Quantum Drinfeld--Sokolov Reduction and the McRae--Yang Tensor Functors

Abstract: Let $p,q\geq2$ be relatively prime and set $k=-2+p/q$ and $c_{p,q}=1-6(p-q)<sup>2/(pq)$. McRae and Yang constructed a right exact braided tensor functor [ F_{p,q}:KL^{k}(\mathfrak{sl}_2)\longrightarrow\mathcal O_{c_{p,q}} ] sending the standard Weyl module $V_2$ to the Virasoro Kac module $\mathcal K_{2,1}$, and conjectured that it is exact and naturally isomorphic to quantum Drinfeld--Sokolov reduction. We prove this conjecture. We first determine the images of every indecomposable projective under $F_{p,q}$ without assuming exactness of $F_{p,q}$. The proof combines the projective tensor recursions in $KL<sup>{k}(\mathfrak{sl}_2)$, exact fusion by $\mathcal K_{2,1}$, and generalized conformal-weight blocks. Projective faithfulness and the non-semisimple affine twists then imply that these images are logarithmic. Using Nakano's logarithmic Virasoro extension theorem and his computation of the relevant $\operatorname{Ext}<sup>1$-groups, we identify the projective images and compute their Hom spaces, proving full faithfulness on projectives. Quantum Drinfeld--Sokolov reduction has the same projective images. Compatibility with the non-standard affine twist fixes the scalars needed to construct a natural isomorphism on projectives, and uniqueness of right exact extension then gives [ F_{p,q}\cong H^0_{DS,+}\big|_{KL^{k}(\mathfrak{sl}_2)}. ] In particular, $F_{p,q}$ is exact. The analogous assertion holds after interchanging $p$ and $q$.

• The paper proves the exactness of the McRae–Yang functor in logarithmic conformal field theory.
• Link between Drinfeld-Sokolov reduction and tensor functors established using a specific twist-locking argument.
• Detailed steps include using tensor recursions, Nakano extension theory, and compatibility of transforms to naturality.

Overview

This paper proves a conjecture of McRae and Yang concerning the relationship between two non-semisimple braided tensor categories arising in logarithmic conformal field theory. For relatively prime integers $p,q \geq 2$, with level $k = -2 + p/q$ and central charge $c_{p,q} = 1 - 6(p-q)^2/(pq)$, McRae and Yang constructed a right exact braided tensor functor

$$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$

from the Kazhdan–Lusztig category of finite-length grading-restricted generalized modules for the universal affine vertex algebra $V^k(\mathfrak{sl}_2)$ to the category $\mathcal{O}_{c_{p,q}}$ of finite-length Virasoro modules, sending the standard Weyl module $V_2$ to the Virasoro Kac module $K_{2,1}$. They conjectured that $F_{p,q}$ is exact and naturally isomorphic to quantum Drinfeld–Sokolov reduction. The main theorem establishes this conjecture: $F_{p,q}$ is exact and there is a natural isomorphism of $k = -2 + p/q$0-linear functors

$k = -2 + p/q$1

with the analogous statement holding after interchanging $k = -2 + p/q$2 and $k = -2 + p/q$3.

The proof faces a genuine technical obstacle. The source projective $k = -2 + p/q$4 admits a Weyl filtration whose left-hand injection would not be preserved by a merely right exact functor, so the images of projectives cannot be read off from that filtration. The argument instead recovers all projective images from tensor recursions, then uses Virasoro extension theory (Nakano's results on triplet $k = -2 + p/q$5-algebra modules) to compute Hom spaces and establish full faithfulness on projectives. A "twist-locking" argument using the nilpotent part of the non-standard affine twist fixes the remaining scalars, and uniqueness of right exact extension completes the proof.

Preliminaries and the Nakano extension input

The paper assembles three bodies of prior results. First, McRae–Yang's structure theory for $k = -2 + p/q$6: the category has enough projectives; for $k = -2 + p/q$7 and $k = -2 + p/q$8, one has $k = -2 + p/q$9, so $c_{p,q} = 1 - 6(p-q)^2/(pq)$0 is simple-projective; and for $c_{p,q} = 1 - 6(p-q)^2/(pq)$1, $c_{p,q} = 1 - 6(p-q)^2/(pq)$2, the projective $c_{p,q} = 1 - 6(p-q)^2/(pq)$3 has composition factors $c_{p,q} = 1 - 6(p-q)^2/(pq)$4 and admits a non-split Weyl filtration. Crucially, each non-wall projective is logarithmic, its endomorphism algebra is $c_{p,q} = 1 - 6(p-q)^2/(pq)$5 with $c_{p,q} = 1 - 6(p-q)^2/(pq)$6, and the non-standard twist acts as

$c_{p,q} = 1 - 6(p-q)^2/(pq)$7

Second, McRae–Sopin's results on Virasoro Kac modules $c_{p,q} = 1 - 6(p-q)^2/(pq)$8 at charge $c_{p,q} = 1 - 6(p-q)^2/(pq)$9: fusion by the rigid module $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$0 is exact, with $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$1 fitting into a short exact sequence involving $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$2 and $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$3 that splits exactly when $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$4.

Third, Nakano's extension theory. The paper works throughout inside Nakano's category $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$5 of finite-socle-length generalized Virasoro modules with simples in the Felder weight set — an extension-closed subcategory, so all Ext groups are Yoneda Ext groups in precisely that category. A lemma matching Felder labels shows that for $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$6, $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$7, $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$8, there is a triple $$F_{p,q}: KL^k(\mathfrak{sl}_2) \longrightarrow \mathcal{O}_{c_{p,q}}$$9 with ordered weights $V^k(\mathfrak{sl}_2)$0 and with $V^k(\mathfrak{sl}_2)$1, $V^k(\mathfrak{sl}_2)$2. Nakano's Theorem 5.11 and Proposition 5.12 then supply the two facts used later: quotients of logarithmic extensions are indecomposable, and the relevant extension space $V^k(\mathfrak{sl}_2)$3 is one-dimensional.

A methodological point deserves emphasis: the paper explicitly avoids assuming any socle description or self-contragredient duality for general Nakano modules $V^k(\mathfrak{sl}_2)$4, since Nakano himself notes that the general socle statement is not determined by his construction. All Hom computations below proceed without such input.

Exactness of restricted Drinfeld–Sokolov reduction

Before comparing functors, the paper establishes that the restriction $V^k(\mathfrak{sl}_2)$5 is well-defined and exact. This requires care because objects of $V^k(\mathfrak{sl}_2)$6 need not have semisimple affine conformal Hamiltonian. Using Arakawa's BRST vanishing theorem — which in type $V^k(\mathfrak{sl}_2)$7 requires no block condition, per Remark 9.1.5(ii) of Arakawa's $V^k(\mathfrak{sl}_2)$8-algebra memoir — vanishing of $V^k(\mathfrak{sl}_2)$9 ($\mathcal{O}_{c_{p,q}}$0) holds on simples and propagates by induction on Jordan–Hölder length through long exact sequences. Hence $\mathcal{O}_{c_{p,q}}$1 is exact, without any semisimplicity hypothesis on the objects.

Two identifications follow. For Weyl modules, $\mathcal{O}_{c_{p,q}}$2: reduction sends the affine Verma module to the Virasoro Verma module, the character computation forces cyclicity, and the kernel is generated by the unique level-$\mathcal{O}_{c_{p,q}}$3 singular vector. For simples, $\mathcal{O}_{c_{p,q}}$4, verified by matching the transformed highest weight $\mathcal{O}_{c_{p,q}}$5 against the Feigin–Fuchs weight $\mathcal{O}_{c_{p,q}}$6 by direct expansion. Faithfulness of $\mathcal{O}_{c_{p,q}}$7 follows immediately, since $\mathcal{O}_{c_{p,q}}$8 is exact and nonzero on every simple.

