How far can 3D printing take microfluidics?

Greg: So it doesn’t have a functional purpose in terms of an end application; this is simply a test design to let us evaluate the effects of turning some of the knobs we’ve talked about. For example, we’ve developed a new uniformity correction method, and this design, placing valves uniformly across the entire projected image area, is a good way to test how effective that correction is. The results are: yeah, it works great. There’s also a range of other issues this design lets us evaluate, in terms of how effective different changes are at getting a uniform result.

 

Robin: What other microfluidic features have you experimented with? I’d imagine droplet generators, regular channels?

 

Greg: You name it, we’ve probably tried it, and a bunch of things that are probably on nobody else’s radar. For example, we’ve made 3D-printed superhydrophobic surfaces out of hydrophilic materials; the native material is hydrophilic, but because of how we pattern it, it becomes superhydrophobic, with a water contact angle of 140 degrees or more.

 

We’ve also made this tiny cone, barely visible to the eye, with a little ball on top, and a channel running up through it. Because it’s a hydrophilic material, if you wet the surrounding area, it draws liquid up through the cone, spills it out over the ball, and coats it uniformly in a couple of microns of water, which then flows down the side; as you get evaporative loss, it’s self-replenishing. Why would anyone want a tiny micro-fountain like that? We have another group we collaborate with looking at whispering-gallery-mode lasers; you can take that thin water film, bring a tapered fiber up close, couple light into the water, and it gets trapped, circulating around and around through total internal reflection off the surface. They’re exploring some interesting laser ideas, materials questions, and so on, and because we have this fabrication capability, we can make these kinds of unusual shapes, and they work.

 

With the same group, we also published on something else: they have small chips, maybe a square centimeter, with silicon photonic devices on them, ring resonators, Mach-Zehnder interferometers, whatever, and they want to control the refractive index in the optical paths. We created 3D-printed microfluidic devices with integrated gaskets that compression-attach to these silicon photonic chips, and just by changing the salinity of the water, you can change the refractive index the light sees in the waveguides, and do a lot of tuning; it’s almost a passive process, taking almost no energy. The only other approach shown to be effective for that kind of tuning is heating, which draws a lot of power. So, at any rate, these are the off-the-wall kinds of things; we’ve done tons of traditional microfluidic devices, components, and integrations, but also plenty of things that are pretty far out there.

 

Adam: This is the part I care a lot about, since we use these devices to run bioassays, detecting molecules related to disease. We’ve created combinations of pumps, valves, and chambers that let us collect and concentrate certain types of molecules, based on simple chemical interactions or affinity-based interactions, like DNA base pairing. We’ve also designed 3D-printed devices that create chemical gradients, so the concentration of a given chemical varies with position within the device, which is useful for studying how cells respond to chemical cues.

 

We’ve used pumps, valves, and a diffusive mixing chamber to do serial dilution, a really common process in biochemistry labs, where you want to study a compound’s effect as a function of concentration. Normally you’d have some poor student pipetting a bunch of concentrations across several orders of magnitude, then testing each one in some downstream system. With 3D-printed pumps, valves, and diffusive mixing, which only really works at very small size scales, hence the need for truly microfluidic devices, we can bring in a concentrated solution and water, flow them together in a controlled way, and generate ten different concentrations spanning about a factor of a thousand, purely through a designed flow system, stabilizing within about a minute. So there are a lot of useful and interesting applications for these devices, and a lot of off-the-wall ones too, probably some we haven’t even thought of yet.

 

Hemdeep: In terms of the material you’ve used to build all these different devices, is it the exact same material across all these platforms?

 

Adam: We have one that’s sort of a go-to, the common denominator: polyethylene glycol diacrylate, PEGDA. We chose it for a few reasons; my lab had used it previously with more conventional cleanroom-based photopolymerization to make devices. It’s also both a monomer and a cross-linker, since it has two reactive groups, which makes it useful as the polymerized material. It also carries polyethylene glycol groups, which is important, because that chemical group, once formed into a material, resists nonspecific adsorption of things like proteins; for bioassays, you really don’t want stuff sticking to your device material. That base material has been our go-to, though we’re always exploring new additives and materials; we’ve probably innovated more in the UV absorber space, and done some work with different photoinitiators too. But we do have a standard go-to resin, and if a student’s coming in to try something new, that’s what we recommend first.

 

Greg: We should also mention these resin formulations are all available in our papers; anyone can pick one up, see the components, order them, mix them up, and have the same materials to work with.

