FDES, a GPU-based multislice algorithm with increased efficiency of the computation of the projected potential.

Area of Science:

• Computational Physics
• Materials Science
• Electron Microscopy

Background:

• Traditional multislice algorithms for electron scattering simulations face computational challenges.
• Reciprocal-space calculations exhibit quadratic scaling with atom number per slice, while real-space calculations show linear scaling.

Purpose of the Study:

• To develop a more efficient computational strategy for simulating electron scattering phenomena.
• To improve the speed and performance of high-resolution transmission electron microscopy (HRTEM) image and diffraction pattern simulations.

Main Methods:

• Introduction of a hybrid computational strategy with a theoretical complexity of O(AlogA).
• Implementation of this strategy in a new program named forward dynamical electron scattering (FDES).
• Leveraging graphics processing units (GPUs) to further accelerate calculations.

Main Results:

• The hybrid approach demonstrably outperforms both pure reciprocal-space and real-space methods.
• Measured performance shows an order of magnitude improvement over reciprocal-space and a significant factor over real-space calculations.
• FDES successfully simulates HRTEM images, diffraction patterns, and convergent beam electron diffraction (CBED) patterns.

Conclusions:

• The hybrid strategy offers a substantial advancement in computational efficiency for electron scattering simulations.
• FDES provides a powerful, open-source tool for researchers in electron microscopy and materials science.
• GPU acceleration further enhances the practical utility of the FDES program.