PHY 5246: Theoretical Dynamics, Fall 2015

Assignment # 4, Solutions

1 Graded Problems
Problem 1
(1.a)
The problem has spherical symmetry and is therefore naturally solved using spherical coordinates
(see figure). The fixed length of the pendulum gives the constraint: r = l and reduces the number
of generalized coordinates to two: {θ, φ}.
The kinetic and potential energies of the pendulum are
1 2 2 
T = m l θ̇ + l2 sin2 θφ̇2 , (1)
2
V = mgl(1 − cos θ) , (2)

and the Lagrangian is:

1  
L = T − V = m l2 θ̇2 + l2 sin2 θφ̇2 − mgl(1 − cos θ) . (3)
2
Since the Lagrangian does not depend on φ, the Euler-Lagrange equation of motion for φ states
the conservation of the angular momentum with respect to the vertical (or z) axis:

d  2 2  pφ
ml sin θφ̇ = 0 −→ ml2 sin2 θφ̇ = constant = pφ −→ φ̇ = . (4)
dt ml2 sin2 θ
On the other hand, the Euler-Lagrange equation of motion for θ reads

l2 θ̈ − l2 φ̇2 sin θ cos θ + gl sin θ = 0 . (5)

Substituting in Eq. (5) φ̇ in terms of pφ , as derived in Eq. (4), the equation for θ reduces to a one
dimensional equation in the only variable θ (and its derivatives) of the form

p2φ g
θ̈ − 3 cos θ + sin θ = 0 . (6)
m2 l4 sin θ l

(1.b)
The case φ = φ0 , i.e. φ̇ = 0 is trivial and reduced to the plane simple pendulum. Indeed, Eq. (5)
reduces to
g
θ̈ + sin θ = 0 . (7)
l
(1.c)
The case θ = θ0 corresponds to the case of a conical pendulum, since the pendulum describes a
conical surface during its motion. θ = θ0 is an equilibrium point, and therefore needs to verify
the condition θ̈ = θ̇ = 0. Imposing this condition in Eq. (5) we get that, in order to realize the
condition of conical pendulum we need to have φ̇ = φ̇0 such that
g g 1 g
φ̇20 sin θ0 cos θ0 = sin θ0 −→ φ̇20 = = sec θ0 . (8)
l l cos θ0 l
θ = θ0 is indeed a stable equilibrium position, as we can prove by studying the motion of the
pendulum for small displacements (in θ) from the equilibrium angle θ = θ0 . In order to do
that, let’s keep φ̇ = φ̇0 and let’s consider a small displacement from the equilibrium position, i.e.
θ = θ0 + δ, for δ infinitesimal. Plugging this into Eq. (5), we get

g g sin4 θ0 cos(θ0 + δ)
δ̈ + sin(θ0 + δ) − =0 . (9)
l l cos θ0 sin3 (θ0 + δ)

Expanding in δ and keeping only up to the linear terms in δ, we find

   
sin θo
g sin θ0 cos θ0 1 − δ cos θ0 
4
δ̈ + sin θ0 + cos θ0 δ −  =0 , (10)
l cos θ0 sin3 θ0 1 + δ cos θ0 3

sin θ0
" !#
g sin θ0 cos θ0
δ̈ + sin θ0 + cos θ0 δ − sin θ0 1 − δ − 3δ =0 ,
l cos θ0 sin θ0
g
δ̈ + (3 cos θ0 + sec θ0 ) δ = 0 ,
l
which is the equation of an harmonic oscillator with frequency
g
ω2 = (3 cos θ0 + sec θ0 ) . (11)
l
So, a conical pendulum rotates about the vertical axis with constant velocity φ̇0 (given in Eq. (8))
while it performs small oscillations about the θ = θ0 position with frequency ω (given in Eq. (11)).
The period of small oscillations in θ is
s
2π l
T = = 2π . (12)
ω g(3 cos θ0 + sec θ0 )

