Chapter 6

Rigid Body Dynamics

6.1 Introduction
In practice, it is often not possible to idealize a system as a particle. In this section, we construct a more
sophisticated description of the world, in which objects rotate, in addition to translating. This general
branch of physics is called ‘Rigid Body Dynamics.’

Rigid body dynamics has many applications. In vehicle dynamics, we are often more worried about
controlling the orientation of our vehicle than its path – an aircraft must keep its shiny side up, and we don’t
want a spacecraft tumbling uncontrollably. Rigid body mechanics is used extensively to design power
generation and transmission systems, from jet engines, to the internal combustion engine, to gearboxes. A
typical problem is to convert rotational motion to linear motion, and vice-versa. Rigid body motion is also
of great interest to people who design prosthetic devices, implants, or coach athletes: here, the goal is to
understand human motion, to protect athletes from injury or improve their performance, or to design
devices that replicate the complicated motion of a human joint correctly. For example, Professor Crisco’s
orthopaedics lab at Brown studies human motion and the forces they generate at human joints, to help
understand how injuries occur and how they can be prevented.

The motion of a rigid body is often very counter-intuitive. That’s why there are so many toys that exploit
the properties of rigid bodies: the motion of a spinning top; a boomerang; the ‘rattleback’ and a Frisbee can
all be explained using the equations derived in this section.

Here is a quick outline of how we analyze motion of rigid bodies.

1. A rigid body is idealized as an infinite number of small particles, connected by two-force members.
2. We already know the equations of motion for a system of particles (Section 4 of the notes):
dp d N
The force-momentum equation ∑ Fiext= = ∑ mi vi
dt dt i =1
i

ext dh d N
The moment – angular momentum equation ∑ i i dt =dt ∑ ri × mi vi
r × F =
i i =1
dT d N 1
The work-kinetic energy equation ∑ Fiext ⋅ vi = = ∑ mi vi ⋅ vi
dt dt i =1 2
i
3. These equations tell us how a rigid body moves. But to use them, we would need to keep track
track of an infinite number of particles! To simplify the problem, we set up some mathematical
methods that allow us to express the position and velocity of every point in a rigid body in terms of
the position rG , velocity vG and acceleration aG of its center of mass, and its rotation tensor
R(quantifying its orientation) and its angular velocity ω , and angular acceleration α . This allows
us to write the linear momentum, angular momentum, and kinetic energy of a rigid body in the form
1 1
p = MvG h= rG × MvG + IG ω T
= MvG ⋅ vG + ω ⋅ IG ω
2 2
where M is the total mass of the body and IG is its mass moment of inertia.
4. We can then derive the rigid body equations of motion:
∑ Fiext= MaG ∑ ri × Fiext= MrG × aG + IG α + ω × [ IG ω ]
i i
2

6.2 Describing Motion of a Rigid Body

We describe motion of a particle using its position, velocity and acceleration. We can describe the position
of a rigid body in the same way - we could specify the position, velocity and acceleration of any convenient
point in the body (we usually use the center of mass). But we also need a way to describe the orientation of
a rigid body, and its rotational motion.

In this section, we define the various mathematical quantities that we use to describe rotation, angular
velocity, and angular acceleration.

k
6.2.1 Describing rotations: The Rotation Tensor (or matrix)

Rotations are quantified by a mathematical object called a rotation

tensor. It is defined as follows: B
1. Choose some convenient initial orientation of the rigid body (eg pB-pA
for the rectangular prism in the figure, we chose to make the faces A j
perpendicular to the {i, j, k} directions.
2. When the body is rotated, every line in the body (eg the sides)
i
moves to a new orientation, without changing its length. We can k
describe this orientation change as a mapping. Let A and B be two
arbitrary points in the body. Let p A , p B be the initial positions of
these points, and let rA , rB be their final positions. We introduce B
the ‘rotation tensor 1’ R which has the property that rB-rA
rB − rA= R (p B − p A )
A j
When we solve problems, we always express vectors as components in i
some basis. When we do this, R becomes a matrix. For example, if
p B − p A = x0 i + y0 j + z0k rB − rA = xi + yj + zk
we would write
 x   Rxx Rxy Rxz   x0 
 y = R R R
 
   yx yy yz   y0 
 z    
 R yz Rzy Rzz   z0 
Here, R11 , R12 ,... are a set of nine numbers (or sometimes formulas). Following the usual rules of matrix-
vector multiplication, this is just a short-hand notation for
x = Rxx x0 + Rxy y0 + Rxz z0
y = R yx x0 + R yy y0 + R yz z0
z = Rzx x0 + Rzy y0 + Rzz z0
The subscripts on R are meant to you help remember what each element in the matrix does – for example,
Rxx maps the x0 onto x, Rxy maps the y0 onto x, and so on.

1
By definition, a ‘second order tensor’ maps a vector onto another vector. In actual calculations R is always just a
matrix, but ‘tensor’ sounds better.
 3

So when we solve a problem, how do we go about finding R? Let me count the ways:

Rotations in two dimensions: B

j pB-pA
Life is simple in 2D. In this case our rigid body must lie in the i,j plane, so
we can only rotate it about an axis parallel to the k direction. A counter- A
clockwise rotation through an angle θ about the k axis is produced by 2 i
cos θ − sin θ 
R=  B
 sin θ cos θ 
For example, a vector Li that start parallel to the i axis is mapped to j rB-rA
cos θ − sin θ   L   L cos θ  θ
 sin θ cos θ=   0   L=  L cos θ i + L sin θ j
     sin θ 
A i
Rotation about a known axis

3D is a bit more difficult. Any rotation can always be expressed as a rotation through some angle θ about
some axis parallel to a unit vector n (we always use the right hand screw convention). In some problems
you can see what n and θ are: then you can write down a unit vector parallel to n
n = nx i + n y j + nz k
and then use the ‘Rodriguez Formula’
 cos θ + (1 − cos θ )nx2 (1 − cos θ )nx n y − sin θ nz (1 − cos θ )nx nz + sin θ n y 
 
R = (1 − cos θ )nx n y + sin θ nz cos θ + (1 − cos θ )n 2y (1 − cos θ )n y nz − sin θ nx 
 
 (1 − cos θ )n n − sin θ n (1 − cos θ )n n + sin θ n 2 
cos θ + (1 − cos θ )nz 
 x z y y z x 
(This formula is impossible to remember – that’s what Google is
for). n
k
If you are given a rotation matrix R, and need to find n and θ , you
can use the formulas:
1 + 2cos θ = Rxx + R yy + Rzz j
i
1 
= n
2sin θ 
( ) (
Rzy − R yz i + ( Rxz − Rzx ) j + R yx − Rxy k ) 

j

The second formula blows up if sin(θ ) = 0 . If θ is zero or 2π you

can simply set R = 1 (the identity), and n can be anything you like.
For θ = π you can use
Rxx − cos θ R yy − cos θ Rzz − cos θ
n= i± j± k
1 − cos θ 1 − cos θ 1 − cos θ
The signs of the square roots have to be chosen = so that nx n y R= xy / 2 nx nz R= xz / 2 n y nz R yz / 2

2
(Tip: it’s easy to remember this but it’s hard to remember where to put the negative sign. You can always
figure this out by noting that a 90 degree counter-clockwise rotation maps a vector parallel to the i direction
onto a vector parallel to the j direction.)
4

In robotics, game engines, and vehicle dynamics the axis-angle representation of a rotation is often stored as
a quaternion. We won’t use that here, but mention it in passing in case you come across it in practice. A
quaternion is four numbers [q0 , q x , q y , q x ] that are related to n and θ through the formulas:
q0 = cos(θ / 2)
= sin(θ / 2)
q x nx= y sin(θ / 2)
q y n= q z nz sin (θ / 2 )

k
Mapping the coordinate axes

In some problems we might know what happens to vectors that are

parallel to the {i,j,k} directions in the initial rigid body (eg we might
know what happens to the sides of our rectangular prism). For
example, we might know that {i, j, k} map to (unit) vectors a, b, c . j
In that case we can write down each of a, b, c as components in i
{i, j, k}
k
a =ax i + a y j + az k b =bx i + by j + bz k c =cx i + c y j + cz k
and use the formula c
 ax bx cx 
 
R = a y by cy 
a a
 z bz cz  j
i b

A sequence of rotations

Suppose we rotate an object twice (perhaps about two different axes). How do we describe the result of
two rotations? That’s not hard. Suppose we do the first rotation with one mapping
rB −=rA R (1) (p B − p A )
Now we rotate our body again – this maps rB − rA onto some new vector u B − u A :
(u B − u A=) R (2) (rB − rA )
We can therefore write
− u A ) R (2) R (1) (p B − p A )
(u B=
We see that Sequential rotations are matrix products
R = R (2) R (1)

Health warning: Matrix products (and hence sequences of rotations) do not commute
R (1) R (2) ≠ R (2) R (1)
For example, the figure below shows the change in orientation caused by (a) a 90 degree positive rotation
about i followed by a 90 degree positive rotation about k (the figure on the left); and (b) a 90 degree
positive rotation about k followed by a 90 degree positive rotation about i (the figure on the right).
 5

(1) i rotation
(2) k rotation (1) k rotation
(2) i rotation

j
i
k
k
j
b

j
j i
i

Orthogonality of R

The rotation tensor (matrix) has a very important property:

If you multiply R by its transpose, the result is always the identity matrix.
Another way to say this is that
The transpose of R is equal to its inverse

Let’s try this with the 2D rotation matrix

cos θ − sin θ   cos θ sin θ  cos 2 θ + sin 2 θ 0  1 0 
=RRT =    =   
 sin θ cos θ   − sin θ cos θ   0 sin 2 θ + cos 2 θ   0 1 
 cos θ sin θ   cos θ − sin θ  1 0 
= RT R =    
 − sin θ cos θ   sin θ cos θ   0 1 
A matrix or tensor with this property is said to be orthogonal.

Why is this? It turns out that a length-preserving mapping must be an orthogonal tensor. To see this, let’s
calculate the length of the rotated vector rB − rA= R (p B − p A ) . We need to remember two vector/matrix
operations:
1. We can calculate the length of a vector by dotting it with itself and taking the square root
u ⋅ RT Ru
2. For a vector u and a matrix R, we know (or can show!) that ( Ru ) ⋅ ( Ru ) = ( )
This means
( rB − rA ) ⋅ ( rB − rA )= {R (p B − p A )} ⋅ {R (p B − p A )}= { }
(p B − p A ) ⋅ RT R (p B − p A )
But we want the length of rB − rA to equal the length of p B − p A , which means we need R to satisfy
6

{
(p B − p A ) ⋅ RT R (p B − p A ) =} (p B − p A ) ⋅ (p B − p A )

{ }
⇒ (p B − p A ) ⋅ RT R (p B − p A ) − (p B − p A ) ⋅ {1(p B − p A )} =
0

⇒ (p B − p A ) ⋅ {( RT R − 1) (p B − p A )} =
0
where 1 is the identity tensor (we normally use I for the identity tensor, but rigid body dynamics uses I to
denote the mass moment of inertia so it’s already been taken….). With a bit of busy work, we can show that
the last line can only be satisfied if RT R = 1 . In fact, a rigorous mathematical derivation of rotations
starts with the statement that R must preserve the length of all vectors, and then derives all the other
material in this section from that statement. This is not easy to follow the first time around, but will
probably be the approach used in more advanced courses.

Examples: B
1. Write down the rotation matrix for the 2D rotation shown in B rB-rA
the figure
j pB-pA j
A A
The object rotates 90 degree counterclockwise about the k axis,
i
so i
cos θ − sin θ  0 −1
=R =   
 sin θ cos θ  1 0 

2. The object shown in the figure is first rotated 90 degrees

about the i axis, and then 180 degrees about the j axis. Find
the rotation tensor.

We can construct the two rotations using the Rodriguez

formula. For the first rotation θ = π / 2
n =⇒i nx = 1 ny = nz = 0
1 0 0 
= R (1)0 0 −1
 
0 1 0 
For the second rotation θ = π n =⇒
j ny = 1 nx =nz =0

 −1 0 0 
R (2)
=  0 1 0 
 0 0 −1
The total rotation is therefore
 7

 −1 0 0  1 0 0   −1 0 0 
=R R =
R  0 1 0  0 0 =
(2) (1)
−1  0 0 −1
 
 0 0 −1 0 1 0   0 −1 0 

3. Find the axis-angle representation for the combined rotation in problem (2).

We can calculate the axis and angle of this rotation using the
formulas
1 + 2cos θ =Rxx + R yy + Rzz ⇒ 2cos θ =−2 ⇒ θ =π

Rxx − cos θ R yy − cos θ Rzz − cos θ

n= i± j± k
1 − cos θ 1 − cos θ 1 − cos θ
−1 − (−1) 0 − (−1) 0 − (−1) 1
= i± j± k =( j ± k )
1 − (−1) 1 − (−1) 1 − (−1) 2

To decide which of these two choices to use we notice that

R yz = −1 , which tells us that n y nz < 0 . The answer is
therefore
1
=θ π=
, n (j − k)
2
It is incredibly difficult to visualize the effect of a rotation about an arbitrary axis (at least for me). In fact
this formula looks wrong – how can a 180 degree rotation end up tipping the box on its side? But the
answer is right, as the animation (which will only show up in the html version of the notes) shows.

6.2.2 Describing rotational motion: The angular velocity vector and spin tensor

We described the location of a particle in space using its

position vector, and its motion using velocity. We need to
come up with something similar to velocity for rotations.

Definition of an angular velocity vector Visualize a

spinning object, like the cube shown in the figure. The box
rotates about an axis – in the example, the axis is the line
connecting two cube diagonals. In addition, the object turns
through some number of revolutions every minute. We
would specify the angular velocity of the shaft as a vector n Axis of
rotation
ω , with the following properties:

1. The direction of the vector is parallel to the axis of

the shaft (the axis of rotation). This direction would
be specified by a unit vector n parallel to the shaft.

2. There are, of course, two possible directions for n.

By convention, we always choose a direction such
that, when viewed in a direction parallel to n (so the vector points away from you) the shaft appears
 8

to rotate clockwise. Or conversely, if n points towards you, the shaft appears to rotate
counterclockwise. (This is the `right hand screw convention’)

Viewed along n Viewed in direction opposite to n

3. The magnitude of the vector is the angular speed dθ / dt of the object, in radians per second. If
you know the revs per minute n turned by the shaft, the number of radians per sec follows as
dθ / dt = 120π n . The magnitude of the angular velocity is often denoted by ω = dθ / dt

dθ
=
The angular velocity vector is then ω = n ωn .
dt

Since angular velocity is a vector, it has components ω = ω x i + ω y j + ω z k in a fixed Cartesian basis.

