Professional Documents
Culture Documents
Landera2021 Modelo Predicción Reid Vapor Pressure
Landera2021 Modelo Predicción Reid Vapor Pressure
Fuel
journal homepage: www.elsevier.com/locate/fuel
A R T I C LE I N FO A B S T R A C T
Keywords: The growing demand for inexpensive, environmentally benign fuels has sparked research into fuel properties
Reid vapor pressure that affect fuel production, handling and engine efficiency. Reid vapor pressure (RVP) has been shown to be one
RVP such important fuel property. Having models that can accurately predict the RVP of a fuel rapidly and robustly
Oxygenates can aid in the design of de novo fuels and increase the value proposition to the refinery. Conventional fossil fuels
Alcohols
typically consist of 100 s of highly non-ideal components, especially when oxygenates are involved, and current
Blend models
methods often fail to predict RVP correctly. The large number of components can render conventional modelling
approaches burdensome and inaccurate. To overcome these barriers, a robust methodology for the rapid and
accurate prediction of the RVP of complex base fuels blended with oxygenated biofuels via surrogate models was
developed. The aim was to minimize the number of components necessary to accurately represent the complex
mixture, whilst retaining the ability to accurately predict RVP of the base fuel. Three models were developed
using the SAFT-γ approach, which varied in complexity and accuracy of prediction. The RVP for a suite of
oxygenates, blended at 10, 20, and 30%, by weight into blendstocks for oxygentated blending (BOB) was studied
in order to evaluate which oxygenate families have potential to reduce RVP. The majority of oxygenates eval-
uated decreased the RVP of the mixture, with the branched alcohols and the ketones showing the largest de-
crease. These results have promising implications for biofuels development and biorefinery economics.
1. Introduction year 2025 [11]. It is therefore imperative that strategies are developed
to reduce the CO2 intensity of transport fuels, such that GHG emissions
Whilst the transportation sector is evolving the general consensus is are reduced.
that liquid fuels will remain dominant for the next several decades [1]. One such strategy is to concentrate on the design of the fuel, its fuel
Although commercial sales of electric cars have steadily increased, to properties, and how they directly impact its combustion characteristics.
date, in the United States, they still account for less than 2% of all auto The vapor pressure of a fuel is an important fuel property, as it de-
sales [2]. Indeed, annual growth rates for electrified cars has been de- scribes the facility of fuel volatilization once it is injected into the en-
clining for each of the past 5 years [3]. Globally, the electric car stock is gine cylinder. When the vapor pressure of a fuel is measured at 37.8 °C,
equivalent to just 0.2% of all Passenger Light-Duty Vehicles (PLDV) [4]. in a container with a V/L ratio of 4:1, it is referred to as the Reid vapor
A number of studies have attempted to forecast the market share of pressure (RVP), and is the industry standard measure for vapor pressure
electric vehicles in the United States [5–9]. Forecasts vary, but by the [12]. Low RVP fuels are associated with cold start problems and higher
year 2030, the most optimistic forecasts show a market share, in the soot production, while high RVP fuels are associated with vapor lock
United States, of only 30%. In 2017, the transportation sector ac- [13]. In the United States, in order to be distributed, the maximum RVP
counted for 29% of all the greenhouse gas (GHG) emissions in the allowed is within a range of 54 and 103 kPa, as a function of location
United States. About half of that came from passenger cars and light (primarily altitude) and season [14]. Therefore, fuels vary in compo-
duty trucks [10]. The production of crude oil in the United States is sition based on the region and season in which they are being used. A
expected to increase for the next several years, plateauing out in the ‘summer’ fuel, for example, has a low RVP due to higher day time
⁎
Corresponding author.
E-mail address: angeorg@sandia.gov (A. George).
https://doi.org/10.1016/j.fuel.2020.118624
Received 30 December 2019; Received in revised form 29 June 2020; Accepted 3 July 2020
0016-2361/ © 2020 Published by Elsevier Ltd.
A. Landera, et al. Fuel 283 (2021) 118624
temperatures, while a ‘winter’ fuel is allowed to have a higher RVP. Although, some work has been aimed at reducing emissions from
The primary test for RVP is governed by ASTM-D5191, which refineries, this has been primarily focused on improving seals, and
measures the RVP of a petroleum product at a V/L ratio of 4:1 [12]. eliminating leaks. An additional way of reducing fugitive losses is to
RVP tests are costly and time-consuming to perform, and therefore the focus on the refinery product itself, via reducing its vapor pressure.
development of a high-throughput screening approach for the reliable Recently, for example, it was found that adding ethyl acetate in con-
description of RVP is of considerable value for the development of novel junction with ethanol decreased the RVP of gasoline [31]. In addition,
fuels in a high-throughput fashion. A key challenge in this context is the the ethyl acetate increased the RON and MON of the original gasoline,
complexity of the fuel mixtures. Due to the complex nature of fuels, the while keeping the OS constant. Ratcliff et al. evaluated the combustion
development of surrogate fuels for a variety of research applications is a characteristics of a 12% i-butanol/7% ethanol blend [32]. Their find-
promising strategy. These surrogates generally have a reduced number ings show that the resulting RVP does not increase (as it would for
of components, but are designed to be representative of the fully ethanol only blends), showing the benefit of adding other oxygenates to
complex fuel, for the attribute being evaluated; in this study, RVP. a petroleum fuel in an effort to decrease fugitive losses. Lastly, An-
The Research and Motor Octane Number, or RON and MON re- dersen et al. tested a 50/50 mixture of ethanol and 1-butanol blended
spectively, are used to describe the knocking tendency of a fuel for use with gasoline and showed that the resulting RVP curve features a much
in a Spark Ignition (SI) engine [15–18]; fuels with a high RON, and low smaller increase in RVP, than ethanol alone. [24]. Although, the au-
MON are considered to have the best anti-knocking behavior. Whilst thors did not evaluate the impact of RVP lowering oxygenates and their
engine configuration is also important, a fuel that has the aforemen- refinery applications, 1-butanol may be used as a refinery additive to
tioned qualities is said to display a large Octane Sensitivity (OS), which lower the fugitive loss of hydrocarbons. Developing a model, which is
is the difference between RON and MON measurements. Therefore, OS capable of high-throughput screening of fuels is therefore key to ac-
is a useful metric for the evaluation of potential fuels. celerating progress in this space.
