Scholarlink

<p>This study introduces a hybrid computational framework that integrates the quantum-mechanically rigorous Boundary-Corrected Four-Body Continuum-Intermediate-State (BCIS-4B) method with a deep neural network (DNN) to create a high-fidelity surrogate model for predicting electron capture cross-sections in ion-atom collisions. The model was trained on a curated dataset of 4820 theoretical data points, generated from both new BCIS-4B calculations and published results, covering a wide range of projectile energies (50–1000&#xa0;keV) and scattering angles (0.1–3.5mrad). The DNN architecture, optimized through systematic hyper parameter tuning, demonstrates exceptional performance as a surrogate, achieving a coefficient of determination (R<sup>2</sup>) of 0.998 ± 0.001 on independent test data. When validated against experimental measurements, the hybrid BCIS-4B-ML model shows superior predictive accuracy compared to standard BCIS-4B and CB1-4B methods, particularly in the critical intermediate angular range where it reduces the mean absolute percentage error by up to 73.3%. Beyond accuracy enhancement, the surrogate model delivers a computational speedup of several orders of magnitude, enabling near-instantaneous cross-section predictions while bypassing the computationally expensive numerical integrations of the traditional quantum method. This work establishes a new paradigm for physics-informed surrogate modeling in atomic collision theory, providing both computational efficiency and enhanced physical insight into complex four-body charge transfer processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A machine learning-based surrogate model for enhanced four-body charge exchange calculations in ion-atom collisions

  • 期刊论文

  • 发表日期：

    2026年5月17日

查看全文
  • Hoda Ghavaminia

摘要

This study introduces a hybrid computational framework that integrates the quantum-mechanically rigorous Boundary-Corrected Four-Body Continuum-Intermediate-State (BCIS-4B) method with a deep neural network (DNN) to create a high-fidelity surrogate model for predicting electron capture cross-sections in ion-atom collisions. The model was trained on a curated dataset of 4820 theoretical data points, generated from both new BCIS-4B calculations and published results, covering a wide range of projectile energies (50–1000 keV) and scattering angles (0.1–3.5mrad). The DNN architecture, optimized through systematic hyper parameter tuning, demonstrates exceptional performance as a surrogate, achieving a coefficient of determination (R2) of 0.998 ± 0.001 on independent test data. When validated against experimental measurements, the hybrid BCIS-4B-ML model shows superior predictive accuracy compared to standard BCIS-4B and CB1-4B methods, particularly in the critical intermediate angular range where it reduces the mean absolute percentage error by up to 73.3%. Beyond accuracy enhancement, the surrogate model delivers a computational speedup of several orders of magnitude, enabling near-instantaneous cross-section predictions while bypassing the computationally expensive numerical integrations of the traditional quantum method. This work establishes a new paradigm for physics-informed surrogate modeling in atomic collision theory, providing both computational efficiency and enhanced physical insight into complex four-body charge transfer processes.