This course gives you an introduction to modeling methods and simulation tools for a wide range of natural phenomena. The different methodologies that will be presented here can be applied to very wide range of topics such as fluid motion, stellar dynamics, population evolution, ... This course does not intend to go deeply into any numerical method or process and does not provide any recipe for the resolution of a particular problem. It is rather a basic guideline towards different methodologies that can be applied to solve any kind of problem and help you pick the one best suited for you. The assignments of this course will be made as practical as possible in order to allow you to actually create from scratch short programs that will solve simple problems. Although programming will be used extensively in this course we do not require any advanced programming experience in order to complete it.
Microdynamics of LGA
Loading...
来自 University of Geneva 的课程
Simulation and modeling of natural processes
62 个评分
从本节课中
Cellular Automata
This module defines the concept of cellular automata by outlining the basic building blocks of this method. Then an insight of how to apply this technique to natural phenomena is given. Finally the lattice gas automata, a subclass of models used for fluid flows, is presented.
与讲师见面
Bastien ChopardFull Professor
Computer Science
Jean-Luc FalconeResearch Associate
Computer Science
Jonas LattSenior Lecturer
Computer Science
Orestis MalaspinasResearch Associate
Computer Science Department
[MUSIC]
Hello everyone.
We are back in to our course on simulation modeling of natural processes.
In this chapter about cellular auto modeling and now I would like to
start this module on microdynamics of lattice gasses.
So I would like to know how we can describe the detailed
movement of our particle in a gas.
So you remember that these particle they are moving in a regular lattice with
some discrete possible velocity.
And when they hit each other there's a collision, and so
they take another direction.
So this is illustrated in this slide.
Again, where you can see here, several particles meeting at the same place,
colliding, then going out with a new direction and then moving to the nearest
neighbor where more particles will be met and new collisions will happen.
So, this picture tells you that you can actually divide the process two sub steps,
one is the collision, and one is the propagation or streaming step.
So [COUGH] last time we introduced this occupation number and
I telling us where there is or not a particle with velocity,
v i entering the side at the given time.
So now we just put another subscript which is in or
out to distinguish between the particle entering from the particle going out.
And then after you go out you are moving into the nearest neighbor side, okay?
So with this, you can formulate
the microdynamics of this particle with these two equations.
So first you get a new distribution of out going particle
from the in going particle plus a collision term which are actually
interaction of all the particle at this position.
And then this propagation telling you that If you are out
with a velocity in direction i at position r and t,
you will be in with still a velocity in the same direction i.
But on the nearest neighbor along that direction v and i,
at the next time slot, okay?
So, basically that's the two equations that describe the system, and
all the collision process is hidden in this function that we will
express in a few moments, okay?
So of course you can combine these two equation in only one and
you get what is the very well known equation of evolution of the population
at [COUGH] every lattice point.
Where we just used this notation that when you don't put any subscript,
it mean that it's actually an incoming particle.
And if the collision term would be zero,
it would mean that just we have a free streaming of the particle.
It'd just keep moving straight.
But that wouldn't be a very interesting problem.
So, of course, we want this collision term to do something interesting.
So I remind you that for this HPP gas, which is probably the simplest
we can consider, there are two important collision, is these two here which is
two hadron collision, which mean that if a particle in the direction, let's say i.
Meets a particle in the direction i plus two, meaning just the opposite,
we have four directions.
So I plus two is the opposite direction of I.
Okay, it will create new particles in the direction i plus 1 and i plus 3,
okay, so that's basically what we want to formulate in our equation.
So for this HPP model I have only four possible
velocity which in a Cartesian coordinate are those, and
then this is the expression of what I was just telling you in the previous slides.
Saying that the outgoing particle in direction i is one or zero
and if it were freeze framing it would be the same as the incoming particle.
So if this one is 1, this one is 1 too, this one is 0, this one is 0 too.
But now suppose it's 1.
So this particle can be deviated to another channel, another direction
provided you have exactly a particle, i + 2 in the opposite direction.
And no other particle in the vertical direction.
So in that case, you will destroy this particle in the direction i and you will
on the other hand create a particle in the other direction perpendicular to it.
So you can that you do it yourself,
quietly show that this expression is a Boolean expression,
so the quantity can be 0 or 1.
And exactly reproduce the collision scheme that I show you in the previous
slide, okay.
So, now we would like to make the connection with physics, and
in physics we like to talk about density and velocity.
So, what is the density?
Well, it's the sum of all the particles that arrive at the same position.
So if you sum up for the four direction, the particle is enters
the site on the side t, you get the density of the number of particles,
and that even site, okay?
And actually you can show that the outgoing mass,
which is the sum of all the particles that leave that site,
is exactly conserved by the rule.
And this is a property of collision terms, so
you can explicitly compute that from this expression by
summing this over i, and you will see that most of the terms, they cancel out.
You can also compute the momentum of a site, which we call j.
Which is by definition the product of the density
by the velocity field of that site.
And it's just the sum of the momentum of all the particles.
So each particle has a momentum in the direction of its velocity.
So in case of mass equal 1,
this is just the sum of the momentum of all the particle entering the side, okay?
And that can give me quantity j, but
I also got the quantity rho from my previous sum, so
I can extract the velocity associated to that specific cell by dividing j by rho.
And it's also easy to show that the momentum before collision and
the momentum after collision are exactly the same just by computing this
expression for the macrodynamics of previous slides.
