1016 Ind. Eng. Chem. Res.

1988,27, 1016-1024
Behie, L. A.; Bergougnou, M. A.; Baker, C. G. J.; Bulani, W. Can. J. Wu, C.-S.; Whiting, W. B. Chem. Eng. Commun. 1988, in press.
Chem. Eng. 1970,48, 158. Yang, W. C.; Keaims, D. L. Ind. Eng. Chem. Fundam. 1979,18,317.
Filla, M.; Massimilla, L. Ind. Eng. Chem. Fundam. 1984, 23, 131. Yang, W. C.; Keairns, D. L. in Fluidization; Grace, J. R., Matsen, J.
Flemmer, R. L. C. Ind. Eng. Chem. Fundam. 1984,23, 113. M., Eds.; Plenum: New York, 1980; p 305.
Geldart, D. Powder Technol. 1973, 7, 285. Yates, J. G.; Bejeck, V.; Cheesman, D. J. In Fluidization V; Proc.
Hirsan, I.; Sishtla, C.; Knowlton, T. M. Paper presented at the 73rd Eng. Found. Conf. Fluid.; Oestergaard, K., Soereneen, A., Eds.;
Annual AIChE Meeting, Chicago, IL, November 16-20, 1980. Engineering Foundation: New York, 1986; p 79.
Knowlton, T. M.; Hirsan, I. In Fluidization; Grace J. R., Matsen, J. Yutani, N.; Ho, T. C.; Fan, L. T.; Walawender, W. P.; Song, J. C.
M., Eds.; Plenum: New York, 1980; p 315. Chem. Eng. Sci. 1983, 38, 575.
Raghunathan, K. “A Probe Technique To Study Jetting in a High-
Temperature Gas-Fluidized Bed”. M.S. Thesis, West Virginia Received for review June 16, 1987
University, Morgantown, 1984. Reuised manuscript received December 14, 1987
Wen, C. Y.; Deole, N. R.; Chen, L. H. Powder Technol. 1982,31,175. Accepted January 5, 1988

Transport Processes in Thin Liquid Films during High-Vacuum

Distillation
Maruti Bhandarkar and John R. Ferron*
Department of Chemical Engineering, University of Rochester, Rochester, New York 14627

A description of the liquid-phase transfer processes in high-vacuum distillation is presented. T h e

thin liquid film on a conical, rotating evaporator is modeled by the mass, momentum, and energy
balance equations. T h e equations are solved by a finite-difference method. Effects of various
parameters upon the distillation rates and the separation efficiencies are studied for binary systems.
A simple procedure to scale-up the process is suggested. Experimental data for evaporation of pure
compounds and distillation of a binary mixture are compared with the present theoretical analysis
assuming t h e Langmuir (i.e., maximum) rate of evaporation. On the basis of experimental data,
we may conclude that the present analysis provides the first step toward the complete characterization
of distillation under vacuum and t h a t the analyses of vapor-phase and condenser processes need
also t o be included for a fuller treatment.

Commercial as well as laboratory distillation under high mass-transfer resistances would be expected to be quite
vacuum (pressures ranging from 0.1 to 15 dyn/cm2) is large.
typically carried out in the so-called centrifugal molecular The most recent analysis of the evaporation of a pure
still, in which the distilland flows over a rotating, truncated liquid in high vacuum was by Kaplon et al. (1986). The
conical surface. We have studied the transport processes analysis was for situations in which the liquid is preheated
occurring in thin liquid films formed in such situations as and flows over an insulated rotating disc. All the heat
part of a wider study of the vacuum distillation process. necessary for evaporation came from feed itself. The
The flow of liquid as a thin film over a rotating disk may model describing heat transfer was considerably simplified
be considered as a special case where the cone half-angle and can lead to an overestimation of evaporation rates as
is 90°. we show later.
Greenberg (1972) considered an approximate model of In commercial as well as laboratory high-vacuum dis-
the flow of a pure liquid over a rotating conical surface (see tillation using the centrifugal molecular still, we usually
Figure 1). The heat-transfer analysis was limited to the encounter feeds relatively colder than the distillation
cases in which the feed and mean distilland residue tem- temperature. This means that the distilland heats up as
peratures were roughly equal. This is valid when the feed it travels across the evaporator, often with wide variations
is preheated and the evaporative heat losses are small. in the material properties, and previous analyses assuming
Greenberg’s theoretical analysis agreed with his experi- that the feed is at distillation temperature are inadequate.
mental data for the evaporation of dibutyl phthalate, di- It would be useful from a design point of view to simulate
butyl sebacate, isooctyl phthalate, and isooctyl sebacate. situations involving a variety of operating parameters. In
However, for glycerol the calculated rates were higher. In this work, a general treatment of the heat and mass
all cases the maximum evaporation rate (Langmuir, 1913) transport in the liquid flowing over rotating surfaces and
calculated for the liquid surface temperature was used to distilling under vacuum is presented. The analysis includes
determine the rate. evaporators shaped as rotary cones as well as disks. Sit-
Analysis of the centrifugal still during distillation of a uations in which the feed is preheated, in which the rotary
binary liquid mixture was carried out by Ruckenstein et surface is heated, or combinations of these are considered.
al. (1983) with a view to studying the effect of diffusional Theoretical results are compared with literature data for
resistance among other things. The heat transport analysis pure materials and with binary data obtained in our lab-
was again simplified by assuming that the feed is a t the oratory. Finally, a simple procedure has been described
distillation temperature and that convective heat transfer for scaling up the results from a laboratory still to a larger,
is negligible compared to conductive heat transfer. For commercial still.
the relatively large diffusion coefficient used (4 x Throughout the treatment we assume that the evapo-
cm2/s at 500 K), diffusional effects were negligible. Most rative processes are independent of the vapor-phase
liquid solutions subjected to high-vacuum distillation have properties and the condenser behavior. These assumptions
diffwivities in the range of 104-106 cm2/s, for which cases are valid in the limit of very low pressures; hence, of small
0888-5885/88/2627- 1016$01.50/0 0 1988 American Chemical Society
 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988 1017
The equation describing mass transfer in a binary mixture
Evaporator surface is written as

