The first four elements form the “distinguished fiber” \(\mathcal
F\). That is, those who, map to the (conjugacy class of) the identity
involution in the Weyl group. The 0 in the last collumn next to
the # sign tells us that these \(K\backslash G/B\) elements are in
the Compact Cartan subgroup and, up to conjugacy by \(K\), parametrize the
Borel subgroups containing the Compact Cartan subgroup. So, these parametrize
the discrete series of \(Sp(4, \mathbb R)\) with a fixed
infinitesimal character. That is, if we fix \(x_b\),
\[x=wx_b \rightarrow \text{discete series with Harish-Chandra parameter} \ w\rho.\]
There are a couple of commands that will
give you discrete series:
The first command deals with a single principal series. We want the last command that will list all discrete series with a fixed infinitesimal character. First let us setup a show function:
Again, as in the example of \(SL(2,R)\), these all have
lambda=rho and nu=0. The only difference is the element
x. In order to identify them with the usual parameters that we
know we do the following. We first fix a base element
\(x_b\). This determines an identity component of \(K\) and we set this as our \(K\)
This is a way to go from atlas parameters to the
Harish-Chandra parameters expressed, in the usual way, with
respect to the fixed base element. The one corresponding to x=2 is
the holomorphic discrete series, the one for x=3 is the
antiholomorphic one and the other two are the large discrete series.
This gives us more information about each representation. Namely, the status of the simple roots for the corresponding x.
The software always chooses, for the quasisplit group, x=0 to be
the large Borel; that is, both of the simple roots are non compact. In
this case the simple roots are \(e_1 + e_2\) and
\(2e_2\). Similarly, for x=1. So these correspond to the large
discrete series. And since we chose the base element to be x=2 and
the simple root for \(K\) is \([1,-1]\), then [2,1] is the usual
parameter for this choice of simple roots. The first simple root is
compact. so this corresponds to the holomorphic case.
Now to go the other way we use:
atlas> whattype discrete_series ?
Overloaded instances of 'discrete_series'
(KGBElt,ratvec)->Param
(RealForm,ratvec)->Param
atlas>
atlas> discrete_series (G, [2,1])
Value: final parameter(x=0,lambda=[2,1]/1,nu=[0,0]/1)
atlas> discrete_series (G, [2,-1])
Value: final parameter(x=2,lambda=[2,1]/1,nu=[0,0]/1)
atlas>
Or we could use the other format using the KGBElt:
In other words, the software conjugates the Harish-Chandra parameter [1,-2] to
[2,1] and conjugates, via the reflection on the long simple root,
the base element to x=0.
To find the elements in W that do this we do the following:
This is the entire Weyl group of type C2 in the form of a list of
pairs rootdatum,productofsimpleroots, starting from the
identity and ending in the long element of the Weyl group. Now to
find out how these elements act on x_b=2 we do:
This lists the cross action of each element of \(W\) on
x_b=2. The Id takes x=2 to itself, the simple reflection by
root [0], which is the compact root, also fixes it. The other
simple root, [1] sends x=2 to x=0 and so on.
Note that the action of \(W\) on this set is transitive and the stabilizer of x_b is \(W_K\)
Also, by contrast notice the action on the element x=10:
and suppose we use x=0 as our base point. Then we get a strange
set of parameters because the compact root is now [1,1] instead of
the usual one. So, the Harish-Chandrra parameters are not what we
expect:
That is what we get if we decide to define our group where \(K\)
is given by x=0. The compact root in this case is [1,1] and if
you we start with [2,1] you apply the Weyl group, modulo the
action of \(W_K\) to it, you get the above representatives.