Cent. Eur. J. Chem.

• 11(1) • 2013 • 57-62

DOI: 10.2478/s11532-012-0131-1

Central European Journal of Chemistry

Measurement and prediction

of binary mixture flash point
Research Article

Mariana Hristova*
Department of Fundamental Chemical Technology
University of Chemical Technology and Metallurgy, Sofia, Bulgaria

y
Received 7 July 2012; Accepted 23 August 2012

op

Abstract: The flash points of three binary mixtures, containing n-heptane, o-xylene, m-xylene and ethylbenzene, were measured by
Pensky-Martens closed cup tester. The experimental data were compared with the calculated values using Liaw`s Model with the
application of Raoult`s Law and Wilson equation. These equations were in good qualitative agreement.
Keywords: Flash point • Binary mixture • Pensky-Martens Closed Cup Tester
© Versita Sp. z o.o.
rc
1. Introduction The measurement of a flash point is defined in test
methods that are maintained by standardization bodies
Flammability is an important factor to consider in such as the Energy Institute in the UK, ASTM in the
ho

developing safe methods for storing and handling solids USA, CEN in Europe and ISO internationally. In the
and liquids [1]. Laboratories and industries commonly last decade, new standards and completely automated
use flammable substances. Corresponding mixtures are methods for measuring flash point were introduced, but
used to carry out certain experiments and processes. older methods are still accepted and are widely used
It is important to take note of the physical properties of throughout the world.
ut

the substances to avoid any of the hazards associated Experimental flash point data for pure substances
with them. Liquids that have a flashpoint of 93°C or less are readily available in the literature. As one of the
are subject to the hazard class flammable liquids of the major safety data items, the flash point is specified in a
GHS (Globally Harmonized System of Classification and standard material safety data sheet (MSDS). The flash
A

Labelling of Chemicals). The flashpoint is the lowest points for pure substances can also be found in many
temperature of the liquid, corrected to a barometric engineering handbooks and databases. The flash point
pressure of 101.3 kPa, at which the application of a test data corresponding to various mixtures appear to be
flame causes the vapour of the liquid to ignite momentarily scarce in the literature and are valid for a specific mixture
and a flame to propagate across the surface of the liquid composition.
under the specified conditions of test [2]. This means, The experimental measurement of the flash point is
the lower flammability limit (LFL) is exceeded at the expensive and time consuming and therefore, several
flashpoint. authors attempted to develop predictive models. By
The flashpoint is determined by standardized using these models, the flash points of various liquid
equilibrium or non-equilibrium methods. Normally a solutions would be predicted relatively easily from a
closed-cup method is used because its results tend to be limited number of initial data.
more reliable. Test results from open-cup measurements Affens and McLaren [3] have developed a model
are not always reliable since volatile compounds may based on Raoult’s Law to predict the flash point for
escape from the measuring equipment. Therefore, they hydrocarbon solutions. White et al. [4] reduced Affens
are acceptable only in special cases with an explicit and McLaren’s model to a simpler equation by ignoring
justification. the temperature effect upon the LFL for prediction of the

* E-mail: [email protected]
57
 Measurement and prediction of binary mixture flash point

flash points of two aviation-fuel mixtures. Gmehling and

Rasmussen [5] proposed the model according to the
non-ideal liquid phase solution and LFL dependence on
temperature. Crowl and Louvar [6] suggested that the
flash point of a liquid solution with only one flammabale
component could be estimated using Raoult’s Law. This
model is only adequate for a composition’s range in
which the flammable substance composition lies close
to unity. Catoire et al. developed a simple formulation
for the estimation of flash points of miscible combustible
solvents mixtures [7]. It consists of an equation
which was previously validated for pure compounds
[8].
A new trend to estimate the flash temperature is
based on the quantitative structure property relationship Figure 1. Photograph of the experimental apparatus.

