Orthogonal Slicing (H-OMA) Insights

• Orthogonal Slicing (H-OMA) is a technique that partitions spaces or resources into mutually independent subspaces, enabling clear decomposition in algebra, data analysis, and wireless communications.
• It decomposes complex structures into manageable slices that facilitate detailed mathematical investigation, robust visualization, and optimized resource allocation.
• While offering strong structural insights and targeted analysis, H-OMA may compromise global optimality and efficiency due to its rigid, orthogonal constraints.

Orthogonal Slicing (H-OMA) refers to a broad class of techniques across algebraic geometry, high-dimensional data analysis, surface reconstruction, and wireless communications in which a space, structure, or resource is partitioned into mutually orthogonal (i.e., non-interfering or independent) "slices" or subspaces. This yields decompositions, optimization, or visualization strategies that exploit the mathematical and combinatorial properties of orthogonality. Across applications, H-OMA serves the dual purposes of simplifying analysis via isolation and providing structural insights, often at the cost of introducing rigidity and potentially forfeiting global optimality or efficiency.

1. Algebraic Geometry: Orthogonal Slicing of Special Orthogonal Groups

In the context of real algebraic geometry, particularly the study of the special orthogonal groups $SO(n)$, orthogonal slicing (H-OMA) encodes a decomposition by imposing specific zero-patterns on matrix entries to define linear subspaces whose intersection with $SO(n)$ yields reducible or irreducible real algebraic varieties.

For $SO(4)$, the hollow-diagonal slice consists of $4 \times 4$ real special orthogonal matrices whose diagonal entries are zero:

• Hollow-diagonal: $m_{ii} = 0$ for all $i$,
• Orthogonality: $MM^T=I_4$,
• Special: $\det M = +1$.

This variety, $HSO(4)$, is cut out in $\mathbb{R}^{16}$ by 4 linear equations (hollowness), 10 quadratics (orthogonality), and a quartic (unit determinant). The resulting space decomposes into 14 irreducible real surfaces: 8 spheres (degree 2), each isomorphic to $S^2$, and 6 tori (degree 4), each isomorphic to $S^1\times S^1$, corresponding respectively to the triangular and square faces of the cuboctahedron. The incidence and intersection lattice of these components is combinatorially matched to the face-lattice of the cuboctahedron: components intersect in a circle if the corresponding faces share an edge, or in points if they share a vertex (Brysiewicz et al., 2 Feb 2026).

By imposing two further generic linear equations, one obtains a totally real witness set of $40$ real points for $SO(4)$, resolving Conjecture 7.1 for $n=4$ regarding the existence of such slices. For $SO(5)$, a unique (up to symmetry) pattern of eight zeros partitions $SO(5)$ into $64$ irreducible surfaces correlating to a 3-polytope with $16$ facets, $18$ vertices, and $48$ edges. No analogous decomposition exists in $SO(6)$, due to the combinatorial and rank constraints of zero-patterns, indicating a deep connection to algebraic matroid theory.

2. High-Dimensional Data Analysis: Orthogonal Slice Tours

Within high-dimensional exploratory data analysis, H-OMA underpins interactive techniques for detecting hidden nonlinear structures (such as hollowness or concavities) that evade direct projection. Given a centered data matrix $X\in\mathbb{R}^{n\times p}$, one projects onto a $k$-dimensional subspace (via $A\in\mathbb{R}^{p\times k}$) while examining slices orthogonal to the projection:

• The orthogonal complement $U\in\mathbb{R}^{p\times(p-k)}$ (basis for the $(p{-}k)$-dimensional orthogonal subspace) is computed via QR or SVD decompositions of $I_p-AA^T$.
• For a given offset $z\in\mathbb{R}^{p-k}$, the slice hyperplane is $S(z) = \{ x\in \mathbb{R}^p : U^T(x-\mu) = z \}$.
• In practice, a slab $S_\epsilon(z)$ of width $h$ around $S(z)$ is defined, with $h = \epsilon^{1/(p-k)}$ calibrated by the desired fraction $\epsilon$ of data to lie within the slab.

The algorithm generates a geodesic interpolation on the Grassmannian manifold between two projection bases $A_1$, $A_2$, ensuring smooth transitions of both $A(t)$ and its orthogonal complement $U(t)$. This slicing is implemented in the animate_slice and display_slice functions of the R package tourr, permitting dynamic visualization of complex data features through slabs that dynamically evolve as the basis rotates (Laa et al., 2019).

3. Surface Reconstruction from Orthogonal Slices

In computational geometry and imaging, H-OMA denotes an approach for reconstructing surfaces from sets of mutually orthogonal 2D cross-sectional contours. Whereas classical methods typically use parallel planar slices (e.g., stacks of $z$-slices), H-OMA leverages three families of orthogonal planes ($x$, $y$, $z$), yielding the following algorithmic structure:

1. Construct axis-aligned grid $M$ of intersection cells from the three families.
2. Each intersection point of contours from orthogonal families is marked as a node point, which unambiguously pins down surface correspondence.
3. Surface-crossing cells are determined by enumerating grid edges crossed by the contours, recording intersection points.
4. A graph is built with node points as vertices and contour-following edges.
5. For each edge, local cell-wise cycles are extracted, triangulating each polygonal surface patch fan-wise around a computed centroid.

