M8 Sol

Kevin Zhou Physics Olympiad Handouts
Mechanics VIII: Synthesis

Three-dimensional rotation is covered in chapter 7 of Kleppner and chapter 9 of Morin. For further
discussion and examples, see chapter I-20 of the Feynman lectures, and this awesome video. There
is a total of 101 points.
1 3D Rotation
In M6, we considered mostly two-dimensional rotation. Now we will tackle the full three-dimensional
case, starting with the general description of rigid body motion.
Idea 1: Chasles’ Theorem

The instantaneous velocity of a three-dimensional rigid body can always be decomposed in
one of two ways. First, for any given point, it can be written in terms of a translational
velocity plus a pure rotation about an axis going through that point. In practice, this point
is almost always chosen to be the center of mass, giving the decomposition
v = vCM + ω × (r − rCM ).
Alternatively, there always exists an axis so that the motion can be written as rotation about
that axis, plus a translational velocity parallel to the axis, giving
v = v0 + ω × (r − r0 )
where v0 and ω are parallel. (This is known as a “screw” motion.) These two decompositions
are analogous to the two we saw in M5, though in the three-dimensional case the first tends
to be much more useful.
Example 1
You have a small globe, which is mounted so that it can spin on the polar axis and can be
spun about a horizontal axis (so that the south pole can be on top). Give the globe a quick
spin about the polar axis, and then, before it stops, give it another quick spin about the
horizontal axis. Are there any points on the globe that are at rest?
Solution
The first spin gives the angular velocity a vertical component ω1 . The second spin gives the
angular velocity an additional horizontal component ω2 . The globe now rotates about its
center of mass with angular velocity ω1 + ω2 . Precisely two points on the globe are stationary,
namely the points that are parallel and antiparallel to this vector.
[2] Problem 1 (Morin 9.3). A cone rolls without slipping on a table; this means that all the points
of the cone that touch the table are instantaneously stationary. The half-angle of the vertex is α,
and the axis has length h.
1
Let the speed of the center of the base, point P in the figure, be v.
(a) Compute the angular velocity ω by thinking of the motion as pure rotation about some axis.
(b) Compute the angular velocity ω by thinking of the motion as translation of P , plus rotation
about an axis passing through P .
(c) The apex of the cone is fixed, and the cone continues to rotate. As this motion goes on, the
angular velocity vector rotates uniformly, keeping a constant magnitude. Find the angular
velocity Ω of the angular velocity vector.
In the precession problems below, it’s important to keep track of the difference between ω and Ω.
Solution. (a) The points of the cone that touch the table form a line. This whole line is
instantaneously stationary, so the motion is pure rotation about this axis. Let d be the
distance from P to the ground. Then the speed of point P is v = ωd, which gives
v v
ω= = .
d h sin α
(b) The point P has speed v. The point directly below P is at rest, and its velocity can also be
written as v − ωd. Then v = ωd, giving the same answer as below.
(c) The center of the base moves in a circle of radius h cos α with speed v, and hence completes
one cycle in time 2πh cos α/v. Hence the angular velocity of the angular velocity is
v
Ω=− ẑ
h cos α
where we used the right hand rule to find the sign. Note how this is distinct from the angular
velocity: not only do they have totally different magnitudes, they point in totally different
directions! The angular velocity vector only describes what a body is doing right now. It
doesn’t include Ω, which is about what the body will do in the future.
Most of our statements about rotational dynamics from M6 remain true. The main new aspect is
that angular momentum is not necessarily parallel to angular velocity.
Example 2: KK Example 7.4
Consider a rigid body consisting of two particles of mass m connected by a massless rod of
length 2ℓ, rotating about the z-axis with angular velocity ω as shown.
2
Find the angular momentum of the system.
Solution
We simply add r × p for both masses. Let the rod lie in the xz plane at this moment. Then
for the top left mass,
r = −ℓ cos α x̂ + ℓ sin α ẑ.
The momentum is
p = mv = mω × r = −mωℓ cos α ŷ.
Then the angular momentum is
L = r × p = mωℓ2 cos α (sin α x̂ + cos α ẑ) .
The other mass has the opposite r and p and hence the same L, so the total angular
momentum is
L = 2mωℓ2 cos α (sin α x̂ + cos α ẑ) .
It is directed perpendicular to the rod, and in particular, it isn’t parallel to the angular velocity!
Here is another way to derive the same result. We can decompose the angular velocity vector
into a component along the rod, and a component perpendicular to the rod. The former
contributes no angular momentum, because rotating about the rod’s axis doesn’t move the
masses. The latter contributes all the angular momentum. So the angular momentum is
L = I⊥ ω⊥ = (2mℓ2 )(ω cos α)
directed perpendicular to the rod, which is what we just saw explicitly.
We can summarize the lessons drawn from this example as follows.
Idea 2
For a three-dimensional object, L is not necessarily parallel to ω. In general, for pure rotation
about an axis passing through the origin, we have L = I ω where I is a 3 × 3 matrix called
3
the “moment of inertia tensor about the origin”. In components, this means that
X
Li = Iij ωj .
j
While this is simple and general, the Iij are a pain to calculate. You can learn more in the
reading, but to my knowledge, no Olympiad problem has ever required computing a general
moment of inertia tensor.
For the purposes of Olympiad problems, there is a better way to think about the angular
momentum. We use the second decomposition of idea 1, and think of the motion as translation
plus rotation about the center of mass. If the object has an axis of symmetry, which it will
in almost all Olympiad problems, then the angular velocity can then be decomposed into a
component parallel to the axis, and perpendicular to the axis,
ω = ω∥ + ω⊥ .
The key is that, in such situations, the spin angular momentum has two pieces, which are
each parallel to the corresponding piece of the angular velocity,
L∥ = I∥ ω∥ , L⊥ = I⊥ ω⊥
where I∥ and I⊥ are ordinary moments of inertia about the center of mass. For example, for
a flat uniform disc, I∥ = M R2 /2 and I⊥ = M R2 /4.
The total angular momentum about the origin is then
L = rCM × M vCM + I∥ ω∥ + I⊥ ω⊥
where the first term is from the motion of the center of mass, and the next two are from
rotation about the center of mass. Note that this is exactly the same as what we saw in M5,
except that the “spin” angular momentum is broken into two parts.
Idea 3
Sometimes it can be hard to visualize ω, so here are two tricks. First, if any two points
on the object are stationary, then ω must be parallel to the axis connecting the two points.
Second, if the rotation is complicated, one can use rotating frames to simplify the analysis.
If a body has angular velocity ω1 in one frame, and that frame has an angular velocity ω2
with respect to a second frame, then the body has angular velocity ω1 + ω2 with respect to
the second frame.
Idea 4
The rate of change of angular momentum is
dL
=τ
dt
4
where the torque τ is defined as in M5. The kinetic energy is

1 2 1 1
K = M vCM + I∥ ω∥2 + I⊥ ω⊥
2
.
2 2 2
[3] Problem 2 (KK 7.1). A thin hoop of mass M and radius R rolls without slipping about the z-axis.
It is supported by an axle of length R through its center, as shown.
The axle circles around the z-axis with angular speed Ω, so that the bottom point of the wheel
traces out a circle of radius R. Let O be the pivot point of the rod, i.e. the point where the rod
meets the z-axis.
(a) Find the instantaneous angular velocity ω of the hoop.
(b) As the motion continues, the angular velocity vector rotates in a circle. Find the angular
velocity of the angular velocity vector of the hoop.
(c) Find the instantaneous angular momentum L of the hoop, about the point O.
(d) Find the instantaneous torque on the hoop about the contact point with the ground, and
verify that τ = dL/dt.
The solution to this question will be the basis for all the rest, so ask if you have any questions!
Solution. (a) Since both O and the bottom point of the hoop are stationary, the angular velocity
must be parallel to the line joining them,
ω ∝ ẑ − ŷ.
In fact, the motion can be thought of as pure rotation about the line joining them. Now
consider the motion of the rod. The y-component of the angular velocity doesn’t affect the
rod, while the z-component makes it rotate about the z-axis. We are already given that the
rod rotates with angular speed Ω about the z-axis, so we must have ωz = Ω, and hence
ω = Ωẑ − Ωŷ.
Alternatively, this part can be done using an intermediate rotating frame. We first go to the
frame rotating with angular velocity Ωẑ. In this frame the rod is frozen in place, while the
wheel turns in place, with angular velocity −Ωŷ. So the angular velocity in the original frame
is the sum, giving the same answer.
5
(b) Since the wheel is rigidly attached to the rod, this is simply Ωẑ.
(c) We apply the result
L = rCM × M vCM + I∥ ω∥ + I⊥ ω⊥ .
We have I∥ = M R2 and I⊥ = M R2 /2, so
M R2

