1460 I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY Vol. 43, No.

6
encountered in continuous operation. It might be well to point
out, however, that if somewhat higher conductivities are to be
measured, a frequency of 60 cycles second can be applied.
This modification leads to a considera le reduction in equipment,
since the alternating current power line can then be used directly
instead of the oscillator. Increased sensitivity to changes of
specific resistance a t lower levels can be obtained by suitably
shunting t,he voltmeter, thereby reducing the value of r in Equa-
tion 9.
impurity is present, by xnea,ns of a measurement of specific
resistance alone.
One additional fact should be noted concerning the calibration
points shown in Figure 3. Although a comparison of one set
of samples (different sets are indicated by different symbols in
the figure) with another wa.s in error as much as 1.50/, a t high
purities, the accuracy of the calibration for any one set was more
consistent. A change in purity was always accompanied by a
change of meter reading in the correct direction. Thus changes
in purity were indicated to an accuracy much better than 1 . 5 ~ o ,
and the indicated purity always varied as it should with changes
in still temperature and reflux ra.te.

98t
96-
It) is of interest tha,t. the conductance of methyl vinyl ether
measured under flow conditions was consistently lower than that
of an equivalent immobile sample. A t first this effect was as-
a cribed t o a cleansing of the cell by the flowing liquid, but the
Y findings could also be explained in wcordance with Plumley’s
Li 94-
F! theory ( 5 ) . If favorable orientation in the electric field counter-
52 92-
acts recombination of the ions formed by spontaneous dissociation
c
8
one may expect a lawer conductivity in a flowing medium since
Y the ions are being carried away by the moving liquid.
2 90-
ACK8OW LEDGMENT ,
88 - \
\ The authors would like to thank H. Beller, new products
A. \ mansger of the General Aniline Works Division and his associ-
ates a t the Graseelli plant, for their wholehearted cooperation ;
0 50 I00 150 200 250 3, they are also grateful to R. M. Fuoss who made several valuable
RECORDER DEFLECTION (ARBITRARY SCALE1
suggest,ions on reading the manuscript. The chemical analyses
Figure 3. Recorder Scale Calibration of the methyl vinyl ether eamplrs vere performed by S. Siggis
and R . 1,. Edsberg.
In the h a l installation provisions were made for continuous LITERATURE CITED
recording by modifying the output of the electronic voltmeter (1) Fuoss, R. M., J . Am. Chem. SOC.,55, 21 (1933).
so t h a t it actuated a conventional temperature recorder. Figure (2) Jones, G., and Josephs, R. C., Ibid., 50, 1049 (1928).
(3) Onsager, L., J . Chem. Phys., 2, 599 (1934).
3 shows the correlation between the readings obtained on the (4) Pao, C. S., Phys. Rel;.,64, 60 (1943).
recorder and the values for methyl vinyl ether content deter- (5) Plumley, H. J., Ibid.,59, 200 (1941).
mined by chemical analyses of identical samples. The methods (6) Rosenthal, R., Instruments, 23, 664 (1950).
described by Siggia (8) and Siggia and Edsberg (9) were used for (7) Schildknecht, C. E., Zoss, A. O., and McKinley, C . , IND.ENO.
CHEM.,39, 180 (1947).
these determinations. As shown in Figure 3, in most cases the ( 8 ) Siggia, S., Anal. Chem., 19, 1025 (1947).
correlation was found to be quite good, Of course, since the (9) Siggia, S., and Edsberg, R. L., Anal. C h m . , 20,762 (1948).
same effect of decreasing the resistance may be obtained from RECEIVEDOatober 23, 1950. Presented before the Gordon Research Con-
many different kinds of impurity, it is impossible t o tell which ference on Instrumentation (A.A.A.S.), New London, N. H. (Auguet 1950)

Enginnyring ifisance of Liqui

Process
development nts in Gas Absor
I

P. V. DANCKWERTS
UNIVERSITY D E P A R T M E N T OF C H E M I C A L E N G I N E E R I N G , T E N N I S C O U R T R O A D , C A M B R I D G E , E N G L A N D

D
T HE conventional picture of the process of absorption of a gas
into a n agitated liquid, under conditions such that (‘the liquid
film is controlling,” is t h a t there exists a stagnant “film” of liquid
R = - (c*
XL
- GO) (1)

at the interface, similar to the laminar sublayer formed when a
fluid flows past a solid. The concentration of the bulk of the
liquid beneath this film is kept uniform by turbulent mixing, while
the surface concentration has a t all times the saturated value c“.
The rate of absorption (per unit area) will then be given by the
expression
where R is the rate of absorption per unit area, c, is the bulk con-
centration of the solute, D is its diffusivity, and X L i? “the effec-
tive film thickness” (9).
It seems doubtful, however, whether the conventional picture
bearP a very close relationship to the actual mechanism of absorp-
tion. The conditions reauirrd to maintain a stagnant film at the
 June 1951 INDUSTRIAL AND ENGINEERING CHEMISTRY 1461

