Conditional Mutual Information and Information-Theoretic Phases of Decohered Gibbs States

Yifan Zhang, Sarang Gopalakrishnan · 2025

Read Conditional Mutual Information and Information-Theoretic Phases of Decohered Gibbs States by Yifan Zhang, Sarang Gopalakrishnan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-13210.txt.

Introduction

Yifan Zhang and Sarang Gopalakrishnan find phase transitions made of pure information in a 2025 quantum information paper. Orderly states called Markov networks reveal their structure by showing zero conditional mutual information between distant parts, but adding local noise can smear information across the whole system.

The authors prove the noisy case splits by temperature. At high temperature, the shared information between far regions still decays exponentially, for classical systems under any local noise and for a broad quantum class. At low temperature, long-range information can persist. Strikingly, these information-theoretic transitions appear in models with no ordinary thermal transition at all.

Catalog record

CREATORYifan Zhang, Sarang Gopalakrishnan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE132 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-13210

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Conditional Mutual Information and Information-Theoretic Phases of Decohered Gibbs States Yifan Zhang, Sarang Gopalakrishnan Abstract Classical and quantum Markov networks—including Gibbs states of commuting local Hamiltonians—are characterized by the vanishing of conditional mutual information (CMI) between spatially separated subsystems. Adding local dissipation to a Markov network turns it into a hidden Markov network , in which CMI is not guaranteed to vanish even at long distances. The onset of long-range CMI corresponds to an information-theoretic mixed-state phase transition, with far-ranging implications for teleportation, decoding, and state compressions. Little is known, however, about the conditions under which dissipation can generate long-range CMI. In this work we provide the first rigorous results in this direction. We establish that CMI in high-temperature Gibbs states subject to local dissipation decays exponentially, (i) for classical Hamiltonians subject to arbitrary local transition matrices, and (ii) for commuting local Hamiltonians subject to unital channels that obey certain mild restrictions. Conversely, we show that low-temperature hidden Markov networks can sustain long-range CMI. Our results establish the existence of finite-temperature information-theoretic phase transitions even in models that have no finite-temperature thermodynamic phase transitions. We also show several applications in quantum information and many-body physics. Preprint: APS/123-QED It is generally believed that long-range correlation quantities cannot be generated under finite-time dynamics because of the Lieb-Robinson bound. However, such a belief does not hold true in conditional mutual information (CMI), a multipartite correlation measure. Unlike standard measures of entanglement and mutual information, the CMI does not obey a light cone [ 1 ] , in quantum or classical systems, and can be generated using local channels. In the quantum context, the ability of local channels to generate long-range CMI can be seen as a form of generalized teleportation: informally, if Bob separately shares a Bell pair with Alice and one with Claire, he can implement a Bell measurement and induce entanglement between Alice and Claire that is conditional on the measurement outcome [ 2 ] . The generation of long-range CMI under local or finite-time dynamics is closely related to quantum error corrections and mixed-state phase transitions. Under local noisy dynamics, error accumulates in the code and destroys the codeword after an O(1) time threshold. This threshold transition challenges the standard paradigms of statistical mechanics as correlation length cannot diverge in O(1) time. However, noise accumulation can generate long-range CMI, which operationally reflects the failure of quasi-local decoders after passing the threshold. Similar argument also shows that long-range CMI is a crucial ingredient in all mixed-state phase transitions [ 3 ] . However, very little is systematically known about the states that can generate long-range CMI under local decoherence. The bulk of this work is dedicated to establishing that local decoherence cannot induce long-range CMI in a wide class of high-temperature Gibbs states. Our considerations apply to both quantum and classical CMI. The classical states we consider are Markov networks , states with zero CMI, but local dissipation generates CMI and turns them into hidden Markov networks . we call the analogous quantum states quantum hidden Markov networks . In both cases, we construct low-temperature states that provably generate long-range CMI under decoherence. Together, our results imply the existence of nonzero-temperature information-theoretic transitions in the structure of Gibbs