Article

pubs.acs.org/jced

Thermodynamic Model of the Urea Synthesis Process

Alexey L. Voskov* and Gennady F. Voronin
Department of Chemistry, Lomonosov Moscow State University, 119991, Moscow, Russia
*
S Supporting Information

ABSTRACT: A thermodynamic model of the ammonia−carbon

dioxide−water−urea system at urea synthesis conditions, that is,
at t = (135 to 230) °C, p = (3.5 to 45) MPa, L = nN/nC = (2 to
5.5) and W = ( nH2O − n (NH2)2CO) = nO/nC − 2 = (−0.75 to 1.2)
was developed. A liquid phase was described by the UNIQUAC
model including urea, ammonium carbamate, and ammonium
bicarbonate as compounds; the gas phase was described by a virial
equation of state. Bubble point pressures and carbon dioxide to
urea conversion data were used for the model parameters
optimization. The unique features of the model are the correct
description of the saddle azeotrope and intensive use of existing thermodynamic data about constituents and binary subsystems
vapor−liquid equilibria data.

■ INTRODUCTION
Urea is widely used as a fertilizer and produced commercially
from ammonia and carbon dioxide by means of Bazarov
reaction:
2NH3 + CO2 = NH4COONH 2
NH4COONH 2 = (NH 2)2 CO + H 2O (1)
where (NH2)2CO is urea and NH4COONH2 is ammonium
carbamate. That makes a thermodynamic description of the
NH3−CO2−H2O−(NH2)2CO system an actual task that can
be useful for industrial process optimization. Many thermody-
namic models of the urea synthesis process are described in the
literature. They can be divided into three groups: based on ideal
solutions and gases, based on real solutions and gases and based
on empirical correlations.
The simplest group of models1−7 uses the ideal solution
model and ideal gas equation for liquid and gas phases,
respectively. Most of them1−6 were proposed before 1980. A
review of some earlier models of this class was made by
Lemkowitz et al.5 These models are usually taking into account
either bubble point pressure or CO2 to urea conversion. They
are simple to calculate but either inaccurate or describe only
one class of data (i.e., conversions or pressures).
The second group of models8−13 uses a nonideal liquid
(eUNIQUAC model in refs 8−11,13 and Wilson model in ref
12) and real gas equations of state (usually an equation of state
of Nakamura14). Their main drawback is bad reproducibility
due to complexity and omitting of some parameters in the
articles.10,11,13 A notable exception is the model of Isla,
Irazoqui, and Genoud8,9 that is based on eUNIQUAC and
Nakamura EOS and contains all required parameters for its
reproduction.
Another approach is using empirical correlations for
pressures and conversions. Most of papers15,16 use polynomial
dependencies but other approaches are also possible: at least in
one case17 an artificial neuron network model is used.
It should be noted that models from refs 8, 9, and 13 use the
empirical correlations of Gorlovskii15 for the conversion and
the equilibrium pressures that do not take into account the
possible coexistence of liquid and gas phases inside an
autoclave.
Another type of models18−21 is describing only the NH3−
CO2−H2O ternary system without urea. One of them18 was
used for construction of a urea synthesis model.10
The aim of this work is construction of a thermodynamic
model of a liquid phase in the urea synthesis process based on
the most accurate experimental data (that take into account the
possible coexistence of gas and liquid phases inside an
autoclave) and known thermodynamic properties of individual
substances, binary subsystems, and an accurate model of the gas
phase.
■ THERMODYNAMIC MODEL OF THE GAS PHASE
We considered a gas phase consisting of three components:
ammonia, carbon dioxide, and water. Existence of urea and
ionic species such as ammonium carbamate in the gas phase has
been neglected. Although urea is not an ionic compound,
experimental data22−25 about vapor pressure of solid urea and
quantum mechanical modeling26 confirm that its vapor pressure
can be neglected. Experimental data about urea vapor pressure
from refs 23−25 are in agreement with each other within the
uncertainty of the experiment (around 5%) and differ in ref 22
at 5−20%. According to refs 24 and 25 data vapor pressure over
Special Issue: Proceedings of PPEPPD 2016
Received: June 30, 2016
Accepted: October 12, 2016

© XXXX American Chemical Society A DOI: 10.1021/acs.jced.6b00557

J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

solid urea at T = 405 K (near the melting point) is 32 Pa. If All algebraic expressions and coefficients values for the eq 2
these data are extrapolated for liquid urea at T = 453 K using can be found in Supporting Information in the form of
ΔmH and Tm for urea from refs 27 and 28 we obtain ps = 435 MATLAB computer program that implements this virial EOS.

■
Pa, which is not more than 5 × 10−5 of the equilibrium pressure
over the liquid phase at urea synthesis conditions. THERMODYNAMIC MODEL OF THE LIQUID PHASE
The virial equation of state for the NH3−CO2−H2O ternary
gas mixture by Voronin et al.29 has been used for calculating Constituents Properties. A thermodynamic model of
fugacities in gas phases. It is valid for T = (423 to 573) K and p liquid used in this work includes six constituents: ammonia
= (0.1 to 28) MPa and is designed for urea synthesis conditions NH3, carbon dioxide CO2, water H2O, urea (NH2)2CO,
(but not for description of critical points and phase equilibria ammonium carbamate NH2COONH4, and ammonium bicar-
near them). Although most urea synthesis process models8,9,11 bonate NH4HCO3. To evaluate components stability parame-
using nonideal gas use the equation of Nakamura14 for ters and equilibrium constants of reactions in the liquid phase
fugacities, this EOS is much less accurate than that suggested analytic expressions for Gibbs energies of individual com-
by Voronin et al. According to ref 29 the relative error of molar pounds are required.
volume description for the NH3−CO2−H2O gas mixtures does For ammonia, carbon dioxide, and water Gibbs energy of
not exceed 10 and 30% for Voronin et al.29 and Nakamura14 formation for t = (135 to 230) °C (i.e., urea synthesis
respectively. For binary systems maximal relative errors are 5.5 conditions) can be accessed directly from experimental data.
and 87%, respectively, and for pure components they are 0.3 In the case of water thermodynamic properties of the liquid
and 72%, respectively. phase at T = (298.15 to 600) K and p = 0.1 MPa (and were
The virial equation of state by Voronin et al.29 has the layout treated as pressure independent) were taken from NIST-
JANAF thermochemical tables.30,31 Vapor−liquid equilibrium
pV A(2) A(3) A(k) data (VLE) such as saturated vapor pressure ps,H2O and molar
Z= =1+ + 2 + ... + k − 1
RT V V V (2) volumes of liquid water Vm,H2O were taken from IAPWS32 for T
where Z is compressibility coefficient, V is molar volume, A(k) = (300 to 500) K. These data are required for estimation of
(where k = 2 to 8) are temperature- and composition- Gibbs energy of vaporization ΔvGNH3 at different pressures.
dependent virial coefficients. According to ref 29 only A(2) and For ammonia and carbon dioxide the situation is more
A(3) correspond to true second and third virial coefficients, complicated because their critical temperatures (405.6 and
respectively; A(k) values for k > 3 obtained by the least squares 304.3 K respectively) are lower than at urea synthesis
method are usually far from true virial coefficients because of a conditions. Thermodynamic properties of NH3 and CO2 in
linear dependency between them. In the case of pure the state of ideal gas have been taken from NIST-JANAF
components, A(k) coefficients depend only on temperature: thermochemical tables.30,31
In the case of ammonia we decided to extrapolate ps,NH3 and
A(k) = a0(k) + a1(k)T −1 + a 2(k)T −2 + a3(k)T −3 (3)
Vm,NH3 reference data by Golubev et al.33 from T = (298 to 350)
(k)
For two- and three-component systems A coefficients from K up to urea synthesis conditions (i.e., above the critical
eq 2 have the next layout temperature) to estimate ΔvGNH3. For carbon dioxide data
k−1 about the CO2−H2O binary system (see below) will be used
(k)
AAB = AA (T )yAk + AB(T )yBk + (i , j)
∑ AAB (T )yAi yBj for estimation of CO2 Henry constant and corresponding
i,j=1 (4) ΔvGCO2.
(k) (k) (k) (k)
Urea and ammonium carbamate decompose above their
AABC = AAB + AAC + ABC − AA (T )yAk − AB(T )yBk melting points that are below temperatures typical for urea
k−2 synthesis conditions. To estimate their ΔfG for t = (135 to 230)
− A C(T )yCk + ∑ (r , i , j)
AABC (T )yAr yBi yCj °C we used existing ΔfH298.15
° , S298.15
° and ΔmH298.15 values and
r ,i,j=1 (5) Cp(T) temperature dependencies. These data were extrapolated
to higher temperatures using the truncated equation of Berman
A(i,j) (i,j)
AB and AABC from eq 4 and 5 have the same temperature and Brown:34
dependent form as A(k) in eq 3. The a(k)
i parameters from eq 3
for A(k), A(i,j) (i,j)
AB , and AABC were found in ref 29 by approximating Cp(T ) = a + eT −0.5 where e<0 (8)
existing pVT data by eq 2.
Fugacity coefficients of the entire mixture φ and partial For solid urea enthalpy of formation and standard entropy
fugacities φi of NH3, CO2, and H2O φi can be calculated using were taken from ref 35, and temperature and enthalpy of
eqs 6 and 7, respectively. melting and Cp(T) dependence suitable for extrapolation were
1 V0 taken from refs 27 and 28. For solid ammonium carbamate
ln φ = Z − 1 − ln Z −
RT
∫∞ (p − pid ) dV
(6)
Δf H298.15
° °
and S298.15 were taken from ref 36, heat capacities
were taken from ref 37, and enthalpy and temperature of
3 ⎛ ∂φ ⎞ melting were taken from ref 38. In the case of ammonium
ln φi = ln φ − ∑ (xj − δij)⎜⎜ ⎟⎟ bicarbonate its thermodynamic properties above 60 °C are
j=2 ⎝ ∂xj ⎠ (7)
unknown because of its decomposition. And only the
equilibrium constant for its formation from liquid ammonia
id
where p is pressure, p = RT/V is pressure of ideal gas at and carbon dioxide will be evaluated.
volume V and temperature T, δij = 0 for i ≠ j, and δij = 1 for i = To describe heat capacities of all constituents except
j is Kronecker delta. NH4HCO3 the next polynomial function will be used:
B DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

