Reflective Intelligent Surfaces (RISs)

• Reflective Intelligent Surfaces (RISs) are engineered metasurfaces with subwavelength elements whose reflection phase can be dynamically tuned to control wireless propagation.
• They utilize electronically tunable components like varactors and PIN diodes to enable precise beam steering and optimized signal modulation.
• RISs improve wireless performance by extending coverage, enhancing SNR and energy efficiency, and boosting physical-layer security in future networks.

Reflective Intelligent Surfaces (RISs) are engineered metasurfaces comprised of subwavelength elements whose individual reflection phase (and sometimes amplitude) response can be dynamically tuned to modify electromagnetic wave propagation in wireless environments. RISs serve as low-power, programmable interfaces—typically realized with planar periodic arrays of meta-atoms—capable of spatially shaping signal reflection to enhance coverage, capacity, and security in modern and anticipated future wireless networks, notably 6G and beyond. Their operation leverages hardware-level controllable phase shifts to steer, focus, and modulate reflected signals, offering a new degree of freedom in wireless environment reconfiguration (Xu et al., 2022, Shabanpour et al., 2023, Liu et al., 2020).

1. Electromagnetic Principles and Unit Cell Design

At the fundamental level, each RIS element acts as a subwavelength scatterer whose response is characterized by a complex reflection coefficient, typically denoted as

$\Gamma(\theta, f) = A(\theta, f) e^{j\varphi},$

where $A(\theta, f)$ is the (tunable) amplitude and $\varphi$ is the programmable phase shift (Costa et al., 2021, Huang et al., 2022). Transmission-line equivalent circuit models accurately describe the interplay between lumped element values (varactor capacitance, substrate inductance, parasitic resistance), incidence angles, and the resulting reflection phase and amplitude (Costa et al., 2021).

Most practical RISs use phase-only control (unit-modulus reflection), realized through electronically tunable loads such as PIN diodes or varactors connected across gaps in metallic patches. The circuit-level programmability of these elements facilitates per-cell tuning with dynamic ranges that, depending on technology and frequency, span up to 360° of phase (continuous varactor), or binary/multibit phase states with quantization–magnitude penalties (e.g., 1–3 bits) (Fara et al., 2021, Sayanskiy et al., 2022, Inácio et al., 28 Apr 2025).

Accurate RIS design requires “angular stability” of the reflection phase with respect to the potential range of incident angles. The “reflection locality” (RL) approximation—treating each non-uniform cell response as that of a uniform array loaded with the same impedance—remains valid only when RL and angular stability coincide, i.e., when the local phase response is approximately invariant to angle (Shabanpour et al., 2023). Deviations in phase uniformity across angles (non-angular-stable designs, such as mushroom metasurfaces) lead to severe errors in beamsteering and efficiency.

2. System and Channel Modeling

The canonical end-to-end input–output model for a RIS-assisted single-antenna link is

$$y = \left(h_d + \mathbf{h}_r^T\boldsymbol{\Theta}\mathbf{g}\right)s + n,$$

where $h_d$ is the direct link, $\mathbf{g}\in\mathbb{C}^{N\times 1}$ the BS–RIS channel, $\mathbf{h}_r\in\mathbb{C}^{N\times 1}$ the RIS–user channel, $n$ noise, and $$\boldsymbol{\Theta} = \mathrm{diag}(e^{j\theta_1},...,e^{j\theta_N})$$ the diagonal phase control matrix (Xu et al., 2022, Liu et al., 2020).

The goal is to jointly optimize $\{\theta_i\}$ to maximize received SNR: $$\max_{\theta_1,...,\theta_N} |h_d + \mathbf{h}_r^T\boldsymbol{\Theta}\mathbf{g}|^2.$$ In the absence of a direct path ($h_d=0$), the optimal phase alignment imparts an SNR gain scaling as $N^2$ in the ideal line-of-sight scenario. Iterative algorithms such as alternating optimization, semidefinite relaxation, and manifold methods are used to find locally optimal phase configurations due to the nonconvex nature of this problem (Xu et al., 2022, Pan et al., 2020).

RIS-assisted channels display path loss and beamforming gains distinct from classical propagation. Large-scale path loss for far-field beamforming follows a $1/(d_1 d_2)^2$ scaling (with $d_1$ and $d_2$ the Tx–RIS and RIS–Rx distances) (Huang et al., 2022). Small-scale statistics are often modeled as Rician fading, with spatial/temporal and frequency correlations set by physical geometry and element configuration.

