How can quantum Darwinism and decoherent histories be unified?

Two record paradigms that never implied each other

Decoherent histories and quantum Darwinism both ask which events in an isolated quantum system leave records, but they ask it in different ways. Decoherent histories uses the decoherent-histories condition to decide which past events can be unambiguously discriminated in the final wave function; it identifies formal records without asking how observers access them. Quantum Darwinism instead describes a mechanism by which information about preferred pointer states of an open system is copied into many environment fragments, so that the current state becomes redundantly encoded.

The paper states plainly that neither framework implies the other, and that connections between them have been scarcely investigated. That gap is the problem the paper addresses: a record, by the ordinary meaning of the word, refers to something that happened earlier, yet the standard definitions of redundancy are written as single-time quantities describing spacelike correlations in a state. Once the global state at time t is given, those definitions can be evaluated without any reference to dynamics or to two-time correlations.

Records point to the past, not the present

The paper's motivating observation is that records are used to infer what happened, not what is happening. The cosmic microwave background carries a fingerprint of the Universe during the recombination epoch around 13.5 billion years ago, even though the Universe no longer looks that way. Dinosaur fossils testify that dinosaurs existed, not that they exist now. The paper extends the same logic to quantum experiments, whose outcomes refer to states of quantum systems in the past, and to seeing an object now, which actually reveals its state a time d/c ago for distance d and speed of light c.

This matters because the three reviewed definitions of quantum Darwinism, including the redundancy plateau of quantum mutual information, the spectrum broadcast structure, and the branching-form wave function, do not imply the formation of records about the past even when evaluated sequentially at multiple times. A generalized CNOT-fan-out circuit does create past redundancy, but rather trivially. The interesting claim is the converse direction: a larger class of situations can generate past redundancy without quantum Darwinism.

What past redundancy adds to standard definitions

Past redundancy is defined with respect to event and record random variables rather than reduced density matrices. That choice removes two assumptions that standard treatments carry: the definition does not rely on a system-environment tensor product structure, and it does not rely on conditional independence of the kind built into the spectrum broadcast structure or the branching form. Instead it directly asks for classically accessible records of past events, and it can treat multiple events at multiple times.

The framework keeps three features the paper identifies in quantum Darwinism. Individual records need not be ideal and may reveal little about the event, because what matters is an extensive number of redundant records. A redundancy plateau in the classical mutual information gives the notion an information-theoretic footing without the problem that quantum mutual information can contain correlations that are practically inaccessible. And where quantum Darwinism is linked to the pointer basis produced by environmentally induced decoherence, past redundancy is linked to the existence of decoherent histories.

Fidelity links redundancy to decoherence of histories

The technical bridge between the two paradigms is the Bhattacharyya coefficient, also called fidelity. The paper connects redundancy to decoherence of histories and to the probability sum rule through this quantity, which lets the same object measure how distinguishable two history branches remain and how much redundant information fragments carry.

The scattered-photon calculation shows how the pieces fit. For a fragment F of n modes, an optimal discrimination strategy can be implemented with a passive optical network of beam splitters and phase shifters, followed by a Helstrom measurement, and it does not require a globally optimized measurement on the fragment; a local operations and classical communication protocol suffices. The success probability for distinguishing the two coherent-state branches is (1 + sqrt(1 - e^{-4||lambda_F||^2}))/2, and the classical mutual information plotted in the paper is built from these conditional probabilities. The decoherence between the two history branches, after dividing out the probabilistic prefactors, falls off as exp(-2|alpha(t)|^2 - 2||lambda_F'||^2). Because the quadratic rotating-wave Hamiltonian conserves the total squared coherent-state amplitude, a fragment of size n carries on average ||lambda_F(t)||^2 = (n/N)|alpha(0)|^2 at long times when alpha(t) is approximately zero. Redundancy and decoherence are therefore read off the same amplitude budget rather than treated as separate phenomena.

Redundancy without relentless cloning of pointer states

The paper separates two mechanisms that can generate past redundancy. One is the quantum-Darwinism mechanism that Zurek summarized as the environment relentlessly cloning information about the einselected pointer states. That mechanism also produces past redundancy, but it is not the only route, and it depends on a well-defined pointer basis while potentially producing fragile records that formally fill the redundancy plateau while remaining practically undetectable.

The other route runs through macroevents. Photons in the cosmic microwave background were released during electron-proton recombination about 13.5 billion years ago, and no relentless cloning of photons was involved; energy conservation forbids it in general, and conservation laws put severe restrictions on redundant cloning. Dinosaur bones encode biological information about an extinct species unrelated to any pointer basis, and the paper cites a fossil recovery rate for Tyrannosaurus rex between 1 per 16,000 and 1 per 80,000,000 individuals. Destructive quantum measurements such as photodetectors work through a metastable cascade: absorption of a single photon triggers a macroscopic electric signal, no photon remains at the end, the quantum mutual information between photon and environment is zero for practically all times, and the pointer basis plays no role. The paper also notes that generating redundancy and decoherent histories in its macroscopic-superposition example requires a large coherent-state amplitude alpha(0).

Why fine-tuning made redundancy look rare

The paper argues that past redundancy is a more widespread phenomenon than the fine-tuned models required for quantum Darwinism suggest. Perfect quantum Darwinism imposes strong constraints: for qubit system and environment, it essentially enforces an almost unique operator structure on the system-environment Hamiltonian, and relaxing the qubit assumption still entails an extensive number of commutator constraints. The generalized CNOT-fan-out unitary is popular in measurement theory and quantum foundations, but very few real, non-engineered physical systems obey it; the paper notes that one reason Serge Haroche received the Nobel prize was that his lab could engineer a quantum non-demolition measurement obeying that form, with a microwave cavity as system and atoms flying through it as fragments, and even there deviations are relevant.

Freeing records from the requirement that they be redundantly correlated with the current system state widens the class of models that generate redundancy. The paper reports that many models supporting only temporal quantum Darwinism support eternal past redundancy, and that thermalization is no longer opposed to the emergence of redundancy. This is the paper's answer to the question that motivated the project: how generic is redundancy? The answer is framed as a step forward rather than a complete classification, since the class of models generating imperfect, approximate quantum Darwinism is stated to be currently unknown.

Redundancy is necessary but not sufficient for objectivity

The richer picture comes with a sharper claim about what redundancy buys. The paper emphasizes that redundancy is necessary but not sufficient for objectivity. Redundancy can arise from mechanisms other than quantum Darwinism, including macroevents and cascade structures, so observing a redundancy plateau does not by itself establish that pointer-state cloning produced it, nor that the records are robust or accessible in practice.

The scattered-photon case is where the paper presses this point. Photons are an ideal candidate for quantum Darwinism because they naturally form a non-interacting environment, yet the paper closes with a critical discussion of that case rather than treating it as the default. The practical priority that follows is to ask which mechanism generated a given redundancy and whether the records are stable and classically accessible, not merely whether a plateau appears. The paper points toward metastable quantum structures and quantum cascade systems as directions for understanding macroevents and redundancy, a direction it says has received little attention from the community working on foundational questions.

Sources used in this answer