Vous êtes sur la page 1sur 13

CHU ct a1.




Two methods are developed for the rstimation of plate effiacncy in distilling cohunnt and &os
absorbers. The fint method, which is the more rigorous, deals with the m e h i o n of local Murphree

plate efficiency, and a find working formula is derived.

It is interesting to note that surface tension has no direct effect in determinin the &en
of plate
efficiency, but enters, together with density of liquid and vapour, slot width tad fei t, t c m p n n n c
and pressure of the system, into the expression for the calculation of critical &m rate
dividcr flow
into turbulent and streamline region for choosing the right equation to dculnte the bubble size which,
in turn, is an important factor determining plate efficiency.
The mold ratio of liquid to vapour rate, L V, is introduced d u r i q either the axlrrenion of local
Murphree plate efficiency into overall plate efficiency or the calmlation of yx\1pI number of plates
through the plate-to-plate merhod with Murphree plate efficiency. The d o L/V might pko enter
into a function as a correction factor for the bubble-six cslculation if the column is q e r a t e d at e m e l y
high d u e of the LjV ratio.
The second method correlates the overall plate efficiency of fradOMtiDg columns with L / V ratio,
effective submergence, relative volatility of kcy components and v i d t y of the feed.
Two equations are derived whch, although more empirical, represent a convenient route to
compute the actual number of plates in distilling columns or gas abeorbm.


In the design of commercial distilling columns and gas absorbers, it is necessary to cstimete
the number of actual contacting plates equivalent to the calculated number of theoretical plates.
Estimation of this conversion factor, plate efficiency, is a problem of considerable practical
importance. Plate efficiency of distilling columns or gas absorbas has bcen expressed and
reported in either one of the following three terms (see ' Nomenchtun ' on p. 540 for the
explanation of the symbols used in this paper).
(I) Overall plate efficiency (Eo)-the ratio of rhe number of theoretical plates to the number
of a d plates required for the separation.
(2) Murphree plate efficiency (EMP)
(Murphree, 1925)-the d e g m of approach of the vapour
or liquid to equilibrium with the liquid or vapour from an a d late.
The Murphree late efficiency can be expressed in terms o a amcentration unit of ather
1b.-moles per cubic oot of the fluid or of mole fraction. Since the clmngc of temperature and
pressure across any two successive plates is small under normal opcraung conditions, it can be
shown that Murphree plate efficiencies expressed in either units are almost the same, giving


(Cv,u).v represents the average concentration of vapour leaving the nth plate; (CV,
"), refers to
the vapour composition in equilibrium with liquid leaving the nth plate, 1.e. it is not the average
liquid com osition on the nth plate. Similarly, Murphree plate efficiency can also be expressed
in terms o liquid composition.
(3) Local Murphree efficiency EL^) (Lewis, 1936). The Murphtte late effiaeacy dclincd
above employs either average vapour or liquid concentration and inch es thmfon the eftect
of concentration gradient across the plate. In terms of vapour compositiba, the local Murphtcc
effidencv can be exDressed as
ELM=zz C"
~ --" c~n+1~................................
- -

cut 1


where C,, IS the equdibrium vapour composition leaving a point en lrth plate which would be
in equilibrium with the liquid at that point ; C u + ,and Cuare the actuel compositions of vapour
entering and leaving that point respectively.
In this paper two methods for evaluating plate efficiencies leading to the final calculation
of the number of actual plates are developed and described. The h t method, which is more
rigorous, is developed accordmg to similar basis employed by Geddes (~946)in his correlation.
The second method, although approximate, involves the estimation ef w e d plate efficiency
from minimum data by equations of simpler form and is more conveniently applied. A great
deal of experimental work on plate efficiencies in the literature reported few imponant variables
of operation (Brown & Lockhart, 1943 ; Carey, Griswold, Lewis & McAdams, 1934 ;Drickamer
& Bradford, 1943 ; Gerster, Colburn, Bonnet & Carmody, 1949 ; Gerster, Koffolt & Withrow,
1945 ;Grohse, McCartney, Hauer, Gerster & Colburn, 1949 ;Keys & Byman, 1941 ;Kirschbaum,
1933, 1934; O'Connell, 1946; Peavy, 1937; Walter & Shenvood, I M I ; Williams, Stigger &
Nichols, 1950). This makes the extensive test of both methods di&dt, and correlations seem

J. appl. Chem.,


December, 1951




I. Rigorous approach to the plate-efficiency correlation

A. Local Murphree plate efficiency
Setting up the differential equation for the rate of mass transfer and integrating across the
plate at one point, Geddes (1946) derived the following equation to relate the local Murphree
plate efficiency with the overall mass-transfer coefficient of both liquid and vapour phases (R),
time of contact between gas bubbles and liquid ( t ) , surface area of contact (A), and the volume
of gas bubbles ( v ):
ln(r - ELM)= - K - .............................
- (2)

This equation may be used for the vapour flow from a single slot, an entire cap, or a group of
caps. As the interest is centred on the Murphree efficiency at a local point, and the performance
of an average single bubble is assumed to represent the average performance of all bubbles formed
at a single spot, equation ( 2 ) will be further developed by evaluating the behaviour of a single
representative bubble.
The recent study on bubble from the rectangular slots (Jackson & Ceagleske, 1950) indicated
that bubbles from slots are rarely spherical. The bubbles emerged at first nearly spherical at
all gas rates, but quickly became flat or dished at the base and wrinkled at the upper surface.
The shapes were generally reproducible in all bubbles formed under the sime conditions of
orifice size and gas flow rate. The effect of liquid properties has not been certain.
In the absence of definite quantitative information with regard to the shape of bubbles from
the slots of a bubble cap, the bubble is assumed to be spherical to carry on the development of
equation ( 2 ) . For the spheres, V I A in equation ( 2 ) is equal to one-sixth of diameter (S). Upon
substitution into equation ( 2 ) , the latter reduces to :
ln(1 -- EL,,) = - -s ...........................
Overall mass-transfer coefficient
K in equation (3) is the overall mass-transfer coefficient accounting for both liquid and vapour
phase resistances and is related to the individual liquid and vapour coefficients by

I / K = I!kg

+ m/kL.. ...........................


