Fuel 252 (2019) 439–447

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Application of Peng-Robinson equation of state for modelling the multiphase T

equilibrium properties in Athabasca bitumen/ethane mixtures
⁎
Ting Suna, Ali Takbiri-Borujenib, , Hossein Nouroziehc, Ming Gub
a
China University of Petroleum, Beijing, China
b
West Virginia University, United States
c
Computer Modelling Group Ltd, Canada

A R T I C LE I N FO A B S T R A C T

Keywords: Phase behavior of (bitumen + solvent) mixtures and their properties are essential for surface upgrading methods
Phase behavior and enhanced oil recovery applications. The objective for this work is to systematically model the phase par-
Bitumen titioning and component distribution in the (bitumen + solvent) mixtures. To perform equation of state mod-
Peng-Robinson equation of state elling, compositional analysis of bitumen is utilized to characterize the bitumen sample and its components up to
Solvent
C100. The full range of components is utilized in Peng-Robinson equation of state (PR-EOS) to predict the
equilibrium phase properties, such as composition and density, in vapor-liquid and liquid-liquid regions. Based
on the results, PR-EOS accurately captures the composition of phases, component distribution in each phase, and
phase densities. Thereafter, an algorithm is proposed to lump the components on the basis of two-phase envelope
generation. The results indicate that Athabasca bitumen and its phase behavior in (bitumen + solvent) mixtures
can be well represented in multi-phase regions by four pseudo-components.

1. Introduction (bitumen + solvent) mixtures.

Determining phase behavior and estimating the thermodynamic
properties for complex fluid systems or ill-defined fluids, such as (bi-
tumen + solvent) mixtures, present significant challenges. Despite the
importance of phase behavior data for production, pipeline transpor-
tation, upgrading, and refining of such fluids, limited experimental data
for bitumen-containing mixtures have been reported in the literature.
This is due to the fact that experimental measurements for these sys-
tems are usually expensive and time-consuming. This issue is even more
pronounced when multiple phases form at equilibrium conditions.
Modelling is a credible alternative for studying the phase behavior and
the interaction of (bitumen + solvent) systems.
The thermodynamic modelling of (bitumen + solvent) systems is
limited compared to simple hydrocarbon systems and conventional oil.
This might be due to complex nature and lack of a known character-
ization method for heavy oil and bitumen. Although different char-
acterization schemes, as stated by Kumar and Okuno [1], have been
proposed for different reservoir fluids [2–6]; for example, gas con-
densates [7–10], volatile oils [11,12], near-critical fluids [13–16], the
methods for heavy oil and bitumen characterizations are limited
[1,17–18]. This led to variety of models in literature for pseudo-com-
ponent determination in the phase behavior study of
Mehrotra et al. [19] utilized Bishnoi et al. [20] characterization
method to develop five pseudo-component in order to study the phase
behavior and solubility of carbon dioxide and ethane in Athabasca and
Peace River bitumens using Peng-Robinson equation of state (PR-EOS).
The critical properties calculated from Kesler and Lee [21] correlation
provided a good agreement with the experimental solubility data. Fu
and Puttagunta [22] and Fu et al. [23] considered a pseudo-component
in the vapour-liquid equilibrium calculations of (bitumen + solvent)
systems for the sake of simplicity. They developed a temperature-de-
pendent correlation for binary interaction parameters to represent the
vapour-liquid data. Mehrotra and Svrcek [24,25] and Mehrotra et al.
[26] utilized PR-EOS along with Kesler and Lee [21] correlation for
calculating critical properties to estimate the nitrogen, carbon mon-
oxide, methane, carbon dioxide, and ethane solubility in Athabasca,
Cold Lake, and Wabasca bitumen at the temperatures up to 110 °C and
at the pressures up to 10 MPa. The bitumen was characterized using
three pseudo-components which represent distillable maltene, un-
distillable maltene, and asphaltene that could adequately match the
experimental data. Frauenfeld et al. [27] estimated the methane,
ethane, propane, and carbon dioxide solubility in Lloydminster Aber-
feldy oil and a blended Cold Lake/Lloydminster oil using PR-EOS. The
binary interaction parameters were adjusted to tune the model.

⁎
Corresponding author.
E-mail address: Altakbiri@mail.wvu.edu (A. Takbiri-Borujeni).