3. Performance Scaling Laws, Information-Theoretic Capacity, and Applications

RISs can deliver dramatic link- and system-level enhancements:

• Coverage extension: RISs form programmable virtual line-of-sight paths, restoring connectivity in blocked conditions. The probability of coverage can scale as $\mathcal{O}(N^2)$ (statistical CSI) or $\mathcal{O}(N^4)$ (full CSI) (Xu et al., 2022).
• Reliability and throughput: The end-to-end SNR grows as $N^2$, enabling either increased data rates (additional $2\log_2 N$ bits/s/Hz at high SNR) or $1/N^2$ transmit power reduction for fixed rate (Xu et al., 2022, Zhang et al., 2020).
• Energy efficiency: Passive RISs operate at milliwatt-level controller power, orders of magnitude below relay-based or active array solutions, even when quantized phase control is employed (Xu et al., 2022).
• Physical-layer security: Optimized RIS reflection can increase secrecy rates and sharply reduce the secrecy outage probability. Under optimal configuration, the legitimate SNR grows as $N^2$ while eavesdropper SNR grows only as $N\log N$ on average (Xu et al., 2022), leading to unbounded secrecy rate growth in $N$ under certain conditions.
• Functionality expansion: Besides classical beam steering, RIS surfaces can perform (a) direct waveform modulation via dynamic phase or ON/OFF switching—Reflection Pattern Modulation (RIS-RPM), Quadrature Reflection Modulation (RIS-QRM), or group-based “ON/OFF” PBIT superposition for additional degrees of freedom, (b) surface-wave assisted beamforming via excitation of evanescent modes for enhanced pattern shaping and wide-angle steering (Ataloglou et al., 8 Apr 2025).

Prominent applications include mmWave/THz coverage, IoT backscatter, multicell interference nulling, integrated sensing and communications, and physically secured links (Pan et al., 2020, Liu et al., 2020, Alexandropoulos et al., 22 Jul 2025).

4. Hardware Architectures and Technological Variants

Reflective RIS hardware consists of a programmable central controller and an array of passive meta-atoms. Tuning mechanisms vary:

• Varactor-based phase shifters: Provide continuous or multibit phase tuning by voltage bias, enabling fine-grain beamforming and group-delay or dispersion engineering (Fara et al., 2021, Abbas et al., 2024).
• Schottky diodes and PIN diodes: Deployed for rapid, low-loss binary state switching, especially at mmWave/sub-THz frequencies, with typical state-averaged insertion losses $\sim$0.8–1.5 dB at 140 GHz and sub-ps switching speeds (Inácio et al., 28 Apr 2025).
• Memristors, phase-change materials, RF-SOI switches, liquid metals: Emerging alternatives offering nW static power, ultracompact footprints, multi-state programmability, or mechanical tunability at D-band frequencies (Inácio et al., 28 Apr 2025).
• Wave-controlled and optically-controlled RISs: Transmission-line biasing or infrared addressable blocks augment scalability and wiring simplicity, supporting modular 2D phase control without per-cell hardwired lines (Itzhak et al., 2024, Sayanskiy et al., 2022).
• Passive, pulse-width-modulated, and self-adaptive RISs: Surface elements can switch reflection behavior between specular and anomalous states based only on incident waveform structure (e.g., pulse width), without external bias or synchronization (Omori et al., 24 Oct 2025).

Table: Representative Switching Technologies for Reflective RISs at D-Band (140 GHz) (Inácio et al., 28 Apr 2025)

Technology	Insertion Loss (dB)	Phase Range (°)	Switching Speed	Power	Reliability (cycles)
Schottky Diode	0.8–1.5	160–200	<1 ps	0.1 mW	$>10^{12}$
Memristor	1.2–2.0	150–190	<10 ns	nW	$\sim 10^9$
RF-SOI Switch	0.5–1.0	180	1–10 ns	µW	$>10^{10}$

5. Surface-Level Optimization, System Integration, and Advanced Schemes

RIS deployment in a wireless system entails several interconnected design and optimization layers:

• Joint precoder and phase optimization: Large-scale, nonconvex problems, particularly in MIMO, multi-user, or wideband regimes, with alternating optimization, semidefinite relaxation, manifold optimization, and data-driven (deep/reinforcement learning) methods as primary solvers (Xu et al., 2022, Liu et al., 2020, Pan et al., 2020).
• Real-time control and feedback: Hardware constraints, finite switching speeds, and phase quantization dictate trade-offs between achievable rate, adaptation speed, and complexity. Hybrid (active) RISs—embedding a small number of RF chains for partial baseband sensing or local computation—support autonomous channel estimation and self-configuration (Alexandropoulos et al., 22 Jul 2025).
• Channel estimation bottlenecks: The need to estimate high-dimensional cascaded BS–RIS–user channels drives research into ON/OFF pilot schemes, compressive sensing (when applicable), and tensor-decomposition-based overhead reduction (Xu et al., 2022, Liu et al., 2020).
• Physical-layer security and information-theoretic joint design: Integration of artificial noise, transmit covariance shaping, and RIS phase control for secrecy-rate maximization has been analytically shown to achieve unbounded asymptotic secrecy rates in $N$ (Xu et al., 2022).
• Holographic and near-continuous surfaces: High-density arrays enable advanced wavefront engineering (e.g., 2D holography, group-delay synthesis), but mutual coupling and finite element sizes require refined electromagnetic and circuit models (Shabanpour et al., 2023, Abbas et al., 2024).

