1
$\begingroup$

Suppose $W$ is a subspace of the finite-dimensional vector space $V$ and $\dim W = \dim V=n$

Let $\{\overrightarrow w_1,\ldots,\overrightarrow w_n\}$ be a basis for $W$ and $\{\overrightarrow v_1,\ldots,\overrightarrow v_n\}$ be a basis for $V$.

The set $\{\overrightarrow w_1,\ldots,\overrightarrow w_n,\overrightarrow v_1,\ldots,\overrightarrow v_n\}$ still spans $W$. Since $W$ is a subspace of $V$, $W\subset V$ and all the $w_i$'s can be written in terms of $\{ \overrightarrow v_{1},\ldots,\overrightarrow v_n\}$.

$\therefore$ We can remove one by one all the $w_i$'s from the set $\{\overrightarrow w_{1},\ldots,\overrightarrow w_n,\overrightarrow v_1,\ldots,\overrightarrow v_n\}$ while still having it span $W$, which leaves us with the set $\{\overrightarrow v_1,\ldots,\overrightarrow v_n\}$. We know it to be linearly independent since it is a basis for $V$. Therefore the basis of $V$ is also a basis of $W$. From there on, it is trivial to prove double inclusion $W\subset V$ and $V\subset W$ which gives us $W=V$.

Is my approach correct?

$\endgroup$
  • $\begingroup$ Yes, this is correct! $\endgroup$ – Jon Warneke Mar 27 '16 at 23:22
  • $\begingroup$ Easier: every linearly independent set in $V$ can be extended to a basis. A basis of $W$ is linearly independent. Can you add vectors to it getting again a linearly independent set? $\endgroup$ – egreg Mar 27 '16 at 23:23
  • $\begingroup$ I think it is. You could also go with this route: There exists a linear mapping that is a bijection between the basis of $W$ and $V$. This mapping is an isomorphism, and then you can use it to prove that $V\subset W$. $\endgroup$ – Marra Mar 27 '16 at 23:24
  • 1
    $\begingroup$ Please consider using the LaTeX command \dim to get $\dim$ instead $dim$. $\endgroup$ – Fly by Night Mar 27 '16 at 23:39
  • 1
    $\begingroup$ I do not think this argument is correct, since you seem to be assuming that the $\vec{v_i}$ are in W when you claim that the set of $\vec{w_j}$ and $\vec{v_i}$ span W. The fact that the $\vec{v_i}$ are a basis for V implies that every vector in W can be written as a linear combination of the $\vec{v_i}$, which seems to be what you are showing. $\endgroup$ – user84413 Mar 27 '16 at 23:45
3
$\begingroup$

Your approach is incorrect: the set $\{\overrightarrow w_1,\ldots,\overrightarrow w_n,\overrightarrow v_1,\ldots,\overrightarrow v_n\}$ spans $V$ and saying that it spans $W$ is the same as assuming $W=V$.

You should rather consider a basis $\{\overrightarrow w_1,\ldots,\overrightarrow w_n\}$ of $W$ and assume, by contradiction, that $W\ne V$. Then there exists $\vec{v}\in V$ with $\vec{v}\notin W$. It's easy to prove that $\{\overrightarrow w_1,\ldots,\overrightarrow w_n,\overrightarrow v\}$ is linearly independent and now this is a contradiction, because we found a linearly independent set in $V$ with more elements than $\dim V$.

$\endgroup$
  • $\begingroup$ Why am I assuming that $W=V$? I just added vectors to the basis of $W$, the spanning of W should not be affected. We can then remove the dependent vectors one by one. $\endgroup$ – Omrane Mar 28 '16 at 12:10
  • $\begingroup$ @Sadem You also add vectors that span $V$, so the span is $V$, not $W$. Your argument would work exactly the same if $w\in\mathbb{R}^2$ and $\{v_1,v_2\}$ is a basis of $\mathbb{R}^2$: from the set $\{w,v_1,v_2\}$ you can surely remove $w$. $\endgroup$ – egreg Mar 28 '16 at 12:14
1
$\begingroup$

Your approach is fine and correct.

The easiest way to prove this is to notice that $W$ is a subspace of $V$, thus it is a subset of $V$. Thus, if $dimV = n$, and we have $n$ linearly independent vectors in $W$ that span $W$, these must also be linearly independent vectors in $V$. But then we have $n$ of them so these must span $V$ as well! So the vectors span $W$ AND $V$, so they are equal.

$\endgroup$
  • 1
    $\begingroup$ The approach is not correct. While is is true that $\{w_1, \ldots, w_n, v_1, \ldots, v_n\}$ spans $W$, this is only because $W$ and $V$ are equal. Sadem's explanation on the other hand is completely wrong. $\endgroup$ – Dominik Mar 28 '16 at 8:26
  • $\begingroup$ @Dominik, {$\overrightarrow w_{1},...,\overrightarrow w_{n}$} spans W, therefore adding any vector to the set will not affect the spanning of W, if anything it will make it larger. $\endgroup$ – Omrane Mar 28 '16 at 12:08
  • $\begingroup$ @Sadem: Your definition of spanning is colloquial, not the mathematical definition. A set of vectors is said to span a vector space, if the span of these vectors is precisely the space - not if it is bigger than the space. Note that your argumentation doesn't even require at any point that the dimensions of $W$ and $V$ are equal. $\endgroup$ – Dominik Mar 28 '16 at 12:12
  • $\begingroup$ It actually does, suppose if $W$ had a lower dimension than $V$, we would end up with a linearly independent set of $n$ elements for $W$ while its dimension is less than $n$. We could therefore not declare it a basis of $W$ which does not allow us to solve the problem. You are right about how I see spanning though, I don't know how to phrase it in another way though. $\endgroup$ – Omrane Mar 28 '16 at 12:16
  • 1
    $\begingroup$ This doesn't correct your argumentation. In your comment, you've just shown that $\operatorname{dim} W \ne \operatorname{dim} V$ implies $W \ne V$. At no point does your argumentation use the fact that the dimensions are the same [cf. the comments under egreg's answer]. Your proof is quite simply false. $\endgroup$ – Dominik Mar 28 '16 at 15:19
1
$\begingroup$

Your proof is incorrect. You first choose a colloquial understanding of the word "spanning" and at a later point the mathematically correct understanding [which changes the meaning of the word!].

You say that $\{w_1, \ldots, w_n, v_1, \ldots, v_n\}$ spans $W$, but you can't deduce this directly. What you mean is that $W$ is a subset of $\operatorname{span}(\{w_1, \ldots, w_n, v_1, \ldots, v_n\})$. Since the $w_i$ can be represented as linear combinations of the $v_i$, you can succesively remove the $w_i$ without changing the span, i.e. $$\operatorname{span}(\{w_1, \ldots, w_n, v_1, \ldots, v_n\}) = \operatorname{span}(\{v_1, \ldots, v_n\})$$

Now at this point in your argumentation, you make the switch from the colloquial term "spanning" to the mathematically rigorous by saying that $\{v_1, \ldots, v_n\}$ spans $W$ and is therefore a basis for $W$. But this is incorrect, as you've only shown that $\{v_1, \ldots, v_n\}$ "overspans" $W$, i.e. $W \subset \operatorname{span}(\{v_1, \ldots, v_n\})$. But this inclusion is trivial, since the span of the $v_i$ is the whole of $V$.

Note that you've at no point in your argumentation actually used the fact that the dimensions of $V$ and $W$ are the same. Simply go throgh the proof and try to point at the part where you've used this assumption.

$\endgroup$

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.