3D printing a culture device: what the resolution really is, and whether the resin kills your cells

3D printing a culture device is genuinely faster than moulding one, and it fails in two specific ways that soft lithography does not: the resolution is worse than the marketing implies, and the resin can be cytotoxic. Both are manageable. Neither is optional to think about.

The honest summary is that printing is excellent for chambers, manifolds, holders, moulds and anything with features above a few hundred microns, and it is a research project in its own right below about 100 µm.

The resolution ceiling, stated properly

The number printer vendors publish is pixel size, and pixel size is not channel size. The Form 4, a representative current desktop machine, publishes a 50 µm pixel size, XY dimensional tolerance of ±0.15%, and constant layer thicknesses of 25, 50 or 100 µm. That does not mean you can print a 50 µm channel.

Three things intervene between pixel size and a working channel:

1. Light bleeds into the void. Resin cures where light lands and a little beyond. A channel that is nominally at the resolution limit closes.
2. Uncured resin must physically leave the channel. This is the binding constraint for closed channels and it gets harder as the channel gets longer and narrower. A partially cleared channel looks fine externally and is blocked.
3. Post-cure continues to cure whatever is still inside. Resin left in a channel during post-cure becomes permanent.

The literature is candid about where this leaves the field. The 2017 paper that first pushed into the genuinely microfluidic regime states that up to that point, achieved feature sizes had not been in the truly microfluidic regime (<100 µm) — and it achieved 18 µm × 20 µm flow channels using a custom DLP-SLA printer with a specifically designed resin. Later work from the same group reports 2 µm channels using multi-resolution printing.

Read those results correctly. They are proof that vat photopolymerisation can reach microfluidic scale, and they were obtained on purpose-built hardware with purpose-built chemistry. Neither the printer nor the resin is something you buy. On a commercial desktop machine with a commercial resin, plan for open features and channels comfortably above 100 µm, and treat anything smaller as an experiment to be validated before it is designed into a programme.

The resin problem

This is the failure that costs a batch of cells rather than a print.

Many photopolymerising resins contain cytotoxic components, and the applied literature on biomicrofluidic printing treats this as a primary obstacle rather than an edge case, alongside print resolution, mechanical stability and low optical compatibility for imaging. The cytotoxicity arises from unreacted monomer and from photoinitiator residues, both of which leach into aqueous medium.

The published mitigation is direct: deliberately leaching the toxic components out of printed materials after post-curing increases cytocompatibility. In practice this means an extended solvent soak or aqueous extraction after post-cure — a step that costs a day and is routinely skipped because the part already looks finished.

Two second-order consequences worth knowing:

• The problem transfers through moulds. Work on 3D printed moulds used to cast PDMS found leachates from the printed mould appearing in the cast PDMS, with propylene glycol identified as a species requiring toxicity investigation before such devices are used in vitro. Printing the mould rather than the part does not remove the chemistry question; it moves it one step back and makes it harder to see.
• Post-cure conditions are part of the material specification. Formlabs’ own data for BioMed Clear is stated for samples washed in a Form Wash for 20 minutes in 99% isopropyl alcohol and post-cured at 60 °C for 60 minutes. Change the wash or the cure and you no longer have the material that was tested.

What “biocompatible resin” certifies, and what it does not

Certified medical resins exist and they are a real improvement over generic resin. They are also certified for something other than what you are doing.

Formlabs publishes for BioMed Clear:

Three observations that decide whether this material suits your device:

The intended contact is skin and mucosa, not cell culture. BioMed Clear is described as suitable for applications requiring long-term skin or mucosal membrane contact. ISO 10993-5 cytotoxicity is a well-defined extract test — it is genuine evidence and it is not the same as “iPSC-derived neurons differentiate normally on it for 28 days”. No published head-to-head study comparing cell viability across the major biocompatible resins under matched long-term culture conditions was found. Run your own extract test with your own cell type before committing a programme.

