PPA suitability

[asanchez]

Hi all,

I've been reading older posts on the subject (such as viewtopic.php?f=14&t=55) but I still have many doubts so I thought I'd start a new topic. Apologies if this is not the proper subforum.

My main question is: How can I know whether the PPA is a suitable approximation for my system?

I've read around that one should look at the EELS and: "If you see peaks with an energy lower then the band width of you metal, then you know that the PPA may fail." (not sure what band width means in this context)

So for starters I have the following questions:

- Would an RPA calculation do or should I go for BSE? (Would excitons be important for this?) Should I set "QpntsRXd" to anything other than "1 | 1 |"?
- I'm studying nanowire systems so the direction of the electric field is going to change things a lot. I suppose I should check that the conditions are satisfied for both parallel and perpendicular fields, right?
- Some of my systems are semiconductors, some are metallic. What should I look for in both cases? peaks below the DFT gap?

Any help would be greatly appreciated. Thanks!

[Daniele Varsano]

Dear Alfonso,
PPA is an approximation, and it is not easy to affirm when it is safe to use it or not. As stated by Andrea in the post you linked a first check to do is look at the eels peaks.
If you find the main peak at low energy, ie, lower than the energy range you are interested in calculating qp corrections, so it can fails. Bandwidths stands for the energy dispersion.
Let me also add, if the eels spectrum has a structured shape, then tae into account that the PPA will fit it with a single pole, so you will be missing effects that could arise from other peaks. Nonetheless, this approximation permits you to perform calculations in a reasonable time and it works in many case.

- Would an RPA calculation do or should I go for BSE? (Would excitons be important for this?) Should I set "QpntsRXd" to anything other than "1 | 1 |"?

Here there is some confusion: PPA is an approximation for the dynamical screening for GW approximation (quasi particle energies). BSE is used to calculate absorption neutral excitations (and usually the screening used is static). When you calculate qp energies you need to integrate the screening in the BZ, so it is calculated for all the q -points. The screening is alway calculated at RPA level. You can look also at finite q eels, as also these components enters in the screening for the GW, anyway in general you should not have a very big dispersion of the plasmon peak wrt the q in the BZ. I would do the check for the q=0 only.

- I'm studying nanowire systems so the direction of the electric field is going to change things a lot. I suppose I should check that the conditions are satisfied for both parallel and perpendicular fields, right?

Well, anysotropy effects can be important, you can choose the direction of the filed when calculating the screening. A good choice could be to indicate the direction along the axis of the wire, or something like |1|1|1|, so a kind of average. The correct way to treat that should be to consider all the dielectric tensor but at the moment it is not implemented.

- Some of my systems are semiconductors, some are metallic. What should I look for in both cases? peaks below the DFT gap?

Usually for semiconductor PPA it works, next consider that for metals you do not have gap, for semiconductor you cannot have peak below the gap unless you include excitonic effects, but the screening is calculated at RPA level. What you expect is that the eels peak is found at high energies.

Best,
Daniele