Affine Current Algebras and WZW Models

An affine current algebra extends Virasoro symmetry by holomorphic spin-one currents. Its level fixes both the central extension and, through the Sugawara construction, the stress tensor. In a Wess–Zumino–Witten model the allowed affine representations are integrable, current Ward identities become Knizhnik–Zamolodchikov equations, and fusion is truncated by the level. Every formula below uses one Lie-algebra metric so that the level shift and conformal weights can be checked without hidden factors of two.

Required background. The Virasoro algebra and the stress tensor provide the mode and OPE conventions. Lie groups, Lie algebras, the exponential map, and the adjoint action provide roots, invariant forms, and Casimirs.

Helpful background. Wess–Zumino and WZW terms explain extension independence, quantization, and global-form qualifications of the action.

Affine currents and normalization

Let $\mathfrak g$ be a compact simple Lie algebra. Normalize its invariant form so that long roots have squared length $2$, and choose Hermitian representation matrices obeying

$$[t^a,t^b]=if^{ab}{}_{c}t^c.$$

The holomorphic currents satisfy

$$J^a(z)J^b(w) \sim \frac{k\kappa^{ab}}{(z-w)^2} +\frac{if^{ab}{}_{c}J^c(w)}{z-w}.$$

With $J^a(z)=\sum_nJ_n^az^{-n-1}$ and counterclockwise contours,

$$[J_m^a,J_n^b] =if^{ab}{}_{c}J_{m+n}^c +km\kappa^{ab}\delta_{m+n,0}.$$

For radial quantization of the compact real form,

$$(J_n^a)^\dagger=J_{-n}^a,$$

and a positive-energy unitary representation has nonnegative integer $k$. Other real forms and nonunitary representations require different adjoint conditions; the same complex affine algebra does not by itself choose a Hilbert-space structure.

For a finite-dimensional highest weight $\lambda$, define the quadratic Casimir by

$$\kappa_{ab}t^at^b=2C_2(\lambda)\mathbf1.$$

This explicit factor of two is paired with the Sugawara coefficient below. In particular, $C_2(j)=j(j+1)$ for $\mathfrak{su}(2)$.

Sugawara construction

Let $h^\vee$ be the dual Coxeter number, fixed by the same invariant form. The Sugawara tensor is

$$T(z)=\frac{1}{2(k+h^\vee)} \kappa_{ab}:\!J^aJ^b\!:(z).$$

Contracting one current with the normal-ordered pair and using

$$f^{a}{}_{cd}f^{bcd}=2h^\vee\kappa^{ab}$$

in this convention gives

$$T(z)J^a(w) \sim\frac{J^a(w)}{(z-w)^2} +\frac{\partial J^a(w)}{z-w}.$$

Thus every current has weight one. A second contraction yields

$$c=\frac{k\,\dim\mathfrak g}{k+h^\vee}.$$

For an affine primary $\Phi_\lambda$,

$$J^a(z)\Phi_\lambda(w) \sim\frac{t^a_\lambda\Phi_\lambda(w)}{z-w},$$

and the zero-mode part of Sugawara gives

$$h_\lambda=\frac{C_2(\lambda)}{k+h^\vee}.$$

These equations are mutually consistent only because $\kappa_{ab}t^at^b=2C_2$. If instead $t^at^a$ is called the Casimir, the displayed factors must be changed together. The complete derivation and affine-module conditions are given in Di Francesco, Mathieu, and Sénéchal 1997, §§14.1–15.3.

WZW models and integrable representations

For a compact group $G$ and a map $g:\Sigma\to G$, the Euclidean WZW action can be written, with a compatible trace normalization, as

$$S_{\rm WZW}[g] =\frac{k}{8\pi}\int_\Sigma \operatorname{Tr}(g^{-1}\partial_\mu g\,g^{-1}\partial^\mu g) +\frac{ik}{12\pi}\int_B \operatorname{Tr}(g^{-1}dg)^3,$$

where $\partial B=\Sigma$ and orientations are fixed together. Independence of $e^{-S}$ from the choice of extension $B$ quantizes $k$ for compact simply connected $G$; quotients of $G$ can impose stronger conditions. The classical left and right $G$ symmetries become independent affine algebras. Witten’s exact analysis relates this action, current algebra, and non-Abelian bosonization in Witten 1984, §§2–4, pp. 457–472.

At nonnegative integer level, an affine highest weight is integrable when

$$\lambda\ \text{is dominant integral}, \qquad (\lambda,\theta)\le k,$$

where $\theta$ is the highest root. This finite set closes under affine fusion. For $SU(2)_k$,

$$j=0,\frac12,\ldots,\frac{k}{2}, \qquad c=\frac{3k}{k+2}, \qquad h_j=\frac{j(j+1)}{k+2}.$$

The fusion rule is

$$j_1\times j_2 =\sum_{j=|j_1-j_2|}^{\min(j_1+j_2,\,k-j_1-j_2)}j,$$

in integer steps. It is the ordinary tensor-product range truncated by integrability. At level one, the sectors are $j=0,1/2$, with $c=1$, $h_{1/2}=1/4$, and $1/2\times1/2=0$.

Knizhnik–Zamolodchikov equations

Sugawara implies the affine null relation

$$\left[ L_{-1}-\frac{1}{k+h^\vee} \kappa_{ab}J_{-1}^at^b_i \right]|\Phi_i\rangle=0.$$

Insert it in a chiral correlator and deform the current contour around the other primaries. The result is

$$(k+h^\vee)\partial_{z_i}\mathcal F =\sum_{j\ne i} \frac{\kappa_{ab}t_i^at_j^b}{z_i-z_j}\,\mathcal F.$$

The sign assumes the current-primary OPE displayed above and counterclockwise contour deformation. Reversing the sign in that OPE reverses the representation matrices in the KZ connection. Flatness of the multi-point connection follows from Lie-algebra invariance and is the integrability condition for simultaneous equations. The original derivation is Knizhnik and Zamolodchikov 1984, §§2–3, pp. 89–101.

As an exact solution, take two conjugate primaries paired to a singlet. On that invariant tensor,

$$\kappa_{ab}t_1^at_2^b=-2C_2(\lambda),$$

because $(t_1+t_2)^2=0$. The KZ equation integrates to

$$\mathcal F_{\lambda\lambda^*}(z_1,z_2) =C\,z_{12}^{-2C_2(\lambda)/(k+h^\vee)} =C\,z_{12}^{-2h_\lambda}.$$

For two $SU(2)_1$ fundamentals, $C_2(1/2)=3/4$ and $k+h^\vee=3$, so $\mathcal F\propto z_{12}^{-1/2}$, exactly the weight-$1/4$ two-point law. Four or more insertions produce a matrix differential equation on the finite-dimensional space of invariant tensors; its monodromy encodes braiding, but a full correlator still needs an antiholomorphic pairing.

Chiral data versus a full WZW CFT

An affine algebra and its integrable modules determine chiral characters, conformal blocks, and fusion. They do not uniquely determine the full theory. One must choose a left–right modular invariant, impose any global-group selection rule, and verify local sewing. Distinct global forms of a group can share the same Lie algebra while differing in allowed representations and topological sectors.

For a diagonal invariant one schematically has

$$\mathcal H_{\rm full} =\bigoplus_{\lambda\in P_k^+} \mathcal H_\lambda\otimes\overline{\mathcal H_\lambda},$$

but charge-conjugation or exceptional pairings can differ. A solution of the holomorphic KZ equation is a chiral block with possible monodromy, not a single-valued Wightman or Euclidean full correlator.

Common pitfalls

Forgetting which metric defines the level. Rescaling $\kappa$ rescales $k$, structure constants with raised indices, and Casimirs. The Sugawara denominator cannot be copied independently of the current OPE.

Using every finite-dimensional representation at level $k$. Positive-energy WZW modules must satisfy $(\lambda,\theta)\le k$. Ordinary tensor products therefore overcount affine fusion.

Identifying a KZ block with a full correlator. KZ equations constrain holomorphic blocks. Single-valuedness requires a compatible antiholomorphic pairing and sector completion.

Exercises

Compute the $SU(2)_2$ central charge, primary weights, and the fusion $1\times1$.

Solution

Here $c=3k/(k+2)=3/2$. The allowed spins are $0,1/2,1$, with

$$h_0=0, \qquad h_{1/2}=\frac{3}{16}, \qquad h_1=\frac12.$$

For $j_1=j_2=1$, the upper fusion bound is

$$\min(2,\,2-1-1)=0,$$

so $1\times1=0$, not the ordinary $0+1+2$ tensor product.

References

• Di Francesco, Philippe, Pierre Mathieu, and David Sénéchal. Conformal Field Theory. Graduate Texts in Contemporary Physics. New York: Springer, 1997. DOI.
• Knizhnik, Vadim G., and Alexander B. Zamolodchikov. “Current Algebra and Wess–Zumino Model in Two Dimensions.” Nuclear Physics B 247, no. 1 (1984): 83–103. DOI.
• Witten, Edward. “Non-Abelian Bosonization in Two Dimensions.” Communications in Mathematical Physics 92, no. 4 (1984): 455–472. DOI.