Membrane capacitive deionization for biomass hydrolysate desalination

Membrane capacitive deionization for biomass hydrolysate desalination

Separation and Purification Technology 118 (2013) 33–39 Contents lists available at SciVerse ScienceDirect Separation and Purification Technology jour...

478KB Sizes 1 Downloads 63 Views

Separation and Purification Technology 118 (2013) 33–39

Contents lists available at SciVerse ScienceDirect

Separation and Purification Technology journal homepage: www.elsevier.com/locate/seppur

Membrane capacitive deionization for biomass hydrolysate desalination Celine Huyskens a,b,⇑, Joost Helsen b, Wim J. Groot c, André B. de Haan c a

Institute for Sustainable Process Technology (ISPT), Groen van Pinstererlaan 37, 3818 JN Amersfoort, The Netherlands Flemish Institute for Technological Research (VITO), Boeretang 200, 2400 Mol, Belgium c Purac Biochem, Arkelsedijk 46, 4206 AC Gorinchem, The Netherlands b

a r t i c l e

i n f o

Article history: Received 30 April 2013 Received in revised form 20 June 2013 Accepted 20 June 2013 Available online 2 July 2013 Keywords: Capacitive deionization Biomass hydrolysates Fermentation Electrochemical treatment

a b s t r a c t Biomass hydrolysates are rapidly gaining interest as low-cost non-food renewable feedstocks for fermentation processes. However, since high concentrations of salt such as sodium and potassium can act toxic to microorganisms, there is a need to remove these salts to maintain high biochemical productivity. In this study, the electrochemical treatment of biomass hydrolysates by membrane capacitive deionization (MCDI) was considered as an auxiliary chemical free, lower cost alternative in comparison to the commonly used ion-exchange processes. Model experiments performed with a commercial bench-scale MCDI set-up and model solutions indicated that none of the most abundant hydrolysate components (sugars, organic acids and furans) prohibited the implementation of MCDI for this application, although performance was lowered by the competition for electro-sorption between the protons deriving from organic acid dissociation and the cations. Such an effect was not observed during MCDI treatment of a real biomass hydrolysate sample. Instead, the results achieved in terms of Na and K removal and energy usage were very comparable to the ones for a model solution with equal conductivity and sugar concentration. As such, this study clearly demonstrates the technical feasibility of MCDI for process streams such as biomass hydrolysates, hereby considerably broadening its potential application field. Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction High production costs remain one of the main obstacles for the replacement of conventional chemical synthesis by microbial fermentation processes to produce biofuels and chemicals like ethanol, butanol, acetone and organic acids. These high costs are mainly associated with the use of relatively expensive carbohydrate sources such as glucose, sucrose and starch [1–3]. Moreover, there is an increasing desire to use carbohydrates obtained from non-food sources such as cellulosic and lignocellulosic biomass [1,4,5]. Consequently, biomass hydrolysates are rapidly gaining interest as low cost non-food renewable feedstocks for fermentation processes [1,6]. However, the application of such secondary feedstocks encompasses one important disadvantage. Extensive pre-treatment of the hydrolysates is required due to their higher complexity as compared to refined sugars. [2,3,7–9]. An example are high concentrations of salts such as sodium and potassium present in the biomass hydrolysates, which can act toxic to the microorganisms [3,10,11]. As such, there is a strong need to remove these salts in order to maintain high biochemical productivity. The commonly used ion-exchange technology [12,13] entails ⇑ Corresponding author. Address: Flemish Institute for Technological Research (VITO), Separation and Conversion Technology, Boeretang 200, 2400 Mol, Belgium. Tel.: +32 14 33 69 49; fax: +32 14 32 65 86. E-mail address: [email protected] (C. Huyskens). 1383-5866/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.seppur.2013.06.032

high operational costs and the generation of a secondary waste stream through the use of chemicals for regeneration [14,15]. Electrochemical pre-treatment of the hydrolysates by membrane capacitive deionization or MCDI could provide a lower cost alternative with minimal waste generation. In conventional CDI (without ion-exchange membranes), salt ions are removed from an influent stream by creating an electric field over two parallel porous carbon electrodes separated by a spacer. As a result, the anions and cations in the influent stream flowing through the spacer are attracted and electro-sorbed onto the anode and cathode respectively (purification step), and a desalinated effluent flows out of the cell. Once the CDI electrodes are saturated, they are regenerated by shorting them so that the ions are released from the electrodes and collected in the spacer. After that, they are flushed from the cell in a highly concentrated waste stream [16,17]. In the more recently developed MCDI process, ionexchange membranes or coatings are positioned in front of the electrodes to prevent co-ion expulsion and enhance CDI efficiency. A second advantage is that these ion-exchange layers make it possible to reverse the polarity over the electrodes during regeneration instead of simply shorting the electrodes, resulting in a better recovery of the carbon electrodes’ sorption capacity [18– 20]. Due to the low voltages applied, MCDI consumes little energy. Moreover, the process is environmentally friendly: the salt ions are merely concentrated in a smaller volume fraction of the original

