Hydrogen Embrittlement

How hydrogen embrittlement works

Hydrogen is the smallest atom there is. A single proton with one electron, it slips into the gaps between iron atoms — the interstitial sites of the steel lattice — more easily than any other element. Once inside, it doesn't sit still. Atomic hydrogen in body-centred-cubic iron is astonishingly mobile, diffusing roughly as fast at room temperature as carbon does at 900 °C. That mobility is the whole story: hydrogen goes looking for the most highly stressed place in the part, gathers there, and weakens the metal exactly where it can least afford to be weak.

The damage happens in three acts. First, atomic hydrogen gets into the steel — from acid pickling, electroplating, electrolytic cleaning, welding with damp flux, or cathodic protection that over-charges the surface. (It must be atomic; the H₂ molecule is too big to enter. Only nascent H from a chemical or electrochemical reaction at the surface can be absorbed.) Second, under sustained tensile load the triaxial stress field ahead of a notch, thread root, or crack tip lowers hydrogen's chemical potential there, so dissolved hydrogen drifts toward that hotspot and its local concentration climbs far above the bulk average. Third, once the local hydrogen exceeds a critical level, atomic bonds across grain boundaries or cleavage planes give way, a brittle crack jumps forward, the stress field moves with it, and the cycle repeats. The part fails with no warning and no visible deformation.

Why does hydrogen weaken the bond? Two mechanisms dominate the textbooks, and both are usually at play. HEDE (hydrogen-enhanced decohesion) says dissolved hydrogen lowers the cohesive energy of the lattice and grain boundaries, so a smaller stress pulls atoms apart. HELP (hydrogen-enhanced local plasticity) says hydrogen makes dislocations glide more easily on confined slip planes, concentrating plastic flow into a thin band that then fails in a brittle-looking way. Real fractures show a mix — the practical takeaway is the same: hydrogen plus a hard microstructure plus tensile stress equals delayed brittle cracking.

The governing physics

Three relationships set the behavior: how fast hydrogen moves, how much gets in, and how it piles up at stress.

Diffusion follows Fick's law with an Arrhenius-temperature-dependent diffusivity. This is why baking works — raising temperature exponentially speeds hydrogen's escape:

Flux:           J = -D · dC/dx          (Fick's first law)

Diffusivity:    D = D₀ · exp(-Q / R·T)

  D₀ ≈ 1.5×10⁻⁷ m²/s   (lattice H in α-iron, trap-free)
  Q  ≈ 6.7 kJ/mol       (activation energy, trap-free lattice)
  R  = 8.314 J/mol·K
  → trap-free lattice D(25 °C) ≈ 1×10⁻⁸ m²/s

But real steel has traps (dislocations, carbides, grain
boundaries) that slow H. The EFFECTIVE diffusivity is far lower:
  At 25 °C:  D_eff ≈ 1×10⁻¹¹ m²/s  (ferritic/martensitic steel)
  At 25 °C:  D_eff ≈ 1×10⁻¹⁶ m²/s  (austenitic 316 — 100,000× slower)

Characteristic out-diffusion time for a part of half-thickness L:
  t ≈ L² / D
A 5 mm-thick bolt (L = 2.5 mm) at 200 °C, D_eff ≈ 5×10⁻¹⁰ m²/s:
  t ≈ (0.0025)² / 5×10⁻¹⁰ ≈ 12,500 s ≈ 3.5 h  → matches a bake spec

How much hydrogen dissolves at the surface follows Sieverts' law — the dissolved concentration scales with the square root of the hydrogen partial pressure (or fugacity), because the H₂ molecule must dissociate into two atoms to enter:

Sieverts' law:  C = K_s · √(f_H₂)

  C    = dissolved atomic-H concentration
  K_s  = solubility constant (temperature dependent)
  f_H₂ = hydrogen fugacity at the surface

In high-pressure H₂ service (e.g. 70 MPa fuelling), fugacity
can exceed pressure, so √f drives surface concentration up fast.

Finally, the piling-up. Hydrogen's local concentration C ahead of a crack tip is set by the hydrostatic (triaxial) stress σ_H through an Oriani-equilibrium Boltzmann factor — the higher the tensile triaxiality, the more hydrogen accumulates:

C(x) = C₀ · exp( σ_H · V_H / (R·T) )

  σ_H = hydrostatic stress at the point
  V_H = partial molar volume of H in iron ≈ 2.0×10⁻⁶ m³/mol
  C₀  = bulk lattice concentration

Failure occurs when the time-dependent K reaches a threshold:
  crack grows once  K_I ≥ K_TH
  K_TH falls as diffusible H content rises — the design number to beat.

