On October 9, 2026, *Nature Physics* published a paper titled *Order from chaos with adaptive circuits on quantum hardware*. A team from Rutgers and IBM, together with City College of New York, Penn State, Iowa State, and the Flatiron Institute, used up to 100 qubits on IBM's 156-qubit Heron processor to do one thing: scramble information across qubits while repeatedly measuring and resetting to push the scrambling back.
Each of the two operations was applied nearly 5,000 times.
What scrambled it, and what held it down
The scrambling came from a quantum version of the Bernoulli map — the famous chaotic card-shuffling transformation of classical dynamics. Each iteration folds information into neighboring qubits, progressively destroying any initial ordering. The experiment ran on a torus, with a periodically driven circuit applying conditioned single-qubit rotations (momentum flips). At the end of each period, measurement in the computational basis produced a random field.
What held it down was mid-circuit local measurement and reset, resetting onto a designated quantum state entangled with the local qubits. Measurement and reset were inserted into the circuit as non-unitary evolution with conditional feedback.
The competition between the two constitutes a dynamical phase transition.
| Item | Value | |---|---| | Processor | IBM 156-qubit Heron | | Qubits used | Up to 100 | | Entangling gate operations | ~5,000 | | Non-unitary mid-circuit operations | ~5,000 | | Paper length | pp. 1–9 | | Data availability | No public repository; request from corresponding authors |
The abrupt change near 50/50
The transition occurs where the two operations' frequencies are roughly equal. When scrambling dominates, the system stays chaotic; when measurement-and-reset is more frequent, it is driven to the selected fixed-point ordered state. Around 50/50, a small change in the relative frequency flips the entire system's behavior.
The reported order parameter is the quantum fluctuation of magnetization, quantified via KL divergence, with a finite mean-field critical description: when perturbations are uniformly suppressed, the response's derivative with respect to the control rate diverges with system size — the signature of a finite-temperature transition in a finite-size system.
The experiment verified the theory and measurement-condition-dependent transition on the largest classically solvable systems, then extrapolated beyond classical simulation scale. The abstract states that the quantum processor gave accurate estimates of universal critical properties.
The headline large-scale result: the transition sharpens on larger systems — collective behavior strengthens with size.
How far is this from fault tolerance?
The paper does not claim to have built a fault-tolerant quantum computer, and that claim cannot yet be made. Its conclusion is that controlling quantum chaos requires measurement and feedback that survive noise and decoherence — now demonstrated on a superconducting processor, revealing a critical transition between quantum and classical dynamics.
Maika Takita, IBM's principal research scientist for quantum error correction and dynamic circuits experiments, and Jedediah Pixley, Rutgers professor of physics and Flatiron Institute member, put it concretely: building a fault-tolerant quantum computer requires performing many such checks and corrections within a single computation, not just one or two. This experiment is a milestone marker on that road.
Co-author Justin Wilson, associate professor at Louisiana State University, offered an analogy: chaos is the butterfly effect — small changes producing unpredictable large consequences — and this unruly behavior can be suppressed through random interventions, with a clearly defined intervention threshold. As system size grows, the transition approaches the universal critical exponents predicted by equilibrium fluctuations.
The theory rests on three supports: noisy simulations, matrix product states, and mappings to statistical mechanics models used to identify the transition's universality class. The authors referenced work such as Krbálek and Šeba on urban traffic flow and random-matrix statistics to find correspondences for universal exponents.
Two caveats: the experimental datasets and numerical simulations were too large to deposit in a public repository and can only be requested from the corresponding authors; the code likewise requires a request.
--- Source: *Nature Physics*, "Order from chaos with adaptive circuits on quantum hardware," online 2026-10-09, by Bibek Pokharel, Haining Pan, Kemal Aziz, Luke C. G. Govia, Sriram Ganeshan, Thomas Iadecola, Justin H. Wilson, Barbara A. Jones, Abhinav Deshpande, Jedediah H. Pixley, Maika Takita, et al. DOI: 10.1038/s41567-026-03470-6. The abstract and four figure captions were taken from the Nature page's ld+json metadata; institutional roles, the KL-divergence criterion, and the finite mean-field critical description come from the abstract and figure captions. The dataset/code availability statement notes neither is public. Figures 1–4 cover, respectively, the schematic principle, the quantum-to-classical control transition in the long-time steady state, the dynamical properties of the transition, and the system-size scaling of magnetization-fluctuation distributions and KL divergence.