The effect of the deformation-induced martensite (α') with varying carbon content on the toughness was investigated in metastable γ-phase weld metals, which were Fe–0.05C–27Ni–3Cr and Fe–0.44C–25Ni–3Cr. The toughness was measured using a Charpy impact test and a three-point bending test. The microstructures were characterized by means of scanning electron microscope. The correlations between α' phase, crack propagation and toughness studied using electron backscatter diffraction. The crack initiation was suppressed by the formation of α' phase, and the toughness improved at −100°C in Fe–0.05C–27Ni–3Cr weld metal. The toughness was reduced by the brittle fracture of the hard α' phase in the Fe–0.44C–25Ni–3Cr weld metal. It is clear that the formation of the α' phase particles with high toughness is important to improve toughness by transformation induced plasticity in the weld metals.

In the Fine Grain Heat Affected Zone (FGHAZ) of Mod. 9Cr–1Mo steel, the shape of grain boundaries is complex. It becomes simpler with creep and grain growth. The complexity of the grain boundaries is expressed with fractal dimension. Prior research found that the fractal dimension of FGHAZ decreased with creep damage, but then saturated or slightly increased, which might imply dynamic recrystallization.

The objectives of this research are to clarify the cause of the phenomenon and to confirm that the fractal dimension of grain boundaries can be used as the indicator of creep damage and dynamic recrystallization.

In this research, creep interruption tests were conducted with the specimens of simulated FGHAZ of Mod. 9Cr–1Mo steel. Two types of specimens with different stress multiaxiality and the same maximum principal stress were used. After the tests, the microstructures were observed by EBSD to measure the fractal dimension of grain boundaries, GOS average, the area ratio of recrystallized grain, KAM average, and the average grain size. Vickers hardness was also measured.

In the specimen of the higher stress multiaxiality, the average grain size and the area ratio of recrystallized grain increased and other indicators decreased with creep. However, in the specimen of the lower stress multiaxiality, these trends were different at mid-life. They were considered to represent dynamic recrystallization.

It was concluded that the fractal dimension of grain boundaries can be used as one of indicators of creep damage and dynamic recrystallization and that the dynamic recrystallization was affected by stress multiaxiality.

The effect of hydrogen concentration on the hydrogen diffusion coefficient was investigated by analyzing the relationship between hydrogen permeation current density and hydrogen diffusion coefficient. The hydrogen diffusion coefficients were obtained using the half-rise time method, which analyzes the build-up behavior of the hydrogen permeation current densities, and were correlated with the steady-state values of the hydrogen permeation coefficients. The hydrogen diffusion coefficient decreased as the hydrogen permeation coefficient decreased, indicating that the diffusion coefficient depends on the hydrogen concentration. Subsequently, a hydrogen diffusion coefficient model was developed based on the Fermi–Dirac distribution to express this concentration dependence. The experimental relationship between the hydrogen permeation and hydrogen diffusion coefficients was fitted using this model. Assuming the presence of more than two hydrogen trap sites was necessary to achieve an accurate fit. Therefore, two types of trap sites were assumed in this study. The hydrogen trap energies at these trap sites were determined using the Choo–Lee method under the same hydrogen charging conditions as those used in the hydrogen permeation tests. The hydrogen occupancy at each trap site was treated as a fitting parameter in this model. The estimated values from the model agreed well with the experimental data. Furthermore, validation under varying temperature conditions exhibited a similar good agreement between the experimental data and the model predictions.