Electromagnetically Reconfigurable Fluid Antenna Systems

• ER-FAS are fluid antenna systems that dynamically reconfigure intrinsic electromagnetic properties—like radiation pattern, polarization, and frequency response—to adapt in real time.
• The technology enables joint optimization of beamforming and electromagnetic state selection, improving spectral efficiency and localization in 6G and beyond.
• Experimental prototypes demonstrate notable gains (up to 4.5 dB enhancement and 1.5 bps/Hz spectral efficiency increase) validating the practical potential of ER-FAS.

Searching arXiv for recent ER-FAS papers and related FAS work. arxiv_search(query="Electromagnetically Reconfigurable Fluid Antenna Systems ER-FAS fluid antenna systems", max_results=10, sort_by="submittedDate")

Electromagnetically Reconfigurable Fluid Antenna Systems (ER-FAS) are antenna architectures in which software-controlled fluidic, conductive, or dielectric structures dynamically alter antenna shape, position, and electromagnetic behavior so that the effective radiating or receiving state can be optimized for the instantaneous radio environment. In the broad fluid-antenna literature, this extends the idea of a fluid antenna system (FAS) from spatial flexibility alone to direct control of intrinsic electromagnetic characteristics such as radiation pattern, polarization, size, and sometimes frequency response. ER-FAS therefore generalizes the conventional fixed-position antenna (FPA) and also goes beyond spatially reconfigurable FAS or movable-antenna formulations, because the element itself becomes a tunable electromagnetic object rather than a fixed-pattern radiator merely displaced in space (Wu et al., 2024, Wang et al., 27 Feb 2025).

1. Conceptual definition and scope

Within the FAS literature, “fluid” does not necessarily denote literal liquid. It denotes a highly adaptable and continuously reconfigurable antenna state that can switch to the most favorable port, position, or radiation state in response to the channel. Survey treatments accordingly place liquid-based antennas, pixel-reconfigurable antennas, mechanically actuated antennas, flexible metamaterial structures, movable antennas, pinching antennas, and reconfigurable waveguide technologies in the same conceptual family (Hong et al., 16 Jun 2025).

The defining distinction of ER-FAS is the shift from spatial reconfigurability to electromagnetic reconfigurability at the element level. Spatially reconfigurable FAS and movable antennas primarily change position or orientation while keeping the element radiation pattern essentially fixed. ER-FAS instead allows each element to reshape its radiation direction, beamwidth, front-to-back ratio, and related electromagnetic properties. A concise formulation used across the literature is that conventional FAS changes where the antenna is, whereas ER-FAS changes how it radiates (Wang et al., 27 Feb 2025).

A second conceptual distinction concerns the system view of the antenna. In the FAS framing, a fixed-position antenna provides access to a signal point, whereas a fluid antenna gives access to a signal function in space. ER-FAS preserves that broader FAS viewpoint but interprets the accessible states more generally: the selectable state may correspond to a port location, a shape, a polarization, a radiation pattern, or a metasurface configuration. This is why electronically reconfigurable antennas, including dynamic metasurface antennas (DMAs), are treated as practical embodiments of the fluid-antenna idea: changing the radiation pattern is taken to be functionally equivalent to moving the device through different virtual-port states (Wu et al., 2024, Ramírez-Espinosa et al., 23 Jul 2025).

2. Electromagnetic and communication models

A standard ER-FAS communication model is a single-user MIMO link with one RF chain at each side, written as

$$y = \mathbf{w}^H \mathbf{H} \mathbf{f}\, s + \mathbf{w}^H \mathbf{n},$$

where $\mathbf{f}$ is the analog transmit precoder, $\mathbf{w}$ is the analog receive combiner, $\mathbf{H}$ is the channel, and $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$. The corresponding spectral efficiency is

$$R = \log_2\!\left(1+\frac{P_T |\mathbf{w}^H \mathbf{H}\mathbf{f}|^2}{\sigma^2}\right).$$

ER-FAS modifies this model by making $\mathbf{H}$ depend not only on geometry and multipath gains but also on the selected radiation state of each element (Wang et al., 27 Feb 2025).

In far field, the literature uses Saleh–Valenzuela-type models in which each antenna element selects one radiation pattern from a finite dictionary. If $\bar{\mathbf{G}}(\theta)$ denotes the dictionary of available element patterns and $\mathbf{b}_{t,i}$, $\mathbf{b}_{r,j}$ are one-hot selection vectors, then the selected gains are represented as

$\mathbf{f}$0

The resulting channel is then rewritten in compact electromagnetic-domain form as

$\mathbf{f}$1

so that spatial array responses and electromagnetic-state selection enter the channel jointly rather than separably (Wang et al., 27 Feb 2025, Wang et al., 4 May 2025).