U -
a2c, 1 aC cot 4 -
ac, + u pac, = D a b -+-a_- ac,
xO ax aY ay2 x ax x ay
2
where C, is the mass fraction of component a in the
mixture of a and b. The appropriate boundary conditions
are
C, = CaO at x = xo (feed location) (9)
aC,/ay = 0 at y = 0 (10)
and
X

Figure 1. Sketch of the evaporator and its coordinate systems.

evaporation rate, and for a cold, highly efficient condenser. (11)

An objective of the paper is to scout the limits of these Here, E , and Eb are the evaporative fluxes (g/cm2/s) of
assumptions. a and b, respectively. We neglect free convection in both
mass and heat transfer, again assuming small temperature
Mathematical Model and concentration changes through the film.
The equation of continuity for the liquid flowing as a The energy balance equation is
thin film in the ( x , y ) system in Figure 1 is
u-+u-=cy
ax
aT ay
aT (
a2T

ay2
1 aT
x ax
cot 4 aT
x ay
) (12)

with the boundary conditions

where u and u are velocities in x and y directions, re-
spectively. Axial symmetry has been assumed here. T = To (feed temperature) at x = xo (13)
(Equation 1may be obtained from a cylindrical form based
on the (z,r) system by rewriting the differential operators 7' = T , (rotor temperature) (14)
in the (x,y) system. Here x = z cos 4 + r sin 4 and y = or
z sin 4 - r cos 4.) The boundary conditions are
a T / a y = 0 (insulated rotor) at y = 0 (15)
u = uo a t x = xo (feed location) (2)
and
and
u=O at y = O (3)
The equation of motion is simplified by assuming that (1) (16)
the liquid film has a fully developed, steady-state velocity
profile for x > xo; (2) liquid pressure gradients are negli- where AH, and AHb are heats of vaporization. (Note that
gible; (3) gravitational effects are negligible; (4)variations radiative flux may be neglected because the surfaces of oils
of temperature and composition, hence of mixture vis- usually distilled in high-vacuum equipments have low
cosity, through the film (in the y direction) are small; we emissivities ( ~ 0 . 5 and
) the temperatures involved are
employ the average kinematic viscosity, usually moderate. For instance, for glycerol at 100 "C, the
maximum possible radiative flux is less than 3% of the
heat flux due to vaporization.)
In our analysis we assume that evaporation rates at the
liquid surface are given by the Langmuir (1913) equation:
(Even though there are small variations of concentration
and temperature over the film, gradients are substantial
because the film is thin. Densities and transport properties
other than viscosity are assumed to be relatively insensitive This assumption is valid if the number of vapor molecules
to changes in composition and temperature. The validity returned to the evaporator due to collisions is negligibly
of this assumption was verified for special cases.) small. This is true when the operating pressures are about
Certain simplifications (Bhandarkar, 1985) also result 1 dyn/cm2 or less and when the condensation of distillate
from considering that the liquid film is very thin compared is complete on a very cold condenser surface. We take the
to the size of the evaporator (6(x) << L ) ,and we may write vapor pressure-temperature relationship to be smoothed
the simplified x-momentum equation as adequately by
F(X) a2u/ay2 = --fi2x sin2 4 (5) Pa = exp(A - B / T ) (18)
with the boundary conditions Further Development and Solution of Model
u=O at y = O (6) The model equations presented in the previous section
may be rewritten in dimensionless forms. We define the
following parameters:
1018 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988

Reynolds number, Re =
602Q sin2 4 d{
_.- - - T d ‘ ( l - C + CS) - -2{ (36)
UO
dX C2ReXS X
dimensionless initial film thickness, X = b0/L (20) with { = 1at X = Xo,the feed location. Here the following
dimensionless parameters have been used:
dimensionless local film thickness, { = 6 ( ~ ) / 6 ~ (21)
s=-(-)
Pa Ma ‘I2
dimensionless viscosity, 9 = ij/uO (22)
Pb Mb
Here 6, is film thickness at the feed location, uo is viscosity = XaMf2 (37)
of the feed, and L is radius of the evaporator.
We also define the following dimenionless variables: where &b is the relative volatility of an ideal liquid solution
of a and b, and M is the ratio of their molecular weights.
Also,

x =x/L
and where 8, = T(x,6)/Tw, and B’ = B/T,. B comes from the
vapor pressure expression for the more volatile component
y = Y/60 ( a ) ,eq 18. C is the ratio of mole fractions given by
The x-momentum balance equation (eq 5) can now be
written in dimensionless form as (39)

a2U/aY2= - R e X / q (27) r is a measure of the ratio of maximum evaporation rate

of the more volatile component (ca = 1 in eq 17) to total
with the boundary conditions
feed rate and is termed the evaporation number, a function
U=O at Y=O (28) of temperature:
and r = -Ea (40)
aU/aY=o at Y={ (29) PQ60

The solution of eq 27, with 9 constant (at its average value) The energy balance equation (eq 12) can be written in
over 0 5 Y 5 5; is dimensionless form by defining
0 = T/T, (41)
(30) and a Nusselt number (the ratio of convective to con-
ductive heat flux)
and the mean U velocity a t any X is
620
NU = -
LY

Then eq 12 may be written as

The continuity equation (eq 1)provides a solution for The boundary conditions are
velocity (V):
e = eo
= To/T, at X = X o (44)
0 = 1 at Y = O (45)
where {’ is the X derivative of film thickness, {. and
Next, we define a dimensionless feed rate, F , as
-1 ae
F=
fo
(33)
Xo2L3poRsin 4
where fo is the feed rate (g/s), X o is the feed location, and
p o is the density of the feed (g/cm3). A mass balance at
the feed location is Here, 0 is a substance parameter (the dimensionless en-
tropy of vaporization a t the rotor temperature) given by
f o = (QUX06,pL2)(2asin 4) (34)
P = AHv,a/CpaTw (47)
from which we may determine that the initial film thick-
ness, eq 20, is and

A=- 3F
(35)
27rRe
A differential equation for local film thickness, { ( X ), is Heat capacities of pure liquids a and b at T, are, respec-
obtained by a similar mass balance: tively, Cpaand Cpb.
 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988 1019
Condenser distillation chamber. Many important features of the still
To vacuum pump
are similar to those of the CMS-5 still described by
Hickman (1944) and more recently by Greenberg (1972)
f H Rotating conical
evaporator ~E;oollng COIIS
and Rees (1980). The differences include a controller for
the evaporator heater to maintain a constant residue
temperature, isolation of the distillation chamber from the
feed tank, and a flat, steel condenser with cooling coils for
improved condensation. We added thermocouples to ac-
Heating curately measure feed and residue temperatures. (Know-
Residue
COllS ing the feed and the residue temperatures allows us to
calculate the rotor wall temperature by a trial-and-error
Figure 2. Sketch of the distillation chamber in the experimental method described later.) The evaporator size is 6.5 cm
system. from the center to periphery. It rotates at a speed of 1450
rpm. The distance from evaporator to condenser is 10.5
Finally, the mass balance equation (eq 8) is written in cm. Condenser diameter is 25.4 cm. The vacuum system
dimensionless form by defining has both mechanical and diffusion pumps and is capable
of maintaining a pressure of well under 1dyn/cm2for cold,
(49) dry conditions, but the value measured at the entrance to
the vacuum line during distillation may be as high as 7-10
and a Sherwood number (the ratio of convective to diffu- dyn/cm2, presumably because of slight leakage when the
sive mass flux) system is hot.
Primary parameters varied during distillations were the
residue temperatures, feed concentration, the rate of
The eq 8 may be written as coolant flow to condenser, and the feed rate. Rates of
distillations were measured as the volumetric rate (cm3/
min) at which condensate was collected during steady
state. Feed and distillate compositions were determined
by using a gas chromatograph.
The boundary conditions are
A = 1 at X = X o (52)
Calculated Results
In order to limit the number of parameters studied, we
aa/au=o a t Y = O (53) used the following constants in our computations. The
fictitious material parameters are similar to those of a
solution of 2,2-diethylhexyl phthalate (EHP) and 2,2-di-
ethylhexyl sebacate (EHS). (The relative volatility was
assumed to be a constant, however.) The equipment pa-
cShI’( C - A( C - 9)) at Y = < (54)
rameters are based on a CVC-LAB5 still: density, p = 1.0
g/cm3; relative volatility, X a b = 3.0; substance parameter,
p = 0.4; ratio of p’s, H = 1.0; ratio of molecular weights,
The equations of change in our system, represented by M = 0.9; vapor pressure parameter, B’ = 30.0 (Perry and
eq 1, 30, 36, 43, and 51, are coupled: Weber, 1949); viscosity parameter, Bu/T, = 10.326
The solution for velocities U and V is obtained if the (Mellan, 1963), where Bu comes from the expression for
average kinematic viscosity ( 7 ) and the film thickness (0 kinematic viscosity, = exp(Au + B u / T ) ;feed concen-
ZJ