The slope of the equilibrium curve which is the vapour concentration in equilibrium with
the liquid divided by the liquid concentration, (dC',/dCL), is different at different compositions
of the liquid. However, within relatively narrow range of liquid composition, some equilibrium
curve may be almost a straight line and a constant value can be assumed for its slope. Since
most vapour-liquid equilibrium data are expressed in terms of mole fraction ( Y - X), it is desirable
to express m in terms of Y and X,instead of the molar concentration C.
Assigning an average value of 0.95 for the compressibility of vapours, the molar concentration
ofvapour in the unit of1b.-moles per cubic foot is related to its mole fraction through the following
generalized equation of state :
C', = P Y'/o. 693T ...........................
*( 5 4
The molar concentration in the liquid is related to the mole fraction, molecular weight and density
of liquid by :
CL = (pL/ML)X .............................
Dividing equation (5a) by equation (gb), one obtains the expression for the slope of equilibrium
dC' . = .PML
M = -v
- - .........................
dCL o . 6 9 3 T p ~* d X
For the binary systems of relatively constant relative volatility x , the vapour-liquid equilibrium
data of which are represented by equation (6),

The slope of the equilibrium curve can be calculated from the liquid composition only b y :

m = - dC',


+ ( a - I)C}

- - -. -


................ ( 6 4
+ (z - I) PLX


J. appl. Chem.,







Through the analogy between heat and mass transfer (Bedingfield & Drew, 1950 ; Chilton
& Colburn, 1934 ;Chu, 1950 ;Jackson & Ceagleske, 1950 ;Perry, 1950 ; Shims, Sarno & Sonders,

1948), the mass-transfer coefficient in the gas phase can be similarly correlated as heat-transfer
Among the analogies mentioned in the literature, the most convenient form which agrees
quite closely with most experimental values of mass transfer is one proposed by Chilton & Colburn'
(1934) :
................................ (7)
where v, =; vapour velocity
NRe= Reynolds number, dimensionless
NsC = Schmidt
Chilton & Colburn's analogy refers to a turbulent flow in a round ipe. There seems to be
some technical difficulty in applying equation ( 5 ) to the mass trans er between bubbles rising
from the slots of bubble caps and the surrounding liquid. The flow of vapour through the slots
and liquid on the plate at first seems in no way to resemble the flow through a round pipe. In
evaluating the Reynolds' number in equation ( 5 ) , it is difficult to determine the values to be used
for the diameter of the tube through which the flow of vapour takes place, and the linear velocity
of the vapour flow.
Visual observation (van Krevelen & Hoftijzer, 1950) of formation and rising of gas bubbles
in liquids from a single immersed orifice indicated the existence of two types of bubbling. At
low rate of flow of gas, separate buobles are formed from the slots and rise through the liquid
in a vertical direction. As the flow rate increases and passes a critical value, the frequency of
the bubble formation will be increased to such extent that the bubbles rise through the liquid
one after another in series, like a continuous chain. The vertical slots in a bubble cap might
give a picture different from that of a single horizontal orifice, and the presence of a number of
slots in a bubble cap and the proximity of the slots of other caps on an actual plate may create
the collision between different bubbles, changing their size and, consequently, their ascending
velocity with possible deviation of their direction of flow from the original vertical line. However
to facilitate the estimation of the mass-transfer coefficient inside gas bubbles on the plate, bubbles
formed from one slot are considered as chain-like bubbles arising in a vertical position without
change of its size during their travel to the top level of the liquid. In the operating range of
vapour rate this assumption of chain-bubbling, which is valid in commercial columns, will lead
to a icture that each bubbling chain resembles a flow of gas through a round tube of diameter
equa to that of the spherical bubbles (S)and make the application of Chilton & Colburn's analogy
possible. A complete expression for the mass-transfer coefficient in the vapour phase can be
obtained by substituting terms of physical properties into two dimensionless groups in equation (7),

where S = diameter of bubbles, ft., calculated by equations (21), (23) or (21a), (23u), the derivation
of which will be discussed later
density of the mixed vapour rising through the bubble caps in distilling columns,
lbs. ~ f t . ~
Although D,, pv and a vary with temperature and pressure, the combination of these three
properties in the Schmidt number, p;p,D,, makes the latter independent of pressure and
insensitive to the change of temperature as indicated by both theory and experiments.
Diffusivity for binary systems may be estimated from the intermolecular potential energy
by the general theory developed by Chapman & Cowling. By employing a specific form
for the intermolecular potential energy, Hirschfelder, Bird & Spotz (1949) developed a
satisfactory correlation for diffusivities of simpler compounds.
For multi-component systems, Wilke ( 1 9 5 0 6)
~ ~developed the following expression for the
diffusivity of compound A with respect to a multi-component mixture containing compounds
A, By C . . . etc. :
DvA:: - -. - .
'1Y 5 - - -_
Y s ' D v . 4 ~T Y c . ' D v i c - YDiDVAD
. . . . . ..............(94
Application of this relation, which relates the diffusivity in a multi-component mixture to those
of different binary systems, was studied by comparison with exact solutions of the diffusion
equations for various types of diffusion and was supported by experimental evidence.


J. appl. Chem.,


December, 1951




For the mass transfer in the liquid phase (kL), one can imagine mass transfer between a
cylinder created by chain-bubbling and its surrounding liquid. The recent work of Linton &
Sherwood (1950) explored the Chilton-Colburn analogy to mass transfer in turbulent flow of
liquids, representing a xooo-fold extension of the experimental range of the Schmidt group.
Their experimental data were obtained on the rate of solution of cast tubes, cylinders, plates and
spheres of benzoic acid, cinnamic acid and 2-naphthol in water in both streamline and turbulent
flow. Under the same assumption of turbulent flow as in the gaseous diffusion, k~ can thus
be calculated by a similar equation to equation (8), substituting p~ for pv etc. :

The experimental value of the liquid diffusivity is used in evaluating the Schmidt number
wherever it is available in the literature. Owing to the difficulty of eliminating natural convection
and the occurrence of association and dissociation in the liquid state, the present method'for
the liquid diffusivity has not been satisfactory. Wilke (1949)developed a method for the estimation
of diffusion coefficients based upon the relationships suggested by the theory of absolute reaction
rates and Stokes-Einstein equation. Correlation has been made through the equation :
DL =: 3'884T .................... ..........(10)
Within the limits of available data, F is independent of the temperature for a givm liquid system.
For the diffusion of various solutes i s a given solvent, F is a function of the molal volume of the
solute. Curves of F versus the molal volume of the solute have been developed for the diffusion
through water, methanol and water. The experimental observed data gave an average deviation
of IOO/~ from equation (10). Diffusion factor in any other solvent may be estimated if it is assumed
that the data will correlate on a curve having a constant solvent factor [see Fig. 4, Wilkes (1949)].

