# higher Casimirs for $\mathfrak{sl}$

The Wikipedia universal enveloping algebra suggest a way to obtain higher Casimir operators (e.g. generators of the center of $\mathfrak{U(g)}$ for $\mathfrak{g}$ semisimple) by evaluating certain determinant:

$$\det\, (t\mathrm{I} - \mathrm{ad}_X) = \sum_{i=0}^{\dim \mathfrak{g}} p_i(X) t^i$$

However if one does that for $\mathfrak{sl}_3$ then the resulting polynomial has only degrees (in $t$) 0,1,2,4,6,8 and it's coefficients seem to be just powers of the quadratic Casimir operator. If one tries to do the same for the defining representation $\mathbb{C}^3$ (replacing $\mathrm{ad}_X$ by $\rho(X)$ in the above formula) then one obtains quadratic as well as cubic invariant polynomials.

In the proof of Harish-Chandra isomorphism as presented e.g. in (1) there is construction of elements of $Z(\mathfrak{U(g)}$ using traces of matrices from representations of $\mathfrak{g}$. Something like $$\sum\mathrm{tr}(\rho(X_{i_1})\rho(X_{i_2})\ldots \rho(X_{i_n}))X_{i_1}^*X_{i_2}^*\ldots X_{i_n}^*$$ where $X_i$ form basis for $\mathfrak{g}$ and $X_i^*$ form dual basis with respect to Killing form.

Q1: What is going on here?

Q2: Is it true that for a semi-simple complex Lie algebra and it's smallest nontrivial representation one obtains in this way all generators of the $Z(\mathfrak{U(g)}$?

Q3: Does the approach through determinant give the same operators that appear in the proof of the H-Ch isomorphism?

(1) Cohomological Induction and Unitary Representations by Knapp, Vogan

• Related questions: 1. cubic Casimir for sl3 - mathoverflow.net/questions/188266/… Apr 26 '19 at 13:53
• Related questions 2. calculating central elements - math.stackexchange.com/questions/1417897/… Apr 26 '19 at 13:53
• None of the above have a full answer for calculating all central elements for all semisimple Lie algebra yet. Apr 26 '19 at 13:54

For $\mathfrak{sl}(3)$ the coefficients are not powers of the Casimir, though the polynomial is one in the square of $t$.
• To be completely honest we haven't tried all coefficients but we checked that the highest degree $p_i$ is indeed a power of Casimir. I've checked again and I discovered I've missed linear term that is there. But all the other odd degrees of $t$ are missing. Sep 18 '18 at 17:15
• Does "the polynomial is one in the square of $t$" mean "the polynomial has non-$0$ terms only in even degrees"? Sep 18 '18 at 19:38
• The coefficients are naturally elements of $\mathcal S(\mathfrak g^*)$ (and not $\mathcal U(\mathfrak g^*)$). If you symmetrize the coefficient of $t^6$ and evaluate it at the dual basis, you obtain the quadratic Casimir. However, the cubic one is missing, since the coefficient of $t^5$ is $0$. There is a non-zero coefficient next to $t$ (which is of degree $7$), so cubic Casimir might be hidden there? Sep 19 '18 at 1:01