Group Contribution Method For Henry's Law Constant of Aqueous Hydrocarbons
Group Contribution Method For Henry's Law Constant of Aqueous Hydrocarbons
Group Contribution Method For Henry's Law Constant of Aqueous Hydrocarbons
Introduction
Reliable knowledge of solubility and vapor-liquid parti- waste. Importantly, many of these operations and processes
tioning of hydrocarbons in aqueous systems over a wide range occur at superambient conditions, thus creating a need for
of temperature and pressure is essential for conception of convenient estimation methods allowing to predict the re-
industrial operations, as much as for understanding environ- quired thermodynamic data. The property of high practical
mental and geological processes. These include, among oth- interest is Henry’s Law constant ŽHLC. which is a direct mea-
ers, modification of petroleum composition by oil field waters sure of volatility of a given solute in a solution. This property
in sedimentary basins, interaction between groundwaters and expressed as a function of temperature and pressure is also
reservoir fluids, pollution of aquatic and atmospheric envi- useful for correlating or predicting solubility of a hydrocar-
ronment during storage, transport and treatment of bon in water provided that the fugacity of the coexisting or-
petroleum and natural gas, depollution of petrochemical ef- ganic phase can be calculated.
fluents by air stripping, and the use of high-temperature wa- HLC is used in various contexts by physical chemists,
ter as a medium for decomposition of hazardous organic chemical and environmental engineers, and geochemists.
Each community developed its own terminology and ap-
Correspondence concerning this article should be addressed to V. Majer. proaches specific for the given field and a variety of HLC
tions and it is not always clear from literature how HLC was
determined. In the case of aqueous organic compounds Žand where f s is fugacity of the solute and x s is its molar fraction.
hydrocarbons particularly ., it is useful to adopt a functional This one-phase definition of HLC can be rewritten in two
group additive scheme. The reason is obvious: while the simple relationships. When considering the standard state of
number of these compounds is huge, they consist of just a ideal gas for the solute, it follows that
few functional groups. In the chemical thermodynamics of di-
lute aqueous solutions, this principle has been utilized with K H Ž T , p . s pT
s Ž2.
success at the temperature of 298 K, and several methods are
available that vary in the set of selected functional groups where T s is the fugacity coefficient of solute at infinite dilu-
and in the type of thermodynamic property expressed. tion, which can be calculated from an equation of state for a
Extension of estimation techniques to elevated tempera- solution. Similarly, the introduction of the standard state of
tures faces more difficulty for two reasons. First, there is a pure liquid leads to relationship
lack of reliable experimental results for various solutes in a
range of conditions, and second, there is a need of a good
K H Ž T , p . s f s ␥s⬁
䢇
Ž3.
thermodynamic model for the description of temperature and
pressure dependence of HLC. For aqueous hydrocarbons, lit-
tle systematic experimental work was performed at superam- where f s and ␥s⬁ stand for fugacity of pure liquid solute Žreal
䢇
bient conditions with the exception of the major contribution or hypothetical . and the limiting activity coefficient in sym-
from Tsonopoulos and collaborators ŽTsonopoulos and Wil- metrical standard state convention, respectively. This latter
son, 1983; Heidman et al., 1985; Economou et al., 1997; parameter can be expressed from models characterizing non-
Tsonopoulos, 1999, 2001.. Attempts to correlate the data in a ideality of a solution in terms of molar excess properties.
range of conditions remained mainly on empirical grounds. HLC is also simply linked with the Gibbs free energy of
Only in the past decade or so, new approaches have ap- hydration ⌬GThyd corresponding to the transfer of a solute
peared inspired by statistical thermodynamics and the theo- from an ideal gas state at the reference pressure Ž pref s 0.1
ries of near critical dilute solutions that allowed to build MPa. to the standard state of infinitely dilute solution at
high-temperature correlations on a more solid basis. Also, the pressure p. ŽThe standard state adopted for aqueous species
high-temperature models developed by geochemists for pre- is unit activity in a hypothetical solution of unit molar frac-
dicting the standard molar properties of aqueous solutes can tion referenced to infinite dilution, denoted here with super-
be adopted for calculation of HLC. script o.. The relationship is obtained by writing relationships
The purpose of this work was to establish a new group for the chemical potential of a solute using the standard state
contribution scheme for aqueous hydrocarbons that would of infinite dilution and that of ideal gas and combining them
take into account all types of experimental results for these with the definition of K H ŽEq. 1.
solutes available in literature. Use of a thermodynamic model
related to fluctuation solution theory makes possible reliable K H ŽT , p.
sGTs Ž T , p . yGsi g Ž T , pref . s ⌬GThyd
calculation of HLC up to the temperature of 573 K and pres-
sures to 100 MPa, and provides realistic predictions in the
RT ln
ž pref / Ž4.
to use relationships giving a realistic description of the where w is the compressibility of water and n s is the mole
derivative hydration properties ŽEqs. 5 to 7.. Several authors number of solute in the volume of solution V. This dimen-
presented empirical equations for correlating HLC of gases sionless parameter changes much less with density of solvent
in water along the saturation line of solvent respecting nega- than the partial molar volume and is well behaved in the crit-
tive divergence of the temperature derivative of HLC at the ical region of water. It can be used for expressing the fugacity
critical point of solvent Žsee, for example, Krause and Benson coefficient of solute at infinite dilution ŽO’Connell et al., 1996.
Ž1989., Fernandez-Prini and Crovetto Ž1989., and Crovetto
allowing to calculate HLC from Eq. 2. When rewriting the
Ž1991...
ATs,w parameter in the form of a virial expansion in density,
Harvey and Levelt-Sengers Ž1990. added an empirical term we get
to Eq. 8, which allowed correlation of HLC down to room
temperature. They have also shown ŽHarvey et al., 1991. that
this relationship is not restricted to the solvent saturation line, ⭸ Ž pVrRT . 2
ATs,w s lim w Bs,w q . . .
but can be applied for correlating HLC and hydration prop-
erties as a function of temperature and pressure
ns™ 0 ž ⭸ ns / T ,V
s1q
ž /
Mw
Ž 16.
T ln w K H Ž T , p . rf w x s Aq B Ž w y c,w .
where Bs,w is the second cross virial coefficient. An analo-
qCTw exp w Ž 273.15yT . r50 x Ž 12. gous procedure can be adopted for pure water where
solvent-solvent interactions are expressed in terms of a mod-
This simple three-parameter equation is not, however, flexi- ulus A w,w
ble enough for allowing quantitative description of the stan-
dard derivative hydration properties using Eqs. 5 to 7 ŽHarvey, 2
w Bw ,w q . . .
1998; Sedlbauer and Majer, 2000.. For engineering applica-
tions, Harvey Ž1996. proposed a simplified form of Eq. 12 for
A w,w s1q
ž /
Mw
Ž 17.
ATs,w s dA w ,w q Ž 1y d . q w aq b Ž exp w w x y1 . where the superscripts pref and T indicate the constant vari-
able at which the integration is performed. The two refer-
q c exp w ⌰rT x q ␦ Ž exp w w x y1 . 4 Ž 19. ence state terms are expressed from an independent scheme
available in literature Žlit. ŽPlyasunov and Shock, 2000a., while
where a, b, c, d are four adjustable parameters, ␦ s 0.35a the sum of three integrals is obtained from our temperature
and ®, ⌰ and are generalized constants that were obtained dependent model Žmod.. Then one can simplify Eq. 22 to
earlier ŽSedlbauer et al., 2000. from analysis of a large amount
of data on various aqueous solutes. This relationship has a
correct ideal gas limit for low densities and is analytically in- ⌬GThyd Ž T , p . s ⌬GThyd
lit
Ž Tref , pref . y Ž T yTref . ⌬ SThyd
lit
Ž Tref , pref .
tegrable to obtain properties on the Gibbs free energy level
Žsee the Appendix.. The particular functional form has been q ⌬GThyd
mod
Ž T , p . y ⌬GThyd
mod
Ž Tref , pref .
inspired by an older model of O’Connell et al. Ž1996.. Al-
though this development was originally made in connection y Ž T yTref . ⌬ SThyd
mod
Ž Tref , pref . Ž 23 .
with an application in geochemistry, we found this approach
appealing also as a basis for prediction of HLC over a wide
The entropy of hydration ⌬ SThyd
lit
ŽTref , pref . is calculated simply
range of temperature and pressure.
as
ment of predictive capabilities of the new group contribution calculated using the corresponding state method of Lee and
scheme Žsee below.. Kesler Ž1975. as described by Degrange Ž1998.
Deri©ati©e properties ⌬ HT
hyd s lim Ž ⌬ Hsolrx s . q ⌬ Hres . Ž 29.
xs™ 0
The derivative properties ⌬ HT
hyd , ⌬CTp,hyd
and VsT
are ob-
tained from results of calorimetric and volumetric experi-
ments with dilute aqueous solutions. The data for aqueous Values of partial molar volumes at infinite dilution VsT
hydrocarbons are rare at near ambient conditions where their ŽTable 6. obtained mainly from density measurements on vi-
determination is complicated due to low solubility of hydro- brating tube densimeters are numerous and present over 20%
carbons in water; they are, however, quite frequent for aque- of the database with the data both at near ambient and high-
ous alcohols. Particularly, a limited amount of high tempera- temperature conditions. Heat capacities of hydration were
ture data is available due to measurements performed at the calculated as a difference between the experimental partial
University of Delaware ŽWood and collaborators. and at the molar heat capacity at infinite dilution CTp, s and the heat ca-
Blaise Pascal University in Clermont-Ferrand, France ŽMajer pacity of an ideal gas C p,igs estimated from the group contri-
and collaborators.. The data on derivative properties ex- bution method of Joback, as presented in the monograph of
tended in a few cases Žhexane, cyclohexane, benzene and Reid et al. Ž1988.. Most data for CTp, s resulted from flow
toluene. to 643 K and supercritical pressure of 30 MPa. The calorimetric experiments on the Picker-type flow calorime-
purpose of their inclusion was to guide correlation of HLC ters performed at near ambient conditions and for three
correctly in the direction of the critical point; however, no compounds also at elevated temperatures ŽTable 7.. For both
near critical experimental results were considered. VsT and CTp, s , we have mainly included primary data sources
Most data on enthalpies of hydration ŽTable 5. at near am- listing measurements at several temperatures. Multiple
bient conditions were derived from articles by the groups of sources presenting the data at 298.15 K only were avoided
I. Wadso¨ at the University of Lund and S. J. Gill at the Uni- with an exception of four compounds for volumes and 12
versity of Colorado. They obtained ⌬ HT hyd combining unique compounds for heat capacities where we also included values
calorimetric data on dissolution of hydrocarbons in water with from representative compilations by Hoiland Ž1986. and Ca-
enthalpies of vaporization. The high-temperature values were bani et al. Ž1981.. These secondary data were used for com-
pounds including group contributions that are only scarcely terest, such as
present in the data set, namely the quaternary and ternary
carbon functional groups and the unsaturated bond, in order N
to improve representation of these functional groups in the as Ý n i ai and so on Ž 30.
correlation. is1
2
O ⌬GThyd
,exp
y ⌬GThyd
,calc
Fs Ýž
is1 ⌬GThyd / i
2 2
P ⌬ HThyd
,exp
y ⌬ HThyd
,calc Q
⌬CTp,exp
,hyd y ⌬C p ,hyd
T,calc
q Ýž ⌬ HT / q Ý ž ⌬CTp ,hyd /
js 1 hyd j k s1 k
T,exp T,calc 2
R Vs yVs
q Ý
ls1
ž VsT / l
Ž 31 .
y ⌬ HT mod
hyd Ž Tref , pref . Ž 32.
for maintaining consistency with the enthalpic data at the ref- temperature and, thus, the distinction is made for orthodi-
erence conditions generated from the group contributions alkylbenzenes and cis-dialkylcycloalkanes in our high-tem-
determined by Plyasunov and Shock Ž2000a.. Expressions used perature predictions.
for CTp,hyd and VsT are directly the model equations listed The estimated uncertainties YT in Eq. 31 play a role of
in the Appendix, without any constraint at the reference weighting factors reflecting reliability of experimental data
conditions. In this way we obtained five parameters of the and error margin within which we want to reproduce the given
temperature-dependent SOCW model for ten hydrocarbon property. In assignment of these weighting factors it is in-
functional groups in Table 2 Žthe value of the OH group is evitable to also take into account the inherent error of the
considered only as provisional.. Compared to the reference functional group additivity assumption. The weighting scheme
state group contribution scheme, we omitted the correction employed here intended to yield the most reliable predictions
for proximity effect IŽC-C. due to an insufficient amount of over the entire temperature and pressure range, and to bal-
data at elevated conditions for relevant hydrocarbon struc- ance the impact of different properties on the overall fit. The
tures. However, the correction at reference conditions also uncertainties were therefore assigned to each property in such
probably accounts for most of the proximity effect at higher a way that, on average, the data on this property were de-
scribed by our model at the limits of the estimated uncer-
tainty. In the case of ⌬GThyd the data sources on HLC were
divided into five classes depending on data quality Žwhich
sometimes changed with temperature . and the assignment of
uncertainties was done using a simple relation ⌬GThyd s 0.2
q0.1 ⴢ C wkJrmolx where C is the uncertainty class ranging
from 1 to 5. This error margin corresponds to about 10᎐40%
relative error in K H . The uncertainties assigned to derivative
properties were in most cases somewhat higher than the ex-
pected experimental errors in calorimetric and volumetric
data. The data of highest accuracy were those available for
standard molar volumes and heat capacities of aqueous alco-
hols near ambient conditions. However, directly using experi-
mental uncertainties associated with these data in our simul-
taneous correlation would lead to overweighting of derivative
properties at the expense of a good fit to HLC. The values of
YT for derivative properties were finally set to 8% for
⌬ HT Ž . T
hyd but at least 1.2 kJrmol , 5% for ⌬C p,hyd at least 13
Ž
T 3
JrKrmol., and 0.8% for Vs Žat least 0.5 cm rmol.. It should
Figure 3a. Henry’s Law constant of homologous series be noted that the adopted weighting scheme at the same time
of alkylbenzenes: test of methylene group provides an assessment of the expected prediction uncertain-
addition. ties of K H and of the derivative hydration properties for
Displayed are experimental data from major sources. aqueous hydrocarbons which were treated by our group con-
tribution procedure, or which are reasonably similar by their along the saturation line of water are given in Figure 1.
structure to our training set used in correlation. Structural elements governing total ⌬GThyd for hydrocarbons
The plots of temperature evolution of the functional group Žmethyl, methylene, c-CH 2 and CH ar groups. exhibit a maxi-
contributions to ⌬GThyd and to its first derivatives calculated mum in the range from 400 K to 450 K that is consistent with
the occurrence of maximum in HLC for hydrophobic solutes. pounds. This is, however, not surprising due to an easy polar-
The trend is reversed mainly for the C sC contribution of ization of the double bond. The temperature evolution of
alkenes, which mimics behavior of polar Žhydrophilic. com- functional groups for derivative properties also follows ex-
w RT
ž
q d Gw yGwig y RT ln
Mw pref / ⌬ HT ½
hyd s e Ž 2Tc y ⌽ .Ž Tc yT . q1r2 Ž T yTc .
cor 2 2
b
q RT w Ž aq c exp w rT x y by ␦ . q
½ Ž exp ww x y1.
T y⌽
q
␦
Ž exp w w x y1. 5 Ž A2.
2
q Ž Tc y ⌽ . ln
Tc y ⌽ 5 Ž A7.
⌬GThyd
cor
s ⌬ HT cor T cor
hyd yT⌬ S hyd Ž A8.
Appropriate derivations of ⌬GThyd lead to other thermody-
namic properties of hydration
2
e Ž T yTc .
⌬ HT T cor ⌬CTp,hyd
cor
s Ž A9.
hyd s ⌬ H hyd q RT Ž T␣ y1. q d Hw y Hwig ŽT y ⌽ .
w
y RT Ž T␣ y1 . x q RT c exp w rT x
T
⭸w where ⌽ s 228 K is a general constant. Correction functions
y RT 2 ž / aq b Ž exp ww x y1. q c exp w rT x are zero by definition at T GTc . All thermodynamic proper-
⭸T p
ties that appear in the above equations should be applied in
q ␦ Ž exp w w x y1 . 4 Ž A3. their basic SI units when using parameters from Table 2 in
the text.
⌬ SThyd s Ž ⌬ HT T
hyd y ⌬G hyd . rT Ž A4.
⭸␣ w ⭸␣ w
⌬CTp,hyd s ⌬CTp ,hyd
cor
q 2 RT␣ w q RT 2 ž / / ½ y R q d C p ,w yC pig,w y 2 RT␣ w q RT 2
ž ⭸T p
ž /
⭸T p
yR
°2 R ⭸w
aq b Ž exp w w x y1 . q c exp w rT x q ␦ Ž exp w w x y1 4 y c exp w rT x
¶
ž /Ž
⭸T p T
w ⭸w 2
Manuscript recei®ed May 21, 2002, and re®ision recei®ed Sept. 17, 2002.