Metal AM Defects: Types, Causes, and How to Eliminate Them

The parameter window for dense, defect-free metal AM is narrower than most machine vendors imply. Every defect type has a distinct cause, a distinct appearance under inspection, and a distinct remedy. Treating all porosity as "too much/too little energy" is the fastest way to spend weeks chasing the wrong variable.

This article maps each defect to its mechanism and gives you the practical response.

Defect classification

Metal AM defects split into three origin classes:

Process-induced defects arise from beam/powder interaction — porosity, balling, delamination. These respond to parameter changes.

Material-induced defects arise from alloy chemistry — hot cracking, solidification cracking, liquation cracking. Parameter changes help but alloy selection or modification is often required.

Geometric defects arise from part design interacting with the process — staircase roughness, elephant footing, support witness marks. These respond to design or orientation changes.

This article focuses on the first two categories, which drive scrapped builds and rework in production.

1. Lack-of-fusion (LOF) porosity

Mechanism: When volumetric energy density (VED) is insufficient, the laser or electron beam does not fully melt the powder layer. Adjacent melt tracks do not overlap enough to fuse. Solid particles are encapsulated by partially melted material. The result is an irregular void network aligned with the scanning pattern.

Volumetric energy density is the first-order predictor:

VED = P / (v × h × t)

Where P = power (W), v = scan speed (mm/s), h = hatch spacing (mm), t = layer thickness (mm). Units: J/mm³.

LOF typically appears below:

• 40–50 J/mm³ for 316L stainless steel
• 45–60 J/mm³ for IN718
• 50–65 J/mm³ for Ti-6Al-4V

These are indicative — the exact threshold is parameter-set and machine-specific.

Appearance:

• Irregular, angular voids (not spherical)
• Aligned with the scan direction or layer boundaries
• Partially melted powder particles visible inside the void (XCT or polished cross-section)
• Often found in clusters where adjacent tracks failed to fuse

Detection: XCT resolves LOF pores well because they are large (50–500 µm) and irregular. Cross-section metallography with etching makes the partially melted powder particles visible. Archimedes density is insensitive unless LOF is severe (> 0.5% porosity).

Remedies (in order of effectiveness):

1. Increase laser power while holding other parameters constant — highest single-lever impact
2. Decrease scan speed (increases dwell time per unit length)
3. Decrease hatch spacing (increase track overlap to ≥ 20%)
4. Decrease layer thickness (less powder to melt per layer)
5. Re-qualify powder PSD — coarse or bimodal PSD requires more energy to fully melt
6. Check packing density — low apparent density in the powder bed reduces effective coupling

The porosity predictor and volumetric energy density tool can help you identify whether a given parameter set is in the LOF regime before running a build.

2. Keyhole porosity

Mechanism: The opposite extreme from LOF. At high VED, the laser creates a deep, narrow vapour channel (the "keyhole") through the melt pool. The keyhole is dynamically unstable — surface tension and vapour pressure forces compete. When the keyhole collapses, it traps a bubble of metal vapour that cannot escape before solidification. The result is a spherical pore below the surface.

Normalised enthalpy is a better predictor of keyhole onset than VED alone:

ΔH/hs = (A × P) / (π × ρ × Cp × Tm × √(α × v × D³))

Where A = absorptivity, ρ = density, Cp = specific heat, Tm = melting temperature, α = thermal diffusivity, D = beam diameter. Keyhole onset typically occurs at ΔH/hs > 6–10 (material-dependent).

Appearance:

• Spherical, smooth-walled pores (distinguishes from LOF)
• Deeper than one layer thickness from the surface — often found at 100–500 µm depth
• Random spatial distribution, not aligned with scan pattern
• Pore diameter typically 20–150 µm

Detection: XCT is definitive. The smooth spherical shape is clearly distinguishable from LOF porosity. Cross-section metallography works but orientation relative to the slice plane matters. Archimedes density picks up keyhole porosity if it exceeds 0.2%.

