Electromagnetic Channel Statistics for Continuous-Aperture Array (CAPA) Systems

Chongjun Ouyang, Boqun Zhao, Xingqi Zhang, Yuanwei Liu

TL;DR

As wireless antennas evolve from discrete arrays to continuous surfaces, traditional matrix-based math is no longer enough. This paper provides a rigorous electromagnetic (EM) framework to characterize the channel statistics of Continuous-Aperture Arrays (CAPA), proving that their performance is governed by a finite number of "degrees of freedom" and demonstrating a clear capacity advantage over conventional MIMO.

What Problem Does This Paper Address?

In the quest for 6G and beyond, antenna arrays are becoming larger, denser, and higher-frequency. We are moving away from discrete "points" of radiation (like standard MIMO) toward Continuous-Aperture Arrays (CAPA)—essentially a single, large antenna surface with a continuous current distribution.

The challenge is that our standard mathematical toolkit for MIMO relies on matrices and vectors. Because a CAPA is continuous, it requires an operator-based model rooted in electromagnetic theory. While researchers have studied CAPA in simple Line-of-Sight (LoS) scenarios, the statistical behavior of these systems in complex, scattering "fading" environments has remained a major knowledge gap. This paper fills that gap by deriving the statistical distribution of the signal-to-noise ratio (SNR) and the average channel capacity for CAPA systems.

Methods

The authors move away from discrete linear algebra and instead use integral-based signal models. Their approach follows several key steps:

  1. EM Multipath Modeling: They model the spatial response as a continuous random field, considering isotropic scattering (where signals arrive from all directions equally).
  2. Landau’s Eigenvalue Theorem: This is the core analytical "engine" of the paper. The researchers use this theorem to analyze the autocorrelation of the EM fading channel. It allows them to determine how many independent "communication pipes" (Degrees of Freedom) a continuous surface actually provides.
  3. Statistical Derivation: By treating the received SNR as a sum of eigenvalues, they derive a closed-form probability density function (PDF) for the SNR.
  4. Numerical Validation: They use numerical solvers for Fredholm integral equations to verify that their elegant mathematical formulas actually match simulated physical reality.

Datasets

As a theoretical and analytical paper, this work does not use external training datasets. Instead, the authors evaluate their models using numerical simulations. They compare their CAPA model against a "Conventional MIMO" benchmark consisting of a discrete uniform linear array with half-wavelength spacing.

Key Insights and Ideas

  • The "Step-Like" Behavior: One of the most important findings is how the eigenvalues of the channel behave. For a CAPA of length $L$, the eigenvalues stay nearly constant (near 1) for the first few indices and then drop precipitously to zero. The "cutoff" point is defined by the Degrees of Freedom (DOF), calculated as $2L/\lambda$.
  • Infinite to Finite: Even though a CAPA is theoretically "infinite-dimensional" because it is continuous, Landau's theorem proves that it effectively behaves like a finite-dimensional system. This allows the complex EM channel to be approximated as a finite sum of simple, exponentially distributed random variables.
  • Spatial Resource Maximization: Unlike discrete arrays that "sample" the field at specific points, a CAPA fully utilizes the entire aperture surface. This allows for much finer control over the current distribution, leading to superior performance.

Key Results

  • Mathematical Accuracy: The derived closed-form expressions for the SNR distribution and average capacity were shown to be a near-perfect match for numerical simulations.
  • Capacity Superiority: The results consistently show that CAPA achieves a higher average capacity than conventional MIMO arrays of the same physical length. This gap remains significant across different power levels.
  • High-SNR Performance: The paper characterizes the "high-SNR slope" (multiplexing gain), showing that while the slope remains 1 for the single-user case, the power offset (the efficiency of that gain) is significantly improved in CAPA systems.

Limitations and Future Work

While this paper provides a foundational statistical framework, it does have some limitations:

  • Simplified Scattering: The analysis assumes isotropic scattering. Real-world environments are often non-isotropic, with signals clustered in specific directions.
  • Uni-polarized Model: The study focuses on a uni-polarized linear CAPA. Future work could expand this to tri-polarized or planar (2D) surfaces.
  • Single-User Focus: The current model is point-to-point. Extending these EM statistics to multi-user scenarios (where interference becomes a factor) is a logical next step for 6G research.
  • Hardware Constraints: The model assumes "free control" of current distribution, but practical implementations will be limited by the number of available RF chains. Integrating these hardware constraints into the EM model is an open challenge.

Summary updated

Abstract

The channel statistics of a continuous-aperture array (CAPA)-based channel are analyzed using its continuous electromagnetic (EM) properties. The received signal-to-noise ratio (SNR) is discussed under isotropic scattering conditions. Using Landau's theorem, the eigenvalues of the autocorrelation of the EM fading channel are shown to exhibit a step-like behavior. Building on this, closed-form expressions for the probability distribution of the SNR and the average channel capacity are derived. Numerical results are provided to validate the accuracy of the derivations.
BibTeX
@misc{ouyang2025electromagnetic,
  author = {Chongjun Ouyang and Boqun Zhao and Xingqi Zhang and Yuanwei Liu},
  title = {{Electromagnetic Channel Statistics for Continuous-Aperture Array (CAPA) Systems}},
  year = {2025},
  eprint = {2502.06980},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2502.06980},
  howpublished = {arXiv},
}

Loading related content...

Figure