1

Ind. Eng. Chern. 50,1026-32 (1958)

Thermal Diffusion in Liquids.

Measurements and a Molecular Model
The liquid thermal diffusion constant (Soret coefficient)
has been measured by a static method for three nonideal
binary mixtures of organic compounds, over the concentra-
tion range from 0.1 to 0.9 average mole fraction at an aver-
age temperature of 35 0 C.
A molecular model is proposed which relates the net heat
of transport, Q*, to the partial molar energies of activation
and hole formation. The model was tested against the
three systems .studied in this work and five systems studied
by Drickamer, Tichacek, and Kmak.
Agreement between predicted and observed values of
the thermal diffusion constant is only qualitative, although
concentration dependence can be predicted satisfactorily.
Because of the large error present in the predicted values,
it is difficult to evaluate the molecular model. With the
exception of one case, the predicted thermal diffusion con-
stants are of the right order of magnitude and sign.
STEPHEN WHITAKER and R. L. PIGFORD
University of Delaware. Newark, Del.

Reprinted from INDUSTRIAL AND ENGINEERING CHEMISTRY, Vol. 50, Page 1026, July 1958
Copyright 1958 by the American Chemical Society and reprinted by permission of the copyright owner
 STEPHEN WHITAKER and R. L. PIGFORD
Department of Chemical Engineering, University of Delaware, Newark, Del.

Thermal Diffusion in Liquids

Measurements and a Molecular Model

A theoretical model is proposed to explain the effect of several vari-

ables on thermal diffusion coefficients. The presentation represents
an interesting approach to a better understanding of molecular move-
ment in the liquid state

ACCORDING to the theory of irreversible manner similar to that of de Groot (5), where ｾ｜＠ and V2 are the partial molar
processes (5), the mass flow of component but have developed equations for a volumes; V is the molar volume,
1 in a binary mixture is given by coordinate system fixed in the laboratory Xlf't+ X2V2; Jl.l is the molar chemical
rather than at the center of mass. With potential (partial molar Gibbs free
｛Ｍｾ＠ Ｍｾｊ＠
...
J l = -pD grad Xl - -r
aX 1X2
grad T (1) this approach a is given by energy) of component 1; and Xl is the
mole fraction component 1. The Q*'s
where a is the thermal diffusion con- are defined by
stant. Drickamer, Tichacek, and Kmak
=
(6) have studied the relationship between a (2) Jh = L(h k + Qk*fh (3)
a and the net heats of transport in a k

when the temperature gradient is zero.

J h is the enthalpy flux, Jk is the mass flux
of component k, and hk is the partial
molar enthalpy of component k. Hence,
in the absence of a temperature gradient,
Ok* is the energy, over and above the
partial molar enthalpy, transported into
a region when a molecule moves into
that region as a result of diffusion.
Equation 3 gives a framework for
developing a molecular model.
llH h* First, the work of Eyring and others
should be reviewed, with regard to the
energy effects connected with molecular
Figure 1. The process of motion (7). Because experimental data
energy transport can be permitting activation energies to be
described in terms of the computed from diffusion coefficients
four steps pictured are lacking, this discussion is concerned
only with viscosity activation energies;
but according to the theory of absolute
reaction rates the same arguments
hold true for diffusion activation energies.
CllH: TRANSPORTED According to Eyring, the viscosity can
INTO REGION Y be described by
OSCSI

'1 = (Nh)
V exp (. t:..F*)
RT (4)

where h = Planck's constant, N =

Avogadro's number, V = molar volume,
t.F* = free energy of activation, R =
gas constant, and T = absolute tem-
perature. The total enthalpy of activa-
tion for viscous flow is given by

ＨｉＭｃｬｈｾ＠ (5)

