The Wess-Zumino-Witten action and the level-k extension

Intuition Beginner

Most field theories assign a number to each point of space and time. The Wess-Zumino-Witten model assigns instead a group element: at every point of a two-dimensional surface sits a rotation, or a unitary matrix, or some other symmetry. The field is a map from the surface into a group.

The energy of such a field has a familiar piece that measures how fast the group element changes from point to point. Smooth maps cost little; rapidly varying ones cost a lot. On its own this piece would give an ordinary, somewhat dull theory.

What makes the model special is a second piece, a topological term. It does not measure stretching. It measures a kind of winding of the field through the group, the way a path can wrap around a hole.

Visual Beginner

Picture the surface as a sheet and the group as a curved room with handles. The field paints each point of the sheet with a location in that room. The ordinary energy term is unhappy when neighbouring points land far apart. The topological term instead counts how the painted image wraps around the handles of the room.

Because winding is a whole-number affair, the strength of this term cannot be tuned freely. It must come in integer steps. That integer is called the level.

Worked example Beginner

Take the group of unit-length complex numbers, the circle. A field is a phase at each point of the sheet. Walk around a small loop on the sheet and watch the phase: it may return to its start having gone around the circle once, twice, or not at all.

The whole-number count of those windings is exactly the kind of data the topological term records. You cannot have half a winding. If you tried to weight this term by a fraction, two ways of measuring the same field would disagree, and the theory would give two different answers for one question.

Insisting the answer be single-valued forces the weight to be a whole number. For richer groups the same logic returns, and the whole number is the level.

Check your understanding Beginner

Formal definition Intermediate+

Let $G$ be a compact, simple, simply connected Lie group with Lie algebra $\mathfrak{g}$ and dual Coxeter number $h^{\vee }$, and let $\Sigma$ be a closed oriented surface. A field is a smooth map $g:\Sigma \to G$. Write $\theta =g^{-1}dg$ for the pulled-back left Cartan-Maurer form, a $\mathfrak{g}$-valued one-form satisfying $d\theta +\theta \wedge \theta =0$.

The kinetic term is the sigma-model energy

$$S_{\mathrm{kin}}[g]=\frac{k}{8\pi }{\int }_{\Sigma }\mathrm{tr}(\theta \wedge \star \theta ),$$

with $\star$ the Hodge star and $\mathrm{tr}$ the Killing form normalised so the highest coroot has length squared two.

The defining feature is the Wess-Zumino term. Choose a three-manifold $B$ with $\partial B=\Sigma$ and an extension $\tilde{g}:B\to G$ of $g$, and set

$${\Gamma }_{\mathrm{WZ}}[\tilde{g}]=\frac{1}{12\pi }{\int }_B\mathrm{tr}({\tilde{g}}^{-1}d\tilde{g}{)}^{\wedge 3}=\frac{1}{12\pi }{\int }_B\mathrm{tr}{\theta }^3.$$

The full WZW action at level $k\in \mathbb{Z}$ is $S_k[g]=S_{\mathrm{kin}}[g]+k{\Gamma }_{\mathrm{WZ}}[\tilde{g}]$ ^{[Witten 1984]}.

The integrand $\omega =\frac{1}{12\pi }\mathrm{tr}{\theta }^3$ is a closed, bi-invariant three-form on $G$. Its cohomology class is the canonical generator of $H^3(G,\mathbb{Z})\cong \mathbb{Z}$; this integrality of $\omega$ is precisely the statement that it is the curvature of the level-one WZW gerbe ^{[Brylinski Ch. 7]}.

Key theorem with proof Intermediate+

Theorem (well-definedness forces the level to be an integer). The exponentiated amplitude $\exp (ik{\Gamma }_{\mathrm{WZ}}[\tilde{g}])$ is independent of the chosen filling $(B,\tilde{g})$ if and only if $k\in \mathbb{Z}$.

