As explained in Section 4.9.2, the restricted root system of the real form can be either reduced or non-reduced. If it is reduced, it corresponds to one of the root systems of the finite-dimensional simple Lie algebras. On the other hand, if the restricted root system is non-reduced, it is necessarily of -type  (see Figure 19 for a graphical presentation of the root system).
By definition, the restricted Weyl group is the Coxeter group generated by the fundamental reflections, Equation (4.55), with respect to the simple roots of the restricted root system. The restricted Weyl group preserves multiplicities .
Although multiplicities are an essential ingredient for describing the full symmetry , they turn out to be irrelevant for the construction of the gravitational billiard. For this reason, it is useful to consider the maximal split “subalgebra” , which is defined as the real, semi-simple, split Lie algebra with the same root system as the restricted root system as (in the -case, we choose for definiteness the root system of to be of -type). The real rank of coincides with the rank of its complexification , and one can find a Cartan subalgebra of , consisting of all generators of which are diagonalizable over the reals. This subalgebra has the same dimension as the maximal noncompact subalgebra of the Cartan subalgebra of .
By construction, the root space decomposition of with respect to provides the same root system as the restricted root space decomposition of with respect to , except for multiplicities, which are all trivial (i.e., equal to one) for . In the -case, there is also the possibility that twice a root of might be a root of . It is only when is itself split that and coincide.
One calls the “split symmetry algebra”. It contains as we shall see all the information about the billiard region . How can be embedded as a subalgebra of is not a question that shall be of our concern here.
The purpose of this section is to use the Iwasawa decomposition for described in Section 6.4.5 to derive the scalar Lagrangian based on the coset space . The important point is to understand the origin of the similarities between the two Lagrangians in Equation (5.45) and Equation (7.8) below.
The full algebra is subject to the root space decompositionnot the full Cartan subalgebra since the compact part of belongs to .
This implies that when constructing a Lagrangian based on the coset space , the only part of that will show up in the Borel gauge is the Borel subalgebra
More specifically, an (-dependent) element of the coset space takes the form
By comparing Equation (7.8) with the corresponding expression (5.45) for the split case, it is clear why it is the maximal split subalgebra of the U-duality algebra that is relevant for the gravitational billiard. Were it not for the additional sum over multiplicities, Equation (7.8) would exactly be the Lagrangian for the coset space , where is the maximal compact subalgebra of (note that ). Recall now that from the point of view of the billiard, the positive roots correspond to walls that deflect the particle motion in the Cartan subalgebra. Therefore, multiplicities of roots are irrelevant since these will only result in several walls stacked on top of each other without affecting the dynamics. (In the -case, the wall associated with is furthermore subdominant with respect to the wall associated with when both and are restricted roots, so one can keep only the wall associated with . This follows from the fact that in the -case the root system of is taken to be of -type.)
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 2.0 Germany License.