Cubosimplicial Subdivision: Polyhedral Perspectives

Updated 6 July 2026

Cubosimplicial subdivision is a method where every cell is a product of simplices, unifying cubical and simplicial subdivision theories.
It leverages common local invariants, decomposition formulas, and poset-theoretic frameworks to create a bridge between different subdivision paradigms.
The approach finds applications in marked order polytopes, toric geometry, and lattice refinements, offering actionable insights in both algebraic and computational contexts.

In the literature considered here, cubosimplicial subdivision appears in two closely related senses. In one sense, it is a polyhedral subdivision in which every cell is a product of simplices; this is explicit for the subdivision $K_{P,\lambda}$ of a marked order polytope. In the other sense, it is a broader viewpoint in which cubical and simplicial subdivision theories are developed in parallel, with common local invariants, common decomposition formulas, and a common poset-theoretic framework for formal subdivisions of locally Eulerian posets (Melikhova, 18 Jul 2025, Athanasiadis, 2010).

1. Terminology and conceptual scope

A polyhedral complex is cubosimplicial if every cell is a product of simplices (Melikhova, 18 Jul 2025). This definition is literal in the setting of marked order polytopes, where the cells of the subdivision $K_{P,\lambda}$ are products of simplices, and it is also literal in the finite subdivision rule for the $n$ -torus, whose tile types are $I^q \times \Delta^{p-1}$ with $p+q=n$ (Rushton, 2011).

A broader use of the term is suggested by the cubical analogue of Stanley’s local $h$ -theory. That theory does not define a literal “cubosimplicial subdivision,” but it does set up a point-by-point parallel between simplicial subdivisions and cubical subdivisions, and then extends both to formal subdivisions of locally Eulerian posets (Athanasiadis, 2010). In that setting, simplicial and cubical phenomena are no longer separated by cell shape alone; they are organized by the same local-to-global mechanism.

This broader viewpoint is reinforced by results on cubical complexes embeddable in cubes. Cubical barycentric subdivisions of simplicial complexes are listed among the classes of cubical complexes that embed in a cube, so a simplicial input can produce a cubical output without leaving a common ambient framework (Rowlands, 2019). A plausible implication is that “cubosimplicial” is best understood not as a single rigid cell type, but as a family of compatible subdivision formalisms that move between simplicial, cubical, and product-of-simplex geometries.

2. Local $h$ -theory and the cubical–simplicial bridge

For an abstract cubical complex $K$ , a cubical subdivision is a cubical complex $K'$ together with a surjective subdivision map

$\sigma:K'^*\to K^*$

such that, for each face $K_{P,\lambda}$ 0, the restriction $K_{P,\lambda}$ 1 is a ball of dimension $K_{P,\lambda}$ 2, and $K_{P,\lambda}$ 3 is exactly the set of interior faces of that ball (Athanasiadis, 2010). This is the direct cubical analogue of simplicial subdivision.

The corresponding global enumerative invariant is the short cubical $K_{P,\lambda}$ 4-polynomial

$K_{P,\lambda}$ 5

which plays the role of the simplicial $K_{P,\lambda}$ 6-polynomial. For a cubical subdivision $K_{P,\lambda}$ 7 of a $K_{P,\lambda}$ 8-cube $K_{P,\lambda}$ 9, the short cubical local $n$ 0-polynomial is

$n$ 1

These local polynomials satisfy the decomposition theorem

$n$ 2

so the global cubical $n$ 3-polynomial of a subdivision splits into local cubical contributions multiplied by simplicial $n$ 4-polynomials of links (Athanasiadis, 2010).

The short cubical local $n$ 5-polynomial is symmetric,

$n$ 6

and for locally quasi-geometric cubical subdivisions its coefficients are nonnegative. Every geometric cubical subdivision is locally quasi-geometric. The same work extends the theory to formal subdivisions of locally Eulerian posets, where generalized $n$ 7-polynomials and local $n$ 8-polynomials recover the usual simplicial $n$ 9-polynomial, the short simplicial $I^q \times \Delta^{p-1}$ 0-polynomial of Hersh–Novik, and the short cubical $I^q \times \Delta^{p-1}$ 1-polynomial in one common framework (Athanasiadis, 2010). This is the main structural basis for a cubosimplicial viewpoint in the broader sense.

3. Canonical subdivision operations

A canonical cubical subdivision operation is cubical barycentric subdivision. For a cubical complex $I^q \times \Delta^{p-1}$ 2, the vertices of $I^q \times \Delta^{p-1}$ 3 are the barycenters of nonempty faces of $I^q \times \Delta^{p-1}$ 4, and its higher-dimensional faces are the convex hulls of barycenters of all faces in a closed interval $I^q \times \Delta^{p-1}$ 5 of the face poset $I^q \times \Delta^{p-1}$ 6. The face poset of $I^q \times \Delta^{p-1}$ 7 is isomorphic to the poset of closed intervals in $I^q \times \Delta^{p-1}$ 8, so $I^q \times \Delta^{p-1}$ 9 is again a cubical complex (Savvidou, 2010).

