Report on Computational Science Support and
Development activities at Daresbury Laboratory
This report describes work done for the EPSRC under the Service Level Agreement “Computational
Science Support and Development Activities at Daresbury Labor
atory”. Its structure follows the headings in
the SLA, namely CCP support, parallel supercomputing, distributed computing and the industrial
monthly newsletter is produced which includes information of relevance to all HPC and computational
science activities in the UK, including: CCPs, Parallel Supercomputing and HPCI and distributed
computing. Editions over the past year include:
May 97: highlighting aerospace simulations using unstructured meshes
July 97: highlighting the UK Chemical Da
September 97: highlighting direct numerical simulations of turbulence and distributed numerical
November 97: highlighting the ASCI Blue Pacific System and comparing mesh decomposition
March 98: highlighting calculations of
localised electronic states in solids.
In subsequent years HPC Profile will be published on a quarterly basis.
Work in Support of the CCP Programme
The following subsections describe work done in support of the CCP programme. This can broadly be
egorized as administrative support and applications support. Details of meetings organised and
publications (including newsletters) are given. Extensive user manuals are provided for the major codes
supported. Note that the priorities for the work of each
CCP, including its permanent DL support, if any, is
determined by its Working Group.
The CCP Steering Panel
At the request of EPSRC, Daresbury convened a new CCP Steering Panel, to complement the CCP Review
Panel, with the following terms of reference:
receive regular reports on the progress of each Project;
to review the work of permanent Daresbury staff in support of the CCPs;
to collate the views of CCP members on strategic computing issues;
to discuss mechanisms for collaborative working between Proj
to promote the work of CCPs among the wider scientific and engineering community;
to provide an annual report covering the above items to relevant research council officials for their
to examine CCP proposals prior to submission to the p
eer review process and provide feedback to the
to serve, if requested, as a peer review “college” to advise on grant applications in the area of
computational science and engineering.
The Steering Panel, under the overall chairmanship of Prof.
D.J. Tildesley and comprising the Chairmen of
all current Projects, met on 24 October 1997. The minutes of this meeting are attached as an appendix. The
main issues addressed were the various reviews of the EPSRC Computational Science Programme at
ry Laboratory and EPSRC’s internal review of its own computing delivery mechanisms. A proposal
to consolidate information concerning collaborative projects co
ordinated and/or supported by CLRC was
discussed. The data would be accessed over the web, using
DataWeb technology, in which pages are
generated dynamically when served to a client from content stored in a database. This mode of organisation
has the potential to make much more sophisticated data retrieval possible, in particular allowing the
ion of summary data concerning the CCP and related programmes. The current generation of web
pages is generated independently for each project, with little common structure and as such, extensive
manual collation is needed to generate cross
CCP reports. Ot
her benefits of the database oriented system
would be the ability to cross
reference the scientific work, publications and software information more
fully, allowing those unfamiliar with the CCP programme to find the resources they require. During the
rting year steps to install a new server to provide some of these capabilities were initiated.
Molecular Electronic Structure
Permanent staff support was provided by Drs. M.F. Guest, V.R. Saunders and P. Sherwood.
A CCP1 working group was held on 28 May 1997, at the University of
None issued during current reporting period.
Many of the CCP1 pages were converted to the DataWeb format under the DCI web.
None organised during the reporting period
None organised during reporting period.
Travel and Subsistence.
Applications Development and Support
Developments are targeting the quantum mechanical stu
dy of the electronic structure of molecules, with a
focus on hybrid Quantum Mechanical Molecular Mechanical (QM/MM) techniques and Density Functional
Theory (DFT). Much of this work is conducted within the framework of the GAMESS
UK, CRYSTAL and
Density Functional Theory (DFT) Project
Following the formal completion date of the DFT flagship project, and the departure of the funded RA (Dr.
P.E. Young), subsequent developments have been carried out by Daresbury staff . These have included:
mplementation of more efficient quadrature schemes for large molecules.
Continuing development of the code from CRYSTAL to compute long
range coulomb interactions
using a multipolar approximation.
code is currently integrated into the GAMESS
de for exploitation on large
applications with CCP1 working group members.
The successful renewal proposal for CCP1 included funding for an RA to work at the University of
Manchester on hybrid QM/MM techniques. CLRC is represented on th
e technical steering committee (by
PS) for the project and developments at Daresbury (see
section below) will be made available to
UK is the
molecular electronic structure program for performing
gradient calculations, together with a variety of techniques for post Hartree Fock calculations. Development
and core capabilities are described on the WWW at http://www.dl.ac.uk/CFS. Code development and
exploitation is a joint project bet
ween groups at Daresbury, Utrecht, ICI Wilton and Norsk Hydro
The following features were made available in Version 6.1 of the code, released in October 1997:
An initial release of the Density Functional Module developed under the auspices of CCP
functionality includes energies and gradients for closed and open shell systems, with B3
and LDA functionals.
A variety of features to augment present capabilities for treating excited states. These included both
conventional and direc
RPA (random phase approximation) calculations of transition energies and
oscillator strengths, and a MCLR (Multi
configurational Linear Response) module (C. Fuchs, V.
Koutecky and J. Koutecky, J.Chem.Phys. 98 (1993) 3121)
The calculation of mole
cular properties was extended to include f
function basis sets. An interface was
also provided to the Bader Analysis codes.
The range of ECPs available were significantly extended, following the work of Cundari and Stevens.
These included ECPs and associa
ted bases sets for the Lanthanides (T.R. Cundari and W.J. Stevens, J.
Chem. Phys. 98 (1993) 5555)
Additions to the built
in basis set libraries included the generally contracted basis sets due to Dunning
pvdz, pvtz, pvqz and pv5z etc.).
The original limitation to cartesian basis sets was extended through the provision of spherical harmonic
basis sets for all options within the programme.
Open shell Direct
SCF treatments for ROHF, UHF and GVB wavefunctions.
The ChemShell Software for Q
Quantum Mechanical/Molecular Mechanical techniques offer a cost
effective route to the computational
study of the properties and reactivity of complex chemical systems. They are applicable where a small part
of the system participates in
the chemical reaction, but the course of the reaction is influenced by
interactions with the rest of the system (referred to as the environment). The basic assumption is that the
environment, and it's interaction with the reactive site, can be treated usin
g a classical approximation,
exemplified by the Molecular Mechanics method.
The most important examples of applications are in the domain of chemical catalysis, including zeolites,
enzymes and organometallic catalysis. In all cases the objective is to redu
ce the cost, and increase the
reliability, of calculations on these systems, compared to the use of QM methods alone. The user
community for these methods includes all those engaged in the modelling of reactivity of complex chemical
our locally developed QM/MM implementation (ChemShell) is currently used in a number of
collaborations with the UK academic community and in collaboration with industry. The code, while still
under development, is available to the community under licence
and the first user licences have been issued,
including the Royal Institution, Universities of Bristol, Bath, and Southampton and Aberystwyth. The code
forms the basis for a CEC
funded industrial project Quantum Simulation in Industry, (QUASI), which
ed in Jan 98) which will extend the applicability of the QM/MM techniques to massively parallel
computers. The consortium is international, but includes three partners from the UK: CLRC, the Royal
Institution and ICI.
Apart from early design and developme
nt work on QUASI, activities during the 97
98 SLA period focused
on providing visualisation facilities (for normal modes, orbitals etc.) based on OpenGL.
The principal means for distributing the applications developments discussed above a
re through frequent
releases of the GAMESS and CRYSTAL packages. A full account of developments around the CRYSTAL
code can be found in the CCP3 section of this report, and is not repeated here.
Recent workstation implementations of the GAMESS package incl
ude those on the IBM P2SC RS6000,
AXP (EV5), Sun (UltraSPARC 1 & 2), HP
9000 Series (PA8200) and SGI R10000
and Octane. Ports during 1997 included those to the SGI Origin 2000 and Origin 200, plus the Cray T3E,
and to the Intel x86 fami
ly (e.g. Pentium and Pentium Pro) under a variety of operating systems (Linux,
FreeBSD and Windows NT).A significant effort was made to enhance the quality of user documentation
and support, in particular, through the development of hypertext pages and em
ail discussion lists featuring
Publications and Presentations
The following lists publications by staff supporting the CCP1 project at Daresbury Laboratory during the
reporting period. Publications by V R Saunders are discussed in
the CCP3 section.
“The equilibrium molecular structures of cross
conjugated alkenes with particular reference to  and
 Dendralene from
studies of various conformers”. M.H. Palmer. J.A. Blair
Fish and P.
J. Molec. Struct
“Computer Simulation of Zeolite Structure and Reactivity using Embedded Cluster Methods”. P.
Sherwood, A.H. de Vries, S.J. Collins, S.P. Greatbanks, N.A. Burton, M.A. Vincent, I.H. Hillier,
“A theoretical study of t
like reaction between ditungsten hexaalkoxides and alkynes”. S.
Fantacci, N. Re, M. Rosi, A. Sgamellotti, M.F. Guest, P. Sherwood and C. Floriani,
J. Chem. Soc.,
. 1997, 3845.
“The Molecular and Electronic states of 1,2,4,5
zine studied by VUV absorption, near
loss spectroscopy and ab initio multi
reference configuration interaction calculations.'',
M.H. Palmer, H. McNab, D. Reed, A. Pollacchi, I.C. Walker, M.F. Guest and M.R.F. Siggel,
, (1997) 191
“Application of Combined Density Functional Theory and Molecular Mechanics to Small Water
Clusters: a Test of Methodology'', X. Long, J.B. Nicholas, M.F. Guest and R. Ornstein
J. Mol. Struct.
“The Molecular and
Electronic states of 1,2,3
triazine studied by and VUV absorption, near
loss spectroscopy and ab initio multi
reference configuration interaction calculations.'',
M.H. Palmer, M.F. Guest and I.C. Walker
triazine'', M.H. Palmer, S. Parsons, S. Smith, A.J. Blake and M.F. Guest
Performance of Various Computers in Computational Chemistry'', M.F. Guest, `` Daresbury Technical
UK: User's Guide and Reference Manual'', M.F. Guest, J.V. van Lenthe, J. Kendrick, K.
Schoeffel and P. Sherwood, Computing for Science Ltd., Version 6.0 (1997) Parts 1
“Massive Parallelism; The Hardware for Computational Chemistry” M.F. Guest P. Sh
erwood and J.A.
Nichols. In proceedings for HPCI98 & proceedings for the International Symposium on
Supercomputing: New Horizions in Computational Science (1998).
“Halogen Nuclear Quadropole Coupling Constants: Comparison of ab
initio Calculations which
nclude Correlation with Experiment” M.H. Palmer, J.A. Blair
Fish, P. Sherwood and M.F. Guest. Z.
, High Performance Computational Chemistry: Fully Distributed Parallel Algorithms
presented at the Colloquium on Applic
ations of High Performance Computing, Gregynog House,
Newtown, Powys, Wales, 8 April 1997.
, The GAMESS
UK Electronic Structure Code; Developments and Applications
at the 11th RAPS (Real Applications on Parallel Systems) Workshop, GMD
, St. Augustin, 12
The NWChem and GAMESS
UK Electronic Structure Codes; Developments and
, presented at the Department of Chemistry, Universite Louis Pasteur, Strasbourg, 19 June,
The Hardware for Computational Chemistry?
, Workshop of the
Chemistry D3 Action "Theory and Modelling of Chemical Systems and Processes" Perugia
30 June 1997
The Hardware for Computational Chemistry?
presented at the IBM
Workshop on Trends and Directions for HPC RS/6000 SP systems, Daresbury, 10 July 1997
The Hardware for Computational Chemistry?
, presented at the
International Symposium on Supercomputing
n of Computational Science, Tokyo
(Japan), September 1
Benchmarking and Computational Chemistry
, presented at the 8th Distributed Computing
Machine Evaluation Workshop, Daresbury, November 11, 1997.
The Hardware for Computational Chemistry?
, presented at the
SARA HPCN Users Meeting, University of Amsterdam, November 12, 1997.
Parallel Computing: Vector MIMD Architectures
, presented at The Department of
Crystallography, Birkbeck College, 1
9 November, 1997.
High Performance Computational Chemistry; NWChem, HPC'97 and Massive
, presented at The Department of Chemistry, University of Sheffield, 3 December, 1997.
Continuum States of Atoms and Molecules
ent staff support was provided by Dr. C.J. Noble and Dr. R.J. Allan.
Working Group Meetings:
A CCP2 Working Group Meeting, was held at UCL on 10 January 1997.
None issued in reporting period.
and maintenance of a CCP2 Web page. Development and maintenance of a
web page for the (e,2e) sub
project and its international collaborators. The latter page is fully integrated into
the Departmental DataWeb system described in the CCP1 section.
The support staff contributed to the organisation of the following conferences
group workshop, Daresbury Laboratory, 3
7 April 1997.
Molecular scattering sub
group meeting, Belfast, 22 May 1997.
tering Conference, Nottingham, 16
18 July 1997.
A workshop entitled SPINUK was organised jointly with the Synchrotron Radiation Department to
focus the current advances of the UK theory and experimental community in spin
studies. This m
eeting was supported by the EPSRC and the Institute of Physics.
None during reporting period.
Handling of travel expenses for meetings.
Applications Development and Support
Development of Parallel Electron
Recent work on parallelizing this package is described in the Parallel Supercomputing section. The code is
currently being proven by performing calculations of electron scattering by Ni
ions. Several symmetry
values have been completed.
Work is also being carried out to determine whether complex energy
techniques can be used to reduce the number of scattering energies that must be calculated in order to map
the complicated resonance structure observed for iron
of Codes to Model Multiphoton Ionization of Atoms and Molecules
Construction of a new direct R
Floquet calculation is underway. The package is based on the use
the RM2 package which will enable calculations to be carried out on two
So far transition matrix elements have been added to RM2 and the integral generation code, RAD, has been
completely rewritten. The new package uses shifted
Legendre radial basis functions rather than model
scattering functions. This chan
ge corresponds to a shift from fixed to arbitrary functions. This has a number
of advantages but the most significant of these is that the Floquet calculations can be carried out entirely
within the velocity gauges (Buttle corrections become unnecessary).
The program is being generalized to
treat arbitrary (elliptic) laser polarizations.