As always, in two dimensions, everything is very simple. In this case objects can only rotate about the k
axis, and we can write the angular velocity vector as
dθ
ω= k
dt
where θ is the counterclockwise angle of rotation of any line embedded in the body.

Writing down angular velocities:

For 2D problems, we always know the direction of the angular velocity and can just use ω = ω z k to write it
down (of course if we know the value or a formula for ω z we can use it).

For 3D problems, we can often use vector addition to write down ω . We can illustrate this with a simple
example:

Example: The propeller on the aircraft shown in the

figure spins (about its axis) at 2000 rpm. The
aircraft travels at speed 200 km/hr in a turn with
radius 1 km. What is the angular velocity vector
of (i) the body of the aircraft, and (ii) the propeller?
Express your answer in the normal-tangential-
vertical basis.

(i) The circumference of the circle is

= s 2= π R 2π km . The airplane completes a full circle in t = s / V = (2π / 200) × 3600 = 36π sec . A full
turn is 2π radians, so the aircraft body turns at a rate 2π / (36π ) = (1 / 18)k rad/s about the k axis.
 9

(ii) The propeller turns at 2000 rpm relative to the body of the plane. The angular velocity of the prop with
respect to a stationary observer is therefore the vector sum of the 2000 rpm about the t axis, plus the angular
velocity of the body. This gives
1 10π 1
ω prop = [ 2000 × 2π / 3600] t + k = t + k rad/s
18 9 18

Relation between the rotation matrix and the angular velocity vector: the spin tensor

We might guess that the angular velocity vector is the derivative of the rotation tensor. This is sort of
correct, but the full story is a bit more complicated. The relationship between R and ω is constructed as
follows:
1. We define the spin tensor W as
dR T
W= R
dt
2. The spin tensor is always skew ( W = − WT ), and we can read off the angular velocity vector by
looking at its components. Specifically, if ω = ω x i + ω y j + ω z k then
 0 −ω z ω y 
 
= W  ωz 0 −ω x 
 −ω 0 
 y ωx
We can use this formula in two ways: (1) Given R, we can calculate W and then read off the angular
velocity vector components. Alternatively, if we know ω , we can calculate R by first constructing W, then
integrating the formula
dR
= WR
dt

Angular velocity-rotation relations in 2D

We can check this for the special case of a 2D rotation:

 dθ dθ   dθ 
− sin θ − cos θ  0 −
cos θ − sin θ  dR T  dt dt  cos θ sin θ   dt 
= R   ⇒ dt= R   =   
 sin θ cos θ   dθ cos θ − dθ sin θ   − sin θ cos θ   dθ 0 
 dt dt   dt 

dθ
As expected, we find that ω z = .
dt

This means that in 2D, angular velocity and the angle of rotation θ are related by the same formulas as
distance traveled and speed for position. We can use all the same rules of calculus to go back and forth
between them.

Angular velocity-Spin tensor formula

10

There is an important formula relating W and ω . Let rB − rA be a vector joining any two points in a rigid
body. Then
W (rB − rA ) =ω × (rB − rA )

You can see this by just multiplying out the definition of W and comparing the result to the cross product: if
rB − rA = xi + yj + zk , then

 0 −ω z ω y   x  ω y z − ωz x 
   
W (rB − rA )=  ω z 0 −ω x   y  =  ωz x − ωx z 
 −ω 0   z  ω y − ω x 
 y ωx  x y 
Hopefully you can see that this is the same as the cross product!

6.2.3 The angular acceleration vector

Angular acceleration is the time derivative of angular velocity

dω
α=
dt
For 3D, we can use
dωx dω y dωz
= αx = αy = αz
dt dt dt
For 3D, we can’t express the angular accelerations or velocities as derivatives of rotation angles, because
these can’t be defined for a general motion.

For a 2D problem, the direction of angular velocity and acceleration are known, so we have
= α α= zk ω ωzk
The components are related by
d ω z d 2θ dωz
αz
= = = ωz
2
dt dt dθ
For 2D problems, we can use all the usual rules of calculus to go from angular acceleration to angular
velocity to angle, and vice-versa (just like distance-speed-acceleration formulas for straight line motion).

6.2.3 Relative velocity and acceleration of two points in a rigid body

We now know how to describe rotational motion. Our next order of business is to discuss a couple of very
important formulas that we use to analyze the motion of a system of
rigid bodies, and also to derive formulas for the angular momentum k
and kinetic energy of a rigid body.. vB,aB
ω,α
Consider a rigid body: B
Let ω be the (instantaneous) angular velocity of the body, rB-rA
and W the corresponding spin tensor
A j
i
vA,aA
 11

Let A and B be two arbitrary points in a rigid body, and let rA , rB and v A , v B , a A , a B be their
(instantaneous) position, velocity and acceleration vectors.

Then the relative position and velocity of A and B are related by

v B − v A =ω × (rB − rA )
v B − v A= W (rB − rA )

The relative acceleration of A and B are related their relative positions and velocity by
aB − a A = α × (rB − rA ) + ω × ( v B − v A ) =α × (rB − rA ) + ω × [ω × (rB − rA )]

For 2D problems only: we can simplify these, because we know ω is vB,aB

always parallel to the k direction. Therefore j B
v B − v A = ω z k × (rB − rA )
a B − a A = α z k × (rB − rA ) − ω z2 (rB − rA )
rB-rA
θ ωz α z
A i
vA,aA
Proof: These fornulas are easy to prove. Remember the mapping:
d dR
rB − rA= R (p B − p A ) ⇒ v B − v A= (rB − rA=
) (p B − p A )
dt dt
Also,
rB − rA = R (p B − p A ) ⇒ RT (rB − rA ) = RT R (p B − p A ) = (p B − p A )
Hence
dR T
v B − v A= R (rB − rA= ) W (rB − rA )
dt
Remember that W (rB − rA ) =ω × (rB − rA ) , so the acceleration formula then follows as
d dω d
aB − a A = ( vB − v A ) = × (rB − rA ) + ω × (rB − rA ) = α × (rB − rA ) + ω × ( v B − v A )
dt dt dt

6.3 Analyzing motion in connected rigid bodies

The formulas in 6.2.3 are used to analyze motion in

machines. A typical problem is illustrated in the figure.
An actuator moves point B on the car jack shown in the
figure horizontally with constant velocity V. What are the F
velocity and acceleration of the platform (CF)? D L C
You could probably solve this rather simple example with j L
elementary trig, but we need a more systematic method for A θ BV i
general problems, especially to analyze 3D motion. Here’s
the general procedure
1. Define variables to denote the unknown angular
12

velocities and angular accelerations of each rigid body in the system

2. Write down all the known velocities in the system
3. Use the rigid body formulas
v B − v A =ω × (rB − rA )
to write down equations relating velocities of the connections, joints, or contacts on each rigid body
4. Write down constraint equations relating velocities of the two connected rigid bodies at each
connection, joint, or contact
5. Solve the equations for unknown velocities of connections, and the angular velocities of the rigid
bodies.
6. Finally, once the velocities are known, write down equations for the accelerations of pairs of
joints/contacts/connections on each rigid body
a B − a A = α × (rB − rA ) + ω × ( v B − v A )
7. Write down constraints equations for accelerations at connected points
8. Solve the equations in 6,7 for unknown accelerations and angular accelerations.

This all sounds terribly complicated, so let’s solve a few examples to show how it works in practice.

Example 1: In the figure shown the link AB rotates counter-

clockwise with constant angular speed 4 rad/s. Point C on B j
member BC is constrained to move horizontally. Calculate the
velocity and acceleration of point C. i
2m
Calculating the velocity: C
A
• We know A is stationary, and are given the angular velocity of
AB. We can use the rigid body formula to find the velocity of 2m
B:
v B − v A =ω zAB k × (rB − rA ) = 4k × 2 j
⇒ vB =
−8i

• We don’t know the angular velocity of BC, so we introduce ω zBC as an unknown, and use the rigid
body formula for member BC to write down an equation for the velocity of C
vC − v=B ω zBC k × (rC − rB=) ω zBC k × (2i − 2=
j) 2ω BC i + 2ω BC j
⇒ vC =−8i + 2ω BC i + 2ω BC j

• We know that C can only move horizontally. This means that its j component of velocity must be
zero. This shows that
ωBC = 0, v C = −8i

Calculating the acceleration:

• We know A is stationary, and are given the angular velocity and angular acceleration of AB. We
can use the rigid body formula to find the acceleration of B:
2
aB − a A = α zAB k × (rB − rA ) − ω zAB (rB − rA ) =
−32 j
⇒ aB = −32 j
 13

• We don’t know the angular acceleration of BC, so we introduce α zBC as an unknown and use the
rigid body formula for member BC to write down an equation for the acceleration of C
2
B α zBC k × (rC − rB ) − ω zAB (rC − rB=
aC − a= ) α zBC k × (2i − 2 j) − 0
⇒ aC = −32 j + 2α BC i + 2α zBC j

• Point C can only move horizontally, so it can’t have any vertical acceleration. This means
that the j component of acceleration is zero:
2α zBC − 32 =0 ⇒ α zBC =16
⇒ aC =
32i

Example 2: For a more complicated example, we can solve

the car jack problem posed at the start of this section. An F
actuator moves point B on the car jack shown in the figure D L C
horizontally with constant velocity V. What are the velocity j L
and acceleration of the platform (CF)?
A θ BV i
The system contains 3 rigid bodies (AC, BD, CF ). We 3

don’t know the angular velocities or accelerations of any of

them, so we denote them by unknowns ω zAC ,
ω zBD ω zCF , α zAC , α zBD α zCF

Calculating the velocity:

• We start at point(s) with known velocity: A is stationary, and the velocity of B is given:
= v A 0= v B Vi
• Point E lies on both member AC and on member BD. We use the rigid body formulas to write
down an equation for the velocity of E on each member (notice we use the 2D equations):
A ω zAC k × (rE − rA )
v E − v=
v E − v=B ω zBD k × (rE − rB )
• The two members AC and BD are pinned together at E and so must have the same velocity. We
can eliminate v E and write out the position vectors in i,j components
ω zAC k × ( L cos30i + L sin 30 j)=
− Vi ω zBD k × (− L cos30i + L sin 30 j)
(−ω zAC L sin 30 − V )i + ω zAC L cos30 j = −ω zBD L sin 30i − ω zBD L cos30 j
The i,j components give two equations for ω zAC , ω zBD
−ω zAC L sin θ − V = −ω zBD L sin θ
ω zAC L cos θ = −ω zBD L cos θ
⇒ −ω zAC 2 L sin θ cos θ − V cos θ =0
⇒ ω zAC = −V / (2 L sin θ ) ω zBD = V / (2 L sin θ )

• We can now use the rigid body formulas for members AC and BD to find the velocities of C and D

3
You may be wondering why only a single point was defined at C and E, but there are two points at D and F. That’s
because at C and E the members are pinned together, but there is a roller at D. At E, members AC, BD always have
the same velocity and acceleration – we can just use a single variable to denote the velocity of this point. The same is
true at C. Members CF and BD touch at F and D, but point D on AB does not have the same horizontal velocity as
point F CF, so we need to be able to distinguish between them.
14

−V
vC − v A =ω zAC k × (rC − rA ) ⇒ vC = k × (2 L cos θ i + 2 L sin θ j) =Vi − V cot θ j
2 L sin θ
V
vD − vB = ω zBD k × (rD − rB ) ⇒ v D =+
Vi k × (2 L cos θ i + 2 L sin θ j) =
−V cot θ j
2 L sin θ

• We can use the rigid body formula for CF to relate the velocities of C and F
v F − vC= ω zCF k × (rF − rC )
vF = Vi − V cot θ j − ω zCF 2 L cos θ j
• Point D on CD and point F on CF must have the same vertical velocity (the roller at D allows their
horizontal velocities to differ). This can be expressed as
v F ⋅ j = v D ⋅ j ⇒ −V cot θ − ω zCF 2 L cos θ = −V cot θ
⇒ ω zCF = 0
• All points on CF therefore have the same velocity (equal to the velocity of C)
vCF= Vi − V cot θ j

Calculating the acceleration.

• We can now calculate the accelerations. We start at a known point: Points A and B have zero
acceleration.
• We can use the rigid body formula to calculate the acceleration of E on each of AC and BD:
2
A α zAC k × (rE − rA ) − ω zAC (rE − rA )
a E − a=
2
B α zAD k × (rE − rB ) − ω zAD (rE − rB )
a E − a=
• The two members are connected at E and so must have the same acceleration there. This shows
that
2
α zAC k × ( L cosθ i + L sin θ j) − ω zAC ( L cos θ i + L sin θ j)
2
=α zAD k × (− L cos θ i + L sin θ j) − ω zAD (− L cos θ i + L sin θ j)
2
⇒ α zAC ( L cos θ j − L sin θ i ) − ω zAC ( L cos θ i + L sin θ j)
2
= α zAD (− L cos θ j − L sin θ i ) − ω zAD (− L cos θ i + L sin θ j)
• The i,j components give two equations for the unknown angular accelerations:
2 2
−α zAC L sin θ − ω zAC L cos θ =
−α zAD L sin θ + ω zAD L cos θ
2 2
α zAC L cosθ − ω zAC L sin θ =
−α zAD L cos θ − ω zAD L sin θ
2
⇒ −2α zAC L sin θ cos θ = ω zAD L(cos 2 θ − sin 2 θ ) + ω zAC
2
L ⇒ α zAC = −V 2 cos θ / (4 L2 sin 3 θ )
α zAD = −α zAC = V 2 cos θ / (4 L2 sin 3 θ )
• We can use the rigid body acceleration formulas to calculate the velocities of D and C:
 15

2
A α zAC k × (rC − rA ) − ω zAC (rD − rA )
aC − a=
V 2 cos θ V2
aC =
− (2 L cos θ j − 2 L sin θ i ) − (2 L cos θ i + 2 L sin θ j)
4 L2 sin 3 θ 4 L2 sin 2 θ
V 2 cos θ V2 1
aC =
− i− j
L sin 2 θ 2 L sin 3 θ
2 V2 1
aD − aB = α zAD k × (rD − rB ) − ω zAD (rD − rB ) ⇒ a D = − j
2 L sin 3 θ
• We can use the rigid body formula to relate the accelerations of C and F
2
C α zCF k × (rF − rC ) − ω zCF (rF − rC )
a F − a=
V 2 cos θ V2 1
⇒ a F =− i− j + α zCF k × (−2 L cos θ i )
L sin 2 θ 2 L sin 3 θ
• Finally, we know that D and F must have the same vertical acceleration (so they remain in contact).
Their horizontal accelerations may differ, because of the roller attached to D. This gives
a D ⋅ j= a F ⋅ j
V2 1 V2 1
⇒− − α zCF 2 L cos θ = − ⇒ α CF = 0
2 L sin 3 θ 2 L sin 3 θ
• Since CF has zero angular velocity and angular acceleration, all points on CF have the same
acceleration (which must equal that of point C). Therefore
V 2 cos θ V2 1
aCF =
− i− j
L sin 2 θ 2 L sin 3 θ

6.3.1 Summary of constraint equations at joints and contacts

As the examples in the preceding section show, the keys to analyzing motion in a system of connected rigid
bodies are: (1) the formulas for relative velocity and acceleration of two points in a rigid body, and (2)
constraints that relate the velocities and accelerations on two bodies at points where they touch.

