Reactions and Separations

Apply Solubility
Theory for Process
Improvements
For solid compounds that dissolve in solvents
Michael J. Gentilcore
Tyco Healthcare but do not ionize, the thermodynamics of
solubility can provide valuable predictions
and insights.

ANY CHEMICAL PROCESSES DEPEND on The activity of a solid in solution is described by

M the solubility of solids in liquid solvents to de-

fine key process steps. Some common examples
include crystallization to isolate a product at good yields,
the following thermodynamic relationship (3):

recrystallization to remove impurities, reaction where a
solid reactant must be dissolved in order to participate in
a chemical conversion, and filtration where a dissolved
product solution is filtered from a solid catalyst. Solubili-
ty is often a major determinant of the economic effec-
tiveness of a particular process step.
Typically, solubility data are provided as experimental
data summarized in tabular or graphical form, and they
are often extremely limited in scope. Solubility theory
can easily be applied to many compounds that do not
ionize or dissociate to provide and predict solubility over
a wide range of conditions. This article reviews the ba-
sics of solubility theory and discusses how to apply solu-
bility theory to process improvement.
Solubility theory of non-electrolytes
The activity of a component in a solution is a multi-
ple of its mole fraction (3):
an = nxn (1)
The multiple is called the activity coefficient. This
is the same activity coefficient used to correct Raoults
law for predicting partial pressure above a solution:
pn = nxn pn0 (2)
d ln ans HnF
= (3)
dT RT 2
This relationship is similar to the Clausius-Clapey-
ron equation for vapor pressure of a pure liquid:
d ln p 0 H V
= ( 4)
dT RT 2
Substituting Eq. 1 into Eq. 3 and integrating, assum-
ing that the heat of fusion is constant from the melting
point temperature (Tm) to any other temperature (T), the
equation for the solubility of solids in solution is (3):
HnF 1 1
ln n x n = (5)
R T Tm
Equation 5 is applicable to compounds that do not
ionize or dissociate in solution and where the solid phase
is pure. It cannot be used for electrolytes like salts, or-
ganic acids and amines. Derivations are available for
cases where the heat of fusion is not assumed to be con-
stant (3). Systems where the solid phase is not pure are
those where the solvent has a true solubility in the solid
phase as opposed to where the solvent is simply en-
trapped or otherwise unremoved as free solvent that con-
taminates the solids. For extreme pressure changes (more

38 CEP March 2004 www.cepmagazine.org

 the scope of this article. Many of these methods require
Nomenclature computer subroutines to fit experimental data, and there-
fore are somewhat cumbersome when used for solubility
a = activity of a component in solution, dimensionless problems. Thus, this article will present only the van Laar
A = van Laar correlation parameter, cal/gmol correlation for simple two-component systems, which con-
B = van Laar correlation parameter, cal/gmol sists of the following two equations (1):
H = molal enthalpy, kcal/gmol
2
MW = molecular weight, g/gmol
A x1
p = partial pressure of component above solution, mmHg RT ln 1 = A1 + ( 9)
p0 = vapor pressure of pure liquid, mmHg B x2
R = universal gas constant = 1.987 cal/gmol/K 2
B x2
S = molal entropy, cal/gmol/K RT ln 2 = B1 + (10)
T = temperature, K A x1
V = molal volume, cm3/gmol
x = mole fraction of a component in solution, dimensionless By a few mathematical transforms, the van Laar corre-
X, Y = simplifying terms for van Laar correlation lation is especially useful in correlating solubility data
Greek Letters using linear regression. Equation 9 can be rearranged to:
= activity coefficient, dimensionless
1 A x1 1
= density, g/cm3
= + (11)
Subscripts RT ln 1 B x2 A
m = melting point
n = nth component Defining the following two mathematical transforms:
1 = first component
x1
2 = second component X= (12)
Superscripts x2
F = fusion, i.e., melting 1
L = liquid Y= (13)
S = solid
RT ln 1
V = vaporization
0 = standard state, pure liquid gives the equation:

A 1
than 100 atm), the change in solubility and melting points
Y= X+ (14)
B A
with pressure cannot be ignored. Derivations are available
for the change in solubility with pressure (3). Equation 14 is in the form of y = mx + b. A plot of trans-
The entropy of fusion is given by the equation (2): formed experimental data (Y vs. X) will be linear with inter-
cept 1/A0.5 and slope (A0.5)/B. If only a single solubility data
HF = TmSF (6) point exists, the van Laar equation can be reduced to a one-
parameter form by assuming the following relationship (1):
Substituting Eq. 6 into Eq. 5 give the following equiva-
A V1L
lent equation for solubility (1): = (15)
B V2L
S Eq.
F
Tmm6 into
1 Eq. 5 gives the following equiv-
F
ln Substituting
nn x nn =
nn
(7)
alent equation R for solubility
T (1): For the solid component, the molal volume of the pure
liquid melt is the proper volume to estimate i.e., not the
Finally, for ideal solutions, the activity coefficient is set specific gravity of the solid or the bulk density of the
to unity to give the ideal solubility equation (2): solid. The one-parameter simplification works best for
non-polar solvents and solutes (1). It is not reliable for
SnF Tm
ln x n = 1 (8) polar components.
R T
Examples
Activity coefficient correlations The following examples illustrate the general use of
Numerous methods to calculate an activity coefficient these equations.
are available (e.g., the Margules, Wilson, NRTL, UNI- Example 1 ideal solubility estimate. What is the ideal
QUAC and UNIFAC methods), but these are beyond the solubility at 25C of a compound that melts at 58C with a

