Intelligent Reflecting Surfaces

• Intelligent Reflecting Surfaces are engineered metasurfaces with programmable passive elements that control electromagnetic waves for deterministic channel reconfiguration.
• They enable passive beamforming, virtual line-of-sight creation, and interference suppression to boost spectral and energy efficiency in diverse network deployments.
• Optimization methods like alternating optimization and semidefinite relaxation tailor phase shifts to improve performance, security, and energy efficiency under real-world constraints.

Intelligent Reflecting Surfaces (IRS) are engineered metasurfaces comprising large arrays of sub-wavelength passive elements whose local electromagnetic boundary conditions can be electronically programmed. They introduce a new physical-layer paradigm by transforming the wireless propagation environment from a stochastic to a deterministic, programmable entity. IRSs enable fine-grained control of electromagnetic waves, facilitating passive beamforming, virtual link creation, interference suppression, and wireless channel reconfiguration—thus offering a promising route to boost spectral and energy efficiency, network coverage, and security in next-generation communication, sensing, and computing systems.

1. Physical Principles and System Models

An IRS typically consists of $N$ electronic or photonic meta-elements (e.g., patch antennas, resonators, varactor-loaded cells) arranged as a two-dimensional aperture on environmental surfaces such as building façades, lamp posts, or indoor panels. Each element applies a programmable complex reflection coefficient,

$$\Phi_n = \beta_n e^{j\theta_n}, \quad 0 \leq \beta_n \leq 1, \; \theta_n \in [0,2\pi),$$

where $\theta_n$ is the local phase shift and $\beta_n$ the (possibly quantized) reflection amplitude, often unity for lossless passive surfaces (Mahbub et al., 2022). Illumination by an incident wave yields re-radiated fields whose aggregate superposition, sculpted by the programmed $\{\theta_n\}$, enables constructive focusing towards intended receivers and destructive nulling for interference management.

For a basic single-input single-output (SISO) scenario, the baseband equivalent channel including direct and IRS-reflected paths is: $$h_{\text{eff}} = h_d + \mathbf{h}^T \boldsymbol{\Theta} \mathbf{g}$$ where $h_d$ is the direct link, $\mathbf{g}$ (BS-to-IRS) and $\mathbf{h}$ (IRS-to-user) are $N$-dimensional link vectors, and $$\boldsymbol{\Theta} = \mathrm{diag}(e^{j\theta_1}, \ldots, e^{j\theta_N})$$. For transmission of symbol $x$ with power $P_t$,

$$y = h_{\text{eff}} \sqrt{P_t}\,x + w,\quad w \sim \mathcal{CN}(0,\sigma^2)$$

Spectral and energy efficiency metrics follow directly; for instance,

$$\mathrm{SNR} = \frac{|h_\text{eff}|^2\,P_t}{\sigma^2}, \quad R = \log_2(1 + \mathrm{SNR}), \quad \mathrm{EE} = R/P_\mathrm{total},$$

with careful accounting of IRS control and signaling power (Mahbub et al., 2022).

In general multiuser and MIMO deployments, IRSs provide an additional layer of spatial multiplexing by reshaping the effective channel matrix. For $M$ elements,

$$h_\text{eff} = h_d + \sum_{m=1}^{M} g_m \beta_m f_m,$$

with $f_m$ and $g_m$ the respective Tx-to-IRS and IRS-to-Rx channels per element, enabling programmable array gain and link engineering (Bilgen et al., 30 Nov 2025). In multi-IRS chains or cascaded deployments, the system model generalizes to products of phase-shift matrices interleaved with channel matrices (Wu et al., 15 Jan 2025).

2. IRS Architectures, Deployment Modalities, and Channel Scaling

Architectures. Passive IRSs (unit-modulus, no RF chains or power amplifiers) offer negligible hardware noise and high energy efficiency, whereas active IRSs incorporate gain elements at the cost of higher noise figures and power consumption (Wu et al., 15 Jan 2025). Multistage (multi-IRS cascade) deployments augment the system's end-to-end degrees of freedom, with each hop contributing a phase-shift matrix and associated path loss.

Deployment:

• Point-to-Point: SISO/MIMO links with a single IRS panel can be analytically optimized in placement; high-performance is often achieved when the IRS is close to either the transmitter or receiver, maximizing aggregate channel gain via minimized double-hop path loss (Wu et al., 15 Jan 2025).
• Distributed/Networked: Partitioning $N$ elements among several panels enables fine-grained area coverage and spatial multiplexing, particularly effective in high-mobility or IoT device-dense scenarios (Cao et al., 2024, Bilgen et al., 30 Nov 2025).
• Virtual Link Creation: IRSs are uniquely effective in furnishing virtual line-of-sight (LoS) links for otherwise NLoS-blocked users or sensors, extended further via drone-mounted or mobile IRS platforms.

Channel Scaling: Passive array gain under LoS approximates $$\mathrm{SNR} \propto N^2/(d_{t\to\mathrm{IRS}}^2\,d_{\mathrm{IRS}\to r}^2)$$. With $M$ cascaded IRS stages, the potential passive gain grows as $O(N^{2M})$—subject to compounded path loss factorization—yielding dramatic potential capacity scaling, especially in mmWave/THz regimes (Wu et al., 15 Jan 2025, Wu et al., 20 Jun 2025).

3. Optimization Algorithms and Control Architectures

Beamforming and Phase-Shift Optimization: The principal IRS control problem is to jointly optimize the active transmitter beamformer and the IRS phase shifts to maximize spectral or energy efficiency under amplitude/phase quantization constraints. The standard problem formulations are: $$\max_{\{\theta_n\}} \log_2\left(1 + \frac{|h_d + \mathbf{h}^T \boldsymbol{\Theta} \mathbf{g}|^2 P_t}{\sigma^2}\right)$$ subject to $|e^{j\theta_n}|=1$ $\forall n$ (Mahbub et al., 2022), and analogously for network-level energy efficiency.

Solution Methodologies:

• Alternating Optimization (AO): Iteratively fix all but one block of variables and update the remaining via closed-form (single-element phase-alignment) or gradient-based rules (Mahbub et al., 2022, Björnson, 2021).
• Semidefinite Relaxation (SDR) / Majorization-Minimization (MM): Employ convex relaxations or surrogate constructions to handle non-convex unit-modulus constraints, often with rank-reduction or projection (Mahbub et al., 2022, Elbir et al., 2022).
• Successive Convex Approximation / Projected Gradient Descent: For large $N$, block-wise updates or manifold-based optimization for scalable real-time computation (Cao et al., 2024).
• Genetic/Evolutionary Algorithms: For high-combinatorial settings (e.g., NOMA-VLC with joint IRS-signal ordering), adaptive restart genetic algorithms can approach optimal performance (Abumarshoud et al., 2021).

Control and Protocol Architectures: At network scale, the Internet of IRS (IoIRS) paradigm embeds IRSs as IP-addressable network entities coordinated by IRS Stations (IRSS) and Servers. Control messages manage discovery, registration, scheduling, and coordinated phase reconfiguration, while user data traverse the IRS-augmented links transparently (Bilgen et al., 30 Nov 2025).

Strategy	Features	References
AO + Single-IRS	Joint beamforming, closed-form	(Mahbub et al., 2022)
SDR/MM	Non-convex constraints	(Mahbub et al., 2022)
AO + Distributed IRS	Channel rank boosting, low overhead	(Cao et al., 2024)
IoIRS IP Control Plane	Protocol suite, multi-IRS-mgmt	(Bilgen et al., 30 Nov 2025)

4. Performance Benefits and Experimental Validation

Field trials and large-scale simulations document substantial IRS gains across a variety of settings:

• Spectral Efficiency: Up to 60% improvement at the cell edge for $N=50$ elements in dense IoT microcells (Mahbub et al., 2022). Multi-IRS deployments exhibit capacity scaling up to $O(N^4)$ in double-reflection chains (Wu et al., 15 Jan 2025).
• Coverage and User Association: RSRP CDFs shift upward by 10–15 dB in network trials (2.6 GHz), and downlink throughput increases by factors 2–4 in field trials at 26 GHz, compared to non-IRS baselines (Wu et al., 15 Jan 2025).
• Latency and Energy Efficiency: Uplink end-to-end delay reduced by up to 35%, energy efficiency gains of 40% in IoT scenarios, primarily due to passive operation (Mahbub et al., 2022).
• Reliability in High-mobility/V2X: Position-aware phase computation and IRS element grouping provide multi-bps/Hz rate gains in vehicular networks with practical pilot overhead (Dampahalage et al., 2020).
• ISAC/Radar: IRS-assisted ISAC systems achieve 10 dB radar-SINR gains and 20–50% sum-rate increase in urban microcells; deep-RL IRS control halves target localization MSE in cluttered environments (Elbir et al., 2022, Meng et al., 2022).

5. Security, Privacy, and Sensing Applications

IRSs are potent enablers of advanced physical-layer security and covert communication:

• Secrecy Enhancement: By judicious phase and amplitude design, IRSs increase average secrecy rates and restrict information leakage by destructively interfering with eavesdropper channels. Notably, setting all $\beta_n=1$ is suboptimal; amplitude adaptation suppresses unintended propagation (Yan et al., 2021, Chen et al., 2019).
• Covert Communication: Precise IRS control increases the covertness probability, allowing transmitters to remain undetectable to wardens (Willies), with deep learning-based passive CSI estimation mitigating pilot-leakage (Yan et al., 2021).
• Sensing and ISAC: IRSs enable multi-target radar via programmable beampattern synthesis, supporting hybrid time-division or signature sequence sensing schemes that balance sensing frequency and beam gain. Joint beamforming and phase optimization deliver robust multi-target detection under mutual interference constraints (Meng et al., 2022, Elbir et al., 2022, Hua et al., 2022).

6. Implementation Challenges and Future Directions

Key Implementation Issues:

• Channel Estimation Overhead: Acquiring full CSI for IRSs with large $N$ is prohibitive; compression (e.g., compressed sensing, on-off element protocols), grouping, or model-based estimation are essential (Mahbub et al., 2022, Dampahalage et al., 2020).
• Hardware Impairments: Discrete phase quantization (2–3 bits sufficient), mutual coupling, and non-ideal losses impact real-world gains; robust optimization strategies are required (Mahbub et al., 2022, Wu et al., 15 Jan 2025).
• Control Latency, Synchronization, and Feedback: Low-latency, reliable control links are prerequisites for reconfiguration in dynamic environments, especially for mobile IRSs and adaptive networks (Mahbub et al., 2022, Bilgen et al., 30 Nov 2025).
• Environmental Dynamics: Blockage, time-variant scattering, weather, and aging necessitate adaptive placement, real-time reconfiguration, and robust materials (Wu et al., 15 Jan 2025).

Frontiers:

• AI/ML for IRS control and channel prediction, including federated and reinforcement learning frameworks (Wu et al., 15 Jan 2025, Cao et al., 2024).
• Distributed and cell-free IRS networks for ultra-dense connectivity (Bilgen et al., 30 Nov 2025).
• Joint IRS design for sensing, backhaul, and computation offloading (Elbir et al., 2022, Siavoshani et al., 2021).
• Near-field regime exploitation for multi-user spatial multiplexing (Wu et al., 20 Jun 2025).
• Optical and THz implementations, including passive beamforming for OWC/VLC/FSO and programmable sounders at >100 GHz (Wu et al., 20 Jun 2025, Sun et al., 2022, Hamad et al., 2024).

7. IRS in Optical Wireless and THz Communications

IRS concepts extend directly to the optical (OWC, VLC, FSO) and terahertz domains:

• FSO/VLC Integration: IRS mirrors or programmable optical meta-surfaces enable non-line-of-sight link formation, mitigate building sway, and substantially increase summative achievable rates in block-prone laser-based OWC environments (Hamad et al., 2024, Sun et al., 2022, Abumarshoud et al., 2021, Najafi et al., 2020, Najafi et al., 2019).
• THz Systems: THz IRSs exploit advanced reconfiguration mechanisms (electronic, optical, PCM, MEMS) and account for beam squint and near-field propagation. Liquid–crystal-based prototypes at 220 GHz deliver $>$15 dB power gain and sub-10% EVM for multi-user 16-QAM, validating IRSs as critical enablers in ultra-high-frequency wireless networks (Wu et al., 20 Jun 2025).

---

IRSs fundamentally alter the design landscape for wireless, sensing, and computation infrastructure by providing a low-cost, energy-efficient, and software-defined means to program the electromagnetic environment. Their integration across the communication stack, from physical to network layers, and their application across mmWave, THz, and optical domains, position them as a cornerstone of 6G and beyond (Mahbub et al., 2022, Bilgen et al., 30 Nov 2025, Wu et al., 15 Jan 2025, Wu et al., 20 Jun 2025, Cao et al., 2024).