You are on page 1of 6

Effect of Concentration Level on

Mass Transfer Rates


1. E. WESTKAEMPER and ROBERT R. WHITE
University of Michigan, Ann Arbor, Michigan

Carbon tetrachloride was evaporated into a stream of air over ranges of gas concentration
from 0 to 0.70 mole fraction carbon tetrachloride, Reynolds number from 600 to 15,000,
and Schmidt number from 0.23 to 1.17. The data were correlated by an equation and also
by a computer solution to the differential equation describing mass transfer from values
of eddy viscosity and eddy diffusivity obtained from the literature.

Investigations of mass transfer have Liquid enters the section through a 1-in. %-in, equilateral triangle centers. Liquid
resulted in the development of methods pipe a t A and is distributed over the test leaves the test section a t overflow weirs B
for correlation and prediction of mass area by the perforated plate PP. The plate and C and flows into a holding tank. Test-
transfer rates in a variety of situations. is perforated by 1/16-in. holes spsced on section details are given in Figure 2. The
The effect of Reynolds number, Schmidt
number, and similar variables has been
analyzed for many systems; however,
the effect of concentration level has
received little attention.
Various terms in mass transfer calcu-
lations contain concentration level as a
variable. For example, the j factor for
mass transfer has been defined by LIQUID FLOW
Colburn ( 2 ) as
Fig. 1. Test section.

where p,, is a log-mean concentration


term.
Experimental studies on the effect of
concentration level on mass transfer
rates have been meager. Cairns and Roper
( 1 ) studied the concentration-level vari- r
.nl.
i y
able in the air-water system and corre-
f7Y
OVERFLOW WElR

lated their data by the following equation: ~

2-3d
2
7

= 0 . 0 2 1 ~(I)
~ ~ ~ ~ ~ I
LIQUID
OUTLLT

112" PIPE
the ratio (pB,/P) was varied from 0.15
to 0.99 by Cairns and Roper, and con- Fig. 2. Test section details; 1/8 in. sheet steel throughout inlet cross section and cross
centration level enters their correlation section for gas flow over liquid: 5 / 8 - by 4 1/16-in. length, 48 in. between weirs.
as an important variable.
The objectives of the present study
were to collect mass transfer rate data
in the air-carbon tetrachloride system
over a wide range of concentration level
and to study various methods of corre-
lating the data.

EXPERIMENTAL WORK

Liquid carbon tetrachloride was evapo-


rated from a plane surface in the bottom
of a rectangular duct into a turbulent air
stream passing through the duct.
Gas concentrations were varied from 0.00
to 0.70 mole fraction carbon tetrachloride.
Reynolds number varied from 600 t o 14,700
and the Schmidt number from 0.23 to 1.17.
Gas temperatures varied from 68 t o 91"C.,
and all experiments were conducted at
atmospheric pressure. The ratio (pB,/P)
varied from 0.29 to 0.81.
The test section is shown in Figure 1.
L. E. Westkaemper is at presenb with Rohm and
Haas Company, Pasadena, Texas. Fig. 3. Schematic flow sheet.

