1. Introduction
Dairy production generates effluents with variable composition depending on the final product. Acid whey is generated from the manufacture of Greek-type yoghurts, fresh cheeses (cream or cottage cheeses), and caseins resulting from milk acidification. In the past few years, the demand for those dairy products has continued to increase, leading to the production of significant volumes of acid whey. Considered as the most contaminated waste generated in dairy production because of its high organic content [
1,
2], this co-product can be valorized in different ways to limit its environmental impact [
2]. Since whey contains compounds of interest, such as lactose and whey proteins, with high nutritional value [
3,
4], it is processed into powder by spray drying [
5] to facilitate its use and transportation. However, the high calcium and lactic acid contents in acid whey impact the quality of the powder generated and the energy consumption of the drying process [
6,
7]. In fact, as the concentrations of calcium and lactic acid in whey increase, the glass transition temperature of lactose decreases, affecting its properties [
6,
7]. Consequently, lactose is mainly present in crystalline phase, leading to a sticky powder that reduces process efficiency. Considering those facts, acid whey needs to be demineralized and deacidified to optimize the drying process and limit wastes [
5].
Many processes and experimental conditions have been tested to realize this critical preliminarily step such as nanofiltration (NF) [
8], ion-exchange resin [
9] and ion-exchange resin coupled to electrodialysis (ED) [
10]. However, NF allows partial demineralization and deacidification of acid whey (46–60% demineralization and 30% reduction in lactic acid) [
11], and ion-exchange resin produces a large amount of effluents during regeneration, which needs to be treated. Consequently, ED without coupling ion-exchange resins would be a very interesting alternative in terms of ecoefficiency, since ED is able to separate simultaneously minerals and organic acid anions [
9]. However, scaling (mineral fouling) formation at the surface of the ion-exchange membrane is still a major problem that needs to be solved to preserve membrane integrity and optimize the valorization of acid whey. Different solutions are already available to diminish or control ED scaling to some extent, such as pre-treatment of the feed solution [
12,
13], the control of hydrodynamic conditions [
14,
15], reversal ED [
16] and modifications of the membrane surface properties [
14,
17]. However, a cleaning-in-place procedure, which employs chemicals and generates effluents, is still needed to recover the process performance over extended periods [
18]. Recent investigations showed that an overlimiting current condition improved ED performances [
19,
20]. Indeed, in overlimiting conditions, the electroconvective vortices (EVs) generated facilitate the transport of ions toward the ion-exchange membrane (IEM) surface and limit fouling formation. Hence, Bukhovets et al. [
21] reported the influence of electroconvection on organic fouling prevention by the “washing out” effect of EVs, while Mikhaylin et al. [
22] demonstrated the positive effect of electroconvection on IEM scaling mitigation. To the best of our knowledge, the application of overlimiting current condition has only been tested on the demineralization of model solution with calcium and magnesium compounds [
22] and on whey protein concentrate [
23]. No information is available in the literature concerning the influence of this non-conventional current condition on the simultaneous acid recovery and demineralization of complex solutions such as acid whey.
In addition, the use of charged filtration membranes stacked in an ED cell was demonstrated to be effective for demineralization of solutions. Hence, in 2011, Bazinet and Moalic [
24] were the first to use a nanofiltration membrane (NF) in an ED stack for the demineralization of sea water. They demonstrated that the use of NF membrane can also allow the selective separation of cations. More recently, Ge et al. [
25] exemplified such a selective cation fractionation of EDNF with H
+/Zn
2+ and Na
+/Mg
2+ systems and demonstrated that an NF membrane can increase the limited current density in ED. In the case of acid whey demineralization and the deacidification process, these membranes could allow a better mass transfer due to their larger pore size and their surface charge, particularly for lactic acid. However, the use of nanofiltration membrane in an ED cell for the deacidification of acid whey has never been reported in the literature.
In this context, the main goal of this study was to explore the use of NF membrane in the replacement of anion-exchange membrane (AEM) stacked in the ED cell as well a combination of the overlimiting current condition in order to potentially optimize lactic acid recovery and the demineralization of acid whey and prevent the formation of scaling.
2. Materials and Methods
2.1. Whey
The acid whey samples were obtained from the Parmalat Canada (Winchester, Ontario, Canada) dairy processing plant and transported at 4 °C until they reached the laboratory. They were aliquoted, stored at −30 °C, and thawed at 4 °C before each experiment. Their composition, as described in
Table 1, was in accordance with the literature [
6,
26,
27].
