Laser Powder Bed Fusion: 3D-Printing Metal One Melt Pool at a Time

The build cycle: a part is a stack of welded cross-sections

An LPBF machine turns a CAD model into a solid by slicing it into thousands of horizontal cross-sections and welding each one out of powder. The build chamber is first sealed and purged with inert gas. A dose piston raises a reservoir of fine metal powder; a recoater — a hard blade, soft polymer lip, or counter-rotating roller — sweeps a uniform layer, typically 20–60 µm thick, across the build plate. A galvanometer scanner steers a focused laser beam through an f-theta lens, tracing the part outline and then filling the interior with closely spaced parallel lines (the hatch). Wherever the beam dwells, powder melts into a small pool that fuses to the solidified layer beneath and refreezes. The build plate then lowers by exactly one layer thickness, the recoater spreads fresh powder, and the process repeats.

The numbers make the tedium concrete: a part 100 mm tall built at a 30 µm layer needs roughly 3,300 recoat-and-scan cycles, and full builds run from several hours to a few days. Recoating a layer takes on the order of ten seconds; laser exposure per layer scales with the cross-sectional area. The feedstock itself is engineered: gas- or plasma-atomized spherical powder with a tight particle-size distribution around 15–45 µm, chosen so it flows and packs into a smooth, dense bed. Coarse, satellited, or moisture-laden powder spreads poorly and seeds defects.

Inside the melt pool: the physics the laser exploits

The heart of the process is a pool of liquid metal roughly 100–200 µm wide and tens of microns deep that races along at the scan speed, so any given point stays molten for well under a millisecond. Coupling the laser energy in is not trivial: polished metal reflects most near-infrared light (absorptivity of a flat surface at 1 µm is only ~0.3 for many alloys). The powder bed rescues the process — light bounces between particles and is trapped by multiple reflections, pushing effective absorptivity to ~0.5–0.7.

Inside the pool, two forces dominate the flow. Marangoni convection arises because surface tension varies with temperature (for a clean metal, dσ/dT is negative), dragging liquid from the hot center outward and stirring the pool. Recoil pressure from metal vaporizing at the beam center pushes the free surface down. At modest intensity the pool stays in conduction mode — a shallow, hemispherical bead. Crank up the power density and you enter keyhole mode: the recoil pressure gouges a deep vapor cavity whose walls trap light like a near-blackbody (absorptivity ~0.8+), giving deep penetration but risking gas bubbles pinched off at the keyhole tip — keyhole porosity. Because the surrounding solid is a giant heat sink, solidification is violent: cooling rates of ~10⁵–10⁶ K/s produce very fine, non-equilibrium microstructures. Ti-6Al-4V, for instance, quenches straight to needle-like α′ martensite, and grains grow epitaxially from the layer below into columns along the build direction — the root of LPBF's mechanical anisotropy.

Energy density and the printable process window

The single most useful knob for tuning a material is the volumetric energy density, E = P / (v·h·t), where P is laser power, v is scan speed, h is hatch spacing, and t is layer thickness. A representative recipe — 280 W, 1,200 mm/s, 0.14 mm hatch, 0.03 mm layer — lands near 55 J/mm³, and most alloys print in a window of roughly 40–120 J/mm³. That window is bounded on both sides by distinct failure modes.

• Too little energy → lack-of-fusion porosity. The pool fails to fully melt the powder or penetrate into the previous layer, leaving irregular voids and unmelted particles between tracks. These flat, jagged pores are the worst kind for fatigue.
• Too much energy → keyhole porosity and evaporation. Deep keyholing traps spherical gas pores, and volatile alloying elements (aluminum, magnesium, zinc) boil off, shifting composition away from spec.
• High speed, low power → balling. A thin, elongated melt track becomes capillary-unstable and breaks into a string of beads (a Plateau–Rayleigh instability), leaving a rough, poorly bonded surface.

Adjacent tracks overlap by ~30–50% of their width so hatch lines merge into a solid; each new layer must remelt into the one below for through-thickness fusion. Vapor and hot gas also entrain nearby powder, clearing a denudation zone and ejecting spatter that can fall back onto the bed. Every alloy on every machine gets its own tuned parameter map (power versus speed), which is why qualified LPBF parameters are guarded like recipes.

Residual stress, warping, and scan strategy

Rapid melting and freezing against cold, already-solid metal is a recipe for locked-in stress. Each freshly deposited track wants to contract as it cools but is restrained by the rigid material beneath — the thermal gradient mechanism — leaving the top in tension and the layers below in compression. Accumulated over thousands of layers, this residual stress curls edges upward, cracks brittle alloys, and can literally peel a part off its plate mid-build, wrecking the recoater on the next pass.

