Extraction in Spray and

Packed
Columns
T. IC. SHERWOOD, J. E. EVANS,
AND J. V. A. LONGCOR
Massachusetts Institute of Technology,
Cambridge, Mass.

Data are presented on liquid-liquid extraction from single

drops which indicate that the interior of the drop is not
stagnant but is considerably agitated. Data on extraction in
spray and packed columns show an initial increase in extraction coefficient with increase in rate of flow of either continuous or discontinuous phase, presumably due to increase
in interfacial surface as the holdup increases. A subsequent
decrease in the coefficient at the highest flow rates is explained as the result of drop coalescence, this being especially
noticeable at high rates of flow of the continuous phase.
The coefficient is greater if the discontinuous phase does not
wet the packing. The coefficients are largest in the packed
column but the flooding rates are greatest in the spray (unpacked) column.

ECAUSE of its effectiveness as a complement to distilla-

tion in the separation of materials, liquid-liquid extraction has in recent years assumed considerable importance as one of the unit operations of chemical engineering.
Within a relatively few years the solvent refining of petroleum
products has become common practice, and much has been
done in developing the equipment necessary for large-scale
operation.
The development of the more theoretical aspects of extraction as a unit operation has been relatively slow as compared
with the widespread adoption of the process on an industrial
scale. Hunter and Nash (11-14) have described both the
diffusional basis of extraction and the graphical methods of
computation which are of value in making the necessary
stoichiometric calculations. The latter have also been described with unusual clarity by Evans (8). The analogy between distillation and extraction and the meaning of reflux
in extraction are discussed by Saal and Van Dyck (17) and
by Thiele (go), and have been presented particularly well
by Varteressian and Fenske (22, 23).
Relatively little information is available in the literature
with regard to the performance of extraction equipment.
Fallah, Hunter, and Nash (9) and Strang, Hunter, and Nash
(19) report data on extraction in a wetted-wall column, and
on the flow conditions in such a column. Elgin and Browning
(7) and Appel and Elgin (2) report investigations of countercurrent extraction in a spray column; acetic acid, isopropyl
ether, and water were used in the first case, and benzoic acid,
toluene, and water in the second. The latter investigation included a study of the operation of a packed column, but since
0.5-inch Berl saddles were used in a 2.03-inch i. d. column,
channeling along the wall was doubtless large. Sherwood (18)
reports data of Demo and Ewing (6) on extraction of acetic
acid from water by benzene in a 3.55-inch i. d. tower packed
with 0.5-inch carbon rings. Varteressian and Fenske (21) report data on extraction in the system benzene-ethyl alcoholwater in a 0.55-inch column packed with small metal chain
and nickel wire rings. Rushton (16) describes the results of
experiments in which oils were treated by countercurrent
extraction with nitrobenzene in a 216/16-in~htower packed
with various rings and saddles,
to 1 inch in size.

It is apparent from the limited literature on performance of

countercurrent columns that no general correlation of the
data for design purposes may be expected until more published information is available. The present article presents
the results of a continuation of the work begun by Demo and
Ewing (6) on extraction in a 3.55-inch column packed with
0.5- and 1-inch carbon rings and 0.5-inch Berl saddles, as well
as with an unpacked spray column. The investigation is
analogous to those of Elgin and Browning and of Appel and
Elgin in that definite three-component systems were used,
and similar to the work of Rushton in that several packings
were investigated. Although the tower was larger than those
used by these investigators, the ratio of tower diameter to
packing size may have been too small in the case of 1-inch
packing. It is sometimes assumed that this ratio should be 8
or larger if the results are to be considered typical of the performance of towers of large cross section.
Simultaneously with the study of the packed tower, an experimental investigation on extraction from single drops was
undertaken, This is also described, as it throws light on the
mechanism of diffusion into the dispersed phase. I n both
studies the solute was acetic acid which was extracted from
water by benzene and by methyl isobutyl ketone.
,

Procedure for Extraction from Single Drops

Solvent containing acetic acid was introduced through a glass

nozzle mounted vertically at the bottom end of a glass column,
1.74 inches i. d. and 60 inches tall. The solvent drops formin
at the nozzle tip rose through water which filled the unpackes
column. The solvent feed was controlled by dropping from an
analytical buret into a side tube connected t o the glass nozzle.
The level in this side tube was maintained constant by close
observation and regulation of the buret cock. This provided a
uniform feed rate and an accurate measure of the amount of
solvent introduced t o the column.
The to of the column was fitted with a cork stopper carrying
a 0.55-incR bent glass tube from which the solvent was withdrawn
t o a measuring buret. The under side of this stopper was hollowed to form a cone-shaped receiver for the solvent drops arriving at the top of the column and so preventing any holdup of
drops under this stopper. A small amount of water was introduced at the bottom of the column t o force some water out with
the solvent leaving the top and thus to retain in the combined
top product all acid present in the solvent. Both phases removed
from the top were titrated, and the acid found was assumed to
1144

SEPTEMBER, 1939

INDUSTRIAL AND ENGINEERING CHEMISTRY

have been present in the solvent drops arriving at the top. I n

using the ketone a correction was made for the small acid concentration of the water phase in the tower, but with benzene this correction was quite negligible. I n both cases the water in the
tower was changed after each run. Total material balances
checked within 2 per cent in all tests.
The acid content of inlet and outlet solvent phases, outlet
water phase, and water in the column a t the end of the run were
obtained by titration with standard sodium hydroxide using the
method described below for the samples from the packed column.
The water in the column at the end of the run was sampled at
both top and bottom of the column in the runs with the ketone.
Rate of solvent and amounts of water fed and of solvent and
aqueous products were obtained from the buret readings. The
rate of drop formation was obtained by counting the drops formed
in 5 minutes. "Time of contact was taken as the average time of
drop formation and rise obtained by several measurements with
a stop watch on single drops.
Several nozzles were used to obtain different drop sizes; the
smallest nozzle was a 0.0010-inch i. d. stainless steel hypodermic
needle. The other nozzles were fire-polished glass tubes. The
effective height of the column was varied from 2.0 to 57.7 inches
by adjusting the position of the lower stopper carrying the nozzle.
All runs were made a t 22-28' C. (71.6-82.4' F.).

1145

or lowered as desired. The elevation of this overflow determined

the position of the interface between phases in the tower, which
could be controlled easily in this way. This overflow device in
the raffinate line is indicated in Figure 1.
Both feed liquids were pumped continuously from storage vessels at floor level to head boxes situated on a platform about 14
feet above the floor. The feed to the column was by gravity from
these constant-head supply vessels, the overflow in each case being returned to the storage vessels. Calibrated orifices were inserted in each feed line, as Figure 1indicates, and fitted with petcocks to remove air from the manometer leads. Dyed butyl
phthalate was used as a manometer fluid. Glass carboys were
used for acid storage, with a copper head box and 0.59-inch glass
lines. Steel drums were used for the solvent, with a/,-inch iron
pipe lines.

Packed Tower
Solvent and aqueous streams were contacted in a 3.55-inch i. d.
Pyrex glass tower, 66 inches long and mounted vertically. This
was fitted with headers and distributing nozzles at both ends and
operated empty as a spra tower or acked with one of three
FIGURE
2. BRASSHEADER
packing materials. I n a&ition to t%e tower, the necessary
auxiliaries included storage vessels, feed and product lines, orifice
meters, pumps, and an overflow device to control the interface
The solvent extract was strip ed of acid for re-use by contactin
between the two phases in the tower (Figure 1).
with water or with a dilute sol%ion of sodium hydroxide. Acia
In all runs the direction of diffusion was from aqueous layer
raffinate was made up to 6.0 per cent for re-use by adding glacial
to solvent layer-i. e., an aqueous solution, initially 6.0 per cent
acetic acid. A layer of solvent was maintained a t all times above
acetic acid, was extracted by either benzene or methyl isobutyl
the solution in the acid storage carboys, and a layer of water was
ketone. The aqueous layer, either feed or raffinate, will be rekept in the solvent drum. Since these vessels were agitated by
ferred to as acid. Since the flow of one phase past the other is
the continuous circulation to the head boxes, each feed liquid was
necessarily by gravity, the heavier acid phase entered the tower
maintained saturated with the other phase.
a t the top and was withdrawn a t the bottom, while the solvent
Before each run the feed liquids were recirculated through the
passed in the reverse direction.
head boxes for ap roximately one hour in order to saturate each
At each end of the glass tower was a brass header consisting of
phase with the otRer layer. The continuous phase was admitted
a cylindrical chamber 3 inches high and 3.5 inches i. d. The inuntil the column was about three quarters full, and the second
comin liquid was fed into the side of the chamber whence it enfeed was then started. The position of the interface was adtered &e tower through six 0.120-inch i. d. brass tubes extending
justed by means of the overflow control, and the flows were set
2 inches into the tower and 2 inches into the header chamber.
and held a t the desired rates. The
These were spaced symmetrically
position of the interface was ap roxi
at a radius of 1 inch from the
mately level with the ends of tge si;
center of the tower. The outgoing
small feed tubes either at top or
liquid was withdrawn through a l/4bottom header, depending on which
inch brass pipe leading from a hole
phase was dispersed. After about
in the center of the header plate.
four com lete changes of the conOne of these headers is shown in
tinuous p\ase in the column, judged
Figure 2. The six small feed tubes
sufficient to obtain steady state, a
extend into the header chamber,
set of four samples was obtained.
with the supply tube feeding the
A second set was then obtained from
header at the right and the brass
5 to 15 minutes later, depending on
pipe through which liquid was withthe flow rates, and the test was ended.
drawn a t the left. The cover plate
Since it was possible to adjust the
(shown removed) was fitted with a
temperature of the room, the t:sts
petcock and a glass thermometer.
were all made a t 25" * 2 " C. (77 *
I n most of the runs with packing
3.6 ' F.).
the packed height was 54 inches, with
The acid samples were titrated
6-inches free space above and below
with 1N sodium hydroxide, b means
the packing. The packing rested on
a S/~-inch-meshnickel wire grid. I n
of thymol blue indicator. T%e bena few runs only 20 inches of packing
zene extract was analyzed by shaking
were used, with the free space above
50-cc. portions with an equal amount
the packing increased to 40 inches.
of water and titrating the mixture
with 0.1 N caustic; thymol blue indiI n order to avoid the occurrence of
appreciable extraction in the large
cator was used and the mixture was
free space above the packing in these
shaken violently until the end point
tests, the six upper feed tubes were
was reached.
c
extended by 0.24-inch glass tubing to
A faint blue in the water layer was
E
taken as an end point. As B check
introduce the acid a short distance
above the packing. A wood spacer
it was found that the same end
point was obtained when sufficient
maintained the glass extensions in
the same relative position as the
ethyl alcohol was added to make
short brass tubes previously dethe two phases completely miscible.
scribed.
The ketone extract was titrated in a
PUMP
PUMP
The aqueous raffinate leaving the
similar manner with 1 N caustic.
column passed from the bottom
The benzene feed was titrated with
header through a swivel pi e to an
0.01 N caustic; the ketone feed with
overflow vessel which could {e raised
FIGURE
1. DIAGRAM
OF APPARATUS
0.1 N caustic.

