The Periodic Table - 2016-04-21

« first day (1456 days earlier) ← previous day next day → last day (3230 days later) »

Loong

00:33

The laws of physics are fun to try to understand, but as an organism with incredibly delicate eyes who evolved in a world full of sharp objects, I have an awful lot of trust in biology's calibration of my flinch reflex.

5 hours later…

CowperKettle

05:41

> In those circumstances, if you carried out electrophoresis on the unmodified solution, there would be a slight drift of amino acid towards the positive electrode (the anode).

But isn't the positive electrode called "cathode"?

2 hours later…

user116211

07:51

user image

3

user116211

Book - An Introduction To Algebraic Topology by Joseph J. Rotman

user116211

@CowperKettle There is difference between Electrochemical cell and Electrolytic cell.

Freddy

08:29

@MAFIA36790 You can check out NIT's, some of them have dual degree in physics!

user116211

@Freddy I'm aspiring for IISER.

user116211

And NISER also.

Freddy

aha great

Do they have their own entrance exam?

user116211

And NISER also.

user116211

@Freddy ammm... JEE advanced.....

user116211

08:33

They also conduct an Aptitude test.....

Freddy

oh okay, Good luck for advance! :-)

user116211

@Freddy :))

user116211

09:09

@Ramanewbie: o/

Ramanewbie

09:29

Hey !!greet/@MAFIA36790

user116211

09:40

@Ramanewbie Chem is not here :(

skill patrol

10:00

:(

2 hours later…

Ramanewbie

11:46

@MAFIA36790 Yes, why ?

ramsay

i have a project in chemistry and i have two options either i have to do the project given to me by my professor or i should do on my own(*i don't have any idea*)
..the project given to me by my sir is to estimate calcium in toothpastes(*this also i don't know how to do*)
can anyone give me an idea for high school project? Please

user116211

@Ramanewbie I'm not @hippa :P

Ramanewbie

@MAFIA36790 Alright, I'll ask him.

user116211

@Ramanewbie But wasn't he supposed to remain 24x7?

Brian

13:02

@Mart I can point you to some of the original Angeli papers on NEVPT2 if you're interested.

It may be formally intruder-state-free, but I've observed root energies jumping around wildly in some of the systems I've poked at with it.

Martin - マーチン

@Brian How did you know that we were just talking about that...

Brian

<chuckle>

By reading the transcript

user116211

@Martin-マーチン Telepathy.

Brian

("Reading")

( equals "Google Translating")

user116211

damn....

Martin - マーチン

13:04

haha...

Loong

:-D

user116211

@Brian is not mutant

Martin - マーチン

Well then, point away :D

Brian

rummages around in lit folder

!!doi/10.1063/1.1361246

UNHAPPINESS! @CHEMOBOT IS NOT HERE!

dx.doi.org/10.1063/1.1361246

dx.doi.org/10.1007/s00214-006-0207-0

dx.doi.org/10.1063/1.1778711

Angeli was also very friendly and responsive when I emailed him about a question I had about the method.

user116211

@Brian hmm... @mart do you know why he is not here?

Brian

13:09

@Hipp's the guy to ask.

Martin - マーチン

I'll have a look. Currently I have no ORCA installed and I am working on CASPT2 with GAMESS right now, but I'll look into it asap

No idea why chemobot is on break.

Brian

I'm really excited for the new ORCA release -- I'm pretty sure it'll roll out their DLPNO method to a whole bunch of stuff: -CCSD(T), -MP2, -MP2-F12, and I think -NEVPT2, too.

(based on their recent publications, anyways)

'course, my available hardware isn't big enough to really take advantage, but... {sigh}

Martin - マーチン

I feel just comfy enough to work with gamess now...

@Brian i know the feeling...

i need ram...

pH13 - Yet another Philipp

@Brian plz always feel free to join the g-block if you are interested in our conversation :)

Brian

@Martin-マーチン I would have replied right into g-block, but despite a bit of Duolingoing, noch nicht spreche Deutsch gut auf allen, at least not without a lot of Google Translate.

skill patrol

13:18

exactly :-D

user116211

One day I would learn German....

user116211

::sighs::

Martin - マーチン

we occasionally switch to english, too ;)

it's not forbidden there

Brian

@MAFIA36790 Highly recommend duolingo.com

It's working really well for me in Spanish

I've got German started, but my wife had some Spanish and that's what our boys are learning, so Spanish is the main focus.

@Martin-マーチン Didn't know that -- didn't want to de-Deutsch the mojo.

Will feel free in future. :-)

user116211

::gives !!ice-cream to @Brian::

Martin - マーチン

13:21

Haha... de-mojo...

well... the mojo is the chemistry ;)

skill patrol

the mistry is the chemistry without the mojo

Brian

@Martin-マーチン What happens in CASPT2 when intruder states get in the way? Does it foul the results? Or, does it foul convergence of the method itself?

I've not used the method, and I've never been able to clearly find out.

Martin - マーチン

I honestly have no idea...

I'm still fighting with my active space

before having even the prerequisites to perform mp2 on it

Brian

pictures @Mart grappling with an orbital isosurface

pH13 - Yet another Philipp

@Brian I should have paid more attention during the resp. lectures :D

Martin - マーチン

13:25

we never had lectures about it...

Brian

Can you run MP2 on a MC wavefunction?

Martin - マーチン

and when I "grew up" caspt2 was still new and only for the advanced

@Brian well... it's not moller-plesset anymore ;)

Brian

{smacks forehead}

Indeed, nevermind.

Martin - マーチン

but pt2 is essentially the same formalism just with a different basis

basis meaning reference wf

Brian

{nodnod}, I got hung up on the "MP2" term itself.

pentavalentcarbon

13:28

@Brian Both.

Brian

@pentavalentcarbon Gotcha, thanks.

@pent @mart I think there may still be intruder state-style behavior in NEVPT2, it just doesn't affect convergence.

In the handful of NEVPT2 calculations I've run, I've never had convergence problems.

pentavalentcarbon

I've had convergence problems doing MR-MP2.

Brian

My systems may not have been challenging from a convergence perspective... no idea.

pentavalentcarbon

Metals. Always metals.

Brian

<nod> Only thing I've done is copper, which seems to be nicely single-configurational in ground state.

Martin - マーチン

13:32

Jep... exactly

Cr/Mo....

Brian

(most of the time)

Martin - マーチン

CAS(19,19)

Brian

>.<

Huge.

Martin - マーチン

needs about 20GB per core...

and weeks

and that's just CAS!!!

pH13 - Yet another Philipp

:D

pentavalentcarbon

13:34

@Brian I'm finding this might not be true.

Brian

ORCA has Chan's DMRG implemented, but I don't think you can do the NEVPT2 atop the result.

pH13 - Yet another Philipp

@Brian implemented or an interface? I read about that a little

Brian

@pentavalentcarbon I accidentally wandered into looking at the gas-phase Coulomb explosion of monoaquocopper(II), and I'm pretty sure the dissociation into Cu(I) and H2O(+) is appreciably multiconfigurational.

I just assumed it was mainly due to the water cation radial, though.

@pH13-YetanotherPhilipp Depends what you mean. BLOCK is integrated into the binaries.

pH13 - Yet another Philipp

fancy

Brian

But the connections were kind of flaky last time I tried to use it.

But, that was at v3.0.1 or v3.0.2. It might be more robust in v3.0.3.

pentavalentcarbon

13:39

@Brian I wouldn't be surprised if the dominant CSF was doubly-occupied Cu(I), but as soon as you start making assumptions, you know what happens...

