In the book Linear Algebra by Werner Greub, at page $112$, it is given that,

Let $E^*, E$ be a pair of dual vector spaces and $\Delta^* \not = 0, \Delta \not = 0$ be determinant functions in $E^*$ and $E$. It will be shown that $$\Delta^*(x^{*1}, \ldots, x^{*n}) \Delta(x_1, \ldots, x_n) = \alpha \det(\langle x^{*i}, x_j \rangle), \quad x^{*i} \in E^*, x_j \in E,$$ where $\alpha$ is a scalar.

Consider the function $\Omega$ of $2n$ vectors defined by $$\Omega(x^{*1}, \ldots, x^{*n}; x_1, \ldots, x_n) = \det(\langle x^{*i}, x_j \rangle).$$

Then it follows from the properties of the determinant of a matrix that $\Omega$ is linear with respect to each argument. Moreover, $\Omega$ is skew symmetric with respect to the vectors $x^{*i}$ and with respect to the vectors $x_i$ (i= 1...n). Hence the uniqueness theorem (sec. 4.3) implies that $\Omega$ can be written as $$\Omega(x^{*1}, \ldots, x^{*n}; x_1, \ldots, x_n) = \phi (x^{*1}, \ldots, x^{*n}) \Delta(x_1, \ldots, x_n)$$

However, the uniqueness theorem is valid(from the its statement) only if $\Omega$ take only one of the $n$ vectors as its input, i.e

enter image description here

So how can we prove that $\Omega(x^{*1}, \ldots, x^{*n}; x_1, \ldots, x_n)$ can be written as $$\Omega(x^{*1}, \ldots, x^{*n}; x_1, \ldots, x_n) = \phi (x^{*1}, \ldots, x^{*n}) \Delta(x_1, \ldots, x_n)$$


The author is arguing as follows: for any fixed $n$-tuple $(x^{*1},\ldots,x^{*n})$ of covectors, the function $$(x_1,\ldots,x_n) \mapsto \Omega(x^{*1},\ldots,x^{*n};x_1,\ldots,x_n)$$ is multilinear and skew-symmetric (i.e., it is a determinant function). Therefore by the uniqueness theorem of determinant functions, there exists a scalar $\Phi(x^{*1},\ldots,x^{*n})$ (depending on our fixed $n$-tuple) such that our determinant function is just a multiple of the fixed non-trivial determinant function $\Delta$, i.e., $$\Omega(x^{*1},\ldots,x^{*n};x_1,\ldots,x_n) = \Phi(x^{*1},\ldots,x^{*n}) \Delta(x_1,\ldots,x_n).$$

Stated differently: the set of all determinant functions on $E$ actually forms a vector space. (If you're familiar with exterior algebra, it is the space $(\Lambda ^n E)^*$.) The uniqueness theorem above amounts to the statement that this space is one-dimensional, spanned by $\Delta$. Therefore there is an isomorphism $\phi:(\Lambda ^n E)^* \to K$ characterized by $\phi(\Delta) = 1$, where $K$ denotes the base field. The author is implicitly defining a map which I will call $\psi:(E^*)^n \to (\Lambda ^n E)^*$, by the rule $$\psi(x^{*1},\ldots,x^{*n}) = [(x_1,\ldots,x_n) \mapsto \Omega(x^{*1},\ldots,x^{*n};x_1,\ldots,x_n)].$$ Then his $\Phi$ is precisely the composition $\Phi = \phi \circ \psi$.

  • $\begingroup$ I mean I can see the result intuitively, but I think your whole statement is still not rigorous. $\endgroup$ – onurcanbektas Aug 4 '17 at 17:32
  • $\begingroup$ @onurcanbektas The argument is solid. Tell me what part bothers you and I'll gladly explain in more detail. $\endgroup$ – Alex Provost Aug 4 '17 at 17:53
  • $\begingroup$ I will take a look at it in the morning with a fresh head, then I will write you an explanation in detail. $\endgroup$ – onurcanbektas Aug 4 '17 at 17:59
  • $\begingroup$ Do you mean $\bigwedge$ ("\bigwedge") or $\Lambda$ ("\Lambda")? $\endgroup$ – Eric Towers Aug 4 '17 at 21:14
  • 1
    $\begingroup$ @onurcanbektas It maps the $n$-tuple of covectors $(x^{*1},\ldots,x^{*n})$ to the determinant function that maps the $n$-tuple of vectors $(x_1,\ldots,x_n)$ to $\Omega(x^{*1},\ldots,x^{*n};x_1,\ldots,x_n)$. $\endgroup$ – Alex Provost Aug 5 '17 at 5:55

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.