A BAYESIAN METHOD FOR ESTIMATING THE PARTICLE SIZE DISTRIBUTION OF POLYMER DISPERSIONS FROM MULTIANGLE DYNAMIC LIGHT SCATTERING MEASUREMENTS

LUIS A. CLEMENTI; HELCIO R.B. ORLANDE; LUIS M. GUGLIOTTA; JORGE R. VEGA

Los Cocos

Simposio; V Simposio Argentino-Chileno de Polímeros; 2009

Universidad Nacional de Cordoba

Dynamic light scattering (DLS) is frequently used for estimating particle size distributions (PSD) of dilute latexes. A single-angle DLS measurement only provides a small amount of information on the PSD, and consequently the estimated PSD exhibits a poor resolution. In contrast, multi-angle DLS (MDLS) measurements allow an increase in the information content, thus improving the PSD estimate. Typically, the scattering phenomenon is modeled on the basis of Mie theory. The PSD estimation requires solving a linear ill-conditioned inverse problem (ICIP); and to this effect regularization methods must be used. Recently, the MDLS problem was stated as a formally simpler and more robust (but nonlinear)ICIP, and a technique based on artificial neural networks (ANN) was developed and successfully validated when applied to several latexes. In this work, we propose a statistical method for solving the non-linear ICIP, based on a Bayesian inference approach. The proposed method was successfully evaluated on the basis of synthetic and experimental examples, corresponding to PS latexes. It was able of adequately recovering PSDs of different shapes and with different range of diameters. In all analyzed examples, the method predicted PSDs better or similar than those obtained through classical regularization methods.