PaperlayerView exact revision on arXiv

Source-linked AI summary

Emerging Technologies in Intelligent Metasurfaces: Shaping the Future of Wireless Communications

Jiancheng An, Mérouane Debbah, Tie Jun Cui, Zhi Ning Chen, Chau Yuen

arXiv:2411.19754v2cs.ITeess.SPmath.IT

TL;DR

This survey addresses the limited emphasis on 3D SIM and FIM in surveys that have primarily focused on 2D RIS. It reviews RIS experiments and applications, SIM principles and prototypes, and FIM technology, concluding with SIM's analog-processing potential and challenges for FIM integration.

  • Problem

    Prior surveys primarily focused on 2D RIS, while this paper addresses emerging 3D SIM and FIM technologies in wireless communications.

    Source: §I · p. 2

  • Method

    The paper surveys RIS experiments and applications, SIM operating principles and prototypes, and FIM technology and integration challenges.

    Source: §I · p. 2, §V · p. 14

  • Results

    The review presents numerical results illustrating SIM's potential for analog signal processing and surveys FIM's impact and integration challenges.

    Source: §V · p. 14

  • Takeaways & Limitations

    The survey offers a perspective on emerging RIS, SIM, and FIM technologies and identifies challenges for integrating FIMs into wireless networks.

    Source: §V · p. 14

    SIM inference is limited by its inherently linear transfer function, and precise modeling and device design for benign nonlinear responses remain to be investigated.

    Source: 1) Nonlinear Activation Function · p. 11

Abstract

from arXiv · show

Intelligent metasurfaces have demonstrated great promise in revolutionizing wireless communications. One notable example is the two-dimensional (2D) programmable metasurface, which is also known as reconfigurable intelligent surfaces (RIS) to manipulate the wireless propagation environment to enhance network coverage. More recently, three-dimensional (3D) stacked intelligent metasurfaces (SIM) have been developed to substantially improve signal processing efficiency by directly processing analog electromagnetic signals in the wave domain. Another exciting breakthrough is the flexible intelligent metasurface (FIM), which possesses the ability to morph its 3D surface shape in response to dynamic wireless channels and thus achieve diversity gain. In this paper, we provide a comprehensive overview of these emerging intelligent metasurface technologies. We commence by examining recent experiments of RIS and exploring its applications from four perspectives. Furthermore, we delve into the fundamental principles underlying SIM, discussing relevant prototypes as well as their applications. Numerical results are also provided to illustrate the potential of SIM for analog signal processing. Finally, we review the state-of-the-art of FIM technology, discussing its impact on wireless communications and identifying the key challenges of integrating FIMs into wireless networks.

I. INTRODUCTION

The introduction presents RIS, SIM, and FIM as distinct approaches to manipulating wireless signals or environments, then outlines a survey that emphasizes 3D SIM and FIM alongside RIS. The RIS section begins with experimental and modeling progress and considers four development perspectives.

I. INTRODUCTION

  • RIS coordinates amplitude and phase shifts to manipulate radio propagation on demand, offering lower power consumption than conventional relays.The introduction also describes RIS as lightweight and compact for integration with structures such as buildings and ceilings.Source: §I · pp. 1–2
  • SIM stacks programmable metasurfaces to process information-carrying electromagnetic waves directly, without digital storage, preprocessing, or external computing power.The introduction cites prototypes for MIMO beamforming, radar sensing, and image classification.Source: §I · p. 2
  • FIM morphs its 3D surface to adapt antenna positions to dynamic channels, helping signal copies from different paths add constructively.The introduction identifies millimeter-wave and terahertz communications as especially relevant because their channel coherence distance is relatively small.Source: §I · p. 2
  • The survey reviews RIS experimental testing and modeling, SIM principles, prototypes, applications and numerical advantages, and emerging FIM’s potential wireless-communication impacts; unlike prior surveys focused on 2D RIS, it emphasizes 3D SIM and FIM.Source: §I · p. 2

II. RECONFIGURABLE INTELLIGENT SURFACE (RIS)

  • The RIS section covers experimental testing and modeling, then examines applications from four perspectives.The introduction figure depicts the development trend from 2D RIS to 3D SIM and FIM.Source: §II · p. 2, Figure 2 · p. 3

A. Experiment and Modeling

This section surveys RIS prototypes and field trials, modeling approaches, development directions, and integration with transceivers. The examples include measured gains and proposed capabilities across laboratory and communication settings.

