- Email: [email protected]
o Pickering emulsion ultrafiltration
Author’s Accepted Manuscript Feasibility of w/o Pickering emulsion ultrafiltration Tina Skale, Lena Hohl, Matthias Kraume, Anja Drews
www.elsevier.com/locate/memsci
PII: DOI: Reference:
S0376-7388(17)30327-7 http://dx.doi.org/10.1016/j.memsci.2017.04.006 MEMSCI15170
To appear in: Journal of Membrane Science Received date: 2 February 2017 Revised date: 4 April 2017 Accepted date: 5 April 2017 Cite this article as: Tina Skale, Lena Hohl, Matthias Kraume and Anja Drews, Feasibility of w/o Pickering emulsion ultrafiltration, Journal of Membrane Science, http://dx.doi.org/10.1016/j.memsci.2017.04.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Feasibility of w/o Pickering emulsion ultrafiltration Tina Skale1*, Lena Hohl2, Matthias Kraume2, Anja Drews1 1
HTW Berlin, Dept. 2, Process Engineering in Life Science Engineering, Wilhelminenhofstr. 75A, 12459 Berlin, Germany 2
Technische Universität Berlin, Fak. 3, Process and Chemical Engineering, Fraunhoferstr. 33-36, 10587 Berlin, Germany *Corresponding author: Tina Skale ([email protected]; +49 (0)3050193605) Abstract: This study shows the possibility of separating w/o Pickering emulsions (PE) via ultrafiltration and consequently enables a continuous process concept for catalysed L/L reactions in Pickering emulsions which are currently receiving increased attention. In two types of filtration experiments, the stability of PE against coalescence caused by the applied filtration pressure and stirring was shown. The PE could be concentrated up to 80 % water phase fraction. In pressure stepping experiments, permeabilities between 3 and 10 L/(m2 h bar) could be achieved depending on the drop size distribution of the PE. For the used 1 kDa PVDF membrane, an unexpected overproportional behaviour of the fluxes at higher pressures was observed. The presented results show that Pickering emulsions can be regarded as a promising alternative to conventional dispersions also in continuous L/L catalysis.
Keywords: Pickering emulsion, ultrafiltration, drop size distribution, organic solvents
1
Introduction
Pickering emulsions (PE) are emulsions in which the dispersed phase is stabilised by solid particles as opposed to surfactants [1]. They were named after S.U. Pickering who in 1907 described them first [2]. It took 80 years, however, until Pickering emulsions were rediscovered for applications in the pharmaceutical, cosmetics and food industries because of their high resistance against coalescence without the need for using harmful surfactants [3]. Since 2011, Pickering emulsions have been studied as possible systems for reactions in liquid-liquid (L/L) multiphase systems [4]. Chemical and enzymatic catalysts are often soluble in the aqueous phase while the reactant and the product are in the organic phase. In their minireview from 2015, Pera-Titus et al. [5] gave an overview about Pickering interfacial catalysis systems. According to the authors, despite the potential for applications in industrial processes, the aspect of catalyst recovery or recycling which would be required for an 1
economically feasible process has not been solved yet [5]. The aim of this work thus is to tackle this challenge and to investigate if membrane filtration for L/L separation of w/o Pickering emulsions and thereby catalyst recycling is possible. While the separation of surfactant emulsions by membrane filtration has been studied frequently, separating Pickering emulsions using membranes has not been done before. The advantage of phase separation via membrane filtration is that the medium does not change in a chemical, thermal or biological way. To approach this task, first, a solvent resistant hydrophobic ultrafiltration membrane has to be found to which the used nanoparticles have less affinity than to the L/L interface. The second question to be solved is if the Pickering emulsions are stable enough against coalescence during ultrafiltration. If the filtration is possible, the influence of parameters that could affect the PE stability like solid particle content and drop size distribution on the filtration behaviour has to be studied. And finally, for the recirculation in a continuous industrial process the maximal possible concentration of PE in the reactor needs to be determined. As a model system for the industrially relevant rhodium catalysed hydroformylation of alkenes, 1-dodecene as a representative of longchain olefins is used as the reactant. The catalyst is dissolved in the water drops that are stabilised by silica particles. The envisaged separation principle is shown in Figure 1. Figure 1: Separation principle of w/o Pickering emulsion (PE)
2
State of the Art
2.1 Applications and properties of PE and their stabilising nanoparticles In the early 2000s, Pickering emulsions came more and more into the focus of investigations because of their high stability against coalescence and other promising chemical properties. Pickering emulsions are more resistant against changing chemical parameters like, e.g., pH and salt content [3]. Applying functional solid nanoparticles, Pickering emulsions can also be made temperature-sensitive or pH-sensitive as required [6]. Much research has been done, e.g., on the influence of particle content and particle properties on the coalescence behaviour [7] and on the rheology of Pickering emulsions [8-11]. Binks et al. [1] gave an overview about particles as stabilizers compared to surfactants. They showed that the formation of either an o/w or w/o emulsion depends on the hydrophobicity of the solid particles, because this is related to the wettability and thereby to the extent to which the particles penetrate each liquid phase. They also found that with increasing particle hydrophobicity and water phase fraction, a w/o emulsion inverted to an o/w emulsion. The hydrophobicity of the particles also influences the stability against coalescence. Particles that are very hydrophobic or very hydrophilic form emulsions with larger drops over 100 µm that are unstable against coalescence [7]. To obtain a coalescence stable emulsion and smaller drop diameters, the stabilizing particles have to have an intermediate hydrophobicity. The size of the solid stabilized drops also depends on the size of the solid particles. With nanoparticles of about 100 nm, micron-sized droplets can be stabilized. Silica nanoparticles are often used because they can easily be surface modified to make them hydrophobic [6]. Since harmful surfactants which can lead to skin irritations can be avoided, a large advantage of PE applications in the cosmetics and pharmaceutical industries is seen [6]. For industrial applications, Frelichowska et al. [12] compared the drug release of a w/o Pickering emulsion with that of a classical emulsifier-stabilized emulsion. The Pickering emulsions showed higher permeation rates of 2
caffeine through the skin because of the stronger adhesion of the water droplets on the skin surface. Further information on the history and properties of Pickering emulsions and their applications in industry can be found in a review from 2013 [6].
2.2 L/L Separation of (Pickering) Emulsions To the best of our knowledge, there are barely any studies about separating Pickering emulsions which would be required to withdraw the product in a continuous process and to reuse the catalyst [13-15]. Zhang et al. recently applied PE in the form of a packed bed for different continuous reactions. However, with 100 h and more, the residence times they could realise with this set-up were very high and would be inefficient for the hydroformylation studied here. To decrease the residence time, the pressure would have to be increased which would most likely deform the drops in the packed bed, in turn would decrease the flow rate even further (self-accelerating problem) and might even lead to coalescence. In addition, once packed inside a column, drops could not be exchanged or renewed when catalyst activity decreases. Demulsification by centrifugation with successive redispersion for the L/L separation of reactive Pickering emulsions was studied by Wei et al. [14]. This, however, would require repeated power input. In addition to centrifugation, typical demulsification techniques of conventional emulsions include the addition of chemical demulsifiers, pH adjustment, gravity settling, filtration, heat treatment and electrostatic demulsification [16]. Dyab studied the destabilisation of w/o silica Pickering emulsions by changing the pH which changes the wettability of the silica particles. However, for applications of Pickering emulsions as a reaction system, changing the pH could be adverse. [17-19] pursue the strategy to separate Pickering emulsions by controlled shear-induced coalescence and as a result interrupting the network between particles and drops. Another approach for the separation of Pickering emulsions is the use of magnetic particles for the stabilisation and to separate the emulsion drops from the continuous liquid via a magnet [20], but there is no literature about the influence of magnetic particles on the reaction system and the industrial scale realisation. Therefore, in this study, membrane filtration for the continuous separation of w/o Pickering emulsions as a less energy intensive method which enables drop exchange and shorter residence times, and which does not change the reaction environment was investigated. The use of membranes for L/L separation of surfactant stabilised emulsions has been studied frequently before [e.g. 21,22]. During ultrafiltration of surfactant stabilized oil in water (o/w) emulsions at pressures between 1 and 2.5 bar, the water drops were found to coalesce in the gel-layer possibly due to viscous drag caused by the permeate flow [21]. Another explanation for the coalescence in the gel-layer could be the loss of surfactant to the permeate. There, both smaller and larger drops than present in the gel-layer were found which implied that oil drops passed through the membrane and during this passage broke and coalesced. During concentrating experiments of o/w emulsions the fluxes hardly showed any changes in the beginning (initial oil fractions of 5 vol%) but when oil fractions of around 10 vol% (depending on the oil type) were reached, the fluxes decreased strongly to a maximum possible oil fraction of about 30 % oil phase [21]. Recent studies showed that during ultrafiltration of surfactant stabilized o/w emulsions with ceramic membranes, the permeate flux depended on the charge of the membrane surface and the emulsion drops. 3
Electrostatic attraction effected an adsorption of the surfactant on the membrane surface and the surfactant could pass the membrane. The electrostatic repulsion caused the opposite, the surfactant could not adsorb which prevented its membrane passage and led to decreased concentration polarisation and cake layer formation [22]. Most literature on membrane filtration of emulsions could be found for o/w emulsions using hydrophilic membranes. To increase the permeate flux without increasing the membrane fouling, Yi et al. tested a modified PVDF membrane for the separation of a surfactant stabilised oil in water emulsion. With increasing oil concentration, the permeate flux (normalised with the initial flux) decreased sharply. The used fouled membranes showed a high flux recovery of 88% and 94% after having been washed with pure water [23]. Using their modified PVDF ultrafiltration membranes, Yi et al. found that for o/w emulsions the cake filtration model did not predict the filtration behaviour over the whole filtration time but was followed by intermediate pore blocking [24]. Headen et al. found that oil drops from a model o/w emulsion deformed with increasing filtration pressure, but nevertheless the filter cake stayed permeable [25]. Tummons et al. developed a method to get real-time images of the membrane surface during microfiltration of o/w emulsions. With this method they could observe three characteristic stages of membrane fouling by oil. At first, the droplets attached and clustered, then they deformed and thirdly coalesced. Varying the surface tension by increasing the surfactant concentration showed a change in the drop sizes followed by a change in the coalescence behaviour. The authors concluded that the membrane fouling depends on droplet coalescence and crossflow shear. Oil droplets with diameters slightly above the pore diameter permeated through the membrane [26].
