The Periodic Table

4:22 AM

hey @pentavalentcarbon! Thanks again for your tips the other day about CASSCF calculations. They were incredibly useful, and I used your thought process to successfully run CASSCF jobs for FeS and Fe2S with no problem! However, I'm having some issues with FeS-, and I can't diagnose why. I've switched to Molpro, and the job has just been running for days with 9 active orbitals (the equivalent job for FeS took minutes).

@pentavalentcarbon: If you have a few minutes, would you be able to take a quick look and point me in the right direction if you see anything that stands out? The CASSCF output is here, the input is here, and the completed output of a HF and UHF job can be found here in case that gives a bit more insight into the orbitals.

@pentavalentcarbon: and like I've already said, I can't thank you enough for the really thoughtful advice. It was a huge help

This might take me a while to figure out, since I am no Molpro expert, but in the CASSCF calculation I already see a problem before the iterations even start.

Node minimum: 2.359 MB, node maximum: 8.389 MB

My god that's a tiny amount of memory.

user116211

@pentav o/

o/

user116211

I've got an interesting book today ;)

o/ :D
@pentavalentcarbon: don't worry about it if it ends up being too cryptic to sift through. I can try upping the allotted memory -- usually it gives me an error message when it needs more

user116211

4:27 AM

@pentavalentcarbon QM is fundamentally non-local ;)

@AndrewR. Ok, that's fine, I'll let you know if I get confused. 975 variational parameters for CASSCF is tiny, so iterations would be slow if your orbital space is bad.

Which I think it might be, because the gradient w.r.t. the CI coefficients is blowing up.

0.10553764 in the last iteration.

user116211

I've got an old piece of Mathematical Physics by Courant and Hilbert without a single page being torn \o/

Hmm, UHF natural orbitals aren't really useful. You'd learn a lot more from MP2 natural orbitals.

@pentavalentcarbon ah, okay, that's good to know. my gut feeling is that the active space isn't great, but I wasn't really sure how to determine that. I'll play around with the active space some more

user116211

Did I welcome you @Andrew?

user116211

4:29 AM

!!greet/@AndrewR.

Chemobot

Welcome to The Periodic Table @AndrewR.! Here are our chat guidelines and it's recommended that you read them. If you want to turn Mathjax on, follow the instructions in this answer. Happy chatting!

user116211

;D

also I have no idea what a natural charge orbital is, I'll just look at the UHF MOs...

hi @MAFIA36790! thanks for the greeting! :) and for the chem bot fella

@pentavalentcarbon oh okay. I was told to do a UHF calculation, but honestly I don't know enough about why that might not be best to look at

The reason we try to avoid UHF orbitals (though DFT might be a little better) is because they're missing what the MCSCF is trying to capture: correlation.

Oh, the natural charge orbitals are the UHF NOs, ok...

user116211

4:33 AM

Oh! man!!! I've to borrow this.... amazing pages on Lebasque integral ;}}}

@pentavalentcarbon ah, okay, yeah that definitely makes sense. I was using it just to get a quick first-glance, so to speak, but if MP2 would still be pretty quick but more insightful I'll try using that in the future

Also, MO coefficients give much less insight than a population analysis, because the population analysis comes from the electron density. MO coefficients are less clear. 7.1, 8.1, 9.1 11.1, 12.1, 13.1, 4.2, 5.2, 4.3, 5.3, and 1.4 all have large d coefficients for UHF NOs. That's 11 orbitals, but again, not super clear because the values might change once you form $P_{\mu\nu} = \sum_{i}^{\textrm{occ MOs}} C_{\mu i} C_{\nu i}$.

A couple of other things.

user116211

@pentavalentcarbon You should write a book when you are old; you have the natural grasp of highlighting a fact in the best way; you need to consider that :)

Once I graduate I'm done with science.

user116211

!!greet/@Gems9

Chemobot

4:39 AM

Welcome to The Periodic Table @Gems9! Here are our chat guidelines and it's recommended that you read them. If you want to turn Mathjax on, follow the instructions in this answer. Happy chatting!

user116211

@pentavalentcarbon you wouldn't be in academia o.O

@MAFIA36790 I don't care, I'm completely over it at this point.

user116211

oohhh... :((

@AndrewR. 1. Molpro is...not easy to use. I personally find the manual, examples, and output unintuitive. I mean, maybe it's because I've only used it for a handful of things, but I think it's user-hostile.

2. It's a very good habit to run all your MCSCF-type calculations in at least two steps, one for generating your starting orbitals that you look at and to serve as a good guess for your MCSCF, and a second for the MCSCF itself.

I think I went into much more detail about that in the post. I cannot emphasize that enough.

Let me post a figure that shows what can happen even when you're being careful with choosing your active orbitals.

@pentavalentcarbon good to know about the MO coefficients. that's indeed what I was using. will try a population analysis as well. as for molpro, yeah... the web resources are terrible compared to orca at least. and a lot of things are very counter-intuitive. I'm having an issue with orca where I can't read in molecular orbitals, and I think it may be a bug of some sort, so I tried something else. will probably switch back to orca at some point

ah, yes, so what I did was I ran a UHF calculation, looked at the orbitals, and then used that information to choose my active space. and that help…

4:46 AM

> I'm having an issue with orca where I can't read in molecular orbitals, and I think it may be a bug of some sort, so I tried something else. will probably switch back to orca at some point

I have some choice vulgarities for this. If you need help with it, I can tell you what to do, since I've done it so many times.

If you don't do everything just right, you get garbage like this.

