Gus. Sep. h-if. Vol. 10, No. 3, pp.

181-186, 1996 Copyright ;C= 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0950-4214/96 $15.00 + 0.00

ELSEVIER

PII:SO950-4214(96)00019-9

A simple model for prediction solubilities in alkanolamines

Mark L. Posey, Kevin G. Tapperson

of acid gas

and Gary T. Rochelle

of Texas at Austin,

Department of Chemical Engineering, Austin, TX 78712-l 062, USA Received 22 August

The University

1995; revised 7 April

1996; accepted 75 April

1996

A simple model has been created for predicting acid gas vapor-liquid equilibrium (VLE) in alkanolamines. The model is simple enough to use in a hand held calculator, but its structure is derived from theory. Model parameters were obtained by regression of experimental VLE data. The model is valid for total gas loadings from 0.003 to 0.8 and over a wide range of temperatures and amine concentrations. Partial pressure predictions are shown to agree with a more complex model over seven orders of magnitude in pressure. Heat of absorption values derived from the model are also shown to agree with literature sources. Parameters are given for the MDEA-H20-H$-CO2 and DEA-H20-H2S systems. Copyright @ 1996 Elsevier Science Ltd Keywords: alkanolamines; equilibrium modeling; acid gas treating

Introduction Natural gas streams and refinery process streams often contain hydrogen sulfide (H2S) and carbon dioxide (C02) gases that must be removed. The traditional treatment method is an absorption/stripping system using aqueous solutions of alkanolamines. Two of the most commonly used amines are methyldiethanolamine (MDEA) and diethanolamine (DEA). Accurate vaporliquid equilibrium (VLE) models for H$ and CO2 in alkanolamine solutions are needed to properly predict operating characteristics and behavior of absorption/ stripping systems. Several experimenters have collected data for the MDEA-CO*, MDEA-HzS and DEA-HZS systems. Fewer data are available for the mixed gas, MDEA-H2S-CO2 and DEA-H2S-CO2 systems. Literature data have been collected over different ranges of temperature, amine concentration and acid gas loading, leaving no possibility for direct comparisons between sources. In order to determine which data sets agree, and which data sets or points are far from the norm, a simple model was constructed. Kent and Eisenberg developed a simple model for predicting VLE that neglects activity coefficients. Their model is based on several equilibrium constants and the Henrys Law relationship. They regress the amine protonation and carbamate reversion equilibrium constants to fit experimental VLE data. Our model is
To whom correspondence should be addressed. Tel: 471 7230; Fax: + 1 (512) 471 7060 + 1 (512)

similar in that it does not have activity coefficients, but our model assumes a single combined reaction instead of the five or six reactions in the Kent-Eisenberg model. This model has an explicit relationship for calculating partial pressure where the Kent-Eisenberg model requires the solution of a set of equations. Also, where this model only needs four values to predict H# partial pressure in the MDEA-H# systems, Kent-Eisenberg uses 22 values. Note, however, that we regressed four parameters for the H2S systems while Kent and Eisenberg only regressed two. Austgen et al2 and Deshmukh and Mather have created more sophisticated models for predicting VLE in these systems. These models explicitly account for activity coefficients and all of the possible ionic species. However, these models are complex and require solution of a set of non-linear equations which is computationally time consuming.

Model

description

The reaction of acid gases with alkanolamines can frequently be approximated by a single equilibrium, Equation (1) or (2) AmineH++ AmineH++ HCO, w Amine + COz(aq.) + Hz0 HS- ej Amine + H$S(aq.) (1) (2)

These two expressions neglect the presence of carbonate (CO:-), sulfide (S2-), and hydroxide (OH-) ions since
181

182

Simple model for VLE of acid gas-alkanolamine

solutions: M.L. Posey et al.

their concentrations will be small. From Equation (l), the equilibrium constant can be written as follows, with concentrations in mole fractions Kko = (CO2 (as.)) (Amine) z (HCO,)(AmineH+)

where: Pcoz = acid gas partial pressure (kPa) Xcoz = acid gas liquid mole fraction
co2

(3)

Amine + Hz0 + H2S + CO2 Kcoz = equilibrium constant for COz(kPa)

=

Rearranging, we obtain Equation (4). Notice that the direction Equations (1) and (2) are written, I& is a desorption equilibrium constant (HCO;)(AmineH+) (Amine)

defined above LT = total loading = T = temperature mole H2S + mole CO2 mole amine (Kelvin)

P2(asJ)

=Go,

X,+mine= gas free amine mole fraction Amine = Amine + Hz0 Derivation of the H2S system is identical and H$S predictions are made by replacing the CO2 subscripts in Equations (10) and (11) with H2S subscripts. Table 1 gives the regressed parameters for CO* and H2S to be used with Equations (10) and (11). Coefficients A and B in Equation (11) represent the standard temperature dependence for reaction equilibrium constants. The total loading terms in Equation (11) approximate an ionic strength dependence suggested by Astarita et ~1.~. The I-Lr term serves to account for the reduced gas absorption capacity of the solution as gas loading increases. This is the term where mixed acid gas systems are taken into account. For example, as H2S loading increases, LT increases which increases the CO2 partial pressure. The mole fraction of the acid gas, Xs,,, accounts for the specific acid gas loading and the gas free amine concentration. Model regression

The (Amine) and (AmineH+) terms can be approximated on a basis of total loading, LT, as follows (Amine) = (1 - LT)XAmine (AmineH+) = (LT)XOAmine (5) (6)

where XAmineis the amine concentration neglecting the presence of the acid gases. Equation (7) can be formed by combining Equations (4), (5) and (6)

(Codaq.)) = K~02(HC0;) FrLr, (1

Henrys Law can be expressed as

tCO2Caq.))

= -

pco*
(8) into Equation (7) we get

Hcoz
Experimental values of the equilibrium constants, K&s and Kco,, were calculated for each experimental point from Equation (12)

Substituting Equation Equation (9)

Pco, = &&ko~ (HCO,) (1 ?r)

We then combine Kko2 and HCQ into a single constant, I&o*. We assume that the bicarbonate mole fraction, (HCO,), accounts for all forms of the absorbed CO2 and replace its symbol with Xoo2 to get the simple, semiempirical model form given in Equation (10)

ln (&o,~~~) = ln

PC02exp(lLT) &O&T

(12)

PC02 JkoJco* =

(1

f;,)

The equilibrium constant, &oz, in Equation (10) is a function of temperature, acid gas loading and amine concentration as given below in Equation (11) In Kco, = A + 4 + CLTXObine + D(LTXAmine)o5 (11)

The natural logarithm of the equilibrium constant was fit by linear regression using Equation (11). The resulting parameters and 95% univariate confidence intervals for the H$S and CO* regressions for MDEA and the H$S regression for DEA are given in Table 1. The parameters are better than implied by the confidence intervals because of the correlation between them. The parameters have already been rounded off to an acceptable accuracy level and additional rounding of these parameters will introduce error into the model predictions.
MDEA-CO2

The four parameters for each acid gas were regressed separately using equilibrium data with the specific acid gas, including mixed acid gas data. The MDEA-CO2

Simple model for VLE of acid gas-alkanolamine

Table 1 Regressed parameters for the acid gas equilibrium constants A MDEA-CO2 MDEA-H2S DEA-HpS 32.45 + 0.43 24.97 kO.64 21.54kO.92 B -744Ok146 -5554_+207 -4452&314

solutions:

M.L. Posey et al.

183

c 33+5 -16.8_+7.0 11.84+11.2

D -18.5_+1.6 -1.8+2.1 -8.25k3.6

parameters were determined by regression of 212 data points. We excluded 23 outlying data points, about 10% of the total data points, from the regression because the ratio of predicted and measured pressure was greater than 2.0 or less than 0.5. The average absolute percent error in CO? partial pressure was 20% for the MDEACOz model. Figure 1 compares calculated and measured CO2 partial pressures in the MDEA system. Only the data sets from Ho and Eguren and Jou et al. contain mixed gas equilibrium data. All of the data were fit within a factor of two over several orders of magnitude in loading and pressure. Figure 2 shows CO1 partial pressure predictions by two models for a 50 wt% MDEA solution. The Austgen model generally agrees with this work over the entire range of loadings and temperatures.
MOEA-H$

contains mixed acid gas data and it shows the greatest scatter at higher loadings. Figure 4 compares H2S partial pressure predictions of this work with Austgen2. The two models generally agree down to a loading of 0.01. Agreement below this point is not expected since the model presented here was regressed on data above a loading of 0.003, and has no theoretical justification for its predictions below this
8

For the MDEA-H2S regression, no data points were omitted, but one possible, unpublished data source was omitted due to strong positive deviations from the norm at H$ loadings less than 0.1. The 191 data points that were regressed were fit with an average absolute percent error in H2S partial pressure of 22%. Figure 3 compares the experimental to calculated H$ partial pressure for the MDEA-H$ system. Only Jou et ~1.~

50~1% MDEA

10-4 t 0.00 1

,I
0.01
0. I

,.I
1

CO, loading

(mole CO, / mole MDEA) Figure 2 Comparison of model predicted CO* partial pressures for 50 wt% MDEA solutions as a function of temperature and Con loading

3.0

0.3

I
0.01

I
0.1

J
1

0.3

I
0.01 Total Acid

I
0.1 Gas Loading, L,

x
1

0.001

0.001

Total acid gas loading, L, (moles H,S+CO,/ moles MDEA)

Figure 1 Comparative fit of experimental vapor-liquid equilibrium data to model calculated values for the MDEA-Cop system. The data include 25-12OC, 35-50 wt% MDEA, and 0.001-0.8 total acid gas loading

(moles H,S+CO, I moles MDEA)

Figure 3 Comparative fit of experimental vapor-liquid equilibrium data to model calculated values for the MDEA-H2S system. The data include 25-12OC, 35-50 wt% MDEA, and 0.001-0.8 total acid gas loading

184

Simple model for VLE of acid gas-alkanolamine

solutions: M.L. Posey et al.

point. Predictions by Austgen2 below a loading of 0.01 are higher than predicted by this model. Calculations were not made for the Austgen2 model at 80C. The resulting simple equilibrium model is valid for MDEA concentration between 35 and 50 wt%, temperature from 25C to 120C and total acid gas loading between 0.003 and 0.8. This model does not account for the solubility of unreacted H2S or COz, therefore the model should not be used above an acid gas loading of 0.8.
DEA-H$
5Gwl% MDEA

111111~
0.001 0.01
H,S

1
I

The regression of the DEA-H$ model included 349 data points, 48 of which were mixed acid gas points from Lawson and Garst**. For this case no data points were omitted. The average absolute percent error in H.$ partial pressure was 60%. This increase compared with MDEA-H2S is due to the wider range of amine concentrations and temperatures in the DEA-H2S regression. Figure 5 compares calculated and measured H2S partial pressure. The model cuts through the middle of the scattered data and all of the data were fit within a factor of three. Figure 6 compares H$ partial pressure predictions by this work with predictions from Austgen2. The models generally agree at 40C and 80C above a loading of 0.01. At 12OC, this model predicts consistently lower pressures than Austgen2 over the entire loading range. Again, below 0.01, predictions by this model are lower than those by Austgen2. The DEA parameters are valid for DEA concentrations between 5 and 55 wt%, temperature from 25C to 150C and total acid gas loading between 0.005 and 0.8. We have not included a set of DEA-CO2 model parameters because of the expected importance of the DEA carbamate species, which violates our single equilibrium assumption. Heat of absorption Parameter B in Equation (11) is directly related to the heat of absorption (AH,,J of CO2 and H$. From ._ thermodynamics, we obtain the relationship between the temperature dependence of the equilibrium constant and the heat of reaction given in Equation (13)
-=

0.1

loading (mole H,S I mole MDEA)

Figure 4 Comparison of model predicted HIS partial pressures for 50 wt% MDEA solutions as a function of temperature and HIS loading

0.01

0.1

Total acid gas loading, L, (moles H,S+CO,I moles DEA)

Figure 5 Comparative fit of experimental vapor-liquid equilibrium data to model calculated values for the DEA-Hz system. The data include 25-l 5OC, 5-55 wt% DEA, and 0.003-0.8 total acid gas loading

10000 3 & d v? X 2 : & 2 f loo0 100 10 1 0.1 0.01 0.001

10-4

d ln Kdes d;

-- AH,,,

(13)

We can take the derivative of Equation (11) and equating it to Equation (13) we arrive at Equation (14) AHdes = -BR In order to compare with literature written AH in terms of absorption AHai,s = BR (14) values, we have

0.001

0.01

0.1

H,S loading (mole H,S / mole DEA)

Figure 6 Comparison of model predicted H$3 partial pressures for 30 wt% DEA solutions as a function of temperature and HIS loading

(15)

Simple model for VLE of acid gas-alkanolamine

Table 2 Heat of acid gas absorption

solutions: ML. Posey et al.

185

Heat of absorption Author This model Jou et a1.6 Jou et aLg Merkley eta/.16 Oscarson et al.* Oscarson et a/. Lee et al.13 -38 30 -38 -38 - 29 -28 wt%, Lr= 0.3 wt%, Lr = 0.3 wt%, LT=0.3 wt%, LT=0.3 wt%, 77C wt%, L,=0.27 Conditions MDEA-CO2 -61.9 -57.6 -60.0 - 53.4 MDEA-H2S -46.2 -40.7

(kJ/gmol

gas) DEA-H>S -37.0

-36.5 -43.6 -46.0

The calculated values of heat of absorption are given in Tubfe 2 and correspond to the average experimental conditions of the data. The average MDEA-CO2 data conditions are 38.4 wt% MDEA, 0.31 total acid gas loading and a temperature of 56.8C. For MDEA-H2S the averages are 38.8 wt% MDEA, 0.31 total acid gas loading and a temperature of 51.6C. The average conditions for DEA-H$ are 29.1 wt% DEA, 0.27 total acid gas loading and a temperature of 75.5C. Merkley et ~1.~ and Oscarson et ai.17,18 report experimentally measured heats of absorption. These references provide simple equations relating AH,,, to temperature and amine concentration which were used to calculate AH,,, at our average experimental data conditions. Both of these references showed as much as f 10% scatter in their data. Jou et al.hs9 and Lee et a1.13 predict AH,,, by modeling experimental vapor-liquid equilibrium data. The predictions of Jou et al. and Lee et a1.13had to be interpolated to our average conditions. Table 2 shows that our predictions reasonably match the experimental data and other model predictions given the accuracy of the data and the error introduced in using average experimental conditions.

model agree with those of much more complicated models over many orders of magnitude in pressure. Although originally created as a tool for comparing data sets, the model presented here is also useful for quick desktop calculations, checking the consistency of experimental data and as a good initial guess for more complex equilibrium calculations. The model structure is general and can be applied to other simple alkanolamine systems. Acknowledgements The authors gratefully thank Huntsman Corporation, the Gas Research Institute and Gas Processors Associa#5092-260-2495, for funding their tion, contract research. References
1 2 Kent, R. and Eisenherg, B. Better data for amine treating Hydrocarbon Processing (1976) 55 87-90 Au&en. D.. Rochelle. G. and Chen. C.-C. Model of vanor-liauid
I I _ I

equicbria for aqueous acid gas-alkanolamine systems. 2. Representation of H$ and CO2 solubility in aqueous MDEA and CO2 solubility in mixtures of MDEA with MEA or DEA Znd Eng Chem
3 Res (1991) 30 543-555 Deshmukh, R. and Mather, A. A mathematical model for equilibrium solubility of hydrogen sulfide and carbon dioxide in aqueous alkanolamine solutions Chem Eng Sci (1981) 36 355-362

Conclusions A simple model has been developed for predicting the solubility of hydrogen sulfide and carbon dioxide in aqueous alkanolamine solutions. The model uses four parameters for each acid gas which are regressed from experimental data. This model is valid over wide ranges of amine concentrations, temperatures and acid gas loadings and can be used for mixed acid gas systems. All of the data regressed for the MDEA system, which includes mixed acid gas data, were predicted with an average absolute percent deviation of 20% for CO2 and 22% for H$ over seven orders of magnitude in pressure. The DEA model predicts all of the data with an average absolute percent deviation in H2S pressure of 60%, also over seven orders of magnitude. From theory, we have related model parameters to the heat of absorption of acid gases. The heat of absorption values calculated from the model are consistent with experimental data and predictions by other models. We have also shown that partial pressure predictions by this

4 5

Astarita, G., Savage D.W. and Bisio, A. Gas Treating with Chemical Solvents Wiley, New York (1983) Ho, B. and Eguren, R. (Amoco Production Company) Solubility of acidic gases in aqueous DEA and MDEA solutions 1988 Spring National Meeting qf the American Institute of Chemical Engineers 6-10 March 1988 Jou, F-Y., Mather, A. and Otto, F. Solubility of H$ and CO2 in aqueous methyldiethanolamine solutions Znd Eng Chem Process Des Dev (1982) 21 539-544 Jou, F-Y., Carroll, J., Mather, A. and Otto, F. Solubility of

mixtures of hydrogen sulfide and carbon dioxide in aqueous N-methyldiethanolamine solutions / Chem Eng Data (1993) 38
15-71 Jou, F-Y., Carroll, J., Mather, A. and Otto, F. Solubility of carbon

dioxide and hydrogen sulfide in a 35 wt% aqueous solution of methyldiethanolamine Can J Chem Eng (1993) 71 264-268 Jou, F-Y., Otto, F. and Mather, A. Vapor-liquid equilibrium of carbon dioxide in aqueous mixtures of monoethanolamine and methyldiethanolamine Znd Eng Chem Res (1994) 33 2002-2005 DB Robinson Research Ltd. GRI Project 1121.002 contract #5092-260-2343 Gas Research Institute Annual Report JanuaryDecember 1994 Atwood, K., Arnold, M. and Kindrick, R. Equilibria for the system of ethanolamines-hydrogen sulfide-water Znd Eng Chem (1957) 49(9) 1439-1444

Simple model for ViE of acid gas-alkanolamine

12 Lawson, J. and Garst, A. Gas sweetening data: equilibrium solubility of hydrogen sulfide and carbon dioxide in aqueous monoethanolamine and aqueous diethanolamine solutions J Chem Eng Data (1976) 21(l) 20-30 13 Lee, J., Otto, F. and Mather, A. Solubility of hydrogen sulfide in aqueous diethanolamine solutions at high pressures J C/rem and Eng Data (1973) 18(l) 71-73 14 Lal, D., Otto, F. and Mather, A. The solubility of H2S and CO1 in a diethanolamine solution at low partial pressure Can J Chem Eng (1985) 63 681-685 15 Prausnitz, J., Lichtenthaler, R. and Azevedo, E. Molecular

solutions: M.L. Posey

et al.

Thermodynamics of Fluid-Phase Equilibria 2nd Edn, PrenticeHall, Inc., New Jersey (1986) 16 Merkley, K., Christensen, J. and Jzatt, R. Enthalpies of absorption of carbon dioxide in aqueous methyldiethanolamine solutions Thermochimicu Acta (1987) 121 437-446 17

Oscarson, J., Van Dam, R., Christensen, J. and Izatt, R. Enthalpies of absorption of hydrogen sulfide in aqueous diethanolamine solutions Thermochimicu Actu (1989) 154 89-95 18 Oscarson, J., Van Dam, R. and Izatt, R. Enthalpies of absorption of hydrogen sulfide in aqueous methyldiethanolamine solutions
Thermochimicu Actu (1990) 170 235-241