3
$\begingroup$

Consider two (possibly infinite-dimensional) representations $\rho$, $\pi$ of a semisimple Lie algebra $\mathfrak{g}$, with $\rho$ irreducible and $\pi$ indecomposable but not irreducible (i.e., not simple). Are there any circumstances under which the product representation $\rho\otimes\pi$ can be semisimple (isomorphic to a direct sum of irreducible representations)?

In other words: is the product of a simple and a non-semisimple representation always non-semisimple?

Motivation

I'm working in particular with $\mathfrak{spin}(3,1)$. The situation is the following: I have a class of irreducible finite-dimensional representations (the $\rho$ above) that I know how to couple with generic irreducible representations $\sigma$ (that is, I know under which conditions the resulting representation is semisimple or not).
I need to find out if the product $\rho_1\otimes\rho_2\otimes\sigma$ is semisimple: if $\rho_2\otimes\sigma$ is, then I can consider the product with $\rho_1$ for each one of the irreducibles in the decomposition to obtain the answer; if, however, $\rho_2\otimes\sigma$ is indecomposable but not irreducible, can I be sure that $\rho_1\otimes\rho_2\otimes\sigma$ is not semisimple as well?
I'm expecting the resulting representation to be not semisimple, but I need an answer to my question to be able to prove it.

Update #1

As Will Sawin pointed out, if $\rho$ is finite dimensional and $\rho\otimes\pi$ happens to be semisimple, then it must necessarily be $\rho^*\otimes\rho\otimes\pi$ non-semisimple, since $\rho^*\otimes\rho$ is semisimple and contains the trivial representation in its decomposition. Unfortunately this doesn't contradict anything, as it is possible for the product of two semisimple representations to be non-semisimple.

$\endgroup$
7
  • $\begingroup$ I'm not sure what the answer is, but does it have consequences (either way)? There tend to be lots of exotic infinite dimensional representations of such Lie algebras, so it's hard to predict what is possible. There don't seem to be examples in the BGG category of such semisimple tensor products with $\rho$ finite dimensional and $\pi$ of course not; but this category of modules is very restrictive. If I had to guess, I'd expect to find no examples of the type you describe. $\endgroup$ Commented Jun 23, 2015 at 23:25
  • $\begingroup$ @JimHumphreys: I added some motivation for my question, to explain the consequences. I'm expecting not to find any example as well, but I need to be sure. $\endgroup$ Commented Jun 24, 2015 at 3:38
  • 1
    $\begingroup$ If $\rho$ is finite-dimensional, the reverse has to happen also - $(\rho \otimes \pi)$ is semsimple but $\rho^\vee \otimes (\rho \otimes \pi)$ is not. $\endgroup$
    – Will Sawin
    Commented Jun 24, 2015 at 13:15
  • $\begingroup$ @WillSawin: Is $\rho^\vee$ the dual representation in your notation? Can you elaborate on why this has to happen? $\endgroup$ Commented Jun 24, 2015 at 15:06
  • 2
    $\begingroup$ Yes, that's the dual. Because $\rho$ is finite dimensional, $1$ is a summand of $\rho^\vee \otimes \rho$, so $\pi$ is a summand of $\rho^\vee \otimes \rho \otimes \pi$. $\endgroup$
    – Will Sawin
    Commented Jun 24, 2015 at 17:36

0

You must log in to answer this question.

Browse other questions tagged .