Vol. 3, No. 1 A.1.Ch.E. Journal Page 69


wetted area of the test section is 1.35 sq. ft. I I I I I 1 would best be determined by calculation
The remaining experimental apparatus is 45
from the measured evaporation rate, the
identified in the flow sheet shown in Figure inlet-gas rate, and the inlet-gas composition.
3. The entrance section preceding the test 30
Check runs showed that evaporation
section is a 36-in. length of rectangular 24
rates could be reproduced within lo%, as
duct of cross section 4 1/16 by % in., may be seen by comparing runs 12 and 14,
which is identical to the cross section for I8 2 and 4, and 18 and 18a in Table 1.
gas flow above the liquid in the test section. Additional experimental details and data
The saturator shown in Figure 3 is a I2 are available elsewhere (6).
36-in. length of 12-in. steel pipe provided
with two 1,500-watt immersion heaters
and a thermoregulator. A perforated plate CORRELATION OF THE DATA
6
in the base distributes incoming air through Dimensionless Groups
the liquid carbon tetrachloride in the REYNOLDS NO.
saturator, and the air leaves saturated with Various methods of correlating the
carbon tetrachloride vapors. Fig. 4. Correlation of the data with a d a t a have been used b y previous investi-
The bottom of the liquid holding tank is concentration term. gators. F o r convenience in comparing the
made from 12-in. steel pipe. A sight glass results with those from other sources, the
was placed on a length of 2-in. pipe forming d a t a were correlated b y plotting
the top of the tank to increase the sensitivity period, and if it was steady as the gas [(k,RTd)/D,].p,,/P. Scn-1 as a function
of volume readings. velocity was increased to operating condi- of Reynolds number. This term is equal
A steel top was used on the test section tions, the run was continued. t o (j). (Re) where j = (k,pB,Sc*)/G,.
during the experimental runs. However, a The compositions reported as experi-
glass top was used in preliminary runs to Figure 4 shows a correlation of the data
mental data were determined as follow.
observe the maximum gas velocities with The inlet composition was calculated from for this investigation. The equation of
stable gas-liquid interfaces. It was observed the temperature and pressure of the gas by the line shown is
that ripples appeared on the liquid surface assuming that it was a t its dew point
when gas velocities exceeded 10 t o 12 ft./sec. when it left the saturator. The validity of
A small increase in gas velocity a t this this assumption was verified in preliminary
level caused the ripples to grow and become runs by a n analysis of the gas made by
unstable; i.e., the ripples grew until en- quantitative charcoal absorption of the T h e standard deviation of the error is
trainment of liquid in the gas resulted. carbon tetrachloride in a measured gas 16%.
When the surface was unstable, the en- sample. Figure 5 shows the da ta plotted in a
trained liquid was carried out into the exit The change in the composition of the manner similar t o that of Figure 4,
lines, and thus the pressure drop through gas passing through the test section was except t h a t the term ( p B m / P )has been
the vent lines was more than doubled. relatively small; therefore error in gas
In the evaporation experiments the omitted from the ordinate. The equation
analysis is magnified when the change in
pressure drop through the vent lines was composition is taken. For this reason, it of the line shown is
carefully observed during the start-up was felt that the exit-gas composition
~

