Favorable Propagation Metric

Updated 19 January 2026

Favorable propagation metric is a quantitative measure that assesses how desired channel decorrelation occurs, crucial for minimizing interference in systems like massive MIMO.
It utilizes normalized variance and inner product calculations to characterize the degree of orthogonality in wireless channels and cascading networks.
Applied across wireless communications, semi-supervised learning, and cascading failure analysis, these metrics guide deterministic and statistical optimization of system performance.

A favorable propagation metric is a quantitative measure used in a variety of scientific and engineering domains to characterize how desirable propagation phenomena—such as orthogonality in wireless channels, label proliferation in semi-supervised learning, or generation decay in cascading failures—occur in systems comprised of many interacting elements. In massive MIMO wireless systems, "favorable propagation" refers specifically to the near-orthogonality of channel vectors between users, enabling spatial multiplexing and low inter-user interference. The metric is typically constructed as a scalar or normalized functional quantifying the degree to which a propagation environment or similarity metric promotes the desired property. Deterministic and statistical versions exist, and their closed forms depend on array geometry, channel models, and system parameterizations. These metrics are critical for the design, analysis, and optimization of complex systems in wireless communications, semi-supervised learning, cascading failure assessment, and even general relativity.

1. Formal Definition and Mathematical Formulations

In multiuser MIMO and related fields, a system exhibits favorable propagation when channel vectors $\mathbf{h}_i$ and $\mathbf{h}_j$ of distinct users satisfy

$\lim_{M \to \infty} \frac{1}{M}\mathbf{h}_i^H\mathbf{h}_j = 0 \qquad \forall\,i \neq j$

where $M$ is the number of antennas or system degrees of freedom. For finite $M$ , metrics quantifying proximity to this ideal take the form

$\mathsf{FP}_{ij} = \frac{\mathrm{Var}(\mathbf{h}_i^H\mathbf{h}_j)}{\mathbb{E}[\|\mathbf{h}_i\|^2]\mathbb{E}[\|\mathbf{h}_j\|^2]}$

or, equivalently, as the mean-squared normalized inner product

$\kappa^{\rm FP} = \frac{1}{M^2}\mathbb{E} \left[ | \mathbf{h}_i^H\mathbf{h}_j |^2 \right ]$

For RIS-assisted systems, the aggregated channel $z_k=u_k + H\Phi g_k$ yields

$\mathsf{FP}_{k\ell} = \frac{\mathrm{Var}(z_k^H z_\ell)}{\mathbb{E}[||z_k||^2]\,\mathbb{E}[||z_\ell||^2]}$

Each metric converges to zero as $M,N$ grow, provided model assumptions (far field, continuous angle/spread distributions) are satisfied (Li et al., 2020, Chien et al., 2021, Chen et al., 12 Jan 2026, Miller et al., 2020, Ngo et al., 2014).

In cascading propagation analysis, e.g. utility line outages, the System Event Propagation Slope Index (SEPSI) quantifies propagation via a Zipf distribution: $P[G = k] = \frac{1}{\zeta(s)}\,\frac{1}{k^s}$ where $s=\text{SEPSI}$ is fitted via maximum likelihood to the observed generation counts, and decreasing $s$ corresponds to heavier tails (more severe propagation) (Dobson, 2018).

2. Physical and Statistical Interpretations

Favorable propagation metrics measure inter-element (typically inter-user) channel correlation or orthogonality. In wireless systems, low values indicate high orthogonality and hence minimal interference, unlocking spatial multiplexing. Statistically, these metrics reflect how array geometry, environmental scattering, propagation models, and deployment (e.g., near vs. far field for IRS) affect channel decorrelation:

ULA/UPA models with continuous angle distributions promote channel orthogonality (Li et al., 2020, Miller et al., 2020).
In the presence of LoS links (rank-deficient channels), favorable propagation can be engineered via near-field deployments and array aperture tuning (Chen et al., 12 Jan 2026).
The convergence rate to the favorable propagation regime is generally $O(1/M)$ or $O(\log M/M)$ , robust to cluster sharing in realistic scenarios (Miller et al., 2020).
For cascading outages, the SEPSI metric physically encodes the system's propensity for cascade growth, with environmental factors (e.g., storms) shifting the propagation strength (Dobson, 2018).

3. Deterministic and Statistical Metrics: Closed-Form Expressions

Deterministic closed-form favorable propagation metrics allow rigorous comparison and optimization. For RIS/IRS-assisted massive MIMO,

$g_{jk} \approx \frac{1}{M^2\,N^2} \sum_{s,t} w(s,t)f_M(\Delta_{s,t})$

where $f_M(\cdot)$ is a squared Dirichlet kernel and $\Delta_{s,t}$ encodes phase increment induced by geometric configuration (Chen et al., 12 Jan 2026). For Rayleigh fading,

$\mathsf{FP}_{k\ell} = \frac{\mathrm{tr}[(R_k + \mathrm{tr}\Theta_k R_s)(R_\ell + \mathrm{tr}\Theta_\ell R_s)]}{(M\beta_k + M\beta_s \mathrm{tr}\Theta_k)(M\beta_\ell + M\beta_s \mathrm{tr}\Theta_\ell)}$

and in the i.i.d. case,

$\mathsf{FP}_{k\ell} = \frac{\beta_k\beta_\ell + N(\beta_k\xi_\ell + \beta_\ell\xi_k) + N^2\xi_k\xi_\ell}{M(\beta_k + N\xi_k)(\beta_\ell + N\xi_\ell)}$

(Chien et al., 2021). For generic ULA/HURA/UCA systems,

$\kappa^{\rm FP} = \beta^{(l)}\beta^{(l')}\left( (1-p_\text{sh})K_c + p_\text{sh}K_s \right)$

with $K_c$ and $K_s$ capturing shared vs. distinct cluster effects (Miller et al., 2020).

Array topology, user clustering, and intelligent surface deployment strongly influence favorable propagation. Key points:

ULAs generally outperform UCAs and HURAs in azimuth-diversity, achieving lower mean-squared distances (MSD) in $\kappa^{\rm FP}$ (Miller et al., 2020).
Cluster sharing increases correlation for finite arrays but the asymptotic decay $O(1/M)$ is preserved (Miller et al., 2020).
IRS deployment in the near field allows channel decorrelation via spherical wavefront exploitation, overcoming rank-deficiency in cascaded LoS links. Criterion: IRS position and array geometry must be tuned to align phase increments with Dirichlet kernel nulls, yielding $g_{jk}<2/M$ for square IRS with sufficient BS aperture (Chen et al., 12 Jan 2026).
RIS phase optimization can further minimize $\mathsf{FP}$ , and rich scattering environments accelerate convergence to orthogonality (Chien et al., 2021).

Metric Symbol	Expression	Description
$\mathsf{FP}_{ij}$	$\frac{\mathrm{Var}(\mathbf{h}_i^H\mathbf{h}_j)}{\mathbb{E}[\|\|\mathbf{h}_i\|\|^2]\mathbb{E}[\|\|\mathbf{h}_j\|\|^2]}$	Normalized variance; RIS/MMIMO applications
$\kappa^{\rm FP}$	$\frac{1}{M^2}\mathbb{E}[\|\mathbf{h}_i^H\mathbf{h}_j\|^2]$	Mean-squared distance; topological/channel dependency
$g_{jk}$	Given above (Chen et al., 12 Jan 2026)	Near-field IRS design; deterministic decorrelation metric

5. Methodologies for Metric Calculation and Optimization

Favorable propagation metrics are calculated using statistical moment computations, numerical optimization for system deployment, and closed-form summations in spatially regular arrays. In practice:

Analytical methods (Dirichlet kernel summations, characteristic functions) for on-array topologies (Li et al., 2020, Miller et al., 2020, Chen et al., 12 Jan 2026).
Rayleigh fading and correlated channel models allow tractable expressions via covariance matrices (Chien et al., 2021).
For cascading propagation analysis, empirical counting and maximum-likelihood fitting establish Zipf parameters, followed by goodness-of-fit testing (Dobson, 2018).

Optimization involves IRS phase selection, BS aperture tuning, and user scheduling (cluster sharing minimization). For IRS/BS joint design, geometry-driven deployment rules ensure $g_{jk}\to0$ , while alternating optimization or fractional programming can maximize ergodic sum-rates under favorable channel statistics (Chen et al., 12 Jan 2026, Chien et al., 2021).

6. Practical Applications and System-Level Impact

Favorable propagation metrics are essential for:

Massive/MU-MIMO systems: Maximizing spatial multiplexing gains and minimizing inter-user interference. The metric guides antenna array design, user scheduling, deployment strategies, and phase optimization (Ngo et al., 2014, Li et al., 2020, Miller et al., 2020, Chien et al., 2021, Chen et al., 12 Jan 2026).
RIS/IRS-assisted communication: Facilitating channel decorrelation in rank-deficient environments, especially with strategic near-field deployment (Chen et al., 12 Jan 2026, Chien et al., 2021).
Semi-supervised learning and metric-based label propagation: Transferred similarity metrics promote "favorable propagation" of labels from small labeled sets to large unlabeled pools; globalized spectral methods outperform local voting under limited supervision (Liu et al., 2018).
Cascading propagation in utility networks: The SEPSI metric quantifies propagation risk, system vulnerability, and event-size probabilities under varied stress conditions (Dobson, 2018).
Gravitational light propagation: Metric structure at 2PN/RM order completes relativistic modeling for high-precision astrophysical experiments, with potential to probe scalar-tensor theory parameters (Minazzoli et al., 2010).

7. Limitations, Failure Modes, and Directions for Future Research

Favorable propagation is not guaranteed for all topologies or under certain parameterizations:

Non-continuous or clustered user angles can preclude convergence to orthogonality (Li et al., 2020, Miller et al., 2020).
Cluster sharing in practical arrays increases correlation but large-system limits remain robust (Miller et al., 2020).
Far-field LoS propagation is inherently rank-deficient; mitigation requires engineered near-field deployments (Chen et al., 12 Jan 2026).
In RIS-assisted systems, phase-shift optimization under imperfect CSI is a practical challenge; deterministic metrics allow optimization over second-order statistics but may omit instantaneous channel effects (Chien et al., 2021).
In semi-supervised learning, transferability of metrics is highly dependent on the pretraining domain's semantic richness and alignment with the target domain (Liu et al., 2018).

Continued research seeks new array topologies, more robust deployment criteria, improved metric learning strategies for data propagation, and more general distributional convergence theorems for propagation metrics in random environments.

Favorable propagation metrics are central to both theoretical analysis and practical design in wireless communications, learning, cascading systems, and relativistic astrophysics, offering rigorous tools for predicting, optimizing, and interpreting propagation phenomena across complex multi-element systems.