Ferkans — Interactive Telecom Tutor

From Arrays to Continuous Apertures

Every MIMO system treated so far in this book assumes a discrete array: $N_t$ antennas at $\lambda/2$ spacing, index $m$ running from $1$ to $N_t$ . Holographic MIMO pushes the abstraction one level up: instead of an array of distinct radiating elements, it models the transmitter as a continuous current distribution $\mathbf{J}(\mathbf{r})$ on a 2D or 3D surface. The unknowns are not discrete beamforming weights but functions on a continuous domain. The degree-of-freedom count becomes a functional-analytic question, not a linear-algebra one.

Pizzo, Marzetta, and Sanguinetti (2020) answered the fundamental question: for an aperture of physical size $L$ and typical communication distance $d$ , the number of usable DoF scales as $(L/L_F)^4$ where $L_F = \sqrt{\lambda d}$ is the Fresnel scale. That scaling is four times as favorable as the $\lambda/2$ -array baseline in terms of $L$ , because both dimensions of a 2D aperture contribute. The open question is whether that theoretical promise survives the transition from idealized physics to manufacturable hardware.

,

Definition:
Holographic MIMO Surface

A holographic MIMO surface is a 2D aperture $\mathcal{A}$ of physical area $|\mathcal{A}| = L_x L_y$ that radiates (or receives) via a continuous current distribution $\mathbf{J}: \mathcal{A} \to \mathbb{C}^3$ . The radiated field at a point $\mathbf{r}$ in space is

$\mathbf{E}(\mathbf{r}) = \int_{\mathcal{A}} \mathbf{G}(\mathbf{r} - \mathbf{r}')\, \mathbf{J}(\mathbf{r}')\,\mathrm{d}\mathbf{r}',$

where $\mathbf{G}$ is the free-space dyadic Green's function at wavelength $\lambda = c/f_0$ . Practical implementations discretize $\mathbf{J}$ by tiling $\mathcal{A}$ with sub-wavelength elements (typical spacing $\lambda/8$ to $\lambda/4$ ) — many more than the $\lambda/2$ -spaced elements of a classical array, but not truly continuous.

The holographic approximation takes the discretization limit: the number of physical elements $N_{\text{HMIMO}}$ satisfies $N_{\text{HMIMO}} \gg (L/\lambda)^2$ , so the spacing becomes negligible relative to $\lambda$ and the continuous-aperture approximation is accurate.

Holographic MIMO is not the same as XL-MIMO or near-field MIMO, although they overlap. XL-MIMO is about aperture physical size; holographic MIMO is about element density at fixed size. A $1$ -meter aperture with $\lambda/2$ spacing is XL-MIMO at 3.5 GHz but not holographic. A $5$ -cm aperture with $\lambda/16$ spacing is holographic but not XL-MIMO.

,

Theorem: Pizzo-Marzetta-Sanguinetti DoF Scaling

Consider a holographic aperture of dimensions $L_x \times L_y$ transmitting to a receiver at distance $d$ with receive aperture $L_{Rx}$ , at wavelength $\lambda$ . Define the Fresnel length $L_F = \sqrt{\lambda d}$ . As $L_x, L_y, L_{Rx} \to \infty$ and the apertures remain in each other's radiative near field, the number of effective communication DoF scales as

$\mathrm{DoF}(L_x, L_y, L_{Rx}) \sim \frac{L_x L_y L_{Rx}^2}{(L_F)^4} \sim \left(\frac{L}{L_F}\right)^4,$

where $L \sim L_x \sim L_y \sim L_{Rx}$ is the common geometric scale. The constant is $\mathcal{O}(1)$ and depends on aperture shape.

In contrast, a $\lambda/2$ -spaced classical array of the same physical size in the far field offers $\mathrm{DoF}_{\text{classical}} \sim (L_x L_y)/\lambda^{2} \sim (L/\lambda)^2$ — a $(L_F/\lambda)^{-2}$ -smaller DoF count per unit area.

In the far field, only angular (direction-of-arrival) diversity is available, and that scales with aperture area in wavelength units. In the near field (holographic regime), both angular and radial diversity are available because wavefront curvature varies measurably across the aperture. The four power of $L$ comes from two powers for angular and two for radial; the Fresnel length $L_F$ sets the scale below which radial resolution vanishes.

Show Hint

Parameterize the received field in terms of an angular spectrum and count the effective modes below a given cutoff.

Apply Weyl's law to the Helmholtz equation on the aperture to relate the mode count to the aperture area.

The near-field correction enters through the wavefront curvature: $\Delta \phi \approx L^2/(\lambda d)$ radians across the aperture.

Proof

Angular spectrum decomposition

The field on the aperture can be expanded in a 2D Fourier basis of plane waves with wavevector $(k_x, k_y)$ constrained by $k_x^2 + k_y^2 \leq k^2 = (2\pi/\lambda)^2$ (propagating modes). The number of such modes is proportional to the area times $k^2/(4\pi) \sim L^2/\lambda^{2}$ — the classical antenna-theory count.

