tag:blogger.com,1999:blog-9633767.post754008062200760061..comments2024-03-25T09:11:17.877-07:00Comments on The Curious Wavefunction: Nobelist John Pople on using theories and models the right wayWavefunctionhttp://www.blogger.com/profile/14993805391653267639noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-9633767.post-40800484216441991052015-04-20T16:15:35.601-07:002015-04-20T16:15:35.601-07:00Oh, that Gaussian could be one of those culled!
T...Oh, that Gaussian could be one of those culled!<br /><br />The price is high for academics and worse for those who can't prove academic affiliation. The input format is fairly dreadful to write or read by hand. The parallel scaling is poor though serial performance is quite good, if you are one of the five computational chemists in the world who does not yet have a multicore computer.<br /><br />On the plus side it has a lot of capabilities.<br /><br />Its biggest advantage isn't broad capabilities though. A bevy of programs have (e.g.) the ability to do a geometry optimization followed by frequency calculation with popular DFT functionals and basis sets. Buying Gaussian for that is like buying a full Photoshop license to put captions on funny cat pictures.<br /><br />No, the seductive appeal of Gaussian is ease-to-replicate: its software defaults are the defaults of the thousands of prior publications that used Gaussian. How are core orbitals chosen for freezing in an MP2 calculation? How are grids generated for DFT? It's too much work to replicate prior Gaussian results to the microhartree in another program. Just buy Gaussian and don't think about it. Everyone else does.<br /><br />On the one hand it irks me that Gaussian so dominates published results even when calculations could have been done in many programs. On the other hand I can understand that non-specialists in particular may not wish to save a few thousand dollars at the expense of a lot of additional learning. And on the gripping hand I wonder how robust results are when cross-software replication is attempted less frequently than it should be, and implicit software choices lead to different results.Matthttps://github.com/mattbernstnoreply@blogger.com