We introduce numerical methods and algorithms to estimate the main parameters of fractal-like particle aggregates from their optical structure factor (i.e. light scattering diagrams). The first algorithm is based on a direct and simple method, but its applicability is limited to aggregates with large size parameter and intermediate fractal dimension. The second algorithm requires to build calibration curves based on accurate particle agglomeration and particle light scattering models. It allows analyzing the optical structure factor of much smaller aggregates, regardless of their fractal dimension and the size of the single particles. Therefore, this algorithm as well as the introduction of a criterial curve to detect the different scattering regimes, are thought to be powerful tools to perform reliable and reproducible analysis.

BC (Black Carbon), which can be found in the atmosphere, is characterized by a large value of the imaginary part of the complex refractive index and, therefore, might have an impact on the global warming effect. To study the interaction of BC with light often computer simulations are used. One of the methods, which are capable of performing light scattering simulations by any shape, is DDA (Discrete Dipole Approximation). In this work its accuracy was estimated in respect to BC structures using the latest stable version of the ADDA (vr. 1.2) algorithm. As the reference algorithm the GMM (Generalized Multiparticle Mie-Solution) code was used. The study shows that the number of volume elements (dipoles) is the main parameter that defines the quality of results. However, they can be improved by a proper polarizability expression. The most accurate, and least time consuming, simulations were observed for IGT_SO. When an aggregate consists of particles composed of ca. 750 volume elements (dipoles), the averaged relative extinction error should not exceed ca. 4.5%.

We report on the first application of the graphics processing units (GPUs) accelerated computing technology to improve performance of numerical methods used for the optical characterization of evaporating microdroplets. Single microdroplets of various liquids with different volatility and molecular weight (glycerine, glycols, water, etc.), as well as mixtures of liquids and diverse suspensions evaporate inside the electrodynamic trap under the chosen temperature and composition of atmosphere. The series of scattering patterns recorded from the evaporating microdroplets are processed by fitting complete Mie theory predictions with gradientless lookup table method. We showed that computations on GPUs can be effectively applied to inverse scattering problems. In particular, our technique accelerated calculations of the Mie scattering theory on a single-core processor in a Matlab environment over 800 times and almost 100 times comparing to the corresponding code in C language. Additionally, we overcame problems of the time-consuming data post-processing when some of the parameters (particularly the refractive index) of an investigated liquid are uncertain. Our program allows us to track the parameters characterizing the evaporating droplet nearly simultaneously with the progress of evaporation.