Modular Functor: Topology and Applications

• Modular functor is a mathematical structure that assigns finite-dimensional vector spaces to labeled surfaces, capturing anyon fusion, braiding, and topological spin.
• It uses gluing axioms and mapping class group representations to translate topological manipulations into coherent algebraic operations.
• Categorical formulations express modular functors as symmetric monoidal 2-functors, linking TQFT, conformal field theory, and quantum topology.

Searching arXiv for recent and foundational papers on modular functors. Searching for "modular functor" on arXiv. A modular functor is, roughly speaking, the topological package that turns the algebra of anyons into a genuinely two-dimensional theory; in its classical form, it is a coherent system of projective representations of mapping class groups of surfaces on spaces of conformal blocks, together with compatibility under gluing (Burton, 2016, Brochier et al., 2022). In the anyon setting, it explains how fusion, associativity, braiding, and topological spin arise from the topology of surfaces rather than from particles arranged on a line. In conformal field theory, it organizes the spaces attached to surfaces and the way they transform under sewing. In categorical formulations, it is expressed as a modular algebra over an extension of the surface operad, or as a symmetric monoidal $2$-functor or double functor. Across these formulations, the common structure is the assignment of linear data to labeled surfaces and of coherent maps to topological manipulations of those surfaces.

1. Classical geometric meaning and basic axioms

In the anyon and topological-order setting, the basic intuition is that anyons in a $2$D topologically ordered system should be modeled not merely as algebraically ordered labels but as holes or boundary components on a compact oriented $2$D manifold with boundary. A modular functor assigns to each such labeled surface a finite-dimensional complex vector space $H(M)$ of fusion states, and to each allowed topological map $f:M\to N$ a unitary map

$H(f):H(M)\to H(N),$

depending only on the isotopy class of $f$ and functorial under composition (Burton, 2016).

The unit axioms encode the vacuum and antiparticle rules. For a finite label set $A$ with vacuum $I$ and involution $a\mapsto \widehat a$, the empty surface satisfies $2$0; a disc with vacuum label satisfies $2$1; a disc with non-vacuum label satisfies $2$2 for $2$3; an annulus with matching labels satisfies $2$4; and a mismatched annulus has $2$5 for $2$6. Disjoint union is monoidal: $2$7 Orientation reversal gives the dual space,

$2$8

The central geometric operation is gluing. If two boundary components labeled $2$9 and $2$0 are glued to form a new surface $2$1, the gluing axiom gives an isomorphism

$2$2

depending only on the isotopy class of the seam. This is the topological origin of summing over internal charge sectors. The seam also defines projectors

$2$3

so that simple closed curves on the surface become observables measuring total anyonic charge inside the curve. A Dehn twist around an annulus labeled $2$4 acts by a phase $2$5, giving the geometric origin of topological spin.

2. Fusion, braiding, mapping class groups, and coherence

The familiar anyon operations $2$6, $2$7, and $2$8 acquire a geometric meaning once surfaces are decomposed into elementary pieces. Pair-of-pants decompositions supply the local building blocks. Comparing two such decompositions of the same surface yields the $2$9-move,

$H(M)$0

which is the topological form of reassociating fusion, $H(M)$1. A half-twist of two boundary components yields the $H(M)$2-move

$H(M)$3

so braiding is realized by actual surface topology rather than by formal exchange symbols (Burton, 2016).

String diagrams and skeins encode these operations, but their deeper interpretation is geometric: the strings are world-lines of anyons in $H(M)$4 dimensions. This is why the braid group, rather than the symmetric group, governs exchange in two dimensions. The mapping class group enters through isotopy classes of diffeomorphisms,

$H(M)$5

with half-twists exchanging punctures appearing as mapping class group elements. In a Schrödinger-picture view, braids act on states; in a Heisenberg-picture view, mapping class group elements act on observables. A modular functor packages both viewpoints into a single topological formalism (Burton, 2016).

In the classical conformal-field-theoretic language, this same structure is described as a coherent system of projective representations of mapping class groups of surfaces on spaces of conformal blocks, together with compatibility under gluing. The coherence conditions are the usual pentagon and hexagon equations: the pentagon arises from comparing pair-of-pants decompositions of a four-punctured sphere, and the hexagon expresses compatibility of braiding with fusion. These relations ensure that all decompositions and re-decompositions of a surface produce compatible linear maps (Brochier et al., 2022).

3. Axiomatizations and categorical classification

The literature contains multiple inequivalent axiomatizations of modular functors, and the precise bijection between anyon theory, modular tensor categories, and any one formulation of modular functor is not completely settled in all details. For that reason, physically motivated and useful formulations coexist with more categorical and operadic definitions (Burton, 2016).

