/sci/ - Science & Math » Thread #13817119

305KiB, 500x463, dft.png

View Same Google iqdb SauceNAO

ab initio chemistry

Anonymous Thu 04 Nov 02:13:51 2021 No.13817119 View Reply Original Report

Quoted By: >>13817215 >>13818812 >>13819010 >>13822277

CC vs DFT vs CI vs QMC?

Anonymous

Anonymous Thu 04 Nov 2021 02:59:52 No.13817215 Report

Quoted By: >>13817492 >>13818063

>>13817119
what the fuck is pic-related

Anonymous

Anonymous Thu 04 Nov 2021 05:27:36 No.13817492 Report

Quoted By:

>>13817215
bump

Anonymous

Anonymous Thu 04 Nov 2021 11:28:01 No.13818063 Report

Quoted By:

>>13817215
it's electron density with some threshold for color

Anonymous

Anonymous Thu 04 Nov 2021 16:19:56 No.13818812 Report

Quoted By: >>13819629

>>13817119
just checking in to say CC superiority. CI is literally just a shitty CC that's worse in every way, any DFT that has a decent level of accuracy relies on some empirical parameter fitting, and QMC is meme tier made to push out papers quickly

Anonymous

Anonymous Thu 04 Nov 2021 17:01:38 No.13818929 Report

Quoted By: >>13818940

CCSD(T) is rightfully the gold standard. DFT a shit, choose your own functional.

More seriously, depends on the system and its features that are of interest to you.

Anonymous

Anonymous Thu 04 Nov 2021 17:06:18 No.13818940 Report

Quoted By: >>13818946

>>13818929
i want the thing that's most generally computable, meaning i don't have to tweak a billion things to get reasonable results on an arbitrary but small molecule. QMC seems like it's what i want

Anonymous

Anonymous Thu 04 Nov 2021 17:08:26 No.13818946 Report

Quoted By: >>13818980

>>13818940
and the features of interest to me is electronic structure and its dynamics of course.

Anonymous

Anonymous Thu 04 Nov 2021 17:22:12 No.13818980 Report

Quoted By: >>13819515

>>13818946
weak and strong correlation have to both work without changing anything manually. excited states have to work as well as ground states.
for those things i'm happy to compromise on doing tens of atoms instead of hundreds or thousands.

Anonymous

View Same Google iqdb SauceNAO 1608823331041.jpg, 49KiB, 850x400

Anonymous Thu 04 Nov 2021 17:32:30 No.13819010 Report

Quoted By: >>13819020 >>13819601

>>13817119
Daily reminder that Physishits lack any sort of rigor and do not understand the math they are using. This of course holds also true for physical chemistry.

Anonymous

Anonymous Thu 04 Nov 2021 17:35:59 No.13819020 Report

Quoted By: >>13819039

>>13819010
Show us your electronic structure algorithm Anon! :3

Anonymous

View Same Google iqdb SauceNAO 1635589201047.png, 32KiB, 740x308

Anonymous Thu 04 Nov 2021 17:41:36 No.13819039 Report

Quoted By: >>13819116 >>13819515

>>13819020
Quantum mechanics can be rigorously defined using operator algebras ( $C^*$ -algebras). Any application of it is trivial (any second year math student could do it) and lacks the elegance and beauty of pure mathematics.

Anonymous

Anonymous Thu 04 Nov 2021 18:06:23 No.13819116 Report

Quoted By:

>>13819039
>I don't really care about the real world
Well fine!

Anonymous

Anonymous Thu 04 Nov 2021 20:05:10 No.13819515 Report

Quoted By:

>>13818980
Maybe MRCC with F12 basisfunctions could be something for you.

>>13819039
Being a mathematican can be quite the advantage I guess, but one also needs a understanding of the applications. If you don't know the chemistry or physics involved, you won't get far with the math alone.
In that way, it could be more advantageous to study physics first and do a bit of math (as you said, second year) in addition.
Also, mathematical rigor can be a disadvantgae. Sometimes shortcuts are fine.

Anonymous

Anonymous Thu 04 Nov 2021 20:33:12 No.13819601 Report

Quoted By: >>13822267

>>13819010
God damn it.

