betaFeedback

WorksEngineering

Space-Time Stacked Intelligent Metasurfaces for Doppler-Resilient LEO Broadcasting

Abstract

Stacked intelligent metasurfaces (SIMs) cascade programmable layers in front of a feed. In the SIM designs reported so far, including those proposed for low-Earth-orbit (LEO) satellites, the layer coefficients are space-only controls or are varied randomly in time to support opportunistic scheduling, so that every radiated beam carries the frequency of the feed. This paper introduces space-time SIMs (ST-SIMs), whose layers are modulated in time and use the temporal dimension of the stack deterministically: each spatial mode of the aperture is paired with its own prescribed frequency translation. The application is LEO broadcasting, where a common waveform reaches terminals with different Doppler shifts, and any common Doppler correction leaves residuals of up to half the Doppler span. We derive the harmonic-domain model of the stack, solve the minimum-energy space-frequency synthesis in closed form, and prove that full column rank of the steering matrix is necessary and sufficient for decoupling K terminal-specific translations, a result that extends to groups of terminals. We further propose a phase-only synthesis algorithm with guaranteed stationarity, together with quantization and control-rate laws. Simulations at 20 GHz over 50 random drops show that a four-layer ST-SIM with a single radio-frequency chain removes the Doppler floor of a static SIM for footprint radii above about 45 km and, under the ideal phase-only model and at equal radiated peak energy, approaches ideal digital per-terminal precompensation within 0.5 dB for resolvable terminals, whereas, in the evaluated geometries, a single time-modulated layer cannot synthesize the required modes.

Introduction

-Earth-orbit (LEO) constellations are a cornerstone of fifth- and sixth-generation non-terrestrial networks (NTNs) owing to their global coverage, moderate latency, and standardized integration with terrestrial radio access [1, 2, 3]. Their orbital speed of about 7.57.5 km/s is also their most persistent impairment, since the satellite–terminal range changes by kilometers per second and the Doppler shift scales with the carrier frequency. At 2020 GHz, the Doppler shift exceeds ±400\pm400 kHz, i.e., more than ten subcarrier spacings of a 30-kHz orthogonal frequency-division multiplexing (OFDM) numerology, and it drifts at rates of several kHz/s [10, 11, 9]. An uncompensated carrier-frequency offset (CFO) of this size destroys subcarrier orthogonality and generates inter-carrier interference (ICI) [16, 17], and Doppler handling permeates every NTN procedure [12, 13, 15]. Since orbital motion is deterministic, the predominant remedy is precompensation, whereby the transmitter, i.e., the gateway for a transparent and the satellite for a regenerative payload, shifts the transmitted carrier by the Doppler predicted from the ephemeris [9, 13, 14]. Such a correction is exact for a single direction. A broadcast beam, however, serves terminals with different line-of-sight (LoS) directions and radial velocities, and, as shown in § 3, the best common translation leaves a worst-terminal residual equal to half the Doppler span, a floor independent of the transmit power, the noise level, and any digital processing of the common waveform. For a 200-km footprint radius at 2020 GHz this span exceeds 300300 kHz, i.e., ten subcarrier spacings, and the residual is left to the terminals [15, 12].

Stacked intelligent metasurfaces (SIMs) cascade LL programmable metasurface layers of NN meta-atoms each in front of a feed, so that inter-layer diffraction implements a high-dimensional transformation of the incident wave [22, 21, 5]. Each layer applies a diagonal phase profile to the field impinging on it, and the free-space propagation between consecutive layers mixes the outputs of all meta-atoms, so that the cascade realizes a dense N×NN\times N linear operator with only LNLN phase controls and without any digital multiplication. The stack can thus perform beamforming, precoding, or a discrete Fourier transform in the wave domain at the speed of light, which reduces the number of RF chains, the baseband load, and the power consumption, while its depth LL and its NN meta-atoms per layer provide degrees of freedom that a single-layer surface lacks. SIMs have been studied for multiuser beamforming [23], channel estimation [27], physically consistent modeling [24], LEO satellite downlinks [26, 4], wideband MIMO-OFDM [28], unmanned aerial vehicles (UAVs) [7, 6], and doubly-dispersive channels [29, 25]. In all of these works, the layer coefficients are space-only control variables: the stack shapes the spatial or frequency-selective response of a beam, but every beam carries the frequency of the feed, and temporal modulation of the layers has not been exploited as a deterministic signal-processing degree of freedom.

State-of-the-Art

Doppler compensation in NTNs has been studied extensively. The 3GPP NTN study [9] relies on ephemeris-assisted precompensation at the network and on CFO estimation from reference signals at the terminal [13], and dedicated solutions have been proposed for random access [12], narrowband Internet-of-Things (IoT) uplinks [14], multibeam initial synchronization [15], and massive multiple-input multiple-output (MIMO) LEO downlinks with per-user Doppler and delay precompensation in the digital domain [11]. Doppler-robust waveforms [18, 19] relocate the problem to the receiver equalizer, and reconfigurable intelligent surfaces (RISs) have mitigated Doppler in high-mobility links [20] as externally illuminated reflectors rather than as the satellite front end.

In parallel, the electromagnetics literature has established that time-varying surfaces manipulate frequency. Time-modulated arrays were used decades ago to shape sidelobes and generate harmonics [31], and a single-layer time-modulated array has been shown to separate downlink users by pre-shifted carriers with one RF chain [47]. Space–time gradient metasurfaces [32] and spacetime metamaterials [33] showed that joint spatial and temporal variation yields frequency conversion and nonreciprocity, and space–time-coding digital metasurfaces [34] demonstrated programmable control of the propagation direction and harmonic power distribution, independent control of harmonic amplitude and phase [35], joint multi-frequency beam shaping and steering [48], Doppler-effect emulation and frequency conversion [37, 38, 39], space- and frequency-division multiplexing [36], sideband-free frequency conversion in a waveguide-fed antenna [49], and frequency conversion with beam shaping on a 1-bit metasurface antenna [40, 41]. Closest to the present work, a single-layer 1-bit spatiotemporal reflecting metasurface on the propagation path offsets the Doppler shift of one high-mobility link by steering one frequency-translated harmonic toward the user, with a switching sequence obtained by a neural network [45]. Stacks of spatially uniform time-varying metasurfaces realize nonreciprocal isolation and frequency circulation [46], and in the SIM literature temporal variation has so far served only to randomize the channel for opportunistic scheduling [30].

Work
layer
varying
shift
per beam
matched
laws
Space-only SIM [22, 23, 26]–––––
Randomized ST-SIM [30]––––
Stacked TV surfaces [46]–––
Time-modulated array [47]–––
STC metasurfaces [34, 48, 49]–––
Doppler-offset STM [45]––1 link–
This work

Table 1. Positioning With Respect to Time-Varying Metasurfaces and SIMs

Table 1 positions this paper. In single-layer time-modulated surfaces, the beam frequencies are harmonics that serve as channels [47, 48, 36, 49], or a single translation offsets the Doppler of one link [45], so that terminals with different Doppler shifts retain half their Doppler span (Proposition 2); in multilayer stacks, the time dimension has served randomization [30] or, with uniform layers, nonreciprocity [46]. The deterministic pairing of distinct spatial modes of a stack with Doppler-matched translations, and the conditions under which a finite-depth, finite-resolution, and finitely fast phase-only stack realizes it, have not been investigated.

Contributions

This paper proposes space-time (ST-) SIMs for deterministic space–frequency processing at the LEO satellite (Figure 1). The spatial modes and the temporal translations cannot be designed independently, since every frequency-translated mode must be spatially nulled at the unintended terminals and the resulting field must be synthesized by a finite-depth, finite-resolution, and finitely fast phase-only stack. The main contributions of our work are:

  • We present the ST-SIM architecture and derive its harmonic-domain model, which shows that stacking convolves the temporal spectra of the layers while mixing their spatial fields, and that desired modes and unwanted sidebands share one passive energy budget.

  • We develop the space–frequency synthesis theory: we solve the minimum-energy multiuser synthesis in closed form with its complete solution set, geometry-dependent energy bounds, and long-run peak energy; we prove that full column rank of the steering matrix is necessary and sufficient for exactly decoupling KK terminal-specific translations, a condition that grouped compensation relaxes; we show that any common translation, hence any static SIM or single-translation time-modulated surface, leaves half the Doppler span; and we quantify in closed form the OFDM interference caused by residual Doppler and ephemeris errors.

  • We propose a realizable design: a phase-only synthesis algorithm with closed-form gain scaling, reusable multilayer gradients, guaranteed monotone decrease to stationary points, and quantization-aware refinement, and we derive explicit quantization and control-rate laws.

  • We demonstrate the architecture in a link-level study at 2020 GHz over 5050 random drops, in which we compare the ST-SIM with a single-layer time-modulated aperture, a static SIM with optimal common precompensation, with the same payload combined with per-terminal CFO correction, and with ideal digital per-terminal precompensation, and we map the operating regimes in footprint radius, aperture size, number of terminals, and grouping.

Figure 1. User-dependent Doppler precompensation in the electromagnetic domain with an ST-SIM. A single feed illuminates an LL-layer stack of time-varying phase-only metasurfaces on a LEO satellite, which transforms the common waveform into KK spatial modes directed toward the KK terminals; mode kk is translated in frequency by −fD,k-f_{D,k} so that terminal kk receives a carrier-Doppler-free signal.

Paper Organization: § 2 presents the system model, § 3 the space–frequency synthesis theory, and § 4 the phase-only synthesis algorithm and its hardware laws. § 5 reports the numerical results, and § 6 concludes the paper.

Notation: Boldface lowercase and uppercase letters denote vectors and matrices. (⋅)T(\cdot)^T, (⋅)H(\cdot)^H, (⋅)∗(\cdot)^*, and (⋅)†(\cdot)^\dagger denote transpose, Hermitian transpose, conjugation, and Moore–Penrose pseudoinverse; ∥⋅∥2\left\lVert \cdot\right\rVert_2 and ∥⋅∥F\left\lVert \cdot\right\rVert_F the Euclidean/spectral and Frobenius norms; λmin⁡(⋅)\lambda_{\min}(\cdot), λmax⁡(⋅)\lambda_{\max}(\cdot) the extreme eigenvalues of a Hermitian matrix; diag⁡(x)\operatorname{diag}(\mathbf{x}) the diagonal matrix built from x\mathbf{x}; IN\mathbf{I}_N the N×NN\times N identity; en\mathbf{e}_n the nnth canonical vector; W2π(⋅)\mathcal W_{2\pi}(\cdot) the entrywise phase wrap to [−π,π)[-\pi,\pi); and sinc⁡(x)=sin⁡(πx)/(πx)\operatorname{sinc}(x)=\sin(\pi x)/(\pi x).

System Model

We consider a 6G multicast–broadcast service in which one LEO beam delivers a common data stream to KK registered terminals whose positions are known to the network from their registration reports.

LEO Propagation and Doppler

Let uu denote transmit time and tt receive time. In an Earth-centered inertial frame, the satellite and terminal kk have positions rs(⋅)\mathbf{r}_s(\cdot) and rk(⋅)\mathbf{r}_k(\cdot), a wavefront emitted at uu is received at t=u+τk(u)t=u+\tau_k(u), and the one-way light time τk(u)\tau_k(u) solves the light-time equation c τk(u)=∥rk(u+τk(u))−rs(u)∥2c\,\tau_k(u)=\left\lVert \mathbf{r}_k(u+\tau_k(u))-\mathbf{r}_s(u)\right\rVert_2 [8]. Let x(u)x(u) be the unit-power complex envelope applied to a common RF feed at carrier fcf_c, ℓk(u)≥0\ell_k(u)\ge0 the slowly varying amplitude gain, ak(u)∈CN\mathbf{a}_k(u)\in\mathbb{C}^N the LoS steering vector of the radiating aperture, and g(u)∈CN\mathbf{g}(u)\in\mathbb{C}^N the aperture field produced by the ST-SIM. After ideal downconversion at fcf_c, the received baseband signal is

yk(t)=ℓk(u) akH(u)g(u) x(u) e−j2πfcτk(u)+nk(t),\begin{equation} y_k(t)=\ell_k(u)\,\mathbf{a}_k^H(u)\mathbf{g}(u)\,x(u)\,e^{-\mathrm{j}2\pi f_c\tau_k(u)}+n_k(t), \tag*{(1)} \end{equation}

with uu the retarded transmit time and nkn_k additive noise. Model (1) does not introduce Doppler as a separate impairment: the carrier Doppler is the time derivative of the delay-induced carrier phase. Appendix § 7 shows that, with uk\mathbf{u}_k the unit line-of-sight (LoS) vector and vkrel=r˙k−r˙s\mathbf{v}_k^{\rm rel}=\dot{\mathbf{r}}_k-\dot{\mathbf{r}}_s the relative velocity, the instantaneous carrier Doppler is fD,k=−(fc/c)ukTvkrelf_{D,k}=-(f_c/c)\mathbf{u}_k^T\mathbf{v}_k^{\rm rel} up to a relative correction below 2.5×10−52.5\times10^{-5} for LEO motion, Over one OFDM block of duration TbT_b centered at t0t_0, the carrier phase is linear in tt up to π∣f˙D,k∣Tb2/4\pi\left\lvert \dot f_{D,k}\right\rvert T_b^2/4, the delay varies by at most ∥vkrel∥2Tb/c\left\lVert \mathbf{v}_k^{\rm rel}\right\rVert_2T_b/c, and the LoS direction rotates by at most ∥vkrel∥2Tb/dk\left\lVert \mathbf{v}_k^{\rm rel}\right\rVert_2T_b/d_k, with dkd_k the range1. Under these conditions, (1) reduces to the blockwise model

yk(t)=αkej2πfD,k(t−t0)akHg(t−τk)x(t−τk)+nk(t),\begin{equation} y_k(t)=\alpha_k e^{\mathrm{j}2\pi f_{D,k}(t-t_0)}\mathbf{a}_k^H\mathbf{g}(t-\tau_k)x(t-\tau_k)+n_k(t), \tag*{(2)} \end{equation}

where αk\alpha_k absorbs the constant phase and path gain. Model (2) is a LoS model; multipath Doppler, phase noise, and oscillator offsets remain for the terminal’s synchronization. The ST-SIM thus performs the coarse synchronization by removing the orbital Doppler of the LoS path, and the terminal performs only the fine synchronization.

