Dichotomy for Holant* Problems with a Function on Domain Size 3

Holant problems are a general framework to study the algorithmic complexity of counting problems. Both counting constraint satisfaction problems and graph homomorphisms are special cases. All previous results of Holant problems are over the Boolean domain. In this paper, we give the first dichotomy theorem for Holant problems for domain size $>2$. We discover unexpected tractable families of counting problems, by giving new polynomial time algorithms. This paper also initiates holographic reductions in domains of size $>2$. This is our main algorithmic technique, and is used for both tractable families and hardness reductions. The dichotomy theorem is the following: For any complex-valued symmetric function ${\bf F}$ with arity 3 on domain size 3, we give an explicit criterion on ${\bf F}$, such that if ${\bf F}$ satisfies the criterion then the problem ${\rm Holant}^({\bf F})$ is computable in polynomial time, otherwise ${\rm Holant}^({\bf F})$ is #P-hard.

💡 Research Summary

The paper establishes a complete complexity dichotomy for Holant* problems defined on a domain of size three with a single symmetric ternary function F. Holant* is a general counting framework where each vertex of a graph is assigned a constraint function and each edge carries a variable taking values from a finite domain; the total weight of an assignment is the product of the vertex functions, summed over all assignments. While the Boolean (domain‑2) case has been extensively studied—culminating in a series of dichotomy theorems based on holographic reductions and matchgate techniques—no comparable classification existed for larger domains.

The authors focus on complex‑valued symmetric functions of arity three, which can be represented as a 3‑dimensional tensor with 27 entries (reduced to 10 independent parameters by symmetry). They introduce a systematic analysis of the holographic transformation group O(3)·ℂ* (orthogonal 3×3 matrices together with non‑zero complex scalars). Under a transformation M∈O(3)·ℂ*, the tensor of F is mapped to (M⊗M⊗M)·T(F). Crucially, certain tensor invariants—tensor rank, eigenvalue multiplicities, and a set of cubic invariants—are preserved by this group. By exploiting these invariants the authors derive an explicit normal‑form classification for any symmetric ternary function on a three‑element domain.

Two families of normal forms turn out to be tractable:

1. Low‑rank tensors (rank ≤ 2). After an appropriate holographic transformation, F can be expressed as a sum of at most two rank‑one tensors, i.e.,
   T(F) = a·u⊗u⊗u + b·v⊗v⊗v,
   where u and v are linearly independent vectors in ℂ³. This structure is exactly the one that underlies matchgate computations. The authors adapt the classical FKT/Pfaffian algorithm to the three‑valued setting, showing that Holant* (F) can be evaluated in polynomial time for any F of this type.

2. Affine‑type tensors. These are functions whose non‑zero entries depend only on linear relations among the three indices, for example
   T(F){i,j,k} = α·δ{i+j+k≡0 (mod 3)} + β·δ_{i=j=k} + γ·(linear terms).
   After transformation, the problem reduces to counting homomorphisms from the input graph to a fixed three‑vertex weighted graph—a problem known to be solvable in polynomial time via matrix powering or dynamic programming. The authors give a concrete algorithm that evaluates Holant* (F) by interpreting the constraints as affine equations over ℤ₃ and solving them efficiently.

If a given F does not fall into either of these two families, the paper proves #P‑hardness. The hardness proof proceeds by constructing gadgets that simulate Boolean NAND, OR, and weighted negation gates using copies of F together with auxiliary vertices. By carefully arranging the gadgets and applying a sequence of holographic transformations, the authors ensure that the resulting Holant instance encodes an arbitrary #CSP instance (in particular, a weighted version of #3‑SAT). An interpolation argument over a complex parameter λ extracts the exact counting value, establishing a parsimonious reduction. The reduction respects the holographic invariants, guaranteeing that no hidden polynomial‑time algorithm can bypass the construction.

A notable technical contribution is an algorithm that, given the 10 independent coefficients of a symmetric ternary function, decides in polynomial time whether the function can be transformed into one of the two tractable normal forms. This decision procedure examines the tensor rank (via singular‑value decomposition of appropriate flattenings) and checks for the existence of an affine relation among the entries, all within O(1) field operations.

The paper also discusses broader implications. It is the first full dichotomy for Holant* problems beyond the Boolean domain, demonstrating that holographic reductions remain powerful when the transformation group expands to O(3). The discovery of the low‑rank tractable class reveals a previously unknown polynomial‑time island that does not correspond to any known tractable CSP or graph homomorphism class. Moreover, the affine‑type class connects Holant* to classical homomorphism counting problems, providing a unified view of several counting frameworks.

Finally, the authors outline future directions. Extending the classification to domain size four or higher will likely require new invariants (e.g., higher‑order tensor symmetries) and possibly richer transformation groups (such as the unitary group U(d)). They conjecture that the dichotomy pattern—tractable low‑rank or affine structures versus #P‑hardness—persists for all finite domains, but proving this will demand deeper algebraic tools. The paper thus opens a promising research avenue at the intersection of algebraic complexity, holographic algorithms, and tensor analysis.