https://doi.org/10.1016/j.fuel.2019.04.106
Received 28 February 2019; Received in revised form 11 April 2019; Accepted 16 April 2019
Available online 01 May 2019
0016-2361/ © 2019 Elsevier Ltd. All rights reserved.
T. Sun, et al.
Kariznovi et al. [28] modelled the phase behaviour of (bi-
tumen + solvent) mixtures using PR-EOS for four bitumens (Athabasca,
Peace River, Wabasca, and Cold Lake) mixed with four different sol-
vents (nitrogen, carbon dioxide, methane, and ethane). They used the
lab data of Mehrotra and Svrcek [24,25,29] and Svrcek and Mehrotra
[30,31] and considered three, five and six pseudo-components which
were defined on the basis of different fractions of bitumen. A universal
set of pseudo-components was also proposed applicable for different
Alberta bitumens.
Saber and Shaw [32] and Saber et al. [33] used a group contribution
method to estimate the critical properties of bitumen pseudo-compo-
nents for equation of state phase behavior modelling of (Athabasca
bitumen + n-decane) mixtures. Diaz et al. [34] studied the phase be-
haviour for the mixtures of Athabasca bitumen, propane, and carbon
dioxide using different characterization methods. They utilized simu-
lated distillation (SimDis) data along with six characterization ap-
proaches to predict the saturation pressures and liquid-liquid phase
boundaries of pseudo-ternary systems.
Agrawal et al. [35] modelled the saturation pressure of live Peace
River bitumen at different condensate concentrations with PR-EOS
using the characterization method proposed by Diaz et al. [34] through
the adjustment of binary interaction parameters. They fitted the sa-
turation pressure and asphaltene onset data with temperature-depen-
dent binary interaction parameters. The developed model reasonably
predicted the saturation pressures of methane/bitumen, ethane/bi-
tumen, carbon dioxide/bitumen, and n-pentane/bitumen. Li and Yang
[36] modelled the phase behaviour of propane/n-butane/heavy-oil
systems at the temperatures up to 123 °C and the pressures up to
5030.0 kPa with a volume-translated PR-EOS coupled with a modified
alpha function. The experimental data was accurately matched with the
binary interaction correlations and the volume-translated equation of
state, while prediction accuracy was decreased near the pure solvent
critical temperature.
Badamchi-Zadeh et al. [37] proposed a bitumen characterization
method with an equation of state model for the carbon dioxide, pro-
pane, and Athabasca bitumen mixtures. Four pseudo-components were
defined for bitumen based on conversion of the SimDist data of as-
phaltene free bitumen to a normal boiling point distribution. Then, the
binary interaction parameters were tuned to match the saturation
pressure. Yang et al. [38] modelled the phase behavior of methane/
propane/heavy-oil (Lloydminster) mixture using the volume-translated
PR-EOS along with a modified alpha function at the pressures up to
10 MPa and the temperature range of 22–50 °C. A new correlation for
the binary interaction parameter was proposed in order to match the
measured swelling factors and saturation pressures. Ghasemi and
Whitson [39] proposed a systematic approach to model the phase be-
haviour of the mixtures containing light solvents (nitrogen, carbon di-
oxide, carbon monoxide, methane, and ethane) and Athabasca bitumen
using cubic equations of state. Gamma distribution was used to fit the
bitumen distillation data; then, single-carbon-number (SCN) fractions
were defined, and Twu [40] correlation was used to calculate the cri-
tical properties of fractious. Finally, SCN fractions were lumped into
five pseudo-fractions.
Gao et al. [41] observed the multiphase behavior for (bitumen + n-
butane) mixtures in their phase behavior studies at pressures up to
10 MPa and temperatures up to 160C. Eghbali and Dehghanpour [42]
presented experimental and modelling study for multiphase behavior of
(Clearwater bitumen + butane) and (Clearwater bitumen + CO2) mix-
tures. They calibrated PR-EOS to predict (bitumen + solvent) phase
behavior. Kumar and Okuno [43] proposed a method for character-
ization of bitumen by improving the fluid characterization method
which was developed based on perturbation from n-alkanes (PnA). The
authors showed that LLE data needed positive interaction coefficients
between (bitumen + solvent) while VLE was not sensitive to the in-
teraction coefficients used. Azinfar et al. [44] fractionated Athabasca
bitumen to four cuts by using modified vacuum distillation. The PR-EOS
440
Fuel 252 (2019) 439–447
was then tuned to match solubility data of propane in each bitumen cut.
The bitumen was also characterized using the boiling point or carbon
number distribution and the generalized model was then proposed to
obtain the solubility of propane in the bitumen samples.
Most of the above-mentioned studies have focused on the modelling
of the vapor-liquid equilibrium and solubility data. The phase behavior
of (bitumen + solvent) mixtures is still challenging when the multi-
phase equilibrium occurs at specific conditions. The objective of this
study is thus to develop a characterization method for heavy oil and
bitumen. The method is applicable for both the multiple-phase equili-
brium conditions, and the development of PVT models for the purpose
of simulation study. In addition, it tests the applicability of equation of
state for phase behavior modelling study of (bitumen + solvent) mix-
tures. The characterization is based on the simulated distillated data
and physical properties of bitumen coupled with available correlations
for the calculation of critical properties. WinProp software from
Computer Modelling Group (CMG) version 2017 was used to model the
phase behaviour of the (bitumen + solvent) mixtures. The phase be-
havior predictions were examined with the available literature data for
bitumen/ethane mixtures including vapor-liquid (VLE) and liquid-li-
quid equilibrium (LLE) data. The compositional analysis, phase com-
positions, and phase densities are compared against the experimental
data and the phase diagrams are generated. These information are ex-
tremely important for surface upgrading, production, pipeline trans-
portation, and refining.
2. Equilibrium calculation
State of the system is determined by fugacity of each component in
each phase at equilibrium conditions. The equilibrium condition for
species i in phase I and II is presented as,
fiI = fiII (1)
In terms of the fugacity coefficients, Eq. (1) can be re-written as,
yi φi ̂ P = x i φiIÎ P,
I
(2)
where the partial fugacity coefficients are calculated using,
∞
∫v ⎛ ⎡ ∂∂nPi ⎤
⎜ −
RT ⎞
v
⎟ dv
ln φi ̂ = ⎝ ⎣ ⎦T , v, nj ≠ i ⎠ − ln Z ,
RT (3)
where P is pressure, T is temperature, v is volume, R is gas universal
constant, and Z is compressibility factor. Equation of state (EOS) is
typically utilized to express the relationship between the pressure,
temperature and volume. Peng-Robinson (Robinson and Peng [45])
EOS, a cubic EOS, provides acceptable results for the hydrocarbon
mixtures and petroleum systems and thus, it was employed in this study
to calculate the equilibrium properties of bitumen containing mixtures.
The general form of Peng-Robinson EOS is shown as,
RT aλ (T )
P= − ,
v−b v (v + b) + b (v − b) (4)
where,
0.45724R2Tc2
a=
Pc (5)
0.07780RTc
b= ,
Pc (6)
λ (T ) = [1 + κ (1 − Tr1/2)]2 , (7)
κ = 0.379642 + 1.48503ω − 0.164423ω2 + 0.016666ω3 (8)
where P is pressure, T is temperature, v is volume, a represents the
attractive forces between molecules, b accounts for the volume of mo-
lecules, Tc, and Pc are the critical temperature and pressure. a and b are
T. Sun, et al. Fuel 252 (2019) 439–447

calculated from the critical properties of the component for a single Table 1
component system. However, for multicomponent mixtures, a and b are Bitumen properties [58].
defined based on the individual components’ a and b using the van der Molecular Weight (g/mol) 512.5 ± 6.9
Waals mixing rules:
SARA Fraction (wt%)
am = ∑ xi Di , Saturates 11.76
i (9) Aromatics 57.00
Resins 21.61
Di = ai ∑j xj (1 − δij ) aj , and
(10)
Asphaltenes 9.62

bm = ∑ xi bi , Component volume shifts can be evaluated by finding the difference

