Professional Documents
Culture Documents
If the system is subjected to constrains then the 3N Cartesian coordinates are not
independent variables. If n is the least number of variables necessary to specify the
most general motion of the system, then the system is said to have n degrees of
freedom.
The configuration of a system with n degrees of freedom is fully
specified by n generalized position-coordinates q1, q2, … , qn
Typically L=T-V, where T is the kinetic energy and V is the potential energy of
the system. For example, for a particle of mass m moving in a potential V(x,t), the
Lagrangian is,
L (x, x , t) = (1/2)m x 2 - V(x,t) (3)
Then, for a couple of fixed end points (a,t1) and (b, t2) the classical action S is
defined as,
( b, t 2 )
Notice S varies depending on the arbitrarily selected path q that joins the ending
point a and b, at the corresponding times t1 and t2. See figure 1 below.
That is, out of all possible paths by which the system could travel from an initial position
at time t1 to a final position at time t2, it will actually travel along the path for which the
integral is an extremum, whether a maximum or a minimum. The figure below shows
the case for the one-dimensional motion.
x(t)
xC
b
xC(t)
a
t1 t2 t
Figure 1. For a fixed points (a, t1) and (b, t2), among all the
possible paths with the same end points, the path xC makes
the action S an extremum.
Let’s assume xC is the path that makes the value of S extremum. One way to obtain
a more explicit form of xC is to probe expression (6) with a family of trial paths that are
infinitesimally neighbors to that particular path. Let’s try, for example,
x( = xC + h (7)
Function
Scalar
(parameter)
which means
x(( t) = xC(t) + h(t) (7)’
where h(t) is an arbitrary function subjected to the condition,
h(t1)= h(t2) =0 (8)
and is an arbitrary scalar parameter,
In expression (7), when choosing small values for , x( becomes a path just a
“bit” different than the path xC.
Note: In the literature the path difference [x( - xC ] is sometimes called x.
Here we are opting to use h ( a number, and h a function) instead, just
to emphasize that [x(-xC] is a difference between two paths (and not
the difference of just two numbers).
x(t)
b xC(t)
h(t)
t1 t2 t
Figure 2. For an arbitrary function h, a parametric family of trial paths x(t ) =
xC(t ) + h(t ) is used to probe the action S given by expression (6).
From (6),
( b, t 2 )
a, t
dS
{
1
2
m 2 [ x C(t) + h (t)] h (t) –V ’h(t) }dt
dα ( 1)
d 0
a, t
( 1)
{ m x C (t) h (t) – V ’(xC) h(t) }dt (11)
The first term on the right side of the equality above can be integrated
by parts,
( b, t 2 ) ( b, t 2 )
a, t a, t
t2
(
x C (t) h (t) dt = x C (t) h(t ) t - x C (t) h(t) dt
1) 1 ( 1)
dS ( b, t 2 )
– { m x C (t) + V ’(xC)} h(t)dt (12)
d 0 ( a, t 1 )
Since this last expression has to be equal to zero for any function h(t), it must happen
that m x C (t) + V ’(xC) =0, or,
d
t1
{
i
L
xi
hi } dt +
i xi
hi -
t1
{ (
i
d L
dt xi
)hi } dt
t1
Since h (t1) = h (t2) = 0, one obtains
dS t2 t2
d
t1
{
i
L
xi
hi } dt - t1
{ (
i
d L
dt xi
)hi } dt
t
t2
L d L
= { -( ) } hi dt
i 1 xi dt xi
L d L
In the last expression and ( ) are evaluated at x (t) + h(t ) .
xi dt xi
dS L d L
t
t2
{ -( ) } hi (t ) dt
d 0 i 1 xi dt xi
L d L
In the last expression and ( ) are evaluated at x (t).
xi dt xi
d L L
0 for i=1,2, …, n. The Lagrange Equations (18)
dt xi xi
These are second order equations. The motion is completely specified if the initial
values of the n coordinates xi and the n velocities xi are specified. That is, the xi and xi
form a complete set of 2n independent variables for describing the motion.
Remark:
Hamilton’s variational principle involves physical quantities T
and V, which can be defined without reference to a particular
set of generalized coordinates. The set of Lagrange equations is
therefore invariant with respect to the choice of coordinates. !
