Yambo q dependent TDDFT kernel

[hongtang2023]

Dear Davide,

I am still trying TDDFT run for EELS for q not zero. I meet some questions I need your help.

1. for original Yambo.5.1.1 code, I do
"
tddft %% [R][K] Use TDDFT kernel
Chimod= "ALDA"
% QpntsRXd
1 | 10 | %% [Xd] Transferred momenta
%
"
I want to do EELS for q1 to q10. The ALDA kernel is q independent, it assumes q=0. So, with above input, I found that Yambo first calculate the Fxc kernel (q=0), then it calculates the eels for q1 to q10. The ALDA kernel ONLY was calculated for one time, before calculating eels for q1. For q2 to q10, Yambo just uses the first calculated kernel (q=0) for EELS for q2 to q10.
So, if I input
" % QpntsRXd
5 | 10 | %% [Xd] Transferred momenta
% "
Yambo still first calculate ALDA kernel with q=0,then do the EELS for q5 to q10.

Is the above understanding correct?



2. I modified tddft_alda_g_space.F file, in order to input a q dependent kernel MCP07. However, in the input for Yambo run, I still use ( but ALDA was modified with MCP07 in code)
"
tddft
Chimod= "ALDA"
"
This time, if I run for
"
% QpntsRXd
1 | 10 | %% [Xd] Transferred momenta
%
"
I found that Yambo first calculate MCP07 kernel with q=0, then do EELS for q1 to q10. But Yambo use the kernel evaluated with q=0 for Chi calculations for q2, q3, ...,q10 calculations ( note that q1=0,but q2 to q10 are not zero). This is not my purpose. I want Yambo calculate MCP07 kernel for a given q (q not zero).



3. continue with the above 2.
If I input
"
% QpntsRXd
5 | 5 | %% [Xd] Transferred momenta
%
"
then, I found that it looks that Yambo is calculating MCP07 kernel with q5, because I was able to code to get the value of q5 and send it to subroutine tddft_alda_g_space. However, after few seconds, Yambo met errors and stop.
The error are "
srun: error: nid004080: tasks 0-6: Segmentation fault
srun: Terminating StepId=25230019.0
slurmstepd: error: *** STEP 25230019.0 ON nid004080 CANCELLED AT 2024-05-06T01:51:30 ***
srun: error: nid004089: tasks 9-15: Segmentation fault
srun: error: nid004091: tasks 24-27,29-31: Segmentation fault
srun: error: nid004090: tasks 17-23: Segmentation fault
srun: error: nid004080: task 7: Terminated
srun: error: nid004090: task 16: Terminated
srun: error: nid004089: task 8: Terminated
...


"



Please let me know why I got this kind of errors.

I appreciate your help very much!

Best,
Hong
Dr. Hong Tang
Tulane University
New Orleans, LA
70118, USA

[Daniele Varsano]

Dear Hong Tang,

1) Yes, ALDA is local and can be evaluated once for all.
2-3) Please note, it is impossible for us to provide you advise and suggestion not knowing the new features you coded.

As a general suggestion

2. I modified tddft_alda_g_space.F file, in order to input a q dependent kernel MCP07. However, in the input for Yambo run, I still use ( but ALDA was modified with MCP07 in code)

As a good practice, I suggest you to leave the existing subroutine as it is, also for reference, and write a new one with your purpose. In this way you can have under control all you need to build your kernel and also understand what is going wrong.

Best,
Daniele