Remedies:

1. Reduce laser power (most direct lever)
2. Increase scan speed (reduce energy per unit length)
3. Increase beam diameter (defocus) — spreads energy, reduces peak intensity
4. Check for focal plane drift — keyhole porosity that appears late in long builds can indicate Z-axis focus shift
5. For multi-laser machines: verify beam overlap zones have consistent parameters, as over-exposure at stitch zones generates keyhole pores

Important: Keyhole pores are spherical and have lower stress concentration than LOF pores, but they nucleate fatigue cracks efficiently. Even at 0.1% apparent porosity, a 100 µm spherical pore in a high-stress region will dominate fatigue life.

3. Solidification (hot) cracking

Mechanism: This is a material-induced defect. During solidification, the alloy passes through a "mushy zone" where liquid and solid coexist. If the solidification shrinkage + residual stress exceeds the strength of the partially solidified material (which is near zero at the solidus), the material tears. The result is intergranular cracks — cracks that follow grain boundaries, where the last liquid to solidify creates a continuous liquid film.

Alloys susceptible to hot cracking in LPBF:

Hot cracking in nickel superalloys is the primary reason that many high-temperature alloys remain difficult or impossible to process by LPBF without modification.

Appearance:

• Intergranular: cracks follow columnar grain boundaries
• Present even in samples with otherwise low porosity
• Often propagate across multiple layers
• Visible as branched networks on polished cross-section (etched)

Remedies:

1. Preheat — reducing thermal gradients reduces the stress on the mushy zone. EBM's high preheat is why EBM IN718 and CoCrMo are less susceptible. Limited preheat in LPBF (to 200°C) helps but rarely eliminates cracking in susceptible alloys.
2. Grain refiner additions — nanoparticle inoculants (ZrO₂, TiC, WC) added to the powder promote equiaxed nucleation, which is less susceptible to hot cracking than columnar grains. This is an active research and commercialisation area.
3. Scan strategy modification — shorter scan vectors, island scanning, beam shaping to modify melt pool geometry. Reducing the length of columnar grains reduces the crack propagation path.
4. Alloy reformulation — reduced aluminium + titanium content in Ni superalloys (e.g., IN625-like compositions) reduces γ' fraction and widens the process window. Alloy 713LC, modified CM247, and SB-CoNiCrAlY variants have been developed specifically for AM.
5. Accept + HIP — for alloys where low crack density is achievable, HIP at appropriate temperature and pressure can close cracks. This works for short, isolated cracks but not for interconnected crack networks.

4. Delamination (layer cracking)

Mechanism: Residual stress in LPBF builds in the Z-direction (perpendicular to the build plate). When accumulated tensile stress exceeds the interlayer bond strength, the layers peel apart. This is distinct from hot cracking — it is a mechanical failure of the built part, not a solidification defect.

Conditions that cause delamination:

• Scan vectors aligned in the same direction across many layers (no rotation) — stress accumulates uniaxially
• High aspect ratio features (tall, thin walls) where stress builds over many layers without geometric relief
• Poor inter-layer bonding from LOF conditions (low VED) at layer interfaces
• Material with high thermal expansion coefficient and low conductivity (austenitic steels, titanium)

Appearance:

• Horizontal or near-horizontal cracks aligned with the build plane
• Often visible on the part exterior after build completion
• In severe cases, the part visually separates from the base plate or splits internally
• XCT shows planar, low-aspect voids oriented perpendicular to the build direction

Remedies:

1. Island scanning with 67° rotation — the standard mitigation. Divide each layer into small islands (3×3 mm to 7×7 mm), scan each island independently, rotate the scan direction 67° per layer. This distributes residual stress across multiple directions and prevents uniaxial accumulation.
2. Platform preheat — even 100–200°C reduces thermal gradient and inter-layer stress.
3. Redesign geometry — reduce wall aspect ratio, add stiffening ribs or section breaks for tall thin walls.
4. Reduce scan vector length — shorter scan vectors reduce the stress accumulated per scan track. Achievable via island scanning or by orienting the part so the largest dimension is not perpendicular to the beam travel direction.
5. Post-build stress relief before base plate removal — mandatory for high-aspect-ratio features.

5. Balling

Mechanism: The melt pool minimises surface energy. If the ratio of surface tension force to inertial/gravitational force (the Ohnesorge number) is too high, the melt pool spheroidises rather than wetting and spreading across the powder layer. The result is a ball of solidified metal, disconnected from the surrounding powder, rather than a continuous track.