Multilayer ST-SIM Transfer Function

The satellite carries an LL-layer ST-SIM with NN meta-atoms per layer, illuminated by one RF feed, which radiates the OFDM signal at carrier fcf_c with bandwidth B=MΔfB=M\Delta f, through a fixed coupling vector f∈CN\mathbf{f}\in\mathbb{C}^N, ∥f∥2=1\left\lVert \mathbf{f}\right\rVert_2=1. The time-varying phase matrix of layer ll is

Φl(u)=diag⁡ ⁣(ejθl,1(u),…,ejθl,N(u)),\begin{equation} \mathbf{\Phi}_l(u)=\operatorname{diag}\!\big(e^{\mathrm{j}\theta_{l,1}(u)},\ldots,e^{\mathrm{j}\theta_{l,N}(u)}\big), \tag*{(3)} \end{equation}

and Wl∈CN×N\mathbf{W}_l\in\mathbb{C}^{N\times N}, l=2,…,Ll=2,\ldots,L, is the fixed propagation operator from layer l−1l-1 to layer ll. The aperture field is

g(u)=G(u)f,G(u)=ΦL(u)WL⋯Φ2(u)W2Φ1(u).\begin{equation} \mathbf{g}(u)=\mathbf{G}(u)\mathbf{f},\qquad \mathbf{G}(u)=\mathbf{\Phi}_L(u)\mathbf{W}_L\cdots\mathbf{\Phi}_2(u)\mathbf{W}_2\mathbf{\Phi}_1(u). \tag*{(4)} \end{equation}

Let νmax⁡=sup⁡k∣fD,k∣\nu_{\max}=\sup_k\left\lvert f_{D,k}\right\rvert and let τstack\tau_{\rm stack} be the transit time across the stack.

Assumption 1 (Frequency-flat inter-layer propagation). νmax⁡/fc≪1\nu_{\max}/f_c\ll1, so a common Wl\mathbf{W}_l applies to all generated sidebands.

Assumption 2 (Negligible modulation during transit). νmax⁡τstack≪1\nu_{\max}\tau_{\rm stack}\ll1, so all layers are evaluated at the same control time uu in (4).

For held control, the modulation spectrum of the layers extends to the update rate fuf_u and its harmonics rather than to νmax⁡\nu_{\max}, so both conditions must hold with νmax⁡\nu_{\max} replaced by the control bandwidth. For a passive stack we assume

∥Wl∥2≤κl≤1,l=2,…,L,\begin{equation} \left\lVert \mathbf{W}_l\right\rVert_2\le\kappa_l\le1,\qquad l=2,\ldots,L, \tag*{(5)} \end{equation}

with equality for lossless propagation; the Wl\mathbf{W}_l are moreover assumed nonsingular, as the Rayleigh–Sommerfeld operators of § 5 are. Equations (3)–(4) constitute the ideal phase-only wave-domain model of the SIM literature [22, 23]: diagonal unit-modulus layers, fixed inter-layer operators, and no multiple reflections, state-dependent or amplitude–phase coupling, or bias-network parasitics. In this model, ∥g(u)∥22\left\lVert \mathbf{g}(u)\right\rVert_2^2 is the energy of the element excitations, which equals the radiated power only up to the radiation matrix of the aperture; the accuracy of the single-pass cascade depends on the inter-layer spacing and on the element design [24, 5], and the transmission efficiency ηL\eta_L is a property of the normalized model rather than of a device.

Harmonic-Domain Representation

Suppose first that each layer is periodic with modulation period Tm=1/fmT_m=1/f_m, so that Φl(u)=∑p∈ZCl[p]ej2πpfmu\mathbf{\Phi}_l(u)=\sum_{p\in\mathbb{Z}}\mathbf{C}_l^{[p]}e^{\mathrm{j}2\pi pf_mu} with diagonal Fourier coefficients Cl[p]=Tm−1∫0TmΦl(u)e−j2πpfmudu\mathbf{C}_l^{[p]}=T_m^{-1}\int_0^{T_m}\mathbf{\Phi}_l(u)e^{-\mathrm{j}2\pi pf_mu}du.

Lemma (Harmonic convolution across the stack). G(u)=∑p∈ZG[p]ej2πpfmu\mathbf{G}(u)=\sum_{p\in\mathbb{Z}}\mathbf{G}^{[p]}e^{\mathrm{j}2\pi pf_mu} with G[p]=∑p1+⋯+pL=pCL[pL]WL⋯C2[p2]W2C1[p1], \mathbf{G}^{[p]}=\sum_{p_1+\cdots+p_L=p}\mathbf{C}_L^{[p_L]}\mathbf{W}_L\cdots\mathbf{C}_2^{[p_2]}\mathbf{W}_2\mathbf{C}_1^{[p_1]}, and the ppth aperture harmonic is g[p]=G[p]f\mathbf{g}^{[p]}=\mathbf{G}^{[p]}\mathbf{f}.

Stacking thus convolves the temporal spectra of the layers while the inter-layer operators mix the associated spatial fields. The reconfiguration capabilities of the meta-atoms enter (6) only through the Cl[p]\mathbf{C}_l^{[p]}: a state held over a reconfiguration period TuT_u weights the harmonics by sinc⁡(pfmTu)\operatorname{sinc}(pf_mT_u) and replicates them at multiples of fu=1/Tuf_u=1/T_u, so that only tones with ∣ν∣<fu/2\left\lvert \nu\right\rvert<f_u/2 can be synthesized (Proposition 7), and a BϕB_\phi-bit reconfiguration accuracy diverts energy into the unwanted harmonics est(u)\mathbf{e}_{\mathrm{st}}(u) (Proposition 6). Lemma (?) also specifies which modal fields the stack can produce. Let the QQ tones to be synthesized, ν1,…,νQ\nu_1,\ldots,\nu_Q, lie on a grid of spacing δf\delta_f (cf. § 4) and set fm=δff_m=\delta_f. Each tone is then a harmonic of the stack, νq=−pqfm\nu_q=-p_qf_m with pq∈Zp_q\in\mathbb{Z}, and the aperture field decomposes as

g(u)=∑q=1Qψqe−j2πνqu+est(u),ψq=g[pq],\begin{equation} \mathbf{g}(u)=\sum_{q=1}^{Q}\boldsymbol{\psi}_qe^{-\mathrm{j}2\pi\nu_qu}+\mathbf{e}_{\mathrm{st}}(u),\quad \boldsymbol{\psi}_q=\mathbf{g}^{[p_q]}, \tag*{(7)} \end{equation}

where est(u)=∑p∉{p1,…,pQ}g[p]ej2πpfmu\mathbf{e}_{\mathrm{st}}(u)=\sum_{p\notin\{p_1,\ldots,p_Q\}}\mathbf{g}^{[p]}e^{\mathrm{j}2\pi pf_mu} collects the unwanted harmonics generated by (6). The number of modes QQ is the number of beams, each of which carries one frequency translation, whether it serves one terminal or several: Q=KQ=K for one beam per terminal and Q=GQ=G for one multicast beam per group in the grouped compensation. Synthesis therefore amounts to concentrating the harmonic content (6) on the indices p1,…,pQp_1,\ldots,p_Q with the prescribed spatial profiles; § 4 performs this in the time domain. For tones that do not share a common grid, the control is quasi-periodic, and (6) holds with the integer indices replaced by the frequencies generated by the layers and index addition replaced by frequency addition.

Lemma 2 (Spectral-energy constraint). Let the inter-layer operators satisfy (5) and let ∥f∥2=1\left\lVert \mathbf{f}\right\rVert_2=1. Then, for every control trajectory {θl(u)}l=1L\{\mathbf{\theta}_l(u)\}_{l=1}^L, the aperture field (4) satisfies ∥g(u)∥2≤1,∀u∈R, \left\lVert \mathbf{g}(u)\right\rVert_2\le1,\qquad \forall u\in\mathbb{R}, with equality for all uu if Wl\mathbf{W}_l is unitary for every l=2,…,Ll=2,\ldots,L. Moreover, if the control is periodic with period TmT_m, the aperture harmonics g[p]\mathbf{g}^{[p]} of Lemma (?) satisfy ∑p∈Z∥g[p]∥22=1Tm∫0Tm∥g(u)∥22 du≤1. \sum_{p\in\mathbb{Z}}\left\lVert \mathbf{g}^{[p]}\right\rVert_2^2=\frac{1}{T_m}\int_0^{T_m}\left\lVert \mathbf{g}(u)\right\rVert_2^2\,du\le1.

Lemma 2 shows that temporal modulation redistributes the aperture energy across spectral modes rather than adding power: ∑q∥ψq∥22+Tm−1∫0Tm∥est(u)∥22du≤1\sum_q\left\lVert \boldsymbol{\psi}_q\right\rVert_2^2+T_m^{-1}\int_0^{T_m}\left\lVert \mathbf{e}_{\mathrm{st}}(u)\right\rVert_2^2du\le1: the desired modes and the unwanted sidebands share the same energy budget. Moreover, a lossless stack has constant energy ∥g(u)∥2=1\left\lVert \mathbf{g}(u)\right\rVert_2=1, whereas a multi-tone field has time-varying energy (Lemma 3); the difference is carried by est(u)\mathbf{e}_{\mathrm{st}}(u) or absorbed by loss.

Space–Frequency Synthesis and Doppler Compensation

Minimum-Energy Multiuser Synthesis

Within one block, collect the steering vectors in A=[a1,…,aK]∈CN×K\mathbf{A}=[\mathbf{a}_1,\ldots,\mathbf{a}_K]\in\mathbb{C}^{N\times K} and let RA≜AHA\mathbf{R}_A\triangleq\mathbf{A}^H\mathbf{A}. To cancel the Doppler of terminal kk, one aperture mode ψk\boldsymbol{\psi}_k is assigned the temporal factor e−j2πfD,kue^{-\mathrm{j}2\pi f_{D,k}u}:

g(u)=ΨdD(u),  dD(u)=[e−j2πfD,1u,…,e−j2πfD,Ku]T ⁣,\begin{equation} \mathbf{g}(u)=\mathbf{\Psi}\mathbf{d}_D(u),\ \ \mathbf{d}_D(u)=\big[e^{-\mathrm{j}2\pi f_{D,1}u},\ldots,e^{-\mathrm{j}2\pi f_{D,K}u}\big]^T\!, \tag*{(10)} \end{equation}

with Ψ=[ψ1,…,ψK]\mathbf{\Psi}=[\boldsymbol{\psi}_1,\ldots,\boldsymbol{\psi}_K]. Terminal kk must observe only mode kk, i.e., akHψj=βkδk,j\mathbf{a}_k^H\boldsymbol{\psi}_j=\beta_k\delta_{k,j}, or

AHΨ=Dβ≜diag⁡(β1,…,βK),\begin{equation} \mathbf{A}^H\mathbf{\Psi}=\mathbf{D}_\beta\triangleq\operatorname{diag}(\beta_1,\ldots,\beta_K), \tag*{(11)} \end{equation}

where βk≠0\beta_k\ne0 prescribes the aperture gain toward terminal kk. Among all decoupling matrices we select the one of minimum modal energy,

P1:min⁡Ψ∈CN×K ∥Ψ∥F2s.t. AHΨ=Dβ,\begin{equation} \mathcal P_1:\quad\min_{\mathbf{\Psi}\in\mathbb{C}^{N\times K}}\ \left\lVert \mathbf{\Psi}\right\rVert_F^2\quad\mathrm{s.t.}\ \mathbf{A}^H\mathbf{\Psi}=\mathbf{D}_\beta, \tag*{(12)} \end{equation}

where ∥Ψ∥F2\left\lVert \mathbf{\Psi}\right\rVert_F^2 is also the long-time average energy of (10) for distinct mode frequencies, because the temporal exponentials are orthogonal in time average.

