# Constraints the elements of the 3D-rotation matrix must satisfy

1. Aug 4, 2014

### ELB27

1. The problem statement, all variables and given/known data
Taken from "Introduction to Electrodynamics" by David J. Griffiths p.12 problem 1.8 (b):
What constraints must the elements $R_{ij}$ of the three-dimensional rotation matrix satisfy, in order to preserve the length of vector A (for all vectors A)?

2. Relevant equations
The 3D rotation matrix around an arbitrary axis:
$$\begin{pmatrix} \bar{A_x}\\ \bar{A_y}\\ \bar{A_z}\\ \end{pmatrix} = \begin{pmatrix} R_{xx} & R_{xy} & R_{xz}\\ R_{yx} & R_{yy} & R_{yz}\\ R_{zx} & R_{zy} & R_{zz}\\ \end{pmatrix} \begin{pmatrix} A_x\\ A_y\\ A_z\\ \end{pmatrix}$$
Thus,
$\bar{A_i} = \sum_{j=1}^3 R_{ij}A_j$
where the index 1 stands for x, 2 for y and 3 for z.

3. The attempt at a solution
Since the length of the vector before and after transformation must be equal:
$\sum_{i=1}^3 (\bar{A_i})^2 = \sum_{i=1}^3 (A_i)^2$
$\sum_{i=1}^3 (\bar{A_i})^2 = \sum_{i=1}^3 (\bar{A_i})(\bar{A_i}) = \sum_{i=1}^3 \left( \sum_{j=1}^3 R_{ij}A_j \right) \left( \sum_{k=1}^3 R_{ik}A_k \right) = \sum_{i=1}^3 (A_i)^2$
The one before the last equality I obtain from substituting the above equation for $\bar{A_i}$ .
I'm not sure how to proceed from here. I guess that I lack understanding of the summation notation. I tried looking it up but I still can't understand how to continue.

Any help will be greatly appreciated!

2. Aug 4, 2014

3. Aug 4, 2014

### vela

Staff Emeritus
First thing you should do is learn Einstein's summation convention (Einstein's greatest contribution to physics): repeated indices imply summation. It'll save you a lot of tedious writing. Using this convention, you have
$$\bar{A}_i \bar{A}_i = (R_{ij}A_j)(R_{ik}A_k).$$ No annoying sigmas to keep writing.

The relationship has to hold for all vectors, so try using $A = (1, 0, 0)$, for example, to show that the first column of $R$ has to be a unit vector. For other choices of $A$, you can show that different columns are orthogonal to each other.

4. Aug 5, 2014

### ELB27

Thanks, I didn't know about these matrices!

Thank you! So if I understand correctly, I just pick 2 convenient vectors like $\vec{A} = <1,0,0>$ and $\vec{B} = <1,1,0>$ and generalize the results I get for all choices? Am I allowed to do it without proving these relations in general?

EDIT: After some more readings I think I finally get it. Thanks again.

Last edited: Aug 5, 2014