Bi-fidelity fitting and optimization.

Area of Science:

• Computational chemistry and physics
• Materials science

Background:

• Investigating chemical and physical properties often involves computations at varying accuracy levels.
• Less accurate computations provide quick insights but are typically discarded for more accurate ones.

Purpose of the Study:

• To demonstrate a method for simultaneously incorporating computations at different accuracy levels.
• To reduce overall computational effort for studying chemical and physical phenomena.

Main Methods:

• Correlating results from inexpensive, less accurate calculations with a smaller set of expensive, highly accurate calculations.
• Applying the method to fitting potential energy surfaces using electronic structure calculations at two accuracy levels.
• Optimizing the optical response of metallic nanostructures using calculations at two grid resolutions.

Main Results:

• Potential energy surfaces were successfully fit by combining data from two electronic structure calculation accuracy levels.
• The optical response of metallic nanostructures was efficiently optimized by correlating calculations at different grid resolutions.
• The integrated approach significantly reduced the computational resources required compared to using only the most expensive calculations.

Conclusions:

• Simultaneously using computations at different accuracy levels offers a more efficient approach to studying complex systems.
• This method allows for a focused and cost-effective investigation of chemical and physical phenomena.
• The demonstrated applications highlight the broad applicability of this computational strategy.