Take the 2-minute tour ×
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.

I have a question: If $P$ and $Q$ are projections. Does it follow that $P$ and $Q$ commute? If yes, how can we prove it? If not, what is the relationship between $PQ$ and $QP$?

share|improve this question

2 Answers 2

up vote 2 down vote accepted

In $\Bbb R^2$ let $P$ be the projection to the $x$-axis and $Q$ the projection to the subspace generated by $\langle 1,1\rangle$. Let $v=\langle 1,1\rangle$. What are $P(Q(v))$ and $Q(P(v))$? Are they the same?

share|improve this answer
    
following your example, they are not the same –  M.Krov Apr 12 '12 at 20:25
    
@M.Krov: That’s correct. –  Brian M. Scott Apr 12 '12 at 20:31
1  
@M.Krov: Let $V=\operatorname{Im}(P),W=\operatorname{Im}(Q)$. Let $U$ be the orthogonal complement of $V\cap W$. You want $V\cap U$ and $W\cap U$ to be orthogonal. –  Brian M. Scott Apr 12 '12 at 23:30

Hint. The image of $PQ$ lies in $\mathrm{Im}(P)$, and the image of $QP$ lies in $\mathrm{Im}(Q)$. So we need $P(\mathrm{Im}(Q))\subseteq \mathrm{Im}(Q)\cap\mathrm{Im}(P)$, and $Q(\mathrm{Im}(P)) \subseteq \mathrm{Im}(P)\cap\mathrm{Im}(Q)$. So we need $\mathrm{Im}(Q)$ to be $P$-invariant, and $\mathrm{Im}(P)$ to be $Q$-invariant. When does that occur?


To answer the question you had asked in comments and since deleted, under what conditions do we have $QP=PQ$?

As noted in the hint, $\mathrm{Im}(P)$ must be $Q$-invariant, and $\mathrm{Im}(Q)$ must be $P$-invariant. A similar computation shows that $\mathrm{ker}(P)$ must be $Q$-invariant, and that $\mathrm{ker}(Q)$ must be $P$-invariant. The key observation to make is:

Proposition. Let $V$ be a vector space, and let $T$ be a projection on $V$ (that is, $T$ is a linear transformation $V\to V$ and $T^2=T$). If $W$ is a subspace of $V$ such that $T(W)\subseteq W$ (that is, if $W$ is $T$-invariant), then $W=(W\cap\mathrm{Im}(T)) + (W\cap \mathrm{ker}(T))$.

Proof. Let $w\in W$. Then $T(w)\in W$, hence $w-T(w)\in W$. Since $w-T(w)\in W\cap\mathrm{ker}(T)$ and $T(w)\in W\cap \mathrm{Im}(T)$, and $w=T(w)+(w-T(w))$, we conclude that $W\subseteq (W\cap\mathrm{Im}(T))+(W\cap\mathrm{ker}(T)) \subseteq W$, giving equality. $\Box$

In fact, the converse also holds (prove it).

Use the above to conclude that $PQ=QP$ if and only if $$V=(\mathrm{Im}(P)\cap\mathrm{Im}(Q)) +(\mathrm{Im}(P)\cap\mathrm{ker}(Q))+(\mathrm{ker}(P)\cap\mathrm{Im}(Q)) + (\mathrm{ker}(P)\cap\mathrm{ker}(Q))$$ in which case, the sum is a direct sum.

share|improve this answer
    
Isn't the last sum always a direct sum? –  user26770 Apr 13 '12 at 0:40
1  
@Boston: Yes, because $\mathrm{Im}(P)\cap\mathrm{ker}(P) = \mathrm{Im}(Q)\cap\mathrm{Ker}(Q) = \{0\}$. –  Arturo Magidin Apr 13 '12 at 2:01

Your Answer

 
discard

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.