Classical physics

Prof. V. Balakrishnan
Department of Physics
Indian Institute of Technology, Madras

Lecture No. # 12

We ended last time by writing down the Liouville's Arnold criterion for integrability of a
Hamiltonian system and I pointed out that, when a system is fully integrable you suppose to have
for an n degree of freedom system, n constants of a motion in involution with each other, and
then the statement was this is necessary and sufficient for you to find a canonical transformation,
which would to action angle variables, after which the problem is in principle solved. Now, what
I am going to do now is to give you number of examples and we apply this criterion and ask is
this problem integrable or not integrable. And that will decide for us; what we should expect in
the general case? So, let us start with the simplest of problems.

(Refer Slide Time: 01:54)

Example one, a Hamiltonian with one degree of freedom H of q p given, equal to p square over 2
m plus V of q say for example; so in this problem n equal to 1, 1 degree of freedom. Is this an
integrable system or not?
Yes, it is. I know that H of q p is a constant of the motion and n is equal to 1 and I need just one
of them, it is of course an involution with itself and that is it. So, every one degree of freedom
problem is integrable, is solvable in principle. You can always write down the phase trajectories
with simply our H of q p equal to constant; those are the constant energy curves. What about this
example two, with n equal to 2, I have an H of q 1, q 2, p 1 and p 2, and this happens to be in the
form H 1 of q 1 p 1 plus H 2 of q 2 p 2. So, it is a two degree of freedom system, which you
could regard as two particles, for example, moving on an axis moving on the x axis for instance.
But the Hamiltonian happens to be the sum of two Hamiltonians; one of which has nothing to do
with the other pair of variables and vice-versa.

Is this an integrable system?

Yes, when are the two constants of the motion in involution with each other?

H one and H two, because it will turn out since q one pass on bracket q two is zero, q one with p
two is zero, p one with q two is zero and p one with p two is zero. This set of this function here
and that functions there are in involution, the pass on bracket it is a guaranteed to be zero. So,
your guaranteed that H one, H two is equal to zero. So, you have two independent constants of
the motion and this is sufficient, it is just like saying I have two separate particles, one of them
here and one of them somewhere else and each of them is one degree of freedom system and its
solved completely. What about a generalization to general n, the same thing H of q p equal to a
summation i equal to one to n, H i of q i, p i.

Is this solvable? Is this integrable?

Yes, it is just n uncoupled one degree of freedom systems and this is immediately integrable.
Where are the n constants of the motion in involution?

The different H i's are all in involution with each other and we are guaranteed, this is trivial and
completely solvable. Of course, you see that there were gone to be a problems once you have
interaction terms, where the different q’s interact to combine with each other and there are
functions which involve q one and q two and so on. Then, it is different story. So, all such
separable problems are integral, so it is no problem. Now, let us look at some other examples of
interest.
(Refer Slide Time: 05:19)

Let us take a free particle, so this is an n equal to two cases. Let us take a free particle moving
inside a square box in two dimensions. Now, you got constraints and so on. So, the particles
inside the box and let us say this box is some kind of zero to L. For example, this is the x
direction, that is the y direction and that is L and this particle is confined to remain inside this
box and it is free, it moves on this plane and it is free no forces on it. What is the Hamiltonian of
this problem? So, this is x y and let me just call it p x and p y this is q one, q two, p one and p
two just the Cartesian components of the momento. And what is the Hamiltonian?

It is just kinetic energy, there is no potential energy. So, this is equal to p x square over two m
plus p y square over two m, this is example three for instance. Is this the total Hamiltonian or is
there some other term?

Pardon me

There is infinite potential outside the box; it cannot get out of the box. So, we assuming it are in
a box with perfectly reflecting walls coefficient of restitution is unity. There is elastic collusions
with the walls of the box, but it is not allowed to go out of the box this is forbidden. So, those are
constraints it says in this problem zero less then equal to x less then equal to L, zero less then
equal to y less then equal to L. Those are constraints, but they are not holonomic constraints they
do not decrease the number of degrees of freedom. So, you leave them as they are and then the
problem really has a potential also but then you have to say, the potential is zero inside the box
and infinite outside the box. So, there is a penitential V of x comma y and this guy is equal to
zero inside box and infinite outside box. Now, where we interested in the motion inside the box?

Well, we simple saying it cannot go outside that is it. So, I do not want to have situation where I
have infinite only on a line and may be it will tunnel through and so on we do not want to do
anything, only this is just infinite outside. Now, is this an integrable system? We, interested in
the motion inside the box nothing more than that; so, what would happen physically if I started
with a particle here and I gave it an initial velocity in the y direction for instance, it start from
here what would its subsequent path be?

