In Axler's Linear Algebra Done Right there is the posit that every linearly independent list of vectors in a finite-dimensional vector space can be extended to a basis of the vector space. This seems fine, intuitively, but I don't understand his proof. He states:

Suppose a linear independent list, $u_1$ , ... , $u_m$ exists in vector space V. V also has a basis $w_1$ , ... , $w_m$.

1) If one creates a list $u_1$ , ... , $u_m$ , $w_1$ , ... , $w_m$, that means you now have another list that spans V.

2) Further, he says that if you remove any $w_j$ elements within the new list, you're able to create a basis as long as you don't remove any $u_j$ elements, as that original list of $u_j$ elements is linearly independent.

My questions are regarding 1) and 2). For 1): How is this new list a span of V? And for 2): Why does he emphasize that you need to remove $w_j$ elements and not $u_j$? Since the $w_j$ elements were originally from a basis, aren't they linearly independent, also?*


Consider the list $u_1,\dots,u_m,w_1,\dots,w_n$ (I'll call it the big list), where $u_1,\dots,u_m$ is linearly independent (I'll call it the $u$-list) and $w_1,\dots,w_n$ is a basis.

Clearly the big list is a spanning set. If it is already linearly independent, then you're done (actually the $u$-list would be empty, but it's irrelevant).

Otherwise the big list is linearly dependent, so we can write $$ \alpha_1u_1+\dots+\alpha_mu_m+\beta_1w_1+\dots+\beta_nw_n=0 $$ where not all the coefficients are zero. Actually one among $\beta_1,\dots,\beta_n$ must be non zero, otherwise we'd have $\alpha_1u_1+\dots+\alpha_mu_m=0$ and linear independence of the $u$-list would imply also $\alpha_1=\dots=\alpha_m=0$.

Without loss of generality, we can assume $\beta_n\ne0$ (reordering a basis still gives a basis). Hence also the list $u_1,\dots,u_m,w_1,\dots,w_{n-1}$ is a spanning set.

Proceed in the same way until the set you have is linearly independent, which must happen in at most $n$ steps. Since at every stage the list after the removal is still a spanning set, the final list will be linearly independent and a spanning set, hence a basis.

Your questions.

  1. Why is the big list a spanning set? Because every list containing a spanning set is a spanning set as well.

  2. Why emphasizing that elements of the $w$-list are removed? Because if we remove items from the $u$-list we won't get a list that extends the $u$-list.

Note that in the context of your second question, you cannot remove any item in the $w$-list, but you have to choose one in a suitable way, as outlined above.

  • $\begingroup$ Perfect answer. Thanks. $\endgroup$ – Shocked Jul 12 '17 at 15:07
  • $\begingroup$ @Shocked If you are satisfied with this answer, then don't forget to accept it. $\endgroup$ – Our Jul 12 '17 at 15:26

For the question $1-)$: Let $x\in V$, so $x= \sum_v \lambda_v w_v$ and also $x= \sum_v \lambda_v w_v + \sum_k 0 \cdot u_k$, so the list spans $V.$

For the question $2-)$: Since you want to create a basis containing the vectors $u_1, ..., u_m$, you don't remove them basically.

Actually, the author kind of using Steinitz exchange theorem, so you might want to check it out.

  • $\begingroup$ so it's not that the list with $w_j$ elements automatically becomes linearly dependent or anything, right? It just seemed that way from his wording. $\endgroup$ – Shocked Jul 12 '17 at 14:24

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.