Multi-Mode PASS: Flexible Antenna Systems

• Multi-Mode PASS is a flexible antenna architecture that utilizes multiple guided modes within dielectric waveguides combined with movable pinching antennas for enhanced wireless communications and sensing.
• The system employs coupled-mode theory and advanced beamforming techniques to enable dynamic spatial multiplexing, robust interference management, and high-capacity data transmission.
• Joint digital and analog optimization in PASS improves spectral efficiency and reconfiguration speed, supporting integrated sensing, secure communications, and scalable multi-user operations.

Multi-Mode Pinching-Antenna Systems (PASS) are a class of flexible antenna architectures wherein multiple guided modes within a single or multiple dielectric waveguides are leveraged for optimized wireless communications, sensing, or integrated applications. By exploiting both the modal diversity of dielectric guides and the spatial agility of movable pinching antennas (PAs), multi-mode PASS platforms provide dynamically reconfigurable high-capacity links, robust near-field and far-field beamforming, and enhanced multi-user support across a variety of use cases and signal environments.

1. Physical Principles and Modal Multiplexing

A multi-mode PASS consists of one or more dielectric waveguides supporting $M > 1$ orthogonal guided modes (e.g., TE$_{01}$, TE$_{11}$, TM$_{01}$) at the operational frequency. Each waveguide is physically loaded with a set of PAs, which are radiating elements that can often be mechanically or electronically positioned along the longitudinal axis of the guide. Depending on the PA design—mode-selective versus mode-combining—each PA either couples predominantly to one specific guided mode (via phase-matching of their propagation constants) or can extract energy from several guided modes with programmable weights (Liu et al., 26 Jan 2026, Xu et al., 28 Jan 2026).

The circuit and physical interactions within multi-mode PASSs are modeled via coupled-mode theory (CMT). For a waveguide with $M$ modes, the evolution of the complex amplitudes of the guided and radiated fields along the $z$-coordinate is:

$$\frac{d a_m}{dz} = -j \beta_m a_m(z) - j \kappa_{m,i} b_i(z), \qquad \frac{d b_i}{dz} = -j \beta_i^{\rm PA} b_i(z) - j \kappa_{m,i} a_m(z)$$

where $a_m(z)$ denotes the amplitude of mode $m$, $b_i(z)$ that of the $i$th PA eigenmode, $\beta_m$/$\beta_i^{\rm PA}$ the propagation constants, and $\kappa_{m,i}$ is the electromagnetic coupling factor determined by field overlap integrals (Liu et al., 26 Jan 2026).

Under weak-coupling and single-mode selectivity (for each PA), these $M$-mode waveguides can be treated as $M$-dimensional multiplexing media, supporting $M$ independent spatial data streams or “mode-domain beams.” The PA placement and modal-coupling parameters thus jointly control both overall array gain and the effective beam patterns delivered per stream or per group of users (Liu et al., 26 Jan 2026, Xu et al., 28 Jan 2026).

2. Transmission Structures and Signal Models

Multi-mode PASS operation in multi-user scenarios can be realized via several transmission structures, distinguished by how data streams and waveguides are mapped (Zhao et al., 20 Aug 2025):

• Waveguide Multiplexing (WM): All $K$ data streams are digitally precoded and simultaneously injected into $K$ waveguides, each with $N$ optimized PAs. The PA positions across all waveguides are jointly optimized. All waveguides cooperate in the spatial multiplexing of user data, supporting both unicast and multicast modes. The radiated signal is

$$\tilde{\mathbf{s}}^{\rm WM} = \mathbf{G}(\mathbf{X}) \sum_{k=1}^K \mathbf{w}_k s_k,$$

where $\mathbf{G}(\mathbf{X})$ stacks all in-guide PA responses.

• Waveguide Division (WD)/WDMA: Each data stream is assigned a dedicated waveguide (with per-guide PAs), and only power allocation and pinching beamforming are optimized, avoiding waveguide-to-waveguide digital coupling. The radiated signal is

$$\tilde{\mathbf{s}}^{\rm WD} = \mathbf{G}(\mathbf{X}) \cdot \mathrm{diag}(\sqrt{p_1},...,\sqrt{p_K}) \mathbf{s}.$$

• Waveguide Switching (WS): Data streams are time-multiplexed (TDMA-like): in each time slot, only one stream is transmitted, and both baseband and PA positions are optimized slot-wise, resulting in interference-free, fully decoupled per-user design at the expense of tighter synchronization requirements.

The resultant system model supports formulation of group or user SINR and rate constraints using the composite channel: $$y_{k,g}^{\rm WM} = \mathbf{h}_{k,g}(\mathbf{X}) \mathbf{G}(\mathbf{X}) \mathbf{w}_k s_k + \sum_{k'\neq k}\mathbf{h}_{k,g}(\mathbf{X}) \mathbf{G}(\mathbf{X}) \mathbf{w}_{k'} s_{k'} + n_{k,g}$$ with

$$\mathrm{SINR}_{k,g}^{\rm WM} = \frac{|\mathbf{h}_{k,g} \mathbf{G} \mathbf{w}_k|^2}{\sum_{k'\neq k} |\mathbf{h}_{k,g} \mathbf{G} \mathbf{w}_{k'}|^2 + \sigma^2}$$

and analogous expressions for WD and WS (Zhao et al., 20 Aug 2025, Zhao et al., 25 Feb 2025).

In mode-domain architectures, $M$ data streams are mapped into $M$ guided modes, each individually excited and radiated at multiple PA locations, supporting true $M$-stream spatial multiplexing on one physical waveguide (Xu et al., 28 Jan 2026).

3. Joint Optimization: Beamforming and Pinching Antenna Placement

The core technological challenge in multi-mode PASS is the joint nonconvex optimization of digital (baseband) beamforming and analog (pinching/placement) beamforming for maximal throughput, fairness, or energy efficiency. The unified max-min fairness (MMF) or sum-rate maximization problems involve constraints on per-user rates, total or per-guide power, PA position domains, and minimum PA separation (Zhao et al., 20 Aug 2025, Wang et al., 9 Feb 2025, Xu et al., 28 Jan 2026). Mathematically:

$$\max_{\mathbf{X},\{ \mathbf{w}_k \}, \{ p_k \}, ...} R_{\min} \;\;\text{s.t.}\;\; R_{k,g}^{Y} \geq R_{\min} \; \forall k,g,$$

with structure-specific power constraints and physically feasible PA movements.

Solution methods include:

• PDD-based alternating optimization: Fractional SINR constraints are transformed with weighted MMSE or penalty dual-decomposition (PDD) methods. The Lagrangian is augmented by auxiliary variables, enforcing equality/phase constraints introduced by highly coupled digital-analog beamforming (Zhao et al., 20 Aug 2025).
• Gradient-based and population-based metaheuristics: Penalty-based gradient ascent (GAA), particle swarm optimization (PSO), and evolutionary algorithms (e.g., SHADE) efficiently explore the multi-modal, nonconvex PA placement landscape. For example, PSO-ZF (outer PSO on PAs, inner ZF on digital beamforming) efficiently finds near-optimal solutions (Xu et al., 28 Jan 2026).
• Zeroforcing and low-complexity designs: Closed-form ZF precoders reduce algorithmic complexity with minimal loss, especially in WDMA and multiple-mode/WDMA settings (Wang et al., 9 Feb 2025, Xu et al., 28 Jan 2026).
• Discrete vs. continuous PA placement: Discrete activation (finite set of allowed PA slots) yields performance close to continuous optimization when the number of candidate positions is moderate-to-high (~20–300 per waveguide) (Zhao et al., 25 Feb 2025).

Mode-domain multi-mode designs require, in addition, careful grouping of PAs to modes (“PA grouping”), enforcement of channel orthogonality conditions (zero mutual interference via multi-scale PA placement constraints), and robust handling of mode leakage and inter-mode crosstalk (Xu et al., 28 Jan 2026, Liu et al., 26 Jan 2026).

4. Capacity, Performance, and Scaling Law Insights

Comprehensive analytical and numerical studies confirm several critical benefits and design laws for multi-mode PASS (Xu et al., 28 Jan 2026, Liu et al., 26 Jan 2026, Ouyang et al., 17 Jun 2025, Shan et al., 31 May 2025):

• Throughput and capacity region: Multi-mode PASS strictly enlarges the multiuser capacity region relative to conventional fixed-antenna systems, with up to 2–3$\times$ spectral efficiency gain at moderate SNR; the gap widens as spatial user/device spread increases.
• Multicast and unicast optimization: WM yields the highest MMF rate for multicast scenarios; WS is superior for isolated unicast transmissions due to per-user spatial tailoring; WD is beneficial in geographically separated groups with low complexity.
• Scalability: MMF and sum-rate scale with number of PAs $N$ nearly linearly (rate increase of 20–27% from $N=4$ to 12), subject to diminishing returns governed by noise accumulation. For discrete PA placements, matching >90% performance to continuous regimes is feasible with $A=20$–$200$ slots (Zhao et al., 25 Feb 2025).
• Robustness: Unlike MIMO, PASS is robust to user dispersion and LoS blockage since PAs can be repositioned to restore path-loss-dominated links.
• Reconfiguration speed: Mode-selective PAs allow microsecond-level mode switching; mode-combining PAs (varactor/MEMS tuned) support rapid adaptation in tens of microseconds (Liu et al., 26 Jan 2026).
• Capacity region bounds: Achievable regions are tightly approximated by alternating optimization inner bounds, with TDMA/FDMA schemes nearly achieving capacity in high PA-count regimes (Ouyang et al., 17 Jun 2025).

