Simplicial Virtual Resolutions

• Simplicial Virtual Resolutions are graded free S-complexes built from labeled simplicial complexes that encode both combinatorial and homological data to resolve monomial and Stanley–Reisner ideals.
• They enable the construction of shorter, computationally efficient complexes even when classical Cohen–Macaulay conditions fail, permitting controlled excess homology on the irrelevant locus.
• These resolutions bridge combinatorial techniques with toric and projective geometry, introducing novel operations like virtual shelling and stellar subdivisions to optimize resolution structures.

A simplicial virtual resolution is a refinement of classical combinatorial commutative algebra methods for resolving monomial and, specifically, Stanley–Reisner ideals over polynomial and Cox rings, central to the study of toric and projective geometry. It allows the construction of free resolutions of such ideals in the context where full Cohen–Macaulayness is absent, by encoding torsion and auxiliary geometric information at the sheafified level, and links the combinatorics of simplicial complexes with homological properties of their associated algebraic objects. Developed principally by Berkesch, Erman, Smith, and collaborators, as well as subsequent work by Stucky, Yang, Van Tuyl, and others, the formalism gives new criteria and construction techniques for short, computationally efficient complexes that are no longer strictly minimal free resolutions, permitting controlled excess homology that vanishes on the irrelevant locus. These tools yield a notion of virtual Cohen–Macaulayness, new classes of combinatorially tractable complexes, and a suite of operations (such as stellar subdivision) to manipulate and improve resolutions in this framework (Yang et al., 2023, Stucky et al., 26 Jan 2026, Berkesch et al., 2020, Kenshur et al., 2020).

1. Definition and Structure of Simplicial Virtual Resolutions

Let $S=k[x_1,\dots,x_n]$ denote a polynomial ring or a Cox ring associated to a smooth toric variety $X$, graded by $\mathbb{Z}^n$ or the class group. For a finitely generated graded $S$-module $M$, a (graded) free $S$-complex

$F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$

is called a virtual resolution of $M$ if, upon sheafification over $X$, the complex

$$F_\bullet \otimes_S \mathcal{O}_X\colon \cdots \to \widetilde{F}_2 \to \widetilde{F}_1 \to \widetilde{F}_0 \to \widetilde{M} \to 0$$

is a locally free (vector bundle) resolution of the corresponding coherent sheaf $X$0 on $X$1 (Yang et al., 2023).

More concretely, for a monomial ideal $X$2, the Stanley–Reisner ideal of a simplicial complex $X$3, the complex $X$4 arises from a (labelled) simplicial complex $X$5, where $X$6 is an lcm-labeling compatible with the face lattice. The length $X$7 satisfies $X$8, and in the optimal situation is exactly $X$9. The modules $\mathbb{Z}^n$0 are direct sums of shifts $\mathbb{Z}^n$1 determined by the combinatorial types of faces (Yang et al., 2023, Stucky et al., 26 Jan 2026).

2. Combinatorial and Homological Construction

Given a labeled simplicial complex $\mathbb{Z}^n$2, one forms $\mathbb{Z}^n$3, a $\mathbb{Z}^n$4-graded free $\mathbb{Z}^n$5-complex,

$\mathbb{Z}^n$6

with boundary operators defined on oriented simplices by

$\mathbb{Z}^n$7

For a monomial $\mathbb{Z}^n$8, define the induced subcomplex $\mathbb{Z}^n$9. The key connection is: $S$0 Thus, the vanishing of reduced homology (in appropriate subcomplexes) is equivalent to the required exactness conditions (Stucky et al., 26 Jan 2026).

A fundamental characterization for when $S$1 is a virtual resolution of $S$2 is: there exists $S$3 such that for all $S$4 (where $S$5 is the irrelevant ideal), $S$6 for all $S$7. This enables precise combinatorial criteria for verifying and constructing virtual resolutions (Stucky et al., 26 Jan 2026).

3. Virtual Cohen–Macaulayness and Sufficient Criteria

For a simplicial complex $S$8 on the rays of $S$9, $M$0 is called virtually Cohen–Macaulay if $M$1 admits a virtual resolution of length $M$2.

The following theorem gives a constructive sufficient criterion (Yang et al., 2023): If there exists a simplicial complex $M$3 and a surjective simplicial map $M$4 such that

1. $M$5 is Cohen–Macaulay,
2. For every face $M$6, $M$7,
3. Every $M$8 with $M$9 lies in the irrelevant complex $S$0,

then $S$1 is virtually Cohen–Macaulay. This allows for "duplication" of faces lying entirely in the irrelevant locus, augmenting $S$2 to a Cohen–Macaulay cover $S$3, without altering the geometric support.

