Passive Intelligent Transmitting Surface (ITS)

• Passive Intelligent Transmitting Surface (ITS) is a reconfigurable electromagnetic aperture that reradiates incident signals by imposing controlled phase shifts without active RF chains.
• Methodologies involve illuminating a large passive surface with a few active antennas and employing optimization techniques such as MI, OMP, and ZF-WF for precise beamforming.
• The design trade-offs include challenges in phase resolution, path loss, and geometric constraints while offering a low-power, low-complexity alternative to active systems.

A passive intelligent transmitting surface (ITS) is a reconfigurable electromagnetic aperture that reradiates an incident field after imposing controlled phase shifts, while avoiding active RF chains or on-surface amplification. In the transmitter-oriented ITS architecture, a small number of active antennas illuminate a nearby large passive surface over the air, and the surface forms the useful radiated beam; in closely related formulations, the surface is integrated into a base station, embedded in a wireless power beacon, or used as a passive information transmitter by encoding data in its phase states (Jamali et al., 2019). The literature also includes a waveform-selective passive ITS/RIS that autonomously switches between reflection states without external DC power or active control circuitry, showing that “passive” does not imply static behavior (Omori et al., 24 Oct 2025).

1. Conceptual scope and terminology

The ITS literature uses several closely related terms. “Intelligent Transmitting Surface (ITS)” denotes a passive transmitting or refractive aperture illuminated by active feeders, while “Transmitting Intelligent Surface (TIS)” appears in satellite-navigation work for a window-mounted surface that redirects outdoor signals into indoor space (Jamali et al., 2019, Guan et al., 26 May 2026). In transmitter-array work, the ITS is described as a large passive surface of many low-cost transmitting elements that retransmit an incident field after element-wise phase adjustment; in base-station-integrated work, a few active antennas illuminate a large passive transmissive aperture to create a highly directional effective link with few RF chains (Ramezani et al., 27 Jan 2026).

A second branch of the literature uses reflection-based surfaces as passive transmitters. In these formulations, the surface does not generate RF energy; instead, it embeds information into reflected carrier waves or reradiated fields by choosing its reflection coefficients symbol by symbol. This appears both in symbol-level precoding formulations and in coded modulation/message-passing formulations, where the RIS phases themselves carry data (Liu et al., 2020, Jiang et al., 2022). A plausible implication is that passive ITS should be understood functionally rather than only geometrically: the common feature is passive reradiation with programmable phase control, regardless of whether the surface is explicitly framed as transmissive, reflective, or waveform-selective.

The distinction between passive and active intelligent surfaces is structurally important. In phase-only passive models, the surface reflection matrix is

$$\mathbf{\Psi}=\boldsymbol{\Theta}=\operatorname{diag}(e^{j\phi_1},\ldots,e^{j\phi_N}),$$

with unit amplitude at every element, no amplification noise, and coherent combining gain, but also product-distance path loss (You et al., 2021). Active IRS models, by contrast, include amplification gain and amplification noise. This contrast recurs throughout deployment, precoding, and WIT/WPT analyses.

2. Electromagnetic architecture and canonical signal models

A canonical ITS transmitter architecture comprises $N$ active antennas, each with a dedicated RF chain, and $M$ passive transmitting elements on the surface. The active antennas illuminate the surface over the air, and the overall precoder has the structured form

$\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$

where $\mathbf{B}\in\mathbb{C}^{N\times Q}$ is the digital baseband precoder, $\mathbf{T}\in\mathbb{C}^{M\times N}$ is the fixed over-the-air coupling matrix from active antennas to passive elements, and $\mathbf{D}$ is the diagonal phase-shift matrix of the surface (Jamali et al., 2019). The coupling matrix explicitly includes free-space attenuation, antenna gains, surface efficiency, and propagation phase. This structure differs from conventional hybrid beamforming because the analog part is not a lossy RF feed network but a physical electromagnetic coupling through the passive surface.

