Casimir connections, braid monodromy and chiral WZW interfaces

Let $\mathfrak{g}$ be a finite-dimensional complex semisimple Lie algebra, let $B$ be its Killing form, and let $(x_a)$ and $(x^a)$ be $B$-dual bases. The tensor $\mathrm{\Omega }:={\sum }_a{x}_a\otimes x^a\in \mathfrak{g}\otimes \mathfrak{g}$ corresponds to the identity map under ${\mathrm{\Phi }}_B(u\otimes v)(y):=uB(v,y)$, and is invariant under the diagonal adjoint action:

$([x,\cdot ]\otimes 1+1\otimes [x,\cdot ])\mathrm{\Omega }={\sum }_a[x,x_a]\otimes x^a+x_a\otimes [x,x^a]=0(x\in \mathfrak{g})\mathrm{.}$

Let $\mathfrak{g}$ be a finite-dimensional complex simple Lie algebra with Killing form $B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}$. Choose a Cartan subalgebra $\mathfrak{h}$ and a positive root system ${\mathrm{\Phi }}^{+}$ with highest root $\theta$, and put $\rho :=\frac{1}{2}{\sum }_{\alpha \in {\mathrm{\Phi }}^{+}}\alpha$. For each $\lambda \in {\mathfrak{h}}^{\ast }$ let $H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}\in \mathfrak{h}$ be the unique vector satisfying $B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}(H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}},H)=\lambda (H)$ for every $H\in \mathfrak{h}$, and define the induced form on ${\mathfrak{h}}^{\ast }$ by $(\lambda ,\mu {)}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}:=B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}(H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}},H_{\mu ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}})$. This extends the supplied convention $B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}(H_{\alpha ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}},H)=\alpha (H)$ for roots $\alpha$. Set $c:=2\mathrm{/}(\theta ,\theta {)}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}$ and $B_0:=B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}\mathrm{/}c$. For each $\lambda \in {\mathfrak{h}}^{\ast }$, let $H_{\lambda ,0}\in \mathfrak{h}$ satisfy $B_0(H_{\lambda ,0},H)=\lambda (H)$ for every $H\in \mathfrak{h}$, and put $(\lambda ,\mu {)}_0:=B_0(H_{\lambda ,0},H_{\mu ,0})$. Then $(\theta ,\theta {)}_0=2$. Define the coroot ${\theta }^{\vee }:=2H_{\theta ,0}\mathrm{/}(\theta ,\theta {)}_0=H_{\theta ,0}$ and the dual Coxeter number $h^{\vee }:=1+\rho ({\theta }^{\vee })$. Then

$B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}=2h^{\vee }{B}_0\mathrm{.}$

Let $\mathfrak{g}$ be a finite-dimensional complex semisimple Lie algebra with Killing form $B$, let $n\ge 2$, and let $V_1,\dots ,V_n$ be finite-dimensional complex vector spaces. A representation here is a complex-linear Lie algebra homomorphism ${\rho }_i:\mathfrak{g}\to {\mathrm{End}}_{\mathbb{C}}(V_i)$, so $[{\rho }_i(x),{\rho }_i(y)]={\rho }_i([x,y])$. Let $F_n(\mathbb{C})=\{(z_1,\dots ,z_n)\in {\mathbb{C}}^n:z_i\mathrm{\ne }{z}_j\text{ for }i\mathrm{\ne }j\}$ be the ordered configuration space and put $V:=V_1\otimes \cdots \otimes V_n$ in the indicated factor order. For any pair of $B$-dual bases $(x_a),(x^a)$ define ${\mathrm{\Omega }}_{ij}:={\sum }_a{\rho }_i(x_a{)}^{(i)}{\rho }_j(x^a{)}^{(j)}\in {\mathrm{End}}_{\mathbb{C}}(V)$, where each superscript means that the operator acts on that tensor factor and as the identity on all other factors. For $i<j$ put ${\eta }_{ij}:=(d{z}_i-d{z}_j)\mathrm{/}(z_i-z_j)$, a globally defined complex-valued one-form on $F_n(\mathbb{C})$. For $\kappa \in \mathbb{C}\setminus \{0\}$, the Casimir Knizhnik–Zamolodchikov connection on the trivial vector bundle $F_n(\mathbb{C})\times V\to F_n(\mathbb{C})$ is

${\mathrm{\nabla }}_{\kappa }^{\mathrm{K}\mathrm{Z}}:=d-{\kappa }^{-1}{\sum }_{i<j}{\mathrm{\Omega }}_{ij}{\eta }_{ij}\mathrm{.}$

Here a connection on this trivial bundle is a complex-linear operator on smooth $V$-valued sections satisfying $\mathrm{\nabla }(fs)=df\otimes s+f\mathrm{\nabla }s$ for every smooth scalar function $f$ and section $s$. The associated quadratic Casimir is $C_B:={\sum }_a{x}_a{x}^a$ in the universal enveloping algebra; with the Killing-form convention above it is central, and ${\mathrm{\Omega }}_{ij}$ is its tensor-level Casimir counterpart.

Ordered configuration spaces $F_n(X)$

The Killing form of a semisimple Lie algebra

The quadratic Casimir element is independent of the choice of dual bases

The quadratic Casimir element

The quadratic Casimir element is central

Let $n\ge 2$, let $\mathfrak{g}$ be a finite-dimensional complex semisimple Lie algebra with Killing form $B$, let $V_i$ be finite-dimensional complex $\mathfrak{g}$-modules, and let $\kappa \in \mathbb{C}\setminus \{0\}$. Fix a basepoint $q\in F_n(\mathbb{C})$. Let $D^2:=\{z\in \mathbb{C}:\mathrm{\mid }z\mathrm{\mid }\le 1\}$, $D^{\circ }:=\{z\in \mathbb{C}:\mathrm{\mid }z\mathrm{\mid }<1\}$, $h(z):=z\mathrm{/}(1+\mathrm{\mid }z\mathrm{\mid })$, and $\tilde{q}:=(h(q_1),\dots ,h(q_n))\in F_n(D^{\circ })$. On $F_n(\mathbb{C})=\{(z_1,\dots ,z_n):z_i\mathrm{\ne }{z}_j\text{ for }i\mathrm{\ne }j\}$ let ${\mathrm{\nabla }}_{\kappa }^{\mathrm{K}\mathrm{Z}}=d-{\kappa }^{-1}{\sum }_{i<j}{\mathrm{\Omega }}_{ij}{\eta }_{ij}$ be the connection defined from the pairwise Casimir tensors ${\mathrm{\Omega }}_{ij}={\sum }_a{\rho }_i(x_a{)}^{(i)}{\rho }_j(x^a{)}^{(j)}$ and forms ${\eta }_{ij}=(d{z}_i-d{z}_j)\mathrm{/}(z_i-z_j)$, using $B$-dual bases. Then ${\mathrm{\nabla }}_{\kappa }^{\mathrm{K}\mathrm{Z}}$ is flat. Its sheaf of local horizontal sections is a local system with right monodromy action $T$ of the fundamental group of the ordered configuration space based at $q$. With the library convention that $[\alpha ][\beta ]$ traverses $\alpha$ first and $\beta$ second, the associated left representation is $[\gamma ]\mapsto T_{[\gamma ]}^{-1}$; on piecewise smooth loops, $T_{[\gamma ]}$ is horizontal parallel transport $P_{\gamma }$. The coordinatewise map $F_n(\mathbb{C})\to F_n(D^{\circ })$, $(z_1,\dots ,z_n)\mapsto (h(z_1),\dots ,h(z_n))$, sends $q$ to $\tilde{q}$. Composed with the inclusion-induced isomorphism at the same interior basepoint, it identifies ${\pi }_1(F_n(\mathbb{C}),q)$ with $P{B}_n={\pi }_1(F_n(D^2),\tilde{q})$. If all $V_i$ are the same module with the same action, simultaneous label and tensor-factor permutation makes the connection equivariant: for $\sigma \in S_n$, set $\sigma \cdot (z_1,\dots ,z_n)=(z_{{\sigma }^{-1}(1)},\dots ,z_{{\sigma }^{-1}(n)})$ and ${\tau }_{\sigma }(v_1\otimes \cdots \otimes v_n)=v_{{\sigma }^{-1}(1)}\otimes \cdots \otimes v_{{\sigma }^{-1}(n)}$, and use $\sigma \cdot (z,v)=(\sigma \cdot z,{\tau }_{\sigma }v)$. The quotient bundle is $(F_n(\mathbb{C})\times V^{\otimes n})\mathrm{/}{S}_n$ under this action, and the connection descends to the unordered configuration space, giving the corresponding full configuration braid-group representation based at $[q]\in C_n(\mathbb{C})$. The coordinatewise map on unordered configurations takes $[q]$ to $[\tilde{q}]$; composing its induced fundamental-group isomorphism with the inclusion-induced isomorphism at that orbit identifies the descended representation with $B_n^{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{f}}={\pi }_1(C_n(D^2),[\tilde{q}])$. No unitarity or equivalence with a quantum-group ribbon category is asserted.

