Exploration of the Global Minimum and Conical Intersection with Bayesian Optimization.

Area of Science:

• Computational Chemistry
• Quantum Computing
• Chemical Physics

Background:

• Molecular geometry optimization typically relies on energy gradients from quantum chemical calculations.
• Quantum computing introduces noise into energy calculations, hindering gradient-based optimization.
• Gradient-free methods are needed to address noise in quantum-assisted chemistry calculations.

Purpose of the Study:

• To develop and evaluate a Bayesian optimization (BO) strategy for molecular geometry searches.
• To identify suitable acquisition functions (AFs) for BO in exploring critical molecular structures.
• To locate the global minimum (GM) and the most stable conical intersection (CI) points.

Main Methods:

• Proposed a geometry search strategy using Bayesian optimization (BO).
• Investigated appropriate acquisition functions (AFs) for exploring global minima and conical intersections.
• Applied the BO strategy to two test molecules, simulating noisy quantum computer measurements.

Main Results:

• Successfully located the global minimum (GM) and most stable conical intersection (CI) geometries with high accuracy for both molecules.
• Demonstrated the robustness of the BO strategy even with the addition of artificial random noise to energies.
• Validated the potential of BO for geometry optimization using noisy quantum computer outputs.

Conclusions:

• Bayesian optimization provides a viable gradient-free approach for molecular geometry searches.
• The proposed BO strategy effectively handles noisy energy data, simulating quantum computation limitations.
• This method accurately identifies key molecular structures, paving the way for quantum-enhanced chemical simulations.