Floquet calculations on the multiphoton ionization of Li
ions exploiting an earlier version of the
code have been completed in collaboration with the group at QUB
Development of Codes to Model Positron
A general computer program for treating positron
atom collisions in a one
electron model has been
developed. The Schrodinger equation is solved using a least squares method which permits a wi
of basis functions to be use. Studies have been carried out using Laguerre, Legendre, STO and oscillatory
basis functions. The use of shifted Legendre functions has been found to be particularly effective. The
approach has been checked by compar
ison with previously published 6
state results for positron
New results have been obtained for positronium formation in positron collisions with positive ions. The
results agree well with Coulomb
Born calculations at high scattering e
nergies but are in poor agreement
with optical model results published by Mellissa and Gianturco. Calculations including pseudostates are
now underway to determine to what extent the discrepancy depends on the model of the collision system.
assistance of rearrangement processes are currently underway. This work has
been carried out in collaboration with the Durham group.
Programs are distributed via the Computer Physics Communications Library.
Publications and Pr
D. Glass, P.G. Burke, H.W. van der Hart and C.J. Noble, R
Floquet Theory of Multiphoton
Processes VIII, A New Direct Method with an Application to Helium, J. Phys. B: At. Mol. Opt. Phys.
D. Glass, P.G. Burke, C.J. Noble
and G.B. Woeste, R
Floquet Theory of Multiphoton
Processes XIV, Multiphoton detachment of Li
, J. Phys. B: At. Mol. Opt. Phys. (1998).
A. G. Sunderland, P.G. Burke, V.M. Burke and C.J. Noble, Parallelization of Atomic R
Scattering Progams, to
be published in
High Performance Computing
, editors R.J. Allan, A. Simpson
and D.A. Nicole, (Plenum, 1998).
H.Y. Mussa, J. Tennyson, C.J. Noble and R.J. Allan, Molecular Rotation
Vibration Calculations using
Massively Parallel Computers, to be published i
High Performance Computing
, editors R.J. Allan, A.
Simpson and D.A. Nicole, (Plenum, 1998).
J.J.M. Michiels, J.E. Inglesfield, C.J. Noble, V.M. Burke and P.G. Burke, Spin and Symmetry in Low
energy Electron Energy
loss Spectroscopy of Transition Metal Ox
ides}, J. Phys. C: Cond. Matter
H.Y. Mussa, J. Tennyson, C.J. Noble and R.J. Allan, Rotation
Vibration Calculations using Massively
Parallel Computers, Comp. Phys. Commun.
A.G. Sunderland, J.W. Heggarty, C.J. Noble and N
.S. Scott, Parallelization of R
Methods on Distributed Memory Computers}, Comp. Phys. Commun., in press, (1998).
D.W. Busby, P.G. Burke, V.M. Burke, C.J. Noble and N.S. Scott, VisRes: a GRACE tool for
Displaying and Analysing Resonances}
, Comp. Phys. Commun., in press, (1998).
M.A. Hayes and C.J. Noble, Inner Valence Levels and Rydberg States of CO+, J. Phys. B: At. Mol.
Opt. Phys. (1998).
B.H. Bransden and C.J. Noble, Application of least squares methods to atomic rearrangement
s, submitted to J. Phys. B: At. Mol. Opt. Phys. (1998).
Permanent staff support was provided by V.R. Saunders (CRYSTAL development) and N.M. Harrison
Working Group Meetings:
A CCP3 w
orking Group meeting was held on 17 December 1997 at Daresbury
Newsletters were published in July 1997 and January 1998. A survey requesting feedback from
members on electronic distribution of the newsletter was included in the Janu
ary newsletter and an excellent
response (80% of membership) was returned. About 25% of users stated that they required a paper copy of
the newsletter. The working group meeting in December decided that in future electronic distribution via
the WWW pages a
mail would be used. Copies of the January issue and previous newsletters were
made available on the web
site. In order to enhance the visibility of CCP3 in the UK summaries of
newsletter contents were sent to members of the IOP’s Thin Films and Surfa
ces Group and the Gas Surface
News mailing lists. An electronic registration systems was provided on the web
site and some 50 extra
members have subscribed to CCP3.
The CCP3 web
site was transferred to the Data
Web system in March 1998. Elec
registration coupled to an ODBC database for the membership has been implemented.
At the working group meeting in December CCP3 asked Profs. Inglesfield and
Weightman to organise a workshop on the “Optical Response of Surfa
ces” during 1998.
Dr H. Tutuncu, Turkey, visited the Exeter Group for 4 weeks in November 1997 and
studied the vibrational properties of the AlSb (110) surface using the adiabatic bond charge method. This
work was presented at the IOP’s
The Working Group:
New members proposed and elected to the working group in this period were;
R. Godby (Physics, University of York), Dr S. Parker (Chemistry, University of Bath), Prof. R. Smith
(University of Loughborough), Prof. M
Finnis (Queen’s University Belfast) and Dr. John Blackman
(University of Reading). These members extended the expertise of CCP3 into the area of classical
simulations at surfaces and added to the existing interest in first principles simulations of surface
Travel and subsistence for working group meetings.
Applications Development and Support
A graphical user interface to CRYSTAL was developed by Searle and Harrison
and placed on beta releas
e in March 1998. The CCP3 groups in St. Andrews, Cambridge and at
Daresbury tested the interface in surface studies of semiconductors and oxides. The interface supports
the visualisation and editing of surface geometry, the calculation of surface energies,
and of the
electronic and magnetic structure of the surface. Charge densities and electrostatic potentials can be
displayed using 3D graphics. After feedback from beta
testers the interface will be refined and released
with CRYSTAL98. It is expected that
the interface will play a key role in allowing non
in academe and industry to take advantage of the CRYSTAL software.
The CCP3 flagship project (GR/K90661) concerns the development of analytic gradients (forces) in the
CRYSTAL software. This
will allow the CCP3 community to perform rapid structural optimisation of
complex surface structures. The detailed formalism and implementation strategy was developed by
Saunders and Harrison during the period from June
October. Implementation of the sche
commenced in November with the arrival of Dr. K. Doll the PDRA on this project. Analytic gradients
of the mono
electronic integrals (kinetic, nuclear attraction) have been completed for 0, 1 2 and 3
The newly developed hybrid
sity functional theory has been demonstrated to provide more accurate
ground state energies, magnetic couplings and band gaps than either density functional or Hartree
formalisms. Saunders implemented a very flexible scheme for arbitrary hybridisation
, including the
popular B3LYP formulation, which was placed on beta
release in November. An immediate
application to studies of alkali metal adsorption on TiO
by the Manchester group yielded surface
geometries and electronic structure in excellent agreeme
nt with photoelectron spectroscopies. This
formalism will be released in CRYSTAL98.
The Surface Science Shell
Harrison has supervised the development of a graphical user interface to surface science codes within a
collaboration with the synchrotron radiati
on department at Daresbury and CCP3. The interface is based on
the AVS5.0 environment that is available to UK academics on a wide variety of computer platforms at
relatively low cost and is intended to provide a flexible, low cost alternative to the commer
interface. Support for surface editing and visualization along with interfaces to PHOTON, XANES and
CRYSTAL is provided. The package has been made freely available via the CCP3 web pages and updated
on a regular basis. In the period to March f
ifteen UK groups downloaded the package and provided
feedback on its use.
Electronic and physical structure of oxide surfaces
Mr J. Muscat is a CASE QUOTA student (Ref: 95563789) registered with Prof. G. Thornton, Chemistry,
University of Manchester and s
ponsored by Daresbury. He has worked at Daresbury during this period
under the supervision of Dr N.M. Harrison. He has used CRYSTAL during this period to perform studies of
the geometrical and electronic structure of oxide surfaces and helped to interpret
experiments performed on
the SRS and in Manchester.
Oxide Surface Chemistry
Dr. P. Lindan worked at Daresbury as a PDRA funded by a EPSRC under the CRAY sponsorship scheme
until 1/10/97. This collaborative project between CCP3 and CCP5 was co
N.M. Harrison at
Daresbury and Prof. M. Gillan at the University of Keele. He used plane wave density functional methods
to perform studies of chemical interactions at oxide surfaces.
Strongly Interacting Systems
The exact exchange interaction implemente
d in the CRYSTAL software allows first principles theory to be
applied to strongly interacting systems. This extension of the applicability of such methods will allow
studies of the surface structure and chemistry of transition metal insulators to be studi
ed. A number of
studies demonstrating this and exploring the accuracy of the new formalism have been conducted by N.M.
Harrison in collaboration with the CCP3 groups in Cambridge, St
Andrews, UMIST and the Royal
During this pe
riod the CCP3 working group considered the usage and utility of the extensive program
library in detail and decided to restructure the library in order to focus support on the key codes which find
wide use in the community. A new category of “unsupported
package” was introduced. The library now
contains the following supported codes ( N/M
Active UK Usage (Groups) March 98
DLVisualizer (nee SS
Apart from DLV the codes are all written in standard Fortran and are easily ported to a wide variety of
platforms. Source code and compilation tools are provided with the package. DLV is currently limited to
s which support AVS 5.0 which include the major Unix workstations (IBM, SGI, DEC,
Support and Documentation
The CRYSTAL package has extensive documentation and support if provided via web
pages and an E
support line ‘email@example.com’. The te
nsor LEED package is documented and supported by Dr A Wander
in the Cambridge group.
Publications and Presentations
Lindsay, P.L. Wincott, G. Thornton, N.M. Harrison “The Electronic Structure of Si(100)2x1
reinterpreting ARP measurements”,
J. Muscat, N.M. Harrison and G. Thornton “The Effects of Exchange Correlation and Numerical
Approximations on the Computed Properties of the Rutile 100 Surface”, Phys. Rev. B (In Press).
P.J.D. Lindan, N.M. Harrison, M.J. Gillan, “Mix
ed dissociative and molecular adsorption of water on
the rutile (110) surface”
Phys. Rev. Lett.
P.J.D. Lindan, J. Muscat, S.vBates, N.M. Harrison, M. Gillan “The Ab Initio Simulation of Molecular
Processes at Oxide Surfaces”
Soc. Faraday Trans.
P.J.D. Lindan, N.M. Harrison, M.J. Gillan, J. A. White “First
polarised calculations on
the reduced and reconstructed TiO
Phys. Rev. B
C. Rehbein, F.
Michel, N.M. Harrison, A. Wander “Ab initio total energy studies of the alpha Cr
(0001) and (01
Surf. Rev. Let.
W.C. Mackrodt, E.A. Simson, N.M. Harrison “An ab initio Hartree
Fock study of the electron
gap states in oxygen
deficient rutile TiO
3) p.192 (1997).
F. Cora, A. Patel, N.M. Harrison, C. Roetti, C.R.A. Catlow “An ab initio Hartree
Fock study of alpha
J. Mater. Chem.
N.M. Harrison, V.R. Saunders, R.
Dovesi, W.C. Mackrodt “Transition Metal Materials: A First
Principles Approach to the Electronic Structure of the Insulating Phase”,
Transition in Macroscopic and Microscopic Systems
87 P.P. Edwards, CNR Rao, RL
n, D.P. Tunstall (Eds), Phil. Trans. A., (1997).
N.M. Harrison “Electronic Structure Methods” C.R.A. Catlow (Eds
), Computer Modelling in Inorganic
, Academic Press, 1997.
N.M. Harrison, B.G. Searle, E.A. Seddon “An Ab Initio Study of the
Magnetic Coupling in Bi
Chem. Phys. Lett.
D.R. Collins, W. Smith, N.M. Harrison, T.R. Forrester “Molecular dynamics study of the high
temperature fusion of TiO2 nanoclusters”
J. Mat. Chem.
(12) p.2543 (1997)
Exact Exchange Potential In Magnetically Ordered Insulators, 9
International Conference on
Quantum Chemistry, Atlanta, June 9
14 1997. N.M. Harrison
Oxide Surface Chemistry from First Principles, Surface Science User Group Meeting, Daresbury, 9
ptember, 1997, N.M. Harrison
Excited States at Surfaces and Hybrid Exchange Approximations, Seminar, University of Oxford, 20
Jan, 1998, N.M. Harrison
On the Best Use of a Synchrotron Through Computer Simulation, CLRC Surface Science Meeting,
Jan 1998, N.M. Harrison
Computational Physics in Practice, Meeting of the Computational Physics Advisory Committee,
University of Salford, 3
March, 1998, N.M. Harrison
A Commercial Interface to CRYSTAL98, Seminar, Turin, 24
March 1998, N.M. Harri
MPP Progress and Prospects, Seminar, Milan, 25
March 1998, N.M. Harrison
Fast Structural Optimisation in CRYSTAL98, Seminar, Bologna, 26
March 1998, N.M. Harrison and
Water Chemistry at Oxide Surfaces, Surface Reactivity Meeting of
the Mineralogical Society,
April 1998, N.M. Harrison, J. Muscat, P.J. Lindan
Current Progress on the First Principles Simulation of Surfaces, Surface Science User Group Meeting,
April 1998, N.M. Harrison
Computer Simulations of Condensed Phases
Permanent Daresbury staff support was provided by Drs W. Smith and M. Leslie.
Working Group Meetings:
The CCP5 Executive Committee met on 21/3/97, 17/9/97, 28/1/98 and
19/3/98. A meeting
of the Steering Panel, open to all registered CCP5 members, was held on 16/9/97.
Maintenance of CCP5 electronic mailing lists.
The CCP5 web page (www.dl.ac.uk/|CCP/CCP5/main.html) and the electronic CCP5
k/CCP/CCP5/newsletter_index.html) and its associated archive of back
(ftp://ftp.dl.ac.uk/ccp5.newsletter) is maintained at Daresbury Laboratory by M. Leslie.
The DL_POLY web page (www.dl.ac.uk/TCS/Software/DL_POLY/main.html) and the associated
entation (accessible through the web page) is also maintained at Daresbury and is managed by W.
Two workshops took place in the reporting period: “Molecular modelling and
Neutron scattering: techniques for large scale systems
” (Cosener’s House, 2/4/97) and “Liquid crystals”
(Sheffield Hallam, 26/11/97). The CCP5 Annual Meeting was on the subject “Materials, Microstructures
and Simulation” (UCL, 15
17/9/97). CCP5 sponsored the Spring School in March 1997 at Bristol. The
chools are aimed at postgraduates commencing studies in molecular simulation.
