Working out the electronic properties of materials from the atomic configuration is routine, if grueling work. The real trick would be to go the other way--start with the electronic and optical properties you want and then work out an atomic structure that would give it to you. This kind of reverse engineering of technical materials on the atomic structure level would be a valuable tool in developing both electronic and optronic devices.
The trick has been turned by researchers at the National Renewable Energy Lab. Their inverse approach finds the atomic configuration that produces a prescribed electronic structure. It should answer such questions as: "For a given superlattice orientation, what is the layer sequence that has the maximum bandgap?" Or even, "Which structure would give a specified bandgap (such as 2 eV)?" Or, "What crystal structure will have a band structure that maximizes Auger carrier multiplication?" You would also be able to address inverse problems in the context of vibrational and photonic properties and for molecules and low-dimensional systems.
The idea of reverse engineering these structures becomes much more enticing because such cutting-edge crystal growth techniques as molecular beam epitaxy or metal-organic chemical vapor deposition can produce what you came up with. They can make prescribed crystal structures, sometimes even in defiance of equilibrium, bulk thermodynamics.
The problem has been that the number of possible combinations is so vast and electronic properties are so sensitive to details of crystal structure that you could never get there with simple trial and error, even the very fast highly automated methods of combinatorial synthesis. In one example, the number of possible configurations would be 100 billion. You must know what you are doing to narrow the field down to what's doable.
The NREL theoretical method could narrow the quest to well below the stab-in-the-dark level. Specifically, it addresses the problem of finding an atomic configuration of a complex, multi-component system having a target electronic structure property. The method uses simulated annealing to focus on the configuration that matches the properties. They tried it with semiconductor alloys and superlattices, but the same algorithm could be applied to many other types of compounds.
The simulated annealing algorithm can learn the system relatively quickly by retaining only those configurations that are conducive to the target electronic structure. Coupled with a fast "Order N" solution of the Schroedinger equation, the NREL method allows them to determine the target configuration of systems containing a few hundred atoms.
It deals with substitutional systems described by an underlying lattice structure of a specified number of sites. Then a set of electronic-structure properties is defined for every configuration and the corresponding target properties. The distance between the electronic structure of the configuration and what is wanted is determined and minimized by varying the configuration variables, using electronic structure theory.
You need a fast learning method of sampling the configuration space, and a numerically efficient yet physically accurate method of calculating the electronic structure of a given atomic configuration. This they have. The simulated annealing technique is an efficient algorithm for finding the global minimum of a multivariable, multivalley function. It's good for exploring the configuration space.
It is also good at determining the electronic structure of a given atomic configuration. It does this through atomic relaxations, realistic Hamiltonians, and fast diagonalization. Since they are interested in electronic properties that involve band-edge energy level and wave function, the Hamiltonian is diagonalized using a computationally efficient method that focuses on an energy window around the bandgap.
NREL found a non-intuitive relation between the reciprocal-space band structure and the real-space atomic configuration that makes any guess made ahead of time unlikely to work. Some of the structures they predict are unsuspected on the basis of the normal insights underlying band theory.
They applied their method to semiconductor alloys and superlattices, but the algorithm could be applied to the
optimization of the electronic structure of complex molecules and clusters as well.
Copyright 1999, John Wiley & Sons, Inc., New York, NY 10158
From: Nanotech Alert, Nov 5, 1999.