Suppose $f$ $-$ isometry of the affine space. $f=Ax+b$ where $A \in O(n)$ and $b$ $-$ is a vector. Proposition. f has a fixed point $\iff b$ is orthogonal to space of eigenvectors with eigenvalue equal to $1$.

Is it right? How can I show this?


Without loss of generality we may suppose that the matrix $A$ is diagonal with elements $(c_1,\ldots,c_k,\overbrace{1,1,\ldots,1}^{\text{ n-k times}})$ on the diagonal with respect to a basis $(e_1,\ldots,e_k, e_{k+1},\ldots,e_n)$. The subspace $V_1$ spanned by the vectors $(e_1,\ldots,e_k)$ is orthogonal to the subspace $V_2$ spanned by $(e_{k+1},\ldots,e_n)$, the former consisting of eigenvectors corresponding to eigenvalues $\neq 1$, the latter to eigenvectors with eigenvalues $=1$. The equation for a fixed point $x$ gives us $\begin{cases} (c_1-1)x_1 + b_1 & = 0\\ (c_2-1)x_2 + b_2 & = 0\\ \ldots \\ (c_k-1)x_k + b_k & = 0\\ b_{k+1}& = 0\\ \ldots\\b_n & = 0\end{cases}$.

This shows that the system has a solution $\iff$ $b$ lies in the subspace $V_1$ and is thus orthogonal to the subspace $V_2$.

  • $\begingroup$ What is "the equation for a fixed point"? It's the main difficulty for me. I would say that $Ax=x$ but $A$ acts on the vectors meanwhile $x$ a point. $\endgroup$ – iou Feb 1 '17 at 17:01
  • $\begingroup$ It's $Ax+b-x =0$, expressing that $f(x)=x$, so that x is a fixed point. $\endgroup$ – Marc Bogaerts Feb 1 '17 at 17:03
  • $\begingroup$ @iou Note that in the plane the eigenspace with eigenvalue 1 is the zero subspace so every vector $b$ is "perpendicular" to it, so every isometric affine transformation in the plane has a fixed point. In 3d space the condition means that $b$ should be perpendicular to the axis of rotation of $A$. $\endgroup$ – Marc Bogaerts Feb 1 '17 at 17:15
  • $\begingroup$ let $f=Ax+b$ be the isometric transformation of $E^n$ and $b$ isn't orthogonal to space of eigenvectors with eigenvalue 1, then $f$ has a invariant line, is it right? $\endgroup$ – iou Feb 1 '17 at 17:54
  • $\begingroup$ That's quite a different problem where I have no answer for. I think this is a nice question to pose. $\endgroup$ – Marc Bogaerts Feb 1 '17 at 18:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.