Let $a_1, \ldots, a_n$ for $n \geq 2$ vectors in an $\mathbb{F}$-vector space $V$.

Show the following:

$a_1 - a_n, \ldots, a_{n-1} - a_n$ linearly independent $\Leftrightarrow$ $a_2 - a_1, \ldots, a_n - a_1$ linearly independent

My (flawed) idea

I tried to show the above by using linear (in-)dependence of $a_1, \ldots, a_n$ as an argument. First of all, let's try to linearly combine the above two systems to zero, i.e. check their linear (in-)depence:

\begin{align} & \lambda_1(a_1 - a_n) + \ldots + \lambda_{n-1} (a_{n-1} - a_n) &= 0 & \qquad(1.1)\\ \Leftrightarrow & \lambda_1 a_1 + \ldots + \lambda_{n-1} a_{n-1} - (\lambda_1 + \ldots + \lambda_{n-1}) a_n &= 0 &\qquad(1.2) \end{align}

\begin{align} & \mu_1(a_2 - a_1) + \ldots + \mu_{n-1} (a_{n} - a_1) &= 0 & \qquad(2.1)\\ \Leftrightarrow & -(\mu_1 + \ldots + \mu_{n-1}) a_1 + \mu_1 a_2 + \ldots + \mu_{n-1} a_{n} &= 0 &\qquad(2.2) \end{align}

If $a_1, \ldots, a_n$ are linearly independent, we follow from (1.2) and (2.2) that $\lambda_1 = \ldots = \lambda_{n-1} = 0$ and $\mu_1 = \ldots = \mu_{n-1} = 0$ and thus the above two systems are both linearly independent.

If $a_1, \ldots, a_n$ are not linearly independent, then there's a non-trivial solution, i.e. $$ \exists i \in \left\{1, \ldots, n\right\}: \kappa_1 a_1 + \ldots \kappa_n a_n = 0,\ \kappa_i \neq 0 \qquad (3) $$

But unfortunately, we can't directly follow that the above two systems must now be linearly dependent. Let's say $(1, 0, \ldots, 0)$ is a solution for (3) (and as the solutions of a homogeneous system of equations forms a subvectorspace, I reckon all $(x, 0 \ldots, 0), x \in \mathbb{F}$ would be a solution in that case). Then we still wouldn't be able to find a non-trivial solution for (2.2), as $\kappa_1 = -(\mu_1 + \ldots + \mu_{n-1}) = 0$.

The question

So I ask you: is my above proof-idea salvageable? What's a good way to prove the statement to be shown?


Note: You tried to argue by applying the definition of LI to both the set of vectors together and that is why you are unable to complete the proof.

Let $S=\{a_i-a_n\}_{i=1}^{n-1}$ and $T=\{a_{j+1}-a_1\}_{j=1}^{n-1}$. We have to show that S is LI $\iff$ T is LI.

Assume that $S$ is LI. Let us consider the linear combination of vectors in $T$ that adds up to the zero vector. \begin{align*} c_1(a_2-a_1)+c_2(a_3-a_1)+\dotsb+c_{n-1}(a_n-a_1) & =0\\ c_1(a_2-a_{\color{red}{n}})+c_2(a_3-a_{\color{red}{n}})+\dotsb+c_{n-1}(a_n-a_1) & =\left(\sum_{k=1}^{\color{blue}{n-2}}c_i\right)(a_1-a_{\color{red}{n}})\\ c_1(a_2-a_n)+c_2(a_3-a_n)+\dotsb-\left(\sum_{k=1}^{\color{red}{n-1}}c_i\right)(a_1-a_n) & =0 \end{align*} Now we use the fact that the set $S$ is LI. This means these coefficients must be zero. Thus we have $$c_1=c_2=\dotsb=c_{n-2}=0 \quad \text{ and } \left(\sum_{k=1}^{n-1}c_i\right)=0.$$ This gives us $c_i=0$ for all $i \in \{1,2,3, \ldots, n-1\}$. Thus the set $T$ is LI.

We can reverse this argument ot get the other implication as well.

  1. I'm not sure your proof can be easily finished for the case when $a_1,\dots, a_n$ are linearly dependent.
  2. Let $b_i:=a_i-a_n$ (with $i<n$) and $c_i:=a_i-a_1$ (with $i>1$).
    Then we have $c_2=b_2-b_1,\ c_3=b_3-b_1,\dots$ and $c_n=-b_1$.
    But, this is a linear basis transformation, with inverse $b_1=-c_n$ and $b_i=c_i-c_n$ for $i>1$.

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.