Spatiotemporal Nonreciprocal Metasurfaces

• Spatiotemporal metasurfaces are ultrathin interfaces whose boundary parameters vary in space and time via traveling modulation, breaking conventional Lorentz reciprocity.
• They employ formalisms such as GSTCs, impedance models, and Floquet–Bloch expansions to analyze and achieve nonreciprocal scattering and frequency conversion.
• Experimental realizations across microwave to optical regimes demonstrate one-way transmission, asymmetric beam steering, and emerging quantum and thermal photonic functionalities.

Spatiotemporally modulated nonreciprocal metasurfaces are ultrathin wave-engineering interfaces whose effective boundary parameters—surface susceptibilities, impedance, conductivity, refractive index, or related quantities—vary in both space and time, typically as a traveling perturbation. In these structures, wave scattering is governed by simultaneous energy and momentum exchange, so that frequency and tangential wavevector are redistributed through channels such as $\omega_n=\omega_0+n\omega_m$ and $k_{x,n}=k_x+n\beta_m$. Because time invariance is explicitly broken, the forward and backward scattering problems are no longer constrained by conventional Lorentz reciprocity, enabling one-way transparency, asymmetric frequency conversion, nonreciprocal beam steering and focusing, and direction-dependent conversion between free-space radiation and surface-bound states without magnetic bias (Hadad et al., 2015, Gupta et al., 2017, Cardin et al., 2019).

1. Foundational definition and reciprocity breaking

A spatiotemporally modulated metasurface is a subwavelength-thin boundary whose effective response varies jointly in space and time. In a susceptibility-based formulation, a traveling modulation along $+x$ can be written as

$$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$

with modulation velocity $v_{\text{mod}}=\omega_m/\beta_m$ (Gupta et al., 2017). In conductivity-based descriptions, the same idea appears as

$$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$

where $\phi(\mathbf r)$ supplies the spatial phase profile and $\Omega$ supplies the temporal pumping frequency (Cardin et al., 2019).

The central reciprocity-breaking mechanism is the coexistence of a directional perturbation and a wave component parallel to the modulation. In the electromagnetic sheet model of Taravati and Eleftheriades, nonreciprocity emerges under oblique incidence because the transverse momentum $K_x=k_0\sin\theta_0$ is parallel to $\beta_m$ in one direction and anti-parallel in the other, changing the phase-matching conditions for excitation of space-time harmonics (Gupta et al., 2017). Hadad, Sounas, and Alù expressed the same principle in impedance language: adding a temporal gradient to a gradient metasurface breaks the symmetry constraints that hold for a static, linear surface and produces a non-reciprocal classical analogue to electromagnetic induced transparency, with a narrow window of one-way transmission in an otherwise opaque surface (Hadad et al., 2015).

This mechanism is distinct from three frequently conflated effects. First, it is not magneto-optical nonreciprocity, since no static magnetic bias or ferrite gyrotropy is required. Second, it is not nonlinear nonreciprocity, because the underlying response may remain linear in the signal field while being time-variant in the medium parameters. Third, it is not mere structural asymmetry: static asymmetric geometries can yield directional asymmetries while remaining reciprocal in the Lorentz sense (Gupta et al., 2017, Cardin et al., 2019).

2. Governing formalisms: GSTCs, impedance models, and Floquet–Bloch expansions

Two modeling frameworks dominate the literature. The first is the generalized sheet transition condition formulation, where the metasurface is idealized as a zero-thickness discontinuity at $k_{x,n}=k_x+n\beta_m$0. In time domain, the field jumps satisfy

$k_{x,n}=k_x+n\beta_m$1

$k_{x,n}=k_x+n\beta_m$2

with surface polarizations related to average fields through dispersive surface susceptibilities (Gupta et al., 2017). In the matched-sheet TE configuration treated there, longitudinal polarizations are neglected, $k_{x,n}=k_x+n\beta_m$3, and the space-time dependence is inserted through a Lorentzian resonant frequency modulation,

$k_{x,n}=k_x+n\beta_m$4

This preserves a causal dispersive model while generating harmonic coupling across both $k_{x,n}=k_x+n\beta_m$5 and $k_{x,n}=k_x+n\beta_m$6 (Gupta et al., 2017).

The second framework uses an effective impedance or conductivity boundary. In the microwave reflectarray demonstration of dynamic free-space nonreciprocity, the metasurface is modeled via a time-dependent 2D conductivity,

$k_{x,n}=k_x+n\beta_m$7

implemented by varactor-loaded resonators with column-dependent RF bias phases (Cardin et al., 2019). In the earlier impedance-sheet theory of Hadad, Sounas, and Alù, the surface operator is

$k_{x,n}=k_x+n\beta_m$8

which supports a space-time leaky-mode picture and the non-reciprocal EIT analogue (Hadad et al., 2015).

Because the modulation is periodic in time and usually periodic or smoothly phased in space, the scattered fields are expanded in Floquet series. A canonical transmitted field expansion is

$k_{x,n}=k_x+n\beta_m$9

with

$+x$0

The corresponding propagation angles obey

$+x$1

so the same harmonic may be radiative in one channel and evanescent in another (Gupta et al., 2017). In the generalized Bloch–Floquet treatment of dynamic reflectarrays, this becomes the local rule

$+x$2

which remains valid for smooth but nonuniform $+x$3 (Cardin et al., 2019).

3. Physical mechanisms of nonreciprocal scattering

The most elementary mechanism is directional phase matching. For an incident oblique plane wave, the traveling modulation alters the relative sign between the incident transverse momentum and the modulation momentum. In the GSTC/Lorentzian sheet model, this sign change reshuffles which harmonics are efficiently excited, so the fundamental transmitted coefficient $+x$4 and the generated sidebands differ between forward and backward propagation (Gupta et al., 2017). At normal incidence, $+x$5, and the same model restores reciprocity: the harmonic spectra are identical for modulation along $+x$6 and $+x$7 (Gupta et al., 2017).

A second mechanism is selective routing between radiative and surface-wave channels. In the dynamic reflectarray platform, a linear spatial phase $+x$8 and temporal modulation $+x$9 define the generalized space-time grating equation

$$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$0

If the resulting $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$1 exceeds $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$2, the scattered harmonic becomes evanescent rather than radiative. The experimentally significant consequence is that a forward free-space beam can be converted into a propagating steered harmonic, whereas the time-reversed input couples mainly into a surface wave and does not reconstruct the original beam (Cardin et al., 2019).

A third mechanism is leaky-mode interference. In the space-time gradient metasurface of Hadad, Sounas, and Alù, the unmodulated surface is opaque at its Lorentzian resonance, but spatial modulation produces a narrow leaky-mode transmission window. Temporal modulation then shifts the leaky-mode cutoffs differently for opposite propagation directions, splitting the forward and backward transparency conditions and yielding a one-way EIT-like response (Hadad et al., 2015).

Other realizations recast the same principle in different physical languages. Mechanically rotating dielectric cylinders realize a moving-medium analogue in which the spinning motion induces a bi-anisotropic Tellegen-type response and lifts the degeneracy of clockwise and counterclockwise multipole modes through a Sagnac-type effect. That in turn produces a non-symmetric scattering matrix and direction-dependent coupling to substrate modes (Yang et al., 2024). A related, but conceptually distinct, regime has been identified in short spatiotemporally modulated systems where transmitted amplitudes can be equal in both directions while the transmitted phases differ. This suggests that phase-only nonreciprocity is a genuine operating regime rather than a perturbative artifact, and it is directly relevant to phase-engineered metasurface functions such as beam synthesis and interferometric isolation (Wu et al., 2024).

4. Representative implementations and demonstrated functionalities

The first rigorous electromagnetic sheet treatment with space-time-varying susceptibilities demonstrated non-reciprocal transmission numerically under oblique TE incidence using GSTCs and causal Lorentzian surface susceptibilities. With $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$3, $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$4, $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$5, $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$6, and $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$7, the computed spectra $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$8 and $$\chi_{ee}(x,t)=\chi_{ee}(\omega_m t-\beta_m x),\qquad \chi_{mm}(x,t)=\chi_{mm}(\omega_m t-\beta_m x),$$9 differ significantly at $v_{\text{mod}}=\omega_m/\beta_m$0 and $v_{\text{mod}}=\omega_m/\beta_m$1, including the fundamental coefficient $v_{\text{mod}}=\omega_m/\beta_m$2 (Gupta et al., 2017).

