Band Inversion Flips the Winding of Bound States in the Continuum

Technical Summary: Band Inversion Flips the Winding of Bound States in the Continuum

Problem Statement
Bound states in the continuum (BICs) in photonic slabs and metasurfaces manifest as polarization singularities in momentum space, characterized by an integer winding number ($w_{\text{BIC}}$). This winding is traditionally viewed as a robust topological label, preserved under smooth structural deformations. However, in open photonic structures, the band structure is intrinsically non-Hermitian due to radiative leakage, leading to complex eigenfrequencies. A central question remains: which topological aspects of BIC polarization vortices remain robust when the band structure itself is engineered in the presence of radiative loss? Specifically, does the winding number of a band-edge BIC remain invariant when the underlying band ordering is altered via band inversion?

Methodology
The authors employ a multi-faceted approach combining general theoretical derivation, analytical modeling, numerical simulation, and experimental verification:

1. General Two-Band Theory: A general description of non-Hermitian band inversion in open periodic photonic structures is developed. The theory models two hybridized lattice resonances ($|\Psi_1\rangle$ and $|\Psi_2\rangle$) that exchange radiative loss through coupling to a common continuum. The far-field polarization is analyzed via a linear projection of the collected Jones vector, allowing for the derivation of the polarization orientation angle $\phi(\mathbf{k})$ and its winding number near the BIC.
2. Analytical Model (1D Grating): A minimal effective non-Hermitian Hamiltonian is derived for a 1D subwavelength grating near the $\Gamma$-point. This model explicitly tracks the mixing ratio $r(\mathbf{k})$ of the two basis modes and the resulting far-field components ($E_\parallel, E_\perp$) to demonstrate how the Jacobian of the polarization map behaves across a band inversion.
3. Numerical Simulations: Rigorous Coupled-Wave Analysis (RCWA) and finite-element eigenmode simulations (COMSOL) are performed on:
   • A tunable 1D grating to verify the minimal model.
   • Multiband rectangular and triangular photonic-crystal slabs (TiO$_2$) to test the rule in complex, multi-band scenarios, including cases with coexisting symmetry-protected and Friedrich–Wintgen BICs.
4. Experimental Demonstration: Angle-resolved reflectivity-contrast spectroscopy and polarization-resolved tomography are performed on fabricated gold gratings on dielectric Bragg mirrors. The grating filling ratio is tuned to induce band inversion, and the momentum-space polarization textures are reconstructed to image the vortex circulation.

Key Contributions and Results

• Theoretical Proof of Winding Flip: The authors prove that band inversion at a band-edge BIC reverses the local far-field polarization map, flipping the winding number $w_{\text{BIC}} \to -w_{\text{BIC}}$. This occurs without any defect motion in momentum space. The mechanism is identified as the exchange of modal character between two hybridized branches: as the branches swap their dominant character across a gap closing (defined by the real part of the complex eigenfrequencies), the slope of the mixing ratio $\nabla_k |r|$ changes sign. Since the winding is determined by the sign of the Jacobian determinant of the polarization map, which depends on this slope, the winding flips.
• Geometric Interpretation: At the critical point of inversion, the polarization texture collapses from a first-order vortex into a rank-deficient, locally shear-like pattern where the winding is undefined. As the parameter is tuned past the inversion, the vortex re-forms with opposite handedness.
• Experimental Verification: In a tunable 1D grating, the authors experimentally observe the inversion of a non-radiative dark branch with its radiative bright partner. Polarization tomography directly images the reversal of the BIC vortex circulation across the transition, confirming the theoretical prediction.
• Multiband Validity: Numerical simulations on rectangular and triangular lattices confirm that the rule holds in multiband non-Hermitian structures. The winding flip occurs branch-by-branch. The paper notes that in high-symmetry lattices (e.g., $C_4$ or $C_6$) where a dark branch crosses a degenerate pair of bright modes, two successive flips occur, resulting in a net unchanged winding. However, when degeneracy is lifted (e.g., by breaking symmetry), the individual sign reversals are resolved.
• Coexistence of BICs: The rule is shown to apply even when multiple BICs (symmetry-protected and Friedrich–Wintgen) coexist on a single branch; the total branch-resolved winding flips sign upon inversion.

Significance and Claims
The paper establishes that the winding number of a band-edge BIC is not an inversion-stable topological label, contrasting with Hermitian topological band theory where band inversion redistributes Berry curvature but conserves total Chern numbers. Instead, in non-Hermitian open systems, the invariant is carried by a single mode and is reversed by the inversion of the band ordering.

The authors claim that band inversion emerges as a key mechanism governing polarization-singularity topology in non-Hermitian photonic band structures. This finding provides a deterministic recipe for reconfiguring polarization-vortex patterns in momentum space through simple parameter tuning (e.g., filling fraction or slab thickness). The work suggests potential implications for engineering high-Q resonances and shaping far-field radiation, specifically for vortex-beam shaping and the on-demand spawning of circularly polarized states, though the paper focuses primarily on the fundamental topological mechanism rather than specific device applications.