Failure modes and how to read them

The egg-shell effect (substrate too soft for the coating)

This is the most common preventable failure in coated tooling, and the diagnostic is straightforward: the coating looks intact for a while, then suddenly the tool starts shedding metal at unexpected rates. Pull the part. Look at the substrate underneath any visible coating wear or chip. Is the substrate galled, deformed, or showing plastic flow? If yes — substrate was too soft. The hard coating sat on a yielding core; under load, the substrate deformed, the coating cracked across the deformed area, and the failure cascaded.

Investigation order: check the heat-treat hardness FIRST, before chasing the coating. Pull a witness coupon (or test an unworn area of the part if the geometry allows). Per industry standards, most PVD applications need 55-58 HRC minimum substrate; heavy-load stamping often needs 60-62. If the substrate reads 48 HRC, you found your problem and the coating is innocent.

Delamination (coating coming off in sheets)

A nearby variant: the coating peels or flakes off, leaving clean substrate exposed. Per Dayton Tool Grind & Coat and the wear-performance literature (kirj.ee paper on PVD-coated tool steels), the three big root causes are:

1. Surface prep failure at coating time. Oils, particles, or oxides were still on the surface when the coating was deposited. The bond was weak from day one and finally let go.
2. Substrate compatibility / prior treatments. The part had a prior incompatible treatment (nitriding compound layer, black oxide, plating) that wasn't fully removed before coating. The coating bonded to that layer, not the substrate, and the underlying layer eventually let go.
3. Operating temperature exceeded the coating's thermal limit. TiN's oxidation onset is commonly cited around 500-600°C in coating manufacturer datasheets. If the application gets hotter than that, the coating degrades and detaches. The fix is a higher-temperature coating like AlTiN (which forms a protective aluminum-oxide layer at temperature). Always check the specific oxidation onset on your coating supplier's datasheet for the grade you're running.

Compound-layer cracking on nitrided parts

The white layer in nitriding is hard but brittle. Per the Cracking Resistance paper in the finishing-and-coating literature and ScienceInsights' process notes, when compound-layer-cracking shows up the root cause is one of:

• Layer too thick (above ~10 µm, the compound layer becomes brittle enough to crack under load)
• Substrate not pre-stress-relieved before nitriding (residual stresses propagate into the layer)
• Impact loading exceeding the layer's tolerance (the compound layer hates point loads and shock)

The fix: thin or polish off the compound layer (the diffusion zone underneath still provides the hardness), or switch to plasma nitriding with a recipe that produces a thinner or compound-free layer.

Same die, different plant

This is the diagnostic story every rep eventually tells: two identical dies, same coating, same heat treat. Plant A runs the die for 200K hits. Plant B runs the die for 40K. Why? The die is identical. The system is not.

Variables to investigate on the plant side:

• Lubricant chemistry and quantity. Different oils have different coefficient-of-friction profiles. Different application rates (light, heavy, none) change wear mode entirely.
• Press speed and tonnage profile. Faster cycles = more heat in the die. Different tonnage = different elastic deformation.
• Material variation. Incoming stock has hardness variation by supplier and lot. A harder-than-spec coil is a more aggressive wear partner.
• Die temperature management. If Plant B isn't cooling the die between hits, temperature creeps up, lubricant breaks down, wear accelerates.
• Setup quality. Alignment, hit position, parallelism — all change the wear vector. A subtly misaligned die wears asymmetrically.
• Operator practice. Anyone running it dry "just for a few hits"? Anyone leaving the press loaded overnight in a humid plant?

The honest answer that customers need to hear: coating performance is system performance, not part performance. Two plants running the same die can get very different lives, and the difference is rarely the die.

What this means on the shop floor

• When a die fails, photograph the failure surface BEFORE doing anything else. Save the photos with the part history. Most failure analyses fail because the evidence was destroyed by handling and cleanup.
• Distinguish failure modes by the substrate condition, not the coating condition. The coating tells you the symptom. The substrate tells you the cause.
• Build a failure log. Date, part number, coating type, hardness spec, hits to failure, failure mode photo, lubricant in use, suspected cause. After 20 entries you'll see patterns nobody at the plant has noticed.

Next: how to talk to your vendors so they take you seriously and give you their best work.