PNNL

The Science

Quantum simulations are made significantly more challenging by the need to rebuild the quantum computer instructions for each individual molecular calculation. Researchers developed COMPOSER, an architecture that separates the quantum circuit from the numerical calculations of chemical and materials systems. This approach allows scientists to use a single overarching set of quantum instructions for a specific molecular model, even as the molecular system and specific calculation change. Testing on water and other molecules validated the approach, reproducing the results of trusted reference calculations.

The Impact

Quantum simulations often repeat the same type of molecular calculation as atoms move during a reaction or study different electrons. Each of these changes can require rebuilding the quantum circuit. COMPOSER offers a new approach to design the circuit once for a defined family of calculations and then make changes through adjustable settings. By reusing a validated circuit, studies that compare numerous molecular structures could become more consistent and involve less repeated preparation work. COMPOSER could help researchers explore chemical bonding, reactions, and materials in quantum simulations more efficiently. 

Summary

Quantum chemistry workflows often evaluate families of closely related problems: a bond length changes, an active space grows, or an effective-Hamiltonian generator is refined. These updates can change operator term lists and trigger repeated circuit construction and routing. COMPOSER reorganizes standard block-encoding tools so that a declared orbital pool, branch library, selector map, ancilla layout, routing plan, maximum branch set, and normalization define one logical circuit topology. For problems within those fixed conditions, a new geometry, active-space mask, or truncation will update the coefficients, rotation angles, and classical masks without requiring structural recompilation.

Researchers performed numerical studies to test the structural premises of this architecture rather than simply claim an end-to-end hardware speedup. Across the molecular benchmark, Hamiltonian branch pools empirically grew near-linearly at fixed thresholds and the median overlap between second-order perturbation and coupled-cluster low-rank excitation subspaces was at least 0.85 in all three basis sets. Water tests kept selector widths fixed during a geometry scan and reused a 54-slot generator pool across active-space masks. The active-space Hamiltonians matched independent complete active-space configuration interaction calculations to numerical roundoff. The chemistry-scale similarity-sandwich demonstration used exact dense matrix exponentials to validate the operator workflow and topology-invariance premise.

Contact

Bo Peng, Pacific Northwest National Laboratory, peng398@pnnl.gov 

Funding

B.P. acknowledges the support from the Early Career Research Program through the U.S. Department of Energy, Office of Science under Grant No. FWP 83466. Y.L. acknowledges the support of the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research program under contract number DE-SC0025384. K.K. acknowledges the support of the Quantum Science Center, a National Quantum Information Science Research Center of the U.S. Department of Energy (under FWP 76213).