Page 88 - Partition & Adsorption of Organic Contaminants in Environmental Systems

CORRELATIONS OF PARTITION COEFFICIENTS 79

solute solubility. The solvent pairs expected to comply with these requirements
include, for example, hexane versus heptane (or other higher alkanes), n-
butanol versus i-butanol, n-pentanol versus s-pentanol, and n-octanol versus
triolein. In those systems, a should be very close to 1, largely independent of
the types of solutes included in the set, and b should be small (i.e., close to
zero). For other systems, the degree of the linear fit between K sw,1 and K sw,2
would vary to different extents with the polarities of individual solutes and
with the compositions of the solvents, and therefore a and b derived from the
regression of logK sw,2 against logK sw,1 could vary widely.
As we have seen from the logK ow and logK hw data presented earlier, a
reasonably good linear relationship exists between logK ow and logK hw if we
restrict our analysis to a group of relatively nonpolar solutes (i.e., the ones
with relatively large logK ow values). The relationship becomes meager when
polar solutes are included in the group because they respond very differently
to apolar hexane compared to weakly polar octanol. Therefore, if a statistical
analysis of logK hw against logK ow is attempted for a mixed set of nonpolar and
polar solutes, the results will not yield a good linear fit, and the resulting a and
b values will be ambiguous. A good way to rectify this problem is to divide
the mixed set of solutes into two or more subsets according to their polarities
(or specific modes of molecular actions) (Leo et al., 1971). This treatment
improves the correlation fit for each subset and allows for a better interpre-
tation of the resulting a and b values for each solute type. Therefore, if one is
to predict the partition coefficient of a test solute with an organic phase of
interest from its partition coefficient with a reference solvent, it is essential
that the test solute belong to the same or similar class of solutes for which a
previous correlation has been established.
If the two organic phases with logK sw,1 and logK sw,2 contain similar polar
group(s) but differ significantly in their overall polarities, linear correlations
may then be observed to encompass many solute classes. Leo et al. (1971)
showed, for example, that the logK sw,2 for a diversity of solutes with weakly
and moderately polar solvents, such as oleyl alcohol, methylisobutyl ketone,
ethyl acetate, n-, s-, and t-pentanol, cyclohanone, and n-butanol, could all
be reasonably correlated with logK ow as the reference. Let us consider
logK bw (but-water) versus logK ow(oct-water) in some detail. It is shown earlier
that the K bw is “shrunk” progressively with increasing K ow for a series of low-
polarity chlorinated benzenes, or alternatively that the difference between
(g* w /g* o ) oct-water and (g* w /g* o ) but-water increases with increasing K bw . The correlation
of logK bw (as logK sw,2 ) with logK ow (as logK sw,1 ) for the series of solutes yields
a < 1 (about 0.7) (Leo et al., 1971). This finding is expected because the log
K bw values for a group of solutes would fall into a shorter range than the cor-
responding logK ow values, owing to the higher polarity of butanol (over
octanol) and the greater butanol–water mutual saturation. In other words, as
the solute water solubility decreases, the solute solubility in butanol-saturated
water decreases (or the g* w increases) less rapidly than that in octanol-
saturated water because of the high butanol content; concomitantly, the solute