Ind. Eng. Chem. Process Des. Dev., Vol. 18, No.

3, 1979 391

Double Salt Solubilities

H. P. Meissner
Massachusetts Institute of Technology, Cambridge, Massachusetts

C. L. Kusik’
Arthur D. Little, Inc., Cambridge, Massachusetts 02140

Solubility products of double salts in aqueous solutions are predicted using previously published methods for estimating
water activities and activity coefficients of the constituent simple strong electrolytes over the temperature range
of 0 to 250 OC or higher. Anhydrous and hydrated double salts such as glaserite (Na2SO4.3K2SO4)and schoenite
(MgS04.K2S04-6H20) are considered. Calculations at 25 OC are presented for the system Na,S04-K,S04-H20,
determining saturated solution compositions and identifying the associated equilibrium solid phases.

Objective for glaserite. Formulas per equivalent of contained ma-

Using previously published methods of predicting ac- terial for other typical double salts are listed in the last
tivity coefficients (Kusik and Meissner, 1978), solubility column of Table I. An advantage of this choice of one
limits in multicomponent aqueous solutions have been equivalent is that it keeps a, 0, and n relatively small, and
calculated for simple strong electrolytes, such as repre- thus avoids the appearance of large exponents in the
sented by the formula A,,B,;nH,O, where UA and vB are activity equations (below). Moreover, since one equivalent
stoichiometric coefficients and n represents the number of material must contain one equivalent of cations and also
of moles water of hydration, if any (Meissner and Kusik, one equivalent of anions to maintain electrical neutrality,
1973). Of equal interest are solubility limits of double salts the following always applies to a four ion solution
formed from three or more ions in solution. The gener- CYZAYA + P Z C Y C = ~ZBVB+PZDUD= 1 (2)
alized formula and dissociation reaction of such a double
salt formed from four ions is as follows S o l u b i l i t y P r o d u c t . For strong electrolytes, the log-
~(AuABuB).P(Cu,D,,).~HzO
9
arithm of the activity a, of a salt in solution equals the sum
of the logarithms of its constituent ions (and water activity)
avAAZA + avBBZB+ /3vcCzC + PvDDZ” + nH,O (1) suitably weighted by the stoichiometric coefficients. Per
Here, A and C represent cations, B and D represent anions, equivalent of double salt it follows directly from eq 1 that
and cy, P, and n are numerical constants, respectively a, =
representing the moles of simple salts AUABvB, CUCDYD, and (,A)”’”( mB)“”(mC)B”C( mD)B~(yAB)a(YA+YB)(yCD)B(uc+~)(a,)n
water per mole of double salt. The term Z represents the
absolute value of the charge on the ions identified by (3)
subscripts A, B, C, or D. Typical double salts containing where m and y represent the molality and the mean ionic
3, 4, and 5 ions are listed in Table I. . The object here is activity coefficient of the simple salt indicated by subscript
to present a method of calculating solubility limits of such while a and P are constants chosen so that eq 2 is satisfied.
salts. In saturated solutions, the activity of any electrolyte
F o r m u l a Selection becomes identical with the so-called thermodynamic
In a double salt like glaserite, which is derived from two solubility product K , The solubility product is usually
cations and one anion, the anions represented by B and considered as applicaKle to sparingly soluble salts such as
D in eq 1 are identical. This double salt may have been BaS04 since at low concentrations, activity coefficients of
written in various ways, three of which are as follows: (a) the ions in solution are unity. However, the concept is
NaZSO4.3K2SO4;(b) 1/3Na2S04.K2S04; and (c) l/sNaz- equally applicable to electrolytes of higher solubility,
S04.3/,K2S04. These three expressions are equally valid, providing that appropriate values of activity coefficents
being simple multiples of each other. In all three formulas, are used (Denbeigh, 1955).
it is obvious that uA, vB, uc, and UD are respectively 2, 1, 2, Equation 3 appears formidable upon first inspection, but
and 1, with the ratio P / a equalling 3. is easily applied to determine activities and solubility
Obviously, the calculated molality of glaserite in a products. Thus, for one equivalent of glaserite (l/sNaz-
particular solution depends on whether formula a, formula S04.3/sKzS04), in which n is zero since this double salt
b, or formula c is chosen for reference. Thus, a given carries no water of crystallization,
solution which is 0.1 m in glaserite by formula a is 0.3 m K,, = ~ ~ ~ ~ 2 ~ 8 ~ ~ 6 ~ 8 ~ s ~ ~ sol.
~ s ~ 3 ~ s
by formula b, and 0.8 m by formula c. Ion molalities in
this solution, however, are the same with all three formulas, (4)
being 0.2 m in sodium ion, 0.6 m in K+ ion, and 0.4 m in T o illustrate, consider a solution of 25 “C (D’Ans, 1933)
S042-. Clearly, in defining double salt molalities, it is containing only K+, Na+, and SO4’- ions, having respective
necessary to stipulate a, 0, and n of the salt under dis- molalities of 1.52, 0.94, and 1.23, which is saturated with
cussion as well as its molality, to avoid misunderstanding. both glaserite and arcanite (K2S04).The ionic strength
In dealing with double salts, it is convenient to choose of this solution is 3.7, while values of Y N ~ ~ Sand
O ~ Y K ~ S Oin
~
that formula which contains one gram-equivalent of this multicomponent solution are each 0.193 as calculated
material. In this discussion, therefore, formula c is chosen by methods previously published (Kusik and Meissner,
0019-7882/79/1118-0391$01 .OO/O 0 1979 American Chemical Society
392 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

