Graded Unitarity in Vertex Operator Algebras

• Graded unitarity is a refined framework for vertex algebras that incorporates a half-integral filtration and quaternionic structure to recover a positive-definite Hermitian form from four-dimensional SCFTs.
• The construction utilizes a nondegenerate invariant bilinear form, Möbius symmetry, and Poisson vertex algebra structures, offering a robust algebraic foundation.
• This approach organizes the classification, BRST reductions, and refined index theory in VOAs, bridging two-dimensional algebra with four-dimensional unitarity constraints.

Graded unitarity is a generalization of the usual notion of unitarity for vertex algebras and vertex operator algebras, designed to capture the “hidden” Hilbert space structure present in chiral algebras arising from supersymmetric quantum field theories even when those algebras are not unitary in the conventional two-dimensional CFT sense. It was introduced in the setting of the SCFT/VOA correspondence, where unitary four-dimensional $\mathcal N=2$ superconformal field theories often produce vertex algebras with negative central charge or logarithmic features, but nevertheless inherit a positive-definite structure after refining the state space by an additional half-integral grading or filtration (Beem et al., 12 Sep 2025, Ardehali et al., 31 Jul 2025).

1. Algebraic framework

Graded unitarity is formulated for Möbius vertex algebras. Such an algebra is a half-integer graded vertex algebra

$$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$

equipped with operators $L_{-1},L_0,L_1$ satisfying the $\mathfrak{sl}_2$ commutation relations and acting compatibly with the vertex operators. The construction does not require a full Virasoro action, although ordinary VOAs form an important subclass (Beem et al., 12 Sep 2025).

A central input is a nondegenerate invariant bilinear form $(\cdot,\cdot)$, in the sense of Frenkel–Huang–Lepowsky. For homogeneous states it satisfies

$$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$

together with

$(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$

When $\mathcal V$ is of CFT type and the invariant form is nondegenerate, the algebra is automatically of strong CFT type and simple (Beem et al., 12 Sep 2025).

The additional structure distinguishing graded unitarity from ordinary unitarity is a half-integral filtration refining the parity decomposition. One writes

$\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$

and equips each parity sector with an increasing filtration

$$\cdots\subseteq \mathcal V^\alpha_{p+\alpha}\subseteq \mathcal V^\alpha_{p+1+\alpha}\subseteq\cdots,\qquad \mathcal V^\alpha=\bigcup_p \mathcal V^\alpha_{p+\alpha}.$$

The OPE must respect this filtration, and the filtration is called good if positive modes lower filtration degree by at least one. Li’s theorem then implies that the associated graded space $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$0 carries a natural Poisson vertex algebra structure (Beem et al., 12 Sep 2025).

If the invariant bilinear form is nondegenerate on each filtered piece, the filtration can be refined to a bigrading

$$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$1

The second grading variable is the $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$2-degree. A key structural consequence is a shortening condition: modes of a field of $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$3-degree $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$4 can change $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$5-degree only within a bounded range, implying in particular that $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$6 for $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$7. The filtration is therefore truncated (Beem et al., 12 Sep 2025).

2. Quaternionic structure, Hilbert space, and adjoints

To turn the invariant bilinear form into a Hermitian inner product, graded unitarity introduces an anti-linear conjugation $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$8 compatible with the vertex algebra structure and with the Möbius operators. If a filtration is present, $$\mathcal V=\bigoplus_{h\in \frac12\mathbb Z}\mathcal V_h$$9 must preserve it. One also defines an operator $L_{-1},L_0,L_1$0 on the refined bigrading by

$L_{-1},L_0,L_1$1

A quaternionic structure is then a conjugation satisfying

$L_{-1},L_0,L_1$2

where $L_{-1},L_0,L_1$3 is the parity operator. The Hermitian form is

$L_{-1},L_0,L_1$4

and graded unitarity requires this form to be positive definite (Beem et al., 12 Sep 2025).

A weak graded-unitary vertex algebra is therefore a quadruple

$L_{-1},L_0,L_1$5

consisting of a Möbius vertex algebra with compatible $L_{-1},L_0,L_1$6-grading, a nondegenerate invariant bilinear form, a nondegenerate half-integral filtration, and a quaternionic structure such that $L_{-1},L_0,L_1$7 is positive definite. If the filtration is good in Li’s sense, the algebra is called graded-unitary (Beem et al., 12 Sep 2025).

This structure produces an orthogonal Hilbert-space decomposition

$L_{-1},L_0,L_1$8

with finite-dimensional $L_{-1},L_0,L_1$9-eigenspaces. A fundamental consequence is the graded spin–statistics relation

$\mathfrak{sl}_2$0

for any nonzero homogeneous $\mathfrak{sl}_2$1. Grassmann parity is thus rigidly tied to conformal weight and $\mathfrak{sl}_2$2-degree (Beem et al., 12 Sep 2025).

Standard unitary Möbius vertex algebras occur as a special case. One may take trivial $\mathfrak{sl}_2$3-grading and trivial filtration, after which the ordinary anti-linear involution and positive-definite Hermitian form induce a weak graded-unitary structure. Graded unitarity therefore extends, rather than replaces, the classical notion of VOA unitarity (Beem et al., 12 Sep 2025).

Canonical examples are the symplectic boson and symplectic fermion VOAs, each endowed with a natural $\mathfrak{sl}_2$4-grading by field number and an explicit anti-linear conjugation. In both cases the resulting inner product is positive and the mode adjoints exchange creation and annihilation operators in the expected way (Beem et al., 12 Sep 2025).

3. Physical origin in the SCFT/VOA correspondence

The main motivation for graded unitarity is the SCFT/VOA correspondence. Vertex algebras arising from four-dimensional $\mathfrak{sl}_2$5 SCFTs inherit strong vestiges of four-dimensional reflection positivity, but they are generally not unitary in the ordinary two-dimensional sense. Their central charge can be negative, and their representation theory can be logarithmic. Graded unitarity is designed precisely to encode the residual positivity structure that survives this passage from four to two dimensions (Beem et al., 12 Sep 2025).

The filtration degree $\mathfrak{sl}_2$6 is physically tied to $\mathfrak{sl}_2$7. In the four-dimensional description, twisted-translated Schur operators of $\mathfrak{sl}_2$8-charge $\mathfrak{sl}_2$9 can only produce components whose twisted $(\cdot,\cdot)$0-weights lie between $(\cdot,\cdot)$1 and $(\cdot,\cdot)$2. This is the physical origin of the shortening condition and the truncation of the $(\cdot,\cdot)$3-filtration (Beem et al., 12 Sep 2025).

The role of the quaternionic structure is likewise physical. The Hermitian conjugate of a four-dimensional Schur operator is generally not itself Schur; rather, it lies in the same $(\cdot,\cdot)$4 multiplet as another Schur operator. Graded unitarity packages this effect algebraically: the anti-linear map $(\cdot,\cdot)$5 implements the relevant conjugation, while $(\cdot,\cdot)$6 supplies the phase needed to convert the invariant bilinear form into a genuine positive-definite Hermitian form (Beem et al., 12 Sep 2025).

When an internal $(\cdot,\cdot)$7-grading $(\cdot,\cdot)$8 is present, one obtains a triple grading $(\cdot,\cdot)$9, and the BPS inequality

$$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$0

appears as the two-dimensional shadow of a four-dimensional BPS bound. This suggests that graded unitarity is the natural vertex-algebraic avatar of four-dimensional unitarity together with $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$1-symmetry, rather than a deformation of ordinary two-dimensional positivity (Beem et al., 12 Sep 2025).

A common misconception is that graded unitarity simply weakens positivity. The actual structure is more rigid: positivity is restored after introducing the correct $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$2-filtration, quaternionic structure, and refined bigrading. The non-unitarity of the underlying chiral algebra is not ignored; it is reorganized.

4. Semi-infinite cohomology and the Kähler package

A major structural advance is the stability of graded unitarity under relative semi-infinite cohomology. Consider a weak graded-unitary vertex algebra $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$3 carrying a chiral quantum moment map for an affine current algebra at level $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$4. The associated relative BRST complex can be written as

$$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$5

and carries two commuting differentials $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$6 related by an outer $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$7-symmetry (Beem et al., 12 Sep 2025).

Under a “good” Hamiltonian action, each $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$8 decomposes into pieces of definite $$(Y(x,z)u,v)=(-1)^{|x||u|}\bigl(u,Y(e^{zL_1}(e^{i\pi}z^{-2})^{L_0}x,z^{-1})v\bigr),$$9-degree, and the BRST complex acquires a package of operators closely analogous to the $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$0, $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$1, adjoint, and Laplacian operators on differential forms of a compact Kähler manifold. One has Kähler-type identities

$(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$2

with $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$3 positive semidefinite and self-adjoint. There is also an $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$4-triple $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$5, with $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$6, playing the role of Lefschetz operators (Beem et al., 12 Sep 2025).

This yields a Hodge decomposition on each finite-dimensional conformal-weight subspace and a quartet mechanism for non-harmonic states. The ensuing $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$7-lemma implies

$(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$8

Because both numerator and denominator are preserved by the quaternionic data, the cohomology inherits graded unitarity. In particular, if $(L_m u,v)=(u,L_{-m}v),\qquad m=-1,0,1.$9 is graded-unitary, then so is its relative semi-infinite cohomology (Beem et al., 12 Sep 2025).

The same formalism yields an outer $\mathcal V$0 action on cohomology. This action extends the natural $\mathcal V$1 action coming from cohomological degree, but it does not preserve that grading. The analogy with the Lefschetz action on Kähler cohomology is explicit, though here the symmetry acts by outer automorphisms rather than by vertex operators (Beem et al., 12 Sep 2025).

The differential graded vertex algebra underlying the BRST complex is also formal: it is quasi-isomorphic to its cohomology endowed with zero differential. This is the vertex-algebraic analogue of the Deligne–Griffiths–Morgan–Sullivan formality theorem. The same Kähler package descends to associated Poisson vertex algebras and leads to formality results for related derived Poisson reductions (Beem et al., 12 Sep 2025).

5. Classification, geometric consequences, and index-theoretic applications

Graded unitarity has rapidly become a selection principle for VOAs expected to arise from four-dimensional SCFTs. For Virasoro and affine Kac–Moody algebras, under natural assumptions on the $\mathcal V$2-filtration, only the $\mathcal V$3 central charges for Virasoro VOAs and the boundary-admissible levels for $\mathcal V$4 and $\mathcal V$5 affine vertex algebras can be compatible with graded unitarity; these are precisely the cases known to arise from four dimensions (Ardehali et al., 31 Jul 2025).

A parallel analysis for $\mathcal V$6 algebras reaches an analogous conclusion. Under the assumption that the $\mathcal V$7-filtration is weight-based with respect to the usual strong generators, all central charges other than those of the $\mathcal V$8 minimal models are incompatible with four-dimensional unitarity. These $\mathcal V$9 algebras are exactly those obtained by principal Drinfel’d–Sokolov reduction from boundary-admissible $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$0 affine current algebras, and they are the known chiral algebras of the $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$1 Argyres–Douglas theories (Beem et al., 17 Feb 2026).

In the semi-infinite-cohomology setting, graded unitarity also has geometric consequences. The associated graded Poisson vertex algebra inherits the same Kähler-type package, and the Hall–Littlewood sector closes under the induced Poisson operations. This leads to a Hall–Littlewood chiral ring and to a formal derived Poisson reduction for large classes of BRST reductions, including examples built from symplectic bosons and symplectic fermions (Beem et al., 12 Sep 2025).

A further development concerns indices. The Schur index of a four-dimensional SCFT is the vacuum character of the associated VOA, but a general VOA-intrinsic derivation of the Macdonald index had remained elusive. Using graded unitarity, a recent proposal recovers a special non-Schur limit of the Macdonald index directly from the VOA by counting positive- and negative-norm states with respect to the graded Hermitian form. The construction extends to surface defects and suggests an analogue of graded unitarity for defect modules (Jiang, 31 Mar 2026).

These developments suggest that graded unitarity is not merely an internal consistency condition. It also organizes classification, BRST reduction, Poisson geometry, and refined index theory.

6. Conceptual status, related terminology, and open directions

Graded unitarity should be distinguished from ordinary unitary VOA theory. Ordinary unitarity demands a positive-definite Hermitian form directly compatible with vertex-operator adjunction. Graded unitarity instead inserts an additional half-integral filtration and a quaternionic structure before positivity emerges. The two notions coincide only in the trivial-filtration limit (Beem et al., 12 Sep 2025).

There is also a distinct operator-algebraic usage of the term in the literature on holomorphic vertex operator superalgebras. In that setting, graded unitarity refers to compatibility of the $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$2-graded superstructure with a positive-definite Hilbert-space form and, in favorable cases, with strong graded locality, leading to graded-local conformal nets. This is conceptually adjacent but not identical to the SCFT/VOA notion based on the $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$3-filtration (Gaudio, 2024).

Several open directions have been identified. One concerns the physical interpretation of the outer $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$4 action appearing on semi-infinite cohomology; suggested connections include dimensional reduction and four-dimensional duality. Another concerns enhanced automorphism groups, since multiple $\mathcal V=\mathcal V^0\oplus \mathcal V^{1/2},$5 actions can appear in some constructions and may combine into larger symmetries. A third concerns the relation between graded unitarity and hyperkähler geometry, especially twistor families of Higgs branches (Beem et al., 12 Sep 2025).

More broadly, graded unitarity has shifted the study of chiral algebras from a purely two-dimensional viewpoint to one controlled by hidden higher-dimensional structure. It provides a mechanism by which vertex algebras that are manifestly non-unitary as two-dimensional theories can nevertheless inherit a precise Hilbert-space interpretation, survive large classes of BRST reductions, and satisfy sharp classification constraints dictated by four-dimensional unitarity.