Table 1. Thermodynamic Properties of Ammonia, Carbon Dioxide and Watera

Δf H °298.15 S °298.15 a b·103 c·10−3 d·106

J·mol−1 J·(mol·K)−1 J·(mol·K)−1 J·mol−1K −2 J·mol−1K J·mol−1K −3
NH3(i.g.)b
−45898 192.774 18.0330 54.6963 254.1506 −17.3993
CO2(i.g.) −393522 213.795 25.2353 51.7574 −125.0052 −23.9485
H2O(i.g.) −241826 188.831 29.1146 10.6804 103.4182 1.4302
H2O(liq)b −285830 69.950 166.3716 −390.1360 −1490.8915 472.88
a
a, b, c, d coefficients are given for Cp(T) in J·(mol·K)−1; e = 0. b(i.g.) means ideal gas, (liq) means liquid.

Cp(T ) = a + bT + cT −2 + dT 2 + eT −0.5 (9) 2NH3(liq) + CO2 (liq) = (NH 2)2 CO(liq) + H 2O(liq) (I)
2NH3(liq) + CO2 (liq) = NH 2COONH4(liq) (II)
The first four terms have been used to describe Cp(T) for NH3,
CO2, and H2O. The eT−0.5 term is taken from eq 8 and used NH3(liq) + CO2 (liq) + H 2O(liq) = NH4HCO3(liq) (III)
only for (NH2)2CO and NH4COONH2. (13)
Expressions H(T)−H298.15 and S(T) are
It should be noted that the NH3−CO2−H2O−urea system can
H(T ) − H298.15 = h(T ) − h(T0); be considered as a ternary reciprocal system and it is possible to
1 3 fix amounts of components using only two independent
h(T ) = aT + 0.5bT 2 − cT −1 + dT + 2eT 0.5 variables.
3 (10)
Three types of coordinates were used in the work for liquid
S(T ) − S298.15 = s(T ) − s(T0); phase description:
• Molar concentrations of constituents xi (i = 1, 2, 3, 4, 5, 6
s(T ) = a ln T + bT − 0.5cT −2 + 0.5dT 2 − 2eT −0.5 for NH3, CO2, H2O, (NH2)2CO, NH4COONH2, and
(11) NH4HCO3, respectively)
where T0 = 298.15 K. • L and W ratios for the liquid phase:
Expression for Gibbs energy (i.e., G(T) − H298.15) is
° 3 /nCO
L = n NH ° 2 (14)
G(T ) − H298.15 = (H(T ) − H298.15) − TS(T )
W = n H° 2O/nCO
° 2 (15)
= −[h(T0) − T (S298.15° − s(T0))] + aT (1 − ln T )
where n°i is amount of a component i and ni is amount of
bT 2 cT −1 dT 3
− − − + 4eT 0.5 a constituent i; that is,
2 2 6 (12)
° 3 = n1 + 2n4 + 2n5 + n6
n NH (16)
where H(T) − H298.15 and S(T) are expressions from eq 10 and
eq 11, respectively. ° 2 = n2 + n4 + n5 + n6
nCO
Obtained coefficients for eqs 9, 10, 11, and 12 are given in (17)
Tables 1 and 2. Equation 12 for the constituents will be used n H° 2O = n3 − n4 + n6 (18)
Table 2. Thermodynamic Properties of Urea and For molten urea L = 2 and W = −1, for molten
Ammonium Carbamatea ammonium carbamate L = 2 and W = 0.
• zi are coordinates of components for the NH3−CO2−
Δf H °298.15 S °298.15 a e H2O−(NH2)2CO reciprocal system (i = 1, 2, 3 for
J·mol−1 J·(mol·K)−1 J·(mol·K)−1 J·mol−1·K −0.5
(NH2)2CO(s) b
−332753 104.6 253.64 −2763.3 2NH3, CO2, and H2O, respectively):
(NH2)2CO(liq)b −321827 130.0 287.59 −2763.3 ° 3 /nΣ°
z 2NH3 = 0.5n NH
NH2COONH4(s) −645047 133.5 311.85 −3137.5
NH2COONH4(liq) −624323 180.3 346.37 −3137.5 ° 2 /nΣ°
z CO2 = nCO
a
a and e coefficients are given for Cp(T) in J·(mol·K)−1; b = c = d = 0.
b
(s) means solid, (liq) means liquid. z H2O = n H° 2O/nΣ° (19)
where
° 3 + nCO
nΣ° = 0.5n NH ° 2 + n H° 2O (20)
for the estimation of equilibrium constant at urea synthesis
Note that z3 can be negative but ∑izi = 1 is always valid.
conditions. This approach is more reliable than trying to
The use of zi coordinates to represent the considered
optimize these constants using solely data about pressures and reciprocal system is illustrated in Figure 1 and in the part
conversions. of the article devoted to the saddle azeotrope
UNIQUAC Model. There are six constituents in the liquid composition.
phase: ammonia NH3, carbon dioxide CO2, water H2O, urea Nonelectrolyte UNIQUAC model39 was used for the
modeling of the liquid phase:
(NH 2 ) 2 CO, ammonium carbamate NH 2 COONH 4 , and
ammonium bicarbonate NH4HCO3. Three reactions are Gmex G ex,comb G ex,res
= m + m
occurring in this system: RT RT RT (21)

C DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

due to urea hydrolysis and supercritical state of ammonia

at these temperatures.
3. It is simpler to use the equilibrium constants from eq 13
and 29 calculated for molten compounds.
The UNIQUAC ri and qi parameters used are given in Table
3. They were taken from ref 40 for urea and from ref 19 for the

Table 3. UNIQUAC ri and qi Parameters for the

Constituents of the Liquid Phase
i constituent ri qi
1 NH3 1.4397 2.0918
Figure 1. Coordinates of some points in the zi reciprocal system 2 CO2 5.7410 6.0806
coordinates (see eq 19). ◇, NH2COONH4; □, NH4HCO3; ▽, 3 H2O 0.92 1.40
(NH4)2CO3; +, W = 0, L = (3 to 8) with step 1, ×, W = 0.5, L = (3 to 4 (NH2)2CO 2.1408 2.4860
8) with step 1, ∗, W = 1, L = (3 to 8) with step 1, ○, water solution of 5 NH2COONH4 9.1911 9.7386
urea, x(NH2)2CO = (0.6 to 0.9) with step 0.1. 6 NH4HCO3 12.8910 13.2834
m m
Gmex,comb ϕi z θi
= ∑ xi ln + ∑ qixi ln
RT i=1
xi 2 i=1
ϕi (22)
rest of constituents. Although ref 19 uses the eUNIQUAC
m m
Gmex,res (electrolyte UNIQUAC) model it is possible to use ri and qi for
= −∑ qixi log(∑ θτ
j ji); τji
RT i=1 j=1
nonelectrolytes without changes because eUNIQUAC turns
into UNIQUAC when ions are absent. In the case of
⎛ aji ⎞
= exp⎜ − ⎟ electrolytes, that is, NH4HCO3 and NH2COONH4, we used
⎝ T⎠
ri and qi values for NH+4 , HCO−3 , NH2COO− ions as group
⎛ a(0) + a(1)T ⎞ contributions: r = r+ + r− and q = q+ + q−. Such an
= exp⎜⎜ − ⎟
ji ji
T ⎟ approximation is equivalent to consideration of the electrolytes
⎝ ⎠ (23)
as ion pairs formed by solvated ions. All UNIQUAC a(0) ij and
rx
i i
qx
i i
(1)
aij parameters were optimized in this work from existing
ϕi = m ; θi = m
∑ j = 1 rjxj ∑ j = 1 qjxj (24)
experimental data (see Parameters Optimization Procedure
section for details).
where aji, ri and qi are adjustable model parameters, z = 10 is To find the composition of liquid eq 29 was solved
UNIQUAC coordination number. UNIQUAC binary inter-
numerically by means of Levenberg−Marquardt method
action parameters aij were considered as temperature-depend-
ent ones. implemented in the levmar library:41
UNIQUAC equations for activity coefficients of constituents ⎧ ψ + ψ − 2ψ − ψ + ln γ + ln γ
are given below: ⎪ 4 3 1 2 4 3
= ln KI
⎪ − 2ln γ1 − ln γ2
ln γi = ln γicomb + ln γi res ⎪
(25) ⎪ ψ5 − 2ψ1 − ψ2 + ln γ5 − 2ln γ1 − ln γ2 = ln KII
m ⎪
ϕi z θ ϕ ⎪ ψ − ψ − ψ − ψ + ln γ − ln γ − ln γ = ln KIII
ln γicomb = ln + ln i + li − i ∑ xjlj ⎪ 6 1 2 3 6 1 2
xi 2 ϕi xi ⎪
⎪ − ln γ3
j=1 (26) ⎨ 6
m m ⎪ ψi
θτ
j ji ⎪∑ e =1
ln γi res = −qi ln(∑ θτ
j ji) + qi − qi ∑ m ⎪ i = 1
∑ θτ
k = 1 k kj ⎪ ψ
j=1 j−1 (27) ψ ψ
⎪ e 1 + 2e 4 + 2e 5 + e 6
ψ
=L ∑ e ψi
⎪ i = 2,4,5,6
where ⎪ ψ
z ⎪ e 3 − e ψ4 + e ψ6 =W ∑ e ψi
⎪
lj = (rj − qj) + (rj − 1) ⎩ i = 2,4,5,6 (29)
2 (28)
The nonelectrolyte version of UNIQUAC was applied because where ψi = ln xi is unknowns to be found by Levenberg−
of three reasons Marquardt method, γi = γi(ψ⃗ ) is the activity coefficients (see eq
1. The number of parameters is smaller than in the 25, 26, and 23) and Ki is equilibrium constants (see eq 13 for
electrolyte UNIQUAC version which is an advantage the equations of the reactions). Usage of logarithmic
in the case of the absence of components activities data coordinates is necessary to provide stable convergence during
for the urea synthesis conditions.
2. The lack of data about the dielectric constant of the the numerical solving of eq 29.
NH3−H2O−(NH2)2CO mixed solvent (these three The ln KI and ln KII equilibrium constants in eq 29 (see
nonionic compounds are present in the liquid phase in reactions (I) and (II) in eq 13) were calculated from Gibbs
comparable concentrations) at urea synthesis conditions energies of substances using coefficients in Tables 1 and 2 and
D DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

Table 4. CV,
i
(j)
and CE,
i
(j)
Coefficients for eq 35 and eq 36, Respectively
i (index of constituent)
coefficient 1 2 3
CE,
i
(1)
/(J·mol−1) 38258.1 ± 286 60904.5 ± 1500 56781 ± 53
−CE,i
(2)
/(J·mol−1·K−1) 471.14 ± 5.65 1040.41 ± 29 404.71 ± 0.95
Ci /(J·mol−1·K−1)
E, (3)
56.995 ± 0.825 135.971 ± 4.1 42.66 ± 0.14
CV,
i
(1)
/(cm3·mol−1) −(0.50 ± 0.80) 45.6a 21.89 ± 0.23
102·CV, i
(2)
/(cm3·mol−1 K−1) 9.54 ± 0.25 0 −(3.101 ± 0.12)
105·CV, i
(3)
/(cm3·mol−1 K−2) 0 0 (5.981 ± 0.14)
a
Note: this value is taken from Edwards et al.42 (partial molar volume of carbon dioxide in dilute aqueous solution at 150 °C).
5
,(k)
of the Vm,i and ΔvGi dependencies. In the case of ammonia,
Δr Gi(p , T ) = −RT ln K i = ∑ νij(Δf H °j ,298.15 existing data about ps,NH3 and Vm,NH3 from ref 33 for T = 298−
j=1
350 K were used for the construction of Vm,i and ΔvGi
2
k) k) dependencies.
+ G(j ,298.15(T ) − H (j ,298.15)− ∑ νijΔv Gj(p , T ) In the case of CO2, its critical temperature is much lower
j=1 (30) than that under urea synthesis conditions and it is reasonable to
where ΔrGi(p,T) is the Gibbs energy of the ith reaction (i = I, use Henry constant for carbon dioxide in pure water for the
II, III, see eq 13) νij is the stoichiometric coefficients of jth estimation of
constituent in ith reaction. The thermodynamic properties of * /p )
Δ v G2 = −RT ln(HCO (37)
substances used are given in Tables 1 and 2. 2 0
The ln KIII equilibrium constant in eq 29 (see reaction (III) 42
Edwards et al. estimated the Henry’s constant of carbon
in eq 13) has been estimated during the optimization procedure
dioxide for T = 273−523 K. After conversion from molal to
as
molar fraction scale it becomes
ln KIII = A + BT −1 − (RT )−1(Δ v G1 + Δ v G2) (31)
* /p ) = − 6789.04T −1 − 11.4519 ln T
ln(HCO2 0
where A and B are adjustable parameters because there are no
data about thermodynamic properties of liquid NH4HCO3. 1000
− 0.010454T + 94.4914 + ln
Vapor−Liquid Equilibria. In the case of coexistence of gas M(H 2O)
and liquid phase vapor−liquid equilibrium can be calculated by (38)
means of the next system of equations based on the equality of −1
where M(H2O) is molar mass of water in g·mol . However, eq
i = μi ):
chemical potentials of components (μgas liq
38 has been obtained using14 the equation for the description
Δ v Gi(p , T ) of the gas phase and use of a model for the liquid phase
ln pyi + ln φi = − + ln xi + ln γi different from UNIQUAC. In this work the carbon dioxide
RT (32)
Henry’s constants in pure water were reevaluated together with
where i = (1 to 3), yi is the molar fraction of the ith constituent UNIQUAC binary interaction parameters for the CO2−H2O
in the gas phase, φi is the fugacity coefficient of the ith binary system using corresponding experimental data. Co-
constituent calculated using eqs 2, 6, and 7, γi is the activity efficients CV, (j)
and CE, (j)
obtained in this work are given in
i i
coefficient of the ith constituent calculated using the Table 4.
UNIQUAC model (see eq 25, 26, and 27). Gibbs energies of Although the liquid phase in the urea synthesis process is
vaporization can be expressed as very far from pure water, the use of Henry’s constant for carbon
Δ v Gi(p , T ) = Δ v Gi(p = p0 , T ) + Vm, i(p − p0 ) dioxide in H2O is possible because of the presence of
UNIQUAC interaction parameters for binary systems including
for i = 1, 3 (33) CO2 (e.g., NH3−CO2, CO2−(NH2)2CO, etc.)

Δ v Gi(p , T ) = Δ v Gi(p = ps (T ), T ) + Vm, i(p − ps (T ))

for i = 2
where i = NH3, CO2, H2O, ΔvGi, and Vm,i are Gibbs energy of
vaporization (i.e., transformation of real liquid to ideal gas) and
molar volume of a liquid component respectively, p0 = 101.325
KPa, ps(T) is pressure of saturated water vapor.
The next expressions have been used for the ΔvGi and Vm,i
temperature dependencies:
Vm, i = CiV,(1) + CiV,(2)T + CiV,(3)T 2
Δ v Gi = CiE,(1) + CiE,(2)T + CiE,(3)T ln T
In the case of water, molar volumes of liquid at p = 15 MPa and
saturation pressures from IAPWS32 were used for construction
(34)
■ PARAMETERS OPTIMIZATION PROCEDURE
Two groups of experimental data exist for the NH3−CO2−
H2O−urea system: properties of binary subsystems and
properties of the liquid phase.
Binary Subsystems Data. The group of data is for the
binary subsystems. In the urea synthesis conditions only, in
three binary subsystems constituents are not reacting with each
other: NH3−H2O, CO2−H2O, NH3−(NH2)2CO. In NH3−
CO2 and other systems including urea, ammonium carbonates
and ammonium carbamate constituents are reacting with each
(35) other and no isolated experimental data about them can be
obtained. In the case of the CO2−(NH2)2CO system existing
(36) thermodynamic data43 include only four points of bubble point
pressures at xCO2 = 0.106. The authors also mentioned that
pressures are not stable due to the next reaction:
E DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

2(NH 2)2 CO + CO2 = 2HCNO + NH4COONH 2 (39)

The next reference data will be used for three binary

subsystems mentioned above:
• For the NH3−H2O binary system reference data by
Tillner-Roth and Friend44 for T = (293 to 523) K and p
= (0.012 to 14.7) MPa have been used.
• For the CO2−H2O system vapor−liquid equilibrium data
for T = (323 to 523) K and p = (0.15 to 40) MPa from
refs 45−48 have been used. It should be noted that many
other works with experimental data exist (e.g., see the
review of Hu et al.49) but we used only reliable data that
correspond to urea synthesis conditions.
• For the NH3−(NH2)2CO system bubble point data for T
= (413 to 473) K and p = (1.2 to 99) MPa from ref 43 Figure 2. Compositions of experimental data points. +, Inoue et al.;3,4
have been used. ∗, Kawasumi;51−53 ×, Lemkowitz et al.5,6,43,55,56
Although there are many (about 1000 points) experimental
data for the H2O−(NH2)2CO binary subsystem that are s= ∑ si = ∑ ωi2(fi (β) − f iexp )2
reviewed and used for its thermodynamic modeling by Kosova i (40)
et al.50 all of them are for t < 135 °C. That temperature is lower
where ωi, f exp
i , and f i(β) are statistical weights, experimental
than urea synthesis conditions and we did not use them in the
model parameters optimization. value, and modeling function for ith point, respectively; β is
Quaternary System Data. The second group of model parameters to be optimized. ωi is selected manually and
experimental data is data about composition of the liquid used either for scaling of data series (to compensate usage of
phase and vapor−liquid equilibrium (mainly bubble point different units of measurement) and/or to vary their weight
pressures) for the entire NH3−CO2−H2O−urea system. during parameters optimization.
There are experimental works2−4,15,16,51−53 that investigate Confidence intervals for coefficients and for values calculated
equilibrium conversion of carbon dioxide. And the data from using the model can be obtained using the next relationships:
different sources are not in a good agreement. A possible reason ⎛ ∂f ⎞
for such a disagreement is not taking into account the existence
Δβ = t0.05, n − k ·σ diag[(J T J )−1] ; Jij = ωi⎜⎜ i ⎟⎟ ;
of the gas phase at some values of autoclave loading density.
⎝ ∂βj ⎠
The most popular empirical correlation obtained by Gorlovskii
and Kucheryavyi15 based on experimental data from refs 3, 4, s
σ2 =
15, and 51−53 used in refs 8, 9, and 54 also ignores the n−k (41)
possibility of heterogenity of the system. In this work we will
use experimental points from the works of Inoue3,4 and t0.05, k ·σ
Kawasumi51−53 that take into account heterogenity of the Δfi (β) = Ji (J T J )−1JiT ;
system. ωi
Existing bubble point pressure data.15,16,43,55,56 are also not in ⎛ ∂f (β) ∂f (β) ⎞
Ji = ⎜⎜ i ⋯ i ⎟
∂βk ⎟⎠
good agreement with each other. Possible reasons are the
presence of such inert gases as N2 and not taking into account ⎝ ∂β1 (42)
the heterogenity of the system mentioned above. In this work
we will use data series obtained by Inoue,3,4 Kawasumi51−53 and where n and k are numbers of experimental points and number
Lemkowitz et al.43,55,56 owing to the control of the purity of the of model parameters, respectively, J is Jacobian, t0.05,n−k is the
components and direct control of the liquid phase composition. two-sided t-distribution quantile for α = 0.05 and (n − k)
For the optimization of the model parameters we have degrees of freedoms.
selected experimental points that satisfy the next conditions: t = All coefficients obtained in this work by the least-squares
(135 to 230) °C, p ≤ 45 MPa, L = (2 to 6) (see eq 14), W = method are intentionally given with an excessive amount of
(−0.8 to 1.5) (see eq 15). These conditions correspond to urea digits (in comparison to confidence intervals from eq 41. These
synthesis conditions (when W ≥ 0) or urea solutions in the digits are required to reproduce the model because of a linear
NH3−CO2−H2O system (when W < 0). dependence between the coefficients that is typical for both
Compositions of the experimental points that will be used in linear and nonlinear regression.
the optimization of thermodynamic model are shown in Figure Because of a high number of adjustable model parameters,
2. that is, 60 UNIQUAC a(k) ij parameters (see eq 23), 2 parameters
It should be noted that many data and models exist for (A and B) for ln KIII (see eq 31) and 3 parameters (CE, 2
(j)
) for
lower-temperature conditions t < 135 °C where urea is either CO2 Henry’s constant temperature dependencies (see eq 37),
not forming or not reacting (see Kuranov et al.20 as an their optimization was divided into five steps.
example) but they will not be considered in this work. At the first step 12 UNIQUAC parameters for the NH3−
Optimization Procedure. Parameters optimization proce- H2O, CO2−H2O, and NH3−(NH2)2CO binary subsystems and
dure used in this work is based on nonlinear least-squares three CE,2
(j)
parameters for the Henry’s constant were optimized
method. Generic target function s (sum of squares of residuals) using the existing bubble point data (see above). The next
minimized by this method can be written as objective function was used:
F DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

Table 5. Accuracy of the Optimized UNIQUAC Model Parameters (see Tables 6 and 7) in the Case of Equilibrium Pressures of
the NH3−H2O, CO2−H2O, and NH3−(NH2)2CO Binary Subsystems
system u(p)/MPa max[Δp]/MPa p/MPa T/K points
NH3−H2O 0.12 0.31 0.012−14.7 293−523 110
CO2−H2O 0.62 3.0 0.15−40 323−523 120
NH3−(NH2)2CO 0.11 0.31 1.2−99 413−473 28

s= ∑ (piexp − picalc )2 (ωi(p,bin))2 of optimization was used. To avoid a stop in a local minimum, a
i genetic algorithm was used after each run of lsqnonlin.
Details of its implementations are shown in the Supporting
+ ∑ (yiexp − yicalc )2 (ωi(y))2 Information.
i (43) All program code required to reproduce the optimization
where pexp and pcalc are experimental and calculated values of procedure described in this work is given in the Supporting
i i
bubble point pressures, respectively, yiexp and yicalc are Information. It is written in MATLAB and C++ programming
experimental and calculated values of molar fraction of NH3 languages and requires levmar library41 for its compilation.
or CO2, respectively, and ω(p,bin)
KPa), ωi = 100.
(y)
i = p−1
0 (where p0 = 101.325

It should be noted that in the case of NH3−(NH2)2CO

■ RESULTS AND DISCUSSION
The results of the model parameters optimization procedure
binary subsystem the second term in the eq 43 (objective are values of 65 parameters. Fifteen of them were obtained as a
function) is equal to zero. result of the NH3−H2O, CO2−H2O and NH3−(NH2)2CO
At the second step the rest of UNIQUAC parameters not binary subsystems modeling: 12 UNIQUAC aij(k) binary
including aij with i = 6 or j = 6 (aij for i = 6 or j = 6 have been interaction parameters and 3 parameters (CE, (j)
* 2
) for HCO
2
set to zero) and two ln KIII temperature dependency
parameters have been optimized using the next objective description. The rest of them are 48 UNIQUAC parameters
function: for the NH3−CO2−H2O−(NH2)2CO quaternary system
obtained from equilibrium conversion and equilibrium bubble
s= ∑ (Y iexp − Y icalc)2 (ωi(Y ))2 + ∑ (piexp − picalc )2 (ωi(p))2 point pressure data for the NH3−CO2−H2O−(NH2)2CO
i i system.
Binary Subsystems. Optimized UNIQUAC parameters for
+ ∑ (pilim,exp − pilim,calc )2 (ωi(p ,lim))2 + λ ∑ βi2 the NH3−H2O, CO2−H2O, and NH3−(NH2)2CO binary
i i (44) systems are given in Tables 6 and 7. Temperature and pressure
where Yexp and Ycalc are experimental and calculated values of intervals and model uncertainties are given in Table 5. These
i i
carbon dioxide to urea conversion, respectively, ω(Y) = 10/6, uncertainties are lower than for bubble point pressures in the
ω(p) = 1/(180 p0), ω(p,lim)
i
= 1/(720 p0) (where p0 = 101.325 NH3−CO2−H2O−urea system (see below), and that means
i i
KPa); λ ≥ 0 (used value is λ = 10−8) is Tikhonov regularization that obtained models of binary subsystems are good enough for
(ridge regression) parameter, and βi is generalized notation for our purposes, that is, their accuracy is not lower than in the
normalized thermodynamic model parameters (i.e., a(k) model of the entire quaternary system.
ij , A and
B). The coefficients used for the parameters normalization are Values of Henry constant of carbon dioxide in pure water can
βl = a(0) be obtained using eq 36, eq 37, and the CE, (j)
coefficients given
ij , βl = 1000aij , βl = 10A, βl = 0.1B, βl = C2
(1) E, (j)
. pilim,exp 2
and pi lim,calc
correspond to experimental points from ref 56 for L in Table 4.
= 1.98, W = 0, and p = (19 to 97) MPa. Most of the values are Comparison of eq 38 obtained by Edwards et al.42 and eq 37
beyond the pressure range of the virial equation of state by used with the coefficients from Table 4 (obtained in this work)
Voronin et al.29 but they were included with lowered weights to is shown in Figure 3. Obtained HCO * 2 temperature dependency
ensure absence of physically incorrect pressure maxima at L = has a similar form with the dependency obtained from ref 42.
2.0−2.2 in the optimization results.
The term of eq 44 containing the λ parameter limits the
growth of the model parameter during an optimization
procedure that improves convergence at initial stages of the
procedure.
At the third step a(k) ij parameters with i = 6 or j = 6 (i.e., for
ammonium bicarbonate NH4HCO3) were included into the
optimization, and the results of the second step were used as
initial approximation.
The fourth step is finishing the optimization with λ = 0.
The fifth step is to exclude statistically nonsignificant
parameters, that is, Δβi > βi and go to step four. The
optimization procedure finishes in the case of either excluding
all parameters with Δβi > βi or if further exclusion of statistically
nonsignificant parameters causes at least 30% increase of the
target function (see eq 40). Figure 3. Solid line is H*CO2 obtained in this work (see eq 37 and
For all four steps the lsqnonlin function from MATLAB coefficients in Table 4), dashed line is H*CO2 by Edwards et al.42 (see eq
Optimization Toolbox that uses a trust-region dogleg method 38).

G DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

Equations 38 and 37 are in a good agreement for T < 350 K,

but our values are lower. Different values for T > 350 K can be
connected to different thermodynamic models used for the
liquid phase in the CO2−H2O: UNIQUAC in this work and
Pitzer model used by Edwards et al.42 Also we did not take into
account dissociation of carbon dioxide (it was considered in ref
42).
Model of the Liquid Phase. The optimized parameters for
ln KIII constant (see eq 13) are
A = (3.1934 ± 0.29) ·102 ; B = (1.1274 ± 0.13)· 104
(45)
Optimized UNIQUAC binary interaction parameters a(0)
ij and
a(1)
ij are given in Tables 6 and 7 respectively. Figure 4. Composition of the liquid phase for L = (2 to 4), W = 0 and
t = 180 °C: ▽, NH3; ○, CO2; □, H2O; ◇, (NH2)2CO; ×,
Table 6. UNIQUAC Model a(0)
ij Binary Interaction NH2COONH4; +, NH4HCO3 .
Parametersa
i j a(0)
ij /K Δa(0)
ij /K

1 2 159.8 800
1 3 3214.6 410
1 4 1664.2 700
1 5 −826.39 480
1 6 1288 2100
2 1 −304.9 530
2 4 441.25 220
3 1 −874.62 26
3 2 63.393 8.6
3 5 290.6 340
4 1 −560.18 87
4 3 −211.96 19
4 6 733.43 270 Figure 5. Composition of the gas phase in the bubble point for L =
(2.1 to 4), W = 0, and t = 180 °C. Solid line, NH3; dashed line, CO2;
5 3 538.17 250
dotted line, H2O.
5 4 59.254 33
6 1 −266.32 140
6 4 657.62 360
6 5 −374.31 340 Agreement of the optimized model parameters with
a
Δa(0)
ij was estimated using eq 41. experimental data will be discussed below.
Carbon Dioxide to Urea Conversion. The accuracy of
Table 7. UNIQUAC Model a(1)
ij Binary Interaction the equilibrium conversions of CO2 to urea (Y or YCO2)
Parametersa description is essential for correct modeling of the urea
i j a(1)
ij Δa(1)
ij
synthesis process. Results of YCO2 modeling are shown in Figure
1 3 −4.3216 0.76 6. Standard absolute deviation is 2.5%; maximal absolute
1 5 2.7073 1.1 deviation is 7.5%. Limits of accuracy of the model are
3 1 0.91464 0.071 connected mainly to accuracy of equilibrium conversion data,
3 5 −1.2096 0.70
4 1 0.380 0.024
4 6 −2.215 0.52
6 3 1.3333 0.19
6 5 1.1849 0.85
a
Δa(1)
ij was estimated using eq 41.

Compositions of the liquid phase and gas phase over it at

bubble point pressure for L = (2 to 4), W = 0 and t = 180 °C
have been calculated to test the thermodynamic model.
Obtained compositions of the liquid and gas phase are shown
in Figure 4 and Figure 5, respectively. As we can see molar
fractions of CO2 and NH4HCO3 species are increasing when L
decreases. In the gas phase the molar fraction of water is low
and rich with CO2 at L = 2.1 ( L = (2 to 2.1) points were not Figure 6. Carbon dioxide to urea conversion: experimental vs
calculated because pbubbl > 35 MPa for them) and rich with calculated values. Points, experimental data; ○, Kawasumi et al.;51−53
NH3 at L = 4. ×, Inoue et al.3,4

H DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

and accuracy is not higher when empirical functions and Table 8. 100·YCO2Equilibrium Conversion of CO2 to Urea
nonlinear regression is used for the YCO2 description. at t = (160 to 200) °C and Different Compositions of the
Although as it has been mentioned above that our model Liquid Phasea
takes into account the possible heterogenity of the autoclave
W
content during the experiment, it is reasonable to compare the
results with those existing in literature. A comparison of L t/°C 0.0 0.2 0.4 0.6 0.8 1.0
calculated conversion with empirical correlation by Gorkovskii 2.5 160 55.1 50.9 47.7 45.1 43.0 41.2
annd Kucheryavyi15 and the model of Isla, Genoud et al.8,9 at t 180 58.9 54.7 51.2 48.3 45.8 43.7
= 190 °C is given in Figure 7. As we can see the obtained values 200 62.1 57.6 53.7 50.3 47.3 44.6
of equilibrium conversions of carbon dioxide are in agreement 3.0 160 63.2 59.4 56.4 53.9 51.8 50.0
with the results from both of them. 180 66.0 62.2 58.9 56.2 53.8 51.7
200 68.0 63.9 60.4 57.3 54.5 52.0
3.5 160 70.6 67.2 64.4 62.0 59.9 58.1
180 72.6 69.1 66.1 63.4 61.1 59.0
200 73.5 69.8 66.5 63.6 61.0 58.5
4.0 160 76.6 73.7 71.2 69.0 67.0 65.3
180 78.0 75.0 72.3 69.8 67.6 65.6
200 78.2 74.9 71.9 69.1 66.6 64.2
4.5 160 81.1 78.7 76.6 74.7 72.9 71.3
180 82.2 79.6 77.2 75.1 73.1 71.2
200 81.8 78.8 76.1 73.6 71.2 69.0
5.0 160 84.3 82.4 80.6 79.0 77.5 76.1
180 85.1 83.0 81.0 79.2 77.4 75.7
200 84.3 81.7 79.3 77.0 74.8 72.7
5.5 160 86.6 85.0 83.6 82.3 81.0 79.8
180 87.2 85.5 83.8 82.2 80.7 79.2
Figure 7. Conversion of calculated conversion at t = 190 °C with other 200 85.9 83.6 81.4 79.3 77.4 75.5
models. Red symbols (○, L = 3.5; △, L = 4; ◇, L = 5; ▽, L = 5.5) a
represent the correlation of Gorlovski et al.,15 blue symbols (●, L = Expanded uncertainties (k = 2): U(YCO2) = 0.03 were estimated using
3.5; +, L = 4; ×, L = 4.5; ∗, L = 5) represent the model of Isla, Genoud eq 42.
et al.,8,9 lines represent our model. L = 3.5, 4, 4.5, 5.
Table 9. Equilibrium Bubble Point Pressure at t = (160 to
Calculated conversions and bubble point pressures for L = 200) °C and Different Compositions of the Liquid Phasea,
(2.5 to 5.5), W = (0.0 to 1.0) and t = (160 to 200) °C are given MPa
in Tables 8 and 9, respectively. W
Bubble Point Pressures. Results of equilibrium bubble
point pressures modeling are shown in Figure 8. Standard L t/°C 0.0 0.2 0.4 0.6 0.8 1.0
absolute deviation is 0.89 MPa, maximal absolute deviation is 2.5 160 6.9 6.9 6.9 6.8 6.7 6.7
3.8 MPa. 180 12.9 13.3 13.5 13.6 13.6 13.6
Calculated bubble point curves at W = 0 and t = (160 to 200) 200 26.5 27.8 28.5 28.9 29.2 29.5
°C and experimental points from refs 55 and 56 are shown in 3.0 160 7.9 7.6 7.3 7.0 6.8 6.5
Figure 9. As we can see calculated curves are in good agreement 180 12.3 12.1 11.8 11.6 11.4 11.1
with experimental data and reproduce the pressure minimum at 200 20.3 20.5 20.6 20.6 20.5 20.4
L = (2.2 to 3.5) for different temperatures. 3.5 160 8.9 8.6 8.2 7.8 7.5 7.2
One of the most important features of obtained model is its 180 13.3 12.9 12.5 12.1 11.7 11.3
possibility to reproduce the saddle azeotrope found by 200 20.2 20.0 19.7 19.4 19.1 18.8
Lemkowitz et al.6,43 Empirical coordinates of the azeotrope 4.0 160 9.8 9.3 8.9 8.5 8.1 7.8
from refs 6 and 43 are in good agreement with the results of 180 14.2 13.7 13.2 12.7 12.3 11.9
calculation (see Table 10). The parts of bubble- and dew-point 200 20.7 20.3 19.9 19.5 19.1 18.7
surfaces near the saddle azeotrope at t = 180 °C are shown in 4.5 160 10.5 10.0 9.5 9.1 8.7 8.3
Figure 10. 180 14.9 14.4 13.8 13.3 12.8 12.4
For an extra control of the saddle azeotrope description 200 21.2 20.7 20.3 19.8 19.3 18.9
accuracy bubble and dew point curves passing through the 5.0 160 11.2 10.6 10.1 9.6 9.2 8.8
azeotrope point at t = 160 and 180 °C have been calculated and 180 15.7 15.0 14.4 13.9 13.4 12.9
compared with experimental data from ref 43. It should be 200 21.7 21.2 20.6 20.1 19.7 19.2
noted that these data were digitized from diagrams from ref 43 5.5 160 11.8 11.2 10.6 10.2 9.7 9.3
but were not included in the parameters optimization 180 16.5 15.7 15.0 14.4 13.9 13.4
procedure and can be considered as a test data set. 200 22.2 21.6 21.0 20.5 20.0 19.5
a
Compositions of the sections can be approximately described Expanded uncertainties (k = 2): U(p) = 0.1 MPa estimated using eq
as zH2O = 1.16−2.09z2NH3 at 160 °C and zH2O = 1.17−2.00z2NH3 42.
at 180 °C (zH2O = (−0.7 to 0.7)). Results of calculations of the
sections and experimental data are shown in Figures 11 and 12. experimental data, that is, deviations are not higher than
Calculated bubble point curves are in a good agreement with standard deviations for bubble point pressure description. In
I DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

Figure 8. Bubble point pressure: experimental vs calculated values.

Symbols, experimental data: ∗, Kawasumi et al.,51−53 +, Inoue et al.;3,4
×, Lemkowitz et al.5,6,43,55,56

Figure 10. Bubble and dew point surfaces near the saddle azeotrope at
t = 180 °C.

Figure 9. Bubble point pressures at at W = 0 and t = (160 to 200) °C.

Symbols are experimental points from refs 55 and 56 ▽, (160 ± 1)
°C; ◁, (170 ± 1) °C; ◇, (180 ± 1) °C; ▷, (190 ± 1) °C; △, (200
± 1) °C. Curves are calculated in this work.
the case of dew point curves agreement is worse, especially near
the azeotrope point.
In the case of azeotrope coordinates at 160 and 180 °C
(Table 10) calculated pressures are within confidence intervals
and calculated values of z2NH3 and zCO2 are within confidence
intervals or near to its borders (not further than 0.005). In the
case of zH2O disagreement clearly exceeds the confidence
intervals.
Figure 11. Bubble and dew point curves near the saddle azeotrope at t
= 160 °C and t = 180 °C. Symbols, experimental data: +, bubble points
at 160 °C; ×, dew points at 160 °C; ○, bubble points at 180 °C; ◇,
dew points at 180 °C. Lines, calculated bubble and dew point curves.
Experimental data have been digitized from the diagram from ref 43
(see the original work for the composition of liquid and gas phase).
to urea in the liquid phase. Such description requires correct
work of the model for W < 0 liquid compositions.
It is important that all experimental works used in the model
optimization take into account a possibility of gas phase

■ CONCLUSIONS
Obtained thermodynamic model of the NH3−CO2−H2O−
(NH2)2CO system is valid for the urea synthesis conditions,
that is, for t = (135 to 230) °C, p = (3.5 to 45) MPa, L = (2 to
5.5), and W = (−0.75 to 1.2). It correctly describes vapor−
liquid equilibrium, that is, bubble and dew point pressures and
saddle azeotrope formation, and equilibrium conversion of CO2
formation inside an autoclave. Such formation can cause
differences between the composition of liquid and the ratios of
components inside the closed vessel. This feature is important
for correct estimation of urea yield.
Another important feature of the model is UNIQUAC
parameters optimized for the description of vapor−liquid
equilibrium in the NH 3−CO 2 , CO 2 −H 2 O and NH 3 −
(NH2)2CO binary subsystems.

Table 10. Coordinates of the Saddle Azeotrope Calculated in This Work (calc), and Experimental Data (exp) by Lemkowitz et
al.6,43

t/°C p/MPa z2NH3 (%) zCO2 (%) zH2O (%)

exp calc exp calc exp calc exp calc
160 7.2 ± 0.2 6.9 52.5 ± 2 53.9 40.0 ± 2 42.7 7.5 ± 2 3.3
180 12.1 ± 0.2 12.0 54.0 ± 2 56.7 38.5 ± 2 40.7 7.5 ± 2 2.6

J DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data
Figure 12. Coordinates of experimental and calculated points at t =
180 °C shown in Figure 11 and azeotropes. Measured compositions:
+, bubble point; ▽, dew point; △, saddle azeotrope. Calculated
compositions: ○, saddle azeotrope (see Table 10). Dashed line, linear
approximation of bubble point measurements.
Possible extensions of the model are the addition of such
constituents as biuret and inerts (i.e., N2, O2, H2, etc.) and
generalization to the lower-temperature (t < 135 °C) region.
■
ASSOCIATED CONTENT
*
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.jced.6b00557.
Supporting Information contains all source code written
in MATLAB and C++ languages code required to
reproduce the model parameters optimization procedure
described in this article. To extract all files it should be
launched in MATLAB interpreter. All source code can be
read by human without the extraction procedure(PDF)
■ AUTHOR INFORMATION
Corresponding Author
*E-mail: alvoskov@gmail.com. Tel.: +74959392280.
Funding
This work was financially supported by URALCHEM Holding
P.L.C.
Notes
The authors declare no competing financial interest.
■
NOMENCLATURES
Latin Letters
A, B = coefficients in KIII equilibrium constant (see eq 31 and
45)
A(k) = coefficients of the virial equation of state (see eq 2)
aij, a(0) (1)
ij , aij = UNIQUAC binary interaction parameters
between constituents i and j (see eq 23)
CE,
(j)
i = coefficients in the Gibbs energy of vaporization
temperature dependence (see eq 36 and Table 4) for ith
constituent.
CiV, (j) = coefficients in the molar volume temperature
dependence (see eq 35 and Table 4) for ith constituent.
H*CO2 = Henry constant of carbon dioxide in pure water
L = nNH ° 3/nCO° 2 = (see eq 14)
M(H2O) = molar mass of water, g·mol−1
n°i = molar amount of component: NH3, CO2 or H2O. Urea
is represented as negative amount of water (see Reaction (I)
in eq 13).
ni = molar amount of constituent
K DOI: 10.1021/acs.jced.6b00557
Article
p = pressure
pid = RT/V = pressure of ideal gas at volume V and
temperature T
ps,i = saturated vapor pressure over pure ith constituent
qi = UNIQUAC surface parameter for ith constituent
ri = UNIQUAC volume parameter for ith constituent
T = temperature, K
t = temperature, °C
Vm,i = molar volume of ith constituent
W = n°H2O/n°CO2 = (see eq 15)
xi = molar fraction of ith constituent in the liquid phase
Y, YCO2 = equilibrium conversion of carbon dioxide to urea in
the liquid phase
yi = molar fraction of ith constituent in the gas phase
Z = compressibility coefficient (see eq 2)
zi = molar fractions of component in special scale (suitable
for reciprocal system representation): i = 1, 2, 3 for 2NH3,
CO2, H2O, respectively. Urea is represented as negative
amount of water (see eq 19 and reaction (II) in eq 13).
z = UNIQUAC coordination number; z = 10.
Greek Letters
βi = generalized notation for normalized parameters of the
thermodynamic model of the liquid
γi = activity coefficients of ith constituent in the liquid phase
(see eq 25, 26 and 27)
ΔrGi = Gibbs energy of ith reaction.
ΔvGi = Gibbs energy of vaporization of i-th constituent.
δij = Kronecker delta
λ = Tikhonov regularization parameter (see eq 44)
νij = stoichiometric coefficient of ith constituent for a
reaction i
ψi = ln xi = logarithmic coordinates for numerical solving of
the eq 29.
τij = UNIQUAC binary interaction parameters based on aij
(see eq 23)
θi = UNIQUAC surface fraction of ith constituent (see eq
24)
ϕi = UNIQUAC volume fraction of ith constituent (see eq
24)
φ = fugacity coefficient of the gas phase (see eq 7)
φi = partial fugacity coefficient of ith constituent in the gas
phase (see eq 6)
ω = statistical weight in least-squares method (see eq 40)
Superscripts
ex = excess
exp = experimental
calc = calculated
comb = combinatorial
id = ideal
res = residual
■
REFERENCES
(1) Frèjacques, M. Les buses théoriques de la synthèse industrielle de
l’urée. Chim. Ind. 1948, 60, 22−35.
(2) Mavrovic, I. Find equilibrium urea yield. Hydrocarbon process.
1971, 161−162.
(3) Inoue, S.; Kanai, K.; Otsuka, E. Equilibrium of Urea Synthesis. I.
Bull. Chem. Soc. Jpn. 1972, 45, 1339−1345.
(4) Inoue, S.; Kanai, K.; Otsuka, E. Equilibrium of Urea Synthesis. II.
Bull. Chem. Soc. Jpn. 1972, 45, 1616−1619.
(5) Lemkowitz, S. M.; De Cooker, M. G. R. T.; Van Den Berg, P. J.
An empirical thermodynamic model for the ammonia-water-carbon
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data
dioxide system at urea synthesis conditions. J. Appl. Chem. Biotechnol.
1973, 23, 63−76.
(6) Lemkowitz, S. M.; Verbrugge, P.; van den Berg, P. J. A phase
model for the gas-liquid equilibria in the ammonia-carbon dioxide-
water-urea system in chemical equilibrium at urea synthesis conditions.
III. Comparison of the phase model with an empirical thermodynamic
model. J. Appl. Chem. Biotechnol. 1977, 27, 349−353.
(7) Piotrowski, J.; Kozak, R.; Kujawska, M. Thermodynamic model of
chemical and phase equilibrium in the urea synthesis process. Chem.
Eng. Sci. 1998, 53, 183−186.
(8) Isla, M. A.; Irazoqui, H. A.; Genoud, C. M. Simulation of a urea
synthesis reactor. 1. Thermodynamic framework. Ind. Eng. Chem. Res.
1993, 32, 2662−2670.
(9) Irazoqui, H. A.; Isla, M. A.; Genoud, C. M. Simulation of a urea
synthesis reactor. 2. Reactor model. Ind. Eng. Chem. Res. 1993, 32,
2671−2680.
(10) Bernardis, M.; Carvoli, G.; Santini, M. Proceedings of the Fifth
International Conference UREA-NH3−CO2−H2O: VLE Calculations
using an extended uniquac equation. Fluid Phase Equilib. 1989, 53,
207−218.
(11) Zhang, X.; Zhang, S.; Yao, P.; Yuan, Y. Modeling and simulation
of high-pressure urea synthesis loop. Comput. Chem. Eng. 2005, 29,
983−992.
(12) Hamidipour, M.; Mostoufi, N.; Sotudeh-Gharebagh, R.
Modeling the synthesis section of an industrial urea plant. Chem.
Eng. J. (Amsterdam, Neth.) 2005, 106, 249−260.
(13) Zendehboudi, S.; Zahedi, G.; Bahadori, A.; Lohi, A.; Elkamel, A.;
Chatzis, I. A dual approach for modelling and optimization of
industrial urea reactor: Smart technique and grey box model. Can. J.
Chem. Eng. 2014, 92, 469−485.
(14) Nakamura, R.; Breedveld, G. J. F.; Prausnitz, J. M.
Thermodynamic Properties of Gas Mixtures Containing Common
Polar and Nonpolar Components. Ind. Eng. Chem. Process Des. Dev.
1976, 15, 557−564.
(15) Gorlovskii, D. M.; Kucheryavyi, V. I. Equation of determination
of the equilibrium degree of CO2 conversion during synthesis of urea.
Zh. Prikl. Khim. (S.-Peterburg, Russ. Fed.) 1980, 53, 2548−2551.
(16) Piotrowski, J.; Zołotajkin, M.; Przybysławska, M. Equilibrium
conversion and pressure in the urea synthesis process. Polym. J. Appl.
Chem. 1992, 36, 277−284.
(17) Piotrowski, K.; Piotrowski, J.; Schlesinger, J. Modelling of
complex liquid-vapour equilibria in the urea synthesis process with the
use of artificial neural network. Chem. Eng. Process. 2003, 42, 285−289.
(18) Bernardis, M.; Carvoli, G.; Delogu, P. NH3−CO2−H2O VLE
calculation using an extended UNIQUAC equation. AIChE J. 1989,
35, 314−317.
(19) Thomsen, K.; Rasmussen, P. Modeling of vapor−liquid−solid
equilibrium in gas-aqueous electrolyte systems. Chem. Eng. Sci. 1999,
54, 1787−1802.
(20) Kuranov, G. L.; Smirnova, N. A. Calculation of phase behavior
and chemical transformations in H2O−NH3−CO2 system using a
modified hole quasichemical model. Russ. J. Appl. Chem. 2010, 83,
219−231.
(21) Xu, Y.; Wang, Z.; Liu, X.; Jin, B. Modeling of the NH3−CO2−
H2O vapor-liquid equilibria behavior with species-group Pitzer activity
coefficient model. Int. J. Greenhouse Gas Control 2014, 31, 113−120.
(22) Suzuki, K.; Onishi, S.-i.; Koide, T.; Seki, S. Vapor Pressures of
Molecular Crystals. XI. Vapor Pressures of Crystalline Urea and
Diformylhydrazine. Energies of Hydrogen Bonds in these Crystals.
Bull. Chem. Soc. Jpn. 1956, 29, 127−131.
(23) De Wit, H. G. M.; Van Miltenburg, J. C.; De Kruif, C. G.
Thermodynamic properties of molecular organic crystals containing
nitrogen, oxygen, and sulphur 1. Vapour pressures and enthalpies of
sublimation. J. Chem. Thermodyn. 1983, 15, 651−663.
(24) Krasulin, A. P.; Kozyro, A. A.; Kabo, G. Y. Saturated vapour
pressure of urea at 329−403 K temperature interval (in Russian). Zh.
Prikl. Khim. (S.-Petersburg, Russ. Fed.) 1986, 60, 104−108.
(25) Kabo, G. Y.; Miroshnichenko, E. A.; Frenkel’, M. L.; Kozyro, A.
A.; Simirskii, V. V.; Krasulin, A. P.; Vorob’eva, V. P.; Lebedev, Y. A.
L DOI: 10.1021/acs.jced.6b00557
Article
Thermochemistry of alkyl derivatives of urea. Bull. Acad. Sci. USSR,
Div. Chem. Sci. (Engl. Transl.) 1990, 39, 662−667.
(26) Kuznetsov, A. V.; Stolyarov, A. V. Nonempirical studies of the
molecular properties and thermodynamic functions of urea in the ideal
gas state. Russ. J. Phys. Chem. A 2009, 83, 270−275.
(27) Voskov, A. L.; Babkina, T. S.; Kuznetsov, A. V.; Uspenskaya, I.
A. Phase Equilibria in the Urea-Biuret-Water System. J. Chem. Eng.
Data 2012, 57, 3225−3232.
(28) Dzuban, A. V.; Voskov, A. L.; Uspenskaya, I. A. Phase Diagram
of HCOOK−(NH2)2CO−H2O System. J. Chem. Eng. Data 2013, 58,
2440−2448.
(29) Voronin, G. F.; Genkin, M. V.; Kutsenok, I. B. Virial equations
of state for gaseous ammonia, water, carbon dioxide, and their mixtures
at elevated temperatures and pressures. Russ. J. Phys. Chem. A 2015, 89,
1958−1970.
(30) Chase, M. W., Jr. J. Phys. Chem. Ref. Data. Monogr. No. 9. NIST-
JANAF Thermochemical Tables. Part I. Al−Co, 4th ed.; American
Institute of Physics, 1998.
(31) Chase, M. W., Jr. J. Phys. Chem. Ref. Data. Monogr. No. 9. NIST-
JANAF Thermochemical Tables. Part II. Cr−Zr, 4th ed.; American
Institute of Physics, 1998.
(32) Wagner, W.; Pruß, A. The IAPWS formulation 1995 for the
thermodynamic properties of ordinary water substance for general and
scientific use. J. Phys. Chem. Ref. Data 2002, 31, 387−535.
(33) Golubev, I. F.; Kiyashova, V. P.; Perel’shtein, I. I.; Parushin, E. B.
Thermophysical Properties of Ammonia; Izd-vo Standartov: Moscow,
1978; in Russian.
(34) Berman, R. G.; Brown, T. H. Heat capacity of minerals in the
system Na2O−K2O−CaO−MgO−FeO−Fe2O3−Al2O3−SiO2−TiO2−
H2O−CO2: representation, estimation, and high temperature
extrapolation. Contrib. Mineral. Petrol. 1985, 89, 168−183.
(35) Yorish, V. S.; Yungman, V. S. Thermal constants of substances
database (in Russian). [web-page] 2008; http://www.chem.msu.ru/
cgi-bin/tkv.pl (accessed on 19 May 2016).
(36) Wagman, D. D.; Evans, W. H.; Parker, V. B.; Schumm, R. H.;
Halow, I.; Bailey, S. M.; Nuttall, R. L. The NBS tables of chemical
thermodynamic properties. Selected values of inorganic and C1 and C2
organic substances in SI units. J. Phys. Chem. Ref. Data; American
Chemical Society, 1982; Vol. 11, Suppl. No 2.
(37) Clusius, K.; P, H. Ü ber die spezifischen Wärmen einiger fester
Kö r per bei tiefen temperaturen. Z. Phys. Chem., Stoechiom.
Verwandtschaftsl. 1928, 134, 243−263.
(38) Efremova, G. D.; Leont’yeva, G. G. Compressibility of ammonia
and carbon dioxide mixtures and equilibrium of carbamide synthesis
reaction. Khim. Prom-st. (St.Petersburg, Russ. Fed.) 1962, 742−747. In
Russian.
(39) Abrams, D. S.; Prausnitz, J. M. Statistical thermodynamics of
liquid mixtures: A new expression for the excess Gibbs energy of partly
or completely miscible systems. AIChE J. 1975, 21, 116−128.
(40) Kuramochi, H.; Noritomi, H.; Hoshino, D.; Nagahama, K.
Representation of activity coefficients of fundamental biochemicals in
water by the UNIFAC model. Fluid Phase Equilib. 1997, 130, 117−
132.
(41) Lourakis, M. levmar: Levenberg-Marquardt nonlinear least
squares algorithms in C/C++. [web page] Jul. 2004; http://www.ics.
forth.gr/~lourakis/levmar/ (accessed 24 Aug. 2016).
(42) Edwards, T. J.; Maurer, G.; Newman, J.; Prausnitz, J. M. Vapor-
liquid equilibria in multicomponent aqueous solutions of volatile weak
electrolytes. AIChE J. 1978, 24, 966−976.
(43) Lemkowitz, S. M.; Vet, E.; van den Berg, P. J. A phase model for
the gas-liquid equilibria in the ammonia-carbon dioxide-water-urea
system in chemical equilibrium at urea synthesis conditions. II.
Experimental verification. J. Appl. Chem. Biotechnol. 1977, 27, 335−
348.
(44) Tillner-Roth, R.; Friend, D. G. A Helmholtz Free Energy
Formulation of the Thermodynamic Properties of the Mixture Water
+ Ammonia. J. Phys. Chem. Ref. Data 1998, 27, 63−96.
(45) Müller, G.; Bender, E.; Maurer, G. Das Dampf-Flüssigkeits-
gleichgewicht des ternären Systems Ammoniak-Kohlendioxid-Wasser
J. Chem. Eng. Data XXXX, XXX, XXX−XXX
Journal of Chemical & Engineering Data Article

bei hohen Wassergehalten im Bereich zwischen 373 und 473 K. Ber.

Bunsen-Ges. 1988, 92, 148−160.
(46) Takenouchi, S.; Kennedy, G. C. The binary system H2O−CO2
at high temperatures and pressures. Am. J. Sci. 1964, 262, 1055−1074.
(47) Bamberger, A.; Sieder, G.; Maurer, G. High-pressure (vapor
+liquid) equilibrium in binary mixtures of (carbon dioxide+water or
acetic acid) at temperatures from 313 to 353 K. J. Supercrit. Fluids
2000, 17, 97−110.
(48) Zawisza, A.; Malesinska, B. Solubility of carbon dioxide in liquid
water and of water in gaseous carbon dioxide in the range 0.2−5 MPa
and at temperatures up to 473 K. J. Chem. Eng. Data 1981, 26, 388−
391.
(49) Hu, J.; Duan, Z.; Zhu, C.; Chou, I.-M. PVTx properties of the
CO2−H2O and CO2−H2O−NaCl systems below 647 K: Assessment
of experimental data and thermodynamic models. Chem. Geol. 2007,
238, 249−267.
(50) Kosova, D. A.; Voskov, A. L.; Kovalenko, N. A.; Uspenskaya, I.
A. A Water-Urea-Ammonium Sulfamate system: Experimental
investigation and thermodynamic modelling. Fluid Phase Equilib.
2016, 425, 312−323.
(51) Kawasumi, S. Equilibrium of the CO2−NH3−H2O System
under High Temperature and Pressure. II. Liquid-Vapor Equilibrium
in the Loading Mole Ratio of 2NH3 to CO2. Bull. Chem. Soc. Jpn. 1952,
25, 227−238.
(52) Kawasumi, S. Equilibrium of the CO2−NH3−Urea−H2O
System under High Temperature and Pressure. III. Effect of Water
Added on Liquid-Vapor Equilibrium. Bull. Chem. Soc. Jpn. 1953, 26,
218−222.
(53) Kawasumi, S. Equilibrium of the CO2−NH3−Urea−H2O
System under High Temperature and Pressure. V. Liquid-Vapor
Equilibrium in the Presence of Excess Ammonia or Carbon Dioxide.
Bull. Chem. Soc. Jpn. 1954, 27, 254−259.
(54) Satyro, M. A.; Li, Y.-K.; Agarwal, R. K.; Santollani, O. J.
Modeling Urea Processes: A New Thermodynamic Model and
Software Integration Paradigm. Clearwater Convention on Phosphate
Fertilizer & Sulfuric Acid Technology. Sheraton Sand Key Resort,
Clearwater Beach, Florida, USA, 2000.
(55) Lemkowitz, S. M.; Goedegebuur, J.; Berg, P. J. Bubble-point
measurements in the ammonia-carbon dioxide system. J. Appl. Chem.
Biotechnol. 1971, 21, 229−232.
(56) Lemkowitz, S. M.; Zuidam, J.; van den Berg, P. J. Phase
behaviour in the ammonia-carbon dioxide system at and above urea
synthesis conditions. J. Appl. Chem. Biotechnol. 1972, 22, 727−737.

M DOI: 10.1021/acs.jced.6b00557
J. Chem. Eng. Data XXXX, XXX, XXX−XXX