Molecular

Copyright Distillation
© 2002 by Humana Press Inc. 1187
All rights of any nature whatsoever reserved.
0273-2289/02/98-100/1187/$15.00

Molecular Distillation
Rigorous Modeling and Simulation
for Recovering Vitamin E from Vegetal Oils

C. B. BATISTELLA, E. B. MORAES,
R. MACIEL FILHO, AND M. R. WOLF MACIEL*
Separation Process Development Laboratory (LDPS).
Faculty of Chemical Engineering, State University of Campinas,
(UNICAMP), CP 6066, 13081-970, Campinas-SP, Brazil.
E-mails: cesarb@ncap.com.br and wolf@feq.unicamp.br

Abstract
In this work, important results from simulations are presented, showing
the potentiality of the molecular distillation process for recovering vitamin
E from vegetal oils. Two types of molecular distillators are considered: fall-
ing film and centrifugal. The results emphasize the degree of recovery and
factors that influence substantially the performance of the molecular
distillators, such as feed flow rate, residence time, and process temperature.
Moreover, they show that each type of molecular distillator enables one to
operate under specific residence time and temperature. Therefore, a careful
analysis must be made in order to determine the best equipment and oper-
ating conditions for obtaining products with high quality and concentration,
and reduced problems of material thermal decomposition. Vitamin E (toco-
pherols) from vegetal oils, more specifically, from the deodorizer distillate of
soya oil, was the studied case.

Index Entries: Molecular distillation; vegetal oils; vitamin E; soya oil.

Introduction
The substitution of conventional materials used in the nutrition,
pharmaceutical, and cosmetic areas by natural products has gained interest
and importance. An example is the product derived from refined vegetable
oils, as the deodorizer distillate of vegetable oils (DDVO); palm oil for
obtaining provitamin A (1,2), oils of rice for the oryzanol recovery (3),
among others, can also be cited.

*Author to whom all correspondence and reprint requests should be addressed.

Applied Biochemistry and Biotechnology 1187 Vols. 98–100, 2002

1188 Batistella et al.
Molecular distillation is a peculiar case of evaporation, which hap-
pens in extremely low pressures and, therefore, in low temperatures (4).
Another important characteristic of molecular distillation is the reduced
time in which the material is submitted to the process temperature, usually
from 1 s to 1 min. This reduces, considerably, the effects of thermal decom-
position. In a general way, these peculiarities show the high potential of
this process in the separation, purification, and/or concentration of natural
products, usually constituted by complex and thermally sensitive mol-
ecules (1). Furthermore, this process can take advantage of other techniques
that use solvents as the separating agent, avoiding problems with toxicity.
Taking all of this into consideration, in this work, the molecular dis-
tillation process for recovering tocopherols (for obtaining vitamin E), using
the deodorizer distillate of soya oil (DDSO) with the conditions and the
quality demanded by the market, is being proposed.
Molecular distillation is a process that requires a great deal of knowl-
edge of its performance before it can be carried out. Small variations in the
process conditions can lead to considerable alterations in the characteris-
tics of the product streams. Therefore, it is very important to have simula-
tion results first, to guide the experimental work (5). Modeling and
simulations of the process were then developed for recovering tocopherols
(vitamin E) from DDSO, in order to determine the feasibility of the process
and the best experimental conditions.
Tocopherols are a mixture of α-, β -, γ -, δ -tocopherols, which can be
found in several proportions and concentrations, and they are present in
vegetable oils such as soy, sunflower, canola, cotton, corn, palm, and rice
(6). They are particularly present in large amounts in the deodorizer distil-
late of soya oil (by product of the production of the vegetable oil) (7). The
deodorizer distillate is composed, basically, of fatty acids, triglycerides,
diglycerides, monoglycerides, hydrocarbon, terpenols, and other materi-
als, besides tocopherols and fitosterols.

Liquid Phase Mathematical Modeling

The mathematical modeling of molecular distillation is developed
in two parts: first, the equations governing the liquid phase are consid-
ered, and, then the equations representing the vapor phase. The final
modeling considering both phases permits through simulation the deter-
mination of important variables representing the system in study
throughout the process.

