Team up and get creative!

Submitted by admin on Thu, 08/15/2019 - 14:01

Collaboration on-site brings added value

Volker Deringer from the University of Cambridge filed a transnational access application in the very first call of HPC-Europa3 to visit Miguel Caro in Aalto University, Finland. Dr. Caro leads the computational simulation efforts in Prof. Tomi Laurila’s group at Aalto. The group’s work focuses on studying the properties of amorphous carbon, among other research topics. Together, the knowledge of the materials and access to the new methods could be used to solve the long lasting debate between competing growth mechanisms of amorphous carbon films and to gain new understanding on the high-tech material that has enormous industrial and scientific importance.

To understand the properties of disordered materials, including the amorphous phases that are made in Laurila’s lab, powerful computational tools are needed. They need to combine accuracy (to describe the diverse atomic structures) and speed (to achieve realistic simulation system sizes). The researchers use a novel simulation technique based on machine learning (ML) from quantum-mechanical data, called “Gaussian Approximation Potentials”, or “GAP” in short. The GAP method is developed by Prof. Gábor Csányi and his group at Cambridge, with whom the team collaborate extensively. As a Leverhulme Early Career Fellow in the same department, Dr. Deringer aims to apply this methodology to study and understand a wide range of amorphous solids.

And this mix of novel simulation tools and intriguing applications got Deringer and Caro talking. During an initial visit in December 2017, they combined their knowledge, built the computational tools and designed computational experiments to simulate the amorphous carbon film growth process. The results unambiguously confirmed the less widely supported "peening" mechanism for the growth of high sp3-hybridized carbon local structure, which is responsible for the mechanically superior properties of the carbon films. The article describing this work was published in Physical Review Letters already in April and was promptly selected by the American Physical Society as their feature article of the week. Another article also based on the first visit's work is coming soon.

"Nothing beats working with colleagues on site!", tweeted Dr. Deringer confirming that in challenging projects with collaborators with complementing skills there really is added value to work together at the same location. Indeed, this idea is at the very core of the programme. It turned out that the collaboration and visit were so successful that Dr. Deringer applied for another HPC-Europa3 visit and is currently at Aalto university. The collaborators are now extending the computational method to make new materials amenable for atomic level simulations. The current version of the GAP can be used to simulate materials composed only of carbon, and the team are now aiming to expand their approach to other chemical species – getting even closer to the real materials that are made in the laboratory.

After a visit, the parties are asked for feedback. This is one of the things Dr. Deringer wanted to say: "This was a most successful visit - we have developed an initial collaboration, which led to a whole range of new ideas. We are planning and working on several publications, and hope to apply for HPC-Europa3 support again in one of the next rounds. " Dr. Caro highlights another benefit of the visit in his feedback: "The collaboration with Volker is invaluable to us. It expands our expertise in electronic structure methods by adding machine learning potentials. "  The new visit, new ideas and new projects give compelling evidence of new fruitful collaboration - supported by HPC-Europa3.

Read more: