Collection Mémoires et thèses électroniques
AccueilÀ proposNous joindre

Chapitre 4 Caractérisation de la morphologie et de la composition du colmatage des membranes échangeuses d’anions formé pendant la déminéralisation par ÉD conventionnelle

Table des matières

Tel que mentionné dans la présentation du chapitre précédent, la présence de deux types de colmatage, protéique et minéral, a été identifiée au cours des travaux décrits dans le deuxième chapitre. Comme pour les MEC, la présence ou l’absence de ces colmatages sur les MEA dépendait des conditions du milieu. Ce chapitre, rédigé sous forme d’article avait comme but la confirmation de la nature des colmatages formés sur les MEA ainsi que la caractérisation et la description morphologique du colmatage minéral. Ces travaux complètent le projet de recherche orienté à la compréhension du phénomène du colmatage des membranes d’ÉD utilisées pour le traitement des aliments et produits nutraceutiques.

L’objectif de ces travaux était l’étude de l’effet du pH du concentrat et de la composition minérale d’une solution de protéines de lactosérum traitée par ÉD conventionnelle sur le colmatage des MEA. Une attention spécifique a été portée aux possibles mécanismes responsables de la formation du colmatage et leur relation avec la disposition de l’empilement d’ÉD. Un colmatage de phosphate de calcium avec une forme de filaments cylindriques a été caractérisé sur les membranes utilisées pour la déminéralisation du diluat contenant seulement du CaCl2 avec un concentrat à pH 7. Un autre colmatage de phosphate de calcium, mais en agrégats spongieux irréguliers a été caractérisé pour les mêmes conditions minérales mais avec un concentrat à pH 12. Finalement, les membranes utilisées pour les traitements avec un concentrat à pH 2 ou 7, indépendamment de la composition minérale du diluat, ont été colmatées par une couche de gel protéique.

The aim of this work was to study the effect of the concentrate solution pH value and of the composition in calcium, carbonate and protein of the diluate solution to be treated by conventional electrodialysis on the fouling of anion-exchange membranes (AEM). It appeared that after demineralization of solutions containing CaCl2 using a concentrate solution maintained at a pH value of 7 and 12, a mineral fouling appeared on the AEM surface in contact with the concentrate. The mineral deposits presented a cylindrical filament-shape for condition with a concentrate solution pH value of 7, while, for a pH value of 12, the mineral deposit had a crumbly and sponge texture formed by irregular aggregates. The nature of the fouling was identified as a calcium phosphate with or without calcium hydroxide. In addition, a gel-like protein fouling was detected on the AEM surface in contact with the diluate, after demineralization procedures using a concentrate pH value of 2 or 7, regardless of the mineral composition of the diluate.

Conventional electrodialysis (ED) is a separation method that uses cation-exchange membranes (CEM) and anion-exchange membranes (AEM), which are permeable to cationic and anionic species, respectively. Recently, ED has gained attention due to the attractive and potential applications of ED as a modification, separation and purification technique for the nutraceutical and food industry (Vera-Calle et al., 2002; Bazinet et al., 2004; Labbé et al., 2005). These applications require special conditions and membranes in order to avoid fouling formation, due to the complex composition of foodstuffs. In the case of ED treatment of liquid foodstuffs, the presence of proteins contributes to such limitation of AEM. Proteins constitute a particularly delicate component, due to the possibility of undesired isoelectric precipitation at unfavourable pH conditions due to water dissociation at the membrane-solution interface and proton leakage through the AEM (Gomella, 1967; Roualdes et al., 2002). Furthermore, several food proteins tend to foul AEM since they are often negatively charged when the pH of the medium is close to neutrality, like in the case of most milk proteins (Cayot and Lorient, 1998). Although research has allowed the development of improved AEM with lower fouling tendency, applications are usually limited to desalination of brackish water (Lee et al., 2002; Kim et al., 2002; Park et al., 2003), where protein is considerably less abundant than in milk or whey products. Not many studies aimed specifically to investigate protein fouling of AEM during ED have been done, so research in order to better understand this type of fouling is justified.

This study was the complement to a previous work on CEM (Ayala-Bribiesca et al., submitted-b), where a deposit, presumably of a protein-nature was found on certain CEM, as well as a mineral fouling composed of calcium carbonate and/or hydroxide. Our objectives in the present work were to identify the nature and to characterize the morphology of AEM fouling during conventional electrodialysis demineralization of a whey protein isolate solution with different compositions in calcium and carbonate, using different pH conditions of the concentrate solution. The integrity of the membranes was analysed by membrane parameters and membrane surface elemental analysis and mapping.

Original AMX-SB membrane characteristics were measured on unused membranes freshly cut from the sheet and then compared to the membranes used for the demineralization procedure. Electrodialysis was carried out three times in 100-minute consecutive batches to ensure the fouling or not of the membranes. In order to maintain the system under the initial current density of 13mA/cm2, the voltage of the power supply (model HPD 60-5SX, Xantrex, Burnaby, BC, Canada) was adjusted accordingly at the beginning of each batch. The voltage necessary to stabilize such initial current density was then kept constant and the current was allowed to drop freely until the end of each demineralization batch (Ayala-Bribiesca et al., submitted-a). The concentrate solution pH value (pH(concentrate)) was maintained constant during the demineralization process at 2, 7 or 12 by the addition of HCl and/or NaOH. The diluate solution pH value was adjusted at 6.5 with HCl 0.1M. Each pH(concentrate) value was used for each mineral condition, where the model diluate solution (800 mg/l KCl and 32.67g/l WPI) was enriched with CaCl2 (800mg/l), Na2CO3 (1000mg/l) or CaCl2 + Na2CO3 (800 and 1000 mg/l, respectively) (Ayala-Bribiesca et al. submitted-a). After the three consecutive 100-minute demineralization batches the ED stack was dismounted and the anion-exchange membranes were photographed with a binocular microscope equipped with a digital camera. Membrane characteristics were determined by measuring electrical conductivity and thickness. Electron-microscopy images, coupled to X-ray elementary mapping, were taken from both sides of the membranes to complete the physical and chemical measurements. Attenuated Total Reflection-Fourier Transform Infrared (ATR-FTIR) spectra of the membrane surface in contact with the diluate solution were collected to detect the presence of a protein fouling.

Membrane electrical conductivity varied according to the pH(concentrate) and the mineral composition of the model diluate solution. In general, a slight rise in the electrical conductivity was found for the AEM, except for conditions at a pH(concentrate) of 2 with only CaCl2 added (4.46 ± 0.28 mS/cm) and Na2CO3 added alone (4.91 ± 0.87 mS/cm) which presented similar values to the control AEM (4.88 ± 0.04 ms/cm) and conditions with a pH(concentrate) of 7 and CaCl2 added, with or without Na2CO3, for which a conductivity reduction (3.29 ± 0.77 and 3.39 ± 1.02 mS/cm, respectively) was found. Conditions treated with a pH(concentrate) of 12 slightly incremented their conductivity in relation to the control (5.21 ± 0.13, 5.57 ± 0.21 and 5.75 ± 0.24 mS/cm, respectively for CaCl2, Na2CO3 and CaCl2 + Na2CO3 added), the same as conditions at a pH of 2 with CaCl2 + Na2CO3 added (6.00 ± 0.44 mS/cm) and at a pH of 7 with Na2CO3 added alone (5.55 ± 0.34 mS/cm). From these results, it can be seen that conductivity values did not show any particular tendency explaining the differences for the overall lot of membranes on the ions acting as counter-ions, as in our previous study on CEM (Ayala-Bribiesca et al., submitted-b). Differences could be mainly due to the presence and/or the nature (i.e. protein or mineral) of a heterogeneous fouling layer on the surface of the membrane and possibly the limited surface available for the analysis, as mentioned for membrane thickness.

Photos of AEM allowed distinguishing two different types of deposits. The control AEM (not shown) presented a clean and smooth texture, where the woven structure of the membrane was not easily visible. This was also the case for membranes from conditions at a pH(concentrate) of 12 with addition of Na2CO3 alone and CaCl2 + Na2CO3, for which no deposit was observed (Figure 22). In the case of the condition at a pH(concentrate) of 12 when CaCl2 was added alone, a mineral-like deposit was observed, as well as at a pH(concentrate) of 7 for the same mineral condition (Figure 22). No other mineral-like deposits were found, but a gel-like protein deposit was found on membranes from conditions with a pH(concentrate) of 2 and 7, regardless of the mineral condition (Figure 22).

Absorbance values obtained were very low (< 0.04), but were close to a ten-fold increase with respect to values obtained for CEM in our previous study (Ayala-Bribiesca et al., submitted-b). Results at the Amide I band region (Figure 23) confirmed the conditions for which a gel-like deposit had been found. All conditions where a pH(concentrate) of 12 was used presented the closest absorbance (from 0.006 to 0.014) to that of the control AEM (0.003). No deposit was observed on the diluate side these membranes, so the difference would be given by normal protein residues after the ED procedure, since no cleaning protocol was employed. On the other hand, conditions with a pH(concentrate) of 2 and 7 did present a fouling on the diluate side of the membrane, which was confirmed to be of a protein-nature, since absorbance at the amide I region between 0.021 and 0.036 were found for membranes from such conditions. Only condition with a pH(concentrate) of 2 and CaCl2 + Na2CO3 added presented an absorbance value of 0.015, which was low when compared to the other protein-fouled membranes. Nonetheless, an evident gel-like fouling had been previously observed on such membrane (Ayala-Bribiesca et al., submitted-a). The lower absorbance could be explained by the detachment, during previous analysis, of the protein layer originally present when the electrodialysis stack was dismounted, and to a low residue remaining on the membrane surface.

Amide II region results confirmed those obtained for the amide I region (Figure 23). A similar absorbance value was found for conditions with a pH(concentrate) of 12 (<0.003) to that of the control AEM (0.001), while conditions with pH(concentrate) of 2 and 7 yielded absorbances between 0.006 and 0.015, except for conditions at a pH(concentrate) of 2 and CaCl2 + Na2CO3 and at a pH(concentrate) of 7 with Na2CO3 added alone, which presented closer values (0.003 ) to the control AEM. Once again, these differences would be due to the extension of the fouling layer on the membrane and their surface ratio when covering the ATR-crystal. Nonetheless, FTIR succeeded in characterizing the protein-nature of the gel-like deposits observed on the optical microscopic photographs and as hypothesized in a previous study (Ayala-Bribiesca et al., submitted-a).

These results confirm the higher fouling tendency of AEM with respect to CEM, especially by protein. WPI is mainly composed of β-lactoglobulin and α-lactalbumin, which have respective isoelectric points of 5.2 and 4.8 (Cayot and Lorient, 1998). This implies they were negatively charged in the model diluate solution at a pH of 6.5, hence, attracted towards the AEM. The influence of pH(concentrate) on protein fouling has also been confirmed. These results comply with the phenomenon of H+ leakage migrating through the AEM, which occurs due to the presence of proton acceptors (e.g. water molecules and acidic species contained in the treated solutions) (Roualdes et al., 2002). H+ leakage had been previously considered as a cause for membrane fouling (Ayala-Bribiesca et al., submitted-a).

SEM images allowed observing the protein layer on this side of membranes. Membranes from conditions demineralized at a pH(concentrate) of 12 (Figure 24) were similar to the control AEM (not shown), except when only CaCl2 had been added, which presented traces of a deposit resembling a mineral fouling (Figure 24). Condition at pH(concentrate) of 7 where Na2CO3 was added alone also presented a comparable surface to the control membrane, although a protein fouling had been observed in a previous study (Ayala-Bribiesca et al., submitted-a). For all the other conditions demineralized at a pH(concentrate) of 7 and 2, the protein layer was evident (Figure 24), hence confirming the results obtained by ATR-FTIR.

Whatever the pH and salt conditions, results for potassium, phosphorous and calcium on the AEM surface in contact with the diluate presented similar values to those obtained for the control AEM (Table 9). The same case for sodium and sulphur occurred, except for membranes from conditions at a pH(concentrate) of 2 with CaCl2 + Na2CO3 added and Na2CO3 added alone, where a unexpectedly high relative concentration was found (Table 9). No particular explanation is given, apart the possible protein-origin of the sulphur detected, since whey proteins contain several cysteine groups (Cayot and Lorient, 1998). Chlorine results were comparable to the control AEM for all conditions at a pH(concentrate) of 12 and for condition at a pH(concentrate) of 7 with Na2CO3 added (Table 9). For the other conditions, chlorine relative concentration decreased due to the masking of the membrane surface by the protein fouling, as observed on the mapping image for condition at a pH of 2 and Na2CO3 added alone (Figure 25, column 1). In such image, the distinct composition of the protein fouling and the membrane was clearly shown, since chlorine constitutes an intrinsic component of the membranes hidden by the deposit on the membrane surface. Presence of oxygen on AEM surface in contact with the diluate clearly increased with respect to the control for membranes from all conditions (Table 9). Conditions presenting the highest oxygen concentration were those demineralized at a pH(concentrate) of 2 (18.1 ± 6.6 %) (Table 9), followed by those demineralized at a pH(concentrate) of 7 (9.7 ± 3.4 %) (Table 9). Although conditions with a pH(concentrate) of 12 also showed an increase in oxygen relative concentration (3.9 ± 0.3 %), the values were considerably lower in relation with the other conditions. This overall increase in oxygen concentration was particularly marked for conditions presenting a protein fouling, which supports the results obtained by ATR-FTIR. This is further supported by mapping images, showing the higher content of oxygen in the deposits, with respect to the non-fouled areas of the membrane (Figure 25, column 1). Carbon, on the other hand, did not show important variations (79.4 ± 2.4 %) with respect to the control AEM (83.49 %) for the overall lot of membranes. Mapping images for this element did, however, coincide with the results found for oxygen, as its higher concentration could be distinguished on the protein-fouled region (Figure 25, column 1). Finally, the deposit found on membrane from condition at a pH(concentrate) of 12 and CaCl2 added alone would be a protein residue on the membrane surface, since oxygen and carbon appeared to be the major constituents, and no calcium was present on the mapping images (Figure 25, column 2).