Proposition 1 (Solution set, minimum-energy solution, and geometry bounds). Let N≥KN\ge K and rank⁡(A)=K\operatorname{rank}(\mathbf{A})=K. Then (12) is feasible, every feasible matrix has the form Ψ=Ψ⋆+PA⊥Z,Ψ⋆=ARA−1Dβ, \mathbf{\Psi}=\mathbf{\Psi}^\star+\mathbf{P}_A^\perp\mathbf{Z},\qquad \mathbf{\Psi}^\star=\mathbf{A}\mathbf{R}_A^{-1}\mathbf{D}_\beta, with Z∈CN×K\mathbf{Z}\in\mathbb{C}^{N\times K} arbitrary and PA⊥=IN−ARA−1AH\mathbf{P}_A^\perp=\mathbf{I}_N-\mathbf{A}\mathbf{R}_A^{-1}\mathbf{A}^H the orthogonal projector onto null⁡(AH)\operatorname{null}(\mathbf{A}^H), and Ψ⋆\mathbf{\Psi}^\star is the unique minimum-Frobenius-norm solution, with qk≜[RA−1]k,kq_k\triangleq[\mathbf{R}_A^{-1}]_{k,k}, ∥Ψ⋆∥F2=tr⁡ ⁣(DβHRA−1Dβ)=∑k=1Kqk∣βk∣2, \left\lVert \mathbf{\Psi}^\star\right\rVert_F^2=\operatorname{tr}\!\big(\mathbf{D}_\beta^H\mathbf{R}_A^{-1}\mathbf{D}_\beta\big)=\sum_{k=1}^Kq_k\left\lvert \beta_k\right\rvert^2, ∥Dβ∥F2λmax⁡(RA)≤∥Ψ⋆∥F2≤∥Dβ∥F2λmin⁡(RA). \frac{\left\lVert \mathbf{D}_\beta\right\rVert_F^2}{\lambda_{\max}(\mathbf{R}_A)}\le\left\lVert \mathbf{\Psi}^\star\right\rVert_F^2\le\frac{\left\lVert \mathbf{D}_\beta\right\rVert_F^2}{\lambda_{\min}(\mathbf{R}_A)}.

Equation (15) gives terminal geometry a direct physical meaning: when two steering vectors become nearly collinear, λmin⁡(RA)→0\lambda_{\min}(\mathbf{R}_A)\to0 and the energy needed to separate their frequency translations grows without bound. Spatial separability and Doppler compensation are therefore fundamentally coupled. Passivity (Lemma 2) constrains the instantaneous rather than the average energy, which motivates the following definition.

Lemma 3 (Long-run peak energy of the modal field). For any Ψ\mathbf{\Psi} and rationally independent fD,kf_{D,k}, the trajectory dD(u)\mathbf{d}_D(u) is dense in the phase torus, so that Ppk(Ψ)≜sup⁡u∥ΨdD(u)∥22=max⁡ϕ∈[0,2π)K∥Ψejϕ∥22 P_{\rm pk}(\mathbf{\Psi})\triangleq\sup_u\left\lVert \mathbf{\Psi}\mathbf{d}_D(u)\right\rVert_2^2=\max_{\boldsymbol{\phi}\in[0,2\pi)^K}\left\lVert \mathbf{\Psi }e^{\mathrm{j}\boldsymbol{\phi}}\right\rVert_2^2 satisfies ∥Ψ∥F2≤Ppk(Ψ)≤(∑k∥ψk∥2)2≤K∥Ψ∥F2\left\lVert \mathbf{\Psi}\right\rVert_F^2\le P_{\rm pk}(\mathbf{\Psi})\le\big(\sum_k\left\lVert \boldsymbol{\psi}_k\right\rVert_2\big)^2\le K\left\lVert \mathbf{\Psi}\right\rVert_F^2, with the long-time average equal to ∥Ψ∥F2\left\lVert \mathbf{\Psi}\right\rVert_F^2.

The phases of βk\beta_k do not change PpkP_{\rm pk}, since the trajectory sweeps every phase combination.

Perfect Blockwise Multiuser Compensation

For arbitrary temporal frequencies ν=[ν1,…,νK]T\mathbf{\nu}=[\nu_1,\ldots,\nu_K]^T let gν(u)=Ψdν(u)\mathbf{g}_{\mathbf{\nu}}(u)=\mathbf{\Psi}\mathbf{d}_{\mathbf{\nu}}(u) with dν(u)=[e−j2πν1u,…,e−j2πνKu]T\mathbf{d}_{\mathbf{\nu}}(u)=[e^{-\mathrm{j}2\pi\nu_1u},\ldots,e^{-\mathrm{j}2\pi\nu_Ku}]^T. Substituting into (2) and absorbing fixed delay phases into the coefficients yields the space–frequency coupling identity

yk(t)−nk(t)αkx(t−τk)=∑j=1KHk,j ej2π(fD,k−νj)tej2πνjτk,\begin{equation} \frac{y_k(t)-n_k(t)}{\alpha_kx(t-\tau_k)}=\sum_{j=1}^KH_{k,j}\,e^{\mathrm{j}2\pi(f_{D,k}-\nu_j)t}e^{\mathrm{j}2\pi\nu_j\tau_k}, \tag*{(17)} \end{equation}

where H≜AHΨ\mathbf{H}\triangleq\mathbf{A}^H\mathbf{\Psi}. The identity factorizes the two design dimensions: Hk,jH_{k,j} is the spatial leakage of temporal mode jj into terminal kk, and fD,k−νjf_{D,k}-\nu_j is the residual temporal rotation of that mode.

Theorem 1 (Perfect blockwise multiuser Doppler compensation). Let βk≠0\beta_k\ne0 for all kk, rank⁡(A)=K\operatorname{rank}(\mathbf{A})=K, and suppose the ST-SIM realizes g⋆(u)=Ψ⋆dD(u)\mathbf{g}^\star(u)=\mathbf{\Psi}^\star\mathbf{d}_D(u) with Ψ⋆\mathbf{\Psi}^\star from (13). Then every terminal has zero residual Doppler and zero inter-mode leakage: yk(t)=αkβkej2πfD,kτkx(t−τk)+nk(t),k=1,…,K. y_k(t)=\alpha_k\beta_ke^{\mathrm{j}2\pi f_{D,k}\tau_k}x(t-\tau_k)+n_k(t),\quad k=1,\ldots,K. Conversely, exact decoupling AHΨ=Dβ\mathbf{A}^H\mathbf{\Psi}=\mathbf{D}_\beta with a nonsingular diagonal Dβ\mathbf{D}_\beta requires rank⁡(A)=K\operatorname{rank}(\mathbf{A})=K; if the fD,kf_{D,k} are distinct, the rank condition is also necessary for (18) to hold for all kk.

Theorem 1 identifies the sufficient aperture field and the rank condition for its existence; whether a finite-depth phase-only stack synthesizes it is addressed in § 4.

Remark 1 (Grouped compensation). The rank condition concerns KK mutually decoupled translations. If the terminals are partitioned into G≤KG\le K groups Kq\mathcal K_q sharing one tone νq\nu_q and one multicast mode ψq\boldsymbol{\psi}_q each, decoupling requires AHΨ=DβE\mathbf{A}^H\mathbf{\Psi}=\mathbf{D}_\beta\mathbf{E} with the group-indicator matrix E∈{0,1}K×G\mathbf{E}\in\{0,1\}^{K\times G}, which is consistent whenever rank⁡(A)=K\operatorname{rank}(\mathbf{A})=K and possibly otherwise (e.g., for collinear terminals in one group), with minimum-norm solution (AH)†DβE(\mathbf{A}^H)^\dagger\mathbf{D}_\beta\mathbf{E}, i.e., ψq=∑k∈Kqψk⋆\boldsymbol{\psi}_q=\sum_{k\in\mathcal K_q}\boldsymbol{\psi}_k^\star under full rank. By Proposition 2 per group, each terminal keeps at most half the Doppler span of its group; G=1G=1 is the common translation and G=KG=K is Theorem 1 (§ 5.5). Terminals should thus be grouped contiguously in Doppler, i.e., by along-track position: a group of Doppler span Δq\Delta_q keeps a residual ϵ=Δq/(2Δf)\epsilon=\Delta_q/(2\Delta f) with interference-limited SIR ≈3/(π2ϵ2)\approx3/(\pi^2\epsilon^2) (Proposition (?)), whereas every additional group costs modal energy (Proposition 1), a trade-off evaluated in § 5.5.

Irreducible Residual of a Common Frequency Translation

A conventional payload with one RF chain and one mixer, a static SIM, or a time-modulated surface that applies one translation to all beams [47, 45], imposes a single translation νc\nu_c on the common waveform, with worst-terminal residual ϵ∞(νc)=max⁡k∣fD,k−νc∣\epsilon_\infty(\nu_c)=\max_k\left\lvert f_{D,k}-\nu_c\right\rvert.

Proposition 2 (Optimal common precompensation). With fDmin⁡=min⁡kfD,kf_D^{\min}=\min_kf_{D,k} and fDmax⁡=max⁡kfD,kf_D^{\max}=\max_kf_{D,k}, the minimax translation and residual are νc⋆=12(fDmax⁡+fDmin⁡),ϵ∞⋆=12(fDmax⁡−fDmin⁡), \nu_c^\star=\tfrac12\big(f_D^{\max}+f_D^{\min}\big),\qquad \epsilon_\infty^\star=\tfrac12\big(f_D^{\max}-f_D^{\min}\big), and the least-squares translation is νcLS=K−1∑kfD,k\nu_c^{\rm LS}=K^{-1}\sum_kf_{D,k}. Both minimize a residual rather than the SINR; the SINR-optimal translation νcSINR=arg⁡max⁡νcmin⁡kγk(νc)\nu_c^{\rm SINR}=\arg\max_{\nu_c}\min_k\gamma_k(\nu_c), a one-dimensional search over [fDmin⁡,fDmax⁡][f_D^{\min},f_D^{\max}], is evaluated in § 5.3.

Whenever the Doppler span is nonzero, no common translation makes every residual vanish; Theorem 1 removes this floor by pairing a different translation with each spatial mode, and the comparison is architectural rather than algorithmic. Since the payload contains a frequency synthesizer, the numerical results apply νc⋆\nu_c^\star to the feed and let the ST-SIM synthesize only the differential tones fD,k−νc⋆f_{D,k}-\nu_c^\star; all results hold for arbitrary tone frequencies, and this hybrid mode reduces the largest synthesized tone from max⁡k∣fD,k∣\max_k\left\lvert f_{D,k}\right\rvert to ϵ∞⋆\epsilon_\infty^\star, by a factor of 1.051.05 in the baseline geometry and more for clustered terminals.

OFDM Consequences and Residual Errors

Consider MM subcarriers with spacing Δf\Delta f and useful duration T=1/ΔfT=1/\Delta f, and let terminal kk have residual frequency error δfk=fD,k−νk\delta f_k=f_{D,k}-\nu_k with normalized value ϵk=δfk/Δf\epsilon_k=\delta f_k/\Delta f. After cyclic-prefix removal, the DFT-domain relation is Yk[p]=hk∑qCp−q(ϵk)Xk[q]+Nk[p]Y_k[p]=h_k\sum_qC_{p-q}(\epsilon_k)X_k[q]+N_k[p] with

Cr(ϵ)=1M∑m=0M−1ej2π(ϵ−r)mM=ejπ(M−1)(ϵ−r)MMsin⁡[π(ϵ−r)]sin⁡[π(ϵ−r)M].\begin{equation} C_r(\epsilon)=\frac1M\sum_{m=0}^{M-1}e^{\mathrm{j}2\pi(\epsilon-r)\frac mM} =\frac{e^{\mathrm{j}\pi\frac{(M-1)(\epsilon-r)}{M}}}{M}\frac{\sin[\pi(\epsilon-r)]}{\sin[\frac{\pi(\epsilon-r)}M]}. \tag*{(20)} \end{equation}

The CFO matrix is unitary, so ∑r∣Cr(ϵ)∣2=1\sum_r\left\lvert C_r(\epsilon)\right\rvert^2=1, and under Theorem 1 ϵk=0\epsilon_k=0, C0(0)=1C_0(0)=1, and the carrier-Doppler-induced ICI vanishes for every terminal. For imperfect compensation, the fraction of desired-mode power leaving the intended subcarrier is

χM(ϵ)≜1−∣C0(ϵ)∣2=1−∣sin⁡(πϵ)Msin⁡(πϵ/M)∣2.\begin{equation} \chi_M(\epsilon)\triangleq1-\left\lvert C_0(\epsilon)\right\rvert^2=1-\left|\frac{\sin(\pi\epsilon)}{M\sin(\pi\epsilon/M)}\right|^2. \tag*{(21)} \end{equation}

Proposition (Local ICI law). As ϵ→0\epsilon\to0, χM(ϵ)=π23(1−1M2)ϵ2+O(ϵ4)\chi_M(\epsilon)=\frac{\pi^2}{3}\big(1-\frac1{M^2}\big)\epsilon^2+\mathcal O(\epsilon^4).