1) Prototype and Field Trials:

  • A 256-element RIS at 28.5 GHz achieved 19.1 dBi antenna gain, and a 1,100-element prototype provided a 26 dB power gain over a copper surface in an indoor test.The section also reports an outdoor average SNR improvement of more than 6 dB and a 3-bit meta-atom with stable responses over ±60° incidence angles.Source: 1) Prototype and Field Trials · p. 3
  • Other prototypes demonstrate space-time wavefront manipulation and metasurfaces that generate beamforming, double beams, and dynamic vortex beams.The reported space-time prototype has less than 0.6 dB insertion loss across a 200 MHz bandwidth.Source: 1) Prototype and Field Trials · p. 3

2) Modeling:

  • RIS modeling work includes physics-based tile models, physically feasible fading and reflection models, and free-space path-loss models with angle-dependent loss.Group- and fully-connected impedance-network architectures were reported to increase received signal power by more than 60% compared with single-connected networks.Source: 2) Modeling · p. 3

B. Development Trend

  • The four RIS development perspectives are deployment location, signal-enhancement capability, coverage area, and practical protocol.The figure frames these as environmental-to-transceiver, passive-to-active, reflective-to-transmissive, and estimation-to-training shifts.Source: B. Development Trend · pp. 3–4

1) From Environmental RIS to Transceiver RIS:

  • Transceiver-integrated RIS enables holographic MIMO, with near-field spatial degrees of freedom and a continuous aperture supporting spatial diversity.For 2D terminal deployments in a line-of-sight setting, the cited analysis gives a maximum spatial DoF of π/λ^2 per square meter.Source: 1) From Environmental RIS to… · p. 4

2) From Passive RIS to Active RIS:

The section surveys RIS changes intended to address severe two-hop path loss, limited coverage from reflective surfaces, and channel-acquisition overhead. These directions include active amplification, simultaneous transmission and reflection, and codebook-based training.

2) From Passive RIS to Active RIS:

  • Active RIS can amplify incoming signals as well as adjust phase, addressing severe two-hop path loss without relying solely on more passive elements.An experimentally verified active-RIS model supported a reported potential to double the sum rate for multiuser MISO communications.Source: 2) From Passive RIS to Active RIS · p. 4

3) From Reflective RIS to Transmissive RIS:

  • STAR-RIS and intelligent omnisurfaces transmit and reflect signals into the surrounding space, extending coverage beyond reflective-only RIS.The cited STAR-RIS protocols are energy splitting, mode switching, and time switching.Source: 3) From Reflective RIS to Transmissive… · p. 4

4) From Channel Estimation to Channel Training:

  • Codebook-based channel training configures RIS patterns from a predesigned codebook, estimates the composite channel, and uses the best-performing codeword for data transmission.This approach is presented to reduce pilot overhead and implementation complexity as estimation overhead grows with the number of RIS elements.Source: 4) From Channel Estimation to Channel… · pp. 4–5

III. STACKED INTELLIGENT METASURFACE (SIM)

The SIM section introduces its operating principle and surveys architectures, prototypes, functions, applications, and case studies. Its layered metasurfaces form an electromagnetic neural network whose configured transmission coefficients shape wave propagation and computation.

