0
$\begingroup$

I have found this question in an old exam for linear algebra. Specifically the last question here: http://math.ucdenver.edu/~langou/5718/Spring2015-MATH5718-exam1.pdf . I can however not convince myself of the truth of this result. Would not

$ V = R^3 $, U := {(x,0,0) |$x \in R $}, W := {(0,0,z) | z$\in R$}

be a counterexample of this claim?

The only explanation i have for this is difference in notation where i am used to $\oplus$ referring specifically to a direct sum whereas they may use it as a normal plus sign. Then however would the result not be trivial by choosing S = V?

$\endgroup$
  • $\begingroup$ No, it is not a counterexample. Consider $S:=\lbrace (x,y,x) | (x,y) \in R^2 \rbrace$. And yes, $\oplus$ here means direct sum not just any sum. $\endgroup$ – Ewan Delanoy Sep 16 '16 at 11:37
2
$\begingroup$

It is not a counterexample, since you can take $$ S = \{(x,x+z,z) \mid y\in\mathbb{R}\}. $$ Let us show that this subspace satisfies $U\oplus S=\mathbb{R}^{3}$ (the other is similar).

A basis for $U$ consists of the vector $(1,0,0)$. A basis for $S$ consists of $(1,1,0)$ and $(0,1,1)$.

You can easily chech that $A=(1,0,0)$, $B=(1,1,0)$ and $C=(0,1,1)$ forms a basis for $\mathbb{R}^{3}$. This can be done by showing that $$ (0,1,0) = B - A, \qquad (0,0,1) = A+C-B. $$ Therefore, the standard basis is generated by $A,B,C$, proving that any element of $\mathbb{R}^{3}$ is generated by $A$, $B$, $C$, i.e., is a basis.

$\endgroup$
1
$\begingroup$

The assertion in title is true in the more general context of finite dimensional spaces.

$\Rightarrow$ is true because $\dim V= \dim U+\dim S=\dim W+\dim S$.

$\Leftarrow$: if $\dim U=\dim W<\dim V$, we have to find a subspace $S$ such that $\;U\oplus S=W\oplus S=V$.

By the avoidance lemma for vector spaces, one can find a vector $v_1$ not in $U\cup W$. If $\DeclareMathOperator\codim{codim}\codim U=\codim W=1$, setting $S=\langle v_1\rangle$ solves the problem.

If $\codim U=\codim W=r>1$, set $U_1=U\oplus\langle v_1\rangle$, $W_1=W\oplus\langle v_1\rangle$, an easy induction on $r$ shows one can find a subspace $S_1$ such that $\;U_1\oplus S_1=W_1\oplus S_1=V$. Then $S=S_1\oplus\langle v_1\rangle$ solves the problem.

$\endgroup$
0
$\begingroup$

To prove the statement note that $W\neq U\cup W$ (set theoretic union). To see this we have $U \not \subseteq W$ since they have the same dimension. (assuming they are not equal in which case its easy). Similarly $W\not \subseteq U$ so let $u\in U-W$ and $w\in W-U$ then $u+w\not \in U\cup W$. So chose $s_1\in V-(U\cup W)$. Now repeat the argument for $U\oplus <s_1>$ and $W\oplus <s_1>$.

$\endgroup$

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.