Falling Film Equipment

A typical apparatus for the falling film molecular distillation is shown
in Fig. 1. The materials flow throughout the equipment and the geometric
coordinates are shown schematically in Fig. 2 (8). The main part of the
installation consists of a cylindrical evaporator surrounded by a condenser
jacket. The liquid to be distilled is transported from a storage tank through

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1189

Fig. 1. Falling film molecular distillator.

Fig. 2. The coordinate system and velocity distribution for the falling film.

a preheater to the surface of the unheated evaporator. It is also possible to

heat the evaporator internally. In the case of a heated evaporator, the dis-
tillation rate is faster, but the separation factor decreases at higher tempera-
tures. Figure 2 presents the velocity distribution in a film flowing down the
evaporator.
Surface Evaporation Rate (High Vacuum)
The rate of surface evaporation, Ei (kg/m2s), is obtained from the
kinetic theory of gases, taking into account the anisotropic properties of
the vapor [number of collisions is too low (9)]. Then, the evaporating rate
equation is given by (8):
1 n
sat Mi 2 –h
Ei = CiSPi 1– 1–F 1–e kβ (1)
2 π RgTS

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1190 Batistella et al.
where CiS is the mole fraction in the liquid surface, Pisat is the vapor pressure
(Pa), Mi is the molecular weight (g/gmol), Rg is the gas constant (J/kmol·K),
TS is the surface temperature of the liquid phase (K), and F, n, k, h, and β are
constants, given by (8).
Velocity and Thickness Profiles
The velocity is the one corresponding to a laminar and isothermal film
with a smooth surface. Under such conditions, the velocity profile in the
film is given by (8):
2
gρ 2 r – R 1 r – R
WZ = S – (2)
η S 2 S
where Wz is the axial velocity (m/s), ρ is the mean liquid density (kg/m3),
η is the viscosity (Pa·s), and S is the thickness of the liquid film on the
evaporator.
The feed flow rate and the continuity equations for the evaporating
film give the film thickness:
1
z 3
m0
S = 3η – 1 ∫ Σ Ei dz (3)
2 2
2 πRgρ g ρ z0
where S is the film thickness (m), m0 is the feed flow rate (kg/s), and dz is
the integration step (m).
Temperature Profile
The temperature in the liquid, T (K), obeys the Fourier–Kirchhoff
equation, which has the following form in cylindrical coordinates for sta-
tionary heat flux (solely vertical liquid flow and negligible axial heat trans-
fer) (8):
2
∂T ∂T ∂ T
Wz = α 1 + (4)
∂z r ∂r
∂T
2

where α is the thermal diffusivity (m2/s).

The initial and boundary conditions are:
1) T = T0 for z = 0 and R ≤ r ≤ R + S (5)

∂T
2) = 0 (in the condition of insulated evaporator) (6)
∂r
or I = Iw (in the condition of heated evaporator) (7)
for r = R and 0 ≤ z ≤ L

vap
∂T Σ Ei∆H i
(3) =– for r = R + S and 0 ≤ z ≤ L (8)
∂r λ
Applied Biochemistry and Biotechnology Vols. 98–100, 2002
Molecular Distillation 1191
where Tw is the temperature on the evaporator surface (K), ∆Hivap is the
evaporation enthalpy (J/kg), and λ is the thermal conductivity (W/mK).
Concentration Profile
The concentration profile, Ci , in the liquid layer for a multicomponent
mixture can be expressed by the following equation, which is valid for
negligible radial and azimuthal flows and negligible axial diffusion (8):
2
∂Ci ∂Ci ∂ Ci
Wz = Di 1 + (9)
∂z r ∂r ∂r
2

where Di is the mass difusivity (m2/s).

The initial and the boundary conditions are:
(1) Ci = Ci0 for z = 0 and R ≤ r ≤ R + S (10)

