3
$\begingroup$

Let $E_{ij}(x)\in \mathrm{Mat}_{7\times7}(\overline{\mathbb{F}}_2)$ be the matrix with zeros everywhere, except for the value $x$ at $ij$. Set $$a(x)=1+E_{12}(x)+E_{34}(x)+E_{56}(x),\quad b(y)=1+E_{23}(y)+E_{45}(y)+E_{67}(y),$$ for two indeterminates $x,y$. Set $$G_8=\langle a(x), b(y) \mid x, y\in\overline{\mathbb{F}}_2\rangle.$$ Is $G_8$ Zariski closed and connected?

For easier visualization, here are the generating matrices:

$$ a(x)=\begin{pmatrix} 1&x& & & & & \\ &1&0& & & & \\ & &1&x& & & \\ & & &1&0& & \\ & & & &1&x& \\ & & & & &1&0\\ & & & & & &1 \end{pmatrix} ,\ b(y)=\begin{pmatrix} 1&0& & & & & \\ &1&y& & & & \\ & &1&0& & & \\ & & &1&y& & \\ & & & &1&0& \\ & & & & &1&y\\ & & & & & &1 \end{pmatrix} $$

To put things on a more general footing, given such a description of a subgroup of an algebraic group, how does one show that it is connected?

$\endgroup$
  • 3
    $\begingroup$ It's generated by the union of two connected (1-parameter) subgroups so is clearly Zariski-connected. $\endgroup$ – YCor Mar 27 at 8:56
  • 2
    $\begingroup$ To see whether it's Zariski-closed... probably one can describe the group, at the price of a few computations (I'd compute various commutators such as $[a(x),b(y)]$, $[a(x),[a(y),b(z)]]$, etc.) $\endgroup$ – YCor Mar 27 at 8:59
  • 2
    $\begingroup$ ... and I'd start beforehand solving the analogous problem in size $<7$. It gives the projections to the northwest and southeast blocks, so is certainly useful. $\endgroup$ – YCor Mar 27 at 9:46
  • 5
    $\begingroup$ Springer LAG Prop. 2.2.6 answers both questions: The union of irreducible subvarieties, each containing 1, generates a closed connected subgroup. $\endgroup$ – Friedrich Knop Mar 27 at 19:35
4
$\begingroup$

Good news and bad news here: This group is connected and Zariski closed for nice general reasons. However, I assume that Dror considered this group because $a(1)$ and $b(1)$ generate a subgroup isomorphic to the dihedral group of order $16$, and are thus relevant to my question. It turns out that the group generated by $a(x)$ and $b(y)$ has dimension at least $9$. The issue is that "the $\mathbb{F}_2$ points of the group generated by $X$" can be much larger than "the group generated by the $\mathbb{F}_2$ points of $X$".

Good news: See Section 2.h of Milne, Algebraic Groups for the following results. Let $G$ be an algebraic group scheme over $k$, let $X$ be a geometrically connected, geometrically reduced, subvariety of $G$ over $k$ and let $\phi: X \to G$ be a map defined over $k$ such that $\phi(X)$ is closed under inversion.

Then there is a smallest algebraic sub-group-scheme $H$ of $G$ such that $\phi$ factors through $H$, and this group is connected. Define $\phi_n : X^n \to G$ by $(x_1, \ldots, x_n) \mapsto \phi(x_1) \cdots \phi(x_n)$, then $\phi^n : X^n \to H$ is dominant for $n$ sufficiently large.

Furthermore (this part isn't in Milne, but seems right to me), let $n$ be large enough that $\phi^n : X^n \to H$ is dominant. Then I claim that $\phi^{2n} : X(k^{\mathrm{alg}})^{2n} \to H(k^{\mathrm{alg}})$ is surjective. Proof: Let $U$ be an nonempty open set within the image of $\phi^n$. Since $H$ is irreducible, for any $h \in H(k^{\mathrm{alg}})$, the open sets $U$ and $U^{-1} h$ have nonempty intersection, so we can find some point $u_1 \in H(k^{\mathrm{alg}})$ which is both in $U(k^{\mathrm{alg}})$ and of the form $u_2^{-1} h$ for $u_2 \in U(k^{\mathrm{alg}})$, and thus $h = u_1 u_2$.

So we can literally think of $H$ as the subgroup of $G$ generated by $X$, as long as we work with points in the algebraic closure.

Bad news: Let $G$ be the group of $7 \times 7$ upper triangular matrices with $1$'s on the diagonal. Define $$G_k = \{ g \in G : g_{ij}=0 \ \mbox{for} \ i < j < i+2^k \}.$$ So $G = G_0 \supset G_1 \supset G_2 \supset G_3 = \{ \mathrm{Id} \}$. Each $G_k$ is normal in $G$, and each $G_k/G_{k+1}$ is a product of copies of $\mathbb{G}_a$. Let $H_k = H \cap G_k$. I will verify that $\dim H_0/H_1 = 2$, $\dim H_1/H_2 = 3$ and $\dim H_2/H_3 \geq 2$. So the group $H$ has dimension at least $7$. Details follow. Note that we always have $g_{ij} = g_{(i+2)(j+2)}$ for any $g \in H$.

Computation of $H_0/H_1$ Every product in $H$ is of the form $$\begin{bmatrix} 1 & x & \ast & \ast & \ast & \ast & \ast \\ 0 & 1 & y & \ast & \ast & \ast & \ast \\ 0 & 0 & 1 & x & \ast & \ast & \ast \\ 0 & 0 & 0 & 1 & y & \ast & \ast \\ 0 & 0 & 0 & 0 & 1 & x & \ast \\ 0 & 0 & 0 & 0 & 0 & 1 & y \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 \\ \end{bmatrix}$$ and every $(x,y)$ occurs, so $H_0/H_1 \cong \mathbb{G}_a^2$.

Computation of $H_1/H_2$ Let $h=\prod_j \left( a(x_j) b(y_j) \right) \in H_1$. Then $\sum x_i = \sum y_j=0$. Then $h_{13} = h_{35} = h_{57} = \sum_{i<j} x_i y_j$ and $h_{24} = h_{46} = \sum_{i \geq j} x_i y_j$. We moreover have $h_{13} + h_{24} = \sum_{i,j} x_i y_j = \left( \sum_i x_i \right) \left( \sum_j y_j \right) = 0$, so $h_{13} = h_{24}$. So we have $h_{13} = h_{35} = h_{57} = h_{24} = h_{46}$. Also, $h_{14} = h_{36}$ and $h_{25} = h_{47}$. So all matrices of $H_1$ are of the form $$\begin{bmatrix} 1 & 0 & p & q & \ast & \ast & \ast \\ 0 & 1 & 0 & p & r & \ast & \ast \\ 0 & 0 & 1 & 0 & p & q & \ast \\ 0 & 0 & 0 & 1 & 0 & p & r \\ 0 & 0 & 0 & 0 & 1 & 0 & p \\ 0 & 0 & 0 & 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 \\ \end{bmatrix}$$ and $H_1/H_2$ embeds into $\mathbb{G}_a^3$. We compute that $a(x) b(y) a(x) b(y)$ has $(p,q,r) = (xy, x^2 y, x y^2)$ and every vector in $(\mathbb{F}_2^{\mathrm{alg}})^3$ is a sum of vectors of the form $(xy, x^2 y, x y^2)$, so $H_1/H_2 \cong \mathbb{G}_a^3$.

Computation of $H_2/H_3$ At this point, I already know that $\dim H \geq 5$, so it is not a $4$-dimensional unipotent group. I wasn't able to fully compute $H_2$. For $h \in H_2$, we definitely have $h_{15}=h_{37}$, so $\dim H_2 \leq 5$. Every element of $H_2$ that I have found obeys the additional equation $h_{15} = h_{26}$. I can find enough elements to generate the $4$-dimensional subspace cut out by these equations. So $H_2$ is a subgroup of $\mathbb{G}_a^5$ which contains at least $\mathbb{G}_a^4$.

$\endgroup$
  • $\begingroup$ Nice! Thanks for putting lots of details in the computation. Good luck with your question, even if this direction is a dead end :) $\endgroup$ – Dror Speiser Mar 27 at 22:48
  • $\begingroup$ Thanks. I found enough elements now to get my bound for $\dim H_2$ up to $4$. $\endgroup$ – DES-SupportsMonicaAndTransfolk Mar 27 at 23:23

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.