Representation theory

Dynkin diagrams

A Dynkin diagram is a picture of a Cartan matrix. Its type, finite, affine, hyperbolic or indefinite, is decided by a symmetric form. For the roots themselves, in rank two and three, see Root systems.

Dynkin and Coxeter diagrams

Cartan matrix

The Cartan matrix

A Cartan matrix is a square matrix A=(aij)A=(a_{ij}) of integers with aii=2a_{ii}=2, with aij≤0a_{ij}\le0 for i≠ji\ne j, and with aij=0a_{ij}=0 exactly when aji=0a_{ji}=0 (a generalized Cartan matrix in [2, Ch. 1]). The simple roots α1,…,αn\alpha_1,\dots,\alpha_n of a root system give a Cartan matrix, with entries

aij=2(αi,αj)(αi,αi).a_{ij}=\frac{2(\alpha_i,\alpha_j)}{(\alpha_i,\alpha_i)}.

The Dynkin diagram of a Cartan matrix has a node for each ii. Nodes ii and jj are joined when aij≠0a_{ij}\ne0, by max⁡(∣aij∣,∣aji∣)\max(\lvert a_{ij}\rvert,\lvert a_{ji}\rvert) lines, with an arrow pointing to node ii when ∣aij∣>1\lvert a_{ij}\rvert>1; for a root system, the arrow points to the shorter root. When aijaji>4a_{ij}a_{ji}>4 they are joined instead by one thick line labelled ∣aij∣,∣aji∣\lvert a_{ij}\rvert,\lvert a_{ji}\rvert [2, Ch. 4]. So a single line is aij=aji=−1a_{ij}=a_{ji}=-1, and a double line with its arrow pointing to node ii is aij=−2a_{ij}=-2 and aji=−1a_{ji}=-1.

A Coxeter diagram may also have bonds labelled mm, which stand for aij=aji=−2cos⁡(π/m)a_{ij}=a_{ji}=-2\cos(\pi/m) [1, Ch. 5]. The reflections si(αj)=αj−aijαis_i(\alpha_j)=\alpha_j-a_{ij}\alpha_i generate the Weyl group of a Cartan matrix [2, Ch. 3], or the Coxeter group of a Coxeter diagram.

Finite, affine and indefinite

The eigenvalues of AA decide the type of a connected diagram. It is of finite type when they are all positive, of affine type when one is 00 and the others are positive, and of indefinite type otherwise [1, Ch. 6; 2, Ch. 4].

This holds when AA is symmetrizable, that is, when there are positive numbers did_i with diaij=djajid_ia_{ij}=d_ja_{ji}. Then (αi,αj)=diaij(\alpha_i,\alpha_j)=d_ia_{ij} is a symmetric bilinear form on the span of the simple roots, the eigenvalues of AA are real, and their signs give the signature of the form. When a matrix is not symmetrizable, its eigenvalues need not be real, and it is of indefinite type.

For a diagram of affine type the kernel of the form is spanned by δ=∑iciαi\delta=\sum_ic_i\alpha_i, with positive integers cic_i, the numbers on the nodes. A diagram of indefinite type is hyperbolic when every proper connected subdiagram is of finite or affine type [2, Ch. 5].

The Coxeter plane

For a diagram of finite type the Weyl or Coxeter group WW is finite. A Coxeter element c=s1s2⋯snc=s_1s_2\cdots s_n, the product of the simple reflections in any order, has order hh, the Coxeter number, and its eigenvalues are

e2πi m1/h, …, e2πi mn/h,\mathrm{e}^{2\pi\mathrm{i}\,m_1/h},\ \dots,\ \mathrm{e}^{2\pi\mathrm{i}\,m_n/h},

where m1≤⋯≤mnm_1\le\dots\le m_n are the exponents [1, Ch. 3]. The Coxeter plane is the plane on which cc acts by the rotation through 2π/h2\pi/h. The nhnh roots fall into nn orbits of cc, each of hh roots, and each orbit projects to the plane as a regular hh-gon [1, Ch. 3]. Each pair of exponents mm and h−mh-m gives a plane on which cc turns by 2πm/h2\pi m/h; from rank 33 on, the figure projects the roots to one of these planes, and for depth to one more direction.

Heights and exponents

A positive root β=∑iciαi\beta=\sum_ic_i\alpha_i has height ∑ici\sum_ic_i. For a diagram of finite type the numbers of positive roots of heights 1,2,3,…1,2,3,\dots form a partition, and the conjugate partition is formed by the exponents [3]. So there are nn roots of height 11, the simple roots, and the highest root has height mn=h−1m_n=h-1.

For a diagram of affine type a real root plus rδr\delta is again a real root, where rr is the rr of the name Xℓ(r)\mathrm{X}_\ell^{(r)}, so the numbers repeat. For affine and indefinite types there are infinitely many real roots, and the figure counts them up to height 3030.

References

  1. J. E. Humphreys, Reflection groups and Coxeter groups. Cambridge Studies in Advanced Mathematics 29, Cambridge University Press (1990)
  2. V. G. Kac, Infinite-dimensional Lie algebras. 3rd ed., Cambridge University Press (1990)
  3. B. Kostant, The principal three-dimensional subgroup and the Betti numbers of a complex simple Lie group. Amer. J. Math. 81 (1959), 973–1032