Optimisation of multi-component phase mapping with machine learning and automation

点击翻译按钮获取中文摘要

Understanding the behaviour of complex chemical and physical soft matter systems is important from fundamental and applied (i.e. industrial) points of view. A way of understanding complex chemical and physical systems is characterising their thermodynamic behaviour. This can be described with the aid of phase diagrams. However, constructing a phase diagram is a laborious task and can take years to fully characterise a given system.Our group is interested in probing and phase-mapping complex polymer and surfactant mixtures with small angle neutron scattering (SANS) and other scattering techniques by coupling these techniques to microfluidic platforms.1-3 Recently, the advent of microfluidics has accelerated phase mapping.4-6 In addition, there have been examples in current literature on optimisation of phase mapping with machine learning.7-10With that in mind, the rationale behind combining microfluidics, in-line analytics (SANS and other scattering techniques) and active learning optimisation algorithm is described. First, the concept of soft matter and microfluidics is briefly introduced. Second, ways of probing soft matter, particularly with scattering techniques, described. A brief theory of multi-dimensional phase diagrams is covered and literature examples for optimisation of phase mapping given.