Applied Thermal Engineering 73 (2014) 539e548

Contents lists available at ScienceDirect

Applied Thermal Engineering

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

An equation for the estimation of alcohol-air diffusion coefficients for

modelling evaporation losses in fuel systems
Magín Lapuerta a, *, Juan Pablo Herna

ndez b, John R. Agudelo b
a
Escuela T

ecnica Superior de Ingenieros Industriales, Universidad de Castilla La-Mancha, Avda. Camilo Jos

e Cela s/n, 13071 Ciudad Real, Spain
b
Departamento de Ingeniería Meca
nica, Universidad de Antioquia, Calle 70, No. 52-21, Medellín, Colombia

h i g h l i g h t s

Diffusion coefficients are needed for predicting evaporation losses in fuel systems.

Nine equations were applied to the binary gaseous alcohol-air mixtures.

Experimental results were obtained at temperatures 25e50  C for C1 to C5 alcohols.

Results were compared with data from literature and from equations predictions.

A new four-parameter equation is proposed showing good agreement with experiments.

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

Article history: Alcohols (not only ethanol but also methanol, propanol, butanol and pentanol) are receiving increasing
Received 7 May 2014 attention as components for blending with conventional fuels (gasoline and diesel or biodiesel fuels)
Accepted 2 August 2014 because they can be an efficient means to increase the renewable fraction of fuels and to reduce emis-
Available online 9 August 2014
sions. Mathematical models for predicting evaporation losses in storage and fuel distribution systems
require information about diffusion coefficients, among other characteristic parameters of the alcohols
Keywords:
and of the alcohol-air gaseous mixtures. Nine equations in the literature have been described and applied
Evaporation losses
to the binary gaseous mixtures alcohol-air to estimate the diffusion coefficient. Other equations were
Fuel tanks
Alcohols
discarded because they are not applicable to alcohol-air systems. The resulting diffusion coefficients were
Biofuels plotted against temperature for the five mentioned alcohols, and the results were compared with
Diffusion coefficient experimental data obtained from measurements with capillary glass tubes immersed in a controlled
temperature glycerine bath at 25  C, 30  C, 40  C and 50  C. The diffusion coefficients were determined
from the variation of the height of the liquid column and the vapour pressure. From comparing the
standard deviations with respect to the experimental data, although the best fit for all alcohols was
reached with the Arnold equation (0.0247 cm2/s), it was concluded that there is not a unique optimal
formula for all alcohols. Additionally, all the revised equations underestimate the effect of temperature
for methanol, ethanol and propanol. Specific logarithmic equations were proposed in this work for
estimating the diffusion coefficient for each alcohol, and finally, a four-parameter equation was also
proposed for the five alcohols studied, which provided excellent agreement (standard deviation
0.0045 cm2/s).
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Alcohols are receiving increasing interest as fuels for various
reasons. On the one hand, new processes are being developed for
the production of alcohols from renewable feedstocks [1].
* Corresponding author. Tel.: þ34 926295431; fax: þ34 926295361.
E-mail address: Magin.Lapuerta@uclm.es (M. Lapuerta).
Therefore, blends of gasolines or diesel fuels (or even diesel bio-
diesel blends) with alcohols are becoming one of the most efficient
means to introduce a renewable fraction in the fuel, to improve its
sustainability, to reduce its viscosity [1,2] and to provide some
oxygen content, which will be essential to reduce soot emissions
from spark ignition [3,4] or diesel engines [5e7]. However, different
challenges will have to be faced for the future extension to the
large-scale use of these alcohols as fuel components. One of these
challenges is to address the fuel storage and handling difficulties

http://dx.doi.org/10.1016/j.applthermaleng.2014.08.009
1359-4311/© 2014 Elsevier Ltd. All rights reserved.
540 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548
derived from the high volatility of alcohols. Previous experiences
with ethanol blends with either gasoline or diesel fuels have proved
that different technical modifications in the fuelling system are
necessary, such as the installation of seals for fuel tanks, flame ar-
resters, security valves on both vehicle fuel tanks and fuel station
tanks, special systems to recover fuel vapours on vehicles and fuel
stations and to avoid vapour locks, etc. [8,9]. These restrictions
constitute a barrier for the massive use of these blends in private
vehicles, and have limited their practical use to captive fleets, such
as those of urban buses or public construction vehicles and civil
machinery.
The design of the most adequate storage and fuelling systems
for each blend requires detailed knowledge about the evaporation
losses from the liquid phase and the diffusion along the gas phase.
Diffusion models often assume constant values for the diffusion
coefficient neglecting the effect of temperature and pressure [10],
or use polynomial fits as a function of temperature [11]. One of the
most challenging points in this modelling process is the selection of
the equation for estimating the diffusion coefficient alcohol-air.
This work reviews the equations reported in the literature and
proposes their application to different alcohol-air mixtures, the
alcohols ranging from methanol to pentanol. This is a first step for
the modelling of the evaporation losses from alcohols in tanks, lines
and nozzles and their diffusion into habitable spaces.
The first approach for the estimation of the diffusion coefficient
was proposed by Maxwell and Stefan [12] who used previous
classical works by Meyer [13] and Jeans [14]. One of the key studies
is the theory proposed by Chapman and Enskog, as cited in [12],
who solved the Bolztmann equation. Based on this theory, different
approaches have been proposed to solve the diffusion coefficient.
Among these, the methods proposed by Chapman and Enskog [12]
and Wilke and Lee [15] make use of the collision integral, whereas
other equations do not include this parameter, either because they
neglect the molecular interactions between the mixed gases or
because they have an empirical origin. The equations reported in
the literature for the determination of the diffusion coefficient were
fitted to a variety of blends of inorganic and organic compounds
[15e32], none of them being specific for a family of substances.
Additionally, there is a wide dispersion among the experimental
values reported in the literature for the diffusion coefficient of C1 to
C5 alcohols with air, which may be attributed to differences in the
experimental methods [33]. Moreover, these results are scarce for
temperatures above 25  C and for alcohols other than n-butanol y
ethanol [18,19].
This motivated the present study, where the equations pro-
posed in the literature are compared with experimental data for
diffusion coefficients reported for C1 to C5 linear alcohols, and
with experimental results obtained in this work from the variation
of the liquid height and the vapour pressure of the alcohols in a
capillary glass tube subjected to different temperatures. The final
objective of this study is the selection of an existing equation or
the proposal of a new one for the determination of the diffusion
coefficient, with the aim to be applied to the estimation of alcohol
evaporation losses in fuel systems operating under atmospheric
conditions.
2. Diffusion of alcohols in air
The diffusion coefficient is defined as the ratio between the
mass (or molar) flux of a component in a gaseous mixture and its
concentration gradient, in absence of mechanical or convective
forces. This coefficient is necessary to mathematically formulate the
Fick's law, which expresses the linear relationship between diffu-
sion flux and concentration gradient [34,35]. If this is applied to an
alcohol (A) in a mixture of alcohol vapour (coming from
evaporation) and air (AIR), inside a vessel with no circulation of air,
the Fick's law can be expressed as:
00 00 dYA
m_ A ¼ YA $m_ A  r$DAAIR (1)
dx
00
with m_ A being the mass flow rate per unit area (with units kg/s m2)
of alcohol in the vertical direction, x, and YA being the alcohol mass
fraction in the vapor phase. The first term in Eq. (1) represents
the convective transport whereas the second represents the diffu-
sion transport, with DAAIR being the binary diffusion coefficient
(in m2/s).
Diffusion transports alcohol mass from the interface towards the
open section of the fuel tank or the security valve. Right after the
tank filling or after any increase in temperature, diffusion is
enhanced because there is no alcohol content in the vessel, the
gradient of concentration being maximal. When the quasi-steady
regime is finally reached, the mass of alcohol transported from
the interface almost stabilizes providing steady alcohol concen-
tration profiles, which permit to estimate the evaporation loss rate.
To build up a mathematical model for predicting the alcohol vapor
transportation, information about the mass fraction of alcohol at
the liquidevapor interface is needed, for which properties of the
alcohol (vapor pressures) and of the binary gaseous mixture (ac-
tivity coefficients) are also needed. Especially in the case of alcohol-
diesel (or biodiesel) blends, vapor pressure of the latter component
can be neglected as a consequence of its low volatility. After the
boundary conditions at the interface being determined, diffusion
can be calculated.
For the determination of the diffusion coefficients of alcohols in
their binary alcohol-air gaseous mixtures different approaches
have been proposed, which are revised below. Some of them
include the collision integral.
3. The collision integral
The collision integral (UD ) accounts for the interactions between
molecules of the gases being mixed and it depends on temperature.
To simulate the intermolecular forces, molecular potential energies
must be modelled, either following the Lennard-Jones potential
function [36] or the Stockmayer one [37]. From the potential
function, an expression can be derived for the collision integral.
This would be equal to unity when molecules are considered as
rigid spheres.
The use of the mentioned potential functions led to the proposal
by Hirschfelder et al. [38] of a table relating the non-dimensional
ratio between the product of the Boltzmann constant and the
temperature, and the maximum attraction energy, kT/εAAIR, with a
function associated with the intermolecular forces, Hirschfelder
function, which can be identified as half of the collision integral.
Later on, Neufeld proposed the following equation [39,40] to esti-
mate the collision integral, UD. This equation accurately fitted the
data listed in Hirschfelder's table, multiplied by two:
1:06036 0:193 1:03587
UD ¼
0:1561 þ
þ

kT kT
exp 0:47635 εAAIR kT
exp 1:52996 εAAIR
εAAIR
1:76474
þ

kT
exp 3:89411 εAAIR
(2)
For the determination of ε/k, Eq. (3) was used for each compo-
nent, with Tb being the normal boiling temperature in K. These
results are then used for obtaining the average value for each
 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548 541

alcohol, εAAIR/k, by means of Eq. (4). In diffusion systems involving where sAAIR is the mean molecular radius of the mixing gases,
air, the following values are usually considered for air: s which can be estimated as:
(air) ¼ 3.62 Å, and ε/k (air) ¼ 97.0 K.
sA þ sAIR
ε sAAIR ¼ (9)
¼ 1:15Tb (3) 2
k
For each particular gas, the molecular radius is considered to be
εAAIR εA εAIR 1=2 proportional to the specific volume of the saturated liquid at the
¼ (4)
k k k boiling temperature, vb. Therefore:
A slightly different fitting of Hirschfelder's table was proposed 1=3
by Chen [41], which led to the following equation for the collision s ¼ 1:18vb (10)
integral:
1:075 2 4.3. Chen-Othmer equation
UD ¼
0:1615 þ
0:74 log10 10ε kT
(5)
kT AAIR
kT
10 εAAIR This equation simplifies the calculation process by integrating
εAAIR
the collision integral as a function of the critical temperature (see
Differences obtained between Neufeld and Chen equations Eq. (5)) in the equation for the determination of the diffusion co-
would remain below 1% in case of using the same method for the efficient [16]:
estimation of ε/k. However, Chen and Othmer [16] proposed to use
1:81
0:5
the latter equation with the ε/k function being estimated from the 0:43 T 1 þ W1AIR
100 WA
critical temperature, rather than the boiling temperature: DAAIR ¼ (11)

0:1405
ε p T10000
cA TcAIR
s2AAIR
¼ 1:276Tc0:9061 (6)
k
Since the equations described above were derived for nonpolar In this case, parameter sAAIR is obtained from the critical vol-
gases, a correction was proposed by Browak [42]. This correction is umes (differently to the WilkeeLee equation, where it is obtained
not described here because air is a nonpolar gas, and the correction from the specific volume at boiling temperature) and the exponent
is therefore nil. is experimentally fitted to 0.4 [16]:
v 0:4 v 0:4
cA cAIR
4. Equations of the diffusion coefficient sAAIR ¼ þ (12)
100 100
4.1. ChapmaneEnskog theory
4.4. Fuller-Schettler-Giddings equation
This theory is derived from the solution given to the Boltzmann
equation independently obtained by Chapman and Enskog [12]. This equation assumes a) the hypothesis of rigid spheres
This is applicable to binary mixtures of gases at low to moderate (UD ¼ 1), b) that the collision diameters are proportional to the
pressure, but with sufficient molecular mean free path to neglect cubic root of the molecular volumes, and c) the additivity of atomic
the molecular collisions with the container surface [33]. From this volumes, v*, which allows to estimate the molecular volumes from
theory, the binary diffusion coefficient for an alcohol-air system the atomic volumes of the elements composing the diffusing
would be: molecules [18,19], both molecular and atomic volumes being
quantified in liter per kmol. These hypotheses lead to the following

0:5 expression [20]:
0:001858 T 3=2 1
WA þ W1AIR

0:5
DAAIR ¼ (7)
ps2AAIR UD 0:001T 1:75 1
WA þ W1AIR
DAAIR ¼ h (13)
The equation proposed by Hirschfelder, Bird and Spotz [17] is P * 1=3 P * 1=3 i2
p vA þ vAIR
identical to the ChapmaneEnskog equation, the only difference
being that they use the Hirschfelder function (through the data
In the case of air-alcohol mixtures, the involved atomic volumes
provided in the Hirschfelder's table) instead of the collision
are shown in Table 1. In this table air is included as a pseudo-atom.
integral.

