March, 1935 INDUSTRIAL AND ESGIKEERING CHEMISTRY 255

PRESENT DEWELOPMENT~ ACKNOWLEDGMENT

Xotwithstanding the economy of the present operation, a Thanks are due the late M. C. Burt, president of the Gray
modification has recently been made, and is the basis of a Chemical Company, for permission to use data obtained a t
patent application, which makes an important saving in the this plant, as well as to T. 0. Wentworth of the Vulcan Copper
size (and consequent cost) of the equipment and of the steam and Supply Company, engineers and builders of the distilla-
consumption. This improvement is now being incorporated tion equipment.
in the operations and has cut the steam cost as given above
some 28 per cent t o about 25 pounds steam per pound anhy- LITERATURE
CITED
drous acid from 6 per cent liquor, or 20 pounds steam per (1) Keyes, D.B., IND.ENG.CHEM.,21, 998 (1929).
pound anhydrous acid from 8 per cent liquor. The capacity of (2) Othmer, D.F.,Chem. & M e t . Eng., 39, 136 (1932).
(3) Ibid., 40, 631 (1933).
the azeotropic unit has a t the same time been increased some (4) Othmer, D.F.. IND.ENG.CHEX.,20,743 (1928).
45 per cent. This large increase in capacity (or decrease in size ( 6 ) Othmer, D.F., Trans. Am. Inst. Chem. Engrs., 30,299 (1933)
necessary for a given throughput) coupled with the fact that, (6) Othmer, D. F.,U S. Patent 1,930,146(Oct. 10, 1933).
many simplifications of equipment have been found possible (7) Young, Sidney, German Patent 142,502 (June 25, 1903).
will make additional units much cheaper to install and operate RBCEIYED
January 18, 1935.

Distillation and Absorption in Packed

Columns
A Convenient Design and Correlation Method
T. H. CHILTONAND A. P. COLBURX,E. I. du Pont de Nemours & Company, Wilmington, Del.

T HE widespread use of columns packed with various

materials for the absorption of soluble gases and for
distillation, principally analytical but also on a limited
commercial scale, makes desirable the development of a simple
continual evaporation of liquid and condensation of vapor at
the interface proceed a t a rate so great that the surface layer
of vapor is always in equilibrium with the liquid, and the only
resistance to the exchange of material is that of diffusion in
method for determining the height to perform the required the vapor phase. Owing to the diffusion of the low-boiling
seDaration. Since emerimental data on such columns are component in the vapor, the concentration of this compo-
reiatively few and cover for the nent in the vapor increases con-
most part only narrow ranges t i n u o u s l y as the vapor passes
of the variables, it is particularly The proposal is made that the dificulty of upward, rather than in steps as
d e s i r a b l e that these s h o u l d separation in a packed column be expressed, in a plate c o l u m n . A treat-
be c o r r e l a t e d in ajs general a instead of in terms of the “number of theoretical ment of the transfer of material
manner as possible, so that their plaies,” as the number of “transfer units” de- in a p a c k e d c o l u m n must be
reliability can be estimated and based, therefore, on a considera-
that they can be extended as fined, f o r distillation, as tion of the differential nature
widely as possible for the solu- of the changes.
tion of various design problems. It can b e c o n s i d e r e d that
It is the purpose of this paper to over a differential element of
describe a convenient method height, d H , the partial pressure
for the d e s i g n of packed col- and, f o r absorption, as of the low-boiling c o m p o n e n t
umns and, conversely, for the in the vapor is increased by the
correlation of experimental data differential a m o u n t , d p , and
from such columns. since in distillation t h e t o t a1
moles of vapor passing successive
L4PPLICATION TO DISTILLATION The eficiency of a column would then be expressed points remains essentially con-
as the “height of a transfer unit” ( H . T. U.) in- stant, the d i f f e r e n t i a l rate of
In a p a c k e d column u s e d increase of moles of low-boiling
for distillation, an a s c e n d i n g stead of the “height equicalent to a theoretical component in the vapor is given
stream of vapor passes counter- plate” (H. E. T. P.). The d u e s so obtained are by the equation:
c u r r e n t t o t h e liquid reflux. directly related to coeflcients f o r the transfer of
At all points the concentration material by diffusion, but the calculation of the
of the lower boiling component
in the vapor a t the ,vapor-liquid necessary height of column to effect a gicen sepa-
i n t e r f a c e is greater than that ration is made more convenient by the use of the
in t h e v a p o r s t r e a m . As a transfer unit concept. This rate must be equal to the
result, there is a diffusion of the r a t e of t r a n s f e r by diffusion
It is shown that f o r many cases of distillation which may be expressed:
lower boiling component away
from the l i q u i d s u r f a c e and
the H . E. T. P. and H. T. U.are nearly identical
of the higher boiling component but that there may be large differences between dw = K A p d A = K o A p S d H
t o it. It is assumed that the themfor certain absorption problems. (2)
256 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 27, No. 3

