The day in conversation for 2026-11-09

discussion

Transcript

The discussion: Kai: That bulk-band topology approach for creating those photonic exceptional topological bound states seems incredibly promising for robustness, but I’m still concerned about the sensitivity of the exceptional point itself to fabrication tolerances. Mira: I see the potential for localization, yes, but focusing on that specific parameter space in a continuum model doesn't tell us much about what happens when we move from ideal geometry to real-world etching errors. Have you considered how much error tolerance those robust states actually need to handle before the topological protection breaks down? Lev: From my perspective on metrology, the real test isn't just theoretical robustness against perturbations, but whether we can measure that localization accurately enough to confirm the topological nature in experiment. If the measurement uncertainty is too high, we lose the distinction between a topologically protected state and a standard localized state. Kai: That ties directly into what I was reading last week about how slight material variations affect band structure calculations; if the bulk band topology isn't sufficiently separated from non-topological bands, that gap narrows rapidly. What kind of threshold in the perturbation strength are we anticipating before that separation collapses? Mira: It depends heavily on the specific symmetry we exploit. If we merge two symmetry-protected states at an exceptional point, the protection mechanism should be quite strong, but I wonder if introducing any parasitic coupling from nearby modes might bypass that topological constraint entirely. Does your simulation account for those off-diagonal coupling terms adequately? Lev: The simulation focuses on the dominant terms first, but I think we need to look at the spectral density near the exceptional point. If the non-Hermitian aspects of these states are significant, even small imaginary components can introduce leakage that undermines our localization claims. Are we relying solely on real-mode analysis

More episodes

← Home