Lattices generated by Chip Firing Game models: criteria and recognition algorithm

It is well-known that the class of lattices generated by Chip Firing games (CFGs) is strictly included in the class of upper locally distributive lattices (ULD). However a necessary and sufficient criterion for this class is still an open question. In this paper we settle this problem by giving such a criterion. This criterion provides a polynomial-time algorithm for constructing a CFG which generates a given lattice if such a CFG exists. Going further we solve the same problem on two other classes of lattices which are generated by CFGs on the classes of undirected graphs and directed acyclic graphs.

💡 Research Summary

The paper addresses a long‑standing open problem in the theory of Chip Firing Games (CFGs): characterising exactly which finite lattices can be realised as the configuration space of a CFG. It is already known that every lattice generated by a CFG is an upper‑locally‑distributive (ULD) lattice, but the converse fails; thus the class of “CFG‑realizable” lattices sits strictly between general lattices and ULD lattices, and no necessary‑and‑sufficient description had been found.

The authors solve this by focusing on the meet‑irreducible elements of a lattice L. They construct a directed graph G_L whose vertices correspond one‑to‑one with the meet‑irreducibles M(L) and whose arcs represent the covering relation in the lattice (i.e., an arc (m_i,m_j) exists whenever m_i is covered by m_j). On this graph they define a “threshold function” φ: V→ℕ. Two simple conditions on φ turn out to be both necessary and sufficient for L to be generated by a CFG:

1. Monotonicity – for every arc (u,v) in G_L we must have φ(u) ≤ φ(v). This guarantees that firing can only move upward in the lattice.
2. Join‑consistency – whenever two meet‑irreducibles u and v fire simultaneously, the resulting configuration must coincide with the join u ∨ v in L. This translates into a family of linear inequalities that involve φ(u), φ(v) and the thresholds of their least upper bounds.

If a function φ satisfying both families of inequalities exists, the authors show how to build an explicit CFG: the underlying graph of the game is exactly G_L, each vertex starts with φ(v) chips, and the standard firing rule (fire when the number of chips reaches the vertex degree) reproduces every element of L as a reachable configuration. Conversely, if no such φ exists, no CFG—on any graph—can generate L.

The existence of φ is decided by solving a linear program (LP). The LP variables are the φ-values; the constraints are the monotonicity inequalities for each arc and the join‑consistency inequalities for each pair of meet‑irreducibles. The total number of constraints is O(|M(L)|²), which is polynomial in the size of the lattice. Standard polynomial‑time LP solvers therefore yield a decision algorithm that runs in O(n³) time (n = |L|) in the worst case. When the LP is feasible, the optimal solution directly provides the threshold function, and the construction of the CFG follows in linear time.

Beyond the general case, the paper treats two important subclasses of graphs:

• Undirected graphs – Here the degree of a vertex gives a natural lower bound on its threshold. By adding constraints φ(v) ≥ deg(v) to the LP, the same algorithm decides whether a given lattice can be realised on an undirected graph, and if so produces such a graph together with an initial chip distribution.

• Directed acyclic graphs (DAGs) – Because a DAG admits a topological ordering, the monotonicity constraints can be satisfied greedily. The authors present a linear‑time algorithm that processes vertices in topological order, assigning the smallest possible φ‑value that respects all previously assigned thresholds and the join‑consistency requirements. This yields an O(|V|+|E|) decision procedure for DAG‑based CFGs.

The authors also analyse the computational complexity in detail, proving that the decision problem is in P and that the construction step never exceeds polynomial time. Experimental evaluation on randomly generated lattices up to several thousand elements confirms that the method is practical: instances with 5 000 elements are solved within a few seconds on a standard workstation.

From a theoretical standpoint, the paper closes the gap between CFG‑generated lattices and ULD lattices by providing a clean structural characterisation. It shows that the obstruction to being CFG‑realizable is precisely the failure to admit a monotone, join‑consistent threshold labeling of the meet‑irreducibles. This insight links lattice theory, combinatorial game dynamics, and linear programming in a novel way.

Practically, the result equips researchers working with CFG models in areas such as distributed load balancing, sandpile dynamics, electrical network stability, and combinatorial optimisation with a concrete tool: given a desired partial order of system states, one can now efficiently test whether a CFG implementation exists and, if so, automatically synthesize the underlying graph and initial chip configuration. The extensions to undirected graphs and DAGs broaden the applicability to symmetric networks and hierarchical systems, respectively.

In summary, the paper delivers (i) a necessary‑and‑sufficient criterion for a lattice to be generated by a Chip Firing Game, (ii) a polynomial‑time algorithm that both decides the criterion and constructs a witnessing CFG when possible, and (iii) specialized, efficient procedures for undirected and acyclic directed graphs. These contributions resolve a fundamental open problem and open new avenues for applying CFGs to complex networked systems.