Theoretical Dynamics October 01, 2010

Homework 4
Instructor: Dr. Thomas Cohen Submitted by: Vivek Saxena

Goldstein 9.7
Part (a)
F1 (q, Q, t) F2 (q, P, t)

F1
Pi = (1)
Qi
F2 (q, P, t) = F1 (q, Q, t) + Pi Qi (2)

F1 (q, Q, t) F3 (p, Q, t)

F1
pi = (3)
qi
F3 (p, Q, t) = F1 (q, Q, t) pi qi (4)

F1 (q, Q, t) F4 (p, P, t)

F1
pi = (5)
qi
F1
Pi = (6)
Qi
F4 (p, P, t) = F1 (q, Q, t) pi qi + Pi Qi (7)

F2 (q, P, t) F3 (p, Q, t)

F2
pi = (8)
qi
F2
Qi = (9)
Pi
F3 (p, Q, t) = F2 (q, P, t) pi qi Pi Qi (10)

F2 (q, P, t) F4 (p, P, t)

F2
pi = (11)
qi
F4 (p, P, t) = F2 (q, P, t) pi qi (12)

4-1
F3 (p, Q, t) F4 (p, P, t)

F3
Pi = (13)
Qi
F4 (p, P, t) = F3 (p, Q, t) + Pi Qi (14)

Part (b)
For an identity transformation, F2 = qi Pi and by equation (7), the type 4 generating
function is

F4 (p, P, t) = F2 (q, P, t) pi qi (15)

= qi P i p i qi
F2
= 0 as pi = = Pi
qi
For an exchange transformation, F1 = qi Qi and by equation (4), the type 3 generating
function is

F3 (p, Q, t) = F1 (q, Q, t) pi qi (16)

= qi Q i p i qi
F1
= 0 as pi = = Qi (17)
qi

Part (c)
Consider a type 2 generating function F2 (q, P, t) of the old coordinates and the new mo-
menta, of the form

F2 (q, P, t) = fi (q1 , . . . , qn ; t)Pi g(q1 , . . . , qn ; t) (18)

where fi s are a set of independent functions, and gi s are differentiable functions of the old
coordinates and time. The new coordinates Qi are given by
F2
Qi = = fi (q1 , . . . , qn ; t) (19)
Pi
In particular, the function
fi (q1 , . . . , qn ; t) = Rij qj (20)
where Rij is the (i, j)-th element of a N N orthogonal matrix, generates an orthogonal
transformation of the coordinates. Now,
F2 fi g g
pj = = Pi = Rij Pi (21)
qj qj qj qj
This equation can be written in matrix form, as
f g
p= P (22)
q q

4-2
where p denotes the N 1 column vector (p1 , . . . , pN )T , g/q denotes the N 1 column
vector (g/q1 , . . . , g/qn )T , and f
q denotes the N N matrix with entries
 
f fi
= = Rij (23)
q ij qj

From (22), the new momenta are given by

1 
 
f g
P = p+ (24)
q q
 
g
= R1 p + (25)
q
= R1 (p + q g) (26)

As R is an orthogonal matrix, RRT = RT R = I, so R1 = RT is also an orthogonal

transformation.

This gives the required result: the new momenta are given by the orthogonal transformation
(R1 ) of an n-dimensional vector (p + q g), whose components are the old momenta (p)
plus a gradient in configuration space (q g).

Goldstein 9.25
Part (a)
The given Hamiltonian is  
1 1
H= + p2 q 4 (27)
2 q2
The equation of motion for q is
H
q = = pq 4 (28)
p

Part (b)
Suppose we let Q2 = 1/q 2 and P 2 = p2 q 4 . Then, Q = 1/q and P = pq 2 . Now,

{Q, P } = {1/q, pq 2 }
= {q 1 , pq 2 }
= {q 1 , p}q 2 + p{q 1 , q 2 }
 1
q p q 1 p 2

= q +p0
q p p q
 
1
= 2 q2
q
= 1

4-3
So, the signs on both Q and P cannot be identical. We take
1
Q = (29)
q
P = pq 2 (30)

which is a valid canonical transformation. This gives the Hamiltonian,

1
H(Q, P ) = (P 2 + Q2 ) (31)
2
The equations of motion are
H
Q = =P (32)
P
H
P = = Q (33)
Q

So, Q + Q = 0, the solution to which is of the form Q = A cos t + B sin t. This gives
P = Q = B cos t A sin t. Now,
1
q = = (A cos t + B sin t)1 (34)
Q
p = P Q2 = (B cos t A sin t)(A cos t + B sin t)2 (35)

so,
q = (A cos t + B sin t)2 (A sin t + B cos t) (36)
and hence

pq 4 = (B cos tA sin t)(A cos t+B sin t)2 (A cos t+B sin t)4 = (B cos tA sin t)(A cos t+B sin t)2 = q
(37)
So, the solution to the transformed equation for Q satisfies the original equation of motion
for q.

Problem 1
Part (a)

{X, Px } = {x + , px }
= {x, px }
= 1 (38)

{Y, Py } = {y, px }
= 1 (39)

4-4
 {Z, Pz } = {z, pz }
= 1 (40)

{X, Py } = {X, Pz } = {Y, Px } = {Y, Pz } = {Z, Px } = {Z, Py } = 0 (41)

{X, X} = {Y, Y } = {Z, Z} = {Px , Px } = {Py , Pz } = {Pz , Pz } = {Px , Py } = {Py , Pz } = {Pz , Px } = 0
(42)
So, this is a canonical transformation. It corresponds to a translated canonical coordinate
system (translation along the x-direction in phase space).

dX
= [X, Px ] = 1 (43)
d
So Px is the generator of the canonical transformation.

Part (b)

{X, Px } = {x cos  + y sin , px cos  + py sin }

= cos2 {x, px } + sin2 {y, py }
= 1 (44)

{Y, Py } = {x sin  + y cos , px sin  + py cos }

= sin2 {x, px } + cos2 {y, py }
= 1 (45)

{Z, Pz } = {z, pz }
= 1 (46)

Using properties of the Poisson Bracket, we also have

{X, Py } = {X, Pz } = {Y, Px } = {Y, Pz } = {Z, Px } = {Z, Py } = 0 (47)

{X, X} = {Y, Y } = {Z, Z} = {Px , Px } = {Py , Pz } = {Pz , Pz } = {Px , Py } = {Py , Pz } = {Pz , Px } = 0

(48)
So, this is a canonical transformation. It corresponds to a rotation about the z-axis in phase
space.
dX
= x sin  + y cos  (49)
d
whereas
{X, Lz } = {x cos  + y sin , xpy ypx } = x sin  y cos  (50)
So,
dX
= {X, Lz } (51)
d
Therefore, Lz is the generator of the canonical transformation.

4-5
Part (c)

{X, Px } = {x, px + }
= 1 (52)

{Y, Py } = {y, py }
= 1 (53)

{Z, Pz } = {z, pz }
= 1 (54)

Using properties of the Poisson Bracket, we also have

{X, Py } = {X, Pz } = {Y, Px } = {Y, Pz } = {Z, Px } = {Z, Py } = 0 (55)

{X, X} = {Y, Y } = {Z, Z} = {Px , Px } = {Py , Pz } = {Pz , Pz } = {Px , Py } = {Py , Pz } = {Pz , Px } = 0

(56)
So, this is a canonical transformation. It corresponds to a translation along the px direction
in phase space. Now,
 
Px X Px X dPx
{Px , X} = =1= (57)
qi pi pi qi d

Therefore, X is the generator of the canonical transformation.

Part (d)

{X, Px } = {(1 + )x, (1 + )1 px }

= 1 (58)

{Y, Py } = {(1 + )y, (1 + )1 py }

= 1 (59)

{Z, Pz } = {(1 + )z, (1 + )1 pz }

= 1 (60)

Using properties of the Poisson Bracket, we also have

{X, Py } = {X, Pz } = {Y, Px } = {Y, Pz } = {Z, Px } = {Z, Py } = 0 (61)

{X, X} = {Y, Y } = {Z, Z} = {Px , Px } = {Py , Pz } = {Pz , Pz } = {Px , Py } = {Py , Pz } = {Pz , Px } = 0

(62)

4-6
So, this is a canonical transformation. It is a scaling transformation, which preserves the
volume element in phase space. Suppose g is the generator of the scaling transformation.
Then,
X
= x = [X, g] = [(1 + )x, g] (63)

which implies
x x g x g
= [x, g] = (64)
1+ x px px x
that is,
x g
= (65)
1+ px
the solution to which is
xpx
g= + f (y, z, py , pz ) (66)
1+
As dY /d = y = [(1 + )y, g] and dZ/d = z = [(1 + )z, g], following a similar argument for
Y and Z (or by symmetry) we get
xpx ypy zpz
g= + + + constant (67)
1+ 1+ 1+
as the generator of the scaling transformation.

Problem 2
As is a canonical transformation, we have
i
= {i , g} (68)

So,
H H i
=
 i 
H
= {i , g}
i
H i g
= Jjk (as is a canonical transformation)
i j k
H i g
= Jjk (as Poisson Brackets are invariant under canonical transformations)
i j k
H g
= ij Jjk
i k
H g
= Jij
i j
= {H, g}
= g (69)
H
But since H is conserved,  = 0 and hence g = 0. Therefore, g is conserved.

4-7
Problem 3
The quantity , which was found to be an invariant of the system, can be expressed in
terms of the canonical coordinates x, y, px , py as
1 2 1
(x, y, px , py ) = (px p2y ) + m 2 (x2 y 2 ) (70)
2m 2
As is the conserved generator of a family of canonical transformations parametrized by
an infinitesimal parameter , we must have
x = {x, } (71)
y = {y, } (72)
px = {px , } (73)
py = {px , } (74)
(75)
We consider each condition separately below.
x = {x, }
1 2 1
= {x, (px p2y ) + m 2 (x2 y 2 )}
2m 2
p2x
= {x, }
2m
px
= (76)
m

y = {y, }
1 2 1
= {y, (p p2y ) + m 2 (x2 y 2 )}
2m x 2
p2y
= {y, }
2m
py
= (77)
m

px = {px , }
1 2 1
= {px , (px p2y ) + m 2 (x2 y 2 )}
2m 2
1 2 2
= {px , m x }
2
= m 2 x (78)

py = {py , }
1 2 1
= {py , (px p2y ) + m 2 (x2 y 2 )}
2m 2
1 2 2
= {py , m y }
2
2
= m y (79)

4-8
Now, let  = where is a parameter. Then, the above equations become
dx px
= (80)
d m
dy py
= (81)
d m
dpx
= m 2 x (82)
d
dpy
= m 2 y (83)
d
So,
d2 x
+ 2x = 0 (84)
d2
d2 y
+ 2 y = 0 (85)
d2
The solutions to which are
x = A cos() + B sin() (86)

y = C cos( ) + D sin( ) (87)
and correspondingly
px = mA sin() + mB cos() (88)

py = m C sin( ) m D cos( ) (89)
Using the subscript 0 to denote the initial coordinates and momenta, we have
x0 = A (90)
y0 = C (91)
px 0 = mB (92)

py 0 = m D (93)
So,
px
x = x0 cos() + 0 sin()
m
py 0
y = y0 cos( ) sin( )
m
px = mx0 sin() + px 0 cos()

py = m y0 sin( ) + py 0 cos( )
Reverting to the notation in which x0 , px 0 , y0 , py 0 denote the original coordinates and
X, Y, Px , Py denote the canonically transformed coordinates, we get the form of the canon-
ical transformation as
px
X = x cos() + sin() (94)
m
Px = px cos() mx sin() (95)
py
Y = y cos( ) sin( ) (96)
m

Py = py cos( ) + m y sin( ) (97)

4-9
where is an arbitrary parameter, such that = 0 corresponds to the untransformed coor-
dinates. This canonical transformation is composed of two rotations in the 4-dimensional
phase space (one involving X and Px and the other involving Y and Py ), and its generator
is the conserved quantity .

Problem 4
Part (a)
P 2t
 
1 2
F2 (q, P, t) = q + gt (P mgt) (98)
2 2m
Now,
F2
p = = P mgt (99)
q
F2 1 Pt 1 pt
Q = = q + gt2 = q + gt2 gt2 (100)
P 2 m 2 m
So, the canonical transformation is

P = p + mgt (101)
pt 1 2
Q = q gt (102)
m 2

Part (b)

pt 1 2 pt 1 2
{Q, Q} = {q gt , q gt } = 0 (103)
m 2 m 2
{P, P } = {p + mgt, p + mgt} = 0 (104)
pt 1 2
{Q, P } = {q gt , p + mgt}
m 2
Q P Q P
=
q p p q
 
t
= (1)(1) (0)
m
= 1 (105)

So, the transformation satisfies the canonical Poisson Bracket relations.