There are three common types of connection between rigid bodies:

1. A pin joint: the two connected members must have the same
velocity and acceleration at the connected point A B
= v B v= A aB a A

2. A slider joint: the two connected members must have the A n

same velocity and acceleration normal to the slider
vB ⋅ n = v A ⋅ n aB ⋅ n = a A ⋅ n B
16

3. Contact between two objects without relative slip (sliding) at t

the contact (friction forces must act to prevent the slip, in
general): The velocities of the touching objects must be n
equal at the contact point. The tangential components of A B
acceleration must also be equal (the normal components of
acceleration differ)
vB= v A a B ⋅ t= a A ⋅ t

6.3.2 The Rolling Wheel

Wheels are everywhere. They can be analyzed using the general rigid body equations, but it’s helpful to be
able to avoid all the tedious cross products. In this section we summarize special formulas for velocity and
acceleration of points on a wheel.

Motion of a wheel rolling without slip on a stationary surface

It is surprisingly difficult to visualize the motion of a wheel. The figure above might help: it shows the
trajectory of one point on the circumference of the wheel. The point traces quite a complicated path. The
important thing to notice is:

If a wheel rolls without slip on a stationary surface, the point touching the surface is stationary

Each point is only in contact with the ground for an instant, and while it touches the ground it has a large
vertical acceleration, but it is instantaneously stationary. We know this from the list of constraints in Sect
6.3.1, of course, but it’s still not an easy thing to visualize.

More generally, the ground need not necessarily be stationary (or the wheel could touch another surface).
In this case we know that the contacting points on two bodies in rolling contact have equal velocity at the
contact.

Angular velocity-linear velocity formula: With this insight, we can use

the rigid body formulas to calculate the instantaneous velocity vector for
ωz
any point on the wheel. Assume that R vxO
• The wheel rolls with angular velocity ω = ω z k counterclockwise
rotation is positive. j O
• The center of the wheel moves with velocity vO = vxO i
The rolling wheel formula gives
vxO = −ω z R i C
 17

To see this, you can simply use the rigid body formula to go from the contact point (which is stationary) to
O
vO − vC = ω × (rO − rC ) ⇒ vO =ω z k × ( Rj) =
−ω z Ri

More generally, we can calculate the velocity of any point on the wheel
we might be interested in. In fact, we can just write down the velocity ωz D
of any point in the wheel by noticing that instantaneously all points are
in circular motion about the contact point (just imagine the disk is R
rotating about C). See if you can show all the following:
• vA = −ω z R (i + j) j A O
B
• v D = −ω z 2 Ri θ
• vB = −ω z R (i − j)
Notice that the direction of the velocity at each point is always
i C
perpendicular to the line connecting to the point to C.

Angular acceleration-linear acceleration formula: Assume that

• The wheel rolls with angular acceleration α = α z k ωz
counterclockwise rotation is positive.
αz
R axO
• The center of the wheel moves with acceleration aO = a xO i
The rolling wheel formula gives j O
a xO = −α z R

i C
You can derive this formula in two different ways:
(1) Differentiate the velocity formula vxO = −ω z R with respect to time
(2) Use the rigid body formula:
α (rO − rC ) − ω z2 (rO − rC )
(aO − aC ) =×
⇒ aO = aC − α z Ri − ω z2 Rj
We know that the i component of acceleration at point C has to be the same as the i component of
acceleration of the ground (i.e. zero). (The j components don’t have to be equal). We also know
that O has no j acceleration, because it remains at the same height above the ground. Therefore
a xO i = a yC j − α z Ri − ω z2 Rj

−α z R
⇒ a xO = ω z2 R
a yC =
We can calculate the acceleration of any other point on the disk using the rigid body formula.

Example: The block AB has horizontal

acceleration a and horizontal speed v. A B v,a
ωz
Calculate the angular velocity and angular αz R R
acceleration of the rollers. Then, calculate
the linear velocity and acceleration of O j O O
To solve problems like this we use two
ideas: (1) the formulas relating velocity and i C
accelerations of points on the disk; and (2)
the tangential velocity and acceleration of contacting points are equal.
18

Here, we know the tangential velocity at C is zero; the tangential velocity at A is vi . We can use the
wheel formulas
vxA = −2ω z R ⇒ ω z =−v / (2 R )
Similarly, the tangential acceleration at A is ai . The rolling wheel formula gives
a xA = −2α z R ⇒ α z =− a / (2 R )

To find the velocity and acceleration at O, we can use

vxO = −ω z R =
v/2
−α z R =
a xO = a/2

6.3.3 Gears

Gears can be analyzed in much the same way as a rolling

wheel. Gears are used to increase or decrease angular
velocities (they act like mechanical amplifiers): for example,
in the animation the small gear is rotating at twice the angular
rate of the large one. They also modify the torques (or
moments) applied to the gears: if a gear system increases
angular velocity, it reduces torque by the same factor (so the
torque on the small gear in the animation is half that on the
large one). Some clever gear systems can even be used to add
angular velocities (see the discussion of epicyclic gears below.

There are many different gear designs. Here, we focus only on two-dimensional ‘spur gears’. Spur gears
have a rather complicated geometry, which we don’t have time to discuss in detail in this course. They are
designed to behave like two wheels which roll against each other with no slip at the contact. The wheel
radius is equal to the ‘pitch circle radius’ of the gears (which is slightly smaller than physical diameter of
the gears, because the teeth have to overlap). Gear manufacturers often specify the number of teeth on a
gear instead of its radius. The number of teeth and the radius have to be related, because the teeth have to
be the same circumferential distance apart for the gear pair to mesh.

We analyze motion of gears using two ideas:

(1) Two meshed gears must have equal velocities at the point where they touch.
(2) The rigid body formula, relating the velocity of points on the
circumference of the gear to the velocity of its center:
vC = vO + ω z k × (rC − rO )
ωz B
In practice we don’t usually bother doing the cross product, and R
instead just write down the velocity on the circumference
directly using the figure provided: j C O A
• v= A v O + ω z Rj θ
• v= B v O − ω z Ri
• v= C v O − ω z Rj
i D
• v= D v O + ω z Ri
You don’t have to remember these – just visualize every point on the gear moving in circular
motion (counterclockwise) around O, and write down the vectors (be careful with signs!).
 19

Example 1: The left gear in the figure rotates with counterclockwise

angular velocity ω z1 . The large gear has radius R1 and N1 teeth, the ω z1 ω z2
small one has radius R2 and N 2 teeth. Calculate the angular velocity
C R2
of the smaller gear. O
j O
Note: R1
• The velocities of the two touching gears are equal at C
• The gear rotation/velocity formula gives i
R
ω z1R1j = − 2
−ω z 2 R2 j ⇒ ω z 2 =
R1
Notice that we assume both gears rotate counterclockwise. The formula tells us that the second gear has a
negative angular velocity – this means that it is actually rotating clockwise. The animation at the top of
this section confirms that this indeed is the case.

Example 2: An ‘epicyclic’ gearbox is a special arrangement of gears that

has many applications. The sketch shows a simple example. The
gearbox can be driven in three different places: one drive shaft is
connected to the central sun gear (A); the other is attached to the ‘planet
carrier’, which is joined to the center of the ‘pinion gears’ B,C and D.
The outer gear (E – called the ‘ring gear’) can also be driven separately.

Epicyclic gearboxes are used in all automatic vehicle transmissions.

They are also very useful in ‘split power’ drives, where two motors need
to be connected together to drive a single axle. Hybrid vehicles, which
have both an electric motor and an internal combustion engine driving
the same axle, are one example. You can find a very nice description of
the Toyota Prius split power transmission here: the website includes a Flash animation that lets you change
the speeds of the motors in the system and visualize the motion of the gears.

The figure shows a schematic diagram illustrating the general

geometry and motion of the system. We have four rigid Planet
bodies: Planet gear
Carrier
• The central sun gear, radius RS , N S teeth, rotating at ω zP
angular velocity ω zS ω zS ω zPC
• The planet carrier, angular velocity ω zPC
RS
• The ring gear, radius RR , with N R teeth, angular Sun gear r
velocity ω zR
RR
• The planet gear, radius= r ( RR − RS ) / 2 , ω zR
N P ( N R − N S ) / 2 teeth, rotating at angular velocity
=
ω zP Ring gear

In any application, we are given the angular velocity of two of the drive shafts (any two of
ω zS , ω zPC , ω zR ), and must calculate the third. The planet gear is not connected to any drive shaft, so we
usually don’t care very much about its angular speed, but we will need to find ω zP to solve for the unknown
one of ω zS , ω zPC , ω zR .
20

This seems a terribly difficult problem, but it can be solved in a very simple way with a trick.

We start by solving a simpler version of the problem. Suppose

that the planet carrier is stationary ( ω zPC =0) and the sun gear
rotates with angular speed ω zS (see the animation). What is the
angular velocity of the ring gear?

The sun gear and the planet gear are just a standard gear pair so
we know that
R
ω zS RS = −ω zP r ⇒ ω zP = −ω zS S
r
The two touching points on the planet gear and the ring gear must
have the same velocity, so (using the rotating gear formula)
r
ω zP=rj ω zR RR j ⇒ ω= R ω zP
RR
We can eliminate ω zP to get the answer:
R
ω zR = −ω zS S
RR

Now let’s try the harder problem. The animation shows a general situation, where ω zS , ω zPC are both
nonzero. How can we find ω zR now?

This is difficult to analyze because the center of the planet gear is

not fixed, so it’s hard for us to visualize the motion, and the
standard gear formulas don’t work. But we can simplify the
problem by analyzing motion in a reference frame that rotates
with the planet carrier. For example, imagine attaching a
videocamera to the planet carrier – this camera would show the
planet carrier to be stationary, with the surrounding world rotating
in the opposite direction. The angular velocity of the planet
carrier would be subtracted from all the other angular velocities.
In this reference frame, we can use the result we just calculated:
(ω zR − ω zPC ) R
= − S
(ω zS − ω zPC ) RR

This result is general, and can be re-arranged to tell you the

angular velocities for any given combination of ω zS , ω zPC and
ω zR .
 21

6.4 Linear momentum, angular momentum and kinetic energy of rigid bodies
In this section, we determine how to calculate the angular momentum and kinetic energy of a rigid body,
and define two important quantities: (1) the center of mass of a rigid
body (which you already know), and (2) the Inertia tensor (matrix) of a
rigid body. m 4

To keep things simple, we won’t consider a general rigid body right m3

away. Instead, we will calculate the linear momentum, angular
momentum, and kinetic energy of a system of N particles that are d4
connected together by rigid, massless links. d3
m1
Definitions of inertial properties: For this system, we will define d2 r G d1
N
The total mass M = ∑ mi
k
i =1
r1
1 N m2 r 2
The position of the center of mass rG = ∑ mi ri i j
M i =1
The position vector of each mass relative to the center of mass di= ri − rG
dr
The velocity of the center of mass vG = G
dt

The mass moment of inertia about the center of mass (a tensor, which can be expressed as a matrix
if we choose a coordinate system and set di = dix i + diy j + diz k )

 I Gxx  diy2 + diz2 − dix diy − dix diz 

I Gxy I Gxz   
  N
IG =  I Gyx I Gyy I Gyz  = ∑ mi  − dix diy dix2 + diz2 − diy diz 
  i =1  
 −d d 2
 I Gzx I Gzy I Gzz  2
 ix iz − diy diz dix + diy 


The mass moment of inertia is sometimes also written in a more abstract but very compact way as
N
IG
= ∑ ( mi di
i =1
2
1 − mi di ⊗ di )
Here, 1 is the identity tensor, and di ⊗ di is a tensor with components dix dix , dix diy , dix diz , etc (the
symbol ⊗ is called the ‘diadic product’ of two vectors).

Formulas for linear and angular momentum and kinetic energy: We will show that:

The total linear momentum is p = MvG

The total angular momentum (about the origin) is h =

rG × MvG + IG ω

1 1
The total kinetic energy is =
T MvG ⋅ vG + ω ⋅ IG ω
2 2
22

These are actually general results that hold for all rigid bodies, as long as we use a more general definition
of M and IG .

Simplified formulas for two dimensions: For planar problems, diz = 0

(since all the masses are in the plane), and ω = ω z k . In this case, we can m2 m3
use
d2 d3
The total linear momentum is p = MvG
j r2
The total angular momentum (about the origin) is rG
h= rG × MvG + I Gzzω z k d1
The total kinetic energy is =
T
1 1
MvG ⋅ vG + I Gzzω z2
i
2 2
A r1
Here I Gzz is just the bottom diagonal term of the full inertia matrix (i.e. just a m1
single number)
N N
( )
=i 1 =i 1
∑ mi di
I=
Gzz
2
∑ mi dix2 + d=
2
iy

Example 1: A simple 3D assembly of masses is shown in the

figure.
k
mz
(1) Find the mass moment of inertia.

By symmetry, the COM is at the origin. The inertia tensor is Lz

mx Ly m
therefore y j

 (
 2 m L2 + m L2
y y z z ) 0 0 
 Lx
Lx
IG

 0 (
2 mx L2x + mz L2z ) 0


my Ly
mx
i
  mz Lz


0 0 ( 2
2 mx Lx + m y L y)
2 


(2) Assume that the COM is stationary (i.e. the assembly rotates about the origin). Find formulas for the
angular momentum and kinetic energy of the system, in terms of the angular velocity components
ωx ,ω y ,ωz

The formula gives the angular momentum

 23

 y y (
 2 m L2 + m L2
z z ) 0 0 
 ω x 
h=rG × MvG + IG ω =



0 (
2 mx L2x + mz L2z ) 0
 
 ω y 
 


0 0 (  )ω
2 mx L2x + m y L2y   z 

= 2 ( m y L2y + mz L2z ) ω x i + 2 ( mx L2x + mz L2z ) ω y j + 2 ( mx L2x + m y L2y ) ω z k

Note that h is a vector. Importantly, h is not generally parallel to the angular velocity vector, as this
example shows.

