Fuel 117 (2014) 206217

Contents lists available at ScienceDirect

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

Understanding the polydisperse behavior of asphaltenes during

precipitation
Mohammad Tavakkoli a,b,, Sai R. Panuganti a, Vahid Taghikhani b, Mahmoud Reza Pishvaie b,
Walter G. Chapman a
a
Department of Chemical and Biomolecular Engineering, Rice University, Houston, USA
b
Department of Chemical and Petroleum Engineering, Sharif University of Technology, Tehran, Iran

h i g h l i g h t s

 The PC-SAFT equation-of-state is used to study asphaltene phase behavior.

 The effect of asphaltene polydispersity on asphaltene phase behavior is studied.
 Results of monodisperse and polydisperse asphaltenes modeling are compared.
 A wide range of crude oils are considered for the study.
 An explanation for the observed behavior is provided based on FloryHuggins theory.

a r t i c l e i n f o a b s t r a c t

Article history: Asphaltenes are a polydisperse fraction of the crude oil, the phase behavior of which is signicantly
Received 28 June 2013 affected by the changes in pressure, temperature and composition. The focus of this study is to model
Received in revised form 16 September the polydisperse asphaltenes precipitation onset condition and the amount of precipitate from sol-
2013
vent-diluted crude oils using the Perturbed Chain form of the Statistical Associating Fluid Theory (PC-
Accepted 19 September 2013
Available online 5 October 2013
SAFT) over a wide range of crude oil density. Heavy oil and bitumen production can involve diluting with
parafnic solvents. Different fractions of the polydisperse asphaltenes thus precipitated are predicted and
when compared to the experimental data show a remarkable matching for different solvents. A compar-
Keywords:
Asphaltenes
ison of monodisperse and polydisperse modeling is also performed. This work illustrates the successful
Polydisperse application of PC-SAFT for predicting the phase behavior of polydisperse asphaltenes and in particular
Phase behavior from heavy oil and bitumen.
PC-SAFT 2013 Elsevier Ltd. All rights reserved.

1. Introduction Thermodynamic models to understand the asphaltenes precip-

itation fall under one of two molecular thermodynamic frame-
Asphaltenes, a polydisperse mixture of the heaviest and most works, mirroring the two prevalent schools of thought regarding
polarizable fraction of the crude oil; are dened according to their how asphaltenes are stabilized in crude oil. The rst approach as-
solubility properties as being soluble in aromatic solvents, but sumes that asphaltenes are solvated in crude oil and that these
insoluble in light parafn solvents. Asphaltenes are well-known asphaltenes will precipitate if the quality of the oil solvent drops
for their tendency to precipitate and deposit during oil production below a certain threshold. The FloryHuggins-regular-solution
because of changes in pressure, temperature, and composition. It is based models and the equation of state based models are some
desirable to prevent or mitigate asphaltene ow assurance prob- examples of this approach [13]. Thermodynamic models in the
lems, because remediation techniques are a signicant cost factor. second category take a colloidal approach to describe asphaltenes
Therefore, as a rst step it is necessary to predict the onset and behavior. In this approach resins stabilize the asphaltenes in the
amount of asphaltenes precipitation. crude oil [4,5]. The second approach was benecial in the early
stages of understanding asphaltene behavior. However, aspects
like addition of toluene (not a resin) which can redissolve asphalt-
Corresponding author at: Department of Chemical and Biomolecular Engineer-
enes made the asphaltene research community tend towards the
ing, Rice University, Houston, USA. Tel.: +1 713 348 4900.
solubility approach (rst approach).
E-mail addresses: mohammad.tavakkoli@rice.edu (M. Tavakkoli), sai@rice.edu
(S.R. Panuganti), taghikhani@sharif.edu (V. Taghikhani), pishvaie@sharif.edu (M.R. In this work, the rst approach is adopted and a SAFT based
Pishvaie), wgchap@rice.edu (W.G. Chapman). equation of state is used to model the asphaltenes behavior [6].

0016-2361/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.fuel.2013.09.069
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 207

Previous work by Ting et al. showed that the phase behavior of Table 1
asphaltenes is different when considered as monodisperse and PC-SAFT parameters correlations for asphaltenes.

polydisperse [7]. Ting et al. considered the effect of asphaltenes PC-SAFT Correlation
polydispersity for only model oil systems. Although Victorov and Parameter
Smirnova modeled asphaltenes polydispersity in real crude oil sys- M (1  c) (0.0223 Mw + 0.751) + c (0.0101 Mw + 1.7296)
tems, they could not obtain a good match with the presented data r (A) (1  c) (4.137738.1483/Mw) + c (4.616993.98/Mw)
[8]. Wang and Buckley presented experimental data showing the e/k (K) (1  c) (0.00436 Mw + 283.93) + c (508234100/
(Mw)^1.5)
effect of dilution ratio on different fractions of asphaltenes [9].
Good knowledge of the polydisperse behavior of the asphaltenes
is important because different fractions precipitate in different
amounts under different conditions. In this regard, limited model- components tendency to behave as a poly-nuclear-aromatic
ing work is available in the literature. Hence, this article deals with (c = 1) or as a benzene derivative (c = 0).
PC-SAFT modeling of real crude oil systems containing polydis- It should be mentioned that in most cases of polydisperse
perse asphaltenes. A detailed explanation on their behavior is also asphaltenes, the correlations in Table 1 give the right values for
presented based on the FloryHuggins theory. asphaltene SAFT parameters. But, in some cases like Lagrave oil
in Section 3.1, we need to treat the parameter r as a separate t-
ting parameter to nd a good match. One single value of r should
2. Modeling procedure be used for all asphaltenes sub-fractions and the correlation in Ta-
ble 1 gives us a good initial estimate for r. So in this case, we have
The focus of this work is to model the polydisperse asphaltenes one aromaticity and one r besides molecular weights of asphalt-
based on the experimental data for titration of dead oils with var- enes sub-fractions which need to be tted to reproduce asphaltene
ious n-alkanes. The PC-SAFT characterization methodology is de- instability experimental data. We will see later that even in the
scribed elsewhere in detail [10]. Hence, this section discusses case of using the correlation in Table 1 for calculating the parame-
only about the asphaltenes and the parameter estimation of this ter r, the obtained values for all asphaltenes sub-fractions are
polydisperse fraction. equal if one rounds them off to two decimals which is a result of
assuming the same aromaticity for each asphaltenes sub-fraction.
2.1. Asphaltenes and their sub-fractions
3. Results and discussion
For treating asphaltenes as polydisperse, asphaltenes are split
into sub-fractions based on the precipitating solvents used. For 3.1. Lagrave oil
example, when n-C10, n-C7 and n-C5 are used as the titrating
agents, asphaltenes are modeled as n-C10+ sub-fraction, n-C710 3.1.1. Asphaltenes polydispersity
sub-fraction, and n-C57 sub-fraction. Here, the n-C10+ asphaltene In the experiments performed by Wang, the asphaltenes for La-
sub-fraction represents n-decane insoluble asphaltenes. Asphalt- grave oil are rst separated into various solubility fractions using
enes from the combination of n-C10+ and n-C710 sub-fractions rep- excess n-pentane, n-heptane, and n-pentadecane precipitants
resent the n-heptane insoluble asphaltenes. Finally, the [14]. These sub-fractions are called n-C5 insoluble asphaltenes,
combination of n-C10+, n-C710 and n-C57 sub-fractions represent n-C7 insoluble asphaltenes, and n-C15 insoluble asphaltenes,
the n-pentane insoluble asphaltenes. respectively. The asphaltenes instability onsets for mixtures of
asphaltene, toluene, and n-alkanes (at ambient condition and an
2.2. Asphaltenes parameter estimation asphaltenes/toluene ratio of 1 g per 100 ml toluene) are then mea-
sured for each asphaltenes fraction.
Asphaltene molecules aggregate forming nano-aggregates even Within the PC-SAFT model, polydisperse asphaltenes are mod-
in good solvents such as toluene. Along with polarizability, polarity eled as three pseudo-components: n-C15+ sub-fraction, n-C715
of the asphaltenes may also play a role in forming the nano-aggre- sub-fraction, and n-C57 sub-fraction. The parameters of molecular
gates, but steric effects are thought to prevent further aggregation weight and aromaticity are t for the n-C15+ asphaltenes sub-
of asphaltenes in solution [11]. Precipitation is a phenomenon fraction to reproduce the experimental data on the minimum vol-
dominated by polarizability creating microdomains of asphaltene ume fraction precipitant needed to induce asphaltenes instability
rich phase [12]. Because of this, the association term in SAFT can uppt
v for mixtures of n-C15 insoluble asphaltenes, toluene, and var-
be safely neglected in asphaltenes modeling work. ious n-alkanes. It should be mentioned that the tting parameters
For every non-associating component, PC-SAFT has just three were optimized to match all the experimental data points for all n-
parameters, the temperature independent diameter of each molec- alkanes. The asphaltenes made from the combination of n-C15+ and
ular segment (r), the number of segments per molecule (m), and n-C715 sub-fractions represent the n-C7 insoluble asphaltenes. The
the segmentsegment dispersion energy (e/k). When considering molecular weight is t for the n-C715 sub-fraction to reproduce
a compound which is not well dened like asphaltenes, these three (together with the previously tted n-C15+ sub-fraction) the exper-
parameters as well as asphaltenes molecular weight are t based imental uppt
v data for a mixture of n-C7 insoluble asphaltenes, tolu-
on experimental data. But, when asphaltenes are considered as ene, and n-alkane. Finally, the molecular weight parameter of n-C5
polydisperse, the number of parameters increases in multiples of 7 sub-fraction is t so that the combination of the n-C15+ (previous
the number of sub-fractions. To reduce the number of parameters, t), n-C715 (previous t), and n-C57 sub-fractions represent the n-
the PC-SAFT parameter correlations for Aromatics + Resins pseudo- C5 insoluble asphaltenes and will reproduce the experimental uppt v
component presented by Gonzalez et al. (Table 1) are used in this data for a mixture of n-C5 insoluble asphaltenes, toluene, and n-
work to calculate the asphaltenes parameters [13]. The aromatic- alkanes.
ity parameter of asphaltenes is kept constant for all the asphaltene
sub-fractions of one oil (because the pseudo-components should 3.1.2. Parameters
have the same nature), and only the molecular weight of each In this section polydisperse asphaltenes are modeled as three
sub-fraction along with the aromaticity are t. The parameter of pseudo-components, hence four adjustable parameters are used,
aromaticity (c) used in these correlations determines the one aromaticity value and three molecular weights for each of
208 M. Tavakkoli et al. / Fuel 117 (2014) 206217

0.70 the carbon number of added n-alkane to the mixture of Lagrave

asphaltenes + toluene to nd the instability onset of the solution.
0.65
precipitant volume fraction

n-C15 asph. (Exp.)

0.60 n-C7 asph. (Exp.) 3.1.3. Comparison to previous results
0.55 n-C5 asph. (Exp.) As seen in the Fig. 1, the agreement between PC-SAFT calculated
n-C15 asph. (This work) and measured uppt v is remarkable. In the modeling results pre-
0.50 n-C7 asph. (This work) sented by Ting for the precipitation onsets with n-alkane extracted
n-C5 asph. (This work)
0.45 asphaltenes, SAFT did not accurately describe the experimental
n-C15 asph. (Ting, 2003)
0.40
uppt
v data in cases where larger n-alkanes (undecane and higher)
n-C7 asph. (Ting, 2003)
are used to induce asphaltenes precipitation [15]. Fig. 1 compares
n-C5 asph. (Ting, 2003)
0.35 the results obtained in this work and those predicted by Ting. The
0.30 modeling results for n-C15 asphaltenes, obtained by Ting, show a
5.00 7.00 9.00 11.00 13.00 15.00 better match in comparison to this work modeling results. But,
n-alkane carbon number for two cases of n-C5 and n-C7 asphaltenes, this work represents
a better match. The average error percentages for the modeling re-
Fig. 1. Comparison of the PC-SAFT modeling results from Ting and the current work sults presented by Ting are 2.53%, 4.54% and 1.83% for n-C5, n-C7
with respect to the experimental precipitant volume fraction at asphaltenes
instability onset for Lagrave asphaltenes-toluene-n-alkane mixtures (20 C and
and n-C15 asphaltenes, respectively. The error percentages for this
1 bar). Experimental data are from Wang [14]. work modeling results are 1.31%, 2.71% and 2.90% for n-C5, n-C7
and n-C15 asphaltenes, respectively. The average error percentage
Pn exp model
kx x k
is calculated using the equation: i1 i n i . The experimental
Table 2
data for n-C7 asphaltenes in Fig. 1 is different than those from n-
PC-SAFT parameters for various asphaltenes sub-fractions of Lagrave oil.
C5 and n-C15 asphaltenes. For the details of experimental procedure
Asphaltenes Asphaltenes Mw (g/ Wt. r M e/k (K) as well as the discussion on obtained experimental data, kindly re-
sub-fractions aromaticity (c) mol) % (A)
fer to Ref. [14].
C15+ 0.4 1475 45 4.00 26.84 375.76 It should be mentioned that although we have used less num-
C715 0.4 1200 30 4.00 22.05 374.44
bers of tting parameters, the modeling results show a better
C57 0.4 1150 25 4.00 21.18 374.17
match in comparison to modeling results obtained by Ting. Ting
used three asphaltenes sub-fractions and tted three molecular
weighs, three m parameters, three e/k parameters, and one r
parameter [15]. Also, he tted the amount of each asphaltenes
the asphaltenes sub-fractions. Since we are going to investigate sub-fraction to reproduce the experimental data on asphaltene
just the effect of asphaltene polydispersity and compare it to instability onset. But, in this study we tted one aromaticity, one
monodisperse asphaltenes, we kept other parameters affecting r parameter and three molecular weights for three asphaltenes
asphaltene phase behavior, the same for both cases of polydisperse sub-fractions. We also tted the amount of each asphaltenes sub-
and monodisperse asphaltenes. Also, to reduce the number of t- fraction to reproduce the experimental data on asphaltene instabil-
ted parameters, all binary interaction parameters between asphalt- ity onset. So, the improvement of this study over the Tings work is
enes and other species are set to zero. The binary interaction in using less parameters as well as nding a better modeling result
parameters, kijs, between all species are also set to 0 because SAFT with less error percentage in comparison to experimental data.
adequately describes the VLE behavior of these systems with zero
kij [15]. In next sub-sections we will discussed about the effect of 3.1.4. Maxima
kijs on model predictions for two real oil systems. One of the interesting points in Fig. 1 is the existence of a max-
Furthermore, in light of the parameter uniqueness issues dis- imum in the curve of the experimental precipitant volume fraction
cussed by Ting in his Ph.D. thesis (2003), the segment diameters at asphaltenes instability onset versus precipitant agent carbon
of all asphaltenes pseudo-components (r) are set to the same value number. Fig. 1 shows that the PC-SAFT predictions also present this
of 4.00 [15], while the other two parameters, m and e/k, are calcu- maximum. The onset of precipitation is increasing as the precipi-
lated using correlations presented in Table 1. Also, we previously tant agent carbon number is increasing up to around n-C9 (around
mentioned that for Lagrave oil we need to treat parameter r as a n-C11 for n-C7 asphaltenes experimental data). After n-C9, the pre-
separate tting parameter and we found the value of 4.00 for r will cipitant volume fraction at asphaltenes instability onset will de-
result in a good match between modeling and experimental re- crease by increasing n-alkane carbon number. This observation
sults. A comparison of the equation of state tted and the experi- on the trend of precipitation onset data is in line with Wiehe
mental uppt v data is shown in Fig. 1, with the tted PC-SAFT et al. who discussed the solubility parameter difference between
asphaltenes parameters listed in Table 2. As seen in Table 2, the n-parafn and asphaltenes in the oil and also the ratio of the molar
values of parameters e/k for different asphaltenes sub-fractions volume of asphaltenes to the molar volume of n-parafn [16].
are very close to each other which is a result of assuming the same Upon increasing the carbon number of n-alkane that is blended
aromaticity for each asphaltenes sub-fraction. Since we have used with oil, the solubility parameter difference between n-parafn
the experimental data for Lagrave asphaltenes, not Lagrave oil, the and asphaltenes in the oil becomes smaller and the ratio of the mo-
characterization data for Lagrave asphaltenes is presented in this lar volume of asphaltenes to the molar volume of n-parafn in the
study. For more information about Lagrave oil, kindly refer to ref- entropy of mixing becomes smaller. Both of these effects decrease
erence 14. the free energy of mixing and make the n-alkane and asphaltenes
For more clarication of Fig. 1, it should be mentioned that the more compatible. This is why, at small-carbon-number n-parafns,
experimental carbon numbers in the legend, i.e. n-C5, n-C7 and n- the volume of n-parafn at the occulation point increases as the
C15 asphaltenes, show three different types of asphaltenes, ex- n-parafn carbon number increases. However, as the carbon num-
tracted from Lagrave oil using three different n-alkanes, i.e. n-C5, ber of the n-parafn increases, the rate of increase of the solubility
n-C7 and n-C15, and used for preparation of Lagrave asphalt- parameter decreases, while the molar volume increases linearly.
enes + toluene mixtures. The carbon number on the x-axis shows Thus, the enthalpy of mixing increases as the molar volume of n-
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 209

10 0.70
mass fraction of each asphaltene sub-fraction

n-C15 asph. (Exp.) n-C7 asph. (Exp.) n-C5 asph. (Exp.)

9 0.65 n-C15 asph. ( =0.4) n-C7 asph. ( =0.4) n-C5 asph. ( =0.4)

precipitant volume fraction

n-C15 asph. ( =0.3) n-C7 asph. ( =0.3) n-C5 asph. ( =0.3)
8 0.60
n-C15 asph. ( =0.2) n-C7 asph. ( =0.2) n-C5 asph. ( =0.2)

7 0.55

6 0.50
(gr/100 ml oil)

n-C5 asph.
5 n-C7 asph. 0.45

4 n-C15 asph. 0.40

3 0.35

2 0.30
5.00 7.00 9.00 11.00 13.00 15.00
1 n-alkane carbon number
0
Exp. This work Ting, 2003 Fig. 3. Comparison of the PC-SAFT equation of state tted for different values of
asphaltenes aromaticity and the experimental precipitant volume fraction at
Fig. 2. Comparison of mass distribution of each asphaltenes sub-fraction in Lagrave asphaltenes instability onset for Lagrave asphaltenes-toluene-n-alkane mixtures
oil. Experimental data are from Wang [14]. (20 C and 1 bar). Experimental data are from Wang [14].

Table 3 imental data for total amount of precipitated asphaltenes using dif-
PC-SAFT parameters for various asphaltenes sub-fractions at different values of ferent n-alkanes.
asphaltenes aromaticity. The amount of high molecular weight asphaltenes (n-C15
Asphaltenes Asphaltenes Mw (g/ Wt. r M e/k (K) asphaltenes) could not be matched as observed from Fig. 2. This
sub-fractions aromaticity (c) mol) % (A) can be attributed to several factors. Experimentally, the measured
C15+ 0.4 1475 45 4.00 26.84 375.76 amount of asphaltenes precipitated may contain not only asphalt-
C715 0.4 1200 30 4.00 22.05 374.44 enes that should precipitate with the precipitant used, but also
C57 0.4 1150 25 4.00 21.18 374.17 small amounts of entrained resins, oil, and other asphaltenes frac-
C15+ 0.3 2280 45 4.00 43.54 357.46
C715 0.3 1500 30 4.00 29.00 354.52
tions. Variations in the amount of asphaltenes precipitated exist
C57 0.3 1450 25 4.00 28.07 354.30 even for the same oil using the same extraction procedure. On
C15+ 0.2 3800 45 4.00 76.41 341.80 the modeling side, because there are insufcient data to uniquely
C715 0.2 2200 30 4.00 44.64 335.96 t all of the model parameters for polydisperse asphaltenes, cer-
C57 0.2 2100 25 4.00 42.65 335.58
tain approximations have to be made. For instance, the segment
diameter of all asphaltenes sub-fractions is set to 4.00 Angstroms,
because we could not nd a good match using the correlation in
Table 1 for parameter r.
parafn increases and soon overcomes the effect of the decreasing
solubility parameter difference between n-parafn and asphalt-
3.1.6. Sensitivity analyses
enes in the oil. This causes the free energy of mixing to increase
To investigate the effect of asphaltenes aromaticity on phase
with increasing n-parafn carbon number at higher carbon num-
behavior, different sets of asphaltene aromaticity with molecular
bers, and results in decrease in the volume of n-parafn at the oc-
weight of the sub-fractions are t to experimental uppt v data shown
culation point, relative to increasing n-parafn carbon number. The
in Fig. 1. Table 3 shows the asphaltenes parameters tted to exper-
consequence is a maximum in the volume of n-parafn at the oc-
imental upptv data at different values of asphaltenes aromaticity.
culation point as one increase the carbon number of n-parafn
Fig. 3 shows a comparison of the equation of state tted with
blended with an oil-containing asphaltenes.
different values for asphaltenes aromaticity and the experimental
uppt
v data. A reasonable match is obtained when aromaticity value
is in the range of 0.20.4 for Lagrave asphaltenes. From Table 3, one
3.1.5. Mass distribution may conclude that there is wide range for asphaltene SAFT param-
A comparison of the PC-SAFT predicted and measured mass dis- eters m and e/k. Here, it is necessary to say that the issue of unique-
tribution of the asphaltenes sub-fractions is shown in Fig. 2. For the ness of the asphaltene SAFT parameters has already been
column labeled experimental, the mass of each asphaltenes sub- investigated by Ting [15]. Different values for asphaltene molecular
fraction is inferred from the difference in the measured amount of weight and aromaticity result in various asphaltene SAFT parame-
asphaltenes precipitated with different precipitants. Wang used ters. So, it is possible to nd reasonable results using different sets
excess amount of n-pentane, n-heptane, and n-pentadecane pre- of tting parameters as shown by Ting (2003). The phase behavior
cipitants to separate asphaltenes from Lagrave oil and measure calculated using these different sets of asphaltene parameters are
the amount of precipitated asphaltenes [14]. Within the PC-SAFT similar and we believe the effects of the nonuniqueness of asphalt-
model, the amount of each asphaltenes sub-fraction is optimized enes SAFT parameters on asphaltene phase behavior in oil should
to t the experimental upptv data in Fig. 1. It is interesting to note be small [15]. For the best t, one needs to choose the set of tting
that the modeled mass distribution in this work is closer to exper- parameters which shows the minimum error percentage between
imentally measured values in comparison to those predicted by modeling results and experimental data. For Lagrave asphaltenes,
Ting [15]. For Lagrave oil, it is necessary to say that we could not the best t happens when the value of 0.4 is selected for asphaltene
nd a good match in the case of using the experimental mass of aromaticity.
each asphaltenes sub-fraction and the reason will be discussed in When aromaticity is decreased, higher molecular weights of
the next paragraph. But, for the other oils in this study, asphaltenes asphaltenes sub-fractions are needed to obtain a reasonable match
will be divided into a number of sub-fractions based on the exper- with the experimental data. Also, it should be mentioned that in all
210 M. Tavakkoli et al. / Fuel 117 (2014) 206217

cases the weight percent of asphaltenes fractions is: 45%, 30% and that when higher asphaltenes aromaticity is used, a lower value for
25% for C15+, C715 and C57 sub-fractions, respectively. asphaltenes molecular weight is needed to obtain a reasonable
In the above cases with different values of asphaltenes aroma- match with experimental onset data. So, here the effects of both
ticity and molecular weight, asphaltenes PC-SAFT parameters are asphaltenes aromaticity and asphaltenes molecular weight are
checked to be in the acceptable range (between pure saturates important. In the following paragraphs the contributions of
and pure poly-nuclear-aromatics). Also, in each case the asphalt- asphaltenes aromaticity and molecular weight in the predicted
enes solubility parameter calculated using the PC-SAFT equation precipitation amount is discussed. It is necessary to say that if
of state is found to be in the range 19 and 24 MPa0.5 reported in one increases asphaltene aromaticity while the value of asphaltene
the literature [15]. The asphaltene molecular weight is still under molecular weight is constant, the modeling results will show more
debate in the literature [17]. Asphaltenes selfassociation along amounts of precipitated asphaltenes as well as a lower dilution ra-
with asphaltenes polydispersity are the reasons why there is no tio for asphaltene precipitation onset. The same result can be seen
range in the reported molecular weight and structural properties for increasing asphaltene molecular weight while the value of
of asphaltenes. asphaltene aromaticity is constant.

