Article

pubs.acs.org/est

Correlation of Chemical Evaporation Rate with Vapor Pressure

Donald Mackay*, and Ian van Wesenbeeck

Environmental and Resource Studies, Trent University, 1600 West Bank Drive, Peterborough, Ontario K9J 7B8, Canada
Dow AgroSciences, 9330 Zionsville Road, Indianapolis, Indiana 46268, United States

ABSTRACT: A new one-parameter correlation is developed for the evaporation rate (ER) of
chemicals as a function of molar mass (M) and vapor pressure (P) that is simpler than existing
correlations. It applies only to liquid surfaces that are unaected by the underlying solid substrate
as occurs in the standard ASTM evaporation rate test and to quiescent liquid pools. The
relationship has a sounder theoretical basis than previous correlations because ER is correctly
correlated with PM rather than P alone. The inclusion of M increases the slope of previous log ER
versus log P regressions to a value close to 1.0 and yields a simpler one-parameter correlation,
namely, ER (g m1 h1) = 1464P (Pa) M (g mol1). Applications are discussed for the
screening level assessment and ranking of chemicals for evaporation rate, such as pesticides, fumigants, and hydrocarbon carrier
uids used in pesticide formulations, liquid consumer products used indoors, and accidental spills of liquids. The mechanistic
signicance of the single parameter as a mass-transfer coecient or velocity is discussed.

INTRODUCTION
Evaporation plays an important role in the fate of chemicals
applied for agricultural and other purposes. It is useful to have an
estimate of the potential for evaporation as a function of the
physicalchemical properties of the substance [notably vapor
pressure (VP)] as measured under idealized laboratory
conditions from a plane surface. Empirical laboratory test data
can be used for screening level assessments of chemical volatility
and to rank chemicals for the tendency to evaporate. Woodrow
et al.1,2 showed that the evaporation rate (ER) from inert surfaces
expressed as g m2 h1 correlates well (r2 of 0.989) with measured
vapor pressure. Using these data along with data reported by Guth
et al.3 and data from the standard ASTM D 3539-87 evaporation
rate test,4 van Wesenbeeck et al.5 derived several similar correlations
for a total of 82 substances and showed that the linear relationship
between ln(ER) and ln(VP) holds over a range of 15 orders of
magnitude of pressure (from 1010 to 105 Pa). ER is a particularly
important factor for fumigants, in which the product may be applied
as a liquid, but the desired toxic exposure occurs via the vapor phase.
There are several incentives for obtaining a robust correlation
of ER with vapor pressure. In general, it is desirable when
selecting chemicals for specic applications that their evaporation characteristics be fully appreciated. For example, Zeinali
et al.6 have recently discussed the need to assess the evaporation
rate of components of hydrocarbon carrier uids used in
pesticide formulations and, thus, their potential to contribute to
ozone formation potential (OFP) using additional data on the
maximum incremental activity of the specic substances. There is
also interest in assessing the relative evaporation rates and, hence,
inhalation exposures of chemicals used indoors. Examples
include personal care products, cleaning agents, and pesticides.79 Relative ER data are also of interest when assessing the
fate and exposures of volatile substances resulting from spills,
such as petroleum products.10
We emphasize that, under environmental and especially
agricultural conditions, absolute evaporation rates from soil
2014 American Chemical Society

and vegetation surfaces are strongly inuenced by a variety of

factors, such as soil or vegetation properties, solar radiation,
humidity, and wind speed. Additional factors inuence
evaporation when the liquid lm is very thin and sorptive
interactions with the surface inuence volatility. The proposed
correlation is not intended to address such conditions.
Here, we discuss the theoretical basis of correlations of the
evaporation rate versus vapor pressure and suggest an improved
and simpler approach based on a sounder theoretical treatment of the
evaporation process. The proposed correlation is regarded as being
particularly suitable for assessing the relative, as distinct from absolute,
rates of evaporation of a variety of liquids and can contribute to
selection of substances on the basis of evaporation potential.

THEORY
In this derivation, we use SI units of kilograms, meters, and
seconds and molar mass as kg/mol. The pressure is expressed
in Pascals with fundamental units of kg m1 s2. If a pure liquid
of molar mass M (kg mol1) and vapor pressure P (Pa)
(corresponding to the ambient temperature) is present on a
solid, non-absorbing surface, it can be assumed that the air
immediately in contact with the liquid surface achieves a partial
pressure of P (Pa). Applying the ideal gas law, this corresponds to
a concentration of P(RT)1 (mol m3) or PM(RT)1 (kg m3) (also
referred to as the saturated vapor concentration at temperature T),
where R is the gas constant (8.314 Pa m3 mol1 K1) and T is the
absolute temperature.
The evaporation rate can be estimated as the product of the
saturated vapor concentration and a mass-transfer coecient or
velocity k (m s1), which is essentially the velocity at which the
saturated air layer is conveyed from the surface. This velocity can
Received:
Revised:
Accepted:
Published:
10259

June 15, 2014

August 5, 2014
August 8, 2014
August 8, 2014
dx.doi.org/10.1021/es5029074 | Environ. Sci. Technol. 2014, 48, 1025910263

Environmental Science & Technology

Article

Table 1. Properties of the Chemicals at 25 C, Reported Evaporation Rates and Rate Coecients Calculated Using Units
Employed in Equation 5
chemical name

VP (Pa)

MW (g/mol)

ER (kg m2 s1)

Emass/(P MW)

acetone
benzene
isobutyl acetate
n-butyl acetate (99%)
sec-butyl acetate (90%)
isobutyl alcohol
n-butanol
isobutyl isobutyrate
cyclohexanol
cyclohexanone
diethylene glycol monobutyl ether
diethylene glycol monoethyl ether
diethylene glycol monomethyl ether
diacetone alcohol
diethyl ketone
diisobutyl ketone
ethyl acetate
ethanol (100%)
ethyl amyl ketone
ethylbenzene
ethylene glycol monobutyl ether
ethylene glycol monoethyl ether
ethyl lactate
n-hexane
isophorone
mesityl oxide
methanol
methyl ethyl ketone
methyl isobutyl ketone
methyl isopropyl ketone
methyl n-propyl ketone
nitroethane
nitromethane
1-nitropropane
n-octane
n-propyl acetate
isopropyl alcohol
n-propyl alcohol
tetrahydrofuran
toluene
p-xylene
dodecane
n-octanol
tridiphane
triuralin
pendimethalin
2,4-D
diazinon
toxaphene
dieldrin
pp-DDT
mean

3.08 10
1.27 104
2.41 103
1.54 103
2.95 103
1.53 103
9.46 102
6.27 102
9.20 101
5.39 102
2.80 100
2.00 101
3.75 101
1.65 102
4.71 103
2.23 102
1.26 104
7.86 103
2.60 102
1.28 103
7.29 101
7.66 102
5.00 102
2.02 104
5.73 101
1.46 103
1.70 104
1.21 104
2.69 103
6.74 103
4.72 103
2.80 103
4.77 103
1.37 103
1.85 103
4.49 103
5.69 103
2.83 103
2.17 104
3.80 103
1.17 103
1.23 101
1.73 101
2.93 102
1.47 102
4.00 103
2.67 103
1.49 103
5.33 104
6.59 104
4.40 105

58.08
78.11
116.16
116.16
116.16
74.12
74.12
144.22
100.16
98.15
204.27
134.18
120.15
116.16
86.13
142.24
88.11
46.07
128.22
106.17
104.11
132.16
118.13
86.18
138.21
98.15
32.04
72.11
100.16
86.13
86.13
75.07
61.04
89.09
114.23
102.13
60.1
60.1
72.11
92.14
106.17
170.34
130.23
320.43
335.29
281.31
221.04
304.35
413.82
380.91
354.49

5.04 102
3.62 102
1.36 102
8.94 101
1.62 102
4.50 101
3.19 101
4.30 101
7.05 100
2.89 101
2.55 101
8.41 101
1.09 100
8.20 100
1.93 102
1.60 101
4.00 102
1.19 102
2.24 101
6.82 101
3.03 100
2.38 101
1.80 101
5.87 102
2.17 100
7.23 101
1.37 102
3.40 102
1.35 102
2.37 102
1.91 102
1.21 102
1.57 102
7.62 101
1.17 102
1.92 102
7.56 101
6.00 101
4.68 102
1.82 102
7.00 101
2.75 101
3.75 101
1.98 103
4.41 104
2.89 104
1.86 104
2.52 104
5.53 105
4.78 105
1.27 105

1015
1315
1744
1795
1709
1426
1635
1712
2755
1967
1602
1128
872
1537
1710
1815
1298
1181
2424
1812
1438
845
1096
1215
988
1820
905
1401
1799
1472
1690
2078
1936
2247
1993
1504
796
1271
1078
1873
2024
474
599
759
322
924
1135
2004
902
685
2935
1463.9

also be regarded, as in Ficks law, as a ratio of the diusivity in the

air layer D (m2 s1) and the diusion path length Y (m); i.e., k is
DY1. The evaporation rate E can then be expressed on either a
molar or mass basis as
Emolar = Pk(RT )1

(mol m2 s1)

or
Emass = PMk(RT )1

(kg m2 s1)

(2)

There is a clear near-linear dependence of E upon P as exploited

in the successful correlations described earlier. An issue then
arises about the selection of the units of E. If E is expressed as a

(1)
10260

dx.doi.org/10.1021/es5029074 | Environ. Sci. Technol. 2014, 48, 1025910263

Environmental Science & Technology

Article

mass ux (Emass), such as kg m2 s1, or a similar quantity, then

the correlation should be with the chemical property PM and not
P. However, if E is expressed in units of mol m2 s1 (Emolar), then
the correlation should be with P alone. Whether correlating Emolar
[Pk(RT)1] with P or correlating Emass [PMk(RT)1] with PM,
care must be taken to use consistent SI units.
The Woodrow et al.1,2 and van Wesenbeeck et al.5 correlations
are in principle of Emass versus P and not PM. The slopes of the
correlations of log E versus log P are consistently less than 1.0,
ranging from 0.85 to 0.93 in both studies. This may be due in part
to the absence of M in the correlation. Substances of low vapor
pressure tend to have higher molar masses; thus, correlating the
evaporation rate in mol m2 s1 instead of kg m2 s1 has the eect
of lowering the evaporation rate more for less volatile chemicals,
thus increasing the slope. To test this assertion, we modify the
correlation to express E as mol m2 s1 (i.e., Emolar) and determine
if the resulting slope of the plot is closer to 1.0. If the slope is close
to 1.0, there may be no need to use log quantities and a simple
correlation of Emolar versus P can be suggested, in which the single
parameter is the slope and there is no intercept.
The slope can be determined in two ways. The simpler way is
to determine the average value of the ratio Emolar/P or
equivalently Emass/PM. Alternatively, a least-squares regression
can be used. Both methods should give similar results. Any
systematic dependence of the slope or ratio can be revealed by a
plot of the ratio Emolar/P versus P.
When the van Wesenbeeck et al.5 correlations were developed,
three data sets were used and regressed separately. The regression
coecients were similar, and the parameters dierences were
statistically insignicant. Accordingly, the standard procedure of
dividing the data into a training and validation set was not needed.
In essence, the correlation involves obtaining one tted parameter
that has a sounder theoretical basis.

versus ln P for the compounds used in this study resulted in a

slope of 0.93 and intercept = 12.3 (gure not shown), very similar
to the regression parameters obtained for the data set with 82
compounds reported by van Wesenbeeck et al.5 This is expected
because the 51 compounds studied here are a subset of the 82
compounds. A plot of the regression of ln Emolar versus ln P is
given in Figure 1. Changing from mass to molar unit rates

Figure 1. Plot of ln(Emolar) versus ln(P).

increases the slope to 1.02 (r2 = 0.99). This is much closer to 1.0
than the regression of ln Emass versus ln P in the studies by
Woodrow et al.1,2 and van Wesenbeeck et al.,5 where the slope
was consistently between 0.87 and 0.93. This suggests that the
concept of expressing E on a molar basis does indeed compensate
for the eects of lower volatility of high-molecular-weight
compounds and increases the linearity of the relationship. The
regression of ln Emass versus ln(PM) results in an identical slope
(1.02) and a negligible intercept and is not shown. As noted
above, the fact that the slope of the line is very close to 1.0
obviates the need for performing the logarithmic transformation
on E and P. As a result, Emass can be plotted against PM (Figure 2)

METHODS

The data set used in this study contains 51 of the 82 compounds

from the data sets analyzed by van Wesenbeeck et al.5 for which
molar mass (M) information could be easily obtained. The other
31 compounds were commercial products of uncertain molecular
structure and identity, being subject to commercial condentiality. The data at 25 C for evaporation rate E (g m2 h1) and
vapor pressures reported in that study were used directly and
expanded to include molar mass (M), and E was recalculated on a
molar basis (Emolar) using eq 1 with units of mol m2 s1 and
plotted against P. Alternatively and equivalently, as shown in the
previous section, the mass evaporation rate (Emass) with units of
kg m2 s1 using eq 2 could be plotted against PM (Pa kg mol1)
or PM(RT)1 (kg m3), with the latter quantity being the vapor
density of the substance in the air phase. Both approaches were
tested.
It is noteworthy that, in the ASTM D3539 test,4 the
evaporation rate of the liquid is measured from a lter paper
disk 9 cm in diameter using a liquid volume of 0.7 cm3; thus, the
liquid lm has a thickness of approximately 0.1 mm, which is
sucient that any sorptive inuences from the substrate are
negligible. Under agrochemical conditions of high soil surface
area and low application rates, such as kg/ha, the lms are much
thinner and sorption can be a controlling factor.

Figure 2. Plot of Emolar versus P.

or Emolar could be plotted against P to obtain the relationship

between the two. If a slope of 1.0 is forced, the ratio of Emass to PM
or the ratio of Emolar to P can be regarded as constant and a simple
one-parameter correlation results.
This one parameter is most readily estimated as either the
mean ratio Emass/(PM) or the identical ratio Emolar/P. The
average value of this ratio is 4.07 107, and the standard
deviation is 1.55 107; thus, there is 95% condence that the

RESULTS
Table 1 contains the chemical parameters at 25 C for the 51
compounds used in this study. A linear regression of ln Emass
10261

dx.doi.org/10.1021/es5029074 | Environ. Sci. Technol. 2014, 48, 1025910263

Environmental Science & Technology

Article

value of the ratio lies between 3.6 107 and 4.5 107. There is
a trend for the 9 compounds with vapor pressures above 5000 Pa
to exhibit a lower slope. To explore if there is a systematic
dependence upon P, the ratio is plotted against P for the 51
chemicals in Figure 3, conrming this trend, but the overall eect

proposed here can be used as a reality check that estimated

values or values obtained from eld studies are of the correct
order of magnitude. In their models of pesticide evaporation,
Davie-Martin et.al.11 expressed the ux correctly, using PM
rather than P only.
The eect of the temperature is readily assessed by adjusting
the vapor pressure using the ClapeyronClausius equation. An
advantage of the linear relationship is that doubling P doubles
ER, whereas this does not apply if the loglog slope is less than
1.0. It is often convenient to express evaporation rates as halftimes or rate constants; however, evaporation from a pure liquid
surface is a zero-order process, and half-times can only be
estimated if the quantity of the evaporating substance is known.
For a dened quantity of liquid, the time for complete evaporation
is (quantity of liquid, g m2)/(evaporation rate, g m2 h1).
Comparing eqs 1 and 5 to the tted parameter of 4.07 107
shows that this parameter is a mass-transfer velocity or coecient
k (m s1) divided by RT; thus, k is 4.07 107RT. At 20 C, k is
then 9.91 104 m s1 or 3.57 m h1, which is in fair agreement
with measured mass-transfer coecients in the air phase at low
wind speed conditions (<1 m s1), which range from 3.6 m h1 12
to 10.8 m h1 13 to 18 m h1.14 These values apply to wind tunnel
or outdoor conditions at near-zero wind velocities rather than
laboratory conditions in the ASTM evaporometer cabinet of a
very low air ow rate of 0.048 L/min, corresponding to an
approximate velocity of 0.075 m h1. If information is available
on the coecient expected under specic conditions, then the
slope coecient can be adjusted accordingly. This applies only to
volatilization of the pure uid substance and does not apply if
there is sorption to soils or other substrates or if the chemical is in
solution. In such cases, the correlations of Woodrow et al.1,2 are
superior.
A possible application of the correlation is to the assessment
and selection of hydrocarbon carrier uids used in emulsiable
concentrates in pesticide formulations, as part of eorts to
minimize ozone formation potential (OFP), as discussed by
Zeinali et al.6 The proposed California Department of Pesticide
Regulation cited in that study is a framework to evaluate the OFP
of pesticide formulations. Essentially, this is calculated from the
mass of product applied, the mass fraction that volatilizes, and the
incremental OFP reactivity of the hydrocarbon. The fraction
volatilized is estimated by thermogravimetric analysis (TGA) of
the product under dened conditions to a nal temperature of
115 C. This necessarily involves extrapolation from 115 C to
ambient temperatures and, as discussed,6 is not a straightforward
method and could be misleading. A more direct method is to use
the volatilization rate directly, as estimated from a known vapor
pressure using the present correlation. Dierent hydrocarbons
and other uids can then be compared directly for volatilization
potential at a relevant temperature. Because the slope coecient
has physical signicance as a mass-transfer coecient (MTC),
the environmental volatilization rate could be determined from
an estimated or measured MTC under eld conditions. This
more direct method is likely to be more accurate, thus avoiding
TGA and subsequent extrapolation. The correlation does not
strictly apply if the liquid is a mixture because the ER depends
upon the changing composition of the chemical species present
and their vapor pressures and activity coecients. If, however,
the components are of similar structure and volatility, as is the
case with certain petroleum distillates, the average evaporation
rate can be measured and estimated. An example is the
pioneering 1986 study of evaporation of weed oil mixtures by
Woodrow et al.15

Figure 3. Plot of Emolar/P versus P. Mean ratio = 4.07 107.

is not judged to be signicant. The proposed equivalent

correlations are as follows:
Emolar (mol m2 s1) = 4.07 107P (Pa)

(3)

Emass (kg m2 s1) = 4.07 107PM (Pa kg mol1)

(4)

A more convenient version for agrochemical, indoor, and other

purposes is
Emass (g m2 h1) = 1464P (Pa) M (g mol1)

(5)

In eq 5, the mean slope coecient of 1464 shown in Table 1 is

modied by a factors of 109 (g/kg), 3600 (s/h), and 103 kg/g
and the molar mass M is now expressed in Table 1 in the
common units of g mol1. The higher vapor pressure
compounds, including volatile liquids, such as acetone, show
reduced evaporation rates, probably because of evaporative
cooling of the liquid surface, i.e., the wet bulb eect.

DISCUSSION
The methodology proposed in this study is useful in that it
substantially reduces the nonlinearity of the E versus P
relationship by either expressing E for the chemical on a molar
basis (Emolar) and regressing against P or alternatively by
expressing E on a mass basis (Emass) and regressing against PM.
This eliminates the need for logarithmic transformations of the
E and P data (which automatically require two parameters) and
allows for a simple and robust one-parameter linear regression of
non-log-transformed parameters for estimating E using readily
obtained vapor pressures. Further, the single parameter is shown
to have signicance as a mass-transfer coecient or velocity,
which is not possible with a logarithmic regression.
Under environmental conditions, the evaporation rate is also
inuenced by wind speed, temperature as inuenced by solar
radiation and the wet bulb eect, fetch, partitioning between
soil and water and soil and air, and gas diusivity of the chemical
species. Woodrow et al.1,2 and Davie-Martin et al.11 have shown
that including partitioning parameters (KOW, KOC, and KOA)
improves the correlation between chemical properties and
cumulative evaporation; however, including these quantities
greatly complicates the correlation. The simple correlation
10262

dx.doi.org/10.1021/es5029074 | Environ. Sci. Technol. 2014, 48, 1025910263

Environmental Science & Technology

Article

Under indoor conditions, air ow rates are generally slow, thus

cleaning products and solvents applied to surfaces evaporate
under fairly quiescent conditions to which the present
correlation may apply. The coecients can be increased by an
appropriate factor to reect increased rates of air circulation, but
this does not apply if there is a measurable air velocity. The time
required for a liquid coating to evaporate completely can be
readily estimated by dividing the quantity applied (g m2) by
the evaporation rate (g m2 h1).

AUTHOR INFORMATION

Corresponding Author

*Telephone/Fax: 705-740-2911. E-mail: [email protected].

Notes

The authors declare no competing nancial interest.

REFERENCES

(1) Woodrow, J. E.; Seiber, J. N.; Dary, C. Predicting pesticide

emissions and downwind concentrations using correlations with
estimated vapor pressures. J. Agric. Food Chem. 2001, 49, 38413846.
(2) Woodrow, J. E.; Seiber, J. N.; Baker, L. W. Correlation techniques
for estimating pesticide volatilization flux and downwind concentrations. Environ. Sci. Technol. 1997, 31, 523529.
(3) Guth, J. A.; Reischmann, F. J.; Allen, R.; Arnold, D.; Hassink, J.;
Leake, C. R.; Skidmore, M. W.; Reeves, G. L. Volatilisation of crop
protection chemicals from crop and soil surfaces under controlled
conditionsPrediction of volatile losses from physicochemical properties. Chemosphere 2004, 57, 871887.
(4) ASTM International. ASTM D3539-87. Evaporation Rates of
Volatile Liquids by Shell Thin-Film Evaporometer; ASTM International:
West Conshohocken, PA, 1987.
(5) van Wesenbeeck, I. J.; Driver, J.; Ross, J. Relationship between the
evaporation rate and vapor pressure of moderately and highly volatile
chemicals. Bull. Environ. Contam. Toxicol. 2008, 80, 3153.
(6) Zeinali, M.; McConnell, L.; Hapeman, C.; Nguyen, A. Volatile
organic compounds in pesticide formulations: Methods to estimate
ozone formation potential. Atmos. Environ. 2011, 45, 24042412.
(7) Bennett, D. H.; Furtaw, E. J., Jr. Fugacity-based indoor residential
pesticide fate model. Environ. Sci. Technol. 2004, 38 (7), 21422152.
(8) Singer, B. C.; Destaillats, H.; Hodgson, A. T.; Nazaroff, W. W.
Cleaning products and air fresheners: emissions and resulting
concentrations of glycol ethers and terpenoids. Indoor Air 2006,
No. 16, 179191.
(9) Montemayor, B. P.; Price, B. B.; van Egmond, R. A. Evaporative fate
of cyclopentasiloxane (D5) from personal care products during product
use: Antiperspirants, skin care products and hair care products.
Chemosphere 2013, 93, 735740.
(10) Fingas, M. F. Modeling oil and petroleum evaporation. J. Pet. Sci.
Res. 2013, 2 (3), 104115.
(11) Davie-Martin, C. L.; Hageman, K. J.; Yu-Ping Chin, Y.-P. An
improved screening tool for predicting volatilization of pesticides
applied to soils. Environ. Sci. Technol. 2013, 47, 868876.
(12) Mackay, D.; Yuen, A. T. K. Mass transfer correlations for
volatilization of organic solutes from water. Environ. Sci. Technol. 1983,
17, 211216.
(13) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M.
Environmental Organic Chemistry; John Wiley and Sons, Inc.: Hoboken,
NJ, 1993.
(14) Liss, P. S. Processes of gas exchange across an airwater interface.
Deep Sea Res. 1973, 20, 22238.
(15) Woodrow, J. E.; Seiber, J. N.; Kim, Y.-N. Measured and calculated
evaporation losses of two petroleum hydrocarbon herbicide mixtures
under laboratory and field conditions. Environ, Sci. Technol. 1986, 20,
783789.

10263

dx.doi.org/10.1021/es5029074 | Environ. Sci. Technol. 2014, 48, 1025910263