The kinetic energy is

ω x  
 2 m L2 + m L2
y y z z ( ) 0 0 
 ω x 
T
=
1
2
2 1
M vG + ω ⋅ IG ω
2
=
1  
ω y  ⋅ 
2  
0 (
2 mx L2x + mz L2z ) 0
 
 ω y 
 
ω z  

0 0 ( ) ω
2 mx L2x + m y L2y   z 

( ) ( ) (
T = m y L2y + mz L2z ω x2 + mx L2x + mz L2z ω y2 + mx L2x + m y L2y ω z2 )
These results help us understand what the formulas are predicting. Note, for example, that:
• The mass moment of inertia always has the form mass*length2. It has units of kg-m2
• The mass moment of inertia is a measure of how mass is distributed about the center of mass. An
object has a large inertia if the mass is far from the COM, and a small one if the mass is close to the
COM.
• The matrix-vector products in the formulas for h and T are
really just a way of calculating the velocity of each particle
in the system in a quick way. For example, suppose we
rotate our assembly of masses about the k axis with
angular velocity ω z (see the animation). Let’s calculate
the kinetic energy of the system, but without using the
rigid body formulas. The two blue masses are stationary,
so they have no KE. The red and green mass are both
moving in a circle about the origin. The circular motion
formula says their speed is V = Rω z We can calculate the
total kinetic energy using the usual formula
1
T= ∑ 2
miVi2
i
1 2 1
) ( )
2
=
2
2mx ( Lxω z ) + 2m y L yω z =
2
( mx L2x + m y L2y ω z2
This explains why the formula for I Gzz contains Lx and L y - the I Gzz component keeps track of how
much energy or momentum is produced by a rotation about the z axis. The energy and momentum depend
on the distances of the masses from the z axis – which of course depends on Lx and L y .

Finally, note that we can interpret the two terms in the formulas for momentum and KE as quantifying
(separately) the effects of translation and rotation
 24

= h rG × MvG + IG ω
Angular momentum
Translational + Rotational

1 1
= T MvG ⋅ vG + ω ⋅ IG ω
Kinetic energy is 2 2
Translational + Rotational
This helps explain why we can often idealize a system as a particle. If the rotational term is negligible, the
angular momentum and kinetic energy of a rigid body is just the same as that of a particle located at the
COM.

6.4.1 Deriving the linear momentum formula

N
By definition p = ∑ mi vi . We can re-write this as follows:
i =1
N N
drd N d
=p ∑=
mi vi ∑=
mi i
dt dt
∑
= mi ri
dt
( MrG ) MvG
=
=i 1 =i 1 =i 1
(we used the definition of the COM to get the last result)

6.4.2 Deriving the angular momentum formula

N
Start with the definition:=h ∑ ri × mi vi
i =1
i rG + di and recall the relative velocity formula v i − v G =ω × (ri − rG ) =ω × di . This means
Note that r=
we can re-write the angular momentum as
N
h
= ∑ (rG + di ) × mi ( vG + ω × di )
i =1
  N N  N  N
=  ∑ mi  rG × vG +  ∑ mi di  × vG + rG × ω ×  ∑ mi di  + ∑ mi di × ω × di
     
= i 1  = i 1  =  i 1=  i 1
Note that
N N N N
∑ mi di = ∑ mi ( ri − rG ) = ∑ mi ri − rG ∑ mi = MrG − MrG = 0
=i 1 =i 1 =i 1 =i 1

Finally, recall the dreaded triple cross product formula

a × b × c = (a ⋅ c)b − (a ⋅ b)c
This means that
di × ω × di = (di ⋅ di )ω − di (di ⋅ ω)
This gives us the result in compact notation directly
N N N
∑( )

∑
mi di ×=ω × di ( mi (di ⋅ di )ω − mi di (di ⋅ ω) ) +  mi di 2 1 − mi di ⊗ d=
∑ i  ⋅ ω IG ω
=i 1 =i 1 =  i 1 

where we used the compact formula for the mass moment of inertia about the COM:
 25

N
IG
= ∑ ( mi di
i =1
2
1 − mi di ⊗ di )
If you don’t like the compact formula, we can also get the matrix version by expand out the triple cross
product
ω x (dix2 + diy2 + diz2 ) 
   dix 
 
(di ⋅ di )ω − di (di =
  ( )
⋅ ω) ω y (dix2 + diy2 + diz2 )  −  diy  ω x dix + ω y diy + ω z diz
ω (d 2 + d 2 + d 2 )   diz 
 z ix iy iz 

 diy2 + diz2 − dix diy − dix diz 

  ω x 
  
= − dix diy dix + diz − diy diz  ω y 
2 2
 
 −d d 2 2  ω z 
 ix iz − diy diz dix + diy 


This again shows that

N
∑ mi di × ω × di =IG ω
i =1

Finally collecting terms gives the required answer

N  N  N
h = ∑ mi  rG × vG +  ∑ mi di  × vG + ∑ mi di × ω × di =rG × MvG + IG ω
   
= i 1 =  i 1=  i 1

6.4.3 Deriving the kinetic energy formula

N
1
=T ∑ 2 mi vi ⋅ vi
i =1
We can use vi − vG =ω × (ri − rG ) =ω × di
N
N
1 1
∑ 2 i i i ∑ 2mi ( vG + ω × di ) ⋅ ( vG + ω × di )
m
= v ⋅ v
=i 1 =i 1
N N  1 N
1
= ( vG ⋅ vG )∑ mi + vG ⋅ ω ×  ∑ mi di  + ∑ mi (ω × di ) ⋅ (ω × di )
  2
=i 1 2 =  i 1=  i 1
Recall that
N 
 ∑ mi di  = 0
 
 i =1 
and expand the dot product of two cross products using the formula
(a × b) ⋅ (c × d) = (a ⋅ c)(b ⋅ d) − (b ⋅ c)(a ⋅ d)
This shows that
2
(ω × di ) ⋅ (ω × di ) = ( ω ⋅ ω ) ( di ⋅ di ) − ( ω ⋅ di )
As for the derivation of the angular momentum, this can be rearranged using the compact notation as
26

N N N
∑( ) ω

∑( )
2
∑ mi (di ⋅ di )ω ⋅ ω − mi (di ⋅ ω)2 + ω ⋅ 
mi (ω × di ) ⋅ (ω=
× di )
 i 1
mi di 1 − mi di ⊗ di
i 1 =i 1 =
= ω ⋅ IG ω

Alternatively, we can get the matrix version of the formula as

2
(ω × di ) ⋅ (ω × di ) = ( ω ⋅ ω ) ( di ⋅ di ) − ( ω ⋅ di )
 2 2 2
0 0   d2 dix diz   
ω x   dix + diy + diz  ω x  ω x   ix
dix diy
 ωx
   ω  − ω  ⋅  d d  
= ω y  ⋅  0 dix2 + diy2 + diz2 0 diy2 
diy diz ω y 
   y   y   ix iy 
ω   2 2 2  ω z  ω z   2  ω z 
 z  0 0 dix + diy + diz  d d diy diz diz 
   ix iz 
 2 2
− dix diz   
ω x   diy + diz − dix diy
 ωx
   
= ω y  ⋅  − dix diy dix2 + diz2 − diy diz  ω y 
 
ω    
 z   − dix diz − diy diz dix2 + diy2  ω z 
 
= ω ⋅ IG ω

Finally, collecting all the terms gives the required answer

N
1 1 1
T
= ∑ 2 mi vi ⋅=
vi
2
MvG ⋅ vG + ω ⋅ IG ω
2
i =1

6.4.4 Calculating the center of mass and inertia of a general rigid body

It is not hard to extend the results for a system of N particles to a general rigid body. We simply regard the
body to be made up of an infinite number of vanishingly small particles, and take the limit of the sums as
the particle volume goes to zero. The sums all turn into integrals.

3D problems: For a body with mass density ρ (mass per unit volume) we have that

The total mass is M = ∫ ρ dV

V
1
M ∫
The position of the center of mass is rG = rρ dV d
V k rG
r
The mass moment of inertia about the center of mass is
 d y2 + d z2 − d x d y −d x d z  j
 
IG = ∫ ρ  − d x d y d x2 + d z2 − d y d z  dV i
 
V  2 2
−d x d z −d y d z d x + d y 
 
where d= r − rG
 27

For 2D problems: We know the COM must lie in the i,j plane and we don’t need to calculate the whole
matrix.

For a body with mass per unit area µ we can therefore use the formulas
rG d
j
The total mass is M = ∫ µ dA r
A
1 i
M ∫
The position of the center of mass is rG = rµ dA
A
1 2 2
The mass moment of inertia about the center of mass
= is I Gzz
M ∫ µ (d x + d y )dA
A
where d= r − rG

Example 1: To show how to use these, let’s calculate the total mass,
center of mass, and mass moment of inertia of a rectangular prism k
with faces perpendicular to the i, j, k axes:

First the total mass (sort of trivial)

cba j
=M ∫=
∫ ∫ ρ dxdydz ρ abc c
000
i a
b
Now the COM
cba
1 1 1 2 1 2 1 2  1
rG=
ρ abc ∫ ∫ ∫ ( xi + yj + zk ) ρ dxdydz=  a bci + a b cj + ab c k =
abc  2 2 2

 2
(ai + bj + ck )
000

And finally the mass moment of inertia

 28

( y − b / 2) 2 + ( z − c / 2) 2 −( x − a / 2)( z − c / 2) 
−( x − a / 2)( y − b / 2)
cba 
IG ∫ ∫ ∫  sym ( x − a / 2) 2 + ( z − c / 2) 2 −( y − b / 2)( z − c / 2)  ρ dxdydz
000
 
 sym sym ( x − a / 2)2 + ( y − b / 2)2 
 
1 3 1 3 
12 ab c + 12 abc 0 0 
 
 1 3 1 3 
ρ 0 a bc + abc 0
 12 12 
 
1 3 1
 0 0 a bc + ab3c 
 12 12 
b 2 + c 2 0 0 
M  
= 0 a2 + c2 0 
12  
 0 0 a 2 + b2 
 

Example 2: As a second example, let’s calculate the mass moment

of inertia of a cylinder with mass density ρ , length L and radius a.
We have to do the integrals with polar coordinates. For example, the k
inertia matrix is
a L/2
 d y2 + d z2 −d x d y −d x d z 
L /2 2π a   z
IG = ∫ ∫ ∫ − d x d y d x + d z − d y d z  ρ rdrdθ dz
Cylinder  2 2
  L/2
− L /2 0 0  2 r
−d x d z 2
−d y d z d x + d y  i θ
 
Now (in polar coordinates, and assuming that the COM is located at
the center of the cylinder)
= d x r= cos θ d y r sin
= θ dz z .
j
We can have Matlab do all the integrals for us:
29
30

Example 3: Let’s finish up with a 2D example. Find the mass, center of

mass, and out of plane mass moment of inertia of the triangle shown in
the figure. x=a(1-y/b)
b a (1− y / b )
1 b j
The total
= mass is M ∫=
∫ µ dxdy 2
µ ab
0 0
The position of the COM is
b a (1− y / b )
i
2 1 a
rG =
µ ab ∫ ∫ ( xi + yj) µ dxdy =
3
(ai + bj)
0 0
The 2D mass moment of inertia is
b a (1− y / b )
I Gzz
= ∫ ∫
 a 2 b 2
 ( x − ) + ( y − )  µ dxdy
 3 3 
=
abµ 2
36
2
a + b=
M 2
18
(
(a + b 2 ) )
0 0

This is all a big pain, and you may be contemplating a life of crime instead of an engineering career.
Fortunately, it is very rare to have to do these sorts of integrals in practice, because all the integrals for
common shapes have already been done. You can google most of them. The tables below give a short list
of all the objects we will encounter in this course.

Table of mass moment of inertia tensors for selected 3D objects

k b
b 2 + c 2 0 0 
Prism c M  
0 a2 + c2 0 
M = ρ abc j 12  
i  0 0 a 2 + b2 
a  

k
a L/2
1 + 3a 2 / L2 0 0 
Solid Cylinder ML 
2 
L/2 0 1 + 3a 2 / L2 0 
M = πρ a 2 L 12  
i  0 0 6a 2 / L2 
 

j
 31

h 1 + h 2 / (4a 2 )
k 0 0
Solid Cone  2 
3Ma 
π i h/4 0 1 + h 2 / (4a 2 ) 0 
M= ρ a2h j 20  
3  0 0 2
a  

k
Solid Sphere 1 0 0 
2 Ma 2 
4 i a 0 1 0 
M = πρ a3 5 
3 0 0 1 
j

k c
b 2 + c 2 0 0 
Solid Ellipsoid
M  
4 0 a2 + c2 0 
M = πρ abc
3
i a b
5  
 0 0 a 2 + b2 
j  

k
a
b L/2
 L2 + 3( a 2 + b 2 ) 0 0 
Hollow Cylinder M 
L/2  0 L2 + 3( a 2 + b 2 ) 0 
12  
M πρ (b 2 − a 2 ) L
=
i  0 0 6( a 2 + b 2 ) 
 
j
32

Table of mass moment of inertia about perpendicular axis for selected 2D objects

j
a M 2
Square i I Gzz
= (a + b2 )
12

a
j
M 2
Disk I Gzz = R
2
R i

Thin ring I Gzz = MR 2

R i

j
a
Hollow disk I Gzz
=
M 2
2
(
a + b2 )
i
b
j
M 2
Slender rod i I Gzz = L
12
L
 33

6.4.5 The Parallel Axis Theorem

In all the previous calculations we have been calculating the mass moment of inertia about the center of
mass. This is what always appears in the general angular momentum formula. But we sometimes want to
find the mass moment of inertia about a different point (not the COM). For example, if a body happens to
be rotating about a fixed point, we can sometimes find its angular momentum and kinetic energy more
quickly by first finding the mass moment of inertia about the fixed point, and then using special simpler
formulas the angular momentum and kinetic energy (see section 6.4.10). We also sometimes want to find
the combned mass moment of inertia of several bodies that are connected together. When we do this, we
usually find the center of mass of the collection of bodies, and then add up the mass moments of inertia of
all the separate bodies about the COM of the assembly (see section 6.4.6). To be able to do this, we need
to be able to calculate the mass moment of inertia of a body about and arbitrary point, i.e. not the COM of
the body.