Small alcohols, such as methanol and ethanol, blended into BOBs Efforts to develop models capable of predicting the vapor pressure
have several advantages over their non-oxygenated counterparts. They of complex mixtures date back to the mid-20th century, with varying
tend to have better RON, MON, and OS, making them far more suitable degrees of success. In the 1960s, three methods were available for es-
for use in SI engine designs [19]. Chemical kinetics modeling has shown timating the vapor pressure of petroleum fuels at high temperature and
that this can be explained by their inability to undergo RO2 chain pressure, namely the British Petroleum, Maxwell, and Edmister
branching chemistry, which would give rise to Negative Temperature methods [33–35]. The British Petroleum method mimics the distillation
Coefficient (NTC) behavior [20]. It is for this reason that methanol and process by using theoretical plates. Using empirical relationships, it
ethanol have an OS of 20, and 19, respectively. In addition, alcohol then calculates the partial pressure for each theoretical plate. The sum
fuels, on the basis of ppm in the exhaust, show a decrease in carbon of these partial pressures is then considered to be the True vapor
dioxide (CO2), and hydrocarbon (HC) emissions, with an increase in NO pressure (TVP). Alternatively, the Maxwell and Edmister method both
emissions, as vehicle load increases [21,22]. It may, however, be worth rely on converting an ASTM distillation curve into either a True Boiling
noting that whilst these fuels show a decrease in CO2 emissions on a Point (TBP) distillation curve, or a (pressure, volume, temperature) PVT
mass basis, owing to their reduced energy density relative to conven- phase diagram, respectively. While these methods were reasonably
tional gasoline fuels, alcohol fuels have an increased level of CO2 accurate, they relied on a large number of conversion tables, and were
emissions on the basis of energy service delivered, and thus, in order to limited to high temperature and pressure estimations. In 1980, Eg-
be consistent with a deep decarbonization scenario, the carbon source gertsen, Nygard, and Nickoley used a gas chromatograph to obtain the
for these fuels needs to be ultimately atmospheric [23,24]. In addition, mole fraction of each component in a fuel [36]. They were then able to
it has been previously shown that small linear alcohols show marked estimate the vapor pressure of petroleum distillate fractions by applying
non-linear blending behavior, with respect to RVP, when blended into Dalton’s, and Raoult’s laws. For each fuel component, Antoine’s vapor
BOBs. In 2010, Andersen et al. measured the RVP of several alcohol- pressure correlation was used to obtain the saturation pressure. In
gasoline blends from 5 to 85 vol% alcohol [25]. Their results confirmed 1984, Wilhelm and Prausnitz used perturbed hard-chain equation of
the non-linear effects of linear alcohols, with the non-linear effects state (EoS) methods to estimate the vapor pressure of crude oil fractions
decreasing as the size of the alcohol increases. As the size of the alco- [37]. Their predictions were in reasonable agreement with experi-
hols increases, they have characteristics more closely resembling linear mental vapor pressure data. In 1985, Furey examined the volatility
alkanes, rather than alcohols, and the non-linear effects caused by the characteristics of oxygenates dissolved in different fuels [38]. Furey
hydroxyl group are diminished. Several oxygenates beyond alcohols was particularly interested in the sharp boost in RVP that occurs when
have been tested as possible additives into fuels. In 2011, Christensen some small alcohols are used as the oxygenate. From his work, he de-
and others looked at the fuel properties of furans, levulinates, and al- veloped a simple relationship, RVP = kCn, which can model the early
cohols blended into three BOBs [26]. They observed that most oxyge- rise in RVP, which occurs when the first 1–2% of alcohol is blended. In
nates decreased the RVP of a BOB, making them suitable for blending his relationship, C is the alcohol concentration, while k and n are
with high RVP BOBs. In addition, McCormick et al. identified several constants dependent on the neat gasoline and alcohol [38]. A few years
oxygenates whose properties offer high promise for improving the later, Chen developed a volatility model for gasoline based on a 10-
performance of high-octane advanced SI engine fuel, when blended into component surrogate [39]. The ten components are fuel dependent but
a conventional or reformulated blendstock for E10 blending [27]. These are based on grouping similar-sized species together, and selecting the
oxygenates were identified as ethanol, 2-butanol, and isobutanol. compound with the largest wt. % in that group. The surrogate was then
Oxygenates which can decrease the vapor pressure of fuels may also used in isothermal flash calculations, using the Peng-Robinson EoS. The
find utility in refinery applications, where high vapor pressures can lead surrogate performed well for non-oxygenated fuels but led to inaccurate
to non-negligible losses, known as fugitive losses, of refinery products results when oxygenates were involved. For methanol, the ten-compo-
via, for example, evaporative emissions from storage tanks and pipes nent surrogate underestimated the initial rise in RVP by approximately
[28,29]. In 2008, a team studied the emissions of hydrocarbons from a 10 kPa, or an error of 12.5%. The authors suggest this may be due to the
typical refinery, and concluded that 0.17% of the mass of the refinery polar nature of methanol. In 1999, Gahbri et al. used a Neural-Network
hydrocarbon were lost due to emissions, a loss in profits by $3.1 million model to estimate the bubble point pressure of crude oil and gas mix-
per year [30], on the basis of a $40 per barrel market value. Minimizing tures. Their neural network was trained using 5200 PVT datasets, and
fugitive losses can therefore clearly improve the economics for re- obtained a correlation coefficient of 0.9891 [40]. The maximum abso-
fineries, decrease pollutants, and increase the efficiency of refinery lute relative error was reported to be 19.93%, while the average ab-
processes. solute relative error was 6.48%. In 2014, Albahri used a neural network
2
A. Landera, et al. Fuel 283 (2021) 118624
algorithm to predict the RVP of 362 different petroleum fractions. Of precision required. These models differ in the number of components
these, 300 were used in the training set, and 62 were used as a testing chosen to represent a given BOB, with the small model employing the
set. For the training set, the average deviation was 2.29 kPa, and the fewest number, and the large model employing the largest number of
maximum deviation was reported to be 13.85 kPa. For the testing set, components. These surrogate models can then be used with the SAFT-γ-
the average deviation was 2.24 kPa, and the maximum deviation was mie EoS, which has been shown to reliably predict thermophysical
reported to be 13.85 kPa. In total, Albahri’s work resulted in a corre- properties of complex fluid mixtures. These models are validated using
lation coefficient of 0.995, equivalent to a relative deviation of 5.09% seven BOBs, with varying chemical compositions, and RVPs. This study
[41]. However, petroleum fractions do not contain oxygenates, and it is then goes on to evaluate the performance of these models in predicting
as yet uncertain how well this method will fare when including oxy- the RVP of oxygenated BOBs. To this end, the models are validated
genated blendstocks. Later, Adlene and coworkers used a fitting pro- using two BOBs (a summer and a winter), each with data on blends with
cedure to predict the RVP of a fuel. They validated the resulting model several oxygenates. A physical interpretation of the interaction of aro-
using a variety of fuels, with a correlation coefficient of 0.87 [42]. Plots matics with alkanes and alcohols is then given. Appealing to previous
of the measured RVP vs. the predicted RVP show large deviations, with excess thermodynamic measurements reported in the literature, an
the largest deviations being about 50%. In 2019, several oxygenates explanation for why these interactions are important in understanding
were blended into a CARBOB (California reformulated BOB), with RVP is made. Lastly, predictions are then made on the RVP of several
subsequent RVP measurements taken [43]. Predictions using five EoS oxygenates blended into a Summer BOB, at 1 vol% increments, ending
methods demonstrated reasonable accuracy in being able to reproduce at 30% by volume. These predictions are made using the large surrogate
experimental measurements for oxygenated blends. Of the five EoS model.
methods used, the two best performing were UNIFAC-Dortmund, and The novelty of this contribution lies in the development of a gen-
the CPA (Cubic Plus Association), which have a Root Mean Square erally applicable approach for the development of reliable surrogate
Errors (RMSE) of 5.2 and 5.5%, respectively. While these RMSE are models, including when oxygenates are present. These surrogate
relatively good, it is interesting to note that all of the methods tested models will aid the development of novel fuels by providing a platform
show deteriorating results towards the high end of their RVP mea- for high- throughput screening, and also help identify oxygenates with
surements. In addition, when blended with ethanol, neither of the five promising refinery applications.
EoS produce accurate results, with the best performing EoS being de- Theory/calculation
viating by more than 10% (~5.1 kPa). In 2011, the Vapor Liquid The method used to predict the RVP is the SAFT-γ-Mie EOS. SAFT-γ-
Equilibrium (VLE) of binary and ternary mixtures of long-chain me- Mie’s robustness relies on an efficient and reliable way to construct the
thylesters was modeled using the SAFT-VR EoS [44]. Predictions Helmholtz free energy of a system. Here, only a brief overview is pro-
slightly underestimated the experimentally derived VLE of the methyl vided, and interested readers are directed to the original papers for
palmitate/methyl linolenate binary mixture, and nearly perfectly re- more details [47–49]. SAFT-γ-Mie calculates the Helmholtz free energy
produced the experimentally derived VLE of the methyl oleate/methyl of a given fluid by decomposing it into specific and meaningful com-
linolenate binary mixture. A residue curve map was generated for the ponents. This is represented below as:
ternary mixture, consisting of methyl palmitate, methyl linolenate, and
A = Aideal + Amono + Achain + Aassoc (1)
methyl oleate. Predictions of the bubble pressure, using the SAFT-VR
EoS, are nearly identical to previous predictions using an empirical EoS. where Aideal
is the Helmholtz free energy obtained through the ideal gas
In 2014, Oliveira and coworkers published predictions on the vapor law, Amono is the residual accounting for monomeric Mie segment in-
pressure of several pure Fatty Acid Methyl Esters (FAME), using the teractions, Achain is the residual accounting for molecule formation, and
soft-SAFT EoS [45]. Overall 14 different FAMEs were chosen. Predic- Aassoc is a term that accounts for association interactions. With the ex-
tions were nearly identical to experimental measurements at tempera- ception of the ideal term, these terms can be obtained by using per-
tures beyond about 350 K but tended to slightly over predict the vapor turbation theory as applied by Barker and Henderson [50], or Wer-
pressure at lower temperatures. A %AAD total for all of their vapor theim’s first order perturbation theory (TPT1) [51–54]. The Mie
pressure data was calculated to be 5.8%. It is suggested by the authors potential, Φ , which offers more flexibility than the Lennard Jones po-
that one reason for the low temperature deviation may be due to the tential, describes the interaction between segments. The interaction
low inherent vapor pressures of FAMEs, which may be accompanied by potential between two identical segments is given by the following
larger experimental errors. In the same year, Perdomo, Millan, and formula:
Aragon published vapor pressure predictions of several saturated and r
λkk a
λkk
unsaturated FAMEs, using the SAFT-γ EoS [46]. While their predictions ⎡ σ σ ⎤
Mie
Φkk (rkk ) = Ckk εkk ⎢ ⎛ kk ⎞
⎜ ⎟ − ⎛ kk ⎞
⎜ ⎟
⎥
were always in good agreement with the available research, they al- r ⎝ rkk ⎠
⎣ ⎝ kk ⎠ ⎦ (2)
ways slightly underpredicted the vapor pressure by a few kPa. Predic-
tions were slightly worse as the degree of unsaturation increased, and where rkk is the center of mass to center of mass distance between the
the authors propose that the inclusion of more unsaturated FAMEs into two segments, σkk is the diameter of each segment, εkk is the depth of the
r a
their objective function could improve predictions. potential well. λkk , and λkk are repulsive and attractive factors. Ckk is a
Most previous work in this area has focused on developing models normalization factor designed to ensure that the magnitude of the po-
suitable for only very specific fuels. Each new fuel requiring a new tential minimum coincides with εkk . It is defined as:
model. This is laborious and inefficient. If a general model could be r
λkk
use, fuels could be screened in a high-throughput fashion. Novel fuels λkk − λkk ⎝ λkk ⎠ (3)
could be screened in silico before production begins, and lead times to
Values for Mie potential parameters can be obtained by regression
more functional fuels can be reduced.
to appropriate pure component data, such as density, or vapor pressure.
The present study aims to develop a set of rules, which will enable
Unlike interactions (also known as cross interaction parameters), i.e.,
customized RVP prediction models for a target fuel. Three sets of rules
those interactions between two segments of different types, may be
were developed. The first set of rules leads to the creation of a “small
obtained by simple combining rules, which involve the individual like
surrogate model”, while the second and third set of rules correspond to
interactions. These combining rules are summarized below.
a “moderate” and “large” surrogate model, respectively. These models
increase in complexity and precision, therefore for each target blend- σkk − σll
σkl =
stock one of the three models can be used depending on the level of 2 (4)
3
A. Landera, et al. Fuel 283 (2021) 118624
3 3
σkk σll
εkl = εkk εll
σkl3 (5)
4
A. Landera, et al. Fuel 283 (2021) 118624
components identified in a BOB along with the quantity of each com- Table 1
ponent present. In addition, a particular BOB is described in terms of Vapor pressure at 37.8 °C for a homologous series of par-
the proportion of chemical classes present. Typically, these chemical affin, i-paraffin, aromatic, olefin, and naphthene chemical
classes are n-paraffins, i-paraffins, aromatics, olefins, and naphthenes. groups. Vapor pressure is given in units of kPa.