4.5. Arnold equation

4.2. Wilke-Lee equation
Sutherland proposed a model for the determination of the vis-
From experiments with different gas-vapor mixtures at pres-
cosity of gases, which can be considered as a particular case of the
sures and temperatures around atmospheric conditions Wilke and
ChapmaneEnskog model [12]. He defined the Sutherland constant
Lee [15] proposed the following equation, derived from the
(given in Kelvin) to correct the effect (previously underestimated)
Hirschfelder-Bird-Spotz equation. Therefore, this equation can be
of intermolecular forces on the collision integral and thus on the
used for low reduced pressure and temperature conditions. After
substituting the Hirschfelder function by the collision integral, the
equation can be written as: Table 1

1=2 
1=2 Atomic volumes of elements involved in air-alcohol diffusion problems. The value
2:104 10:7  2:46 WWAAþW AIR
WAIR T 3=2 WA þWAIR
WA WAIR
given for air corresponds to its mean molecular volume.
DAAIR ¼ Atom C H O Air
ps2AAIR UD
v* (L/kmol) 15.9 2.31 6.11 19.7
(8)
542 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548

diffusion coefficient. For a pure gas he proposed to estimate this 4.8. Huang equation
effect as proportional to the boiling temperature:
From new experiments for different ambient pressures and
S ¼ 1:47Tb (14) temperatures, Huang revised the Arnold and Gilliland equations,
For a mixture of gases, such as that of an alcohol with air, he and re-adjusted the equation by including an exponent for pressure
proposed the following mixing rule: higher than unity, increased the exponent for temperature (turning
back to that proposed by Fuller-Schettler-Giddings) and modified
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SAAIR ¼ 1:47F TbA TbAIR (15) the prefactor [22]. The proposed equation was, after converting to
SI units:
The data provided for F have been adjusted here to a polynomial

0:5
equation, as a function of the molar volumes at saturated liquid
conditions, leading to: 0:0009986T 1:75 1
WA þ W1AIR
DAAIR ¼ h i (21)


 1=3 1=3 2
vbA 3 vbA 2 p1:286 vbA þ vbAIR
SAAIR ¼ 1:47 0:0000747  0:000323
vbAIR vbAIR
! 4.9. Othmer-Chen equation
v pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 0:0266 bA þ 1:03 TbA TbAIR (16)
vbAIR
The same authors who proposed the Chen-Othmer equation
(see Section 4.3) proposed, nine months later, a modified equation
Arnold adopted this parameter for the estimation of the diffu-
for the diffusion coefficient in which the viscosity of air is included
sion coefficient [18]. Since the correction provided for intermolec-
[23]. This equation is only acceptable for air systems, and should be
ular forces depend on temperature, the adjustment to experimental
corrected in other cases where air is not the main gas. They pro-
data of diffusion coefficients led to a modification of both the
posed that the effect of the intermolecular forces is physically
prefactor and the exponent of temperature. Finally, Arnold pro-
included in the viscosity of air. In fact, the latter can also be
posed to estimate the molecular volumes from the molar volumes
calculated as a function of the Sutherland constant. Based on their
of the mixing components at saturated liquid conditions:
previous experimental tests, they kept the exponent for the specific

0:5 volume in 0.4:
0:00837T 1:5 þ W1AIR1
2
0:5 3
WA
DAAIR ¼ h i
 (17) 1 þ W1AIR
1:23
1=3 1=3 2
p vbA þ vbAIR 1 þ SAAIR 2:52,107 m2:74
AIR 6 WA
7
T DAAIR ¼ 4 2 5 (22)
p
v0:4
bA
þ v0:4
bAIR
For the determination of saturated volumes, in case of air, the
experimental value (33.33 L kmol1) is proposed here, whereas for Consistently with the approach by Othmer and Chen, for the
alcohols, the equation proposed by Tyn and Calus [40], as a function determination of the viscosity of air (in cP), the equation proposed
of the critical volume, is proposed: by Sutherland is recommended here (including the Sutherland
constant, which can be estimated from the boiling temperature as
vbA ¼ 0:285v1:048
cA (18) shown above):

4.6. Gilliland equation T 1:5

mAIR ¼ 0:0014972 (23)
T þ SAIR
Gilliland proposed a simplification of the Arnold equation by
including the thermal effect of the Sutherland constant in the 4.10. Other methods to determine the diffusion coefficient
prefactor of the diffusion equation [19]. Consequently, the Gilliland
equation (proposed to be applied to mixtures of gas and organic Some other methods have been used for the determination of
vapours) eliminates the explicit effect of the Sutherland constant the diffusion coefficient. All of them propose relations with pres-
and modifies the prefactor with respect to the Arnold equation. sure and temperature, but use different parameters related to the

0:5 gas components. Slattery et al. [24], proposed an equation based on
0:0043T 1:5 1 þ W1AIR the critical properties of the components, by least square regres-
WA
DAAIR ¼ (19) sion. This equation is not displayed here because the authors
h i
1=3 1=3 2 excluded it for mixtures including air, hydrogen, helium and water.
p vbA þ vbAIR
Hashim et al. [25] proposed to calculate the diffusion coefficient as
a function of the molecular weight and the Pitzer's acentric factor
4.7. SingheSingh equation (trying to simulate deviations with respect to the law of corre-
sponding states). However, the errors obtained in some systems
Singh and Singh proposed a different re-adjustment of the with this equation reach up to 38%. Eslamlouyan et al. [26] esti-
prefactor and exponent of the diffusion equation [21], trying to mated the diffusion coefficients with a neural network. In all three
improve the simulation of the effect of the Sutherland constant, cases, the proposed correlations avoid using the collision integral or
which, as Arnold did, was also explicitly eliminated. The proposed the Sutherland constant, and consequently they can be considered
equation was: mathematical methods rather than phenomenological ones.
Some other equations are biased by the experimental method