Time of contact
Time of contact of a given bubble represented by t in equation (3) is equal to the distance
of travel through the fluid divided by the ascending velocity of the bubble through the fluid. In
a commercial fractionating tower the ascending velocity of the mass of bubbles is apparently
greater than that for an isolated bubble in a large amount of liquid. According to Geddes (1946),
the effective time of contact for a gas bubble taken as the mean slot submergence with respect to
the weir crest divided by the ascending velocity for an isolated bubble is equal to the actual time
of contact of a bubble in the mass of bubbles through large amount of liquid. This assumption
approaches the actual mechanism in a tower operating at higher pressure.
Similar conditions have been observed on a bubble plate. The increased bubble rate is
compensated by the increased height of the liquid level for the bubbles to travel. This is
equivalent to saying that the actual time of contict of vapour bubbles with the liquid on a bubble
plate in operation is just equal to time of contact of an isolated bubble with a quiescent liquid,
t = (bw + h~ i- 0.5&)/V. .........................

The sum of hw, h~ and h, represents the average depth of liquid in quiescent state through which
a bubble travels.
3, Bubble size and its ascending velocity
Although superficial observations give the size of the bubble about 4 in. diameter for an
ordinary round cap with narrow slots, few data on sizes of bubbles from plates in operation have
been published in the literature. Davidson & Amick (1948) bubbled air at slow rate through an
orifice and rectangular slots immersed in water and p-qmene and reported that slot-width and
flow rate have little effect on bubble size. Geddes (1946) adapted the theoretical equation for
the size of a bubble that was developed slowly and under equilibrium condition from a vertical
slot and modified the value of constant from 0.78 to unity in the equation empirically to account
for the condition of rapid flow on an actual plate, giving:
i hw) l 3
s - <:a(h,
- -- .- ..............................
; PL - Pv
Davidson & Amick (1948)found that this equation gave a bubble radius only about one-third
the size formed in their work under the same flow condition. At the rates of flow prevailing
in normal operation on bubble plates, the effective bubble size is, indeed, smaller than it would
be at low flow rates. The decrease in size is created by the turbulence and agitation in the liquid.
Recent work on plate efficiency in extractive distillation gave poor quantitarive agreement with

J. appl. Chem.,


December, 1951



Geddes correlation (Grrster ci al., 1949'; Grohse et al., 1949). As the basis of Geddes correlation,
in principle, was generally sound, the discrepancy was explained by a pronounced calming effect
in the plate action with respect to spleshing and entrainment for given vapour velocities. The
use of a new value for the empirical constant in equation (12) might result in satisEoctory agreement
with the experimental values. This constant, as guessed by Grohse (1949), m y be a function
of the liquid rate which was not included in the Geddes method.
The recent contribution on the study of formation and rising of gas bubbles in liquid (van
Krevelen & Hoftijzer, 1950) is used here to evaluate the radius of the bubbles as well as its ascending
velocity through liquid on a bubble plate. Chain-bubbling of gas through the liquid, which is
the actual phenomenon on a plate in a fractionating column or gas absorber, taka phce in two
regions, turbulent and streamline, on the basis of experimental results. Experimental results
were graphically correlated by plotting the expansion group, which is a dimensionless group
represented by gS(pL - pp)/w.2 pr,, versus the Reynolds number. In the streamline region, the
expansion group is inversely proportional to Reynolds number, giving

N E X== 18;NR,..


which, upon the substitution of the dimensionless groups, gives the bubble diameter :

For the spherical bubble, the ascending velocity through liquids




is related to



S2V,,. ....... :. ................(15)

Molal flow rate of vapour is used more frequently in the distillation calculation. The volumetric
flow rate is converted to the molal flow rate of vapour by the following generalized equation of
state if an average value of 0.95 is assigned to the compressibility factor of vapours :

o . 6 9 3 T V , ' P . ..........................


Combining equations (14),( I S ) and ( 1 6 )one obtains the final expression for the bubble diameter
in the streamline region :



V T p L _ -. l l r

. p"),"



This equation indicates that the bubble diameter is a function of liquid viscosity, vapour rate,
and the difference between liquid and vapour density. Although the. equation is derived from
the experimental results of an orifice, the diameter or the dimension of the orifice gives no effect
on the bubble size. For this reason, the same equation is recommended for the calculation of
the size of bubbles from slots of a bubble cap in the absence of additional information concerning
the latter.
In the turbulent region, the expansion group is a constant and independent of t$e Reynolds
number, giving :

s =- d P 1 .





which, upon substitution in equation (IS), gives the bubble diameter in the turbulent region:

This equation, which is derived from experimental results of orifices, indicates the independence
of the bubble diameter from the dimension of an orifice, and is recommended for the calculation
of the diameter of bubbles from slots of a bubble cap.
To determine which is the right equation to use-(17) or (rg),-the critical value of flow rate
of the vapour dividing the turbulent or streamline region of flow should be calculated first. At
the critical value the bubble diameter in chain-bubbling was given by the following equation :

(PI. -- P J) 1 3
J. appl. Chem.,


December, 1951




From equations (19)and

an orifice as :



one obtains- the critical value of flow rate of the vapour through




-- PVJ2


Equation (21) is derived by considering that the buoyancy of the bubble just balances the
surface force holding the bubbles to the orifice at the moment when a bubble is just on the verge
of breaking away from an orifice. For the rectangular slots, the length along which the surface
tension is acting equals twice the sum of slot height and width. For this reason d i n equation (21)
should be replaced by z(hs $- h,)/z to give the critical value of flow rate from a rectangular slot.
The resulting equation is further transformed with the substitution of equation (16)~
giving the
critical flow rate:
P [ a5(k
-i 4 5

v = 3.33 i<


---pv> .....................



