2576 Ind. Eng. Chem. Res.

1991,30, 2576-2582
The fall in overall rate constant value may be attributed process with well-defined kinetics. The control of ring
to the fact that at higher acid concentration the side re- expansion is possible by judicious choice of reaction tem-
actions (viz., furanization and some ring-opening reactions perature and acid concentration. Total control of acido-
(Gelling, 1983; Burfield et al. (1984)) predominate over the lytic ring opening of the epoxy group is possible. From
epoxidation process, resulting in a sharp fall in the overall the experimental curves (Figure 6) it is possible to predict
rate constant value of epoxidation. the overall rate constant and maximum extent of ep-
The maximum extent of epoxidation was also found to oxidation for a given acid concentration.
vary with the acid concentration (Figure 6). When the Registry No. Formic acid, 64-18-6; acetic acid, 64-19-7.
concentration of the acid was very high, e.g., 5 mol/dm3,
the maximum extent of epoxidation was about 40%, and Literature Cited
when the acid concentration was 3.6 mol/dm3, the maxi-
mum yield was about 57%. When reactions were contin- Badran, B. M.; El-Souckary, M. A.; Aggour, S. H. In Situ Ep-
oxidation of Elastomers. Elastomerics 1980, April, 40-42.
ued beyond these maxima, the products contained mainly Burfield, D. R., Lim, K. L.; Law, K. S. Epoxidation of Natural
hydrofuran rings and ring-opened groups such as formate Rubber Lattices: Methods of Preparation and Properties of
and hydroxyl. Modified Rubbers. J. Appl. Polym. Sci. 1984, 29, 1661-1673.
When the acid concentration was about 2.3 mol/dm3, Davey, J. E.; Loadman, M. J. R. A Chemical Demonstration of the
70% epoxidation was achieved. However, further decrease Randomness of Epoxidized Natural Rubber. Br. Polym. J. 1984,
in the acid concentration resulted in a much slower rise 16, 134.
Gelling, I. R. Epoxidized cis-1,4-Polyisoprene Rubber. Br. Pat. GB
in the level of maximum epoxidation. This observation 2,113,692, 1983.
may be explained by the phenomena of hydrofuranization Gelling, I. R. Modification of Natural Rubber Latex with Peracetic
and ring opening of epoxy groups, both of which were Acid. Rubber Chem. Technol. 1985,58,86-96.
found to be accelerated in the presence of excess acid Mairs, J. A.; Todd, J. The oxidation of Caoutchouc, Gutta-percha
(Perera et al., 1988; Burfield et al., 1984). It was also and Belata with Hydrogen Peroxide. J. Chem. SOC.1932,386-399.
explained by Gelling (1985) and Perera et al., (1988) that Ng, S. C.; Gan, L. H. Reaction of Natural Rubber Latex with Formic
Acid. Eur. Polym. J. 1981, 17, 1073-1077.
the higher the block length of epoxy groups the greater is Pummer, R.; Burkhard, P. A. Uber Kautschuk. Ber. Dtsch. Chem.
the tendency of furanization, and this was the reason that Ges. 1922,55, 3458-3472.
the epoxidation rate at higher than 70% was very slow Perera, M. C. S.; Elix, J. A.; Bradbury, J. H. Furanized Rubber
even by further lowering of acid concentration. Studied by NMR SDectroscow. _ - J. Polvm. Sci.,. Polvm.
- Chem. Ed.
The point of intersection of the two curves at the highest 1988, 26, 637-651.'
position i.e., corresponding to 3.3 mol/dm3 acid concen- Roux, C.; Pautrat, R.; Cheritat, R.; Ledran, F.; Danjard, J. C. Mod-
ification de Polymeres Insatures par Epoxydation. J. Polym. Sci.,
tration, should give the highest rate of conversion with the Part C 1969, 16, 1687.
lowest extent of side reaction. However, experimentally Saffer, A,; Johnson, B. L. Measurement of Internal Double Bonds
k was found to be 1.166 X m i d and (CE)max/CRo to in Polymers by Perbenzoic Acid Addition. Ind. Eng. Chem. 1948,
be 0.634 with a negligible amount of hydroxyl and about 40, 538.
5% hydrofuran rings, but there was no IR absorption due Terry, R. W.; Jacobs, A. F. Modified Diene Polymers. Br. U.K. Pat.
to formates (1720-1725 cm-'). Appl. 2008,125, 1979.
Tobalsky, A. V.; Mesrobion, R. B. Organic Peroxides; Interscience:
Conclusion New York, 1954; p 35.
In conclusion, it can be summarized that the epoxidation Received for review January 22, 1991
of NR with performic acid produced in situ is a feasible Accepted July 1, 1991

Group Contribution Method for the Prediction of Liquid Densities as a

Function of Temperature for Solvents, Oligomers, and Polymers
Helle S. Elbro, Aage Fredenslund,* and Peter Rasmussen
Engineering Research Center, IVC-SEP, Department of Chemical Engineering, The Technical University of
Denmark, DK-2800 Lyngby, Denmark

A new group contribution method, GCVOL, for the prediction of liquid densities as a function of
the temperature is proposed. A table for 36 different group volume parameters is given. Using the
proposed method liquid densities for even strongly polar solvents can be predicted with a n error
of approximately 1% in the temperature range between the melting temperature and the normal
boiling point. Densities of amorphous polymers are predicted with an average error of 2.7%. T h e
proposed method is able to describe densities for the complete molecular range, covering solvents,
oligomers, and polymers.

