Sie sind auf Seite 1von 29

Sect. 4.

7: Finite Rotations
So far: Have used various representations to
describe the relative orientation of 2 Cartesian coord
systems with common origin: Weve shown the
transformation matrix A in terms of:
The Euler Angles: A = A(,,)
The Cayley-Klein parameters: A = A(,,,)
Euler parameters: A = A(e0,e1,e2,e3)
Single rotation about a given direction A = A()
Here: We seek another representation: A vector
relationship between an initial vector r & the vector
r which results from the orthogonal transformation
(rotation) r = Ar. Representation in terms of the
rotation angle & direction cosines of rotation axis.
Get result using figure &
vector algebra. Treat
transformation as active
Counterclockwise coord
rotation = Clockwise
rotation of vectors in the
fixed coord system.
Fig: Initial position of vector r = OP. Final position
r = OQ. Rotation axis along ON. Unit vector along ON
= n. Rotation angle = . |ON| = nr ON = n(nr).
NP = OP - ON = r - n(nr). |NP| = |NQ| = |rn|
OQ = r = ON + NV + VQ
Taking all of these together:
r = n(nr) + [r - n(nr)]cos + (rn)sin Or:
r = r cos + n(nr)(1 - cos) + (rn)sin
The Rotation Formula
Holds for any rotation!
Worth mentioning: Can easily express rotation angle
in terms of Euler angles ,, by comparing
representations of the transformation matrix A in
terms of one & the other.
For A = A(,,) we had:

For A = A() we had:


cos sin 0
A= -sin cos 0
0 0 1
The trace of A is invariant under a change of
representation:
TrA() = TrA(,,)

Use some trig identities (student exercise!):


cos[())] = cos[()( + )]cos[()]

Sign fixed by requiring 0


as ,, 0
Sect. 4.8: Infinitesimal Rotations
Summary: Representations to describe relative
orientation of 2 Cartesian coord systems with
common origin: r = Ar
Euler angles: A = A(,,)
Cayley-Klein parameters: A = A(,,,)
Euler parameters: A = A(e0,e1,e2,e3)
Single rotation about a given direction A = A()
Direct vector relationship between r & r
r = r cos + n(nr)(1 - cos) + (rn)sin
In all cases discussed for the transformation A, the
number of matrix elements is > # independent
variables (=3) Had a number constraints between
variables to reduce # indep ones to 3.
In proof of Eulers Theorem weve shown: Any given
orientation r = Ar can be obtained by a single
rotation about some axis.
It would be nice if we could associate a vector (with
3 independent components) with a finite rotation of a
rigid body about a fixed point.
Direction of vector is obvious: Direction of axis of
rotation. That is, direction of eigenvector R.
Magnitude: A suitably chosen function of rotation
angle could possibly be used as magnitude of this
vector.
However, we cannot do this (for finite rotations!).
Why?
Let A & B be 2 vectors associated with
transformations A & B.
Vectors Addition must be commutative:
A + B = B + A
But, weve seen, the addition of 2 rotations (one,
followed by a 2nd) corresponds to product AB
Matrix multiplication is NOT commutative:
AB BA
A & B are not commutative & cannot be
represented as vectors.
That is, the sum of finite rotations depends on the
order of rotations. See figure for illustration:
A finite rotation cannot be represented by a vector.
However, we can show that infinitesimal rotations CAN
be represented by a vector!
Infinitesimal rotation An orthogonal transformation
of coordinate axes x = Ax in which components of the
vector are almost the same with respect to both sets of
axes (infinitesimal change).
That is, xi xi + small corrections (i = 1,2,3)
Write: xi = xi + i1x1 + i2x2 + i3x3 (i = 1,2,3)
Where the ij are infinitesimals. In calculations only
terms linear in ij are kept, higher powers are neglected.
Using summation convention, write infinitesimal
transformation as: xi = (ij+ ij)xj (i = 1,2,3)
In matrix notation (defining infinitesimal matrix &
relating it to transformation matrix A) this is:
x (1+ )x Ax
That is: (1+ ) A
Note that the sequence of operations doesnt matter
for infinitesimal transformations 1 + 2 (they commute
because powers of higher than 1st are neglected):
(1+ 1)(1+ 2) = 12 + 11 + 21 + 121
Or: (1+ 1) (1+ 2) 1 + 1 + 2
The order of infinitesimal rotations doesnt matter!
Infinitesimal rotations: (1+ ) A
In terms of Euler angles ,,, orthogonal matrix
A is represented as:

For infinitesimal Euler angles: d,d,d:


cos(d) cos(d) cos(d) 1
sin(d) d, sin(d) d, sin(d) d
For infinitesimal rotations:
1 (d + d) 0
A = - (d + d) 1 d
0 - d 1
Infinitesimal rotations: (1+ ) A, = A - 1
Where: 1 (d + d) 0
A = - (d + d) 1 d
0 - d 1

0 (d + d) 0
= A - 1 = - (d + d) 0 d
0 - d 0

Euler angle representation of


General properties of & A: = A - 1, 1+ = A
Clearly, A-1 = = 1-
Proof: AA-1 = (1+ )(1- ) = 12 + 1
Also: = - . is antisymmetric!
Proof: = 1- = 1 + = -
Convenient to write in form:
0 d3 - d2
= - d3 0 d1
d2 - d1 0
d1, d2, d3 3 infinitesimal parameters
identified with 3 independent parameters
needed to specify rotation.
We had: r (1+ )r Ar (or x = (1+ )x = Ax)
r - r = r dr
Using the matrix, this becomes:
dx1 = x2d3 - x3d2
dx2 = x3d1 - x1d3
dx3 = x1d2 - x2d1
These expressions should remind us of components
of the cross product of 2 vectors.
Define the vector: d vector with components
d1, d2, d3
Text emphasizes that d differential of a vector! Instead,
it is a vector of differential magnitude. That is a vector
which has a differential d does not exist!!!!
So, we can write: dr = r r d
r is a vector Transforms under an orthogonal
transformation B according to r = Br or:
xi = bijxj
If d is a vector, it must also transform like this
under an orthogonal transformation B. Will now
look at the transformation properties of d
To look at this, first look at transformation
properties of matrix under the transformation B.
Previous discussion = BB-1
It is shown in Prob. 3, p 180, that the property that
is antisymmetric is preserved in this similarity
transformation.
So similar to , can also be written as:

0 d3 - d2
= -d3 0 d1
d2 - d1 0

Skipping some tedious steps, can show that di are


related to dj by (|B| determinant of B) :
di = |B|bijdj

This transformation is ALMOST the same as for a


vector (xi = bijxj) but differs by the factor of |B|.
Before discussing the details of the difference between
d and a vector, look at its properties another way.
For FINITE rotations, we had:
r = r cos + n(nr)(1 - cos) + (rn)sin
The Rotation Formula (1)
To get the present INFINITESIMAL rotation case
from this, let d in (1).
cos(d) 1, sin(d) d
(1) becomes: r = r + (rn)d
Or: r - r dr = (rn)d
We had: dr r d d nd
Now, discuss transformation properties of
d = nd under the orthogonal transformation B:
We had: di = |B|bijdj (1)
Ordinary vectors transform like: xi = bijxj (2)
Define: (property under the transformation B):
Polar Vector Any quantity for which the
components transform according to (2)
Axial Vector Any quantity for which the
components transform according to (1)
(also called a pseudovector)
Properties under inversion Sij = -ij:
Polar Vector: All components change sign
Axial Vector: Components do not change sign.
Sect. 4.8: Polar & Axial Vectors
Example of an axial vector: A vector which is the
cross product of 2 polar vectors. If D, F are polar
vectors: V* = D F is an axial vector.
Clearly, Vi* = DjFk -DkFj (i,j,k cyclic)
By definition of polar vectors, components Dj, Fk, Dk,
Fj change sign on inversion Vi* does not.
Physical quantities which are polar vectors:
Position r, velocity v, momentum p = mv,
acceleration a, force F = ma, electric field E,...
Physical quantities which are axial (pseudo)
vectors: Angular momentum L = r p, torque
N = r F, magnetic field B, ...
Define a Parity Operator P (actually ~ same as
inversion operator of earlier): P converts x, y, z to -x,-y, -z
S = scalar, V = vector, V* = axial (pseudo) vector
PS = S, PV = -V, PV* = V*
Suppose, we have a scalar (like) quantity S* VV*
PS* = P(VV*) = -(VV*) = -S*
Pseudoscalar Any otherwise scalar quantity
which behaves like S* under inversion.
Any quantity which can be written as the scalar product
of a polar vector and an axial (pseudo) vector.
Physical quantities which are pseudoscalars:
Magnetic flux, .
Polar vectors: The vector itself remains unchanged on
inversion, the components change sign.
Axial vectors: Carry a handedness property with
them (consistent with the cross product! ). Cross product (the
vector itself!) clearly changes sign with coordinate
inversion.
Back to infinitesimal rotations:
We had: di = |B|bijdj (1)
Ordinary vectors transform like: xi = bijxj (2)
Also: dr r d. Both r & dr are polar vectors &
transform according to (2). d must be an axial
vector, transforming according to (1).
d nd. Inversion n changes direction.
Some comments on cross product. C = D F
Introduce the Levi-Civita density ijk
Write Cartesian components of C:
Ci j,k ijkDjFk
ijk permutation symbol or Levi-Civita density
ijk 0, if any 2 indices equal
1, if i, j, k are even permutation of 1,2,3
-1, if i, j, k are an odd permutation of 1,2,3
122 = 313 = 211 = 0 , etc.
123 = 231 = 312 =1, 132 = 213 = 321 = -1
Formalism developed so far (r = Ar) to represent the
orientation of a rigid body. Transformation involves
rotation of the coordinate system. That is, usually
assumed passive interpretation of transformation A.
Counterclockwise rotation about some axis.
Often need active interpretation of r = Ar.
Physical system & the associated vectors are
rotating in the clockwise direction.
Often also, need rotation formulas explicitly for this
but for vectors rotating in counterclockwise direction.
If careful, can use formulas just derived. But, need to
let direction of rotation be opposite & also axis of
rotation to be opposite.
Figure illustrates passive, active transformations:
Formulas for Counterclockwise Rotations
(note sign changes in , di)
The Rotation Formula

r = r cos + n(nr)(1 - cos) - (rn)sin
Infinitesimal rotations: dr = - (rn)d
Antisymmetric, infinitesimal rotation matrix
0 - d3 d2
= d3 0 - d1
- d2 d1 0
Or, using d nd
0 -n3 n2 (ni = components of n)
= n3 0 -n1 d
n2 n1 0
Using dr = - (rn)d we can immediately get
the derivative of r with respect to :
(dr/d) = - (rn) (1)
Define the matrix N: Nr (rn)
(dr/d) = -Nr (2)
Using the cross product definition (&
summation convention):
(rn)i ijk xjnk
Nij = ijk nk
Another (sometimes) useful representation for matrix :
Start from 0 -n3 n2
= n3 0 -n1 d (ni = components of n)
n2 n1 0
Define matrices Mi Infinitesimal rotation generators
0 0 0 0 0 1 0 -1 0
M1 0 0 -1 , M2 0 0 0 , M3 1 0 0
0 1 0 -1 0 0 0 0 0
Can show: MiMj - MjMi [Mi,Mj] = ijk Mk (cyclic i,j,k)
[Mi,Mj] Commutator Lie Bracket
These relations define the Lie algebra of the rotation group.
Relations like these come up all the time in QM (where the Ms
are matrix representations of angular momentum)!
Finally can write: = niMid (summation convention!)

Das könnte Ihnen auch gefallen