🔢 mathematics

Dirac operators on periodic planar graphs: the right definition advocated by flat bands

This paper investigates massive discrete Dirac-type operators on periodic planar graphs, demonstrating that neglecting faces yields spurious flat bands and establishing a spectral reduction that links the operator's spectrum to vertex and face Laplacians to correctly characterize flat bands on Archimedean and Lieb lattices.

Original authors: Koichi Kato, Daniel Parra, Serge Richard

Published 2026-10-09

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Koichi Kato, Daniel Parra, Serge Richard

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of quantum physics, the behavior of electrons is often mapped onto geometric shapes called lattices. Imagine a vast, repeating grid where atoms sit at the intersections. When physicists study how energy moves through these grids, they look for something called a "band structure," which describes the possible energy levels an electron can hold. Usually, these energy levels change smoothly as the electron moves, creating a flowing spectrum. However, sometimes a lattice produces a "flat band," a rare condition where the energy level remains exactly the same regardless of where the electron is or how it moves. This lack of movement, or dispersion, is fascinating because it can trap electrons in place, leading to exotic states of matter like superconductivity or magnetism. For decades, scientists have understood these flat bands when looking only at the points where atoms sit, known as vertices. But real materials are more complex; they have connections between atoms, and those connections enclose empty spaces, or faces. The question remained: does including these edges and faces change the story of flat bands, or does the simple view of the vertices tell the whole tale?

A team of researchers set out to answer this by building a more complete mathematical model of these grids. They focused on a specific type of operator, a mathematical tool used to describe how particles move, which they constructed to account for the vertices, the edges connecting them, and the faces enclosed by those edges. In their work, they compared two different ways of defining this tool. One version, which they call the Gauss-Bonnet operator, ignores the faces and only looks at the points and the lines connecting them. The other, which they call the Dirac-Hodge operator, includes the faces as a fundamental part of the structure. By using a technique that breaks the infinite grid down into smaller, repeating units, they were able to calculate the energy bands for both versions across many different types of lattices.

Their investigation revealed a significant flaw in the simpler, face-ignoring approach. They found that the Gauss-Bonnet operator consistently produced a flat band at a specific energy level for every single graph they tested, regardless of the graph's actual shape or geometry. This result was a mathematical artifact, a false signal created simply by cutting off the model at the edges and ignoring the faces. It suggested that flat bands were everywhere, even in lattices where they did not physically exist. In contrast, the Dirac-Hodge operator, which respects the full geometry including the faces, did not produce this spurious result. Instead, it showed that a flat band only appears if the underlying geometry of the vertices or the faces naturally supports one. This distinction allowed the researchers to identify the correct definition for studying these systems: the one that includes the faces is the only one that reveals the true, geometry-dependent physics.

With the correct tool in hand, the team applied it to a famous family of repeating patterns known as Archimedean tilings. These are grids made of regular polygons, such as squares, triangles, and hexagons, arranged in specific ways. Previous studies had shown that only two of these patterns, the Kagome lattice and the Super-Kagome lattice, possessed flat bands when looking only at the vertices. The researchers wanted to see if adding the faces would reveal new flat bands in the other patterns. They calculated the energy spectra for all eleven Archimedean tilings. The results were clear: the faces did not create any new flat bands. The only patterns that supported flat bands were still the Kagome and Super-Kagome lattices. In these two cases, the faces actually contributed their own flat bands, doubling the number of flat energy levels available, but they did not introduce flat bands into any of the other nine patterns.

To be certain their findings were not limited to this specific family of patterns, the researchers also examined a different grid called the Lieb lattice. This lattice is known to have a flat band when looking at the vertices. Their analysis confirmed that the Dirac-Hodge operator inherited this flat band from the vertices. However, unlike the Kagome case, the faces of the Lieb lattice did not contribute any new flat bands; their energy levels varied smoothly. This contrast proved that the behavior of the faces is not a universal echo of the vertices. Sometimes the faces add new flat bands, and sometimes they do not, depending entirely on the specific geometry of the grid.

The paper concludes by showing how to find the actual, localized states of these flat bands. They demonstrated that for every flat band found in the vertices or the faces, there is a corresponding pair of stable states in the full system. These states are not spread out across the entire grid but are confined to small, finite regions, such as a single hexagon or a cluster of triangles. By constructing these states explicitly, the researchers provided a concrete picture of what these flat bands look like in practice. Their work establishes that to correctly understand the spectral properties of periodic planar graphs, one must use the operator that includes the faces. Without this inclusion, the results are contaminated by artificial flat bands that have nothing to do with the physical reality of the material. The study confirms that while the geometry of the faces can sometimes create new flat bands, it never creates them where the underlying structure does not already support them.

Technical Summary: Dirac operators on periodic planar graphs: the right definition advocated by flat bands

Problem Statement
The paper investigates the spectral theory of massive discrete Dirac-type operators on periodic planar graphs. While the spectral theory of the graph Laplacian (a second-order operator acting solely on vertices) is well-established, particularly regarding band structures and flat bands in periodic settings, the analysis of first-order operators acting on higher-degree cochains (edges and faces) remains less developed. A central challenge addressed is the ambiguity in defining discrete Dirac operators on arbitrary graphs. Unlike the Laplacian, which has a canonical definition via the coboundary map $d_0$ , first-order operators can be constructed in multiple ways depending on whether the two-dimensional information carried by faces is retained. The authors aim to determine the correct definition of a discrete Dirac operator that avoids spurious spectral features and to analyze the resulting flat band phenomena on Archimedean and other periodic lattices.

Methodology
The authors employ a framework based on discrete differential calculus and combinatorial Hodge theory. The core methodology involves:

Operator Construction: Defining two massive first-order operators on planar graphs with bounded geometry:
- The Gauss–Bonnet operator ( $G_m$ ), acting on vertices and edges (truncating the cochain complex at degree 1).
- The Dirac–Hodge operator ( $D_m$ ), acting on the full chain of vertices, edges, and faces.
  Both are constructed using the coboundary maps $d_0$ (vertices to edges) and $d_1$ (edges to faces) and their adjoints, combined with a mass term $m$ .
Floquet–Bloch Decomposition: For periodic graphs, the authors utilize the Floquet–Bloch theorem to decompose the operators into fiber-wise matrix-valued functions $D_m(\xi)$ and $G_m(\xi)$ over the Brillouin torus $\mathbb{T}^2$ .
Spectral Reduction: By exploiting the block structure of these operators, the authors derive a spectral reduction that expresses the spectrum of the first-order operators in terms of the spectra of scalar Laplacians on vertices ( $-\Delta_V$ ) and faces ( $-\Delta_F$ ).
Case Studies: The theoretical results are applied to specific periodic tilings, including the 11 Archimedean lattices and the Lieb lattice. The analysis involves computing the characteristic polynomials of the face Laplacians $-\Delta_F(\xi)$ to determine the existence of flat bands (eigenvalues independent of the quasi-momentum $\xi$ ).

Key Contributions and Results

Spectral Reduction Theorem: The primary theoretical result establishes that the spectrum of the Dirac–Hodge operator $D_m$ is entirely determined by the spectra of the vertex Laplacian $-\Delta_V$ and the face Laplacian $-\Delta_F$ . Specifically, every band of $D_m$ takes the form $\pm\sqrt{\lambda + m^2}$ , where $\lambda$ belongs to the spectrum of either $-\Delta_V$ or $-\Delta_F$ . Consequently, $D_m$ possesses a flat band if and only if at least one of these scalar Laplacians possesses a flat band.
Identification of Spurious Flat Bands: The analysis reveals that the Gauss–Bonnet operator $G_m$ (which ignores faces) exhibits a "spurious" flat band at energy $-m$ with multiplicity equal to the number of faces. This band is an artifact of truncating the cochain complex and is unrelated to the graph's geometry. In contrast, the Dirac–Hodge operator $D_m$ is free of this spurious feature. The authors argue that $D_m$ is the appropriate definition for studying geometry-dependent flat bands.
Flat Bands on Archimedean Lattices: Applying the reduction to the 11 Archimedean tilings, the authors find that flat bands for $D_m$ $D_{m}$ arise only in the Kagome and Super-Kagome lattices.
- For the Kagome lattice $(3.6)^2$ , flat bands appear at $\pm\sqrt{6+m^2}$ (from $-\Delta_V$ ) and $\pm\sqrt{3+m^2}$ (from $-\Delta_F$ ).
- For the Super-Kagome lattice $(3.12^2)$ , flat bands appear at $\pm\sqrt{5+m^2}$ (from $-\Delta_V$ ) and $\pm\sqrt{3+m^2}$ (a coincidence where both $-\Delta_V$ and $-\Delta_F$ contribute).
- For the remaining nine Archimedean lattices, neither $-\Delta_V$ nor $-\Delta_F$ supports flat bands; thus, $D_m$ has no flat bands.
The Lieb Lattice Contrast: The authors examine the Lieb lattice, which is known to have a flat band for the vertex Laplacian $-\Delta_V$ . Their analysis shows that for this lattice, the face Laplacian $-\Delta_F$ is purely dispersive (no flat bands). Consequently, $D_m$ on the Lieb lattice inherits exactly two flat bands from the vertex sector, with no new contributions from the faces. This demonstrates that face contributions do not automatically mirror vertex contributions.
Duality: The paper notes a duality between the face Laplacian of a planar graph and the vertex Laplacian of its dual graph, allowing for simplified analysis of certain lattices (e.g., the square, triangular, and hexagonal lattices).

Significance and Claims
The paper claims that the Dirac–Hodge operator $D_m$ is the correct discrete analogue of the continuous Dirac operator for periodic planar graphs because it retains the full cochain structure (vertices, edges, and faces). The authors argue that neglecting faces, as done in the Gauss–Bonnet operator, leads to unphysical, geometry-independent flat bands. By establishing the spectral reduction, the authors reduce the complex problem of analyzing a first-order operator on three degrees of freedom to the more tractable problem of analyzing two independent scalar Laplacians.

The study of flat bands on Archimedean lattices reveals that while higher-order information (faces) introduces new spectral data, it does not broaden the class of Archimedean tilings supporting flat bands beyond those already identified for the vertex Laplacian. The work provides a rigorous mathematical foundation for distinguishing between genuine geometric flat bands and artifacts of operator definition in discrete quantum systems.

More like this