3.1.7. Molar volume and solubility parameter

3.1.6.1. Asphaltene aromaticity. To see the effect of asphaltenes aro-
Table 4 presents the values of asphaltenes solubility parameter
maticity on the asphaltene precipitated amount, model oil system
and molar volume, calculated by PC-SAFT, for C15+ asphaltenes pre-
1 is dened consisting of monodisperse Lagrave n-C15+ asphalt-
sented in Table 3. Table 4 shows that the asphaltenes solubility
enes, the PC-SAFT parameters of which are t to the experimental
parameter increases with aromaticity and the molar volume in-
uppt
v using various values for asphaltene aromaticity. In this model
creases as molecular weight increases. To understand the contribu-
system 1, toluene is used as the model oil and the asphaltenes/tol-
tions of both asphaltenes molar volume and solubility parameter
uene ratio while modeling is taken the same as the experiment by
in the amount of precipitation, the FloryHuggins theory is used.
Wang, 1 g/100 ml [14]. Standard conditions of 20 C and 1 bar are
It is necessary to say that asphaltene aromaticity and molecular
assumed. Fig. 4 shows the amount of precipitated asphaltenes from
weight control the values of asphaltene SAFT parameters. These
model system 1 predicted by PC-SAFT using the parameters in the
SAFT parameters can be used for calculating asphaltenes solubility
Table 3. N-pentadecane is used as the precipitating agent. It can be
parameter and molar volume by PC-SAFT equation of state. So, we
found from Fig. 4 that lower values for asphaltenes aromaticity will
are relating asphaltene aromaticity and molecular weight to
result in more asphaltenes being precipitated. Also, asphaltenes re-
asphaltene solubility parameter and molar volume and one may
dissolution begins at lower precipitant volume fractions and with a
nd the effect of aromaticity and molecular weight by investigat-
higher re-dissolution amount when higher values for asphaltenes
ing the effect of solubility parameter and molar volume. The Flo-
aromaticity and therefore lower values for molecular weight are
ryHuggins theory may give us a good insight into the effect of
used. To understand better the modeling results presented in
solubility parameter and molar volume.
Fig. 4, precipitant dilution ratio is used in Fig. 5 instead of precipi-
For any two phases at equilibrium, the fugacities of each com-
tant volume fraction. Dilution ratio is based on the volume of pre-
ponent in existing phases are equal. For liquidliquid equilibrium,
cipitating agent in cc per 1 g of original oil. It should be mentioned
the equilibrium ratio, Ki can be written as
that the format of Fig. 5 will be used for many other gures in the
  L
present paper because in many cases we need to compare the xHi vi vL vL 2
modeling results at high amount of added precipitant agent. Also, K HL
i exp ln  i i dL  dL
xLi v Lm v Lm RT i m
this format is the common format used in both industrial and aca-  H 
vi v Hi v Hi H H 2
demic applications [1,18].  ln  d  d 1
Higher asphaltenes aromaticity means that the nature of
v Hm v Hm RT i m

asphaltenes in the system is closer to poly-nuclear-aromatics and where i is the component, L and H are the asphaltenes lean and
as a result they are more incompatible with the maltenes in the asphaltenes rich phases, respectively, x is the mole fraction, v is
crude oil. We would expect that higher asphaltenes aromaticity the molar volume, R is the universal gas constant, T is the absolute
would result in more asphaltene precipitating which is not in line temperature, the subscript m stands for the mixture, and d is the
with the modeling results presented in Figs. 4 and 5. The reason is

Model oil system 1, precipitant = n-C15

Model oil system 1, precipitant = n-C15 1.2
1
Asph. aromaticity ( ) = 0.2
0.9 1
Asph. aromaticity ( ) = 0.3
asphaltene mass fraction
asphaltene mass fraction

0.8 Asph. aromaticity ( ) = 0.4

0.7 0.8

0.6
0.6
0.5
0.4 Asph. aromaticity ( ) = 0.2
0.4
0.3
Asph. aromaticity ( ) = 0.3
0.2 0.2
0.1
Asph. aromaticity ( ) = 0.4

0 0
0 0.2 0.4 0.6 0.8 1 0 20 40 60 80 100
precipitant volume fraction dilution ratio, cc/gr

Fig. 4. Precipitated asphaltenes mass fraction versus n-pentadecane volume Fig. 5. Precipitated asphaltenes mass fraction versus n-pentadecane dilution ratio,
fraction at different values of asphaltenes aromaticity for the model oil system 1. cc/g, at different values of asphaltenes aromaticity for the model oil system 1.
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 211

Table 4
Values of solubility parameter and molar volume for C15+ asphaltenes sub-fraction at different values of asphaltenes aromaticity calculated by PC-SAFT.

Asphaltenes Asphaltenes aromaticity (c) Molecular weight (g/mol) Molar volume (cm3/mol) Solubility parameter (MPa0.5)
C15+ 0.4 1475 1069.06 23.29
C15+ 0.3 2280 1747.48 22.49
C15+ 0.2 3800 3091.05 21.76

solubility parameter. The general derivation of Eq. (1) can be found =0.4, Mw=1475 g/mol
14
elsewhere [2]. =0.3, Mw=2280 g/mol
Assuming that only asphaltenes partition to asphaltenes rich =0.2, Mw=3800 g/mol
12

molar volume ratio difference

phase, the above equation will result in the same equation pro-

(light phase -heavy phase)

posed by Hirschberg et al. for modeling asphaltenes precipitation 10
by combining the FloryHuggins theory for polymer solution and
Hildebrand solubility concept [1]. But, this assumption is not accu- 8
rate for asphaltenes thermodynamic modeling using PC-SAFT, be-
cause although asphaltenes rich phase consist mainly 6
asphaltenes, there are considerable amount of saturates as well
as resins and aromatics in this phase [19]. Rearrangement of Eq. 4
(1) results in
        2
xH v Li v Hi v Li v Hi
K HL iL exp ln  ln   vL  vH
i
xi v Lm v Hm v Lm v Hm 0
0 100 200 300 400
2
dilution ratio, cc/gr
where v is the FloryHuggins interaction parameter
Fig. 6. Molar volume ratio difference between light and heavy phases calculated
vi 2 using PC-SAFT versus n-pentadecane dilution ratio at different values of asphalt-
v di  dm 3 enes aromaticity and molecular weight.
RT
Based on Eq. (2), when vL  vH , increases the amount of pre-
 vL vH 
cipitation will increase, while v Li  v Hi has a reverse effect. Figs. 6
Flory-Huggins interaction parameter difference

m m

and 7 represent the PC-SAFT calculated values for two expressions 35

 vL vH 
v L  v H and v  v , respectively, for the three cases presented
i i L H
m m 30
(light phase-heavy phase)