It would just bounce off and come back, it would keep doing this if I start here, it would of
course go down and bounce and then by the law of reflection it would do this and it would keep
going. So, it can execute fairly complicated trajectories inside depending on what the initial
conditions are; but the question asked is this integrable? By which I mean if I specify the initial
positions and momenta at p equal to zero, can I predict analytically can I write down what the
solution is at an arbitrary instant of time, no matter how long in the future. For this it is necessary
and sufficient that you must have two constants of the motion that are in involution with each
other. Are they two set constants?

p x square is not a constant of p x is not a constant of the motion, p x is not because as soon as
you hit this wall, the vertical wall p x is reversed in sign, if you hit this wall p y is reversed in
sign. So, p x and p y are not constants of the motion.

But, p x squared and p y squared are constants of the motion. So, certainly this is true and they
are analytical constants of the motion. So, we know that p x square and this is sufficient, it is
integrable this is of going to be of some interest. Because, you can know change the situation just
a little bit and the system will become chaotic.

Suppose, the Hamiltonian is no longer differentiable then you are in trouble, I assume that all
these are analytic constants of the motion. Yes, excellent. In this problem itself, the Hamiltonian
is not differentiable it is got infinite discontinuities, that does not bother us so much. The real
problem is what happens, if I shoot the particle directly into that corner. What happens now?
How do I apply the law of reflection? So, this problem is set to be pseudo integrable, because
there are sets of measure zero, initial conditions with a sets of measure zero for which you cannot
write down what the solution is. You assume then for simplicity that anything that hits the corner
is absorbed and that is the end of it; so, apart from that technicality this problem is solvable, it is
integrable and so on.

Since, he does not like the idea of sharp corners; Let us look at a circle and put the particle inside
a circular stadium this is like carom coin. So, let us look at example four, circular. These are
called stadia, this is called circular stadium and you have a particle inside a circular box and it is
confined to stay inside here. Now, the Hamiltonian is still p x squared over two m plus p y
squared over two m, but what is there is there a problem integrable? Because, neither p x square
nor p y square is going to be a constant of the motion; is this integrable? Because if you hit this is
going to do that, and then it is going to do crazy things. So, at each stage you got to find the
normal and then you have to find out what there angle of reflection is and so on. There may be
some special trajectories some special initial conditions where this guy would just go through a
diameter back and forth or it will go in an equilateral triangle and so on. But, in general of course
for arbitrary initial conditions that is not guaranteed at all. Is this integrable?

Pardon me

We should go to polar coordinates, plain polar coordinates. Then, the problem appears to become
integrable but what is the other constant of the motion? The Hamiltonian is a constant of the
motion; the energy is conserved of course. Is the distance from the centre constant?

Not quite; not quite close but not quite. So, what is constant in this problem? Let us look at it
from first principles; this problem has no potential it is a free particle. Is angular momentum
conserved? Angular momentum about the centre is conserved, because this problem the
boundary has circular symmetry. So, the boundary also matters if the potential is zero inside and
infinite on the boundary; the boundary has circular symmetry then this problem has circular
symmetry you can actually rotate the coordinate axis and nothing will change. So, what is the
other constant of the motion? One of them is H, which is p x squared plus p y squared over two
m that is equal to the total momentum squared over two m, that is a constant of the motion. What
is the other constant of the motion? It is the angular momentum about the origin and what is that?

But, can you be write it in a Cartesian coordinates?

Since, it is a plainer problem angular momentum has only one component; in two dimensional
angular movement is got only one component, it is not a vector there is no z direction, it is just a
x p y minus y p x that is it. Now, I leave you to check that x p y minus y p x in this problem
actually commutes with personal commutes with p x squared plus p y squared over two m. We
have two constants of the motion and therefore this problem is solvable. Now, he mentioned
something about the distance from the center; now the angular momentum as you can see the
magnitude the speed of this particle is not going to change. So, if this is the distance of closest
approach then you can actually find the magnitude of the momentum by multiplying this impact
factor multiplied by the speed time’s m that is going to be constant.

So, it is evident that no matter what your initial condition is this particle either will have a close
trajectory or will go on doing this. So, at some stage it will do this it will keep bouncing off and
there would be an inner circle into which the particle can never come and outside like those
thread work things you have seen the pins struck on the board and then you have an envelope
curves. So, this circle inside will form like an envelope curve. But, the problem is integrable; it is
solvable completely. So, the circular stadium is solvable.

Now, you can play this game and ask what happens, if you have an elliptic stadium. Next thing is
to ask what happens, if you have an elliptic stadium. So, let us look at that, it is a non trivial
problem really. So, I have a particle moving in an elliptic stadium with two foci here. Do you
think this is an integrable problem? Angular momentum about the center of the ellipse is not
conserved, definitely not because in this problem you definitely do not have circular symmetry.

The sum of the angular momenta with respect to each of the foci, this is conserved in this
problem. So, this stadium is also solvable this thing is solvable. What happens if I do this, so
another example I do not want to number it, because it is not something I am going to discuss
now.
(Refer Slide Time: 16:18)

I have a square stadium and I put a circular obstacle inside at the center and I am not allowed to
go through into that. So, whenever it hits the particle hits that obstacle it bounces off, by the laws
of reflection. Do you think this is problem is solvable? It does not have Cartesian symmetry,
because this obstacle does not have Cartesian symmetry it is got circular symmetry, but then the
boundary does not have circular symmetry. So, there is a conflict here between circular
symmetry and Cartesian symmetry. Do you think this problem is solvable?

You need further information, this is not an integrable system this simple looking; thing is not an
integrable system and let me tell you the mechanism by which chaos appears here. We are not
going to discuss it in great detail right now and that is the following. What happens if the system
is integrable? Is that we saw you could go the action angle variables and once you go the action
angle variables, we saw the action remains constant and the angle variable increase linearly in
time; which means that if you start with two face trajectories one of them here and one of them
here with an enabling initial condition, the distance between these two can only increase linearly
in time in the angle variable. And when you go back to the original variables, it would increase
in some prescribe fashion some known fashion, but the fact is errors do not amplify in this
problem not exponentially any way, you just increase linearly.

But, in this problem a very simple physical consideration shows you can be in deep trouble,
because if I showed a particle at it like this it bounces back, but I shot it a little bit to the right,
ever so little to the right. Then, the next time it does this and then it does this and then it does this
etc, but ever so little to the left would cause it to take a totally different history and this spreads
out. And, any error in that initial angle can actually become as big as a system size itself, any
separation initially can become as big as a system itself. Due to the fact, that this Cartesian
symmetry is not come insulate with this circulate symmetry here and it does deep focusing effect
here. And because of this, the system becomes chaotic; it is not predictable. There are lots of
trajectories in initial condition, which you can predict things. For example, if I shoot it like this it
would do this or if I shoot it in this fashion, it would keep going in a trajectory of this kind no
problem, but there are sets of non-zero measure initial conditions for which the trajectory in
future cannot be predictable; it is not computable and this stadium has chaos.

Now, you might say this is very easy, because every time you have something like this a
defocusing effect at a convex lens you are in good shape, you can immediately say this is going
to defocus and produce chaos. So, to disabuse you of that notion let me point out that the
following is going to happen. We saw that the square stadium was integrable and the circular
stadium was integrable. So, suppose I take a circular stadium I cut it in two and separate the two
pieces and then I have a semicircle and then a straight segment here and straight segment here.
The slope at this point and slope at this point, there is no discontinuity. But the curvature has a
discontinuity.

Because the straight line has zero curvature and this is finite curvature. So, this stadium very
famous one; it is called a bunimovich stadium and it has no convex surface, it has two concave
lenses. If you like concave mirrors here and then straight mirrors, plane mirrors, but this problem
is also chaotic. As long as this ratio of this straight line segment to the radius of the circle is non
zero, you have chaotic behavior. There is defocusing here in this problem and to give you an
optics analogy, the defocusing occurs not because of the defocusing of a convex mirror, but it
occurs due to other optical abrasions. So, you could have astigmatism for instances and you
could have spherical abrasion.

So, those effects could because this is like ray optics as you can see and that produces chaos in
this problem. And then, of course you can categorize various kinds of stereotracy which would
produces chaos and which would not and most of the time systems are chaotic. So, even two
degree of freedom systems because the face pace is four dimensional, there are four differential
equations can produce chaos. But, these are sort of mathematical models let us go the physical
models, let us go and ask next what happens and that is going to be our example five, I believe or
four example five.

(Refer Slide Time: 21:32)

Example five; particle in a central potential in three dimensions; so now we trying to look for
physical problems and so we look at the physical problems of a particle in three dimensions
moving in a central potential. What is the Hamiltonian going to be? It is the function of r and p
and this is equal to the kinetic energy plus a potential v, which is a function of little r alone no
angular dependences. So, there are two cyclic coordinates here to start with and the potential
there is no r depend, theta dependence, no pie dependence, but remember in p square there is
theta dependence.

If I wrote this out in spherical polar coordinates, this is p r square over two m plus p theta square
plus p pie square over sin square theta over two m r square plus V of r; this is what you call the
square of the angular momentum. So, there is one constant of the motion, how many degrees of
freedom does this problem have?

Three degrees of freedom; so you need three constants of the motion in involution with each
other to solve this problem. What are they? The Hamiltonian is one. What are the others? Since,
it has spherical symmetry there is no talk on this particle. Therefore, angular momentum is
conserved. So, it is looks like you have an H and you have angular momentum itself L and that is
got three constants of the motion in it L x, L y and L z. Are they in involution with each other?
No, unfortunately no. So, that is not true. They are not in involution with each other. So, tell me
do we have Constance of the motion in involution?

You can choose one component, what would you like to choose?

The z component, the e component does not have to be any component. So, let us choose z
component so there is H there is L z, L x L y are out because as soon you choose this there are
the two not in involution. Therefore, you cannot choose that. Anything else? Think of quantum
mechanics in the hydrogen atom, you have this principle quantum number, you have the orbital
angle momentum quantum number and you have this magnetic quantum number. Quantum
numbers are like the analogs of constants of the motion here in classical physics. So, what is the
other Constance of a motion?

