Holographic Simple Tree Graph Model

• Holographic simple tree graph model is a mathematical framework that represents entropy vectors and partition functions on tree-structured graphs using normal factor graphs and holographic transformations.
• It enables efficient, linear-time computation of partition functions while bridging combinatorial and algebraic methods to analyze quantum and classical correlation constraints.
• The model is crucial in applications such as holographic entropy cones, network theory, and p-adic AdS/CFT, offering explicit constructions for entropy vector realizability.

A holographic simple tree graph model is a mathematical framework for representing and computing entropy vectors and partition functions on tree-structured graphs, rooted in the theory of normal factor graphs (NFGs), holographic transformations, and their application to problems in holographic entropy and tractable inference. These models play a key role in bridging combinatorial and linear-algebraic representations of quantum and classical correlations, with applications to holographic entropy cones, network theory, and the study of quantum and classical marginal constraints for multipartite systems (Al-Bashabsheh et al., 2010, Hubeny et al., 21 Dec 2025).

1. Definitions and Formal Structure

A holographic simple tree graph model is a specialized normal factor graph (NFG) $G=(V,E,\emptyset,\{f_v\})$ whose underlying undirected graph is a tree without dangling edges. Each vertex $v \in V$ is assigned a local function $f_v$ defined over the product of finite alphabets associated to the incident edges $\partial v$. The partition function (Holant) is

$Z = \sum_{x_E} \prod_{v\in V} f_v(x_{\partial v}),$

where $x_E$ denotes the collections of variables associated to all edges. In the entropy vector context, a simple $n$-party graph model $G=(V,E,w,\partial V,\ell)$ consists of a weighted tree where boundary vertices are bijectively labeled by parties and weights are non-negative. For a nonempty subset $I\subseteq\{1..n\}$, the entropy $S_I(G)$ is determined by the minimal boundary cut, with the entropy vector defined as $\{S_I(G)\}_{\emptyset\ne I\subset[n]}$ (Hubeny et al., 21 Dec 2025).

2. Holographic Transformations and the Holant Theorem

A holographic transformation assigns, for each half-edge $(v,e)$, an invertible linear map $T_{v,e}$ on the alphabet of $e$, with the property that for internal edges $(u,v)$, $T_{u,e}$ and $T_{v,e}$ are mutual inverses. Each local function is replaced with

$$f'_v(y_{\partial v}) = \sum_{x_{\partial v}} f_v(x_{\partial v}) \prod_{e\in\partial v} T_{v,e}(x_e, y_e),$$

yielding a transformed NFG $G^H$, with the notable property that for trees and no dangling edges, the partition function is strictly invariant:

$$\sum_{x_E} \prod_{v} f_v(x_{\partial v}) = \sum_{y_E} \prod_{v} f'_v(y_{\partial v}).$$

This transformation is central to tractable evaluation of partition functions and underpins belief propagation (sum-product) as a sequence of local holographic transformations eliminating leaves in the tree. The partition function is thus computable in linear time on the number of vertices (Al-Bashabsheh et al., 2010).

3. Characterization by Entropy Vectors and Chordality Condition

A central question is which entropy vectors can be realized by a holographic simple tree graph model. Let $S$ be an entropy vector for $n$ parties, and $P$ its pattern of marginal independence (PMI). The key result is that $S$ is realizable by such a model if and only if the line graph $L$ of its correlation hypergraph $H(P)$ is chordal (i.e., all cycles of length $\ge4$ have a chord). This Buneman–Gavril–Walter theorem provides a necessary and, conjecturally, sufficient condition. The construction of $H$ encodes subsets of parties with nontrivial mutual information. The chordality condition is efficiently checkable and leads directly to an explicit model when satisfied (Hubeny et al., 21 Dec 2025).