i (11)
between the saturated liquid molar volume of a pure component ob-
where δij represents the binary interaction between the hydrocarbons in tained from the EOS and the saturated liquid molar volume calculated
the mixture, which is typically defined as a function of the critical using a modified Rackett‘s Z-factor equation at any specific temperature
properties. In WinProp, the binary interaction parameters are related to [52],
critical volume of components through the following equation [46,47], 2/7
[1 + (1 − Tr ) ]
vs = (RTc / Pc ) ZRA (17)
1/6 1/6 n
⎛ 2vci vcj ⎞
δij = 1 − ⎜ v1/3 + v1/3 ⎟ Rackett’s compressibility factors are back-calculated from the spe-
⎝ ci cj ⎠ (12) cific gravity and critical properties with the assumption that the specific
Oellrich et al. [48] showed that a value of 1.2 for n provides an gravity is roughly equal to the saturated liquid density at 60°F.
acceptable match with the experimental results for the paraffin-paraffin
interaction coefficients. This equation satisfactorily reproduces the 4. Viscosity calculation
methane interaction coefficients reported by Katz and Firoozabadi [49]
(Mehra [46]). In this study, n was considered as a tuning parameter to In WinProp, Pedersen viscosity correlation [53] is used to calculate
fit the experimental VLE and LLE data. Worth to mention that, Diaz the viscosity of high viscous fluids such as heavy oil and bitumen. This
et al. [34] defined the binary interaction parameters as a function of correlation is based on the corresponding state method, and it calcu-
components’ critical temperatures to match the saturation pressure data lates the viscosity of mixtures with respect to a reference substance at
of (Athabasca bitumen + solvent) mixtures. the same reduced temperature and pressure using the following rela-
tion,
3. Density calculation −1/6 2/3 1/2
μmix (T , P ) Tc,mix ⎞ ⎛ Pc,mix ⎟⎞ ⎛ MWmix ⎞ ⎛⎜ βmix ⎞⎟,
= ⎜⎛ ⎟ ⎜ ⎜ ⎟

μref (Tref , Pref ) ⎝ Tc,ref ⎠ ⎝ Pc,ref ⎠ ⎝ MWref ⎠ ⎝ βref ⎠ (18)

EOS can also be utilized to determine the volumetric properties of
petroleum/hydrocarbon mixtures. Peneloux et al. [50] proposed a vo- where Tc, and Pc are the critical temperature and pressure. The mole-
lume translation technique to be applied to Peng-Robinson EOS in order cular weight (MW) is calculated using both weight and molar averaged
to improve the volumetric calculations without impacting the other values as,
properties. In the volume shift technique, the volume is corrected by the
following relation, MWmix = d1 (MW wd2 − MWnd2) + MWn (19)

v cor = v − c, (13)
cor
where v is corrected molar volume and c in translation parameter. For
multi-components systems, c is defined as,
N
c= ∑ x i ci ,
i=1 (14)
where xi is the mole fraction of component i in the mixture and ci is
defined as
ci = Ωb vsi RTc / Pc (15)
where Ωb is a constant specific to the EOS (for Peng-Robinson EOS, it is
0.077796), vs is the volume translation, Tc is the critical temperature,
and Pc is the critical pressure. vs is also considered as a tuning para-
meter that commonly obtained through matching the lab volumetric
data (e.g. densities, relative volumes, and etc.). Thus, the volume
translation technique is available to modify the calculated volumetric
and also to match the bitumen and equilibrium phases’ densities.
In WinProp, three methods are available for the volume shift: con-
stant volume shift, temperature dependent volume shift correlation,
and linear temperature dependent volume shift. Using constant volume
shift option, the densities of the equilibrium phases are inaccurately
modeled with the EOS particularly when a wide range of temperatures
is considered. Thus, it is essential to define temperature-dependent
volume shifts. Pedersen et al. [51] proposed the following equation to
calculate the temperature-dependent volume shifts for each component,
S = S0 + S1 (T − Tref )
(16)
The rotational coefficient (β) represents the deviation from simple
corresponding states, and it is defined as,
d4
β = 1 + d3 (ρ / ρc )ref MW d5 (20)
where ρc, and ρ are the critical density and mixture density. The visc-
osity of a mixture calculated using the Pedersen model is dependent on
the critical pressures, critical temperatures, molecular weights of each
component, and the coefficients di (as shown in the above equations). A
modified version of Pedersen correlation for methane viscosity gives a
better correlation of the viscosity data [54].
5. Bitumen characterization
Before applying an EOS for a phase behavior study, the fluid sam-
ples should be thoroughly characterized. The characterization of heavy
fractions such as heavy oil and bitumen is generally on the basis of the
boiling point distribution from atmospheric and vacuum distillation.
Crude oils and distillate fractions contains thousands of components
with complex molecular structures that typically show a continuous
boiling point distribution. Two commonly used approaches for oil
characterization are the characterization of the crude oil or its fractions:
(i) in terms of pseudo-components that are defined by hydrocarbon
type, e.g., paraffinic, olefinic, naphthenic or aromatics, or an average
boiling point (ii) the continuous mixture approach, in which the oil is
represented by a continuous distribution function with respect to
boiling point, molecular mass, and/or some other measurable proper-
ties. The continuous mixture approach is adequate when the oil consists
of one type of hydrocarbon groups, such as paraffins or naphthenic;

441
T. Sun, et al. Fuel 252 (2019) 439–447

3
Exp
Model

Composition (wt%) 2

0
5 15 25 35 45 55 65 75 85 95
Single Carbon Number
Fig. 1. Component distribution in raw Athabasca bitumen; experimental data from Kariznovi et al. [57].

1000 Table 2
Pseudo-components properties for PR-EOS; MW, molecular weight; Pc, critical
pressure; Tc, critical temperature; SG, specific gravity; w, acentric factor.
800
Component MW (g/g-mol) Pc (MPa) Tc (°C) SG ω
Boiling Point (°C)

C6-19 210.2 2.234 461.9 0.892 0.575

600 C20-34 360.4 1.482 587.4 0.952 0.877
C35-99 806.2 0.833 746.4 1.021 1.289
C100+ 2034.0 0.542 959.5 1.153 1.590

400
however, heavy oil and bitumen are a complex mixture composed of
different hydrocarbons types (e.g. saturates, aromatics, etc.).
200
In this study, the pseudo-component approach is employed for the
bitumen characterization. The pseudo-components are primarily de-
fined, and their critical properties are estimated and along with the
0
0 20 40 60 80 100 component’s composition are applied into the Peng-Robinson EOS. The
multi-phase equilibrium is then calculated by a flash method combined
Cumulative Distilled (wt%)
with a phase stability check. In this study, WinProp software from
Fig. 2. Boiling point temperature with respect to weight percent distilled for Computer Modelling Group (CMG) was used to model the phase be-
raw Athabasca bitumen (Δ) taken from Kariznovi et al. [57].and proposed haviour of the bitumen containing mixtures. The tangent-plane stability
characterization scheme (line). criterion is used in WinProp [55]. This method is on the basis of a stage-
wise procedure in which the number of phases gradually increases. The
2500 equilibrium equation (Eq. (3)) is solved with Quasi-Newton Successive
Substitution (QNSS) [56].
The compositional analysis of Athabasca bitumen sample was al-
2000
ready reported by Kariznovi et al. [57], and was used to define the
pseudo-components and characterize the bitumen. Table 1 summarizes
Pressure (kPa)

1500 the bitumen properties [58]. Based on the lab distillation data, com-
ponents were defined for the distillable fraction of Athabasca bitumen,
which accounts for about 80 percent (in mass) of the fluid, and also
1000 matches the component distribution. Riazi [59] developed the fol-
lowing correlation to relate the component’s boiling point to corre-
sponding carbon number,
500
Tb = 1090 − exp(6.9955 − 0.11193Nc2/3), (21)

0 where Tb is the boiling point in Kelvin, and Nc is carbon number. Riazi

0 200 400 600 800
[59] showed that the data of Katz-Firoozabadi [49] is represented by
Temperature (°C) this equation with an average absolute deviation (AAD) of 0.2%.
Fig. 3. Pressure-temperature diagram for full characterization (solid line) and
Riazi [59] also utilized the physical properties of hydrocarbons
lumped schemes (dash line). (C6–C22) to develop correlations for the estimation of molecular weight
as a function of normal boiling point, and the prediction of specific
gravity as a function of molecular weight.
The molecular weight of each single carbon number, as proposed by
Riazi [59], can be estimated by the following equation,

442
T. Sun, et al. Fuel 252 (2019) 439–447

10 10 Fig. 4. The pressure-composition (P-x)

diagram (a) and coexisting equilibrium
phase compositions (b) for the mixture

Equilibrium Pressure (MPa)

8 Single L 8 of (Athabasca bitumen + ethane) at
ambient temperature; ○, Mehrotra and
LLE LLE Svrcek [29] data; ■, Nourozieh et al.
Pressure (MPa)

6 6 [65] data; solid line, equation of state

using full characterization scheme;
dashed line, equation of state using
4 4 lumped scheme.

VLE VLE
2 2

(a) (b)
0 0
0 25 50 75 100 0 25 50 75 100
Solvent Overall Composition (wt%) Equilibrium Compositions (wt%)

100 100 Fig. 5. Effect of overall solvent com-

Co-excisting Phase Composition
Co-excisting Phase Composition

position on the co-existing phase com-

position for mixtures of (Athabasca bi-
75 75 tumen + ethane) at 5 MPa (a) and
9 MPa (b) and ambient temperature;▲,
solvent-enriched phase; □, asphaltene-
(wt%)
(wt%)

50 50 enriched phase (data from Nourozieh

et al. [65]); solid line, equation of state
using full characterization scheme; da-
25 25 shed line, equation of state using
lumped scheme.
(a) (b)
0 0
20 40 60 80 100 20 40 60 80 100
Overall Solvent Composition (wt%) Overall Solvent Composition (wt%)

MW = [[6.97996 − ln(1080 − Tb)]/0.01964]3/2 (22) calculating the critical properties of Alberta bitumen and they con-
cluded that the Kesler and Lee [21] correlations provided the most
where Tb is the boiling point. The specific gravity of each component is acceptable results for bitumen/gas systems. Kariznovi et al. [28] and
also related to molecular weight by, Diaz et al. [34] also examined different correlations to calculate critical
SG = 1.07 − exp[3.56073 − 2.93886 × MW 0.1] (23) properties and acentric factor of non-distillable fractions. They came
into the conclusion that the phase boundaries are better predicted with
where MW is the molecular weight. Eqs. (21)–(23) were used to cal- Lee-Kesler correlation compared to other available methods. Thus, in
culate the boiling points, molecular weights, and specific gravities of this study, the Kesler and Lee [21] correlation was used to calculate the
SCN groups. The boiling points of the components along with their critical properties and acentric factor of the components.
compositions reasonably represent the experimental distillation data of Kesler and Lee [21] proposed the following equations,
Athabasca bitumen. The un-distilled fraction of Athabasca bitumen was
represented by a single component and assigned an average molecular
Tc = ⎛341.7 + 811SG + [0.76392 + 0.21132SG] Tb
⎜

weight based on Diaz et al.’s [34] work. Alboudwarej et al. [60] pro- ⎝
posed the following equation for asphaltene fractions to relate the (0.25939 − 1.81239SG ) × 105 ⎞
specific gravity and molecular weight, + /1.8 ⎟

Tb ⎠ (25)
SG = 670 × MW 0.0629 (24)

Eq. (24) along with the average molecular weight of asphaltene was Pc = exp[8.3634 − 0.0566/ SG − (0.24244 + 2.2898/ SG + 0.11857/SG2)
used to estimate the specific gravity of non-distillable fraction in this × 10−3 (1.8 × Tb)+(1.4685 + 3.648/ SG + 0.47227/SG2)
study. Thus, on the basis of the developed characterization scheme, the × 10−7 (1.8 × Tb)2 − (0.42019 + 1.6977/SG2) × 10−10 (1.8 × Tb)3]/14.7
distillable fraction of bitumen is represented by single carbon numbers,
(26)
and the non-distillable fraction by one pseudo-component (composition
of C100+ is 18.76 mass percent). The comparison of experimental and
developed characterization scheme in terms of the component dis-
ω=
(ln P − 5.92714 +
cr
6.09648
Tbr
+ 1.28862 ln Tbr − 0.169347Tbr6 ) for T
br ⩽
tribution and boiling point curve of Athabasca bitumen is presented in
Figs. 1 and 2. (15.2518 − 15.6875
Tbr
− 13.4721 ln Tbr + 0.43577Tbr6 )
The critical properties and acentric factor of components are typi- 0.8 (27)
cally correlated through the components’ molecular weight, specific
gravity, and boiling point. Three commonly used correlations are ω = −7.904 + 0.1352Kw − 0.007465K w2 + 8.359Tbr
available in WinProp to calculate these properties: Kesler and Lee [21],
(1.408 − 0.01063Kw )
Twu [40], and Riazi and Daubert [61]. Svrcek and Mehrotra [30,31] + for Tbr > 0.8
Tbr (28)
and Mehrotra et al. [19] evaluated a number of these correlations for

