You are on page 1of 19

Journal of Molecular Liquids 400 (2024) 124441

Contents lists available at ScienceDirect

Journal of Molecular Liquids


journal homepage: www.elsevier.com/locate/molliq

Modeling H2S solubility in aqueous MDEA, MEA and DEA solutions by the
electrolyte SRK-CPA EOS
Niloufar Mehdizade a, Mohammad Bonyadi a, *, Parviz Darvishi a, Mohammad Shamsi a, b
a
Chemical Engineering Department, Faculty of Engineering, Yasouj University, Yasouj, Iran
b
Department of Process Engineering, Faculty of Chemical Engineering, Tarbiat Modares University, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, an electrolyte cubic plus association equation of state (e-CPA EOS) has been proposed to predict the
e-CPA vapor–liquid equilibrium of (H2S) in aqueous solutions of monoethanolamine (MEA), diethanolamine (DEA), and
Alkanolamine methyldiethanolamine (MDEA). This EOS is composed of repulsive forces, short-range interactions, short-range
Electrolyte
ionic interactions, association term, long-range ionic interactions, and Born term. The Soave– Redlich– Kwong
H2S
Genetic algorithm
EOS is used as a non-electrolyte part. A new iterative procedure based on the Jacobi method is introduced to
obtain the compositions of all species in the liquid phase. The genetic algorithm (GA) is used to find the optimal
values of binary interaction coefficients between molecules and ions. Compared with experimental data, the e-
CPA EOS model successfully describes the vapor–liquid equilibrium of H2S in the aqueous solutions of MEA,
DEA, and MDEA. The average absolute deviations for 313 data of MEA-H2S-H2O, 156 data of DEA-H2S-H2O, and
145 data of MDEA-H2S-H2O systems were obtained as 8.20 %, 8.04 %, and 9.41 %, respectively. In addition, for
some similar cases, the capability of the e-CPA EOS was compared with the Clegg–Pitzer, N-Wilson-NRF, e-NRTL,
and e-CTS models. The results indicated that the e-CPA EOS has less error than the four other models.

Numerous authors such as Deshmukh and Mather, Austgen et al., Li and


1. Introduction Mather, and Haghtalab and Shojaeian have used this approach [3–6].
Finally, the last approach is the φ-φ approach, in which both the
H2S and CO2 are acid gases that must be removed from the gas liquid and vapor phases are represented with an EOS. Huttenhuis et al.
stream in refinery operations due to their toxicity and corrosiveness, as [7] showed that to calculate the VLE in high-pressure systems, utilizing
well as to prevent the poisoning of catalysts. Removal of acid gases using e-EOSs in the φ-φ approach can lead to more realistic results than ac­
alkanolamines is a conventional process in the oil and gas industry. The tivity models. Some of the previous studies in this field are as follows:
most commonly used alkanolamines in industrial plants are MEA, DEA, Vallee et al. [8] applied the electrolyte EOS proposed by Fürst and
and MDEA. So far, several thermodynamic models have been presented Renon [9] to predict the solubility of H2S and CO2 in an aqueous DEA
to model the vapor–liquid equilibrium (VLE) of acid gases in different solution. Huttenhuis et al. [7] proposed an electrolyte equation of state
alkanol amine solutions, which can be classified into empirical, ɣ-φ, and to predict the vapor–liquid equilibrium of CO2–H2O–MDEA–CH4 sys­
φ-φ approaches. Most of the empirical models are simple; however, the tems. In this model, the modification of the Redlich-Kwong equation of
accuracy of these models is limited to a specific range of pressure, state was used as the molecular part, and a short-range ionic, a long-
temperature, and acid gas loading [1]. One of the famous models in the range ionic, and a Born term were added to this equation as an ionic
empirical category is the Kent-Eisenberg model [2]. term.
In the ɣ-φ approach, the gas and liquid phases can be modeled with Zoghi et al. [1] presented the mPR-CPA model to calculate the sol­
an EOS and activity coefficient, respectively. In this approach for elec­ ubility of CO2 in an aqueous N-methyldiethanolamine solution. They
trolyte solutions, the effect of electrostatic forces in the liquid phase improved the EOS presented by Huttenhuis et al. [7] by adding an as­
caused by ionic species has been considered by adding short-range and sociation term and using the modified Peng-Robinson EOS as a cubic
long-range ionic interaction terms. Usually, the Debye and Hückel term of EOS.
expression or a modification thereof is used for long-range interaction, Poormohamadian et al. [10] used the Peng-Robinson and Soa­
and an activity model like NRTL is used for short-range interaction. ve–Redlich–Kwong EOS with random and non-random mixing rules to

* Corresponding author.
E-mail address: bonyadi@yu.ac.ir (M. Bonyadi).

https://doi.org/10.1016/j.molliq.2024.124441
Received 2 April 2023; Received in revised form 22 January 2024; Accepted 7 March 2024
Available online 16 March 2024
0167-7322/© 2024 Elsevier B.V. All rights reserved.
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Nomenclature Greek Letters


εAiBj association energy of interaction between sites A and B
Ar molar residual Helmholtz free energy (J/mol) βAiBj association volume of interaction between sites A and B
a(T) attraction parameter of e-CPA EOS (J m3/mol) ξ3 parameter defined in Eq. (22)
a0 parameter of e-CPA EOS η packing fraction
a(i) coefficient in Eq. (38) and (39), i = 1, … 5 γ activity coefficient
AAD% average absolute percent deviation φ fugacity coefficient
b covolume parameter of e-CPA EOS (m3/mol) ΔAiBj strength of interaction between sites A and B
C(i) coefficient in Eq. (5), i = 1, … 4 Γ shielding parameter (m− 1)
C1 parameter of e-CPA EOS δi ionic/molecular diameter (m)
D dielectric constant of solution σ pa anionic Pauling diameter
Ds dielectric constant of solvents ρ molar density (molm− 3)
Dm molecular pure component dielectric constant υ molar volume (m3 mol -1)
d(i) coefficients in Eq. (30) ε0 vacuum electric permittivity (C2 J− 1 m− 1)
e electronic charge (1.60219 × 10-19) (C) αLR long-range parameter in ionic interaction term (m)
g(ρ) radial distribution function λ1,λ2 fitting parameter for e-CPA equation of state
Kx equilibrium constant in mole fraction scale
kij binary interaction parameter between species i and j Subscripts
LH2S H2S loading (moles of acid gas per mole of amine) a anion
Mj number of associating sites per molecule j c critical property
n the number of moles i component i
Np number of data points j component j
NA Avogadro’s number (6.02205 × 1023) (mol− 1) m molecular component
O. F objective function mix mixture
P pressure r residual
R gas constant (8.314 J mol− 1 K− 1) s -saturated
T temperature (K) t -total
v molar volume (m3/mol) Born Born term in Eq. (8)
wij binary interaction parameter between species i and j LR long-range in Eq. (8)
xi mole fraction of component i RF repulsive force in Eq.8
xAi fraction of sites A on molecule i that do not form bonds SR1 nonelectrolyte short-range in Eq. (8)
with other active sites SR2 ionic short-range in Eq. (8)
xBj fraction of sites B on molecule j that do not form bonds
Superscripts
with other active sites
l Liquid
zi charge number of ionic species i
r Residual
v Vapor

model the phase equilibria of H2S and CO2 in an aqueous MDEA solu­ decreases reboiler duty. However, an 80 % increase in solvent flow rate
tions. Shirazi and Lotfollahi [11] used the perturbed chain statistically is required for the same acid gas absorption performance.
associated fluid theory (PC-SAFT) EOS, which combined with a mean Since alkanolamines are associative compounds and the dissolution
spherical approximation (MSA) and Born terms to predict the solubility of H2S in alkanolamine solutions produces ionic species, accordingly in
of H2S in MDEA aqueous solution. Using this model and adjusting the this study, an e-CPA EOS is proposed to predict the solubility of H2S in
binary interaction coefficients between ions and molecules, the average aqueous MEA, DEA, and MDEA solutions in the φ-φ approach. Also, a
absolute error of 30 % was obtained for the partial pressures of H2S. new iterative procedure based on the Jacobi method is introduced to
Plakia and Voutsas [12] modeled the solubility of (H2S + CO2) and obtain the compositions of all species in the liquid phase.
H2S in MEA and MDEA solutions using extended combines Peng-
Robinson EOS with UNIFAC via universal mixing rules (eUMR-PRU). 2. Thermodynamic model
The VLE data are used to determine the binary interaction coefficients.
They showed that the eUMR-PRU model could predict the solubility of 2.1. Chemical equilibrium
acid gases in the MEA and MDEA aqueous solutions with an accuracy
close to the Kent-Isenberg and the NRTL electrolyte models. When hydrogen sulfide reacts with an aqueous alkanolamine solu­
Cleeton et al. [13] applied the e-PC-SAFT EOS to predict the ab­ tion, the following chemical equilibrium reactions occur [15]:
sorption of CO2 and H2S in MDEA aqueous solutions. The results showed Water Dissociation:
that for binary absorption in aqueous MDEA solutions up to 50 % amine
content, this model performs well and predicts partial pressures of CO2 2H2O↔H3O+ + OH– (1)
and H2S with average absolute deviations of 57.18 % and 79.32 %.
Hydrogen bisulfide formation:
Gonzales et al. [14] used an electrolyte NRTL to investigate the ab­
sorption of CO2 and H2S in aqueous mixtures of MDEA and ethylene H2O + H2S↔H3O+ + HS– (2)
glycol. Equilibrium data were used to determine the parameters of the
model. They used the proposed thermodynamic models in Aspen Plus to Sulfide formation:
simulate the adsorption process and compare it with an aqueous solution H2O + HS-↔H3O+ + S2- (3)
containing MDEA. The simulation results showed that ethylene glycol
Protonated alkanolamine dissociation:

2
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

H2O + RR’R“N H+↔H3O+ + RR’R”N (4) Waals mixing rule:


∑∑
where RR’R“N is referred to alkanolamines (MEA, DEA and MDEA). amix = xi xj (ai aj )0.5 (1 − kij ) (11)
The thermodynamic equilibrium constants for reactions (1) to (4) as a i j

function of temperature can be defined as [15]: ∑


( ) bmix = xi bi (12)
∏n
Cj(2) i
Kj = (xi γ i )υi = exp Cj +
(1)
+ Cj(3) ln(T) + Cj(4) T (5)
T
i=1
where kij is a binary interaction coefficient.
Wertheim’s association theory [24] is a useful approach for
In this equation, xi is the mole fraction, γi is the activity coefficient and ʋi
explaining association interactions in association-based equation-of-
is the stoichiometry coefficient of species i in reaction j. For reactions
state (EOS) models, including SAFT and CPA models.
(1)–(4), the values of C (1) through C (4) are shown in Table 1.
In some studies, the nonideality of the solution is explicitly consid­
The activity coefficient of water is calculated according to the
ered due to associations between ions and solvent species. Lin et al. [25]
following equation [1]:
described the ion associations under the association theory framework
φw (T, P, xw ) without additional parameters to improve their eNRTL model’s accu­
γ water = (6)
φ∞
0w (T, P, xw →1)
racy for strongly associating electrolyte solutions (aqueous salt solu­
tions). Water and counterions are associated simultaneously with all
where P is the total pressure of the system and φ is the fugacity coeffi­ types of ion pairs, including double-solvent-separated ions (2SIP),
cient. For the other species, the unsymmetrical activity coefficient is solvent-shared ions (SIP), and contact ions (CIP).
calculated as follows: Also, Yang et al. [26] in the electrolyte polar perturbed chain sta­
tistical associating fluid theory (ePPC-SAFT) model for aqueous alkali
φi (T, P, xi )
γ* i = (7) halide solutions used the Wertheim association term for describing the
φ∞ i (T, P, xi →0)
short-range ion-ion and ion–solvent interactions. As in previous works,
where “i” refers to all of the species except water. In this work, the the number of association sites is set to 7 for the cation, and 6 for the
fugacity coefficient of molecules and ions are calculated by the e-CPA anion [27,28]. While the e-CPA model can accurately describe the as­
EOS. sociation behavior of molecules in a mixture, it may not explicitly
consider the association between ions and the molecules of the ions. The
e-CPA model is primarily designed to account for the association be­
2.2. Electrolyte CPA EOS tween molecules in a mixture, particularly for systems involving polar
and associating compounds.
In this work, an electrolyte CPA EOS is proposed to represent the VLE However, it is worth noting that the e-CPA model can be extended or
of alkanolamines-H2S-Water systems. The proposed e-CPA EOS can be modified to include ion-ion and ion–molecule associations if desired.
expressed as the sum of two contributions: The electrolyte part, which is This would typically involve incorporating additional terms or param­
based on the e-EOS offered by Fürst and Renon [9] plus the Born term eters to account for these specific interactions.
[23], and the non-electrolyte part, which is based on the SRK EOS by In this study based on Wertheim’s association theory, the association
considering the association term. In this model, the molar residual term in Eq. (8) is defined by the following equation [24]:
Helmholtz energy (Ar) can be expressed as a combination of different ( r) ( ( ) )
∑ ∑
contributions, including repulsive forces (RF), short-range interaction A
= xj lnX Aj −
X Aj
+
Mj
(13)
(SR1), association interaction (Asso), short-range ionic interaction RT Asso. j Aj
2 2
(SR2), long-range ionic interaction (LR) and the Born term.
where xj is the mole fraction of molecular species and Mj is the number
Ar Ar Ar Ar Ar Ar Ar
= ( )RF + ( )SR1 + ( )Asso. + ( )SR2 + ( )LR + ( )Born (8) of association sites in molecule j. XAj represents the fraction of sites A on
RT RT RT RT RT RT RT
molecule j that do not form bonds with other active sites, which the
The first three would be considered for molecules, and the next three following equation can obtain:
would be for ionic species. The Helmholtz energy arising from repulsive ∑
XAj = (1 + ρ XBj ΔAiBj )− 1 (14)
forces (RF) and short-range interactions (SR1) is based on the SRK EOS Bj
and can be expressed as follows:
Ar Ar b a(T) b In Eq. (14), ρ is molar density and ΔAB is the association strength defined
( )RF + ( )SR1 = − ln(1 − ) − ln(1 + ) (9) by Eqs. (15) and (16):
RT RT υ RTb υ
εAiBj
In this equation ʋ is the molar volume and a(T) is calculated from the ΔAB = g(ρ)(exp( ) − 1)bijβAiBj (15)
RT
following equation:
bi + bj
a(T) = a0 (1 + c1 (1 − (T/Tc )0.5 ))2 (10) bij = (16)
2
The parameters ‘a’ and ‘b’ are extended to mixtures by applying van der

Table 1
Temperature dependence of equilibrium constants for reactions (1) to (4). Kj = exp (C(1)
j + C(2) (3)
j /T + Cj ln(T) + C(4)
j T) j = 1,…,4.
Equation C(1)
j C(2)
j C(3)
j C(4)
j T (◦ C) Ref.

(1) 132.899 13445.9 –22.4773 0.0 273.15–498.15 [16]


(2) 214.582 − 12995.4 –33.5471 0.0 273.15–423.15 [16]
(3) –32.0 − 3338 0.0 0.0 287.15–343.15 [17,18]
(4) RR’R“N = MDEA − 9.4165 − 4234.98 0.0 0.0 298.15–413.15 [19,20]
(4) RR’R“N = MEA 2.1211 − 8189.38 0.0 − 0.00748 273.15–393.15 [21,22]
(4) RR’R“N = DEA − 6.7936 − 5927.65 0.0 0.0 273.15–393.15 [21,22]

3
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

where εAiBj and βAiBj are association energy and association volume, In this equation λ1 and λ2 are fitting parameters and σpa is anionic Pauling
respectively. Also, g(ρ) is the simplified radial distribution function diameter. The value of λ1 and λ2 are 1.6 × 10-7 and 3.005 × 10-6,
given by Eq. (17) [29]: respectively [35]. Also, the value of σpa for the two anions HS- and S2- are
1 3.6 and 3.68 ◦ A, respectively [34].
g(ρ) = (17) The long-range ionic interaction (LR) term is expressed by the
1 − 1.9η
simplified Mean Spherical Approximation (MSA) model [9].
η = 0.25bρ (18)
Ar − α2LR ∑ xion z2ion Γ Γ3 υ
( )LR = ( )+ (25)
The values of parameters a0, c1, b, ε and β for the molecules are adopted RT 4π ion 1 + Γδion 3πNA
from the literature [30,31] and shown in Table 2.
For mixtures, parameters of εAiBj and βAiBj are determined via a where Γ (Shielding parameter) and αLR can be calculated by the
combining rule proposed by Kontogeorgis [32]: following equations:

εAiBi + εAjBj ∑xion ( zion )2


εAiBj = (19) 4Γ2 = α2LR NA (26)
2 ion
υ 1 + Γδion

βAiBj = (βAiBi .βAjBj )0.5 (20) e2 NA


α2LR = (27)
ε0 DRT
According to previous investigations [30,31,33] which indicate that the
four-site (4C) association scheme is the best scheme for H2O, H2S, MEA, In Eq. (27), D is dielectric constant and defined by the following equa­
DEA and MDEA, the four-site (4C) association scheme is adopted for all tion [7]:
molecules in the present study.
The SR2 term in Eq. (8) is expressed as follows: 1 − ξ˝3
D = 1 + (Ds − 1)( ) (28)
1 + ξ˝3 /2
Ar ∑∑ xi xj wij
( )SR2 = − (21)
RT i j
υ(1 − ξ3 ) Parameter ξ˝3 can be calculated by Eq. (22) but ionic species must be
used in the summation. In Eq. (27), ε0 is the vacuum electric permittivity
where wij is an interaction coefficient between ion-ion and molecule-ion and parameter “e” is the electron charge. In Eq. (28), the solvent
and ξ3 is the packing factor which can be obtained by the following dielectric constant (Ds) is defined by the following equations [7]:
equation: ∑
m xm Dm
πNA ∑xi δ3i Ds = ∑ (29)
ξ3 = (22) m xm
6 i υ
d(1)
Dm = d(0) + + d(2) T + d(3) T 2 + d(4) T 3 (30)
where i is related to all species such as molecules and ions, NA is Avo­ T
gadro’ number and δi is the diameter of molecule or ionic species. In this
study the diameter of MDEAH+, MDEA, H2O and H3O+ are taken from Table 2 shows the values of parameters d(0) through d(4). In Eq. (26)
[1] and the H2S diameter is adopted from [34]. Also, the anion diameter parameter Γ can be determined using Newton-Raphson iteration
can be calculated by the following equation [9]: method. Also, the Born term in Eq. (8) can be calculated by the following
√̅̅̅̅̅̅̅̅̅ equation:
3 6bi
δi = (23) Ar NA e2 ∑xi z2
NA π ( )Born = (1/Ds − 1) i
(31)
RT 4πε0 RT i
δi
- 2−
The covolume of HS and S are calculated by the following equation
[9]: The Born term is an effective parameter in vapor–liquid equilibrium
calculations in the systems containing ions. It is also a strong function of
bi = λ1 (σpa )3 + λ2 (24) the dielectric constant of the solvent and is used as a correction factor for
the standard state of ions [7]. The values of ionic species diameters are

Table 2
The parameters of pure components in e-CPA equation of state.
H2O H2S MDEA MEA DEA

Tc (K)a 647.3 373.2 677.8 638 715


MWa 18.02 34.08 119.16 61.08 105.14
δ(Ȧ) 2.52b 3.49b 4.5b 3.7 4.92
d (0)a − 19.29 2 − 8.17 − 17.5544 − 5.953
d (1)a 2.98 × 104 0.0 8.99 × 103 14,836 9277
d (2)a − 1.97 × 10-2 0.0 0.0 0.0 0.0
d (3)a 1.32 × 10-4 0.0 0.0 0.0 0.0
d (4)a − 3.11 × 10-7 0.0 0.0 0.0 0.0
Type of associationd 4c 4c 4c 4c 4c
a0(pa.m6.mol− 2) d 0.12277 0.396977 2.1659 1.4112 2.0942
Cd1 0.667359 0.53703 1.3371 0.7012 1.5743
bd 1.4515 × 10-5 2.95 × 10-5 0.11145 × 10-3 5.656 × 10-5 9.435 × 10-5
ε ε(Pa.m3/mol) d 16,655 3726.34 16,159 18,177 16,159
βd 0.0692 0.04745 0.0332 0.00535 0.0332

a. Parameters are taken from [3].


b. Parameters are taken from [34].
d. Parameters are taken from [30,31].

4
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

shown in Table 3. ⎡ ⎤
f1 (x1 , x2 , ..., xn )
⎢ f2 (x1 , x2 , ..., xn ) ⎥
2.3. Phase equilibrium ̅̅→ ⎢ ⎥
F(x) = ⎢ ⎥
⎢ f3 (x1 , x2 , ..., xn ) ⎥ (38)
⎣ ⋮ ⎦
Vapor-liquid equilibrium in the φ-φ approach is expressed by the fn (x1 , x2 , ..., xn )
following equation:
̅̅→
ϕvi yi = ϕli xi (32) By using the first two terms of Taylor Series expansion, F(x) can be
written as follows:
where φvi and φli are the vapor and liquid phase fugacity coefficient, and ⎡
∂f1 ∂f1 ∂f1

yi and xi are the vapor and liquid phase mole fraction, respectively. Mole ⎢ ∂x1 dx1 ∂x2 dx2 ⋯∂xn dxn ⎥
⎡ ⎤ ⎢ ⎥
balance is needed to determine the moles of compounds in the liquid → ⎡ → ⎤ ⎢ ⎥
→ ⎢ ∂f ∂ f ∂f ⎥
phase. In the VLE calculation we assume that the vapor phase contains ⎢ f1 ( xi + dx) ⎥ f (x ) ⎢ 2 2 2

⎢ → → ⎥ ⎢ 1 →i ⎥ ⎢ ∂x1 dx1 ∂x2 dx2 ⋯∂xn dxn ⎥
H2S, water and alkanolamine. However, the material balance must ⎢ f2 ( xi + dx) ⎥ ⎢ f2 ( xi ) ⎥ ⎢


→ ⎥ =⎢ → ⎥ ⎢ ⎥
(39)
determine the composition of the liquid phase. The balance equations in ⎢ f3 (→
⎢ x + dx ) ⎥
⎥ ⎣⎢ f3 ( xi ) ⎥ + ⎢ ∂f3 ∂ f3 ∂f3 ⎥
i
⋮ ⎦ ⎢ dx1 dx2 ⋯ dxn ⎥
the liquid phase are as follows: ⎢
⎣ ⋮ ⎥


⎢ ∂x1 ∂x2 ∂xn ⎥

fn (→
→ fn (→xi ) ⎢ ⎥
Water: xi + dx) ⎢ ⋮ ⎥
⎢ ⎥
⎣ ∂fn ∂fn ∂fn ⎦
nH2 O,t = nH2 O + nHS− + nS2− + nOH− (33) dx1 dx2 ⋯ dxn
∂x1 ∂x2 ∂xn
Alkanolamine:
The above expression can be written as the following form:
nRR′R″N,t = nRR′R ″N + nRR′R ″NH+ (34) ⎡ ⎤
⎡ (→i ̅→) ⎤ ⎡ (→) ⎤
∂f 1 ∂f 1 ∂f 1

⎢ ∂x1 ∂x2 ∂xn ⎥ ⎡
f 1 x + dx f 1 xi ⎤
Hydrogen sulfide: ⎢ (→ ̅→) ⎥ ⎢ (→) ⎥ ⎢
⎢ ⎥
⎥ dx1
⎢ f xi + dx ⎥ ⎢ i ⎥ ⎢ ∂f 2 ∂f 2 ∂f 2 ⎥ ⎢ dx ⎥
nH2 S,t = LH2S nRR’R˝N,t = nH2 S + nHS− + nS2− (35) ⎢ 2
⎢ (→
⎥ ⎢ f2 x ⎥ ⎢
⎥ ⎢ ⎥ ⎢ ⋯ ⎥ ⎢ 2⎥

⎢ ̅→) ⎥ ⎢ (→) ⎥ ⎢ ∂x1 ∂x2 ∂xn ⎥ ⎢ dx3 ⎥
⎢ f 3 xi + dx ⎥ ⎢ f 3 xi ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥=⎢ ⎥ + ⎢ ∂f 3 ∂f 3 ∂f 3 ⎥ × ⎢ . ⎥
Charge balance: ⎢ . ⎥ ⎢ . ⎥ ⎢ ⋯ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ∂x1 ∂x2 ∂xn ⎥ ⎢ . ⎥ ⎢

⎢ . ⎥ ⎢ . ⎥ ⎢ ⎥ ⎣
nRR′R ″NH+ + nH3 O+ = nHS− + 2nS2− + nOH− (36) ⎢
⎣ ( .
⎥ ⎢
⎦ ⎣
⎥ ⎢
⎦ ⎢ ⋮⋮⋱⋮


. ⎦
→ ̅→) . ⎢ ⎥ dxn
→ ⎣ ∂f n ∂f n ∂f n ⎦
f n xi + dx f n ( xi )
where LH2S is H2S loading equal to the mole ratio of absorbed H2S per ⋯
∂x1 ∂x2 ∂xn
mole of amine, and the subscript “0” represents the initial moles.
Chemical reactions (1)–(4) coupled with Eqs. (33)–(36), form a set of (40)
nonlinear equations that should be solved to calculate the mole fraction
The Jacobian matrix is a key component for solving a set of nonlinear
of all species and molecules in the liquid phase.
equations and is defined as a matrix of first-order partial derivatives:
Based on the Gibbs energy minimization, Smith and Missen [36]
⎡ ⎤
proposed a procedure to find the equilibrium compositions. In this study ∂f1 ∂f1 ∂f1

a new simple and fast approach has been proposed to solve these ⎢ ∂x1 ∂x2 ∂xn ⎥
⎢ ⎥
nonlinear equations based on the Jacobi method. The details of the ⎢ ⎥
⎢ ∂f2 ∂f2 ∂f2 ⎥
⎢ ⋯ ⎥
proposed model are as follow: ⎢ ∂x1 ∂x2 ∂xn ⎥
⎢ ⎥
Consider the following nonlinear equations: ⎢ ⎥
J = ⎢ ∂f3 ∂f3 ∂f3 ⎥ (41)
⎧ ⎢ ⋯ ⎥
f 1 (x1 , x2 , ⋯, xn ) = 0 ⎢ ∂x1 ∂x2 ∂xn ⎥

⎪ ⎢ ⎥
⎪ f (x , x , ⋯, x ) = 0
⎪ ⎢ ⎥

⎪ 2 1 2 n ⎢ ⋮⋮⋱ ⎥

⎪ f 3 (x1 , x2 , ⋯, xn ) = 0 ⎢ ⎥
⎨ ⎣ ∂fn ∂fn ∂fn ⎦
. (37) ⋯

⎪ ∂x1 ∂x2 ∂xn

⎪ .


⎪ .

⎩ The Eq. (40) can be rewritten more concisely as follows:
f n (x1 , x2 , ⋯, xn ) = 0
̅̅̅̅̅̅
→ ̅→ ̅̅̅→

→ →
̅̅→ F( xi + dx) = F( xi ) + Jdx (42)
Let F(x) be defined as follow:
̅̅̅→ → ̅→
Since xi + 1 = xi + dx so Eq. (40) can be written as the following form:
̅̅̅̅→ ̅̅̅→
̅→ → →
F(xi+1 ) = F( xi ) + Jdx (43)
Table 3
Ionic species diameters.
̅̅̅̅→
̅̅→
It should be noted that F(xi+1 ) = 0, so the following equation can be
Ionic species δ(◦ A)
obtained:
MDEAH +
4.5a
̅̅̅→
MEAH+ 3.7b → →
DEAH+ 4.92b
F( xi ) = − Jdx (44)
H3O+ 2.52a
HS- 3.219c A summary of the steps required for determining “x” are as follow:
S2- 3.27b
OH– 3.52a 1. Guess initial values for x:
a. Parameters are taken from [7].
b. Parameters are calculated in this work.
c. Parameters are taken from [34].

5
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

⎡ ⎤
Then Eqs. (54) to (62) can be written as functions equal to zero:
⎢ x10 ⎥
⎢ 0⎥
⎢x ⎥ F1 = xH2 O,0 − xHS− − xS= − xOH− − xH2 O = 0 (62)
→0 ⎢⎢ ⎥
2⎥
x = ⎢ x30 ⎥ (45) F2 = xRR′R˝N,0 − xRR′R˝NH+ − xRR′R˝N = 0 (63)
⎢ ⎥
⎢⋮⎥
⎢ ⎥
⎣ 0⎦
xn F3 = xH2 S,0 − xHS− − xS= − xH2 S = 0 (64)

F4 = xHS− + 2xS= + xOH− − xH3 O+ − xRR′R˝NH+ = 0 (65)


̅̅̅̅→
̅̅→ [ ]2
2. Calculate F(xi+1 ) F5 = K1 xH2 O γH2 O − [xH3 O+ γH3 O+ γOH− ][xOH− ] = 0 (66)
3. Calculate the Jacobian matrix from Eq. (41).
̅→
4. Calculate dx from Eq. (44). F6 = K2 [xH2 O γH2 O ][xH2 S γH2 S ]− [xH3 O+ γH3 O+ γHS− ][xHS− ] = 0 (67)
̅̅̅→ → ̅→
5. Calculate new “x” values with this equation: xi + 1 = xi + dx .
i+1 i -10 F7 = K3 [xH2 O γH2 O ][xHS− γHS− ] − [xH3 O+ γH3 O+ γS= ][xS= ] = 0 (68)
6. Repeat steps 2 to 5 until convergence (|x -x |≤10 ).
F8 = K4 [xH2 O γH2 O ][xRR′R˝NH+ γRR′R˝NH+ ] − [γH3 O+ xRR′R˝N γRR′R˝N ][xH3 O+ ] = 0
Based on the above descriptions, a brief procedure for obtaining
(69)
mole fraction of molecules and species in the liquid phase can be
described as follows: The next step is to differentiate each equation with respect to the mole
The mole balance equations in the liquid phase are as follows:
fraction of all molecules and species to form Jacobian matrix. The Ja­
xH2 O,0 = xH2 O + xHS− + xS= + xOH− (46) cobian matrix defined as follows:
⎡ ⎤
∂F1 ∂F1 ∂F1
xRR′R ″N ,0 = xRR′R ″N + xRR′R ″NH+ (47) ⎢ ∂x1 ∂x2 ⋯∂x8 ⎥
⎢ ⎥
⎢ ⎥
xH2 S,0 = LH2 S xRR′R˝N,0 = xH2 S + xHS− + xS= (48) ⎢ ∂F2 ∂F2 ∂F2 ⎥
⎢ ⋯ ⎥
⎢ ∂x1 ∂x2 ∂x8 ⎥
⎢ ⎥
⎢ ⎥
xRR′R ″NH + + xH3 O+ = xHS− + 2xS= + xOH− (49) J = ⎢ ∂F3 ∂F3 ∂F3 ⎥ (70)
⎢ ⋯ ⎥
⎢ ∂x1 ∂x2 ∂x8 ⎥
⎢ ⎥
The equilibrium constants of reactions (1) to (4) can be written as ⎢
⎢ ⋮⋮⋱⋮


follows: ⎢
⎣ ∂F8 ∂F8 ∂F8 ⎦


[xH3 O+ γH3 O+ ][xOH− γOH− ] ∂x1 ∂x2 ∂x8
K1 = (50)
[xH2 O γH2 O ]2
Index 1–8 are related to H2O, H2S, RR’R“N, RR’R”NH+, H3O+, HS-, S=
and OH–, respectively.
[xH3 O+ γH3 O+ ][xHS− γHS− ]
K2 = (51) The procedure for calculating the mole fraction of molecules and ions
[xH2 O γH2 O ][xH2 S γ H2 S ]
with the proposed method is shown in Fig. 1.
In modeling the liquid–vapor equilibrium for the H2S-water-alka­
[xH3 O+ γH3 O+ ][xS= γ S= ]
K3 = (52) nolamine system we need to calculate the chemical equilibrium in the
[xH2 O γH2 O ][xHS− γ HS− ]
liquid phase and the phase equilibrium between liquid and vapor pha­
ses. Therefore, we need two inner and outer loops. Chemical equilibrium
[xH3 O+ γ H3 O+ ][xRR′R ″N γ RR′R ″N ]
K4 = (53) calculations (Fig. 1) are placed in the inner loop and vapor–liquid
[xH2 O γH2 O ][xRR′R ″NH+ γRR′R ″NH+ ]

The activity coefficients can be obtained by Eqs. (6) and (7). To calculate
the mole fraction of molecules and ions, first an initial guess for the mole
fraction of all species and molecules is taken into account. Then Eqs.
(46)–(53) per mole of molecules and ions will be rewritten and through
these equations, the amount of moles of each component will be cor­
rected. These equations are defined as follow:
xH2 O = xH2 O,0 − xHS− − xS= − xOH− (54)

xMDEA = xMDEA,0 − xMDEAH+ (55)

xH2 S = xH2 S,0 − xHS− − xS= (56)

xRR′R˝NH+ = xHS− + 2xS= + xOH− − xH3 O+ (57)


[ ]2
xOH− = K1 xH2 O γH2 O /[xH3 O+ γH3 O+ γOH− ] (58)

xHS− = K2 [xH2 O γH2 O ][xH2 S γH2 S ]/[xH3 O+ γH3 O+ γHS− ] (59)

xS= = K3 [xH2 O γH2 O ][xHS− γHS− ]/[xH3 O+ γH3 O+ γS= ] (60)

xH3 O+ = K4 [xH2 O γH2 O ][xRR′R˝NH+ γRR′R˝NH+ ]/[γH3 O+ xRR′R˝N γRR′R˝N ] (61)


Fig. 1. Flowchart for calculating the mole fraction of molecules and ions in the
liquid phase.

6
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

equilibria calculations are placed in the outer loop. The output of the Table 4
internal loop at a certain temperature and loading is the mole fraction of Parameters used in the genetic algorithm.
molecules and ions in the liquid phase, and the overall output of the Population size 20
program is the partial pressure of hydrogen sulfide and the mole fraction
Selection function Roulette wheel
of molecules in the vapor phase. The calculations continue until the Crossover fraction 0.402
partial pressure of hydrogen-sulfide converges and the error is mini­ Elite count 2
mized. The details of the algorithms are provided in the appendix. Maximum generation 100
Function tolerance 1e-4

2.4. Binary interaction coefficients


using the TABLE CURVE software version 2.11, the following function
Binary interaction coefficients (BIC) significantly affect the results of has been found suitable:
phase equilibrium calculations. As described in Section 2.1, in the kij = a1T4 + a2T3 + a3T2 + a4T + a5 (72)
alkanolamine-H2S-H2O systems, there are three molecules (alkanol­
amine, H2S and H2O) and five ionic species (RR′R″NH+, H3O+, HS-, S= (73)
4 3 2
wij = a1T + a2T + a3T + a4T + a5
and OH–). So, there will be 64 binary interaction coefficients in these
For the systems H2O-H2S-MDEA, H2O-H2S-MEA and H2O-H2S-DEA,
types of systems.
the coefficients a1 to a5 are given in Tables 5–7, respectively.
In this study, as described by Chunxi and Fürst [34] the following
assumptions are made to reduce the number of interaction parameters:
3. Results and discussion
1. Self-interaction coefficients are considered equal to zero.
2. The interaction coefficients between S2- and other species are In order to examine the validity of the proposed e-CPA EOS model,
neglected due to the lower solvation energy of anions compared to the predicted results of solubility of H2S in aqueous MDEA, MEA and
cations. Also, it should be noted that the dissociation constant of Eq. DEA solutions have been compared with experimental data.
(3) is three or four orders of magnitude smaller than Eq. (2).
3. The interaction coefficients between H3O+ and OH– with other spe­
cies in liquid phase can be neglected due to low concentrations of 3.1. MDEA solution
H3O+ and OH–.
4. The binary interaction coefficients are symmetrical (kij = kji and wij Figs. 2–4 show predicted results and experimental data of H2S partial
= wji). pressure versus H2S loading for 23.3 wt%, 35 wt% and 48.8 wt%
5. The binary interaction coefficients are temperature dependent. aqueous MDEA solutions at two different temperatures of 313.15 K and
373.15 K. As can be observed in Figs. 2–4, at very low ranges of H2S
These assumptions reduce 64 binary interaction coefficients to 7 loading, the partial pressure of H2S does not significantly change with
binary interactions. In this work, firstly, the binary interaction co­ temperature. However, at 373.15 K, increasing the partial pressure of
efficients of H2S-H2O and H2S-alkanolamine are determined by using H2S with H2S loading is much higher than that of 313.15 K. The average
binary vapor–liquid equilibrium (VLE) data of Selleck et al., Clarke and absolute deviations of partial pressure were 7.03 %, 13.28 %, and 8.63
Glew, Chang et al., Voutsas et al. and Park et al. [37–40] then the other %, respectively. Also, Fig. 5 represents the predicted results of total
binary interactions such as ion-water and ion-alkanolamine are deter­ pressure versus H2S loading at various temperatures for 18.68 wt%
mined by the solubility data of H2S in aqueous alkanolamine solutions. aqueous MDEA solution. The average absolute deviation is 9.8 %. As can
Genetic Algorithm (GA) is used to obtain the binary interaction co­ be seen, reasonable results are obtained with the e-CPA EOS.
efficients by minimizing the following objective function: In Figs. 6 and 7 for an aqueous solution of 23.3 wt% MDEA, the
equilibrium mole fractions of H2S, MDEA, MDEAH+, and HS- in the
AAD =
1 ∑NP |EXPi − Cali |
( ) (71) liquid phase are plotted versus H2S loading at two different tempera­
Np i=1 EXPi tures of 313.15 K and 373.15 K. As described by Chunxi and Fürst [34],
in the liquid phase of the MDEA-H2S-H2O system, the primary ions
AAD is the average absolute deviation of calculated VLE data from the produced are protonated MDEAH (MDEAH+ ) and bisulfite (HS-). As
experimental data. In this work experimental data of Riegger et al. [22], seen for all ranges of H2S loading, the concentration of MDEAH+ and
Atwood et al. [41], Lawson and Garst [42], Jou et al. [43], Jou et al. HS- are the same because the amount of S2- is near zero. Comparison
[44], Lal et al. [45], MacGregor and Mather [46], Li and Mather [5], between Figs. 6 and 7 indicates that at 313.15 K, the concentration of
Kuranov et al. [19], Sidi-Boumedine et al. [47], Barreau et al. [48] are H2S in the liquid phase starts to increase at about H2S loadings greater
used to adjust the binary interaction coefficients between ion-water and than 0.5, while at 373.15 K, the concentration of H2S in the liquid phase
ion-alkanolamine. starts to increase at about H2S loading greater than 0.3. So, at constant
H2S loading and amine concentration, the partial pressure of H2S in­
2.4.1. Genetic algorithm creases with increasing temperature.
The Genetic algorithm (GA) is a popular strategy that is widely used
to optimize non-linear systems with a large number of variables based Table 5
on the process of natural selection. In the genetic algorithm, an initial Coefficient of Eq. (38) for temperature-dependent binary interaction parameters
population is created then the fitness of each individual is evaluated. in the system of H2O-H2S-MDEA. kij = a1T4 + a2T3 + a3T2 + a4T + a5, wij = a1T4
Based on the fitness score, the individuals (parents) are selected and + a2T3 + a3T2 + a4T + a5.
undergo genetic evolution by crossover and mutation. With parental BIP a1 a2 a3 a4 a5
mating a new population is created. These steps are repeated until a
kH2 O− H2 S − 3 × 10-8 4 × 10-5 − 0.0213 5.08 − 452.14
stopping criterion has been achieved [49].Table 4 shows various GA
kH2 O− MDEA − 2 × 10-8 3 × 10-5 − 0.0158 3.7607 333.62
parameters used in this study. The optimal values of interaction kH2 S− MDEA 4 × 10-7 − 6 × 10-4 0.3038 − 72.981 6547.4
parameter are defined by genetic algorithm and generalized in terms of wMDEAH+ − MDEA 10-10 − 2 × 10-7 10-4 − 0.0283 2.601
temperature: wMDEAH+ − H2 O 0 9 × 10-9 − 10-5 0.0041 − 0.5067
The optimum values of the interaction parameter were obtained at wMDEAH+ − H2 S 6 × 10-11 − 9 × 10-8 5 × 10-5 − 0.0111 1.0014
wMDEAH+ − HS− − 10-10 10-7 − 7 × 10-5 0.0167 1.4317
various temperatures by minimizing the objective function, and then by

7
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Table 6
Coefficient of Eq. (38) for temperature-dependent binary interaction parameters
in the system of H2O-H2S-MEA. kij = a1T4 + a2T3 + a3T2 + a4T + a5 , wij = a1T4
+ a2T3 + a3T2 + a4T + a5.
BIP a1 a2 a3 a4 a5
-7 -4
kH2 O− H2 S − 10 2 × 10 − 0.0959 21.89 − 1865.2
kH2 O− MEA − 10-7 10-4 − 0.074 16.568 − 1381.8
kH2 S− MEA − 7 × 10-7 9 × 10-4 − 0.4649 106.24 − 9077.4
wMEAH+ − H2 0 0 10-9 10-6 − 0.0005 0.0536
wMEAH+ − H2 S 0 − 10-8 10-5 − 0.0053 0.6171
wMEAH+ − HS− 0 − 6 × 10-8 7 × 10-5 − 0.0238 2.7949

Table 7
Coefficient of Eq. (38) for temperature-dependent binary interaction parameters
in the system of H2O-H2S-DEA. kij = a1T4 + a2T3 + a3T2 + a4T + a5. wij = a1T4 +
a2T3 + a3T2 + a4T + a5. Fig. 4. Solubility of H2S in 35 wt% MDEA aqueous solution at (◆) 313.15 K
and (▴) 373.15 K. (Symbols are experimental data [44]).
BIP a1 a2 a3 a4 a5

kH2 O− H2 S − 10− 7 0.0002 − 0.0959 21.89 − 1868.2


kH2 O− DEA − 10-7 − 10-4 − 0.074 16.568 − 1381.8
3.2. MEA solution
kH2 S− DEA − 7 × 10-7 9 × 10-4 − 0.4649 106.24 − 9077.4
wDEAH+ − DEA 2 × 10-9 − 3 × 10-6 0.0015 − 0.3361 28.515 Fig. 8 demonstrates the predicted and experimental H2S partial
wDEAH+ − H2 0 0 2 × 10-10 − 2 × 10-7 7 × 10-5 0.0076 pressure against the H2S loading.
wDEAH+ − H2 S − 7 × 10-11 9 × 10-8 5 × 10-5 0.0107 0.913
For a 15.27 % MEA aqueous solution at various temperatures. The
wDEAH+ − HS− 3 × 10-11 − 4 × 10-8 10-5 − 0.0025 0.1408
average absolute pressure was found to be 8.09 %.
Also, Figs. 9 and 10 show the effect of MEA concentration on the
partial pressure of H2S at 298.15 K and 318.15 K, respectively. At similar
concentrations of MEA (0.6, 1.5, and 4 molal), the average absolute
deviations were found to be 5.94 % and 4.02 % for isotherm of 298.15 K
and 318.15 K, respectively.
As can be seen, the error of the e-CPA EOS is reduced with increasing
temperature. Also, the figures illustrate at a specified H2S loading, the
partial pressure of H2S increases with MEA concentration. Figs. 11 and
12 show the calculated composition of liquid phase versus H2S loading
at 298.15 K and 318.15 K, respectively. As can be seen at H2S loading
greater than 0.3 the amount of H2S, MEAH+, and HS- in the liquid phase
increased with increasing temperature. The total average absolute de­
viation of partial pressure was found to be 7.12 %, which represents the
excellent accuracy of the e-CPA model.

Fig. 2. Solubility of H2S in 23.3 wt% MDEA aqueous solution at (◆) 313.15 K 3.3. DEA solution
and (▴) 373.15 K. (Symbols are experimental data [43]).
Fig. 13 shows the comparison of partial pressure of H2S in 25 wt%
aqueous DEA solution at two different temperatures of 366.48 K and
339 K. Also, Figs. 14–17 illustrate the solubility of H2S in 0.5, 2, 3.5 and
5 normal aqueous solution of DEA at various temperatures. The total
average absolute deviation was found to be 5.41 %, which indicates that
the predicted partial pressures are matched well with experimental data.
In Figs. 18 and 19 for an aqueous solution of 21 wt% DEA, the
equilibrium mole fractions of H2S, DEA, DEAH+, and HS− in the liquid
phase are plotted versus H2S loading at 313.15 K and 373.15 K. As can
be seen, the trend of changes in composition with H2S loading is similar
to MDEA and MEA solutions.
Fig. 20 illustrates the deviation analysis of the partial pressure versus
H2S loading for the MEA-H2S- H2O, DEA-H2S-H2O, and MDEA-H2S-H2O
systems. As can be observed the ratio of the calculated to the experi­
mental equilibrium of H2S partial pressure at high H2S loading is closer
to unity than at low H2S loading. However, the predicted results of the e-
CPA are in good agreement with experimental data and the deviations
Fig. 3. Solubility of H2S in 48.8 wt% MDEA aqueous solution at (◆) 313.15 K
are low.
and (▴) 373.15 K. (Symbols are experimental data [43]).
So far, different thermodynamic models have been used to model the
phase equilibrium of H2S-alkanolamine-H2O systems. In this section, the

8
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 5. Solubility of H2S in 18.68 wt% MDEA aqueous solution at 313.15 K, 333.15 K, 373.15 K, 393.15 K, and 413.15 K. (Symbols are experimental data [19]).

in the aqueous solution of MDEA. It can be seen for all of the concen­
trations of MDEA, the e-CPA is superior to the Clegg–Pitzer and N-Wil­
son-NRF model. Table 9 shows the average absolute deviations of the e-
CPA, e-CTS and e-NRTL models for the solubility of H2S in the aqueous
solution of DEA. As can be observed, the accuracy of the e-CPA is more
than that of the other two models. Also, Table 10 compares the average
absolute deviations of the e- CTS, N-Wilson-NRF and e-CPA models for
the solubility of H2S in the aqueous solution of MEA. Results indicate
that the proposed model has less error than the other two.

4. Conclusion

In this work, the electrolyte version of CPA EOS has been proposed to
predict the solubility of H2S in aqueous solutions of MEA, DEA and
MDEA in a wide range of pressure (0.0026–3866.5 kPa), temperature
Fig. 6. Calculated mole fractions of H2S, MDEA, MDEAH+ and HS- versus H2S (298.15–413.15 K) and H2S loading (0.0725–1.56). The electrolyte part
loading, in an aqueous solution of 23.3 wt% MDEA at 313.15 K. of the e-CPA EOS includes the Fürst and Renon model plus the Born
term, and the non-electrolyte part is based on the SRK EOS.
capability of the e-CPA EOS has been compared with the Clegg–Pitzer Also, in this study the four-site (4C) association scheme is used for all
[5], N-Wilson-NRF [6], e-NRTL [48], and e-CTS [50] models for some the molecules. A new iterative procedure based on the Jacobi method
similar cases. Table 8 shows the average absolute deviations of the has been proposed to obtain the composition of all molecules and species
Clegg–Pitzer, N-Wilson-NRF and e-CPA models for the solubility of H2S in the liquid phase. With the help of the genetic algorithm, the binary

Fig. 7. Calculated mole fractions of H2S, MDEA, MDEAH+ and HS- versus H2S loading, in an aqueous solution of 23.3 wt% MDEA at 373.15 K.

9
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 8. Solubility of H2S in 15.27 wt% MEA aqueous solution at different temperatures. (Symbols are experimental data: [22]).

Fig. 9. Comparison of solubility of H2S in MEA solution at various amine concentrations at 298.15 K. (Experimental data:[22]).

Fig. 10. Comparison of solubility of H2S in MEA solution at various amine concentrations at 318.15 K. (Experimental data: [22]).

10
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 11. Calculated mole fractions of H2S, MEA, MEAH+ and HS- versus H2S loading, in an aqueous solution of 15.27 wt% MEA at 313.15 K.

Fig. 12. Calculated mole fractions of H2S, MEA, MEAH+ and HS- versus H2S loading, in an aqueous solution of 15.27 wt% MEA at 393.15 K.

Fig. 13. Solubility of H2S in the aqueous solution of 25 wt% DEA at 339 K and 366.48 K. (Experimental data: [48]).

11
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 14. Solubility of H2S in the 0.5 N aqueous solution of DEA at various temperatures. (Experimental data: [50]).

Fig. 15. Solubility of H2S in the 2 N aqueous solution of DEA at various temperatures. (Experimental data:[51]).

Fig. 16. Solubility of H2S in the 3.5 N aqueous solution of DEA at various temperatures. (Experimental data: [51]).

12
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 17. Solubility of H2S in the 5 N aqueous solution of DEA at various temperatures. (Experimental data: [51]).

Fig. 18. Calculated mole fractions of H2S, DEA, DEAH+ and HS- versus H2S loading in the aqueous solution of 21 wt% DEA at 313.15 K.

Fig. 19. Calculated mole fractions of H2S, DEA, DEAH+ and HS- versus H2S loading in the aqueous solution of 21 wt% DEA at 373.15 K.

13
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Fig. 20. Comparison of the calculated and the experimental data for H2S equilibrium partial pressure over aqueous MEA, DEA and MDEA solution (◆ MEA-H2S-H2O
[22,52], (+DEA-H2S-H2O [42,45,51], (▴ MDEA-H2S-H2O [19,43,44]).

Table 8
Comparison between results of the e-CPA, Clegg–Pitzer and N-Wilson-NRF models for the solubility of H2S in the MDEA aqueous solutions.
MDEA (wt%) Temperature (K) Np Clegg-Pitzer AAD% N-Wilson-NRF AAD% e-CPA AAD% Ref.

11.8 298.15 7 30.20 50.97 10.30 [53]


20.0 310.96, 338.75, 388.75 25 15.23 42.76 9.12 [53]
23.3 313.15 21 43.5 30.71 9.19 [46]
23.3 313.15, 373.15 17 51 22.12 7.03 [43]
35 313.15, 373.15 21 35 15.74 13.28 [44]
48.8 313.15 14 41.6 28.54 10.63 [44]
48.8 298.15, 313.15, 343.15, 373.15, 393.15 40 26.5 27.31 8.12 [43]
Total 145 36.41 26.66 9.41

Table 9
Comparison between results of the e-CPA, e-CTS and e-NRTL models for the solubility of H2S in the DEA aqueous solutions.
MDEA (wt%) Temperature (K) Np e-NRTL AAD% e-CTS AAD% e-CPA AAD% Ref.

25 338.71 17 5 4.68 0.61 [42]


10,25,41.78,50 310.93, 313.17, 324.82, 338.71, 352.59, 366.48, 373.15, 380.37, 394.26, 408.15, 422 156 – 12.8 8.04 [41,42,47]

with the Clegg–Pitzer, N-Wilson-NRF, e-NRTL and e-CTS models. The


Table 10 results indicate that the proposed e-CPA model has the best accuracy
Comparison between results of the e-CTS, N-Wilson-NRF and e-CPA models for among the other models.
the solubility of H2S in the MEA aqueous solutions.
MDEA (wt Temperature (K) Np e-CTS N-Wilson- e-CPA Ref. CRediT authorship contribution statement
%) AAD% NRF AAD AAD
% %
Niloufar Mehdizade: Conceptualization, Formal analysis, Investi­
3.5,8.4,23.4 298.15 27 12.8 – 5.50 [22] gation, Software, Validation, Writing – original draft. Mohammad
15.2 313.15, 333.15, 35 8.58 5.99 5.12 [42]
Bonyadi: Conceptualization, Investigation, Methodology,Project
353.15, 373.15,
393.15 administration, Supervision, Writing - Review & Editing. Parviz Dar­
vishi: Formal analysis, Methodology, Validation, Writing – review &
editing. Mohammad Shamsi: Formal analysis, Methodology, Writing -
interaction coefficients have been generalized in terms of temperature. Review & Editing.
Comparisons with experimental data show that the proposed model
successfully described the solubility of H2S in three wildly used Declaration of competing interest
alkanolamines.
The average absolute deviations for 313 data of MEA-H2S-H2O, 156 The authors declare that they have no known competing financial
data of DEA-H2S-H2O and 145 data of MDEA-H2S-H2O systems were interests or personal relationships that could have appeared to influence
found to be 8.20 %, 8.04 %, and 9.41 %, respectively. Also, the accuracy the work reported in this paper.
and validity of the proposed model have been examined and compared

14
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

Data availability

Data will be made available on request.