34

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

influent, without forming a secondary waste stream [21,22]. The technology is also assumed to be rather insusceptible to fouling and scaling because the driving force is an electric field which is reversed at fixed time intervals, and not a pressure gradient [23]. These advantages suggest that MCDI could be a useful technology for a broad range of applications. However, the process is currently only commercialized for ‘relatively clean’ water streams, such as domestic water softening and demineralization of make-up water for cooling towers [24]. In this study, the technical feasibility of MCDI for the removal of Na+ and K+ from the much more complex biomass hydrolysates containing high concentrations of sugars and other organics, is investigated. It should be noted that the implementation of MCDI for process streams is not an entirely novel concept. Jung et al. [25] and Kim et al. [26] have already to a certain extent demonstrated the suitability of MCDI for insulin purification (removal of ZnCl2) and hydrolysate treatment (removal of acetic acid and suplhuric acid) respectively. However, these studies were limited to artificial solutions, which are not always fully representative for real-life streams. In contrast, to our knowledge, this study describes for the first time the implementation of a commercial MCDI system for a real complex process stream. In a first series of experiments, the influence of the most abundant matrix components present in the biomass hydrolysate on MCDI operation were examined using model solutions mimicking the biomass hydrolysate. The next step consisted of the MCDI treatment of the real hydrolysate, targeting the removal of Na+ and K+ to concentrations below 100 mg l1. The achievable treatment performance, the stability of the MCDI process and the required energy usage during this experiment were thoroughly examined and compared to results obtained for a model solution with equal conductivity and sugar content. The experiments performed in this study serve as a true proofof-concept for the application of MCDI for complex process streams and could thus pave the way for a broader application field for MCDI technology. 2. Experimental This paper describes the implementation of MCDI to a real process stream, i.e. a hydrolysate obtained from Purac Biochem and produced from cellulosic biomass. In this hydrolysate, sugar was predominantly present as glucose, whereas Na+, K+ and Cl were the most abundant salt ions that needed to be removed during the MCDI process. The hydrolysate also contained a number of impurities, such as organic acids and furans. Prior to the treatment of the hydrolysate, a number of MCDI tests using model solutions were performed to examine the effects of the most abundant hydrolysate components on MCDI performance. The substances under examination were sugars, organic acids and furans. Their effects were evaluated separately as well as in combination with each other. The next phase consisted of the MCDI treatment of the real biomass hydrolysate. 2.1. MCDI set-up All experiments were performed using a commercially available bench-scale MCDI installation, i.e. the ESD400 supplied by Enpar Technologies Inc. (Canada). A simplified process scheme is shown in Fig. 1. The ESD400 consists of a single MCDI cell with a total activated carbon electrode surface area of 0.7 m2 and employs ion-exchange membranes in front of the electrodes. The system is controlled by a PLC with a touch screen control panel to specify the desired operational settings. During operation, the ESD400 automatically passes through cycles of alternating purification

Power supply

PLC Conductivity controller

MCDI cell Pump

Influent tank

Conductivity sensor

Effluent tank

Waste tank

Fig. 1. Simplified process scheme for the Enpar ESD400.

and regeneration steps. In the purification steps, water is pumped through the cell by a centrifugal pump and a constant voltage is applied. After on-line measurement of its conductivity, the water exiting the cell is directed towards the effluent tank by a threeway valve. Typically, the conductivity will rise in time as the electrodes become increasingly saturated with ions. Regeneration of the electrodes is accomplished by a recharge and purge step. During the recharge step, the cell is shorted, followed by polarity reversal. This way, the ions are repelled from the electrode and collected in the spacer channels. The purge step is subsequently used to flush these desorbed ions from the spacers in a highly concentrated waste stream, whilst the polarity reversal is maintained. During the purge step, the three-way valve is switched so that the waste exiting the cell is directed towards the waste tank. The effluent and waste conductivity, the applied cell voltage and the electric current are continuously logged at a 1 s1 rate by a DaqPRO data logging system (Fourier Technologies, USA). 2.2. Model experiments 2.2.1. Effect of sugars on MCDI performance The effect of sugars was examined by performing MCDI experiments on 1 g l1 NaCl solutions without and with increasing concentrations of glucose or sucrose, i.e. 10%, 30% and 40% w/v. All solutions were prepared in distilled water and stirred throughout the MCDI test to ensure a homogeneous composition. The conductivity and pH of each influent sample was measured prior to the MCDI treatment. The experiments were performed in duplicate, albeit with slightly different settings. During the first test series, a voltage of 1 V during purification and 1 V during recharge and purge, a flow of 0.2 l min1 during purification and purge (no flow during recharge), and respective purification, recharge and purge times of 240 s, 95 s and 35 s were applied. The conditions during the second series of tests were identical except for the longer recharge time of 130 s. All tests were continued until four reproducible purification–regeneration cycles were obtained. Only the data from these cycles were used to ensure steady-state results. At the end of each test, the conductivity and pH of the collected effluent and waste was measured. In between the different tests, the MCDI cell was thoroughly rinsed with distilled water. MCDI performance was evaluated in terms of the achievable effluent NaCl concentration, the ion removal efficiency (gi), the product recovery (gP) and the energy consumption expressed per volume of effluent produced (EV). Eqs. (1)–(3) were used to calculate these variables, with Ve and Vw the effluent and waste volume, Ci and Ce the molar influent and effluent NaCl concentration, I the current, t the time and V the cell voltage. NaCl concentrations were calculated from the measured conductivity using linear calibration

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

curves. Because the presence of sugars affects solution viscosity and hence the measured conductivity, different calibration curves were set up for the different sucrose and glucose concentrations. The term product recovery is used instead of the more generally accepted water recovery, because this paper describes the implementation of MCDI for a process/product stream and not an aqueous solution. It should be noted that the energy usage includes the energy required for desalination and electrode regeneration, but not the energy consumption of the supplementary equipment (e.g. pumping energy).

