Microstructure Effects on Hydrogen Embrittlement in Austenitic Steels: A Multidisciplinary Investigation

S. Evers,¹ T. Hickel,² M. Koyama,³ R. Nazarov,² M. Rohwerder,¹ J. Neugebauer,² D. Raabe,³ M. Stratmann ¹

1  Department of Interface Chemistry and Surface Science

2  Department of Computational Materials Design

3  Department of Microstructure Physics and Alloy Design

Hydrogen atoms, which can be absorbed into steel during production and service, often have a detrimental embrittling effect on the mechanical properties of iron and steels. It is meanwhile known that hydrogen embrittlement (HE) is also affected by the microstructure of the material. Consequently, previous indications that hydrogen atoms are trapped by vacancies, dislocations, and grain boundaries led at MPIE to investigations of superabundant vacancy formation, hydrogen-enhanced local plasticity (HELP), and hydrogen-enhanced decohesion (HEDE). Despite these efforts, any proof of a HE mechanism to be active in a given steel sample has so far been a formidable task, which cannot be achieved by a single method. A direct experimental observation of hydrogen impurities is difficult due to the low solubility and high mobility of hydrogen in steels, whereas pure theoretical investigations are challenged by the complexity and diversity of microstructural features present in steels.

We therefore follow a multidisciplinary strategy to derive a deeper understanding of HE in steels.
This strategy combines novel potentiometric methods based on the Kelvin probe technique to detect the local hydrogen content in materials (GO department), ab initio determination of the same quantities including the local behaviour at grain boundaries (CM department), and characterization of hydrogen induced materials failure (MA department) using orientation-optimized electron channelling contrast imaging (ECCI). Selected findings of these investigations and their relevance for austenitic steels are summarized in the following:

The crucial idea for the new hydrogen detection method is the observation that hydrogen dissolved in a palladium matrix leads to the formation of a hydrogen electrode on the palladium surface, even in dry atmospheres. The origin is the presence of a nanoscopic water layer adsorbed on the surface, enabling the formation of a corresponding electrochemical double layer [1, 2]. As the electrode potential for the hydrogen electrode depends logarithmically on the activity of H in Pd, this potentiometric method is extremely sensitive especially at low activities.

The idea can be employed for the investigation of steels (and various other materials) by evaporating a thin film of Pd on their surface. Since the chemical potential of H in iron-based materials is much higher than in Pd, H diffuses into the Pd film. Time dependent measurements of this accumulation can be used to perform extremely sensitive and laterally resolved measurements of H permeation through and its presence in materials. In the latter case an effective “activity” of H is measured, providing information about depth and density of traps sites. Main challenges of this method are the need for an exact calibration of the potential-concentration correlation for H in the evaporated Pd films, the precise calibration of the Kelvin probe tip in the dry nitrogen measurement atmosphere, as well as its long term stability.

As an example the measurement of H in a H-charged austenitic steel sample, comprised of mainly austenitic and ferritic grains, is shown in Fig. 1. It can be seen that the austenite contains much more H, as the potential decreases much faster over the austenite grains. Especially active sites are located at boundaries between ferrite and austenite.

An additional insight into the relevance of the different phases and their boundaries has been obtained by ab initio calculations based on density functional theory (DFT). They clearly confirm the increased solubility of H in austenite grains as compared to the ferrite grains. Mn yields further increase of the austenite solubility by straining the lattice (volume effect). One of the new insights obtained by the calculations is that small amounts of further alloying elements (like Ca, Nb, Si, Ti, and in particular Mo) considerably enhance the preference of H for austenite [3].

In order to understand the experimental results on microstructures, we have additionally used DFT to study the solubility and diffusion of hydrogen in austenite twin and grain boundaries [4]. We generally find that the solution energy of H strongly depends on the local coordination and that it is in this case only moderately correlated with the actual volume of the interstitial site. Within open structures, such as the Σ₁₁[1-10](113) fcc grain boundary, various different interstitial sites are favorable for the incorporation of H atoms, providing effective trapping centres (Fig. 2). Only if these traps are filled by other H atoms, efficient diffusion channels along (113) planes might become active. We further find that the critical strain required to fracture the material is reduced by the presence of hydrogen in this grain boundary. For twin boundaries, the DFT calculations show that interstitial H atoms are actually slightly repelled. As origin for this unusual and unexpected behaviour the structural similarity between the octahedral interstial configurations in the twins and in austenitic bulk has been identified.

These theoretical insights are highly relevant for experiments, which investigate the fracture mode in austenitic steels. For this purpose the recently developed orientation-optimized ECCI method has proven to be particularly useful to reveal deformation twins and complex dislocation substructures in TWIP steel. The actual measurements have been performed for a H charged Fe–18Mn–1.2C austenitic steel [5], for which the tensile ductility was drastically reduced by H charging during tensile testing. The central region of these samples, which have not been reached by hydrogen, showed a ductile fracture surface. In contrast, a brittle fracture surface was observed from the surface down to about 150 µm. The facet size of the brittle fracture areas is about 50 µm, which corresponds to the grain size, indicating that intergranular fracture was caused by H charging.

An advantage of the employed ECCI method is that in addition to the cracks primary and secondary deformation twins on (11-1) and on (1-11) planes become visible with bright contrast (Fig. 3). The measurements therefore revealed that cracks typically occur at grain boundaries with intercepting primary deformation twins. The stress concentration at these points and the reduction of the cohesive energy by hydrogen loading apparently yields crack initiation. While the primary fracture mode is intergranular, one additionally observes crack propagation following primary and secondary deformation twin boundaries (Fig. 3). Since the ab initio calculations predict that perfect twin boundaries are not sensitive to H, the stresses due to the interception of twins with grain boundaries or of primary with secondary twins need to be responsible for such a transgranular fracture along twin boundaries. Being crucially important, because deformation  twinning is essentially required to achieve the superior mechanical properties of TWIP steels, further investigations of this effect are currently performed.