Input File Description

Program: all_currents.x / QEHeat / Quantum ESPRESSO (version: 7.3.1)

TABLE OF CONTENTS

INTRODUCTION

&ENERGY_CURRENT

delta_t | file_output | trajdir | vel_input_units | eta | n_max | first_step | last_step | step_mul | step_rem | ethr_small_step | ethr_big_step | restart | subtract_cm_vel | add_i_current_b | save_dvpsi | re_init_wfc_1 | re_init_wfc_2 | re_init_wfc_3 | three_point_derivative | n_repeat_every_step | n_workers | worker_id | continue_not_converged

INTRODUCTION

Program to compute energy current given the atomic configuration and the velocities of the atoms.

Note that a very small conv_thr must be given in the ELECTRONS namelist, in the order of 1.D-11.
The numerical derivative is very sensitive to this parameter and to delta_t. Careful convergence
tests are needed. Note that if too relaxed values are chosen, the result can depend on the algorithm
used to diagonalize the hamiltonian a lot (the 4th/3rd digit can be wrong). Options that allows
estimating the variance are provided, to reinitialize the wavefunctions and repeat each step many
times ( n_repeat_every_step re_init_wfc_1 re_init_wfc_2 re_init_wfc_3 ).
Performance of the calculation can be tuned a little bit with the parameters ethr_small_step
and ethr_big_step, that can avoid the waste of some iterations in the diagonalization of the
hamiltonian in the first scf step of every scf calculation (the program does 2 scf for each step).
Note that in order to read atomic velocities, in the namelist CONTROL you must set calculation='md',
and in the namelist IONS you must set ion_velocities='from_input'. Algorithm for computing finite
difference derivatives can be set with the option three_point_derivative .

This program implements

Marcolongo, A., Umari, P. & Baroni, S.
Microscopic theory and quantum simulation of atomic heat transport.
Nature Phys 12, 80-84 (2016). https://doi.org/10.1038/nphys3509

and was originally written by Aris Marcolongo in 2014 at SISSA for his PhD thesis
( https://iris.sissa.it/handle/20.500.11767/3897 )
The all_current driver program was rewritten from scratch by Riccardo Bertossa at SISSA in 2020.
Other contributions are from Davide Tisi (SISSA), Loris Ercole (SISSA - EPFL ) and Federico Grasselli (SISSA).
Details of the implementation are discussed in
Marcolongo, Bertossa, Tisi, Baroni, https://arxiv.org/abs/2104.06383 (2021)

All the namilist but ENERGY_CURRENT are the same as the program pw.x

Structure of the input data:
===============================================================================

&ENERGY_CURRENT
  ...
/

&CONTROL
  MUST SET calculation='md'
  ...
/

&SYSTEM
  ...
/

&ELECTRONS
  you may want startingwfc = 'random' (for better standard deviation estimation)
  ...
/

[ &IONS
  MUST SET ion_velocities='from_input'
  ...
 / ]

[ &CELL
  ...
 / ]

ATOMIC_SPECIES
 X  Mass_X  PseudoPot_X
 Y  Mass_Y  PseudoPot_Y
 Z  Mass_Z  PseudoPot_Z

ATOMIC_POSITIONS { alat | bohr | crystal | angstrom | crystal_sg }
  X 0.0  0.0  0.0  {if_pos(1) if_pos(2) if_pos(3)}
  Y 0.5  0.0  0.0
  Z O.0  0.2  0.2

ATOMIC_VELOCITIES
  X 0.0  0.0  0.0
  Y 0.5  0.0  0.0
  Z O.0  0.2  0.2

K_POINTS { gamma }
if (gamma)
   nothing to read

[ CELL_PARAMETERS { alat | bohr | angstrom }
   v1(1) v1(2) v1(3)
   v2(1) v2(2) v2(3)
   v3(1) v3(2) v3(3) ]
   

Namelist: &ENERGY_CURRENT

delta_t REAL
Default: 1.D0
Small timestep used to do the numerical derivative needed
in order to compute some parts of the current. Note that is in the pw.x units.
         
file_output CHARACTER
Default: 'current_hz'
The program will write the output in file_output and file_output  + '.dat'.
In the latter file the format of the output is:

   NSTEP t_ps J_x J_y J_z Jele_x Jele_y Jele_z v_cm(1)_x v_cm(1)_y v_cm(1)_z ...

where J_x, J_y, J_z are the three components of the DFT energy current,
and can be easily post-processed by other external programs.
Jele_* are the components of the electronic density current that may be used
for decorrelation and better data analysis or for calculating the electric current.
v_cm(1) ... v_cm(nsp) are the center of mass velocities for each atomic species.