1026 INDUSTRIAL AND ENGINEERING CHEMISTRY

 Molecular motion-the jump of a
molecule from one equilibrium position
to another-occurs when a molecule is
suitably activated for flow and when a
nearby hole is available for this activated
molecule to flow into. Therefore 1:1H"*
must be the sum of the enthalpy required
to activate the molecule-the flow
activation enthalpy-and the enthalpy
required for the formation of a hole.
To obtain information on how the total
activation enthalpy, t::.H"*, is distributed
between the flow activation, enthalpy,
!::.H;"*, and the hole formation enthalpy,
1:1Hh "*, consider a liquid consisting of
molecules and holes. At a given volume
the number of holes and the "free
volume" in the liquid should be very
nearly independent of pressure and
temperature, if the molecules are not
Ｚｾ＠ ｴＺｏｩ｣ｾ＠
should represent only the flow activation
enthalpy, as none of the activation
enthalpy is used to form a hole and all
of it is used to detach the molecule from
its equilibrium environment. Hence
The hole formation enthalpy is now given
by
The total activation energy can be
expressed as
｛ｾＨＩｊｰ＠
R =
RT2 ＨｾＩｖ＠
Therefore
!1H
h
"* = RT2 ＨｾｰＩ＠
ｾｔ＠
The process of energy transport
during diffusion can now be described in
terms of the four steps pictured in
Figure 1.
1. Consider a subregion, R, in a
fluid at rest. R is contained within two
regions, X and Y, which divide the fluid
into two parts.
2. A molecule in region X is activated
and simultaneously a hole is formed in
region Y.
3. The activated molecule in X
jumps to Y, carrying with it some fraction
of the flow activation enthalpy it received
when it was freed from its neighbors.
4. The activated molecule fills the
hole in Y and the fraction, C, of !::.H;"*
transported is given up to the region.
The hole in X collapses and t::.Hk"* is given
up to this region. The flow activation
enthalpy that was not transported from
X to Y is given up to X and subregion R
assumes its initial state.
Some fraction of !::.H;"* is transported in
the direction of molecular motion and
1:1Hk"* is "transported" in the opposite
direction. We are considering what
happens during molecular motion on the
average. The net heat of transport-
defined as positive when energy flows
in the direction of molecular motion-
is the difference between the two trans-
ported quantities.
Q.* = C!1H;."* - !1Hhi"*
where 0 ｾ＠ C ｾ＠ 1
Wirtz (27) has obtained a similar
expression for Qi*.
This model oversimplifies the energy
effects connected with molecular motion.
The liquid does not consist only of
molecules, activated molecules, and
holes, but contains molecules occupying
ＺｬﾷＡｃｾ＠
a nearly continuous range of energy
levels. Presumably, the distribution of
hole sizes or free volumes is similar to the
distribution of molecular energies. Cer-
tainly there must be some relationship
between the degree of activation and the
hole size necessary for molecular motion.
A highly activated molecule might possess
enough energy to form its own hole;
hence the energy transported would be
larger than that given by Equation
10. On the other hand, if an especially
large hole were nearby, a molecule
would need only a small activation
energy, to make the jump.
Table I. Flow Activation and Hole
Formation Enthalpies
!1H;"*. "*.
｛ｾｮＨＩｊｶＫ＠
!1Hh
Cal./G. Cal./G. T.
R Substance Mole Mole cC.
Water 4060 -2 25
｛ｾｬｮｖＩｊｔ＠ (8)
Ethanol
3580
2120
22
1070 25
40
2370 1090 40
Methanol 1240 1040 40
Benzene 670 1600 40
V
｛ｾｬｮ＠
ｾｰ＠
("IV!]
T
(9) Carbon tetrachloride
Cyclohexane
600
210
1680 40
2300 40
Fortunately, these problems are
avoided to some extent by using values
of the activation enthalpies based . on
experimentally observed processes in
which similar effects of distributed
quantities are present. Whether the
average energy quantity derived from
viscous flow data, for example, is the
same average needed to calculate the
heat of transport is open to conjecture.
Diffusion and viscosity activation ener-
gies are seldom precisely the same and
sometimes differ considerably, an in-
dication that the activated molecules
influence the two phenomena differ-
ently.
The viscosity data and P-V-T data
used for the calculations in Table I
were obtained from the literature (3, 9,
17, 72). The values of 1:1H;"* for water
may be in error by as much as 1%.
For ethanol and methanol t::.H;"* may be
in error by as much as 2%; the values
for benzene, carbon tetrachloride, and
cyclohexane may be in error by as much
as 200 cal. per gram mole. Bondi (2)
and Kmak (73) have also calculated
values of 1:1Hh"* for ethanol and methanol
and found them to be in the range 400
to 600 cal. per gram mole. At present
this discrepancy cannot be explained,
although various methods of differentiat-
(10) ing numerical data can lead to different
numerical results, even when the same
data are employed.
The nonassociated liquids require a
relatively small amount of energy for
flow activation; the strongly associated
alcohols and water require a much
larger energy (Table I). This is ex-
pected, because the flow or jump of a
molecule requires the breaking of some,
though not necessarily all, of the bonds
between it and its neighbors. The
energy required to break the hydrogen
bond between two alcohol molecules is
large compared to that required to
break the weaker bond between two
benzene molecules. Bondi (2) has in-
vestiga ted the terms 1:1H;"* and 1:1Hh"* for
many liquids and discussed the effect of
molar volume and association on these
terms.
Associated substances have relatively
high values of 1:1H;"* compared to their
values of 1:1Hh"*. Thus, although large
amounts of energy must be supplied to
associated molecules to break them
away from their neighbors, this is not
true for hole formation. Apparently a
hole can be formed in a liquid with less
disturbance of the liquid structure than
that required for the movement of a
molecule through the liquid.
ｾ＠
ｾ＠ ...J
;3 sl-I---+---t}.
'"
ｾ＠
ii:
a ＰｲＭｾ｟Ｋ＠
r:z·
w
ｾ＠
i=
§!
ｾＧ＠
ｾ＠ ＲｲＭｴ｟ＫＴｾ＠
§:;;
0.2 0.0 0.8 0.1 1.0
MOLE FRACTION ETHANOL
Figure 2. 1:1H J + and 1:1Hh"* for the
ethanol (2)-water (1) system at 25° C.
as predicted by Equations 15 and 17
VOL. 50, NO.7. JULY 195. 1027
 I
I
I

