Ultra-Large-Scale Arrays: System Innovations

• Ultra-large-scale arrays are engineered systems that deploy hundreds to thousands of coordinated sensing, communication, or processing elements across expansive apertures.
• They overcome classical electromagnetic modeling limitations by employing novel near-field, metasurface-enabled, and distributed signal processing strategies.
• These arrays drive advancements across diverse fields including radio astronomy, wireless communications, AI acceleration, remote sensing, and space applications.

Ultra-large-scale arrays are engineered systems comprising an extremely large number of closely coordinated sensing, communication, or processing elements—most commonly antenna elements—arranged over physical apertures large enough that classical modeling and design assumptions often fail. These arrays underpin next-generation radio astronomy, wireless communication, high-performance computing, remote sensing, and space-borne applications. As physical scale and element counts increase (reaching hundreds, thousands, or more), system architectures, signal processing, power delivery, and deployment methodologies must be fundamentally rethought to address new regimes of spatial, computational, and electromagnetic complexity.

1. Scientific Motivation and Domains of Application

Ultra-large-scale arrays have emerged as critical instruments in several advanced research and engineering fields:

• Radio Astronomy: The push to paper phenomena such as the cosmic dark ages and the epoch of reionization, by detecting redshifted 21 cm hydrogen lines well below 100 MHz, motivates the deployment of space-based ultra-long wavelength (ULW) arrays. Ground-based observations below 30 MHz are untenable due to ionospheric opacity and pervasive man-made RFI, making space-based arrays the necessary solution (Rajan et al., 2015).
• Wireless Communications: The relentless demand for increased data rates, reliability, and spectral efficiency in 5G/6G and beyond has driven the exploration of extremely large-scale antenna (ELAA/XL-array) systems, sometimes realized via continuous electromagnetic surfaces or modular distributed arrays (Lu et al., 2021, Li et al., 2022).
• Astroparticle Physics: Detection of ultra-high-energy cosmic air showers relies on vast ground-based arrays of radio antennas or other detectors, where layout optimization and computational tractability necessitate specialized simulation and design methodologies (A et al., 2 Jan 2024).
• High-Performance Computing and AI Acceleration: The exponential growth in computing and memory bandwidth needs in artificial intelligence and big data motivates wafer-scale integration, systolic array acceleration, and scalable power delivery for ultra-large-scale logic and memory arrays (Yüzügüler et al., 2022, Safari et al., 2022).
• Space Applications: The geometric constraints of launch vehicles for deploying large apertures in orbit or on the lunar surface require flexible and robust deployable array technologies, such as pop-up composite antenna arrays (Mizrahi et al., 2023).

2. Physical Modeling and Electromagnetic Regimes

As physical aperture scales become comparable to or exceed the operational wavelength and to typical signal propagation distances, classical approximations break down:

• Near-field effects dominate, making it essential to model spherical rather than planar wavefronts. Key spatial parameters to model include distance-dependent phase, amplitude, and projected aperture variations across the array (Lu et al., 2021, Li et al., 2022, Zheng et al., 2023, You et al., 2023, Wang et al., 5 Aug 2025).
• Uniform-Power Distance (UPD): Beyond standard Rayleigh (phase) criteria demarcating far- and near-field, UPD is introduced as a regime where amplitude variation across the array is negligible. For extremely large arrays, both phase and amplitude non-uniformity must be considered to determine modeling validity domains (Lu et al., 2021).
• Metasurface-enabled architectures: By employing transmissive reconfigurable metasurfaces as large phased antenna apertures, phase control and array function can be achieved “over the air,” drastically simplifying wiring and phase shifter requirements (Wang et al., 5 Aug 2025). The physical model relies on electromagnetic field propagation via surface equivalence, where the effective response is characterized at the individual unit cell level.
• Array factor and spatial resolution: The half-power beamwidth (HPBW), frequently used as a spatial resolution metric, depends on the effective electrical aperture, the phase coherence across the array, and, in modular and metasurface implementations, the details of feed and cell coupling (Wang et al., 5 Aug 2025).

3. System Architectures, Deployment, and Scalability

Ultra-large-scale arrays require novel architectural and deployment paradigms:

• Centralized vs. Distributed Architectures: Classical architectures collect all signals at a mothership or centralized processing node, potentially leading to a single point of failure and a prohibitive scaling of interconnect and data movement. Distributed architectures partition the array into clusters or modules, each with its own local processing, synchronization, and limited inter-cluster communication—significantly reducing required bandwidth and enhancing robustness (Rajan et al., 2015, Xu et al., 23 Jul 2024).
• Space-Based and Modular Designs: Deployment in space, including lunar far side and Lagrange points, is preferred for radio quietness in ULW astronomy (Rajan et al., 2015). Modular arrays with larger inter-module separation facilitate deployments on terrestrial structures, but require accurate near-field modeling for performance prediction (Li et al., 2022).
• Popup and Flexible Arrays: Mechanically deployable arrays fabricated on flexible substrates, engineered for mass-producibility, extreme thermal and mechanical stress resistance, and long-term stowage reliability, support scalable apertures for space-based applications (Mizrahi et al., 2023).
• Wafer-Scale Integration and Power Delivery: Ultra-large logic and memory arrays integrated on silicon interconnect fabrics (Si-IF) demand specialized power topologies (peripheral, backside, hybrid) to sustain kilowatt power levels with acceptable voltage drops and low area overhead (Safari et al., 2022).

4. Distributed Signal Processing and Inference

For ultra-large-scale arrays, distributed signal and image processing is indispensable:

• Distributed Channel Estimation and Uplink Equalization: Arrays are partitioned into clusters or distributed nodes (DNs), each with local baseband processing to handle their portion of signals. Aggregation and consensus (e.g., via ADMM, BCD) allow global inference with orders-of-magnitude reduction in bandwidth and computational resources (Xu et al., 23 Jul 2024).
• Reduced Data Transfer via Local Processing: In massive interferometric imaging (e.g., LOFAR, SKA), intra-cluster correlations are performed locally, and only low-dimensional reference signals are shared for high-frequency (fine-resolution) information, drastically curtailing data transfer (Ferrari et al., 2015).
• ADMM-Based Optimization: Consensus-based distributed optimization methods, such as ADMM, reconcile local array’s estimate of sky images (for astronomy) or received signals (for MIMO systems), ensuring consistency while maintaining scalability (Ferrari et al., 2015).
• Physical Layer Security, ISAC, and Emerging Paradigms: Distributed architectures also enable finer-grained spatial control for applications in security, backscatter communication, and integrated sensing and communication (Xu et al., 23 Jul 2024).
• Inference Accelerator Arrays: In AI accelerators, multi-pod systolic arrays with careful sizing (e.g., 32×32 processing elements per pod), scalable interconnect (butterfly networks), and optimal data tiling enable both peak throughput and high utilization. The focus is on mapping realistic DNN workloads onto the hardware with minimal arithmetic underutilization and memory stalls (Yüzügüler et al., 2022).

5. Beam Management, Calibration, and Codebook Design

The scale and operating regime of ultra-large-scale arrays pose distinct challenges for beamforming and calibration:

• Near-Field Beam Management: Beam training, tracking, and scheduling in the near field must account for joint angle and range resolution—a departure from pure angular domain processing. Algorithmic innovations include polar-domain codebooks, hierarchical codebook design, and Kalman filter-based tracking (You et al., 2023).
• Codebook Quantization for Hybrid Near/Far Field: Classical DFT codebooks are insufficient; instead, new codebooks based on ellipsoid (UPA) or ellipse (ULA) fitting to codeword correlation performance are proposed. Dislocation (hexagonally packed) codebooks achieve near-optimal quantization with minimal feedback overhead (Zheng et al., 2023).
• Calibration, Synchronization, and Ranging: Joint affine clock modeling and two-way message exchanges enable tight time/frequency synchronization and ranging across distributed satellite nodes or baseband units, crucial for coherent combination over vast apertures (Rajan et al., 2015).

6. Manufacturing, Mechanical, and Power Constraints

Physical realization of ultra-large-scale arrays necessitates scalable solutions for manufacturing, thermal management, and power delivery:

• Automated and Co-Cured Composite Manufacturing: Scalable co-curing of conductive and mechanical layers, precise alignment, and the avoidance of adhesives deliver mechanical integrity and repeatability for massively deployed space-borne antenna elements (Mizrahi et al., 2023).
• Mechanical and Thermal Qualification: Arrays must pass vibration, thermal cycling, and long-term stowage tests simulating orbital environments, ensuring both electromagnetic and mechanical robustness over mission lifetimes (Mizrahi et al., 2023).
• Power Topologies for Wafer-Scale Arrays: Peripheral and backside delivery methods (including on-wafer and PCB-mounted converters, through-wafer vias, hierarchical decoupling) are matched to application profiles—balancing area, power loss, and complexity—to support applications from AI acceleration to neuromorphic computing (Safari et al., 2022).

7. Future Directions and Technological Outlook

The trajectory for ultra-large-scale arrays is governed by advances in:

• Metasurface-Enabled and Reconfigurable Architectures: Over-the-air phase control via passive transmissive metasurfaces promises even greater scalability and lower hardware cost, enabling efficient channel estimation even in hybrid near-/far-field scenarios, with spatial resolution approaching classical architectures (Wang et al., 5 Aug 2025).
• Advanced Distributed Algorithms: Exploiting near-field channel sparsity, spatial nonstationarity, and dynamic clustering for scalable, robust signal processing remains an open challenge (Xu et al., 23 Jul 2024).
• Low-Power and Hybrid Analog/Digital Solutions: Integrating low-resolution ADCs/DACs, hybrid beamforming, and advanced lens or fluid antennas may further reduce power and cost at extreme scales.
• Machine Learning and Self-Optimization: The high-dimensional, non-linear parameter space of ultra-large arrays in near-field regimes is likely to see the adoption of deep learning and reinforcement learning frameworks for adaptive beam management, calibration, and self-healing operation (You et al., 2023).
• Universal Simulation and Layout Tools: Simulation-recycling and pruning methods offer a universal route for exploring and optimizing array layouts across domains, from air-shower detection to environmental sensing (A et al., 2 Jan 2024).

Ultra-large-scale arrays thus synthesize innovations in system architecture, physical layer modeling, distributed processing, and scalable manufacturing, forming the backbone of several ongoing and future advances in scientific exploration, communication, sensing, and high-performance computing.