QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: lknife on October 8, 2017, 04:28

Title: Is this EDP curve correct?
Post by: lknife on October 8, 2017, 04:28: Hello,

I made a device configuration and calculated the EDP curve of it. However, there is something I cannot understand. Please see the attached figure of the EDP plot and the .py file for it, notifying the Y value of the plot. Can anybody here tell me if the EDP curve is correct? I used the ATK 2016.4 to do the calculation.

Thank you very much for your help!
Title: Re: Is this EDP curve correct?
Post by: Petr Khomyakov on October 9, 2017, 10:03: It might be that you have chosen Sum instead of Average for Projection type while plotting the potential in the 1D Projector. This could in principle give you such huge numbers for the potential.
Title: Re: Is this EDP curve correct?
Post by: lknife on October 9, 2017, 10:41: Thank you very much for your reply.
However, I do chose the "average" for the Projection type. If I choose the type to "Sum", the figure will be shown as the attached file.
Title: Re: Is this EDP curve correct?
Post by: Jess Wellendorff on October 10, 2017, 08:19: Looks like your NEGF calculation did not converge. Check the ATK log file.
Title: Re: Is this EDP curve correct?
Post by: Daniele Stradi on October 10, 2017, 09:14: Could you please have a look at the density matrix of the last SCF loop in your output file? In particular, the two columns DM (density matrix) and DD (density difference).

+------------------------------------------------------------------------------+
| Density Matrix Report DM DD |
+------------------------------------------------------------------------------+
| 0 In [ -0.000 , 0.000 , 0.000 ] 3.52719 0.52719 |
| 1 As [ 1.520 , 1.520 , 1.520 ] 4.47281 -0.52719 |
+------------------------------------------------------------------------------+
| 3 E = -2.67569 dE = 2.502366e-04 dH = 9.448242e-05 |
+------------------------------------------------------------------------------+

My first guess is that the charge density has been drained away from the central region due to inappropriate simulation parameters. In this case, the column DM should be 0.0.
Regards,
Daniele
Title: Re: Is this EDP curve correct?
Post by: lknife on October 10, 2017, 16:26: Quote from: Jess Wellendorff on October 10, 2017, 08:19
Looks like your NEGF calculation did not converge. Check the ATK log file.

You are right: I have checked the .log file and the calculation did not converge. Another question: how to make it converge? What parameters should I tune, or I need to modify my device configuration?

Thank you very much for your kind help!
Title: Re: Is this EDP curve correct?
Post by: lknife on October 11, 2017, 03:31: Quote from: Daniele Stradi on October 10, 2017, 09:14

+------------------------------------------------------------------------------+
| Density Matrix Report DM DD |
+------------------------------------------------------------------------------+
| 0 In [ -0.000 , 0.000 , 0.000 ] 3.52719 0.52719 |
| 1 As [ 1.520 , 1.520 , 1.520 ] 4.47281 -0.52719 |
+------------------------------------------------------------------------------+
| 3 E = -2.67569 dE = 2.502366e-04 dH = 9.448242e-05 |
+------------------------------------------------------------------------------+

...In this case, the column DM should be 0.0.

Thank you very much for you help! I have a question about the words marked in blue: if DM=0.0, is it correct or incorrect?
Title: Re: Is this EDP curve correct?
Post by: Jess Wellendorff on October 11, 2017, 08:10: DM means DensityMatrix, so if DM=0.0 there are no valence electrons on that atomic site, which is almost always very bad and indicates poor SCF convergence. You need to inspect the SCF convergence in order to figure out how to fix it. Perhaps it will converge in 20 more iterations? Perhaps electrons are depleted from the central region - electrode boundaries, in which case you perhaps need to use longer electrodes, etc. It's impossible to give a completely general recipe for this kind of thing.
Title: Re: Is this EDP curve correct?
Post by: lknife on October 11, 2017, 14:51: Thank you very much for your explanation!

The followings are part of the .log file about the SCF iteration (the full file is too big to be posted here). Could you please have a loot at it to see what's wrong with it?

Thanks a lot again for your kind help!

------------------------------------------------------------

|--------------------------------------------------|
Calculating Density Matrix : ==================================================
+------------------------------------------------------------------------------+
| Density Matrix Report DM DM[D] DD |
+------------------------------------------------------------------------------+
| 0 Te [ 16.799 , 0.875 , 0.801 ] 7.29316 7.29194 -1.41489 |
| 1 Te [ 13.695 , 2.624 , 1.739 ] 0.11796 0.11840 -15.76363 |
| 2 W [ 15.353 , 0.875 , 3.303 ] 6.29553 6.29539 -1.40908 |
| 3 Te [ 17.473 , 2.624 , 3.891 ] -0.09463 -0.09465 -16.18927 |
| 4 Te [ 13.022 , 0.875 , 4.836 ] 6.81351 6.81183 -2.37466 |
| 5 W [ 15.142 , 2.624 , 5.509 ] 5.36223 5.36642 -3.27135 |
| 6 Te [ 16.799 , 0.875 , 7.111 ] 4.26804 4.26763 -7.46433 |
| 7 Te [ 13.695 , 2.624 , 8.048 ] 4.52539 4.52769 -6.94692 |
| 8 W [ 15.353 , 0.875 , 9.612 ] 9.01083 9.01866 4.02949 |
... ...
| 83 W [ 14.628 , 2.624 , 88.339 ] 4.99012 5.02517 -3.98470 |
+------------------------------------------------------------------------------+
| 498 E = -747.46 dE = 5.595637e+01 dH = 2.538735e+02 |
+------------------------------------------------------------------------------+

|--------------------------------------------------|
Calculating Density Matrix : ==================================================
+------------------------------------------------------------------------------+
| Density Matrix Report DM DM[D] DD |
+------------------------------------------------------------------------------+
| 0 Te [ 16.799 , 0.875 , 0.801 ] 6.79753 6.84718 -2.35530 |
| 1 Te [ 13.695 , 2.624 , 1.739 ] 0.17277 0.17969 -15.64754 |
| 2 W [ 15.353 , 0.875 , 3.303 ] 6.41454 6.41024 -1.17522 |
| 3 Te [ 17.473 , 2.624 , 3.891 ] -0.09635 -0.09908 -16.19542 |
| 4 Te [ 13.022 , 0.875 , 4.836 ] 6.15977 6.15823 -3.68200 |
| 5 W [ 15.142 , 2.624 , 5.509 ] 5.34169 5.33755 -3.32076 |
| 6 Te [ 16.799 , 0.875 , 7.111 ] 5.57543 5.56182 -4.86275 |
| 7 Te [ 13.695 , 2.624 , 8.048 ] 4.90536 4.89295 -6.20169 |
| 8 W [ 15.353 , 0.875 , 9.612 ] 9.02604 9.02771 4.05374 |
... ...
| 83 W [ 14.628 , 2.624 , 88.339 ] 2.57530 2.48627 -8.93844 |
+------------------------------------------------------------------------------+
| 499 E = -667.988 dE = 7.947253e+01 dH = 2.527983e+02 |
+------------------------------------------------------------------------------+
################################################################################
# #
# Warning: The calculation did not converge to the requested tolerance! #
# #
################################################################################
Title: Re: Is this EDP curve correct?
Post by: Anders Blom on October 11, 2017, 16:21: What version of ATK is this?
Title: Re: Is this EDP curve correct?
Post by: Jess Wellendorff on October 12, 2017, 09:00: Perhaps you can also attach the ATK script?
Title: Re: Is this EDP curve correct?
Post by: lknife on October 12, 2017, 14:18: I am using ATK2016.4.

After increasing the k-point sampling from (1 7 50) to (1 9 100) and tuning some parameters such as dampling factor (from 0.1 to 0.3) and history step (from 20 to 12), the EDP curve is now converged. I think the insufficient k-point sampling may be the main reason to the problem.

Thanks a lot to all of you!
Title: Re: Is this EDP curve correct?
Post by: Anders Blom on October 12, 2017, 16:32: Esp. in the C direction. It's typically not sufficient to use 50 points, unless you have a very long electrode in the first place.