I have the following (real) quantities (which are from a Classical Mechanics problem):
$$A_1=\frac 1 4(x^2 +p_x^2-y^2-p_y^2 ) \quad A_2=\frac 1 2(x y +p_x p_y)$$
$$A_3=\frac 1 2(x p_y - y p_x )$$
This should be a Poisson algebra with the following Lie bracket:
$$\{f , g\} =\frac{df}{dx}\frac{dg}{dp_x}+\frac{df}{dy}\frac{dg}{dp_y}-\frac{dg}{dx}\frac{df}{dp_x}-\frac{dg}{dy}\frac{df}{dp_y}$$
And if we define: $H=x^2 + y^2 + p_x^2+p^2_y $
I calculated the following relations: $$\{A_i,A_j\}=\sum_k\epsilon_{ijk}A_k$$
$$\{A_i,H\}=0$$
I see that $\epsilon_{ijk}$ are the structure constants, so this Poisson algebra is isomorphic to $\mathfrak{su}(2)$ and $\mathfrak{so}(3)$.
Shouldn't $H$ be a Casimir operator? But what is the universal enveloping algebra where it belongs to? I tried to construct a Casimir operator using the Killing form but it was not $H$.
And does this Lie algebra have an associated group? If $A$ where matrices, I could do $e^{A\theta}$, to obtain the group elements.
I think that my confusion arises because this is not a matrix algebra, where the Lie bracket is just the commutator.