3D printed metal lattice structures: lighten, cool and filter
A lattice is a repeating structure that fills a volume with very little material. It's one of the few things additive manufacturing can do that no other technology can: there's no way to mill, cast, or machine an internal mesh of half-millimeter bars inside a closed part.
And it is also where most projects get stuck, almost always for the same reason, which is not the design but the powder that remains inside.
The three types, and what each one is used for
| Family | What it is | What it's good for |
|---|---|---|
| Bar (cubic, octet, diamond) | Nodes joined by straight bars | Lighten while maintaining rigidity. Absorb energy in an impact |
| Surface (TPMS: gyroid, diamond, primitive) | A continuous, curved surface that divides the volume into two non-touching channels | Exchange heat. High surface area per unit volume and no stress-concentrating nodes |
| Stochastic (foam) | Irregular cells, like a sponge | Filter, cushion, promote bone ingrowth in implants |
The choice isn't about aesthetics. A lattice of bars optimized for rigidity is poor at heat exchange, and a gyroid is an excellent heat exchanger but a poor shock absorber.
What can really be manufactured
The numbers that follow are orders of magnitude for thin-film LPBF with a well-parameterized alloy. They change with the material, the machine, and the orientation, and some shop floors exceed these values because they have qualified their process to do so.
| Feature | What works | The limit | What happens if you go too far |
|---|---|---|---|
| Bar diameter | 0.6-1 mm | ~0.3-0.4 mm | Incomplete bars, unfused nodes, lattice crumbles when dusting |
| Wall thickness in TPMS | 0.4-0.8 mm | ~0.3 mm | Holes and leaks between the two circuits |
| Cell size | 3-8 mm | ~1.5 mm | Powder no longer comes out below |
| Bar angle | ≥ 45° to the horizontal | ~30-35° if short | Rough bottom face and local deformation |
| Layer thickness | 30-50 µm | — | With a thick layer, a thin bar takes two or three passes and comes out poorly |
The rule that saves the most trouble: horizontal bars are the enemy. A classic cubic cell has bars at 0°, which are pure cantilevers and turn out badly. An octet or diamond cell places almost all the bars above 45° and is therefore much easier to manufacture, with the same mass.
The problem no one warns about: powder
An internal lattice with 5-millimeter cells inside a sealed part remains filled with unmelted powder. And that powder:
- It's heavy, and often the lattice is used precisely to reduce weight;
- It loosens over time and with vibrations, and ends up where it shouldn't;
- It prevents the part from being certified, because it cannot be proven that it is empty;
- Expensive material is lost every time it gets caught.
Bringing it out requires three things from the design stage, not afterwards:
- Drainage openings in the correct position: open to the outside and positioned so that the powder falls by gravity in the dedusting direction.
- Open cells connected to each other. A closed cell is a permanent powder capsule.
- A dedusting procedure that sometimes includes vibration, compressed air, or immersion, and which must be tested on a test part before manufacturing the final product.
And one honest check at the end: weigh the part. If it weighs more than the model number indicates, there's powder inside. It's the cheapest check there is, and almost no one does it systematically.
The three uses that do pay
Reduce weight where it matters. Fill the interior of a thick part with a lattice structure instead of leaving it solid. This saves material, reduces laser cutting time, and minimizes residual stresses in a large mass. It's the most straightforward application.
Heat exchange. This is where additive manufacturing wins hands down. A TPMS offers a heat exchange surface area per unit volume that no plate or tube heat exchanger can match, with the two circuits intertwined and without a single weld. It's the same argument as conformal cooling in moulds, taken to the extreme.
Filter and diffuse. Controlled porosity metal foams for filters, silencers, gas diffusers, and burners. They replace sintered parts that require assembly and are integrated directly into the housing.
In implants there is a fourth use —surface porosity to promote osseointegration— which is a world apart with its own regulations.
How long does it take to manufacture them
A word of caution to avoid budget surprises: a lattice structure isn't always faster than the solid part it replaces. Laser cutting time depends on the amount of contour that needs to be tracing, and a lattice structure is almost entirely contour. A part lightened to 30% less mass can take the same amount of time as the solid part, or even longer.
What it does always save is material and weight. If the goal was to save machine time, it's worth checking beforehand: the quote calculator gives the estimated time with the actual geometry, and you'll see it right away.
In summary
Lattice structures are among the most valuable aspects of this technology, and they demand the most discipline. The sequence that works is always the same: choose the family based on its function, not the photo; keep the bars above 45°; respect the minimum thicknesses; and decide how the powder will be distributed before drawing the first cell.
If you want to see actual parts with lattices and internal geometries, they are in the parts mosaic. And the general rules that also apply here are in design rules for LPBF.
Business9 September 2026
How to order a printed metal part without making a mistake
Sustainability10 September 2026
Eco-design and digital product passport: what awaits those who manufacture in metal
Process31 August 2026