Principles of Gas Absorption

By W. K. Lewis and W. G. Whitman

MASSACHUSETTS INSTITUTE OF

TECHNOLOGY, CAMBRIDGE, ?.IASS

permit proper estimates and designs for new operations.

Further progress in this field can be made most effectively

through
SI study of the mechanism of absorption phenomena
and of the factors which control them. When the basic prinXGINEERIPUG information on the operation of gas
absorption equipment is at present quite inadequate to
ciples have been definitely established, they can then be
applied to specific problems by the engineer in the same manner
as he now employs heat transfer coefficients, friction
factors, and the like.
The object of this paper is to outline certain of the more
fundamental conceptions which have been developed from the
available data and from the general laws of reactions between
two phases.
The basis of all processes involving the absorption or the
escape
of gas lies in the fact that a liquid-gas system which is
not in equilibrium tends to approach equilibrium conditions.
Thus,
if the liquid is not saturated with gas under the existing conditions, absorption occurs, whereas if it is supersaturated the reverse
is true. The escape of gas, as applied in
this case, is merely negative absorption, and for simplicity the
discussion will be confined to absorption alone.
Equilibrium or saturation represents the ultimate state
which the system tends to assume, and is the first of the primary characteristics of absorption phenomena to be considered. The other fundamental is the rate at which the system approaches equilibrium, and in many cases the rate
is
more important than the equilibrium itself. These two factors
are not independent of each other and, in general, the rate is
greater the further the system is from equilibrium.
NATURE OF ABSORPTION PROCESSES
Existing data clearly indicate that in almost all the cases
so far investigated the rate of absorption of a solute from a
gas by a liquid is limited by processes of diffusion. Any
other reactions which may take place are
so rapid relative
to the speed of diffusion that they have no appreciable effect
upon the absorption rate.
It is now becoming generally
recognized that wherever a liquid and
a gas come into contact there exists on the gas side of the interface
a layer of gas
in which motion by convection is slight compared to that in
the main body
of the gas, and that similarly on the liquid
side of the interface there is a surface layer of liquid which is
practically free from mixing by convection. This phenomenon

is frequently expressed by assuming the existence of

stationary films of gas and liquid respectively on the two sides
of the interface. This phraseology will be employed because
of its simplicity, although one should not infer that there
is a
sharp line of demarcation between stationary liquid or gas
films on the surfaces and the major bodies of the fluids.
In the main body of either liquid or gas, except under special conditions which will not be considered here, mixing by
convection is
so rapid that the concentration of solute in the
luid is essentially uniform at all points. (The exception
is illustrated in Beckers paper, p.
1220 this issue, in experiments with quiescent liquids. Differences in gas or liquor
concentration due to channeling in absorption apparatus
are not involved in this concept.) On the other hand, the
surface films are practically free from convection currents
and consequently any transfer
of solute through these films
must be effected by the relatively slow process of diffusion.
These films, therefore, offer the controlling resistances to transfer of a material from one phase to another.
Diffusion through the gas film proceeds at a rate that is
proportional to the difference between solute concentrations
in the gas on the outside and inside of the gas film. (Resistance to diffusion due to a gas film is, of course, nonexistent
in the special case where an absolutely pure gas
is being absorbed. This problem
is very rarely encountered in practice, however, since the presence of very small amounts of
inert gas, which will concentrate at the liquid surface, is
sufficient to create an effective gas film.) Diffusion through
the liquid film, on the other hand, is controlled by the difference between the concentration of solute in the liquid at
the interface and
its concentration on the other side of the
liquid film-i. e., in the main body of the liquid..* Since the
surface films are very thin, the actual amount of solute
contained in them at any one time is usually negligible compared to the amount diffusing through them.
It follows,
therefore, that
all the solute which passes through one film
must also pass through the other, and the two films may be
considered as two diffusional resistances in series. Under
certain circumstances the importance of one of the films
may be
so much greater than that of the other that the second
film may be neglected and the problem treated
as if only one
film existed.
The concentration difference through either of these films
represents the potential or driving force that is causing
diffusion to take place.
If the concentration differences be
expressed in identical units-e. g., gram mols per liter-it
is readily seen that the specific conductivity

