El investigador del Grupo de Fusión del BSC, el Dr. Julio Gutiérrez, presentó los progresos recientes en el trabajo del proyecto FusionCAT sobre «Large-scale ab-initio study of tungsten metal from linear-scaling density functional theory methods» en el Simposio de Computación en Química (COMP) de la División de American Chemical Society ‘s (ACS) sobre Ciencia de materiales centrado en el desarrollo de métodos / aprendizaje automático / propiedades de materiales (Papel ID 3529923).
Ab initio density functional theory (DFT) is probably the most widely-used method for electronic structure calculations and has been key to elucidate the characteristics of the smallest defects in a wide range of materials. To simulate large defects in structures accurately and reproduce typical experimental defect concentrations, the unit cell should be large enough to correct the periodic interactions between repeating point defects. Unfortunately, DFT‘s intrinsic cubic scale limits accessible systems to a few hundreds of atoms. In this work, we show that DFT calculations for large systems are possible with the so-called linear scaling (LS) algorithms, meaning that doubling the number of atoms in a system leads to a computation time only twice as large.
Looking into specific applications, one of the big challenges towards the safe use of nuclear fusion is the design of appropriate reactor components. Tungsten (W) is one of the best plasma-facing candidate materials for fusion; however, defects produced by radiation may have a detrimental effect on its structure and mechanical behaviour. Understanding the atomic-level structure and the electronic features of the defective material and their connection with the mechanical properties is fundamental to build predictive models of microstructure evolution under irradiation.
Our calculations have demonstrated that the LS version of BigDFT can handle bulk metallic W structures with up to 3456 atoms, surface models, grain boundaries, and reproduce defective systems with realistic vacancy concentrations, yielding results the same quality as those from traditional DFT. These calculations show the possibility of addressing the challenge of simulating realistic large-scale defective metallic systems with DFT calculations, which will be used to develop appropriate materials for fusion applications.