CCP5 invited two overseas visitors in 1997:
Dr. F. Mueller
Plathe (MPI Polymerforschung, Mainz, Germany) visited Cambridge, Daresbury and the
on (June 19997)
Prof. D. Rapaport (Bar
Ilan University, Israel) visited Edinburgh, Daresbury, UMIST and Reading (August
Travel and subsistence for committee members and visiting speakers
Applications Development and Support
most important code under development at Daresbury is DL_POLY, however some effort is also
directed to supporting the lattice code THBREL, which is designed to assist development of accurate
DL_POLY is a general purpose paral
lel molecular dynamics simulation code developed by Daresbury
Laboratory to support CCP5's scientific work. 1997 saw its adoption by the Materials Consortium under
Prof. Catlow at the Royal Institution, in which it was used to look at the diffusion of hydr
zeolites. The code is developed, documented and managed by W. Smith.
The main technical development for DL_POLY in 1997 was the Graphical User Interface, which is based
on the CERIUS 2 visualiser from Molecular Simulations Inc. The GUI was succ
essfully employed in the
CCP5 Spring School at Bristol in March 1998, where it was used to train students in molecular dynamics. A
series of training exercises and a user manual was written to accompany the exercises. Work was also done
(by Dr. Leslie) to
incorporate multipolar electrostatics.
In 1997 we took delivery of a pre
release version of the DLPROTEIN code, which was developed from
DL_POLY by the university of Rome. Its release to CCP5 was scheduled for summer 1998.
The DL_POLY code was exploited in
a number of collaborative scientific studies which were published in
benzene diffusion in silicalite
antibiotic activity of valinomycin with J.H.R. Clarke (UMIST);
the structure of glasses with G.N. Greaves (Wales) and M.J.
the structure of zeolites with D.A. Faux (Surrey) ; and,
the properties of titanium dioxide nanoclusters .
Dr. Leslie collaborated with Dr. Ji
Chen Li and student C. Burnham at UMIST in a CASE award study of
in water [1,2]. Dr. Smith collaborated with Prof. R. Lynden
Bell and student P. Smith
(Queen's Belfast) in a CAST award study of polymers at interfaces.
Dr. Smith submitted a proposal to EPSRC with J. Penfold (RAL), D. Heyes (Surrey) and D. Barlow (King's
College, London) to initiate a collaborative project on surfactants in solution. The project was unfunded.
Dr. Smith also submitted a proposal with D. Heyes (Surrey), L.V. Woodcock (Bradford), ICI Technology,
BNFL, ECC and Oxford Materials to the Foresight
Link scheme to model powders. The project passed the
first round in December 1997. DL_POLY was intended to be an integral component of these projects.
THBREL is a static lattice, energy minimization program designed to compute the structure and mec
properties of crystalline materials. The main technical development in 1997
98 was the introduction of a
new polarisation model incorporating anisotropic shell relaxation. It was applied to magnesium oxide.
Dr. Leslie submitted two proposals with D
r S. Price (UCL) to use the code THBREL to model molecular
crystals. The first proposal (to EPSRC) was to determine the structure of molecular crystals and the second
(to Zeneca) was to model pharmaceuticals. Both were successful and due to start in Decemb
The CCP5 Program Library contains 70 codes, most of which are molecular dynamics codes. Almost all are
written in FORTRAN. There are also some 20 educational codes from the book Computer Simulation of
Liquids, by M.P. Allen and D.J
. Tildesley. The codes are easily ported to all machine architectures (except
parallel) and the bulk are available over the internet. The exceptions are the code DL_POLY, which is free
but requires a licence, and the static lattice codes (THBREL, CASCADE e
tc.) which are also licenced.
Two new releases of the code appeared in 1997 (versions 2.8 and 2.9) with two additional bug
fixes. A total
of 273 licences had been registered for the code up to March 30 1998. Training in the use of DL_POLY is
now a feature
of the CCP5 Schools. Students gain hands
on experience using the code in actual
applications, as well as learning about the underlying algorithms.
Documentation for some of the older codes in the program library are available as hard copy from
boratory. Detailed documentation for DL_POLY is available over the web. Advice or
assistance in the use of these programs may be requested by e
mail or by telephone.
Publications and Presentations
J.C. Li and M. Leslie, Some basic properties of ice dynami
cs, J. Phys. Chem. B 101 (1997) 6304
C.J. Burnham, J.C. Li and M. Leslie, Molecular dynamics calculations for ice Ih, J. Phys. Chem. B 101
T.R. Forester and W. Smith, Blue moon simulations of benzene in silicalite
1, Faraday Trans.
T.R. Forester, W. Smith and J.H.R. Clarke, Antibiotic activity of valinomycin, Faraday Trans. 93
W. Smith, T.R. Forester, G.N. Greaves, S. Hayter and M.J. Gillan, Molecular dynamics simulation of
alkali metal diffusion i
n alkali metal disilicate glasses, J. Matter. Chem. 7 (1997) 331
G.N. Greaves, W. Smith, E. Guilotto and E. Pantos, Local structure, microstructure and glass
properties, J. Noncryst. Solids 222 (1997) 13
D.A. Faux, W. Smith and T.R. Forester, Mole
cular dynamics studies of hydrated and dehydrated Na+
4A, J. Phys. Chem. B 101 (1997) 1762
D.R. Collins, W. Smith, N.M. Harrison and T.R. Forester, Molecular dynamics study of high
temperature fusion of TiO
nanoclusters, J. Matter. Chem. 7 (
M.R. Wilson, M.P. Allen, M.A. Warren, A. Sauron and W. Smith, Molecular dynamics calculation of
elastic constants in Gay
Berne nematic liquid crystals, J. Comput. Chem. 18 (1997) 478
T.R. Forester and W. Smith, SHAKE, RATTLE and Roll: Effi
cient algorithms for rigid body
dynamics, J. Comput. Chem. submitted (1997).
W. Smith, Alkali diffusion in silicate glasses, presented at the CCP5 Annual Meeting September 1997.
W. Smith, Parallel algorithms for molecular dynamics, CCP5 Spring School Marc
Heavy Particle Dynamics
Permanent Daresbury staff support was provided by Dr. R.J. Allan.
Working Group Meetings:
These are held approximately twice per year at University College London.
They are supplemented by
meetings held at conferences thereby reducing the cost of travel.
Produced annually. Issue 20 in January 1998. Contents of recent issues are on the Web.
Web pages are maintained at CLRC on behalf of CCP6. These were created by R
contributions from the working group. URL=http://www.dci.clrc.ac.uk/Activity/CCP6.
CCP6 contributes to running a number of specialist workshops and also
international conferences. Events during the current reporting period
Nonlinear Dynamics and Spectra of Molecules, University of Warwick, 18
John Murrell 65th Birthday Meeting, University of Sussex, 5/12/97.
Joint CCP6/High Resolution Spectroscopy Meeting, University of Exeter, 17
ning a Model: Optimization methods in Chemical Physics, University of Durham, 24
CCP6 publishes booklets (via Daresbury) following its specialist workshops. Last
year this included: "Fashioning a Model: Optimization methods
in chemical physics" ed. A.Ernesti,
J.M.Hutson and N.J.Wright, (CCP6, Daresbury Laboratory, 1998) ISBN 0
Dr. S.K. Gray, Argonne National Laboratory.
CCP6 travel funds are maintained by Dr. C.J. N
oble at Daresbury
Applications Development and Support
Assistance in the management of the ChemReact HPCI Consortium. Collaboration w
ith Prof. G.G. Balint
Kurti (Bristol) to develop a new version of the parallel TDRS codes based on a new methodology using
purely real wavepacket. This methodology was developed during a visit to Bristol of Dr. S.K. Gray
(Argonne) in 1997
A distributed program library is maintained via the Web pages. Individual packages are kept on servers at
their home i
NoLLS and MOLSCAT are maintained via the ChemReact HPCI Consortium.
Publications and Presentations
H.Y.Mussa, J.Tennyson, C.J.Noble and R.J.Allan, Rotation
vibration calculations using massively
parallel computers, Comp. Phys. Comm. 108: 29
Gabriel G. Balint
Kurti, Stephen K. Gray, Evelyn M. Goldfield and Robert J. Allan, Time
Quantum Theory of Reactive Scattering, Faraday Discussion (1998) submitted.
R.J.Allan and G.G.Balint
Kurti, HPCI Grand Challenge Consortium
p article for the CCP6
Newsletter, Jan'98 pp4
Electronic Structure of Solids
Permanent Daresbury staff support was provided by Dr. W.M. Temmerman and Dr. Z. Szotek
Working Group Meetings:
held on December 22, 1998
k Newsletters published in February, April, June, August, October and December
installed on the Web at ://www.dl.ac.uk/TCSC/HCM/PSIK/main.html. Maintenance of mailing list.
development and maintenance of web pages for CC
K and Interface Magnetism
On behalf of CCP9: Co
ordinator of the TMR Network: Interface Magnetism (1/9/1996
This Network provides CCP9 with an extra staff year for the development and application of
magnetism codes. It also allows CCP9 to participate in the Network of the most outstanding European
groups in the field of quantum mechanical calculations for magnetic properties of solids. The annual
meeting in Vienna June 5
7,1998 was attended by 81 par
ticipants, of which 15% came from the UK.
On behalf of CCP9: Node leader in TMR Network: Electronic structure calculations of materials
properties and processes for industry and basic sciences (1/3/1998
28/2/2002). The two UK CCP9
nodes are Belfast and Da
resbury. Daresbury’s role is disseminator of band theory knowledge. This
happens through the Psi
k Newsletter, hands
on courses and workshops. ELECTRONIC STRUCTURE
CALCULATIONS FOR INDUSTRY AND BASIC SCIENCES, Vienna June 3
4 1998. Organized by
W. Temmerman, E. Wimmer and A. Svane; MATERIALS UNDER PRESSURE: FROM
ICES TO METALS, Daresbury Laboratory (UK) 17
18 December 1998. Organized by Richard
Nelmes and Walter Temmerman.
On behalf of CCP9: Secretary to the ESF Programme on Electronic Structure
Elucidating the Complex Atomistic Behaviour of Solids and Surfaces (1/1/1998
programme funds CCP9’s collaborations with the rest of Europe and provides CCP9 with funds for the
organisation of workshops and the psi
CCP9 supports only UK based collaborative visits. All foreign collaborative visits
are supported by ESF and TMR.
The finances of CCP9, ESF and TMR are managed.
Applications Development and Support
Support for the CCP9 Flagship Project on Surface Magnetism: The CCP9 RA is based in Bristol and is
applying surface magnetism codes to the calculation of the Bloch wall thickness in Co. At Daresbury,
complementary LMTO and KKR codes have al
so been developed. In particular, Daresbury is developing a
layer KKR and photoemission code for the study of quantum well states in magnetic multilayers and for
magnetostriction studies, and a real space KKR code using sparse matrix technology (in collabo
consistent layer KKR code has been developed and is now being tested. This code will provide us
with the potentials needed in the photoemission code. This layer KKR and photoemission codes are now
being tested by groups in
Cambridge and Keele in a study of the photocurrent on the different faces of Cu.
A study of the convergence of the magnetic properties of Fe with the screened real space multiple scattering
code is also underway. This study has shown that one needs to con
sider more than 500 atoms to converge
the magnetic and orbital moments of Fe. Fortunately, these calculations are feasible with the screened KKR
which gives rise to sparse matrices. Utilising the sparse matrix packages of the Harwell Subroutine Library,
e can run in excess of 1,000 atoms on a DEC EV5 500 MHz workstation. This work is part of a
collaboration with the group at Sheffield.
Interaction Correction Code
Adding new functionality to the Self
support of (a) the Keele work on valency in rare
earth compounds; (b) the Sheffield work on Colossal
Magnetoresistance materials and the Bristol work on high T
materials. These codes are being exploited in a
number of projects.
of the magnetic and electronic properties of rare
earth monochalcogenides and
ytterbium monopnictides. The short conclusion of this study is that we can identify the number of
electrons with the valency in these compounds. Two manuscripts are i
LSD study of the magnetic and electronic properties of YBa
. This study
is in the process of being written up. The main conclusions are that, for the first time, an ab initio
electronic structure calculation can des
cribe the anti
ferromagnetic ground state of YBa2Cu3O
upon oxygenation the magnetic energy gain gradually decreases and that for YBa
magnetic states are very close in energy. We are now studying the implications of this o
magnetic susceptibility data in collaboration with Steve Hayden (Bristol).
de Gennes Codes
Support for the CCP9 Project on Bogoliubov
de Gennes: development of all the codes to solve the
de Gennes equations, to calculate the sp
ecific heat and the penetration depth of high T
materials. This work is undertaken in collaboration with the group at Bristol.
The development of TB
de Gennes formalism. In this formalism, the variation of T
versus doping, and the speci
fic heat in the superconducting state were studied. Several papers were
published on this subject. The aim of these studies is to determine which electrons, on which sites and on
which orbitals are responsible for the pairing. The purpose is to make contac
t with as many experiments as
possible and therefore our next study is on the penetration depth. This work is done in collaboration with
the group at Bristol.
Support of the UK’s experimental work on magneto circular dichroism on the SR
S: adding new
functionality to the existing codes, for example, the PBE formalism was implemented. The FP
John Wills (Los Alamos) and a real space KKR dichroism code from Hubert Ebert (Muenchen) were
imported. This was needed for calculations on Gd
/Mo multilayers to investigate some very surprising
results obtained by the Birkbeck group on the SRS.
Effort was also put in to incorporating an improved treatment for electron
positron correlations in electron
positron momentum density codes. These impro
vements are necessary for the analysis of the experimental
positron annihilation data obtained for example in Bristol.
A suite of LMTO and KKR based codes is distributed. Formal additional support is not provided.
Publications and Presenta
B.L. Gyorffy, Z. Szotek, W.M. Temmerman, O.K. Andersen and O. Jepsen, “On the Quasi
Spectra of High Temperature Superconductors”, Phys. Rev. B 58 (1998) 1025
S.L. Dudarev, G.A. Botton, S.Y. Savrasov, Z. Szotek, W.M. Temmerman and A.P. S
structure and elastic properties of strongly correlated metal oxides from first principles: LSDA+U, SIC
LSDA and EELS study of UO
and NiO”, phys. stat. sol. (a) 166 (1998) 429.
