Hao Li, Chaoqun Ji, and Mengbo Fu of University of Chinese Academy of Sciences have developed a method for building quantum circuits that bypasses a major computational hurdle. The researchers use a graph to represent the circuit construction problem, encoding based on the antisymmetrized two-electron Hamiltonian coupling. For a canonical Hartree-Fock reference, the initial ADAPT gradient magnitude is exactly the corresponding Hamiltonian coupling, allowing for the construction of quantum circuit operators. Across benchmarks of eight to twenty qubits, this approach achieves chemical accuracy while minimizing the need for repeated, resource-intensive quantum measurements.
Heterogeneous-Graph Encoding of Static Hamiltonian Coupling
The initial magnitude of the ADAPT gradient, a key metric in variational quantum eigensolver (VQE) methods, precisely matches the antisymmetrized two-electron Hamiltonian coupling for a canonical Hartree-Fock reference. This surprising correspondence, detailed in recent work, allows for a novel approach to quantum circuit construction by establishing a direct link between a complex quantum calculation and a fundamental Hamiltonian property. This encoding enables the construction of quantum circuit operators, eliminating the need for repeated quantum-gradient measurements that traditionally bottleneck ADAPT methods.
The team demonstrated that this approach achieves chemical accuracy, a benchmark for reliable quantum chemistry results, while minimizing the number of quantum gates required. Across benchmarks using systems of eight to twenty qubits, incremental circuit construction with this method yielded operator counts comparable to those achieved with full gradient-based selection. This efficiency is particularly notable as it avoids the iterative quantum screening process inherent in conventional ADAPT techniques.
The researchers tested the effectiveness of learned corrections by comparing their method against supervised and residual models, finding that static learned readouts did not consistently improve performance. However, state-conditioned features did enhance ranking metrics, though they required additional information and did not fully close the loop on operator count reduction.
As a further refinement, the team derived an exact shared-residual formulation that reduces the number of Hamiltonian applications needed for triple- and quadruple-excitation calculations. This advancement offers a complementary benefit, particularly for larger, more complex systems, and suggests a pathway toward more efficient and scalable quantum simulations of molecular properties.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