Accordingly, the vector expression (22) can be expressed in term of the components,
m A A
x + x ( x ) A = 1, 2, 3
q x t x
The second and the fourth terms on the right side constitute a
exact differential of A
A dA
+ x (24)
x x dt
We have to figure out the Lagrangian L, such that the Lagrangian equations (18) lead to
(24). If there were no magnetic field, L would be L (1/2)m x q( x ) . The magnetic
2
L f
mx
x x
d L f f f
mx ( x ) ( x )
dt x x x x x t x
where stands for ( , , )
x x1 x 2 x3
stands for ( , , )
x x1 x 2 x 3
d L L
0 implies
dt x x
f f f f
mx q ( x ) ( x ) (26)
x x x x x x t x
which should be compared with (24)
A dA
m x q + qx q
x x dt
Notice the right side of (26) contains a term in x , which is not present in (24). Therefore
we require
f
( x ) =0
x x
Or, equivalently,
f
for = 1, 2, 3 (27)
x x
1
L( x , x ,t ) m x 2 q ( x ) q x A( x ,t ) (31)
2
Lagrangian for a particle in an electro-
magnetic field described by a scalar
potential and vector potential A.
A
E
t
B A
L d L
From the Lagrange equation , the previous equation can be written as
xi dt xi
dL d L dxi L dx i
dt
[ i
( )
dt xi dt
+
x i dt
]
d L dxi L dx i
i
[ ( )
dt xi dt
+
x i dt
]
L
i
(
d
dt
( xi )
xi
L
d
dt
i
xi
xi
This implies,
L
i
xi
xi
L = constant (34)
A more rigorous definition of the Hamiltonian function will be given in the following
sections.
Expression (34) indicates that, when L does not depend explicitly on time, the
Hamiltonian quantity H is a constant of motion.
=
k
1
2
mk x k
2
V(x) (36)
L
= mi xi
x i
L
i
xi
xi
= i
mi xi
2
= 2 T (x)
L
H ≡
i
xi
xi
L
= 2 T ( x ) L
= 2 T ( x ) ( T( x ) V(x) )
= T( x ) V(x) = Energy (37)
Thus, when the potential does not depend on the the velocity, and L does not depend
on the time explicitly, H is the energy of the system and at the same time a constant of
motion.
L(x, y) B(x, v)
(38)
B(x, v) ≡ L(x, y) – y v
= L(x, y) – y L/ y
dB = u dx – y dv
With this background, we will see below that the transformation to transit from the
Lagrangian formulation to the Hamiltonian formulation is of the Legendre’s type. (Just
identify the x variable with the y variable we have used in this section.)
L H
H ( x, p, t ) x i pi - L( x, x , t ) (40)
i
xi has to be expressed in terms of ( x, p, t )
x i ( x, p, t ) pi - L( x, x ( x, p, t ), t )
i
On the other hand, the expression for H ( x, p, t ) xi pi - L( x, x , t ) given in (40)
i
implies,
L L L
dH = p dx x dp
i
i i
i
i i
i xi
dxi i xi
dxi
t
dt
L
(d ) pi
dt xi
( d pi )
dt
p i
Thus, the first and fourth sum-terms cancel out
L
dH = x dp
i
i i
i
p i dxi
t
dt (42)
x i ( x, p, t ) pi - L( x, x ( x, p, t ), t )
i
H
xi Hamilton canonical
pi equations
H (2n first-order equations) (43)
p i
xi
i = 1, 2, …, n
and
H L
(44)
t t
Notice if L does not depend on t explicitly, neither does H.
Recipe for solving problems in mechanics
i) Set up the Lagrangian L( x, x , t ) .
L
ii) Obtain the canonical momenta pi ( x, x, t )
xi
iii) Obtain H ( x, p, t ) xi pi - L( x, x , t ) in terms of x and p .
i
Example. Two dimensional motion of a particle in a central potential V( r ) =V(r).
1
L( r , , r,) )= T - V = m ( r 2 r 2 2 ) - V(r).
2
Using the definition (39),
L L 2
pr = m r and p = m r
r
pr p
pr 2
p 1 m ( ( pr )2 r 2 ( p2 )2 )
m mr 2 m mr
2
p 2
H (r, , pr , p ) = pr V(r).
2m 2mr 2
Using (30) one obtains the equation of motion for each of the 4 independent variables
r , , pr , p .
H pr H p
r = =
pr m p mr 2
H V H
p r = pr =0
r r
- p i xi
The first and second sum-term cancel out, which proves the statement. That is,
L L
One obtains,
xi
= 2ki xi and i
xi
xi
2 ki ( xi )2 = 2T
i
L
H= i
xi
xi
L = 2T – L= 2T – ( T – V )= T + V .
t=to
phase ( x, p )
space
Alternatively, the Hamiltonian determines all the possible motions the system can
perform in phase space, the initial conditions picking out the particular motion
which is the solution to a particular problem.
H H
-
pi xi
F H F H F
i xi pi i pi xi t
dF F H F H F
x (48)
dt i i pi pi xi t
The first term on the right side is called the Poisson bracket between F and H (to be
described in more detailed in the sections below).