443
T. Sun, et al. Fuel 252 (2019) 439–447

10
L1 Exp.
(a)
L1 EOS
8 L2 Exp.

Composition (wt%) L2 EOS

0
5 15 25 35 45 55 65 75 85 95
Single Carbon Number

10
L1 Exp.
(b)
L1 EOS

8 L2 Exp.
L2 EOS
Composition (wt%)

0
5 15 25 35 45 55 65 75 85 95
Single Carbon Number
Fig. 6. Component distribution in ethane-free liquid samples, extract (black and green) and residue (blue and red), from LLE study of (Athabasca bitumen + ethane)
mixtures at 5 MPa (a) and 9 MPa (b) and a constant overall ethane composition of 40 wt%; blue and black, experimental data from Kariznovi et al. [57]; red and
green, predictions by PR-EOS using full characterization scheme. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

(1.8Tb)1/3 rule [63]. This scheme calculates the number of components using the
Kw =
SG (29) following equation,

where P is pressure in atm, T is temperature in K, Pcr and Tbr are reduced Ng = Int [1 + 3.3log10 (N − n)] (30)
critical pressure and reduced boiling point temperature, respectively.
As discussed earlier, the components’ critical volumes are needed in Eq. where, n and N are the minimum carbon number and the maximum one
(12) to calculate the binary interaction parameters between the com- in the sample, respectively. WinProp has two options for lumping into
ponents. In this study, Twu’s correlation [40] was employed to calcu- hypothetical components: default and user defined lumping scheme.
late the critical volumes of the components using their boiling point, The default lumping is according to components log (k) values and the
molecular weight, and specific gravity. k-values are calculated using Wilson’s equations. The Lee-Kesler [64]
mixing rule are then used to calculate the critical properties of the
components,
6. Component lumping
1
vc =
8
∑ ∑ xi xk (vci1/3 + vck1/3 )3
Following the characterization of Athabasca bitumen, lumping or i j (31)
grouping of the components into fewer numbers of components is re-
quired because large number of pseudo-components is not commonly 1
practical for modelling study. The accuracy of predictions could be Tc =
8vc
∑ ∑ xi xk (vci1/3 + vck1/3 ) Tci Tcj
i j (32)
reduced during lumping step while the calculation time is significantly
improved. The adequate number of pseudo-components mainly de-
pends on the mixture/system under study. Whitson [62] proposed a ω= ∑ xi ωi
i (33)
lumping approach that is developed originally on the basis of Sturges’

444
T. Sun, et al. Fuel 252 (2019) 439–447

100 80 10000000
Cumulative Carbon Number Ditsribution (L1)

Cumulative Carbon Number Distribution (L2)

1000000
80
60

Bitumen Viscosity (mPa.s)

100000

60
10000
40

40 1000

20 100
20
10

0 0
0 20 40 60 80 100 1
0 50 100 150 200
Single Carbon Number
Temperature (°C)
Fig. 7. Cumulative carbon number distribution in gas-free liquid samples from
LLE study of (Athabasca bitumen + ethane) mixtures at 5 MPa (Δ, ▲) and Fig. 9. Comparison of calculated and measured bitumen viscosities; ■,
9 MPa (■, □) and a constant overall ethane composition of 40 wt%; dots, Nourozieh et al. [65] data; solid line, corresponding state using full char-
Kariznovi et al. [57] data; lines, predictions by PR-EOS using full character- acterization scheme; dashed line, corresponding state using lumped scheme.
ization scheme.
10
1030

8
Equilibrium Pressure (MPa)

1005 LLE
Bitumen Density (kg/m3)

980 6

955 4

VLE
930
2

905
0
0 200 400 600 800 1000
880
0 50 100 150 200 Density of Equilibrium Phases (kg/m3)
Temperature (°C) Fig. 10. The coexisting equilibrium phase densities for the mixture of
(Athabasca bitumen + ethane) at ambient temperature; ○, Mehrotra and
Fig. 8. Comparison of calculated and measured bitumen densities; ▲,
Svrcek [29] data; ■, Nourozieh et al. [65] data; solid line, equation of state
Nourozieh et al. [65] data; solid line, equation of state using full character-
using full characterization scheme; dashed line, equation of state using lumped
ization scheme; dashed line, equation of state using lumped scheme.
scheme.

Zc RTc (0.2905 − 0.085ω) RTc

Pc =
vc
=
vc
∑ ∑ xi xk (vci1/3 + vck1/3 ) Tci Tcj 7. Results and discussions
i j

(34) As previously mentioned, the compositional analysis of Athabasca

bitumen sample was used to characterize the oil. The compositions and
where vc is the critical volume, Tc is the critical temperature, w is properties of components were applied in PR-EOS. In order to fit the
acentric factor, and Pc is the critical pressure. While the modelling re- experimental data, the measured composition data of VLE and LLE was
sults with full number of components might be acceptable, the number converted into k-values, and the exponent of binary interaction para-
of pseudo-components can be reduced for computational efficiency. In meter equation was adjusted to match the k-values. The best fitted
this study, the two-phase envelope (pressure-temperature diagram) is exponents (n in Eq. (12)) for un-lumped and lumped schemes are
generated with PR-EOS using full number of components. The defined 0.4964 and 0.5034, respectively. Using these values, the pressure-
pseudo-components are then lumped into fewer number of pseudo- composition (p − x) diagrams for mixtures of (Athabasca bi-
components using user defined method while the two-phase envelope is tumen + ethane) were generated at the temperature of 22 °C as shown
generated and then compared with full characterization scheme. The in Fig. 4. The VLE data are from Mehrotra and Svrcek [29] study and
minimum number of pseudo-components resulting in the same two- the LLE data from Nourozieh et al. [65] study. As depicted in the figure,
phase envelope as full characterization scheme is the optimum number the LLE was observed at overall ethane compositions greater than 18 wt
of components. As depicted in Fig. 3, the Athabasca bitumen sample % and at pressures larger than 4 MPa. Two schemes (lumped and un-
can be represented by four pseudo-components. The lumped pseudo- lumped) accurately predict the VLE and LLE regions using the equation
components properties are listed in Table 2. of state. The lumped scheme predicts the LLE separation in slightly
lower overall ethane composition compared to un-lumped scheme.