Its enumerative effect is explicit. If $p+q=n$ 0 is a $p+q=n$ 1-dimensional cubical complex, then

$p+q=n$ 2

and

$p+q=n$ 3

From these formulas, symmetry and nonnegativity of the short and long cubical $p+q=n$ 4-vectors are preserved under cubical barycentric subdivision, and real rootedness of the short cubical $p+q=n$ 5-polynomial is preserved as well (Savvidou, 2010).

An explicit cubosimplicial finite subdivision rule is provided for the $p+q=n$ 6-torus. Using the standard simplicial decomposition of the hypercube, the tile types are

$p+q=n$ 7

and each such tile is subdivided into one tile of the same type $p+q=n$ 8 and $p+q=n$ 9 tiles of type $h$ 0 (Rushton, 2011). Here the simplex factor comes from the simplicial decomposition of the “rank $h$ 1” directions, while the remaining directions stay cubical. This is a literal cubosimplicial subdivision rule.

4. Marked order polytopes and the subdivision $h$ 2

For a finite poset $h$ 3, a subset $h$ 4 containing all extremal elements, and an order-preserving map $h$ 5, the marked order polytope

$h$ 6

consists of all $h$ 7 such that $h$ 8 if $h$ 9, and $h$ 0 for every $h$ 1 (Melikhova, 18 Jul 2025). The cubosimplicial subdivision $h$ 2 is indexed by $h$ 3-admissible chains of order ideals

$h$ 4

To such a chain $h$ 5 one associates a polyhedron $h$ 6 consisting of all $h$ 7 such that $h$ 8, $h$ 9 is constant on each difference $K$ 0, and the corresponding floor-values form a nondecreasing sequence. If $K$ 1 is $K$ 2-admissible, then $K$ 3 is nonempty, lies inside $K$ 4, and the correspondence $K$ 5 is bijective onto the nonempty cells of this form. The complex

$K$ 6

is a polyhedral complex, and its face poset is the poset of $K$ 7-admissible chains ordered by inclusion (Melikhova, 18 Jul 2025).

Geometrically, $K$ 8 coincides with the hyperplane subdivision obtained by intersecting $K$ 9 with the hyperplanes $K'$ 0 for incomparable $K'$ 1 and $K'$ 2 for incomparable $K'$ 3, $K'$ 4. It is cubosimplicial because each maximal cell $K'$ 5 is a product of simplices (Melikhova, 18 Jul 2025).

The same construction supports a cohomological computation of face numbers. If $K'$ 6 is the $K'$ 7-vector space with basis all $K'$ 8-admissible chains of dimension $K'$ 9, and $\sigma:K'^*\to K^*$ 0 is the differential defined by densifications that come in conjugate pairs, then the resulting cochain complex $\sigma:K'^*\to K^*$ 1 is isomorphic to the geometric complex attached to the subdivision $\sigma:K'^*\to K^*$ 2, and

$\sigma:K'^*\to K^*$ 3

Thus the $\sigma:K'^*\to K^*$ 4-vector of a marked order polytope is recovered from a purely combinatorial complex built from the cubosimplicial subdivision (Melikhova, 18 Jul 2025).

For cubical homology manifolds embedded as cubical subcomplexes of a cube, moderately high skeleta determine the entire cubical structure. If $\sigma:K'^*\to K^*$ 5 is a $\sigma:K'^*\to K^*$ 6-dimensional cubical homology manifold without boundary, then $\sigma:K'^*\to K^*$ 7 is determined by its $\sigma:K'^*\to K^*$ 8-skeleton. Under additional hypotheses, including the case of cubical spheres, the bound improves to the $\sigma:K'^*\to K^*$ 9-skeleton for $K_{P,\lambda}$ 00 (Rowlands, 2019). This reconstruction theory is relevant to cubosimplicial structures because cubical barycentric subdivisions of simplicial complexes are among the cubical complexes that embed in cubes, and because the recognition of subcomplexes isomorphic to $K_{P,\lambda}$ 01 is a cubical analogue of the simplicial face-recognition problem (Rowlands, 2019).