2 2 3
L = M R Ωẑ − M R Ωŷ + Ωẑ = M R2 Ω ẑ − ŷ .
2 2
(d) The reason we take the force about the contact point is so that we won’t have to deal with the
friction and normal forces. The axle pulls with force M Ω2 R for the centripetal force, giving a
torque of M Ω2 R2 x̂ about the contact point.
To find dL/dt, note that the vertical component of L won’t change; only the ŷ component
will change at rate Ω. Thus at that instance, the magnitude of dL/dt is M R2 Ω2 . By thinking
about how L rotates about the the z-axis, we see that the direction of dL/dt is x̂. Thus we
see that dL/dt = M Ω2 R2 x̂ = τ .
Finally, we need just a few more facts to get going with dynamics problems.
Idea 5
In any dynamics problem, there are many choices you can make in the setup. For example,
if you’re using an inertial frame, you need to choose where the origin is; usually it’s best
to place it along the axis of symmetry if possible. You are also free to use a noninertial
frame with acceleration a. The only difference is that there will be a fictitious force
−M a acting at the center of mass. For that reason, it’s usually best to have the acceler-
ating frame follow the center of mass, keeping it at its origin, so no new torques are introduced.
However, you should avoid rotating reference frames for dynamics problems. Not only will
there be position-dependent Coriolis forces, but they’ll add up and contribute a Coriolis
torque, which is a pain to calculate, as you saw in M6. In general, rotating frames are only
good for getting a handle on the kinematics, as mentioned in idea 3.
Example 3: KK Example 7.5
Calculate the magnitude of the torque on the rod in example 2.
Solution
We recall that the angular momentum was
L = 2mωℓ2 cos α (sin α x̂ + cos α ẑ).
The rod as a whole rotates with angular velocity ωẑ. In particular, the angular momentum
vector rotates with this angular velocity as well; its horizontal component moves in a circle
with angular velocity ω. Then
dL
|τ | = = ωLx = 2mω 2 ℓ2 cos α sin α = mω 2 ℓ2 sin(2α).
dt
6
It might be surprising that there needs to be a torque given that ω is constant, but that’s
just because L and ω aren’t necessarily parallel. Conversely, there can be situations where
there is no torque, yet ω changes over time.
[2] Problem 3 (Morin 9.10). A stick of mass m and length ℓ spins with frequency ω around an axis
in zero gravity, as shown.
The stick makes an angle θ with the axis and is kept in its motion by two strings that are perpen-
dicular to the axis. Find the tension in the strings.
Solution. The component of angular velocity parallel to the stick contributes no angular momentum,
1
and the component perpendicular to the stick is ω sin θ, so it contributes L = 12 mℓ2 ω sin θ (this
can be formalized with the inertia tensor). The direction is perpendicular to the stick. We see then
that it traces out a cone, so to find L̇, we note that only the horizontal component of L is changing,
so its rate of change is ωL cos θ. The torque is 2T (ℓ/2) cos θ, so
1
T ℓ cos θ = ω mℓ2 ω sin θ cos θ,
12
1
or T = mℓω 2 sin θ .
12
[2] Problem 4 (KK 7.4). In an old-fashioned rolling mill, grain is ground by a disk-shaped millstone
which rolls in a circle on a flat surface driven by a vertical shaft. Because of the stone’s angular
momentum, the contact force with the surface can be greater than the weight of the wheel.
Assume the millstone is a uniform disk of mass M , radius b, and width w, and it rolls without
slipping in a circle of radius R with angular velocity Ω. Find the contact force. Assume the millstone
is closely fitted to the axle so that it cannot tip, and w ≪ R. Neglect friction.
7
Solution. Take torques about the point where the vertical and horizontal rods meet. The only
torque is from gravity and the normal force,
τ = (N − M g)R.
The angular momentum due to the motion of the center of mass is vertical and constant, so it
doesn’t matter. As in problem 2, the angular velocity connects this point and the contact point, so
R
ω= Ωx̂ + Ωẑ.
b
The vertical component of the angular velocity yields another constant vertical component of the
angular momentum, so it also doesn’t matter. The only part of the angular momentum that changes
is the part due to the horizontal component of the angular velocity of the axle. The whole system
precesses with angular velocity Ωẑ, so

RΩ 1 2
τ= Ω Mb .
b 2
Setting this equal to our other expression for torque and solving gives
bΩ2

N = Mg 1 + .
2g
[2] Problem 5 (Morin 9.29). A uniform ball rolls without slipping on a table. It rolls onto a piece
of paper, which you then slide around in an arbitrary (horizontal) manner. You may even give
the paper abrupt, jerky motions, so that the ball slips with respect to it. After you allow the ball
to come off the paper, it will eventually resume rolling without slipping on the table. Show that
the final velocity equals the initial velocity. (Hint: this remarkably simple result is because of a
conservation law. We saw a lower-dimensional version of this problem in M5.)
Solution. Let our system be the ball plus the disk, and let the origin be the center of the disk.
We first claim that the component of angular momentum in the plane of the disk is conserved.
To show this, it suffices to show that all torques are perpendicular to the plane of the disk. First,
the torque at the center of the disk that keeps it spinning is clearly in the perpendicular direction,
since the angular momentum of the disk is always in the perpendicular direction, and τ = dL dt . Now,
let us find the torque on the ball.
Let R be the vector from the origin to the center of the ball, and r the vector from the contact
point of the ball to the center of the ball. Now, the normal force and gravity are equal and opposite,
but gravity acts at the center, and the normal force acts at the contact point. Thus, the total torque
due to these two is
R × N + (R + r) × mg = 0
since g ∥ r, and as noted before, mg = −N. Finally, there is the friction force, which acts at the
contact point. The torque due to it is R × f , which points in the perpendicular direction, so the
only torques are in the perpendicular direction.
Now, we just compute the parallel component of the angular momentum before and after leaving.
This is simply
((R + r) × v + I ω)∥ = (I + mr2 )ω,
so ω doesn’t change, so the ball continues on its original course.
8
When a system such as a gyroscope is given a high angular momentum, it can exhibit precession. In
this case, the angular velocity ω precesses (i.e. rotates) with a small angular velocity Ω. Precession
is a famously counterintuitive phenomenon, but it can be handled using just the same principles
we’ve laid out above.
Example 4: KK 7.3
A gyroscope wheel is at one end of an axle of length ℓ. The other end of the axle is suspended
from a string of length L.
The wheel is set into motion so that it executes slow, uniform precession in the horizontal
plane. The wheel has mass M and moment of inertia I0 about its center of mass, and turns
with angular speed ωs . Neglect the mass of the shaft and string. Find the angle β the string
makes with the vertical, assuming β is very small.
Solution
Let T be the tension in the rope. The entire system precesses with associated angular velocity
Ωẑ. (That is, this is the angular velocity of the angular velocity and angular momentum
vectors.) Since the center of mass does not accelerate vertically, and the center of mass moves
in a horizontal circle,
T cos β = M g, T sin β = M Ω2 (ℓ + L sin β).
We’ll work to lowest possible order in β everywhere, which means approximating cos β ≈ 1
and ignoring the L sin β term, giving
T = M g, T β = M Ω2 ℓ.
Combining these equations gives the precession frequency

r
gβ
Ω= .
ℓ
This is as far as we can go with forces alone.
Now we use τ = dL/dt. First we need to determine the angular velocity of the wheel. Note
that if we went in the rotating frame with angular velocity Ωẑ, the wheel would just spin
with angular speed ωs in place. So the angular velocity vector in the original frame is
ω = Ωẑ + ωs x̂
9
where x̂ is directed along the rod. The center of mass of the wheel moves with speed
(ℓ + L sin β)Ω in a horizontal circle.
Since the precession is assumed to be slow, Ω is much smaller than ωs . Thus, we can simply
ignore contributions to the angular momentum proportional to Ω. That is, we can take
ω ≈ ωs x̂
for the purposes of computing angular momentum, giving
L ≈ I0 ωs x̂.
This rotates in a horizontal circle with angular speed Ω, so
|τ | ≈ I0 Ωωs .
The torques in the relevant direction come from gravity and the vertical component of the
tension force, |τ | = M gℓ. Equating these, we have
M gℓ = I0 Ωωs .
Plugging the result for Ω in above and solving for β gives
m2 gℓ3
β= .
ω 2 I02
Note that assuming Ω was small kept the equations simple. This is what Kleppner calls the
gyroscope approximation. It can be applied in some, but not all, of the below problems.
Remark
In most gyroscope problems, such as the one above, we assume the motion has reached a
steady state, but you might wonder just how it gets started in the first place. For example,
suppose we had the same setup as the previous problem, with the wheel spinning and the axle
horizontal. For simplicity, let’s get rid of the string and suppose the end of the axle is held
at a fixed support. Now suppose the axle and wheel are released with no translational motion.
The following chain of events ensues:
• Of course, the axle starts to tip downward because of the weight of the wheel. (Rotational
mechanics is counterintuitive, but not that counterintuitive!)
• This produces a downward component of angular momentum, which is balanced by the