Brian

! RHF def2-QZVP def2-QZVP/J RIJCOSX
! PAL8
%casscf
nel 13
norb 10
mult 4,2
nroots 20,30
MaxIter 450
NEVPT2 True
nev_qdtype 1
end

3.7 Angstroem distance between Cu and O

pH13 - Yet another Philipp

nroots 20,30

:D

Brian

<shrug> Lot of five-fold degenerate (or nearly) roots

LOWEST ROOT (ROOT 0 ,MULT 4) = -1714.196696638 Eh -46645.664 eV

STATE ROOT MULT DE/a.u. DE/eV DE/cm**-1
1: 1 4 0.000021 0.001 4.7
2: 0 2 0.000135 0.004 29.7
3: 1 2 0.000157 0.004 34.4
4: 2 4 0.000375 0.010 82.2
5: 3 4 0.000403 0.011 88.4
6: 2 2 0.000513 0.014 112.5
7: 4 4 0.000521 0.014 114.3
8: 3 2 0.000541 0.015 118.6
9: 4 2 0.000662 0.018 145.3

---------------------------------------------
CAS-SCF STATES FOR BLOCK 2 MULT= 2 NROOTS=30
---------------------------------------------

ROOT 0: E= -1714.1965613886 Eh
0.54202 [ 4612]: 2221221100 <---
0.23604 [ 2434]: 1222221100 <---
0.09243 [ 4345]: 2212221100
0.08213 [ 4708]: 2222121100
0.03746 [ 3722]: 2122221100
ROOT 1: E= -1714.1965399780 Eh 0.001 eV 4.7 cm**-1
0.83016 [ 3722]: 2122221100
0.08422 [ 4708]: 2222121100
0.04954 [ 4612]: 2221221100
0.02562 [ 2434]: 1222221100

---------------------------------------------
CAS-SCF STATES FOR BLOCK 1 MULT= 4 NROOTS=20
---------------------------------------------
ROOT 0: E= -1714.1966966383 Eh
0.54202 [ 3833]: 2221221100 <--
0.23645 [ 2099]: 1222221100 <--
0.09237 [ 3646]: 2212221100
0.08206 [ 3889]: 2222121100
0.03741 [ 3163]: 2122221100
ROOT 1: E= -1714.1966752320 Eh 0.001 eV 4.7 cm**-1
0.83030 [ 3163]: 2122221100
0.08403 [ 3889]: 2222121100
0.04970 [ 3833]: 2221221100
0.02573 [ 2099]: 1222221100

See CSFs marked with "<--"

They put unpaired electrons in Cu 3d, Cu 4s, and O 2p.

pentavalentcarbon

So this is very dangerous.

You're averaging 20 and 30 states together for each spin multiplicity, as you know.

Brian

Mmph. Point.

I think I might have one ground-state only

(What I get for searching in a hurry)

pentavalentcarbon

13:51

My guess is that there is decent enough separation between the single determinantal GS and the 1st excited states, but the states in the ES manifold are close together.

And are even closer together once you move to many-particle CSF space.

I bet there's quite a difference in energy between your 20-,30- state-averaged GS and a SS-CASSCF.

ROOT 1: E= -1714.1965399780 Eh 0.001 eV 4.7 cm**-1

5 wavenumbers??!!?!?

Brian

Searching for GS-only run

pentavalentcarbon

That gap can't be real.

Brian

This is why I only ever say I play with CAS. :-)

I dunno if I have a run at a dissociated spacing, all weight on 2'[0] root, with large enough active space.

pentavalentcarbon

@Brian Yes, but that's how you have to start, is to play with it. Running MCSCF calculations is an art.

Brian

Oh, I know!

That's why I'm doing i t

pentavalentcarbon

13:57

I can run it.

Brian

It's also why I won't be publishing any MC work for a looooooong (not Loong, though) time.

pentavalentcarbon

I have a gray hair for every friggin orbital I've rotated for MRCI calculations.

Brian

:-)

This is the geometry I used for that 20,30 run:
4
Coordinates from ORCA-job CuII_1aq_157
Cu -2.59939669435660 -0.84509895826102 -0.84501841743074
O 1.50134480833304 0.46527768809885 0.46526086140404
H 2.04901383579977 1.20803422473467 0.07174471940928
H 2.04903705022377 0.07178704542750 1.20801383661741

pentavalentcarbon

Damn it, I just prepared my geometry run...

Brian

Oh, wait, I might've found an okay one/

Averaged only of first five doublet states

CAS(13,10)

This one's at 4.5 Angstron Cu-O distance

Loong

14:01

@Jan hmm

Brian

Hm. This one seems to be converged to the Cu(II) + H2O(0) solution.

~Degenerate single occupancy in the Cu 3d orbitals

I don't know if I have a worthwhile one on a Cu(I) + H2O(+) wavefunction.

Ah well. Thanks, @pent, for diving down the rabbit hole with me.

pentavalentcarbon

This is gross. Aaaaand the CI stalled.

As usual.

Brian

ORCA CI stall?

Loong

@Jan nice problem. ChemDraw and ACD/ChemSketch both create a wrong name.

pentavalentcarbon

Yes, in the unfolded two-step procedure, first you take the "macro" step, where the MO coefficients are updated, then the "micro" steps, where the CI coefficients are updated.

Sometimes the default CI converger has trouble.

Brian

14:10

Oh, I know the CI stall.

The sweet, sweet inchoate fury of the CI stall.

pentavalentcarbon

It usually isn't a problem, but it's a good indicator that your active space choice is bad.

Brian

Really? It's always seemed random to me.

I thought that a bad AS lent more to an excessive number of macro-iterations.

pentavalentcarbon

Yes, this is true, but there can be multiple indicators.

I do think in this case it's more likely that ORCA's CI algorithm is...weak.

Another indicator is "Rot=207,0"...

And "N(occ)= 1.99991 1.99971 1.99972 1.99970 1.99971 1.99971 1.00000 0.00141 0.00003 0.00010"

Brian

Heh. Deep inactive/virtual swaps, yeah

I take it you find the system to be strongly single-configurational?

:-)

pentavalentcarbon

Yes, when you do a single-state calculation.

And you don't know what MOs those are.

pH13 - Yet another Philipp

14:14

also the many nearly 2/nearly 0 orbitals are not good for convergence, afaik

pentavalentcarbon

@pH13-YetanotherPhilipp That's my point.

Brian

@pH Yep

@pent Are you running single-root GS CAS?

pentavalentcarbon

@Brian Yeah, that won't converge without some previous MO coefficient inspection or DFT/natural orbitals.

pH13 - Yet another Philipp

may I compare a cas(4,5) energy with a cas(6,7) energy for two isomers?

one isomer 4,5, the other 6,7

pentavalentcarbon

No.

(I don't think so. Comparing energies is dangerous.)

pH13 - Yet another Philipp

14:18

I think the cas needs the same orbitals

Brian

<nod>, at the very least I'd keep the same active space dimensions

Ideally the active space orbital composition would be substantially the same for each, too.

Even then... probably qualitative at best?

CAS itself is not that great for energies, I don't think.

pH13 - Yet another Philipp

nevpt2 corrected

Brian

Critical to keep the same active space with NEVPT2

Take a look at these: chat.stackexchange.com/transcript/message/29163579#29163579

"It should be remarked
that in the strongly contracted approach the zero order
Hamiltonian is not invariant under rotation of the core
or virtual orbitals; in our codes we make use of the
canonical orbitals for the SC-NEVPT variant. Furthermore,
it isworth remembering thatNEVPTis not invariant
under rotations between active and inactive orbitals
and care must be devoted to a proper choice of the
active space so as to avoid possible exchanges of the
identities of the orbitals whichmight occur, for instance,

@pH From p747 of dx.doi.org/10.1007/s00214-006-0207-0

Brian

14:32

@pentavalentcarbon Sure, I figure it's just a degenerate root, except for some numerical noise?

pentavalentcarbon

@Brian You mean the singlet-state GS calculation?

Brian

No, see the specific line I'm replying to, it's back at the silly 20,30 output.

pentavalentcarbon

@Brian Is there a good way of doing that? I can only hover and check for the highlighted message, and it doesn't work when it's off-screen.

Brian

Click the arrow to the left of the message

(the upside-down 'return' symbol)

pentavalentcarbon

@Brian But which one is it? i.imgur.com/9oqUhSO.png

Wait, I see it now.

Not the down arrow, the return symbol, like you said.

Maybe I should learn to read.

Brian

14:40

Nah, there was a bit of a lag before I typed the 'return symbol' thing. Probably you'd already started hunting elsewhere before I sent it.

pentavalentcarbon

@Brian As much as I don't trust this program, I would expect a degenerate root to show up with the exact same energy.

So I don't think it's degenerate, just artificially close due to the state averaging procedure.

The true test that I know of to see if SA-MCSCF is appropriate would be to increase/decrease the number of states you average over and see how the energy spectrum changes.

Brian

Huh. You don't think the presence, even distant, of the water molecule might slightly perturb the energies of the various Cu 3d orbitals?

pentavalentcarbon

Yes, they're quite sensitive due to the core contraction, and water would love to donate those lone pairs.

Brian

Ok, yeah, so that's where I figure they're near-degenerate, but not exactly

An isolated Cu(II) gives very tightly degenerate roots, as expected.

---------------------------------------------
CAS-SCF STATES FOR BLOCK 2 MULT= 2 NROOTS= 7
---------------------------------------------

ROOT 0: E= -1638.1914715666 Eh
0.89301 [ 4740]: 12222000000
0.08413 [ 882]: 21222000000
ROOT 1: E= -1638.1914715666 Eh 0.000 eV 0.0 cm**-1
0.88578 [ 882]: 21222000000
0.08298 [ 4740]: 12222000000
0.00875 [ 112]: 22122000000
ROOT 2: E= -1638.1914715665 Eh 0.000 eV 0.0 cm**-1
0.96003 [ 112]: 22122000000
0.00888 [ 7]: 22212000000

That was with def2-TZVP, SA-CAS(9,11) on those five lowest doublet roots.

pentavalentcarbon

@Brian This is probably correct, based on our chemical intuition about free atoms (no octahedral field), but HF/DFT will show a sizable first GS->ES gap.

Brian

14:51

Oh, sure.

pentavalentcarbon

And SS-CASSCF might too, for individually-optimized states.

might

Brian

<grin>

user116211

Wikipedia to moon!

user116211

meta.wikimedia.org/wiki/Wikipedia_to_the_Moon

pentavalentcarbon

@Brian (In case you couldn't tell, I study the spectroscopy of copper(II) compounds.)

Brian

15:00

Hehe, I was starting to suspect something like that.

<--- will not be disputing @pent's expertise on copper any time soon

Jan

@Loong Which one?

Loong

@Jan the last one; the chloride fluoride

Jan

Oh that one.

Brian

@pent The SUPER-CI is struggling to converge on the single-state CAS(9,11)

Oscillations upon nearing convergence

pentavalentcarbon

I forget what the other CI convergers are.

Brian

15:12

i know there's a POPLE

Those might be the only two?

Jan

I'm voting to close this question as off-topic because this is not a question. — Jan 53 secs ago

*feels strong* x3

pentavalentcarbon

checking the manual now

DIIS, KDIIS, SuperCI, perturbative SuperCI, SOSCF, Newton-Raphson

Actually, I usually use the NR solver.

Brian

15:30

Feh, not POPLE. That must be for the CP-SCF

@pent I think this might have been why I ended up doing the SA on the first five roots -- convergence was terrible otherwise.

pentavalentcarbon

@Brian Yes, CPSCF is either Pople's algorithm, DIIS, or conjugate gradient.

Brian

Would HF/CIS be a faster & more effective approach here, with the CAS convergence problems?

« first day (1456 days earlier) ← previous day next day → last day (3230 days later) »

Transcript for

Apr '1621

The Periodic Table

Haikus are awesome / Chemistry's even better / So pull up a chair

blowing-the-lab-up german-stuff hullo love-you-all martin table-flipping tre welcome

join this room
about this room

00:00

06:00

12:00

18:00

all times are UTC

site design / logo © 2025 Stack Exchange Inc; legal

mobile