A different geometric realization of cube-based refinement is the hierarchical subdivision of the simple cubic lattice. Starting from the simple cubic lattice with lattice constant $K_{P,\lambda}$ 02, interstitial points are inserted by the maximum-distance rule, equivalently at vertices of Voronoi cells. Level 1 adds body-centered points and produces the BCC lattice, whose Voronoi cell is a truncated octahedron of volume

$K_{P,\lambda}$ 03

Level 2 adds $K_{P,\lambda}$ 04-points, the Voronoi vertices of the truncated octahedron, and yields two Voronoi-cell types with volumes

$K_{P,\lambda}$ 05

Level 3 adds $K_{P,\lambda}$ 06-points along space diagonals and yields three Voronoi-cell types with volumes

$K_{P,\lambda}$ 07

(Mathar, 2013). Via Voronoi–Delaunay duality, this hierarchy yields a dual simplicial refinement of the cubic mesh, so it naturally fits a cube-to-simplex, hence cubosimplicial, interpretation (Mathar, 2013).

6. Algebraic, homotopical, and toric frameworks

Subdivision theory for spline spaces provides an algebraic framework that extends to polyhedral, hence cubosimplicial, meshes. For a full-dimensional simplicial complex $K_{P,\lambda}$ 08, subdividing a maximal cell $K_{P,\lambda}$ 09 by a local refinement $K_{P,\lambda}$ 10 yields a split subdivision $K_{P,\lambda}$ 11 when the smoothness ideals on the interior boundary faces are unchanged. In that case there is a short exact sequence of complexes, and, under the vanishing hypothesis $K_{P,\lambda}$ 12, one obtains

$K_{P,\lambda}$ 13

Moreover, when $K_{P,\lambda}$ 14 and $K_{P,\lambda}$ 15 are free, the module on the refinement is free, and the splitting takes the form

$K_{P,\lambda}$ 16

The authors state that all results “generalize easily to the polyhedral case,” so a cubosimplicial mesh fits this framework directly (Schenck et al., 2016).

At the homotopical level, the canonical map

$K_{P,\lambda}$ 17

from the Kan subdivision of a product of finite simplicial sets to the product of the Kan subdivisions is a simple map, meaning that its geometric realization has contractible point inverses (Fjellbo et al., 2014). The proof uses Barratt nerves, path posets in $K_{P,\lambda}$ 18, and iterated mapping cylinders. This suggests that product-type subdivision comparisons can often be controlled by contractible fibers, a feature that is structurally relevant whenever product cells of the form $K_{P,\lambda}$ 19 are present.

In toric and polyhedral geometry, smooth combinatorial cubes furnish another cube-based setting with subdivision-like slice structure. A smooth combinatorial cube is a polytope whose face poset is in bijection with the face poset of the unit cube, and the paper proves that every such cube has two parallel facets, hence is a prismatoid. Its slices are Minkowski equivalent lower-dimensional smooth cubes, and Minkowski equivalent smooth $K_{P,\lambda}$ 20-dimensional cubes form an IDP pair. As a consequence, every smooth combinatorial cube is IDP, establishing Oda’s conjecture for this class (Curtis, 3 Sep 2025). The proof is recursive in dimension and proceeds by slicing the cube into parallel lower-dimensional cubes.

7. Significance and open directions

Across these settings, cubosimplicial subdivision is characterized by three recurring structural features: product cells such as $K_{P,\lambda}$ 21, local-to-global decomposition formulas, and recursive control by lower-dimensional faces or slices. In cubical local $K_{P,\lambda}$ 22-theory, the short local $K_{P,\lambda}$ 23-polynomial is symmetric and nonnegative for locally quasi-geometric subdivisions, but full nonnegativity of the long local $K_{P,\lambda}$ 24-polynomial is open and is known to fail for general non-locally-quasi-geometric subdivisions; the analogous monotonicity for the long cubical $K_{P,\lambda}$ 25-vector is also open except in low dimensions (Athanasiadis, 2010).

For cubical manifolds, a central open question is what skeleton determines an arbitrary $K_{P,\lambda}$ 26-dimensional cubical manifold not necessarily embeddable in a cube (Rowlands, 2019). For marked order polytopes, the cohomological model $K_{P,\lambda}$ 27 gives a combinatorial computation of the $K_{P,\lambda}$ 28-vector, and the natural next questions concern explicit formulas or recurrences for those $K_{P,\lambda}$ 29-vectors, as well as relations to other regular or barycentric subdivisions and to Ehrhart theory (Melikhova, 18 Jul 2025). For smooth combinatorial cubes, the paper proves the weakest property in the hierarchy it discusses, namely IDP, and a plausible next step is the stronger existence of unimodular triangulations or unimodular covers in higher dimensions (Curtis, 3 Sep 2025).

Taken together, these developments show that cubosimplicial subdivision is not a single construction but a research program linking cubical subdivisions, simplicial subdivisions, products of simplices, formal poset subdivisions, Voronoi–Delaunay refinements, and polyhedral models arising in toric, enumerative, and computational geometry.