A microwave experimental milestone was the varactor-loaded spatio-temporally modulated reflectarray. It used a $v_{\text{mod}}=\omega_m/\beta_m$3 FR-4 platform, about $v_{\text{mod}}=\omega_m/\beta_m$4 at $v_{\text{mod}}=\omega_m/\beta_m$5 GHz, with unit cells containing two SMV1405 varactor diodes and two fixed capacitors of $v_{\text{mod}}=\omega_m/\beta_m$6 pF. The modulation parameters were $v_{\text{mod}}=\omega_m/\beta_m$7 V, $v_{\text{mod}}=\omega_m/\beta_m$8 V, $v_{\text{mod}}=\omega_m/\beta_m$9 kHz, and $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$0 GHz. In beam steering, the $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$1 harmonic was steered continuously from $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$2 to $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$3 by varying $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$4, with conversion efficiency to the $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$5 harmonic $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$6. For $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$7, the forward $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$8 harmonic appeared at $$\sigma(\mathbf r,\omega;t)=\sigma_0(\omega)+\Delta\sigma(\omega)\sin[\phi(\mathbf r)-\Omega t],$$9, whereas the reverse excitation produced a reflected beam at $\phi(\mathbf r)$0, not $\phi(\mathbf r)$1. For $\phi(\mathbf r)$2, the reverse harmonic became evanescent and the experiment demonstrated perfect free-space isolation at that harmonic (Cardin et al., 2019).

At optical frequencies, an ultrathin nonlinear dielectric metasurface operating at wavelengths around $\phi(\mathbf r)$3 nm used a traveling-wave Kerr-induced phase modulation

$\phi(\mathbf r)$4

on a $\phi(\mathbf r)$5 nm amorphous-silicon active layer above Ag and SiO$\phi(\mathbf r)$6. Heterodyne interference of two $\phi(\mathbf r)$7 nm pump beams generated a traveling intensity pattern with $\phi(\mathbf r)$8 THz and $\phi(\mathbf r)$9. The resulting sidebands obeyed

$\Omega$0

The experiment observed completely asymmetric reflections in forward and backward propagations within a sub-wavelength interaction length of $\Omega$1 nm, with an operational nonreciprocal bandwidth of about $\Omega$2 THz (Guo et al., 2019).

A practical route that avoids deeply subwavelength in-plane modulation is the space-time modulated loaded-wire metagrating. In that architecture, only three wires per supercell are assigned distinct modulation phases $\Omega$3, which synthesizes a traveling modulation across electrically large cells. For incidence at $\Omega$4, with $\Omega$5, $\Omega$6, $\Omega$7 pF, $\Omega$8, and $\Omega$9, the specular order was nearly suppressed and almost all the incident power was redirected into the $K_x=k_0\sin\theta_0$0 order at $K_x=k_0\sin\theta_0$1, reflected at $K_x=k_0\sin\theta_0$2; for $K_x=k_0\sin\theta_0$3, the same structure behaved nearly as a specular reflector (Hadad et al., 2019).

The same design logic extends beyond electromagnetism. An acoustic realization used 24 ultrathin membrane units modulated at $K_x=k_0\sin\theta_0$4 Hz with a $K_x=k_0\sin\theta_0$5 Hz incident wave and a phase gradient $K_x=k_0\sin\theta_0$6. Forward propagation converted $K_x=k_0\sin\theta_0$7 Hz normal incidence into a blue-shifted $K_x=k_0\sin\theta_0$8 Hz beam at $K_x=k_0\sin\theta_0$9, whereas the backward process drove the $\beta_m$0 Hz harmonic to $\beta_m$1, i.e. an evanescent state bound to the surface. The same platform also demonstrated nonreciprocal blue-shift focusing (Chen et al., 2021).

5. Extensions across quantum, thermal, moving-medium, and stacked architectures

The topic has expanded from classical free-space scattering to quantum state control. "Space-Time Quantum Metasurfaces" defines dielectric or plasmonic metasurfaces with

$\beta_m$2

or equivalently time-dependent effective polarizabilities

$\beta_m$3

Within that framework, the traveling synthetic phase $\beta_m$4 biases photon momentum, while rotating phases bias orbital angular momentum. The paper states that the scattering matrix for the dynamical Casimir process becomes asymmetric via the spatio-temporal modulation, reflecting that Lorentz reciprocity is broken at the level of quantum vacuum fluctuations, and it identifies nonreciprocal photon propagation for free-space quantum isolation as a target functionality (Kort-Kamp et al., 2021).

A more specific quantum direction is spatiotemporal photon blockade for nonreciprocal quantum absorption. There the superconducting slab is modulated with a space-time periodic current density, and the critical operating point is $\beta_m$5. At that point, forward-propagating higher-order harmonics cluster at $\beta_m$6, allowing the incoming photon to transfer energy into guided harmonics and be absorbed, while the backward-propagating case lacks the same coherent coupling and remains transmitting (Taravati, 2024).