Near-field modeling requires a spherical-wave description because different transmit–receive pairs experience distinct path lengths, amplitudes, and angles. The near-field ER-FAS literature therefore emphasizes beam focusing rather than beam steering alone. One representative LoS entry is modeled as

$\mathbf{f}$2

which makes explicit that propagation geometry and state-dependent radiation response are co-determined (Wang et al., 27 Feb 2025).

A more rigorous electromagnetic abstraction is provided for metasurface-based embodiments of FAS. There, a DMA is modeled as a multiport admittance network in which the wireless channel is embedded in the mutual-admittance block $\mathbf{f}$3, and the configuration-dependent response appears through the metasurface loads. The equivalent communication response for configuration $\mathbf{f}$4 is expressed through a configuration-dependent DMA response $\mathbf{f}$5, so the effective channel is the inner product of the propagation channel and $\mathbf{f}$6. The covariance across configurations becomes

$\mathbf{f}$7

showing that the effective spatial correlation depends simultaneously on environmental covariance and designer-controlled metasurface response (Ramírez-Espinosa et al., 23 Jul 2025).

3. Physical architectures and implementation families

The ER-FAS literature spans several material and control families. One widely cited taxonomy classifies FAS by material type, shape type, and dynamic control: liquid metal, non-metallic liquid, and metallic pixels; filament, planar, and 3D structures; and controllable liquid flow, pattern-controlled liquid, amount-controlled liquid, and electronic switching control. ER-FAS lies at the intersection of these axes because the attainable radiation state depends jointly on material platform, geometric embodiment, and control mechanism (Wu et al., 2024).

A representative liquid-based ER-FAS architecture uses three stacked layers per array element: a front parasitic fluid-metal layer acting as directors, a middle excitation layer formed by a planar monopole antenna on FR4, and a back parasitic fluid-metal layer acting as reflectors. In the proof-of-concept element, each parasitic layer contains three independent fluid channels. Because each channel is binary, the two layers jointly realize

$\mathbf{f}$8

reconfigurable radiation states. The corresponding array demonstrator is a 1D 12-element ER-FAS array operating at 3.55 GHz with half-wavelength spacing, driven by a 1-to-12 power divider and phase shifters (Wang et al., 27 Feb 2025).

Family	Reconfiguration mechanism	Representative notes
Liquid metal	Fluid motion changes parasitic geometry, shape, or position	Flexibility and low resistive loss; switching latency and actuation matter
Non-metallic liquid / water	Liquid level or conductive-fluid control	Useful for resonance tuning and frequency agility
Metallic pixels / PRA	Planar pixel matrix with electronic switching	Fast, fully electronic switching; a 12-state prototype is reported
DMA / metasurface-based	Electronic switching of metasurface configuration	Planar, scalable, reconfigurable with p-i-n diodes or varactors
Mechanical / movable	Motors, tracks, or physical repositioning	Fine spatial resolution but slow actuation and structural constraints

The pixel-reconfigurable antenna (PRA) is frequently treated as the clearest fast-switching ER-FAS embodiment in the FAS literature. A planar matrix of pixels interconnected by switches can realize multiple reconfigurable states, with each state behaving like an antenna at some position with a certain shape, polarization, and orientation; because switching is electronic, the delay is described as negligible, and a prototype with 12 states is reported. At the same time, tutorial treatments stress that mechanical, metamaterial-based, and pixel-based devices each involve distinct tradeoffs in switching speed, fabrication complexity, and attainable state resolution (Wu et al., 2024, Yang et al., 16 Oct 2025).

The liquid-based prototype work also gives specific fabrication details. The element prototype uses Galinstan liquid metal alloy with composition 68.5% Ga, 21.5% In, and 10% Sn and conductivity $\mathbf{f}$9. Fluid channels are formed from PMMA and PET, laser-cut and bonded with adhesives, while a programmable pumping system injects or extracts liquid metal at about 50 $\mathbf{w}$0l/min. For the 12-element array prototype, parasitic fluid metal is implemented with DM-SIP-3072S fluid silver paste, screen-printed onto 125-$\mathbf{w}$1m PET and cured at 70°C for 2 hours; the feeding network uses 2-bit phase shifters realized with microstrip line lengths (Wang et al., 27 Feb 2025).