2.3 Membranes in oil and w/o applications The vast majority of studies about membrane filtration deals with aqueous systems and fewer with oily/organic systems. The literature on oily systems narrows down to o/w emulsions, e.g., the cleaning of oily wastewater with ultrafiltration and microfiltration [27-32]. Since the advent of organic solvent nanofiltration, several solvent resistant nanofiltration membranes have become commercially available, but commercial ultrafiltration membranes are still mostly designed for aqueous applications. This narrows the choice of possible membrane materials for the ultrafiltration of the w/o Pickering emulsions. Some companies (MICRODYN-NADIR GmbH Germany) offer membranes with solvent resistant material for microfiltration. Alfa Laval’s ETNA series is composed of an active PVDF layer on a cellulose support. PVDF has a high hydrophobicity and resistance against most solvents [33], but because of the usual usage of ultrafiltration in aqueous systems like for wastewater treatment or filtration of protein solutions, the hydrophobic PVDF membranes were mostly hydrophilized to avoid membrane fouling [33,34]. In the cleaning of phospholipids from a crude soybean oil/hexane mixture, PVDF permeate fluxes were up to threefold larger than the fluxes of a polyimide membrane under the same operating conditions [35]. Lencki and Williams studied the influence of non-aqueous solvents on ultrafiltration and found that while some solvents increased the permeability, others decreased it. They stated that the resistance of polymeric membranes depends on the membrane structure (isotropic/ anisotropic) and on the interactions between solvent and membrane (solubility parameter) [36]. 4
For the preparation of o/w emulsions via membrane emulsification, Giorno et al. studied the effect of different solvents on polymeric tubular ultrafiltration membranes of different MWCO [37]. Depending on the solvent and the MWCO, the pore structure changed, so in order to achieve reproducible flux levels, an individual optimal pre-treatment of the used membrane and solvent needs to be found which, due to the frequent use of membranes in aqueous media, has not been investigated exhaustively yet. Penha et al. studied the influence of different solvents and pre-treatment durations on commercial polymeric ultrafiltration membranes [38]. They found that the permeability of the polymeric membrane depends not only on the pore size but also on the interaction between solvent and polymer. An increase in permeate fluxes was reached with an increase of the immersion times in a combination of different solvents. This special treatment of the membranes increased the contact angle and increased the hydrophobicity of all tested membranes without changing the membrane structure as described by Lencki and Williams [36] and Giorno et al. [37].
3 3.1
Materials and methods Preparation of the emulsion
As a model system, water-in-dodecene Pickering emulsions were investigated. 30 mL of emulsion were prepared for one batch which consisted of 1-dodecene with a purity of >94% by Merck Chemicals GmbH, Germany and deionised water (10.0 µS/cm at 25 °C) at a volumetric phase ratio of 3:1 (dodecene:water). The emulsion was stabilised with 0.5 wt% (with respect to the total mass of the emulsion) of hydrophobic silica nanoparticles HDK H20 by Wacker Chemie AG, Germany. The particles form elongated aggregates with a size of about 10 x 150 nm [38] and have a residual silanol content of 50%. First, the silica particles were wetted completely with 1-dodecene. Then, the deionised water was added. The still separated mixture was emulsified using an ultrasonic homogeniser (Bandelin HD 3200, Germany) for 5 minutes under manual stirring in an ice bath. To prepare emulsions with different drop size distributions, the amplitude and the pulse time of the ultrasonic homogeniser were varied. The amplitude and thus the energy input was reduced from 50 to 15% to achieve larger drops. In experiments on the influence of the solvent type on PE filtration behaviour, 1-dodecene was replaced by n-decane (99 %, Thermo Fisher (Kandel) GmbH, Germany), 1-decene (>96 %, Merck Chemicals GmbH, Germany) or toluene (99.9 %, Carl Roth GmbH + Co. KG, Germany). 3.2
Characterisation of the emulsions
To check the reproducibility of preparing the emulsion with the ultrasonic homogeniser and to characterise the influence of filtration on the emulsion, drop sizes were measured before and after the filtration experiments. 25 pictures (five pictures of each of the five samples of each emulsion) were made by microscopy (Carl Zeiss AG, Germany, Axio Imager 2). The pictures were analysed manually and by an image evaluation software (Sopat GmbH, Germany). The number of counted drops per emulsion was between 2000 and 6000 drops for the image evaluation software and between 600 and 6000 drops for the manual analysis. 600 drops were found to be statistically sufficient because the Sauter mean diameter stayed constant despite further counted drops.
5
3.3
Pre-treatment of the membrane
For the filtration experiments, a hydrophilized PVDF membrane (Alfa Laval Denmark, ETNA 01PP) with a MWCO of 1 kDa was used [34]. Other solvent resistant membranes (PuraMemS600 by Evonik, Germany, and GR95PP by Alfa Laval, Denmark) showed no relevant 1-dodecene fluxes at the applied pressure and were therefore not used in further trials. Despite micrometre scale emulsion droplets, a low MWCO membrane was chosen to also retain possible free silica agglomerates. For the pre-compaction of the membrane and to ensure complete wetting, the membranes were pre-treated before every pressure stepping experiment by washing with pure solvent for 90 min (unless indicated otherwise) at 2.5 bar. A new membrane was used for each experiment.
Figure 2: Experimental set up for pressure stepping experiments at constant phase fraction
3.4
Experimental set-up
For all filtration experiments, a solvent resistant stirred cell XFUF04701 by Merck Millipore with a working volume of 91.5 mL was used. 3.4.1 Concentration experiments The objective of the concentrating experiments was to find out to what extent the emulsions could be concentrated, i.e., at which L/L ratio drops would break and water would begin to pass through the membrane. The pressure p was set to 2.5 bar and stirrer speed to 500 rpm which corresponds to wtip = 1 m/s. The flux was logged by reading an electronic balance. 3.4.2 Pressure stepping experiments with constant phase fraction To maintain a constant phase ratio while analysing the influence of pressure, 1-dodecene was continuously dosed into the stirred cell from a surge tank (see Figure 2). When an almost constant flux was achieved, the pressure was increased in a stepwise manner from 1 to 4 bar after which it was decreased again from 4 to 1 bar. Each pressure level was kept for 1 h during pressure raise and 30 min during pressure descent. The flux was logged by reading an electronic balance.
4 4.1
Results and discussion Reproducibility of the emulsion preparation method
Prior to membrane filtration experiments, the reproducibility of PE properties is a prerequisite. For a validation of the required drop size distribution measurements, see 4.2. Figure 3 (left) shows the cumulative distribution function (cdf) Q0 of the drop diameters dd for 3 emulsions prepared separately but with the same settings of the ultrasonic homogeniser (15 % amplitude, 2 s pulse on and 3 s pulse off for 5 min). The drop size distributions were nearly the same with little differences in the smallest and the biggest drop sizes. To avoid these little differences and to increase the reproducibility of the filtration experiments, the separately prepared emulsions were mixed and separated again into 30 mL batches. The results of the drop size measurements are shown in Figure 3 (right). The cumulative distribution functions still show little differences in the medium drop sizes. The maximum deviation between the three individually prepared emulsions was 2.02 µm for the arithmetic 6
mean diameter d1.0 and 4.39 µm for the Sauter mean diameter d3.2. In contrast, the difference between the determined minimum and the maximum arithmetic mean diameter of the mixed emulsion samples was only 0.24 µm and 3.25 µm for the Sauter mean diameter. Hence, for further filtration experiments, batches of PE were always mixed. n
d 3.2
d
3 di
d
2 di
i 1 n i 1
(4.1)
n
d1.0
d i 1
di
(4.2)
n
Figure 3: Reproducibility of the ultrasonic homogeniser preparation: left: cdf of 3 separately prepared emulsions, and right: cdf of 3 separately prepared and then mixed emulsions
Figure 4 (left) shows the drop size distribution of two PEs prepared with equal settings but with two different ultrasonic homogenisers of different power. The emulsion prepared with the low power device (HD70, 70 W) contained both a higher amount of larger drops and more small drops below 10 µm. The PE was highly polydisperse with a Sauter mean diameter d3.2 of 18.35 µm. The emulsion prepared with HD3200 (200 W) was totally different, almost monodisperse with almost halved drop sizes (d3.2 = 11.18 µm). To visualise the difference of the two emulsions, microscope images are also shown in Figure 4 (right). With a variation in the type and in the settings of the ultrasonic homogenizer, the drop sizes and the drop size distribution could be influenced to achieve the desired wide or narrow distribution for studying their influence on flux and their stability during filtration.
