Quantum Darwinism explains the objectivity of classical reality through redundant records of a system's pointer states imprinted on fragments of its decohering environment. Touil, Yan, and Zurek (arXiv:2503.14791, 2025) quantified the agreement — consensus — between two observers accessing distinct environment fragments using the refined mutual information between the fragments, I(F:F′∣S)=I(F:F′)−I(F:F′∣S)\mathfrak{I}(\mathcal{F}:\mathcal{F}'|\mathcal{S})=I(\mathcal{F}:\mathcal{F}')-I(\mathcal{F}:\mathcal{F}'|\mathcal{S}), and its normalized version C\mathfrak{C}. They proved that fragments carrying enough information about the system force observer agreement on the pointer state, and that I\mathfrak{I} is nonnegative — but both results hold only within the branching-state family, with Theorem 1 additionally assuming branches completely decohered by the rest of the environment. When environment subsystems carry preexisting correlations or interact with one another, the state is no longer branching, I\mathfrak{I} can become negative, and numerical simulations show that scrambling eventually destroys consensus; the authors state that a detailed analysis of such cases is beyond the scope of their work. The open problem is a complete characterization of inter-fragment consensus measures — positivity conditions, the operational interpretation of negative refined mutual information, and dynamical behavior — in generic non-branching, interacting, or inadequately decohered environments.

Background

Decoherence theory explains how a quantum system S\mathcal{S} coupled to an environment E\mathcal{E} loses its quantum coherences: the interaction selects preferred 'pointer' states that are robust against monitoring by the environment, a process called einselection. Quantum Darwinism (QD) goes further by recognizing that the decohering environment also acts as a communication channel: many observers can each access a small fragment F\mathcal{F} of the environment and independently infer the state of S\mathcal{S}. Objectivity — the fact that many observers agree on one classical state — is then identified with the redundancy of records, usually quantified by the mutual information I(S:F)=HS+HF−HSFI(\mathcal{S}:\mathcal{F})=H_{\mathcal{S}}+H_{\mathcal{F}}-H_{\mathcal{SF}} (with HH the von Neumann entropy), which for redundant records saturates the system entropy HSH_{\mathcal{S}} for fragments of modest size.

A central simplifying assumption in this program is the branching structure of the global state,

∣ΨSE⟩=∑n=0DS−1qn ∣sn⟩∣Fn⟩∣Fn′⟩∣E/FFn′⟩,|\Psi_{\mathcal{SE}}\rangle=\sum_{n=0}^{D_{\mathcal{S}}-1}\sqrt{q_{n}}\,|s_{n}\rangle|\mathcal{F}_{n}\rangle|\mathcal{F}'_{n}\rangle|\mathcal{E}_{/\mathcal{F}\mathcal{F}'_{n}}\rangle,

where the system is in orthogonal pointer states ∣sn⟩|s_{n}\rangle and, conditioned on each branch, the two fragments F\mathcal{F} and F′\mathcal{F}' and the remainder of the environment E/FF′\mathcal{E}_{/\mathcal{F}\mathcal{F}'} carry pure record states. Branching states are exactly the states with zero quantum discord (a measure of genuinely quantum, non-classical correlations) between the fragments and the rest, and small discord forces the state to be close to branching form — so generic interacting environments lie outside this family.

Touil, Yan, and Zurek (2025) asked what this record structure implies about agreement between observers, which they call consensus. Their first measure is the normalized mutual information between two distinct fragments, c(F:F′)=I(F:F′)/HS\mathfrak{c}(\mathcal{F}:\mathcal{F}')=I(\mathcal{F}:\mathcal{F}')/H_{\mathcal{S}}. Their Theorem 1 shows that for completely decohered branching states, if I(F:S)=(1−δ)HSI(\mathcal{F}:\mathcal{S})=(1-\delta)H_{\mathcal{S}} and I(F′:S)=(1−δ′)HSI(\mathcal{F}':\mathcal{S})=(1-\delta')H_{\mathcal{S}}, then I(F:F′)=(1−δ^)HSI(\mathcal{F}:\mathcal{F}')=(1-\hat{\delta})H_{\mathcal{S}} with consensus deficit δ^=δ+δ′−δ~\hat{\delta}=\delta+\delta'-\tilde{\delta} and min⁡(δ,δ′)≥δ~≥0\min(\delta,\delta')\ge\tilde{\delta}\ge0: fragments that each know enough about S\mathcal{S} inevitably agree on its pointer state. However, I(F:F′)I(\mathcal{F}:\mathcal{F}') can also contain correlations that have nothing to do with S\mathcal{S} — for instance, correlations created by direct interaction between the environment subsystems. To filter these out, the authors introduce the refined mutual information

I(F:F′∣S)=I(F:F′)−I(F:F′∣S),\mathfrak{I}(\mathcal{F}:\mathcal{F}'|\mathcal{S})=I(\mathcal{F}:\mathcal{F}')-I(\mathcal{F}:\mathcal{F}'|\mathcal{S}),

the three-party quantity that information theorists call the interaction information (its sign convention varies across the literature), and the normalized consensus measure C(F:F′∣S)=I(F:F′∣S)/HS\mathfrak{C}(\mathcal{F}:\mathcal{F}'|\mathcal{S})=\mathfrak{I}(\mathcal{F}:\mathcal{F}'|\mathcal{S})/H_{\mathcal{S}}, which can also be conditioned on a chosen observable Λ\Lambda of S\mathcal{S}. For branching states with perfect records, I(F:F′∣S)=0I(\mathcal{F}:\mathcal{F}'|\mathcal{S})=0 and I\mathfrak{I} reduces to I(F:F′)I(\mathcal{F}:\mathcal{F}'); their Theorem 2 proves I≥0\mathfrak{I}\ge0 for singly-branching states — those requiring perfect, mutually orthogonal record states across branches only in F\mathcal{F} and the remainder E/FF′\mathcal{E}_{/\mathcal{F}\mathcal{F}'}, not in F′\mathcal{F}' — independently of what is measured on S\mathcal{S}. They test the measures analytically in the c-maybe model (a solvable qubit model of imperfect copy gates with a non-interacting environment) and numerically in a many-body spin model with all-to-all intra-environment couplings, H=σSz⊗∑idiσiz+∑j≠kgjkσjz⊗σkzH=\sigma^{z}_{\mathcal{S}}\otimes\sum_{i}d_{i}\sigma^{z}_{i}+\sum_{j\neq k}g_{jk}\sigma^{z}_{j}\otimes\sigma^{z}_{k}.

The limitation emerges in the interacting model. There, the two regimes the theorems cover no longer apply, and numerically the mutual information shared by two fragments is increasingly dominated by correlations the fragments acquire about each other through intra-environment interactions — a process related to quantum information scrambling, the rapid delocalization of local information. For large fragments (∣F∣=∣F′∣≥4|\mathcal{F}|=|\mathcal{F}'|\ge4) and late times, when the remainder of the environment is too small to decohere SFF′\mathcal{S}\mathcal{F}\mathcal{F}', the refined mutual information itself dips to negative values, while the ordinary mutual information between the fragments keeps growing. Consensus rises as decoherence disseminates records and falls when relaxation and scrambling destroy them. Prior to this work, Chisholm, Innocenti, and Palma (Quantum 7, 1074, 2023) had already distinguished redundancy from consensus in quantum objectivity; the 2025 paper supplied the branching-regime theorems but explicitly left the non-branching regime — pre-correlated or interacting environment subsystems, and ineffective decoherence by the environment's remainder — as a stated open direction.

Problem Statement

Extend the quantitative theory of consensus between environment fragments beyond the branching-state regime of Touil, Yan, and Zurek (arXiv:2503.14791). The common objective is a workable, interpretable theory of inter-fragment consensus — one that recovers the branching-state theorems as a limit — for global states that are not branching, with three jointly supporting deliverables. (1) Characterization: for environments whose subsystems carry preexisting correlations before interacting with the system, or interact with one another during or after record formation (so the global state is no longer branching), establish conditions — analogues of the source's Theorems 1 and 2 — under which correlations between two fragments F\mathcal{F} and F′\mathcal{F}' certify that observers accessing them agree on the pointer state of S\mathcal{S}, and determine for which state classes the refined mutual information I(F:F′∣S)=I(F:F′)−I(F:F′∣S)\mathfrak{I}(\mathcal{F}:\mathcal{F}'|\mathcal{S})=I(\mathcal{F}:\mathcal{F}')-I(\mathcal{F}:\mathcal{F}'|\mathcal{S}) and the normalized consensus C\mathfrak{C} are guaranteed nonnegative or can be repaired (e.g., by an improved measure), including the regime where the remainder of the environment E/FF′\mathcal{E}_{/\mathcal{F}\mathcal{F}'} is too small or too slow to decohere SFF′\mathcal{S}\mathcal{F}\mathcal{F}'. (2) Interpretation: give an operational meaning to the sign of refined mutual information — when negative I\mathfrak{I} reflects spurious fragment–fragment correlations (classical shared noise, or entanglement generated by scrambling) that are unrelated to what the fragments know about S\mathcal{S}, and when it signals genuine absence of consensus about S\mathcal{S} or about a chosen observable Λ\Lambda. (3) Dynamics: describe quantitatively the rise and fall of these measures in interacting models such as the all-to-all spin Hamiltonian of the source (dependence on fragment size, observable tilt relative to the pointer basis, decoherence time, and relaxation/scrambling time), and either prove the source's numerical observation that I\mathfrak{I} and C\mathfrak{C} remain effective indicators of consensus as long as decoherence is effective, or determine the exact conditions on decoherence strength and fragment size under which I\mathfrak{I} and C\mathfrak{C} cease to indicate consensus and state the corresponding corrected criterion. A resolution must address all three deliverables: each is independently refutable, and a partial result on any one of them is not by itself a resolution of the task. An answer may additionally propose and justify a modified consensus measure, provided it reduces to I\mathfrak{I}/C\mathfrak{C} behavior on branching states and its adequacy is demonstrated in the non-branching regimes above.

References

  1. A. Touil, B. Yan, W. H. Zurek, "Consensus About Classical Reality in a Quantum Universe", arXiv:2503.14791 (2025), DOI 10.48550/arXiv.2503.14791, https://arxiv.org/abs/2503.14791 (primary source)
  2. A. Touil, F. Anza, S. Deffner, J. P. Crutchfield, "Branching States as The Emergent Structure of a Quantum Universe", Quantum 8, 1494 (2024), DOI 10.22331/q-2024-10-10-1494, https://quantum-journal.org/papers/q-2024-10-10-1494/
  3. D. A. Chisholm, L. Innocenti, G. M. Palma, "The meaning of redundancy and consensus in quantum objectivity", Quantum 7, 1074 (2023), DOI 10.22331/q-2023-08-03-1074, https://quantum-journal.org/papers/q-2023-08-03-1074/
  4. A. Touil, B. Yan, D. Girolami, S. Deffner, W. H. Zurek, "Eavesdropping on the Decohering Environment: Quantum Darwinism, Amplification, and the Origin of Objective Classical Reality", Phys. Rev. Lett. 128, 010401 (2022), DOI 10.1103/PhysRevLett.128.010401, https://doi.org/10.1103/PhysRevLett.128.010401
  5. J. Gu, "Hamiltonian thresholds for objective records", Phys. Rev. A 114, L020201 (2026), DOI 10.1103/fm9j-m9lh, arXiv:2608.00365, https://arxiv.org/abs/2608.00365
  6. A. Kiely, D. A. Chisholm, A. Touil, S. Deffner, G. Landi, S. Campbell, "Metrological approach to the emergence of classical objectivity", Phys. Rev. A 113, 022403 (2026), DOI 10.1103/hn78-7xx3, arXiv:2510.12313, https://arxiv.org/abs/2510.12313
  7. B. Ferté, D. Farci, X. Cao, "Decoherent Histories with(out) Objectivity in a (Broken) Apparatus", Phys. Rev. Lett. 136, 090404 (2026), DOI 10.1103/8qzx-xpfz, arXiv:2508.16482, https://arxiv.org/abs/2508.16482
  8. E. Doucet, S. Deffner, "From compatibility of measurements to exploring Quantum Darwinism on NISQ", Quantum Sci. Technol. 11, 035008 (2026), DOI 10.1088/2058-9565/ae753f, arXiv:2601.05350, https://arxiv.org/abs/2601.05350
  9. S.-I. Tang, E. Doucet, A. Touil, S. Deffner, A. Sone, "Information processing in quantum thermodynamic systems: An autonomous Hamiltonian approach", Phys. Rev. E 114, 024144 (2026), DOI 10.1103/rwkz-dc5z, arXiv:2511.08858, https://arxiv.org/abs/2511.08858
  10. H. Arora, B. K. Das, B. Suri, V. Madhok, "Quantum-to-classical transition and the emergence of trajectory level Darwinism with measurements distributed in time: a path integral approach", arXiv:2505.16889 (2025), https://arxiv.org/abs/2505.16889
  11. X. Cao, Z. Nussinov, "Redundancy from subsystem thermalization", arXiv:2603.15743 (2026), https://arxiv.org/abs/2603.15743