Partitionable and interval-decomposable complexes
DefinitionRank-\(k\) interval decomposition with a subcomplex of bottoms#✓fully proved
For a natural number \(k\), a finite family \(K\) of finite subsets of \(\mathbb{N}\) has a rank-\(k\) interval decomposition with a subcomplex of bottoms if there are a finite family \(B\) of finite subsets of \(\mathbb{N}\), closed under taking subsets, and an assignment of a finite set \(\mathrm{top}(F) \subseteq \mathbb{N}\) to each finite set \(F \subseteq \mathbb{N}\), such that every \(F \in B\) satisfies \(F \subseteq \mathrm{top}(F)\), \(|\mathrm{top}(F)| = |F| + k\) and \(\mathrm{top}(F) \in K\), and every \(\sigma \in K\) lies in the interval \([F, \mathrm{top}(F)] = \{G : F \subseteq G \subseteq \mathrm{top}(F)\}\) for exactly one \(F \in B\).
DefinitionStanley's interval-decomposition conjecture for complexes with acyclic links#✓fully proved
The statement that for every integer \(k \ge 1\) and every finite simplicial complex \(K\) on the vertex set \(\mathbb{N}\) (a finite family of finite subsets of \(\mathbb{N}\) closed under taking subsets): if the link \(\mathrm{lk}_K(F) = \{G \in K : G \cap F = \emptyset,\ G \cup F \in K\}\) of every face \(F \in K\) with fewer than \(k\) vertices (including \(\mathrm{lk}_K(\emptyset) = K\)) is acyclic over \(\mathbb{Q}\) (all of its reduced rational homology vanishes), then \(K\) has a rank-\(k\) interval decomposition with a subcomplex of bottoms: there are a family \(B\) of finite subsets of \(\mathbb{N}\), closed under taking subsets, and faces \(\mathrm{top}(F) \in K\) for \(F \in B\) with \(F \subseteq \mathrm{top}(F)\) and \(|\mathrm{top}(F)| = |F| + k\), such that every face of \(K\) lies in the interval \([F, \mathrm{top}(F)]\) for exactly one \(F \in B\).
This is Stanley's Problem 5 (Positivity Problems; Conjecture 2.4 of Stanley, Discrete Math. 120 (1993)) with \(k = j + 2\): the faces with fewer than \(k\) vertices are the faces of dimension at most \(j\), and for an interval \([F, F']\) with \(F' = \mathrm{top}(F)\), the condition \(|F'| = |F| + k\) means \(\dim F' - \dim F = k\).
def acyclicLinksIntervalDecomposition : Prop :=
∀ k : ℕ, 1 ≤ k → ∀ K : Finset (Finset ℕ), IsSimplicialComplex K →
(∀ F ∈ K, F.card < k → IsRationallyAcyclic (link K F)) → HasRankIntervalDecomposition K k
def IsSimplicialComplex (K : Finset (Finset ℕ)) : Prop :=
∀ σ ∈ K, ∀ τ ⊆ σ, τ ∈ K
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:40
def IsRationallyAcyclic (K : Finset (Finset ℕ)) : Prop :=
∀ k, reducedBetti K k = 0
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:64
def link (K : Finset (Finset ℕ)) (F : Finset ℕ) : Finset (Finset ℕ) :=
K.filter fun G => Disjoint G F ∧ G ∪ F ∈ K
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:68
def HasRankIntervalDecomposition (K : Finset (Finset ℕ)) (k : ℕ) : Prop :=
∃ (B : Finset (Finset ℕ)) (top : Finset ℕ → Finset ℕ), (∀ F ∈ B, ∀ G ⊆ F, G ∈ B) ∧
(∀ F ∈ B, F ⊆ top F ∧ (top F).card = F.card + k ∧ top F ∈ K) ∧
∀ σ ∈ K, ∃! F, F ∈ B ∧ F ⊆ σ ∧ σ ⊆ top F
WantedExt/Combinatorics/AcyclicLinksIntervalDecompositionAMR156Problem0004Wanted.lean:46
TheoremStanley's interval-decomposition conjecture for complexes with acyclic links is false#●ready to prove
It is not true that for every integer \(k \ge 1\) and every finite simplicial complex \(K\) on the vertex set \(\mathbb{N}\) (a finite family of finite subsets of \(\mathbb{N}\) closed under taking subsets) such that the link \(\mathrm{lk}_K(F) = \{G \in K : G \cap F = \emptyset,\ G \cup F \in K\}\) of every face \(F \in K\) with fewer than \(k\) vertices is acyclic over \(\mathbb{Q}\) (all of its reduced rational homology vanishes), there are a family \(B\) of finite subsets of \(\mathbb{N}\), closed under taking subsets, and faces \(\mathrm{top}(F) \in K\) for \(F \in B\) with \(F \subseteq \mathrm{top}(F)\) and \(|\mathrm{top}(F)| = |F| + k\), such that every face of \(K\) lies in the interval \([F, \mathrm{top}(F)]\) for exactly one \(F \in B\).
Doolittle and Goeckner constructed a complex with \(k = 2\) (a \(2\)-fold acyclic complex) that cannot be partitioned into intervals \([F, F']\) with \(|F'| = |F| + 2\) at all; the case \(k = 1\) is Stanley's theorem.
public theorem_wanted acyclicLinksIntervalDecomposition_refuted :
¬ acyclicLinksIntervalDecomposition
def acyclicLinksIntervalDecomposition : Prop :=
∀ k : ℕ, 1 ≤ k → ∀ K : Finset (Finset ℕ), IsSimplicialComplex K →
(∀ F ∈ K, F.card < k → IsRationallyAcyclic (link K F)) → HasRankIntervalDecomposition K k
WantedExt/Combinatorics/AcyclicLinksIntervalDecompositionAMR156Problem0004Wanted.lean:55
Sources
- R. P. Stanley, Positivity Problems and Conjectures in Algebraic Combinatorics (1999), Problem 5 (interval decompositions of complexes with acyclic links); resolution: J. Doolittle and B. Goeckner, Resolving Stanley's conjecture on k-fold acyclic complexes, FPSAC 2019 (Sém. Lothar. Combin. 82B, Article 46), arXiv:1811.08518, §3: Theorem 3.2 (gluing N copies of Δ along Γ preserves k-fold acyclicity and non-decomposability for N > ℓ/2^k) with Theorem 3.4 (Φ = (Δ, Γ) is not rank 2 boolean decomposable); "Ω₁₇ is a counterexample to Conjecture 1.3" (Stanley's Conjecture 2.4 = Problem 5 of the 1999 list, k = 2), https://arxiv.org/abs/1811.08518.
Uses 1
DefinitionCohen–Macaulay complexes over \(\mathbb{Q}\)#✓fully proved
A finite family \(K\) of finite subsets of \(\mathbb{N}\) is Cohen–Macaulay over \(\mathbb{Q}\) when for every member \(F\) of \(K\) (the empty face included, if it belongs to \(K\)) the link \(L = \operatorname{lk}_K F = \{G \in K : G \cap F = \emptyset,\ G \cup F \in K\}\) satisfies \(\tilde\beta_i(L) = 0\) for every integer \(i\) with \(-1 \le i < \dim L\). Here \(\dim L\) is one less than the largest number of elements of a member of \(L\), taken to be \(-1\) when \(L\) has no nonempty member, and \(\tilde\beta_i(L)\) is the number of members of \(L\) with \(i + 1\) elements minus the ranks of the augmented rational simplicial boundary maps (vertices ordered as in \(\mathbb{N}\)) out of and into the span of these members; when \(L\) is a simplicial complex, \(\tilde\beta_i(L) = \dim \tilde H_i(L; \mathbb{Q})\).
For a simplicial complex this is Reisner's criterion.
def IsCohenMacaulayOverRat (K : Finset (Finset ℕ)) : Prop :=
∀ F ∈ K, ∀ k, k < topCard (link K F) → reducedBetti (link K F) k = 0
def topCard (K : Finset (Finset ℕ)) : ℕ :=
K.sup Finset.card
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:72
def link (K : Finset (Finset ℕ)) (F : Finset ℕ) : Finset (Finset ℕ) :=
K.filter fun G => Disjoint G F ∧ G ∪ F ∈ K
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:68
noncomputable def reducedBetti (K : Finset (Finset ℕ)) (k : ℕ) : ℤ :=
((facesOfCard K k).card : ℤ) - (if k = 0 then 0 else ((boundaryMatrix K (k - 1)).rank : ℤ)) -
(boundaryMatrix K k).rank
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:58
Uses 3 · Used by 1
DefinitionPure complexes#✓fully proved
A finite family \(K\) of finite subsets of \(\mathbb{N}\) is pure when all its facets (maximal members) have the same number of elements.
def IsPure (K : Finset (Finset ℕ)) : Prop :=
∃ d : ℕ, ∀ F, IsFacet K F → F.card = d
def IsFacet (K : Finset (Finset ℕ)) (F : Finset ℕ) : Prop :=
F ∈ K ∧ ∀ G ∈ K, F ⊆ G → G = F
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:76
Uses 1 · Used by 1
DefinitionPartitionable complexes#✓fully proved
A finite family \(K\) of finite subsets of \(\mathbb{N}\) is partitionable when there is a map \(r\) assigning to each facet \(F\) of \(K\) a set \(r(F) \subseteq F\) such that for every member \(\sigma\) of \(K\) there is exactly one facet \(F\) of \(K\) with \(r(F) \subseteq \sigma \subseteq F\).
