### Abstract:

Chemical equilibrium is a major factor in many natural and
industrial systems. The ability to predict ionic equilibrium is
extremely valuable in industry in solving problems without the costs
involved in full scale experiments. An example of an application in
the pulp and paper industry is the prediction of non-process element
solubility in process streams.
A computer program, ISIS, was developed to estimate the solubility
of inorganic salts in aqueous, inorganic solutions. The model
incorporates a two step Gibbs free energy minimization algorithm and
Pitzer's method for ionic activity coefficient prediction.
The program, ISIS, was tested on three cases. The test cases
were: KCl solubility in a NaCl and water solution, NaCl solubility
in a MgCl₂ and water solution, and CaSO₄ solubility in a NaCl and
water solution. The equilibrium predictions in the test cases were
very good, with the mean of the absolute value of the relative error
ranging from 4.3% for the first case, 17% for the second case, to
3.8% in the third case. ISIS accurately predicted the solid phase
amount and chemical composition in each of the test cases.
A parametric study was conducted on the three test cases to
examine the effects of the activity coefficient predictor, and of
uncertainty in chemical potentials, and third virial coefficients and
mixing terms in the Pitzer activity coefficient prediction model.
The effects of the activity coefficient predictor were determined by
comparing predictions made with the Pitzer model versus an extended
Debye-Huckel equation and an ideal solution assumption. The differences
were great, with the other activity coefficient predictors
resulting in errors greater than five times the error with the Pitzer
activity coefficient model. The effect of the chemical potentials
was large, especially in the trace species case, CaSO₄ in a NaCl and
water solution. A relative change of less than a tenth of a percent
in the solid species chemical potential resulted in an increased
error of ten times the original error. The effect of assuming the
third virial coefficients and the mixing terms to be zero in the
Pitzer activity coefficient predictor can be large. Errors up to 90%
in the mean activity coefficient were found.
It was concluded that the computer program ISIS could predict
accurately the solubility of inorganic salts in aqueous, inorganic
solutions. The accuracy of the prediction would be greatly affected
by the accuracy of the chemical potentials and the availability of
the third virial coefficient and mixing terms in the Pitzer activity
coefficient predictor. Future work is recommended in collecting
accurate chemical potentials and Pitzer interaction terms to increase
the database for ISIS and similar programs. The inclusion of ISIS
into a steady state chemical process simulator is also recommended.