axle/wheel system twisting about its center of mass, rotating slightly about the z-axis.
• This twist tries to push the end of the axle out of the page, causing the support to exert
a force on the axle pointing into the page. That force propagates down the axle as an
internal shear stress, eventually causing the center of mass of the wheel to start moving
into the page, starting the precession.
10
• In reality, the process overshoots and overcorrects, leading to oscillations called nutation
on top of the precession. (For more details, see Note 2 of chapter 7 of Kleppner.)
• For a typical pivot, energy can be dissipated at the pivot point, but the angular momentum
of the system stays roughly the same. Assuming this is the case, the oscillations will
eventually damp away, leaving the uniform precession.
Notice that in this example, the initial angular momentum is perfectly horizontal. The final
angular momentum includes an upward component due to the uniform precession, which
implies that the axle must tilt slightly downward, by an angle of order (ω/Ω)2 . Therefore,
if you want to set up uniform precession with the axle perfectly horizontal, as in the above
example, you should point the axle slightly upward when releasing it from rest.
[2] Problem 6 (KK 8.5). An “integrating gyro” can be used to measure the speed of a vehicle.
Consider a gyroscope spinning at high speed ωs . The gyroscope is attached to a vehicle by a
universal pivot. If the vehicle accelerates in the direction perpendicular to the spin axis at rate a,
then the gyroscope will precess about the acceleration axis, as shown.
The total angle of precession is θ. Show that if the vehicle starts from rest, its final speed is
Is ωs
v= θ
Mℓ
where Is ωs is the gyroscope’s spin angular momentum, M is the total mass, and ℓ is the distance
from the pivot to the center of mass.
Solution. Work in the accelerating reference frame where the pivot is at rest. Using the gyroscope
approximation, we see that L̇ = Is ωs ω where ω = θ̇, and τ = M aℓ. Thus,
Is ωs θ̇
a= .
Mℓ
Integrating yields the desired result.
[3] Problem 7 (KK 7.5). When an automobile rounds a curve at high speed, the weight distribution
on the wheels is changed. For sufficiently high speeds, the loading on the inside wheels goes to
zero, at which point the car starts to roll over. This tendency can be avoided by mounting a large
spinning flywheel on the car.
(a) In what direction should the flywheel be mounted, and what should be the sense of rotation, to
help equalize the loading? (Check your method works for the car turning in either direction.)
11
(b) Show that for a disk-shaped flywheel of mass m and radius R, the requirement for equal
loading is that the angular velocity ω of the flywheel is related to the velocity of the car v by
ML
ω = 2v
mR2
where M is the total mass of the car and flywheel, and L is the height of their center of mass.
Solution. (a) If the car is turning with a radius of curvature of r at velocity v, then the frictional
force must provide the centripetal force of f = M v 2 /r. This will exert a torque of f L on the
car about the center of mass, where L is the height of the center of mass. The torque points
forward for turning left, and backwards for turning right.
Normally, a difference in the normal forces between the wheels will provide the opposing
torque to keep the car from rolling over. To keep an equal loading, the flywheel must provide
the opposing torque. Another way to think about it is to have the torque from friction to
cause precession of the flywheel instead of turning the car (the equal and opposite ”reaction
torque” will keep the car stable).
The key is that as the car is turning, the flywheel will also turn with the car at angular velocity
v/r, thus the direction of its spin angular momentum Ls will change. From a top view, that
means the forwards torque for turning left must turn Ls counterclockwise, and the backwards
torque for turning right must turn Ls clockwise. This works when Ls is pointing to the right
with respect to the car’s motion (the flywheel spins in the opposite direction than that of the
wheels).
(b) In order for the torque from friction to turn the flywheel, τ = f L = Ls (v/r). For a disk-shaped
flywheel with angular momentum Ls = 12 mR2 ω, putting in f = M v 2 /r yields
M v2L 1
= mR2 ωv/r
r 2
which gives
ML
ω = 2v
mR2
as desired.
[3] Problem 8 (KK 7.7). A thin hoop of mass M and radius R is suspended from a string through a
point on the rim of the hoop. If the support is turned with high angular velocity ω, the hoop will
spin as shown, with its plane nearly horizontal and its center nearly on the axis of the support.
The string makes an angle α with the vertical.
12
(a) Find, approximately, the small angle β between the plane of the hoop and the horizontal.
(b) Find, approximately, the radius of the small circle traced out by the center of mass.
Solution. (a) Let the tension in the string be T . To balance gravity, T cos θ = M g. The angle
between the string and the plane of the hoop is π − (α + π/2 − β) = π/2 − α + β, so the
torque on the hoop about the center of mass is τ = T R sin(π/2 − α + β) = T R cos(α − β).
This torque rotates the horizontal component of the angular momentum about the center of
mass, M R2 ω sin β, with angular velocity ω. Thus
T R cos(α − β) ≈ T R(cos α − β sin α) ≈ M R2 ω 2 β
Rω 2 β = g(1 − β tan α)
g
β= 2
.
Rω − g tan α
(b) The horizontal component of the tension should provide a centripetal force,
T sin α = M ω 2 r
giving an answer of
g tan α
r= .
ω2
[4] Problem 9 (KK 7.6, Morin 9.23). With the right initial conditions, a coin on a table can roll in a
circle.
As shown, the coin leans inward, with its axis tilted to the horizontal by an angle ϕ. The radius of
the coin is b, the radius of the circle it follows on the table is R, and its velocity is v.
(a) Assuming the coin rolls without slipping and b ≪ R, show tan ϕ = 3v 2 /2gR.
(b) No longer assuming b ≪ R, show that the described motion is only possible if R > (5/6)b sin ϕ.
Solution. (a) We work in the nonrotating, but noninertial frame whose origin follows the center
of mass. In this frame, the only part of the angular momentum that changes is the horizontal
component of the spin angular momentum. The coin spins by “rolling” and “turning”, along
ωs and ω2 respectively:
13
Since the motion of the coin is the combination of “rolling” and going around in a circle, the
total angular velocity should be ωs + Ω where Ω describes the circular motion/turning of the
coin and points vertically upwards. The components of Ω are Ω cos ϕ and −Ω sin ϕ along ω2
and ωs respectively. The moments of inertia about the coin in the ωs and ω2 directions are
1 2 1 2
2 M b and 4 M b respectively. Note that we can’t simply ignore Ω because it is vertical; this
is because the angular momentum from Ω does not point in the same direction as Ω (they
point in the same direction only along the principal axes).
With L = Iω along those principal axes, the horizontal components of the angular momenta
is Lx = 12 M b2 (ωs − Ω sin ϕ) cos ϕ + 14 M b2 Ω cos ϕ sin ϕ. The no slip condition is that ωs b = ΩR.
With b ≫ R, we can approximate Lx ≈ 21 M b2 ωs cos ϕ. The torque, τ = ΩLx , about the center
of mass is (N sin ϕ − f cos ϕ)b where f = M v 2 /(R − b sin ϕ) ≈ M v 2 /R is the frictional force,
and N = M g is the normal force. The velocity of the CM is v = Ω(R − b sin ϕ) ≈ ΩR. Then
1 1 M v2b
τ = ΩLx = M b2 (ΩR/b)Ω cos ϕ = M Ω2 bR cos ϕ = cos ϕ
2 2 2R
but we also know that
M v2b
τ = M gb sin ϕ − cos ϕ
R
from which we conclude
3v 2
tan ϕ = .
2gR
(b) Now we will do the calculations above without b ≪ R.

1 2 1 2
τ =Ω M b (ωs − Ω sin ϕ) cos ϕ + M b Ω cos ϕ sin ϕ
2 4
M v2b
= M gb sin ϕ − cos ϕ.
R − b sin ϕ
Doing the replacements of ωs = ΩR/b and v = Ω(R − b sin ϕ), dividing the above equation by
M Ω2 b yields
1 1 1 g
R cos ϕ − b sin ϕ cos ϕ + b sin ϕ cos ϕ = 2 b sin ϕ − (R − b sin ϕ) cos ϕ
2 2 4 Ω
14
3 5 g
R − b sin ϕ = 2 b tan ϕ.
2 4 Ω
Since tan ϕ > 0 in order for the motion to make sense, 32 R − 54 b sin ϕ > 0, thus we get
R > 56 b sin ϕ.
[4] Problem 10 (Morin 9.24). If you spin a coin around a vertical diameter on a table, it will slowly
lose energy and begin a wobbling motion. The angle between the coin and the table will gradually
decrease, and eventually it will come to rest. Assume this process is slow, and consider the motion
when the coin makes an angle θ with the table, as shown.
You may assume that the center of mass is essentially motionless. Let R be the radius of the coin,
and let Ω be the frequency at which the contact point on the table traces out its circle. Assume
the coin rolls without slipping.
(a) Show that the angular velocity of the coin is ω = Ω sin θ x̂2 , where x̂2 always points upward
along the coin, directly away from the contact point.
p
(b) Show that Ω = 2 g/R sin θ.
(c) Show that the face on the coin appears to rotate, when viewed from above, with angular
frequency (1 − cos θ)Ω.
Solution. (a) Since the center of mass is essentially motionless and thee coin is rolling without
slipping, the center of mass and the contact point are both stationary. Thus the angular
velocity must pass through those lines, and is pointing along x̂2 . Let k̂ be a vertical unit
vector. The angular velocity can be seen as the sum of the rotation about the center of mass
and pointing along k̂ (turning of the coin’s orientation) with angular velocity ωk = Ωk̂, and
rotation about −vxˆ 3 with angular velocity ω3 to roll without slipping. Thus ω = ωk + ω3
Since x̂2 and x̂3 are perpendicular, ω = ωk sin θ, thus
ω = Ω sin θx̂2
(b) The torque about the contact point from gravity is M gR cos θ, and points horizontally to
change the horizontal component of the angular momentum Lx = Iω cos θ at a rate of Ω. The
moment of inertia about x̂2 is 14 M R2 , which gives
1
M gR cos θ = M R2 Ω2 sin θ cos θ.
4
r
g
Ω=2
R sin θ
The same result can be found be taking torques about the center of mass (no friction).
15
(c) From part (a), we found that ω = ωk + ω3 and x̂2 and x̂3 are perpendicular which gets ω3 =
−Ω cos θ. Consider a point on the coin from the top view. ωk makes it rotate counterclockwise
with angular velocity Ω, and ω3 rotates it clockwise with angular velocity Ω cos θ. Thus the
face of the coin appears to be rotating with angular velocity Ω(1 − cos θ).
Another way to do this is to consider the difference between the radius of the coin and the
radius of the traced out circle. In a full rotation of the contact point in time T = 2π/Ω, a
distance of 2πR cos θ was covered by the coin. Since the coin didn’t slip, that same distance
was covered along the coin’s edge, so the initial contact point will be a distance of 2πR(1−cos θ)
ahead of the new contact point. Thus the angle change is 2π(1 − cos θ) in time T = 2π/Ω,
giving an apparent angular velocity of (1 − cos θ)Ω.
Remark: Bivectors
Vector quantities defined by the cross product have some unusual properties. For example,
under a spatial inversion, which flips the signs of r and p, the sign of L = r × p doesn’t
get flipped, so L transforms differently from other vectors. The same applies to the veloc-
ity ω and magnetic field B. All three of these quantities are “pseudovectors”, not true vectors.
The underlying reason is that all of these quantities are fundamentally a different kind
of mathematical object. They are really rank 2 differential forms, also called bivectors
in three dimensions. While a vector is specified by an arrow with magnitude and di-
rection, a bivector is specified by a planar tile with area and orientation. The following
figure, taken from this paper, shows how it can be constructed visually from the cross product.
In three dimensions, we can always convert between bivectors and pseudovectors using
the right-hand rule, so any calculation can be done with either form. Bivectors have the
advantage of visually representing rotational quantities: the angular velocity bivector lies
along an object’s plane of rotation, while the magnetic field bivector lies along the plane in
which it makes charged particles circularly orbit. However, it is easier to add vectors, both
visually and mathematically, which also makes it easier to think about decomposing vectors
into components. This advantage is so important in practice that I don’t recommend using
bivectors at all for three-dimensional problems.
On the other hand, when you work in higher-dimensional spaces, the differential form per-
spective becomes indispensible. In general, in d dimensions the angular velocity has d2

16
components, corresponding to the rotation rate in each independent plane.
• Of course, when d = 1 there is no such thing as rotation at all, while when d = 2 the
angular velocity has one component, so we treat it as a scalar.
• When d = 3 the angular velocity has three components, so we treat it as a vector.
• When d = 4 the angular velocity has four components, so we can’t even pretend it’s a
vector; we have to use the differential form description.
By the way, if you want to look into this material more, be sure to steer clear of “geometric
algebra”, which dominates the Google search results. Geometric algebra is a strange internet
cult which recruits unsuspecting young people by telling them about bivectors, which are
indeed cool. Once they have your attention, they’ll claim that “mainstream” physics has
hit a dead end because it refuses to go beyond vector notation, and then you’ll spend years
relearning all of physics in their wacky alternative notation. The truth is that physicists don’t
teach geometric algebra because it’s not that useful when d = 3, while in higher dimensions
we use tensor calculus and differential forms, which are much more powerful than either
vectors or geometric algebra. More generally, if a physics movement has tons of internet
presence but no actual textbooks or novel results, it’s not worth paying attention to.
Example 5: IIT JEE 2016

√
Two thin circular discs, with radii a and 2a, are connected by a rod of length ℓ = 24 a
through their centers. This rigid object rolls without slipping on a flat table.
The center of mass of the object rotates about the z-axis with an angular speed of Ω. The
angular speed of the object about the axis of the rod is ω. How are Ω and ω related?
Solution
This is the most famous problem ever set on the IIT JEE (condensed for clarity), cel-
ebrated by generations of students for its difficulty. But it’s also an example of how
not to write a 3D rotation problem. Under the standard definition of angular velocity,
none of the options provided in the question were correct, while the intended answer
requires a nonstandard, arbitrary definition. You can find a detailed explanation of this
here, by one of the former top scorers on the JEE, and I’ll give a condensed explanation below.
17
First, let’s figure out what’s going on. The kinematics of this problem isn’t any different
from problem 1. Defining the x-axis to be horizontal in the figure above, the instantaneous
angular velocity is ω = ω x̂,√while the “angular velocity of the angular velocity”, describing
the precession, is Ω = (ω/ 24) ẑ. The hard part is figuring out what the question writers
meant by “the angular speed ω about the axis of the rod”.
If we’re only talking about the object’s instantaneous motion, then the only possible answer
is ω = ω · n̂, where n̂ is the unit vector pointing along the rod. In that case we have
Ω/ω = 5/24, which wasn’t an answer choice in the exam. On the other hand, if we are
comparing the object’s orientation at different times, then there isn’t a unique answer. At
a finite time later, the object will be in a different place, and computing a relative angle
requires defining a convention for comparing orientations.
Here’s what the problem authors meant. We work in the frame rotating with angular
velocity Ω. In this frame, the system is spinning in place, with angular velocity ω + Ω
parallel to n̂. The definition of ω is |ω + Ω|, which gives Ω/ω = 1/5, the intended answer.
Another way of saying this is that when we compare the orientation of the system at one
moment to its orientation at another moment, we bring them to the same position by rotating
about the z-axis, at which point they differ by a rotation about n̂. But this procedure is
totally arbitrary, and not specified by the problem. To pose the problem properly, the writers
could have either defined ω explicitly in the rotating frame mentioned above, or replaced
it with a quantity with equivalent but unambiguous physical meaning, such as the interval
between times a given point on the rim of a disc touches the ground. Fortunately, you’ll
almost never see problems this ambiguous on Olympiads.
2 Composite Rotation
These are rotational dynamics problems like the ones you saw in M5, but more complex.
[3] Problem 11 (PPP 60). A uniform thin rod is placed with one end on the edge of a table in a
nearly vertical position and then released from rest. Find the angle it makes with the vertical at
the moment it loses contact with the table. Investigate the following two extreme cases.
18
(a) The edge of the table is smooth (friction is negligible) but has a small, singe-step groove.
(b) The edge of the table is rough (friction is large) and very sharp, which means the radius of
curvature of the edge is much smaller than the flat end-face of the rod. Half of the end-face
protrudes beyond the table edge, so that when it is released the rod pivots about the edge.
Solution. By energy conservation, we have that

11 2 2 ℓ 3g
mℓ ω = mg (1 − cos θ) =⇒ ω 2 = (1 − cos θ).
23 2 ℓ
Therefore, the centripetal acceleration of the CM is ac = ω 2 ℓ/2 = 32 g(1 − cos θ). Differentiating the
equation for ω 2 , we learn that
3g
2ω ω̇ = (sin θ)θ̇,
ℓ
3g
so ω̇ = 2ℓ sin θ. Thus, the tangential acceleration of the CM is at = 34 g sin θ. We can solve (b) now.
(b) We see that in this case, the normal force points along the rod. Therefore,
N − M g cos θ = −M ac ,
5 3
. This becomes 0 at θ = cos−1 (3/5) .