The mass moment of inertia about an arbitrary point is defined exactly

the same way as the inertia about the COM, except that we use the
distances from our arbitrary point instead of the distance from the
COM. k
rO-rG
 ry2 + rz2 − rx ry − rx rz 
 
 2 2
− ry rz  dV r j
IO = ∫ ρ − rx ry rx + rz
  i O
V  2 2
 − rx rz − ry rz rx + ry 

i
It’s painful to have to re-do all these integrals, however. If we already know IG , the parallel axis theorem
lets us calculate IO directly. Define the vector d that points from G to O
d = rO − rG = d x i + d y j + d z k
Then for a 3D object with mass M
 d y2 + d z2 −d x d y −d x d z 
 
IO =IG + M  − d x d y d x2 + d z2 −d y d z 
 
 2 2
 − d x d z −d y d z dx + d y 


For 2D we have a simpler result j

rO-rG
I Ozz =I Gzz + M (d x2 + d y2 ) r
O i

Example: Let’s find the mass moment of inertia of a cylinder about k

axes that pass through one end of the cylinder (O), instead of the O
COM. a L/2
z
L/2
r
i θ

j
34

L L
Here, d = k ⇒ dx = d y = 0 dz =
2 2

The formula gives

 d y2 + d z2 −d x d y −d x d z 
 
IO =IG + M  − d x d y d x2 + d z2 −d y d z 
 
 −d d −d y d z 2 2
dx + d y 
 x z 
1 + 3a 2 / L2 0 0   L2 / 4 0 0
ML 2    
=  0 1 + 3a 2 / L2 0 +M  0 L2 / 4 0 
12    
 0 0 6a 2 / L2   0 0 0
   
2
 4 + 3a / L 2
0 0 
ML2  
=  0 4 + 3a 2 / L2 0 
12  
 0 0 6a 2 / L2 
 

Proof of the parallel axis theorem

Let rO be some arbitrary point in space, and let rG be the position of

the COM. Define d = rO − rG as the vector from the COM to O, as
shown in the figure. k
rO-rG
Then let r denote the position vector of an infinitesimal volume
b j
element in the rigid body relative to O, and let b denote the position O
vector of the same volume element relative to the COM G. r i
Then r= b − d .
i
We also know that (by definition)

∫ρ(r )
2
=IO 1 − r ⊗ r dV
V

∫ ρ ( b 1 − b ⊗ b ) dV
2
=IG
V

∫ ρbdV 0=
=
V
∫ ρ dV M V
2
We can make use of r = r ⋅ r and then substitute r= b − d into (1). Expand the dot and dyadic product
of b − d , note d is a constant and use the identities on the last line above, as follows
 35

∫ρ(r ) ∫ ρ ( (b − d) ⋅ (b − d) − (b − d) ⊗ (b − d) ) dV
2
IO
= 1 − r ⊗ r dV
=
V V

= ∫ ρ ( (b ⋅ b)1 − b ⊗ b ) dV + ∫ ρ dV ( (d ⋅ d)1 − d ⊗ d )
V V

     
−  2d ⋅ ∫ ρ bdV  1 +  ∫ ρ bdV  ⊗ d + d ⊗  ∫ ρ bdV 
 V  V  V 

∫ρ(b ) ( ) ( )
2 2 2
= 1 − b ⊗ b dV + M d 1 − d ⊗ d = I G + M d 1 − d ⊗ d
V

6.4.6 Calculating moments of inertia of complex shapes by summation

The most important application of the parallel axis theorem is in calculating the mass moment of inertia of
complicated objects (which don’t appear in our table) by adding together moments of inertia for simple
shapes. We can illustrate this with a couple of simple examples.

Example 1: Two spheres with radius 3a are connected by a rigid

k
cylinder with length 6a and radius a to create a dumbbell. All 3a
objects have the same mass density ρ . Calculate the total mass
moment of inertia of the dumbbell.

The general approach is 2a 3a

(1) Find the COM of the entire assembly
(2) Find the mass moment of inertia of each shape (the spheres i 3a
and the cylinder) about its own COM
(3) Use the parallel axis theorem to find the moment of inertia
of each shape about the combined COM j
(4) Add all the moments of inertia

For our problem

(1) We know the COM is at the origin by symmetry, so we
don’t need to calculate it
(2) The inertia matrices of each object (cylinder + sphere) about their own COM are:
( 3a )2 0 0 
 
sphere 2  4π 3  2
IG =  ( 3a ) ρ   0 ( 3a ) 0 
5 3  
 0 2
 0 ( 3a ) 
(6a )2 + 3a 2 0 0 
 
= cylinder
IG
1
12
6π a3 ρ 

0 ( ) (6a ) 2 + 3a 2 0 

 0 0 6a 2 
 
 36

(3) We don’t need to use the parallel axis theorem for the cylinder, because its COM is already at the
same place as the COM of the assembly. For the spheres, we need to move the COM a distance 6a
parallel to the k direction. This means that d=
x d= y 0, d= z 6a in our formula. Therefore
( 3a )2 0 0  0 0
  0 
2  4π 3   4π 3  
sphere
ICOM  ( 3a ) ρ   0 ( 3a )2 0  +
  3
( 3a ) ρ  0 0 0 
5 3   
2
 0
 0 ( 3a )  0 0 ( 6a )2 

(4) We can add everything up (note that there are two spheres). Its best to use Mupad. The answer is
929a 2 0 0 
1  
= ICOM = M 0 929a 2 0  M 42π a3 ρ
140  
 0 0 9514a 2 
 

a
a
a a
j j j
2a j 2a j

i i i
2a 2a
Example 2: Things are a lot simpler in 2D. The procedure is the same, but we only need to calculate I zz .
For example, to calculate the mass moment of inertia for a square 2ax2a plate with a hole with an axa
square cut out from the top corner we would use the following approach.

Start by calculating the total mass and the position of the COM. We can regard the cut-out section as a
square with negative density inside a larger 2ax2a square.

2 2
The total mass is therefore M= ρ ( 2a ) − ρ a= 3ρ a 2
1  2 2 3a 3a  5
The position of the COM is rG =  4a ρ (ai + aj) − a ρ ( i + j) = a (i + j)
M  2 2  6

The mass moment of inertia of the 2ax2a square and the axa square are
1 8 4
Large square=I Gzz 4 ρ a 2 (4a 2 +
= 4a 2 ) ρa (COM at a (i + j))
12 3
1 1 3
− ρ a 2 (a 2 + a 2 ) =
Small square I Gzz = − ρ a4 (COM at a (i + j))
12 6 2

We now use the parallel axis theorem to find the moment of inertia of each square about the combined
1 1 2 2
COM. For the large square:
= dx = a dy a . For the small square,
= dx = a dy a . The total
6 6 3 3
mass moment of inertia is therefore
 37

8 4  1 1  1 4 4  11 4 11
total
I Gzz = ρ a + 4a 2 ρ  a 2 + a 2  − ρ a 4 − ρ a 2  a 2 + a 2  = ρa = Ma 2
3  36 36  6  9 9  6 18

6.4.7 Rotating the inertia tensor

k
All the curious properties of spinning objects – a gyroscope; a
boomerang; the rattleback – are consequences of the fact that
mz
the mass moment of inertia of an object changes when it is
mx Lz
rotated. We can see this very easily by re-visiting our
assembly of masses. In the original calculation, the red, green Ly m
and blue masses were located on the i,j,k axes. We calculated y j
the inertia tensor to be
Lx
Lx
 (
 2 m L2 + m L2
y y z z ) 0 0 

my Ly
mx
i
Lz
IG



0 (
2 mx L2x + mz L2z ) 0



mz


0 0 ( )
2 mx L2x + m y L2y 

k
mz
Lz
Now suppose we rotate the assembly through 90 degrees
Lx m x j
about the k axis. The red masses now lie on the j axis, and the my Ly
green ones line up with the i axis. It is not hard to see that the
new mass moment of inertia is now
Ly
Lx my i
mx
(
2

mx L2x + mz L2z ) 0 0 
 mz Lz
IG 


0 (
2 m y L2y + mz L2z ) 0





0 0 ( )
2 mx L2x + m y L2y 

( I xx , I yy have switched positions)

This seems like a huge problem – if we needed to re-calculate the mass moment of inertia from scratch
every time a rigid body moves, analyzing rigid body motion would be nearly impossible.

Fortunately, we can derive a formula that tells us how the mass moment of inertia of a body changes when
it is rotated.
38

Rotation formula for moments of inertia: Consider the rectangular prism

0 k
shown in the figure. Let IG denote the mass moment of inertia with the 0
prism oriented so the faces are perpendicular to i,j,k (i.e. the inertia given in IG
the table in Sect 6.4.5). B
pB-pA
Suppose the body is then rotated by a tensor R. A j
The mass moment of inertia after rotation is given by i
0 T
IG = RIG R k
IG
Example: The prism shown in the figure is rotated by 45 degrees about the B
rB-rA
k axis. Calculate the mass moment of inertia after the rotation
A j
Start by calculating the rotation (use the formulas from 6.2.1) i
cos(45) − sin(45) 0 
R =  sin(45) cos(45) 0 
 0 0 1 

We know the inertia tensor of the prism before it is rotated is

b 2 + c 2 0 0 
M  
0
= IG 0 a2 + c2 0 
12  
 0 0 a 2 + b2 
 

We can use Matlab to do the tedious matrix multiplications

Note that the inertia tensor is no longer diagonal.

Rotation formula for 2D motion: Fortunately, 2D

is simple j j
Rotating a 2D object about the k axis does not a IGzz
change I Gzz i IGzz i
a
 39

Proof of the rotation formula: Consider a system of N particles. Suppose that before rotation, the
particles are at positions di0 relative to the COM. The initial inertia tensor is
N 2
 
= 0
IG ∑  mi di0 1 − mi di0 ⊗ di0 

i =1
Now rotate the system, so the particle s move to new positions di = Rdi0 . The new inertia tensor is
N
IG
= ∑ ( mi di
i =1
2
1 − mi di ⊗ di )
Recall that RRT = 1 and recall that a rotation R does not change lengths so di0 = di . Therefore
N 2
 
=IG ∑  mi di0 RRT − mi (Rdi0 ) ⊗ (Rdi0 ) 

i =1
It is easy to show (just write out the matrix products) (Rdi0 ) ⊗ (Rdi0 ) =R (di0 ⊗ di0 )RT , which shows that
N N
 02 0 T  02 
=IG ∑ m d
 i i

RR T
− mi R (d 0
i ⊗ d
= i ) R 

R

∑ 0
di0  RT RIG
 mi di 1 − mi di ⊗=

0 T
R
i 1 =i 1

6.4.8 Time derivative of the inertia tensor

When we analyze motion of a rigid body, we will need to calculate the time derivatives of the linear and
angular momentum. Linear momentum is no problem, but for angular momentum, we will need to know
how to differentiate IG with respect to time. There is a formula for this:
dIG
= WIG − IG W
dt
dR T
where W = R is the spin tensor (see sect 6.2.2)
dt

Proof:
0 T
• Start with IG = RIG R and take the time derivative
T
dI G dR 0 T 0 dR
= IG R + RIG
dt dt dt

dR T dRT dRT dR T
• Recall that RRT = 1⇒ R +R = 0⇒ − RT
= R =− RT W
dt dt dt dt
• Finally note that dR / dt = WR and therefore
dIG 0 T 0 T
=WRIG R + RIG R W= WIG − IG W
dt
40

6.4.9 Time derivative of angular momentum

To use the angular momentum conservation equation, we will need to know how to calculate the time
derivative of the angular momentum. When we do this for a 3D problem, we need to take into account that
the mass moment of inertia changes as the body rotates. We will prove the following formula:
dh
= rG × MaG + IG α + ω × ( IG ω )
dt
For 2D planar problems this can be simplified to:
dh
= rG × MaG + I Gzzα z k
dt

Proof: We start by taking the time derivative of the general definition of h

dh d
= ( rG × MvG + IG ω )
dt dt

We can go ahead and do the derivative with the product rule:

dh drG dv dI dω
= × MvG + rG × M G + G ω + IG
dt dt dt dt dt
We can simplify this by noting that drG / dt = vG and of course the cross product of vG with itself is zero.
We can also use the definition of angular acceleration: dω / dt = α . This gives
dh dI
= rG × MaG + G ω + IG α
dt dt
Finally, substitute for dI G / dt from the formula in the previous section, and recall that Wu= ω × u for all
vectors u, and that a vector crossed with itself is zero to see that:
dh
= rG × MaG + ( WIG − IG W ) ω + IG α
dt
= rG × MaG + IG α + ω × IG ω − IG ω × ω
= rG × MaG + IG α + ω × IG ω

6.4.10 Special equations for angular momentum and KE of bodies that rotate about a
stationary point

We often want to predict the motion of a system that rotates about a fixed pivot – a pendulum is a simple
example. These problems can be be solved using a useful short-cut for the angular momentum or KE of a
body rotating about a fixed point. The short-cut will give the same answer as the general formulas.

For an object that rotates about a fixed pivot at the origin:

The total angular momentum (about the origin) is h = IO ω k

rO-rG
1
The total kinetic energy is=
T ω ⋅ IO ω O
2 r j
i

i
 41

Here IO is the mass moment of inertia about O (calculated, eg, using the parallel axis theorem). Note that
the special formulas do not include the term involving the velocity of the COM – that’s been automatically
included by using IO instead of IG .

For 2D rotation about a fixed point at the origin we can simplify these to

The total angular momentum (about the origin) is j

h = I Ozzω z k rO-rG
1 r
The total kinetic energy is T = I Ozzω z2 O i
2

Proof: It is straightforward to show these formulas. Let’s show the two dimensional version of the kinetic
energy formulas as an example. For fixed axis rotation, we can use the rigid body formulas to calculate the
velocity of the center of mass (O is stationary and at the origin)
vG =ω × rG =ω z k × rG
The general formula for kinetic energy can therefore be re-written as
1 1 2 1 1
= T MvG ⋅ vG + I Gzzω =
z M ω z2 (k × rG ) ⋅ (k × rG ) + I Gzzω z2
2 2 2 2
=
1
2( 2
)
M rG + I Gzz ω z2 =
1
2
I Ozzω z2
The other formulas can be proved with the same method – we simply express the velocity or acceleration of
the COM in the general formulas in terms of angular velocity and acceleration, and notice that we can re-
arrange the result in terms of the mass moment of inertia about O.

The 3D proof is the same. Start with the general formula

1 1
=T MvG ⋅ vG + ω ⋅ IG ω
2 2

and use the kinematics formula to find v G (noting that O is stationary and at the origin)
v G= ω × rG
1 1
T
= M ( ω × rG ) ⋅ ( ω × rG ) + ω ⋅ IG ω
2 2
Remember the vector formula (a × b) ⋅ (c × d) = (a ⋅ c)(b ⋅ d) − (b ⋅ c)(a ⋅ d) , which shows that
2
(ω × rG ) ⋅ (ω × rG ) = ( ω ⋅ ω ) ( rG ⋅ rG ) − ( ω ⋅ rG )
We can re-write the kinetic energy as
= T
1
2
( ) 1
M (rG ⋅ rG )ω ⋅ ω − (rG ⋅ ω) 2 + ω ⋅ IG ω
2
1 2 1 1
= ω ⋅ M  rG 1 − rG ⊗ rG  ω + ω ⋅ IG ω = ω ⋅ IO ω
2 
 
 2 2
using the parallel axis theorem.