gP ¼

Ve  100% Ve þ Vw

ð1Þ

gi ¼

Ci  Cw  100% Ci

ð2Þ

EV ¼

V It Ve

ð3Þ

2.2.2. Effect of organic acids and furans on MCDI performance Apart from sugars, biomass hydrolysates contain lower concentrations of other organic components, such as organic acids and furans. The effect of organic acids was evaluated by adding 8 g l1 of acetic acid to a solution containing 1 g l1 of NaCl and 10% w/v glucose. The influence of furans was tested using furfural as a model component. First, 5 g l1 furfural was added to a solution containing solely 1 g l1 NaCl and 10% w/v glucose. Afterwards furfural and acetic acid were tested in combination at respective concentrations of 5 g l1and 8 g l1 in the presence of 1 g l1 NaCl and 10% w/v glucose. All solutions were prepared in distilled water and stirred throughout the experiment. Their conductivity and pH were measured prior to the start of the MCDI tests. The experimental settings were a voltage of 1 V during purification and 1 V during regeneration, a flow of 0.2 l min1 and purification, recharge and purge times of 240 s, 145 s and 25 s respectively. In each experiment, at least four cycles of purification and regeneration were executed. The data from the first cycle were omitted to ensure steady-state results. Conductivity and pH were measured in the effluent and waste tank after each experiment. In between the different experiments, the MCDI cell was thoroughly rinsed with distilled water. The effluent Na+ concentration, ion removal efficiency, product recovery and energy consumption for the different solutions were compared. Because the solution now contained a mixture of various ions, Na+, K+ and Cl concentrations could no longer be determined from the conductivity. Instead, they were measured by ion chromatography (IC).

35

1 V during regeneration, a flow of 0.15 l min1 and purification, recharge and purge times of 250 s, 300 s and 30 s respectively. Similar settings were applied during the second MCDI passage except for the longer purification time of 350 s. After each MCDI passage, the conductivity and pH of the collected effluent and waste were measured. Again, MCDI performance was evaluated in terms of the achievable product recovery, the effluent ion concentration, the ion removal efficiency and the energy consumption. In addition, the sorption capacity for each ion was calculated using Eq. (4). The sugar concentration was measured by high pressure liquid chromatography (HPLC) and Na+, K+ and Cl concentrations by ion chromatography (IC).



ðC i  C e Þ  V e A

ð4Þ

The two-passage MCDI-experiment was repeated for a model solution with an equal sugar concentration of 8% w/v, present as glucose. NaCl was added to this model solution until a conductivity similar to the conductivity of the real hydrolysate was obtained. This way, the achievable treatment performance, the process stability and the required energy usage for a model and a real hydrolysate could be compared. Na+ and Cl concentrations for the model solution were calculated from the measured conductivity according to a linear calibration curve. 3. Results and discussion 3.1. Model hydrolysate experiments 3.1.1. Effect of sugars on MCDI performance Fig. 2 shows the evolution of conductivity at the MCDI cell outlet during the last two treatment cycles of the MCDI process for streams with different glucose and sucrose concentrations. The purification, recharge and purge steps are respectively indicated by the white-, light grey-, and dark grey-shaded areas. The data are plotted for the first test series with a recharge time of 95 s. Sim-

2.3. Real hydrolysate experiment The next phase consisted of the MCDI treatment of the real biomass hydrolysate. This hydrolysate was diluted with distilled water to a sugar concentration of 8% w/v and pre-filtered at 5 lm through a cartridge filter to prevent clogging of the MCDI spacer channels. The conductivity and pH of the 14 kg hydrolysate sample were measured prior to the start of the experiment and the sample was continuously stirred throughout the experiment to ensure a homogeneous composition. Sugar was predominantly present as glucose, whereas Na+, K+ and Cl were the most abundant salt ions. Because a deep desalination down to Na+ and K+ concentrations below 100 mg l1 was required, a two-passage MCDI process was configured. This means that the effluent obtained after the first passage through the MCDI cell was further treated by MCDI during a second passage through the cell. The applied settings during the first MCDI passage were a voltage of 1 V during purification and

Fig. 2. Evolution of the conductivity at the MCDI cell outlet during the last two cycles of the MCDI experiments at different glucose (a) and sucrose (b) concentrations. Data are shown for the first test series.