 

Robin: I remember, in a previous conversation when we first talked about doing this podcast, you mentioned something very interesting: 3D printing in a vacuum. That concept alone sounds crazy; can you talk more about what you’re trying to do there, and whether you’re still pursuing it?

 

Adam: I can give some initial background, and then let Greg get into the details. Part of the rationale is that 3D printing uses a polymerization reaction, initiated by light from your optical engine, and there are a lot of things that inhibit polymerization; oxygen is one of them. So the initial thought experiment was, well, if we do it in a vacuum, there’s no oxygen. It turns out that if you do it in a vacuum, other things happen too.

 

Greg: We’re actually putting together the data for a big paper to reveal the whole story on this. But the bottom line is: when you 3D print at atmospheric pressure, in a normal lab, the resin you’re photopolymerizing has some dissolved gas in it, one component of which is oxygen, which actually serves a useful function as an inhibitor; we’ll get into why in the paper. But that dissolved gas also causes problems, because when you’re polymerizing against the bottom of your resin tray, typically a Teflon-like FEP film, your polymerized material adheres to that film, and when you go to do the next layer, you need to pull it off the film, come up, and come back down to whatever your layer thickness is, often around 10 microns, though we do use other thicknesses too.

 

As you’re pulling off, you create this suction effect, a space where resin needs to rush in very rapidly, dramatically changing the local pressure conditions in the resin. It appears that causes some of that dissolved gas to come out of solution as little bubbles. Those bubbles never do anything good; they tend to get attracted to, and trapped at, exactly the features you’re trying to create, which wipes out your result. We tried so many different things to avoid generating those bubbles, and none of them were entirely, or even mostly, successful. And depending on the time of year and atmospheric conditions, bubble generation actually varied; there was one time of year where we knew our yield was just going to drop, because we’d get more bubbles.

 

So, kind of as a measure of desperation, a few years ago, since we didn’t have anywhere else to go, I said, well, let’s just print in a vacuum, completely degas the resin so there’s no gas left to come out and cause the bubble problem. Over a couple of years, we got funding together and designed this monstrosity of a 3D printer, with a vacuum system attached; it really is a monstrosity, but a beautiful one. Now we can actually 3D print in vacuum. We’ve begun exploring that space, and we’re far enough into it now that we have a pretty good handle on what’s going on, and how to mitigate various issues that come with it. In fact, the 1,600-valve design I mentioned was printed in vacuum. That’s the overview; we’ll have all the details in the paper, hopefully out in the next couple of months.

 

Hemdeep: I’m curious what scale the bubbles are at, relative to the scale of the features you’re printing, since that’s clearly caused you a lot of angst to go to these lengths.

 

Adam: I think the problem is that the bubbles are near the scale of the features, or larger. As they form, they tend to stick to surfaces, and our 3D printing happens at a surface, so the bubbles stick right to the features. You end up with a bubble in place of a feature, and the bubble is much larger; if you’re trying to print 1,600 valves, and you have a bubble that’s about the size of ten of them, and you get a few of those connected together, instead of having all 1,600 working, you end up with a lot less.

 

Greg: The size of the bubbles varies widely; they can be as small as 20 or 30 microns, but when squished into a single 10-micron layer, they can spread out to hundreds of microns. We’ve developed, again not yet published, various types of bubble traps, since bubbles like to get caught on edges and corners; if you place those traps near the features you actually care about, they’re pretty effective at capturing bubbles. So we’ve built up a set of tools to deal with this, and vacuum printing is one of them. But it turns out if you remove the oxygen, and oxygen’s your inhibitor, you cause yourself all kinds of other problems: spontaneous polymerization, for one. We’ve worked through all of that now, and it works out really well, as evidenced by the chip I showed earlier.

 

But maybe we should also mention that vacuum printing has been one offshoot of our 3D printer development. We have another direction, too: multi-resolution printing, where we use two optical engines to print a single device. Those two engines sit on an XY stage, so we can bring either to bear as needed, in different regions of the print. The lower-resolution engine projects 15-micron pixels, so you generate the bulk of your device at that resolution, which is still very high compared to commercial offerings, and then we have a second engine that projects 0.75-micron pixels, sub-micron, and use that to create super-high-resolution structures only where you need them, fully embedded within the rest of the print.

