4.2.5 Angle-Preserving Mappings

The euclidean motions we have just described as $ v\mapsto R\cdot v + w$, where $ R$ is a rotation or reflection and $ w$ is a translation vector, are length and angle preserving. Now we try to identify those mappings which are only angle preserving.


Lemma (Linear conformal mappings).
Let $ f:\protect\mathbb{R}^n \to \protect\mathbb{R}^m$ be linear. Then the following statements are equivalent:

  1. $ f$ respects angles, i.e. is CONFORMAL;
  2. $ \exists \lambda >0:
\langle f(x),f(y) \rangle=\lambda \langle x,y \rangle$ for all $ x,y\in \protect\mathbb{R}^n$;
  3. $ \exists \mu >0$: $ \mu \, f$ is an isometry.

Proof. $ (2\Leftrightarrow 3)$ is obvious with $ \lambda \,\mu ^2=1$.

$ (1\Leftarrow 2)$ let $ \alpha $ be the angle between $ x$ and $ y$ and $ \alpha '$ the one between $ f(x)$ und $ f(y)$. Then

$\displaystyle \cos\alpha '=\frac{\langle f(x)\vert f(y) \rangle}{\Vert f(x)\Ver...
...y \rangle}{\sqrt \lambda \Vert x\Vert\sqrt \lambda \Vert y\Vert}=
\cos\alpha .
$

Thus $ \alpha =\alpha '$, and $ f$ respects angles.

$ (1\Rightarrow 2)$ We define $ \lambda (v)\geq 0$ implicitly by $ \langle f(v)\vert f(v) \rangle=:
\lambda (v)\langle v\vert v\rangle$.

Let $ v,w$ be orthonormal vectors, then $ (v+w)\perp(v-w)$. Since $ f$ is conform, we have:

$\displaystyle 0=\langle f(v+w)\vert f(v-w) \rangle =
\langle f(v)\vert f(v) \rangle-\langle f(w)\vert f(w) \rangle
=\lambda (v)-\lambda (w).
$

Thus $ \lambda $ constant on the orthonormal basis $ (e_1,\ldots,e_n)$ von $ \protect\mathbb{R}^n$. We put $ \lambda :=\lambda (e_1)=\ldots=\lambda (e_n)$. For an arbitrary vector $ v\in \protect\mathbb{R}^n$ we have:

  $\displaystyle \phantom{\Rightarrow \quad} v=\sum _{i=1}^n v^i e_i$    
  $\displaystyle \Rightarrow \quad \lambda (v)\sum (v^i)^2=\lambda (v)\left \langle \sum v^i e_i\vert\sum v^i e_i \right \rangle=$    
  $\displaystyle \qquad=\left\langle f\left({\sum _i v^i e_i}\right), f\left({\sum _j v^j e_j}\right)\right\rangle$    
  $\displaystyle \qquad=\sum _{i,j} v^iv^j\underbrace{\langle f(e_i)\vert f(e_j) \rangle} _{\lambda \delta _{i,j}} = \lambda \sum _i (v^i)^2$    
  $\displaystyle \Rightarrow \quad \lambda (v)=\lambda \quad\forall v\in \protect\mathbb{R}^n.$    

The lemma now follows from:

$\displaystyle 2\langle v\vert w \rangle$ $\displaystyle =\langle v+w\vert v+w \rangle-\langle v\vert v \rangle- \langle w\vert w \rangle$    
$\displaystyle 2\langle f(v)\vert f(w)\rangle$ $\displaystyle =\langle f(v)+f(w)\vert f(v)+f(w) \rangle- \langle f(v)\vert f(v) \rangle - \langle f(w)\vert f(w) \rangle$    
  $\displaystyle =\lambda \langle v+w\vert v+w \rangle-\lambda \langle v\vert v \rangle- \lambda \langle w\vert w \rangle$    
  $\displaystyle =2\lambda \langle v\vert w \rangle.\quad\qed$    

Andreas Kriegl 2003-07-23