Mapping results for this side of the AEM were different to those obtained for the diluate side, clearly depicting the mineral nature of the fouling formed on the concentrate side of AEM (Table 9). As for the diluate side, potassium and sulphur were not present on the membranes, maintaining their similarity to the control. This would suggest the protein-origin of sulphur previously found on the diluate side of certain AEM. Sodium also was absent in the AEM surfaces, except for membranes from conditions at a pH(concentrate) of 7 and 12 when CaCl2 was added alone (1.28 % and 0.14 %, respectively) as well as condition at a pH(concentrate) 7 with CaCl2 + Na2CO3 added (1.05 %). As expected, the presence of monovalent cations was not abundant on AEM surfaces, given its affinity for anions. The presence of sodium on certain membranes would be due to its accumulation in the foulings, in the case of conditions with only CaCl2 added, or to a residual presence after the conductance analysis. Phosphorous, was similar (0.03 ± 0.02 %) to the control (0.04 %) for all conditions at a pH(concentrate) of 2 and all conditions with only Na2CO3 added. On the other hand, it did show interesting differences for the other membranes. Conditions at a pH(concentrate) 7 and 12 with only CaCl2 added presented a relatively high concentration of phosphorous (2.3 and 2.5 %, respectively). This element, present in the diluate solution as an intrinsic component of the WPI, would have precipitated as it entered concentrate stream, presumably as a form of calcium phosphate. These two membranes presented a mineral fouling, as previously discussed. The presence of a mineral fouling on the concentrate side of the membrane from condition at a pH(concentrate) of 12 and CaCl2 added alone, as observed by mapping images (Figure 25, column 4) and SEM (Figure 26), may have caused an increase on the electrical resistance of this membrane during the ED process. This would have caused the production of H+ from water dissociation (Simons, 1985). These electrogenerated H+ would be responsible for the presence of protein on this membrane, by precipitation of the latter. In the case of CaCl2 + Na2CO3, also at a pH(concentrate) of 7 and 12, higher amounts of phosphorous (0.23 and 0.47 %) than in the control were found. Although no mineral fouling or deposit on these membranes was detected, it could indicate the migration of phosphate across the AEM and/or a fouling in its first formation stages. Calcium was present in the same membranes where phosphorous was found, which would confirm the hypothesis of a calcium phosphate fouling, since it was not detected on the other membranes. A high concentration of calcium was found for both conditions with a mineral fouling (pH(concentrate) of 7 and 12 with CaCl2 added alone), and a moderately high concentration for conditions where a fouling would have begun its formation (pH(concentrate) 7 and 12 with CaCl2 + Na2CO3 added). The delayed formation of such fouling for the latter conditions would have been due to higher concentration in monovalent cations, specifically sodium, which allowed a delay in calcium ions migration from the diluate compartment through the CEM, as previously observed during the study of cation electromigration evolution (Ayala-Bribiesca et al., submitted-a). Once calcium ions began to cross the CEM, they would have rapidly mixed with the rest of the concentrate solution, eventually entering in contact with the AEM due to stream recirculation. Since phosphate ions would be simultaneously passing through the AEM, they would have precipitated as they entered in contact with calcium ions at the respective pH(concentrate) conditions. In the case of conditions with a visible mineral fouling, monovalent cations were scarce with respect to the equivalent conditions with CaCl2 + Na2CO3 added. Since calcium migration began earlier for all conditions without Na2CO3 added (i.e. with respect to those with Na2CO3 added) (Ayala-Bribiesca et al., submitted-a), a higher amount of calcium ions would have crossed the CEM, and, hence, entered in contact with the AEM. This would explain the presence of a more advanced mineral fouling for conditions without Na2CO3. These results are in accordance with those previously found during the mass balance for migration in our previous study (Ayala-Bribiesca et al. submitted-a), where calcium migration was delayed as long as monovalent ions were still abundant in the diluate solution. This is due to the slower mobility of calcium ions with respect to potassium and sodium ions (Pérez et al., 1994; Kabay et al., 2003). Chlorine results for the concentrate side of AEM (16.07 ± 1.09 %) did not show any difference with respect to the control (15.99 %), except for a slight decrease in membranes from conditions presenting a visible mineral fouling, (i.e. pH(concentrate) of 7 and 12 with only CaCl2 added, with 14.19 % and 13.72 %, respectively). This reduction would be due to the masking of the membrane surface by the fouling layer, as shown by mapping images (Figure 25, columns 3 and 4, respectively). In the case of oxygen content, the overall batch of membranes resulted in higher values with respect to the control membrane. Membranes with no sign of fouling (2.19 ± 0.33) were slightly above the control (0.78 %). Membranes from conditions with a pH(concentrate) of 7 and 12 with CaCl2 added alone presented 52.92 % and 46.77 %, respectively. Oxygen determination was useful in order to characterize the mineral deposits as a calcium phosphate, since it confirmed oxygen as a component of the fouling. This was supported by the mapping images, which show oxygen localization within the fouling structures concurring with phosphorous and calcium. In the case of conditions with a pH(concentrate) of 7 and 12 with CaCl2 + Na2CO3 added, no fouling was observed, although oxygen did follow the fouling tendency previously explained. Both membranes had an oxygen concentration between non-fouled and fouled membranes (23.46 % and 23.98 %, respectively), supporting the hypothesis of a fouling in its first stages of formation. Mapping images for these conditions were not included in this report since fouling was not visually apparent. Finally, carbon remained close (82.38 ± 0.70 %) to the value obtained for the control AEM (83.49 %), except for conditions with a visible mineral fouling (25.93 and 31.99 % respectively for conditions with a pH(concentrate) of 7 and 12 with CaCl2 added alone) and those with a fouling in its first stages of formation (56.99 and 58.85 %, respectively for pH(concentrate) 7 and 12 with CaCl2 + Na2CO3 added). The differences for carbon between fouled and non-fouled membranes clearly showed the masking effect of the fouling layer on the membrane and, when coupled to the mapping images, discard the possibility of a carbonate fouling on the AEM surface in contact with the concentrate. The possibility of a calcium hydroxide presence in the mineral fouling could not be discarded, since the EDS X-ray analysis is not suitable for hydrogen determination.

Higher magnification (x1500) images of mineral fouling showed some structural characteristics of the precipitates. AEM from conditions with CaCl2 added alone and with a pH(concentrate) of 7 in a cylindrical filament-shaped fouling (Figure 27). For the same mineral condition, but with a pH(concentrate) of 12, the mineral deposit had a crumbly and sponge texture formed by irregular aggregates without any apparent crystalline shape or pattern, suggesting the formation of a low solubility calcium phosphate form, such as hydroxyapatite, Ca5(PO4)3OH (Visser and Jeurnink, 1997) (Figure 28). A less drastic precipitation (i.e. at a pH of 7 versus 12) might have allowed the formation of a more organized structure, such as the filaments observed in contrast to the spongy aggregates.