36

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

ilar graphs were obtained for the duplicate experiments with a longer recharge time. Ions were attracted from the influent stream, transported through the ion-exchange membranes and stored into electrical double layers on the porous electrodes during purification. As a result, the conductivity at the MCDI cell outlet was considerably lower than in the influent. Due to the constant voltage operation, the conductivity of the treated effluent started to rise gradually at the end of each purification step as the electrodes’ limited sorption capacity became increasingly saturated with ions. The electrodes were regenerated by polarity reversal in two consecutive steps: recharge and purge. During recharge, the conductivity at the cell outlet remained constant, because the flow through the cell was interrupted and the solution thus stood still in the tubing. The flow was resumed during the purge step to flush all released ions from the cell in a highly concentrated waste, resulting in sharp conductivity peaks. These peaks reached their maximum value at 10 mS cm1, but this was merely a consequence of the set measurement range of the conductivity probe. In reality, even higher instantaneous conductivity values were achieved. It can be seen from Fig. 2 that MCDI performance was equally stable without or with glucose or sucrose, as evidenced by the very reproducible purification–regeneration cycles for all tests. At first sight, MCDI performance even seemed to improve with increasing sugar concentration, because lower conductivities were measured at the MCDI cell outlet during purification. Moreover, the conductivity peaks were narrower during electrode regeneration, suggesting a more efficient ion desorption. However, it should be kept in mind that sugars affect solution viscosity and thus also the measured conductivity. This already became apparent from the different conductivities measured for the different influent solutions (Table 1), despite the similar 1 g l1 NaCl concentration. To account for this effect, different calibration curves were used to calculate the NaCl concentration from the on-line conductivity measurements for the various sugar concentrations. This resulted in very similar effluent and waste NaCl concentrations for the reference sample as well as the different glucose and sucrose concentrations. In all experiments in the first test series, the concentration of NaCl was reduced from 1000 mg l1 l in the influent to 54–90 mg l1 after MCDI treatment, corresponding to very high single step ion removal efficiencies of 91–95%. Accordingly, the NaCl concentrations measured in the waste ranged from 6.1 to 6.9 g l1 independent of the amount of sugar present. This highly concentrated waste stream is a direct consequence of the very efficient electrode regeneration, consisting of a long recharge step without flow followed by a much shorter and intensive flush of the MCDI cell during purge. The lack of an effect of glucose and sucrose on MCDI performance was confirmed by the duplicate experiments (Table 1), in

Table 1 Influent conductivity, ion removal efficiency and energy usage achieved during the duplicate MCDI experiments at different glucose and sucrose concentrations. Influent conductivity (lS cm1)

Ion removal (%)

Energy usage (kW h m3)

Test 1

Test 2

Test 1

Test 2

1983

94

83

0.61

0.53

Glucose (% w/v) 10 1560 30 985 40 801

93 95 95

86 86 79

0.62 0.58 0.56

0.58 0.57 0.59

Sucrose (% w/v) 10 1510 30 947 40 741

92 93 91

81 81 84

0.62 0.57 0.54

0.53 0.55 0.50

Reference

which the achievable ion removal varied between 79% and 86% for all influent solutions (without or with sugars), corresponding to effluent NaCl concentrations of 140–240 mg l1. The lower and slightly more variable ion removal efficiencies obtained during the second test series are quite surprising because the MCDI settings were identical as in the first tests. Only the recharge time was prolonged in the second test series, but this is rather expected to result in a more efficient electrode regeneration and hence a higher ion removal efficiency. It is possible that the lower ion removal during the second test series is attributed to the higher temperatures at that time, i.e. 21–22 °C vs. 16–17 °C during the first test series. Similarly, a lower ion removal at higher temperature was found by Mossad and Zou [27]. These authors ascribed this effect to the lower sorption capacity of the carbon electrodes, the higher tendency of ions to escape from the electrode surface back into solution and/or the strengthened affinity of carbon for interfacial hydrated ions at higher temperatures. A relatively high product recovery of 87% was obtained, which was identical for all experiments because of the similar purification and purge times applied. The desalination energy usage varied between 0.50 and 0.62 kW h per m3 of produced effluent, independent of the amount of glucose or sucrose present. In contrast, the pressure over the MCDI cell considerably differed between the various experiments. Higher pressures were required to reach the desired flow rate at higher sugar concentrations due to the higher solution viscosity. This was, however, not problematic because the pressure remained constant throughout each experiment and was well within the normal operational pressure range for the ESD400 (<1.5 bar) for all tested glucose and sucrose concentrations. In addition, the effect of the sugars on the operating pressure was clearly reversible: the pressure immediately dropped to lower values upon rinsing the cell with water. It can thus be concluded that glucose and sucrose did not induce any adverse effects on the MCDI process up to concentrations of 40% w/v, neither during purification, nor during regeneration. In fact, the lack of an effect of high sugar concentrations on MCDI performance is somewhat surprising. After all, it is well-known that the effect of these sugars on solution viscosity also lowers the mobility of the ions present. Because the MCDI process is based on the movement of ions under the influence of an electric field, one would therefore also assume an effect on the achievable ion removal efficiency. Of course, it is possible that this effect was simply not strong enough to impact MCDI performance at the tested sugar concentrations of 10–40% w/v. Another possible explanation is that the sugars induced a second opposite effect. Polar sugar molecules are known to strongly attract and bind water molecules when added to an aqueous solution, thereby lowering water activity. Because the available water reservoir is limited, this will in turn result in a decline in the absolute number of water molecules in the hydration shells of the ions present, i.e. Na+ and Cl [28]. It is plausible that less hydrated ions are more easily transported through the ion-exchange membranes [29], penetrate deeper into the small pores of the carbon electrodes [30] and/or form more dense electrical double layers [31]. All these mechanisms would contribute to improved ion removal during MCDI treatment and could thus offset the possible adverse effect of the higher solution viscosity in the presence of sugars. 3.1.2. Furans and organic acids Apart from sugars, biomass hydrolysates typically contain some other organics at low concentration levels [32]. Table 2 summarizes the results of the tests performed to investigate the effects of acetic acid and furfural as model components for organic acids and furans. Because of the variety of ions present in the model solutions, Na+ and Cl concentrations could no longer be determined from the conductivity measured at the MCDI cell outlet.