Element	Description	Reference
Partition func	$Z = \sum_{x_E} \prod_v f_v(x_{\partial v})$	(Al-Bashabsheh et al., 2010)
Chordality	All cycles in $L$ of length ≥4 have a chord	(Hubeny et al., 21 Dec 2025)
Entropy vector	$S_I(G) =$ cost of minimal $I$-cut	(Hubeny et al., 21 Dec 2025)

4. Construction Algorithms and Complexity

Given $S$ that passes strong subadditivity and subadditivity and whose PMI is chordal, the construction proceeds as follows:

1. Compute $P$ and the correlation hypergraph $H$.
2. Form the line graph $L$ of $H$.
3. Construct the clique graph $K$ of $L$ using maximal cliques.
4. Compute a maximum-weight spanning tree $\tilde{T}$ of $K$; this serves as the model's tree structure.
5. Attach boundary leaves for each party.
6. Assign edge weights to capture the desired cut costs (entropies). The total algorithmic complexity is polynomial in the size of the MI-poset, typically $O(2^{2n})$, but much smaller in practice for realistic $n$ due to the sparsity induced by the chordality condition. The construction provides explicit control over realizable entropy vectors and can detect non-realizability efficiently (Hubeny et al., 21 Dec 2025).

5. Correlation Hypergraphs, Coarse-Graining, and Fine-Graining

To generalize the construction and analyze marginal constraints, the correlation hypergraph toolkit is employed. Coarse-graining (a surjective map $\Theta$ from $n'$ to $n$ parties) induces maps on entropy vectors, PMI, and hypergraphs, and is compatible with weak-union closure at the hypergraph level. Fine-graining involves lifting PMI and expanding the hypergraph such that each party is replaced with a block of parties (“blowing up” vertices). Every non-simple tree model on $n$ parties is the coarse-graining of some simple tree model on $n'>n$ parties. This correspondence enables a systematic approach to realize entropy vectors on more general (possibly non-simple) holographic tree graph models and to algorithmically search for fine-grainings that realize a given vector (Hubeny et al., 21 Dec 2025).

6. Applications, Examples, and Detection of Unrealizability

Holographic simple tree graph models are foundational in representing entropy vectors in the context of holographic entropy cones, network models, and marginal constraints. A concrete example for $n=4$ demonstrates the construction: the cut structure yields an explicit model that reproduces all $S_I$ by direct min-cut computation. If a given $S$ fails the chordality test for any coarse- or fine-graining, it is concluded not to be tree-realizable. Moreover, the framework independently verifies unrealizability without relying on an a priori list of entropy inequalities—a significant advance for entropy cone analysis (Hubeny et al., 21 Dec 2025).

Further, in contexts such as $p$-adic and discrete holographic models, simple tree graphs underpin the construction of bulk-to-boundary correlators and the tractable computation of partition functions and Green's functions. For example, the specific structure of biregular trees in $p$-adic AdS/CFT is directly related to holographic simple tree models through their semihomogeneous geometry and computation of boundary correlators (Mondal et al., 28 Jul 2025).

7. Connections and Theoretical Significance

Holographic simple tree graph models cohesively unify the NFG formalism, holographic transformations, efficient inference (including message-passing and belief propagation), and the combinatorics of entropy vector realization. The framework encapsulates:

• Holographic invariance: any choice of invertible $T_{v,e}$ leaves $Z$ unchanged.
• Linear-time computability of $Z$ on trees, owing to the absence of cycles.
• Closed-form solutions for special choices of $T$ (e.g., Hadamard, Fourier) and situations where $f_v$ admit matchgate or diagonalizable realizations.
• A conceptual and algorithmic toolkit (correlation hypergraphs, chordality, fine/coarse-graining) connecting combinatorial objects (graphs, hypergraphs) to fundamental quantum information constraints. This establishes holographic simple tree graph models as a central tool in the design, construction, and analysis of entropy cones, holographic codes, and tractable inference on tree-structured systems (Al-Bashabsheh et al., 2010, Hubeny et al., 21 Dec 2025).