---
title: Graver's Maximality Conjecture in Rigidity Matroids
url: https://www.emergentmind.com/topics/graver-s-maximality-conjecture
type: topic
---

# Graver's Maximality Conjecture in Rigidity Matroids

Graver's Maximality Conjecture is a conjecture in rigidity matroid theory concerning abstract \(d\)-rigidity matroids on the edge set of the complete graph \(K_n\), ordered by the weak order. In its standard form, it asks whether, for fixed \(n\) and \(d\), this family has a unique maximal element, and whether that maximal element is the generic \(d\)-dimensional rigidity matroid \(\mathcal{R}_d(K_n)\). The conjecture is proved for \(d=1\) and \(d=2\), false for \(d \geq 4\), and structurally most significant in dimension \(3\), where uniqueness and identification with \(\mathcal{R}_3\) separate into distinct issues. Recent work also recasts the conjecture through matroid duality in terms of symmetric tensor matroids, while a distinct strand of the Graver-complexity literature uses the same label for a maximality statement about Lawrence liftings and Graver complexity [2503.14780, 1911.00205, 1102.4674].

## 1. Conjecture in the language of abstract rigidity matroids

For matroids \(\mathcal{M}_1,\mathcal{M}_2\) on the same ground set \(E\), the weak order is defined by
$$
{\cal M}_1 \preceq {\cal M}_2
\quad \text{iff} \quad
r_{{\cal M}_1}(X) \leq r_{{\cal M}_2}(X)
\text{ for all } X \subseteq E.
$$
Graver's conjecture asks for a unique maximal abstract \(d\)-rigidity matroid with respect to this order, and conjecturally identifies it with \(\mathcal{R}_d(K_n)\) [1911.00205].

An abstract \(d\)-rigidity matroid \(M\) on \(E(K_n)\), for \(n \ge d+1\), is defined by closure axioms. If \(E_1,E_2 \subseteq E(K_n)\), then:

- if \(|V(E_1)\cap V(E_2)| \le d-1\), one requires
  $$
  \cl_M(E_1 \cup E_2)\subseteq K(V(E_1))\cup K(V(E_2));
  $$
- if \(\cl_M(E_i)=K(V(E_i))\) for \(i=1,2\) and \(|V(E_1)\cap V(E_2)| \ge d\), one requires
  $$
  \cl_M(E_1\cup E_2)=K(V(E_1\cup E_2)).
  $$
These are the gluing and separation axioms that abstract the behavior of generic rigidity [2503.14780].

The same family admits more operational characterizations. A matroid \(M\) is an abstract \(d\)-rigidity matroid if and only if it satisfies equivalent rank, circuit, cocircuit, and extension properties:
\[
\rank M = dn-\binom{d+1}{2},
\]
every copy of \(K_{d+2}\) is a circuit, every copy of \(K_{1,n-d}\) is a cocircuit, and if \(G \subseteq K_n\), \(v\) has degree \(d\), and \(G-v\) is independent, then \(G\) is independent [2503.14780]. In the special case \(d=3\), Nguyen's characterization used in subsequent work identifies abstract \(3\)-rigidity matroids with \(K_5\)-matroids of rank \(3n-6\) [1911.00205].

## 2. Dual formulation via symmetric tensor matroids

A major recent reformulation uses matroid duality. Writing \(t=n-d-1\), an abstract symmetric \(t\)-tensor matroid on \(E(K_n)\), for \(n \ge t+1\), is defined by the dual conditions
\[
\rank M = \binom{t+1}{2},
\]
every copy of \(K_{1,t+1}\) is a circuit, and every \(K_{n-t+1}\) is a cocircuit [2503.14780].

The key lemma is exact dual equivalence: a matroid \(M\) is an abstract symmetric \(t\)-tensor matroid if and only if its dual is an abstract \(d\)-rigidity matroid for \(d=n-t-1\). In particular, the geometric symmetric \(t\)-tensor matroid belongs to this family. Theorem-level duality results highlighted by this literature state that, for all \(n,d\), the generic rigidity matroid \(\mathcal{R}_d(K_n)\) is dual to the symmetric \(t\)-tensor matroid \(\mathcal{S}_{n-d-1}(K_n)\) [2503.14780].

This reformulation produces a dual version of Graver's conjecture: among all abstract symmetric \(t\)-tensor matroids on \(K_n\), there should be a unique maximal matroid. Because uniqueness on one side implies uniqueness on the other by duality, the rigidity and symmetric-tensor versions are equivalent as maximality statements. This dual viewpoint also connects the conjecture to matrix and tensor completion matroids, rather than only to rigidity-theoretic closure phenomena [2503.14780].

## 3. The three-dimensional case and Whiteley's cofactor matroid

The case \(d=3\) is distinguished by the absence of a purely combinatorial characterization of \(\mathcal{R}_3\), and by Whiteley's analogous conjecture from spline theory. Whiteley observed a close similarity between the generic \(d\)-dimensional rigidity matroid and the generic \(C_{d-2}^{d-1}\)-cofactor matroid, and conjectured that the latter is the unique maximal abstract \(d\)-rigidity matroid for all \(d \ge 2\). For \(d=3\), this becomes the statement that the generic \(C_2^1\)-cofactor matroid \(\mathcal{C}_2^1(V)\) is the unique maximal abstract \(3\)-rigidity matroid on \(K(V)\) [1911.00205].

That \(d=3\) case is proved. The main theorem establishes that \(\mathcal{C}_2^1(V)\) is the unique maximal abstract \(3\)-rigidity matroid on \(K(V)\), and in a stronger form that it is the unique maximal \(K_5\)-matroid of rank \(3n-6\) on \(K_n\). This proves the first part of Graver's conjecture for \(d=3\): uniqueness of the maximal abstract \(3\)-rigidity matroid. It does not resolve whether
\[
{\cal R}_3 = {\cal C}_2^1,
\]
so the identification of the maximal matroid with generic \(3\)-dimensional rigidity remains open [1911.00205].

The proof is inductive and graph-constructive. Bases of the generic cofactor matroid are generated from \(K_4\) using \(0\)-extension, \(1\)-extension, \(X\)-replacement, and double \(V\)-replacement, and the technically decisive step is that double \(V\)-replacement preserves \(C_2^1\)-independence. Whiteley had identified this as the final unresolved step for the \(d=3\) cofactor setting. The result therefore closes the cofactor side of the maximality problem in dimension \(3\), while leaving the rigidity-side equality question open [1911.00205].

## 4. Positive results for small tensor order and for \(n \le d+6\)

A broader affirmative range is now known through the symmetric-tensor formulation. For \(1 \le t \le 5\) and \(n \ge t+1\), the symmetric tensor matroid \(\mathcal{S}_t(K_n)\) is the unique maximal abstract symmetric \(t\)-tensor matroid on \(K_n\). By duality, this is equivalent to the statement that, for all \(n \le d+6\), the rigidity matroid \(\mathcal{R}_d(K_n)\) is the unique maximal abstract \(d\)-rigidity matroid [2503.14780].

The proof proceeds by a new circuit-based sparsity criterion. Independence is tested by a count condition determined by a family \({\cal C}_{n,t}\) of graphs, refined by isomorphism class rather than just cardinality. For \(t \le 5\), a graph is independent in \(\mathcal{S}_t\) if and only if it is count independent with respect to this family, and all maximally independent sets in any other abstract symmetric tensor matroid coincide with those of \(\mathcal{S}_t\). Because this criterion is matroidal, it yields uniqueness rather than only extremality [2503.14780].

This result has two distinct consequences. First, it confirms Graver-type unique maximality in a substantial finite-codimension range, namely \(n-d \le 6\). Second, through duality, it provides a new noncomputational route to previously computer-assisted rigidity results in that range. A plausible implication is that the obstruction to further extension lies not in the basic dual framework, but in the breakdown of the count characterization beyond small \(t\).

## 5. Failure modes, boundaries, and related families

The conjecture does not hold uniformly. It is proved for \(d=1\) and \(d=2\), but for \(d \ge 4\) it is false, with non-uniqueness established by Thurston and Whiteley [1911.00205]. In the symmetric-tensor language, the \(t \le 5\) theorem marks a positive zone, while for \(t \ge 6\) the count characterization can fail, and with it uniqueness [2503.14780].

A related misconception is that uniqueness is robust under modest enlargement of the matroid family. The recent literature shows the opposite. If one weakens the axioms and considers \(K_{1,t+1}\)-matroids, meaning matroids on \(K_n\) in which every copy of \(K_{1,t+1}\) is a circuit, then uniqueness survives only for \(t \le 3\). For \(t \ge 4\) and sufficiently large \(n\), there are at least two distinct maximal \(K_{1,t+1}\)-matroids on \(K_n\). The same phenomenon passes to the family of second quasi symmetric powers of the uniform matroid \(U_{t,n}\): they do not have a unique maximal element for \(t \ge 4\) and sufficiently large \(n\) [2503.14780].

| Setting | Maximality status | Source |
|---|---|---|
| Abstract \(d\)-rigidity, \(d=1,2\) | Proved | [1911.00205] |
| Abstract \(3\)-rigidity | Unique maximal matroid exists; \(\mathcal{C}_2^1\) is that maximum | [1911.00205] |
| Identification \(\mathcal{R}_3=\mathcal{C}_2^1\) | Open | [1911.00205] |
| Abstract \(d\)-rigidity, \(n \le d+6\) | \(\mathcal{R}_d(K_n)\) uniquely maximal | [2503.14780] |
| Abstract \(d\)-rigidity, \(d \ge 4\) | False in general | [1911.00205] |
| \(K_{1,t+1}\)-matroids, \(t \ge 4\) | No unique maximal element for large \(n\) | [2503.14780] |

The boundary results matter because they show that the exact choice of family is essential. Positive results for abstract rigidity matroids and abstract symmetric tensor matroids do not automatically extend to larger circuit-defined families.

## 6. A distinct Graver-complexity usage of “maximality”

In the Graver-complexity literature, the name “Graver's Maximality” is also attached to a conjecture about Lawrence liftings and complete bipartite incidence matrices. In this setting, for the incidence matrix \(A_{t,r}\) of \(K_{t,r}\), the conjecture is that for any positive integers \(t,r\), the sequence
\[
g(A_{t,r}^{(h)})
\]
is non-decreasing in \(h\) and is maximized by the Lawrence lifting \(A_{t,r}\) [1102.4674].

Kudo and Takemura give an explicit exponential lower bound for the Graver complexity of \(A_{t,r}\). For \(4 \le t \le r\),
\[
g(A_{t,r}) \ge (t-1)^{r-t}\left( b_t + \frac{1}{t-2} \right) - \frac{1}{t-2},
\]
where
\[
b_t = (t-2)! \left( 15 + \sum_{i=1}^{t-4} \frac{i+4}{(i+2)!} \right).
\]
For fixed \(t \ge 4\), this yields
\[
g(A_{t,r}) = \Omega((t-1)^r)
\quad \text{as } r \to \infty.
\]
It generalizes the Berstein–Onn bound for \(t=3\),
\[
g(A_{3,r}) \ge 17\cdot 2^{r-3}-7,
\]
and is presented as evidence supporting, though not proving, the maximality conjecture in that setting [1102.4674].

A related recursive construction for \(M\)-fold matrices strengthens the lower bound for \(A_{3\times M}\) to
\[
g(A_{3\times M}) \ge 24\cdot 2^{M-3}-21, \qquad M \ge 4,
\]
and more generally gives
\[
g(A^{(M)}) \ge \frac{g-1}{g-2}(g-1)^{M-(g-1)}-\frac{1}{g-2}.
\]
The construction turns primitive relations on Graver basis elements of \(A^{(M)}\) into primitive relations for \(A^{(M+1)}\), and the paper explicitly states that the bound for \(g(A_{3\times M})\) is not tight [1311.3853]. This suggests that, in the complexity-theoretic usage of the term, maximality questions are accompanied by genuinely explosive combinatorics.

## 7. Current obstacles: implied nonedges, nucleations, and algorithmic relevance

The current \(3\)-dimensional frontier is shaped by the conjectural isomorphism
\[
\mathcal{R}_3 \cong \mathcal{M}_{K_5},
\]
where \(\mathcal{M}_{K_5}\) is Whiteley's cofactor matroid, described as the unique maximal matroid over the edge set of \(K_n\) in which all graphs isomorphic to \(K_5\) are circuits [2508.12417]. Recent work isolates a class of graphs that appear to be the main obstruction to either proving or refuting this identification.

Two notions are central. An implied nonedge is a nonedge \(f\) of a graph \(G\) such that \(G \cup \{f\}\) contains a circuit including \(f\). A nucleation is a rigid subgraph on at least \(5\) vertices, and a graph is nucleation-free if it has no such subgraph. The 2025 constructions produce independent graphs with implied nonedges but no non-trivial rigid subgraphs, and some of the constructions apply to \(3\)-dimensional abstract rigidity matroids as well [2508.12417].

One explicit family is given by rings of butterflies: for \(m \ge 7\) butterflies, the ring is independent and nucleation-free, but the hinges are implied nonedges. The paper emphasizes that none of these examples refutes the maximality conjecture, because the relevant properties hold in both \(\mathcal{R}_3\) and \(\mathcal{M}_{K_5}\). Their significance lies elsewhere. They identify the “especially intractable” flexible circuits that obstruct a simple combinatorial characterization and complicate any polynomial-time independence algorithm for the maximal matroid. The paper also singles out a potential counterexample pattern: a graph \(G\) with a nonedge \(f\) that is implied in \(\mathcal{R}_3\) but is not contained in any \(K_5\) in the closure \(CL_{\mathcal{R}_3}(G)\) [2508.12417].

Taken together, these developments place Graver's Maximality Conjecture at the intersection of rigidity theory, matroid duality, spline/cofactor methods, and Graver-complexity growth. The conjecture is no longer a single undifferentiated assertion: uniqueness, explicit identification of the maximal matroid, robustness under family enlargement, and algorithmic recognizability have emerged as separate but tightly coupled problems.

Source: https://www.emergentmind.com/topics/graver-s-maximality-conjecture