For this case, where diffusion occurs equally in both direc- tained from correlations of mass transfer factors. Proce-
tions, mass transfer coefficient, K , is theoretically independ- dures for integrating Equation 6 are also given in detail later.
ent of the partial pressure of the component considered (6). As a matter of interest, the relationship between H. T. U.
and K is seen from Equations 4 and 5 to be:
H G
H.T . U. = -
n
= -
KaIIM,,, (7)

This concept of the height of a transfer unit is, therefore,

similar in some respects to that proposed several years ago by
Peters (9),the "height equivalent to a theoretical plate."
Continued use of this latter concept since it was formulated
has demonstrated its value, even though it has been recog-
nized that there are no real "plates" or integral steps in a
packed distillation column. Under certain special conditions,
as will be shown later, these concepts become identical.
APPLICATION >t
TO ABSORPTION

The process of absorption of soluble gases is very similar

' I , ~ ~
to that of rectification, in that a gas stream passes counter-
ovO,l &9 0.3 0.4 0.5 0.6 0.7 0.8 0.9 current to the absorbing liquor, and in the absence of an ap-
FIGURE1. MCCABEAND THIELEDIAGRAM preciable liquid-film resistance, the resistance to transfer is
FOR RECTIFICATION
OF ETHANOL-WATER
(EQUI- entirely one of diffusion from the main gas stream to the gas-
LIBRIUM DATAFROM CAREY) liquid interface. For this case, however, the total number of
Composition of feed, 10 mole per cent ethanol; moles passing successive points in the column changes, and
composition of waste, 0.1 mole per rent ethanol: com-
the rate of absorption must be written:
-
position of product, 80 mole per cent ethanol, reflux
ratio, 2.5 moles liquid run-bac,k per mole of iroduct:
L / V above feed 0.715. dw = d(pG/M,)S/E (8)

Equations 1 and 2 can be combined and rearranged to give As previously shown (2), this equation becomes, upon differ-
the relation : entiation of the product, (pG/M,) :
dpGS
dw = - (9)
pa -vm

Furthermore, for
which becomes on integration: this case diffusion
occurs chiefly in
(4) o n e direction, and
by the Stefan diffu-
Thus the required height of column for any problem could sion e q u a t i o n (6)
be determined by a solution of Equation 4, provided that the t h e r a t e is greater
value of K can be predicted. t h a n for two-di-
There is another way of looking a t the problem, however, r e c t i o n a l diffusion
which has proved to be more convenient. The quantity, by t h e factor
rI/pu, where pu is
L: can be considered a measure of the difficulty of a
desired separation, just as the required number of theoretical
the logarithmic
m e a n of (n - p )
and (II--p*). Thus
plates is a measure of the difficulty of a separation to be the value of K in
carried out in a plate column. Because of the similarity of E q u a t i o n 2 is no
the two concepts, as is later discussed in detail, the authors longer independent
have called the solution of the above integration "the number of concentration but
of transfer units." Thus the number of transfer units, n, is: varies inversely
with p g f , and it is
n = J,p2 3 (5) the product, Kpd,
which is independ-
FIGURE 2. GRAPHICAL DETER- e n t of concentra-
This equation can be expressed in terms of mole fraction in MINATION OF NUMBER OF TRANSFER tion. Accordingly,
the vapor, y, as well as partial pressures: UNITSFOR RECTIFICATIONOF ETHA- from Equations 9
NOL-WATER(DATAFROM FIGURE 1)
and 2 the following
Y Y* AV I/ AY e q u a t i o n is ob-
0.001 0,009 0.008 125
0.01 0.035 0.025 40 tained:
0.05 0.16 0.11 9.1
0.10 0.27 0.17 5.9
By this means, without any consideration being given to the 0.20 0.38 0.18 5.56
calculation of the total number of moles being transferred, 0.30
0.4
0.44
0.55
0.14
0.15
7.14
6.67
any problem can be directly expressed as a required number 0.6 0.607 0.107 9.13
of transfer units. Then if experimental data are available as 0.6 0.662 0.062 16.0
0.66 0.696 0.046 22.2
values of height per transfer unit (H. T. U.), the determina- 0.7
0.75
0.732 0.032
0.7725 0 0225 44.5
31.0
(10)
tion of column height involves merely the multiplication of 0.8 0.817 0.017 59.0
n by H. T. U. It is shown later that the values of H. T. U. Area under the curve above feed, 9.3 The integrated form
transfer units; area under the curve below
can' be conveniently correlated with vapor velocity or ob- feed, 2.76 tranafer units. is :
March, 1935 INDUSTRIAL AND ENGINEERING CHEMISTRY 257

For absorption, the general solution of Equation 12 is simi-

lar to the general solution of Equation 6. Values of p are
calculated a t various values of z by means of a material bal-
For this case the number of transfer units is then defined: ance, and then values of p* are obtained a t the same values
of z from equilibrium relations, or values of A p are read from
an operating diagram similar to the McCabe and Thiele dia-
gram for distillation (8). At each point p , and pol must be
calculated, and then a plot is made of p ~ / ( A p ) p vs.
, p , and
The relation between the H. T. U. and the controllingvari- the area under the curve is equal to the number of transfer
ables for absorption are considered later. The relationship units. An example illustrating this procedure is shown by
with K is, from Equations 12 and 11: Figures 3 and 4.

DETERMIXATION OF NUMBER OF TRAKSFEIL UNITS

In continuous rectification a relationship can be derived
between the vapor composition, y, and the liquid composi-
tion, z, leaving any cross section above the liquor feed level o om ato a5 of0 0.25 0.m
as follows: x
FIGURE3. DI.4GR.4M FOR ABSORP-
TION OF BENZENE IN STR.4W O I L
Concentration of benzene in inlet gas, 15
per cent by volume; concentration of ben-
and below the feed by the equation zene in outlet gas, 1.5 per cent by volume;
total reasure 1 atmosphere: temperature,
43’ 8.
Aa&e benzene solution follows
Raoult’s law. Vapor pressure of benzene
a t 43O C. = 200 mm. = 0.263 atmosphere.
Benzene content of oil entering srrubber =
1.0 mole per cent: benzene content of oil
The reflux ratio, L / V , can be taken as constant. although the leaving scrubber (about 50 per cent of satu.
values above and below the feed point are different, on the ration) = 30 mole per cent.
assumptions that the components of the mixture have ap-
proximately equal latent heats and negligible heat of mixing, For cases There the mixture is dilute, so that pa/ does not
and that the temperature change through the column and differ much from p,, and There the range of absorption is such
heat lossps from the column can be neglected. These equa- that the equilibrium relationship follows Henry’s law (i. e.,
tions, with the curve representing the equilibrium composi- p* is proportional t o x, or else the values of p* are negligible),
tion of vapor from the liquid, permit the solution of Equa- Equation 12 becomes:
tion 6 for the general case. This solution was developed by
Thormann (11). It involves choosing a number of values of
z between zv and z p , calculating values of y corresponding
to each from Equation 14 or 15, and then a t the same values where Aplm is the logarithmic mean of the terminal values
of x obtaining values of y* from the equilibrium curve. A of Ap, and (pufr,’pv)m is the average of the terminal values
plot is then made of l/Ay vs. y; and a graphical integration of this ratio. For cases where p* is entirely negligible, Equa-
of the area under the curve between the limits of yI [the com- tion 12 can be simplified to Equation 19.
position of vapor entering the column (i. e., in equilibrium
with z,)] and y2 [the composition of vapor leaving (i. e., the
product composition) 1 gives the total number of transfer
units. The values of Ay can also be directly obtained from
a McCabe and Thiele diagram (7) as illustrated by Figure 1.
The resulting graphical integration is shown by Figure 2.
The customary procedure in testing packed columns is to
operate a t total reflux. This facilitates steady-state opera-
tion and eliminates uncertainties in knowing the reflux ratio
exactly. Under this condition, L/V = 1, and by Equations
14 and 15, y = 2. If the equilibrium relation between y*
and z can be expressed mathematically, a general solution of
Equation 6 can be made for this case. Such a relation, in
terms of a,the relative volatility of the components, is:
ax o aoz O.M om a08 0.10 0.12 014

y* = 1 + (a - l)x (16)
FIGURE4.
MI5

GRAPHICAL DETERMINA-
P a15

TION OF h TOF TRANSFER

~ ~
UNITS ~ ~ ~
Where a can be assumed constant over the distillation range, FOR ABSORPTION OF BENZENE
(DATA
Equation 6 can be integrated to give: FROM FIGURE3)
P ui
z P P* AP PO PO( PAP
0.30 0.150 0,0789 0.0711 14.1 0.85 0.92 15.2
0.25 0.1227 0.0658 0.0569 17.9 0.877 0.94 18.6
This equation holds, however, only for total reflux and for 0.20 0.097 0.0526 0.0444 22.5 0.903 0.952 23.7
0.15 0.0732 0,0394 0.0398 30.4 0.927 0.963 30.7
cases where a is constant; this is substantially true for mix- 0.10 0.0514 0.0263 0.0251 39.8 0.949 0.974 41.0
tures which do not deviate appreciably from Raoult’s lav 0.05 0.0313 0.0132 0.0181 55.2 0,989 0.964 66.1

- -
0.01 0.0150 0.0283 0.01237 80.8 0.985 0.992 81.3
and which have boiling points close together, and for some
other mixtures. Area under curve 4.48 transfer units; by the approximate method
(Equation 18) n 4.18 transfer unita.
 258 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 27, No. 3

n = 2.3 (E)m (E)

log since, by definition, the equivalent diameter D, = 4 S H / A .
For solid packings the following Reynolds number has been
successfully applied for friction correlations (1):
OF VALUESOF H. T. U.
CORRELATION
With the problem expressed in terms of a required number
of transfer units, to determine the necessary height of column,
the height per transfer unit (H. T. U.) must be predicted from where D, = average diameter of packing particle. This is
existing experimental data. Fortunately, a previously recom- the definition used by Fallah (4) in a recent paper. It is not
mended procedure for correlation of data can be directly known which basis will prove the more satisfactory for cor-
utilized for this purpose. relating data; but whichever definition of Reynolds number
It was proposed some time ago (3) that forced convection is applied, it must naturally be the same for design as used
data on heat transfer be calculated as heat transfer factors and for correlation.
these factors be correlated by plotting against the Reynolds
number in a manner analogous to that used for friction fac- H. T. U. OF AN TUBE
UXPACKED
tors. The heat transfer factor, j, can be expressed in terms
either of the temperature conditions or of the heat transfer For the case of rectification or absorption in an unpacked
coefficient, h, as follows: tube, where the liquid completely wets the wall, the mass
transfer factor relation with Reynolds number previously ob-
tained (2, 3) can be applied. This relation, for the turbulent
region, is approximated by the equation,
Extensive correlations were made of j vs. Re for data for flow
inside tubes, across tubes and tube banks, and parallel to j = 0.023 (--)
DG -0.2

plane surfaces. At the same time it was suggested, and later

confirmed (2) that mass transfer data be expressed as mass where D = inside diameter of tube. Since for an empty tube
transfer factors, j, according to the analogous relation (for a = 4/D:
rectification) :
H. T. U. = 10.9 D

