Estimating SAXS profiles correlation by a Hierarchical Non Negative Matrix Factorization algorithm

This work aims at studying the crystallization process of Hydroxyapatite samples in three different chemical environments (Cit, Glr, CitOH), as a function of time and temperature (25°C, 37°C or biomimetic temperature, 60°C and 80°C) . In particular non-crystalline and/or precursor states (SAXS) are expected to play a key-role in this analysis. Due to the huge amount of data collected at Synchrotron facilities, a preliminary correlation evaluation is needed in order to extract the most representative curves showing significant modification in shape and/or in the regions of interest. An algorithm based on Hierarchical Non-Negative Matrix Factorization (intensity SAXS profiles are positive) has been developed and applied in order to select 2^n profiles (n==number of bisections of the original data set). The comparison of the algorithm findings to the known particle morphologies (SAXS fitting) has spotted the HA crystallization dynamics (time resolved) beneath, both at different temperatures and chemical environments.