Open Shell Systems

[vitoversace]

Dear Daniele,

I have done for LiF a BSE run for singlet and triplet in 3.3.0 version of Yambo. But strangely i get for both absolutely the same optical spectrum ??? What am i doing here wrongly ?

As you see in my bse input file, i say, that it should write the excitonic wave functions to the disc but i no where see an output with excitonic wave functions !!!
ypp -e w
I get the ypp.in

What is the next step to do to get the excitonic wave functions?
Bests

[vitoversace]

By the way many thanks for your answers and patience.

[Daniele Varsano]

Dear Vito,
please take attention to the syntax of the input file.It is different if you use 3.3 or 3.4, have a look to the manual, there are two different section: older and 3.4.

In 3.3, in order to calculate the singlet you should consider the variable:

Code: Select all

BSresKmod= "xc"
BScplKmod="xc"

while to to have the triplet excitaion:

Code: Select all

BSresKmod= "c"
BScplKmod="c"

see the paper: PHYSICAL REVIEW B VOLUME 62, NUMBER 8 (2000)

In 3.4:

Code: Select all

BSKmod= "SEX"

for the singlet and:

Code: Select all

 
BSKmod= "TRIPLET"

for the triplet.

In your input file you are mixing 3.3. and 3.4 syntax. In order to not make confusione I suggest you to generate inout file by using the yambo command line.

As I told you in the 3.4, you have to apply the patch and it has still to be extensively tested.
In bulk system I do not expect big differences anyway between singlet and triplet excitations.

About the excwf: the WRbsWF flag write the eigenstates needed for the exc wf in the databases.
type:
> ypp -s
you will have the excitation energies sorted by intensity of by energy and the relative index.
Then
> ypp -e w
generate the input file. Edit it:
Choose the excitations you want to consider (seen in the output sorted file) and put the index in:

Code: Select all

States= "1 - 1"

if you want just one, say number 2, out "2 -2", if you want more "min - max".
Then in ypp.in choose the format you want the exc-wf, the direction of the plot (123) for a 3d plot etc.
Choose the number of repetition of the unit cell along each direction (this has to be an odd number contrary to what written in the legend) and choose where to put the hole in the unit cell.
Then run
>ypp
I suggest you to have a look to the tutorial, for instance in the tutorial "you-hydrogen" in the plot section there are examples of excitonic wavefunctions calculations and you can follow the procedure.

Best,
Daniele

[vitoversace]

Dear Daniele,
Thanks a lot, it works now, I see small deviations in intensity and position of peaks in LiF between singlet and triplet when using full diagonalization instead of Haydock. For Haydock they are identical.

1 ) BUT when i type : ypp -s for H2 in the tutorial, program tells me : it is an INVALID option. Then i tried ypp -e s as it's been said here http://www.yambo-code.org/input_file/ypp/index.php , but it does not generate the output file after running ypp .

command like : ypp -s s or ypp -s w , etc work perfectly but those for excitons do not produce for me the output file containing exciton energies and etc after running ypp !!! (very strange !!!)
(I am using right now yambo version 3.3.0.)

2 ) How can i know how large BSENGBlk and NGsBlkXs matrices should be ? Only by testing ?

Bests

[Daniele Varsano]

Dear Vito,

Thanks a lot, it works now, I see small deviations in intensity and position of peaks in LiF between singlet and triplet when using full diagonalization instead of Haydock. For Haydock they are identical.

Well, the diagonalization and Haydock in principle should give the same outcome if the Haydock procedure is well converged.

1 ) BUT when i type : ypp -s for H2 in the tutorial, program tells me : it is an INVALID option. Then i tried ypp -e s but it does not generate the output file after running ypp .

Sorry, it ws a mistake, the right syntaz is ypp -e s. It produces the o.exc_E_sorted and o.exc_I_sorted, ie, the excitation sorted by energy or intensity. You do not have to run ypp after. It works only after diagonalization (not Haydock as you need the (n)db.BS_diago database). Haydock procedure it is iterative and does not store eigenvalues and eigenfunction.

command like : ypp -s s or ypp -s w , etc work perfectly but those for excitons do not produce for me the output file containing exciton energies and etc after running ypp !!! (very strange !!!)
(I am using right now yambo version 3.3.0.)

The same it is true for the exciton wavefunction, you need to diagonalize the matrix having activated the flag WRbsWF, then ypp -e w will produce the input you post last time. After having edited it you need to run ypp.

2 ) How can i know how large BSENGBlk and NGsBlkXs matrices should be ? Only by testing ?

Yes, they are variables that have to be converged, you can set them to same values ie you calculate NGsBlkXs vectors in the screening and use them in the direct term in the BSE (BSENGBlk).

Best,
Daniele

[vitoversace]

Dear Daniele,

1) From what ever reason those commands in ypp tool for exciton's energy and intensity extraction just don't work in yambo 3.3.0 but other commands are working very well.
2) Is in yambo 3.3.0 the NON-self-consistent cohsex approximation implemented ? As you see in my o-ppa.qp file i have GW SC iterations : 0 !!! Or is there any keyword specifying the iteration number in the cohsex loop? I could not find any.
3) Is it okay that i get for all k points and band Z=1 in cohsex for LiF?
4) How can i get a cohsex + G0W0 in yambo?
5) Is it in yambo possible to use another better approximation for gw band structure calculation than HF_and_locXC as in tutorial ? (so instead of exact exchange self-energy (or HF) a self energy which contains correlation self energy, too ?)
6) Is exciton life time also implemented in the 3.4.1 version ?
7) I registered for SVN, does it take a long time to gain access to repository ? (I would like to test the 3.4.1 version very soon.)


With best wishes

[Daniele Varsano]

Dear Vito,

1) From what ever reason those commands in ypp tool for exciton's energy and intensity extraction just don't work in yambo 3.3.0 but other commands are working very well.

Very strange, it should work, if you want post your input and report files and I will take a look, otherwise (suggested) download the latest version.

2) Is in yambo 3.3.0 the NON-self-consistent cohsex approximation implemented ? As you see in my o-ppa.qp file i have GW SC iterations : 0 !!! Or is there any keyword specifying the iteration number in the cohsex loop? I could not find any.

Yes, it is. As it is non self consistent, but first order, no iteration are needed. The SC implementation it is not in the GPL release.

3) Is it okay that i get for all k points and band Z=1 in cohsex for LiF?

I would say yes, it means somehow that for LIF the KS wfs are a good approximation, indeed the NON-SC GW gives good results

4) How can i get a cohsex + G0W0 in yambo?

As I told you beofre, it is not possible directly with Yambo as the SC cohsex it is not present in the gpl version. To do that may be you should resort to Abinit COhsex electronic structure, but to have the KSS you should contact Abinit people and most probably some testing is needed.

5) Is it in yambo possible to use another better approximation for gw band structure calculation than HF_and_locXC as in tutorial ? (so instead of exact exchange self-energy (or HF) a self energy which contains correlation self energy, too ?)

I do not get the point, HF_and_locXC calculate <Sigma_x>-<Vxc> which is a part needed for the GW calculation. As starting point you have the KS band structure and not the HF.

6) Is exciton life time also implemented in the 3.4.1 version ?

No, and it is not very clear to me how to calculate it in this framework

7) I registered for SVN, does it take a long time to gain access to repository ? (I would like to test the 3.4.1 version very soon.)

No it should be very fast.

Best,
Daniele

[vitoversace]

Dear Daniele,

I switched now to the 3.4.1 devel version.

1 ) I think there is probably some thing wrong with patch file which u sent me.

patching file src/interface/init.F
Hunk #1 succeeded at 428 (offset 5 lines).
Hunk #2 succeeded at 774 (offset 6 lines).
Hunk #3 succeeded at 821 (offset 6 lines).
Hunk #4 succeeded at 872 (offset 5 lines).
Hunk #5 FAILED at 1038.
1 out of 5 hunks FAILED -- saving rejects to file src/interface/init.F.rej
patching file src/modules/mod_drivers.F
patching file src/modules/mod_stderr.F
patching file src/modules/std_presets.F
Hunk #1 succeeded at 234 (offset 1 line).

2 ) There is also a small error in the devel version of yambo :

type(levels) :: E_k_p_q

Error: Object 'e_k_p_q' at (1) must have the SAVE attribute for default initialization of a component
make[1]: *** [mod_ELPH.o] Error 1
make[1]: Leaving directory `/home/vito/my_local_copy/devel/src/modules'
make: *** [yambo_ph] Error 2

3 ) For taking in bse calculation the coupling term in 3.4.1 into account, should i include both BSEmod = coupling AND WehCpl or is it enough to have one of them ?
4 ) For HF_and_locXC i am just wondering where the correlation part of self energy of gw approximation is going to be calculated. HF_and_locXC : this is only exact exchange self energy - exchange-correlation potential of DFT, so where is correlation self energy ?

5) I tried to run a bse calculation on LiF using this time em1d (dynamic dielec. matrix) but it seems that yambo only works with the static one (em1s), am i right ???
6) I think:

imaginary part of poles of 4-point Green's function gives exciton life time.

7) What are the keywords to do a sc gw and sc cohsex calculation for both energies AND wave functions in 3.4.1 devel version ?
8) I saw in some input files that in the static cohsex approximation one uses plasmon pol model !!! Why is this the case, so if cohsex is static, so there is no frequency integration any more.

With best wishes

[Daniele Varsano]

Dear Vito,

1 ) I think there is probably some thing wrong with patch file which u sent me.

The patch was done for the 4.0 so it is possible that it is not compatible. Let me check, also with other people and I will update the source directly. It is better than working with patches. Coming soon.

2 ) There is also a small error in the devel version of yambo :

We will have a look at it.

3 ) For taking in bse calculation the coupling term in 3.4.1 into account, should i include both BSEmod = coupling AND WehCpl or is it enough to have one of them ?

For a complete coupling you should put both of them. BSEmod = coupling activate the coupling part of the matrix, WehCpl include the direct term also in the coupling.

For HF_and_locXC i am just wondering where the correlation part of self energy of gw approximation is going to be calculated. HF_and_locXC : this is only exact exchange self energy - exchange-correlation potential of DFT, so where is correlation self energy ?

For a GW calculation, generate an input file as:
yambo -x -p p -g n for having the correlation part calculated in plasmon pole approximation.
yambo -x -d -g n for a direct integration in the real frequency axis (needed for the lifetimes, hard to converge with respect the integration).

I tried to run a bse calculation on LiF using this time em1d (dynamic dielec. matrix) but it seems that yambo only works with the static one (em1s), am i right ???

Yes, only static screening in the BSE. Dynamical BSE it is not in the GPL version.

imaginary part of poles of 4-point Green's function gives exciton life time.

Not implemented. The BSE equation it is recast in an eigenvalue problem: you do not have access to the lifetime, and the imaginary part of the excitation it is taken into account with an artificial broadening. Broadening effect due to electron-phonon interaction (temperature) can be included, see this paper (Phys. Rev. Lett. 101, 106405 (2008).) and this tutorial.

What are the keywords to do a sc gw and sc cohsex calculation for both energies AND wave functions in 3.4.1 devel version ?

Sorry for not having been clear. The devel version it is still GPL and self consistency it is not included. It is in the non-GPL version of the code. We will release it as GPL, but still work and testing is needed, so I cannot say anything when it will be done, hope very soon.

I saw in some input files that in the static cohsex approximation one uses plasmon pol model !!! Why is this the case, so if cohsex is static, so there is no frequency integration any more.

As you say, cohsex is static, so you need the static dielectric matrix (em1s). If you calculated the plasmon pole previously, it contains the 0 frequency screening so you can read it from the PP instead of recalculating the em1s.

Hope it helps,
Daniele

[vitoversace]

Dear Daniele,

Why is in the tutorial the acell for H2 molecule in the cube so large ( acell = 25 Bohr) ?
Should it not be much lower in order to reach the minimum of the total energy as a function of acell or is the value just a random number without any further meaning?

Bests