Effects of process conditions on organic fouling behaviour in electrodialysis

EFFECTS OF PROCESS

CONDITIONS ON ORGANIC

FOULING BEHAVIOUR IN

ELECTRODIALYSIS

Dries Flamée

Supervisor(s): prof. dr. ir. Ingmar Nopens, prof. dr. ir. Arne Verliefde

A dissertation submitted to Ghent University in partial fulfilment of the requirements for the degree of master in Master of Science in Biochemical Engineering Technology.

Universiteitsbibliotheek Gent, 2020.

This page is not available because it contains personal information.

Ghent University, Library, 2020.

DANKWOORD

De laatste dagen, weken en maanden stond enkel het afwerken van deze thesis cen-traal. Ook al ging dit gepaard met de nodige druk, toch kijk ik met veel trots terug op mijn thesis. Ik zie dit als de kers op de taart van de weg die ik als student heb afgelegd de afgelopen jaren. Het schrijven in het Engels en het leren werken met latex en python waren een zeer leerijke ervaring. Ook was het zeker en vast een speciale ervaring om een thesis af te werken in een lockdown.

Ik wil vooral Bram De Jaegher bedanken voor alle hulp tijdens mijn thesis. Hij was vanaf het begin zeer geïnteresseerd in dit onderzoek en uitte de nodige dosis positiviteit. Hierdoor kreeg ik veel motivatie, wat het labowerk en het schrijven zeker ten goede kwam. Door de goede communicatie kon ik steeds met vragen of andere problemen bij hem terecht voor een snel en duidelijk antwoord.

Ook wil ik alle PhD’s, labo assistenten en medethesisstudenten bedanken die mij geholpen hebben tijdens mijn praktisch werk in het labo.

PREAMBULE

Het doel van deze thesis is de invloed van verschillende proces parameters op de vervuiling van anion wisselende membranen te onderzoeken. De gekozen parameters zijn temperatuur, stroomsterkte, stroomsnelheid en membraantype. Deze worden elk individueel onderzocht om hun specifieke impact op de snelheid en intensiteit van deze vervuiling te bestuderen. Het onderzoek naar hoe deze parameters de vervuilingsgraad en snelheid ervan beïnvloeden is uitgevoerd en beschreven in deze thesis. Het achterhalen en verduidelijken van de effectieve oorzaak hiervan bleef echter wat achterwege door de stopzetting van de labo-proeven. Hierdoor zijn toch echter enkele proeven en analyses niet uitgevoerd die hieronder worden toegelicht. Tijdens het laatste vervuilingsexperiment (hoofdstuk 3.3) werden stalen genomen van diluaat, concentraat en de electrode spoeling voor een TOC analyse. Deze TOC analyse is echter niet volledig uitgevoerd voor het diluaat en electrode spoeling. Met deze waarden kon in functie van de tijd de concentratie aan SDS is de verschil-lende stromen aangetoond worden en een massabalans opgesteld worden. Deze had kunnen aantonen hoeveel SDS er na de test in elke stroom overbleef waaruit de mogelijkse hoeveelheid SDS die achterblijft in de membranen berekend worden. Dit zou mogelijks een mooie verduidelijking geven bij de permanente vervuiling die geob-serveerd wordt in de voorgaande proeven.

De bekomen TOC grafiek van het concentraat compartiment (figuur 3.17) toont aan dat SDS capabel is om door het anionwisselend membraan te migreren. Deze bekomen resultaten gaven op hun beurt aanleiding naar extra vervuilingstesten met andere or-ganische componenten gelijkaardig aan SDS, maar met een verschillende lengte van de koolstofketen. Hierdoor kon de impact van de lengte koolstofketen en molecu-ulgrootte op de vervuiling geanalyseerd worden. Er werden 2 nieuwe componenten besteld: Sodium octadecyl sulfate (18C) en Sodium octyl sulfate (8C). Deze zouden analoog gebruikt worden in vervuilingstesten waarvan dan opnieuw de TOC waarden gemeten worden. Deze konden dan vergeleken worden met de resultaten van SDS (12C) en zou aanvullende informatie geven over de impact van de molecuulgrootte.

CONTENTS

Dankwoord i

preambule iii

Contents vi

Nederlandse samenvatting vii

Summary ix 1 Literature study 1 1.1 Introduction . . . 1 1.2 Electrodialysis . . . 2 1.2.1 The stack . . . 3 1.2.2 Ion-exchange membranes . . . 5

1.2.3 Concentration polarisation and limiting current density . . . 6

1.2.4 Applications . . . 8 1.3 Fouling in ED . . . 9 1.3.1 Organic fouling . . . 10 1.3.2 Colloidal fouling . . . 12 1.3.3 Scaling . . . 13 1.3.4 Biofouling . . . 13

1.3.5 Characterisation and detection of fouling . . . 14

1.3.6 Cleaning of membranes . . . 15

1.3.7 Fouling prevention . . . 16

2 Materials & methods 19 2.1 Electrodialysis setup . . . 19

2.2 The stack . . . 21

2.3.2 Membranes . . . 22

2.3.3 Pumps & tubing . . . 22

2.4 Experiments and calculations . . . 23

2.4.1 Limiting current density . . . 23

2.4.2 Continuous fouling tests of SDS . . . 23

2.4.3 Batchwise fouling tests of SDS . . . 25

2.4.4 Analysed parameters . . . 26

2.4.5 Theoretical description of ED . . . 27

3 Results and discussion 31 3.1 Limiting current density . . . 31

3.2 Continuous fouling of SDS . . . 32

3.2.1 The effect of SDS concentration . . . 32

3.2.2 The effect of the crossflow velocity . . . 36

3.2.3 The effect of the current . . . 37

3.2.4 The effect of the membrane type . . . 38

3.2.5 pH and temperature . . . 39

3.3 Batch-wise fouling of SDS . . . 41

3.3.1 The effect of fouling in a batch process . . . 42

3.3.2 pH . . . 43 3.3.3 Conductivity . . . 44 3.3.4 TOC . . . 45 3.4 Theoretical description of ED . . . 47 4 Conclusion 51 Bibliography 53

SAMENVATTING

Desondanks de vele voordelen die elektrodialyse te bieden heeft, blijft het vervuilen van de ionuitwisselingsmembranen een groot probleem van deze technologie. Vervuilen kan omschreven worden als de ongewenste aanhechting van verscheidene compo-nenten aan deze membranen die voor een significante weerstandsverhoging zorgen van de membranen zelf en zo op hun beurt de efficiëntie van het elektrodialyse proces drastisch verlagen. Beter begrijpen hoe deze vervuiling tot stand komt en hoe deze kan voorkomen worden is van uiterst belang om deze techniek relevant te houden en ze te exploiteren naar andere industrieën.

In deze thesis werd het effect van verschillende factoren die het vervuilend gedrag van SDS beïnvloeden onderzocht. Er werd voor SDS gekozen omdat in de literatuur al meerdere malen aangetoond is dat SDS ionenuitwisselingsmembranen gemakke-lijk vervuilt. SDS-concentratie, crossflow, stroom en membraantype waren de vier gekozen parameters om te onderzoeken hoe ze elk het vervuilingsgedrag beïnvloe-den. Dit werd gevolgd door een batchgewijs experiment om de impact van een vervuilende component beter te bestuderen in een meer realistische opstelling. De conductiviteit, pH- en TOC-waarden werden gemeten om een beter inzicht te krijgen in het proces zelf. Finaal werd het batchgewijs experiment theoretisch benaderd. De resultaten hiervan kwamen vrij goed overeen met het echte experiment ondanks de vele onbekenden. Dit gaf aan dat de opstelling correct werkte.

