---
title: Higher Berry Invariant
url: https://www.emergentmind.com/topics/higher-berry-invariant
type: topic
---

# Higher Berry Invariant

The higher Berry invariant is a higher-degree topological quantity attached not to a single gapped Hamiltonian but to a **family** of gapped quantum systems over a parameter space. In the Kapustin–Spodyneiko framework, a \(d\)-dimensional parametrized gapped system carries a closed \((d+2)\)-form \(\Omega^{(d+2)}\); its cohomology class is the higher Berry or KS invariant, while integration over a closed \((d+2)\)-cycle gives a quantized higher Berry number. In one spatial dimension this degree-three structure governs phenomena such as Chern-number pumping, admits gerbe-theoretic and tensor-network realizations, and in free-fermion settings can be detected from boundary scattering data [2001.03454] [2112.07748] [2602.21301].

## 1. Terminology and conceptual scope

A parametrized gapped system is a continuous family of finite-range gapped lattice Hamiltonians
\[
H:X\to GH,
\]
with fixed spatial dimension, local Hilbert-space type, symmetry class, and parameter space \(X\). When \(X=\mathrm{pt}\), one recovers an ordinary gapped phase; when \(X\neq \mathrm{pt}\), the family itself can carry global topology even if each individual Hamiltonian is topologically trivial as a phase over a point [2112.07748].

The literature distinguishes several related objects. In one standard usage, **higher Berry curvature** is a local closed differential form on parameter space, **higher Berry invariant** or **KS invariant** is the corresponding cohomology class, and **higher Berry number** or **KS number** is the quantized integral of that form over an appropriate closed cycle [2112.07748]. A closely related formulation defines the higher Berry invariant as the de Rham cohomology class of a closed \((D+2)\)-form \(\Omega^{(D+2)}\) for a \(D\)-dimensional gapped family [2001.03454]. Other papers, especially in free-fermion boundary or tensor-network settings, sometimes use “higher Berry invariant” for the quantized closed-cycle value itself; this terminological variation is intrinsic to the current literature [2602.21301].

The conceptual motivation is that ordinary phase classification detects connected components of the space of gapped Hamiltonians, whereas parametrized families probe higher homotopy and higher cohomology. In one-dimensional families, a nonzero higher Berry invariant signals a gerbe structure of the ground-state family, generalizing the line-bundle structure underlying ordinary Berry phase. In the free-fermion boundary-scattering formulation, a nontrivial invariant also obstructs a globally continuous MPS tensor over parameter space and diagnoses a quantized Chern-number pump rather than an ordinary charge pump [2602.21301] [2405.05327].

## 2. Mathematical formulations

For finite-dimensional quantum mechanics, the ordinary Berry curvature may be written in resolvent form as
\[
\Omega^{(2)}=\frac{i}{2}\oint \frac{dz}{2\pi i}\,\mathrm{Tr}\bigl(G\, dH\, G^2\, dH\bigr),
\qquad
G(z)=\frac{1}{z-H},
\]
and satisfies \(d\Omega^{(2)}=0\). In extended systems, the total trace diverges, so the construction is replaced by a local/descent formalism built from local Hamiltonian terms \(H=\sum_{p\in\Lambda}H_p\) and chains \(F^{(n)}\) obeying
\[
dF^{(n)}=\partial F^{(n+1)}.
\]
For a \(D\)-dimensional lattice family, the higher Berry form is
\[
\Omega^{(D+2)}(f_1,\dots,f_D)
=
\bigl(F^{(D+2)},\delta f_1\cup\cdots\cup\delta f_D\bigr),
\]
where the \(f_\mu\) are step-like functions selecting coarse spatial directions. This form is closed, and changing the \(f_\mu\) only changes it by an exact form, so the cohomology class \([\Omega^{(D+2)}]\in H^{D+2}(M,\mathbb R)\) is well defined [2001.03454].

In one dimension, the KS higher Berry curvature is a closed \(3\)-form
\[
\Omega^{(3)}(f)=\frac{1}{2}\sum_{p,q\in\mathbb Z}F_{pq}^{(3)}\,(f(q)-f(p)),
\]
with \(f\) interpolating from \(0\) at the far left to \(1\) at the far right. For a step function \(f(p)=\Theta(p-a)\), this becomes
\[
\Omega^{(3)}(f)=\sum_{p<a,q>a}F_{pq}^{(3)}.
\]
Its cohomology class is phase-invariant, and for invertible systems Kapustin and Spodyneiko argued that spherical periods are quantized:
\[
\int_{S^3}\phi^*\Omega^{(3)}(f)\in 2\pi \mathbb Z.
\]
If \(X\) is an oriented closed \(3\)-manifold, the integrated KS number is
\[
KS=\int_X\Omega^{(3)}(f)\in 2\pi\mathbb Z.
\]
More generally, for a \(d\)-dimensional system over a \((d+2)\)-manifold, the higher Berry number is \(KS=\int_X\Omega^{(d+2)}\) [2112.07748].

For translationally invariant \(1d\) free fermions with three parameters \(\lambda_1,\lambda_2,\lambda_3\), the same cohomology class is represented by the Brillouin-zone integral of the degree-four Chern character of the occupied-band Berry curvature:
\[
\Omega^{(3)}/2\pi
\sim
\int_{S^1_{\mathrm{BZ}}}\mathrm{Ch}_2(F),
\qquad
\mathrm{Ch}_2(F)=-\frac{1}{8\pi^2}\mathrm{Tr}(F\wedge F).
\]
This identification ties the higher Berry invariant to characteristic classes familiar from higher-dimensional band topology [2001.03454].

## 3. Physical meaning: flow, pumping, and bulk–boundary correspondence

The main physical interpretation developed in the KS literature is that higher Berry curvature measures a **flow** of lower-degree Berry data. In one spatial dimension, the higher Berry curvature is a flow of ordinary Berry curvature to or from the boundary:
\[
\Omega_{\infty/2}^{(3)}(f)=d\omega^{(2)},
\]
where \(\omega^{(2)}\) is a boundary \(2\)-form. In \(d\) dimensions, \(\Omega^{(d+2)}\) is a flow of \((d-1)\)-dimensional higher Berry curvature. A nonzero bulk KS number therefore forces anomalous boundary behavior—isolated Weyl-like singularities or more general boundary gap closings—and the boundary flux of the lower-dimensional invariant equals the bulk KS number [2112.07748].

This flow interpretation leads to a pumping picture. In the solvable \(1d\) model over \(S^3\), the integrated higher Berry curvature is
\[
KS=\int_{S^3}\Omega^{(3)}(f)=2\pi,
\]
and a boundary supports a single isolated Weyl point in parameter space. In the related \(1d\) family over \(S^2\times S^1\), the same number is expressed as a jump in boundary Chern number,
\[
KS
=
\lim_{\epsilon\to 0}\Big[C(-t_0+\epsilon)-C(-t_0-\epsilon)\Big],
\]
so the higher invariant is a **Chern-number pump**: during one cycle, a quantized Chern number \(2\pi\) is pumped between bulk and boundary [2112.07748].

A complementary free-fermion formulation detects the same invariant from boundary scattering. For a semi-infinite gapless lead attached to a semi-infinite gapped \(1d\) system, the boundary reflection matrix \(R(\lambda)\in U(N)\) defines a map \(R:X\to U(N)\). For a closed \(3\)-manifold \(X\), the scattering invariant is
\[
\nu_3(R)=\frac{1}{24\pi^2}\int_X \mathrm{Tr}(R^\dagger dR)^{\wedge 3}\in\mathbb Z.
\]
In the continuum \(S^3\) Dirac model and in lattice realizations, this higher winding number equals the bulk higher Berry invariant; in the explicit examples,
\[
\nu_3(R)=-1.
\]
The approach is robust against disorder as long as the family remains gapped and, in the examples studied, diagnoses a quantized Chern-number pump without direct access to the bulk many-body wavefunction [2602.21301].

## 4. Tensor-network, wavefunction, and iMPS realizations

In one-dimensional tensor-network language, the higher Berry invariant is encoded by a gerbe rather than a line bundle. For parameterized MPS, the gerbe cocycle is extracted from the triple overlap
\[
c_{\alpha\beta\gamma}^{(0)}
=
\mathrm{tr}\!\left(\Lambda_{\alpha\gamma}^{L}\Lambda_{\alpha\beta}^{R}\Lambda_{\beta\gamma}^{R}\right),
\]
and the higher Berry connection consists of \(1\)-forms \(w_{\alpha\beta}^{(1)}\) on double overlaps and \(2\)-forms \(B_\alpha^{(2)}\) on patches, satisfying
\[
(\delta w^{(1)})_{\alpha\beta\gamma}=d\log c_{\alpha\beta\gamma}^{(0)},
\qquad
(\delta B^{(2)})_{\alpha\beta}=dw_{\alpha\beta}^{(1)}.
\]
The global curvature is
\[
H^{(3)}=dB_\alpha^{(2)},
\]
and the integral invariant is
\[
\int_{M_3}\frac{H^{(3)}}{2\pi i}\in\mathbb Z.
\]
In this framework, constant-rank MPS cannot realize a nontrivial free integral higher Berry invariant; nonconstant-rank or essentially normal MPS are required for the nontrivial integral class [2405.05327].

