[gmx-users] von Mises stress-strain curve

[Anh Vo]

I’m running GROMACS to simulate a simple phospholipid bilayer membrane
(POPC) under deformation, and plot the von Mises stress-strain curve to
determine yielding point.



I have got the membrane deformed and tried to get the stress-strain curve
by extract output pressure from .edr file (option "PRESS X, PRESS Y, PRESS
Z"). The general shape of the curve is correct; however, the pressure value
fluctuates widely in GROMACS and the yield stress/ yield strain is wrong. Why
did I get the wrong value and where/ which parameter files should I address
to solve this problem?


1. I got a lower yield stress and a higher yield strain than it should be
(compared to previous research findings on similar problem using LAMMPS).
Why did I get this problem, and what should I fix to get the right yielding
stress/strain values?


2.  Is the wrong value due to the incorrect bonding or constraint between
the molecules? If this is the case, which files or which parameters should
I look into to modify the bond or constraints? Does lower yield stress and
higher yield strain means that the bond in GROMACS is more ductile?


3.  In the edr file, which options should I get as the output pressure or
stress value? Is it “PRESS-XX, PRESS-YY, PRESS-ZZ” or  “VIR-XX, VIR-YY,
VIR-ZZ” options? Or what else should I get to calculate stress values? Can
I directly use that output pressure as stress values in X, Y, Z dimension,
or I need to perform further calculation on the pressure I get?



 4.   Sometimes I run the same simulation several times without changing
anything to double check. I can get similar results most of the time, but
sometimes the simulation stop faster and an error is reported like this
“The domain decomposition grid has shifted too much in the Z-direction
around cell 0 0 3. This should not have happened. Running with fewer ranks
might avoid this issue”.

What does this error mean? What is domain decomposition? Is the Domain
Decomposition automatically set in GROMACS for running simulation?


5. How did you choose your time step? What is the logic for deciding the
time and time steps to run an MD simulation? I’ve read that larger time
step permits better sampling without significant errors in observed
thermodynamic, structural or dynamic properties

   (e.g. https://www.sciencedirect.com/science/article/pii/001046558690113X;
 https://udel.edu/~arthij/MD.pdf;  etc.)

 ==> What does “sampling” mean? Why does the larger time step allow better
sampling?


 Just some extra questions about some terms that I’m not very clear when
reading the materials. I have looked them up and read more but I’m still
confused. Please help me to clarify them.



     - What is improper vs. proper dihedral?



     - It is said that “If a dynamical system is ergodic, the ensemble
average becomes equal to the time average”.

       What does it means if a system is ergodic? Why does ensemble
sampling becomes similar to time sampling if a system is ergodic?


Thank you very much for sharing your time.


Sincerely,

Anh Vo