De vervuilende experimenten toonden aan dat elke onderzochte factor een grote impact heeft op het gedrag van de vervuiling. Ook werd aangetoond dat SDS de anionenuitwisselingsmembranen permanent vervuilt. De TOC-metingen tijdens het batchgewijs experiment toonden aan dat SDS capabel is om door het anionen-uitwisselingsmembraan te migreren. Dit fenomeen was iets wat niet verwacht was aangezien SDS een relatief grote molecule is. Het meeste onderzoek naar SDS-vervuiling in elektrodialyse beschrijft de vorming van een ‘gel-laag’ op het oppervlak van het membraam, niet het doordringen of het permanent vervuilen hiervan. Dit was een zeer significante ondervinding in deze thesis.

SUMMARY

Despite the many advantages electrodialysis applications have to offer, one of the major setback of this technology is the fouling of the ion-exchange membranes. Fouling of the ion-exchange membranes can be described as the unwanted attachment of different components to the membranes which results in a significant decrease in the overall efficiency of the ED process. This reduces the capability to compete with other (membrane) separation techniques considerably. A better under-standing of the fouling process and how it can be prevented is of great importance to keep this technology viable and exploit it to other industries.

In this thesis, the effect of different parameters that influence the fouling behaviour of SDS are analysed in a continuous ED setup. SDS is chosen as a foulant because it is known to foul anion-exchange membranes quite easily and it is easy to work with. Foulant concentration, crossflow velocity, current and membrane characteristics are the parameters chosen to investigate their individual impact on fouling behaviour. This is followed by a batch-wise ED setup to further analyse the impact of the foulant on a more realistic desalination process. By measuring conductivity, pH and the TOC values during the process, more data is acquired to help shed light on the fouling be-haviour during desalination. Finally, the batch-wise experiment is theoretically recre-ated to attribute the different resistance components. This theoretical approach of the last experiment came quite close to the data acquired of the fouling test itself, indicating a good working stack.

The fouling experiments indicate that the analysed parameters all have a great im-pact on fouling and stack behaviour and thas SDS is capable of permanently fouling the anion exchange membrane. The TOC measurements during the batch-wise exper-iment show that SDS is capable of migrating through the anion-exchange membrane. This was something that was not expected from SDS as it is a quite large molecule. Furthermore, all research on SDS fouling of IEMs only describes the formation of a gel layer on the IEM surface, not the penetration and poisoning of the membranes, which is a significant finding of this thesis.

LIST OF FIGURES

1.1 conventional ED stack setup, adapted from (13). . . 2

1.2 schematic representation of the stucture of an electrodialysis stack (10). 4 1.3 standard reduction potentials of half-reactions. (16) . . . 5

1.4 illustration of the structure of a CEM (15). . . 6

1.5 graphical representation of concentration polarisation close to the sur-face of the IEM (adapted from (19)). . . 7

1.6 graphical representation of the limiting current density with; (I) - ohmic region, (II) - plateau region and (III) - over limiting current density region (adapted from (21)). . . 8

1.7 the structural formula of sodium dodecyl sulfate molecule. . . 10

1.8 the formation of a gel layer on the AEM, adapted from (38). . . 11

1.9 structure of a micelle (40). . . 11

1.10representation of an electrical double layer around a negatively charged particle (adapted from (43)). . . 13

1.11formation of a bio fouling layer on a surface (45). . . 14

1.12influence of fouling (scaling) on the resistance of IEM (33). . . 15

1.13by switching polarity, the electric field pulls the foulants away from the membrane. . . 16

2.1 experimental electrodialysis setup in the lab. . . 19

2.2 setup where concentrate and diluate are coming from the same container (continuous). . . 20

2.3 setup where concentrate and diluate are coming from a different con-tainer (batch). . . 20

2.5 side view of a gasket/spacer of a diluate compartimen that is placed

between membranes. . . 22

3.1 linear relation between the current and the potential. . . 32

3.2 fouling behaviour of different SDS concentrations. . . 33

3.3 relative potential increase of the different fouling curves compared to the baselines. . . 34

3.4 fouling of the first and second run of 1.00 g/l SDS. . . 34

3.5 reversals after fouling of the different SDS concentrations. . . 35

3.6 fouling of 0.50 g/l SDS at different flow rates. . . 36

3.7 reversals after fouling 0.50 g/l SDS at different flow rates. . . 37

3.8 fouling of 0.50 g/l SDS at different currents. . . 38

3.9 fouling of 0.50 g/l SDS on different membranes. . . 39

3.10temperature of diluate/concentrate during some fouling tests. . . 40

3.11pH of the diluate/concentrate during some fouling tests. . . 41

3.12batch fouling test of 0.75g/l SDS and its baseline. . . 42

3.13netto potential increase of 0.75 g/l SDS in a continious and batch setup. . 43

3.14pH progress during the baseline. . . 44

3.15pH progress during fouling. . . 44

3.16conductivity progress during the baseline and fouling. . . 45

3.17TOC values of the concentrate during fouling. . . 46

3.18evolution of NaCl concentration in concantrate/diluate reservoirs. . . 47

3.19comparison between the theoretical approach of the conductivity’s with temperature correction (Tcorr) and the aquired data of the conductivity during the test (dashed). . . 48

3.20representation of the individual calculated channel resistance over time and their summation (left) and the comparison between the theoretically approached stack resistance and the practical test (right). . . 49

LIST OF TABLES

2.1 overview of test parameters used during fouling and reversal cycles. . . . 24 2.2 different SDS concentrations where fouling was analysed. . . 24 3.1 overview of the potential and relative increase of the different fouling

LITERATURE STUDY

1.1

Introduction

The ever increasing pressure to decrease fossil fuel usage and the shift towards a complete environmentally friendly and circular economy is something industries can-not ignore. This shift is key to develop a sustainable industrial economy that is bio-based, energy independent and can effectively manage the greenhouse gas emis-sions. Those bio-based alternatives are mostly derived out of complex streams that are not yet suited for direct replacement in the respective application. Therefore cor-responding separation techniques are needed to isolate those valuable components to help this transition (1).

Membrane separation is one of those techniques attracting a lot of attention over the last couple of decades and continues to increase in popularity. They are being used to complement or replace the previous, mostly chemical, used technique like distil-lation or extraction. This all because of selectivity and efficiency regarding energy and the usage of chemicals (2; 3). Some prominent examples of membrane sepa-ration technologies are ultrafiltsepa-ration (UF), nanofiltsepa-ration (NF), reverse osmosis (RO) and electrodialysis (ED) (3). These are all membrane-based separations but differ in membrane structure, composition and working principle (4). UF, NF and RO are all pressure-driven processes, unlike ED that uses electricity as driving force.

ED is quite common in various food-based industries with applications in the produc-tion of drinking water, desalinaproduc-tion of cheese whey, deacidificaproduc-tion of fruit juices and in wastewater treatment (5). Next to this, other potential applications are exploit-ing ED such as the detoxification of hemicellulose hydrolysates (6) or the isolation of amino acids (7). However, for these applications in the bio-based industry, fouling of the ion-exchange membranes (IEM) tends to occur due to the presence of foulants in these process streams. Fouling leads to changes in membrane properties and can rapidly decrease the efficiency of the process and increase the overall operating costs of an ED setup. Fouling can render the process infeasible and is a major setback for applying ED in the bio-based industry. Research to better understand and remediate

this fouling is necessary to improve the process efficiency and make it possible to take full advantage of the potential of ED (8).

In this thesis, the organic fouling behaviour of IEMs is closely analysed by chang-ing parameters that are known to influence foulchang-ing behaviour to better understand their part. Foulant concentrations and composition, electrical current, flow rate and membrane composition are the main variables that are changed throughout the ex-periments. Before those practical experiments, a literature study is given to illustrate and explain the main factors considering ED technology.

1.2

Electrodialysis

ED is a technology that uses ion-exchange membranes (IEM) to separate ions from a liquid stream using an electrical field (9). It has been used for desalination of seawater and to remove heavy metals for over 60 years (10; 11). ED uses alternating anion-exchange membranes (AEM), cation-anion-exchange membranes (CEM) separated by flow spacers while an electrical potential is generated over two encapsulating electrodes to achieve the separation (12). The whole system of IEMs, spacers and electrode is called an ED stack. This is shown in figure 1.1. A variety of stack designs with different flow paths exist (10) but are not discussed here.

Alternating AEMs and CEMs create the diluate and concentrate compartments be-tween them. They are fixed bebe-tween the electrodes that provide an electric potential supplied by an external power supply. A conductive feed solution containing disso-ciated electrolytes is pumped to the membrane stack. After the application of an electric potential, ions start to migrate to the electrode of opposite charge as a result of the electric field and migrate through the membranes. This results in a formation of two different streams also shown in figure 1.1. One stream from which ions have been removed, the diluate stream, and one where they migrate to, the concentrate stream. This simple depleting and concentrating of ions in two separate streams is the basic working principle of every ED setup (14; 10).

1.2.1

The stack

The heart of an ED setup is the membrane stack as it is directly responsible for the actual separation of ions. A lot of research and development went into those mem-branes to enhance overall performance. They can be produced out of a variety of building materials like resins, zeolites or betonites. Sometimes a specific membrane composition or building material is required for the best results but overall the func-tion of AEM and CEM are the same no matter what structure or composifunc-tion is used (15). There exist other kinds of membranes like bipolar- and mosaic exchange mem-branes but those are not applicable here. Depending on the intended usage of the IEM, different properties are needed, like low electrical resistance, high permselectiv-ity and chemical or mechanical resistance (see chapter 1.2.2) (15).

In figure 1.2 a basic stack design is shown. Spacers, which are placed in the gasket between the membranes, separate the membranes and prevent any possible contact between IEMs. The gasket’s orientation ensures that the diluate, concentrate and electrode rinse do not mix and make sure that there is a watertight seal. The flow spacers mostly consist out of a plastic inert mesh that keeps the membranes from sticking together and keep a free path for liquid to flow. This mesh also creates tur-bulence on the surface of the membranes and enhances liquid uniformity throughout the stack that decreases concentration polarisation (see chapter 1.2.3) (10; 14). As seen in figure 1.1 and 1.2, the two outer compartments do not contain a diluate or a concentrate stream but an electrode rinse solution. This is a conductive salt solution separated from the diluate/concentrate solutions by the outer membranes. This solution plays an important role in the electrochemical reactions that occur at the electrode surface and carry the current to the diluate/concentrate channels. The choice of the electrode rinsing components determines which chemical species are formed at the electrode surface and is an important factor to avoid the corrosion of

Figure 1.2: schematic representation of the stucture of an electrodialysis stack (10). the electrodes and the production of toxic gases (10). To keep the electrical resistance of the stack low, the rinse solution has to be conductive. A simple NaCl solution suffices but chlorine gas will start forming on the anode which is corrosive, toxic and will shorten the lifespan of the electrodes (14; 15). Na2SO4 or NaNO3 are commonly

used for electrode rinsing.

All possible reactions that can occur on the surface of the electrodes (electrolysis) can be found in a redox potential table (figure 1.3), depending on the electrolytes present in the solutions. In a setup where NaCl is used as background electrolyte and Na2SO4

as electrode rinse the highlighted reactions can take place. Note that this states that two CEMs are placed on the outsides of the stack to prevent Cl- ions to migrate to the electrodes compartment.

The electrodes are capable of delivering a direct current over the stack creating the driving force of the separation. Formation of hydrogen gas on the cathode and oxygen gas on the anode is inevitable due to the presence of water. This gas will cause parts of the electrode to get isolated from the rinsing solution and will thereby increase the overall resistance, which in turn decreases the performance of the stack. Keeping the rinse solution at a decent flow rate will remove those gases and keep the electrodes cool (14; 17).

All previously-mentioned variables of the ED stack are crucial and have to work to-gether in perfect harmony to have a successful setup. Membranes, electrolyte com-position and concentration, pH, temperature, flow rate and current are the most im-portant factors to control and alter during ED. A variation in one of them can result in some serious changes throughout the process (12).

Figure 1.3: standard reduction potentials of half-reactions. (16)

1.2.2

Ion-exchange membranes

The most fundamental part of the ED stack are the IEMs. They are directly responsible for the separation of the charged species. A closer look at the structure of the IEM can give insight on how an ED setup is capable of selective separation.

Figure 1.4 shows the structure of a CEM. Anions are fixed on the polymer matrix of the membrane and are in equilibrium with mobile cations called counter-ions. Mo-bile anions are called co-ions and are far less present than counter-ions due to the electrostatic repulsion. Due to this structure and the lack of co-ions, a CEM is (al-most) only permeable for cations. In an AEM the fixed ions are positively charged giving the opposite effect than the CEM by only letting anions pass. “The selectivity of ion-exchange membranes results from the exclusion of co-ions from the membrane phase” (12). The characteristics of different IEM are determined by a couple of as-pects, such as polymer network density and its hydrophobic and hydrophilic proper-ties, the type and concentration of the fixed ions and heterogeneity and homogeneity of the membranes. All those different aspects of those IEM contribute to their overall capabilities (10; 18).

Figure 1.4: illustration of the structure of a CEM (15).

An important parameter of an IEM is its permselectivity. It can be seen as the ability to distinguish ions of opposite charge. When the permselectivity of an IEM is 100%, no co-ions migrate through the membrane and all the charge through the membrane is carried by the counter-ions. A high permselectivity is therefore highly desirable to obtain the best separation. High permeability, low electrical resistance and long-life capabilities are other main factors that are also desirable (12).

1.2.3

Concentration polarisation and limiting current density

In an ED setup, the total migration of ions from the diluate solution through the IEM to the concentrate solution is directly related to the current applied by the electrodes. Within the feed solution, the current is carried by both anions and cations but in the IEM the current is mostly only carried by the counter-ions as the passage of co-ions is restricted. Close to the surface of the membranes the overall ion concentration lowers in a diluate compartment and rises in a concentrate compartment due to the transport through the membranes. This phenomenon is called concentration polarisation and is schematically shown in figure 1.5 (19).

The natural diffusion of ions in a solution will automatically try to counter this concen-tration gradient. A lower electrical current and a higher ion concenconcen-tration in the bulk solution will also reduce the concentration polarization as fewer ions are removed on one side of the membrane and more are present in the bulk solution (20).

Because of this concentration polarisation, a lack of counter-ions to transport through the membranes can occur. This causes the overall resistance of the stack to rise and

Figure 1.5: graphical representation of concentration polarisation close to the surface of the IEM (adapted from (19)).

will result in a higher potential to maintain the same ion flux. This higher potential leads to even greater polarization and potential increase and so forth until a critical point is reached, this is called the limiting current density (m) and is shown on figure

1.6. At this moment the concentration of counter-ions tends to 0 at the surface of the membranes resulting in not enough ions to carry the current. At this point, a fur-ther current increase will result in a great potential increase and an almost unaltered current density or ion transport because there are far fewer ions to carry the current (plateau) (21).

When increasing the electric potential even further than this plateau (over limiting current density), two phenomena take place that leads to an increase in current-carrying capacity. (1) Water dissociation reactions start to occur to produce extra ions (H+ _{and OH}-_{) that help carry the excess current through the membranes. (2) Fluid}

vortices form in the diffusion boundary layer that disrupts the boundary layer and electrolytes are carried from the bulk fluid to the membrane, leading to an increase in current carriers. The effect of water splitting is often not desired as this will influence the pH and it is a fraction of the current (cost) that is not being used for separation. An optimal ED setup should be operated in the ohmic zone region where an increase in potential results in a linear increase in ion transport, below the limiting current density (22; 12; 20; 23).

Figure 1.6: graphical representation of the limiting current density with; (I) - ohmic region, (II) - plateau region and (III) - over limiting current density region (adapted from (21)).

1.2.4

Applications

The production of potable water is still the main application of ED technology for 60 years due to its efficiency within a certain concentration range. Since only 2.5- 3.5% of the water on earth is freshwater, and 0.3 – 0.8% of this fraction is available to us (13), an efficient technology for continuous production of potable water is still needed to provide enough water for the population. ED competes directly with reverse os-mosis for desalinating brackish water but it has been shown that in between certain limits of salt concentrations, ED is more economically feasible than other desalination techniques as it is not limited by osmotic pressure (18). ED is not only used to make potable water but also to produce a lot of industrial process water that can be used in different applications. Some examples are desalinating cleaning water for re-use as boiler feed water. Also in different food industries, ED has found its way: desalination of cheese way or skimmed milk, stabilisation of wine or deacidification fruit juices are common practices where it can be found (10). Even industrial wastewater is being treated with ED to remove salt or heavy metals.

All of the applications above demand for ions to be removed from a stream that is then ready for application or consumption, but streams can also be concentrated to obtain high purity and concentration of various products. Some examples are the pre-concentration of salts such as NaCl, KBr and KI from wastewater streams (13; 11). A pre-concentration stream to eventually produce table salt (NaCl) can be achieved by

ED. It is superior to evaporation as it does not require as much effort and energy and is also produced when seawater is treated to produce potable water (24; 25; 5; 10). Over the last decades, ED has come a long way and evolved from simple water purifi-cation related processes to more complex applipurifi-cations mentioned above. Continuous research and development can help to keep expanding this technology to other areas and enhance the current operations (3; 18; 13; 10).

1.3

Fouling in ED

With all the possible applications and benefits ED has to offer, there are still some ma-jor downsides and problems. A problem that is pretty common and very well known in membrane separation technology is membrane fouling (10; 26). Fouling has a big influence on the membrane properties and negatively influences the separation pro-cess. “Fouling of IEMs can be described as the unwanted attachment of different components to the IEMs which results in a significant decrease in the overall effi-ciency of the ED process” (26). These foulants accumulate on the surface of the IEM or enter the membranes, hereby hindering the transport of ions through these IEMs resulting in a significant increase of membrane resistance and a decrease in ion-flux (27; 28). AEMs are more sensitive to fouling compared to CEMs. This is easily ex-plained because the most common foulants in wastewater and other watery streams are negatively charged (29). However, there are also cationic components that foul the CEM but these are far less common (28).

The increase in resistance of the IEM caused by fouling is one of the reasons why the cost of using and operating ED setups increases. This is because an increasing resistance will require a higher electrical potential to maintain the same ion flux during the process. If the potential is unaltered and maintained at a constant value, the current that goes through the membranes, directly impacting the flux of ions, will decrease in time. Membrane degradation because of fouling is a major factor to consider. When performing a purification or separation with ED, foulants will settle in or on the membranes until the membranes eventually come to a point where they are too polluted to function correctly and have to be cleaned or even replaced. Cleaning membranes can be a quick and easy solution if the foulants are not irreversibly bound to the membrane to mostly regain its original properties. If they are irreversible, they will have to be replaced after a certain amount of time. Downtime, cleaning and buying new membranes can easily make up to 50% of the production cost of the desired product (30), thus extending the lifespan and reducing the frequency

of cleaning those membranes is definitely of great importance to keep ED a viable separation technique (30).

Fouling components can be classified into several groups, namely organic fouling, col-loidal fouling, scaling, membrane poisoning and biofouling (26). These will be further elaborated upon along with the detection methods and potential fouling remediation techniques. Each fouling type has its negative influence on the ED process and as of now, there are a couple of techniques that can be applied to prevent or reduce fouling as much as possible.

1.3.1

Organic fouling

Fouling of IEMs by organic foulants is an important research topic because a lot of pro-cess streams that are treated contain some kind of organic substance, even surface waters or natural streams (31; 29). If those foulants are organic and soluble, this foul-ing type is called organic foulfoul-ing. Organic foulants are mostly negatively charged in nature and therefore foul the AEMs (29; 32). Some examples are proteins (33), humic acid (34), natural organic matter (NOM) (28), surfactants, amino acids (35), sugars (26) and organic acids (36; 29). Those organic compounds tend to stick to the mem-brane and/or migrate into the memmem-brane and severely increase the resistance of the IEM even at very low concentrations. The interaction of the organic component with the membranes is mostly determined by the combination of electrostatic interactions and affinity interactions with the used membranes (26).

A well-known studied organic fouling component is sodium dodecyl sulfate (SDS). SDS is a surface-active agent or surfactant that is frequently used in biochemistry and microbiology to disrupt cell membranes, emulsify fats and solubilize proteins (37). SDS has a characteristic 12C long aliphatic chain and a hydrophilic head (S-atom). Because of this bipolarity, it is still soluble in water at room temperature and can, therefore, solubilize proteins, fats and disrupts membranes (37).

In solution, SDS dissociates leaving the aliphatic chain negatively charged. The chain is therefore affected by the electrical field applied in ED setups. When performing ED the negative aliphatic chain will migrate in the direction of the anode, towards the AEM. The presence of SDS during ED is challenging as it has a strong affinity for the AEM due to the long aliphatic chain and the electrical charge (29). It is deemed that the large size of the aliphatic chain blocks SDS from migrating through the AEM. As a consequence, SDS accumulates at the membrane surface which results in a gel-fouling layer forming on the surface of the membrane which is illustrated by figure 1.8 (26; 28).

Figure 1.8: the formation of a gel layer on the AEM, adapted from (38).

Another factor to consider when working with surfactants is the critical micelle con-centration (CMC). When surpassing this CMC in a polar solvent, surfactants will start to self-associate and form micelles (39). The hydrophobic aliphatic chains will direct themselves to each other in the middle of the micelle, away from the solvent. The polar ‘heads’ will face outwards creating a round micelle (see figure 1.9) (40). The higher the concentration the more micelles there will be, but there will always be a mixture of free surfactants and micelles (39).

The formation of a gel layer and micelle formation caused by SDS molecules at the AEM are possible reasons for the considerable fouling potential of SDS. Because of all those aspects, SDS is frequently used to analyse fouling behaviour (34; 41; 40).

1.3.2

Colloidal fouling

Colloidal foulants are finely dispersed solids that are not dissolved and are present in a lot of natural streams and vary in size from 1 nm to 2 μm. Some examples are metal oxides like iron, aluminium and manganese inorganic colloids such as silica, large or-ganic molecules and smaller molecules in aggregations (26; 42). In contrast to oror-ganic fouling, colloidal foulants are not dissolved and exhibit different fouling behaviour. Or-ganic foulants are classified as orOr-ganic particles that are charged and are quite big in size relative to other molecules in solution, but as the organic structure gets larger and larger, solubility decreases and the organic compound aggregates to a colloid. It will adopt more properties of colloidal substances and will, as a result, behave as such rather than as organic foulants (26). SDS is a good example of this transition from organic to colloidal fouling behaviour. As mentioned above (chapter 1.3.1) if the concentration of SDS surpasses the CMC, micelles (aggregates of molecules) will start to form and tend to adopt colloidal foulant properties. The defined line between col-loidal and organic compounds is not completely clear but is maybe better described as a transition area. Most of the research concerning IEM fouling focuses on organic or colloidal fouling region (31).

The main property of colloidal particles is that they are characterised by a high net surface charge and are therefore surrounded by an electric double layer. This elec-tric double layer consists of two layers, an inner and outer layer. The inner layer (Stern Layer) solely contains immobile counter-ions and neutralises a part of the ini-tial charge of the particle. The outer layer (diffusion layer) counters the rest of the charge with a mobile layer with an elevated concentration of counter-ions. The elec-tric double layer prevents colloidal coagulation (43; 26).

Until a certain distance into the diffusion layer, the ions can be considered bound to the colloid. A slipping plane occurs in the diffusion layer where the colloid and the charges encapsulated by the slipping plane, determine the net charge of the colloid. Because of this charge, the colloid will be affected by the electrical field and can easily foul the IEM because of their overall larger size (26). Humic acid is an example of a commonly researched colloidal foulant. Note that this is also described as an organic foulant as research found that it has colloidal as well as organic fouling behaviour (28).

Figure 1.10: representation of an electrical double layer around a negatively charged particle (adapted from (43)).

1.3.3

Scaling

Scaling is a type of fouling that occurs when salts, present in the solution, precipitate on the membranes. Some examples of those salts are CaCl2, MgCl2, barium salts,

bi-carbonates, and sulfates (26; 44). This scaling can occur when the salt concentration rises above solubility concentration and starts to precipitate. This is facilitated close to the surface of the IEM because of a few phenomena that contribute to this scaling. Concentration polarization at the surface of the membranes can lead to higher local salt concentrations. A pH gradient can occur because of the transport and generation of H+ or OH- ions through/in the membrane. This pH change has a direct influence on the solubility of the salts. Working with an adequate pH and a reasonable con-centration of your diluate and concentrate streams can partially suppress this effect (26).

1.3.4

Biofouling

Another type of membrane fouling is biofouling. It is the least researched fouling type due to its time scale (time needed to grow this biofouling film by the micro-organisms) as it is mostly outweighed by other ‘faster’ fouling components. It does still occur in some setups and is an important factor to take into account for specific process streams. It is a problem commonly encountered when working with mem-brane filtration or reverse osmosis (26).

Biofouling in ED can best be described as the unwanted growth of micro-organisms (MO) on the surface of the IEM. This biofilm becomes a physical barrier for ions to pass

and therefore decreases the efficiency of an ED setup. The formation of this biofilm can be shortly described in 5 stages and is illustrated in figure 1.11.

Figure 1.11: formation of a bio fouling layer on a surface (45).

Free dispersed MOs attach to the membrane through a sticky mix, the exopolysac-charide layer (EPS). It is produced by the MOs themselves and contains sacexopolysac-charides, proteins and nucleic acids (45). When one MO is attached it starts to grow and di-vide from an increasing lump of cells until a mature biofilm. From this mature biofilm, some MOs detach and can start a new biofilm layer elsewhere (45).

Removing or lowering key nutrients from the solution that is being treated or adding biocides are some easy solutions and will decrease the growth of those organisms and therefore the biofouling. In case of a solution where it is not possible to add or remove components then increasing flowrate could prevent bacteria from sticking to the membranes caused by the turbulent streams at the membrane surface (26).

1.3.5

Characterisation and detection of fouling

Because of the wide variety of potential foulants that occur in different operational conditions, a lot of ED setups are being influenced by fouling. Predicting the fouling rate or the fouling potential of certain process streams is a key element in this branch of research and requires knowledge on the composition of the solution and the state of the ED setup which need to be monitored (26).

An easy way to know that some kind of fouling is occurring in a stack, without dis-mantling or visually inspecting it, is the increase in overall resistance. Foulants collect on the surface of the IEM and block the easy passage of ions or other charged soluble components. This blockage directly results in a decreased flux and account for this increase in resistivity seen in figure 1.12 (46). As mentioned in the theoretical chapter of ED, a higher resistance of the membranes will lead to greater power consumption

and a decrease of efficiency. Unfortunately, it hardly gives information about the na-ture of fouling the system is dealing with. Increasing the electrical field that is being applied can be a solution to sustain a constant ion flux through the membranes (10).

Figure 1.12: influence of fouling (scaling) on the resistance of IEM (33).

To obtain more detailed information on the fouling layer composition and structure the stack can be disassembled to visualise the fouling layer and characterise by a multitude of visualisation techniques like (electron) microscopy and photo imaging (26). Note that disassembling an ED stack can be very time-consuming.

1.3.6

Cleaning of membranes

In most cases, a considerable amount of fouling can be removed from the mem-branes by cleaning them. This is possible because a lot of components are reversibly bond with the membrane surface and can be removed with some effort. Cleaning membranes can be quite a time-consuming and expensive procedure but is still con-sidered valuable as continuously replacing is even more expensive. This cleaning can be done physically, chemically or by reversing the polarity of the stack (26; 47). Changing the polarity of the stack can be a quick and easy solution to discard foulants from the IEM. Reversing the direction of the electrical field (polarity) will cause a coulomb-force in the opposite direction, away from the membrane surface (see figure 1.13). Because of this change, some foulants that are stuck on the membrane and are not irreversibly bound to it can detach. The main disadvantage of ED reversal is the downtime for the system while it is ‘reversing’. Modern stacks can switch the diluate and concentrate streams depending on the polarity which results in a continuous flow with no downtime or contamination of the purified stream (10). Each switch of the

electrical field can be executed preventively every couple of minutes to hours instead of switching when resistance surpasses a certain level (26). Reversing the polarity can be somewhat compared to backwashing a normal membrane or sand filter.

Figure 1.13: by switching polarity, the electric field pulls the foulants away from the membrane.

Mechanical cleaning is also a valid option but is used less frequently because of the fragile membranes. Vibration, air sparging or ultrasound can alter the membrane properties and is, therefore, a lesser-used method of membrane cleaning. Studies show that in some cases ultrasound can be a very helpful fouling remediation tech-nique (48). Membranes can also be chemically cleaned by various agents ranging from strong acids to enzymes, EDTA, detergents and disinfectants (48). When using some of those agents it has to be considered that they can react with the membranes and decrease their performance, like alkali solutions can degrade the ion-exchange groups (26).

1.3.7

Fouling prevention

It makes sense to use membranes as long as possible before replacing them with new ones to reduce the running costs significantly. To postpone this membrane degrada-tion, a couple of things can be done to prevent this fouling from happening in the first place. Solution pre-treatment, process conditions optimisation, turbulent cell com-partments, alternating polarity and membrane properties are the most commonly used methods (48).

Pre-treating the solution to remove the most disturbing foulants before it enters the ED setup can be quite easy if the composition of the stream and foulant is known. The methods most commonly used for this are other membrane filtration techniques like an ultra- or nano-filtration, centrifugation or activated carbon filters (26). Optimizing the process conditions like temperature and pH flow rate can also make a significant

difference (see scaling). It adds an extra expense to the operation cost but can still be viable overall.

MATERIALS & METHODS

2.1

Electrodialysis setup

In figure 2.1 the setup used for all experiments is shown. Consisting from left to right out of pumps and their respective tubing, pH/temperature/conductivity meter, concentrate/diluate/electrode rinse containers and the stack itself.

Figure 2.1: experimental electrodialysis setup in the lab.

Figure 2.2 and 2.3 show a schematic representation of the ED stack in two configu-rations to better illustrate the flow paths of the different streams. Whilst performing experiments 2 different setups are used, one where the concentrate and diluate are coming from the same container and are mixed (figure 2.2). This setup tries to rep-resent a continuous process where the liquid going in has a constant concentration during the experiment. The second setup is one where diluate and concentrate are separated (figure 2.3) to simulate a batchwise setup. The explanation of experiments further on will always mention which setup was used.

Figure 2.2: setup where concentrate and diluate are coming from the same container (continuous).

Figure 2.3: setup where concentrate and diluate are coming from a different container (batch).

2.2

The stack

The two outsides of the stack are made out of chemical inert plastic and house the electrodes. The outsides are bolted together squeezing the membranes, gaskets and spacers tightly together. Two CEM on the outside and an AEM on the inside were always used to create one concentrate channel and one diluate channel. The stack itself is placed at a slight angle to make sure no air is trapped inside. In figure 2.4, a top view of the stack is shown to give a better insight into the composition of the stack.

Figure 2.4: schematic top-view of the electrodialysis stack.

The gaskets/spacers used to separate concentrate, diluate and electrode rinse are illustrated in figure 2.5. It gives the dimensions of the spacers used from a side-view perspective of the stack. On the left and right-hand side, 2 square cutouts can be seen. They are the main channels where the diluate and concentrate are pumped into. From here on out the diluate or concentrate (depending on the orientation of the spacer) flows in the centre cutout through the tiny holes in the gaskets. By al-tering the orientation of this spacer each time, alternating diluate and concentrate compartments are created. (spacers are not shown in figure 2.4).

The centre cutout is the same size as the electrodes of the stack giving them an effec-tive surface of 200 cm2. The distance between membranes is equal to the thickness of the spacer 3 mm (not shown in image 2.5) making the volume between two mem-branes 60 cm3 or 60 ml. Inside the centre cut-out, a spacer mesh is placed to ensure good mixing of the solution within the stack.

Figure 2.5: side view of a gasket/spacer of a diluate compartimen that is placed be-tween membranes.

2.3

Equipment

2.3.1

Potentiostat

To deliver direct current over the stack an automated potentiostat is used (Biologic VSP - France). It is capable of delivering a constant direct current to 12V or 5A. All parameters can be set in the respective computer program (EC-lab) that will automat-ically perform the pre-programmed test. No human interactions were needed during any test.

2.3.2

Membranes

In all the experiments three kinds of membranes were used: ˆ PCA-SK (CEM)

ˆ PCA-SA (AEM)

ˆ Fujifilm Type 2 (AEM and CEM)

PCA membranes are homogeneous membranes whereas the Fujifilm membranes are heterogeneous. The membranes were cut by hand from big sheets at the size of the outer edge of the spacer/gasket (550 mm by 120 mm). Before using them in experiments, they are prepared by soaking them for 24 hours in a similar liquid that will be used in the experiments. Here, this was 0.1 M NaCl.

2.3.3

Pumps & tubing

To maintain a constant flow rate throughout the stack peristaltic pressure pumps were used. Watson-Marlow pumps with complementary tubing were able to deliver 530 ml/min at 220 rpm (max RPM).

2.4

Experiments and calculations

2.4.1

Limiting current density

First of all, the limiting current density was determined. Knowing this LCD value is crucial for future fouling tests because below this value the ED stack works efficiently as all current delivered by the potentiostat is used for ion migration and not for water splitting for example. Surpassing this value would be undesirable and would make future ion transport calculations a lot more difficult. The LCD was determined with PCA membranes at a flow rate of 530ml/min and a concentration of 0.1 M NaCl for concentrate and diluate and 0.1 M Na2SO4 for electrode rinse. Those parameters

are chosen as they are very similar to future fouling tests. The LCD was determined by applying a constant potential for two minutes. The last 30 seconds of those two minutes were used to take an average value of the current measured over the stack. Between 0 to 9V at every interval of 200 mV, the current was measured and plotted to visually determine the LCD.

2.4.2

Continuous fouling tests of SDS

To analyse the fouling behaviour of SDS at different process conditions, multiple ex-periments were performed with different SDS concentration, current, flow rate and membranes. A basic fouling test consists out of applying a constant current and mon-itoring the potential over time. After this fouling test, the diluate and concentrate compartment are rinsed with 1l 0.1 M NaCl and a reversal without SDS is performed and also monitored to analyse if the fouling of SDS is (ir)reversible. For all the foul-ing tests the setup where diluate and concentrate are mixed is used (see figure 2.2). This was primarily opted to eliminate the potential (voltage) increase that would be caused by the desalination of the diluate compartment. By mixing both canals, the lit-tle desalination of the diluate after one passage through the stack was compensated with the concentrated stream by mixing them back together resulting in a constant salt concentration during the whole fouling test. For each set of fouling and reversal, a new AEM was installed to assure identical setups with minimal changes and varia-tions. Constant parameters of those fouling and reversal experiments are shown in table 2.1. Varying parameters will be specified for each test separately. Note that dilu-ate and concentrdilu-ate composition is the solution where SDS is added and all reversals are always the same.

Table 2.1: overview of test parameters used during fouling and reversal cycles.

Parameter Fouling Reversal

Dil./conc. composition 0.1 M NaCl 0.1 M NaCl

Dil./conc. volume 2 L 2 L

Dil./conc. Flow rate variable 530 ml/min Elec. Rinse composition 0.1 M Na2SO4 0.1 M Na2SO4

Elec. Rinse volume 1 L 1 L

Elec. Rinse flow rate 530 ml/min 530 ml/min

Duration 180 min 30 min

Current variable -1 A

SDS concentration variable none Membranes variable Identical to fouling

The effect of SDS concentration

First, the fouling of 5 different SDS concentrations was analysed by adding a certain amount of SDS (shown in table 2.2) in the diluate/concentrate container. During those fouling tests, pumps were all running at 530 ml/min. The constant current applied for fouling was 1A and PCA-SA and PCA-SK membranes were used. A baseline (blank) experiment was also performed for comparison.

Table 2.2: different SDS concentrations where fouling was analysed. Fouling concentrations of SDS Baseline 0.50 g/l 0.75 g/l 0.82 g/l 0.92 g/l 1.00 g/l

After each fouling concentration, the stack was rinsed and the process was followed by a reversal.

The effect of flow rate

All pumps normally operate at maximum RPM (530 ml/min). To evaluate fouling be-haviour at a lower flow rate the pump of the diluate and concentrate was lowered to 75% (397.5 ml/min). The pumps of the electrode rinse remained unaltered and stayed at 530 ml/min. Diluate and concentrate were coming from the same 2 L con-tainer containing 0.5 g/l SDS. The constant current applied for fouling was again 1 A. After fouling, the diluate & concentrate compartments were again rinsed with 1 L 0.1 M NaCl and a 30-minute reversal was performed.

The effect of current

To evaluate the influence of the current on fouling, the constant current during fouling was increased to 1.5 A instead of the usual 1 A. Pumps were all operating at maximum RPM (530ml/min). Diluate and concentrate were coming from the same 2 L container containing 0.75 g/l SDS. After fouling, the diluate and concentrate compartment were again rinsed with 1 L 0.1 M NaCl and a 30-minute reversal was performed.

The effect of membranes

In this test, only the homogeneous PCA-AEM was changed with a heterogeneous Fuji-film Type two AEM. Diluate and concentrate were coming from the same 2 L container containing 0.75 g/l SDS. The constant current applied for fouling and reversal were again 1 A and -1 A respectively. After fouling, the diluate and concentrate compart-ment were again rinsed with 1 L 0.1 M NaCl and a 30-minute reversal was performed.

2.4.3

Batchwise fouling tests of SDS

To further analyse fouling behaviour and the impact of this fouling of SDS, the diluate and concentrate compartments were separated to achieve real concentration and dilution, as it gives a better comparison to real batchwise industrial ED setups (see figure 2.3 in chapter 2.1 for the schematic overview). In the following tests, the goal was to analyse the influence of SDS on ED separation like it would be, more or less, in a practical ED setup. Tracking the potential over the stack and measuring the pH, conductivity and TOC values of diluate, concentrate and electrode rinse during those fouling tests gives a lot of useful information to better understand the impact and the progression of this fouling throughout the experiment. First of all, a baseline was established to have an idea of the normal behaviour of the stack when there is no foulant present in a batch-wise setup. The concentrate and dilute composition both consisted out 1 L of 0.1 M NaCl, the electrode rinse remained the same as in the previous experiments with 1 L of 0.1 M Na2SO4. 1A was applied over the stack

for three hours with all pumps running at a flow rate of 530 ml/min. All membranes were PCA membranes. After this baseline, the same test was performed but with SDS added only to the dilute compartment at the beginning of the test (0.75 g/l). All measured parameters remained the same.

2.4.4

Analysed parameters

Different parameters were used to analyse the fouling behaviour of SDS during differ-ent experimdiffer-ents. The equipmdiffer-ent used to measure or analyse those differdiffer-ent compo-nents are summarised here.

Potentiometry

To visualise the fouling of the SDS a constant current was applied over the stack with a potentiostat. The resulting potential was plotted overtime during the fouling and reversal tests. The potential increase gives a good real-time value that is influenced by fouling.

Conductivity

Measuring the conductivity of the different liquid streams can give insight into the migration and concentrations of ions during the separation. Thus measuring conduc-tivity during fouling tests is also a solid value to take into account. A conducconduc-tivity probe (Consort) was used and calibrated with three solutions KCL of 1, 0.1 and 0.01 M before usage.

pH - temperature

Tracking the pH of the different streams during fouling can give additional information about the process. Measuring the temperature is of great importance to conductivity and pH values as it is influenced quite significantly by it. Both pH and temperature were measured by a HANNA – HI 5522. The device was calibrated daily with stock buffers.

Total Organic Carbon (TOC) – mass balance

To track the migration of SDS molecules throughout the stack, TOC measurements were taken during fouling (Shimadzu Toc-V series TOC - Japan). They give good infor-mation on how many SDS molecules migrate through or potentially get stuck inside the membrane during fouling. When taken samples of all possible streams a mass balance can be achieved that gives insight into the migration of the SDS molecules during the fouling experiments. Vials that were used to collect samples were baked for five hours at 500 °C to remove any residual carbon.

2.4.5

Theoretical description of ED

With the help of some basic equations, it is possible to predict the behaviour of the concentration and conductivity of all the different streams in an ED setup during de-salination. This can be an interesting thing to do to check if the ED setup works properly and does not show any major deviations compared to what can be expected. Here the batch desalination of NaCl is replicated that is performed as a baseline in section 2.4.3.

Simulation of ion concentration

First the transport of ions (flux) during the test is calculated with equation 2.1:

Jm=

ε 

F (2.1)

with Jm the flux of salt (mol/s), the current efficiency ε (-),  the electric current (A or

C/s) and F Faraday’s constant (C/mol).

As F is a constant and A is equal to 1 during the experiment, only ε remains unidenti-fied. This is a value between 0 and 1 but is not exactly known as it depends on a lot of different factors. Here ε is chosen to be 0.985 as it gives the best fit to the acquired data.

This ion flux is used to calculate the time-evolution of the ion concentration in the reservoirs. It is assumed that the reservoirs are continiously stirred tanks. A mass-balance over the reservoirs leads to the following equation,

dC dt =

ε 

F V (2.2)

With C the concentration of a reservoir (mol/m3_{) and V the volume of that reservoir}

(m3_{). Integration of this equation leads to,}

Ct= C0− t

ε

F V (2.3)

This equation can be plotted overtime to give the ion concentration of the diluate and concentrate. The electrode rinse is assumed to stay at a constant ion concentration.

Simulation of the conductivity

Assuming a constant molar conductivity of NaCl and Na2SO4, which is not entirely

true (49), the conductivity κ is computed by plugging the concentration arrays into the conductivity-method,

κ= Λmc (2.4)

with Λm the molar conductivity of a certain electrolyte

Simulation of the conductivity with tempearture-dependance

The conductivity depends on the temperature and by applying a non-linear tempera-ture compensation to the conductivity prediction one can greatly improve the accu-racy of the prediction. To take into effect the temperature evolution a polynomial is fit to the data of a typical temperature profile during the ED experiments.

The non-linear temperature corection from (50) is used:

Λm,T= (Λm,25− 0.0545) (1 + 0.02 (T − 25)) + Λ,T (2.5)

with Λm,T the molar conductivity of the electrolyte at temperature T (◦C), Λm,25 the

conductivity at 25◦C (tables) and Λ,T the conductivity of pure water at temperature

T_{. For the latter the following correlation with the temperature is used:}

Λ,T = 0.0545 0.55 e0.0363 T− 0.356

(2.6) The constant molar conductivity in equation 2.5 can be replaced with the temperature corrected conductivity to include the effect of the temperature during the desalination process.

Simulation of stack resistance

Now that the conductivity can be simulated in the ED system, The final step is to simulate the total stack resistance Rt (the stack has one cell pair).

Rt= Rer+ 2 Re+ UD+ Rc+ Rd+ 2 RCEM+ RAEM

with:

ˆ Rer the resistance as a result of electrode reactions and other mass transfer

phenomena at the electrode surface,

ˆ Re the electrolytic resistance of the electrode rinsing solution,

ˆ UDthe Donnan potential, but the net-effect is zero so this can be removed,

ˆ Rc the electrolytic resistance of the concentrate channel,

ˆ Rd the electrolytic resistance of the diluate channel,

ˆ RCEM the resistance of the CEM,

ˆ RAEM the resistance of the AEM.

Rer is very hard to estimate but since this is the only parameter left undetermined,

it can easily be estimated from the data. Re, Rc, Rd are easily calculated with the

conductivity computed with equation 2.4,

R=



κ AmAs (2.7)

with  the channel thickness, Am the membrane surface and As the fraction of the

membrane that is not blocked by spacers. Image analysis showed that for our spacers this is about 85%. Finally, the resistance of the IEMs are obtained from the manufac-turer. This equation is implemented as the resistance-function and plotted over time. Equation 2.1 - 2.7are implemented in Python. The poly1D function is used to fit a 2nd order polynomial to the temperature data.

