My Community

Hello,

I am using VASP 6.5 to calculate the transport properties using constant relaxation time closely following the wiki. I have some questions regarding the "N" steps at the end of the OUTCAR file:

1. What do the N steps supposed to represent?
2. The transport properties calculated for each N step are different. Is this normal? Should the transport properties be converging as N values increase?
2.1. If so, what the best practice to achieve that?
2.2. If not, which N value should I take the properties from? The last step?
3. There seem to be 13 N steps only. Is this the case for my calculations only or are the steps number fixed for all calculations? Is there a parameter that allows changing the number of N steps?

I have checked the wiki for explanation. But, I could not find information regarding it.
Thank you

Hello,

Thanks for reaching out. Since the electron-phonon code is still extremely new, it is possible that some parts of the wiki are not as well documented as we would like. We are constantly trying to improve our wiki documentation based on user feedback.

1. What do the N steps supposed to represent?

The N that you refer to is counting so-called "accumulators". Each accumulator represents one electron-phonon calculation. The different accumulators have different computational parameters, such as chemical potential or which scattering approximation is used. So for example, if you specify

ELPH_SCATTERING_APPROX = CRTA SERTA
ELPH_SELFEN_CARRIER_DEN = -1e17 -1e16 -1e15

that means you run two different scattering approximations and three different carrier densities. You will end up with 6 accumulators (so 6 electron-phonon calculations), one for each combination of computational parameters, so N goes from 1 to 6. For each N, the corresponding parameters are reported in a list at the end of the OUTCAR file. The first entry may look like this:

Transport calculator N =  1
----------------------------------
 transport driver: 2 ! Gauss-Legendre integration
 Scattering approximation: constant relaxation-time approximation (CRTA)
 Static self-energy: F
 Transport number of points:   501
 temperature:  100.000 K
   Transport energy range:   [   9.699:   9.907]  wich corresponds to    0.208 eV
   Average relaxation time:    1.0000E-14 s
   Number of electrons:    0.0000E+00
   Number of holes:        1.1296E-06

Transport calculator N =  2
----------------------------------
 transport driver: 2 ! Gauss-Legendre integration
 Scattering approximation: self-energy relaxation-time approximation (SERTA)
 Static self-energy: F
 Transport number of points:   501
 temperature:  100.000 K
   Transport energy range:   [   9.699:   9.907]  wich corresponds to    0.208 eV
   Average relaxation time:    2.1451E-13 s
   Number of electrons:    0.0000E+00
   Number of holes:        1.1296E-06

Here, you see that N=1 corresponds to the CRTA and N=2 corresponds to SERTA. Below that table is a list of output for the transport calculation for each N.

The reason we run multiple electron-phonon calculations in a single VASP run is efficiency. This way, you only need to calculate the expensive part (electron-phonon matrix elements) once. Accumulating all the matrix elements to obtain a physical observable is usually comparatively cheap, so we try to do this multiple times in a single VASP run.

2. The transport properties calculated for each N step are different. Is this normal? Should the transport properties be converging as N values increase?

This is normal as explained above. N is not a convergence variable but simply a means to distinguish between the different electron-phonon calculations performed in the same VASP run.

2.2. If not, which N value should I take the properties from? The last step?

You should look at the table at the end of the OUTCAR that lists the parameters for the different N and pick the one that suits your case. You can also do this using the binary vaspout.h5 HDF output file, where the different accumulators are organized as different HDF5 groups.

3. There seem to be 13 N steps only. Is this the case for my calculations only or are the steps number fixed for all calculations? Is there a parameter that allows changing the number of N steps?

This is specific to your calculation. The total number of accumulators depends on the number of independent electron-phonon calculations performed in the same VASP run, and therefore on the INCAR parameters. If you send me your INCAR or OUTCAR file, I will be able to tell you exactly how the number 13 came to be.

I hope that I was able to answer your questions regarding the electron-phonon output. I will work on a wiki page that tries to explain these things clearly and in more detail.

Thank you for your reply Mr. Manuel. This was very helpful. I appreciate your support. I have attached both INCAR and OUTCAR.

I'm happy to help. After looking at your INCAR file, I can confirm that what you see is normal and expected behavior. When setting ELPH_MODE = transport, the default value of ELPH_SELFEN_CARRIER_DEN is set to 13 different values as documented on the ELPH_MODE wiki page. Each of the N accumulators corresponds to one of these numbers in the same order as they appear in the ELPH_SELFEN_CARRIER_DEN tag. We might consider adding additional information to the OUTCAR to make this output easier to read in the future.