Dear Oscar,
I was trying to use your shock simulation script i.e flyer plate, but it suffers from so called “flying ice cube” problem. I have confirmed it by computing temperature from “ke/atom” and “ave/spatial”, but it significantly differs from temp computed by “temp/com”, and this difference increases drastically as soon as the shock front have covered up a considerable portion (50%) of the system, i.e a translatory motion along shock direction drifts the whole simulation domain. As the matter of fact like temperature, most of the variables would be severely affected. In this context, I have seen recently the “fix recenter” and “fix momentum”, and have studied the documentation. Would it be a solution if I use “fix momentum” for the group “bulk” by setting the total momentum
of “bulk” to zero, or is it reasonable to use “fix recenter” as this fixes does not affect the dynamics ?
Secondly can I define more groups within the group “bulk” and after that can I set bulk type 1 ? For post-processing purpose more groups have to be defined for monitor the shock more accurately.
I have no idea what you're talking about .... (NOTE: It doesn't mean I
don't want to Help you.)
I have never tried to calculate the temperature rise due to the
passing of the shock wave, because I still haven't figure it out how
to do it Also your message and previous messages just contains a lot
of misconceptions and it reveals you don't probably understand well
the Fundamentals concepts. It would required a lot of my effort and
patient from my part to just give some enlighten . In a Nutshell what
I try to say is: I Wish I could help, but I have no idea how to do
it. Please be a sure that once i figure it out I will Tell you, but
also understand that i'm a very Lazy person and a slow learned....
PS: More probably another person from the LAMMPS community will be
able to share it expertise
Oscar’s setup does not create the ‘flying ice cube’. Examine the velocity distribution of the individual atoms, particularly in the directions orthogonal to the shock.
The ‘flyer plate’ method of generating shocks is a NON EQUILIBRIUM method. That is to say that the state variables (P,T,rho) at one part of the cell will be different to those in another part of the cell. Therefore Axel’s comment earlier about global quantities being meaningless holds in this case.
Secondly, if one starts with atoms with a CoM velocity of zero, then a shock wave passes, the CoM velocity must necessarily change. This applies to a small region as much as a large one. It is important to understand that with the flyer plate, the user is CONTINUALLY adding kinetic energy to the system. Moreover, that energy is applied in one dimension only. Eventually the whole system will be travelling with the velocity of the flyer plate. Indeed, if the simulation is allowed to continue, and the atoms are allowed to expand in the shock direction, the atoms at the end furthest from the flyer plate will travel at twice the flyer velocity, in the shock direction.
Dear Nigel,
I have done several experiments regarding this shock simulation by using “flyer plate” method, and with my realization, I am completely agree with you that during the simulation the “piston” is continuously giving kinetic energy to the sample along the shock direction. As the matter of fact, the sample getting compacted and it is obvious that the COM will be shifted and having some non-zero velocity along shock direction. And as the natural behavior of NEMD there will be surely some spatial gradient of the thermodynamic state variable i.e T, P and rho.
In some experiment, I have seen that if I use fix recenter to group all with INIT INIT INIT, it does not disturb the dynamics but at all steps as the com is recented so the continual drift of my total simulation box is now constrained without affecting the TD variables ( have attached the temperature variation with time with and without fix recenter).
In the two curves (please see the attachment) temperature is continuously increases, and as per I know it is pretty obvious there is no confusion, but the confusion is that when I tried to compute temperature with E = 3/2 KT i.e using ke/atom and compute the temperature with temp/com to be confirm how far the com velocity affects the temperature, and it’s clearly showing in the curve that they differs significantly, i.e. why I thought that this is “flying ice cube” problem.
(Although it is not reasonable to take the whole simulation box as a group in this NEMD simulation and compute the average temperature in each timestep because of non-equilibrium states of the surface atoms, yet I have taken the whole group "bulk " and averaged it over, just to see the nature of the curves )
Dear Nigel,
I have done several experiments regarding this shock simulation by using “flyer plate” method, and with my realization, I am completely agree with you that during the simulation the “piston” is continuously giving kinetic energy to the sample along the shock direction. As the matter of fact, the sample getting compacted and it is obvious that the COM will be shifted and having some non-zero velocity along shock
Shock physics tell us that this “some non-zero velocity” will be exactly the same as the impact velocity (or the velocity of the flyer plate) after reaching steady-state. Different regions (pre-shocked, shocked, steady) have different CoM velocities.
direction. And as the natural behavior of NEMD there will be surely some spatial gradient of the thermodynamic state variable i.e T, P and rho.
In some experiment, I have seen that if I use fix recenter to group all with INIT INIT INIT, it does not disturb the dynamics but at all steps as the com is recented so the continual drift of my total simulation box is now constrained without affecting the TD variables ( have attached the temperature variation with time with and without fix recenter).
Fix recenter, as far as my understanding goes, constrains the CoM position of a group, but does not remove its CoM velocity, hence the kinetic energy given by E=0.5mv^2 still contains CoM velocity. This is may be one of the reason that your red curves are higher in T than the black ones. When computing T with E=3/2KT = 1/2mv^2, the CoM velocity should be removed.
In the two curves (please see the attachment) temperature is continuously increases, and as per I know it is pretty obvious there is no confusion, but the confusion is that when I tried to compute temperature with E = 3/2 KT i.e using ke/atom and compute the temperature with temp/com to be confirm how far the com velocity affects the temperature, and it’s clearly showing in the curve that they differs significantly, i.e. why
Does not this tell you exactly “how far the CoM velocity affect the T”? One with CoM and one without? Although please note that estimating a thermodynamic state function for a non-equilibrium system is not a good idea, and the estimation can be meaningless and bogus.
I thought that this is “flying ice cube” problem.
You are mis-understanding and mis-using this term.
I have never used this fix; there is no reason why drift of the simulation cell should be a problem (think about the practical experiment you are simulating).
Have you compared 3/2KT against 1/2mv^2 for a small region, as the shock passes?
Secondly, if one starts with atoms with a CoM velocity of zero, then a shock
wave passes, the CoM velocity must necessarily change. This applies to a
small region as much as a large one. It is important to understand that with
the flyer plate, the user is CONTINUALLY adding kinetic energy to the
system. Moreover, that energy is applied in one dimension only. Eventually
the whole system will be travelling with the velocity of the flyer plate.
Indeed, if the simulation is allowed to continue, and the atoms are allowed
to expand in the shock direction, the atoms at the end furthest from the
flyer plate will travel at twice the flyer velocity, in the shock
direction.
Nigel
If the kinetic energy keeps increasing then the temperature should
also increase.
More probably you dont want the piston to RUN forever. You use the
piston to give a little push to the system, then the piston should be
STOP , and then let the simulation run . You could also try a RAMP
compression e.g Let the piston start at rest and then increase it
speed gradually to a determined value in a determined time range ...
If you stop the piston, a release wave will be generated which will propagate forward to cause the shock to decay.
Ramp compression is really interesting, because the material stays closer to the isentrope (so-called ICE method), but because the compression wave steepens up to a shock, the simulation cell needs to be very long.