So, I'd like to show some demos and
maybe give you some information about this demo.
And the first one
I'd like to show you is how does it look like this HPP gas.
When you simulate that, and
the my accumulation will be to have a box with a lot of particles.
And in the middle of this box I will put a higher density of particle.
So will see that mostly the particle are uniformly distributed over space.
Except in the middle, would mean that I put a high density in my gas,
and this high density will propagate as a sound wave if you want a pressure wave.
So we'll run that, and we'll see that of course you see this wave going on.
If I run it a second time you see that the high density I had
in the middle propagates away as a wave.
But it has a very weird shape.
Okay, you will expect a round shape for
a wave and here you see more like a square or a diamond shape.
And that's of course a sign that perhaps something is not optimal in this model.
So if you do the same now with this FHP model which has more symmetry,
you see that things are going much better.
So you still see a slight hexagonal structure.
But if you're far enough from the center, you are also very close to a circle.
And you can show that to second order of accuracy, this is a perfect circle.
So you can see it slightly as an ellipse because when you do
hexagonal lattice on a computer, you have to do some deformation, and
then the horizontal and vertical axes are not the same scale.
That's why you have a slight deformation which is truly artificial in the picture.
But still you see that by adding more symmetry to your space
you get also a much better symmetry of your macroscopic imaging behavior.
Now what's interesting also with these lattice gases is that they
can compute the movement of a lot of particle in a totally exact way,
there is no rounding error, no loss of information.
So this is illustrated with this experiment.
So you see, this is my gas HPP, so
is this a square symmetry and you could see a very artificial way
during how the particle which were continuous the left part of the system,
they start to invade the empty right compartment.
They can only move horizontally until they start hitting the wall and then mixing.
But still, what I'm claiming here is that the dynamic I've obtained here
is totally deterministic without any errors, meaning that
actually I could use this property of physics, which is time reversibility.
Know that the equation of Newton, we are invariant by reversing time,
if you reverse the velocity the particles.
Here it means that if I take all the particles, and
those going left, I told them to go right.
Those going up, I told to go down, and so on.
And I just run the exact same rule while I should actually retrace my own past.
Okay, that's the time we will simulate.
Let's check whether this is true.
Now I've started from my previous
configuration by just reversing the velocity.
And you'll see that the system evolve in a very unlikely situation
back to its initial state.
Okay so this is actually a very sensitive simulation.
If you do a very small error, like rounding a truncation in your arithmetic
but this is not the case here because we are just doing boolean, and
the computer can do boolean arithmetic exactly.
But assume that now I do the same experiment but I add just a little error,
so it mean that I added just one particle somewhere, okay.
And then you see that I don't go back to my initial state.
All the particle left in this
compartment are those that actually interacted with my extra particle.
So it mean that in practice,
even though Newton mechanics is a time reversal in variant.
There's no chance to produce it because there's no chance to know the system with
enough accuracy to change all of it velocity with enough accuracy.
If you have any perturbation or
error you will lose this property of exact time reversible.
But in this fully discrete world you can still play with that.
So, now I'd like to show you some other example.
Like for instance, this is another lattice gas,
which of course illustrates the motion of a sand grain and
this mimics the hourglass.
So, it's just a tool of what you can do with this type of a model.
That's another example where you combine the free dynamics
with some grain particle which you have here is in the snow.
So the snow falls and piles up and you can have
a Christmas card out of your simulation, if you want.
You can, of course, do other then free dynamics, for instance, you can
ask your particle to have collision that do not conserve momentum, only mass,
but produce what's called diffusion, as you've seen in one of the chapter.
And here's a diffusion with a special case,
where you have in the middle, a particle at rest.
And each time, when another particle sticks to it,
it will stay at rest and form this cluster.
Okay, so this is called a DLA.
Diffusion limited aggregation, and it's a very common pattern in
nature that you can reproduce with this simple system.
You can also combine several different species
in your system and do some chemical reaction,
meaning that if a species A meets species B it can turn into D and so on.
And this is an example of a famous process called
band, where you have a test tube [COUGH] with a gel and some substance.
And at the inlet of the test tube,
you inject some chemicals, which will diffuse into the test tube.
And by diffusing, it precipitate with the substance that was already
in the test tube, forming separate bands, and that's what of course was
extremely surprising to the chemist who found out at the end of the 1800.
Is why is it not a continuous process and
he also found there's a nice geometrical rule describing this.
And this is an example where I show you can simulate this with and
get all this property right, because you put,
write mesoscopic approximation of the reaction diffusion process.
Okay, [COUGH] so I think what we have seen in this lattice
class is that you may have wrong behavior from since
we saw this anisotropy in the lattice class, and
this is due to not enough symmetry in your lattice.
For other process like diffusion, that's not a problem.
Again it means that if you want to control this type of thing
you need to understand that you cannot do everything or anything you want.
You should be sure that physic is properly put into your model.
For instance in this HPP gas, you can see that yes, you conserve momentum.
But actually you conserve too much momentum.
You conserve momentum along each of the line of the lattice or each of the column.
And that's not very good.
You also have this checkerboard invariant which means that everything which with
a time key was on a let's say, white cell of a checkerboard.
And the next time the step will be on the black one, and so you don't interact with
all a particle you actually just partition your system in two sub-system and
that's may be not what you want.
Okay, so with this I finish
my module on the Microdynamic of Lattice Gases and some example.
And this also terminates our chapter on modeling.
Thank you for your attention.
[MUSIC]