In multi-user ITS-aided array formulations, the effective transmitted field is shaped by the product $\mathbf{D}\mathbf{T}\mathbf{B}$, and user $k$ observes

$$y_k = \mathbf{h}_k^T \mathbf{D}\mathbf{T}\mathbf{B}\mathbf{s} + n_k,$$

with

$N$0

The corresponding objective is the weighted sum rate

$N$1

under either a radiated-power constraint $N$2 or a transmitted-power constraint $N$3 (Beerten et al., 9 Apr 2026).

Angle-domain models offer a complementary description. In hierarchical passive beamforming, generalized Snell’s law is used to impose a structured phase profile,

$N$4

where the phase gradients $N$5 determine reflection direction and $N$6 is a reference phase (Cai et al., 2021). This reduces design from $N$7 element-wise variables to $N$8 structured variables for $N$9 URA RISs. In another integrated-BS formulation, the surface vector is decomposed as

$M$0

and the passive precoding gain magnitude for low-mobility users is invariant to the common phase-shift $M$1, a property later exploited for transmit diversity (Zheng et al., 2024).

3. Passive control mechanisms and self-reconfiguration

Passive ITS need not rely on external biasing or digital control lines. A direct experimental demonstration uses a waveform-selective metasurface described as a passive ITS/RIS that reacts autonomously to the pulse width of the incident waveform and switches between two singular reflection angles without any external DC power supply or active control circuitry (Omori et al., 24 Oct 2025). The unit cell contains four diodes arranged as a diode bridge, together with a capacitor $M$2, a resistor $M$3, and an additional inductor $M$4. For a short pulse, the capacitor does not fully charge, the bridge behaves effectively like a short circuit, and the supercell phase profile is nearly flat, producing specular reflection. For a continuous wave or sufficiently long pulse, the capacitor becomes fully charged, the steady-state impedance changes, and the added inductance produces a designed phase gradient that yields anomalous reflection.

The same work provides impedance and beamforming expressions. The effective impedance is modeled as

$M$5

with

$M$6

For beam steering, generalized reflection-law relations are given for the phase profile across the surface. The paper also notes that the printed reflection-coefficient equation

$M$7

appears to contain a typographical error (Omori et al., 24 Oct 2025).

Experimentally, the prototype uses Rogers RO3003, HSMS 286x diodes, $M$8 nF, $M$9 k$\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$0, and a 10-cell supercell with different $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$1 values. In the rigorous simulation and experiment, the short-pulse state reflects toward $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$2, while the continuous-wave state redirects power toward $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$3. The paper reports that the received magnitude can vary by a factor of ten; the CW anomalous-reflection peak is about 19.2 dB larger than the SP reflection at the same anomalous angle in the simplified model; and BPSK communication characteristics vary by 7 dB or more, with correct reception around the expected $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$4 and $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$5 phase states (Omori et al., 24 Oct 2025). This establishes a passive route to dual-state beam routing driven by waveform duration rather than by external control.

4. Beamforming, precoding, and channel tracking methodologies

Optimization for passive ITS is shaped by unit-modulus constraints, the fixed coupling matrix $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$6, and often sparse angular structure. In mmWave ITS transmitter design, two precoders were proposed: an MI-based precoder that directly maximizes

$\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$7

subject to $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$8 and $\mathbf{F}=\mathbf{D}\mathbf{T}\mathbf{B},$9, and an OMP-based precoder that approximates the unconstrained fully digital precoder under the structural constraint $\mathbf{B}\in\mathbb{C}^{N\times Q}$0 (Jamali et al., 2019). The design is justified by sparse mmWave channel models in which the row space of the channel is spanned by a small set of steering vectors.

For structured passive beamforming, the HPB-SPP algorithm first chooses phase gradients by strongest-path pairing and then aligns per-surface reference phases by successive convex approximation with update

$\mathbf{B}\in\mathbb{C}^{N\times Q}$1

The main simplification is that complexity depends on the number of surfaces rather than the number of elements, avoiding the $\mathbf{B}\in\mathbb{C}^{N\times Q}$2 scaling noted for traditional SDR-based optimization (Cai et al., 2021).

