The experimental part
of the work had two main directions of focus: multicomponent filtration
experiments (sections 4.1, 4.2 and 4.5) and electrochemical impedance
spectroscopy (EIS) measurements (sections 4.3 and 4.4). The combination of both
types of experiments enabled obtainment of a wider picture of membrane
separation behavior with information dominated by either slow (filtration) or
fast (EIS) ions in the membrane.
A widespread approach
used in modeling membrane transport uses a mean field theory. Models assume
that transport of salt and water takes place in nanopores or a homogneous gel,
from which the salt is excluded by a mean uniform potential caused by a
combination of Steric, Donnan, and diElectric exclusion mechanisms (SDE model).
This approach predicts certain dependence of salt permeability on salinity,
however, filtration data and other independent measurements of ion partitioning
indicate that it is not consistent with experiments and fails to predict
membrane performance for different salts and concentrations.
A surprisingly good
agreement between filtration experiments and model was obtained simply by
assuming constant ion permeabilities, e.g., separation of seawater ions by
nanofiltration (section 4.1). A still better agreement was obtained if some
variation may be allowed for divalent cations only, provided coupling of
transport of all ions is appropriately addressed. Using a similar model, pH
variations in a desalination process were modeled (section 4.2), and an
extremely high permeability of hydronium and hydroxyl ions to the RO membrane
was demonstrated, indicating a different exclusion mechanism for the latter
ions. The dependence of salt permeability on solution composition was further
examined in a system with different combinations of NaCl and CaCl2
at different pH values (section 4.5). Results indicated that ion permeability was affected by
ion specificity and different membrane affinity towards different ions. These
data suggest that treatment of ionic exclusion
mechanisms using conventional mean-field approaches is problematic.