The selection and recovery of two or more solvents (i.e. co-solvents) are studied because of their importance in environmentally responsible processing. A particularly challenging problem arises when the contaminants (or solutes) are comprised of a mixture of volatile and nonvolatile compounds. The nonvolatile components present difficulties in solvent selection while the volatile components give rise to solvent recovery challenges. Traditional approaches such as the Kauri–Butanol test and predictions based on solubility parameters from regular solution theory can result in poor solvent selection when some of the contaminants are nonvolatile. A simple, graphical and rigorous procedure based on Gibbs energies of mixing for all binary mixtures of solvents and contaminants is proposed for choosing solvent/contaminant pairs. This selection procedure is based on the choice of solvent resulting in the lowest solvent/contaminant Gibbs free energy of mixing for a given contaminant. Nonvolatile contaminants are often easily separated in a single-stage flash vessel. Volatile contaminants, on the other hand, can form homogeneous or heterogeneous azeotropes with commonly used solvents, give rise to distillation boundaries and result in a challenging co-solvent recovery problem. A systematic procedure for the synthesis, design and economic analysis for co-solvent recovery is presented. Residue curve maps are used to identify distillation boundaries and to generate a conceptual process flow sheet for co-solvent recovery. Equipment sizes for the separators and auxiliaries (i.e. condensers, reboilers, decanters, etc.) are determined and capital investment, installation and operating costs are calculated. The removal of oil–water emulsions from the surface of machined metal parts using n-propyl bromide (NPB) and isopropyl alcohol (IPA) in a closed-loop degreaser is used to illustrate the proposed methodology of co-solvent selection and recovery. Analysis shows that the Kauri–Butanol test and the use of solubility parameters can result in poor solvent selection. In contrast, the proposed methodology shows that NPB is competitive with banned solvents like chlorodifluoro, fluorodichloroethane and trichloroethane and highly regulated solvents such as perchloroethylene, trichloroethylene and is a good solvent for the removal of machining oils while IPA is a good solvent for removing water. Separation of mixtures of NPB/IPA/water is difficult because of the presence of three binary and one ternary azeotrope that create three distinct distillation regions as well as regions of vapor–liquid–liquid behavior. It is shown that these distillation boundaries preclude direct water separation and that a two-distillation sequence with decantation is required to cross boundaries and recover both solvents. A variety of geometric illustrations are presented to elucidate key concepts.