Another way to prove the result is just to calculate the KE of the body from scratch, by summing the KE of
the infinitesimal particles in the rigid body, and noting that they are all in circular motion about O.
42

The proof of the angular momentum formula is just the same – start with the general formula for h and then
simplify it using v G= ω × rG . You might like to try this as an exercise.

Example: In the planetary gear system shown, the sun gear has
Planet
radius RS and mass m , the ring gear has radius 3RS , while the Planet gear
planet gear has mass m and the planet carrier has mass m/2 . Carrier
ω zP
The sun gear rotates with angular speed ω zS and the ring gear ω zS ω zPC
is stationary.
RS
Sun gear r

Find a formula for the total angular momentum of the assembly RR

ω zR
about the center of the sun gear, in terms of ω zS , RS and m.
Treat the gears as disks, the planet carrier as a 1D rod and
Ring gear
assume there’s only one planet gear as shown to keep things
simple; this would be a rather unusual gear system but adding more gears just makes the problem tedious
without illustrating any new concepts...

The 2D formula for angular momentum of a rigid body (about the origin) is

h=rG × mv G + I Gzzω z k

where rG is the position vector of the COM of the body relative to the origin.

We need to find the angular speed of all the moving parts: using the gear formulas
ω zP − ω zPC R ω zR − ω zPC R
=
− S =
− S RS + 2 RP
RR =
ω zS − ω zPC RP ω zS − ω zPC RR
we see that
0 − ω zPC R R  R 
− S ⇒ ω zS S =
= ω zPC 1 + S 
ω zS − ω zPC RR RR  RR 
1  1 1
⇒ ω zS= ω zPC 1 +  ⇒ ω zPC
= ω zS
3  3 4
and
ω zP − ω zPC 2 RS ω − ω zPC
=
− ⇒ zP −1
=
ω zS − ω zPC RR − RS ω zS − ω zPC
1
⇒ ω zP =
ω zPC − (ω zS − ω zPC ) =
− ω zS
2

The COM of the planet carrier is half way along its length; its COM is in circular motion with
speed V = ω zPC RS

Similarly the COM of the planet gear is in circular motion with speed V = ω zPC 2 RS
 43

Now we can add up all the angular momenta:

1
1. Sun h S = mR 2ω zS k
2
1 1 1 1 1 1
2. Planet carrier h PC =
RS i × ( mRS ωZs ) j + m(2 RS ) 2 ω zS k = mRS2ω zS k
2 4 12 2 4 6
Notice that the planet carrier rotates about the center of the sun. So, if we want, we could also use
the special formula for angular momentum of an object rotating about a fixed point
h PC = I Ozzω zPC k
where I Ozz is the mass moment of inertia of the planet carrier about the fixed point, which must be
calculated using the parallel axis theorem
1 m m 2
I Ozz =I Gzz + Md 2 = (2 Rs ) 2 + RS2 = mRS2
12 2 2 3
(where we noted that the length of the bar is 2 RS ). We know ω zPC = ω zS / 4 so
1
= h PC I=Ozzω zPC k mRS2ω zS k
6
as before.

1 1 1 3
3. Planet gear h P =
2 RS i × (m2 RS ωZs ) j + m( RS ) 2 (− ω zS )k =mRs2ω zs k
4 2 2 4
Note that we can’t use the special formula for rotation about a fixed point for the planet gear,
because although there is a fixed point on the planet gear (where it touches the ring), we were asked
to find the angular momentum about the center of the sun. This is not a fixed point on the planet
gear.

17
Sum everything h = mRS2ω zS k
12
44

6.5 Rotational forces – review of moments exerted by forces and torques

You can find a detailed discussion of forces and moments, with lots of examples, in Section 2 of these
notes. Moments and torques don’t come up very often in particle dynamics, but play a very important role
in rigid body dynamics. We therefore review the most important concepts related to torques and moments
here.

You need to remember, and understand, these ideas:

(1) A moment is a generalized force that causes an object to rotate (see section 2).
(2) A force can exert a moment on a rigid body. The moment of a force (about the origin) is defined
as
M= r × F
(3) In general, a force causes a rigid body to accelerate, and will also induce an angular acceleration (so
it influences both translational and rotational motion).
(4) A ‘torque’ or ‘pure moment’ is a special kind of generalized force that causes an object to rotate,
but has no effect on its translational motion. As an example, a motor shaft (eg the bit on a power-
driven screwdriver!) will exert a torque on the object connected to it.
(5) A torque or pure moment is a vector quantity – it has magnitude and direction. The direction
indicates the axis associated with its rotational force (following the right hand screw convention);
the magnitude represents the intensity of the rotational force. The magnitude of a torque has units of
Newton Meters. A moment is often denoted by the symbols shown in the figure

6.5.1 Rate of work done by a torque or moment: If a torque Q = Qx i + Qy j + Qz k acts on an object that
rotates with angular velocity ω , the rate of work done on the object by Q is
P = Q ⋅ ω = Qxω x + Qyω y + Qzω z

6.5.2 Torsional springs

A solid rod is a good example of a torsional spring. You could

take hold of the ends of the rod and twist them, causing one
end to rotate relative to the other. To do this, you would apply
a moment or a couple to each end of the rod, with direction
parallel to the axis of the rod. The angle of twist increases
with the moment. Various torsion spring designs used in
practice are shown in the picture – the image is from

http://www.mollificio.lombardo.molle.com/springs/torsion_spr
ings.html
 45

More generally, a torsional spring resists rotation, by exerting equal and opposite moments on objects
connected to its ends. For a linear spring the moment is proportional to the angle of rotation applied to the
spring.

The figure shows a formal free body diagram for two objects
connected by a torsional spring. If object A is held fixed, and object
B is rotated through an angle θ about an axis parallel to a unit θ
vector n, then the spring exerts a moment
Q = −κθ n j
on object B where κ is the torsional stiffness of the spring.
B n
A
Torsional stiffness has units of Nm/radian. i -Q
The potential energy of the moments exerted by the spring can be
determined by computing the work done to twist the spring through Q
an angle θ .
1. The work done by a moment Q due to twisting through a
very small angle dθ about an axis parallel to a vector n is B
dW= Q ⋅ dθ n Q
A
2. The potential energy is the negative of the total work done
by M, i.e.
-Q
θ θ θ
1 2
V= − ∫ ( −κθ n ) ⋅ dθ n =
− ∫ Q ⋅ dθ n = ∫ κθ dθ = κθ
2
0 0 0

A potential energy cannot usually be defined for most concentrated moments, because rotational motion is
itself path dependent (the orientation of an object that is given two successive rotations depends on the
order in which the rotations are applied).

6.6 Dynamics of rigid bodies

We predict the position and velocity of a particle by integrating F=ma.
For a rigid body, we need to predict both its position and orientation. v1 Q(2) F(3)
F(1)
We use the following equations to do this.
v2
(i ) Q(1) vG
The figure shows a rigid body subjected to several forces F and
(i )
torques (pure moments) Q . During a representative time interval k rG
t0 < t < t1 the forces exert a linear impulse ℑ and angular impulse A, v3
and do total work on the rigid body ∆W . F(2)
j
The body has total mass M and mass moment of inertia IG about the
center of mass.
i

Let rG , vG , aG denote the position, velocity and acceleration of the center of mass, and let ω, α denote the
angular velocity and acceleration.
46

The linear and angular momentum (about the origin) of the rigid body follow as p = MvG ,
1 2 1
h = MrG × vG + IG ω , and its kinetic energy =
is T M vG + ω ⋅ IG ω .
2 2

The equations of motion are then

(i )
Force-acceleration relation: ∑F = MaG
i

(i )
Moment – angular velocity/acceleration relation ∑ ri × F + ∑ Q( j=) MrG × aG + IG α + ω × [ IG ω ]
i j

(i ) dp
Force-momentum and impulse-momentum relation: ∑F =
dt
ℑ= p1 − p0
i

(i ) dh
Moment – angular momentum relation: ∑ ri × F + ∑ Q( j ) =dt A =h1 − h0
i j

(i ) dT
Power – work – kinetic energy relation ∑F ⋅ v (i ) + ∑ Q( j ) ⋅ ω = dt
∆W = T1 − T0
i j

For 2D planar motion we can use the simplified formulas

(i )
∑F = MaG
i
(i )
∑ ri × F + ∑ Q( j ) = MrG × aG + I Gzzα z k
i j

Derivations: It is possible to obtain the equations of motion for a rigid

body from Newton’s laws for a particle – the basic idea is to assume that F(3) m4
a rigid body consists of an infinite number of particles connected by
rigid massless links – but this isn’t really a rigorous proof, because we m3
have to assume that the links are two-force members, and there is no way
to prove that this is a realistic description of matter. Another viewpoint d4
is to accept conservation of linear momentum and angular momentum as d3
two separate physical laws (the linear momentum is just Newton’s law, m1
and the angular momentum equation is sometimes referred to as Euler’s F(2) d2 r G d1
law). We can then ‘prove’ that a rigid body can be represented as a
bunch of particles connected by two force members. We’ll show the k
first approach here. r1
m2 r 2 F(1)
The figure shows a system of particles connected by rigid massless links. i j
 47

The length of the link between the ith and jth particle will be denoted by Lij We assume that all the links
are two-force members.

The particles are subjected to a set of external forces F (i ) . We denote the magnitude of the force in the
member connecting the ith and jth particle by Rij (by convention a positive Rij represents an attractive
force between the particles). Note that the Rij = R ji because the two particles exert equal and opposite
forces on each other. The vector valued force exerted on the ith particle by the jth follows as
r j − ri
R ij = Rij
Lij
(to see this note that (r j − ri ) / Lij is a unit vector from the ith to the jth particles)

We can start the derivation with the force-linear momentum relation for a single particle. For example, for
the ith particle (see section 4 of the notes)
(r j − ri ) d
Fi + ∑Rij
Lij
=
dt
mi vi
j ≠i
Sum this over all particles

(r j − ri ) d
∑ Fi + ∑∑ Rij Lij
∑ =
dt
mi ri
i i j ≠i i
But we know that ∑ mi vi = MvG , and since
= Rij ji , Lij L ji the second term on the left hand side is
R=
i
zero. Therefore
dv G dp
∑=
Fi M =
dt
M
=aG
dt
i
Since this is independent of the number of particles, it must also apply to a rigid body. This shows that the
force-momentum and force-acceleration for a rigid body can be derived from Newton’s law for a particle.

We can derive the angular momentum relation for a rigid body using the same idea. For one particle we
have the angular momentum equation
(r j − ri ) r j dh i
ri × F (i ) + ri × Rij∑ Lij
∑
= ri × F (i ) + ri × Rij
Lij
=
dt
j ≠i j ≠i
where we have noted that ri × ri =
0 . We can sum this over all the particles
ri × r j d
∑
ri × F (i ) + ∑∑
Rij
Lij
= hi
dt i
∑
i i j ≠i
The second term here is zero, because ri × r j =−r j × ri=
and Rij R= ji Lij L ji (just write out the sum
term by term for some finite number of particles – eg two – if you don’t see this). The term on the right
hand side is clearly just the total angular momentum of the system. If we replace some subset of the forces
with a statically equivalent torque and force, we obtain the moment-angular momentum equation.
48

6.7 Summary of equations of motion for rigid bodies

In this section, we collect together all the important formulas from the preceding sections, and summarize
the equations that we use to analyze motion of a rigid body.

We consider motion of a rigid body that has mass density ρ during some time interval t0 < t < t1 , and
define the following quantities:

6.7.1 Forces, torques, impulse, work, power

v1 Q(2) F(3)
F (1)
The total force acting on the body ∑F (i )

i v2
Q(1) vG
The total linear impulse exerted by forces during the time interval
t1
k rG
∫ ∑F
(i )
ℑ= (t )dt v3
t0 i
F(2)
The total moment (including torques) acting on the body
(i ) j
∑ ri × F + ∑ Q( j )
i
i j
The tot al angular impulse exerted on the body during the time
t1  
∫∑
interval A =  ri × F (i ) (t ) + Q( j ) dt
 
∑
t0  i j 
The rate of work done by forces and torques acting on the body =
P ∑ F (i ) ⋅ v i + ∑ Q ( j ) ⋅ ω
i j
t1
The total work done by forces and torques on the body during the time interval W = P (t )dt ∫
t0

6.7.2 Inertial properties

The total mass is M = ∫ ρ dV

V
1
M ∫
The position of the center of mass is rG = rρ dV
d
V
k rG
The mass moment of inertia about the center of mass r
 d y2 + d z2 − d x d y −d x d z 
  j
IG = ∫ ρ  − d x d y d x2 + d z2 − d y d z  dV i
 
V  2 2
 − d x d z −d y d z d x + d y 

where d= r − rG
 49

For a 2D body with mass per unit area µ we use

The total mass is M = ∫ µ dA rG d

j
A r
1
M ∫
The position of the center of mass is rG = rµ dA
A
i
1 2 2
The mass moment of inertia about the center of mass
= is I Gzz
M ∫ µ (d x + d y )dA
A
where d= r − rG

6.7.3 Describing motion

The rotation tensor (matrix) maps the vector

connecting two points in a solid before it moves k k
to its position after motion
rB − rA= R (p B − p A ) B B
pB-pA rB-rA
The spin tensor is related to R by
dR T dR A j A j
= W = R WR
dt dt i i
Rotation through an angle θ about an axis
parallel to a unit vector
n = nx i + n y j + nz k is
 cos θ + (1 − cos θ )nx2 (1 − cos θ )nx n y − sin θ nz (1 − cos θ )nx nz + sin θ n y 
 
R = (1 − cos θ )nx n y + sin θ nz cos θ + (1 − cos θ )n 2y (1 − cos θ )n y nz − sin θ nx 
 
 (1 − cos θ )n n − sin θ n (1 − cos θ )n n + sin θ n 2 
cos θ + (1 − cos θ )nz 
 x z y y z x 
The angular velocity vector ω = ω x i + ω y j + ω z k is related to W by
 0 −ω z ωy 
 
=W  ωz 0 −ω x 
 −ω ωx 0 
 y

dω
The angular acceleration vector is α =
dt

The velocities of two points A and B in a rotating rigid body are related by
v B − v A =ω × ( rB − rA )
50

The accelerations of A and B are related by

aB − a A =α × (rB − rA ) + ω × ( v B − v A ) =α × (rB − rA ) + ω × [ω × (rB − rA )]

6.7.4 Momentum and Energy

The total linear momentum is p = MvG

The angular momentum (about the origin) is h =

rG × MvG + IG ω

1 1
The total kinetic energy is =
T MvG ⋅ vG + ω ⋅ IG ω
2 2

For 2D planar problems, we know ω = ω z k . In this case, we can use

The total linear momentum is p = MvG
The total angular momentum (about the origin) is h =rG × MvG + I Gzzω z k
1 2 1
The total kinetic energy
= is T M vG + I Gzzω z2
2 2

6.7.5 Conservation laws

(i ) dp
Linear momentum ∑F =
dt
ℑ= p1 − p0
i

(i ) dh
Angular momentum ∑ ri × F + ∑ Q( j ) =dt A =h1 − h0
i j

(i ) dT
Work-Power - Kinetic Energy relation ∑F ⋅ v (i ) + ∑ Q( j ) ⋅ ω = dt
∆W = T1 − T0
i j

d
Energy equation for a conservative system (T + V ) =0 T0 + V0 =T1 + V1
dt

6.7.6 Linear and angular momentum equations in terms of accelerations

The linear and angular momentum conservation equations can also be expressed in terms of accelerations,
angular accelerations, and angular velocities. The results are
 51

(i )
∑F = MaG
i
(i )
∑ ri × F + ∑ Q( j=) MrG × aG + IG α + ω × [ IG ω ]
i j

For 2D planar motion we can use the simplified formulas

(i )
∑F = MaG
i
(i )
∑ ri × F + ∑ Q( j ) = MrG × aG + I Gzzα z k
i j

6.7.7 Special equations for analyzing bodies that rotate about a stationary point

We often want to predict the motion of a system that rotates about a fixed pivot – a pendulum is a simple
example. These problems can be solved using the equations in 6.6.5 and 6.6.6, but can also be solved using
a useful short-cut.