so N = M g 2 cos θ − 2
(a) This is identical to the falling ladder problem from M5, and hence has the same answer. But
we can also solve the problem directly here. We have
3
Nx = M (at cos θ − ac sin θ) = M g sin θ(3 cos θ − 2)
3
and
1
Ny = M g − M (ac cos θ + at sin θ) = M g(3 cos θ − 1)2 .
4
The first one to go to 0 is Nx , and this happens at θ = cos−1 (2/3) .
19
[3] Problem 12 (Cahn). A tall, thin brick chimney of height L is slightly perturbed from its vertical
equilibrium position so that it topples over, rotating rigidly about its base B until it breaks at a
point P .
(a) For concreteness, we will model the internal forces in the chimney as shown below. Assume
throughout that r is very small.
We assume that each piece of the chimney experiences a shear force F and longitudinal
tension/compression forces T1 and T2 from its neighbors. Find the point on the chimney with
the greatest |T1 | or |T2 |, assuming the chimney is very thin.
(b) Find the point on the chimney experiencing the greatest shear force F .
(c) At what point is the chimney most likely to break? Do you think the limiting factor is the
chimney’s maximal compressive strength, tensile strength, or shear strength?
01^
Solution. See the solution here.
[3] Problem 13. IPhO 2014, problem 1A.
[3] Problem 14 (PPP 14). A bicycle is supported so that it can move forward or backwards but
cannot fall sideways; its pedals are in their highest and lowest positions.
20
A student crouches beside the bicycle and pulls a string attached to the lower pedal, providing a
backward horizontal force.
(a) Which way does the bicycle move?
(b) Does the chain-wheel rotate in the same or opposite sense as the rear wheel?
(c) Which way does the lower pedal move relative to the ground?
In particular, be sure to account for the gearing of the bike! To check your answer, watch this video.
Solution. (a) This actually depends on the gearing of the bicycle. If the string pulls with force
F0 , and the pedals have distance rp from the pedal axle, it will exert a torque F rp about the
pedal axle. That torque is then used for the gear with radius rg about the pedal axle, and
the chain wheel will feel a force of F0 rp /rg . This force is then transferred to the back wheel.
If the wheel gear has a radius of Rg , then the torque on the wheel from the string will be
F0 rp Rg /rg .
Now the force of friction exerts a torque of f R, where R is the radius of the back wheel. The
conditions for moving forward are f > F0 (net force points forward), and F0 rp Rg /rg > f R
(the net torque about the center of the back wheel will make it turn forward). Thus the
forward criterion is
Rrg
f > F0 > f
rp Rg
R rp
< .
Rg rg
The opposite is true for backwards movement:
R rp
> .
Rg rg
Typically, the gearings on bikes will be set so that the bike will move backwards.
(b) The chain-wheel should rotate in the same sense as the rear wheel. The chain rotates the axle,
and the wheel rotates with the axle.
(c) The lower pedal will move backwards. This is because positive work must be done on the bike
01r
to let it move, so the force and displacement should go in the same direction.
[4] Problem 15. APhO 2005, problem 1B. A problem on parametric resonance, an idea we first
encountered in M4. The problem was so subtle that the APhO problem writers themselves could
not agree on what the correct answer was!
21
01m
Solution. See solution 1B of the official solutions.
[4] Problem 16. INPhO 2020, problem 5. A tough angular collision problem.
01h
Solution. See the official solutions here.
[5] Problem 17. EuPhO 2019, problem 2. A tough problem about the motion of an rigid body
in a magnetic field.
3 Frictional Losses
These miscellaneous problems are grouped under the theme of friction or energy dissipation.
[2] Problem 18 (Kalda). A plank of length L and mass M lies on a frictionless horizontal surface; on
one end sits a small block of mass m.
The coefficient of friction between the block and plank is µ. The plank is sharply hit and given
horizontal velocity v. What is the minimum v required for the block to slide across the plank and
fall off the other end?
Solution. If the block barely is able to slide off, then right before it does, it has relative velocity
mv
of 0 with the plank. By momentum conservation the velocities are m+M , so the energy loss is
2
1 2 1 mv 1 mM 2
∆E = mv − (m + M ) = v .
2 2 m+M 2m+M
p
But this is also µmgL, so v = 2µgL(1 + m/M ).
[3] Problem 19 (BAUPC). A uniform sheet of metal of length ℓ lies on a roof inclined at angle θ,
with coefficient of kinetic friction µ > tan θ. During the daytime, thermal expansion causes the
sheet to uniformly expand by an amount ∆ℓ ≪ ℓ. At night, the sheet contracts back to its original
length. What is the displacement of the sheet after one day and night?
Solution. When the sheet expands/contracts, it should do so about a certain point that doesn’t
move by continuity (the opposite ends move in opposite directions). Additionally, the forces from
the expansion/contraction should balance so the point remains stationary.
If the center of expansion is a distance x up from the bottom of the sheet, then the compressional
force balance for a sheet with linear mass density ρ will be
µxρg cos θ − ρgx sin θ = µ(ℓ − x)ρg cos θ + ρg(ℓ − x) sin θ
ℓ ℓ
µ − tan θ = µ − µ + tan θ − tan θ
x x
µ + tan θ
x= ℓ.
2µ
22
For contraction, the tension at the stationary point a distance y up from the bottom of the sheet
has a force balance equation of
µyρg cos θ + ρgy sin θ = µ(ℓ − y)ρg cos θ − ρg(ℓ − y) sin θ

µ − tan θ
y =ℓ−x= ℓ
2µ
When the sheet expands by an amount ∆ℓ, the distance each point moves is proportional to the
distance away from the stationary point since the expansion is uniform. The stationary point for
contraction is a distance of x − y = ℓ tan θ/µ away from the stationary point for contraction, and
will move a distance of ∆ℓ(x − y)/ℓ down (away from the expansionary point), and stay stationary
for contraction. The net displacement for all the points is this distance, which is
tan θ
∆ℓ.
01r
µ
[3] Problem 20. APhO 2010, problem 1A. A question about a different kind of inelastic collision.
[5] Problem 21. 01h IdPhO 2020, problem 2. A nice problem on anisotropic friction.
4 Ropes, Wires, and Chains

Example 6: MPPP 78
A uniform flexible rope passes over two small frictionless pulleys mounted at the same height.
The length of rope between the pulleys is ℓ, and its sag is h. In equilibrium, what is the
length s of the rope segments that hang down on either side?
Solution
The problem can be attacked by differential equations, but there is an elegant solution using
only algebra. We let our unknowns be s, the tension T1 = (T1,x , T1,y ) in the rope at the
pulley, and the tension T2 at the lowest point.
23
Considering the entire sagging portion as the system, vertical force balance gives
2T1,y = λℓg, T1,y = λℓg/2.
Now consider half of the sagging portion as the system. Horizontal force balance gives
T2 = T1,x .
Finally, consider one of the hanging portions as the system. Then
T1 = λgs.
We hence have three equations, but four unknowns.
For the final equation, we need to consider how the tension changes throughout the rope.
This would usually be done by a differential equation, but there is a clever approach using
conservation of energy. Suppose we cut the rope somewhere, pull out a segment dx, and
reattach the two ends. This requires work T dx, where T is the magnitude of the local
tension. Now suppose we cut the rope somewhere else, separate the ends by dx, and paste
our segment inside. This requires work −T ′ dx. After this process, the rope is exactly in the
same state it was before, so the total work done must be zero.
This would seem to prove that T = T ′ , which is clearly wrong. The extra contribution is that
if the two locations have a difference in height ∆y, then it takes work λg(∆y) dx to move the
segment from the first to the second. So in equilibrium, for any two points of the rope,
∆T = λg ∆y.
Therefore, we have
T1 − T2 = λgh.
Now we’re ready to solve. We have
T12 − T22 = (λℓg/2)2
from our first three equations, and dividing by this new relation gives
ℓ2
T1 + T2 = λg .
4h
This allows us to solve for T1 , which gives
T1 h ℓ2
s= = + .
λg 2 8h
This is a useful result in real engineering projects: it means that the tension in a cable can
be estimated by seeing how much it sags.
24
Example 7: Kalda 27
A wedge with mass M and acute angles α1 and α2 lies on a horizontal surface. A string has
been drawn across a pulley situated at the top of the wedge, and its ends are tied to blocks
with masses m1 and m2 .
There is no friction anywhere. What is the acceleration of the wedge?
Solution
This is a classic example of a problem best solved with the Lagrangian-like techniques of M4.
By working in generalized coordinates, we won’t have to solve any systems of equations.
Let s be the distance the rope moves through the pulley, so that both blocks have speed ṡ in
the noninertial frame of the wedge. The “generalized force” is
dV
Feff = − = (m1 sin α1 − m2 sin α2 )g.
ds
Now, the kinetic energy in the lab frame will be of the form
1
K = Meff ṡ2
2
which means that, by the Euler–Lagrange equations,
Feff
s̈ = .
Meff
Our task is now to calculate Meff . Since the center of mass of the system can’t move
horizontally, the wedge has speed
m1 cos α1 + m2 cos α2
vw = ṡ.
M + m1 + m2
Now, it’s a bit annoying to directly compute the kinetic energy K in the lab frame, but it’s
easy to compute the kinetic energy in the frame of the wedge: it’s simply (m1 + m2 )ṡ2 /2.
But the two are also related simply,
1 2 1
K + (M + m1 + m2 )vw = (m1 + m2 )ṡ2 .
2 2
Using this to solve for K, we conclude
(m1 cos α1 + m2 cos α2 )2
Meff = m1 + m2 − .
M + m1 + m2
Finally, the desired answer is
m1 cos α1 + m2 cos α2
aw = s̈.
M + m1 + m2
25
[3] Problem 22 (Kalda). A rope of mass per unit length ρ and length L is thrown over a pulley so
that the length of one hanging end is ℓ. The rope and pulley have enough friction so that they do
not slip against each other.
The pulley is a hoop of mass m and radius R attached to a horizontal axle by light spokes. Find
the force on the axle immediately after the motion begins.
Solution. Let the distance the rope moves along the pulley be represented by the coordinate q. The
kinetic energy of the rope is 12 ρLq̇ 2 since every section of the rope moves with velocity q̇. Without
slipping, the kinetic energy of the pulley is 12 mq̇ 2 . Another consequence of no slipping is that energy
is conserved, so dK/dt = −dU/dt.
When the rope moves along a small distance of dq, the change in potential energy can be
calculated by considering a segment dq moving from one end to another, having a difference in
vertical height of L − πR − 2ℓ. Thus dU = −ρgdq(L − πR − 2ℓ).
dK
= (ρL + m)q̇ q̈ = ρg q̇(L − πR − 2ℓ)
dt
ρ(L − πR − 2ℓ)
q̈ = g .
ρL + m
P
The vertical normal force can be determined by mi (ay )i of the rope. The acceleration of the
parts moving up and down will cancel out, so the acceleration of the center of mass can be found
by considering the ”extra” segment of length L − πR − 2ℓ. The net vertical force on the system is
then ρ(L − πR − 2ℓ)q̈, so the force on the axle N satisfies (m + ρL)g − N = ρ(L − πR − 2ℓ)q̈.
(ρL + m)2 − ρ2 (L − πR − 2ℓ)2