2.2. Electrodialysis Cell
An MP-type ED cell (ElectroCell AB, Täby, Sweden) with an effective membrane surface area of 100 cm
2 was used for the experimentations in this work. This cell included a dimensionally stable electrode (DSA-O
2) as the anode and a stainless steel electrode as the cathode. The first membrane configuration “CACAC” (
Figure 1a), a conventional industrial ED configuration for demineralization, combined three cation-exchange membranes (CMX-fg, Astom, Tokyo, Japan) and two anion-exchange membranes (AMX-fg, Astom) as described by Chen et al. [
26] and Dufton et al. [
27]. In the second configuration “CNfCNfC” (
Figure 1b), the AEMs were replaced by NF membranes with a molecular weight cut-off of 500 Da (NFX, Synder Filtration, Vacaville, USA) to obtain a similar configuration to the one described by Bazinet and Moalic [
24]. According to the manufacturer information, the NF membranes were positively charged at acid whey’s pH.
NaCl 5.5 g/L (Fisher Scientific, Ottawa, ON, Canada) was used as the recovery solution and Na
2S0
4 20.0 g/L (Anachemia, VWR International, Mississauga, ON, Canada) was used as the electrode rinsing solution. First, 500 mL of whey and NaCl solutions circulated in two different loops to create two deacidification/demineralization units (dotted line,
Figure 1a,b). In underlimiting current density ED experiments, a 15A/60V power supply (Model 9110, BK Precision, Yorba Linda, CA, USA) was used to generate the potential difference, while a 5A/100 V power supply (Ionics Inc, Watertown, MA, USA) was used in the overlimiting current condition.
2.3. Protocol
Prior to ED experiments, membrane samples were characterized in terms of electrochemical characteristics (current density, resistance, thickness of the boundary layer, and transition time). In parallel, the limiting current density (LCD) of the ED stack was determined by the Cowan and Brown method [
28]. Briefly, whey, NaCl, and Na
2SO
4 were circulated in the ED stack (
Figure 1) as for the main experiments. The voltage was gradually increased from 0 to 40 V by increments of 0.5 V, and the resulting current intensity was monitored. The values of intensity and voltage obtained were plotted as resistance (U/I) as a function of reciprocal current (1/I). LCD was determined by the intersection of the tangents of the curves (
Figure 2). Since the LCD of each studied ED stack was very close (
Figure 2), the control ED experiments were carried out at a current density corresponding to 70% of this value.
This underlimiting current condition corresponds to a constant current of 0.7 A (current density of 7 mA/cm2) and was applied during 1 h with no temperature control and at flow rates of 400 mL/min in whey and NaCl compartments. A volume of 10 mL of samples was collected every 15 min during the process. They were frozen at −30 °C until further analyses such as lactic acid, lactose, and mineral contents. To complete this protocol, both configurations were carried out in an overlimiting current condition at 3.0 A (corresponding to 30 mA/cm2, largely over the limiting current value determined) during 30 min with no temperature control and the same flow rates as in the underlimiting conditions. In this study, the temperature was deliberately not controlled. Indeed, since the overlimiting current would increase the temperature, by controlling the temperature increase, its potential effect on the molecules’ transfer and energy efficiency would have not been taken into account. Furthermore, in a context of sustainability, controlling the temperature increase during ED at the overlimiting temperature would use lots of energy, and the potential energy saving related to the increase in temperature would be lost. So, to test both processes in real conditions and to be more close to the reality, the temperature was consequently not controlled. The ED duration was fixed to 30 min since during preliminary tests, it was observed, as expected, that the demineralization was faster and did not need to reach 1 h as for the underlimiting conditions. A volume of 10 mL of samples was also collected at 0, 5, 10, 20 and 30 min. Then, they were frozen until further analyses. All ED combinations of current conditions and configurations were repeated three times.
2.4. Analysis Methods
2.4.1. Membrane Electrochemical Characteristics
Current–voltage (CV) curves as well as chronopotentiograms (ChP) were determined according to Villeneuve et al. [
29] and Mikhaylin et al. [
22]. Briefly, AEM and NF membranes were placed between two Haber–Luggin’s capillaries with an external diameter of 0.8 mm. A 0.02 M NaCl solution was circulated at flow rates of 32.9 mL/min for AEM and 32.4 mL/min for NFX. The current density, resistance, and thickness of the boundary layer were determined according to the CV curves, while the transition time representing the time required to reach the limiting current was determined according to the ChP.
2.4.2. pH
The pH of both compartments (NaCl and whey) was measured using a Symphony pHmeter (Model SP70P, VWR, West Chester, PA, USA) equipped with a temperature probe to compensate for temperature changes.
