Novice user, trying to optimise + analyse structure.

Find answers or ask questions regarding Dalton calculations.
Please upload an output file showing the problem, if applicable.
(It is not necessary to upload input files, they can be found in the output file.)

Post Reply
pcxgj
Posts: 2
Joined: 12 Aug 2016, 16:22
First name(s): Joe
Last name(s): Jones
Affiliation: University of Nottingham
Country: United Kingdom

Novice user, trying to optimise + analyse structure.

Post by pcxgj » 12 Aug 2016, 16:45

Hello,

I'm a novice computational chemist looking to support their empirical data with some DFT and I have been pointed in the direction of Dalton as a tool that could help me. Previously I have been using the ADF program running on a single machine but I am looking at Dalton as an option since I have access to my university's HPC and while geometry optimisation of my structures has been reasonably ok on ADF (only a couple of days) trying to do things like calculate EPR values has been very expensive.

With the help of a friend I have built Dalton on the HPC with MPI and passed the tests and have started my first attempts on one of my structures which appears to be running but I would be interested if someone with more knowledge of this sort of thing could suggest any glaring errors I might have made, or have any useful tips for someone who is looking at the wealth of options and not really understanding what will save time vs what will work well but quadruple the computation time etc. I'm looking for a good balance between accuracy and computation time (aren't we all)?

The system is between 90 and 120 atoms, (Ni, Fe, N, C, S, H) and begins from an XRD structure as a 2+ ion (anions removed). My aim with the system is to reduce it in silico to +1 and look at the geometry change/IR/EPR that I have observed empirically when doing the same thing chemically/electrochemically.

Someone suggested to me that 100 atoms might be too big for Dalton. In that mind, I have tried to build LSDalton on the HPC using the intel MPI compilers, but it fails 90+ of the tests after completion, so at the moment I am sticking wth Dalton but I can try to get that sorted if it would be much more suited to that program.

Edit: I should add, if it helps, that the empirical data supports a Ni (I/II) couple with the frozen EPR data exhibiting a mostly axial system with a small amount of rhombic character (no hyperfine, since the system is not 61Ni enriched, although the plan is to synthesise the enriched version to look at this data).

My input file is attached if anyone has any pointers, modified from Avogadro's output (mainly used to generate the coordinates list from my xyz file).
Attachments
NiFetBuRed.dal
(6.01 KiB) Downloaded 243 times

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: Novice user, trying to optimise + analyse structure.

Post by taylor » 12 Aug 2016, 17:50

I am a very old user: so old that my telepathic skills have declined with the years and I cannot really understand what it is you are posting about. But a molecule with 100 atoms and any sort of vaguely sensible/useful basis set is likely to be well out of the reach of Dalton, if that is your question. Even on a parallel platform (and not all of Dalton is parallelized!).

I am very surprised that you have major problems with your LSDalton build. It would be helpful to you to try to rectify this (post details of your build (as attachments, not inline!)) and maybe one of the LSDalton mob (of whom I am not one, by the way) will be able to help. Usually, because of its cleaner, more modern code, LSDalton is easier to build. At least in my experience.

Best regaards
Pete

arnfinn
Posts: 231
Joined: 28 Aug 2013, 08:02
First name(s): Arnfinn
Middle name(s): Hykkerud
Last name(s): Steindal
Affiliation: UiT
Country: Norway
Location: UiT The Arctic University of Norway

Re: Novice user, trying to optimise + analyse structure.

Post by arnfinn » 12 Aug 2016, 20:48

If you compile lsdalton with MPI and then run each test on one core, then several test cases will fail because they have to be run with multiple cores. That might be the reason for the failing lsdalton test cases. I think you can

Code: Select all

export DALTON_NUM_MPI_PROCS=4
before you run the test cases to run each test in parallel on 4 cores.

pcxgj
Posts: 2
Joined: 12 Aug 2016, 16:22
First name(s): Joe
Last name(s): Jones
Affiliation: University of Nottingham
Country: United Kingdom

Re: Novice user, trying to optimise + analyse structure.

Post by pcxgj » 12 Aug 2016, 20:56

Thanks for the reply Pete,

Sorry if I didn't explain my aims very well - I have a 100-atom cation that is +2 charge that adopts a particular geometry, but when it is reduced the geometry changes with the formation of a bridging isocyanide between the nickel and iron centres in the system. My starting geometry is that +2 state from the crystal structure. What I have fed into Dalton is that set of coordinates but with a +1 charge to simulate the reduction and I want to see if it will optimise into the bridge structure and to see if the simulated IR and EPR properties match up.

I have seen this effect experimentally and am looking to see if Dalton will also do it. I have had some success with the ADF program in doing that optimisation - it takes a couple of days on a 4 core machine, but the requirements for calculating the ESR properties go through the roof so I was looking at alternatives.

My problems with LSDalton appear to be related to the compiler I am using - I am using a 2015 version of the Intel MPI ones and some preliminary googling on this forum suggested that some other people have also had problems with it. It builds fine but fails about 90 tests. I will export that variable as arnfinn suggests and try the tests again.

Is lsdalton the only real option for a system of the size I am attempting?

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: Novice user, trying to optimise + analyse structure.

Post by taylor » 13 Aug 2016, 00:46

I m afraid, now I understand your problem better, that you will not find LSDalton helpful. You talk of both a +1 and +2 oxidation state: one of these must be open-shell and I am pretty sure LSDalton does not currently handle open shells.

Realistically, I think 100 atoms, with reasonable basis sets (say an average of 10 per atom assuming you have plenty of hydrogens, otherwise it will be more) you end up with 1000 basis functions. Does the system have symmetry? Given that LSDalton does not appear to be an option it is certainly worth a go with Dalton, but you may be disappointed.

I can't really comment on other codes, less because I don't use other codes (I do) but more because I don't do anything this large except with LSDalton. In fact, I hardly ever do anything this large anyway!


Best regards
Pete

Post Reply

Who is online

Users browsing this forum: No registered users and 12 guests