Ny = g .
ρL + m
Since rope is transferred to the right, there is also a horizontal component of force on the axle.
P
When the rope moves along a distance dq, mi dxi is essentially ρdq(2R) since it can be seen as a
segment of length dq moving to the other side. Thus Fx = ρq̈(2R).
ρ(L − πR − 2ℓ)
Nx = 2ρRg .
ρL + m
[3] Problem 23 (French 5.10). Two equal masses are connected as shown with two identical massless
springs of spring constant k.
26
Considering only motion

√ in the
√ vertical direction, show that the ratio of the frequencies of the two
normal modes is ( 5 + 1)/( 5 − 1).
Solution. Let y1 denote the displacement of the upper mass and y2 for the lower mass. The
equations of motion are
mÿ1 = −ky1 − k(y1 − y2 ) = −2ky1 + ky2 , mÿ2 = −ky2 + ky1
For normal modes, the particles will oscillate at the p same frequency. Guessing a form y1 =
Aei(ωt+ϕ1 ) = Ãeiωt and y2 = B̃eiωt and defining α = ω/ k/m, the equations are
−ω 2 Ã = −2ω02 Ã + ω02 B̃, −ω 2 B̃ = −ω02 B̃ + ω02 Ã.
Dividing these equations gives

Ã 1 1 − α2
= = .
B̃ 2 − α2 1
Solving for α, we find √
4 2 3± 5 2
α − 3α + 1 = 0, α =
2
Then the ratio of the two normal mode frequencies is
s √ √
α1 1+5+2 5 5+1
= √ =√
α2 1+5−2 5 5−1
as desired.
[3] Problem 24 (Kalda). A massless rod of length ℓ is attached to the ceiling by a hinge which allows
the rod to rotate in a vertical plane.
The rod is initially vertical and the hinge is spun with a fixed angular velocity ω.
(a) If a mass m if attached to the bottom of the rod, find the maximum ω for which the configu-
ration is stable.
27
(b) [A] Now suppose another mass m and rod of length ℓ is attached to the first mass by an
identical hinge that turns in the same direction, as shown above. Find the maximum ω for
which the configuration is stable. (Hint: the configuration is unstable if any infinitesimal
change in the angles of the rods can lower the energy.)
Solution. (a) In the rotating reference frame rotating with angular velocity ω, the effective
potential from the centrifugal force mω 2 r is − mω 2 rdr = − 12 mω 2 r2 , where r is the distance
R
from the vertical axis through the hinge. In this setup, r = ℓ sin θ where θ is the angle between
the rod and the vertical. The potential energy from gravity is mgℓ(1 − cos θ). For small angles,
the potential energy is

1 2 1 1
U ≈ mgℓ 1 − 1 − θ − mω 2 ℓ2 θ2 = θ2 (mgℓ − mω 2 ℓ2 )
2 2 2
The system is stable when U ′′ (θ) > 0, so the maximum value of ω for stability is
p
ωmax = g/ℓ.
(b) Let the angles between the vertical and the upper, lower rods be θ1 , θ2 ≪ 1 respectively. The
gravitational potential energy of the lower mass is mgℓ(1 − cos θ1 ) + mgℓ(1 − cos θ2 ), and the
Taylor expansion gives U2 = 12 mgℓ(θ12 + θ22 ). Using the rotating reference frame again, the
potential energy from the centrifugal force is 12 mω 2 r2 , where r = ℓ(sin θ1 + sin θ2 ) ≈ ℓ(θ1 + θ2 )
since the hinges go in the same direction. The total potential of the system (same potential
for the first mass) is then
1 1 1
U (θ1 , θ2 ) = mgℓθ12 + mgℓθ22 − mω 2 ℓ2 θ12 − mω 2 ℓ2 (θ1 + θ2 )2
2 2 2
2 2 2 2 1 2 2 2 2
= mℓ (ω0 − ω )θ1 + (ω0 − ω )θ2 − ω θ1 θ2 .
2
Where ω02 = g/ℓ. To be stable, we want the potential energy to be at a local minimum
near that point. For this multivariable function, we will do the second derivative test (local
minimum if fxx fyy − (fxy )2 > 0 and fx x > 0, where fx = ∂f /∂x). That means the condition
is
2(ω02 − ω 2 )2 − ω 4 > 0 ω < ω0 .
Solving for the roots in the first equation yields
√
ω 2 = ω02 (2 ± 2).
Since ω < ω0 , we take the smaller root to find the upper bound for a stable configuration:
√
r
g
ωmax = (2 − 2)
l
[3] Problem 25 (PPP 104). A flexible chain of uniform density is wrapped tightly around two cylinders
as shown.
The cylinders are made to rotate and cause the chain to move with speed v. The chain suddenly
slips off the cylinders and falls vertically. How does the shape of the chain vary during the fall?
28
Solution. Firstly, notice that the problem is equivalent to the following question: Given a chain
in the plane moving such that each point has velocity directed along the chain and speed v, what
happens to the chain. Note that there is no gravity since all the interactions are happening in the
xy plane, so we can transform into the CM frame.
Interestingly enough, the answer is that the chain maintains its shape. Consider a little piece of
chain dx, and suppose the radius of curvature there is r. We see that if the tension there is T , then
T dθ is the centripetal force, so T dθ = (λdx)v 2 /r, so T = λv 2 where λ is the linear mass density. In
particular, this is independent of r. Therefore, all the pieces are in equilibrium, and so stay there,
as they have no reason to move.
[4] Problem 26 (PPP 106). A long, heavy flexible rope with mass ρ per unit length is stretched by a
constant force F . A sudden movement causes a circular loop to form at one end of the rope.
The center of the loop moves with speed c as shown.

(a) Calculate the speed c, assuming gravity is negligible.
(b) Find the energy E carried by a loop rotating with angular frequency ω.
(c) Show that the momentum p carried by the loop obeys E = pc. This is true for waves in
general, as we’ll see in W1.
(d) Find the angular momentum carried by the loop.
Solution. (a) By balancing forces on a small piece of the rope,
ρ(c2 /R)(Rdθ) = F dθ
p
which gives F = ρc2 , so c = F/ρ.
(b) The mass of the loop is m = 2πRρ. Splitting the energy into center of mass energy and
rotational energy, we have
1 1 2πF c
E = mc2 + (mR2 )ω 2 = mc2 = 2πRF = ,
2 2 ω
since c = ωR.
(c) Since the loop as a whole moves with speed c and has mass m, we have p = mc. Since
E = mc2 , we have E = pc as desired.
(d) The orbital angular momentum is
mc2
Lo = mcR =
ω
while the spin angular momentum is
mc2
Ls = Iω = mR2 ω = .
ω
Thus the total angular momentum is 2mc2 /ω.
29
5 [A] Advanced Mathematical Techniques

The following problems were cut from earlier problem sets because they required more advanced
math; however, they illustrate some very neat and important ideas.
[3] Problem 27. In P1, you found a general expression for the period of a pendulum oscillating with
amplitude θ0 in terms of an integral, then approximated the integral to find
θ02

