Let $S =\{u_1, u_2,...,u_n\}$ be a finite set of vectors. Prove that $S$ is linearly dependent iff $u_1 = 0$ or $u_{k+1} ∈ span(\{u_1, u_2,...,u_k\})$

509 Views Asked by At

Let $S =\{u_1, u_2,...,u_n\}$ be a finite set of vectors. Prove that $S$ is linearly dependent if and only if $u_1 = 0$ or $u_{k+1} ∈ \text{span} (\{u_1, u_2,...,u_k\})$ for some $k\space (1 ≤ k<n).$

I solved the problem as follows: If $u_1=0$ then $S$ is linearly independent as $1.0=1.u_1=0.$

If $u_{k+1} ∈ \text{span} (\{u_1, u_2,...,u_k\})$ then, $\exists$ scalars $c_1,...,c_k$ such that $$u_{k+1}=c_1u_1+...+c_ku_k\implies u_{k+1}-c_1u_1-...-c_ku_k=0,$$ which means $S$ is linearly dependent.

Now, for the other part we assume that $S$ is linearly dependent. If $u_1=0\in S$ then we are done. Else, from the definition of linear dependent sets we can find a finite set of vectors $a_1,...,a_{k+1}$ and scalars $c_1,...,c_{k+1}$ not all zero, such that $c_1a_1+...+c_{k+1}a_{k+1}=0$. Assuming $c_{k+1}\neq 0$ we have, $$a_{k+1}=c_{k+1}^{-1}c_1a_1+c_{k+1}^{-1}c_2a_2+...+c_{k+1}c_ka_k\implies a_{k+1}\in\text{span} (\{a_1, a_2,...,a_k\}).$$ Let $u1=a_1,u_2=a_2,...,u_k=a_k,u_{k+1}=a_{k+1}$ and then, we have, $u_{k+1} ∈ \text{span} (\{u_1, u_2,...,u_k\}).$

This completes the proof.


I think the question is ambiguous because if $u_{k+1} ∈ \text{span} (\{u_1, u_2,...,u_k\})$ is the claim (when $u_1\neq 0$), then they implicitly assume, that the coefficient of $u_1$ is always equal to $0$ if $u_1\neq 0.$ This is because, it was also a possible case, that $u_2$ was a linear combination of the vectors $$u_1,...,u_n(\implies u_2\in \text{span}(\{u_1,u_3,...,u_n\}))$$ and in addition, we have $u_2\notin \text{span}(\{u_1\}).$ Then the claim in the question would've been incorrect. The question is thus, wrong if we examine it carefully. Am i correct?

3

There are 3 best solutions below

2
On BEST ANSWER

You are really overthinking this. This answer is inspired from the comment of the user ancient mathematician.

Your first part which shows that if $u_1=0$ or if for some $k\in [1,n)$ such that $u_{k+1}\in\text{span}(\{u_1,u_2,...,u_{k-1}\})$ then $S$ is linearly dependent is correct. Also all's perfect upto the point where you showed, if $S$ is linearly dependent and $u_1=0$ then we are done.

The problem arises with the way you tried to prove, the statement:

If $S=\{u_1,u_2,...,u_n\}$ is a linearly dependent set then $\exists k$ satisfying $1\leq k\lt n$ such that, $u_{k+1}\in\text{span}(\{u_1,u_2,...,u_{k-1}\}).$

We give a proof of this (above) mentioned statement below:

Given $S=\{u_1,u_2,...,u_n\}.$ Since, $S$ is linearly dependent, there exist scalars $a_1,...,a_n$ not all zero such that $$a_1u_1+a_2u_2+...+a_nu_n=\sum_{i=1}^na_iu_i=0.$$ We choose the largest $1\leq k\lt n$ such that $a_{k+1}\neq 0.$ Then, $a_{k+1}u_{k+1}=a_1u_1+a_2u_2+...+a_{k}u_{k}+a_{k+2}u_{k+2}+...+a_nu_n\implies a_{k+2}=a_{k+3}=\cdots=a_n=0$ as $k$ is the largest of all the integers such that $a_{k+1}\neq 0.$ So, $$a_{k+1}u_{k+1}=a_1u_1+a_2u_2+...+a_{k}u_{k}+a_{k+2}u_{k+2}+...+a_nu_n=a_1u_1+a_2u_2+...+a_{k}u_{k}+0.u_{k+2}+\cdots +0.a_n=a_1u_1+a_2u_2+...+a_{k-1}u_{k-1}\implies u_{k+1}=a_{k+1}^{-1}a_1u_1+...+a_{k+1}^{-1}a_{k-1}u_{k-1}\implies u_{k+1}\in\text{span}(\{u_1,u_2,...,u_{k-1}\}).$$


This is all to it! With this modification, the proof in the original post should be fixed.

2
On

(1) The Claim at hand is totally Correct & unambiguous.

(2A) I am not sure why & how you state that the textbook Proof assumes that the co-efficient of $u_1$ must be $0$.
(2B) You also state that $u_2$ must be in the span of all other elements , which might be true but that is not related to the Claim at hand.

(3) The Claim at hand is about putting the elements in some order & then saying that there is at least 1 element $u_{k+1}$ which is in the span of the previous $k$ elements. It is not about whether some element $u_{k+1}$ is in the span of all other $(n-1)$ elements.
IMPORTANT : Here , the $u$ Element values are fixed , though the Order is not , because it is a Set.

(4) In other words , this might be easy to see like this :
Put the elements in order. Either $1^{st}$ element is $0$
Or $2^{nd}$ element is in span of Previous element
Or $3^{rd}$ element is in span of Previous $2$ elements
Or $4^{th}$ element is in span of Previous $3$ elements
Or $5^{th}$ element is in span of Previous $4$ elements
Or $6^{th}$ element is in span of Previous $5$ elements
....
Or $n^{th}$ element is in span of Previous $(n-1)$ elements

Here , if all the Cases are not true , it means that we have linearly InDependent Elements. If at least 1 Case had been true , it means that we have linearly Dependent Elements & that Case gives the least $k$ value.

Important : When the Order is changed , that $k$ value might change.

(5) There may or may not be other valid $k$ values , the Claim at hand is about getting at least 1 such $k$ value.
When we put the elements in some other Order , the $k$ value(s) might change.

This Example should clear things further ...

Let $S=\{(0,0),(0,1),(0,2),(1,0),(1,1)\}$

I put in this order :
$u_1=(0,0)$
$u_2=(0,1)$
$u_3=(0,2)$ .... $u_n=(1,1)$
Then Claim is true : $u_1=0$

I put in this order :
$u_1=(0,1)$
$u_2=(0,2)$
$u_3=(1,1)$ .... $u_n=(0,0)$
Then Claim is true : $u_2=2u_1$

I put in this order :
$u_1=(0,1)$
$u_2=(1,0)$
$u_3=(1,1)$ .... $u_n=(0,0)$
Then Claim is true : $u_3=u_1+u_2$

Elements are fixed , Order is not fixed.
We choose 1 Order & then go with the Claim , to get the $k$ value according to that Order.

Note: When we put in "Some Order" & get a $k$ value , then we have linearly Dependent Elements.

If we put in "Some Other Order" , then we MUST get a $k$ value ( might or might not be Same $k$ value ) , other-wise we would have linearly Independent Elements !

1
On

For future reference, I believe the cleanest and most efficient approach is to take the largest $k$ with $1\le k<n$ such that $\{u_1,\dots,u_k\}$ is linearly independent. (Note that we're using the hypothesis that $u_1\ne 0$ here.) This means that $\{u_1,\dots,u_k,u_{k+1}\}$ is linearly dependent and you can check immediately that $u_{k+1}\in\text{Span}(u_1,\dots,u_k)$.