Conditions that promote balling:

• Low scan speed combined with low power — very small melt pool with high surface tension
• Oxide layer on powder particles — reduces wettability and increases surface tension of the melt
• Re-used or contaminated powder — higher oxygen content
• First layer on a cold build plate — poor thermal coupling

Consequences:

• Surface roughness spikes (Ra doubles or triples)
• Balled material interferes with the recoater blade on subsequent layers
• In severe cases, the recoater crashes into the solidified balls and crashes the build
• Inter-layer voids where balls fail to provide a continuous melting substrate for the next layer

Remedies:

1. Optimise power/speed combination to maintain continuous track formation — run single-track experiments at varying power/speed and map the regime boundaries
2. Verify inert atmosphere quality — oxygen content in the build chamber should be < 0.1% (< 1000 ppm) for most alloys, < 100 ppm for titanium
3. Use fresh or freshly sieved powder — measure apparent density and flowability; degraded powder balls more readily
4. Set appropriate first-layer parameters — many machines have a "base layer" parameter set with higher power and lower speed for the first 1–5 layers

6. Satellite formation and surface roughness

Mechanism: At the melt pool boundary, partially melted powder particles adhere to the solidifying surface. They are not fully melted but are sufficiently heated to bond. The result is satellite particles attached to the as-built surface, increasing Ra and creating stress concentration sites on fatigue-loaded surfaces.

Downskin and upskin surfaces have different satellite formation characteristics:

• Downskin (over-hanging loose powder): melt pool sags, contacts more surrounding powder, higher satellite density
• Upskin (top surfaces): better thermal support, fewer satellites

Remedies:

1. Contour exposure parameters — a separate parameter set for the outer perimeter of each layer that uses lower power, lower speed, and a smaller offset from the nominal edge. This refines the melt pool at the boundary and reduces satellite attachment.
2. Downskin parameters — dedicated parameter set for down-facing layers: lower power, reduced speed, possibly reduced layer thickness for the first 2–5 layers above a down-facing surface.
3. Post-processing: shot peening, abrasive flow machining, electropolishing, or vibratory finishing to reduce surface Ra before fatigue or sealing applications.

7. Defect detection methods compared

Recommendation for production qualification: Use Archimedes density as a fast process control metric (measurable on witness coupons in 10 minutes), but validate the process window with XCT on at least one full sample per new alloy or parameter set. Cross-section metallography is essential for hot crack assessment.

8. Defect impact on fatigue life

Not all defects are equally damaging. In fatigue loading, cracks initiate from the largest, sharpest stress concentration in the highest-stress region.

LOF pores are the most damaging per unit volume: their irregular geometry (sharp corners, large effective diameter) and alignment with layer planes create high stress concentrations. A 200 µm LOF pore reduces fatigue life by a factor of 3–10× compared to a defect-free sample in Ti-6Al-4V.

Keyhole pores are spherical — much lower stress concentration factor. A 100 µm keyhole pore has approximately half the fatigue impact of a 100 µm LOF pore. However, sub-surface keyhole pores are not detectable by dye penetrant and can escape inspection.

Hot cracks are the most damaging: sharp intergranular cracks with minimal radius of curvature are essentially pre-existing fatigue cracks. Even 100 µm intergranular cracks in high-cycle fatigue applications (aerospace turbine blades, medical implants) are disqualifying.

Design guidance: For fatigue-critical parts, specify the acceptable defect type and size in the build acceptance criteria. "Less than 0.5% total porosity" is not sufficient — a single 500 µm LOF pore in a high-stress region is worse than 0.1% uniformly distributed keyhole porosity.

Further reading

• Porosity predictor — predict porosity regime from power/speed/hatch/layer parameters
• Volumetric energy density tool — calculate VED and check against LOF/keyhole thresholds
• Melt pool geometry tool — model melt pool depth and width to assess LOF/keyhole risk
• NDT selector — choose the right inspection method for your defect type and part geometry
• Laser parameter optimizer — systematic parameter development with defect regime mapping