feed streams, concentration using capacitive deionization could be an interesting alternative for biotechnological applications.

1. Introduction Against a background of ecological awareness, applications of socalled white biotechnology are expected to continuously grow in their industrial importance [1]. A well-known example of white biotechnology is the production of small organic acids, like fumaric acid or crotonic acid by fermentation [2]. These organic acids are interesting building blocks for more complex commodity chemicals; however, until today biotechnological production in many cases is not economically competitive. Besides precipitation followed by solid-liquid separation, ion exchange chromatography is a frequently used technique for concentration and purification of dilute bioproducts which are present in an ionic form [3]. However, conventional ion exchange chromatography requires large amounts of buffer solutions for elution and for regeneration of the stationary phase. In contrast, CDI is a separation technology in which the retention of ionic substances can be altered by an electrical potential applied to the stationary phase without the need to change the chemical composition of the fluid phase [4]. A CDI system is basically an electrochemical cell which contains a working electrode (WE), a counter electrode (CE) and in case of three-electrode systems, also a reference electrode (RE) [5]. With the help of a potentiostat or a conventional power supply, a cell potential is applied between the WE and CE electrodes. According to the resulting surface loading, ions from

the electrolyte solution flowing through the CDI cell adsorb into the electrochemical double layer of the electrodes. Highly efficient CDI electrodes are available in form of flat sheets, resulting in a stack design of CDI cells with alternating working and counter electrodes separated by a porous, nonconductive spacer [6]. Besides the design with the electrolyte passing a narrow gap between the electrodes there exist also CDI systems with flow through electrodes [7] or slurry electrodes [8,9]. For efficient operation, CDI systems require conductive stationary phases, which are mainly based on commercially available carbon modifications such as activated carbon (AC) or porous graphitic carbon (PGC) [10]. However, in recent years also more sophisticated carbon structures with tailored surface modifications have been tested for CDI [11–13]. Besides desalination CDI becomes increasingly popular for the removal of other inorganic contaminants like nitrate, phosphate or heavy metals [14–16]. Another type of tunable electrosorption systems, so-called potential controlled chromatography (PCC), has also been tested for analytical separations of organic compounds, ranging from short chain carboxylic acids [17,18], aromatic sulfonates [19,20], monosubstituted benzenes [21], up to benzodiazepines [22], and corticosteroids [23]. While this list shows the potential of electrosorption for analytical separations of charged organic compounds, the authors are not aware of publications, which report the use of such systems for preparative or industrial separations. The main reason for this

Corresponding author. E-mail addresses: [email protected] (E. Hack), [email protected] (D. Hümmer), [email protected] (M. Franzreb). Received 16 July 2018; Received in revised form 5 August 2018; Accepted 26 August 2018 Available online 03 September 2018 1383-5866/ © 2018 Elsevier B.V. All rights reserved.

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

thickness, Panasonic, Japan) and dried for 30 min at 50 °C. The final electrode mass was calculated from the weight of the cut electrodes minus the weight of the graphite foil and deduction of the PVDF amount. The effective electrode surface area and pore size distribution were determined via gas adsorption according to the BET-model [29] with argon at 87°K in an Autosorb-1MP (Quantachrome GmbH & Co. KG, USA). Scanning Electron Microscopy images were taken with a Philips XL 30 FEG (Philips, Germany) under high vacuum mode and at a beam strength of 15 kV. The electrical capacity of the electrode material was measured via cyclic voltammetry (CV) by cycling between −5 and +5 mV with a feed rate of 0.01 mV·s−1 and 100 mM NaClO4 as electrolyte solution. Capacity was calculated according to Hamann and Vielstich [30].

limitation are the small achievable binding capacities of the used PCC systems. However, CDI systems using highly porous carbon materials are known to show higher salt adsorption capacities, making it worthwhile to study the separation and concentration of charged organic compounds by CDI. Against this background, we demonstrate the separation and concentration of the commercially relevant organic substance crotonic acid by the application of CDI technology. The influence of important parameters, such as applied voltage, feed concentration and pH, is investigated, and based on the results a first estimation of the expected performance and costs is given. 2. Experimental 2.1. Chemicals and reagents

2.3. Experimental procedure The following chemicals were used for the experiments: crotonic acid 98% (CA) (Sigma Aldrich, USA), 32% NaOH (VWR Analyticals, Germany), purified water of Milli-Q® grade (Millipore, Billerica, USA). The capacitive electrodes were prepared from activated carbon Norit SX ultra (Norit, Japan), polyvinylidenfluoride (PVDF) Solef® 5130 (Solvay, Belgium), carbon black C-NERGY™ Super C65 (Timcal, Switzerland), and N-methyl-2-pyrrolidone (NMP) (Sigma Aldrich, USA)

For the experiments on concentrating CA from dilute solutions CA (pKa = 4.69 at 25 °C [31]) was diluted with purified water to the target concentration and pH was adjusted by adding NaOH. The CDI cell was connected to a fast protein liquid chromatography (FPLC) system ÄKTApurifier10, (GE Healthcare, Sweden). This system allows accurate control of the flow rates as well as online measurement of pH and conductivity. Knowing the equivalent conductivities of all involved species as well as the dissociation constants of crotonic acid and water, these two parameter allow an unambiguous calculation of the concentration of all ions in the effluent (details of this calculation are given in the Supporting Information). For application of an electrical potential the CDI cell was connected in two-electrode mode to a potentiostat Reference 600 (GAMRY Instruments, Germany). During the experiments a rectangular periodic potential, switching between a fixed voltage and zero voltage was applied to the CDI cell. The adsorption/ desorption interval was repeated for each set of parameters at least three times in order to reach stationary conditions. An initial conditioning period of 120 mL at 0 V was applied in order to reach constant conductivity and pH signals before the actual experiment starts. The flow rate was adjusted to a constant value of 1 mL·min−1 during all adsorption as well as desorption intervals of 60 min duration. For voltage dependent experiments, voltages between +400 and +1400 mV were applied for the adsorption and 0 mV for the desorption steps; CA solution was kept at 5 mM, pH = 7.5. In the case of concentration dependent experiments, a fourfold repeated adsorption/desorption cycle was executed with a constant adsorption voltage of +1200 mV and desorption voltage of 0 mV. Concentrations were varied from 2 to 50 mM CA, each adjusted to pH = 7 by the addition of caustic. As stated, the effluent concentration of crotonic acid was calculated from the effluent values of conductivity and pH. The vast majority of CDI experiments reported for desalination use dissolved NaCl in the feed solution. In these cases there exists a simple relationship between the measured conductivity and the concentration of sodium and chloride

2.2. Design of the CDI cell, electrode preparation and specification Our cell design orientates on CDI cell designs reported for laboratory scale desalination experiments using NaCl as target like basically suggested by Farmer et al. [24,25] and afterwards frequently used by e.g. Suss et al. or Zhao et al. [26,27]. The cell chamber as well as the ladder like spacer for isolation of the electrodes were produced by rapid prototyping, using the acrylate based 3D printing materials VeroClear and VeroWhitePlus, printed with an Objet Eden260V (Stratasys, USA). The seal was hand cut from a sheet of soft polyvinylchloride (PVC-P, trade name Mipolam). Fig. 1 displays a sketch and a photo of the 3D printed CDI cell. The liquid phase enters through the top lid and flows through a narrow gap between the WE and CE to a transversely arranged outlet in the bottom lid. The carbon based capacitive electrodes are separated by a ladder-like spacer and contacted via titanium clamps to the potentiostat. As both electrodes are made of the same material and have the same capacitances, the cell is symmetric and the polarity of the connection to the potentiostat plays no role. The total void volume of the CDI chamber was determined to be 1.15 mL. The capacitive electrodes were prepared following the method of Porada et al. [28] with 85% activated carbon Norit SX ultra, 10% polyvinylidene fluoride (PVDF) and 5% carbon black. From these components, a slurry was prepared by carefully mixing them with the solvent N-methyl-2-pyrrolidone (NMP) up to a solid content of 22.5% (calculated without PVDF). The slurry was raked with a layer thickness of 600 µm onto a graphite foil (PGS graphite sheet EYG type, 100 µm

Fig. 1. Exploded 3D sketch and photo of the 3D printed CDI cell, housing dimensions (l × w × h = 180 × 50 × 40 mm). 659

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

activated carbon particles in the range of 20 µm (see Fig. 2a). Micropores with pore sizes below 2 nm represent 87% of the cumulative pore volume, whereas mesopores (pore sizes between 2 and 50 nm according to IUPAC [32]) account for only 13%. Mesopores over 36 nm are rare. By means of the mass specific capacity determined by CV experiments and the measured BET surface a surface specific capacity of the electrodes of 8 µF·cm−2 was calculated (for details see SI). The specific BET surface area was be expected to be higher according to the manufactureŕs data of 1200 m2·g−1 for activated carbon and 62 m2·g−1 for carbon black. A possible explanation could be that the PVDF binder used for electrode preparation may have blocked some of the pores rendering them inaccessible and thus reducing the specific surface area in the BET measurement, like it is described by Kinoshita [33]. A specific capacity of 8 µF·cm−2 is within the normal capacitance range of carbon based electrodes which is published by Ji et al. between 4 and 5 µF·cm−2 for highly porous electrodes with specific surfaces above 1.500 m2·g−1 [34] and between 5 and 20 µF·cm−2 by Kinoshita in his overview over carbon materials [33].