It is interesting to note that the surface tension and the dimension of slots appear only in the
expression of critical flow rate of vapour. They are important in locating the region of the flow,
streamline or turbulent, but have no effect in determining the bubble size in either streamline
or turbulent region. It will be shown later that the ascending velocity of the vapour bubble
is a function of bubble size only and therefore independent of the dimensions of slots as well
as surface tension of the liquid.
I n deriving equations (17) and (IS),the circulation of liquid except by the action of the bubbles
is assumed to be absent. In an actual column, especially in extractive distillation where a large
amount of liquid solvent flows from plate to plate, it is believed that a factor which may be a
function of liquid rate, or possibly a ratio of liquid to vapour rate, should be used as a multiplying
factor in equations (17) and (IS),giving the bubble diameter.
Streamline region :

Turbulent region :
S =- o.o2418f(L,V)~-

- - -

........ ............(19.)

f(L/V) is unknown at present for the lack of experimental data.

The ascending velocity for bubbles formed in series has been found independent of the
dimension of-orifice and the same equation is therefore recommended for its calculation, giving :





the value of S can be calculated from either equation (17a) or (19a),depending on whether the
flow is streamline or turbulent.

B. Relation between two Murphree plate efficiencies and overall plate e5ciency
Although it is unreasonable to have local Murphree plate efficiency greater than 1003/0, it
has long been recognized that the presence of a liquid concentration gradient across a plate may
yield the value of plate Murphree efficiency greater than 100% on the basis of the average vapour
composition and the composition of the liquid leaving the plate. Lewis (1936)and Nord (1946)
gave mathematical analyses of three cases for binary systems with respect to the condition of
mixing of vapor and liquid streams from the plate, and numerical comparison between plate and
local Murphree plate efficiencies. Three cases considered by Lewis are :
(I) Liquid flow is in opposite direction on successive plates and unmixed on the plate ;
vapours leaving the plate are completely mixed :
= e ELdumV/L l:(mViL) .........................
(2) Liquid

flows in the same direction without mixing on plates ; vapour flows unmixed ;

J. appl. Chem.,


December, 1951



(Y,,),, = mole fraction of the more volatile component in the vapour leaving nth plate at
a point where the liquid overflows to the next plate.
(3) Liquid flows in opposite direction on successive plates without mixing ; vapour flow is
also unmixed.
(Ynr)E M p = -.............................
f(ELM,Ynr) -- I

For Y,,, > I, the sign in the parenthesis is positive ;

for Yfir< I, the sign is negative.
The analytical solutions of these equations are tedious and difficult. For this reason, Lewis

(1936)gave numerical solutions.

In distilling columns operating at constant molal overflow from plate to plate in each section,
the ratio V / L is equal to ( R ;- I ) / R in the enriching section and to R'/(R' I ) in the stripping
section. R, the reflux ratio, is defined as the molal ratio of reflux liquid going back to the column
to the roduct taken off overhead ; R', the re-boil ratio or boil-up ratio, is defined as the molal
ratio o vapour rising from the re-boiler to the liquid taken off as bottom product. As indicated
in these equations, the difference between two Murphree efficiencies becomes greater as the
reflux ratio approaches the minimum value, and the re-boil ratio reaches the maximum value
required for a given separation.
Graphical solutions of equations (24), (25),(27)and (28)have been presented by Colburn &
Pigford (see Perry, 1950,p. 552) to facilitate the calculation of local Murphree plate efficiency.
In these figures, the Murphree plate efficiency is plotted versus local Murphree plate efficiency
using parameter J representing the term m(R 4 I ) / R .
As pointed out by Carey (see Perry, 1g50), the majority of commercial towers fall between
cases I and 3 in the actual mechanism of flow. That in which the bottom plate which receives
the well mixed vapour from the re-boiler belongs to case I. For plates well up in the upper part
of the column, the vapour flow becomes unmixed and is close to case 3. It might be pointed
out here that the graphs given by Colburn & Pigford, as well as Lewis' relations which are
represented by equations (24)and (28), are limited to the binary systems with liquid flow unmixed
on each plate, and no split flow is involved. Although the local disturbance may cause some
local circulation of liquid, the assumption of non-mixing of liquid in its direction of flow is not
far from the actual realization in a commercial tower.

C. Conversion of Murphree plate efficiency to overall plate efficiency

The Murphree plate efficiency is the same as overall plate efficiency in either one of the
following conditions :
( I ) The average composition of the vapour from the plate is in equilibrium with that of
the liquid from the same plate. This means that the plate efficiency is roo%.
( 2 ) The ratio of vapour rate to liquid rate is always equal to the slope of equilibrium curve,
In enriching section :
R / ( R -7 I ) = m...............................
* (29)
In stripping section :
R';(R' I) = m.. .............................


For a system operated in a column under universal assumptions imposed by McCabe-Thiele's

method, the operating line is a straight line with its slope, R / ( R I) equal to a constant value.
Lewis' relations may be extended to the multi-component mixture if the procedure suggested by
Hengstebeck is followed (Chu, I 950). Multi-component systems are treated as binary systems
of the key components with liquid and vapour quantities in the distilling column adjusted to
' effective ' quantities which are calculated assuming (I) negligible concentrations in the stripping
section of components more volatile than the key components, (2) negligible concentration in the
rectifying section of components less volatile than the key components, and (3) values for the
' constant concentrations ' of the components lighter and heavier than the keys. The ' effective '
liquid and vapour quantities, i.e. quantities of the keys present in the liquid and in the vapour,
are approximated by subtracting the ' constant ' quantities calculated from the total liquid and
vapour quantities. Composition of feed, overhead and bottom products, in terms of mole fraction
of the light key, are also converted into ' effective mole fraction ' of the light key. These values
of ' effective quantities ' are used then to determine the required number of theoretical plates by
a McCabe-Thiele construction in which the keys are treated as a binary system.

J. appl. Chem.,


December, 1951




The necessary condition to satisfy equations (29) and (30) is that the equilibrium curve must
be a straight line parallel to the operating line in the range of composition prevailing in the column
from the bottom of the plate to top. Of course, this may be possible only when the concentration
range is rather narrow.
In most cases where any of these conditions do not exist, the overall plate efficiency and
Murphree plate efficiency are not equal ; their relation derived by Lewis (1936) is as follows :

E - log(I -E
iM P ) ( ~ V -/ LI )
..................... *(3I)
O - Gg(1
E L M ) (mV/L - I) * *
This equation is strictly applied to the binary system with constant slope of equilibrium
curve operated under constant molal overflow. The equation can be applied to the enriching
and stri ping section separately to evaluate the overall plate efficiency for each section. If the
slope o the equilibrium curve changes across each section, the arithmetic mean between the top
and the bottom of the section can be used as an approximation to evaluate the overall plate
efficiency by equation (31) provided the change is not too large. Hengstebeck's method may also
be employed here to extend the procedures to multicomponent systems (Chu, 1950).
The Murphree plate efficiency can be directly used in the plate to plate calculation to find
the actual number of plates without going through the overall plate efficiency. This would be
the only procedure to find out the number of actual plates if the slope of the equilibrium is not
contant and changes are appreciable across each section of distilling columns. For binary systems,
the McCabe-Thiele stepwise procedure on an x-y diagram can be used with slight modification
The graphlcal procedure starts from the point on the diagonal where the abscissa x equals X,,
the composition of the bottom product, by drawing a vertical line to the equilibrium curve.
Locate a point on this vertical line at a height equal to EMP% of the total height between the
equilibrium curve and the starting point. Next draw a horizontal line from this point, in
accordance with the ordinary McCabe-Thiele procedure, to the operating line. Continue to
draw rectangular steps between the operating line and the points which approach the equilibrium
curve within EMr.o/.;until the top of the column is reached. With this modified procedure, the
actual number of plates can be found even if the Murphree plate efficiency varies considerably
from tray to tray in the column. A similar procedure can also be used when the molal overflow
is not a constant and Ponchon's method is employed.

D. Summary of working equations

The final working equations used for evaluating local Murphree plate efficiency and the
procedures for calculating the actual number of plates from the latter have been given in equations
(31, (41, (6)Y (6Q)Y(81, (96)Y (11) and (23).
Combining these equations, the final expression for local Murphree plate efficiency is obtained,
giving :

T o calculate S in equation (32), the critical value of molal flow rate of vapour ( V ) dividing
the streamline and turbulent flow regions should first be computed from equations (22), (17) and

For extremely high flow rate of liquid, as in the extractive distillation, the actual bubble
diameter might be greater than those calculated by equations (17) and (19). In the absence of
definite experimental information, the correcting factor is guessed to be a function of the molal
ratio of liquid to vapor rate of flow, f(LW).
Local Murphree plate efficiency can be converted into Murphree plate efficiency by equations
(24)-(28), the choice of which depends upon the condition of mixing of vapour and liquid streams
from the plate. Among the cases considered by Lewis, only cases I and 3 represent the conditions
prevailed in the majority of commercial towers.
The plate-teplate calculation for theoretical plates can be slightly modified to evaluate the
actual number of plates if Murphree plate efficiency for each plate is known. This modification
is also possible in the graphical solutions, such as in the McCabe-Thiele and Ponchon methods.
An approximate and more rapid computation of the actual number of plates is to use the Murphree
plate efficiency for a plate representing the average conditions of all plates in one column or in
each section of the column. Then the overall plate efficiency can be calculated from the Murphree
plate efficiency by equation (31). The number of theoretical plates divided by overall plate
efficiency gives the actual number of plates.

J. appl. Chem.,


December, 1951



II. Approximate method: estimate of overall plate efficiency

Overall plate efficiency represents the average value of plate efficiency of a great number of
plates in each section of the tower or the whole column. Since physical properties of the fluids
as well as the conditions of hydraulics vary from plate to plate, the correlation of overall plate
efficiency is impossible to be rigorous. As overall plate efficiency is defined as the ratio of number
of theoretical to actual plates, the computation of actual number of plates can be much simplified
if the overall plate efficiency is given, and the estimation of the latter even by an approximate
method of estimation is highly desirable. The procedure employed in computation is simpler
and can be easily carried out at a sub-professional level.
Several correlations of overall plate efficiency have appeared in the literature. Drickamer
& Bradford (1943) presented a correlation of plate efficiency with viscosity for hydrocarbon
separations of about the same relative volatility. This is a logical extension of Walter & Sherwoods
(1941) experimental correlation of local Murphree plate efficiency. OConnel (1946) proposed
relative volatility of key components as an additional factor to extend the correlation of Drickamer
& Bradford. The correlation has been developed for fractionating towers from test data on both
commercial and laboratory columns, By simpllfying the correlation of Walter & Sherwood for
laboratory plate absorbers, OConnel expressed the efficiency of gas absorbers graphically in terms
of viscosity of liquid and solubility of gas in liquid.
Although an average deviation of 10% is claimed, the method inherent in the nature of overall
plate efficiency is empirical and has all limitations of empirical relationship. It is believed that
some important factors which are known to affect plate efficiency are missing. It is the purpose
of this part of the paper to develop a correlation which would include more of the known factors
and be still simple enough for quick computation at a sub-professional level.
In terms of overall plate efficiencies, the important variables omitted by OConnell appear
to be the slope of the operating line, LIV, and the effective submergence, h~ hhs. The effect
of varying slot width is considered to be negligible over the usual ranges encountered.
There is considerable disagreement as to the effect of L / V ratio on overall plate efficiency.
Some people viewed that this ratio has no effect on overall plate efficiency. We, however, believe
that L / V ratio is an important factor and should be included in the correlation of overall plate
efficiency. The reasons for this choice are :
( I ) As discussed in the first part of this paper, the LIV ratio is an important factor in
determining the degree of turbulence on a bubble plate. In extractive distillation, large volumes
of solvent flowing down the column exercises a calming effect upon the flow on the plates. The
bubble diameters are larger than those formed on the plates operating in the normal range of
L / V ratio. As the bubble diameter is an important factor in determining local Murphree plate
efficiency, there should be no doubt about the effect of LIV ratio on overall plate efficiency.
(2) In converting local Murphree plate efficiency to overall plate efficiency, the L / V ratio
is introduced in all three cases of conditions of flow with respect to the mixing of vapour and liquid
flow. The effect of L / V ratio on overall plate efficiency introduced through conversion of
Murphree plate efficiency is op osite to that due to the change of bubble size. Therefore the
net effect of L / V ratio on overal plate efficiency depends upon the relative order of magnitude of
the two opposing factors.
(3) Experimental evidence is contributed by Gerster et al. (1945) ;by Williams et al. (rgjo),
and by Brown & Lockhart (1943).
Effective submergence employed in this paper is defined as the s u m of ( h -t-~ &j, the effect
of liquid head over the weir is thrown into the L / V ratio. The treatmknt has special advantages
in adopting the data of Williams el al. in which the ratio L / V was varied in their work by changing
liquid rate at practically constant vapour rate.
In order to evaluate the effect of these two additional variables, it was necessary to fit an
analytical expression to OConnells graphical correlation. A satisfactory fit to the fractionating
column data was obtained with the simple equation on the basis of calculated results in Table I
and the plot in Fig. I :
E, = 4 9 * 2 ( f ~ a ) - ~. .. .~. ~. ... . . . . . . . . . . . . . . . . . (33)
Since Brown & Lockhart (1943) had shown practically a linear proportion-between overall
efficiencyand L / V , and since Williams et al. (1950) had shown a definite increase in overall plate
efficiency with increasing L / V , it is decided to assume that E is proportional to some power of
LIV. No data in addition to those cited could be found for testing the proposed equation.
Using these cited data and assuming h to be constant, the exponent of L / V was determined
to be 0.295 by means of the results given by Williams et al. (1950) and Brown & Lockhart (1943)
and the plot in Fig. 2 . The proposed correlation then assumed the following form :
E, = 54 * I ( L / V ) O (pL)&ii5:.
. . . . . . . . . . . . . . . . . . . . . .(34)

. .

J. appl. Chem.,


December, 1951








FIG.I . Comparison between O'Connell's correlation and the proposed

correlation for commercial absorber dara

Table I
0'Connell's fractionating-column data

E, 9b





log E














FIG.2. Eficiency correction

facior as a funcrion of

effective liquid deprh

It is likely that the exponent of L / V will be a function of the length of liquid travel across
the plate and mechanical design of the plate. While the L / V correction for a single cap per liquid
path should be small, the maximum L / V correction should be required for some larger number
of caps per liquid path. Between these two limits, a variation in the LIV correction might be
expected. Lewis (1936) found that the influence of m=/V on the plate efficiency for a given local
efficiency was a function of column flow conditions. Thus the reflux ratio may affect the plate
efficiency differently for Merent types of distilling column. The arrangement of caps on each
plate in the distilling column was not given in the literature. It is therefore impossible at the
present time to evaluate the exponent of LIP' as a function of the length of liquid travel across
the plate and the number of caps on the same.
However, from the published information up to date, the effect of the length of liquid path on
plate efficiency is not pronounced for liquid path from 6 in. to 5 ft. Above 5 ft. the efficiency
increases and for liquid path of 12to 20 fi., the efficiency would be increased by 40 to 50% over
that obtained from the present correlation.
As reviously mentioned, available data on the effect of the effective submergence are
essentialfy limited to those presented by Carey, Griswold, Lewis & McAdams (1934)and those
presented by Peavy & Baker (1937). The data of Carey et al. were obtained in a single bubble-cap
plate and therefore are not applicable to the present work. Peavy & Baker used a multi-cap
plate, but, unfortunately, their original data have not been presented. The lack of data made
it impossible to substitute values directly into the present equation in order to determine the
effect of effective submergence. However, a fair idea of this effect can be obtained if the average
ethanol concentration in Pcavy & Baker's work is assumed to be 30 moles-%. It is used only
because most feed concentrations are close to this value. To eliminate the effect of this mismatching, a new constant, B, is used instead of 54-1in equation (31) and calculated as a function
of effective submergence using Peavy & Baker's data, by the following equation :

...................... (35)

J. appl. Chem.,


December, 1951




The ratio BIB,, was then calculated, and represented by C'. B,, was arbitrarily taken as
the value of B equivalent to an effective submergence of 0.75 in. This value was chosen because
9 and the effective submergency
the effective submergency in the work of Williams et al. was o ~ 6 in.
in Brown & Lockhart's work is roughly estimated to have been I in. The value C' is seen to
be a factor for correcting calculated overall plate efficiencies when effective submergence differs
from 0-75 in., and is therefore defined as efficienq correction factor. A semi-log plot of C'
versus effective submergence is presented in Fig. 2.
It can be seen from Fig. 2 that the effect of variation in effective submergence between Brown
& Lockhart's data and the data of Williams et al. affects overall plate efficiency by less than 20,/,.
The previous assumption of a constant effectivesubmergence for these two sets of data is therefore
well within the accuracy of the correlation. The plot of C' versus effective submergence, h,,
can evidently be fitted by an equation of a straight line :
- 0.06
0.092he.. , . . . . . , . . . . . . . . . . . . . .(36)

.. .

With the incorporation of efficiency correction factor by means of equation (36), equation (34)
can be transformed into the following form which -is the final proposed equation for overall plate
efficiency :
E, = 54. I( 10)- 0 ~ ~ + ~ ~ o e r ~ ~ ( ~ / V ) 0.'. .2.9. .5. ( .~. .~ o .c .) .~. .~ (37)
Lilie all empirical equations, equation (37)cannot be extended beyond the limit of experimental data which have been used for the evaluation of its empirical constants. The equation
is applicable for the calculation of overall plate efficiency within the limit indicated, as follows :
L / V ratio, between 0.4 and 8 ; effective submergence, not higher than 1 - 5in. This same
limitation will be applied to equation (39) for gas absorbers.
The effect of submergence on overall plate efficiency, as indicated by Fig. 2, is further tested
by the experimental data of Gerster, Koffolt
Table II
& Withrow (1945). Since the feed compositions
are not given in the original paper, the value Effect of d o t submergence on Murphree plate eficiency.
Data from Peavy 6 Baker (1937)
of @a) for the feed is calculated through
the plate efficiency of 96.2 at L / V of unity to
be equal to 0.558. The values of abscissa submergence, efficiency
corcecuonfactor C'
represented by E are calculated at a Merent
L / V ratio with this value of @a) for the feed,
and compared with those values read from the
I -25
straight line in Fig. 2. The agreement between
two values, as illustrated in Table 11, is good.
The effect of submergency might be more pronounced at lower efficiencies. For instance,
in a column with a point efficiencyof 30%~ doubling the submergence might materially increase
the efficiency, but, on the other hand, if the point efficiency was high-say over go0,40-doubling
the submergence might result in little increase in the efficiency. However, in the absence of
definite experimental information, the effect of submergence which is evaluated at high values
of efficiency, is assumed to be independent of the efficiency.
T o apply a general equation of the type developed for fractionating columns to the absorber
data collected by O'Connell, the assumption was made that the exponent of LIV would be the
same for both fractionating and absorbing columns. With this assumption it w a s possible to
fit the absorber data with the following equation :