Introduction well-known techniques, do not have any meaning. Group

Most methods for estimating saturated liquid densities contribution, on the other hand, can be used to estimate
are based upon the law of corresponding states or upon polymer densities. Unfortunately existing group contri-
group contributions. The estimation of density using the bution tables are usually only valid either for polymers o r
principle of corresponding states generally requires for low-molecular-weightsolvents. No method exists which
knowledge of the critical properties. For polymers, critical can successfully predict densities for oligomers (with mo-
properties, even though they can be estimated using lecular weights in between solvents and polymers). Fur-
0888-5885/91/2630-2576$02.50/0 @ 1991 American Chemical Society
 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991 2577
thermore, most group contribution tables for predicting It was especially developed to be used for calculating the
saturated densities are only given at one temperature solubility parameter for polymers. The density of solventa
(usually 25 "C). and polymers is calculated using eq 2. Fedors lists a very
The purpose of this work is to develop a group contri- large table of group volume increments, also covering a
bution method to predict densities as a function of tem- number of inorganic metals. Fedors does not list any
perature for the whole molecular weight range, covering comprehensive analysis of the accuracy of his method.
solvents, oligomers and polymers.
GCVOL: The New Group Contribution Method
Background for Predicting Saturated Liquid Density
Corresponding-State Models. Various models exist Model. We propose the following simple model for
for predicting saturated liquid densities as a function of predicting the molar volume of a liquid:
the temperature using the principle of corresponding V = CniAui (4)
states. Traditionally these models have been used in either
of two forms: the Rackett equation (Rackett, 1970) and where ni is the number of group i , and the temperature
the method of Bhirud (1978) are examples of predictive dependency of the molar group volume, Aui, is given by
methods that only depend on critical properties or some the following simple polynomial function:
other commonly tabulated property such as the normal
boiling temperature. The modified Rackett equation
+
A u ~= Ai + BiT C i P (5)
(Spencer and Danner, 1972) and the HBT method (Han- where the temperature, T/K, can vary between the melting
kinson and Thomson, 1979), on the other hand, both re- point and the normal boiling point if the model is used to
quire an extra characteristic parameter adjusted to already predict densities of solvents, and between the glass tran-
known density data. In the latter case one can often sition temperature and the degradation temperature if
predict the density with errors below 1%, whereas the used to predict density data for amorphous polymers. To
errors in the first case can get as high as 5-10%. The predict density data for polymers, only groups present in
difference between these two approaches lies more in the the repeat unit of the polymer need to be included in eq
use of some experimental density data than in differences 4, whereas all groups used to define the chemical formula
in the functional form of the predictive model. Our model of low-molecular-weightsolvents and oligomers have to be
is as the Rackett equation a purely predictive model. included. Group volume temperature constants are given
In the Rackett equation (Rackett, 1970), the saturated for 36 different groups in Table I.
liquid volume, V, is found from the following equation: Determination of Group Volumes. Group volumes
were optimized independently at temperatures spaced 10
v = v~~(1-77,)*/7 (1) K apart in the temperature range 200-500 K using liquid
where 2, is the critical compressibility, V, is the critical volume data for solvents below their normal boiling tem-
volume, and T,is the reduced temperature. In the cal- perature taken from the DIPPR data compilation (Daubert
culation of the V, we use critical data taken from the and Danner, 1989).
DIPPR compilation (Daubert and Danner, 1989). In the following paragraph the mathematical framework
Group Contribution Methods. The best known group for the calculation of the group volumes will be given. The
contribution method for predicting the density of amorp- below-mentioned procedure avoids nonlinear optimization,
hous polymers is the one given by Van Krevelen (Van which otherwise would be the obvious choice. Using
Krevelen and Hoftyzer, 1972). The density, p, of a polymer normal statistics the following equation can be set up for
is calculated from the following simple equation: each temperature:
MW MW CT-W-GAV = dr-WW"P (6)
p=-=- (2)
v CniAui where the matrix G contains the group specification of each
solvent included in the optimization, one solvent for each
where the molecular weight, MW, and the molar volume, row (GTis the transposed matrix); Av is the molar group
V, both apply to the repeat unit of the polymer. The molar volume parameter vector; vexPis the experimental molar
volume is calculated from a sum over all group volume volume; and W is a diagonal matrix, containing the
increments, Au? Van Krevelen gives a table of 20 different weighing factors used for each pure component. These
group volume increments encompassing alkane, aromatic, factors were obtained from the errors in the experimental
ether, ester, chloride, fluoride, and sulfur groups. Van volumes listed in the DIPPR data compilation (Daubert
Krevelen assumed that it was impossible to use the same and Danner, 1989);high errors lead to low weighing factors.
table of group volume increments both for polymers and This equation reduces to a quadratic set of linear equa-
for solvents, and the given group volume increments were tions, where the number of unknowns corresponds to the
therefore optimized from experimental densities at 25 "C number of molar group volumes to be found. These
for 40 amorphous polymers. The error in reproducing equations can be solved using simple Gauss elimination.
these densities using the set of optimized group volume The resulting temperature-dependent molar group
increments is 1.5%. volumes, Av, were then fitted to a polynomial function (eq
The temperature dependency of the polymer densities 5) using the lowest degree possible.
is described by Van Krevelen quite empirically by the The primary groups to be determined were the alkyl
following equation: groups, which were determined from 56 different noncyclic
V(t) = V(25 "C)+ ( t - 25 "C)V, X (3) alkanes. Secondly, the arene groups were determined from
37 different alkylbenzenes. The remaining groups were
where V(t) is the molar volume at the temperature t , V(25 determined from solvents, which could be defined using
"C) is the molar volume at 25 "C calculated using the group only alkyl or arene groups and one additional functional
volume increments, and V, is the van der Waals volume group.
as listed by Bondi (1968). Selection of Temperature Range. The range 2OMOO
Another group contribution method for calculating the K was selected because many solvents are in liquid form
density of liquids at 25 "C is the method of Fedors (1974). within this range.
2578 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991
20.0 6.0
(A A
19.0 - I
No restriction in temperature range
-
.2. 5.0
18.0 - ._
0 No restriction in temperature
$; 0
.-
E 17.0 -
v $n 4.0
D o . ~ n ~ ~ o ~ .
o m n o
2 16.0 - . a
a a 0 '
0

