There was one day the bus was late so I drove in with a grad student who did density functional theory calculations of MOFs and asked him "How do you make a MOF?" and he said "Beats me, I'm a theorist" so I figured that I wanted a quick answer to that one myself and it turned out to be "mix up the ingredients and bake them in the oven"
I'll say grad students today often seem a bit sheepish and inarticulate. They're facing a very competitive market so I think it behooves the theorists to be able to talk about experiment a bit and vice versa. On the other hand they have a big hill to climb to just be successful with DFT.
One of my few regrets in grad school is that I didn't take a course in DFT, not like I was really going to use it, but DFT is an example of the kind of very complex calculation which takes a lot of care to apply. I got a little of this art from Sethna's class in renormalization groups and such but it was really
by Chandrasekhar that taught me how to organize the kind of complex calculations that might involve numeric integration differential equations, using a computer, etc -- extracurricular for a cond-mat PhD but really a lot of fun.
I just made a breakthrough in selfobject technology (enough of a reformulation that I can take back the ideas that my evil twin published in such a way that I couldn't ever publish them under my name) and managed to get the evil out of my evil twin and I've been practicing "radiance drills" that get me into a state where I can really draw out 30y+ people but how it works with "kids these days" is an open question because since the pandemic grad students mostly seem like damp squibs -- I gotta give it a try.
I do regret I didn't figure this out much sooner (if I had I wouldn't have some things on my chart I do now) but right now I having so much fun I think other people should be jealous.
Have you asked LLMs about the VASP codebase? With special reference to Tomas Arias' flow-chart(s) [0]
Grad students these days will need to figure out their Altman-informed version of radiance (maybe easier if they spend more time on some dank corners of HN).
I'm jealous you found Chandrasekhar :) going to look at this to train understanding of human and/or AI pedagogy.
I too have my favorite Dover books (Lanczos, Yaglom, the "Green books") but I can't say I got any general techne out of them (yet)
These are the underappreciated infra born of the preAI cold-war times (doubt postsoviet mathematicians have an equivalent. Especially not "Bourbaki" =P)!
Worth recommending to gen-A/Z so they have an inkling what semifunctional social infra looked like before *verflow/*stackexchange
