Lees Edwards Boundary Conditions and the Choice of a Thermostat

Dear LAMMPS users,

I am working on a project, where we are trying to simulate 2D test case with Lees-Edwards Boundary Conditions using LAMMPS with overall constant temperature of 20 Lenard-Jones units.
From the results, we expect to obtain a linear shear profile (please, see attachment - the black line shows the analytical solution of the problem while the red line represents the velocity profile
obtained from the LAMMPS simulation). However, this is not the case and we suspect that our LAMMPS input script does not describe correctly the system, we are trying to simulate:

# 2d Lennard-Jones melt
dimension ��� ��� ���� 2
units��� �� �� ��� ��� ��� � lj
atom_style��� �������� atomic
timestep��� ��� ��� ��� � 0.0001
variable��� ��� ��� ��� �� output_folder universe /local/disk/LEbc
boundary��� ��� ��� ��� pp pp pp

# Initial data structure
read_data�� ��� ��� �� initStruct_1000.input
mass��� ��� ��� ��� ��� �� 1 1.0
velocity�� �� ��� ��� ��� all create 20 87287 loop geom

pair_style�� �� ��� ��� lj/cut 2.5
pair_coeff�� �� ��� �� 1 1 1.0 1.0 2.5

neigh_modify��� � delay 0 every 1 check no

# Lees-Edwards Boundary Conditions
fix��� ��� ��� ��� ��� ��� � 1 all nvt/sllod temp 20.0 20.0 20.0���
fix��� ���� ��� ��� ��� ���� 2 all deform 1 xy erate 0.01 remap v

thermo������� ��� ��� 1000
thermo_style ��� custom step dt temp pe ke

dump��� �� ��� ��� ��� 2 all custom 1000 ${output_folder}/dump_LEbc_termostat_t20_out1000.relax id type xs ys zs vx vy vz
run��� � ��� ��� � ��� ��� 5000000

We are not confident in our implementation of the Lees-Edwards BC, even though all the suggestions on the LAMMPS mailing list were to use a combination

of the commands fix nvt/sllod and fix deform.
We assume that our choice of the thermostat might be wrong, which leads also to wrong results.
Is it possible to give us some advice how to solve this issue?

We are using a LAMMPS version from 16 Mar 2018.

I am looking forward to your answer. Thank you in advance.

Sincerely,
Aleksandra Pachalieva

PhD Student at the TUM

LEbc_vel_profile.JPG

Fix deform and nvt/sllod are all you should need for what is
efffectively LE BC. You might try shrinking the thermostat
damping constant. 20 tau is a long time constant. Also
varying the system size.

Steve