/
f--------
/
/
--I

THERMAL
DIFFUSION
CELL

PHOTOGRAPHIC
f- NEGATIVE

Figure 3. A'stationary concentration gradient is measured in a thermal diffusion cell by means of an optical system

Calculation of Thermal
Diffusion Constant
To calculate the thermal diffusion
constant from Equation 2, the values of
* *
!lil;i and t::.Hhi must be known for both
components at the concentration of the
mixture in question-Le., partial molar
flow activation and hole formation
enthalpies must be ｫｮｯｾ＠ for each
component as indicated in Equation 11.
.. \
Viscosity data are available (4; S-l1,
13, 17, 19) from which the total activa-
tion enthalpy can be computed as a
function of concentration and from which
the partial molar activation enthalpy,
ilJli* , can be obtained, where
The deviation of !::.H+ from values pre-
dicted by Equation 13 is less than 15%
for these three systems. If a system
obeyed Equation 13 it might be assumed
for trial that
tlH;* = X]tlHi'*
tlHh* = x,tlHh,*
To attempt a test of Equation 11 for
the estimation of Qi*' some assumption
must be made concerning the distribu-
tion of t::.H* between t::.H;* and t::.Hh+ for
miXtures that do not obey Equation 1:5.
Excess total enthalpy of activation is
defined by
tlH* E = tlH* - (x,tlH,*
The problem is to determine how much
of this excess can be assigned to ilJl;*
and how much to ilJlh *.
Solutions which are nonideal in the
thermodynamic sense usually contain
one or more associated components.
Although nonideality can be caused
+ x2tlH;2* (14) simply by difference in molar volumes
+ X2tlHh2'f' (15) and by dipole interaction, highly non-
ideal solutions are generally not found,
except when one of the components is an
associated or hydrogen-bonded sub-
stance. Nonideality of the enthalpy of
activation for viscosity is similarly related
to hydrogen bonding-Le., t::.H* E is
relatively large for a mixture containing
an associated component. Table I
shows that associated components have
+ x2tlH2*) high values of !lil;* . With this in mind
(16) we reason as follows:
Solutions containing one or two associ-

Unfortunately, the necessary data are

not available to calculate the individual
terms of the right side of Equation 12.
To calculate both quantities on the right,
thermal expansion and compressibility
z
.
'-
'-'
10

8
V
/0
T Figure 4. Dif-

V
a: CELL GAP = 0.36 IN.
ference between
data are needed, as well as viscosity or ｉＮＺｾ＠ temperature gra-
diffusion data as a function of pressure Z
lIJ 6 dient at top and
and temperature for mixtures. For the '-'
°

ｾY ＧｭＮｯ＠
center of cell as a
mixtures studied only t::.H* and pure function of the
component values of !lil;;* and t::.Hhi* average tem-
are known. . perature gradient
For ideal solutions the linear relation-
ship , '" VO Although convective
° Jt. currents were not elimi-

1__ -L.---ＡＮＭｾ｣ｲ＠
nated, their effect
tlH* = x,tlH,* + x2tlH2* (13) was small
0
might be assumed. This is very m;arly
o
true for the benzene-carbon tetrachlo- o 4 8 12 16 20 24
ride, cyeIohexane-benzene, and cyeIo-
hexane-carbon tetrachloride systems. AVERAGE TEMPERATURE GRADIENT, ·C/IN.

1028 INDUSTRIAL AND ENGINEERING CHEMISTRY

 MOLECULAR PHYSICS

ated components have large values of

AlI=*' E. Associated compounds have high A
values of AlI;.=*', presumably caused by
the necessity of breaking hydrogen bonds
in order that a molecule may flow.
The intermolecular forces which cause
high values of AlI;.=*' apparently are also
the cause of high values of AlI=*' E. It is
concluded that the excess activation
enthalpy, AlI=*' E, can be attributed
mainly to the nonideality of AlI;=*', while
fj.Hh+ is nearly ideal and obeys Equation
15. The flow activation enthalpy of the
mixture may be written
AH;=*' = XIAH;.=*' + x2AH;2 =*' + AH* E A
(17 )
FRONT VIEW SECTION A-A
Figure 2 shows the result of the appli-
cation of Equations 15 and 17 to the
ethanol-water system at 25 0 C.
With the aid of Equations 15 and 17,
Qi* can be calculated from Equation 11,
if C is known. Unfortunately, all that is
known of C is that its limiting values are
o and 1 and it is a measure of how much ｾＭ［Ｃｌｄｉｓｔｏｒｎ＠ OF THE .LIGHT BEAM
of the flow activation enthalpy is carried DUE POSSIBLY TO A HORIZONTAL
TEMPERATURE GRADIENT IN
with an activated molecule when it THIS REGION.
jumps from one equilibrium position to
another. Its value will depend on the
forces existing between the activated
molecule and its neighbors, and hence it
should be dependent on the system and
the concentration. In the absence of
any direct measurement or theoretical
estimate of C, no rigorous test of Equation DETAIL A
11 can be obtained. Nevertheless, the Figure 5. After three cells had been tested, that chosen for measuring thermal
empirical determination of the value of diffusion had minimum convection
C that gives the closest agreement be-
tween the experimental data and the diffusion constant was readily obtained the second cell. Figure 4 shows the
values predicted from Equations 2 and 11 (16). Thomaes (18) improved this effect of temperature gradient and cell
makes it possible to appraise the pro- type of optical system by using a double gap on the amount of convection present,
posals at least tentatively. For some cases slit to reduce the interference fringes as indicated by the difference in the
the predicted values of a are essentially and a traveling eyepiece to measure thermal gradients at the top edge and
independent of the choice of C, and these the deflection of the light beam. The center of the cell. Although convective
cases can be considered valid tests of the simpler optical system employed origi- currents were not entirely eliminated in
molecular model. nally by Tanner was adequate for this the final cell, their effect must have been
study. small. This is indicated by the fact
Experimental Method Three cells, having cell gaps, /!;.r, of that the experimental values of a were
0.36, 0.18, and 0.0853 inch, were tested. independent of the temperature gradient
The apparatus used in this investiga- Considerable convection was present in over the range from 18 0 to 42 0 C. per
tion was similar to that developed by the first cell and, to a lesser extent, in inch. Figure 5 shows a detailed drawing
Tanner (16) in 1927. The technique
permits measurement of a stationary
concentration gradient within the ther- Table II. Smoothed Values of Thermal Diffusion Constants Agree with Drickamer
mal diffusion cell by means of an optical and Others
system. A schematic drawing of the Investigator
system is shown in Figure 3. This work, Drickamer Prigogine Thomaes
The experimental procedure consisted Mole Frac. 35 0 C., et al. (6), etal. (14), (18),
of maintaining the cell at a constant Methanol ａｔｾＲＮＵﾰ＠ C. ａｔｾＱＵﾰ＠ C. AT = 20 0 C. AT = 23.5 0 C.
temperature while the isothermal posi- Methanol-Benzene
tion of the light beam was recorded. 0.8 -0.92 -0.80 -1.72
The cell was then heated on the top 0.6 -0.75 -0.24 -1.92 -0.008
and cooled on the bottom by water 0.4 0 +0.67 -1.94 -0.076
0.2 +1.80 -1.37 -0.010
pumped from constant-temperature 0.1 0 +0.015
baths. After 1 hour (99% of the steady-
Methanol-Carbon Tetrachloride
state concentration gradient was ob-
ａｔｾＲﾰ＠ C. ａｔｾＴﾰ＠ C.
tained in an hour or less) the final 0.2 -4.0 -2.8
deflection of the beam was recorded. 0.5 -5.2 -4.9
This deflection was caused by both the 0.8 -3.2 -3.0
temperature gradient and the concen- ex is negative when methanol moves toward hot wall.
tration gradient and from it the thermal

VOL. 50, NO.7. JULY 1958 1029

 2.o.---;--_;-_--:,-_--,__-, 2.0.---;---,---,----,,---1 Tichacek, and Kmak (6) is very good.
In general thermogravitational columns,
1.0 t---t---i-------1"-------1-----
such as that used by Prigogine, De
Brouchere, and Amand (14), have never
"
,:
:z agreed well with static methods. It
;;en
is possible that the higher average tem-
:z
0
0 :: \ I:; I:; I:; perature would account for Prigogine's
ｾ＠
:z
0
-4.0 -1.0 t--\-t----t----.. 1:;,...-1-- higher values, but normally ex should
in
decrease with temperature, owing to the
"
It 8z .. ____ ｾｬＬＮＭ＼Ｚ＠ ｅｧｾ｡ＲｯＢ＠
ｾＺ＠
0
..J .... _- increase of the XI in Equation 2.
"'" -6.0
ｾ＠ -2.0 t---\-t---t---t---t----i
0:
ｾ＠ It is more likely that these high values
"'
:I:
I- o are caused by the uncertain fluid dy-
..J ,
-8,0
:l -3.0 r--- \ i - - - i - - - i - - - - t -./"""C;;-! namics of the column.
ffi 1\ ｾ＠ ECS.2,1I ", Thomaes' results are rather erratic.
ｾ＠

