Let $V$ be a finite-dimensional inner-product space. Let $U$ be a subspace of $V$ and $(e_1, . . . , e_m)$ an orthonormal basis of $U.$ The orthogonal projection of $V$ onto $U$ is the operator $P_U \in \mathcal L(V)$ defined by $$P_U v = \langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m$$ (a) Use the above definition to prove that $U^{\perp} \subseteq nullP_u$ and $nullP_u \subseteq U^{\perp}$ thus conclude $nullP_u =U^{\perp}$

What I have: $u \in P_uv$ by definition. $(e_1,...e_m)$ is a basis, so $P_uv$ is linearly independent. This implies that if $P_uv=0$ (the $nullP_u$) then $\langle v,e_1 \rangle=\langle v,e_2 \rangle= \space... \space=\langle v,e_m\rangle =0$ But as this is an inner product space, this implies that $v=e_j$ by definiteness. $U^{\perp}=\{v \in V:\langle v,u \rangle=0 \space \forall u\in U \}$ since $u \in P_uv$, $U^{\perp}=\{v \in V:\langle v,\langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m \rangle=0 \space \forall u\in U \}$ I'm stuck on where to go after this though.

b)Use the above definition to prove that $P_u^2 = P_u$

What I have so far

$$P_U(P_Uv) = \langle \langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m,e_1 \rangle e_1 +\dots+\langle \langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m,e_m\rangle e_m = (\langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m)(\langle 1,e_1\rangle e_1 +\dots + \langle 1,e_m\rangle e_m = (\langle v,e_1 \rangle e_1 +\dots+\langle v,e_m\rangle e_m)(\langle e_1, e_1 \rangle +\dots + \langle e_m, e_m\rangle)$$


Here we have $\;U\;$ is a subspace of $\;V\;$ and $\;E\;=\;\{e_{1},e_{2},...e_{m} \}\;$ is an orthonormal basis for $\;U\;.\;$

(a).$\;$Here, $\;\;P_{U}:\;V \to\;U\;\;\text{and}\;\;P_{U}(v)\;=\;\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle e_{i}\;\;$

For any $\;v \in U^{\perp}\;$ we have $\;\langle v,e_{i} \rangle =0\;\;\forall\;i=1,2,..m\;.\;$Hence $\;\;P_{U}(v)\;=\;\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle e_{i}\;=0\;\;$ Hence $\;v \in N\big(P_{U}\big)\;.\;$

Conversely, $\;$for each $\;v \in N\big(P_{U}\big)\;$ we have$\;\;P_{U}(v)\;=\;0\;\;$and hence $\;\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle e_{i}\;=0\;$ which implies $\;\langle v,e_{i} \rangle =0\;\;\forall\;i=1,2,..m\;,\;$by linear independence of $\;E\;.\;$ Hence $\;v \in U^{\perp}\;.\;$

Thus $\; N\big(P_{U}\big)\;=\; U^{\perp}\;.\;$

(b). Note that by orthonormality of $\;E\;$ we get $\;\langle e_{i},e_{j} \rangle \;=\;\delta_{ij}\;=\;\begin{cases}1\;\;\;\text{if}\;\;\;i=j\\0\;\;\;\text{if}\;\;\;i\ne j \end{cases}\;$

For each $\;v \in V\;\;$we have $\;\;P^{2}_{U}(v)\;=\;P_{U}\;\left(\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle e_{i}\right)\;=\;\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle \; P_{U} \big(e_{i}\big)\;$

$\;\;=\;\;\displaystyle \sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle \; \;\displaystyle\sum_{j=1}^{j=m}\;\langle e_{i},e_{j} \rangle e_{j}\;=\;\;\displaystyle \sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle \;\displaystyle\sum_{j=1}^{j=m}\;\;\delta_{ij} e_{j}\; \;=\;\displaystyle\sum_{i=1}^{i=m}\;\langle v,e_{i} \rangle e_{i}\;=\;P_{U}\big(v\big)\;.\;$

Hence $\;\;P^{2}_{U}\;\equiv \;P_{U}\;$


I cannot make sense of most of your arguments:

  • You write "$u\in P_uv$; this makes no sense as $P_Uv$ is a vector.

  • You say "$P_Uv$ is linearly independent"; linear independence is a property of sets of vectors. You could say that a single vector is linear independent, but that's exactly zero new information about the vector: every nonzero vector is linearly independent.

  • Your notation with inner products has a mix of numbers and vectors that again makes no sense. For example, what would "$\langle 1,e_1\rangle$" be?

Let $v\in U^\perp$. Then $\langle v,e_j\rangle=0$ for all $j$ (since $e_j\in U$), and so $PUv=0e_1+\cdots0e_m=0$. So $U^\perp\subset \text{null}\,P_U$. Conversely, if $P_Uv=0$, then $$\langle v,e_1\rangle\,e_1+\cdots+\langle v,e_m\rangle\,e_m=0.$$ As $e_1,\ldots,e_m$ are linearly independent, we get that $\langle v,e_1\rangle=\cdots=\langle v,e_m\rangle=0$. Then $v\in U^\perp$ and so $\text{null}\,P_U\subset U^\perp$.

For the second part, since $P_U$ is linear $$ P_U(P_Uv)P_U(\langle v,e_1\rangle\,e_1+\cdots+\langle v,e_m\rangle\,e_m) =\langle v,e_1\rangle\,P_Ue_1+\cdots+\langle v,e_m\rangle\,P_Ue_m =\langle v,e_1\rangle\,e_1+\cdots+\langle v,e_m\rangle\,e_m=PUv. $$ As this works for all $v$, we get $P_U^2=P_U$.


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.