It's coming along, but I haven't really gotten much done since about Thursday of last week. Didn't run anything over the weekend.
Haven't tried running anything with dispersion again yet. I think I may try a different flavor of dispersion, on the off chance that it's something about the BJ version that's giving it trouble.
Ultimately though, I probably need both a higher level of theory, as well as anharmonic corrections, in order to get to quality enough of spectra for my purposes.
Neither of which are particularly feasible at the moment.
<nod>, I'm certainly inclined to anticipate it's being useful.
The problems it was having with dispersion seemed reminiscent of difficulties I've had with COSMO in ORCA... numerical instabilities due to rounding errors or somesuch in the method.
With COSMO it's the slight differences in the cavity construction as the geometry adjusts bit by bit.
hmmm... solvent models tend to oscillate because they are applied post-calc and then reoptimised... that shouldn't happen with dispersion, which is added directly to the energy
I just had a project which started about 3 years ago. I used theory without dispersion to keep it consistent... a couple of days I restarted the whole project as non-disp geometries are just bad and offer a completely different interpretation
@hBy2Py Theoretically you converge the SCF, then you add dispersion corrections, then you determine gradient etc. This should not affect convergence much...
maybe dispersion is a bit more sensitive to small changes in geometry
Oh, I guess the big thing is, I never did rerun a dispersion optimization after rigging up a little widget to plot the gradient and energy values over the course of the optimization.
it's entirely possible that being a Cartesian optimization, it was just taking a very roundabout path to get to the minimum. I should recheck that.
After our discussion, I've been doing everything in cartesians.
It's been weird, after working in ORCA for so long, having to reexamine all of my acquired assumptions about the best practices for running computations.
yes. I always suggest to run with the defaults first, that's what the designers/developers optimise first. Gamess is a nice piece of software, but it carries around a whole bunch of old code and probably incomplete features...
I remember reading somewhere that using RI in gamess actually slows down calculations, which is somewhat absurd since RI was developed to speed things up
it has something to do with the implementation of grids and such... I don't remember...
being in academia, i can pick the best axe to battle a problem (usually) for free
A software that uses arrays as direct input is not meant to be understood by the black box user...
did you read my post on fixing things in gamess?
I found it incredible, that you use a single array to define all the redundant coordinates.... who programs stuff like that???
I guess 20 years ago you didn't really have a choice, but why don't you update these things?
@hBy2Py Oh the "manual"... The formatting alone is terrible... it's like a readme that's five thousand lines long... sometimes I feel like they grepped out all comments of the source code to compile the manual...
Of course, I have no idea how often people respond off list because, well, I don't get those. But there definitely have been times when someone has effectively forwarded a message back to the list that someone sent offline, and so I've seen those.
I'd like to be able to say that people put anything contentious off list, but there's been more than enough... Shall we say, raucous conversation... That has happened on-list, so...
@Martin-マーチン I was a bit sad that that happened when LordStryker took a new position and somewhat vanished... occasionally though he comes by and drops a hi, which is nice...
It's much more an ennui thing for me, though, unfortunately. Too many questions that I look at and just can't bear the thought of trying to type out an answer, only to have comments dropped that indicate that my effort hasn't really clarified anything.
If I wanted to tutor someone, I would tutor a kid in my neighborhood, or at least in my town. Stack exchange Q&A is pretty much the worst place for that.
If the overlords saw fit to add a privilege so that I could invite sub-20 reppers to private chat rooms, tutor them there, in real time... That might change the calculus a bit.
@Martin-マーチン paracetamol almost died of starvation while Jan was in transition
@hBy2Py I'm always interested in learning new ways to do things, though, but right now I'm more interested in getting my results in before christmas and taking my gf on a ski trip
@Martin-マーチン @hBy2Py if either of you has time for a quick question...
@gannex Gaussian is a fine suit for many purposes. There are others that do better in some areas, but no need to run off and learn every available suit. Also, if you have questions regarding Gaussian, you are also welcome to ask them here. Just that we are on gamess right now doesn't mean we can't diverge...
I've done a bunch of TD-DFT calculations to generate theoretical uv-vis spectra to compare to some titrations I've done (forming complexes between Cu(I) and various potentially redox noninnocent ligands)
I'm having a few problems
for one thing, I can't get the TD-DFT to work with PCM, so I've used geometries that were optimized with PCM (for tetrahydrofuran), but when I put scrf=solvent=tetrahydrofuran in the TD-DFT calculations, they just never converged
but my prof says he's happy with the results withought the pcm
so I just want to generate pictures of the electron densities of the most intense transition for each calculation
a former colleague made these things called "difference electron density sketches", but I don't know how she generated them
my understanding is that it's supposed to show you where the electrons go during the transition. The difference between the electron density of the ground state and the excited state.
1. choose > tools > orbitals > render molecular orbitals, and pick the orbitals you need
2. then you will have in the right column the picker (pick a orbital to display it), where you have the display options, like countour value, style, etc... (I send a screenshot)
sadly I just realized I don't actually have chemcraft installed on this computer (it's on my laptop, but I left it at work), so I'll have to do that first :- /
pick one mo on the left, and another on the right, and then you will probably want to use the operation "subtraction of squares", click okay and the you should have a new item in the picker
if I make these figures, all I have to do is go clean up my fumehood and then there's no research until after I go skiing with a beautiful lady from germany
Mont Tremblant or Le Massif, or Possibly Jay Peak in Vermont. These are very shitty mountains compared to Europe or the West Coast though, but the snow is beautiful
I'm more into cross country skiing though. That way I can ski around on a small mountain behind our chalet and then go inside and light a fire
yes, cross-country would be nice, but I have to go way north for that. I can downhill in Fukushima or Tochigi, that's only a couple of hours away
the resorts are not very challenging, and rather small, but it's awesome to go out of town and into the mountains just for a break of the city noise
i learned cross country when i was a tiny boy of about 5/6 years... back then i hated it... no action, too boring... now it would be heaven
❤
but I also have a vacation coming up, so that'll be fun, too... part of it will be in Tokyo, and part will be in Hiroshima, and then celebrating new year at home, and the rest i have not yet figured out :D Something will turn up
same here. Small mountains. Everybody spends their time getting good at sous-bois (glades) or freestyle since the trails are not that long
I think cross country is the sort of thing that becomes better as you get older. The best skiers I know are usually middle aged people. I'd love to learn to shoot though and do biathlons or something.
But having a nice trail really helps. In Montreal, we have a mountain with nice cleared tracks alongside a wide, flat path. So you can actually skate ski, then switch to classic if you get tired.
there's no better feeling than skating downhill along the path on the way down the mountain
just because I read Montreal, it reminded me; the next WAToC is in Canada - although on the other side in Vancouver in roundabout three years...
If you have a chance you should go :D
It's fun
@hBy2Py btw. don't know if you noticed already, the Japanese course on DouLingo was moved to desktop... I only found out today, because I kinda stopped learning O.o
so I just have to figure out where the admin stores the gaussian binaries and then put that directory in my $PATH, and I should be able to run formchk locally?
@gannex I made a mistake earlier, sorry. Check if echo $PATH | grep -i g09 gives any result. It worked for me, because my versions are stored in /home/chemsoft/gaussian/{g09d01,g09e01,g16}. Alternatively, just inspect your path and look for anything that could conflict
@Martin-マーチン No!!! Did not notice, thanks for the heads-up! I have smertfon now, so have just been on the mobile version, but will gladly try PC. Does PC have speaking exercises? Mobile doesn't, a/o the last time I used it.
Yep. And, really, you NEVER run only one calculation...unless you really enjoy pestering people, it's going to be an awkward process getting anywhere useful.
@Mart Do you see a big difference between the zero-damped and BJ versions of Grimme dispersion?
Also, I assume I should be using hybrids for opt of organics?
@Mart Huh. Re-OPT of PBE-D3BJ/PCseg-1 from the PBE/PCseg-1 optimized geometry converged in 6 iterations. Perhaps it would have eventually made it there running PBE-D3BJ from the outset.
= PCseg-n - Polarization Consistent basis sets.
n = 0,1,2,3,4 indicates the level of
polarization. (n=0 is unpolarized, n=1
is ~DZP, n=2 is ~TZ2P, etc.). These
provide a hierarchy of basis sets
suitable for DFT and HF calculations.
Available for H-Kr.
Doesn't say whose they are
So, I'm guessing it's approximately like def2-SVP
I believe they're the segmented-contracted versions of the pc-# bases.
Optimi(s|z)ing optimi(s|z)ing
Optimi(s|z)ing optimi(s|z)ing