4
ω = ω0 1 − + O(θ0 )
16
p
where ω0 = g/L. In this problem, we will show a different way to get the same answer, by solving
the equation of motion approximately. We write the solution θ(t) as a series in θ0 . The overall
solution is of order θ0 , and the corrections only depend on θ02 , so we can write
θ(t) = θ0 f0 (t) + θ03 f1 (t) + θ05 f2 (t) + . . .
where all the functions fi (t) are of order 1. Then we plug this expansion into Newton’s second law,
θ̈ + ω02 sin θ = 0, and expand it out order by order in θ0 .
(a) A naive first guess is to set f0 (t) so that it cancels precisely the order θ0 terms in this equation,
then set f1 (t) to cancel the order θ03 terms, and so on. Using this guess, show that
ω2f 3
f¨0 + ω02 f0 = 0, f¨1 + ω02 f1 = 0 0
6
where the first equation has solution f0 (t) = cos(ω0 t).
Unfortunately, this decomposition is not very useful. The problem is that two things are going on
at once: the oscillations are not quite sinusoidal, and they have an angular frequency lower than
ω0 . The expansion we’ve done would be useful if we only had the first effect, because then f1 (t)
would just capture the small, non-sinusoidal corrections to f0 (t). But our method can’t account
for the frequency shift; by construction, f0 (t) always oscillates at angular frequency ω0 . Over time,
the real oscillation θ(t) gets out of phase with f0 (t). This manifests itself as a “secular growth” in
f1 (t), i.e. it increases in magnitude every cycle until it has a huge value, of order 1/θ02 , and our
perturbative expansion breaks down.
(b) Write the right-hand side of the differential equation for f1 (t) as a sum of sinusoids, and show
that it contains a term proportional to cos(ω0 t). This resonantly drives f1 (t), causing the
secular growth.
(c) We can salvage our perturbative expansion using the method of “renormalized” frequencies.
We impose by fiat that f0 (t) oscillates at the true angular frequency, letting
f¨0 + ω 2 f0 = 0, ω = ω0 (1 − cθ02 + O(θ04 )).
Because of this choice, the differential equation for f1 (t), which contains all terms at order
θ03 , will be altered. The correct choice of ω is precisely the one for which this eliminates the
secular growth of f1 (t). Using this idea, show that c = 1/16.
30
If you keep going, you’ll find the next term f2 (t) still has secular growth. We can remove it by
having both f0 (t) and f1 (t) oscillate at frequency ω0 (1 − θ02 /16 + dθ04 ), where d is chosen to cancel
the secular growth of f2 (t). In this way, the frequency can be found to any order in θ02 . (This
technique is called the method of strained coordinates. It’s an example of multiple-scale analysis.)
Solution. (a) Plugging everything in and using sin θ = θ − θ3 /6 + O(θ5 ), we have

¨ 3¨ 2 3 3 1 3 3 5
θ0 f0 + θ0 f1 + ω0 θ0 f0 + θ0 f1 − θ0 f0 + O(θ0 ) = 0.
6
Collecting the order θ0 and θ03 terms gives the desired result.
(b) The easiest way to do this is to use the definition of cos(ω0 t) in terms of complex exponentials,
3
eiω0 t + e−iω0 t e3iω0 t + 3eiω0 t + 3e−iω0 t + e−3iω0 t

1 3
cos3 (ω0 t) = = = cos(3ω0 t)+ cos(ω0 t).
2 8 4 4
Another way is to remember the cosine triple angle identity, but who knows that?
(c) Adjusting ω0 to the renormalized frequency for f0 moves terms between the two differential
equations, so that now we have
3
¨ 2 ¨ 2 2 f0 2
f0 + ω f0 = 0, f1 + ω0 f1 = ω0 − 2cf0 + O(θ0 ) .
6
The part of the right-hand side that oscillates at angular frequency ω0 is

2 13
ω0 cos(ω0 t) − 2c cos(ω0 t)
64
from which we conclude c = 1/16.
[3] Problem 28. You might be wondering how we can solve the weakening spring problem from M4
without anything fancy like the adiabatic theorem. There is a general technique to solve linear
differential equations whose coefficients are slowly varying. First, write the equation of motion as
ẍ + ω 2 (t)x = 0.
Then expand x(t) as

x(t) = A(t)eiϕ(t) , ϕ̇(t) = ω(t).
The point of writing x(t) this way is that pulling out the factor of eiϕ(t) will automatically account
for the rapid oscillations. The factor A(t) only varies slowly, so it’s easier to handle by itself.
(a) Evaluate ẍ(t) and plug it into the equation of motion.
(b) Using the fact that A(t) and ω(t) vary slowly, throw out small terms in your equation from
part (a), until you get a differential equation you can easily integrate. This is a simple example
of the WKB approximation.
(c) Show that this gives the expected final result for a weakening spring.
31
Solution. (a) Just carrying out the time derivatives using the product rule gives
ẍ = Äeiϕ + 2iω Ȧeiϕ + iω̇Aeiϕ − ω 2 Aeiϕ .
Plugging this back into the equation of motion, the last term cancels, and we can cancel an
overall factor of eiϕ to get
Ä + 2iω Ȧ + iω̇A = 0.
(b) Let’s think carefully about how big each of these terms is. If the total time it takes for the
spring to weaken is T , where ωT ≪ 1, then each time derivative on A or ω multiplies the
magnitude of the term by roughly 1/T . So the first term is of order A/T 2 , while the other
two are of order ωA/T ≫ A/T 2 . Therefore, we can throw out the first term and to get
2Ȧ ω̇
=−
A ω
which is equivalent to
d log(A2 ) d log(1/ω)
= .
dt dt
(c) The above result tells us that A2 ω is constant, so A ∝ k −1/4 as found in M4.
[4] Problem 29 (BAUPC 1996). A mass M is located at the vertex of an angle θ ≪ 1 formed by
two massless sticks of length ℓ. The structure is held so that the left stick is initially vertical, then
released. The right stick hits the ground at time t = 0. The structure then rocks back and forth,
coming to a stop at time t = T .
(a) Prove the identity

1 1 1 π2
1+ + + + . . . =
32 52 72 8
2 = π 2 /6.
P
using the result n≥1 1/n
(b) Calculate T to leading order in θ. Performing the expansion is rather tricky and requires the
previous identity.
[3] Problem 30. In this problem, we’ll go through Laplace’s slick derivation of Kepler’s first law.
Throughout, we assume the orbit takes place in the xy plane, with the Sun at the origin.
(a) Show that

γx γy
3
ẍ = −
, ÿ = − 3
r r
where γ is a constant that depends on the parameters.
(b) Show that

d 3 d 3
(r ẍ) = −γ ẋ, (r ÿ) = −γ ẏ.
dt dt
32
(c) Show that

d 3
(r r̈) = −γ ṙ.
dt
(Hint: this can get messy. As a first step, try showing the left-hand side is equal to
(r2 /2) d3 (r2 )/dt3 . You will have to switch variables to x and y and then switch back; for
these purposes it’s useful to use the results of part (a), and the definition r2 = x2 + y 2 .)
(d) Define ψ(t) = r(t)3 . In parts (b) and (c), we have shown that the differential equation

d du
ψ(t) = −γu
dt dt
has three solutions, namely ẋ, ẏ, and ṙ. Any second-order linear differential equations only
has two independent solutions. If ẋ and ẏ are not independent, the orbit is simply a line,
which is trivial. Assuming that doesn’t happen, they are independent, so ṙ must be a linear
combination of them,
ṙ = Aẋ + B ẏ.
Use this result to argue that the orbit is a conic section.
Solution. (a) This just follows from F = ma. In terms of the usual parameters, γ = GM .
(b) This immediately follows from clearing denominators in the results of part (a) and differenti-
ating both sides.
(c) Following the hint, we have
d 3 ... 1 d3 d2
(r r̈) = r3 r + 3r2 ṙr̈ = r2 3 (r2 ) = r2 2 (rṙ).
dt 2 dt dt
At this point, we switch back to x and y. By differentiating r2 = x2 + y 2 ,
rṙ = xẋ + y ẏ.
Plugging this in gives
d 3 d2 d
(r r̈) = r2 2 (xẋ + y ẏ) = r2 (xẍ + y ÿ + ẋ2 + ẏ 2 ).
dt dt dt
We see that we’ll have a lot of factors involving ẍ and ÿ, but we know how to handle these
using part (a). Using part (a) several times, we have
γ γ γ ṙ
ẋẍ + ẏ ÿ = − 3
(xẋ + y ẏ) = − 3 (rṙ) = − 2
r r r
and
γ 2 γ
xẍ + y ÿ = − 3
(x + y 2 ) = − .
r r
Plugging these results in, we find

d 3 2 d −γ 2γ ṙ
(r r̈) = r − − 2 = −γ ṙ
dt dt r r
just as desired.
33
(d) Integrating both sides,

r = Ax + By + C.
But then squaring both sides shows that the equation of the orbit is just a quadratic in x and
y, which is precisely the form of a conic section in Cartesian coordinates. You can also show
that the focus is at the origin, though this requires a bit more knowledge about conics.
This question was inspired by this paper, which has a few more derivations of Kepler’s first law.
6 Mechanics and Geometry