Table I. Typical Double Salts

common formula formula per g-equiv
burkeite 2Na,S0,~Na,C03 ’ Na SO .l/sNa,CO
carnallite KC1.MgC1,*6H20 /3kCf I / 3hgC1,.2H2d
darapski te NaNO,.Na,SO, I / 3 NaNO,.’/ Na,SO,
glaserite Na,S0,~3K,S04 1/8Na2S04.3/8KZS04
hanks i te 9Na,SO,. 2Na,C03.KCl 9/~3Na,S0,~2/23Na,C03~1/~3KCI
kainite KCl.MgS0,.3H2O l/sKCl*l/~MgSO,~H,O”
mendozite 0.5Na,SO4~O.5A1,(SO,),~11H,O 1/~Na,S0,~1/8Al,(S0,)3~~~/4H,0
pol yhalite K,SO,*MgSO,. 2CaS0,. 2H,O ‘ I S K , S O , ~*MgSO,.I/
~/ CaSO,.’/ ,H,O
teepleite Na,B,04~2NaC1~4H,0 ‘ I 4Na,B,0,. l/ ,NaCl.H,O
As with all double salts containing 4 ions or more, kainite’s formula per equivalent could have been written in an alterna-
tive way such as: 1/6K2S04.1/6MgS04.1/6MgC12.Hz0. It follows that alternative ways of writing a, (eq 3) for kainite
include: a, = m K 1 i 3 m M g 1 i 3 m c 1mS0,’1 i 3 ’ 3(YKC1)2’ ’(7MgSO, )” ’(awl = K I ’ ’mMg I ’ %l” ’“SO, I ’ ’(7K,SO, )’ ’ ’(7MgSO, ) I ’ ’-
( y M g C 4 ) 1 / 2 ( a w ) .With either formula values of a, are equivalent if y and ow are obtained by equations for multicomponent
solutions presented earlier (Kusik and Meissner, 1978).
Table 11. Calculated Thermodynamic Solubility Product for Glaserite Saturated Solutions (25 “C)from Eq 4
glaserite :
molalitiesa ionic calcd values
solid phase(s) in strength of K , at 25 “ C
addition t o elaserite“ K’ Na’ SO”,+ MCZ+ c1- I (eauiv basis)
K2S04 1.52 0.94 1.23 0 0 3.69 0.127
Na,SO,.lOH,O 1.02 4.48 2.75 0 0 8.25 0.136
NaC1, Na,SO, 1.10 6.50 0.81 0 5.99 8.41 0.117
KC1, Naci ~ 2.22 5.40 0.24 0 7.13 7.86 0.114
KCl, K,SO, 3.90 1.80 0.15 0 5.41 5.86 0.107
Na2S04, 0.84 5.81 1.99 0 2.67 8.64 0.120
Na,SO,~lOH,O
KC1, NaCl, Scc 1.58 2.56 0.78 1.97 6.52 10.8 0.138
Ast,b NaCl, 0.91 5.32 1.22 0.88 5.54 10.1 0.117
Na,SO,
KC1, K,SO,, Scc 2.64 1.02 2.56 1.42 5.42 12.5 0.133
KC1, Scc 1.96 1.97 0.65 1.73 6.08 9.77 0.136
Ast, NaC1, Scc 1.17 3.06 1.03 1.94 6.06 11.1 0.128
average 0.125
a D’Ans (1933). Sc = Schoenite, 1/4KZS04.1/4MgS04.3H20.
Ast = Astrakanite, ]/4Na2S0,.1/4MgS04.2Hz0.
1978). Substituting into the above expression, a, and K,, gram-equivalent of total anions (or cations) in solution; the
for glaserite a t 25 “C is found to be 0.127. fraction “f,” of the total equivalents of cations present in
Inspection shows that the composition of the solids (dry solution which is represented by each cation; and the
basis) dissolved in the saturated solution just mentioned fraction “f,” of the total equivalents of anions present
is not the same as that of glaserite itself. Indeed, at 25 “C which is represented by each anion. The total equivalents
it is found experimentally that the composition of a so- of anions present, of course, equal the equivalents of the
lution on a dry basis is never that of glaserite when cations.
glaserite is the equilibrium solid phase. As a result, Regardless of the number of ions present, such a solution
glaserite is said to be “incongruently” soluble a t this at a given temperature having a known set of values of w,
temperature. This finding is not an uncommon one for f,, and f, can be readily tested for saturation with a given
double salts and in no way prevents the use of eq 4 for compound. This involves calculation of the compound’s
calculating saturated solution compositions from known activity in this solution by substituting its appropriate ion
values of K , . molalities and activity coefficients into eq 3. If the value
The fact t i a t the solubility product (and activity) of a so obtained is smaller than this compound’s activity
double salt at a given temperature remains the same in all (solubility product) a t saturation, then the solution being
saturated solutions, regardless of the type and concen- tested is less than saturated. When a, and Kspare nu-
tration of any other ions present, is illustrated in Table merically equal, then the solution is saturated. When a,
11, where values of (K,,) for glaserite are listed in various is greater than Ksp,then the solution is supersaturated, and
solutions saturated a t 25 “ C with this double salt and other a t equilibrium, the salt will precipitate.
salts. The solubility products in each of these saturated Various techniques can be used to present isothermal
solutions were calculated from eq 4 by the method just data on compositions of saturated solutions (Meissner,
described. Inspection shows that despite the wide vari- 1971). Solutions containing three ions, such as Na’, K’,
ation in both the compositions and ionic strengths of these and SO:-, are conveniently represented by a diagram such
saturated solutions, calculated values of K,, for glaserite as Figure 1. Thus the experimentally determined com-
a t 25 “C in all these systems remain relatively constant, positions a t 25 “ C of saturated solutions for the K+-
varying within 15% of the average value of 0.125. Na+-SO?--water system are shown by the triangular and
diamond shaped points of Figure 1 (D’Ans, 1933; Seidell
Phase Diagrams and Linke, 1965) and contrasted with the (circular) points
Whether saturated or not, solution compositions are as calculated by the method outlined below. Glaserite is
conveniently expressed as gram-equivalents of each ion per the stable phase along the dotted line (which is drawn
1000 g of water. Obviously, a t a fixed temperature, a through the calculated points) with K2S04, the stable
solution containing more than two ions can be charac- phase on the solid line indicated on the left-hand side of
terized by: the value of w , the grams of water per the figure and Glauber’s salt on the solid line to the right
 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 393

