1
$\begingroup$

Let $\mathbb{F}$ be a field of characteristic $2$ and define $S$ to be the set of all triples $(i,j,k)\in\lbrace 1,\dotsc,n\rbrace^3$ with $\left|i-j\right|=1$, $\left|i-k\right|>1$, and $\left|j-k\right|>1$. Define $V=\bigoplus\limits_{(i,j)\in\lbrace 1,\dots,n\rbrace^2}\mathbb{F}\sigma_{i,j}$ to be the vector space with basis the quotient of $\lbrace \sigma_{i,j}\rbrace_{(i,j)\in\lbrace 1,\dots,n\rbrace^2}$ by the relation $\sigma_{i,j}=\sigma_{j,i}~\forall~(i,j)\in\lbrace 1,\dots,n\rbrace^2$. We can consider the vector subspace $V_S=\sum_{(i,j,k)\in S}\mathbb{F}(\sigma_{i,k}+\sigma_{j,k})$ (I don't know if this is an standard notation, but I want to represent the sum of subspaces). It is obvious that the above sum is not direct, since for every tuple $(i-1,i,j,j+1)$ with $i<j$ and $\left|j-i\right|>1$ we have $$(\sigma_{i-1,j}+\sigma_{i-1,j+1})+(\sigma_{i,j}+\sigma_{i,j+1})+(\sigma_{i-1,j}+\sigma_{i,j})+(\sigma_{i-1,j+1}+\sigma_{i,j+1})=0.$$ My question is if we can find the number of summands needed in $V_S$ such that the vector space is represented as a direct sum, which is equivalent to finding $\dim(V_S)$. It is easy to prove that $\left|S\right|=(n-2)(n-3)$ and with a bit of extra work one can see that the problem mentioned above can be solved by deleting $\frac{(n-3)^2}{4}$ summands in case $n$ is odd and $\frac{(n-2)(n-4)}{4}$ summands in case $n$ is even. However, this is not enough, for the case $n=6$ there is one extra summand to delete which does not correspond to the above situation. I guess there must be many more extra summands to delete for bigger $n$.

This problem appeared when trying to compute the homology of a certain family of groups, and I am not an expert on combinatorics, so I'm quite lost.

Any hints or help will be appreciated.

$\endgroup$
11
  • $\begingroup$ The dimension is $n(n-3)$. $\endgroup$ Commented Mar 23, 2023 at 12:10
  • $\begingroup$ @DaveBenson Can you elaborate a bit more? Also, how you can get $n(n-3)$ if $\left|S\right|=(n-2)(n-3)<n(n-3)$. I'm pretty sure that the computation of $\left|S\right|$ is ok. $\endgroup$
    – Marcos
    Commented Mar 23, 2023 at 12:15
  • $\begingroup$ I'm sorry, I counted everything twice, because I also used $\sigma_{k,i}+\sigma_{k,j}$. It should be $n(n-3)/2$ according to my computation. Actually, this also isn't true. Back soon. $\endgroup$ Commented Mar 23, 2023 at 12:40
  • $\begingroup$ Okay, so now that I'm not counting everything twice, I don't understand the relevance of your displayed relation. Are you sure the answer isn't just $(n-2)(n-3)$? $\endgroup$ Commented Mar 23, 2023 at 12:50
  • 1
    $\begingroup$ @DaveBenson Don't worry, do it when you have some spare time, thanks. $\endgroup$
    – Marcos
    Commented Mar 23, 2023 at 15:11

1 Answer 1

1
$\begingroup$

The dimension is $n(n-3)/2$. The argument is as follows. The set of pairs $(i,j)$ with $i$ and $j$ between $1$ and $n$, and $j-i \geqslant 2$ has cardinality $\binom{n-2}{2}=\frac{(n-1)(n-2)}{2}$. The $\sigma_{i,j}$ with these indices are the only ones involved. So let $W$ be the vector space spanned by these. For each element of $S$, we get a vector in $W$ such that the sum of the coordinates is zero. This gives us a subspace of $W$ of codimension one, and dimension $n(n-3)/2$. I claim that the images of elements of $S$ span this subspace. To see this, order the pairs $(i,j)$ lexicographically. Then for each $(i,j)$ apart from the first (which is $(1,3)$), there is an element of $S$ giving a vector whose last non-zero coordinate is $\sigma_{i,j}$. These span the codimension one subspace.

$\endgroup$
4
  • 1
    $\begingroup$ I suppose the point is that you were trying to count the dimension of the domain modulo the kernel. It's much easier to count the dimension of the image. $\endgroup$ Commented Mar 23, 2023 at 20:44
  • $\begingroup$ Yes, I've already counted the dimensión of the image, thanks. One question, why for each element of $S$ we have an element of $W$ such that the sum of the coordinates is zero? Is the only par I don't understand. $\endgroup$
    – Marcos
    Commented Mar 24, 2023 at 7:32
  • $\begingroup$ You're in characteristic two. There are two coordinates $1$, the rest are zero. $\endgroup$ Commented Mar 24, 2023 at 8:36
  • $\begingroup$ True, I was thinking that my vectors have three coordinates $1$ since I've defined $S$ in ters of triplets... Sorry and thanks. $\endgroup$
    – Marcos
    Commented Mar 24, 2023 at 9:15

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .