The well-developed physics of semiconductors has given scientists fast computers and experimental equipment capable of measurements with high resolution. So it is fitting that, by applying these very tools, physicists have discovered that the domain of semiconductor crystal structures, and the processes responsible for them, is far richer than indicated by the models upon which the advances were built. First, there were new high-pressure angle-dispersive X-ray measurements, which showed that some of the previously predicted and apparently observed structures did not exist1. Now, writing in Physical Review Letters, Ozoliš and Zunger2 show that calculations from first principles may give the wrong structures when such calculations are only based upon the ‘usual suspects’. The authors used fast computers to incorporate the effects of phonons — which are quanta of crystal lattice vibrations — into the theory, and discovered that the old standby structures do not stand up well to atomic vibrations.
The prototypical semiconductor is part of the family of A NB 8–N compounds, where A and B are elements in the N th and (8−N)th main groups of the periodic table. For example, gallium arsenide (GaAs) and indium phosphideare A IIIB V compounds. Observations and simple theoretical models based on concepts of ionic size, as well as the relative electronegativity (power of the atoms to attract electrons) and hence the degree of covalency or ionicity of the chemical bonds, yield three dominant structure types for these compounds3. Those in which the energy benefits of forming four, fully saturated covalent bonds outweigh the competing need for efficient atomic packing, adopt tetrahedrally coordinated structures. These ‘covalent’ structures include the diamond lattice of elemental semiconductors such as silicon and germanium, and the zinc blende or wurtzite structures of many A IIIB V and A IIB VI compounds. Those ‘ionic’ compounds with ionicities that exceed about 80% tend to adopt the octahedral coordination of the sodium chloride (NaCl) structure. The semimetallic tin adopts a third ‘metallic’ structure, β-tin. This is also octahedrally coordinated, although the four equatorial bonds are somewhat distorted towards tetrahedral corners, suggesting a significant covalent influence. Pseudopotential calculations from first principles, which involve solutions of Schrödinger's wave equation without any empirically adjustable parameters (only carefully developed approximations of a ‘mean field’ type are allowed to convert the intractable many-electron problem to a single-electron problem),have largely supported the three-structure model4.
This is a preview of subscription content, access via your institution