free surface of an agitated liquid appear to be lacking, and it tested by experiment. It may well be found that the processes
seems more probable that turbulence extends to the surface and occurring in gas-absorption equipment are too complicated to be
that there is no laminar boundary layer. I n particular, if we treated successfully by such simple methods as those presented
consider the liquid flowing over the packing in an absorption here, but the author hopes that these may constitute a step in the
tower, it appears most unlikely that the surface layer of liquid right direction.
maintains its identity throughout all the discontinuities of flow.
ABSORPTION INTO A STAGNANT LIQUID
The fictitious nature of the “liquid film” is probably widely
suspected; nevertheless, it is constantly referred to as though it Consideration is limited in the first place to the case where
actually existed. This may be regarded, for many purposes, as a the surface of the liquid is at all timea saturated with the solute
harmless and convenient usage, as measured absorption rates gas, and there is no chemical reaction between solute and solvent.
appear to conform to the expression If the normal diffusion law is obeyed in the liquid, the rate of
* absorption (quantity per unit area per unit time) into a stagnant
R = ~ L ( c *- eo) (2) liquid of infinite depth is given by
where k ~the , liquid-film mass-transfer coefficient, is constant for a
pjven liquid and gas under given conditions. However, if the
$(e) = ( C * - e o ) (3)
U film model is in fact an un-
where 6’ is the time of ex-
realistic one, it may lead t o
T h e present theory of gas absorption is based on a pic- posure to the gas and the
erroneous results if it is used
as the basis for theories which ture of the absorption mechanism which is probably un- initial concentration is uni-
seek to relate k~ to the con- realistic. For many practical purposes this is nnimpor- formly equal to co (see dis-
ditions of operation. tant, but the theory is likely to be misleading when used cussion of derivations of R
It may be of value, there- to predict the behavior of systems for which no experi- below).
fore, to show t h a t the proved mental data are available. A more plausible picture of the I t can be seen that unless
usefulness of Equation 2 in absorption process is therefore suggested. the surface is renewed by
no way provides support for The usual assumption of a “stagnant film” of liquid at stirring or by convection, the
the stagnant-film hypothesis, the interface is abandoned. Instead it is supposed that rate of absorption becomes
as this expression can be the surface is continually being replaced with fresh liquid, v e r y slow a f t e r a time.
derived from what is con- and a theory of absorption rates is developed on this basis. Stagnant water will normally
sidered t o be a more realistic It appears to agree generally with experience, but it must absorb in 1 hour a quantity
picture of the processes occur- be regarded as tentative until confirmed experimentally. of gas equivalent to a satu-
ring during absorption into
an a g i t a t e d l i q u i d . T h e
treatmeDt is here extended to
If the detailed features of the theory are confirmed, it
offers a powerful method of predicting the performance of
absorption equipment. Failing this, it may still provide
(taking D -
rated layer about 2 mm. thick
sq. em. pcr
second). If the “depth of
cases in which gas-film resist- suggestions for a new method of approach to the subject. penetration” is arbitrarily de-
ance or surface resistance fined as the depth a t which the
plays a part. Provided no rise in concentration is l/loa
chemical reactions occur, the expressions derived for the rate of ab- that a t the surface, it will be equal to 3.6 d E (see Equation
sorption are similar in form to those based on the film theory, but 27), with a value of about 6 mm. at 1 hour. (In the case, for
a different significance is attached t o the liquid-film coefficient. instance, of carbon dioxide and water, $ will conform to Equation
On the other hand, when the absorbed gas undergoes a reaction in 3 only for small values of 8; thereafter convection currents will
solution, the two methods of treatment give different expressions arise because the density of a solution of carbon dioxide is greater
for the relationship of the rate of absorption to such quantities as than that of water, and the predicted values of $ will be exceeded.)
the liquid-film coefficient without reaction, the reaction-velocity Under certain conditions Equation 3 can be applied as a close
constant, and the diffusivities of the components of the system.’ approximation to: (a)liquid layers of restricted depth, and ( b )
The expressions derived in this paper for the rate of a b s o r p liquid moving parallel to the surface with a velocity that varies
tion, R, under various conditions contain a quantity, 8 , which with the depth. The necessary condition for ( a ) is that the time
. relates to the rate of renewal of liquid surface. A complete de-
velopment of the subject would require further discussion of the
of exposure should be so short that the depth of penetration is
less than the depth of the liquid; for ( b ) it must be so short that
relation of s to the hydrodynamics and geometry of the system. the depth of penetration is less than the depth at which the veloc-
For the present s is regarded as a quantity which must be deter- ity is appreciably different from that a t the surface.
mined experimentally for any given system. The various expres-
* sions for R in terms of s are thus likely to be of use mainly in com- ABSORPTION INTO SURFACE O F TURBULENT LIQUID
paring rates of absorption in systems of equal 8. For instance, When a liquid is in turbulent motion it is a mass of eddies
suppose the rate of absorption of carbon dioxide into water is which incessantly change their conformation and position.
measured in a packed tower. With the same packing and flow These eddies are pictured as continually exposing fresh surfaces to
rate, and with a liquid of virtually the same density and vis- the gas, while sweeping away and mixing into the bulk parts of
cosity, s will have the same value, and under these conditions it the surface which have been in contact with the gas for varying
may prove possible to use the measurements on carbon dioxide lengths of time. The assumption will be made that during the
and water to predict the rate of absorption of, say, sulfur dioxide time of exposure of any portion of the liquid it absorbs gas a t a
into water or carbon dioxide into dilute alkali. (The possibility rate given by Equation 3, as in case ( b ) above. This is equiva-
of doing so depends in part on factors discussed under “Nonhomo- lent to assuming that the “scale of turbulence” is much greater
geneous Distribution of Surface Ages.”) than the “depth of penetration” of the solute diffusing from the
No attempt has been made to compare the expressions derived surface, so that relative motion of the liquid a t different levels
here with published experimental measurements. I n spite of close beneath the surface may be disregarded. Velocity gradi-
the enormous number of these, very few are suitable for an analy- ents no doubt exist beneath the surface, but while an increase in
sis of this sort, and in any case a knowledge of some of the essen- intensity of turbulence will accentuate these, it will a t the same
tial physical constants is at present lacking (particularly of ks,the time decrease the depth of penetration by shortening the period
“surface resistance” coefficient). for which any part of the liquid is exposed to the gas before being
The theory must be regarded as speculative until it has been submerged once more.
1462 I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY Vol. 43, No. 6
Consider a liquid which is maintained in turbulent motion by Equation 6 t h a t for a series of solutes 1 / k ~should be linear in
stirring a t a steady rate. The total area of the surface which is l/dB.
exposed to the gas will be taken as equal t o unit and the average
rate of absorption is uniform over the area. &e motion of the Further complications‘will be introduced if the solute undergocs
liquid will continually replace with fresh surface those parts
which are older, in the sense that they have been exposed for a chemical reaction in the liquid, or if there is a gas-film resistance
finite length of time, The mean rate of production of fresh sur- or surface resistance. These are considered in connection with
face will be constant and equal to s, and the chance of a n element absorption in packed columns, and the expressions obtainrd are in
of surface being replaced within a given time is assumed to be in- general applicable to stirred liquids also.
dependent of its age; hence the fractional rate of replacement of
the elements belonging to any age group is equal to s.
Let the area of surface comprising those elements having ages
between 0 and (0 +
de) be +(B)d6. At steady state this does not
vary with time. Hence in a short time interval equal to de the
PACKED ABSORPTION TOWERS

Observation of the behavior of filaments of colored liquid sug-

.
area entering the age group 0 . . , (8 4- de) from the age group
gests that the liquid running over the packing in a tower is i n
(e - de) . . , 8 will be +(B)dB. This is also equal to the area in age
group (0 - . .
de) , . e less the portion of this which is replaced by some places flowing in a thin laminar film, while a t others it is in
fresh surface in a time equal to de-i.e., turbulent motion caused by discontinuities in the wetted surface.
Most of the exposed surface appears to belong to liquid in laminar
+@)dB =+(e - de) dO(1 - sd8) rather than turbulent flow-at least provided the rate of flow is
Therefore de) = +(0) - 2 de - s+(e)de
not too great.
K e consider first the case in which virtually all absorption it.
assumed to take place into the surface of t’hose areas in which t,he
a n cl liquid is in laminar flow. The system is assumed to be a t steady
state.
Since also
Consider a thin horizontal section of the packing. The section
is supposed to have a volume great enough to offer a representa-
+dB = 1 tive sample of flow conditions, but its height must be sufficiently
J;;“ small t’o allo~vthe bulk concentration of the liquid to be regarded
as uniform. Where a stream of liquid in laminar flow encounters
we have a discontinuity in the packing, tarbulent conditions prevail and
the liquid is assumed to be thoroughly mixed before passing on to
9 = se-sff (4) another laminar region. Fresh liquid surface, with a concentra-
For areas other than unity, +(@)deis equal to the fractiori of the tion equal (subject to local variations which will be ignored) to
area which at steady state is in the age group e . . , . (e de).
,$(e) is referred to here as the surface-age distribution function.
+, the mean concentration of the liquid stream, is thus being gen-
erated at. the top of these laminar regions and destroyed a t the
The rate of absorption into those elements of surface having same rate a t the points of discontinuiby. The time taken by the
age e and combined-area s e - d d e is found from Equation 3 to be surface of the liquid to travel from the preceding point of discon-
tinuit’y to some fixed point on the wetted surface of the p‘acking is
termed the “age” of the surface at this point.
(c” - c o ) s e - d
l/.s
de. Hence the mean rate of absorption
per unit area of turbulent surface is
It’is assumed t h a t the local l a t e of absorption of gas is the samc
as for a stagnant liquid of infinite depth which has been exposetl
to the gas for a time equal to the local age of the surface. A coni-
plete analysis of the assumption cannot he given here, but thc
following considerations show t’hat it probably represents a
reasonable approximation. A solution has been obtained (by
graphical methods) for the rate of absorption of gas into a
= (c* - c,) dns (5) laminar layer of liquid flowing under the influence of gravity down
a vertical wall. It is assumed that the parabolic velocity profile is
Comparing this with Equation 2 we see that 4%can be fully developed, and that the surface is constantly saturated.
identified with k ~ . The ratio of the amount of gas absorbed by an element. of moving
iurface within a time e of its formation to the amount absorbed in
It has been tacitly assumed in the above calculation that time 0 by a,n equal area of surface of stagnant, infinitely deep
immediately underlying all freshly formed surface is liquid with liquid depends only on the quantity L = - 1DM where d is the
the same concentration, ca, of dissolved gas. I n identifying 4 /09 ypd4
with kL, co is identified with the mean bulk concentration of the thickness of t,he liquid layer and I is the distance which the ele-
ment of surface moves in time e. The difference between the
liquid; this implies t h a t the mean concentration does not change quantities absorbed in the two cases amounts to less than 575,
with time, and that the bulk of the liquid is of uniform concentra- provided L < 0.1. To give an idea of the significance of this con-
tion. The first condition is automatically fulfilled in a continu- dition, d will be given the rather low value of 0.01 em., with I = 5
ous process operating a t steady state. The extent t o vihich the em., and D = 10-6 sq. em. per second. L is then about, 0.05 (for
water).
second is fulfilled will depend on the depth and flow chnracteris-
tics of the liquid. Although this analysis ignores the fact that the parabolic veloc-
ity distribution is established only after a considerable distance
Consider a case in which a t steady state the mean concentra-
tion a t a level A , close to the surface, is C A , while t h a t a t a greater of travel, and that the surfaces of the packing will not generally be
depth, B , is maintained a t the bulkor discharge concentration, co vertical, it does indicate t h a t the assumption is likely to he justi-
(the depth of A being comparable with the scale of turbulence). fied under conditions of practical importance. While the dis-
Solute is carried from A to B by eddy diffusion, and foi steady crepancy increases with the age of the surface, absorption into
state
L L ~ l d e surfaces
r” is relat,ively unimportant in determining the
(CX - ca) 4%= (CA - c,)ka = (c* - Co)kL mean rate of absorption.
The packing is supposed to be arranged in a completely random
or
manner, so t h a t there is no correlation between the age of any
element of moving surface and the probability of its being de-
stroyed within a given time (the validity of this assumption is
discussed under “hTonhomogeneous Distribution of Surface
where k . is~ the coefficient of mass transfer from A to B by eddy Ages”). Under these circumstances the age-distribution func-
diffusion. s and k~ will be functions only of the physical proper-
ties and flow conditions of the liquid and the geometry of the tion +(e) is easily shown to be the same as t h a t already found for
vessel. Hence, provided these remain constant, we see from the stirred liquid (Equation 4)-namely, se -so. The surface-
 June 1951 INDUSTRIAL AND ENGINEERING CHEMISTRY 1463