The engineering parameter that ties this together is the threshold stress intensity K_TH (also written K_IH or K_ISCC): the stress-intensity below which a hydrogen-charged crack will not grow no matter how long you wait. For a clean high-strength steel K_TH might be 50–90 MPa·√m, but charge it with hydrogen and K_TH can collapse below 20 MPa·√m — meaning a flaw a fraction of the size that was previously safe now propagates.

Worked example: a Grade 12.9 bolt baked vs un-baked

Take an M10 Grade 12.9 socket-head cap screw — tensile strength 1220 MPa, hardness about 39–44 HRC, right on the susceptibility cliff. It's zinc-electroplated for corrosion protection. Plating is a cathodic process: hydrogen is generated at the part surface and a fraction of it is absorbed into the steel.

Bolt:            M10 × 1.5, Grade 12.9
Proof load:      ~57 kN     (88% of yield, per ISO 898-1)
UTS:             1220 MPa
Hardness:        39–44 HRC  → above the ~40 HRC threshold
Plating:         8 µm zinc, acid bath (high H pickup)

Absorbed diffusible hydrogen after plating:  ~3–6 ppm
Critical level for this strength:            ~1 ppm
→ Un-baked, the bolt is ~3–6× over the danger line.

Loaded to proof at the thread root the stress-concentration factor is roughly 3, so local stress reaches ~2,000 MPa with strong triaxiality — a perfect hydrogen trap. Un-baked, this bolt commonly fails by delayed fracture 24 to 72 hours after torque-up, snapping at the first engaged thread with a flat, granular, deformation-free face. There is a documented industry pattern of new-construction galvanized or zinc-plated high-strength bolts (ASTM A490, F3043) failing this way days to weeks after installation.

Now bake it. Per ASTM B850 / AMS 2759/9, start the bake within 4 hours of plating, hold at 200 ± 10 °C for the duration set by strength class:

Bake at 200 °C, D_eff ≈ 5×10⁻¹⁰ m²/s
Bolt half-thickness L ≈ 2.5 mm (root)

Out-diffusion time  t ≈ L²/D ≈ 12,500 s ≈ 3.5 h (single time-constant)
Spec for ≥1200 MPa class: 23 h hold (several time-constants for margin)

Result: diffusible H drops from ~4 ppm to <0.5 ppm.
K_TH recovers from ~15 MPa·√m back toward ~60 MPa·√m.
→ Baked, the same bolt passes a 200-hour sustained-load test (ASTM F519).

One subtlety the numbers hide: zinc and cadmium coatings act as a hydrogen barrier, slowing the escape and forcing the long 23-hour holds. And baking only helps before cracking starts — once a crack has initiated, no amount of baking will heal it.

What raises and lowers susceptibility

Factor	Raises risk	Lowers risk	Why
Strength / hardness	>1000 MPa (~32 HRC), worse ≥40 HRC	<1000 MPa, <30 HRC	No plastic relaxation at crack tip; hard martensite can't blunt cracks
Crystal structure	BCC (ferritic/martensitic)	FCC (austenitic)	FCC diffuses H ~10⁵× slower, so it can't concentrate
Microstructure	Untempered martensite	Tempered martensite, bainite	Tempering reduces residual stress and traps H more benignly
Stress state	High triaxiality (notches, threads)	Pure shear / compression	Hydrostatic tension draws H in; compression pushes it out
Loading	Sustained / static	Short-duration, no dwell	H needs time to diffuse to the tip
Temperature	Near room temp (~−50 to +100 °C)	Cryogenic or hot (>150 °C)	Worst when H is mobile but the lattice is still brittle; hot lets H escape
Hydrogen source	Acid pickle, electroplate, sour service	Mechanical clean, dry environment	Cathodic and acid reactions liberate absorbable atomic H

Note the temperature inversion: hydrogen embrittlement is worst near room temperature. Go cold and hydrogen freezes in place (can't diffuse to the tip); go hot and it diffuses straight back out. This is the opposite of most embrittlement modes and routinely surprises people.

Hydrogen embrittlement vs related failure modes

	Hydrogen embrittlement	Stress corrosion cracking	Corrosion fatigue	Liquid metal embrittlement	Temper embrittlement
Driving agent	Atomic H in lattice	Corrosive env. + tensile stress	Corrosion + cyclic stress	Molten metal (Zn, Cd, Ga) contact	P/Sn segregation to grain boundaries
Load type	Static / sustained	Static / sustained	Cyclic	Static (under wetting)	Static (impact toughness loss)
Timing	Delayed (hours–days)	Delayed (weeks–months)	Progressive	Near-instant once wetted	Reveals on later impact
Needs active corrosion?	No (works in dry air)	Yes	Yes	No (needs molten metal)	No
Fracture appearance	Intergranular / quasi-cleavage	Branched, intergranular/transgranular	Beach marks + corrosion	Intergranular, metal film on facets	Intergranular, no metal film
Reversible by heat?	Yes, if baked before cracking	No	No	No	Partly (re-temper + fast cool)
Worst temperature	~Room temp	Service temp	Service temp	Above coating melt point	~370–575 °C exposure