OF H. T. U.
COMPARISON H. E. T. P.
AND
These mass transfer factors were also to be correlated by
The suggestion made by Peters (9) was that a given separa-
plotting against the Reynolds number. The H. T. U. can,
tion be treated as though it were carried out in a plate column
therefore, be expressed in terms of these j factors, as follows :
and be expressed as a required number of theoretical plates.
G 1
H.T.U.=-
KaII.Mm
=-(’)
ja pkd
a /a The height of the packed column was then divided by this
number to give the height equivalent to a theoretical plate
(H. E. T. P.). The theoretical plate concept implies that the
It is thus apparent that experimental values of H. T. U. can change in composition of the vapor passing through the plate
be readily expressed as mass transfer factors and plotted is equal to the difference between the composition of the en-
against the Reynolds number and also that, from mass trans-
fer correlations, values of H. T. U. are easily obtained.
Likewise, for absorption, I

Since the relationship between H. T. U. and j is, by Equa-

tions 22 and 23, identical for both rectification and absorp-
tion, it will be of great interest t o compare, when they become
available, data for each of the processes using the same
packing.
For absorption of only slightly soluble gases there exists a
resistance to mass transfer in the liquid, and the above equa-
tions cannot be directly applied. If it is known that a large
proportion of the resistance is due to the gas film, the H. T. U. FIGURE
5. RELATION OF H. T. U. AND
calculated by Equation 23 should be multiplied by the ratio H. E. T. P.
of the gas-film mass transfer coefficient to the over-all value;
this ratio is equal to [l + ( K / K L ) ]where
, K and KL are the
One transfer unit gives an enrichment corre-
sponding to Ub’ - go, where Ub’
age of U * b -
g b and Y*O - YO.
-
Z/O = aver-
One theoreti-
gas and liquid film coefficients, respectively. ea1 plate gives an enrichment corresponding
Correlations of mass transfer factors with Reynolds num- to up - yo.
bers have been made (9) only for the cases of flow inside tubes,
across tubes, and parallel to plane surfaces, for which cases tering vapor and the corresponding equilibrium vapor, or
they were found to be in line with heat transfer factors. For yp - yo = y*o - yo (Figure 5). On the other hand, a
the cases studied, the surface was 100 per cent effective, transfer unit is so defined that yb’ - y. = (y* - y),,. If
whereas in packed columns the useful area appears to vary the mean value, (y* - y)m, is equal to y*. - yo, the two con-
with the liquor rate, and this variable may have as much im- cepts are identical. This obtains when the equilibrium curve
portance as the vapor velocity, or even more. For packed is parallel to the y vs. x or “operating” line, a condition rarely
columns the Reynolds number may be calculated as follows: occurring for appreciable ranges of y. A demonstration of
this relationship for the process of absorption (in terms of
mole ratio units) has been given by Sherwood and Gilliland
(24) (IO). If values of Av increase as the vapor passes through
March, 1935 INDUSTRIAL AND ENGINEERIKG CHEMISTRY 259