y
(QSPR), which incorporates in its procedure theoretical
molecular calculations for geometrical, topological, refrigerated chamber, and then allowing the temperature

op
quantum mechanical and electronic properties [9-12]. In of the test sample to rise slowly as the experiments
general, the QSPR method requires special computer began. The ramping rate was reduced to about
programs. Moreover, the training set of the QSPR 1ºC min-1 to test the assumption of thermal equilibrium in
procedure should contain large number of compounds the tester cup. The ISO 2719 gradient was 1.5ºC min-1.
with different molecular structures to obtain suitable The temperature control was sustained by electrical
rc
results. heating.
Since 2002, Liaw et al. have developed many models A tester thermometer with a range from -7 to +110ºC
for total miscible flammable mixtures, partially miscible was used. The ambient barometric pressure was
and aqueous-organic mixtures [13-20]. observed and recorded at the time of the test. When the
This study measured and predicted the flash pressure differed from 760 mm Hg (101.3 kPa), the flash
ho

point of three binary mixtures, n-heptane+o-xylene, point was corrected as follows:

n-heptane+m-xylene, n-heptane+ethylbenzene. The
flash points were measured by Pensky-Martens closed Corrected flash point = T0 +0.25(101.3 – P)
cup tester, and compared with the values calculated
by using the Liaw Model, Raoult’s Law and Wilson where T0 is observed flash point (ºC); P is barometric
equation. pressure (kPa).
ut

The mole fraction of each component was

determined by measuring the mass using the Sartorius
2. Experimental procedure digital balance (sensitivity 0.0001 g, maximum load
100 g). The sample was prepared and transferred to the
A

2.1. Materials cup of the apparatus at least 10ºC below the expected
All components were purchased from the Merck flash point. The sample was not stirred while the flame
Company (Germany) with a minimum purity of 99.5%. was lowered into the cup. The flash point was the
Three mixtures were selected for the samples: temperature at which the test flame application caused
n-heptane (1)+o-xylene (2), n-heptane (1)+m-xylene (2) a distinct flash in the interior of the cup. The measured
and n-heptane (1)+ethylbenzene (2). value was the mean of the two measurements which did
not differ more than 2ºC.
2.2. Experimental Apparatus and Procedure
The experimental data were obtained using the Pensky-
Martens Closed Cup tester (Fig. 1) model PM 1, SUB 3. Theory
(Berlin, Germany).The closed cup tester was operated
according to the standard test method, BDS EN ISO 3.1. Calculation model
2719 [21]. A modification of this standard test method The properties of mixtures are difficult to predict because
was implemented because most of the investigated flash a simple mixing rule will not work when interactions
points were below room temperature. The modification among the mixture components are strong. The flash
consisted of cooling the testing cup with a sample in a point of a liquid mixture depends on the composition

58
M. Hristova

of the mixture and on the types of chemicals involved identical, and the mixture is in the ideal state; if the
[16]. activity coefficient is away from unity, the mixture is in
The flash point of a flammable liquid is the the non-ideal state. The concept of activity coefficient
temperature at which the vapor pressure of the liquid is is used for the liquid phase. The liquid-phase activity
such that it produces a concentration of vapor in the air coefficient models include the Margules equation [24],
that corresponds to its lower flammable limit (LFL) [1]. van Laar equation [24], Wilson equation [28], NRTL
Thus, at the flash point of a liquid solution, Le Chatelier’s equation [29], UNIFAC equation [30], and so on. In this
rule, that describes the lower flammable limit of a gas work, the Wilson equation was used to calculate activity
mixture, is followed: coefficient:
yi
1= ∑ (1) (6)
i LFLi
where yi is the mole fraction of the flammable substance
I in the vapor phase, LFLi is its lower flammable (7)
limit.

y
From the definition of flash point [1], it is the
temperature at which the vapour pressure is equivalent (8)

op
to the lower flammable limit composition in air:
where Vi L is the molar volume of pure liquid i , is
(2) an energy parameter characterizing the interaction of
molecule i with molecule j , R is the universal gas
where is the vapour pressure at the flash point constant, T is the temperature (K).
rc
temperature and P is the ambient pressure. If the mixture is an ideal, Eq. 4 becomes:
Another assumption of the model is that the liquid
and vapour phases of the mixture are in equilibrium [13]. (9)
Thus, the vapor-phase composition, yi, can be derived
from the vapor–liquid equilibrium (VLE) between a non- The temperature that satisfies Eqs. 4 or 9 is the flash
ho

ideal liquid and a perfect gas as point temperature of the mixture.

(3) 3.2. Data input

The parameters required for the calculation include
Substituting Eqs. 2 and 3 into Eq. 1 the flash point of the pure components flash points and the Antoine
binary mixture can be calculated: coefficients.
ut

The measured flash points of the pure components

(4) are compared with values provided by the supplier and
are listed in Table 1.
where xi , , and Pi sat are the mole fraction in liquid phase, The experimental flash points for n-heptane and
A

activity coefficient and vapour pressure at temperature ethylbenzene are identical to values reported by the
T, of the mixture components, respectively. supplier. The values for o-xylene and m-xylene are
The vapour pressure Pi sat can be estimated from almost identical to those provided by the supplier.
an equation such as Antoine’s equation, if the required Experimental flash points of pure components were
constants are known: used (Table 1) for the calculations in this study.
Bi The Antoine coefficients and binary interaction
log Pi sat = Ai −
T + C i (5) parameters of the Wilson equation were taken from the
literature and are listed in Tables 2 and 3.
where Ai, Bi and Ci are the parameters of compound i.
The parameters for the Antoine equation can be obtained
from different collections [23,24]. 4. Results and discussion
From the classic thermodynamics, we know that
the activity coefficient is introduced as the revision and The flash points of of n-heptane(1)+o-
judgment for the non-ideality of the mixture [25-27]. If xylene(2), n-heptane(1)+m-xylene(2) and
the activity coefficient is equal to unity, the interactions n-heptane(1)+ethylbenzene(2) mixtures were tested
between dissimilar or same molecules are always over the entire composition range. The results

59
 Measurement and prediction of binary mixture flash point

Table 1. Molar volume, experimental flash points and flash points from supplier for pure components.
Substance CAS № V L (cm3 mol-1) Flash point Exp. (ºC) Flash point from suppl. (ºC)

n-heptane 142-82-5 147.6 -4 -4

o-xylene 95-47-6 120.8 32 30
m-xylene 108-38-3 122.3 26 25
Ethyl benzene 100-41-4 123.07 15 15

Table 2. Antoine coefficients of the components* [24].

Substance A B C

n-heptane 6.89385 1264.37 216.635

o-xylene 6.99891 1474.679 213.686
m-xylene 7.00908 1462.266 215.105
Ethyl benzene 6.95719 1424.255 213.206

y
op
Table 3. Binary parameters of the Wilson equation* [31].
System A12 A21

n-heptane (1)+o-xylene (2) -84.2388 235.4754

n-heptane (1)+m-xylene (2) -139.8292 250.8485
Figure 2.
rc
Comperison of the flash point prediction curves with
n-heptane (1)+ethylbenzene (2) -266.22 454.42 the experimental data for n-heptane (1) + o-xylene (2)
system.

Table 4. Experimental flash points and predictions for n-heptane (1)

+ o-xylene mixture.

DTfp / oC DTfp / oC
ho

X1 Exp. (ºC) Ideal Wilson

0 32.0 32.00 0.00 32.00 0.00

0.1 20.5 23.53 3.03 21.73 1.23
0.2 15.0 17.13 2.13 15.09 0.09
0.3 11.0 12.41 1.41 10.58 0.42
ut

0.4 6.9 8.72 1.82 7.23 0.33

0.5 4.5 5.72 1.22 4.60 0.10
0.6 2.5 3.20 0.70 2.42 0.08
0.7 1.0 1.04 0.04 0.56 0.44
A

0.8 -0.9 -0.84 0.06 -1.09 0.19

0.9 -1.9 -2.51 0.61 -2.59 0.69
1.0 -4.0 -4.00 0.00 -4.00 0.00
Figure 3. Comperison of the flash point prediction curves with
the experimental data for n-heptane (1) + m-xylene (2)
Table 5. Experimental flash points and predictions for n-heptane (1)
system.
+ m-xylene mixture.

X1 Exp. (ºC) Ideal DTfp / oC Wilson DTfp / oC

from experiments and predicted by Liaw’s model
are presented in Tables 4-6 respectively, where
0 26.0 26.00 0.00 26.00 0.00 .
0.1 17.0 19.47 2.47 18.5 1.50 In Figs. 2-4 the flash point variation between the model
0.2 14.0 14.57 0.57 13.36 0.64
predictive curves and the experimentally-derived data
0.3 9.0 10.72 1.72 9.55 0.55
for the binary solutions are compared. For all mixtures,
0.4 6.5 7.57 1.07 6.57 0.07
0.5 3.9 4.94 1.04 4.14 0.24
the predicted values based upon an ideal solution
0.6 2.5 2.68 0.18 2.10 0.40 assumption and Wilson equation (non-ideal behaviour)
0.7 0.2 0.71 0.51 0.33 0.13 are accurate with experimental measurements. As
0.8 -0.9 -1.04 0.14 -1.24 0.34 confirmed by the activity coefficient vs. composition
0.9 -2.0 -2.59 0.59 -2.67 0.67 (Table 7), all mixtures reveal a positive deviation from
1.0 -4.0 -4.00 0.00 -4.00 0.00 that of an ideal solution. The activity coefficients are very

60
M. Hristova

Table 6. Experimental flash points and predictions for n-heptane (1)

+ ethyl benzene mixture.

X1 Exp. (ºC) Ideal DTfp / oC Wilson DTfp / oC

0 15.0 15.00 0.00 15.00 0.00

0.1 11.0 11.94 0.94 11.25 0.25
0.2 8.5 9.28 0.78 8.26 0.24
0.3 5.5 6.94 1.44 5.81 0.31
0.4 3.2 4.87 1.67 3.77 0.57
0.5 2.5 3.02 0.52 2.03 0.47
0.6 0.8 1.35 0.55 0.54 0.26
0.7 -0.3 -0.17 0.13 -0.77 0.47
0.8 -1.7 -1.55 0.15 -1.93 0.23
0.9 -2.9 -2.82 0.08 -3.00 0.10
1.0 -4.0 -4.00 0.00 -4.00 0.00 Figure 4. Comperison of the flash point prediction curves with the
experimental data for n-heptane (1) + ethylbenzene (2)

y
system.
Table 7. Calculated activity coefficients of binary mixtures.
Table 8 includes average absolute deviation (A.A.D.)
X1 n-hepnane(1) n-hepnane(1) n-hepnane(1)+

op
+ o-xylene(2) + m-xylene(2) ethylbenzene(2) for three binary solutions:
γ1 γ2 γ1 γ2 γ1 γ2

0.0 1.26 1.00 1.17 1.00 1.19 1.00 (10)
0.1 1.21 1.00 1.14 1.00 1.16 1.00
0.2 1.17 1.01 1.12 1.01 1.14 1.00
A.A.D. is a measure of agreement between the
rc
0.3 1.13 1.02 1.09 1.01 1.12 1.01
0.4 1.10 1.04 1.07 1.02 1.09 1.02
experimental data and the calculated values.
0.5 1.07 1.06 1.05 1.04 1.07 1.04 In the prediction model, it was assumed that the
0.6 1.05 1.09 1.03 1.06 1.05 1.07 vapour phase and the liquid phase of a solution were in
0.7 1.03 1.14 1.02 1.09 1.03 1.10 equilibrium. The predicted data was only adequate for
0.8 1.01 1.19 1.01 1.13 1.02 1.15 the data determined by the closed cup test method, and
ho

0.9 1.00 1.25 1.00 1.17 1.00 1.23 may not be appropriate to apply to the data obtained
1.0 1.00 1.33 1.00 1.23 1.00 1.34
from the open cup test method because of its condition
having deviated from the vapour-liquid equilibrium.
Table 8. Average absolute deviation (A.A.D.) between calculated
and experimental flash point for binary mixtures of
n-heptane, o-xylene, m-xylene and ethylbenzene.
5. Conclusions
ut

Solution Ideal Wilson

n-heptane + o-xylene 1.02 0.37 The flash points of three binary mixtures containing
n-heptane + m-xylene 0.78 0.44 n-heptane, o-xylene, m-xylene and ethylbenzene,
A

n-heptane + ethylbenzene 0.58 0.19 were measured by the Pensky-Martens closed cup
tester. The experimental data were compared with the
close to unity; therefore it can be concluded that these values calculated by using Raoult`s Law and the Wilson
mixtures are almost-ideal solutions. This is the reason equation. Good qualitative agreement was obtained
for the small differences between the flash points that with these equations. However, the calculated values
were predicted as an ideal solution and by using the based on the Wilson equation were found to be better
Wilson equation. than those based on Raoult`s Law.

References

[1] F.P. Lees, Loss Prevention in the Process Industries, [3] W.A. Affens, G.W. Mclaren, J. Chem. Eng. Data,
2nd edition (Butterworth-Heinemann, Oxford, UK, 17, 482 (1972)
1996) vol.1 [4] D. White, C.L. Beyler, C. Fulper, J. Leonard, Fire
[2] CCPS/AIChe, Guidelines for Engineering Design Safety. J. 28, 1 (1997)
for Process Safety (American Institute of Chemical [5] J. Gmehling, P. Rasmussen, Ind. Eng. Chem. Fund.
Engineers, New York, USA, 1993) 21(2), 186 (1982)

61
 Measurement and prediction of binary mixture flash point

[6] D.A. Crowl, J.F. Louvar, Chemical Process safety; [20] H.-J. Liaw, V. Gerbaud, C.-Y. Chiu, J. Chem. Eng.
Fundamentals with Applications (Prentice-Hall: Data 55, 134 (2010)
Upper Saddle River, NJ, USA, 2002) [21] EN ISO 2719. Standard Test Methods for Flash
[7] L. Catoire, S. Paulmier, J. Phys. Chem. Ref. Data Point by Pensky-Martens Closed Cup Tester.
35, 9 (2006) [22] M. Vidal, W.J. Rogers, M.S. Mannan, Process Saf.
[8] L. Catoire, V. Naudet, J. Phys. Chem. Ref. Data 33, and Envir.Prot. 84, B1, 1 (2006)
4, 1083 (2004) [23] T. Boublik, V. Fried, E. Hala, Vapour Pressure
[9] F. Gharagheizi, R. Faraghi, QSAR&Combinatorial of Pure Substances: Selected Values of the
Science 27, 6, 679 (2008) Temperature Dependence of the Vapour Pressures
[10] A.R. Katritzky, R. Petrukhin, R. Jain, M. Karelson, of Some Pure Substances in the Normal and Low
Journal of Chemical Information and Computer Pressure Region, 2nd edition (Elsevier Science Ltd:
Sciences 41,6, 1521 (2001) Amsterdam, New York, 1984)
[11] A.R.  Katritzky, I.B. Stoyanova-Slavova, [24] B.E. Poling, J.M. Prausnitz, J.P. O’Çonnell, The
D.A. Dobchev, M. Karelson, J. of Molecular Grafhics Properties of Gases and Liquids, 5th edition
and Modeling 529-536, (2007) (McGraw-Hill: New York, 2001)

y
[12] A. Khajeh, H. Modarress, J. Hazard. Mater. 179, [25] J.M. Prausnitz, Molecular Thermodynamics of
715 (2010) Fluid-Phase Equilibria (Prentice Hall: New Jersey,

op
[13] H.-J. Liaw, Y.-H. Lee, J. Loss Prev. Process Ind. 15, 1969)
429 (2002) [26] J.M. Smith, H.C. Van Ness, Introduction to Chemical
[14] H.-J. Liaw, Y.-Y. Chiu, J. Hazard. Mater. A101, 83 Engineering Thermodymanics (McGraw-Hill, New
(2003) York, 1975)
[15] H.-J. Liaw, Y.-Y. Chiu, J. Hazard. Mater, A137, 38 [27] S.I. Sandler, Chemical Engineering Thermodynamics
rc
(2006) (Wiley, New York, 1989)
[16] H.-J. Liaw, T.-P. Lee, J.-S. Tsai, W.-H Hsiao, [28] G.M. Wilson, J. Am. Chem. Soc. 86(2), 127 (1964)
M.-H. Chen, T.-T. Hsu, J. Loss Prev. Process Ind. [29] H. Renon, J.M. Prausnitz, AIChE Journal, 14, 135
16, 173 (2003) (1968)
[17] H.-J. Liaw, S.-C. Lin, J. Hazard. Mater. 140, 155 [30] A. Fredenslund, R.L. Jones, J.M. Prausnitz, AIChE
ho

(2007) Journal, 21, 1086 (1975)

[18] H.-J. Liaw, W.H. Lu, V. Gerbaud, C.C. Chen, [31] C. Díaz, J. Tojo, J. Chem. Thermodynamics, 34,
J. Hazard. Mater. 153, 1165 (2008) 1975 (2002)
[19] H.-J. Liaw, C.T. Chen, V. Gerbaud, Chem. Eng. Sci.
63, 4543 (2008)
ut
A

62