A cost-effective formula for large-scale computation using the Internet is proving its worth, David Bradley reports
When the announcement came four years ago that the University of London Computer Centre's computational chemistry service would cease in 1996, there was only one rational response.
Julie Altmann, secretary of the Computational Chemistry Working Party (CCWP) explains: "It was obvious that the large chemistry community using ULCC would have to find alternative facilities for computing which was to include a combination of machine power, specialist software and expert consultancy which was unique to the centre".
The chemists, concerned that access to computers powerful enough to model everything from anticancer drugs to explosions would vanish, together with software and expert advice, submittedt a proposal collectively through the CCWP - with Cambridge professor Nicholas Handy as its chair - to lever the Engineering and Physical Sciences Research Council into providing a cost-effective alternative.
The aim would be to reboot the UK's chemistry computing system and give computational chemistry a kick-start. "The proposal included 28 mini-proposals for new science encompassing many areas of chemistry," Altmann explains. The proposal was subjected to the usual peer review process as a large science initiative. At the end of 1996, the EPSRC granted four years' funding for a "facility for computational chemistry in the UK" amounting to several thousand hours of central processing time on Columbus, one of the supercomputers at the Rutherford Appleton Laboratory.
Roger Evans, manager of supercomputing services at the RAL, explains: "The grant also included a lot of money for staff effort and application software and the sharing of Columbus was in lieu of a dedicated chemistry machine."
Altmann, the chemistry software expert who runs the facility, says at the end of the first year of full service there were 65 subprojects and 80 users. New project ideas are still being received and the number of users is growing.
"All branches of chemistry are represented," says Altmann. "The software supported means chemists don't have to purchase expensive third-party packages and hire staff to maintain them." This, she adds, has been most important for young and upcoming chemists who have neither the time nor the money to start from scratch. Many of the users are theoretical chemists who work on realistic models of chemicals and their reactions and the facility provides the power, software and expert advice. There are those chemists who have powerful workstations but Evans points out that a Columbus (a DEC 8400), amounts to ten processors running together with six gigabytes of memory and is "about ten times what any individual could have. More processing power and memory equates to larger molecules or more precise calculations on a given molecule".
The chemists access Columbus via the Internet and submit their jobs to a queuing system. Scientists can carry out interactive work but most jobs are too big for humans to intervene although they still need to spend time on the Internet setting up and keeping track. It is important, therefore, that the Joint Academic Network (JANET) is fast enough to cope with the traffic.
An average academic site can carry only§ 10§ megabits per second - far faster than a desktop modem but not enough for a supercomputer connection. Altmann points out that the SUPERJANET 3 upgrade scheduled for next month will provide 155 megabits pe s ond shared bandwidth across the UK.
But what do chemists do with all this power? A typical project might involve carrying out several experiments, for example, generating highly unstable chemical species found in the atmosphere, stratosphere or deep space, or being created by complex chemical processes and recording their spectra. The interpretation of these spectra is rather complicated and requires very large calculations with sophisticated software. With a supercomputer program running on several processors in parallel, scientists can interpret their spectra more quickly. "They (chemists) can eliminate much of the trial-and-error work costing a great deal of time and money. They can concentrate on a handful of routes with the most probable positive outcome," says Altmann.
Columbus is running software to study the electronic structure and the bonding between atoms in many kinds of molecules. One of these is a group of organic nitrogen-containing compounds known as terpyridines. These compounds can bind to the so-called lanthanide and actinide metals, such as gadolinium and lutetium. Very little is known about the formation of these metal complexes but they are attracting a great deal of interest from chemists because they can selectively extract these metals from aqueous mixtures and might be used to remove toxic heavy metals from industrial waste streams.
Such remediation technology would allow industry to recycle rare and expensive heavy metals and would prevent the metals entering the environment.
Michael Drew of Reading University is using the supercomputing facility to model the interaction of terpyridines with metals. The supercomputer model of the actinide and lanthanide complexes will allow him and his team to fine-tune the properties of the terpyridines so that they can be made yet more specific in their extracting abilities so that non-hazardous ions in the dirty water are ignored.
After a year in operation and some major upgrades, is the facility fulfilling its role? Altmann thinks so: "We now have an impressive software holding on Columbus, which enables studies using quantum chemical methods and simulations of all sorts of molecular processes. This is why so many chemists can use the facility - they can choose the right software to suit their research interest."
The first year interim report to the EPSRC has been assessed by independent scientists. Altmann says: "It appears that they were pleased with the science, some of which had been published in learned journals and the rest is expected to be published when the study is completed. Our success is measured in terms of producing science worthy of publication."