ions, given by the molar conductivity which can be calculated in good approximation by Kohlrausch’s law [30]. However, this relationship only holds for strong electrolytes. In the case of weak organic acids the relationship is more complicated because, depending on the pH, the organic acid might be only partly dissociated. Nevertheless, if conductivity and pH are known, the ion concentrations in the effluent of our experiments can be readily calculated (see supporting material for details). Another prerequisite for the calculation is the knowledge of the molar conductivities of the ions at infinite dilution. While this is known for common ions like H+, OH− and Na+, etc. the value for crotonate had to be determined. Hence, we measured this value according to Hamann-Vielstich [30]. The pH-dependent conductivity curves with different crotonate concentrations were recorded and interpolated. With these data, the conductivity at infinite dilution of crotonate could be calculated to be 40 mS·L·cm−1·mol−1 (see supporting material for details). For our analysis of the voltage dependent adsorption of CA, we evaluated mainly two parameters: (i) the experimental capacitive loading of the electrode expressed in mmol bound salt (in our case sodium crotonate) per g of electrode material. As explained in detail in the supporting information, the course of species in the effluent can be derived from the courses of conductivity and pH. Afterwards, the integration of the course of the effluent concentrations during the desorption step gives the number of bound ions. Divided by the electrode mass, the capacitive loading of ions per gram electrode can be calculated. (ii) The second parameter, the theoretical loading, is calculated by the course of the measured current and the assumption of 100% charge efficiency. Integrating the current occurring during the desorption step and assuming that the entire resulting electrical charge (qelec) is generated by desorbing ions, the measured qelec can be converted into a theoretical loading of ions with the use of Faraday’s constant (F) (Eq. (1)).

qCA, th =

3.2. Capacitive adsorption of crotonic acid Fig. 3 shows a typical result of a multi-cycle CDI experiment for capacitive adsorption of crotonic acid species. A constant voltage operation scheme is used, alternating between a period when a voltage of 1.2 V is applied and zero voltage periods. When the voltage is applied anions of the dissociated organic acid are adsorbed at the working electrode while sodium ions of the sodium hydroxide solution used for pH adjustment are adsorbed at the counter electrode, leading to a sharp decrease of the effluent concentrations of the respective ions. In consequence, exactly speaking it is not neutral crotonic acid to be adsorbed but the salt sodium crotonate. Over time, the electrode capacities start to exhaust and the concentrations of the ions slowly increase although a constant voltage is applied. When the voltage is switched back to zero, the adsorbed ions desorb and the ion concentration in the effluent of the cell shows a sharp increase followed by an exponential decline. The left plot of Fig. 3 shows the observed trends of conductivity and pH for repeated voltage dependent ad- and desorption cycles. The application of a voltage of 1.2 V results in a fast reduction of the conductivity in the effluent of about 35%. Simultaneously the pH value in the effluent droped from 6.8 to 6. At the beginning of each adsorption cycle, the conductivity showed a small sharp increase before it starts to drop. This phenomenon has also been observed in conventional CDI cells for desalination and is most probably caused by functional groups of the activated carbon material with ion exchange functionality [35]. Functional groups of this sort can be produced by partial oxidation if carbon electrodes are stressed by higher voltages and used for longer times [36–38]. When the potential is switched on, ions fixed by these

qelec F


For both parameters the values from the desorption step were taken, because during the adsorption step several effects such redox reactions may occur, leading to currents that are in no coherence with the ion adsorption at the electrode (see also [5]). 3. Results and discussion 3.1. Electrode specification

Differential pore volume (cm3 g-1 Å)

The carbon content of the overall electrode mass (counter- and working electrode) used in the experiments was calculated to 932 mg. The produced electrodes are brittle and have to be handled with care. The specific BET surface is 302 m2·g−1 with an average pore width of 1.4 nm. The electrodes show a rough surface with microporous


0,06 0,05 0,04 0,03 0,02 0,01 0,00 0






Pore diameter (Å) Fig. 2. Activated carbon electrode material; (a) ESEM picture of the activated carbon based capacitive electrode; (b) pore size distribution determined by BET measurement. 660

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

Fig. 3. Four times repeated ad- and desorption cycles of CA in the CDI cell. Feed: 5 mM CA adjusted to pH = 7 using NaOH; voltage: 1.2 V; interval time: 60 min; flow rate: 1 mL·min−1; Left: black line: Conductivity in mS∙cm−1; blue line: pH-value; Right: Calculated sum concentration of all crotonic acid species, grey bars indicate the periods with constant voltage of 1.2 V. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

could be, that at 1400 mV we start to see first signs of water electrolysis or other unwanted redox reactions. The experimental SAC approach 50–80% of the theoretical SAC predicted from the accumulated electric charge. Therefore, it can be concluded that a high percentage of the current delivered by the potentiostat is used for charging the capacitance of the electrode and that the respective surface charge of the carbon electrode is compensated by adsorbing crotonate and sodium ions in the capacitive double layer. The second y-axis of Fig. 4b shows the fraction of CA bound with respect to the total amount of CA available during the adsorption cycle. In the best case (1400 mV) 20% of the CA flowing through the CDI cell over a period of 60 min was captured. The relatively low number results from the fact, that the CDI cycles were run until almost complete equilibrium in order to measure maximum SAC values. However, as can be seen from Fig. 3 at the end of the cycle the salt adsorption efficiency drops and most of the ions of the feed solution pass the electrodes without being adsorbed. Nevertheless, shorting cycle times and optimizing CDI design by applying larger electrode surfaces the fraction of CA bound could be substantially increased to more than 50% or higher.