III. STACKED INTELLIGENT METASURFACE (SIM)

  • A SIM stacks programmable metasurface layers that act as hidden layers, with meta-atoms behaving like neurons whose transmission coefficients are trained by error backpropagation.The section situates this architecture within a broader survey of SIM prototypes, functions, applications, case studies, and future directions.Source: A. Operating Principle · p. 5, §III · p. 5

A. Operating Principle

  • Each meta-atom acts as a secondary source whose spherical wave illuminates the next layer, and each layer's field is the weighted sum of diffracted fields from the previous layer.The weights combine complex-valued transmission coefficients with propagation coefficients.Source: A. Operating Principle · p. 5
  • The SIM transfer function composes inter-layer propagation matrices and layer transmission matrices, and these matrix operations occur automatically as waves propagate.This lets SIMs perform complex signal-processing tasks without digital storage, transmission, or preprocessing, and supports parallel calculations for input electromagnetic signals.Source: A. Operating Principle · p. 5

B. Hardware Prototypes

SIM hardware prototypes span passive and reconfigurable devices for tasks from image recognition and logic to sensing and communication. In wireless systems, SIMs support wave-domain beamforming and channel estimation, with fully analog and hybrid designs addressing different system constraints.

B. Hardware Prototypes

  • SIM prototypes demonstrate image classification, logic operations, multi-beam focusing, DOA estimation, ISAC, and holographic image generation.Examples include a five-layer reconfigurable prototype and a visible-spectrum device using polarization multiplexing for parallel recognition.Source: B. Hardware Prototypes · pp. 5–6

C. Functionalities of SIM in Wireless Communications

  • SIM wave-domain processing is presented as a way to reduce communication-system hardware costs and energy consumption.The surveyed functionalities include beamforming-related matrix operations, channel estimation, and machine-learning-based design.Source: C. Functionalities of SIM in Wireless… · p. 6

1) MIMO Beamforming:

  • Fully analog SIM pairs perform MIMO precoding and combining in the wave domain, while hybrid architectures combine SIMs with low-complexity digital processors.The fully analog design targets a diagonal end-to-end channel; hybrid designs address scalability when more data streams require a larger surface aperture.Source: 1) MIMO Beamforming · pp. 6–7

C. Functionalities of SIM in Wireless Communications

  • A six-layer SIM with four RF chains achieved the same channel-estimation accuracy as a fully digital estimator with 64 RF chains.The surveyed approaches also include subspace-based estimators using channel statistics and sparse recovery for mmWave systems.Source: 1) MIMO Beamforming · p. 7

3) ML-based Design:

Machine-learning approaches are used to configure SIM phase shifts and power allocation in dynamic wireless settings. The section also points to Table II as a summary of SIM applications in wireless communication systems.

3) ML-based Design:

  • TD3-based joint phase-shift and power-allocation optimization achieves a higher sum rate than DDPG, while adding SIM layers yields diminishing returns.A separate DRL approach jointly optimizes phase shifts and transmit power for downlink multiuser systems, using whitening to improve robustness.Source: 3) ML-based Design · pp. 7–8
  • The surveyed ML methods also use meta-learning to adapt RIS-SIM resource allocation to user mobility while meeting terminal QoS requirements.The resource-allocation problem is modeled as a Markov decision process and optimized with a TD3 policy-gradient agent.Source: 3) ML-based Design · p. 8

D. Promising Applications

1) Cell-Free (CF) Massive MIMO:

The surveyed applications use SIMs for cell-free, near-field, satellite, security, and sensing-communication systems. They describe potential benefits including reduced processing delay, interference mitigation, and physical-layer security, alongside approaches tailored to each setting.

1) Cell-Free (CF) Massive MIMO:

  • SIM-enhanced cell-free massive MIMO studies target uplink spectrum and energy efficiency while addressing the hardware costs, energy use, and backhaul overhead of growing access-point deployments.Approaches optimize local SIM coefficients and receiver combiners, derive spectrum-efficiency expressions, and use two-stage capacity maximization.Source: 1) Cell-Free (CF) Massive MIMO · p. 8

2) Near-Field Communications (NFC):

  • In near-field multiuser MISO systems, a SIM-integrated base station can mitigate user interference while reducing processing delay.The design leverages spherical-wave channel characteristics; another study maximizes weighted sum rate by adjusting transmit power and SIM phase shifts.Source: 2) Near-Field Communications (NFC) · pp. 8–9

3) Low-Earth Orbit (LEO) Satellite:

  • A proposed LEO-satellite SIM would provide analog multiuser beamforming with lower processing delay and computing load than traditional digital beamforming.A separate aerial-SIM concept encodes geomorphic images into transmission coefficients and sends recognition results to a ground station.Source: 3) Low-Earth Orbit (LEO) Satellite · p. 9

4) Physical Layer Security (PLS):

  • A SIM can jointly perform modulation, beamforming, and artificial-noise generation to mitigate information leakage without excessive RF chains.The surveyed design fits desired output signals and iteratively optimizes phase shifts and transmit power.Source: 4) Physical Layer Security (PLS) · p. 9

5) ISAC:

  • Wave-domain ISAC precoding can generate a desired sensing beam pattern as multiple data streams pass through the SIM.Other designs optimize spectrum efficiency under sensing-power constraints or minimize the Cramér-Rao bound while meeting communication-user SINR requirements.Source: 5) ISAC · p. 9

6) DOA Estimation:

SIM applications include DOA estimation, image-classification semantic encoding, and hologram generation. The case study also reports that increasing SIM layers improves MIMO channel diagonalization and that SIM-based waveforms maintain constant-envelope behavior.

6) DOA Estimation:

  • A SIM designed to compute the 2D DFT lets a receiver array observe the angular spectrum and estimate DOA by probing its energy peak, without RF chains at each antenna.The surveyed work reports −40 dB MSE, and another approach reports angular resolution exceeding the diffraction limit.Source: 6) DOA Estimation · pp. 9–10

7) Semantic Communications:

  • A SIM performs source and semantic encoding in the wave domain for image recognition, achieving over 90% accuracy after training its meta-atom transmission coefficients.At the receiver, the image class is recognized by measuring signal magnitude across the receiving array.Source: 7) Semantic Communications · p. 10

8) Hologram Generation:

  • Over 90% of radiated power can be concentrated in a desired area with three continuous-valued SIM layers; two-valued phase shifts require nine layers for the same concentration.A tunable SIM with 96% transmission efficiency is also reported for multichannel information transmission and holographic image generation.Source: 8) Hologram Generation · p. 10

E. Case Studies

  • With four metasurface layers per SIM, the case study forms an almost perfectly diagonal MIMO channel, while SIM-based waveforms remain at 0 dB PAPR versus over 8 dB for conventional MIMO transmitting 16 streams.The PAPR comparison uses binary phase-shift keying and Rayleigh fading; the SIM result is attributed to the waveforms' constant-envelope property.Source: E. Case Studies · pp. 10–11

2) Semantic Encoding:

The section illustrates SIM-based semantic encoding and discusses limits on nonlinear inference and on evaluating SIM performance. It also identifies a need to characterize the tradeoff between multilayer signal loss and increased depth.

2) Semantic Encoding:

  • A four-layer SIM at 28 GHz directs image-bearing signals to corresponding receiving-array antennas, allowing image classes to be recognized from energy peaks.The EMNIST evaluation uses six letters and reports that this approach reduces hardware cost while improving processing speed.Source: 2) Semantic Encoding · p. 11

1) Nonlinear Activation Function:

  • SIM inference is limited by its linear transfer function; a hybrid optical-electronic neural network (HOENN) can use SIM preprocessing and receiver envelope detectors to support an electronic neural network.Integrating nonlinear components into meta-atoms is another proposed route, but precise modeling and device design with a benign nonlinear response need further investigation.Source: 1) Nonlinear Activation Function · p. 11

2) Performance Evaluation:

  • Evaluating SIM’s fundamental performance limits requires theoretical frameworks that characterize structure-function relationships and the tradeoff between multilayer signal loss and increased depth.The passage identifies semi-classical analytical models, heuristic optimization, and topology optimization as methods for characterizing EM systems.Source: 2) Performance Evaluation · p. 11

IV. FLEXIBLE INTELLIGENT METASURFACE (FIM)

FIMs use flexible substrates or active actuation to change their surface shape, with potential applications in wireless communications and other fields. The section reviews hardware prototypes, wireless functions, numerical results, and deployment challenges.

IV. FLEXIBLE INTELLIGENT METASURFACE (FIM)

  • The section reviews FIM prototypes, wireless functionalities, numerical case studies, and practical challenges for deployment in wireless networks.FIMs are presented as an approach to the practical limitations of conventional flat, rigid metasurfaces.Source: §IV · pp. 11–12

2) Active Morphing:

  • FIM morphing can position radiating elements to attain the highest channel gain, while the figures present diversity gain versus morphing range and channel gain versus eigenchannel index.The figure captions identify the comparisons but do not provide numerical outcomes.Source: Figure 10 · p. 12, Figure 11 · p. 12, Figure 12 · p. 12

1) Passive Morphing:

  • Passive FIMs can alter electromagnetic responses while conforming to curved surfaces, including reducing specular reflection and absorbing microwaves across multiple bands.Reported examples include a 10 dB radar-cross-section reduction across the X band on a coated cylindrical metal surface and 99% near-perfect absorption across Ku, K, and Ka bands when bent.Source: 1) Passive Morphing · p. 12

2) Active Morphing:

  • Active FIM prototypes use voltage-driven elastomer electrodes, liquid-metal phase transitions, programmed Lorentz forces, or ionic actuators to reconfigure surface shape.The Lorentz-force design uses digital control and in-situ stereo-imaging feedback to reach target shapes in milliseconds.Source: 2) Active Morphing · p. 12

B. Functionalities of FIM in Wireless Communications

FIM surface-shape optimization can improve MIMO capacity, reduce transmit power in multiuser systems, and increase diversity and multiplexing gains. The reported benefits depend on morphing range and system configuration.

1) MIMO Capacity Enhancement:

  • Under certain system configurations, FIM-aided MIMO channel capacity can double compared with conventional rigid arrays.The study jointly optimizes transmitting and receiving FIM shapes and transmit covariance under power-budget and morphing-range constraints.Source: 1) MIMO Capacity Enhancement · p. 13

2) Multiuser Interference Mitigation:

  • FIMs reduce required base-station transmit power by about 3 dB versus conventional rigid 2D arrays while maintaining the same data rate.The multiuser design alternately optimizes surface shape and transmit beamforming subject to user QoS and morphing-range constraints.Source: 2) Multiuser Interference Mitigation · p. 13

1) Diversity Gain:

  • An FIM morphing range of 0.8λ boosts diversity gain by 3 dB, while achieving a 6 dB gain may require 7.2λ, a substrate-elasticity challenge.In the evaluated correlated Rayleigh fading scenario, diversity gain increases logarithmically with morphing range.Source: 1) Diversity Gain · p. 13

2) Multiplexing Gain:

  • With a 0.1λ morphing range, weak eigenchannels gain over 40 dB versus rigid arrays; increasing the range to 0.5λ raises eigenchannel gain by about 20 dB.The larger range enables more eigenchannels to be used for data transmission.Source: 2) Multiplexing Gain · p. 13

D. Open Challenges

The paper identifies channel estimation, surface-shape morphing, and practical deployment as key challenges for FIMs in wireless networks. Its conclusion situates these challenges within an overview of RIS, SIM, and FIM technologies.

1) Channel Estimation:

  • FIM channel estimation is challenging because it spans continuous space and performance gains depend heavily on perfect CSI; estimation protocols still need development.Compressed sensing and space-alternating generalized expectation maximization are identified as possible techniques, and shape optimization may improve estimation accuracy.Source: 1) Channel Estimation · p. 13

2) Surface-Shape Morphing:

  • Practical FIM morphing may be constrained by hardware limitations and may add energy consumption, motivating efficient methods and robust shape optimization.For rapidly time-varying channels, optimizing ergodic performance across multiple coherence blocks may be beneficial.Source: 2) Surface-Shape Morphing · p. 13

3) Practical Deployment:

  • Network integration presents considerable challenges, and FIM material resilience, performance, and effective deployment protocols require further assessment.The paper mentions clothing, parachutes, and hot-air balloons for outdoor use, and curtains as indoor reflectors or access points.Source: 3) Practical Deployment · pp. 13–14

V. CONCLUSION

  • The paper surveys RIS experiments and applications, SIM principles and prototypes, and FIM technology, including numerical results for SIM and FIM integration challenges.The conclusion presents the overview as a perspective intended to inspire research toward efficient communication and computing.Source: §V · p. 14
Loading 2411.19754v2…