in Table 4.
From Figs. 6 and 7 it can be concluded that for the case of c = 0.2 25
and Mw = 3800, both molar volume ratio difference and value of
20
FloryHuggins interaction parameter difference are more for this
case in which the effect of FloryHuggins interaction parameter 15
difference is higher than the molar volume ratio difference and
=0.4, Mw=1475 g/mol
would result in more asphaltenes precipitation for the case of 10
=0.3, Mw=2280 g/mol
c = 0.2 and Mw = 3800. In other words, although asphaltenes solu-
bility parameter has a lower value for c = 0.2 and Mw = 3800, the 5 =0.2, Mw=3800 g/mol

molar volume for this case is more than two other cases which re-
0
sults in a higher value for the FloryHuggins interaction parameter 0 100 200 300 400
and this case shows less asphaltenes re-dissolution by increasing dilution ratio, cc/gr
precipitant volume fraction.
Fig. 7. FloryHuggins interaction parameter difference between light and heavy
phases calculated using PC-SAFT versus n-pentadecane dilution ratio at different
3.1.8. Monodisperse vs polydisperse
values of asphaltenes aromaticity and molecular weight.
To investigate the roles of polydispersity on asphaltenes phase
behavior, the solubility behavior of two model oil mixtures with
monodisperse and polydisperse asphaltenes are compared. In and C57) and the PC-SAFT parameters for each asphaltenes sub-
fraction are t to the experimental uppt
v data of the fractionated
these model systems 2 and 3, toluene is used as the model oil (with
different amounts of n-alkane precipitants added) and the asphalt- asphaltenes. See Table 2 for asphaltenes sub-fractions parameters.
All binary interaction parameters between all species used in the
enes parameters are t to experimental uppt v data for Lagrave
asphaltenes. The asphaltenes/toluene ratio while modeling is taken PC-SAFT calculations in this section are set to zero.
The effects of n-alkane addition on the amount of asphaltenes
the same as the experiment by Wang (2000), 1 g/100 ml. Standard
conditions of 20 C and 1 bar are assumed for both the model sys- precipitated at 20 C and 1 bar for the two model oils are shown
in Figs. 8 and 9. The y-axis shows the precipitated asphaltene
tems. Model system 2 is a monodisperse system consisting of La-
grave n-C5 asphaltenes, the PC-SAFT parameters of which are t amount as a fraction of total asphaltene content of the oil. Since
same asphaltenes have been used in both model systems and the
to the experimental uppt
v for n-C5 asphaltenes in Fig. 1. The aroma-
ticity and molecular weight of these monodisperse n-C5 asphalt- PC-SAFT parameters of asphaltenes are t to the experimental
enes are 0.4 and 1375, respectively. Also, the value of 4.00 was
uppt
v data, we can see the same asphaltene precipitation onset in
used for parameter r. In model system 3, polydisperse asphaltenes both cases of monodisperse and polydisperse asphaltenes. How-
ever, a change in the amount of precipitated asphaltenes vs. preci-
with 3 subfractions: C57, C715 and C15+ are used. The weight
percent of asphaltenes fractions is: 45%, 30% and 25% (C15+, C715 pitant volume fraction can be seen for the case of polydisperse
212 M. Tavakkoli et al. / Fuel 117 (2014) 206217

precipitant = n-C5 Model oil system 3, precipitant = n-C5

1 1

Model oil system 2 Total asphaltene precipitation

asphaltene mass fraction

0.8 0.8

asphaltene mass fraction

Model oil system 3 C15+ asphaltene precipitation
C7-15 asphaltene precipitation
0.6 0.6 C5-7 asphaltene precipitation

0.4
0.4

0.2
0.2

0
0 0.2 0.4 0.6 0.8 1 0
0 0.2 0.4 0.6 0.8 1
precipitant volume fraction
precipitant volume fraction
Fig. 8. Asphaltenes mass fraction versus n-pentane volume fraction for model oil
Fig. 10. Distribution of the asphaltenes sub-fractions in the precipitated phase as a
system 2 (monodisperse asphaltenes) and model oil system 3 (polydisperse
function of n-pentane volume fraction in the model oil system 3 (polydisperse
asphaltenes) at 20 C and 1 bar. Model oil contains 1 gram of Lagrave oil
asphaltenes).
asphaltenes in 100 ml toluene.

3.1.9. Fractions of individual precipitation

precipitant = n-C15 To see the role of each asphaltenes sub-fraction in the total
1 amount of precipitated asphaltenes, plots of the mass distribution
of polydisperse asphaltenes sub-fractions as a function of precipi-
Model oil system 2
0.8 tant volume fraction are shown in Figs. 10 and 11. From the Figs. 10
asphaltene mass fraction

Model oil system 3 and 11 it is observed that near the initial asphaltenes instability
onset, the precipitated phase is composed mostly of the heaviest
0.6 asphaltenes fractions (n-C15+ sub-fraction). As the amount of preci-
pitant is increased further, lower molecular weight asphaltenes
will also precipitate. From Figs. 10 and 11, one can also see that
0.4
the amount of re-dissolution is more for higher carbon number
precipitant agents following the explanation earlier in this section.
0.2 Unfortunately there is no experimental data on the precipitated
amount of each asphaltenes sub-fraction in the literature to com-
pare with modeling results obtained by PC-SAFT. Part of the reason
0
is the difculties in extracting asphaltenes sub-fractions from
0 0.2 0.4 0.6 0.8 1
crude oils. Hence, the results presented here for the role of each
precipitant volume fraction
asphaltenes sub-fraction are theoretical predictions.
Fig. 9. Asphaltenes mass fraction versus n-pentadecane volume fraction for model
oil system 2 (monodisperse asphaltenes) and model oil system 3 (polydisperse
asphaltenes) at 20 C and 1 bar. Model oil contains 1 gram of Lagrave oil
asphaltenes in 100 ml toluene. 3.2. Iranian oil

The model results discussed thus far are based on the asphalt-
asphaltenes. By treating asphaltenes as a polydisperse species, the
enes data for Lagrave oil. As further tests/proof of general applica-
amount of asphaltenes precipitated increases more gradually with
bility of the model, we will compare with experimental data for
precipitant addition. Also, the total amount of precipitated asphalt-
several other oils including both light and heavy crudes.
enes in polydisperse system is less than the total precipitated
Titration experimental results presented by Hirschberg et al.
amount in monodisperse system. In a polydisperse system, some
(1984) are one of the well-known asphaltene studies available in
amount of lower molecular weight asphaltenes (C57 and C715) ex-
the literature [1]. The oil used is an Iranian crude oil with an n-hep-
ists in the mixture, and a signicant amount of asphaltenes will
tane asphaltenes content of 1.9 weight percentage in the stock tank
stay in solution even at high precipitant volume fractions. Thus,
oil. Table 5 presents PVT properties of this crude. Table 6 shows the
more asphaltenes can be precipitated using lower molecular
results of titration experiments performed on stock tank oil pre-
weight n-alkanes.
sented in Table 5.
Asphaltene redissolution at high dilution ratios is a phenome-
For modeling of Iranian oil titration experiments, asphaltenes
non reported in literature [9]. It can be seen from Figs. 8 and 9 that
are divided into three sub-fractions: n-C57, n-C710 and n-C10+
asphaltenes re-dissolution begins at lower precipitant volume frac-
asphaltenes, based on the experimental amount of n-C5, n-C7 and
tions and with a higher re-dissolution amount for polydisperse
n-C10 asphaltenes. These experimental amounts were measured
model system (system 3). Existence of low molecular weight
by adding excess amount of n-alkane to the tank oil and then lter-
asphaltenes sub-fractions (C57 and C715) which can dissolve in
ing the mixture with 0.45 micrometer lter paper and nally
the system better than high molecular weight asphaltenes (C15+)
washing the lter paper with hot n-heptane [1]. Accurate amount
is the reason for this re-dissolution. Also, the amount of re-dissolu-
of precipitation prediction was not possible using only one asphal-
tion is more for higher carbon number precipitant agents, because
tene fraction, i.e. n-C5 asphaltenes for all the titration experiments
they are better solvents for asphaltenes than lower carbon number
presented in Table 6. Hence, using polydisperse asphaltenes is the
n-alkanes.
rst key point for modeling of these titration experiments.
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 213

