This lecture describes our techniques for three dimensional visualization of our Molecular Dynamics modelling. A brief introduction to Molecular Dynamics is given in the supplement. An earlier version of some of the material described below was presented in
This preprint can be obtained from our ftp site ( ftp (AHKRS)). Our approach has been developed by the members of the Computational Physics group at the Technion over the last few years to visualize a range of different systems. A simplified version is taught in our graduate Computational Physics course.
DESIRABLE ATTRIBUTES: We desire that our methods should be efficient, capable of providing full three-dimensional drawing capabilities, cheap to install and maintain (public domain wherever possible) and as light as possible on computer systems so that each group member has the possibility to use interactive visualization constantly on their desktop. Only when everyone has facilities on their desktop or on each station in the computer classroom can visualization be possible for each calculation without hesitation. We also consider it important that the systems used should be widely available on different platforms and expected to remain so for the forseeable future. The catch is that cheap does not usually combine well with efficient three-dimensional graphics.
SOFTWARE SELECTION: We found that neither PGPLOT nor MATLAB has the three dimensional capability that we desired; although we continue to use them for two dimensions. The three-dimensional ouput from Silicon Graphics' original GL library (which requires special hardware) or from AVS seemed to be just what we wanted, but the former required expensive hardware and the latter expensive software so neither was practical to place at each workplace. After brief flirts with GL, AVS, Silicon Graphics Explorer and Scian we decided to move on. (We are in no way saying that these are not good systems, they are in fact excellent but too expensive in our local environment in either their software or hardware demands for us.) After much testing we selected OpenGL which is Silicon Graphics' replacement for GL. It gives the option of graphics acceleration in hardware or a software only implementation and is widely available for both Unix and windows platforms. However while OpenGL is not proprietary, specific implementations make certain hardware/software demands that are difficult for us to provide on every existing X-terminal and LINUX box. Thus we gravitated to Mesa, the public domain (GNU) clone of OpenGL that is really free, does not require any special hardware, and works on every X-terminal and LINUX box. We have found the two systems to be truly interchangeable for our purposes: we use Mesa for development and revert to OpenGL if needed to obtain optimum performance for presentations and video recording.
DEVELOPMENT CREDIT: The OpenGL routines that we use today were developed by two of my graduate students, David Saada and Adham Hashibon (shown below, with David seated at the computer) with support from Dr Batia Pery from the Technion Computer Centre consultation group .
PROGRAMMING STYLE: OpenGL and Mesa are called from c programs (in a similar style to the calls to PGPLOT) and the images can be manipulated by mouse or keystroke commands at will. For the larger systems we prefer to work from data files previously generated rather than interactively, since most of the programs have to be run in parallel mode on either or our LINUX cluster or the Israeli IUCC's SP2. (A cluster of LINUX nodes built from off the shelf components and running MPI makes a very nice budget parallel supercomputer. We obtain a lot of CPU from ours.)
EXAMPLES: All our projects in this area relate to modelling projects carried out in collaboration with experimental researchers. We discuss a few related to carbon in detail below; others can be found in the AHKRS manuscript mentioned above.
The first set of calculations used Molecular Dynamics with Tersoff's carbon potentials. In the manuscript, D. Saada, J. Adler and R. Kalish, ``Transformation of the Diamond (sp^3) to Graphite (sp^2) bonds by ion-impact'', (1998) International Journal of Modern Physics, C, 9, 61, which can be found on our ftp site ( ftp (SAK)), we described our T=0 study of the formation of point defects in diamond induced by an energetic displacement of a carbon atom from its lattice site and the relaxation of the thereby disrupted crystal. The displacement energy for Frenkel pair creation was calculated to be 52 eV, in agreement with experiments. It was found that the <100> split-interstitial, with a bonding configuration which resembles graphite, was created by many different bombardments. In the figure below left, a section of a 5000 atom carbon sample ordered in a diamond lattice is shown, before one atom (drawn in green) is kicked out of position. In the figure below right, the lattice is shown at a later stage when the atom has moved to the right and displaced the surrounding ones.
The conclusions from this study include the further observation that ehe disrupted region around the defects was found to be rich in sp^2-like (graphitic) bonds and to extend several nanometers. This is likely to be the elementary electrically conductive cell experimentally found in radiation-damaged diamond. Recent extensions to this project include temperature dependence and bombardment by up to 12 particles with subsequent annealing.
MOVIE: A selection of visualizations from these two projects including the temperature dependence of defect structures, an animation of the energetic displacement at T=0 and visualization of configurations after bombardment and after annealing at finite temperatures is contained in our movie by I. Rosenblum, D. Saada, S. Brandon and J. Adler, entitled ``Simulations in Diamond''. This movie has three parts, a general introduction to carbon and molecular dynamics for carbon, a section on ``Irradiation of Diamond'', by D. Saada, J. Adler and R. Kalish and a section on ``Stress in Diamond'', by I. Rosenblum, J. Adler, S. Brandon and A. Hoffmann.
The preliminary development for the research projects and video was done using Mesa under LINUX, but for the recording we switched to OpenGL with the change of a few include statements and the Makefile. The movie was recorded using the Galileo system on a Silicon Graphics computer, and also includes some vrml frames based on our still frames. Some sample visualization programmes used in the Diamond movie are avaliable by ftp and give the compilation command, c routine containing the OpenGL commands, datafile and OpenGL executable (prepared on the O2 Silicon Graphics of the Technion Visualization center) to draw the image below:
To the previous page: Lecture 1.
To the next page: Lecture 3.