(247d) Computerized Analysis & Optimization of Liquid-Liquid Extraction (CAO-LLE): A Systematic Approach

Liquid-liquid extraction (LLE) is the most common post reaction step in small molecule pharmaceutical synthesis. LLE is typically used either as a purification step or as an isolation step. In either case, the chemist would like to achieve both high selectivity and high yield at the same time. Despite the relevance of LLE, the development of liquid-liquid extraction process steps is still essentially empirical, generally relying on the chemist’s experience rather than on a systematic data rich approach. We hypothesize that a combination of extraction understanding coupled with a large amount of data from extractions carried out using a real process solution sample can lead to unique conditions which greatly benefit chemical process efficiency. Importantly, the utilization of optimization algorithms would allow to optimize for a variety of different objectives depending on what is most critical to the specific project (i.e. cost, purity, greenness of the solvents, etc...).

To this end, we discuss our efforts on the development of a systematic approach to the identification of optimal conditions for liquid-liquid extraction. Our plan is to rely on experimental data obtained in a high throughput manner while conditions are changed following a search algorithm that changes both continues and discrete variables leveraging a combination of DOE and multivariable optimization.

Our device is capable of carrying out hundreds of LLE experiments in series at the microliter scale. Each experiment uses a different solvent or solvent combination. For each LLE experiments, liquids are combined in a specifically designed mixing chamber, then separated via Zaiput’s membrane separation technology. After separation, sample volumes are measured and then appropriately diluted for injection into an HPLC. An algorithm is used to extract and analyze all the data collected with the HPLC in order to calculate partition coefficients of the desired product and the impurities, as well as to guide optimization towards a user defined goal.

We report our efforts on the design of both system hardware and software, and we discuss device characterization and LLE extraction optimization results obtained so far.