Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Linear independence

  1. Aug 26, 2005 #1
    Hi, can someone help me with the following question?

    Q. Show that if [tex]\left\{ {\mathop {v_1 }\limits^ \to ,...,\mathop {v_k }\limits^ \to } \right\}[/tex] is linearly independent and [tex]\mathop {v_{k + 1} }\limits^ \to \notin span\left\{ {\mathop {v_1 }\limits^ \to ,...,\mathop {v_k }\limits^ \to } \right\}[/tex] then [tex]\left\{ {\mathop {v_1 }\limits^ \to ,...,\mathop {v_k }\limits^ \to ,\mathop {v_{k + 1} }\limits^ \to } \right\}[/tex] is linearly independent. Use this to prove that the non-zero rows of a matrix in row-echelon form are linearly independent.

    Here is my attempt.

    Write [tex]\alpha _1 \mathop {v_1 }\limits^ \to + .... + \alpha _k \mathop {v_k }\limits^ \to + \beta \mathop {v_{k + 1} }\limits^ \to = \mathop 0\limits^ \to ...\left( 1 \right)[/tex]

    \beta \mathop {v_{k + 1} }\limits^ \to = - \left( {\alpha _1 \mathop {v_1 }\limits^ \to + .... + \alpha _k \mathop {v_k }\limits^ \to } \right)

    If [tex]\beta \ne 0[/tex] then [tex]\mathop {v_{k + 1} }\limits^ \to = - \left( {\frac{{\alpha _1 }}{\beta }\mathop {v_1 }\limits^ \to + ...\frac{{\alpha _k }}{\beta }\mathop {v_k }\limits^ \to } \right)[/tex] but this is impossible since [tex]\mathop {v_{k + 1} }\limits^ \to \notin span\left\{ {\mathop {v_1 }\limits^ \to ,...,\mathop {v_k }\limits^ \to } \right\}[/tex]

    So beta is equal to zero and equation one reduces to [tex]\alpha _1 \mathop {v_1 }\limits^ \to + .... + \alpha _k \mathop {v_k }\limits^ \to = \mathop 0\limits^ \to [/tex] where all of the a_i are equal to zero by hypothesis. Is that enough to show the given result?

    I can't think of a way to tackle the second part with the matrix. Seeing as that's the case I'll just write out whatever I can think of.

    I think the key idea is that in row echelon form, each time I 'move up' one row, the vector(represented by a row in the matrix) has at least one additional non-zero component. So let A be the n by k (n columns and k rows) matrix in row echelon form whose rows are the vectors v_i where i = 1,...,k and each of the vectors has at least one non-zero component.

    Starting at the bottom of the matrix and moving up to the first non-zero row I a vector which has c non-zero components call it v_1 and {(v_1)} is linearly independent since it consists of a non-zero single vector. Moving up to the next row I get another vector call it v_2 which has at least c + 1 non-zero components. Since v_2 has more non-zero components than v_1 then {v_1, v_2} is linearly independent. From here I'd probably just continue with the same argument. The problem is that what I've said is a pretty clumsy explanation. I wasn't really sure how to do this question either. So can someone please help me with this?

    Edit: Ok my attempt for the second part is completely incorrect because I could have something like v_1 = (0,0,1,0,0) and v_2 = (1,0,0,0,0). Help would be appreciated.
    Last edited: Aug 26, 2005
  2. jcsd
  3. Aug 26, 2005 #2


    User Avatar
    Homework Helper

    Do the second part by induction.
    n=1 case
    the vector is not zero so is linearly independent
    n+1 case
    n vectors are independent
    if (n+1)st vector is in span(first n vectors) the matrix is not in row-echelon form
    therfore (n+1)st vector is not in span
    thus n+1 vectors are linearly independent
  4. Aug 27, 2005 #3
    Thanks for te help lurflurf. However, I am not sure how my answer should be worded. The question refers to a matrix so would I need to make some reference to the matrix? If so how would I do it in a clear and concise manner? For example, do I need to mention the position of the vectors(represented as rows) in the matrix?
  5. Aug 27, 2005 #4


    User Avatar
    Homework Helper

    The matrix is a representation. You can consider the set of the nonzero row vectors. The position of the vectors is not important.
    Just say something like
    let v1,...,vn be the nonzero row vectors
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook