ComputationalChemistryActivity | /OurEvents | /OurCollaborations
ComputationalChemistryActivity
The Problem | The Solution | The System | The Software | The Applications | The Expertise | The People
The Problem
Computational chemistry is in the process of becoming a major component in almost all forms of chemistry research. Key research papers across the field now routinely involve the results of computation alongside experimental ones. Unfortunately, particularly in chemistry, a boundary still exists between experimental science and computational science. This Activity will provide a route through this barrier to progress by allowing the wider chemistry community access to computational chemistry expertise and cutting-edge facilities.
The Solution
A Research Computing Officer (RCO), AndrewTurner, has been appointed in the School of Chemistry at the University of Edinburgh to provide a conduit into computational chemistry for non-specialist researchers. An 72-processor, Beowulf-type cluster has also been purchased to allow these computational tools to be used on systems of experimental interest. In addition, our efforts to develop a resource in Edinburgh are being mirrored by our partners in EaStCHEM at the University of St. Andrews. If the two resources are viewed as two nodes of a network, the development quickly acquires the critical mass to support a wide range of activities through coverage of a broader range of software and to achieve sustainability in the face of absence of key personnel. The complete set of resources - personnel and hardware - are known as the EaStCHEM Research Computing Facility.
The System
The AMD Opteron based Linux cluster was purchased from Streamline Computing and is organised as:
- Twelve 4-processor compute (thin) nodes with 8GB memory and 160GB disk space.
- Two 8-processor compute (fat) nodes with 16GB memory and 400GB disk space.
- Two 4-processor interactive use nodes with 4GB memory and 800GB disk space.
- One 4TB RAID disk array.
This set up gives us the flexibility to use parallel codes that run best with shared memory (fat nodes) and the those that run well with distributed memory (thin nodes) simultaneously.
The Sun Grid Engine will be used to manage queues on the system and to combine our resource with the sister EaStCHEM resource in St. Andrews. We plan to explore the possibility of using this close link between the sister machines to develop data mirroring for - in the first instance - off-site redundency. To this end, we are investigating using the QCDgrid software in collaboration with edikt2 partner GeorgeBeckett.
We plan to develop and run a web interface to the cluster that will allow browser based monitering of the queues and submission of batch jobs. This may well result in collaborations with the EdiktPortalActivity, EPCC, NeSC and EUCS.
The Software
Gaussian 03 - Multipurpose electronic structure suite.
MOLPRO - Specialises in highly accurate electronic structure calculations on smaller systems.
CASTEP - Electronic structure calculations on periodic and condensed phase systems.
DL_POLY - Molecular dynamics simulations using an empirical description of chemical interactions.
The Applications
The recently acquired ability to perform accurate electronic structure calculations on large molecules or in the condensed phase and to simulate the thermal behaviour of large assemblies of (bio)molecules is changing the way in which chemical research is undertaken. Physical measurements on systems of previously unimaginable complexity may now be interpreted by using calculations to provide the values of key quantities. The drive to synthesise new materials is guided by the ability to do realistic calculations – to identify the key interactions responsible for a particular magnetic order, for example. In the biochemical field, simulations play a key role in enhancing the fragmentary picture available from experiment, as in turning a static diffraction picture of an ion-channel protein into an understanding of its mode of operation.
The Expertise
There are four research groups within the School of Chemistry specialising in computational and theoretical methods:
- Madden Group - Condensed phase dynamical simulations that include electronic structure.
Camp Group - Stochastic and dynamical simulations of condensed phase systems, liquid crystals and magnetic fluids.
Morrison Group - Electronic structure calculations of crystals and periodic systems.
Barran Group - Molecular modelling of proteins, peptides and other natural products.
AndrewTurner, the RCO, brings his own expertise in both electronic structure calculations and classical simulations of the condensed phase and provides a dedicated resource for the non-specialist researchers.
The People
People involved with this activity:
Publications
See Publications