TABLE1. CALCTTLATED DATA


The standard deviation of the error is
Run U,,,, Gm, Re S C N, k,, 15%. T h e ordinate of- Figure 5 is equal
ft./sec. lb. mole/ lb. mole/ %mjP lb. moles/
(hr.)(sq. ft.) hr. (hr.)(atm./sq. ft.) to (j) (Re) when j = (k,PSco 56)/G,.
Data taken b y Cairns a nd Roper for
2 6.90 53.5 11,900 0.385 0.121 0.533 0.585 the evaporation of water into air in a
3 6.80 53.0 11,800 0.386 0.116 0.529 0.538
4 6.72 52.3 12,200 0.382 0.114
wetted-mall column m a y be plotted in
0.531 0.588
5 6.15 47.5 11,200 0.364 0.106 0,514 0.530 a manner similar to t h a t shown in Figure
6 5.37 40.0 9,800 0.350 0.097 0.499 0.457 4. A plot of Cairns and Ropers data is
7 5.20 39.5 9,500 0.355 0.085 0.511 0.471 shown in Figure 6. The equation of the
8 6.44 49.8 11,800 0.348 0.074 0.502 0.421 line shown is
9 5.87 45.0 12,200 0.315 0.076 0.466 0.489
10 5.21 39.3 11,400 0.295 0.078 0.423 0.473
11 5.56 42.0 11,600 0.308 0.079 0.651 0.583
12 4.26 32.7 5,980 0.471 0.0476 0.630 0.218
13 4.35 33.0 6,020 0.470 0.0586 0.625 0.273 T h e standard deviation of the error is
14 4.25 32.4 5,790 0.478 0.0507 0.625 0.228 217,. Cairns and Roper improved the
15 6.02 46.3 7,800 0.505 0.072 0.642 0.317
16 3.49 27.6 3,750 0.638 0.0249 0,722
correlation of their da ta as shown in
0.116
17 5.32 41.6 5,570 0.640 0.0510 0.734 0.262 Figure 7 and used the following equation:
18 2.71 21.3 2,690 0.690 0.0218 0.732 0.093
18a 2.81 21.9 2,770 0.630 0.0231 0.733 0.098
19 4.24 32.6 3,770 0.722 0.0380 0.748 0.158
20 3.54 27.3 3,460 0.675 0.0300 0.743 0.127
21 4.74 36.4 4,470 0.700 0.0416 0.747 0.202 = 0 . 0 2 1 ~ ~(1)~ . ~ ~
22 2.74 21.3 1,620 1.16 0.0312 0,810 0.077
23 1.70 13.5 1,110 1.11 0.0270 0.798 0,070 Here the standard deviation of the error
24 0.75 6.03 600 1.02 0.0276 0.782 0,060 is reduced to 10%. If the form shown by
25 3.25 25.3 1,910 1.17 0.0342 0.810 0.082 Equation (3) is used to correlate Cairns
26 2.28 17.7 1,400 1.13 0.0309 0.802 0.076 and Ropers data, the correlation is
27 6.88 53.0 11,800 0.432 0.116 0,471 0.519 quite poor.
28 4.62 34.1 13,100 0.230 0.0440 0.287 0.760 Equations (I) t o (4) were determined
29 5.60 42.0 14,700 0.242 0.079 0.312 0.613
30 2.43
from the d a t a b y the method of least
18.5 4,690 0.333 0.0185 0.520 0.190
31 4.05 30.9 7,490 0.353 0.0304 0.535 0.312
squares.
32 5.47 41.6 9,660 0.368 0.0467 0.551 0.411 Equations (2) and (3) show that the
definition of the .j factor influences the
Area of evaporating surface, 1.35 sq. ft. exponent of the Reynolds number in the

Page 70 A.1.Ch.E. Journal March, 1957


correlation. An exponent of 1.08 appears dc d dc d(W)
100

on the Reynolds number in Equation (2), -


dt
= -(D"
dx + E ) z - 75
where j = (kVpB,Sc09/Gm.
An exponent of 1.42 appears on the d dc d(vc)
v) 50

Reynolds number in Equation (3), where + - dY


(D,+E)dy--- aY 40

j = (k,PSc0.56)/Gm. 30
Cairns and Roper were able to focus a a(wc)
attention on the (p,,/P) term, since the +-
dx
(D,+
dc
E ) - - --
dx dx
(11) 20

Schmidt number was held constant in


their research. The ( p B m / P ) must be Equation (11) relates gas concentration
included in the correlation of their data. to time, gas velocity, eddy diffusivity, in
2006- 4 6 8 10,000 2O.WO
The air-carbon tetrachloride data of this molecular diffusivity, and position in the
paper may also be correlated by use of system. Under steady state conditions REYNOLDS NO.

the ( p B m / P )term, as is shown by Equa- for turbulent gas flow between parallel Fig. 5. Correlation of the data without a
tion (2). Hence the form shown by plates in the II: direction and mass transfer concentration term.
Equations (2) and (4) seems to be the in the y direction normal to the plates,
more general, as the form given by
Equation (3) will not correlate the data
of Cairns and Roper. Since Equations
(2) and (4) correlate the data of this
dc/dt = 0 and w = 0. For the experi-
ments of this investigation du/dz, d/dz
[(D, + E)(dc/dz)],and v are negligible,
and so Equation (11) simplifies to
-
" 100

80

50
investigation with deviations of 16 and
7" 40
15% respectively, Equation (1) used by VI