Figure 4: left: drop size distribution of two emulsions prepared with different ultrasonic homogenisers with different power, right: microscopic pictures of the two differently produced emulsions
4.2
Characterisation of the emulsion
In order to investigate the influence of drop size distribution on filtration behaviour and of filtration on drop size distribution, the drop sizes have to be measured. To facilitate the image evaluation and to reduce the possibly subjective error of the manual measurements, an automatic image evaluation software by SOPAT GmbH (Germany) was used. To decide whether the software can be applied to evaluate these pictures, the two measurement methods were compared by measuring the drop sizes of the same pictures manually and with the software. The results are shown in Figure 5 for two different emulsions. The cumulative density functions of the emulsions were averaged from at least three (left) and four (right) individually prepared emulsions. Table 1 shows the total number of counted drops for both measurement methods. 7
Figure 5: Comparison of the two drop size distribution measurement methods (manual and software), left: 15% amplitude PEs and right: 50% amplitude PEs
For the 50% amplitude emulsion, the number of the automatically counted drops was nearly as high as the number of the manually counted drops (same order of magnitude). The plotted cumulative density functions of both measurement methods were nearly the same with only little deviations. The drop size distributions for the 15% amplitude emulsions showed small variations in the small and medium drop sizes. However, the deviation between software and automatic evaluation of the Sauter mean diameter was only 0.62 µm and of the arithmetic mean diameter even only 0.12 µm. For the less polydisperse 50 % amplitude emulsion, the difference of both mean diameters was below 1 µm. Since the maximum deviation between manual and automatic analysis is less than 0.7 µm in both the Sauter and the arithmetic mean diameter, the automatic evaluation can be applied for the characterisation of the emulsions. The accuracy and resolution of the automated image analysis software was studied by repeating the search parameter estimation of the same microscopic pictures. A deviation of ± 1 µm from the Sauter mean diameter was normally observed [39,40]. Table 1: Number of counted drops of the two measurement methods for two different emulsion types (different drop size distributions because of two different energy inputs during the PE preparation) and average drop diameters
15% amplitude no. of counted drops [-] d1.0 [µm] d3.2 [µm] 4.3
manual 1972 9.63 15.43
software 19609 9.75 14.81
50% amplitude manual 16924 8.72 10.7
software 32560 9.41 11.18
Pre-treatment of the membranes
The results of the membrane washing with 1-dodecene showed a rapid flux decline within the first 20 minutes from over 180 L/(m2 h) to about 15 L/(m2 h). This rapid flux decline could be explained by the anisotropic structure of the ETNA01PP membrane where the surface layer has another swelling behaviour than the support layer. According to [36], the active layer swells more than the support layer which would lead to an increased membrane resistance. This change in membrane structure caused by the solvent seemed to be different for the used membrane samples, so to allow reproducible filtration experiments, membranes with strong deviations (±100%) from the average washing flux were discarded. This flux variation despite the same pre-treatment is a typical behaviour of ultrafiltration membranes [36]. The deviations in the pre-treatment of the membranes followed no pattern (e.g., position on the membrane sheet from which samples were cut). Membranes with high washing fluxes did not have higher fluxes in the pressure stepping experiments than membranes with lower washing fluxes. This might be caused by the dominating resistance of the filter cake. 4.4
Concentration experiments
Figure 6 shows the results of the concentration experiments. The flux is plotted against the current aqueous phase fraction in the stirred cell for three 15 % amplitude emulsions and for different membrane pre-treatment times. The aqueous phase fraction of the emulsion in the stirred cell φw was calculated according to eqs. (4.3) and (4.4.) from the initial dodecene 8
volume of the emulsion Vdod,0 and the accumulated permeate mass measured on the balance at each point in time mdod(t). As long as no water passes through the membrane, the water volume VW remains constant throughout the experiment.
j w (t ) =
Vw (Vw +Vdod,0 -Vdod (t))
Vdod (t) =
(4.3)
mdod (t)
(4.4)
rdod
The wetting of the membrane influenced the maximum concentration point. After 90 minutes of pre-treatment, the phase fraction could be increased from 22 % to about 80 %. This maximum phase fraction could not be increased by longer pre-treatment of 120 minutes. The individual flux levels of the concentration experiments with dry membranes differ from each other and show steeper decreases during the experiments. The pre-treatment led to more reproducible and stable fluxes and was therefore applied in pressure-stepping experiments. The achieved maximum phase fraction of about 80 %, which is much higher than the maximum phase fraction achieved in surfactant emulsions [21], shows the high stability of the Pickering emulsion against pressure, shear force and changing phase ratio, and allows compact reactors and membrane modules to be used.
Figure 6: Three 1-dodecene PE concentration experiments each after different membrane pre-treatment times (0, 20, 90 and 120 min)
4.5
Pressure stepping experiments
Figure 7 (top left) shows results of the pressure stepping experiments with two Pickering emulsions prepared with HD70 and 50 % amplitude in comparison to two pure 1-dodecene experiments. Apart from a slightly larger deviation at 4 bar, experiments can be seen to be reasonably reproducible. The pure solvent experiments show the expected disproportional behaviour at higher pressure due to changing resistances caused by, for example, membrane compaction. However, the results of the PE filtration were unexpected: The fluxes of the PE were higher than the fluxes of the pure 1-dodecene experiments at 3 and 4 bar and increased disproportionally high at higher pressures. To explain these unexpected results, three hypotheses were investigated. One was that abrasion of the membrane surface occurs caused by the silica particles. Figure 8 shows SEM pictures of the fresh membrane surface and two membrane surfaces after the PE pressure stepping experiments. No abrasion of the membrane surface but only a silica layer located in the region below the stirrer could be observed.
9
Figure 7: Pressure stepping experiments with constant phase fraction for four different solvents (1dodecene, 1-decene, n-decane and toluene, pure solvent and PE, respectively)
A second assumption was that residual traces of silica particles from previous experiments (despite thorough cleaning) in the stirred cell could block the membrane surface during the pressure stepping experiments with pure 1-dodecene. During the experiments with the PE, the water drops would adsorb these residual silica particles thus freeing the membrane surface. However, repeating the pure solvent experiments in a new silica-free stirred cell resulted in the same lower flux levels. So higher resistances due to silica on the membrane can be ruled out as the reason for the low pure solvent fluxes.
Figure 8: SEM pictures of fresh and used membrane samples
To investigate if the lower pure solvent fluxes are specific to the used 1-dodecene because of, e.g., stabilisers present in the 1-dodecene, other solvents that are similar to 1-dodecene in chain length and hydrophobicity but have different purities were tested. Figure 7 (bottom left) shows the results of two experiments with PE prepared with n-decane and two experiments with pure n-decane. The fluxes of the PE with n-decane also show the disproportionately high permeability at higher pressure, but in this case, the fluxes of the pure n-decane were higher than the fluxes of the PE at every pressure step. For toluene and 1decane (Figure 7, bottom), similar flux levels for pure solvent and PE filtration were found. Little differences could be caused by intrinsic variations of the membrane samples. In conclusion, the unexpected much higher fluxes of the PE prepared with 1-dodecene compared to the pure solvent seem to be specific to 1-dodecene and the used membrane. For all solvents it could be seen that the flux differences during pressure raise and pressure descent were larger for the pure solvents than for the PE. This could also been seen in the raw data (not shown). The fluxes of the PE filtration reached the steady state faster than the fluxes of the pure solvent where the fluxes only reached steady state during pressure decrease. Among the PEs prepared with different solvents, the 1-decene PE was the least stable emulsion, so during the pressure stepping experiments, small amounts of water continuously broke through the membrane. Nevertheless, the fluxes showed the same hysteresis as the fluxes of the stable emulsions. One possible explanation for the disproportionate increase of the fluxes at higher pressure could be the coalescence of the water drops and the resulting larger hydraulic diameter of 10
the filter cake. In our previous study, drop coalescence during the pressure-stepping experiments was observed [15]. The extent of drop coalescence was found to depend on the drop size distribution and the stability of the Pickering emulsion [15]. Figure 9 shows the results of the pressure stepping experiments with PEs of different drop size distributions. The maximum deviation in the cdf can be seen between the PE prepared with HD70 and the emulsions prepared with HD3200. The higher fluxes of the PE prepared with HD70 might have been caused by the bigger drops and the consequently higher hydraulic diameter of the filter cake layer. The reduction of the power input by an amplitude reduction from 50% to 15% (HD3200) hardly changed the drop sizes. An influence of the drop size distribution on filtration could only be seen for emulsions with many larger drops such as the emulsion prepared with HD70 and an emulsion with a narrow drop size distribution with smaller drop sizes such as the emulsions prepared with HD3200. The smaller amounts of larger drops as present in the 15% emulsion did not influence the flux level significantly compared to the emulsion prepared at 50% amplitude. Therefore, the larger drops might coalesce at higher pressure so that the fluxes become higher, but a critical amount of larger drops is needed before a significant influence on the flux levels can be seen. Also, the observed differences in the washing fluxes of the membranes made it harder to discover fine differences in the flux levels and therefore in the influence of changing drop sizes. However, one obvious difference caused by the different drop sizes was that small water drops passed through the membrane during the pressure stepping experiments with the PE prepared with HD3200 and 50% amplitude. It seemed that because of the low hydrophobicity of the silica particles, a phase inversion and a formation of a double emulsion during the permeation of water, solvent and particles through the pores occurred. Small water drops containing several even smaller drops could be observed in the permeate.
Figure 9: Pressure stepping experiments with 1-dodecene PE with different drop size distributions, bottom right: average cdf of the PE used for these filtration runs
5
Conclusions
For the first time it could be shown that L/L separation of different w/o Pickering emulsions by membrane filtration is possible. The two liquid phases could be separated without destroying either the membrane or the emulsion. The nanoparticles did not adhere to nor block the membrane. The Pickering emulsions showed an unexpected flux behaviour with permeability increasing with higher pressures. Fluxes between 3 and 40 L/(m2 h) were achieved for pressures between 1 and 4 bar. The emulsions could be concentrated up to a water phase fraction of 80 %. The drop size distribution of the emulsion had no significant influence on the flux level except for emulsions with a larger amount of larger drops. The presented results show that Pickering emulsions can be regarded as a promising alternative to conventional dispersions also in continuous L/L catalysis. 11
6 Acknowledgement Financial support by the German Research Foundation DFG (collaborative research centre "Integrated Chemical Processes in Liquid Multiphase Systems" InPROMPT TRR63, B6) is gratefully acknowledged. We thank Alfa Laval for kindly providing free membrane samples.
