Intelligent Reflecting Surface (IRS)

Updated 30 March 2026

Intelligent Reflecting Surface (IRS) is a programmable 2D metasurface composed of sub-wavelength, tunable elements that dynamically control phase and amplitude to shape wireless signals.
IRS applications span 6G communications, sensing, energy transfer, and security, enabling passive beamforming and significant improvements in link quality and coverage.
Optimization methods such as alternating optimization, semidefinite relaxation, and deep learning ensure efficient IRS configuration and real-time adaptation in complex wireless environments.

An intelligent reflecting surface (IRS) is a planar two-dimensional metasurface comprising an array of sub-wavelength, electronically configurable elements. Each element can dynamically control the phase and/or amplitude of incident electromagnetic waves, enabling programmable manipulation of the wireless propagation environment through passive beamforming, wavefront shaping, or scattering control. IRS technology is forecast to be a foundational enabler for 6G and beyond, with applications spanning wireless communications, sensing, energy transfer, security, and integrated wireless systems.

1. Physical Principles and Mathematical Modeling

IRSs consist of $N$ sub-wavelength elements (meta-atoms), typically metallic/dielectric patches loaded with tunable components such as PIN diodes, varactors, or phase-change materials. Each element imposes a complex reflection coefficient $\theta_n = \beta_n e^{j\phi_n}$ , where $\beta_n \in [0,1]$ denotes amplitude and $\phi_n \in [0, 2\pi)$ is the programmable phase shift (Yan et al., 2021, Shao et al., 2023, Wu et al., 14 Nov 2025). In most practical designs, $\beta_n$ is set to unity for lossless operation.

The IRS interaction with the radio channel is typically modeled by a diagonal reflection matrix: $\Theta = \mathrm{diag}( \beta_1 e^{j\phi_1}, \ldots, \beta_N e^{j\phi_N} )$ For a transmitter–IRS–receiver link in a narrowband far-field regime, the composite scalar channel is: $h_\text{eff} = \sum_{n=1}^N h_{r,n} \beta_n e^{j\phi_n} h_{t,n}$ where $h_{t,n}$ and $h_{r,n}$ are the channels from transmitter to $n$ th IRS element, and from IRS to receiver, respectively (Wu et al., 14 Nov 2025). In matrix notation for MIMO systems: $\mathbf{y} = \left( \mathbf{h}_d + \mathbf{H}_r \Theta \mathbf{h}_v \right) x + \mathbf{n}$ Here, $\mathbf{h}_d$ is the direct path, $\mathbf{H}_r$ and $\mathbf{h}_v$ represent BS–IRS and IRS–user channels, and $\Theta$ is the IRS phase matrix (Dampahalage et al., 2020).

IRS phase and amplitude configuration can be continuous or discretized; 2–3 bits phase quantization typically suffices for >90% performance recovery (Dampahalage et al., 2020, Cao et al., 2024).

2. Implementation Architectures and Design Variants

2.1 Hardware Architectures

Passive IRS: All elements offer tunable amplitude/phase but no active RF chains, achieving lowest energy consumption (Noor-A-Rahim et al., 2022).
Semi-Passive IRS: A subset of elements includes sensors for limited RF reception, enabling in-situ channel acquisition for enhanced sensing (Shao et al., 2023).
Active IRS: Meta-atoms feature low-noise amplification to enable amplification and programmable gain, at increased power and hardware complexity (Wu et al., 14 Nov 2025).
THz-IRS Architectures: At terahertz frequencies, IRS can be based on semiconductor (CMOS, Schottky), optical (laser-pumped), phase-change (VO₂, GST), or MEMS-based technologies, each with specific trade-offs in bandwidth, switching speed, and loss (Wu et al., 20 Jun 2025).

2.2 Topological and Functional Variants

Distributed IRS arrays: Multiple spatially distributed IRSs collaboratively extend coverage and multiplexing rank (Cao et al., 2024).
Target-mounted IRS: IRS is installed on moving objects to increase radar cross-section or modulate echo for secure sensing (Wei et al., 2022, Wu et al., 14 Nov 2025).
STAR-IRS and BD-IRS: Advanced forms support full-space coverage (simultaneous transmission and reflection) and sectorized/multisector architectures for full-angle operation (Wu et al., 14 Nov 2025).
Holographic IRS: Continuous-phase profiles yield holographic control, approaching the continuous aperture limit (Wu et al., 14 Nov 2025).

3. Optimization and Signal Processing Methodologies

Joint design of IRS phase shifts and active beamforming is generally a non-convex, high-dimensional problem due to the unit-modulus and coupled nature of the IRS coefficients.

3.1 Alternating and Manifold Optimization

Alternating Optimization (AO): Iteratively updates the IRS phase shifts and active beamformers; each subproblem is convex (SOCP) or lies on a Riemannian product of circles (manifold optimization) (Cai et al., 2021, Wu et al., 14 Nov 2025, Cao et al., 2024).
Semidefinite Relaxation (SDR): Lifts quadratic optimization over phases to semidefinite form, then extracts near-optimal unit-modulus solutions via randomization (Yan et al., 2021, Chen et al., 2019).
Quadratic Transform and Penalty Approaches: Decompose multivariable-coupled problems in IRS-aided ISAC for tractable solutions (Zhu et al., 2022, Hua et al., 2022).
Meta-Learning and Deep Reinforcement Learning: Applied for reduced-latency, near real-time control in dynamic environments (Noor-A-Rahim et al., 2022, Wu et al., 14 Nov 2025).

