The structural integrity of steel can be compromised by hydrogen atoms that are able to diffuse through the lattice. The steel then snaps suddenly, while it is operating well within its supposed elastic limits. There exist explanations.
These theories are assessed here in the context of Karl Popper's criterion of falsifiability, where philosophical failure is avoided by testing whether the science can make predictions that risk being disproven by a counter-example.
H. K. D. H. Bhadeshia, Advanced Engineering Materials (2026) e202503067.
Audio podcast
Study guide
More 
There clearly is a dire need to formulate quantitative models that enable engineering design as a function of the specific levels of hydrogen concentration expected to enter steel in the circumstances of application. The obvious way forward is to focus on fracture toughness determinations ($K_{\text{IC}}$) as a function of $[\text{H}]$, and express the data empirically in terms of the effective cleavage surface energy. This is analogous exactly to the current practice for plane-strain conditions where:
where $\gamma_{\text{p}}$ is an effective surface energy that accounts for some level of localised plasticity, $E$ is the Young's modulus, and $\nu$ is the Poisson's ratio. The $K_{\text{IC}}$ parameter can then be used directly in engineering design so in that sense represent fundamental material properties.
The true surface energy $\gamma$ is the purely thermodynamic energy required to break the atomic bonds across a plane to create two new, stress-free surfaces. For iron this would be between $2.3\text{--}2.5\,\text{J m}^{-2}$ for $\{100\}_{\alpha}$, $\{110\}_{\alpha}$, and $\{111\}_{\alpha}$ surfaces [Spencer, 2002]. Failure in this way would represent a process in which no energy is dissipated through secondary mechanisms. The effective surface energy $\gamma_{\text{p}}$ accounts for all the energy dissipated from the crack tip per unit area of crack advance, primarily the localised plastic work per unit area, $w_{\text{p}}$, associated with a finite plastic zone at the crack tip, $\gamma_{\text{p}}=\gamma+w_{\text{p}}$, that can be much larger. And yet, the effective surface energy is able to function semi-empirically as a robust parameter in safe design against rapid fracture when hydrogen is not present in the steel. There is no reason why such an approach cannot be exploited for hydrogenated steel.
This approach clearly relies on hydrogen affecting the fracture-surface energy so has similarities to the original thinking on decohesion, Section decohesion. The caveat is that the $K_{\text{IC}}$ tests can be rigorous representations of potential reductions in toughness due to hydrogen. They can be used directly in engineering design, and unlike most assessments based on simple tensile tests, accurately represent unstable crack propagation. The function $K_{\text{IC}}=f\{[\text{H}]\}$ can in principle even be applied to estimate the safe-life of a component subjected to fatigue crack growth.
Once the dependence $K_{\text{IC}}=f\{[\text{H}]\}$ is established, the quantitative variation in $\gamma_{\text{p}}$ with hydrogen can potentially provide imaginative, though hopefully falsifiable fodder for further mechanistic investigations. If this hypothesis is accepted, then the following testing methods become less relevant:
- Tensile tests: These are associated with benign embrittlement accompanied by substantial bulk plasticity. Such tests do not represent phenomena such as the catastrophic, unpredictable fracture of pressure vessels.
- Routine experimentation: The standard interpretation of rudimentary embrittlement experiments with respect to unfalsifiable, descriptive hydrogen embrittlement models lacks predictive validity.
One aspect not covered here due to time limitations, is the widespread application of Sieverts' law to estimate the atomic hydrogen concentration within a steel due to pressure from molecular hydrogen gas. This is particularly urgent in the context of the imminent use of existing pipelines, originally designed for fossil-gas, to transmit hydrogen. Sieverts' law [Sieverts, 1911]; [Sieverts, 1929] is assumed without adequate attention to the physical necessity for hydrogen dissociation at the steel surface. Johnson pointed out in 1988 that there are substantial surface impedances, such as native oxides, which hinder hydrogen entry and exit from the steel at ambient temperatures [Johnson, 1988]. A palladium sputter-coating leads to true Sieverts' law behaviour, whereas the square root dependence on gas pressure is highly unlikely to be justified otherwise.
