H€uckel method (Section 4.4). In the hands of Hoffmann, to whom we owe the
EHM in its current form [ 112 ], extended H€uckel calculations have given powerful
insight into the structures of these compounds. Wide-ranging corroboration of this
assertion is seen in Hoffmann’s Nobel lecture [ 100 ]. Some other examples are a
polymeric rhenium compound [ 113 ], manganese clusters [ 114 ], and iridium [ 115 ]
and nickel [ 116 ] coordination compounds.
Unlike the extended H€uckel method, AM1 and PM3 are useful for optimizing
geometries and (less reliably) calculating relative energies of organic compounds,
a purpose for which they were primarily designed. For TM compounds, a version
of PM3, PM3(tm), available in Spartan [ 31 ] (in later versions of the program,
not explicitly called PM3(tm) but parameterized for several transition metals)
was developed. This is very fast and has been quite extensively used, with mixed
results. Buda et al. compared PM3(tm) with ab initio (MP2 on HF geometries) and
DFT for 30 complexes and found that PM3(tm) reproduced the crystallographic
data in 80% of the cases, compared to 87% for MP2//HF and hybrid DFT, and 90%
for pure DFT [ 117 ]. Cooney et al. found it accurate enough as far as steric factors
go, for predicting novel properties of rhodium phosphines [ 118 ], and Zakharian and
Coon reported that “In general, the PM3(tm) method in Spartan shows promise for
predicting adsorption sites and vibrational frequencies of molecules on metal [i.e.
nickel] surfaces” [ 119 ], while Goh and Marynick found it to be inadequate for
energies, although its geometries were accurate enough for “energetics at a higher
level” (they refer to isodesmic reaction energies) with compounds of Cr, Mo, W and
Co [ 120 ], and Bosque and Maseras obtained geometries ranging from excellent to
very poor by comparison with literature X-ray and neutron diffraction and with
ab initio and DFT calculations, with compounds of Pd, W and Ti [ 121 ]. The TM
parameterization of PM3 is discussed by McNamara et al. [ 122 ]. With this varia-
bility in performance great care is clearly needed in judging the appropriateness and
reliability of PM3(tm) calculations: results for model systems might be compared
with experiment, or, because of its speed, the method could be used in a large,
suggestive survey. Semiempirical approaches to the computation of geometries and
energies (e.g. bond energies, heats of formation) of transition metal compounds
have not reached the same level of reliability that has been attained for organic
compounds with the normal (full) first-row (C, H, N, O, F) elements (Chapter 6).
Some may not regard this as a serious problem in view of the speed of DFT over
high-level ab initio methods like CCSDT, the availability of improved functionals,
and the reliability of pseudopotentials.
8.4 Summary................................................................
For some purposes gas-phase calculations are unrealistic, e.g. for understanding
some solution-phase reactions, or even almost useless, e.g. for the prediction of pKa
in solution. For introducing the effects of solvation, there are two methodologies
552 8 Some “Special” Topics