3.2 Channel Estimation and Low-Overhead Protocols

Element Grouping: Neighboring IRS elements are grouped and controlled jointly to reduce pilot overhead from $O(N)$ to $O(G)$ , with $G \ll N$ (Dampahalage et al., 2020).
Position-Based Passive Beamforming: For slowly moving devices, geometric knowledge is used to infer main path angles, minimizing training (Dampahalage et al., 2020).
Hierarchical Codebook and Beam-Training: Coverage of potential directions with a small number of preset phase patterns for rapid acquisition (Wu et al., 2021).
Deep Learning for Passive Channel Estimation: Scene images can be mapped to channel estimates via CNNs for pilot-free operation (Yan et al., 2021).

4. Application Domains and Performance Benchmarks

IRSs are critical in several emerging application classes, each with domain-specific design considerations.

4.1 Integrated Sensing and Communication (ISAC)

IRS enables simultaneous communication and high-resolution radar sensing by providing additional spatial DoF and power gains (Wu et al., 14 Nov 2025, Wei et al., 2022). Mounting an $M$ -element IRS on a target increases monostatic ISAC radar SNR by $20 \log_{10} M$ dB and reduces CRLBs for range and velocity by $1/M^2$ (Wei et al., 2022).

4.2 Vehicular and mmWave/THz Communications

In vehicular mmWave systems, large IRSs ( $16\times16$ or $256$ elements) provide tens of dB link improvement for vehicles in NLoS positions. Low-phase-resolution designs (2–3 bits) suffice (Dampahalage et al., 2020). In THz bands (220 GHz), prototype systems attain $>15$ dB SNR gains and enable multi-user QAM demodulation (Wu et al., 20 Jun 2025).

4.3 Secure and Covert Communications

IRS enhances physical layer security through energy-focusing (for the legitimate receiver) and energy-nulling (for the eavesdropper), significantly boosting secrecy rates. The covertness of communications is quantifiably improved, with achievable covert rates and distances increasing with IRS size and noise uncertainty (Lu et al., 2019, Yan et al., 2021, Chen et al., 2019, Zheng et al., 2024). Optimization can include both amplitude and phase of IRS elements for secrecy, with amplitudes $\beta_n<1$ optimal under stringent covertness constraints (Yan et al., 2021).

4.4 Energy Transmission, SWIPT, and WPCN

IRSs are highly effective in wireless energy transfer, with received power scaling as $N^2$ for $N$ elements, extending practical range and improving SWIPT throughput regions (Wu et al., 2021). In wireless powered communication networks (WPCN), IRS control enables efficient time and energy allocation protocols.

4.5 Smart Manufacturing and Industry 5.0

IRSs are seen as essential enablers for URLLC, supporting collaborative robotics, digital twins, and AR maintenance, by achieving coverage improvements of $10\!-\!20$ dB, sub-100 $\mu$ s latencies, and ultra-high reliability targets (Noor-A-Rahim et al., 2022).

4.6 Free Space Optical (FSO) Systems

IRS can be adapted as a passive phase-control surface for FSO, replacing LOS path requirements by anomalous reflection, with controlled geometric and misalignment loss models for robust system design (Najafi et al., 2020, Takimoto et al., 2024).

5. Protocols, Networking, and Large-Scale Deployment

As IRS deployment scales into thousands of units, networking protocols evolve to support real-time, scalable control (Bilgen et al., 30 Nov 2025). The Internet of Intelligent Reflecting Surfaces (IoIRS) framework introduces a layered architecture, allocating centralized and distributed control, physical-layer configuration, and application-layer service requests. Protocols are tailored for discovery, resource allocation, and adaptation under network control, achieving up to $1.6\times$ sum-throughput, sub-ms reconfiguration, and energy/bit reductions (Bilgen et al., 30 Nov 2025). Multi-IRS deployment enables cooperative mesh, resilience to blockages, and multi-hop coverage.

6. Design Challenges and Future Directions

IRS research faces several open challenges:

Channel Acquisition and Hardware Impairments: Acquiring high-dimensional cascaded channels remains a bottleneck; techniques include low-overhead training, grouping, sensing on the IRS, and learning-based approaches (Dampahalage et al., 2020, Wu et al., 2021, Wu et al., 14 Nov 2025).
Control Complexity & Optimization: High-dimensional, non-convex AO/BCD or SDR formulations are computationally intensive. Heuristic, closed-form, or partitioned (e.g., dual-beam) schemes are proposed for real-time adaptation (Jiang et al., 2022, Wu et al., 20 Jun 2025, Cao et al., 2024).
Resource Allocation and Network Coordination: IoIRS and ML-aided distributed control mechanisms become essential as networks densify and embrace non-terrestrial (UAV, satellite) deployments (Bilgen et al., 30 Nov 2025).
Robustness and Security: Designs must handle CSI uncertainty, phase quantization, mutual coupling, and adversarial scenarios such as eavesdropping or active attacks (Yan et al., 2021, Hua et al., 2022).
Integration with Sensing and Powering: ISAC, ISCAP, and 6D-IRS (joint position, orientation, and phase control) are active areas for system-level integration and cross-modal joint optimization (Wu et al., 14 Nov 2025).
Prototype and Standardization: Advances in cost-effective (e.g., inkjet-printed paper-based) IRSs, as demonstrated at GHz frequencies, pave the way for broad, scalable physical-layer deployments (Takimoto et al., 2024).

Ongoing and anticipated research aims to resolve these issues through novel architectures, AI-driven control, robust optimization, and experimental validation, shaping IRS as a central technology in programmable, intelligent wireless environments for 6G and beyond.