Milk protein fractionation by microfiltration (MF) membranes is a still growing field in dairy technology and a lead technology for the valorisation of complex food materials such as milk by making single fractions available with their unique individual functional properties. It is generally known that, in membrane filtration, a considerable accumulation of retained material on the membrane surface occurs, in particular during milk protein fractionation by MF.
During MF of food systems, material accumulations at the membrane surface mainly consist of biopolymers such as proteins or polysaccharides [1
]. This applies even under crossflow conditions, where the wall shear stress only reduces the amount of deposited material, but a complete prevention of material accumulation cannot be achieved [5
]. This is referred to deposit formation [8
]. In applications in dairy technology, the deposit primarily consists of casein micelles, the main protein component in milk [9
]. Previous studies showed that the total filtration resistance considerably increases as a function of time and processing conditions owing to the deposit layer formation and cross-linking of proteins in the deposit, which leads to a steep decline in flux with filtration time in the very first minutes after filtration started [13
]. In protein fractionation by MF, deposit formation by the fraction of the larger proteins (casein micelles) has a decisive negative effect since it hinders the desired permeation of the smaller whey proteins [18
]. Therefore, the deposit layer formation correlates with a loss in separation efficiency and throughput [4
]. This is not only of relevance for the overall fractionation result of the whole module. Since in cross-flow membrane filtration the transmembrane pressure decreases along the flow path from module inlet to outlet, the extent of deposit formation varies accordingly, while the shear stress almost stays the same as at the inlet. As found by Hartinger et al. [5
], Piry et al. [19
], and Kulozik & Kersten [4
] using custom-made module systems able to measure flux and permeate composition in segregated modules separately, the targeted maximum of whey protein permeation is achieved only at or towards the end of the module. At this point, the ratio of transmembrane pressure and shear stress shifts to the benefit of shear forces imposed on already deposited or depositing material. Deposit formation is thus reduced where the transmembrane pressure is gradually reaching lower levels, while shear forces transporting deposited material away from the membrane surface stay the same (if one neglects the minimal amount of permeate reducing the volume flow from module inlet to outlet). Kühnl et al. [16
] and Steinhauer et al. [20
] have further extended these insights by indirectly studying the effect of colloidal interactions between particles in the deposited layer at the membrane. There is room for optimizing the efficiency of, e.g., milk protein fractionation by creating more or less open porous deposits as a result of varying attractive or repulsive forces between particles as a function of pH and ionic strength. It would be of interest to be able to assess these structural details in-situ
, but a capable method does not yet exist.
To investigate the accumulated material with regard to their chemical or structural properties, indirect or destructive techniques are widely used. Because the deposit layer changes its appearance, composition and density upon pressure release, the deposit layer should ideally be investigated in-situ and non-destructively. Furthermore, changes in height and inner structure are expected because the deposited layer expands due to relaxation under ambient pressure conditions. In addition, ex-situ measurements require the removal and destruction of the membrane in order to get access to the deposited material.
In comparison to previously investigated clear model fluids multiple complications arise: (1) milk is an in-transparent fluid and (2) membrane systems, such as tubular ceramic or polymeric hollow fibres that are operated in inside-out filtration mode, are difficult to be analysed by direct observation methods, e.g., by optical means.
In previous studies [21
], nuclear magnetic resonance imaging (MRI) was used as a non-invasive and non-destructive technique to assess the growth of a deposit layer during filtration of diverse materials as a function of filtration time. These works mostly studied model systems consisting of inorganic particles or defined hydrocolloids such as alginate in aqueous solution. Skim milk as an in-transparent feed medium and especially the deposit layer formation in ceramic hollow fibre membranes was also studied [27
] to prepare the methodology of the MRI measurements. MRI was applied on skim milk as a complex medium and to establish the required sample preparation to achieve the necessary contrast difference for MRI analysis and to develop the mathematical data processing protocol. When using MRI, spatially and time-resolved information about density and height of the deposit layer can therefore be obtained for skim milk filtration in inaccessible, optically in-transparent module housings to directly and in-situ
assess deposit formation.
Beyond this, membrane filtration performance depends on the processing conditions such as pressure, temperature and filtration time. However, it is unclear whether the time-dependent filtration performance depends on the effect of the change in deposit height because of protein accumulation or change in porosity due to deposit compaction. When trying to better understand these effects and changes of the deposit layer it is obvious that this gap in knowledge needs to be filled to explain the impact of deposit layer formation on flux reduction. Furthermore, an answer to questions regarding the impairment of permeation of solutes through the combined layers of membrane and deposit material is of interest because the deposit layer acts as an additional filtration resistance in the form of a secondary membrane.
Regarding processing conditions it is known that a loss in static pressure not only occurs along the flow path (ΔpL
), i.e., along a membrane, but also inside the deposited material when the permeate flows towards the membrane surface because of frictional pressure losses, thus reducing the transmembrane pressure (ΔpTM
]. For a layer of compressible and deformable casein micelles, the pressure loss results in an increase in compaction and, thus, in higher deposits´ densities or reduced porosities [20
]. A reduction of flux with filtration time during MF of whey proteins has been attributed to an increase of the deposit layer height alone [30
]. Apart from that, some studies were performed to qualitatively and quantitatively elucidate the structure of the deposit [31
]. However, a correlation between the chemical analysis of the deposit layer and in-situ
measured layer has not been shown so far. The methods developed in [26
] are applied to get more insight into the details of the deposit´s composition regarding the ratio of casein and whey protein. Additionally, there is still a lack of understanding of how pressure affects the behaviour of the deposit’s chemical composition and height.
Apart from ΔpL
, temperature ϑ
is a variable in membrane filtration that results in increasing flux and, therefore, a higher convective transport of protein towards the membrane surface at higher temperatures. Its influence on deposit formation and filtration efficiency is still the subject of current research [37
]. In industrial applications, the MF process of skim milk is operated either at rather low ϑ
(10–15 °C) [38
] or rather high temperatures of about 45–55 °C [39
]. Higher temperatures yield higher fluxes mainly due to the lower product viscosity, but may promote denaturation and also microbial growth to some extent, especially in long-term operations between cleaning cycles [1
]. Meanwhile, lower temperatures keep microbial growth better under control, but yield low fluxes and induce a migration of β-casein into the permeate, thus reducing the purity of the whey protein fraction [40
]. Despite the fact that cold filtration is by now more frequently applied in industrial application, little is known about the effect of low temperature in milk MF on the deposition of casein micelle layers.
Many models exist that try to determine the filtration phenomena by evaluating the macroscopic data such as the permeate flux or pressure drop [42
]. The deposit formation is affected by more than only one of these effects: pore blocking, pore constriction, cake formation, solute adsorption and concentration polarization [43
]. In case of the in-situ
measurement by MRI the deposit formation during skim milk filtration is observed on a microscale where it was already shown that parts of the deposit layer behave differently as the pressure is released as a part of it diffuses back into the membrane lumen [25
In summary, the purpose of this study is to apply MRI in-situ to assess milk protein deposits on MF membranes applied for milk protein fractionation under processing conditions relevant for practical situations. MRI is currently in ongoing development to be applied in various filtration processes. MF was at first operated in dead-end mode to measure the impact of ΔpTM, filtration time, and filtration temperature. To correlate MRI results with chemical composition and the permeation of proteins, reversed-phase high-performance liquid chromatography (RP-HPLC) analysis of the deposit layer was performed. These data allow a compositional characterisation of the deposit layer and could thus lead to a deeper understanding of mechanisms responsible for the reduction of flux and protein permeation caused by deposited materials on membrane surfaces.