IQM and Volkswagen report up to 124x faster classical step in QC-AFQMC chemistry method
IQM Quantum Computers and Volkswagen Group Innovation describe advances in Quantum-Classical Auxiliary-Field Quantum Monte Carlo (QC-AFQMC) in a study titled Quantum-Classical Auxiliary-Field Quantum Monte Carlo at the Edge of Practicability. Using Aitken's block transformation on singular Pfaffians and automatic differentiation for force bias, they cut per-step classical post-processing scaling from Õ(N^5.5) to Õ(N^4.5), which gave up to a 124x classical runtime reduction for a 50-orbital benchmark and an estimated 248x at 100 orbitals. On IQM's 54-qubit (physical) Emerald superconducting processor, the team computed the ground-state energy of a stretched H8 hydrogen chain, using matrix product state filtering to discard noisy measurements. In classical simulation, they modeled the rearrangement of a lithium superoxide dimer (Li2O4; 26 electrons in 20 orbitals) in a 40-qubit active space, which the report says captured the activation barrier and transition-state geometry. The article calls the work a peer-reviewed preprint on arXiv; a preprint is not yet peer reviewed.