About the project
Through QCIF, a Griffith University renewable energy researcher has taken his work to the next level with all-important access to national high-performance computers.
Before contacting QCIF, Dr Yun Wang of Griffith’s Centre for Clean Environment and Energy was facing the very real possibility of having to stop computational research due to a lack of HPC resources. This would have been a major setback as supercomputers are essential to his research, using large-scale computations in order to understand the properties of complicated materials at the atomic level.
Thankfully, friends at the University of Queensland recommended Yun contact QCIF for help.
Back in 2013, QCIF helped his research team, Yun’s Group, gain more time on the National Computational Infrastructure’s (NCI) supercomputer via QCIF’s share, and Yun has been using this ever since.
QCIF’s share of the Canberra-based NCI’s HPC is available for access to all QCIF members, and accepts applications for its use all year round. This is especially helpful for researchers who either missed NCI’s annual call for applications (usually in October) through its National Computational Merit Allocation Scheme (NCMAS) for time on its HPC, or whose application was unsuccessful in the competitive process, or who, like Yun’s Group, have exhausted their NCMAS allocation.
Yun has had an NCMAS grant since 2010, and has received top-up time from QCIF’s NCI share since 2013, if his project runs out of its time allocation. This usually occurs a few times each year, sometimes more, and enables his research to continue at pace.
On NCI, allocations are fixed, and in most cases, researchers cannot run anymore jobs once they are over the limit. However, QCIF can give additional time out of its share for researchers to continue to run their jobs.
Yun also inquired about other HPC options, and one of QCIF’s UQ-based eResearch Analysts, Dr Marlies Hankel (a HPC expert), suggested he apply to the Perth-based Pawsey Supercomputing Centre to use its HPC, Magnus, and he was successful there too.
Yun’s Group had three high-impact publications in Nature Publishing Group journals in 2016 from the work using NCI and Pawsey HPCs (published 18 January, 5 October and 28 November), and four papers published in leading journals in 2018—three in Advanced Materials, the flagship Materials Science journal (published 28 May, 18 July and 28 November) and one in JACS, the flagship journal in Chemistry.
“Only HPCs can enable us to do our research,” said Yun. “The larger supercomputers and clusters can greatly accelerate our research. We use NCI’s HPC every day to compute the properties of materials in my project.
“The insightful information from the computations can benefit the development of the renewable energy industry, as well as environmental monitoring and remediation.”
Marlies helps Yun’s Group “a lot”, according to Yun, and he has also recently received assistance from QCIF’s Griffith University-based eResearch Analyst, Amanda Miotto. In March, Amanda provided Yun with an overview of QCIF’s other HPC options should he need them, such as Awoonga and FlashLite for compute, and QRIScloud for data storage.
As a result of that meeting, Amanda organised a HPC workshop for Griffith’s Gold Coast Campus on Thursday, 18 April 2019, as Yun mentioned a few researchers in his group were interested in learning about supercomputing. Members of Yun’s Group also attended the Software Carpentry Python workshop that Amanda had already organised for 15–17 April 2019, at Griffith’s Nathan Campus.
Yun began his research project solo in 2010. Today, seven people are working on the Yun’s Group project at Griffith’s Gold Coast Campus.
The central focus of their research is on computer-aided materials design for next-generation energy and environmental applications. Currently, the focus is on the development of novel low temperature and environmentally-benign applications, including: clean hydrogen fuel generation via water splitting, fuel cells, selective reduction and oxidation of hydrocarbons, CO2 reduction, environmental remediation, sensors, and solar cells.
“Understanding the structure, property and performance relationship of materials holds the key to designing novel materials,” said Yun.
“The research strategy we generally utilise is to identify the correlation between materials structure and their properties at the atomic level by incorporating the fundamental knowledge and data into first-principles electronic structure calculations. Thanks to the recent progress in theory and supercomputer areas, we can develop approximate solutions to the exact geometric, electronic, magnetic and optical structure problem to be obtained with reasonable accuracy, compared to experimental data.
“We can further study a variety of important elementary reactions at solid-solid, solid-liquid or solid-gas interfaces with the consideration of the mass transport, charge transfer, phonon transport and reaction conditions by virtue of sophisticated computational algorithms.
“Consequently, the reactivity, selectivity and stability of materials can be systematically and comprehensively evaluated through the calculation of the corresponding thermodynamics and kinetics that are not accessible to experiments.
“We can then design materials by tailoring their properties to improve their practical performance for specific applications in strong synergetic interactions with experiments. More importantly, our outcomes can provide theoretic foundations to guide new experimental efforts towards unexplored territory.”
Yun plans to continue focusing on materials design for clean energy with the assistance of big data science and making the most of HPC resources and other technologies as they continue to improve.
NCI and Pawsey are supporting the computational aspects of the Yun’s Group project with the assistance of QCIF.
Dr Yun Wang
Centre for Clean Environment and Energy
- NCI’s HPC
- About 50 cores to do parallel computation for each task
- About 200 GBs to store the computational results.
- Pawsey’s Magnus HPC
- 56 cores to do parallel computation for each task
- 10 TBs to store the computational results.