renewal factor, s, will in this case depend on the physical proper-
ties and rate of flow of the liquid and the geometry and size of the
packing. If the rate of absorption into a stagnant liquid, nThich
has been exposed t o the gas for a time e, is denoted by $(e), the
mean rate of absorption per unit area of wetted surface in the
section of packing will be
(7)
values of k ~ but
, this requires confirmation. Further investiga-
tion may show the phenomenon to be one of considerable impor-
tance in determining rates of gas absorption.
Until more information is available, it is not possible to say
whether ICs is a function of the age of the surface, although this
seems unlikely. Assuming it to be constant, we have by analogy
with Equation 9
e* - eo
R =
1 1

The function $ will take different forms according t o the be-

7E+;
a
havior of the liquid and gas which are under consideration. I n if there is no gas-film resistance, and
principle it may be found by solving the appropriate diffusion
equations with the required boundary conditions (see discussion
of derivation of expressions for R). However, the necessity for
* this can usually be avoided, and the mathematical treatment
simplified, by making use of the f a c t that the expression for R/s,
as can be seen from Equation 7, is the Laplace transform of $, when there is gas-film resistance of appreciable magnitude.
The transform is considerably easier t o find than $ itself. SOLUTE DESTROYED BY FIRSFORDERREACTION WITH SOLVENT.
Mean Absorption Rates for Various Types of System. The
derivation of the expressions given here is discussed a t the end
of this paper.
SVRFACE CONSTANTLY SATURATED, NO CHEMICAL REACTION where T is the first-order reaction velocity constant. This may
IN SOLUTION.
be put in the form
R = (c* - c,) d& (8)
As in the analogous case of absorption into a turbulent liquid
(when k E is large), fican be identified with kL. If the solvent
if fl is identified with k ~ the , hypothetical liquid-film coeffi-
and the flow conditions remain unchanged, k~ should be propor-
cient for t h e absorption of the same gas in the same solvent ~ i t h -
tional to 4 0 for a series of solutes, provided they behave toward
out reaction. The value of s might be determined from measure-
the solvent in the manner postulated.
ments of k~ for a gas which did not react with the solvent.
GAS-FIU RESISTANCE, No CHEMICAL REACTION.
If there is a gas-film resistance t o absorption, Equation 12 he-
comes
(9)
c* - co
R = 1 (14)
In deriving this expression it has been assumed t h a t the system H
obeys Henry’s law and t h a t the gas-film mass-transfer coefficient, r;, + d D m )
kc, has a constant value independent of the local rate of absorp- H / k c should be replaced by l / k s if only surface resistance is
tion into the liquid. The latter is almost certainly an over-
simplification, but the approximation is probably no more drastic
important, and by ( l / k s +
H / k c ) if both are of appreciable
magnitude.
than t h a t involved in the normal treatment of gas-film coefficients, The term involving co may be omitted when conditions are
and the result should give at least a general indication of the be- such t h a t the bulk concentration of unreacted absorbed gas is
havior to be expected from this type of system. It is interesting negligibly small. This will be the case if a t all points in the
t o find t h a t Equation 9 is similar in form to t h a t derived by a absorption tower AR<< Vrc*, where A is area of wetted surface
4
very different path from the simple two-film theory. It is and V is volume of liquid, per unit volume of packing,
shown in the discussion of derivation of expressions for R t h a t Equation 13 may be compared with the expressions derived
the conventional rule for the addition of “resistances in series” from the conventional film theory (11).
can be applied t o all the types of systems here considered. INSTANTANEOUS REACTION BETWEEN DISSOLVED GAS A N D
SURFACE RESISTANCE,No CHEXICALREACTION.Even when REAGENT IN SOLUTION. When there is n o surface or gas-film
s
there is no effective gas-film resistance t o absorption-for in- resistance
stance, when a pure gas is absorbed-there is some evidence that
the gas and the liquid may not attain equilibrium a t the interface
immediately on being brought into contact (4, 7 ) . Such a
phenomenon may be explained by assuming that only a very
small fraction of the gas molecules striking the surface enters the where p is defined by
liquid, while the remainder is reflected (this hypothesis has been
discussed more fully elsewhere, 4). An alternative explanation
might be that there is a region in the immediate neighborhood of
the liquid surface in which surface forces restrict the mobility of c,’ being the bulk concentra- of the reagent in the solution and
the solute molecules, so t h a t their diffusivity is much lower than D’ its diffusivity. c* and c, are expressed in chemical equivalents
the bulk of the liquid. In either case, the rate of absorption a t per unit volume. Graphs which will help in the solution of
any time will be given by a n expression of the form (c* - c s ) k s , Equation 16 have been published ( 6 ) .
where cs (the concentration at the surface or immediately be- If the diffusivities of the dissolved gas and of the reagent are
neath it) is a function of the time of exposure. Only in the case equal (which will often be approximately true) we find from
of carbon dioxide and water is there any experimental evidence as Equations 15 and 16
to the value of k s ( 7 ) . The apparent value (ea. 0.05 em. per
second) is in this case of the same order of magnitude as normal R = (c* + e:) dFs (17)
 1464 INDUSTRIAL AND ENGINEERING CHEMISTRY
which is similar to the expression derived (with D = D’)from
the conventional hypothesis (IO).
If a surface or gas-film resistance of appreciable magnitude,
1lk,is present Equation 15 becomes:
SECOND-ORDER REACTION BETWEEN DIMOLVED GAS A N D
REAGENTIN SOLUTION. The exact solution of this problem
offers considerable mathematical difficulties. Under certain
conditions, however, Equation 12, 13, or 14 may be used as an
approximation if r is replaced by r‘c;, where c i is the bulk concen-
tration of the reagent and r’ is the second-order velocity con-
stant-k., the rate of reaction a t any point in the liquid is r’w’,
where c’ is the local concentration of the reagent. The error ui-
volved depends on the ratios S / T ’ C ; and cL/c*. We are interested
mainly in cases where the former is of the order of magnitude of
unity-Le., diffusion and chemical reaction both play appreriable
parts in controlling the rate of absorption. Putting s/r‘ci = 1
for purpose of illustration, we find t h a t the error will be less than
10% when c,’/c* is equal to 50. (In other words, the concentJa-
tion of the reagent in the absorbent solution must be about 50
times as great as the concentration of a saturated solution of the
gas in the pure solvent.) The accuracy of the appioximation
will be lessened if either of the above-mentioned ratios is dimin-
ished. (The bulk concentration of unreacted absorbed gas IS
assumed to be much less than cc.)
NONHOMOGENEOUS DISTRIBUTION OF SURFACE AGES
The expression previously derived (Equation 7) in terms of \I.
for the mean rate of absorption in a section of a packed columri
was based on the assumption that the probability of an element of
moving surface being destroyed by remixing waR independent. of
its age. However, there are two types of system in which this
condition will not be fulfilled.
Nonrandom Packing. Any regularity in the arrnngement,
of the packing is likely to affect the age-distribution function,
$(e). For instance, if the packing consists of vertically stacked
rings, virtually none of the surface will reach an age greater than
themtime,Be, t.aken for the surface of the liquid to move from the
topato the bottom of a ring, and there is no chance of destruction
before thie age has been reached. If we ignore the acceleration
of the surface during the descent from one discontinuity t.0 an-
other, we have approximately
+(e) = I/&, 0 < e < ec
+(e) = 0, e > e,
In a simple case such as this it is, of course, easy if rC. is known
to determine R from the expression
n = Jo’” +(e).rC.(e).de (19)
which is of general applicability.
I n a dumped pac,king, even of elements of uniform size, may be
expected t o approximate much more closely to the form of Equs-
tion 4, since the linear rate of flow of the liquid will vary widely
with the inclination of the surface, and there will be considerable
variation in the distance of travel between discontinuities. It
may be objected t h a t there is bound to be a finite upper limit to
the age of any element of surface. However, this is not lilrely t.0
lead t o serious error in itself, provided the distribution of the
“younger” parts of the surface does not depart too widely from
t h a t given by Equation 4. This is because most of the absorp-
tion takes place into surface which has been exposed for only a
short time, and the upper limit of the integral in Equation 7 can
be given a finite value without greatly altering the value of the
Vol. 43, No. 6
integral. For instance, in the simplest type of system, in which
there is n o surface or gas-film resistance or chemical reaction, the
integral is (cf. Equation 5):
If the upper limit of this integral is changed to ec, it becomes on
rearrangement d&erj aC, which is within 10% of 1 / 5 8 if
soc > 1.5. Tentative values of s, calculated from published
values of k~ for carbon diovide arid water in packed towers (8)
are of the order of magnitude 5 see.-’, indicating that less than
one tenth of the observed absorption rate is due to surfaces older
than 0.33 second.
Nonuniform Value of s. If an appreciable fraction of the
absorption in a pachcd column takes place into the surfacr of
turbulrnt liquid a t the pointa of discontinuity, the form of
ill again be modificd We may picture the effective area of
surfacc a8 being made up of a number of areas, a, each having a
burface-age distribution similar to Equation 4 but with its own
value, sa, for 8 . This leads to the exprefision
If for either oi the reasons discussed above, thc function
expressing the distribution of surface ages in an absorption syp-
tem differs from se-@, and if its actual form isundetermined, sys-
tematic investigations or attempts to predict absorption ratrs in
various types of system may be rendered difficult.
I n the simplest case, in which there is no chemical reaction or
gas-film or surface resistance, we find from Equations 8 and 19
for nonrandom packing, and from Equations 8 and 20
for nonuniform values of s. Similarly, the analogs of Equation 15
are
Equations 21 to 24 resemble Equations 8 and 15 in that they
express R as the product of a term containing known (or inde-
pendently measurable) physical constants, and another which de-
pends only on flow conditions, etc., and which must be deter-
mined empirically. Thus, provided the latter term remains con-
stant, there should be a simple and predictable relationship be-
tween the value of R and the values of the physical constants in
the first term.
However, if there is any gas-film or surface resistance or if a
first- or second-order reaction takes place, we find that the reyultr
ing expressions no longer have the relative simplicity which marks
the equations referring to systems in which $ = 8e-&. For in-
stance, Equation 9 (simple gas-film resistance) becomes
when $ is undetermined (cf. Equation 29), and
 June 1951 INDUSTRIAL AND ENGINEERING CHEMISTRY 1465

when the value of s is not homogeneous. Thus even a system 3. First-Order Reaction between Dissolved Gas and Sol-
with constant flow conditions, and for which k a is known, will vent.
show no simple relationship between R and D . All t h a t can be c = ca,x > 0, e = o
said ie t h a t R/(c* - e,) will be a unique function of D for a series C,L)&+r”
ae ax
c = C * , X = 0, e> 0
of solutes if H / k c is kept constant, and t h a t it will be a unique c < c ~ , x =~ . , e > o
function of H/SGfor a series of systems in which D is constant If conditions are such t h a t the concentration of unreacted ab-
(liquid-flow conditions assumed constant throughout). Such sorbed gas in the bulk of the liquid is negligibly small (c, = 0),
considerations indicate a basis for the correlation of experimental the solution of the above is, as has been shown elsewhere (6),
results, but are not of great practical valut-. Similar conclusions
may be reached regarding systems of the types referred to in
Equations 10 t o 14.
whence
dERlVATION OF EXPRESSlONS FOR R IN VARIOUS TYPES OF
SYSTEM

Two methods may be used. One is to solve the appropriate This result could have been reached much more directly by the
partial differential equation to find +(e), the rate of absorption per
unit area of surface into a stagnant liquid which has been exposed application t o the eneral case where eo 0. *
Laplace transform method, which will be illustrated by ita
Multiplying eacx term in the differential equation by e -80,
to the gas for time e. $(e) may then be multiplied by se-sO and integrating with respect t o e between 0 and 03, and making use of
integrated. as already explained, t o give R. This method is the first boundary condition, TVC find ( 3 )
used in cases 1, 2, and 4 below. Alternatively, the Laplace
dzE
transform may be applied t o the diffusion equation and the re- sC - co = D-
dx2
- rC
sulting ‘*subsidiary equation” solved for i ( ~s), , the Laplace
transform of c ( r , e) ( 3 ) . R is equal to - D s ( d z / d x ) , = 0. The while the second and third boundary conditions become
latter method, which is illustrated in cases 3 and 6 below, is
generally considerably simpler than the former, but is applicable
only when the‘ diffusion equation is linear.
A general rule for the “addition of resistances” is derived be-
low, so t h a t it is not necessary to consider the effect of an added
gas-film or surface resistance for each type of system separately.
where C, the Laplace transform of c, is defined by
1. No Reaction. Surface Constantly Saturated.
C,= c,,, x >0, e= o
C = C * , X = O , B > O
c = c ~ , x =0 0 , e > o The solution of the above is
The solution t o t,his is (1) -8
c o t p * \ s )- ~._le . -
c = co + (e* - c,)er.fc [A]
2de0 e =
(r +4 (33)
whence Now $(e), the rate of absorption at time e, is equal t o
- D (z),
bC
= whence it can be seen from the definition of C that

R = sJ.;” e-%(e)de = -sD J m e--s8 (2) x = o dB

= (c* - co) d5s (28)

2. Surface or Gas-Film Resistance. This system illustrates
the rule for the )addition of resistances given in Case 6. so that R is immediately found from Equation 33:
,The differential equation is the same as in case 1, but the
boundary conditions become
(34)
L
C=G, X>O, e=o
4. Instantaneous Reaction between Absorbed Gas and
-D
(3
- = k(c* - c), x = 0, e >0 Reagent in Solution. The method of solution for t h e stagnant
liquid has been given (6). There is a steadily deepening zone be-
neath the surface which contains absorbed gas but no reagent.
c=c., x = m , e > O This is bounded by a plane a t which the concentration of both is
zero, and beneath this the liquid contains reagent alone. The
The second condition states t h a t the rate of absorption is a t all absorbed gas and the reagent obey the normal diffusion equations,
times proportional to the difference between the surface concen- but the boundary conditions are somewhat complicated. The
tration and the saturated concentration. k stands for k s or solution is
k G / H . The solution is known (.e) and may be put in the form
-
$(e) = k(c* - c,) = k(c* -
eo)ek20/Qwfc Lk J;]
(29)
(35)

whence from which Equation 15 follows.

