Setting proper pararmeter in the input file for GW

[Bramhachari Khamari]

Dear developers,
I am running GW calculation of LiF example using Yambo. Two of the input parameters are as follows,

EXXRLvcs= 3719 RL # [XX] Exchange RL components
VXCRLvcs= 3719 RL # [XC] XCpotential RL components

Could you tell me the meaning of RL in the parameters (EXXRLvcs and VXCRLvcs).

[Daniele Varsano]

Dear Bramhachari,
RL are Reciprocal Lattice vectors.
You can assign these variables in RL or as energy cutoff.
Have a look here for variables meaning:
http://www.yambo-code.org/wiki/index.ph ... s#EXXRLvcs

Best,
Daniele

[andrea.ferretti]

Dear Bramhachari Khamari,

RL stands for reciprocal lattice vectors, so

Code: Select all

EXXRLvcs= 3719 RL # [XX] Exchange RL components
VXCRLvcs= 3719 RL # [XC] XCpotential RL components

means that you are using 3719 G vectors to represent both the non-local exchange operator and the local exchange-correlation
potential.

Another possibility is to give these numbers in energy units (as a kinetic energy cutoff)

Code: Select all

EXXRLvcs= aaa Ry # [XX] Exchange RL components
VXCRLvcs= bbb Ry # [XC] XCpotential RL components

multiple energy units ara available.

In general, yambo picks as default value for these two variables the number of plane waves used to represent the charge density
(corresponding to ecutrho in Quantum ESPRESSO). Personally, I would leave these number unchanged as much as possible,
resorting to lower them with care only when actually needed.

take care
Andrea

[Bramhachari Khamari]

Dear Daniele, Andrea
Thanks a lot for providing me the details.

[Bramhachari Khamari]

Dear Daniele,
I need some clarification for convergence parameter for GW calculation for a material. Can I choose a particular k point and bands around the fermi level to converge the gap with respect to different parameters like BndsRnXp, GbndRnge. While changing these parameters to a higher value, Do I need to do a new DFT calculation for these higher values or simply, I can change in the input file of GW run.

[Daniele Varsano]

Dear Bramhachari,
The best strategy is to do a DFT calculation with an enough large number of bands in order to no recalculate them.
You can fix the k points and tests the other parameters: note that BndsRnXp and NGsBlkXp are not independent.
Note that %QPkrange can be fixed to just a few points (even just one band) to check the convergence.
Finally, in order to test the k point sampling, you need to run several DFT calculations increasing the k points grid.
For a guided tour, I suggest you have a look at this tutorial:
http://www.yambo-code.org/wiki/index.ph ... rial:_h-BN

Best,
Daniele

[Bramhachari Khamari]

Dear Daniele,
Thank you sir for the reply.