Chemistry 324 - Spring 2008

A few days ago I showed you how to model an energy profile for internal rotation. As part of that exercise, we tried (quickly) calculating a profile using AM1 and I said the results weren’t believable. I have also asked you to calculate some profiles using HF/3-21G. Are these more believable? Let’s see…

Truth = MP2/6-31G*

I have calculated energy profiles for internal rotation in 1,3-butadiene using 5 different tools: MMFF (a molecular mechanics force field), AM1 (semi-empirical QM), HF/3-21G and HF/6-31G (two versions of Hartree-Fock QM), and MP2/6-31G* (a post-Hartree-Fock QM tool). Of these, MP2 should give the most reliable results, so let’s just assume the results are reliable.

E relative (kcal/mol) vs. CCCC dihedral angle (^o) for MMFF, AM1, & MP2

MMFF energies are pretty good. AM1 are not. The AM1 curve is in the right ballpark, but all kinds of important details are incorrect: no skewed minimum, no cis (eclipsed) maximum, trans → skew barrier much too small.

E relative (kcal/mol) vs. CCCC dihedral angle (^o) for HF/3-21G, HF/6-31G* & MP2

The 3 models agree very closely. The agreement is especially good at large angles (180 > dihedral > 120), then subtle disagreements (error ~0.3 kcal/mol) appear.

The behavior of both Hartree-Fock models is very reassuring, but I have a sad fact to relate: the reliability of any approximate tool (and HF is definitely an approximation) varies from one molecule to the next. What works out so well for 1,3-butadiene, might or might not work for acrolein.