Hints on procedure: single molecule spectra

[bob]

Hi folks,

I need a bit of assistance from the experts here in the forum with the following task.

I have a single molecule, e.g. perylene diimide, for which I need to determine
a) the single-particle spectrum
b) excitonic levels.

So far, I did the following:

1. run quantum-espresso (might better be replaced by abinit due to missing nonlocal PP contributions in BSE)
2. convert using p2y
   Here, I already realized that the calculations use a lot of Gvectors

   Code: Select all

   G cutoff =10945.1917  ( 624469 G-vectors)     FFT grid: (216,135, 90)
   G cutoff = 7004.9227  ( 319643 G-vectors)  smooth grid: (180,108, 72)

   so it takes a while to complete the shell finding.
3. run a GW calculation with

   Code: Select all

   yambo -g n -p p

4. get the statically screened e-e interaction with

   Code: Select all

   yambo -b

5. solve the BSE with

   Code: Select all

   yambo -o b -y d

I did this without changing any of the default parameters. During step 5, I get the following output:

Code: Select all

 <---> [M 0.015 Gb] Alloc g_vec k_pt dl_sop wf_nc_k wf_igk (0.015)
 <---> [M 0.026 Gb] Alloc RL_Gshells RL_Eshells (0.011)           
 ___ __  _____  __ __  _____   _____                              
|   Y  ||  _  ||  Y  ||  _  \ |  _  |                             
|   |  ||. |  ||.    ||. |  / |. |  |                             
 \_  _/ |. _  ||.\_/ ||. _  \ |. |  |                             
  |: |  |: |  ||: |  ||: |   \|: |  |                             
  |::|  |:.|:.||:.|:.||::.   /|::.  |                             
  `--"  `-- --"`-- --"`-----" `-----"                             

 <---> [01] Job Setup
 <---> [02] Input variables setup
 <---> [02.01] Unit cells        
 <---> [02.02] Symmetries        
 <---> [02.03] RL shells         
 <---> [02.04] K-grid lattice    
 <---> [02.05] Energies [ev] & Occupations
 <---> [03] Transferred momenta grid      
 <---> X indexes |####################| [100%] --(E) --(X)
 <---> SE indexes |####################| [100%] --(E) --(X)
 <---> [04] Bethe-Salpeter Kernel                          
 <---> [BSE] Kernel dimension    : 1824                    
 <---> [04.01] Screneed interaction header I/O             
 <---> Filling Estimation |####################| [100%] --(E) --(X)
 <---> [P 01] Kernel filling [o/o] 100.0000                        
 <---> [WARNING]Exchange FFT size is too big. RL vectors reduced to 1
 <---> [M 0.163 Gb] Alloc WF (0.127)                                 
 <---> [FFT-BSK] Mesh size:  84   54   35                            
 <---> [WF loader] Wfs (re)loading |####################| [100%] --(E) --(X)
 <---> [04.02] Main loop                                                    
 <---> [M 0.190 Gb] Alloc BS_mat (0.027)                                    
 <---> BSK |####################| [100%] --(E) --(X)                        
 <---> [M 0.163 Gb] Free BS_mat (0.027)                                     
 <---> [M 0.036 Gb] Free WF (0.127)                                         
 <---> [05] BSE solver(s)                                                   
 <---> [WARNING] Missing non-local pseudopoential contribution              
 <---> [05.01] Diagonalization solver                                       
 <---> [M 0.090 Gb] Alloc BS_mat (0.053)                                    
 <---> BSK diagonalize |####################| [100%] --(E) --(X)            
 <---> BSK   residuals |####################| [100%] --(E) --(X)            
 <---> BSK res epsilon |####################| [100%] --(E) --(X)            
 <---> [M 0.037 Gb] Free BS_mat (0.053)                                     
 <---> [06] Game Over & Game summary    

The line
<---> [WARNING]Exchange FFT size is too big. RL vectors reduced to 1
is a bit worrisome. I also found it quite weird that the spectrum in "o.eps_q001-bd" is the same whether or not I do the GW calculation or not. Could that be related to the above warning?

If anyone has advice on how to chose appropriate parameters for such a calculation, I'd greatly appreciate any assistance.

Cheers,
Bjoern

[Daniele Varsano]

Dear Bjoern,

careful with using default values. The input files generated by Yambo are meant
to be edited. In all the steps you performed there are parameters that
have to be carried out to convergences. As for instance the number of bands
that enter in the calculation of the calculation of the screening, in the
GW calculations and in the BSE. As well the Gblk, that by default are
set to 1, and have to be increased until convergence.

About the QP correction, when running the BSE you have to use
the verbosity option -V qp. Then in the input you can insert the Qp correction,
for instances
KfnQPdb= "E < ./SAVE/ndb.QP" # [EXTQP BSK BSS] Database


Anyway in general, I strongly suggest you to take your time to learn how
to run converged calculation and get meaningful results with Yambo by looking
at the theory and run the tutorial you can find in the webpage.

Cheers,

Daniele

[bob]

Dear Daniele,

I am sorry if I was a bit diffuse. I did have a look at the tutorials but after reading it, I wasn't quite sure though whether I got the steps in principle right or not. Also I was unsure whether or nor these single molecule systems require special treatment.

Anyway, I tried to read the QP corrections as you mentioned. I did not have nbd.QP in the SAVE directory but only db.QP (I guess because I am not using netcdf) but then yambo stops with the error

Code: Select all

<---> [04] External QP corrections (K)At line 379 of file mod_IO.F (unit = 42, file = '')
Fortran runtime error: File already opened in another unit
[\code]

I am a bit confused about what is happening now - and also a bit embarrassed for asking so many questions... 

Cheers,
Bjoern

[Daniele Varsano]

Dear Bjoern,
please have a look to
this post by Claudio.
Anyway I strongly suggest you to link netcdf library.

and also a bit embarrassed for asking so many questions...

don't worry, this is why we opened the forum.

Cheers,

Daniele

[andrea marini]

Dear Bjoern,

regarding the I/O problem I am not sure whether the problem is the same reported in the post linked by Claudio. In any case that problem has been solved in the latest revision (6) available from the SVN repository. If you want to access this revision you need to register on the qe-forge portal as explained in the download page. Otherwise in 7/10 days we will release the tar-ball of revision 6.

Please report us of the problem persists also with the latest revision.

Regarding the other issues I agree with Daniele's post. Please consider that Yambo pre-writes the input file following the command line to suggest the NAME of the variables to be used, not their value. The latter must be carefully decided and checked by the user on the basis of the system under study. TO help you in understanding the meaning of these variables try re-running the tutorials without using the provided input files but using yambo to create them from scratch.

[bob]

Hi guys,

first, thanks for all the help so far.

regarding the I/O problem I am not sure whether the problem is the same reported in the post linked by Claudio. In any case that problem has been solved in the latest revision (6) available from the SVN repository. If you want to access this revision you need to register on the qe-forge portal as explained in the download page. Otherwise in 7/10 days we will release the tar-ball of revision 6.

Please report us of the problem persists also with the latest revision.

The patch worked. I tried to checkout the SVN version anyway, but it fails with "svn: Network connection closed unexpectedly", so I stick to the patched code, for now.

I have been looking into the settings. From the tutorials, I guess that the H2 example is the most relevant for my problem. However, there is no example of obtaining QP corrections for such a system (correct me, if I missed it). After I ran yambo without changing parameters (just to see whether it works), I increased NGsBlkXd and tried to rerun. However, then I get another I/O error:

Code: Select all

<---> [06] Dyson equation: Newton solverAt line 379 of file mod_IO.F (unit = 42, file = '')
Fortran runtime error: File already opened in another unit    

Also the error persists when I set NGsBlkXd=1 again. I assume that yambo reads data from the previous crashed run from somewhere. Is there a command to clean this old data?

Cheers,
Bjoern

[andrea marini]

The patch worked. I tried to checkout the SVN version anyway, but it fails with "svn: Network connection closed unexpectedly", so I stick to the patched code, for now.

Ok. As you prefer. To use the svn repo you need to register first to the qe-forge portal and follow the instructions to become member of the Yambo project. Only in that case your connection to svn will be accepted. If you want you will be more than welcome.

I have been looking into the settings. From the tutorials, I guess that the H2 example is the most relevant for my problem. However, there is no example of obtaining QP corrections for such a system (correct me, if I missed it).

This is correct. There is no tutorial about GW, so far

After I ran yambo without changing parameters (just to see whether it works), I increased NGsBlkXd and tried to rerun. However, then I get another I/O error:

Code: Select all

<---> [06] Dyson equation: Newton solverAt line 379 of file mod_IO.F (unit = 42, file = '')
Fortran runtime error: File already opened in another unit    

Also the error persists when I set NGsBlkXd=1 again. I assume that yambo reads data from the previous crashed run from somewhere. Is there a command to clean this old data?

Yambo does not read old creshed data. However, as I am not sure about all consequences of the patch provided by Claudio I cannot tell if that patch introduced other problems. But, in any case, this problem should be related to the

Code: Select all

KfnQPdb= "E < ./SAVE/ndb.QP" # [EXTQP BSK BSS] Database

Try to remove it. If the problem persists please provide us will all necessary informations (listed in this post) and we will do our best to solve it.

[bob]

After I ran yambo without changing parameters (just to see whether it works), I increased NGsBlkXd and tried to rerun. However, then I get another I/O error:

Code: Select all

<---> [06] Dyson equation: Newton solverAt line 379 of file mod_IO.F (unit = 42, file = '')
Fortran runtime error: File already opened in another unit    

Also the error persists when I set NGsBlkXd=1 again. I assume that yambo reads data from the previous crashed run from somewhere. Is there a command to clean this old data?

Yambo does not read old creshed data. However, as I am not sure about all consequences of the patch provided by Claudio I cannot tell if that patch introduced other problems. But, in any case, this problem should be related to the

Code: Select all

KfnQPdb= "E < ./SAVE/ndb.QP" # [EXTQP BSK BSS] Database

Try to remove it. If the problem persists please provide us will all necessary informations (listed in this post) and we will do our best to solve it.

It does not occur at the BSE level, though. I was redoing the GW calculation when this error occurred. Looks like I should really checkout the SVN repo. I did register and activate my qe-forge account before but I wasn't member of the Yambo project. Is it required to be a member to access the repo?

[andrea marini]

I did register and activate my qe-forge account before but I wasn't member of the Yambo project. Is it required to be a member to access the repo?

I have already activated your membership. So you should already be able to checkout the code. Does it work ?

[bob]

I did register and activate my qe-forge account before but I wasn't member of the Yambo project. Is it required to be a member to access the repo?

I have already activated your membership. So you should already be able to checkout the code. Does it work ?

Nope. I tried it also on a different machine and there the error message was slightly more helpful:

Code: Select all

>svn checkout svn+ssh://bob@scm.qe-forge.org/svn/yambo
Permission denied (publickey).
svn: Connection closed unexpectedly

But even after uploading my public keys and waiting some time, it doesn't accept it.