14. I get a bad structure from ZINDO/1.

It turns out that ZINDO/1 often does a better job with 2nd row transition metals than 1st row transition metals. This seems to be a "feature" of Michael Zerner's INDO/1 method (called ZINDO/1 in HyperChem) that was originally implemented in the ZINDO program from Zerner's group. Some ZINDO experts use different sigma-sigma or pi-pi overlap weighting factors to get better results. One user has reported problems that are due to a poor initial guess for the wavefunction; he got a reasonable structure when he started from one with short bond lengths, but a subsequent single point did not converge. We are continuing to investigate this behaviour, but in general the ZINDO methods are not as robust for transition metals as AM1 or PM3 are for non-transition metal complexes.

We have tried very hard (with lots of generous assitance from Tom Cundari and David Baker) to reproduce results of INDO/1 or INDO/S calculations that use Michael Zerner's ZINDO program (including the choice of the initial guess for the wavefunction). The optimization algorithms and wavefunction convergence acceleration options may differ from the ZINDO program, but we are confident that an INDO/1 (or INDO/S) potential surface will be same whether calculated from HyperChem, ZINDO or Argus (Mark Thompson's program).

While not very satisfactory, we would advise caution when using ZINDO/1 for geometry optimization. In some cases the model builder, MM+, or even "hand-editing" will give better structures. ZINDO/S (INDO/S) was not developed for geometry optimization and should not be used for any geometry optimization.

One should check, via a log file or by monitoring the status line, that all iterations in the optimization actually reached wavefunction convergence. Sometimes better results are observed by turning "Accelerate convergence" off (or on!).

[Products] [Sales] [Support] [Science] [News] [Corporate] [Search] [Home]