VOL. 31, NO. 9

INDUSTRIAL AND ENGINEERING CHEMISTRY

1146

Equilibrium data for the system benzene-acetic acid-water

were available in the literature (6). Corresponding data for
the system methyl isobutyl ketone-acetic acid-water were
obtained experimentally. Whereas the acetic acid concentration in water is roughly thirty times as great as in the benzene
phase in equilibrium, it was found that the distribution of
acetic acid between methyl isobutyl ketone and water corresponds to about equal concentrations in both phases. The
ketone is obviously a much better solvent than benzene for
the removal of acetic acid from water. The mutual solubility
and equilibrium data for the ketone system are given in
Table I.
TABLEI. EQUILIBRIUM
AND MUTUAL
SOLUBILITY
DATAFOR
SYSTEM ACETICACID-WATER-METHYL
ISOBUTYL KETONE AT
25" c.

-Mutual

Soly. Data-

Ketone

Water

Acid

1.55
3.7
10.5
17.4
26.0
37.6
51.6
66.4
81.6
97.9

98.45
76.8
57.5
48.4
39.6
29.1
19.2
12.0
6.5
2.12

19.5
32.0
34.2
34.4
33.3
29.2
21.6
11.9

-Equilibrium
-Ketone LayerAcetic
acid
Density

G./cc.

1.87
8.9
17.3
24.6
30.8
33.6

0.798

0.804
0.807
0.809
0.811
0.812

Dat-Water
Acetic
aoid

%
2.85
11.7
20.5
26.2
32.8
34.6

LayerDensity
G./cc.

0.995
0.996
0.998
0,999
1.000
1,001

Extraction from Single Drops

The quantities varied were drop size, column height, inlet
concentration of acid in solvent, and solvent feed rate. Both
benzene and methyl isobutyl ketone were used. Six runs were
made with an acid concentration of 0.0755 pound mole per
cubic foot, and five runs a t a concentration of 0.0474 pound
mole per cubic foot in the inlet ketone, all with a column
height of 57.7 inches. Thirteen runs were made a t an inlet
concentration of 0.0603 pound mole per cubic foot, with the
column height varied from 2 to 57.7 inches. Several inlet
nozzles were employed with each inlet acid concentration of
the ketone. Drop diameters, calculated from the measured
feed rate and the drop count on the assumption that the
drops were spheres, varied from 0.0745 to 0.137 inch. This
range of drop diameters is quite small compared with the
range of nozzle diameters used. The smaller drops appeared
to be spherical, but the larger drops were noticeably flat,
with horizontal axes perhaps twice the vertical axes.
Seven runs were made with benzene containing 0.00576
pound mole of acetic acid per cubic foot a t a column height
of 57.7 inches. Twelve runs were made with 0.00374 pound
mole of acid per cubic foot in benzene with column heights
from 2 to 57.7 inches. The benzene drops were larger than
those obtained with the ketone, the calculated diameters
varying from 0,109 to 0.221 inch. The drop diameters were
calculated on the basis of spherical drops, the drop volume
being obtained by dividing the volumetric feed rate by the
number of drops formed per unit time. In the runs with benzene the equilibrium concentrations in benzene corresponding
to the observed water-phase concentrations were negligible, and
the actual benzene concentrations could be taken as equal to
the over-all concentration driving forces on the benzene basis.
The calculated transfer coefficients are plotted against drop
diameter on Figure 3; all the data shown are for the 57.7-inch
column height. The values of K were obtained from the equation :
where L

= ketone or benzene

rate, cu. ft./hr.

Cp, C1 = acid concentrations of drops at exit and inlet, respectively, lb. mole acetic acid/ cu. ft.
A = calculated area term, in sq. ft., of total drops in column
at any time, obtained from measured feed rate, drop count (no.
of drops per min.), and measured time of formation and rise of
drops from bottom to top.

The calculation of A is based on the assumption that the drops

are spherical. The logarithmic mean driving force, AC, m.
is based on C1 and Cz and the equilibrium concentrations in
ketone or benzene corresponding to the observed concentrations
in the aqueous phase a t the bottom and top of the column.
From Figure 3 it is apparent that K increases with drop
size for both systems, and that for the same drop size K Kis
greater than K B . Variations in acid content of the inlet
ketone have no effect on K K ,but two curves are obtained for
K Bfor the two inlet acid concentrations in benzene. The correlation is essentially the same if K is plotted against Reynolds
number for the rising drops. Since the liquid-phase diffusivities are probably almost equal in the two cases, it might be
expected that K K and K B would not differ greatly. The interfacial tension for the benzene-water system (10) is between
33 and 35 dynes per cm., whereas that for the ketone (1) varies
from 8.8 for ketone-water with no acid to 3.0 for the feed containing 0.075 pound mole of acid per cubic foot. Although the
interfacial tension should influence the drop size, it is difficult
to see how it might affect K for a given drop size. Except for
minor differences in velocity of rise, the conditions outside
the drop were essentially the same for both ketone and benzene, and it seems logical to look within the drop for an explanation of the observed differences in K for the two systems.