b 15.0 -
L
g3.0

ll.O
10.0
1 ~ I , , I I I I I , I , , , / , , , , , I / < , , , , , ,500.0
,I 0.0
200.0 300.0 400.0 200.0 300.0 400.0 50
Temperature, K Temperature, K

20.0 6.0
B j B
19.0 Temperature less than Tb

I
b 15.0
L

a 13.0 ;!
._

P
2.0

E!!14'012.0
ll.O 1 0.0 .~
200.0 300.0 400.0 500.0 200.0 300.0 400.0 50
Temperature, K Temperature, K
Figure 1. Group volume for CH2 as a function of the temperature. Figure 2. Errors in the prediction of the liquid volumes as a func-
(A) All liquid solvent data contributed to the determination of the tion of the temperature. (A) All liquid solvent data contributed to
group volume. (B) Only solvents below their normal boiling tem- the determination of the group volume. (B)Only solvents below
perature contributed to the determination of the group volume. their normal boiling temperature contributed to the determination
of the group volume.
At first no restriction was made regarding the state of
the solvent; Le., the solvent could be close to the critical better correlation of the data. On the basis of this in-
point and its volume data would still be included in the vestigation it was therefore decided only to include volume
determination of the corresponding group volumes. data below the normal boiling point into the determination
However, optimization of the alkane group volumes (CH3, of the corresponding group volumes.
CH2, CH, and C) showed that this simple group contri- Basis of Group Volumes. The group volume incre-
bution scheme could not be used in the complete range up ment vtilues wili always to some extent be correlated. A
to the critical temperature. This can be seen in Figure 1, basis was therefore established using density data for a
which shows the molar group volume for. the CH2 group linear polyethylene since it only contains CH2 groups, the
as a function of the temperature under two different CH, end-groups making insignificant contributions. The
conditions: In Figure 1A volume data for alkanes in the temperature dependence of the CH2 group between 140
near-critical region have been included and in Figure 1B and 180 OC was found from density data for a completely
only volume data for alkanes below their normal boiling amorphous polyethylene, and subsequently the three al-
temperature have been included in the determination of kane group volumes (CH,, CH, and C) were adjusted to
the CH2 group volume. In Figure 2 the corresponding density data for low-molecular-weight alkanes.
errors in the prediction of the alkane volumes are given Using the polyethylene data seems legitimate since it
for the same two conditions. One would expect the molar results in a physically realistic value for the CH2 group
volume for the CH2 group to increase as a function of the volume. This can be seen from Figure 3, where the CH2
temperature; therefore the results from condition A do not group volume at 20 "C is given for a series of cycloalkanes
appear physically valid. Furthermore, the error in pre- as a function of the number of carbon atoms. As the
dicting the alkane volume data for condition A is an in- number of carbon atoms in the cycloalkane increases, the
creasing function of temperature. For condition B the cycloalkane loses the strained conformation which is
errors are constant until around 400 K, where it drops dominant especially for the lower cycloalkanes, and the
below 1%. A drop in the error can readily be understood, molar volume for a CH2unit goes asymptotically toward
since higher temperatures will result in fewer alkanes being the value we have selected using the data for linear poly-
below their normal boiling temperature, and therefore ethylene. From Figure 3 it is also clear why the proposed
 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991 2579

Table I. Group Contributions for Saturated Molar Volume

group volume temp const
A, IO~B, 105c,
no. cm3/mol cm3/(mol K) cm3/(mol K2) sample ~ O U Dassimment
1 18.960 45.58 0 butane: 2 CH,, 2 CH2
2 12.520 12.94 0 2-methylpropane: 3 CH,, 1 CH
3 6.297 -21.92 0 2,2-dimethylpropane: 4 CH,, 1 C
C 4 1.296 -59.66 0
ACH 5 10.090 17.37 0 benzene: 6 ACH
ACCH, 6 23.580 24.43 0 toluene: 5 ACH, 1 ACCH,
ACCHz 7 18.160 -8.589 0 ethylbenzene: 1 CH,, 5 ACH, 1 ACCH,
ACCH 8 8.925 -31.86 0 cumene: 2 CH,, 5 ACH, 1 ACCH
ACC 9 7.369 -83.60 0 tert-butylbenzene: 3 CHI, 5 ACH, 1 ACC
CH,= 10 20.630 31.43 0 1-hexene: 1 CH,, 3 CH,, 1 C H p , 1 CH=
CHI 11 6.761 23.97 0 2-hexene: 2 CH,, 2 CHz, 2 CH=
C= 12 -0.3971 -14.10 0 2-methyl-2-butene: 3 CH,, 1 CH=, 1 C=
CHzOH 13 39.460 -110.60 23.31 1-hexanol: 1 CH,, 4 CH,, 1 CHzOH
CHOH 14 40.920 -193.20 32.21 2-hexanol: 2 CH,, 3 CH,, 1 CHOH
ACOH 15 41.200 -164.20 22.78 phenol: 5 ACH, 1 ACOH
CH&O 16 42.180 -67.17 22.58 methyl ethyl ketone: 1 CH,, 1 CH,, 1 CHSCO
CHzCO 17 48.560 -170.40 32.15 diethyl ketone: 2 CH,, 1 CH2, 1 CHzCO
CHCO 18 25.170 -185.60 28.59 diisopropyl ketone: 4 CH,, 1 CH, 1 CHCO
CHO 19 12.090 45.25 0 I-hexanal: 1 CH,, 1 CH,, 1 CHO
CHSCOO 20 42.820 -20.50 16.42 butyl acetate: 1 CH,, 3 CH2, 1 CH3CO0
CHzCOO 21 49.730 -154.10 33.19 butyl propionate: 2 CH,, 3 CH2, 1 CHzCOO
CHCOO 22 43.280 -168.70 33.25 methyl isobutyrate: 3 CH,, 1 CHCOO
coo 23 14.230 11.93 0 poly(methy1methacrylate): 2 CH,, 1 CHz, 1 C, 1 COO
ACCOO 24 43.060 -147.20 20.93 methyl benzoate: 1 CH,, 5 ACH, 1 ACCOO
CH,O 25 16.660 74.31 0 ethyl methyl ether: 1 CH,, 1 CH,, 1 CH,O
CHZO 26 14.410 28.54 0 diethyl ether: 2 CH,, 1 CH,, 1 CHzO
CHO 27 35.070 -199.70 40.93 diisopropyl ether: 4 CH,, 1 CH, 1 CHO
co 28 30.120 -247.30 40.69 di-tert-butyl ether: 6 CH,, 1 C, 1 CO
CHzCl 29 25.29 49.11 0 1-chlorobutane: 1 CH,, 2 CH,, 1 CHzCl
CHCl 30 17.40 27.24 0 2-chloropropane: 2 CH,, 1 CHCl
CCl 31 37.62 -179.1 32.47 2-chloro-2-methylpropane:3 CH,, 1 CCl
CHC1, 32 36.45 54.31 0 1,l-dichloroethane: 1 CH,, 1 CHCl,
CCl, 33 48.74 65.53 0 l,l,l-trichloroethane: 1 CH,, 1 CCl,
ACCl 34 23.51 9.303 0 chlorobenzene: 5 ACH, 1 ACCl
Si 35 86.71 -555.5 97.90 hexamethyldisiloxane: 6 CH,, 1 SiO, Si
Si0 36 17.41 -22.18 0 poly(dimethylsi1oxane): 2 CH3, 1 Si0

25.0

I"
0
19.0 1 Cycloalkanes 20.0

15.0
1
4

4
1 -HEXANOL

15.0 15 I
6
I
7
I
8 9
I
10
I I
11
Number of carbons in cycloolkane
I
12 13
-5.0
O'OI
0.0
, , ,
0.5
________________
______-_____--_______________________

