Marek T. Michalewicz writes:
At the Spring Foresight Senior Associate Meeting I pledged to give my electronic structure code for public use. Long overdue, but here it is.
I hope the program will be useful to some.
Marek T. Michalewicz
Quantum Precision Instruments Pty. Ltd."
Read more for details on the code and how to access it on the web. Marek T. Michalewicz writes:
"Please be forgiving – it is a research piece of code in Fortran90. Not much of embellishments, rather dirty, not a lot of comments. However, I'm happy to guide and explain details. It was written for NEC SX-4 vector-parallel supercomputer, and I managed to brake few nice records with it.
The code computes the electronic density of states of very large samples (in my case up-to nearly 8 million atoms of TiO2). It is very fast, very well optimized for this type of architecture (67% theoretical peak speed of this machine, or 43 GFLOPS sustained speed on a 64 GFLOPS, 32CPUs, machine).
You can download it from: www.quantum-pi.com/programs/Eq_Of_Mot.tar.gz
Background: The program computes the electronic density of states (local, per atom, per orbital resolved, and total) using the Equation-of-Motion method. For references see: http://www.quantum-pi.com/Publications.html
It is based on semi-empirical Tight-Binding Hamiltonian. The advantage of using this method is the ability to treat point defects, vacancies, surfaces, and very large systems (nanocrystallites) with arbitrary morphology. I believe it will be possible to compute some 30 million atoms on a current SX-5 or a massively parallel machines. Mesoscopic systems, nanocrystallites of arbitrary morphology, carbon structures, f-orbitals, and other improvements are strightforward. If anyone needs hints or suggestions, I will be happy to help.
The program can be used to model very large networks of carbon structures, nanotube networks etc. It was developed for disordered transition metal oxides, where it is often difficult to apply ab-initio methods.