Although the method was originally developed to search for metal hydrides, it is equally effective in finding M-H-X bonds (e.g. M-H-B or M-H-C type "agostic" interactions), or even ordinary C-H bonds if the esd restriction is tight enough! (NOTE though that SHELXL provides a far more satisfactory placement of such bonds, and this method should NOT be used for this purpose).
A conjugate-gradient technique is used (derived from that of Fletcher and Reeves, Computer J. 1964, 7, 149) which finds local minima of the potential energy function (including bond length constraints) appropriate to the hydride specified. In an initial search, interactions are evaluated for all those atoms within a restricted "search radius" (see test output) of an initial estimate of the hydride position. Twenty six starting hydride positions around this estimate are refined and each optimised position checked for suitability. Positions too close to a non-bonded atom are rejected, as are those where the M-H bond length is too far from that specified. When convergence of a particular optimisation is slow and therefore incomplete, a message to this effect is output. The unique "suitable" positions are further refined and their coordinates and potential energies (excluding the bond length term) output.
This technique ensure location of separated multiple hydride sites in cases where more than one hydride is bound to a vertex, edge or face of a cluster. The output table of inter-minima distances enables the user to check for short hydride-hydride contacts (inter-hydride distances as short as 1.85A are known). The co-refinement of hydride positions is possible using the HLOC command and is recommended in cases where the hydrides are sufficiently close to one another to have significant non-bonded interactions (less than say 3A apart say). Multiple hydride sites may well not be detected in an initial search.
Typical hydride site energies are in the range 0.5-3.5 units wheras false minima have higher energies greater than 10 units. Energies are not on an absolute scale, but show only slight variation between different molecules and site types. Thus typically terminal hydride sites have higher energy than edge bridging or triply bridging sites. The program has been tested on numerous examples. Initial testing was done against neutron diffraction data and showed r.m.s disagreement with neutron determined hydride positions of ca 0.05A. (see A. G. Orpen, J. Chem. Soc., Dalton Trans., 1980, 2509). Difficulties are most likely to arise when the non-hydride ligand framework is sparse (i.e. polyhydrides) or the ligand distribution is far from spherical, as in square- planar complex.