Let $G$ be a connected, linear algebraic group over $\mathbb{C}$. Let $\rho:G\to \mathrm{GL}(V)$ be a rational representation. Does $\rho$ have at most countably many subrepresentations (up to isomorphism)?

Cross-posted https://math.stackexchange.com/questions/4797754/subrepresentations-of-a-representation-of-a-unipotent-algebraic-group

  • 2
    $\begingroup$ Does "up to isomorphism" mean the isomorphism type of the sub-representation, or up to conjugation by a $\rho$-equivariant automorphism of $V$? $\endgroup$ May 15 at 14:48
  • $\begingroup$ Up to isomorphism as representations of $G$. $\endgroup$
    – Doug Liu
    May 15 at 14:53
  • 3
    $\begingroup$ Side note: a reductive group has only countably many representations up to isomorphism. And so does the 1-dimensional unipotent group. $\endgroup$
    – YCor
    May 15 at 14:56
  • 1
    $\begingroup$ @PseudoNeo, re, Jantzen - Representations of algebraic groups is, I think, a standard reference. $\endgroup$
    – LSpice
    May 15 at 19:23
  • 1
    $\begingroup$ @KentaSuzuki I should have said "constructible", not "Zariski-closed"; it is enough to have this finite-continuum dichotomy. Let $E$ be the (Zariski-closed) subset of the Grassmanian of $V$ consisting of invariant subspaces. The set of triples $(W_1,W_2,f,g)$ where $W_1,W_2\in E$ and $G$-invariant, and $f,g$ are inverse equivariant homomorphisms $W_1\to W_2$, $W_2\to W_1$, is Zariski-closed in $E^2\times\mathrm{Mat}(V)^2$. So its projection to $E^2$ is constructible. This is exactly the equivalence relation. $\endgroup$
    – YCor
    May 15 at 22:46

1 Answer 1


No. Let $G$ be the 2-dimensional unipotent abelian group, and its 3-dimensional representation given by $$(t,s)\mapsto \begin{pmatrix}1 & t & s\\ 0 & 1 & 0\\ 0 & 0 & 1\end{pmatrix}.$$

Let $G_u$ be the subgroup of $G$ of those $(t,ut)$, $t\in\mathbf{C}$. Then the fixed point set $V_u$ of $G_u$ consists of triples $(x,-uz,z)$, $x,z\in\mathbf{C}$. Then $V_u$ is a subrepresentation of $V$ whose kernel is exactly $G_u$. In particular, the $V_u$, for $u\in\mathbf{C}$, are pairwise non-isomorphic representations.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.