Physicists affiliated with Los Alamos National Laboratory, the Joint Quantum Institute, NIST, and the University of Maryland have proposed a feedback-driven quantum engine that can statistically reverse the arrow of time in microscopic systems. Published in Physical Review X in February 2026, the theoretical framework treats the disturbance caused by quantum measurement not as unwanted noise but as a thermodynamic resource (one that can be harnessed, redirected, and even run backward). The result challenges a bedrock assumption, that measurement always pushes quantum systems irreversibly forward in time.
Measurement as a Thermodynamic Engine
Most popular accounts of quantum physics treat measurement as a simple readout. But continuously monitoring a quantum system does something far more physical: it injects energy and entropy into the system with every interaction. A 2013 experiment on weak back-action in a superconducting qubit showed this directly, demonstrating that even a gentle observation disturbs the system in a measurable, tunable way. That disturbance is not a bug. According to the new paper, it is the fuel for a kind of engine.
The core proposal, detailed in the engine framework, introduces explicit quantum-control constructions that replicate the stochastic trajectories a continuously monitored system would follow. More strikingly, the authors design a feedback loop that can reverse the effect of monitoring, generating trajectories that are statistically consistent with a reversed arrow of time. In plain terms, the engine does not literally rewind a clock or reconstruct a broken egg. Instead, it produces measurement records whose statistical fingerprints look identical to what you would see if time ran backward.
Technically, the protocol works by interleaving short periods of continuous measurement with carefully chosen control operations that depend on the entire measurement history. The feedback law is tuned so that, when you compare the probability of any observed path with the probability of its time-reversed counterpart, the ratio is inverted relative to ordinary monitoring. Where standard measurement drives entropy production, the feedback-assisted dynamics can effectively consume previously generated entropy, at least within the small, controlled quantum system.
What the Arrow of Time Means at Quantum Scales
The concept of time’s arrow in quantum measurement has a precise technical meaning, rooted in a 2008 theoretical framework by Crooks and Feng. Their Physical Review Letters paper defined the length of the arrow using likelihood ratios that compare the probability of a process running forward against the probability of the same process running in reverse. When those probabilities are equal, time has no preferred direction. When they diverge sharply, irreversibility dominates.
In that language, the “length” is essentially a log-likelihood: how strongly a given record of events points to one temporal direction over the other. For macroscopic processes such as mixing cream into coffee, this quantity is enormous, reflecting the overwhelming improbability of spontaneous unmixing. For microscopic quantum systems undergoing continuous measurement, the length can be modest, fluctuating from run to run and even temporarily favoring the backward direction.
A 2019 experiment brought this idea into the lab. Researchers working with a superconducting qubit directly measured an arrow-of-time statistic by comparing forward and backward-in-time path probabilities during continuous quantum measurement. That experiment also confirmed consistency with both detailed and integral fluctuation theorems, mathematical constraints that govern how entropy is produced and exchanged in small systems far from equilibrium. The new feedback-engine proposal builds on exactly this statistical scaffolding, arguing that if you can quantify time’s arrow, you can also reshape it by engineering the measurement and control protocol.
From Qubits to Ultracold Atoms
One reasonable objection to any single-qubit result is that it might be a quirk of one platform. That concern weakened considerably when a separate team linked the measurement arrow to fluctuation relations in a many-body setting, using spin measurements of ultracold atoms. The fact that the same arrow-of-time statistics appear in both superconducting circuits and atomic ensembles suggests the phenomenon is general, not an artifact of a particular technology.
Earlier foundational work also matters here. A landmark experiment demonstrated real-time trajectories of individual quantum states under continuous measurement in superconducting qubits, proving that single-run stochastic paths could be tracked and reconstructed. Without that capability, the feedback constructions in the new paper would have no experimental basis, because the engine’s operation depends on knowing, and reacting to, the evolving state in real time.
Separately, quantum state diffusion trajectories were demonstrated via heterodyne detection, establishing measurement-unraveling tools for continuous monitoring and feedback contexts. Those techniques let theorists and experimentalists translate between abstract master equations and concrete streams of detector clicks or voltage traces. Together, these experiments form the technical backbone that makes the proposed engine physically plausible rather than purely mathematical, and they outline a path toward implementing similar feedback loops in future devices.
What the Engine Actually Does, and Does Not Do
A natural question is whether this engine violates the second law of thermodynamics. It does not. As reporting and prior theoretical work have emphasized, the second law remains intact when all sources of thermodynamic cost are accounted for. The feedback loop that reverses time’s arrow requires energy to operate, information to process, and a controller that itself generates entropy. The engine extracts useful work or order from measurement disturbances, but it pays for that work elsewhere in the system, much like Maxwell’s demon ultimately pays in information-processing costs.
The distinction matters because it reframes what “reversing time” means in this context. The engine does not create perpetual motion or free energy, nor does it make a macroscopic movie run backward. What it does is demonstrate that the irreversibility introduced by quantum measurement is not a fixed, immovable feature of reality. It is a controllable quantity. By tuning feedback protocols, the researchers show that the statistical arrow of time can be stretched, compressed, or flipped, all while respecting fundamental thermodynamic limits.
The Los Alamos release quoted a team member explaining that, unlike everyday phenomena, most microscopic laws of physics are indifferent to the direction of time. The engine exploits that symmetry: by combining time-reversal-invariant dynamics with measurement and feedback, it carves out a small domain in which apparent irreversibility can be dialed up or down rather than simply endured.
Why This Challenges Standard Coverage
Most coverage of feedback-based quantum devices has framed them as steps toward quantum batteries or nanoscale engines. That framing is not wrong, but it skips the deeper tension. The paper’s real contribution is conceptual: it treats measurement-induced irreversibility as a thermodynamic variable, akin to temperature or pressure, that can be engineered. In doing so, it blurs the line between “observer” and “machine,” recasting the act of looking as a potential source of work.
This perspective also highlights the role of information infrastructures that make such research possible. The preprint version of the engine proposal appears on community-supported archives, reflecting how large laboratories and universities collectively underwrite open access to cutting-edge results. That ecosystem accelerates the feedback loop between theory and experiment. Ideas about time’s arrow move from blackboard to arXiv to the lab and back again in a matter of months, not years.
Looking ahead, the most immediate impact is likely to be methodological rather than technological. By providing explicit control recipes for reversing time’s arrow in a statistical sense, the authors give experimentalists a target for testing the limits of fluctuation theorems, quantum trajectory reconstruction, and measurement-based feedback. If realized in hardware, such engines could become precision probes of how quantum information, energy, and entropy flow in regimes where classical thermodynamics offers only coarse averages.
At a broader level, the work underscores a shift in how physicists think about measurement. Instead of treating it as the end of a process (a final readout that collapses possibilities into facts), the feedback-engine view treats measurement as an active, tunable interaction that shapes the very direction of time experienced by a system. In that sense, the engine does not merely exploit quantum weirdness; it forces a reconsideration of what it means for time to “flow” at the smallest scales, and how much of that flow is written into nature versus engineered by the ways we choose to look.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