Near-field correction

At finite range $d$ , wavefront curvature $\delta(\mathbf{r}) = |\mathbf{r}|^2/(2d)$ introduces a position-dependent phase that decorrelates plane waves. The effective spatial coherence length shrinks to $L_F = \sqrt{\lambda d}$ . This adds $(L/L_F)^2$ radial modes on top of the $(L/\lambda)^2$ angular modes.

Multiplication

The total usable DoF is the product $(L/\lambda)^2 \times (L/L_F)^2 = (L^2/\lambda)(L^2/\lambdad) \sim L^4/(\lambda^{2} d)$ . Dividing by $L_F^4 = \lambda^{2} d^2$ gives the $(L/L_F)^4$ scaling after normalizing per unit range. The rigorous argument via Slepian prolate spheroidal wave functions and the Landau-Pollak-Slepian eigenvalue theorem appears in Pizzo et al. (2020), Theorem 1. $\blacksquare$

Holographic DoF vs Aperture Size

Compare the Pizzo-Marzetta-Sanguinetti $(L/L_F)^4$ scaling to the classical $(L/\lambda)^2$ far-field scaling. Tune the carrier frequency and communication distance to see where holographic gains become material and where they vanish.

Parameters

Carrier

f_0

(GHz)28

Link distance

d

(m)50

Max aperture

L

(m)2

Example: Holographic vs Classical DoF at 28 GHz

Compare the effective DoF of a $1\,\text{m} \times 1\,\text{m}$ holographic surface to a classical $\lambda/2$ -spaced array of the same size, at $f_0 = 28$ GHz, with a link distance $d = 50$ m to a similar receive aperture.

Solution

Wavelength

$\lambda = c/f_0 = 3 \times 10^8 / 28 \times 10^9 \approx 1.07$ cm.

Fresnel length

$L_F = \sqrt{\lambda d} = \sqrt{0.0107 \cdot 50} \approx 0.73$ m. Since $L = 1$ m $> L_F$ , we are deeply in the near field.

Classical DoF (far-field array)

Classical array count: $(L/\lambda)^2 = (1/0.0107)^2 \approx 8720$ elements. Per 2D aperture: $8720 \times 8720 \approx 7.6 \times 10^7$ potential "DoF", but only those that the receiver can distinguish matter. In the far field, the achievable DoF is limited by the angular resolution of the link, not by the element count. Effective DoF: $\sim (L L_{Rx})/(d \lambda) \approx (1)(1)/(50 \cdot 0.0107) \approx 1.9$ .

Holographic DoF (near-field correction)

$(L/L_F)^4 = (1/0.73)^4 \approx 3.5$ . Rough estimate: $3$ - $4$ near-field DoF per pair of apertures, compared to $\approx 2$ in the far-field limit.

Interpretation

The holographic improvement is a factor of $\approx 2$ in DoF for this geometry — not dramatic, but meaningful. The improvement grows rapidly as $L/L_F$ increases: at $L = 2$ m, $(L/L_F)^4 \approx 56$ DoF, versus $\approx 8$ classical. The scaling asymmetry in aperture size is what drives the holographic argument. $\blacksquare$

🔧Engineering Note

Manufacturing Constraints on Holographic Surfaces

The theoretical DoF scaling assumes a truly continuous current distribution with arbitrary amplitude and phase per unit area. Real holographic surfaces are built from metamaterial unit cells at sub-wavelength spacing ( $\lambda/8$ is typical) and controlled via varactors, PIN diodes, or MEMS switches. Each unit cell offers a small finite set of states $(1$ - $4$ bits per cell), not continuous phase control. Mutual coupling between adjacent cells makes the effective per-cell steering vector depend on its neighbors — the "isolated element" assumption that classical array theory relies on breaks. Heat dissipation across a meter-scale surface at $100$ MHz switching rates is also non-trivial; a $1$ m $^2$ RF-reconfigurable surface at $28$ GHz can dissipate $\sim 10$ W of control-circuit power, comparable to a classical AAU.

Practical Constraints

•
Unit-cell spacing: target $\leq \lambda/4$ at $f_0$ ; currently $\lambda/8$ to $\lambda/2$ in prototypes
•
Per-cell phase resolution: typically 1-4 bits (not continuous)
•
Mutual coupling: unmodeled couplings degrade achievable DoF by 3-6 dB
•
Power dissipation: $\sim 1$ - $10$ W/m $^2$ for active control, $\sim 0.1$ W/m $^2$ for passive RIS-style designs
•
Calibration: per-cell state mapping must account for temperature and bias drift

📋 Ref: IEEE 1918.1 Working Group on Holographic MIMO (pre-standard activity, 2024-)

,

Historical Note: Prolate Spheroidal Wave Functions: 1960 Origins

1961-present

The eigenvalue theory that underlies holographic MIMO was developed by David Slepian, Henry Landau, and Henry Pollak at Bell Labs between $1961$ and $1964$ in a series of Bell System Technical Journal papers on "Prolate Spheroidal Wave Functions, Fourier Analysis and Uncertainty." Their result: the number of nearly orthogonal functions that can simultaneously be time-limited ( $|t| \leq T$ ) and bandwidth-limited ( $|f| \leq W$ ) is approximately $2TW$ . The identical mathematical structure governs the number of nearly orthogonal spatial modes on an aperture of size $L$ at wavelength $\lambda$ : it is the $\lambda$ -domain version of the "space-bandwidth product."

