Physics simulations with novel computing architectures: Optically bound swarms and light-driven micro-machines

点击翻译按钮获取中文摘要

The aim of this PhD project will be to use computer simulations to explore the optical trapping and binding of large numbers of shaped colloidal particles. Further, this complex system will be used as a test-bed for investigating the influence of processor architecture on computational efficiency in physics-based simulations. The project will be a new collaboration between ARM Holdings plc and the School of Physics, University of Bristol. Optical trapping has been studied in the School of Physics for more than 10 years. With tightly-focused laser beams, it is possible to manipulate micron-sized particles to apply forces and assemble complex nanostructures. A related phenomenon is known as optical binding, in which mutual scattering of light causes microscopic particles to attract each other. While most research in this area involves spherical particles, recent work from the Bristol group examines the phenomenon in the case of nanowires [1]. It is clear from this work that the reduction in symmetry affords opportunities for orientational and positional ordering, as well as light-driven translations and rotations.The potential for self-organisation of particles using light alone is exciting, and points to a range of possible applications, including optically-bound swarms, self-organised metamaterials, ultra-sensitive force sensors and light-driven micromachines. The aim of this PhD project will be to simulate a range of possibilities within this field, taking two main lines: (1) to study the influence of symmetry on the configurations and motions of optically bound nanostructures, and (2) to search for emergent phenomena when large numbers of particles are considered - a form of optically driven active matter. Simulations will require hydrodynamic and light-matter interactions to be calculated; both are computationally demanding and involve a range of numerical techniques, including Langevin dynamics, Cholesky and LU decompositions, fast Fourier transforms and Krylov sub-space optimisations. The collaboration with ARM will enable us to use this system as a test-bed for exploring the influence of different computer architectures on the performance of physics-based simulations. As well as using the new GW4 ARM-based supercomputer, Isambard, the project will employ hardware emulators, courtesy of ARM, to allow the effects of different processor architectures to be assessed, including e.g. super-wide vectorisation, reproducible long wordlength floating point accumulation and variable numbers of cores. Optimisations of the numerical methods will ultimately be fed back into the ARM maths libraries.During the PhD project, the student will perform most of their research at the University of Bristol, but it is anticipated that they will spend two 3-month periods at ARM in Cambridge, towards the ends of the first and second years of study, to receive training in the use of the ARM emulators, and to learn more about the ARM design philosophy for high performance computing.This PhD project relates to a number of EPSRC strategic themes. In the physics realm, it supports both the "Light matter interaction and optical phenomena" theme and the "Biophysics and soft matter physics" theme. The opportunity to design light-driven machines relates to the "Robotics" theme, while the testing and optimisation of computer architecture supports EPSRC's "Microelectronics design" theme. [1] Simpson, S.H. et al., Nano Letters, 17, 3485-3492 (2017).