For an object that rotates about a fixed pivot at the origin:

The total angular momentum (about the origin) is h = IO ω k

rO-rG
1
The total kinetic energy is=
T ω ⋅ IO ω O
2 r j
The equation of rotational motion is i
(i )
∑ ri × F + ∑Q ( j)
= IO α + ω × [IO ω ]
i j i
Here IO is the mass moment of inertia about O (calculated, eg, using
the parallel axis theorem)

For 2D rotation about a fixed point at the origin we can simplify these to

The total angular momentum (about the origin) is j

h = I Ozzω z k rO-rG
1 r
The total kinetic energy is T = I Ozzω z2 O i
2
The equation of rotational motion is
(i )
∑ ri × F + ∑ Q( j ) =
I Ozzα z k
i j

Proof: It is straightforward to show these formulas. Let’s show the two dimensional version of the kinetic
energy formulas as an example. For fixed axis rotation, we can use the rigid body formulas to calculate the
velocity of the center of mass (O is stationary and at the origin)
52

vG =ω × rG =ω z k × rG
The general formula for kinetic energy can therefore be re-written as
1 1 2 1 1
= T MvG ⋅ vG + I Gzzω =
z M ω z2 (k × rG ) ⋅ (k × rG ) + I Gzzω z2
2 2 2 2
= (
1
2
2
)
M rG + I Gzz ω z2 =
1
2
I Ozzω z2
The other formulas can be proved with the same method – we simply express the velocity or acceleration of
the COM in the general formulas in terms of angular velocity and acceleration, and notice that we can re-
arrange the result in terms of the mass moment of inertia about O.

6.8 Examples of solutions to problems involving motion of rigid bodies

The best way to learn how to use the equations in section 6.6 is just to work through a series of examples.

6.8.1 Solutions to 2D problems

Example 1: A solid of revolution (eg a cylinder or sphere) with mass

M and mass moment of inertia about its COM I Gzz is released from
rest at the top of a ramp. It rolls without slip. Calculate its velocity at
the bottom of the ramp.

• The system is conservative, so we can solve the problem using

energy conservation. The energy equation tells us that the sum
of kinetic and potential energy of the cylinder is constant:
T0 + V0 =T1 + V1
• We can take the datum for potential energy to be the position of the COM at the bottom of the
ramp. The initial potential energy is therefore V0 = Mgh ; the final potential energy is zero.
• The initial kinetic energy is zero, because the cylinder is stationary. The final kinetic energy is
1 1
= T Mvx2 + I Gzzω z2 .
2 2
1 1
• The energy equation gives T0 + V0 = T1 + V1 ⇒ Mgh = Mvx2 + I Gzzω z2
2 2

• Finally, since the cylinder rolls without slip, we know that vx = − Rω z .

Hence
I
= Mvx2 + Gzz vx2
2 Mgh
R2
2 gh
⇒ vx =
1 + I Gzz / ( MR 2 )
 53

This formula predicts that an object with a smaller inertia I Gzz will move faster than an object with a large
inertia. A sphere rolls down the ramp more quickly than a cylinder, for example, and a solid cylinder rolls
more quickly than a ring.

Example 2: For the problem treated in the preceding section, calculate

the critical value of friction coefficient necessary to prevent slip at the
contact.

If we want to learn about forces, we have to use the linear and angular
momentum equations. This problem can be solved with the 2D
formulas in terms of accelerations:
(i )
∑F = MaG
i
(i )
∑ ri × F + ∑ Q( j ) = MrG × aG + I Gzzα z k
i j

• The figure shows a free body diagram for the cylinder (or sphere)
• We know that the COM is always a constant height above the ramp, so R
the acceleration must be parallel to i. The linear momentum equation
Mg
gives
( Mg sin α − T )i + ( N − Mg cos α ) j =
MaGx i µN N
• We can use the angular momentum equation – it is convenient to take
moments about the contact point C. (There are no torques in this problem).
(i )
∑ ri × F + ∑ Q( j ) = MrG × aG + I Gzzα z k
i j
⇒ − RMg sin α k = Rj × aGx i + I Gzzα z k = − MRaGx k + I Gzzα z k
• Finally, we can use the rolling wheel formula for accelerations aGx = − Rα z .
• The preceding results give:
a
− RMg sin α = − MRaGx − I Gzz Gx
R
MgR sin α g sin α
= ⇒ aGx =
MR + ( I Gzz / R ) 1 + I Gzz / ( MR 2 )
• Finally, substituting back into the i components of (1):
= T Mg sin α − MaGx
I Gzz / ( MR 2 )
Mg sin α
Mg sin α −
= = Mg sin α
1 + I Gzz / ( MR 2 ) 1 + I Gzz / ( MR 2 )
• The j component of (1) gives N = Mg cos α
I Gzz / ( MR 2 )
• For no slip T ≤ µ N ⇒ µ ≥ tan α
1 + I Gzz / ( MR 2 )
54

The formula shows that objects with large values of I Gzz / MR 2 are more likely to slip. If the inertia is
very small, slip will never occur. A ring will slip on a lower slope than a cylinder, which will slip on a
lower slope than a sphere.

Example 3: A vertical mast can be idealized as a slender rod with length L

and mass M, which is held in an inverted position by a torsional spring with
stiffness κ at its base. Find the equation of motion for the angle θ in the M j
figure, and hence determine the natural frequency of vibration of the mast. θ
This is a conservative system. Also, the mast rotates about a fixed point. We
can analyze the problem using energy methods, and use the special formulas L
for rotation about a fixed point. i
κ O
• The kinetic energy formula for planar motion is
1
T = I Ozzω z2
2
• For planar motion we know that
dθ
ωz =
dt
• We can use the parallel axis theorem to calculate the mass moment of inertia of a rod about one
end:
2
1 L 1
I Ozz =I Gzz + Md = ML2 + M   = ML2
2
12 2 3
• Gravity and the torsional spring both contribute to the total potential energy of the system. The
total potential energy is
L 1
= V Mg cos θ + κθ 2
2 2
• Energy conservation means that
d
T +V= const ⇒ (T + V= ) 0
dt
d 1 1 
⇒  I Ozzω z2 + MgL cos θ + κθ 2  = 0
dt  2 2 
dωz dθ dθ
⇒ I Ozz ω z − MgL sin θ + κθ = 0
dt dt dt
dθ
• Recall that ω z = so
dt
d 2θ
⇒ I Ozz − MgL sin θ + κθ = 0
dt 2
• We assume that θ is small enough that sin θ ≈ θ , so
I Ozz d 2θ
⇒ +θ =0
κ − MgL dt 2
• This is a standard ‘Case I’ undamped vibration EOM, so we can just read off the natural
frequency
 55

κ − MgL 3(κ − MgL)

=ωn =
I Ozz ML2

Example 4: A thin uniform disk of radius R, mass m and mass moment of ω0

inertia mR 2 / 2 is placed on the ground with a positive velocity v0 in the
horizontal direction, and a counterclockwise rotational velocity (a m v0
backspin) ω0 . The contact between the disk and the ground has friction
coefficient µ . The disk initially slips on the ground, and for a suitable R
range of values of ω0 and v0 its direction of motion may reverse. The
goal of this problem is to calculate the conditions where this reversal will
occur.

General discussion of slipping contacts: Solving problems with sliding

at a contact is always tricky, because we have to draw the friction forces in
the correct direction. Before tackling the example, we will summarize the ωz
general rules. We will consider a wheel as an example, but the rules apply
α z

to contact between any object and a stationary surface. The figure shows R
a wheel that spins with angular velocity ω = ω z k while the center moves j
O
with speed vO = vox i . The direction of the friction force is determined
by the direction of motion of the point on the wheel that instantaneously
touches the ground, which can be calculated from the formula i C
=vC (vOx + ω z R )i
Friction always acts to try to bring point C to rest – if C is moving to the right, friction acts to the left; if C
is moving to the left, friction acts to the right.

There are three possible cases:

Forward slip: vOx + ω z R > 0 Point C moves in the positive i direction over the ground

Slip occurs at the contact, R

We have to use the friction law T = µ N
Point C is moving to the right, so friction must act to the left O
j C
T=µ N
i
N
56

Pure rolling vOx + ω z R =

0 . Point C is stationary.

No slip occurs at the contact.

In this case T < µ N R
We can draw the friction force in either direction at the contact (if O
we choose the wrong direction, our calculations will just tell us that T
is negative). It is usually convenient to choose T to act in the positive
j C
i direction, but this is not necessary. T
i N

Reverse slip: vOx + ω z R < 0 Point C moves in the negative i direction over the ground

Slip occurs at the contact,

R
We have to use the friction law T = µ N O
Point C is moving to the left, so friction must act to the right
j C
T=µ N
N
i
Now we return to the example.

4.1 Draw a free body diagram showing the forces acting on the disk just after it hits the ground.

We are given that vx 0 and ω z 0 are both positive so we have

vOx + ω z R > 0 . This is forward slip, and we use the corresponding FBD.
mg R

µN N
 57

4.2 Hence, find formulas for the initial acceleration a and angular acceleration α for the disk, in terms of
g , R and µ . Note that the contact point is slipping.

The equations of motion are

(i ) (i )
∑F = MaG ∑ ri × F + ∑ Q( j ) = MrG × aG + I Gzzα z k
i i j

− µ Ni + ( N − mg ) j =
ma x i
− µ NRk =
I Gzzα z k

Solving these and using I Gzz = mR 2 / 2 :

1
N= mg µN =
− ma x mR 2α z =
− µ NR
2
⇒ ax =−µ g αz = −2 µ g / R

4.3 Find formulas for the velocity and angular velocity of the disk, during the period while the contact point
is still slipping.

The acceleration and angular acceleration are constant, so we can use the constant acceleration
formulas:

v0 − µ gt
vx = ωz =
ω0 − 2µ gt / R

4.4 Find a formula for the time at which the disk will reverse its direction of motion.

Velocity is reversed where v=0. From the previous part, v = v0 − µ gt ⇒ t = v0 / µ g at the reversal.

4.5 Find a formula for the time at which the disk begins to roll on the ground without slip. Hence, show
that the disk will reverse its direction only if v0 < ω0 R / 2

Rolling without sliding starts when vxO = −ω z R . We have that

ωz =
ω0 − 2µ gt / R v0 − µ gt
vx =
− (ω0 R − 2 µ gt )
⇒ v0 − µ gt = −ω z R
when vx =
⇒ t= (v0 + ω0 R ) / 3µ g

The reversal will only occur if rolling without slip occurs after the reversal of velocity. This means
58

(v0 + ω0 R ) / 3µ g > v0 / µ g ⇒ v0 < ω0 R / 2

Example 5: The ‘Sweet Spot’ on a softball or baseball bat, or tennis or squash racket is a point that
minimizes the reaction forces acting on the athlete’s hand when the ball is struck. In fact, any rigid body
has a sweet spot – the magic point is called the ‘center of percussion’ of a rigid body.

For baseball and softball bats in particular, there is a standard ASTM test that can be used to measure the
position of the sweet spot. The test works like this: the bat is suspended from the knob on handle, so it
swings like a pendulum. The period of vibration τ of the swinging bat is then measured. ASTM say that
the center of percussion is then a distance
τ 2g
d=
4π 2
from the end of the handle. Why does this work? It seems that this test has nothing
whatever to do with a ball hitting the bat! j
We will solve this problem in two parts. First, we will calculate a formula for the
period of vibration in the ASTM test. Then we will calculate the position of the O i
center of percussion. We will see that the ASTM test does indeed make the correct
prediction.
L
We can calculate the period using the energy method. The figure shows the ASTM
pendulum test. We assume that
θ
• The bat has a mass moment of inertia about its COM I Gzz
• The COM is a distance L from O

The bat pivots about O, so we can use the fixed axis rotation formula for the kinetic
energy
2
1  dθ 
T= I Ozz  
2  dt 
2
Ozz I Gzz + ML (using the parallel axis theorem).
Here I=

The potential energy is V = − MgL cos θ .

Energy conservation gives

d
+ V constant ⇒
T= (T =
+V ) 0
dt
d 2θ dθ dθ
⇒ I Ozz + MgL sin θ 0
=
dt 2 dt dt

If θ is small then sin θ ≈ θ so the equation of motion reduces to

I Ozz d 2θ
+θ =
0
MgL dt 2
MgL
This is a standard ‘Case 1’ EOM. The natural frequency is ωn = so the period is
I Ozz
 59

2π I Ozz
τ
= = 2π
ωn MgL

Next, we find the position of the ‘sweet spot’. We can do this by calculating the
reaction forces on the handle when the bat is struck, and finding the impact point j
that minimizes the reaction force. Ry
Rx
The figure shows an impact event. We assume that: O i
• The bat rotates in the horizontal plane (so gravity acts out of the plane of
the figure). ω
• The bat rotates about the point O
• The ball impacts the bat a distance d from the handle. L
• The ball exerts a (large) force Fimpact on the bat d
• Reaction forces Rx , R y act on the handle during the impact.

