An HPCQC-Tailored Approach for Scalable Measurement of Physical Observables

This research paper presentation from the Chair of Computer Architecture and Parallel Systems at the Technical University of Munich addresses scalable measurement of physical observables in hybrid HPC-quantum computing (HPCQC) systems. The work targets a primitive that dominates many hybrid quantum-classical workflows: evaluating the expectation value of a generic operator with respect to a system state. Because complex operators consist of non-commuting terms, each term generally requires a separate quantum circuit execution; even a simple 12-qubit lithium hydride Hamiltonian demands hundreds of circuit runs, making efficient orchestration of this volume the central challenge. The approach assumes an architecture shared by several existing HPCQC systems, with a central authority, a classical worker layer acting as device proxies for QPU-specific classical operations, and a heterogeneous fabric of quantum backends. This decoupling enables a pipeline of optimizations that the authors tie together into one coherent workflow, aiming to make observable estimation a first-class citizen of the HPCQC software stack. Evaluation on molecular energy estimation problems shows that a custom batching and parallelization strategy improves makespan, while a non-uniform shot allocation strategy improves accuracy, with compound benefits when multiple QPUs are available. The speaker notes planned future work on real-world deployments and validation on physical quantum systems.