## Data output

Various technical topics such as parallelism and efficiency, netCDF problems, the Yambo code structure itself, are posted here.
hplan
Posts: 58
Joined: Thu Mar 26, 2009 2:26 pm

### Data output

Dear developers:

I want to do some manipulation over response function related data such as, $\chi_0$ or $\chi$. Which subroutines should i read and modify first ? I just learn that O_driver.F accounts for output of the dielectric function. I wonder whether i could do some data porting here ? And Does the variable 'X_mat' contain the exact information of the response function ? Is there any detailed formulation for this variable ? In fact, i think it is the response function i am looking for !

Regards,
Hai-Ping
Hai-Ping LAN,
Department Of Electronics,
Peking University, 100871,Beijing, CHINA

myrta gruning
Posts: 308
Joined: Tue Mar 17, 2009 11:38 am
Contact:

### Re: Data output

Hallo Hai-Ping
I want to do some manipulation over response function related data such as, $\chi_0$ or $\chi$. Which subroutines should i read and modify first ? I just learn that O_driver.F accounts for output of the dielectric function. I wonder whether i could do some data porting here ? And Does the variable 'X_mat' contain the exact information of the response function ? Is there any detailed formulation for this variable ? In fact, i think it is the response function i am looking for !
First, make sure that you really need to touch the code and that you cannot just manipulate the output files.
then yes the $\chi_0$ or $\chi$ are calculated in O_driver.F. For a given q (there is a loop on q) $\chi_0(G,G, \omega)$ comes out after the call to X_os and $\chi(G,G, \omega)$ after the call of X_s.

cheers
m
Dr Myrta Grüning
Atomistic Simulation Centre
Queen's University Belfast - Northern Ireland

http://www.researcherid.com/rid/B-1515-2009

hplan
Posts: 58
Joined: Thu Mar 26, 2009 2:26 pm

### Re: Data output

Dear Myrta:
Thank you!
myrta gruning wrote:
First, make sure that you really need to touch the code and that you cannot just manipulate the output files.
then yes the $\chi_0$ or $\chi$ are calculated in O_driver.F. For a given q (there is a loop on q) $\chi_0(G,G, \omega)$ comes out after the call to X_os and $\chi(G,G, \omega)$ after the call of X_s.
I want to get the information of response function $\chi$ without other calculation to be performed. It seems all output data by yambo are related to dielectric function. By the way, i just found that the energy loss function calculated by yambo is not directly via $\im[\chi_{G=0,G'=0}]$. Why does yambo adopt other approach ? Is there any other consideration here ?

Best Wishes

Hai-Ping
Hai-Ping LAN,
Department Of Electronics,
Peking University, 100871,Beijing, CHINA

myrta gruning
Posts: 308
Joined: Tue Mar 17, 2009 11:38 am
Contact:

### Re: Data output

Hallo Hai-Ping
i just found that the energy loss function calculated by yambo is not directly via $\im[\chi_{G=0,G'=0}]$. Why does yambo adopt other approach ? Is there any other consideration here ?
Does it regard the use of Kramers Kronig relations when you have advanced or retarded ordering? or else?

cheers
m
Last edited by myrta gruning on Fri Jul 31, 2009 3:30 pm, edited 1 time in total.
Dr Myrta Grüning
Atomistic Simulation Centre
Queen's University Belfast - Northern Ireland

http://www.researcherid.com/rid/B-1515-2009

hplan
Posts: 58
Joined: Thu Mar 26, 2009 2:26 pm

### Re: Data output

Dear Myrta:

myrta gruning wrote: What is exactly your question?
Does it regard the use of Kramers Kronig relations when you have advanced or retarded ordering? or else?
I want to repeat a calculation for the loss function Al metal, -$\Im[\chi_{G=0,G'=0}(q,\omega)]$. However, i found the data i obtained from yambo for o.eel_qxxx were quite different from a published data. Then , i wonder there should be some different definitions for the loss function. That is why i want to do some data manipulation over EEL.

Best Wishes,

Hai-Ping
Hai-Ping LAN,
Department Of Electronics,
Peking University, 100871,Beijing, CHINA

andrea marini
Posts: 445
Joined: Mon Mar 16, 2009 4:27 pm
Contact:

### Re: Data output

hplan wrote: I want to repeat a calculation for the loss function Al metal, -$\Im[\chi_{G=0,G'=0}(q,\omega)]$. However, i found the data i obtained from yambo for o.eel_qxxx were quite different from a published data.
There are tons of possible reasons why you do not reproduce the published data. Please check carefully that your calculations are converged (Al is quite hard to converge) and post the reference of the paper you are comparing with.
hplan wrote: Then , i wonder there should be some different definitions for the loss function. That is why i want to do some data manipulation over EEL.
Different definitions of the loss ? As far as I know (and I coded the optical response of Yambo) there is only one definition of the loss for a solid. What do you mean as alternative definition ?
Andrea MARINI
Istituto di Struttura della Materia, CNR, (Italy)

myrta gruning
Posts: 308
Joined: Tue Mar 17, 2009 11:38 am
Contact:

### Re: Data output

Hallo Hai-Ping,

just one addition to what Andrea said.
You are referring to the loss function as -$\Im[\chi_{G=0,G'=0}(q,\omega)]$, but the definition is $-Im(eps^{-1}(q,\omega)) = -v(q)Im[\chi_{G=0,G'=0}(q,\omega)$ (v Coulomb interaction), and this is the quantity you get from yambo.

cheers

m
Dr Myrta Grüning
Atomistic Simulation Centre
Queen's University Belfast - Northern Ireland

http://www.researcherid.com/rid/B-1515-2009

hplan
Posts: 58
Joined: Thu Mar 26, 2009 2:26 pm

### Re: Data output

Dear Andrea:
andrea marini wrote: There are tons of possible reasons why you do not reproduce the published data. Please check carefully that your calculations are converged (Al is quite hard to converge) and post the reference of the paper you are comparing with.
I used the larger energy cutoff 10 Hartree and the same kgrid sampling (20x 20 x20) as described in the paper, Computer Physics Communications 180 466 to get kss data. I then checked the number of FFT for $\chi_0$, and also bands , the dimension of $\chi_0$ size for inversion (NGsBlkXd) for convergent data. I also tried to change the number of energy steps to remove artificial features. After all , my curve for the loss function was still quite different the published data .
lossAl.png
rpa_loss-Al.png
Different definitions of the loss ? As far as I know (and I coded the optical response of Yambo) there is only one definition of the loss for a solid. What do you mean as alternative definition ?
I was also very surprised when i looked the loss function for the high transfer momenta. I then browsed the routines for final data, and found the energy loss function was not really coded by $\Im[\chi_{G=0,G'=0}]$ which is why i wonder about this definition.

Thanks again!

Cheers,
Hai-Ping
You do not have the required permissions to view the files attached to this post.
Hai-Ping LAN,
Department Of Electronics,
Peking University, 100871,Beijing, CHINA

hplan
Posts: 58
Joined: Thu Mar 26, 2009 2:26 pm

### Re: Data output

Dear Myrta:
Thanks for reminding and clarification!
myrta gruning wrote:
just one addition to what Andrea said.
You are referring to the loss function as -$\Im[\chi_{G=0,G'=0}(q,\omega)]$, but the definition is $-Im(eps^{-1}(q,\omega)) = -v(q)Im[\chi_{G=0,G'=0}(q,\omega)$ (v Coulomb interaction), and this is the quantity you get from yambo.
In fact, i noticed this difference between my expression and the definition, which is also the reason i want to manipulate over the data to get the exact the curve for -$\Im[\chi_{G=0,G'=0}](q,\omega)$. But after i examined the formula and the data curve, i found things cannot be changed even i get -$\Im[\chi_{G=0,G'=0}](q,\omega)$ curve from yambo, since the curves for high transfer momenta are quite different from the published data . My data present some noice oscillations for hign transfer momenta even i increased energy steps and included more bands . The most important difference i found is that the so-called plasmon energy is not the same: mine is around 22 eV for (0.40, 0., 0.) (iku) while the paper giving is about 15.8 eV for (0.40, 0.0 ,0.0) $2\pi/a$.
By the way what does iku exactly stands for ? In my situation, i just thought (0.40, 0., 0.0)(iku) was the same as
(0.40, 0.0 ,0.0) $2\pi/a$ .

Thanks again for time and patience!

Cheers,
Hai-Ping
Hai-Ping LAN,
Department Of Electronics,
Peking University, 100871,Beijing, CHINA

andrea marini
Posts: 445
Joined: Mon Mar 16, 2009 4:27 pm
Contact:

### Re: Data output

hplan wrote: The most important difference i found is that the so-called plasmon energy is not the same: mine is around 22 eV for (0.40, 0., 0.) (iku) while the paper giving is about 15.8 eV for (0.40, 0.0 ,0.0) $2\pi/a$.
This sounds very strange to me.
hplan wrote:By the way what does iku exactly stands for ? In my situation, i just thought (0.40, 0., 0.0)(iku) was the same as
(0.40, 0.0 ,0.0) $2\pi/a$
Please attach any report file of your calculations and the input file of abinit. Be carefull the iku stands for internal K units, and check the the alat used by Yambo is the same of the one used in the CPC paper. I already wrote somewhere else that alat in Yambo is a simple internal parameter used to rescale the k-pt vectors.
Andrea MARINI
Istituto di Struttura della Materia, CNR, (Italy)