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

37

Table 2 Conductivity, pH, Na+ concentration in influent, effluent and waste, ion removal efficiency, product recovery and energy usage for the different experiments. Glucose

Glucose + acetic acid

Glucose + furfural

Glucose + acetic acid + furfural

Influent Na+ concentration (mg l1) Conductivity (lS cm1) pH

393 1644 5.5

393 2061 2.6

393 1638 4.6

393 2120 2.7

Effluent Na+ concentration (mg l1) Conductivity (lS cm1) pH

16 73 4.8

344 1039 3.6

16 73 4.6

335 1015 3.6

Waste Na+ concentration (mg l1) Conductivity (lS cm1) pH

3310 11,860 3.8

1500 15,210 <1

3343 13,120 3.0

1370 17,590 <1

CDI performance Ion removal (%) Product recovery (%) Energy usage (kW h m3)

96 91 0.96

13 91 1.56

96 91 0.97

15 91 1.52

Instead, the Na+ concentration was determined using ion chromatography. Again, neither of the organics under investigation influenced the process stability, because very reproducible purification– regeneration cycles were obtained throughout all experiments and the pressure over the cell remained constant and below the maximum allowable pressure of 1.5 bar (data not shown). However, the presence of acetic acid significantly decreased the Na+ removal efficiency from 96% down to 13%. This effects is due to the partial dissociation of acetic acid into acetate ions and protons when added to water. As a result, the total amount of ions in the influent increased, as evidenced by its higher conductivity and lower pH. The protons derived from the acetic acid competed with the Na+ ions for electro-sorption onto the carbon electrodes during the MCDI process. Due to this competition and the higher mobility of protons compared to Na+ ions, a much smaller amount of Na+ was removed as compared to the test with only NaCl and glucose. Accordingly, the amount of Na+ ions desorbed during electrode regeneration was also lower. In support of this hypothesis, the pH increased from 2.6 to 3.6 during purification, pointing at a reduction of the amount of free protons. In addition, extremely low pH values (below 1) were measured in the waste, indicating high concentrations of protons desorbed during cell regeneration. In contrast, in the experiments without acetic acid, the pH tended to decrease during both purification and regeneration, presumably due to parasitic reactions occurring in the MCDI cell (despite the fact that the applied cell voltage was only 1 V). In a recent literature review on CDI, Porada et al. [33] have attempted to summarize all reactions and processes occurring in CDI cells. Apart from capacitive storage, both non-Faradaic (i.e. ion kinetics and the pH-dependent surface charge of the carbon electrodes) and Faradaic phenomena (i.e. carbon redox reactions, water chemistry and carbon oxidation) were suggested as possible electrochemical processes that take place in CDI electrodes and might cause pH fluctuations. Furfural, on the other hand, did not decrease the Na+ removal at the tested concentration of 5 g l1. Both the effluent and waste Na+ concentration were highly similar to the experiment with only glucose. Moreover, when acetic acid and furfural were added in combination, the observed effects were comparable to the situation in which only acetic acid was present: the presence of furfural did not mitigate or aggravate the adverse effect of acetic acid. Again, the total amount of ions and hence the influent conductivity was increased due to acetic acid dissociation. And again, this resulted in competition for electro-sorption with the acetic acid protons and

thus a lower overall Na+ removal and subsequent lower Na+ concentrations in the generated waste. Apart from its effect on Na+ removal, acetic acid also increased the required desalination energy usage from 0.96 kW h m3 to 1.6 kW h m3. This can be also related to the higher conductivity of the acetic acid-containing influent solution, resulting in a higher electric current during purification at constant cell voltage. Furfural, on the other hand, did not affect MCDI energy usage. In conclusion, the experiments demonstrate that MCDI treatment of solutions containing glucose, acetic acid and furfural is technically feasible, but that the protons derived from acetic acid dissociation significantly lower the removal of Na+ as a result of competition for electro-sorption on the porous carbon electrodes. In real hydrolysates, the latter effect is not expected to be this pronounced, because the pH is usually close to neutrality, implying that a much lower amount of free protons is present. On the other hand, real hydrolysates do not only contain Na+ and Cl, but a mixture of various ions at different concentrations. A number of literature studies have revealed the preferential removal of some ions by the MCDI process, depending on ion concentration, charge, hydrated radius, mass and/or time scale [23,27,31,34,35]. Consequently, the Na+ and K+ removal achievable with the MCDI process for real hydrolysates will depend on which other ion species are present. 3.2. Real hydrolysate experiments Since no detrimental effects on MCDI performance were observed for the most abundant organic components present in biomass hydrolysates in the model experiments, the technology was subsequently applied for the real hydrolysate sample. This hydrolysate sample was diluted with distilled water to a sugar concentration of 8% w/v, almost exclusively present as glucose. The diluted sample had a conductivity of 3650 ls cm1 and contained 800 mg l1 of Na+, 201 mg l1 of K+ and 1210 mg l1 of Cl. Other ions might have been present as well, albeit at much lower concentrations. To prevent clogging of the spacer channels, the diluted sample was prefiltered in a cartridge filter at 5 lm. Because a deep desalination down to Na+ and K+ concentrations below 100 mg l1 was set as the target for this test, a two-passage MCDI experiment was performed. The conductivity, ion concentrations and removal efficiencies after each passages are listed in Table 3. No clogging of the MCDI cell was observed throughout the test, indicating that no further pre-treatment of the hydrolysate was required besides the 5 lm pre-filtration. In addition, purification and

38

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

regeneration cycles were very stable without any noticeable deterioration in time. After its first passage through the MCDI system, the conductivity of the hydrolysate was decreased to 939 lS cm1 and Na+, K+ and Cl concentrations were lowered to 185 mg l1, 39 mg l1 and 259 mg l1. This corresponds to respective removal efficiencies for these ions of 77%, 81% and 79% and sorption capacities of 24, 4 and 24 mmol m2. The much lower sorption capacity of K+ compared to Na+ is in line with the findings of Xu et al. [23] and Mossad and Zou [27] that the selectivity of carbon electrodes in multi-ionic solutions is primarily based on the molar feed concentration. After all, K+ was present in the influent in much lower concentrations than Na+. It can be seen that the sum of electro-sorbed Na+ and K+ (28 mmol m2) is higher than the sorption capacity for Cl (24 mmol m2), suggesting that anions other than Cl were also removed during MCDI treatment (but these were not quantified). The effluent of the first MCDI step was further treated during a second MCDI passage with similar settings. Only the purification time was prolonged, because slower electrode saturation was expected due to the lower ion concentration in the feed of this second MCDI step compared to the original hydrolysate sample. This second MCDI passage resulted in a further reduction of conductivity to 200 lS cm1, corresponding to Na+, K+ and Cl concentrations of 32 mg l1, 7 mg l1 and 66 mg l1. Ion removal efficiencies and sorption capacities for Na+, K+ and Cl were 83%, 82% and 74% and 6, 1 and 5 mmol m2 respectively. Again, sorption capacities were determined by the molar ion concentration in the feed, which was higher for Na+ than for K+. And again, the amount of electrosorbed Na+ and K+ (7 mmol m2) was higher than Cl (5 mmol m2), implying the removal of other anions. Moreover, the found sorption capacities during this second MCDI step were much lower than in the first MCDI passage, despite the longer purification time applied. This can be explained by the much higher feed concentration in the first MCDI step, resulting in a lower elec-

Table 3 Conductivity, Na+, K+ and Cl concentrations in the real and model hydrolysate and in the effluents obtained after the consecutive CDI treatment steps. Ion removal efficiencies are shown in between brackets. Product recovery and energy usage are given for each separate MCDI passage as well as for the overall process. Real hydrolysate

Model hydrolysate

Influent Sugar concentration (g l1) pH Conductivity (lS cm1) Na+ (mg l1) K+ (mg l1) Cl (mg l1)

80 4.7 3650 800 201 1210

80 5.7 3660 826

Effluent after first CDI step Effluent pH Effluent conductivity (lS cm1) Effluent Na+ (mg l1) Effluent K+ (mg l1) Effluent Cl (mg l1) Product recovery (%) Energy usage (kW h m3)

5.8 939 185 (77%) 39 (81%) 259 (79%) 89 1.8

5.6 824 186 (77%)

Effluent after second CDI step Effluent pH Effluent conductivity (lS cm1) Effluent Na+ (mg l1) Effluent K+ (mg l1) Effluent Cl- (mg l1) Product recovery (%) Energy usage (kW h m3)

5.8 200 32 (83%) 7 (82%) 66 (74%) 92 0.7

5.1 107 24 (87%)

Global CDI performance Product recovery (%) Energy usage (kW h m3) Energy usage (kW h kg1 sugar)

82 2.7 0.03

82 2.3 0.03

1274

287 (77%) 89 1.6

37 (87%) 92 0.5

trical double layer thickness. Consequently, electrical double layers could also be formed in smaller pores and the electrical double layer overlapping effect between adjacent pores was reduced, making a larger portion of the carbon pores accessible for the salt ions. It is clear from the obtained results that the concentrations achieved for Na+ and K+ after the two-passage MCDI process were well below the aimed target of 100 mg l1, demonstrating the technical feasibility of MCDI for hydrolysate desalination. In a next step, it was investigated whether the MCDI performance for this real hydrolysate was comparable to results achievable for a model solution containing equal amounts of sugar. NaCl was added to this solution until a conductivity similar to the one of the real hydrolysate was obtained. Like the real hydrolysate, this model solution was treated by a two-passage MCDI process. The achieved MCDI performance is represented in Table 3. In the first step, the conductivity decreased from 3660 lS cm1 to 824 lS cm1, corresponding to Na+ and Cl concentrations of 186 mg l1 and 287 mg l1 respectively. The NaCl removal was 77% and the sorption capacity for Na+ and Cl was 25 mmol m2. These values are in the same order of magnitude as for the real hydrolysate. Likewise, MCDI performance for the model solution was similar to the real hydrolysate in the second step, with an NaCl removal efficiency of 88% and a sorption capacity of 6 mmol m2. A final conductivity of 107 lS cm1 and Na+ and Cl concentrations of 24 and 37 mg l1 were reached. These values are slightly lower than for the real hydrolysate. However, the real hydrolysate contained ionic species other than Na+ and Cl (K+, other anions, etc.) which were also removed to some extent during the MCDI process. Table 3 lists the product recoveries and energy usages for MCDI treatment of the real and model hydrolysate, for each step separately as well as for the overall MCDI process. Of course, the similar MCDI settings applied for both solutions resulted in a similar overall product recovery of 82%. It can be noted that this is lower than for the one-step MCDI processes in the previous tests, because each step generates its own waste stream during electrode regeneration. The product recovery for the second MCDI passage was slightly higher than for the first due to the longer purification time (while the purge time remained the same). The energy usage required for the removal of Na+ and K+ from the real hydrolysate was 2.7 kW h per m3 of treated hydrolysate or 0.03 kW h per kg of desalinated sugar. This is slightly higher than the energy required for the desalination of the model solution, while the final conductivity was higher. So apparently, the MCDI process was slightly less energy efficient for the real hydrolysate. Again, it should be borne in mind that the hydrolysate contained ions other than Na+, K+ and Cl, which contributed to the measured conductivity and were also removed to some extent during the process. For example, small amounts of divalent ions might have been present, which are preferentially removed over monovalent ions according to Seo et al. [36]. Previous results [37] have indicated that the removal of these divalent ions is more energy-intensive due to their higher conductivity and thus higher current at constant voltage operation. In addition, the hydrolysates are known to contain organic acids, which are partially dissociated into ions. Just like all other charged species, the conjugate bases of these organics acids (e.g. acetate ions) will be subjected to the electric field in the MCDI cell and attracted towards the electrodes. However, due to their large size, these conjugate bases are less easily transported through the ion-exchange membranes and stored into the electrical double layers in the carbon electrodes, which could also lead to a higher energy usage. From the found results, it can be concluded that MCDI treatment of real biomass hydrolysates is technically feasible, with comparable removal efficiencies and product recoveries as for model solutions. This finding considerably broadens the potential

