A Practical Guide to Membrane Transfer
We’ve made a few academic inquiries so far into the theory which governs our nanoporous membrane transfer, but engineers as we are, the pace of our practical knowledge has outstripped that of our theoretical understanding. Here, I want to share a summary of what we know about these transferred membranes from a practical perspective, oriented towards application to Ottowa’s prefilter needs and other projects and to be supplemented with theory where appropriate.
Wetting:
We’ve identified that wetting underneath the transferred membrane plays a key role in its behavior. Recall from my previous post on this topic how we observed an air-water interface interacting with a transferred membrane on glass by rolling in a sheet underneath it, evidently by capillary action. We take this as evidence that the membrane rests very close to its substrate, but that it is not bound to it by any particularly strong force. For obvious enough reasons we’d rather the transferred membranes remain where they’re placed when wetted, so we set to work on keeping them down through the wetting process.
Two changes in particular have proven to be effective in this goal: replacing the glass substrate with silicon nitride (a more likely substrate for real applications including Ottowa’s anyway) and wetting the membrane rapidly rather than slowly. From the wet-paper-towel-like behavior of our initial experiments, we’ve progressed to something much more stable:
This membrane stays down very well, and microparticles in the water can be seen zipping past it as a visualization of the high flow which this membranes is evidently capable of withstanding.
We can better understand the reason for the speed helping through this next video:
This membrane is wetted once rather slowly, but does not come up except at one corner, and is seen to be capable of withstanding significant shear. When the water is then withdrawn and some time is allowed for drying before the reintroduction of the water slowly, we see how the interface peels up the previously-loosened corner, even though it had withstood much higher shear forces just moments before.
Thus, we come to the working hypothesis that what is really important in keeping a membrane down is an air-water interface which passes over the membrane on a timescale much shorter than that scale on which water infiltrates beneath the membrane. Imperfections in the membrane, like a peeling corner or a wrinkle, promote water infiltration and thus tend to disturb it more readily. This hypothesis may also serve to explain why the change from glass to nitride substrates seems to enhance the transferred membranes stability: nitride is less hydrophilic than glass, and thus is relatively resistant to capillary-like infiltration of water between the membrane and the substrate.
In summary, we expect that the keys to membrane stability during wetting are: a rapid air-to-water transition, a relatively hydrophobic substrate, and a smooth, relatively defect-free transfer. Which brings us to…
Transfer:
Our transfer procedure has been the subject of several iterations since Greg’s discovery that breathing on the membranes promotes their transfer. We’ve finally come to a transfer procedure which we believe is “optimal” for our purposes, though no doubt some small tweaks may improve it still.
We’ve gone through three types of attempted transfers: manual breaking away of the membrane, UV/Ozone or plasma bonding, and vapor transfer. We quickly found that the first was undesirable due to a tendancy to result in wrinkling or other membrane imperfections post-transfer, and more recently that the second simply doesn’t work, possibly due to a missing organic group (more on this later.) Vapor transfer on the other hand has proven very effective, if not difficult to grasp theory-wise.
As we understand it now, we believe that when the membrane and its substrate are pressed against one another and exposed to water vapor, wetting occurs through the same capillary action which drives delamination from the substrate when it’s wetted slowly post-transfer, although in this case the water is introduced through the nanopores rather than under the edge of the membrane. The water in this new capillary resists the motion of pulling the donor chip away from the substrate (which is intuitive — imagine trying to lift a piece of glass off a countertop once a thin layer of water has gotten between the two surfaces,) and this resistance is sufficient to break the membrane away from its frame.
The advantage of a vapor transfer is that the mechanical disturbance to the membrane is minimal, resulting in a very clean, wrinkle-free transfer compared to breaking it our manually. However, significant wrinkling still occurs when the vapor is applied evenly across the surface, perhaps due to condensation nucleating at several different points under the membrane and resulting in colliding capillary fronts. Much more smooth transfers have been observed when the vapor is applied at a shallow angle to the membrane (roughly less than or equal to 45 degrees up from the plane of the membrane.)
The most obvious problem with vapor transfer is that, as mentioned before, we’re unable to get an actual bond directly between the membrane and the substrate. This was the rationale behind our attempt at UV/Ozone or plasma bonding: we wanted to covalently secure the two surfaces together to keep them at a constant distance from one another and well-adhered.
Unfortunately, it seems that plasma bonding is incompatible with this application, where two inorganic silicon surfaces are to be adhered. We’ve developed a successful procedure for bonding silicon nitride to silicone via plasma by first growing a layer of silicon oxide on the nitride surface via rapid thermal treatment; however, this method fails to adhere the nitride to another oxide surface. This may be because, unlike when bonding an oxide to silicone, no organic group exists to be sacrificed in the Si-O-Si bond formation — two Si-OH surfaces will not react to form H2, but an Si-CH3 and an Si-OH surface will react to form CH4. This leaves us only with anodic bonding for chemical bonding of these two surfaces — however, it may prove unnecessary for our applications.
In summary, we find that our best transfer process is this: clamp the donor and substrate chips together firmly, and then expose the donor chip to fine water vapor at a low angle and separate them from one another as smoothly as possible. There is no covalent bond between the surfaces, but if properly wetted they should remain extremely close to one another nonetheless.
Drying:
A footnote to the wetting and transfer process which may prove useful in understanding the mechanics at work here: I ran a simple experiment in which I performed two vapor transfers of NPN onto nitride, placed one into the oven and one at room temperature, and checked for their ability to wet after about 4 hours by breathing on them and observing Newton’s colors (or the lack thereof.) The sample kept in the oven did not wet, while the one at room temperature did. Thus, we can conclude that the vapor which condenses under the membrane remains there for at least 4 hours at room temperature. This information is also useful if one for some reason wants to perform a transfer and then wet the membrane after a delay.
What remains to be seen now is whether these membranes will prove valuable for Ottowa’s application. We have shown that they remain in close proximity to their substrate when wetted properly and outlined methods and rationale for ideal wetting and transfer. We’re still working on demonstrating that transferred membraned can act properly as a filter above an oxide patterned chip to separate small and large particles, but if past performance of our nanomembranes is anything to go off of, this should present no trouble. If we can get enough to show together, there may be another post about that specifically in the near future.