Disproof of the tree product conjecture via the Heisenberg group

Abstract: Product structure theory aims to understand complex graphs by embedding them into products of simpler graphs. In this direction, Campbell, Distel, Gollin, Harvey, Hendrey, Hickingbotham, Mohar and Wood (2022) put forth the conjecture that all graphs of degree-$d$ polynomial growth (i.e., where balls of radius $r$ have $\mathcal{O}(r<sup>d)$ vertices) can be embedded into the strong product of $d$ trees, each with linear growth, and a constant-size clique. In this paper, we disprove this conjecture for $d = 4$. The counterexamples are finite subgraphs of a Cayley graph of the discrete $3$-dimensional Heisenberg group $\mathbb{H}(\mathbb{Z})$. These graphs were first proposed by Huang and McCarty as potential counterexamples to the conjecture. A key technical tool of our proof is the ''quantitative central collapse'' theorem due to Cheeger, Kleiner and Naor (2011), guaranteeing that every Lipschitz map from the continuous Heisenberg group $\mathbb{H}$ to the function space $L_1$ collapses along a central line.

• The paper presents a rigorous disproof of the tree product conjecture by constructing a counterexample using the Cayley graph of the Heisenberg group, demonstrating embedding obstructions for polynomial-growth graphs.
• It leverages combinatorial properties and Lipschitz embedding arguments, including quantitative collapse, to reveal intrinsic limitations in decomposing nilpotent group graphs as products of trees and cliques.
• The results significantly impact graph structure theory by challenging existing decomposition methods and prompting the search for alternative frameworks to handle complex polynomial-growth structures.

Disproof of the Tree Product Conjecture Using the Heisenberg Group

Introduction and Context

The paper "Disproof of the tree product conjecture via the Heisenberg group" (2607.03041) presents a rigorous disproof of a central conjecture in the product structure theory of graphs. This conjecture, posed by Campbell, Distel, Gollin, Harvey, Hendrey, Hickingbotham, Mohar, and Wood (2022), postulated that any finite graph with degree-$d$ polynomial growth can be embedded into a strong product of $d$ trees of linear growth and a constant-size clique. The authors construct a counterexample for $d = 4$ by leveraging the metric and combinatorial properties of Cayley graphs of the discrete 3-dimensional Heisenberg group $\mathbb{H}(\mathbb{Z})$.

This contribution settles a widely circulated question on the optimal deconstruction of large-scale graph classes with polynomial growth and reveals intrinsic obstructions to embedding such graphs as strong products of simple structures. The result connects the theory of graph products with geometric group theory and functional analysis, utilizing subtle arguments from the Lipschitz embedding literature.

Theoretical Framework and Main Result

Product structure theory seeks to represent graphs with complex structure as subgraphs of products of well-understood, low-complexity graphs, typically trees or graphs of bounded treewidth. While the planarity-based product structure theorem and its ramifications provide strong tools for minor-closed classes, the extension of similar representations to polynomial-growth graphs has been open.

Let $G$ be a graph with balls of radius $r$ containing $\mathcal{O}(r^d)$ vertices. The conjecture precisely stated that every such graph is isomorphic to a subgraph of $T_1 \boxtimes ... \boxtimes T_d \boxtimes K_C$, with each $T_i$ a linear-growth tree ($\mathcal{O}(r)$ vertices in balls of radius $d$0) and $d$1 denoting a clique of constant size.

The central theorem of the paper asserts the following: There exists a Cayley graph of $d$2 with growth at most $d$3 such that, for any choice of four finite or infinite trees of linear growth and any constant $d$4, no subgraph of their strong product with a clique of size $d$5 is isomorphic to this Cayley graph or even any of its (sufficiently large) finite subgraphs.

Techniques and Key Arguments

Analysis of the Heisenberg Group

The authors analyze the discrete Heisenberg group $d$6 and its Cayley graph constructed with standard generators. This group is nilpotent of step two and is a canonical example of an object with strictly polynomial growth, realized as $d$7 for metric balls of radius $d$8 in the associated word metric.

The combinatorial structure of the Cayley graph is carefully described, and its growth properties are established explicitly. The authors exploit the non-commutative nature of the group and the presence of a non-trivial center, which is crucial for the embedding obstructions.

Connection to Lipschitz Embeddings and $d$9

The analytical core of the disproof relies on the "quantitative central collapse" theorem of Cheeger, Kleiner, and Naor (2011), which asserts that any Lipschitz map from the continuous Heisenberg group to $d = 4$0 collapses distances along central directions. The authors discretize this phenomenon and show that any potential $d = 4$1 embedding of large subgraphs (even after normalization and extension) must dramatically collapse a positive measure subset along the center.

Obstruction to Tree Product Decompositions

Infinite trees (and their finite subtrees) admit isometric embeddings into $d = 4$2, and so do strong products thereof. If the conjecture were true, the Cayley graph could be embedded into such a strong product, and consequently, there would exist a Lipschitz $d = 4$3 embedding with well-controlled distortion. The quantitative collapse theorem, however, shows that in the Heisenberg group, large sets of vertices in the same central coset must have pairwise $d = 4$4 distances arbitrarily small compared to their combinatorial distances, which is impossible for products of trees with linear growth.

A counting argument, grounded in the group structure and the behavior of coset representatives, provides an explicit contradiction—bounding below the number of distinct central elements and the size of their orbits under certain product operations—incompatible with the hypothetical embedding.

Compactness Argument for Finite Subgraphs

The authors extend their result from infinite graphs to finite subgraphs using a compactness argument. They show that if every finite subgraph could be so embedded, then a countable union argument plus diagonalization would yield a global embedding of the full Cayley graph, contradicting the main theorem. Therefore, there exist large finite subgraphs of $d = 4$5 witnessing the failure of the conjecture.

Numerical and Structural Highlights

• Growth Estimate: The constructed Cayley graph has explicit polynomial growth with balls of radius $d = 4$6 containing at most $d = 4$7 vertices.
• Collapse Bound: For any $d = 4$8-Lipschitz $d = 4$9, for arbitrarily large $\mathbb{H}(\mathbb{Z})$0, there exist central cosets of size $\mathbb{H}(\mathbb{Z})$1 where all pairwise $\mathbb{H}(\mathbb{Z})$2 distances are $\mathbb{H}(\mathbb{Z})$3—a sharp deviation from tree-product behavior.
• Obstruction Holds Uniformly: The non-embeddability argument applies for any fixed linear-growth bound (uniformly in the parameter $\mathbb{H}(\mathbb{Z})$4 of the clique), showing the obstruction is robust and not sensitive to the auxiliary clique factor.

Implications and Future Directions

This disproof fundamentally limits any attempt to decompose graphs of polynomial growth (even $\mathbb{H}(\mathbb{Z})$5) into low-complexity strong products of trees and bounded cliques. It shows that the combinatorial and geometric complexity of even highly structured nilpotent groups escapes such decompositions, indicating the need for more nuanced invariants and more flexible product structures—potentially involving other basic graphs or additional parameters.

The result emphasizes the crucial role played by Lipschitz extension and embedding obstructions, and strongly ties graph product structure theory, geometric group theory, and the theory of metric embeddings. It signals that similar negative results may hold for higher-degree polynomial growth and potentially for a wider range of algebraic structures.

Practically, this forces a reevaluation of algorithms and meta-theorems (that depend on tree products) for large-scale graphs in metric geometry, group theory, and combinatorial optimization, as their applicability to nilpotent structures is now definitively limited.

Future directions include:

• Investigating alternative decomposition theorems involving more general factors.
• Developing new invariants in structural graph theory distinguishing the Heisenberg-type growth from product-embeddable structures.
• Understanding to what extent randomized or approximate decompositions can bypass the collapse phenomenon.
• Analyzing whether related obstructions exist for other classes of groups or metric spaces, and clarifying the boundary between embeddable and non-embeddable polynomial-growth graphs.

Conclusion

The paper delivers a definitive negative solution to the tree product conjecture for polynomial-growth graphs, providing a rigorous construction and analytic framework showing non-embeddability of certain Cayley graphs into products of trees and bounded cliques. This advances the understanding of the structure and limitations of product-theoretic approaches in graph theory and indicates new directions for the decomposition of large-scale finite and infinite graphs beyond tree products.