Cairns and Roper also will correlate the ' 30

data as well.
In the experimental work only small
X
Differential Equation changes in gas composition occurred in
In the case of molecular transport the 4-ft. length of the test section.
I I l l I I
processes, the transfer rates for momen- Temperature changes in the gas were 2000 4 6 8 10.000 20,000
tum, heat, and mass are given by small, as were changes in the gas volume. REYNOLDS NO.
Hence the assumptions that = 0
7-
du
= p- and (a/&) [ ( D v + E)(ac/dz)]= 0 are Fig. 6. Correlation of Cairns and Roper.
dY justifiable.
The assumption that v = 0, however,
is valid only over certain ranges of
operation. This assumption was necessary
in separating the variables in Equation
(11). The comparison of the evaporation
rates calculated by use of Equation (12)
with the experimental rates is shown in
Here the conductance terms are vis- Table 2. Mathematical details are avail-
cosity, thermal conductivity, and molec- able elsewhere (6). The calculated values
ular diffusivity. The potential terms are are seen to be in error a t low Reynolds
the gradients of velocity, temperature, numbers. Previous workers have shown 2000 4 6 8 10nOOe 20,000
and concentration. Each potential term that the presence of a velocity normal
is in differential form. to the wall is important in momentum
REYNOLDS NO.
Similar treatment is made in the case transfer studies. Fig. 7. Data of Cairns and Roper.
of turbulent transfer processes. Expand- Yuan and Finkelstein (9) have shown
ing the molecular transfer equations that the presence of a velocity normal
to the surface 1/100 as great as the main- locity based on these data is shown in
stream velocity increases the friction by Figure 8.
85%. The actual value of o in this work Since the variation of eddy diffusivity
is given by N A / c , , the moles transferred with distance from the wall has not been
per unit time per unit area divided by reported in the literature, this informa-
the total gas concentration in moles per tion must be obtained indirectly.
unit volume. When main-stream velocity It was noted that eddy viscosity, eddy
is lowered, this term becomes increasingly conductivity, and eddy diffusivity were
important for a given evaporation rate. defined by Equations (8) to (10) in the
The foregoing equations serve to define Consequently, it is felt that the poor hope that these quantities would be
em, eddy viscosity; ec, eddy conductivity; correlation of the data a t low Reynolds equal or proportional. Various researchers
and E, eddy diffusivity. I n each case numbers is caused by neglecting v. have conducted experiments to determine
the units are (distance)z per unit time. The solution of Equation (12) requires the relationship between the eddy prop-
The definitions take this form in the that velocity u and eddy diffusivity E erties.
hope that the similarity of the definitions be known as a function of position. The Woertz (8) measured concentration
will result in simple relationships between variation of velocity with distance from distribution and mass transfer rates of
eddy properties; that is, it is hoped that the wall has been measured by Corcoran water through air between parallel
the eddy properties will be equal or et al. (3) for air passing between parallel plates. Eddy-diffusivity values for the
proportional to one another. plates. The velocity distribution was center region between the plates were
In this work the primary interest is in measured in an isothermal as well as in obtained by differentiation of the con-
the mass transfer rate equation (10). a nonisothermal system. Temperature centration data and use of Equation (10).
From this equation and a material differences between the plates of 20F. Woertz also measured pressure drop
balance, the equation for turbulent mass did not appreciably alter the velocity and velocity distribution while the mass
transfer is derived: distribution observed. Dimensionless ve- transfer process took place. Eddy-

