Large-Aperture MIMO Architectures

Updated 13 January 2026

Large-aperture MIMO architectures are systems that deploy extensive, dense antenna surfaces to exploit near-field, non-planar wavefronts for enhanced spatial multiplexing gains.
They leverage holographic modeling, continuous-surface representations, and multiple polarizations to achieve full vector-field diversity and scalable beamforming.
Advanced hardware designs such as metasurfaces, hybrid analog-digital schemes, and subarray partitioning reduce complexity while maximizing degrees of freedom and spectral efficiency.

Large-aperture MIMO architectures deploy electrically large, densely packed antenna surfaces to exploit non-planar, near-field wavefronts for enhanced spatial multiplexing gains, especially in line-of-sight (LoS) environments and high-frequency bands. Modern large-aperture MIMO leverages either very dense discrete arrays, continuous holographic surfaces, or metasurfaces supporting multiple polarizations and advanced analog combining. The fundamental shift from far-field scattering to near-field vector-wave regime enables multiple spatial and polarization degrees of freedom (DoF), analytic characterization via continuous-aperture models, and novel hardware architectures for scalable beamforming and combining.

1. Holographic Array Modeling and Continuous-Surface Channel Representations

The foundational modeling step for large-aperture MIMO is the holographic approximation, in which a dense array of point-dipole elements on a surface $S$ converges toward a continuous radiating aperture as the inter-element spacing vanishes. Under this limit, the MIMO channel becomes an integral operator:

$H = \iint_S \frac{\xi}{\lambda}\frac{e^{-j2\pi d(r)/\lambda}}{d(r)} P_\perp(r)\, dS$

where $P_\perp(r) = I - u(r)u(r)^H$ projects each local polarization onto the plane orthogonal to the propagation direction $u(r)$ , and $d(r)$ is the element-to-receiver distance. Reactive near-field terms ( $1/r^2$ , $1/r^3$ ) contribute $O(1/d^3)$ errors and can be neglected for practical aperture-to-user distances (Agustin et al., 2024). Each surface "pixel" can host up to three orthogonal dipoles, supporting full vector-field diversity.

Large-aperture MIMO with multiple polarizations forms a block-matrix channel:

$H \in \mathbb{C}^{r_{\text{pol}} \times (|S|\cdot t_{\text{pol}})}$

for $t_\text{pol}$ Tx and $r_\text{pol}$ Rx polarizations. Singular-value decomposition (SVD) yields $r_\text{pol}$ spatial/polarization modes, enabling up to three simultaneous streams for full vector-field excitation. Water-filling optimizes power allocation across modes given perfect channel state information (CSI) (Agustin et al., 2024). In the continuous (holographic) limit for large arrays, analytic expressions for Gramian eigenvalues underpin spectral efficiency calculations and DoF estimation.

3. Asymptotic Eigenvalue Distribution: Linear vs. Planar Geometries

For uniform linear arrays (ULAs) and uniform planar arrays (UPAs) in near-field LoS, analytic closed-form Gramian matrices and their eigenvalues can be computed as functions of the geometry and Rx position.

For an extra-large ULA of length $2L$ and Rx at distance $D$ with elevation $\theta$ , normalized Gramian $W$ depends on $\rho = L/D$ and admits three eigenvalues derived from high-order trigonometric series (Agustin et al., 2024).
For a UPA of size $2L_x \times 2L_y$ , the normalized Gramian is obtained via surface integrals over the continuous array coordinates. Its eigenvalues again directly determine the asymptotic MIMO capacity.

This approach, also employed in "LoS MIMO-Arrays vs. LoS MIMO-Surfaces" (Renzo et al., 2022), quantifies the spatial DoF as $R \approx (L_T L_R)/(\lambda D)$ for parallel 1D apertures, with surface geometry controlling the available orthogonal modes.

4. Optimal Aperture Size, SNR Scaling, and DoF Maximization

Spectral efficiency in large-aperture MIMO is a sensitive function of aperture size, Rx distance, and SNR. Under a normalized total power budget $P \propto 1/N$ , the reference SNR at distance $D$ is

$\text{SNR}_0 = \frac{P}{\sigma^2} \left| \frac{\xi}{\lambda} \right|^2 \frac{1}{D^2}$

For each geometry:

ULA: For every elevation angle, there exists a unique $\rho^* = L^*/D \approx 0.9-1.0$ that maximizes capacity; oversizing yields diminishing returns due to near-field reactive edges.
UPA: The optimum diagonal scales linearly with $D$ : $\Lambda^*/D \approx 1.4-2.6$ , reflecting the higher spatial resolution of a 2D aperture (Agustin et al., 2024).

Planar holographic arrays support a density of up to $3(\pi/\lambda^2)$ DoF per m $^2$ with three polarizations, in contrast to $2/\lambda$ per m for dense 1D arrays.

5. Hardware Architectures for Large-Aperture MIMO

Large-aperture operation can be realized through several hardware paradigms, each with its own trade-offs:

Holographic Metasurfaces: Achieve continuous-aperture performance via sub-wavelength elements and local amplitude/phase control, enabling spatial and polarization DoF saturation (Agustin et al., 2024).
Subarray Partitioning: Extremely large arrays are partitioned into subarrays, each equipped with separate analog or hybrid combining, capitalizing on user "visibility regions" to facilitate decentralized processing and reduce the complexity of joint detection (Amiri et al., 2018, Yang et al., 2019).
Switches and Constant Phase-Shifters: Hybrid architectures using switch networks and constant phase dictionaries achieve near-full-digital combining performance with drastically reduced RF chains and phase-shifter counts (Alkhateeb et al., 2016).
Reflect-Array / Transmit-Array Architectures: Passive large apertures illuminated by a few active feeds deliver energy-efficient scaling and OMP-based or mutual-information maximizing precoder designs (Jamali et al., 2018, Jamali et al., 2019).

Advanced analog architectures such as microwave linear analog computers (MiLACs) provide graph-theoretic reductions in circuit complexity—from fully-connected $\mathcal{O}(N^2)$ to stem-connected $\mathcal{O}(N)$ topologies—while provably maintaining the channel SVD and achieving capacity (Nerini et al., 18 Jun 2025). This enables practically feasible analog domain beamforming for "gigantic" ( $N \gg 1000$ ) arrays.

6. Measurement Campaigns and Empirical Scaling Laws

Measurement-based studies confirm theoretical scaling laws:

Increasing array aperture from $5\lambda$ to $40\lambda$ yields a $2\times-3\times$ increase in usable DoF for 8 users; $120\lambda$ approaches i.i.d. channel behavior in LoS/NLoS blends (Martínez et al., 2015).
In "grouped LoS" scenarios (closely clustered users), aperture increase is particularly critical for restoring inter-user orthogonality and channel resolvability.

The performance benefit saturates beyond $\Lambda \approx 100$ (in wavelengths), emphasizing the importance of installation constraints and hardware trade-offs.

7. Practical Design Guidelines and Future Directions

Geometry Selection: Planar (2D) holographic surfaces are preferred for maximizing spatial and polarization DoF, but ULA designs may suffice in azimuthally restricted deployments.
Polarization Diversity: Full vector-field exploitation (three orthogonal polarizations) maximizes DoF; dual-polarization may suffice at lower complexity when Rx elevation is small.
Aperture Sizing: Optimal aperture scales linearly with user distance, SNR, and polarization count. Oversizing yields diminishing returns due to edge inefficiency.
Multi-User Scheduling: Exploiting user visibility regions and spatial non-stationarity reduces multiuser detection complexity and energy consumption.
Hardware-Complexity Reduction: Advanced analog beamforming (MiLAC, switches), and passive metasurfaces facilitate scalability; energy-efficient IRS/ITS architectures maintain sum-rate performance at nearly flat power cost as passive aperture size increases.

As the field progresses, integration of ultra-wideband, sub-terahertz, and tight mutual-coupling designs promises further bandwidth gain and DoF scaling (Akrout et al., 2022). The theoretical framework presented in (Agustin et al., 2024) establishes the basis for analytic sizing, capacity prediction, and DoF quantization of large-aperture multi-polarized holographic MIMO, guiding future 6G and beyond deployments in ultra-dense and high-frequency environments.