AbstractThis thesis consisted of two projects. Firstly the study of the dimers Al2H6, Ga2H6, A1BH6, GaBH6 and GaA1H6, and XB2H9 and the series of XnB4-nH1O isomers (X=Al or Ga, n~1) by ab initio (DZP and 321G*) and, for comparitive purposes, pseudopotential methods (where the Hays-Wadt ECP and basis set was compared directly to 321G* results, as the valence basis was the same, and likewise for DZP/Stevens-Basch-Krauss (SBK**) calculations) at the RHF level. The gallium structures have all been studied previously, but the aluminium borohydrides have not. It was found for aluminium borohydrides that the aluminium preferred, at this level of theory, to not bond directly to other heavy atoms (like Boron), the opposite situation to that found for gallium. If it must, in certain isomers bind to a heavy atom, it prefers to bind to another aluminium, rather than a boron atom. Gallium too appears to favour bonding to another gallium atom over boron.
Structurally, the pseudopotential basis sets reproduced ab initio values to a high
degree of accuracy. The bis(diboranyl) structures were better reproduced than
arachno structures due to the lesser steric strain. Generally, the HW method yielded results in better agreement with its ab initio counterpart, than the SBK did with DZP. However, this may be attributed to the greater difference in basis set between SBK and DZP, since the SBK basis set assigns TZ representation to the 3d orbitals. The SBK method also frequently overestimated boron-associated parameters. Timing studies of ab initio against pseudopotential, revealed little advantage in using these methods, probably due to inefficient coding. A small SBK basis was also tested (with 3d electrons in the ECP) and found to be a little better. The I-lW method generally yielded results in better agreement with contracted DZP (the larger ab initio basis set) structures, than did the 321G* model. This is probably due to the ECP underestimating nuclear charge. Frequencies were also well calculated by pseudopotential means, although the zero frequencies were not well reproduced.
The second project was to parameterise the semiempirical program ZINDO, for
elements gallium to bromine. This was done for both an sp and spd (d polarisation) basis set. Gallium was also tested with 3d electrons in the valence basis set, but this was not found to be successful. A series of reference compounds were compared, geometrically, with ZINDO and more sophisticated PM3 results. ZINDO was found to be more successful than PM3 at reproducing structures for these elements, particularly using the sp basis, since d polarisation tended to "unbalance" results.Costwise, this method was very effective, with calculations always taking under 2 minutes for geometry optimisations. A series of aluminium-chlorine calculations were also performed for comparison (these being already parameterised for spectral calculations) and showed dreadful results, indicating that re-parameterisation for geometry optimisation might produce even better ZINDO results for the galliumbromine.
|Date of Award||1994|