Vol. 3, No. 1 A.1.Ch.E. Journal Page 71


1.0 Reynolds numbers. This comparison is
shown in Figure 9, where the ratio of
0.8 eddy viscosity to eddy conductivity is
\ plotted vs. distance from the wall with
E parameters of Reynolds number.
0.6 0.6
There has been no experimental study
relating eddy conductivity to eddy
11 0.4 0.4 diffusivity; however, a relationship may
0.5 0.6 0.7 0.8 0.9 1.0
be inferred from the studies of Sage and
0.2 Y/ 2 A Woertz. Figure 9 shows that the ratio of
eddy viscosity t o eddy conductivity is
Fig. 9. Ratio of eddy viscosity to eddy approximately 0.7 for the center of the
0.50 060 070 080 090 100 conductivity, (3). duct a t a Reynolds number of 10,000.
Y/ 2 A
Hence,
Fig. 8. Velocity distribution between parallel eJe, = 0.7 (14)
plates, (3).
The work of Woertz &on-ed that e,/E =
0.62 and hence
EJE = 1.1 z1 (15)
6.0
Equation (15) suggests that eddy con-
ductivities may be taken as equal to
5.0
eddy diffusivities when these quantities
are defined by Equations (9) and (10).
The units in each case are (distance)* per
4.0 unit time. Figure 10 shows the eddy
conductivity values calculated by Sage
for air. The characteristic diameter in the
3.0
Reynolds number used by Sage was
twice the hydraulic radius, and this basis
2.0 is used in Figure 10. Equation (15) and
Figure 8 were used to calculate eddy
diffusivity values for air from eddy con-
I.0
ductivity values for air.
I n this research, however, the gas was
0.0 not air, but a mixture of air and carbon
0.50 0.60 0.70 0.80 0.90 1.00 tetrachloride. Consequently, a method of
Y/2A predicting the effect of composition on
Fig. 10. Eddy conductivity values (3). eddy diffusivity must be used. As sug-
gested by Deissler (4), this was done
with the following expression:
eddy diffusivity
= constant (16)
kinematic viscosity
viscosity values for the center region
between the plates were obtained by
calculating the shear stress from the
pressure drop, differentiating the velocity
data, and using Equation (8). TABLE
2. COMPARISON
OF COMPUTED
AND EXPERIMENTAL
RESULTS.
Woertz found that the ratio of the
eddy viscosity to eddy diffusivity in the Run (Re) E F N N % Dev.
center of the duct for Reynolds numbers calc. exp. calc. exp.
up to 60,000 was 0.62; i.e., 2 11,900 0.614 0.600 0.0110 0.0123 -10.6
4 12,200 0.608 0.595 0.0113 0.0115 -1.7
6 9,800 0.598 0.627 0.0107 0.0099 8.1
7 9,500 0.561 0.615 0.0101 0.0086 17.4
Further study of eddy properties was 8 11,800 0.623 0.687 0,0093 0.0075 24.0
made by Sage and associates (S), who 9 12,200 0.631 0.626 0.0077 0.0077 0.0
considered the relation between eddy 11 11,600 0.629 0.635 0.0083 0.0081 2.5
14 5,790 0.190 0.690 0.0155 0.0050 210
viscosity and eddy conductivity. 16 3,750 0.507 0.757 0.0052 0.0025 108
Sage determined eddy viscosity for air 17 5,570 0.221 0.682 0.0139 0.0051 172
passing between parallel plates by use of 18 2,690 0.227 0.806 0.0080 0.0022 264
Equation (8), as shown above. Sage also 21 4,470 0.497 0.707 0.0085 0.0047 81
measured heat flux and temperature 28 13,100 0.675 0.633 0.0039 0.0044 -11.4
distribution. Eddy conductivity values 29 14,700 0.670 0.684 0.0080 0.0079 1.9
were obtained by differentiating the 30 4,690 0.865 0.530 0.0005 0.0018 -73
temperature data and using Equation (9). 31 7,490 0.576 0.612 0.0036 0.0031 17.4
Sage obtained more detailed distribu- N , evaporation rate, std. cu. ft./sec.
tions of velocity and temperature than E, dimensionless outlet concentration, (c - c,)/(co - c,)
did Woertz. Hence Sage compared eddy % Dev, (exp. IV - calc. N)/(exp. N ) (100)
viscosity to eddy conductivity for various (Re),Reynolds number, characteristic radius equal to the clearance between plates

Page 72 A.1.Ch.E. Journal March, 1957


A similar relation was used by Woertz, 0.9
who found that the product (Ep) was
the same for air, carbon dioxide, and
helium at the same Reynolds number.
Since the viscosities of air, carbon dioxide,
9r
wwu
0.8

0.7
and helium do not vary greatly, Woertz
could have used the product ( E p / p ) = 0.6 0.7 B.8 0.9 1.0
D,+El,

(E/v) as Deissler suggested.