of the gas film

will be much greater than that
of the liquid. The only
resistance encountered by the diffusing molecules is due to
their collisions with the interfering molecules of the gas
or
liquid through which diffusion is taking place. Because of
the greater density of the liquid the collisions in the liquid are
much more frequent and the diffusional resistance is correspondingly greater than in the gas. However, the resistances
to absorption are dependent also upon the film thicknesses.
It is to be expected that the gas film would be somewhat
thicker than the liquid, because under comparable conditions
of disturbance the major factor in determining film thickness is probably the ratio of viscosity to density and this
ratio is greater for gases than for liquids. Despite this fact, the
greater specific conductivity of the gas film far more than
overbalances the greater film thickness,
so that the actual
diffusional resistance encountered in the liquid film is much
greater than that in the gas.
The amount of solute absorbed per unit time by diffusion
through the two
films is de. (See table of nomencla* Numbers in text refer to bibliography at end of article
ure for resume of all symbols.) The amount of diffusion
is obviously proportional to the surface of the interface,
A,
and it is therefore convenient to refer to the diffusional
current density, Ade. dW This quantity is equal to the diffusion coefficient per unit area (k, for a gas film or k, for a liquid
film) multiplied by the concentration difference available as
a driving force-i. e.,
dW
Ade = K,(P, = kL(Ci

- CL)

where P = gas concentration

and the subscripts g, i, and L refer respectively to conditions
C = liquid concentration
at the outside of the gas
fiIm, at the interface, and at the inside of
the liquid film.

It is clear that the numerical values of the diffusion coefficients

k, and kL will depend upon the units in which the gas
and liquor concentrations are expressed. The choice
of these
units is purely arbitrary. In this paper solute concentration
in the gas is expressed as its partial pressure in fractions of
an atmosphere and concentration in the liquid as grams per
cubic centimeter because of the convenience
of these units.
This makes the separate values
of k, and kL of the same
order of magnitude for many cases and offers a good working
basis which will be employed throughout the paper.
TABLS OF NOYRNCLATURB
W = weight of solute, in grams
8 = time, in hours

E = rate of absorption, in grams per hour

de
A = area of liquid-gas interface, in square centimeters
k, = diffusion coefficient through gas film
kL = diffusion coefficient through liquid film
P = concentration of solute in gas, in atmospheres
C = concentration of solute in liquid, in grams per cubic centimeter
Subscript
g applies to conditions in main body of gas
Subscript
i applies to conditions at liquid-gas interface
Subscript
L applies to conditions in main body of liquid
H = solubility coefficient = liquid concentration in grams per cubic
centimeter divided by equilibrium gas concentration in atmospheres
KO = over-all diffusion coefficient, based on gas concentration
KL = over-all diffusion coefficient, based on liquid concentration
M = molecular weight of solute

The data at present available indicate that at the true

interface between liquid and gas the two phases are
at substantial equilibrium
(Pi in equilibrium with Ci), even though
there may be rapid diffusion and therefore high concentration gradients through the
films on the two sides of that interface. This fact puts us in a position to visualize clearly what
takes place
in an absorption process. (Figs. 1 and 2) Conditions at the outside of the gas
film and at the inside of the
liquor film are the same as those in the main bodies of gas and
liquid, respectively. Conditions at the interface are determined by two factors: first, the equilibrium between gas and
liquor concentrations; and second, the fact that all the
solute diffusing through the gas film must
also diffuse
through the liquid film. Thus,
this function being the solubility relationship;
Equation 1
k,(P, - Pi) = k~(Ci - CL)
(3)

If the values of k, and of kL are known, the values of Pr and

Ci, the interfacial concentrations, are at once determined by
these two considerations. If, for example, k, should be
just equal to
kL, (Pa - Pi) would have to equal (Cs