Yes, L square. Absolutely, L square H L z L vector square this pass on commutes with every
component you know that L x L y L z do not commute with each other, but each of them
commutes with L square and you have to chose any one of them. Now, we have chosen an L z
purely out of convenience, you could have chosen any other component could be having done
this could we have chosen L dot some arbitrary unit vector n. Yes indeed. Any one direction in
space you could have chosen that along with L square. So, these three form three independent
Constance of the motion in involution with each other, they are functionally independent of each
other; specifying two of them do not specify the third completely and therefore this problem is
integrable. The central force problem is integrable in three dimensions.

Notice, I am not made use of the fact that this is of the form one over r it does not have to be
every central force problem is integrable, the one over r and r square potential will be very
special they have extra symmetries over and about this. There trajectories would have very very
special properties, but in principle every central force problem in three dimensional is a single
particle is integrable. The rest is beated writing down the solution, choosing the proper
coordinates and generalizes momenta and so on. How about two particles interacting in the
following way, so that is example six.
(Refer Slide Time: 26:17)

Two particle system, now I have H of r 1, r 2, p 1 and p 2 and let us suppose it is of the form p
one square over two m one plus p two square over two m two plus a potential V which is a
function of r one and r two in general arbitrary potential. Do you think this is integrable?
Remember V is not a function of r one alone plus a function of r two alone then of course, it is
separable. But, it is an arbitrary function of both these vectors do you think this is integrable first
of all what is n in this problem, n is six; you need twelve dimensional face pace and you need six
constants of the motion. We have one, the Hamiltonian is angular momentum concerned.

No, no about what the question you have asked is about what? What would you do to reduce this
problem?

You go to the center of math’s coordinates, you go to a system where you change variables from
r one, r two to the center of math’s and the relative coordinate. So, perhaps you do the following,
you write R equal to m one r one plus m two r two divided by m one plus m two and little r equal
to r one minus r two. Is this going to help? Is this going to help? It is going to help here, so what
is happens to this thing? What happens to the kinetic energy, if I do that?

Well, conjugate to this R you would have total momentum P which is p one plus p two and
conjugate to this you would have a relative momentum p which is equal to p one minus p two
and then what happens to the kinetic energy? You change variables to these guys and then what
happens, whether a certainly a contribution which is p square over twice total momentum
squared no not reduce mass; this is the total system moving. Total mass M. We put M equal to
m one plus m two, plus you see you cannot change the number of variables, does not r one and r
two and p one and p two now you got a little r and big r little p and big p, that is it.

What happens to the remaining terms, this is p squared divided by twice capital m. This is the
relative momentum squared, reduce mass here this is reduce the mass. So, it is simple exercises,
you know how the reduced mass is defined?

m one m two divided by m one plus m two. So, that part is fine plus what happens to v you
should do these exercises at some stage, you should take is not you know change these variables
plug it in see, what happens; what happens to V? It some arbitrary functions something else
some U of little r and capital R, there is nothing u can do about it; that is it. Is really nothing you
can do about it in general? So, would you say this problem is integrable? You gone to the center
of mass coordinate, but it has not produced any particular simplification. But, suppose V of r one
comma r two equal to a function of the distance r one minus r two alone. Suppose, that is true
then what happens? It instantly says this becomes V of r implies this. Is there cyclic coordinate?
What is that? Capital R is the cyclic coordinate. So, capital R implies capital R is a cyclic
coordinate.

What constancy of what constant of the motion does that yield at once? Capital P, that says the
entire centre of mass moves either at constant speed forever or stays at rest depending on the
initial condition; inside something is going on. So, this immediately implies P is a COM. Is this
problem integrable? Would you say this problem is integrable? Where the Hamiltonian is the
constant of the motion, p is the constant of the motion. So, is that integrable? Why not? How
many constants of the motion you need now? So, you got six of see this problem with the capital
P is completely solvable, it is totally solved then you have left with this because this is like
saying I have two decoupled Hamiltonians; one which involves a capital variable, one which
involves small variables, the relative variables. The capital variables is a free particles the center
of max acts like free particles of course you can solve it. Hence, immediately solved it is just
gone a move into a straight line at constant speed, whatever be the initial speed and that is the
end of it.

So, this is like reducing it in to a decoupled system. This is done it is finish, and then you left
with a single degree of freedom in three dimension. So, three degrees of freedom here
corresponding to little hours components and now this problem here is a central force problem,
which we saw a solved. So, the orbital angle of the momentum is a constant.

So, that Hamiltonian together with the remaining two that L square the corresponding L square
and the single component is like a central force problem and it solved. So, this is solvable, it is
integrable. As long as the potential is a function of the difference between the distances between
the two particles it is solved, it is integrable completely. Because, two body problems is reduced
to two one body problems one of which is free motion and other is just simple central force. So,
this is very much integrable.

So, I am going to leave it to you an exercise to figure out, where are the constants of the motion
you must checked there in involution with each other, six independent constants of the motion in
involution. So, the hint I am giving you is that it is like two one particle problem and this set of
three coordinates and that side of three coordinates are nothing do with each other anymore and
therefore it solvable completely. So, this is very much integrable.