6. Experimental Demonstrations, Measurement, and Deployment Guidelines

A broad set of experimental validations confirms key RIS concepts:

• Prototype beamsteering RISs at C-band demonstrate up to $\pm 60^{\circ}$ steering with average illumination efficiency of $\sim$95%, sidelobe level $\leq-11$ dB, and measured patterns matching full-wave simulations within 2° and 1 dB (Ataloglou et al., 8 Apr 2025).
• Large apertures of 400 bits (20$\times$20 arrays, 5.2 GHz) programmable via infrared light exhibit measured directivity near 26 dBi, with efficiency losses $\sim$4.7 dB and >25 dB link gain improvements (Sayanskiy et al., 2022).
• Passive pulse-width-modulated RISs achieve 7 dB link state differentiation without external power, motivated for low-complexity, low-power scenarios (Omori et al., 24 Oct 2025).
• Measurement campaigns across 4–33 GHz demonstrate beamforming, MIMO rank enhancement, channel reciprocity, and link gains up to 30 dBi in both indoor and outdoor environments (Huang et al., 2022).

Deployment studies—e.g., RISs on unmanned aerial vehicles—show RIS size, carrier frequency, and altitude impact coverage, secrecy, and flight time. For sub-6 GHz, $N=100$–$200$ elements balance performance vs. payload, while mmWave bands require $N\geq 300$ (Brighente et al., 2022). Roadside RISs modeled via stochastic geometry indicate rapid growth in vehicular coverage with modest RIS densities (1–5/km), especially at moderate SNR thresholds and vehicle-favored path-loss exponents (Choi et al., 2024).

7. Challenges, Limitations, and Design Implications

RIS implementation and deployment face significant open challenges:

• Channel estimation and feedback: In large arrays, CSI acquisition is a critical barrier; compression, hybrid, or self-sensing solutions are vital (Xu et al., 2022, Alexandropoulos et al., 22 Jul 2025).
• Hardware impairments and quantization: Discrete phase levels, amplitude-phase coupling, insertion losses, and switching latency limit practical array gain (Xu et al., 2022, Williams et al., 2024).
• Wave-optical constraints: Power-conservation and passivity impose fundamental restrictions on the amplitude-phase response; not all phase shifts are achievable with constant amplitude, and conventional cascaded channel models may fail for single/multi-user configurations with complex polarization or angular structures (Williams et al., 2024).
• Computational complexity: Realtime optimization for massive arrays, multi-user, multi-RIS settings remains a computational bottleneck. Data-driven (deep RL, federated learning) strategies offer promise but require domain-adapted architectures (Liu et al., 2020).
• Physical layer security, mutual coupling, and RCS limitations: Large element counts can alleviate amplitude-phase coupling and restore standard channel estimation, but mutual coupling and angular/polarization design must be rigorously considered for optimal array efficiency (Williams et al., 2024, Shabanpour et al., 2023).
• Integration and deployment: Issues such as wiring, power distribution for large panels, control signaling, and hybrid (reflect/sensing) operation must be systematically addressed, especially at mmWave/sub-THz regimes (Inácio et al., 28 Apr 2025, Alexandropoulos et al., 22 Jul 2025).

A summary of recommended practices:

• Use hardware with angularly stable phase response to ensure local reflection locality and efficient anomalous reflection (Shabanpour et al., 2023).
• Quantify insertion losses and phase-amplitude limitations according to the physical load and electromagnetic substrate.
• Design system-level solutions (precoder, phase set, allocation) cognizant of hardware constraints, overhead, and control speed.
• Exploit large $N$ and spatial coding to recover idealized performance in the large-array regime, while accommodating amplitude and phase variance at moderate $N$.
• For integrated reflection and sensing, hybrid RISs provide a path to autonomous self-configuration, reduced pilot overhead, and integrated communications/sensing with joint optimization (Alexandropoulos et al., 22 Jul 2025).

Reflective Intelligent Surfaces constitute a foundational technology for 6G wireless networks, enabling unprecedented control over the propagation environment. Realizing the full potential of RISs will require coordinated advances in electromagnetic design, hardware miniaturization, system-level optimization, intelligent control, and rigorous attention to physical constraints and scalability at all architectural layers.