During one complete oscillation in θ the pendulum sweeps an angle φ1 > π, since
s
g l 2π
r
φ1 = φ̇0 T = sec θ0 2π =√ >π . (13)
l g(3 cos θ0 + sec θ0 ) 1 + 3 cos2 θ0
If we imagine to project the motion of the pendulum on the (x, y) plane, it will look like a curve
that could close after m turns if the condition 2πn = mT is verified. On the other hand, if we
imagine to trace the motion of the pendulum on a sphere, we would see it oscillating between two
circles (one lower and one higher of the θ = θ0 circle) describing a sort of sinusoidal curve on the
surface of the sphere.
Problem 2
The problem is planar (given the symmetry in the azimuthal angle), and can be treated using
polar coordinates {r, θ}. Normally, we would impose the constraint r = a and work with just one
generalized coordinate, θ. However, the problem asks for the angle at which the particle leaves
the hemisphere, which corresponds to the angle at which the force of the constraint vanishes. We
then need to determine the force of the constraint, keeping the full set of coordinates, {r, θ}, and
using the method of Lagrange undetermined multipliers.
We have one constraint, r − a = 0, and therefore we introduce one undetermined multiplier,
λ. Comparing the equation of the constraint, r − a = 0, to the generic constraint equation in
differential form
ar dr + aθ dθ = 0 we deduce: ar = 1 , aθ = 0 .
The kinetic and potential energies of the particle are
1
T = m(ṙ 2 + r 2 θ̇2 ) , (14)
2
V = mgr cos θ ,
and the Lagrangian is
1
L = T − V = m(ṙ 2 + r 2 θ̇2 ) − mgr cos θ . (15)
2
The equation of motions will then be, for r,
d ∂L ∂L
− = ar λ −→ mr̈ − mr θ̇2 + mg cos θ = λ , (16)
dt ∂ ṙ ∂r
and for θ,
d ∂L ∂L
− = aθ λ −→ mr 2 θ̈ + 2mr ṙ θ̇ − mgr sin θ = 0 . (17)
dt ∂ θ̇ ∂θ
We will solve them for r(t), θ(t) and λ, together with the constraint equation r = a (i.e. ṙ = r̈ = 0).
After using the equation of the constraint, Eqs. (16) and (17) become:
maθ̇2 − mg cos θ + λ = 0 (18)
mga sin θ − ma2 θ̈ = 0 . (19)
Noticing that
dθ̇ dθ̇ dθ dθ̇
θ̈ = = = θ̇ , (20)
dt dθ dt dθ
we can derive, from Eq. (19), that
g dθ̇
θ̈ = sin θ = θ̇ , (21)
a dθ
which, upon integration, yields
1 2 g g
θ̇ = − cos θ + , (22)
2 a a
where we have assumed that θ̇ = 0 at t = 0, when θ = 0. Substituting θ̇2 into Eq. (18) we get
λ = mg(3 cos θ − 2) , (23)
and therefore the particle leaves the hemisphere when
2
 
λ = 0 −→ θ = arccos . (24)
3
Problem 3 (Goldstein 2.14)
There are two constraints in this problem: (1) the distance from
the center of the cylinder to the center of the hoop is R+a, and (2)
a there is no slipping, i.e. the velocity of the contact point between
the cylinder and the hoop is zero. Describe this motion using the
coordinates of the center of mass of the hoop and the rotation angle
of the hoop about the center of mass. The constraints are thus
R
r =R+a (25)

(R + a)θ̇ − aφ̇ = 0 (26)

We will use the method of Lagrange multipliers to find the normal force on the hoop. The value
of θ for which the force is zero will correspond to the point at which the hoop falls of the cylinder.
The generalized coordinates will be {r, θ, φ}.
The equations of constraints are

r = R + a → dr = 0 → a1,r = 1, a1,θ = a1,φ = 0

φ
(R+a)θ̇−aφ̇ = 0 → (R+a)dθ−adφ = 0 → a2,r = 0, a2,θ = R+a, a2,φ = −a.
The Lagrangian is
θ ~r
1 1
L = m(ṙ 2 + r 2 θ̇2 ) + (ma2 )φ̇2 − mgr cos θ,
2 2
and the equations of motion (via Euler-Lagrange) are

d ∂L ∂L
− = λ1 a1,r ⇒ mr̈ − mr θ̇2 + mg cos θ = λ1 ,
dt ∂ ṙ ∂r
d ∂L ∂L
− = λ2 a2,θ ⇒ 2mr ṙθ̇ + mr 2 θ̈ − mgr sin θ = (R + a)λ2 ,
dt ∂ θ̇ ∂θ
d ∂L ∂L
− = λ2 a2,φ ⇒ maφ̈ = −aλ2 ,
dt ∂ φ̇ ∂φ

where the Lagrange multipliers are λ1 , λ2 . These need to be solved together with the equations of
constraint (25) (which implies ṙ = r̈ = 0) and (26). We find

−m(R + a)θ̇2 + mg cos θ = λ1 , (27)

m(R + a)2 θ̈ − mg(R + a) sin θ = (R + a)λ2 , (28)

ma2 φ̈ = −aλ2 . (29)
Note that

• λ1 corresponds to the r-coordinate and is the normal force on the hoop. Thus, the condition
we are looking for is λ1 = 0.

• λ2 corresponds to the rotation of the hoop on the cylinder (in either θ or φ), and is thus the
tangent friction that make the “no-skipping” condition possible.
Solving the system of equations:
R+a
(29): ma2 θ̈ = −aλ2 → λ2 m(R + a)θ̈,
a

(28): m(R + a)2 θ̈ − mg(R + a) sin θ = −m(R + a)2 θ̈ → 2(R + a)θ̈ − g sin θ = 0
g sin θ
⇒ θ̈ = .
2(R + a)
Multiply this last equation by θ̇ and rewrite it as
d 1 2 g g d
 
θ̇ = sin θθ̇ = (cos θ)
dt 2 2(R + a) 2(R + a) dt
1 g
⇒ θ̇2 = cos θ + const.
2 2(R + a)
Setting θ = 0 and θ̇ = 0 as the initial conditions we find that constant term is
g
const. = −
2(R + a)

which gives us the equation of motion

g
θ̇2 = (cos θ − 1).
R+a
Now plug this into (27):
mg(cos θ − 1) + mg cos θ = λ1 .
This gives us the Lagrange multiplier (and force of constraint)

λ1 = mg(2 cos θ − 1),

and setting this to zero we can find the angle at which the constraint is violated:
1 π
λ1 = 0 → cos θ = → θ = = 60◦ .
2 3

Problem 4 (Goldstein 2.21)

(a)+(b)
The laboratory frame (x, y) and rotating frame (R, r) coordinates are defined as,
y n x = R cos θ − r sin θ
(x, y) (30)
y = R sin θ + r cos θ
r ⇓

n R = x cos θ + y sin θ
(31)
r = −x sin θ + y cos θ
R
(32)
with θ = ωt .
θ = ωt x
Using lab. frame coordinates (x, y), we can write the kinetic and potential energies as,
1
T = m(ẋ2 + ẏ 2 ) , (33)
2
1 2 1 1 1
V = kr + K (R − R0 )2 = k (−x sin(ωt) + y cos(ωt))2 + K (x cos(ωt) + y sin(ωt) − R0 )2 ,
2 2 2 2
such that the Lagrangian is,
1 1 1
L = T − V = m(ẋ2 + ẏ 2 ) − k (−x sin(ωt) + y cos(ωt))2 − K (x cos(ωt) + y sin(ωt) − R0 )2 , (34)
2 2 2
The energy function or Jacobi integral is,
∂L ∂L
h= ẋ + ẏ − L = 2T − L = T + V = E , (35)
∂ ẋ ẏ
as expected since the potential energy does not depend on the velocities and the kinetic energy is
a homogeneous function of second degree in the velocities. We can then use that,
dh ∂L
=− , (36)
dt ∂t
to prove that the energy of the system is not conserved, since,
dE dh ∂L
= =− 6= 0 , (37)
dt dt ∂t
given the explicit time dependence in the Lagrangian (via θ = ωt). This was expected since, in
order to keep the steady rotational motion of the system some work is done on the system and
this goes into the balance of the mechanical energy of the system, which, by itself (i.e. without
including the work done on the system) is not conserved. Stated differently, the mechanical system
of the two springs is not isolated.
For completeness, we will also derive here the equations of motion in the lab. frame coordinates:
d ∂L ∂L
− = 0 −→ mẍ = k(y cos θ − x sin θ) sin θ − K(x cos θ + y sin θ − R0 ) cos θ (38)
dt ∂ ẋ ∂x
mẍ = kr sin(ωt) − K(R − R0 ) cos(ωt)
d ∂L ∂L
− = 0 −→ mÿ = k(−y cos θ + x sin θ) cos θ − K(x cos θ + y sin θ − R0 ) sin θ
dt ∂ ẏ ∂y
mÿ = −kr cos(ωt) − K(R − R0 ) sin(ωt)

(c)
In the (non-inertial) system of the rotating carriage, the two springs oscillate in the x and y
direction respectively, and we chose to call the corresponding coordinates R and r. The kinetic
energy is them simply given by,
1 1
T = mṙ 2 + mṘ2 . (39)
2 2
The force acting on m in the rotating frame (F ~ rot ) is,

~ rot = F
F ~ lab − m~ω × (~ω × ~r) − 2m~ω × ~v , (40)
where F~ lab is the force acting on m in the laboratory frame, ~ω is the vector angular velocity
(orthogonal to the plane of motion), while ~r = (R, r) and ~v = (Ṙ, ṙ) are the position and velocity
vectors of m in the rotating frame. The last two terms in Eq. (40) are often referred to as
centrifugal and Coriolis forces. As explained in Sec. 1.5 of Goldstein’s book, for forces Qj that
also contain terms dependent on the velocities, a generalized potential function U(qj , q̇j ) can be
defined such that it satisfies,
d ∂U ∂L
− = Qj . (41)
dt ∂ q̇j ∂qj
In our case, it is easy to show that the generalized potential function is of the form,
1 1 1
U = kr 2 + K(R − R0 )2 − mω 2 (r 2 + R2 ) − mω(Rṙ − r Ṙ) . (42)
2 2 2
In the rotating system the Lagrangian is then,
1 1 1 1 1
L = mṙ 2 + mṘ2 − kr 2 − K(R − R0 )2 + mω 2 (r 2 + R2 ) + mω(Rṙ − r Ṙ) . (43)
2 2 2 2 2
From this Lagrangian one can derive the equation of motions corresponding to the (R, r) system
of coordinates, i.e.
d ∂L ∂L
− = 0 −→ mR̈ = −K(R − R0 ) + mω 2 R + 2mω ṙ , (44)
dt ∂ Ṙ ∂R
d ∂L ∂L
− = 0 −→ mr̈ = −kr + mω 2 R − 2mω Ṙ ,
dt ∂ ṙ ∂r
and verify that they indeed corresponds to the equations of motions generated by the non-inertial
force in Eq. (40). Moreover one can verify that they correspond to the equation of motion in the
laboratory frame, when the equations in Eq. (38) are transformed to the rotating frame (using
the formalism explained in the Note at the end of this problem).
Finally, in the rotating frame the Jacobi integral is,
∂L ∂L
h = Ṙ + ṙ − L (45)
∂ Ṙ ṙ
= m(ṙ + Rω)ṙ + m(Ṙ − rω)Ṙ − L
= E − mRω(ṙ + Rω) + mrω(Ṙ − rω) .
Since L is independent of time,
dh ∂L
=− = 0, (46)
dt ∂t
but since T is not a homogeneous function of the velocities (of second degree) then h 6= E and we
cannot infer anything from the time (in)dependence of h. Indeed, as seen in (a)+(b), the energy
is not conserved and this holds always, independently of the choice of frame or coordinates. It is
possible to check this explicitly (this is not needed but it can be an extra check of the consistency
of the formalism derived in this part of the problem),
dE dh d h i
= + mRω(ṙ + Rω) − mrω(Ṙ − rω)
dt dt dt
d h i
= mω(Rṙ − Ṙr) + mω 2 (r 2 + R2 )
dt
= mω(Rr̈ − r R̈) + 2mω 2 (RṘ + r ṙ). (47)
This expression can be simplified further by using the two equations of motion:
d h i
m(Ṙ − rω) m(ṙ + Rω)ω + K(R − R0 ) = 0
dt
⇒ mR̈ = 2mṙω + mω 2 R − K(R − R0 )
d
[m(ṙ + Rω)] + m(Ṙ − rω)ω + kr = 0
dt
⇒ mr̈ = −2mω Ṙ + mω 2 r − kr . (48)
Plugging these two equations into (47) we find that many things cancel and we are left with
dE
= −kωrR − Kωr(R − R0 ) 6= 0 . (49)
dt