- C,)

to satisfy Equation
EFFECT OF SOLUBILITY
Certain simplifications of the general equation (1) are permissible when the conditions become such as to make one of
the two films negligible in importance as compared with the
other. Since these conditions are primarily determined by
the solubility
of the solute, it is desirable to consider as three
separate cases the very soluble gases, the slightly soluble
gases, and those of intermediate solubility.
CASE I. VERY SOLUBLE GASES-The absorption of hydrogen chloride in water is typical for the very soluble gases.
In dilute solutions this gas is
so soluble that even a 20 per
cent solution
(0.18 gram per cubic centimeter) exerts a

vapor pressure of hydrogen chloride of less than

0.2 mm. Hg
(0.00026 atmosphere) at 30" C. In this case the solubility
of the gas makes it possible
to build up high concentration
gradients through the liquid film, although the concentration
in the gas at the interface may be very small. In other words,
once it reaches the interface, the hydrogen chloride is rapidly
sucked through the liquid
film and the gaseous concentration
Pi is maintained at a very low figure. The absorption rate for
this extreme case will be determined almost solely by the rate
of diffusion through the gas film under a difference in gaseous
concentration substantially equal to
P, since the value of
Pi is negligibly small in comparison with P,. The general
formula can therefore be simplified
for very soluble gases to
the form
and the problem treated purely as a case of gas film diffusion.
The same result can be reached by a more mathematical
treatment. Consider the absorption of hydrogen chloride by
water from an air mixture in which the partial pressure of
hydrogen chloride is
0.1 atmosphere. The gas at this pressure would be in equilibrium with a saturated solution containing 0.39 gram hydrogen chloride per cubic ;'- bl
centimeter at 30' C.
The liquid concentration
at the interface,
Ci,

would therefore be
somewhere between the
limits of zero (that of
the main body
of liquid)
and
0.39 (that of a saturated solution). This
concentration would be
definitely fixed by Equations
2 and 3. For example, if the absorption
conditions were such
that
k, = kL, the equations would be satisfied
by C< = 0.1 and Pr
0.000021; if ko = 2k~, I
Ci = 0.2 and Pi =
0.00037; and if k, =
0.5 k~, Ci = 0.05 and
Pi = 0.0000054. It will
=

be observed that in the three examples the value of PI is so

low compared to P, that it can be neglected in the expression,
from
(1)

(4)
L /ow0

-0
G#ES 0 H/Cn JOLVB/L/TY

fNc,l
FIQ. 1

and we may substitute the simpler form of Equation 4.

Fig. 1 shows the film conditions diagrammatically, with the
concentrations prevailing at the interface when
k, = k~.
The sudden rise in concentration at the interface is due to the
solubility characteristics of this solute-i. e., a low partial
pressure in the gas is in equilibrium with a high liquid concentration because the solute has
a great affinity for the water
phase.
It will be noted that the concentration gradient
through the gas film is essentially the same as it would be if
no liquid film existed.
GASES OF Low SoLuBILmY-with gases of low
CASE 11.
solubility the rate of absorption is low because only very
small concentration differences can be established across the
liquid film. The solute diffuses
so slowly through the
liquid film that only a small concentration difference is required across the gas film. As a result the liquid at the
interface is substantially saturated with solute at the pressure
P, and it is unnecessary to consider the gas film in the
calculataons.
This may be shown mathematically for the absorption of
oxygen at 0.1 atmosphere by oxygen-free water at 30" C. Saturation under these conditions corresponds to only 0.0000037
gram per cubic centimeter, and the value of
Ci must therefore be between this saturation value and the zero concentration of the main liquid.
As in the earlier example, Ci is determined by Equations
2 and 3. If k, = k~, Ci = 0.0000037
and
PC = 0.0999963; if IC, = 2k~, CC = 0.0000037 and Pi
= 0.0999981; and if k, = 0.5 kt, Ci = 0.0000037 and Pi =
0.0999926. It is evident that in all cases Pi is essentially the
same as Po-i. e., the interfacial conditions are practically the
same as those existing in the main body of the gas. The value
of Ci is therefore the same as that of a liquid saturated with
oxygen at
P,, and may be expressed as C,. It then becomes
possible
to express the rate equation as
and eliminate any consideration of the gas film.
The film conditions for oxygen are represented in Fig. 2,
with the concentrations prevailing when kg = IC,. The
sudden drop in concentration at the interface is due to the low
solubility
of this solute-i. e., its slight affinity for the water
phase.
It should be noted that the relative slopes of the gas
and liquid gradients are the same as in Fig.

1, since
p, - pi Ci - CL kp

( = 1 for both figures)

In the diagram it is necessary to magnify the slopes of the

gradients, since they would appear as horizontal lines
if drawn
to scale.
CASE 111. GASES OF INTERMEDIATE SoLuBILITY-There
are many important cases where the gas is too soluble to permit one to neglect the gas film (Equation 5)) but is nevertheless not soluble enough to allow the liquid film to be disregarded (Equation
4). General Equation 1 must be used as
the basis of calculation
for such gases.
Fortunately, however, an over-all coefficient, combining
the two
film coefficients, may be used whenever Henry's
law holds over the concentration range in question. The
concentration difference to be used with this over-all coefficient is the total difference between gas and liquid, expressed
in comparable units. Thus, if gaseous concentration is
expressed as partial pressure,
Po, the corresponding liquid
concentration must be given as the partial pressure,
P,, in
equilibrium with the liquid concentration,
CL. Correspondingly, the gaseous concentrations must be expressed as C,
(the liquid in equilibrium with Po) if the liquid concentration,
CL, is sed.^.^.^
If the Henry's law relation is shown as
C = HP (6)

the absorption equation becomes

(5)
(7)

or
dW KL(G -

CL)

(8)

Ade =
and the over-all coefficients, Kg and KL, can be obtained from
the individual film coefficients,
k, and k~, by algebra, using
Equations
1,6, wd 7 or 8.
Thus
and
,a

KL =

Hk&,
= Hk,

+ k,

- HkL +"

k,

Kg - kLk
From the nature of the over-all coefficients it will be seen
that the over-all coefficient
KI, can never be greater than the

liquid
film coefficient k~, and similarly, that K, can never
be greater than
k,.
It should be noted that
Equations 7 and 8 can
theoretically the over-all *be applied only to
isothermal absorptions,
since the coefficient of
Henry's law changes
with temperature. In
practice, however, it is
permissible to employ
this method over narrow
GrN

ranges of temperature.
The treatment for
soluble gases which do
not approximatbly obey
Henry's law over the
range being studied, and
yet are not soluble
enough to be treated
under Case
I, may become quite complicated.
C;

0 099996

In such cases, one must

resort to the original (Oxygen)
equation and work out
CC

= 0.0000037

c,= 0

Gases of Low Solubilify

a method of integration which is determined by the character of the solubility

curve. The absorption of hydrogen chloride in solutions

of
strong hydrochloric acid (which exert a vapor pressure of
HC1) represents this type of problem. Fortunately, most
industrial problems can be classified under one of the three
simplified treatments.
Fro. 2

OTHER VARIABLES
The discussion has so far considered only the effect of solubility of the gas
as affecting absorption rate per unit area.
The other important valuables are the diffusion coefficients
k, and k~, which can be varied considerably by the character
of the absorption process. Any factors which tend to cut
down the thickness of the surface films should increase the
coefficients and correspondingly speed up the absorption rate.
Thus, agitation of the liquid increases diffusion through the
liquid film, while higher gas velocities past the surface cause
more rapid diffusion through the gas
film.
The effect of such factors will be somewhat dependent
upon whether the process is controlled by gas film

or liquid
film diffusion. Thus, a change that decreased the thickness
of the gas
film but did not affect the liquid film would increase
the absorption of hydrogen chloride but would not affect the
rate of oxygen absorption.
Changes in temperature affect several factors, and it is
necessary to distinguish clearly between the effect of temperature on the equilibrium and its effect on the coeffioients of
diffusion. Increase in temperature makes the gas less soluble, thus tending to lower the rate of absorption.
The
diffusion coefficients themselves may, however, be either
raised
or lowered by temperature, depending upon changes
in film thickness and in specific diffusivity.
Liquid film
coefficients,
lc~, rise rapidly with increasing temperature,
both because of decreased film thickness due to lowered
viscosity and because of greater diffusivities. On the other
hand, Haslam, Hershey, and KeanK show that the gas film
coefficients,
k,, decrease somewhat as the temperature rises
because the ratio of density to viscosity for the gas decreases
and therefore the gas film becomes thicker. Although little
is known regarding the effect of temperature on specific
diffusivities through gas films,
it seems evident that this
effect is less than that due to an increase in the film thickness.
A type of apparatus in which gas bubbles up through a
liquid would represent vitally different conditions from one
where liquid drops are sprayed through a gas. In the first
instance, a rising gas bubble would continuously expose
fresh liquid surface and the conductivity of the liquid film
would be high, whereas the gas film would be relatively
undisturbed. Such a device should therefore be most satisfactory for absorbing the less soluble gases where liquid film
diffusion is controlling. In the other case, a falling drop
might be expected to have only a very thin gas film but a
fairly thick liquor film. Apparatus of a spray type might
therefore be well suited for absorbing very soluble gases.
From these suggestions it is evident that the ratio of film

