An affine Lie algebra is obtained by allowing a finite-dimensional Lie algebra to oscillate around a loop, then adding the central term required by the loop cocycle and a derivation that records Fourier mode. Its representations are organised by level, affine roots and integrability.
The construction explains why affine Kac–Moody algebras sit simultaneously in representation theory, conformal field theory, vertex operator algebras, integrable systems and geometric Langlands. The underlying finite Lie algebra remains visible in the zero mode, while the nonzero modes create an infinite tower of new operators.
This guide specialises the previous Kac–Moody Algebra article. It also connects to Vertex Operator Algebras, where positive-energy affine modules become major examples.
Loop algebra · Central extension · Level · Affine roots · Highest weights · Affine sl2 · Characters · Practice
Start with Laurent-polynomial loops
Let g be a finite-dimensional complex simple Lie algebra. Its algebraic loop algebra is
L(g)=g⊗C[t,t−1].
Write xm=x⊗tm. The bracket is defined coefficientwise:
[xm,yn]=[x,y]m+n.
Thus Fourier-mode indices add exactly as exponents of Laurent monomials add.
Why the zero mode recovers the original Lie algebra
The subspace g⊗1 consists of the mode-zero elements x₀. Since
[x₀,y₀]=[x,y]₀,
it is a copy of g inside the loop algebra.
Every affine representation therefore contains a finite-dimensional Lie algebra acting through its zero modes, though the full module can be infinite-dimensional because of the nonzero modes.
Positive and negative modes behave like two infinite halves
Modes with n>0 and n<0 play roles analogous to raising and lowering directions after a choice of affine Borel. Negative modes can create arbitrarily many descendants from a highest-weight state.
This is why a highest-weight affine module is naturally infinite-dimensional even when every zero-mode weight belongs to a familiar finite-dimensional representation of g.
The loop algebra has a canonical one-dimensional central extension
Choose an invariant symmetric bilinear form (·,·) on g. The untwisted affine central extension adds a central element K and changes the bracket to
[xm,yn]=[x,y]m+n+mδm+n,0(x,y)K.
The extra term can appear only when m+n=0. Modes of opposite frequency can therefore produce a central scalar in addition to the ordinary Lie bracket.
K commutes with every element. It acts by a scalar on every irreducible module, and that scalar is the level.
The cocycle is antisymmetric
Define ψ(xm,yn)=mδm+n,0(x,y). If m+n=0, then n=−m. Therefore
ψ(yn,xm)=n(x,y)=−m(x,y)=−ψ(xm,yn).
The invariance of the bilinear form supplies the Lie-algebra 2-cocycle identity needed for the Jacobi relation.
This central extension is not an arbitrary decoration. It is the nontrivial extension that turns loop symmetry into affine Kac–Moody symmetry.
Add the degree derivation
Introduce d with
[d,xn]=n xn, [d,K]=0.
The operator d measures Fourier degree. It separates modes with the same finite-dimensional g-weight but different loop frequency.
The full untwisted affine algebra is therefore
ĝ=L(g)⊕CK⊕Cd.
Some authors call the central extension without d the derived affine algebra and reserve the full notation for the version including d. State the convention before comparing formulas.
Level is the central charge of the affine Lie algebra action
If M is an irreducible ĝ-module, Schur’s lemma gives
K·v=k v
for one scalar k. This k is the level of the representation.
Representations at different levels cannot be isomorphic as affine modules because the central element acts differently.
The level is not the eigenvalue of d and not the conformal weight of a state. These are separate quantities.
Level zero versus nonzero level
Ordinary evaluation representations of the loop algebra give important level-zero examples. Fix a∈C× and a finite-dimensional g-module V. Define
(x⊗tn)v=anxv.
This is a representation of L(g). It extends across the central element by letting K act as zero. A fixed evaluation point is generally not preserved by the degree derivation d, so the extension to the full algebra including d requires extra structure.
Positive-level integrable highest-weight modules are very different: the central term is essential and the modules are necessarily infinite-dimensional.
Worked affine sl2 commutators
Let e,f,h satisfy [e,f]=h, [h,e]=2e and [h,f]=−2f, and normalise the invariant form by (e,f)=1 and (h,h)=2.
Then the affine modes satisfy
- [em,fn]=hm+n+mδm+n,0K;
- [hm,hn]=2mδm+n,0K;
- [hm,en]=2em+n;
- [hm,fn]=−2fm+n.
For m=1,n=−1,
[e₁,f−1]=h₀+K.
On a level-k module this acts as h₀+kI. The level has entered an ordinary operator calculation directly.
Affine root systems repeat along the null root δ
For an untwisted affine algebra associated with finite root system Δfin, the real affine roots have the form
α+nδ, α∈Δfin, n∈Z.
The imaginary roots are nonzero multiples nδ.
The null root δ vanishes on the finite Cartan directions and on K, while recording degree against d in the standard realization.
This creates an infinite vertical repetition of the finite root system. Every finite root appears at infinitely many loop modes.
Imaginary-root multiplicity in the untwisted case
For a simple finite algebra g of rank r, the root space for nδ, n≠0, is essentially hfin⊗tn. Hence
mult(nδ)=r
in the untwisted affine case.
This contrasts with real roots, whose multiplicity is one. The distinction becomes visible in the affine denominator formula and character theory.
The affine Weyl group
The affine Weyl group extends the finite Weyl group by root-lattice translations. Geometrically it acts by affine reflections across parallel families of hyperplanes.
For rank-one finite g=sl₂, the affine Weyl group is infinite dihedral. Two simple reflections generate an infinite sequence rather than closing after finitely many elements.
This infinite reflection symmetry is why affine characters naturally involve infinite Weyl sums.
Affine highest-weight modules
Choose the standard affine Borel. A highest-weight vector vΛ is annihilated by all positive simple-root generators and is an eigenvector for the affine Cartan subalgebra.
The highest weight Λ records:
- its finite simple-coroot coordinates;
- the affine simple-coroot coordinate;
- the level Λ(K);
- optionally the d-eigenvalue, depending on normalization.
The corresponding Verma module is universal, while its simple quotient L(Λ) may or may not be integrable.
Integrable highest weights
For an affine Kac–Moody algebra, a highest-weight simple module is integrable when every Chevalley ei,fi acts locally nilpotently.
The highest weight must be dominant integral:
Λ(hi)∈Z≥0
for every affine simple coroot.
In the standard untwisted normalization, integrable highest-weight modules have nonnegative integer level.
At a fixed nonnegative level there are only finitely many dominant integral highest weights, even though each corresponding module is infinite-dimensional.
Worked affine sl2 level classification
For A₁(1), let Λ₀ and Λ₁ be the affine fundamental weights. A dominant integral highest weight has the form
Λ=m₀Λ₀+m₁Λ₁, m₀,m₁∈Z≥0.
The central element is K=h₀+h₁ in the standard realization, so the level is
k=Λ(K)=m₀+m₁.
At level k the possibilities are
kΛ₀, (k−1)Λ₀+Λ₁, …, Λ₀+(k−1)Λ₁, kΛ₁.
There are exactly k+1 integrable highest weights at level k.
Level-one affine sl2
At k=1 the only integrable highest weights are Λ₀ and Λ₁. These give two irreducible integrable highest-weight modules.
At k=2 the weights are 2Λ₀, Λ₀+Λ₁ and 2Λ₁. The finite number of sectors at fixed positive integer level is one reason affine Lie algebras feed naturally into rational conformal models.
The modules themselves are nevertheless infinite-dimensional because negative loop modes generate infinitely many descendants.
Finite zero-mode content is not the full affine module
The zero modes form a copy of sl₂. Inside an affine module one can decompose states into finite-dimensional sl₂ weight spaces at each grade.
But applying modes such as e−1,h−2 or f−3 changes the affine degree. These states do not belong to one finite-dimensional sl₂ multiplet alone.
Thus “the finite Lie algebra representation at grade zero” is only the first layer of the affine representation.
The energy grading
The derivation d obeys [d,xn]=n xn. If v has d-eigenvalue r, then x−mv has eigenvalue r−m in this algebraic convention.
In conformal-field conventions one often defines an energy operator with the opposite sign or shifts by the lowest conformal weight, so descendants created by negative current modes have positive energy grade.
Do not identify d and L₀ without translating conventions. Sugawara L₀ is built quadratically from currents and differs by normalization and lowest-weight shifts.
The Sugawara bridge
At noncritical level, affine currents can be combined quadratically to construct Virasoro operators through the Sugawara construction.
For simple g with dual Coxeter number h∨, a standard normalization gives central charge
c=k dim(g)/(k+h∨)
when k≠−h∨.
For sl₂, dim g=3 and h∨=2, so at level k=1 the Sugawara central charge is c=1.
This provides the direct bridge from affine Lie representations to the VOA conformal vector discussed in Batch 07.
Critical level is structurally different
The level
k=−h∨
is called the critical level in the standard normalization. The Sugawara denominator vanishes there, and the completed enveloping algebra develops a large center.
Feigin and Frenkel identify that center with functions on opers for the Langlands dual group. This is a major entry point into the local geometric Langlands story.
The critical-level category is not obtained by blindly substituting k=−h∨ into positive-level rational formulas. It requires a different representation-theoretic analysis.
Affine characters and the Weyl–Kac formula
Integrable highest-weight affine modules satisfy the Weyl–Kac character formula from the preceding article. Because the affine Weyl group and positive-root system are infinite, the formal character contains infinitely many terms.
After suitable specialisation, affine characters often become q-series with strong modular properties. This is one source of the modular transformation matrices later used in affine VOAs and conformal field theory.
For affine sl₂ at fixed positive integer level, the integrable characters form a finite family even though each individual character is an infinite q-series.
Fusion truncates ordinary tensor-product intuition
At positive integral level, the corresponding rational affine VOA has a finite set of integrable simple modules. Their fusion product is not simply the unrestricted tensor product of finite-dimensional g representations.
For affine sl₂ at level k, only highest weights 0,1,…,k are allowed. Fusion follows a level-truncated version of the Clebsch–Gordan pattern.
This is where affine highest-weight theory meets the Fusion Categories article and the VOA article.
Twisted affine algebras
If g has a finite-order diagram automorphism, one can form twisted loops whose Fourier modes satisfy an equivariance condition. Their central extensions produce twisted affine Kac–Moody algebras.
The root systems and mode gradings then involve fractional-looking shifts before the algebra is reorganised into its standard integer root lattice.
This guide focuses on untwisted affine algebras. Results stated using α+nδ and multiplicity rank(g) for nδ should not be copied unchanged into every twisted type.
A representation workflow
- 1. Fix the invariant form. Central-term coefficients depend on normalization.
- 2. Write xn=x⊗tn.
- 3. Verify the loop bracket before adding K.
- 4. Add the cocycle mδm+n,0(x,y)K.
- 5. Record whether d is included.
- 6. Read the level from K.
- 7. For a highest weight, compute every affine simple-coroot coordinate.
- 8. Check integrability through nonnegative integers.
- 9. At fixed level, enumerate the dominant weights before calculating characters.
- 10. Separate generic, positive-integral and critical levels.
Common mistakes
- Dropping the central extension: the loop algebra and affine Kac–Moody algebra are not the same object.
- Calling K the derivation: K is central; d measures mode.
- Confusing level with conformal weight: level is the K-eigenvalue.
- Forgetting normalization: central cocycle coefficients scale with the invariant form.
- Assuming evaluation modules have nonzero level: ordinary evaluation representations are level zero.
- Assuming integrable means finite-dimensional: positive-level affine highest-weight modules are infinite-dimensional.
- Using positive-level Sugawara formulas at critical level: the critical denominator vanishes.
- Copying untwisted root formulas into twisted types without checking.
Practice questions
1. Compute [x₂,y−1] in the loop algebra. 2. When can the central term in [xm,yn] be nonzero? 3. What is [d,x₅]? 4. If K acts by 3, what is the level?
5. In affine sl₂, compute [e₂,f−2] on a level-k module. 6. What are the real untwisted affine roots? 7. What are the imaginary roots? 8. How many integrable affine sl₂ highest weights are there at level 4?
9. List those level-4 weights. 10. What is the Sugawara central charge for sl₂ at k=2? 11. Why is k=−h∨ exceptional? 12. Why does a finite list of integrable highest weights not imply finite-dimensional modules?
Worked answers
1. [x₂,y−1]=[x,y]₁. The mode sum is 1, so no central term appears.
2. Only when m+n=0.
3. [d,x₅]=5x₅.
4. Level 3.
5. [e₂,f−2]=h₀+2K, so on level k it acts as h₀+2kI.
6. α+nδ with α a finite root and n∈Z.
7. Nonzero multiples nδ.
8. Five, because k+1=5.
9. 4Λ₀, 3Λ₀+Λ₁, 2Λ₀+2Λ₁, Λ₀+3Λ₁ and 4Λ₁.
10. c=2·3/(2+2)=3/2.
11. The Sugawara denominator k+h∨ vanishes, and the completed enveloping algebra develops a large centre.
12. Each allowed highest weight generates infinitely many descendants through negative loop modes.
Sources and further study
[1] Kevin Wray, Affine Lie Algebras, for loop algebras, cocycles, central extensions and the derivation. [2] Edward Frenkel, Langlands Correspondence for Loop Groups, for affine Kac–Moody representation theory and the critical-level Langlands architecture. [3] Edward Frenkel’s vertex-algebra chapter develops affine modules and their connection with vertex algebras. [4] Victor Kac, Infinite-Dimensional Lie Algebras, remains the standard reference for affine root systems, integrable highest weights and Weyl–Kac characters.
Representation Mathematics — Batch 08
Begin with the general matrix construction in Kac–Moody Algebras. Reduce quantum highest-weight modules to combinatorial graphs in Crystal Bases. Compute finite-type tensor multiplicities in Littlewood–Richardson Coefficients, Honeycombs and Saturation. Return to the BTT Mathematics Learning Hub.