At that stage, we were still developing the algorithms needed, using local Oxford machines, planning to move the work on to the CRAY at the appropriate stages within 3 to 6 months of the start date. In the event, the removal of the Y-MP and installation of the Cray J90 overtook this plan. We had several difficulties with the J90: in the first place, code which previously ran on the Y-MP was rejected by the new compilers, and caused us some hassle to overcome the problems; the second was that turnaround was not good, and the system seemed difficult to use. As a result, we continued with the local facilities, and more or less abandoned the use of the RAL machines when we discovered that the code ran faster on our local, slightly elderly, DEC Alpha workstation. Our subsequent use of the RAL allocation was therefore minimal.
From the scientific point of view, the project has been extremely successful. We have more than achieved all the targets that we set ourselves in the Spring of 1995. In particular, we discovered how to design very fast algorithms for the key relativistic molecular integrals late that year, completely overturning all our earlier ideas on how to do this. As a result, calculations for which we had expected to use supercomputers could be transferred to desktop workstations. A further major breakthrough came when we realized in the Spring of 1997 that it was possible to set up the relativistic molecular structure problem in a manner compatible with QED (Quiney, Skaane and Grant, 1997, J. Phys. B: At. Mol. Opt. Phys. 30 L829-834)). These major breakthroughs have enabled us to run polyatomic calculations on the DEC Alphastation 500/400 purchased with the aid of an EPSRC grant GR/L55155 as well as calculations on diatomics as originally envisaged. We have also adapted other modern ideas for studying molecules with a large number of atoms to run Dirac-Hartree-Fock calculations for such molecules as Germanocene (Ge(cp)2), with 21 atomic centres, and for atomic clusters with up to about 20 atoms.
Something of the range of our work is indicated in the list of
publications attached. Two things should be noted: the first is that
we have still to write up the details of our new BERTHA code (the last
two items), and that the articles with Dr. Scott do not reflect our
current thinking, but rather what we were doing early in 1995.
Nevertheless, Dr. Scott was highly influential in using the algebraic
computer system, MAPLE, to explore a range of algorithms that might be
used for relativistic molecular integral generation and we are certain
that we shuld not have got as far as we have done without using
PUBLICATIONS and ARTICLES IN PREPARATION
H. M. Quiney, H. Skaane and I. P. Grant, Ab initio relativistic quantum chemistry: four components good, two-components bad! Adv. Quant Chem. 32, 1-49 (1998). (Thematic volume: Quantum Systems in Chemistry and Physics Guest Editors: S. Wilson, P. J. Grout, R. McWeeny, J. Maruani and Y. G. Smeyers).
T. C. Scott, M. B. Monagan, I. P. Grant and V. R. Saunders, Numerical computation of molecular integrals via optimized (vectorized) FORTRAN code. Nucl. Instr. and Methods in Physics Research A 389 117-120 (1997)
H. M. Quiney, H. Skaane and I. P. Grant, Relativistic calculation of electromagnetic interactions in molecules, J. Phys. B: At. Mol. Opt. Phys. 30 L829-834 (1997).
H. M. Quiney, H. Skaane and I. P. Grant, Hyperfine and PT-odd effects in YbF . J. Phys. B: At. Mol. Opt. Phys. 31 L85-95 (1998).
T. C. Scott, I. P. Grant, M. B. Monagan and V. R. Saunders, Generation of Optimized FORTRAN Code for Molecular Integrals of Gaussian-Type Functions, MapleTech Vol. 4, No. 2, pp. 15-24 (Boston, Birkhäuser, 1997).
H. M. Quiney, H. Skaane and I. P. Grant, Relativistic, quantum electrodynamic and many-body effects in the water molecule, Chem. Phys. Letts., 290, 473-480, (1998).
I. P. Grant, H. M. Quiney and H. Skaane, BERTHA: an ab initio relativistic molecular electronic structure program
H. M. Quiney, H. Skaane and I. P. Grant, Relativistic molecular integrals with Dirac spinors