∂Ci
(2) = 0 for r = R and 0 ≤ z ≤ L (11)
∂r

n Ej
Ei – Ci Mi ∑
∂Ci j=1 Mj
(3) =– for r = R + S and 0 ≤ z ≤ L (12)
∂r ρD i
Using condition 3, we can guarantee that the mass balance is com-
pletely satisfied.
These equations are modified to a convenient form to apply the finite
differences method and solved by an implicit finite differences method
(11). The film thickness was divided in 200 equal intervals, and the evapo-
rator length, L, was divided into 100 equal intervals. For each value of z,
the value of r will vary from r = R (evaporator wall) to r = R + S (evaporation
surface). So, the system of equations is solved in the following way:
1. Solution of equation 1.
2. Solution of equation 3.
3. Solution of equation 2.
4. Solution of the equation system through finite differences generated
by equations 4 to 8.
5. Solution of the equation system through finite differences generated
by equations 9 to 12.
6. Increase the value of r with ∆r.
7. Return to step 4 until r = R + S.
8. Increase the value of z with ∆z.
9. Return to step 1 until z = L.
10. End.
The mathematical method considered is highly stable and the results
were convergent after using sufficiently reduced integration step. In this
Applied Biochemistry and Biotechnology Vols. 98–100, 2002
1192 Batistella et al.

Fig. 3. Centrifugal molecular distillator.

Fig. 4. Coordinate system of the evaporator.

case, ∆r was equal to S/200 and ∆Z equal to L/100. The solution was con-
sidered correct after the mass global balance reach values above 99.5%.

Centrifugal Equipment
A typical apparatus for centrifugal molecular distillation is shown in
Fig. 3 (conventional scheme shown in the literature). The material fluxes
through the apparatus are shown schematically in Fig. 4 (12). The liquid
to be distilled is heated until the feed temperature is reached and goes to
the evaporator center through pumping. Therefore, the liquid by cen-
trifugal force is uniformly spread on the evaporator until its border is a
thin film, where it is partially vaporized, and which vapor is condensed
in the condenser.

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1193
Surface Evaporation Rate under High Vacuum
In the present analysis, it is assumed that the evaporation rates at the
liquid surface are given by the Langmuir equation:
1
sat Mi 2
Ei = Ci SPi (13)
2 πRgTS

where the nomenclatures are analogous to the falling film distillator.

Velocity and Thickness Profiles
The continuity equation for the liquid flowing as a thin film in the
coordinate system, as illustrated in Fig. 4, is given by (12)
∂u ∂v u v cot φ
+ + – =0 (14)
∂x ∂y x x
where u and v are the velocity profiles in the directions x and y (axial and
diagonal coordinates, respectively) (m/s) and φ is the half angle of the
evaporator.
The boundary conditions are
u = u0 at x = x0 (feed) (15)
and v = 0 at y = 0 . (16)
The film thickness S(X)(m) is given by
1
x n 2
∑ ∑
3
m0 – Ei π∆x sin φ 2x + ∆x
x = x0 i = 1
S(x) = (17)
2 2
πρ Ω xsin φ
2 x sinφ – Scosφ
3µ

where µ is the cinematic viscosity (m2/s), Ω is the angular velocity of the

evaporator (rad/s), and n is the total number of components.
Temperature Profile
The energy balance equation gives the temperature profile, T, in the
liquid film on the evaporator (12)
2
∂T ∂T ∂ T 1 ∂T cot φ ∂T
u +v =α + – (18)
∂x ∂y ∂y
2 x ∂x x ∂y

with the initial and boundary conditions

T = T0 at x = x0 (feed) (19)

T = Tw (rotor temperature) at y = 0 (20)

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1194 Batistella et al.
1 vap
∂T 2 2 Σ Ei ∆ H i
and = – 1 + (S') at y =S(x) (21)
∂r λ
where S’ is written as: S' = ∆S (22)
∆x

Concentration Profile
The equation describing the mass transfer in a multicomponent mix-
ture gives the concentration profile in the liquid film, C, and is written as (12)
2
∂Ci ∂C ∂ Ci 1 ∂Ci cot φ ∂Ci
u + v i = Di + – (23)
∂x ∂y ∂y
2 x ∂x x ∂y

with the initial and boundary conditions

Ci = Ci0 at x = x0 (24)
∂Ci
= 0 at y = 0 (25)
∂y
n
Ei
1
Ei – CiMi ∑
∂Ci 2 2 M
j=1 j
and = – 1 + (S') at y = S(x) (26)
∂r ρD i

These equations are modified to the form of finite differences and

solved by an implicit finite difference method (11). This method has shown
good stability for this application. The film thickness was divided in 200
equal intervals, and the evaporator length, L, was divided in 100 equal
intervals. For each value of x, the value of y will vary from y = 0 (evaporator
wall) to y = S (evaporation surface). Therefore, the system of equations is
solved in the following way:
1. Solution of equation 13.
2. Solution of equation 17.
3. Solution of equations 14 to 16.
4. Solution of the equation system through finite differences generated
by equations 18 to 22.
5. Solution of the equation system through finite differences generated
by equations 23 to 26.
6. Increase the value of y with ∆y.
7. Return to step 3 until y = S.
8. Increase the value of x with ∆x.
9. Return to step 1 until x = L.
10. End.
In this case, also, the implicit finite difference method also was highly
stable and the results are convergent after using a sufficiently reduced
Applied Biochemistry and Biotechnology Vols. 98–100, 2002
Molecular Distillation 1195
integration step. In this case, ∆y was equal to S/200 and ∆x equal to L/100.
The solution was considered correct after the mass global balance reach
values above 99.6%.
Therefore, these equations consider the phenomena of the liquid
phase, but do not take into consideration the vapor phase. However, at this
point, all conditions at the surface of the liquid film, such as concentration
and temperature, where the evaporation phenomena occurs, are defined.
Therefore, it is possible, now, to continue with the modeling of the vapor
phase. It can be said that the Langmuir equation does not represent the
vapor phase realistically on the molecular distillation. A vapor phase mod-
eling has to be considered for this process, as described below.

Vapor Phase Mathematical Modeling

With the conditions at the liquid evaporation surface determined, it is
possible to analyze the phenomena that occur in the vapor phase.
The surface evaporation rate on the evaporator for both equipment
(falling film and centrifugal) is given by the Langmuir equation, as already
mentioned [eq. (1)]:
1
sat Mi 2
Ei = CiSPi (27)
2 πRgTS

This expression provides exact values of the distillation rate, but only
when there is no return of the molecules from the vapor phase to the liquid
phase in the evaporator. However, collisions happen in the vapor phase
and part of the molecules come back to the evaporator. Thus, studying the
molecular dynamics of the vapor phase allows evaluating the amount of
molecules that come back to the liquid phase and, consequently, the real
deviation of the values given by equation (27). Therefore, the evaporation
efficiency is defined as the ratio between the real and the ideal values (this
one is given by the Langmuir equation). The smaller the efficiency is, the
larger will be the operating time and, therefore, the eventual problems with
thermal decomposition of the material. Thus, it is important to determine
layouts and operating conditions of the molecular distillation equipment
that maximize the evaporation efficiency.

Methodology
The real behavior of the molecules in the vapor phase under high
vacuum can be described by the Boltzmann equation (13):
∂fi n
uzi = ∑ J ij (28)
∂z j=1

This equation refers to a planar, one-dimensional and steady-state

flow, where i represents the species of the vapor mixture, fi is the distribu-
tion function and uzi is the molecular velocity (m/s) in the vapor phase of
Applied Biochemistry and Biotechnology Vols. 98–100, 2002
1196 Batistella et al.
component i in the direction z. The symbol Jij represents the integral of the
collision for the interaction between species i and j.
To avoid the complex solution of Boltzmann’s equation, Bird (14) has
applied the direct simulation Monte Carlo method (DSMC) in problems
involving the dynamics of rarefied gases. Therefore, the Monte Carlo
method was considered for the vapor phase modeling and not Boltzmann’s
equation, because the modeling requires large flexibility due to the com-
plex geometry of the molecular distillator.
The Direct Simulation Monte Carlo Method
Monte Carlo method allows the direct simulation of molecular motions
and of intermolecular collisions over small time intervals in the vapor phase.
Molecular motions are modeled deterministically, while the collisions are
treated statistically. In this method, the simulation is carried out on a bat-
tery of cells, which is obtained, by the division of the path that the molecules
will pass between the evaporator and the condenser.
The main variables, involving the simulation for this method, are the
time interval for the simulation in the vapor phase, the size of each cell, the
evaporation rate, the molecular motions, and the intermolecular collisions.
Time Interval
Molecular motions and intermolecular collisions in the vapor phase
are computed in each interval of time dt: if the vapor phase is to be simu-
lated precisely, dt must be nearly infinitesimal, necessitating a large amount
of computer time. It is convenient, then, to choose a finite value of dt, how-
ever, it should be small enough to avoid introducing excessive distortions
of properties. In ref 15 it was noted that the value of dt can be determined
by equation (29).
To evaporate a molecule, a minimum evaporation area equal to the
area of the molecular section is necessary, or πd2/4, which, multiplying by
the absolute rate equation, gives
–1
4
E πd
dt = a (29)
4
where dt is the time interval (s), d is the molecular diameter (m), and Ea is
given by equation:
1

