8
$\begingroup$

I am trying to optimize the geometry of a Cs4CuSb2Cl12 Stoichiometric Quantum dot and I have checked the input multiple times, but the opt+freq calculation is not converging. The quantum dot has about 80 atoms and is charge neutral furthermore I don't see any dangling bonds as such.

Here is my input file -

%NProcShared=16
%mem=24GB
%chk=Checkpoint_dbperovskite_geo_2.chk
#p opt freq cam-b3lyp/gen nosymm int=(grid=ultrafine,acc2e=12) pseudo=read

 geometry opetimization_cs4cusb2cl12_stochiometric

0 1
Cs         3.89312        0.00000       10.49322
Cs         1.85782        3.66335        7.48003
Sb        -0.41300        0.00000        9.15427
Cu         4.07868        0.00000        6.03024
Cl         5.00209        1.62404        7.36968
Cl         1.78933        0.00000        7.65648
Cl         0.56747        1.74654       10.53435
Cl        -2.53224        0.00000       10.58850
Cs         4.26424        0.00000        1.56726
Cs        -2.61973        3.66335       10.49322
Cs        10.77709        3.66335        1.56726
Cs         6.29954        3.66335        4.58045
Cs         8.37067        0.00000        7.48003
Cs        -0.21331        0.00000        4.58045
Sb         8.57036        0.00000        2.90621
Sb         6.09985        3.66335        9.15427
Sb         2.05751        3.66335        2.90621
Cu        -2.43417        3.66335        6.03024
Cl         3.15527        1.62404        4.69080
Cl         3.15527        5.70266        4.69080
Cl         5.00209        5.70266        7.36968
Cl        -1.51076        5.28739        7.36968
Cl         9.66812        5.28739        4.69080
Cl         9.66812        2.03931        4.69080
Cl        -1.51076        2.03931        7.36968
Cl         6.36803        0.00000        4.40401
Cl         8.30218        3.66335        7.65648
Cl        -0.14482        3.66335        4.40401
Cl         7.58989        1.74654        1.52613
Cl         7.58989        5.58016        1.52613
Cl         0.56747        5.58016       10.53435
Cl         7.08032        5.40989       10.53435
Cl         1.07704        5.40989        1.52613
Cl         1.07704        1.91681        1.52613
Cl         7.08032        1.91681       10.53435
Cl        10.68959        0.00000        1.47198
Cl         3.98061        3.66335       10.58850
Cl         4.17674        3.66335        1.47198
Cs        -0.97523        0.00000       22.55371
Cs        -3.01053        3.66335       19.54052
Sb        -5.28135        0.00000       21.21475
Cu        -0.78967        0.00000       18.09072
Cl         0.13374        1.62404       19.43016
Cl        -3.07902        0.00000       19.71696
Cl        -4.30088        1.74654       22.59483
Cl        -7.40058        0.00000       22.64898
Cs        -0.60410        0.00000       13.62774
Cs        -7.48808        3.66335       22.55371
Cs         5.90875        3.66335       13.62774
Cs         1.43119        3.66335       16.64093
Cs         3.50232        0.00000       19.54052
Cs        -5.08166        0.00000       16.64093
Sb         3.70201        0.00000       14.96670
Sb         1.23150        3.66335       21.21475
Sb        -2.81084        3.66335       14.96670
Cu        -7.30252        3.66335       18.09072
Cl        -1.71308        1.62404       16.75129
Cl        -1.71308        5.70266       16.75129
Cl         0.13374        5.70266       19.43016
Cl        -6.37911        5.28739       19.43016
Cl         4.79977        5.28739       16.75129
Cl         4.79977        2.03931       16.75129
Cl        -6.37911        2.03931       19.43016
Cl         1.49968        0.00000       16.46449
Cl         3.43383        3.66335       19.71696
Cl        -5.01317        3.66335       16.46449
Cl         2.72155        1.74654       13.58662
Cl         2.72155        5.58016       13.58662
Cl        -4.30088        5.58016       22.59483
Cl         2.21197        5.40989       22.59483
Cl        -3.79130        5.40989       13.58662
Cl        -3.79130        1.91681       13.58662
Cl         2.21197        1.91681       22.59483
Cl         5.82125        0.00000       13.53247
Cl        -0.88773        3.66335       22.64898
Cl        -0.69160        3.66335       13.53247

Cs Sb Cu 0
LANL2DZ
****
Cl 0
6-31+g(d)
****

Cs Sb Cu 0
LANL2DZ

If there is anything that could be causing the convergence issues please let me know. And if the input is correct does that mean our quantum dot system is not stable. Because the PhD student who I am working under thinks that our input as well as system is correct but Gaussian's basis set are not compatible with this and all the literature regarding the substance actually uses VASP or some other plane wave based dft code.

