After an extensive survey of what a two phase capillary pumped cooling system is, key physical properties and their implication on the CPL process operation were identified. This information was used to refit the set of fluid requirements in terms of target property values and to define property weights according to their importance in the performance function. This function aggregates the properties of each candidate fluid. The performance of existing fluids was then calculated. It showed that none was satisfactory.
The influence of chemical functions on each property in the performance function was investigated to provide useful guidelines about which chemical fragments to expect in candidate fluids. It enabled to draw the fingerprint of the ideal candidate fluid. It further exemplified that finding a new fluid would be difficult to achieve because some property targets were found contradictory. Consequently adjustments to the performance function were made.
The search of a candidate fluid was run under a reverse engineering approach which combined a bottom-up and a top-bottom approach.
The bottom-up approach first started with the exploration of possible fluids issued from renewable materials, with the help of GRASS® computer aided synthesis software for sustainable solvents developed at ENSC Lille (France). This way would have ensured that candidate fluids were readily synthesizable within the green chemistry paradigm and would comply with some sustainable growth issues. It proved moderately successful as existing, and non-satisfactory, fluids were found and new fluids had a poor performance. This prompted us to extend further the search for fluids among databases both in the literature and on the web. For many molecules, experimental data were missing, preventing the proper calculation of their performance. Solutions were projected in two ways. One way was to restrict the performance function to properties that bore the highest weight in the overall performance function. Another way was to fill in missing data with predicted data from the IBSS® property package developed at INP Toulouse (France). These solutions led to the proposal of a first list of candidate pure fluids.
The top-down approach consisted in using the IBSS® CAMD tool developed at INP Toulouse (France). A set of target property values matching the project requirements was defined and the complete performance function was implemented. Second a set of chemical fragments was selected, incorporating all chemical functions for which the fluid performance was found sensitive. Then the CAMD search was run with the help of the IBSS CAMD in-house genetic algorithm to build new molecular structures. For each, the IBSS® property package was used to predict their properties and evaluate their performance. This led to a second list of candidate pure fluids based on a purely predictive approach. For each of them the full and the restricted performance was computed.
Comparison of the two pure fluid lists over the restricted performance showed similar molecules. The bottom-up approach was more efficient for small molecules, confirming an intrinsic weakness of fully predictive top-down approaches to accurately predict small molecules properties and properly evaluate their performance. The top-down approach was efficient to propose innovative molecules. Although no pure fluid candidate reached the 100% mark over the set of specifications, some solutions showed a better performance value than existing fluids.
The next step aims at improving the performance of candidates by using mixtures. The IBSS® CAMD was run aiming at finding those mixtures but it failed because most solutions were found unstable and because over the limit uncertainty was propagated by non-linear mixture property models which accuracy is somewhat too erratic. This led to a refocus of the mixture search over a set of selected molecules taken from the pure fluid search lists. This is currently running.