Good day Dear Researchers, I just read a paper that was published in 1998, where the electronic states of a porphyrin containing iron was investigated via Hartree-Fock, using STO-3G for carbon and hydrogen, and 3-21G for the other atoms including iron.
The calculations were performed in a super-computer of that time, which nowadays that may be only just a node or less of a current local cluster of a QM research group.
The authors explained differences between calculated and crystal geometries, and attributed errors to the lack of polarization functions in the basis set.
Also, they analyzed spin states, and concluded based on energy differences.
It was interesting the way they wrote in that time, very similar to the way we write today:
"... our HF/STO-3G method reproduces well experimental results ..."
Today reviewers of a journal are more strict in how we use the methods, but it seems a common practice then in the 90s and now:
"... our QM approach reproduces well the experimental evidence ..."
Apparently being this phrase the one that measures the quality of our work, no matter if we use HF/STO-3G or modern DFT with extended and refined basis sets.
I guess people know what
happend with the duality problem:
Two camps, one has wave and particle together, the other
Seperately.However its clear both are there, whatever relation.
There are two things that I usually think about the comparison of theory and experiments: 1) nature considers every variable we forget, since, we are just trying to crudely model what is happening around us. 2) for most problems nature has better statistical ensembles than we could ever model.
Anyways, we should be humble enough to remember that the ultimate goal of theory is describing and predicting phenomena and that nature is the dataset we should aim to reproduce (whenever possible).
As it can be easily verified, I wrote papers in the 90s and I write papers now And both then and now the papers I work on, which are always founded on a an experimental basis, largely rely on a computational interpretation.
Over the past decades, I must say that the framework of calculations has not changed that much. My papers at the time (end of the 90s) already used DFT with hybrid functionals and basis sets with polarisation and diffuse functions. These have not been introduced so recently as younger researchers may think. And I add that I used Gaussian (G94) as well as I use Gaussian (G16) today. It was not common at that time to use HF calculations and even less STO-3G basis sets (it never occurred to me, for example), so I guess that the paper mentioned by Pablo Mtz is not very representative of the time. The intrinsic limits of HF calculations and those of poor basis sets were well-known even at the end of the past century, and very-well explained in theoretical chemistry courses.
What has changed, and it has changed a lot, is the time required for calculations and therefore the possibility to model large systems without introducing often-drastic oversimplifications.
What surprises me is that even today the good old B3LYP functional, which I used when I was younger, and that has well-known limitations especially when transition metals are concerned, is still so largely adopted…
- Badges
- Science topic