CCP5 - Environment from the Molecular Level (e-Minerals)

Project Overview

Many environmental problems require an understanding of fundamental mechanisms and processes at a molecular level to which computer simulations can give a considerable advantage. Developments in atomistic simulation tools must now be linked with Grid technologies in order to facilitate simulations that can be performed with realistic conditions and which can scan across a wide range of chemical and physical parameters.

Objectives

This proposal brings together simulation scientists, applications developers and computer scientists to develop UK e-science and Grid capabilities for molecular simulations of issues critical for our environmental. A common set of simulation tools will be developed for a wide range of applications and the Grid environment will be established:

• Web portals and link to Grid technology;
• Enhanced simulations using distributed resources;
• Enhanced data manipulation, metadata creation and analysis;
• Collaborative working.

e-Minerals Science Testbed Runs on NW-GRID

The e-Minerals project is making "a comprehensive assault on the issue of transport of pollutants in the environment". This includes studies of heavy metal poisonous waste, toxic organic molecules and nuclear waste encapsulation. Codes used include OSSIA, GULP, SIESTA, CASTEP and DL_POLY 3.

The following text is based on informatino provided by Martin Dove and Rik Tyer.

Compressibility of Amorphous Silica using DL_POLY 3.

Density is not linear with applied pressure, the curvature yields the bulk modulus of the material as it varies with pressure. The BM has a minimum (material is "softest") at around 2GPa.

To understand this phenomon three different configurations of atoms and two different inter-atomic potentials were considered in runs on NW-GRID. Using a parameter sweep with the RMCS software enabled a large number of points to be calculated on the curve which could be differentiated accurately to give the BM.

Over 600 simulations were run in this analysis producing around 2GB of data which was stoed in SRB. This was done by a 3rd year project student. The data could be shared and discussed by colleagues.

Compression of the mineral Diopside using CASTEP.

Diopside, CaMgSi2O6, is found in the Earth's mantle and consists of chains of SiO4 tetrahedra connecting MgO6 octahedra and irregular CaO8 polyhedra. Results were computed for volume vs. pressure to determine the equation of state of this material. This enabled the inter-atomic spacing to be determined as a response to pressure. The Ca-O distance (the longest) changes most, i.e. are highly compressible and this phenomenon is currently being understood through related investigations.

Cation ordering in Clay layer silicates using OSSIA.

Energies of different neighbours are obtained using QM calculations. There is a site-ordering phase transition which ocurrs on heating. This is difficult to study experimentally because clays are unstable and contain large amounts of water. Ordering is therefore studied using the OSSIA Monte Carlo code running as a function of temperature over experimentally inaccessible values. Graphs are produced showing order vs. temperaturs, stiffness cs. temperature, heat capacity vs. temperature. These all clearly show the phase transition.

In this study both QM and empirical calucations were used to study the effects of water, the surface-molecula interactions, the structure of species in solution and the ssurface structure of the mineral.

Pollutant adsoeption on mineral surfaces using SIESTA.

Organic molecules such as Dioxins and PCBs are present in the environment as a side-effect of human industrial or agricultural activities. Unfortunately they do not degrade, but pass into the food chain and are starting to cause illnesses such as reproductive problems, skin, liver and stomach conditions and thyroid problems. These componds leach into water and may settle on clay-like minerals such as pyrophyllite or smectite or chalks such as calcite.

Dioxin = C12O2HxCl8-x PCB = C12HxCl10-x

Methodology using Siesta is as follows:

• SIESTA quantum mechanics (DFT) code: demanding!
• Compute energies of isolated molecules and mineral surfaces first
• Scan molecule over surface to identify possible adsorption sites
• Compute energy minimisation, and subtract isolated molecule and
  • mineral surface energies to give adsorption energy
• Repeat for all congeners (ideally)

Technology Used

The requiremements from the above science were as follows:

• Require a client environment that provides integrated compute,
  • data and metadata functionality
• Want to be able to interact with the grid from the desktop, i.e.:
  • o Tools need to be `firewall friendly'

    o Tools need to be lightweight' and self-contained' o Tools should be integrated into native package management systems

• RMCS is the client environment used by eMinerals for desktop
  • job submission and job management

RMCS ia a genuine desktop Grid client. There is no need for VDT, Globus etc. to be installed locally. It is firewall friendly because it uses Web services. Functionality is either provided as a set of client tools or a library as required. It provides a relatively simple interface to integrated compute, data and metadata functionality.

RMCS principally implements a met-scheduling functionality. It polls to find spare CPUs and submits jobs to them. A temporary working directlry is created and input files transferred from SRB. Jobs are submitted to thebatch queue via SGE on NW-GRID. Output is marked up using XML (using Toby White's FoX library). This allows key data to be extracted across a range of output files for post processing using the AgentX software. This data can be graphed, e.g. using RGem.

Rcommands, used with RMCS, are shell tools and associated Web services that provide meta-data recording and editing functionality from user desktop or from compute resources. The meta-data schema has been designed to meet the e-Minerals requirements.

RGem searches across multiple data sets and extracts a particular parameter via AgentX and RCommands the meta-data database. This can easily enable plots of different parameters to be produces, e.g. as a function of temperature or pressure as described above.

Summary

A mix of capabilities accessible vie Grid middleware is important. A wide range of computations can be done as illustrated above.

Parameter sweep studies are a "killer" feature for the Grid. There are periods when large numbers of processors are required interspersed with quiet perios for analysis. A single point of access, the RMCS meta-scheduling function is required. Similarly a single point of access for data using SRB and book-keeping using RCommands to record the meta-data with a URI to the data location. This integrates compute, data and meta-data in one environment. This is key to running many hundred jobs and extracting their significant output.

Outputs

A demonstration of a Monte Carlo study with RMCS (Monty) and the OSSIA2004 application was run at the

NW-GRID Seminar on Materials Modelling, Daresbury, 20/3/07.

Several papers on work of the e-Minerals project exist, for instance as presented at the e-Science All Hands Conference, Nottingham, September 2006.