0:5
1 used for validation. A review of these experimental methods was
0:00279T 1:622 WA þ W1AIR
presented by Marrero et al. [33]. Karaiskakis et al. [27] reviewed the
DAAIR ¼ h i (20) equations proposed for the diffusion coefficient when chromatog-
1=3 1=3 2
p vbA þ vbAIR raphy methods are used for experimental validation, and showed
 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548 543

that the equations depend on the specific chromatographic method
used, and therefore they should not be used from the component
properties. Elliott et al. [28] proposed another equation to be used
for mixtures of hydrocarbons in air. The LeonardeJones potential
function used was determined from viscosity measurements for
such hydrocarbons. The proposed equation includes two specific
constants depending on the class of hydrocarbons (alkanes and
alkenes), but there is no proposal for alcohols.
Among the equations studied here, six of them (Arnold, Gilli-
land, WilkeeLee, Fuller-Shettler-Giddings, CheneOthmer and
OthmereChen) were included in the set of methods which Lugg
[32] compared with his experimental results. He measured the
diffusion coefficient of 147 organic compounds and other vapours
into air at 25  C and 1 atm, by determining the rate of fall of the
meniscus level of the fluid contained in a uniform base glass tube
subjected to a conditioned air stream at a rate of 100 mL min1. The
equations which fitted best to the experimental results obtained for
light alcohols (C1 to C5) were those by CheneOthmer (CeO) and
the WilkeeLee (W-L), which in most of the cases provided results
within 5% differences with respect to his measured values. He also
found that the equation of Hirschfeider-Bird-Sportz fitted better for
heavier alcohols.
where h is the height from the liquid interface to the top of the glass
capillary (diffusion path) and t  t0 is the time from the start of the
experiment in seconds. In all cases, the measurements started 24 h
after reaching a stable height. From this time onwards, measure-
ments were taken every 24 h for a period between 8 and 15 days for
all alcohols. In all cases, heights were measured with a sliding
gauge with an accuracy of 0.05 mm. This, together with the accu-
racies of the temperature and pressure sensors, led to the experi-
mental uncertainties presented below for all alcohols tested. The
mean diffusion coefficient was determined from the regression line
obtained when the height measurements were plotted in front of
the measurement times.
In order to compare the results obtained (at 86 kPa) with those
reported in the literature at the same temperature (in all cases at
101.325 kPa), the following equation has been used to convert the
diffusion coefficient results [32] (temperature was not varied for
this conversion):


  T2 n p2
DAAIR p1 ; T1 ¼ DAAIR p1 ; T2 (25)
T1 p1
6. Results and discussion

5. Experimental 6.1. Results from the equations proposed in the literature

A 1.5 L beaker was filled with glycerin (99.9% purity) up to 2 cm Previously to the diffusion coefficients, the resulting values for
below the top. Also, five thin glass tubes with 20 cm height and the collision integral are shown for different alcohol-air systems.
5 mm internal diameter were placed vertically into the beaker. The differences observed in Fig. 2 between the collision integral
They were supported with a bored acrylic plate and filled with estimated with Neufeld and Chen equations are not derived from
different alcohols leaving 4.5 cm unfilled from the liquid surface to the equations but from the different estimation of the ratio ε/k
the top of the glass tube. A heating iron with magnetic agitation was (calculated from the boiling temperature and from the critical
used to raise the temperature of the glycerin bath and an EBCHQ temperature respectively), which leads to substantial differences
58008 temperature controller was coupled to the heating device especially in the case of methanol and ethanol, with significant
(Fig. 1). The temperatures controlled in the glycerin bath were 25,
30, 40 and 50  C, and the alcohols were methanol (Honeywee 99.9%
purity), ethanol (Panreac, 99.5%), n-propanol (Merck, 99.5%), n- Neufeld equation Methanol
butanol (Panreac, 99.5%) and n-pentanol (Panreac, 98%). The envi- Neufeld equation Ethanol
Neufeld equation Propanol
ronmental conditions were controlled at 22  C ± 1  C and 45% ± 5%
Neufeld equation n-Butanol
humidity. All alcohols experienced an initial increase in the liquid
Neufeld equation n-Pentanol
level due to thermal dilatation especially at 30, 40 and 50  C, but in 1.28 Chen equation Methanol
the case of methanol, ethanol and n-propanol, such dilatation was Chen equation Ethanol
almost compensated by the evaporation, and no modification of the Chen equation Propanol
liquid level was detected. In the case of n-butanol and n-pentanol, Chen equation n-Butanol
the rate of evaporation was insufficient to compensate the dilata- Chen equation n-Pentanol
tion and an increase was observed in the liquid level (h0) during the 1.24
first hours. This increase was considered in the equation, proposed
Collision integral

by Erbil and Avci [43], for the determination of the diffusion co-
efficients at each temperature according:
  1.2
h2  h20 rRT
DAAIR ¼   (24)
p
2ðt  t0 ÞWA p ln pps

1.16

1.12
0 10 20 30 40 50
Temperature (ºC)

Fig. 2. Estimation of the collision integral for different alcohol-air systems from
Neufeld equation (with ε/k obtained from boiling temperature) and from Chen equa-
Fig. 1. Experimental setup for the measurement of diffusion coefficients. tion (with ε/k obtained from critical temperature).
544 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548

Experimental data Experimental data

Chapman-Enskog Chapman-Enskog Propanol-Air
Wilke-Lee Methanol-Air Wilke-Lee
Chen-Othmer Chen-Othmer
Fuller-Schettler-Giddings Fuller-Schettler-Giddings
Arnold Arnold
0.24 Gilliland 0.16 Gilliland
Singh-Singh
Singh-Singh
Huang
Huang
Othmer-Chen
Othmer-Chen

0.14

Difussion coefficient (cm2/s)

Difussion coefficient (cm2/s)

0.2

0.12

0.16
0.1

0.08
0.12 0 10 20 30 40 50
0 10 20 30 40 50 Temperature (ºC)
Temperature (ºC)
Fig. 5. Diffusion coefficients vs temperature for propanol-air mixtures from different
Fig. 3. Diffusion coefficients vs temperature for methanol-air mixtures from different equations.
equations.
differences in boiling temperature but very similar critical tem-
peratures. The Neufeld equation was used to estimate the collision
integral in Eqs. (7) and (8) because in the original literature ε/k was
primarily related to boiling temperature rather than to critical
Experimental data temperature.
Chapman-Enskog Ethanol-Air
Wilke-Lee
Experimental data
Chen-Othmer
Chapman-Enskog n-Butanol-Air
Fuller-Schettler-Giddings Wilke-Lee
Arnold Chen-Othmer
Gilliland Fuller-Schettler-Giddings
Singh-Singh Arnold
0.18 Huang Gilliland
Othmer-Chen 0.14 Singh-Singh
Huang
Othmer-Chen

0.16
Difussion coefficient (cm2/s)

0.12
Difussion coefficient (cm2/s)

0.14
0.1

0.12
0.08

0.1 0.06
0 10 20 30 40 50 0 10 20 30 40 50
Temperature (ºC) Temperature (ºC)

Fig. 4. Diffusion coefficients vs temperature for ethanol-air mixtures from different Fig. 6. Diffusion coefficients vs temperature for butanol-air mixtures from different
equations. equations.
 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548 545

Experimental data However, the selection of an equation for the diffusion coeffi-
Chapman-Enskog cient to be recommended for use in air-alcohol diffusion applica-
Wilke-Lee n-Pentanol-Air tions should be supported with experimental data.
Chen-Othmer
Fuller-Schettler-Giddings 6.2. Comparison between experimental results and those reported
Arnold
in the literature
Gilliland
0.14 Singh-Singh
Huang
The experimentally obtained diffusion coefficients for
Othmer-Chen methanol-air, ethanol-air, n-propanol-air, n-butanol-air and n-
pentanol-air systems at 86 kPa (from 25 to 50  C) were firstly
0.12 compared with those reported in the literature, after being trans-
lated to 101.325 kPa by means of Eq. (25) (Table 3). From this
Difussion coefficient (cm2/s)

comparison, it can be observed that a good agreement was ob-

tained for methanol at 25  C. No data were found in the literature
0.1
for this alcohol at other temperatures. For ethanol, the differences
with reported values decreased with increasing temperature.
Finally, the results obtained for n-propanol, n-butanol and n-pen-
0.08 tanol were significantly lower than those reported at any temper-
ature. No previous diffusion coefficient results were found at 50  C
for any alcohol. Nevertheless, for an appropriate comparison, the
differences between the experimental methods used must be
0.06 considered. The literature reports experimental methods, such as
those using gas chromatography, which are based on the broad-
ening of the chromatographic bands [22,27,33], the open-tube
0.04
evaporation method with air flow over the top end of the diffu-
sion path by forced convection [32,33] and those based on the
0 10 20 30 40 50
Temperature (ºC) Stefan tube without air flow over the top end [29,33]. The method
used affects the results obtained, which explains the differences
Fig. 7. Diffusion coefficients vs temperature for pentanol-air mixtures from different observed.
equations.

6.3. Comparison between experimental results and equation

The diffusion coefficients obtained from all the equations pre-
sented in Section 4 are compared in Figs. 3 to 7 for different alcohol
e air mixtures in a temperature range between 0 and 50  C, which
was selected as representative of real storage conditions in vehicle
tanks. The properties of air and alcohols used in these equations are
listed in Table 2.
As observed in Figs. 3e7, all equations predict almost linear
increases for the diffusion coefficient with temperature, showing
that the range selected for temperatures is not large enough to
make the potential increase evident. The equations predict values
for the diffusion coefficient in all cases with differences lower that
0.04 cm2/s, except for methanol-air mixtures where differences
may reach 0.07 cm2/s. It is also noticeable that SingheSingh
equation overestimates the diffusion coefficient with respect to the
rest of equations, except in the case of methanol and especially in
the case of large carbon-chain alcohols, and that the Othmer-Chen
equation provides higher rates of increase of the diffusion coeffi-
cient than the rest of equations, especially in the case of light
alcohols.
predictions
The experimental results obtained in this work were secondly
compared with the results obtained from the equations proposed in
the literature (Table 4 and Figs. 3e7). The experimental values
measured at 86 kPa are shown in Table 4 together with the pre-
dicted ones (from all equations). For each equation, the standard
deviation between predicted and experimental values was deter-
mined. It can be concluded that there is not a unique optimal for-
mula for all alcohols. The lowest standard deviation for methanol
was obtained with the Fuller-Schettler-Giddings equation
(0.0116 cm2/s), for ethanol with the Chapman-Enskog equation
(0.0128), for propanol, butanol and pentanol with the Gilliland
equation (0.0005, 0.0172 and 0.0305, respectively). The best fit for
all alcohols was obtained with the Arnold equation (0.0247 cm2/s).
Other equations show standard deviations higher than 0.026 cm2/s.
The worst predictions are obtained with the SingheSingh equation
(0.0401 cm2/s). In any case, none of the equations is recommended
for the determination of the diffusion coefficient for the whole
collection of alcohol-air systems. It must be considered that any

Table 2
Main properties of the pure alcohols used in the diffusion coefficient equations.

Properties Air Methanol Ethanol n-Propanol n-Butanol n-Pentanol

Molecular weight, W (kg/kmol) 28.85 32.0419 46.0684 60.0950 74.1216 88.1482

Mean molecular radius, s (Å) 3.62 4.1109 4.6422 5.1022 5.5246 5.8629
Boiling temperature, Tb (K) 78.8 337.69 351.8 370.93 390.88 411.16
Critical temperature, Tc (K) 132.164 512.64 513.92 536.78 563.05 588.15
Critical volume, vc (L/kmol) 81.32 118 167.1 219 275 326
Volume of saturated liquid at boiling 33.3 42.284 60.887 80.841 102.628 122.658
temperature, vb (L/kmol)
Ratio maximum attraction energy/ 97 388.34 404.57 426.57 449.51 472.83
Boltzmann constant, ε/k (K)
Sutherland constant, S (K) 115.836 496.404 517.146 545.267 574.594 604.405
546 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548

Table 3
Comparison of experimental data with respect to those reported in the literature.

Alcohol T( C) Experimental Derived from Eq. 25 Literature Abs. Error (%)

p ¼ 86 kPa p ¼ 101.325 kPa p ¼ 101.325 kPa

DAAIR (cm2/s) Uncertainty (cm2/s) DAAIR (cm2/s) DAAIR (cm2/s) Ref.

Methanol 0 N.A N.A 0.132 [30]

25 0.1777 0.0009 0.1529 0.152 [32] 0.56
0.1591 [29] 3.92
30 0.1840 0.0008 0.1582 e e
40 0.2194 0.0007 0.1887 e e
Ethanol 0 N.A N.A 0.102 [30]
25 0.1264 0.001 0.1087 0.135 [20] [15] 19.48
0.1318 [29] 17.52
0.1181 [32] 7.95
30 0.1288 0.0008 0.1107 e e
40 0.1636 0.0007 0.1407 0.147 [30] 4.29
42 N.A N.A 0.145 [18]
50 0.1719 0.0005 0.1478 e e
n-Propanol 0 N.A N.A 0.0818 [30]
25 0.0899 0.0016 0.0773 0.0865 [29] 10.65
0.0993 [32] 22.16
30 0.0906 0.001 0.0779 0.101 [30] 22.88
40 0.1030 0.0006 0.0886 e e
50 0.1240 0.0001 0.1067 e e
n-Butanol 0 N.A N.A 0.703 [30]
25 0.0736 0.0037 0.0633 0.0865 [29] 26.86
0.087 [19] 27.28
30 0.0749 0.0008 0.0644 e e
40 0.0815 0.0013 0.0701 0.092 [19] 23.78
50 0.0841 0.0017 0.0723 e e
n-Pentanol 0 N.A N.A 0.0589 [30]
25 0.0532 0.0011 0.0457 0.0716 [32] 36.13
30 0.0545 0.0014 0.0469 e e
40 0.0596 0.0033 0.0512 e e
50 0.0594 0.0042 0.0511 e e

N.A: Not available.

equation proposed for the diffusion coefficient intends to predict its 6.4. Proposal of an equation for the diffusion coefficient of alcohols
variation with pressure and temperature for a wide collection of
substances, which leads to inaccuracies in the prediction of the Consequently, specific linear logarithmic equations (following
coefficients for any specific substance or type of substances. the variation pattern proposed by Erbil [43], as expressed in

Table 4
Comparison between experimental and equation predictions (86 kPa). Equations are identified by initials: C-E: ChapmaneEnskog; W-L: WilkeeLee; CeO: CheneOthmer; FeS-
G: Fuller-Schettler-Giddings; A: Arnold; G: Gilliland; SeS: SingheSingh; H: Huang; OeC: OthmereChen.

Alcohol T ( C) Dexp (cm2/s) Dpredicted (cm2/s)

CeE WeL CeO FeS-G A G SeS H OeC

(Eq. (7)) (Eq. (8)) (Eq. (11)) (Eq. (13)) (Eq. (17)) (Eq. (19)) (Eq. (20)) (Eq. (21)) (Eq. (22))

Methanol 25 0.1777 0.1677 0.1817 0.1791 0.1865 0.1581 0.1471 0.1912 0.1482 0.1983
30 0.1840 0.1730 0.1875 0.1845 0.192 0.1633 0.1508 0.1965 0.1526 0.2055
40 0.2194 0.1838 0.1993 0.1957 0.2032 0.1739 0.1583 0.2071 0.1615 0.2201
Methanol standard deviation 0.0223 0.0120 0.0137 0.0116 0.0310 0.0438 0.0128 0.0416 0.0172
Ethanol 25 0.1264 0.1347 0.1467 0.1419 0.1429 0.1268 0.1195 0.1553 0.1204 0.1491
30 0.1288 0.1390 0.1514 0.1462 0.1472 0.1310 0.1225 0.1596 0.1239 0.1544
40 0.1636 0.1478 0.1609 0.1551 0.1558 0.1396 0.1286 0.1682 0.1312 0.1654
50 0.1719 0.1567 0.1707 0.1642 0.1646 0.1483 0.1348 0.1770 0.1386 0.1767
Ethanol standard deviation 0.0128 0.0153 0.0130 0.0135 0.0169 0.0259 0.0214 0.0235 0.0173
n-Propanol 25 0.0899 0.1141 0.1246 0.1186 0.1195 0.1073 0.1025 0.1333 0.1033 0.1204
30 0.0906 0.1178 0.1286 0.1222 0.1230 0.1109 0.1051 0.1369 0.1063 0.1247
40 0.1030 0.1252 0.1367 0.1296 0.1302 0.1181 0.1103 0.1443 0.1125 0.1336
50 0.1240 0.1328 0.1450 0.1372 0.1376 0.1256 0.1157 0.1519 0.1189 0.1427
n-Propanol standard deviation 0.0019 0.0042 0.0027 0.0029 0.0009 0.0005 0.0065 0.0005 0.0034
n-Butanol 25 0.0736 0.0996 0.1089 0.1021 0.1045 0.0935 0.0905 0.1177 0.0912 0.1010
30 0.0749 0.1028 0.1124 0.1052 0.1076 0.0966 0.0928 0.1209 0.0939 0.1047
40 0.0815 0.1093 0.1195 0.1116 0.1138 0.1030 0.0974 0.1274 0.0994 0.1121
50 0.0841 0.1159 0.1268 0.1181 0.1203 0.1095 0.1021 0.1341 0.1050 0.1198
n-Butanol standard deviation 0.0285 0.0385 0.0308 0.0331 0.0222 0.0172 0.0465 0.0189 0.0310
n-Pentanol 25 0.0532 0.0896 0.0981 0.0911 0.0938 0.0841 0.0824 0.1071 0.0830 0.0884
30 0.0545 0.0925 0.1013 0.0939 0.0966 0.0869 0.0845 0.1100 0.0854 0.0916
40 0.0596 0.0983 0.1077 0.0996 0.1023 0.0926 0.0887 0.1160 0.0904 0.0982
50 0.0594 0.1044 0.1143 0.1054 0.1080 0.0985 0.0930 0.1221 0.0956 0.1049
n-Pentanol standard deviation 0.0397 0.0488 0.0409 0.0436 0.0340 0.0305 0.0572 0.0320 0.0393
Alcohols standard deviation 0.0267 0.0333 0.0276 0.0290 0.0247 0.0270 0.0401 0.0267 0.0286
 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548 547

Fig. 9. Logarithm of diffusion coefficients vs logarithm of temperature for alcohol-air

Fig. 8. Logarithm of diffusion coefficients vs logarithm of temperature for alcohol-air mixtures at 86 kPa. Unique equation proposed for all alcohols (lines) and experi-
mixtures at 86 kPa. Specific equations for each alcohol (lines) and experimental re- mental results (symbols).
sults (symbols).

Eq. (25)) are proposed for each alcohol-air mixture, based on cor-
with n being the number of carbon atoms in the alcohol. The
relations from the experimental results obtained at 86 kPa, which
optimal fit was obtained with A ¼ 5.386; B ¼ 0.686; C ¼ 32.138 and
predict the variation of the diffusion coefficient with temperature.
D ¼ 3.604. Results from this equation are compared in Table 5 and
These equations are shown in Fig. 8. In all cases, the correlation
Fig. 9 with the experimental ones (from this work). As can be
coefficients (R2) remain above 92% except for n-pentanol (86%). As
observed, the standard deviation from the proposed equation is
can be observed, the exponents affecting temperature result in
much better (0.0045 cm2/s) than those obtained from all other
4.44, 4.28, 4.11, 1.79 and 1.51 for methanol, ethanol, n-propanol, n-
equations from the literature, and the correlation coefficient
butanol and n-pentanol, respectively. These results show that all
(R2 ¼ 0.9917) is much better than those of the specific equations
the equations proposed in the literature (with exponents affecting
proposed for each alcohol.
temperature ranging between 1.37 and 1.81, as shown in Section 4)
underestimate the effect of temperature for methanol, ethanol and
propanol. In an attempt to obtain a single equation for all alcohols 7. Conclusion
(from C1 to C5), the following is proposed:
Nine different equations among those proposed in the literature
to estimate the diffusion coefficient have been tested for binary
ln DAAIR ¼ ðA  BnÞln T  ðC  DnÞ (26)
gaseous mixtures alcohol-air, alcohol being either methanol,
ethanol, n-propanol, n-butanol or n-pentanol. The two former
equations (the original one from ChapmaneEnskog and the Wilke-
Table 5
Lee equation), were expressed as a function of the collision integral,
Comparison between experimental results and those predicted with Eq. (26).
which accounts for the molecular interactions. Subsequent equa-
Alcohol T ( C) Dexp (cm2/s) Dpredicted (cm2/s) tions (CheneOthmer and Fuller-Schettler-Giddings) eliminate
(Eq. (26))
explicitly the collision integral, and some other equations introduce
Methanol 25 0.1777 0.1724 a correction for the effect of the intermolecular forces by including
30 0.1840 0.1865 the Sutherland constant, either explicitly (Arnold) or implicitly
40 0.2194 0.2172
Methanol standard deviation 0.0036
(Gilliland, SingheSingh, Huang). Finally, the Othmer-Chen equation
Ethanol 25 0.1264 0.1270 accounts for the effect of intermolecular forces by means of the air
30 0.1288 0.1358 viscosity, which can also be estimated from the Sutherland con-
40 0.1636 0.1547 stant. Among the equations studied, the Arnold equation provided
50 0.1719 0.1755
the best predictions for the diffusion coefficients of all the alcohols
Ethanol standard deviation 0.0060
n-Propanol 25 0.0899 0.0936 tested. However, none of them led to a good overall prediction, and
30 0.0906 0.0989 the inaccuracies were especially high for n-butanol and n-pentanol.
40 0.1030 0.1102 Additionally, experimental values for the diffusion coefficients of
50 0.1240 0.1223 the alcohols tested show a wide dispersion in the literature, prob-
n-Propanol standard deviation 0.0058
n-Butanol 25 0.0736 0.0689
ably due to the different measurement methods used and to the
30 0.0749 0.0720 experimental difficulties. The main variable contributing to such
40 0.0815 0.0785 difficulties is temperature, since increasing temperature leads to
50 0.0841 0.0852 faster liquid evaporation together with higher liquid expansion. A
n-Butanol standard deviation 0.0032
set of experiments made at different temperatures (from 25 to
n-Pentanol 25 0.0532 0.0508
30 0.0545 0.0524 50  C) at 86 kPa have led to the proposal of five specific equations
40 0.0596 0.0559 (one for each tested alcohol), which show that the exponent
50 0.0594 0.0594 affecting temperature was clearly underestimated for lightest al-
n-Pentanol standard deviation 0.0024 cohols. A four-parameter equation for the estimation of the diffu-
Alcohols standard deviation 0.0045
sion coefficient of all the alcohols tested has also been proposed,
548 M. Lapuerta et al. / Applied Thermal Engineering 73 (2014) 539e548

