Fragile Topological Phases

Updated 23 June 2026

Fragile topological phases are crystalline states characterized by a Wannier obstruction that vanishes upon the addition of specific trivial bands.
Diagnostic techniques such as Wilson loop winding, Euler class evaluation, and symmetry-indicator mismatches precisely reveal fragile topology.
Prototypical models include honeycomb and kagome lattices, twisted bilayer graphene, and engineered metamaterials exhibiting robust defect-bound modes.

Fragile topological phases constitute a distinct class in the modern taxonomy of crystalline topological matter. Unlike stable (or "strong") topological phases, which are protected by additive $\mathbb{Z}$ or $\mathbb{Z}_2$ invariants robust under stacking with any number of trivial bands, fragile topological phases are characterized by Wannier obstructions that are resolved by the addition of certain trivial (“atomic”) degrees of freedom. Initially demonstrated in models of electronic band structures, the concept of fragile topology has since been generalized to phononic, photonic, mechanical, interacting bosonic/fermionic, and Floquet systems. This article provides a technical overview of the mathematical definition, diagnostic tools, material realizations, and emergent phenomena associated with fragile topological phases, emphasizing rigorous connections to band representation theory, homotopy and cohomology, Wilson loop invariants, and experimental signatures.

1. Fundamental Definition and Mathematical Structure

Fragile topology is defined via an intermediate notion of Wannier representability. A collection of isolated bands $\mathcal{B}$ is Wannier-representable (atomic/trivial) if and only if there exists a set of exponentially localized, symmetry-respecting Wannier functions spanning $\mathcal{B}$ (Po et al., 2017). Stable topological bands (e.g., those with nonvanishing Chern number or Fu–Kane–Mele $\mathbb{Z}_2$ invariant) exhibit a Wannier obstruction that cannot be removed, regardless of stacking with any trivial bands.

A set of bands $\mathcal{B}$ is fragile topological if:

$\mathcal{B}$ alone is non-Wannierizable: no symmetric, exponentially localized Wannier functions exist for $\mathcal{B}$ .
There exists a set of (possibly disconnected) trivial (atomic) bands $\mathcal{T}$ such that $\mathcal{B} \oplus \mathcal{T}$ is Wannier-representable.

Formally, in terms of elementary band representations (EBRs) (Song et al., 2019):

Trivial: $\mathbb{Z}_2$ 0, all $\mathbb{Z}_2$ 1
Stable: $\mathbb{Z}_2$ 2 not rationally expressible as sums and differences of EBRs.
Fragile: $\mathbb{Z}_2$ 3, $\mathbb{Z}_2$ 4, but not as a sum alone.

Topological quantum chemistry (TQC) provides a practical framework for classification, employing the symmetry-data vector of little-group irreducible representation multiplicities at high-symmetry points. Polyhedral and monoid structures (Hilbert bases, Smith decomposition) precisely enumerate all fragile phases within a given space group (Song et al., 2019).

Mathematically, fragile topology can also be viewed through equivariant 2-cohomotopy/homotopy theory, where band topology in the presence of crystalline symmetry is classified by connected components of equivariant mapping spaces $\mathbb{Z}_2$ 5, which are generally finer than $\mathbb{Z}_2$ 6-theory invariants and capture unstable "difference" classes (Sati et al., 31 Dec 2025).

2. Diagnostic Invariants: Wannier Obstruction, Wilson Loops, and Euler Class

The key physical diagnosis of fragility is a Wannier obstruction: an absence of symmetric, exponentially localized Wannier functions for a given set of bands. In practical settings, this is detected via:

Wilson loop winding: For rank-2 (or higher) band subspaces, one constructs the Wilson loop operator along closed paths in momentum space (e.g., $\mathbb{Z}_2$ 7). Nontrivial winding (gapless spectrum or quantized crossings) signals an obstruction (Bradlyn et al., 2018, Po et al., 2017). In $\mathbb{Z}_2$ 8-symmetric systems, the nontrivial winding is linked, via the Plücker embedding and Grassmannian geometry, to the Euler class $\mathbb{Z}_2$ 9 of the associated real rank-2 bundle ( $\mathcal{B}$ 0 for orientable bundles) (Bouhon et al., 2020, Zhao et al., 2022, Becker et al., 5 Feb 2025).
Symmetry-indicator mismatch: The pattern of irreducible representations at high-symmetry momenta in the Brillouin zone is compared to all possible atomic insulators (sums of EBRs) (Song et al., 2019, Mañes, 2019). Mismatches that can be resolved by subtracting (but not adding) EBRs signal fragility.
Real-space invariants: Charge (or local irrep) imbalances at Wyckoff positions, induced by the site-symmetry group, provide a set of real space invariants (RSIs) that classify fragile phases and predict gapless spectral flow under twisted boundary conditions (Song et al., 2019).

Notably, fragile phases can be invisible to standard symmetry-indicator methods: in twofold-rotation ( $\mathcal{B}$ 1) symmetric crystals, all atomic insulators have identical symmetry eigenvalues at the high-symmetry points, yet Berry phase and nested Wilson loop invariants may distinguish distinct fragile classes (Kooi et al., 2019).

3. Prototypical Models and Material Realizations

Fragile topology has been exhibited in a wide variety of physical models and material systems:

Honeycomb lattice models: The original explicit construction (Po et al., 2017) uses spin-orbit-coupled $\mathcal{B}$ 2 orbitals in space group $\mathcal{B}$ 3 with time-reversal symmetry ( $\mathcal{B}$ 4), where a four-band elementary band representation splits into two trivial and two fragile conduction bands.
Triangular/kagome lattices: Disconnected EBRs lead to fragile Wilson loop winding and nontrivial entanglement spectra, showcasing the synergy of TQC and real-space diagnostics (Bradlyn et al., 2018, Chen et al., 2022).
Twisted Bilayer Graphene (TBG): Empirically, the magic-angle flat bands of TBG carry $\mathcal{B}$ 5-protected fragile topology, as deduced from symmetry representations, Wilson loop winding, and both real and momentum space diagnostics (Becker et al., 5 Feb 2025, Sati et al., 31 Dec 2025, Lee et al., 5 Mar 2025).
Phononic, photonic, mechanical platforms: Fragile topology has been realized in honeycomb-based phononic and photonic crystals, kagome lattices, and metamaterials, allowing direct experimental access to bulk-defect correspondence and domain-wall-protected modes (Mañes, 2019, Azizi et al., 2023, Wu et al., 2024).
Ab-initio predictions: Exhaustive eigenvalue-based classification (Song et al., 2019) and direct DFT calculations (Könye et al., 16 Jun 2026) have identified hundreds of real materials with robust, well-isolated fragile topological bands, e.g., CsAu₃S₂, Bi₂Ru₂O₇, MoOBrCl, MoAg₂Te₄, PdIr₃S₄Br₄.
Floquet systems: Periodic driving (Floquet engineering) can induce fragile phases protected by crystalline symmetries, featuring higher-order boundary effects such as robust corner charges (Zhang et al., 2020).

Fragile phases have also been rigorously explored in purely interacting (correlated) settings, where parton constructions generate fragile topological insulators of electrons and bosons, with effective negative charge assignments at Wyckoff centers (Latimer et al., 2020, Else et al., 2018).

4. Physical Manifestations and Experimental Signatures

The lack of stable bulk-boundary correspondence is a fundamental property of fragile topology: adding trivial bands can trivialize the phase and destroy any associated edge modes (Po et al., 2017, Sati et al., 31 Dec 2025, Mañes, 2019). Nevertheless, robust experimental signatures do exist:

Filling anomaly and fractional corner/edge charges: Fragile topology manifests as a "filling anomaly"—a mismatch between electrons and Wannier centers per unit cell—which can induce quantized fractional charges at boundaries, corners, or defects. This is notably observed as robust domain wall modes in bi-domain kagome lattices (Azizi et al., 2023), altogether absent for generic edges but protected at engineered interfaces.
Bulk Dirac points enforced by fragile invariants: In systems with $\mathcal{B}$ 6 and $\mathcal{B}$ 7 symmetry, fragile-indexed band subspaces guarantee the existence of symmetry-protected Dirac crossings in the bulk spectrum; these crossings are robust to arbitrary band structure complexity (arbitrary number of bands), and are directly observable in ARPES and spectroscopic measurements (Könye et al., 16 Jun 2026).
Bulk-defect correspondence: Experimental work demonstrates the emergence of 1D bound states at screw dislocations, serving as a unique probe of fragile topology and filling anomaly, even absent any observable boundary states (Wu et al., 2024).
Twisted boundary flows: Under engineered twisted boundary conditions, fragile real-space invariants predict protected level crossings associated with the obstruction, which can be measured in synthetic lattices or metamaterials (Song et al., 2019).
Second-order boundary phenomena: Stacked Chern insulator models with fragile topology exhibit zero-dimensional (corner) modes and a filling anomaly driven quantization of boundary and corner charges, distinguished from strong SOTIs by the lack of a Wannier gap in the Wilson spectrum (Shang et al., 2020).

5. Theoretical Frameworks and Unified Classification

Topological Quantum Chemistry (TQC) and the formalism of EBRs and their algebraic combinations underpin the exhaustive diagnosis of fragile phases (Song et al., 2019). The affine monoid structure (via Hilbert bases of integer cones) classifies fragile roots and their material realizations on a group-theoretic footing.

Homotopy and Cohomology: Fragile phases are classified in unstable (non-additive) homotopy theory rather than stabilized $\mathcal{B}$ 8-theory. For $\mathcal{B}$ 9- or $\mathcal{B}$ 0-invariant bands, the obstruction to trivialization is measured by the (integral) Euler class $\mathcal{B}$ 1 for real rank-2 bundles; this obstruction is "fragile" because it vanishes upon adding a trivial band, yet the $\mathcal{B}$ 2 (Stiefel–Whitney) class can persist (Becker et al., 5 Feb 2025, Bouhon et al., 2020).

Spectral Localizer Approach: Matrix homotopy theory provides a $\mathcal{B}$ 3 energy-resolved topological marker constructed purely from the real-space Hamiltonian and matrix representations of symmetry, enabling direct diagnosis of fragile topology in finite, disordered, or non-periodic systems, with a local gap providing a quantitative robustness criterion (Lee et al., 5 Mar 2025).

6. Applications, Robustness, and Prospects

Fragile topology has implications for both fundamental understanding and device engineering:

Nonlinear optics and transport: Generic Dirac points enforced by fragile topology act as sources of Berry curvature under symmetry breaking, leading to robust second-order (Berry dipole) nonlinear responses (Könye et al., 16 Jun 2026).
Reconfigurable metamaterial design: Fragile-induced domain-wall modes are omnidirectional and robust to interface geometry and disorder, opening pathways for configurable waveguiding, energy localization, and sensing in mechanical, acoustic, and photonic systems (Azizi et al., 2023).
Correlated phases and quantum simulation: Parton-constructed fragile phases extend the notion of fragile topology to interacting bosonic and fermionic systems, with cold-atom and quantum simulator implementations enabling dynamical measurement of topological invariants and dynamical signatures (skyrmion-antiskyrmion pairs, Hopf defects) (Zhao et al., 2022, Latimer et al., 2020).
Materials discovery: Comprehensive theoretical and computational classification generates vast catalogs of fragile phases in real materials, suggesting directions for novel functionalities and quantum hardware leveraging the interplay of momentum-space anyons, quantum gates, and defect statistics (Sati et al., 31 Dec 2025).

7. Extensions, Generalizations, and Open Directions

The fragile topology framework is rapidly expanding:

Beyond symmetry indicators: Homotopy, cohomology, and geometric analysis of higher-dimensional band manifolds (Grassmannians, flag varieties) capture fragile phases inaccessible to standard TQC (Bouhon et al., 2020).
Floquet and non-Hermitian fragile phases: Periodic driving or non-Hermitian system extensions produce entirely new fragile phases, involving quasienergy bands, exceptional points, and amplified/nonreciprocal response (Zhang et al., 2020).
Higher-order and defect-bound states: Fragile topology is deeply linked to higher-order phenomena (corners, dislocations), with bulk-defect correspondence(s) replacing classic bulk-boundary mechanisms (Malavé et al., 2023, Wu et al., 2024).
Robustness and disorder: Fragile topology persists under strong disorder and even allows re-entrant (disorder-induced) transitions, indicating nontrivial "fragile Anderson transitions" (Lee et al., 5 Mar 2025).

Continued theoretical innovation and experimental realization promise further expansion of the fragile topological paradigm, driven by advances in crystalline quantum chemistry, synthetic platforms, and symmetry-protected entanglement.