Part (c)
The Lagrangian is
1 2
L(q, q) = mq mgq (106)
2

4 - 10
The canonical momentum is
L
p =
q
= mq (107)

So the Hamiltonian is

H = pq L
p2
= + mgq (108)
2m
pt
Now, Q = q m 12 gt2 , so

pt 1 2 p2
 
{Q, H} = q gt , + mgq
m 2 2m
p2
   
pt
= q, , mgq
2m m
1
= {q, p2 } gt{p, q}
2m
p
= + gt (109)
m
Also
Q p
= gt (110)
t m
So,
dQ Q
= + {Q, H} = 0 (111)
dt t
Also, P = p + mgt, so

p2
{P, H} = {p + mgt, + mgq}
2m
= mg{p, q}
= mg (112)

and
P
= mg (113)
t
So,
dP P
= + {P, H} = 0 (114)
dt t

4 - 11
Part (d)

P2
 
F2 1 2
= gt(P mgt) + q + gt (mg)
t 2 2m
3 P 2
= P gt mg 2 t2 mgq
2 2m
3 (p + mgt)2
= (p + mgt)gt mg 2 t2 mgq
2 2m
p 2
= mgq mg 2 t2 (115)
2m
So, the Hamiltonian associated with Q, P is
F2
K = H+
t
p2 p2
= + mgq mgq mg 2 t2
2m 2m
= mg 2 t2 (116)

So, the Hamiltonian K is zero up to time-dependent constant term mg 2 t2 , but it is not a

function of P and Q (which are constant with time, since {Q, H} = {P, H} = 0 as shown
above).

Part (e)
Q and P are conserved quantities, that equal the initial position and the initial momentum
respectively. They are constant with time, as q and p vary:

q(t = 0) = Q
p(t = 0) = P

Part (f )

F2
= P mgt = p (117)
q
F2 mg 2 t2 P2
= P gt mgq (118)
t 2 2m
p2
H = + mgq
2m
F2 2
 
1
= + mgq (119)
2m q

So,
F2
K = H+ = mg 2 t2 (as shown in part d)
t

4 - 12
implies
 
F2 F2
H q, + = mg 2 t2 (120)
q t
So, the Hamilton-Jacobi equation is satisfied, except for a time-dependent constant term
appearing on the right hand side.

Part (g)

f (Q, P, t) = F2 (q(Q, P, t), P, t)

P 2t
 
1 2
= q + gt (P mgt)
2 2m
P 2t
 
pt
= Q+ + gt2 (P mgt)
m 2m
P 2t
 
(P mgt)t
= Q+ + gt2 (P mgt)
m 2m
2
 
Pt P t
= Q+ (P mgt)
m 2m
2
P t
= QP + Qmgt gP t2 (121)
2m
So,
f P2
= mgQ 2P gt (122)
t 2m
Also, p = mq = P mgt. So,
p2
L(q, q) = mgq
2m  
1 2 Pt 1 2
= (P mgt) mg Q + gt
2m m 2
1 1
= (P 2 + m2 g 2 t2 2P mgt) mgQ P gt + mg 2 t2
2m 2
P2
= 2P gt mgQ + mg 2 t2
2m
f (Q, P, t)
= + mg 2 t2 (123)
t
f (Q,P,t)
So, L(q(Q, P, t), q(Q, P, t)) = t up to a time-dependent term mg 2 t2 .

Problem 5
The Hamilton-Jacobi equation, as expressed in the form
S(q, P )
H(q, S(q, P )) + =0 (124)
t

4 - 13
was obtained by constructing a generating function of the form

F = F2 (q, P, t) Qi Pi

where F2 denotes a generic type-2 generating function. For such a choice of F , the Hamil-
tonian K = H + Ft is zero.

Now, consider a type-3 generating function F3 of the old momenta and the new coordi-
nates, such that the Hamiltonian K is zero. Therefore,
K
Qi = =0 (125)
Pi
K
Pi = =0 (126)
Qi
Now,
F3
qi = = (p F3 )i (127)
pi
so,
F3
H(q(Q, p), p, t) + (Q, p, t) = 0 (128)
t
where the old coordinates q have been expressed in terms of the old momenta and the new
coordinates using equation (127). This is a PDE in (n + 1) variables p1 , . . . , pn , t. Let S
denote the solution of this PDE. Then, a solution of the form,

F3 S = S(p1 , . . . , pn ; 1 , . . . , n+1 ; t) (129)

where Qi = i are the constants of motion (for i = 1, . . . , n), is consistent with equa-
tion (125). Here the constant n+1 must be a constant of integration, so the physically
meaningful solution is of the form

S = S(p1 , . . . , pn ; 1 , . . . n ; t) (130)

So, in terms of S, equation (128) can be written as

S
H(p S, p, t) + (Q, p, t) = 0 (131)
t
which is of the desired form.

4 - 14