CEP March 2004 www.cepmagazine.org 39

Reactions and Separations
heat of fusion of 8.1 kcal/gmol? Solution: Tm = 273 + 58 =
331 K and SF = (8.1/331)(1,000) = 24.5 cal/gmol/K. Solve
Eq. 8 for x: x = exp{(24.5/1.987)[331/(25 + 273) 1]} =
0.255 mole fraction. Comment: The ideal solubility is inde-
pendent of the solvent. The solvents molecular weight will
change the solubility on a weight-percent basis, but not the
mole fraction.
Example 2 activity coefficient estimate. For the com-
pound in the first example, the observed solubility at 25C
is 0.05 mole fraction. What is the activity coefficient? So-
lution: a = 0.255 and x = 0.05. Solve Eq. 1 for : =
0.255/0.05 = 5.1. Comment: Activity coefficients are de-
pendent on real data. The solvent was not stated in this ex-
ample, but the calculated activity coefficient is unique to
that solvent in combination with the solid compound.
Example 3 one-parameter van Laar correlation. For
the compound in the previous examples, the solvent is
CCl4 (MW = 154 g/gmol, = 1.58 g/cm3). The solid is
nonpolar (MW = 218 g/gmol). The density of the liquid
from the solid upon melting is 0.90 g/cm3. What are the
van Laar parameters for the solvent and for the solid?
Solution: Assign component 1 to be the solid and compo-
nent 2 to be the solvent. V1 = 218/0.90 = 242.2 cm3/gmol
and V2 = 154/1.58 = 97.5 cm3/gmol. Per Eq. 15, A/B =
242.2/97.5 = 2.484. RTln1= (1.987)(25 + 273)[ln 5.1] =
964.7 cal/gmol. x1 = 0.05 and x2 = 1 0.05 = 0.95. Solve
Eq. 9 for A: A = (964.7)(1 + 2.484[0.05/0.95])2 = 1,233.4.
B = A/(A/B) = 1,233.4/2.484 = 496.5. Thus, A = 1,233.4
cal/gmol and B = 496.5 cal/gmol. Comment: The same pa-
rameters are used for both the solid and the solvent. In this
example, Eq. 9 is used for the solid and Eq. 10 for the sol-
vent. The assignment of component 1 and 2 is arbitrary. If
more than one data point is available, obtain the van Laar
parameters by plotting 1/(RTln1)0.5 vs. x1/x2.
Example 4 partial pressure estimate. At 25C, the
vapor pressure of CCl4 is 114 mmHg. What is the vapor
pressure of a CCl4 solution at 25C (298 K) saturated with
the solid in the previous examples? Solution: First, find 2
using Eq. 10: 2 = exp {[496.5/(1.987 298)][1 +
(496.5/1,233.4)(0.95/0.05)]2} = 1.011. Then, Eq. 2 gives p2
= (1.011)(0.95)(114) = 109.5 mm Hg. Comment: Note that
solubility data have been used to estimate a vapor-liquid
equilibrium quantity. Activity coefficients from solubility
data can be used to estimate any property based on solution
theory such as boiling points or osmotic pressures (3).
Example 5 solubility estimate. What is the solubility
of the compound in the previous examples at 40C (313 K)
in CCl4? Solution: The following trial-and-error procedure
is used: guess x1, calculate 1 by Eq. 9, calculate x11 and
compare it to the value of x11 calculated from Eq. 7. If x1
= 0.33, then x2 = 1 0.33 = 0.67, and 1 =
exp{[1,233.4/(1.987 313)][1 + (2.484)(0.33/0.67)]2} =
1.49; x11 = (0.33)(1.49) = 0.49; by Eq. 7, x11 =
exp{[24.5/1.987][331/313 1]} = 0.49. Comment: Com-
mon ways to deal with trial-and-error calculations include
40 CEP March 2004 www.cepmagazine.org
using the goal seek feature in a spreadsheet or writing a
user-defined function. Notice how the solubility has dra-
matically increased for this compound (from 0.05 mole
fraction at 25C to 0.33 mole fraction at 40C) as it ap-
proached its melting point of 58C.
Physical properties
Only two physical properties are required to employ
solubility theory, heat of fusion and melting point. Thermal
analysis systems that perform differential scanning
calorimetry (DSC) and thermogravimetric analysis (TGA)
easily measure these properties if they are not otherwise
available in standard references.
DSC/TGA testing is quite common in research and pro-
cess development as part of the initial safety testing
screening of new compounds. Typically, the focus of these
tests is to assess thermal stability risks for compounds by
quantifying decomposition temperatures, mass changes on
decomposition, heat of the decomposition reactions, and
decomposition reaction kinetics.
With the predominant focus on collecting safety data, heat
of fusion and melting points may be observed but not report-
ed. If they are reported, their value in predicting solubility
and guiding process development is often overlooked.
Some good advice ask specifically for these proper-
ties when DSC/TGA work is performed.
Application of theory
As can be seen in the examples, a very minimal amount
of data (the melting point, the heat of fusion and a single
solubility data point) can quickly be used to predict the en-
tire solubility curve. When available, additional data points
showing temperature dependence will increase confidence
in the estimated solubility curve.
A solubility curve could be used to estimate yield losses
for material balance for a process with a crystallization/fil-
tration step. The ratio of the solubility of an impurity to
the solubility of a product could be used to define the best
crystallization temperatures and solvent amounts. A solu-
bility curve could also be used to define minimum purge
amounts of impurities required in a recycle process.
When optimizing solvent charges and crystallization
temperatures for a process, fresh solvents should not be ar-
bitrarily added to a process to prevent slurry concentra-
