200 Chem. Eng. Technol.

16 (1993) 200-205

Fluid Phase Equilibria in the System Hydrogen Sulphide-Water

at High Pressure.
Application of an Extended Henry’s Law

John J. Carroll and Alan E. Mather”

Henry’s law is often used to correlate the solubility of a gas in water. Such an approach
is frequently employed to calculate the solubility of hydrogen sulphide in water. However,
Wright and Maass [l] clearly demonstrated that the strict Henry’s law did not apply to the
solubility of hydrogen sulphide in water. That is, the solubility of hydrogen sulphide is not
directly proportional to its partial pressure, even at seemingly low pressures. By accounting
for the non-idealities in the vapour phase, Carroll and Mather [2] showed how a simple
modification of Henry’s law could be used for moderate pressures (up to 1 MPa). In this
paper, a further extension of Henry’s law is used to model the vapour-liquid equilibrium
at higher pressures and temperatures. It is also applied to liquid-liquid and vapour-liquid-
liquid equilibria, areas where Henry’s law is rarely used.

1 Introduction 2 Henry’s Law

Hydrogen sulphide is an important, non-hydrocarbon com- Carroll [7] presented several forms of Henry’s law and
ponent of petroleum. The toxicity, offensive odour and cor- some guidelines for their application. Unlike hydrocarbons,
rosiveness of H2S require that it should receive special which are only sparingly soluble in water, hydrogen
handling. Water is almost always present in petroleum sulphide has a significant solubility. Therefore, it is
reservoirs. Not only does water occur naturally, but it is necessary to include the effect of the non-idealities in the
also often injected as part of a secondary recovery scheme. liquid phase in the solubility model. This includes both the
Thus, an understanding of the phase equilibrium in the Poynting correction, the effect of pressure on the reference
system H2S-H20 is important to the understanding of fugacity, and activity coefficients, to account for deviations
phase equilibria in petroleum systems in general. from ideality due to composition. That is, the so-called
ensemble form of Henry’s law should be used. Thus, for the
An understanding of the phase equilibria in the system solute
H2S-H20 is also important in the field of geochemistry.
Hot, aqueous (hydrothermal) reservoirs that contain y2x2H21
exp [O,” (P- P?)/R TI = y2&2P (1)
hydrogen sulphide are important in the formation of some
sulphide and sulphate minerals.
where y2 is the activity coefficient, x2 mole fraction of the
The system hydrogen sulphide-water has been studied ex- solute in the liquid phase, HZlthe Henry’s constant for
tensively. Detailed reviews of the literature have been pub- solute 2 in solvent 1, O
r is the partial molar volume of the
lished and will not be repeated here. The reader is referred solute at infinite dilution, P is the total pressure, Py is the
to the works of Carroll and Mather [2-41 and Carroll [5] vapour pressure of the solvent, R is the gas constant, T is
for more information. However, the work of Selleck et al. the absolute temperature, y2 is the mole fraction of the
[6] was considered to be the benchmark investigation for solute in the vapour and 62is the fugacity coefficient of the
this system at high pressure. Carroll and Mather [3, 41 dem- solute in the vapour.
onstrated some problems with the results reported by
Selleck et al. [6]. In the light of these problems, Carroll and
Mather [4] presented a new interpretation of the fluid phase
equilibria, based on an equation of state. The method
presented in this paper is consistent with the new description
where y1 is the activity coefficient, xI is the mole fraction
of phase equilibria, but it uses a different approach, namely
of the solvent in the liquid, 47is the fugacity coefficient of
Henry’s law. In keeping with the observations of Carroll
the pure solvent at the temperature of interest and its
and Mather [4], whenever the data of Selleck et al. [6] are
vapour pressure, u1 is the molar volume of the pure sol-
referenced, it is the raw data that are used.
vent, y 1 is the mole fraction of the solvent in the vapour
and ilis the fugacity coefficient of the solvent in the
* Dr. J. J. Carroll and Prof. A.E. Mather, Department of Chemical
vapour. Since the model was used at a high temperature,
Engineering, University of Alberta, Edmonton, Alberta, Canada
T6G 2G6. and thus the vapour pressure was quite high, the reference
1)List of symbols at the end of the paper. fugacity for the solvent was the fugacity of the pure,

