Modern nanophotonics has established a powerful multilevel design framework, spanning local amplitude and phase control [
1], resonant modal engineering [
2], multiport scattering optimization [
3], and computational inverse design [
4]. This mature toolkit enables precise control over the amplitude and phase of transmitted and reflected fields [
1], polarization and mode conversion [
1,
4], and optical splitting and routing [
3]. Yet in compact multifunctional devices, these observables are often governed by a shared set of resonant modes [
5] and a limited number of radiative channels [
6]. Tuning a single structural parameter may shift resonance frequencies and decay rates while simultaneously reorganizing several scattering coefficients [
6]. The design challenge therefore extends beyond realizing an isolated optical response to coordinating modal frequencies, decay dynamics, eigenstate relations, and the coupling between internal modes and external channels. Non-Hermitian design approaches this coordination by incorporating loss, gain, radiative leakage, and complex coupling as design variables within the modal and scattering architecture [
7]. Engineering these open-system interactions and coupling relations can restructure eigenvalue spectra, eigenstate geometry, and mode–channel coupling, allowing interaction with the environment to participate directly in generating the desired optical response [
8]. Under appropriate conditions, such organization gives rise to eigenvalue and eigenstate coalescence at exceptional points [
9], pronounced eigenstate nonorthogonality [
10], Riemann surface structure [
11], complex-spectral winding [
12,
13], and dissipation-driven spectral reorganization [
14]. These modal and spectral structures, in turn, reshape how internal excitations couple to, interfere through, and redistribute among external channels, producing channel-selective scattering [
15,
16], parameter-dependent phase evolution [
17], and switching between distinct scattering states [
18]. When engineered non-Hermitian modal and scattering relations become integral to the functional mechanism, openness becomes function itself.
This design challenge rests on the correspondence between internal open-system modes and external scattering channels, as schematically illustrated in Fig. 1. Here, modal–scattering correspondence denotes how the complex spectrum and eigenstates of an open photonic system are projected onto measurable scattering responses through mode–channel coupling and interference. In a representative two-mode, two-port system, incident amplitudes
A1,2 excite two internal resonant modes characterized by resonance frequencies
ω1,2 and decay rates
Γ1,2. The modes interact through the intermodal coupling μ and exchange energy with the external channels through coupling coefficients
κ1,2, collectively determining the outgoing amplitudes
B1,2 [
6]. Non-Hermitian design acts on this correspondence by engineering both the internal modal structure and its coupling to external channels. In finite systems, for example, the effective non-Hermitian modal operator can be tuned toward a resonant (modal) exceptional point, where its complex eigenfrequencies and eigenstates coalesce [
9]. This modal degeneracy should be distinguished from a scattering exceptional point, which instead corresponds to the coalescence of eigenvalues and eigenchannels of the scattering matrix. Because the scattering matrix depends on both the internal modal dynamics and their coupling to external channels, the two exceptional-point conditions need not coincide [
6]. Consequently, the observable scattering consequence of a modal exceptional point is determined not by the internal singularity alone, but also by how it is coupled to and read out through the external channels. More generally, non-Hermitian eigenstates can become strongly nonorthogonal, thereby altering their excitation, interference, and redistribution under a common external drive [
10,
15]. In extended or periodic systems, non-Hermitian coupling can instead generate point-gap spectral winding, where the complex spectrum
E(
k) encircles a reference point
E0 with
W(
E0) ≠ 0 [
12,
13]. This spectral winding is distinguished from parameter-space encircling of an exceptional point [
13,
17], as it is intrinsically associated with the global structure of the complex spectrum and can lead to pronounced boundary sensitivity [
19] and non-Hermitian skin localization [
20−
22]. Across these different regimes, the central design task is therefore not merely to create non-Hermitian modal or spectral structures, but to translate them into predictable and controllable external responses. This requirement is motivating increasingly systematic strategies based on mode-channel engineering [
6], dissipative control [
14,
18], and complex-coupling design [
23].
Recent studies are increasingly translating non-Hermitian modal and spectral structures into experimentally accessible control parameters and measurable photonic responses, as summarized in Fig. 2. At the level of modal physics, generalized Petermann-factor criteria [
10], direct measurements of left−right eigenvectors [
24], and point-gap descriptions of non-Hermitian spectra [
12,
13] have provided quantitative access to eigenstate nonorthogonality, biorthogonal structure, and spectral winding. The emphasis is now shifting from identifying these non-Hermitian properties to systematically controlling how they participate in device operation. Periodic modulation, for example, can reshape the gain-loss balance and shift the parity−time-symmetric (
PT-symmetric) phase boundary of optical couplers [
25], whereas complex-index
PT modulation can introduce a directional momentum channel for asymmetric conversion into selected second-harmonic guided modes [
26]. In a distinct scattering architecture, non-Hermitian epsilon-near-zero photonic doping can decouple transmission and reflection, enabling independent control over their amplitudes and phases and extending this principle toward multiport signal manipulation [
27]. More direct modal-to-response mappings have also emerged in metasurfaces. Encircling a scattering exceptional point can map a closed parameter trajectory onto a complete 2π phase winding within a selected cross-circularly polarized scattering channel [
6,
17], while optically induced dissipation can switch a bound-state-in-the-continuum-derived (BIC-derived) exceptional ring and thereby reconfigure beam deflection [
18]. These examples reveal a common design principle: gain, loss, complex coupling, and radiative leakage are no longer treated merely as perturbations to an underlying resonance, but are deliberately organized to determine which modal pathway is activated, how energy is redistributed among scattering channels, and which operating state the photonic system ultimately occupies.
The next step is therefore not simply to pursue stronger amplitude, phase, polarization, or wavefront responses, many of which are already attainable through established resonant and inverse-design approaches [
1,
4]. A more consequential question is which functionalities arise specifically from the non-Hermitian organization of modal decay, complex coupling, and external scattering channels. This distinction is already evident in recent metasurface studies. Pronounced spin-dependent responses, for example, can already be realized using conventional quasi-BIC resonances [
28], without requiring an exceptional-point-based mechanism. By contrast,
PT-symmetric metamaterials can exploit exceptional points to enhance and asymmetrize photonic spin Hall responses [
29], while other exceptional-point and exceptional-ring platforms connect polarization conversion [
15], phase evolution [
17], and beam switching [
18] to prescribed reorganizations of the underlying modal or scattering structure. The significance of non-Hermitian design therefore lies less in generating an unusual optical response than in establishing a controllable physical mechanism through which that response is selected, redirected, or reconfigured. We argue that non-Hermitian engineering is particularly compelling in regimes where controlled dissipation, mode−channel coupling, or scattering-state selection directly participates in the target functionality. Reconfigurable scattering systems can exploit controlled dissipation to switch between distinct modal states [
18]; multiport or directional-scattering devices can use engineered mode−channel relations to redistribute, enhance, or suppress selected outputs [
27,
30]. The key question is therefore not whether a non-Hermitian system produces a distinctive response, but whether engineering these open-system relations provides a more direct, robust, or reconfigurable route to the desired functionality under the same physical constraints as conventional photonic design.
PDF
(3678KB)
