Fully quantum inflation: quantum marginal problem constraints in the service of causal inference

1. Problem Statement

The paper addresses the Fully Quantum Causal Compatibility Problem. In classical causal inference, this problem involves determining whether a given probability distribution over observed variables is compatible with a specific causal structure (represented by a Directed Acyclic Graph, or DAG) containing latent (unobserved) variables.

The authors extend this to the quantum domain:

• Context: A multipartite quantum system where the "observed" nodes are quantum systems (visible nodes) and the "unobserved" causes are also quantum systems (latent nodes).
• Goal: Given a joint quantum state $\rho$ on the visible subsystems, determine if it can be generated by a quantum causal network with a specific structure (e.g., the triangle network).
• Challenge: Unlike the classical case, quantum systems cannot be broadcast (No-Broadcasting Theorem), and the parameters of the model involve quantum channels and states rather than stochastic maps and probability distributions. Existing methods often rely on converting quantum states to classical distributions via measurement, which may lose information about the state's incompatibility.

2. Methodology: Fully Quantum Inflation

The authors introduce and formalize Fully Quantum Inflation, a technique that leverages the Quantum Marginal Problem to derive necessary conditions for causal compatibility.

Core Concepts

• Inflation Technique: The method maps the problem of checking if a state $\rho$ is compatible with a causal structure $G$ to checking if a set of marginals of $\rho$ is compatible with an "inflated" structure $G'$.
• The Inflation Map ($G \to G'$):
  • The inflated DAG $G'$ contains multiple copies of the original nodes.
  • Non-Fanout Restriction: The authors restrict attention to non-fanout inflations. In a fanout inflation, a latent node influences multiple copies of the same visible node. This is impossible in the fully quantum case due to the No-Broadcasting theorem. Therefore, only non-fanout inflations (like the "Cut" inflation) are valid for deriving constraints in the fully quantum regime.
  • Parameter Mapping: Parameters (states and channels) in $G$ are mapped to $G'$ such that the state on a latent node in $G'$ is the reduced state of the original latent state on the relevant subsystems.
• Injectable and Expressible Sets:
  • Injectable Sets: Subsets of visible nodes in $G'$ whose ancestral subgraph is isomorphic (up to copy indices) to a subset in $G$.
  • Expressible Sets: Unions of injectable sets that are ancestrally independent in $G'$.
  • Corollary 1: If a state is compatible with $G$, then the family of states formed by the marginals on the expressible sets of $G'$ must be compatible with $G'$.

Leveraging the Quantum Marginal Problem

The key innovation is using constraints from the Quantum Marginal Problem (specifically operator inequalities derived by Hall) on the inflated network $G'$ to generate incompatibility witnesses for the original network $G$.

• Hall's Inequality: For a set of quantum systems $V$ with odd cardinality, a set of marginals $\{\sigma_X\}$ is compatible with a global state only if a specific operator $\Delta(\{\sigma_X\})$ is positive semi-definite (PSD).
• Deriving Witnesses: By substituting the causal constraints of the inflated network (e.g., factorization of marginals for d-separated nodes) into Hall's inequality, the authors derive a new operator inequality $I(\rho) \geq 0$. If $I(\rho)$ is not PSD, the state $\rho$ is incompatible with the causal structure.

3. Key Contributions

1. Formalization of Fully Quantum Inflation: The paper provides the first rigorous framework for applying inflation techniques where both visible and latent nodes are quantum, explicitly handling the constraints imposed by the No-Broadcasting theorem.
2. Complete Classification of Pure Three-Qubit States:
   • The authors prove Theorem 1: A pure three-qubit state is compatible with the triangle scenario if and only if it is biseparable (factorizable across some bipartition).
   • This implies that any genuinely tripartite entangled pure state is incompatible with the triangle scenario.
3. Distribution-Witnessability Analysis:
   • The paper investigates whether incompatibility of a quantum state can be inferred solely from the incompatibility of the probability distribution obtained by local measurements (distribution-witnessability).
   • Proposition 3: They construct a family of pure states that are triangle-incompatible, yet their incompatibility cannot be witnessed by the same Cut inflation using any local measurement distribution. This proves that fully quantum inflation is strictly more powerful than distribution-based tests for certain states.
4. Extension to Mixed States and Higher Dimensions:
   • The method is applied to mixed states, showing that separable states can be triangle-incompatible (unlike the pure state case where biseparability implies compatibility).
   • The technique is generalized to higher-dimensional systems (qutrits, ququarts), demonstrating that the method is not limited to qubits.
5. Generalization to Other Networks:
   • The authors extend the method beyond the triangle scenario to a family of networks $G(n, l)$, including the Pentagon and Woven Hexagon scenarios.
   • They demonstrate how to construct incompatibility witnesses for these larger networks by defining composite systems as elements of the marginal problem set $V$.

4. Key Results

• Triangle Scenario (Pure States): The boundary between compatible and incompatible pure states coincides exactly with the boundary between biseparable and genuinely multipartite entangled states.
• Triangle Scenario (Mixed States): The boundary is more complex. There exist separable (non-entangled) mixed states that are triangle-incompatible. Conversely, there are entangled states that are triangle-compatible (e.g., certain Werner-like states).
• Incompatibility Witnesses: The paper provides explicit operator inequalities (e.g., $I_{AB}(\rho) \geq 0$) that serve as witnesses. Violation of these inequalities (non-PSD result) proves incompatibility.
• Computational Efficiency: The authors note that while the general quantum marginal problem is hard, using specific operator inequalities (like Hall's) allows for incompatibility testing with complexity $O(D^3)$ (eigendecomposition) rather than the $O(D^{13/2})$ complexity of general Semidefinite Programming (SDP) hierarchies, offering a practical speedup for specific cases.
• Counter-Examples:
  • GHZ and W States: Both pure GHZ and W states are proven triangle-incompatible.
  • Non-Distribution-Witnessable States: A specific family of "tri-Bell" states is shown to be incompatible, but no local measurement basis yields a distribution that violates the Cut-inflation inequalities.

5. Significance

• Foundational Understanding: The work deepens the understanding of the relationship between quantum entanglement and causal structures. It clarifies that "genuine multipartite entanglement" (defined via Local Operations and Shared Randomness with 2-Way Shared Entanglement, LO2WSE) is equivalent to incompatibility with the triangle network for pure states.
• Experimental Verification: The method provides a practical tool for experimentalists. Given a tomographically reconstructed quantum state from a network experiment, one can computationally test if the state could have been generated by the assumed causal structure. A failure indicates a deficiency in the experimental setup or understanding of the causal links.
• Theoretical Hierarchy: It establishes that fully quantum causal inference is strictly more powerful than classical causal inference applied to measurement statistics. There are quantum states whose causal incompatibility is "hidden" from classical observers (who only see measurement distributions) but is revealed by the quantum marginal constraints.
• Scalability: By generalizing the technique to arbitrary network topologies (Pentagon, Hexagon, etc.), the paper opens the door to analyzing complex quantum networks relevant to the future quantum internet.

In summary, this paper bridges the gap between quantum information theory and causal inference, providing a robust, computationally tractable, and theoretically rigorous framework for certifying the causal origins of quantum states.