A discrete numerical formulation replaces differential forms by simplicial data on a triangulated parameter space. For neighboring MPS, the dominant eigenvector \(V_{p_0p_1}\) of the mixed transfer matrix plays the role of discrete transition data. On a triangle \((p_0p_1p_2)\), one defines
\[
\phi(p_0p_1p_2)
=
{\rm Arg}\,
[\Lambda_{p_0}^{2/3}V_{p_0p_1}\Lambda_{p_1}^{2/3}V_{p_1p_2}\Lambda_{p_2}^{2/3}V_{p_2p_0}],
\]
and on a tetrahedron \(\Delta^3=(p_0p_1p_2p_3)\),
\[
F(\Delta^3)
=
\phi(p_1p_2p_3)-\phi(p_0p_2p_3)+\phi(p_0p_1p_3)-\phi(p_0p_1p_2).
\]
The integrated invariant is
\[
\nu(|\Sigma_3|)=\frac{1}{2\pi}\sum_{\Delta^3\in |\Sigma_3|}\sigma(\Delta^3)\,F(\Delta^3)\in\mathbb Z.
\]
For the \(S^3\) example studied numerically, this invariant is \(1\) [2305.08109].

Beyond one dimension, wavefunction-based constructions define a closed \((d+2)\)-form
\[
\Omega^{(d+2)}=\langle F^{(d+2)},b\rangle
\]
for locally parameterized short-range-entangled states, and in \(d=2\) an exactly solvable family over \(S^4\) satisfies
\[
\int_{S^4}\Omega^{(4)}=2\pi
\]
[2405.05323]. In PEPS language, the \(2d\) higher Berry phase is extracted from a quadruple inner product on fourfold overlaps, giving a Čech \(3\)-cocycle
\[
[c_{\alpha\beta\gamma\delta}]
\in
\check H^3(X;\mathrm{sh}(\mathbb C^\times))
\simeq
H^4(X;\mathbb Z),
\]
with an explicit nontrivial family on \(\mathbb{R}P^4\) realizing the nontrivial \(\mathbb Z_2\) class in \(H^4(\mathbb{R}P^4;\mathbb Z)\) [2405.05325].

A recent iMPS application makes the relation to familiar band-topological invariants explicit. Rewriting the \(4d\) lattice Chern insulator as a family of translationally invariant infinite chains over \(T^3\), one computes a DDKS number
\[
\nu=\frac{1}{2\pi}\sum_{\Delta^3\in |T^3|}F(\Delta^3).
\]
The resulting phase diagram as a function of the Dirac mass is exactly congruent to the known phase diagram of the second Chern number \(\mathrm{ch}_2\), and with fixed orientation the paper states
\[
\mathrm{ch}_2=\nu.
\]
This shows that higher Berry curvature can compute second Chern numbers in a manifestly quantized way [2606.26096].

## 5. Boundary conformal field theory and field-theoretic generalizations

A BCFT formulation introduces higher Berry geometry on a **boundary conformal manifold**, the moduli space of conformal boundary conditions connected by exactly marginal boundary deformations. The basic data are no longer pairwise overlaps but **triple** overlaps of boundary conditions, encoded in the phase of the disk three-point function of lightest boundary-condition-changing operators,
\[
c(\alpha,\beta,\gamma)=|c(\alpha,\beta,\gamma)|e^{i\phi(\alpha,\beta,\gamma)}.
\]
From this phase one defines a local \(2\)-form higher Berry connection
\[
\mathcal{B}
=
-\frac{i}{2}
\left[
\frac{\partial^2 \phi(\alpha,\beta,\gamma)}{\partial \beta^i\partial\gamma^j}
-(i\leftrightarrow j)
\right]_{\beta=\gamma=\alpha}
d\alpha^i\wedge d\alpha^j,
\]
with curvature
\[
\mathcal H=d\mathcal B.
\]
The gauge law is that of a gerbe connection,
\[
\mathcal B\to \mathcal B+d\lambda.
\]
In compact-boson and WZW examples, \(\mathcal B\) coincides with the NS–NS \(B\)-field and \(\mathcal H\) with the Wess–Zumino \(3\)-form [2507.12525].

A related BCFT analysis makes the pumping interpretation fully explicit. For a two-flavor Dirac BCFT with \(S^3\)-parametrized boundary conditions, the occupied Fock sea carries a total ordinary Berry curvature
\[
\omega^{(2)}
=
\left(\frac{\alpha}{\pi}-1\right)\frac{\sin\theta}{2}\,d\theta\wedge d\phi,
\]
whose derivative gives the higher Berry curvature
\[
\Omega^{(3)}
=
d\omega^{(2)}
=
\frac{1}{2\pi}\sin\theta\,d\alpha\wedge d\theta\wedge d\phi.
\]
Its integral is quantized,
\[
\int_{S^3}\Omega^{(3)}=2\pi.
\]
Here the higher invariant measures a **Chern-number pump in Fock space**: multi-parameter spectral flow transports ordinary Berry curvature across the BCFT spectrum, and the same structure appears in entanglement Hamiltonians of parameterized gapped states [2507.12546].

In relativistic field theory, higher Berry curvature arises as the local response of parameterized invertible theories. For massive free Dirac fermions with spacetime-dependent mass parameters, the phase of the regulated partition function is expressed by an APS \(\eta\)-invariant, and the local higher Berry curvature is
\[
\mathcal{B}
=
\int_{\widetilde M_0=0}^{\widetilde M_0=\infty}
\left[\operatorname{ch}(\mathcal F)-\operatorname{ch}(\mathcal F^{(0)})\right]\wedge \hat A.
\]
This differential-form part need not exhaust the invariant: when \(\mathcal B=0\), the theory may still carry a **torsional Berry phase**, a bordism invariant detected by exponentiated \(\eta\)-invariants on nontrivial backgrounds [2205.02188].

## 6. Ambiguities, limitations, and neighboring notions

The phrase “higher Berry invariant” does not have a single fixed meaning across the literature. In the KS many-body framework it usually denotes the cohomology class \([\Omega^{(d+2)}/2\pi]\), with a separate “KS number” for the integrated period. In gerbe-based MPS and BCFT formulations, closely related local data may instead be presented as a \(2\)-form connection \(\mathcal B\), a \(3\)-form curvature \(H^{(3)}\) or \(\mathcal H\), or a Čech cocycle such as \(c_{\alpha\beta\gamma}^{(0)}\). Encyclopedia treatments therefore require explicit attention to which layer—local form, cohomology class, or integrated integer—is meant in a given source [2112.07748] [2405.05327] [2507.12525].

Quantization and robustness also depend on hypotheses. The higher-form construction exists for interacting lattice systems quite generally, but quantization is argued most cleanly for short-range-entangled or invertible families, especially on spherical cycles [2001.03454]. In the KS program, quantization on general closed oriented manifolds is expected but not proved in full generality [2112.07748]. In BCFT, orbifold singularities can make \(\frac{1}{2\pi}\int \mathcal H\) appear fractional unless one passes to the appropriate orbifold cohomology [2507.12525]. In boundary scattering, the explicit detection formula is established for \(1d\) gapped free-fermion families, while the interacting extension is stated as a conjecture rather than a theorem [2602.21301].

Several neighboring constructions are “higher” in a broader sense but are distinct from the KS higher Berry invariant of parametrized gapped families. The \(\mathbb Z_Q\) Berry phase used for higher-order SPT phases is a quantized local-twist invariant for irreducible \(Q\)-site clusters, not a degree-\((d+2)\) cohomological invariant of a family space [1906.00218]. The three-point Bargmann invariant
\[
P^{(3)}(x_1,x_2,x_3)=\operatorname{tr}[P(x_1)P(x_2)P(x_3)]
\]
captures extrinsic geometry of projective-state manifolds beyond ordinary Berry curvature and quantum metric, but it is a different geometric program [2205.15353]. Likewise, the Berry-curvature dipole
\[
D_{ab}=\int_{\mathbf{k}} f_0(\mathbf{k})\,\partial_a \Omega_b(\mathbf{k})
\]
and the quantum-metric-dipole terms governing second-order dc transport are gauge-invariant higher-order Berry-geometric response tensors, not quantized higher Berry invariants in the strict topological sense [1508.00571] [2606.22359].

Taken together, these works establish the higher Berry invariant as a family-level topological datum of gapped quantum matter. Its characteristic signatures are a closed higher-degree curvature form, a gerbe or higher-connection structure rather than an ordinary line bundle, quantized periods on suitable parameter cycles, and physical realizations as Chern-number pumps, anomalous boundary flow, scattering windings, and higher-dimensional characteristic classes [2001.03454] [2112.07748] [2602.21301]

Source: https://www.emergentmind.com/topics/higher-berry-invariant