Where it shows up — real systems and failures

• High-strength fasteners. The classic case. ASTM A490 and F3043 structural bolts, Grade 12.9 / 10.9 cap screws, and aerospace AN/MS bolts. The 2013 San Francisco–Oakland Bay Bridge had 32 of its A354 BD anchor rods fracture within weeks of tensioning — a hydrogen/stress-corrosion event traced to galvanizing and trapped moisture. Failures cluster days to weeks after final torque.
• Electroplated landing-gear and airframe steel. 300M and 4340 steel landing gear is heat-treated to ~1900 MPa — extremely susceptible. Cadmium plating, the standard corrosion finish, is a strong hydrogen barrier, which is exactly why ASTM F519 and AMS 2759/9 mandate a 23-hour bake at 190 °C within hours of plating.
• Oil-and-gas "sour service." Wells with H₂S generate atomic hydrogen at the steel surface; this is sulfide stress cracking (SSC), a hydrogen-assisted mode. NACE MR0175 / ISO 15156 caps hardness at 22 HRC for carbon-steel components in sour service precisely to stay under the embrittlement threshold.
• Hydrogen fuel and pipeline infrastructure. 70 MPa fuel-cell vehicle tanks, electrolyzers, and hydrogen-blended natural-gas pipelines all expose steel to high hydrogen fugacity. ASME B31.12 governs hydrogen piping and forces conservative stress-design-factor derating because of embrittlement.
• Galvanized prestressing and suspension cables. High-strength bridge wire and post-tensioning strand can suffer hydrogen-assisted cracking where galvanizing, moisture, and high sustained stress combine.
• Welding. Hydrogen from damp electrode flux or surface moisture causes "cold cracking" (also called HAZ hydrogen cracking) hours after a weld cools — the reason low-hydrogen electrodes are stored in heated ovens and preheat is specified for hardenable steels.

How engineers prevent it

• Stay below the strength threshold when you can. If a part doesn't need >1000 MPa, don't specify it. Sour-service rules cap hardness at 22 HRC for this reason.
• Bake after plating. 190–220 °C, started within 1–4 hours of plating, held 3–23 hours by strength class (ASTM B850, AMS 2759/9). The single most effective production control for plated high-strength steel.
• Choose low-hydrogen finishing. Mechanical zinc plating, zinc-flake (Dacromet/Geomet) dip-spin coatings, ion-vapor-deposited aluminium (IVD Al) and physical vapor deposition put little or no diffusible hydrogen in. Many specs now prefer these over electroplated cadmium.
• Pick a forgiving microstructure. Fully tempered martensite or bainite over untempered martensite; austenitic stainless where corrosion allows; reduce phosphorus/sulfur to keep grain boundaries strong.
• Cut the hydrogen source. Low-hydrogen welding electrodes baked and stored hot, preheat for hardenable steels, avoid over-aggressive acid pickling, and limit cathodic-protection potentials so you don't over-charge the surface.
• Test the process, not just the part. Qualify every plating and cleaning line with ASTM F519 sustained-load specimens or ASTM F1624 incremental step-load tests; measure incoming hydrogen content by melt extraction.

Common misconceptions and pitfalls

• "It's just rust / corrosion." No. Internal hydrogen embrittlement needs no active corrosion and happens in dry air. The hydrogen is already inside the metal from manufacturing.
• "Baking always fixes it." Only if done before cracking starts and while hydrogen is still mobile. Once a crack has initiated, baking can't reverse it — and waiting too long after plating (past the 4-hour window) lets hydrogen reach traps you can no longer empty.
• "The proof test passed, so it's safe." The most dangerous myth. A hydrogen-charged part can survive an immediate overload test and then fail under a lower sustained load days later, because embrittlement is time-dependent. This is why sustained-load testing exists.
• "More plating thickness gives more protection." Zinc and cadmium coatings act as hydrogen barriers — thicker coatings trap hydrogen in and force longer bakes. The coating that protects against corrosion can worsen embrittlement if the bake isn't matched to it.
• "Mild steel is at risk too." Below ~1000 MPa, steel yields and blunts cracks faster than hydrogen can embrittle them. The threat is specific to high-strength, high-hardness microstructures.
• "Stainless is immune." Austenitic 304/316 is highly resistant, but martensitic and PH stainless (410, 17-4 PH, 440C) at peak hardness are very much susceptible.