Using the DHA of a fuel, three sets of simple rules for the inclusion Paraffin
of components into an RVP model were developed. They lead to the Chemical name VP, 37.8 °C
development of a small, moderate, and large surrogate model. These propane 1301.96
butane 354.65
models developed also give insight into the predicted RVP as more
pentane 107.45
components are added to the model. The rules used to develop the hexane 34.20
models are as follows, with a detailed procedure shown in the supple- heptane 11.18
mentary materials: I-Paraffin
Chemical name VP, 37.8 °C
2.2.1. Rules for small surrogate model i-butane 498.90
The small surrogate model includes all those components which i-pentane 142.86
i-hexane 46.69
make at least a 1 mol% contribution to the composition of the BOB. The i-heptane 15.67
composition of the components, within a chemical class, is then nor-
Aromatics
malized to the fraction of the chemical class present in the fuel, ac-
Chemical name VP, 37.8 °C
cording to the DHA. The chemical classes used here are n-paraffins, i- benzene 22.24
paraffins, aromatics, naphthenes, and olefins. This normalization step toluene 7.12
ensures that the proportion of each chemical class is preserved. In m-xylene 2.24
general, implementing the small surrogate models will include ~10–15 Olefin
components. Chemical name VP, 37.8 °C
1-hexene 41.45
1-heptene 14.36
2.2.2. Rules for the moderate surrogate model
The moderate surrogate model, includes all those components Naphthene
Chemical name VP, 37.8 °C
which make at least a 0.5 mol% contribution to the composition of the
cyclopentane 68.41
BOB, and normalization undertaken as with the small surrogate model. methylcyclopentane 31.07
In general, these models include ~30 components. In general, the small
and moderate models are easy to implement, and represent a very small
fraction of the total number of components present in a typical BOB. If Table 2
this threshold is decreased further from 0.5 mol% to 0.1 mol% a typical Predicted and measured values for the seven BOBs used in this study, using the
model would include more than 80 components. Such a model would small, moderate, and large surrogate model. Values are in kPa.
be cumbersome to produce, and difficult to calculate; and so, a slightly Small Moderate Large Measured
different set of rules for obtaining the large surrogate model have been
developed. BOB 1 35.21 38.35 38.47 37.43
BOB 2 104.60 98.01 87.34 89.71
BOB 3 23.76 26.38 25.76 27.64
2.2.3. Rules for the large surrogate model BOB 4 57.25 51.72 50.35 52.46
The large surrogate model includes those n-paraffins and i-paraffins BOB 5 66.97 61.52 58.62 54.16
that make at least a 0.1 mol% contribution to the BOB. A 1 mol% BOB 6 43.35 42.19 40.12 41.45
threshold is used for aromatics, and only the most abundant naphthene BOB 7 57.82 58.73 54.84 56.44
%AAD 10.59 5.302 4.36
is included. Olefins are included, as α-olefins, and only those that make
at least a 1 mol% contribution are included. In order to ensure that
some aromatics are included in each model the three isomers of xylene used in each model, for each BOB, can be found in the supplementary
are treated as one component, and the same normalization procedure materials section of this paper. For convenience, a Standard Operating
followed in the development of the small and moderate surrogate Procedure (SOP) for the development of the models outlined in this
model is used. On average, the large surrogate model employs 39.6 section has been written. It can be found in the Supplementary
components. This is only slightly larger than the 30 components re- Materials Section.
quired to use the moderate model. The advantage of the large surrogate
model is not only in the number of components used to model RVP, but
also in the type of components used. It draws heavily on paraffins, 2.3. Model validation
while de-emphasizing aromatics. This works because the range of RVPs
present in paraffins is much greater than that which is present in aro- Fig. 2 graphically illustrates the performance of these three models
matics. Table 1 illustrates this idea by showing the RVP of a homo- using the seven BOBs available for validation. Table 2 shows both the
logous series of paraffins, i-paraffins, aromatics, olefins, and naph- predicted, and measured RVP values. All three models perform well,
thenes [79]. In this study, when the terms small surrogate model, with the %AAD of the moderate and large surrogate model both cal-
moderate surrogate model, or large surrogate model are used, it is in culated to be < 6%. The small surrogate model is less accurate, with a
reference to the model developed from the corresponding set of rules. %AAD of 10.59%. On average, 19.8 components are included in the
When analyzing Table 1, it can be observed that for any surrogate small surrogate model. This means that the small surrogate model can
model to succeed at predicting RVP it must contain most of the low produce modestly accurate results, with very few components.
molecular weight paraffins and i-paraffins. Propane, butane, and pen- By percentage, all models performed less well when trying to model
tane all have RVPs above 100 kPa. If they are not included large errors BOB 5. BOB 5 is anomalous, because it contains a very large amount of
in RVP predictions are likely to occur. For aromatics, and naphthenes, naphthenes, 29.98 mol%. The dispersion energy parameter (εkk) for the
the difference between successive members of the homologous series is cyCH2 (cyclic methene) group was not optimized for RVP predictions,
much smaller thereby making them less important in RVP surrogate and was kept at its default SAFT-γ-Mie value. Optimization of the
models. Although the smallest olefins do have large RVPs, the propor- cyCH2 εkk parameter for RVP predictions would likely improve the
tion of olefins in most BOBs are very small. Tables of the mole fractions performance of the model when applied to BOB5. However, the large
5
A. Landera, et al. Fuel 283 (2021) 118624
Fig. 3. Performance of the small and moderate surrogate model validated using
oxygenate blending data from a summer and winter BOB. Values are in kPa. The
data show that these models tend to over-predict RVPs for oxygenate containing
BOBs.
6
A. Landera, et al. Fuel 283 (2021) 118624
Fig. 5. Excess enthalpy data for a series of n-alkane + toluene binary systems.