For a simplicial complex this says that its faces are partitioned into the intervals \([r(F), F] = \{\sigma : r(F) \subseteq \sigma \subseteq F\}\), one for each facet \(F\).
def IsPartitionable (K : Finset (Finset ℕ)) : Prop :=
∃ r : Finset ℕ → Finset ℕ, (∀ F, IsFacet K F → r F ⊆ F) ∧
∀ σ ∈ K, ∃! F, IsFacet K F ∧ r F ⊆ σ ∧ σ ⊆ F
def IsFacet (K : Finset (Finset ℕ)) (F : Finset ℕ) : Prop :=
F ∈ K ∧ ∀ G ∈ K, F ⊆ G → G = F
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:76
Uses 1 · Used by 1
DefinitionStanley–Garsia partitionability conjecture#✓fully proved
Every finite simplicial complex on the vertex set \(\mathbb{N}\) that is pure and Cohen–Macaulay over \(\mathbb{Q}\) is partitionable.
def cohenMacaulayPartitionable : Prop :=
∀ K : Finset (Finset ℕ), IsSimplicialComplex K → IsPure K → IsCohenMacaulayOverRat K →
IsPartitionable K
def IsSimplicialComplex (K : Finset (Finset ℕ)) : Prop :=
∀ σ ∈ K, ∀ τ ⊆ σ, τ ∈ K
WantedExt/Combinatorics/RationalSimplicialHomologyShared.lean:40
def IsPure (K : Finset (Finset ℕ)) : Prop :=
∃ d : ℕ, ∀ F, IsFacet K F → F.card = d
WantedExt/Combinatorics/CohenMacaulayPartitionabilityAMR156Problem0005Wanted.lean:50
def IsCohenMacaulayOverRat (K : Finset (Finset ℕ)) : Prop :=
∀ F ∈ K, ∀ k, k < topCard (link K F) → reducedBetti (link K F) k = 0
WantedExt/Combinatorics/CohenMacaulayPartitionabilityAMR156Problem0005Wanted.lean:46
def IsPartitionable (K : Finset (Finset ℕ)) : Prop :=
∃ r : Finset ℕ → Finset ℕ, (∀ F, IsFacet K F → r F ⊆ F) ∧
∀ σ ∈ K, ∃! F, IsFacet K F ∧ r F ⊆ σ ∧ σ ⊆ F
WantedExt/Combinatorics/CohenMacaulayPartitionabilityAMR156Problem0005Wanted.lean:55
Sources
- R. P. Stanley, Positivity Problems and Conjectures in Algebraic Combinatorics, Problem 6 (partitionability conjecture of Stanley and Garsia).
Uses 4 · Used by 1
TheoremThe Stanley–Garsia partitionability conjecture is false#●ready to prove
There is a finite simplicial complex on the vertex set \(\mathbb{N}\) that is pure and Cohen–Macaulay over \(\mathbb{Q}\) but not partitionable.
Duval, Goeckner, Klivans and Martin constructed explicit pure \(3\)-dimensional Cohen–Macaulay simplicial complexes that are not partitionable.
public theorem_wanted cohenMacaulayPartitionable_refuted : ¬ cohenMacaulayPartitionable
def cohenMacaulayPartitionable : Prop :=
∀ K : Finset (Finset ℕ), IsSimplicialComplex K → IsPure K → IsCohenMacaulayOverRat K →
IsPartitionable K
WantedExt/Combinatorics/CohenMacaulayPartitionabilityAMR156Problem0005Wanted.lean:61
Sources
- R. P. Stanley, Positivity Problems and Conjectures in Algebraic Combinatorics (1999), Problem 6 (Stanley–Garsia partitionability conjecture); resolution: A. M. Duval, B. Goeckner, C. J. Klivans and J. L. Martin, A non-partitionable Cohen-Macaulay simplicial complex, Adv. Math. 299 (2016), 381-395, arXiv:1504.04279, Theorem 3.5 (arXiv numbering): the complex C₃ obtained by gluing three copies of X along A is Cohen-Macaulay and non-partitionable (Theorem 3.4: likewise C₂₅, by the gluing Theorem 3.1); abstract: "We disprove the conjecture by constructing an explicit counterexample", https://arxiv.org/abs/1504.04279.
Uses 1
Loading dependency explorer…
The dependency explorer needs JavaScript and its graph data.