This is a planar problem, so we can use the 2D equations of motion. The equation Fimpact
for translational motion gives
( Rx − Fimpact )i + R y j =
MaG

For the rotational equation we can also use the short-cut for fixed axis rotation

(i )
∑ ri × F + ∑ Q( j ) =
I Ozzα z k
i j
Fimpact d
⇒ − Fimpact dk = I Ozzα z k ⇒ αz = −
I Ozz
We can relate aG to α z using the rigid body formula:
aG = α z k × rG − ω z2rG = α z Li + ω z2 Lj

We therefore see that

( Rx − Fimpact )i + R
= ( 2
y j M α z Li + ω z Lj )
x Fimpact + M α z L
⇒ R=
 MdL 
⇒
= Rx Fimpact 1 − 
 I Ozz 

The sweet spot is at the position that makes Rx = 0 , which shows that
I
d = Ozz
ML

For comparison, the ASTM formula gives

2
τ 2g
g  I  I Ozz
=d =  2π Ozz=
4π 2 4π 2  MgL  ML
60

Example 6. The ‘Cubli’ is used to develop control algorithms used to stabilize aircraft and spacecraft. It
consists of a cube whose attitude can be controlled by spinning a set of
reaction wheels inside the cube.

This simplified 1-D version of the device is used to test the algorithm that
stands the cube up on one edge. The goal of this problem is to do the
preliminary design calculations needed to set up the system.

Idealize the rectangular frame as four rods with length L and combined
mass M and the spinning wheel as a ring with radius R and mass m. The
corner at O is supported by a frictionless bearing.

Part 1: Find formulas for the mass moments of inertia of the frame and the wheel (about the center of the
wheel).

L j

The ring is easy – we can use the formula I R = mR 2

The frame is made up of four rods of mass M/4. The moment L

R
1 M 2 ω0
of inertia of one rod about its center of mass is L . We
12 4 O
need to shift the COM by a distance of L/2 to the center of the i
frame. The total mass moment of inertia of the frame is
therefore

1 M 2
M L  1
IF = 4 L2 +    = ML2
 12 4 4 2  3
 

Part 2: The frame is at rest and the wheel is spun up (clockwise) to an angular speed ω0 . Find the total
angular momentum of the system about the corner at O.

The formula for angular momentum is hO = ∑ r × mvG + ∑ Iω

Since the frame is not moving only the second term contributes
and we get h = − mR 2ω0k
j

Part 3: Thee wheel is then braked quickly, which causes the frame to
rotate about the corner O at angular speed ω f , while the motor driving
the ring spins at (clockwise) angular speed ω1 (note that this is relative
to the frame). Write down the angular momentum of the system about
O
i
ωf
 61

O.

Note that the frame rotates about O so the COM of the ring and frame are both in circular
motion about O. We know the speed of their COMs are therefore ω f L / 2

Use the formula again

h= ∑ rG × mvG + IGzzωz k
1 L L L L 
−  ML2ω f +
= M ω f + mR 2 (ω1 + ω f ) + m ωf k
3 2 2 2 2 
5 L2 
−  ML2 + m( R 2 + )  ω f k − mR 2ω1k
=
6 2 


We could also use the fixed axis rotation formula for the frame (using the mass moment of inertia
about O) but this would not work for the ring, because O is not a stationary point on the ring.

Part 4: Explain why angular momentum is conserved about O during

the braking. Use momentum conservation to find an equation relating
ω f to (ω1 − ω0 )

The external forces acting on the frame and ring together are
(1) gravity and (2) reaction forces at O. We assume that the ω1
speed change of the rotor takes place over a very short time
interval. The force of gravity is constant and exerts a
negligible impulse on the system during this time interval. The
reactions exert a finite impulse, but if we take moments about O
O the external angular impulse about O on the system vanishes.
This means angular momentum must be conserved.

5 L2 
h1 − h 0 = 0 ⇒ −  ML2 + m( R 2 + )  ω f k − mR 2ω1k + mR 2ω0k = 0
6 2 

mR 2 (ω0 − ω1 )
⇒ωf =
5 2
2 2 L 
 ML + m( R + ) 
6 2 


Part 5: For the special case ω1 = 0 show that the critical value of ω0 required to flip the frame (and ring)
into the stationary vertical configuration is

5 L2  gL ( 2 −1)
ω=
0  ML2 + m( R 2 + )  (m + M )
6 2  mR 2

62

Energy is conserved as the frame rotates up onto its edge.

The formula for the kinetic energy of a system of rigid bodies is

1 2 1
=T ∑ 2 m vG + ∑ 2 ω ⋅ IG ω
1
For 2D problems we can replace the last term by I Gzzω z2
2

Assume that the frame is at rest in the upright state. The total potential and kinetic energy in the
upright state is therefore

1 L
T1 =
+ U1 mR 2ω12 + (m + M ) g
2 2

In the initial state

2
1  L  1 2 11 L
T0 + U 0 =
2
(m + M ) 
 2  2
(
ω f  + mR 2 ω1 + ω f +
23
)
ML2ω 2f + (m + M ) g
2
15 L2  1 L
=  ML2 + m( R 2 + )  ω 2f + mR 2ω1ω f + mR 2ω12 + (m + M ) g

26 2  2 2

Energy conservation gives

15 L2  1 L 1 L
 ML2 + m( R 2 + )  ω 2f + mR 2ω1ω f + mR 2ω12 + (m + M=
)g mR 2ω12 + (m + M ) g

26 2  2 2 2 2

15 L2   1 1
⇒  ML2 + m( R 2 + )  ω 2f + mR 2ω1ω f − (m + M ) gL  − = 0

26 2    2 2

For ω1 = 0 we get

(m + M )
=ωf
5 2
gL ( 2 −1)
2 2 L 
 ML + m( R + ) 
6 2 


From part 4 we get

5 2
2 2 L 
 ML + m( R + ) 
6
 2  5 L2  gL ( )
2 −1
ω=
0 ω=
f  ML2 + m( R 2 + )  (m + M )
mR 2 6 2  mR 2

 63

6.8.2 Solutions to 3D problems

Example 1: The figure shows a wheel spinning on a frictionless axle.

The axle is supported on one side (at A) by a pivot that allows free k Ω
rotation in any direction. If the wheel were not spinning, it would j
simply swing about A like a pendulum. But if the angular speed is
high enough, the axle remains horizontal, and the wheel turns slowly A B
about the vertical axis. This behavior is called ‘precession’ and is a n i
bit mysterious – why does spin somehow hold the wheel up? The
goal of this example is to explain this, and to calculate a formula for ν
the rotation rate of the axle.
R
We will do this by showing that steady precession satisfies all the
equations of motion.
d
1.1 Let n be a unit vector parallel to the axle. Consider the disk at the
instant when n = i , and assume that
• the disk spins at constant rate about the axle at ν radians per second,
• the disk rotates slowly at constant rate about k at Ω radians per second
Find the angular velocity and angular acceleration at the instant shown in the figure

The angular velocity is easy – we just add the two vectors: ω= ν n + Ωk= ν i + Ωk

The angular acceleration is harder. Both ω and Ω are constant. But this does not mean that the
angular velocity vector is constant, because the axle is rotating about the k axis. The direction of
the angular velocity is changing, even though the magnitude is not. We can calculate the rate of
change of n by using the rigid body formula
d
(rB − rA ) =v B − v A =ω × (rB − rA )
dt
If we choose A and B to be a unit distance apart, then (rB − rA ) =n and therefore
dn
= ω×n
dt
We can now calculate the angular acceleration
dω d dn
α= = (Ωk + ν n) =ν =ν (Ωk + ν n) × n =νΩj
dt dt dt

1.2 Find a formula for the acceleration of the center of mass of the disk

We can use the rigid body formula

α (rG − rA ) + ω × [ω × (rG − rA )]
aG − a A =×
= νΩj × (di ) + (ν i + Ωk ) × (ν i + Ωk ) × (di ) 
= −Ω 2 di
64

A quicker way is to notice that the COM is in circular motion around A and use the circular motion
formula, with the same result.

1.3 Draw a free body diagram showing the forces acting on the wheel

N
A B

T i
Mg

d
1.3 Write down the equations of translational and rotational motion for the disk
(i )
∑F =MaG ⇒ Ti + ( N − Mg )k =− Md Ω 2 i
i
(i )
∑ ri × F + ∑Q ( j)
= MrG × aG + IG α + ω × [ IG ω ]
i j
 I Gxx 0 0   0  ν   I Gxx 0 0  ν 
   
⇒ di × (− Mg
= k ) M (di ) × (−Ω 2 di ) +  0 I Gyy 0  νΩ  +  0  ×  0 I Gyy 0   0 
 0 0 I Gzz   0  Ω   0 0 I Gzz  Ω 


Working through the cross products and the matrix-vector products we get
Mgd
= j I GyyνΩj − ( I Gzz − I Gxx )νΩj
We see that steady precession can indeed satisfy all the equations of motion. Moreover, for a disk (or any
solid of revolution) I Gzz = I Gxx , so we can calculate the precession rate
Mgd
= j I GxxνΩj
Mgd
⇒Ω=
I Gxxν

Example 2: The prism shown in the figure floats in space (no gravity).
At time t=0 its faces are perpendicular to the i,j,k axes as shown. It is k b
then given an initial angular velocity
= ω ω z 0k + ω y 0 j with ω y 0 << ω z 0
(i.e. we set the body spinning about the k axis, but give it a very small
disturbance) . Investigate the nature of the subsequent motion, with
c
j
both hand calculations and by writing a MATLAB script that will i
animate the motion of the prism. a
No forces or moments act on the prism. We can use the equations of motion
 65

0= MaG 0= MrG × aG + IG α + ω × [ IG ω ]
The angular momentum equation can be written out explicitly
 I Gxx 0 0   d ω x / dt    I Gxx 0 0  ω x  
     
 0 I Gyy 0   d ω y / dt  + ω x , ω y , ω z  ×   0 I Gyy 0  ω y  
 0   
 0 I Gzz   d ω z / dt   0 0 I Gzz  ω z  
(we could substitute values for I Gxx , I Gyy , I Gzz in terms of a,b,c and M but it is clearer to leave them)
Expanding out the matrix products and cross product gives
dωx
I Gxx
dt
(
+ I Gzz − I Gyy ω yω z = )
0

dω y
I Gyy − ( I Gzz − I Gxx ) ω xω z =
0
dt
dωz
I Gzz
dt
(
+ I Gyy − I Gxx ω xω y = )
0

At time t=0 ω x is zero and ω y is small. They might increase, but we will only consider behavior while
they remain small. In this case ω xω y is extremely small so we can assume d ω z / dt ≈ 0 . We can then
decouple the first two equations like this:
d 2ω y dωx
1. Differentiate the second equation with respect to time I Gyy − ( I Gzz − I Gxx ) ωz0 =
0
2 dt
dt
2. Now we can substitute for d ω x / dt using the first equation, and divide by I Gyy
d 2ω y ( IGzz − IGxx ) ( IGzz − IGyy )
2
+ ω z20ω y =
0
dt I Gyy I Gxx
This is an equation of the form
d 2ω y
+ λω y = 0
dt 2
We recognize this as an undamped vibration equation (case I or case II from our table of solutions). Its
solution depends on the sign of λ :
1. For λ > 0 the solution= is ω y A sin λ t + B cos λ t where A, B are constants. This is stable
motion - ω y remains small.
2. For λ < 0 the solution is ω=
y A exp λ t + B exp( − λ t ) . This is unstable motion - ω y will
become very large.

( )
The sign of λ is determined by the product ( I Gzz − I Gxx ) I Gzz − I Gyy . There are three possible cases:
1. I Gzz is greater than I Gxx , I Gyy (the k axis has the maximum inertia). Motion is stable
2. I Gzz is less than I Gxx , I Gyy (the k axis has the minimum inertia). Motion is stable
3. I Gzz is between I Gxx , I Gyy . Motion is unstable.

We can learn more about the motion by using MATLAB to solve the equations of motion for us. Since
there is no motion of the center of mass, we only need to consider rotational motion. We know that we can
66

describe the orientation of the prism by the rotation tensor R and its rate of change of orientation by the
angular velocity ω . The orientation and angular velocity are governed by the differential equations
dR
= WR
dt
dω
IG + ω × [IG ω ] =
0
dt
0 T
where IG = RIG R is the rotated inertia tensor for the block, and W is the spin tensor
 0 −ω z ω y 
 
= W  ωz 0 −ω x 
 −ω 0 
 y ωx

We need to set up the MATLAB ‘ode’ solver to calculate R and ω as functions of time by integrating these
equations.

We can store the unknown rotation matrix and the angular velocity vector in a MATLAB vector:
w =  Rxx , Rxy , Rxz , R yx , R yy , R yz , Rzx , Rzy , Rzz , ω x , ω y , ω z 
We need to write a MATLAB function that will calculate the time derivatives of this vector, given its
current value. The calculation involves the following steps:
(1) Assemble the vectors ω and the rotation tensor R from the Matlab solution vector w. Matlab has a
useful function that will automatically convert a matrix to a vector, and vice-versa. For example,
R (a 3x3 matrix) can be converted to w (a 1x9 column vector) using
w = reshape(transpose(R),[9,1]))
To transform w (as a column vector) back to R, you can use
R = transpose(reshape(w,[3,3]))
(2) Calculate the spin tensor W
0 T
(3) Calculate the rotated inertia tensor IG = RIG R (Matlab will multiply the matrices for us)
(4) Solve the equations for the angular acceleration α
(5) Calculate dR / dt = WR
dw
(6) Assemble the matlab vector =  R xx , R xy , R xz , R yx , R yy , R yz , R zx , R zy , R zz ,α x ,α y ,α z 
dt 

This sounds complicated but actually MATLAB is great at doing this sort of calculation efficiently. Here’s
a function:

function dwdt = rigid_body_eom(t,w)

Rvec = w(1:9); % Rotation matrix, stored as a vector
omega = w(10:12); % Angular velocity
R = transpose(reshape(Rvec,[3,3]));
II = R*I0*transpose(R); %Current inertia tensor, in fixed coord system
W = [0,-omega(3),omega(2);omega(3),0,-omega(1);-omega(2),omega(1),0];
alpha = -II\(cross(omega,II*omega)); % Angular accel
Rdot = W*R; % Rate of change of rotation matrix
Rdotvec = reshape(transpose(Rdot),[9,1]);
dwdt = [Rdotvec;alpha];
end

We just need to set up ode45 to integrate (numerically) the differential equation:

 67

omega0 = [0.,0.01,1]; % Initial angular velocity

a = [4,1,2]; % Dimensions (a,b,c) of the prism
time = 20;
initial_w = [1;0;0;0;1;0;0;0;1;transpose(omega0(1:3))];
I0 = [a(2)^2+a(3)^2,0,0;0,a(1)^2+a(3)^2,0;0,0,a(1)^2+a(2)^2];
options = odeset('RelTol',0.00000001);
sol = ode45(@(t,w) rigid_body_eom(t,w,I0),[0,time],initial_w,options);
animate_rigid_body(sol,a,[0,time])

You can download the full script here.

The figures below show animations of the predicted behavior for the three possible types of behavior

I zz is the maximum inertia – rotation is stable

I zz is the intermediate inertia – rotation is unstable (the block tumbles)

68

I zz is the smallest inertia – rotation is stable