WE HAVE ALREADY SOLVED COMPUTATIONAL CHEMISTRY ANALYTICALLY. Guess what? It isn't remotely time efficient, essentially exponential, and in many cases factorial. So unless mathematicians have a god send algebraic simplification, we're stuck optimizing CC or waiting till fucking quantum algorithms turn out to be useful for more than lining google's journalism fund.

Anonymous

Anonymous Thu 04 Nov 2021 20:39:54 No.13819629 Report

Quoted By: >>13824122

>>13818812
>QMC is a meme
It shits all over your CC shit when the electron correlation really kicks in to high gear. But yes it is a meme unless you have a cluster of > 1024 nodes

Anonymous

View Same Google iqdb SauceNAO ct-2017-00479z_0009.jpg, 114KiB, 800x404

Anonymous Thu 04 Nov 2021 20:44:18 No.13819642 Report

Quoted By: >>13819659 >>13822150

>Gives you DFT tier results at the fraction of walltime
>Nothing personal kiddos, just solving problems that matter

Anonymous

Anonymous Thu 04 Nov 2021 20:50:00 No.13819659 Report

Quoted By:

>>13819642
>Implying people want DFT tier results for anything other than time dependent frequency calculations for obscure crystal strucures

Anonymous

Anonymous Fri 05 Nov 2021 14:19:02 No.13822150 Report

Quoted By:

>>13819642
thanks a lot anon, I just lost no fap november because of you.

Anonymous

Anonymous Fri 05 Nov 2021 15:08:15 No.13822267 Report

Quoted By: >>13822519

>>13819601
>WE HAVE ALREADY SOLVED COMPUTATIONAL CHEMISTRY ANALYTICALLY
How can you say you solved it when the computer shit itselfs over few atoms, how unoptimized and stupid the algorithm has to be?
Talk to me when you actually use unique electron forces to compute chemistry.

Anonymous

Anonymous Fri 05 Nov 2021 15:10:25 No.13822277 Report

Quoted By: >>13823600 >>13824122 >>13824880

>>13817119
Molecular dynamics is literally trash-tier, i don't care to see pointless molecules vibrating forever. You can't even simulate molecule reactions or electricity, useless.

Anonymous

Anonymous Fri 05 Nov 2021 16:48:59 No.13822519 Report

Quoted By:

>>13822267
well hypothetically with an infinitely large array of computers and an infinite amount of time, any chemistry problem can be proven to be solvable :^)

Anonymous

Anonymous Fri 05 Nov 2021 19:06:56 No.13822969 Report

Quoted By: >>13823095 >>13824122

Bump

Will someone come up with a kinetic energy density functional ?

Will orbital free DFT ever free it self from meme status ?

Anonymous

Anonymous Fri 05 Nov 2021 19:52:04 No.13823095 Report

Quoted By:

>>13822969
With the current AI and computational progress and the amount scientists working on this, it's unbelievable that our solutions are still baby-level. Some of our fundamental understandings or assumptions must be totally wrong, computers can't solve this for us if we present it with wrong problem, AI can't do experiments itself.

Anonymous

Anonymous Fri 05 Nov 2021 22:47:44 No.13823600 Report

Quoted By: >>13825291

>>13822277
Yeah go ahead and model protein-protein interactions with QM

Anonymous

Anonymous Sat 06 Nov 2021 02:29:13 No.13824122 Report

Quoted By: >>13824820

>>13819629
show me all your efficient derivatives for arbitrary perturbations, anon

>>13822277
dubs are wasted on you, there are dozens of ab initio md methods

>>13822969
nope, both kinds of dft are fucked, but at least with kohn-sham these stupid machine learning simps will get us better answers at the expense of increasingly stupid ccsd(t)-based training sets

Anonymous

Anonymous Sat 06 Nov 2021 10:05:24 No.13824820 Report

Quoted By:

>>13824122
now we just need a training asic you can buy that's not a gpu and not in the jewgle cloud

Anonymous

Anonymous Sat 06 Nov 2021 10:46:48 No.13824880 Report

Quoted By:

>>13822277
how do we simulate electricity quantum chemistrybros?

Anonymous

Anonymous Sat 06 Nov 2021 14:00:55 No.13825291 Report

Quoted By:

>>13823600
that's why MD monkeys are for anon ;)

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches