Higher Berry Invariants in Quantum Systems

• Higher Berry Invariants are topological invariants that generalize the ordinary Berry curvature by encoding geometric information through higher-order forms in gapped quantum systems.
• They extend Berry connections to include higher forms via gerbes and tensor-network formulations, providing a framework to study localized Berry curvature transport.
• Quantized through wave-function and transfer-matrix methods, these invariants reveal robust bulk–boundary correspondences and facilitate the detection of topological phase transitions.

Searching arXiv for papers on higher Berry invariants and related formulations to ground the article in current literature. Higher Berry invariants are topological invariants associated with families of gapped many-body quantum systems whose parameter dependence carries geometric information beyond the ordinary Berry curvature. In the now standard many-body formulation, a $d$-dimensional parametrized gapped system is assigned a closed $(d+2)$-form on parameter space, called the higher Berry curvature, whose cohomology class generalizes the role of the ordinary Berry 2-form. In one spatial dimension this higher object is a 3-form, physically interpreted as a flow of ordinary Berry curvature through the system; mathematically it is naturally associated with gerbes rather than line bundles (Kapustin et al., 2020, Wen et al., 2021, Artymowicz et al., 2023).

1. Conceptual definition and scope

The ordinary Berry connection and curvature for a smooth family of normalized states $\ket{\psi}$ over parameter space $X$ are

$$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$

For extended many-body systems, however, the total Berry curvature is generally extensive and therefore is not the appropriate invariant. The higher-Berry program replaces this extensive quantity by a localized transport object. In one dimension, Kapustin–Spodyneiko and subsequent work formulate a closed 3-form $\Omega^{(3)}$ that measures the spatial flow of ordinary Berry curvature, and in general spatial dimension $d$ they construct a closed $(d+2)$-form $\Omega^{(d+2)}$ on parameter space (Kapustin et al., 2020, Wen et al., 2021).

In the one-dimensional lattice formulation, a decomposition $d=\sum_p d_p$ of parameter variation into local pieces allows one to define local Berry-curvature densities and higher descendants. With a cut at $(d+2)$0, the higher Berry curvature is the net 3-form flow from sites left of the cut to sites right of the cut,

$(d+2)$1

This is the sense in which higher Berry curvature is a direct generalization of the usual Berry curvature: the transported quantity is no longer charge but Berry curvature itself (Sommer et al., 2024).

The framework applies most cleanly to infinite-volume gapped systems, short-range entangled states, and invertible phases. Several distinct but related notions occur in the literature under similar names. The many-body higher Berry invariant of parametrized gapped systems should be distinguished from nested Wilson-loop constructions for higher-order topology, and also from symmetry-quantized cluster Berry phases such as $(d+2)$2 Berry phases for HOSPTs, which are different invariants with different geometric input (Lo et al., 24 Feb 2026, Araki et al., 2019).

2. Gerbes, integral classes, and quantization

The mathematical structure underlying ordinary Berry curvature is a line bundle with connection; the corresponding higher structure in one-dimensional many-body systems is a $(d+2)$3-gerbe with connection. For smooth families of invertible 1D states, the higher Berry curvature

$(d+2)$4

defines a de Rham class that admits an integral refinement: $(d+2)$5 Equivalently, the 3-form is the curvature of a gerbe with connection, just as ordinary Berry curvature is the curvature of a line bundle (Artymowicz et al., 2023).

This integral refinement is constructive. On a good cover $(d+2)$6, one obtains Deligne–Beilinson data

$(d+2)$7

with

$(d+2)$8

and $(d+2)$9, depending on conventions. The higher invariant is then the integral period of the gerbe curvature over a closed 3-manifold in parameter space (Artymowicz et al., 2023).

For translationally invariant uniform MPS, the gerbe structure can be written explicitly in tensor-network language. On overlaps of parameter patches, the injective parts of the MPS are related by unitary gauge transformations $\ket{\psi}$0, and one can identify a Čech cocycle

$\ket{\psi}$1

together with a 1-connection

$\ket{\psi}$2

a 2-connection $\ket{\psi}$3, and a 3-curvature $\ket{\psi}$4. The resulting invariant is

$\ket{\psi}$5

This gives a concrete gerbe realization of the higher Berry invariant in the MPS setting (Sommer et al., 2024).

A torsion version also exists. For $\ket{\psi}$6-dimensional bosonic invertible-state families parametrized by spaces such as $\ket{\psi}$7 and $\ket{\psi}$8, the relevant invariant lives in the torsion part of $\ket{\psi}$9 and is formulated using smooth Deligne cohomology and higher holonomy on torsion 2-cycles. In these models the invariant detects $X$0 and $X$1 higher pumps that are invisible to de Rham curvature alone (Ohyama et al., 2023).

3. Wave-function and tensor-network formulations

A major development was the direct wave-function formula for infinite 1D systems. If the state has Schmidt decomposition across a cut,

$X$2

then the higher Berry curvature is

$X$3

This formula exhibits the higher Berry curvature as the parameter-space variation of Schmidt weights coupled to the ordinary Berry-curvature-type 2-form of the left Schmidt states. It is gauge invariant under overall Schmidt-state phases and under unitary rotations within degenerate Schmidt sectors; in degenerate sectors it reduces to $X$4 times the trace of the non-Abelian Berry curvature (Sommer et al., 2024).

The same paper derives transfer-matrix formulas for left-canonical MPS. In the left tangent-space gauge, the higher Berry curvature across a cut simplifies to

$X$5

and for a translationally invariant uMPS to

$X$6

These formulas were used both in an exactly solvable dimerized spin-$X$7 chain and in iDMRG calculations for perturbed models; for $X$8 the integrated invariant remained quantized to within about $X$9 (Sommer et al., 2024).

A complementary discrete tensor-network approach defines a triangle phase

$$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$0

from mixed transfer-matrix eigenvectors $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$1, and then defines a tetrahedral curvature

$$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$2

Summing over a closed triangulated 3-manifold yields an integer invariant

$$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$3

This discrete formulation is explicitly gauge invariant and was implemented with iDMRG for nontrivial $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$4-parametrized MPS families (Shiozaki et al., 2023).

The wave-function construction has also been extended beyond one dimension for locally parameterized short-range entangled states. Introducing the simplex-exterior derivative $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$5, one defines

$$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$6

where $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$7 is a closed non-exact $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$8-cochain on the lattice. In $$\mathcal A=-\Im\braket{\psi|d\psi}, \qquad \Omega^{(2)}=-\Im\braket{d\psi|d\psi}=d\mathcal A.$$9, exactly solvable PEPS-like examples realize

$\Omega^{(3)}$0

providing explicit higher Berry classes in two spatial dimensions (Sommer et al., 2024).

A connection-level tensor-network formulation has also been given. For parameterized MPS families, one introduces a 1-form overlap connection $\Omega^{(3)}$1, a patchwise 2-form $\Omega^{(3)}$2, and curvature

$\Omega^{(3)}$3

thereby making the gerbe connective structure explicit at the MPS level. In the examples treated, $\Omega^{(3)}$4 yields the expected integer invariant (Ohyama et al., 2024).

4. Physical interpretation, transport, and bulk–boundary correspondences

The basic physical interpretation is as a pump of Berry curvature. In one dimension, a nontrivial $\Omega^{(3)}$5 means that ordinary 2-form Berry curvature is transported through the system during a multi-parameter adiabatic cycle. Bulk–boundary correspondence then relates the bulk higher invariant to lower-dimensional boundary Berry data. For semi-infinite systems, the boundary 2-form $\Omega^{(3)}$6 satisfies asymptotically

$\Omega^{(3)}$7

so that a nonzero bulk higher Berry invariant forces anomalous boundary behavior such as isolated Weyl points in parameter space (Wen et al., 2021).

In the 2D $\Omega^{(3)}$8-equivariant setting, the analogous object is the higher Thouless pump, a closed 2-form $\Omega^{(3)}$9 that admits an integral refinement in invertible phases: $d$0 It can be identified with the excess Berry curvature generated by flux insertion, through

$d$1

This expresses the higher Thouless pump as Berry-curvature transport generated by a flux defect (Artymowicz et al., 2023).

A boundary-scattering detection principle has been established for 1D gapped free-fermion families. Coupling a gapless lead to the boundary gives a reflection matrix $d$2, and the higher Berry invariant is detected by the 3D winding number

$d$3

In the explicit $d$4-parametrized models treated there, $d$5, and the invariant remains robust under moderate disorder (Lo et al., 24 Feb 2026).

A further development recasts the 4D Chern insulator as a three-parameter family of translationally invariant infinite 1D chains. Computing the higher three-form Berry curvature with iMPS yields the Dixmier–Douady–Kapustin–Spodyneiko number

$d$6

whose phase diagram as a function of mass is exactly congruent to the known second Chern number phase diagram of the 4D model. This shows that higher Berry curvature can be used to compute second Chern numbers in a manifestly quantized manner (Heinsdorf et al., 24 Jun 2026).

Boundary conformal field theory provides another transport interpretation. For parametrized BCFT boundary conditions induced from gapped systems in a nontrivial higher Berry class, one defines a Fock-space 2-form $d$7 from the Berry curvatures of occupied modes and then a higher curvature

$d$8

In the explicit $d$9-family of boundary conditions studied, the result is

$(d+2)$0

interpreted as a Chern-number pump in BCFT Fock space (Wen, 16 Jul 2025).

5. Variants, related constructions, and conceptual distinctions

One important adjacent line of work concerns “information hierarchy” rather than higher quantized invariants. Starting from the Resta formula

$(d+2)$1

the many-body Berry phase can be expanded as

$(d+2)$2

where $(d+2)$3 are cumulants of $(d+2)$4. The result is a no-go statement: for generic interacting systems, no finite set of local correlators suffices to determine the many-body Berry holonomy in the thermodynamic limit. This does not define a tower of separately quantized higher Berry invariants; rather, it identifies an infinite hierarchy of geometric information underlying the ordinary Berry phase (Watanabe, 1 Jun 2026).

Another neighboring construction uses tensor Berry connections in Bloch-band settings. There a 2-form tensor Berry connection

$(d+2)$5

has 3-form curvature $(d+2)$6, and the corresponding topological invariant is a Dixmier–Douady invariant

$(d+2)$7

This framework reinterprets, for example, the 3D class-AIII winding number as a DD invariant and identifies 4D tensor monopoles as sources of $(d+2)$8-flux. It is closely related geometrically, but it is formulated in momentum-space band theory rather than as a many-body higher Berry curvature for parametrized gapped lattice families (Palumbo et al., 2018).

A field-theoretic generalization for free Dirac fermions with spacetime-dependent masses yields a higher Berry phase identified with the partition function of a parameterized invertible field theory. Its local curvature is a degree-$(d+2)$9 form

$\Omega^{(d+2)}$0

derived using the APS index theorem for superconnections. When this curvature vanishes, a torsional Berry phase can still remain, detected by $\Omega^{(d+2)}$1-invariants on defect-localized modes (Choi et al., 2022).

BCFT has also acquired an intrinsic higher Berry geometry. On a boundary conformal manifold, the phase $\Omega^{(d+2)}$2 of the 3-point function of lightest boundary-condition-changing operators defines a local 2-form

$\Omega^{(d+2)}$3

with curvature $\Omega^{(d+2)}$4. When the boundary conformal manifold is a D-brane moduli space, $\Omega^{(d+2)}$5 coincides with the NS–NS $\Omega^{(d+2)}$6-field, and in WZW examples $\Omega^{(d+2)}$7 is the usual Wess–Zumino 3-form (Choi et al., 16 Jul 2025).

6. Limitations, non-canonicity, and open directions

Several limitations are explicit in the literature. First, the direct wave-function formula in terms of Schmidt data is developed explicitly only for $\Omega^{(d+2)}$8, and the strongest quantization statements are currently established for invertible phases and, concretely, for translationally invariant or essentially injective MPS families (Sommer et al., 2024, Artymowicz et al., 2023).

Second, the local distribution of higher Berry curvature is not canonical. The decomposition $\Omega^{(d+2)}$9 is non-unique, so local higher-curvature densities can change under different choices even when the integrated invariant is unchanged. This explains why different Hamiltonian-based, wave-function-based, and tensor-network-based constructions can differ locally while agreeing on the global topological class (Sommer et al., 2024, Sommer et al., 2024).

Third, finite systems without explicit edge bookkeeping do not generally support nontrivial higher Berry invariants of the same kind. In the one-dimensional wave-function formulation,

$d=\sum_p d_p$0

for finite systems, so its integral over a closed parameter space vanishes unless edges are treated explicitly. The genuinely new invariant is therefore an infinite-system or boundary-sensitive phenomenon (Sommer et al., 2024).

Fourth, not every use of “higher Berry” refers to a quantized invariant. The cumulant hierarchy of the many-body Berry phase is a cautionary case: it shows that higher-order geometric information may be organized as an infinite hierarchy of nonlocal descriptors without yielding a finite ladder of robust quantized invariants (Watanabe, 1 Jun 2026).

Open problems recur across the literature. These include extending gerbe and higher-connection structures beyond translationally invariant tensor networks, clarifying Hall-response and magnetoelectric interpretations, understanding experimental detection in interacting settings, generalizing to higher-dimensional tensor networks, and developing a full global gerbe formulation for BCFT boundary conformal manifolds (Sommer et al., 2024, Sommer et al., 2024, Lo et al., 24 Feb 2026, Choi et al., 16 Jul 2025). A plausible implication is that “higher Berry invariant” will remain an umbrella term covering several layers of structure: integral $d=\sum_p d_p$1-form classes for invertible many-body families, torsion Deligne-holonomy invariants, and field-theoretic or conformal realizations of gerbe-valued geometric transport.