Figure 1was constructed by the procedure just outlined.

Inspection of the calculated points shows these to be close
to the experimental points for solutions saturated with one
or two solid phases. Similar procedures can obviously be
used for strong electrolyte systems containing more than
three ions and at temDeratures other that 25 "C (Meissner
et al., 1972, 1977).
Solubility Product Expression Derivation. The
3 following three-step isothermal reversible process is in-
volved in showing that eq 3 is the thermodynamic solu-
bility product of one gram-equivalent of a double salt in
its saturated solutions, and in showing this product to be
constant.
Step 1. Transfer 1g-equiv of the double salt (eq 1) into
the solution saturated with the double salt in question.
This involves adding CY moles of the simple salt AvABQand
P moles of simple salt C,,D, from their respective hy-
pothetical standard state soyutions, in which their re-
spective activities are, of course, unity. The Gibbs free
energy change for the transfer of CY moles of A, B, is now:
aRT In ((mA)YA(mB)Q(y)U~I and for P moles of &,D, is PRT
In ( ( m c ) u ~ ( m D ) ~ ( y c ~The
) u ~quantities
~). TAB and YCD are
the activity coefficients of the indicated electrolytes in this
Figure 1. Comparison of predicted and experimental solubilities at saturated multicomponent solution, as calculated by the
25 "C for the Na2S04-K2S04system. methods mentioned earlier.
Step 2. At the same time, transfer n moles of water
of the figure. The calculated solution compositions which from its standard state (pure liquid) into the saturated
can coexist with two solid phases (K2S04-glaserite, and solution is question. The free energy change is nRT In (a,)
glaserite-Glauber's salt) are shown by the two unshaded where a, can be calculated by published methods (Kusik
circles. and Meissner, 1978). Obviously, if the solid double salt
General procedure in calculation is now illustrated with carries no water of hydration, then n is zero.
an arbitrarily chosen dilute solution at 25 "C, for which Step 3. Simultaneously allow one equivalent of the
fNa+ and f F + are to be 0.75 and 0.25, respectively, while fso42- double salt formula (1) to precipitate. The Gibbs free
of course is 1.0. Such a solution can be made by dissolving energy change of this step is zero.
in water a small amount of salt mixture containing a ratio Upon combining terms, the sum of the free energy
of 8 mol of Na2S04to 1 mol of glaserite (Na2S04.3K2S04). changes to steps 1, 2, and 3 becomes
Upon evaporation, the object is to find the grams of water,
w, when this solution first reaches saturation of 25 "C, and .IG = R T In { m A a u ~ m B a v ~ m C ~ v ~ m D ~ ~( 5y) ~ ~ a u ~ ~ y
to identify the solid phase(s) which forms, recognizing that For a fixed temperature and regardless of any other ions
values of K,, for glaserite, arcanite, and Glauber's salt are present, the free energy change of eq 5, per gram-equiv-
0.125, 0.14, and 0.38, respectively. alent of double salt, is always the same from the standard
As a first step, arbitrarily pick a value of w,such as 833 state solutions to all solutions saturated with the double
g of water per equivalent of cations (or anions) in solution. salt in question. The quantity in brackets in eq 5 is,
The sodium ion molality mNa is then 0.75 X 1000/833 or therefore, also a constant for all these saturated solutions,
0.9, while mK is 0.3, and mso, is 0.60. Using the methods and is the solubility product Kspper equivalent of double
described earlier, values of y for K2S04and Na2S04are salt. For these saturated solutions it is equal to the
estimated in this multicomponent solution as 0.247 and electrolyte activity a, as presented in eq 3.
0.248, respectively. Substituting into eq 3, a, is found to Note that this development applies or can be extended
be 0.04 for glaserite. Similarly, after determining the water to all double salts, regardless of whether they involve three
activity by earlier published methods (Kusik and Meissner, ions (carnallite and glaserite), four ions (kainite and PO-
1978) the values of a, for arcanite (K2S04)and Glauber's lyhalite), five ions (hanksite), etc., and regardless of
salt are 0.03 and 0.08, respectively, in this solution. whether waters of hydration are present or not.
Since in each case a, is far smaller than the corre- Precision. Errors in estimating water activities or
sponding Ksp,the solution is far from being saturated with activity coefficients of electrolytes in multicomponent
any of these three salts. An arbitrary amount of water is solutions are generally within 20% with occasional values
therefore next removed, so that w is now 750 g of water showing larger deviations (Meissner and Tester, 1972;
per equivalent of cations making mNaequal to 0.999 and Meissner and Kusik, 1972, 1973). Errors of this magnitude
mK equal to 0.333 mol/ 1000 g of water. Estimating values can significantly affect calculated values of (Ksp),especially
of y for K2S04and Na2S04as before, determining a,, and for double salts, since these exist in equilibrium with mixed
calculating a, for all three salts, electrolyte activities a, for (multicomponent) solutions. Solid solutions could in-
each salt are still found t o be smaller than the corre- troduce another source of potential error. Thus, the re-
sponding solubility products Ksp. Finally, when w is chosen lationships proposed here should be used only for ori-
as 500 g of water, the mNaand mKare respectively 1.5 and entation purposes when direct experimental evidence is
0.5 and the value of a, for glaserite in this solution is found not available.
to be almost identical with its solubility product, namely
0.125. Calculations, therefore, indicate that this solution Nomenclature
is saturated with glaserite, but not with K2S04or Glauber's a, = activity of a dissolved electrolyte
salt, because values of a, for these latter two salts lie well a, = activity of water, namely p l p " , where p is the water vapor
below their respective solubility products ITsp, pressure over the salt solution and p" is the water vapor
394 Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979