A. Svane, Z. Szotek, W. M. Temmerman, J. Lægsgaard and H.
Winter, “Electronic Structure of Cerium
Pnictides under Pressure”, J. Phys.: Condens. Matter 10 (1998) 5309
Z. Szotek, B.L. Gyorffy, W.M. Temmerman and O.K. Andersen, “The Van Hove Scenario and the
Band’ Model for High T
s. Rev. B 58 (1998) 522
Rubaszek, Z. Szotek and W. M. Temmerman, “Non
Positron Correlations in Solids
within the Weighted Density Approximation”, Phys. Rev. B 58 (1998) 11285 (1998)
S. V. Beiden, W. M. Temmerman, Z. Szotek, G. A. Gehri
ng, G. M. Stocks, Yang Wang, D. M. C.
Nicholson, W. A. Shelton and H. Ebert, “Real Space Approach to the calculation of Magnetocrystalline
Anisotropy in Metals”, Phys. Rev. B 57 (1998) 14247
W.M. Temmerman, Z. Szotek, A.C. Jenkins, A. Svane, H. Winter, S.V
. Beiden and G.A. Gehring, “An
Unified Bandstructure Description of Localized and Delocalized F States”, accepted for publication in
the proceedings of 190. WE
Seminar on Magnetism and Electronic Correlations in Local
Moment Systems: Rare
ements and Compounds, Berlin, 15
18 March 1998
A.C. Jenkins and W.M. Temmerman, “On the nature of the magnetic coupling at the Gd (0001)
surface”, submitted to MML’98, Vancouver 15
A.C. Jenkins and W.M. Temmerman, “Calculated properties o
f Gadolinium/Molybdenum multilayers”,
submitted to MML’98, Vancouver 15
Ernst, W.M. Temmerman, Z. Szotek, M. Wood and P.J. Durham, “Real
photoemission”, submitted to Second Annual Meeting of TMR Network “Interface M
L. Petit, S.V. Beiden, W.M. Temmerman, Z. Szotek, G.M. Stocks and G.A. Gehring, “Screened Real
Space Study of the Magnetic Properties of Ni”, submitted to Second Annual Meeting of TMR Network
“Interface Magnetism” Vienn
a June 5
A.C. Jenkins, W.M. Temmerman, V. Popescu and H. Ebert, “Magnetic Circular Dichroism in Gd
Metal: The L
edges”, submitted to Second Annual Meeting of TMR Network “Interface Magnetism”
Vienna June 5
Rubaszek, Z. Szotek and
W. M. Temmerman, “Non
Positron Enhancement Factors in
Solids”, submitted for publication to Acta Physica Polonica (1998).
A. Svane, W. M. Temmerman and Z. Szotek, “Theory Of Pressure Induced Phase Transitions in
Cerium Chalcogenides”, submi
tted for publication to Phys. Rev. B.
W.M. Temmerman, A.Svane, Z. Szotek and H. Winter, “Applications of Self
Interaction Corrections to
Localized States in Solids”, in "Electronic Density Functional Theory: Recent Progress and New
Directions" Eds J. F. D
obson, G. Vignale and M. P. Das, Plenum, NY (1998) ISBN 0
Z. Szotek, W.M. Temmerman, A.Svane, H. Winter, S.V. Beiden, G.A. Gehring, S.L. Dudarev and A.P.
Sutton, “Applications of Self
Interaction Corrections to Localized States in Solids”,
W. M. Temmerman, B. L. Gyorffy, Z. Szotek, O. K. Anderson and O. Jepsen, “On the Quasi
Spectra of YBa
”, accepted, Plenum Press (1998)
W.M. Temmerman, A.Svane, Z. Szotek, H. Winter and S. Beiden, “On the implem
entation of the Self
Interaction Corrected Local Spin Density Approximation for d
electron systems”, Springer
Verlag (1998) Submitted.
HPCI Conference, Manchester 12
14 January 1988: ‘On the quasi
particle spectra of high
Seminar at the Physics Department of University of Sheffield, February 16: ‘From Quantum Mechanics
to Designing Materials on a Workstation’.
CLRC Magnetism Workshop, Daresbury Laboratory, February 24
25: ‘The magnetic properties of Ce
Seminar on Magnetism and Electronic Correlations in Local
Earth Elements and Compounds, Berlin, 15
18 March: ‘A Unified Bandstructure Description of
Localized and Delocalized f States’.
Theoretical Magnetism 1998 (O
rganised by the Magnetism Group of the Institute of Physics), Oxford
July 6: ‘A Unified Bandstructure Description of Localized and Delocalized f States’.
Mini Symposium on “Gadolinium and its f’s”, Uppsala, August 19
21: ‘The SIC method and magnetic
ing in Gd’.
LMTO Meeting, Mont St. Odile, 2
5 October: ‘An Unified Bandstructure Description of
Localized and Delocalized States’.
Seminar at IFF in Juelich, November 16: ‘An Unified Bandstructure Description of Localized and
Bristol Workshop on the Bogoliubov
de Gennes Equations, Bristol, November 20
Interaction Corrected LDA for YBCO’.
High Performance Computing in Engineering
Permanent Daresbury staff support was provided by Drs. D.R. Emerson,
Y.F. Hu and R.J. Blake.
Working Group Meetings:
CCP12 Steering Group meetings were held on 6/6/1997 and 14/11/1997. The
combustion consortium met on 29/7/1997. The Marine CFD consortium met on 2/10/1997. Meeting on
12/1/1998 to cons
ider formation of a consortium in the area of flutter.
none (a number of articles were submitted to HPCProfile).
a brief section updating CCP12 programme and sections on the engineering consortia.
AC Course on High Performance Parallel Computing for Fluid Dynamic Applications
16 May, 1997, Daresbury Laboratory (18 attendees, 9 from UK, 4 non
Parallel CFD ’97, 19
21 May, 1997, Manchester (150 attendees)
This conference provided an ideal
opportunity to showcase much of the CFD work that was being done on
the Cray T3D. As the HPCE project organised the event it was possible to have keynote invited speakers
from the CCP12 community, which included Prof. M.A. Leschziner, Prof. K. Morgan and
Sandham. A number of other UK researchers also took the opportunity to present a range of high quality
papers. The conference provided an ideal forum to promote the UK’s work on parallel CFD. The
proceedings have now been published by Elsevier and
copies have been sent to appropriate Programme
Managers at EPSRC.
The HPCE project co
ordinates the ERCOFTAC Special Interest Group (SIG) 19 on the topic of
Parallel CFD. In this capacity, two important themes from parallel CFD are being organised for
COMAS98 including a session on Dynamic Load Balancing and a Special Technological Session on
Industrial Design Using Parallel CFD. The session on Dynamic Load Balancing is a follow
up of a
very successful meeting held at Daresbury Laboratory on 27/11/1995.
The HPCE project will be
submitting papers in each session.
CFD Theme Day:
Dr. Emerson was invited to be a panel member to assess the current state of CFD
in the UK today. The CFD Theme Day was held on the 18
March, 1998. The panel identified that
fficient HPC resources were being made available to a number of key researchers and that Large
Eddy Simulation will play a important role in CFD applications in the future. These views were
endorsed by an international expert. The emerging LES consortium w
ill aim to be the corner stone of
the UK activity and the HPCE project will be fully involved in making this successful.
None in reporting period.
Travel and Subsistence.
Applications Development and Support
nd Challenge Consortia
The HPCE (CCP12) project played a key role in initiating the four
to exploit the Cray T3D and has been extensively involved in supporting these consortia in the development
of parallel applicatio
ns for massively parallel architectures and particularly the Cray T3D and T3E
machines at EPCC. The work has involved:
developing optimised iterative routines, such as conjugate gradient and multigrid algorithms, and
optimised communication routines using
asynchronous message passing;
developing portable versions of codes within PVM and MPI; and,
investigating the performance of a variety of solvers on a number of distributed memory and shared
memory architectures, including novel cluster systems such as
the Beowulf machine at Daresbury, so
as to provide accurate advice to the academic community.
The HPCE project has also sought to broaden the applications base and assisted in the establishment of a
consortium looking at Marine CFD problems. The consort
ium agreed to work with CCP12 and submit a
proposal for pump
priming time on the Cray T3D.
Benchmark CFD Applications
The HPCE project has developed representative CFD benchmark codes to assist in the evaluation of any
machines under consideration in a pr
ocurement. The codes contain algorithms and methodologies that are in
common use within the engineering community. The applications have been developed to be robust (tested
on many architectures), portable (using MPI), highly optimised (using BLAS librarie
s) and easy to use by
vendors. They have been selected to fully test any machine’s, vector or cache
based, capability. The codes
: The production code is used for the direct numerical simulation of turbulent combustion. It uses a
e formulation and employs highly optimised iterative and direct solvers. These include a
conjugate gradient solver written in level 1 BLAS, a multigrid solver and a FFT solver. The conjugate
gradient routine will test a machine’s floating point performance
and also its global and near neighbour
communications. The multigrid solver will severely test a machine’s communication performance and
particularly its latency. The code was selected for use in the HPC97 procurement and has recently been
used to assess
the performance of Beowulf
class systems and other cluster platforms.
: The code solves the Euler equations for flow over arbitrary geometries. It uses a finite element
approach with unstructured (tetrahedral) meshes. Due to the connectivity of the m
eshes, it is necessary to
use indirect addressing for the elements. This places significant demands on the memory bandwidth of any
architecture. The code is very representative of applications using the finite element method in fluids,
CCP12 does not maintain and distribute a program library in the way that many of the other CCP projects
do (e.g. CCP5’s DL
POLY). This situation arises because many of the academic groups that CCP12 works
with have a
very close relationship with industry and much of their research is supported by industry. Much
of the work is therefore done
. However, any algorithms developed and the experience and
expertise gained from these collaborations is readily tran
sferable to other areas.
Publications and Presentations
Introduction to Parallel Computers: Architecture and Algorithms, D.R. Emerson, High Performance
Parallel Computing for Fluid Dynamic Applications, ERCOFTAC course, Daresbury Laboratory, May
High Performance Parallel Computing for Fluid Dynamic Applications, D.R. Emerson, Report
published in ERCOFTAC Bulletin 34.
Parallel CFD ’97, Manchester, May 19
21, 1997, Editor of conference proceedings.
Introduction to Parallel Computing Using Adv
anced Architectures and Algorithms, D.R. Emerson
NATO course, September 10
12, Istanbul, Turkey.
An Optimal Migration Algorithm for Dynamic Load Balancing, Y.F. Hu, R.J. Blake and D.R.
Emerson, Concurrency: Practice and Experience, V10(6), 467
Chemical Process Simulation on Parallel Computers, R.J. Blake and D.R. Emerson, Proceedings of
The Optimal Property of Polynomial Based Diffusion
Like Algorithms in Dynamic Load Balancing,
Y.F. Hu and R.J. Blake, Proceedings of ECCOMAS 98.
Evaluation of Cost Effective Parallel Computers for CFD
D.R. Emerson, K. Maguire, K. Takeda and D. Nicole, Submitted to Parallel CFD98.
Computational Engineering in the UK, D. R. Emerson, Presentation to Science & Technology Agency,
9, Tokyo, Jap
Parallel Computing in CFD, D.R. Emerson, Presentation on activities of ERCOFTAC Special Interest
Parallel Supercomputing Activities
In this section we describe work undertaken over the past year aimed at providing ope
rational support for
the IBM SP2 system and the extensive scientific support for EPSRC users of parallel supercomputing
facilities. The latter includes the development of new very high performance scientific applications and
assistance in porting and optim
ising codes and developing parallel programming tools.
Parallel Computing Service
There have been a number of changes in the parallel supercomputing resources offered by the Programme.
First, the long serving Intel iPSC/860 was finally switched off at th
e end of 1997, but most of the activities
were wound up gradually throughout the year, and so in the end the event passed off rather quietly. Second,
the IBM SP/2 was upgraded to 26 nodes in July 1997 involving a number of modifications to the
n of the system as summarised below:
the 14 Power2 Thin Nodes (66 MHz) on the old system were replaced by 14 Power2 Superchip (P2SC,
120 MHz) nodes;
10 new P2SC nodes were added;
the 2 Power2 Wide Nodes were kept for the login pool (see below); and,
e High Performance Switches (and adapters) were replaced by the SP Switches.
In order to support a wide range of applications with various requirements and computing activities, the 26
nodes are divided into 3 Pools as briefly described below:
this consists of the two wide nodes. Apart from being a gateway to the SP system, this also
provides a platform for users to prepare, submit and monitor their jobs.
this consists of 4 P2SC nodes. The word ‘shared’ here means that the no
des in this pool
may be shared by more than one job at any one time. Apart from allowing more users to access the
machine, this also makes users feel that their jobs are executed immediately (which most of the users are
used to on a personal workstation).
This facility is particularly useful for users who are
developing/testing new applications, requiring very fast turnaround of their jobs.
this consists of the remaining 20 P2SC nodes. Most of the parallel applications are
designed to maxim
ise the performance of the applications on the underlying architecture by balancing
the workload evenly across all of the nodes. This assumes dedicated access to the resources and so we
have to make the necessary provision for this computational requiremen
t. Although it is possible to run
any size of job on the machine (up to a maximum limit that is presently 16), the system is best deployed
with a mix of 4, 8 and 16 processor jobs. A scheduling system, developed locally, is used to ensure that
an optimal m
ix is maintained.
It is no surprise that the SP is very heavily used by computational scientists and engineers at the Laboratory
working on the various CCP projects. However, we have also found that a number of consortia, which have
access to the T3D/T3E
national facilities, are very active on the machine. The computational activities of
the CCPs are described elsewhere is this report, the work of some of these external consortia is described
briefly in Appendix 2.
