Higher-Order Spatial Moments

• Higher-order spatial moments are algebraic and statistical measures that encode spatial data geometry by evaluating powers and products of positions.
• They play a crucial role in modeling anisotropy in heavy-ion collisions, characterizing image shapes, and establishing scaling limits in lattice models.
• Computational techniques like the discrete Radon transform and orthogonal polynomial methods ensure numerical stability and robust extraction of these moments.

Higher-order spatial moments are algebraic or statistical quantities that encode information about the structure, geometry, and fluctuations of spatially distributed data by measuring powers and products of positions (or functions thereof). These moments appear across mathematical physics, stochastic processes, image analysis, and condensed matter systems, serving as descriptors, constraints, and order parameters. Their asymptotics, computation, transformation properties, and physical consequences are central to both theory and application.

1. Formal Definitions and Mathematical Foundations

Let $\{X_i\}_{i=1}^n$ be random variables (often vector-valued), or consider a spatially indexed function (e.g., an image intensity $f(x,y)$) or a lattice occupation field. The higher-order moment of order $r$ typically refers to

$$\mathbb{E}\left[ X_{i_1}^{\alpha_1}\cdots X_{i_k}^{\alpha_k} \right]\,,\qquad \sum_{j=1}^k \alpha_j = r\,,$$

or, in continuous/discrete integrals,

$m_{pq} = \iint x^p y^q f(x,y)\, dx\,dy$

for bivariate spatial data.

On lattices or in field theories, multipole moments are defined as sums over powers of coordinates weighted by the field:

• Charge ($0$th): $Q = \sum_i n_i$
• Dipole ($1$st): $P_i = \sum_j j_i n_j$
• Quadrupole ($2$nd): $Q_{ij} = \sum_k k_i k_j n_k$
• General $N$th: $$Q_{N,d}^{(\alpha)} = \sum_\ell M_{N,d}^{(\alpha)}(\ell) n_\ell$$ where $M_{N,d}^{(\alpha)}$ enumerates the symmetric monomials of degree $N$ in $d$ dimensions (Wang et al., 2024).

In spatial anisotropy analyses of heavy-ion collisions, higher-order moments $\epsilon_n$ quantify the n-fold harmonic deformation of a distribution of points (participants), defined as

$$\epsilon_n = \frac{ \sqrt{ \langle r^2 \cos(n\phi) \rangle^2 + \langle r^2 \sin(n\phi) \rangle^2 } }{ \langle r^2 \rangle }$$

after recentring to eliminate translation (Nagle et al., 2010).

Moment tensors provide a coordinate-free encapsulation of higher-order structure, for instance,

$$T = \mathbb{E}[ X_1 \otimes \cdots \otimes X_n ] \in V^{\otimes n}$$

where $V$ is a real vector space (Mathews, 2018).

2. Physical and Statistical Significance

Higher-order spatial moments capture aspects of geometry, fluctuations, and symmetries not accessible to lower-order descriptors:

• In lattice models (percolation, contact processes, trees), the finiteness and scaling of high moments (e.g., sixth, eighth) are essential for proving scaling limits (super-Brownian motion convergence). Specifically, for spread-out oriented percolation and the contact process in high dimensions, all moments exist and obey

$M_r(n) = \sum_x |x|^r P_n(x) \leq C_r n^{r/2}$

for every $r\geq0$, with asymptotic constants tied to the normal law via $E[|Z|^r]$ (Sakai et al., 2018).

• In heavy-ion physics, the angular moments ($\epsilon_n$, $\psi_n$) of the initial spatial distribution determine the magnitude and orientation of anisotropic flow components $v_n$, mapping to final-state observables. Correlations between even/odd moments reflect participant number fluctuations and geometric biases, directly impacting the interpretation of measured harmonics (Nagle et al., 2010).
• In condensed matter (fracton phases, superfluidity), conservation of higher moments (dipole, quadrupole, etc.) constrains allowed dynamics, alters critical dimensions for symmetry breaking, and engenders new phases (e.g., fractonic superfluids with hybridized moment conservation). The structure of Goldstone branches and order–disorder transitions is controlled by the order of the conserved moment(s) (Wang et al., 2024).
• In spatial statistics and morphometrics, the decomposition of higher-order moment tensors via representation theory (Schur–Weyl) yields coordinate- and group-invariant shape descriptors, essential for robust object comparison and classification (Mathews, 2018).

3. Computational Methods and Algorithms

Efficient extraction of higher-order spatial moments across dimension and domain is a technical bottleneck:

• Image analysis (2D/3D moments): The Discrete Radon Transform (DRT) enables fast computation of arbitrary high-order geometric moments by relating 1D projections at various integer slopes to the binomial expansion of the desired moments. By inverting a small Vandermonde-like system per order, moments up to any $N$ are accessible with computational complexity $\mathcal{O}((N+1)MN)$, where $M$ and $N$ are image dimensions (Diggin et al., 2020).
• Orthogonal moments (Hahn polynomial moments): For discrete object representations, higher-order moments based on Hahn polynomials avoid the redundancy and poor conditioning of geometric moments. However, the standard closed-form initial values become numerically unstable for large order or parameter ($\alpha,\beta$), due to gamma-function overflow/underflow. A stabilized approach employs logarithmic gamma functions for robust initialization and blends $x$– and $n$–direction recurrences with adaptive cutoffs to control error propagation. Computational work is $\mathcal{O}(N^2)$, and reliability is demonstrated up to $N > 10^4$ (Mahmmod et al., 2021).
• Moment tensor transformation: The Schur transform projects the raw moment tensor onto irreducible $S_n \times GL(V)$ components, requiring formation of (potentially) large projectors but yielding robust, group-theoretic features for statistical learning or shape quantification (Mathews, 2018).

4. Group-Theoretic and Statistical Structure

The symmetrization and decomposition of higher-order spatial moments underlie their interpretation and application:

• Schur–Weyl duality: The space $V^{\otimes n}$ admits decomposition into irreducible representations indexed by partitions $\lambda\vdash n$, allowing systematic extraction of symmetric, alternating, and mixed-symmetry features from moment tensors. The projection onto each subspace is constructed via character sums (analogous to Fourier projection), and coefficients serve as quantitative invariants for the data (Mathews, 2018).
• Cumulants and higher moments: Central moments (subtracting means) and cumulants (Möbius inversion along set partitions) provide hierarchical/separable information, with cumulants isolating genuine multi-point dependencies.
• Anisotropy and event-plane moments: For spatial anisotropy, the participant-plane angles $\psi_n$ are defined modulo $2\pi/n$, necessitating specialized metrics (e.g., RMS of angular differences modulo periodicity) for correlation analysis (Nagle et al., 2010).

5. Domain-Specific Applications

The role of higher-order spatial moments is pervasive:

• Image and shape analysis: Moments robustly characterize global shape, local deformations, and symmetries. Higher moments discriminate finer features (e.g., kurtosis, n-fold rotations), critical for classification, retrieval, and medical imaging (Diggin et al., 2020, Mathews, 2018).
• Lattice critical phenomena: In percolation/contact processes/trees, high moments control the probability tail for large excursions, enabling rigorous scaling limits and connections to continuum superprocesses (Sakai et al., 2018).
• Fractonic condensed matter: Higher-moment conservation laws define classes of fractonic superfluids, where restricted mobility and Goldstone dispersion ($\omega \propto |k|^{N+1}$) arise. Hybridization (e.g., dipole–quadrupole) in multi-species systems further structures phase diagrams and critical dimensions for order (Wang et al., 2024).
• Heavy-ion geometry and hydrodynamics: Spatial anisotropy moments frame the mapping between initial geometry and collective flow harmonics, with higher-order correlations constrained by both statistics (finite participant number) and geometry (Nagle et al., 2010).

6. Asymptotics, Scaling, and Stability

The viability and interpretation of higher-order spatial moments depend on control of large-order/large-size limits:

• Scaling in lattice models: Explicit asymptotic formulas connect moments to the standard normal distribution,

$$M_r(n) = c_r n^{r/2} + o(n^{r/2})\,,\qquad c_r = A_0 (A_2/(A_0 d))^{r/2} E[|Z|^r]$$

(Sakai et al., 2018).

• Numerical stability: For orthogonal polynomial moments at high order, standard closed-form expressions become unstable. Stabilized algorithms combine log-gamma initial values and divide computation into domains where different recurrences are numerically robust, using adaptive thresholds to trim propagation of noise (Mahmmod et al., 2021).
• Critical dimensions and condensation: In fractonic systems, the allowed dimensions for symmetry breaking and off-diagonal long-range order (ODLRO) scale with the order $N$ of the conserved moment, e.g., pure $N$th moment conservation yields Goldstone mode $\omega \sim |k|^{N+1}$, and true ODLRO appears only for $d > N+1$ (Wang et al., 2024).

7. Interactions, Correlations, and Misconceptions

Empirical and theoretical results correct prevailing assumptions about the independence and alignment of higher-order spatial moments:

• Contrary to the widespread assumption that odd anisotropy moments ($\epsilon_3, \epsilon_5$) are uncorrelated with even ones in nuclear collisions, measurable correlations persist in peripheral events due to strong finite-size fluctuations, only vanishing (to ≲1%) in central events (Nagle et al., 2010).
• Even–even moment plane alignment (e.g., between $\psi_2$ and $\psi_4$) persists across a broad centrality range, with implications for extracting collective flow harmonics.
• In statistical shape analysis, naive moment vectors are not invariant under group actions; only properly projected Schur amplitudes support robust statistical inference across samples (Mathews, 2018).

---

For further technical and domain-specific details, including explicit recurrence formulas, implementation pseudocode, and explicit asymptotic constants, see (Mahmmod et al., 2021, Nagle et al., 2010, Mathews, 2018, Sakai et al., 2018), and (Wang et al., 2024).