Defects and Impurity Properties of VN precipitates in ARAFM Steels: Modelling using a Universal Machine Learning Potential and Experimental Validation

TL;DR

VN precipitates in ARAFM steels are challenged by dissolution under Fe irradiation; this work combines TEM, APT, DFT, and a finetuned universal neuroevolution potential (NEP) to model point defects and solute substitutions in VN and to simulate irradiation damage. The approach enables large-scale Monte Carlo and collision cascade simulations that reveal vacancy ordering at operating temperatures and how solutes disrupt order, potentially accelerating dissolution. The study demonstrates the importance of local reference states for defect stability and provides insights into how and when VN precipitates may resist irradiation damage, informing materials design for fusion reactors.

Abstract

VN precipitates used to strengthen ARAFM steels for fusion applications dissolve under high Fe ion irradiation (100 dpa at 10^-3 dpa s^-1, 600 C). This study examined point defects and solute substitutions using atom probe tomography, machine learning interatomic potentials, and density functional theory. Combined with transmission electron microscopy, results show N-vacancies and substitutional Cr exist in VN precipitates before irradiation. Monte Carlo simulations and collision cascade simulations confirm ordered vacancies at operating temperatures help mitigate irradiation damage. However, solute additions disrupt vacancy ordering and enhance irradiation-induced damage, potentially accelerating dissolution.

Figures (9)

• Figure 1: The lattice parameters calculated using VASP for defects in rocksalt VN as a function of concentration are compared to the experimental value (unknown stoichiometry) from TEM highlighted with a horizontal line. The effect of vacancies, interstitial, and antisites are compared in a). The effect of substitutional C, Fe, and Cr on lattice parameter as calculated by DFT are plotted in b) and compared to an experimental value from a single measurement at an unknown stoichiometry (the dashed horizontal line).
• Figure 2: The convex hull calculated with 6519 enumerated structures calculated with the finetuned NEP potential. The red line indicates the convex hull with molecular N and BCC V as reference states, while the blue line is the convex hull where V and N as dilute solutes in Fe are used as reference states.
• Figure 3: a) The concentration profiles of decomposed V, N, and Cr through the interface of a 2.2% isoconcentration surface of an APT reconstruction averaged over 3 precipitates. b) The concentration profiles of decomposed C, Si, W and Mn through the interface of a 2.2% isoconcentration surface of an APT reconstruction. For both plots, the distance is the distance from the interface of the isosurface, with positive values inside the precipitate and 0.0 nm being the interface marked by a vertical dashed line.
• Figure 4: The ordering present taken from the final states from on-lattice Monte Carlo simulations. a) shows the SRO parameters as a function of initial vacancies (X) on the N sublattice at a temperature of 933 K. b) shows the SRO parameters as a function of temperature at a vacancy concentration of 25%. c) an illustration of nearest neighbour shells of N atoms in a slice in the [111] plane viewed down the [111] direction. d) shows the SRO parameters as a nearest neighbour shells including solutes at a temperature of 933 K.
• Figure 5: The on-lattice Monte Carlo atomic configurations taken from the final state of a 5% vacancy concentration at 933 K with and without solutes. a) and c) show the vacancies in $\mathrm{VN}_{0.95}$ condition coloured by species, a), or by local vacancy concentration (within 3 Å). b) and d) show the vacancies in $\mathrm{VN}_{0.95}$ containing solutes condition coloured by species, a), or by local vacancy concentration (within 3 Å).