The Australian Partnership for Advanced Computing<br>National Facility Annual Report

Car-Parrinello Simulations of Preferred Orientation

This project aims to determine the reasons for the formation of preferred crystal faces in the small crystals which make up thin film wear resistant coating for machine tools. The project has been expanded to look at the effects of crystal orientation in ion implantation, important for semiconductor and quantum computing applications. We have also begun investigating silicon carbon alloy structures in a collaboration with Case-Western University, USA.

Principal Investigator Marcela Bilek Applied and Plasma Physics Group School of Physics University of Sydney	Project d94
Co-Investigators Nigel Marks David McKenzie Applied and Plasma Physics Group School of Physics University of Sydney	RFCD Codes 240202, 240203, 250605, 250699

Significant Achievements, Anticipated Outcomes and Future Work
Crystal Nucleation and Orientation in Titanium Nitride
Titanium nitride is an industrially important alloy which adopts a variety of crystal orientations when prepared using energetic beams. We performed Car-Parrinello liquid quench simulations to explain why titanium nitride does not have an amorphous phase such as observed in other alloys of nitrogen. In both the liquid and solid states a strong preference for chemical order was found, and the amorphous solids displayed a high degree of crystalline character. Reducing the quenching rate was found to increase the fraction of atoms in crystalline sites. We propose to carry out further calculations employing constrained molecular dynamics in which a crystal seed is located within the liquid sample to aid crystal nucleation. This will be a crucial step upon the path of using the Car-Parrinello method to explain how biaxial stress controls the orientation of titanium nitride films.

Simulation of the Fabrication of the Quantum Computer
The Kane concept of the quantum computer in the top down process requires the implantation of isolated phosphorus atoms into a silicon host at a chosen depth and location. We have simulated the implantation process for a single phosphorus atom being implanted into silicon at an energy of 1 keV using the environment-dependent interaction potential (EDIP) for silicon. We have obtained important conclusions concerning the benefit of using a channelling versus a non channelling direction for silicon. We found that the yield of acceptable implantations is higher for the <111> direction than for the other channelling directions and higher than for a non channelling condition. We presented the results at the national quantum computing workshop in Mt Victoria NSW, September 2001. A journal article on the same subject is under preparation for the Journal of Computational Science.

The other main process for fabricating the Kane concept involves the growth of a silicon overlayer on a silicon surface onto which isolated phosphorus atoms have been placed. The silicon is grown by a CVD process in which silicon-containing radicals deposit silicon onto the surface. As a first step in studying this process, the UNSW quantum computing group has studied the surface of silicon which has been hydrogen terminated. We have used the Car-Parrinello method to study surface reconstructions and defect states in order to interpret the scanning tunnelling microscope images obtained by the experimental group. An excellent reproduction of the images has been obtained from first principles molecular dynamics simulations on the APAC National Facility computer. This work is under preparation for publication.

Wannier Function Analysis of Bonding in Silicon-Carbon Alloys
Wannier functions provide a means of obtaining maximally localised representations of the electron states in solids. The density functional method upon which the Car-Parrinello method is based provides Kohn Sham states as normal output. These can be transformed to the Wannier functions which provide the most localised states possible for representing the electron density. The location of the function is represented by the coordinates of its centre and the degree to which the state is localised is measured by a spread parameter. The Wannier functions have been used to study the details of the bonding in the silicon carbon alloys. Puzzling bonding arrangements in these alloys in which a silicon atom is apparently five coordinated have been studied and a means of determining the true coordination of the atom has been worked out using Wannier functions. In addition, an unusual "flipping" process has been observed in which the Wannier function describing a bond moves to a new location as a result of the thermal motion of the atoms in the vicinity. We have a paper in preparation for Physical Review B describing this work.

Structure of Silicon-Carbon Alloys
As part of a collaboration with Case Western Reserve University in the USA we have simulated the structure of a range of silicon—carbon alloys using the liquid quench method. The silicon carbon alloy system has many applications as an electronic material and as a wear resistant coating. We have prepared a range of compositions of these alloys for comparison with equivalent materials prepared by sputtering from elemental targets at Case Western Reserve University. We have studied the effect on the chemical order of various levels of hydrogenation and computed a range of quantities including the vibrational density of states, the radial distribution function and the electronic density of states for comparison with experiment. This work is under preparation for publication in Journal of Physics (Condensed Matter).

Computational Techniques Used

The activities of our group are focused on molecular dynamics simulations of solids. We have a suite of programs capable of carrying out a wide range of molecular dynamics calculations, based on either empirical potentials (EDIP method) or on ab initio techniques for quantum mechanical calculation of the bonding forces (Car-Parrinello method). Our codes run well on the APAC National Facility and the reductions in computation time have enabled us to simulate more complex systems than previously possible.