2.4.3. Lactic Acid and Lactose Contents
Lactic acid and lactose concentrations in collected samples during ED were determined with high-performance liquid chromatography (HPLC) with a Waters chromatograph (Waters Corp., Milford, MA, USA) equipped with a Hitachi differential refractometer detector L-7490 (Foster City, CA, USA), a 600E controller, a column oven, and a thermostated 717 Plus autosampler. An ICSep ICE-ION-300 column (Transgenomic, Omaha, NE, USA) was used with 8.5 mM of H2SO4 (180 µL H2SO4/L) solution as the mobile phase and at a flow rate of 0.4 mL/min. The column temperature was kept constant at 40 °C. Samples were centrifuged during 5 min at 5000 rpm (AllegraTM 25R Centrifuge, Beckman Coulter, Brea, CA, USA) and filtered (0.22 µm nylon; CHROMSPEC Syringe Filter; Chromatographic Specialties, Brockville, ON, Canada) before injection (15 µL). Each sample was analyzed during a 36 min run time. Lactose anhydrous (PHR1025) and L-(+)-lactic acid (L1750) from Sigma-Aldrich (Saint-Louis, MO, USA) were used as external standards.
The deacidification or lactic acid recovery rate (DaR) related to each treatment was calculated by considering the acid lactic concentration detected in the samples collected at the beginning (C
i in ppm) and the end (C
f in ppm) of ED treatments according to Equation (1).
2.4.4. Conductivity
Acid whey and NaCl conductivities were measured with an YSI conductivity meter (Model 3100, Yellow Springs Instrument, Yellow Springs, OH, USA) equipped with an immersion probe (Model 3252, cell constant K = 1.0/cm). Then, the demineralization/mineralization rate based on the conductivity measurement was calculated with the same equation than for the deacidification rate (Equation (1)), considering the conductivity of whey and NaCl at the beginning (Ci in mS/cm) and the end (Cf in mS/cm) of each treatment.
2.4.5. Mineral Content
As described by Dufton et al. [
27], the samples collected during ED treatment were thawed at 4 °C before their dilution 1:20 in Milli- Q water to reach a final volume of 10 mL. Calcium, potassium, magnesium, sodium, and phosphorus concentrations were determined by optical emission spectrometry with inductively coupled plasma as an atomization and excitation source (ICP-OES Agilent 5110 SVDV Agilent Technologies, Victoria, Australia), using the following wavelengths (nm): 393.366; 396.847; 422.673 (Ca), 766.491 (K), 279.553; 280.270; 285.213 (Mg), 588.995; 589.592 (Na), 177.434; 178.222; 213.618; 214.914 (P). The analyses for all ions were carried out in axial and/or radial view. The demineralization rate based on the mineral concentration for each treatment was calculated using Equation (1) while considering the total ion concentration in acid whey at the beginning (C
i in ppm) and at the end (C
f in ppm) of each treatment.
2.4.6. Overall System Resistance
The voltage (U) and current intensity (I) directly obtained from the two power supplies (Model 9110, BK Precision, Yorba Linda, CA, USA and Ionics Inc, Watertown, MA, USA) were used to calculate the global system resistance (R) according to Ohm’s law (U = R·I).
2.4.7. Energy Consumption
The energy consumption (in Wh) was calculated according to Equation (2) where I is the current intensity (in A), U(t) tis he voltage (in V) as a function of time, and t is the duration of the ED treatment (min.)
Energy consumption for demineralization was calculated according to the concentration of K
+ migrated from whey, since it is the ion having the higher migration rate [
27].
2.5. Statistical Analyses
Data were subjected to one-way and two-way, and three-way analyses of variance (ANOVA). A Tukey multiple comparison test was performed to compare treatments together at a probability level of p < 0.05 (SAS software, version 6.3 for Windows, SAS Institute, Inc., Cary, NC, USA).
4. Conclusions
It was the first time that NF was used as an AEM in an ED stack for the treatment of acid whey and compared with a classic AEM-containing ED stack while applying conventional underlimiting and emerging overlimiting current conditions. The results showed that the overlimiting condition led to greater deacidification rates with CNfCNfC configuration (40%) while it had a more significant impact on demineralization with a CACAC configuration (87%). However, these two optimal conditions needed more energy than treatments performed with the underlimiting current condition. Their conductivity, pH, and global resistance were also more influenced by the water splitting phenomena that occurred while the limiting current was reached. Furthermore, it appeared that the selectivity for sodium was drastically decreased or suppressed in the overlimiting condition for the NF membrane, while no change was observed for other cations.
Further experiments also have to be carried out on the CNfCNfC membrane configuration to study the generation of EVs and their development on both AEM and NF membrane types. In addition, applying a pulsed electric field in both underlimiting and overlimiting current conditions could be interesting in order to compare their impact on acid whey treatment efficiency and to potentially enhance the demineralization and deacidification processes.