Reduced temperature, T,
,
1 .o
I
Figure 3. Molar group volume for the CH2 group as a function of Figure 4. Percent error in the prediction by the GCVOL model of
the number of carbons in the cycloalkaneat 20 OC. The dashed line the liquid volume of 1-hexanol as a function of the reduced tem-
corresponds to the value for the CH2 group used in this work. Ex- perature. Experimental data are taken from the DIPPR data com-
perimental data are from the Polymer Handbook (Brandrup and pilation (Daubert and Danner, 1989).
Immergut, 1989).
boiling point the errors are within the experimental error.
model is not K?co"ended to be used to describe density As the temperature reaches the near-critical area, the er-
data for cycloalkanes. rors become unacceptably high. This behavior is typical
for the GCVOL model, and we therefore recommend to
Prediction of Liquid Densities use the model onlv in the temwrature range between the
Solvents. Figure 4 is a graph of the error in the pre- melting point and the norm4 boiling point.
diction of the saturated liquid density for 1-hexanol as a In Table I1 we have compared the GCVOL model with
function of the temperature. Up to around the normal the widely used Rackett (1970) model in the calculation
2580 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991

Table 11. Average Mean Deviation' between Experimental

and Predicted Saturated Specific Volumes for
Low-Molecular-Weight Solvents. Experimental Data Are
from the DIPPR Data Compilation (Daubert and Danner,
1989)
h
1.4 - Siloxane oligomers

oi
Rackett this work \
R
chemical no. of T, - 298.15 T,- 298.15
family compds Th K Tb K
alkanes 55 4.4 4.5 1.4 1.1
aromatics 33 2.5 2.5 1.1 0.8
alkenes 36 5.3 5.3 1.0 0.7
alcohols 38 7.5 4.3 0.8 0.7
ketones 8 5.4 5.4 0.7 0.7
aldehydes 9 6.9 6.9 1.0 0.7
esters 22 4.1 4.1 1.2 1.1
esters' 20 3.3 3.3 1.8 1.3 Experimental doto Hurd 1946
0 ~ 0 0 0

ethers 13 6.7 7.7 1.2 0.5 Experimental data [Hein& at a!, 1983)
Prediction with Group Contribution
chlorides 23 4.1 4.0 0.9 0.5
siloxanesb 5 na na 0.1 0.0
0.0 2.0 4.0 6.0 0
a Average mean deviation is calculated as Inverse Molecular Weight (1 OOO/MW)
Figure 6. Specific volume at 25 O C for a series of siloxanes from
hexamethyldisiloxane to poly(dimethylsi1oxae).

where n, is number of compounds and np is number of points for 1972) and the Fedors (1974) methods. As foreseen by Van
each compound (usually spaced 10K apart). bExperimental data Krevelen, his model cannot reproduce the relationship
are from Hurd (1946) in temperature range between 0 and 80 O C . between the molecular weight and the specific volume, and
Using only the -COO- group. the Fedors method has some problems in the polymer end
1.6 I
of the molecular weight. As we fixed the value of the group
1 volume for CH2 to the value in an amorphous linear
polyethylene, it is not surprising that the GCVOL method
1.5 - will reproduce correctly the trend in the specific volume.
In Figure 6 the specific volume of a siloxane series going
h
oi
\
from hexamethyldisiloxane to poly(dimethylsi1oxane) is
R 1.4 - given as a function of the inverse molecular weight. In this
v
i series only the experimental data for the five low-molec-
ular-weight siloxanes were used in the determination of
__---_ _ - - the group volume parameters. It is therefore reassuring
that the trend in the specific volume as a function of the
r molecular weight is readily reproduced by the GCVOL
Lo
model.
0 0 0 0 0
-This
Experimental data
work
Only density data for compounds with one functional
Fedors group different from alkyl and arene groups were included
1,l ______ Van Krevelen in the determination of the group volumes. It is therefore
1.0
1I , , 8 . . . . . . . . . . . . . . . . . . . . . . . . . . .
I encouraging to see how well the GCVOL model predicts
the densities for solvents with several different functional
0.0 5.0 10.0 15.0
Inverse Moleculor Weight (1000/MW) groups present. The average mean deviation between
experimental and predicted data for diethylene glycol,
Figure 5. Specific volume at 25 O C for a series of linear alkanes from triethylene glycol, tetraethylene glycol, and polyethylene
hexane to amorphous polyethylene. The ordinate intercept corre-
sponds to polyethylene. Experimental data for low-molecular-weight glycol-400 (containing alkyl groups, an ether group, and
alkanes is taken from the DIPPR data compilation (Daubert and an alcohol group) is only 1%. The Fedors (1974) model
Danner, 1989), and data for the polymer are taken from Orwoll and predicts the same data with an average error of 7%. Ex-
Flory (1967). perimental data are taken from Muller and Rasmussen
(1991).
of the saturated densities for a number of different chem- Polymers. The density of a polymer is dependent on
ical families. Most of the density data used in Table I1 whether the polymer is amorphous, glassy, or crystalline.
went into the optimization of the group volume parame- The density of an amorphous polymer is the one that most
ters, but as the number of solvents used in the parameter closely resembles the liquid density of a solvent, and is
estimation is very big, we believe that one would obtain therefore the only kind that can be predicted accurately
the same quality for solvents not included in the data base. by the proposed GCVOL model.
The errors in using the Rackett model are typically around Table I11 shows a comparison between amorphous
5%, whereas our proposed group contribution model only polymer densities at 25 OC predicted by the three group
has errors around 1%. The larger error in the Rackett contribution methods: Van Krevelen (Van Krevelen and
predictions of the liquid densities might be caused by Hoftyzer, 1972), Fedors (1974), and the GCVOL model.
erroneous critical data. Considering that the same limited amount of density data
Specific Volumes for a Complete Molecular Weight went into the optimization of the Van Krevelen group
Range. In Figure 5 the specific volume of an normal volumes, the predictions by the proposed GCVOL model
alkane series going from hexane to polyethylene is given are satisfactory. It is interesting to note that the Fedors
as a function of the inverse molecular weight. The GCVOL model always overpredicts the density of the polymers,
model is compared with two other group contribution which might point to a general deficiency in the group
models: the Van Krevelen (Van Krevelen and Hoftyzer, volumes.
 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991 2581

