The most commonly used alternative to cross-flow filtration is dead-end filtration. This section gives an overview of the general principle as well as advantages and disadvantages of this technique. In dead-end filtration flow from the feed is perpendicular to the membrane. The permeate flux through the membrane, again, comes from a pressure differential over the membrane. As there is no tangential flow over the membrane, particles will stack up on the membrane and form the so-called cake layer. This leads to plugging of the membrane which decreases the permeate flux over time. To overcome this issue the cake layer has to be flushed away regularly. One advantage over cross-flow filtration, is that a retentate does not exist. A disadvantage is that filtering can only take place as a batch process as cleaning with water is a reoccurring issue 11. For the topic of blood plasma separation dead-end filtration is relatively uncommon, with a huge number of publications rather concentrating on cross-flow filtration. When it is used though, some advantages over cross-flow applications are found. For example can the high flow rates of cross-flow devices lead to hemolysis, which is not the case in dead-end filtration 25. Also the application is very simple with little to no extra equipment needed in most cases 25.
After reviewing the general principle of cross-flow filtration, this section will layout the adaption of the technique for microfluidic blood separation purposes. The first major issue is enabling a sufficient tangential flow over the membrane while at the same time keeping the pressure of the blood below 0.5 bar 5. The reason for this restriction is, that research found that higher pressure levels will lead to hemolysis and therefore impede plasma quality 5. At this point in time there are several approaches to this issue. Some publications use syringe pumps 5, 49 or peristaltic pumps 6, while others are driven passively by capillary forces 44. These techniques come with big drawbacks though. The use of pumps requires big amounts of sample, as well as complicating the handling. There is also a problem with how precise these systems can be used. Passive systems on the other hand are too slow in most applications and are not flexible in their speed. A promising solution for this are the already explained microfluidic valves, which come in different forms like steel balls in the paper by Cheng et al. 7. The use of valves for driving the fluid allows precise and fast handling speed.
The way the cross-flow filtering is realised also differs between different publications. The simplest option is to run a channel over a membrane. In this case different geometries can be found to maximize the filtering surface. For example a spiral will maximize the surface area 5, 49, another common option is the use of a meandering channel over the membrane 7. The drawback though is that blood can react to sharp changes in channels by coagulating. This can only be avoided by using a straight channel like we are for this project. When using straight channels a common technique is to use several capillaries parallel on a membrane 19, 44. An issue with parallel capillaries can be uneven loading, with air bubbles that block channels.
To unclog the cross-flow filter there are a number of strategies. The most obvious ones are to dilute the blood 5, which leads to a depletion of biomarkers or to increase flowing speed, which can cause damage to the blood by high strain rates 49. More sophisticated approaches are the use Dielectrophoresis 44 to actively unblock the membrane or the use of back flushing with microfluidic valves 7. The drawback of Dielectrophoresis is the high degree of complexity. The easiest and presumably most effective way to unblock the membrane is therefore to use back flushing by reprogramming the order in which the valves open and close.