Ea = N av P
sat 1 2
(30)
2πMmRgTS
where Ea is the absolute rate (molecules/m2 s), Nav is the Avogadro’s number,
Mm is the medium mass of a single molecule (kg), and Rg is the gas constant.
The fact is that, many times, the value determined by equation (29)
is high, inducing considerable deviations in the simulation results, mainly
in the cells where there is high collision density. Thus, it was assumed that
the best dt value will be determined when the total number of colliding
molecules inside the cell with the largest molecular density (next to the

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1197

Fig. 5. Computational cells for using in the Monte Carlo simulation.

evaporator) reaches 30 – 50% of the total number of molecules in this cell.

Simulation results showed that the dt value calculated in this manner
gave results with good precision and reasonable simulation times.
The Size of the Cells
The space between the evaporator and the condenser is divided into
cells, according to Fig. 5. In practical terms, the size of a cell must not exceed
the value of the mean free path of the molecules for a correct application of
the Monte Carlo method. It is known that the largest concentration of
molecules occurs near the evaporator. In ref. 16, it was suggested that the
size of the cell should be equal to β/3.
Evaporation Surface Area
The mean free path, β, can be obtained considering a gas in equilibrium:
Rg T S
β= (31)
1/2 2
2 πd N av P
where P is the system pressure (Pa).
It is necessary now to determine the area (plane perpendicular to the
flow of the molecules) of this cell. In practical terms, there are 100 – 200
molecules for this cell, so that the simulation is not too slow (14).
For the calculation of this area, it was considered the cell nearest to
the evaporator. Starting from the expression of Ea (absolute rate, mol-
ecule/m2s) and from the time interval dt, the areas (and therefore the vol-
umes, once the axial dimension is known) of the cells containing 100 – 200
molecules can be determined. If the total area of the evaporator is different
from the total area of the condenser, a correction factor should be intro-
duced in the volume for the areas of the subsequent cells until the last cell.
Kinetics of the Molecules (Vapor Phase)
The molecules leave the evaporator (or condenser) and present the
following velocity components (17):
Velocity (m/s) in the axial direction of the flow:
1
2kT InU1 2
u= (32)
Mm

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1198 Batistella et al.
Velocities (m/s) in the diagonal directions of the flow:
v = B cos(φ) (33)
and w= B sin(φ) (34)
1
2 kT S InU2 2
where B= ; φ = 2 πU 3
Mm
k = Boltzman constant and Mm = mass of a single molecule (kg).
These equations are applied for each molecule that evaporates, and
they depend on random numbers U1 , U2 , and U3 uniformly distributed in
the interval (0,1). These equations derive from the Maxwel’s equations and
present a normal distribution for the velocity.
It is also necessary to determine the new velocities of the molecules
after the collisions. In ref. 18 collisions were considered hard-sphere, where
any direction can be taken by the molecules after the collisions. Thus, this
isotropic condition can be used to determine the vectorial velocities of the
molecules 1 and 2 (both with same weight) after the collisions. These values
(u1’ and u2’) are shown in ref. 4. Therefore, considering a value of time dt,
it is possible to determine the displacement and the new position of each
molecule in the vapor space. It is also possible to know the destination of
each molecule, in addition to knowing when it reaches the condenser or it
comes back to the evaporator. If two molecules collide, the velocity vectors
are substituted according to the above equations. If collisions do not occur,
the molecules’ velocities remain unchanged.
Collisions
Applying the Monte Carlo method for rarefied vapor flow in case of
hard-sphere collisions, we have (4):
Nm 2
Nc = πd ng d (35)
2
where Nc is the number of collisions in the cell, n is the number of mol-
ecules in the cell, Nm is the number of molecules inside a cell, and g is the
average relative velocity (m/s). Then, for each interval of time dt, the total
of collisions in a cell will be accepted until to sum Nc collisions. The col-
lision partners are chosen randomly.
Sampling
The macroscopic properties of the vapor phase, as temperature, pres-
sure, and mean free path are calculated through the molecular velocity,
are as presented in ref. 4.
Monte Carlo Method (Vapor Phase)
The simulation of the vapor phase occurring in an interval of time
dt, follows:

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1199
1. Molecular vaporization on the evaporator inside the closest cell,
given by Langmuir’s equation (equation 27) in the time interval dt
(equation 29).
2. Molecular vaporization on the condenser inside the closest cell, also
using Langmuir’s equation and the consideration of item 1.
3. Determining molecular velocities of the evaporating molecules
through equations 32, 33, and 34.
4. Intermolecular collisions occur inside each cell until they reach a
value given by equation 35. The molecules that collide acquire new
velocity vectors given by equations u’, v’ and w’.
5. Displacement of the molecules for a time dt according to the their
velocity vectors.
6. Evaluation of each molecule, in function of its new position (new
cell, evaporator or condenser surfaces).
The calculations are carried out sequentially in each cell, and when a
cycle (battery of cells) is completed, the total time of simulation is increased
by dt. This process is repeated until the time that the simulation reaches a
preestablished value or a certain variable reaches the steady state (distilla-
tion rate, concentration, separation selectivity, etc). The distillation rate is
given by the number of molecules that definitely reach the condenser (go
out the last cell, next to the condenser). The ratio between the quantity of
a particular molecule (component) and the total of molecules that definitely
reach the condenser gives the distillate concentration for this species.

DISMOL Simulator
The simulator DISMOL (1) presents the architecture described bel-
low. For each integration step (∆z -falling film; ∆x centrifugal) of the liq-
uid phase modeling, the temperature and concentration conditions in the
evaporation surface are obtained (item 2). These variables, being known
Langmuir’s equation is used to determine the amount of molecules that
will vaporize. These molecules will be the start-up for the Monte Carlo
method, described for the vapor phase (item 3). Finally, results as the
distilled rate, compositions, etc. (inside the integration step carried out in
the liquid phase) are obtained. A new integration step is carried out in the
liquid phase, and the procedure above is repeated. This process will fin-
ish when the integration reaches the end of the evaporator. At this point,
a global analysis of the distillate and concentrate rates is made and the
final separation is determined.

Simulations for Recovering Tocopherols

When DDSO was used directly in the molecular distillator, it was not
possible, a priori, to obtain a product with high tocopherol concentrations.
The process, then, requires a pretreatment to transform heavier compo-
nents into components of smaller molecular weights. Several methods for

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1200 Batistella et al.
recovering tocopherols from vegetable oils have been studied. The DDSO
can be saponified and acidulated to convert glycerides and sterol esters to
free fatty acids and free alcohols. The free fatty acids are esterified with a
monohydric alcohol of low molecular weight in the presence of a mineral
acid catalyst. The sterols are precipitated and crystallized by the addition
of water to the mixture, and the tocopherols are concentrated by removing
fatty acid esters by molecular distillation (19).
After the molecular distillation process, two flows are generated: one
rich in the most volatile component, such as ethyl esters, and another one
rich in glycerides and tocopherols. In this work, this method of concentra-
tion is being considered, and the conversion of fatty acids to ethyl esters
was considered as being of 95%.
Then, the mixture containing 71% ethyl esters (ethyl–palmitate, ole-
ate, stearate, and linoleate), 5% fatty acids not converted (palmitic, oleic,
stearic, linoleic acids), 21% glycerides (tri-, di- and monoglycerides), 3%
fitosterols (β-sitosterol, stigmasterol) was considered for the modeling
and simulation of the falling film and centrifugal molecular distillators.
Simulations for the falling film molecular distillator were carried out at
different operating temperatures (from 150 to 170°C) and feed flow rate
(from 0.3 to 3.0 kg/h) and for simulations of the centrifugal molecular
distillator, the operating temperatures (from 180 to 220°C) and the feed
flow rate (from 0.3 to 5.0 kg/h) were considered.
Simulations were carried out with variations in the feed flow rate
for each operating temperature. The results of the simulations are shown in
Figs. 6 and 7. In the process, there are two product streams: the concen-
trated, rich in tocopherols and in heavier components such as fatty acids,
glycerides, and fitosterols, and the distilled, rich in the most volatile com-
ponent such as ethyl esters and monoglicerides (poor in tocopherols). The
presented results refer to the concentrated product stream (Figs. 6 and 7).
Figure 6 shows results of simulations for the falling film molecular
distillator for different operating temperatures and feed flow rates. The
comparative analyses between the quantity of tocopherol molecules and
the total number of molecules that reach the condenser gives the concentra-
tion of tocopherols in the distillate flow. It is possible to observe that, for the
same temperature, reducing the feed flow rate will increase the tocopherol
concentration. This behavior is repeated for all temperatures studied.
Moreover, for a fixed feed flow rate, increasing the operating temperature,
increases the tocopherol concentration. It can also be verified that the toco-
pherol concentrations converge to a maximum value, about 38%, in the
cases presented.
Figure 7 shows the simulation results for the centrifugal molecular
distillator. The same behavior mentioned for the falling film distillator can
be observed. However, in this equipment, the concentration converges to
values of about 41%, since the operating temperature is higher. The use of
temperatures comparatively larger in the centrifugal distillator is due to
the fact that this distillation has shorter residence time (on the evaporator)

