Problem with convergence test

[barnali]

covergence.Zip

Respected Sir (Daniele Varsano),

I am a new user of YUMBO. I want to do the convergence test for my further calculation. As for my knowledge, the convergence test for "yambo -g n -p p" are: k-point, EXXRLvcs, BndsRnxp, NGsBlkxp, GbndsRange. First of all I relax the structure of graphene (vc-relax with QE) and trying to perform the yambo convergence tests for k-point. First of all I have done scf and nscf calculation at gamma point with QE. I have followed the same procedure as described in Si bulk tutorials. In the tutorial by running GoWo_gap.sh (YAMBO_TUTORIALS\Solid_Si\WorkSpace\scripts) I am able to get the correct data. But I met some difficulties when I am trying to use this script for my structure. Then, I checked each part of the script and compare both the result. I have found that in case of Si, in the scrip, t it is written that ‘grep 0.000000 ../../YAMBO/k_grid_Gamma/Results/o-GoWo_PPA_HF7Ry_X0Ry-nb10_nb10.qp | head -n 2 | tail -c 21 | head -c 10’ and it grep 1.203355 ( d_GoWo_PPA_v ) from the .qp file
……………………………………………………………………………………………………………………….
K-point Band Eo E-Eo Sc(Eo)
………………………………………………..
1.000000 2.000000 0.000000 1.203355 4.983801
1.000000 3.000000 0.000000 1.203355 4.983801
1.000000 4.000000 0.000000 1.203355 4.983801
………………………………………………………….
…………………………………………………………………………………………………………..
Then written the command ‘ grep -A 1 0.000000 ../../YAMBO/k_grid_Gamma/Results/o-GoWo_PPA_HF7Ry_X0Ry-nb10_nb10.qp | head -n 4 | tail -c 21 | head -c 10’ , which gives 1.342991 (d_GoWo_PPA_c1).
But in my case, in the o.qp there are only one 0.000000 value in Eo.

Could you please guide me to get (d_GoWo_PPA_v) and (d_GoWo_PPA_c1) value from any o.qp file of any structure?
Here I have attached my input and output file and script and qp file for Si.

I am waiting for your valuable suggestion. Thanking you.

Sincerely

Barnali Bhattacharya
Ph.D student
Department of Physics,
Assam University, India.

[Daniele Varsano]

Dear Barnali,

The script it is just used to automatize the analysis of the convergence, but it is not at all required. My suggestion is that you analyze your results one by one, checking your output and your reports. In this way you can recognize easily if something went wrong. Or alternatively you can write on your own the script in order to do what you aim.

In any case, the convention of Yambo is to set as 0 energy the last occupied Kohn-Sham level, this is because in the script the 0.00000 value it is searched.
So, you also have the 0.000000 value in your output:
which is found here:

Code: Select all

 1.000000   4.000000   0.000000  -2.791427   5.976181

Anyway the script it is thought for that particular output as it is based on the number of line of the file (head -xx, tail --yy etc...) so it is not at all general, and it is practical for us to be used in school and to make that specific tutorial.
So I really suggest you to not use the script, and copy the the values in your run in an external file in order to analyze the convergences. Or, if you want to use the script you need to prepare one for your case.

Best,
Daniele

[barnali]

Capture.JPG

2.JPG

Respected sir (Daniele Varsano),

Thank you for your valuable suggestion. According to your suggestion, I have analyzed my report and .qp file. I have found that, the E-Eo value (marked by red circle in my attached capture.jpg) corresponding to the 0.00000 energy of Kohn-Sham level (Eo) is my V value and the E-Eo value just below that value is my C1 value (marked by blue circle in my attached capture.jpg). Then I have searched for DFT indirect band gap in Eo (shown in the green circle in capture.jpg) and taken the corresponding value of E-Eo as my C2 value.

Now my direct band gap and indirect band will be-
DFT direct gap – V +C1 = Direct gap. (5.491357)
DFT indirect gap –V +C2= indirect gap.(5.491357)

1) Could you please tell me is my way of getting GW corrected Direct and indirect band gap is right or wrong?
2) If wrong, then how can I get this value?

I am waiting for your reply. Thank you.

Sincerely
Barnali Bhattacharya
Department of physics,
Assam University, India.

[Daniele Varsano]

Dear Barnail,
yes it looks me ok. Of course, in this case direct and indirect are the same. Even if you have a system with different direct and indirect gap, you cannot see that when dealing with gamma point only.

Best,
Daniele

[barnali]

k-point.rar

Respected Sir (Daniele Varsano),

Thank you for your quick response. I again met some difficulties about k-point convergence test. I have not found convergence for graphene when I proceed for k-point gamma to 14 *14*1. Again I have found negative indirect and direct band gap for 6*6*1. First of all I have relaxed (vc-relax) the structure with 12*12*1 k-point with occupation ‘smearing’. For convergence test, then I have done the scf and nscf calculation for gamma to 14 *14*1 k –point in different folder. After that, entering in each .save file I have use following command-
 p2y –N
 yambo –D
 yambo –i
 yambo
 yambo –g n –p p
 yambo –M
I have found r_setup and o.qp in each .save file and calculated the direct and indirect band gap. I have already discussed in my previous mail how I have calculated the direct and indirect band gap from r_setup and o.qp file.
I am attaching all the input and output file here, along with jpg file.
Could you please tell me where I have made mistakes?
And how can I overcome this problem?
I am waiting for your kind response. Thanking you in advance.

Sincerely
Barnali Bhattacharya
Department of physics,
Assam University, India

[Daniele Varsano]

Dear Barnali,
you should always post the yambo report files in order to inspect it.
Some considerations:
1) Grapehene has zero gap, you should look at the k-points having zero gap, so you need to include it in your sampling (1/3, 1/3,0) or at least k-point closed to the Dirac point.
2) From your data, looking at the DFT results, you are looking at different k-point in different sampling (of course if you look at gamma, or few k-point you do not include it), so it is not a real convergence test.
3) DFT gap has to be the same you find in QE, you can check it.
4) In order to look at always at the same k-point you can check the convergence of the correction at Gamma, as it is present in all your samplings, taking care that near the Dirac point the gap approaches to zero, in order to be sure there are not problems related to your geometry, DFT calculation etc.
5) negative gap is a nonsense by definition, in the sampling 6x6x1 and 12x12x1 it looks you are looking at the Dirac point, or closed to it, as you have a DFT gap very small,
the correction is also very small, ie nearly zero, which is consistent. From what you say, consider that you are adding a smearing, so you are also populating the conduction band according to the smearing, try to put the smearing as small as you can.


So check the DFT results first and then pass to the GW correction. Then choose the k-point you want to inspect (possibly the same for all the samplings) and calculate thte corrections for that points (in this way you will also reduce the computation time ( this is controlled by %QPkrange ) and you can put is for just one k-point and the bands you are interested in, of course the k-point index will change in different samplings.

Other hint looking at your input:
1) Chimod= "" , set it to Hartree, you can do it by command line yambo -p p -k hartree -g n
2) 10 bands in BndsRnXp and GbndRnge could be very few, but this is something you need to check.

Best,
Daniele

[barnali]

input.rar

Respected Sir,
Thank you for your suggestion. Still, I have some confusion about the convergence test for k-point in Yambo. Please guide me to fix this. Now I am taking graphyne as my system which is a semiconductor having a direct band gap of 0.45 eV. In QE I have checked the cutoff and k-point (in terms of energy) and found that 6*6*3 (automatic) k-point is a good assumption. Taking this k-point and cutoff I have done ‘VC-relax’ calculation and again done scf, nscf and bands calculation and plot it, I have found the band gap of 0.42 eV and the gap situated at M point of brilliant sun.
Now I want to do the convergence test for k-point in yambo. I have built 8 folders including gamma. In each folder I have done scf calculation with k-point 6*6*3 (as I see in Si Tutorial in each folder, including gamma the scf file contain 4*4*4 k-point ). In nscf file I have just changed the k point (for gamma 1*1*1). After performing scf and nscf calculation I have use the command in .save
p2y –N,
yambo –D,
yambo –i,
yambo respectively.
I have seen in Si example the DFT direct gap is same for all k-point sampling including gamma. But in my case the gap in gamma is too large and after that it is decreasing.
I am confused am I wrong in the DFT calculation or can I proceed for GW correction.?
Could you please explain more elaborately how can I proceed for convergence test?
In your previous mail you have written “ 2) From your data, looking at the DFT results, you are looking at different k-point in different sampling (of course if you look at gamma, or few k-point you do not include it), so it is not a real convergence test.” Please explain more elaborately.


I am waiting for your kind reply. Thanking you.

Sincerely

Barnali Bhattacharya
Department of Physics,
Assam University, India

[Daniele Varsano]

Dear Barnali,
the attachment is not present in your mail.
Anyway, if the gap is at any k-point different from gamma (K,M...) how can you calculate it, if you include gamma only in the calculation?
Check your band structure, and pay attention to focus on a k point which is common in all the sampling. This can be easily done looking at the band structure and just
the "Direct/Indirect gap" in the report, which is the gap available according to the k-point sampling (for instance at gamma if you have gamma only).
Best,
Daniele

[barnali]

QE.rar

Respected Sir,
By checking the band structure, and report file I have found that Gamma point (0.00000 0.00000 0.00000) is common for all k –point sampling. And the 4.194218 value is common in all the r_setup file (when I checked the direct /indirect gap in r_setup ). Please guide me with this information, how can I do convergence test for k-point?
I have attached all the input and output file yambo and QE.
I am waiting for your suggestion. Thanking you.

YAMBO.rar

Sincerely
Barnali Bhattacharya
Ph.D student, Assam University

[Daniele Varsano]

Dear Barnali.
sorry, but I cannot get what is your problem.
I inspected your setup files and looking at the reports you have for instances looking at gamma (bands 24 and 25)

Code: Select all

ksampl  b24     b25             gap
111 0.000000 4.194218 4.19422
221 -1.594256 2.599961 4.19422
441 -1.594256 2.599961 4.19422

this means that at DFT level the gap is converged. Of course in the sampling with only gamma the band 24 is set to zero because with this sampling, this is the Fermi level. Next you need to do GW calculation on top of that. If you want to check directly the real band gap value at M point you need to include that points in your samplings. Sorry, but I did not get which difficulty you are facing.

Best,
Daniele