0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1993 0930-7516/93/0306-0200 $5.00+ .25/0

Chem. Eng. Technol. 16 (1993) 200-205 201

saturated solvent at the temperature of interest and thus the al. [ I 11 at 25 “C. They obtained 0; = 35 cm3/mol, a value
inclusion of @: in Eq. (2). typical of a small gas molecule in water (Moore et al. [12]).
To extrapolate this value to higher temperatures, the
An activity coefficient model that has been applied to the following approximation is often used
solubility of moderately soluble gases is the two-suffix
Margules equation. Unfortunately, this equation did not
provide sufficient accuracy. This was particularly true for
the high pressure range. A logical extension was to use the where Tref is a reference temperature at which 0 ; is
three-suffix Margules equation. Thus, for the solute known. This expression will be used here. It is further
assumed that 0 ; is independent of pressure.
RTlny2=(A-3B)x:+4Bx:-A-B . (3)
The vapour pressure of water, Py, was also calculated us-
For the solvent ing the correlation of Saul and Wagner [9].The values from
the correlation are in excellent agreement with “The Steam
R T l n yI = (A+3B)x;-4Bx; (4) Tables” (Haar et al. [ l o ] ) .

where A and B are empirical parameters (Margules con- Finally, the vapour phase fugacity coefficients were
stants). These equations are slightly different from the calculated using the Peng-Robinson [I31 equation of state
usual equations because of the unsymmetrical convention. (PR). It was necessary to employ an equation of state that
The proposed model is quite complex and will be justified could be used: (1) in the very high pressure range where the
by the predictions made using it. vapour has a liquid-like density and (2) for the H2S-rich
liquid. As will be shown, the PR equation does an excellent
job in both regions. The fugacity of pure water at its vapour
2. I Parameters in the Model pressure was also calculated using the PR equation. Only at
high temperatures, did this value deviate significantly from
Henry’s constants for temperatures lower than or equal to unity.
90 “ C were taken from Carroll and Mather [2].For higher
temperatures, the experimental data of Lee and Mather [8] As inputs to the equation of state, the critical point and
were used to obtain the Henry’s constants. The Henry’s acentric factor for each component are required. The
constants were then correlated as a function of temperature
following were used: water: T, = 647.3 K , P, = 22.09 MPa,
to obtain the following expression: 0=0.334; H2S: Tc=373.4K, P c = 8 . 9 6 M P a , w=O.IOO.
Also, the mixing rules require an interaction parameter. In
lnH2, = 312.831 +0.05303T- 12224.4/T-49.78251n T ( 5 ) this study, dij = 0.12 was used. This value was obtained by
improving the prediction of the H2S-rich phase, particular-
where T is in K and 273 < T<453 K (0< t < 180°C) and H21 ly the vapour, compositions.
is in MPa, for concentrations in mole fraction.
Thus, the model used to correlate the solubilities makes very
The Margules constants were obtained by considering most few assumptions. The key assumptions are: ( 1 ) the density
of the vapour-liquid equilibrium data. The following cor- of the solvent is independent of pressure, (2) the partial
relation was obtained: molar volume of the solute at infinite dilution varies direct-
ly with the molar volume of the solvent and ( 3 ) the activity
A / R T = 2.5 , (6) coefficients are accurately modelled by the three-suffix
Margules equation.
B / R T = 11.751 -0.012T-2557/T . (7)

In an attempt to keep the model as simple as possible, A was 2.2 Calculations Involving an H2S-Rich Liquid
held constant. Then, B was fitted as a function of tempera-
ture, largely based on the data of Lee and Mather [ 8 ] . For The correlation outlined here was also applied to vapour-
some combinations of A and B , a converged solution was liquid-liquid and liquid-liquid equilibria (the second liquid
not obtained in the high pressure-high temperature region. being a n H2S-rich liquid). In order to do so, the fugacity of
Thus, the high temperature solubility predictions were the components in the second liquid must be obtained. The
sacrificed in order to obtain reasonable results at high PR equation was used to calculate the fugacity of the com-
pressure. ponents in the second liquid. Thus, in the liquid-liquid
region, the equilibrium equations become
The molar volume of water, required for the Poynting cor-
rection for the solvent, was calculated using the correlation y2x2H2,exp [U;)o ( P - P:)/R TI = w2&P (9)
of Saul and Wagner [9],which is in good agreement with
“The Steam Tables” (Haar et al. [ l o ] ) .The partial molar where w2 is the mole fraction of H2S in the H2S-rich liquid
volume of H2S in water at infinite dilution, for the Poyn- and q?iis the fugacity coefficient of H2S in the H2S-rich
ting correction for the solute, was measured by Barber0 et liquid.
202 Chem. Eng. Technol. 16 (1993) 200-205

For the solvent From Fig. 1 it is clear that model 1, the method just
presented, is the best fit of the experimental data. Neglect-
ing the activity coefficients, model 2 underpredicts the
solubility. At 5.165 MPa, the point where a second liquid is
where w1 is the mole fraction of water in the H2S-rich li- predicted to form, the difference between models 1 and 2 is
quid and 6k is the fugacity coefficient of water in the H2S- 17.8%. Therefore, extending Henry's law with just the
rich liquid. Poynting correction does not yield a good prediction of the
solubility. The model of Carroll and Mather [2] is an im-
For calculations performed in the three-phase region, the provement over the second model. At 5.165 MPa, the dif-
liquid-liquid and vapour-liquid equations are solved simul- ference between models 1 and 3 is 12.4%. Initially, this is
taneously. a surprising result since the method of Carroll and Mather
[2] neglects both the activity coefficients and the Poynting
factor. However, the reason for the improvement is a for-
3 Results tunate cancellation of errors. The product of the activity
coefficient and the Poynting term is approximately one.
Fig. 1 shows the calculated solubilities of hydrogen sulphide The higher the pressure, the larger the deviation from unity.
in water at 71 "C. Four correlations are presented. The first Hence, neglecting both the Poynting correction and the ac-
of these is the model detailed in the present discussion. In tivity coefficients yields a better prediction than neglecting
Fig. 1 this curve is labelled 1. The second model is similar either term individually. Finally, the strict Henry's law
to that discussed above except that, in this model, it is overpredicts the solubility. At 5.165 MPa, the difference be-
assumed that the activity coefficients are unity. Thus: tween the strict Henry's law and the complete version
presented in this paper is about 16%. Note that the strict
x2H21
exp [OF(P-Py)/RT] = y2q?2P . (1 1) Henry's law results in a prediction which is slightly more ac-
curate than that including only the Poynting correction and
This model is labelled 2 in Fig. 1. The third model is that vapour-phase non-idealities (model 2).
used by Carroll and Mather [2] where both the activity coef-
ficients and the Poynting correction are neglected. This Fig. 2 shows a comparison between the calculated and ex-
model is labelled 3 in Fig. 1. Finally, the fourth model is the perimental solubilities of hydrogen sulphide in water for six
strict Henry's law. That is, the solubility is assumed to be isotherms. For the 30"C, 60°C and 90°C isotherms, the
directly proportional to the partial pressure of the solute. calculations end at the three-phase point. The other three
Thus: isotherms extend to 7 MPa, approximately the limit of the
experimental values. The prediction of the solubility is quite
good, except at 180 "C where the predicted solubilities are
slightly lower than the experimental values. The reason for
This model is labelled 4 in Fig. 1. Also shown in Fig. 1 are this was mentioned in the discussion of the model. It was
the experimental data of Lee and Mather [8]. The calcula- these high temperature solubilities that were sacrificed in
tions are terminated at 5.165 MPa, which is the point where order to obtain better results at higher pressures.
a third phase, an H2S-rich liquid, is predicted to form. The
liquid-liquid and vapour-liquid-liquid equilibria will be Lee and Mather [8] measured a few vapour compositions at
discussed in detail later in this paper. 90"C, 120°C and 150°C. These are shown in Fig. 3 along

55 r 2 3 1
50 -

45 -

40 ~

m m
a 35 - a
B
v
z
v
30 -
a,

25 ~

20 - i.
a

0 50
O I
0 5 10 15 80 25 30 35 40 0 5 10 15 20 25 30 35 40

H,S Mole Fraction ( x 1000) H,S Mole Fraction ( ~ 1 0 0 0 )

Fig. 1. Solubility of hydrogen sulphide in water at 71 "C as predicted Fig. 2. Solubility of hydrogen sulphide in water at 30°C, 6 0 ° C , 9O"C,
by four different models (data from Lee and Mather [8]). 120C", 150'C and 180°C (data from Lee and Mather [8]).
Chem. Eng. Technol. 16 (1993) 200-205 203

agreement between the model and the experimental values

is very good.

Fig. 5 shows the predicted vapour compositions for four

temperatures 37.8"C, 71.1 "C, 93.3"C and 104.4"C along
with the raw data of Selleck et al. [6] and the data of
Gillespie and Wilson [14]. These are the vapour composi-
tions in equilibrium with the solubility values shown in
Fig.4. From this figure, it must be concluded that the
model is an excellent prediction of the vapour compositions
in this region.

Fig. 6 is similar to Fig. 4 except that, this time, the high tem-
0 4 8 12 16 20 24 28 perature region is shown. Three isotherms, 137.8 "C,
Mole F r a c t i o n of W a t e r in Vapour 1x100) 148.9"C and 171.1 "C, are shown. Data from both Selleck
Fig. 3. Dew points of hydrogen sulphide-water mixtures at 9 0 ° C
et al. [6] and Gillespie and Wilson [I41 are also shown.
120°C and 150°C (data from Lee and Mather [8]). Fig. 7 shows the predicted vapour compositions for three
temperatures 137.8"C, 148.9"C and 171.1 "C along with
the raw data of Selleck et al. [6] and the data of Gillespie
and Wilson [14]. These are the vapour compositions in

- m 6
9 -

8 -
A

a
2
- 5
GI
i
5

L
IU

aa,
L
a
00 05 1 0 15 20 25 30 35 40 45

Mole F r a c t i o n o f H,S in Liquid 1x100)

Fig.4. Solubility of hydrogen sulphide in water at 37.8"C, 71.1 " c ,

0 2 4 6 8 10 12
93.3"C and 1 0 4 T (data: full symbols from Selleck et al. [6], blank
symbols from Gillespie and Wilson [14]). Mole F r a c t i o n of W a t e r in Vapour (x100)

Fig. 5. Dew Points of hydrogen sulphide-water mixtures at 37.8"C,

71.1 "C, 93.3"C and 104.4'C (data: full symbols from Selleck et al.
with the prediction from the model. Unlike the solubility [6], blank symbols from Gillespie and Wilson [14]).

I
data of Lee and Mather [8], these values were not used to
obtain the parameters of the model. Thus, this represents a
true prediction. From this diagram, it can be seen that, in
general, the model predicts a higher concentration of water
in the vapour than shown by the experiment. However,
;:: 0 13780C
1489°C
171 1°C
17 ! / 8731 1°C

there is a large amount of scatter in the data and Lee and -m 150
Mather [8] reported difficulties in obtaining these values. a
Thus, this is a satisfactory fit. -
r
W
k+
125

5 100
In the remainder of the discussion, the phase behaviour will v1

be considered in four regions: (1) low temperature a

z 75

(30 " C < t< 105 "C) vapour-liquid equilibria, (2) high tem- 50
perature (105 "C < t < 180 "C) vapour-liquid, (3) vapour-liq-
uid-liquid equilibria and (4) liquid-liquid equilibria. 25

O O Y ' ' 1 ' 1 ' 1 '

Fig. 4 shows the solubilities at four temperatures 37.8 "C, 0 1 2 3 4 5 6 7 8

71.1 "C, 93.3 "C and 104.4"C. In this diagram, the correla- Mole F r a c t i o n o f H,S in Liquld 1x100)
tion is compared with the raw data of Selleck et al. (61 and Fig.6. Solubility of hydrogen sulphide in water at 137.8OC, 148.9"C
the data of Gillespie and Wilson [14]. All the isotherms and 171.1 "C (data: full symbols from Selleck et al. [6], blank symbols
shown in this plot end at three-phase points. In general, the from Gillespie and Wilson Lt40
204 Chem. Eng. Technol. 16 (1993) 200-205

, I J
5 10 15 20 25 30 40 50 60 70 80 90 100 110
Mole Fraction o f Water in Vapour 1x1001 Temperature ("C)

Fig. 7. Dew points of hydrogen sulphide-water mixtures at 137.8 "C, Fig. 8. Vapour-liquid-liquid locus for system hydrogen sulphide-water
148.9"C and 171.1 "C (data: full symbols from Selleck et al. [ 6 ] ,blank (data from Carroll and Mather [3]).
symbols from Gillespie and Wilson [14]).

equilibrium with the solubility data in Fig. 6. For pressures

_i
of less than about 20 MPa, the experiment and the predic- 1'
tion are in good agreement. At higher pressures, there is a
-
0

significant difference, particularly at 171.1 "C. -0 4

L",
I
@-/
- 3
3. I Vapour-Liquid-Liquid Equilibrium
C
0
._
c
The model was then used to predict the vapour-liquid-liquid : : 2
L
equilibrium for the system water-hydrogen sulphide. Fig. 8 LL

shows the pressure-temperature locus along which the three -

al
0
phases coexist. Although the prediction is consistently at z 1

lower pressure than the experimental data, the model is an

excellent prediction. 0
30 40 50 60 70 80 90 100 110

Fig. 9 shows the calculated compositions of the phases Temperature ("C)

along the three-phase locus together with some experimen- Fig. 9. Compositions of coexisting phases along the vapour-liquid-
tal values. Considering the slight difference in pressure be- liquid locus for system hydrogen sulphide-water (data from Carroll and
tween the prediction and the experiment and the uncertainty Mather [3]).
in the experimental data, this is concluded to be a good fit.

It is worthy to note that none of the vapour-liquid-liquid

equilibrium data were used to establish the parameters in
the model. Hence, this represents a prediction. 25 1 3 7 8°C 1 71 l'C,/ 93 3.f ,/ 98 9°C
3.2 Liquid-Liquid Equilibrium

At pressures above the three-phase locus, the equilibrium is

between an aqueous liquid and an H,S-rich liquid. Fig. 10
shows the calculated aqueous phase compositions along
with the data of Gillespie and Wilson [14]. Fig. 11 shows the
-
m
a
I
v

'x

2
20
15

10 -
t i.
i@
a,
4
H2S-rich liquid phase compositions along with data from a
both Selleck et al. [6] and Gillespie and Wilson [14]. 5 -
Although there is some scatter in the data and some
systematic deviations between the experimental data and the
predictions, it is concluded that this is a good fit. It must be
OLJ J15 410 .I; 510 5l5
pointed out that the Henry's law approach is usually only
applied to vapour-liquid systems and, as a rule, only at low M o l e Fraction of HZS ( ~ 1 0 0 )
pressure. Not only is this liquid-liquid equilibrium, but the Fig. 10. Liquid-liquid equilibrium in system hydrogen sulphide-water
pressures are quite high as well. - aqueous phase compositions (data from Gillespie and Wilson [14]).
Chem. Eng. Technol. 16 (1993) 200-205 205

total pressure

./
25
37 8°C

I
7 1 1°C
I 04 4°C critical pressure
vapour pressure of solvent (water)
20 .i gas constant, R = 8.314 J/molK

-a
a I
absolute temperature
critical temperature
Celsius temperature
-
r 15
molar volume of solvent (water)
partial molar volume of solute (H2S) at infinite dilution
mole fraction of component i in H2S-rich liquid
mole fraction of component i in aqueous liquid
mole fraction of component i in vapour

:L 1 2 3 4 5 6 7
Greek symbols

yi activity coefficient of component i

Mole Fraction of HO
,
6:: interaction Darameter in Peng-Robinson equation of state
Fig. 11. Liquid-liquid equilibrium in system hydrogen sulphide-water
- H2S-rich liquid phase compositions (data: blank symbols from
ii fugacity coefficient of pure solvent at its vapour pressure
fugacity coefficient of component i in vapour
Selleck et al. [6], full symbols from Gillespie and Wilson [14]). ti
@i fugacity coefficient of component i in H2S-rich liquid
wi acentric factor of component i

4 Conclusions Subscripts

A new model for the solubility of hydrogen sulphide in i component i

water is presented in this paper. The model is based on ref reference state
Henry’s law but includes non-idealities in both the vapour 1 solvent, water
and liquid phases. It is demonstrated that the model is an 2 solute, hydrogen
excellent fit of the experimental data for temperatures up to
180°C and pressures up to about 10MPa. At higher
pressures, the model deviates from the experimental data,
References
but it continues to give realistic results. Thus, the model can
be used to obtain order of magnitude estimates at higher
[l] Wright, R.H., Maass, O., Can. J. Res. 6 (1932) pp. 94- 101.
pressures. 121 Carroll. J.J., Mather, A.E., Geochim. Cosmochim. Acta 53

..
liquid. It is demonstrated that the model yields accurate pp. 468 - 470.
results in the vapour-liquid-liquid and liquid-liquid regions. PI Carroll, J . J., Mather, A.E., Can. J. Chem. Eng 67 (1989)
These are areas where the Henry’s law approach is rarely pp. 999- 1003.
[5] Carroll, J. J., Thesis, Univ. Alberta 1990.
used. [6] Selleck, F. T., Carmichael, L.T., Sage, B. H., Znd. Eng Chem. 44
(1952) pp. 2219-2226.
[7] Carroll, J. J., Chem. Eng Prog. 87 (1991) No. 9, pp. 48 - 52.
Acknowledgement [8] Lee, J. I., Mather, A.E., Ber. Bunsenges. Phys. Chem. 81 (1977)
pp. 1021 - 1023.
This work was supported by the Natural Sciences and [9] Saul, A., Wagner, W., J. Phys. Chem. Ref. Data 16 (1987)
pp. 893 - 901.
Engineering Research Council of Canada (Strategic Grant [lo] Haar, L., Gallagher, J. S., Kell, G. S., NBS/NRC Steam Tables,
100777). Their financial assistance is gratefully acknowl- Hemisphere Publishing Corp., Washington 1984.
edged. [ll] Barbero, J.A., McCurdy, K. G., Tremaine, P. R., Can. J. Chem.
60 (1982) pp. 1872- 1880.
Received: January 22, 1992 [CET 4671 [12] Moore, J.C., Battino, R., Rettich, T. R., Handa, Y. P., Wilhelm,
E., J. Chem. Eng Data 27 (1982) pp. 22-25.
[13] Peng, D.-Y., Robinson, D.B., Ind. Eng Chem. Fundam. 15
(1976) pp. 59-64.
Symbols used [14] Gillespie, P. C., Wilson, G.M., Vapour-Liquid and Liquid-Liq-
uid Equilibria: Water-Methane, Water-Carbon Dioxide, Water-
A Margules constant Hydrogen Sulfide, Water-n-Pentane,Water-Methane-n-pen tune,
B Margules constant Gas Processors Association, Research Report RR-48, Tulsa/OK
H,, Henry’s constant 1982.