Yes

Yes, that is a vector. So, this stands for P x P one x P one y P one z etc it is moving in three-
dimensional space.

I am not able to write bold face on the board so I put an arrow but then I put a square there I so
stands for p dot p, which is a scale. So, it is a certainly a scale. So, this is solved. Now, suppose
you have three particles this is the very important problem, because you have three particles and
then let us assume that there interacting with each other by gravitational force for example,
which depends on the distance between the two particles, so what would you say is happening,
example seven.
(Refer Slide Time: 35:58)

Let us, in fact look at the n body problem. So, you look at the n body problem this Hamiltonian
is a function of r one to r N, p one to p N and that is equal to a summation i equal to one to n p i
square is put a bracket here over two mi that is the kinetic energy plus and now, yeah.

That is right pardon me, what is an potential yes No, the two particles experience potentials
only due to each other, so there is no external potential added.

This is a function of r one and r two if you like you can add to this a constant potential, that is not
gone to effect things. It might break the symmetry of course if I add a gravitational field might
break the symmetry then you have a problem. So, you may not be able to right this anymore, but
the assumption I made was that this is a set of particles interacting with each other by pair voice
forces, in fact in this case just two of them and the force is entirely a function of the two
coordinates.

Of course, I assume no velocity dependent forces then that was not good enough so, I said the
force between these two particles is derived from a potential. So, it is a conservative force and
that potential depends on the mutual distance between the particles. No non central force here
and that immediately gave me a scalar it said this potential depends only on this distance and that
got reduce to a central force form for the relative coordinates. And therefore, the problem was
solvable. Now, with this understanding that we know this problem is reduce to in this fashion I
try to do the n body problem.
So, now let us assume I know that the n body problem with the arbitrary forces is not integrable;
even two body problem is not. So, let us assume that this potential here is V which is a function
of modulus r i minus r j summed over i j equal to one to N and i not equal to j, simply to ensure
that I am not adding unnecessary terms which corresponds to self interaction. Each point
particles sees the force is due to all the other particles, but their pair voice interactions particle
one interacts with two, three, four, etc always depend on the distance between the given particle
and the other particle.

Excellent question, Are there two three body forces in, the answer is yes. This is a heart problem
nuclear physics is happens all the time then of course, this whole things goes out of the window.
There are three body true genuine three body effects to occur, we can give probabilistic
arguments about how often they occur so on and but they do, they do happen.

Yes

Any number, yes this is possible in principle; yes it is possible an N body problem in quantum
physics this is quite routine you really have N body potentials, but there is also another
possibility the force between two particles need not depend on the just the distance between
them. You could also depend on the vectors r one and r two, can you give me simple example of
this? A simple example, with which you already familiar.

Yes, in electro magnetism where even an electro statics this such as a simple force, what happens
if you got two points dipoles electric dipoles P one and P two? What happens to the potential
energy between them so, you got a P one here to use the word the symbol P one, let us call the
electrical dipole D one so, let D one here and little D two here and the distance between them is
r. So, let us say this guy is at coordinate r one and this is at coordinate r two and what is the
potential energy between these two dipoles; they are point dipoles. So, you assume that the
length of the dipole goes to zero the charge goes to infinity. So, the product is one point is
infinite, what is the potential energy?

What is a potential energy between two dipoles, I am studied electrostatic so there suddenly a
term which is d one dot d two incidentally this must be a scalar, it must depend on the distance
between then and so on. So, d one dot d two divided by r one two cube apart from one over four
pi epsilon not which is matter of units. So, this is proportional to this term here it must linearly
depend on d one, it must linearly depend on d two we got d one dot d two over r cube r one two
cube, but that is not enough because I know that the force between two dipoles like this is
different from the force between two dipoles like that.

So, it cannot just depend on the distance and certainly this force is different from that force. So,
what is the next term? There is also a term which is three d one dot r d two dot r over let me just
call r, let us call this vector r the difference between the two and what is the power here?

Five, because I put to factors is here so you see this depends on the relative orientation of the
vector r, the distance between these loop the vector joining one to the other with d one and d two.
It must be symmetric under the interchange of d one and d two, it must linear in d one, linear in d
two and you got three vectors to play with d one d two and r, we cannot depend on the overall
origin but only on r.

So, these are the only possibilities. So, it is linear dependents and incidentally it should be the
same whether you check r to minus r or not. Because, whether you call r the vector going from
dipole one to two or two to one does not matter and that too is preserved here; this is not a
central force.

Well, lots a question of you know it is a question of resolution, if I have an atom then on distance
is much larger then the atom it would certainly have a dipole could have a dipole moment or
magnetic dipoled moment. So, it is not a question of whether something exists physically or not.
I know for instance if I have an arbitrary charge distribution I could resolve it the potential due to
this from, as if you had an effective monopole and then a dipole and a Quadra pole and so on and
so forth.

