Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

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

This question is a follow-up/extension to this question. Suppose $P$ is an orthogonal operator on a finite dimensional inner product space $V$. By definition, this means that $$ \langle Pu, Pv \rangle = \langle u, v \rangle $$ for all $u,v \in V$. I want to show that $P^TP = I$.

By definition of the transpose,

$$ \langle u, v \rangle = \langle Pu, Pv \rangle = \langle u, P^T(Pv) \rangle $$

and by bilinearity of the inner product it follows that $$ \langle u , (I - P^TP)v \rangle = 0 $$ If $v \neq 0 \neq u$, I want to use the fact that the inner product is nondegenerate to conclude that $I - P^TP = 0$ which would prove the claim. However, I'm not so sure that this applies here for the nondegenerate criterion is a statement about all vectors and not just a particular vector. That is, I don't believe it necessarily true that $\langle x, y \rangle = 0$ and $x \neq 0 \implies y = 0$.

On the other hand, without invoking that the inner product is nondegenerate one can see from inspection that $P^TP = I$ satisfies $$ \langle u, v \rangle = \langle u , (P^TP)v \rangle $$
Is this enough?

share|cite|improve this question
up vote 5 down vote accepted

You're almost there. Notice that in fact, given $v \in V$ $$\langle u, (I - P^TP)v \rangle = 0$$ for all $u \in V$. In particular, setting $u = (I - P^TP)v$ gives

$$\|(I - P^TP)v\|^2 = 0$$ which implies $$(I - P^TP)v = 0$$

share|cite|improve this answer
Beaten by 2 seconds... :-) – Henning Makholm Jul 2 '12 at 1:26

Since you know $\langle u,(I-P^tP)v\rangle=0$, for arbitrary $u$ and $v$, you can fix any $v$ and then chose $u$ to be $(I-P^tP)v$. Then you know $\langle (I-P^tP)v,(I-P^tP)v\rangle=0$ for every $v$, and because the inner product is positive definite, that must mean $(I-P^tP)v=0$ for all $v$...

share|cite|improve this answer

In general it is indeed true that if $\langle u, w \rangle=\langle v,w\rangle$ for all $w$ then $u$ must equal $v$. The other answers show one way of seeing this, and here is another: take $w$ to be each element of an orthonormal basis in turn, say $e_1\dots e_n$, then $\langle u, e_i \rangle$ is the $i^\mathrm{th}$ component of $u$ with respect to this basis. Hence, corresponding components of $u$ and $v$ are equal, so of course $u$ and $v$ are equal.

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.