lecture 8

Topics:
Where are we now?
When is F the energy?
Symmetries and transformations
Example: space translations for one particle
Space translations for two particles
Space translations for many particles
Finding Symmetries
Rotations
Noethers theorem
Momentum conservation from Noethers theorem
More on rotations
What functions are invariant?
Example of functionals - Soap bubbles
Where are we now?
We have seen how conserved quantities can arise as the generalized momenta associated with vari-
ables that do not appear in the Lagrangian. Here we will discuss a generalization of this fact, that
is an even more important principle. In Lagrangian mechanics, continuous symmetries lead to
conserved quantities. We have already seen one example of this in our discussion of energy, and
we will begin by making the connection with time translation invariance. We will go on to dis-
cuss symmetries more generally, and also in more detail the specic example of space translation
symmetry, which leads to the conservation of total momentum.
When is F the energy?
Last time we talked about the quantity F, dened by
F =

j
q
j
L
q
j
L (1)
which when evaluated for solutions to Euler-Lagrange equations satises
d
dt
F =
L
t
(2)
We noted that F is often the energy. I said this in lecture 7, but I dont think it got into the notes, so
I thought it worth saying again. Of course, to even get started on this, we have to restrict ourselves
to situations where there are no frictional forces, and therefore nothing is converting kinetic and
potential energy into heat. Since heat is just kinetic energy of random particle motion, this really
just means that we need to look at a system at a sufciently fundamental level to see all the kinetic
1
energy explicitly. For now we will take care of this by restricting ourselves to systems described
by Lagrangians of the form T V . Then the rst answer is that F is the energy whenever the
kinetic energy T is quadratic in the velocities, q and the potential energy V does not depend on the
velocities. The reason is that the differential operator
z

z
(3)
counts the degree in z, because
z

z
z
n
= z nz
n1
= nz
n
(4)
That is, this operator acting on a term that is some power of z just gives the term back multiplied
by the power. When we sum over all the velocities, the differential operator

j
q
j

q
j
(5)
adds up the powers of all the velocities. When this acts on a function quadratic in the velocities,
every term in the function just gets multiplied by 2. For example

j
q
j

q
j
q
2
1
= q
1

q
1
q
2
1
= 2 q
2
1
(6)
and

j
q
j

q
j
q
1
q
2
= q
1

q
1
q
1
q
2
+ q
2

q
2
q
1
q
2
= q
1
q
2
+ q
1
q
2
= 2 q
1
q
2
(7)
Thus if T is quadratic in the velocities and V is independent of the velocities,

j
q
j
L
q
j
=

j
q
j

q
j
(T V ) = 2T (8)
and then
F =

j
q
j
L
q
j
L = 2T (T V ) = T + V = E (9)
and as promised, F is the energy.
At the fundamental level, the kinetic energy is always quadratic in the velocities, because it is
just a sum over all the parts of the system of
1
2
mv
2
. But what can go wrong with this can be seen
in examples like the bead on a rod rotating with xed angular velocity . Sometimes, we have to
separate the variables that describe the system into those that are dynamical - like the position of
the bead along the rod - and those that are imposed by some external constraint - like the angle
of the rod, which is xed whatever is causing the rod to rotate. Then we include include in our list
of coordinates only the dynamical coordinates and put the effect of the others into the Lagrangian
by hand. Then if our external constraint is causing the motion, some of the velocities that appear
2
in the kinetic energy do not correspond to dynamical coordinates in our Lagrangian, and they are
not included in the sum,

j
q
j

q
j
(10)
In such a case, (9) is not correct, and F is not the energy.
In a system with time translation invariance, however, there can be no time dependent external
constraint. Such a thing would look different at different times and break time translation invari-
ance. Thus time time translation invariance does two things. Not only does it ensure that the
Lagrangian does not depend explicitly on time - which implies that F is conserved. But is also
ensures that the kinetic energy is quadratic in the dynamical velocities, which ensures that F is the
energy. In fact, time translation invariance is the more general answer to the question. Even if L is
not in the form T V , time translation invariance implies that F is the energy.
Symmetries and transformations
What is a symmetry? We have talked about several examples, so perhaps we should dene pre-
cisely what we mean by it. Symmetry is a mathematical statement of some very specic regularity
in a system. A system has a symmetry if there is some transformation you can make that leaves the
system looking exactly as it did before. We tend to regard things with many symmetries as pretty,
like the kalaidoscope that we saw at the beginning of lecture, which has many planes of symmetry.
In the case of mechanics, we have an even more specic meaning in mind. Let us now consider
a class of symmetries in which we make some transformation of the coordinates describing a
system at a xed time. What this means mathematically is that we dene a new set of coordinates
as functions of the original coordinates. The transformation is then a symmetry if the physics looks
exactly the same in terms of the new coordinates as it did in the old coordinates.
We talked briey about such a transformation when we discussed the double pendulum with
two equal masses in lecture 2. The Lagrangian for the double pendulum for small oscillations
looks approximately like
m
2
_
x
2
1
+ x
2
2
_

(x
2
1
+ x
2
2
)
K
2
(x
1
x
2
)
2
(11)
This has the property that it is unchanged if we interchange x
1
and x
2
. This is the mathematical
statement of the obvious physical symmetry of the system.
The symmetry of the double pendulum is an example of a discrete symmetry, so-called because
the symmetry is an all or nothing sort of thing. The transformation cannot be made bigger or
smaller - it is xed by the structure of the symmetry.
It is more even interesting to consider symmetries in which the symmetry transformation can be
made arbitrarily small. Such a thing is called a continuous symmetry, because the transformation
can change the system continuously. By putting arbitrarily small transformations together, we can
get a whole set of transformations which, unlike the symmetry of the double pendulum, depend on
a parameter that can be continuously varied.
3
An example is translations. We think that space probably looks the same everywhere, and we
could describe this by saying that there is a symmetry in which we move everything by the same
arbitrary vector and we would end up with a completely equivalent physical system.
Here is the general theoretical setup (well discuss examples in more detail shortly). Consider
a system of n degrees of freedom described by coordinates q
j
for j = 1 to n. Lets assume
that there is a symmetry in which each of the coordinates changes only a tiny bit, proportional
to an innitesimal parameter, and that the changes involve only the current conguration of the
system. What we mean precisely by innitesimal is that is sufciently small that we can always
ignore terms of order
2
. Translating what we have just assumed into mathematics, we consider a
symmetry in which the coordinates q
j
are transformed as follows:
q
j
q
j
= q
j
+
q
j
(q) . (12)
That is each of the coordinates changes by times a function
q
j
(q) of the qs. The
q
j
(q) tells you
how the variable q
j
changes under the transformation. The transformation (12) is a symmetry of
the Lagrangian if
L( q,

q) = L(q, q) (13)
Example: space translations for one particle
Here is a simple (perhaps even boring) example. Consider a particle with mass m moving along
the x-axis in a potential. When does this system have a symmetry under the innitesimal transfor-
mation
x x = x + ? (14)
This transformation has the form of (12) with (x) = 1. The Lagrangian looks like this:
L(x, x) =
1
2
m x
2
V (x) (15)
The condition that (14) is a symmetry is then
L( x,

x) = L(x, x) (16)
Because (x) is just a constant, we have

x =
d
dt
x = x (17)
and so (16) becomes
L(x + , x) = L(x, x) (18)
for innitesimal . Because the x doesnt change and kinetic energy is the same on both sides, so
this condition only effects V -
V (x + ) = V (x) (19)
4
But because this is supposed to be true for any innitesimal , we can use the Taylor expansion
(surprise, surprise) to rewrite (19) as
V (x + ) = V (x) + V

(x) +O(
2
) = V (x) (20)
If this is to be satised for innitesimal , we must have
V

(x) = 0 (21)
so that V (x) is just a constant and the particle has no force on it at all. In this case, mv is a con-
served momentum. We will see how this connection between symmetry and conserved momentum
generalizes to more complicated (and more interesting) situations.
Another reason that I wanted to look at this simple system in detail is to emphasize the dif-
ference between continuous and discrete symmetries. Suppose that instead of being constant, the
potential in (15) is
V (x) = E
0
cos(x/) (22)
where E
0
and are constants. This system also has a symmetry under space translations of the
form
x x = x + 2n (23)
for any integer n. But here we clearly cannot conclude that V (x) is constant because we started
with an example with the symmetry that is not constant. Except at special points where the particle
is in equilibrium, there is a force on it. There is no conserved momentum (though energy is still
conserved because the Lagrangian does not depend explicitly on t). The difference between this
and the previous example is that this is a discrete symmetry. The changes in x that leave the
Lagrangian invariant are a discrete set. They cannot be varied continuously, and they cannot be
made innitesimally small. Thus we cannot use the Taylor expansion argument to conclude that
V (x) is constant.
Space translations for two particles
Space translation symmetry becomes interesting and important when there is more than one par-
ticle. Let us now consider a one-dimensional system of 2 particles, with positions x
1
and x
2
, so
that q
j
= x
j
for j = 1 to 2. A space translation in the x direction just adds the same innitesimal
constant, , to both x
1
and x
2
so the transformation has the form
x
1
x
1
= x
1
+ , x
2
x
2
= x
2
+ . (24)
Notice that this satises (12), with q
1
= x
1
, q
2
= x
2
and
x
1
(x) =
x
2
(x) = 1. This is a symmetry
of any Lagrangian that depends only on x
1
and x
2
and the difference between x
1
x
2
, for example
L(x, x) =
m
1
2
x
2
1
+
m
2
2
x
2
2
V (x
1
x
2
) (25)
5
The transformation (24) is a symmetry of the kinetic energy because is a constant, so that

x
1
= x
1
,

x
2
= x
2
. (26)
It is a symmetry of the potential energy because the s cancel when we subtract one coordinate
from another, so that
x
1
x
2
= (x
1
+ ) (x
2
+ ) = x
1
x
2
. (27)
Putting (26) and (27) together implies
L( x,

x) = L(x, x) . (28)
This system has a conserved momentum,
p = m
1
x
1
+ m
2
x
2
(29)
because the forces that come from the potential energy obey Newtons third law. This in turn is
related to the fact that the potential energy depends only on x
1
x
2
, which in turn is related to the
symmetry. We will see that this connection between a continuous symmetry and the existence of a
conserved momentum is a general thing.
Space translations for many particles
Let us now consider a one-dimensional system of n particles, with positions x
j
, so that q
j
= x
j
for
j = 1 to n. A space translation in the x direction adds the same innitesimal constant, , to each
x
j
so the transformation has the form
x
j
x
j
= x
j
+ n. (30)
Notice that this satises (12), with q
j
= x
j
and all of the
x
j
(x) = 1. This is a symmetry of any
Lagrangian that depends only on x
j
and differences between two x
j
s, for example
L(x, x) =

j
m
j
2
x
2
j
V (x
1
x
2
, x
2
x
3
, , x
j
x
j+1
, , x
n1
x
n
) . (31)
Again, (30) is a symmetry of the kinetic energy because is a constant, so that

x
j
= x
j
. (32)
It is a symmetry of the potential energy because the s cancel when we subtract one coordinate
from another, so that
x
j
x
j+1
= (x
j
+ ) (x
j+1
+ ) = x
j
x
j+1
. (33)
Here the total momentum,

j
m
j
v
j
(34)
is conserved because the forces that come from the potential energy obey Newtons third law.
6
Finding Symmetries
So far, we have looked at systems in which it is pretty obvious what the symmetry transformation
is. But when come upon some Lagrangian, you may want to nd what transformation the La-
grangian is invariant under. The way to do this is to write down how your Lagrangian trasforms
under a general inntesimal transformation of the form
q
j
q
j
= q
j
+
q
j
(q) . (35)
and then require that it be invariant - that is that the coefcient of the term in
L( q,

q) L(q, q) (36)
vanishes. Here is a simple example, where you could probably guess the answer, but it will illus-
trate the technique.
L(x, x) =
m
1
2
x
2
1
+
m
2
2
x
2
2
V (x
1
+ 2x
2
) (37)
Does this Lagrangian has a symmetry? To see, we want to nd L( x,

x) where
x
j
x
j
= x
j
+
x
j
(x) . (38)
L( x,

x) =
m
1
2
_
x
1
+
x
1
_
2
+
m
2
2
_
x
2
+
x
2
_
2
V (x
1
+
x
1
+ 2x
2
+ 2
x
2
) (39)
=
m
1
2
_
x
2
1
+ 2x
1

x
1
+
_
+
m
2
2
_
x
2
2
+ 2x
2

x
2
+
_
V
_
x
1
+2x
2
+(
x
1
+2
x
2
)
_
(40)
The term will vanish if

x
1
=
x
2
=
x
1
+ 2
x
2
= 0 (41)
One way to satisfy the rst two conditions is to take the s to be constant independent of x (if they
depend on x, the time derivatives would be nonzero). Then the last condition implies that we can
take

x
1
= 2
x
2
= 1 (42)
which gives a transformation of the form
x
1
x
1
2 x
2
x
2
+ (43)
You see that it is sometimes easy to nd symmetry transformations. It is harder to show that you
have found them all. This depends on details, in this case the precise form of the function V . We
wont talk about that now.
7
Rotations
Another simple example can be found in the example we discussed in lecture 7 of the mass m
1
sliding on a horizontal frictionless table, connected to a string that goes through a hole at the origin
to a mass m
2
hanging below the table. In terms of the length, , of string on the table and the angle,
, of the string on the table from the x axis, the Lagrangian for this system looks like
L(, ,

,

) =
1
2
(m
1
+ m
2
)

2
+
1
2
m
1

2
m
2
g (44)
This physical system has a symmetry under rotations about the z axis, which add a constant to the
angle without changing ,


= ,

= + . (45)
This symmetry is responsible for the fact that the Lagrangian does not depend on at all, but only
on

, because the condition for symmetry, that
L(

,

,

) = L(, ,

,

) (46)
in this case becomes
L(, + ,

,

) = L(, ,

,

) (47)
This just says that if we make a little change in in the function, nothing happens, so the function
must not depend on . If we wanted to say the same thing in fancier mathematics, we could say
that because is innitesimal, we can reliably Taylor expand the left-hand-side of (47) and keep
only the rst two terms,
L(, + ,

,

) = L(, ,

,

) +

L(, ,

,

) (48)
Putting (48) into (47) implies

L(, ,

,

) = 0 (49)
which as promised is the statement that L does not depend on .
As we saw in lecture 7, this system also has a conserved quantity, the angular momentum
L = m
1

2

. (50)
So again, we see a connection between a symmetry and a conserved quantity.
Noethers theorem
Having seen a correlation between symmetry and conservation laws in a couple of examples, let
us now consider how this works in more generality and see if we can pin down the connection
precisely. Putting (12) into (13) and Taylor expanding gives
L(q, q) = L(q + , q + ) = L(q, q) +

q
j

q
j
L
q
j
+

q
j

q
j
L
q
j
. (51)
8
Thus the condition that (12) is a symmetry of the Lagrangian is equivalent to

q
j

q
j
L
q
j
+

q
j

q
j
L
q
j
= 0 (52)
For example, for space translations, (30),

x
j
= 1 (53)
for all j, because all the x
j
s are translated in the same way. Thus = 0 and the condition of
symmetry becomes

x
j

x
j
L
x
j
=

x
j
L
x
j
= 0 (54)
This is equivalent to the statement that L depends only on differences x
j
x
k
.
Now return to the general case and consider the quantity

q
j

q
j
L
q
j
(55)
Consider the time derivative of this quantity
d
dt

q
j
_

q
j
L
q
j
_
=

q
j

q
j
L
q
j
+

q
j

q
j
d
dt
L
q
j
(56)
If we apply the Lagrange equations of motion to the second term, it becomes
d
dt

q
j
_

q
j
L
q
j
_
=

q
j

q
j
L
q
j
+

q
j

q
j
L
q
j
= 0 (57)
which vanishes because of the condition (52) that the Lagrangian is symmetric. This is a very
important general theorem. It is the precise connection between a continuous symmetry and a
conservation law. For every continuous symmetry of the form (12), there is a conservation law
the quantity of the form (55) is constant.
Now we can nd the conserved quantity for any symmetry. For example for the Lagrangian of
(37), where the s are given by (42), the conserved quantity is

x
1
L
x
1
+
x
2
L
x
2
= 2m
1
x
1
+ m
2
x
2
(58)
This theorem is a special favorite of mine because a lot of my own work in particle physics
is based on it. It was worked out early in this century by the great woman mathematician and
theoretical physicist Emmy Noether.
1
The quantity
L
q
j
(59)
1
For more information about Noether, see The Life and Times of Emmy Noether by Nina Byers
http://xxx.lanl.gov/abs/hep-th/9411110 and on the 16 website in handouts/noether.pdf.
9
that appears in Noethers theorem is a very important one in classical mechanics. It is the gen-
eralized momentum we talked about earlier, corresponding to the coordinate q
j
, we discussed in
the last lecture. Of course, if q
j
is a normal space coordinate (like x
j
), and the kinetic energy has
the standard form, it is the momentum (or a component of it). In fact, one way of thinking about
(57) is to recognize that in a sense (which I will not explain in detail) (55) is the generalized mo-
mentum associated with, , the innitesimal variable that describes how all the q
j
s change under
the symmetry.
Momentum conservation from Noethers theorem
For space translation symmetry, because all the
x
j
are equal to 1, the quantity (55) becomes

x
j

x
j
L
x
j
=

x
j
L
x
j
(60)
which is just the sum over the momenta of all the individual particles. Space translation invari-
ance implies that the total momentum is conserved.
This analysis can obviously be extended to three dimensions, where the coordinates and the
corresponding momenta become vectors.
More on rotations
Here is a more involved example that we can discuss if we have time. When we thought about
rotations for the bead sliding on the table, we were already using polar coordinates. This makes
rotations easy, because an innitesimal rotation is just a translation of ,

= + . (61)
This is an example with two particles in Cartesian coordinates. Consider two particles, with masses
m
1
and m
2
, which move in a plane with coordinates r
1
= (x
1
, y
1
) and r
2
= (x
2
, y
2
). Consider the
following Lagrangian:
1
2
m
1
_
x
2
1
+ y
2
1
_
+
1
2
m
2
_
x
2
2
+ y
2
2
_
V
_
x
2
1
+ y
2
1
+ x
2
2
+ y
2
2
2(x
1
x
2
+ y
1
y
2
)
_
(62)
This Lagrangian is invariant under the following innitesimal transformation:
x
1
x
1
= x
1
y
1
y
1
y
1
= y
1
+ x
1
x
2
x
2
= x
2
y
2
y
2
y
2
= y
2
+ x
2
(63)
If you have never seen anything like this before, it is not obvious at all, but you can see it by explicit
calculation. For example, look at the m
1
terms in the kinetic energy. First note that
x
1
x
1
= x
1
y
1
y
1
y
1
= y
1
+ x
1
(64)
10
then substitute

x
2
1
+

y
2
1
=
_
x
1
y
1
_
2
+
_
y
1
+ x
1
_
2
= x
2
1
+ y
2
1
+O(
2
) (65)
The
2
terms do not cancel, but we dont care about them because is innitesimal - we only
consider linear terms in Noethers theorem. So this is good enough. All the other terms work
similarly. You can see what the rotation looks like pictorially in the gure below.
........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................ . . . . . . . . . . . . .
. . . . . . . . . . . . . .
.....................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
. . . . . . . . . . . . . .
... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.

( x, y)

. . . . . . . . . . . . . . . . . . . . . . . . ..........................
......................................................................................................................................................................................................................................................................................................................................................................................................................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .

(x, y)
The heavy lines represent the vector r = (x, y) from the origin to the original point, and the smaller
vector from r to

r = ( x, y). Because is innitesimal, these two are nearly perpendicular, and the
11
ratio of their lengths is about . It is helpful to complete these two lines into a similar triangle.
........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................ . . . . . . . . . . . . .
. . . . . . . . . . . . . .
.....................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
. . . . . . . . . . . . . .
... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.

( x, y)

. . . . . . . . . . . . . . . . . . . . . . . . ..........................
......................................................................................................................................................................................................................................................................................................................................................................................................................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .

(x, y)
......................................................... . . . . . . . . . . . . . . . . . . . . . . . . . .
.......................................................................................................................................................................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...........................................................................................................................................................................................................................................................................................................................................................................................................................................
Then blowing up the box, you can see that the change in x under the rotation is y and the change
in y is x.
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
...

x
2
+ y
2
..............................................................................................................................................................................................................................................................................................................................................................................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..................................................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.............................................................................................................................................................................................................................. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
y
x

Now lets see what Noethers theorem looks like. The s are

x
1
= y
1
,
y
1
= x
1
,
x
2
= y
2
,
y
2
= x
2
. (66)
and the momenta are
L
x
j
= m
j
x
j
,
L
y
j
= m
j
y
j
. (67)
12
Thus the conserved quantity is

j
m
j
(x
j
y
j
y
j
x
j
) (68)
As we will see in more detail later in the course, this is the angular momentum.
Notice also that the complicated function of the variables that appears in V can be written as
(r
1
r
2
) (r
1
r
2
) = |r
1
r
2
|
2
(69)
just the square of the length of the difference r
1
r
2
.
What functions are invariant?
We have seen some examples of functions that are invariant under transformations, such as the
potential in (31),
V (x
1
x
2
, x
2
x
3
, , x
j
x
j+1
, , x
n1
x
n
) . (70)
It always straightforward to check that a given function is invariant under a specic transforma-
tion. And we have seen how to nd transformations that are symmetries, if they exist. It is also
important to be able to go the other way and to construct the most general function invariant under
a transformation or set of transformations.
To illustrate what I am talking about, lets show that a function of F(x
1
, x
2
) that is invariant
under the symmetry transformation
x
j
x
j
+ (71)
for j = 1 and 2 actually only depends on the difference, x
1
x
2
. One way to do this is to change
variables to include the variable x
1
x
2
, and eliminate x
1

y x
1
x
2
x
1
= x
2
+ y (72)
Then we can dene a new function
G(y, x
2
) F(x
2
+ y, x
2
) (73)
in terms of the original function. But now, in terms of the new variables the transformation (71) is
y y , x
2
x
2
+ . (74)
Now since F is invariant, G must be also, because we have only relabeled things. Thus
G(y, x
2
+ ) = G(y, x
2
) (75)
for innitesimal . In words, this says that making an innitesimal change in x
2
doesnt affect the
fucntion, so it is probably obvious that this means that G(y, x
2
) is independent of x
2
. But if we
want to be more formal about it, we can use the Taylor expansion to get
G(y, x
2
+ ) = G(y, x
2
) +

x
2
G(y, x
2
) + = G(y, x
2
) (76)
13
and thus

x
2
G(y, x
2
) = 0 (77)
which means that G doesnt depend on x
2
, so we can take x
2
to be anything in G. Thus using (72)
F(x
1
, x
2
) = G(x
1
x
2
, x
2
) = G(x
1
x
2
, 0) = f(x
1
x
2
) . (78)
It is easy to extend this proof to show that the most general function invariant under the transfor-
mation (71) for n variables, x
j
for j = 1 to n is given by (70).
Notice that what we are doing here makes Noethers theorem seem a little trivial. If all we
do with invariance is to show that there is some variable that the function doesnt depend on,
then we could have changed variables rst and then found the conserved quantity by just using
the statement that the generalized momentum associated with a variable that doesnt appear in
the Lagrangian is conserved. And in fact, in this course, you can always do that. But there is
actually more to Noethers theorem, because transformations that leave a system invariant have an
additional interesting property. They form what mathematicians form a group. The group property
is quite powerful and often allows you to extend the innitesimal transformations that we start with
to a much larger set. For example, in the case of translations, group theory can be used to show
that if things are invariant under (71) for innitesimal , they are also invariant for nite . Then
we dont need the Taylor expansion any more. For example, in (75), we could rst set x
2
to zero
to get
G(y, ) = G(y, 0) (79)
and then relabel x
2
to get
G(y, x
2
) = G(y, 0) (80)
which directly gives (78). We wont pursue this approach much in the course but I may not be able
to restrain myself from talking more about it, because it forms the basis of much of the research
that I have done in my scientic career.
Example of functionals - Soap bubbles
I thought that it would be interesting to discuss further the pretty mathematics of calculus of vari-
ations in a slightly different context. I hope this may give you a better feel for it. There are many
examples of the use of the calculus of variations. One nice one is to the shape of soap bubbles. The
connection here is that soap bubbles have a surface tension, so they want to minimize their surface
area. But the surface area is a functional of the function that describes the shape of the bubble. So
this is a job for the calculus of variations. Heres a specic problem that is kind of fun. Consider a
bubble formed between two circular loops of wire, both centered on the z axis, one in the z = z
1
plane with radius r
1
, and the other in the z = z
2
plane with radius r
2
. Well assume that z
1
< z
2
.
14
So it looks something like this:
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......................................................................................................................................................................................................................................................................................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ............................................................................................................................................
............................................... . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...............
..............
............... . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . ...............
..............
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
. . . . . . . . . . . . .
z
z
1
z
2
r
1
r
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(81)
The soap bubble will be some surface of revolution, because of the cylindrical symmetry of the
system , indicated by the dotted line in (81). Thus we can specify the shape of the bubble by giving
the radius as a function of z, r(z), subject to the constraint
r(z
1
) = r
1
r(z
2
) = r
2
(82)
which is just the physical requirement that the bubble is attached to the frame but it looks just
like the condition we impose on trajectories in Hamiltons principle. Thus we can compute the
shape exactly by solving an Euler-Lagrange equation.
We want to minimize the area of the surface of revolution described by r(z) from z = z
1
to
z = z
2
subject to the constraint (82). First we must compute the area. Consider the area of a small
band of surface between z and z + dz. The circumference of the band is approximately 2r(z)
(the difference between r(z) and r(z + dz) does not matter here). The width of the band can be
computed by looking at a slice through the z axis, which looks like this:
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .....................................................................................................................................................................................................................................................................................................................................................................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........................................................................................................................................................................................................................................................................................................................................................................................................................................
z
z + dz
..................
..................
r(z)
......................................
r(z + dz)
r(z) + r

(z) dz
......................................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.................................................................................
.............
.............
.............
.............
.............
.............
.............
.............
.............
.............
. . . . . . . . . . . . .
z
(83)
Evidently,the width of the slice is approximately

dz
2
+ dr
2
= dz
_
1 + r

(z)
2
(84)
15
Thus the area of the slice is
dz
_
r(z), r

(z)
_
where (r, r

) 2r

1 + r
2
(85)
and the total area is
A[r] =
_
z
2
z
1

_
r(z), r

(z)
_
dz = 2
_
z
2
z
1
r(z)
_
1 + r

(z)
2
dz (86)
Now we can use the calculus of variations. The mathematics of this problem is the same as in
nding the vanishing variation the action, S[x], so the condition is the analog of the Lagrange
equation,
A[r]
r(z)
= 0 (87)
We can do the functional differentiations in the same way as above to the analog of the Lagrange
equation

r
(r, r

)
d
dz

(r, r

) = 0 (88)
But rather than trying to solve the Euler-Lagrange equation directly, it is useful to extend the
analogy with Lagrangian mechanics and notice that because does not depend explicitly on z,
there will be an analog of energy that is independent of z:
q
L
q
L r

(89)
r

= r

_
2r
_
1 + r

2
_
2r
_
1 + r

2
(90)
= r

2r
r

1 + r

2
2r
_
1 + r

2
(91)
= 2r
_
r

1 + r

2

1 + r

1 + r

2
_
=
2r

1 + r

2
(92)
r

1 + r

2
= r
0
for some constant r
0
(93)
We could mess around with this but it easier to guess the answer.
r(z) = r
0
cosh
zb
r
0
(94)
r

(z) = r
0
1
r
0
sinh
zb
r
0
= sinh
zb
r
0
(95)
Now the key step that might actually cause you to make this guess is the next one.
1 + r

(z)
2
= 1 + sinh
2 zb
r
0
= cosh
2 zb
r
0
(96)
r

1 + r

2
=
r
0
cosh
zb
r
0
cosh
zb
r
0
= r
0
(97)
16
In the demo I will show you, both rings have the same radius, r
1
= r
2
= R. Then we dene the
z coordinate system so that z = 0 is right in the middle. Then the fact that cosh(x) = cosh(x)
implies that b = 0 so that
r(z) = r
0
cosh(z/r
0
) (98)
and the situation looks like the diagram below (the red lines represent a side view of the rings and
the blue curves are a slice through the bubble):
................................................................................................................................................................................................................................................
. . . . . . . . . . . . . .
r(z)
..................................................................... . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...............
..............
......................................................... . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...............
.............. r
0
z
Z
....................................................
. . . . . . . . . . . . . .
....................................... . . . . . . . . . . . . .
..............
.....................................................................................................................
. . . . . . . . . . . . . .
....................................... . . . . . . . . . . . . .
..............
............................................................................................................................... . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...............
..............
R
r(z) = r
0
cosh z/r
0
1 = R = r(Z) = r
0
cosh Z/r
0
solve for r
0
in terms of Z (units of R)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
One can analyze this analytically, but it is complicated, and since we have Mathematica around,
we may as well just nd an numerical solution for r
0
as a function of the ring separation Z for
xed ring radius R. This is shown below and in the notebook soapbubble.nb.
0.0 0.1 0.2 0.3 0.4 0.5 0.6
0.6
0.7
0.8
0.9
1.0
(99)
Note that something happens if you try to make Z/R big. The maximum is 0.6627 R, for which
17
the bubble looks something like the blue curves:
For large Z/R (as you can see analytically in the appendix), there is no real solution for R/r
0
,
which means that we cannot actually minimize the area in this way. Physically, what is going on
is that for large z it is always energetically favorable to just keep narrowing in the center until
the bubble actually breaks into two separate bubbles on the two rings. We should be able to see
this in the demo. You can also get a sense of what is happening by looking at soapbubble.nb,
increasing Z/R to its maximum value, and then clicking on color 0, which toggles diagonal lines
that represent a cross-section of the cone through the two circles. The system is unstable when the
the bubble gets beyond that cone.
A Analytic treatment of the bubble
Now
R/r
0
= cosh(Z/r
0
) = cosh
_
(R/r
0
)(Z/R)
_
(100)
is a transcendental equation that can be solved numerically for aR in terms of Z/R. Because, from
(98),
r(0) = r
0
(101)
the physical interpretation of r
0
is that it is the radius of the bubble at its narrowest point, in the
center. When we are nished messing with the math, we will express things in term of r
0
/R,
because that is easier to think about physically.
It is actually much easier to solve for Z/R in terms of R/r
0
, because (100) can be rewritten as
18
a quadratic equation for e
(R/r
0
)(Z/R)

R/r
0
= cosh((R/r
0
)(Z/R)) =
_
e
(R/r
0
)(Z/R))
+ e
(R/r
0
)(Z/R))
_
/2
e
2(R/r
0
)(Z/R)
2R/r
0
e
(R/r
0
)(Z/R)
+ 1 = 0
(102)
with solution
e
(R/r
0
)(Z/R)
= R/r
0
+
_
(R/r
0
)
2
1 (103)
or
Z/R =
ln
_
R/r
0
+
_
(R/r
0
)
2
1
_
R/r
0
(104)
Expressing this in terms of r
0
we have
Z
R
=
r
0
R
ln
_
R/r
0
+
_
(R/r
0
)
2
1
_
(105)
This shows a plot of Z/R versus r
0
/R
0.2 0.4 0.6 0.8 1
0.1
0.2
0.3
0.4
0.5
0.6
Z/R
r
0
/R
The rst thing to notice about thus graph is that there are two possible values of r
0
/R for each
values of Z/R. Physically, this doesnt make sense. We can understand what is going by looking at
the ends of the r
0
/R axis on the rst graph. The rightmost point on the graph corresponds to Z = 0
and r
0
= R which is the right solution because in the limit of very small Z the area is obviously
minimized by an almost at ribbon between to two rings. Conversely, the leftmost point, z = 0
and r
0
= 0 is obviously not a physical solution. While the variation vanishes there, the area is
not minimized. The leftmost point is some kind of saddle point. Now as we go up along the Z/R
axis, the same situation obtains. The point on the right corresponds to the minimum. Thus the
physical region of the rst graph is only the part to the right of the maximum, which occurs for
r
0
= r
min
0.5524 R.
19
Now that we know what is happening, we can display just the physical solution by plotting
r
0
/R versus Z/R from r
min
/R to 1.
0.1 0.2 0.3 0.4 0.5 0.6
0.6
0.7
0.8
0.9
1
r
0
/R
Z/R
This is the analytic version of (99).
20