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1201

Fig. 6. Tocopherol concentrations in the distillate stream. Falling film distillator.

Fig. 7. Tocopherol concentrations in the distillate stream. Centrifugal distillator.

than the falling film distillator, as shown in Figs. 8 and 9. The residence time
is determined by the axial velocity equations considering the evaporator
length (equation 2 for the falling film and equations 14, 15, and 16 for the
centrifugal distillator).
At reduced flows, the residence time, for both equipment, are very
long: the residence time of the centrifugal distillator reaches 3.4 s, as is
shown in Fig. 8, whereas for the falling film distillator, the residence time
is about 370 s (Fig. 9).
Components can be damaged preferentially at high temperatures or at
long expositure times. Products that are damaged when they are exposed
for a long time must be handled by the centrifugal distillator, whereas for
the products that are more sensitive at high temperatures, the falling film
is preferred.

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1202 Batistella et al.

Fig. 8. Residence time vs feed flow rate. Centrifugal distillator.

Fig. 9. Residence time vs feed flow rate. Falling film distillator.

Figures 10 and 11 show the residence time profiles under different

tocopherol concentrations in both equipment. The higher the concentration
is, the longer the contact time of the material on the evaporator will be. Note
that for the centrifugal distillator, the residence time is much more sensitive
than the falling film at medium tocopherol concentrations for different
temperature levels due to higher evaporation rate (high operating tem-
perature) of the centrifugal distillator. This is necessary because the evapo-
ration area is reduced compared to the falling film.
Figures 12 and 13 show the tocopherol concentrations given by equa-
tions 9 to 12 for the falling film and 23 to 26 for the centrifugal distillator on
evaporation surface in the liquid film: z is the distance for the falling film
and x is the distance for the centrifugal distillator. At the beginning, the
concentration profile of the centrifugal distillation shows low sensitivity
due to the reduced evaporation area near the center of the evaporator;

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1203

Fig. 10. Residence time vs tocopherol concentration. Falling film distillator.

Fig. 11. Residence time vs tocopherol concentration. Centrifugal distillator.

Fig. 12. Tocopherol concentrations in the evaporation surface. Falling film distillator.

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1204 Batistella et al.

Fig. 13. Tocopherol concentrations in the evaporation surface. Centrifugal distillator.

however, the concentrations acquire similar values to the falling film as the
liquid reaches the end of the evaporator.

Concluding Remarks
The use of modeling and simulation applied to the tocopherols recov-
ery one allow to evaluate, a priori, the potential of the molecular distillation
process for this application. It was possible to obtain concentrations of
tocopherol with 40%, starting from a DDSO with 8% of tocopherol, that is,
it was possible to concentrate five times the initial material, using just one
step of distillation.

Acknowledgments
The authors are grateful to FAPESP (Fundação de Amparo à Pesquisa
do Estado de São Paulo) for the financial support for this project (99/04656-
9, 99/03550-2, 98/14384-3).