the column, fewer transfer units than theoretical plates are Another example may be given, as applied to absorption.
required; if values of Ay decrease, the reverse is true. Suppose a concentration of 5 per cent (by volume) water
Because of the ease of “stepping off” theoretical plates on vapor in air is to be reduced to (a) 0.05 per cent or (b) 0.005
a McCabe and Thiele diagram compared with the trouble re- per cent by absorption in sulfuric acid so concentrated that
quired to make an additional plot of l / A y vs. y and to de- the equilibrium partial pressure of water vapor is negligible.
termine the area in order to find the number of transfer units, The number of transfer units can be calculated by Equation
the following procedure is suggested for distillation problems 19. A comparison with the required number of theoretical
where extreme accuracy is not necessary. On the McCabe plates is as follows:
and Thiele diagram, transfer units can be “stepped off,” if T R A N S F E R THEORETICAL
care is taken to make the magnitude of enrichment afforded UNITS PLATEB
by each unit, yb‘ - Ya, equal to the average of the values, 0,057 HzO in exit air 4.6 0.99
y*. - yo and g*b - y b . That is, to choose a point b (Fig-
0.005k H 9 0 in exit air 6.9 0.999

ure 5) on the operating line so located that

OF USE OF H. T. U.
ADVASTAGES OVER Kn
(US’ - ya) := 0.5 [(2/*a - Ua) + (y*b - ?/a)]
It has been general practice in the past to compute for a
While this involves a little cut-and-try work, the time re- given absorption problem the m-eight of solute to be absorbed
quired is not a t all great. If the equilibrium curve and the and the mean driving force, and then from experimental
operating line were both straight (though not parallel) over values of R a to calculate the required tower volume. This
the change yb - ya, and the logarithmic mean value of procedure could, of course, be extended to distillation to
y*a - ya and y*b - yb were used, the method would be exact. eliminate the errors which might be incurred by use of the
For most problems the inaccuracy involved by the curvature H. E. T. P. concept. On the other hand, the authors have
of either or both of these lines and by the use of arithmetic found the use of transfer units and H. T. U.’s to be not only
instead of logarithmic mean is of little consequence. completely reliable but also particularly coiivenient for both
TABLEI. COMPARISON OF NUMBER OF TRANRFER UNITS distillation and absorption. The H. T. U. is independent of
WITH NUMBER OF THEORETICAL PLATESFOR MIXTUREP the concentration of inert gas and, for cases limited by the
WlTH CONSTAXT CY UNDER TOTAL REFLUX gas film resistance, varies in the turbulent flow range only
TRANBFERTHEORETICAL slightly with mass velocity of the gas, and often even less
V
”.I S u?
-- b (2c UNITS PLATES
0.01 0.50 1.1 46.5 47.5 with throughput rate when the liquor-gas ratio is kept con-
0.50 0.99 1 1 50.0 47.5 stant, since the decrease with liquor rate offsets the increase
0.01 0.99 1.1 96.5 95.0
0.01 0.50 3.0 3.0 3.3 with gas velocity. The H. T. U. has only one dimension,
0.50 0.99 3.0 3.9 3.3
0.01 0 09 3.0 6.9 6.6 length; and, since it is expected in many cases to vary but
a yl = mole fraction low-boiling component in vapor entering column slightly with velocity, approximate values can be easily kept
b ys = mole fraction low-boiling component in vapor leaving column.
6 LI = relative volatility in mind. On the other hand, the mass transfer coefficient,
Ka, varies markedly with velocity, average molecular weight,
In case the equilibrium curve and operating line diverge and concentration of inert gas (in absorption). It is ex-
over part of the distillation range and converge over another pressed in terms of four different quantities, and since eaoh
part, the total number of transfer units may be practically investigator may make a different choice for one or more of
equal to the total number of theoretical plates. Such a con- these from different systems of units, the conversion of units
dition exists for total reflux with ideal mixtures with low rela- is often laborious. The range of values may be a million fold
tive volatilities, if the distillation range is about equally great from one common set of units to another, so that it is difficult
on each side of 0.5 mole fraction. This is shown in Table I to keep typical values in mind or to make comparisons.
by examples in which the number of transfer units is calcu- Finally, as contrasted with the separate use of absorption
lated by Equation 17, and the number of theoretical plates coefficients, Ka,for absorption, and H. E. T. P. for distilla-
ii: calculated by the following stepwise equation derived by
tion, it is a simplification to use the same procedure for dis-
Friiske ( 5 ) : tillation] absorption, and other mass transfer processes in
packed columns, and to have existing data from all the proc-
esses correlated on the same basis. At is apparent, for ex-
No. theoretical plates =
log a (28) ample, that the same methods can be applied to countercur-
rent liquid-liquid extraction processes when these are carried
Since many column tests, where results have been expressed as o u t in packed columns.
H. E. T. P., were carried out under the special conditions
stated above, the values of H. T. U. are practically equal to r\TohfENCLATURE
those of H. E. T. P.
In many cases, however, the number of transfer units differs Any set of self consistent units may be used; those of the
more noticeably from the number of theoretical plates and foot-pound-hour-” C. system are given for illustration; total
the application of the latter concept might lead t o an unsat- and partial pressures are, however, expressed in atmospheres:
isfactory design. For the example shown by Figure 1, both
the number of transfer units and the number of theoretical A = interface area, sq. ft.
D = diam., ft.
plates have been determined and are as follows: D, = equivalent diam., ft.
TRANBFER THEORETICAL D, = diam. of packing particle, ft.
UNITS PLATXB G = mass velocity of gas or vapor, lb,/(hr.) (sq. ft.)
Below feed 2.76 4.25 H = height of column, or packed section, ft.
Above feed 9.3 8.1 H. E. T. P. = height equivalent to a theoretical plate, ft.
H. T. U. = height of one transfer unit, ft.
If, under comparable conditions, the H. T. U. and H. E. T. P. K = absorption coefficient (gas film), Ib. moles/(hr.)
for the packing to be employed were found to be, for example, (sq. ft.) (atm.)
one foot, it is apparent that a design based on the H. E. T. P. Ka = absorption coefficient (gas film), lb. moles/(hr.)
concept would have the feed a t too high a point. If only the (cu. ft. packed volume) (atm.)
K L = absorption coefficient (liquid film), lb. moles/(hr.)
enriching operation were required, the column based on (sq. ft.) (atm.)
H. E. T. P. would be considerably too short. L = liquor rate, Ib. moles/hr.
260 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 27, No. 3

Mm av. mol. weight of vapor stream y* = equilibrium mole fraction of diffusing component
Re Reynolds number out of liquid = p * / n
s cross-sectional area (over which G is measured), Ay = y - y*
sq. ft. n = total pressure, atm.
= relative volatility = Y* (1 - 2)
V vapor rate, lb. moles/hr.
a surface area of packing per unit of packed volume, (1 - Y*) 2
a ft./cu. ft. = viscosity, lb./(hr.) (ft.), taken at “film” conditions
C
d
sp. %eat at constant pressure, P. c. u./(lb.) (” C.)
differential operator
p” = density, lb./cu. ft.
h filyocge\fficient
of heat transfer, P. c. u./(hr.) (sq. ft.) LITERATURE C I T E D
b.1
j heat transfer or mass transfer factor (Equations 20 (1) Chilton, T. H., and Colburn, A. P., IND.ESG.CHEW,23, 913-19
and 21) (1931); Trans.Am. Inst. Chem. Engrs.,26,178-96 (1931).
k thermal conductivity, P. c. u./(hr.) (sq. ft.) (” C./ft.) (2) Chilton, T. H., and Colburn. A. P., IND.ENG.CHEM.. 26, 1183-7
diffusion coefficient, sq. ft./hr. (1934).
kd
n number of transfer units (Equations 5 and 12) (3) Colburn, A. P., Trans. Am. Inst. Chem. Engrs., 29, 174209
partial pressure of diffusing component, atm. (1933).
P (4) Fallah, R., J.SOC.Chem. I d , 53,262-6T (1934).
P* equilibrium partial pressure of diffusing component ( 5 ) Fenske. M. R.. IND.ENG.CHEM.. 24.482-5 (1932).
out of liquid, atm. (6) Lewis, W. K., and Chang, K. C., Trans. Am‘. Inst. Chem. Engrs.,
AP P - P* 21, 127-38 (1928).
APm true mean value of ( p - p * ) over height of column
APm loearithmic mean of terminal values of (2) - m*) (7) McCabe,
.---
ESQ. CHEM.,17, 605-12
W. L., and Thiele, E. W., IND.
Pa y k a l pressure of inert component (atm.1 ri - p (1Y26).
(8) Murray, I. L., Ibid.. 22, 165-7 (1930).
Po, ogaritimic mean of (n - p ) and (n- p * ) , atm. (9) Peters, W. A., Jr., Ibad., 14,476-9 (1922).
t temp., C. (10) Sherwood, T . K., and Gilliland, E. R., Ibid., 26, 1093-6 (1934).
Atm true mean temp. difference, ’ C.
W rate of transfer of diffusing component, lb. moles/hr. (I1) K.* l4>61-4
X mole fraction of diffusing component in liquid RECEIVED November 9, 1934. Contribution 149 from the Experiments1
Y mole fraction of diffusing COmpOnent in gas = p/n Station of E. I. du Pont de Nemours BE Company.

