HPC simulations of complex solids and clusters using static lattice techniques

Lead Research Organisation: University College London
Department Name: Chemistry

Abstract

We are proposing to develop and implement new software on HPC platforms which will enable new wide-ranging scientific applications in materials simulations using static lattice techniques. The project will initiate new developments in the GULP (General Utility Lattice Program) which over the last decade has become the standard code for lattice simulations, with a very substantial national and international user base of several thousand users. Current versions of the code are, however, limited to single processor or small cluster platforms, which prevents applications to the type of complex problems and systems which are addressed by materials chemistry and physics. The project will develop new software, which will be based on (i) an efficiently parallelised version of GULP; (ii) a new integrated version of GULP bringing together developments from different groups; and (iii) a new master code KLMC (Knowledge Led Master Controller) that is able to setup novel complex simulations, span multiple GULP jobs, and analyse results in order to achieve the following main application types:(a) mapping energy landscapes as a route to complex simulations of solid state reactions, enumeration and sampling of local configurations in disordered systems;(b) ion ordering in solid solutions, which show unique magnetic, superconducting, optical and catalytic properties;(c) interaction and clustering of multiple defect centres in solid state systems, for example, materials exposed to radiation;(d) structure prediction and properties for complex solids with large unit cells and large clusters or nanoparticles;(e) surface and interface structure and property determination and prediction;(f) free energy calculations of a phase transitions and calculation of diffusion paths and rates;(g) crystal growth of nanoparticles and surfaces.The project is a collaboration between multiple developers as well as academic and industrial users.

Planned Impact

The extensive user base of the GULP lattice simulation code will ensure a wide and substantial impact for this project. The code currently has been installed at over 4000 academic and industrial locations and is the standard simulation package for static lattice simulations of materials. By enabling new scientific applications via developing new functionality and by enabling GULP to scale efficiently on multi-core HPC platforms, the impact of the lattice simulation techniques will be greatly extended. The GULP code has a substantial academic user base; but is also supported and distributed to industrial users by Accelrys Ltd as a module of their Materials Studio package. Materials simulation using static lattice methods are extensively used in industry in screening materials properties and in calculating thermodynamic and defect properties. The existing large impact of the code will be greatly increased by the capability to model more complex problems and systems, and to run large scale simulations on HPC platforms. Impact will be further enhanced by the effective dissemination of the outputs of the project via the CCP5 project, the British Association of Crystal Growth, and the HPC Materials Chemistry Consortium, all of which have a large academic and industrial membership. Moreover, all registered users (of which there are over 4000) will be notified by electronic mail when the HPC implementation of GULP is available for download.

Publications

10 25 50

publication icon
Deacon-Smith DE (2014) Interlayer cation exchange stabilizes polar perovskite surfaces. in Advanced materials (Deerfield Beach, Fla.)

publication icon
Escher S (2017) Structure prediction of (BaO) n nanoclusters for n ? 24 using an evolutionary algorithm in Computational and Theoretical Chemistry

publication icon
Lazauskas T (2018) Thermodynamically accessible titanium clusters Ti, N = 2-32. in Physical chemistry chemical physics : PCCP

publication icon
Mora-Fonz D (2017) Development of Interatomic Potentials for Supported Nanoparticles: The Cu/ZnO Case in The Journal of Physical Chemistry C

publication icon
Mora-Fonz D (2017) Why Are Polar Surfaces of ZnO Stable? in Chemistry of Materials

publication icon
Sokol AA (2014) Double bubbles: a new structural motif for enhanced electron-hole separation in solids. in Physical chemistry chemical physics : PCCP

publication icon
Woodley S (2013) Structural and Optical Properties of Mg and Cd Doped ZnO Nanoclusters in The Journal of Physical Chemistry C

publication icon
Xie Z (2019) Donor and acceptor characteristics of native point defects in GaN in Journal of Physics D: Applied Physics

 
Description Following our earlier in-house development of the Knowledge Led Master Code (KLMC), this funding allowed us to parallelise and port onto multiprocessor computers. Thus exploitation of National High Performance Platforms has become possible and crucially new science enabled. This development is aimed at resolving the problems occurring in modelling complex materials, their structure and properties, that in particular require massive samplings of configurational spaces and/or hierarchical simulations. KLMC automates many tasks traditionally performed by the user of a range of third party codes; enable a multistage approach where the KLMC code learns on the fly and refines input files that are submitted for new calculations - hence the name knowledge led; and is able to exploit massively parallel computer platforms for a more general set of applications that may require statistical sampling. Currently KLMC is capable of updating a simple database of structures (or solutions); performing post analyses (e.g. computing radial distribution functions, ensemble average - Boltzmann weighted - properties); generating and reading input and outputs files for commonly used materials codes (GULP, FHI-AIMS, VASP and NWCHEM), and running these either on the same platform (either through system calls or as library files) or remotely (KLMC runs itself on a local machine and many individual calculations, or tasks on larger resources elsewhere). Applications include: simple task farming (screening structures from the database through a third party code); structure prediction of nano-sized clusters, surfaces and bulk phases using a range of global optimisation techniques based on basin hopping and genetic algorithms; exploration of ergodic regions (application of the energy lid or threshold algorithm); and statistical sampling of solid solutions or multiple point defects in a crystalline solid, allowing for statistical sampling not only of potential energy, but also more generally any thermodynamical potential of interest.

