(23b) Computational Scheme for Rational Solvent Design in Extraction of Pharmaceutical Ingredient | AIChE

(23b) Computational Scheme for Rational Solvent Design in Extraction of Pharmaceutical Ingredient


Madakashira, H. - Presenter, Indian Institute of Technology Bombay
Adhikari, J., Indian Institute of Technology, Bombay.
Noronha, S. B., I.I.T. Bombay, Mumbai
K, Y. R., Indian Institute of Chemical Technology

The pharmaceutical industry is one of the largest users of organic solvents in processes such as reaction, formulation, separation and cleaning of equipment. Solvents constitute the major part of the mass in manufacture of an active pharmaceutical ingredient (API) with typical E-factors (kg of waste per kg of product) ranging from 25 to 100.1  An attempt to minimize the amount and number of solvents in a process is essential with regard to cost, environmental, health and toxicological concerns. It is also desirable for these solvents to be green towards reducing their impact on the environment. An experimental trial and error approach to finding an appropriate solvent is time consuming and expensive; computer aided molecular design2 (CAMD) can be an alternative strategy. CAMD utilizes structure property correlations and optimization technique to find a novel or potential solvent for a physical separation process provided the property estimates are sufficiently accurate. This approach has been applied to a case study where an aliphatic solvent is designed for efficient extraction of an API, R-phenylacetylcarbinol, from an aqueous phase. Current industrial practice involves the use of the solvent toluene.

            The prime properties that are of interest for liquid-liquid extraction include distribution coefficient, solvent loss, solvent power, selectivity, viscosity, boiling point and toxicity. These properties have been predicted using group contribution and connectivity index based structure property correlations. The constraints on properties have been framed by choosing toluene as a reference solvent. A variant of genetic algorithm based optimization approach, namely non-dominated sorting genetic algorithm3 (NSGA-II) along with a UNIFAC4 (universal functional activity coefficients) subroutine has been employed for optimization. A total of 64 UNIFAC groups (including repetitions) were considered for the current study whose combinations lead to the design of different type of solvent molecules such as aldehydes, ketones, ethers and esters. UNIFAC based structural constraints are employed to design a structurally feasible molecule. A multi objective function minimizing solvent loss and maximizing distribution coefficient have been used to arrive at globally best molecules. The CAMD approach yielded a pareto set of optimal solvents including, amyl acetate and 2-methyl penta 1,3-diene.

             Isobaric-isothermal Gibbs ensemble Monte Carlo simulations have been carried out on the solvents identified above, using MCCCS towhee package5. Simulation results confirm the property estimation predictions and hence, the short-list of solvents chosen for further investigation. Experiments are performed with these solvents for determination of the extent of their suitability for R-PAC recovery.


1.           Grodowska, K. & Parczewski, A. Organic solvents in the pharmaceutical industry. Acta poloniae pharmaceutica 67, 3–12 (2010).

2.           Harper, P. M. , Gani, R., Kolar, P., Ishikawa, T.  Computer-aided molecular design with combined molecular modeling and group contribution. Fluid Phase Equilibria 158-160, 337–347 (1999).

3.           Deb, K., Member, A., Pratap, A., Agarwal, S. & Meyarivan, T. A Fast and Elitist Multiobjective Genetic Algorithm. IEEE transactions on evolutionary computation 6, 182–197 (2002).

4.           Magnussen, T. UNIFAC Parameter Table for Prediction of Liquid-Liquid Equilibria. Industrial  Engineering Chemistry Process Design and Development 20, 331–339 (1981).

5.           MCCCS Towhee. Available at: http://towhee.sourceforge.net.