New Design Calculation for Multi-

component Rectification
E. R. GILLILAND,Massachusetts Institute of Technology, Cambridge, Mass.

T HE separation of mate-
rials is a problem often
requiring the attention
of the chemist and engineer. If
A new design method f o r the distillation of
multicomponent mixtures is proposed. The
calculations are based o n the use of “relative
volatility” and “relative operability” of the
method of calculation is to be
able to go from the composition
of the liquid leaving any plate to
that of the liquid leaving the
the materials consist of a mix- subsequent plate. I t is often
t u r e of l i q u i d s , rectification components in the mixture. The procedure is convenient and advantageous
f u r n i s h e s o n e of t h e m o s t rapid and simple, and f o r the condition of infinite t o d i v i d e t h i s step into two
common and convenient means r e f u x the calculation reduces to Underwood’s portions. The first portion con-
of performing this separation. sists in going from the liquid on
Thus, the design of rectifying modification of Fenske’s “relative volatility” the plate to the vapor over it,
equipment is one of the funda- system. For the case where separation between and the latter portion involves
mental problems with which the adjacent components is not relatively complete, going from this vapor to the
chemical engineer is repeatedly the determination of suitable design conditions by liquid from the plate above.
confronted. previous design methods has been tedious. The The composition of the vapor
It would be fortunate indeed above the liquid is determined
if a method of calculation for proposed procedure offers a valuable method for f r o m e q u i l i b r i u m data and
these designs would r a p i d l y facilitating such calculations. The method is plate efficiencies. The e q u i -
give results of high accuracy, useful in determining the proper design condi- librium data may be obtained
but such methods are usually tions, such as dependent terminal concentrations from published results or, for
l a b o r i o u s and time-consuming. h y d r o c a r b o n s , may be esti-
and number of equilibrium plates. The ap- mated by the fugacity plots of
As a great many of the prob-
lems to be solved are proposed plicability of the method is demonstrated by Lewis and co-workers (6,8):
designs, i t is desirable- to have illustrative problems.
Y’fr = Z f P (1)
~

a method of calculation that ?- L?‘

will give a satisfactory solution with a minimum of effort. For detailed use of these methods the reader is referred to
Commercial rectifying columns consist essentially of two the original articles. The latter portion of the step, that of
main types, the plate tower and the packed tower. The former going from the vapor to the liquid on the next plate, is ac-
predominates in industrial rectification, and the majority of complished by a material balance. The assumption of con-
proposed design methods are for this type of column. stant molal overflow leads to considerable simplification; in
Both algebraic and graphical methods of design have been general, it is well within the accuracy of the method of design
used to simulate the successive contact of liquid with vapor and will be used in all following derivations. The relation be-
characteristic of the plate column. The graphical method is tween liquid and vapor becomes:
essentially an attempt to reduce the tediousness of the alge- Above feed plate
braic calculations. The fundamental requirement of a O”X, +
P X P = vs-1 y n - 1 @A)