Dear developers,
I worked on the Kremer-Grest model, and I imposed the shear by “fix nvt/sllod” with “fix deform”:
fix 1 all nvt/sllod temp 1.0 1.0 1.0 tchain 1
fix 2 all deform 1 xy erate ${Rate} units box remap v
For some reasones, I want to calculate the velocity profile along the gradient direction.
At first, my simulation box was 727248 (LJ unit), and I found the velocity of center mass of simulation box is nonzero which is closed to V/2 (V=“shear rate”“gradient size”). Thus, I thought the velocity profile from bottom to top is 0 to V, which is closed to the experiement.
However, when I changed the box size into 14472*48 (double the flow direction size), I found the velocity of center mass of simulation box is 0. This means the velocity profile from bottom to top is -V/2 to V/2.
Lammps user manual just states: “the points at bottom of the box (low) have a small x velocity, while points at the top box (hi y) have a larger x velocity”. So, I can not find a reasonable answer to my problem.
In summary, my questions are:
(1) what it the velocity profile when I impose the shear by the sllod algorithm?
(2) Is there something default setting which I missed?
If my expression is not clear, please let me know.
Dear developers,
I worked on the Kremer-Grest model, and I imposed the shear by “fix nvt/sllod” with “fix deform”:
fix 1 all nvt/sllod temp 1.0 1.0 1.0 tchain 1
fix 2 all deform 1 xy erate ${Rate} units box remap v
For some reasones, I want to calculate the velocity profile along the gradient direction.
At first, my simulation box was 727248 (LJ unit), and I found the velocity of center mass of simulation box is nonzero which is closed to V/2 (V=“shear rate”“gradient size”). Thus, I thought the velocity profile from bottom to top is 0 to V, which is closed to the experiement.
However, when I changed the box size into 14472*48 (double the flow direction size), I found the velocity of center mass of simulation box is 0. This means the velocity profile from bottom to top is -V/2 to V/2.
instantaneous values will fluctuate a lot, so the first result matching your expectations is likely just a coincidence.
Lammps user manual just states: “the points at bottom of the box (low) have a small x velocity, while points at the top box (hi y) have a larger x velocity”. So, I can not find a reasonable answer to my problem.
you have to look at the explanations in both, the fix nvt/sllod and the fix deform documentation.
In summary, my questions are:
(1) what it the velocity profile when I impose the shear by the sllod algorithm?
the velocity profile should follow the box deformation, fix nvt/sllod and using “remap v” with fix deform should see to it.
you can easily output the profile for yourself. if I add, for example, the lines
compute ybins all chunk/atom bin/1d y lower 0.1 units reduced
fix 3 all ave/chunk 10 1000 10000 ybins vx file yprof.dat
and plot the collected data from the yprof.dat file I get the expected (noisy) profile from the attached image.
(2) Is there something default setting which I missed?
I suspect you are missing the impact of much data fluctuates when when looking at rather tiny systems in MD simulations when compared to (macroscopic) experiments.
Hello Axel,
Thanks for answering my problem again. I apologize that I should express my problem more clearly.
I know the flow velocity is often smaller than the thermal flucutation for polymer system, so I calculated the flow field by adjacent coordinates (not velocity). For example, I stored the coordinates at strain 0 and strain 0.1. Of course, I “ignore” the monomer diffusion in the gradient direction and introduced a correction when the monomer cross the box boundary. This method can be checked easily when the shear rate is large enough (the flow velocity is comparable to the peculiar velocity).
In other word, I get the problem when I used the coordinates to calculate the velocity profiles. I can shift my result to get a velocity field from “-V/2” to “V/2”, but I want to confirm this result is occasionally or not since I used 8 independt sample.
From your attached image, I want to ask another problem: I guess the abscissa is the gradient coordinate of “ybin”, and the ordinate is the velocity of “ybin”. Is it right?
If the answer is yes, the veloctiy profile from bottom to top is -0.05 to 0.1 (approximatively). In other word, the velocity is asymmetric. I think it is counterintuitive.
this is too short a simulation and too noisy data for converged results. the plots are merely usable for a qualitative assessment, not for the kind of quantitative evaluation that you are applying. it also does not reserve time for equilibration etc.
as I already alluded to, my impression is that you are not properly accounting for what the statistical errors in your numbers are.
Hello Axel,
Thanks for answering my problem again. I apologize that I should express my problem more clearly.
I know the flow velocity is often smaller than the thermal flucutation for polymer system, so I calculated the flow field by adjacent coordinates (not velocity). For example, I stored the coordinates at strain 0 and strain 0.1. Of course, I “ignore” the monomer diffusion in the gradient direction and introduced a correction when the monomer cross the box boundary. This method can be checked easily when the shear rate is large enough (the flow velocity is comparable to the peculiar velocity).
In other word, I get the problem when I used the coordinates to calculate the velocity profiles. I can shift my result to get a velocity field from “-V/2” to “V/2”, but I want to confirm this result is occasionally or not since I used 8 independt sample.
From your attached image, I want to ask another problem: I guess the abscissa is the gradient coordinate of “ybin”, and the ordinate is the velocity of “ybin”. Is it right?
If the answer is yes, the veloctiy profile from bottom to top is -0.05 to 0.1 (approximatively). In other word, the velocity is asymmetric. I think it is counterintuitive.