The twist compatibility is proved directly at the level of the BRST complex. Writing out the diagonal $\mathcal{O}_{c_{p,q}}$9-action and Arakawa's new Hamiltonian yields

$V_2$0

and since the last term has integral spectrum while $V_2$1-eigenvalues lie in $V_2$2 on degree-$V_2$3 objects, exponentiation gives $V_2$4. This compatibility is what later locks naturality scalars.

Projective images without exactness

The first main step determines $V_2$5 for every indecomposable projective, using only right exactness of $V_2$6. For small Weyl modules, Krull–Schmidt cancellation applied to the split affine fusion sequences gives $V_2$7 for $V_2$8.

For the general projective-image theorem, the argument is a simultaneous induction along strips between $V_2$9-walls. Tensoring the known filtration for $K_{2,1}$0 with $K_{2,1}$1 — exact by rigidity — and combining with the source projective recursions produces four Kac factors that separate into two congruence classes modulo $K_{2,1}$2 of generalized conformal weights. The key arithmetic input is that certain weight differences are integral while crossed differences carry nonintegral parts $K_{2,1}$3, so the block projection lemma (which is exact, since Virasoro modes shift generalized weights by integers) cleanly isolates each block. Comparing Jordan–Hölder lengths against already-known summands forces the new summands to have the asserted filtrations:

$K_{2,1}$4

The paper stresses that no injection here is obtained by applying $K_{2,1}$5 to the source Weyl filtration; every injection comes from exact tensoring or exact block projection. This independence from exactness of $K_{2,1}$6 is essential to the logical structure of the proof.

Faithfulness of $K_{2,1}$7 on the projective subcategory is proved separately for $K_{2,1}$8 and $K_{2,1}$9. For $F_{p,q}$0, the kernel pulls back along the monoidal equivalence between the projective subcategory and the quantum $F_{p,q}$1 tilting category at $F_{p,q}$2 to a tensor ideal of the tilting category. Quantum Schur–Weyl duality identifies Hom spaces between tensor powers with Temperley–Lieb morphisms, and Goodman–Wenzl's ideal theorem forces any proper nonzero tensor ideal to contain the identity of the critical tilting object $F_{p,q}$3 — hence of $F_{p,q}$4, contradicting $F_{p,q}$5. The case $F_{p,q}$6 falls outside the range of the Goodman–Wenzl theorem (the loop parameter vanishes), and the paper supplies a separate argument: identities survive because all projective images are nonzero; adjacent zigzag arrows survive via rigidity adjunctions normalized compatibly with the functor; and radical endomorphisms survive because preservation of the non-semisimple twist forces $F_{p,q}$7 and linearly independent from the identity.

Logarithmicity and full faithfulness on projectives

With faithfulness in hand, logarithmicity of the non-wall projective images follows at once: applying $F_{p,q}$8 to the twist formula preserves the non-semisimple nilpotent part, so $F_{p,q}$9 has non-semisimple $F_{p,q}$0, and consequently the filtration above cannot split. Nakano's extension theory then identifies these images precisely: the quotient $F_{p,q}$1 represents a class in the correct Yoneda Ext group, is indecomposable by Nakano's Theorem 5.11, and the one-dimensionality of the relevant Ext space forces $F_{p,q}$2.

The Hom-space computation proceeds along reflection chains $F_{p,q}$3, $F_{p,q}$4. Long exact sequence arguments, together with the elementary fact that a non-split length-two module with distinct composition factors has scalar endomorphisms, yield the target Hom table within a chain:

Pair	Dimension
$F_{p,q}$5, $F_{p,q}$6	2
$F_{p,q}$7	1
$F_{p,q}$8, $F_{p,q}$9	1
$k = -2 + p/q$00, $k = -2 + p/q$01	0

Upper bounds come from the filtrations; lower bounds come from faithfulness of $k = -2 + p/q$02 applied to the known nonzero source morphisms. Cross-chain vanishings require attention only when $k = -2 + p/q$03, where the weight equality $k = -2 + p/q$04 creates collisions between distinct chains; these are resolved by direct analysis of the two non-split sequences sharing the common simple $k = -2 + p/q$05. Wall objects $k = -2 + p/q$06 are simple, and a weight-collision check excludes any Hom with non-wall images. Since the source projective Hom algebra has identical zigzag dimensions, and injectivity of $k = -2 + p/q$07 on each Hom space follows from faithfulness, the restriction $k = -2 + p/q$08 is fully faithful.

Comparison and the natural isomorphism

Exactness of $k = -2 + p/q$09 lets one apply it to the source Weyl filtration, producing the same filtration shape for $k = -2 + p/q$10; the same Nakano argument, using twist preservation to certify logarithmicity, gives objectwise isomorphisms $k = -2 + p/q$11 for every indecomposable projective.

Naturality is the delicate point, and the twist-locking lemma resolves it. Any Virasoro-module isomorphism $k = -2 + p/q$12 commutes with the target twist; since both functors preserve the source minus twist, canceling scalar and identity terms forces

$k = -2 + p/q$13

where $k = -2 + p/q$14 spans the radical of $k = -2 + p/q$15. On a reflection chain, one constructs $k = -2 + p/q$16 inductively: forward naturality is arranged by rescaling in the one-dimensional adjacent Hom space, and the reverse direction introduces an unknown scalar $k = -2 + p/q$17; composing both relations and invoking twist locking forces $k = -2 + p/q$18. Since identities, adjacent arrows, and radical loops span all nonzero Hom spaces, the family is natural on the whole projective subcategory.

McRae–Yang's right-exact extension theorem states that a $k = -2 + p/q$19-linear functor on the projective subcategory extends uniquely, up to natural isomorphism, to a right exact functor on the entire abelian category. Both $k = -2 + p/q$20 and $k = -2 + p/q$21 are right exact extensions of naturally isomorphic restrictions, hence $k = -2 + p/q$22 on $k = -2 + p/q$23, and exactness of $k = -2 + p/q$24 follows from exactness of $k = -2 + p/q$25. Finally, since every step was proved for an arbitrary ordered coprime pair, applying the result to $k = -2 + p/q$26 — with the identification $k = -2 + p/q$27 via uniqueness of singular vectors in the shared Verma module — yields the transposed statement, completing Conjecture 7.16 of McRae–Yang.

Limitations and open questions

The paper is explicit about the boundaries of its methods. The $k = -2 + p/q$28 case required a separate faithfulness proof because the Goodman–Wenzl ideal theorem does not cover the zero loop parameter; the resulting argument relies on explicit adjunction normalizations rather than the general ideal classification. The natural isomorphism obtained is one of $k = -2 + p/q$29-linear functors only: transporting the braided tensor structure of $k = -2 + p/q$30 equips the restricted reduction with some braided monoidal structure, but whether this agrees with any independently constructed monoidal structure on Drinfeld–Sokolov reduction is left as a separate question not addressed here. The Hom computations deliberately avoid socle and duality statements for general Nakano modules $k = -2 + p/q$31, which remain undetermined in Nakano's framework; any future simplification of the argument via such structure would require resolving that gap independently.

Conclusion

The paper settles Conjecture 7.16 of McRae–Yang: the braided tensor functor $k = -2 + p/q$32 from the admissible-level Kazhdan–Lusztig category for affine $k = -2 + p/q$33 to the Virasoro module category at charge $k = -2 + p/q$34 is exact and naturally isomorphic to the principal quantum Drinfeld–Sokolov reduction, in both orientations of the parameter pair. The proof combines projective tensor recursions, exact block projections by generalized conformal weight, Nakano's one-dimensional extension spaces, and a twist-locking mechanism that converts preservation of the non-standard affine twist into naturality. Beyond confirming the conjecture, the result identifies the images of all indecomposable projectives as explicit Nakano modules $k = -2 + p/q$35 and computes the full Hom algebra of these images, providing a concrete bridge between the representation theory of affine $k = -2 + p/q$36 at admissible levels and logarithmic Virasoro minimal models.