Quantum Processors Steer Random Dynamics with 100 Qubit Control

Ruizhe Shen and Ching Hua Lee at National University of Singapore integrated mid-circuit measurements with conditional operations to guide quantum dynamics, simulating systems of up to 100 qubits on IBM superconducting quantum processors. Transforming measurements into active control signals generated asymmetry in random circuit simulations and observed strong, noise-resilient effects distinct from previously known phenomena. Feedback functions as a programmable set of tools for managing non-unitary quantum control, potentially enabling the engineering of complex quantum behaviours and tunable open-system dynamics on larger quantum processors.

Emulating open quantum systems via programmable measurement dynamics

Quantum computers are rapidly evolving into flexible experimental platforms capable of engineering and probing complex quantum dynamics at large scales. Recent progress in device coherence, control precision, and qubit connectivity has enabled quantum processors to implement long-depth circuits, high-fidelity entangling operations, and mid-circuit measurements, establishing a new frontier for simulating non-unitary physics on programmable architectures. This progress opens new avenues for engineering measurement-based dynamics, non-equilibrium phenomena, and tunable open-system behaviour on large-scale quantum hardware.

In these settings, various non-unitary phenomena, previously associated with open systems under dissipation or decoherence, can now be strongly emulated and controlled. At the heart of this emerging direction lies the realization that quantum measurements are a powerful source of controllable non-unitarity. Each projective measurement collapses the wavefunction, generating stochastic yet tunable evolution that can profoundly reshape entanglement and information flow.

Repeated or selective application of these measurements can induce collective dynamical effects, such as measurement-induced phase transitions, crossovers between volume-law and area-law entanglement, and other non-unitary critical behaviour. This technological progress has given rise to a new generation of hybrid quantum, classical control protocols, where measurement outcomes are processed by a classical controller determining the next quantum operation in real time. Consequently, mid-circuit control has emerged as a practical route to embedding non-unitary evolution within programmable quantum circuits, bridging theoretical concepts of non-unitary dynamics with experimentally realizable platforms.

Measurements have conventionally been associated with stochastic wavefunction collapse. However, when spatially structured and combined with feedback-based control, mid-circuit measurements can additionally generate a new class of feedback-directed quantum dynamics. In this setup, feedback-based circuits implement an explicitly record-dependent quantum channel, promoting the measurement record to an active control signal encoding classical information into the quantum evolution.

As such, the arrow of information flow within the circuit can be dynamically engineered and controlled, rather than passively inferred. By coupling the quantum state to its classical measurement record, feedback introduces an operational form of spacetime nonlocality: future operations condition past outcomes, correlating distant events across the circuit spacetime. When embedded within chaotic (random) unitary circuits, these feedback-based primitives provide a strong and scalable mechanism for steering the evolution of local observables, establishing feedback as a practical resource for engineering directionality and controllable dynamics on quantum hardware.

Large-scale digital quantum simulations performed on state-of-the-art superconducting devices on the IBM Quantum platform, using up to 100 qubits, validate this feedback-directed quantum dynamical framework, a scale far beyond previous attempts in related directions. Despite the inherent noise in current hardware, strong signatures of feedback-directed scrambling are observed: information propagation and local observables exhibit a clear directional asymmetry induced by conditional measurement and control. Quantum circuits for feedback-based control, in the absence of nonunitarity, are composed of random unitary gates exhibiting statistically symmetric spreading of local observables such as occupation in either direction.

Over time, information initially localized in excitations disperses nearly uniformly across the system. Introducing feedback-based non-unitary operations demonstrates a five-fold increase in gate fidelity, resulting in a strong directional flow of information emerging from the interaction between circuit randomness and feedback. To describe the effect of feedback-based control, the projector acting on qubit x is introduced: Px[m] = |m⟩⟨m|. A conditional operation is applied if the outcome is m for the x-th qubit.

The associated Kraus operator is: Kx[m] = Vx[m]Px[m]. In the quantum trajectory picture, the density matrix evolves as follows: ρ →Kx[m]ρK†x[m] p[m] with p[m] = Try[Px[m]ρPx[m]]. This formalism makes the role of feedback explicit: once the system collapses onto a branch labelled by m, the deterministic control action Vx[m] is applied. Without feedback, Vx[m] = I, reducing the dynamics to random projections without directed influence. To incorporate a directional effect arising from feedback, Vx[m] or Kx[m] can be chosen asymmetrically across sites.

Two broad approaches for effecting direction-asymmetric feedback-based control are described: position-dependent feedback and feedback-conditioned operator. Position-dependent feedback implements spatial asymmetry by assigning feedback strengths that depend on the qubit position x: Kx[m] = f(x)Vx[m]Px[m], where f(x) encodes position-dependent feedback strength and Vx[m] is the conditional operation. Alternatively, asymmetry can be embedded directly into the feedback-conditioned operation itself: Kx[m] = a[m]Ox,x+Px[m], where Ox,x+1 is a two-site operator, and the coefficient a[m] depends explicitly on the measurement outcome m. The explicit asymmetry of Ox,x+1 (linking x to x + 1 but not x −1) also serves to asymmetrically steer transport and information flow correlated with the observed outcome.

Measurement-conditioned operations have been implemented previously for tasks such as state preparation and teleportation, but here they are employed as the central dynamical ingredient, thereby enabling a broader level of control over the flow of quantum information. These two feedback-based control approaches are demonstrated by progressing through the four circuit setups. Each panel shows one iteration of the circuit configuration, where the input state is scrambled by random unitaries (blue) before undergoing various measurement schemes (red, purple, or green). The random circuit ansatz serves to effect information scrambling and consists of repeated layers of unitary gates denoted as Uunitary(i) = [Urand]i. Each layer comprises alternating even, odd arrangements of controlled-Z (CZ) gates and random single-qubit rotations about the X axis, expressed as Urand = Yj∈even CZj,j+1RXj RXj+1 Yj∈odd CZj,j+1RXj RXj+1, where RXj = e−iθjX/2 denotes a random rotation about the X axis on qubit j.

Feedback control steers asymmetry in large-scale random quantum circuits

For the first time, simulations utilising up to 100 qubits have demonstrated strong asymmetry in random quantum circuits; previous attempts were limited to much smaller scales. This advance establishes feedback as a programmable resource for controlling non-unitary quantum dynamics, a previously inaccessible level of control for steering random dynamics on digital quantum processors. Researchers at the National University of Singapore achieved this by integrating mid-circuit measurements, measurements taken during a quantum computation, with real-time conditional operations, effectively turning passive readouts into active control signals.

This directional flow of information, induced by feedback, differs from the well-known non-Hermitian skin effect and opens new possibilities for engineering complex quantum behaviours and exploring non-equilibrium phenomena. Spatially structured mid-circuit measurements, combined with real-time conditional operations, actively steer quantum dynamics within random circuits. The team implemented this framework on IBM superconducting quantum processors, observing noise-resilient signatures of feedback-induced asymmetry. These findings open avenues for engineering complex quantum behaviours and exploring non-equilibrium phenomena; however, current results do not yet demonstrate sustained control beyond the timescales limited by qubit coherence, representing a significant hurdle to practical applications.

Mid-circuit measurements induce asymmetry and control in quantum systems

Quantum dynamics can now be steered using mid-circuit measurements, transforming passive observation into active control of quantum systems. This breakthrough enables directional information flow within complex circuits, generating asymmetry previously unseen in larger quantum simulations. Maintaining coherence, the delicate state required for quantum computation, is vital to the team’s success, and current limitations in qubit stability present a significant challenge.

Acknowledging current qubit instability is key; maintaining the delicate quantum state known as coherence for extended periods remains a considerable engineering hurdle. Nevertheless, demonstrating controlled asymmetry in these systems, even with limited coherence, validates the principle of ‘measurement-based’ control as a viable route towards more complex quantum devices. This ability to actively steer quantum behaviour, rather than simply observe it, unlocks possibilities for designing new algorithms and simulating complex physical phenomena.

This demonstration of programmable, non-unitary control via mid-circuit measurements establishes a new model for manipulating quantum systems. By transforming standard passive measurements into active control signals, scientists have achieved directional information flow within quantum circuits, generating asymmetry not previously observed at this scale. This work validates measurement-based control as a viable technique, despite limitations imposed by maintaining qubit coherence, the fragile quantum state essential for computation. The ability to steer quantum dynamics opens avenues for exploring complex phenomena such as non-equilibrium states and tunable open-system behaviour, moving beyond simply observing quantum processes.

Researchers demonstrated that quantum dynamics can be actively steered using mid-circuit measurements on up to 100 qubits. This transforms passive observation into a method of controlling quantum systems and generates asymmetry within complex circuits. The findings validate measurement-based control as a viable technique for manipulating quantum behaviour, even with current limitations in maintaining qubit coherence. The authors suggest this approach enables the exploration of non-equilibrium phenomena and tunable open-system behaviour.