Proof. Let $(B_1,{\tilde{g}}_1)$ and $(B_2,{\tilde{g}}_2)$ be two fillings of the same boundary data $g:\Sigma \to G$. Glue $B_1$ to the orientation-reversal of $B_2$ along $\Sigma$ to form a closed oriented three-manifold $X=B_1{\cup }_{\Sigma }\overline{B_2}$, carrying a map $G:X\to G$ that restricts to ${\tilde{g}}_1$ and ${\tilde{g}}_2$ on the two pieces. The difference of the two Wess-Zumino integrals is then a single integral over the closed manifold,

$${\Gamma }_{\mathrm{WZ}}[{\tilde{g}}_1]-{\Gamma }_{\mathrm{WZ}}[{\tilde{g}}_2]=\frac{1}{12\pi }{\int }_X\mathrm{tr}(G^{-1}dG{)}^3={\int }_X{G}^{\ast }\omega .$$

Now $\omega$ represents the integral generator of $H^3(G,\mathbb{Z})$, so its pairing with the fundamental class of any closed oriented three-manifold is a whole number: ${\int }_X{G}^{\ast }\omega \in \mathbb{Z}$. The two fillings therefore differ by an integer, ${\Gamma }_{\mathrm{WZ}}[{\tilde{g}}_1]-{\Gamma }_{\mathrm{WZ}}[{\tilde{g}}_2]\in \mathbb{Z}$. Multiplying by $k$ and exponentiating, $\exp (2\pi ikm)$ equals one for every such integer $m$ exactly when $k$ itself is an integer. Conversely, taking $X=S^3$ with a degree-one map realises the integer $1$, so non-integer $k$ produces an honest ambiguity. $\square$

Bridge. This integer-level constraint builds toward the entire algebraic skeleton of the rational conformal field theory the model defines, and the same quantization condition appears again in 03.06.07 as the integrality of the Chern-Simons level. The foundational reason is that the Wess-Zumino three-form is the de Rham image of an integral cohomology class, so this is exactly the Dirac quantization argument in cohomological dress: a wavefunction on a space with non-exact flux is single-valued precisely when the flux period is integral, and here the relevant flux lives in 03.06.09 as the curvature of the WZW gerbe. The bridge is that the boundary ambiguity of ${\Gamma }_{\mathrm{WZ}}$ generalises the multivaluedness of the magnetic potential, and putting these together identifies the level $k$ with the Dixmier-Douady class of that gerbe, an integer by construction.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none — Mathlib can express Lie groups and differential forms separately, but not the integral three-form generator of $H^3(G,\mathbb{Z})$ or the bounding-manifold construction of the Wess-Zumino term.

-- Pseudocode only: no WZW structure is available in Mathlib.
axiom WZForm (G : Type) [LieGroup G] : DifferentialForm G 3   -- (1/12π) tr θ³
axiom WZForm_integral
    (G : Type) [LieGroup G] (X : ClosedOriented3Manifold) (φ : X → G) :
    integral X (pullback φ (WZForm G)) ∈ (Int : Set Real)

A faithful formalization would build the bi-invariant Cartan three-form, prove it generates the integral third cohomology, and then realise the level as the Dixmier-Douady class of the associated gerbe. Each step needs infrastructure not yet present.

Advanced results Master

The factorised equations of motion expose the symmetry. The combined kinetic-plus-topological action satisfies $\partial (\bar{\partial }g{g}^{-1})=0$ and $\bar{\partial }(\partial g{g}^{-1})=0$, so left- and right-moving currents $J=k\partial g{g}^{-1}$ and $\bar{J}=-k{g}^{-1}\bar{\partial }g$ are separately conserved. The Polyakov-Wiegmann identity, expressing ${\Gamma }_{\mathrm{WZ}}[g_1{g}_2]$ in terms of the individual terms plus a bilinear cross piece, is the geometric origin of these conservation laws and of the chiral splitting ^{[Witten 1984]}.

Expanding the holomorphic current in modes, $J^a(z)={\sum }_n{J}_n^a{z}^{-n-1}$, gives the affine Kac-Moody algebra at level $k$,

$$[J_m^a,J_n^b]=i{f}^{ab}{}_c{J}_{m+n}^c+km{\delta }^{ab}{\delta }_{m+n,0},$$

which is exactly the level-$k$ central extension $\widetilde{LG}$ of the loop group acting on the state space 03.11.05. The structure constants $f^{ab}{}_c$ are those of $\mathfrak{g}$; the central term is governed by the same level fixed in the action.

The Sugawara construction builds the stress tensor from a normal-ordered bilinear in the currents,

$$T(z)=\frac{1}{2(k+h^{\vee })}\sum _a:J^a{J}^a:(z),$$

and a direct computation of $T(z)T(w)$ recovers the Virasoro algebra of 03.10.02 with central charge

$$c=\frac{k\dim G}{k+h^{\vee }}.$$

The current correlators obey the Knizhnik-Zamolodchikov equations, first-order holomorphic differential equations whose solutions are the conformal blocks ^{[Knizhnik-Zamolodchikov 1984]}. With finitely many integrable highest-weight modules at each level, the WZW model is the prototype of a rational conformal field theory.

Synthesis. The level $k$ is a single integer that organises the whole theory, and tracing its role is the central insight of this unit. This is exactly the same number in three guises: the curvature period of the WZW gerbe 03.06.09, the central term of the affine algebra 03.11.05, and the Chern-Simons coupling 03.06.07 whose boundary theory the WZW model is. The bridge is transgression: the three-dimensional Chern-Simons functional restricts to the two-dimensional Wess-Zumino term on the boundary, so the WZW action generalises a boundary value of a topological bulk theory, and the affine symmetry is dual to the bulk gauge invariance. Putting these together, the foundational reason the central charge takes the Sugawara value $c=k\dim G/(k+h^{\vee })$ is that the same level controls both the normalisation of the current two-point function and the shift by the dual Coxeter number coming from the currents' own back-reaction, and this is exactly why rational conformal field theories are classified by representation-theoretic rather than Lagrangian data.

Full proof set Master

Proposition (the Wess-Zumino three-form is closed and bi-invariant). On a Lie group $G$, the form $\omega =\frac{1}{12\pi }\mathrm{tr}{\theta }^3$, with $\theta =g^{-1}dg$, is closed and invariant under both left and right translations.

Proof. Left invariance is immediate: $\theta$ is the left Cartan-Maurer form, so $L_a^{\ast }\theta =\theta$ for every $a\in G$, hence $L_a^{\ast }\omega =\omega$. For right translation $R_a$ one has $R_a^{\ast }\theta ={\mathrm{Ad}}_{a^{-1}}\theta$, and the trace is conjugation-invariant, so $R_a^{\ast }\omega =\frac{1}{12\pi }\mathrm{tr}({\mathrm{Ad}}_{a^{-1}}\theta {)}^3=\omega$. Closedness uses the Maurer-Cartan equation $d\theta =-\theta \wedge \theta$. Differentiating,

$$d\mathrm{tr}{\theta }^3=3\mathrm{tr}(d\theta \wedge \theta \wedge \theta )=-3\mathrm{tr}(\theta \wedge \theta \wedge \theta \wedge \theta ).$$

Cyclicity of the trace together with the odd degree of each $\theta$ gives $\mathrm{tr}({\theta }^4)=-\mathrm{tr}({\theta }^4)$, so $\mathrm{tr}({\theta }^4)=0$ and $\omega$ is closed. $\square$

Proposition (chiral conservation of the WZW current). For a critical point of $S_k=S_{\mathrm{kin}}+k{\Gamma }_{\mathrm{WZ}}$, the current $J=k\partial g{g}^{-1}$ is holomorphically conserved, $\bar{\partial }J=0$.

Proof. Vary $g\to g{e}^{ϵ}$ with $ϵ$ a small $\mathfrak{g}$-valued function. The kinetic term contributes a total-derivative variation proportional to $\partial (g^{-1}\bar{\partial }g)+\bar{\partial }(g^{-1}\partial g)$, while the Wess-Zumino term, by the Polyakov-Wiegmann identity, contributes the antisymmetric combination $\partial (g^{-1}\bar{\partial }g)-\bar{\partial }(g^{-1}\partial g)$. At the conformal point the relative normalisation is tuned so the two combine into a single chiral equation $\bar{\partial }(\partial g{g}^{-1})=0$. Multiplying by the level gives $\bar{\partial }J=0$, so $J$ is a function of the holomorphic coordinate alone. $\square$

Connections Master

• CFT basics 03.10.02 — the WZW model is the worked, non-abelian realisation of the conformal-symmetry framework introduced there; its Sugawara stress tensor reproduces the Virasoro algebra, and the central charge $c=k\dim G/(k+h^{\vee })$ specialises the abstract central charge to an explicit, level-dependent value.

• WZW gerbe and integral 3-curvature 03.06.09 — the integrality of $\frac{k}{12\pi }\mathrm{tr}{\theta }^3$ that quantizes the level is exactly the statement that this three-form is the curvature of a gerbe with integral Dixmier-Douady class, the geometric object resolving the filling ambiguity of ${\Gamma }_{\mathrm{WZ}}$.

• Level-k loop-group extension 03.11.05 — the affine ${\hat{\mathfrak{g}}}_k$ symmetry generated by the chiral current is the Lie-algebra shadow of the level-$k$ central extension $\widetilde{LG}$, so the same integer governs the action, the cohomology class, and the projective representation on the state space.

• Chern-Simons transgression 03.06.07 — the Wess-Zumino term is the boundary transgression of the three-dimensional Chern-Simons functional, making the WZW model the chiral edge theory of a topological bulk and tying the level of one to the level of the other.

Historical & philosophical context Master

The topological term entered physics through Wess and Zumino's 1971 study of anomalous Ward identities, where it encoded the consequences of chiral anomalies for pion interactions ^{[Wess-Zumino 1971]}. At that stage the term was a local functional with a puzzling normalisation. Witten's 1984 paper recast it geometrically: by writing the term as an integral over a bounding three-manifold he made its quantization condition visible and identified the resulting theory as an exactly solvable conformal field theory equivalent, for suitable groups, to free fermions ^{[Witten 1984]}. In the same year Knizhnik and Zamolodchikov derived the differential equations governing its correlators, completing the model's solution ^{[Knizhnik-Zamolodchikov 1984]}.

Philosophically the WZW model is a clean instance of how topology constrains physics. A coupling constant that one might expect to vary continuously is instead locked to the integers by the demand that a quantum amplitude be single-valued. The same arithmetic discreteness recurs across the subject, from magnetic charge to the Chern-Simons level, and the geometric language of gerbes makes plain that these are one phenomenon seen in different settings ^{[Brylinski Ch. 7]}.

Bibliography Master

@article{Witten1984,
  author  = {Witten, Edward},
  title   = {Non-abelian bosonization in two dimensions},
  journal = {Communications in Mathematical Physics},
  volume  = {92},
  pages   = {455--472},
  year    = {1984}
}

@article{WessZumino1971,
  author  = {Wess, J. and Zumino, B.},
  title   = {Consequences of anomalous Ward identities},
  journal = {Physics Letters B},
  volume  = {37},
  pages   = {95--97},
  year    = {1971}
}

@article{KnizhnikZamolodchikov1984,
  author  = {Knizhnik, V. G. and Zamolodchikov, A. B.},
  title   = {Current algebra and Wess-Zumino model in two dimensions},
  journal = {Nuclear Physics B},
  volume  = {247},
  pages   = {83--103},
  year    = {1984}
}

@book{Brylinski1993,
  author    = {Brylinski, Jean-Luc},
  title     = {Loop Spaces, Characteristic Classes and Geometric Quantization},
  publisher = {Birkh\"auser},
  year      = {1993}
}

@book{PressleySegal1986,
  author    = {Pressley, Andrew and Segal, Graeme},
  title     = {Loop Groups},
  publisher = {Oxford University Press},
  year      = {1986}
}

---

Produced autonomously; connects the conformal-field-theory side of the curriculum to the gerbe and loop-group constructions. CFT review pending.