RIM convergence

[Daniele Varsano]

You can also use a slightly smaller value of the cutoff, but if you are at the edge of the convergence could be not easy to recognise. A safer option would be to consider a larger cell with the cutoff slightly smaller than the box.

In general anyway, you can save time considering a large enough box. How much is large enough it is explained in this thread as indicated in the FAQ.

Best,
Daniele

[Laura Caputo]

Dear Daniele,

I have re-calculated the band gap deleting the previous databases. However, I still have a much higher band gap compared to the calculation without the truncated potential.

In particular, with a RIM calculation, I obtain 2.4eV GW band gap, while for a truncated potential I obtain 3.12eV. I have done the same GW calculation for the same system by using another code (VASP) and from that I obtain 2.5eV, which is very similar to the calculation without the usage of the truncated potential.

To what may this be due?

[Daniele Varsano]

Dear Laura,

I have re-calculated the band gap deleting the previous databases. However, I still have a much higher band gap compared to the calculation without the truncated potential.

This is expected, it is indeed a vacuum effect.

In particular, with a RIM calculation, I obtain 2.4eV GW band gap, while for a truncated potential I obtain 3.12eV. I have done the same GW calculation for the same system by using another code (VASP) and from that I obtain 2.5eV, which is very similar to the calculation without the usage of the truncated potential.

As far as I know VASP does not apply truncated potentials.
If you want to make a test you can try to perform calculations with larger cells without using the truncated potentials and see the effect.
Best,
Daniele

[Laura Caputo]

Dear Daniele,

Thank you once again for your quick response.
So, as I understand, it's normal to have such higher band gap compared to the 'non-truncated' calculation. However, it should be better to take the value of the calculation with the truncated coulomb potential since the underestimation is not correct, right?

For the test, I could do a calculation with higher vacuum spacing and see if the gap is enhanced like when we apply the truncated potential?

One last tip, if I may: I am now doing the convergence for the GW calculation. Is it correct if I just do all the calculations with the truncated coulomb potential for BndsRnXp, GbndRnge and NGsBlkXp and see the convergence from there?

Sincerely,
Laura

[Daniele Varsano]

Dear Laura,

However, it should be better to take the value of the calculation with the truncated coulomb potential since the underestimation is not correct, right?

Yes, alternatively you can try to converge wrt the vacuum, but this is quite expensive.

For the test, I could do a calculation with higher vacuum spacing and see if the gap is enhanced like when we apply the truncated potential?

Yes, note that the larger the cell, the larger is the number of gvectors (so it is better to assign the exchange and block matrix for the screening in energy and not in RL). Note also the number of bands needed to converge increase with the cell size.

One last tip, if I may: I am now doing the convergence for the GW calculation. Is it correct if I just do all the calculations with the truncated coulomb potential for BndsRnXp, GbndRnge and NGsBlkXp and see the convergence from there?

Yes, these should be anyway quite independent on the use to the truncated potential.

Best,
Daniele

[Laura Caputo]

Dear Daniele,

I use this post to ask you a quick question always concerning 2D systems.

Regarding the LongDwXp, I read that one should choose this tag depending on the axis on which the system is lying. In my case, I have a non-periodicity along z, so is it correct to select 1.00 | 1.00 | 0.00 ?

Moreover, I have read that it should not change much for GW gap but could be important in absorption spectra calculation, right?

Sincerely,
Laura

[Daniele Varsano]

Dear Laura,

yes, it is correct with the assumption that the (1,1,0) is somehow equivalent to an average of the 2 direction.
If the system present large anisotropy you can study the difference in absorption along the (100) and (010) direction.
In order to take the anisotropy into account in a more precise way you can have a look at step 7 of this tutorial:
http://www.yambo-code.org/wiki/index.ph ... rial:_h-BN

but usually it is not needed.
Best,
Daniele

[Laura Caputo]

Dear Daniele,

Thank you. I read from the tutorial that it's present in Yambo 4.6, has this been implemented in Yambo 5?

Moreover, is the convergence of the bands and the response block size strongly affected by the direction of the electric field?

Sincerely,
Laura Caputo

[Daniele Varsano]

Dear Laura,

yes, it should be for yambo > 4.6
anyway, we compared the anisotropy treatment wrt (1,1,0) for an highly anisotropy material and the difference was negligible.
No, the convergence is not affected as you are modifying there only the G=G'=0 component of the dielectric matrix..
Best,
Daniele

[Laura Caputo]

Dear Daniele,

I have a last question about 2D systems within the truncated potential approach.

I am doing GW calculations with a truncated Coulomb potential on a graphene-like sheet in 0.25 on the z axis. For a 18.8973a.u. cell along z, I am using 18.7973 a.u. box. I have converged the parameters on that system and calculated the GW gap which was 3.12eV.

I understand that the position of the sheet should not matter since the Coulomb potential is a function of |r-r'|. To try, I have done a calculation on the same system but in 0.0 on the z axis, using the parameters converged with the system on 0.25. However, the final result changes (3.06eV instead of 3.12eV).
So, does it matter where it is placed inside the box along the non-periodic axis (0.0 or 0.25 for example) if I select the truncated potential? If yes, I have done all calculations considering a 0.25 coordinate along the z axis. Should I re-do all the convergence tests? If no, should I always select a z box slightly smaller than the actual box size?
In attachment the report file for the system in 0.0 and in 0.25.

Moreover, in the Abinit calculation for the WFK file, I have considered the k point sampling that I used for the optimization after a convergence study on the final energy and I am calculating the gap on Gamma. Is it good this way or should I do different Abinit runs with different k-points meshes to check the GW gap? The last option could take a lot of time.

Thanks,
Laura