Innovative Spatially and Angularly-Resolved Diffuse Reflectance Spectrometer for Dense Media

There has been increasing interest and emphasis from chemical process industries on inline monitoring of vital product related indicators, such as chemical concentration and particle size. These indicators (particularly particle size) are commonly measured offline. Conventional probe-based inline spectroscopic instrument collect and combine spectral response of the samples from multiple optical fibers to produce a final spectrum. Results are often unreliable and inaccurate for samples containing particles. This is due to the non-linear light scattering effect due to the particles which alters the path of light travelling in the sample and give rise to complexity in the collected spectra.
We have developed a spatially and angularly-resolved diffuse reflectance (SARDR) spectrometer for measuring systems with high particle content. The probe-based instrument collects multiple reflectance individually from optical fibers placed at various distances from the source fibers delivering light at different incident angles. Polystyrene particle suspensions of various particle size and concentration were used to evaluate the performance of the SARDR system. Multivariate regression analysis was applied to the spectra collected by the SARDR probe and by the reference spectrometer (Varian Cary-5000) for off-line analysis. Two approaches to utilise the information contained in the SARDR spectra were employed to construct datasets for building calibration models. The first approach combined the individual spectra from each fiber in the probe to form a dataset of ‘total reflectance’ spectra for each sample. The second approach keeps the different SARDR spectra separate but restructures the original dataset into a new matrix for the regression model building step. The same procedures for PLS regression analysis to estimate particle size and concentration were applied to SARDR datasets constructed using the different approaches, and the diffuse transmittance spectra obtained from the reference instrument.
The performance of the regression model for estimating particle size and concentration were compared to evaluate the SARDR spectrometer performance. The results show that the probe-based SARDR spectrometer has similar performance in estimating particle concentration to the reference, off-line spectrometer, and performs better in estimating particle size than the reference instrument. It also suggests that the use of the second approach to construct the dataset for building regression model would lead to better model performance. Further investigation suggests that the difference in the information collected between fibers of different optical arrangement could be utilised in building the regression model. The different spectral response due to the spatial and angular arrangement of the fibers provides additional information for the regression model to produce more accurate estimates of sample properties.