# $3×3$ orthogonal matrix $T$ fixes 2 points in unit sphere

Let $$T:\mathbb{R^3} \rightarrow \mathbb{R^3}$$ be an orthogonal transformation such that $$\det T = 1$$ and $$T\neq I$$. Let S be the unit sphere in $$\mathbb{R^3}$$. I need to show that $$T$$ fixes exactly two points on S.
What I think is, if I can show $$T$$ has eigenvalue $$1$$ with geometric multiplicity $$1$$, then I am done.The possible eigenvalues of $$T$$ are:
1)$$1,-1,-1$$
2)$$1,a+ib,a-ib$$ where $$a,b\in\mathbb R$$ and $$a^2+b^2=1$$
3) all the eigenvalues are $$1$$.
Now I am struggling with the case 3. How can I show that the case 3 is not possible? I think I have to use the fact $$T\neq I$$, but don't know how.

• Any real orthogonal matrix is normal so by spectral theorem all eigenvalues (that exist) are semi-simple -- i.e. geometric multiplicity equal algebraic multiplicity Commented Apr 30, 2021 at 17:40
• But that's for complex orthogonal matrix,right? Commented Apr 30, 2021 at 18:07
• You need to be a lot more careful with definitions -- the answer is no. The matrices $\in O_n(\mathbb C)$ are not normal in general. On the other hand $U_n(\mathbb C)$ only has normal matrices in it --i.e. unitary matrices are normal. And real orthogonal matrices may be viewed as a special case of unitary. Commented Apr 30, 2021 at 19:34
• I know when we consider T as a complex matrix, then by spectral theorem it’s diagonalizable, i.e, algebraic and geometric multiplicity is same for any eigenvalue of T. But how are you saying the same thing for real orthogonal matrix? Please explain Commented May 1, 2021 at 1:35
• 3x3 orthogonal matrix T has real eigenvalue 1 with algebraic multiplicity 3. When we consider T as complex matrix, we know it’s diagonalizable by spectral theorem . That's geometric multiplicity is also 3. Suppose the independent eigenvectors { (1,0,0),(0,i,0),(0,0,i)}. But how are you saying that T as real matrix will also have geometric multiplicity 3? Commented May 1, 2021 at 1:55

If all eigenvalues are $$1$$, then there are three linealy independent vectors $$v_1$$, $$v_2$$ and $$v_3$$ such that$$T(v_i)=v_i\text{ for each }i\in\{1,2,3\}.\tag1$$But then $$\{v_1,v_2,v_3\}$$ is a basis of $$\Bbb R^3$$ and it follows from $$(1)$$ that $$T$$ is the identity map.
• For $3\times3$ matrices in general, yes; for orthogonal matrices, no. Commented Apr 30, 2021 at 16:49
• Suppose that $T(v)=v$, for some $v\ne0$. Consider the set $v\perp=\{w\in\Bbb R^3|mid\langle v,w\rangle=0\}$. Then (since $T$ is orthogonal), $T\left(v^\perp\right)\subset v^\perp$. But $\dim v^\perp=2$ and $T|_{v^\perp}$ is orthogonal and it has determiant $1$. Therefore, it is a rotation, with respect to some angle $\theta\in[0,2\pi)$. If $\theta=0$, then $T=\operatorname{Id}$. Otherwise, $1$ is not an eigenvalue of $T|_{v^\perp}$, and therefore the geometric multiplicity of $1$ is $1$. Commented Apr 30, 2021 at 16:57