7
$\begingroup$

Where can I find a proof of the following general Steinitz exchange lemma:

Let $B$ be a basis of a vector space $V$, and $L\subset V$ be linearly independent. Then there is an injection $j:L\rightarrow B$ such that $L\cup(B\setminus j(L))$ is a disjoint union and a basis of the vector space $V$.

Or can someone suggest a proof of this result? Probably using Zorn's lemma.

Thank you!

$\endgroup$
  • 1
    $\begingroup$ @Censi LI: The axiom of choice tag is for questions asking specifically about the role of the axiom of choice in a proof, or what sort of statement are consistently true without the axiom. This question does not seem to fit into that tag. Similarly, this is not a set theory question. $\endgroup$ – Asaf Karagila Jun 2 '15 at 21:36
  • $\begingroup$ @AsafKaragila I'm so sorry... $\endgroup$ – Censi LI Jun 2 '15 at 21:39
  • 1
    $\begingroup$ @Censi LI: No need to apologize. Just for future reference. $\endgroup$ – Asaf Karagila Jun 2 '15 at 21:40
  • $\begingroup$ Here you can find a blogpost about a very general replacement theorem of Steinitz for modules. $\endgroup$ – Pedro Tamaroff Jul 1 '17 at 21:19
2
+200
$\begingroup$

Let $W$ denote the subspace generated by $L$, $\mathcal B$ denote the set of subsets of $B$ whose images in $V/W$ are linearly independent. There is an obvious partial order on $\mathcal B$, namely the one induced by inclusion. Apply Zorn's Lemma to obtain a maximal element of $\mathcal B$, denoted by $C$, then the disjoint union $C\cup L$ forms a basis of $V$, just as $B=C\cup(B\setminus C)$ do. So there must be a bijection $j:L\xrightarrow{\sim}B\setminus C$. This $j$ will fulfill the requirement.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ It is not true that every chain in $\mathcal{L}$ has an upper bound--see my answer at math.stackexchange.com/questions/2341188/…. The problem is that at each step of the chain you are removing new elements of your spanning set since $j(L')$ is getting larger, and the limit may end up not spanning. $\endgroup$ – Eric Wofsey Jul 1 '17 at 17:06
  • $\begingroup$ @EricWofsey Thank you, I've reworked the proof. $\endgroup$ – Censi LI Jul 1 '17 at 21:15
0
$\begingroup$

Choose (by using Zorn's Lemma) a subset $B'$ of $B$ maximal with the property that $\langle L\rangle\cap \langle B'\rangle=0$. Then $\langle L\rangle+\langle B'\rangle=V$: if $b\in B-B'$ then $\langle L\rangle\cap \langle B'\cup\{b\}\rangle\ne0$ hence there are $l_i\in L$, and $b_i'\in B'$ such that $\sum\alpha_il_i=\sum_i\beta_ib_i'+\beta b$ with $\beta\ne0$, so $b\in \langle L\rangle+\langle B'\rangle$.

Now use that any two bases of a vector space are equipotent. Let $V'=\langle B'\rangle$. The quotient vector space $V/V'$ has the following bases: $L$ and $B\setminus B'$. Then there is a bijection $f:L\to B\setminus B'$ which composed by the inclusion $i:B\setminus B'\to B$ gives us the desired injection, that is, $j=i\circ f$.

| cite | improve this answer | |
$\endgroup$

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.