Author Topic: convergence problem  (Read 2289 times)

0 Members and 1 Guest are viewing this topic.

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
convergence problem
« on: June 17, 2022, 13:39 »
After putting gas molecule in device configuration, does simulation converge?
I put multiple gas molecule in device configuration and running in super computer by using 240 core but from yesterday it has not calculated i-v curve for a single point, it seems nothing is happening.
« Last Edit: June 17, 2022, 13:50 by Ambika kumari »

Offline AsifShah

  • QuantumATK Guru
  • ****
  • Posts: 144
  • Country: in
  • Reputation: 2
    • View Profile
Re: convergence problem
« Reply #1 on: June 18, 2022, 09:28 »
Depending on number of k points & size of device, transport calculations usually take long time.

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
Re: convergence problem
« Reply #2 on: June 18, 2022, 11:25 »
total 250 atoms
for test I took k value 1*1*1 still same.

Offline AsifShah

  • QuantumATK Guru
  • ****
  • Posts: 144
  • Country: in
  • Reputation: 2
    • View Profile
Re: convergence problem
« Reply #3 on: June 18, 2022, 12:00 »
Sometimes it doesn't converge because K points are not enough. 1x1x1 could be one problem. Check log file and save it then run with higher k points for same time & compare two log files to see if K point convergence is the issue

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
Re: convergence problem
« Reply #4 on: June 18, 2022, 12:24 »
In log file I am getting nothing, it just mention 16 task will be calculated.
Kindly see the attachment

Offline AsifShah

  • QuantumATK Guru
  • ****
  • Posts: 144
  • Country: in
  • Reputation: 2
    • View Profile
Re: convergence problem
« Reply #5 on: June 20, 2022, 14:39 »
I see you are using Gold (as electrode I guess). From my experience with Gold adsorption calculations they taken long to optimize (converge). I normally run from command line & not from user interface & I see two files apart from .py during run. One is log file & another is name something like optimizationxxxxxx.log . Most infiormation of optimization were found in this optimizationxxx.log file.

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
Re: convergence problem
« Reply #6 on: June 21, 2022, 14:03 »
Optimization converged.
Now I am calculating I-V curve.  K point is 2*5*60. At 1st point calculation converged in 40 steps, at 2nd point it didnot converged. Please see the attachment. When K point was 2*5*50, then at 1st point calculation conversed in 38 stepst and at 2nd point calculation converged in 95 steps and at 3rd point calculation did not converged.  Attachment is for k=2*5*60 How to currect it
.
Thanks

Offline AsifShah

  • QuantumATK Guru
  • ****
  • Posts: 144
  • Country: in
  • Reputation: 2
    • View Profile
Re: convergence problem
« Reply #7 on: June 21, 2022, 14:11 »
If you are using Semiempirical calculator, then in that go to iteration control & if you have unticked no scf iteration then below it there will be maximum steps written as 100, change it to 400 & recompute. I am not sure but there is a way I guess to recompute only not converged points instead of rerunning whole calculations.

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
Re: convergence problem
« Reply #8 on: June 21, 2022, 14:16 »
400 is too high it will take long time.

Offline AsifShah

  • QuantumATK Guru
  • ****
  • Posts: 144
  • Country: in
  • Reputation: 2
    • View Profile
Re: convergence problem
« Reply #9 on: June 21, 2022, 14:17 »
You need to increase it thats the only way. I faced same issue some time ago but got results.
Try 300 maybe

Offline Ambika kumari

  • Heavy QuantumATK user
  • ***
  • Posts: 32
  • Country: in
  • Reputation: 0
    • View Profile
Re: convergence problem
« Reply #10 on: June 21, 2022, 14:18 »
ok Thanku. I will check.