Sie sind auf Seite 1von 27

This article was downloaded by: [Universidad De Concepcion]

On: 17 June 2014, At: 08:29


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK
Drying Technology: An International Journal
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/ldrt20
Optimization of the Freeze-Drying Cycle: A New
Model for Pressure Rise Analysis
P. Chouvenc
a

b
, S. Vessot
a
, J. Andrieu
a
& P. Vacus
b
a
Laboratoire dAutomatique et de Gnie des ProcdsLAGEP-UMR Q 5007 , CNRS UCB
Lyon1-CPE , Villeurbanne, Cedex, France
b
Aventis Pasteur , Campus Mrieux , Marcy LEtoile, France
Published online: 06 Feb 2007.
To cite this article: P. Chouvenc , S. Vessot , J. Andrieu & P. Vacus (2004) Optimization of the Freeze-Drying Cycle:
A New Model for Pressure Rise Analysis, Drying Technology: An International Journal, 22:7, 1577-1601, DOI: 10.1081/
DRT-200025605
To link to this article: http://dx.doi.org/10.1081/DRT-200025605
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
DRYING TECHNOLOGY
Vol. 22, No. 7, pp. 15771601, 2004
Optimization of the Freeze-Drying Cycle:
A New Model for Pressure Rise Analysis
P. Chouvenc,
1,2
S. Vessot,
1
J. Andrieu,
1,
*
and P. Vacus
2
1
Laboratoire dAutomatique et de Ge nie des
Proce de sLAGEP-UMR Q 5007, CNRS UCB Lyon1-CPE,
Villeurbanne, Cedex, France
2
Aventis Pasteur, Campus Me rieux, Marcy LEtoile, France
ABSTRACT
The principal aim of this study was to evaluate the Pressure Rise
Analysis (PRA) method as a nonintrusive method for monitoring the
product temperature during primary drying of the freeze-drying
process of model pharmaceutical formulations. The principle of
this method, based on the MTM method initially published by
Milton et al.
[1]
consisted in interrupting rapidly the water vapor flow
from the sublimation chamber to the condenser chamber and by
*Correspondence: J. Andrieu, Laboratoire dAutomatique et de Ge nie des
Proce de sLAGEP-UMR Q 5007, CNRS UCB Lyon1-CPE, Ba t. 308G, 43 Bd
du 11 Novembre 1918, 69622 Villeurbanne, Cedex, France; E-mail: andrieu@
lagep.cpe.fr.
1577
DOI: 10.1081/DRT-200025605 0737-3937 (Print); 1532-2300 (Online)
Copyright & 2004 by Marcel Dekker, Inc. www.dekker.com
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
analyzing the resulting dynamics of the chamber total pressure
increase. A new physical model, named PRA model, based on
elementary heat and mass balance equations and on constitutive
equations expressing the elementary fluxes, was proposed and
validated in this study for interpreting the experimental pressure
rise data. It was possible to identify very precisely the values of some
key parameters of the freeze-drying process such as the ice
sublimation interface temperature, the mass transfer resistance of
the dried layer and the overall heat transfer coefficient of the vial.
The identified ice front temperatures were compared with experi-
mental data obtained from vial bottom temperatures measured by
thin thermocouples during freeze-drying runs of 5% w/v mannitol
solutions. These two sets of data were found consistent with a maxi-
mum difference of no more than 2

C. The dried layer mass transfer


resistance increased linearly as a function of its thickness, and the
values were coherent with the few literature data published for this
system. The method also led to reliable values of the vial overall heat
transfer coefficient of approximately 20 Wm
2
K
1
in accordance
with the published data for this type of vials and these experimental
freeze-drying conditions.
Key Words: Freeze-dryer control; PRA model; MTM model;
Pharmaceuticals freeze-drying; Ice sublimation temperature.
INTRODUCTION
During freeze-drying processes for pharmaceuticals, the precise
determination of the mean product temperature is the key-parameter
for the optimization of the process and its control. Then, it is important
to have a precise and significant estimation of this parameter in order to
maintain the product temperature as high as possible during the sublima-
tion period without going beyond the collapse temperature.
[13]
Insertion
of thermocouples or Pt100 sensors in a few vials is a widely used method
to measure this mean product temperature. Nevertheless, it is well known
that this sensor introduction modifies appreciably the elementary
phenomena of nucleation and of ice crystals growth,
[2]
by lowering the
degree of supercooling that leads to an increase of the average ice crystal
size in the frozen matrix. As a consequence, the mass transfer resistance
to water vapor flux through the dried layer in monitored vials is
considerably reduced and their drying kinetics are quite different from
the whole batch. Moreover, these monitored vials, for technical reasons,
are, most of the time, placed close to the loading door and, thereby,
1578 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
submitted to wall effects. So, all these mechanisms tend to make the
monitored vials dry faster, preventing them from being representative of
the entire batch. Besides, this method required a manual process that may
compromise the product sterility when manufacturing pharmaceutical
products.
This is why the Pressure Rise Analysis method (PRA method),
derived from the MTM method originally proposed by Milton et al.
[1]
and modified recently by Obert,
[3]
appears to be a very promising
method. Liapis et al.
[4]
have also published a similar method for deter-
mining mean product temperatures and temperature profiles in the frozen
layer. Indeed, it is a rapid, simple to implement, noninvasive, and
averaging temperature measurement method that requires a freeze-dryer
equipped with an external condenser and a very fast closing separating
valve with closing times lower than 1 s. Process parameters such as the
temperature at the ice sublimation interface, T
i
, the resistance to mass
transfer of the dried layer, R
p
, and the overall heat transfer coefficient,
K
v
, could be identified by fitting the experimental pressure rise data
obtained after the closing of the separating valvewith a transient
pressure response model.
Finally, this method has been proposed as a more precise predictive
way for determining the end of primary drying. One of the objectives of
this study was to point out the critical temperature increase of the
product during the PRA measurement at the end of primary drying
period and suggest an alternative way to determine the switch over from
primary drying to secondary drying period.
MATERIALS AND METHODS
Materials
Mannitol (reagent grade) was used as received from Sigma Aldrich
(Saint Louis, MO). Unstoppered 7 mL glass tubing vialshaving a
total mass of 8 gmanufactured by VerreTubex (France) were used for
this study.
Equipment and Methods
The freeze-dryer used was a model SMH45 manufactured by
Usifroid (France). The freeze-drying chamber contained three shelves
providing a total shelf area of 0.45 m
2
. The total pressure was
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1579
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
measured using a capacitance manometer MKS Baratron 622 (MKS
Instruments) having a typical time response as low as 4 ms. The water
vapor pressure measurements were realized by using a humidity
sensor PANAMETRICS M2LR (Chicago), the principle of which was
based on capacitance determination and providing data converted in
dew point temperature values in a range lying from 110

C to 20

C,
with a precision of 1

C. Calibration curves for all the sensors were


supplied by the manufacturers. The temperature at the bottom of the
product inside 6 vials and the gas temperature inside the sublimation
chamber were monitored with K type thermocouples. A pneumatic
butterfly valve, with a closing time approximately equal to 0.5 s, was
mounted in the cylindrical duct connecting the freeze-drying chamber
and the condenser chamber. During the PRA runs, the dynamic pressure
rise was monitored with a data acquisition systemmultimeter 2700
from Keithley (Keithley Instruments)allowing a sampling period as
low as 10 ms.
A 5% mannitol w/w solution was prepared with distilled water. The
whole batch for standard freeze-drying runs contained 324 vials, each
containing 2 mLof solution corresponding to an initial product thickness L
0
of 1 cm. As recommended in the literature, the ratio chamber volume/ice
sublimating front area was equal to 1.28m. After freezing the product down
to 45

C with a cooling rate of 1

C/min, the product was freeze-dried at a


shelf temperature of 0

Cand with a total pressure equal to 38Pa. Condenser


temperature was kept constant during all experiments at around 60

C.
The mean experimental product temperature was calculated by
averaging the temperature of five monitored vials. The temperature of
the vial located close to the Altuglass door of the freeze-dryer was
discarded from the calculation, because it was always 25

C higher than
the mean product temperature due to door radiation effects. The chamber
leak rate was determined with an empty chamber by measuring the
pressure increase 2 h after closing the butterfly separating valve and a
mean leak rate equal to 0.13 Pa s
1
was obtained.
MODELING OF PRESSURE RISE KINETICS
Literature Review
The first modeling of the pressure rise dynamics after rapidly closing
the separating valve was the MTM method proposed by Milton et al.,
[1]
1580 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
expressed by the following equation state:
Pt P
i
T
i
P
i
T
i
P
c
: exp
N:A:R:T
v
M
H
2
O
:V:R
p
: t
_ _
,..,
TERM1

