The Development of 3D Printed Membranes: How Printing Changes Filtration

• 1
  Key takeaways
• 2
  Why the development of 3D printed membranes drew $6.9 million
• 3
  How printing builds a pore network
• 4
  What printed geometry actually changes
• 5
  Materials and where printing stops
• 6
  The machined parts around a printed membrane
• 7
  Cast membrane vs 3D printed membrane vs machined housing

Evove is a UK separation technology company. It raised $6.9 million to speed up the development of 3D printed membranes and the modules built around them. The bet is simple: membrane chemistry has improved for decades, but the physical shape of a membrane has barely changed.

Most filter membranes still come from a process developed in the 1960s. A polymer solution is cast into a thin film, then submerged in a bath. The polymer separates into a porous skin and a spongy support layer. Pore size is set by bath chemistry, temperature and draw speed. It is a fast, cheap way to make flat sheet.

The limit is control. Phase inversion gives you a random pore network. You cannot place a pore at a coordinate, you cannot taper a channel along its length, and you cannot thicken one region to survive a pressure spike. For water treatment that was acceptable. For high-value separations, it leaves performance on the table.

Printing offers a different starting point. The pore network is drawn in CAD and deposited layer by layer. That does not automatically make a better membrane, but it makes geometry a variable instead of a constant. Once geometry is a variable, engineers can attack fouling, flux and pressure drop separately.

• 1
  Cast filmRandom pores, low cost per square metre, limited geometry control.
• 2
  Printed membraneDesigned pores and channels, higher cost, geometry tuned per application.

Two printing routes dominate. The first is direct extrusion of a polymer dope through a fine nozzle. Lines of material are laid down and fused, leaving gaps that become pores. The second prints a sacrificial template, then backfills it with the membrane polymer and dissolves the template away.

Extrusion printing gives pore sizes roughly in the 0.1–10 μm band. Below that, nozzle diameter and polymer rheology fight back. The dope must be viscous enough to hold a printed line, yet fluid enough to exit a small tip. Those two demands pull in opposite directions.

Template printing reaches smaller features. A printed lattice can be backfilled and then removed, leaving channels set by the template rather than the nozzle. Feature sizes down to a few micrometres are reported. The trade-off is extra steps: print, backfill, cure, dissolve, rinse.

Print speed is the other lever. A membrane module needs a large surface area, often several square metres. A single nozzle writing that area line by line is slow. Multi-nozzle arrays and continuous roll-to-roll printing are the routes being pursued to bring unit cost down.

• 1
  Direct extrusionFewer steps, pore size set by nozzle and dope viscosity.
• 2
  Sacrificial templateFiner features, more process steps, higher scrap risk.
• 3
  ThroughputMulti-nozzle or roll-to-roll printing is needed for square-metre output.

Fouling is the first target. In a flat sheet, foulant collects on a flat plane and the deposit grows evenly. A printed membrane can carry a herringbone or chevron pattern that keeps the feed moving across the surface. Moving feed means the deposit is swept away instead of building up.

Pressure drop is the second. A printed support layer can taper its channels so the flow path widens toward the outlet. That spreads pressure more evenly across the sheet, which reduces the risk of a localized rupture. Cast supports are uniform because the casting process has no way to vary thickness locally.

Mechanical strength is the third. A printed membrane can carry a ribbed backing printed in the same pass as the separation layer. The ribs take the load, so the active layer can be thin. A thin active layer means lower resistance and higher flux at the same driving pressure.

None of this is free. A printed membrane has more internal surface area, and dead zones between printed lines can trap stagnant liquid. If the pattern is wrong, fouling gets worse, not better. Geometry helps only when it is matched to the feed stream.

• 1
  Patterned surfaceHerringbone or chevron features reduce deposit buildup.
• 2
  Tapered supportWidening channels spread pressure and cut rupture risk.
• 3
  Integrated ribsLoad goes into printed ribs, so the active layer stays thin.

Printable membrane polymers are still a short list. PVDF, PES, polyethersulfone blends and some polyamide chemistry print well. Ceramic-filled resins exist for higher temperature duty but shrink during sintering, which shifts pore size. Shrinkage of a few percent across a whole module is hard to correct after the fact.

Solvent choice limits the printer more than the polymer. Many membrane polymers need aggressive solvents that attack print heads, seals and tubing. Switching to a milder solvent usually costs pore uniformity.

This is where CNC work enters the picture. A printed membrane is a soft, thin, chemically specific component. It has to sit inside something rigid that holds it flat, seals it, and takes the plumbing load. That something is a machined housing, and it is usually the part that sets the module's dimensional accuracy.

Housing materials are ordinary by comparison: 6061-T6 aluminium for lab and pilot rigs, 316L stainless for anything with aggressive feed or clean-in-place cycles. Both machine cleanly to ±0.005 mm, which is what a face seal needs to hold at pressure.

• 1
  Printable polymersPVDF, PES, polyamide blends, some ceramic-filled resins.
• 2
  Solvent constraintAggressive dopes limit print head and seal life.
• 3
  Machined housing6061-T6 or 316L carries the seal and plumbing loads.

A pilot membrane module is a stack. Feed and permeate plates alternate, with the printed membrane between them. Each plate needs flat sealing faces, port holes, and often a channel milled into its face. Those plates are 3-axis or 4-axis milling work, and flatness matters more than any other feature.

Seal grooves are the tightest feature. An O-ring groove typically runs 2–4 mm wide and 1.5–3 mm deep, with the groove walls square to the floor. A groove that is 0.05 mm too deep will not compress the O-ring enough, and the module weeps under pressure.

Manifolds and end caps often need ports on more than one face. A 5-axis machine can cut an angled port and its sealing counterbore in one setup, which keeps the port axis true to the flange. Repositioning the part between setups is where angular error creeps in.

For long plates, 4,000 mm travel machines handle full-length parts without a splice. That matters when a plate carries a continuous distribution channel, because a joint in the middle becomes a leak path.

• 1
  Seal grooves2–4 mm wide, 1.5–3 mm deep, square walls, depth-critical.
• 2
  Angled ports5-axis cuts port and counterbore in one setup.
• 3
  Long plates4,000 mm travel avoids spliced distribution channels.