0
$\begingroup$

Let $R$ be a finite-dimensional irreducible representation of $U(N)$, with the set of weights $W_R$. Each element of $W_R$ is a vector of length $N$ with integer entries.

Firstly, I would like to know if it is true that $$\sum_{\mu\in W_R}m_\mu\mu=A(R)(1,\ldots,1),$$ where $m_\mu$ is the multiplicity of the weight $\mu$ and $A(R)$ is some integer that varies with the representation. Note that we are considering $U(N)$ rather $SU(N)$, because I think the sum of weights of an irrep vanishes for $SU(N)$.

Secondly, I would like to know if there is a simple formula for $A(R)$, or if there is an existing name for it?

$\endgroup$
10
  • $\begingroup$ You are just taking a bare sum of weights, without any multiplicity (according to the dimension of the weight space)? $\endgroup$
    – LSpice
    Jan 25, 2023 at 22:37
  • 2
    $\begingroup$ @LSpice Sorry I meant to include multiplicities; the question has been edited. $\endgroup$ Jan 25, 2023 at 22:49
  • $\begingroup$ Since $m_\mu$ is constant on Weyl orbits (because we're dealing with a representation of $G$), $\sum m_\mu\mu$ is Weyl-fixed. The Weyl-fixed sublattice of the character lattice of $\operatorname{SU}(N)$ is trivial, which is why you always get $0$ there; and is spanned by the character $\operatorname{det} = (1, \dotsc, 1)$ you indicate for $\operatorname U(N)$, so, indeed, you always get your desired equality. Possibly $A(R)$ can be computed in terms of the highest weight by the Kostant multiplicity formula. $\endgroup$
    – LSpice
    Jan 25, 2023 at 23:15
  • $\begingroup$ @LSpice Thanks for the reply! Do you know a simple proof (or reference) for why the weyl-fixed character lattice of $SU(N)$ is trivial? $\endgroup$ Jan 26, 2023 at 0:08
  • 1
    $\begingroup$ @LSpice Thanks! now it's obvious... $\endgroup$ Jan 26, 2023 at 0:32

1 Answer 1

2
$\begingroup$

As discussed in the comments, your sum is a Weyl-fixed character, so trivial for $G = \operatorname{SU}(N)$ and a multiple of $\det = (1, \dotsc, 1)$ for $\operatorname U(N)$.

To be concrete, as I guessed in the comments, one sees that your sum $\sum_\mu m_\mu\mu$ (the sum taking place in the character lattice $X^*(T)$ of the (implicitly chosen) maximal torus $T$, not in $\mathbb C[X^*(T)]$ as in the Weyl character formula) is precisely the character of $\det \circ R$. Again, since $\det \circ R$ is Weyl invariant and $\det$ spans the Weyl-invariant part of the character lattice, your integer $A(R)$ is precisely the integer $n$ such that we have $\det \circ R = \det(\cdot)^n$.

Inspired by your comment, I realise we can be a little more explicit. (On having written this, I realise that it was actually exactly what you were saying; I mistook your $\lambda_i$ for an indexing of weights, rather than a component of $\lambda$. Oops, sorry!)

Let $\lambda$ be the highest weight of $R$. Then all weights $\mu$ of $R$ agree on the centre of $\operatorname U(N)$ with $\lambda$. Specifically, they all act as $z I_N \mapsto z^\ell$ for some integer $\ell$. (If we think of $\lambda$ as an element of $\mathbb Z^N$, then $\ell$ is the sum of the components.) Thus, we have that $R$ agrees on the centre with $z I_N \mapsto z^\ell I_{\dim(R)}$, so $\det \circ R$ agrees on the centre with $z I_N \mapsto \det(z^\ell I_{\dim(R)}) = z^{\ell\dim(R)} = \det(z I_N)^{\ell\dim(R)/N}$, so $A(R)$ equals $\ell\dim(R)/N$. As you point out, we can use the Weyl dimension formula to compute $\dim(R)$ in terms of $\lambda$ if desired.

$\endgroup$

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.