The Signal
Researchers report 85% distillation probability on superconducting processor
A four-round protocol on IBM’s ibm_aachen processor produced an estimated 85% probability of extracting an entangled pair from a W state, according to a paper accepted by Physical Review Applied. The study also estimates lower entanglement of formation for four-round extractions than for single-round ones.

A four-round protocol on IBM’s ibm_aachen superconducting processor produced an estimated 85% probability of extracting an entangled pair from a W state, according to a paper accepted by Physical Review Applied. The paper, by Alexander C. B. Greenwood, Jackson Russett, Hoi-Kwong Lo and Li Qian, was accepted on October 5, 2026.
Source journals.aps.org
From a three-party state to a pair
The experiment used a three-qubit W state, the multipartite state from which the protocol tries to extract a pair shared by two of the three parties. In the protocol, the three parties—described as Alice, Bob and Charlie—each weakly couple a system qubit to an ancilla, or probe qubit, and measure the ancillas between rounds.
When exactly one party’s ancilla returns a 1, the other two parties share the extracted EPR pair. If all three outcomes are 0, the W state remains and the protocol can continue; if two or more outcomes are 1, it ends without distillation. A final strong measurement on Alice’s qubit provides another opportunity to extract a pair after the weak-measurement rounds.
Four rounds on ibm_aachen
The experiment ran on ibm_aachen, a 156-qubit Heron r3 system, using dynamic circuits whose operations depend on mid-circuit measurement results. The paper reports circuit depths from 24 for a single-round implementation to 60 for a four-round implementation, counting basis gates and measurements; each round required an average of 12 operations.
The researchers estimated the success probability by counting measurement outcomes and used matrix-free measurement error mitigation, known as M3. They report that the measured success rate tracked the theoretical prediction within 1-sigma error for the first two rounds. After the third round, the experiment diverged from theory, which the authors attribute to a decline in W-state fidelity at the start of each round.
Mid-circuit measurements also leave qubits idle while readout takes place. The paper gives a measurement time of about 1 microsecond and reports idle periods of 836 ns, compared with gate operations lasting 20–60 ns; the authors identify natural qubit dephasing as the dominant error mechanism.
Success probability and entanglement
The four-round success probability was 85%, compared with 75% in prior work and 66% for specific-party distillation on the same processor, according to the paper. These percentages describe the probability of obtaining a pair, rather than how much entanglement the resulting pair retains.
To account for pair quality, the researchers also estimated entanglement of formation. Their lower bound was at least 0.6379 pairs per W state for a single-round implementation and at least 0.5404 for four rounds. The authors say the lower four-round estimate reflects reduced fidelity of the extracted pairs; they note that their measure is an upper bound on distillable entanglement, rather than a direct measurement of it.
From preprint to journal acceptance
The study first appeared as an arXiv preprint submitted on August 12, 2025. The accepted-paper listing identifies it as a Physical Review Applied paper; its abstract describes implementation of up to four rounds and a distillation rate of approximately 0.85 pairs per W state.
Sources
Related stories
- Researchers define states whose entanglement dimension resists change
- IQM Entangles All 54 Qubits on Its Emerald Processor
- Researchers introduce a framework to upper-bound entanglement from a state's spectrum
- Three Researchers Derive a Unique Quantum Bayes Rule
- Moth claims Tessa record for quantum data encoding and readback