Tripartite Mutual Information (TMI)

• Tripartite Mutual Information (TMI) is a multipartite measure that quantifies total correlations among three subsystems, distinguishing pairwise from synergistic interactions.
• TMI diagnoses information scrambling by revealing whether mutual information is delocalized, with negative values indicating nonlocal encoding of quantum information.
• TMI finds applications in many-body physics, holographic dualities, and measurement-induced phase transitions, offering insights into quantum chaos and entanglement dynamics.

Tripartite mutual information (TMI) is a multipartite extension of quantum mutual information that characterizes how total correlations—quantum and classical—are distributed among three subsystems of a composite quantum state or region. It plays a central role as an operator-independent diagnostic for information delocalization, quantum scrambling, and the structure of correlations in quantum many-body systems, open-system dynamics, and holographic dualities.

1. Formal Definition and Equivalent Expressions

Given a tripartite quantum state $\rho_{ABC}$, the tripartite mutual information is defined as

$I_3(A:B:C) = I(A:B) + I(A:C) - I(A:BC),$

where the bipartite mutual information is

$I(X:Y) = S(\rho_X) + S(\rho_Y) - S(\rho_{XY}),$

and $S(\rho) = -\mathrm{Tr}[\rho \ln \rho]$ is the von Neumann entropy. $I_3$ may be expanded fully symmetrically: $$I_3(A:B:C) = S(A) + S(B) + S(C) - S(AB) - S(AC) - S(BC) + S(ABC).$$ By construction, $I_3$ is finite and symmetric under permutations of $A,B,C$; it quantifies whether correlations among subsystems are merely pairwise (redundant or extensive), or possess “synergy” where information about one subsystem is encoded in a genuinely multipartite fashion (Han et al., 2022, Monaco et al., 2023, Balasubramanian et al., 2011).

2. Physical Interpretation and Scrambling Diagnostics

The operational significance of $I_3$ lies in its sign and magnitude:

• $I_3 < 0$ (Super-extensive / Monogamous): The whole ($BC$) contains strictly more mutual information with $A$ than either $B$ or $C$ alone, i.e., information about $A$ is delocalized—it can only be retrieved by joint access to $B\cup C$. This serves as a direct signature of quantum information scrambling in closed systems; under many-body dynamics, locally encoded information spreads irreversibly into nonlocal degrees of freedom and becomes inaccessible by local measurements alone (Han et al., 2022, Iyoda et al., 2017, Kuno et al., 2022).
• $I_3 > 0$: Pairwise redundancies dominate, as in classical information or non-scrambling quantum dynamics.
• $I_3 = 0$: No three-way effect beyond what's determined by pairwise correlations.

In the context of quantum channels, evolution with substantial $I_3 < 0$ is evidence of information scrambling. Channel–state duality relates $I_3 < 0$ to nonzero out-of-time-ordered correlators (OTOCs), making $I_3$ an operator-independent witness of scrambling behavior. However, $I_3$ includes both quantum and classical correlations and may misdiagnose scrambling in certain open or measurement-induced settings (Han et al., 2022, Monaco et al., 2023).

3. Behavior in Closed and Open Quantum Systems

Closed Unitary Systems:

In unitary many-body dynamics, $I_3$ robustly diagnoses scrambling: starting from a locally entangled state (e.g., initial $A$ maximally entangled with $B$), evolution leads to $I_3 \to -\text{const} < 0$, with the negativity quantifying irreducible nonlocal encoding. This behavior holds across integrable/non-integrable Hamiltonians, as $I_3$ is logically independent of conventional quantum chaos as defined by level statistics or ETH (Iyoda et al., 2017, Kuno et al., 2022).

Open Quantum Systems:

In open systems, especially under non-Markovian environments, $I_3$ can become negative even in the absence of genuine quantum entanglement. For example, in local dephasing channels, steady-state $I_3<0$ can survive after all quantum correlations have decayed, merely reflecting classical delocalization of correlations. Conversely, in some dissipative channels, $I_3$ may remain non-negative even when transient quantum delocalization occurs. To resolve these issues, one can instead use the tripartite logarithmic negativity (TLN), defined analogously to $I_3$ but replacing quantum mutual information with logarithmic negativity between pairs of subsystems. Unlike $I_3$, TLN captures only genuine quantum entanglement, providing an unambiguous scrambling diagnostic in open-system or noisy contexts (Han et al., 2022).

4. Accessible Tripartite Mutual Information and Measurement Constraints

Standard $I_3$ is based on von Neumann entropy and quantum mutual information, which are not, in general, operationally accessible via local measurements. This leads to several pitfalls:

• Non-accessibility: Quantum mutual information does not always correspond to statistics of local measurement outcomes.
• Encoding-basis independence: $I_3$ can depend on the basis in which information is encoded, failing to capture operationally relevant scrambling in certain bases.
• Blindness to pure-state structure: $I_3$ vanishes for all pure tripartite states, but these can have widely varying structures of correlation (e.g., GHZ vs. W states). To address these defects, (Monaco et al., 2023) defines the accessible tripartite mutual information, $I_3^\text{acc}$, which replaces quantum mutual information by its maximized classical version over local POVMs: $$I_3^\text{acc}(R;C;D) = I_\text{acc}(R:C) + I_\text{acc}(R:D) - I_\text{acc}(R:CD).$$ This measure captures only correlations extractable by local measurement and thus faithfully diagnoses operationally meaningful scrambling, providing a more robust tool especially for dynamical channels and scenarios where basis-choice and measurement access constrain the correlation structure.

5. Application Domains: Many-Body Systems, Holography, and Measurement-Induced Transitions

Disordered and Many-Body Localized Phases:

In ergodic regimes ($I_3 < 0$), TMI characterizes fast scrambling; in many-body localized (MBL) phases, $I_3$ remains near zero or turns negative only extremely slowly, reflecting inhibited information delocalization. In random spin chains with topological order or self-duality properties, the long-time TMI tracks different MBL phases, and its value is sensitive to the spatial structure of stabilizers and phase transitions in ergodicity (Kuno et al., 2022, Orito et al., 2022).

Holographic Quantum Field Theories:

In large-$N$ strongly-coupled CFTs with gravitational duals, $I_3$ is computed via Ryu–Takayanagi or Hubeny–Rangamani–Takayanagi extremal surface prescriptions. A hallmark of holographic theories is the strict monogamy property: $I_3 \leq 0$ for all choices of regions—implying mutual information cannot be simultaneously shared extensively across multiple parties. This is observed universally in computations for AdS/CFT (Balasubramanian et al., 2011), flat-space holography (Asadi et al., 2018), and non-conformal holographic backgrounds (Ali-Akbari et al., 2019, Asadi et al., 2018).

Measurement-Induced Phase Transitions and Random Circuits:

The critical behavior and saturation value of $I_3$ are sensitive to measurement protocols and effectively locate measurement-induced transitions in random quantum circuits, with TMI discriminating between phases with area-law or volume-law entanglement scaling (Orito et al., 2022).

6. Structural Properties: Monogamy, Lower Bounds, and Connections to Entanglement Measures

$\,I_3$ encodes subtle constraints on how correlations are shared:

• Monogamy: $I_3 < 0$ is equivalent to monogamy of mutual information: $I(A:B)+I(A:C) \leq I(A:BC)$. This is exact for holographic theories and is closely tied to the extensivity or synergy of correlations (Balasubramanian et al., 2011, Maziero, 2013).
• Bounds and Residual Correlation: For arbitrary tripartite mixed states, $$I(\rho_{123}) \geq \frac{2}{3}(I_{12}+I_{23}+I_{31})$$. The residual correlation $$I_r = I(\rho_{123}) - \frac{2}{3} (I_{12} + I_{23} + I_{31})$$ is always nonnegative but does not accurately capture genuine three-way entanglement in all cases (Maziero, 2013).
• Limitations: In perfect multipartite-entangled (MMES) states, $I_3$ can be maximally negative for certain partitions. However, the behavior under small deformations and in disconnected composite systems indicates that averaging over permutations or using more refined diagnostics is necessary for robustly characterizing scrambling and multipartite entanglement (Rota, 2015).

7. Limitations and Contextual Sensitivity

$\,I_3$ is a powerful but context-sensitive observable:

• Classical vs Quantum Delocalization: Negative $I_3$ can be produced by classical correlations, especially in open systems where all quantum entanglement has decayed. Logarithmic negativity-based analogs are necessary to distinguish genuine quantum scrambling in the presence of decoherence (Han et al., 2022).
• Permutational Symmetry Constraints: In fully permutation-symmetric systems (e.g., symmetric Dicke ensembles, quantum kicked top), $I_3$ is generically positive, even when operator scrambling occurs as signaled by OTOCs. The “area-law” scaling of entanglement in such systems inherently precludes the negativity of TMI. This shows TMI and operator scrambling can decouple under strong symmetry constraints (Seshadri et al., 2018).
• Operational Accessibility: Measurement restrictions, encoding choices, and channel structure can render $I_3$ non-informative about the actual information retrievability in experimental settings, motivating the accessible mutual information and its multipartite extensions (Monaco et al., 2023).
• Dynamical and Environmental Factors: In non-Markovian open quantum systems, environment memory enhances transient negativity of $I_3$ and genuine entanglement delocalization but ultimately damps scrambling at long times; Markovian vs. non-Markovian regimes lead to qualitatively different TMI dynamics (Han et al., 2022).

---

References

• "Quantum information scrambling in non-Markovian open quantum systems" (Han et al., 2022)
• "Quantum scrambling via accessible tripartite information" (Monaco et al., 2023)
• "Thermalization of mutual and tripartite information in strongly coupled two dimensional conformal field theories" (Balasubramanian et al., 2011)
• "Information spreading and scrambling in disorder-free multiple-spin interacting models" (Kuno et al., 2022)
• "Scrambling of Quantum Information in Quantum Many-Body Systems" (Iyoda et al., 2017)
• "Distribution of Mutual Information in Multipartite States" (Maziero, 2013)
• "Quantum information spreading in random spin chains with topological order" (Orito et al., 2022)
• "Tripartite mutual information, entanglement, and scrambling in permutation symmetric systems with an application to quantum chaos" (Seshadri et al., 2018)
• "Delocalization of quantum information in long-range interacting systems" (Wanisch et al., 2021)
• "Tripartite information of highly entangled states" (Rota, 2015)
• "Negative tripartite mutual information after quantum quenches in integrable systems" (Caceffo et al., 2023)
• "Scrambling in Ising spin systems with periodic transverse magnetic fields" (Shukla, 2023)
• "Revealing Tripartite Quantum Discord with Tripartite Information Diagram" (Lee et al., 2017)
• "Holographic Calculation of BMSFT Mutual and 3-partite Information" (Asadi et al., 2018)
• "Effect of acceleration on information scrambling" (Ming, 2023)
• "Holographic Mutual and Tripartite Information in a Non-Conformal Background" (Ali-Akbari et al., 2019)
• "Holographic Mutual and Tripartite Information in a Symmetry Breaking Quench" (Asadi et al., 2018)
• "A tripartite entanglement in de Sitter spacetime" (Bak et al., 2019)
• "Tripartite information at long distances" (Agón et al., 2021)