This method ensures robustness against ambiguous correspondences and topological correctness, provided sampling intervals ($\Delta x, \Delta y, \Delta z$) are smaller than the smallest relevant feature size. H-OMA substantially reduces both average and maximal Hausdorff reconstruction error over classical (parallel-slice-only or volume-based) techniques and is inherently insensitive to disconnected surface patches or holes unless a slice direction is undersampled (Svitak et al., 2023).

4. Radio Access Network and Communication: Orthogonal Slicing (H-OMA) for Wireless Slices

In 5G and beyond, orthogonal slicing (H-OMA) describes the allocation of non-overlapping time-frequency resources ("slices") to heterogeneous services—enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC):

• Resources, modeled as an $F\times S$ grid of resource elements (REs), are partitioned: fractions $\alpha_B, \alpha_U, \alpha_M$ of the REs are assigned to eMBB, URLLC, and mMTC respectively, with $\alpha_B + \alpha_U + \alpha_M \leq 1$.
• Each service's link, throughput, and reliability expressions are derived in isolation from the others, given the removal of mutual interference under H-OMA.
• Resource split optimization equates to allocating just enough $\alpha_i$ to meet each service's reliability and throughput constraints, with remaining resources going to the most throughput-intensive service.

The resulting feasible region in performance space (e.g., $(R_B, R_U, \lambda_M)$) is convex and "simplex-shaped," with all trade-offs being linear in the resource fractions. While this delivers per-service isolation and predictable guarantees, it introduces inefficiency due to resource waste (e.g., idle slices during bursty URLLC or sporadic mMTC traffic) and precludes exploiting "reliability diversity" or opportunistic reuse by more flexible non-orthogonal multiple access (H-NOMA) (Popovski et al., 2018).

5. Orthogonal Slicing in Hybrid NOMA-OMA Schemes for Satellite Networks

In massive LEO constellations, the hybrid NOMA-OMA (H-OMA) scheme combines intra-group non-orthogonal multiple access with inter-group orthogonal slicing:

• The full resource pool $\mathcal{R}$ (e.g., time-frequency blocks) is divided into $G$ orthogonal slices assigned to groups of satellites.
• Within each slice, group members (satellites) employ NOMA via superposition coding and successive interference cancellation (SIC).
• Grouping strategies optimize either Doppler diversity (anticlustering in frequency shift) or fairness (maximizing Jain’s index), addressing the fairness-capacity trade-off inherent in pure NOMA versus pure OMA.
• In realistic Starlink-type LEO constellations (e.g., $L=19$ satellites, $S=8$), H-OMA schemes achieve $30+$ bits/s/Hz higher sum-rate and sharply better fairness (e.g., Jain index up to $0.997$) compared to pure-OMA, while sacrificing only $\approx 20\%$ of the sum-rate of pure-NOMA (Darsena et al., 2023).

Scheme	Sum-Rate (bits/s/Hz)	Jain Fairness Index
Pure-NOMA	76.645	0.316
Pure-OMA (opt-DoF)	28.237	0.136
H-OMA (max-fair, unif)	54.856	0.997

Compared to both extremes, H-OMA furnishes a controllable trade-off between throughput and fairness through intelligent group partitioning and DoF allocation.

6. Structural, Combinatorial, and Algorithmic Insights

Across all domains, H-OMA is unified by the strategy of decomposing a space, dataset, or resource into mutually orthogonal subspaces to enable tractable analysis, enhance robustness, or ensure fairness and isolation. In algebraic settings, the method reveals deep connections between the decomposability of group varieties ($SO(n)$) and the combinatorics of polytopes—specifically, face lattices encoding intersection patterns. In data analysis and visualization, H-OMA ensures local geometric properties of high-dimensional structures are directly inspectable. In communications, orthogonal slicing yields tractable optimization under isolated service constraints but is intrinsically limited by its linear trade-off structure and rigidity—failing to exploit opportunities for resource sharing or adaptive interference management.

The boundaries of this approach are marked where the underlying algebraic, geometric, or combinatorial structure does not support a dimension-generic and degree-generic decomposition (as for $SO(6)$ and above), or where practical inefficiencies arise due to underutilized resources or lack of global coordination.

7. Implications and Future Directions

The limits of H-OMA in certain settings (e.g., failure to find suitable decompositions in $SO(6)$, spectral inefficiency in bursty wireless scenarios) suggest two directions:

• Algebraic and combinatorial refinement: Seeking new zero-patterns, slice structures, or connections to matroid theory to extend the method's scope in higher-dimensional or more complex algebraic varieties (Brysiewicz et al., 2 Feb 2026).
• Relaxing strict orthogonality: In communications and networks, hybrid schemes (e.g., hybrid NOMA-OMA) and reliability-diversity-aware access leverage non-orthogonality to balance service isolation with dynamic resource reallocation and improved fairness or throughput (Darsena et al., 2023, Popovski et al., 2018).

Ongoing research explores the interplay between combinatorial design, algebraic invariants, orthogonality, and the structure of optimal partitions in both mathematical and applied settings.