<

0
L

0.04

0.08

0.12

DROP DIAMETER

0.18

0.20

0.24

IN

FIGURE3. EFFECTOF DROPSIZE ON ExTRACTION FROM SIR'GLE

DROPS

If the fluid within the drop is quite stagnant as the drop

rises through the column, it should be possible to calculate
the amount of acid transferred from the laws of unsteadystate diffusion in a sphere. Xewman (16) gives values calculated from the theoretical equation for a sphere with which
the experimental results may be compared. The ratio A of
unextracted solute to total solute which it is possible to extract may be taken as approximately equal to the ratio of
AC a t the top of the column to AC a t the bottom of the
column. This is related by the theoretical equation to the
ratio D 8 / R 2 , where D is the diffusivity for the solute in
the drop, 8 is the time of contact of drop with water, and R is
the drop radius. The correlation of diffusivities given by
Arnold (3)indicates that D a t 25" C. (77" F.) should be about
square inch per second (1.9 X
sq. cm. per
0.3 X
second). For the first run with ketone, A is 0.0072, for which
the theoretical value of D 8 / R 2 is 0.45. Since the drop radius
was 0.05 inch and the time of formation and rise 14.4 seconds,
the actual value was:

SEPTEMBER, 1939

D e - 0.3
_
R2

INDUSTRIAL AND ENGINEERING CHEMISTRY

X

X 14.4 = o.0173
(0.05)

1147

TABLE11. COMPARISON
OF SINGLE-DROP
DATAWITH DIFFUSION
THEORY
(COLUMN
HEIGHT57.7 INCHES)

Methyl Isobutyl Ketone

Benzene
Therefore, the extraction proceeds a t a rate correspondRatio of
Ratio of
ing to an effective diffusivity larger than the molecular
Drop
eflective
Drop
effective
diffusivity in the ratio of the two values of D9/R2, which
to actual
diameter,
to actual
diameter,
in.
diffusivitv
In.
diffusivitv
is 26. The value
0.075
17.7
0.109
11.4
0.10
26.0
0.153
26.0
of D of 0.3 X
I .o
0.115
23.0
0.185
36.4
square inch
0 .a
0.137
36.6
0.210
43.3
per second is an
0.6
estimate only but
is probably good
0.4
sizes, and the location of the curves would vary with drop size.
within 50 per cent.
This would be due not only to the variation in conditions deT h e theoretical
termining diffusion into the rising drop but also to the fact
values from New3
that the ratio of time of drop formation to time of rise increases
a
0.2
man assume no
L
rapidly with drop size. The importance of the latter ratio is
surface resistance
8
also apparent when the feed rate is varied. Two runs with
to diffusion in the
E
the ketone under quite similar conditions resulted in drops
water phase; if
E 0.1
0.10 inch in diameter, but one feed rate was twice the other.
such a resistance is
0.08
For the higher feed rate, A was 0.088, whereas in the other,
allowed for the
0.06
A was 0.044. The longer time of drop formation resulted in
ratio of the two,
e
a n appreciable increase in amount of solute extracted.
values of D 0 / R 2
3 0.04
Continuous Extraction i n Column
would be even
larger. Similar
The experimental data and calculated results are sumc a l c u l a t i o n s for
0.0 2
marized in Tables 111 and IV. The quantities varied were
other typical runs
flow rate of dispersed phase, L D ,flow rate of continuous phase,
are summarized in
Lc, nature of solvent, type of packing, phase dispersed, and
Table 11. It may
0.0 I
number of feed nozzles. In tests with benzene, LD (cubic feet
be seen that the
0
50
100
150
per
hour per square foot of tower cross section) was varied
ratio of effective
COLUMN HEIGHT, CM
from 10 to 40, and LC from 3 to 60. With methyl isobutyl
to actual diffusiviFIGURE4. FRACTION
EXTRACTED
us.
ketone, L Dwas varied from 10 to 70, and LCfrom 10 to 100.
COLUMN
HEIGHT,ILLUSTRATING
EXties increases with
TRACTION DURING DROP
FORMATION The packings used were half-inch and one-inch carbon Raschig
drop size, and that
for drops somewhat smaller than those employed,
TABLEI11 EXTRACTION
OF ACETIC
ACIDFROM WATERBY BENZENE
the theoretical diffusion equation
Flow Rates
might be expected to hold. The
Concn., Lb. Mole Acetic Acid/Cu. Ft.
KBa, Lb.
Cu. Ft./(Hr.)
Packed Mole/(Hr.)
(Sq.Ft.)
most apparent explanation of the
Height
Run
Acid
Acid
Benzene
Benzene
Benzene Acid
Ft.
in
out
In
out
No.
results with larger drops is that the
Half-Inch Carbon Rings, Benzene Dispersed
interior of the drop is fairly well
0 .000081
0.00159
4.42
1
0.0611
0.0585
3.1
3.2
10
5.9
mixed and not stagnant. As the drop
0.00192
4.42
0.000128
2
0.0649
0.0621
10
10
4.9
2.0
0.00190
4.42
0.000061
3
10
20
0.0631
0.0618
5.1
2.0
rises there is a frictional drag on
n. no0107
0.00166
4.42
4
10
30
0.0604
0.0600
2.8
3.6
0.00170
4.42
0 . ooooss
5Fa
10
40
0.0606
0.0604
2.7
3.7
the top half inducing convection
0.00167
4.42
0.000050
6Fa
13.2
40
0.0609
0.0602
5.4
2.5
down the sides and up the central
0.000080
0.00144
4.42
0.0605
0.0514
20
3
6.0
3.3
0,000064
0.00169
4.42
0.0614
0.0602
SF,
22
30
7
.
7
2
.9
axis. Viscosity damps these currents
0.00149
0.000120
4.69
0.0616
0.0532
30
3
9
6.5
4.6
0.00187
4.42
0.000085
10
30
10
0.0646
0,0595
in very small drops, but in the
11.3
2.6
0.00151
0.000048
4.42
11
30
20
0.0611
0.0590
8.3
3.6
larger drops they are of sufficient
0.00150
0.000044
4.69
12
30
20
0.0598
0.0576
8.7
3.4
0.00141
0.000047
4.69
13
0.0615
0.0516
38
3
7.9
4.8
magnitude to transfer solute much
0.00180
4.42
0.000070
14
38
10
0.0644
0.0577
13.7
2.8
0.00157
4.42
0.000058
15Fa
more rapidly than would be possible
38.5
20
0.0620
0.0592
10.8
3.6
Rings,
Aci
IHalf-Inch
Carbon
d
Dispersed
by molecular diffusion.
0.000053
0.00162
4.69
30
0.0614
0.0608
10
2.9
3.4
The variations in K between ketone
0,000053
0,00138
20
30
0.0605
0,0595
4.69
17
4.8
4.2
0.000052
0.00131
4.69
30
0.0614
0.0602
18
30
5.8
6.2
and benzene shown in Figure 3 are
19
40
30
0.0615
0,0599
0.000046
0,00121
4.69
7.4
5.4
presumably due to variations in
Spray Column, Benzene Dispersed
0.0580
0.000019
0.00160
5.02
20
30
10
0.0619
7.7
3.9
convection currents within the drop,
0.000010
0.00164
5.01
21
30
20
0.0626
0.0604
7.9
3.8
caused by differences in density,
0.000061
0.00169
5.20
22
30
40
0.0621
0.0609
8.4
3.6
0.000010
0.00177
5
.
2
1
23
30
60
0.0617
0.0609
1
0
.
0
3
.0
viscosity, and shape of the liquid
Half-Inch Berl Saddles, Benzene Dispersed
drops. Another factor of some im0.000074
0.00167
24
30
3
0.0618
0.0488
4.65
10.7
2.8
0.000040
0.00183
4.65
25
30
6
0.0619
0.0530
13.8
2.2
portance, however, is the appreciable
0,0567
0,000018
26
30
10
0.0617
0.00189
4.65
13.8
2.2
extraction occurring as the drop is
20
0.0624
0.0601
0.000048
0.00179
27
30
4.65
9.9
3.0
0.000197
0.00187
28
306
10
0.0619
0.0576
4
.
6
5
1
2
.
1
2
.5
formed before its release from the
0.000177
0.00181
4.65
29F
30b
20
0.0618
0.0592
11.9
2.5
nozzle. This effect is illustrated by
One-Inch Carbon Rings, Benzene Dispersed
30
10
0.00172
4.75
10.1
30
0.0572
3.0
Figure 4, in which A is plotted
30
0.00171
4.75
0.0610
31
40
10.6
2.8
against column height for both ketone
0.00187
30
4.75
0.0611
32
60
12.4
2.4
30
0.00176
4
.
7
5
0.0574
33
1
0
.
5
b
10
2.8
and benzene. It is apparent that
0.00174
4.75
0.0893
34
20
30b
10.1
3.0
0.00209
4.75
35
306
40
10.5
0.0670
2.8
40-45 per cent of the solute is ex0.00173
4.75
0.0691
30b
36
60
10.5
2.8
tracted before the drop leaves the
These runs were approximately at the flooding point.
nozzle. The data shown in Figure 4
b Using three instead of six inlet nozzles for benzene.
are for approximately constant drop
'

INDUSTRIAL AND-ENGINEERING CHEMISTRY

1148

VOL. 31, NO. 9

OF ACETIC
ACIDFROM WATERB Y METHYL
ISOBUTYL
KETONE
TABLEIV. EXTRACTION

Run
No.
1
2
3
4
5
6
7
8

9
10
11
12
13Fa
14
15
l6Fa

Flow Rate,
Cu. Ft./
(Hr.) (Sq.Ft.)

--*-

Ketone
10
in

10
60
10
60
80

10

20
30
40
50
60
70
98.5
10
30
37.3

40
40

17
18
19
20

40
40
40
40

21

40
40

22

Acid

28
40
24
40
25
40
a These runs

Concn., Lb. Mole Acetic Acid/Cu. Ft.

h

A,cid
Acid
Ketone
Ketone
in
out
in
out
Half-Inch Carbon Rings, Ketone Dispersed
0.0614
0.0618
0.0649
0.0610
0.0611
0.0606
0.0610
0.0611
0.0612
0.0611
0.0694
0.0618

....

....

Kica,

c&

20
40
60
80
100
Lo FLOW RATE OF CONTINUOUS PHASE, FT./HR.

FIGURE5. CAPACITY
CQEFFICIENTS
FOR
KETONE,
KETONE
DISPERSED,
IN HALF-INCH
CARBON
RINGS
rings and half-inch stoneware Berl saddles. The apparatus
was also operated as a spray column with no packing. For
the benzene-acetic acid-water system the distribution ratio,
CW/CB,is roughly 33, indicating that the acetic acid strongly
favors the aqueous phase. For the ketone system the corresponding ratio, CFV/CK,is about 2. For the toluenebenzoic acid-water system studied by Appel and Elgin,
C W / C Tis in the vicinity of 0.1. Tables I11 and IV give the
calculated values of both Ka and H. T. U. although the following discussion is based on the values of Ka obtained:
H. T.U. = L / K a
where Ka is calculated from concentrations in the same phase
to which L refers. Both Ka and H. T. U. are calculated from
over-all concentration differences based on the solvent phase,
since the solubility relations indicated that the major diffu-

1.05
0.55
1.50
0.62
0.69
1.87
1.15
0.87
0.78
0.69
0.60
0.59

1.66

....

0.0104
0.0061
0.0582
0.0243
0.0077
0.0624
0.0271
0.0153
0.0612
Half-Inch Berl Saddles. Ketone Disuersed
10
0.0006
0.0036
0.00077
0.0152
0,00222
0.0307
0,0144
25
0,0624
0.00569
0.0344
0.0265
40
0,0618
0,00638
0,0279
0,0405
70
0,0521
One-Inch Carbon Rings, Ketone Dispersed
0.00052
0.0151
0.0050
10
0.0820
0.00069
0.0315
0.0295
40
0,0591
Spray Column, Ketone Dispersed
40
0.0607
0.0350
0.00113
0.0265
70
0,0564
0,0405
0.00108
0.0305
90
0,0607
0.0459
0.00071
0,0344
were approximately at the flooding point.

Lb.

Packed Mole/(Hr.)
Hei&t,
Cu: Ft.) H. T . U.,
nit AC)
Ft.

1915
53.5
57.1

3:6
1.31
1.22

4.50
4.50
4.50
1.66

19.8
52.5
88.3
87.3

2.0
0.76
0.45
0.46

4.62
4.62

21.6
40.5

1.85
0.99

L c - FLOW RATE OF CONTINUOUS PHASE

FT/ HR

5.15
5.04
5.24

16.1
31.5
46.3

2.5
1.27
0.86

FIGURE6. CAPACITY
COEFFICI~NTB
FOR BENZENE,
BENZENE
DISPERSED,
IN HALF-INCH
CARBON
RINGS

sional resistance might be in the solvent rather than in the

water phase. Although Ka is expressed in terms of concentrations in the solvent phase, it is an over-all coefficient and may
be expected to vary with changes in conditions in either
phase. Logarithmic mean driving forces are employed, since
the equilibrium curve is essentially straight over the range
of concentrations encountered in any one run. The material
balances checked within 10 per cent in all but a few cases,
and within 5 per cent in all of the ketone runs except four
(runs 2, 4, 5, and 20) in which the concentration of acid in
the ketone leaving the column was not measured but was
calculated by a material balance. I n several runs the column
was on the point of flooding.
Photographs and visual observation of the operation
helped considerably in interpreting these results. In the case
of both solvents, drops were formed by the division of the
solvent stream issuing from the inlet nozzles. At low acid
and benzene rates the drops formed were not spherical but
were of roughly uniform size. The drops issuing from the top
of the packing were definitely of larger size and were less uniform in both size and shape, Some holdup was evident, and
some drop coalescence was obviously occurring in the tower.
As the flooding point was approached, a layer of drops accumulated beneath the packing, obscuring the nozzles. As
the rate of either phase was increased sufficiently, this accumulation of drops would coalesce to form a slug of solvent
which was carried out of the bottom of the tower with the
raffinate. T h a t considerable coalescence was taking place,
even before this final flooding, was apparent from a comparison of the sizes of the drops a t the bottom and top of the
column.
The tendency of the disperse phase to coalesce suggests a n
explanation of the results of Figures 5 and 6. At a constant
LD,Ka increases with increasing Lc,primarily because of the
increased holdup and a corresponding increase in interfacial
area a. This increase is approximately proportional to Lc as
long as drop size remains constant, but coalescence a t the
higher flow rates reduces Q and hence Ka. In the ketone system the drops were smaller and holdup less than with benzene. Accordingly, since the tendency to coalesce was less,
the maximum in the curve of Ka os. LCis much less apparent

INDUSTRIAL AND ENGINEERING CHEMISTRY

SEPTEMBER, 1939

-t

-E

100

U
3

60

-5

3
V

L
c

D:
I

u?
-I

G 6

20

d
I

9
m

IO

20
30
40
50
00
L,- FLOW RATE OF CONTINUOUS PHASE

70
, F T /HR.

80

n
rn

90

for ketone than for benzene. It seems evident that the

principal effect of increasing the flow rate of either phase is
to increase the holdup until flooding occurs. Ka increases
with holdup until coalescence causes an actual reduction in a.
Figures 7 and 8 illustrate the results obtained with the
three packing materials and with no packing (spray tower).
The four curves for ketone are quite similar, but for benzene
the two curves for half-inch packings are quite different from
those for the one-inch rings and for the spray-tower operation.
The ratio of tower diameter to packing size was 7 for the
smaller packings and only 3.5 for the one-inch rings. The
latter served only to increase the holdup somewhat as compared with the operation without packing. It should be
noted that although Ka is smallest for the spray tower a t a
given flow rate, the flooding rates are much greater than with
any of the three packings.
Several runs were made on half-inch rings with acid dispersed and benzene the continuous
phase. The results are compared in
Figure 9 with corresponding runs for
benzene dispersed. Ka is smaller
with acid dispersed except a t very
5
10

L
3
4

2 8

-5
-.
v

c 6

LL

U
3

FIGURE7. COMPARISON
OF CAPACITY
COEFFICIENTS
FOR
VARIOUS PACKINGS
AT LD = 40 FEETPER HOUR,KETONE
DISPERSED

10

40

I2

14

.
i

80

3
m

1149

L,-

10
20
30
40
50
60
FLOW RATE of CONTINUOUS PHASE, F T / HR.

FIGURE
8. COMPARISON
OF CAPACITY
COEFF~CIENTS FOR VARIOUSPACKINGS
AT LO =
30 FEETPER HOUR,BENZENE
DISPERSED

appearance, whereas with acid dispersed and wetting

the packing, there was little visual evidence of motion of any
kind.
Comparable data on the various packings for both benzene
and ketone are
on the H.T. u. basis in Figure
The
show the same results as the graphs of Kanamely, that extraction is
rapid with the ketone than
with the larger benzene drops, and that the saddles and small
rings give better extraction than the large rings or the spray
tower.

Conclusion
The results support the conclusions of Appel and Elgin (2)
on various points. As these authors emphasized, variations
in a appear to mask variations in K ,
and Ka for any given system is determined primarily by the interfacial
surface obtained. Ka increases roughly
in proportion to L D , coalescence being more noticeable with varying LO
than with varying LO. The drop size
in the packing, and hence the value of
Ka, depends more on the packing than
on the size of the drops entering the
packing. Coalescence a t high values
of LC tends to offset the large holdup,
and Ka may go through a maximum
as LC is increased. Flooding rates
are much higher for the spray column
than for the half-inch packings.

_J
v)

Acknowledgment

q 2
m
_J

20
BENZENE FLOW RATE, F T / H R

20

40

60

80

100

FIGURE
9. EFFECTOF INVERTING
L C - FLOW RATE OF CONTINUOUS PHASE , FT. / HR.
PHASES
(ACID RATE 30 FEETPER
HOUR)IN HALF-INCH
CARBON
RING
FIGURE
10. COMPARISON
O F DATA ON
PACKING
H. T. U. BASIS

small benzene rates, but flooding did not occur with acid dispersed in the range studied. The acid wet the carbon ring
packing, flowing down in narrow rivulets which did not appear
to be moving. Under such conditions relatively high rates of
flow of the continuous benzene phase did not entrain acid
drops to cause flooding. With benzene dispersed, the motion
of benzene drops through the packing gave the tower a

Thanks are due R. Ewing, J. E.

Demo, and A. W. Barry (4) who constructed the atmaratus. The Berl
saddles were su&&ed by the Maurice
A. Knight Company, and the methyl
isobutyl ketone by the Carbide and
Carbon Chemicals Corporation.

Nomenclature
a = interfacial surface of contact between phases, sq.
ft./cu. f t . packed volume
A = interfacial surface of contact, sq. ft.
CB = concentration of acetic acid in benzene, Ib. mole/cu.

CK

ft.

= concentration of acetic acid in ketone, lb. mole/cu.

ft.

INDUSTRIAL AND ENGINEERING CHEMISTRY

1150

concentration of benzoic acid in toluene phase, lb.

mole/cu. ft.
concentration of benzoic or acetic acid in water, Ib.
mole/cu. f t .
acid concentration of feed, lb. mole/cu. ft.
acid concentration of drops leaving
- column, lb.
mole/cu. f t .
diffusivity of solute in liquid, sq. in./sec., or sq.
cm./sec.
height of packing equivalent to 0netransfer unit =
L / K n , where K a is based on concentration in
the same phase to which L refers
extraction coefficient, Ib. mole/(hr.) (sq. ft.) (unit
AC)
extraction coefficient based on concentrations in
benzene phase, lb. mole/(hr.) (sq. ft.) (unit AC)
extraction coefficient based on concentrations in
ketone phase, lb. mole/(hr.) (sq. ft.) (unit AC)
flow rate of benzene, ketone, or water layer, cu.
ft./hr.
flow rate of continuous phase, cu. ft./hr.
flow rate of dispersed phase, cu. ft./hr.
drop radius, in. or cm.
ratio of solute in raffinate to solute which would be
extracted if raffinate came to equilibrium with
extracting liquid
logarithmic mean over-all driving force, lb. mole/cu.
ft.
time of contact, sec.

VOL. 31, NO. 9

Literature Cited

(4) Barry, A. W., S.M. thesis in chem. eng., M. I. T., 1937.

123,2430 (1923) ; International
(5) Brown and Bury, J . Chem. SOC.,
Critical Tables, Table 111, p. 404, New York, McGraw-Hill
Book Co., 1928.
(6) Demo, J. J., and Ewing, R., S.M. thesis in chem. eng., M. I. T.,
1936.
(7) Elgin, J. C., and Browning, F. M., T r a n s . Am. Znst. Chem.
Engrs., 31, 639 (1935); 32, 105 (1936).
(8) Evans, T. W., IND.
ENG.CHEM.,26, 439 (1934).
(9) Fallah, R., Hunter, T. G., and Nash, A. W., J . SOC.Chem. I n d . ,
54, 49T (1935).
(10) Harkins and McLaughlin, J. Am. Chem. SOC.,47, 1610 (1925).
(11) Hunter, T. G., and Nash, A. W., IND.ENO. CHEM.,27, 836
(1935).
(12) Hunter, T. G., and Nash, A. W., J. SOC.Chem. I n d . , 51, 285T
(1932).
(13) Ibid., 53, 95T (1934).
(14) Hunter, T. G., and Nash, A. W., World Petroleum Congr.,
London, 1933, Proc. 2, 340.
(15) Newman, Trans. Am. I n s t . Chem. Engrs., 27, 203 (1931).
(16) Rushton, J. H., IND.
ENG.CHEM.,29, 309 (1937).
(17) Saal, R. N. J., and Van Dyck, W. J. D., World Petroleum Congr.,
London, 1833, Proc. 2, 352.
(18) Sherwood, T. K., Absorption and Extraction, New York,
McGraw-Hill Book Co., 1937.
(19) Strang, L, C., Hunter, T. G., and Nash, A. W., IND.ENO.CHEM.,
29, 278 (1937).
(20) Thiele, E . W., Ibid., 27, 392 (1935).
(211 Varteressian. K. A., and Fenske. M. R.. Zbid.. 28, 928 (1936)
(22) Ibid., 28, 1353 (1936).
(23) Ibid., 29, 270 (1937).

(1) Andreas and Tucker, Sc.D. theses, M. I. T. dept. chem. eng.,

1938.
(2) Appel, F. J., and Elgin, J . C., IND.
ENG.CHEM.,29, 451 (1937).
(3) Arnold, J . Am. Chem. SOC.,52, 3937 (1930).

PR~SENTE
before
D
the meeting of the American Institute of Chemical
Engineers, Akron, Ohio. Abstracted from the doctors thesis of J. E.
Evans and the masters thesis of J. V. A. Longoor, Department of Chemical
Engineering, M. I. T., 1938.

CT =

cw =

c1
cz

=
=

D =
H. T. U. =

K =

KB

KK =
L =
Le =

Lo =
R =
A =

Unidirectional Drying of Wood

ERNEST BATEMAN,
JOHN P. HOHF,
AND

ALFRED J. STAMM
Forest Products Laboratory, Madison, Wis.

HE drying of wood is a complicated phenomenon which

has thus far defied rigorous theoretical analysis. Most
of the evidence indicates that it is a t least in part a diffusion phenomenon. Even this might be questioned, however,
on the basis of the recent findings of Ceaglske and Hougen ( I )
that the drying of granular nonhygroscopic solids is controlled
entirely by capillary forces rather than by diffusion. Tuttle
(8),Sherwood (6),and Kollmann (4) showed that the moisture gradients obtained in drying wood under definite boundary conditions can be theoretically reproduced by Fourier
analysis methods (3) with a fair degree of accuracy by assuming that the phenomenon is one of simple diffusion over
the complete moisture-content range. Hawley ( 2 ) ,however,
pointed out that diffusion would not be expected to take place
above the fiber saturation point on the basis that the fiber
saturation point is the moisture content in equilibrium with
unit relative vapor pressure. Further, in simple diffusion,
the diffusion constant and the diffusivity in the Fourier form
of the equation (3) should be independent of the moisture
content. This is not the case for transverse drying of wood
according to the moisture transfusion measurements of Martley (5) in which the equilibrium moisture gradients set up
under steady-state drying conditions were determined.

Measurements were made of the rate of

drying from a single face of small cylinders
of Sitka spruce at different temperatures
and under different relative humidity and
atmospheric pressure conditions. Moisture
gradients were determined on the specimens prior to the complete removal of free
water. Drying in all cases gave weight
losses that varied directly with the square
root of the time. Values for the mean effective diffusion per unit moisture gradient
were calculated from the rate of drying and
the moisture gradients up to the fiber saturation point. The values increase slightly
with an increase in the relative humidity
effective in the drying, and increase to a
greater extent with an increase in the drying temperature, a decrease in the atmospheric pressure, and a decrease in the specific gravity of the wood.
These complications undoubtedly arise from the complex
nature of the capillary structure of wood ( 2 ) . Water is held
with an appreciable reduction in vapor pressure within the
cell walls of wood as surface-bound and capillary-held water
(7) and within the microscopically visible capillary structure
with only a small reduction in vapor pressure. The fiber