07.04.2013 Views

Essentials of Computational Chemistry

Essentials of Computational Chemistry

Essentials of Computational Chemistry

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

6.3 KEY TECHNICAL AND PRACTICAL POINTS OF HARTREE–FOCK THEORY 181<br />

6.3.1 SCF Convergence<br />

As noted in Chapter 4, there is never any guarantee that the SCF process will actually<br />

converge to a stable solution. A fairly common problem is so-called ‘SCF oscillation’. This<br />

occurs when a particular density matrix, call it P (a) , is used to construct a Fock matrix F (a)<br />

(and thus the secular determinant), diagonalization <strong>of</strong> which permits the construction <strong>of</strong> an<br />

updated density matrix P (b) ; this is a general description <strong>of</strong> any step in the SCF cycle. In<br />

the oscillatory case, however, the diagonalization <strong>of</strong> the Fock matrix created using P (b) (i.e.,<br />

F (b) ) gives a density matrix indistinguishable from P (a) . Thus, the SCF simply bounces back<br />

and forth from P (a) to P (b) and never converges. This behavior can be recognized easily by<br />

looking at the SCF energy for each step, which itself bounces back and forth between the<br />

two discrete values associated with the two different unconverged wave functions defined<br />

by P (a) and P (b) .<br />

In more pathological cases, the SCF behaves even more badly, with large changes occurring<br />

in the density matrix at every step. Again, observation <strong>of</strong> the energies associated with<br />

each step is diagnostic for this problem; they are observed to vary widely and seemingly<br />

randomly. Such behavior is not uncommon for the first three or four steps <strong>of</strong> a typical SCF,<br />

but usually beyond this point there is a ‘zeroing-in’ process that leads to convergence.<br />

In the abstract sense, converging the SCF equations is a problem in applied mathematics,<br />

and many algorithms have been developed for this process. While the technical details are<br />

not presented here, the process is quite analogous to the process <strong>of</strong> finding a minimum<br />

on a PES as described in Chapter 2. In the SCF problem, instead <strong>of</strong> a space <strong>of</strong> molecular<br />

coordinates we operate in a space <strong>of</strong> orbital coefficients (so-called ‘Fock space’), and there<br />

are certain constraints beyond the purely energetic ones, but many <strong>of</strong> the search strategies<br />

are analogous. Similarly analogous is the degree to which they tend to balance speed and<br />

stability. Usually the default optimizer in a given program is the fastest one available, while<br />

other methods (e.g., quadratically convergent methods) typically take more steps to converge<br />

but are less likely to suffer from oscillation or other problems. Thus, one option for dealing<br />

with a system where convergence proves difficult is simply to run through all the different<br />

convergence schemes <strong>of</strong>fered by the electronic structure package and hope that one proves<br />

sufficiently robust.<br />

In general, however, it is more efficient to solve the problem using chemistry rather than<br />

mathematics. If the SCF equations are failing to converge, the problem lies in the initial guess<br />

(this is, <strong>of</strong> course, something <strong>of</strong> a truism, for if you were to guess the proper eigenfunction,<br />

obviously there would be no problem with convergence). Most programs use as their default<br />

option a semiempirical method to generate a guess wave function, e.g., EHT or INDO. The<br />

resulting wave function (remember that a wave function is simply the list <strong>of</strong> coefficients<br />

describing how the basis functions are put together to form the occupied MOs) is then used<br />

to construct a guess for the HF calculation by mapping coefficients from the basis set <strong>of</strong> the<br />

semiempirical method to the basis set for the HF calculation.<br />

When the HF basis set is minimal, this is fairly simple (there is a one-to-one correspondence<br />

in basis functions) but when it is larger, some algorithmic choices are made about<br />

how to carry out the mapping (e.g., always map to the tightest function or map based on<br />

overlap between the semiempirical STO and the large-basis contracted GTO). Thus, it is

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!