4. Control, beamforming, and codebook design

ER-FAS introduces a joint optimization problem over analog phases and electromagnetic states. In the fluid-antenna system perspective, the controller selects the best state based on channel conditions, and the optimization variables can include position, orientation, polarization, and beamforming. For element-pattern ER-FAS, this becomes a mixed discrete–continuous problem over transmit precoder $\mathbf{w}$2, receive combiner $\mathbf{w}$3, and pattern-selection matrices $\mathbf{w}$4 and $\mathbf{w}$5 (Wu et al., 2024, Wang et al., 27 Feb 2025).

A low-complexity design proposed for liquid-based ER-FAS uses block-coordinate descent. The algorithm alternates between optimizing $\mathbf{w}$6, $\mathbf{w}$7, $\mathbf{w}$8, and $\mathbf{w}$9 until convergence. The phase updates have closed form, while the discrete state-selection problem is handled by greedy pruning. On the transmit side, the objective is reduced to

$\mathbf{H}$0

and the removal rule is

$\mathbf{H}$1

The same greedy idea is then applied to the receiver side, providing a polynomial-time approximation in place of exhaustive combinatorial search (Wang et al., 27 Feb 2025).

Localization-oriented ER-FAS work extends this optimization logic to hybrid baseband and electromagnetic precoding. Two paradigms are treated: a synthesis model, in which each antenna generates a desired beampattern from a finite set of EM basis functions, and a finite-state model, in which each antenna selects one pattern from a predefined pattern library. A central structural result is that the optimal covariance lies in the 3-dimensional subspace spanned by the channel vector and its angular derivatives, so only three beams are fundamentally required: the nominal beam, the elevation derivative beam, and the azimuth derivative beam. In the finite-state case, the EM pattern-selection problem is handled by a low-complexity block-coordinate-descent algorithm with complexity $\mathbf{H}$2 for all three codewords (Fadakar et al., 29 Aug 2025).

Related network-level formulations show how ER-FAS-type ideas integrate with reconfigurable environments. In a liquid intelligent metasurface (LIM)-assisted downlink multi-user MISO system, sum-rate maximization is jointly optimized over BS beamforming, LIM phase shifts, and the positions of fluid antennas and liquid reflecting elements. The solution uses alternating optimization together with successive convex approximation and the penalty convex-concave procedure, illustrating that electromagnetic and spatial reconfigurability can be co-optimized in a single design loop (Shen, 22 Jul 2025).

5. Measured gains, validated performance, and application domains

The empirical case for ER-FAS combines full-wave simulation, fabricated hardware, and over-the-air trials. For the 12-element liquid-based array, analytical and ANSYS HFSS beampatterns match closely. At $\mathbf{H}$3, the reported analytical gain enhancement is 3.4 dB and the full-wave enhancement is 4.5 dB; at $\mathbf{H}$4, the analytical enhancement is 2.5 dB and the full-wave enhancement is 1.7 dB. In a $\mathbf{H}$5 far-field MIMO setting at 3.55 GHz, the main reported spectral-efficiency gain is about 1.5 bps/Hz over a conventional fixed-pattern array. Measured $\mathbf{H}$6 remains below $\mathbf{H}$7 dB from 1.8 to 7.3 GHz, corresponding to a 120.9% fractional bandwidth. In indoor LOS-dominant SDR trials using two Analog Devices ADALM-Pluto SDRs, ER-FAS provides about 1–3 dB higher received power than a conventional antenna, and with 4-QAM and transmit power from 0.2 to 0.5 mW it yields lower BER across nearly the entire power range, especially at medium and low SNR (Wang et al., 27 Feb 2025).

Broader FAS studies place these gains in a 6G system context. In a SWIPT case study with a BS equipped with 4 FAs, one FA at the information receiver and one FA at the energy receiver, and minimum FA spacing

$\mathbf{H}$8

the FAS scheme is reported to achieve a 73.1% enhancement over the FPA scheme in communication rate. In a RIS-assisted case study under rich-scattering Rayleigh fading, increasing the number of FAS ports from $\mathbf{H}$9 to $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$0 significantly reduces outage probability, and example received-SNR values of 10 dB and 11.0231 dB are reported for $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$1 and $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$2, respectively. These results are presented as consequences of spatial diversity, interference nulls, and the ability to select the best radiating or receiving state even with a small number of RF chains (Wu et al., 2024).

