For any Lie algebra symmetrization maps Poisson center of S(g) to center of U(g). Consider g=gl_n, the Poisson center of S(gl_n) is isomorphic to algebra of symmetric polynomials in eigenvalues. Consider some element c \in S(gl_n), which corresponds to some symmetric polynomial P(l1,...ln).

Question: is there something known about value of symmetrization(c) in irrep of highest weight (w1,...wn), i.e. what is the Harish-Chandra image of symmetrization(c) ?

We should expect that this value is P(w1,...,wn)+correction, but it is quite difficult to calculate this "correction".

For sl(2) the answer is known - see Kirillov's paper "Merits and demerits of orbit method" section 3.5 (http://www.ams.org/bull/1999-36-04/S0273-0979-99-00849-6/S0273-0979-99-00849-6.pdf). Answer and proof are rather non-trivial and obtained with the help of the Duflo map...


Well, long time passed and no answer, I also asked by the e-mail M. Duflo, A. Kirillov, E. Vinberg, G. Olshanski, so I got the impression that it seems indeed uknown.

  • $\begingroup$ You should not be too presumptious based on the lack of response at MO: I, for one, missed your question the first time around. Some things are certainly known, e.g. didn't Okounkov and Olshanskii investigate this question ("shifted Schur functions")? $\endgroup$ – Victor Protsak Dec 12 '11 at 16:23
  • $\begingroup$ @Victor Okounkov Olshanskii is not related to this question. They consider "special" symmetrization, in opposite to the "standard" one I am asking about. The "special" symmetrization introduced by Olshaskii can be seen in two ways: 1) consider representation of GL_n by right-invariant vector fields E=XD^t and sym(E...E)= WickOrdering(XD) i.e. put "X" to the left "D" to the right - that is why it is related to Capelli identities 2) see second comment $\endgroup$ – Alexander Chervov Dec 12 '11 at 17:20
  • $\begingroup$ 2) "general" symmetrization (in the sense of Olshanki) comes from ANY local (nearby origins) identification of g and G - then U(g) as delta functions at "e" gots identified with S(g). Standard symmetrization comes from exponential mapping. "Special" comes from X-> 1+X for gl_n. For other semisimple it appears to be more tricky with "special" one. $\endgroup$ – Alexander Chervov Dec 12 '11 at 17:24
  • $\begingroup$ Ok, I think that I now understand your question (it wasn't clear from the original post). Let $$\sigma: S(\mathfrak{g}) \to U(\mathfrak{g})$$ be the standard symmetrization map, $$HC: S(\mathfrak{h})^W \to Z(\mathfrak{g})$$ be the Harish-Chandra homomorphism and $$HC_{qc}: S(\mathfrak{h})^W \to S(\mathfrak{g})^{\mathfrak{g}}$$ be its quasi-classical version. Your question, then, concerns an explicit form of the composite map $$HC^{-1}\circ\sigma\circ HC_{qc}: S(\mathfrak{h})^W \to S(\mathfrak{h})^W.$$ I am busy this week (exams), but if it's what you want, I'll think about it later. $\endgroup$ – Victor Protsak Dec 12 '11 at 21:47
  • $\begingroup$ @Victor. Yes absolutely correct. You are welcome to edit my post to make it more readable. As a small remark, I think it is necessary to emphasize that there are two actions of W on S(h) - standard one and shifted by \rho - the image of HC is invariant with respect to shifted action. If you have some interest in the question I can share some calculations and the letter by M. Duflo how to approach this. But I am NOT sure this is "good" question with "nice" answer... May be Kirillov's "nice" answer only for sl_2. $\endgroup$ – Alexander Chervov Dec 13 '11 at 5:59

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.