at every X are known. The film thickness is influenced tration, CaO= 0.5 mass fraction; feed temperature, eo =
by the mass flux out of the film, a function of temperature, 0.8; initial film thickness, X = 0.008; rotor temperature, T,
and concentration gradients. = 430 K (except in Figure 9, where it is 410 K); evaporator
The solution for temperature and concentration profiles radius, L = 6.5 cm; rotor cone angle, 24 = 165O; rotor speed,
requires the knowledge of velocity profiles which determine Q = 160 rad/s.
the convective heat and mass fluxes. The remaining parameters are as noted on each figure.
The solution is accomplished by a numerical technique. The representative values of some of the variables used
A finite-difference method is suitable. The solution me- are as follows: feed rate, fo = 1.6 g/s; thermal diffusivity,
thod (Bhandarkar, 1985) involves the following steps: a! = cm2/s; mass diffusivity, Dab = IO+ cm2/s.
1. Divide the interval 0 i X i 1into 1000 intervals. At Figure 3a shows temperature profiles. There are sig-
each value of X , determine the velocities, U and V, as- nificant gradients with respect to both X and Y for the
suming that the value for temperature- and concentra- inner half of the evaporator, after which the liquid is closer
tion-dependent average viscosity is the same as (or a to the rotor temperature, T,,at all points. Near the feed
suitable extrapolation of) that at ( X - A X ) , where AX is point, as illustrated by the curve for X = 0.1, near-surface
1/1000. temperature may drop below the feed temperature owing
2. Use these velocities to determine the film thickness to evaporative cooling.
a t X , by using eq 36. Concentrations, as shown in Figure 3b, vary throughout
3. Determine the temperature and the concentration the liquid. Depletion of the more-volatile component by
profiles a t X by using eq 43 and 51, respectively. about 6% of the inlet concentration occurs at the heater
4. Improve the solution by repeating steps 1-3, this time surface for the conditions shown. The undistilled residue
using viscosity values updated for local temperature and (at X = 1.0) has lost about 17% of its original content of
concentration. more-volatile component.
Film-surface conditions are illustrated by Figure 4. We
Experimental Work see that as feed rate is changed from F = 0.01 to 0.02, a
Distillations were carried out in a CVC-LAB5 centrifugal larger fraction of the evaporator radius is used to heat the
molecular still. Figure 2 shows a schematic sketch of the surface liquid and to reduce its concentration of more-
1020 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988

'\

I EVAPORATION NO (AT ROTOR TEMPERATURE:, r = 0 0 2

EVAPORATION NO (AT ROTOR TEMP), r = o oi
OT'IER PARAMETERS ARE AS SHOWN ON FIGURE 5
0
h OTHER PARAMETERS ARE AS SYOWN CN FIGURE 5 '"I
00 02 04 06 08 10 0.0 0.2 0.4 0.6 0.8 1.0
DISTAKCE FROM EVAPORATOR SURFACE, Y/C DISTANCE FROM EVAPORATOR CENTER, X

=EE3 RATE

...... F = 002

F = 0 015

EVAPORATION NO. (AT ROTOR TEMP)), r = c c: 1 EVAPORATION NO (AT ROTOR TEMP), = 0 02

'
~

OTHER PARAMETERS ARE AS SFOWN ON FIGURE 5

OTHER PARAMETERS ARE 48 SH?WN ON '[GLIRE 5
:i
00 02 04 06 08 10 00 02 04 0.6 0.8 10
DISTANCE FROM EVAPORATOR SURFACE, Y/< DISTANCE FROM EVAPORATOR CENTER, X
Figure 3. (a, top) Temperature profiles in film a t different axial 'l-\
0

positions. (b, bottom) Concentration profiles in film at different .........

