Topological-Topological Flat Bands Reframe Band Touching as a Removable Singularity

How topological flat bands became singular

The baseline for this work is the physics of compact localized states (CLSs) in flat-band lattices. In Lieb and kagome models, CLSs are strictly local zero modes of the non-interacting Hamiltonian, and adjacent CLSs overlap such that a Stokes-like theorem holds: the sum of CLSs over a region equals the sum of boundary operators [1]. This first topological condition implies two non-contractible loop states in real space and an unavoidable band touching with dispersive bands in momentum space [1]. The touching point is singular: the Bloch state is ill-defined there, and so is any topological invariant for the entire flat band [1]. Earlier work classified such singular flat bands and studied their winding, Berry phase, and quantum distance [1].

This singularity is not a minor technicality. It precludes a well-defined projector and therefore blocks the definition of Chern numbers, Z2 invariants, and topological-crystalline invariants for the flat band itself [1]. The new paper's central move is to ask whether the singularity can be removed by imposing an additional condition on the loop states, rather than treating the touching as an unavoidable feature [1].

A second topological condition makes the touching removable

The proposed second topological condition states that loop states along different directions must be linearly dependent: Θy = λΘx with nonzero imaginary part of λ [1]. When this holds, the two directional limits of the Bloch state as k approaches the touching point represent the same physical state, so the flat-band projector P(k) becomes continuous at k = 0 [1]. The paper shows that this enforces a nonzero Chern number C = sign(ℑλ) = ±1 for the flat band [1]. The singularity is thus removable: the touching remains, but it no longer obstructs a well-defined topological invariant [1].

The authors call the resulting bands topological-topological flat bands, or top²-flat bands, where the first 'topological' refers to the topological loop states and the second to the nonzero invariant [1]. They construct explicit CLSs for a 2D Chern band and for 2D and 3D Z2 bands protected by time-reversal symmetry, with D4/D6 and cubic symmetry respectively [1]. In 3D, the second condition requires that three pairs of membrane states span the same linear subspace, leading to a 3D Z2 = 1 invariant [1].

How top² flat bands compare with earlier topological flat-band constructions

Earlier work on topological flat bands often focused on gapped flat bands with fragile topology or on nearly flat Chern bands in moiré systems. For example, rhombohedral graphene superlattices have been shown to host integer and fractional quantum anomalous Hall effects, with the moiré potential playing a crucial role in forming flat Chern bands [3]. That approach relies on moiré engineering to create a topological flat band, and the fractional Chern insulator state at ν = 2/3 survives phase transitions [3]. The top²-flat band construction differs in that it starts from exactly flat bands with CLSs and imposes a condition that makes the unavoidable touching removable, rather than relying on a gap or on moiré tuning [1].

A competing approach in cold-atom systems uses SU(N) non-Abelian gauge fields on a square lattice to generate large-Chern-number flat bands. In the SU(4) model, the lowest and highest bands carry Chern number C = −4, while the two middle bands touch at 16 Dirac points and have a combined Chern number C = 8 [4]. There, the band touching is between the two middle bands, and the topological invariant is defined for the combined bands or for the isolated bands, not for a single flat band with a removable singularity [4]. The top² construction instead enforces a condition that makes the projector continuous at the touching point, allowing a well-defined invariant for the flat band itself [1].

The new paper also connects to the broader program of constructing topological crystalline insulators from layer constructions. It shows that with the 2D Z2 top²-flat band as a building block, one can obtain top²-flat bands for all topological crystalline states in 218 of the 230 space groups, and for the remaining 12 groups via topological-crystal tilings [1]. This extends the reach of the construction beyond the elementary Chern and Z2 cases, but it remains a theoretical construction in tight-binding models [1].

Interactions, correlated phases, and what remains uncertain

The paper argues that under small, generic interactions, top²-flat bands flow to correlated topological insulators with a symmetric mass term [1]. For the Chern case, Hubbard attraction or repulsion of the proper sign dynamically generates a mass term that separates the two states at the touching point, and if the lower state is occupied, the band becomes a Chern insulator [1]. The authors also show that specially designed interacting models can have top²-flat bands as exact zero modes, using a parameter-counting argument for four-fermion interactions [1]. This suggests that the topological invariants are stable under interactions, but the analysis is perturbative and model-specific [1].

The evidence boundary is important. The construction is theoretical, demonstrated in 2D and 3D tight-binding models, and the correlated topological insulator phases are discussed only under specific interactions [1]. No material realization is proposed or tested [1]. In contrast, moiré systems such as twisted TMDs and rhombohedral graphene superlattices provide experimental platforms for flat-band topology, but their flat bands are not exactly flat and their topology arises from different mechanisms [3][5]. The top²-flat band proposal thus offers a conceptually distinct route to well-defined invariants in exactly flat bands, but its experimental relevance remains to be established [1].

Another open question is whether the second topological condition can be satisfied in realistic lattice models beyond the explicit examples given. The paper provides CLSs for square, hexagonal, and cubic lattices, but the general conditions for finding such CLSs in arbitrary lattices are not fully explored [1]. The interaction-driven phases are also discussed only for fully occupied bands, leaving partial filling and fractional states as future directions [1]. These limitations do not undermine the central result—that band touching can be reframed as a removable singularity—but they define the current scope of the claim [1].

About These Sources

This research page is built on 5 studies (4 peer-reviewed, 1 preprint) — published from 2023 to 2026, 4 from 2024 or later — selected as the most relevant from 11 studies that passed quality screening, drawn from 52 papers retrieved from a database of over 500 million.

Sources used in this answer