Btw note that Gaussian has J functions between I (L=6) and K. Many other programs skip J, and set K as L=7 whereas in Gaussian J is L=7 and K is L=8. — Susi Lehtola 2 days ago

Thanks @SusiLehtola ! Long time no see! I've edited my answer to say that Gaussian doesn't allow treatment of L-type orbitals and beyond, while open source codes like OpenMolcas, MRCC and Psi4 do treat them! — Nike Dattani 2 days ago

I enjoyed reading this answer. Thank you for this. — dval98 2 days ago

@dval98 thanks so much! I look forward to your first post here if the opportunity ever arises! — Nike Dattani 2 days ago

@NikeDattani I just recently stumbled across matter modeling. I hope to become more involved in the community. — dval98 2 days ago

@dval98 That would be great! We look forward to welcoming you warmly! — Nike Dattani 2 days ago

I would highly disagree about interest / enrollment in chemistry and physics. Maybe for you, but here, there is continually increases in enrollment. Yes, CS and Statistics are increasing, but we have a hard time creating enough lab sections for students. — Geoff Hutchison 2 days ago

For what it's worth, while VASP and Gaussian may have unusual licenses, I don't think either of them are in financial difficulty. Indeed, I think many people pay their licenses precisely because they include GUI and support. — Geoff Hutchison 2 days ago

Matlab has the huge issue of being proprietary and very expensive. They might sell it to universities at a reduced price, but that is "to get people hooked". In addition, Matlab is slow for calculations. Massive multi-purpose software often does everything and nothing very well... And you have forgotten ORCA in the QC software list :). — DetlevCM 2 days ago

I'm from a different field but you'd be surprised... OP has a point here. It absolutely DOES make a difference to students whether they're greeted with user-friendly GUI stuff or have to learn all that CLI black magic - aka the reason they're not lining up there and going for something more glamorous instead. I know a whole bunch of dropouts who dropped out solely because of having to code more than they've bargained for, or at all. You might thing "to hell with those weaklings", but this is quite a systemic problem in academia which industry never gets because... priorities. — Lodinn yesterday

Thanks so much for your comment @Lodinn, what is your field? — Nike Dattani yesterday

@GeoffHutchison I was worried when I typed that part about decreasing interest in chemistry and physics, that I didn't have a good reference for that. Someone told me a few years ago something like "it used to be that the top high school students went into physics, but now they go into CS and economics", and I have at least noticed here that interest in CS (and in my class, surprisingly, also actuarial science) is growing far faster than I had previously imagined it would. The requirements to get into CS here are now at around 98% (high school average) whereas it's closer to 70% for physics. — Nike Dattani yesterday

I do not think that Gaussian being or not a state-of-art is an issue in the given context. People who need a GUI to enter a basic Gaussian input are not the people who need CCSDT or things like that. The speed of calculations will also be mostly bottlenecked by the incompetence of the researcher (wrong models, wrong approaches), not the actual speed of the code. — Greg yesterday

@Lodinn My experience shows otherwise. Glorius fancy colorful GUIs are nice, but 99% of our students after these labs have only a vague recollection of what exactly they have done. They strongly enforce the usual "I don't know and I don't care" attitude of experimental chemists to details of computational work while boosting their false self-confidence on the topic. Graphical programs have a role in creating initial geometries or review results, especially for beginners, but fully automated solutions have very limited educational value imho. — Greg yesterday

Matlab has the huge issue of being proprietary and very expensive. They might sell it to universities at a reduced price, but that is "to get people hooked". In addition, Matlab is slow for calculations. Massive multi-purpose software often does everything and nothing very well... And you have forgotten ORCA in the QC software list :). — DetlevCM 2 days ago

@Greg I strongly feel like setting the bar at "these are tools of the trade, learn them or you're out" is harmful for at least modern physics. Old-timers went from the field-specific things to programming (I need X so gotta learn how to Y), students see it the other way around. This is indeed an educational issue from both sides, and I don't think one-button eye candy solutions are great, either, but some fields have figured out both. X-ray powder diffraction, genomics pipelines... Again, obviously you know your field better but I still don't think it should be a must for all labs/subfields. — Lodinn 13 hours ago

@NikeDattani Now I work in spectroscopy/remote sensing, but seen this in theoretical physics and astronomy previously as well. Students are fine with Mathematica and a-fine with Matlab but some would quit literally over having to run FORTRAN simulation code manually. — Lodinn 13 hours ago

@Lodinn Then what is missing from the aforementioned (visual) tools? Learning Mathematica at any decent level is not easier than most command-line tools, neither much cheaper than most commercial GUI (actually many, like Avogadro is for free). I would argue that any theoretical physicist or astronomer who thinks that running command-line software is beyond their skillset and considers it a significant challenge to handle is like a surgent who is afraid of blood and want to play only on dolls. — Greg 7 hours ago

I'm from the much less tech literate (compared to physics) field of biology, and I think most of the teaching effort here is to make the case to students that doing this hard thing (learning programming skills) pays dividends in the rest of their career. I'd also argue that, at the very least, a GUI that doesn't export a script to include in your paper doesn't meet the standards we're trying for with result reproducibility. — lupe 5 hours ago

@lupe "at the very least, a GUI that doesn't export a script to include in your paper doesn't meet the standards we're trying for with result reproducibility" what does that mean? — Nike Dattani 4 hours ago

@NikeDattani sorry, was writing pre coffee, and wasn't the clearest. My argument is that, without a clear record of what was run, your results are not reproducible. A list of selected settings on a GUI in your paper has a much, much higher chance of human error, either by omission ("this setting isn't important to my experiment") or by forgetfulness. In contrast, a script is pretty fixed. It will probably run the same way if you run it on the same software version with the same raw data — lupe 3 hours ago

@lupe I agree that a script is better than a list of GUI settings. Are you arguing in favor of command-line vs GUI for biology? — Nike Dattani 3 hours ago

@NikeDattani sort of :P I'm saying that there are standards that should be met when using GUIs, and that many GUIs I've seen fall short of the "Allow you to easily access, save, and add the script to your paper" standard. I saw someone mention genomics pipelines further up, and I think these are a great example, Many of them are a black box, making it hard to work out the assumptions they're using, and become accepted in the field, because they "just work". My standard is "can you easily rerun your data when an automated pipeline is found to have a critical bug", and most don't meet that. — lupe 3 hours ago