Classification of complex semisimple Lie algebras using Dynkin diagrams August Geelmuyden University of Oslo Part I Root systems I. Symmetry of a vector Definition 1. Let α 6= 0 be a vector in a vector space V. A Symmetry of the vector α is a linear transformation sα ∈ Aut(V) satisfying (S1) sα (α) = −α. (S2) sα fixes a hyperplane H. Proposition 1. There exist at most one sα . Proof. Let α ∈ V be a non-zero vector and let s and s 0 be two symmetries for α. Since α extends the basis for H to a basis for V every vector x ∈ V can be written in the form x = β + Cα where β ∈ H and C is a constant from the underlying field. Then s(x) = s(β) + Cs(α) = β − Cα since s is a symmetry for the vector α. Since s 0 is also a symmetry for α, we have s 0 (x) = β − Cα and thus s 0 (x) = s(x). Since x was arbitrarily chosen, the proof is complete. Proposition 2. Every sα is of the form sα (x) = x − hα∗ , xiα. Proof. Since there exist at most one sα , it suffices to show that sα (x) = x − hα∗ , xiα is a symmetry for α. First note that the map x 7→ hα∗ , xiα is a linear map by definition. Since hα∗ , βi = 0 for all β ∈ H it follows that sα fixes H. Finally, since hα∗ , αi = 2, we have sα (α) = −α. II. Root systems Definition 2. A set R ⊂ V is a root system if (R1) R does not contain 0, is finite and spans V. (R2) For all α ∈ R there is a symmetry sα with sα (β) ∈ R for all β ∈ R. (R3) For all α, β ∈ R there is some n ∈ Z so that sα (β) − β = nα. One one-dimensional, four two-dimensional Definition 3. A root system R is said to be irreducible if for all α ∈ R the only distinct vector being proportional to α is −α. 1 III. Weyl group Proposition 3. If R is a root system, then the set W of all symmetries sα of vectors from R form a subgroup of Aut(V) called the Weyl group. Proposition 4. There exist a symmetric, positive definite bilinear form (., .) that is invariant under the Weyl group. p Definition 4. The length of an element α ∈ R is defined as |α| = (α, α). The angle θ between two elements α and β of a root system R is given by the relation (α, β) = |α||β| cos θ. hα∗ , xi = 2 (α, β) ∈Z (α, α) Definition 5. n(α, β) = 2 (α, β) . (α, α) (1) Proposition 5. Any non-empty subplane of a root system R is itself a root system. Proposition 6. There is exactly seven distinct planar root systems. Proposition 7. If α and β are elements in a root system R with (α, β) > 0, then α − β ∈ R. Proposition 8. If R is a root system and α, β ∈ R spans an angle φ 6 π then there exist a pair of elements γ, ρ ∈ R spanning an angle nφ for n ∈ Z. IV. Bases Definition 6. A set B ⊂ R is called a basis for the root system R if it spans V and for all P α ∈ R there is a set of integers {nb } with the same sign satisfying α = b∈B nb b. Choose t ∈ V ∗ so that ht, αi = 6 0∀α ∈ R and define R+ t = α ∈ R ht, αi > 0 . + + It is clear that this induces a separation R = (R+ t ) ∪ (−Rt ) and that Rt spans the same space as R. Definition 7. A root α ∈ R+ t is said to be decomposable if there exists β, γ ∈ R so that α = β + γ. A root that is not decomposable is said to be indecomposable. Proposition 9. The set St of all indecomposable roots is a base for R. Isomorphic systems and well-defined-ness Let S and S 0 be bases for the two root systems R and R 0 respectively, being subsets of the two vector spaces V and V 0 . Let Φ : S → S 0 be a bijection satisfying n(Φ(α), Φ(β)) = n(α, β) for all α, β ∈ S. Then a unique isomorphism f : V → V 0 exists which is an extension of Φ to R. First note that if α, β ∈ S then sΦ(α) ◦ f (β) = sΦ(α) ◦ Φ(β) = Φ(β) − n (Φ(α), Φ(β)) Φ(α) = Φ(β) − n (α, β) Φ(α) and (f ◦ sα ) (β) = f (β − n(α, β)α) = f(β) − n(α, β)f(α) = Φ(β) − n(α, β)Φ(α). That is, f ◦ sα = sΦ(α) ◦ f. If W and W 0 are the weyl groups of R and R 0 respectively, then we may note that R = WS and R 0 = W 0 S 0 . Thus R 0 = W 0 S 0 = W 0 f(S) = f(WS) = f(R). Hence R ' R 0 . 2 V. Cartan matrix Let S = {αi }i∈I be a base for a root system R. Then the matrix n(αi , αj ) determines R. VI. Coxeter graph Let nij be the cartan matrix and define Nij = nij nji . Let Nij denote the number of vertices connecting the nodes i and j. VII. Dynkin diagram Part II Semisimple Lie algebras Lie algebra Definition 8 (Lie algebra). A lie algebra g is a vector space over a field F equipped with a bilinear binary operation [., .] : g × g → g, called the lie bracket, satisfying i) (Alternativity) [x, x] = 0 for all x ∈ g. ii) (Jacobi identity) [x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0 for all x, y, z ∈ g. Bilinearity may be used together with alternativity to show that the lie bracket satisfies anticommutativity [x, y] = −[y, x] for all x, y ∈ g. Definition 9 (Adjoint representation). If g is a lie algebra over a field F then the representation ad : g → End(g) given by x 7→ adx is called the adjoint representation of the algebra g. On End(g), which is the set of all homomorphisms from g to g, the lie bracket is defined to be [adx , ady ] = adx ◦ ady − ady ◦ adx . For lie algebras g of finite dimension, the fact that End(g) is isomorphic to the general linear group gl(g) is often used. Definition 10 (Adjoint action). If x ∈ g is an element in a lie algebra over a field F then the adjoint representation of x is the adjoint action adx : g → g where adx (y) = [x, y]. It may well be noted that the adjoint action is, by the jacobi identity, a derivation adx [y, z] = [x, [y, z]] = [y, [x, z]] + [[x, y], z] = [adx (y), z] + [y, adx (z)] 3 Definition 11 (subalgebra and ideal). A vector subspace s of a Lie algebra g is a subalgebra if [x, y] ∈ s for all x, y ∈ s. If a Lie subalgebra s satisfies [x, y] ∈ s for all x ∈ g and s, then s is said to be an ideal of g. Semisimplicity Definition 12 (Central series). For a lie algebra g the central series is a descending series of ideals {Cn g}n>1 so that C1 g = g Cn g = g, Cn−1 g It follows that C2 g = [g, g] and [Cn g, Cm g] ⊂ Cn+m g. Definition 13 (Nilpotent). If there exist a positive integer n so that Cn g = 0 then g is said to be nilpotent. More precisely g is said to be nilpotent of class 6 n. Definition 14 (Derived series). For a lie algebra g the derived series is a descending series of ideals {Dn g}n>1 so that D1 g = g Dn g = Dn−1 g, Dn−1 g Definition 15 (Solvable). If there exist a positive integer n so that Dn g = 0 then g is said to be solvable. More precisely g is said to be solvable of length 6 n. Notice that if g is solvable of order 2, then g is abelian. Theorem 1. Every quotient, subalgebra and extension of a solvable lie algebra is solvable. Theorem 2. There exists a maximal solvable ideal. Definition 16 (Radical). The maximal solvable ideal of a lie algebra g is called the radical r of g. Definition 17 (Semisimplicity). A lie algebra g is said to be semisimple if its radial r is zero. Definition 18 (Invariance). A symmetric bilinear form B : g × g → F is said to be invariant if B ([x, y] , z) + B (y, [x, z]) = 0 ∀x, y, z ∈ g. The killing form B(x, y) = Tr(adx ◦ ady ) is both invariant and symmetric. Definition 19 (Orthogonal space). The orthogonal space a⊥ of a subset a ⊆ g relative to a bilinear form B : g × g → F is given as the set a⊥ = x ∈ gB(x, y) = 0 for all y ∈ a . Definition 20 (Degenerate). A bilinear form i said to be degenerate on a Lie algebra g if there is a nonzero element x in g⊥ . If g⊥ = 0, then the bilinear form is said to be nondegenerate. Theorem 3 (Cartan-Killing Criterion). A Lie algebra is semisimple if, and only if, its Killing form is nondegenerate. 4 Cartan subalgebras Definition 21 (Cartan subalgebra). A subalgebra h of g is called a Cartan subalgebra if it is nilpotent and its own normalizer. Definition 22 (nilspace). Let g be a complex Lie algebra and fix an element x in g. If λ ∈ C the nilspace of adx − λI is the space gλx = y ∈ g∃n ∈ N : (adx − λI)n y = 0 . The space g0x is referred to as the generalized nullspace of x. This is a decomposition of g into invariant subspaces of adx . That is, for any x ∈ g we may write M g= gλx . λ∈C Definition 23 (regular element). An element x of a complex Lie algebra g is said to be regular if its generalized nullspace has minimal dimension. It should now be noted that the generalized nullspace g0x of x is a subalgebra of g. If x is regular, then g0x is a Cartan subalgebra og g. Factorization of complex semisimple Lie algebras For a semisimple, complex Lie algebra any Cartan subalgebra is abelian and its own centralizer. It so turns out that any Cartan subalgebra is a maximal abelian subalgebra. Let x be an element of a Cartan subalgebra h of a complex semisimple Lie algebra g. Then the endomorphism adx is diagonalizable. Definition 24 (eigen-subspaces). Let g be a complex, semisimple Lie algebra and h a Cartan subalgebra of g. For α ∈ h∗ let gα = x ∈ g[y, x] = α(y)x for all y ∈ h An element α ∈ h∗ is said to be a root of g if α 6= 0 and gα 6= 0. Theorem 4. If R is the set of all roots of a complex, semisimple Lie algebra g, then M g'h⊕ gα . α∈R Proof. The endomorphisms ady of g for y ∈ h are diagonalizable. 5