7
References
[1] B.P. Binks, Particles as surfactants – similarities and differences, Current Opinion in Colloid & Interface Science 7 (2002) 21-41. [2] S.U. Pickering, Emulsions, J. Chem. Soc. 91 (1907) 2001-2021. [3] G. Lagaly, O. Schulz, R. Zimehl, Dispersionen und Emulsionen, Springer-Verlag, Berlin Heidelberg, 1997. [4] C. Wu, S. Bai, M.B. Ansorge-Schumacher, D. Wang, Nanoparticles Cages for Enzyme Catalysis in Organic Media, Adv. Mater. 23 (2011) 5694-5699. [5] M. Pera-Titus, L. Leclercq, J.-M. Clacens, F. de Campo, V. Nadrello-Rataj, Pickering Interfacial Catalysis for Biphasic Sytems: From Emulsion Design to Green Reactions, Angew. Chem. Int. Ed. 54 (2015) 2006-2021. [6] Y. Chevalier, M.-A. Bolzinger, Emulsions stabilized with solid nanoparticles: Pickering emulsions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 439 (2013) 23-34. [7] B.P. Binks, S.O. Lumsdon, Influence of Particle Wettability on the Type and Stability of Surfactant-Free Emulsions, Langmuir 16 (2000) 8622-8631. [8] J. Chen, R. Vogel, S. Werner, G. Heinrich, D. Clausse, V. Dutschk, Influence of the particle type on the rheological behavior of Pickering emulsions, Colloids and Surfaces A: Physicochemical and Engineering Aspects 382 (2011) 238-245. [9] T. Fuma, M. Kawaguchi, Rheological response of Pickering emulsions prepared using colloidal hydrophilic silica particles in the presence of NaCl, Colloids and Surfaces A: Physicochemical and Engineering Aspects 465 (2015) 168-174. [10] A. Nesterenko, A. Drelich, H. Lu, D. Clausse, I. Pezron, Influence of a mixed particle/Surfactant emulsifier system on water-in-oil emulsion stability, Colloids and Surfaces A: Physicochemical and Engineering Aspects 457 (2014) 49-57. [11] L. Hohl, S. Röhl, D. Stehl, R. v. Klitzing, M. Kraume, Influence of nanoparticles and drop size distributions on the rheology of w/o Pickering emulsions, Chem. Ing. Tech. 88 (2016) 1815-1826. [12] J. Frelichowska, M.-A. Bolzinger, J.-P. Valour, H. Mouaziz, J. Pelletier, Y. Chevalier, Pickering w/o emulsions: Drug release and topical delivery, International Journal of Pharmaceutics 368 (2009) 7-15. 12
[13] M. Zhang, L. Wei, H. Chen. Z. Du, B. P. Binks, H. Yang, Compartmentalized droplets for continuous flow liquid-liquid interface catalysis, J. Am. Chem. Soc. 138(32) (2016) 10173– 10183. [14] L. Wei, M. Zhang, X. Zhang, H. Xin, H. Yang, Pickering emulsion as an efficient platform for enzymatic reactions without stirring, ACS Sustainable Chem. Eng. 4 (12) 2016, 6838– 6843. [15] T. Skale, D. Stehl, L. Hohl, M. Kraume, R. v. Klitzing, A. Drews, Tuning Pickering Emulsions for Optimal Reaction and Filtration Conditions, Chem. Ing. Tech. 88(11) (2016) 1827–1832. [16] A.K.F. Dyab, Destabilisation of Pickering emulsions using pH, Colloids and Surfaces A: Physicochem. and Eng. Aspects 402 (2012) 2–12. [17] C.P. Whitby, F.E. Fischer, D. Fornasiero, J. Ralston, Shear-induced coalescence of oilin-water Pickering emulsions, Journal of Colloid and Interface Science 361 (2011) 170-177. [18] C.P. Whitby, M. Krebsz, Coalescence in concentrated Pickering emulsions under shear, Soft Matter 10 (2014) 4848-4854. [19] C.P. Whitby, H.K. Anwar, J. Hughes, Destabilising Pickering emulsions by drop flocculation and adhesion, Journal of Colloid and Interface Science 465 (2016) 158-164. [20] Z. Lin, Z. Zhang, Y. Li, Y. Deng, Magnetic nano-Fe3O4 stabilized Pickering emulsion liquid membrane for selective extraction and separation, Chemical Engineering Journal 288 (2016) 305-311. [21] P. Lipp, C.H. Lee, A.G. Fane, C.J.D. Fell, A fundamental Study of the ultrafiltration of oilwater emulsions, Journal of Membrane Science 36 (1988) 161-177. [22] M. Matos, G. Gutiérrez, A. Lobo. J. Coca, C. Pazos, Surfactant effect on the ultrafiltration of oil-in-water emulsions using ceramic membranes, Journal of Membrane Science 520 (2016) 749-759. [23] X.S. Yi, S.L. Yu, W.X. Shi, N. Sun, L.M. Jin, S. Wang, B. Zhang, C. Ma, L.P. Sun, The influence of important factors on ultrafiltration of oil/water emulsion using PVDF membrane modified by nano-sized TiO2/Al2O3, Desalination 281 (2011) 179-184. [24] X.S. Yi, S.L. Yu, W.X. Shi, S. Wang, N. Sun, L.M. Jin, C. Ma, Estimation of fouling stages in separation of oil/water emulsion using nano-particles Al2O3/TiO2 modified PVDF membranes, Desalination 319 (2013) 38-46. [25] T.F. Headen, S.M. Clarke, A. Perdigon, G.H. Meeten, J.D. Sherwood, M. Aston, Filtration of deformable emulsion drops, Colloid and Interface Science 304 (2006) 562-565. [26] E.N. Tummons, V.V. Tarabara, J.W. Chew, A.G. Fane, Behavior of oil droplets at the membrane surface during crossflow microfiltration of oil-water emulsions, Journal of Membrane Science 500 (2016) 211-224. [27] B. Chakrabaerty, A.K. Ghoshal, M.K. Purkait, Ultrafiltration of stable oil-in-water emulsion by polysulfone membrane, Journal of Membrane Science 325 (2008) 427-437. 13
[28] B. Chakrabaerty, A.K. Ghoshal, M.K. Purkait, Cross-flow ultrafiltration of stable oil-inwater emulsion, Chemical Engineering Journal 165 (2010) 447-456. [29] R.Jamshidi Gohari, F. Korminouri. W.J. Lau. A.F. Ismail, T. Matsuura, M.N.K. Chowdhury, E. Halakoo, M.S. Jamshidi Gohari, A novel super-hydrophilic PSf/HAO nanocomposite ultrafiltration membrane for efficient separation of oil/water emulsion, Separation and Purification Technology 150 (2015) 13-20. [29] W. Chen, J. Peng, Y. Su, L. Zheng, L. Wang, Z. Jiang, Separation of oil/water emulsion using Pluronic F127 modified polyethersulfone ultrafiltration membranes, Separation and Purification Technology 66 (2009) 591-597. [30] D.J. Miller, S. Kasemset, L. Wang, D. R. Paul, B. D. Freeman, Constant flux crossflow filtration evaluation of surface-modified fouling-resistant membranes, Journal of Membranes Science 452 (2014) 171-183. [31] L.T. Choong, Y.-M. Lin, G.C. Rutledge, Separation of oil-in-water emulsions using electrospunfiber membranes and modeling of the fouling mechanism, Journal of Membrane Science 486 (2015) 229-238. [32] T. Raijasekhar, M. Trinadh, P.V. Babu, A.V. S. Sainath, A.V.R. Reddy, Oil-water emulsion separation using ultrafiltration membranes based on novel blends of poly(vinylidene fluoride) and amphiphilic tri-block copolymer containing carboxylic acid functional group, Journal of Membrane Science 481 (2015) 82-93. [33] F. Liu, N.A. Hashim, Y. Liu, M.R. Moghareh Abed, K. Li, Progress in the production and modification of PVDF membranes, Journal of Membrane Science 375 (2011) 1-27. [34] J. Wei, G.S. Helm, N. Corner-Walker, X. Hou, Characterization of a non-fouling ultrafiltration membrane, Desalinisation 192 (2006) 252-261. [35] C. Pagliero, N. Ochoa, J. Marchese, M. Mattea, Degumming of crude soybean oil by ultrafiltration using polymeric membranes, JAOCS 78 (2001) 793-796. [36] R.W. Lencki, S. Williams, Effect of nonaqueous solvents on the flux behavior of ultrafiltration membranes, Journal of Membrane Science 101 (1995) 43-51. [37] L. Giorno. R. Mazzei, M. Oriolo, G. De Luca, M. Davoli, E. Drioli, Effects of organic solvents on ultrafiltration polyamide membranes for the preparation of oil-in-water emulsions, Journal of Colloid and Interface Science 287 (2005) 612-623. [38] F.M. Penha, K. Rezzadori, M.C. Proner, V. Zanatta, G. Zin, D.W. Tondo, J.V. de Oliveira, J.C.C. Petrus, Influence of different solvent and time of pre-treatment on commercial polymeric ultrafiltration membranes applied to non-aqueous solvent permeation, European Polymer Journal 66 (2015) 492-501. [39] D. Stehl, L. Hohl, M. Schmidt, J. Hübner, M. Lehmann, M. Kraume, R. Schomäcker, R. von Klitzing, Characteristics of stable Pickering emulsions under process conditions, Chem. Ing. Tech. 88 (2016) 1806-1814.
