---
title: Non-Hermitian Topological Band Theory
url: https://www.emergentmind.com/topics/non-hermitian-topological-band-theory
type: topic
---

# Non-Hermitian Topological Band Theory

Non-Hermitian topological band theory is a generalization of conventional band topology that encompasses complex eigenvalue spectra, exceptional point singularities, and non-orthogonal eigenstates arising in systems governed by non-Hermitian Hamiltonians. Unlike Hermitian systems, where band structures admit a real energy spectrum and unitary evolution, non-Hermitian physics naturally describes open, dissipative, gain/loss, or non-reciprocal systems, prominent in photonics, cold-atom settings, active metamaterials, and beyond. This framework necessitates new mathematical definitions of band gaps, novel topological invariants, and the inclusion of phenomena such as the non-Hermitian skin effect and exceptional degeneracies, leading to an enriched classification of topological phases and boundary states.

## 1. Gap Notions and Band Structure in the Complex Plane

**Line-Gap vs. Point-Gap.** In non-Hermitian systems, two non-equivalent notions of a spectral gap are central: a *line-gap* excludes a prescribed line (e.g., $\mathrm{Re}E=0$) from the spectrum at every $\mathbf{k}$, while a *point-gap* excludes a reference point $E_0\in\mathbb{C}$—i.e., $E_0\notin\mathrm{Spec}[H(\mathbf{k})]$ for all $\mathbf{k}$ [1706.07435, 1911.12748, 2504.15154]. The point-gap notion is critical for capturing genuinely non-Hermitian phenomena (e.g., the skin effect), whereas conventional Hermitian invariants are associated with line-gap topology.

**Separability and Band Gaps.** A non-Hermitian band is *separable* if its eigenvalue locus does not cross those of any other band on the Brillouin zone [1706.07435]. In the complex plane, isolated bands—i.e., those whose energy images are surrounded by gaps—generalize the concept of insulating bands. The most general setting involves separable bands with non-degenerate spectra across the Brillouin zone, or a prescribed separation between compact subsets of bands (separation gaps), which further refines possible invariants [2309.14416].

## 2. Topological Invariants: Winding, Chern, and Braid Group

**1D Winding/Braid Group.** In one dimension, the key invariant is the winding of the complex eigenvalues (or their differences) around a base point $E_0$. For $N$-band systems, the robust topological invariant is the conjugacy class of a braid in $B_N$ (the Artin braid group), reflecting how eigenvalues permute as $k$ winds over the Brillouin zone [1911.12748, 2111.09977]. For two bands, this reduces to a familiar integer winding number; for $N>2$, the invariants are intrinsically non-Abelian [2212.13736].

**2D Chern Numbers and Fractional Winding.** The non-Hermitian analogue of the Chern number employs the biorthogonal basis of left/right eigenvectors. The Chern number is constructed from the biorthogonal Berry curvature:
\[
C = \frac{1}{2\pi}\int_{BZ} \epsilon^{ij} \langle \partial_{k_i} \psi_n^L|\partial_{k_j}\psi_n^R\rangle \, d^2k
\]
which remains quantized as long as the relevant band remains isolated [1706.07435, 1901.01408]. In sectors with band permutation (e.g., via nonseparable bands connected by PHLs), fractional winding numbers $W_m/n$ arise, reflecting $n$-cycle exchanges and associated with robust topological phase phenomena in the absence of exceptional points [2405.17749].

**Higher-Dimensional Fragile and Torsion Invariants.** In 2D, the presence of non-trivial braid-group sectors can reduce the possible topological invariants to cyclic groups $\mathbb{Z}_n$ or $\mathbb{Z}_2$, making topological structure "fragile": the invariant can be trivialized by the addition of trivial bands [1911.02697, 1911.12748]. In 3D, Hopf-type invariants protected by point gaps emerge for two-band systems, such as the non-Hermitian Hopf insulator with $\mathbb{Z}$– or $\mathbb{Z}_2$–valued topological charge depending on symmetry and dimensionality [2504.15154].

**Green Function and Gauge-Free Approaches.** In continuum models, the non-Hermitian Chern number can be cast as a frequency-momentum space integral involving the retarded Green's function,
\[
C = -\frac{1}{2\pi^2} \int_{BZ} dk_x dk_y \int_{\omega_{\mathrm{gap}}-i\infty}^{\omega_{\mathrm{gap}}+i\infty} d\omega \;\epsilon^{ij}\;\mathrm{Tr}\left[ G(\mathbf{k},\omega)\,\partial_{k_i}G^{-1}(\mathbf{k},\omega)\, G(\mathbf{k},\omega)\,\partial_{k_j}G^{-1}(\mathbf{k},\omega)\right]
\]
where the analyticity of $G$ in a vertical strip of the complex-$\omega$ plane defines the generalized band gap [1901.01408].

## 3. Exceptional Points, EP Braiding, and Pseudo-Hermitian Lines

**Exceptional Points and Braiding.** Non-Hermitian systems generally allow robust (codimension-2) exceptional point (EP) degeneracies, where eigenvalues and eigenvectors coalesce. Their presence is encoded by braid-group data in the momentum space: for example, in 2D, the fundamental group of the Brillouin zone punctured by $k$ EPs leads to constraint equations relating the braids around cycles and the EPs themselves [2111.09977]. Braiding of bands at EPs underlies nontrivial phase transitions and can generate non-Hermitian topological phase transitions not present in Hermitian settings [2212.13736].

**Pseudo-Hermitian Lines (PHLs).** Topological features can persist without EPs, mediated by PHLs along which the Hamiltonian is pseudo-Hermitian (i.e., $\eta H \eta^{-1} = H^\dagger$ for a fixed invertible $\eta$) [2405.17749]. In 2D, a PHL can be non-contractible on the torus, allowing nontrivial topology even in the absence of band-touching points. The permutation group $S_N$ organizes the possible exchange classes of bands, and fractional winding numbers result from the action of PHLs.

**Symmetry Engineering and Indicator Formulas.** Non-Hermitian symmetries generalize the Hermitian Altland-Zirnbauer classes to a rich Bernard-LeClair symmetry taxonomy [1812.10490]. Generalized inversion and chiral symmetries lead to topological indicators expressible in terms of simple functions at inversion-invariant momenta, facilitating the diagnosis of skin effects and EP-related topological transitions [2103.05141].

## 4. Bulk–Boundary Correspondence and the Non-Hermitian Skin Effect

**Non-Hermitian Skin Effect (NHSE).** In contrast to Hermitian systems, the bulk spectrum under periodic (PBC) and open (OBC) boundary conditions can differ dramatically: a nonzero point-gap winding predicts the macroscopic accumulation of bulk states at the system boundary under OBC—the skin effect. The OBC spectrum collapses onto a region of the complex plane corresponding to zero point-gap winding [2109.01328, 2507.09447]. In models with translational symmetry, this is encoded by non-Bloch band theory, where boundaries are incorporated by shifting momenta into the complex plane; in disordered or aperiodic systems, the Lyapunov exponent formulation generalizes this physics [2507.09447].

**Bulk–Edge Correspondence.** The number of protected edge states is controlled by the change in the appropriate topological invariant (e.g., winding number or Chern number) across a boundary or domain wall. For lossy systems described by Maxwell's equations, the non-Hermitian Chern number remains quantized and edge modes traverse the complex-frequency band gaps as in Hermitian systems, provided the analyticity strip is preserved [1901.01408].

**Flat Bands and Higher-Order Skin Modes.** Real-space decimation methods elucidate the formation of flat bands and compact localized states, as well as higher-order skin effects, where not only edges but also corners accumulate exponentially localized states, signaled by corresponding topological invariants in the real-space framework [2310.04834].

## 5. Interplay of Symmetry, Fragility, and Topological Classification

**Symmetry-Protected Phases and Fragility.** Non-Hermitian Bernard-LeClair symmetries and point-gap/line-gap distinctions result in a "38-fold periodic table" for symmetry-protected topological phases [1812.10490]. However, certain invariants are only stable as long as the band number and symmetry structure are fixed; adding trivial bands can trivialize non-Hermitian “torsion” (fragile) invariants [1911.02697]. For two-band systems, non-Hermitian Hopf-type invariants (in 3D and 4D) exist, which do not extend to higher bands [2504.15154].

**Homotopy and Braid-Frame Classifications.** The homotopy-theoretic approach reveals a two-level hierarchy: eigenvalue braiding (non-Abelian and Abelian braid-group structure) classifies sectors, while eigenvector (frame) topology further distinguishes subclasses in each sector [1911.02697, 2111.09977, 2309.14416]. This framework naturally unifies and extends K-theoretic line-gap schemes and captures both stable and fragile topological phenomena unique to non-Hermitian band physics.

**Pseudo-Hermitian Topologies.** Pseudo-Hermitian operators support phases where the spectrum remains real for arbitrarily large non-Hermiticity, circumventing the limitations of conventional PT-symmetric models and giving rise to robust topological phases with well-defined bulk-boundary correspondence [2111.02701, 2106.09648].

## 6. Physical Realizations, Experimental Platforms, and Applications

**Optics, Photonics, Atomic, and Electronic Metamaterials.** Non-Hermitian band topology is realized in systems with balanced gain/loss, radiative decay, nonreciprocal couplings, and engineered dissipation, such as photonic crystals, cold-atom arrays, acoustic devices, and topological lasers [1901.01408, 2601.00487].

**Atomic Lattices and Dirac-Type Models.** Long-range radiative coupling in bipartite atomic lattices leads to non-Hermitian Dirac equations with complex Fermi velocities and biorthogonal bulk and edge state topology [2601.00487]. Analytic edge-state solutions verify bulk-edge correspondence under non-Hermitian deformations.

**Twisted Graphene Systems.** The combination of non-reciprocal hopping and sublattice-staggered mass in twisted bilayer graphene aligned with hBN produces non-Hermitian valley Hall phases, gapless nodal-line physics across extended parameter regions, and robustness of Dirac points against symmetry-breaking perturbations [2505.00566].

**Measurement Protocols.** Experimental access to non-Hermitian geometric tensors, Berry curvature, and quantum metric is feasible via dilated Hermitian embeddings and modulation spectroscopy in platforms such as superconducting qubits, NV centers, and trapped ions [2106.09648].

## 7. Outlook and Open Directions

Non-Hermitian topological band theory has established new paradigms for classifying and understanding topological matter beyond Hermitian constraints. Open directions include the systematic classification of intrinsic/extrinsic point-gap invariants for higher-band and higher-dimensional systems, the full characterization of exceptional degeneracy braiding in real multi-band structures, robust detection protocols for fragile invariants, and the integration of interaction and Floquet effects. The interplay of symmetry, disorder, non-Hermitian topology, and dynamical phenomena continues to drive rapid developments at the intersection of condensed matter physics, optics, quantum engineering, and mathematical physics [1911.12748, 2309.14416, 2504.15154].

Source: https://www.emergentmind.com/topics/non-hermitian-topological-band-theory