Twisted Fourier–Mukai Partner

• Twisted Fourier–Mukai partner is a cohomological notion that generalizes classical derived equivalences to settings with Brauer twists and Gₘ-gerbes.
• It plays a key role in algebraic geometry, connecting invariants of K3 surfaces, abelian varieties, Enriques surfaces, and special cubic fourfolds via twisted derived categories.
• The construction relies on explicit twisted kernels that preserve modified Mukai lattices and B-field invariants, providing a framework for classifying derived equivalences.

A twisted Fourier–Mukai partner is a fundamentally cohomological notion in the theory of nonlinear dualities for derived categories. It generalizes the classical concept of Fourier–Mukai partners to the setting where derived categories of twisted sheaves—specifically, sheaves twisted by Brauer classes or $\mathbb{G}_m$-gerbes—are considered. The theory underpins advances in the study of derived equivalences in algebraic geometry, especially in the context of K3 surfaces, abelian varieties, Enriques surfaces, cubic fourfolds, and noncommutative deformations.

1. Definition and Formalism

Let $X$ be a smooth projective variety over an algebraically closed field (or more generally, over $\mathbb{C}$) and $\alpha\in \mathrm{Br}(X)$ a Brauer class. The bounded derived category $D^b(X,\alpha)$ of $\alpha$-twisted coherent sheaves generalizes $D^b(X)$ by admitting objects whose transition data are nontrivially twisted by $\alpha$ in the étale topology. Given two such pairs $(X,\alpha)$ and $(Y,\beta)$, a twisted Fourier–Mukai partner is defined by the existence of an exact linear equivalence

$$\Phi_P: D^b(X,\alpha)\xrightarrow{\sim} D^b(Y,\beta)$$

realized by an object $P \in D^b(X\times Y, \alpha^{-1}\boxtimes \beta)$, called the twisted Fourier–Mukai kernel. The integral transform is

$$\Phi_P(-) = \mathbf{R}\pi_{Y*}(\pi_X^*(-)\otimes^{\mathbf{L}} P)$$

where $\pi_X, \pi_Y$ are the projections. Such equivalences must respect the twisting data: the kernel $P$ is simultaneously twisted along both factors $(X, \alpha^{-1})$ and $(Y, \beta)$ (Lane, 23 Jan 2026, Li, 24 Nov 2025).

This definition encompasses a significant range of geometries: K3 surfaces and their moduli, abelian varieties, Enriques surfaces, special cubic fourfolds with associated (possibly twisted) K3 surfaces, and even noncommutative complex tori (Pertusi, 2016, Addington et al., 2018, Lane, 23 Jan 2026, Okuda, 2023, Li, 24 Nov 2025).

2. Cohomological and Lattice-Theoretic Invariants

The classification and existence of twisted Fourier–Mukai partners are governed by deep lattice-theoretic and Hodge-theoretic phenomena. For surfaces such as K3s, the structure of twisted Mukai lattices and the behavior of their weight-two Hodge structures under B-field twists are central. Given a B-field $B\in H^2(X, \mathbb{Q})$ with $\exp(B)=\alpha$, the total cohomology $H^*(X, \mathbb{Z})$ is endowed with the “twisted” Hodge structure $H^{2,0}(X, \alpha) = \exp(B)\cdot H^{2,0}(X)$. The algebraic part $$N(X,\alpha) = H^{1,1}(X,\alpha) \cap H(X,\alpha,\mathbb{Z})$$ generalizes the Néron–Severi lattice, and its orthogonal $T(X,\alpha)$ is the twisted transcendental lattice.

For abelian varieties, the Mukai lattice approach persists, but the main invariants are isogeny classes of the Picard variety and the existence of isogenies whose kernels have specific built-in symplectic or isotropic structure, often of the type $\oplus(\mathbb{Z}/m\mathbb{Z})^2$ (Li, 24 Nov 2025, Lane, 23 Jan 2026).

Special cubic fourfolds with discriminant $d$ admit an associated (twisted) K3 surface $(X,\alpha)$ if there is a Hodge isometry between the Mukai lattice of Kuznetsov’s K3-like subcategory of $Y$ and $H(X,\alpha, \mathbb{Z})$ (Pertusi, 2016).

3. Twisted Fourier–Mukai Partners in Key Geometric Settings

A. Abelian Varieties

Any twisted Fourier–Mukai partner of an abelian variety is necessarily an abelian variety, and two such twisted abelian varieties $(X,\alpha)$ and $(Y,\beta)$ are equivalent if and only if there exists an isogeny of dual abelian varieties $\widehat{X} \to Y$ with kernel of the form $\oplus_{i=1}^r(\mathbb{Z}/m_i\mathbb{Z})^2$, and the corresponding Brauer classes are compatible via the induced gerbe (Lane, 23 Jan 2026, Li, 24 Nov 2025). The main theorem classifies twisted equivalences cohomologically and in terms of semi-homogeneous vector bundles and Poincaré-type kernels.

B. K3 and Enriques Surfaces

For K3 surfaces, a twisted Fourier–Mukai partner $(Y, \beta)$ of $(X,\alpha)$ arises precisely when there is a Hodge isometry of their twisted Mukai lattices. In characteristic $p$ and tame settings (i.e., $p\nmid$ order of $\alpha$), the number of such partners is finite and can be counted via lattice-theoretic data (Srivastava et al., 2021). For Enriques surfaces, the situation is rigid: a complex Enriques surface $(X,\alpha)$ with nontrivial Brauer class $\alpha$ has no nontrivial twisted Fourier–Mukai partners except itself with the same class (Addington et al., 2018).

C. Special Cubic Fourfolds

A smooth cubic fourfold admitting an associated twisted K3 surface $(X,\alpha)$ allows a precise count of its twisted Fourier–Mukai partners. For discriminant $d = k^2c$, the number $m'$ of twisted partners of order $k$ is computed by explicit lattice formulas involving the Euler totient $\varphi(k)$ and the number of odd prime factors of $c$, with a dichotomy depending on $d$ mod 6:

• If $d \equiv 2 \pmod{6}$, $|FM(Y)|=m'$,
• If $d \equiv 0 \pmod{6}$, $|FM(Y)| = \lceil m'/2 \rceil$ (Pertusi, 2016).

D. Noncommutative Complex Tori and Gerbes

Given a complex torus $X$ and a torsion $2$-cocycle $\lambda \in Z^2(\widehat{T},\mu_N)$, the category of coherent sheaves on the corresponding non-commutative torus $X_\lambda$ is derived equivalent to the category of $\alpha$-twisted sheaves on the dual torus $\widehat{X}$, with $$\alpha = \iota_P(\lambda)\in \mathrm{Br}(\widehat{X})$$. The equivalence is implemented by a twisted Poincaré kernel, making $(\widehat{X}, \alpha)$ a twisted Fourier–Mukai partner of the noncommutative $X_\lambda$ (Okuda, 2023).

4. Enumerative and Classification Results

The enumeration of twisted Fourier–Mukai partners can be explicit in favorable settings:

• For very general special cubic fourfolds with associated twisted K3 surfaces of rank-one Néron–Severi lattice, the number

$$m' = \begin{cases} \frac12\varphi(k)2^{h-1} & (k>2) \ 2^{h-1} & (k=2) \ 1 & (k=1) \end{cases}$$

where $k^2c = d$ and $h$ is the number of distinct odd prime divisors of $c$ (Pertusi, 2016).

• In positive characteristic, the number of tame twisted partners of an ordinary untwisted K3 surface equals the count over $\mathbb{C}$ given by Ma’s formula involving orbits of isotropic elements in the discriminant form of the Néron–Severi lattice and their genera (Srivastava et al., 2021).

5. Implications, Special Cases, and Further Directions

Twisted Fourier–Mukai equivalences have powerful invariance and rigidity properties. They preserve key structures such as isogeny classes (abelian varieties), Newton polygons (K3 surfaces in positive characteristic), and transcendental data via Mukai lattices and B-field twistings.

Rigidity results:

• Enriques surfaces with non-trivial Brauer class are uniquely determined up to isomorphism by their twisted derived category (Addington et al., 2018).
• For abelian varieties, the isogeny class of the dual is a derived invariant even in the twisted setting (Lane, 23 Jan 2026).
• In generic cases for complex tori with Brauer class of prime order, only duality and identity (up to isomorphism) exhaust the twisted partners (Li, 24 Nov 2025).

A plausible implication is that twisted derived equivalence can be used to probe or classify fine arithmetic and geometric invariants where untwisted equivalence is too coarse, particularly in the study of derived invariants and the generalized global Torelli theorem for K3 and related varieties.

6. Methodological Tools and Kernel Construction

Central to all settings is the explicit construction of the Fourier–Mukai kernel $P$ implementing the equivalence:

• For abelian varieties, $P$ is often a semi-homogeneous vector bundle or a twisted Poincaré bundle, classified up to isogeny and compatible slope data (Li, 24 Nov 2025).
• For K3 surfaces, B-fields and cohomological correspondence guarantee the existence of a universal twisted sheaf on the product, which acts as kernel (Srivastava et al., 2021).
• In the case of noncommutative tori, the kernel arises from equivariant descent of the Poincaré bundle along finite group covers and the assignment of the Brauer class (Okuda, 2023).

The action on cohomology—preserving the twisted Mukai pairing and appropriately shifted Hodge filtration—is both necessary and sufficient for the existence of a twisted Fourier–Mukai equivalence (Li, 24 Nov 2025).

7. Typical Examples and Illustrations

Examples highlight both flexibility and constraints:

• A very general special cubic fourfold of discriminant $50$ admits exactly $2$ non-isomorphic twisted Fourier–Mukai partners $(X, \alpha)$ with $\alpha$ of order $5$ and Néron–Severi generator of self-intersection $4$, as dictated by the explicit formula (Pertusi, 2016).
• For a complex torus $A$ with a Brauer class of order $n$, the only twisted partners are itself and its dual with the same character, except in certain degeneracies (Li, 24 Nov 2025, Okuda, 2023).
• For an ordinary untwisted K3 in positive characteristic, all tame twisted partners arise as moduli spaces of twisted sheaves (Srivastava et al., 2021).

These results collectively illustrate the power and reach of the twisted Fourier–Mukai framework as a unifying and classifying tool in algebraic and derived geometry.