Analysis of noisy dynamic light scattering data using constrained regularization techniques.

Dynamic light scattering (DLS) from colloidal particles often contains noise, which makes inversion of the correlation function to obtain the particle size distribution (PSD) unreliable. In this work, poor-quality correlation function data with baseline error were analyzed using constrained regularization techniques. The effect of baseline error was investigated, and two strategies were proposed to compensate for baseline error. One strategy is based on edge proportion detection of spurious peaks at large size in the PSD, and the other is based on the solution norm. Results from simulated and experimental data demonstrate the effectiveness of our proposed strategies. The L-curve rules for standard Tikhonov and for constrained regularization, the generalized cross-validation (GCV) rule, and the robust GCV rule were investigated for determination of the regularization parameter. A comparison of these rules was done using both simulated and experimental data. It is shown that correction of baseline error with baseline compensation as well as a reasonable regularization parameter choice improves the accuracy of PSD recovery in poor-quality DLS data analysis.