\'W"l
a25
-10.0
0 0.2
- - - ESTIMATED ERROR

MOLE FRACTION ETHANOL
Figure 6. Predicted thermal diffusion
values for ethanol (2)-water (1) and
methanol (2)-water (1) systems are 5
to 10 times too large
Left.
of the final cell. The dimensions of this
cell were (inches):
Cell gap ＨｾｲＩ＠
Cell length
Cell width
This experimental method has the
advantages of short relaxation times and
small temperature differences. The re-
laxation time was less than 1 hour and
the use of small temperature differences
eliminated the error involved in the use
of an average temperature. The dis-
advantage of this method is its tendency
to be unstable. When the less dense
1.0
-4.01--\ \ 'rZ: / For the methanol-benzene system his
f
0 .•
values are much lower than those ob-
" /."/ tained in this work and by Drickamer
-0.0 _II
" :.-"
,,/ .1'--:--'::-::::-:::::-::===-1
/ ll.
'
DATA OF DRICKAMER
and others. For the benzene-cyclo-
\ .... __ .... 1 /1 TICHACEK AND KMAK
hexane system Thomaes apparently ob-
\ 1/ (REF .• )
' ..... __ ..-"i - - - ESTIMATED ERROR served no separation, whereas Drickamer
-"°0ｾＭＰＢＧＺＮＲ［Ｔ .•---0-
.•- - - I , . O reported ex ｾ＠ 0.4. For the cyclohexane-
MOLE fRACTION METHANOL carbon tetrachloride system Thomaes
Ethanol·water 25 0 c. Right. Methanol-water 40 0 c. found ex ｾ＠ 1.5, which is in excellent
agreement with Drickamer's value of
approximately 1.3. Although the lower
temperature might account for these
component concentrated at the hot side, differences, it would seem that Thomaes,
no trouble was encountered in the who used a cell having a cell gap of
narrowest cell (cell gap of 0.0853 inch), 0.674 inch, must have had considerable
0.0853 convection in his cell, which was operated
but when it concentrated at the cold
2.090
2.5 wall, convection currents were appar- with a temperature difference of 23.5 0 c.
ently set up. For this reason ex could Even though convective flow is present,
not be measured for methanol mole it does not necessarily destroy the con-
fractions less than 0.39 in the methanol- centration gradients completely. It was
benzene system. found in this work that concentration
Smoothed values of the experimental gradients were set up even when con-
data are compared with the results of vection was present. Furthermore, the
other investigators in Table II. For the gradients set up were not proportional to
methanol-benzene and methanol-carbon the measured temperature gradients and
tetrachloride systems the agreement existed in the center of the cell where the
between this work and that of Drickamer, convective flow must have been the

Figure 7. Observed thermal diffusion constants at 35 0 C. agree reasonably well with theory for these systems
1.0

I
/--J---l I ...., ....
.... 2 .0
/- ,
•. 0.---:,-;....\_...,._ _ _ _ _ _ _ _-, I:;
0 THIS WORK, 35°C.
DATA OF DRICKAMER
"- ....
.... ........
--- I
/
/ \
\
,"-
----
-1.0 TlCHACEK AND KMAK I
I \ 0 THIS WORK, 35°C.
(REF.6), 40°C.
0
....
" .....

ｾ＠
\ 6 DATA OF DRICKAMER --- ESTI MATED ERROR
3.0 -f- -II ｔｉｃｈａｾＺｦ＠ Ｔｾｋ＠
" / --1--

-
.. -2.0
f \ ---
J
I-

JO
0
ESTIMATEa ERROR

,J {:::'
Z
;; Eas 2,11 "
"
,: 2.0 -,"-
I
I:; ,--\+---,----.----1
\
'"
:z
8
c=o 10
1\(I:;.,L
,:
ｾ＠ ｾ＠ E QS.2,1I
c= 0,0
-s.O

--- .....

i'Lf
(I) -1.0
:z

\, \ \ .L
ＬＮｾ］Ｍ＠ \f\>,-_,
z I'0 oo ,,-
o
in
VO :z /
/

ｾ＠
o
-4.0
, "\J
--I t---
I o
ｾ＠ -2.0 t
/

..J

""
1
I ( I:; c I
I
I It
o
.J
, 0
0

i \-1-
0:

｜ｾ＠
W
F
-s.O 1 I ｾ＠ -3.0
, ,
_I
I

,,
1 0
I 0: 0
I
I /, I 0 1
i="' I
!l!
,
1 0
ｾＬＭ ,
-1.0 \ 0 I- 0

,,
I- \L EQS, 2," I -6.0 t-! -4.0 I
\ C·O.30 I 1 I
, ;/ I 1 f
·-2.0/---+---+-,', ..... __ ...... ,,"'I--
-7.0
I
t-I I
1
-5.0
r-\-,
\
I
I
I
I
I
I
I
I
,, I
I
0 THIS WORK, 35°C.
- - - ESTIMATED ERROR
"5.° " ---o.l.
.•- - - o.....- - - o.....---o..l..•- - - J I . O I
0 -8.0 -6.0
o 0.2. 0,4 0.6 0.8 1.0 o 0.2 0.4 0.6 0 .• 1.0
MOLE fRACTION METHANOL MOLE FRACTION METHANOL MOLE FRACTION ETHANOL

Methanol (2)-benzene (1) system Methanol (2)-carbon tetrachloride (1) system Ethanol (2)-carbon tetrachloride (1) system

1030 INDUSTRIAL AND ENGINEERING CHEMISTRY

 MOLECULAR PHYSICS

2.0 1.0 2.0

"
..:
A DATA OF DRICKAMER
TICHACEK AND KMAK

- - - ESTIMATED ERROR
IREF.6 )
1
z
f'!
V>
"3----- Lr - ---T z
..:"
A
J f-A
A
A

---l
;'
z A ;'
0
u
A f'!VI L

"
J --
z
0 1.0
z
o
u
..:
z 1.0
u;

ｾ］ＺＢＭ＠
::> EQS, 2,11 z
o ｾ＠
"-
"-
e
,.a:.
--'
J C = 0.20
ｾ＠
u;
ｾ＠
e"-
Z
o
u
z
---------------
1 ｉＧｻｯＮｾ＠ o
u;
UJ
:I:
t-
,.«--' -1.0
ｾ＠ EQS.2,'I

o --- - -- -1
0:
UJ
...
:I:

---("/"l e
--'
<[

ｾ＠ 0
\ c= 0.20

o 0.2

r---
0.4 0.6 0.8 1.0 A DATA OF DRICKAMER UJ

MOLE FRACTION CARBON TETRACHLORIDE

TICHACEK AND KMAK ｾ＠
IREF.6 )
- - - ESTI MATED ERROR
Carbon tetrachloride (2)-cyclohexane (J) system -2. 0
0.2 0.4 0.6 0.8 1.0

MOLE FRACTION CYCLOHEXANE

Cyclohexane (2)-benzene (J) system

-1---1--- ｾ＠ DATA OF DRICKAM£R
TICHACEK AND KMAK
(REF.£' )
- - - ESTIMATED ERROR
-I. 0
0.2 0.4 0.6 0.8 1.0
MOLE FRACTION CARBON TETRACHLORIDE

Carbon tetrachloride (2)-benzene (1) system

Figure 8. These systems present the best test of the molecular model

greatest. Apparently the static cell with quantities that are ｮ･､ｾＮ＠ ｏｲｾｩｮ｡ｹ＠
convective flow acts somewhat like a
thermogravitational column, where con-
1J(a,N--a)2
l '
was calculated. An esti- diffusion measurements, 1f ava1lable,
would give a better indication of the
mate of the standard deviation, ob- energy effects connected with molecular
centration gradients are enhanced by
tained by estimating the errors in the motion of the kind involved in thermal
countercurrent convective flow.
experimental method, is given for each diffusion.
The average values of the experi-
concentration. A more complete dis-
mental data for this work are listed in
cussion of this method is given elsewhere The observed a values for methanol-
Table III. Where enough data were
(20). benzene, ethanol-carbon tetrachloride,
taken the standard deviation, S =
and methanol-carbon tetrachloride sys-
tems agree reasonably well with the
Discussion of Results theory, considering the error involved in
Table III. Experimental Results Agree The agreement between the predicted the predicted values. Pure-component
Reasonably Well with Theory and measured values of a is shown in
Mole Estd. Calcd. Figures 6 through 8. Figure 9 shows
Frac. Std. Std. the dependence of the predicted a on C
Alcohol Av.a Dev. Dev. for the ethanol-water system.
10 -
I
Ethanol-Carbon Tetrachloride System. For the ethanol-water and methanol-
35° C. water systems, the agreement is poor; ,
0.10
0.17
0.32
-0.14
-0.10
-3.70
0.07
0.03
0.24
the predicted values are five to ten
times too large. The theory does predict
the change of sign of a at approximately .
,
0.39 -3.79 0.24 0.18
0.50 -3.69 0.24
the right concentration for both of these .....
0.21 z
0.64
0.77
0.90
-3.23
-2.62
-2.36
0.21
0.17
0.38
0.08 systems, if a large enough value of C
is chosen. In both systems it is not
f'!
'"
z
:3 - 8
t e = 0.0
V i-
1/
Methanol-Carbon Tetrachloride System.
difficult to imagine appreciable changes
of C with concentration which might
z
o
u;
ｾ＠ ｾ＠ V V V
/ il
35 0 C. improve the agreement considerably. "- -I 6
./
e
0.09
0.17
-0.25
-2.90
0.24 Probable Causes of Lack of Agree- <i,. \ ｣ｾ＠
0.15 ffi
0.34 ment between Predicted and Experi- -2 41----
-5.48 0.18 ｾ＠
mental Values. I t was assumed that the
ｾ＠
0.49 -5.29 0.17
/
0.43

, v-
0.58 -4.90 0.16 excess enthalpy of activation can be attri-
0.70 -3.69 0.12 0.38 buted solely to t!Ji;*. While this is a -3 2

0.81 -3.46 0.11 reasonable assumption, some error must c= 0.7

0.90 -2.64 0.17 be incurred because of it.
The energy effects connected with -4
°
Methanol-Benzene System. 35° C. molecular motion for alcohol-water sys- ｾ＠
0.39 0.0 tems are undoubtedly more complicated 0.2. 0.4 0.6 0.8 1.0

0.52 -0.46 0.11 MOLE FRACTION ETHANOL

than they have been assumed to be.
0.68 -0.89 0.21 The use of a single value of C for both
0.84 -0.92 0.22 Figure 9. Effect of C on predicted
components and all concentrations is not value of a for ethanol (2)-water (1)
a is negative when alcohol moves toward realistic.
hot wall. Viscosity activation energies probably
system at 25° C.
do not precisely represent the energy Predicted values are 5 to 10 times too large

VOL. 50, NO.7. JULY 1958 1031

values of AHh'* and AH;.* were used for
the ethanol-carbon tetrachloride system,
because viscosity data as a function of
temperature and concentration were not
available. [Thermodynamic data for
this system were obtained from the work
of Barker, Brown, and Smith (1).]
The methanol-benzene and methanol-
carbon tetrachloride systems are good
examples of the error present in the term
Xl ＨｾＺＩＮ＠ In the methanol-benzene sys-
tem the predicted ex is 3.3 and the
estimated error is ±3.0 at a methanol
mole fraction of 0.2. In the methanol-
carbon tetrachloride system the pre-
dicted ex is -26.5 and the estimated
error is ±26.0 at a methanol mole
fraction of 0.2. In both cases the high
value of ex and the large estimated error
are due to small values of Xl ＨｾＺＩ＠
the large percentage error in this term.
The peculiar behavior of the theoretical
curve in the central concentration region
for the latter system is also due to the
unlikely variation of Xl ＨｾＺＩ＠
concentration. Both systems were
studied very carefully by Scatchard and
Ticknor (15) and enough of the original
data were presented to permit a fairly
good estimate of the error. This was not
possible in the other systems studied and
in these the estimate was always con-
servative.
The benzene--carbon tetrachloride,
cyclohexane-benzene, and carbon tetra-
chloride--cyclohexane systems present
the surest test of the molecular model.
In these three systems the value of !:1H;.*
is much smaller than !:1Hhi*; hence the
value of C affects Q,* to a small extent.
The predicted ex for the carbon tetra-
chloride-cyclohexane system (mole frac-
tion of 0.50) varies from 0.62 to 0.81
as C varies from 0 to 1. There is no
appreciable error present in the term
Xl ＨｾＺＩＬ＠ as it deviates only slightly from
RT and the total activation energy is
closely approximated by Equation 15.
Little error is incurred in determining
partial molar activation enthalpies.
There is a single important error in each
of these three systems: the calculation of
!:1Hh,* or !:1H;,* . For benzene and car-
bon tetrachloride AH;;* has been meas-
ured directly (12). The standard devi-
ations of !:1H;,* for benzene and carbon
tetrachloride were 195 afld 243 cal. per
mole, respectively. The data of Bridg-
man (3) were used to calculate!:1Hh,* for
cyclohexane. Data were listed for only
two pressures (1 and 1000 kg. per sq. cm.)
at 75 0 C. From these data and similar
data for other compounds t:.Hhi* at 40 0 C.
was estimated. The estimated error was
taken to be 200 cal. per mole. In all
three systems the predicted values of ex
are of the proper order of magnitude.
In each system the value of C was arbi-
trarily chosen to be 0.20.
1032 INDUSTRIAL AND ENGINEERING CHEMISTRY
The carbon tetrachloride--cyclohexane
system shows the best agreement with
theory. While the predicted values are
approximately one half to two thirds of
the experimental values (which can be
considered accurate to ±10%), they are
within the estimated error bounds.
The theory predicts the wrong sign for
the carbon tetrachloride--cyclohexane
system, but the concentration depend-
ence is properly predicted. For each of
these three systems an error of 30% in the
prediction of one of the Q*'s can cause
an error of approximately ±0.9 in the
predicted value of ex. It appears that
