For emulsification, many micro-structured devices have been used such as membranes, and various microfluidic devices, but even the nozzles of high pressure homogenizers may qualify as such (for recent extensive reviews, see, e.g., [1
]). Various process parameters have been investigated and scaling relations have been derived [3
], and up-scaling has also been considered [4
It should be mentioned that there is a lot of contradicting information in the literature. In some cases, it is reported that a certain emulsion can be made very efficiently with a specific membrane, while other authors that use seemingly the same combination are not able to make the emulsion, or only for a very short time [5
]. Besides, there are considerable differences between the modes of operation; in some cases, cross-flow emulsification works well, while pre-mix emulsification does not work at all while trying to make the same emulsion with the same membrane; specific examples will be given in the respective sections.
We think that these discrepancies can be explained by the fact that the emulsion components interact with surfaces and interfaces (both liquid and solid), since this leads to the energetically most favorable option for the system as a whole. Because emulsification involves the creation of a lot of interface, and through that an increase in the Gibbs free energy, the surface active components will reduce the interfacial tension and therewith also the Gibbs free energy.
is the Gibbs free energy, σ is the interfacial tension, and is dA
the increase in interfacial area. This is a well-known fact, and has been discussed in many papers, also in relation to micro-structured devices, including dynamic interfacial tension effects [6
]. A good overview of surfactant effects in microfluidics was performed by Baret [9
Besides, there is another interface that plays a big role in emulsification, and that is the solid interface. What is normally postulated is that you need a membrane/microfluidic device that is non-wetting for the to-be-dispersed phase to prepare an emulsion with a certain continuous phase (e.g., [10
]). In cross-flow membrane emulsification and microfluidic devices this implies that a hydrophilic (membrane) surface is needed to make an oil-in-water emulsion and a hydrophobic (membrane) surface for water in oil emulsions. In premix emulsification, a hydrophilic surface is used to refine an oil-in-water emulsion, and a hydrophobic one for water in oil emulsification [16
]. In some rare cases, a very concentrated premix emulsion can be phase reversed by pushing it through a membrane with polarity equal to the dispersed phase [17
These rules of thumb may seem to be very easy to use, but what is mostly not considered is that if a combination of, e.g., hydrophilic surface/membrane and hydrophobic oil is used, this also implies an effect on the Gibbs free energy. The internal area of any micro-structured device can be very large, and putting this in contact with a non-wetting phase implies a large increase in Gibbs free energy. The system as a whole will strive for a reduction of the Gibbs free energy, and this also implies that there is a driving force for either surfactant-solid interface, or liquid-solid interactions. This may change the wettability of a system completely as demonstrated in experimental work by, e.g., Keurentjes et al
] for emulsion separation, and Hüsken et al
] for two-phase membrane bioreactors. Further, in simulations, the effect of a change in contact angle on emulsification with microsieves has been demonstrated to result in sticking of the to-be-dispersed phase, thereby leading to loss of droplet formation [20
An illustrative example on the rigorous effects of wettability changes on process operation can be found in the early work of one of the authors on emulsion separation with membranes. The emulsion contained the enzyme lipase that was used to hydrolyze vegetable oil into free fatty acids, and the emulsion was separated with a hydrophobic polypropylene membrane [21
]. When starting the experiment, the flux immediately decreased very rapidly due to binding of lipase that is able to form very strong interactions with any hydrophobic interface. Besides, the pressure that could be used to separate the emulsion decreased very rapidly; in first instance, only the oil phase permeated, but within a day water also permeated the membrane at very low pressure, indicating that the wettability of the membrane had changed drastically (due to adsorption and diffusion of lipase into the membrane pores [22
]). Only when the membrane was modified with a specific block copolymer that prevented lipase adsorption [23
], while still keeping its wettability [24
], the membrane could be operated at high pressure, and for long periods of time (tests of up to six weeks were performed). The underlying mechanism is schematically represented in Figure 1
; in the brush configuration, the surface has protein repelling properties due to the presence of the long hydrophilic polyethylene oxide buoy groups, while it is anchored by the hydrophobic polypropylene oxide group, which allowed retention of the hydrophobic nature of the membrane. If adsorbed onto a hydrophilic surface, the block co-polymer assumes a pancake configuration that is not protein repelling.
This effect has not only been demonstrated with enzymes, but also with two phase systems used for down-stream processing of microbial bioconversions [19
], and for surfactant containing emulsions [18
]. In one phase systems these effects are not that relevant; e.g., in membrane separation this would be termed fouling, or flux loss, but the process as such would not be lost because of component surface interactions.