states, even when there is no thermodynamic phase transition at nonzero temperature: dissipation generates long-range CMI only when the initial state is below a critical temperature. Figure 1: (a) An example of partitioning a Markov network exhibiting conditional independence. {\color[rgb]{0,1,0}Y} separates {\color[rgb]{0.06,0.89,1}X} and {\color[rgb]{1,0.9,0.16}Z} and we have I({\color[rgb]{0.06,0.89,1}X}:{\color[rgb]{1,0.9,0.16}Z}|{\color[rgb]{0,1,0}Y})=0 . (b) An example of hidden Markov chain. WXYZ forms a Markov chain and each site is subject to a transition matrix. (c) An example of quantum hidden Markov chain. \rho_{ABCD} forms a quantum Markov chain and each site is subject to a quantum channel. Note that we use the double leg to denote the bra and ket, so fixing the index mn corresponds to choosing \ket{m}\bra{n} . (d) An example of classical hidden Markov network. Grey dashed lines separate sites and we partition the system into {\color[rgb]{0.06,0.89,1}X} , {\color[rgb]{0,1,0}Y} , and {\color[rgb]{1,0.9,0.16}Z} . The white boxes represent a channel that takes two bits as input and outputs their parity. (e) An example of quantum hidden Markov network. Grey dashed lines separate sites and we partition the system into {\color[rgb]{0.06,0.89,1}A} , {\color[rgb]{0,1,0}B} , and {\color[rgb]{1,0.9,0.16}C} . The circuit rotates the computation basis to the Bell basis and then measures it. Definitions .—We start with the classical case. The CMI I({\color[rgb]{0.06,0.89,1}X}:{\color[rgb]{1,0.9,0.16}Z}|{\color[rgb]{0,1,0}Y}) is the mutual information between the two random variables {\color[rgb]{0.06,0.89,1}X} and {\color[rgb]{1,0.9,0.16}Z} , conditioned on a third variable, {\color[rgb]{0,1,0}Y} . In terms of Shannon entropies, I({\color[rgb]{0.06,0.89,1}X}:{\color[rgb]{1,0.9,0.16}Z}|{\color[rgb]{0,1,0}Y})=H({\color[rgb]{0.06,0.89,1}X}{\color[rgb]{0,1,0}Y})+H({\color[rgb]{1,0.9,0.16}Z}{\color[rgb]{0,1,0}Y})-H({\color[rgb]{0,1,0}Y})-H({\color[rgb]{0.06,0.89,1}X}{\color[rgb]{0,1,0}Y}{\color[rgb]{1,0.9,0.16}Z}) . Note that this definition also carries over to the quantum setting if one replaces all Shannon entropies with Von Neumann entropies. If I({\color[rgb]{0.06,0.89,1}X}:{\color[rgb]{1,0.9,0.16}Z}|{\color[rgb]{0,1,0}Y})=0 , a property known as conditional independence, the variables {\color[rgb]{0.06,0.89,1}X} , {\color[rgb]{0,1,0}Y} , and {\color[rgb]{1,0.9,0.16}Z} are said to form a (classical or quantum) Markov chain . Intuitively, this means that all correlations between {\color[rgb]{0.06,0.89,1}X} and {\color[rgb]{1,0.9,0.16}Z} are mediated by {\color[rgb]{0,1,0}Y} . In higher dimensions, Markov chains generalize to Markov networks , where a lattice of random variables exhibits zero CMI when partitioned such that {\color[rgb]{0,1,0}Y} separates {\color[rgb]{0.06,0.89,1}X} and {\color[rgb]{1,0.9,0.16}Z} (see Fig. 1 (a)). Quantum Markov chain and Markov networks are analogously defined [ 4 , 5 ] . Because Markov networks represent distributions without multipartite correlations, they are highly structured and computationally tractable. They have been extensively studied across physics, probability theory, and machine learning. A key property of classical Markov chains is the Hammersley-Clifford theorem, which establishes the equivalence between Markov networks and Gibbs states of local Hamiltonians. However, while quantum Markov networks and Gibbs states of commuting Hamiltonians on Hypercube lattices are equivalent [ 6 ] 1 1 1 There exist quantum Markov networks on graphs with triangles that cannot be written as Gibbs states of commuting local Hamiltonians [ 6 ] . However, this non-commuting property disappears after “coarse-graining” by combining multiple adjacent sites into a super-site. To our knowledge, no examples exist of quantum Markov networks that are not Gibbs states of commuting local Hamiltonians even after coarse-graining. , generic Gibbs states of local quantum Hamiltonians do not obey the Hammersley-Clifford theorem. Local dissipation does not preserve the Markov chain property: instead, it turns Markov chains (networks) into hidden Markov chains (networks) (Fig. 1 (b)), for which the CMI is generically positive. In classical probability, hidden Markov networks are powerful tools for modeling complex distributions due to their ability to encode non-local, multipartite correlations. Quantum Hidden quantum Markov chains/networks are defined analogously (Fig. 1 (c)) 2 2 2 We note that the notion of Hidden Quantum Markov Model is defined in the literature [ 45 ] . Their quantum generalization of the hidden Markov model is also well-motivated but is different from ours. To our best knowledge, the quantum generalization of hidden Markov networks has not been discussed in literature , and again have nonvanishing CMI. The existence of long-range CMI in the quantum setting is associated with phenomena that have no classical counterpart, such as quantum teleportation and measurement-induced entanglement [ 9 , 10 , 11 , 12 , 13 , 14 , 15 , 2 , 16 ] .

Read the complete text