Since the common-compensation residual scales linearly with fcf_c and the ICI fraction quadratically with the residual normalized by Δf\Delta f, differential Doppler becomes critical as fc/Δff_c/\Delta f grows (§ 5.3). Two residuals remain after carrier-Doppler compensation. First, the envelope in (1) is time-scaled by 1+αk1+\alpha_k, αk≈fD,k/fc\alpha_k\approx f_{D,k}/f_c, which leaves the band-averaged ICI fraction χˉk≈(π2/36)(MfD,k/fc)2\bar\chi_k\approx(\pi^2/36)(Mf_{D,k}/f_c)^2, i.e., −47-47 and −38-38 dB at 160160 and 450450 kHz; this sampling-clock-offset effect is common to all carrier-Doppler schemes and is removed by the timing tracking of an OFDM receiver. Second, ephemeris and terminal-position errors leave a residual tone. With the first-order model fD,k=−(fc/c)ukTvkrelf_{D,k}=-(f_c/c)\mathbf{u}_k^T\mathbf{v}_k^{\rm rel}, suppose the controller uses ρ^k=ρk+δρk\widehat{\mathbf{\rho}}_k=\mathbf{\rho}_k+\delta\mathbf{\rho}_k and v^krel=vkrel+δvkrel\widehat{\mathbf{v}}_k^{\rm rel}=\mathbf{v}_k^{\rm rel}+\delta\mathbf{v}_k^{\rm rel} to predict f^D,k\widehat f_{D,k}, with ρk=dkuk\mathbf{\rho}_k=d_k\mathbf{u}_k the range vector, and let Pk⊥=I3−ukukT\mathbf{P}_k^\perp=\mathbf{I}_3-\mathbf{u}_k\mathbf{u}_k^T.

Proposition (Residual Doppler under state errors). If ∥δρk∥2<dk\left\lVert \delta\mathbf{\rho}_k\right\rVert_2<d_k, then ∣f^D,k−fD,k∣≤fcc[∥δvkrel∥2+2∥vkrel∥2∥δρk∥2dk−∥δρk∥2], \left\lvert \widehat f_{D,k}-f_{D,k}\right\rvert\le\frac{f_c}{c}\left[\left\lVert \delta\mathbf{v}_k^{\rm rel}\right\rVert_2+\frac{2\left\lVert \mathbf{v}_k^{\rm rel}\right\rVert_2\left\lVert \delta\mathbf{\rho}_k\right\rVert_2}{d_k-\left\lVert \delta\mathbf{\rho}_k\right\rVert_2}\right], and, to first order, δfD,k≜f^D,k−fD,k=−fcc[ukTδvkrel+dk−1(vkrel)TPk⊥δρk]\delta f_{D,k}\triangleq\widehat f_{D,k}-f_{D,k}=-\frac{f_c}{c}\big[\mathbf{u}_k^T\delta\mathbf{v}_k^{\rm rel}+d_k^{-1}(\mathbf{v}_k^{\rm rel})^T\mathbf{P}_k^\perp\delta\mathbf{\rho}_k\big].

If the ST-SIM applies νk=f^D,k\nu_k=\widehat f_{D,k}, the residual CFO is −δfD,k-\delta f_{D,k} and, by Proposition (?), the ICI fraction is π23(1−M−2)(δfD,k/Δf)2\frac{\pi^2}{3}(1-M^{-2})(\delta f_{D,k}/\Delta f)^2, which exposes the critical factor fc/Δff_c/\Delta f of high-frequency LEO OFDM links. Only the transverse component of a position error matters, scaled by ∥vrel∥2/dk\left\lVert \mathbf{v}^{\rm rel}\right\rVert_2/d_k.

Phase-Only ST-SIM Synthesis

The ideal field of Theorem 1 need not belong to the set of fields realizable by a finite-depth phase-only stack. This section formulates the physical synthesis problem and derives an efficient first-order method with explicit guarantees.

Trajectory Synthesis Problem

Sample a predictable control horizon at transmit times ur=u0+rTuu_r=u_0+rT_u, r=0,…,R−1r=0,\ldots,R-1, where fu=1/Tuf_u=1/T_u is the control update rate, and let gr⋆\mathbf{g}_r^\star be the desired field generated by (10); since Ψ⋆\mathbf{\Psi}^\star has full column rank and dD(u)≠0\mathbf{d}_D(u)\ne\mathbf{0}, ∥gr⋆∥2>0\left\lVert \mathbf{g}_r^\star\right\rVert_2>0 for every rr. Let θ[r]∈RLN\mathbf{\theta}[r]\in\mathbb{R}^{LN} collect all layer phases at slot rr and Θ=[θ[0],…,θ[R−1]]\Theta=[\mathbf{\theta}[0],\ldots,\mathbf{\theta}[R-1]]. A passive stack cannot reach an arbitrary absolute amplitude (Lemma 2), so a common complex scaling α\alpha, which absorbs the transmission gain of the stack, is optimized jointly with the phases. The finite-resolution design problem is

P2: min⁡Θ,α J(Θ,α)=1R∑r=0R−1ωr∥αG(θ[r])f−gr⋆∥22∥gr⋆∥22+λsS(Θ)s.t. θl,n[r]∈QBϕ,\begin{equation} \begin{aligned} \mathcal P_2:\ \min_{\Theta,\alpha}\ &\mathcal J(\Theta,\alpha)=\frac1R\sum_{r=0}^{R-1}\omega_r\frac{\left\lVert \alpha\mathbf{G}(\mathbf{\theta}[r])\mathbf{f}-\mathbf{g}_r^\star\right\rVert_2^2}{\left\lVert \mathbf{g}_r^\star\right\rVert_2^2}\\ &\qquad\qquad+\lambda_s\mathcal S(\Theta)\\ \mathrm{s.t.}\ &\theta_{l,n}[r]\in\mathcal Q_{B_\phi}, \end{aligned} \tag*{(23)} \end{equation}

where QBϕ={2πq/2Bϕ}q=02Bϕ−1\mathcal Q_{B_\phi}=\{2\pi q/2^{B_\phi}\}_{q=0}^{2^{B_\phi}-1} is the BϕB_\phi-bit phase alphabet, ωr>0\omega_r>0 are trajectory weights, and

S(Θ)=1R−1∑r=1R−1∑i=1LN[1−cos⁡(θi[r]−θi[r−1])]\begin{equation} \mathcal S(\Theta)=\frac1{R-1}\sum_{r=1}^{R-1}\sum_{i=1}^{LN}\Big[1-\cos\big(\theta_i[r]-\theta_i[r-1]\big)\Big] \tag*{(24)} \end{equation}

penalizes switching between consecutive slots. The penalty (24) is 2π2\pi-periodic and smooth, unlike a penalty on wrapped phase differences, so it does not affect the differentiability assumed below; its gradient is ∂S/∂θi[r]=1R−1[sin⁡(θi[r]−θi[r−1])−sin⁡(θi[r+1]−θi[r])]\partial\mathcal S/\partial\theta_i[r]=\frac{1}{R-1}[\sin(\theta_i[r]-\theta_i[r-1])-\sin(\theta_i[r+1]-\theta_i[r])]. A hard slew limit, if required by the drivers, can be imposed afterwards by projecting each increment onto [−Δmax⁡,Δmax⁡][-\Delta_{\max},\Delta_{\max}]. Problem (23) is nonconvex because of the cascaded phase products and the discrete alphabet. We study its continuous relaxation, then quantify the effect of quantization.

Optimal Scaling and Reusable Multilayer Gradient

For one slot (index rr suppressed), define the field incident on layer ll and the transfer from the output of layer ll to the aperture, with u1=f\mathbf{u}_1=\mathbf{f},

ul=WlΦl−1ul−1,  Vl={IN,l=L,ΦLWL⋯Φl+1Wl+1,l<L,\begin{equation} \mathbf{u}_l=\mathbf{W}_l\mathbf{\Phi}_{l-1}\mathbf{u}_{l-1},\ \ \mathbf{V}_l=\begin{cases}\mathbf{I}_N,&l=L,\\ \mathbf{\Phi}_L\mathbf{W}_L\cdots\mathbf{\Phi}_{l+1}\mathbf{W}_{l+1},&l<L,\end{cases} \tag*{(25)} \end{equation}

so that g=VlΦlul\mathbf{g}=\mathbf{V}_l\mathbf{\Phi}_l\mathbf{u}_l for every ll.

Lemma (Optimal scaling and phase derivatives). For fixed Θ\Theta, with gr=G(θ[r])f\mathbf{g}_r=\mathbf{G}(\mathbf{\theta}[r])\mathbf{f} and wr=ωr/∥gr⋆∥22w_r=\omega_r/\left\lVert \mathbf{g}_r^\star\right\rVert_2^2, the minimizer of J\mathcal J over α\alpha is α⋆(Θ)=∑rwrgrHgr⋆∑rwr∥gr∥22, \alpha^\star(\Theta)=\frac{\sum_rw_r\mathbf{g}_r^H\mathbf{g}_r^\star}{\sum_rw_r\left\lVert \mathbf{g}_r\right\rVert_2^2}, and J~(Θ)≜J(Θ,α⋆(Θ))\tilde{\mathcal J}(\Theta)\triangleq\mathcal J(\Theta,\alpha^\star(\Theta)) has gradient ∇ΘJ~=∇ΘJ∣α=α⋆\nabla_\Theta\tilde{\mathcal J}=\nabla_\Theta\mathcal J|_{\alpha=\alpha^\star}. For meta-atom nn of layer ll, ∂g/∂θl,n=jejθl,n[ul]nVlen\partial\mathbf{g}/\partial\theta_{l,n}=\mathrm{j}e^{\mathrm{j}\theta_{l,n}}[\mathbf{u}_l]_n\mathbf{V}_l\mathbf{e}_n, and for F(θ)=∥αg(θ)−g⋆∥22F(\mathbf{\theta})=\left\lVert \alpha\mathbf{g}(\mathbf{\theta})-\mathbf{g}^\star\right\rVert_2^2, ∂F∂θl,n=2ℜ ⁣{jejθl,n[ul]n [VlHε]n∗},ε≜α∗(αg−g⋆). \frac{\partial F}{\partial\theta_{l,n}}=2\Re\!\left\{\mathrm{j}e^{\mathrm{j}\theta_{l,n}}[\mathbf{u}_l]_n\,[\mathbf{V}_l^H\mathbf{\varepsilon}]_n^*\right\},\quad \mathbf{\varepsilon}\triangleq\alpha^*(\alpha\mathbf{g}-\mathbf{g}^\star).

All LNLN derivatives of a slot follow from one forward sweep for {ul}\{\mathbf{u}_l\} and one backward sweep for {VlHε}\{\mathbf{V}_l^H\mathbf{\varepsilon}\}, i.e., 2(L−1)2(L-1) matrix–vector products of size NN, without differentiating the full cascade for every meta-atom.

Algorithm and Convergence

Algorithm alg:synthesis summarizes the proposed synthesis. Step 1 (lines 2–8) performs wrapped gradient descent on the continuous relaxation, where wrapping only changes the coordinate representation of a point on the phase torus; the trial step is the Barzilai–Borwein estimate of the inverse local curvature [43], safeguarded to [μmin⁡,μmax⁡][\mu_{\min},\mu_{\max}] and then reduced by Armijo backtracking until sufficient decrease is achieved. Step 2 (line 9) projects onto QBϕ\mathcal Q_{B_\phi}, and Step 3 (lines 10–12) performs quantization-aware coordinate refinement, in which each meta-atom is assigned the best of its 2Bϕ2^{B_\phi} states given all others, using Vl\mathbf{V}_l and ul\mathbf{u}_l so that each trial costs O(N)\mathcal O(N). With α⋆\alpha^\star fixed, the coordinate cost of θl,n[r]\theta_{l,n}[r] collects all terms of J\mathcal J that depend on it,

Jl,n,r(ϕ)=wrR∥α⋆Vl(Φlul+Δϕ[ul]nen)−gr⋆∥22+λsR−1[2−cos⁡(ϕ−θ^−)−cos⁡(θ^+−ϕ)],\begin{equation} \begin{aligned} \mathcal J_{l,n,r}(\phi)&=\tfrac{w_r}{R}\big\|\alpha^\star\mathbf{V}_l(\mathbf{\Phi}_l\mathbf{u}_l+\Delta_\phi[\mathbf{u}_l]_n\mathbf{e}_n)-\mathbf{g}_r^\star\big\|_2^2\\ &\quad+\tfrac{\lambda_s}{R-1}\big[2-\cos(\phi-\widehat\theta^-)-\cos(\widehat\theta^+-\phi)\big], \end{aligned} \tag*{(28)} \end{equation}

where Δϕ=ejϕ−ejθ^l,n[r]\Delta_\phi=e^{\mathrm{j}\phi}-e^{\mathrm{j}\widehat\theta_{l,n}[r]}, θ^±=θ^l,n[r±1]\widehat\theta^\pm=\widehat\theta_{l,n}[r\pm1], and the switching terms are omitted at r=0r=0 and r=R−1r=R-1. For λs=0\lambda_s=0 the slots decouple and are updated in parallel; for λs>0\lambda_s>0 the even and the odd slots are updated in alternate half-sweeps, so that every update sees fixed neighbors.