This clearly indicates the importance of surfactant interactions, but in membrane and microfluidic emulsification they are not often taken as a starting point for process design. It could be that if a combination does not work, it can also not be found in literature because people have moved on to try other options. We will report on membrane emulsification, both cross-flow (or direct) and premix emulsification, and specifically compare with the spontaneous droplet formation technique called EDGE (Edge-based Droplet Generation; [26
]). This technique was chosen, for its large viewing area in which wettability changes can be observed. Besides, the pressure dependency of droplet formation allows charting wettability issues, as will be explained in the next section, that starts with a general introduction of the EDGE technology.
It is good to mention that there are other microfluidic devices that use spontaneous droplet formation, and for which wettability effects are relevant. For microchannel emulsification that was proposed by Kawakatsu et al.
], and its’ up-scaled relative straight through microchannels [30
], the droplet formation mechanism was described by Sigiura and coworkers [33
], and various comments were made related to wettability effects [34
]. Besides, in the system proposed by Dangla et al
], in which a tapered droplet formation unit is used, wettability plays an important role.
3. Membrane Emulsification
Contrary to EDGE emulsification in which spontaneous droplet formation takes place, in membrane emulsification shear based droplet formation occurs. For overviews on shear based membrane emulsification, please consult the work of Joscelyne and Trägårdh ([52
] review food applications), Charcosset and coworkers ([53
] general review), Vladisavljevic and Williams ([54
] general overview with many products), Lambrich and Schubert ([55
] microstructured devices), Spyropoulos et al
] general reviews), van der Graaf and coworkers ([58
] double emulsions), Charcosset ([11
] specific for food), and Nazir et al
] premix emulsification). The patent literature was reviewed by Piacentini et al
We will focus on the classic shear based systems in which droplets are made by pushing a liquid through a membrane pore or tiny channel into a cross-flowing other liquid. Besides, premix emulsification is discussed. Both processes are schematically shown in Figure 5
. Alternative shearing methods have been described in literature such as rotation [60
], vibration and other dynamic effects, and these techniques are handsomely reviewed by Jaffrin [62
3.1. Cross-Flow or Direct Droplet Formation
As schematically indicated in Figure 5
a, during cross-flow emulsification the to-be-dispersed phase is pushed through a membrane. The liquid that emerges from the membrane surface is sheared of as droplets. Cross-flow emulsification has been investigated extensively; the most popular membranes for oil in water emulsions are made of Shirasu porous glass (SPG) that was discovered by Nakashima and Shimizu [63
], and ceramic materials such as alumina oxide (e.g., [15
]). Besides, work was done using α-alumina- and zirconia-coated membranes [64
], macro-porous silica glass membranes [65
], and micro-fabricated metal membranes [67
]. A special category are the microsieves, which are microfabricated devices with extremely uniform pores covered with a thin silicon nitride layer [69
], that have been investigated by e.g., Geerken et al
], Abrahamse et al
] and Zhu and Barrow [72
]. A comparison in relation to process design can be found in Gijsbertsen Abramhamse et al
For water in oil emulsification, PTFE (polytetrafluoroethylene) membranes, and hydrophobized SPG membranes have been used [17
]. To be complete, it should be mentioned that micro-fabricated metal membranes [77
] and hydrophobized silicon nitride microsieves [78
] have also been successfully applied, thereby indicating the relevance of wettability by the continuous phase.
Based on the wettability requirement described earlier, it can be expected that the contact angle plays a role in the droplet formation mechanism. However, early investigations on droplet formation with membranes (e.g., [15
]) have identified mechanisms that are based on force balances (critical capillary number, which is independent of contact angle). Further, droplet formation has been investigated, through experimentation with microfluidic devices (e.g., [79
]), and simulations (e.g., [82
], as recently reviewed by Muijlwijk and colleagues [84
]). Only in simulations, the contact angle can be varied systematically, and this was done in the simulations of Abrahamse and coworkers [20
], who found large effects on droplet release; thereby, clearly signifying its importance. Quantifying this effect in an experimental setting is far from trivial, and in the process guidelines section, we suggest how these effects can be assessed.
It was reported that SPG membranes can make droplets in the absence of cross-flow (e.g., [85
]), using a droplet formation mechanism similar to that of microchannels (introduced in 1997 by Kawakatsu and coworkers [29
]). These membrane are rather thick, and the porous structure is well connected [86
], which allows for local continuous phase intrusion, leading to uniform droplets (e.g., Sugiura et al
]). For this scenario, wettability by the continuous phase is also a requirement.
3.2. Premix Droplet Formation
In Figure 5
b, premix emulsification is schematically depicted, with the coarse premix being pushed through the membrane from the top, which leads to emulsion refinement upon passing the membrane. The technique was introduced by Suzuki et al
], and as mentioned previously, mostly a membrane is used that is wetted by the continuous phase, and for oil in water emulsions SPG membranes are mostly used as was recently reviewed by Nazir and coworkers [16
]. Also α-alumina, polycarbonate, polyamide and cellulose have been reported, together with glass filters for oil in water emulsions, and for water in oil emulsions, PTFE, PE, and hydrophobized SPG membranes were applied. Besides single emulsions, more complex structures were reported to be produced successfully, including double emulsions, polymeric beads, capsules, etc.
(reviewed in [16
]). As mentioned previously, if the membrane is wetted by the dispersed phase, phase inversion can take place, leading to very high disperse phase volume fractions [17
The mechanism of droplet formation during premix emulsification is hardly investigated. Van der Zwan and coworkers [87
] used a microfluidic device and mentioned that many different droplet mechanisms operate at the same time during pre-mix emulsification, ranging from snap-off due to localized shear forces, break-up due to interfacial tension effects (Rayleigh and Laplace instabilities), and break-up due to steric hindrance between droplets. Surprisingly enough, the droplet size distributions are narrow and in the same range as the pore size distributions for SPG membranes [1
3.3. Comparison with Literature
In the review of Nazir and coworkers [16
], a list is given of the various combination of emulsifiers and oils that have been reported to work for membrane emulsification. The list comprises both cross-flow and premix emulsification, and oil in water (O/W) and water in oil (W/O) emulsions, and some related products. When taking the findings with EDGE as a starting point, it is expected that when targeting oil in water emulsions with SPG membranes (or any other glass surface), the combination of alkane and SDS, Tween, or protein would work. When using a vegetable oil, it is expected that SDS would not work but the other emulsifiers would work. For water in oil emulsification it is expected that combinations with PGPR would work.
Of the 30 entries in the table that relate to emulsions, only two report that SDS and vegetable oil was used [85
]. In the work of Kukizaki and Goto [88
], a new SPG membrane is presented with a rather complex structure that deviates from the standard SPG membrane, and also contains alumina oxide. If this last material dominates emulsification, the result can be understood since the interaction of SDS with this ceramic material is much better than with glass. It should also be mentioned that the pore activation is very low, 2%–4%, and this could also be indicative for adverse wettability effects taking place; normally the porosity of SPG membranes is very high, as was reported by Vladisavljevic and coworkers [86
]. In the study of Yasuno and colleagues [85
], in which actually no cross-flow was applied and spontaneous emulsification was allowed to take place, very low pore activation (0.3%–0.5%) was found, which could very well be caused by wettability changes that prevent inflow of the continuous phase into the membrane.
When focusing on polymeric membranes, it is even clearer how membrane properties influence the emulsification process. Polycarbonate membranes could not be successfully employed to produce stable emulsions with Tween 20 or BSA [89
] regardless of the pressure applied, and most probably this is caused by the hydrophobic nature of the construction material. An alternative explanation is that the membrane is very thin, and that could induce less droplet break-up than thicker membranes.
For polyamide, polyethersulfone or nitrocellulose mixed ester membranes, with thicknesses between 100 and 200 μm, the ability to produce a stable emulsion depended largely on the type of emulsifier (see also Table 2
). When using 0.8 μm polyethersulfone (PES) and a nitrocellulose mixed ester (MCE) membranes with different surface porosity and thickness, the droplet size (d3,2
) was about 1 μm when using Tween 20. When using BSA, the cross-flow velocity had to be increased, while the final droplet size was three to five times the pore diameter depending on the BSA concentration and membrane used (Table 2
). The equilibrium interfacial tension cannot solely explain the observed differences, as can be seen in Table 2
where the interfacial tension values and the droplet size at the end of emulsification are presented. Through FTIR analysis performed on membranes [90
] used for emulsification with BSA, it was found that most of the surface fouling was oil, which indicates that sunflower oil replaces BSA at the surface, and a film is formed that hampers droplet formation [45
]. Besides, the ability of the protein to stabilize the oil-water interface will also depend on the protein maintaining its functionality, since if it is lost, oil droplets will coalesce and deposit on the hydrophilic membrane surface, hindering the emulsification process, which is in line with the findings reported for EDGE emulsification.
Alternatively, also complexes can be considered as stabilizing agent; soluble whey protein-polysaccharide complexes can stabilize the O/W and the W/O interface of single and double emulsions, respectively [92
]. These more complex interfacial structures did not affect the final droplet size of the single and double emulsions produced by premix membrane emulsification with SPG membranes, compared to the ones obtained using whey protein alone. Since favorable interaction is formed between both components, they are also less likely to interact with the membrane surface. When using these complexes, an increase in viscosity of the continuous phase will occur and that will increase the pressure needed for premix emulsification to take place, which is as expected.
In summary, it can be said that combinations of construction materials and emulsion components that work well in microfluidic devices are in general also successful for membrane emulsification. We expect that this similarity is due to favorable wetting conditions, that facilitate droplet release through prevention of the to-be-dispersed phase sticking to the wall. How this can be translated to process design is presented in the next section.
3.4. Process Guide Lines
The summary above sounds very simple, but how to achieve these conditions, or how to predict which conditions will occur beforehand? The core to the answer can be found in the contact angle, as is illustrated in Figure 6
. The contact angle is a resultant of three surface free energy terms, one between droplet and surface (red), one between continuous phase and surface (blue) and one between droplet and liquid (green). If the arrow is small, this indicates a low surface free energy, which is indicative of compatibility between the two phases. In Figure 6
, the surface free energy between continuous phase and surface is stepwise decreased while keeping the other two constant, and this results in a systematic increase in droplet contact angle, thereby going from a droplet to a continuous phase wetted surface. From this it is clear that wettability is very sensitive to changes in surface interactions.
In practice, not just one of the surface free energies will change, mostly more would change and sometimes even all. If the surface gets different properties this would influence both the interaction with the continuous phase and with the droplet phase. If the liquid/liquid interfacial tension decreases as a result of the presence of a surface active component, this mostly also implies that the component will adhere to the solid interface, and changes the surface free energies. What the resulting contact angle will be is hard to predict beforehand, especially when the liquid/liquid interfacial tension is changing very rapidly, as happens during high throughput emulsification processes.
Ideally the wettability for both liquid phase does not change in the presence of the surface active component that is needed to make and stabilize the emulsion, and this implies that the surface active component should be kept from adsorbing to the surface, or that the component forms a layer that does not change in time (so cannot be replaced by any of the phases). It has been demonstrated to be feasible to prevent components from adsorbing, through tailor made modifications, also for construction materials that are used for microfluidic devices (e.g., [23
]; however, this is very specialized work. It seems to be more realistic to tailor the surface differently, as, e.g., demonstrated by Sahin and Schroën [5
] who used a layer of pre-adsorbed protein that cannot be replaced by the oils of choice, and in that way create constant surface properties. Obviously, this layer should be strongly bound and not replaced by either of the liquid phases (e.g., sunflower oil that replaced SDS), or by the emulsifier (see also outlook section).
To check wettability changes, contact angle measurement may be used but for this a specific cell would need to be applied that allows measurement of the contact angle in a liquid system (not against air). The measurement should be done in the absence and presence of surface active components, and steady values should be found. If this is not the case, this is a clear indication of interaction of one of the liquids with the solid surface, or that changes induced by the surface active components take place. These effects become even more important when trying to scale-up microfluidic devices, or using membrane emulsification at large scale [23
As illustrated in Figure 6
, not just the solid surface is important, also the liquid/liquid interface, and more specifically that value of the interfacial tension given a certain droplet formation time. The surface active components will be transported and diffuse toward the newly formed interface, thereby decreasing the interfacial tension (e.g., [92
]). The actual value of the interfacial tension is highly time dependent, and will range between the value of a completely empty interface (fast droplet formation), and that at equilibrium with the bulk solution (slow droplet formation), and it is not trivial to access the values for the fast droplet formation times that are used in emulsification devices. Starting from a microfluidic Y-junctions, a method that allows estimation of interfacial tension at even sub-millisecond range, has been developed [96
]. In follow-up research in our labs, other surface active components have also been tested, and it is clear that these are elaborate measurements to carry out, although steady progress is made.
The approach described in the previous section has been shown to work for simple emulsions made by direct and premix emulsification. When more complex emulsion-based products are made, such as double emulsions or capsules, all surface active components can be tested as described above. The one with the strongest surface interaction while not changing the wettability of the continuous phase should be chosen to modify the surface. Component evaluation may become a rather impossible task when using mixtures, on the other hand, the natural competition between the surface active components will lead to preferential adsorption of one, and that would also be the one that needs to be tested on replacement by the two liquids. So in effect it is expected that the contact angle test could still be of use, but now it would not be possible to pinpoint the observed effects to a specific component.
Membrane emulsification was also demonstrated to work in the preparation of many solid and hollow particles [98
], for a good review please read [54
]. Also here the same rules apply as mentioned before, and in case of rapid solidification, also interactions of the solvent (that is needed to be able to work with polymers), with all available interfaces need to be considered. Away from solidification occurring when still in the membrane or microfluidic device, the wettability changes that are related to mass transfer of the solvent to the continuous phase can induce drastic wettability changes as we observed in unpublished work done with the systems described by Sawalha et al
In conclusion, wettability changes occur and can destroy emulsification processes when carried out with microstructured devices such as membranes and microfluidics. This paper will help guide researchers in finding the optimal combination of construction material and the components that can be used safely in the emulsion formulation. Obviously, this would also need to be combined with other process requirements to be successful, but we find that these considerations should be the core of any new process design.