I think it is possible that it is possible that there is something wrong with the structure too, so here is a snipit.

enter image description here

enter image description here

The log file gives the conversion error at the end

>>>>>>>>>> Convergence criterion not met.
 SCF Done:  E(RCAM-B3LYP) =  -23237.5859709     A.U. after  129 cycles
             Convg  =    0.1362D-02             -V/T =  2.0258
 KE= 2.265279142009D+04 PE=-1.242564502738D+05 EE= 4.383249239695D+04
 Convergence failure -- run terminated.
 Error termination via Lnk1e in /scf-data/apps/gaussian/g09/l502.exe at Sun Dec 26 05:00:55 2021.
 Job cpu time: 39 days  3 hours 26 minutes 55.5 seconds.
 File lengths (MBytes):  RWF=   1493 Int=      0 D2E=      0 Chk=    442 Scr=      2

Although the same convergence error is not seen in the out.txt file but rather some weird error shows up,

Error: segmentation violation
   rax 0000000000000000, rbx ffffffffffffffff, rcx ffffffffffffffff
   rdx 0000000000016888, rsp 00007ffc6e52a518, rbp 00007ffc6e52ab00
   rsi 000000000000000b, rdi 0000000000016888, r8  00007f90d2e81740
   r9  0000000000000000, r10 00007ffc6e529920, r11 0000000000000206
   r12 0000000000000000, r13 0000000000000000, r14 00007ffc6e52ab48
   r15 00000000000003e6
  --- traceback not available

Here is the link to the log file

Improve this question
$\endgroup$
5
  • $\begingroup$ To keep things from getting too long here, I moved comments to chat $\endgroup$
    – Tyberius ♦
    Commented Jan 21, 2022 at 21:45
  • $\begingroup$ Perhaps you could try first doing a calculation with PBEPBE and then, if that works, read in the wavefunction as a guess for CAM-B3LYP. $\endgroup$
    – Andrew Rosen
    Commented Jan 22, 2022 at 5:02
  • 1
    $\begingroup$ From where did you got the actual geometry? Did you just cut out of something from a unit cell or know the stable surfaces, morphologies? Your choice of tool is also way too slow for these tasks - you couldn’t have a single opt step in one months. $\endgroup$
    – Greg
    Commented Jan 22, 2022 at 12:42
  • $\begingroup$ @AndrewRosen I will try that but I think it is not that the structure is too far away from the stable geometry but something is actually wrong somewhere so will try your suggestion but once I am sure about this $\endgroup$
    – Parmeet Singh EP 066
    Commented Jan 22, 2022 at 15:11
  • $\begingroup$ @Greg I am not sure about the actual geometry as such. Like I got the cif file for the unit cell from a paper and then simply made a small structure using that. Now this is what I have been thinking about and there is a paper which talks about spherical nano particles of the same structure so maybe I should try that instead of rectangular ones. Furthermore I am not sure what do you mean exactly when you say my choice of tools. If you are talking about computational resources then these are the best I can get sadly $\endgroup$
    – Parmeet Singh EP 066
    Commented Jan 22, 2022 at 15:15

2 Answers 2

Reset to default
6
$\begingroup$

OK, I do not want to be that guy, but I will be that guy.

Some general remarks

  • You should plan better your calculation projects. You got 129 SCF cycle calculated in about 3 days. Even if this converges and everything goes right, it can easily take a month to complete a single “opt freq” calculation. That is way too much time to “just trying out something” or fiddling with parameters.
  • Especially if you have a slower calculation like this, it worth to pre-optimize your structure with something similar but faster: eg with a pure functional.
  • Whenever you have an SCF convergence problem, check: geometry, if you have a small HOMO-LUMO gap, how does the SCF energies look like. Sometimes you just need a few steps more, sometimes those are wildly oscillating. It is always a good idea to play a little bit with DIIS and dumping, it often cures smaller issues.

Remarks specific to nanoparticles

  • Nanoparticles, oxides, halogenides etc are notoriously hard to calculate with local basis sets.
  • Geometry of the nanoparticle does matter. Try to find the dominant faces and morphologies. Try to find out what are the typical coordinations on the surface (from experimental data). When you create models, do not just randomly cut out something from the bulk. Generally, models that has no huge dipole moment, somewhat symmetric, all metal ions are mostly coordinated and shapes not too elongated are the easiest to converge. Compared to that you have many Sb and Cu with low coordination numbers. I would also expect dipole moment, too, just from how asymmetric arrangement of positive and negative ions.
Improve this answer
$\endgroup$
4
  • $\begingroup$ Thanks for being that guy!! I have changed the structure of the Quantum Dot from that slab looking thing, to a more symmetric spherical structure, there is literature which says they were able to make spherical nano particles for the said structure. One thing I wanted to ask was that currently I am using Avogadro for making the structure and avogadro has an option to do energy minimization using UFF force field. Would that be enough as pre-optimization?? $\endgroup$
    – Parmeet Singh EP 066
    Commented Jan 24, 2022 at 7:52
  • 1
    $\begingroup$ @ParmeetSinghEP066 It is hard to tell without trying. If you have time, it worse a shot to optimize the geometry with UFF or some tight-binding method, which may help you to identify problematic atoms. $\endgroup$
    – Greg
    Commented Jan 24, 2022 at 7:54
  • $\begingroup$ Can you please elaborate what do you mean about a problematic atom ?? $\endgroup$
    – Parmeet Singh EP 066
    Commented Jan 24, 2022 at 7:59
  • $\begingroup$ @ParmeetSinghEP066 For example an atom/ion that is very far from the equilibrium position. Of course, UFF can be very misleading, too. $\endgroup$
    – Greg
    Commented Jan 24, 2022 at 8:05
5
$\begingroup$

I didn't deal with perovskite before, but I got this problem when I used metal atom in my complex.

So, from my point of view, I think your calculation needs more cycles to get converged so what about increasing the number of cycles using keyword SCF(MaxCycle=1000) and if it didn't work well, you can use keyword SCF=XQC.

XQC means you will use quadratically convergent SCF procedure in case the first-order SCF has not converged. It will need more time, so my advice, you can try with small basis set to see if it worked well, then you can use your basis set.

Improve this answer
$\endgroup$
4
  • 1
    $\begingroup$ My concern with adding more cycles is that it can paper over more fundamental issues with the input geometry, method, etc. Unless the energy and density are close to convergence (or at least making steady progress), adding hundreds of additional SCF iterations might just be prolonging a broken calculation. XQC or YQC can be helpful with systems containing metals, but I agree with your suggestion to try it on a smaller basis first to see if that actually solves the problem, because for me these issues tend to arise from a faulty input geometry. $\endgroup$
    – Tyberius ♦
    Commented Jan 23, 2022 at 15:46
  • 1
    $\begingroup$ @Tyberius I agree with you, but if you saw the output, the criteria to get converged, all are No. hence, increasing the number of cycles, it will just try to find a good geometry to start again and later maybe the large number of cycles will not be needed, and the program can converge the system in a less than 129 cycles. Hence, increasing the number of cycles is just to overcome this problem. I am not sure $\endgroup$
    – Abd-Elazeem Mohamed
    Commented Jan 23, 2022 at 16:01
  • $\begingroup$ If I use STO-3G directly for making the computation time smaller, I know that there will be deviation from the real values of molecular properties. Wouldn't this lead to problems in convergence?? And according to you would STO-3G do the work ?? And thanks for the other remarks $\endgroup$
    – Parmeet Singh EP 066
    Commented Jan 24, 2022 at 7:45
  • $\begingroup$ @ParmeetSinghEP066 I think, using STO-3G is not good for this system. At least you can use 6-31G with XQC keyword and if you get convergence and the problem has been solved then take the last geometry or read checkpoint to start a new calculation with your final basis set. $\endgroup$
    – Abd-Elazeem Mohamed
    Commented Jan 25, 2022 at 13:12

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .