A September study used IBM quantum hardware to prepare and probe a 100-site model of quantum matter, demonstrating both the promise of programmable experiments and the limits noise still imposes.
A quantum computer has been used to prepare and examine a model of matter made of 100 linked sites—an achievement that brings a difficult problem in condensed-matter physics into the realm of programmable experiments. The result is promising, but a key signal weakened when researchers measured across longer stretches of the chain.
The study appeared September 1 in npj Quantum Information. George Pennington of the STFC Hartree Centre led the work with collaborators from IBM Quantum and IBM Research Cambridge, IBM’s Thomas J. Watson Research Center, IBM Research Europe, the National Physical Laboratory and Oxford. Specialist coverage on November 6 renewed attention to the study; it was not newly published that day.
The team used IBM quantum hardware to prepare four ground states of a bond-alternating Heisenberg chain, a theoretical model of interacting spin-½ sites. Such models let physicists explore quantum phases that can be difficult to isolate in ordinary materials. The prepared states had reported fidelities of 98.9%, 98.9%, 99.0% and 97.9% against reference calculations. Their circuits used between 18 and 39 layers of CNOT gates, which operate on pairs of qubits.
The researchers were looking for symmetry-protected topological, or SPT, order. This form of organization is not readily revealed by inspecting individual sites. Instead, physicists look for patterns linking parts of the system, including string order, entanglement features and protected edge states. The team extracted several such diagnostics, with string-order measurements extending to 20 sites.
That matters because a quantum processor can serve as a controllable laboratory, not just a calculator. Researchers can program it to create and study states that are difficult to investigate directly in materials, potentially opening new ways to test how quantum systems behave. The work strengthens the case for quantum simulation, while stopping short of showing that today’s processors can reliably model any material of interest.
The caution lies in the measurements. The hardware’s string-order values declined as the measured string grew longer and did not converge to a constant. The authors say the cause remains unsettled: accumulated readout error could be distorting the result, or device noise could be physically reducing the order. Distinguishing those explanations will matter for judging how faithfully a processor can preserve and reveal the quantum properties it is meant to simulate.
For the odd-Haldane phase, the team extracted an edge-state correlation length of 1.76 ± 0.20 without zero-noise extrapolation and 1.81 ± 0.21 with it. A density-matrix-renormalization-group calculation, a classical numerical method, gave 1.97 ± 0.03. The comparison provides a check against established computation, while the difference points to room for improvement. One circuit compilation took seven days of wall-clock time, illustrating the work required to translate a theoretical state into an experiment.
A separate result from RIKEN offers context for the broader effort to scale quantum devices. In October, the institute reported high-fidelity parallel control in a five-qubit silicon array using a shared microwave line, after researchers identified and mitigated an error caused by simultaneous operations. It is a distinct platform and study, but it addresses a related challenge: maintaining control as quantum operations become more complex.
The next steps for the IBM team are specific: determine how much of the weakening string-order signal comes from measurement error, improve control and readout, and test whether the diagnostics remain dependable in larger or more demanding simulations. The achievement is a sizeable, programmable model of quantum matter. The unresolved signal is a reminder that building the model and seeing its properties clearly are separate challenges.