Since, the resolution into spherically into components whose transformation properties under
rotations are known to be a question of approximation; I mean a ceiling fan is a dipole in some
approximation, because what was the force line due to the fan due to a dipole look like; they
come out like this and they go back in this fashion is not it. This is what dipole force is going to
look like. The field lines are going to look like this they go out this way and they come in this
way. So, you could imagine the center of the ceiling fan it is sucking in air from an above and
pushing it through and then going back.

So, and some approximation it is really like a dipole source of a velocity field. So, this is a non
central, this is an example of a non central. But, our interest now is in trying to find out whether
this problem is solvable or not. Because now we made all assumptions we need you said you got
n particles and a c forces is only due to each and the force between any pair of particles is
dependent only on the distance between them. It is directed along the line joining these two
particles, so that is a scalar v of mode r i minus r is scaled. And these whole thing is spherically
symmetrical in the sense that if I a rotate the coordinate system mode r i minus r j does not
change. Therefore, the potential does not change, so it has spherical symmetry. Would you say
this is integrable?

In the two body case, it was but in the three body case let us right down all the constants of the
motion that we can right down. So, COM’s the first one is a Hamiltonian itself and the second
one would be the total angular momentum of this system, that is certainly a constant of the
motion. What is the total angular momentum? Summation over i equal to one to n r i cross p i r
cross p is the angular momentum for each particle, the orbital angular momentum about some
origin incidentally; this Hamiltonian is invariant under shift of the origin.

Because, it only depends on the distances so, you can choose the overall origin wherever you like
and certainly L equal to r cross p is a constant of the motion. There are three Constance of the
motion there; three components. Anything else is a constant of the motion? How about the total
momentum of the system? Is that? That is a constant that is external force so that certainly a
constant generalization of Newton’s third law totals momentum. Because, there is no external
force on the system, so three capital P equal to summation i equal to one to N P i that is a
constant of the motion. How many do we have now? We got one plus three four plus three seven.
What is the little N equal to?

No, then I am only labeling now right now, I am only listing the constants of the motion after
that we got examine how many of them are going to be in involution with each other. So, that is
for integrability you need that but what is n equal to three N. Keep that as the back of our minds
we found grand total of seven so far. What about you see I know this guy, I know this is the
constant of the motion and therefore I know that R of t equal to R of zero plus P t over M. I
certainly know that I know the center of mass of this entire system is going to move at uniform
speed a constant velocity simply, because there is no external force on it and that is depends on
the initial condition. So, other constants of the motion there?
There are. So, we have four R minus P t over M the time dependent, but then we yesterday we
saw that we could have time dependent constants of the motion. How many are there now? Ten.
They call the ten Galilean constants of the motion; the Galilean invariance. In the absence of
further information, that is it is really nothing else you can do. Now, we still have to worry about
the her problem which is how many of them are going to be involution with each other, it is clear
the three components of L are not in involution with each other. They have to be involution with
each and with the Hamiltonian of course, each of them is involution with Hamiltonian as a
constants of the motion.

There is the three components of P they would certainly be in involution with everything. p with
L is not true immediately it is called in any case you need 3 N constants of the motion, because
we talk of integrability and you have a grand total of ten an even they are not involution with
each other. For n equal to three, the three body problem only nine constants of the motion and
you are far below that, even for three particles. Let us forget about ten to the twenty-three
particles. So, the system is badly chaotic even the three body problem is not integrable except in
very special cases, then of course you have ten to the twenty-three particles is gone. We saw
infact that if you have a single particle inside a stadium with a fix scatterer who is got a different
symmetry in two dimensions even, that is chaotic.

So, certainly what is happening to the gas in this room is highly chaotic there is no possibility of
solving it. So, the reason why you need statistical mechanics partly is not just that you have ten
to the twenty three particles and even god cannot write all those equation down. But even if you
were three particles the situation is bad enough it is gone; it is cannot be integrated in general.

It is a very good question, how do I know I found out everything. It is possible to find out we can
make a for the statement, the point is it is a very good question because we have to ask where are
these constants of the motion coming from? What is the represent and the answer lies in theorem
which I am going to do tomorrow, mention this. This constant of the motion associated with a
symmetry of the system always symmetry. We saw if got spherical symmetry or circular
symmetry, it had angular momentum as a constant of the motion. So, symmetry is going to imply
invariance which is going to imply a conservation principle or conservation law and that is going
to give you conserved quantities. So, this is where the constants of the motion come from and
then you have to ask for what is the symmetry of this Hamiltonian and the most general
Hamiltonian does not have any much more symmetry than spherical symmetry over all that is it.

So, really you cannot do much more so we have to now go back and ask what simple
Hamiltonian looks would like. What kind of symmetry they do have and then we take lesson
from that and see what happens in the general case. So, we now back track and let us go back to
a problem which we can slow completely which is integrable and look at what is symmetry is
and how that symmetry can be broken and the simplest of these examples is to go back and look
at two simple harmonic oscillators, because one simple harmonic oscillator is doable completely.