For dessert, we’ll consider a few cute problems that relate statics to geometry.
Example 8
Given a triangle ABC, the Fermat point is the point X that minimizes AX + BX + CX.
Design a machine that finds the Fermat point.
Solution
We take a horizontal plane and drill holes at points A, B, and C. A mass M on a rope is
fed through each hole, and the three ends of the rope are tied together at point X. The
gravitational potential energy is proportional to AX + BX + CX, so in equilibrium X lies
on the Fermat point. Moreover, since the tensions in each rope are all equal to M g, force
balance requires ∠AXB = ∠BXC = ∠CXA = 120◦ .
[1] Problem 31. Using similar reasoning, design a machine that finds the point X that minimizes
(AX)2 + (BX)2 + (CX)2 . What geometrical property can you conclude about this point?
Solution. Attach springs at each of A, B, C with spring constant k, then X minimizes the PE, so
it is in equilibrium. Therefore, k(X − A) + k(X − B) + k(X − C) = 0, so X is the centroid of ABC.
Alternatively, we could have set the gradient of AX 2 + BX 2 + CX 2 to be 0.
Example 9
Show that the incenter of a triangle (i.e. the meeting point of the angle bisectors) exists.
Solution
Apply six forces at the vertices of a triangle as shown.
These forces clearly balance, and also produce no net torque on the triangle. Now combine
the forces applied at each vertex, yielding three forces that point along the angle bisectors.
34
By the principles of M2, the torques of these forces can only balance if their lines of action
meet at a point. Therefore the angle bisectors are concurrent, so the incenter exists.
Example 10
Let AB be a diameter of a circle, and let a mass be free to slide on the circle. The mass is
connected to two identical straight springs of zero rest length, which are in turn connected
to points A and B. At what points C can the mass be in static equilibrium?
Solution
The potential energy of the system is proportional to (AC)2 + (BC)2 . Since ABC is a right
triangle, this is just equal to (AB)2 by the Pythagorean theorem. Since the potential energy
doesn’t depend on where the mass is, it can be at static equilibrium at any point on the
circle. Alternatively, you can show that the mass is in static equilibrium by force balance,
and use the reasoning in reverse to derive the Pythagorean theorem.
[1] Problem 32. Consider a right triangle ABC filled with a fluid of uniform pressure. Using torque
balance, establish the Pythagorean theorem.
Solution. Suppose ∠C = 90, and suppose the pressure is p. Taking torques about C, we see that
pa · (a/2) + pb · (b/2) − pc · (c/2) = 0, or a2 + b2 = c2 .
[1] Problem 33. Shown below is a setup due to the physicist Stevin.
One might argue that because there are more masses on AB than on BC, this is a perpetual motion
machine that turns counterclockwise. By using the fact that perpetual motion machines don’t
actually exist, prove the law of sines.
Solution. For each mass on AB, the component of gravity along AB is proportional to sin ∠BAC.
Furthermore, the number of masses is proportional to AB. This must be balanced by the masses
along BC, giving
AB sin ∠BAC = BC∠BCA
which after minor rearrangement is the law of sines.
[2] Problem 34. Consider the n-sided polygon P of least possible area that circumscribes a closed
convex curve K. Prove that every tangency point of K with a side of P is the midpoint of that side.
(Hint: begin by supposing that the area outside P is filled with a gas of uniform pressure, with a
vacuum inside P .)
35
Solution. The minimum energy is achieved when the gas takes up the largest possible area, i.e. when
the polygon P has minimum area. Let’s model the polygon as being formed by n infinite rods, which
don’t push on each other. Now, in equilibrium, the torque on each rod must be zero, but the only
forces on the rod are the uniform pressure along the part of the rod making up the corresponding
polygon side, and the normal force at the contact point. Taking torques about the contact point
shows that it must be the midpoint.
[2] Problem 35. In this problem we’ll derive Kepler’s first law yet again, using no calculus, but a bit
of Euclidean geometry. As usual, we suppose a planet of mass m orbits a fixed star of much greater
mass M . Placing the star at the origin, let ϕ be the angle between r and v for the planet.
(a) Write down the quantities E and L in terms of G, M , m, v, r, and ϕ, and show that
L2

2 GM m
r + r sin2 ϕ = .
E 2mE
(b) Now consider an ellipse with semimajor axis a and eccentricity e, meaning that the distance
between the foci is 2ae, with one of the foci F at the origin. Consider a point P on the ellipse,
so that the angle between the tangent to the ellipse at P and F P is ϕ. If r = |F P |, show that
(r2 − 2ar) sin2 ϕ = −a2 (1 − e2 ).
You will have to use the geometrical property that a light ray sent from one focus will reflect
at the ellipse to hit the other focus.
(c) By comparing your results for (a) and (b), conclude that the orbit is an ellipse with
r
GM m 2EL2
a=− , e = 1 + 2 2 3.
2E G M m
Solution. (a) By definition, we have

1 GM m
L = mrv sin ϕ, E = mv 2 − .
2 r
Solving the second equation for v gives
r
2E 2GM
v= + .
m r
Plugging this into the first equation, squaring, and rearranging gives the desired result.
(b) Refer to the below diagram, where T T ′ is the tangent to the ellipse and N N ′ is the normal.
36
By the law of cosines,
(2ae)2 = (2a − r)2 + r2 − 2r(2a − r) cos ψ.
By the geometrical properties of the ellipse, N N ′ is the angle bisector of ∠F ′ P F , so
cos ψ = cos(π − 2ϕ) = − cos(2ϕ) = 2 sin2 ϕ − 1.
Plugging this into the law of cosines and rearranging gives the desired result.
(c) This follows immediately, from inspection. Note that this all breaks down for E ≥ 0, since
in that case the trajectory isn’t an ellipse, but similar derivations can be performed for the
parabola and hyperbola.
37

M8 Sol

Uploaded by

Copyright:

Available Formats

M8 Sol

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

M8 Sol

Uploaded by

Copyright:

Available Formats

Kevin Zhou Physics Olympiad Handouts

Mechanics VIII: Synthesis

Idea 1: Chasles’ Theorem

Example 2: KK Example 7.4

Find the angular momentum of the system.

L = r × p = mωℓ2 cos α (sin α x̂ + cos α ẑ) .

L = I⊥ ω⊥ = (2mℓ2 )(ω cos α)

directed perpendicular to the rod, which is what we just saw explicitly.

We can summarize the lessons drawn from this example as follows.

The total angular momentum about the origin is then

where the torque τ is defined as in M5. The kinetic energy is

Example 3: KK Example 7.5

Calculate the magnitude of the torque on the rod in example 2.

L = 2mωℓ2 cos α (sin α x̂ + cos α ẑ).

T cos β = M g, T sin β = M Ω2 (ℓ + L sin β).

Combining these equations gives the precession frequency

for the purposes of computing angular momentum, giving

This rotates in a horizontal circle with angular speed Ω, so

Plugging the result for Ω in above and solving for β gives

The following chain of events ensues:

• This produces a downward component of angular momentum, which is balanced by the

The string makes an angle α with the vertical.

T R cos(α − β) ≈ T R(cos α − β sin α) ≈ M R2 ω 2 β

(b) Now we will do the calculations above without b ≪ R.

components, corresponding to the rotation rate in each independent plane.

• When d = 3 the angular velocity has three components, so we treat it as a vector.

Example 5: IIT JEE 2016

Solution. By energy conservation, we have that

[3] Problem 13. IPhO 2014, problem 1A.

Solution. See the official solutions here.

µxρg cos θ − ρgx sin θ = µ(ℓ − x)ρg cos θ + ρg(ℓ − x) sin θ

µyρg cos θ + ρgy sin θ = µ(ℓ − y)ρg cos θ − ρg(ℓ − y) sin θ

4 Ropes, Wires, and Chains

2T1,y = λℓg, T1,y = λℓg/2.

Finally, consider one of the hanging portions as the system. Then

We hence have three equations, but four unknowns.

T12 − T22 = (λℓg/2)2

There is no friction anywhere. What is the acceleration of the wedge?

(ρL + m)2 − ρ2 (L − πR − 2ℓ)2

Considering only motion

mÿ1 = −ky1 − k(y1 − y2 ) = −2ky1 + ky2 , mÿ2 = −ky2 + ky1

−ω 2 Ã = −2ω02 Ã + ω02 B̃, −ω 2 B̃ = −ω02 B̃ + ω02 Ã.

Dividing these equations gives

The center of the loop moves with speed c as shown.

5 [A] Advanced Mathematical Techniques

θ(t) = θ0 f0 (t) + θ03 f1 (t) + θ05 f2 (t) + . . .

f¨0 + ω 2 f0 = 0, ω = ω0 (1 − cθ02 + O(θ04 )).

Solution. (a) Plugging everything in and using sin θ = θ − θ3 /6 + O(θ5 ), we have

The part of the right-hand side that oscillates at angular frequency ω0 is

from which we conclude c = 1/16.

Then expand x(t) as

(a) Evaluate ẍ(t) and plug it into the equation of motion.

ẍ = Äeiϕ + 2iω Ȧeiϕ + iω̇Aeiϕ − ω 2 Aeiϕ .

(a) Prove the identity

Solution. See the official solutions here.

(a) Show that

(b) Show that

(c) Show that

(c) Following the hint, we have