On the emergence of quantum memory in non-Markovian dynamics

Alexander Yosifov, Aditya Iyer, Vlatko Vedral, Jinzhao Sun · 2025

Read On the emergence of quantum memory in non-Markovian dynamics by Alexander Yosifov, Aditya Iyer, Vlatko Vedral, Jinzhao Sun (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-2507-21907.txt.

Introduction

Alexander Yosifov and co-authors ask when the memory inside non-Markovian quantum dynamics must itself be quantum. Open quantum systems often evolve with memory, since their surroundings feed past states back in. That memory could be plain classical bookkeeping, or it could ride on quantum links.

The team studies a minimal model called the quantum homogenizer. With entanglement gauges and math proofs backed by simulation, they show the memory type turns on the reservoir's starting entanglement. How non-classical links spread inside the reservoir decides the matter, not the dynamics alone. For split reservoir starts, interaction strength maps onto entanglement growth in a clean way. The authors give a threshold rule for when quantum memory switches on. The rules guide builders of quantum devices that live with noisy surroundings.

Catalog record

CREATORAlexander Yosifov, Aditya Iyer, Vlatko Vedral, Jinzhao Sun
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-21907

Browse the full library · About the Tower of Babel

Full text

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

On the emergence of quantum memory in non-Markovian dynamics Preprint: APS/123-QED Alexander Yosifov, Aditya Iyer, Vlatko Vedral, Jinzhao Sun Abstract Quantum systems are often hindered by decoherence due to impact from the environment. While memoryless Markovian collision models are commonly used to approximate such evolution, non-Markovian dynamics (with memory) is typical in practice, with memory effects being harnessed as a resource for many tasks like quantum error correction and information processing. Yet, the type of memory—classical or quantum—necessary to realize the dynamics of many collision models is not known. In this work, we extend the quantum homogenizer to the non-Markovian regime by introducing intra-ancilla interactions mediated by Fredkin gates, and study the nature of its memory. Using entanglement measures and relying only on the local dynamics as a witness, we prove the model can be realized with either classical or quantum memory, depending on the initialization of the reservoir and the propagation of non-classical correlations within the reservoir. We further explore how quantum memory emerges across a wide range of practical scenarios. The results shed light on the origin of memory in open quantum systems and can advent the design of near-term quantum technologies for a variety of applications. Open quantum systems tend to decohere and lose information, which is a major roadblock to the scalability of quantum hardware 1 ; 2 ; 3 ; 4 . While usually captured by a Lindbladian master equation 5 , collision models (CMs), such as quantum homogenization maria ; ziman ; vlatko , offer an efficient way to study the dynamics of such systems by ensuring completely positive trace-preserving (CPTP) evolution without active control; with applications to engineered dissipation alexander2 ; steady , quantum control alexander ; ieee , and state stabilization stabilization ; stabilization2 . Under the Born-Markov approximation, such CMs are typically Markovian, assuming no backreaction from the bath after interacting with the system and large separation between their characteristic time scales NM . Most realistic scenarios, however, are non-Markovian NM2 , endowing the reservoir with memory that has been proposed as a resource for improving quantum control protocols in solid state and NMR devices qcontrol1 ; qcontrol2 ; qcontrol3 , quantum communication qcomm , and mitigation of off-resonant gate errors NMhelp . Still, to make full use of it, it is crucial to determine the nature of the memory (classical or quantum) and its relation to the initialization of the reservoir and its entanglement structure, as the distinction significantly impacts the design, implementation, and performance of noisy hardware memory1 ; memory2 ; memory3 ; memory4 . Establishing clear conditions to determine the type of memory necessary to realize certain non-Markovian dynamics can thereby enable the use of non-Markovian CMs as novel scheme for tasks like continuous quantum error correction oo , coherence retention in superconducting processors coherence , quantum teleportation qteleport , quantum key distribution qkd , and quantum feedback control control . Surprisingly, despite being essential for the advent of near-term devices (especially for solid-state hardware), the nature of the memory, and its intrinsic sensitivity to the entanglement structure of the environment, are not known for the homogenizer. In this work, we answer the question by studying the dynamics of a novel non-Markovian extension of the homogenizer, where we can on demand control the memory effect. The homogenizer offers a particularly relevant and physically grounded model for exploring memory effects because it directly mirrors interactions found in many solid-state and superconducting architectures, such as exchange-based gates or SWAP -type couplings. Moreover, its structure enables precise control over the system–reservoir coupling and timing, making it ideally suited to probe the boundary between Markovian and non-Markovian dynamics. By introducing controlled intra-reservoir interactions into this model, we gain access to a tunable, analytically transparent platform for studying how memory (classical or quantum) arises and propagates in open-system dynamics. As a witness to the nature of the memory, we use a new criterion memory that relies on the local dynamics alone. By considering the concurrence as an entanglement measure, we examine the conditions under which the non-Markovian dynamics can be realized quantumly. We show that whether true quantum memory is needed depends not only on the dynamics itself, but also on the initialization of the reservoir, i.e., its entanglement structure. Our analysis covers practical scenarios in which the reservoir contains initial asymmetric entanglement —common in many realistic noise models, see, e.g., asymm1 ; asymm2 ; asymm3 . In those cases, we show how the physical realization becomes nontrivial, and why genuine quantum memory is essential. Our findings reveal insight into the nature of memory in non-Markovian CMs and can thus inform the development of advanced quantum memory designs as well as quantum error correction protocols for quantum devices. Setup. —Following recent physical insight about quantum trajectories ziman2 ; physical and the role of quantum interference in CMs vlatko2 , the intrinsic memory effects in our model emerge dynamically . Like the standard homogenizer alexander ; alexander2 , the system-bath interactions are described by a series of two-qubit partial- SWAP operations which implement probabilistic exchange of states between the system and bath ancillas. Yet, unlike the classical protocol, here we restrict the partial- SWAP to couple the system to only odd bath ancillas, where assuming the ancillas are grouped in even-odd pairs, this corresponds to the Stinespring dilation dilation of the quantum operation. To engineer the non-Markovianity, on the other hand, the ancilla-ancilla interactions are mediated by controlled- SWAP (Fredkin gates). This way, the embedded quantum information propagates across the bath, driven by the controlled- SWAP , such that the state of still unperturbed ancilla depends on the past bipartite interactions. The use of a controlled- SWAP gate to mediate reservoir interactions simulates a conditional memory mechanism. Such dynamics are relevant for understanding correlated noise and may provide a useful abstraction for studying non-Markovian effects in quantum error correction. Although this interaction does not generalize to all possible non-Markovian dynamics, especially those involving adaptive control, measurements, or non-unitary feedback, its structure offers a useful abstraction for studying memory effects relevant to quantum error correction and system-environment correlations. Quantum homogenizer. —Intuitively, the homogenizer (see maria ; ziman ; vlatko ; alexander2 ) is a quantum machine composed of a set of N qubits in state \xi^{\otimes(k)} \forall k\in N , i.e., a reservoir, denoted as {R} . Where at the k th timestep, the system interacts with the reservoir state \xi^{(k)}=\xi . Generally, a qubit S , initialized in some arbitrary state \rho_{S} , interacts with the reservoir qubits (one at a time) via the unitary partial- SWAP \mathcal{U}_{{S},{k}}=\text{cos}\left(\eta\right)\mathbb{1}_{{S}{k}}+i\,\text{sin}\left(\eta\right)\mathbb{S}_{{S}{k}}, (1) where \mathbb{S}_{{S}{k}} is the SWAP transformation between S and the k th qubit in R (i.e. the k th timesteps), and \mathbb{1}_{{S}{k}} is the identity; with \mathbb{S}\ket{\phi}\ket{\psi}=\ket{\psi}\ket{\phi},\forall\ket{\phi},\ket{\psi} , and \mathbb{1}:\ket{\psi}\rightarrow\ket{\psi} . Here, \eta_{k}=\eta\in\left[0,\pi/2\right] denotes the coupling strength in the k th step, and is drawn from a Gaussian distribution with mean value \eta : the closer it is to \pi/2 , the higher the probability is that the channel \mathcal{U}_{{S},{k}} will act as the SWAP operation, Fig. 1 . Figure 1: The Markovian setup. (a) Schematic of the homogenizer, where a qubit S in state \rho_{S} (green square) interacts sequentially ( k=1,...,N ) with the

Read the complete text