Basis-Free Study of Confined Atoms and Molecules Based on a Neural Network Approach

Area of Science:

• Computational chemistry
• Quantum mechanics
• Materials science

Background:

• Analyzing cavity systems often requires constructing specialized basis sets that meet boundary conditions.
• This process involves time-consuming variational optimization for each system, potential, and cavity size.

Purpose of the Study:

• To present a novel method for analyzing cavity systems that bypasses the need for basis sets.
• To offer a computationally efficient and versatile approach for molecular calculations in varying cavity sizes.

Main Methods:

• Utilizing the variational quantum Monte Carlo (VMC) method.
• Implementing a neural network approach within the VMC framework to avoid basis set construction.
• Applying the method to systems in cavities of variable sizes.

Main Results:

• The proposed method requires no basis set, enabling "out-of-the-box" application across diverse systems and potentials.
• Achieved results comparable to highly accurate variational estimates for small systems.
• Demonstrated superior performance over traditional Gaussian basis set methods for many-electron systems.

Conclusions:

• The neural network-based VMC method provides an efficient and adaptable solution for cavity system analysis.
• This approach is particularly advantageous for studying high-pressure effects in systems with variable cavity sizes.
• The method offers a significant improvement over existing techniques, especially for complex many-electron systems.