Thermal photonics provides another major extension. A graphene-based mid-infrared integrated photonic structure modulated at $\beta_m$7 GHz with a three-pixel phase progression $\beta_m$8 experimentally demonstrated nonreciprocal reflection at thermal wavelengths. The theoretical result is the channelwise relation

$\beta_m$9

which ties unequal emissivity and absorptivity directly to nonreciprocal scattering under spatiotemporal modulation and thereby to breakdown of the spectral directional Kirchhoff’s law (Efimov et al., 30 Aug 2025).

Two newer architectural generalizations show how the space-time paradigm is being reformulated. "Stacked Time-Varying Metasurfaces" replaces in-plane traveling modulation by cascades of spatially uniform but temporally modulated sheets. With only two layers and optimized phase-shifted modulation of plasma and collision frequencies, the work reports about $k_{x,n}=k_x+n\beta_m$00 dB isolation at the fundamental frequency and proposes a temporal analogue of circulators (Movahediqomi et al., 28 Aug 2025). "Pulse-driven photonic transitions and nonreciprocity in space-time modulated metasurfaces" shows that a single-period ultrafast traveling pulse can mimic periodic-modulation-driven transitions when metasurface dispersion produces narrow density-of-states peaks; in the reported metasurface examples, this yields isolation ratios of about $k_{x,n}=k_x+n\beta_m$01 dB and $k_{x,n}=k_x+n\beta_m$02 dB, depending on the pulse profile and relaxation regime (Hayran et al., 26 Jan 2026).

6. Design principles, misconceptions, and outlook

Several design rules recur across the literature. First, modulation frequency and modulation wavevector determine the spacing of the accessible Floquet channels; $k_{x,n}=k_x+n\beta_m$03 sets the harmonic ladder in frequency, while $k_{x,n}=k_x+n\beta_m$04 or $k_{x,n}=k_x+n\beta_m$05 sets the momentum ladder and therefore the radiative or evanescent character of each generated order (Gupta et al., 2017, Cardin et al., 2019). Second, operating near a resonant dispersion feature—Lorentzian surface resonance, guided-mode resonance, quasi-BIC, or hyperbolic mode—amplifies harmonic conversion and therefore the nonreciprocal contrast (Gupta et al., 2017, Hayran et al., 26 Jan 2026). Third, modulation phase is not merely a bookkeeping parameter: short-system analyses show that the difference between transmitted phases can be the main contributor to breaking reciprocity, even when transmitted amplitudes are equal. This suggests that nonreciprocal phase-gradient metasurfaces deserve to be treated as a distinct class rather than as a weak form of amplitude isolation (Wu et al., 2024).

Several misconceptions also recur. One is that any asymmetric transmission implies nonreciprocity; the literature repeatedly distinguishes directional asymmetry in static structures from genuine Lorentz-nonreciprocal scattering in time-varying ones (Hadad et al., 2015, Cardin et al., 2019). Another is that magnetless nonreciprocity must rely on material nonlinearity; the susceptibility-, impedance-, conductivity-, and moving-medium formulations all show linear time-variance as an independent route (Gupta et al., 2017, Yang et al., 2024). A third is that in-plane traveling modulation is the only practical architecture. The stacked approach shows that out-of-plane cascades of temporally modulated sheets can realize magnet-free isolation with far fewer time-varying elements, although that comes with different trade-offs in angular control and synthetic-momentum engineering (Movahediqomi et al., 28 Aug 2025).

Implementation remains constrained by modulation depth, synchronization, loss, and bandwidth. At microwave frequencies, synchronized programmable RF bias lines, varactors, and reflectarray or metagrating architectures are already sufficient for dynamic free-space demonstrations (Cardin et al., 2019, Hadad et al., 2019). At terahertz, infrared, and optical frequencies, the recurring material routes are carrier injection, optical pumping, electro-optic modulation, fast graphene conductivity control, transparent conductive oxides, and semiconductor index transients (Guo et al., 2019, Efimov et al., 30 Aug 2025, Hayran et al., 26 Jan 2026). This suggests that the field is evolving from proof-of-principle nonreciprocal scattering toward integrated functions in quantum isolation, thermal radiation control, synthetic-frequency routing, and nonreciprocal wavefront processing.