A-TDEP ascii input files

Dear all,

I am using abinit v 9.6.2. I would like to use the tdep executable to obtain the IFC from a dynamic simulation. As an input, I would like to provide three separated ASCII files fcart.dat, xred.dat and
etot.dat as described here

However, in the same guide, I read

TODO : Explain the extra input variables when the 3 ASCII files are employed.

Please, could you tell me what these variables are and what is the exact format of such ascii files? Is it necessary to provide some header in those files or it is enough to grep out forces, positions and energies and concatenate them?
Thanks a lot in advance!

Best
Antonio

Dear Antonio,

Sorry. The documentation is not up to date… In fact you can use ASCII files in TDEP as follows :

  • build 3 files corresponding to the AIMD trajectory : xred.dat, fcart.dat and etot.dat are the reduced coordinates, cartesian forces and total energies as they are given in the ABINIT output file.
  • add 5 lines in the .abi input file. If you consider the v8/Input/t37.abi input file (corresponding to the reading of a NetCDF trajectory), and read the “echo” at the beginning of the v8/Refs/t37.abo output file, you can infer the 5 lines to add. See below in bold these lines (for the v8/t37 test case) :

NormalMode
#Unit cell definition
brav 7 0
natom_unitcell 5
xred_unitcell 0.000000000e+00 0.000000000e+00 0.000000000e+00 5.000000000e-01 5.000000000e-01 5.000000000e-01 0.000000000e+00 5.000000000e-01 5.000000000e-01 5.000000000e-01 0.000000000e+00 5.000000000e-01 5.000000000e-01 5.000000000e-01 0.000000000e+00
typat_unitcell 3 2 1 1 1
ntypat 3
amu 15.9994000000 101.0700000000 87.6200000000
#Supercell definition
rprimd 14.8529572410 0.0000000000 0.0000000000 0.0000000000 14.8529572410 0.0000000000 0.0000000000 0.0000000000 14.8529572410
multiplicity 2.00 0.00 0.00 0.00 2.00 0.00 0.00 0.00 2.00
natom 40
typat 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1 3 2 1 1 1
temperature 495.05
#Computation details
nstep_max 101
nstep_min 1
Rcut 7.426
#Optional inputs
Ngqpt2 2 2 2
TheEnd

I hope it will work for you. Let me know.
Regards,
François Bottin

Hi François,
thanks a lot for the quick answer. Following your indications the code reads my input. However, it stops quickly with the error

— !ERROR
src_file: m_tdep_utils.F90
src_line: 297
mpi_rank: 0
message: |
The basis of atoms written in input.in file does not appear in the MD trajectory

— !ERROR
src_file: m_tdep_utils.F90
src_line: 298
mpi_rank: 0
message: |
Perhaps, you can adjust the tolerance (tolmotif)

The error message is not clear to me: what does it refer to? I checked the source code and it looks like that it is related to the average position of the atoms but I am not an abinit developer and I don’t know how this part is handled and why this check is necessary. Also, I don’t know how to set tolmotif, I don’t find it in the list of variables here

https://docs.abinit.org/topics/a-TDEP/#related-input-variables

Thanks a lot for your help.
Best
Antonio

Hi Antonio,
There are may reasons to have such error (the input file, the MD trajectory…).

  • It likely come from your input file : are you sure that your motif (xred_unitcell) and the multiplicity reproduce the atom coordinates (xred.dat) in your supercell (rprimd)? I mean that the ideal reduced positions computed using xred_unitcell and multiplicity could be not equal to the average of the real reduced positions included in xred.dat.

  • It could also come from your MD trajectory : is there any diffusion of your atoms ? I mean that some atoms do not oscillate around average positions but move freely inside the supercell. That would say, this system does not have any average positions and you cannot compute phonons.

See if one of these answers solve your problem and let me know if not. In this case I will probably ask you to send me your data to see what’s going on.

Regards,
François

1 Like

Hi François,

I have been doing several test to be sure about the origin of the problem. I am afraid that the atomic positions of the supercell that I have are shifted with respect to those generated by a-tdep by using the multiplicity. Most probably, also the scheme of replica might be different - in my case, the first atoms is replicated along a, then b then c, then second one and so on. How is the supercell created by the a-tdep code? Finally, I don’t have any diffusion.

Thanks for your help.
Best
Antonio

Dear Antonio, dear François,

It seems that I am having the same issue than Antonio. Did you solve it?

Some context: I had an unstable BaTiO3 structure where ions were diffusing away from their initial positions. Thanks to the comment of François, I stabilized it, run a Noseé-Hoover MD simulation again and attempted to run aTDEP based on the MD HIST.nc file of 1) a 5 atoms unit cell and 2) a 40 atoms supercell.

Unfortunately, I am still getting the 2 same errors than Antonio for both the 1) unit cell and 2) supercell case scenarios: ‘The basis of atoms written in input.in file does not appear in the MD trajectory’ and ‘Perhaps, you can adjust the tolerance (tolmotif)’.

I plotted the input original atomic positions (in atdep.abi) as well as the xred_average.xyz output positions for the unit cell to investigate what was going on. Interestingly, all positions were correct modulo the lattice cell parameter. In other words, some atoms ended up outside of the unit cell, at a position which corresponds to their periodic boundary translation in the neighbouring cell. Is this taken into account in the code or could it be the source of a distorted atoms basis, thus leading to the first error message?

I thank you for your consideration,

Virginie
Master student at EPFL/ETH

atdep.abi.txt (778 Bytes)
atdep.abo.txt (4.3 KB)
log_atdep.txt (1002 Bytes)
xred_average.xyz.txt (358 Bytes)

Dear Virginie,

Concerning the first system (only a unitcell), it seems natural that aTDEP does not work. There is no interaction shell so you cannot define any interatomic force constants.

Concerning the second one (the supercell), even it the number of shells will be really low, I don’t understand why aTDEP does not work. Please, send me everything needed to run aTDEP (the abi and nc files). If the trajectory is too long, you can reduce it to a few hudreds MD time steps. Here, we don’t care about the physics, just a number of configurations sufficient to build an average atomic structure.

You can do it here or by email : francois.bottin@cea.fr

Regards,
Francois

Dear Antonio,

Sorry for the delay, I was really busy. Please, send me all your input files (abi and nc files) here or by email (francois.bottin@cea.fr). If the trajectory is too long, you can reduce it to a few hudreds MD time steps. Here, we don’t care about the physics, just a number of configurations sufficient to build an average atomic structure. Thanks.

Regards,
Francois

Dear Francois,

I am not using abinit to generate the trajectory so there is no abi and nc files for my simulation.

Best
Antonio

Dear Antonio,
OK. So, please send me:

  • the 3 data files : xred.dat, fcart.dat and etot.dat (with a few hundreds MD time steps)
  • the TDEP input file

Thanks.
François