Schr"odinger and Heisenberg non-Markovianity in quantum information tasks

arXiv:2606.30751 · quant-ph · Submitted 2026-06-29 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: I'm Kai, and with me are Mira and Lev, guest researcher.

Mira: Today's paper: "Schr"odinger and Heisenberg non-Markovianity in quantum information tasks".

Kai: Quantum non-Markovianity has been widely studied and connected to memory effects in open system dynamics,

Mira: First, who's behind it and why it matters.

Paper summary: Kai: So, we're diving into this paper, "Schr"odinger and Heisenberg non-Markovianity in quantum information tasks," which looks at how memory effects show up differently depending on whether you look at the Schrödinger or Heisenberg picture. Mira, can you give us the main idea of what the authors are trying to establish here?

Mira: Absolutely, Kai. The central thesis of this paper is that while quantum non-Markovianity is often discussed in one picture, this work investigates which specific type of memory—Schrödinger or Heisenberg—actually matters for different quantum information tasks. They distinguish between divisibility properties in the Schrödinger and Heisenberg pictures, essentially looking at how the dynamics behave when viewed through these two lenses.

Lev: That distinction between the pictures sounds like it could be really tricky when we try to translate this into actual hardware experiments; I'm curious if this theoretical separation translates cleanly to what we can measure on a real system.

Kai: Right, Lev, that's exactly what I wanted to ask. The paper sets up these concepts of Schrödinger CP divisible and Heisenberg CP divisible, and SHCPD versus SHnCPD, which essentially categorize the memory effects into four different types based on how the propagators behave.

Mira: Exactly. They link these divisibility properties directly to things like monotonicity in time for certain norms; for instance, SPD is equivalent to a monotonic decrease in the guessing probability between states seven, and HPD relates to monotonic behavior of guessing probability between effects eleven.

Lev: Monotonicity conditions are always the hard part when you're trying to move from theory to practice, because real noisy channels rarely exhibit perfect monotonicity. Does the paper provide any concrete examples of how these different divisibility types manifest in typical open system dynamics we see in experiments?

Kai: They do mention some explicit examples, like a process that is not CP-divisible in one picture while the propagator in the other describes unitary evolution thirty-five, which really highlights why having both pictures is necessary to fully describe the situation <ref:2606.30751#pg2,not CP-divisible in one picture while the propagator in the other>.

Mira: And they establish necessary conditions for non-Markovianity in both pictures by looking at dynamics described in just one picture, showing that a specific witness for Schrödinger non-Markovianity actually proves SHnCPD simultaneously three <ref:2606.30751#pg0>. This necessity stems from conditions related to monotonicity violations, like if a subset of states is preserved by local CPTP maps but the evolution takes it outside that set.

Lev: That sounds like a strong constraint on the dynamics; if those monotonicity violations are necessary for SHnCPD, it suggests we need very specific types of non-unitary evolution to get these memory effects we're looking for.

Kai: It seems they then provide stronger criteria too, like Proposition three which states that if the dynamics is SnPD and certain conditions on the maximal distance between states are met, it must be SHnPD three <ref:2606.30751#pg0>. This is a key piece of the puzzle for defining when both pictures are involved.

Paper summary: Mira: And they tie this together by introducing a combined witness for non-Markovianity in both pictures, specifically for SHnCPD, by combining the witnesses derived from SnPD and HnPD, which relies on SnPD and HnPD implying contractivity of the guessing probability between states or effects.

Lev: Combining those two different types of divisibility conditions into a single measure is ambitious; it would make testing this really challenging to verify experimentally, but theoretically that's where the real power lies.

Kai: This paper then moves into applying these findings to specific quantum information tasks, showing that different types of memory are required for revivals in task performance, which is where things get concrete.

Mira: For instance, when looking at capacity measures—the classical, entanglement-assisted classical, and quantum ones—they show they are contractive under SCPD dynamics twenty-eight, but the non-Markovianity measure reveals that achieving revivals actually requires both SnCPD and HnCPD dynamics simultaneously <ref:2606.30751#pg2>.

Lev: So, for an error correction scenario, this implies that if we want to achieve a certain performance revival, we can't just rely on one type of memory effect; we need the structure to be non-Markovian in both pictures at the same time.

Kai: They also look at channel distinguishability, showing that postprocessing with SCPD dynamics reduces distinguishability between channels, while preprocessing with an HCPD dynamics also diminishes it, but not if the map is SCPD but HnCPD under postprocessing Example four <ref:2606.30751#pg0>.

Mira: And they extend this analysis into classical stochastic dynamics, setting up a rigorous analogy between Schrödinger and Heisenberg divisibility in the classical world. This suggests that the mathematical framework isn't just quantum-specific.

Lev: That classical analogy is important; if we can map these structural properties to a classical channel, it gives us a better idea of what kind of memory structure we are looking for when designing systems, even if we can't build the exact quantum state dynamics yet.

Kai: The overall implication here is that understanding non-Markovianity requires considering the dual nature of descriptions in both pictures, rather than just picking one convenient way to look at the evolution.

Mira: Precisely. The paper "Schr"odinger and Heisenberg non-Markovianity in quantum information tasks" moves beyond simply identifying memory effects and starts classifying *which* memory structures are required for specific computational goals.

Lev: For error correction, this suggests that designing error correcting codes needs to account for the dual nature of how we model the system's evolution to ensure we capture all necessary non-Markovian features.

Kai: It really underscores that when building experimental setups, we have to be careful about whether we are measuring state evolution or operator evolution, because the memory effects can look completely different in each case.

Mira: And for anyone working on open quantum systems, this paper provides a framework for precisely identifying the kind of memory needed by demanding non-Markovianity in both pictures simultaneously when certain information tasks are being performed.

Conclusion: Kai: So, to wrap up this discussion, we're looking at the paper "Schr"odinger and Heisenberg non-Markovianity in quantum information tasks," which really digs into how memory effects show up differently depending on whether you look at the Schrödinger or Heisenberg picture.

Mira: Exactly, Kai. The authors are mapping out which specific type of memory—Schrödinger versus Heisenberg—is actually relevant for different quantum information tasks by looking at divisibility properties in both pictures.

Lev: From a theoretical standpoint, that means they're trying to figure out if the dynamics need memory effects in just one picture or if they have to be non-Markovian in both simultaneously for certain goals.

Kai: Right, and the results show that some tasks only require memory effects in one picture, while others demand non-Markovianity from both pictures at the same time.

Mira: That distinction is crucial because it tells us exactly what kind of dynamical structure we need to engineer into a system to achieve a desired outcome in quantum computation or information processing.

Lev: If this holds up under real-world conditions, it gives us a much sharper tool for designing error correction codes, telling us precisely what kind of noise correlation we need to anticipate.

Kai: It really shifts the focus from just saying "it's non-Markovian" to specifying *which* picture and *which* type of non-Markovianity is required for the specific task at hand.

Mira: And if the authors' findings on capacity measures and channel distinguishability are correct, it means our current models for how information flows through noisy channels might be missing some structural requirements.

Lev: That suggests that whatever real hardware we build, we can't just pick a general non-Markovian model; we have to target the specific dual-picture structure they identify.

Kai: So, the big picture here is that this paper gives us a much more nuanced roadmap for understanding and engineering quantum memory effects.

Mira: It frames the entire field of open system dynamics by showing how the choice of mathematical representation directly impacts what physical memory structure is needed for a given task.

Lev: It means for researchers in error correction, the path forward isn't just finding *a* non-Markovian process, but finding one that satisfies these dual requirements derived from the Schrödinger and Heisenberg pictures.

Kai: We'll be looking closely at how this theoretical structure translates into measurable quantities in our next segment.

Mira: That’s what we’ll do next, connecting these abstract divisibility concepts to the concrete dynamics we observe in the lab.

Department of Physics and Astronomy, University of Turku · Dipartimento di Fisica “Aldo Pontremoli”, Universit`a degli Studi di Milano, Istituto Nazionale di Fisica Nucleare, Sezione di Milano, Institute of Physics, Faculty of Physics, Astronomy and Informatics, Nicolaus Copernicus University

quant-ph

Submitted: 2026-06-29

Updated: 2026-10-07

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 79/100

The gist: Quantum non-Markovianity has been widely studied and connected to memory effects in open system dynamics, but this work investigates which specific type of memory—Schrödinger or Heisenberg

Key concepts

Schrödinger CP divisible (SCPD)
This property relates to the evolution of quantum states described by a master equation. A map is SCPD if its propagator is completely positive (CP). This condition is linked to the monotonic decrease in guessing probability between states, which implies non-monotonic behavior for certain state norms over time.
Heisenberg CP divisible (HCPD)
This property concerns the evolution of quantum operators in the Heisenberg picture. A map is HCPD if its propagator is CP. This condition is equivalent to a monotonic behavior in guessing probability between effects, suggesting that arbitrary incompatibility monotone functions are contractive under completely positive unital maps.
SHnCPD
This signifies non-Markovianity in both the Schrödinger and Heisenberg pictures simultaneously. It arises from conditions related to monotonicity violations, such as when a subset of states is preserved but evolution leads outside that set. It is a stronger condition than requiring non-Markovianity in just one picture.

Terminology

Summary

Quantum non-Markovianity has been widely studied and connected to memory effects in open system dynamics, but this work investigates which specific type of memory—Schrödinger or Heisenberg picture—is relevant for different quantum information tasks. The core finding is that some tasks require memory effects only in one picture, while others necessitate non-Markovianity in both pictures simultaneously.

How it works

The paper distinguishes between non-Markovianity defined through divisibility properties of the dynamical map in the Schrödinger and Heisenberg pictures. The dynamics are described by a completely positive and trace preserving (CPTP) map, denoted as a CPTP map, in the Schrödinger picture where states evolve according to the master equation. Alternatively, one can use the Heisenberg picture where operators evolve according to a different master equation.

The paper introduces several concepts related to divisibility:

  1. Schrödinger CP divisible (SCPD) if the propagator is CP, and Schrödinger non-CP divisible (SnCPD) otherwise.

  2. Heisenberg CP divisible (HCPD) if the propagator is CP, and Heisenberg non-CP divisible (HnCPD) otherwise.

  3. SHCPD if both propagators are CP, and SHnCPD if none of the two is CP.

These divisibility properties are linked to monotonicity in time of suitable norms:

(7)

SPD is equivalent to a monotonic decrease in time of the guessing probability between states, which is equivalent to the non-monotonic behavior of the trace norm for at least a pair of states.

(11)

HPD is equivalent to a monotonic behavior of the guessing probability between effects, which is equivalent to an arbitrary incompatibility monotone being contractive under CPU maps.

How it works

The paper establishes necessary conditions for non-Markovianity in both pictures by considering dynamics in just one picture. It shows that the previously considered witness of Schrödinger non-Markovianity in terms of the volume of accessible states does indeed witness non-Markovianity in both pictures at the same time, and this is shown to be a witness for SHnCPD.

The necessity of SHnCPD arises from conditions related to monotonicity violations:

  1. Proposition 1 concerns properties of quantum states: If a subset of states is preserved by local CPTP maps, and there exists a state where the evolution leads outside that set, then the dynamics must be SHnCPD.

  2. Proposition 2 concerns monotone functionals: If a functional is monotonic under CPTP maps, and its maximum value increases over time, then the dynamics must be SHnCPD.

Furthermore, stronger criteria are provided:

(3)

Proposition 3 states that if the dynamics is SnPD and certain conditions on the maximal distance between states are met, then it must be SHnPD. Analogous conditions hold for HnPD based on the maximal operator norm of differences between effects.

How it works

The paper provides a combined witness for non-Markovianity in both pictures, specifically for SHnCPD, by combining the witnesses derived from SnPD and HnPD. This witness relies on the fact that SnPD and HnPD imply contractivity of the guessing probability between states or effects.

The key measure introduced is:

(28)

The norm of the channel is defined as:

∥Φ∥1→1 B

max

∆∈S1 max E∈S∞ tr Φ[∆] E, where S1 is the unit sphere in the trace norm and S∞ is the set of effects.

(29)

If this norm is non-monotonic in time, i.e., if ∥Φt∥1→1 > ∥Φs∥1→1 for t > s, then the dynamics must be SHnCPD. This result holds for both SCPD and HCPD dynamics.

How it works

The paper applies these findings to relevant quantum information tasks, demonstrating that different types of memory are required for revivals in task performance.

The analysis covers:

  1. Capacity: The capacity measures (classical, entanglement-assisted classical, and quantum) are contractive under CP maps (SCPD), and the non-Markovianity measure shows that revivals require both SnCPD and HnCPD dynamics.

  2. Channel Distinguishability: Postprocessing with a SCPD dynamics diminishes distinguishability between channels, while preprocessing with an HCPD dynamics also diminishes it, but not under postprocessing if the map is SCPD but HnCPD (Example 4).

How it works

The analysis is extended to classical stochastic dynamics, establishing a rigorous analogy between Schrödinger and Heisenberg divisibility in the quantum and classical settings.

For classical processes:

Improvements for AI systems

Based on the provided scientific paper, here are the specific improvements that could be made to AI systems, categorized by their application domain:

) Improved AI System Capabilities:

The paper provides a mathematical framework for understanding and characterizing non-Markovianity in both quantum and classical stochastic dynamics across two equivalent but distinct pictures (Schrödinger and Heisenberg). This knowledge can be leveraged to build AI systems with enhanced memory, better predictive capabilities in complex environments, and more robust decision-making under uncertainty.

Here are specific improvements:

  1. --- Enhanced Predictive Modeling for Open Quantum Systems (Quantum AI):

  2. --- Improved Channel Capacity Estimation and Optimization (Quantum Communications/AI):

  3. --- Robust State Discrimination and Classification (Quantum Machine Learning):

  4. --- Novel Witness-Based Anomaly Detection in Dynamic Systems (System Diagnostics/Control Theory).

) Specific Improvements:

  1. --- Enhanced Predictive Modeling for Open Quantum Systems (Quantum AI):

  2. This improvement involves integrating non-Markovianity analysis into the core of quantum AI models designed to simulate or control open quantum systems (e.g., those modeling molecular dynamics, quantum chemistry, or trapped ion systems).

  3. The system can now explicitly model and predict the memory effects in system evolution. Instead of assuming Markovian decay (which is often inaccurate), the AI can utilize the divisibility properties identified in Section II and III to determine if a specific quantum task (like entanglement revival) requires memory from both the Schrödinger and Heisenberg pictures.

  4. The resulting AI model will be superior at predicting long-term system behavior where information backflow occurs, allowing for more accurate simulations of processes that exhibit temporal correlations, such as decoherence in realistic noisy environments.

  5. --- Improved Channel Capacity Estimation and Optimization (Quantum Communications/AI):

  6. This improvement focuses on developing AI agents that optimize communication protocols through quantum channels (e.g., in quantum networks or secure key distribution). The system can now distinguish between tasks requiring memory only in one picture versus those requiring memory in both, as detailed in Section IV.

  7. The AI system can dynamically select the optimal processing strategy—whether to use post-processing (sensitive to Schrödinger non-Markovianity) or preprocessing (sensitive to Heisenberg non-Markovianity)—to maximize channel capacity metrics like Classical Capacity, Entanglement-Assisted Classical Capacity, or Quantum Capacity. This leads to optimized protocols that adapt their information flow based on the current dynamical regime.

  8. --- Robust State Discrimination and Classification (Quantum Machine Learning):

  9. This improvement applies the concepts of divisibility (SCPD vs. SnCPD) to quantum machine learning tasks, such as classifying quantum states or identifying specific features in noisy data representations.

  10. The AI can be designed to detect when its input data is evolving under dynamics that are not completely positive (non-CP), which is a signature of non-Markovianity. This allows the system to flag anomalous or highly correlated data points that indicate memory effects, leading to more reliable classification in noisy quantum datasets.

  11. --- Novel Witness-Based Anomaly Detection in Dynamic Systems (System Diagnostics/Control Theory):

  12. This improvement involves creating diagnostic tools for complex physical or classical stochastic systems (e.g., financial markets, sensor networks, or chemical reactions) that exhibit memory effects. The system can use the derived witnesses:

  13. The Volume of Accessible States witness (Eqs. 32-34) to detect non-Markovianity in both pictures simultaneously, providing a highly sensitive indicator of dynamic memory effects that traditional Markovian models miss.

  14. The Channel Distinguishability witness (Eqs. 51-61) for classical channels can be used to monitor the stability or predictability of stochastic processes, allowing AI controllers to detect when the underlying process dynamics shift from a predictable (Markovian) regime to a memory-dependent one, triggering adaptive control measures.

Sources

Related papers