(2)

conductivities
varies with the type of apparatus and
, -.
with the conditions of operation. It is therefore quite possible
that the absorption of a gas of intermediate solubility might
be controlled primarily by liquid film diffusion in one piece
of equipment and by gas film diffusion in another. This
point is illustrated later by the case of sulfur dioxide, which
is almost entirely governed by the liquid film when passed
over a free surface of liquid, but which is largely affected by
the gas film when absorbed by bubbling through water.
It is possible to predict comparative performances of different solutes in the same equipment on the basis of this
general concept. Under similar operating conditions the
effective film thicknesses will in most cases be independent

of
the solute that is being absorbed.
It has been shown6 that for low concentrations of the
diffusing component there is a definite relationship between
the coefficient of diffusion of material through a gas film and
the coefficient of diffusion of heat through the same film. In
the units here employed this relationship requires that the
gas film coefficient be proportional to the molecular weight of
the diffusing substance. The total gas pressure and character
of the inert gas are also involved in such a way that the
coefficient varies inversely with the product of the molecular
weight of the inert gas and the total pressure. This relationship is based upon a thermodynamic analysis, which is believed to be valid
so long as the concentration of the diffusing
component relative to that of the interfering gas is low. In
other words, we have dependable information as to the
change of the gas film coefficient with changes in solute,
inert gas, and total pressure. Furthermore, the influence of
such other variables as velocity, temperature, and the like
must be identical for different diffusing substances.
Our knowledge as to liquid film coefficients is incomplete.
It is known that as the molecular weight of the diffusing
material becomes high its diffusivity decreases, but the data
show clearly that the change in diffusivity is small compared with that in molecular weight. In the absence of
more definite information the authors have elected to asume
that the liquid film coefficient follows the same rule as that
just given for the gas film. In the units here employed
this means that the liquid film coefficient,
k~, under conditions
of equal film thickness, is the same for all substances.
This approximation is somewhat justified by the
fact that
diffusivity data given in the literature show a surprisingly
small variation for various weak electrolytes and other
aqueous solutions of low molecular weight.
CORRELATION OF DATA
Certain of the published data on absorption rates illustrate
the applicability of these principles and the effects of certain
factors and types of apparatus. The experimental methods
employed may be grouped in two general classes-those in
which the gas was bubbled through the liquid, and those in
which the gas was passed above the liquid surface. Since
there
is a marked difference in the absorption characteristics
under these conditions, the discussion will consider each one
separately.
For purposes of comparison, wherever possible, results are
expressed, not only as diffusion coefficients, but also (for
liquid film processes) in terms of effective film thicknesses.
The calculation of fllm thickness
is based on Hufner's data for
the specific rate of diffusion
of oxygen through water. Hufner's figures show that
0.067 gram of oxygen diffuse per hour

through an area of 1 sq. cm. when the diffusion layer is 1 cm.