groups will be suddenly released and will result in a brief increase of conductivity. For the calculation of the cumulative loading during adsorption, this release counts as negative, thus reducing the result of the achievable loading. The pH decrease from 6.8 to 6 during the adsorption step seems in accordance with the expectations. In the feed at pH = 6.8 most of CA is in the form of crotonate, however a small percentage (1%) of undissociated acid is left. When crotonate is removed from the solution by capacitive deionization, the dissociation equilibrium is disturbed and some of the CA has to dissociate in order to readjust the equilibrium. However, on closer inspection, it becomes clear that the system does not follow this simple picture. The total concentration of CA in the effluent is reduced only about 35%, therefore the effluent concentration is approx. 3.3 mM. According to the pKs, at pH = 6 around 5% (0.16 mM) of it is in the form of undissociated CA. This is three times more than the original 0.05 mM (1% of 5 mM) in the feed solution at pH = 6.8. In order to achieve this increase in the concentration of undissociated CA the CDI cell must generate protons when the potential is applied. One possible explanation is the accumulation of hydroxyl ions in the electrical double layer of the anode (WE) followed by the dissociation of water, releasing protons. This shows that the process of capacitive capture of weak organic acids is more complex than the original assumption of the potential driven adsorption of only crotonate and sodium ions at the electrodes.

3.4. Concentration dependent adsorption of CA Cyclic CDI experiments at 1200 mV and a flow rate of 1 mL∙min−1 were repeated with crotonate concentrations up to 50 mM at pH = 7 and pH = 5 (Fig. 5). Looking at Fig. 5 it shows that in contrast to classical Langmuir adsorption isotherms the achievable capacitive loading decreases in case of an increase of the concentration of CA in the feed. In the case of pH = 7 this trend is clearly visible and starts at concentrations as low as 10 mM. While this effect in principle is known for CDI, the extent of the decrease at only moderate concentrations is remarkable. In CDI theory it is normally assumed that if no potential is applied the ion concentrations in the macro- and micropores of the electrode are equal to the respective values in the bulk solution. When a potential is applied, ions having the same charge than the electrode material, so-called co-ions, will be repelled from the pores, thus reducing the deionization effect and the charge efficiency (for more detail see e.g. Andelmann et al. [39] or Biesheuvel et al. [40]). However, at CDI cell potentials of 1.2 V the ion concentration in the pores can be estimated to be up to 500 mM or higher [41]. Therefore, even at bulk concentrations of 50 mM the repelling of co-ions from the pore should not have a major influence. Nevertheless, our results suggest that in case of organic molecules there are additional attractive forces between them and the carbon material of the electrode resulting in higher initial loadings and therefore a stronger detraction of the capacitive deionization. This assumption is supported by the circumstance that activated

3.3. Voltage dependent adsorption of CA Fig. 4 shows the theoretical specific adsorption capacity (SAC) as well as the experimental SAC of CA (in the form of sodium crotonate) in dependence of the applied voltage during adsorption. In case of an ideal capacitor, the accumulated electrical charge and therefore the theoretical adsorption capacity should be linear with respect to the applied voltage. As can be seen in Fig. 4a, this prediction holds for the theoretical loading calculated by integration of the current during desorption. Applying Eq. (1), a theoretical SAC of CA of around 0.1 mmol·g−1 can be calculated at an applied voltage of 1200 mV. The achieved experimental SAC (see Fig. 4b) was determined by the desorbed amount, which itself results from integrating the difference between in- and outflow concentration of CA during the desorption interval. Although in this case an almost linear dependency between capacitive loading and applied voltage holds true between a voltage of 400–1200 mV, however the curve is shifted by about 0.1 mmoll·g−1 towards lower loadings. For voltages higher than 1200 mV the increase of the capacitive loading starts to flatten. This might display a saturation of small pores where even with higher voltage supply no more ions can be adsorbed due to simple sterical hindrance. Another explanation 661

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

Fig. 4. (a) Theoretical SAC calculated from the measured electrical charge during the desorption steps; 5 mM CA solution; pH = 7.5; interval time: 60 min; flow rate: 1 mL·min−1; (b) Influence of the applied voltage onto the experimental SAC and bound fraction of a 5 mM CA solution (pH 7.5) applied for 60 min at a flow rate of 1 mL·min−1. The experimental SAC is calculated via the effluent concentration during the desorption steps.

displays the influence of the adjusted pH value within the CA feed. At pH = 5 only about 67% of the CA is dissociated, or in other words, one third is in the form of uncharged CA and therefore not directly susceptible to capacitive deionization. Nevertheless, at least beyond feed concentrations of 10 mM the concentration of crotonate ions is more than 6 mM and therefore should be enough for efficient capture. However, the maximum SAC reached at pH = 5 is 0.015 mmol∙g−1 at 20 mM CA and the trend for higher concentrations is again towards decreasing loading. We assume that higher concentrations of undissociated CA lead to an even stronger adsorption at zero potential, thus reducing the efficiency of capacitive deionization when a potential is applied. Another observation is that in case of a feed of 2 mM CA the pH-value drops from 5 to 4.2 (data not shown) during the adsorption step. At this pH only about 25% of the CA remains in the form of crotonate. This raises again the questions about the source of the protons required to shift the pH and if the pH within the pores of the working electrode may be even lower, reducing the crotonate concentration further. While a deeper investigation of the processes within the microand mesopores is not within the scope of this work, it can be said, that capacitive deionization of organic acids only seems suitable at neutral or slightly alkaline pH-values.

Fig. 5. Influence of the feed concentration onto SAC of CA at pH = 7 (filled squares) and pH = 5 (empty squares); 1200 mV; interval time: 60 min; flow rate: 1 mL·min−1; the star is a point from the adsorption cycle of the pH = 7 experiment.

carbon is a well-known adsorber for organic molecules and widely used in industrial applications for the purification of exhaustions and pollutants [42–44]. In consequence, even rather low CA concentrations in the bulk can result in high concentrations in the pores at zero potential. If we assume in addition that this initial pore concentration increases with the concentration in the bulk, quickly a point will be reached at steric hindrance and/or an increased co-ion repulsion will strongly diminish the efficiency of capacitive deionization. At this point it must be mentioned that the problem of co-ion repulsion in CDI can be solved by covering the electrodes with thin ion exchange membranes which prevent the co-ions from entering into the bulk solution. The trapped co-ions even enhance the electrode performance as they increase the overall attracting ion concentration behind the membrane and more ions from the bulk can be sorbed through the membrane until an ion equilibrium is reached. The idea of a charge barrier was published first in the patent of Andelmann [39] and further examined by several research groups [45–48]. Another approach to increase the charge efficiency of CDI cells is to functionalize the electrode material with ionic groups. Andelmann realized this with surfactants that irreversibly adsorb to the carbon surfaces of the electrodes. With this modification an increase of the charge efficiency from 45% (unmodified control cell) to 70% was reached [49]. In addition to the concentration influence, Fig. 5 also

3.5. Outline of a concentration process of CA using CDI In a final experiment, all process parameters were changed in a direction intended to improve the degree of CA uptake and concentration. For this, the flow rate was decreased to 0.1 mL/min, the pH was increased to 8.5, and the applied voltage was adjusted to 1600 mV. In addition, the cycle time was increased to 6 h due to the reduced flow rate. Fig. 6a shows the resulting CA effluent concentration over time. As intended, in comparison to Fig. 3 the degree of CA capture was strongly increased, reaching up to 50% between 30 and 60 min after the potential was applied. What is even more important, the concentration factor during the desorption step could be more than doubled, showing a sharp peak followed by some tailing shortly after the applied voltage was reduced to zero. Nevertheless, even with these optimized parameters the used CDI cell was not able to fully capture the CA in the feed. In order to prevent product losses if CA is separated from a fermentation solution we therefore suggest to apply a potential CDI system within a bypass as shown in Fig. 7. This means, solution is continuously withdrawn from the fermentation tank, pumped through the CDI system, and the CA depleted solution is transferred back to the fermentation tank. The suggested scheme shows some analogy to concentration processes using tangential flow filtration (TFF), which are 662

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

Fig. 6. (a) Effluent concentration of the CDI cell optimized for increased capture efficiency. The dashed area shows exemplarily the fractionation of a part of the effluent during the desorption step into a concentrate flow. 5 mM CA; pH = 8.5; 1600 mV; interval time: 360 min; flow rate: 0.1 mL·min−1; (b) CA concentration factor in dependence of the used bypass yield.

