A mapping $f:\mathbb{R}^n \rightarrow \mathbb{R}^n$ is affine if there is an invertible $n$ x $n$ matrix $M$ and a vector $s \in \mathbb{R}$ such that $f(x)=Mx+s$ for every $x \in \mathbb{R}^n$.
- Show that every affine map $f(x) = Mx+s$ is invertible and its inverse is also affine?
Show that the composition of affine maps $\mathbb{R}^n \rightarrow \mathbb{R}^n$ is affine.
For two subsets, $S$, $T$ of $\mathbb{R}^n$, write $S \sim T$ if there exists an affine map $f$ such that $f(S) =T$. We then say that $S,T$ are affine-related. Prove that $\sim$ is an equivalence relation on the set of subsets of $\mathbb{R}^n$.
1.) Let
$f(x) = Mx + s \tag{1}$
be an affine map $\Bbb R^n \to \Bbb R^n$. Suppose
$f(x_0) = y_0 \tag{2}$
for some $x_0 \in \Bbb R^n$. Then we have
$Mx_0 + s = y_0, \tag{3}$
or
$Mx_0 = y_0 - s, \tag{4}$
or, since $M$ is invertible,
$x_0= M^{-1}(y_0 - s) = M^{-1}y_0 - M^{-1}s. \tag{5}$
(4) suggests we set
$t = -M^{-1}s \tag{6}$
and consider the affine map
$g(x) = M^{-1}x + t; \tag{7}$
we have, for any $x \in \Bbb R^n$,
$g(f(x)) = M^{-1}(f(x)) + t = M^{-1}(Mx + s) + t$ $ = M^{-1}Mx + M^{-1}s + t = Ix + M^{-1}s - M^{-1}s = x, \tag{8}$
and
$f(g(x)) = f(M^{-1}x + t) + s = M(M^{-1}x + t) + s$ $ = MM^{-1}x + Mt + s = Ix - s + s = x, \tag{9}$
using (6). This shows that any affine map $f(x)$ has an affine inverse
$g(x) = f^{-1}(x). \tag{10}$
2.) Taking
$g(x) = Nx + r, \tag{11}$
where $N$ is invertible, we see that
$g(f(x)) = Nf(x) + r = N(Mx + s) + r = NMx + Ns + r = NMx + (Ns + r); \tag{12}$
since $M$ and $N$ are both invertible, so is $NM$; thus $g(f(x))$ is affine, showing that the composition of two affine maps is also affine.
3.) This is actually a special case of a much more general observation about collections of invertible maps on arbitrary sets:
Generality: Let $\Sigma$ be any set, and let $\mathcal C$ be any non-empty collection of invertible mappings $f:\Sigma \to \Sigma$ which is closed under both composition and the taking of inverses, that is
$f, g \in \mathcal C \implies f \circ g:\Sigma \to \Sigma \in \mathcal C, \tag{13}$
and
$f \in \mathcal C \implies f^{-1}:\Sigma \to \Sigma \in \mathcal C. \tag{14}$
If, for $\Delta_1, \Delta_2 \subset \Sigma$ we write
$\Delta_1 \sim \Delta_2 \Longleftrightarrow \exists f \in \mathcal C, f(\Delta_1) = \Delta_2, \tag{15}$
then $\sim$ is an equivalence relation on $2^\Sigma$, the power set (set of all subsets) of $\Sigma$.
Proof of Generality: Note $\mathcal C \ne \emptyset$ by hypothesis; denoting the identity map from $\Sigma \to \Sigma$ by $Id$, that is, $Id(x) = x$ for $x \in \Sigma$, we have $Id(\Delta_1) = \Delta_1$ for any $\Delta_1 \subset \Sigma$; furthermore, for $f \in \mathcal C$ and $x \in \Sigma$, since $f^{-1} \in \mathcal C$ by (14) and $f \circ f^{-1}, f^{-1} \circ f \in \mathcal C$ by (13), and
$Id(x) = f^{-1} \circ f(x) = f \circ f^{-1}(x), \tag{16}$
we see that $Id \in \mathcal C$. Thus $\Delta_1 \sim \Delta_1$ showing $\sim$ is reflexive. Now if $\Delta_1 \sim \Delta_2$, there is an $f \in \mathcal C$ with
$f(\Delta_1) = \Delta_2, \tag{17}$
whence
$f^{-1}(\Delta_2) = \Delta_1 \tag{18}$
with $f^{-1} \in \mathcal C$ by (13). Thus
$\Delta_2 \sim \Delta_1 \tag{19}$
and so $\sim$ is reflexive. Finally, suppose further that $\Delta_3 \subset \Sigma$ and
$\Delta_1 \sim \Delta_2, \Delta_2 \sim \Delta_3; \tag{20}$
then there are $f, g \in \mathcal C$ with
$f(\Delta_1) = \Delta_2, g(\Delta_2) = \Delta_3, \tag{21}$
whence
$g \circ f(\Delta_1) = g(\Delta_2) = \Delta_3, \tag{22}$
that is, since $g \circ f \in \mathcal C$ by (13),
$\Delta_1 \sim \Delta_3, \tag{23}$ and we see that $\sim$ is transitive. Thus $\sim$ is an equivalence relation on $2^\Sigma$. End of Proof of Generality.
Applying this Generality to the case at hand, taking $\Sigma = \Bbb R^n$ and $\mathcal C$ to be the set of affine mappings $\Bbb R^n \to \Bbb R^n$, we immediately see that the resulting relation $\sim$ is an equivalence relation on $2^{\Bbb R^n}$, by virtue of items (1.) and (2.) above.