Note
The laboratory-frame and rotating-frame coordinates are related by a time-dependent rotation
(θ = ωt), i.e.
! ! ! ! !
x cos θ − sin θ R R t x
= =M −→ =M (50)
y sin θ cos θ r r y
and consequently,
! ! ! ! !
ẋ R Ṙ x Ṙ
= Ṁ +M =A +M , (51)
ẏ r ṙ y ṙ
where, !
0 −ω
A = ṀM t = . (52)
ω 0
In the same way we can show that,
! ! ! ! !
ẍ x ẋ Ṙ R̈
= Ȧ +A + Ṁ +M (53)
ÿ y ẏ ṙ r̈
! ! ! ! !
x 2 x Ṙ Ṙ R̈
= Ȧ +A + AM + Ṁ +M (54)
y y ṙ ṙ r̈
! ! ! !
x 2 R Ṙ R̈
= Ȧ +A M + 2Ṁ +M (55)
y r ṙ r̈
( ! ! !)
2 R Ṙ R̈
= M A + 2A + , (56)
r ṙ r̈
where we have used that Ȧ = 0, A = M t AM, AM = MA, and A2 M = AMA = MA2 . From the
previous relation we deduce that,
! ! ! ! !
ẍ R Ṙ R̈ −ω 2 R − 2ω ṙ + R̈
Mt = A2 + 2A + = , (57)
ÿ r ṙ r̈ −ω 2 r + 2ω Ṙ + r̈
and therefore we show that the equations of motion in the laboratory frame (see Eq. (38)) transform
into the equations of motion on the rotating frame (see Eq. (44)), since,
! ! !
t ẍ −K(R − R0 ) −ω 2 R − 2ω ṙ + R̈
mM = =m . (58)
ÿ −kr −ω 2 r + 2ω Ṙ + r̈
2 Non-graded Problems
Problem 5 (Goldstein 2.2)
Consider a rotation about an arbitrary axis that we identify with
ẑ. This corresponds to the rotation angle φ. Let us first consider
the one-particle case; if V (ẋ, ẏ, ż) then
z

∂V ∂V ∂ ẋ ∂V ∂ ẏ ∂V ∂ ż
= + +
∂ φ̇ ∂ ẋ ∂ φ̇ ∂ ẏ ∂ φ̇ ∂ ż ∂ φ̇
~r ∂V ∂V
= (−r sin θ sin φ) + (r sin θ sin φ)
∂ ẋ ∂ ẏ
= (~r × ∇~ v V (~r˙ )) · ẑ = ~n · (~r × ∇ ˙
~ v V (~r, ~r).
θ
In the above we have defined ~n := ẑ, and we have used the coor-
dinates
y
( x = r sin θ cos φ
y = r sin θ sin φ
φ z = r cos θ
( ẋ = ṙ sin θ cos φ + r θ̇ cos θ cos φ + r φ̇ sin θ sin φ
x ẏ = ṙ sin θ sin φ + r θ̇ cos θ sin φ + r φ̇ sin θ cos φ
ż = ṙ cos θ − r φ̇ sin φ.
Therefore we find
∂L ∂T ∂V X
~ v V ).
pφ = = − = lφ − ~n · (~ri × ∇ i
∂ φ̇ ∂ φ̇ ∂ φ̇ i

The new term ∂V /∂ φ̇ only exists for V = V (~r), and we have also extended the result to include
many particles. For electromagnetic forces the potential is
q~
V = qΦ − A · ~v
c
so the gradient is
~ v V = − q A.
∇ ~
c
Plugging this in gives us the final result;
qi ~
X  
p φ = lφ − ~n · ~ri × A .
i c
Problem 6 (Goldstein 2.4)
On a spheroidal surface of radius R:
( x = R sin θ cos φ
y = R sin θ sin φ
z = R cos θ
The arc length is given by
q A
ds2 = R2 dθ2 + R2 sin2 θdφ2 ⇒ ds = R dθ2 + sin2 dφ2 . θA

θB B
Thus the distance between two points (A,B) is given by
Z B Z B q
lAB = ds = R dθ2 + sin2 θdφ2 φA
A
v A φB
u !2
φB dθ
Z u
= R dφ sin2 θ +
t
.
φA dφ

Let us define the integrand as

q
f (θ, θ′ ) = sin2 θ + θ′2 .

The variational principle tells us that the minimal dis-

tance is obtained when θ, θ′ satisfy the equation
d ∂f ∂f
− =0
dθ ∂θ′ ∂θ
!
d θ′ sin θ cos θ
√ −√ 2 =0
dθ sin θ + 2
θ′2
sin θ + θ′2
θ′′ θ′ sin θ cos θθ′ + θ′ θ′′ sin θ cos θ
− 2 − = 0.
f f f f
(θ′′ − sin θ cos θf2 − θ′2 (sin θ cos θ + θ′′ ) = 0
(θ′′ − sin θ cos θ)(θ′2 + sin2 θ) − θ′2 (sin θ cos θ + θ′′ ) = 0
θ′′ sin2 θ − 2θ′2 sin θ cos θ − sin3 θ cos θ = 0
h i
sin θ θ′′ sin θ − 2θ′2 cos θ − sin2 θ cos θ = 0
⇒ θ′′ sin2 θ − 2θ′2 cos θ − sin2 θ cos θ = 0. (59)
Now let’s work a little backwards. Great circles (which is what we are trying to prove being the
shortest path between two points on a sphere) are the intersections between a sphere and a plane
going through these two points. If

n̂ = nx x̂ + ny ŷ + nz ẑ

is the vector perpendicular to the plane, the points in the great circle are those such that

n̂ · ~r = 0
 nx R cos θ cos φ + ny R sin θ cos φ + nz R cos θ = 0
R [sin θ(nx cos φ + ny sin φ) + nz cos θ] = 0
Thus we find the condition satisfied for points on a circle is
cos θ
= A cos φ + B sin φ.
sin θ
Now define a function
cos θ(φ)
g(φ) := .
sin θ(φ)
Find the second derivative of this quantity:
! !
d2 g d dg dθ d − sin θ − cos2 θ ′
= = θ
dφ2 dφ dθ dφ dθ sin2 θ
d 1
 
= − 2 θ′
dφ sin θ
2 cos θ ′2 1 ′′
= 2 θ − θ
sin θ sin2 θ
1 2 cos θ ′2
= − 2 θ′′ + θ
sin θ sin2 θ
1 h i
= − 3 θ′′ sin θ + 2 cos θθ′2 .
sin θ
Our minimization condition (59) tell us that the quantity in brackets must be equal to sin2 θ cos θ,
so we have
d2 g 1 2 cos θ d2 g
= − sin θ cos θ = − = −g ⇒ +g =0
dφ2 sin3 θ sin θ dφ2
⇒ g(φ) = A cos φ + B sin φ.
This is exactly the equation for great circles on a sphere.