Sorption and partitioning of neutral and charged organic species: a log-normal Langmuir isotherm model & application of quantum-chemically estimated Abraham solute parameters

Di Toro, Dominic M.The overall goal of this doctoral dissertation is to develop methods to accurately predict the partitioning and sorption behavior of neutral and ionic species from water into liquid organic phases as well as onto various forms of black and organic carbon. For neutral solutes, this includes the development of a novel sorption isotherm – the log-normal Langmuir model – that describes nonlinear partitioning onto carbonaceous sorbents with one sorbate-specific parameter. For charged species, a model has been developed that can generate the Abraham parameters that are the physical-chemical descriptors for charged species using quantum chemical computations. These solute descriptors are then used to predict both the solvent-water partitioning of the ionic species, as well as the sorption of ionic species onto soil organic carbon, using the log-normal Langmuir isotherm model developed for neutral species. To better understand the nature of non-linear adsorption of organic solutes onto black carbon, a non-linear sorption model – the log-normal Langmuir model – was developed in Chapter 2 that utilizes Langmuir isotherms with a log-normal distribution of binding constants and a single maximum sorption capacity. The model has two sorbent-specific parameters: the maximum sorption capacity, qmax, and the standard deviation, σ_κ, of the log of the Langmuir binding constants; and one sorbate-specific parameter, the median Langmuir binding constant, K ̃_L. This is an important advance over previously available sorption isotherm models, for example the Freundlich isotherm, which has two sorbate-specific parameters. The reduction to a single sorbate-specific parameter which has chemical meaning - the median Langmuir binding constant, K ̃_L- is an important advance. In particular it allows quantitative prediction of the isotherm for a new chemical if the single sorbate-specific parameter can be predicted. In Chapter 3, the median Langmuir binding constants are predicted using an Abraham poly-parameter linear free energy relationship (pp-LFERs). For sorption of neutral organic solutes onto graphite, charcoal, Darco GAC, and F400 GAC (n = 13, 11, 14, 44 sorbates, respectively), RMS errors of predicted median binding constants, log⁡(K ̃_L), of 0.129, 0.307, 0.407, and 0.424 were obtained. Predicted isotherms were constructed with RMS errors of the predicted sorbed concentrations, log⁡(q(c)), of 0.0820, 0.1809, 0.183, and 0.220, respectively. This demonstrates that using the LNL isotherm and Abraham pp-LFER models, it is possible to predict the sorption isotherm of a new sorbate from only its molecular structure. In Chapter 4, Abraham parameters for ionic species are estimated directly from quantum chemical (QC) computations of solvent-water partition coefficients and molecular polarizability by extending a method developed by Liang & Di Toro for neutral species. Quantum-chemically estimated Abraham solute (QCAP) parameters are determined for the solvent-water partitioning of a suite of carboxylic acid anions (n = 60) in acetone-, acetonitrile-, dimethylsulfoxide-, and methanol-water systems, as well as a suite of quaternary amine cations (n = 217) in an octanol-water system. Using these QCAP solute parameters, predictions of experimental solvent-water partition coefficients are made for the carboxylate anions with RMS errors of 0.475, 0.512, 0.460, and 0.393 for the four solvent-water systems, respectively. This is an improvement over both direct a priori QC calculations (RMSE = 3.43, 3.71, 0.698, and 2.14, respectively) and predictions made using Absolv-estimated Abraham solute descriptors (AAP) (RMSE = 0.636, 0.59, 1.11, and 0.389, respectively) for the four solvent-water systems. For the quaternary amine cations, the QCAP and AAP methods showed comparable improvements over direct QC computations of the octanol-water partition coefficients (RMSE = 1.16, 2.82, and 0.997, for the QCAP, direct QC, and AAP methods, respectively). In Chapter 5, the log-normal Langmuir (LNL) isotherm is used to model non-linear sorption of a suite of primary through quaternary amines (n = 80) onto natural organic carbon (Pahokee peat). The LNL model can reproduce the sorption data (RMSE = 0.272, N = 80) for both fully ionized, as well as partially ionized, species. The latter are modeled as the sum of a linear isotherm for the neutral species, the usual model, and a LNL model for the charged species, weighted by the fractions neutral and ionized species present at the experimental pH = 4.5 and 6.8. The median Langmuir binding constants for the ionic species are predicted using QCAP solute descriptors (RMSE = 0.526, N = 60) with accuracies comparable to those of linear partition coefficients for neutral species. The parameter for the neutral species linear isotherm model,〖 K〗_OC, is predicted using a previously developed Abraham model. The predicted isotherms, constructed using the QCAP-predicted median binding constants and Abraham-predicted K_OC, demonstrate good agreement with the experimental data (RMSE = 0.457, N = 60). The results of this dissertation research are models that can reproduce non-linear sorption isotherm data of neutral and ionic species with only one sorbate-specific parameter. For the cases considered: samples of graphite, powered charcoal, and activated carbon, and for a natural organic carbon (Pahokee peat), the sorbate specific parameter can be predicted from its molecular structure only. The QCAP parameters can also be used to predict solvent-water partitioning of ionic compounds. This greatly expands the range of compounds that can be analyzed for physical/chemical properties that are used to evaluate the environmental risk posed by new compounds for which only the molecular structure is known.University of Delaware, Department of Civil and Environmental EngineeringPh.D