Two-dimensional pattern reverse Monte Carlo analysis of nanoparticles in polymer matrices using a combination of OpenACC and cuFFT

We propose a program code for reverse Monte Carlo (RMC) modeling of two-dimensional scattering patterns using a combination of OpenACC and cuFFT. The RMC method estimates the three-dimensional positions of nanoparticles (NPs) in polymer matrices to fit scattering data of NPs observed in X-ray and neutron experiments. The scattering data can be calculated as the convolution sum of the three-dimensional Fourier transform of the three-dimensional density distributions of NPs using the particle-mesh approach. To speed up the graphics processing unit (GPU) calculations, we implement a code by combining the OpenACC standard and cuFFT library and minimize the data transfer between the GPU and host. This program is publicly available and can facilitate RMC analysis of two-dimensional scattering patterns to model NP morphologies in various polymer materials.

本研究提出了一种基于 OpenACC 与 cuFFT 结合的二维散射模式逆向蒙特卡洛（RMC）建模的程序代码。逆向蒙特卡洛法可估算聚合物基体中纳米颗粒（NPs）的三维位置，以拟合 X 射线和中子实验中观测到的 NPs 散射数据。通过粒子网格法，散射数据可计算为纳米颗粒三维密度分布的三维傅里叶变换的卷积和。为加速图形处理单元（GPU）的计算，我们通过结合 OpenACC 标准和 cuFFT 库实现了一项代码，并最小化了 GPU 与主机之间的数据传输。该程序代码对外公开，有助于利用 RMC 分析二维散射模式，以模拟多种聚合物材料中纳米颗粒的形态。