C. Huyskens et al. / Separation and Purification Technology 118 (2013) 33–39

application fields for MCDI technology. In the future, an elaborate economical evaluation of MCDI technology, including both capital and operational expenditures, will be performed to evaluate the economical feasibility of the technology for the desalination of process streams. In addition, the technology will be benchmarked against the ion-exchange processes which are applied nowadays for biomass hydrolysate treatment. 4. Conclusions Biomass hydrolysates are rapidly gaining interest as non-food renewable feedstocks for fermentation processes to produce chemicals and fuels. They do, however, require more extensive pretreatment compared to current primary feedstocks due to their higher complexity. An example are the high concentrations of Na+ and K+ present in the hydrolysates, which can inhibit biochemical production and thus need to be reduced. In this study, membrane capacitive deionization technology was considered as a lower cost and more sustainable alternative for the commonly applied ion-exchange processes. Model experiments performed with a commercial bench-scale MCDI set-up and model solutions indicated that none of the most abundant components present in biomass hydrolysates (sugars, organic acids and furans) prohibited the implementation of MCDI for this application. However, the protons derived from organic acid dissociation did compete with the cations for electro-sorption onto the carbon electrodes and thus lowered MCDI performance. Such an adverse effect was not observed during the multiple passage MCDI treatment of the real biomass hydrolysate sample. Instead, the results achieved in terms of Na+ and K+ removal as well as energy usage were very comparable to the ones for a model solution with equal conductivity and sugar concentration. As such, this study clearly demonstrates that the implementation of MCDI for process streams such as biomass hydrolysates is technically feasible. This finding considerably broadens the potential application fields for MCDI technology. Acknowledgements The authors gratefully acknowledge the Institute for Sustainable Process technology (ISPT) for financial support of this research. They would also like to thank Kristof Mijnendonckx and Jan Stroobants (VITO) for technical support and performing some of the experiment, and Enpar Technologies Inc. for their support and advice. References [1] M.A. Abdel-Rahman, Y. Tashiro, K. Sonomoto, Lactic acid production from lignocellulose-derived sugars using lactic acid bacteria: overview and limits, J. Biotechnol. 156 (2011) 286–301. [2] H.G. Joglekar, I. Rahman, S. Babu, B.D. Kulkarni, A. Joshi, Comparative assessment of downstream processing options for lactic acid, Sep. Purif. Technol. 52 (2006) 1–17. [3] H.B. Klinke, A.B. Thomsen, B.K. Ahring, Inhibition of ethanol-producing yeast and bacteria by degradation products produced during pre-treatment of biomass, Appl. Microbiol. Biotechnol. 66 (2004) 10–26. [4] Y.H. Weng, H.J. Wei, T.Y. Tsai, W.H. Chen, T.Y. Wei, W.S. Hwanga, C.P. Wang, C.P. Huang, Separation of acetic acid from xylose by nanofiltration, Sep. Purif. Technol. 67 (2009) 95–102. [5] S. Brethauer, C.E. Wyman, Review: continuous hydrolysis and fermentation for cellulosic ethanol production, Bioresour. Technol. 101 (2010) 4862–4874. [6] J. Doran-Peterson, D.M. Cook, S.K. Brandon, Microbial conversion of sugars from plant biomass to lactic acid or ethanol, Plant J.l 54 (2008) 582–592. [7] C.E. Wyman, B.E. Dale, R.T. Elander, M. Holtzapple, M.R. Ladisch, Y.Y. Lee, Coordinated development of leading biomass pretreatment technologies, Bioresour. Technol. 96 (2005) 1959–1966. [8] E. Palmqvist, B. Hahn-Hägerdal, Fermentation of lignocellulosic hydrolysates. I: inhibition and detoxiÒcation, Bioresour. Technol. 74 (2000) 17–24.