In ITS-aided MU-MIMO, WSR maximization is solved by block coordinate descent combined with WMMSE/fractional programming, alternating over auxiliary variables, the surface phase matrix, and the digital precoder. A lower-complexity ZF-WF alternative maximizes signal power at the ITS, applies zero forcing on the effective channel, and performs water filling. Under the radiated-power constraint, ZF-WF nearly matches WMMSE-BCD; under the transmitted-power constraint, WMMSE-BCD clearly outperforms ZF-WF (Beerten et al., 9 Apr 2026).

Discrete-phase constraints motivate other methods. Spatial first-order $\mathbf{B}\in\mathbb{C}^{N\times Q}$3 modulation has been applied to a passive RIS transmitter with one active antenna and a large ULA RIS, using

$\mathbf{B}\in\mathbb{C}^{N\times Q}$4

Under mild assumptions, the approach allows ZF-based synthesis of RIS reflection phases with negligible complexity. The modulator directly outputs unit-modulus discrete phases, satisfies the constant-envelope constraint by construction, and for $\mathbf{B}\in\mathbb{C}^{N\times Q}$5 achieves performance comparable to continuous-phase operation (Keung et al., 2023).

Passive ITS also reshapes channel-estimation strategy. In an ITS-integrated BS, the UE–ITS line-of-sight channel is modeled per coherence block as

$\mathbf{B}\in\mathbb{C}^{N\times Q}$6

so only the amplitude, phase, and AoA must be tracked. A MAP tracker uses priors centered at previous estimates, closed-form updates for $\mathbf{B}\in\mathbb{C}^{N\times Q}$7 and $\mathbf{B}\in\mathbb{C}^{N\times Q}$8, a one-dimensional AoA search, and only $\mathbf{B}\in\mathbb{C}^{N\times Q}$9 pilots per coherence block. The result is spectral efficiency close to that under perfect CSI (Ramezani et al., 27 Jan 2026).

5. Passive ITS as an information, energy, and sensing platform

Passive ITS can serve as an information transmitter. In symbol-level precoding formulations, the surface maps each desired symbol vector $\mathbf{T}\in\mathbb{C}^{M\times N}$0 to a unit-modulus phase vector $\mathbf{T}\in\mathbb{C}^{M\times N}$1, pushing the received signals into PSK decision regions by constructive interference. The resulting nonconvex design is solved on the complex circle manifold by Riemannian conjugate gradient, with direct quantization, branch-and-bound, and heuristic element-wise refinement for low-resolution phase shifters (Liu et al., 2020). In the joint passive reflection and information transmission architecture, the IRS also supports primary users while embedding a binary secondary stream through two reflection states $\mathbf{T}\in\mathbb{C}^{M\times N}$2 and $\mathbf{T}\in\mathbb{C}^{M\times N}$3.

A related coded formulation is simultaneous active and passive information transfer (SAPIT), in which the transmitter sends conventional symbols while the RIS conveys its own coded data through phase coefficients from a discrete constellation. By introducing auxiliary variables,

$\mathbf{T}\in\mathbb{C}^{M\times N}$4

the original bilinear detection problem is decomposed into two linear models and one entry-wise bilinear model. The receiver uses AMP-style approximations, scalar message passing for the RIS symbols, and state evolution for large-system analysis. The paper reports that SAPIT significantly outperforms passive beamforming in achievable sum rate and, in one example, roughly doubles the sum rate of passive beamforming at large $\mathbf{T}\in\mathbb{C}^{M\times N}$5 (Jiang et al., 2022).

Wireless energy transfer is another major application. A passive ITS-equipped power beacon uses a collocated digital feeder, an ITS with $\mathbf{T}\in\mathbb{C}^{M\times N}$6 passive elements, and nonlinear Doherty HPA models. The received RF power at device $\mathbf{T}\in\mathbb{C}^{M\times N}$7 is

$\mathbf{T}\in\mathbb{C}^{M\times N}$8

and total power consumption includes baseband, transceiver-chain, HPA, and ITS control terms. The resulting nonconvex problem is handled by successive convex approximation plus a feasibility-oriented initializer. The simulations show that the ITS-equipped PB outperforms fully digital and hybrid benchmarks in power consumption, and that nonuniform ITS power distribution can shift a device between near- and far-field regions even with a constant aperture (Rosabal et al., 9 Jul 2025).

Navigation and sensing provide a further extension. In indoor GNSS extension, a TIS mounted on windows creates an extended line-of-sight link from satellite to indoor user. The three-stage TSIPA procedure first localizes TIS arrays from corrected pseudo-ranges, then estimates AoA by maximum likelihood, and finally locates the user by minimizing distances to AoA-defined rays. The work introduces TPDoP,

$\mathbf{T}\in\mathbb{C}^{M\times N}$9

to quantify centroid deviation, and RMSE to quantify compactness of TIS placement (Guan et al., 26 May 2026).

6. Deployment trade-offs, limitations, and recurrent debates

A recurring question is whether passive or active intelligent surfaces are preferable. In a single-AP single-user LoS model with blocked direct link, passive IRS achieves

$\mathbf{D}$0

showing quadratic array gain in $\mathbf{D}$1. Under optimized placement and small altitude $\mathbf{D}$2, the passive IRS is near-optimal when placed above either the transmitter or the receiver, whereas the active IRS tends to be deployed closer to the receiver in downlink. The passive surface is shown to outperform the active counterpart when the number of reflecting elements is sufficiently large and/or the active-IRS amplification power is too small; for weighted uplink/downlink sum-rate, passive IRS is even more likely to achieve superior rate performance (You et al., 2021).

This does not imply universal superiority of passive architectures. In multi-hop WIT/WPT with a cascade of passive IRSs and one active IRS, all-passive chains suffer from severe multiplicative path loss. The analysis shows that passive chains provide cooperative passive beamforming gain but can be inferior when $\mathbf{D}$3 is small or the cascaded path is long; for WPT the optimal active IRS is always the final IRS, while for WIT it is generally in the second half of the chain in the practical regime $\mathbf{D}$4 (Fu et al., 2023). This suggests that passive ITS is most effective as a beam-routing and aperture-expansion mechanism, but not necessarily as the sole remedy for deep multi-hop attenuation.

Hardware and geometry remain decisive. The ITS power model for mmWave ultra-massive MIMO explicitly includes spillover loss, taper loss, aperture loss, and phase-shifter loss, and derives nearly constant total power consumption with growing $\mathbf{D}$5, unlike FD, FC, and PC MIMO whose power grows strongly with array size (Jamali et al., 2019). In ITS-aided MU-MIMO, the choice between radiated-power and transmitted-power constraints changes optimal geometry and the sensitivity to surface loss; under transmitted-power constraints, a randomly configured ITS can even be harmful (Beerten et al., 9 Apr 2026). In waveform-selective passive ITS, the proof-of-concept also depends on pulse width, used stronger-than-ordinary communication signals because of commercial-diode voltage requirements, demonstrated only one steering dimension in a compact measurement space, and relied on a linear phase gradient rather than focusing (Omori et al., 24 Oct 2025).

Another misconception is that passive ITS requires static or slow control. The literature shows several distinct fast-control mechanisms: symbol-level phase modulation for passive information transfer, symbol-dependent common phase rotation that yields diversity order two for high-mobility users without CSI at the BS, and intrinsic waveform-selective state switching driven by capacitor charging dynamics (Liu et al., 2020, Zheng et al., 2024, Omori et al., 24 Oct 2025). The unifying limitation is not programmability per se, but the combination of unit-modulus constraints, surface loss, feeder illumination, and propagation geometry.

Passive ITS therefore occupies a specific design regime: it replaces dense active front ends or lossy analog feed networks with a large reconfigurable passive aperture, but remains governed by structured electromagnetic coupling, finite phase resolution, and geometry-dependent efficiency. Across communication, wireless power, channel tracking, and indoor navigation, the central design tension is consistent: maximize controllable aperture gain while preserving the low-power, low-RF-complexity character that makes passive ITS distinct in the first place.