I'm reading through Treil's Linear Algebra Done Wrong and have a question about proving a remark he leaves as an exercise on p.77. Here it is in my own words:

Consider a function on $n$ vectors $D(v_1,v_2,\ldots,v_n)$. It has the following properties.

  1. $D(v_1,\ldots,\alpha v_k,\ldots,v_n) = \alpha D(v_1,\ldots,v_k,\ldots,v_n)$, for any $\alpha \in \mathbb{F}$, and
  2. $D(v_1,\ldots,v_j + \alpha v_k, \ldots, v_k, \ldots, v_n) = D(v_1,\ldots,v_j,\ldots,v_k,\ldots,v_n)$ for $j \neq k$ and any $\alpha$.

Using only these two properties, I'm supposed to show that $D$ is additive in each argument, that is,

$$D(v_1,\ldots,u_k + v_k, \ldots, v_n) = D(v_1,\ldots,u_k,\ldots,v_n) + D(v_1,\ldots,v_k,\ldots,v_n).$$

So far, I've checked that additivity holds for $u_k \in \operatorname{span}(\{v_i\}_{i=1, i\neq k}^n)$ and $u_k = \alpha v_k$, with the hope of checking each case of $u_k$ or something, but I don't think that's going anywhere. I think there is something clever I'm not seeing — tips or solutions appreciated!


You are close! You just need to show that nothing else can happen besides the cases you've covered.

We first establish a well-known property of the determinant from the two properties. Consider any $v_1,\ldots, v_n$. Suppose the dimension of $\operatorname{span}\{v_i\}$ is less than $n$. Then we claim $D(v_1,\ldots v_n)=0$.

These $n$ vectors are linearly dependent. By applying property 2 repeatedly, you can show that $D(v_1,\ldots, v_n)=D(0,v_2,\ldots, v_n)$, and then apply property 1 to show that this quantity is zero.

We now return to your question. If the dimension of $\operatorname{span}\{v_i\}_{i \ne k}$ is less than $n-1$, then the above shows that the claim you want to show becomes $0=0+0$. So from here on we assume $\operatorname{span}\{v_i\}_{i \ne k}$ has dimension $n-1$.

If $u_k$ is in the span of $\{v_i\}_{i\ne k}$, then $D(v_1,\ldots, u_k,\ldots, v_n)=0$, and you can show that $D(v_1,\ldots, u_k+v_k,\ldots, v_n)=D(v_1,\ldots, v_k, \ldots, v_n)$ using the second property.

If $v_k$ is in the span of $\{v_i\}_{i\ne k}$, a similar argument works.

Finally, if neither $u_k$ nor $v_k$ lie in the span of $\{v_i\}_{i\ne k}$, then $u_k=\alpha v_k$ for some $\alpha$ [edit: see correction in comments] (because $\operatorname{span}\{v_i\}_{i \ne k}$ has dimension $n-1$), and you have shown this case.

  • $\begingroup$ Ah, thank you! I think at your last sentence you meant to say that $u_k = \alpha v_k + \sum_{i=1,i \neq k}^n \beta_i v_i$? Regardless, very clear and helpful answer : ) $\endgroup$ – rorty Oct 31 '15 at 0:51
  • $\begingroup$ @rorty Oops, yes you are absolutely right! $\endgroup$ – angryavian Oct 31 '15 at 1:13

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.