Early Stages of Self-Healing at Tungsten Grain Boundaries from Ab Initio Machine Learning Simulations
Jorge Suárez-Recio, Pablo M. Piaggi, Francisco J. Domínguez-Gutiérrez, Raquel Gonzalez-Arrabal, Roberto Iglesias · 2025
Read Early Stages of Self-Healing at Tungsten Grain Boundaries from Ab Initio Machine Learning Simulations by Jorge Suárez-Recio, Pablo M. Piaggi, Francisco J. Domínguez-Gutiérrez, Raquel Gonzalez-Arrabal, Roberto Iglesias (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-13744.txt.
Introduction
Jorge Suárez-Recio and co-authors simulate how tungsten heals its own radiation damage at grain boundaries. Their 2025 study deals with a metal eyed for use facing hot plasma in fusion devices. Fine grained tungsten may heal because its many boundaries soak up defects. The team ran dynamics simulations with a machine learned force model. The model was trained to match high accuracy quantum calculations for a common boundary type.
The new model beat older ones at predicting defect energies. Simulations showed defects migrating to boundaries at low temperature. They ordered into dumbbell shapes, then settled along boundary grooves. Older models wrongly showed the boundary breaking down at high heat; this one stayed stable. A fitted migration energy of 0.048 electron volts matched experiment. The work supports machine learned models for radiation tolerant materials.
Catalog record
| CREATOR | Jorge Suárez-Recio, Pablo M. Piaggi, Francisco J. Domínguez-Gutiérrez, Raquel Gonzalez-Arrabal, Roberto Iglesias |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-13744
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Early Stages of Self-Healing at Tungsten Grain Boundaries from Ab Initio Machine Learning Simulations Jorge Suárez-Recio, Pablo M. Piaggi, Francisco J. Domínguez-Gutiérrez, Raquel Gonzalez-Arrabal, Roberto Iglesias Abstract Nanostructured tungsten has been reported as a possible alternative plasma-facing material due to its potential ability to self-heal radiation-induced defects, a property that is attributed to its high density of grain boundaries (GB). Here, we study the initial stages of self-healing at tungsten interfaces with molecular dynamics simulations driven by a machine-learning interatomic potential tailored to one of the most common GBs found in experiments. Our model accurately reproduces the ab initio potential energy surface derived from density functional theory (DFT) calculations and outperforms previously reported empirical and machine learning interatomic potentials in predicting defect energetics. The simulations reveal low-temperature defect migration to GBs driven by rapid dumbbell-like ordering and subsequent accommodation along GB grooves. In contrast to empirical potentials, which predict unexpected GB degradation at high temperatures after defect migration, our model maintains stable GB motifs over the investigated temperature range. The temperature-dependent defect counts, evaluated using an Arrhenius-like fit, yield an average interstitial migration energy of 0.048 eV, in agreement with experiment. This work underscores the capabilities of ab initio machine learning simulations in accurately modeling defect-GB interactions and highlights their potential to contribute to the development of radiation tolerant materials. A Preprint Keywords tungsten \cdot self healing \cdot molecular dynamics \cdot machine learning interatomic potential \cdot nanostructured materials 1 Introduction Decarbonizing energy systems has become a defining challenge of the 21st century. Although fission reactors currently supply about 10% of the world’s electricity [ 1 ] and continue to evolve with advanced designs, fusion energy is poised to address the interconnected crises of climate change, energy security, and resource competition. Its high power density, lack of long-lived radioactive waste, and inherent safety (fusion reactions stop naturally under abnormal conditions, eliminating the risk of reactor runaway) make fusion a promising route to next-generation power plants. Realizing this promise on a commercial scale, however, requires the development of robust plasma-facing materials capable of withstanding intense radiation and extreme thermal loads, hence acting as the reactor’s first line of defense. Ideally, they should exhibit healing properties that allow radiation-induced defects to be eliminated, thereby preserving structural and functional integrity. Progress toward such materials requires not only breakthroughs in fabrication and engineering, but also the development of computational methods to predict their behavior under high temperature and high-flux conditions, harsh environments that are still difficult to fully replicate in existing experimental facilities. Irradiation of coarse-grained materials by particles (e.g., ions, neutrons) with energies exceeding the displacement threshold, as occurs in the inertial nuclear reactors operated in the direct drive approach, produces atomic displacements that generate vacancies and self-interstitial atoms (SIAs). At a given temperature, both types of defects migrate towards defect sinks such as grain boundaries (GBs), dislocations, and voids. However, in general, they do so at quite different energies and time scales. These differences mean that they usually do not meet and recombine, leading to a permanent degradation of material properties such as electrical conductivity and mechanical response [ 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 ] . From an atomistic point of view, the situation can be different if vacancies and SIAs are forced to be very close to each other, which can be done by reducing the grain size (nanostructuring) or, in other words, by introducing a high density of GBs. In this case, the recombination probability between vacancies and SIAs increases, leading, under certain conditions, to their mutual annihilation at the surface or GBs, a phenomenon commonly referred to as self-healing, so that the material spontaneously returns to its unirradiated structure. Nevertheless, as previously mentioned, this self-healing behavior only occurs under specific conditions, well captured phenomenologically by the three-temperature (low, medium, and high) model proposed by Beyerlein et al. [ 13 ] . In the low temperature regime, vacancies are immobile, so any damage inflicted on the nanostructured material is greater than on its coarse-grained counterpart [ 14 , 15 ] . This is because highly mobile SIAs migrate along the bulk toward the GBs, leaving immobile vacancies in the grain interior. At intermediate temperatures, where vacancies are still immobile, some of the radiation-induced damage can be annihilated by the re-emission of SIAs from the GBs [ 15 , 16 , 17 ] . Finally, at high temperatures, vacancies become mobile and move toward GBs, where they effectively recombine with SIAs, annihilating the radiation-induced damage. The evolution of defects and the temperature threshold for moving from one temperature regime to another depend on the material composition and microstructure. Even though this model describes the origin of the self-healing behavior, its effectiveness depends strongly on the GB architecture [ 18 ] . Therefore, a proper description of the self-healing capabilities of nanostructured materials requires an understanding of how defects migrate and interact with the specific GB at the atomic scale, which can be achieved by performing computer simulations. Traditionally, two main approaches have been used: ab initio molecular dynamics (AIMD) — typically based on Density Functional Theory (DFT) to compute interatomic forces — and classical Molecular Dynamics (MD) using empirical or semi-empirical interatomic potentials (hereafter IPs). While AIMD offers high precision in capturing atomic interactions, it is constrained by limited time and length scales [ 19 , 20 ] . Conversely, classical MD allows the simulation of larger systems over longer timescales, but relies on IPs that often have limited accuracy and transferability [ 21 , 22 ] . As a result, an IP developed for coarse-grained materials may not be sufficiently accurate to adequately describe radiation-induced damage in nanostructured materials. Indeed, this fact could limit self-healing studies, where the simulation of defect migration requires an extremely precise description of GB energetics as well as relatively large simulation boxes and long simulation times. Recent years have witnessed the rapid development of machine learning interatomic potentials (MLIPs) that leverage large datasets, often derived from quantum-mechanical calculations, to build flexible, highly accurate models of atomic interactions [ 23 , 24 , 25 ] . These advanced methods have already been applied to challenging problems such as radiation damage cascades, surface reactions, and phase transformations under extreme thermomechanical conditions [ 26 , 27 , 28 , 29 ] . Critically, they can achieve near- ab initio accuracy at a fraction of the computational cost, thus enabling simulations on time and length scales relevant to capturing the early stages of self-healing at tungsten interfaces. In this work, we present a computational study of how intrinsic defects migrate and interact with a tungsten GB, with an eye towards having a deeper understanding of the self-healing in nanostructured tungsten, one of the candidates proposed as a plasma-facing material in nuclear fusion reactors [ 30 , 31 , 32 , 33 , 34 ] . We focus on atomistic damage since the thermomechanical effects that PFMs will undergo and that could lead to significant temperature increases and, consequently, grain growth (if this is not somehow stabilized), could be alleviated by using a/or by combining different strategies: lower target yield, employing radiation mitigation strategies, and