I am trying to prove this theorem in my book. Because it is provided without proof, please let me know what you think!

$\mathbf{Theorem:}$ Let V be a finite , $n$ -dimensional vector space and let U be a subspace of V.

Define, $T: V \rightarrow V$ as $$ T(v)=proj_U(v)$$ ( ie, the orthogonal projection)

Then the following are true;

i) T is a linear transformation

ii) $Im(T)=U$ and $kerT=U^{\perp} $ where $U^{\perp}$ represents the orthogonal complement.

iii) $Dim(U)+Dim(U^{\perp})= n$


(i) Let $v , w \in V$ , $k \in \mathbb R$, and let $\mathbb B=\{u_1,…u_n\}$ be an orthogonal basis for V.

then we have $$T(v)= \frac{\langle v , u_1 \rangle}{\langle u_1, u_1 \rangle}u_1 +…+ \frac{\langle v, u_n \rangle}{\langle u_n, u_n \rangle}u_n$$

and $$T(w)=\frac{\langle v , w_1 \rangle}{\langle w_1, w_1 \rangle}w_1 +…+ \frac{\langle v, w_n \rangle}{\langle w_n, w_n \rangle}w_n$$

then the rest follows by invoking the inner product properties of linearity. I have it written but it will take my a long time to type that out , so for that I would just be wondering if that is the correct approach?


Must show that $Im(T) \subseteq U$ and that $U \subseteq Im(T)$

Let $w \in Im(T)$ then there exists a$ v \in V$ such that $T(v)=w$

and we have seen that this would imply w is just the projection i.e. a linear combination of vectors $u_i$ and for $U \subseteq Im(T)$ we can simply take the projection of $u$ onto itself. Thus, showing $Im(T)=U$

Now suppose $w \in U^{\perp}$ then $$T(w)=\frac{\langle w,u_1 \rangle}{\langle u_1, u_1 \rangle }u_1+…+\frac{\langle w , u_n \rangle}{\langle u_n, u_n \rangle}u_n=0$$ ie $w \in Ker(T)$ and now now suppose $w \ in Ker(T)$ then $T(w)=0$ , using the fact that for any $w \in V$ we have $w-T(w) \in U^{\perp}$ then this gives $w-0 \in U^{\perp}$ that is $w \in U^{\perp}$ i.e. $Ker(T) \subseteq U^{\perp}$

(iii) This follows directly from the rank nullity theorem, which states $dim(ImT)+dim(KerT)=dim(V)$

Let me know what you all think, any suggestions etc, thanks!

  • $\begingroup$ What's $\phantom{x}^\perp$? $\endgroup$ – GFauxPas Apr 20 '15 at 20:57
  • $\begingroup$ What is $proj_U$? And how can you assume $V$ has an orthogonal basis? $\endgroup$ – Hagen von Eitzen Apr 20 '15 at 21:03
  • $\begingroup$ @GFauxPas the orthogonal complement $\endgroup$ – Ben Grossmann Apr 20 '15 at 21:04
  • $\begingroup$ @HagenvonEitzen the (orthogonal) projection onto $U$ $\endgroup$ – Ben Grossmann Apr 20 '15 at 21:04
  • $\begingroup$ @HagenvonEitzen It's usually given as a theorem that every inner product space has an orthogonal basis. The proof is essentially given by the Gram Schmidt algorithm. $\endgroup$ – Ben Grossmann Apr 20 '15 at 21:05

I think you meant $\mathbb{B}=\{u_1,…,u_r\}$ is an orthogonal basis for $U$. You can extend it, by adding vectors in the corresponding orthogonal complements, to an orthogonal basis $\{u_1,…,u_r,v_{r+1},…,v_n\}$ for $V$. There is no need to normalize the vectors. Now, if $v=\sum <v,u_i>u_i+\sum <v,v_i>v_i$, then $T(v)=\sum<v,u_i>u_i$. From that explicit formula everything is simpler, for example, $v\in Ker(T) \Leftrightarrow \;<v,u_i>=0,$ for all $i=1,…,r$, which is equivalent to $v \in U^\bot$.


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.