If n_repeat_every_step > 1, an additional file file_output + '.stat' is
written with the following format:

   NSTEP t_ps mean(J_x) mean(J_y) mean(J_z) std(J_x) std(J_y) std(J_z)

only one line per step is printed in this case (in the other output files you will
find every calculation, also repeated ones). std is the standard deviation.
         
trajdir CHARACTER
Default: ''
Prefix of the cp.x trajectory. The program will try to open the files
trajdir .pos and trajdir .vel
The files, for n atoms, are formatted like this:

   NSTEP1 t_ps1
   x(1) y(1) z(2)
   .    .    .
   .    .    .
   .    .    .
   x(n) y(n) z(n)
   NSTEP2 t_ps2
   x(1) y(1) z(2)
   .    .    .
   .    .    .
   .    .    .
   x(n) y(n) z(n)
   ...

the order of the atomic types must be the same of the one provided in the input file.
If the files are not found, only the positions and the velocities from the input file will be used.
Note that the units are specified by the input file. The units of the velocities are the same of
the positions with time in atomic units. If a cp.x trajectory is provided (see vel_input_units )
a factor 2 can be used for the velocities.
         
vel_input_units CHARACTER
Default: 'PW'
This multiplies or not by a factor 2 the velocities given in the input.
                      Available options are:
            
'CP' :
assume velocities are given in cp.x time units (thus multiplying by 2 the velocities)
            
'PW' :
assume velocities are given in pw.x time units
            
eta REAL
Default: 1.D0
 Convergence parameter for Ewald-like sums
         
n_max INTEGER
Default: 5
 Number of images in each direction used to converge some sums.
         
first_step INTEGER
Default: 0
The program will start with step  istep >= first_step.
If greater than zero the input file's positions and velocities will be ignored.
Note that this is not a sequential index but refers to the indexes reported in
the input trajectory file. The index of 0 is assigned to the snapshot described
in the input namelist file.
         
last_step INTEGER
Default: 0
The program will end with step  istep <= last_step.
If 0, it will stop at the end of the trajectory file
Note that this is not a sequential index but refers to the indexes reported in
the input trajectory file.
         
step_mul INTEGER
Default: 1
The program will use the step only if
MOD(step, step_mul) == step_rem.
         
step_rem INTEGER
Default: 0
The program will use the step only if
MOD(step, step_mul) == step_rem.
         
ethr_small_step REAL
Default: 1.D-7
Diagonalization threshold after the small delta_t numerical derivative step.
(the system changed a very little)
         
ethr_big_step REAL
Default: 1.D-3
Diagonalization threshold at the beginning of each step but the first,
for wich the pw.x input value is used.
         
restart LOGICAL
Default: .false.
If true try to read file_output .dat and try to set first_step to the
last step in the file + 1
         
subtract_cm_vel LOGICAL
Default: .false.
If true subtract from the velocities of all atoms for every step
the center of mass velocity for each atomic type.
It help to decorrelate a little the mass flux from the energy flux
         
add_i_current_b LOGICAL
Default: .false.
If true adds to the energy current a part that is correctly implemented only for cubic cells.
This part is in the form of a sum over the atomic types of a constant time the center of mass velocity
of the atomic type. It does not change the value of the thermal conductivity when the formula for the
multicomponent case with the inverse of the Schur complement is used, and in the single component
or solid case this is a non-diffusive contribution.
         
save_dvpsi LOGICAL
Default: .false.
If true allocate the space needed for saving the solution of the linear system betweew every calculation.
The iterative algorithm will always start from there. By default it starts always from scratch.
         
re_init_wfc_1 LOGICAL
Default: .false.
If true initializes, as specified in the ELECTRON namelist of the PW section, the wavefunctions
before the first ground state calculation, then compute the charge density.
 Otherwise use the last calculated wavefunctions.
         
re_init_wfc_2 LOGICAL
Default: .false.
If true initializes, as specified in the ELECTRON namelist of the PW section, the wavefunctions
before the second ground state calculation, then compute the charge density.
Otherwise use the last calculated wavefunctions.
Note that if three_point_derivative is false, this has no effect.
         
re_init_wfc_3 LOGICAL
Default: .false.
If true initializes, as specified in the ELECTRON namelist of the PW section, the wavefunctions
before the third ground state calculation, then compute the charge density.
Otherwise use the last calculated wavefunctions.
         
three_point_derivative LOGICAL
Default: .true.
If true calculates three ground stated: one at t - delta_t /2, one at t and one at t + delta_t/2.
Obviously it needs more computer time, but the derivative should be better.
         
n_repeat_every_step INTEGER
Default: 1
Number of repetition of the full current calculation for each step. If > 1, the file file_output + '.stat'
is written with some statistics. Note that if you don't specify at least re_init_wfc_1 ,this may be useless.
You may want to specify startingwfc = 'random' in the ELECTRONS namelist.
         
n_workers INTEGER
Default: 0
The calculation over all the trajectory is splitted in n_workers chunks. Then to run the code over all
the trajectory you must run n_workers input files each one with a different worker_id,
from 0 to n_workers - 1 . Those inputs can run at the same time in the same folder. The worker_id
will be appended to the outdir folder and to the file_output input variables, so you can safely run all
the inputs in the same directory at the same time.
         
worker_id INTEGER
Default: 0
See n_workers variable
         
continue_not_converged LOGICAL
Default: .false.
If it is not possible to find a ground state for a given frame of the trajectory, go to the next one.
You will not find this step in the output file(s).
         
This file has been created by helpdoc utility on Tue Mar 19 10:18:32 CET 2024.