14
[40] S. Maaß, J. Rojahn, R. Hänsch, M. Kraume, Automated drop detection using image analysis for online particle size monitoring in multiphase systems, Comput. Chem. Eng. 45 (2012) 27-37. Highlights:
Separation of w/o Pickering emulsions by ultrafiltration is possible
Pickering emulsions can be concentrated up to very high dispersed phase fractions
Results promise possibility of continuous (bio)catalysis in Pickering emulsions
Unexpected disproportionate permeability behavior
15
16
17
18
19
20
21
22
www.elsevier.com/locate/memsci
PII: DOI: Reference:
S0376-7388(17)30327-7 http://dx.doi.org/10.1016/j.memsci.2017.04.006 MEMSCI15170
To appear in: Journal of Membrane Science Received date: 2 February 2017 Revised date: 4 April 2017 Accepted date: 5 April 2017 Cite this article as: Tina Skale, Lena Hohl, Matthias Kraume and Anja Drews, Feasibility of w/o Pickering emulsion ultrafiltration, Journal of Membrane Science, http://dx.doi.org/10.1016/j.memsci.2017.04.006 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Feasibility of w/o Pickering emulsion ultrafiltration Tina Skale1*, Lena Hohl2, Matthias Kraume2, Anja Drews1 1
HTW Berlin, Dept. 2, Process Engineering in Life Science Engineering, Wilhelminenhofstr. 75A, 12459 Berlin, Germany 2
Technische Universität Berlin, Fak. 3, Process and Chemical Engineering, Fraunhoferstr. 33-36, 10587 Berlin, Germany *Corresponding author: Tina Skale ([email protected]; +49 (0)3050193605) Abstract: This study shows the possibility of separating w/o Pickering emulsions (PE) via ultrafiltration and consequently enables a continuous process concept for catalysed L/L reactions in Pickering emulsions which are currently receiving increased attention. In two types of filtration experiments, the stability of PE against coalescence caused by the applied filtration pressure and stirring was shown. The PE could be concentrated up to 80 % water phase fraction. In pressure stepping experiments, permeabilities between 3 and 10 L/(m2 h bar) could be achieved depending on the drop size distribution of the PE. For the used 1 kDa PVDF membrane, an unexpected overproportional behaviour of the fluxes at higher pressures was observed. The presented results show that Pickering emulsions can be regarded as a promising alternative to conventional dispersions also in continuous L/L catalysis.
Keywords: Pickering emulsion, ultrafiltration, drop size distribution, organic solvents
1
Introduction
Pickering emulsions (PE) are emulsions in which the dispersed phase is stabilised by solid particles as opposed to surfactants [1]. They were named after S.U. Pickering who in 1907 described them first [2]. It took 80 years, however, until Pickering emulsions were rediscovered for applications in the pharmaceutical, cosmetics and food industries because of their high resistance against coalescence without the need for using harmful surfactants [3]. Since 2011, Pickering emulsions have been studied as possible systems for reactions in liquid-liquid (L/L) multiphase systems [4]. Chemical and enzymatic catalysts are often soluble in the aqueous phase while the reactant and the product are in the organic phase. In their minireview from 2015, Pera-Titus et al. [5] gave an overview about Pickering interfacial catalysis systems. According to the authors, despite the potential for applications in industrial processes, the aspect of catalyst recovery or recycling which would be required for an 1
economically feasible process has not been solved yet [5]. The aim of this work thus is to tackle this challenge and to investigate if membrane filtration for L/L separation of w/o Pickering emulsions and thereby catalyst recycling is possible. While the separation of surfactant emulsions by membrane filtration has been studied frequently, separating Pickering emulsions using membranes has not been done before. The advantage of phase separation via membrane filtration is that the medium does not change in a chemical, thermal or biological way. To approach this task, first, a solvent resistant hydrophobic ultrafiltration membrane has to be found to which the used nanoparticles have less affinity than to the L/L interface. The second question to be solved is if the Pickering emulsions are stable enough against coalescence during ultrafiltration. If the filtration is possible, the influence of parameters that could affect the PE stability like solid particle content and drop size distribution on the filtration behaviour has to be studied. And finally, for the recirculation in a continuous industrial process the maximal possible concentration of PE in the reactor needs to be determined. As a model system for the industrially relevant rhodium catalysed hydroformylation of alkenes, 1-dodecene as a representative of longchain olefins is used as the reactant. The catalyst is dissolved in the water drops that are stabilised by silica particles. The envisaged separation principle is shown in Figure 1. Figure 1: Separation principle of w/o Pickering emulsion (PE)
2
State of the Art
2.1 Applications and properties of PE and their stabilising nanoparticles In the early 2000s, Pickering emulsions came more and more into the focus of investigations because of their high stability against coalescence and other promising chemical properties. Pickering emulsions are more resistant against changing chemical parameters like, e.g., pH and salt content [3]. Applying functional solid nanoparticles, Pickering emulsions can also be made temperature-sensitive or pH-sensitive as required [6]. Much research has been done, e.g., on the influence of particle content and particle properties on the coalescence behaviour [7] and on the rheology of Pickering emulsions [8-11]. Binks et al. [1] gave an overview about particles as stabilizers compared to surfactants. They showed that the formation of either an o/w or w/o emulsion depends on the hydrophobicity of the solid particles, because this is related to the wettability and thereby to the extent to which the particles penetrate each liquid phase. They also found that with increasing particle hydrophobicity and water phase fraction, a w/o emulsion inverted to an o/w emulsion. The hydrophobicity of the particles also influences the stability against coalescence. Particles that are very hydrophobic or very hydrophilic form emulsions with larger drops over 100 µm that are unstable against coalescence [7]. To obtain a coalescence stable emulsion and smaller drop diameters, the stabilizing particles have to have an intermediate hydrophobicity. The size of the solid stabilized drops also depends on the size of the solid particles. With nanoparticles of about 100 nm, micron-sized droplets can be stabilized. Silica nanoparticles are often used because they can easily be surface modified to make them hydrophobic [6]. Since harmful surfactants which can lead to skin irritations can be avoided, a large advantage of PE applications in the cosmetics and pharmaceutical industries is seen [6]. For industrial applications, Frelichowska et al. [12] compared the drug release of a w/o Pickering emulsion with that of a classical emulsifier-stabilized emulsion. The Pickering emulsions showed higher permeation rates of 2
caffeine through the skin because of the stronger adhesion of the water droplets on the skin surface. Further information on the history and properties of Pickering emulsions and their applications in industry can be found in a review from 2013 [6].
2.2 L/L Separation of (Pickering) Emulsions To the best of our knowledge, there are barely any studies about separating Pickering emulsions which would be required to withdraw the product in a continuous process and to reuse the catalyst [13-15]. Zhang et al. recently applied PE in the form of a packed bed for different continuous reactions. However, with 100 h and more, the residence times they could realise with this set-up were very high and would be inefficient for the hydroformylation studied here. To decrease the residence time, the pressure would have to be increased which would most likely deform the drops in the packed bed, in turn would decrease the flow rate even further (self-accelerating problem) and might even lead to coalescence. In addition, once packed inside a column, drops could not be exchanged or renewed when catalyst activity decreases. Demulsification by centrifugation with successive redispersion for the L/L separation of reactive Pickering emulsions was studied by Wei et al. [14]. This, however, would require repeated power input. In addition to centrifugation, typical demulsification techniques of conventional emulsions include the addition of chemical demulsifiers, pH adjustment, gravity settling, filtration, heat treatment and electrostatic demulsification [16]. Dyab studied the destabilisation of w/o silica Pickering emulsions by changing the pH which changes the wettability of the silica particles. However, for applications of Pickering emulsions as a reaction system, changing the pH could be adverse. [17-19] pursue the strategy to separate Pickering emulsions by controlled shear-induced coalescence and as a result interrupting the network between particles and drops. Another approach for the separation of Pickering emulsions is the use of magnetic particles for the stabilisation and to separate the emulsion drops from the continuous liquid via a magnet [20], but there is no literature about the influence of magnetic particles on the reaction system and the industrial scale realisation. Therefore, in this study, membrane filtration for the continuous separation of w/o Pickering emulsions as a less energy intensive method which enables drop exchange and shorter residence times, and which does not change the reaction environment was investigated. The use of membranes for L/L separation of surfactant stabilised emulsions has been studied frequently before [e.g. 21,22]. During ultrafiltration of surfactant stabilized oil in water (o/w) emulsions at pressures between 1 and 2.5 bar, the water drops were found to coalesce in the gel-layer possibly due to viscous drag caused by the permeate flow [21]. Another explanation for the coalescence in the gel-layer could be the loss of surfactant to the permeate. There, both smaller and larger drops than present in the gel-layer were found which implied that oil drops passed through the membrane and during this passage broke and coalesced. During concentrating experiments of o/w emulsions the fluxes hardly showed any changes in the beginning (initial oil fractions of 5 vol%) but when oil fractions of around 10 vol% (depending on the oil type) were reached, the fluxes decreased strongly to a maximum possible oil fraction of about 30 % oil phase [21]. Recent studies showed that during ultrafiltration of surfactant stabilized o/w emulsions with ceramic membranes, the permeate flux depended on the charge of the membrane surface and the emulsion drops. 3
Electrostatic attraction effected an adsorption of the surfactant on the membrane surface and the surfactant could pass the membrane. The electrostatic repulsion caused the opposite, the surfactant could not adsorb which prevented its membrane passage and led to decreased concentration polarisation and cake layer formation [22]. Most literature on membrane filtration of emulsions could be found for o/w emulsions using hydrophilic membranes. To increase the permeate flux without increasing the membrane fouling, Yi et al. tested a modified PVDF membrane for the separation of a surfactant stabilised oil in water emulsion. With increasing oil concentration, the permeate flux (normalised with the initial flux) decreased sharply. The used fouled membranes showed a high flux recovery of 88% and 94% after having been washed with pure water [23]. Using their modified PVDF ultrafiltration membranes, Yi et al. found that for o/w emulsions the cake filtration model did not predict the filtration behaviour over the whole filtration time but was followed by intermediate pore blocking [24]. Headen et al. found that oil drops from a model o/w emulsion deformed with increasing filtration pressure, but nevertheless the filter cake stayed permeable [25]. Tummons et al. developed a method to get real-time images of the membrane surface during microfiltration of o/w emulsions. With this method they could observe three characteristic stages of membrane fouling by oil. At first, the droplets attached and clustered, then they deformed and thirdly coalesced. Varying the surface tension by increasing the surfactant concentration showed a change in the drop sizes followed by a change in the coalescence behaviour. The authors concluded that the membrane fouling depends on droplet coalescence and crossflow shear. Oil droplets with diameters slightly above the pore diameter permeated through the membrane [26].