5. Sensing, ISAC, and Security Applications

Multi-mode PASS naturally supports integrated sensing and communication (ISAC) as well as covert communications due to the fine-grained control over the spatial and modal distribution of radiated energy (Li et al., 27 Aug 2025, Wang et al., 21 May 2025, Jiang et al., 14 Apr 2025, Jiang et al., 7 Sep 2025):

• Integrated Sensing: The joint optimization of digital and pinching beamforming minimizes the CRB for multi-target localization under communication SINR constraints. Full-duplex architectures with PAs for Tx and ULA for Rx achieve superior sensing accuracy over MIMO/UPA and are robust to self-interference (Li et al., 27 Aug 2025).
• Covert Communications: PASS leverages PA reconfiguration to enhance the covertness of links; multi-mode operation uses combined PA placement and artificial noise strategies, exploiting mobility over meters to counter eavesdropper advantages. EKF (extended Kalman filter) + DRL (deep reinforcement learning) frameworks enable adaptive mode and PA control under adversarial uncertainty (Jiang et al., 7 Sep 2025, Jiang et al., 14 Apr 2025).
• Mode-switching for ISAC: By alternating mode allocations (e.g., one mode for comm, one for sensing), PASS enables flexible ISAC tradeoff and real-time adaptation (Liu et al., 26 Jan 2026).

6. Practical Variants and Implementation Considerations

PASS admits multiple hardware realizations and architectural variants (Liu et al., 26 Jan 2026):

Variant	DoF/Features	Notes
S-PASS (single)	1 mode	All PAs radiate identically
C-PASS (center-fed)	2 counter-propagating	Combines two stream DoFb
M-PASS (multi-mode)	$M$ spatial/mode DoF	Mode-selective/combining PAs

• Segmented and Center-fed Architectures: Segmentation introduces per-segment beamforming, C-PASS supports split-streams via counter-propagation.
• Mode-selective vs. Mode-combining: Trade hardware complexity for flexibility; mode-selective is simpler but less adaptive, while mode-combining supports tunable multi-mode extraction per PA at higher hardware and isolation cost.
• PA actuation: Mechanical, MEMS, or varactor tuning enables rapid, accurate PA positioning. Accurate control at the scale of $\sim \lambda/100$ is needed for mmWave systems.
• Discrete vs. continuous activation: A practical number of discrete slots suffice for most performance needs; mechanical precision and mutual coupling avoidance ($\sim \lambda/2$ PA spacing) are critical.

7. Open Challenges and Future Trends

Significant theoretical and practical challenges remain (Liu et al., 26 Jan 2026):

• Mode-beam coordination: Joint, nonconvex optimization across digital and analog/modal domains; ML (graph-based neural nets, permutation-invariant models) is a promising avenue.
• Inter-mode crosstalk: Density of PAs and suboptimal isolation can introduce spurious coupling; advances in electromagnetic meta-materials and circuit co-design are required.
• Wideband and OFDM compatibility: Strong modal dispersion in higher-order modes complicates wideband transmission; robust modal equalization and new training protocols are needed.
• Fabrication scalability: Volume production of multi-mode guides and mode-aware PAs with sub-mm tolerances—additive manufacturing and flexible circuits are under development.
• Real-time implementation: Integration of real-time user tracking, environmental sensing, and distributed optimization remains a practical challenge for large-scale deployments.
• Dynamic mode allocation: Adaptive assignment of sensing and communication functionality across spatial, frequency, and polarization modes enables further efficiency in future ISAC and secure systems.

Multi-mode PASS establishes a framework—underpinned by coupled-mode physical models, mode-group optimization, and algorithmic advances in joint beamforming and placement—that is highly suited to next-generation, dynamic, cost- and energy-efficient wireless networks, including dense access, ISAC, and high-precision sensing environments (Liu et al., 26 Jan 2026, Xu et al., 28 Jan 2026, Zhao et al., 20 Aug 2025, Zhao et al., 25 Feb 2025).