Non-Hermitian Floquet Photonic Lattice

• Non-Hermitian Floquet photonic lattice is a periodic photonic system where engineered gain/loss and asymmetric couplings yield a complex quasienergy spectrum.
• It employs a one-period propagator framework to reveal phenomena such as anomalous skin effects, non-Bloch band structures, and topological edge-corner states.
• These systems enable asymmetric transport and bulk localization, providing robust control for advanced photonic and quantum applications.

A non-Hermitian Floquet photonic lattice is a photonic lattice whose evolution is periodic in time or in a stroboscopic propagation coordinate, and whose one-period evolution operator is non-unitary because of gain and loss, asymmetric couplings, imaginary gauge fields, or modulation-induced non-reciprocal coupling between Floquet sectors. In such systems the primary spectral object is the Floquet operator rather than a static Hermitian Hamiltonian, and the quasienergy is generally complex. Across coupled ring resonators, helical waveguide arrays, time-multiplexed mesh lattices, photonic quantum walks, microwave resonator arrays, and physical-synthetic resonator-frequency lattices, this framework supports anomalous Floquet non-Hermitian skin effects, Floquet skin-topological corner accumulation, real- and imaginary-gap edge states, non-Bloch $\mathcal{PT}$-symmetry breaking, erratic bulk localization at global reciprocity, and complex band structures defined in generalized or complex momentum space (Gao et al., 2022, Sun et al., 2023, Fritzsche et al., 2020, Longhi, 2019, Park et al., 2021, Sun et al., 31 Mar 2026).

1. Floquet formulation and photonic realizations

The common mathematical structure is a one-period propagator. In coupled ring networks the spectral problem is written as

$\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$

with $\phi$ the Floquet quasienergy and $\hat U$ non-unitary in the presence of gain or loss. In continuously modulated waveguide arrays one instead writes

$$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$

where the propagation coordinate $z$ plays the role of time. In time-multiplexed photonic mesh lattices the evolution is discrete,

$\Psi(m+1)=U_F\,\Psi(m),$

and in physical-synthetic lattices a two-step period takes the form

$$U_T=\exp\!\left(-i H_2T/2\right)\exp\!\left(-i H_1T/2\right)$$

(Gao et al., 2022, Sun et al., 2023, Sun et al., 31 Mar 2026, Ning et al., 5 Jun 2026).

Platform	Periodic variable	Non-Hermitian ingredient
Coupled ring resonator lattice	One circulation / round trip	Fixed on-site gain or loss in each ring
Helical optical waveguides	Helix period $Z$	Structured loss; Floquet-induced complex next-nearest-neighbor couplings
Time-multiplexed photonic mesh lattice	One round trip in unequal fiber loops	Channel-dependent gain/loss encoding disordered imaginary gauge fields
Microwave Floquet medium	Time-periodic capacitance modulation	Non-reciprocal coupling between positive- and negative-frequency sectors plus dissipation
Physical-synthetic resonator-frequency lattice	Two-step modulation period $T$	Imaginary gauge fields and real synthetic flux

The physical implementations are correspondingly diverse. One-dimensional coupled ring resonators use directional couplers and ringwise round-trip gain or loss (Gao et al., 2022). Helical optical waveguides realize Floquet driving geometrically through a $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$0-periodic synthetic vector potential (Li et al., 2023, Sun et al., 2023). Time-multiplexed photonic mesh lattices use two fiber loops of unequal length connected by a $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$1 coupler, with time bins providing a synthetic spatial coordinate and round trips providing discrete Floquet time (Sun et al., 31 Mar 2026). A microwave realization uses a waveguide loaded with an array of varactor-tuned resonators to reconstruct both Bloch-Floquet and non-Bloch band structures experimentally (Park et al., 2021). Photonic quantum walks supply another discrete-time formulation, with gain and loss inserted as a dedicated sub-step of the drive (Longhi, 2019).

2. Mechanisms of non-Hermiticity and complex gauge structure

The literature uses several distinct non-Hermitian resources. In the coupled ring resonator lattice, each ring contributes phase accumulation and amplification or attenuation over one round trip through

$\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$2

and the conceptual distinction from earlier ring proposals is explicit: the loading is fixed and ringwise, not segment-resolved (Gao et al., 2022). In helical waveguide arrays, loss is the basic source of non-Hermiticity, while Floquet engineering creates effective complex next-nearest-neighbor couplings $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$3; the combination yields the chiral transport bias required for a Floquet non-Hermitian skin effect (Li et al., 2023).