L
pi ( x1 , x2 , x1 , x2 , t ) xi for i= 1,2 (50)
xi
H ( x, p, t ) xi pi - L( x, x , t )
i
2 2
1 1
p1 p1 p2 p2 ( p1 p2 ) ( x1 x2 )
2 2 2 2
2 2
1 1
H ( x1, x2 , p1, p2 , t ) ( p1 p2 ) ( x1 x2 )
2 2 2 2
(51)
2 2
Without finding the explicit solution, let’s show that expression (48) tell us that the
angular momentum F,
F ( q1 , q2 , p1, p2 , t ) x1 p2 x2 p1 angular momentum (52)
Is a constant of the motion.
First, let’s calculate,
F H F H
p2 p1 x2 x1
x1 p1 p1 x1
F H F H
p1 p2 x1 x2
x2 p2 p2 x2
F
0
t
which gives
2
F H F H F
x
pi xi t
=0
i 1 i pi
b) Cyclic coordinates
According to the definition given in Section 4.1C, a cyclic coordinate xi is one that does
L d L
not appear in the Lagrangian. The Lagrangian equation then implies that
xi dt xi
L
the generalized momentum pi = is a constant of motion. But the Hamiltonian
x i
H H
equation p i implies also that in this case 0 ; that is,
xi xi
a coordinate the is cyclic (i.e. absent in the Lagrangian) (54)
will also be absent from the Hamiltonian
This conclusion can also be obtained from the definition of the Hamiltonian,
H ( x, p, t ) x i pi - L( x, x , t ) . Notice, H differ from – L by x i pi , which does
i i
not involved the coordinates xi explicitly.
It is true that rarely occurs in practice that all the coordinates are cyclic (for, thus, taking
advantage of the easy way to find the solutions.) However, a given problem can be
described by different sets of coordinates. It becomes important then to find a
systematic procedure for transforming from one set of variables to another set of
variables where the solution is more conveniently tractable. That will be the subject of
using canonical transformations, to be described in section 4.4.
4.2D The modified Hamilton’s principle: Derivation of the Hamilton’s equations from a
variational principle.
Hamilton’s canonical equation can be obtained from a variational principle, similar
to the way the Lagrange equations were obtained in Section 4.1B above. However, the
variations will be over paths in the ( x, p) phase-space, which has 2n dimensions, twice
the n dimensions of the x configuration-space. Interestingly enough, the function inside
the integral, upon which the variational principle will be applied, is again the Lagrangian
L, but now considered as a function of ( x, x , p, p ,t ) .
L( x, x , p, p ,t ) x i pi - H ( x, p, t ) (55)
i
As a first step, let’s apply the variational principle to
( x2 , p2 ,,t2 )
S ( x, p) ) ≡ L( x, x , p, p ,t ) dt (56)
( x1 , p1 , t1 )
Inside the integral, we have just written ( x, x , p, p ,t ) for
simplicity, but ( x(t ), x (t ), p(t ), p (t ) ,t ) should be used
instead, respectively.
In applying the variational principle, we realized it is very similar to the case when we
applied it in the configuration space. This time we just have more independent
variables. The result is,
d L L d L L
0 and 0, for i=1,2, …, n. (57)
dt xi xi dt p i pi
Applying (57) to the Lagrangian given in (55), L( x, x , p, p ,t ) xi pi - H ( x, p, t ) ,
i
L L H
pi , -
xi xi xi
d L L H
0 implies p i (58a)
dt xi xi xi
On the other hand,
L L H
0, xi -
p i pi pi
d L L H
0 implies xi (58b)
dt p i pi pi
That is, we obtain in (58) the canonical Hamiltonian equations (43).
That is,
[ x , x ]( x, p ) [ p , p ]( x, p ) 0 (60)
[ x , p ]( x, p ) … ..
Remark.
Notice, the result (60) appears to be trivial. It is. In fact, this is a property of the Poisson
bracket itself, regardless of the existence of a Hamiltonian.
It turns out, however, that if the bracket of the variables x , p were calculated with
respect to other arbitrary variables (Q,P), the value of [ x , p ](Q, P ) would be, in general,
different than .
For an arbitrary transformation
[ x , p ](Q, P ) (61) (61)
(x,p) (Q,P)
But, for a particular type of transformation of coordinates, called canonical
transformations (which are introduced in connection to Hamiltonian description of
motion, to be described in detail in the following sections), the Poisson bracket has the
remarkable property of remaining constant, that is [ x , p ]
( Q, P )
, regardless of
the new canonical variables used to describe the motion. Hence the usefulness of the
Poisson bracket; it helps to identify such important canonical transformations.
Here we derive the chain rule that relates the Poisson brackets evaluated with
respect to different coordinates.
If S=S (Q, P ) and R=R (Q, P )
where Q Q( x, p, t ) and P P ( x, p, t )
S , R( x, p ) S R S R
=
i xi pi pi xi
S Qα S Pα R Qβ R Pβ
( ) ( )
i α Qα xi Pα xi Q β pi Pβ pi
S Qα S Pα R Qβ R Pβ
( ) ( )
i Qα pi Pα pi Q β xi Pβ xi
S R S R
( [Qα ,Q β ]( x, p ) [Qα , Pβ ]( x, p )
Qα Q β Qα Pβ
+ S [ Pα ,Q β ]( x, p ) R S [ Pα , Pβ ]( x, p ) R )
Pα Q β Pα Pβ
Thus,
For S=S (Q, P ) and R=R (Q, P )
where Q Q( x, p, t ) and P P ( x, p, t )
S R S R
[Qα ,Q β ]( x, p ) [Qα , Pβ ]( x, p )
Q Q β Qα Pβ
α β α
S R S R
[ Pα ,Q β ]( x, p ) [ Pα , Pβ ]( x, p )
Pα Q β Pα Pβ
Notice, if a particular transformation of coordinates (x,p) (Q,P) satisfies,
[Qβ , Pα ]( x, p ) = ,
[ Pβ , Pα ]( x, p ) = ,
S,R( x,p ) S,R(Q, P ) Valid only for those particular trans- (64)
formation (x,p) (Q,P) that satisfy (63)
Thus, we have found that those transformation of coordinates (x,p) (Q,P) satisfying
(63) are very special: the bracket of two arbitrary dynamical quantities R and S remain
invariant, independent of whether we use (x,p) or (Q,P) to evaluate the bracket.
4.3C The Poisson bracket theorem: Preserving the description of the classical motion
in terms of a Hamiltonian
a) Example of a motion for which there is not Hamiltonian to describe it
Given a Hamiltonian H ( x, p, t ) we say that the time evolution of the classical system
is generated by H according to the canonical equations (43) and (59).
There are cases, however, in which no Hamiltonian generates a particular motion.
Consider, for example, the following motion:
x ( p x 2 )2 (65)
x
p 1 - 2x ( p x 2 ) 2
2 ( p x2 )
That there is no H whose associated canonical equation ends up in (60) can be seen by
first requiring,
H
x ( p x 2 )2
p
H x
p 1 - 2x ( p x 2 ) 2
x 2 ( p x2 )
P 2( p x 2 )( p 2 xx )
= 2P1/2( p +2Q Q )
according to (65)
x
= 2P1/2 ( 1 - 2x ( p x 2 ) 2 + 2Q Q )
2 (p x )
2
Q
= 2P1/2 ( 1 1/2 - 2 Q P 2Q Q )
2P
But Q P
Q
= 2P1/2 ( 1 1/2 )
2P
P Q
In short, the motion described, in the x,p coordinates, as
x ( p x 2 )2
x
p 1 - 2x ( p x 2 ) 2
2 ( p x2 )
P Q
These equations do have a Hamiltonian. Indeed,
K
setting Q P , gives K(Q,P) = (1/2)P2 + f (Q)
P
K
setting P 2
Q , gives K(Q,P) = (1/2)Q + g (P)
Q
Therefore,
K(Q,P) = (1/2) Q2 + (1/2) P2 (69)
An outline of the proof is given in the Appendix-1, at the end of this chapter.
This theorem is used to verify whether or not a given motion is a Hamiltonian motion. If
we knew x=x(t) and p = p(t), we would apply (70) with the choice of R=q and S=p. If
(70) were not fulfilled, then there will be no Hamiltonian to describe such a motion.
Transformation of coordinates
Consider that a classical motion is described by
x(t ), p(t ) (71)
for which a Hamiltonian exists; i.e. the changes of x (t ) and p(t ) are governed by a
Hamiltonian H:
H H
xi and p i (72)
pi xi
t
t
H
t=to
Transformation t=to
( x, p )
phase space ( Q, P )
phase space
Figure 3. For
Q Q( x, p, t ) ; P P( x, p, t )
S, R ( x , p ) S R S R
i xi pi pi xi
Consider an invertible transformation
Q Q( x, p, t ) ; P P( x, p, t )
H= C +
(c x c' p ) +
1
n n
1 1
+ ( x x ' x p " p p ) (78)
2
1 1 2
Note: In the literature the restricted canonical transformations are often simply called
canonical transformations.
generate restricted canonical transformations will result along the way of attempting to
classify them. Toward this end, let’s analyze first the following expression
x p - Q P
i
i i
i
i i (81)
The strategy is to show that there exists a function F that allows express (81) in the
following alternative form,
F F F
x p - Q P
i i i i = x x p p
i
i
i
i
t
... else (83)
i i i i
To this end, let’s expand the left side in terms of the old coordinates,
i
i
x p - Q P =
i
i
i i
Qi Qi Q
= x p
i
i i - ( x
i k
x k
k pk
p k i ) Pi
t
k
Q i
Qm Qm Qi
xi pi - ( x xi Pm p p i Pm Pi )
i m i i m i i i t
Qm Q Q
( pi Pm ) xi ( m Pm ) p i i Pi (84)
i m xi i m pi i t
i i
The next step is then to identify, term by term, expressions (83) and (84).
