Abstract:Particles with internal inclusions or cores are ubiquitous in the atmosphere. One example is dust particles coated with water-soluble aerosols such as sulfate or nitrate. For these particles, the dust non-sphericity and the core-shell structure of the particle both fundamentally impact the optical properties. However, it is challenging to obtain an accurate and comprehensive parameterization of the optical properties of coated atmospheric particles. Recent studies have found that super-spheroids could provide a robust model for representing the non-sphericity of irregular particles, such as dust, sea salt, and ice particles. This important finding warrants further investigations on coated super-spheroids for the development of mixed-aerosol models. In this study, we developed a GPU version of the invariant imbedding T-matrix program (GPU-IITM) to improve the computational efficiency. This version allows for efficient computation of the optical properties of coated super-spheroids with extensive shape parameters and refractive indices. The GPU-IITM is 8–25 times more efficient than the conventional CPU-IITM, depending on the size parameter of the coated super-spheroids. The single-scattering properties of coated super-spheroids computed from the GPU-IITM allows for close examination of the optical differences between homogeneous and inhomogeneous particles. As an example, we present the linear depolarization ratios (LDRs) of coated super-spheroids, which have implications in active lidar remote-sensing research. We found that the LDR decreases obviously as the real part of the shell’s refractive index increases from 1.3 to 1.6. Greater LDRs occur for particles with a thin coating, while a rounder shape of the particle decreases the LDRs. However, special attention should be given to nearly spherical particles, for which the LDRs could be enhanced. In comparison to homogeneous particles, the degree of enhancement decreases or increases depending on whether the shell’s refractive index is smaller or larger than the core refractive index.

How machine learning approaches are useful in computing the optical properties of non-spherical particles across a broad range of size parameters?

Application of a Neural Network to Store and Compute the Optical Properties of Non-Spherical Particles

Group Contribution Method Supervised Neural Network for Precise Design of Organic Nonlinear Optical Materials

Jacobians of Single-Scattering Optical Properties of Super-Spheroids Computed Using Neural Networks.

Optical Properties of Nonspherical Atmospheric Particles and Relevant Applications

Optical Analog Computing Devices Designed by Deep Neural Network

Machine-learning-based computationally efficient particle size distribution retrieval from bulk optical properties

Applying Machine Learning to the Optics of Dielectric Nanoblobs

Computation of Optical Properties of Core-Shell Super-Spheroids Using a GPU Implementation of the Invariant Imbedding T-Matrix Method

Plasmonic Nanoparticle Simulations And Inverse Design Using Machine Learning

A comprehensive deep learning method for empirical spectral prediction and its quantitative validation of nano-structured dimers

Deep learning-based size prediction for optical trapped nanoparticles and extracellular vesicles from limited bandwidth camera detection

A deep learning approach for quantum dots sizing from wide-angle X-ray scattering data

Deep Neural Networks for Computational Optical Form Measurements

A calculation method for optical properties of yolk shell based on deep learning

Deep-Learning-Enabled Fast Optical Identification and Characterization of Two-Dimensional Materials

Machine Learning for Optical Scanning Probe Nanoscopy

Insights into the Machine Learning Predictions of the Optical Response of Plasmon@Semiconductor Core-Shell Nanocylinders

Deep‐Learning‐Enabled Fast Optical Identification and Characterization of 2D Materials

A photonic chip-based machine learning approach for the prediction of molecular properties