 

So we have multi-resolution in X and Y, from the two different projected pixel sizes, and also in Z, since we’ve put two different optical absorbers in the resin, and shaped the emission spectrum of the two LEDs we use, a 365- and a 405-nanometer LED, tuned so that the 365-nanometer light sees ten times the absorption of the 405-nanometer light. With the 405-nanometer light and 15-micron pixels, we print 15-micron layers; with the 0.75-micron pixels at 365 nanometers, we print roughly 1-micron layers. That’s gotten us some stunning results, embedding these super-high-resolution regions where we need them. One example, from a student of Adam’s who’s really pushing this forward, is embedding very high-surface-area structures at high resolution inside channels, so you can think about 3D printing a chromatography column directly; that’s looking very promising.

Another thing we’ve built is a tiny mixer. In microfluidics, mixing is always a challenge, since you’re at low Reynolds numbers, laminar flow, so it’s mostly diffusion, unless you induce turbulence, which generally needs long channels and structures within the channel. We took a different approach: we built a tiny mixer, maybe 500 microns long, 300 microns wide, 300 or 400 microns tall, that can do complete, beautiful mixing of two flows, very fast, in a tiny volume. The key is that super-high-resolution 3D printing; if you just bring two streams together, you only get diffusion across the interface, so instead we split each stream into six narrower streams, interleave them, so you’ve got streams maybe eight to ten microns wide flowing side by side, and diffusion across eight or ten microns doesn’t take very long. We route that into a very narrow channel, itself maybe 10 microns wide, serpentine it a little over a few hundred microns, and by the time the fluid comes out the other end, it’s completely mixed. It’s beautiful, and it’s all done in a tiny volume.

 

Hemdeep: It sounds like the scale of that mixer would be no more than, what, a thousand microns long, from the initial introduction through the serpentine to the output?

 

Greg: Less than that, maybe 600 microns or so.

 

Adam: And with that, you can stack a whole bunch of them into a single 3D print; that’s one of the powerful things here. If you want to do this at scale, you could make dozens in a single print.

 

Greg: Yeah, this multi-resolution work, I think, is going to be an important innovation, and again, we’d love to see it in the hands of everyone.

 

Robin: What’s the footprint of the multi-resolution 3D printer?

 

Greg: That one’s maybe three feet by three feet, and four or five feet tall, something like that. This is actually our second iteration; we call it MR1.1, multi-resolution 1.1. The first one, MR1, from 2019, we never published anything on. We found that the height of the frame, which used an aluminum strut, meant temperature swings in our lab over the course of a day were enough to throw us out of focus, which made it a real struggle to get consistent results over any period of time; you could get one-off successes, but keeping the whole thing in focus and calibrated was very difficult. With MR1.1, we replaced the aluminum strut with a welded Invar structure, much more thermally stable, and added Keyence confocal distance sensors on both lens barrels of the optical engine, so we can measure our actual working distances in real time and compensate for any residual thermal drift. That seems to work quite well.

 

Hemdeep: So it sounds like moving to this new technology required a whole new round of iterations and issues you’d never have imagined. When you mentioned temperature control, I thought, yes, you’d have a really hard time staying in focus at three-quarters of a micron; you’d be out of focus immediately.

 

Adam: These are really problems you only run into once you start pushing the edges. If you’re making 200-micron features, it doesn’t matter. But if you’re trying to make three-quarter-micron pixels, and get the appropriate Z resolution for that, focus matters a lot, so temperature control, and appropriate design of the materials you build the printer from, suddenly matters a lot too.

 

Greg: And even on our mainstream HR 3.3 series printers, of which we’ve built two, we use Invar in strategic places, just to reduce thermal sensitivity in the prints.

 

Hemdeep: Now that you’ve got these two projects underway, is there anything coming down the pipeline you’d love to share with us, even just a snippet, while we wait for those papers to come out?

 

Adam: Great question. I have to be a bit careful, since there’s some IP associated with some of it, so I’ll be measured in what I share. We’re doing a lot with the ability to make a combination of different feature sizes, using the multi-resolution printer, plus the ability to create many functional valves. We’re looking at improving our ability to carry out chemical separations, simple things like: I have a complex mixture, say blood, and I’m looking for a specific molecule, or class of molecule, within it. In our case, we’ve studied molecules related to risk of preterm birth, and we’ve looked at viral RNA from a mosquito-borne infection.

 

With a complex sample like that, and wanting to isolate just one molecule or class of molecule, we need the ability to control fluid flow, pattern surfaces at high resolution, and attach specific molecules to those surfaces. All of that becomes enabled by the HR 3.3 printers, the MR1, and, I haven’t yet figured out exactly what we’ll do with the vacuum printer, but I’m sure once my chemistry students get access to it, we’ll find good uses; there are always things you didn’t know you needed a 3D printer for, until someone built one and you could actually make them. From my lab, some of the advances will be automating simple wet-lab processes, to improve detection of disease-related molecules, using a combination of these different 3D printers, along with other post-processing techniques we use, to carry out those analyses.

 

Greg: There’s just so much going on. One thing I’ve wanted since the very early days, 2015, 2016, is to be able to simulate the photopolymerization process, layer by layer, to predict what structures you’ll get as a function of how you tweak various operating parameters. For fun, honestly, to get away from administrative work I don’t enjoy, over the last year I taught myself the finite element method, found a paper I really liked from a group at Georgia Tech, and implemented a solution to a set of four coupled partial differential equations, using finite elements, to simulate what happens layer by layer during photopolymerization. That’s been really interesting, though it’s a bit hamstrung by not having as much time as I’d like to work on it. We’re getting relatively close to being able to open-source it as a tool.

 

That work connects to some fundamental studies we’re doing on the relationship between resolution and the irradiance you hit the resin with; we’re finding some interesting effects there, resolution enhancements, single-pixel channels. I’ve got a PhD student working on that, and hopefully she’ll have a paper out sometime this year. We’ve also got some other, frankly weird, phenomena we’ve observed over the years that some undergraduates are trying to figure out.

 

There’s also some really nice work out of Mehmet Toner’s group, and following up on that, Albert Folch’s group at the University of Washington, on fluidic logic, fluidic transistors. The point isn’t that fluidics is going to compete with electronics for computation; it’s that when you build a chip to run a particular assay, it’d be really nice to embed the operational sequencing directly into the chip itself, so that, in the end, a completely untrained person can take the chip, put in a sample, pipette in a couple of reagents, hook it up to a vacuum or pressure source, and the chip generates its own clock, and runs through the entire sequence needed to complete a biomedical assay on its own. That’s really the point of fluidic logic. To go in that direction, you need nonlinear fluidic structures to do the transistor-like operations that fluidic logic requires, and because we can print with such high resolution, and do some fairly crazy things, I’ve got a number of students working on different paths toward fluidic logic with 3D-printed structures. Maybe it’ll fail, but so far it’s looking pretty interesting.

 

Hemdeep: It sounds a lot like an ELISA chip, just at a much larger scale than what you’re probably trying to develop; it seems to fit that same logic.

 

Greg: Yeah, right. Out of David Juncker’s group there, Miguel [reference unclear in the source recording]. Yeah, sure, he’s using the burst-valve approach, which is a really nice way to go about things, and I think has tremendous potential. But this is another avenue toward something similar, where the entire operation of the chip is embedded in the chip structure itself.

 

Hemdeep: Well, we’ve come to the end of our conversation, and it was extremely enlightening, as always; I didn’t expect anything less. Thank you again, Greg and Adam, for taking the time to speak with us, sharing your experience to date, and giving us a glimpse of what’s coming around the corner.

 

Adam: Thanks so much for hosting us; it’s been a delight chatting with you.

 

Greg: Yes, thank you, it really has been. We really appreciate you, and everything you’re doing in the 3D printing and microfluidics space; we love that you’re making things available to people so they can pursue their own ideas and applications.

 

Hemdeep: Thank you very much. And that ends this episode; we’ll see you next time around. Bye for now.

Robin: And with that, we’ve reached the end of our series with Greg Nordin and Adam Woolley from Brigham Young University. Over the past few episodes, we’ve explored their groundbreaking work in 3D printing for microfluidics, from building custom 3D printers and materials to pioneering applications in bioanalysis and diagnostics.

 

Hemdeep: Their dedication to 3D printing innovation continues to push the limits of what’s possible in microscale fabrication. A huge thank you to Greg and Adam for sharing their journey and insights. If you’ve enjoyed this series, be sure to explore their ongoing work and open-source contributions, and stay tuned for more fascinating conversations in upcoming episodes.

 

Robin: You can join us next time on August 4th, when we’ll be joined by Alexandre Leblond, a PhD student at McGill University, for a conversation about mycelium-based biomaterials.

Robin: Thanks for tuning in to Big Ideas in Microscale. If you enjoyed the episode, follow us to stay up to date. You can listen on Apple Podcasts and Spotify, or watch the full video on YouTube. Follow us for updates and behind-the-scenes content on LinkedIn, Instagram, Bluesky, and X: we’re CADWorks 3D across the board. For show notes, paper references, and bonus resources, visit cadworks3d.com.

Hemdeep: Thank you for tuning in, and as always, stay curious, keep exploring, and never stop asking the big questions shaping our world.