Infrared spectral modeling and simulation of multicomponent aerosol clouds

The size of particles typically present in aerosol clouds are in the range of 0.1-10 μm, which is within one order of magnitude of the infrared (IR) wavelengths in the molecular “fingerprint” region (approx. 6-12 μm). This length scale is also close to the optical absorption depth for materials of interest. Consequently, IR scattering signatures of aerosols differ from those associated with reflectance from surfaces of bulk materials. Furthermore, the shape of particles is also a factor that affects IR scattering spectra. Accordingly, both aerosol particle size and morphology must be considered in the development of accurate models and reliable detection algorithms. This report describes recent advances concerning modeling of IR scattering signatures of micron-sized spherical particles (found in liquid aerosols) and irregularly-shaped particles (found in solid aerosols). In our model, spherical particles are modeled using Mie scattering theory while nonspherical ones require numerical modeling. The model inputs involve particle optical constants (complex index of refraction - n and k), concentration of the aerosol components, particle size (diameter) distribution and various shape parameters (for solid aerosols only). Our model describes two detection scenarios: one where the signatures consist of back-scattered light only and the other where the portion of the forward scattered light which (diffusely) reflects off background surfaces is also collected. We discuss the effect of particle size and shape distribution on the IR signatures of aerosol clouds. We also present preliminary results on machine learning approaches to develop detection algorithms capable of detecting multicomponent aerosol clouds that have variable IR signatures due to different component fractions and particle size distributions. Finally, we discuss a simple method to test the robustness of detection algorithms against spectral distortions due to particle shape or other effects. The level of spectral distortions can be quantified using the novel concept of a “complexity score”.