Multiscale Simulation

Unleashing the power of distributed computing infrastructure – workflow automation for accelerated scientific investigation

The multiscale simulation challenge

Real world processes consist of components and interactions that take place on different length and time scales, often crossing the boundaries between different conventional scientific disciplines. In order to model the world with increasing fidelity, today’s scientists and engineers face the challenge of designing, deploying and controlling multiscale systems where processes acting at different scales coexist and interact. Such multiscale studies, when performed in three spatial dimensions and in time, require large scale and, sometimes, extreme scale computing capabilities.

However, executing a multiscale simulation is a non-trivial task. It typically requires access to multiple computational resources, consisting of clusters and supercomputers, together with a host of software services running on top of them. An example of such a scenario occurs in the modelling of interactions between clays and specific types of synthetic polymers, which together produce nanocomposite materials with unexpected and valuable properties. Their applications are wide ranging, from light but strong materials suitable for automotive and aircraft design, through use in drilling for oil and gas, to drug transport. Hierarchical multiscale modelling resolves the interactions at quantum mechanical, atomistic, course-grained and finite element scales, allowing simulation of a material at appropriate length and time scales, safe in the knowledge that it represents a specified physicochemical composition.

The solution

Our approach allows the coupling of models to be done in a standardised and automated manner, running on a suitable e-infrastructure which allows all components to run in optimal fashion and to manage the data flows during the simulation.

We deploy a scientific workflow engine above a standard software stack, to orchestrate the process of running and linking simulations at different scales.  For example, in the clay-polymer case, one can link applications such as the density functional code CPMD, the molecular dynamics package LAMMPS, and the finite element suite in OpenFOAM.

Within the workflow platform, the multiscale application is decomposed into a number of so-called “snippets”. Each snippet can be written in a different programming language; moreover, the workflow engine enables users to execute lengthy sequences of steps or just selected snippets. In this way, time-consuming submodels do not have to be re-run from scratch each time a modification is made during development.

Impact on the discovery process

With these methods, we have been able to design, computationally, new, biodegradable clay-swelling inhibitors which meet the increasingly demanding requirements of the oil industry for exploration and production.  Other applications include the design of novel clay-polymer nanocomposites by simulation, as opposed to the lengthy trial and error processes required for their experimental discovery today.

The ability to automate many common, mundane tasks through the use of workflow tools unleashes the power of multiscale computing on suitable local and/or distributed e-infrastructure, making it far more than the sum of its parts. Workflow automation makes possible the investigation of multiscale models in a routine way, greatly accelerating the scientific and engineering discovery process.

The figure shows the distinct length and time scales in the so-called scale separation map which are coupled in our multiscale approach, in which data is fed from the quantum mechanical level (CPMD), through all atom molecular dynamics to coarse-grained molecular dynamics in order to compute materials properties of interest. All data transfers are performed in a standardised multiscale modelling environment.


J. Suter, P.V. Coveney, R. L. Anderson, H. C. Greenwell, and S. Cliffe, “Rule Based Design of Clay-Swelling Inhibitors”, Energy and Environmental Science, 4, 4572-4586, (2011),  DOI: 10.1039/C1EE01280K

B. Chen, J. R. G. Evans, H. C. Greenwell, P. Boulet, P. V. Coveney, A. A. Bowden, and A. Whiting, “A critical appraisal of polymer-clay nanocomposites”, Chemical Society Reviews, 37, (3), 568-594, (2008), DOI: 10.1039/b702653f.

J. Suter, D. Groen, L. Kabalan, and P. V. Coveney, “Distributed multiscale simulations of clay-polymer nanocomposites.” Materials Research Society Proceedings, 1470, (2012), DOI: 10.1557/opl.2012.1009