Photo-Induced Polymerization Reaction Refining 3D Printing to Sub-Millimeter Resolution

• 1
  How the photo-induced polymerization reaction refining 3D printing builds a solid voxel
• 2
  Two-color inhibition: the trick behind sub-millimeter control in photo-induced polymerization reaction refining 3D printing
• 3
  Where resolution is lost: dose, diffusion and shrinkage
• 4
  When to print it and when to machine it
• 5
  What to specify on a drawing that mixes both processes
• 6
  Photo-induced polymerization reaction refining 3D printing vs CNC machining

A photo-induced polymerization reaction turns liquid resin into a solid network when light hits it. A photoinitiator absorbs photons at a specific wavelength, breaks into radicals, and those radicals open the acrylate or epoxy groups on the monomer. The opened groups link into chains, the chains cross-link, and the liquid turns into a gel within milliseconds. Stopping the light does not stop the chemistry at once. Radicals keep reacting until they meet oxygen, another radical, or a chain that cannot move.

Resolution is set by the size of the region where enough photons arrive to push the conversion past the gel point. Photon flux falls off with depth through the Beer-Lambert law, so the cure front is never a flat plane. It is a gradient. The top of the layer receives more dose than the bottom, and the edges of a projected pixel receive less than the center. That gradient is the real limit on feature size, not the pixel pitch of the light engine.

Two numbers describe the resin at a given wavelength. The penetration depth Dp is the depth where dose drops to 1/e of the surface value, typically 0.05–0.3 mm for acrylate resins in the near-UV. The critical dose Ec is the energy per area needed to reach the gel point. A layer only forms where the local dose exceeds Ec. Everything else stays liquid and gets washed away later.

This is why a printer with a 50 μm pixel can still produce a 30 μm channel but cannot guarantee a 5 μm wall. The wall thickness depends on how sharply the dose falls off across the boundary, and that falloff has a physical floor set by diffusion of radicals and by the wavelength of the light itself.

• 1
  DpDepth where dose falls to 1/e; controls maximum cure depth per exposure.
• 2
  EcCritical dose to reach the gel point; below this the resin washes out.
• 3
  ConversionFraction of reactive groups consumed; drives stiffness and shrinkage.

One light source gives you one knob: turn it on and the resin cures, turn it off and it keeps curing for a short while. That lag is the main source of bleed. The fix, studied at groups such as the Autonomous University of Barcelona, is a second wavelength that deactivates the photoinitiator or quenches the growing chain. One color writes, the other erases. Where both overlap, nothing solid forms.

In practice the writing beam is near-UV at 365–405 nm and the inhibiting beam is a longer wavelength in the visible or near-IR. The inhibitor keeps the radical population below the threshold for gelation in a thin shell around the written voxel. That shell can be as thin as a few micrometers. The result is a sharper wall than a single-color printer can hold, because the boundary is defined by the crossing point of two dose profiles rather than by the tail of one.

The trade-off is hardware and speed. A second beam path, a dichroic mirror and a scanning system that keeps both spots aligned within a few micrometers add cost and calibration work. Writing speed drops because the inhibitor must be replenished by diffusion between exposures. For a small medical component that needs a 40 μm internal channel, that is a good trade. For a 100 mm cosmetic shell, it is not.

Oxygen plays a similar role and is free. Dissolved oxygen scavenges radicals near the resin surface, creating a dead zone of 10–100 μm depending on resin viscosity and oxygen permeability. You can use that zone as a release layer against the window, or you can fight it by purging with nitrogen. Either way, you must know which one your printer is doing, because the same file will print differently in the two modes.

• 1
  Writing beam365–405 nm, sets where the gel point is crossed.
• 2
  Inhibiting beamLonger wavelength, quenches radicals at the voxel boundary.
• 3
  OxygenNatural inhibitor; useful at the window, harmful inside the part.

Three effects set the practical floor on feature size. The first is the dose gradient. A projected pixel has a Gaussian or top-hat profile, and after it passes through the resin the edges are softer than the center. If you want a 20 μm wall, the dose must drop from above Ec to below Ec within that 20 μm. With Dp around 0.1 mm that is a steep ask, and it is why sub-millimeter printing usually means thin layers and low light intensity, not more power.

The second is radical diffusion. A radical that escapes the written voxel before it finds a monomer can start a chain in a neighboring voxel. The distance is small, on the order of a few micrometers in a high-viscosity resin, but it grows with temperature and with low conversion. Warm the resin to lower viscosity and you print faster, but you also blur the edge. This is the main reason production resin printers run in a temperature-controlled enclosure.

The third is shrinkage. Acrylate resins shrink 5–8% by volume on cure, epoxies 2–4%. Shrinkage happens after the gel point, so it pulls on the already-solid network and warps thin features. A 50 μm rib printed on a 2 mm base will bow toward the light source. You can compensate in the CAD file, but the compensation is resin-specific and often directional. Print a test coupon in the actual resin before you commit a production tool.

None of these effects care about the marketing number on the printer. A 25 μm pixel pitch tells you the addressable grid, not the smallest useful wall. The useful wall is two to four times the pixel pitch for a single-color printer, and closer to one to two times for a two-color system with good calibration.

• 1
  Dose gradientEdge softness scales with Dp; thinner layers help.
• 2
  Radical diffusionWarm resin prints faster but blurs boundaries.
• 3
  Shrinkage5–8% for acrylates, 2–4% for epoxies; warps thin ribs.

Print the part when the value is in the geometry, not the material. A microfluidic manifold with 200 μm channels, a lattice that would need five setups on a mill, a housing with an internal path no tool can reach. Those are cases where the photo-induced polymerization reaction refining 3D printing earns its place, because no subtractive process can reach the feature at all.

Machine the part when the value is in the material or the tolerance. A bracket that carries a load, a shaft that runs in a bearing, a cover that must hold ±0.005 mm across a 200 mm bolt pattern. Photopolymers creep under sustained load and lose stiffness above their glass transition temperature, which for common resins is 60–90 °C. A printed part that feels rigid on the bench can sag in a warm enclosure.

The usual answer is both. Print the prototype to check fit and flow, then machine the production parts from aluminium or stainless. We see this constantly on automotive and medical programs: a resin manifold validates the channel layout in a day, and the 5-axis version in 6061 or 316L follows once the geometry is signed off. The printed part is a geometry check, not a material substitute.

If the part has to be both fine-featured and structural, look at metal additive or at a machined assembly. Both cost more than either process alone. Deciding early avoids a redesign after the tooling is cut.

• 1
  Print for geometryInternal channels, lattices, one-off shapes.
• 2
  Machine for materialLoad, temperature, tolerance, surface finish.
• 3
  Print then machineValidate flow and fit, then cut production parts.

A drawing that says '3D print' is not enough. State the process, the resin family, the build orientation and the critical faces. If a 0.15 mm slot must stay open, dimension it and add a note that it will be checked optically. If a face must seat against a machined part, call out the flatness and the finish on that face only. Everything else can stay loose.

For the machined parts, give us the tolerance, the material grade and the finish. We work in 6061-T6, 316L, 17-4PH, Ti-6Al-4V and PEEK, with ±0.005 mm tolerance and finishes from Ra 1.6–3.2 μm as machined down to Ra 0.2–0.8 μm where it matters. A 4,000 mm maximum processing size covers most housings, and 16 simultaneous 5-axis centers handle the contoured faces that a 3-axis setup cannot reach in one pass.

Send the file and a short note on function. We return a quotation and a free DFM analysis within 12 hours, and production can start within 24 hours once the drawing is agreed. Parts ship in 3–5 days. No minimum order quantity, so a single prototype and a 10,000-part run go through the same first-article process.

Keep the resin printer and the mill in the same discussion. The choice is not which process is better. It is which process holds the feature that the assembly actually needs.

• 1
  State the processResin family, orientation, critical faces.
• 2
  Dimension the fine featuresSlots, channels and walls that must stay open.
• 3
  Call out finish per faceOnly the faces that seal, slide or show.