0
$\begingroup$

I am trying to show that a general element of the kth exterior product $\Lambda^kV^*$ (of V an n-dimensional vector space) $$ \alpha = \sum_{i} \alpha_i e_i$$ (where the $\{e_i\}$, for $1\leq i\leq {n\choose k}$ are a basis of the vector space $\Lambda^kV^*$) can be written as : $$\alpha = f_j \wedge \beta_j, $$ where the $f_j$'s are at most $k$ linearly independent elements of $V^*$ and $\beta_j\in \Lambda^{k-1}V^*$ depends on $\alpha$ and on the possible $f_j$. I thought of using a recurrence on $k$ (it is obviously true when $k=1$ and $k=n$), but I don't just don't know how I could use the kth step to proceed up to the (k+1)th step...

The objective of this exercise, is to get a bound on the dimension of the kernel of : $$ x\mapsto x\wedge \alpha$$ when $x$ is a 1-form and $\alpha$ is a given $k$ form. I was able to show that the dimension of the kernel is equal to the number of $f_j$ of the preceding paragraph, and now I'm stuck.

I also tried a less general approach, consisting on counting the different basis elements $e_i$ of $\alpha$ (not including the elements appearing once in every $e_i$) that differ by just one element of the $\Lambda V^*$ basis (those that will give non-trivial equations for $x\wedge \alpha = 0$). This procedure looks however like a big mess.

If anyone could help, I would appreciate any clue.

$\endgroup$
  • $\begingroup$ Actually, I've found what I was looking for (in great detail) in the notes of K. Conrad! (indeed the problem I was working on was related to the "decomposability" of $\alpha$, and the way I looked at the problem was pretty wrong). $\endgroup$ – faero Jan 24 '16 at 0:01
1
$\begingroup$

I find your first question pretty unclear as it is stated, so I will concentrate on the second half.

Consider the basis of $\Lambda^kV^*$ given by $$\{v_{i_1}\wedge\ldots v_{i_k}|1\le i_1<\ldots<i_k\le n\}.$$ We can write $\alpha$ as $$\alpha = \alpha^{i_1\ldots i_k}v_{i_1}\wedge\ldots v_{i_k}.$$ We have $$v_j\wedge\alpha = \left(\prod_{\ell=1}^k\delta_{ji_\ell}\right)(-1)^{\max\{\ell|i_\ell<j\}}\alpha^{i_1\ldots i_k}v_{i_1}\wedge\ldots\wedge v_j\wedge\ldots\wedge v_{i_k}.$$ This transforms your problem in pure linear algebra, namely to finding the dimension of the kernel of a matrix of dimension $k$ times $\binom{n}{k+1}$. Of course, a priori this is not an easy task to do by hand, but in many cases you are interested in $\alpha$'s giving you an easy form for your matrix.


As an example, let $V^*=\mathbb{R}^4$ with basis $\{x,y,z,w\}$ and $\alpha = x\wedge y + z\wedge w$. Then \begin{align} x\wedge\alpha = & x\wedge z\wedge w\\ y\wedge\alpha = & y\wedge z\wedge w\\ z\wedge\alpha = & x\wedge y\wedge z\\ w\wedge\alpha = & x\wedge y\wedge w, \end{align} so that (after choosing an order of the basis of $\lambda^3V^*$) the associated matrix is $$\pmatrix{0&0&1&0\\0&0&0&1\\1&0&0&0\\0&1&0&0},$$ which has maximal rank and is square, and thus has trivial kernel.

$\endgroup$
0
$\begingroup$

This seems false. Consider $R^6$ with coordinates $a,b,c,x,y,z$. Then

$$ da \wedge db + dc \wedge dx + dy \wedge dz $$ which is a 2-form, can't be written (I think!) as a product of 1-forms $$ s \wedge \phi $$ where $s$.

At the very least, you can generalize the number of coordinates to something quite large; if you're restricting to just two $f_j$s, then it's gonna be tough.

By the way, do you mean to write $$ \alpha = \sum_j f_j \wedge \beta_j $$ perhaps? Or is every 2-form supposed to be either $f_1$ wedge something, OR $f_2$ wedge something, etc.?

I think that if you look up "indecomposable 2-forms", you'll see what I'm talking about. In fact, this question gives a criterion for a 2-form to be decomposable. :)

$\endgroup$
  • $\begingroup$ Indeed, your first example is indecomposable (hence, in that case, there is no such $f_j$ and the kernel of the cited application in the question is just $0$). I didn't mean a summation. As you stated after, it is an "or". So it is looking for the maximal number of linearly independent 1-forms that can serve to decompose a k-form separately. $\endgroup$ – faero Jan 23 '16 at 0:53

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.