\caption{Phase-only ST-SIM trajectory synthesis}
\label{alg:synthesis}
\begin{algorithmic}[1]
\REQUIRE $\{\mathbf{g}_r^\star\}_{r=0}^{R-1}$, $\{\mathbf{W}_l\}$, $\mathbf{f}$, $B_\phi$, $0<c<1$, $0<\rho<1$, $I_{\max}$, $n_{\rm ref}$
\STATE Initialize $\Theta^{(0)}$ uniformly at random on $[0,2\pi)^{LN\times R}$
\FOR{$i=0,1,\ldots,I_{\max}-1$}
\STATE Forward sweeps: $\{\mathbf{u}_l[r]\}$, $\mathbf{g}_r$, and $\alpha^\star$ by \eqref{eq:alphastar}
\STATE Backward sweeps: $\mathbf{\Gamma}_i\triangleq\nabla\tilde{\mathcal J}(\Theta^{(i)})$ by \eqref{eq:objectivegradient}
\STATE Trial step $\mu\gets\langle\mathbf{S}_i,\mathbf{S}_i\rangle/\langle\mathbf{S}_i,\mathbf{\Gamma}_i-\mathbf{\Gamma}_{i-1}\rangle$, $\mathbf{S}_i=\mathcal W_{2\pi}(\Theta^{(i)}-\Theta^{(i-1)})$, clipped to $[\mu_{\min},\mu_{\max}]$
\STATE \textbf{while} $\tilde{\mathcal J}(\mathcal W_{2\pi}(\Theta^{(i)}-\mu\mathbf{\Gamma}_i))>\tilde{\mathcal J}(\Theta^{(i)})-c\mu\left\lVert \mathbf{\Gamma}_i\right\rVert_F^2$ \textbf{do} $\mu\gets\rho\mu$
\STATE $\mu_i\gets\mu$;\ $\Theta^{(i+1)}\gets\mathcal W_{2\pi}(\Theta^{(i)}-\mu_i\mathbf{\Gamma}_i)$
\ENDFOR
\STATE Quantize: $\widehat\Theta\gets\arg\min_{\Theta\in\mathcal Q_{B_\phi}^{LN\times R}}\left\lVert \mathcal W_{2\pi}(\Theta-\Theta^{(I_{\max})})\right\rVert_F$
\FOR{$n_{\rm ref}$ rounds, every slot $r$, layer $l$, meta-atom $n$}
\STATE $\widehat\theta_{l,n}[r]\gets\arg\min_{\phi\in\mathcal Q_{B_\phi}}\mathcal J_{l,n,r}(\phi)$ in \eqref{eq:coordcost} \hfill (even $r$, then odd $r$)
\ENDFOR
\ENSURE $\widehat\Theta$, $\alpha^\star$
\end{algorithmic}

Proposition 5 (Stationarity of Armijo descent). Let J~=J(⋅,α⋆(⋅))\tilde{\mathcal J}=\mathcal J(\cdot,\alpha^\star(\cdot)) with λs≥0\lambda_s\ge0 and the smooth penalty (24), which is continuously differentiable with Lipschitz gradient on the compact phase torus (the data term is a smooth function of unit-modulus entries and ∑r∥gr∥22>0\sum_r\left\lVert \mathbf{g}_r\right\rVert_2^2>0 since f≠0\mathbf{f}\ne\mathbf{0} and the Wl\mathbf{W}_l are nonsingular), and let {Θ(i)}\{\Theta^{(i)}\} be generated by Step 1 of Algorithm alg:synthesis with trial steps in [μmin⁡,μmax⁡][\mu_{\min},\mu_{\max}], 0<μmin⁡≤μmax⁡<∞0<\mu_{\min}\le\mu_{\max}<\infty. Then J~(Θ(i+1))≤J~(Θ(i))\tilde{\mathcal J}(\Theta^{(i+1)})\le\tilde{\mathcal J}(\Theta^{(i)}), ∑iμi∥∇J~(Θ(i))∥F2<∞\sum_i\mu_i\left\lVert \nabla\tilde{\mathcal J}(\Theta^{(i)})\right\rVert_F^2<\infty, and every accumulation point is first-order stationary; in particular lim⁡i∥∇J~(Θ(i))∥F=0\lim_i\left\lVert \nabla\tilde{\mathcal J}(\Theta^{(i)})\right\rVert_F=0 whenever the accepted step sizes are bounded away from zero along nonstationary subsequences. Every update of Step 3 is nonincreasing in J(⋅,α⋆)\mathcal J(\cdot,\alpha^\star) for any λs≥0\lambda_s\ge0.

The trial step is well defined for every curvature estimate: when ⟨Si,Γi−Γi−1⟩≤0\langle\mathbf{S}_i,\mathbf{\Gamma}_i-\mathbf{\Gamma}_{i-1}\rangle\le0, the Barzilai–Borwein quotient is replaced by μmax⁡\mu_{\max}, which the backtracking then reduces until sufficient decrease holds, so the guarantee of the proposition is unaffected. In the baseline geometry of § 5, ∥∇J~∥F\left\lVert \nabla\tilde{\mathcal J}\right\rVert_F after 300300 iterations is 2.5×10−52.5\times10^{-5}, 7.5×10−67.5\times10^{-6}, 2.6×10−62.6\times10^{-6}, and 9.2×10−69.2\times10^{-6} for L=2L=2, 33, 44, and 66, and the single layer meets the tolerance 10−610^{-6} after 142142 iterations: its stall at −5.9-5.9 dB is a stationary point. The proposition guarantees monotone decrease and convergence to first-order stationary points, without global optimality, for the continuous relaxation of P2\mathcal P_2; the discrete alphabet is handled by Steps 2–3. Each iteration and backtracking trial cost O(RLN2)\mathcal O(RLN^2).

Finite-Resolution Phase Quantization

Let Φ^l\widehat{\mathbf{\Phi}}_l denote the nearest-neighbor BϕB_\phi-bit quantization of Φl\mathbf{\Phi}_l and G^\widehat{\mathbf{G}} the corresponding stack matrix, and define δBϕ≜2sin⁡(π/2Bϕ+1)\delta_{B_\phi}\triangleq2\sin(\pi/2^{B_\phi+1}), so that ∥Φ^l−Φl∥2≤δBϕ\left\lVert \widehat{\mathbf{\Phi}}_l-\mathbf{\Phi}_l\right\rVert_2\le\delta_{B_\phi}.

Proposition 6 (Multilayer quantization scaling law). With κl=∥Wl∥2\kappa_l=\left\lVert \mathbf{W}_l\right\rVert_2, ∥(G^−G)f∥2≤L δBϕ(∏l=2Lκl)∥f∥2≤L δBϕ∥f∥2, \left\lVert (\widehat{\mathbf{G}}-\mathbf{G})\mathbf{f}\right\rVert_2\le L\,\delta_{B_\phi}\Big(\prod_{l=2}^L\kappa_l\Big)\left\lVert \mathbf{f}\right\rVert_2\le L\,\delta_{B_\phi}\left\lVert \mathbf{f}\right\rVert_2, where the last inequality holds for a passive stack.

The bound ignores cancellation among layers and is loose for coarse quantization, where it exceeds the norm of the field itself (LδBϕ>1L\delta_{B_\phi}>1 for Bϕ≤4B_\phi\le4 with L=4L=4); it is to be read as a scaling law: for fixed depth the field error decays as O(2−Bϕ)\mathcal O(2^{-B_\phi}), i.e., the synthesis error J\mathcal J, which is a squared norm, as O(2−2Bϕ)\mathcal O(2^{-2B_\phi}) or 66 dB per bit, and for fixed resolution its worst-case accumulation is at most linear in LL (Figure 2b).

Control Rate, Switching Harmonics, and Control Bandwidth

If the desired field contains temporal frequencies in [−νmax⁡,νmax⁡][-\nu_{\max},\nu_{\max}] and the control is interpolated ideally, sampling theory requires fu>2νmax⁡f_u>2\nu_{\max}. Practical digital-coding metasurfaces, however, hold each state constant over TuT_u, which generates switching harmonics.

Proposition 7 (Control-rate law for held control). Let the aperture field be held constant over each control slot, gh(u)=g⋆(ur)\mathbf{g}_h(u)=\mathbf{g}^\star(u_r) for u∈[ur,ur+Tu)u\in[u_r,u_r+T_u), and let the ideal field toward terminal kk be the tone βke−j2πfD,ku\beta_ke^{-\mathrm{j}2\pi f_{D,k}u}. Then the held field toward terminal kk contains the tone at −fD,k-f_{D,k} with amplitude βksinc⁡(fD,k/fu)\beta_k\operatorname{sinc}(f_{D,k}/f_u) and images at −fD,k−pfu-f_{D,k}-pf_u, p∈Z∖{0}p\in\mathbb{Z}\setminus\{0\}, with amplitudes βksinc⁡((fD,k+pfu)/fu)\beta_k\operatorname{sinc}\big((f_{D,k}+pf_u)/f_u\big). The energy fraction diverted into the images is 1−sinc⁡2 ⁣(fD,k/fu)=π23(fD,kfu)2+O((fD,k/fu)4), 1-\operatorname{sinc}^2\!\big(f_{D,k}/f_u\big)=\frac{\pi^2}{3}\Big(\frac{f_{D,k}}{f_u}\Big)^2+\mathcal O\big((f_{D,k}/f_u)^4\big), so that, when the images fall inside the signal band, the self-interference ratio of terminal kk is capped at SIRkZOH=sinc⁡2(fD,k/fu)/[1−sinc⁡2(fD,k/fu)]≈3fu2/(π2fD,k2)\mathrm{SIR}_k^{\rm ZOH}=\operatorname{sinc}^2(f_{D,k}/f_u)/[1-\operatorname{sinc}^2(f_{D,k}/f_u)]\approx3f_u^2/(\pi^2f_{D,k}^2).

Equation (30) has the same rectangular-window origin as Proposition (?) and yields an explicit hardware requirement: an in-band SIR of 30 dB for a synthesized tone of 160160 kHz requires fu≳9f_u\gtrsim9 MHz, and when fuf_u exceeds the signal bandwidth the images fall out of band, where they are an emission rather than an interference problem (Figure 2c).

The binding hardware constraint is not the MHz-rate switching of single elements, which is established [34, 36, 40], but the aggregate control bandwidth: streaming every state requires LNBϕfuLNB_\phi f_u bit/s, i.e., 9.29.2 Gbit/s for LN=400LN=400, Bϕ=3B_\phi=3, and fu=7.68f_u=7.68 MHz. If the tones are rounded to a grid δf\delta_f, the target is periodic with period 1/δf1/\delta_f, and each meta-atom can replay a local table of fu/δff_u/\delta_f states, provided that one full period is synthesized offline, since a nonconvex synthesis does not return a periodic trajectory by itself; for δf=100\delta_f=100 Hz and fu=7.68f_u=7.68 MHz, a period comprises 76 80076\,800 slots (3.1×1073.1\times10^7 phases for LN=400LN=400), which is tractable because (23) couples the slots only through the switching penalty and the gain α\alpha: synthesized in parallel blocks of R=274R=274 slots with fixed α\alpha and overlapping boundaries (44 s per block on one core), the period costs about 2020 core-minutes per pass geometry. The tables are refreshed when a rounded tone has drifted by δf/2\delta_f/2, which reduces the streaming rate by δf2/(2max⁡k∣f˙D,k−ν˙c⋆∣)\delta_f^2/(2\max_k\left\lvert \dot f_{D,k}-\dot\nu_c^\star\right\rvert), i.e., with the drift of at most 1.51.5 kHz/s of § 5, by 3.3×3.3\times for δf=100\delta_f=100 Hz and 30×30\times for δf=300\delta_f=300 Hz (residual below 0.005Δf0.005\Delta f).

Numerical Results

In this section, we extensively investigate the performance of the proposed ST-SIM architecture.

Setup, Benchmarks, and Metrics

Table 2 lists the baseline parameters. At t=0t=0, the four baseline terminals at ground offsets (−180,60)(-180,60), (−40,−130)(-40,-130), (90,140)(90,140), and (200,−30)(200,-30) km (along-, cross-track) see Doppler shifts of −143.7-143.7, −32.7-32.7, 72.772.7, and 158.6158.6 kHz. Geometry follows § 2 with static terminals in the free-space LoS channel of (1), with steering vectors and Doppler shifts evaluated at the actual directions and slant ranges; the path gain ℓk\ell_k is absorbed into SNR0\mathrm{SNR}_0 and taken equal for all terminals, since the slant range grows from 600600 km at the beam center to 677677 km at 300300 km from it, i.e., by at most 1.11.1 dB of free-space loss; Wl\mathbf{W}_l and f\mathbf{f} follow the Rayleigh–Sommerfeld model of [22, 42] with ∥Wl∥2=1\left\lVert \mathbf{W}_l\right\rVert_2=1. The payload broadcasts one common waveform, and the control trajectory is synthesized over the OFDM symbol including its cyclic prefix (7272 samples) and held over Tu=4T_u=4 samples, so switching harmonics are included; the ST-SIM synthesizes the differential tones fD,k−νc⋆f_{D,k}-\nu_c^\star. Algorithm alg:synthesis uses [μmin⁡,μmax⁡]=[10−2,105][\mu_{\min},\mu_{\max}]=[10^{-2},10^{5}], λs=0\lambda_s=0, a gradient tolerance of 10−610^{-6}, nref=2n_{\rm ref}=2, and Imax⁡=300I_{\max}= 300. Figures show medians over 5050 random drops unless stated.