Casimir KZ connection on ordered configurations

The Casimir tensor is invariant under the diagonal adjoint action

A flat connection admits local parallel frames

Parallel transport along a piecewise smooth curve

Based loops and the fundamental group

Ordered configuration spaces $F_n(X)$

The pure braid group $P{B}_n$ as the fundamental group of an ordered configuration space

Unordered configuration spaces $C_n(X)$

Ordered configuration spaces cover the unordered ones regularly with deck group $S_n$

The configuration braid group $B_n^{\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{f}}$ as the fundamental group of an unordered configuration space

A chiral Wess–Zumino–Witten (WZW) / vertex-operator-algebra (VOA) completion record concerns one holomorphic current-algebra sector, not a full left-right theory; it is additional mathematical data and theorems that make a comparison with the finite-dimensional Casimir KZ connection precise. This definition is a contract for such a comparison; it asserts no existence or equivalence theorem.

Fix a compact, connected, simply connected Lie group $G$ with finite-dimensional simple real Lie algebra ${\mathfrak{g}}_{\mathbb{R}}$, and put $\mathfrak{g}:={\mathfrak{g}}_{\mathbb{R}}{\otimes }_{\mathbb{R}}\mathbb{C}$. Choose a Cartan subalgebra $\mathfrak{h}$ and positive roots ${\mathrm{\Phi }}^{+}$, let $\theta$ be the highest root, and put $\rho :=\frac{1}{2}{\sum }_{\alpha \in {\mathrm{\Phi }}^{+}}\alpha$. Use the induced-root-form convention of The basic form is the Killing form divided by twice the dual Coxeter number: for each $\lambda \in {\mathfrak{h}}^{\ast }$, $H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}$ is determined by $B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}(H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}},H)=\lambda (H)$ for all $H\in \mathfrak{h}$, and $(\lambda ,\mu {)}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}:=B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}(H_{\lambda ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}},H_{\mu ,\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}})$. Let $c:=2\mathrm{/}(\theta ,\theta {)}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}$ and $B_{\mathrm{b}\mathrm{a}\mathrm{s}}:=B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}\mathrm{/}c$, which is the normalized form $B_0$ of The basic form is the Killing form divided by twice the dual Coxeter number. For each $\lambda \in {\mathfrak{h}}^{\ast }$ define $H_{\lambda ,\mathrm{b}\mathrm{a}\mathrm{s}}$ by $B_{\mathrm{b}\mathrm{a}\mathrm{s}}(H_{\lambda ,\mathrm{b}\mathrm{a}\mathrm{s}},H)=\lambda (H)$ and $(\lambda ,\mu {)}_{\mathrm{b}\mathrm{a}\mathrm{s}}:=B_{\mathrm{b}\mathrm{a}\mathrm{s}}(H_{\lambda ,\mathrm{b}\mathrm{a}\mathrm{s}},H_{\mu ,\mathrm{b}\mathrm{a}\mathrm{s}})$; then $(\theta ,\theta {)}_{\mathrm{b}\mathrm{a}\mathrm{s}}=2$. Put ${\theta }^{\vee }:=2H_{\theta ,\mathrm{b}\mathrm{a}\mathrm{s}}\mathrm{/}(\theta ,\theta {)}_{\mathrm{b}\mathrm{a}\mathrm{s}}=H_{\theta ,\mathrm{b}\mathrm{a}\mathrm{s}}\in \mathfrak{h}$ and $h^{\vee }:=1+\rho ({\theta }^{\vee })$. The scale lemma gives $B_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}=2h^{\vee }{B}_{\mathrm{b}\mathrm{a}\mathrm{s}}$. Let ${\mathrm{\Omega }}_{\mathrm{b}\mathrm{a}\mathrm{s}}$ and ${\mathrm{\Omega }}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}$ denote the pairwise Casimir tensors for these two forms, as in the KZ definition. For a positive integer level $k$, the standard basic-form WZW parameter is ${\kappa }_{\mathrm{b}\mathrm{a}\mathrm{s}}:=k+h^{\vee }$. The KZ definition on this page uses the Killing-form tensor, whose tensor is ${\mathrm{\Omega }}_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}=(2h^{\vee }{)}^{-1}{\mathrm{\Omega }}_{\mathrm{b}\mathrm{a}\mathrm{s}}$; therefore its parameter must be ${\kappa }_{\mathrm{K}\mathrm{i}\mathrm{l}\mathrm{l}}:=(k+h^{\vee })\mathrm{/}(2h^{\vee })$ to give the same connection. This is a normalization conversion, not an identification of affine conformal blocks with the finite-dimensional KZ fibers.

Restrict $B_{\mathrm{b}\mathrm{a}\mathrm{s}}$ to the compact real form and use that real invariant symmetric form as $\kappa$ in the smooth-current central extension defined by the cocycle $\omega (X,Y)=(2\pi {)}^{-1}{\int }_0^{2\pi }\kappa (X(\theta ),Y^{\mathrm{\prime }}(\theta ))d\theta$. Complexify the extension and restrict its loop algebra to Laurent modes $J_m(x):=x{e}^{-im\theta }$, $m\in \mathbb{Z}$. If $K$ is the central generator from the smooth-current extension, set $K_{\mathrm{a}\mathrm{f}\mathrm{f}}:=iK$; then the fixed convention is $[J_m(x),J_n(y)]=J_{m+n}([x,y])+m{\delta }_{m+n,0}{B}_{\mathrm{b}\mathrm{a}\mathrm{s}}(x,y)K_{\mathrm{a}\mathrm{f}\mathrm{f}}$. The derivation $d:=i{\mathrm{\partial }}_{\theta }$ satisfies $[d,J_m(x)]=m{J}_m(x)$ and $[d,K_{\mathrm{a}\mathrm{f}\mathrm{f}}]=0$. The resulting central extension with degree derivation is the affine Lie algebra used here. A representation is at level $k$ when $K_{\mathrm{a}\mathrm{f}\mathrm{f}}$ acts as $kI$. After choosing affine Cartan and positive-root data, an affine highest-weight module is generated by a vector annihilated by the positive-root subalgebra and on which the Cartan acts by a specified weight whose central coordinate is $k$ and whose degree coordinate is recorded; it is integrable when each simple-root Chevalley generator $e_i,f_i$ acts locally nilpotently, meaning that each vector is killed by some positive power of that generator. These are input modules; this definition does not prove their existence or any classification.

For the genus-zero marked curve ${\mathbb{P}}^1$ with distinct ordered points $z_1,\dots ,z_n$ and level-$k$ modules $M_1,\dots ,M_n$, put ${\mathcal{A}}_{\mathbf{z}}:=\mathfrak{g}\otimes \mathcal{O}({\mathbb{P}}^1\setminus \{z_1,\dots ,z_n\})$, where $\mathcal{O}$ denotes rational functions with poles only at the marked points. An interface must specify the action of ${\mathcal{A}}_{\mathbf{z}}$ on $M_1\otimes \cdots \otimes M_n$, including the compatible central-extension lift. Given that action, the coinvariant space is the quotient of the tensor product by the linear span of all action vectors; its algebraic dual (or a specified continuous dual if a completion is used) is the conformal-block space. A top-level $\mathfrak{g}$-module is the specified grade-zero subspace relative to the chosen highest-vector degree, with the action of the constant-loop subalgebra; an interface must state that grading convention and prove finite dimensionality before using it as a KZ fiber. Its identification with these varying conformal-block spaces and the analytic continuation maps must be specified rather than inferred from the word “WZW.”

A sewing datum specifies, for a curve obtained by joining marked local coordinates, the source and target conformal-block spaces, the intermediate modules, the linear gluing map, and any completion topology and convergence condition. Factorization is a theorem about that datum, not part of this definition. On a module where the current modes act as operators, fix normal ordering of their compositions by

$:J_m(x)J_n(y):=\{\begin{array}{ll}{J}_m(x)J_n(y),& m<0,\\ J_n(y)J_m(x),& m\ge 0.\end{array}$

A Sugawara datum is a proposed mode-indexed family of quadratic current operators formed with this ordering. An existence theorem must state its exact coefficients, operator domains, meaning and convergence of any infinite sums, and central charge, and prove the Virasoro relations.

For a quantum-group comparison, specify the category with the exact interface of Braided line representation data: it is a locally finite $\mathbb{k}$-linear abelian rigid monoidal category over a field $\mathbb{k}$, its tensor product is $\mathbb{k}$-bilinear, and $\mathrm{End}(\mathbf{1})\cong \mathbb{k}$; rigid means left and right duals with evaluation and coevaluation maps satisfying the two snake identities. Its braiding $c_{Y,Z}:Y\otimes Z\to Z\otimes Y$ is a natural isomorphism satisfying both hexagon identities with the associator ${\alpha }_{X,Y,Z}:(X\otimes Y)\otimes Z\to X\otimes (Y\otimes Z)$, which satisfies the pentagon identity. For $X$ put $P_0(X)=\mathbf{1}$, $P_1(X)=X$, and $P_{m+1}(X)=P_m(X)\otimes X$; the positive braid generator acts by the supplier's rebracketed adjacent $c_{X,X}$, with the right input passing over the left. These category and coherence hypotheses are part of the input, not consequences of WZW data.

For a ribbon-fusion comparison, use the exact convention of Modular ribbon fusion category and Hopf-link data: $\mathcal{C}$ is a complex finite semisimple rigid tensor category with simple unit and finitely many simple-object classes, equipped with a braiding and ribbon twist $\theta$. The induced pivotal map is $j_X=u_X{\theta }_X:X\to X^{\vee \vee }$, where $u$ is the Drinfeld morphism supplied by A braided rigid category has a Drinfeld morphism, and the associated spherical pivotal structure is part of the data. Its trace is ${\mathrm{tr}}_j(f):={\mathrm{Tr}}_L(j_X\circ f)$ for $f\in \mathrm{End}(X)$, with the left-trace convention of the same supplier. With simple labels $i,j$, the unnormalized Hopf-link entry is $s_{ij}={\mathrm{tr}}_j(c_{j,i}{c}_{i,j})$, and modularity means this matrix is invertible. No positivity of general quantum dimensions is assumed.

Let $N$ be an oriented smooth $3$-manifold and let $L={⨆}_{a=1}^r{S}^1\hookrightarrow N$ be a smooth embedding of a finite disjoint union of oriented circles. Its normal bundle is $\nu L:=TN{\mathrm{\mid }}_L\mathrm{/}TL$. A framing of $L$ is a homotopy class of smooth nowhere-zero sections of $\nu L$. The orientations of $N$ and of each component orient the rank-two normal bundle; a positive unit twist changes the framing by one positive full $2\pi$ rotation as each oriented core circle is traversed, and the ribbon twist specifies its operator. If a semisimplification is used, state the category and quotient ideal: a morphism $f:X\to Y$ is negligible when $\mathrm{tr}(g\circ f)=0$ for every $g:Y\to X$, using the declared pivotal trace. A claimed equivalence must be a specified monoidal equivalence with coherent structure maps preserving the braiding and twist, together with a proof under its exact hypotheses.

The finite-dimensional KZ flatness theorem concerns finite-dimensional $\mathfrak{g}$-modules and does not by itself establish finite-dimensionality of affine conformal blocks, Sugawara well-definedness, sewing/factorization, or a Drinfeld–Kohno or rational-VOA ribbon-category equivalence. The modularity theorem for the stated ${\mathfrak{s}\mathfrak{l}}_2$ root-of-unity quotient establishes modularity of that quotient only; it does not identify its exchanges with WZW monodromy. No generic monodromy unitarity follows.

Infinitesimal braid identities imply KZ flatness and braid monodromy

Loop current Lie algebra and central term

The Schwinger term obeys the Lie cocycle condition

Braided line representation data

The basic form is the Killing form divided by twice the dual Coxeter number

A braided rigid category has a Drinfeld morphism

Modular ribbon fusion category and Hopf-link data

The standard root-of-unity sl2 tilting ribbon quotient is modular