PhD studentship in controlling errors in the simulation of chaotic systems on digital computers

Principal Investigator: Professor P. V. Coveney

Start date: March 2021

UCL Chemistry Department is offering a fully funded studentship for three years to a highly motivated candidate to start in March 2021. The project will involve a collaboration with CBK Sci Con and is co-funded by them. CBK is a sci-tech consultancy devoted to the provision of high end scientific, technical and management advice to businesses in computational science domains. This will enable the student to gain a thoroughly rounded view of the research landscape in computational science, including the vital importance of disseminating their research findings within academia, industry, governmental organisations, and the general public.

The student will investigate and mitigate a newly discovered pathology in the simulation of chaotic dynamical systems on digital computers. This pathology was discovered and quantified for the case of the very simple generalised Bernoulli map, by direct comparison of the exact results from continuum mathematics with those obtained from a floating point representation. These numerical errors make many numerical results wrong, while most users remain completely unaware of them.

The purpose of this PhD studentship is to assess the extent of these errors in other, more complicated, cases in order to restore credibility to predictions arising in diverse areas of computational science, from turbulence to molecular dynamics and reaction-diffusion systems. As such the student will be expected to work closely with two EU-funded projects currently running in the group; CompBioMed is a Centre of Excellence in Computational Biomedicine and VECMA is a Future and Emerging Technology project for the Verification of Exascale Computing in Multiscale Applications.

The applicants should have, or be expecting to achieve, a first or upper second class Honours degree or equivalent and a strong background in one or more of applied mathematics, dynamical systems theory, numerical analysis, computer science and theoretical physics. Familiarity with one or more programming languages would be a distinct advantage (e.g. C/C++, Python, etc.).

Interested candidates should initially contact supervisor p.v.coveney "at" with a degree transcript and a motivation letter expressing interest in this project. Informal inquiries are encouraged. Suitable applicants should complete an electronic application form here. They will be invited for the interview no more than 4 weeks after the application deadline. Any admissions queries should be directed to Dr Jadranka Butorac at j.butorac "at"