the theory predicts the value of Q;* to
some degree of accuracy.
Discussion of Error
and The principal source of error in the
determination of Q* arises from the fact
that experimental data must be differen-
tiated once and sometimes twice, to
obtain the energy quantities desired.
with
The viscosity and P-V-T data were
fitted to a least squares equation using
Gram polynomials and five evenly
spaced points. This equation was then
differentiated to obtain the viscosity
activation energy and the compressi-
bility. The statistical criterion for the
best fit was used to determine the order
of the equation. An estimate of the
error can be obtained by use of
, (3(/'.'
(/' (dy(dx) = Ax (18 )
and
(19 )
where (/'2 (dy/dx) and ,,; are the estimated
variances and where (3 = 0.216 for the
evaluation of the derivative at the central
point and (3 = 0.957 at the end points.
y I is the value predicted by the least
squares equation and y is the experi-
mental value.
Two assumptions are involved in the
derivation of these equations: that the
variance of y is independent of X and
that the polynomial chosen to fit the
data is in fact the true functional rela-
tionship between y and x-i.e., para-
bolic, cubic, etc. The polynomial does
not fit the data exactly only because of
the error in the data.
This method of estimating the error
has one serious drawback. When
smoothed data ar8 being used, the vari-
ance will be smaller than it should be
and will be zero if the da ta were smoothed
by means of a least squares polynomial.
This estimate relies on the scatter of the
original data to predict the error in the
derivative. If estimates of the variance
in the original data are available, they
can be used in place of the variance
PRINTED IN u. s. A.
estimated by Equation 19. In general,
this information was not available and
Equation 19 had to be relied upon for
an estimate of the variance.
In the determination of Q* the major
source of error comes from the pressure
derivatives, [0 ＱｾｖＩｊｔ＠ and ＨｾｔＢ＠
The estimated error in these cases is
relatively large, because it must be
evaluated at the end point of the five
experimental points.
Acknowledgment
Stephen Whitaker would like to
acknowledge the aid of a National
Science Foundation fellowship.
Literature Cited
(1) Barker, J. A., Brown, I., Smith, F.,
Discussions Faraday Soc. 15, 142-50
(1953).
(2) Bondi, A., J. Chern. Phys. 14, 591-
607 (1936).
(3) Bridgman, P. W., Proc. Arn. Acad.
Arts Sci. 48, 309-62 (1921):
49,4-14 (1913); 61,57-99 (1936);
66, 183-233 (1931); 77, 115-46
(1949).
(4) Carr, C., Riddick, J. A., IND. ENG.
CHEM. 43, 692-6 (1951).
(5) De Groot, S. R., "Thermodynamics
of Irreversible Processes," Inter-
science, New York, 1950.
(6) Drickamer, H. G., Tichacek, L. J.,
Kmak, W. S., J. Phys. Chern. 60,
660-65 (1956).
(7) Glasstone, S., Laidler, K. L., Eyring,
J., "Theory of Rate Processes,"
McGraw-Hill, New York, 1941.
(8) Grunberg, L., Trans. Faraday Soc.
50, 1293-303 (1954).
(9) "Handbook of Chemistry and Phys-
ics," 37th ed., Chemical Rubber
Publishing Co., 1955.
(10) Harns, H., Rofler, H., Wolf, K. L.,
z. physik. Chern. B41, 321-64
(1938).
(11) "International Critical Tables," Mc-
Graw-Hill, New York, 1933.
(12) Jobling, A., Lawrence, A. S. C.,
Proc. Roy. Soc. (London) 206A,
257-74 (1951).
(13) Kmak, W. S., Ph.D. thesis in chem-
ical engineering, University of
Illinois, 1956.
(14) Prigogine,I.,DeBrouchere,L.,Amand,
R., Physica 16, 851-60 (1950).
(15) Scatchard, G., Ticknor, L. B., J. Arn.
Chern. Soc. 74, 3724-9 (1952).
(16) Tanner, C. C., Trans. Faraday Soc.
23, 75-95 (1927).
(17) Teitelbaum, B. Ya., Gortalova, T.
A., Ganelina, S. G., J. Gen.
Chern. U.S.S.R. 20, 1481-4 (1950)
(English trans.).
(18) Thomaes, G., Physica 17, 885-98
(1951).
(19) Tichacek, L. J., Ph.D. thesis in chem-
ical engineering, University of
Ilinois, 1956.
(20) Whitaker, S., M.Ch.E. thesis, Uni-
versity of Delaware, 1956.
(21) Wirtz, K., Z. Naturforschung 3a,
672-90 (1948).
RECEIVED for review January 15, 1958
ACCEPTED April 11, 1958
Division of Industrial and Engineering
Chemistry, Symposium on Molecular Phys-
ics in Chemical Engineering, Cleveland,
Ohio, January 1958. Work supported
in part by the U. S. Atomic Energy
Commission.