also operated in a mode that recirculates most of the feed solution entering the TFF module. Also in the case of CDI, the partially depleted solution will be recirculated back into the feed storage during the adsorption step. During the desorption step a part of the solution passing the CDI cell will be fractionated into a concentrate stream. If required, the CA feed could be stopped and the CDI cell flushed with a buffer shortly before and during this fractionation. By this the concentrated organic acid would be obtained in a buffer of choice. Based on the result of Fig. 6a the achievable concentration factors and the corresponding bypass yields were calculated (Fig. 6b). The bypass yield is defined as the fraction of CA that will be transferred into the concentrate stream in relation to the amount of CA pumped through the CDI cell (see supporting material). For the conditions chosen, in order to achieve a CF of 1.8 only around 16% of CA pumped through the CDI cell will end up in the concentrate. However, ratios of 1:6 or even lower between permeate and feed rate are not unusual in TFF [50] and it can be expected that the pressure drop in both systems is in a comparable range. What remains is the question about the energy demand of a hypothetical CDI concentration process. Assuming an energy efficiency of capacitive deionization of 50% a simple calculation shows that the up-concentration of a 1 m3·h−1

feed of CA with 20 mM would require a power of approx. 1 kWh (see supporting materials for details). Because of the flow-by operation the pressure drop within such a CDI system operated in bypass mode would be only moderate and can be estimated to be less than 1 bar. Therefore, in case of feed volumes common in biotechnology the energy consumption for concentration of dilute ionic targets by CDI is rather low and no hindrance to the application of this technique. 4. Conclusions A laboratory scale flow-by CDI cell with microporous carbon electrodes was developed and successfully employed to capture CA via potential control. In contrast to classical ion exchange, CDI switches between adsorption and desorption steps by the simple application of small voltages during adsorption and does not require any special elution or regeneration buffers. Experimental series with different voltages, CA concentrations and pH-values showed the proof-of-concept but also the limitations of this physicochemical separation process. Despite its simple basic principle and its proven efficiency for brackish water desalination, the pH dependence of weak electrolytes such as organic acids results in unexpected effects and challenges not

Fig. 7. Simplified process scheme of the concentration of organic acids from diluted feed stocks. 663

Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

encountered in classical CDI. Nevertheless, it could be shown that even with the simple inhouse fabricated cell and electrodes, concentration factors up to two can be reached and that a functional process could be outlined if the CDI cell is operated in bypass. Future work should concentrate on the development of materials and designs suitable for the concentration of biomolecules. Possible approaches for this include the use of carbon materials with larger pores, thus improving the accessibility and mass transfer kinetics of biomolecules. Another property, which must be carefully optimized, is the tendency of the electrode material to adsorb organic molecules even without a potential applied. This reduces the space available for potential controlled adsorption, and counteracts the deionization effect if partial desorption occurs when a potential is applied. On the other hand our data show that several aspects of CDI of weak electrolytes are not fully understood. Future work should therefore also focus on modeling, taking into account dissociation equilibria, water splitting and possible redox reactions.