. .. .

E, = 4 5 ' 6 ( L / V ) 0 . 2 B 5 ( H P / ~ L. ,).~, ~. .~. . . . . . .. . . . . .. ..(38)

A comparison between this equation and O'Connell's plot is shown in Fig. I. From t h s
plot it can be seen that the agreement between the proposed correlation and the experimental
data is of the same order of magnitude as O'Connell's correlation. , A better agreement with
the experimental data is obtained if efficiency correction factor C' is introduced into equation (38)
to give the following form :


. . . . . ... . .. ..... .(39)

= 45 -6(I O ) - ~

Although it is believed that a direct test of equation (39)with the experimental data is desirable,
the absence of such experimental data in the literature has made the direct proof impossible.
Overall plate efficiency for multi-cap bubble plate are appreciably larger than that for a
plate with a single bubble cap because of liquid concentration gradient across a large plate.
O'Connell's correlation of plate efficiency which is in better agreement with some experimental
data than the complicated Geddes' method accounts only for a few physical properties of the
system to be distilled. In the present work both the slot submergence and the ratio of flow rate

J. appl. Chem.,


December, 1951





of liquid to vapour are important factors in determining the plate efficiency. The ratio of flow
rate of liquid to vapour, which is the slope of the operating line under constant molal overflow
accounts largely for difference of plate efficiency in stripping and enriching columns. On the
basis of existing experimental data, equation (38)is,recommended to estimate an average overall
plate efficiency for the commercial fractionating columns. Equation (39)is for the commercial
absorbing columns. Whereas the corrections for L / V and he in the present work might be too
simple, insufficient data available in the literature which has made a thorough test with experimental
evidence impossible do not justify a more elaborate correlation at the present time.
Too little effort has been devoted to critically analysing the available data to determine
whether it should be used as a basis for correlation. As pointed out by Shiras, Sarno & Souders
(1948)the evaluation of plate efficiency from experimental data is extremely sensitive tothe method
of calculation, the equilibrium data, the analyses of feed and products compositions, and the
conditions under which the test is carried out ;the satisfactory correlation of plate efficiency can
be achieved only when considerable experimental data on plate efficiency accepted by rigorous
evaluation become available. A more fundamental approach of experimental evaluation of plate
efficiency, however, should be the study of the effect of physical properties of the system distilled,
mechanical construction and operating conditions of the distilling column. It is recommended
that a small column be used to study the effect of physical properties to the system distilled in
the absence of additional factors introduced in a large scale commercial column. To evaluate
the variables peculiar to the mechanical installation and the operating conditions of a distilling
column, a large column should be used.
Polytechnic Institute of Brooklyn
Brooklyn 2
New York

September, 1950


Surface area, sq. ft.

Concentration, Ib.-moles/cu. ft.
Diameter of orifice, ft.
D D i f i i v i t y , ft.'/hr.
E , Overall plate efficiency
E m Murphree plate efficiency
ELMLocal Murphree plate efficiency
F Diffusion factor, ("R((hr)*/(lb.)(ft.)
Gravitational constant, 4.18 > I O ft.,'hr.*
he Effective subme'rgence, h~ -t I /zhs
hL Distance from the top of slot to the weir lip, ft.
h, Slot height, ft.
hw Head of liquid over weir, ft.
H Henry's law constant, lb. moles/(cu. ft.)(atm.)
Kg Gas-transfer coefficient, Ib. moles/(sq. ft.)(hr.)
(lb. moles/cu. ft.)
Liquid-transfer coefficient, lb. moles/(sq. ft.)(hr.)
(Ib. moles/cu. ft.)
K Overall mass-transfer coefficient, Ib. molesi(hr.)
(sq. ft.)(lb. moles/cu. ft.)
Flow rate of liquid, lb. molesjhr.
Slope of equilibrium curve
M Molecular weight, lb. Ilb. mole
Reynolds number, dimensionless
NSC Schmidt
Pressure, atm.
P Parachor
Qt. Volumetric flow rate of vapour, cu. ft./hr.


Reflux ratio, L/(V-L),above feed plate

Re-boil ratio, V'(L-V), below feed plate
Temperature, "R
volume, cu. ft.
Va Ascending velocity of bubbles, ft./hr.
V Flow rate of vapour, Ib.-moles!hr.
W Slot width, ft.
X Mole fraction in liquid phase
,, vapour ,,
Z dkpres$bility factor, dimensiontess
Equivalent diameter of bubble, ft.
Density, lb./cu. ft.
Viscosity, lb./hr.ft. [centipoises, equations (33)
and (3911
4 Solvent factor, dimensionless
Surface tension of liquid, Ib./ft.
Relative volatility, dimensionless
C log E, 4-0.246 log (pu) feed

Superscripts and subscripis

average value
A, B, c --compounds
critical state
liquid state
n,n f I plate number counted from the top
at the weir overflow
vapour State

Brown, G . G. & Lockhart, F. (1943). Trans. Amer.
Inst. chena. Engrs 39, 63.
Bedingfield, C. H. & Drew, T. B. (1950). Industr.
Engng Chem. 42, 1164.
Carey, J. S., C~riswotd,J., Lewis, W. K. & McAdams,
W. H. (1934). Trans. Amer. Inst. chem. Engrs 30,
Chilton, T. H. & Colburn, A. P. (1934). Zndusrr.
Engng Chem. 26, 1183.
Chu, J. C. (1950). ' F i h T h e o r y of Fluids,' Encyclopedia of Chemical Technology, vol. VI (New
York : Interscience Publishing Co.).
Chu, J. C. (1948). Chem. Engng Progr. 44, 721.
Chu, J. C. (1950). Chem. Engng Progr. 46 215.

Davidson, L. & Amick, E. H. (1948). ' A Study of

Rubble Size and Local Efficiency,' presented at
AICHE meeting in New York.
Drickamer, H. C. & Bradford, J. R. (1943). Tram.
Amer. Znsi. chem. Engrs 39, 319.
Geddes, K. L. (1946).Trans. Amer. Zmt. cheni. Engrs
42, 79.
Gerster, J. A,, Colburn, A. P., Bonnet, W. E. &
Carmody, T. W. (1949). Chem. Engng Progr. 46,
Gerster, J. A., Koffolt, J. H. & Withrow, J. R. (1945).
Tram. Anier. Insr. chem. Engrs 41, 393.
Gilliland, E. R. (1934). Zndustr. Engng Chem. 22,

J. appl. Chem.,


December, 1951

Grohse, E. W., McCartney, R. F., Hauer, H. J.,
Gerster, J. A. & Colburn, A. P. (1949). Chem.
Engng Progr. 45, 725.
Hirschfelder, J., Bird, R. B. & Spotz, E. L. (1949).
Chem. Rev. 44, 205.
Keys, D. B. & Byman, L. (1941). Univ. of Illinois
Engineering Experiment Station Bulletin No. 328
Kirschbaum, E. (1933). Chem. Fabrik 6, 431.
Kirschbaum, E. (1934). Forsch. Gebiete Zngenieurw. 5 ,
Jackson, M. L. & Ceagleske, N. H. (1950). Industr.
Engng Chem. 42, 1188.
Lewis, W. K., junr, (1936). Zndustr. Engng Chem. 28,
Linton, W. H. & Sherwood, T. K. (1950). Chem.
Engng Progr. 46, 258.
Murphree, E. V. (1925). Industr. Engng Chem. 17,

- .-


Nord, M. (1946). Trans. Amer. Inst. chem. Engrs 42,




OConnell, H. E. (1946). Trans. Amer. Inst. chem.

Engrs. 42, 741.
Peavy, C. C. & Baker, E. M. (1937). Industr. Engng
Chem. 29, 1056.
Perry, J. H. (1950). Chemical Engineers Handbook,
u d ed. (New York : McGraw-Hill Book Co.).
Shirt&, R. N;; Sarno, D. H. & Souders, M., junk.
(1948). A Critical Study of the Evaluation of
Plate Efficiency, Shell Development Co., San
Francisco, Calif., presented at AICHE Meeting
in New York.
Sherwood, T. K. (1940). Trans. Amer. Inst. chern.
Engrs 36, 817.
van Krevelen, D. W. & Hoftijzer, P. J. (1950). Chem.
Engng Progr. 46, 29.
Walter, J. F. & Sherwood, T. K. (1941). Industr.
Engng Chem. 33, 493.
Wilke, C. R. (1949). Chem. Engng Progr. 45, 218.
Wilke, C. R. (1950~2). Chem. Engng Progr. 46, 95.
Wilke, C. R. (1g5ob). Industr. Engng Chem. 42, 471.
Williams, G. C., Stigger, E. K. & Nichols, J. H.
(1950). Chem. Engng Progr. 46, 7.



Sources of germanium and gallium are discussed and descriptions are given of processes for the
recovery of these elements from flue dusts and from germanite. The flue-dust process involves
smelting to recover the rare elements in a metallic regulus, dissolution of this regulus in ferric chloride
solution with the aid of chlorine, distillation of crude germanium tetrachloride from the resulting
solution, fractional distillation of the crude product to remove the bulk of the arsenic, and final
rectification through a column packed with copper turnings to remove the remainder. After hydrolysis
of the resulting tetrachloride germanium oxide is obtained with less than 0 . 1 p.p.m. of arsenic.
Gallium is recovered from the acid liquor in the iirst still by treatment with aluminium to remove
heavy metals and to reduce the iron to ferrous chloride, followed by extraction of the gallium chloride
in a continuous process with isopropyl ether, removal of the ether by distillation, purification of the
aqueous phase with hydrogen sulphide, and conversion of the gallium into sodium gallate for
The germanite process is a modification of that recommended by Sebba & Pugh (1937), combined
with similar refining stages to those used in the treatment of flue dust.

(I) Sources of germanium and gallium

Although germanium was first isolated as long ago as 1886 by Winkler, it was only during
the course of the recent war that any commercial use was found for the element and it is only
within the last few years that satisfactory methods have been developed for the recovery and
refining of germanium on a relatively large scale. Winkler discovered the element in the very rare
mineral argyrodite, which is a silver sulpharsenide in which part of the arsenic is replaced with
germanium. Shortly after the end of the First World War the mineral germanite was discovered
in the copper mines at Tsumeb in South West Africa. This mineral is essentially a copper sulpharsenide c o n t a u 5-10% of germanium and 0 * 3-0 8% of gallium. Small quantities of germanium
also occur in many zinc ores j when the roasted ore is leached with sulphuric acid to provide
a zinc sulphate solution for the production of electrolytic zinc, it dissolves in the leach liquors, and,
unless removed, seriously interferes with the efficiency of zinc deposition. The sludges obtained
during the purification of these liquors, therefore, contain germanium in addition to any cadmium,
gallium, and indium which also occur in the ores. Residues of this type produced from certain
American zinc ores were the first commercial source of germanium and gallium, but the method
employed for the recovery of the metal from these residues has not been disclosed in detail, although,
according to Jaffee, McMullen & Gonser (1946)~the main recovery step involves a distillation of
the germanium as tetrachloride. The American producers claim to be able to obtain a quantity
of the order of 0.5-1 ton of germanium per year from these residues. Gallium is also obtained
from these residues as well as from residues obtained in the treatment of bauxite.

* Read

before the London Section on 4 December, 1950.

J. appl. Chem.,


December, 1951