Representation theory

Highest weight representations

An irreducible finite-dimensional representation of a simple Lie algebra is determined by its highest weight λ\lambda, and denoted V(λ)V(\lambda). Explore its weights and their multiplicities, tensor products, Verma modules, and the representations of affine Lie algebras.

Show
Module
Any weight

Lie algebras

Weights and multiplicities

The irreducible finite-dimensional representations of a simple Lie algebra g\mathfrak g are the V(λ)V(\lambda) with highest weight λ\lambda, one for each dominant integral weight λ\lambda. V(λ)V(\lambda) is the sum of its weight spaces V(λ)μV(\lambda)_\mu, on which a Cartan subalgebra h\mathfrak h acts by the weight μ\mu, and the multiplicity of μ\mu is m(μ)=dim⁡V(λ)μm(\mu)=\dim V(\lambda)_\mu. The weights are the points of λ+Q\lambda+Q, QQ the root lattice, in the convex hull of the orbit WλW\lambda of the Weyl group, and WW permutes them, preserving their multiplicities [4, §21].

So the multiplicities of the dominant weights determine all the others, and Freudenthal’s formula computes them one after another, starting from m(λ)=1m(\lambda)=1 [4, §22.3]:

m(μ)=2∑α>0∑k≥1(μ+kα,α) m(μ+kα)∣λ+ρ∣2−∣μ+ρ∣2,m(\mu)=\frac{2\sum_{\alpha>0}\sum_{k\ge1}(\mu+k\alpha,\alpha)\,m(\mu+k\alpha)}{\lvert\lambda+\rho\rvert^2-\lvert\mu+\rho\rvert^2},

where ρ\rho is half the sum of the positive roots. The multiplicities add up to the dimension, given by Weyl’s formula as a product over the positive roots [4, §24]:

dim⁡V(λ)=∏α>0⟨λ+ρ,α∨⟩⟨ρ,α∨⟩.\dim V(\lambda)=\prod_{\alpha>0}\frac{\langle\lambda+\rho,\alpha^\vee\rangle}{\langle\rho,\alpha^\vee\rangle}.

For sln\mathfrak{sl}_n, with λ=∑iλiϖi\lambda=\sum_i\lambda_i\varpi_i, the λi\lambda_i being the numbers on the nodes of its Dynkin diagram, λ\lambda corresponds to the partition with λi+⋯+λn−1\lambda_i+\dots+\lambda_{n-1} boxes in row ii. The multiplicity of μ\mu is the number of semistandard tableaux of this shape and weight μ\mu, where a tableau with cjc_j entries equal to jj has weight ∑i(ci−ci+1)ϖi\sum_i(c_i-c_{i+1})\varpi_i [2, Lecture 15].

Hexagons and triangles

For sl3\mathfrak{sl}_3 and λ=aϖ1+bϖ2\lambda=a\varpi_1+b\varpi_2, the weights lie on min⁡(a,b)\min(a,b) nested hexagons and, inside them, on triangles. The multiplicity is 11 on the outer hexagon, one more on each hexagon further in, and min⁡(a,b)+1\min(a,b)+1 on every triangle [2, Lecture 13].

The relief

The relief draws the multiplicities of V(λ)V(\lambda), in rank two, as the heights of columns, one on each weight. Its shape has a limit as λ\lambda is scaled up [3]. Let dd be the number of positive roots not orthogonal to λ\lambda, minus the rank. Shrunk by kk across and by kdk^d in height, the relief of V(kλ)V(k\lambda) tends to the graph of a function on the convex hull of WλW\lambda that is a polynomial of degree dd on each of finitely many pieces: the density of the Duistermaat–Heckman measure of the coadjoint orbit through λ\lambda.

For λ\lambda off the walls of the dominant chamber, dd is 11 for sl3\mathfrak{sl}_3, 22 for so5\mathfrak{so}_5 and 44 for g2\mathfrak g_2. For sl3\mathfrak{sl}_3 the limit rises linearly from the hexagon to the triangle and is flat on the triangle, as the multiplicities do.

Tensor products

The tensor product of two irreducible representations is a sum of irreducible ones,

V(λ)⊗V(ν)=⨁κnκ V(κ),V(\lambda)\otimes V(\nu)=\bigoplus_\kappa n_\kappa\,V(\kappa),

and the Brauer–Klimyk rule computes the multiplicities nκn_\kappa from the weights of V(ν)V(\nu) [4, §24]. Each weight μ\mu of V(ν)V(\nu), of multiplicity mν(μ)m_\nu(\mu), is placed at λ+μ\lambda+\mu. A point in the dominant chamber adds mν(μ)m_\nu(\mu) copies of V(λ+μ)V(\lambda+\mu). A point outside it is moved into the chamber by the dot action w⋅x=w(x+ρ)−ρw\cdot x=w(x+\rho)-\rho, and adds ε(w) mν(μ)\varepsilon(w)\,m_\nu(\mu) copies of V(w⋅(λ+μ))V\bigl(w\cdot(\lambda+\mu)\bigr), where ε(w)=det⁡w=±1\varepsilon(w)=\det w=\pm1; a point on a mirror through −ρ-\rho adds nothing. In all,

nκ=∑w∈Wε(w) mν(w⋅κ−λ).n_\kappa=\sum_{w\in W}\varepsilon(w)\,m_\nu(w\cdot\kappa-\lambda).

When λ\lambda lies so far inside the chamber that every λ+μ\lambda+\mu is dominant, nothing moves, and V(λ)⊗V(ν)=⨁μmν(μ) V(λ+μ)V(\lambda)\otimes V(\nu)=\bigoplus_\mu m_\nu(\mu)\,V(\lambda+\mu).

Verma modules

The Verma module M(λ)M(\lambda), for any weight λ∈h∗\lambda\in\mathfrak h^*, is generated by a vector of weight λ\lambda killed by the positive root vectors, with no other relations:

M(λ)=U(g)⊗U(b)Cλ.M(\lambda)=U(\mathfrak g)\otimes_{U(\mathfrak b)}\CC_\lambda.

Its weights are the λ−β\lambda-\beta with β\beta a sum of positive roots, and the multiplicity of λ−β\lambda-\beta is Kostant’s partition function P(β)P(\beta), the number of ways of writing β\beta as such a sum [5, Ch. 1]. M(λ)M(\lambda) has a unique simple quotient L(λ)L(\lambda), which is V(λ)V(\lambda) when λ\lambda is dominant integral.

The dot action w⋅λ=w(λ+ρ)−ρw\cdot\lambda=w(\lambda+\rho)-\rho governs the submodules. If n=⟨λ+ρ,α∨⟩n=\langle\lambda+\rho,\alpha^\vee\rangle is a positive integer for a positive root α\alpha, then M(λ)M(\lambda) contains M(sα⋅λ)M(s_\alpha\cdot\lambda), whose highest weight is λ−nα\lambda-n\alpha; if there is no such root, M(λ)M(\lambda) is simple [5, Ch. 4].

The character of L(λ)L(\lambda) is given by Kazhdan–Lusztig theory [5, Ch. 7–8]. The roots α\alpha with ⟨λ+ρ,α∨⟩\langle\lambda+\rho,\alpha^\vee\rangle an integer form a root system, whose Weyl group WλW_\lambda acts by the dot action. Let λ0\lambda_0 be the lowest weight in the orbit Wλ⋅λW_\lambda\cdot\lambda, and xx the shortest element of WλW_\lambda that takes λ0\lambda_0 to λ\lambda. Then

ch⁡L(λ)=∑y≤x(−1)ℓ(x)−ℓ(y) Py,x(1) ch⁡M(y⋅λ0),\operatorname{ch}L(\lambda)=\sum_{y\le x}(-1)^{\ell(x)-\ell(y)}\,P_{y,x}(1)\,\operatorname{ch}M(y\cdot\lambda_0),

where Py,xP_{y,x} are the Kazhdan–Lusztig polynomials of WλW_\lambda and ≤\le is its Bruhat order. In rank one and two, WλW_\lambda is a dihedral group or smaller, and every Py,xP_{y,x} is 11.

Affine Lie algebras

The affine Lie algebra g^\widehat{\mathfrak g} is the central extension of the loop algebra g⊗C[t,t−1]\mathfrak g\otimes\CC[t,t^{-1}] by a central element KK, with a derivation added [6, Ch. 7]. Its Dynkin diagram is that of g\mathfrak g with one more node, 00, for the simple root α0=δ−θ\alpha_0=\delta-\theta, where θ\theta is the highest root of g\mathfrak g and δ\delta the smallest positive imaginary root.

A weight Λ\Lambda is given, up to a multiple of δ\delta, by its labels λi=⟨Λ,αi∨⟩\lambda_i=\langle\Lambda,\alpha_i^\vee\rangle, one for each node: Λ=∑i=0rλiΛi\Lambda=\sum_{i=0}^r\lambda_i\Lambda_i, with Λi\Lambda_i the fundamental weights. The labels are the numbers on the Dynkin diagram next to the picture, where they can be changed.

The irreducible representation L(Λ)L(\Lambda) with highest weight Λ\Lambda is generated by a vector of weight Λ\Lambda killed by the positive root vectors. There is one for each Λ\Lambda [6, Ch. 9], and KK acts on it as the scalar

k=∑i=0rai∨λi,k=\sum_{i=0}^r a_i^\vee\lambda_i,

the level, where a0∨=1a_0^\vee=1 and θ∨=∑i≥1ai∨αi∨\theta^\vee=\sum_{i\ge1}a_i^\vee\alpha_i^\vee.

The representation L(Λ)L(\Lambda) is integrable when the action of every eie_i and fif_i is locally nilpotent, each vector killed by a power of it. Then L(Λ)L(\Lambda) is a sum of finite-dimensional representations of the copy of sl2\mathfrak{sl}_2 spanned by eie_i, fif_i and αi∨\alpha_i^\vee, for each ii. This is so exactly when Λ\Lambda is dominant integral, every λi\lambda_i a non-negative integer [6, Ch. 10], and then the level is a non-negative integer too; these are the Λ\Lambda the picture allows with the Any weight option off.

The weights of an integrable L(Λ)L(\Lambda) are the conjugates, under the affine Weyl group, of the dominant weights below Λ\Lambda, and Freudenthal’s formula gives their multiplicities, with the imaginary roots nδn\delta counted rr times [6, Ch. 11]. Each weight lies on a string μ−nδ\mu-n\delta, n≥0n\ge0, below a maximal weight μ\mu [6, Ch. 12], and the picture draws μ−nδ\mu-n\delta at nn steps below μ\mu. Every weight μ\mu has (μ∣μ)≤(Λ∣Λ)(\mu|\mu)\le(\Lambda|\Lambda), so the weights lie in a paraboloid [6, Ch. 11].

For g\mathfrak g simply laced, here sl2\mathfrak{sl}_2 and sl3\mathfrak{sl}_3, the multiplicity of μ−nδ\mu-n\delta in the basic representation L(Λ0)L(\Lambda_0), for μ\mu in WΛ0W\Lambda_0, is the number pr(n)p_r(n) of partitions of nn into parts of rr colours [1]:

∑n≥0pr(n) qn=∏m≥11(1−qm)r.\sum_{n\ge0}p_r(n)\,q^n=\prod_{m\ge1}\frac{1}{(1-q^m)^r}.

Verma modules

The Verma module M(Λ)M(\Lambda), for any weight Λ\Lambda, is generated by a vector of weight Λ\Lambda killed by the positive root vectors of g^\widehat{\mathfrak g}, with no other relations:

M(Λ)=U(g^)⊗U(b^)CΛ.M(\Lambda)=U(\widehat{\mathfrak g})\otimes_{U(\widehat{\mathfrak b})}\CC_\Lambda.

Its weights are the Λ−β\Lambda-\beta with β\beta a sum of positive roots, and the multiplicity of Λ−β\Lambda-\beta is the number of ways of writing β\beta as such a sum, each imaginary root nδn\delta counted rr times [6, Ch. 9]. M(Λ)M(\Lambda) has a unique simple quotient L(Λ)L(\Lambda), the integrable one above when Λ\Lambda is dominant integral.

With the dot action w⋅Λ=w(Λ+ρ)−ρw\cdot\Lambda=w(\Lambda+\rho)-\rho of the affine Weyl group: if n=⟨Λ+ρ,α∨⟩n=\langle\Lambda+\rho,\alpha^\vee\rangle is a positive integer for a positive real root α\alpha, then L(sα⋅Λ)L(s_\alpha\cdot\Lambda), whose highest weight is Λ−nα\Lambda-n\alpha, is a subquotient of M(Λ)M(\Lambda). When there is no such root, L(Λ)L(\Lambda) is M(Λ)M(\Lambda) itself, unless the level is critical [7]: k=−h∨k=-h^\vee, where h∨=∑i=0rai∨h^\vee=\sum_{i=0}^r a_i^\vee is the dual Coxeter number, and then (Λ+ρ∣δ)=0(\Lambda+\rho\mid\delta)=0. For sl^2\widehat{\mathfrak{sl}}_2 the critical level is −2-2. For sl^3\widehat{\mathfrak{sl}}_3 the critical level is −3-3. For so^5\widehat{\mathfrak{so}}_5 the critical level is −3-3. For g^2\widehat{\mathfrak g}_2 the critical level is −4-4.

For sl^2\widehat{\mathfrak{sl}}_2 off the critical level the integral Weyl group is dihedral, finite or infinite, its Kazhdan–Lusztig polynomials are all 11, and ch⁡L(Λ)\operatorname{ch}L(\Lambda) is an alternating sum of characters of Verma modules [8]. Otherwise L(Λ)L(\Lambda) is built from its definition, one weight space after another: a vector below Λ\Lambda is zero in L(Λ)L(\Lambda) exactly when every eje_j sends it to zero.

References

  1. I. B. Frenkel, V. G. Kac, Basic representations of affine Lie algebras and dual resonance models. Invent. Math. 62 (1980), 23–66
  2. W. Fulton, J. Harris, Representation theory: a first course. Graduate Texts in Mathematics 129, Springer (1991)
  3. G. J. Heckman, Projections of orbits and asymptotic behavior of multiplicities for compact connected Lie groups. Invent. Math. 67 (1982), 333–356
  4. J. E. Humphreys, Introduction to Lie algebras and representation theory. Graduate Texts in Mathematics 9, Springer (1972)
  5. J. E. Humphreys, Representations of semisimple Lie algebras in the BGG category O\mathcal{O}. Graduate Studies in Mathematics 94, American Mathematical Society (2008)
  6. V. G. Kac, Infinite-dimensional Lie algebras. 3rd ed., Cambridge University Press (1990)
  7. V. G. Kac, D. A. Kazhdan, Structure of representations with highest weight of infinite-dimensional Lie algebras. Adv. Math. 34 (1979), 97–108
  8. M. Kashiwara, T. Tanisaki, Kazhdan–Lusztig conjecture for affine Lie algebras with negative level. Duke Math. J. 77 (1995), 21–62; II: Nonintegral case, Duke Math. J. 84 (1996), 771–813; and, for positive rational weights, Asian J. Math. 2 (1998), 779–832