Table 111. Predictions of Amorphous Polymer Densities at 298.15 K

exPo
density, this work Van Krevelenb Fedors'
polymer g/cm3 p, g/cm3 % de@ D . e/cm3 % de@ ~,g/cm~ Ide$
polypropene 0.864 -1.7 0.857- -0.8 0.866 -1.9
polybutene 0.850 0.862 -0.3 0.852 1.0 0.867 -0.9
polypentene 0.860 0.861 -1.3 0.855 -0.6 0.868 -2.1
polyhexene 0.850 0.861 -0.1 0.855 0.7 0.869 -1.0
polyisobutene 0.860 0.864 -0.4 0.860 0.1 0.878 -2.1
poly(5-phenylpentene) 0.860 1.038 1.1 1.042 0.8 1.085 -3.3
poly(methy1 acrylate) 1.050 1.205 1.2 1.229 -0.7 1.293 -6.0
poly(ethy1 acrylate) 1.220 1.140 -1.8 1.156 -3.2 1.211 -8.1
poly(buty1 methacrylate) 1.120 1.078 -2.4 1.036 1.7 1.092 -3.7
poly(hexy1 methacrylate) 1.053 1.034 -2.7 1.001 0.6 1.049 -4.1
poly(2-ethylbutyl methacrylate) 1.007 1.037 0.3 1.003 3.6 1.047 -0.6
poly(1-methylpentyl methacrylate) 1.040 1.037 -2.4 1.003 1.1 1.047 -3.3
poly(octy1 methacrylate) 1.013 1.005 -3.5 0.997 -0.5 1.019 -5.0
poly(dodecy1 methacrylate) 0.971 0.968 -4.2 0.946 -1.8 0.982 -5.8
poly(viny1 propionate) 0.929 1.221 -19.7 1.109 -8.7 1.211 -18.7
poly(isopropy1 vinyl ether) 1.020 0.925 -0.1 0.954 -3.2 1.015 -9.8
poly(buty1 vinyl ether) 0.924 0.960 -3.5 0.936 -0.9 0.995 -7.3
poly(sec-butyl vinyl ether) 0.927 0.915 1.0 0.938 -1.4 0.992 -7.3
poly(isobuty1 vinyl ether) 0.924 0.964 -3.7 0.938 -0.8 0.992 -6.7
poly(penty1 vinyl ether) 0.930 0.946 -3.0 0.925 -0.8 0.978 -6.5
poly(hexy1 vinyl ether) 0.918 0.935 -1.1 0.916 1.0 0.965 -4.3
poly(octy1 vinyl ether) 0.925 0.920 -0.7 0.905 1.1 0.947 -3.6
poly(2-ethylhexyl vinyl ether) 0.914 0.922 -2.0 0.906 -0.2 0.945 -4.5
av mean dev, 90 3.5 1.5 5.8
OExperimental data are taken from Van Krevelen and Hoftyzer (1972). bCalculated densities are from Van Krevelen and Hoftyzer (1972).
CPredictionswith the Fedors model is calculated in connection with this work. dThe percent deviation between experimental and predicted
densities is calculated as
- pprd
% dev = x 100%

Table IV gives a comparison between the Van Krevelen Table IV. Predictions of Amorphous Polymer Densities as
(Van Krevelen and Hoftyzer, 1972)model and the GCVOL a Function of Temwrature
model for the prediction of the temperature dependency AMP
of the specific volume for a number of different amorphous temp Van this
polymers. The predictions by the proposed GCVOL model polymer range, K ref Krevelen work
compares well with those of the Van Krevelen model. 277-328 a
polybutadiene 0.5 1.4
That the same group contribution table can be used to poly(1-butene), isotactic 407-514 b 1.2 0.2
predict the densities of both solvents and polymers is poly(methy1 methacrylate), 328-463 c 8.1 2.2
contradictory to the observations by Van Krevelen (Van isotactic
Krevelen and Hoftyzer, 19721, and might also seem polypropylene, isotactic 447-571 b 3.2 1.7
somewhat contradictory to the idea of polymers and sol- poly(buty1 methacrylate) 295-473 d 1.0 2.4
vents as having different free volumes. However, according polydimethylsiloxane 298-343 e na 1.0
polyisobutylene 326-383 e 9.4 6.8
to the simple group contribution concept put forward in poly(methy1 methacrylate) 387-432 d 18.3 1.2
this work, a rather large amount of free volume is created poly(4-methyl-1-pentene) 514-592 f 5.5 3.6
by monovalent end-groups. This is probably because polyoxymethylene 462-492 g 6.5 4.8
monovalent groups have a much larger degree of freedom poly(wmethy1styrene) 413471 h 0.3 2.4
than groups with higher valency. An intuitive way of polystyrene 389-469 h 5.4 3.7
realizing that the monovalent groups create more free poly(viny1 acetate) 337-393 e 1.1 4.1
polytetrahydrofuran 337-444 i 0.5 1.1
volume than the groups with higher valency is to compare poly(viny1chloride) 355-370 j 1.5 3.1
the ratio of optimized group volumes and van der Waals
volume (given by Bondi, (1968)).As an example, this ratio total av mean dev, % 4.5 2.7
is much bigger for a CH3 group than it is for a CH2group. Barlow, 1978. Zoller, 1979. 'Quach et al., 1974. dOlabisi and
Accordingly, the difference in the free volume between Simha, 1975. eBeret and Prausnitz, 1975. / Zoller, 1977.
hexane and polyethylene can be explained as an end-group gstarkweather et al., 1988. hQuach and Simha, 1971. 'Tsujita et
effect: Polyethylene has a negligible number of the free- al., 1973. 'Hellwege et al., 1962. &Averagemean deviation is cal-
volume-generating CH3end-groups compared to the num- culated as
ber of bivalent CH2groups, whereas hexane has only twice
the number of CH2 groups compared to the number of CH,
end-groups. Consequently hexane has a much higher de-
gree of free volume than polyethylene. aromatics, alkenes, ketones, ethers, esters, chlorides, and
Conclusion siloxanes) has been developed. The model can be used in
the complete molecular weight range from solvents, over
A new group contribution model for the prediction of oligomers, and to polymers.
liquid densities as a function of the temperature has been The liquid densities for solvents between the melting
proposed. A table of 36 different group volume increments point and the normal boiling point can be predicted with
for a wide variety of chemical families (noncyclic alkanes, an average error of 1%, which compares favorably with
2582 Ind. Eng. Chem. Res., Vol. 30, No. 12, 1991

predictions b y the widely used Rackett model (Rackett, Fedors, R. F. A Method for Estimating Both the Solubility Param-
1970). Also densities for solvents with more than one eter and Molar Volumes of Liquids. Polym. Eng. Sci. 1974, 14,
147.
different functional group can be predicted with very good Hankinson, R. W.; Thomson, G. H. A New Correlation for Saturated
results. Densities of Liquids and Their Mixtures. AZChE J. 1979,25,653.
Densities of amorphous polymers in the temperature Heintz, A.; Neff, B.; Lichtenthaler, R. N. Thermodynamics of Poly-
range of 277-571 K are predicted with an average error of dimethylsiloxane Solutions. 11. Application of an Extended
2.7%, which compares favorably with the group contri- Flory-Theory. Ber. Bunsen-Ges. Phys. Chem. 1983, 87, 1169.
bution models by Van Krevelen (Van Krevelen and Hof- Hellwege, V. K. H.; Knappe, W.; Lehmann, P. Die isotherme Kom-
pressibilitat einiger amorpher und teilkristalliner Hochpolymerer
tyzer, 1972) and Fedors (1974). im Temperaturbereich von 20-250C und bei Drucken bis zu 2000
Registry No. Polypropene, 9003-07-0; polybutene, 9003-28-5; kp/cm2. Kolloid Z. Z. Polym. 1962, 183, 110.
polypentene, 9078-70-0; polyhexene, 50981-41-4; polyisobutene, Hurd, C. B. Studies on Siloxanes. I. The Specific Volume and
9003-27-4; poly(5-phenylpentene), 136489-80-0; poly(methy1 Viscosity in Relation to Temperature and Constitution. J. Am.
acrylate), 9003-21-8; poly(ethy1 acrylate), 9003-32-1; poly(buty1 Chem. SOC.1946,68, 364.
methacrylate), 9003-63-8; poly(hexy1 methacrylate), 25087-17-6; Miiller, E. A.; Rasmussen, P. Densities and Excess Volumes in
poly(Zethylbuty1 methacrylate), 25087-19-8; poly(1-methylpentyl Aqueous Poly(ethy1ene glycol) Solutions. J. Chem. Eng. Data
methacrylate), 136489-81-1; poly(octy1methacrylate), 25087-1&7; 1991, 36, 214.
poly(dodecy1 methacrylate), 25719-52-2; poly(viny1 propionate), Olabisi, 0.; Simha, R. Pressure-Volume-Temperature Studies of
25035-84-1; poly(isopropy1 vinyl ether), 25585-49-3; poly(buty1 Amorphous and Crystallizable Polymers. I. Experimental.
vinyl ether), 25232-87-5; poly(sec-butyl vinyl ether), 28853-85-2; Macromolecules 1975,8, 206.
Orwoll, R. A.; Flory, P. J. Equation-of-State Parameters for Normal
poly(isobuty1 vinyl ether), 9003-44-5; poly(penty1 vinyl ether), Alkanes. Cmrelation with Chain Length. J. Am. Chem. SOC.
71356-03-1; poly(hexy1 vinyl ether), 25232-88-6;poly(octy1 vinyl 1967,89, 6814.
ether), 25232-89-7; poly(2-ethylhexyl vinyl ether), 29160-05-2; Quach, A.; Simha, R. Pressure-Volume-TemperatureProperties and
polybutadiene, 9003-17-2; poly( 1-butene), isotactic, 25036-29-7; Transitions of Amorphous Polymers; Polystyrene and Poly(or-
poly(methy1 methacrylate), isotactic, 25188-98-1; polypropylene, thomethylstyrene). J. Appl. Phys. 1971, 42, 4592.
isotactic, 25085-53-4; poly(methy1 methacrylate), 9011-14-7; Quach, A.; Wilson, P. S.; Simha, R. The Effects of Stereoregularity
poly(4-methyl- 1-pentene), 25068-26-2;polyoxymethylene, 9002- on the Pressure-Volume-Temperature and Related Properties of
81-7; poly(@-methylstyrene),25014-31-7;polystyrene, 9003-53-6; Poly(methy1 Methacrylate). J. Macromol. Sci., Phys. 1974, B9,
poly(viny1 acetate), 9003-20-7; polytetrahydrofuran, 24979-97-3; 533.
poly(viny1 chloride), 9002-86-2; cyclopentane, 287-92-3; cyclo- Rackett, H. G. Equation of State for Saturated Liquids. J. Chem.
hexane, 110-82-7;cycloheptane, 291-64-5; cyclooctane, 292-64-8 Eng. Data 1970, 15, 514.
cyclononane, 3350-30-9; cyclodecane, 293-96-9; cycloundecane, Spencer, C. F.; Danner, R. P. Improved Equation for Prediction of
294-41-7; cyclododecane, 294-62-2; cyclotridecane, 295-02-3. Saturated Liquid Density. J. Chem. Eng. Data 1972, 17, 236.
Starkweather, H. W., Jr.; Jones, G. A.; Zoller, P. The Pressure-
Literature Cited Volume-Temperature Relationship and the Heat of Fusion of
Polyoxymethylene. J. Polym. Sci. 1988, 26, 257.
Barlow, J. W. Measurement of the PVT Behaviour of cis-1,4-Poly- Tsujita, Y.; Nose, T.; Hata, T. Thermodynamic Properties of Poly-
butadiene. Polym. Eng. Sci. 1978, 18, 238. (ethylene glycol) and Poly(tetrahydrofuran). I. P-V-T Relations
Beret, S.; Prausnitz, J. M. Densities of Liquid Polymers at High and Internal Pressure. Polym. J. 1973,5, 201.
Pressure. Pressure-Volume-Temperature Measurements for Van Krevelen, D. W.; Hoftyzer, P. J. Properties of Polymers. Cor-
Polyethylene, Polyisobutylene, Poly(viny1 acetate), and Poly(di- relation with Chemical Structure; Elsevier Scientific: Amster-
methylsiloxane) to 1 kbar. Macromolecules 1975,8, 536. dam, 1972; Chapter 4, pp 41-65.
Bhirud, V. S. Saturated Liquid Densities of Normal Fluids. AIChE Zoller, P. The Specific Volume of Poly(4-methylpentene)-l) as a
J. 1978, 24, 1127. Function of Temperature (30-32OC)and Pressure (@2000kg/cm2).
Bondi, A. Physical Properties of Molecular Crystals, Liquids, and J. Appl. Polym. Sci. 1977, 21, 3129.
Glasses; Wiley: New York, 1968; Chapter 14, pp 450-75. Zoller P. Pressure-Volume-Temperature Relationships of Solid and
Brandrup, J.; Immergut, E. H., Eds. Polymer Handbook, 3rd ed.; Molten Polypropylene and Poly(butene-1). J. Appl. Polym. Sci.
Wiley: New York, 1989; Section IV, p 4. 1979,23, 1057.
Daubert, T. E.; Danner, R. P. Physical and Thermodynamic Prop-
erties of Pure Compounds: Data Compilation; Hemisphere: New Received for review April 25, 1991
York, 1989. Accepted August 7, 1991