445
T. Sun, et al. Fuel 252 (2019) 439–447

1000 1000 Fig. 11. Effect of overall solvent com-

Co-existing Phase Density (kg/m3)

Co-existing Phase Density (kg/m3)

position on the co-existing phase den-

sity for (Athabasca bitumen + ethane)
800 800
mixtures at 5 MPa (a) and 9 MPa (b)
and ambient temperature; ▲, solvent-
600 600 enriched phase; ■, asphaltene-enriched
phase (experimental data from
Nourozieh et al. [65]); solid line,
400 400
equation of state using full character-
ization scheme; dashed line, equation of
200 200 state using lumped scheme.
(a) (b)
0 0
20 40 60 80 100 20 40 60 80 100
Overall Solvent Composition (wt%) Overall Solvent Composition (wt%)

Table 3 equation of state. Further investigation of results (Fig. 7) also shows

The deviations of models for the calculation of mixture densities. that the non-distilled fractions of extract and residue are in close
Model Data AARD (%) a
AAD (kg/m3) b
MAD (kg/m3) c agreement with the lab values. The increase in overall ethane con-
centration or pressure leads in higher extraction of light components
Un-lumped Raw Bitumen 0.34 3.3 6.2 from the bitumen, and this results into an increase in the fraction of
Equilibrium 0.43 3.8 11.7 heavier components in the bitumen-enriched phase (L2) [65].
Lumped Raw Bitumen 0.28 2.7 5.8 The proposed models (un-lumped and lumped schemes) have been
Equilibrium 0.45 4.0 8.9 examined for the solubility data over the temperature and pressure
a range of 20–200 °C and the pressure range of 0.1–10 MPa and they
AARD: average absolute relative deviation.
b provide AARDs of 16.3% and 16.7%, respectively.
AAD: average absolute deviation.
c
MAD: maximum absolute deviation.
The density and viscosity of bitumen is of importance for recovery
processes and their accurate predictions provide better field design and
Table 4
optimization. Although the density and viscosity of bitumen are usually
The deviations of models for the calculation of mixture viscosities. fitted and then predicted with available correlations, the gas-saturated
and solvent-diluted bitumen properties are not accurately predicted
Model Data AARD (%)a
with the correlations over the wide ranges of pressures, temperatures,
Un-lumped Bitumen 14.6 and solvent compositions. This is even worse when the composition of
Equilibrium Data 21.3 original fluid is changing with temperature, pressure, and/or solvent
Lumped Bitumen 23.5 composition due to precipitation or formation of a second liquid phase.
Equilibrium Data 23.3 Thus, it is desired to model the density and viscosity of fluid or equi-
librium phases while the compositional calculation is done. In WinProp,
a
AARD: average absolute relative deviation. the density and viscosity of equilibrium phases were simultaneously
reported with compositional calculations. Using the temperature-de-
Fig. 4 also demonstrates the solvent composition in the equilibrium pendent volume shift correlation, the density of raw bitumen is well
phases at different pressures and at an overall ethane concentration of predicted over the temperature range as shown in Fig. 8 using un-
40 wt%. As anticipated from the figure, the equilibrium compositions lumped and lumped schemes. This indicates that the developed char-
are reasonably predicted with two lumped and un-lumped schemes. acterization and the defined components reasonably represent the bi-
Although the predictions show slightly deviation from experimental tumen. The viscosity of bitumen was also correlated with modified
data at higher pressures, there is a good agreement between the mea- Pedersen’s correlation and the results are presented in Fig. 9. The
sured data and predicted ones. coefficients of viscosity correlation are d1 = 0.000156, d2 = 2.309757,
As it was explained in the study by Nourozieh et al. [65], the solvent d3 = 0.005902, d4 = 2.216300, and d5 = 0.442047. The data is accu-
overall composition affects the solvent composition in equilibrium rately modeled with this correlation over the studied temperature
phases due to the nature of extraction process as well as the solvent range.
affinity to extract light fractions/components. Thus, the equilibrium The density of equilibrium phases with dissolved solvent is also
composition of solvent in two phases is plotted as a function of solvent reasonably represented by equation of state in the VLE and LLE region
overall composition at two different pressures in Fig. 5. The satisfactory as shown in Fig. 10. The effect of solvent overall composition on the
predictions and similar results to experimental data were obtained for equilibrium densities were evaluated and compared with experimental
equilibrium compositions using two schemes. data. The results at two pressures 5 and 9 MPa are plotted in Fig. 11.
It is also worth to compare the component distribution in two liquid The density of light liquid phase in LLE region is slightly overestimated
phases excluding the solvent. This indicates how the solvent extracts at higher pressure (9 MPa). This indicates that the pressure dependency
the light hydrocarbon components into the lighter liquid phase, and of density might be not captured well in the model. Tables 3 and 4
partitions the heavy constitutes into the heavy liquid phase. Fig. 6 summarize the deviation of two models from experimental data.
shows the distribution of components in two flashed-off phases taken at
the equilibrium pressure of 9 MPa and a constant overall ethane com-
position of 40 wt%. The results indicate that the Peng-Robinson EOS 8. Conclusions
coupled with the developed characterization scheme provides a good
agreement not only in the solvent composition in different phases but The VLE and LLE data of (Athabasca bitumen + ethane) mixtures as
also in the component distribution in two phases. The same plot but at a well as the component distribution within each phase are well re-
constant pressure of 5 MPa is also shown in Fig. 6. Again, the compo- presented by PR-EOS and the proposed characterization scheme. The
nent distribution in extract and residue are well represented by the minimum number of components to represent the multi-phase

446
T. Sun, et al.
equilibrium was determined on the basis of two-phase envelope com-
parison. For the system under study, four components were found to be
enough to accurately capture the equilibrium phase compositions and
densities. The calibrated EOS with four components predicted multi-
phase equilibrium compositions, densities and viscosities within ac-
ceptable errors. The proposed models (un-lumped and lumped schemes)
have been examined for the solubility data over the temperature and
pressure range of 20–200 °C and the pressure range of 0.1–10 MPa and
they provided AARDs of 16.3% and 16.7%, respectively. The density
data were predicted with maximum AARD of 0.43%.
References
[1] Kumar A, Okuno R. Characterization of reservoir fluids using an EOS based on
perturbation from n-alkanes. Fluid Phase Equilibr 2013;358:250–71.
[2] Whitson CH. Characterizing hydrocarbon plus fractions. SPE J. 1983;23:683–94.
[3] Stamataki SK, Magoulas, KG, Characterization of Heavy Undefined Fractions, 2001
SPE International Symposium on Oilfield Chemistry, SPE 64996.
[4] Montel F, Gouel PL, A new lumping scheme of analytical data for compositional
studies. In: SPE Annual Technical Conference and Exhibition. Society of Petroleum
Engineers, 1984.
[5] Leibovici C, Stenby EH, Knudsen K. A consistent procedure for pseudo-component
delumping. Fluid Phase Equilibr 1996;117(1–2):225–32.
[6] Lomeland F, Harstad O. Simplifying the task of grouping fluid components in
compositional reservoir simulation. SPE Comp App 1995;7(2):38–43.
[7] Yarborough L. Application of the generalized equation of state to petroleum re-
servoir fluids, equation of state in engineering and research, in: K.C. Choa, R.L.
Robinson Jr. (Eds.). Adv Chem Ser Am Chem Soc 1978;182:386–439.
[8] Lawal AS, Van der Laan ET, Thambynayagam RKM, Four-parameter modification of
the Lawal-Lake-Silberberg equation of state for calculating gas-condensate phase
equilibria. In: SPE Annual Technical Conference and Exhibition. Society of
Petroleum Engineers, 1985.
[9] Pedersen KS, Thomassen P, Fredenslund A, AIChE Spring National Meeting New
Orleans. LA, March 1988, pp. 6–10.
[10] Guo TM, Du L, A New Three-Parameter Cubic Equation of State for Reservoir Fluids.
III. Application to Gas Condensates. SPE-19374-MS, unsolicited, 1989.
[11] Peneloux A, Jain C, Behar, E, Application of the Developed Redlich-Kwong
Equation of State to Predict the Thermodynamic Properties of Condensate Gases. In:
SPE Annual Technical Conference and Exhibition. Society of Petroleum Engineers,
Jan 1979.
[12] Whitson, CH, Torp SB, Evaluating constant volume depletion data. In: SPE Annual
Technical Conference and Exhibition. Society of Petroleum Engineers, Jan 1981.
[13] Zuo JY, Zhang D, Plus fraction characterization and PVT data regression for re-
servoir fluids near critical conditions. In: SPE Asia Pacific Oil and Gas Conference
and Exhibition. Society of Petroleum Engineers, 2000.
[14] Al-Meshari AA, McCain WD, An accurate set of correlations for calculating critical
properties and acentric factors for single carbon number groups. In: SPE Technical
Symposium of Saudi Arabia Section. Society of Petroleum Engineers, Jan 2006.
[15] Hosein R, Dawe R. Characterization of trinidad gas condensates. Pet Sci Technol
2011;29(8):817–23.
[16] Hosein R, Jagai T, Dawe RA. A method for predicting the phase behaviour of tri-
nidad gas condensates. W Indian J Eng 2013;35(2):22–30.
[17] Krejbjerg K, Pedersen KS, Controlling VLLE equilibrium with a cubic EOS in heavy
oil modeling. In: Canadian International Petroleum Conference. Petroleum Society
of Canada, Jan 2006.
[18] Ghasemi M, Alavian SA, Whitson, CH, SPE Heavy Oil Conference and Exhibition
held in Kuwait City. Kuwait, Dec 2011, pp. 12–14.
[19] Mehrotra AK, Sarkar M, Svrcek WY. Bitumen density and gas solubility predictions
using the Peng Robinson equation of state. AOSTRA J Res 1985;1(4):215–29.
[20] Bishnoi PR, Heidemann RA, Shah MK, Calculation of the thermodynamic properties
of bitumen system. In: The oil sands of Canada – Venezuela 1977, editors, David A.
Redford and Alvin G. Winestock, Proceedings of the Canada-Venezuela Oil Sands
Symposium 1977, Edmonton, Alberta, Canada, 248-254.
[21] Kesler MG, Lee BI. Improve prediction of enthalpy of fractions. Hydrocarb Process
1976;55(3):153–8.
[22] Fu C, Puttagunta R. Binary interaction coefficients of the modified soave-redlich-
kwong equation of state for bitumen-containing systems at in-situ conditions. J Can
Petrol Technol 1986;25(5):38–43.
[23] Fu CT, Puttagunta VR, Vilcsak G. Gas solubility of methane and ethane in Cold Lake
bitumen at in situ conditions. J Can Petrol Technol 1988;27(4):79–85.
[24] Mehrotra AK, Svrcek WY. Characterization of Athabasca bitumen for gas solubility
calculations. J Can Petrol Technol 1988;27(6):107–10.
[25] Mehrotra AK, Svrcek WY. Correlation and prediction of gas solubility in Cold Lake
bitumen. Can J Chem Eng 1988;66(4):666–70.
[26] Mehrotra AK, Patience GS, Svrcek WY. Calculation of gas solubility in Wabasca
bitumen. J Can Petrol Technol 1989;28(3):81–3.
[27] Frauenfeld TWJ, Kissel G, Zhou S, PVT and viscosity measurements for
Lloydminster-Aberfeldy and Cold Lake blended oil systems. Paper SPE 2002-110
presented at Canadian International Petroleum Conference, Calgary, Alberta,
Canada, 11-13 June 2002.
[28] Kariznovi M, Nourozieh H, Abedi J. Bitumen characterization and pseudocompo-
nents determination for equation of state modeling. Energ Fuel 2010;24(1):624–33.
[29] Mehrotra AK, Svrcek WY. Viscosity, density and gas solubility data for oil sand
bitumens. part I: Athabasca bitumen saturated with CO and C2H6. AOSTRA J. Res.
1985;1(4):263–8.
447
Fuel 252 (2019) 439–447
[30] Svrcek WY, Mehrotra AK. Gas solubility, viscosity and density measurements for