tions from becoming too thick. The amounts of solvents
added to a process should be based on what is needed to
control impurities based on solubility. Process mother
liquors can be recycled as diluents (rather than fresh sol-
vents) to control the slurry concentration.
Solubility theory could be used to evaluate operating con-
ditions for process improvement potential that are not easily
explored by a bench chemist (e.g., at elevated pressures and
temperatures), prior to committing to an expensive research
program. Higher solubility at elevated temperatures might be
taken advantage of for several reasons, including:
 higher payloads for a reaction where the product is kept
in solution and filtered away from other solids such as het-
erogeneous catalyst, solid absorbents, or byproduct salts
faster reaction kinetics, especially for reactions where
the solids are predominantly out of solution
reduced energy costs, solvent costs and waste disposal
costs, which can be achieved by processing with reduced
amounts of solvent present; this is especially true if the sol-
vent must be subsequently evaporated or distilled, or a reac-
tion byproduct must be stripped to complete the reaction
elimination of reaction solvent by conducting the re-
action of a solid with no solvent in a small excess of a liq-
uid reactant.
Solubility may also be used to estimate vapor/liquid
equilibrium or other properties predicted from solution
theory such as osmotic pressure.
Process improvement example:
Crystallization temperature
A process crystallizes a product from a solvent. The
final crystallization temperature is less than 10C to mini-
mize the yield loss from solubility. The process is run very
dilute to keep an impurity from co-crystallizing. The phys-
ical property information for the system is presented in
Table 1. Do you agree the cold temperatures and extra sol-
vent are the proper approach?
Answer: The solubility of the impurity will increase
faster with temperature than the solubility of the product,
as illustrated by the ideal solubilities shown in Table 2.
Increasing the crystallization temperature from 10C
to 25C and cutting the amount of solvent in the pro-
cess by one-half to two-thirds is the directionally cor-
rect approach to improving the process. The combined
change of warmer crystallization temperatures and half
the solvent will result in no increased product losses
and an improved ability to keep impurities in solution.
Estimating activity coefficients could further refine
these estimates.
Table 1. Physical property data for
crystallization temperature example.
Product Impurity
Melting Point, C 150 165
Heat of Fusion, cal/gmol-C 8 13
Table 2. Ideal solubilities for
crystallization temperature example.
Product Impurity
10C 0.0091 0.00028
20C 0.0185 0.00090
Increase in solubility 2.0 times 3.2 times3
Process improvement example:
Eliminating reaction solvent
A process dissolves a solid S in a solvent and then re-
acts it with a liquid L to create a product P according
to the following stoichiometry: S + 8L P. P has a melt-
ing point of 180C and a heat of fusion of 4 cal/gmol. L
has a normal boiling point of 110C. What is the best-case
usage of L if the reaction is conducted only in L at atmo-
spheric reflux conditions and the reaction must finish as a
uniform solution of P in L? For simplicity, ignore any
boiling point elevations.
Answer: The ideal solubility mole fraction of P at
110C is 0.445. The amount of L needed over stoichio-
metric requirements is (1 0.445)/0.445 = 1.25 moles
L/mole of P. The best-case usage is then 8 + 1.25 = 9.25
moles of L per mole of P, or 15% over theory. This usage
could be compared to that of the current process, the raw
material cost for the usage of original process solvent,
and the operating costs for any solvent recovery steps for
the original solvent.
Final thoughts
Process simulation software has made extensive in-
vestment in solution theory, especially for non-elec-
trolytes, for prediction of vapor/liquid equilibrium. An
activity coefficient prediction based on structure (UNI-
FAC) is a common feature in such software. However,
the extension of activity coefficient to the solubility of
non-electrolytes has often been overlooked in process
simulation software. CEP
MICHAEL J. GENTILCORE is an engineering consultant at Tyco
Healthcare Mallinckrodt (675 McDonnell Blvd., Hazelwood, MO
63042; Phone: (314) 654-4031; Fax: (314) 654-7174; E-mail:
mike.gentilcore@tycohealthcare.com). He has over 25 years of chemical
engineering experience in design, research and development, and
manufacturing, and he holds several patents. He received a BS in
chemical engineering from Clarkson Univ. and an MS in engineering
management from the Univ. of MissouriRolla. He is a member of AIChE
and a licensed P.E. in Missouri.
Literature Cited
1. Reid, R. C., et al., The Properties of Gases and Liquids, 4th
ed., McGraw-Hill, New York, NY (1987).
2. Castellan, G. W., Physical Chemistry, 2nd ed., Addison-
Wesley Publishing Co., Reading, MA (1971).
3. Hildebrand, J. H., and R. L. Scott, The Solubility of Non-
electrolytes, 3rd ed., Reinhold Publishing Co., New York, NY
(1950).
Website of Interest
www.chemstations.net/documents/examples.htm Readers
who use CHEMCAD may find the information at the Predictive
Crystallizer link on this site useful.
CEP March 2004 www.cepmagazine.org 41