Evaluation and Development of Hardware
and Software Technologies
Within this section we outline activities undertaken over the past year targeted at:
the development of a library of core parallel algorithms that capture recent developments within the
HPCI Consortia in the form of robust, porta
ble and efficient routines and provide a foundation for
expanding the base of HPC applications; and,
evaluating migration and performance optimisation strategies for the next generation of HPC systems
based on distributed memory and coupled shared memory
Basic research on parallel algorithms is carried out for routines of generic use in a number of applications in
a series of collaborations between the HPCI Centre and applications specialists. We have developed a
number of parallel FFT library modules that are being evaluated within key applications codes including:
CETEP, DL_POLY, wavepacket simulations and spectral engineering calculations. Eigensolver packages
are currently being evaluated within the DVR3D, GA
UK and R matrix propagator packages. In
order to make available results of work carried out over the last few years, parallel algorithms which have
been developed for specific purposes (e.g. iterative eigensolvers, FFTs, non
linear optimisation and li
equation solvers) will be combined to produce a CLRC library. This will be available for use by UK
academics in a form similar to the Harwell Subroutine Library (HSL). Component routines must pass an
extensive suite of test procedures to ensure specif
ied performance and precision is met within the
guaranteed range of input variables.
Future Systems Performance Evaluation
Over the past year we have explored programming paradigms for the next generation of HPC systems
mory/vector compilation and threads programming in various
possible combinations. Vampir and Dimemas (from PALLAS GmbH) have been identified as important
tools for instrumenting code performance and enabling performance predictions on a range of parallel
latforms and these were installed on the SP2. In a series of collaborations between the applications
specialists and the HPCI staff, these tools have been evaluated and used to develop models of the
performance of key applications codes including GAMESS, D
POLY, CETEP, ANGUS, FLITE3D and
In brief our approach to analysing code scalability and identifying bottlenecks is:
Profiling of communication patterns using VAMPIR;
Projection of performance on alternative computer architectures using DIMEMAS;
Detailed cost modelling of major code steps with extrapolation to larger numbers of processors. Cost
models include architectural features; and,
The suite of cost models may be applied to “virtual benchmarking”.
This methodology goes a long way beyond w
hat is possible by fitting performance results to Amdahl’s or
similar laws. It enables assessment of performance on systems to which we do not easily have access and
will give us the opportunity to prepare a number of Consortium codes for the HPC97 system
probably have more processing elements than previously available. Following code analysis, optimisation
can be carried out to improve scalability and time to solution. Some initial results of this work will be
presented at the RAPS workshop at D
aresbury Laboratory in May 1998.
Leading Edge HPC Applications Development and Support
Atomic and Molecular Physics
A range of enhanced features and new algorithms have been incorporated within the parallel R
electron scattering program. These incl
ude a new channel
splitting method and the parallel accumulation of
thermally averaged collision strengths. Full restart facilities have been added to allow calculations to be
performed in convenient stages. A scaled SVD method has been introduced to cal
cases involving strongly closed channels. PERL scripts have been written to provide overall program
control including error detection. Progress has been made on the development of the computational models
required to dynamically assig
n computer nodes as the size and character of the calculation varies with
values. This mechanism will allow the load
balancing to be improved as timings on a given
parallel computer are accumulated. The parallelization techniques have been adapted by colleagues at
Queen’s University Belfast to the two
dimensional propagator a
pproach used at intermediate scattering
In the area of multi
photon ionization of atoms and molecules, studies of a range of pre
methods have demonstrated the great difficulty in obtaining efficient iterative (conjugate gradient) sol
methods for the case of indefinite matrices that are not diagonally dominant. Presently the focus is on the
use of direct block
diagonal methods which appear adequate for treating a range of atomic multiphoton
ionization calculations using realis
tic laser frequencies.
Quantum Chemistry and Electronic Structure
Version 6.1 of the GAMESS code was distributed in October 1997. The range of parallel platforms
supported was increased to include implementations for the Cray T3E
and Cray T3E
160, Hitachi SR2201, Silicon Graphics Origin
2000 and DEC
highly scalable performance in SCF and direct
MP2 calculations (e.g. a speedup of 115 in large
calculations on 128 processors of
the Cray T3E). Many of the associated developments have been
conducted on machines abroad. The MPP version is currently being used on the Cray T3E by selected
members of the Macromolecular consortium.
In addition to the activities
reported in the CCP1 section, work over the past year has included porting of
the computational shell kernel to parallel platforms, such as the Daresbury IBM SP2 and the Edinburgh
DL_POLY is the general purpose parallel sim
ulation code developed at Daresbury Laboratory to support
CCP5. The code runs on a wide range of platforms including the T3D, where it has been adopted by the
Materials Consortium, and the IBM SP/2. DL_POLY is unique in offering parallel capability and a
broad range of scientific applications with access to source code. It is an established, major tool of
academic research. An important technical development undertaken in 1997 was the production of a
standard benchmark suite of large scale test
to assist with future computer procurement exercises (e.g.
CRYSTAL and CASTEP
Developments involve the computer simulation of material properties with a strong emphasis on
simulation of very large systems using high per
formance computers, and are focused on the CRYSTAL
code, and on Plane Wave DFT methods (CASTEP, CETEP, FEMD, VASP etc.). These codes are under
in particular the parallelisation strategies. Activities at Daresbury include co
supporting the use of the codes, proof of principle calculations and quick response to changing
user requirements for new functionality. In collaboration with the HPCI Centre effort has been put into
developing benchmark versions of these codes and modell
ing their performance.
A version of CRYSTAL suitable for use on massively parallel computers was developed in collaboration
with the HPCI centre at Daresbury. The implementation using standard MPI and ScalaPack libraries and a
Fortran90 interface was c
hosen in order to ensure portability to a wide range of computers. The beta release
was made available to CCP3 groups in February on the IBM SP2 at Daresbury and the Cray T3D machine
at EPCC. For large systems treated within the Hartree
excellent scaling was observed
up to 128 processors. This significantly increases the scale of the surface structures that can be studied
Andrews, Cambridge and Daresbury groups have initiated studies of the structure of TiO
main limitation on the scale of system that can be studied is the memory per node
expected that this limit will be lifted in the proposed HPC97 facility.
Spectral Element Methods in Quantum Mechanics
This work is seeking to apply spectral element
methods developed in engineering applications to the
solution of the electronic Schrödinger equation. Work over the reporting period has focused on a standard
Legendre implementation within hexahedral spectral elements applied to the three
mensional Cartesian form of the Schrödinger equation with hydrogenic potentials for atomic systems. The
behaviour of matrix elements around the singular Coulomb potential has been explored and poor
convergence has been identified. Current work is evaluatin
g whether Tchebyshef polynomials, with their
exact integrals, can improve convergence rates. Implementations within spherical polar or gnomonic
coordinate systems will be explored in the future.
Dynamic Load Balancing
One of the
more sophisticated features of advanced simulation packages is their ability to adapt the
computation based on the current state of the simulation, for example, using extra grid points where the
solution changes rapidly. As a result, the computational loa
d changes in both time and space leading to load
imbalances that degrade the parallel performance and lead to restrictive individual processor memory
requirements. Dynamic load balancing algorithms include a range of techniques to measure, schedule and
distribute the computation and data, with the objective of improving parallel performance. Over the past
year work has focused on the developing algorithms that balance the load whilst minimising the amount of
data that needs to be redistributed. A novel d
ata flow algorithm was proposed which is far more efficient
than any existing algorithm. This algorithm has now been widely adopted by researchers in the area of
dynamic load balancing and appears in a number of successful codes, including codes developed
UK. The optimal property of this algorithm, as well as some other diffusion algorithms, has also been
The ultimate aim of all simulation is to understand and therefore to affect a process through improved
design. The availab
ility of predictive and fast massively parallel simulation capabilities opens up the
opportunity of exploiting numerical modelling applications within an optimisation framework. Over the past
year work has focused on the implementation of control random se
arch and genetic algorithms to handle
efficiently the non
linear inequality and equality constraints found in many engineering applications.
References to publications arising from both of these areas of work can be found within the CCP reports.
ributed Computing Activities
The Distributed Computing Support Programme (DisCo) continues to be coordinated by M.F. Guest, who
provides liaison between the EPSRC Programme Managers and DisCo activities at the Laboratory.
Currently, the coord
ination activity focuses on the assessment of technical merit and value for money in
individual Programme grant applications containing requests for distributed computing equipment. We
currently receive over 100 grant proposals a year, and the coordinator’
s input has resulted in substantial
savings to EPSRC.
A key factor in the success of the Distributed Computing Co
ordination and Technical Support Programme
is that it covers a wide range of activities, providing much needed services to as ma
ny grant holders as
possible, and not just to a small section of users. In a commercial offering, the support function might well
be a simple HelpDesk type of facility, so the Distributed Computing Programme is quite an ambitious
activity in terms of both
its scope and its objectives. Key staff members involved in establishing and
resourcing the programme (Royd Whittington and Tristan Green) have both left the Laboratory over the
past 18 months, and it has to be said that progress during this reporting peri
od has been hampered by on
going recruitment problems. While we have successfully sustained the majority of activities by creative re
deployment of effort from related areas within the SLA, some of the associated areas of support have
arguably not been as
effective as in previous years (courses, site visits etc.). We would stress, however, that
the service to review the computing requirements of research proposals from the various sponsoring
Programmes has not been adversely affected by these manpower limi
tations. Highlights of some of these
activities are listed below.
105 grant line proposals were reviewed with the main bulk still coming from the Materials Programme. In
addition to the grant line, advice was provided to EPSRC Programme Mana
gers on computing
proposals submitted to the 1997 Joint Research Equipment Initiative (JREI), with 24 proposals considered
for the HPCG and 8 for Chemistry. This advice focused on value for money and appropriateness of the
equipment requested. A
meeting with interested Programme Mangers was held at EPSRC (April '98) to
further promote the availability of this service and to quantify and understand the differing requirements
associated with each of the Programme areas. We still feel that in future
years, when the peak of the
demands from all other Programmes is reached, that this could be a very significant activity with probably
over 200 research proposals to be reviewed a year. With about one million pounds of saving suggested by
the review so far
, there is really no reason why the service would not be fully utilised by all of the
Programmes. However, this should not be seen as a money saving exercise, and indeed our aim has also
been to provide better value for money. Some users now choose to disc
uss their computing requirement
with us before submitting the research proposal, reducing the scope of saving. However, in practice, the
price and technology may change very quickly within the normal grant processing period, so the review
should be seen as
a part of a continuous dialogue between the Support, EPSRC and the users.
In providing courses to the EPSRC Distributed Computing community, our goals are to provide:
training material highly focused on the day
today needs of a
primary requirement for a machine is to do science;
on experience which enables users to exploit more of the performance and functions of the
equipment funded; and,
a basic level of expertise in each area on which further support
from the Technical Support team may be
A course on System Administration for Scientists was held at Daresbury on 3 July, 1997. The purpose of
this workshop was to provide a platform for system administrators to exchange experience, to discuss
on problems and to learn from each other. It was a successful event attended by 12 system
administrators. For the EPSRC Distributed Computing Programme, active participation from its users in
this kind of events is clearly essential.
The annual M
achine Evaluation Workshop had been a major event of the Programme since it was
introduced in 1990, bringing together researchers and suppliers in a productive environment. Unlike other
trade shows and exhibitions, this was targeted at the scientists suppo
rted by the EPSRC Programmes, and
the vendors had very good ideas of what to bring to the event in order to meet the demands of this special
community. A total of 17 companies were represented at the 1997 Workshop, including most of the major
such as Hewlett Packard, IBM, Sun and Silicon Graphics. Several newly announced systems
were on show. In the two weeks preceding the Workshop, 31 delegates were registered to use the systems
for benchmarking. Ninety eight delegates attended the workshop w
hich lasted 2 days (11
The EPSRC Distributed Computing Software archive continues to provide a software on demand service to
the research community. Software already in the library may be simply downloaded onto a local
from where it can be run with no further work from the users. If the software is not in the library, the
Support Team will trace the software through various sources and may occasionally recommend an
alternative with more functionality and/or bett
er performance. Any newly found software will be placed in
the library ready for future requests. The major focus during the past twelve months has been on upgrading
the packages to the current OS releases on each of the major platforms. Currently, there a
re 194 packages
maintained in the library, spread evenly across the major platforms, as follows; Silicon Graphics (44), IBM
(34), SUN (41), DEC/Digital (40), and Hewlett Packard (35).
Telephone and Email Support
ordination and technical support are alwa
hand when help is needed. In urgent cases, users may
talk to the Support Team directly on the phone, although in most cases email has proved sufficient. Users
may call the Support Team not only for problems, but also for advice on what is the best bu
y at any point in
time, availability of software and hardware, and for news on the latest developments. About 50 non
calls were received last year. In addition, bulletin board, email and Web were used for urgent messages.
Performance related inquir
ies have been addressed by providing a variety of benchmark results on the
DisCo WWW pages (http://www.dl.ac.uk/TCSC/disco), results that have been updated throughout the year
as access became available to newly released hardware. The DisCo support team ha
s also remained in
regular contact with the leading workstation suppliers through meetings and email so as to stay abreast of
the latest product releases.
Savings to EPSRC
It is not difficult to calculate the saving to EPSRC from each activity. For exam
savings on grant proposals may be estimated on a case by case;
the average cost for a one
day course is about £300;
a support call may cost about £75 per call (it may need more than one call for each problem, and there
may also be a time limit for e
ach call); and,
a fixed sum can be attached to each package transferred.
The total derived in this way is substantial, but this is only a small part of the overall picture. The true
saving is in the time gained through problems being solved more quickly
, less downtime on machines, more
work done through better configured machines, and less headache for the scientists struggling with the
equipment. As a result of the time saved, productivity of the scientists may be improved and more science is
e are the real benefits from the Support Programme over and above the savings reported in
Industrial Collaboration Programme
The SLA funded industrial collaboration programme, originally instigated by the Science and Materials
EPSRC, has the aim of stimulating and facilitating the migration of computational expertise to
industry. The main progress to date has been in establishing a project with Unilever on the modelling of
zeolites using combined QM/MM techniques and the develo
pment of a coherent strategy for developing the
programme. In the following, we report on progress in both of these areas over the past year.
The collaboration with Unilever Port Sunlight Laboratories sought to develop and implement coup
Quantum Mechanical/ Molecular Mechanical (QM/ MM) methods for the study of acid catalysis by
zeolites, and to help Unilever assess the applicability of such modelling techniques for in
house studies. In
our last SLA report, which covered work up until
July 1997, we reviewed work that described the
benchmarking of an appropriate QM/MM model. Point charge models that gave an accurate treatment of
the long range electrostatics were explored and a number of schemes to describe the inner QM/MM
boundary and t
he outer MM termination boundary were investigated.
Over the past year organisational changes within Unilever have caused a lack of substantial input from them
into the project. It was announced that the Crossfields, the part of the Unilever business tha
t would have
benefited most from the research, had been sold to ICI.
This report covers the period July
October 1997. The PDRA associated with the project left at the end of
October 97 (2 months early), which interfered with the completion of the work di
scussed below. In the
absence of any guidance or experimental results from Unilever, the final few months were spent
investigating the applicability of the electronegativity equalisation method (EEM) in collaboration with
Prof. W. Mortier, University of Le
uven, Belgium. The EEM provides a way of parameterising the variation
of atomic charges with topology and structure for a series of related materials. In this case it allows the
charges derived for use in zeolite modelling forcefields (which have formed a
major part of activity under
the project) to be predicted for unknown materials based on a small set of parameters. This led to a draft
manuscript. Publications arising from the current reporting period include:
Computer Simulation of Zeolite Structure a
nd Reactivity using Embedded Cluster Methods. P.
Sherwood, A.H. de Vries, S.J. Collins, S.P. Greatbanks, N.A. Burton, M.A. Vincent, I.H. Hillier,
“Zeolite Structure and Reactivity by combined QM/MM Calculations”. A.H. de V
ries, P. Sherwood,
S.J. Collins, A.M. Rigby, M. Rigutto and G.J. Kramer.
J. Phys Chem,
“Quantum Chemical studies of alkene chemisorption in chabazite: A comparison of cluster and
embedded cluster models”. P.E. Sinclair, A.H. de Vries, P. Sherw
ood C.R.A. Catlow and R.A. van
, J. Chem. Soc., Faraday Trans
The techniques developed within this project are being taken forward within a European project, QUASI,
which is seeking to demonstrate and validate QM/MM methodologies wi
thin a broad base of industrial
Strategy for the Industrial Initiative
In our last report we described the Computational Science and Engineering Programme’s wider industrial
and marketing strategy and the role of the SLA funded activity wit
hin that strategy. The primary focus of
the EPSRC activity was to take the outputs from the core CCP and Parallel Supercomputing programmes
and to demonstrate their applicability to industrial problems though a series of Grand Challenge
ts. Further development and validation against a broader base of industrial applications
and the development of robust and supported commercial software offerings would then be undertaken
through larger scale projects funded, for example, by the European C
ommission. In fact the QM/MM
Unilever project and the subsequent QUASI proposal are a vivid demonstration of this linear development
process working successfully.
At the SLA Review in April 1997 the future plans for the Industrial Initiative were discu
ssed in detail.
There was strong support for the view that there was certainly a need for the type of Industrial Grand
Challenge projects proposed, but EPSRC argued that the SLA may not be the most appropriate place from
which to fund them. At subsequent m
eetings this view was reinforced and it was concluded that these
projects should be pursued through individual grant applications and that the SLA activity should be wound
down as soon as practicable. Over the year a number of other projects were explored
submitted to a number of European programmes and the Technology Foresight programme in the UK. The
programme is in the process of preparing a number of bids into EPSRC’s programmes, in particular, in the
areas of engineering optimisation, po
wder modelling and materials’ property design.
Appendix 1: Minutes of the CCP Steering Panel
C L R C
Department for Computation and Information
Computational Science and Engineering Programme
Minutes of the Meeting of the CCP Steering P
anel, Daresbury Laboratory, Friday 24 October
Prof D J Tildesley (Chairman)
Prof M A Robb (CCP1)
Prof K Taylor (CCP2
Prof J E Inglesfield (CCP3)
Dr N M Harrison (CCP3)
Dr S Bailey (CCP4)
Dr W Smith (CCP5)
Prof G G Balint
Dr W M Temmerman (CCP9)
Dr D Moss (CCP11)
Prof M Leschziner (CCP12)
Dr R Denny (CCP13)
Dr R Cernik (CCP14)
Dr P J Durham (Daresbury)
Mrs S Miller
Apologies were received from Prof N Isaa
cs (CCP4), Prof M P Allen (CCP5), Prof D Flower (CCP7),
Prof B Gyorffy (CCP9), Prof J Squire (CCP13), Dr A Fitch (CCP14), Ms D Harding (BBSRC), Dr A
1. Notes of the last Meeting (CCPSP/96/M2)
The notes of the last meeting were agreed.
2. Matters Arising
Dr Durham referred to the two new proposals presented at the last meeting. Dr Steve Linton’s
proposal was already written and possibly been submitted. He would inquire as to the status of
Professor Roy Chantrell’s proposal.
s an action to suggest industrial and overseas members to expand the membership of the
Steering Panel. Chairman would ask for these when the round table reports on individual CCPs were
3. EPSRC Review of Computational Science at Daresbury
am provided a brief background to this item. There had been a series of reviews conducted
into all computing activities directly supported by EPSRC through SLAs; and this involved permanent
staff at DL who work directly on CCP support. Members of the rev
iew panel were
Dr Alan Robiette
London Business School
Dr Roy Baker
University of London Computer Centre
Dr Louis Lee
University of Westminster
Prof Bryan Richards
University of Glasgow
Prof Michael Robb
King’s College, London
Dr Alison Wall
Ms Anne Rees
Dr Durham felt that the CCPs had faired very well overall with positive comments and no radical
changes. There was a suggestion that the permanent support staff at DL should become more active in
planning and d
irecting the CCPs. However, after general discussion the members of the CCP
Steering panel considered DL staff should actively participate but that direction should come from the
There was a request for EPSRC staff to be invited to futur
e CCPSP meetings and that increased
interaction between CCPs and programme managers be encouraged by way of invitation to CCP
working group meetings. This was felt to be a helpful suggestion.
At the review panel meeting Dr Durham felt that they were look
ing for ways in which DL could be
more flexible in deploying the permanent staff across the CCPs, looking at present needs as opposed
to historical background which had led to more support being given to some CCPs than others. This
would provide an opport
unity to review the needs of individual CCPs.
Professor Tildesley agreed that assessing the individual needs of CCPs with their chairmen was a
useful idea, and would enable support by experts in the relevant fields.
4. EPSRC Review of Computing Deliv
Professor Tildesley reported on a meeting organised by EPSRC seeking views about the handling of
HPC in the EPSRC. The attendees included Professor Tildesley, Prof M Allen, Prof J Tennyson, Prof
C R A Catlow with Dr A Rose, Ms P Backway an
d Dr A Wall from EPSRC.
A number of important issues was addressed including:
i) How, and whether, to fund HPC in the future;
ii) the distribution of computing needs;
The recommendation of the meeting was that there should be HPC provisio
n and support in the future
but there was still uncertainty about the precise nature of the support and the mechanism for funding.
Programme managers would need to look carefully into which form of computing their budgets
should support, looking at value
for money across the full spectrum of computing from the
workstation to the highest performance national facilities. The meeting encouraged Programme
Managers to take a strategic view on this.
HPCG had a major responsibility for the CCPs at present. Pr
ofessor Tildesley believed that the CCPs
had a strong role in developing software for the responsive mode grant line. He sensed that this view
was quite widely held and that some mechanism for the funding of CCPs in the future would be
found. However, it
was important to interact more thoroughly and effectively with programme
managers and make sure we knew who was responsible for programmes. We needed to involve them
more, give them more information and invite them to appropriate meetings. We needed to
Web sites were of high quality and kept up to date.
Dr Durham gave a short overview of the current organisation of EPSRC programmes. At present the
core CCP support came through the HPC cross programme group but in future this might be
lled through individual programmes, which made it even more important to increase contact
with programme managers.
5. Reports on current projects
A tour of the table was made in which all chairmen and representatives present gave a brief progress
on their projects.
Professor Tildesley asked chairmen and representatives to give careful consideration to which
programme their CCP would naturally fall under and to ensure that the relevant programme managers
and assistants were invited to their meet
ings and study weekends. Where programmes overlapped,
managers from both programmes should be asked.
To consider which programme managers would be most closely involved
in considering the renewal of CCPs and to provide some thought
s on how they might be more
involved in the meetings and discussions of the CCPs. This information to be sent to Professor
Tildesley and Dr Durham as soon as possible.
There was an issue about the circulation and production of newsletters by CCPs. Thes
e were mailed
by a variety of methods, snail mail, electronically and on the Web. The need for newsletters on the
Web was emphasised, but the point was made that not everyone could take the time to browse the
Web. Also hard copies were mailed directly to
interested parties whereas the Web was available for
those who opted to look. A hard copy was still preferable with reference to Web addresses.
Professor Tildesley asked the meeting for suggestions for international and industrial members of the
nd the following were offered:
Dr Charles Clark, National Institute for Standards and Technology, Gaithersburg, USA
John Scott, Catalco
Stefano Baroni, former director of CECAM
Andrew Lyle, Glaxo
Mike Stapleton, MSI
Eric Wimmer, MSI
Ole Anderssen, Stuttg
David Parry, Pfizer
Bob Pendlebury ICI
Rick Mallane, Purdue
A Bedford, Unilever
R Miller, ICI
J Kendrick, ICI
T Ebisuzaki,, RIKEN, Japan
6. Collaborations with Japan
Dr Durham provided background to this item which was a project initiated by the Br
itish Embassy and
British Council with a view to stimulating potential research collaborations with the Science and
Technology Agency and its associated institutes in Japan. There was a visitation from a Japanese
delegation here last November followed by
a visit to Japan by Professor Richard Kenway and Dr
Paul Durham. The follow up was a workshop held in Japan in July which was attended by a number
of people from the CCPs. Dr Bob Gatten, NERC, co
ordinated the visit and was expected to prepare a
. However there have been individual follow up contacts with RIKEN, involving CCP5 and the
UKCP consortia. The UKCP consortia hoped to acquire some computer time from RIKEN and to
progress this, Dr Philip Lindan ,DL, is due to visit Japan in the near fu
Dr Temmerman, one of the group who visited Japan, believed there would be a Memorandum of
Understanding with the STA, to be signed in November. He also said it was unclear from the
Japanese what exactly they wanted. Professor Tildesley was concern
ed that there had been a huge
amount of work poured into computational science in the UK which should not be given away lightly.
7. Web pages for the CCPs
Dr Robert Allan and Ms Kerstin Kleese joined the meeting to talk to their paper on Web pages for
CPs. Ms Kleese gave an informative talk and demonstration on the capabilities of this national
information service which could be available for CCPs and HPCI consortia. The concept was to have
one central point containing standardised information relatin
g to CCPs and consortia without the need
to link through many pages. There would also be the facility to update pages externally via the use of
masks. It would also be a useful tool to advertise jobs, dates of events, newsletters, publication lists,
tings and workshops and anything else that would benefit from publicity.
In conclusion Professor Tildesley thought it would be advantageous to have a uniform CCP look and
encouraged chairmen to let Kerstin and Rob know what they would like to see included
. If the pages
were updated with key activities and events it would help with future renewal applications. Professor
Tildesley asked Dr Allan what the cost would be and he said that it would depend on what software
To summarise Professor T
ildesley was in favour of taking this idea forward but we needed to consider
carefully the best way to manage it and the costs involved.
Action CCP Chairmen
To consider the quality of individual CCP web pages and comment on the
delivery of CCP newslett
ers in browsable form across the Web. Comments to Professor Tildesley and
Dr Durham as soon as possible.
8. Any other business
Professor Cernik reported that Dr S Hasnain in the Synchrotron Radiation Department at DL was
working on a new CCP proposal a
nd this would probably be ready for the next meeting.
9. Date and venue of next meeting
The date of the next meeting was agreed as Thursday, 2 April 1998. If possible, the meeting was
hoped to take place at EPSRC, Swindon, and Dr Durham agreed to inv
estigate arrangements for this.
Appendix 2: Application Highlights
External Usage of the SP2
Computational Studies of Early Transition Metal Oxides
Furio Corà and C. Richard A. Catlow
Davy Faraday Research Laboratory
Institution of Great Britain
The work that we performed on the IBM/SP2 machine during the past year, has concerned the
computational study of the structure, reactivity and electronic properties of early transition metal oxides.
The oxides in which we have
been interested contain cations in high formal oxidation state (+4 to +6); the
high oxygen/metal (O/M) ratio leaves empty cationic interstices in the oxygen sub
lattice, a feature which
explains the tendency of oxides such as TiO
ncorporate extra cationic species
and form insertion compounds. The inserted species are generally neutral alkali metal atoms, which ionise
and cede electrons to the band
structure of the host material upon insertion. The insertion compounds are
red to as bronzes, for their metallic conductivity and appearance. They are widely exploited in the
electronics industry, with applications ranging from electro
chromic devices, based on the Mo
and W bronzes, to rechargeable Li batteries, which
, to solar cells and gas sensors based
on partially reduced WO
. Moreover, these transition metal oxides are widely exploited catalysts, usually in
reactions which require the activation of reagents via a redox step during the catalytic c
ycle. The catalytic
properties of the oxides are modified during the insertion process to form the bronzes. Finally, several
structured ternary oxides containing transition metal cations are employed in the electronics
industry for their ferroel
Rationalising the behaviour of transition metal oxides in the applications described above, requires the
understanding of their electronic structure, and of how it is influenced by the solid
state chemistry. Such a
study is particularly
suited to theoretical, quantum mechanical investigations. In our case, we used the
computer code CRYSTAL, jointly developed at the University of Torino (Italy) and at the Daresbury
Laboratory, and available for use on the SP2. The complexity of the materi
als examined and the cost of the
calculations require supercomputing facilities; a parallel machine, with few but powerful nodes, best
matches the technical requirements of the CRYSTAL code. The IBM/SP2 machine in Daresbury has
therefore the ideal combinat
ion of hardware resources for the investigations that we planned to perform.
Our work concentrated on the Mo and W trioxides, and on their alkali
metal bronzes. The bronzes can be
described as composed of an MO
framework, in whose interstices are located
the inserted, or extra
framework, cations. In the structural chemistry of the Mo and W oxides and bronzes, we find several
polymorphs, four of which have been studied, namely the perovskite (PV) structure of WO
, the hexagonal (Hex) phas
e of the K
bronze, the pyrochlore (PY) structure of Cs
the layered (L)
framework. For each of them we have optimised the structure of the host MO
framework (M=Mo,W); we further studied the Na and K bronzes of the PV and Hex structure
s. The results
obtained have enabled us to highlight and quantify the influence of the M
O bonding pattern [1
Coulomb forces [2,3], short
range repulsion effects , and structural requirements  in the solid
chemistry. By comparing the calcul
ated energy of the four polymorphs examined, we have constructed a
scale of relative stability for the host MO
framework. Results are reported in table 1.
Table 1: Relative energy, in eV per formula unit of the MO
framework, for the polymorphs examined.
Moving from the binary oxide to the insertion compounds, the relative stability of the polymorphs is altered;
in particular, a compromise b
etween the Coulomb and steric effects stabilises the structures whose
interstices have dimensions comparable with the size of the inserted ions: these are the PV polymorph for
, Hex for K
, PY for Cs
. We have further calculated the energy profi
le for the migration of
ions in the channels of the Hex tungsten bronzes; comparing the energetics of
the two ions, we estimated in 0.52 eV/K
ion the short
range repulsion energy between the inserted K
the host framework 
. The results of our calculations suggest that steric effects are sufficient to reverse the
relative stability of the MO
polymorphs upon insertion, in agreement with the experimental observation for
their stable crystal structures. The presence of extra
ramework cations in the synthesis medium may
therefore act as an inorganic template, or structure
directing agent, for the forming transition metal oxide
framework. The latter finding is reminiscent of the zeolite synthesis, where organic molecules are
roduced to produce molecular
sized interstices in the SiO
framework. We believe that a similar
template/host approach can be applied to produce novel transition metal oxide frameworks. In particular,
transition metal oxides with larger pore structures wou
ld satisfy the requirements of high ionic conductivity
and structural stability upon ion insertion and migration required for applications in electronics.
Furthermore, if the pore sizes reach the dimension of small organic molecules, the novel materials wo
also allow catalytic applications, and microporous transition metal oxides could combine the molecular
sieve characteristics of zeolites, with the much higher acidic strength and redox properties of transition
Computer modelling allows
in this case an analytic approach to the problem. A strategy has recently been
developed for the de
novo design of templates in zeolites [D.W.Lewis
., Nature, 382 (1996) 604], and
could be applied to the novel transition metal oxide structures: given
a target microporous architecture, the
organic template is gradually grown computationally within the pores of the host, starting from a small
molecular fragment used as seed. At each stage of the tentative template construction, the position,
and configuration of the molecule is altered, to maximise its interaction energy with the host.
The molecules obtained provide optimal space filling and maximum interaction energy with the selected
host, thus helping to direct the synthesis towards the tar
get structure. The method requires the knowledge of
the porous transition metal oxide structure as starting point. This is a challenging task in itself, since
microporous structures built up on MO
octahedra have never been reported. We proceeded by defini
new building block, as a chain of
sharing octahedra. Complete space filling can be obtained
by replacing the single octahedral units of the PV and Hex polymorphs, with alternating chains of
octahedra. Given the method of constru
ction, we call the new structures PV
. A complete series
of new frameworks with larger pores can be obtained by choosing the
We have optimised the structure of the first two members of the H
, with the sa
initio technology employed to study the known MO
polymorphs. The calculated internal energy of the
newly designed structures is compared in table 1 with that of the known polymorphs. The new structures are
unstable, as expected; the energy differenc
e has nonetheless only a marginally larger magnitude than that
between the known structures; we consider therefore possible to overcome this energy difference with steric
effects, using templates specifically designed to match the new structures . We ar
e currently employing
novo method outlined above to refine the template construction, before attempting the synthesis
In a parallel project, we have examined a set of 20 perovskite
structured materials of composition AMO
tand the effect of chemical composition on the structural and electronic properties, with a particular
emphasis towards the ferro
electric (FE/AFE) behaviour [1,3,6]. An accurate analysis of the
structure of the materials has shown that
the driving force for the FE and AFE distortional modes is
represented by the formation of a new,
O bond; the corresponding interaction is symmetry forbidden in
higher symmetry phases . The dispersion of energy levels in reciprocal space can be rat
ionalised in terms
of three parameters: one,
, linked to the intensity of the Coulomb field in the M and O positions, and
directly correlated to the formal oxidation state of the transition metal ion M. The second parameter,
quantifies the intensity o
O covalent interactions in the solid, while the third parameter,
, represents the
oxygen repulsion. The three parameters can be derived from the knowledge of the energy dispersion in
reciprocal space of the band
structure calculated with CRYSTAL
. A privileged role in the electronic
distribution is assumed by the value of
for each M
O interaction; it is directly proportional to the
degree of covalence in the M
O bonding. A local imbalance in
in the first coordination shell around
r an oxygen or a transition metal ion in the perovskite structure will induce a ferroelectric
distortion in the surrounding lattice. The electronic perturbation is then propagated in the neighbouring sites
via a delocalisation of the
O bonding ban
ds, a feature whose effectiveness is proportional to the value
in the host oxide. We have in this way been able to define an electronic parameter, easily
calculated from the ab
initio band structure of the pure oxides, which controls the ferroelec
Local imbalances in
can be introduced in the material by chemical doping, as demonstrated
experimentally by the Nb/Ta substitution in KTaO
, or by intrinsic defects, such as vacancies and surfaces.
To validate the idea, w
e have studied the BaTiO
<001> surfaces. In agreement with the
prediction of our model, the electronic and geometric relaxation induced by the surface couple with the
FE/AFE distortional mode of the bulk material. Ferroelectricity in BaTiO
and antiferroelectricity in
<001> surfaces are considerably enhanced, compared to the bulk materials. By analysing the
geometric constraints imposed by the surface, introduced to satisfy the charge neutrality requirement of the
surface layer, we have
been able to associate different surfaces with a FE or AFE character: the surface
periodicity controls the FE or AFE behaviour in the subsurface layers . Exploiting the hardware capacity
of the IBM/SP2, we have been able to study and fully optimise slab
s up to 15 atomic layers thick
(approximately 30 Å), to quantify the propagation of the surface perturbation in the subsurface region.
Finally, we have designed model surfaces, obtained by covering the BaTiO
<001> surfaces with
a single layer of
. The interface modifies the surface periodicity; we have in this way
been able to combine materials displaying bulk AFE behaviour with FE surfaces, and viceversa. Results of
the calculations performed suggest that in BaTiO
, the surfac
e requirement prevails, and the subsurface
region of the Nb
system has an AFE behaviour; when the substrate is WO
, the surface and bulk
requirements balance, leaving an almost undistorted WO
phase in the subsurface region . The latter
s suggest that surface engineering is a viable route to modify the FE/AFE properties of perovskite
materials containing transition metal ions, a property which could have important implications in the design
of novel ferroelectric materials for the electro
F.Corà, M.G.Stachiotti, C.R.A.Catlow and C.O.Rodriguez, 'Transition Metal Oxides Chemistry:
Electronic Structure Study of WO
', J. Phys. Chem. B, 1997, 101, 3945.
M.G.Stachiotti, F.Corà, C.R.A.Catlow and C.O.Rod
riguez, 'First Principles Investigation of ReO
Related Oxides', Phys. Rev. B, 1997, 55, 7508.
F.Corà, A.Patel, N.M.Harrison, R.Dovesi, C.R.A.Catlow, 'An ab
Fock study of the cubic
and tetragonal phases of Bulk Tungsten Trioxide', J. Am.
Chem. Soc., 1996, 118, 12174.
F.Corà and C.R.A.Catlow, 'Quantum
mechanical investigations of transition metal oxides' in 'How can
ab initio calculations be an effective tool for the study of mineral properties' , B.Winkler (ed), Kiel
(1997) ISBN 3
F.Corà and C.R.A.Catlow, 'Computer design of microporous transition metal oxides', accepted for
publication in the proceedings of the 9th World Ceramic Congress (CIMTEC 98), 1998.
F. Corà and C.R.A.Catlow, Chem. Commun., in preparation.
F.Corà , A.
Patel, N.M.Harrison, C.Roetti and C.R.A.Catlow, 'An ab
Fock study of
', J. Mater. Chem., 1997, 7, 959.
F.Corà and C.R.A.Catlow, 'Design of perovskites for electronics and catalytical applications', accepted
for publication in the proce
edings of the 9th World Ceramic Congress (CIMTEC 98), 1998.
C.R.A.Catlow, L.Ackermann, R.G.Bell, F.Corà, D.H.Gay, M.A.Nygren, J.C.Pereira, G.Sastre, B.Slater
and P.E.Sinclair, 'Computer Modelling as a Technique in Solid State Chemistry', Faraday Discussion
Vol. 106, 1 (1998).
M.Konigstein, F.Corà and C.R.A.Catlow, 'An ab
Fock study of the Enthalpies of Mixing
MnO and CaO
MnO solid Solutions', in press.
F. Corà, Ph.D. Thesis, in preparation.
Vibration energy level
s and wavefunctions of triatomic molecules
H. Y. Mussas and J Tennyson
Department of Physics and Astronomy, University College London
As a part of the ChemReact consortium we have developed a parallel version of the code DVR3D, which is
]. PDVR3D is a general code for finding rotation
vibration energy levels and
wavefunctions of triatomic molecules. We are using the new code to study a number of key molecules, such
, HCP and HN
using both the SP2 at Daresbury and the T3D/
T3E at Edinburgh. The SP2
has been our prime platform for program development because of is fast turnround. Since the CPU time of
the T3D/T3E is scarce, the SP2 has provided us with the opportunity to perform some production work.
Large problem, of course,
have required the Cray T3E/T3D.
: A key molecule, responsible for absorbing 70 percent of sunlight in the Earth’s atmosphere and is
necessary to sustain life. Detailed rotation
vibration calculations performed at UCL using serial computers
sformed our understanding of the spectroscopy of water. Ten of thousands of new transitions have
been assigned by members of the Theoretical Atomic, Molecular Physics and Astronomy (TAMPA) group
here at UCL. However, these assignments are only half way up
to the dissociation energy. Therefore much
remains to be done.
Other groups reported that they had obtained all the vibrational states of the water, but until our recent
calculations for was in its J=2 rotational states using PDVR3D, no one had obtained a
ll the bound states of
a rotationally excited chemically bound molecule. Calculations with J>0 are difficult since they scale as
(J+1) times the J=0 calculation. Using the SP2 and other Massively Parallel Computers, we currently are
calculating for high
er J’s upto the dissociation to enable others here at UCL to continue their pioneering
work of assigning the water spectra. Some of the data obtained so far have also been used to compute
reliably thermodynamic parameters for water.
: Fifteen years
ago Carrington and co
workers observed a very detailed and complicated spectrum of
the molecule in its dissociation region. Despite immense theoretical interest in this problem, since
understanding the behaviour of molecules as they dissociate is the key
to any attempt at detailed control of
chemical reactions, the spectra remains poorly understood.
Our first attempt to model this spectrum only included rotational states up to J=2, but individual
calculations took up to two weeks on a serial computer. We
have now developed an accurate, global
potential for H
, which supports approximately twice as many bound states as the less realistic potentials
as used by previous studies, A version of the PDVR3D for rotationally excited H
, which has particular
echnical problems, has been developed solely on the SP2.
: We have done some work on this very important molecule, but we have been hindered by the lack of an
accurate global potential. However, we are computationally placed to take advantage of any new
potential surfaces available.
: Protonated nitrogen HN
has been detected in interstellar clouds and is well
characterised by microwave and infrared spectroscopy. Theoretical studies on HN
are rather scarce. Using
potential surfaces by
Schmatz and co
workers, we have calculated all its vibrational bound states.
Finally we have looked at, albeit briefly, the HCP molecule as a test of the parallel programs against results
from other people.
H.Y.Mussa, J.Tennyson, C.J.Nobl
e and R.J.Allan, Rotational
vibration calculations using massively
parallel computers, Computer Phys. Comms., 108, 29
H.Y.Mussa and J.Tennyson, All bound vibration
rotation states of water up to dissociation, J. Chem.
Phys. (to be submitted).
J.Harris, S.Viti, H.Y.Mussa and J.Tennyson, Calculated high temperature partition function and
related thermodynamic data for H
O, J.Chem.Phys. (submitted).
R.Prosmiti, O.L.Polyansky and J.Tennyson, A Global potential energy surface for H
, Chem. Phys.
Letts,. 273, 107
H.Y.Mussa, S.Schmatz, M.Mladenovic and J.Tennyson, All bound vibration
rotation states of HN
to dissociation, J Chem. Phys. (in preparation).
R.Jones and C.D.Latham,
Over the past twelve months,
the Aimpro group based at Exeter, Newcastle, and Sussex has continued its
intense programme of work into atomistic modelling of point defects in semiconductors and related topics.
This has resulted in around thirty publications, together with numerous le
ctures and poster presentations at
national and international meetings. Central to our work is the Aimpro computer code: a local spin
functional based method using a real
space Gaussian basis. Recently, we have also developed a
ersion of the code: this is currently undergoing testing. Aimpro works well on both serial
and parallel architecture machines, and the IBM SP2 at Daresbury has been found to be a particularly
convenient platform for the code, especially since the recent up
grade to 26 nodes. The Daresbury facility is
ideal for problems of intermediate size that are too large to perform on workstations, yet not large enough to
warrant using the largest resources where time is severely rationed.
The main focus of
our work is
alloys. Selected details of this follow.
We have also performed exploratory work into modelling other systems such as peptides, and the
behaviour of C
molecules on silicon.
The diamond di
ways, diamond can be considered to be the prototype semiconductor. It also possesses some
remarkable, extreme properties: some are well known such as its famed hardness; others including its very
high thermal conductivity
five times that of copper at roo
are less well known. The unique
situation of diamond means that the study of defects in the material has been of fundamental importance to
science over many decades.
The two simplest native defects are the self
interstitial and the singl
e vacancy. Experimental and theoretical
work performed over many years have determined their electronic, optical, structural and dynamically
properties. Much debate, however, still surrounds the nature of the
semiconductors. The danglin
g bonds give rise to a number of deep levels in the forbidden gap, one of
which is a partially occupied doublet.