Appendix A

A.1. Calculation of the fugacity coefficient of the e-CPA EOS

The fugacity coefficient in terms of Helmholtz energy is expressed as follows:


∫V ( )
∂ nRT ∂Ar (T, V, n)
RT lnφi = − P− dV − RT lnZ = − RT lnZ (A.1)
∂ni V ∂ni

In Eq. (A.1), Z is the compressibility factor and it is obtained from the following relation:
Z = PV/nRT (A.2)

∫V ( )
nRT
Ar (T, V, n) = − P− dV (A.3)
V

( )
∂Ar (T, V, n) nRT
P = − + (A.4)
∂V T,P V

Ar (T, V, n)
F = (A.5)
RT
( )
∂F nRT
P = − RT + (A.6)
∂V T,n V
( )
∂F
lnφi = − lnZ (A.7)
∂ni T,V,nj

The general form of cubic equation of state is as follows:


RT a(T)
P = − (A.8)
ν− b (ν + δ1 b)(ν + δ2 b)

If δ1 = 1 and δ2 = 0, it becomes the SRK EOS, and if δ1 = 1+√2 and δ2 = 1-√2, it becomes the PR EOS, which in the present study the SRK EOS is used.
( )
Ar (T, V, n) D 1 + δ1 B/V
F(T, V, n) = = − n ln(1 − B/V) − ln (A.9)
RT RTB(δ1 − δ2 ) 1 + δ2 B/V

The above equation is used for mixtures in which D and B are amix and bmix respectively and can be obtained by the mixing rule.
The expression of fugacity coefficient in terms of chemical potential [45] is also expressed as follows:
μr,eCPA
i = μr,SRK
i + μr,Asso.
i + μr,LR
i + μr,SR2
i + μr,Born
i (A.10)

By applying the derivative to Eq. (A9) with respect to ni at constant temperature, volume and nj, the residual chemical potential of SRK, Asso., LR, SR2
and Born can be obtained by the following order:
μr,SRK bi amix bi v v − bmix amix 2amixi v + δ1 bmix
i
= − − ln( )− ( )( )(ln ) (A.11)
RT v − bmix bmix RT(v + δ1 bmix )(v + δ2 bmix ) v bmix RT(δ1 − δ2 ) (amix − bi) bmix v + δ2 bmix

μr,Asso. ∑ 1 n∂lng ∑N ∑
i
= ln(XAi )i − xj (1 − (XAj )j ) (A.12)
RT XAi
2 ∂ni j=1 XAj

where,

15
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

n∂lng 0.475bi
= (A.13)
∂ni v − 0.475bmix

μr,SR2 NA πσi 3 W Wi
i
= + (A.14)
RT 6 V2 (1 − ε3 )2 V(1 − ε3 )
∑∑
W= − nk nl wkl (A.15)
k l

∂W ∑
Wi = ( ) = − 2 nk wki (A.16)
∂ni T,V k

( ) ∑ xion z2 Γ α2
μr,LR α2LR ∂D Z2i Γ
i
= ion
− LR (A.17)
RT 4πD ∂ni T,V,n ion 1 + Γδ ion 4 π 1 + Γσi
( ) ( ) 2 ∑ni Z2 ∂DS
μr,Born dFBorn Ne2 1 Z Ne2
i
= = − 1 *i − i
( ) (A.18)
RT ∂ni T,V,n 4πε0 RT Ds σi 4πε0 RTD2s i σ*i ∂ni

In order to calculate the value of the fugacity coefficient, the total volume is required, and to calculate the volume, the e-CPA equation of state must
first be written in terms of pressure.
( )∑ ∑ ∑∑ xk xl wkl ( )
PeCPA 1 amix 1 0.475bmix − α2LR ∂D ∑ xion z2ion Γ Γ3
= − − 1+ xi (1 − XAi ) − 2
− − (A.19)
RT vm − bmix vm (vm + bmix )RT 2vm vm − 0.475bmix i XAi k l vm (1 − ε3 )
2 4πD ∂v ion 1 + Γδion 3πNA

Also, the volume can be determined by trial and error or by the Newton-Raphson method.

A.2. Calculation of shield parameter

In Eq. (26), the Shielding parameter (Γ) can be calculated by the Newton-Raphson method according to the following relations:
F(Γi )
Γi+1 = Γi − (A.20)
F′(Γi )

∑xion ( zion
)2
F(Γ) = α2LR NA − 4Γ2 = 0 (A.21)
ion
υ 1 + Γδion

∑xion ( zion
)
− zion δion
F′(Γ) = − 8Γ + 2α2LR NA ( ) (22.A)
ion
υ 1 + Γδion (1 + Γδion )2

A1. Flowchart for calculating the shield parameter

16
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

A.3. Flow chart for calculating bubble pressure when composition of liquid phase is known

A2. Bubble pressure calculation algorithm with e-CPA equation of state

17
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

A.4. Flowchart for calculating bubble pressure by taking into account both chemical and phase equilibrium simultaneously

A3. Flowchart for calculating bubble pressure by taking into account both chemical and phase equilibrium simultaneously

References [17] W. Giggenbach, Optical spectra of highly alkaline sulfide solutions and the second
dissociation constant of hydrogen sulfide, Inorg. Chem. 10 (1971) 1333–1338.
[18] B. Meyer, K. Ward, K. Koshlap, L. Peter, Second dissociation constant of hydrogen
[1] A.T. Zoghi, F. Feyzi, M.R. Dehghani, Modeling CO2 solubility in aqueous N-
sulfide, Inorg. Chem. 22 (1983) 2345–2346.
methyldiethanolamine solution by electrolyte modified Peng-Robinson plus
[19] G. Kuranov, B. Rumpf, N.A. Smirnova, G. Maurer, Solubility of single gases carbon
association equation of state, Ind. Eng. Chem. Res. 51 (2012) 9875–9885.
dioxide and hydrogen sulfide in aqueous solutions of N-methyldiethanolamine in
[2] R.L. Kent, Better data for amine treating, Hydrocarbon Process 55 (2) (1976) 87.
the temperature range 313–413 K at pressures up to 5 MPa, Ind. Eng. Chem. Res.
[3] D.M. Austgen Jr, A Model of Vapor-Liquid Equilibria for Acid Gas-Alkanolamine-
35 (1996) 1959–1966.
Water Systems, The University of Texas at Austin, 1989.
[20] K. Schwabe, W. Graichen, D. Spiethoff, Physikalisch-chemische Untersuchungen an
[4] R. Deshmukh, A.E. Mather, A mathematical model for equilibrium solubility of
Alkanolaminen, Z. Phys. Chem. 20 (1959) 68–82.
hydrogen sulfide and carbon dioxide in aqueous alkanolamine solutions, Chem.
[21] V.E. Bower, R.A. Robinson, R.G. Bates, Acidic dissociation constant and related
Eng. Sci. 36 (1981) 355–362.
thermodynamic quantities for diethanolammonium ion in water from 0 to 50◦ C,
[5] Y.-G. Li, A.E. Mather, Correlation and prediction of the solubility of CO2 and H2S
J. Res. Natl. Bur. Stand. A 66 (1962) 71–75.
in aqueous solutions of methyldiethanolamine, Ind. Eng. Chem. Res. 36 (1997)
[22] E. Riegger, H. Tartar, E. Lingafelter, Equilibria between hydrogen sulfide and
2760–2765.
aqueous solutions of monoethanolamine at 25, 45 and 60, J. Am. Chem. Soc. 66
[6] A. Haghtalab, A. Shojaeian, Modeling solubility of acid gases in alkanolamines
(1944) 2024–2027.
using the nonelectrolyte Wilson-nonrandom factor model, Fluid Phase Equilib. 289
[23] M. Born Yon, Volumen und Hydratationswärme der Ionen, Zeitschr, Physik, 1920.
(2010) 6–14.
[24] M. Wertheim, Fluids with highly directional attractive forces. III. Multiple
[7] P. Huttenhuis, N. Agrawal, E. Solbraa, G. Versteeg, The solubility of carbon dioxide
attraction sites, J. Stat. Phys. 42 (1986) 459–476.
in aqueous N-methyldiethanolamine solutions, Fluid Phase Equilib. 264 (2008)
[25] Y.J. Lin, C.J. Hsieh, C.C. Chen, Association-based activity coefficient model for
99–112.
electrolyte solutions, AIChE J. 68 (2022) e17422.
[8] G. Vallée, P. Mougin, S. Jullian, W. Fürst, Representation of CO2 and H2S
[26] F. Yang, T.D. Ngo, J.S.R. Pinto, G.M. Kontogeorgis, J.-C. de Hemptinne, Systematic
absorption by aqueous solutions of diethanolamine using an electrolyte equation of
evaluation of parameterization approaches for the ePPC-SAFT model for aqueous
state, Ind. Eng. Chem. Res. 38 (1999) 3473–3480.
alkali halide solutions, Fluid Phase Equilib. 570 (2023) 113778.
[9] W. Fürst, H. Renon, Representation of excess properties of electrolyte solutions
[27] S. Ahmed, N. Ferrando, J.-C. De Hemptinne, J.-P. Simonin, O. Bernard,
using a new equation of state, AIChE J 39 (1993) 335–343.
O. Baudouin, Modeling of mixed-solvent electrolyte systems, Fluid Phase Equilib.
[10] S.J. Poormohammadian, A. Lashanizadegan, M.K. Salooki, Modelling VLE data of
459 (2018) 138–157.
CO2 and H2S in aqueous solutions of N-methyldiethanolamine based on non-
[28] J.S.R. Pinto, N. Ferrando, J.-C. De Hemptinne, A. Galindo, Temperature
random mixing rules, Int. J. Greenhouse Gas Control 42 (2015) 87–97.
dependence and short-range electrolytic interactions within the e-PPC-SAFT
[11] A.R. Shirazi, M.N. Lotfollahi, Modeling H2S solubility in aqueous N-
framework, Fluid Phase Equilib. 560 (2022) 113486.
methyldiethanolamine solution using a new ePC_SAFT-MB equation of state, Fluid
[29] G. Kontogeorgis, E. Voutsas, Yakoumis, D.P. IV Tassios, Ind. Eng. Chem. Res. 35
Phase Equilib. 502 (2019) 112289.
(1996) 4310–4318.
[12] A. Plakia, E. Voutsas, Modeling of H2S, CO2+ H2S, and CH4+ CO2 solubilities in
[30] A.S. Avlund, G.M. Kontogeorgis, M.L. Michelsen, Modeling systems containing
aqueous monoethanolamine and methyldiethanolamine solutions, Ind. Eng. Chem.
alkanolamines with the CPA equation of state, Ind. Eng. Chem. Res. 47 (2008)
Res. 59 (2020) 11317–11328.
7441–7446.
[13] C. Cleeton, O. Kvam, R. Rea, L. Sarkisov, M.G. De Angelis, Competitive H2S–CO2
[31] I. Tsivintzelis, G.M. Kontogeorgis, M.L. Michelsen, E.H. Stenby, Modeling phase
absorption in reactive aqueous methyldiethanolamine solution: prediction with
equilibria for acid gas mixtures using the CPA equation of state. I. Mixtures with
ePC-SAFT, Fluid Phase Equilib. 511 (2020) 112453.
H2S, AIChE J. 56 (2010) 2965–2982.
[14] K. Gonzalez, L. Boyer, D. Almoucachar, B. Poulain, E. Cloarec, C. Magnon, F. de
[32] G.M. Kontogeorgis, G.K. Folas, Thermodynamic Models for Industrial Applications:
Meyer, CO2 and H2S absorption in aqueous MDEA with ethylene glycol: electrolyte
From Classical and Advanced Mixing Rules to Association Theories, John Wiley &
NRTL, rate-based process model and pilot plant experimental validation, Chem.
Sons, 2009.
Eng. J. 451 (2023) 138948.
[33] S.H. Huang, M. Radosz, Equation of state for small, large, polydisperse, and
[15] D.M. Austgen, G.T. Rochelle, C.C. Chen, Model of vapor-liquid equilibria for
associating molecules, Ind. Eng. Chem. Res. 29 (1990) 2284–2294.
aqueous acid gas-alkanolamine systems. 2. Representation of H2S and CO2
[34] L. Chunxi, W. Fürst, Representation of CO2 and H2S solubility in aqueous MDEA
solubility in aqueous MDEA and CO2 solubility in aqueous mixtures of MDEA with
solutions using an electrolyte equation of state, Chem. Eng. Sci. 55 (2000)
MEA or DEA, Ind. Eng. Chem. Res. 30 (1991) 543–555.
2975–2988.
[16] T. Edwards, G. Maurer, J. Newman, J. Prausnitz, Vapor-liquid equilibria in
[35] E. Solbraa, Equilibrium and non-equilibrium thermodynamics of natural gas
multicomponent aqueous solutions of volatile weak electrolytes, AIChE J 24 (1978)
processing, in: Fakultet for ingeniørvitenskap og teknologi, 2002.
966–976.

18
N. Mehdizade et al. Journal of Molecular Liquids 400 (2024) 124441

[36] W.R. Smith, R.W. Missen, Strategies for solving the chemical equilibrium problem [45] D. Lal, F.D. Otto, A.E. Mather, The solubility of H2S and CO2 in a diethanolamine
and an efficient microcomputer-based algorithm, Can. J. Chem. Eng. 66 (1988) solution at low partial pressures, Can. J. Chem. Eng. 63 (1985) 681–685.
591–598. [46] R.J. Macgregor, A.E. Mather, Equilibrium solubility of H2S and CO2 and their
[37] F. Selleck, L. Carmichael, B. Sage, Phase behavior in the hydrogen sulfide-water mixtures in a mixed solvent, Can. J. Chem. Eng. 69 (1991) 1357–1366.
system, Ind. Eng. Chem. 44 (1952) 2219–2226. [47] R. Sidi-Boumedine, S. Horstmann, K. Fischer, E. Provost, W. Fürst, J. Gmehling,
[38] E. Clarke, D. Glew, Aqueous nonelectrolyte solutions. Part VIII. Deuterium and Experimental determination of hydrogen sulfide solubility data in aqueous
hydrogen sulfides solubilities in deuterium oxide and water, Can. J. Chem. 49 alkanolamine solutions, Fluid Phase Equilib. 218 (2004) 149–155.
(1971) 691–698. [48] A. Barreau, E.B. Le Bouhelec, K.H. Tounsi, P. Mougin, F. Lecomte, Absorption of
[39] E. Voutsas, A. Vrachnos, K. Magoulas, Measurement and thermodynamic modeling H2S and CO2 in alkanolamine aqueous solution: experimental data and modelling
of the phase equilibrium of aqueous N-methyldiethanolamine solutions, Fluid with the electrolyte-NRTL model, Oil Gas Sci. Technol.-Rev. De l’IFP 61 (2006)
Phase Equilib. 224 (2004) 193–197. 345–361.
[40] S.-B. Park, H. Lee, Vapor-liquid equilibria for the binary monoethanolamine+ [49] G.C. Onwubolu, B. Babu, New Optimization Techniques in Engineering, Springer,
water and monoethanolamine+ ethanol systems, Korean J. Chem. Eng. 14 (1997) 2013.
146–148. [50] P. Téllez-Arredondo, M. Medeiros, Modeling CO2 and H2S solubilities in aqueous
[41] K. Atwood, M. Arnold, R. Kindrick, Equilibria for the system, ethanolamines- alkanolamine solutions via an extension of the Cubic-Two-State equation of state,
hydrogen sulfide-water, Ind. Eng. Chem. 49 (1957) 1439–1444. Fluid Phase Equilib. 344 (2013) 45–58.
[42] J.D. Lawson, A. Garst, Gas sweetening data: equilibrium solubility of hydrogen [51] J.I. Lee, F.D. Otto, A.E. Mather, Partial pressures of hydrogen sulfide over
sulfide and carbon dioxide in aqueous monoethanolamine and aqueous diethanolamine solutions, J. Chem. Eng. Data 18 (1973) 420.
diethanolamine solutions, J. Chem. Eng. Data 21 (1976) 20–30. [52] J.I. Lee, F.D. Otto, A.E. Mather, Equilibrium in hydrogen sulfide-
[43] F.Y. Jou, A.E. Mather, F.D. Otto, Solubility of hydrogen sulfide and carbon dioxide monoethanolamine-water system, J. Chem. Eng. Data 21 (1976) 207–208.
in aqueous methyldiethanolamine solutions, Ind. Eng. Chem. Process Des. Dev. 21 [53] R.N. Maddox, Gas conditioning and processing. Vol. 4: gas and liquid sweetening,
(1982) 539–544. 1982.
[44] F.Y. Jou, J.J. Carroll, A.E. Mather, F.D. Otto, The solubility of carbon dioxide and
hydrogen sulfide in a 35 wt% aqueous solution of methyldiethanolamine, Can. J.
Chem. Eng. 71 (1993) 264–268.

19

You might also like