Intelligent Reflecting Surfaces (IRS)

• Intelligent Reflecting Surfaces (IRS) are programmable metasurfaces with reconfigurable electromagnetic responses that enable deterministic channel shaping and adaptive beam steering.
• IRS applications span RF to optical domains, supporting energy transfer, security enhancements, and integrated sensing and communication functions.
• Empirical studies demonstrate that IRS prototypes can boost received power by up to 15 dB and improve modulation reliability in high-frequency wireless networks.

Intelligent reflecting surfaces (IRS) are reconfigurable metasurfaces composed of subwavelength elements whose local electromagnetic response (phase, amplitude, and sometimes polarization) can be programmable in real-time. By imposing spatiotemporally varying boundary conditions on incident electromagnetic waves, IRSs enable deterministic channel shaping, beam steering, interference nulling, and interaction with both RF and optical signals, across a wide frequency range from microwave to THz and beyond. This technology is entrenched at the physical layer but, as network densification and scale expands, is also emerging as an integral network node with protocol and control-plane functions.

1. IRS Physical Architectures and Element Models

IRS implementations span several fundamental classes, with specific trade-offs and practical models:

• Electronic tuning (CMOS, HEMT/Schottky, graphene): Voltage-controlled devices offer nanosecond-scale switching, moderate insertion loss at THz (2–5 dB per meta-atom), and large-scale digital programmability (up to 8-bit phase resolution), but require sophisticated biasing and wiring infrastructure. Example: 65 nm CMOS metasurface with 2×2 arrays and 576 meta-atoms allows ±30° phase tuning at 0.3 THz.
• Optical tuning (photoconductive semiconductors): Illumination-induced carrier population enables picosecond reconfiguration and fully contactless control, at the cost of external laser dependence and fine-granularity constraints imposed by optical diffusion.
• Phase-change materials (PCMs): Materials such as VO₂, GeSbTe, and liquid crystals provide reversible amorphous–crystalline state transitions, enabling non-volatile control with sub-millisecond switching, but typically with 2–4 distinct states and significant (>10 mW) heating energy per switch.
• MEMS-based tuning: Micro-cantilevers or moveable plates achieve high-fidelity analog phase control and <1 dB loss but introduce latency (µs–ms) and reliability/practicality issues in scaling.
• Hybrid designs: Electromechanical plus PCM integration can achieve low-loss tuning, non-volatility, and fast operation by combining physical mechanisms.

Each IRS element is described by a frequency-dependent reflection coefficient $\Gamma(f) = \alpha(f) e^{j\theta(f)}$, where $\alpha(f)\in[0,1]$ encodes amplitude (losses) and $\theta(f)\in[0,2\pi)$ is the programmable phase. Frequency dispersion, quantization and switching latency, spatial coupling, and substrate influence are the dominant nonidealities in real-world designs (Wu et al., 20 Jun 2025).

2. Fundamental IRS-Aided Channel Models

The IRS scene is modeled by cascaded MIMO channels between transmitter (e.g., BS) and receiver (e.g., user or target) via the IRS, with explicit frequency, spatial, and propagation-regime dependence:

• Far-field: Both BS→IRS and IRS→user links are plane-wave dominated, with array steering vectors for each node; total link is described by a double outer product and frequency-dependent reflections.
• Near-field: For macro-scale IRSs, spherical wavefronts must be accounted for; each IRS element is uniquely parameterized by distance and relative angle, yielding elementwise channel terms with strong spatial diversity.
• Beam squint: In wideband and THz scenarios, fixing IRS phases for the carrier frequency introduces "beam squint," i.e., frequency-dependent angular deviation and gain degradation, constraining coherent combining only to narrowband regimes (Wu et al., 20 Jun 2025).

The effective end-to-end channel at subcarrier $f_m$ is

$$H_{\rm eff}(f_m) = H_{IU}(f_m) \cdot \text{diag}\{\Gamma_n(f_m)\}_{n=1}^M \cdot H_{TI}(f_m).$$

3. IRS-Driven Beamforming, Channel Estimation, and Optimization

Passive Beamforming and SNR Maximization

The canonical IRS beamforming problem is nonconvex due to the unit-modulus constraints per element:

$$\max_{g, w, \{\Gamma_n\}} |w^H H_{IU} \text{diag}(\Gamma) H_{TI} g|^2, \quad \text{subject to } | \Gamma_n |=1.$$

Closed-form single-stream optimal phases are given by per-element phase alignment:

$$\theta_n = \arg \left( [w^H H_{IU}]_n \cdot [H_{TI} g]_n \right)^*.$$

Multi-stream, multi-user, and wideband cases require alternating optimization (AO) with projected gradient descent or manifold-based updates, and, for wideband, joint phase-time-delay optimization to combat beam squint.

Channel Acquisition

IRS-assisted wireless requires estimation of high-dimensional cascaded channels. Three main families of estimation protocols are recognized:

• Pilot-based: Sequential ON/OFF or phase-state cycling, multiplexed pilots, and compressive sensing for sparse/structured scenarios; pilot overhead scales with number of elements unless element grouping/compression is used.
• Beam training: Codebook-based hierarchical search in spatial domains; applicable when angular parameters are slowly varying.
• Beam-squint-assisted: Leverages natural subcarrier squint (e.g., OFDM) to sense directions and ranges with minimal pilot overhead.

Low-complexity estimation via element grouping or position-informed parametric models is essential in high-mobility, large-scale (especially vehicular) deployments (Dampahalage et al., 2020).

4. Deployment and System-Level IRS Considerations

IRS performance scaling is dictated by:

• Path loss: Double-Friis (cascaded) scaling proportional to $d_1^2 d_2^2$ between BS–IRS and IRS–user/target, and $f^4$ frequency penalty; beamforming gain ideally scales as $M^2$.
• Placement and scale:
  • Large-scale IRS: Macro-area panels deployed near BS or users, distributing IRSs increases rank and spatial diversity but may exacerbate near-field effects and squint.
  • Small-scale IRS: Wavelength-scale, mobile (e.g., drone-mounted) IRSs enhance dynamic blockage mitigation and agile coverage.
  • Dual-scale networking: Hybridization of large fixed and small mobile IRSs supports coverage extension and on-demand NLoS pathway synthesis.
  • Multi-hop IRS relaying: Path selection using graph algorithms and, in some scenarios, active elements for overcoming the compounded attenuation.

5. Prototyping Results: Empirical Performance at THz

An 80-element liquid-crystal IRS operating at $f_0=220$ GHz demonstrated (Wu et al., 20 Jun 2025):

• Single-user: ≥15 dB received power improvement across all steering angles; error-vector magnitude (EVM) improved from >12% (no IRS, unrecoverable for 16QAM) to <8% (with IRS, meeting 16QAM standards).
• Multi-user: Independent sub-beams attained 10–14 dB gain over IRS-off case; EVM improvement sustained with slight penalty vs. single-user case; aperture size remains critical for multi-user separation.
• Aperture/Prototype characteristics: 1 GHz bandwidth, meta-atom pitch ≈ λ/3, switching latency <10 ms, high spatial scanning accuracy.

These real-world results validate the IRS concept for robust, high-capacity THz networking.

6. Beyond RF: Optical and Hybrid IRS Systems

IRSs are applicable not only in RF but also in optical (free-space optical, visible light, laser-based OWC) and hybrid network settings:

• Free-Space Optical (FSO): IRSs act as programmable mirrors, with design optimizations for aperture size, orientation, and placement, while robust channel models quantify geometric and misalignment losses under stochastic sway (Najafi et al., 2019, Najafi et al., 2020).
• Visible Light Communication (VLC): IRSs offer amplitude (and possibly phase) control in arrays to redirect and focus optical signals, with significant performance improvements—even under orientation randomness and link blockage—when joint with multiple access (NOMA) and optimization via metaheuristics (Abumarshoud et al., 2021).
• Laser-based OWC: Phase-aligned passive IRSs in laser-based angle-diversity networks deliver up to 71% sum-rate increase over IRS-free baselines, enhance spatial resilience, and strictly respect passive eye-safety constraints (Hamad et al., 2024).

7. IRS in Advanced Wireless Functions: Energy, Security, and ISAC

IRSs serve as technological building-blocks for emerging high-layer capabilities:

• Wireless energy and SWIPT: IRS-assisted RF power transfer, wireless-powered communication, and joint info/energy optimization depend heavily on deployment geometry, reflection codebook design, and energy-feedback protocols to maximize harvested energy and information rates (Wu et al., 2021).
• Physical-layer security and privacy: IRSs enable programmable attenuation/scattering to maximize secrecy rate and support covert transmission, requiring joint amplitude-phase optimization under both perfect and uncertain CSI; deep learning–based passive channel estimation provides non-pilot covert acquisition (Yan et al., 2021, Chen et al., 2019).
• Integrated sensing and communications (ISAC): IRSs enhance target illumination for radar/classical estimation while enabling joint communication, with AO-based co-design trading off sum-rate and sensing accuracy. Prototype and theoretical studies find up to 80% reductions in radar beampattern error and >30% improvement in communications sum-rate (Wu et al., 14 Nov 2025, Zhu et al., 2022).
• Network-layer integration: The "Internet of IRS (IoIRS)" paradigm generalizes IRSs as hybrid nodes with IPv6-based identification and control, forming a dynamically addressable network fabric for distributed EM environment programming, resource scheduling, and cross-layer optimization (Bilgen et al., 30 Nov 2025).

---

IRSs constitute a foundational enabler for truly programmable, high-frequency, and tightly integrated wireless networks. Advanced metasurface engineering, robust signal processing, and unified optimization frameworks now corroborate both the feasibility and substantial performance gains of IRSs across RF and optical communications, navigation, sensing, energy transfer, and security domains (Wu et al., 20 Jun 2025).