These different forms might suggest different crystal growth conditions and, very likely, a difference in the mineral form of the calcium phosphate (Kim et al., 2005). According to these results, it appeared that protein fouling found on AEM in this experiment depended on the pH of the concentrate solution, regardless of the mineral composition of the diluate solution. Protein precipitation had place on the membrane surface in contact with the diluate solution. H+ leakage, attracted from the concentrate towards the cathode through the AEM, would have accounted for the formation of such protein fouling. This showed the important role of the pH conditions of the concentrate in the formation of this type of fouling. On the other hand, mineral fouling observed depended on the mineral composition of the concentrate solution, and the concentration evolution according to the ions traversing the opposite membrane; this is cation migration through the CEM. In this study, mineral fouling was due to the recirculation of the concentrate solution and the subsequent mixture of stock diluate solution with that enriched with the demineralized ions. Hence, calcium that had crossed the CEM eventually entered in contact with the AEM and precipitated in a form of calcium phosphate, which was present as an intrinsic constituent of the whey protein isolate. The presence of carbonate protected in a certain way the AEM from the formation of a mineral fouling during the ED process. Carbonate ionic forms would have, given their smaller size, demineralized before larger phosphate ions did. Hence, carbonate ions would have delayed the migration of phosphate ions, as in the case of calcium migration delay induced by sodium and potassium (i.e. smaller and monovalent ions with a higher mobility due to their hydration shell (Pérez et al., 1994; Kabay et al., 2003)). In the case of conditions where a mineral fouling was present, (i.e. with only CaCl2 at pH of 7 and 12), the absence of carbonate and the fast depletion of Cl- ions from the diluate solution due to their extremely high mobility would have forced phosphate ions to cross the AEM faster with respect to conditions with membranes showing no fouling. As it entered in contact with a less acidic pH at the membrane interface with the concentrate, it would have formed insoluble salts with calcium. Results indicating the beginning of a fouling on membranes with only Na2CO3 would be given by the depletion of carbonate ions, following that of Cl- ions which were less abundant for this mineral condition, but present anyway, given the presence of KCl (800 mg/l) in all diluate model solutions.

Further studies involving longer ED treatments following phosphate and carbonate migration would be useful in order to confirm the previous hypothesis. Moreover, other morphological analysis for conditions presenting a mineral fouling, like X-ray diffraction, would be required in order to identify the precise composition of the mineral fouling. Several forms of calcium phosphate exist according to pH value, concentration of the different ionic and non-ionic species and temperature conditions during precipitation. The presence of a hydroxide or its phosphate form as hydroxyapatite in the mineral fouling cannot be discarded/confirmed, since hydrogen cannot be determined by the technique employed. X-ray diffraction would also allow further characterization of the fouling and a better description of the conditions prevailing at the membrane-solution interface during its formation.

The present work completes the evaluation of the impact of the concentrate solution pH value on protein fouling and its relation with the presence of calcium and carbonate during ED treatment. It was shown that pH conditions above neutrality caused the formation of a mineral fouling on the concentrate side of AEM, identified as a calcium phosphate with or without calcium hydroxide. Calcium was present due to recirculation of the concentrate solution, which contained the cations that had crossed from the diluate compartment into the concentrate solution through the CEM. Furthermore demineralization procedures using a concentrate pH value of 2 or 7 caused the formation of a protein fouling on the AEM surface in contact with the diluate.

Although the fouling formed during the ED treatment should be cleaned by a cycle of current reversal (i.e. mineral fouling) or a cleaning-in-place (CIP) procedure (i.e. protein fouling), these cleaning procedures are not adapted to simultaneously remove both types of fouling, which would still imply repeated process interruptions. Industrial demands require long demineralization times without the formation of a significant fouling layer. Hence, conditions permitting the integrity preservation of membranes, like the use of a pH(concentrate) of 12 without recirculation of the concentrate or by adding a supplementary concentrate compartment, would reduce the frequency of such cleaning procedures and the eventual dismount of the ED stack for membrane replacement. This preventive measure would avoid both types of fouling described in this study, which would allow substantial benefits from membrane-life optimization.

© Erik Ayala Bribiesca, 2005