New Protocol Reconstructs 96-Qubit Quantum States
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Matteo Votto and collaborators from Université Grenoble Alpes, Technical University of Munich, Max Planck Institute of Quantum Optics, University of Innsbruck, and University of Bologna developed a protocol that learns matrix-product operator (MPO) representations of quantum states from randomized measurements, published in Physical Review Letters.
- The protocol was tested on IBM's superconducting quantum processor Brisbane, where it reliably reconstructed an entangled state across 96 qubits — compared to the previous state-of-the-art for quantum state tomography, which the authors cite as limited to 35 qubits.
- The method works by performing random operations on individual qubits, collecting bit-string data, and then learning a tensor network (MPO) compatible with that dataset — where each qubit's correlations can be inferred from data on just a few neighboring qubits rather than the full system.
- Votto told Phys.org the team was motivated by the observation that noisy systems contain 'less information' and should be easier to analyze, leading them to leverage tensor network states — a tool routinely used to simulate noisy quantum systems — to simplify the learning task.
- The protocol is compatible with existing randomized-measurement workflows and borrows from established tensor network algorithms like the density matrix renormalization group, making it naturally robust to experimental errors and applicable to datasets researchers are already collecting.
- Votto said future work will generalize the protocol to learning quantum channels (the operations a computer performs) and to two-dimensional qubit geometries, since most current quantum computers have 2D connectivity.
Why it matters: State reconstruction is a prerequisite for verifying that quantum computers are actually doing what they should — and 96 qubits is nearly three times the prior ceiling of 35. For the teams building noisy intermediate-scale quantum hardware, a protocol that learns a full state from realistic, already-collected measurement data means faster benchmarking and error correction at scales that were previously out of reach.