Appendix-3 outlines the demonstration that there exists such a scalar function F such
that,
Qm F
pi x
m
Pm
xi
i
Qm F
Pm (85)
m pi pi
and
Qi F
Pi H K
i t t
F F F
x p - Q P
i
i i
i
i i = x x
i
i p p
i
i (
t
+ H - K)
i i
Or
F F F
xi pi - H - Qi Pi K = x x i p p i
t
i i i i i i
xi pi - H - Qi Pi K
dF
= (86)
i i dt
F4 ( p, P, t ) = F ( x( p, P), p, t )
Since the xi and the Qi are independent, then the coefficients of x i and Q i should be
equal,
F1
pi ( x, Q , t ) i= 1, 2, … , n (88a)
xi
These are used to solve for the n variables
Qi as a function of ( x, p)
F1
Pi ( x, Q , t ) i= 1, 2, … , n (88b)
Qi
Since the Qi have been determined in (88a),
this expression gives the n variables Pi as a
function of ( x, p)
which leaves (87) with,
F1
K H (88c)
t
xi pi - H - Qm Pm K = dt ( x, P , t )
dF2
i m
Q Q
where we will replace Q m m xi m Pi
i xi i Pi
Q Q
m
Q m Pm = ( m xi m P i ) Pm
m i xi i Pi
Inverting the order of the summation
Q Q
m
Q m Pm = m Pm xi m Pm P i
i m xi i m Pi
(90)
Since the xi and the Pi are independent in (91), then the coefficients of x i and P i
should be equal,
Q F2
pi m Pm ( x, P , t )
m xi xi
Q F2
m Pm = ( x, P , t ) (92)
m Pi Pi
F2
K -H =
t
It is not straightforward to visualize that from the first equation the Pi can be solved as
a function of ( x, p) , since the Qm are involved there. Hence, we perform an extra step.
First, notice that the first equation in (92) can be written as,
F2
pi
xi
Q
m
m m P
xi
( x, P , t )
pi ( F2 Qm Pm ) (93a)
xi m
Q
m Pm imQm
m Pi m
Q
m Pm Qi
m Pi
(Qm Pm ) F2
Qi = ( x, P , t )
m Pi Pi
Qi = ( F2 Qm Pm ) (93b)
Pi m
At the end, there will be no interest in the function F2. We already found a F2’ that, if
expressed in terms of (x,P) as independent variables, it will generate a canonical
transformation. Hence, let’s just rename the F2’ as F2.
In summary
F2 F2 F2
pi ( x, P , t ) , Qi = ( x, P , t ) , K -H =
xi Pi t
I
( x, p) (Q, P)
G H
Pi pi = pi (dt )
xi xi
Using the Hamiltonian equations (43)
= pi (dt ) ( p i )
But p i (dt ) is the increment of p due to the motion
= pi dpi
G H
Qi xi = xi (dt )
pi pi
Using the Hamiltonian equations (43)
= xi (dt ) ( xi )
= xi dxi
That is, we have obtained:
transformation ( x, p)
(Q, P)
Qi xi dxi (101)
Pi pi dpi
where dxi and dpi are, respectively, the changes
in xi and pi due to the motion governed by the
Hamiltonia H. (70)
F ( t, t' ) F ( t, t" )
x(t’), p(t’) x(t”), p((t”)
x (t), p(t) , …. , (103)
F2 f m ( x t )
pi
xi i
=
m xi
Pm
That is, the new coordinates Qi depend only on the old coordinates and time, but do
not involve the old momenta. Such a transformation is then an example of the class of
point transformation, done in the configuration space, which led to the consideration of
the invariant of the Lagrangian equation with respect to a transformation of
coordinates. In this context, such a point transformation (107) are canonical, and
therefore preserve the Hamiltonian description of the motion. We have therefore a
view of the invariance of the Lagrangian from a canonical transformation of coordinates
perspective.
Now come the time to highlight further the importance of expression (112). It turns out
that replacing (112) in (108), one obtains,
d
L' (η, η ,t ) 0 (113)
dt i i
That d L( ζ , ζ ,t ) 0 and d L' (η, η ,t ) 0 are the same is
dt dt
i i i i
remarkable.
