Mechanics

Physics 151
Lecture 22
Canonical Transformations
(Chapter 9)
What We Did Last Time
Direct Conditions
Necessary and sufficient
for Canonical Transf.
Infinitesimal CT
Poisson Bracket
Canonical invariant
Fundamental PB
ICT expressed by
Infinitesimal time transf. generated by Hamiltonian
Hamiltons equations
,
,
j
i
j i
Q P
q p
p
Q
q P

=





,
,
j
i
j i
Q P
q p
q
Q
p P

=






,
,
j
i
j i
Q P
q p
p
P
q Q

=





,
,
j
i
j i
Q P
q p
q
P
p Q

=






[ ]
,
i i i i
u v u v
u v
q p p q



[ , ] [ , ] 0
i j i j
q q p p = = [ , ] [ , ]
i j i j ij
q p p q = =
[ , ]
u
u u G t
t

= +

Two Points of View

Canonical Transformation allows one system to be
described by multiple sets of coordinates/momenta
Same physical system is expressed in different phase spaces
This is the static view The system itself is unaffected
q
p
Q
P
Is there a dynamic view?
Dynamic View of CT
A system evolves with time
At any moment, q and p satisfy Hamiltons equations
The time-evolution must be a Canonical Transformation!
Static View = Coordinate system is changing
Dynamic View = Physical system is moving
0 0
( ), ( ) q t p t ( ), ( ) q t p t
q
p
This movement
is a CT
Infinitesimal Time CT
Infinitesimal CT
We know that the generator = Hamiltonian
Integrating it with time should give us the finite CT that
turns the initial conditions q(t
0
), p(t
0
) into the configuration
q(t), p(t) of the system at arbitrary time
Thats a new definition of solving the problem
( ), ( ) q t p t ( ), ( ) q t dt p t dt + +
[ , ]
u
du dt u H dt
t

= +

[ , ] q q H =

[ , ] p p H =

Hamiltonian is the generator of the systems

motion with time
Hamiltonian is the generator of the systems
motion with time
Static vs. Dynamic
Two ways of looking at the same thing
System is moving in a fixed phase space
Hamiltons equationsIntegrate to get q(t), p(t)
System is fixed and the phase space is transforming
ICT given by the PB Integrate to get CT for finite t
Equations are identical
Youll find yourself integrating exactly the same equations
Did we gain anything?
Conservation
Consider an ICT generated by G
Suppose G is conserved and
has no explicit t-dependence
How is H (without t-dependence) changed by the ICT?
A transformation that does not affect H
Symmetry of the system
Generator of the transformation is conserved
[ , ]
u
u u G t
t

= +

[ , ] 0 G H =
[ , ] 0
H
H H G t
t

= + =

If an ICT does not affect

Hamiltonian, its generator
is conserved
Momentum Conservation
Simplest example:
What is the ICT generated by momentum p
i
?
Thats a shift in q
i
by spatial translation
If Hamiltonian is unchanged by such shift, then
Momentum p
i
is conserved
This is not restricted to linear momentum
[ , ]
j j i ij
q q p = = [ , ] 0
j j i
p p p = =
[ , ] 0
i
H p =
Hamiltonian is
unchanged by
a shift of a
coordinate q
The generator
of the ICT is
the conjugate
momentum p
p is
conserved
[ , ] 0 H p =
Angular Momentum
Lets consider a specific case: Angular momentum
Pick x-y-z system with z being the axis of rotation
n particles positions given by
Rotate all particles CCW around z axis by d
Momenta are rotated as well
Generator is
( , , )
i i i
x y z
i i i
x x y d

=
i i i
y y x d

= +
d
( , ) x y
( , ) x y

ix ix iy
p p p d

=
iy iy ix
p p p d

= +
i iy i ix
G x p y p =
[ , ]
i i
ix
G
d x G d y d
p

= =

[ , ]
ix iy
i
G
d p G d p d
x

= =

etc.
Angular Momentum
The generator is obviously
i.e. the z-component of the total momentum
Generator for rotation about an axis given by a unit vector n
should be
We now know generators of 3 important ICTs
Hamiltoniangenerates displacement in time
Linear momentumgenerates displacement in space
Angular momentumgenerates rotation in space
i iy i ix
G x p y p = ( )
z i i z
L = r p
G = L n
Integrating ICT
I said we can integrate ICT to get finite CT
How do we integrate ?
First, lets rewrite it as
We want the solution u() as a function
of , with the initial condition u(0) = u
0
Taylor expand u() from = 0
[ , ] u u G =
[ , ] du d u G =
[ , ]
du
u G
d
=
2 2 3 3
0
2 3
0
0 0
( )
2! 3!
du d u d u
u u
d d d



= + + + +
This is [u,G]
0
What can I do with these?
Integrating ICT
Since is true for any u, we can say
Now apply this operator repeatedly
Going back to the Taylor expansion,
Now we have a formal solution But does it work?
[ , ]
du
u G
d
= [, ]
d
G
d
=
2
2
[ , ] [[ , ], ]
d u d
u G u G G
d d
= = [ [[ , ], ], , ]
j
j
d u
u G G G
d
=
2 2 3 3
0
2 3
0
0 0
2 3
0 0 0 0
( )
2! 3!
[ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
du d u d u
u u
d d d
u u G u G G u G G G

= + + + +
= + + + +

Rotation CT
Lets integrate the ICT for rotation around z
Let me forget the particle index i
Parameter is in this case
Lets see how x changes with
Evaluate the Poisson Brackets
Where does this lead us?
y x
G xp yp =
2 3
0 0 0 0
( ) [ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
x x x G x G G x G G G

