Explain why $S$ is not a basis for $\mathbb{R}^3$

$S=\{(1, 3, 0),(4, 1, 2),(-2, 5, -2)\}$

I set this equal to an arbitrary vector $\mathbf{x} = (x_1, x_2, x_3)$

After solving I got the matrix:

$\begin{bmatrix} 1 & 4 & -2 \\ 3 & 1 & 5 \\ 0 & 2 & -2\end{bmatrix}$

And since the $det(A)=0$ this set does not span $\mathbb{R}^3$. However, I don't understand the answer the solution gave. It says:

A basis for $\mathbb{R}^3$ contains three linearly independent vectors. Because $-2(1, 3, 0)+(4, 1, 2)+(-2, 5, -2)=(0, 0, 0)$

$S$ is linearly dependent and is, therefore, not a basis for $\mathbb{R}^3$

However, if I test for linear independence I get the following $rref(A)$:

$\begin{bmatrix} 1 & 0 & 2 & 0 \\ 0 & 1 & -1 & 0 \\ 0 & 0 & 0 & 0 \end{bmatrix}$

Which would imply to me that $S$ is linearly independent given the $0$ column.


Your RREF tells you that $S$ is linearly independent. What you have done is row-reduced the augmented matrix: $$\left(\begin{array}{ccc|c} 1 & 4 & -2 & 0\\ 3 & 1 & 5 & 0\\ 0 & 2 & -2 & 0 \end{array}\right)$$ to the reduced row echelon form $$\left(\begin{array}{ccc|c} 1 & 0 & 2 & 0\\ 0 & 1 & -1 & 0\\ 0 & 0 & 0 & 0 \end{array}\right)$$ What this tells you is that the matrix formed by setting the columns equal to the given vectors of $S$ has non-trivial nullspace, meaning there exist constants $c_1,c_2,c_3$ not all $0$, such that $c_1 (1,3,0) + c_2(4,1,2) + c_3(-2,5,-2) = (0,0,0)$

What we will have done is effectively checked whether the 3 vectors given satisfy the definition of linear independence. Since we have nonzero constants $c_1,c_2,c_3$ so that $c_1 (1,3,0) + c_2(4,1,2) + c_3(-2,5,-2) = (0,0,0)$, they are not linearly independent.

In fact, by simply backsolving and picking $c_3 = 1$, we could choose $c_2 = 1$, $c_1 = -2$

That implies that $\{(1,3,0),(4,1,2),(-2,5,-2)\}$ are linearly dependent, as the solution in the book states. And 3 vectors in $\mathbb{R}^3$ are a basis of $\mathbb{R}^3$ if and only if they are linearly independent, so the 3 vectors given are not a basis of $\mathbb{R}^3$.

I previously said that this is because you have a row of zeros. Technically, this is wrong. The correct explanation has to do with pivot columns, namely that we have a non-leading non-pivot column on the LHS when reducing to row echelon form. A row of zeros is enough if you are dealing with a square matrix, but seeing as that is rarely the case, you will want to know the more general method.

Say you have $4$ vectors in $\mathbb{R}^6$. These are linearly independent if and only if the matrix formed with the vectors has rank $4$. Another way to verify this is that if each and every column on the LHS of the augmented matrix (when reduced to row echelon form) is a pivot column, then the column vectors forming your matrix are linearly independent.

  • $\begingroup$ Thank you for this explanation! This is what I was looking for. $\endgroup$ – hax0r_n_code Apr 23 '14 at 1:53
  • $\begingroup$ You are quite welcome. I went ahead and elaborated a bit on what the mechanics are behind seeing whether 3 vectors in $\mathbb{R}^3$, or for that matter, any set of $n$ vectors in $\mathbb{R}^n$, is a basis of $\mathbb{R}^n$ (equivalently, linearly independent). In practice, set the vectors up as columns in a matrix, and row-reduce to compute the nullspace of the matrix. If the nullspace is nontrivial (i.e., if the nullspace contains a nonzero vector), then the vectors are linearly dependent, and cannot be a basis of $\mathbb{R}^n$. $\endgroup$ – Nicholas Stull Apr 23 '14 at 1:59
  • $\begingroup$ Strictly speaking the explanation given in this question is not correct: it is the non-leading (non-pivot) column on the LHS that shows the vectors are dependent, not the zero row. For example, if two vectors in ${\Bbb R}^3$ lead to an echelon form$$\pmatrix{1&2\cr0&3\cr0&0\cr}\ ,$$then the vectors are independent despite the zero row. If the LHS is a square matrix, as in the OP, it comes to the same thing, but usually it doesn't. $\endgroup$ – David Apr 23 '14 at 2:11
  • $\begingroup$ Corrected. I was admittedly dealing specifically with the case $n$ vectors in $\mathbb{R}^n$ (as in checking whether $n$ vectors form a basis of $\mathbb{R}^n$), but as this usually is not the case, I should not have left that as the main reason for seeing this. It just happens to work if the LHS is square. $\endgroup$ – Nicholas Stull Apr 23 '14 at 2:38


$$ 2(1,3,0) - (4,1,2) = (-2,5,-2) $$

  • $\begingroup$ Was the method I used incorrect? How would I have gotten your answer? $\endgroup$ – hax0r_n_code Apr 23 '14 at 1:45
  • $\begingroup$ A basis is a set of vectors that span the vector space $V$ and is a linearly independent set. In this case $V = \mathbb{R}^3$, and your set of vectors is not linearly independent because of what I wrote. $\endgroup$ – user139708 Apr 23 '14 at 1:47
  • $\begingroup$ Yes, I understand what you're saying, but how would I have gotten to your answer since I didn't see it? Are there mechanics to finding this answer that you used or is it just something you spotted by examining the vectors? $\endgroup$ – hax0r_n_code Apr 23 '14 at 1:48
  • $\begingroup$ there is not a machine to do that. Just see your vectors and play with them. $\endgroup$ – user139708 Apr 23 '14 at 1:49
  • $\begingroup$ your solution is correct. You can put them in a matrix, and if the det is $0$, then they are not linearly independent $\endgroup$ – user139708 Apr 23 '14 at 1:51

This is an amplification of Lemur's answer, expanding on the "see your vectors and play with them" in his comment. We want to know if we can find scalars $\alpha,\beta,\gamma$ such that $$\alpha(1,3,0)+\beta(4,1,2)+\gamma(-2,5,-2)=(0,0,0)\ .$$ But looking at the third component gives immediately $\beta=\gamma$. So the previous equation can be written as $$\alpha(1,3,0)+\beta(2,6,0)=(0,0,0)\ ,$$ and now it is easy to see that $\alpha=2$, $\beta=-1$ is a solution. So $$2(1,3,0)-(4,1,2)-(-2,5,-2)=(0,0,0)\ ,$$ which shows that the vectors are linearly dependent.


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.