\(\pi \)-systems of symmetrizable Kac-Moody algebras (Q2224522): Difference between revisions

A \(\pi\)-system of a symmetrizable Kac-Moody algebra \(\mathfrak{g}\) is a subset \(\Sigma\) of real roots such that the difference between any two roots in \(\Sigma\) is not a root. For finite-dimensional \(\mathfrak{g}\), Dynkin showed [\textit{E. B. Dynkin}, Am. Math. Soc., Transl., II. Ser. 6, 111--243 (1957; Zbl 0077.03404)] that \(\pi\)-systems are the simple roots of regular semisimple subalgebras of \(\mathfrak{g}\), that is, semisimple subalgebras which are invariant under the action of some Cartan subalgebra in \(\mathfrak{g}\). This nicely-written paper studies \(\pi\)-systems in more general Kac-Moody algebras, motivated by relations between Kac-Moody symmetries of dimensionally reduced (super)gravity theories in physics. For example, the hyperbolic Kac-Moody algebra \(E_{10}\) appears in \(11\)-dimensional supergravity, while the Feingold-Frenkel hyperbolic algebra \(A_1^{++}\) appears in \(4\)-dimensional Einstein gravity. Thus inclusion of Einstein gravity into \(11\)-dimensional supergravity should correspond to an inclusion of \(A_1^{++}\) into \(E_{10}\). One result in this paper is that indeed there is a \(\pi\)-system of type \(A_1^{++}\) in \(E_{10}\), and moreover it is unique up to negation and Weyl group conjugation. This suggests that there should be an inclusion of Einstein gravity into \(11\)-dimensional supergravity that is canonical in some sense. In this paper, the authors first clarify when \(\pi\)-systems actually correspond to Kac-Moody subalgebras: Given a linearly-independent \(\pi\)-system \(\Sigma\) in a root system of type \(A\), where \(A\) is a symmetrizable generalized Cartan matrix, the roots in \(\Sigma\) form the simple roots of a subsystem, say of type \(B\), and there is an injective homomorphism \(\mathfrak{g}(B)\rightarrow\mathfrak{g}(A)\) between corresponding Kac-Moody algebras. If the roots in \(\Sigma\) are not linearly independent, then one still gets a not-necessarily-injective map \(\mathfrak{g}'(B)\rightarrow\mathfrak{g}'(A)\) between derived subalgebras. Next, the authors show that any indecomposable \(\pi\)-system of a symmetrizable \(\mathfrak{g}\) is Weyl group conjugate to a \(\pi\)-system consisting entirely of positive or of negative roots; both possibilities occur if and only if \(\mathfrak{g}\) is finite dimensional. The remainder of the paper studies existence and classification of Weyl group orbits of \(\pi\)-systems of certain types. The authors reduce the classification of affine-type \(\pi\)-systems of symmetrizable \(\mathfrak{g}\) to the classification of such \(\pi\)-systems for affine-type subdiagrams of the Dynkin diagram of \(\mathfrak{g}\), and they show that there are always infinitely many Weyl group orbits (or none). They also classify overextended-type \(\pi\)-systems of simply-laced \(\mathfrak{g}\) in terms of overextended-type subdiagrams of the Dynkin diagram of \(\mathfrak{g}\). Specifically, let \(X\) be the (symmetric) Dynkin diagram of \(\mathfrak{g}\) and \(K\) an over-extended-type Dynkin diagram. The authors show that \(\mathfrak{g}\) has a \(\pi\)-system of type \(K\) if and only if \(X\) has an overextended-type subdiagram \(Z\) such that the finite part \(Z^{\circ}\) of \(Z\) has a \(\pi\)-system of type \(K^\circ\). Moreover, up to negation and Weyl group conjugation, there is a unique \(\pi\)-system of type \(K\) in \(\mathfrak{g}\) for each such subdiagram \(Z\) and each Weyl group orbit of \(\pi\)-systems of type \(K^\circ\) in \(Z^\circ\). In the case that \(K\) is the simplest overextended-type diagram \(A_1^{++}\), the last result shows that Weyl group orbits of \(\pi\)-systems of type \(A_1^{++}\) are classified simply by overextended-type subdiagrams of the Dynkin diagram of \(\mathfrak{g}\). For example, the Dynkin diagram of type \(E_{10}\) has a single overextended-type subdiagram, namely itself, so \(E_{10}\) has two Weyl group orbits of \(A_1^{++}\)-type \(\pi\)-systems: one has a representative consisting entirely of positive roots, and the other has a representative consisting entirely of negative roots. Each such \(\pi\)-system yields an \(A_1^{++}\)-subalgebra in \(E_{10}\), and all these subalgebras are conjugate under the action of automorphisms of \(E_{10}\).