A modern categorical formulation describes a modular functor in a symmetric monoidal $H(M)$6-category $H(M)$7 of linear categories as a modular algebra over an extension

$H(M)$8

of the modular surface operad, relative to the handlebody operad $H(M)$9. The extension is required to be reasonable: it must be relative to $f:M\to N$0 and admit insertions of vacua, encoded by the condition that for every oriented embedding $f:M\to N$1, the induced functor

$f:M\to N$2

is an equivalence. In this framework, genus-zero data are exactly a cyclic framed $f:M\to N$3-algebra, equivalently a self-dual balanced braided algebra $f:M\to N$4 with multiplication $f:M\to N$5, braiding $f:M\to N$6, balancing $f:M\to N$7, and non-degenerate pairing $f:M\to N$8, satisfying

$f:M\to N$9

Factorization homology then supplies the extension from genus zero to higher genus via

$H(f):H(M)\to H(N),$0

with connectedness as the criterion that determines whether the genus-zero data extend to a full modular functor (Brochier et al., 2022).

The classification theorem states that the moduli space of modular functors is equivalent to the $H(f):H(M)\to H(N),$1-groupoid of connected self-dual balanced braided algebras in $H(f):H(M)\to H(N),$2, and the equivalence is afforded by restricting modular functors to genus-zero surfaces. The extension, if it exists, is empty or contractible, so a modular functor is completely determined by its genus-zero data once extension is possible. Cofactorizability gives a sufficient criterion for connectedness. This recovers Lyubashenko’s construction from a not necessarily semisimple modular category and also produces examples that do not come from modular categories, including examples from vertex operator algebras and Drinfeld centers of pivotal finite tensor categories (Brochier et al., 2022).

A different but compatible construction starts from an arbitrary modular tensor category and produces a Walker modular functor by adapting Turaev’s framework. The resulting object is a symmetric monoidal functor

$H(f):H(M)\to H(N),$3

equipped with a gluing isomorphism, a duality pairing, and, in the unitary case, a Hermitian structure. The construction depends on chosen isomorphisms $H(f):H(M)\to H(N),$4, but different choices yield quasi-isomorphic modular functors (Andersen et al., 2016).

4. Relations with topological field theory, modular tensor categories, and conformal blocks

Modular functors are topologically rooted constructs and are closely related to $H(f):H(M)\to H(N),$5D topological quantum field theories. In practice, a modular functor can be seen as part of a TQFT, and physicists sometimes use “TQFT” where mathematicians might say “modular functor.” On the algebraic side, anyon theory is the study of braided fusion tensor categories, specifically modular tensor categories; the modular functor is the topological incarnation of that algebraic structure (Burton, 2016).

This relationship is explicit in constructions from three-dimensional field theory. For a not necessarily semisimple modular tensor category $H(f):H(M)\to H(N),$6, one obtains a chiral modular functor as a symmetric monoidal $H(f):H(M)\to H(N),$7-functor from a bordism $H(f):H(M)\to H(N),$8-category to a $H(f):H(M)\to H(N),$9-category of finite linear categories and left exact profunctors. Gluing is expressed by a coend,

$f$0

and the construction recovers Lyubashenko’s modular functor from the non-semisimple $f$1D TFT of De Renzi–Gainutdinov–Geer–Patureau-Mirand–Runkel. Pullback along bordisms with orientation-reversing involution cancels the gluing anomaly, and doubling identifies the resulting theory with the modular functor for the Drinfeld center $f$2 (Hofer et al., 2024).

A parallel comparison appears in non-semisimple string-net theory. For a pivotal finite tensor category $f$3, the finitely cocompleted string-net modular functor built from projective objects is equivalent to Lyubashenko’s modular functor for the Drinfeld center $f$4. The circle category becomes the Kleisli category of the central monad on projectives, and after finite cocompletion one obtains

$f$5

This gives a topological, skein-theoretic model for Lyubashenko’s modular functor far beyond the semisimple spherical case (Müller et al., 2023).

In rational conformal field theory, conformal blocks themselves produce modular functors. For a strongly rational vertex operator algebra $f$6, the spaces of conformal blocks on moduli of stable marked curves with tangent vectors define a modular functor over $f$7. The proof rests on propagation of vacua and gluing or factorization, expressed by isomorphisms such as

$f$8

and on the semisimple factorization of nodal degenerations into sums over simple modules. As consequences, the category $f$9 inherits a ribbon Grothendieck–Verdier structure and then a modular fusion category structure; the modular functor is canonically equivalent, up to contractible choice, to the Reshetikhin–Turaev modular functor of $A$0, extends to a once-extended $A$1-dimensional topological field theory, and admits a factorization-homological description (Damiolini et al., 8 Jul 2025).

Additional constructions emphasize that modular functors need not be vector-space-valued in the classical sense. State-sum constructions for finite tensor categories and their bimodules yield framed modular functors valued in bicategories of finite linear categories and left exact functors, with $A$2-framings replacing pivotal structures (Fuchs et al., 2019). Chain-complex-valued constructions assign to each labeled surface a chain complex satisfying excision via homotopy coends and recover Lyubashenko’s mapping class group representations in zeroth homology (Schweigert et al., 2020). In the string-net approach to full conformal field theory, the modular functor is refined to a symmetric monoidal double functor, and field functors together with universal correlators form a monoidal vertical transformation that is an equivalence (Fuchs et al., 5 May 2026).

5. Generalizations beyond the finite-dimensional semisimple setting

The standard Segal or Atiyah–Moore–Seiberg framework can be extended from finite-dimensional vector spaces to Hilbert spaces of possibly infinite dimension. In this generalized setting, an $A$3-labeled marked surface

$A$4

is sent to a Hilbert space, and factorization is no longer a finite direct sum but a direct integral,

$A$5

The disjoint union axiom remains tensorial, and the gluing isomorphism is required to be associative, compatible with gluing of morphisms, compatible with disjoint unions, and independent of the choice of gluing map. When conformal dimensions $A$6 are added, Dehn twists act by scalars, and the generalized modular functor determines unique wave functions as flat sections. In the Liouville conformal field theory example, normalized gluing conformal blocks define the relevant Hilbert spaces, and the resulting wave functions coincide with eigenfunctions in quantum Teichmüller theory (Ichikawa, 2020).

Other extensions change the target and the notion of locality. One formulation broadens the definition so that super-line-valued theories, Clifford-valued theories, and holomorphic modular functors with nontrivial topological twisting are included. In that framework, the correct categorical language is weak multicategories and $A$7-categories, the source is a multicategory of spin surfaces, and the target may be a weak $A$8-category built from Clifford algebras and bimodules. Topological twisting records the projective scalar ambiguity in gluing and is controlled, up to equivalence, by the central charge; the formal realization lands in twisted $A$9-modules (Kriz et al., 2013).

A related quantization program treats the modular-functor output for families of rational $I$0D CFTs with gauge symmetries parametrized by a $I$1-space $I$2 as a holomorphic sheaf over a universal elliptic curve. Using dominant $I$3-theory for positive-energy representations of the loop group $I$4, one constructs a sheaf $I$5 whose stalks are cohomological functors of $I$6 and whose global sections recover Weyl-invariant theta functions of degree $I$7. This theory is interpreted as a model of equivariant elliptic cohomology (Kitchloo, 2014).

The phrase “algebraic modular functor” also appears in higher Teichmüller theory. For $I$8, the Fock–Goncharov quantizations $I$9 of cluster Poisson varieties satisfy the conjectured cutting property: cutting a marked surface along an essential simple closed curve leads to a canonical, mapping-class-group-equivariant algebra isomorphism that reconstructs the algebra of the original surface from a residue universal Laurent ring associated to the cut surface. Here the modular-functor pattern is preserved, but the objects assigned to surfaces are noncommutative algebras rather than vector spaces (Schrader et al., 4 Sep 2025).

6. Scope, examples, and conceptual issues

A persistent conceptual issue is that “modular functor” names a pattern rather than a single rigid formalism. The common thread is coherent assignment to surfaces and coherent compatibility with cutting and regluing, but the concrete targets vary: finite-dimensional vector spaces, Hilbert spaces, linear categories, profunctors, chain complexes, twisted $a\mapsto \widehat a$0-modules, and noncommutative algebras all occur in the literature surveyed here. This suggests that the invariant core of the notion is the gluing-compatible surface assignment, not a fixed codomain.

The same breadth appears in applications. Permutation equivariant modular functors provide module category structures and modular invariant partition functions: starting from a finite group $a\mapsto \widehat a$1, a finite $a\mapsto \widehat a$2-set $a\mapsto \widehat a$3, and a modular tensor category $a\mapsto \widehat a$4, the associated $a\mapsto \widehat a$5-equivariant modular functor produces the graded category

$a\mapsto \widehat a$6

constructs the action of the neutral component on each graded piece from $a\mapsto \widehat a$7-covers of punctured spheres, and yields the modular invariant

$a\mapsto \widehat a$8

Here modular functors serve as a geometric-to-algebraic bridge between topology of covers and module-category data (Barmeier, 2010).

A second point of scope is that not all modular functors come from modular categories. Examples from $a\mapsto \widehat a$9-cofinite vertex operator algebras and from Drinfeld centers of pivotal finite tensor categories show that connected self-dual balanced braided algebras can define modular functors without arising from modular categories in the semisimple sense. Cofactorizability is sufficient for extension, but not necessary in every example (Brochier et al., 2022).

Finally, the relation between anyon theory and surface topology remains a central interpretive point. The one-dimensional algebraic operations of fusion, reassociation, and braiding are best understood as shadows of two-dimensional topology. A modular functor makes this precise: holes and labeled boundary components replace a line of particles, isotopy replaces formal rewriting, gluing replaces summation over internal channels, and mapping class groups encode the global symmetry of the theory. In that sense, modular functors provide the topological language in which anyon theory, conformal blocks, and low-dimensional field theory become different presentations of the same structural idea (Burton, 2016).