Princeton University

Ph205
Mechanics
Problem Set 3
Kirk T. McDonald
(1988)
[email protected]
http://kirkmcd.princeton.edu/examples/
Princeton University 1988 Ph205 Set 3, Problem 1 1

1. We wish to slide objects, initially at rest, down a straight, frictionless chute that begins
at a vertical walls and ends at distance d from the wall. At what angle θ to the vertical
should the chute be placed to minimize the time of descent?

If the chute were bent into a cycloid, the descent would be the fastest possible. It is
agreeable to our intuition that a cusp of the cycloid must be at the wall, and the end
point at the bottom of the cycloid. For a derivation, see sec. 3-11 of R. Weinstock,
Calculus of Variations (McGraw-Hill, 1952),
http://kirkmcd.princeton.edu/examples/mechanics/weinstock_52.pdf
taking the reference curve to be the wall, x = 0.
Compare the times of descent for a straight chute at the angle θ found above, and for
the cycloidal chute described above.
Princeton University 1988 Ph205 Set 3, Problem 2 2

2. A curve y(x) is rotated about the y-axis to make a surface of revolution bounded by
two disks. What form of y(x) produces the minimum surface area?

Based on our experience with the case of rotation of the curve about the x-axis, p. 55
of http://kirkmcd.princeton.edu/examples/Ph205/ph205l5.pdf, we might expect the curve
to be a catenary, of the form x = A cosh(y − B).
Princeton University 1988 Ph205 Set 3, Problem 3 3

3. (a) Hanging Rope

A rope of length L and uniform mass is hung from two fixed points at the same
height, separated by distance D. Use elementary methods to examine the con-
dition of static equilibrium for a segment of the rope dx long (with the √ x-axis

horizontal, y-axis vertical). First look at Fx , then Fy , to show that y / 1 + y 2
= constant, ⇒ y(x) is a catenary.

(b) Suspension Bridge

A massless cable is strung between two fixed points at the same height, separated
by distance D, and (massless) vertical cables are attached to it to cary a uniform
horizontal load – the bridge. What is the shape of the curving cable such that
the tensions are equal in all vertical cables?

Assume an infinite number of evenly spaced vertical cables.

Use elementary methods to show that y  = constant, ⇒ parabola.
Princeton University 1988 Ph205 Set 3, Problem 4 4

4. (a) Hanging Rope

Use the calculus of variations to show that the form of a hanging rope, Prob. 3(a),
is a catenary.
(b) Suppose the rope of part (a) has one end attached to a fixed point while the other
end drapes over a fixed, frictionless peg at the same height. What is the length l
of the vertical segment of the rope, assuming no friction?
Princeton University 1988 Ph205 Set 3, Problem 5 5

5. Geodesics on a Sphere
Full credit for working either part (a) or (b).
Find the curve on the surface of a sphere that is the shortest distance between two
points.

(a) One approach is to reduce the problem to 2 dimensions. Parameterize the sur-
face by two independent coordinates (u, v). Then, (x(u, v), y(u, v), z(u, v)) is the
surface. Use the equation of the surface of the sphere to find functions P (u, v),
Q(u, v), R(u, v) such that an element of arc length on the sphere can be written
as ds2 = dx2 + dy 2 + dx2 = P du2 + 2Q du dv + R dv 2 .
A 1-dimensional
√ curve on the surface of the sphere can be written as v = v(u) so
ds = P + 2Qv  + Rv 2 du.
Try parameters a = (constant) radius of sphere, u = φ = azimuthal angle, v = θ
= polar angle. You should eventually find a solution of the form,
z
x sin C1 + y cos C1 −  = 0, (1)
a2 /C22 − 1

which is the equation of a plane passing through the center of the sphere. The
intersection of this plane with the surface of the sphere is a great circle, which is
the geodesic on a sphere.
(b) Stay in 3 dimensions and regard the surface of the sphere as a constraint. Let
(x(t), y(t), z(t)) be the desired curved, and g(x, y, z) = 0 be the equation of the
surface
√ of the sphere. Then, ds = f dt is an element of arc length, with
f = ẋ + ẏ 2 + ż 2 .
2

Use the calculus of variations to show that,

     
d ẋ d ẏ d ż d2 x d2 y d2 z
dt f dt f dt f ds2 ds2 ds2
= = ⇒ = = . (2)
∂g/∂x ∂g/∂y ∂g/∂z ∂g/∂x ∂g/∂y ∂g/∂z

Referring to p. 12 of http://kirkmcd.princeton.edu/examples/Ph205/ph205l1.pdf, the

numerators are components of a vector normal to the curve (i.e., n̂ ∝ dŝ/ds). The
denominators are components of ∇g, which is normal to the surface. The equality
of the ratios is consistent with the geometric picture of a Lagrange multiplier given
on pp. 58-89 of http://kirkmcd.princeton.edu/examples/Ph205/ph205l5.pdf.
For a sphere use the left form of eq. (2) to find the geodesic curve.
A trick is to write f/f˙ two ways, giving expressions that can be integrated to
yield logarithms. Rearrange and integrate again to show that x + Ay + Bz = 0
as in part (a).
Princeton University 1988 Ph205 Set 3, Problem 6 6

6. Meanders
A flexible tape is bent into a planar curve by fixing the positions
and slopes at its endpoints. The tape has length l0 and thickness
h0 when undeformed, and spring constant k for linear stretching
along its length.
In the bend, suppose the length along the midline remains l0,
and the thickness h0 , while above the midline (for a bend as in
the figure) the tape is stretched, while below the midline it is
compressed.
The tape assumes whatever shape requires the least work of
deformation. That is, the stored potential energy will be a min-
imum.
Divide the wedge shown in the lower figure into slices, each a tiny
spring. Deduce the spring constant of each slice, and integrate
to show that the total potential energy is,
 l 0  2
kl0h20 dθ
PE = dl. (3)
24 0 dl

We wish to minimize this, subject to a suitable constraint. But, noting that the length
of the tape is l0 won’t help here. Instead, consider that the distance D between the
endpoints is fixed,
 l0
D= cos θ dl = constant. (4)
0

The other constraints, on the angles θ1 and θ2 , can be applied once the general shape
is known.
Use the calculus of variations to find θ(l). For small θ1 , show that θ ≈ θ 1 cos(kl) for
some constant k.
To get a sense for the more exact solution at large θ1, Sketch 1 or 2 periods of the
curve for θ1 = 120◦ . This curve appears on maps – as the shape of meandering rivers.
See, for example,
http://kirkmcd.princeton.edu/examples/mechanics/leopold_sa_214-6_60_66.pdf
http://kirkmcd.princeton.edu/examples/mechanics/einstein_natur_14_223_26_english.pdf
Princeton University 1988 Ph205 Set 3, Problem 7 7

7. Components of Acceleration in a Non-Cartesian Coordinate System

Suppose we use an orthogonal coordinate system (q1, q2, q3) to describe the position of
a point particle. Then, we can write the line element as,

ds2 = ds21 + ds22 + ds23 , where dsi = fi (q1, q2, q3 ) dqi . (5)

The velocity v is just v = ds/dt, so the kinetic energy is,

 2 ⎡  2  2  2 ⎤
1 1 ds 1 dq1 dq2 dq3
T = mv 2 = m = m ⎣f12 + f22 + f32 ⎦. (6)
2 2 dt 2 dt dt dt

Lagrange’s equations of motion provide a quick way to deduce the components of

acceleration in this coordinate system.
One form is,
d ∂T ∂T
− = Qj , (7)
dt ∂ q̇j ∂qj

where for small displacements the generalized force Qj obeys,


Qj δqj = F · δr. (8)
j

If we decompose vectors F and δr in our new coordinate system, we can write,

F = F1 q̂1 + F2 q̂2 + F3 q̂3, and δr = δr1 q̂1 + δr2 q̂2 + δr3 q̂3. (9)

But, by definition of the line element,

 
δrj = δsj = fj δqj , such that Qj δqj = Fj fj δqj , (10)
j j

so,

Qj = Fj fj , (11)

relates the generalized force Q to the ordinary force F.

Meanwhile, Newton tells us that Fj = maj , so the component of the acceleration in
the q̂j direction is,

Qj
d ∂T
dt ∂ q̇j
− ∂T
∂qj
aj = = . (12)
mfj mfj

Use these tricks to compute the forms of the aj in cylindrical and spherical coordinates
systems. Compare with the results given on pp. 9-10 of the Ph205 Lecture Notes.
Princeton University 1988 Ph205 Set 3, Problem 8 8

8. Find the constraint force on a mass m slides without friction along a rod that is
constrained to rotate in a plane with constant angular velocity ω (Ph205 Set 5, Prob. ).
Use both elementary methods, and the method of Lagrange multipliers.
Princeton University 1988 Ph205 Set 3, Problem 9 9

9. The Return of the One-Legged Ice Skater

Consider the model of a one-legged ice skater introduced on p. 70 of the Notes, but
now suppose that the constraint force F is not applied at the center of mass, but at
distance a from it. Recall that the motion of the skate consists of a slide along the
direction of the skate combined with a rotation about the point of application of the
constraint force. We still have that F · vc.m. = 0, so the constraint force does no work,
and the kinetic energy of the ice skater is constant. If vc.m. decreases, θ̇ increases, and
the possibility exists that the skater comes to rest, with a higher angular velocity.

Find the equations of motion of the ice skate by elementary methods, or by Lagrange’s
method (using a Lagrange multiplier to include the non-holonomic constraint force).1
To integrate the equations of motion, you may wish to replace ẋc.m. and ẏc.m. by func-
tions of vF and θ where vF is the velocity of the point of application of the constraint
force (rather than vc.m.). Write the moment of inertia about the center of mass as
I = mb2 and let k 2 = 1 + b2 /a2 . In general, k is not constant, but it suffices to suppose
that it is.
You should find that,
2
ak 2θ̈ + vF θ̇ = 0, and v̇F = aθ̇ . (13)

Differentiate and combine, then multiply by θ̈/θ̇ to find,

 2 2
d 2 θ̈ dθ̇
k =− . (14)
dt θ̇ dt

Integrate twice (and then a third time) to show that,

c θ
θ̇ = = c cos , (15)
cosh(ct/k) k
where c is a constant. Hence,
ct θ
vF = ack tanh = ack sin , (16)
k k
1
The term holonomic was introduced by Hertz on p. 91 of Die Prinzipien der Mechanik (Barth, Leipzig,
1894), http://kirkmcd.princeton.edu/examples/mechanics/hertz_mechanik_94.pdf
Princeton University 1988 Ph205 Set 3, Problem 9 10

which implies that vF and θ go to constant values as t → ±∞.

Show that at t = 0, dxF /dθ = dyF /dθ = d2 yF /dθ 2 = 0, while d2 xF /dθ 2 = 0 and
d3 yF /dθ 3 = 0. This implies that the trajectory has a cusp at the origin at t = 0,
pointing in the negative-x direction.
Some possible trajectories are shown below,2

Can real ice skaters produce these shapes?

You should be able to verify that the kinetic energy is,

ma2c2 k 2 c2 I 2θ
T = , and that F = sin . (17)
2 2k k

2
C. Carathéodory, Der Schlitten, Z. Angew. Math. Mech. 13, 71 (1933),
http://kirkmcd.princeton.edu/examples/mechanics/caratheodory_zamm_13_71_33.pdf
A. Sommerfeld, Mechanics (Academic Press, 1952), p. 251,
http://kirkmcd.princeton.edu/examples/mechanics/sommerfeld_mechanics_52.pdf
Princeton University 1988 Ph205 Set 3, Problem 10 11

10. (a) A particle starts from rest at the top of a frictionless sphere of radius a and slides
down. When it flies off, the normal force vanishes. Use the method of Lagrange
multipliers to show that this occurs when cos θ = 2/3, where θ is the polar angle
of the particle.
This problem was posed on a Ph103 Learning Guide, sans Lagrange.

(b) A uniform sphere of radius b starts at rest from the top of a fixed sphere of radius
a, and rolls without slipping down the latter. Show by any method that the upper
sphere flies off when cos θ = 10/17 = 0.59, so at a larger angle than for case (a).

For the more complicated case where the lower sphere/cylinder can roll without
slipping, see http://kirkmcd.princeton.edu/examples/2cylinders.pdf
Perhaps surprisingly, there are cases where the inital rotations of the two spheres/cylinders
have opposite senses, but change to the same sense before the upper sphere/cylinder
flies off the lower.
Princeton University 1988 Ph205 Set 3, Solution 1 12

Solutions

1. The acceleration of an object down a frictionless chute at angle θ to the vertical is

a = g cos θ, where g is the acceleration due to gravity. If the chute ends at distance d
from the vertical wall, its length is l = d/ sin θ. The time of the descent (from rest at
the top of the chute) is related by l = at2/2, i.e.,

4d dt 8d cos 2θ
t2 = , 2t =− , (18)
g sin 2θ dθ g sin2 2θ

so the time is minimal when dt/dθ = 0, i.e., for cos 2θ = 0, θ = 45◦ . The time of
descent is,

d
t=2 . (19)
g

The least time of descent from the wall to a point at distance x = d from is obtained a
cycloid, with a cusp at the top of the (curved) chute, say (x, y) = (0, 0), and its bottom
at (d, −2b). This cycloid can be parameterized by,

x = b(φ − sin φ), y = −b(1 − cos φ). (20)

The bottom of the cycloid is at φ = π, such that b = d/π.

When
 the object
 is at height y < 0, after starting from rest at y = 0, its speed is
v = 2g |y| = 2gb(1 − cos φ).
The time to slide down a segment ds of the cycloid at height y is,
√ 2 
ds dx + dy 2 b (1 − cos φ)2 + sin2 φ b d
dt = = =  dφ = dφ = dφ. (21)
v 2gb(1 − cos φ) 2gb(1 − cos φ) g πg

The total travel time from φ = 0 to π is,

 π
dt πd d d
t= dφ = = 1.77 <2 . (22)
0 dφ g g g

However, the descent using a cycloid is only 11% faster than that using a 45◦ straight
chute.
Princeton University 1988 Ph205 Set 3, Solution 2 13

2. The surface of revolution of curve y(x) about the y-axis has area,
 x2  
A = 2π x 1 + y 2 dx = 2π f(x, y ) dx. (23)
x1

To minimize the area, we consider the Euler-Lagrange equation for the function f,

∂f d ∂f d xy 
=0= = √ . (24)
∂y dx ∂y  dx 1 + y 2
Hence,

xy  C x
√ = C, y = √ , y = cosh−1 +B, x = C cosh(y − B),(25)
1 + y 2 − C2
x2 C
using Dwight 260.01.
Princeton University 1988 Ph205 Set 3, Solution 3 14

3. (a) Hanging Rope

We consider an element of the hanging rope, as sketched below.

For static equilibrium, the horizontal force equation is,

d
T (x) cos θ(x) = T (x + dx) cos θ(x + dx) ≈ T (x) cos θ(x) + (T cos θ) dx, (26)
dx
which tells us that,

d(T cos θ) = 0, T cos θ = T0. (27)

Similarly, the vertical force equation is, with tan θ = dy/dx = y  ,

Mg 
1 + y 2 dx = T (x + dx) sin θ(x + dx) − T (x) sin θ(x)
L
d dy 
= T0 tan θ(x + dx) − T0 tan θ(x) ≈ T0 (tan θ) dx = T0 dx = T0 y  dx, (28)
dx dx
y  Mg
√ 2
= = constant. (29)
1+y LT0
This the differential equation for a catenary = shape of the hanging rope.
(b) Suspension Bridge
The analysis for part(a) holds for a suspension bridge as well, except that the left
side of eq. (28) is just Mg dx/L for a bridge of length M and mass M which is
uniform in x. Hence y  = constant, and the shape of the long supporting cable is
a parabola.
Princeton University 1988 Ph205 Set 3, Solution 4 15

4. (a) Hanging Rope

To analyze the hanging rope using the calculus of variations, we minimize the
potential energy,
  
V = ρgy 1 + y 2 dx = ρg F dx, (30)

where ρ is the mass density of the rope, subject to the constraint on the length
of the rope,
  
L= 1 + y 2 dx = G dx. (31)

We consider the “Lagrangian” function F ∗(x, y, y ) = ρgF + ρgλG, where λ is a

to-be-determined Lagrange multiplier (written so that we can ignore the constant
factor ρg in the following).
Since ∂F ∗/∂x = 0, we know there exists a conserved (constant) quantity, the
“Hamiltonian”,
∗
y 2(y + λ) 
 ∂F ∗
H =y −F = √ − (y + λ) 1 + y 2, (32)
∂y  1 + y 2

H 1 + y 2 = y 2(y + λ) − (y + λ)(1 + y 2) = −(y + λ), (33)

 dy (y + λ)2 − H 2
y = = , (34)
dx H

dy y+λ
x=H  = H cosh−1 + C, (35)
(y + λ)2 − H 2 H
x−C
y = H cosh − λ, (36)
H
using Dwight 260.01, which is the form of a catenary.
If we had not noticed the “conservation law” (32), use of the Euler-Lagrange
equation would have led to a second-order differential equation, and the same
result (36) after somewhat greater effort. See the handwritten solution at
http://kirkmcd.princeton.edu/examples/Ph205/ph205sol3.pdf
The shape of the rope is symmetric about the line x = D/2, so C = D/2.
y(0) = 0 ⇒ λ = H cosh(D/2H).
H can be determined from the transcendental equation,
 D  D/2
x D
L= 1+ y 2 dx = cosh dx = 2H sinh . (37)
0 −D/2 H 2H

(b) For the rope of total length L with length l < L hanging over a peg with one end
fixed, we minimize the potential energy,
 D  D 
l ρl2g
PE = ρgy dl − ρlg = ρgy 1 + y 2 dx − , (38)
0 2 0 2
Princeton University 1988 Ph205 Set 3, Solution 4 16

subject to the constraint on the length,

 D  D
L= dl + l = 1 + y 2 dx + l. (39)
0 0

For the calculus of variations, we consider,

 D   D
∗  ρl2g
F (x, y, y , l) = ρgy 1 + y 2 dx − + λρg 1 + y 2 dx + λl, (40)
0 2 0

where λ is a Lagrange multiplier to be determined.

The variational analysis in the independent variable l is simply dF ∗/dl = 0, which
implies that λ = l.
The variational analysis in y is now the same as in part (a), so,

x−C x − D/2
y = H cosh − λ = H cosh − l. (41)
H H
We can use eq. (35) with (x, y) = (0, 0) to find,

D
λ = l = H cosh . (42)
2H
The length calculation now gives the transcendental equation,
 D
D D
L − l = L − H cosh = 1 + y 2 dx = 2H sinh . (43)
2H 0 2H
H is less than in part (a).
Princeton University 1988 Ph205 Set 3, Solution 5 17

5. (a) We write the surface of the sphere of radius a as x = (x(θ, φ), y = y(θ, φ, z =
z(θ), φ) where θ and φ are the polar and azimuthal angles of a spherical coordinate
system with z as the polar axis. An element of arc length on the surface is,
ds2 = dx2 + dy 2 + dz 2 . (44)
With,
 2  2
∂x ∂x 2 ∂x 2 ∂x ∂x ∂x
dx = dθ + dφ, dx = dθ + 2 dθ dφ + dφ2 , (45)
∂θ ∂φ ∂θ ∂θ ∂φ ∂φ
etc., we can write,
ds2 = P dθ 2 + 2Q dθ dφ + R dφ2, (46)
where,
 2  2  2
∂x ∂y ∂z
P = + + , (47)
∂θ ∂θ ∂θ
∂x ∂x ∂y ∂y ∂z ∂z
Q= + + (48)
∂θ ∂φ ∂θ ∂φ ∂θ ∂φ
 2  2  2
∂x ∂y ∂z
R= + + . (49)
∂φ ∂φ ∂φ
On the surface of the sphere,
x = a sin θ cos φ, y = a sin θ sin φ, z = a cos θ, (50)
so,
P = a2 sin2 θ, Q = 0, R = a2 . (51)
and for a curve θ = θ(φ) on the surface of the sphere, its arc length is,
  φ 
2
L= ds = a sin2 θ + θ2 dφ, (52)
φ1

where θ = dθ/dφ. To minimize L, we consider the function F (φ, θ ) = sin2 θ + θ2
in the Euler-Lagrange analysis, with φ as the independent variable.
Since ∂F/∂φ = 0, the “Hamiltonian” H is a conserved quantity,

∂F  θ2
H=φ − F =  − sin2 θ + θ2. (53)
∂θ sin2 θ + θ2

sin θ = H sin2 θ + θ2,
2
sin4 θ = H 2 (sin2 θ + θ2 ), (54)

dθ sin4 θ dθ
θ = = − sin2 θ, dφ =  (55)
dφ H2 sin θ sin2 θ
−1
H2

cos θ
dφ = φ + C = − tan−1  , (56)
sin2 θ
H2
−1
Princeton University 1988 Ph205 Set 3, Solution 5 18

using 2.599.6 of Gradshteyn and Ryzhik,

https://physics.princeton.edu/~mcdonald/examples/mechanics/gradshteyn_80.pdf.
Then, recalling that identity that,
a a
tan−1 = sin−1 √ 2 , (57)
b a + b2
we have that
cos θ cot θ
φ + C = − sin−1  2
= − sin−1  , (58)
1
cos2 θ + sin
H2
θ
− 1 H 2 − 1
cot θ
sin(φ + C) = sin φ cos C + cos φ sin C =  . (59)
1
H2
−1

