Hello this is my first post on here and I have a question about an error in a proof. My Linear Algebra professor was covering Linear Independence, Basis, and Dimension for Vector Spaces. We got to a proof from the textbook and he said the proof is wrong. He said to try and figure out what is wrong with it. I have been trying to figure out what is wrong and after thinking about it, talking with peers, and with my professor I have narrowed it down to a single sentence within the proof. I think that it is a wrong word choice. He said it is very subtle. I was hoping some of you can provide some insight.

THM 6.11: Let $W$ be a subspace of a finite-dimensional vector space $V$. Then:
(a) $W$ is finite-dimensional and $\dim W \le \dim V$.
(b) $\dim W = \dim V$ if and only if $W = V$.

He has a problem with the proof of part (a) which is as follows.

Proof (a) Let $\dim V = n$. If $W = \{\mathbf{0}\}$, then $\dim(W) = 0 \le n = \dim V$. If $W$ is nonzero, then any basis $\mathcal{B}$ for $V$ (containing $n$ vectors) certainly spans $W$, since $W$ is contained in $V$. But $\mathcal{B}$ can be reduced to a basis $\mathcal{B'}$ for $W$ (containing at most $n$ vectors), by Theorem 6.10(f). Hence, $W$ is finite-dimensional and $\dim W \le n = \dim V$.

Theorem 6.10(f) states: Let $V$ be a vector space with $\dim V = n$. Then: (f) Any spanning set for $V$ can be reduced to a basis for $V$.

My professor said that me and my peers were right about the error being in the sentence, "But $\mathcal{B}$ can be reduced to a basis $\mathcal{B'}$ for $W$ (containing at most $n$ vectors), by Theorem 6.10(f)." I believe that the error has something to do with with the word reduced. I asked him and he said that I was on the right track but he didn't give me a clear answer. What is wrong with this proof? More specifically the sentence above? Any help would be much appreciated. (Textbook: Linear Algebra: A Modern Introduction by David Poole 4th edition. The proof is found in section 6.2: page 456)

  • 3
    $\begingroup$ Suggestion: it would be worth while improving your use of MathJax. I was going to edit your post but found there are lots of things which are not done well. For example instead of *W*$=$*V* it is much better to write $W=V$. $\endgroup$ – David May 2 '18 at 5:31
  • 2
    $\begingroup$ I apologize this was my first time using Mathjax. I'll try to work on improving it! Thanks for the suggestion! $\endgroup$ – Glacier May 2 '18 at 14:11
  • $\begingroup$ "But B can be reduced to a basis B′ for W (containting at most n vectors" I take it your professor isn't talking about this error? $\endgroup$ – Acccumulation May 2 '18 at 15:08

The error is the following. Theorem 6.10(f) says that if you have a set that spans $V$, you can remove elements ("reduce") until you get a basis. In the proof of 6.11, the elements in the basis of $V$ may not belong to $W$, so the argument does not apply.

For example, maybe $V=\mathbb R^3$, and you take the canonical basis $\{(1,0,0),(0,1,0),(0,0,1)\}$. Then if $W=\{(x,y,z): x=y=z\}$, none of the three elements in the basis belongs to $W$.

The usual argument is that a basis of $W$ will be linearly independent, and so it will also be linearly independent in $V$, and it follows that $\dim W\leq \dim V$.

  • $\begingroup$ Yep, when you say it that way it is clear. +1 $\endgroup$ – Randall May 2 '18 at 2:43
  • $\begingroup$ Oh, that makes sense! Thank you so much! $\endgroup$ – Glacier May 2 '18 at 2:45

Your professor is correct.

The problem is with the word " reduced"

Let us look at an example.

Let W be the subspace of $\mathbb {R}^2$ spanned by $\{(1,1)\}.$

Let $ B=\{(1,0),(0,1)\}$ be a basis for $\mathbb {R}^2$

Then $B$ spans $W$, but we can not reduce $B$ to a basis for $W$

  • $\begingroup$ Thank you for giving me some more insight! $\endgroup$ – Glacier May 2 '18 at 3:04
  • $\begingroup$ Thanks for your attention and understanding. $\endgroup$ – Mohammad Riazi-Kermani May 2 '18 at 3:17

Maybe there is a difference in use of terminology here but I would say that the problem is with the statement "$B$ spans $W$".

The way I have always seen the term used is as follows.

  • We define ${\rm span}(B)=\{\hbox{linear combinations of vectors in $B$}\}$.
  • Then "$B$ spans $W$" means that ${\rm span}(B)$ equals $W$.

In the proof you have quoted, they appear to be saying that "$B$ spans $W$" means that ${\rm span}(B)$ contains $W$.

So it is a question of using the precise meaning of terminology. BTW, I would not agree with your professor that it is a "very subtle" error, IMHO it is a blatant error :)

  • $\begingroup$ It isn't a terminology issue. Even if we take the weird interpretation that "$B$ spans $W$" means "the linear span of $B$ contains $W$", there is still no reason why $B$ can be reduced to a basis of $W$. The author is simply wrong. $\endgroup$ – user1551 May 2 '18 at 10:56
  • $\begingroup$ My professor didn't have a problem with that part of the proof. It was the sentence afterward that he had a problem with. Thank you for your help! $\endgroup$ – Glacier May 2 '18 at 14:16

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.