\
axial positions. ..-.__
-._._
- _- _-.
.._
-..-._
volatile component. Evaporation number, the parameter ........
of Figure 4c, emphasizes the evaporation rate of the <2
more-volatile component as a fraction of liquid feed rate.
The profiles of Figures 3 and 4 relate the measure of
separation and the rate of distillation. Figure 5 illustrates
the effect of increasing distillation rates (for a constant feed \
rate the evaporation number (at rotor temperature), I', is \
proportional to distillation rate) upon the measure of
EVAPORATION 110 \
(AT ROTOR TEMP)
'\
separation, the selectivity, defined as
\
(5)
0 034

(L)
....... F=

selectivity = (55) r = 0.01 \

- condensate feed __ i=O02
\
\'

In Figure 5 and subsequently we utilize the quotient of 1 T H E R PARAMETERS ARE AS SH9WN O N FIGURE 5
selectivity and relative volatility which is assumed to be !
a constant (=3.0).This quotient is limited in its maximum
value by the square root of the ratio of molecular weights, 00 02 04 06 08 10

( M b / M , ) 1 / (Bhandarkar
2 and Ferron, 1986). The undis- DISTANCE FROM EVAPORATOR CENTER, X
tilled fraction of feed shown in the figure is taken to be
the ratio of residue flow rate (g/s) to feed rate (g/s). Figure 4. (a, top) Film surface temperature profiles at different feed
Experimental values of selectivity are obtained from rates. (b, middle) Film surface concentration profiles at different
measured values of c, in the feed and condensate. Se- feed rates. (c, bottom) Film surface concentration profiles at dif-
lectivities predicted by the theory are obtained from the ferent evaporation numbers.
 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988 1021
n 0

7 ?EL VOLATILITY FEED TEMP

h __ 4.0 09

In
2-

--.._
._--.
..--.
-.._

_L,
-..-..
UNDISTILLED ....
FRACTION OF FEED -..-._
-.._
- - _-._
FEED RATE, F = 0.02 -._
-...-.
51 FEED F = O;, 1 , ,,\ NUSSELT NO, Nu = 1 0 0 0
..- - _ _
REYNOLDS NO, Re = 10 REYNOLDS NO, Re = 10
NUSSELT NO, Nu = 1 0 0 0 EVAPORATION NO, r = o 02
SHERWOOD NO, Sh = 1 0 0 0 0 0 (AT ROTOR TEMP)
n
P
2
0.002 0.005 0.008 0.011 0.014 0.017 0.020 1.0 2.8 4.6 6.4 8.2 10.0
EVAPORATION NUMBER (AT ROTOR TEMP), r SHERWOOD NUMBER, Sh YO'
Figure 5. Effect of distillation rate on selectivity. Figure 7. Effect of diffusional resistance on selectivity.

a,
1(SELECTIVITY / REL. VOLATILITY)

FEED ?ATE, F = 0.02

W
NUSSELT NO, Nu = 1 0 0 0
2-
SHERWOOD NO, Sh = 1 0 0 0 0 0

'i
FEED RATE, F = 0.01
EVAPORATION NO, r = 0.01

4\ (AT ROTOR TEMP) SHERWOOD NO, Sh = 1 0 0 0 0 0

REYNOLDS NO, ?e = 10
EVAPORATION NO,
(AT ROTO? TEMP)
r = o 01

UNDISTILLED FRACTION
/ OF FEED

00 40 80 12 0 16 0 20 0 700.0 1obo.o 1300.0 l6bO.O

REYNOLDS NUMBER, R e NUSSELT NUMBER, N u
Figure 6. Effect of viscous resistance in binary distillation. Figure 8. Effect of thermal diffusivity in binary distillation.

assumed feed composition (eq 9) and the condensate is not much (Figure 8). Increasing the thermal resistance
composition derived from the total evaporated amounts (as Nu increases) reduces the rate of distillation because
of components a and b in a unit time (eq 17). The se- the surface cooling effect is more significant. The im-
lectivity decreases as the distillation rate increases pri- portance of the heat- and mass-transfer resistances was
marily because the more volatile component (a, in this also pointed out by Bose and Palmer (1984) who modeled
case) is depleted faster a t the liquid surface. Should the unsteady-state evaporation from tensimeter jets. Ruck-
diffusional resistance (assumed constant in the present enstein et al. (1983) found the diffusional effects to be
illustration) increase owing to, for instance, an increase in negligible in their analysis of molecular stills when they
the liquid viscosity, then the decrease in selectivity is more used diffusivities about lo00 times larger than those used
pronounced as the rate of distillation increases. The in- in our calculations.
crease in viscosity may result from either high surface Computations were carried out to study the effect of
cooling or concentration changes. Varying viscous re- various degrees of preheating of the feed. A dimensionless
sistance does not affect the separation efficiency to any feed temperature, eo,of 0.9 corresponds to an actual feed
great extent (see Figure 6) if the diffusivities are kept temperature of 369 K when the rotor surface temperature
constant. However, the rate of distillation increases with (kept constant) is 410 K in Figure 9. Similarly eo= 1.1
decreasing viscosity (Re going from 1to 20), because the corresponds to a feed temperature of 451 K, i.e., 41 deg
liquid spreads out in a thinner f i i , allowing more efficient above the rotor temperature. The selectivity shows a
heat and mass transport. decrease with the increase in eo. The decrease is more
Figure 7 shows the effect of increasing diffusional re- marked when eois greater than unity, with the attendant
sistance (as Sh increases, diffusivity decreases in the sharper increase in distillation rates. These trends should
present illustration), for two different values of relative be useful in determining an optimum eo, in order to bal-
volatilities. The effects are significant. By the same token, ance the conflicting goals of high rate and high degree of
the effect of varying thermal diffusivity upon selectivity separation. It was noted that the residue temperature was
1022 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988