2.3 Membranes in oil and w/o applications The vast majority of studies about membrane filtration deals with aqueous systems and fewer with oily/organic systems. The literature on oily systems narrows down to o/w emulsions, e.g., the cleaning of oily wastewater with ultrafiltration and microfiltration [27-32]. Since the advent of organic solvent nanofiltration, several solvent resistant nanofiltration membranes have become commercially available, but commercial ultrafiltration membranes are still mostly designed for aqueous applications. This narrows the choice of possible membrane materials for the ultrafiltration of the w/o Pickering emulsions. Some companies (MICRODYN-NADIR GmbH Germany) offer membranes with solvent resistant material for microfiltration. Alfa Laval’s ETNA series is composed of an active PVDF layer on a cellulose support. PVDF has a high hydrophobicity and resistance against most solvents [33], but because of the usual usage of ultrafiltration in aqueous systems like for wastewater treatment or filtration of protein solutions, the hydrophobic PVDF membranes were mostly hydrophilized to avoid membrane fouling [33,34]. In the cleaning of phospholipids from a crude soybean oil/hexane mixture, PVDF permeate fluxes were up to threefold larger than the fluxes of a polyimide membrane under the same operating conditions [35]. Lencki and Williams studied the influence of non-aqueous solvents on ultrafiltration and found that while some solvents increased the permeability, others decreased it. They stated that the resistance of polymeric membranes depends on the membrane structure (isotropic/ anisotropic) and on the interactions between solvent and membrane (solubility parameter) [36]. 4
For the preparation of o/w emulsions via membrane emulsification, Giorno et al. studied the effect of different solvents on polymeric tubular ultrafiltration membranes of different MWCO [37]. Depending on the solvent and the MWCO, the pore structure changed, so in order to achieve reproducible flux levels, an individual optimal pre-treatment of the used membrane and solvent needs to be found which, due to the frequent use of membranes in aqueous media, has not been investigated exhaustively yet. Penha et al. studied the influence of different solvents and pre-treatment durations on commercial polymeric ultrafiltration membranes [38]. They found that the permeability of the polymeric membrane depends not only on the pore size but also on the interaction between solvent and polymer. An increase in permeate fluxes was reached with an increase of the immersion times in a combination of different solvents. This special treatment of the membranes increased the contact angle and increased the hydrophobicity of all tested membranes without changing the membrane structure as described by Lencki and Williams [36] and Giorno et al. [37].
3 3.1
Materials and methods Preparation of the emulsion
As a model system, water-in-dodecene Pickering emulsions were investigated. 30 mL of emulsion were prepared for one batch which consisted of 1-dodecene with a purity of >94% by Merck Chemicals GmbH, Germany and deionised water (10.0 µS/cm at 25 °C) at a volumetric phase ratio of 3:1 (dodecene:water). The emulsion was stabilised with 0.5 wt% (with respect to the total mass of the emulsion) of hydrophobic silica nanoparticles HDK H20 by Wacker Chemie AG, Germany. The particles form elongated aggregates with a size of about 10 x 150 nm [38] and have a residual silanol content of 50%. First, the silica particles were wetted completely with 1-dodecene. Then, the deionised water was added. The still separated mixture was emulsified using an ultrasonic homogeniser (Bandelin HD 3200, Germany) for 5 minutes under manual stirring in an ice bath. To prepare emulsions with different drop size distributions, the amplitude and the pulse time of the ultrasonic homogeniser were varied. The amplitude and thus the energy input was reduced from 50 to 15% to achieve larger drops. In experiments on the influence of the solvent type on PE filtration behaviour, 1-dodecene was replaced by n-decane (99 %, Thermo Fisher (Kandel) GmbH, Germany), 1-decene (>96 %, Merck Chemicals GmbH, Germany) or toluene (99.9 %, Carl Roth GmbH + Co. KG, Germany). 3.2
Characterisation of the emulsions
To check the reproducibility of preparing the emulsion with the ultrasonic homogeniser and to characterise the influence of filtration on the emulsion, drop sizes were measured before and after the filtration experiments. 25 pictures (five pictures of each of the five samples of each emulsion) were made by microscopy (Carl Zeiss AG, Germany, Axio Imager 2). The pictures were analysed manually and by an image evaluation software (Sopat GmbH, Germany). The number of counted drops per emulsion was between 2000 and 6000 drops for the image evaluation software and between 600 and 6000 drops for the manual analysis. 600 drops were found to be statistically sufficient because the Sauter mean diameter stayed constant despite further counted drops.
5
3.3
Pre-treatment of the membrane
For the filtration experiments, a hydrophilized PVDF membrane (Alfa Laval Denmark, ETNA 01PP) with a MWCO of 1 kDa was used [34]. Other solvent resistant membranes (PuraMemS600 by Evonik, Germany, and GR95PP by Alfa Laval, Denmark) showed no relevant 1-dodecene fluxes at the applied pressure and were therefore not used in further trials. Despite micrometre scale emulsion droplets, a low MWCO membrane was chosen to also retain possible free silica agglomerates. For the pre-compaction of the membrane and to ensure complete wetting, the membranes were pre-treated before every pressure stepping experiment by washing with pure solvent for 90 min (unless indicated otherwise) at 2.5 bar. A new membrane was used for each experiment.
Figure 2: Experimental set up for pressure stepping experiments at constant phase fraction
3.4
Experimental set-up
For all filtration experiments, a solvent resistant stirred cell XFUF04701 by Merck Millipore with a working volume of 91.5 mL was used. 3.4.1 Concentration experiments The objective of the concentrating experiments was to find out to what extent the emulsions could be concentrated, i.e., at which L/L ratio drops would break and water would begin to pass through the membrane. The pressure p was set to 2.5 bar and stirrer speed to 500 rpm which corresponds to wtip = 1 m/s. The flux was logged by reading an electronic balance. 3.4.2 Pressure stepping experiments with constant phase fraction To maintain a constant phase ratio while analysing the influence of pressure, 1-dodecene was continuously dosed into the stirred cell from a surge tank (see Figure 2). When an almost constant flux was achieved, the pressure was increased in a stepwise manner from 1 to 4 bar after which it was decreased again from 4 to 1 bar. Each pressure level was kept for 1 h during pressure raise and 30 min during pressure descent. The flux was logged by reading an electronic balance.
4 4.1
Results and discussion Reproducibility of the emulsion preparation method
Prior to membrane filtration experiments, the reproducibility of PE properties is a prerequisite. For a validation of the required drop size distribution measurements, see 4.2. Figure 3 (left) shows the cumulative distribution function (cdf) Q0 of the drop diameters dd for 3 emulsions prepared separately but with the same settings of the ultrasonic homogeniser (15 % amplitude, 2 s pulse on and 3 s pulse off for 5 min). The drop size distributions were nearly the same with little differences in the smallest and the biggest drop sizes. To avoid these little differences and to increase the reproducibility of the filtration experiments, the separately prepared emulsions were mixed and separated again into 30 mL batches. The results of the drop size measurements are shown in Figure 3 (right). The cumulative distribution functions still show little differences in the medium drop sizes. The maximum deviation between the three individually prepared emulsions was 2.02 µm for the arithmetic 6
mean diameter d1.0 and 4.39 µm for the Sauter mean diameter d3.2. In contrast, the difference between the determined minimum and the maximum arithmetic mean diameter of the mixed emulsion samples was only 0.24 µm and 3.25 µm for the Sauter mean diameter. Hence, for further filtration experiments, batches of PE were always mixed. n
d 3.2
d
3 di
d
2 di
i 1 n i 1
(4.1)
n
d1.0
d i 1
di
(4.2)
n
Figure 3: Reproducibility of the ultrasonic homogeniser preparation: left: cdf of 3 separately prepared emulsions, and right: cdf of 3 separately prepared and then mixed emulsions
Figure 4 (left) shows the drop size distribution of two PEs prepared with equal settings but with two different ultrasonic homogenisers of different power. The emulsion prepared with the low power device (HD70, 70 W) contained both a higher amount of larger drops and more small drops below 10 µm. The PE was highly polydisperse with a Sauter mean diameter d3.2 of 18.35 µm. The emulsion prepared with HD3200 (200 W) was totally different, almost monodisperse with almost halved drop sizes (d3.2 = 11.18 µm). To visualise the difference of the two emulsions, microscope images are also shown in Figure 4 (right). With a variation in the type and in the settings of the ultrasonic homogenizer, the drop sizes and the drop size distribution could be influenced to achieve the desired wide or narrow distribution for studying their influence on flux and their stability during filtration.
Figure 4: left: drop size distribution of two emulsions prepared with different ultrasonic homogenisers with different power, right: microscopic pictures of the two differently produced emulsions
4.2
Characterisation of the emulsion
In order to investigate the influence of drop size distribution on filtration behaviour and of filtration on drop size distribution, the drop sizes have to be measured. To facilitate the image evaluation and to reduce the possibly subjective error of the manual measurements, an automatic image evaluation software by SOPAT GmbH (Germany) was used. To decide whether the software can be applied to evaluate these pictures, the two measurement methods were compared by measuring the drop sizes of the same pictures manually and with the software. The results are shown in Figure 5 for two different emulsions. The cumulative density functions of the emulsions were averaged from at least three (left) and four (right) individually prepared emulsions. Table 1 shows the total number of counted drops for both measurement methods. 7
Figure 5: Comparison of the two drop size distribution measurement methods (manual and software), left: 15% amplitude PEs and right: 50% amplitude PEs
For the 50% amplitude emulsion, the number of the automatically counted drops was nearly as high as the number of the manually counted drops (same order of magnitude). The plotted cumulative density functions of both measurement methods were nearly the same with only little deviations. The drop size distributions for the 15% amplitude emulsions showed small variations in the small and medium drop sizes. However, the deviation between software and automatic evaluation of the Sauter mean diameter was only 0.62 µm and of the arithmetic mean diameter even only 0.12 µm. For the less polydisperse 50 % amplitude emulsion, the difference of both mean diameters was below 1 µm. Since the maximum deviation between manual and automatic analysis is less than 0.7 µm in both the Sauter and the arithmetic mean diameter, the automatic evaluation can be applied for the characterisation of the emulsions. The accuracy and resolution of the automated image analysis software was studied by repeating the search parameter estimation of the same microscopic pictures. A deviation of ± 1 µm from the Sauter mean diameter was normally observed [39,40]. Table 1: Number of counted drops of the two measurement methods for two different emulsion types (different drop size distributions because of two different energy inputs during the PE preparation) and average drop diameters
15% amplitude no. of counted drops [-] d1.0 [µm] d3.2 [µm] 4.3
manual 1972 9.63 15.43
software 19609 9.75 14.81
50% amplitude manual 16924 8.72 10.7
software 32560 9.41 11.18
Pre-treatment of the membranes
The results of the membrane washing with 1-dodecene showed a rapid flux decline within the first 20 minutes from over 180 L/(m2 h) to about 15 L/(m2 h). This rapid flux decline could be explained by the anisotropic structure of the ETNA01PP membrane where the surface layer has another swelling behaviour than the support layer. According to [36], the active layer swells more than the support layer which would lead to an increased membrane resistance. This change in membrane structure caused by the solvent seemed to be different for the used membrane samples, so to allow reproducible filtration experiments, membranes with strong deviations (±100%) from the average washing flux were discarded. This flux variation despite the same pre-treatment is a typical behaviour of ultrafiltration membranes [36]. The deviations in the pre-treatment of the membranes followed no pattern (e.g., position on the membrane sheet from which samples were cut). Membranes with high washing fluxes did not have higher fluxes in the pressure stepping experiments than membranes with lower washing fluxes. This might be caused by the dominating resistance of the filter cake. 4.4
Concentration experiments
Figure 6 shows the results of the concentration experiments. The flux is plotted against the current aqueous phase fraction in the stirred cell for three 15 % amplitude emulsions and for different membrane pre-treatment times. The aqueous phase fraction of the emulsion in the stirred cell φw was calculated according to eqs. (4.3) and (4.4.) from the initial dodecene 8
volume of the emulsion Vdod,0 and the accumulated permeate mass measured on the balance at each point in time mdod(t). As long as no water passes through the membrane, the water volume VW remains constant throughout the experiment.
j w (t ) =
Vw (Vw +Vdod,0 -Vdod (t))
Vdod (t) =
(4.3)
mdod (t)
(4.4)
rdod
The wetting of the membrane influenced the maximum concentration point. After 90 minutes of pre-treatment, the phase fraction could be increased from 22 % to about 80 %. This maximum phase fraction could not be increased by longer pre-treatment of 120 minutes. The individual flux levels of the concentration experiments with dry membranes differ from each other and show steeper decreases during the experiments. The pre-treatment led to more reproducible and stable fluxes and was therefore applied in pressure-stepping experiments. The achieved maximum phase fraction of about 80 %, which is much higher than the maximum phase fraction achieved in surfactant emulsions [21], shows the high stability of the Pickering emulsion against pressure, shear force and changing phase ratio, and allows compact reactors and membrane modules to be used.
Figure 6: Three 1-dodecene PE concentration experiments each after different membrane pre-treatment times (0, 20, 90 and 120 min)
4.5
Pressure stepping experiments
Figure 7 (top left) shows results of the pressure stepping experiments with two Pickering emulsions prepared with HD70 and 50 % amplitude in comparison to two pure 1-dodecene experiments. Apart from a slightly larger deviation at 4 bar, experiments can be seen to be reasonably reproducible. The pure solvent experiments show the expected disproportional behaviour at higher pressure due to changing resistances caused by, for example, membrane compaction. However, the results of the PE filtration were unexpected: The fluxes of the PE were higher than the fluxes of the pure 1-dodecene experiments at 3 and 4 bar and increased disproportionally high at higher pressures. To explain these unexpected results, three hypotheses were investigated. One was that abrasion of the membrane surface occurs caused by the silica particles. Figure 8 shows SEM pictures of the fresh membrane surface and two membrane surfaces after the PE pressure stepping experiments. No abrasion of the membrane surface but only a silica layer located in the region below the stirrer could be observed.
9
Figure 7: Pressure stepping experiments with constant phase fraction for four different solvents (1dodecene, 1-decene, n-decane and toluene, pure solvent and PE, respectively)
A second assumption was that residual traces of silica particles from previous experiments (despite thorough cleaning) in the stirred cell could block the membrane surface during the pressure stepping experiments with pure 1-dodecene. During the experiments with the PE, the water drops would adsorb these residual silica particles thus freeing the membrane surface. However, repeating the pure solvent experiments in a new silica-free stirred cell resulted in the same lower flux levels. So higher resistances due to silica on the membrane can be ruled out as the reason for the low pure solvent fluxes.
Figure 8: SEM pictures of fresh and used membrane samples
To investigate if the lower pure solvent fluxes are specific to the used 1-dodecene because of, e.g., stabilisers present in the 1-dodecene, other solvents that are similar to 1-dodecene in chain length and hydrophobicity but have different purities were tested. Figure 7 (bottom left) shows the results of two experiments with PE prepared with n-decane and two experiments with pure n-decane. The fluxes of the PE with n-decane also show the disproportionately high permeability at higher pressure, but in this case, the fluxes of the pure n-decane were higher than the fluxes of the PE at every pressure step. For toluene and 1decane (Figure 7, bottom), similar flux levels for pure solvent and PE filtration were found. Little differences could be caused by intrinsic variations of the membrane samples. In conclusion, the unexpected much higher fluxes of the PE prepared with 1-dodecene compared to the pure solvent seem to be specific to 1-dodecene and the used membrane. For all solvents it could be seen that the flux differences during pressure raise and pressure descent were larger for the pure solvents than for the PE. This could also been seen in the raw data (not shown). The fluxes of the PE filtration reached the steady state faster than the fluxes of the pure solvent where the fluxes only reached steady state during pressure decrease. Among the PEs prepared with different solvents, the 1-decene PE was the least stable emulsion, so during the pressure stepping experiments, small amounts of water continuously broke through the membrane. Nevertheless, the fluxes showed the same hysteresis as the fluxes of the stable emulsions. One possible explanation for the disproportionate increase of the fluxes at higher pressure could be the coalescence of the water drops and the resulting larger hydraulic diameter of 10
the filter cake. In our previous study, drop coalescence during the pressure-stepping experiments was observed [15]. The extent of drop coalescence was found to depend on the drop size distribution and the stability of the Pickering emulsion [15]. Figure 9 shows the results of the pressure stepping experiments with PEs of different drop size distributions. The maximum deviation in the cdf can be seen between the PE prepared with HD70 and the emulsions prepared with HD3200. The higher fluxes of the PE prepared with HD70 might have been caused by the bigger drops and the consequently higher hydraulic diameter of the filter cake layer. The reduction of the power input by an amplitude reduction from 50% to 15% (HD3200) hardly changed the drop sizes. An influence of the drop size distribution on filtration could only be seen for emulsions with many larger drops such as the emulsion prepared with HD70 and an emulsion with a narrow drop size distribution with smaller drop sizes such as the emulsions prepared with HD3200. The smaller amounts of larger drops as present in the 15% emulsion did not influence the flux level significantly compared to the emulsion prepared at 50% amplitude. Therefore, the larger drops might coalesce at higher pressure so that the fluxes become higher, but a critical amount of larger drops is needed before a significant influence on the flux levels can be seen. Also, the observed differences in the washing fluxes of the membranes made it harder to discover fine differences in the flux levels and therefore in the influence of changing drop sizes. However, one obvious difference caused by the different drop sizes was that small water drops passed through the membrane during the pressure stepping experiments with the PE prepared with HD3200 and 50% amplitude. It seemed that because of the low hydrophobicity of the silica particles, a phase inversion and a formation of a double emulsion during the permeation of water, solvent and particles through the pores occurred. Small water drops containing several even smaller drops could be observed in the permeate.
Figure 9: Pressure stepping experiments with 1-dodecene PE with different drop size distributions, bottom right: average cdf of the PE used for these filtration runs
5
Conclusions
For the first time it could be shown that L/L separation of different w/o Pickering emulsions by membrane filtration is possible. The two liquid phases could be separated without destroying either the membrane or the emulsion. The nanoparticles did not adhere to nor block the membrane. The Pickering emulsions showed an unexpected flux behaviour with permeability increasing with higher pressures. Fluxes between 3 and 40 L/(m2 h) were achieved for pressures between 1 and 4 bar. The emulsions could be concentrated up to a water phase fraction of 80 %. The drop size distribution of the emulsion had no significant influence on the flux level except for emulsions with a larger amount of larger drops. The presented results show that Pickering emulsions can be regarded as a promising alternative to conventional dispersions also in continuous L/L catalysis. 11
6 Acknowledgement Financial support by the German Research Foundation DFG (collaborative research centre "Integrated Chemical Processes in Liquid Multiphase Systems" InPROMPT TRR63, B6) is gratefully acknowledged. We thank Alfa Laval for kindly providing free membrane samples.
7
References
[1] B.P. Binks, Particles as surfactants – similarities and differences, Current Opinion in Colloid & Interface Science 7 (2002) 21-41. [2] S.U. Pickering, Emulsions, J. Chem. Soc. 91 (1907) 2001-2021. [3] G. Lagaly, O. Schulz, R. Zimehl, Dispersionen und Emulsionen, Springer-Verlag, Berlin Heidelberg, 1997. [4] C. Wu, S. Bai, M.B. Ansorge-Schumacher, D. Wang, Nanoparticles Cages for Enzyme Catalysis in Organic Media, Adv. Mater. 23 (2011) 5694-5699. [5] M. Pera-Titus, L. Leclercq, J.-M. Clacens, F. de Campo, V. Nadrello-Rataj, Pickering Interfacial Catalysis for Biphasic Sytems: From Emulsion Design to Green Reactions, Angew. Chem. Int. Ed. 54 (2015) 2006-2021. [6] Y. Chevalier, M.-A. Bolzinger, Emulsions stabilized with solid nanoparticles: Pickering emulsions. Colloids and Surfaces A: Physicochemical and Engineering Aspects 439 (2013) 23-34. [7] B.P. Binks, S.O. Lumsdon, Influence of Particle Wettability on the Type and Stability of Surfactant-Free Emulsions, Langmuir 16 (2000) 8622-8631. [8] J. Chen, R. Vogel, S. Werner, G. Heinrich, D. Clausse, V. Dutschk, Influence of the particle type on the rheological behavior of Pickering emulsions, Colloids and Surfaces A: Physicochemical and Engineering Aspects 382 (2011) 238-245. [9] T. Fuma, M. Kawaguchi, Rheological response of Pickering emulsions prepared using colloidal hydrophilic silica particles in the presence of NaCl, Colloids and Surfaces A: Physicochemical and Engineering Aspects 465 (2015) 168-174. [10] A. Nesterenko, A. Drelich, H. Lu, D. Clausse, I. Pezron, Influence of a mixed particle/Surfactant emulsifier system on water-in-oil emulsion stability, Colloids and Surfaces A: Physicochemical and Engineering Aspects 457 (2014) 49-57. [11] L. Hohl, S. Röhl, D. Stehl, R. v. Klitzing, M. Kraume, Influence of nanoparticles and drop size distributions on the rheology of w/o Pickering emulsions, Chem. Ing. Tech. 88 (2016) 1815-1826. [12] J. Frelichowska, M.-A. Bolzinger, J.-P. Valour, H. Mouaziz, J. Pelletier, Y. Chevalier, Pickering w/o emulsions: Drug release and topical delivery, International Journal of Pharmaceutics 368 (2009) 7-15. 12
[13] M. Zhang, L. Wei, H. Chen. Z. Du, B. P. Binks, H. Yang, Compartmentalized droplets for continuous flow liquid-liquid interface catalysis, J. Am. Chem. Soc. 138(32) (2016) 10173– 10183. [14] L. Wei, M. Zhang, X. Zhang, H. Xin, H. Yang, Pickering emulsion as an efficient platform for enzymatic reactions without stirring, ACS Sustainable Chem. Eng. 4 (12) 2016, 6838– 6843. [15] T. Skale, D. Stehl, L. Hohl, M. Kraume, R. v. Klitzing, A. Drews, Tuning Pickering Emulsions for Optimal Reaction and Filtration Conditions, Chem. Ing. Tech. 88(11) (2016) 1827–1832. [16] A.K.F. Dyab, Destabilisation of Pickering emulsions using pH, Colloids and Surfaces A: Physicochem. and Eng. Aspects 402 (2012) 2–12. [17] C.P. Whitby, F.E. Fischer, D. Fornasiero, J. Ralston, Shear-induced coalescence of oilin-water Pickering emulsions, Journal of Colloid and Interface Science 361 (2011) 170-177. [18] C.P. Whitby, M. Krebsz, Coalescence in concentrated Pickering emulsions under shear, Soft Matter 10 (2014) 4848-4854. [19] C.P. Whitby, H.K. Anwar, J. Hughes, Destabilising Pickering emulsions by drop flocculation and adhesion, Journal of Colloid and Interface Science 465 (2016) 158-164. [20] Z. Lin, Z. Zhang, Y. Li, Y. Deng, Magnetic nano-Fe3O4 stabilized Pickering emulsion liquid membrane for selective extraction and separation, Chemical Engineering Journal 288 (2016) 305-311. [21] P. Lipp, C.H. Lee, A.G. Fane, C.J.D. Fell, A fundamental Study of the ultrafiltration of oilwater emulsions, Journal of Membrane Science 36 (1988) 161-177. [22] M. Matos, G. Gutiérrez, A. Lobo. J. Coca, C. Pazos, Surfactant effect on the ultrafiltration of oil-in-water emulsions using ceramic membranes, Journal of Membrane Science 520 (2016) 749-759. [23] X.S. Yi, S.L. Yu, W.X. Shi, N. Sun, L.M. Jin, S. Wang, B. Zhang, C. Ma, L.P. Sun, The influence of important factors on ultrafiltration of oil/water emulsion using PVDF membrane modified by nano-sized TiO2/Al2O3, Desalination 281 (2011) 179-184. [24] X.S. Yi, S.L. Yu, W.X. Shi, S. Wang, N. Sun, L.M. Jin, C. Ma, Estimation of fouling stages in separation of oil/water emulsion using nano-particles Al2O3/TiO2 modified PVDF membranes, Desalination 319 (2013) 38-46. [25] T.F. Headen, S.M. Clarke, A. Perdigon, G.H. Meeten, J.D. Sherwood, M. Aston, Filtration of deformable emulsion drops, Colloid and Interface Science 304 (2006) 562-565. [26] E.N. Tummons, V.V. Tarabara, J.W. Chew, A.G. Fane, Behavior of oil droplets at the membrane surface during crossflow microfiltration of oil-water emulsions, Journal of Membrane Science 500 (2016) 211-224. [27] B. Chakrabaerty, A.K. Ghoshal, M.K. Purkait, Ultrafiltration of stable oil-in-water emulsion by polysulfone membrane, Journal of Membrane Science 325 (2008) 427-437. 13
[28] B. Chakrabaerty, A.K. Ghoshal, M.K. Purkait, Cross-flow ultrafiltration of stable oil-inwater emulsion, Chemical Engineering Journal 165 (2010) 447-456. [29] R.Jamshidi Gohari, F. Korminouri. W.J. Lau. A.F. Ismail, T. Matsuura, M.N.K. Chowdhury, E. Halakoo, M.S. Jamshidi Gohari, A novel super-hydrophilic PSf/HAO nanocomposite ultrafiltration membrane for efficient separation of oil/water emulsion, Separation and Purification Technology 150 (2015) 13-20. [29] W. Chen, J. Peng, Y. Su, L. Zheng, L. Wang, Z. Jiang, Separation of oil/water emulsion using Pluronic F127 modified polyethersulfone ultrafiltration membranes, Separation and Purification Technology 66 (2009) 591-597. [30] D.J. Miller, S. Kasemset, L. Wang, D. R. Paul, B. D. Freeman, Constant flux crossflow filtration evaluation of surface-modified fouling-resistant membranes, Journal of Membranes Science 452 (2014) 171-183. [31] L.T. Choong, Y.-M. Lin, G.C. Rutledge, Separation of oil-in-water emulsions using electrospunfiber membranes and modeling of the fouling mechanism, Journal of Membrane Science 486 (2015) 229-238. [32] T. Raijasekhar, M. Trinadh, P.V. Babu, A.V. S. Sainath, A.V.R. Reddy, Oil-water emulsion separation using ultrafiltration membranes based on novel blends of poly(vinylidene fluoride) and amphiphilic tri-block copolymer containing carboxylic acid functional group, Journal of Membrane Science 481 (2015) 82-93. [33] F. Liu, N.A. Hashim, Y. Liu, M.R. Moghareh Abed, K. Li, Progress in the production and modification of PVDF membranes, Journal of Membrane Science 375 (2011) 1-27. [34] J. Wei, G.S. Helm, N. Corner-Walker, X. Hou, Characterization of a non-fouling ultrafiltration membrane, Desalinisation 192 (2006) 252-261. [35] C. Pagliero, N. Ochoa, J. Marchese, M. Mattea, Degumming of crude soybean oil by ultrafiltration using polymeric membranes, JAOCS 78 (2001) 793-796. [36] R.W. Lencki, S. Williams, Effect of nonaqueous solvents on the flux behavior of ultrafiltration membranes, Journal of Membrane Science 101 (1995) 43-51. [37] L. Giorno. R. Mazzei, M. Oriolo, G. De Luca, M. Davoli, E. Drioli, Effects of organic solvents on ultrafiltration polyamide membranes for the preparation of oil-in-water emulsions, Journal of Colloid and Interface Science 287 (2005) 612-623. [38] F.M. Penha, K. Rezzadori, M.C. Proner, V. Zanatta, G. Zin, D.W. Tondo, J.V. de Oliveira, J.C.C. Petrus, Influence of different solvent and time of pre-treatment on commercial polymeric ultrafiltration membranes applied to non-aqueous solvent permeation, European Polymer Journal 66 (2015) 492-501. [39] D. Stehl, L. Hohl, M. Schmidt, J. Hübner, M. Lehmann, M. Kraume, R. Schomäcker, R. von Klitzing, Characteristics of stable Pickering emulsions under process conditions, Chem. Ing. Tech. 88 (2016) 1806-1814.
14
[40] S. Maaß, J. Rojahn, R. Hänsch, M. Kraume, Automated drop detection using image analysis for online particle size monitoring in multiphase systems, Comput. Chem. Eng. 45 (2012) 27-37. Highlights:
Separation of w/o Pickering emulsions by ultrafiltration is possible
Pickering emulsions can be concentrated up to very high dispersed phase fractions
Results promise possibility of continuous (bio)catalysis in Pickering emulsions
Unexpected disproportionate permeability behavior
15
16
17
18
19
20
21
22