It was obtained not obtained because d L( ζ , ζ ,t ) 0 and
i
dt i
d
L' (η, η ,t ) 0 are equations for the same set of motion in the
dt i i
phase-space.
F ( t, t0 )
( x(t), p(t) ) ( Q , P ) (114)
( x(t0) , p(t0) )
( x0 , p0 )
New coordinates constant in time
F ( t, t0 )
H(x , p ) K( Q, P)
One can automatically ensure that the new variables (Q , P) are constant in time by
requiring that the new Hamiltonian K( Q, P) be identically zero!, for then the equations
of motion are,
K
Q i 0 (115)
Pi
K
P i 0
Qi
K is related to the old Hamiltonian H and the generating function by,
F
K H
t
which will be zero if,
F
H ( x, p, t ) 0 (116)
t
It is convenient to choose ( x, P) as the independent coordinates; that is,
F2 ( x, P, t ) = F ( x, p( x, P), t ) (117)
which gives the
F2
pi ( x, P , t ) (118)
xi
F2
Qi = ( x, P , t )
Pi
Given H, this constitutes an equation for F2. It is customary to denote the solution of
this equation by S, which is called the Hamilton’s principal function.
S S S
H ( x, , .... , , t) 0 (119)
x1 xn t
Hamilton-Jacobi Equation
S S ( x, P , t ) with P = constant
Hamilton’s principal function
When solving (119), the n constants of integration can be taken as the Pi s.
S
Expression pi ( x, P , t ) evaluated at t0 gives a relationship between the
xi
constant value of P and the initial conditions x0 , p0.
F
Then using expression (118), Qi = 2 ( x, P, t ) , one has a relationship to find the
Pi
Qis. In particular the Qis can be obtained by evaluating this expression at t0.
F
Inverting Qi = 2 ( x, P, t ) , one obtains,
Pi
x x(Q, P, t ) (120)
which gives the the coordinates as a function of time and the initial conditions.
In short, when solving the Hamilton-Jacobi equation, we are at the same time obtaining
a solution of the mechanical problem
( )
(Below we demonstrate that (1) implies = )
x p
0 [ , p ] [ x , ]
+ =0
x p
or
( )
=
x p
In other words, (2) does not contradict the quest for finding a function H
that satisfies H/p and H/x .
________________________________________________________________________
Appendix-2
H= C +
(c x c' p )
1
n n
1 1
+ ( x x ' x p " p p ) (2)
2
1 1 2
S , R ( x , p)
d
z
dt
Assuming that the evolution of ( x, p) is governed by
a Hamiltonian, Poisson bracket theorem implies,
S , R ( x , p) z{ [ R , S ]( x , p) [ R, S ]( x , p) }
d
z
dt
Using b) again,
[ R , S ](Q ,P) [ R, S ](Q ,P )
That c) implies b) is more elaborated.
Exploiting the fact that Q Q( x, p, t ) ; P P( x, p, t ) is canonoid with respect to
any quadratic Hamiltonian, we will figure out first the value of
[ x , x ](Q ,P ) = ? , [ x , p ](Q ,P ) =?, and [ p , p ](Q ,P ) =?,
( Q ,P )
which then be used in the result for [ R, S ] obtained in (36).
Lets’ start considering the case that Q Q( x, p, t ) ; P P( x, p, t )
is canonoid with respect to H=C (constant).
H ( x, p, t ) constant implies,
H const H const
xi 0 and p i 0 (3)
pi pi xi xi
Since Q Q( x, p, t ) ; P P( x, p, t ) is canonoid with respect to the
Hamiltonian H ( x, p, t ) =const, it means (according to the Poisson bracket
theorem) that there exist a Hamiltonian (K= K (Q, P) ) according to which a
quantities like [ x , x ](Q ,P ) and [ x , p ](Q ,P ) would evolve in time as,
d
[ x , x ](Q , P ) [ x , x ](Q , P ) [ x , x ](Q , P )
dt
d
[ x , p ](Q , P ) [ x , p ](Q , P ) [ x , p ](Q , P )
dt
d
[ p , p ](Q , P ) [ p , p ](Q , P ) [ p , p ](Q , P )
dt
But according to (40) all the xi 0 and p i 0 , which gives
d
[ x , x ]( Q ,P )
0;
dt
d
[ x , p ](Q , P ) 0 ; and (4)
dt
d
[ p , p ](Q , P ) 0
dt
Let’s put [ x , x ](Q , P ) , [ x , p ](Q , P ) , and [ p , p ](Q , P ) as a
function of ( x, p, t )
[ x , x ](Q , P ) ≡ f ( x, p, t ) (5)
(Q , P )
[ x , p ] ≡ g ( x, p, t )
(Q ,P )
[ p , p ] ≡ s ( x, p, t )
For a Hamiltonian H ( x, p, t ) =const,
[ f , H]
( x , p)
0 , [ g , H ]( x , p) 0 , [s, H ]( x , p) 0 ; and
the time variation of f, g, and s, are given by,
d
f [ f , H ]( x , p ) f = f
dt t t
d
g [ g , H ]( x , p ) g = g
dt t t
d
s [ s, H ] ( x, p)
s = g
dt t t
d d d
Since, according to (4), f 0 , g 0 , and s 0 , one
dt dt dt
obtains,
f [ x , x ](Q , P ) 0 ; .
