Professor Stephan Roche of the Catalan Institute of Nanoscience and Nanotechnology in Spain is currently working on his project 'longitudinal and transverse electronic transport in atomically doped graphene from first principles', for which he has received 22 million core hours from PRACE, the Partnership for Advanced Computing in Europe. He is one of the early adopters of high-performance computing (HPC) in the Graphene Flagship initiative. Here, we speak to him about his experience of working with PRACE.
Roche has fifteen years of experience in developing novel algorithms and computational approaches that can be used to investigate quantum transporting in complex materials such as graphene. His career in the field began at the French Atomic Energy and Alternative Energies Commission, where he was in charge of the computational platform and also helped to set up their global research program for computation in nanoscience. "I have a varied background," explains Roche. "There is my work in microelectronics and my personal development of a few methodologies, but I am also aware of the needs of large European organizations, and how computation and collaboration with industry can help to bring added value to an organization."
After arriving in Barcelona, Roche began developing a new algorithm that could be used to investigate the quantum Hall effect, a phenomenon that is very important for the standardization of electrical resistance in materials due to its property of being extremely precise. "Before our work there was no existing methodology for simulating quantum conductivity in very complex materials. We developed a new and very efficient approach, but even though we had made this breakthrough in terms of the algorithm, we still required access to large computational resources. Being in Barcelona, I was quickly put in touch with some people working at the Barcelona Supercomputing Centre (BSC), one of the four hosting members of PRACE."
Spintronics, also known as spinelectronics or fluxtronics, is an emerging technology exploiting both the intrinsic spin of the electron and its associated magnetic moment, in addition to its fundamental electronic charge, in solid-state devices.
After initially applying unsuccessfully for a PRACE preparatory type B project, Roche consulted with colleagues at the BSC who were able to help out. "You have to be able to demonstrate that your code is transferrable," he explains. "If you get given access to computational resources, you have to make sure your code is fully operational for use in the PRACE infrastructure. That is why the preparatory phase is so important, and with some help from PRACE we were able to make it."
The project was focused on the development of strongly efficient linear scaling methods to compute the Hall conductivity of two-dimensional disordered materials, such as graphene. The computational strategy was based on a real space implementation of transport coefficients, and on the Lanczos approach to compute off-diagonal elements of the Green's function matrix. This enabled simulations of disordered materials with several tens of millions of atoms in geometries of arbitrary complexity, thus reaching a quantitative and predictive power for comparison with experimental data achieved in real materials. With the size and complexity of these systems, the use of supercomputers like the ones proposed by PRACE is necessary in order to stay at the forefront of worldwide scientific competition.
The project entitled 'linear scaling methods for quantum Hall transport simulations' was allocated 200,000 core hours on Curie at GENCI@CEA, France, and 50,000 core hours on Hermit at the High Performance Computing Center Stuttgart (part of the Gauss Centre for Supercomputing), Germany, under the ninth cut-off for PRACE preparatory access (type C).
The project entitled 'longitudinal and transverse electronic transport in atomically doped graphene from first principles' was allocated 14.3 million core hours on Curie under the sixth call for proposals for PRACE project access and 22.1 million core hours under the eighth call.
Roche was pleasantly surprised with the help and support provided by PRACE during the process of developing his code. "We had a Greek engineer who worked alongside us, helping us to take our code and find out ways to maximize its potential in terms of several technical issues," he says. "With this help, we were then able to get accepted for a larger project using PRACE regular calls, which have been renewed this year."
All in all, Roche has been impressed by the professionalism and ease of communication that has come from working with PRACE. "I have been in contact with people from Belgium, France, and Spain who have all been very helpful with their support," he says. He is now working as leader of the Graphene Flagship spintronics work package, with the goal of establishing the ultimate potential of graphene for spintronics, targeting efficient room-temperature spin injection and detection - but also spin gating and spin manipulation - in graphene spintronic devices. Spin transport mechanisms in graphene devices will be explored using different materials, and by comparing the results with realistic theoretical simulations, the group is working towards the demonstration of novel types of functional graphene spintronic devices.
PRACE plans to increase its relations with the graphene research community (it has already supported five projects in this area with over 90 million core hours), and it is now looking to focus on the Graphene Flagship research initiative. The Graphene Flagship is the EU's biggest ever research initiative, tasked with taking graphene from the realm of academic laboratories into European society in the space of ten years. Roche thinks this is an important step forward: "This kind of interaction can bring real value to Europe in the medium term. For instance, in the Graphene Flagship community, we are always trying to network with the best groups and push forward technology transfer and support of our industries."