Review of 'Mechanisms of strength and hardening in austenitic stainless 310S steel: Nanoindentation experiments and multiscale modeling'

Reviewed for PREreview  (https://doi.org/10.5281/zenodo.7371559) 

Overview

The paper presents interesting experimental and molecular dynamics (MD) simulation studies on nanoindention of a face centred cubic Fe-Ni-Cr stainless steel 310S. 310S is high in both Cr (~25wt%) and Ni (~20wt%) compared to other austenitic stainless-steel grades and finds application in areas where high temperature environmental degradation is a concern. A strength of the work is that the high quality of both the nanoindentation experiments and the MD simulations.

The nanoindentation experiments were undertaken with a Berkovich tip with loads in the range 0.25 to 10 mN, so that shallow indents below 200 nm depth resulted. Data for even the smallest <50nm, 0.25nM) indents appeared to be of good quality indicating careful experimentation on difficult measurements.  Results from repeat tests are given to indicate noise levels and scatter in material response.

The MD simulations of the 310S alloy were undertaken using LAMMPS with an embedded-atom method (EAM) potential and models with Fe, Ni and Cr atoms in an initially randomised substitutional solid solution. In line with other literature a fixed layer furthest from the indented surface, and a thermostatic layer allowing for heat dissipation were included with the model which was initially equilibrated at 300 K.  Given that repeat simulations were conducted for multiple orientations the models size was kept as large as reasonable possible and consisted of a total of 8.5-9 million atoms.  

The most significant challenge for the work is in making a strong connection between the experiments and simulations when computational resource prohibits using a larger model, while experimental uncertainties are more marked for smaller indents. The dilemma is perhaps made most evident by comparing the 10 nm tip radius and 5 nm maximum indent depth used in the simulations with the smallest experimental indents of a little under 50 nm. A more fundamental barrier to direct comparison is that the simulations are for tip radius of 10 nm, while the experiments are for much larger value (~100 nm seems likely from load-displacement data though the actual value is not quoted). The larger tip radius in the experiments provides access to much larger indentation strains than is possible in the simulations, while larger strain gradients are in place for the simulations. Finally, there is a large difference in loading rate (and therefore deformation rate).  

Reading the preprint provoked the following comments and questions some of which might be useful to the authors.

Main Points

• Given the challenges above in making direct comparisons the conclusion of excellent agreement between experiment and simulation should perhaps be softened, similarly the work itself does not deal with high temperature behaviour, or effects of irradiation so conclusions regarding the suitability of the alloy for nuclear applications seem out of scope.
• The attempt to estimate GND density is interesting, especially as the strain gradients are extremely high for the simulations, and quite a bit lower for the experiments. Typically, a length scale needs to be set in describing how the total dislocation density is split between GND and SSD densities, but this has not been made explicit here. For the MD simulations it may be that all dislocations have been taken to contribute to the GND density, but it is not clear what volume term has been used (text around eq 12-13 suggests contact diameter may be the indicative lengthscale). For the Ma-Clarke model how was the shear strain calculated (the Ma-Clarke paper was for Berkovich rather than spherical indents)? No details are given for the calculation of GND density from the EBSD map, but the characteristic lengthscale is likely markedly larger than for the MD simulation, and the effects of this should be discussed.  A little more detail in methodologies should be given for all of this analysis.  Fig 11 c) is the only figure where both simulation and experimental data are shown directly on the same plot. Ma & Clarke (and subsequently Nix & Gao) used both SSD and GND densities as contributions to a Taylor hardening expression to link strain gradients to indentation size effects. Have the authors thought of extended their analysis to see if this can consistently link the markedly lower hardness values seen for deeper experimental indents with the much higher hardness reported for MD simulations?
• Some interesting dislocation density-based laws are introduced in eq 8 to 10. These are fit to the MD simulation results for dislocation density in fig 8a. It would be good to state the dislocation mean free path and annihilation constants obtained. As with the point above can a Taylor hardening model then be used to connect to hardness values and then provide a link to the experimental data. [In passing the text describing eq 10 refers to grain size though perhaps indent size is more relevant here].
• Results from simulations and experiments show relatively little anisotropy in either indentation modulus or hardness (eg figs 4b, fig 5, fig 8b) – it seems odd then to state in the conclusions that "…310S indicates anisotropic properties…".

Minor Points

• Caption on fig 4 swaps parts (a) and (b)
• Scale bars should be added to fig 6, fig 7 a and b, fig 9, fig 11a and b, and fig 12
• Fig 9b – the 5nm and unload images seem to be identical though difference are talked about in the main text.