How to simulate a real fluid flow-not nve, nvt, npv...just under natural conditions

Excuse me ,I would like to simulate a fluid flow under natural conditions , and not by period boundary conditions.In nve esenmble,I add an atom every 10fs at the entrance and delete an atom every 10fs at the export.However,this is not true because the temperature of entrance area jump lower and temperature of export area jump higher with the process of simulation. It is not an isolated system but open system in nature.
There is any commond to help me achieve to update the velocity and coordinates of atoms without affecting the energy?Nve,nvt,npv…these commands cannot satisfy the natural fluid fow.
Thank you for your reply!
Youwei Gu
From Tsinghua University

    Excuse me ,I would like to simulate a fluid flow under natural
conditions , and not by period boundary conditions.In nve esenmble,I add an
atom every 10fs at the entrance and delete an atom every 10fs at the
export.However,this is not true because the temperature of entrance area
jump lower and temperature of export area jump higher with the process of
simulation. It is not an isolated system but open system in nature.

this is an ill-posed question (to put it mildly) which ignores a lot
of the realities of physics.
in nature, atoms just don't appear and disappear (unless under rather
extreme conditions, of course).
if you want the kind of "natural flow" that you are describing, you
have to build two large (enough) reservoirs and then have a force push
atoms from the outlet of the first toward the inlet of the other.
remember that on the atomistic scale things are not always quite the
same as on the macroscopic scale.

however, the computational cost of this is insane, and the sane
approach would be to have periodic boundary conditions and placing a
suitably set up langevin thermostat in between outlet and inlet of
your observation area that dissipates perturbations through the
observation area.

this kind of topic has been discussed a few times over the last months
on the list already.

    There is any commond to help me achieve to update the velocity and
coordinates of atoms without affecting the energy?Nve,nvt,npv...these

please look at your question again. how does this make sense? if you
update the velocities of atoms you *do* affect the energy. if you want
atoms to move without regard for their environment, you can use fix
move, but then you are not doing physics, you are creating a movie
(sic!).

axel.

Thank you !However,when I consider convective heat transfer,there will be aimed temperature at the entrance,and natural flow at the export.Periodic boundary conditions cause heat flow from entrance directly to export.I feel it is not suitable so I want to add or delete atoms to zachieve fluid flow.How could I solve these boundary problems especially at the export?I appreciate to your reply!Thanks a lot!

what you feel is your concern. i already described two approaches that
address your issue. what *else* do you want?

I wasn't on the list for these past discussions, so maybe this is old news but there is something called a "dual-control volume grand canonical MD" method which I guess kind of does what you want in a less incorrect way. I don't know much more about, but I'm quite sure LAMMPS won't do it for you. There's some discussion of that method and some other simpler methods for applying flow like Axel described and which you could do in LAMMPS in this review by Alexiadis and Kassinos, Chem. Rev. 108, 5014 (2008).