5. Second-Order Reaction between Absorbed Gas and
Reagent in Solution. Attention will be confined t o the case in
which the concentration of unreacted absorbed gas in the bulk
of the liquid is negligibly small compared to c*.
Putting c(x, e) = concentration of unreacted absorbed gas,
, = concentration of unreacted reagent (both in chemical
~ ’ ( 28)
1466 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 43, No. 6

equivalents per unit volume), and T ’ = the second-order reaction where k is the appropriate mass-transfer coefficient. This trans-
velocity constant, we have for a stagnant liquid forms to
c = C*,X = o,e > o
c = 0, x > 0, e = o
c=O,x= m,8>0
C’ = e;, x > 0,e = o The new solution for Cis
c’=c,I,x= m,e>O
dc’
-=o,x=o,~>o
dX

These equations are nonlinear and no solution has so far been

found. However, Equations 31 and 32 can be used as approxi-
mations, under certain conditions, if r‘c: is substituted for r. Let
us call the expressions for R and $(e) obtained in this way R1 and
$ ~ ( e ) ,while the exact solutions of the above equations would lead whence
to other expressions, Rz and &(e). Now it can be shown (6) that
$2 3 $1 if r‘c*e < < 1. If r‘c*e is taken to be O.O5-i.e., e =
O.OS/r’c*-the difference between and is less than 5%, hence (39)
the approximation
0.05/r ’e* O.O5/r ’ c *
.JI;
has an error less than 5%.
e-s%(e)de = s
JL e-ao$,(B)dB

If in addition it is stipulated that

In other words, if R has the form

R = (c* - A ) B .\/B
( A and B being constants) when the surface is constantly satu-

L0*05/T’c’ Jd” e-+l(e)de > 0.95s

the following approximation can be used with an error of less than
e-ao$l(e)de (36) rated, then for the case where there is a resistance l / k at t h e
surface

R =
e* -A
10%: 1
G5’i
R1= sJdm e-%(e)de

But from Equations 31 and 32

=s
Jd” e-ao$,(e)de = Rz
I n cases such as Equation 15 this rule is ambiguous, and
recourse must be had to the equation for E to determine the effect
of a n added resistance.
r0.05/r’c* These arguments are easily extended to the case where there
Jo f w
e--&$.l(e)de
= eyf
,-
0.05(r’cd
T’p*
+ s) - are two resistances a t the surface.

JU
The value of the ratio is seen to depend both on c:/c* and on
P’c:/s, but because interest centers on systems in which s and
T’c.) are of comparable magnitude-i.e., diffusion and reaction
play roughly equal parts in determining the absorption rate-
we may put r‘c,/s = 1 for purposes of illustration. It is then
found that c : / c * > 50 if Equation 36 is to be complied with.
6 . Rule for Addition of Resistances. I n all types of systems
considered here the equation for C(x,s),the Laplace transform of
the concentration a t x, has the following form:
dzE
_ - aE b probably be retained as a matter of convenience. As regards the
dx2 - D - 5 (37)
where a and b are constants. I n the case where there is no surface
or gas-film resistance, so that the surface of the liquid is con-
stantly saturated with the gas, the boundary condition a t the
surface becomes
whence
CONCLUSION
The considerations advanced in this paper raise two distinct
questions, which can be answered only by experiment.
1. Is the surface-renewal mechanism here postulated closer t o
the truth than the conventional picture of an undisturbed layer
a t the surface of the liquid?
2. If so, does the distribution of surface ages in absorption
equipment of practical interest approximate to Equation 41
It should be possible to shed some light on the hrst by qualita-
tive observations on stirred and flowing liquids, without reference
to absorption. If the answer should prove to be “yes,” then the
conventional picture is misleading, and should not be used as the
basis for theories of absorption although its terminology would
second, careful measurements of rates of absorption, and of the
relevant physical quantities, such as k s , would be required to
determine thP answer; it mould be convenient if this should
prove to be “yes,” because a precise mathematical treatment
could then be employed, as described above. Should the answer
to the second question prove to be “no,” the predictions of the
theory lose a great deal of their precision and utility, but this in
itself would not be a valid reason for preferring the “stagnant
film” hypothesis as a basis for further theoretical developments.