ParameterValueParameterValue
Carrier fcf_c20 GHzAltitude hh600 km
Terminals KK4Footprint radius ρf\rho_f200 km
Layers LL4Meta-atoms/layer NN10×1010\times10
Element spacingλ/2\lambda/2Layer / feed spacingλ\lambda / 2λ2\lambda
Subcarriers MM1024Spacing Δf\Delta f / CP30 kHz / 72 samples
Control rate fuf_u7.68 MHzSlots per symbol RR274
Phase bits BϕB_\phi1–3, ∞\inftyArmijo (c,ρ)(c,\rho)(10−4,0.5)(10^{-4},0.5)
Random drops50Feed SNR SNR0\mathrm{SNR}_025 dB

Table 2. Baseline Simulation Parameters

(a) Convergence of Algorithm~alg:synthesis

(b) Synthesis error vs.\ phase resolution

(c) In-band SIR and out-of-band power vs.\ control rate

Figure 2. Realizability (K=4K=4, L=4L=4, N=100N=100, R=274R=274 unless stated). (a) Objective of Step 1 of Algorithm alg:synthesis for L=1,2,3,4,6L=1,2,3,4,6 (three initializations each), fixed-step, Adam, and layer-wise descent for L=4L=4. (b) Synthesis error with quantization-aware refinement (L=1,2,3,4,6L=1,2,3,4,6) and projection only (L=4L=4), with the scaling law of Proposition 6. (c) In-band SIR (left) and out-of-band power (right) versus fuf_u, with the law of Proposition 7.

Three benchmarks are considered. Benchmark 1 is a space-only SIM of the same geometry [23, 26] synthesizing the max–min multicast beam b⋆=arg⁡max⁡∥b∥2=1min⁡k∣akHb∣2\mathbf{b}^\star=\arg\max_{\left\lVert \mathbf{b}\right\rVert_2=1}\min_k\left\lvert \mathbf{a}_k^H\mathbf{b}\right\rvert^2, computed by multi-start weighted power iteration, which is not certified to be globally optimal but is never worse than the sum Ψ⋆1K\mathbf{\Psi}^\star\mathbf 1_K of the decoupled modes, with its feed shifted by νc⋆\nu_c^\star; it represents every single-translation front end (Proposition 2). Benchmark 2, ideal digital per-terminal precompensation, applies one digital translation per terminal and a lossless zero-forcing precoder, i.e., radiates Ψ⋆dD(u)\mathbf{\Psi}^\star\mathbf{d}_D(u) exactly; it requires KK RF chains and a lossless N×KN\times K multibeam network and upper-bounds every exact realization of Theorem 1. Benchmark 3 is the payload of Benchmark 1 with every terminal correcting its residual CFO, including the error of Moose’s two-symbol estimator, var(ϵ^k)=1/(4π2M SNRk)\mathrm{var}(\hat\epsilon_k)=1/(4\pi^2M\,\mathrm{SNR}_k) [17], and, where stated, two pilot symbols per 14-symbol slot. The no-compensation reference is the space-only SIM without frequency correction. Finally, a single-layer time-modulated aperture is also investigated, g(u)=Φ1(u)f\mathbf{g}(u)=\mathbf{\Phi}_1(u)\mathbf{f}, with the same feed, phase resolution, control rate, and synthesis objective as the ST-SIM (the one-layer case of Algorithm alg:synthesis), isolating the contribution of stacking.

The performance metrics are obtained as follows. The aperture field is evaluated on a four-times oversampled grid, Doppler-shifted, band-limited to ∣f∣≤MΔf/2\left\lvert f\right\rvert\le M\Delta f/2 by an ideal receiver filter, and sampled; for terminal kk, the response sk[m]=ej2πfD,kmTsakHg^[m]s_k[m]=e^{\mathrm{j}2\pi f_{D,k}mT_s}\mathbf{a}_k^H\widehat{\mathbf{g}}[m] over the DFT window gives the ICI matrix Fdiag⁡(sk)FH\mathbf{F}\operatorname{diag}(\mathbf{s}_k)\mathbf{F}^H, whose diagonal equals sk‾\overline{s_k}; hence Pd,k=∣sk‾∣2P_{d,k}=\left\lvert \overline{s_k}\right\rvert^2, and Pi,k=∣sk∣2‾−Pd,kP_{i,k}=\overline{\left\lvert s_k\right\rvert^2}-P_{d,k} is the average interference for i.i.d. data, which includes switching images, quantization noise, and inter-mode leakage. Since the receiver filter acts on the product of the aperture response and the OFDM waveform, this cyclic model was verified by a full waveform simulation (random CP-OFDM symbols through the time-varying aperture, Doppler shift, receiver filter, decimation, and FFT): over the drops of Figure 3b, the per-terminal SIR of the ST-SIM differs by 0.30.3 dB in median (0.70.7 dB at the 95th percentile), the cyclic model being slightly pessimistic, and the worst-terminal SINR by less than 0.10.1 dB; over the control rates of Figure 2c, the average SIR of the full waveform differs from the cyclic model by 0.10.1 to 0.70.7 dB with continuous and three-bit phases, except by −2.3-2.3 dB with continuous phases at 15.3615.36 MHz, where the SIR exceeds 4444 dB. The SINR is γk=SNR0Pd,k/(SNR0Pi,k+N)\gamma_k=\mathrm{SNR}_0P_{d,k}/(\mathrm{SNR}_0P_{i,k}+N) with SNR0=PN/σ2\mathrm{SNR}_0=PN/\sigma^2, i.e., the SNR of a lossless full-aperture beam including the array gain NN (55 dB per meta-atom at 2525 dB for N=100N=100), and, since all terminals decode the same broadcast stream, the figures report the worst-terminal SINR min⁡kγk\min_k\gamma_k and the broadcast spectral efficiency min⁡klog⁡2(1+γk)\min_k\log_2(1+\gamma_k). All schemes are compared at equal long-run radiated peak energy, defined by sup⁡u∥g⋆(u)∥22=Ppk(Ψ)=1\sup_u\left\lVert \mathbf{g}^\star(u)\right\rVert_2^2=P_{\rm pk}(\mathbf{\Psi})=1 for every ideal field g⋆(u)=ΨdD(u)\mathbf{g}^\star(u)=\mathbf{\Psi}\mathbf{d}_D(u), with PpkP_{\rm pk} the long-run peak energy (16), and by g^r=min⁡{1,max⁡r′∥α⋆gr′∥2−1}α⋆gr\widehat{\mathbf{g}}_r=\min\{1,\max_{r'}\left\lVert \alpha^\star\mathbf{g}_{r'}\right\rVert_2^{-1}\}\alpha^\star\mathbf{g}_r for every realized field, i.e., the gain α⋆\alpha^\star of (26) fits the realized field to the ideal target and is reduced if the energy exceeds one within the simulated window, as required by passivity (Lemma 2); the equal feed power figures of merit additionally scale the stack-based schemes by their simulated transmission efficiency ηL=∥Gf∥22‾\eta_L=\overline{\left\lVert \mathbf{G}\mathbf{f}\right\rVert_2^2}.

Realizability: Convergence, Quantization, and Control Rate

In Figure 2a, the objective of Algorithm alg:synthesis decreases monotonically (Proposition 5). The single layer stalls at −5.9-5.9 dB for every initialization: a phase-only layer cannot change the amplitude of any element excitation, whereas the multi-tone target (10) requires element amplitudes that vary in time, so that inter-layer mixing is necessary rather than merely helpful. For L≥3L\ge3 it reaches −50-50 to −61-61 dB after 300300 iterations and the three initializations of each depth end within 44 dB; L=2L=2 stalls near −26-26 dB, lacking the degrees of freedom to shape four frequency-translated modes. Fixed-step and layer-wise descent reach only −23-23 and −32-32 dB; Adam descends faster (−68-68 and −73-73 dB at learning rates 0.10.1 and 0.050.05) but is not monotone (8383 and 3636 ascending steps out of 300300) and must be tuned per problem; since the quantization floor of −13-13 dB lies more than 4040 dB above the stationary points of either method, the monotone, tuning-free descent of Algorithm alg:synthesis is preferable for an autonomous payload.

Phase resolution does not change this picture (Figure 2b): the single layer stays between −5.3-5.3 and −5.9-5.9 dB from three bits on, whereas two layers reach −26-26 dB with continuous phases. In Figure 2b, the refinement of Step 3 lowers the four-layer error after projection from −2.0-2.0, −6.5-6.5, and −12.0-12.0 dB to −8.2-8.2, −13.4-13.4, and −19.1-19.1 dB for 22, 33, and 44 bits, which is worth more than one bit, and all refined curves follow the 66-dB-per-bit slope of Proposition 6 (5.25.2 and 5.75.7 dB per bit from two to four bits for L=4L=4). Beyond L=3L=3 depth barely helps (−12.8-12.8, −13.4-13.4, and −13.9-13.9 dB at three bits for L=3L=3, 44, and 66), whereas ηL\eta_L drops from 0.400.40 to 0.240.24, 0.160.16, and 0.050.05, so the simulated diffraction loss limits the useful depth. In Figure 2c, the in-band SIR with continuous phases follows Proposition 7 within 11 dB up to fu=3.84f_u=3.84 MHz and exceeds it from 7.687.68 MHz on (3535 dB on average, 3131 dB for the terminal with the largest tone), where the receiver filter rejects part of the images, whereas two and three bits saturate at 2222 and 2828 dB, the quantization floor. The gain at high rates is not free: the out-of-band fraction decreases from −26-26 dBc at 0.480.48 MHz to −33-33 dBc at 7.687.68 MHz for continuous phases, but rises again for quantized phases (−27.5-27.5 dBc for two bits at 30.7230.72 MHz), whose quantization noise spreads over a bandwidth proportional to fuf_u; generated in the radiating aperture, beyond any payload filter, these images must be controlled through fuf_u, BϕB_\phi, and the control waveform.

Figure 3a shows the capability that distinguishes the ST-SIM from a static SIM: each terminal direction receives its own tone −(fD,k−νc⋆)-(f_{D,k}-\nu_c^\star) in a beam nulled toward the other terminals, i.e., H≈Dβ\mathbf{H}\approx\mathbf{D}_\beta in (17); at each terminal, the other tones are at least 3737 dB down, so the 2626-dB in-band SIR of the three-bit stack is set by quantization and switching noise, not by inter-terminal leakage. Along the pass, Benchmark 1 attains an average SIR of only −32-32 to +3+3 dB, since the common-translation residual reaches 151151 kHz (5.0Δf5.0\Delta f) near zenith, whereas the ST-SIM keeps 3535–4949 dB (continuous) and 2121–2727 dB (three bits).

Figure 3b shows two regimes. In small footprints, the Doppler span is small, and the terminals are not spatially resolvable: Benchmark 1 performs well (16.816.8 dB at 55 km), whereas the ST-SIM and Benchmark 2, which must separate the terminals to give them different frequencies, pay the penalty of Proposition 1 and Lemma 3 (−42-42 dB at 55 km for the ideal field; the approximate realizations of the stack lie 22–66 dB below it, trading leakage against energy). The single-layer aperture, whose interference floor lies near 13.513.5 dB in the baseline geometry, stays 8.18.1 and 6.36.3 dB below the four-layer stack at 200200 and 300300 km, the gain of stacking at equal radiated peak energy, phase resolution, and control rate. The SINR-optimal translation νcSINR\nu_c^{\rm SINR} leaves Benchmark 1 unchanged up to 5050 km and raises it by 5.05.0, 4.74.7, and 3.23.2 dB at 100100, 200200, and 300300 km, where it exploits the zeros of the ICI law, still more than 2020 dB below the ST-SIM. The curves cross at about 4545 km; beyond it, Benchmark 1 collapses, while the continuous-phase ST-SIM stays within 0.20.2 dB of Benchmark 2 in median from 100100 km on (14.414.4 dB at 200200 km and 17.617.6 dB at 300300 km, 0.70.7–0.80.8 dB less with three bits). The long-run peak-to-average ratio of the ideal field, paid by every scheme, is 2.22.2 dB in median at 200200 km and 4.64.6 dB at 55 km. Benchmark 3 attains 2020–2525 dB at every radius, i.e., 6.06.0 dB more than the ST-SIM at 200200 km, for two reasons inherent to per-terminal translation at the payload, digital or electromagnetic: its static field has no peak-to-average penalty and delivers a common gain to all terminals instead of separating them; it presumes, however, that every terminal has acquired and tracks a residual of several subcarrier spacings. At equal feed power, where every stack-based field is divided by the fitted gain ∣α⋆∣2\left\lvert \alpha^\star\right\rvert^2 of (26) and Benchmark 2 is unchanged, the stack-based schemes lose about 88 dB to the lossless Benchmark 2 (ηL=0.16\eta_L=0.16), the simulated diffraction loss of the normalized cascade model: in the baseline geometry, Benchmark 2 attains 18.718.7 dB, Benchmark 3 10.410.4 dB, and the ST-SIM 10.910.9 and 11.311.3 dB with continuous and three-bit phases, the latter reflecting a smaller fitted gain rather than a benefit of quantization (medians over the drops at 200200 km: 12.812.8, 8.88.8, and 14.514.5 dB for Benchmark 3, the ST-SIM, and Benchmark 2). The lossless single layer (η1=1\eta_1=1) reaches 12.012.0 dB at equal feed power in the baseline geometry, and the four-layer stack overtakes it from SNR0≈27\mathrm{SNR}_0\approx27 dB on, where the interference floor of the single layer dominates. The optimal depth at equal feed power is smaller: in the baseline geometry, the equal-feed worst-terminal SINR at SNR0=15/25/40\mathrm{SNR}_0=15/25/40 dB is 6.6/12.0/13.46.6/12.0/13.4 dB for L=1L=1, 5.0/14.8/27.15.0/14.8/27.1 dB for L=2L=2, 2.8/12.8/26.02.8/12.8/26.0 dB for L=3L=3, 0.9/10.9/24.70.9/10.9/24.7 dB for L=4L=4, and −4.1/5.8/20.4-4.1/5.8/20.4 dB for L=6L=6, whereas L=2L=2 to 66 coincide within 0.40.4 dB at equal radiated energy; two layers, whose −26-26-dB synthesis error lies below the noise at these SNRs, overtake the single layer from SNR0≈20\mathrm{SNR}_0\approx20 dB and are the best equal-feed choice in the normalized model, where each additional layer costs about 44 dB of diffraction loss. The near-digital performance thus holds at equal radiated peak energy only, and the single-layer comparison is conditional on the SNR, on the geometry, and on the diffraction loss of the model. The baseline SNR0=25\mathrm{SNR}_0=25 dB is optimistic for current high-power amplifiers and Ka-band link budgets [9]; at SNR0=15\mathrm{SNR}_0=15 dB, the median worst-terminal SINR at 200200 km is 4.54.5 dB for the ST-SIM and Benchmark 2 and 10.410.4 dB for Benchmark 3, and the crossover with Benchmark 1 moves to about 5959 km, which calls for the larger apertures of § 5.4.

Aperture Size and Number of Terminals

Fig. fig:aperture sweeps the aperture from 6×66\times6 to 16×1616\times16 meta-atoms at 200200 km. The static Benchmark 1 is limited by the Doppler floor at −27-27 dB and does not improve with NN, and Benchmark 3 stays at 2020–2121 dB, whereas the ST-SIM gains 8.48.4 dB from N=36N=36 to 256256, because a larger aperture resolves the terminals with less modal energy (Proposition 1); with L=4L=4 the continuous-phase stack tracks Benchmark 2 within 0.10.1 dB for every NN, three bits cost 0.80.8–1.41.4 dB, L=2L=2 falls 0.20.2–0.50.5 dB short, and L=6L=6 brings no further gain.

The same mechanism governs Fig. fig:sumrate: the broadcast spectral efficiency of the continuous-phase ST-SIM tracks Benchmark 2 within 0.050.05 bit/s/Hz for every KK with a single RF chain, three bits cost at most 0.40.4 bit/s/Hz, and Benchmark 1 collapses for K≥2K\ge2, and the single-layer aperture loses 3.23.2 bit/s/Hz to the four-layer stack at K=4K=4. Benchmark 3 without pilot overhead reaches 6.86.8 bit/s/Hz at K=4K=4 and exceeds the ST-SIM (5.05.0 bit/s/Hz) for K≥2K\ge2, since a static beam does not pay the separation cost that grows with KK; with two pilot symbols per slot it drops to 5.85.8 bit/s/Hz, and the ST-SIM leads unless CFO tracking costs less than 3%3\%, 16%16\%, and 27%27\% of the resources for K=2K=2, 33, and 44. Time-division beam hopping serves each terminal only during its dwell: log⁡2(1+SNR0)/K=2.1\log_2(1+\mathrm{SNR}_0)/K=2.1 bit/s/Hz at K=4K=4.

Grouped Compensation and Robustness

Remark 1 is evaluated with contiguous partitions of the Doppler-ordered terminals, each group receiving the midpoint of its Doppler range. In Figure 3b, adaptive grouping selects for every drop the partition that maximizes the spectral efficiency predicted from the ideal grouped field, which the stack then synthesizes without terminal-side correction. It follows Benchmark 1 in small footprints (G=1G=1 in all drops up to 1010 km and in 96%96\% at 2020 km) and the per-terminal ST-SIM in large ones (G=KG=K in 88%88\% of the drops at 300300 km), and exceeds both in between: at 5050 km, where it mostly selects G=2G=2 or 33, its median SINR of 5.15.1 dB lies 9.29.2 dB above the per-terminal ST-SIM, and at 100100 and 200200 km the gain is still 2.32.3 and 1.11.1 dB, so the mode switch at the crossover becomes a smooth transition. Fig. fig:grouping sweeps GG for K=8K=8 terminals in a 200200-km footprint with the partition that minimizes the largest in-group residual. Without terminal-side correction, the spectral efficiency grows from 0.00.0 bit/s/Hz at G=1G=1 (Benchmark 1) to 2.12.1 bit/s/Hz at G=KG=K, since in-group residuals of several subcarrier spacings are fatal; with correction of the in-group residual, it decreases from 5.25.2 bit/s/Hz at G=1G=1 (Benchmark 3), because separating groups costs energy, while the frequency search of the terminal shrinks from 1717 to 99 and 55 hypotheses on a Δf/2\Delta f/2 grid for G=2G=2 and 33 at 4.74.7 and 4.14.1 bit/s/Hz. For K=4K=4, G=2G=2 reduces the search from 1313 to 55 hypotheses at a 6%6\% loss in spectral efficiency. Grouping thus turns the choice between payload- and terminal-side correction into a continuum.

The SINR loss of the ST-SIM is about 11 dB for terminal-position errors up to 11 km, as for Benchmark 2 and as predicted by Propositions (?) and (?) (Fig. fig:robust), and 0.30.3 dB for satellite-velocity errors up to 1010 m/s. The initial acquisition of an unregistered terminal is outside the scope of the ST-SIM, since no dedicated mode exists before its position is known (§ 2); for the reacquisition of a registered terminal and for tracking during a session, the ST-SIM delivers a carrier-Doppler-free signal, so that a terminal needs a single frequency hypothesis, whereas a common-translation payload leaves 99 (1313) hypotheses in median (95th percentile) at 200200 km and 1111 (1717) at 300300 km; the oscillator offset adds the same range to all schemes.

Conclusion

This paper introduced space–time SIMs, which pair each spatial mode of a multilayer stack with its own frequency translation, and applied them to Doppler-resilient LEO broadcasting with a single RF chain. For the ideal aperture field, stacking convolves the temporal spectra of the layers (Lemma (?)), the minimum-energy space–frequency synthesis has a closed-form solution governed by the terminal geometry (Proposition 1), full column rank of the steering matrix is necessary and sufficient for decoupling KK terminal-specific translations (Theorem 1), grouping relaxes this condition (Remark 1), and any common translation, i.e., any static SIM or single-translation time-modulated surface, leaves half the Doppler span (Proposition 2). Algorithm alg:synthesis decreases the synthesis objective monotonically to stationary points, and explicit laws quantify quantization, control rate, and ephemeris errors. Under the ideal phase-only wave-domain model, a four-layer stack reached synthesis errors of −58-58 dB with continuous and −13-13 dB with three-bit phases, removed the Doppler floor of a static SIM above a footprint radius of about 4545 km, matched ideal digital per-terminal precompensation within 0.50.5 dB and 0.050.05 bit/s/Hz at equal radiated peak energy, and kept an in-band SIR of 2626 dB with three bits at a 7.687.68-MHz control rate, whereas a single time-modulated layer stalled at a synthesis error of −6-6 dB, which identifies inter-layer mixing as the enabling element of the architecture.

The results also delimit the regime of the architecture. Separating terminals costs modal energy that grows with their number and shrinks with the aperture (++8.48.4 dB from 3636 to 256256 meta-atoms per layer). A static beam with terminal-side CFO correction attained a higher SINR and, with dedicated pilots, a higher broadcast spectral efficiency for three or more terminals; the ST-SIM led for two terminals when CFO tracking costs more than 3%3\% of the resources, and always delivered a carrier-Doppler-free signal that removes the frequency search of registered terminals, with grouped compensation providing a continuum between the two approaches. At equal feed power, each layer of the normalized cascade model costs about 44 dB of diffraction loss, which favors two or three layers.

Light-Time Doppler Model

Let Tk(u)≜u+τk(u)\mathcal T_k(u)\triangleq u+\tau_k(u), where τk\tau_k solves c τk(u)=∥rk(u+τk(u))−rs(u)∥2≜dk(u)c\,\tau_k(u)=\left\lVert \mathbf{r}_k(u+\tau_k(u))-\mathbf{r}_s(u)\right\rVert_2\triangleq d_k(u) [8], and let uk\mathbf{u}_k be the unit vector from rs(u)\mathbf{r}_s(u) to rk(u+τk)\mathbf{r}_k(u+\tau_k). Since ∣ukTr˙k∣<c\left\lvert \mathbf{u}_k^T\dot{\mathbf{r}}_k\right\rvert<c and ∣ukTr˙s∣<c\left\lvert \mathbf{u}_k^T\dot{\mathbf{r}}_s\right\rvert<c, the light-time equation has a unique solution, Tk\mathcal T_k is strictly increasing, and the retarded transmit time uk(t)=Tk−1(t)u_k(t)=\mathcal T_k^{-1}(t) is well defined.

Proposition 8 (Light-time LEO Doppler). The instantaneous carrier Doppler frequency of (1) is fD,kex(t)=−fcτ˙k(u)1+τ˙k(u)=−fcukTvkrelc−ukTr˙s,u=uk(t), f_{D,k}^{\rm ex}(t)=-f_c\frac{\dot\tau_k(u)}{1+\dot\tau_k(u)} =-f_c\frac{\mathbf{u}_k^T\mathbf{v}_k^{\rm rel}}{c-\mathbf{u}_k^T\dot{\mathbf{r}}_s},\quad u=u_k(t), so that fD,kex(t)=−fccukTvkrel[1+O(∥r˙s∥2/c)]f_{D,k}^{\rm ex}(t)=-\frac{f_c}{c}\mathbf{u}_k^T\mathbf{v}_k^{\rm rel}\big[1+\mathcal O(\left\lVert \dot{\mathbf{r}}_s\right\rVert_2/c)\big].

For LEO motion the relative correction is at most ∥r˙s∥2/c=2.5×10−5\left\lVert \dot{\mathbf{r}}_s\right\rVert_2/c=2.5\times10^{-5}, so the first-order form is used throughout.

References

  1. [1]Kodheli, O., Lagunas, E., Maturo, N., Sharma, S. K., Shankar, B., Montoya, J. F. M., Duncan, J. C. M., Spano, D., Chatzinotas, S., Kisseleff, S., Querol, J., Lei, L., Vu, T. X., & Goussetis, G. (2021). Satellite Communications in the New Space Era: A Survey and Future Challenges. IEEE Commun. Surveys Tuts., 23(1), 70–109. https://doi.org/10.1109/COMST.2020.3028247
  2. [2]Kumar, S., Sheemar, C. K., Querol, J., Yilmaz, T., Chatzinotas, S., Hammouda, M., Heyn, T., Schlichter, T., Marques, P., Pereira, L., et al. (2023). 5G NTN LEO based demonstrator using OpenAirInterface5G. IET Conference Proceedings CP873, 69–75.
  3. [3]Sheemar, C. K., Iacovelli, G., Khan, W. U., Solanki, S., & Chatzinotas, S. (2026). Beam Tracking for SIM-Assisted LEO Satellites: From Orbital Prediction to Wave-Domain Mobility Management. Engineering Archive. https://doi.org/10.31224/8514
  4. [4]Sheemar, C. K., Iacovelli, G., Solanki, S., Khan, W. U., & Chatzinotas, S. (2026). One-Shot Beam Tracking for Stacked Intelligent Metasurfaces-Assisted LEO Broadcasting. https://arxiv.org/abs/2610.05520
  5. [5]Iacovelli, G., Sheemar, C. K., & Chatzinotas, S. (2026). Stacked Fluid Metasurfaces: Mutual-Coupling-Aware Modeling and Optimization. https://arxiv.org/abs/2609.31591
  6. [6]Sheemar, C. K., Iacovelli, G., Solanki, S., Khan, W. U., & Chatzinotas, S. (2026). Stacked Intelligent Metasurfaces Assisted UAV Communications. https://arxiv.org/abs/2607.24374
  7. [7]Sheemar, C. K., Iacovelli, G., Solanki, S., Khan, W. U., Alexandropoulos, G. C., & Chatzinotas, S. (2026). Energy Efficient Multi-User Beamforming and 3D Position Optimization for SIM-Assisted UAVs. https://arxiv.org/abs/2608.13765
  8. [8]Montenbruck, O., & Gill, E. (2000). Satellite Orbits: Models, Methods, and Applications. Springer. https://doi.org/10.1007/978-3-642-58351-3
  9. [9]3GPP. (2023). Solutions for NR to Support Non-Terrestrial Networks (NTN) (Release 16). In 3rd Generation Partnership Project (3GPP) (Numbers TR 38.821 V16.2.0).
  10. [10]Ali, I., Al-Dhahir, N., & Hershey, J. E. (1998). Doppler Characterization for LEO Satellites. IEEE Trans. Commun., 46(3), 309–313. https://doi.org/10.1109/26.662636
  11. [11]You, L., Li, K.-X., Wang, J., Gao, X., Xia, X.-G., & Ottersten, B. (2020). Massive MIMO Transmission for LEO Satellite Communications. IEEE J. Sel. Areas Commun., 38(8), 1851–1865. https://doi.org/10.1109/JSAC.2020.3000803
  12. [12]Kodheli, O., Astro, A., Querol, J., Gholamian, M., Kumar, S., Maturo, N., & Chatzinotas, S. (2021). Random Access Procedure Over Non-Terrestrial Networks: From Theory to Practice. IEEE Access, 9, 109130–109143. https://doi.org/10.1109/ACCESS.2021.3101291
  13. [13]Lin, X., Lin, Z., Löwenmark, S. E., Rune, J., & Karlsson, R. S. (2021). Doppler Shift Estimation in 5G New Radio Non-Terrestrial Networks. Proc. IEEE Global Commun. Conf. (GLOBECOM), 1–6. https://doi.org/10.1109/GLOBECOM46510.2021.9685184
  14. [14]Kim, M.-G., & Jo, H.-S. (2022). Performance Analysis of NB-IoT Uplink in Low Earth Orbit Non-Terrestrial Networks. Sensors, 22(18), 7097. https://doi.org/10.3390/s22187097
  15. [15]Meshram, A. K., Kumar, S., Querol, J., Andrenacci, S., & Chatzinotas, S. (2025). Reduced Complexity Initial Synchronization for 5G NR Multibeam LEO-Based Non-Terrestrial Networks. IEEE Open J. Commun. Soc., 6, 1528–1551. https://doi.org/10.1109/OJCOMS.2025.3543625
  16. [16]Pollet, T., Bladel, M. V., & Moeneclaey, M. (1995). BER Sensitivity of OFDM Systems to Carrier Frequency Offset and Wiener Phase Noise. IEEE Trans. Commun., 43(2/3/4), 191–193. https://doi.org/10.1109/26.380034
  17. [17]Moose, P. H. (1994). A Technique for Orthogonal Frequency Division Multiplexing Frequency Offset Correction. IEEE Trans. Commun., 42(10), 2908–2914. https://doi.org/10.1109/26.328961
  18. [18]Hadani, R., Rakib, S., Tsatsanis, M., Monk, A., Goldsmith, A. J., Molisch, A. F., & Calderbank, R. (2017). Orthogonal Time Frequency Space Modulation. Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), 1–6. https://doi.org/10.1109/WCNC.2017.7925924
  19. [19]Bemani, A., Ksairi, N., & Kountouris, M. (2023). Affine Frequency Division Multiplexing for Next Generation Wireless Communications. IEEE Trans. Wireless Commun., 22(11), 8214–8229. https://doi.org/10.1109/TWC.2023.3260906
  20. [20]Souza, P. H. C. de ., & Mendes, L. L. (2024). Doppler Shift Compensation Assisted by RIS Systems in High-Mobility Communications. IEEE Commun. Lett., 28(11), 2658–2662. https://doi.org/10.1109/LCOMM.2024.3466562
  21. [21]An, J., Renzo, M. D., Debbah, M., & Yuen, C. (2023). Stacked Intelligent Metasurfaces for Multiuser Beamforming in the Wave Domain. Proc. IEEE Int. Conf. Commun. (ICC), 2834–2839. https://doi.org/10.1109/ICC45041.2023.10279173
  22. [22]An, J., Xu, C., Ng, D. W. K., Alexandropoulos, G. C., Huang, C., Yuen, C., & Hanzo, L. (2023). Stacked Intelligent Metasurfaces for Efficient Holographic MIMO Communications in 6G. IEEE J. Sel. Areas Commun., 41(8), 2380–2396. https://doi.org/10.1109/JSAC.2023.3288261
  23. [23]An, J., Renzo, M. D., Debbah, M., Poor, H. V., & Yuen, C. (2025). Stacked Intelligent Metasurfaces for Multiuser Downlink Beamforming in the Wave Domain. IEEE Trans. Wireless Commun., 24(7), 5525–5538. https://doi.org/10.1109/TWC.2025.3547779
  24. [24]Nerini, M., & Clerckx, B. (2024). Physically Consistent Modeling of Stacked Intelligent Metasurfaces Implemented With Beyond Diagonal RIS. IEEE Commun. Lett., 28(7), 1693–1697.
  25. [25]Sheemar, C. K., Khan, W. U., Solanki, S., Alexandropoulos, G. C., & Chatzinotas, S. (2026). A Survey on Stacked Intelligent Metasurfaces: Fundamentals, Recent Advances, and Challenges. https://arxiv.org/abs/2603.05633
  26. [26]Lin, S., An, J., Gan, L., Debbah, M., & Yuen, C. (2024). Stacked Intelligent Metasurface Enabled LEO Satellite Communications Relying on Statistical CSI. IEEE Wireless Commun. Lett., 13(5), 1295–1299. https://doi.org/10.1109/LWC.2024.3368238
  27. [27]Yao, X., An, J., Gan, L., Renzo, M. D., & Yuen, C. (2024). Channel Estimation for Stacked Intelligent Metasurface-Assisted Wireless Networks. IEEE Wireless Commun. Lett., 13(5), 1349–1353. https://doi.org/10.1109/LWC.2024.3369874
  28. [28]Li, Z., An, J., & Yuen, C. (2026). Stacked Intelligent Metasurface-Enhanced MIMO OFDM Wideband Communication Systems. IEEE Trans. Wireless Commun., 25, 9608–9622. https://doi.org/10.1109/TWC.2025.3636393
  29. [29]Ranasinghe, K. R. R., Sandoval, I. A. M., Rou, H. S., Abreu, G. T. F. de ., & Alexandropoulos, G. C. (2026). Doubly-Dispersive MIMO Channels With Stacked Intelligent Metasurfaces: Modeling, Parametrization, and Receiver Design. IEEE Trans. Wireless Commun., 25, 3801–3817. https://doi.org/10.1109/TWC.2025.3605839
  30. [30]Darsena, D., Iudice, I., Galdi, V., & Verde, F. (2026). Randomized Space-Time Stacked Intelligent Metasurfaces for Massive Multiuser Downlink Connectivity. IEEE Trans. Commun., 74, 8059–8074. https://doi.org/10.1109/TCOMM.2026.3686695
  31. [31]Kummer, W. H., Villeneuve, A. T., Fong, T. S., & Terrio, F. G. (1963). Ultra-Low Sidelobes from Time-Modulated Arrays. IEEE Trans. Antennas Propag., 11(6), 633–639. https://doi.org/10.1109/TAP.1963.1138102
  32. [32]Hadad, Y., Sounas, D. L., & Alù, A. (2015). Space-Time Gradient Metasurfaces. Phys. Rev. B, 92(10), 100304(R). https://doi.org/10.1103/PhysRevB.92.100304
  33. [33]Caloz, C., & Deck-Léger, Z.-L. (2020). Spacetime Metamaterials---Part I: General Concepts. IEEE Trans. Antennas Propag., 68(3), 1569–1582. https://doi.org/10.1109/TAP.2019.2944225
  34. [34]Zhang, L., Chen, X. Q., Liu, S., Zhang, Q., Zhao, J., Dai, J. Y., Bai, G. D., Wan, X., Cheng, Q., Castaldi, G., Galdi, V., & Cui, T. J. (2018). Space-Time-Coding Digital Metasurfaces. Nat. Commun., 9, 4334. https://doi.org/10.1038/s41467-018-06802-0
  35. [35]Dai, J. Y., Zhao, J., Cheng, Q., & Cui, T. J. (2018). Independent Control of Harmonic Amplitudes and Phases via a Time-Domain Digital Coding Metasurface. Light Sci. Appl., 7, 90. https://doi.org/10.1038/s41377-018-0092-z
  36. [36]Zhang, L., Chen, M. Z., Tang, W., Dai, J. Y., Miao, L., Zhou, X. Y., Jin, S., Cheng, Q., & Cui, T. J. (2021). A Wireless Communication Scheme Based on Space- and Frequency-Division Multiplexing Using Digital Metasurfaces. Nat. Electron., 4(3), 218–227. https://doi.org/10.1038/s41928-021-00554-4
  37. [37]Ramaccia, D., Sounas, D. L., Alù, A., Toscano, A., & Bilotti, F. (2020). Phase-Induced Frequency Conversion and Doppler Effect With Time-Modulated Metasurfaces. IEEE Trans. Antennas Propag., 68(3), 1607–1617. https://doi.org/10.1109/TAP.2019.2952469
  38. [38]Taravati, S., & Eleftheriades, G. V. (2021). Pure and Linear Frequency-Conversion Temporal Metasurface. Phys. Rev. Appl., 15(6), 064011. https://doi.org/10.1103/PhysRevApplied.15.064011
  39. [39]Taravati, S., & Eleftheriades, G. V. (2022). Microwave Space-Time-Modulated Metasurfaces. ACS Photon., 9(2), 305–318. https://doi.org/10.1021/acsphotonics.1c01041
  40. [40]Sun, Y., Wu, G.-B., & Chan, C. H. (2025). A Versatile Shaped-Beam Space-Time-Coding Metasurface Antenna Based on 1-bit Phase Control. IEEE Trans. Antennas Propag., 73(11), 8876–8887. https://doi.org/10.1109/TAP.2025.3590874
  41. [41]Gholami, M., Khajavi, S., Neshat, M., Tewes, S., & Sezgin, A. (2026). Realization of Reconfigurable Intelligent Surfaces With Space--Time-Coded Metasurfaces. IEEE Trans. Antennas Propag., 74(5), 4281–4291. https://doi.org/10.1109/TAP.2026.3660044
  42. [42]Lin, X., Rivenson, Y., Yardimci, N. T., Veli, M., Luo, Y., Jarrahi, M., & Ozcan, A. (2018). All-Optical Machine Learning Using Diffractive Deep Neural Networks. Science, 361(6406), 1004–1008. https://doi.org/10.1126/science.aat8084
  43. [43]Barzilai, J., & Borwein, J. M. (1988). Two-Point Step Size Gradient Methods. IMA J. Numer. Anal., 8(1), 141–148. https://doi.org/10.1093/imanum/8.1.141
  44. [45]Zhu, X., Qian, C., Zhang, J., Jia, Y., Xu, Y., Zhao, M., Zhao, M., Qu, F., & Chen, H. (2024). On-demand Doppler-offset beamforming with intelligent spatiotemporal metasurfaces. Nanophotonics, 13(8), 1351–1360. https://doi.org/10.1515/nanoph-2023-0569
  45. [46]Movahediqomi, M., Tretyakov, S., Asadchy, V., & Wang, X. (2026). Stacked Time-Varying Metasurfaces. Adv. Opt. Mater., 14(4), e02831. https://doi.org/10.1002/adom.202502831
  46. [47]He, C., Liang, X., Zhou, B., Geng, J., & Jin, R. (2015). Space-division multiple access based on time-modulated array. IEEE Antennas Wireless Propag. Lett., 14, 610–613. https://doi.org/10.1109/LAWP.2014.2373431
  47. [48]Castaldi, G., Zhang, L., Moccia, M., Hathaway, A. Y., Tang, W. X., Cui, T. J., & Galdi, V. (2021). Joint multi-frequency beam shaping and steering via space--time-coding digital metasurfaces. Adv. Funct. Mater., 31(6), 2007620. https://doi.org/10.1002/adfm.202007620
  48. [49]Wu, G.-B., Dai, J. Y., Cheng, Q., Cui, T. J., & Chan, C. H. (2022). Sideband-free space--time-coding metasurface antennas. Nat. Electron., 5(11), 808–819. https://doi.org/10.1038/s41928-022-00857-0

Paper details

vlnc:W_c2sc5qs4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Engineering

Not recorded for this paper

View statement
Looking for related papers…
Contents