Crystal26/SCANZ at Barossa

I’ve immersed myself for the last 2 days in the Australia/NZ crystallography meeting – about 100 scientists – some old acquaintances. It’s been wonderful. Some of the imaging and related techniques have been awesome – instrumentation has moved on so much. And, of course, this means that data is more and more critical.

It’s been great to see experiments where theory and experiment are compared (this is almost completely unknown in modern chemoinformatics where there is little theory and virtually no experiment).

I’ll pick just one – quantitative diffuse scattering from Harald Reichert (ESRF). Don’t switch off – as it relates to chemistry. Because as atom moves in a crystal it experiencese the chemical and lattice forces of its neighbours. Since diffuse scattering can now be measured with a dynamic range of (IIRC) 5 orders of magnitude we can see places ikn the crystal where atoms occur rarely and this maps out the anharmonic potential. By working in reciprocal space it’s possible to observe second and third-order interatomic forces.

So what does this have to do with chemistry? Well, for years, according to Harald, the DFT experts have been telling the experimentalists that theory (“first principles”) is so good that experiments are irrelevant.

Well the latest synchrotrons are showing the reverse. That theory is all over the shop compared with experiment. And of course it’s not “first principles” – it’s a folklore of basis sets, functionals and pseudopotentials. So maybe the “correct” DFT parameterisation could be better determined from experiment. [Note this is for alloys at present, but in principle it can be done for ionic solids.]

I miss not doing experiments.

This entry was posted in Uncategorized. Bookmark the permalink.

One Response to Crystal26/SCANZ at Barossa

  1. bill brouwer says:

    Peter, an inspired post as always, wish I could have been in the motherland to hear you speak.
    I think you made similar remarks to me at M’soft escience ’08, after I delivered a talk on using machine learning to learn structural models directly from solid state NMR spectra + ab initio data. I think this is the future of spectroscopic sciences; a marriage of experiment(s) and ab initio, with computer science as the celebrant.

Leave a Reply

Your email address will not be published. Required fields are marked *