..
0
0
Table 11. Experimental and Calculated (Langmuir) Rates
and Selectivities for Distillation of Ethylhexyl Phthalate
and Ethylhexyl Sebacate at Partial Vacuum"
0
distillation
: feed (EHP) rate
temp, "C concn, mole cm3/cm2/9 selectivity
feed residue fract. calcd exptl calcd exptl
34.1 107.1 0.497 7.2 3.5 4.27 4.62
36.2 111.6 0.496 10.5 5.3 4.15 4.32
I FEED RATE, F = 0 02 31.6 115.3 0.489 14.2 8.0 4.05 4.28
35.1 127.5 0.483 37.4 16.7 3.70 3.77
SHERWOOD NO, Sh = 1 0 0 0 0 0
35.5 137.2 0.464 75.4 33.9 3.36 3.25
REYNOLDS NO, Re = 10 42.1 145.0 0.491 134.6 64.9 3.02 2.87
NUSSELT NO, Nu = '000
Measured operating pressure = 7-10 dyn/cm2. Feed rate = 0.5
EVAP NUMBER, r = o 01 UNDISTILLED
cm3/s, rotor speed = 1450 rpm.
(AT ROTOR TEMP)

P I
$1 -
.
Glycerol, Feed Rate = 0 42 gm/s
Di-buty; Phthalate Feed Rate = 0 19 gn/s
Present Analysis
0 90 OS4 0198 1.02 106 110
DIMENSIONLESS FEED TEMPERATURE, Bo
Figure 9. Effect of feed temperature in binary distillation.

Jwaporation of Etnyinexyi rhthaiate at rartiai vacuum"

evaporation rate X
temp, "C io5, cm3/cm2/s
feed residue calcd exptl
36.2 107.6 11.8 5.8
1179 27 7 12 Sn
37.0 127.9 61.0 26.7 Data points are experimental
34.7 131.6 80.7 35.8 results of Greenberg (1972)
37.6 135.4 106.5 48.5
35.6 137.2 121.8 51.2 I
Measured operating pressure = 7-10 dyn/cm2. Feed rate = 0.5
a
40.0 50.0 800 70.0 800 900 1600
cm3/s, rotor speed = 1450 rpm. MEAN DISTILLAND TEMPERATURE [DEG C]
Figure 10. Comparison of theory with the experimental evaporative
nearly constant regardless of the extent of preheating. This fluxes of pure liquids.
has implications for the control of the process in com-
mercial equipment. condenser. Recent and current work to accomplish this
Calculations involving variations in the cone angle involves the solution of Boltzmann's equation describing
(155O-18Oo) revealed that the changes in angle did not have vapor-phase dynamics by moment methods (Ferron, 1986)
a significant effect upon either rates or separation. In- and by the Monte Carlo simulations (Bhandarkar and
creasing cone angle aids in the formation of a thinner film. Ferron, 1986; Bhandarkar, 1988). In addition, the con-
This increases rates of evaporation slightly. From a denser behavior is studied by Bhandarkar (1988).
practical point of view, with too large an angle the film The experimental data available in the literature
tends to detach itself from the rotor surface and it is not (Greenberg, 1972; Kaplon et al., 1986) involve the evapo-
desirable. ration of a pure component. Langmuir rates may be as-
In order to compare our present analysis with experi- sumed to apply because background presmres are low (less
mental data, we need to assure ourselves that the distil- than 3 dyn/cm2). Since Greenberg's (1972) data for gly-
lation process is not affected by the processes in the vapor cerol could not be predicted by his approximate theory,
phase. For instance, when collisions in the vapor phase we chose these for comparison. Figure 10 shows that his
(among distilling molecules and among noncondensable data agree well with our analysis. The results for any
gases present) are significant, the net rate of evaporation substance were most sensitive to vapor pressures as shown
is less than that given by eq 17. These collisions can be by Figures 5-8. In the case of glycerol we used vapor
very important even when the measured condenser pres- pressurs given by Stull (1947). Viscosity a t the local tem-
sures are not more than 10 dyn/cm2. Tables I and I1 show peratures, and density, thermal conducitivity, and latent
the results of experiments under vacuum that may be heat of vaporization a t 20 "C were obtained from Weast
expected in commercial distillations when it is neither (1983). Variations of f10% in density, conductivity, or
practical nor economical to maintain very high vacuum. heat of vaporization produced variations of about rtl% or
The data were obtained for operating pressures ranging less in the evaporative fluxes of glycerol. We have also
from 7 to 10 dyn/cm2. It is observed that if effects of the shown the agreement of our analysis with Greenberg's
vapor phase are not included, the present analysis with the (1972) data for dibutyl phthalate (vapor pressures from
assumption of the Langmuir rate of evaporation (eq 17) Perry and Weber (1949)). In these examples it was as-
predicts higher rates, about twice the experimental. The sumed that the mean feed and the residue temperatures
selectivities in experiments are also higher than those were equal. This involved a trial-and-error calculation.
predicted a t the lower temperatures in Table 11. These First an assumption was made for the rotor temperature.
differences clearly point out the need for extending our The computations were carried out as described in an
analysis to include the study of vapor processes and of the earlier section. The mean residue temperature resulting
 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988 1023
0
the film thickness and exposure times should form the
1' Evaporation of Di-butyl Phthalate bases for scaling up the process. While it is more
straightforward to have equal values for the various di-
mensionless parameters given in the present paper, for the
small and the scaled-up process, such an approach is highly
restrictive and leads to unrealistic feed rate and rotor
speed. We will follow a heuristic approach.
We stipulate that for the smaller (subscript 1)and the
larger stills (subscript 2), at the evaporator periphery, the
following conditions hold:
z l
9
b
<$
Gi
Feed Temperature = 352.0 K u velocities are equal u1 = u2 (56)
2 Feed Temperature = 348.2 K