Multiplying this by a sin θ, we obtain an equation for the curve as a function of

(x, y, z),

a cosθ z
0 = a sin θ(sin φ cos C + cos φ sin C) −  x cos C + y −  , (60)
1 1
H 2 − 1 H 2 − 1

which is a plane passing through the origin. The geodesic curve is the intersection
of this plane with the surface of the sphere, i.e., a great circle.
(b) For an analysis that uses the three spatial coordinates (x, y, x), we consider a
curve (x(t), y(t), z(t)) whose length is,
 t2 
L= F dt, where F = ẋ2 + ẏ 2 + ż 2 , (61)
t1

with the constraint,

G(x, y, z) = x2 + y 2 + z 2 − a2 = 0, (62)

that the curve lie on a sphere of radius a centered on the origin. For the Euler-
Lagrange method we consider the function,

F ∗(x, y, y, ẋ, ẏ, ż; t) = F + λG, (63)

where λ is a Lagrange multiplier. Then,

d ∂F ∗ d ẋ ẍ ẋḞ ∂F ∗ Ḟ
= = − 2 = = 2λx, ẍ = ẋ + 2λxF, (64)
dt ∂ ẋ dt F F F ∂x F
together with similar equations for y and z.
We now perform a “hat trick”, and introduce the 3 quantities,

Jx = y ż − z ẏ, Jy = z ẋ − xż, Jz = xẏ − y ẋ, (65)

Princeton University 1988 Ph205 Set 3, Solution 5 19

Their derivatives are, using all three versions of eq. (64),

Ḟ Ḟ Ḟ
J˙x = ẏż + y z̈ − ż ẏ − z ÿ = y ż + 2λyzF − z ẏ − 2λzyF = Jx , (66)
F F F
Ḟ Ḟ
J˙y = Jy , J˙z = Jz . (67)
F F
We can integrate these to find,

ln Ji = ln F + Ci , Ji = Di F, (68)

where the Di are constants. Another trick is to note that,

xJx + yJy + zJz = x(y ż − z ẏ) + y(z ż − xż) + z(xẏ − y ẋ) = 0

= xDx F + yDy F + zDz F. (69)

Finally, we obtain,

Dx x + Dy y + Dz z = 0, . (70)

which is the equation of a plane that passes through the origin. Again, the
geodesic curve is the intersection of such a plane with the sphere = a great circle.
Princeton University 1988 Ph205 Set 3, Solution 6 20

6. The elastic tape an unstretched length l0 , thickness h0 , and spring constant k for
stretching along its length.
If the tape is stretched by force F , an element of the tape of size dh × dl is stretched
by force,
dh
dF = F , (71)
h0

The entire tape is stretched by length Δl = F/k, while the small element is stretched
by,
dl dF
Δ(dl) = Δl =  , (72)
l0 k
where k  is the spring constant of the element,
dF dh l0 dh l0 dh l0
k = =F = kΔl =k . (73)
Δ(dl) h0 Δl dl h0 Δl dl h0 dl

A segment of length dl of the bent tape is illustrated in the figure below, assuming
that the bend lies in a plane. We subdivide this segment into slices of thickness dh, at
height h relative to the centerline of the tape. The stretch/compression of such a slice
is,
 
dl
Δ(dl) = dθ + h − dl = h dθ, (74)
dθ
so the potential energy stored in the slice is,
 2
1 dh l0 dh l0 2 2 kl0 dθ
d PE = k  Δ2(dl) = k
2
h dθ = dh dl. (75)
2 h0 dl h0 dl 2h0 dl

Integrating over the slices in the segment of length dl,

 2  h0 /2  2
kl0 dθ kh2l0 dθ
dPE = dl h dh = 0
2
dl, (76)
2h0 dl −h0 /2 24 dl
and the total potential energy in the bent tape is,
 2
kh2 l0  l0 dθ
PE = 0 dl. (77)
24 0 dl
Princeton University 1988 Ph205 Set 3, Solution 6 21

We can now consider the shape of the bent tape as described by the function θ(l), and
minimize the potential energy subject to the constraint that the ends of the tape are
separated by distance,

D= cos θ dl, (78)

for angle θ measured with respect to the line joining the endpoints of the tape. The
function F is F (θ; l) = θ2 and the constraint function is G(θ; l) = cos θ, so we use the
calculus of variations for the combined function F ∗ θ2 + λ cos θ, where λ is a Lagrange
multiplier. The Euler-Lagrange equation for this is,
d ∂F ∗  ∂F ∗ λ
= 2θ = = −λ sin θ, θ + sin θ = 0, (79)
dl ∂θ ∂θ 2
which equation is familiar from the simple pendulum at large amplitudes.

For small amplitudes, an oscillatory solution has the form,

⎛ ⎞
λ ⎠
θ = θmax cos ⎝ l . (80) 3
2

While the form (80) does not hold exactly for large θmax, it gives
a sense of the solution. The figure to the right is for θmax = 110◦ .

3
http://kirkmcd.princeton.edu/examples/mechanics/leopold_gspp_422-h_66.pdf
4
http://kirkmcd.princeton.edu/examples/mechanics/leopold_sa_214-6_60_66.pdf
Princeton University 1988 Ph205 Set 3, Solution 7 22

7. In an orthogonal coordinate system (q1, q2, q3 ), the line element has the form,

ds2 = f12 dq12 + f22 dq22 + f32 dq32. (81)

In cylindrical coordinates (r, φ, z), the line element is,

ds2 = dr2 + r2 dφ2 + dz 2 , (82)

so, fr = 1, fφ = r, fz = 1. (83)

The kinetic energy of a point particle is,

⎡  2  2  2 ⎤
m ⎣ 2 dq1 dq2 dq3 m 2 2
T = f + f22 + f32 ⎦ = (ṙ + r2 φ̇ + ż 2), (84)
2 1 dt dt dt 2

The components of the acceleration are given by,

d ∂T
dt ∂ q̇j
− ∂T
∂qj
aj = , (85)
mfj
2 r2 φ̈ + 2rṙ φ̇
ar = r̈ − rφ̇ , aφ = = rφ̈ + 2ṙ φ̇, az = z̈. (86)
r

In spherical coordinates (r, θ, φ), the line element is,

ds2 = dr2 + r2 dθ 2 + r2 sin2 θ dφ2 , (87)

so, fr = 1, fθ = r, fφ = r sin θ. (88)

The kinetic energy of a point particle is,

m 2 2 2
T = (ṙ + r2 θ̇ + r2 sin2 θ φ̇ ), (89)
2
The components of the acceleration are given by eq. (85) as,
2 2
ar = r̈ − 2rθ̇ − r sin2 θ φ̇ , (90)
2
r2 θ̈ + 2rṙ θ̇ − r2 sin θ cos θ φ̇ 2
aθ = = rθ̈ + 2ṙ θ̇ − r sin θ cos θ φ̇ , (91)
r
2 2 2 2
r sin θ φ̈ + 2rṙ sin θ φ̇ + 2r sin θ cos θ θ̇φ̇
aφ = = r sin θ φ̈ + 2ṙ sin θ φ̇ + 2r cos θ θ̇φ̇.(92)
r sin θ
Princeton University 1988 Ph205 Set 3, Solution 8 23

8. The constraint force F is parallel to the velocity component vθ = ω × r, and F · vθ

equals the rate of change of kinetic energy of the sliding mass m,

mv 2 1 1
T = = m(vr2 + vθ2) = m(ṙ2 + ω 2 r2 ), (93)
2 2 2
dT
F · vθ = F ωr = = m(ṙr̈ + ω rṙ) = 2mω 2 rṙ,
2
(94)
dt
F = 2mω ṙ, (95)

recalling that the centripetal acceleration is r̈ = ω 2 r.

Taking the generalized coordinates of mass m to be r and θ, the Lagrangian is,
 
1 2
L = T = m ṙ2 + r2 θ̇ , (96)
2

subject to the constraint g = θ̇−ω = 0. We define ar = ∂g/∂r = 0 and aθ = ∂g/∂θ = 1.

Then, in Ferrers’ variant of the method of Lagrange,5 we associate a Lagrange multiplier
λ with the constraint force, and write the equations of motion as,
d ∂L ∂L 2
− = mr̈ − mrθ̇ = Qr = λar = 0, (97)
dt ∂ ṙ ∂r
d ∂L ∂L
− = mr2θ̈ + 2mrṙ θ̇ = Qθ = λaθ = λ. (98)
dt ∂ θ̇ ∂θ
where Qθ = λ is the generalized constraint force associated with coordinate θ. With
θ̇ = ω = constant, θ̈ = 0, we obtain r̈ = ω 2 r and,

Qθ = 2mωrṙ. (99)

This generalized force is a torque, r times the constraint force F = 2mω ṙ.

5
N.M. Ferrers, Extension of Lagrange’s Equations, Quart. J. Pure Appl. Math. 12, 1 (1872),
http://kirkmcd.princeton.edu/examples/mechanics/ferrers_qjpam_12_1_72.pdf
Princeton University 1988 Ph205 Set 3, Solution 9 24

9. We describe the skater by three coordinates, (x, y) of its center of mass, and the angle
θ of the skate to the x-axis, as shown in the figure below.

To deduce
√ 2 a constraint on the motion when the center of mass of the skater moves by
2
dl = dx + dy , we have that,

dx = dl cos(θ + dα) ≈ dl cos θ − dl dα sin θ ≈ dl cos θ − a sin θ dθ, (100)

dy = dl sin(θ + dα) ≈ dl sin θ + dl dα cos θ ≈ dl sin θ + a cos θ dθ, (101)

using the law of sines for the small triangle,

dα sin(dα) sin(π − dθ) dθ
≈ = ≈ . (102)
a a dl dl
Then, the non-holonomic constraint is,
dx + a sin θ dθ dy − a cos θ dθ
dl ≈ ≈ , (103)
cos θ sin θ
dg(x, y, θ) = sin θ dx − cos θ dy + a dθ = 0. (104)

An alternative derivation of the constraint notes that the constraint force F does no
work, so

0 = F · vF = F · (vc.m. − aθ̇ F̂) = F (− sin θ ẋ + cos θ ẏ − a θ̇), (105)

from which eq. (104) follows (more quickly).

We use Ferrers extension of Lagrange’s method, with Lagrange multiplier λ,
1 2 dg dg dg
L = T = (mẋ2 + mẏ 2 + I θ̇ ), ax = = sin θ, ay = = − cos θ, aθ = = a, (106)
2 dx dy dθ
d ∂L ∂L
− = λaj , (107)
dt ∂ q̇j ∂qj
mẍ = λ sin θ, mÿ = −λ cos θ, I θ̈ = λa. (108)

Note that these equations also follow from elementary methods, with the constraint
force F = −λ.
Whatever method used to obtain eq. (108), to go further it is useful to write the
constraint (104) as

ẋ sin θ − ẏ cos θ + a θ̇ = 0. (109)

Princeton University 1988 Ph205 Set 3, Solution 9 25

We also introduce the velocity vF of the point of application of the constraint force,
xF = x − a cos θ, yF = y − a sin θ, (110)
ẋF = ẋ + a sin θ θ̇ = vF cos θ, ẏF = ẏ − a cos θ θ̇ = vF sin θ, (111)
v = ẋ cos θ + ẏ sin θ, (112)
2
v̇F = ẍ cos θ + ÿ sin θ + (−ẋ sin θ + ẏ cos θ) θ̇ = aθ̇ , (113)
where the last equality is obtained using eqs. (108)-(109). For later use we note that,
2 2
vF2 = ẋ2F + ẏF2 = ẋ2 + ẏ 2 + a2θ̇ + 2(a sin θ ẋ − a cos θ ẏ) θ̇ = ẋ2 + ẏ 2 − a2 θ̇ . (114)

Furthermore, we can differentiate the constraint (108) with respect to time,

ẍ sin θ − ÿ cos θ + a θ̈ + (ẋ cos θ + ẏ sin θ) θ̇ = 0, (115)
ẍ sin θ − ÿ cos θ + a θ̈ + (ẋ cos θ + ẏ sin θ) θ̇ = 0, (116)
I θ̈
+ aθ̈ + vF θ̇ = 0, (117)
ma
using eqs. (108)-(109) again. We also write,
b2
I = mb2, k2 = 1 + , (118)
a2
such that the equation of motion (117) becomes,
θ̈
ak 2 θ̈ + vF θ̇ = 0, ak 2 = −vF , (119)
θ̇
We take the time derivative of this,
d θ̈ 2
ak 2 = −v̇F = −aθ̇ , (120)
dt θ̇
recalling eq. (113). Multiply this by θ̈/θ̇,
 2 2
2 θ̈d θ̈ k2 d θ̈ 1 dθ̇
k = = −θ̈θ̇ = − . (121)
θ̇ dt θ̇ 2 dt θ̇ 2 dt
Integrating this,
 2 
2 θ̈ 2 2 kdθ̇
k = c − θ̇ , k θ̈ = θ̇ c − θ̇ ,  = dt, (122)
θ̇ 2
θ̇ c − θ̇
for some constant c. We integrate this using Dwight 341.01,6 setting t0 = 0,
k c c θ
cosh−1 = t + t0 = t,
− θ̇ = = c cos , (123)
c θ̇ cosh(ct/k) k
ct ct θ ct 1
θ = k tan−1 sinh , sinh = tan , cosh = , (124)
k k k k cos(θ/k)
6
http://kirkmcd.princeton.edu/examples/EM/dwight_57.pdf
Princeton University 1988 Ph205 Set 3, Solution 9 26

defining θ(t = 0) = 0. Using eq. (123) in eq. (113) we find (also using eq. (124)),
2 ac2 ct θ
v̇F = aθ̇ = , vF = ack tanh = ack sin , (125)
cosh2 (ct/k) k k
defining vF (t = 0) = 0.
The kinetic energy is constant, using eqs. (114), (123) and (125),
2 2
mẋ2 + mẏ 2 + I θ̇ mvF2 + ma2θ̇ + I θ̇ m 2 2 2 2 ma2c2 k 2
T = = = (vF + a k θ̇ ) = . (126)
2 2 2 2
The constraint force is, using eqs. (108) and (123),
cI θ c2 I 2θ
F = −λ = −I θ̈ = sin θ̇ = sin . (127)
k k 2k k
To characterize the motion for very large times, we note from eq. (125) that
vF (t → ±∞) → ±ack, and hence that θ(t → ±∞) → ±kπ/2. That is, the trajectories
are asymptotically straight lines at a constant angle.
For very small times where θ ≈ 0, we consider various derivatives, more simply for
(xF , yf ) than for (x, y) of the center of mass:
dxF ẋF vF cos θ ak sin(θ/k) ak θ t→0
= = = cos θ = tan cos θ → 0,(128)
dθ θ̇ θ̇ c cos(θ/k) c k
 
2
d xF ak cos θ θ t→0 a
2 = 2
− tan sin θ → = 0,(129)
dθ c k cos (θ/k) k c
dyF ẏF vF cos θ ak sin(θ/k) ak θ t→0
= = = sin θ = tan sin θ → 0,(130)
dθ θ̇ θ̇ c cos(θ/k) c k
 
d2 yF ak sin θ θ t→0
2 = 2
+ tan cos θ → 0,(131)
dθ c k cos (θ/k) k
 
3
d yF ak cos θ 2 sin θ sin(θ/k) cos θ θ t→0 2a
3 = 2
− 2 3
+ 2
− tan sin θ → .(132)
dθ c k cos (θ/k) k cos (θ/k) k cos (θ/k) k c
All these imply the curve has a cusp at the origin, with its tip pointing to the −x axis.
Princeton University 1988 Ph205 Set 3, Solution 10 27

10. (a) The particle moves in a vertical plane through the center of the sphere, subject
to the constraint that,

g(r, θ) = r − a = 0. (133)

The Lagrangian is, for coordinates r and θ,

 
m 2 2
L=T −V = ṙ + r2 θ̇ − mgr cos θ, (134)
2
taking V = 0 at the center of the sphere. Lagrange’s equations for the constrained
motion are,
d ∂L ∂L 2 ∂g
− = mr̈ − mrθ̇ + mg cos θ = λ = λ, (135)
dt ∂ ṙ ∂r ∂r
d ∂L ∂L ∂g
− = mr2 θ̈ + 2rṙ θ̇ − mgr sin θ = λ = 0, (136)
dt ∂ θ̇ ∂θ ∂θ
Applying the constraint (133) to eq. (136), we have,
2
g g θ̇ g
θ̈ = sin θ, θ̇θ̈ = θ̇ sin θ, = (1 − cos θ), (137)
a a 2 a

for the initial conditions that θ(t = 0) = 0 = θ̇(t = 0).

Then, from eq. (135),
2g
λ = −ma (1 − cos θ) + mg cos θ = mg(3 cos θ − 2). (138)
a
Since λ has the physical significance of the constraint force, this vanishes when
2
cos θ = , (139)
3
at which condition the mass flies off the sphere.

(b) A uniform sphere of radius b, mass m and moment of inertia (about its center of
mass) I = 2mb2 /5 starts at rest from the top of a fixed sphere of radius a, and
rolls without slipping down the latter.
Princeton University 1988 Ph205 Set 3, Solution 10 28

The constraint of rolling without slipping can be written as,

a a
bφ = aθ, φ= θ, gφ = φ − θ (140)
b b
where φ is the angle between the original point of contact between the spheres
when θ = 0 and the line of center of the two spheres.
Associated with this constraint is a tangential force F between the spheres at
their point of contact, as well as the normal force N there. The upper sphere flies
off the lower when the normal force vanishes.
Following pp. 114-115 of Routh, Elementary Rigid Dynamics,
http://kirkmcd.princeton.edu/examples/mechanics/routh_elementary_rigid_dynamics.pdf,
we consider the radial and tangential accelerations of the center of mass of the
upper sphere,
2
m ar = m θ̇ (a + b) = mg cos θ − N, (141)
m at = m θ̈(a + b) = mg sin θ − F. (142)

Also, energy E = T + V is conserved in the rolling motion,

2  
m v2 I m θ̇ 2b2 (a + b)2 7 2
T = + (θ̇ + φ̇)2 = (a + b)2 + 2
= m(a + b)2θ̇ , (143)
2 2 2 5 b 10
V = mgy = mg(a + b) cos θ, (144)
7 2 2 10g 1 − cos θ
E = m(a + b)2 θ̇ + mg(a + b) cos θ = mg(a + b), θ̇ = , (145)
10 7 a+b
dE 7 5g sin θ
= 0 = m(a + b)2θ̇θ̈ − mg(a + b) sin θ θ̇, θ̈ = . (146)
dt 5 7(a + b)
From eqs. (141) and (145), the normal force is,
 
10 mg
N = mg cos θ − (1 − cos θ) = (17 cos θ − 10), (147)
7 7
which vanishes when,
10
cos θ = , (148)
17
and the upper sphere flies off. For what it’s worth, the tangential (friction) force
is, using eqs. (142) and (146),
 
5 2mg sin θ
F = mg sin θ 1 − = . (149)
7 7

In a Lagrangian analysis, we deduce the normal force N as a constraint force,

related to a constraint on the radius of the lower sphere,

gr = r − a = 0, (150)
Princeton University 1988 Ph205 Set 3, Solution 10 29

while keeping the radius b of the upper sphere fixed.7 Then, we can use the
constraint (140) as is.8 That is, we use φ = aθ/b in Lagrange’s equations for coor-
dinates r and θ with one Lagrange multiplier, λr , associated with the constraint
(150). Again, θ̇ + φ̇ = (1 + a/b)θ̇, and we write eqs. (143)-(144) as,
 
mv 2 I m 2 2 2 2
T = + (θ̇ + φ̇)2 = ṙ + (r + b)2θ̇ + (a + b)2θ̇ , (151)
2 2 2 5
V = mgy = mg(a + r) cos θ, (152)

Lagrange’s equations for this system are,

d ∂L ∂L 2 ∂gr
− = mr̈ − m(r + b)θ̇ − mg cos θ = λr = λr , (153)
dt ∂ ṙ ∂r ∂r
d ∂L ∂L 2m
− = m(r + b)2θ̈ + m(r + b)ṙθ̇ + (a + b)2 + mg(a + r) sin θ
dt ∂ θ̇ ∂θ r
∂gr
= λr = 0. (154)
∂θ
Having deduced the equations of motion with the constraint (150) relaxed, we
now enforce it, which implies ṙ = 0 = r̈,
2
m(a + b)θ̇ − mg cos θ = λr , (155)
7m
(a + b)2θ̈ + mg(a + b) sin θ = 0. (156)
5
To complete the analysis we note that energy is conserved, so we can use eq. (145)
in eq. (155), where λr is the constraint force associated with coordinate r, i.e.,
the normal force −N on the lower sphere,
10g 1 − cos θ mg
λr = −N = m(a + b) − mg cos θ = (10 − 17 cos θ). (157)
7 a+b 7
The upper sphere flies off when this normal force vanishes,
10
cos θ = , (158)
17
as found more easily by elementary methods.

7
It does not work to use the constraint gr = r − b = 0 on the radius of the upper sphere, which would
make the moment of inertia of that sphere variable when this constraint is relaxed.
8
This is discussed on pp. 374-378 of Symon, Mechanics, 3rd ed.
http://kirkmcd.princeton.edu/examples/mechanics/symon_71.pdf