showing much better agreement to the experimental results than [24] J.C. Slattery, R.B. Bird, Calculation of the diffusion coefficient of dilute gases
and of the self-diffusion coefficient of dense gases, AIChE J. 4 (2) (1958)
those proposed in the literature.
127e141.
Further experiments at different pressures or altitudes would be [25] E.T. Hashim, H.N.A. Al-Shorachi, Prediction of diffusion coefficients for binary
necessary to include the effect of pressure. A thorough selection of gas system, Petroleum Sci. Technol. 25 (8) (2007) 983e989.
the diffusion coefficient is essential for an accurate prediction of [26] R. Eslamloueyan, M.H. Khademi, A neural network-based method for esti-
mation of binary gas diffusivity, Chenometr. Intell. Lab. Syst. 104 (2010)
evaporation losses, and for an appropriate design of fuel systems. 195e204.
[27] G. Karaiskakis, D. Gavril, Determination of diffusion coefficients by gas chro-
matography, J. Chromatogr. 1037 (2004) 147e189.
Acknowledgements [28] R.W. Elliott, H. Watts, Diffusion of some hydrocarbons in air: a regularity in
the diffusion coefficients of a homologous series, Can. J. Chem. 50 (1972)
31e34.
The authors wish to thank the Colombian department for Sci-
[29] K.C. Kwon, T.H. Ibrahim, Y.K. Park, C.M. Simmons, Pseudo-binary molecular
ence, Technology and Innovation (Colciencias) for granting J.P. diffusion of vapors into air, Adv. Environ. Res. 8 (2004) 667e678.
Hernandez graduate scholarship and for financing his stay at the [30] R.H. Perry, D.W. Green, J.O. Maloney, Perry's Chemical Engineers Handbook,
University of Castilla-La Mancha (Spain).The Government of sixth ed., McGraw-Hill, New York, 1984.
[31] E.A. Mason, L. Monchick, Transport properties of polar-gas mixtures, J. Chem.
Castilla-La Mancha is acknowledged for the financial support Phys. 36 (1962) 2746e2757.
through project COMBALT-2 (Ref. POII10-0173-0731). Experimental [32] G.A. Lugg, Diffusion coefficients of some organic and other vapors in air, Anal.
support was granted by the program Sostenibilidad 2013e2014 of Chem. 40 (1968) 1072e1077.
[33] T.R. Marrero, E.A. Mason, Gaseous diffusion coefficients, J. Phys. Chem. Ref.
the Universidad de Antioquia. Data 1 (1) (1972) 3e116.
[34] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, second ed., John
Wiley and Sons, Inc, Nueva York, 2002.
References [35] C.O. Bennett, J.E. Myers, Momentum, Heat and Mass Transfer, second ed.,
McGraw-Hill, New York, 1962.
[1] M. Lapuerta, R. García-Contreras, J. Campos-Fern
andez, M.P. Dorado, Stability, [36] J.E. Lennard-Jones, A.F. Devonshire, Critical phenomena in gases, Proc. R. Soc.
lubricity, viscosity and cold flow properties of alcoholediesel blends, Energy 163 (1937) 53e70.
Fuels 24 (2010) 4497e4502. [37] W.H. Stockmayer, Second virial coefficients of polar gases, J. Chem. Phys. 9
[2] A. Kerihuel, M.S. Kumar, J. Bellettre, M. Tazerout, Use of animal fats as CI (1941) 398.
engine fuel by making stable emulsions with water and methanol, Fuel 84 [38] J.O. Hirschfelder, C.F. Curtiss, R.B. Bird, Molecular Theory of Gases and Liquids,
(2005) 1713e1716. John Wiley and Sons, New York, 1964.
[3] X. Gu, Z. Huang, J. Cai, J. Gong, X. Wua, C. Lee, Emission characteristics of a [39] P.D. Neufeld, A.R. Janzen, R.A. Aziz, Empirical equations to calculate 16 of the
spark-ignition engine fuelled with gasoline-n-butanol blends in combination transport collision integrals U (l,s) for the LennardeJones (12e6), Potential. J.
with EGR, Fuel 93 (2012) 611e617. Chem. Phys. 57 (1972) 1100e1102.
[4] M.B. Celik, Experimental determination of suitable ethanolegasoline blend [40] B.E. Polling, J.M. Prausnitz, J.P. O'Connell, The Properties of Gases and Liquids,
rate at high compression ratio for gasoline engine, Appl. Therm. Eng. 28 fifth ed., Mc-Graw Hill Book Company, New York, 2000.
(2008) 396e404. [41] N.H. Chen, Empirical equation for the collision function on LennardeJones
[5] M. Lapuerta, O. Armas, R. García-Contreras, Effect of ethanol on blending potential for gas diffusion coefficient, Ind. Eng. Chem. 51 (1959) 1494.
stability and diesel engine emissions, Energy Fuels 23 (2009) 4343e4354. [42] R.S. Browak, Predicting transport properties of dilute gases, Ind. Eng. Chem.
[6] O. Armas, R. García-Contreras, A. Ramos, Pollutant emissions from engine Process Des. Dev. 8 (1969) 240e253.
starting with ethanol and butanol diesel blends, Fuel Process. Technol. 100 [43] H.Y. Erbil, Y. Avci, Simultaneous determination of toluene diffusion coefficient
(2012) 63e72. in air thin tube evaporation and sessile drop evaporation on a solid surface,
[7] L. Zhu, C.S. Cheung, W.G. Zhang, Z. Huang, Effect of charge dilution on gaseous Langmuir 18 (2002) 5113e5119.
and particulate emissions from a diesel engine fueled with biodiesel and
biodiesel blended with methanol and ethanol, Appl. Therm. Eng. 31 (2011)
2271e2278.
Nomenclature
[8] N. Weyandt, M.L. Janssens, Flame Arrester Evaluation for E-diesel Fuel Tanks, D: diffusion coefficient (cm2/s)
2003. Report NREL SR-540-34301. h: diffusion path (m)
[9] L.R. Waterland, S. Venkatesh, S. Unnasch, Safety and Performance Assessment k:00Boltzmann constant (J/K)
of Ethanol/diesel Blends (E-diesel), 2003. Report NREL/SR-540-34817. m_ : mass flow rate per unit area (kg/m2s)
[10] K. Okamoto, M. Hiramatsu, H. Miyamoto, T. Hino, M. Honma, N. Watanabe, p: pressure (bar)
Y. Hagimoto, K. Miwa, H. Ohtani, Evaporation and diffusion behavior of fuel T: temperature (K)
mixtures of gasoline and kerosene, Fire Saf. J. 49 (2012) 47e61. Tb: boiling temperature (K)
[11] Y. Guo, M. He, Q. Zhong, Y. Zhang, Mass diffusion coefficients of oxygenated Tc: critical temperature (K)
fuel additives in air, Energy 34 (2009) 1560e1564. t: time (s)
[12] S. Chapman, T.G. Cowling, The Mathematical Theory of Non-uniform Gases, R: gas constant (bar m3/kmol K)
third ed., Cambridge University Press, New York, Melbourne, 1991. S: Sutherland constant
[13] E.O. Meyer, The Kinetic Theory of Gases, second ed., Longmans, Green and Co., vb: specific volume of saturated liquid at boiling temperature (L/kmol)
London, New York and Bombay, 1899. vc: critical volume (L/kmol)
[14] J.H. Jeans, The Dynamical Theory of Gases, C. J. Clay and Sons, London, 1904. v*: atomic volumes (L/kmol)
[15] C.R. Wilke, C.Y. Lee, Estimation of diffusion coefficients for gases and vapors, W: molecular weight (kg/kmol)
Ind. Eng. Chem. 47 (1955) 1253e1257. Y: mass fraction
[16] H.T. Chen, D.F. Othmer, New generalized equation for gas diffusion coefficient, ε: maximum molecular attraction energy (J)
J Chem.. Eng. Data 7 (1962) 37e41. r: density (kg/m3)
[17] J.C. Hirschfelder, R.B. Bird, E.L. Spotz, The transport properties of gases and s: molecular radius (Å)
gaseous mixtures, J. Chem. Phys. 44 (1949) 205e231. m: viscosity (cP)
[18] J.H. Arnold, Studies in diffusion: Idestimation of diffusivities in gaseous UD: collision integral
systems, Ind. Eng. Chem. 22 (1930) 1091e1095.
[19] E.R. Gilliland, Diffusion coefficients in gaseous systems, Ind. Eng. Chem. 26 Subscripts
(1934) 681e685.
[20] E.N. Fuller, P.D. Schettler, J.C. Giddings, A new method for prediction of binary 0: start of the experiments
gas phase diffusion coefficients, Ind. Eng. Chem. 58 (1966) 19e27. A: alcohol
[21] P.C. Singh, S. Singh, Development of a new correlation from binary gas phase AIR: air
diffusion coefficients, Int. Commun. Heat Mass Transfer 10 (1983) 123e140. B: n-Butanol
[22] T. Huang, F.J. Yang, F. Huang, C. Kuo, Measurements of diffusion coefficients by E: ethanol
the method of gas chromatography, J. Chromatogr. Chrom. 6044 (1972) M: methanol
13e24. Pn: n-Pentanol
[23] D.F. Othmer, H.T. Chen, Correlating diffusion coefficients in binary gas sys- Pr: n-Propanol
tems, EC Process Des. Dev. 1 (4) (1962) 249e254. s: saturated vapor