thick and the concentration difference between the ends of
this layer is
1 gram per cubic centimeter. The effective film
thickness in centimeters for an absorption layer is then calculated by dividing Hufner's figure of 0.067 by the value of
k~.
GAS BUBBLED THROUGH LIQUID-Adeney and Becker'
studied the absorption of oxygen, nitrogen, and air by water
from gas bubbles.
The solubilities of these gases are
so low
that they may be classed under Case
I1 as examples of liquid
film diffusion. The absorption rates in these experiments were
very rapid, giving values of
k~ as high as 60 (Equation 5)
expressed as grams abaorbed per hour per square centimeter
when the concentration difference is 1 gram per cubic centimeter. This
k~ of 60, obtained at 20' C. with an elongated
bubble about
15 cm. long, offering 20.5 sq. cm. of surface area,
corresponds to a film thickness of
0.0011 cm. The largest
bubble used in their work had about three times the surface
area of the one noted above and gave a coefficient of 43
corresponding to a film thickness of
0.0016 cm.
The effect
of increasing temperature was to increase the coefficient
almost linearly owing to greater fluidity;
thus the value of
ki, doubled in going from 0" to 35" C.
Ledig and Weaver,s absorbing a bubble of carbon dioxide
in water, report results from which the calculated value of
k~ is about 80, corresponding to a film 0.0008 cm. thick. This
coefficient is greater than the maximum obtained by Adeney
and Becker, but the carbon dioxide bubble was much smaller
and was nearly spherical.
As Adeney and Becker intimated,
the coefficient should be greater under these conditions.
Permang studied rate of escape by bubbling air through
solutions of ammonia, chlorine, bromine, sulfur dioxide, and
hydrogen sulfide in similar experiments.
The data are too
incomplete to permit a calculation of specific absorption
rates, as the rate of air flow and size of bubbles were not
recorded. His work, however, allows a comparison of rates
between gases which vary widely in solubility. Perman's
data are shown in Fig.
3, in which the logarithm of the liquor
concentration is plotted against the time
of bubbling air. The
significance of these data may be analyzed by computing
the over-all diffusion coefficients for the whole apparatus,
AKL and AK, (Equations 8 and 7). If the values of AKL
should be constant for the five gases it would mean that the
process is substantially liquid film diffusion alone for all of

them and hence that the over-all coefficient KL is practically

identical with the film coefficient
kL. On the other hand,
if the values of
AK, were nearly constant, varying only with
the molecular weight as previously noted, it would indicate
that gas film diffusion controls in all five cases, the over-all
coefficient
Kg being equal to the film coefficient k,.
The values of AKL in the second column of Table I were determined directly from the slopes of Fig.
3, with the addiII1

tioiial information that 50 cc. of liquid were employed in

each experiment. The third column gives the solubility
coefficient
H under the experimental conditions. Values of
AK, were calculated by multiplying AK, by H (Equation
lo), and these values are divided by M in the fifth column
to compensate for the effert of molecular weight on the gaseous coefficient.
TABLE I
AKL H A Kg A4 as below
c12 1340
0 0096
12 9 0 18
1090
HzS 1000
0 0048
4.8 0 14
1090
Brz 490
0 238
117 0 73
640

so1
380 0 199
76 12
430
NH?
19
1 90 36
2.1 19
AKL calcd

A K,

Since neither the values of AK, nor those of 7 are constant, it is evident that neither gas nor liquid film alone controls the diffusion for all five gases. The coefficients
AKL
decrease with increasing solubility, showing that the least
soluble gases (chlorine and hydrogen sulfide) most nearly
approach purely liquid film diffusion. For example, ammonia could escape much more rapidly than it actually does if
only the liquid film had to be considered, since the value of
AKL for ammonia would be at least as high as 1000. The
values of

xg

increase with solubility, indicating that with

he most soluble gas, ammonia, the process is largely controlled by the gas film.
The validity of this reasoning may be checked in a semiquantitative way by assuming that the chlorine and hydrogen
sulfide diffusions are entirely liquid film
(AK, = Ak,) and
that the ammonia diffusion is entirely gas film
(AK, =
Ak,). The values of Ak, and Ak, may be taken in round

numbers as
1170 and 2.1 X M, respectively, and used in
Equation
10 to predict the over-all coefficients AK, for the
other gases.
AkLkl7 =
1170 X 2.1M

+ ko
H X 1170 + 2.1M
AKL = Hk,

The coefficients figured from this formula are given in the

sixth column of Table
I and show a reasonable agreement
with the experimental values of
AK, in the secohd column.
GAS PASSED OVER LIQUID SURFACE-Adeney and Beckerlo
studied the aeration of quiet columns of pure and salt water.
They found coefficients,
IC,, for the relatively insoluble atmospheric gas of about
0.4 for pure water and 0.5 for salt water
at
15" C., corresponding to film thicknesses of 0.17 and 0.13
cm., respectively. These rates obtained with quiet liquid
surfaces are about one-hundredth of those realized with gas
bubbles.
When the liquid is stirred the rate, of course, is greatly
increased. Thus
Reeker" obtained coefficients of 5 at 150
r. p. m. of his stirrer and of 15 at 1000 r. p. m. in another set of
experiments with air. Bohr2studied both absorptionandescape
of carbon dioxide by passing gas over a liquid surface, and the
coefficients calculated from his data give
a value of about
7.5 at 250 r. p. m. He also found that the coefficients for
absorption approximately checked those for escape. Davisx2
passed oxygen gas over water and obtained a value of
k, of
about
3.3 at 60 r. p. m.
The effect of velocity on the liquid film is shown in a
qualitative way by the table of results by different investigators.
TABLE I1
R. p. m. INVESTIGATOR Gas kL Cm.
At rest Adeney and Becker 03, Nz 0.4 0.17
3.3 0.020
60 Davis 02
150
Becker 02, Nz 5 0.013
250
Bohr coz 7.5 0.009
1000 Becker 02, Nz 15 0.0045
Film
thickness

It is somewhat fortuitous that the order of absorption rates

so closely follows that of the rates of stirring when different
types
of vessel and stirrer were employed by the various
experimenters. Becker's
own data on rates of stirring,
however, show in
a more quantitative way that the coeffi-

cient increases almost proportionally to the

0.8 power of the
velocity
of rotation.
Davis also absorbed more soluble gases under conditions
similar to those of his oxygen runs. He found the following
over-all rate coefficients
KL and KO:
TABLE IIr
KL Ko
Oz in water 3.3 0.00014
SO2 in water 3.8 0.47
NHa in water 1.6 1.5
HCI in water 0.29 4.0
NHs in 2.3 N HCI .. 2.4

Just as was shown with Perman's data, the values of K, decrease and those of
K, increase with increasing solubility.
Assuming that the absorption of ammonia in hydrochloric
acid solution
is controlled entirely by gaseous diffusion in this
case
(KO = IC,), and that oxygen absorption is a purely liquid
film phenomenon (KL = k, for OZ), one may calculate how
rapidly sulfur dioxide, ammonia, and hydrogen chloride,
which are between these two in solubility, should be absorbed
by water. In a more complete discussion of this work, to
be given in a later paper, it is shown that the predicted absorption rates check the experimental rates within
15 per
cent for all three gases.
It is of interest to note that the ratio of liquid to gas film
coefficients

2 is roughly about

15-30 to 1 in the case of gas

bubbles (estimated from Permans data)

;
whereas the ratio
is only about 1 to 1 for gas passed over a liquid surface (using
Davis data). This difference can probably be attributed
more to the exceedingly thin liquid films obtained around gas
bubbles than to any marked change in gas film thickness.
CONCLUSIONS
Rate of absorption is controlled by rate of diffusion of solute through the surfade films of gas and liquid at the gasliquid boundary. The relative importance of the two
films is
determined primarily by the gas solubility and to a lesser
degree by the conditions of operation.
Gas bubbles give very thin liquid films and are therefore
particularly adapted to the absorption of the less soluble
ga$es. On the other hand, the rates for the more soluble
gases are limited by slow diffusion through the gas, and this
effect becomes increasingly evident with gases above sulfur
dioxide in solubility.
Absorption through free liquid surfaces is enormously
increased by stirring the liquid. The same general considerations hold as for gas bubbles, but the liquid film is
thicker relative to the gas
flm. For this reason the effect

of the gas film is less noticeable with sulfur dioxide, although

with more soluble gases it comes into prominence.

BIBLIOGRAPHY
1-Whitman, Chem. Met. Eng., 29, 146 (1923).
%Bohr,
Ann. Phys. Chem., 68, 500 (1899).
3--lewis,
J. Ind. Eng. Chem.. 8, 825 (1916).
4-Donnan and Masson,
J SOC. Chem. Ind., 89,236 (1920).
5--Haslam, Hershey, and Kean,
Ind. Eng. Chem., 16, 1224 (1924).
6--lewis,
Much. Eng., 44,445 (1922).
7-Adeneyand Becker,
Phil. Mag.,88,317 (1919); S9,385 (1920).
8-Ledigand Weaver,
J. Am. Chem. SOC., 46, 650 (1924).
g-Perman,
J. Chem. SOC. (London), 67,868 (1895).
10-Adeneyand Becker, Phil. Mag., 42,87 (1921).
11-Becker,
Ibid., 46, 581 (1923).
12-Whitman and Davis,
Ind. Eng. Chem., 16, 1233 (1924).

..........
DISCUSSION

In the discussion following this paper, R. E. Wilson suggested that the general absorption equation could be expressed
using partial pressures multiplied by the solubility coefficient in place of liquid concentrations for diffusion through
the liquid film.
Dr. Lewis pointed out that, although this
was quite permissible in those cases where the solubility
followed Henrys law,
it did not really simplify the problem
and could not be used when the solubility relationship was
not linear. The chairman noted that this method of presentation waq employed in a later paper (by Professor Haslam)
and that the choice between the two systems of expressing
liquid concentration would be primarily determined by convenience in using the equations.
Dr. Becker communicated $he following discussion:
It should be pointed out that the film theory can apply only
when the liquid is mixed and in fact that
it will have its most
precise application when the conditions
on both sides of the
interface are distinctly turbulent. When the gas and the liquid
are not kept in motion other factors appear to become of greater
importance and the values of the absorption rates are not then
always proportional to the solubility of the gas.
The conclusion that for the absorption of less soluble gases a
bubble rising through the liquid is the most efficient device is
t)orne out by my experiments
on stirring. When the liquid wa
stirred as rapidly as possible without causing splashing, the value
of the rate of solution was only about half the value obtained
when the absorption took place from
a bubble rising through the
liquid.
A further important point is the form of the bubble,
as this determines the area it exposes for
a given volume, and also
the rapidity with which the film surroundingit is renewed
as it
rises in the liquid. From previous experiments it has been
found that the most effective form is that of

a cylindrical bubble
ascending
a fairly narrow tube, as this gives the maximum of
mixing
of the liquid for a given speed of rise.
The remarkable effect of very slow stirring of the liquid in reducing the effective thickness of the film is
of great importance
when circumstances are such that a relatively insoluble gas has
to be absorbed without bubbling the gas through the liquid, as
by this means it is possible to obtain a very large increase in the
rate
of solution with the expenditure of very little power. Thus
a stirrer running at only
60 r. p. m. increases the value of k~ for
oxygen about twenty-five times, as compared with stationary
water.

Mechanism of Absorption of Moderately Soluble Gases in Water .

By H. G. Becker
STANDARD OIL COMPANY OF NEW JERSEY, ELIZABETH, 16.

J.

WHEN THE LIQUID Is MIXED

vv

thoroughly mixed so that it may be regarded as of

HEN a liquid and a moderately soluble gas are allowed to come in contact and the liquid is kept
uniform composition at all points, the rate of solution of the
gas varies directly
as the degree of unsaturation of the liquid.
This has been shown experimentally by following the course of
the absorption of oxygen and nitrogen in air-free water by
a
method which allowed the liquid surface to remain unbroken
although the liquid was vigorously mixed. The method used
consisted in allowing a large, cylindrical bubble of the gas to
pass up through a tube full of water which was initially airfree, and measuring the reduction in pressure which resulted in
the bubble, due to the absorption of a portion of the gas. After
each traverse of the bubble, the pressure could be restored to
its original value by connecting the gas space with an air reservoir by means of a special tap. The process of absorption
could thus be followed step by step from zero concentration of
dissolved gas to saturation.
The curves shown in Fig. 1 which are for oxygen, show the
regular way in which the percentage saturation of the liquid
with the gas increases with time, and
it is found that when the
rate of solution at various points on the curve is plotted
against the corresponding gas content of the liquid a straight
line results. The rate of solution can therefore be expressed
in the form
Integrating,
dw

- a - bw

-dt

w - a/b = ce-bt

For the experiments with air-free water,

Whent = 0,w = 0,c =- a/b

Therefore w = a/b (1 - e-b8)
and when t = infinity, a/b = w = w8, the saturation value
Hence,

- e-bt)
where w = weight of gas per cubic centimeter of liquid at
time, t
w. = weight of gas per cubic centimeter of liquid at
equilibrium
When the initial gas content wt is not zero, this expression
w = w, (1

becomes
1

Adeney and Becker, Phil. Mag., 88, 317 (191s); 89, 385 (1920)