White layer: the surface damage that passes inspection and fails in service

A part comes back from the shop. Every dimension is in tolerance, the surface finish reads Ra 0.3, the hardness test passes. It goes into service and cracks at a third of its expected life.

This is the failure mode that dimensional inspection cannot see.

What white layer is

When hardened steel is machined with too much heat — a worn edge, excessive speed, inadequate chip evacuation — the surface can reach the austenitising temperature momentarily and then be quenched by the bulk of the cold part beneath it.

That is a heat treatment cycle, happening in microseconds, at the surface only. The result is a thin layer of untempered martensite, typically 1-20 µm deep. It etches white under a microscope, which is where the name comes from.

Directly beneath it sits a dark layer — over-tempered material, softer than the parent metal.

Both matter, and the second is often worse than the first.

Why it fails parts

The white layer is hard and brittle. Untempered martensite can reach 65+ HRC with essentially no toughness. It cracks readily.

Underneath it, the residual stress is tensile. This is the real problem. A machined surface in good condition carries compressive residual stress, which suppresses crack initiation. Thermal damage flips it to tensile, which does the opposite.

The dark layer is soft. A hard brittle skin sitting on over-tempered material is a poor structure to load.

The practical consequence is a large reduction in fatigue life — studies on hard-turned bearing steel have shown rolling contact fatigue life falling by more than half when a continuous white layer is present. On a die that sees millions of cycles, or a shaft under rotating bending, that is the whole design margin.

Where it shows up

• Hard turning of bearing races and seal journals. The classic case, and the reason hard turning had a poor reputation in bearing work for years.
• Hard milling with a worn edge. Tool wear raises cutting temperature sharply. The last part before a tool change is the one at risk.
• Grinding burn. Same mechanism, different process. Inadequate coolant, a glazed wheel, or too aggressive a pass.
• Wire EDM recast layer. Related but distinct — EDM melts and re-solidifies material, leaving a recast layer with tensile stress and often micro-cracks. See wire EDM in hardened dies.

Why inspection misses it

A CMM measures geometry. White layer is 5 µm thick; it does not change any dimension you are measuring.

A surface roughness tester measures texture. A thermally damaged surface can be very smooth. Ra tells you nothing about what is underneath it.

A bulk hardness test measures the wrong thing. A Rockwell C indenter penetrates far deeper than the damaged layer. It reads the parent material.

So the standard proof package — dimensional report, surface finish, hardness — passes a part that is compromised.

How to actually detect it

Nital etch. The practical shop-floor method. The surface is etched with dilute nitric acid in alcohol; thermally damaged areas discolour distinctly. Non-destructive, cheap, well established in gear and bearing work. ASTM E1444-adjacent practices and internal aerospace standards specify it formally.

Metallographic cross-section. Definitive. Cut, mount, polish, etch, look at it under a microscope. Destructive, so it is done on a sample or a sacrificial first article.

Micro-hardness traverse. A series of very light indents from the surface inward reveals both the hard white layer and the soft dark layer below it. Also destructive.

X-ray diffraction residual stress measurement. Measures the thing that actually matters. Expensive and specialist, reserved for critical components and process validation.

How to specify against it

Three approaches, in increasing order of cost:

1. State the requirement in words. “Surface shall be free from thermal damage, untempered martensite and over-tempered material.” This puts the obligation on the shop and gives you recourse. It costs nothing.

2. Require a nital etch on 100% of the critical surfaces, or on a defined sample. Modest cost, catches gross process failures.

3. Require a first-article metallographic section plus process controls — tool change intervals tied to measured wear, documented cutting parameters. This is what regulated industries do.

For a one-off die block that sees moderate load, option 1 is proportionate. For a shaft under rotating bending in a machine that runs continuously, option 3 is cheap insurance.

What to put on the drawing

Mark which surfaces are fatigue critical. This is the single most useful thing you can do, because it tells the shop where to be careful and where it does not matter. A die face that sees only compressive load does not need this treatment; a fillet radius on a rotating shaft does.

Then say what the loading is. “Rotating bending, 10^8 cycles design life” tells a shop far more than a tolerance does, and a good one will adjust the process — fresh edge for the finishing pass, conservative parameters, verification — without being told how.

Put the loading condition in the notes field when you send the RFQ. The inspection guide covers what to ask for in the proof package.