Model oil system 3, precipitant = n-C15 Table 7 presents the characterization results rounded off to two
1 decimals for the Iranian oil using the PC-SAFT methodology. Being
dead oil, there is no gas component in the system. The aromaticity
Total asphaltene precipitation
value for Aromatics + Resins component is 0.38 and is tted to the
asphaltene mass fraction

0.8 C15+ asphaltene precipitation experimental stock tank oil density. The experimental value for
C7-15 asphaltene precipitation stock tank oil density, 0.873 g/cc and the tted value, 0.872 g/cc
0.6 C5-7 asphaltene precipitation are well matched. The aromaticity value for asphaltenes as well
as molecular weights of different asphaltenes fractions are tted
to the experimental data for both onset and amount of precipita-
0.4
tion. A comparison of the equation of state tted and the experi-
mental data for onset and amount of precipitation are shown in
0.2 Figs. 12 and 13, respectively. The x-axis in Fig. 12 shows the carbon
number for the added precipitant agent.
Titration experiments presented by Hirschberg et al. contains
0
0 0.2 0.4 0.6 0.8 1
asphaltenes onset of precipitation data with four different n-al-
kanes as well as precipitation amount for three different titrating
precipitant volume fraction
agents. The PC-SAFT tted parameters for all these experimental
Fig. 11. Distribution of the asphaltenes sub-fractions in the precipitated phase as a results are: molecular weights of the three asphaltenes sub-frac-
function of n-pentadecane volume fraction in the model oil system 3 (polydisperse tions and the aromaticity value for asphaltenes component. These
asphaltenes). four parameters are t to match simultaneously all experimental
data points for onset and amount of precipitation using different
titration agents. The only parameter that can help us to handle
Table 5 each titration experiment individually is the binary interaction
Iranian oil composition and PVT properties [1]. parameter, kij, between asphaltene component and n-alkane. The
Component Composition (mol%) kijs for dead oil, between Saturates, Aromatics + Resins, Asphalt-
enes and different n-alkanes, presented by Gonzalez are used in
Methane 0.10
Ethane 0.48
this work [20]. The kijs between asphaltenes and other compo-
Propane 2.05 nents are equal to the kijs between Aromatics + Resins and the
i-Butane 0.88 other components, as presented by Gonzalez. By slightly changing
n-Butane 3.16 the kij between each n-alkane and asphaltenes component, one can
i-Pentane 1.93
nd a better match for both onset and amount of precipitation. The
n-Pentane 2.58
Hexane 4.32 kij between asphaltenes and saturates controls the onset of precip-
Heptane plus 84.5 itation as well as the shape of the curve right after the precipitation
Saturates (mol%) 50 onset. If one increases the value for the kij between asphaltenes
Aromatics (mol%) 25 and saturates, the modeling results will show a higher slope for
n-C7 asphaltenes (wt%) 1.9
n-C5 asphaltenes (wt%) 3.9
the yield curve, right after the onset of precipitation. Also, the mod-
Asphaltenes + resins (wt%) 18 eling results will demonstrate a lower dilution ratio of the precipi-
Molecular weight 221.5 tant agent for the asphaltene precipitation onset. Thus the kij
Density (g/cc) 0.873 between asphaltenes and saturates is also a matching parameter
besides previously mentioned tting parameters, molecular
weights of three asphaltenes sub-fractions and aromaticity value
for asphaltenes component. Table 8 presents the binary interaction
parameters between all components used for modeling of the Hir-
Table 6 schberg titration experiments.
Iranian stock tank oil titration results [1].

Dilution Ratio (cc diluent/g tank oil) n- n-C7 n-C10 n-C12 n-C16 3.3. Heavy oil modeling
C5
1.0 Onset
In this section, the procedure used for predicting the instability
1.3 Onset b onset data and for modeling the titration experiments of model
1.4 Onset Onset b and light oils is used for modeling of titration experiments for hea-
5 a 1.53 1.34 vy crudes and bitumens available in the literature. The titration
10 3.61 1.82 1.45
experimental results for heavy oils from different sources, Canada,
20 3.79 1.89 1.5
50 3.87 1.87 Venezuela, Russia, and Indonesia, presented by Sabbagh et al. are
a
used in this section [21]. Table 9 demonstrates the SARA analysis
Onset of occulation not determined.
b of these bitumens and heavy oils and measured molar mass and
Amount of precipitation not determined.
density of each fraction.

Table 7
Characterized Iranian stock tank oil as a combination of ve pseudo-components and their PC-SAFT parameters.

Component Mw (g/mol) Mass fraction r (A) M e/k (K)

Saturates 207.43 0.468 3.92 6.17 259.32
Aromatics + Resins (c = 0.38) 219.44 0.493 4.05 4.99 342.30
C10+ Asphaltenes (c = 0.2) 4307.22 0.015 4.22 86.49 343.60
C710 Asphaltenes (c = 0.2) 2950.00 0.004 4.22 59.53 338.74
C57 Asphaltenes (c = 0.2) 2900.00 0.020 4.22 58.54 338.56
214 M. Tavakkoli et al. / Fuel 117 (2014) 206217

Experimental data (Hirschberg et al., 1984) Table 9

1.8 SARA analysis of bitumens and heavy oils from Venezuela, Canada, Russia and
PC-SAFT modeling results
Indonesia, and the measured molar mass and density of each fraction.
1.6
1.4 Heavy oil fraction Content (wt%) Density (kg/m3) Molar mass (g/mol)
dilution ratio, cc/gr

1.2 Venezuela
Saturates 20.5 882 400
1
Aromatics 38.0 997 508
0.8 Resins 19.6 1052 1090
Asphaltenes 21.8 1193 7662
0.6
Solids 0.1
0.4
Canada
0.2 Saturates 23.1 876 482
0 Aromatics 41.7 997 537
4 5 6 7 8 9 10 11 12 13 14 15 16 17 Resins 19.5 1039 859
Asphaltenes 15.3 1181 6660
n-alkane carbon number
Solids 0.4

Fig. 12. Comparison of the PC-SAFT equation of state tted and the asphaltenes Russia
instability onset experimental data for Iranian stock tank oil. Saturates 25.0 853 361
Aromatics 31.1 972 450
Resins 37.1 1066 1108
Asphaltenes 6.8 1192 7065
Titration agent=n-C5 (Exp.) Titration agent=n-C7 (Exp.) Solids 0.0
Titration agent=n-C10 (Exp.) Titration agent=n-C5 (Model) Indonesia
4.5 Titration agent=n-C7 (Model) Titration agent=n-C10 (Model) Saturates 23.2 877 498
precipitated asphaltene, wt%

4 Aromatics 33.9 960 544

Resins 38.2 1007 1070
3.5 Asphaltenes 4.7 1132 4635
3 Solids 0.0

2.5
2
1.5 SAFT parameters of aromatics and resins, using a single aromaticity
1
value for both of them. For asphaltenes fraction, the measured
molecular weights are not used for modeling purpose and are trea-
0.5
ted as a matching parameter for each sub-fraction. When treating
0 asphaltenes as polydisperse, we need the molecular weights for
0 10 20 30 40 50 60
each asphaltenes sub-fraction which are not available in Table 9.
dilution ratio, cc/gr Table 10 presents the parameters used for modeling of heavy
Fig. 13. Comparison of the PC-SAFT equation of state tted and the experimental
oils and bitumens in Table 9. For each crude oil, the aromaticity va-
amount of precipitated asphaltenes for Iranian stock tank oil. Experimental data lue for aromatics and resins components is tted to the experimen-
from Hirschberg et al. [1]. tal stock tank oil density. The aromaticity values for asphaltenes as
well as molecular weights of different asphaltenes fractions are t-
For modeling of heavy oils in Table 9, asphaltenes fraction in the ted to the experimental data for amount of precipitation. Table 11
characterization process is divided into three sub-fractions for the shows the asphaltenes PC-SAFT parameters to the nearest second
oil from Canada and two sub-fractions for the other three oils, decimal.
based on the experimental data for total amount of precipitated Table 12 compares experimental values and modeling results
asphaltenes using different n-alkanes. These experimental for the stock tank oil density showing a good agreement (maxi-
amounts were measured by addition of excess n-alkane to the hea- mum error is less than 2%). A comparison of the equation of state
vy crude [21]. tted and the experimental data for heavy oils and bitumens from
The molecular weights for saturates, aromatics and resins of all Venezuela, Canada, Russia and Indonesia are shown in Figs. 1417
four oils are available, and there is no need to calculate them dur- respectively. In each case there is a good agreement for the amount
ing the characterization process. The composition data of the crude of asphaltenes precipitated upon dilution with different
oils are not available; hence the aromatics and resins fractions precipitants.
could not be combined and are left as individual pseudo-compo- From Tables 10 and 12, one can understand that both asphalt-
nents. The correlations in Table 1 are used for predicting the PC- enes aromaticity, and aromatics and resins aromaticity decrease