RESULTS AND DISCUSSION

3.1

Limiting current density

The LCD marks the point at which the transfer of ions from the liquid to the membrane is limiting. Exceeding this value will cause water splitting and decrease the energy efficiency of the stack since this fraction of the current is not used to remove salt ions. An experimental determination of the LCD was performed prior to the fouling experiments. This to make sure all fouling experiments were performed in ohmic conditions that avoid water splitting. To visually determine the LCD, multiple voltages, chosen within a certain range, are applied on the stack. The corresponding current measured for each potential is then plotted in a graph. In this graph, the LCD can be determined at the onset of a plateau. This means that for an increase in potential the measured current stays about the same which indicates an increase in the resistance of the stack. Figure 3.1 shows that between 0 V to 8 V no LCD is found as no plateau or deviation from the linear curve is seen. In theory, this means that all of the current is used to transport ions through the membranes and no water splitting occurs. The high salt concentration of 0.1M and a high crossflow are very likely responsible for a higher LCD as no depletion of ions occurs at the IEMs. Finding the LCD to a maximum of 8V was more then enough as all fouling tests were run at 1 A. 8V already corresponds to 3A (if no fouling or desalination is present).

Figure 3.1: linear relation between the current and the potential.

3.2

Continuous fouling of SDS

In this series of tests, the contribution of different factors that are likely to influence the fouling behaviour of SDS are analysed. As mentioned in section 1.3.1, SDS is known to foul AEM quite easily due to its negative charge and its aliphatic chain. The effect of fouling is primarily an increase in membrane resistance and will require cleaning or replacing later on. Knowing the effect of each parameter can help to better understand the fouling process and aid in avoiding irreversible damage to the membranes. To investigate the impact of each parameter separately one variable is changed for each test. Parameters that are chosen to alter are SDS concentration, cross-flow velocity, current and membrane type.

To visualise and quantify the amount/severity of this fouling during the test, the elec-tric potential is plotted over time. As the current that is applied during the test is kept constant, the potential increase can be directly linked to the resistance increase via Ohm’s law.

3.2.1

The effect of SDS concentration

First, the impact of the foulant concentration is studied. A baseline and five different SDS concentrations are tested and results are plotted in figure 3.2. The potential of the baseline (blank) stays almost completely flat during the test which is to be ex-pected as no foulant is present. The minor decrease in the potential of the baseline can be caused by a slight increase in temperature due to pumping and heat

gener-ation from the electrodes (Ohmic dissipgener-ation). An increase in temperature raises the conductivity of the electrolyte solutions which lowers the resistance.

At the lowest concentration of SDS (0.50 g/l) fouling is already observed as the po-tential increases approximately 0.5 V after three hours. For the other time series, the trend is clear, an increase in SDS concentration results in an increase in membrane resistance. Only the concentration of 1.00 g/l SDS gives a rather different behaviour. All other concentrations increase gently over time but here a steep increase of the electric potential occurs during the first 30 minutes after which the electric potential stabilises at a lower potential. Clearly, the SDS concentration has a large impact on fouling behaviour, especially near 1 g/l. One possible explanation is that as a result of concentration polarisation, in combination with a high SDS concentration, the CMC is exceeded close to the membranes, leading to a severe fouling effect of the SDS mi-celles (51). It is known that colloidal fouling can give rise to an autocatalytic fouling regime due to water splitting effects. No explanation was found for the decrease in potential and some additional tests have to be performed in the future.

Figure 3.2: fouling behaviour of different SDS concentrations.

In table 3.1, the total resistance and the relative increase after three hours of fouling, compared to the baseline, is given. Plotting those percentages in figure 3.3 shows a quite linear relation for an increase in potential between 0.50 g/l and 0.92 g/l of SDS. Only the value of 1.00 g/l diverges quite seriously from this curve as it increases significantly for such a minor increase in concentration compared to 0.92 g/l.

As the behaviour of 1.00 g/l SDS differs a lot from the behaviour of the other concen-trations, a second run is performed with new AEM and plotted In figure 3.4 together with the baseline and the previous run of 1.00 g/l. Similar behaviour is found for both experiments and, even though there is a slight discrepancy, the occurrence of a peak

Table 3.1: overview of the potential and relative increase of the different fouling curves compared to the baseline.

SDS concentration (g/l) Potential increase (V) Relative increase (%)

0.50 0.49 13.6

0.75 0.98 27.1

0.82 1.07 29.6

0.92 1.19 33.0

1.00 1.96 54.3

Figure 3.3: relative potential increase of the different fouling curves compared to the baselines.

in the electric potential is not an artefact. Both curves generally follow the same shape during the test, only the peak of the second run is higher and occurs later. The decrease after this peak is steeper but remains higher after three hours.

After the fouling tests, a reversal with a 0.1M salt solution is performed to ‘clean’ the membranes and to try and remove the fouling. This is done by reversing the polarity of the electrodes causing the foulants to be pulled in the opposite direction. Figure 3.5 shows all reversal experiments one for each of the performed fouling tests concentration along with the baseline as reference. The same absolute value of the stabilised electric potential is found for the baseline as fouling but now negative due to the polarity change. For the membranes that are fouled by SDS, the potential first drops to a minimum in the first few minutes, to then smoothly decrease and flatten out to a constant value. All the final values of the multiple reversals tend to a very similar value with some minor variation between them. the slight drop in the first few minutes can be explained by the transient behaviour due to the build-up of the concentration polarisation layer and other capacitive factors that need to stabilize. What is especially interesting about this graph is that the value where all the fouling experiments tend to stabilise never comes close to the original value of the baseline. This indicates that SDS is causing some irreversible fouling inside or on the surface of the membranes. The slow and steady decrease most likely indicates the reversible fraction of the fouling layer being removed. There might be a slight indication that the minimum value of all of those curves seem to be lower when fouled with a higher SDS concentration, although not every minimum follows this rule as the minimum of 0.50 g/l is lower than for example 0.82 g/l. The average potential or resistance increase due to irreversible fouling can be computed by subtracting the end value of the baseline from the average steady-state value of the fouling tests. This value is about approximately 0.55 V or 0.55 ohms respectively as the current is kept constant at 1 A.

3.2.2

The effect of the crossflow velocity

The next parameter chosen to alter is the flow rate of the diluate and concentrate streams in the stack. Lowering the crossflow velocity decreases the shear force on the surface of the membranes, increasing fouling rate as this flow is less capable of removing foulants sticking to the surface of the membranes (26).

Figure 3.6 shows the fouling of 0.50 g/l SDS at a maximum rpm (530 ml/min) and 0.50 g/l SDS at 75% rpm together with their respective baselines. When decreasing flowrate without a foulant present, the base behaviour does not change considerably. This means that the depletion of ions is still limited even when the crossflow veloc-ity is reduced by 25%. The slight potential difference could again be accounted to variations in temperature. When comparing the two profiles with foulant present, it is clear that decreasing the crossflow velocity increases the fouling rate considerably. A steep increase in potential is observed for about 75 minutes, after that it decreases slightly to a lower end value. This is a substantial difference compared to the fouling curve of the same SDS concentration but at maximum rpm. When only looking at the end value after three hours of fouling at a decreased flow rate, an increase of 1.07V or 29.3% is observed when comparing it to its respective baseline. Comparing it to the similar fouling test of 0.50 g/l but at the maximum flow rate, an increase of 0.62V or 15.1% is observed due to changing the reduction of the flow rate. It is interest-ing to point out that the foulinterest-ing behaviour of the decreased flow rate is not quite as different than the fouling curve 1.00 g/l SDS curve. Both show a peak followed by a slight decrease. This could indicate that the source of the potential increase might be caused by the same phenomenon, i.e. exceeding the CMC as a result of concentration polarisation but this should require more in-depth research.