Athabasca bitumen. J Can Petrol Technol 1982;21(4):31–8. 387.
[31] Svrcek WY, Mehrotra AK. Properties of peace river bitumen saturated with field gas
mixtures. J Can Petrol Technolol 1989;28(2):50–6.
[32] Saber N, Shaw JM. On the phase behaviour of Athabasca vacuum residue+ n-de-
cane. Fluid Phase Equilibr 2011;302(1–2):254–9.
[33] Saber N, Zhang X, Zou XY, Shaw JM. Simulation of the phase behaviour of
Athabasca vacuum residue+ n-alkane mixtures. Fluid Phase Equilibr
2012;313:25–31.
[34] Diaz OC, Modaresghazani J, Satyro MA, Yarranton HW. Modeling the phase be-
havior of heavy oil and solvent mixtures. Fluid Phase Equilibr
2011;304(1–2):74–85.
[35] Agrawal P, Schoeggl FF, Satyro MA, Taylor SD, Yarranton HW. Measurement and
modeling of the phase behavior of solvent diluted bitumens. Fluid Phase Equilibr
2012;334:51–64.
[36] Li H, Yang D. Phase behaviour of C3H8/n-C4H10/heavy-oil systems at high pres-
sures and elevated temperatures. J Can Petrol Technol 2013;52(1):30–40.
[37] Badamchi-Zadeh A, Yarranton, HW, Maini BB, Phase behavior and physical prop-
erty modeling for vapex solvents: propane, carbon dioxide, and Athabasca bitumen.
Paper SPE 165505 presented at SPE Heavy Oil Conference, Calgary, Alberta,
Canada, 11–13 June 2013.
[38] Yang P, Li H, Yang D, Phase behaviour of CH4-C3H8-heavy oil systems under re-
servoir conditions. Paper SPE 165502 presented at SPE Heavy Oil Conference,
Calgary, Alberta, Canada, 11-13 June 2013.
[39] Ghasemi M, Whitson CH, Phase behavior and viscosity modeling of Athabasca bi-
tumen and light solvent mixtures. Paper SPE 165416 presented at SPE Heavy Oil
Conference, Calgary, Alberta, Canada, 11-13 June 2013.
[40] Twu CH. An internally consistent correlation for predicting the critical properties
and molecular weights of petroleum and coal-tar liquids. Fluid Phase Equilibr
1984;16(2):137–50.
[41] Gao J, Okuno R, Li HA. An experimental study of multiphase behavior for n-butane/
bitumen/water mixtures. SPE J 2017;22(03):783–98.
[42] Eghbali S, Dehghanpour H. An experimental and modeling study of carbon dioxide/
bitumen and C 4/bitumen phase behavior at elevated temperatures using cold lake
bitumen. SPE J 2018.
[43] Kumar A, Okuno R. Reliable characterization of bitumen based on perturbation
from n-alkanes for steam-solvent coinjection simulation. Fuel 2016;182:141–53.
[44] Azinfar B, Haddadnia A, Zirrahi M, Hassanzadeh H, Abedi J. Phase behaviour of
butane/bitumen fractions: experimental and modeling studies. Fuel
2018;220:47–59.
[45] Robinson DB, Peng DY, The Characterization of the Heptanes and Heavier Fractions
for the GPA Peng-Robinson Programs; Gas Processors Association, Research Report
RR-28: Tulsa, Oklahoma, USA, 1978.
[46] Mehra RK. The Computation of Multi-Phase Equilibrium in Compositional Reservoir
Studies. University of Calgary; 1981. Ph.D. Thesis.
[47] Li Y-K, Heavy Fraction Characterization and Hypothetical Component Selection for
Oil and Gas Mixtures, Computer Modelling Group Research Report R12.04, May
1983.
[48] Oellrich L, Ploecker U, Prausnitz JM, Knapp H. Equation-of-state methods for
computing phase equilibria and enthalpies. Int Chem Eng 1981;21(1):1–15.
[49] Katz DL, Firoozabadi A. Predicting phase behavior of condensate crude/oil systems
using methane interaction coefficients. J Petrol Technol 1978;30(11):1649–55.
[50] Peneloux A, Rauzy E, Freze R. A consistent correction for redlich-kwong-soave
volumes. Fluid Phase Equilib 1982;8(1):7–23.
[51] Pedersen KS, Milter J, Sørensen H. Cubic equations of state applied to ht/hp and
highly aromatic fluids. SPE J 2004;9:186–92.
[52] Kokal SL, Sayegh SG. Gas-saturated bitumen density predictions using the volume-
translated peng-robinson equation of state. J Can Petrol Technol 1990;29(05):31–9.
[53] Pedersen KS, Fredenslund A, Christensen PL, Thomassen P. Viscosity of crude oils.
Chem Eng Sci 1984;39(6):1011–6.
[54] Pedersen KS, Fredenslund A. An improved corresponding states model for the
prediction of oil and gas viscosities and thermal conductivities. Chem Eng Sci
1987;42(1):182–6.
[55] Nghiem LX, Li YK. Computation of multiphase equilibrium phenomena of reservoir
fluid. Fluid Phase Equilibr 1984;17(1):77–95.
[56] Nghiem LX, Heidemann RA, General acceleration procedure for multiphase flash
calculation with application to oil-gas-water systems. Paper presented at the 2nd
European Symposium on Enhanced Oil Recovery, Paris, France, 8-10 November
1982.
[57] Kariznovi M, Nourozieh H, Abedi J. Experimental determination of k-values and
compositional analysis of liquid phases in the liquid-liquid equilibrium study of
(Athabasca bitumen + ethane) systems. J Chem Eng Data 2013;58(6):1772–80.
[58] Nourozieh H. Phase Partitioning and Thermo-physical Properties of Athabasca
Bitumen/Solvent Mixtures PhD dissertation University of Calgary; 2013
[59] Riazi MR, Characterization and properties of petroleum fractions. 1st Edition,
Philadelphia, PA, 2005.
[60] Alboudwarej H, Akbarzadeh K, Beck J, Svrcek WY, Yarranton HW. Regular solution
model for asphaltene precipitation from bitumens and solvents. AIChE J
2003;49(11):2948–56.
[61] Riazi MR, Daubert TE. Prediction of the composition of petroleum fractions. Ind Eng
Chem Proc DD 1980;19(2):289–94.
[62] Whitson C. Characterizing hydrocarbon plus fractions. SPE J 1983;23(4):683–94.
[63] Sturges HA. The choice of a class interval. J Am Stat Assoc 1926;21(153):65–6.
[64] Lee BI, Kesler MG. A generalized thermodynamic correlation based on three-para-
meter corresponding states. AIChE J 1975;21(3):510–27.
[65] Nourozieh H, Kariznovi M, Abedi J. Physical properties and extraction measure-
ments for Athabasca bitumen + light hydrocarbon system: evaluation of pressure
effect, solvent-to-bitumen ratio, and solvent type. J Chem Eng Data
2011;56(11):4261–7.