Table 8
Binary interaction parameters used for modeling of Iranian stock tank oil.

Component n-C5 n-C7 n-C10 n-C12 n-C16 Saturates A+R C57 Asph. C710 Asph. C10+ asph.
n-C5 0
n-C7 0 0
n-C10 0 0 0
n-C12 0 0 0 0
n-C16 0 0 0 0 0
Saturates 0 0 0 0 0 0
A+R 0.007 0.0065 0.006 0.006 0.005 0.007 0
C57 Asph. 0.008 0.0065 0.006 0.005 0.005 0.003 0 0
C710 Asph. 0.008 0.0065 0.006 0.005 0.005 0.003 0 0 0
C10+ Asph. 0.008 0.0065 0.006 0.005 0.005 0.003 0 0 0 0
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 215

Table 10
Aromaticity values for aromatics, resins and asphaltenes pseudo-components as well as various asphaltenes sub-fractions molecular weight used for modeling of bitumens and
heavy oils presented in Table 9.

Heavy oil Aromatics and resins Asphaltenes C57 Asph. Mw (g/ C56 Asph. Mw (g/ C67 Asph. Mw (g/ C7+ Asph. Mw (g/
source aromaticity aromaticity mol) mol) mol) mol)
Venezuela 0.3 0.4 2500 2600
Canada 0.25 0.3 3600 3650 3700
Russia 0.23 0.2 4700 5600
Indonesia 0.22 0.2 4100 4300

Table 11 Table 12
Asphaltene PC-SAFT parameters obtained by using the aromaticity and molecular Comparison of experimental stock tank oil density and modeled density using PC-
weight from Table 10 in the equations presented in Table 1. SAFT. Experimental values are from Sabbagh et al. [21].

Heavy oil source Asphaltene fraction r (A) m e/k (K) Heavy oil Experimental stock tank oil Modeled stock tank oil
source density (g/cc) density (g/cc)
Venezuela n-C57 4.30 44.69 379.34
n-C7+ 4.30 46.43 379.65 Venezuela 1.016 0.998
Canada 0.996 0.985
Canada n-C56 4.26 68.14 361.81 Russia 0.982 0.970
n-C67 4.26 69.08 361.97 Indonesia 0.963 0.962
n-C7+ 4.26 70.01 362.13
Russia n-C57 4.22 94.28 344.99
n-C7+ 4.22 112.16 348.16
Indonesia n-C57 4.22 92.30 344.63
n-C7+ 4.22 96.27 345.34 Heavy oil from Venezuela
25
precipitated asphaltene, wt%

when stock tank oil density decreases. Higher aromaticity for cor-
20
relations in Table 1 shows behavior closer to poly-nuclear-aromat-
ics, and so a crude oil with higher aromaticity contains heavier
components. It can be found that as asphaltenes content of heavy 15
oils in Table 9 decreases, the uid density decreases and modeling
results also show a decreasing trend for asphaltenes aromaticity. Precipitant=n-C5 (Model)
10
It is still unknown how to relate aromaticity parameter to phys- Precipitant=n-C7 (Model)
ical properties of a crude oil. It should be mentioned that we used Precipitant=n-C5 (Exp.)
two aromaticity parameters in this study: aromaticity of Aromat- 5
Precipitant=n-C7 (Exp.)
ics + Resins component and Asphaltenes aromaticity. It would be
great if we can relate these two parameters to the crude oil speci- 0
cations and its asphaltenes physical properties, instead of just t- 0 5 10 15 20 25 30
ting them based on the experimental data, and so one could dilution ratio, cc/gr
improve the predictive nature of the model. To do that, more
experiments and modeling studies need to be done. For instance, Fig. 14. Comparison of the PC-SAFT equation of state tted and the experimental
data for heavy oil from Venezuela. Experimental values are from Sabbagh et al. [21].
given the exact values for molecular weight of asphaltenes, one
can focus on the values of aromaticity parameters because asphalt- for modeling of heavy oils, the key point is using zero binary inter-
enes molecular weight and aromaticity are two related parameters action parameters (kij) for asphaltenes, aromatics and resins com-
in the modeling of asphaltenes phase behavior. ponents, except for kij between asphaltenes and saturates which
A key point for modeling of heavy oils presented in Table 9 is needs to be a negative value. Zero and negative values for kij
using zero binary interaction parameters (kij) for asphaltenes, aro- parameters of heavy oils are not uncommon. Also, comparison of
matics and resins components, except for kij between asphaltenes asphaltenes aromaticity and molecular weight of Iranian oil in
and saturates which needs to be a negative value. Negative kij val- Section 3.2 to those of heavy crudes in this section shows that Ira-
ues for asphaltenes are used by Sabbagh et al. to model the same nian crude which is light oil has a low aromaticity as well as low
crude oils using the PengRobinson equation of state [21]. They asphaltene molecular weight.
used zero kij values for all other components. Also, Panuganti For the rst time, this study shows the application of PC-SAFT
et al. used negative value for kij between asphaltenes and saturates for asphaltenes phase behavior modeling of heavy crudes and bit-
for modeling asphaltenes phase behavior using PC-SAFT [10]. Ta- umens. Using polydispersity concept as well as zero kij parameters
ble 13 represents the values for kij parameters between asphalt- for asphaltenes component are the key points of titration experi-
enes and saturates components for all heavy oils and bitumens ments modeling of heavy crudes using PC-SAFT. Negative kij
presented in Table 9. A single value is used for kij parameter be- parameters between asphaltenes and saturates components can
tween saturates and all asphaltenes sub-fractions. This value help to nd a better match for the onset point and the shape of
strongly controls the onset of precipitation as well as the shape the curve right after precipitation onset.
of the curve right after precipitation onset.
A point which is learned from modeling of light oils versus hea-
vy oils is the difference in the values of binary interaction param- 4. Conclusions
eters used to nd a reasonable match. The values for kij parameters
for light crudes are chosen from the available values in literature This work provides a new insight into the polydisperse behavior
which inturn were determined using VLE experimental data. But, of asphaltenes. In this article, the PC-SAFT equation of state is
216 M. Tavakkoli et al. / Fuel 117 (2014) 206217

Heavy oil from Canada Table 13

Binary interaction parameters between asphaltenes and saturates pseudo-compo-
16 nents used for modeling of heavy oils and bitumens presented in Table 9.

