There are a handful of established ways to make a microfluidic chip. Each is good at something and each breaks down somewhere, and the pattern in where they break is the reason this company exists. This page sets out the landscape at the level of what each method can and cannot do.
The matrix compares every route on the same variables. The detail for each follows underneath. Costs and lead times are sourced on the process deep dive and the cleanroom page.
| Route | Capital overhead | Iteration | Production scale | Geometry limits | Organoid housings | Small molecule uptake |
|---|---|---|---|---|---|---|
| PDMS soft lithographyExcels at iteration | Cleanroom, aligner, trained operator | Hours to days in house | Cannot scale to production | Surface relief only, bonded lid | Yes, layered, but the material absorbs media components | High, plus protein adsorption and oligomer leaching |
| Hot embossingExcels at scale | Lab press from a few thousand; commercial systems quote only, roughly $150k to $350k, plus a $5k to $50k insert per design | Weeks per new insert | Yes, from roughly a thousand units | Thin sheet, surface relief only | No, thin sheet only | Low in COC |
| Injection moldingExcels at scale | Steel tool per design | Weeks per new tool | Yes, seconds per part | Surface relief only, draft angles required | No, undercuts cannot demold | Low in COC |
| Glass and siliconSpecialist, neither | Semiconductor facility access | Weeks | Yes, but cost per chip stays high at every volume | Planar, etched | No, planar etch | None |
| Resin printing, Aionik™ cureIteration and scale | Desktop printer and our unit | Print 20 min to 2 hrs by size and printer; cure 1 to 3 min | Same process at any volume | Full 3D, enclosed volumes included | Yes, printed in one piece | Low across printed resins generally; ours not yet measured |
Row tints show what each route is good at: iteration, scale, or both. Within a row, green marks an advantage, oxblood marks a limitation, and grey means neutral or unknown. COC is cyclic olefin copolymer, a glassy thermoplastic used where PDMS absorption is unacceptable. Our own uptake figure is grey because we have not measured it, not because it is good.
Cast silicone against a photolithographic master mold. The academic standard for two decades, and the route most published microfluidics is built on.
A thermoplastic sheet heated just above its glass transition and pressed against a mold insert, then cooled and demolded. The middle route between soft lithography and injection molding.
Melted polymer injected into a hardened steel tool. The route a device takes once the design is frozen and the volume justifies the tooling.
Etched substrates using semiconductor fabrication. The route for chemistry, pressure, or optical requirements that polymers cannot meet.
Print the part on standard SLA hardware, then run our patent pending post process in place of the chemical wash. No master, no mold, no cleanroom for the part itself.
Read the table above column by column rather than row by row and the same shape appears each time. The methods that let a researcher change a design quickly cannot make the part at volume. The methods that make it at volume require the design to stop changing first.
| Route | Can it iterate quickly? | Can it produce at volume? |
|---|---|---|
| PDMS soft lithography | Yes, once the master exists | No |
| Injection molding | No, a revision is a new tool | Yes |
| Glass and silicon | No | At high cost per chip |
| Resin printing with our cure | Yes | Yes, same process throughout |
In practice a lab does not choose one of these. It uses several in sequence: prototype in PDMS because iteration is possible there, then move to hot embossing or injection molding when volume is needed. The routes depend on each other, and that dependency is the expensive part. The handoff costs a redesign for a new process, fresh tooling, and revalidation of a device that is no longer quite the device that was tested. Our process is meant to remove the handoff rather than any one of these routes, by holding the same material and the same process on both sides of it.
Most labs own none of these processes, so they send the design to a contract manufacturer and buy finished chips. That is a sourcing decision rather than a fabrication method, which is why it is not in the matrix: the foundry runs one of the routes above on your behalf, and you inherit its limits along with its capabilities.
What you also inherit is the queue. Buying custom devices takes one to twelve weeks depending on the vendor and the complexity of the design, and every revision starts that clock again. Setup and tooling are charged per design, so cost moves with how unusual your geometry is and supply moves with how busy the vendor is. Researchers routinely wait months to hold a design they drew in an afternoon. The wait, not the invoice, is what actually limits how many questions a lab can ask in a year.
That split is the reason a device gets validated in one material and shipped in another. It is a structural feature of how these processes work, not a failure of anyone using them, and it is the specific gap our process is built to close.
Five points worth stating precisely. Two mark where the process does not reach today, two are predictions we have not tested, and one is a gap in our own evidence. What we have demonstrated is on the homepage.
Today, where a design needs geometry finer than a printer resolves, photolithography and silicon remain the right tools. The hypothesis below is how that boundary could move, not a claim that it already has.
Ours is a post process, so in principle it applies to any resin printed part, including output from the highest resolution microfluidic printers on the market. If that holds, the finest printed geometry becomes curable the same way. We are testing it now.
Our results are our own. Independent third party test data with published methodology is not yet part of the record, and we do not present it as though it were.
Integrated devices such as electrode bearing organoid chips look like a geometry and materials problem rather than new chemistry, which is why they are on our list to test. We have built none and we claim nothing here.
Material class gives us good reason to expect a glassy printed resin takes up far less than an elastomer such as PDMS, and the published comparisons point the same way. That comparison is underway. We do not present our own results as validation.
Every established route sits on one side of it. Holding both sides in one material, with one process, is what we built.
See what we built