Quantum Monte Carlo made easy

Computer simulation is used in many different contexts, but the questions addressed are often surprisingly similar. What happens when many simple objects interact? How does the behaviour of the whole emerge from the simple laws obeyed by its parts? The constituent objects range from electrons and nuclei to girders and cables, but the common aim is to predict the complex large-scale behaviour from the simpler small-scale behaviour. In essence, computers are used to build bridges between length scales.

The use of computers to simulate molecules and solids in terms of their constituent particles is a successful example of this sort of bridge building. The inputs are the identities of the atoms that make up the molecule or solid, and the outputs describe the behaviour of the assembly as a whole. Since the electrons that bond atoms together are quantum mechanical objects and behave more like waves than particles, these simulations have to solve the quantum mechanical Schroedinger equation.

Quantum mechanical simulations are increasingly used across much of the EPSRC's physical sciences portfolio, including all of the following areas: quantum fluids and solids; magnetism and magnetic materials; functional ceramics and inorganics; materials for energy applications; electrochemical sciences; surface science; chemical structure; chemical reaction dynamics and mechanisms; plasmonics; light matter interaction and optical phenomena; photonic materials and metamaterials; cold atoms and molecules; chemical biology and biological chemistry; graphene and carbon nanotechnology; and catalysis. They have also had an impact on parts of the engineering portfolio. The practical importance of quantum mechanical simulation to many of these fields is still quite small, but it is growing fast and its success has been one of the most striking scientific stories of the past twenty years. Few other fields of research have done so much to revitalise and maintain the health of other disciplines.

By far the most common approach to quantum mechanical materials simulation is density functional theory (DFT). This offers a good balance between ease-of-use and precision, but is not quite accurate enough for room-temperature chemistry, biology and biochemistry. This has led to a widespread feeling that it is time to put effort into developing more accurate approaches. One of these is the diffusion Monte Carlo (DMC) method that is the subject of this proposal.

DMC simulations are much slower than DFT simulations but are more accurate and can be used to study large systems of several thousand interacting electrons. Furthermore, unlike DFT codes, a well-written DMC code will run with almost 100% parallel efficiency on even the largest of the world's parallel computers. Large-scale parallelism just happens to suit the method. This proposal seeks support to make CASINO able to run efficiently on the next generation of supercomputers, which will use GPUs (graphics processing units, originally developed for use in graphics cards for computer gamers) for most of the heavy computation. GPU programming presents new challenges and a great deal of work will be required to re-engineer CASINO to use GPUs efficiently. We will also work on making CASINO much easier to use, and further develop our ideas about how to optimize trial wave functions and calculate inter-atomic forces efficiently using DMC. Our aim is to make CASINO into a flexible, efficient, and easy-to-use tool capable of running almost anywhere. CASINO is already the world's most widely used DMC code. We would like the chance to make sure that it retains its lead as the use of DMC takes off over the next decade or two.

In addition to the academic benefits described in the "Academic Beneficiaries" section, our work will benefit the nation by producing highly-trained people. Computational science "has become generally accepted as the third pillar of science, complementing and extending theory and experimentation" [International Review of Research Using HPC in the UK, EPSRC (2005)] and is widely used in industry as well as academia. Despite the obvious importance of the field, we estimate that the UK produces no more than 100 people per year with a deep and high-level understanding of computational (as opposed to computer) science. The researchers employed by this grant will contribute to the UK's small pool of experts.

More specifically, now that the era of continuously increasing computer clock speeds is coming to an end, improvements in computational power are relying almost entirely on increases in parallelism. Over the next 10 years, every enterprise that requires substantial computer power - a list that includes the electronics, pharmaceutical, transport, banking, financial, defence, internet, and energy industries among others - will be forced to come to terms with massively-parallel massively-multi-threaded computing. For the UK economy to remain internationally competitive, it is important that our research and education systems produce an increasing number of people capable of programming computers with many thousands of cores and using them to solve real-world problems. The researchers employed by this grant and the PhD students trained alongside them will be in a position to render an important service to the nation's economy, whether or not they stay in computational science.

The purpose of the CASINO diffusion quantum Monte Carlo code is to model the properties of molecules and solids on the atomic scale, allowing new materials and molecules to be designed and assessed using a computer without wasting money making and testing the real thing. Materials modelling techniques based on quantum mechanics are only now making their way into industry, but their potential is illustrated by the success of the CASTEP density-functional code, which originated as an academic project at Cambridge University but has been sold commercially since 1995 and accumulated sales of over $20 million.

In the longer run, far beyond the end of this grant, we expect that quantum Monte Carlo methods will also be used in industry, just as density-functional methods are today. This will not happen soon, mainly because quantum Monte Carlo simulations take so much longer to run than density-functional simulations, but density-functional theory is not reliably capable of achieving the accuracy required to study room-temperature chemical and biochemical processes and is unlikely in our view ever to do so. This leaves the way open for more accurate methods such as quantum Monte Carlo.