Support from this grant also included parallelisation work of the third party code, GULP, and, moreover, enabled the creation of a suitable library version of the third party code FHI-AIMS for KLMC (and potentially other codes) to utilise.

The overall objective to enable new science, employing the KLMC code, by software developments and enhancements was achieved. Moreover, the progress made in this phase 1 did impress the four referees assigned to assess our application for phase 2 funding, all of whom gave our phase 2 proposal the maximum score.
Exploitation Route Potentially, KLMC software can be used by secondary schools, or colleges, and moreover there is a high potential for future commercialisation of the KLMC software.
The new code enables a host of new complex atomistic simulations to be performed with relative ease and efficiency on modern HPC platforms, and has been distributed to a number of external users, current beta-testers, forming a nucleus of a future KLMC user base. In particular, D. Scanlon (UCL) and P. Ngoepe (Limpopo, South Africa) are currently using KLMC to investigate defect chemistry and doping of ZnO and Li doped manganese dioxide. A. Walsh (Bath) is studying ion exchange and transport in the quaternary Cu-based semiconductors, which adopt the kesterite mineral structure, of relevance to solar energy applications; and, with J. Bristol, open-shell, Fe doping in alumina. C. Schön and M. Jansen (Stuttgart MPI) are working on structure prediction of zinc oxysulphide bulk phases, materials used in desulpharisation of methane in fuel cell devices. Other exciting applications of KLMC include the work undertaken by D. Morá-Fonz (a PhD Student at UCL funded by the Mexican Government) who is using KLMC to predict the atomic structure of ZnO polar surfaces; whereas D. Deacon-Smith (UCL PhD student provided as institutional support) has successfully used KLMC to investigate the structure and composition of external surfaces of perovskite materials that have been reported to support a two-dimensional electronic gas (so called 2DEG); and I. Demiroglu and S. Bromley (Barcelona, Spain) have studied the effect of a metallic surface on the structure of ZnO nanoparticles. It is hoped that the latter collaborative project will be expanded into other application areas. More recently, KLMC has been installed on computer resources in Assam (Tezpur University, India), where it is being used in the research of R.C. Deka's group.
Sectors Chemicals,Education,Electronics,Energy,Environment

URL http://www.ucl.ac.uk/klmc/Software
 
Description The solutions used in the software developed in this project will aid the development of other client-server codes (which need not be in the field of Materials Modelling) that also aim to exploit to High Performance Computers. The methodology implemented within developed software and the results on the prediction of atomic structure of nanoparticles, surfaces and atom and vacancy ordering in solids obtained from the use of the developed software will be of use to the wider computational physics and chemistry community, as well as provide guidance to experimental design. The overall economic and societal impact of the work supported by this grant is hard to judge as such impact will not occur immediately.
Sector Chemicals,Education,Energy
Impact Types Economic

 
Description DTA
Amount £60,000 (GBP)
Organisation University College London 
Sector Academic/University
Country United Kingdom
Start 09/2015 
End 08/2018
 
Description IMPACT
Amount £30,000 (GBP)
Organisation University College London 
Sector Academic/University
Country United Kingdom
Start 09/2011 
End 08/2014
 
Description Beta tester of KLMC software 
Organisation University of Barcelona
Country Spain 
Sector Academic/University 
PI Contribution Provided KLMC software, which was also further developed for predicting the atomic structures of inorganic nanoclusters on a metallic surface.
Collaborator Contribution Provided man-power (student) to test and apply KLMC software, co-authored resulting publication.
Impact Publication of research.
Start Year 2010
 
Title KLMC 
Description KLMC, or more specifically the Knowledge Led Master Code, was created with the desire to: automate many tasks traditionally performed by the user of a range of third party codes; enable a multistage approach where the KLMC code learns on the fly and refines input files that are submitted for new calculations - hence the name knowledge led; and be able to exploit massively parallel computer platforms for a more general set of applications that may require statistical sampling. The KLMC code is witten in Fortran90 and uses MPI to simultaneously exploit more than one processor. KLMC has a number of modules, primarily aimed at predicting the atomic structure using global optimisation schemes (e.g. Basin Hopping (BH), Genetic Algorithm (GA), ...). First show case examples for a particular module include: employment of BH to predict the structures of LaF3 nanoclusters ; prediction of surface reconstruction; employment of employment of GA to predict the structures of ionic, covalent and metallic nanoclusters; and prediction of surface free energies. 
Type Of Technology Software 
Year Produced 2017 
Impact Software is currently being employed in a number of research projects based in academic institutions in the UK, Spain, South Africa and India. It is also being employed in the training given to postgraduate students enrolled on the Masters degree course that is run in the Molecular Modelling and Materials Science Centre for Doctoral Training (UCL). 
URL http://www.ucl.ac.uk/klmc/Software/