You are on page 1of 17

Physics 430: Lecture 23

Inertia Tensor and Principal Axes

Dale E. Gary
NJIT Physics Department
10.3 The Inertia Tensor
 For the case of a body rotating about a fixed axis, we are always free to
arbitrarily call the axis the z axis, so that what we have discussed up to now
(e.g. Ixz, Iyz, Izz) is sufficent.
 However, objects may not be constrained to rotate about a fixed axis, e.g.
a tumbling object in free-fall. We can still take the instantaneous rotation
axis as the z axis, but the axis itself can be changing with time.
 In addition, there are certain spin axis directions for which the angular
momentum L and the rotation axis  are in the same direction. These are
called principal axes.
 Finally, as invariably happens, any symmetries of the object make these
principal axes obvious, and the calculations are a lot easier when
symmetries can be taken advantage of.
 What we will find is that one can calculate a single inertia tensor through
a point that takes account of the shape of the object independent of the
axis of rotation, and once that is established, the angular momentum about
any axis through that point can be determined.
 We will also find that there are certain coordinate axes, the principal axes,
for which the inertia tensor becomes especially simple.

November 24, 2009


Angular Momentum for an Arbitrary
Angular Velocity
 We will write an arbitrary angular velocity vector as ω  ( x ,  y , z ).
 The angular momentum is then
L   m r  v   m r  ω  r .
 For any position r = (x, y, z), the terms r  (ω  r ) can be explicitly written in
the rather ugly form
r  ω  r   [( y 2  z 2 ) x  xy y  xzz ,
 yx x  ( z 2  x 2 ) y  yzz ,
 zx x  zy y  ( x 2  y 2 )z ].
 A double cross-product like this can be written down with the aid of the BAC-
CAB rule ( A  (B  C)  B( A  C)  C( A  B)). Try it.
 Then the general expression for the angular momentum has components
Lx  I xx  x  I xy  y  I xz  z  I xx   m ( y2  z2 )

Ly  I yx  x  I yy  y  I yz  z  . where I x y   m x y .
Lz  I zx  x  I zy  y  I zz  z  etc.
November 24, 2009
Simpler Forms
 We can write this equation for L in simpler forms. Instead of writing x, y, z,
we can use subscripts 1, 2, 3 to get
3
Li   I ij  j . or just Li  I ij  j in Einstein summation notation.
j 1

 Or, we can write it in matrix form L  Iω,


 I xx I xy I xz   x   Lx 
     
where I   I yx I yy I yz  , ω   y  , and L   Ly  .
 I zx I zy I zz   z   Lz 
 
 In this form, I is known as the inertia tensor. To distinguish this from the
identity tensor you may be familiar with, the text uses
1 0 0 
1  0 1 0  .
0 0 1 
 Note that the term tensor refers to a higher-order vector. A vector is written
as a column, as in L and  above, while a tensor is written as a matrix.

November 24, 2009


Properties of the Inertia Tensor
 You can see from the elements of the moment of inertia tensor
I xx   m ( y2  z2 )
I x y   m x y .
etc.
that it has the property that Iij = Iji. The elements Iii are called the diagonal
elements, so we can say that the inertia tensor is unchanged by swapping off-
diagonal elements mirrored about the diagonal.
 Such a swap (replacing Iij with Iji and vice versa) is an operation called taking
the transpose of the matrix, so we can Tsay
II .
 A matrix that is its own transpose is said to be symmetric, and this symmetric
property plays a key role in the mathematical theory of the moment of inertia
tensor.

November 24, 2009


Example 10.2: Inertia Tensor for Cube
 Let’s find the moment of inertia tensor for a solid cube of mass M and side a,
rotating about a corner. By “rotating about,” we mean to take the origin at
the corner, but we do not have to specify a rotation axis. Once we know the
inertia tensor, we can specify any rotation axis ω  ( x ,  y ,  z ), and find L  Iω.
 Since the mass of the cube is evenly distributed, we need to z
convert the sums to integrals. For example, the upper-left
element becomes: y
a a a
I xx   dx  dy  dz ( y 2  z 2 ),
0 0 0

where   M / a denotes the mass density. I use the symbol O


3
x
 to distinguish from the cylindrical coordinate  (although that is not
used here).
 In principal, we have to do 9 such integrals (not a happy prospect), but you
can see by symmetry that Ixx = Iyy = Izz, and similarly for the off-diagonal
elements.
 The above integral is easy to do, and gives
a

0
a

0
a

0
a

0
a

0
a

I xx    dx  y 2 dy  dz   dx  dy  z 2 dz  23  a 5  23 Ma 2 .
0

November 24, 2009


Example 10.2, Cont’d
 The off-diagonal elements have the form
a a a a a a
I xy    dx  dy  dz xy,     xdx  ydy  dz   14  a 5   14 Ma 2 .
0 0 0 0 0 0
 Again, by symmetry, all of the off-diagonal elements have the same value.
Thus, the moment of inertia tensor is
 23 Ma 2  14 Ma 2  14 Ma 2  2
 8 3 3
 2 Ma   . [about a corner]
I    14 Ma 2 2
Ma 2
 1
Ma    3 8  3
3 4
12  
  4 Ma  4 Ma
1 2 1 2 2 2
3 Ma 
  3 3 8

 Once we have the inertia tensor, we can find the angular momentum for
rotation about any axis through this corner.
L not in same
Examples: direction as
 Rotation about x axis ( = (, 0, 0)): L = I = Ma2/12 (8, 3, 3) rotation axis
= Ma2 (2/3, 1/4, 1/4).
 Rotation about diagonal through O ( ω   / 3 (1, 1, 1) ): L is in same
 8 3 3  1 2 direction as
2 2 2
Ma   Ma    Ma
L  Iω   3 8 3  1  2  ω. rotation axis
12 3 12 3   6
 3 3 8 1  2 
November 24, 2009
Example 10.2, Cont’d
 While we are on the cube problem, let’s illustrate an important point. If we
shift the origin to the center of the cube, the diagonal element integrals are
just as easy, simply change the limits, e.g.

I xx    a /2

 a /2
dx 
a /2

 a /2
y dy 
2
a /2

 a /2
dz  
a /2

 a /2
dx 
a /2

 a /2
dy 
a /2

 a /2 
z 2 dz  2 23  a 2 (a / 2)3  16 Ma 2 .

but the off-diagonals are all odd functions, so when we change the limits they
all go to zero, e.g.
a /2 a /2 a /2 a /2 a /2 a /2
I xy    dx  dy  dz xy,     xdx  ydy  dz  0.
 a /2  a /2  a /2  a /2  a /2  a /2

 The inertia tensor is then diagonal, i.e.


2
1 0 0 
Ma   Ma 2
I 0 1 0  1.
6  6
0 0 1 
 Note that, no matter what direction  is, L is always parallel to it:
Ma 2
L  Iω  ω.
6
November 24, 2009
Example 10.3: Inertia Tensor for Cone
 Let’s do one more example—Find the moment of inertia tensor I for a spinning
top that is a uniform solid cone (mass M, height h, and base radius R) spinning
about its tips. Choose the z axis along the axis of symmetry of the cone, as
shown in the figure. For an arbitrary angular velocity , what is the top’s
angular momentum L? R
z
 The Izz element is given by the integral:
I zz   dV  ( x 2  y 2 ), h
V r=Rz/h
where the volume density is

  M / V  3M /  R 2 h .  O y
 This is most easily solved in cylindrical polar coordinates, x
(, , z), where 2 = (x2 + y2). NB: The two rho’s are different!
 Then
4
h 2 Rz / h h Rz / h  R
  h 4 3
I zz    dz  d   d   2  2  dz   3 d      z dz ,  MR 2
.
0 0 0 0 0 2 h 0 10
 The Ixx and Iyy elements are equal, and are
I xx    dV ( y 2  z 2 )  3
20 M ( R 2  4h 2 ).
V

November 24, 2009


Example 10.3: Cont’d
 All of the off-diagonal elements are zero. Note that symmetry about any two
axes guarantees that all of the off-diagonal elements are zero. Then, the
moment of inertia tensor is:
 R 2  4h 2 0 0  1 0 0 
3  
I M 0 R 2  4h 2 0    0 2 0  .
20
 2
 0 0 2 R   0 0 3 

 The last form is just for further discussion. A matrix with all zero off-diagonal
elements is, as we said, called a diagonal matrix. We can then write
L  Iω  (1 x , 2 y , 3 z ).
 What this means is that whenever  points along one of the three coordinate
axes, L and  are parallel. This brings us (finally) to the concept of principal
axes of inertia.

November 24, 2009


10.4 Principal Axes of Inertia
 Whenever L and  point in the same direction, this marks a principal axis
for the body. For principal axes, we can express this parallelism as L = .
But since L = , that means the constant of proportionality is the moment of
inertia about that axis, so  is called a principal moment.
 Recall that when I is diagonal
1 0 0 
I   0 2 0  , thenL  Iω  (1 x , 2 y , 3 z ).
 0 0 3 
 Thus, each of the i are principal moments, and are in fact THE principal
moments for rotations about a particular origin.
 If a body has an axis of symmetry about any origin, then that axis is a
principal axis. Furthermore, there are always two additional principal axes
perpendicular to such a symmetry axis (even though there is no further
symmetry about those additional axes).
 Although symmetry makes it easier to spot a principal axis, it is NOT required.
Any body, symmetrical or not, has three principal axes about any origin, for
which the inertia tensor I is diagonal.
November 24, 2009
Principal Axes of Inertia-2
 The appendix shows how to diagonalize any matrix, and since such a
diagonalized matrix represents three principal axes, any choice of axes for
which the matrix starts out non-diagonal can be transformed into a new set of
axes (called eigenvectors) for which the matrix is diagonal.
 Before going on to discuss this further, we first make a point about
determining the kinetic energy of a rotating body. Our old form of kinetic
energy (valid only when L and  are aligned, which we called the z axis) is
most properly given now as
T  12 I zz  2 .
 The general expression is (see problem 10.33)
T  12 ω  L.
 Using a set of principal axes, we have L  (11 , 22 , 33 ), so the kinetic
energy becomes:
T  12 (112  222  332 ).

