Research

Fugaku linked to Quantinuum Reimei and ibm_kobe in multi-QPU hybrid workflow

Researchers from Quantinuum, the University of Tsukuba and RIKEN R-CCS linked the Fugaku supercomputer to two quantum processors at once: Quantinuum's Reimei trapped-ion system (20 physical qubits, at Wako, about 500 km away) and IBM's ibm_kobe superconducting system (156 physical qubits, at Kobe). They adapted Quantinuum's open-source Tierkreis workflow system to Fugaku's pjsub scheduler and direct sqcsub submission channels, and describe it as the first demonstrated multi-modality quantum-classical workflow across geographically separated QPUs. On a 16-site 2D transverse-field Ising model, ibm_kobe ran 500 VQE iterations with 20,000 shots per circuit, and the parameters were passed classically to seed a single-ancilla phase estimation routine on Reimei (300 shots). The hybrid approach gave an energy error of 0.61 ± 0.36 in units of J, versus 2.47 ± 0.10 for VQE alone on ibm_kobe and 3.45 for a classically optimized product-state baseline; the work is described in an IEEE conference paper for SC26, and the article also links a technical preprint, which is not yet peer reviewed.

Why it mattersHybrid quantum-classical jobs often rely on ad hoc scripts that are hard to scale in production supercomputing settings. The authors say this work shows a classical supercomputer can orchestrate different quantum hardware types without physical quantum interconnects. In a secondary case study, Tierkreis automated 80% of runtime coordination in biomolecular excited-state workflows, according to the article.

Read the full story at Quantum Computing Report →

Summarized by AI from the linked article.