Equation (16) was used t o calculate Y/2A
eddy diffusivity in air-carbon tetra-
Fig. 11. Ratio of total viscosity to total
02 I I I I b
- 0

chloride as a function of position from conductivity (3).


0
5 6 7 8 9 I0
the eddy diffusivities in pure air, except 5 Y
00

for the value of eddy diffusivity at the Fig. 12. Total diiksity values for run 6.
liquid surface. The viscosity of the air-
carbon tetrachloride mixtures was calcu- and
lated by use of the correlation and data
of Tlrilke and Bromley ( 7 ) . The surface velocity of the gas was
A final calculation was necessary to not high enough (a x 0.1) to alter
obtain the eddy diffusivity at the liquid greatly the velocity profile shown in
surface. Since one of the parallel plates Equation (22) gives the total viscosity Figure 8. As this velocity distribution
of this research was actually a liquid (v + r,) a t the surface as a function of was approximated in a computer solution
surface, it could not be considered as a the average velocity U,,, and the velocity by a step function, accuracy was not
fixed surface. The drag of the gas stream gradient a t the surface (du/dy).. improved by adding the correction at the
on the liquid surface causes a surface It was assumed that the velocity wall.
velocity in the liquid, as shown by Tek gradient a t the liquid surface was equal An electronic differential analyzer was
(5); hence the gas velocity does not drop to the velocity gradient a t the wall in used to solve the ordinary differential
to zero a t the liquid surface but instead Sages experiments. Accordingly, the equations (6) resulting from the separa-
drops to a value equal to the surface values of (duldy), were calculated by tion of variables in Equation (12).
velocity of the liquid. differentiating the velocity data shown A summary of the outlet concentra-
Tek has developed equations which in Figure 8 and the total viscosities tions calculated by computer solution
may be used to calculate the liquid surface calculated from Equation (22). of Equation (12) is shown in Table 2,
velocity and the surface shear from the Actually the velocity gradient at the which also shows the experimental outlet
main-stream gas velocity and the prop- liquid surface will be somewhat less than concentration, which was calculated from
erties of the liquid. This calculation the velocity gradient a t a fixed boundary, the inlet-gas concentration, the inlet-gas
involves the use of Equations (17) and and therefore the values of total viscosity rate, and the evaporation rate.
(18), which were developed by Tek ( 5 ) . obtained by the foregoing procedure are Table 2 shows good agreement between
low. As a result, the values of eddy con- the calculated and experimental evapora-
ductivity or eddy diffusivity derived from tion rates for Reynolds numbers over
the eddy viscosity will also be low and 7,000. The average deviation is 10.5%
will tend to decrease the rate calculated for these runs. Even though numerous
from Equation (12). approximations must be made in using
where (dhldx) = ~ T = Usurface
~ ~ ~ ~
Sage reports values of the ratio of total Equation (11) to correlate mass transfer
inclination due to gas velocity, viscosity (v + em) to total conductivity data, the outlet concentrations predicted
and [k/(pc,) + E , ] , and these are shown as a by this equation agree closely with the
function of Reynolds number and position experimental concentrations for Reynolds
in Figure 11. Thus Figure 11 can be used numbers over 7,000, as seen in Table 2.
to evaluate the total conductivity a t the No arbitrary constants are used in the
surface from the total viscosity calcu- equation a t any point. This suggests that
Eliminating (u,)from Equations (17) lated from Equation (22). The total Equation (11) may prove as useful in
and (18) gives diffusivity a t the surface was taken as mass transfer as the Fourier-Poisson
equal to the total conductivity a t the equation has in heat transfer.
surface. Certain predictions may be made
This step is somewhat inconsistent concerning the effect of concentration
with the use of the data of Woertz and level on mass transfer by examination of
Equation (19) may be used to calculate Sage to show that e,/E = 1. Consistency Equation (12). I n addition to the terms
the shear stress at the liquid surface, would require that (k/pc,) be subtracted containing concentration gradient, this
which is also given by Equation (8). For
the case of the liquid surface, in units
from total conductivity values to give equation contains the term (Do + E),
eddy conductivity and eddy diffusivity. which is a function of concentration
consistent. with Equation (19), Equation Total diffusivity would be obtained for level. This term depends on concentra-
(8) becomes use in Equation (12) by adding D, to tion level because of the variation of E
the eddy diffusivity. with concentration, as has been discussed.
However the equation eJE = 1 is Hence Equation (12) suggests that for
verified only for the center region of gas a system in which there is no change in
flow, and the approximation of taking eddy diffusivity with concentration, there
total conductivity equal to total diffu- would be no effect of concentration level.
sivity is probably as accurate as assuming Also if (eddy diffusivity)/(kinematic vis-
that rJE = 1 throughout the flow cosity) = constant, a mixture with a
domain. Once a total diffusivity a t the kinematic viscosity independent of con-
liquid surface is known, i t may be used centration should also have a n eddy
as a limiting value for the total diffu- diffusivity independent of concentration.
sivities (6) in the main gas stream, as The air-water system is one which has
shown in Figure 12. a gas kinematic viscosity largely inde-

Vol. 3, No. 1 A.1.Ch.E. Journal Page 73


TABLE3. CORRELATION OF DATAOF CAIRNS
AND ROPER WITH EQUATION (12) k = thermal conductivity, B.t.u. per
Run Reynolds No. Gas AT, Outlet concentration, E unit time per unit area per degree
OF . Equation (12) Cairns and Roper Fahrenheit per foot
Ic, = mass transfer coefficient, moles
4 6520 251 0.29 0.43 per unit time per unit area per
9 9095 254 0.26 0.38 unit concentration difference
11 3273 342 0.15 0.39 k, = mass transfer coefficient, moles
18 4828 315 0.21 0.31 per unit time per unit area Der
unit pressure difference
kT = a constant
pendent of concentration. At 212F. the coefficients to Equation ( 2 ), lead one to N A = mass transfer rate, moles Per unit
kinematic viscosities of air and steam suspect that functions of this type time per unit area
differ by only 7%. This system would be between the dimensionless groups may p = kkal Pressure
expected to show little effect of concen- be too simple to represent adequately P = gas Partial pressure
tration level on mass transfer rates. mass transfer data over a wide range in PB, = logarithmic mean Partial pres-
Extensive data in the air-water system the level of concentration of the trans- sure of component B
have been taken by Cairns and Roper. ferring component. p, = gas partial pressure a t point i
While these data clearly require a con- I n a sense this is corroborated by the Q = heat flow rate, I3.t.u. per unit
centration variable ( p B m )in their correla- relatively successful result of applying time
tion, this apparent contradiction of the mass equation in differential form 4 = heat flow rate, I3.t.u. Per unit
Equation (12) may be explained. Cairns to the data correlation even though a time per unit area
and Roper operated a wetted-wall column considerable number of assumptions and R = the gas constant
in their investigation, and the tempera- the use of data on eddy properties from Re = the Reynolds number
ture of the gas passing through the many sources were involved. Of particular SC = the Schmidt number
column changed by as much as 340F. interest is the fact that the correlation = temperature
Furthermore, the inlet-gas temperatures by the differential equation begins to fail t = time
varied from run to run. Hence there was exactly where i t might be expected to u,,, = average gas velocity in direction
a kinematic viscosity change, caused by fail a t low flow rates, or Reynolds of flow
temperature change. The variation in the numbers less than 7,000, where the bulk = maximum gas velocity in direc-
kinematic viscosity of the inlet gas for motion normal to the axis of flow and in tion of flow
various runs mas over 30y0 and of the the direction is no longer negligible. u = component of gas velocity in z
gas passing through the test section was This suggests that, as further research direction
as high as SO%. Most of the data taken makes available velocity distributions in VI = kinematic viscosity of liquid
by Cairns and Roper were for Reynolds turbulent-flow regimes of various geome- 2) = component of gas velocity in y
numbers less than 7,000, where Equation tries, the use of computing machines will direction
(12) is not satisfactory. Nevertheless, make feasible the correlation and applica- W = component of gas velocity in z
four runs of Cairns and Roper were cor- tion of mass and heat transfer data in the direction
related by Equation (12) with the differ- form of the differential equations de- z = coordinate axis; distance
ential analyzer. Each calculated dimen- scribing mass and energy balances as a Y/y = !//2a = coordinate axis; distance
sionless outlet concentration for these more basic technique for treating many % = eddy conductivity, (distance)2
runs was lower than that reported by problems now being approached only per unit time
Cairns and Roper (Table 3). Since the through dimensional analysis. em = eddy viscosity, (distance)z per
dimensionless concentration p = (c - unit time
c,)/(co - cW), the mass transfer rate NOTATION
X = kinematic viscosity, (distance)z
predicted by concentrations from Equa- per unit time
tion (12) would be high in each case. I n A = area p = absolute viscosity
the derivation of Equation (12) it was a = one-half the clearance between I#I = density
assumed that the eddy diffusivity parallel plates T = shear stress
was independent of x. The large change al = liquid depth
in kinematic viscosity and eddy diffu- Co = concentration of the inlet gas to
sivity in the direction of flow in the work the test section LITERATURE CITED

of Cairns and Roper makes it unlikely cw = concentration Of the gas at the 1. Cairns, R. C., and G. H. Roper, Chem.
that Equation (12) is adequate to corre- liquid surface Eng. Sci., 3, 97 (1954).
late their data. c = concentration, moles per unit 2. Colburn, A. P., and T. H. Chilton, Znd.
If the air-water system is to be used volume Eng. Ghem., 26, 1183 (1934).
to study the effect of concentration level, F = dimensionless concentration, 3. Corcoran, W. H., F. Page, W. G.
(c - c,)/(co - )c, Schlinger, and B. H. Sage, ibid., 44,
the temperature drop in the gas must be
c, = heat capacity 410 (1952).
held at a minimum. This suggests that 4. Deissler, R. G., Natl. Advisory Gomtn.
an apparatus similar to that used in this D, = molecular diffusivity Aeronaut. Tech. Note 3145, 15 (1954).
study might prove more successful than d = characteristic diameter 5. Tek, M. R., Ph.D. thesis, Univ. Mich.,
a wetted-wall column, where control of E = eddy diffusivity, (distance)z per Arbor (1953).
gas temperature is difficult. unit time 6. Westkaemper, L. E., Ph.D. dissertation,
gc = conversion factor, (32.2 pounds Univ. Mich., Ann Arbor (1955).
mass/pound force)(feet/second)z 7 . Wilke, C. R., and L. A. Bromley, Ind.
CONCLUSION
G, = flow rate, moles per unit time per Eng. Chem., 43, 1641 (1951).
I n summary, the facts that data of this unit area 8. Woertz B. B., and T. K. Sherwood,
investigation may be correlated equally h = heat transfer coefficient zbid., 31, 1034 (1939).
= a modified j
9. Yuan, s. w., and A. 3. Finkelstein,
well by Equations (I), ( 2 ) , and (3) and j factor for General Proceedings of Heat Transfer
that the data of Cairns and Roper may transfer and Fluid Mechanics Institute, Los
be correlated best by Equation (1) and j, = the j factor for transfer de- Angeles, Am. SOC.Mech. Engrs. (1955).
less satisfactorily by Equation (4), fined by Colburn, (krpBmSc2/3)/
which is similar in form but not in G, Presented at A.I.0h.E. New Orleans meeting

Page 74 A.1.Ch.E. Journal March, 1957