Multiscale Materials Modelling on High Performance Computer Architectures

In this project we will combine expertise from leading European supercomputer centres, high­profile academic groups and key industrial users to develop integrated platforms for a multi­disciplinary and multi­scale "in silico" experimentation and simulation and demonstrate the applicability for high­-profile European research and development objectives. We have united groups covering the entire range of materials­ research simulation, from the quantum to the continuum level, with the state of the art modelling techniques. With the support of the CSC and CINECA Supercomputer Centres and the Steinbuch Centre for Computing at the Karlsruhe Institute of Technology we will integrate these methods into scalable simulation protocols based on software standards developed by leading European providers of HPC resources into a powerful platform for multiscale materials modelling. We will deploy a computational science infrastructure through models, tools, algorithms, workflows and services by achieving the following goals:

Workflows and Services

We will solve the challenges of adaptable multi­scale materials simulation by linking multiscale simulation methods into workflows to provide powerful new e-infrastructures for a comprehensive description of macroscopic device function on the basis of their nanoscale building blocks. The resulting e­infrastructure will be an open toolkit based initially on the software codes developed by the modelling groups of this consortium. It will be hosted and maintained as DEISA and PRACE computer centres throughout Europe to provide solutions to materials simulation challenges. Most important, however, the [email protected] framework will be open to additional software providers at any time during this project, and after its termination it will grow into a versatile platform for materials modelling that can address a wide and extensible range of challenges. Realization of this project will provide the European research community with a leading edge software infrastructure that will exploit the highest quality computational and data resources enabling Europe to address the emerging grand challenges in science and engineering. The workflows will be realized on the basis of existing solutions, such as the UNICORE workflow system, which is implementing the workflow concept as a container of integrated services. This concept permits the development, adaptation and maintenance of scientific software on dynamically evolving hardware platforms, because it decouples the maintenance of individual modules (software packages) from the interfaces and workflows within which they are exploited. It is therefore possible to exchange one software package with another for some specific task in a complex workflow when a new computational platform becomes available. The promotion of appropriate software standards both for the workflows and the interfaces will provide scalability, evolution and interoperation in integrated platforms (see WP SA1 and JRA1).

HPC and Application Integration

We will ensure the ability to fully and timely exploit high­ performance and distributed computing capabilities by pairing modelling groups with supercomputer centres to identify the most powerful computational platform for each of the software packages maintained by the partners of this project. In work package SA3, these partners will work together to optimize and parallelize the software packages for the computational architectures that are best suited for them. All modelling groups participating in this project have demonstrated experience in the exploitation of high performance computer architectures, and all involved supercomputer centres will provide services to the modelling groups to realize their integration tasks. We will also define the interfaces that will permit these software packages to be used in the workflows developed in the course of the project, and to integrate our state ­of the art scientific application software in a European e-­infrastructure. We stress that the development of individual simulation modules is not a goal of this project.

Proof-of-Principle Demonstration in Key Applications

Demonstration of the deployment of a computational science infrastructure requires identification of a few key problems of acknowledged relevance to the European research infrastructure and European industrial R&D efforts in the materials sciences and demonstration of successful proof ­of principle simulations for such problems. With the development of e-­infrastructures, we enter a new phase in the deployment of scientific software whose acceptance will be contingent on the perception of the (industrial) users that: (a) such software satisfies pressing needs in their R&D programs, and (b) the barriers to exploit the solutions delivered by this project are sufficiently low to justify the effort.

We have therefore identified four simulation challenges of undisputed importance for European industrial R&D efforts in the materials sciences:

• Development of efficient organic light emitting diodes (WP JRA2), in close cooperation with the EU NMP project MINOTOR, including BASF AG.
• Optimization of lithium ion batteries (WP JRA5).
• Carbon based devices for electronics (WP JRA4, Nokia Research Center).
• Modelling charge transport in novel polymer­based materials from molecular electronics (WP JRA3) with SONY Europe.

For each of these areas of research, we have teamed partners in this consortium comprising modelling groups, resource providers and (most importantly) companies actively involved in industrial R&D efforts in these fields, in order to demonstrate the value of [email protected] for addressing challenging problems in the development of new materials. Together, these partners will perform challenging proof ­of ­principle simulations on the computational architectures provided by the supercomputing partners, demonstrating the usefulness and power of the e-­infrastructure developed in this project. The results of these demonstrations will be a key prerequisite to attract new users and software providers to the [email protected] project at an early stage.

Community Building

In the duration of this project we shall disseminate this concept of an e-­infrastructure to users and providers of materials modelling solutions throughout Europe and actively recruit new groups to nucleate a growth process for the [email protected] initiative. By forming an open and unified community of computational scientists on the [email protected] e-infrastructure this project will strengthen Europe's international role as software producer and user by providing a platform where novel solutions to key aspects of materials modelling can find use far beyond the user­basis that can be reached by an individual scientific group developing the software. We will hold several open workshops where the concept, workflows and modules of [email protected] will be explained and demonstrated to interested scientists from academia and industry. By inviting additional software suppliers (outside of this consortium) and potential users of the [email protected] framework we hope to initiate a fast growth of the exploitation of this e­infrastructure in the academic and industrial materials science communities. We will also present the results of this project using the internet facilities (project's own and DEISA/PRACE sites), and also through publication at selected conferences (such as organizing symposia at the European Materials Research Society EMRS conference).