= + + + +
[ , ] x G y = [[ , ], ] x G G x = [[[ , ], ], ] x G G G y =
[[[[ , ], ], ], ] x G G G G x =
Repeats after this
Rotation CT
Similarly
2 3
0 0 0 0
2 3 4
0 0 0 0 0
2 4 3 5
0 0
0 0
( ) [ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
2! 3! 4!
1
2! 4! 3! 5!
cos sin
x x x G x G G x G G G
x y x y x
x y
x y

= + + + +
= + +

= + +


=

2 3
0 0 0 0
0 0
( ) [ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
cos sin
y y y G y G G y G G G
y x



= + + + +
= +

Free Fall
An object is falling under gravity
Hamiltonian is
Integrate the time ICT
2
2
p
H mgz
m
= +
z
2 3
0 0 0 0
( ) [ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
t t
z t z t z H z H H z H H H = + + + +
[ , ]
p
z H
m
=
[[ , ], ] z H H g = [[[ , ], ], ] 0 z H H H =
2
0
0
( )
2
p g
z t z t t
m
= +
Its easier than it looked
Infinitesimal Rotation
ICT for rotation is generated by
Weve studied infinitesimal rotation in Lecture 8
Infinitesimal rotation of d about n moves a vector r as
Compare the two expressions
Equation holds for any r that rotates
together with the system
Several useful rules can be derived from this
G = L n
d d = r n r
[ , ] d d d = = r r L n n r [ , ] = r L n n r
[ , ] = r L n n r
Scalar Products
Consider a scalar product of two vectors
Try to rotate it
Obvious: scalar product doesnt change by rotation
Also obvious: length of any vector is conserved
a b
[ , ] [ , ] [ , ]
( ) ( )
( ) ( )
0
= +
= +
= +
=
a b L n a b L n b a L n
a n b b n a
a n b a b n
[ , ] = r L n n r
Angular Momentum
Try with L itself
x-y-z components are
These relationships are well-known in QM
They tell us two rather interesting things
[ , ] = L L n n L
[ , ] 0
x x
L L = [ , ]
x y z
L L L = [ , ]
x z y
L L L =
[ , ]
y x z
L L L = [ , ] 0
y y
L L = [ , ]
y z x
L L L =
[ , ]
z x y
L L L = [ , ]
z y x
L L L = [ , ] 0
z z
L L =
[ , ]
i j ijk k
L L L =
Angular Momentum
Imagine two conserved quantities A and B
How does [A,B] change with time?
Poisson bracket of two conserved quantities is conserved
Now consider
If 2 components of L are conserved, the 3
rd
component must
Total vector L is conserved
[ , ] [ , ] 0 A H B H = =
[[ , ], ] [[ , ], ] [[ , ], ] 0 A B H B H A H A B = =
J acobisidentity
[ , ]
i j ijk k
L L L =
Angular Momentum
Remember the Fundamental Poisson Brackets?
Now we know
Poisson brackets between L
x
, L
y
, L
z
are non-zero
On the other hand, , so |L| may be a canonical
momentum
QM: You may measure |L| and, e.g., L
z
simultaneously, but
not L
x
and L
y
, etc.
[ , ] [ , ] 0
i j i j
q q p p = = [ , ] [ , ]
i j i j ij
q p p q = =
PB of two canonical momenta is 0
[ , ]
i j ijk k
L L L =
2
[ , ] 0
i
L L =
Only 1 of the 3 components of the angular momentum
can be a canonical momentum
Phase Volume
Static view: CT moves a point in one phase space to a point
in another phase space
Dynamic view: CT moves a point in one phase space to
another point in the same space
If you consider a set of points, CT moves a volume to
anther volume, e.g.
q
p
Q
P
dq
dp
How does the area change?
Phase Volume
Easy to calculate the J acobianfor 1-dimension
i.e., volume in 1-dim. phase space is invariant
This is true for n-dimensions
Goldstein proves it using simplectic approach
dQdP dqdp = M
Q q Q p
P q P p


=



M
where
[ , ] 1
Q P P Q
Q P
q p q p

= = =

M dQdP dqdp =
Volume in Phase Space is a Canonical Invariant
Harmonic Oscillator
Weve seen it in the oscillator example (Lecture 21)
One cycle draws the same area in both spaces
Thats static view
p
q
2
2E
m
2mE
Q
P
E

2
2 E

Dynamic View
Consider many particles moving independently
e.g., ideal gas molecules in a box
They obey the same EoM independently
Can be represented by multiple points
in one phase space
They move with time CT
q
p
q
p
Time
Ideal Gas Dynamics
Imagine ideal gas in a cylinder with movable piston
Each molecule has its own position
and momentum They fill up a
certain volume in the phase space
What happens when we compress it?
q
p
q
p
Compress
slowly
Extra momenta
Gas gets hotter!
Liouvilles Theorem
The phase volume occupied by a group of particles
(ensemble in stat. mech.) is conserved
Thus the density in phase spaceremains constant with time
Known as Liouvilles theorem
Theoretical basis of the 2
nd
law of thermodynamics
This holds true when there are large enough number of
particles so that the distribution may be considered
continuous
More about this in Physics 181
Summary
Introduced dynamic view of Canonical Transf.
Hamiltonian is the generator of the motion with time
Symmetry of the system Hamiltonian unaffected by the
generator Generator is conserved
How to integrate infinitesimal transformations
Discussed infinitesimal rotation
Angular momentum QM
Invariance of the phase volume
Liouvilles theorem Stat. Mech.
2 3
0 0 0 0
( ) [ , ] [[ , ], ] [[[ , ], ], ]
2! 3!
u u u G u G G u G G G

= + + + +
[ , ] = r L n n r