Excess enthalpies are given in J/mole. Excess enthalpies increase as the n-al-
Fig. 4. Validation of the large surrogate model: Plot of the predicted vs. mea-
kane size increases.
sured RVP of oxygenates blended into the Winter and Summer BOB. RVPs are
given in units of kPa. This plot illustrates the statistically superior performance
of the large surrogate model vs. the small and moderate surrogate model.
2.4.1. Physical explanation Fig. 6. Excess enthalpy data for a series of cycloalkanes + toluene binary
Although having models that correctly predict the RVP of BOBs is mixtures. Excess enthalpies are given in J/mole. Cycloalkanes weaken the
important, it would be complimentarily interesting to understand some aromatic interactions between toluenes.
of the important interactions which give rise to the RVP observed for a
given BOB. To do this, it is important to examine excess thermodynamic
properties. Excess thermodynamic properties are derived from the ap-
propriate excess thermodynamic function and is the difference between
the observed thermodynamic function of mixing and the function for an
ideal solution. Excess properties can give insight into the physical in-
teractions occurring in a mixture. There are many excess thermo-
dynamic properties; here the focus is mainly on the excess enthalpy
(HE). Figs. 5 and 6 show the HE of several n-alkane + toluene, and
cycloalkane + toluene mixtures at 298.15 K [82–91]. The n-alkanes
range from n-hexane to n-hexadecane [82–91]. Interactions with to-
luene are of interest because there are sufficient measurements of to-
luene, and it represents an aromatic with a side chain. Benzene could
have also been considered, but it would not have been a good proto-
typical aromatic found in fuel. From Fig. 5, HE increases with the size of
n-alkane indicating that larger n-alkanes are more capable of disrupting
toluene-toluene interactions. Fig. 7. Excess enthalpies of alcohols + toluene at 298.15 K. Excess enthalpies
In 2011, using two-color resonant two-photon ionization spectro- are in J/mole.
scopy, as well as ab initio calculations, Fujii et al. investigated the non-
covalent interactions of benzene-paraffin clusters [92]. A correlation
This computational study found a correlation between the average
between the strength of the interaction energy, and average polariz-
polarizability of the alkane and the interaction energy. Larger alkanes,
ability of the paraffin was found, and based on this, it was determined
given their higher polarizabilities, are capable of larger dispersion
that these interactions between benzene and paraffins (normal/bran-
forces, and thus larger strengths of interaction. This is in accord with
ched) were mainly due to dispersion. These results are reinforced by
the results shown in Fig. 5. It is important to note that these compu-
computational results carried out on saturated hydrocarbon - benzene
tational results, and the resulting correlation between polarizability and
complexes. The saturated hydrocarbons studied were propane, iso-
strength of interaction were also found to extend to iso-alkanes, and
butane, cyclopropane, cyclobutane, cyclopentane, cyclohexane, cy-
naphthenes. Indeed, these findings are also in agreement with the data
clooctane, and bicyclo[2.2.2]octane [93].
7
A. Landera, et al. Fuel 283 (2021) 118624
Using the large surrogate model, predictions were made for dif- ketones. The overall trend in both of these groups is to lower the RVP
ferent oxygenates blended into the Summer BOB. These predictions will when blended into a BOB. In comparison to alcohols, esters and ketones
aid researchers in their understanding of how different oxygenates, lower the RVP of a BOB by a further 1 to 2 kPa. While still exhibiting
when blended into a BOB, effect the RVP of that BOB. The first set of antagonistic blending, esters and ketones show more linear blending
predictions were performed for alcohols. Fig. 8 shows the results of behavior than alcohols. Fig. 10 shows the same type of predictions, but
these predictions. The RVP predictions are plotted on the left panel, and applied to cyclic ketones. It is clear that synergistic blending occurs for
on the right panel, the predicted RVP minus the RVP taken from the all cyclic ketones, with cyclooctanone exhibiting large synergistic
linear interpolation blend model are plotted. The goal in this study was blending behavior.
to find RVP suppression i.e. that the RVP is lowered with addition of an Overall, it was found that higher molecular weight oxygenates
oxygenate. If values on the right panel are negative, they correspond to tended to decrease the RVP of a BOB more than lower molecular weight
synergistic lowering of the RVP, and the RVP predicted is lower than oxygenates.
what the linear interpolation blend model predicts. Conversely, if those The lowering trend of many of the oxygenates tested shows the
values are positive, they indicate antagonistic blending, i.e. and the potential for reducing the RVP of refinery products, especially cyclic
RVP predicted is higher than what the linear interpolation blend model ketones. As highlighted in Section 1, this lowering can account for in-
predicts. A synergistic lowering of the RVP is favorable for refinery creased profits, and reduced GHG emissions emitted into the atmo-
applications, because lower RVP fuels lead to less fugitive losses. sphere. Even at 10% blend levels, many of the oxygenates tested show
Fig. 8 shows that, in general, only the higher molecular weight al- decreases of the neat BOB RVP by at least 1 kPa, a substantial reduction
cohols tend to decrease the RVP of the BOB they are blended into. All which can reduce costs over the long run. There is also an opportunity
alcohols showed antagonistic blending behavior with 2-pentanone and for some of these oxygenates to be used in the development of better
2-hexanone behaving the most linearly (Fig. 9). performing vehicle fuels. As an additive, they can take the RVP of an
Fig. 8 shows the same type of predictions made for esters and otherwise good performing fuel and lower it so that it meets
8
A. Landera, et al. Fuel 283 (2021) 118624
Acknowledgments
9
A. Landera, et al. Fuel 283 (2021) 118624
[24] Daggash HA, Patzschke CF, Heuberger CF, Zhu L, Hellgardt K, Fennell PS, Bardow Propylbenzene at 313.15 K. Fluid Phase Equilib 1993;1:201–9.
A, Bhave AN, Mac Dowell N, et al. Closing the carbon cycle to maximise climate [56] Chamorro CR, Martin MC, Villamanan MA, Segovia JI. Characterization and
change mitigation: power-to-methanol vs. power-to-direct air capture. Sustainable modelling of a gasoline containing 1,1-dimethylethyl methyl ether (MTBE), dii-
Energy Fuels 2018;2:1153–69. sopropyl ether (DIPE) or 1,1-dimethylpropyl methyl ether (TAME) as fuel oxyge-
[25] Andersen VF, Anderson JE, Wallington TJ, Mueller SA, Nielsen OJ. Vapor pres- nate based on new isothermal binary vapor-liquid data. Fluid Phase Equilib
sures of alcohol-gasoline blends. Energy Fuels 2010;6:3647–54. 2004;1:105–12.
[26] Christensen E, Yanowitz J, Ratcliff M, McCormick RL. Renewable oxygenate [57] Gardeler H, Horstmann S, Tsuboi A, Toba S, Rarey J, Gmehling J. Vapor-liquid
blending effects on gasoline properties. Energy Fuels 2011;10:4723–33. equilibria at six temperatures from 313.15 K to 402.15 K and excess enthalpy data
[27] McCormick R, Fiorini G, Fouts L, Christensen E, Yanowitz J, Poliarpov E, et al. at 363.15 K and 1.65 MPa for the system toluene + 2,2,4-trimethylpentane (iso-
Selection criteria and screening of potential biomass-derived streams as fuel octane). ELDATA Int Electron J Phys Chem Data 1998;1:1–10.
blendstocks for advanced spark-ignition engines. SAE Int J Fuels Lubr [58] Moreau A, Segovia JJ, Villamanan MA, Martin MC. Vapour-liquid equilibria of the
2017;2:442–60. ternary mixture (1-pentanol + 2,2,4-trimethylpentane + heptane) and the binary
[28] McKay G, Holland CR. Energy savings from steam losses on an oil refinery. Eng mixture (2,2,4-trimethylpentane + heptane) at T = 313.15 K for the character-
Costs Prod Econ 1981;3–4:193–203. ization of second generation biofuels. Fluid Phase Equilib 2015;1:101–6.
[29] Rao P, Ankam S, Ansari M, Gavane AG, Kumar A, Pandit VI, et al. Monitoring of [59] Villamanan RM, Martin CM, Chamorro CR, Villamanan MA, Segovia JJ.
hydrocarbon emissions in a petroleum refinery. Environ Monit Assess Thermodynamic of fuels with a biosynthetic component: vapor-liquid equilibrium
2005;1–3:123–32. data for binary and ternary mixtures containing ethyl 1,1-dimethylethyl ether, n-
[30] Chambers AK, Strosher M. Direct measurement of fugitive emissions of hydro- heptane, and toluene at T = 313.15 K. J Chem Eng Data 2006;6:2091–5.
carbons from a refinery. J Air Waste Manage Assoc 2008;8:1047–56. [60] Ashcroft SJ, Clayton AD, Shearn RB. Isothermal vapor-liquid equilibria for the
[31] Amine M, Awad EN, Ibrahim V, Barakat Y. Effect of ethyl acetate addition on phase systems toluene – n-heptane, toluene – propan-2-ol, toluene – sulfolane and
stability, octane number and volatility criteria of ethanol-gasoline blends. Egypt J propan-2-ol – sulfolane. J Chem Eng Data 1979;3:195–9.
Petrol 2017. https://doi.org/10.1016/j.ejpe.2017.08.007. [61] Yiying L, Xie R, Fan Z, Zan J. Isothermal vapor-liquid equilibrium for n-heptane
[32] Ratcliff MA, Luecke J, Williams A, Christensen E, Yanowitz J, Reek A, et al. Impact and methylbenzene. Huaxue Gongcheng Xi’an 1983;6:42–50.
of higher alcohols blended in gasoline on Light-Duty vehicle exhaust emissions. [62] Segovia JJ, Martin MC, Chamorro CR, Montero EA, Villamanan MA. Excess ther-
Environ Sci Technol 2013;23:13865–72. modynamic functions for ternary systems containing fuel oxygenates and sub-
[33] The Calculation of the Vapour Pressures of Aviation Fuels. The British Petroleum stitution hydrocarbons 2. Total-pressure data and GE for methyl ter-butyl ether/n-
Company Limited, BP Research Centre, Petroleum Division, Sunbury-On-Thames, heptane/1-hexene at 313.15 K. Fluid Phase Equilib 1998;2:265–76.
Technical Memorandum No, 110.040, 20 June 1960. [63] McCormick RL, Fioroni G, Fouts L, Christensen E, Yanowitz J, Polikarpov E, et al.
[34] Maxwell JB. Data book on hydrocarbons. 1st ed., Van Nostrand, New Jersey. First George A Selection criteria and screening of potential biomass-derived streams as
Edition; 1950. fuel blendstocks for advanced spark-ignition engines. SAE Int J Fuels Lubr
[35] Edmister WC. Applied Hydrocarbon Thermodynamics. Gulf Publishing Company. 2017;2:442–60.
Volume 1, 1961. [64] Monroe E, Gladden J, Albrecht KO, Bays JT, McCormick R, Davis RW, et al.
[36] Eggertsen FT, Nygard NR, Nickoley LD. Estimation of the vapor pressure of pet- Discovery of novel octane hyperboosting phenomenon in prenol biofuel/gasoline
roleum distillate fractions from gas chromatographic data. Anal Chem blends. Fuel 2019;1:1143–8.
1980;13:2069–72. [65] Song W, Rossky PJ, Maroncelli M. Modeling alkane + perfluoralkane interactions
[37] Wilhelm A, Prausnitz JM. Vapor pressures and saturated-liquid volumes for heavy using all atom potentials: Failure of the usual combining rules. J Chem Phys
fossil fuel fractions from a perturbed hard-chain equation of state. Fuel 2003;17:9145–62.
1985;4:501–8. [66] Moucka F, Nezbeda I. Water-methanol mixtures with non Lorentz-Berthelot com-
[38] Furey RL, Volatility characteristics of gasoline-alcohol and gasoline-ether fuel bining rules: a feasibility study. J Mol Liq 2011;1:47–51.
blends. International Fuels and Lubricants Meeting and Exposition, October 21, [67] Chremos A, Papaioannou V, Lafitte T, Galindo A, Jackson G, Adjiman CS, SAFT-γ
1985-October 24, 1985. Tulsa, OK, United states: SAE International; 1985. group contribution methodology based on Mie segments for branched alkanes.
[39] Chen KC, DeWitte K, Cheng WK, A species-based multi-component volatility model 2014, unpublished.
for gasoline. SAE Technical Paper Series Warrendale, PA: SAE International; 1994. [68] Yaws CL. The Yaw’s handbook of vapor pressures: Antoine coefficients. Saint
[40] Gharbi RB, Elsharkawy AM, Karkoub M. Universal Neural-Network based model Louis: Elsevier Science & Technology; 2015.
for estimating the PVT properties of crude oil systems. Energy Fuels [69] DIPPR 801, published by the Design Institute for Physical Properties (DIPPR) of
1998;13:454–8. American Institute of Chemical Engineers (AIChE), contains experimental data,
[41] Albahri TA. Specific gravity, RVP, octane number, and saturates, olefins, and estimated values, temperature-dependent correlation coefficients, references,
aromatics fractional composition of gasoline and petroleum fractions by neural notes, quality codes, and other information.
network algorithms. Petrol Sci Technol 2014;10:1219–26. [70] Marongiu B, Dernini S, Polcaro M. Thermodynamics of binary mixtures containing
[42] Adlene B, Cherif BA, Souheil S, Souheib N, Nawel O, Abdeslam-Hassen M, cyclic alkanones.1. Excess enthalpies of cyclopentanone and cyclohexanone + n-
Optimization of model parameters for gasoline properties predictions. The 8th alkanes, + cyclohexane, + benzene, and + tetrachloromethane. J Chem Eng Data
International Renewable Energy Congress; 2017. 1986;2:185–9.
[43] Gaspar DJ, Phillips SD, Polikarpov E, Albrecht KO, Jones SB, George A, Landera, A, [71] Gao J, Zhang K, Xu D, Zhang L, Chen N, Li C. Isobaric vapor-liquid equilibrium for
Santosa DM, Baldwin AG, Bays JT, Measuring and predicting the vapor pressure of binary systems of cyclohexanone + benzene, cyclohexanone + toluene, and cy-
gasoline containing oxygenates. 2019;1:630-644. clohexanone + p-Xylene at 101.3 kPa. J Chem Eng Data 2017;7:1948–54.
[44] Perdomo FA, Gil-Villegas A. Predicting thermophysical properties of biodiesel fuel [72] Vittal Prasad TE, Raju AN, Sriram N, Prasad DHL. Bubble temperature measure-
blends using the SAFT-VR approach. Fluid Ph Equilibria 2011;1:124–8. ments on the binary mixtures formed by decane with a variety of compounds at
[45] Oliveira MB, Freitas SVD, Llovell F, Vega LF, Coutinho JAP. Development of simple 95.8 kPa. Fluid Ph Equilibria 2005;1:37–9.
and transferable molecular models for biodiesel production with the soft-SAFT [73] Process Systems Enterprise, gPROMS, www.psenterprise.com/gproms,
equation of state. Chem Eng Res Des 2014;12:2898–911. 1997–2018.
[46] Perdomo FA, Millan BM, Aragon JL. Predicting the physical-chemical properties of [74] Fuel Property Database, https://fuelsdb.nrel.gov/fmi/webd/
biodiesel fuels assessing the molecular structure with the SAFT-γ group contribu- FuelEngineCoOptimization.
tion approach. Energy 2014;1:274–90. [75] Doohyun K, Martz J, Violi A. A surrogate for emulating the physical and chemical
[47] Papaioannou V, Lafitte T, Avendano C, Adjiman CS, Jackson G, Muller EA, et al. properties of conventional jet fuel. Combust Flame 2014;161:1989–11498.
Group contribution methodology based on the statistical associating fluid theory [76] Muller CJ, Canella WJ, Bruno TJ, Bunting B, Dettman HD, Franz JA, et al.
for heteronuclear molecules formed from Mie segments. J Chem Phys 2014;5. Methodology for formulating diesel surrogate fuels with accurate compositional,
054107-1-054107-29. ignition-quality, and volatility characteristics. Energy Fuels 2012;26:3284–303.
[48] Dufal S, Papaioannou V, Sadeqzadeh M, Pogiatzis T, Chremos A, Adjiman CS, et al. [77] Lemaire R, Faccinetto A, Therssen E, Ziskind M, Focsa C, Desgroux P. Experimental
Prediction of thermodynamic properties and phase behavior of fluids and mixtures comparison of soot formation in turbulent flames of diesel and surrogate fuels.
with the SAFT-γ Mie group-contribution equation of state. J Chem Eng Data Proc Combust Inst 2009;32:737–44.
2014;10:3272–88. [78] Ahmed A, Goteng G, Shankar VSB, Al-Qurashi K, Roberts WL, Sarathy SM. A
[49] Gil-Villegas A, Galindo A, Whitehead PJ, Mills SJ, Jackson G, Burgess AN. computational methodology for formulating gasoline surrogate fuels with accurate
Statistical associating fluid theory for chain molecules with attractive potentials of physical and chemical kinetic properties. Fuel 2015;143:290–300.
variable range. J Phys Chem Phys 1997;10:4168–86. [79] Acree Jr WE, Chickos JS. Phase transition enthalpy measurements of organic and
[50] Barker JA, Henderson D. Perturbation theory and Equation of State for fluids. II. A organometallic compounds. In: Linstrom PJ, Mallard WG, editors. NIST Chemistry
successful theory of liquids. J Chem Phys 1967;11:4714–21. WebBook, NIST standard Reference Database Number 69. Gaithersburg MD:
[51] Wertheim MS. Fluids with highly directional attractive forces I. Statistical ther- National Institute of Standards and Technology; 2018.
modynamics. J Statist Phys 1984;1–2:19–34. [80] Today in Energy: Almost all U.S gasoline is blended with 10% ethanol, http://
[52] Wertheim MS. Fluids with highly directional attractive forces. II. Thermodynamic www.eia.gov/todayinenergy/detail.php?id=26092; 2017 [accessed 10 August
perturbation theory and integral equations. J Statist Phys 1984;1–2:35–47. 2019].
[53] Wertheim MS. Fluids with highly directional attractive forces. I. Multiple attrac- [81] Mac Dowell N, Llovell F, Adjiman CS, Jackson G, Galindo A. Modeling the fluid
tion sites. J Statist Phys 1986;3–4:459–76. phase behavior of carbon dioxide in aqueous solutions of monoethanolamine using
[54] Wertheim MS. Fluids with highly directional attractive forces II. Equilibrium transferable parameters with the SAFT-VR approach. Ind Eng Dhem Res
polymerization. J Statist Phys 1986;3–4:477–92. 2010;4:1883–99.