Further, a more specialized notion—virtually shellable simplicial complexes—vividly generalizes classical shellability and inherits virtual Cohen–Macaulayness whenever a suitable virtual shelling (ordered CM cover, duplicating only irrelevant faces) exists (Yang et al., 2023).

Balanced simplicial complexes—each facet including exactly one variable from each multigrading component—are always virtually Cohen–Macaulay (Kenshur et al., 2020). A radical-of-monomial extension criterion further generalizes this to settings where an auxiliary monomial ideal $S$4 can be found so that $S$5 is Cohen–Macaulay (Kenshur et al., 2020).

4. Minimality, Nontrivial Homology, and Homology-Reduction Techniques

Although virtual resolutions permit controlled homology off the irrelevant locus, it is crucial to characterize the minimal supporting structure for such homology. If $S$6 has $S$7 for some $S$8, then any subcomplex $S$9 exhibiting nontrivial homology must be large enough: specifically, the total number of vertices satisfies

$F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$0

In the minimal case where $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$1, $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$2 must be a bipyramid over a simplex of dimension $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$3 (Stucky et al., 26 Jan 2026).

Eliminating unwanted homology is achieved via virtual-compatible stellar subdivisions. A new vertex is added at a chosen face, with label $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$4 compatible with the divisibility and ideal-theoretic saturation relative to the irrelevant ideal. Under such subdivisions, one produces a new complex $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$5 that supports a virtual resolution and in general has strictly less homology, with precise criteria dictating when homology is strictly reduced (Stucky et al., 26 Jan 2026).

5. Comparison with Classical Resolutions and Computational Aspects

Classical simplicial (cellular or Taylor) resolutions are genuine free resolutions, requiring exactness off the defining ideal. Simplicial virtual resolutions relax this, allowing extraneous homology supported entirely on the irrelevant locus. This relaxation can strictly shorten the required length: a typical non-CM complex may admit a virtual resolution of length equal to codimension, while any genuine resolution is longer (Yang et al., 2023).

A combinatorial comparison is illustrated in the following table:

Resolution Type	Length	Homology Location
Minimal Free Resolution	$F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$6 codimension	Vanishes off ideal
Virtual Resolution	codimension	Permitted on irrelevant locus

The combinatorial sufficient tests for virtual resolutions avoid the full strength of e.g., Reisner's criterion. Rather than requiring vanishing of all reduced link homologies, it suffices to produce a Cohen–Macaulay or shellable cover duplicating faces only on the irrelevant subcomplex.

Constructing virtual resolutions is done explicitly:

• by finding a monomial ideal $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$7 with $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$8, and using its minimal free resolution,
• by splitting vertices (duplicating) in $F_\bullet\colon \cdots \to F_2 \to F_1 \to F_0$9 along faces in $M$0 to generate a Cohen–Macaulay cover.

These are computationally tractable, and, in practice, often yield simpler complexes than direct minimal resolutions.

6. Applications, Examples, and Generalizations

Key examples illustrate the construction:

• In $M$1, for a $M$2 not Cohen–Macaulay as in $M$3, a shellable complex $M$4 is formed by splitting a variable, and yields a virtual resolution of minimal length (Yang et al., 2023).
• In $M$5 (the 4-cycle), $M$6 is not CM, but $M$7 makes $M$8 a hypersurface, so the induced $M$9 provides a virtual resolution of minimal length (Kenshur et al., 2020).
• In products of projective spaces, many Stanley–Reisner rings, especially from balanced complexes, are virtually Cohen–Macaulay even in the absence of classical CM properties (Kenshur et al., 2020).

Homological tools, such as the vanishing of sheafified Ext and Tor beyond the length of a virtual resolution, mapping cone constructions, and virtually regular elements (for dimension control), enable further algebraic manipulation and deeper understanding of module-theoretic invariants (Berkesch et al., 2020).

7. Relationships with Cohen–Macaulayness and Open Questions

Three related notions are emphasized:

• Arithmetically Cohen–Macaulay: a module has a free resolution of minimal length (codimension).
• Virtually Cohen–Macaulay: a virtual resolution of minimal length exists.
• Geometrically Cohen–Macaulay: the sheaf is locally Cohen–Macaulay at all relevant primes.

There is a strictly descending implication chain: $X$0 with known strictness—in particular, virtual CM captures modules that are not classical CM but admit virtual resolutions of optimal length (Berkesch et al., 2020, Kenshur et al., 2020).

Open questions concern the dependence of virtual Betti numbers on combinatorial type, the existence of combinatorial criteria analogous to Reisner's for general virtual CMness, and stability under fundamental operations (joins, links, subdivisions) (Kenshur et al., 2020). A plausible implication is that virtual shellability provides a combinatorial framework parallel to, but more flexible than, classical shellability in guaranteeing Cohen–Macaulay-type homological properties (Yang et al., 2023).