A second class of mechanisms is based on imaginary gauge fields. In the time-multiplexed mesh lattice, the target reference model is a disordered Hatano–Nelson chain with asymmetric hoppings

$\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$4

implemented through channel-resolved gain and loss in the four stepwise couplings $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$5. A major conceptual point is the separation of local nonreciprocity from global reciprocity: even with strong linkwise asymmetry, the lattice is globally reciprocal when the mean imaginary gauge field satisfies $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$6 (Sun et al., 31 Mar 2026). In the physical-synthetic resonator-frequency lattice, the synthetic-direction hopping amplitudes

$\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$7

combine imaginary gauge fields $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$8 with real Peierls phases $\hat U |\Psi\rangle=e^{-i\phi}|\Psi\rangle,$9, while the physical-direction step carries both a real synthetic magnetic flux $\phi$0 and a second imaginary gauge field $\phi$1 (Ning et al., 5 Jun 2026).

A third route originates in Floquet-sector coupling itself. In the microwave Floquet medium, the reduced two-band description contains non-reciprocal coupling between an undressed positive-frequency band and a modulation-dressed negative-frequency band, and the resulting effective Floquet Hamiltonian is explicitly non-Hermitian even before ordinary dissipation is added (Park et al., 2021). In non-unitary photonic quantum walks, the non-Hermitian element is a gain/loss operator inserted between two coupling steps, so the one-period propagator becomes non-unitary while preserving a controlled stepwise structure (Longhi, 2019). In the generalized driven Aubry–André–Harper lattice, balanced gain and loss defects are modulated periodically, producing a non-Hermitian Floquet lattice whose static counterpart is generically in the broken phase while the driven system can recover a fully real quasienergy spectrum over a finite parameter region (Blose, 2019).

3. Complex quasienergy, non-Bloch bands, and topological diagnosis

Because the Floquet operator is non-unitary, quasienergies are generally complex. In ring resonator networks, $\phi$2 is the phase accumulated per round trip and $\phi$3 gives the additional net amplification or attenuation required for a self-consistent eigenmode (Gao et al., 2022). In helical-waveguide Floquet systems and mesh lattices, the eigenvalues of the one-period operator similarly encode both oscillatory phase and growth or decay (Sun et al., 2023, Sun et al., 31 Mar 2026).

A central diagnostic is the mismatch between periodic-boundary-condition and open-boundary-condition spectra. In the anomalous Floquet ring lattice, periodic and open spectra can differ qualitatively even when both wrap across the full quasienergy Brillouin zone (Gao et al., 2022). In the loss-induced Floquet NHSE of helical waveguides, periodic spectra form loops in the complex plane while open spectra collapse into their interior, and the generalized Bloch factor $\phi$4 determines the localization direction through $\phi$5 or $\phi$6 (Li et al., 2023). In non-unitary photonic quantum walks, the generalized Brillouin zone is a circle

$\phi$7

and the physically relevant open-boundary quasienergy spectrum is obtained only after this non-Bloch substitution (Longhi, 2019).

The topological diagnostics depend on symmetry class and geometry. For ring-resonator Floquet skin physics, the natural invariant is the winding of the eigenvalues $\phi$8 around a reference point: $\phi$9 For disordered imaginary-gauge-field lattices, a real-space winding number is used,

$\hat U$0

which distinguishes the two oppositely directed skin phases and vanishes at the globally reciprocal erratic point (Gao et al., 2022, Sun et al., 31 Mar 2026). In chiral non-Hermitian Floquet lattices, symmetric time frames produce two windings $\hat U$1, which combine into separate $\hat U$2- and $\hat U$3-gap invariants,

$\hat U$4

with corresponding bulk-edge correspondence for $\hat U$5- and $\hat U$6-edge states (Zhou et al., 2019).

The non-Hermitian Floquet spectrum can also be organized by real and imaginary gaps. In stacked Floquet honeycomb lattices, four non-Hermitian time-reversal-symmetry variants separate phases with counterpropagating edge states in real gaps from phases with a single edge state in an imaginary gap. The latter is topological in the imaginary part of the complex quasienergy spectrum and has flat $\hat U$7, so it does not propagate even though its amplitude grows (Fritzsche et al., 2020). A different but related viewpoint appears in photonic Floquet media, where the experimentally reconstructed Bloch-Floquet bands and non-Bloch bands are described as two measurable subsets of complex eigenfrequency surfaces in complex momentum space (Park et al., 2021).

4. Principal phenomena

The best-known phenomenon in this context is the Floquet non-Hermitian skin effect. In the one-dimensional coupled ring resonator lattice, strong coupling produces an “anomalous Floquet NHSE” in which skin modes occur at essentially every Floquet quasienergy. At perfect coupling $\hat U$8, all open-boundary modes share

$\hat U$9

while their real parts span the entire quasienergy zone; tuning $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$0 through $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$1 reverses the localization direction and reverses the point-gap winding (Gao et al., 2022). A related but distinct mechanism appears in helical waveguide arrays: loss plus Floquet-induced complex next-nearest-neighbor hopping yields a loss-induced Floquet NHSE, and the same mechanism extends to a second-order NHSE in two dimensions (Li et al., 2023).

A second major phenomenon is the Floquet skin-topological effect. In a two-dimensional honeycomb Floquet topological photonic lattice with structured pure loss, the chiral topological edge channels of the Floquet Chern phase acquire point-gap winding in their complex spectrum. Under full open boundaries, the edge states do not remain distributed around the perimeter; instead they accumulate at a selected corner. This corner funneling is not an ordinary higher-order topological corner state and not a bulk two-dimensional skin effect, but a skin effect acting on topological edge states. The same work presents a topological switch: when a sufficiently large sublattice energy difference $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$2 closes the Floquet topological gap, the edge manifold disappears and so does the skin-topological effect (Sun et al., 2023).

A third phenomenon is disorder-enabled bulk localization without boundary accumulation. In the photonic mesh lattice with programmable imaginary gauge disorder, the conventional disordered skin phases occur for $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$3 and $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$4, but at $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$5 boundary accumulation disappears and the eigenstates share a common bulk-localized envelope determined by the disordered imaginary gauge fields. This erratic non-Hermitian skin effect is boundary-independent, bulk in character, and marks the critical point of a disorder-driven transition with $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$6 (Sun et al., 31 Mar 2026).

Non-Hermitian Floquet lattices also support transport channels protected by symmetry rather than by skin accumulation. The non-Hermitian time-reversal-symmetric Floquet ladder has two Kramers-degenerate quasienergy pairs separated by an imaginary gap and realizes a $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$7 phase with two topological transport channels of opposite directionality. The relative phase of two input beams selects which channel is predominantly populated, while the incidence angles determine which channel is suppressed by loss (Höckendorf et al., 2020). In stacked Floquet honeycomb lattices, non-Hermitian time-reversal symmetry can instead protect either two counterpropagating real-gap edge states or a single imaginary-gap edge state that remains spatially localized while its amplitude increases (Fritzsche et al., 2020).

5. Transport, localization, and optical response

A recurring theme is that Floquet non-Hermitian effects translate directly into transport asymmetry. In the ring-resonator lattice, the parameters that yield left-localized skin modes also yield asymmetric transmission: for $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$8, transmission from left to right is continuously attenuated whereas transmission from right to left is essentially unattenuated. In the anomalous regime this persists across the quasienergy zone, and the acoustic analogue displays asymmetric transmission from roughly $$U(Z)=\mathcal{T}\exp\!\left[-i\int_0^Z H(z)\,dz\right], \qquad U_F\psi=e^{-i\beta Z}\psi,$$9 to $z$0, corresponding to a relative bandwidth of about $z$1 (Gao et al., 2022).

In the disordered photonic mesh lattice, transport signatures depend on whether non-Hermitian localization is boundary-driven or erratic. For $z$2 or $z$3, a single-site excitation drifts to the left or right and accumulates at a boundary; at $z$4, the pulse evolution instead settles into a stable bulk-localized pattern without any interface. Adding on-site phase disorder reveals competition with Anderson localization: weak disorder coexists with ENHSE, while strong phase disorder $z$5 produces full Anderson localization and suppresses erratic skin dynamics (Sun et al., 31 Mar 2026).

The physical-synthetic non-Hermitian Floquet lattice separates four issues that are often conflated: topological existence of $z$6- and $z$7-corner states, skin-selected localization of right eigenmodes, local optical visibility, and defective two-period dynamics. The local observable is a balanced interferometric readout of

$z$8

and the relevant matrix element is

$z$9

As a result, the same topological coexistence sector can be bright, skin-dark, or flux-dark in a local measurement. The same complex gauge can also tune an exceptional point of the projected two-period corner propagator, producing a defective doubled-period response with preserved $\Psi(m+1)=U_F\,\Psi(m),$0 alternation and an algebraic Jordan-block envelope (Ning et al., 5 Jun 2026).

