Fluid-Antenna Systems Overview

• Fluid-antenna systems are reconfigurable antennas that use fluids, movable metallic pixels, or meta-atoms to dynamically vary their radiating aperture.
• They leverage architectures like liquid-metal channels and pixel arrays to enable adaptive beamforming, interference avoidance, and optimal channel sampling.
• FAS enhance wireless performance in 6G, IoE, and UAV networks by providing improved diversity, capacity, and energy harvesting through dynamic reconfigurability.

Fluid-antenna systems (FAS) constitute a class of reconfigurable antennas whose radiating elements are realized by conductive or dielectric fluids, movable metallic pixels, or meta-atom arrays. The defining feature is dynamic spatial and/or shape reconfigurability of the radiating aperture, enabling the antenna to alter its position, geometry, or feeding configuration in real time. This flexibility fundamentally distinguishes FAS from traditional fixed-element arrays and underpins a suite of new spatial, spectral, and physical-layer capabilities, including selection-based spatial diversity, adaptive beamforming, interference avoidance, and optimal channel sampling—even within highly compact form factors. FAS are poised to play a pivotal role in future 6G and beyond, IoE, ISAC, SWIPT, and UAV-centric networks due to their ability to combine multiplexing, diversity, and application-adaptable operation within a single, software-controlled RF aperture (Wu et al., 2024, Hong et al., 16 Jun 2025, Lu et al., 6 Jan 2025).

1. Physical Principles and Enabling Architectures

FAS rely on reservoirs, microfluidic channels, arrays of conductive meta-atoms, or programmable pixel layers that can reconfigure the effective aperture by moving (or activating) the radiating medium among a discrete or continuous set of ports (Wu et al., 2024, Zhang et al., 2024). The system may involve:

• Liquid-metal antennas: Conductive fluids (e.g., Galinstan or eutectic gallium–indium) are displaced using micro-pumps, shape-memory-actuator mechanisms, or electrostatic control along millimeter-scale channels, forming the active radiating element at selectable positions (Wu et al., 2024, Lu et al., 6 Jan 2025, Psomas et al., 2023).
• Pixel-based or meta-fluid arrays: Arrays of metallic patches (“pixels” or “meta-atoms”) with electronic switches (e.g., PIN diodes) form a rapid-switching FAS. These can provide dozens to hundreds of distinct radiating states over a λ to few λ aperture, with state transitions on the μs timescale (Zhang et al., 2024, Liu et al., 15 Sep 2025).
• Dielectric- or phase-change media: The effective dielectric loading is adjusted by flowing or displacing fluids of variable permittivity around a metallic trace (Wu et al., 2024).
• Feed control: The excitation is switched among ports/programmed pixels, or spatial feeding profiles are induced, to selectively excite basis eigenmodes of the physical aperture (Lu et al., 6 Jan 2025).

The architecture may be realized in 1D (linear tracks), 2D (planar matrices), or via programmable meta-surfaces (e.g., meta-fluid antennas) (Liu et al., 15 Sep 2025).

2. Mathematical Models and Channel Characterization

FAS channel models must account for small-scale fading, spatial correlation, and fluid/material geometry:

• Jakes-type spatial correlation: For a linear (or 2D) FAS with N ports distributed over length Wλ, the channel vector $\mathbf{h} = [h_1,\ldots,h_N]$ exhibits spatial autocorrelation

$$\mathbb{E}[h_n h_m^*] = J_0\left(2\pi \frac{|n-m| W}{N-1}\right)$$

where $J_0$ is the zeroth-order Bessel function (Hong et al., 16 Jun 2025, Lu et al., 6 Jan 2025).

• Eigenmode expansion: Antenna behavior is governed by eigenmodes $\mathbf{E}_n$ (solutions of the vector wave equation under reconfigurable boundary conditions), with real-time adaptation via spatial boundary or feeding changes (Lu et al., 6 Jan 2025).
• Selection combining gain: The FAS selects the port $n^* = \arg\max_n |h_n|^2$, yielding an effective received SNR $\gamma_{\max} = \max_n |h_n|^2 P / \sigma^2$, and achieving diversity order typically limited by the effective rank of the spatial correlation matrix, not just N (Zhu et al., 10 Sep 2025).
• Continuous motion: For continuous FAS (CFAS), the SIR or SNR process is a stationary random field $g(x)$ with correlation $J_0(2\pi (x-x')/\lambda)$; level-crossing rate (LCR) and fade-duration (AFD) statistics can be derived in closed form (Psomas et al., 2023).