IVOM ENC LATURE

A = area of wetted surface per unit volume of packing
and n = area of those parts of absorbing surface having value sa for s
c = concentration of absorbed gas in liquid a t x, 8
c* = saturated concentration of gas in liquid
co = initial or bulk concentration of gas in liquid
e’ = concentration of reagent in solution
If, however, there is a constant resistance a t the surface, the C: = initial or bulk Concentration of reagent
boundary condition is changed to
E = Laplace transform of c =
D = diffusivity of absorbed gas
Lm- ce @de
 June 1951 INDUSTRIAL AND ENGINEERING CHEMISTRY‘
D‘ = diffusivity of reagent in solution
d = thickness of liquid layer on packing or other solid surface
H = Henry’s law constant
k = generalized mass-transfer coefficient
kr, = liquid-film mass-transfer coefficient
kc = gas-film mass-transfer coefficient (partial-pressure units)
ks = mass-transfer coefficient for surface resistance (concentra-
tion units)
kE = mass-transfer coefficient for eddy diffusion
I = distance traveled in time e by element of surface of moving
liquid
R = mean rate of absorption per unit area of nonstagnant
. $”’
liquid
r = velocity constant for first-order reaction between absorbed
gas and liquid
velocity constant for second-order reaction between ab-
sorbed gas and reagent in solution
s = fractional rate of renewal of surface of liquid
so = value of s for area a
f v = volume of liquid per unit volume of packing
X L = “effective thickness of liquid film”
x = distance beneath surface of liquid
B = quantity defined by Equation 16
e = time for which a liquid surface has been exposed to gas,
“age” of surface
* = rate of absorption into unit area of surface of stagnant
liquid
Air-Water Contact Operations
in a Packed Column
FUMITAKE YOSHIDA
DEPARTMENT OF CHEMICAL ENGINEERING, KYOTO UNIVERSITY, KYOTO, J A P A N
S 1MULTANEOUS interphase transfer of heat and water vapor
between air and water flowing countercurrently in packed
columns is of considerable engineering importance in humidifiers,
dehumidifiers, and water coolers. T h e performance of this
equipment used t o be expressed in terms of the over-all coefficient
of heat or mass transfer, or in the over-all height of transfer unit
(H.T.U.), the bulk-water temperature usually being assumed
equal t o the interfacial temperature. However, this assumption
is indisputably valid only in the case of constant water tempera-
ture humidification, in which the heat given up by air is wholly
consumed for evaporating water and hence no heat transfer takes
place across the water film.
1 T h e present work was intended (I) t o study the effects of the
gas and water rates on the true gas-film coefficients of heat and
mass transfer from the constant water temperature data; (2)
t o investigate how the water-film resistance, if any, was affected
by the gas and water rates from the water-cooling data; (3)
to study whether the correlations for the film coefficients obtained
from the above tw8 operations were applicable to the dehumidi-
fication data.
This work was undertaken before the similar study by Mc-
Adams and coworkers ( a ) was published. Some discrepancies
exist between the results of the two investigations.
EXPERIMENTAL
T h e schematic diagram of the apparatus is shown in Figure 1.
The column (Figure 2) was 10 inches (25 om.) in inside diam-
eter and was dumped-packed with 15-, 25-, or 35-mm. ceramic
Rasohig rings t o a depth of 12.5 inches. It was necessary to
1467
4 = surface-age distribution function
M = viscosity of liquid
p = densityofliquid
erf(z) = 1 - erfc(z) = e -Pdy (numerical values may
be found in tables)
LITERATURE CITED
(1) Carslaw, H. S., and Jaeger, J. C., “Conduction of Heat in
Solids,” p. 43, Oxford University Press, 1947.
(2)Ibid., p. 53.
( 3 ) Ibid., p. 240.
(4) Danokwerts, P. V., Research, 2, 494 (1949).
(5) Danokwerts, P. v., Trans. Faraday Soc., 46,300 (1950).
(6)Ibid., p. 701.
(7) Higbie, R., Trans. Am. Inst. Chem. Engrs., 31, 65 (1935).
(8) Perry, J. H., “Chemical Engineer’s Handbook,” pp. 1179. 1184,
New York, MoGraw-Hill Book Co., 1941.
(9) Sherwood, T. K., “Absorption and Extraction,” p. 61, New York,
McGraw-Hill Book Co., 1937.
(10) Ibid., p. 196.
(11) Ibid., p. 202.
RECEIVED
August 8,1950.
Engrnyring
Process
development
I
AND TATSUO TANAKA
T h e work was undertaken to study the film coefficients
of mass and heat transfer in the three kinds of air-water
contact operations-i.e., constant water temperature
humidification, water cooling, and dehumidification, in
columns packed with ceramic Raschig rings.
The results showed that liquid-film resistance was not
negligible as compared with gas-film resistance and that
the same empirical equations for gas- and liquid-film coef-
ficients could practically correlate the performance of a
packed column throughout the three operations. The gas-
film coefficients of heat and mass transfer were propor-
tional to gas rate and to the 0.2 power of water rate, while
the liquid-film coefficient of heat transfer was proportional
to the 0.8 power of water rate. The ratio of eas-film coeffi-
cient of heat transfer to that of mass transfer nearly
equaled the humid heat of air in the column.
The correlations obtained should make it possible to de-
sign packed column-type air-water contact apparatus on a
sounder basis.
make the packed section relatively short in order to obtain sub-
stantial driving potential at the top of t h e packin . Due
care was taken in the column design t o minimize the e n 3 effects
owing to the sprays above apd below the packed section. As
shown in Figure 2, nineteen overflow pipes from t h e water-
distributing tray were extended down t o t h e top of the packing,
thus eliminating the air-water contact above the packed section.
Below the packing, there were a tray with air risers and an over-
flow ipe through which water was drawn out of the column.
The c h a n c e between the water level on the tray and the bottom