Several models for the
in diamond have been investigated using Aimpro. With reference to
recent experimental work in Oxford we pro
pose a outward
distortion from the unrelaxed D
configuration where atoms occupy their normal lattice positions. The spin density localisation, back bond
angles, overall spin, intradefect transitions and structure all closely match the experi
mental observations. A
paper will be submitted within the next two months.
Intrinsic defects in irradiated silicon
In silicon, defects such as the phosphorus
vacancy pair (
) or the oxygen
vacancy pair (
), as well as the lattice
, are subject of renewed interest by the scientific community. The
performance of heavily irradiated silicon
integrated particle detectors are severely degraded due to the
related defects inside the mate
Some controversy exists concerning the sense of the
distortion. Saito and Oshiyama have
proposed a new model for the structure of the single negatively charged
[Phys. Rev. Lett. 73,
866 (1995)]. This is in clear disagreement
with the model of Watkins and Corbett. From a theoretical
point of view, the modelling of the
is a very difficult problem. Like in diamond, this defect in
silicon is a
system, and therefore requires a large cluster (nearly 400 atoms
) to avoid defect
surface interactions. The earlier theoretical calculations employed
methods to determine the
structure of the defect, and it appears that this has led to an underestimate of the magnitude of the
simulation is necessary to describe the relaxation of this defect correctly.
Preliminary calculations using Aimpro show that the structure proposed by Saito and Oshiyama for a single
is unstable against bond recons
truction by pairing of four of the six dangling
bonds in the unrelaxed structure, thus confirming Watkins' model.
Calculations are also underway to calculate the optical and electrical properties of
related hydrogenated single
omplexes. Some of this work is being done in collaboration with
Nielsen at the Aahrus University in Denmark.
The results of our recent work on point defects in silicon will be presented at the European Materials
Research Society conference in Str
asbourg this summer.
Unlike diamond and silicon, defects in germanium have received far less attention although it was once
used much more widely than silicon. In recent years germanium, and silicon
germanium technology has
a renaissance. New devices are now being produced which make use of so called 'bandgap
engineering' to achieve superior performance over silicon alone. The commercial importance of this
provides strong motivation for basic research into point defects in
germanium and silicon
The four related defects in germanium, V
(n = 1, 4), are being investigated in response to EPR work on
these defects in both silicon and germanium at Aarhhus in Denmark. Past work by the Aimpro group has
successful with these defects in silicon and it is expected that similar results will be obtained. The
first stage of this work is nearing completion, and it is expected that the full results will be reported on at
MRS 1998 spring meeting in Strasb
ourg in June with a publication following. This work has
substantial technological importance as the passivation of vacancy and vacancy
like defects in germanium
becomes an increasingly common problem due to the current revolution in SiGe technology.
ansition metals in germanium
Both substitutional gold and the antimony
vacancy pair are being investigated. Early results for the
substitutional gold are promising, with four electronic levels identified with the defect by DLTS and Hall
s being supported by the work. The triple acceptor nature of gold and other transition
metals is of both fundamental and technological importance with applications associated with the use of
these metals as contacts and minority lifetime controlling centr
es in electronic devices. The preliminary
results on gold were presented at HPCI 1998 conference and were published in the proceedings of that
Hydrogen molecules in silicon
It has long been expected that hydrogen molecules exist as so called '
dark matter' inside silicon but a major
problem has been in devising an experiment which detects them. Very recently, experimental observations
on two distinct types of molecular hydrogen within silicon have been reported. Three classes of candidates
these hydrogenic defects have been examined using Aimpro to calculate their structures and vibrational
modes: i) isolated interstitial sites, ii) bound to oxygen impurities in the silicon, and iii) trapped in voids in
the lattice. The isolated molecule is
found to lie at a T
interstitial site, oriented along  and is infra
active. The rotational barrier for the molecule is found to be at least 0.17 eV. The molecular frequency is a
sensitive function of cage size and increases to the gas value fo
r cages about 40% larger than the T
We find that it is possible for H
molecules to cause an upward shift in the antisymmetric stretch mode of O
defects in the lattice when hydrogen is replaced by deuterium, which could explain the anomalous shift
the 1075 cm
H related local vibrational mode. It is suggested that Raman active modes observed
around 4158 cm
are due to molecules within voids.
hydrogen complexes in crystalline silicon
level transient spectroscopy
(DLTS) measurements on wet
chemically etched samples doped
metal impurities, such as gold, silver, platinum, copper, iron or even titanium, showed a
wealth of 'new' defects in silicon. This results have shown that atomic hydrogen, introd
uced via the etching
process, allows the creation of both active and inactive H
related complexes in the material. The possible
technological applications derived from these findings are far reaching. Despite, being a very powerful
technique, DLTS does n
ot give any information about the chemical composition, structure, stability or
formation mechanisms. In order to calculate the activation energy associated with different defect
structures, Slater's transition argument is applied. Due to the fact that t
he majority of these defects are
systems, relaxation and distortion is of crucial importance in reliable theoretical alculations.
This work has being carried out in close collaboration with Dr. J. Weber's experimental group at the Max
stitut Fur Festkorperforschung in Stuttgart
carbon defects in AlAs and GaAs
There is a need to understand the mechanism which causes degradation of semiconductor devices fabricated
from AlAs and GaAs doped with high concentrations of carbon
. Experimental work supported by EPSRC
and DERA is being performed by several groups in the UK led by R. C. Newman at the Interdisciplinary
Research Centre, Imperial College London. GEC
Marconi are industrial partners in the project, and
asurements of materials are performed by J. Wagner at the Fraunhofer Institute, Freiburg,
Germany. The theoretical support for this work is provided by the Aimpro group at Exeter.
In material which has been given heat treatments two vibrational modes are
observed. These appear to be
due to the C
C stretch vibrations of di
carbon defects. Calculations using Aimpro (mostly performed on
the Daresbury machine) strongly suggest that these di
carbon defects exist in two distinct forms depending
on whether th
e complex is singly positively ionised or in the neutral charge state. The results of the work
on GaAs have been published in Physical Review Letters, and the new work on AlAs in nearly ready for
submission to a leading journal. It will also be presented
at the forthcoming Shallow
Level Centres in
Semiconductors conference being held in Montpellier at the end of July.
In addition to Aimpro, more approximate but faster calculations can be performed using the Density
Functional Tight Binding method (DFTB)
developed by Th. Frauenheim and G. Seifert, originally at
Chemnitz, Germany (now at Paderborn and Dresden). The DFTB code can, for example, be used to sift
through large numbers of structures for interesting candidates to test in detail with Aimpro. In t
can supply accurate potential energy surfaces that are needed to generate and refine the tight
parameters for the DFTB code.
At the end of March, Michael Haugk joined the team at Exeter to work on the project investigating the
n of carbon in heavily doped AlAs and GaAs, using a combination of the DFTB and Aimpro
methods. He has made the necessary modifications to the DFTB program to run it on the IBM SP2 at
Daresbury, and incorporated new code for performing diffusion simulatio
ns. First results indicate carbon
moves through the crystal via bond centred sites, with an activation energy of less than 1.7 eV. We are now
in the process of refining this value using Aimpro: Wada
. report a figure of about 1 eV from their
Our models for the di
carbon defects indicate that hydrogen may react with them to form new complexes
with Raman and infrared active modes. In response to this hypothesis, J. Wagner exposed GaAs and AlAs
specimens containing di
carbon defects to a
hydrogen plasma and measured their Raman scattering
spectra. Several new modes have been found in the expected frequency range. We are presently performing
calculations to identify the exact nature of these new modes.
hydrogen defects in GaP
trogen impurities are important exciton traps in GaP and the observations that hydrogen can complex
with N has stimulated wide interest. Recently, three vibrational modes due to a nitrogen
have been reported and assigned to an NH
However, calculations performed using Aimpro of the
vibrational modes of this defect leads to modes which do not exhibit the observed isotope shifts. We thus
propose that the defect consists of a single H atom bonded with N. The model assumes that one o
f the high
frequency modes is an overtone of a H
bend mode. This new model can account for the observed modes
and their isotope shifts, as well as the effect of photo
excitation which leads to changes in the frequencies.
This section contain
s a selection of recent publications relating to the work described above.
carbon defects in annealed highly doped GaAs, J. Wagner, R. C. Newman, B. R. Davidson, S. P.
Westwater, T. J. Bullough, T. B. Joyce, C. D. Latham, R. Jones, and S. Öberg, Phy. Re
v. Lett, 78, 74
Dynamic properties of interstitial carbon and carbon
carbon pair Defects in Silicon, P. Leary, R. Jones,
S. Öberg, and V. J. T. Torres, Phys. Rev. B., 55, 2188
Substitutional carbon in germanium, L. Hoffmann, J. C. Bach
, B. Bech
Nielsen, P. Leary, R. Jones, and
S. Öberg, Phys. Rev. B, 55, 11167
Light impurities in silicon, R. C. Newman, and R. Jones, Current Opinion in Solid State & Materials
Science, 2, 40
Comment on "Electronic structure of the N
V center in diamond", J. P. Goss, R. Jones, P. R. Briddon,
G. Davies, A. T. Collins, A. Mainwood, J. A. van Wyk, J. M. Baker, M. E. Newton, A. M. Stoneham,
and S. C. Lawson, Phys. Rev. B, 56, 16031
cluster method and the dynamic
s of defects in semiconductors, R. Jones, and P. R.
Briddon, in Identification of Defects in Semiconductors, ed. M. Stavola, Semiconductors and
Semimetals, treatise editors, R. K. Willardson, A. C. Beer, and E. R. Weber, Academic Press.), in press
hydrogen defects in GaAs
a first principles study, J. P. Goss, R. Jones, S. Öberg, and P. R.
Briddon, Phys. Rev. B, 55, 15576
The N2 and N4 optical transitions in diamond: a breakdown of the vacancy model, R. Jones, J. P. Goss,
P. R. Br
iddon S. Öberg, Phys. Rev. B, 56, R1654
modeling of defects in non
metallic systems, R. Jones, to be published Bunsen Society,
carbon interstitial defect in diamond: a conflict between theory and experiment, T. D. Shaw,
R. Briddon, R. Jones, and S. Öberg, in preparation.
Carbon atoms catalyse fullerene growth, B. R. Eggen, M. I. Heggie, G. Jungnickel, C. D. Latham, R.
Jones, P. R. Briddon, Fullerene Science and Technology, 5, 727
Theory of gold
defects in silicon, A. Resende, J. Goss, P. R. Briddon, S. Öberg, R. Jones.,
Mater. Sci. Forum, Vol.258
Breakdown of the vacancy model for impurity
vacancy defects in diamond, R. Jones, J. P. Goss, P. R.
Briddon, and S. Öberg, Mater. Sc
i. Forum, Vol.258
Substitutional carbon in Ge and Si
, L. Hoffmann, J. C. Bach, J. Lundsgaard Hansen, A.
Nylandsted Larsen, B. Bech
Nielsen, P. Leary, R. Jones and S. Öberg, Mater. Sci. Forum, Vol.258
action with silicon interstitials, M. Budde, B. Bech
Nielsen, P. Leary, J. Goss, R. Jones,
P. R. Briddon, and S. Öberg, Proc. of ICDS Conference, Aveiro, in press (1997).
Local vibrational modes of weakly bound O
H complexes in Si, B. Bech
Nielsen, K. Tand
Budde, K. Bonde Nielsen, J. L. Lindström, R. Jones, S. Öberg, B. Hourahine, and P. Briddon, Mater.
Sci. Forum, Vol.258
study of the 90 degree partial dislocation core in diamond P. K. Sitch, R. Jones, S. Öberg,
M. I. Heggie, J. de Physique III, 7, 1381
Cyclic cluster model for calculating defects in solids using the local density approximation, J. Miro, P.
Deak, C. P. Ewels, and R. Jones, J. Phys. Condens. Matter., 9, 9555
Theory of thre
ading edge and screw dislocations in GaN, J. Elsner, R. Jones, P. K. Sitch, Th.
Frauenheim, M. I. Heggie, S. "Öberg, and P. R. Briddon, Physical Review Letters, 79, 3672
Identification of hydrogen
interstitials in silicon and germa
nium, M. Budde, B. Bech
Nielsen, P. Leary, J. Goss, R. Jones, P. R. Briddon, and S. Öberg, Phys. Rev. B, 57 4397
Interaction of hydrogen with substitutional and interstitial carbon defects in silicon, P. Leary, R. Jones,
and S. Öberg, Physical
Review B, 57 3887
First stage of oxygen aggregation in silicon: the oxygen dimer, S. Öberg, C. P. Ewels, R. Jones, T.
Hallberg, J. L. Lindström, L. I. Murin, P. B. Briddon, submitted.
Anomalous shift of the 1075 cm
in silicon B. Hourahine, R. Jones, S. Öberg,
and P. R. Briddon, Mater. Sci. Forum, 258
Hydrogen molecules in silicon located at interstitial sites and trapped in voids, B. Hourahine, R. Jones,
S. Öberg, R. C. Newman, P. R. Briddon, and E
. Roduner Phys. Rev. B rapid comm. in press (1998).
Structural and electrical properties of threading dislocations in GaN, J. Elsner, R. Jones, P. K. Sitch,
Th. Frauenheim, M. I. Heggie, S. Öberg, P. R. Briddon, Mater. Sci. Forum, Vol.258
Theory of nitrogen
hydrogen complexes in GaP, V. J. B. Torres, S. Öberg, R. Jones, R. Briddon,
Mater. Sci. Forum, Vol.258
A first principles study of interstitial Si in diamond, J. P. Goss, R. Jones, S. J. Breuer, P. R. Briddon, S
Öberg, Mater. Sci. Forum, Vol.258
hydrogen defects in silicon, P. Leary, S. Öberg, P. R. Briddon, R. Jones, Mater. Sci.
interstitial in silicon and german
ium, M. Budde, B. Bech
Leary, J. Goss, R. Jones, P. R. Briddon, S. Öberg, S. J. Breuer, Mater. Sci. Forum, Vol.258
APPENDIX 3: FINANCE
The outturn for 1997/98 was as follows: