Molecular Dynamics Simulation Laboratory

Part I: Distributions and Equation of State.

Contents

Introduction

A Molecular Dynamics (MD) simulation is a computer simulation of Newton's Laws for a collection of particles. For a brief introduction to molecular dynamics please see: An introduction to molecular dynamics in Matlab. In part I of this lab, we will focus on systems with fixed Number, Volume (area), and Energy $NVE$. We say these systems are in the microcanonical ensemble. The matlab code we will uses is a script mdNVE000.m, which requires one helper file plotNCirc.m to function.

Matlab has two main ways to store programs---scripts and functions. A script is just a list of commands stored in a file that are run sequentially. A function takes input variables and returns variables, but anything inside is hidden. We will use a script so that we have access to all of the internal variables. You can freely edit mdNVE000.m, but it is setup so that it is not necessary. The code has comments on most lines. Comments are any text after a % sign. Comments can not be nested.

Exercise I

Running the code:

tic; override={'TT',10};mdNVE000; toc; clear('override');

..........Done.
Elapsed time is 0.591410 seconds.

Exercise II

Changing the default behavior:

In the command above notice the use of the cell-list variable override. A cell-list is a variable container in matlab. Cells can hold any objects including other cells. In this case override holds option pairs.

TT=500;           % total simulation time

TT is the total simulation time. The simulation works by moving forward small steps at a time. For example, if a particle is a x=0 at the beginning of the simulation with a velocity of vx=.3 then a short time later it will move to a new location x_new=x_old+vx*dt. In this simulation the small time dt=.01 so the new x after one time step would be x=x+.3*.01=.003. This is a simplification if you are interested in the details look at An introduction to molecular dynamics in Matlab. Then TT=Nt*dt, where Nt is the total number of time steps. # To run the code with a different TT use override={'TT',20}. Notice the curly brackets and the single quotes for variable names. Only single quotes on the variable that changes so:

for nv=1:3; |override={'TT',200*nv}|;mdNVE000;end

will run three simulation with TT=200 then TT=400 then TT=600. Try running the simulation again with this command:

tic; override={'TT',20};mdNVE000; toc; clear('override');

This command tells the code to set TT=20. The override lasts as long a it is defined. In this case, we clear it at the end of the command, but if we wanted to do many runs with TT=20, we could leave it.

tic; override={'TT',20};mdNVE000; toc; clear('override');

..........Done.
Elapsed time is 0.605435 seconds.

Exercise III

Conservation of Energy and More overrides:

You can override more than one variable: override={'TT',1000,'KE0',10}. Here it would increase the simulation time form the default 500 and change the initial total Kinetic Energy KE0 to 10. You can also add on to a previously defined override: override=[override,{'phi_set',.8}] Also change the initial density to 80% coverage. Notice the square bracket to concatenate. The order of the pairs does not matter, but they have to be pairs.

The simulation defines and saves a lot of information about the system. From the code here is a sample:

%% Save Variables
Ek=zeros(Nt,1);     % kinetic energy of particles
Ep=zeros(Nt,1);     % particle-particle potential energy
Ps=zeros(Nt,2,2);   % pressure tensor

if(saveall)
  xs=zeros(Nt,N);    % all particle x-positions
  ys=zeros(Nt,N);    % all particle y-positions
  vxs=zeros(Nt,N);   % all particle x-velocities
  vys=zeros(Nt,N);   % all particle y-velocities
end

Plot KE and PE

t=(1:Nt)*dt;                         % make a time variable from dt to TT
plot(t,[Ek Ep Ek+Ep],'linewidth',3)  % concatenate variable to add to plot
set(gca,'fontsize',20);              % make font bigger
xlabel('Time')                       % axis labels
ylabel('Energy')

Exercise IV

Distributions.

Exercise V

Units and area fraction (density)

In the computer there are no meters, so we are free to set the units however we like. For example the disks have a diameter of 2, but 2 "what?". Until we compare to something in the real world it could be $2\mu m$ or $2$ miles. For this simulation there are only 3 units length, mass, and time. One common way to deal with this is to measure everything in non-dimensional units. So we might measure distances in units of the particle diameter or the box size. We might get a time from the typical collision time $\sqrt{M/K}$. To connect to the real world we could use the particle area fraction, the total disk area divided by the box area:

$\phi=\frac{\pi\sum_n D_n^2}{4 L_x L_y}.$

In the code this determined by phi_set.

Exercise VI

Equation of state:

To non-dimensionalize the energy we need to include the pressure, which is conjugate to A=Lx*Ly or phi_set. Pressure is a measure of momentum flux so there are two ways momentum moves around. Particles carry momentum as they move and they exchange momentum when they collide.

$PA=\frac{1}{2}\sum_{nm}\vec{F}_{nm}\cdot\vec{d}_{nm}+ \frac{1}{2}\sum_n^{N}M_n \vec{v}^2_n= \frac{1}{2}\sum_{nm}\vec{F}_{nm}\cdot\vec{d}_{nm}+K,$

where $\sum_{nm}$ is a sum only over collision between $n$ and $m$, $\vec{F}_{nm}$ is the force between $n$ and $m$ and $\vec{d}_{nm}$ is the vector between $n$ and $m$. The second term is the total kinetic energy $K$. If there are no collisions the first term is zero and $PA=K=Nk_BT$, the ideal gas law equation of state. Since we are trying to get rid of dimensional quantities, we will continue to use $K$ or Ek from the simulation. Then we can define the non-dimensional compressibility factor:

$\mathcal{X}(\phi)\equiv\frac{PA}{K}-1= \frac{\sum_{nm}\vec{F}_{nm}\cdot\vec{d}_{nm}}{2K}= -\frac{A\mathbf{Tr}(\mathbf{\Sigma}_c)}{2K}.$

$\mathcal{X}(\phi)$ is non-dimensional and depends only on a non-dimensional quantity the area fraction $\phi$. In the simulation we measure the collisional pressure (stress) tensor $\mathbf{\Sigma}_c$ as Ps. Momentum flux is a 2x2 tensor so the pressure tensor Ps has 4 components at each time step. We only need the isotropic pressure or the half of the trace of the tensor. The pressure we calculate in the code is the negative of the usual pressure.

If you need some help here is script to do the measurements.

if(~exist('MDLabIdata.mat','file'))  % only run first time
  Nv=40;                  % number of points between .1 and .9 ~10 min
  phlist=(1:Nv)/Nv*.8+.1; % list of 40 numbers between .1 and .9
  Pc=zeros(1,Nv);         % store average Pc for each run
  Ks=zeros(1,Nv);         % " Ek for each run
  Es=zeros(1,Nv);         % " total for each run (check for conservation)
  for nv=1:Nv;            % loop
    override={'phi_set',phlist(nv),'Gb',8/5/sqrt(3)};    % set overrides
    mdNVE000;                                            % run simulation
    % find average value over last half of sim
    Pc(nv)=-mean(Ps(end/2:end,1,1)+Ps(end/2:end,2,2))/2; % average collisional
    Ks(nv)=mean(Ek(end/2:end));                          % kinetic
    Es(nv)=mean(Ek(end/2:end)+Ep(end/2:end));            % total for check
    % visual feedback
    figure(2);
    plot(phlist(1:nv),Lx*Ly*Pc(1:nv)./Ks(1:nv));
    drawnow;
    figure(1);
  end
  save('MDLabIdata.mat');     % save for later
else
  load('MDLabIdata.mat');     % load second time around
end

% nice plot:
plot(phlist,Lx*Ly*Pc./Ks,'o-','linewidth',2);
axis([.1 .9 0 20])
set(gca,'fontsize',20);
xlabel('$$\phi$$','interpreter','latex');
ylabel('$$\mathcal{X}(\phi)$$','interpreter','latex');

m-files

Data files

Published with MATLAB® 7.14