- Present Analysis
film thicknesses are equal a1 = a2 (57)
x-
*I

The first condition requires that the exposure times be

roughly in the same proportion as the evaporator sizes, and
Data points ore experimental
D results of Kaplon et oL(1986)
the second one ensures that the film thickness is not zero
anywhere on the evaporator, thus precluding the danger
4.0 8.0 12.0 16.0 20.0 of thermal hot spots.
FEED RATE [GM/S]
Assuming (for simplicity) the same rotor cone angle for
Figure 11. Comparison of theory with the experimental rate of
evaporation from a rotating disk. the larger and the smaller stills, and by use of the above
conditions (eq 56 and 57), eq 31 results in
Table 111. Results of the Scale-up Procedure for
Evaporation of Glycerola
rotor evaporation rate Next, the film thickness (at the rotor periphery) is related
feed rate, speed, simulat- predict- (see, eq 34) to the mass rate of flow, rotor size, and rotor
still size. cm P/S rvm ed. e/s ed. e/s speed as follows:
Reference Still
6.5 0.75 1600 0.053
Scaled Up Operations OC
(59)
15 1.88 1050 0.28 0.28
30 4.39 750 1.13 1.14
45 7.23 600 2.53 2.56 Here the proportionality constant is a function of the
properties of materials being distilled and the tempera-
"Feed temperature = 60 "C,rotor temperature = 70 "C.
tures. For the same materials and the same average tem-
from calculations was checked for equality with the feed peratures, eq 57 and 59 together give
temperature. If not equal, the rotor temperature was al-
tered, and the computations were repeated until the de-
sired agreement was achieved.
The data of Kaplon et al. (1986) provide an opportunity
to test our analyses for an insulated disk rotating a t low Expecting eq 58 and 60 to give similar surface temperature
speeds. For this the cone half-angle (4) was set to 90°, the profiles in the smaller and the larger operations, we may
feed temperature (0,) was taken to be unity, and the say that average evaporative fluxes are equal:
suitable boundary condition (eq 15) was employed. Figure
11shows results for two of the runs of Kaplon et al. (1986). Q i / L i 2 = Q2/Lz2 (61)
The simplified analysis of heat transfer given by Kaplon
et al. (1986) tends to overestimate evaporation rates. For Here Q is the amount distilled, i.e., Q = f o - f . Combining
instance, for the evaporation of glycerol (feed rate = 8.0 eq 60 and 61, we obtain
g/s, rotational speed = 1650 rpm, feed temperature = 51
"C, and rotor temperature = 96 "C), Kaplon's simplified
model overestimated evaporation rate by about 18%.
A Scale-up Procedure
Thus, if the speed of the larger rotor is determined ac-
As a first step toward predicting distillation rates and cording to eq 58 and the feed rate is determined according
efficiencies for industry-scale equipment (with L as large to eq 62, we may predict the total evaporation rates using
as 0.5 m), the computations were carried out for a wide
range of equipment and operating parameters. It was eq 61. In the case of a mixture, eq 61 may be used to
observed that with constant feed and rotor temperatures predict the distillation rates if Q is taken as the distillative
the average liquid-surface temperatures were lower for flux of a component.
larger evaporators. This is obviously due to the increased Tables I11 and IV show results obtained by using the
initial film thickness and longer exposure times for the scale-up procedure described above. The agreement be-
liquid element (the role of thermal resistance is mentioned tween predicted and actual simulation results shows the
in the previous section). It was also observed that the usefulness of the simple approach. The scale-up equations
variation of film thickness across the evaporator has a tend to slightly overestimate the liquid-surface tempera-
bearing on the surface temperature profile. I t was con- tures. This may actually be desirable for a safer operation,
cluded that, for safe operation without thermal hazards, since the danger of thermal hazard is reduced.
1024 Ind. Eng. Chem. Res., Vol. 27, No. 6, 1988

Table IV. Results of the Scale-up Procedure for uo = u velocity at feed location
Distillation of a Mixture of n -0ctyl Phthalate (NOP)and U = dimensionless velocity in X direction
Isooctyl SebacateO v = velocity in y direction
distillation rate (of V = dimensionless velocity in Y direction
component NOP) x = distance along rotor surface
rotor x o = feed location
feed rate, speed, simulat- predict-
still size, cm g/s rpm ed, d s ed, d s
X = dimensionless distance along rotor surface
y = distance perpendicular to rotor surface
Reference Still Y = dimensionless distance perpendicular to rotor surface
6.5 1.00 1600 0.022
Scaled Up Operations Greek Symbols
15 3.46 1050 1.16 1.20 a = thermal diffusivity
30 11.08 750 4.54 4.79 3( = substance parameter (eq 47)
45 22.61 600 10.03 10.79 r = evaporation number (eq 40)
n-Octyl phthalate/isooctyl sebacate feed concentration = 0.5 6 = local film thickness
mole fraction of NOP. Feed temperature = 160 'C, rotor temper- So = initial film thickness (at x = xo)
ature = 164 "C. AHv = heat of vaporization
6 = quantity defined by eq 38
Acknowledgment +t= dimensionless film thickness (eq 21)
The authors thank Distillation Products Industries of
r = X derivative of film thickness
9 = dimensionless viscosity (eq 22)
the Eastman Kodak Co. and CVC Products, Inc., both of 9 = dimensionless temperature (eq 41)
Rochester, for the loan of molecular stills and for other 8, = dimensionless surface temperature
valuable assistance. Generous computing time provided 8, = dimensionless feed temperature
by the Department of Chemical Engineering, University X = dimensionless initial film thickness (eq 20)
of Rochester, is also gratefully acknowledged. A = dimensionless concentration (eq 49)
v = kinematic viscosity of liquid
Nomenclature p = density of liquid
a = component (more volatile) in the binary mixture 4 = cone half-angle (Figure 1)
b = component in the binary mixture X & = relative volatility of a and b
c, = mole fraction of component a in mixture = rotor speed, rad/s
C = ratio of mole fractions (eq 39)
C, = mass fraction of component a in mixture Literature Cited
C, = heat capacity Bhandarkar, M. Master's Thesis, University of Rochester, Rochester,
Dab = mass diffusivity in the binary mixture NY, 1985.
E , = rate of evaporation of a (eq 17) Bhandarkar, M. Ph.D. Thesis (In Progress), University of Rochester,
Eb = rate of evaporation of b Rochester, NY, 1988.
f = mass flow rate on the evaporator, g/s Bhandarkar, M.; Ferron, J. R. In Modeling and Simulation in En-
fo = feed rate, g/s gineering; Wahlstrom, B., Leiviska, K., Eds.; Elsevier Science:
F = dimensionless feed rate (eq 33) Amsterdam, North Holland, 1986; pp 153-160.
H = quantity defined by eq 48 Bose, A.; Palmer, H. J. Ind. Eng. Chem. Fundam. 1984, 23, 459.
Ferron, J. R. Ind. Eng. Chem. Fundam. 1986,25, 594.
kab = thermal conductivity in the mixture Greenberg, D. B. AIChE J . 1972, 18, 269.
L = size (radius) of the evaporator Hickman, K. C. D. Chem. Rev. 1944, 34, 51.
M = ratio of molecular weights of a and b Kaplon, J.; Kawala, Z.; Skoczylas, A. Chem. Eng. Sei. 1986,41,519.
Ma = molecular weight of component a Langmuir, I. Phys. Reu. Ser. 2 1913, 2, 329.
Mb = molecular weight of component b Mellan, I. Industrial Plasticizers; McMillan: New York, 1963.
Nu = Nusselt number (eq 42) Perry, E. S.; Weber, W. H. J . Am. Chem. SOC. 1949, 71, 3726.
Pa = vapor pressure of component a Rees, G. J. Chem. Eng. Sci. 1980, 35, 837.
Q = rate of distillation, g/s Ruckenstein, E.; Hassink, W. J.; Gaurisankar, S. V. Sep. Sci. Tech-
R = gas constant nol. 1983, 18, 523.
Stull, D. R. Ind. Eng. Chem. 1947, 39, 517.
Re = Reynolds number (eq 19) Weast, R. C., Ed. CRC Handbook of Chemistry and Physics, 64th
S = quantity proportional to X a b (eq 37) ed.; CRC: Boca Raton, FL, 1983.
Sh = Sherwood number (eq 50)
T = temperature Received for review June 16, 1987
T, = evaporator (wall) temperature Revised manuscript received October 19, 1987
u = velocity in x direction Accepted January 20, 1988