1

Exercise: Suppose $V$ is finite-dimensional with $\dim V \ge 2$. Prove that there exist $S,T \in L(V, V)$ such that $ST \ne TS$

$L(V, V)$ denotes the set of all linear maps from $V\to V$.

Proof: Define $T : V\to V$ by $T(a_1v_1+\dots+a_nv_n)=2a_1v_1+\dots+a_nv_n$ and define $S: V\to V$ by $S(a_1v_1+\dots+a_nv_n)=(a_1+\dots+a_n)v_1+\dots+(a_1+\dots+a_n)v_n$. Pick non-zero scalars $a_1,\dots,a_n\in F$.

We see that $(ST)v=S(2a_1v_1+\dots+a_nv_n)=(2a_1+\dots+a_n)v_1+\dots+(2a_1+\dots+a_n)v_n$ and

$(TS)v=T((a_1+\dots+a_n)v_1+\dots+(a_1+\dots+a_n)v_n)=(2a_1+\dots+2a_n)v_1+\dots+(a_1+\dots+a_n)v_n$.

This shows that $ST\ne TS$.

Is this solution correct?

Edit: Let $v_1,\dots,v_n$ be a basis of V.

Seeker
  • 3,594
  • A perspective: It suffices to prove the case $n = 2$, that is, the task is to find two $2 \times 2$ matrices that do not commute. – Mason Aug 04 '22 at 04:21
  • This proof is correct. – Sarvesh Ravichandran Iyer Aug 04 '22 at 04:26
  • @Mason This was a question in the section before talking about matrices so I was trying to find a proof that just uses the machinery developed so far. Also, How would just proving the case for $n=2$ work for every $n$? – Seeker Aug 04 '22 at 04:30
  • 1
    One suggestion: We may define S and T by their images on the basis ${ v_i}$, $i=1,2, \dots n$. For example, $T(v_i)=v_1$ for all $i$. This may lead to simpler example. – Li Kwok Keung Aug 04 '22 at 05:45
  • @LiKwokKeung Yeah I agree. I just saw a solution using exactly what you suggested and that is much simpler. – Seeker Aug 04 '22 at 06:12
  • 1
    The logic is a bit confusing to me.

    The statement to be proved is: $$\forall V, \mathrm{dim} V \ge 2 \implies ( \exists S, T ; \mathrm{s.t.} ;ST \neq TS)$$

    To prove that it is true, one has to consider all finite dimensional $V$, not just dim$V$=2.

    The negation is an existential statement, namely $$\exists V, \mathrm{dim} V \ge 2 ; \mathrm{and} ; ST=TS; \forall ; S, T$$

    To prove the negation is true, an example suffices.

    But we are NOT going to prove that the negation is true. Right?

    – Li Kwok Keung Aug 04 '22 at 09:30
  • @LiKwokKeung Yeah I agree with you. I was also a bit confused by the comment. I also asked another question regarding it and got the same answer. – Seeker Aug 04 '22 at 10:15
  • @RaoulLuqué That is not the statement that has to be disproved. Have a look at the comment by Li Kwock Keung – Seeker Aug 04 '22 at 10:17
  • @Seeker How is my proof? – user264745 Aug 04 '22 at 10:20
  • @user264745 The proof is correct in the sense that for any vector space with dimension $n\ge 2$, the proof still works as the basis would have at least 2 vectors in the basis. – Seeker Aug 04 '22 at 10:25
  • 1
    @Seeker Indeed, I missunderstood the question. Sorry for my missleading comment, I am going to delete it in order not to confuse anymore people – Raoul Luqué Aug 04 '22 at 14:04
  • @RaoulLuqué That's all good. I had the same thing happen to me yesterday and that is what got me to scrutinise the statement so much. – Seeker Aug 05 '22 at 00:33

1 Answers1

1

Your proof is correct but not efficient. Since $\mathrm{dim}(V)=2$, we have $\exists B=\{\alpha_1,\alpha_2\}$ basis of $V$. Linear map is uniquely determined by it’s action on basis $B$. It’s easy to check $\exists !S,T\in L(V,V)$ such that $S(\alpha_1)=\alpha_2$, $S(\alpha_2)=\alpha_1$ and $T(\alpha_1)=\alpha_1$, $T(\alpha_2)=\alpha_1$. Then $S\circ T(\alpha_1)$ $=S(T(\alpha_1))$ $=S(\alpha_1)$ $=\alpha_2$ and $T\circ S(\alpha_1)$ $=T(S(\alpha_1))$ $=T(\alpha_2)$ $=\alpha_1$. Thus $S\circ T(\alpha_1)\neq T\circ S(\alpha_1)$. Hence $S\circ T\neq T\circ S$.

user264745
  • 4,143
  • 1
    I saw the sane proof shortly after I posted my answer. It also works for when $\dim V \ge 2$ as the basis would still consist of two vectors. I just couldn’t see the importance of the result that a linear map can take a vector in its basis to an arbitrary vector in another vector space. But today I saw the importance of it through this exercise and one other one. – Seeker Aug 04 '22 at 11:37
  • 1
    Btw, I solved this problem independently. I have solved similar problems so many times that I knew I have to use hypothesis in that sequence to get desired result. I also check for $S(\alpha_i)=\alpha_i$ and $T(\alpha_i)=\alpha_i$, for all $i\in {1,2}$. But that didn’t work. So I did some try and error before reaching to desired result. The result you’re taking about is extremely important and use countless times, IMO. Most problems which requires to show existence of a linear map with certain property, make use of that result. – user264745 Aug 04 '22 at 11:57