Engineers fight it on four fronts. First, the build plate is preheated (~200 °C for titanium and steels, higher for crack-prone nickel superalloys) to shrink the gradient. Second, the laser follows a scan strategy that never draws long continuous vectors: island or checkerboard patterns break each layer into small randomized tiles, and the hatch direction is rotated a prime-like 67° between successive layers (the EOS default) so stress and grain texture never align. Third, support structures — thin lattice anchors — are added under any overhang shallower than about 45° from horizontal, both to hold the part down against warping and to conduct heat out of features that would otherwise sit on insulating loose powder and overheat. Fourth, the whole build is given a stress-relief anneal while still bolted to the plate, before it is cut free with wire EDM or a bandsaw. Skip that step and the part springs into a taco the instant it is released.

Defects, densification, and qualification

A well-tuned LPBF part is typically ≥99.5–99.9% dense as-built, but 'dense' is not 'flawless.' For fatigue-critical aerospace and medical hardware, parts are further densified by Hot Isostatic Pressing (HIP): argon at ~100–200 MPa and high temperature (≈920 °C for Ti-6Al-4V, up to ~1,150–1,200 °C for nickel alloys) collapses internal pores and simultaneously heals the as-built microstructure. As-built surface roughness runs Ra ~5–20 µm — worse on down-facing (downskin) surfaces — so sealing faces, bearing surfaces, and fatigue-loaded skins are machined or polished, because surface roughness and near-surface pores are potent stress concentrators.

Because the process is stochastic, qualification leans on standards and inspection. The ISO/ASTM 52900 family sets terminology (the neutral name for LPBF is PBF-LB/M) and process/feedstock requirements (52904, 52907); material specs such as ASTM F2924 and F3001 (Ti-6Al-4V and its extra-low-interstitial grade) and F3055 (Inconel 718), plus aerospace AMS specifications, pin down chemistry, density, and mechanical properties. Parts are inspected by X-ray computed tomography for internal porosity, and modern machines add in-situ monitoring — melt-pool photodiodes, high-speed cameras, and layer-by-layer imaging — to flag anomalies as they form. Feedstock is tracked just as carefully: recycled powder picks up oxygen and loses sphericity with each reuse, so oxygen content and particle-size distribution are re-measured before it goes back in the machine.

What only LPBF can build — and where it still hurts

The payoff for all this difficulty is design freedom that subtractive and formative processes cannot match. Because material is added only where it is needed, engineers can route internal conformal cooling channels that curve through a mold or turbine part exactly along the heat, place triply-periodic lattices (like a gyroid) for stiffness at a fraction of the weight, and let topology-optimization software carve a bracket down to the pure load path. Just as valuable is part consolidation: assemblies that once needed dozens of machined pieces, brazes, and fasteners collapse into a single print.

The canonical example is GE Aviation's LEAP fuel nozzle tip, which merged roughly 20 separate parts into one cobalt-chrome print, came out about 25% lighter and around five times more durable, and has been produced in six-figure quantities. GE went further on its Catalyst turboprop, replacing 855 conventional parts with 12 printed ones. SpaceX's SuperDraco engine chamber is printed in Inconel, and titanium brackets and ducts fly on Airbus and Boeing aircraft. In medicine, LPBF prints titanium acetabular cups, spinal cages, and cranial/maxillofacial implants with porous lattice surfaces that encourage bone to grow in and lock the implant in place, plus cobalt-chrome dental frameworks.

The trade-offs are real. Build rates are slow — a single laser deposits only ~5–20 cm³/hour, which is why machines like the SLM Solutions/Nikon NXG XII 600 pack twelve lasers — parts are size-limited to the chamber (commonly ~250–500 mm, and a metre or more tall on the largest systems), and powder, inert gas, and post-processing make it expensive. Anisotropy, residual stress, and the heavy qualification burden mean LPBF wins where geometry or consolidation is worth the price, not where a lathe would do.

Three metal powder-bed families compared: laser fusion versus its electron-beam and binder cousins.

Attribute	LPBF (SLM / DMLS)	Electron Beam Melting (EBM)	Binder Jetting
Energy source	Yb-fiber laser, 1070 nm	Electron beam in high vacuum	Inkjet polymer binder (no melting)
Atmosphere	Inert gas: argon or nitrogen	High vacuum, bed preheated ~700–1000 °C	Ambient; sinter/infiltrate later
Layer thickness	~20–60 µm	~50–100 µm	~50–200 µm
Residual stress	High — steep gradients, needs stress relief	Low — hot bed anneals as it builds	None during printing (thermal in sinter)
Density / finish	≥99.5% as-built, Ra ~5–20 µm	Near-full, rougher surface	Porous 'green' part; shrinks in sinter
Best for	Fine detail, reactive alloys, Ti/Ni/Al/steel	Conductive metals, low-stress Ti/CoCr	High volume, low-cost steels/tungsten