November 24, 2009


10.5 Finding the Principal Axis
Eigenvalue Equations
 The recipe for finding the principal axes through some origin O is as follows:
Find the inertia tensor by choosing some axes. If the inertia tensor is
diagonal, we are done. We have already found the principal axes. If not, we
need to diagonalize the tensor matrix.
 How do we do that? The clue is that the equation
Iω   ω,
must be satisfied for some as yet unknown number . This is called an
eigenvalue equation, and the number  is called the eigenvalue. The vector 
is called an eigenvector, and will be one of our principal axes.
 In practice, we want to know three eigenvalues, and three eigenvectors,
which will give us our three principal axes. As it turns out, we can find all
three at once. The trick is to recognize that we can convert the number  into
a matrix by multiplying by the identity matrix 1. Since
leads to cubic for ,
Iω   1ω,
with three roots
we must have (I   1)ω  0, whose only non-trivial solution is
det(I  1)  0.
November 24, 2009
Example 10.4: Principal Axes for
Cube about a Corner
 Find the principal axes and corresponding moments for the cube of Example
10.2, rotating about its Matlab Symbolic Math
corner.
>> syms x y;
 Solution:
>>We
A = found
[8*x-y,in-3*x,
example-3*x; 10.2
-3*x,that the-3*x;-3*x,
8*x-y, inertia tensor
-3*x, about
8*x-y] the three

edges of the cube were: [ -3*x, 8*x-y, -3*x]


[ -3*x,
2
 8-3*x,3 8*x-y]
3  8 3 3 
>> det(A) Ma   =  3 .
I = 242*x^3-165*y*x^2+24*x*y^2-y^3
ans 3 8 3 8  3
>> factor(det(A)) 12    
 3 3
ans = (2*x-y)*(11*x-y)^2 8  3 3 8
 Using det(I  A
1=) [ 0,
8*x-y, -3*x, -3*x]
we have
8   3 3
3 8   3  (2   )(11   ) 2  0
3 3 8  
 The obvious solutions are
1  2  16 Ma 2 and 2  3  11  12
11
Ma 2 .
 In this case, two of the roots are a repeated root. These are the eigenvalues
(moments), but what are the eigenvectors? Plug back into (I   1)ω  0.
November 24, 2009
Example 10.4: Cont’d
 We have to solve this equation three times, once for each eigenvalue (although
in the present case where two solutions are repeated, we just have to do it
twice). First, insert   2 ,
8   3 3   x   6 3 3  x 
(I  1)ω  0   3 8   3       3 6 3    0.
  y   y
 3 3 8     z   3 3 6   z 
 This yields the simultaneous set of equations: 2 x   y   z  0
  x  2 y   z  0
  x   y  2 z  0.
 The solution to this is x = y = z, hence the unit vector along this direction is
e1  1is
which (1,1,1),
3 the diagonal of the cube.
 The other solution is surprising:  3 3 3  x 
  3 3 3  y   0   x   y   z  0.
 3 3 3  z 
  
Since ω  e1  13  x   y   z  0,this means any axis perpendicular to e1. Any
perpendicular axes e2 and e3 that are also perpendicular to e1 work!
November 24, 2009
10.6 Precession of a Top due to Torque
 We now know enough to do some interesting problems. Take the case of a
spinning top (say the cone we just looked at), spinning freely about its tip.
 We will take the axes x, y, z to be fixed to the ground, and take the principal
axes of the top to be e1, e2 and e3, with e3 along the symmetry axis.
 Let’s first ignore gravity and consider the top to be spinning e3
around the e3 principal axis (i.e.  = e3). Then the angular z

momentum is fixed, as L  3ω  3 e3 . With gravity off, R


Mg
there is no torque, and L = constant. 
 Let’s now assume that there is a small amount of gravity.
O y
What actually happens is that there is now some x
torque, so  starts to change, and  and  are no
longer zero. However, let’s consider the case where the effect of gravity is so
small that we can ignore these changes.
 The torque is R ×Mg, with magnitude RMg sin . Note that the direction is
perpendicular to R (or e3), hence the direction of e3 begins to change, but not
R  Mg  L  3e 3 .
the magnitude of , i.e.
November 24, 2009
Precession of a Top due to Torque-2
 Solving for the changing e3, and noting that R  Re3 and g   gzˆ ,
MgR
e 3  zˆ  e3  Ω  e3 ,
3
MgR
where Ω  zˆ .
3

 This is the angular velocity of precession of the top. When gravity becomes
greater, so that we cannot ignore the changes in  and , then we will find
another wobble of the top, called nutation (see section 10.10).
 The Earth displays exactly these same motions, in this case due to non-central
forces by the Sun and Moon. The precession of the equinoxes takes about
23,000 years, and over time makes the north pole of the Earth point to
different “pole stars.” The current pole star, Polaris, will cease being the
“north star” over time.

November 24, 2009

You might also like