carbonyl groups and electrophilic C/C bonds, radical addition to alkenes, aldol
condensations, and various intramolecular reactions [ 13 ].
3.3.2 To Obtain Good Geometries (and Perhaps Energies)
for Small- to Medium-Sized Molecules
Molecular mechanics can provide excellent geometries for small (roughly C 1 to
about C 10 ) and medium-sized (roughly C 11 to C 100 ) organic molecules. It is by no
means limited to organic molecules, as forcefields like SYBYL and UFF [ 5 ] have
been parameterized for most of the periodic table, but the great majority of MM
calculations have been done on organics, probably largely because MM was the
creation of organic chemists (this is probably because the concept of geometric
structure has long been central in organic chemistry). The two salient features of
MM calculations on small to medium-sized molecules is that they arefastand they
canbevery accurate. Times required for a geometry optimization of unbranched
C 20 H 42 , ofC2hsymmetry, with the Merck Molecular Force Field (MMFF), the
semiempirical AM1 (Chapter 6 ) and the ab initio HF/3–21G (Chapter 5 ) methods,
as implemented with the program SPARTAN [ 15 ], were 1.2 s, 16 s, and 57 min,
respectively (on an obsolescent machine a few years ago; these times would now
by shorter by a factor of at least 2). Clearly as far as speed goes there is no contest
between the methods, and the edge in favor of MM increases with the size of the
molecule. In fact, MM was till recently the only practical method for calculations
on molecules with more than about 100 heavy atoms (in computational chemistry
a heavy atom is any atom heavier than helium). Even programs not designed
specifically for macromolecules will handle molecules with thousands of atoms
on a good PC.
MM energies can be very accuratefor families of compounds for which the
forcefield has been parameterized. Appropriate parameterization permits calcula-
tion ofDHf^0 (heat of formation, enthalpy of formation) in addition to strain energy
[ 1 f]. For the MM2 program (see below), for standard hydrocarbonsDHf^0 errors are
usually only 0–4 kJ mol$^1 , which is comparable to experimental error, and for
oxygen containing organics the errors are only 0–8 kJ mol$^1 [ 16 ]; the errors in MM
conformational energies are often only about 2 kJ mol$^1 [ 17 ]. MM geometries are
usually reasonably good for small to medium-sized molecules [ 4 , 9 a, 18 ]; for the
MM3 program (see below) the RMS error in bond lengths for cholesteryl acetate
was only about 0.007 A ̊[ 4 a]. “Bond length” is, if unqualified, somewhat imprecise,
since different methods of measurement give somewhat different values [ 4 a, 9 ]
(Section5.5.1). MM geometries are routinely used as input structures for quantum-
mechanical calculations, but in fact the MM geometry and energy are in some cases
as good as or better than those from a “higher-level” calculation [ 19 ]. The bench-
mark MM programs for small to medium-sized molecules are probably MM3 and
MM4 [ 4 , 5 ]; the Merck Molecular Force Field (MMFF) [ 20 ] is likely to become
very popular too, not least because of its implementation in SPARTAN [ 15 ].
64 3 Molecular Mechanics