t t
g [ x , p ] (Q ,P )
0 ; and .. (6)
t t
s [ p , p ](Q , P ) 0 …
t t
(Q ,P ) (Q , P )
So, neither [ x , x ] , [ x , p ]
(Q , P )
nor [ p , p ] depend explicitly on the
time variable. .
Note: Notice that, although we have used the Hamiltonian H=constant to
(Q ,P ) (Q , P )
obtain the result (6), neither of the values [ x , x ] , [ x , p ] , or
(Q , P )
[ p , p ] depend on the particular Hamiltonian used. They depend
only on the transformation coordinates.
n
Consider now the Hamiltonian H= (c x c' p )
1
This implies
H H
xi c 'i and p i ci (7)
pi xi
Since Q Q( x, p, t ) ; P P( x, p, t ) is canonoid with respect to this
Hamiltonian H, it means (according to the Poisson bracket theorem) that
there exist a Hamiltonian (K’= K ' (Q, P) ) according to which a quantities like
[ x , x ](Q ,P ) , [ x , p ](Q ,P ) and [ p , p ](Q , P ) would evolve in time as,
d
[ x , x ](Q , P ) [ x , x ](Q , P ) [ x , x ](Q , P ) ,
dt
d
[ x , p ] (Q , P )
[ x , p ](Q ,P )
[ x , p ] (Q ,P )
, and
dt
d
[ p , p ](Q , P ) [ p , p ](Q , P ) [ p , p ](Q , P )
dt
Using (7),
[ x , x ](Q, P ) [c' , x ](Q , P ) 0 ;
[ x , p ] (Q , P )
[ x ,c ] (Q , P )
0 ; and
[ p , p ] (Q , P )
[ p ,c ] (Q , P )
0
Thus,
d d
[ x , x ](Q , P ) f 0
dt dt
d d
[ x , p ](Q , P ) g 0
dt dt
d d
[ p , p ](Q , P ) s 0
dt dt
Hence, in terms of the Hamiltonian H=H ( x, p, t )
d
0 f [ f , H ]( x , p ) f
dt t
d
0 g [ g , H ]( x , p ) g
dt t
d
0 s [ s, H ]( x , p ) s
dt t
Using (6),
[ f , H ]( x , p) 0
n
[ f , (ci xi c'i pi )] 0
i 1
n n
i 1
ci [ f , xi ] c' [ f , p ] 0
i 1
i i
n
f n
f
ck c 'i 0
k 1
pk k 1
xk
This should be true for all arbitrary values of ck and c’k, which implies,
f f
0 and 0 for k = 1, 2, …. , n
xk xk
Similarly, (8)
g g
0 and 0 for k = 1, 2, …. , n.
xk xk
Now we need to prove that all those constant values are of the same
magnitude │z│.
To that effect consider the Hamiltonian
n n
H
xi ( ' i x 2 "i
1
p 2 "i
1
p )
pi
H
p i ( i x 'i p )
xi
0 [ ( 'k xk "k pk ), x ](Q , P )
k
[ x , ( 'k xk "k pk )](Q , P )
k
( 'k [ xk , x ] (Q , P )
"k [ pk , x ] (Q , P )
k
( 'k [ x , xk ] (Q , P )
"k [ x , pk )] (Q ,P )
0 [ ( 'k xk "k pk ), p ](Q , P )
k
[ x ,( (k xk ' k pk )](Q , P )
k
0 ( 'k [ xk p ] (Q ,P )
"k [ pk , p ](Q , P )
k
k [ x , xk ](Q , P ) ' k [ x , pk )](Q , P )
k
k k
0 k [ xk , p ] (Q ,P )
'k [ pk , p ] (Q , P )
k
k [ p , xk ](Q , P ) ' k [ p , pk )](Q , P )
k
S , R (Q,P) S R S R
. … ..
i Qi Pi Pi Qi
S
R S
R
[ x , x ](Q , P ) [ x , p ](Q , P )
x x x p
S
R S
R
[ p , x ](Q , P ) [ p , p ](Q , P )
p x p p
S
R S
R
S , R (Q,P)
x
[ x , p ](Q , P )
p p
[ p , x ](Q , P )
x
S R S R
[ z x ] [ z x ]
x p p x
S R S R
z x
x p p x
which gives,
S , R (Q,P) z xS , R ( x,p)
___________________________________________________________________
Appendix-3
How to generate canonical transformations
A way to generate restricted canonical transformations will result along the way of
attempting to classify them. Toward this end, lets analyze the following expression,
x p - Q P
i
i i
i
i i (1)
x p - Q P
i
i i
i
i i =
expanding the second term
Qi Qi Q
= x p - ( x
i
i i
i k
xk Pi
k pk
p k Pi i Pi )
t
k
Renanimng Renanimng
i m , and i m , and
k i k i
Qm Qm Qi
xi pi - x xi Pm p p i Pm Pi )
i m i i m i i i t
Qm Q Q
( pi Pm ) xi ( m Pm ) p i i Pi (2)
i m xi i m pi i t
i i
The objective is to show that there exists a function F that makes (2) equal to
F F F
x x p p
i
i
i
i
t
i i
Qm
i pi Pm implies
m xi
i Qm 2Qm Q P
( Pm ) ( )Pm m m
xk m xk xi m xk xi m xi xk
Qm
k pk Pm implies
m xk
k Qm 2Qm Q P
( Pm ) ( )Pm m m
xi m xi xk m xi xk m xk xi
which gives,
i k Q P P Qm
m m m {xi , xk }Q,P the latter is the Lagrange bracket
xk xi m xk xi m xk xi
Qm
k Pm implies,
m pk
k 2Qm Q P
( )Pm ( m ) m
pi m pi pk m pk pi
which gives,
i k Q P Q P
( m ) m m ) m ) { pi , pk }Q,P = 0
pk pi m pk pi pi pk
This indicates that there exist a function F such that,
F
(5)
p
Qi
Ψ Pi implies,
i t
Ψ 2Qi Q P Ψ 2Qi Q P
( Pi i i ) and ( Pi i i )
xk i xk t t xk pk i pk t t pk
Qm Qm
k pk Pm implies k Pm implies
m xk m pk
k Qm Q P
2
k 2Qm Q P
( Pm m m ) and ( Pm m m )
t m txk xk t t m tpk pk t
This gives,
k Ψ Q P Q P k Ψ Q P Q P
( i i i i ) and ( i i i i )
t xk i t xk xk t t pk i t pk pk t
(6)
On one hand Q i can be determined by the Hamiltonian K=K (Q, P)
K
Q i
Pi
On the other hand Q i (t ) is a quantity that can also be tracked by the
Hamiltonian H=H ( x, p, t )
Q
Q i [Qi , H ]( x ,p ) i
t
Thus
Qi K
Q i [Qi , H ]( x ,p ) [Qi , H ]( x ,p )
t Pi
K Q H Q H
( i i )
Pi x p p x
Similarly,
Pi K
Pi [ Pi , H ]( x ,p ) [ Pi , H ]( x ,p )
t Qi
K P H P H
( i i )
Qi x p p x
Further,
Qi Pi K Pi Q H Q H Pi
( i i )
t xk Pi xk x p p x xk
Qi Pi Q K Qi Pi H P H
xk t
i
xk Qi xk
(
x
i )
p p x
Upon subtraction
Qi Pi Qi Pi K Pi Qi K
t xk xk t Pi xk xk Qi
Qi Pi H P H Q H Q H Pi
xk
(
x
i ) ( i i )
p p x x p p x xk
Qi Pi Qi Pi K Pi Qi K
t xk xk t Pi xk xk Qi
Pi Qi H P Qi H Q P H Q P H
( i ) ( i i i i )
x xk p p xk x x xk p p xk x
i t x x t x
{ x , x k }
p
k x
{ x , p }
k k k
K H
xk
0
k
x
( K H )
xk
k Ψ ( K H )
(7)
t xk xk
F
But, according to (4), k
xk
F Ψ ( K H )
t xk xk xk
Ψ ( K H ) F
0
xk xk xk t
F
0 (Ψ K H ) (8)
xk t
F
0 (Ψ K H ) (9)
pk t
F
This implies that (Ψ K H ) is a function of t alone. However, (4) and (5)
t
determine F up to an additive function f(t); the latter can be choose such that,
F
Ψ K H (10)
t
Qm F
pi Pm
m xi xi
From (2), which defines i, and (5)
Qm F
Pm
m pi pi
___________________________________________________________________
1
H. Goldstein, Classical Mechanics, Addison-Wesley Publishing (1959).
2
Saletan and Cromer, “Theoretical Mechanics” Wiley, 1971.), page 180.
3
Ref.2, page188.