Application coverage is correspondingly broad. The literature explicitly discusses SWIPT, ISAC, NOMA, RIS-assisted communication, physical layer security, mobile edge computing, AirComp, cognitive radio, backscatter, short-packet communication, full-duplex, OTFS, satellite and non-terrestrial communications, and links to holographic MIMO and stacked intelligent surfaces. Survey work further connects FAS and ER-FAS to FAMA, queueing-aware QoS provisioning, energy-efficient power allocation, and cache-enabled HetNets, arguing that electromagnetic or spatial state selection can improve outage, throughput, delay, and massive connectivity without classical array scaling (Wu et al., 2024, Hong et al., 16 Jun 2025).

Localization is a particularly clear example of an ER-FAS-specific gain mechanism. In downlink localization, finite-state ER-FAS with $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$3 is reported to achieve about 13.98 dB higher peak beampattern gain than a traditional non-reconfigurable array, while the synthesis model with $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$4 achieves about 6 dB higher peak. Both ER-FAS variants outperform the traditional array in RMSE and PEB, with especially strong robustness at low SNR. The study also reports that a $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$5 ER-FAS with sufficiently many states can outperform a much larger $$\mathbf{n}\sim \mathcal{CN}(\mathbf{0},\sigma^2 \mathbf{I})$$6 traditional array, indicating that electromagnetic reconfiguration can substitute for larger aperture in some localization regimes (Fadakar et al., 29 Aug 2025).

6. Practical constraints, misconceptions, and research directions

A recurrent misconception is that ER-FAS is synonymous with liquid antennas. The literature is explicit that FAS and ER-FAS include fluidic, pixel-based, metasurface-based, and mechanical realizations. A second misconception is that ER-FAS is merely a movable-antenna system under another name. The distinction repeatedly drawn is that movable or spatially reconfigurable systems alter location or orientation, whereas ER-FAS alters the electromagnetic radiation state itself; the strongest embodiments may do both simultaneously (Hong et al., 16 Jun 2025, Wang et al., 27 Feb 2025).

The main technical bottlenecks are channel estimation, channel modeling, robust optimization, and hardware realism. Estimating CSI over many FAS states produces large pilot overhead, hardware cost, and latency, and the literature accordingly proposes exploiting spatial correlation, compressed sensing, and AI-based estimation. Existing channel models are also described as insufficient because they typically capture either rich scattering or finite scattering but not full shape reconfigurability, polarization, viscosity, switching delay, or motor noise. Under imperfect CSI, robust optimization, stochastic optimization, and error-bound formulations are identified as necessary for practical deployment (Wu et al., 2024).

Perspective work on bridging theory and practice sharpens this critique. It argues that many theoretical analyses assume near-instant reconfiguration, perfect CSI, static or slowly varying channels, ideal material properties, no electromagnetic or mechanical coupling, and no actuation latency or energy cost. The non-idealities emphasized as most consequential are mutual coupling, impedance change and matching variation, resonance shift and frequency instability, pattern changes that increase interference leakage, mechanical tolerance and alignment errors, packaging and enclosure effects, user-body detuning, fast fading, and stale CSI. The same work stresses that pixel-based and metamaterial antennas adapt better in fast-fading settings because switching is very fast, whereas mechanical and liquid-based systems are more vulnerable because actuation is slower (Yang et al., 16 Oct 2025).

Survey treatments broaden the open-problem set to green FAS, security and privacy, standardization, and deployment. Energy consumption from dynamic switching and circuit power must be modeled realistically; reconfigurability may create attack surfaces such as beam hijacking or expose location information; and protocol support is required for port switching and feedback. These issues matter because the central promise of ER-FAS is not merely hardware novelty but controllable electromagnetic adaptability under realistic network constraints (Hong et al., 16 Jun 2025).

Recent sensing work on Enormous Fluid Antenna Systems (E-FAS) extends the ER-FAS logic to coordinated intelligent surfaces that form a gigantic reconfigurable electromagnetic aperture. In that setting, maximizing coherent routing gain does not necessarily maximize sensing performance: high routing correlation can collapse the sensing rank even as SNR improves. The derived Fisher-information analysis therefore exposes a gain–diversity trade-off, with angular accuracy governed by both routing power and angular sensitivity rather than by power gain alone. This result reinforces a general ER-FAS lesson already visible in smaller systems: electromagnetic reconfiguration is valuable only when routing, radiation, estimation, and control are jointly designed rather than optimized in isolation (Ghadi et al., 22 Jun 2026).