Nomenclature
C Concentration, in mole fraction
Cs Surface concentration, in mole fraction
d Molecular diameter (m)
Di Diffusion coefficient (m2/s)
dt Time interval (s)
E Evaporation rate (kg/m2·s)
Ea Evaporation rate, (molecules/m2·s)
F Surface ratio
fi Distribution function
g Gravitational acceleration (m/s2), molecular relative velocity in the
vapor phase (m/s)

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

Molecular Distillation 1205
h Distance between evaporator and condenser surfaces (m)
Jij Integral of the collision
k Anisotropy of vapor (equation 1), Boltzman constant (vapor phase)
L Evaporator length (m)
m Mass flow rate (kg/s)
Mi Molar mass (kg/kg·mol)
Mm Molecular mass (kg)
n Number of components in the liquid mixture, number of molecules
(vapor phase)
Nav Avogadro´s number
Nc Number of collisions
Nm Number of molecules
P System pressure (Pa)
Pisat Vapor pressure (Pa)
r Radial coordinate (m)
R Outer radius of condenser (m)
Rg Universal gas constant, (J/kmol·K)
S Film thickness (m)
T Temperature (K)
Ts Surface temperature (K)
u Velocity in x direction, liquid film or molecules (vapor phase) (m/s)
v Velocity in y direction, liquid film or molecules (vapor phase) (m/s)
w Velocity of the molecules in z direction (m/s)
Wz Velocity in z (m/s)
x Distance along rotor surface (m)
y Distance perpendicular to rotor surface (m)
z Axial coordinate (m)
Greek Symbols
α Thermal diffusivity (m2/s)
β Mean path of vapor molecule (m)
∆H Evaporation enthalpy (J/kg)
φ Cone half-angle (rad)
λ Thermal conductivity (W/m(K)
η Viscosity (Pa(s)
µ Cinematic viscosity (m2/s)
ρ Density (kg/m3)
Ω Rotor speed (rad/s)

References
1. Batistella, C. B. (1999), PhD thesis, UNICAMP, LDPS Campinas-SP, Brazil.
2. Ooi, C. K., Choo, Y. M., Yap, S. C., Barison, Y., and Ong, A. S. H. (1994), AOCS Press
71, 423 – 426
3. Seetharamaiah, G. S. and Prabhakar, J. V. (1986), J. Food Sci.Technol. 23, 270 – 273.
4. Batistella, C. B., Maciel, M. R. W., and Maciel Filho, R. (2000), Computers Chemical
Engineering 24, 1309 –1315.

Applied Biochemistry and Biotechnology Vols. 98–100, 2002

1206 Batistella et al.
5. Batistella, C. B. and Maciel, M .R. W. (1996), Computers Chemical Engineering 20
(Suppl.), S19 – S24.
6. Eitenmiller, R. R. (1997), Food Technology 51(5), 78 – 81.
7. Ramamuuthi, S. and McCurdy, A. R. (1993), AOCS 70, 287– 295.
8. Kawala, Z. and Stephan, K. (1989), Chem. Eng. Tech. 12, 406 – 413.
9. Kawala, Z. (1983), Kinetik der Oberflachenverdamfung unter den Bedingungen der
Molekulardestillation. Wydawnictwo Politechniki Wroclawskiej, Wroclaw, Poland.
10. Batistella, C. B. (1996), MS thesis, UNICAMP, LDPS Campinas-SP, Brazil.
11. Carnahan, B., Luther, H. A., and Wilkes J. O. (1969), Applied Numerical Methods, Wiley,
NY, pp. 440 – 442.
12. Bhandarkar, M. and Ferron, J. R. (1988), Ind. Eng. Chem. Res. 27, 1016 –1024.
13. Ferron, J. R. (1986), Ind. Eng. Ch. Fund. 25, 594 – 602.
14. Bird, G. A. (1970), Phys. Fluids 13, 2676 – 2681.
15. Bhandarkar, M. and Ferron, J. R. (1991), Ind. Eng. Chem. Res. 30, 998 –1007
16. Bird, G. A. (1994), Molecular Gas Dynamics and the Direct Simulation of Gas Flows,
Clarendon Press Oxford, UK.
17. Lutisan, J. and Cvengros, J. (1995), Chem. Eng. J. 56, 39 – 50.
18. Bird, G. A. (1976), Molecular Gas Dynamics, Clarendon Press, Oxford, UK.
19. Smith, F. (1967), US patent no. 3,335,154.

Applied Biochemistry and Biotechnology Vols. 98–100, 2002