Predicting adsorption isotherms in natural water using polyparameter linear free energy relationships

Activated carbon is widely used in drinking water treatment to remove both natural organic matter (NOM) and trace organic contaminants. Adsorption isotherms for trace contaminants in the presence of NOM are needed to predict the performance of activated carbon for removing these compounds, but such data are not available for most of these contaminants and it would be expensive to carry out isotherm tests for all compounds of interest. This research was conducted to reduce the effort required to determine needed isotherm data by using molecular parameters to predict these data. Polyparameter linear free energy relationships (pp-LFER) for trace contaminants adsorbed on activated carbon from organic-free water have been well developed, and this research was carried out to extend the method to predict the adsorption capacity of activated carbon in natural water. Literature values of molecular descriptors of pharmaceuticals, endocrine disrupting compounds and some industrial solvents were used to predict natural water isotherms. Suwannee River NOM was used to represent natural water organic matter.;The first part of this study used oxidized graphite to represent a typical activated carbon surface. Oxidized graphite was used to eliminate the effect of pores on adsorption capacity so that surface chemical interactions can be better understood. pp-LFERs for trace contaminants adsorbed on oxidized graphite (OG) from organic-free water and water containing Suwannee River natural organic matter (SRNOM) were developed. The pp-LFERs were developed for measured partitioning coefficients Kd (i.e. the ratios of adsorption uptake qe over the equilibrium concentration Ce). The best-fit pp-LFER correlation for predicting single solute log Kd was determined to depend on solute activity a (i.e. the ratio of equilibrium concentration of a specific solute over its water solubility):;log Kd,i = [(13.30 +/- 5.83) + (1.96 +/- 1.22)log ai]V + [(-9.97 +/- 5.37) -- (1.16 +/- 1.11)log ai] B + [(1.02 +/- 3.10) -- (0.16 +/- 0.64)log ai]S + [(0.24 +/- 2.41) + (0.08 +/- 0.51)log ai]E + [(-4.12 +/- 3.87) -- (0.66 +/- 0.81)log ai] A + [(-11.87 +/- 5.20) -- (2.34 +/- 1.08) log ai] .;The best-fit pp-LFER correlation for adsorption from water containing SRNOM was determined to depend on solute equilibrium concentration:;log Kd,i = [(1.57 +/- 0.63) + (0.36 +/- 0.26)log Ce,i]V + [(-2.49 +/- 0.57) -- (0.66 +/- 0.25)log Ce,i] B + [(2.11 +/- 0.25) + (0.30 +/- 0.15)log C e,i]S + [(0.03 +/- 0.23) + (0.36 +/- 0.10)log Ce,i]E + [(-0.51 +/- 0.25) + (0.04 +/- 0.14)log Ce,i] A + [(-1.27 +/- 0.42) -- (1.60 +/- 0.23) log Ce,i] .;The results showed that predictions by the pp-LFER model for adsorption by oxidized graphite are within a factor of two from the experimental values.;In the second part of this study, a pp-LFER was developed to predict activated carbon adsorption isotherms of trace organic contaminants in the presence of Suwannee River natural organic matter (SRNOM) using molecular descriptors of the trace organic contaminants. The trace organic contaminants used to develop the pp-LFER include pesticides, pharmaceuticals, endocrine disrupting compounds and some industrial solvents and the molecular descriptors (V, A, B, S, E) can be found in the literature.;The best-fit pp-LFER equation was determined to be:;log Kd,i = [(-6.79 +/- 3.06) + (-0.71 = +/- 0.77)log Ce,i] V + [(7.84 +/- 2.27) + (1.63 +/- 0.64)log C e,i]B + [(1.08 +/- 0.48) + (-0.14 +/- 0.15)log Ce,i]S + [(-1.26 +/- 0.74) + (-0.83 +/- 0.19)log Ce,i] E + [(0.52 +/- 2.57) + (0.20 +/- 0.62)log C e,i]A + [(8.54 +/- 1.68) + (0.67 +/- 0.45)log Ce,i].;This pp-LFER equation was able to predict activated carbon adsorption isotherms that are within a factor of 1.81 from the experimental values. The equation was also able to predict atrazine adsorption isotherms in two other natural waters that are within a factor of 3 from the experimental values.