[55] Goral M, Asmanova N. Vapor-liquid-equilibria in nonpolar mixtures. Part1. 2,2,4- [82] Shiohama Y, Ogawa H, Murakami S. Molar excess enthalpies of cis-decalin +
trimethylpentane with Benzene, Toluene, o-xylene, p-xylene, Ethylbenzene and benzene, + toluene, + isooctane, and + heptane at 298.15 K. Fluid Ph Equilibria
10
A. Landera, et al. Fuel 283 (2021) 118624
1987;3:249–60. the excess enthalpies and heat capacities of alkane systems. J Chem Soc, Faraday
[83] Wilhelm E, Inglese A, Grolier J-PE. Excess enthalpies of binary mixtures of cy- Trans 1979;1(75):1700–7.
cloheptane or cyclooctane with benzene or toluene or ethylbenzene at 298.15 K. [96] Letcher TM, Scoones BWH. The excess enthalpies of bicyclohexyl + a cycloalkane
Thermochim Acta 1993;1:271–80. and + an alkane at two temperatures. J Chem Thermodynamics 1982;8:703–6.
[84] Hsu K-Y, Clever L. The excess enthalpies of the 15 binary mixtures formed from [97] Ewing MB, Marsh KN. Excess gibbs free energies, excess enthalpies, and excess
cyclohexane, benzene, toluene, 1,4-dimethylbenzene, 1,2,4-trimethylbenzene, volumes for 2,3-dimethylbutane+cyclo-octane. J Chem Thermodynamics
1,3,5-trimethylbenzene at 298.15 K. J Chem Thermodynamics 1975;5:435–42. 1973;5:651–7.
[85] Watson AEP, McLure IA, Bennett JE, Benson GC. Excess properties of some aro- [98] Ewing MB, Levien BJ, Marsh KN, Stokes RH. Excess ethalpies, excess volumes, and
matic-alicyclic systems. I. Measurements of enthalpies and volumes of mixing. J excess gibbs free energies for mixtures of cyclo-octane+cyclopentane at 288.15,
Phys Chem 1965;8:2753–8. 298.15, and 308.15 K. J Chem Thermodynamics 1970;5:689–95.
[86] Woycicki W. Enthalpies of mixing of some binary mixtures benzene, cyclohexane, [99] Kimura F, Benson GC, Halpin CJ. Excess enthalpies of binary mixtures of n-heptane
and pyridine, and their methyl and ethyl derivitives. J Chem Thermodynamics with hexane isomers. Fluid Phase Equilib 1983;3:245–50.
1972;1:1–8. [100] Lifi M, Munoz-Rujas N, Montero EA, Chhiti Y, Aguilar F, Ezzahrae F, et al.
[87] Elkabule AS, Whitman DL, Miller RC. Excess molar enthalpies of (toluene + n- Measurement and modeling of excess molar enthalpies of binary mixtures invol-
heptane) and (toluene + methylcyclohexane) using a new variable-volume dilu- ving hydrocarbon components of fuel. J Chem Eng Data 2020;2:717–24.
tion calorimeter. J Chem Thermodynamics 1988;5:615–20. [101] Kretschmer CB, Wibbe R. Liquid-Vapor equilibrium of ethanol-toluene solutions. J
[88] Mathieson AR, Thynne JCJ. Thermodynamics of hydrocarbon mixtures. Part II. Am Chem Soc 1949;1:1793–7.
The heats of mixing of the binary mixtures formed by benzene, cycloHexane, n- [102] Aguilar F, Alaoui FEM, Segovia JJ, Villamanan MA, Montero EA. Ether + alcohol
heptane, toluene, and n-hexane. J Chem Soc 1956:3708–13. + hydrocarbon mixtures in fuels and biofuels: excess enthalpies of binary mix-
[89] Wang Z, Benson GC, Lu BCY. Excess molar enthalpies of binary mixtures of 1- tures containing dibutyl ether (DBE) or 1-butanal and 1-hexene or methylcyclo-
hexene with some cyclic and aromatic hydrocarbons at 298.15 K. Thermochim hexane or toluene or cyclohexane or 2,2,4-trimethylpentane at 298.15 and 313.15
Acta 2004;1:31–3. K. Fluid Phase Equilib 2012;1:1–8.
[90] Arenosa RL, Menduina C, Tardajos G, Diaz-Pena M. Excess enthalpies at 298.15 K [103] Moreau A, Carmen-Martin M, Chamorro CR, Segovia JJ. Thermodynamic char-
for binary mixtures of toluene + n-alkane. J Chem Thermodynamics acterization of second generation biofuels: vapor-liquid equilibria and excess en-
1979;11:825–8. thalpies of the binary mixtures 1-pentanol and cyclohexane or toluene. Fluid Phase
[91] Fujihara I, Kobayashi M. Excess enthalpies of trans-decalin + benzene, + toluene, Equilib 2012;1:127–31.
+ isooctane, and + heptane at 298.15 K. J Chem Thermodynamics 1983;1:1–6. [104] Singh KC, Kalra KC, Maken S, Gupta V. Excess heat of mixing of 1-propanol or 2-
[92] Fujii A, Hayashi H, Park JW, Kaxama T, Mikami N, Tsuzuki S. Experimental and propanol with benzene, toluene, o-, m-, and p-xylenes at 298.15 K. Thermochim
theoretical determination of the accurate CH/π interaction energies in benzene- Acta 1996;1:271–85.
alkane clusters: correlation between interaction energy and polarizability. Phys [105] Hsu K, Clever HL. Enthalpy of mixing of some hydrocarbon-alcohol and hydro-
Chem Chem Phys 2011;13:14131–41. carbon-nitroalkane mixtures at 25 oC. J Chem Eng Data 1975;3:268–71.
[93] Ran J, Wong MW. Saturated hydrocarbon-benzene complexes: theoretical study of [106] Pedersen MJ, Kay WB, Hershey HC. Excess enthalpies, heat capacities, and excess
cooperative CH/π interactions. J Phys Chem A 2006;31:9702–9. heat capacities as a function of temperature in liquid mixture of ethanol + to-
[94] Islam N, Ibrahim M. Excess volumes and excess viscosities of binary liquid systems: luene, ethanol + hexamethyldisiloxane, and hexamethyldisoloxane + toluene. J
Cyclohexane + toluene, amyl alcohol + toluene, and ethylene glycol + pyridine. Chem Thermodynamics 1975;7:1107–18.
Indian J Chem 1982;1:1026–30. [107] Zeberg-Mikkelsen C, Andersen SI. Density measurements under pressure for the
[95] St De Romain P, Van HT, Patterson D. Effects of molecular flexibility and shape on binary system 1-propanol + toluene. J Chem Eng Data 2005;2:524–8.
11