Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $V$ be a vector space with the orthonormal basis $Q = \{ \vec{q_1},\ldots, \vec{q_n} \}$ and let $\ell:V\to V$ be an orthogonal map. Prove that the matrix $L$ of of $\ell$ with respect to $Q$ is orthogonal.

Note: By orthogonal map I mean that $\ell$ is linear and satisfies $\left\Vert \ell(\vec{x}) \right\Vert = \left\Vert \vec{x} \right\Vert$ for all $x \in V$. By orthogonal matrix I mean that $L$ has orthonormal columns.

I have that $$ L=\left[ \begin{array}{ccc} [\ell(\vec{q_1})]_Q & \cdots & [\ell(\vec{q_n})]_Q \end{array} \right] $$

but I don't have any idea how to proceed.

share|cite|improve this question
If you are in a real vector space do you mean that $l : V \mapsto V$ is a unitary operator? – user38268 Oct 29 '11 at 0:44
up vote 1 down vote accepted

A proof sketch.

  1. Show that $\langle \ell(\vec x), \ell(\vec {y}) \rangle = \langle \vec x , \vec y \rangle$ holds for all $\vec x, \vec y \in V$. For this step, you may need the following hint: $$\langle \vec x, \vec y \rangle = \frac{1}{2} (\| \vec x+\vec y \|^2 - \| \vec x \|^2 - \| \vec y \|^2) .$$

  2. Show that if the $i^{\rm th}$ column of $L$ is $c_i \in \mathbb R^n$, then $$ \ell(\vec {q_i}) = \sum_{k = 1}^n c_{i,k} \cdot \vec {q_k}. $$

  3. Show that for $1 \leq i, j \leq n$, we have $$ \langle \ell(\vec {q_i}) , \ell(\vec {q_j}) \rangle = \langle c_i, c_j \rangle. $$

  4. Using (1.), what can you say about $\langle \ell(\vec {q_i}) , \ell(\vec {q_j}) \rangle$? What does this mean about $\langle c_i, c_j \rangle$?

share|cite|improve this answer
I assumed real base field, but the complex case shouldn't be much more difficult. – Srivatsan Oct 29 '11 at 0:47
I am making no assumptions about the base field. It could be real, complex, finite, etc. I have shown now that the columns are orthogonal, but how can I show they are still length 1? I only used your first step to show orthogonality, and I don't really follow the other steps. – nullUser Oct 29 '11 at 0:56
Did you understand what $c_i$ means? Where exactly are you stuck in (2.)? – Srivatsan Oct 29 '11 at 0:59
I have shown that $[\ell(\vec{q_i})]_Q = \sum_j \langle \ell(\vec{q_i}),\vec{q_j} \rangle \vec{q_j}$ if that helps. – nullUser Oct 29 '11 at 1:05
@Kb100 the base field must be either real or complex. Otherwise the inner product doesn't make sense. – user12014 Oct 29 '11 at 1:25

Suppose $T: V \to V$ is an isometry, i.e. $\|Tx\| = \|x\|$ for all $x \in V$. Then $T$ is clearly injective, so since $V$ is finite-dimensional, it is also surjective. This means $T^{-1}$ exists. By the polarization identity we have $$\langle Tx, Ty \rangle = \frac{1}{2}(\|T(x+y) \|^2- \|Tx\|^2 - \|Ty\|^2)$$ $$ = \frac{1}{2}(\|x+y \|^2- \|x\|^2 - \|y\|^2) = \langle x, y \rangle$$ which implies $$\langle Tx, y \rangle = \langle x, T^{-1}y\rangle$$ Hence $T^{-1} = T^*$. But the matrix for $T^*$ with respect to an orthornormal basis $Q = \{q_1, ..., q_n\}$ is Hermitian transpose of the matrix for $T$: $$[T]_Q^* = [T^*]_Q = [T^{-1}]_Q = [T]_Q^{-1}$$ So $$[T]_Q[T^*]_Q = [T]_Q[T]^{-1}_Q = I$$ Let $(A)_i$ and $(A)^j$ denote the $i$-th row and $j$-th column of $A$ respectively. Then by the definition of matrix multiplication and Hermititan transpose $$[T]_Q{[T^*]_Q}_{ij} = \langle ([T]_Q)^i, {([T]^*_Q)}_j \rangle = \langle ([T]_Q)^i, {([T]_Q)}^j \rangle = \delta_{ij}$$ So the columns of $[T]_Q$ are orthonormal.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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