Quantum chemistry: molecular orbitals
Drug designers would like to compute how strongly a candidate molecule binds its target, to within about 1 kcal/mol, the difference between a drug and a dud. The equation that governs it is just the Schrödinger equation for electrons and nuclei, written down in 1926. Yet getting one drug-sized molecule’s energy right to that accuracy can take a supercomputer, and for some molecules no classical method succeeds at all. This chapter shows where the difficulty comes from, using the smallest molecule there is.
On the plate: the hydrogen molecule, computed two ways.
Born–Oppenheimer and orbitals
Nuclei are thousands of times heavier than electrons, so Max Born and J. Robert Oppenheimer proposed in 1927 to freeze the nuclei, solve for the electrons, and treat the resulting energy as a potential surface on which the nuclei move. Molecular orbitals are then built as linear combinations of atomic orbitals (LCAO). For H₂, two hydrogen 1s orbitals and combine into a bonding orbital , which piles electron density between the nuclei, and an antibonding . In practice each atomic orbital is itself a sum of Gaussians, a basis set; the minimal STO-3G basis uses three Gaussians per orbital, and larger basis sets add more functions, at rapidly rising cost. Established
Hartree–Fock, and where it fails
Hartree–Fock (HF) theory puts each electron in an orbital that feels only the average field of the others, and solves self-consistently. For H₂ in STO-3G, both electrons occupy . The plate computes every integral exactly (overlaps, kinetic energy, nuclear attraction, electron repulsion) and plots the HF energy against bond length 1.40 bohr, alongside the exact answer in the same basis (full configuration interaction, mixing in the doubly-excited configuration). Near equilibrium they differ by 20.6 millihartree, the correlation energy: 13 kcal/mol, more than ten times chemical accuracy. As the bond stretches, HF goes badly wrong: it insists both electrons share one orbital, giving a 50% chance of finding both on the same atom (H⁺ + H⁻), which costs far too much energy. These values match Attila Szabo and Neil Ostlund’s classic textbook. Established
Why correlation is hard
Full configuration interaction is exact in a basis but its size grows combinatorially: the number of ways to place electrons in orbitals. Practical methods trade accuracy for cost: HF scales as about , second-order perturbation theory , coupled cluster CCSD(T), the “gold standard”, as . That gold standard works when one configuration dominates. It fails for strongly correlated molecules, with many nearly degenerate configurations: stretched bonds, transition-metal clusters such as the iron–sulfur and iron–molybdenum cofactors of enzymes. There the number of important configurations is exponential, and that is precisely where quantum computers (Chapter 122, Part 5) might help. Trygve Helgaker, Poul Jørgensen and Jeppe Olsen’s textbook covers the methods in depth. Established
What you now have
The Born–Oppenheimer approximation, LCAO orbitals and basis sets, Hartree–Fock theory, an exactly computed H₂ curve, the correlation energy, HF’s failure on bond breaking, the cost ladder of correlated methods, and strong correlation.
· · ·
Three questions to carry with you.
And one to work out. At bohr in STO-3G, and hartree. What is the correlation energy, in millihartree?
Sources: M. Born and R. Oppenheimer, Ann. Phys. 389, 457 (1927). A. Szabo and N. S. Ostlund, Modern Quantum Chemistry (Dover, 1996). T. Helgaker, P. Jørgensen and J. Olsen, Molecular Electronic-Structure Theory (Wiley, 2000). Checked by code: STO-3G integrals evaluated analytically; E_HF(1.4) = −1.11672, E_FCI(1.4) = −1.13728 Ha, overlap 0.6593, matching Szabo and Ostlund.