Introduction

QUASI (Quantum Simulation in Industry) is a collaborative research project to implement techniques for combined QM/MM (quantum mechanics/molecular mechanics) simulations on a variety of High Performance Computing (HPC) platforms, and to apply the techniques to industrial catalytic chemistry applications.

The project was coordinated by Daresbury and ran for 3 1/2 years, Jan 1998 to June 2001. Software developments were conducted in collaboration with the groups of Prof. Walter Thiel at the Max Planck Institute für Kohlenforschung, Mülheim, Germany, and Prof C.R.A. Catlow at the Royal Institution (London, UK). The demonstration and applications work has involved modelling teams from within three major European chemical companies; Norsk Hydro (Porsgrunn, Norway), BASF (Ludwigshafen, Germany) and ICI (Middlesborough, UK). For full partner contact details please see the project web page [1].

Software and Algorithm Development

The QUASI software environment is based on ChemShell, a Tcl-based computational chemistry environment which is described in full in another section of this report. The QUASI partners have extended ChemShell to support the requirements of the systems to be modelled and to make the software easier to use in an industrial setting These developments have included

• The Solid-state embedding model, in which DFT calculations are interfaced with semi-classical models of ionic materials using the shell model. An interface region is defined, within which effective core potentials are included on the ionic centres to provide a suitable potential to terminate the QM wavefunction. This work was led by RI group [2,3].
• The Adjusted Connection Atom Scheme, a semi-empirical QM/MM termination scheme in which specially parameterised atoms are defined at the QM/MM interface. This was incorporated in the ChemShell MNDO interface in collaboration with the Muelheim group.
• MD Module A range of integration methods, including rigid body and constrained dynamics are available. This module was derived from code within the DL_POLY package. A Monte Carlo driver is also available.
• A Linear Scaling Geometry Optimiser (HDLCOpt) An optimiser, based on the delocalised internal coordinate approach was developed by the Muelheim group, for the treatment of macromolecular systems.[2] The system is divided up into residues, each of which is handles by delocalised internal coordinates. The separate regions are coupled by including the cartesian coordinates in the delocalisation procedure, and an iterative optimisation procedure is employed. It is possible to define a core residue within which a transition state search is conducted, while the remaining residues are minimised, thus enabling the size of the hessian matrices to controlled.
• New Code Interfaces including drivers for the TURBOMOLE and MNDO packages were developed (in collaboration with the BASF and Muelheim groups).
• A Graphical User Interface was constructed. This is based on the Cerius-2 environment of from MSI (now Accelrys) and makes use of the Software Developers Kit (SDK). The control parameters for QM, MM QM/MM and MD modules can be defined, and simple control sequences for creating cluster models

HPC Exploitation

QUASI is available for MPP computer systems, the MPP version allowing efficient operation of the DL_POLY, GAMESS-UK and GULP modules and the QM/MM models derived from them. In future versions this will incorporate MNDO also.

The Industrial applications sub-projects within the QUASI project (see below) have made extensive use of the Alpha-based beowulf system with QSNet interconnect (loki). The code is currently being ported to the Origin system (green) provide by the CSAR service at the University of Manchester for use by members of the Materials Consortium.

Industrial Applications Sub Projects

A major element of the QUASI project was a series of demonstration and application sub-projects that were conducted in collaboration with the industrial partners.

1. Modelling of NO_x Chemistry of Copper-containing Zeolites P. Sherwood and H.G. Schreckenbach, (DL) with Merethe Sjovoll, Elly Karlsen (Norsk Hydro, Porsgrunn, Norway).
   The QM/MM approach to zeolite modelling follows that used in previous studies [5], with a forcefield based on the CFF valence forcefield of Hill and Sauer [6]. We have used the QM/MM model to characterise the structure of the ZSM-5 framework with a single Cu atom adsorbed at a number of sites (ring, interstitial etc). The energies and structures of a number of species occurring in the catalytic decomposition on NO₂ and N₂O at these sites have been characterised, including the transition state for the decomposition of adsorbed Cu-OONN into Cu-O and N₂O. So far it appears that the intersitital sites, with just 2 Cu-O interactions are more catalytically active than the more heavily coordinated ring sites.
2. Understanding the Chemistry of Oxide Surfaces Richard Catlow, Alexei Sokol,Sam French and Stefan Bromley (RI, London) together with John Kendrick, Steve Rogers at ICI.
   The basis of this sub-project was the solid-state embedding scheme incorporating QM calculations performed using GAMESS-UK in an environment modelled using shell model forcefields using the GULP code. The project has characterised the energies of a large number of species implicated in the Cu/ZnO catalysed methanol synthesis process, using an oxygen terminated (000-1) surface of Zincite. The energetics of the involvement of vacant oxygen interstitial sites has also been explored. In a number of cases vibrational frequencies have been characterised and shown to be in good agreement with experiment. A series of calculations on adsorbed Cu clusters of varying sizes have been performed and these will form the basis of further studies on the Cu-catalysed chemistry [2,3].
   
3. Studies of Enzyme Reactivity (Walter Thiel, Frank Terstegen (Muelheim) together with Ansgar Schaefer and Christian Lennartz and BASF) The enzyme work made use of the CHARMM macromolecular force-field (as implemented in DL_POLY) and the MNDO and Turbomole codes. Three systems were studied. The catalytic interconversion of dihydroxyacetone into glyceraldyde-3-phosphate by the Triosphosphate Isomerase (TIM) system has been studied before and was chosen as a demonstration calculation. Using large-scale DFT calculations and varying the size of the QM region the proposed 4-step proton transfer pathway, via an enediol, with involvement of His95 [7]. The search for the transition states of the embedded reaction were made possible by the developments to the HDLCOpt optimiser (vide supra). Similarly, the HDLC optimiser was used, in conjunction with the GROMOS code and MNDO hamiltonian to study enzymatic reactions in hydroxybenzoate hydroxylase (PHBH) [8]. A detailed study of enzymatic reactions in thrombin involved, in addition to geometry optimisation involved extensive MD simulation to characterise the effect of different initial configurations on the energetics of the cleavage of the Arg-Ser bond in the substrate.

References

[1] http://www.cse.clrc.ac.uk/Activity/QUASI

[2] Samuel A. French, S.T. Bromley, A.A. Sokol, C.R.A. Catlow, J. Kendrick, S. Rogers, P. Sherwood. Presented at the 2001 MRS Spring Meeting and due to be published in the proceedings.

[3] S.T. Bromley, A.A. Sokol, C.R.A. Catlow, S. Rogers, F. King, and P. Sherwood.: Angewandte Chemie, In press

[4] S.R. Billeter, A.J. Turner and W. Thiel, Phys. Chem. Chem. Phys 2: (2000) 2177

[5] A.H. de Vries, P. Sherwood, S.J. Collins, A.M. Rigby, M. Rigutto and G.J. Kramer, J. Phys. Chem., 103 (1999) 6133.
[6] J.R. Hill and J. Sauer. J. Phys Chem 99 (1995), 9536.

[7] C. Lennartz, A Schäfer, F. Terstegem, W. Thiel. J. Phys. Chem. A, Submitted.

[8] S.R. Billeter, C.F.W. Hanser, T.Z. Mordasini , et al. Phys Chem Chem Phys 3 (2001) 688-695.