@pentavalentcarbon nvm I see what you're saying. yeah that's essentially what I did with the separate UHF job to get the occupation for each symmetry and to choose my active space.

lol yeah it's been frustrating to say the least. my post on the orca forum is [here](https://orcaforum.cec.mpg.de/viewtopic.php?f=8&t=2244), but I kind of just gave up since everything *seemed* to be right

oh wow. yeah, I figured it'd make a big difference

Your input looks ok. The base name of the *.gbw file isn't the same as the basename for the 2nd calculation by any chance, is it?

nope I made sure of that. totally different basenames

Ok. #1 source of problems starting out...

yeah I had that foresight for that one, thankfully :)

4:55 AM

agh what a dumb POS, I forgot it chokes on orbitals from different versions...even 3.0.2 vs. 3.0.3...

I know!! so frustrating. I just installed 3.0.3 the other day. I can re-run it. probably should do that anyway

Ok, I get the same problem with 3.0.2. The size of the gbw file seems reasonable, but something is definitely wrong. I get an error code of 134, which means nothing to me, but std::bad_alloc could probably happen for a couple of reasons.

phew... I was "concerned" it was going to work perfectly and I was going crazy all this time

I would just re-run everything 3.0.3, because old versions are full of bugs.

That being said 3.0.3 is full of bugs too.

ORCA is full of bugs.

Gems9

@MAFIA36790 Thanks Mafia :)

5:00 AM

yeah I will. I remember trying this again on 3.0.3 and having the same issue but didn't save the files. I should double-check though

@pentavalentcarbon so would you recommend I run an MP2 calculation, look at the natural orbitals, and then set up my active space appropriately and run a CASSCF calculation (taking the MP2 orbitals as a starting guess)? is that a reasonable workflow in this case?

Yes. If the MP2 NOs aren't in the right order, you might need to reorder them too.

okay, and the "right order" being monotonically increasing in energy as occupation decreases?

anyway, that gives me something to play around with. I'll take a stab at that tomorrow and am hopeful it'll provide a bit more insight! thanks again for the help/advice. electrons are weird

@AndrewR. The right order meaning the 1. "chemical identity" (d-orbitals, etc.) of your desired orbitals and 2. orbitals with 0.02-1.98 population, all go in your active space.

ah okay yes that makes a lot more sense and has been what I've been doing. sweet

A: Why does lithium move on the surface of water during the reaction?

7:42 AM

@TIPS Use RSS feeds for that

TIPS

1

This is a very speculative answer to your comment on spinning. The spinning presumably occurs because the under surface of the lithium is not flat and also the metal itself is not a perfect cuboid so will not be levitated evenly by the gas produced but will be tilted w.r.t. the surface. There wil...

@Hippalectryon I want something on chat.

Oh, scratch that.

I demand something on chat.

(┛◉Д◉)┛彡┻━┻

@TIPS Use the chat's RSS feed handler

It will auto post XKCD

chat.stackexchange.com/rooms/info/38172/chemobot?tab=feeds

^ look at this feed

(url is xkcd.com/rss.xml)

user116211

8:38 AM

Hey @hippa.

@MAFIA36790 \o

user116211

So, the Oral monster is still giving shit to you?

My orals are finished :P finally

user116211

9:32 AM

@Hippalectryon WHHHHAT?

user116211

\o/

user116211

!!flip/Oral

Chemobot

(づಥਊಥ)づ︵Oɹɐꞁ

9:44 AM

Results will be out in a few days :(

2spooky

user116211

11:37 AM

@Hippalectryon chill, dude ;)

;_;

user116211

@Hippalectryon what, again?

user116211

stop fearing.

user116211

@hippa so, what's new are you reading?

Nothing really, I'm waiting for the results while boxing most of my stuff

And watching anime

user116211

11:43 AM

Should I name you some interesting topic to read?

user116211

You'll enjoy it ;)

Ok

user116211

12:03 PM

So, presenting before you

user116211

Axiomatic Thermodynamics

user116211

It's a whole lot different than the conventional thermodynamics we learn generally.

user116211

In a word, you can say,

user116211

Thermodynamics has been rigorously put down in mathematical language

user116211

> [...] conceptually the easiest and clearest formulation of the whole subject of thermodynamics was written down by Carathéodory - everything is given by just two axioms, and the subject reduces to differential geometry, with absolute temperature being an integrating factor for heat transfer that makes the change in entropy an exact form.

user116211

12:08 PM

this is really interesting to read.

user116211

You can start from the original work of Carathéodory:

user116211

neo-classical-physics.info/uploads/3/0/6/5/3065888/…

user116211

This is really gonna be amazing to read this @hippa ;))

user116211

The paper is quite easy to read and self-contained.

ok :D

1:25 PM

@MAFIA36790 How come (I'm at the beginning) the theory says that at fixed volume and pressure, the composition of a phase makes it unique ? Doesn't for instance the spacial arrangement of the components have an influence ?

(bottom of page 3- top of page 4)

user116211

@Hippalectryon Temperature, volume, and pressure, for instance, are not independent quantities. That is $$f(p,V,T) =0\;.$$ Any two of the three quantities are sufficient to determine the state of the system.

user116211

Now , you can get the statement in the article.

user116211

You may take Fermi's Thermodynamics as a reference.

user116211

It is explicitly discussed there at the first chapter.

So, the state of a phase is unaffected by - say, for a cristal - its structure as long as the composition in its elements is the same ? (since it does't change either p or V or T)

user116211

1:35 PM

I've to attend a call; coming within 5 mins.

user116211

2:01 PM

sorry, I'm back.

user116211

@Hippalectryon Re your comment, I would say two quantities are enough to determine the state of the system iff the system is homogeneous.