39

[9] P.T. Pienkos, E.M. Zhang, Role of pretreatment and conditioning processes on toxicity of lignocellulosic biomass hydrolysates, Cellulose 16 (2009) 743–762. [10] N. Quresh, B.S. Dien, N.N. Nichols, B.C. Saha, M.A. Cotta, Genetically engineered Escherichia Coli for ethanol production from xylose: substrate and product inhibition and kinetic parameters, Food Bioprod. Proc. 84 (2006) 114–122. [11] S. Walton, A. van Heiningen, P. van Walsum, Inhibition effects on fermentation of hardwood extracted hemicelluloses by acetic acid and sodium, Bioresour. Technol. 101 (2010) 1935–1940. [12] N.O. Nilvebrandt, A. Reimann, S. Larsson, L.J. Jönsson, Detoxification of lignocellulose hydrolysates with ion-exchange resins, Appl. Biochem. Biotechnol. 91–93 (2001) 35–49. [13] S.S. da Silva, A.K. Chandel, D-Xylitol: Fermentative Production, Application and Commercialization, first ed., Springer Berlin Heidelberg, 2012. [14] M.L. Soto, A. Moure, H. Dominguez, J.C. Parajo, Recovery, concentration and purification of phenolic compounds by adsorption: a review, J. Food Eng. 105 (2011) 1–27. [15] S.I. Mussatto, I.C. Roberto, Alternatives for detoxification of diluted-acid lignocellulosic hydrolyzates for use in fermentative processes: a review, Bioresour. Technol. 93 (2004) 1–10. [16] M. Andelman, Flow through capacitor basics, Sep. Purif. Technol. 80 (2011) 262–269. [17] P.M. Biesheuvel, A. van der Wal, Membrane capacitive deionization, J. Membr. Sci. 346 (2010) 256–262. [18] Y.J. Kim, J.H. Choi, Enhanced desalination efficiency in capacitive deionization with an ion-selective membrane, Sep. Purif. Technol. 71 (2010) 70–75. [19] J.Y. Lee, S.J. Seo, S.H. Yun, S.H. Moon, Preparation of ion exchanger layered electrodes for advanced membrane capacitive deionization (MCDI), Water Res. 45 (2011) 5375–5380. [20] H. Li, L. Zou, Ion-exchange membrane capacitive deionization: a new strategy for brackish water desalination, Desalination 275 (2011) 62–66. [21] M.A. Anderson, A.L. Cudero, J. Palma, Capacitive deionization as an electrochemical means of saving energy and delivering clean water. Comparison to present desalination practices: will it compete?, Electrochim Acta 55 (2010) 3845–3856. [22] Y. Bouhadana, E. Avraham, A. Soffer, D. Aurbach, Several basic and practical aspects related to electrochemical deionization of water, AIChE 56 (2010) 779–789. [23] P. Xu, J.E. Drewes, D. Heil, G. Wang, Treatment of brackish produced water using carbon aerogel-based capacitive deionization technology, Water Res. 42 (2008) 2605–2617. [24] E. Driessen, Deionisation-based desal for cooling tower water treatment, Water 21 (2011) 46–47. [25] S.M. Jung, J.H. Choi, J.H. Kim, Application of capacitive deionization (CDI) technology to insulin purification process, Sep. Purif. Technol. 98 (2012) 31– 35. [26] S.J. Kim, J.H. Choi, J.H. Kim, Removal of acetic acid and sulfuric acid from biomass hydrolyzate using a lime addition–capacitive deionization (CDI) hybrid process, Proc. Biochem. 47 (2012) 2051–2057. [27] M. Mossad, L. Zou, A study of the capacitive deionisation performance under various operational conditions, J. Hazard. Mater. 213–214 (2012) 491–497. [28] V. Martorana, L. La Fata, D. Bulone, P.L. San Biagio, Potential of mean force between two ions in a sucrose rich aqueous solution, Chem. Phys. Lett. 329 (2000) 221–227. [29] L. Banasiak, A.I. Schäfer, The removal of boron, fluoride and nitrate in electrodialysis in the presence of bulk organic matter, J. Membr. Sci. 334 (2009) 101–109. [30] Y. Gao, L. Pan, H. Li, Y. Zhang, Z. Zhang, Y. Chen, Z. Sun, Electrosorption behavior of cations with carbon nanotubes and carbon nanofibres composite film electrodes, Thin Solid Films 517 (2009) 1616–1619. [31] C.J. Gabelich, T.D. Tran, I.H. Suffet, Electrosorption of inorganic salts from aqueous solution using carbon aerogels, Environ. Sci. Technol. 36 (2002) 3010– 3019. [32] M.A. Franden, P.T. Pienkos, M. Zhang, Development of a high-throughput method to evaluate the impact of inhibitory compounds from lignocellulosic hydrolysates on the growth of Zymomonas mobilis, J. Biotechnol. 144 (2009) 259–267. [33] 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. (2013). doi: http://dx.doi.org/10.1016/j.pmatsci.2013.03.005. [34] B. van Limpt, Performance relations in capacitive deionization systems, Wageningen University, PhD Dissertation, 2010. [35] R. Zhao, P.M. Biesheuvel, A. van der Wal, Energy consumption and constant current operation in membrane capacitive deionization, Energy Environ. Sci. 5 (2012) 9520–9527. [36] S.J. Seo, H. Jeon, J.K. Lee, G.Y. Kim, D. Park, H. Nojima, J. Lee, S.H. Moon, Investigation on removal of hardness ions by capacitive deionization (CDI) for water softening applications, Water Res. 44 (2010) 2267–2275. [37] C. Huyskens, J. Helsen, A.B. de Haan, Capacitive deionization for water treatment: Screening of key performance parameters and comparison of performance for different ions. Submitted for publication in Desalination.