pressure over pure water at the same temperature v i = stoichiometric coefficient: see eq 1; also vij = v i + vi
fi = fraction of total cations in solution represented by cation AG = Gibbs free energy change
i
f, = fraction of total anions in solution represented by anion Literature Cited
j D'Ans, J., "Die Losungsgleichgewichthe der Systeme der Salzeozeanischer
I = ionic strength = ~ O . 5 m , Z I 2 Salzablagerungen", Verlagsgesellschafl fur Ackerbau, Berlin, 1933.
K,, = solubility product for one equivalent of salt with ex- Denbeii, K. G., "Principles of Chemical Equilibrium", p 308, Cambridge Unhrersity
Press, 1955.
amples shown in Table I Kusik, C. L., Meissner, H. P., AIChE Symp. Ser. 173, 7 4 , 14-20 (1978).
m, = molality, g-mol (of the ion indicated by subscript)/1000 Meissner, H. P., "Processes and Systems in Industrial Chemistry", pp 46-61,
g of water Prentice-Hail. Englewood Cliffs, N.J., 1971.
n = number of moles of waters of hydration (crystalliza- Meissner, H. P., Kusik, C. L., AIChf J., 18, 294 (1972).
Meissner, H. P., Kusik, C. L., Ind. Eng. Chem. Process Des. Dev., 12, 205
tion)/g-equiv of solid double salt (1973).
w = g of water per total equivalents of cations or anions in Meissner, H. P., Tester, J. W., Ind. Eng. Chem. Process Des. Dev., 1 1 , 128
a saturated solution (1972).
2, = ionic charge on ion represented by subscript Meissner, H. P., et ai., AIChE J., 18, 661 (1972).
Meissner, H. P., Kusik, C. L., Field, E. L., "Estimation of Phase Diagrams and
a = number of moles of simple salt A,B,/g-equiv of double Solubilities for Aqueous Multi-ion Systems", presented at 70th Annual AIChE
salt (eq 2 must be satisfied) Meeting, New York, N.Y., 1977.
/3 = number of moles of simple salt C,,D,,/g-equiv of double Seidell. S., Linke, W. F., "Solubilities", Voi. 11, 4th ed, pp 296-316, American
salt (eq 2 must be satisfied) Chemical Society, Washington, D.C., 1965.
y = mean ionic activity coefficient of the ion pair, indicated
by subscript, in the multicomponent solution under con- Received for reuiew December 23, 1977
sideration Accepted January 2, 1979

Use of the Vaporization Efficiency in Closed Form Solutions for

Separation Columns
Phillip C. Wankat" and Joseph Hubert
School of Chemical Engineering, Purdue University, West Lafayette, Indiana 4 7907

Modified forms of the Winn, Fenske, and Kremser equations utilizing a constant vaporization efficiency are derived.
The modified forms of the Smoker and Underwood equations are presented but not derived. These equations allow
the rapid calculation of required number of stages without requiring the assumption of equilibrium stages. They
also allow estimation of the average vaporization efficiencies from data on column performance, and then simulation
of column performance under different operating conditions.

Introduction yAj = EA]KAjXAj (2)

Accurate solutions for staged separation systems can be
obtained on computers by using various stage-by-stage or where KAj = KAj*yAj includes the ideal solution K value
successive approximation methods. Although the modern and the activity coefficient evaluated using the mole
computer techniques are quite rapid, they still require too fractions of the liquid leaving plate j . The vaporization
much computer time for problems such as preliminary efficiencies are bounded and can be estimated from pilot
economic estimates or reactor recycle calculations where or plant tests, or from models of separation plates (Hol-
a large number of iterations through the separator are land, 1975). The vaporization efficiencies also have the
required. For these problems, for preliminary hand cal- advantage that they are easy to incorporate in calculations.
culations, and for simulating existing installations closed For binary systems the vaporization efficiencies are not
form solutions such as the Kremser, Smoker, Fenske or equal unless the stage is an equilibrium stage.
Underwood equations are very useful. These equations In this paper the vaporization efficiency will be used to
are usually derived with the following assumptions: derive modified forms of the Winn and Fenske equations
constant flow rates, constant relative volatility or linear for total reflux in distillation columns and the Kremser
equilibrium isotherms, and equilibrium stages. The last equation for linear equilibria. The modified forms of the
assumption can be relaxed if a stage efficiency can be Smoker and Underwood equations will be presented
incorporated. However, the usual Murphree efficiency is without derivation. The use of these equations to de-
usually difficult to utilize (the Kremser equation is an termine the vaporization efficiencies from operating data
exception (King, 1971)). is also developed. A sample calculation using these
An alternate efficiency which has frequently been used equations will be presented.
in multicomponent separation calculations is the vapor- Total Reflux Equations
ization efficiency. The vaporization efficiency, EAl, can be Equations for total reflux in distillation have been
defined as (Holland, 1975) presented by Winn (1958) and Fenske (1932) (or see King,
1971; or Smith, 1963). Both of these equations are easily
modified to include the vaporization efficiency. The
where the vapor is assumed to form an ideal solution. We derivation will be done for the Winn equation since the
will rewrite eq 1 as Fenske equation is a special case of the Winn equation.
0019-7882/79/1118-0394$01.00/0 0 1979 American Chemical Society