[pen2l]

I wish the paper included more details on what sort of connectors they use to connect the tubes carrying fluids to the microfluidics chips. Fig.6A shows 9 abstract-looking cylinders, but it's not really clear what's going on there or how it works out. Perhaps the connectors themselves were printed, lending themselves to be easily connected with elastic-tubes -- but you'd have to be super careful when handling those.

Anyway, I do wonder about the feasibility of microfluidics in general. It's become quite popular in recent years (hey, Theranos' whole thing was actually microfluidics!!), but I find lack of blowout successes riding on microfluidics to be something of a concern. I don't know if whether that's because actually delivering with a microfluidics chip takes so much resources that any lab taking on a microfluidics project will then be robbed of resources to work on anything else, or if because orchestrating a dance of small quantities of fluids traversing elaborate pathways here and there at controlled rates, measuring quantities and setting up open feedback loop systems and setting up logic systems is just so hard and that's why I don't hear about happy reportings.

[tranzudao]

I’m actually one of the authors. The main thing we are trying to address in microfluidics is he feasibility, which is why we have been focusing on 3D printing as opposed to traditional clean room fabrication techniques because we can make a new design and see if it works in all of like 15 minutes whereas with clean room fabrication it would take closer to weeks. As far as the connections, we talk about those in some of our earlier papers referenced in the article, but it is actually quite simple. Instead of traditional connectors, we just include roughly 1mm diameter output holes where we can epoxy in tubing that has the same width.

[pen2l]

Congrats on the nature publication! And congrats on making the HN frontpage, I think this is one of the few times I've seen a paper (as opposed to a news article about a paper) on the frontpage.

What type of epoxy do you use? I imagine depositing the epoxy in micro-doses must take a steady hand.

Your approach is great for its simplicity. Can you comment on the biocompatibility of the cured resin? Good enough to culture cells in?

[tranzudao]

We were using a UV curing epoxy from Amazon, but we stopped because it had biocompatibility issue. Now we just use whatever is cheapest at the hardware store. It isn't hard at all to make the connections because the tubes go straight into the device and then we just epoxy on the outside and that is good enough.

My main contribution was testing the biocompatibility of the resins actually. The original resin was developed by the engineering team and while it worked well, some of the ingredients were extremely cytotoxic (to the point where if someone got it on their skin it would leave a nasty chemical burn). The resin used in the paper was a newer formulation that is completely biocompatible. The last bit of the paper is a quick dose-response assay with live cells.

[R0b0t1]

Do you list the 3d printing apparatus used?

[tranzudao]

The 3d printer was custom built by our lab. They have since commercialized the design and are sold by the company Acrea 3D.

[knolan]

These are the type of fitting used in many applications.

https://www.idex-hs.com/store/fluidics/fluidic-connections/f...

They are stupid expensive but usually work well. An alternative is to use a biopsy punch in PDMS and simply shove a 1.6mm tube into the resulting hole. The PDMS will hold the tube securely at most pressures.

[pen2l]

Neat. I see a lot of threaded components, is the expectation that at contact point we should have a threaded opening? How the heck does someone manage to do that at these micro-scales?

Also, wowza, it pains me to see non-metric units used there. One would have thought that fields that are so deeply academic in nature would be free of imperial units!

[knolan]

There are lots of different types of fittings. For glass chips where you etch the glass and bond another layer of glass over to seal you often have surface mount ports that you clamp on over a small hole and cure in an oven to bond. These give you a threaded port.

https://darwin-microfluidics.com/products/nanoport-kit-for-1...

It’s not so micro really. The internal flow geometry is around 100um but you’re mostly working with 1.6mm (1/16”) tubing that’s easily handled and used with all these fittings. The fittings are all finger tightening and a lot of the prefabricated chips are easily clamped into manifolds. You can also be creative and use heat shrink to connect tubes and the aforementioned PDMS punching works well, you just leave a large (2mm) region at the start and end of a micro channel to punch before you bond your PDMS to a glass microscope slide.

You’re never really exposing the flow geometry outside the clean room. It’s much like working with an IC on a breadboard, once it’s made you can wire it up easily and not really worry about anything mechanical other than the device working.

Yes, the imperial measurements for fittings is a pain.

My biggest complaint about microfluidics is that the design of the circuits is very ad-hoc. Since the flow rate is typically creeping flow the assumption is that the flow physics are extremely simple so we see these basic linear designs, dramatically different to microfluidic flows we see in biology. People just draw basic shapes on CAD that are easy to fabricate and iterate until the decide works.

I’ve submitted an ERC proposal to take a more computational approach to the design and layout of microfluidic devices. Instead of designing the exact layout of the device geometry the designer expresses their intent for its function, as a circuit diagram or node workflow and we then use various methods (such as numerical simulation and optimisation) to construct the geometry. I guess it’s not too dissimilar to automated PCB layout.

[pen2l]

What about a hybrid approach, as I think folks do right now more or less -- draw it out by eye, and then validate the design with simulation-software (e.g. https://www.comsol.com/microfluidics-module) and iterate as needed from thereon.

Your inclination toward a computational approach makes me think of what folks in mechanical engineering are up to these days. Creating mechanical designs generatively with topological optimization (https://all3dp.com/2/topology-optimization-simply-explained/). There appear to be clear use-cases and advantages: you reduce the amount of material needed, decide exactly what stress points you want to focus on and work to strengthen those particular parts of the structure in interesting ways, etc. Fluid mechanics is a little more complicated though I think, so you're certainly up for a challenge!

I wish you success with your proposal, and I hope to discover your findings published in a nature paper on the frontpage of HN sometime soon ;).

[knolan]

Thank you! Yes, generative design is part of it.

I’m interested in this approach because the current situation (including hybrid) has engineers, often students, designing devices to meet the need of biologists. There a going to be a fundamental challenge here in terms of communication of requirements, expertise and time the engineers have and limitations of available fabrication methods.

This results in the lowest common denominator in terms of design and performance. I want to make tools that the key stakeholders (biologists etc) can use to plot out the functionality they require and obtain a functional geometry. I’ve also got a concept for a rapid prototyping platform that eliminates many of the problems with 3D printing.

[pas]

Moderna and BioNTech/Pfizer both depend heavily on microfluidics, I'd call it an amazing success. All sorts of assays and genetic engineering techniques also depend on it.