Let $r\leq n$ be a natural number and let $\{v_1,v_2,\dots ,v_r\}$ and $\{w_1,w_2,\dots ,w_r\}$ be two linearly independent subsets of $\mathbb{R}^n$ such that $\langle v_i,v_j \rangle = \langle w_i,w_j \rangle \ \forall \ 1\leq i,j \leq r$, where $\langle ,\rangle$ denotes the standard inner product on $\mathbb{R}^n$. Prove that there exists an orthogonal operator $T$ on $\mathbb{R}^n$ such that $T(v_i)=w_i$ for all $1\leq i \leq r$.

An orthogonal mapping preserves inner products, and it is also known that $T$ is orthogonal on $\mathbb{R}^n$ if $\langle T(\alpha),T(\beta)\rangle = \langle \alpha ,\beta \rangle$ for all $\alpha,\beta \in \mathbb{R}^n$. But the question here demands to be proved that the existence of such an orthogonal map. So do I need to find an example of such map or is there another way to prove that in general? Any help will be appreciated.


You define $Tv_j=w_j$ for all $j=1,\ldots,r$, and since both sets are linearly independent, $T$ gets defined by linearity on their span.

But now you have the problem that maybe $r<n$. Expanding both sets to bases does not work right away, because you have no way to preserve the inner product relations. So what you do is take $\{v_{r+1},\ldots,v_n\}^\perp$ to be an orthonormal basis of $\{v_1,\ldots,v_r\}^\perp$, and $\{w_{r+1},\ldots,w_n\}$ an orthonormal basis of $\{w_1,\ldots,w_r\}^\perp$. Then define $Tv_j=w_j$ for all $j=1,\ldots, n$. The inner product relations are preserved because $\{v_j,v_k\rangle=\delta_{kj}=\langle w_j,w_k\rangle$ for $j,k>r$, and when one index is below $r$ and the other above, the inner product is zero by the orthogonality.


Let the subspace

$V = \text{span}(\{ v_1, v_2, \ldots, v_r \}) = \text{span}(\{ w_1, w_2, \ldots, w_r \}); \tag 1$

we define $T$ on $V$ by setting

$Tv_k = w_k, \; 1 \le k \le r, \tag 2$

and extending $T$ by linearity to all of $V$, viz:

$T\left ( \displaystyle \sum_1^r \alpha_i v_i \right ) = \displaystyle\sum_1^r \alpha_i Tv_i = \sum_1^r \alpha_i w_i; \tag 3$

that such an extension is possible follows from the linear independence of the $v_i$ and the $w_i$ as is well-known.


$\alpha, \beta \in V \tag 4$

we may write

$\alpha = \displaystyle \sum_1^r \alpha_i v_i, \; \beta = \sum_1^r \beta_j v_j, \tag 5$


$\left \langle T\alpha, T\beta \right \rangle = \left \langle T \left (\displaystyle \sum_1^r \alpha_i v_i \right), T \left (\displaystyle \sum_1^r \beta_j v_j \right) \right \rangle = \left \langle \displaystyle \sum_1^r \alpha_i Tv_i, \displaystyle \sum_1^r \beta_j Tv_j \right \rangle = \displaystyle \sum_{i, j = 1}^r \alpha_i \beta_j \langle Tv_i, Tv_j \rangle = \sum_{i, j = 1}^r \alpha_i \beta_j \langle w_i, w_j \rangle = \sum_{i, j = 1}^r \alpha_i \beta_j \langle v_i, v_j \rangle$ $= \left \langle \displaystyle \sum_1^r \alpha_i v_i, \displaystyle \sum_1^r \beta_j v_j \right \rangle = \langle \alpha, \beta \rangle, \tag 6$

which proves that $T:V \to V$ is orthogonal.

In the event that $r < n$, we extend $T$ from $V$ to all of $\Bbb R^n$ as follows: let $W \subset \Bbb R^n$ be the orthogonal compliment of $V$, $W = V^\bot$; then we have

$\Bbb R^n = V \oplus W; \tag 7$

we observe that any $x \in \Bbb R^n$ may be represented uniquely as

$x = v + w, \; v \in V, w \in W; \tag8$

existence of such a decomposition follows directly from (7), i.e. from the fact that $W = V^\bot$; to see uniqueness, suppose

$x = v_1 + w_1 = v_2 + w_2, \; v_1, v_2 \in V, w_1, w_2 \in W; \tag 9$


$v_1 - v_2 = w_2 - w_1, \tag{10}$

which forces

$v_1 - v_2 = 0 = w_2 - w_1, \tag{11}$


$v_1 - v_2 \in V, \; w_1 - w_2 \in W = V^\bot, \tag{12}$

and hence

$\Vert v_1 - v_2 \Vert^2 = \langle v_1 - v_2, v_1 - v_2 \rangle = \langle v_1 - v_2, w_2 - w_1 \rangle = 0, \tag{13}$

implying $v_1 = v_2$, and thus $w_1 = w_2$ via (11); so the decomposition (8) is unique; this uniqueness means we may unabiguously define an extension $T_E$ of $T$ from $V$ to $V \oplus W = \Bbb R^n$ by

$T_E(v + w) = Tv + w, \; v \in V, w \in W; \tag{14}$

that is,

$T_E = T \oplus I: V \oplus W \to V \oplus W; \tag{15}$

we can formally show $T_E$ is orthogonal on all of $\Bbb R^n$ if we observe that, for $v_1, v_2 \in V$, $w_1, w_2 \in W$,

$\langle v_1 + w_1, v_2 + w_2 \rangle$ $= \langle v_1, v_2 \rangle + \langle v_1, w_2 \rangle + \langle w_1, v_2 \rangle + \langle w_1, w_2 \rangle = \langle v_1, v_2 \rangle + \langle w_1, w_2 \rangle; \tag{16}$


$\langle T_E(v_1 + w_1), T_E(v_2 + w_2) \rangle = \langle (T \oplus I)(v_1 + w_1), (T \oplus I)(v_2 + w_2) \rangle$ $= \langle Tv_1 + w_1, Tv_2 + w_2 \rangle = \langle Tv_1, Tv_2 \rangle + \langle w_1, w_2 \rangle$ $= \langle v_1, v_2 \rangle + \langle w_1, w_2 \rangle = \langle v_1 + w_1, v_2 + w_2 \rangle, \tag{17}$

which shows that $T_E = T \oplus I$ is orthogonal on $\Bbb R^n$.


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.