Nontrivial extension to finite-scattering (geometric) and wideband models is achieved by representing the FAS as a beamformer over a set of tracked rays or via a 2D/3D field-response (Hong et al., 7 Mar 2025).

3. Performance Limits: Diversity, Capacity, and Outage Behavior

FAS offer unconventional spatial diversity and robust outage performance, subject to spatial correlation and aperture constraints:

• Outage probability: For Rayleigh fading, the FAS post-selection outage is

$$P_{\text{out}}(\gamma_{\mathrm{th}}) = \Pr(\max_n |h_n|^2 P / \sigma^2 < \gamma_{\mathrm{th}})$$

For uncorrelated branches, this yields classic order-N diversity; when ports are highly correlated (small W), the performance saturates to an effective diversity order $N_{\text{eff}}$ set by the channel eigenvalue spectrum (Khammassi et al., 2022, Zhu et al., 10 Sep 2025).

• Ergodic capacity: The selection gain enhances ergodic capacity, with increases up to ~70% over fixed single-port antennas in representative scenarios, especially for moderate N and λ-scale apertures (Wu et al., 2024, Wong et al., 2020).
• Diversity scaling law: Asymptotic analysis establishes that the FAS diversity gain, coding gain, and rate improvement all scale with the effective spatial rank $$\mathbb{E}[h_n h_m^*] = J_0\left(2\pi \frac{|n-m| W}{N-1}\right)$$0 (for normalized aperture W in wavelengths) regardless of additional port density beyond this threshold (Zhu et al., 10 Sep 2025).
• Saturation effect: Increasing N for fixed W ultimately yields diminishing returns; only enlarging the aperture can substantially increase $$\mathbb{E}[h_n h_m^*] = J_0\left(2\pi \frac{|n-m| W}{N-1}\right)$$1 and thereby further enhance diversity and error exponent (Zhu et al., 10 Sep 2025, Khammassi et al., 2022).
• Comparison with MRC: FAS with sufficient N and/or W can outperform conventional multi-antenna MRC in both outage and capacity, using only a single RF chain (Wong et al., 2020, Wong et al., 2020).

4. Adaptive Beamforming and Reconfigurable Eigenmode Control

Unlike rigid phased arrays or RIS, FAS leverages shape, boundary, and feed reconfiguration at the resonant eigenmode level:

• Eigenmode-resonant beamforming: Continuous adaptation of eigenmodes via spatially tuned boundary conditions (Dirichlet/Neumann/Robin) or feeding positions enables flexible beam steering and null placement without resorting to phased array summation, resulting in highly agile and hardware-efficient beam patterns (Lu et al., 6 Jan 2025).
• Parity and modal symmetry: FAS exploit modal parity (even/odd) and symmetry axes as additional resonant degrees of freedom, allowing for tailored far-field patterns (e.g., dipole vs. hoop) and on-demand nulls or multibeam forms with minimal hardware (Lu et al., 6 Jan 2025).
• Prototype demonstrations: Plasma-lamp, microfluidic, or pixel-based FAS designs have validated wide-area radiation pattern reconfiguration (~10–30 dB beam/null agility) across broad frequency bands (2.4–30 GHz) in both liquid and non-liquid implementations (Lu et al., 6 Jan 2025, Zhang et al., 2024).
• Ultra-fast reconfiguration: Pixel/meta-fluid systems with PIN-diode switching achieve μs-scale state changes, supporting packet-to-packet adaptation under rapid channel fading (Zhang et al., 2024, Liu et al., 15 Sep 2025).

5. Application Frameworks: Networking, Sensing, Localization, and SWIPT

FAS unlock a range of novel networking and sensing architectures:

• Fluid antenna multiple access (FAMA): Multi-user interference mitigation via dynamic port selection—slow FAMA (block-based) or fast FAMA (symbol-based)—enables CSI-free, scalable, collision-resilient access in ultra-dense networks (Liu et al., 15 Sep 2025, Hong et al., 16 Jun 2025).
• 6G and ISAC: FAS contribute to 6G integrated sensing & communications by enabling agile tradeoffs between communication and sensing SNR, adaptable ISAC Pareto frontiers, and precise control of angle/range estimation CRB via joint optimization of FAS position, beamforming, and system parameters (Zou et al., 2024, Zhou et al., 2024).
• Energy harvesting and SWIPT: Both joint position–beamforming optimization and continuous element travel allow simultaneous maximization of downlink rate and harvested energy, yielding up to 40% energy-harvesting gains versus fixed arrays (Zhou et al., 2024, Zhang et al., 23 Oct 2025).
• UAV and indoor deployments: Adaptive FAS port/trajectory planning in UAV or indoor environments enables sub-degree precision in multi-target sensing, robust signal/rate gains, and low-latency adaptation in constrained geometries (Zhu et al., 21 Nov 2025, Zhang et al., 26 Sep 2025, Zhang et al., 18 Sep 2025).
• Localization: FAS port correlation structure is directly exploited for high-resolution RSSI-based positioning via MLE-based joint estimation, matching conventional multi-antenna accuracy at reduced hardware and feedback overhead (Liu et al., 2 Mar 2025).

6. Implementation, Design Constraints, and Practical Algorithms

Realizing FAS in practice involves a multidisciplinary set of challenges and corresponding algorithmic innovations:

• Materials and hardware: Fast-switching, low-loss alloys (GaInSn), robust microfluidics, or PIN-diodes for pixel arrays; high spatial precision in fluid/port actuation; field-programmable switch networks; matched input impedance and isolation; loss mitigation at mmWave (Wu et al., 2024, Zhang et al., 2024).
• Aperture/port optimization: Efficient geometric/policy-gradient algorithms (e.g., GRPO, AO, PSO) tailored to non-convex, real-time design of port positions, beamforming, and power, with computational reductions of up to 83% over classical approaches in large-scale indoor layouts (Zhang et al., 18 Sep 2025, Zhou et al., 2024).
• Channel estimation: Compressed-sensing, low-overhead pilot design, and machine-learning models for port selection and channel prediction in the presence of estimation overhead and time-varying conditions (Zhang et al., 2024, Hong et al., 16 Jun 2025).
• Coding and scheduling: Joint port selection and code/beam allocation, robust to switching delays, using coded modulation or space-time rotation schemes to restore or maximize achievable diversity even under port feedback or actuation delays (Psomas et al., 2022, Hong et al., 7 Mar 2025).
• Standardization and control: Control-protocol extensions (e.g., NETCONF/YANG with fluid state variables), integration with SDN/NFV, and compatibility with 5G NR and massive MIMO standards (Wu et al., 2024).

7. Open Challenges and Emerging Research Directions

FAS pose several research and implementation challenges:

• Effective aperture vs. port density: Design should prioritize increasing physical aperture W over simply increasing N, due to diversity saturation effects (Zhu et al., 10 Sep 2025).
• Fluid/material dynamics: Modeling and compensating for nonidealities in fluid flow, mechanical tolerance, conduction loss, and temperature dependence (Wu et al., 2024).
• Control and feedback: Real-time, distributed optimization under sensing/communication/actuation delays, scalable to hundreds of ports, possibly with AI-driven (e.g., deep RL) policies (Wu et al., 2024, Zhang et al., 18 Sep 2025).
• Integration with RIS/XL-MIMO: Joint optimization of FAS with large-scale intelligent surfaces or multi-antenna systems to unlock additional DoF (Wu et al., 2024, Zhang et al., 26 Sep 2025).
• Security and privacy: Protection against beam hijacking and location-based attacks inherent to spatially reconfigurable apertures (Hong et al., 16 Jun 2025).

By fusing physics-based reconfigurability, advanced signal processing, and software-defined control, fluid-antenna systems offer a pathway to ultra-adaptable, high-capacity, and robust wireless architectures for the 6G era and beyond (Wu et al., 2024, Lu et al., 6 Jan 2025, Hong et al., 16 Jun 2025, Lu et al., 6 Jan 2025, Zhu et al., 10 Sep 2025).