Heavy oil source kij Value between asphaltenes and Saturates

precipitated asphaltene, wt%

14
Venezuela 0.009
12
Canada 0.009
10 Russia 0.000
Indonesia 0.008
8

6
Precipitant=n-C5(Model) Precipitant=n-C6(Model)
Precipitant=n-C7(Model) Precipitant=n-C8(Model)  The proposed approach for selection of PC-SAFT parameters of
4
Precipitant=n-C5(Exp.) Precipitant=n-C6(Exp.) polydisperse asphaltenes simplies the process of parameter
2 Precipitant=n-C7(Exp.) Precipitant=n-C8(Exp.) estimation by reducing the number of adjustable parameters.
0
 By treating asphaltenes as a polydisperse species, the amount of
0 5 10 15 20 25 asphaltenes precipitated increases more gradually with precipi-
dilution ratio, cc/gr tant addition than a monodisperse asphaltene species.
 Comparison of monodisperse and polydisperse asphaltenes
Fig. 15. Comparison of the PC-SAFT equation of state tted and the experimental modeling results showed that polydisperse asphaltenes re-dis-
data for heavy oil from Canada. Experimental values are from Sabbagh et al. [21].
solution begins at lower precipitant volume fractions and with
a higher re-dissolution amount. The amount of re-dissolution
is more for higher carbon number precipitant agents because
Heavy oil from Russia they are better solvents for asphaltenes than lower carbon
8
number n-alkanes.
7
precipitated asphaltene, wt%

 For polydisperse asphaltenes, near the initial asphaltenes insta-

6 bility onset, the precipitated phase is composed mostly of the
heaviest asphaltenes fractions. As the amount of precipitant is
5
increased further, lower molecular weight asphaltenes will also
4 precipitate.
 Sensitivity analysis of model parameters shows that higher
3 Precipitant=n-C5 (Model) asphaltenes aromaticity along with lower asphaltenes molecu-
2 Precipitant=n-C7 (Model) lar weight will result in fewer amounts of precipitated asphalt-
Precipitant=n-C5 (Exp.) enes. This can be explained as an effect of the FloryHuggins
1 interaction parameter difference between light and heavy
Precipitant=n-C7 (Exp.)
0 phases which is higher than the molar volume ratio difference.
0 5 10 15 20  Heavy oil modeling results show that aromaticities of asphalt-
dilution ratio, cc/gr enes, aromatics and resins have a decreasing trend when stock
tank oil density decreases. Higher aromaticity values show
Fig. 16. Comparison of the PC-SAFT equation of state tted and the experimental
behavior closer to poly-nuclear-aromatics, and so a crude oil
data for heavy oil from Russia. Experimental values are from Sabbagh et al. [21].
with higher aromaticity contains more amounts of poly-
nuclear-aromatics fractions.
 Binary interaction parameters analysis show that the kij
Heavy oil from Indonesia
between asphaltenes and saturates fractions plays an important
5
role in the modeling of both light and heavy crude oils diluted
4.5
precipitated asphaltene, wt%

with n-alkanes. It can control the onset of precipitation as well

4 as the shape of the titration curve right after the onset point.
3.5
3
2.5 Precipitant=n-C5 (Exp.) Acknowledgements
2 Precipitant=n-C5 (Model)
1.5 The authors thank DeepStar and the R&D Oil Subcommittee of
the Abu Dhabi National Oil Company for their nancial support.
1
The authors also thank Dr. Francisco M. Vargas for his valuable dis-
0.5
cussion about the polydisperse asphaltenes modeling.
0
0 10 20 30 40 50
dilution ratio, cc/gr References

Fig. 17. Comparison of the PC-SAFT equation of state tted and the experimental [1] Hirschberg A, deJong LNJ, Schipper BA, Meijer JG. Inuence of temperature and
data for heavy oil from Indonesia. Experimental values are from Sabbagh et al. [21]. pressure on asphaltene occulation. SPE J 1984;24:28393.
[2] Alboudwarej H, Akbarzedeh K, Beck J, Svrcek WY, Yarranton HW. Regular
solution model of asphaltene precipitation from bitumen. AIChE J
2003;49:294856.
[3] Panuganti SR, Vargas FM, Chapman WG. Modeling reservoir connectivity and
shown to reasonably model the phase stability of asphaltenes in tar mat using gravity-induced asphaltene compositional grading. Energy Fuels
2012;26:254857.
crude oils diluted with n-alkanes. The asphaltenes and crude oils [4] Leontaritis KJ, Mansoori GA. Asphaltene occulation during oil production and
tested are from a variety of sources and over a wide range of oil processing: a thermodynamic colloidal model. SPE 16258. In: SPE international
density including heavy oils and bitumens. The following conclu- symposium on oileld chemistry, San Antonio, February 46 1987.
[5] Victorov AI, Firoozabadi A. Thermodynamic micellization model of asphaltene
sions can be made from the investigations described in this paper. precipitation from petroleum uids. AIChE J 1996;42:175364.
 M. Tavakkoli et al. / Fuel 117 (2014) 206217 217

[6] Panuganti SR, Tavakkoli M, Vargas FM, Gonzalez DL, Chapman WG. SAFT model [14] Wang JX. Predicting asphaltenes occulation in crude oils. Ph.D. thesis 2000.
for upstream asphaltene applications. uid phase equilibria; 2013; <http:// In: Petroleum and chemical engineering, New Mexico Institute of Mining and
dx.doi.org/doi:10.1016/j.uid.2013.05.010>. Technology. Socorro.
[7] Ting PD, Gonzalez DL, Hirasaki GJ, Chapman WG. Application of PC-SAFT [15] Ting PD. Thermodynamic stability and phase behavior of asphaltenes in oil and
equation of state to asphaltene phase behavior. In: Mullins OC, Sheu EY, of other highly asymmetric mixtures. Ph.D. thesis 2003. In: Chemical and
Hammani A, Marshall AG, editors. Asphaltenes, heavy oils, and Biomolecular Engineering, Rice University. Houston.
petroleomics. New York: Springer; 2007. p. 30128. [16] Wiehe IA, Yarranton HW, Akbarzadeh K, Rahimi PM, Teclemariam A. The
[8] Victorov AI, Smirnova NA. Description of asphaltene polydispersity and paradox of asphaltene precipitation with normal parafns. Energy Fuels
precipitation by means of thermodynamic model of self-assembly. Fluid 2005;19:12617.
Phase Equilib 1999;158160:47180. [17] Strausz OP, Safarik I, Lown EM, Morales-Izquierdo A. A critique of asphaltene
[9] Wang J, Buckley J. Effect of dilution ratio on amount of asphaltenes separated uorescence decay and depolarization-based claims about molecular weight
from stock tank oil. J Dispersion Sci Technol 2007;28:42530. and molecular architecture. Energy Fuels 2008;22:115666.
[10] Panuganti SR, Vargas FM, Gonzalez DL, Kurup AS, Chapman WG. PC-SAFT [18] Buenrostro-Gonzalez E, Lira-Galeana C, Gil-Villegas A, Wu J. Asphaltene
characterization of crude oils and modeling of asphaltene phase behavior. Fuel precipitation in crude oils: theory and experiments. AIChE J 2004;50:255270.
2012;93:65869. [19] Wu J, Prausnitz JM, Firoozabadi A. Molecular thermodynamics of asphaltene
[11] Sedghi M, Goual L, Welch W, Kubelka J. Effect of asphaltene structure on precipitation in reservoir uids. AIChE J 2000;46:197209.
association and aggregation using molecular dynamics. J Phys Chem B [20] Gonzalez DL. Modeling of asphaltene precipitation and deposition tendency
2013;117:576576. using the PC-SAFT Equation of State. Ph.D. thesis 2008. In: Chemical and
[12] Buckley JS, Hirasaki GJ, Liu Y, Von Drasek S, Wang JX, Gill BS. Asphaltene biomolecular engineering, Rice University. Houston.
precipitation and solvent properties of crude oils. Pet Sci Technol [21] Sabbagh O, Akbarzadeh K, Badamchi-Zadeh A, Svrcek WY, Yarranton HW.
1998;16:25185. Applying the PR-EoS to asphaltene precipitation from n-alkane diluted heavy
[13] Gonzalez DL, Hirasaki GJ, Chapman WG. Modeling of asphaltene precipitation oils and bitumens. Energy Fuels 2006;20:62534.
due to changes in composition using the perturbed chain statistical associating
uid theory equation of state. Energy Fuels 2007;21:123142.