Other transport-engineering consequences are more direct. In the non-Hermitian driven binary lattice, the quasienergy minibands can satisfy

$\Psi(m+1)=U_F\,\Psi(m),$1

so forward-propagating modes are amplified and backward-propagating ones are attenuated, yielding robust one-dimensional transport against disorder without relying on edge topology (Longhi et al., 2015). In non-Hermitian Floquet invisibility, a harmonically modulated complex impurity with

$\Psi(m+1)=U_F\,\Psi(m),$2

becomes exactly invisible at $\Psi(m+1)=U_F\,\Psi(m),$3 over the interval $\Psi(m+1)=U_F\,\Psi(m),$4, and for $\Psi(m+1)=U_F\,\Psi(m),$5 invisibility extends over the full tight-binding band; the same effect persists for multiple collectively oscillating impurities (Longhi, 2016).

6. Experimental status, methods, and unresolved issues

Experimental work already spans several photonic and adjacent platforms. The photonic Floquet skin-topological effect has been demonstrated in femtosecond-laser-written helical waveguide arrays (Sun et al., 2023). The erratic NHSE has been observed in a time-multiplexed photonic mesh lattice with programmable gain, loss, and phase modulation (Sun et al., 31 Mar 2026). Microwave experiments have reconstructed both Bloch-Floquet and non-Bloch band structures of a photonic Floquet medium and interpreted them as slices of complex eigenfrequency surfaces in complex momentum space (Park et al., 2021). The anomalous Floquet NHSE of the ring-resonator lattice has been validated experimentally in an acoustic analogue, which, while not optical, directly tests the network mechanism and its broadband asymmetric transport (Gao et al., 2022).

The field also contains a substantial proposal literature. Loss-induced Floquet NHSE in helical waveguides, second-order skin effects, and non-Hermitian higher-order Floquet corner responses in physical-synthetic lattices are developed theoretically with explicit implementation routes but without the same level of experimental realization in the cited works (Li et al., 2023, Ning et al., 5 Jun 2026). The non-Hermitian time-reversal-symmetric Floquet ladder is likewise formulated as a photonic waveguide proposal with disorder-robust directional control (Höckendorf et al., 2020). Periodically driven generalized Aubry–André–Harper lattices show that balanced gain/loss defects can recover fully real quasienergies in regimes where the static non-Hermitian system is complex for any nonzero non-Hermitian degree, but this remains a theoretical construction in the cited work (Blose, 2019).

Several unresolved issues are explicit in the literature. The ring-resonator paper formulates the theory directly at the level of the non-unitary Floquet operator and does not provide a closed-form Bloch matrix $\Psi(m+1)=U_F\,\Psi(m),$6 or an explicit generalized Brillouin-zone equation in the main text (Gao et al., 2022). The Floquet skin-topological experiment emphasizes spectral winding and transport phenomenology but does not develop a full non-Bloch analytical theory for the two-dimensional Floquet case (Sun et al., 2023). The stacked honeycomb study leaves the robustness of the imaginary-gap edge state against disorder open, because its protection depends on momentum pinning at $\Psi(m+1)=U_F\,\Psi(m),$7 (Fritzsche et al., 2020). The microwave Floquet-medium analysis is based on a linearized model, and the non-Bloch phase-retardation measurement relies on a single-pass approximation that breaks down at stronger gain (Park et al., 2021).

Methodologically, the topic has begun to absorb data-driven tools. A periodically driven non-Hermitian bipartite lattice study introduces amortized clustering and an algorithm selector to classify Floquet eigenfunctions, emphasizing Floquet $\Psi(m+1)=U_F\,\Psi(m),$8-modes and anomalous hybrid corner-edge or edge-bulk localization patterns (Xia et al., 1 Aug 2025). This suggests a growing need for mode-classification methods as driving protocols, synthetic dimensions, and non-Hermitian control parameters become more elaborate.

Taken together, these works establish the non-Hermitian Floquet photonic lattice as a family of non-unitary periodic photonic systems in which quasienergy periodicity, gain/loss or asymmetric couplings, and complex spectral topology jointly determine transport, localization, and observability. The unifying lesson is not a single mechanism but a structured set of mechanisms: ringwise on-site loading, Floquet-induced complex longer-range hopping, disordered imaginary gauge fields, non-Hermitian coupling between Floquet sectors, and complex-gauge interference all generate distinct versions of non-Hermitian Floquet photonic matter (Gao et al., 2022, Li et al., 2023, Sun et al., 31 Mar 2026, Park et al., 2021, Ning et al., 5 Jun 2026).