The heat deflection temperature is the number to look at twice. At 54 °C under 1.8 MPa and 67 °C under 0.45 MPa, the material is below autoclave temperature by a wide margin — yet the same data sheet lists autoclaving at 121 °C and 134 °C as compatible. Both statements are true because HDT is measured under load and autoclaving is unloaded, but the practical implication is real: a printed part with thin walls, internal stress or any applied load can distort in an autoclave even when the resin is listed as autoclavable. Test a part, not a coupon. See sterilising a custom culture device.

Water absorption is not zero. At 0.54%, a printed part in an incubator for weeks is absorbing medium, and anything that goes in can come back out.

Optical quality: usually the deciding constraint

If your endpoint is fluorescence, printing is often ruled out before any of the above matters.

Any 3D print material exhibits enormous autofluorescence when UV light is used to cure the resin, alongside poor optical quality and residual uncured material in closed channels. The applied tutorial literature lists low optical compatibility for imaging among the principal obstacles to bioanalytical labs adopting resin printing.

The workable pattern is hybrid: print the body, bond a proper optical window. A printed manifold with a glass or COP coverslip bonded over the imaging region gives you the geometry freedom of printing and the optics of the coverslip. This is the configuration most successful printed organ-chip devices actually use, and it should be the default assumption rather than a fallback.

When printing is the right answer

Printing wins decisively in four situations:

1. Iteration speed on features above ~200 µm. This is the core case. The published motivation for the shift is precisely that reliance on microfabrication and soft lithography limits prototyping turnover and scalability. A design change that costs a new photomask and a cleanroom slot costs a print and an afternoon.
2. Geometry that cannot be moulded. True 3D internal structures, crossing channels at different heights, and undercuts are natural for printing and require multi-layer assembly in moulding. This is the one capability printing has that lithography does not.
3. Holders, adapters, manifolds and stage inserts. The unglamorous majority of useful printed parts. No cells touch them, the tolerances are loose, and the alternative is a machine shop.
4. Moulds for PDMS casting. Fast and cheap — with the leachate caveat above. Passivate or thoroughly extract the mould, and check the cast part rather than assuming the chemistry stopped at the mould surface.

When it is the wrong answer

• Sub-100 µm channels on commercial hardware. Achievable in the literature, on custom equipment. Not a purchasing decision.
• Fluorescence imaging through the printed material. Bond a window instead.
• Anything where small-molecule absorption matters. The absorption behaviour of a given printed resin is generally uncharacterised, which is worse than PDMS’s well-characterised bad behaviour — at least with PDMS you know the size of the problem. See the material guide.
• Production volumes. Printing does not amortise. Each part costs roughly the same as the first. Above a few hundred parts, tooling wins, and that crossover is the whole argument of the custom fabrication page.
• Regulated or long-term-contact applications without your own biological validation on the actual part, printed with the actual settings, post-processed the actual way.

A workable protocol

If you are going to print culture-contacting parts, the published guidance converges on roughly this sequence:

1. Choose a resin with published ISO 10993-5 data, and read what else it was tested for.
2. Print with generous channel dimensions. Start at 3–5× the pixel size and reduce empirically.
3. Clear channels aggressively before post-cure. Uncured resin left inside becomes permanent.
4. Wash to the manufacturer’s stated protocol, because the published material data is only valid for that protocol.
5. Post-cure to the stated conditions.
6. Then extract deliberately — an extended soak beyond the wash step, on the published finding that leaching toxic components after post-curing increases cytocompatibility.
7. Sterilise by a method the resin data sheet lists, and check the part for distortion rather than trusting the coupon-level HDT.
8. Run a cytotoxicity control with your own cells on the finished part before the real experiment. This is one plate and it is the cheapest insurance in the whole workflow.
9. Bond an optical window if anything will be imaged.

How we can help

We do not print devices. Where we are useful is the decision that precedes printing: whether the part you have drawn is a print, a mould, a machined part or an injection-moulded one, and what each route costs at your volume. If you have a printed prototype that works and you need it in a material and a process that will survive a real study, that is the transition we can put to several fabricators at once — with the resolution, optical and biocompatibility requirements written into the specification rather than discovered afterwards. Send us the design.