(2015) 46–50. [15] C.-S. Fan, S.-C. Tseng, K.-C. Li, C.-H. Hou, Electro-removal of arsenic(III) and arsenic(V) from aqueous solutions by capacitive deionization, J. Hazard. Mater. 312 (2016) 208–215. [16] Z. Huang, L. Lu, Z. Cai, Z.J. Ren, Individual and competitive removal of heavy metals using capacitive deionization, J. Hazard. Mater. 302 (2016) 323–331. [17] M.W. Knizia, K. Vuorilehto, J. Schrader, D. Sell, Potential-controlled chromatography of short-chain carboxylic acids, Electroanalysis 15 (1) (2003) 49–54. [18] M. Brammen, P. Fraga-García, S. Berensmeier, Carbon nanotubes-A resin for electrochemically modulated liquid chromatography, J. Sep. Sci. 40 (5) (2017) 1176–1183. [19] L.M. Ponton, M.D. Porter, High-speed electrochemically modulated liquid chromatography, Anal. Chem. 76 (19) (2004) 5823–5828. [20] R.S. Deinhammer, E.Y. Ting, M.D. Porter, Electrochemically modulated liquid chromatography (EMLC): a new approach to gradient elution separations, J. Electroanal. Chem. 362 (1–2) (1993) 295–299. [21] E.Y. Ting, M.D. Porter, Electrochemically modulated liquid chromatography (EMLC) as a probe of the adsorption characteristics of monosubstituted benzenes at porous graphitic carbon, J. Electroanal. Chem. 443 (2) (1998) 180–185. [22] E.Y. Ting, M.D. Porter, Separations of benzodiazepines using electrochemically modulated liquid chromatography. Efficient separations from changes in the voltage applied to a porous graphitic carbon stationary phase, J. Chromatogr. A 793 (1) (1998) 204–208. [23] E.Y. Ting, M.D. Porter, Separations of corticosteroids using electrochemically modulated liquid chromatography: selectivity enhancements at a porous graphitic carbon stationary phase, Anal. Chem. 69 (4) (1997) 675–678. [24] J.C. Farmer, Method and Apparatus for Capacitive Deionization and Electrochemically Purification and Regeneration of Electrodes, US005954937A, 1999. [25] J.C. Farmer, Capacitive deionization of NaCl and NaNO[sub 3] solutions with carbon aerogel electrodes, J. Electrochem. Soc. 143 (1) (1996) 159. [26] M.E. Suss, S. Porada, X. Sun, P.M. Biesheuvel, J. Yoon, V. Presser, Water desalination via capacitive deionization: what is it and what can we expect from it? Energy Environ. Sci. 8 (8) (2015) 2296–2319. [27] R. Zhao, Theory and Operation of Capacitive Deionization Systems, Wageningen University, Wageningen, NL, 2013. [28] S. Porada, L. Weinstein, R. Dash, A. Van Der Wal, M. Bryjak, Y. Gogotsi, P.M. Biesheuvel, Water desalination using capacitive deionization with microporous carbon electrodes, ACS Appl. Mater. Interf. 4 (3) (2012) 1194–1199. [29] Y. Peng, J. Liu, Z. Pan, L.D. Connell, Z. Chen, H. Qu, Impact of coal matrix strains on the evolution of permeability, Fuel 189 (1) (2017) 270–283. [30] C.H. Hamann, W. Vielstich, Elektrochemie 4. Auflage, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2005. [31] F.K. Rogers, Crotonic acid, Industrial 52 (1) (1960) 25–26. [32] J. Rouquerol, D. Avnir, C.W. Fairbridge, D.H. Everett, J.M. Haynes, N. Pernicone, J.D.F. Ramsay, K.S.W. Sing, K.K. Unger, Recommendations for the characterization of porous solids (Technical Report), Pure Appl. Chem. 66 (8) (1994) 1739–1758. [33] K. Kinoshita, Carbon: Electrochemical and physicochemical properties, Other Inf. From Rev. by T. Apple, Univ. Nebraska, J. Am. Chem. Soc. Vol. 110, No. 18 (31 Aug 1988), p. Medium: X; Size: Pages: 541, 1988. [34] H. Ji, X. Zhao, Z. Qiao, J. Jung, Y. Zhu, Y. Lu, L.L. Zhang, A.H. MacDonald, R.S. Ruoff, Capacitance of carbon-based electrical double-layer capacitors, Nat. Commun. 5 (1) (2014) 3317. [35] P.M. Biesheuvel, H.V.M. Hamelers, M.E. Suss, Theory of water desalination by porous electrodes with immobile chemical charge, Colloids Interf. Sci. Commun. 9 (2015) (2015) 1–5. [36] Z. Chen, H. Zhang, C. Yang, X. Sun, H. Guo, C. Wu, F. Xue, L. Gao, Effects of ageing and incorporation of ion-exchange membrane on the electrosorption performance of activated carbon based electrodes modules, Desalin. Water Treat. 51 (16–18) (2013) 3489–3496. [37] I. Cohen, E. Avraham, Y. Bouhadana, A. Soffer, D. Aurbach, Long term stability of capacitive de-ionization processes for water desalination: the challenge of positive electrodes corrosion, Electrochim. Acta 106 (2013) 91–100. [38] J.J. Lado, R.E. Pérez-Roa, J.J. Wouters, M.I. Tejedor-Tejedor, C. Federspill, M.A. Anderson, Continuous cycling of an asymmetric capacitive deionization system: an evaluation of the electrode performance and stability, J. Environ. Chem. Eng. 3 (4) (2015) 2358–2367. [39] M.D. Andelmann, Charge Barrier Flow-Through Capacitor, WO02086195A1, 2002. [40] P.M. Biesheuvel, R. Zhao, S. Porada, A. van der Wal, Theory of membrane capacitive deionization including the effect of the electrode pore space, J. Colloid Interf. Sci. 360 (1) (2011) 239–248. [41] P.M. Biesheuvel, Activated carbon is an electron-conducting amphoteric ion adsorbent, arXiv:1509.06354, Sep. 2015, pp. 9. [42] G. McKay, M.J. Bino, A.R. Altamemi, The adsorption of various pollutants from aqueous solutions on to activated carbon, Water Res. 19 (4) (1985) 491–495. [43] A. Benedek, J. Demarco, J.G. Denblanken, F. Fiessinger, R.A. Hyde, M.J. Mcguire, A.A. Stevens, I.H. Suffet, W.J. Weber, Treatment of water by granular activated carbon discussion-III - biological adsorptive interactions, Adv. Chem. 202 (1983) 373–379. [44] C.Y. Yin, M.K. Aroua, W.M.A.W. Daud, Review of modifications of activated carbon for enhancing contaminant uptakes from aqueous solutions, Sep. Purif. Technol. 52 (3) (2007) 403–415. [45] H. Li, L. Zou, Ion-exchange membrane capacitive deionization: a new strategy for brackish water desalination, Desalination 275 (1–3) (2011) 62–66. [46] P.M. Biesheuvel, A. van der Wal, Membrane capacitive deionization, J. Memb. Sci. 346 (2) (2010) 256–262.

Acknowledgements The studies have been part of the cooperation project “Synthesis, Characterization and Application of new Stationary Phases for the Potential Controlled Liquid-Chromatography in the Field of White Biotechnology”. Many thanks are owed to the Arbeitsgemeinschaft industrieller Forschungsvereinigungen (AIF) for the financial support (project No. 17472 N). Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at References [1] R. Ulber, D. Sell (Eds.), White Biotechnology - Advantages in Biochemical Enigneering/Biotechnology, Springer Verlag, 2007. [2] B.L. Lorraine, K.N. Thomas, Fermentation process for carboxylic acids, US4877731A, 1989. [3] J.A. Asenjo, Separation Processes in Biotechnology, Marcel Dekker Inc., New York, 1990. [4] P.M. Biesheuvel, M.Z. Bazant, R.D. Cusick, T.A. Hatton, K.B. Hatzell, M.C. Hatzell, P. Liang, S. Lin, S. Porada, J.G. Santiago, K.C. Smith, M. Stadermann, X. Su, X. Sun, T.D. Waite, A. van der Wal, J. Yoon, R. Zhao, L. Zou, M.E. Suss, Capacitive Deionization – defining a class of desalination technologies, arXiv:1709.05925 [], 2017, pp. 3. [5] S. Porada, R. Zhao, A. Van Der Wal, V. Presser, P.M. Biesheuvel, Review on the science and technology of water desalination by capacitive deionization, Prog. Mater. Sci. 58 (8) (2013) 1388–1442. [6] S. Porada, L. Weinstein, R. Dash, A. van der Wal, M. Bryjak, Y. Gogotsi, P.M. Biesheuvel, Water desalination using capacitive deionization with microporous carbon electrodes, ACS Appl. Mater. Interf. 4 (3) (Mar. 2012) 1194–1199. [7] M.E. Suss, T.F. Baumann, W.L. Bourcier, C.M. Spadaccini, K.A. Rose, J.G. Santiago, M. Stadermann, Capacitive desalination with flow-through electrodes, Energy Environ. Sci. 5 (11) (2012) 9511. [8] S. Jeon, H. Park, J. Yeo, S. Yang, C.H. Cho, M.H. Han, D.K. Kim, Desalination via a new membrane capacitive deionization process utilizing flow-electrodes, Energy Environ. Sci. 6 (5) (2013) 1471. [9] A. Rommerskirchen, Y. Gendel, M. Wessling, Single module flow-electrode capacitive deionization for continuous water desalination, Electrochem. Commun. 60 (2015) 34–37. [10] G. Wang, B. Qian, Q. Dong, J. Yang, Z. Zhao, J. Qiu, Highly mesoporous activated carbon electrode for capacitive deionization, Sep. Purif. Technol. 103 (2013) 216–221. [11] G. Wang, Q. Dong, T. Wu, F. Zhan, M. Zhou, J. Qiu, Ultrasound-assisted preparation of electrospun carbon fiber/graphene electrodes for capacitive deionization: importance and unique role of electrical conductivity, Carbon N. Y. 103 (2016) 311–317. [12] Q. Ren, G. Wang, T. Wu, X. He, J. Wang, J. Yang, C. Yu, J. Qiu, Calcined MgAllayered double hydroxide/graphene hybrids for capacitive deionization, Ind. Eng. Chem. Res. 57 (18) (2018) 6417–6425. [13] T. Wu, G. Wang, F. Zhan, Q. Dong, Q. Ren, J. Wang, J. Qiu, Surface-treated carbon electrodes with modified potential of zero charge for capacitive deionization, Water Res. 93 (2016) 30–37. [14] Z. Chen, H. Zhang, C. Wu, Y. Wang, W. Li, A study of electrosorption selectivity of anions by activated carbon electrodes in capacitive deionization, Desalination 369


Separation and Purification Technology 209 (2019) 658–665

E. Hack et al.

deionization, J. Mater. Sci. Chem. Eng. 02 (03) (2014) 16–22. [50] K.S.L. Schwartz, Introduction to Tangential Flow Filtration for Laboratory and Process Development Applications, Pall Scientific & Technical Report, 2002. [Online] (Available: literature-library/non-gated/id-34212.pdf).

[47] R. Zhao, O. Satpradit, H.H.M. Rijnaarts, P.M. Biesheuvel, A. van der Wal, Optimization of salt adsorption rate in membrane capacitive deionization, Water Res. 47 (5) (2013) 1941–1952. [48] Y.J. Kim, J.H. Choi, Enhanced desalination efficiency in capacitive deionization with an ion-selective membrane, Sep. Purif. Technol. 71 (1) (2010) 70–75. [49] M. Andelman, Ionic group derivitized nano porous carbon electrodes for capacitive