For half a century these results lived in antenna-theory and optical-imaging papers. Pizzo, Marzetta, and Sanguinetti (2020) imported them into MIMO theory, which closed a circle: the same mathematics that Shannon cited in 1949 as the backbone of signal-space dimension counting now bounds the DoF of a holographic wireless link.

Historical Note: The Pizzo-Marzetta-Sanguinetti Paper That Started the Field

2020-present

The modern holographic MIMO literature dates from Pizzo, Marzetta, and Sanguinetti's 2020 IEEE JSAC paper "Spatially-Stationary Model for Holographic MIMO Small-Scale Fading." The paper is notable for two reasons. First, it gave the first clean DoF scaling law for a continuous aperture in the near field, the $(L/L_F)^4$ result. Second, it framed the problem in Shannon-theoretic language that MIMO researchers already understood, removing the wall that had kept electromagnetic-theory results out of information-theoretic MIMO work. The paper has been cited over $500$ times as of 2026 and is the reference point for every subsequent holographic MIMO paper.

The Open Problem, Stated Precisely

The open problem is the achievability gap between theoretical holographic DoF and what a manufacturable surface can realize. Specifically:

Given a manufacturable holographic surface (finite unit cells, finite phase resolution, non-negligible mutual coupling), what fraction of the $(L/L_F)^4$ theoretical DoF is actually achievable?
What is the optimal trade-off between unit-cell spacing, phase resolution, and control power?
Do active surfaces (amplifier per cell) scale differently from passive surfaces (phase-shifter only) in the holographic regime?
For a fixed material budget, is one big holographic surface better than many distributed smaller surfaces?

Questions 3 and 4 are where the section connects to Section 27.5: a distributed holographic network is indistinguishable from an ultra-dense RIS-assisted cell-free network at the right parameter regime.

,

Common Mistake: Holographic MIMO Does Not Replace Massive MIMO

Mistake:

Holographic MIMO is sometimes billed as a successor to massive MIMO that will displace it in 6G networks.

Correction:

The $(L/L_F)^4$ gain requires the link to be in the near field ( $L \gtrsim L_F$ ). For a $1$ -meter aperture at 3.5 GHz, $L_F = 1$ m at only $d = 11$ m. Beyond that range, the gain disappears and classical massive MIMO math takes over. Holographic MIMO is a short-range, high-frequency, indoor/near-cell technology. It does not replace the long-range, low-frequency mmWave and sub-6 GHz massive MIMO arrays that carry most 5G traffic. It is an additional tool in the 6G kit.

Why This Matters: Echo of Chapter 18: Near-Field Meets Holographic

Chapter 18 introduced near-field (Fresnel) communication for XL-MIMO arrays and showed that the Fraunhofer boundary $d_F = 2L^2/\lambda$ moves the "near field" to $\sim 100$ m for typical 5G arrays at 3.5 GHz. Section 27.4 stays in the same near field but pushes the aperture density from $\lambda/2$ to $\lambda/8$ . Chapter 18 gave us the Fresnel machinery; Section 27.4 asks what happens when we use all of it.

Holographic MIMO

A MIMO architecture in which the transmit/receive surface is modeled as a continuous current distribution, approximated in practice by unit cells at spacing much smaller than $\lambda/2$ . Achieves $(L/L_F)^4$ DoF scaling in the radiative near field, where $L$ is the aperture size and $L_F = \sqrt{\lambda d}$ is the Fresnel length.

Fresnel Length

The geometric scale $L_F = \sqrt{\lambda d}$ , where $\lambda$ is the wavelength and $d$ is the communication range. Apertures of size $L \ll L_F$ see planar wavefronts (far field); apertures of size $L \gtrsim L_F$ see curved wavefronts (near field) and can exploit radial diversity. Central to the Pizzo-Marzetta-Sanguinetti DoF theorem.

Quick Check

A holographic aperture has physical size $L$ and a receiver at distance $d$ with the same aperture size. In the radiative near field, the number of effective communication DoF scales as:

$L/\lambda$

$(L/\lambda)^2$

$(L/L_F)^4$ with $L_F = \sqrt{\lambda\,d}$

$(L/L_F)^2$