(Refer Slide Time: 52:41)

So, now let us go back and look at 2 D oscillator and by that I mean the situation in which H of q
one q two p one p two is just two linear harmonic oscillators uncouple to each other. So, you
could regard this as a single oscillator which is got a force in both x and y directions or you could
regard it as two simple harmonic oscillators uncouple from each other does not matter. So, let me
right as H one of q one p one plus H two of q two p two and let simplify things a little bit by
saying this is equal to one over two m p one square; let us right it out p one square over two m
plus one half m omega l square q one square plus I know that the actual property of oscillators
basically the ratio of spring constant to the mass that is the relevant parameter in the frequency.

So, let me call the frequency omega one and let us take the second oscillator to have the same
mass, but perhaps a different frequency q two square this fashion. I can go to action angle
variables this problem is totally solvable, we know this is an independent oscillator that is an
independent oscillator and each of them has a phase trajectory in it is q p plain which is an
ellipse of some kind. But the real system has four a four dimensional phase space and all we can
do we cannot draw the phase trajectories, because I cannot draw four dimensions. But, I can
draw projections of the phase trajectory on to various plains so I could do the following and start
by saying asking what happens in the q one q p one plane.

(Refer Slide Time: 54:32)

So, here is q one p one and what happens in the q 2 p 2 plain well, this guy would go round in
this fashion and this fellow would also go round in this fashion. And, any given instant of time
you are on some point on this ellipse and some point on that ellipse and they need not be in
phase. These two oscillators need not been phase at all; this problem is solvable. I could if with a
little bit effort draw the trajectory in the p one q two plain q two p one plain and so on q one q
two plain and so on and so forth. What you call the trajectory drawn on the q one q two plain?

The two oscillators they could regard them as right angles they like figures. Now, we have got
sophisticated we are not gone look at it in these coordinates, we know I can we can go to action
angel variables and this action angle variables I define, so from here I go to an i one the theta one
i one and form here I go theta two i two and I know that i one and i two are constants. Therefore,
let us just look at it as a function of theta one theta two. What would they look like? What would
the space look like?
(Refer Slide Time: 56:06)

Well, to describe theta one I would like to take an angle and to describe theta two I take another
angel and the space is the space which is the direct product of two circles if you like and that is a
two dimensional torus. So, let us draw this torus in this fashion and as I move along this tube in
this direction, I thought say it is theta one and as move in the transverse direction, I call it theta
two. And the point the phase space point you could go a little further you could say let us take
the diameter the radius of this tube to be proportional to i two and the radius of this guy to be
proportional to i one. So, you can see this entire four dimensional space is laminated by these
torie and each torus specifies what i one and i two are and the motion on the i one direction has
time period two pi over omega one and the other direction has 2 pi over omega 2.

Now, tell me what would the trajectory look like in general? Suppose, omega one equal to omega
two a simplest case then it means that as the representative point moves once around in this
direction, it moves once in this direction also. So, it is sort of winds round and comes back and
what is the figure looks like. If this is the amplitude in the two direction and that is the
amplitude in the one direction what would the Lisa to figure look like.

In general, this would be a periodic curve it closes so what would it look like well, if these two
fellows are in phase then of course, it will just do this. But if they are not in phase, then in
generally it will be an ellipse of some kind, if there exactly 90 degree out of phase it would be an
ellipse of this kind. If there is some arbitrary angel out of phase it would be an ellipse of this
kind, this is when the two frequencies are equal. Is the motion periodic?

Yes, it is periodic. What if one frequency is twice another frequency? What if omega one equal
to twice omega two? Then, it says as the system goes around here once this it makes two curves
on this side, but it would do I cannot draw this two well but do some crazy thing. But it would
come back to the same point and the Lisa to figure in this case again very imperfectly, drawing it
very imperfectly could perhaps do something like this like a figure of eight but it would close on
itself.

What if omega one over omega two is a rational number P times omega one is q times omega
two r and s for example in r times omega one is s times omega two, where r and s are integers.
Then again, the Lisa to figure gets more complicated but it closes on itself. What if the frequency
ration is irrational? What happens then when you see you might have seen these Lisa to figures
they densely fill up this rectangle and the system never returns to it is original point, no matter
where you start. So, it is periodic in q one in q one p one periodic in q two p two, but the two
periods are not commenced with each other. So, therefore the face trajectory on this torus will
never close, it will never close it will be like a bowl of thread going like round and round on top
of this but never coming back to it is original point.

And, we can simplify matters little bit like by saying well let us write this Hamiltonian as p one
square plus over two m plus half m q so, let us put p one square over two m plus half m, I put
omega one equal to one so q one square plus p two square over two m plus half m omega square
q two square. And, this is an irrational number the ratio of frequencies is a rational omega is
irrational, then this Hamiltonian is integrable it is completely solvable just the some two
oscillators; but on this torus the trajectory does not close. Would you say this is periodic motion?
Because for me, periodic motion is when all the phase space variables, come back to their
original values after finite amount of time.

So, the motion is said to be quasi periodic it is not periodic in general. So, omega irrational
omega one omega two irrational implies quasi periodic motion. And, it has strange properties
and you could do the following you could say let me set this frequency equal to one and look at it
as a function of this second angle. So, I put a cross section here and this cross section is called a
Poincare section, is an example of Poincare section. Now, I ask when does the trajectory hit this
circle so what is happing is that is going round and round an every time it hits the circle I note it
down.

Therefore, that circle if I look it at separately it started here then it went round in the other
direction it came back here and then it went round in other direction, it came back some where
here and so on. This is like saying each time you add an irrational number if this as unit
circumference your adding an irrational number to the circumference. Of course, if you add a
rational number then every point will come back to itself after certain amount of time. But, if you
add an irrational number then there is a theorem due to wire stars with says if you take unit
circle, circle of unit circumference add an irrational number.

So, what are you doing you saying theta n plus one equal to theta n plus an irrational number
omega modulo one modulo two pie or one it does not matter this modulo one and this is
irrational. Modulo one means you remove the integer and you come back. So, you start here and
then you go here, you go here and you keep going this you never come back to this point and
then you never over shoot or under shoot it and then it does this and so on.

And what happens is that this point of intersection given enough time will fill up the entire circle
densely and uniformly. No particular point on it is preferred over anything else and this
trajectory on the two dimensional torus will fill up eventually any initial condition will
eventually fill up this entire torus. And the system is now set to be quasi periodic and since any
initial point visits the neighborhood of the entire torus the system is said to be ergotic on the
torus.

You are adding on irrational number.

I do not care what theta is it does not matter, any theta. So, do this experiment on the pocket
calculator on a simple computer. You cannot add an irrational number because you will always
end with a finite precision on your computer. So, you really cannot add an irrational number so
what would you do, this is simple problem start with number theta not between zero and one it
does not matter any number, add two it an irrational number throw away the integer and keep
doing this and keep track of all the iterates and you will discover that they form histogram on the
unit interval which is uniform, completely uniform never comes back.
But now I leave it to as a problem as to how you would add an irrational number on a computer,
because any number you specify on the computer would be a rational number it terminates after
certain stage, the decimal point terminates after certain stage. So, what would you do?

You have finite precision and this now leads us into very certain questions you should try and
add as irrational number as possible, you should add a number which would be rational
eventually, but as irrational as possible by that I mean the periodicity will be very large. In other
words, if you express this number as a fraction you express it completely as a fraction exactly
you are gone but the best approximation to this would be you should be fraction with very large
denominator. Then, of course you know the period is very long, if the period is twenty-five
thousand and six hundred and forty-two then you do not really care on the finite mode of time.

So, you should try to add a number which is very got a very large it is a very irrational square
root of two minus one, good number for a reason I have explain later on. A square root of five
minus one divided by two this is a very good number to add.

So, I wanted to say that it becomes ergodic where that I meant that any initial condition is going
to visit arbitrarily close to every other point on the torus infinitely often as you keep going and
with equal frequency.

No, it is unit circle so otherwise you write modulo two pi add an irrational number. So, that is
why I said one because otherwise it is two pi irrational modulo two pi it becomes so I take a unit
circumference. And, this is a very remarkable theorem were it has it is magic property and you
see how number theory is getting into this whole game, you see how the property of irrational
numbers, rational numbers are getting in to this game. So, you should remember what irrational
numbers are they numbers at cannot have a decimal expansion, which terminates or recurs or if
you like they are numbers which cannot be expresses as a ratio of two integers p over q that p in
q is not zero.

What was that? I like choosing one twice, I am saying add an irrational number I said I choose
the circumference to be equal to one rather than two pie. So, in his question would be equal to I
am saying what happens if I add one, then of course you back but one is not irrational, you got it
choose an irrational number between zero and one; number which is an element this range
between 0 and 1. So, do not choose 0 or 0 1 of course, is trivial. Zero is the identity map, it does
not do anything and one is also the identity map its say every point goes back to itself.
But, any number in between will see there are no periodic orbits. There are no exceptions, there
are no numbers which you come back to themselves no initial conditions will come back to
itself. Now, in general therefore if I have an n dimensional system, n oscillator the same thing
would happen and the big lesson is that if I take a general Hamiltonian system and the system is
integrable, the motion in general is quasi periodic. Bounded motion is in general quasi periodic,
but the big difference between the harmonic oscillator and the general system is that the time
period of oscillation would depend on the energy or the amplitude or the action variables, where
as in this problem the frequency of going around that does not depend on the size of this torus,
does not depend on the action variable.

But, in the general case of course that been non linearity then it will depend on the size. And then
now you have to start imagining I have the n dimensional phase space, two n dimensional phase
space in which I have n dimensional tori. So, the question is what happens in between the tori, if
the system is full integrable then every point is on a torus. But, if it is not there are regions of
chaotic behavior in between and then the next question is can it is region escape and this will
again require a little bit of high dimensional of imagination, but it is not so difficult to come to
that we need this we will get to it. So, let we stop here.