P
i
T
i
:H
s
:M
H
2
O
R:T
2
i
:
T
2
1
8

2
: exp
kappa:
2
L
2
: t
_ _ _ _
,..,
TERM2

P
i
T
i
:H
s
:M
H
2
O
R:T
2
i
:
1
Cp:L:
c
: K
v
T
shelf
T
bottom
: t
,..,
TERM3
F
leak
: t
,..,
TERM4
1
In a semi-empirical approach, this model assumes that four physical
phenomena influence independently the pressure rise inside the chamber.
The total pressure increase was assumed to result from four elementary
contributions: term 1, the ice sublimation at constant sublimation front
temperature, T
i
; term 2, the equilibration of the thermal gradient in the
frozen layer; term 3, an increase in the ice temperature of the frozen
matrix due to the continuous heating during the measurement time
interval; term 4, the leak rate of the freeze-dryer directed from the
external atmosphere to the chamber inner. This model allowed these
authors to identify T
i
and R
p
values from experimental pressure rise
kinetics data obtained with mannitol, lactose, and potassium chloride
solutions; however, Milton et al.
[1]
observed that the values identified for
the overall heat transfer coefficient, K
v
, were twice or three times higher
than the ones obtained by a direct measurement method and, thereby,
could not be identified from their model.
More recently, Obert
[3]
proposed a modification of the previous MTM
model by taking into account two other elementary physical phenomena:
. The desorption of the bound water in the dried layer during
primary drying which contributes to the increase of the total
pressure. It was expressed by the following simplified relationship:
dP=dt D
des
2
where D
des
represents the desorption kinetics constant (Pa s
1
).
. The thermal inertia of the glass wall of the vial which was
introduced in the accumulation term on the heat balance of the
vial (term 3 of Eq. (1)).
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1581
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
The overall heat transfer coefficient, K
v
, was calculated from the
sublimation kinetics data by adopting the heat and mass transfer steady
state hypothesis:
K
v

P
i
P
c
:H
s
R
p
:T
shelf
T
bottom

3
where P
i 0
represents the water vapor partial pressure of water at the
interface before the valve closing. The term corresponding to the thermal
equilibration of the frozen layer after closing the valve was dropped
because it does not correspond to physical reality. So, the complete model
was expressed by the following equation (Eq. (4)):
Pt P
i
T
i
P
i
T
i
P
c
: exp
N:A:R:T
v
M
H
2
O
:V:R
p
: t
_ _

P
i
T
i
:H
S
:M
H
2
O
R:T
2
i
:
P
i
T
i
P
c
:H
S
Cp:L:
C
m
vial
:Cp
vial
=A : R
p
: t
F
leak
: t D
des
: t 4
The aim of our study was to set up a more advanced model taking
into account the increase in water vapor pressure at the sublimation
interface, noted P
i
, due to the overall product temperature increase as well
as the heat transfer coefficient variations during the dynamic pressure rise.
New Pressure Rise Dynamics Modeling: PRA Model
In a more rigorous approach, we propose to delete the arbitrary
hypothesis of additivity of the four mechanisms contributing to the total
pressure increase and to set up a new modeling based on elementary
heat and mass transfer balances coupled with the constitutive equations
expressing the heat and mass fluxes during the desiccation process.
Two gas species contribute independently to the pressure rise dynamics
during PRA runs: the water vapor and the inert gas.
The water vapor accumulated in the product chamber was generated
by the sublimation of ice crystals and, to a lower extent, by the desorption
of the bound water from the dried layer. Assuming a rapid valve closing
and that the gas phase obeyed the ideal gas law, the rate of total pressure
increase was described by the relationship:
dP
water
t
dt

NART
v
M
H
2
O
VR
p
P
i
T
i
P
w
t D
des
5
1582 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
where N represents the number of vials, A the sublimation interface area
for one vial, R the universal gas constant, P
i
and P
w
, respectively the
water vapor pressure in equilibrium with the interface temperature T
i
and
in the sublimation chamber. The rate of desorption of the bound water
from the dried layer, noted D
des
(Pa s
1
), was assumed to remain constant
during the PRA run and Antoines law was used to express P
i
as a
function of T
i
:
lnP
i

A1
T
i
B1 6
where A1 and B1 are constants respectively equal to 6320.1517 and
29.5578 in S.I. units.
[5]
Assuming that there was only water vapor in the
chamber before closing the butterfly valve, the inert gas partial pressure
increase was only due to the leak of the freeze dryer chamber:
dP
inert
t
dt
F
leak
7
where F
leak
represents the constant chamber leak rate.
Finally, the total gas pressure was obtained by adding the partial
pressures of both species:
Pt P
water
t P
inert
t 8
Heat Balance for One Vial
During a PRA experiment, the total heat flux dQ/dt transferred
from the shelves to the product contributed mainly to the ice sublimation
and, to a least extent, to the product and the glass vial temperature
increase and to the desorption of the bounded water. Assuming that
the temperature increase at the interface dT
i
/dt was the same as the
mean product temperature rise, the heat balance for a whole vial was
expressed by:
dQ
dt
AK
v
T
shelf
T
bottom

c
CpLA m
vial
:Cp
vial

dT
i
t
dt

A:H
s
R
p
P
i
T
i
P
w
t
VM
H
2
O
H
des
NRT
v
D
des
9
The total mass of one vial was 8 g but due to the poor thermal
conductivity of glass, it was assumed that the mass contributing to the
overall temperature increase was only the mass of the glass wall in
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1583
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
contact with the frozen matrix. Globally, this mass was estimated to 2 g.
[6]
Combining Eqs. (6) and (9), we finally obtained:
dT
i
t
dt

1

c
CpL
m
vial
Cp
vial
A
K
v
P:T
shelf
T
bottom

H
s
R
p
exp
A1
T
i
t
B1
_ _
P
w
t
_ _ _

VM
H
2
O
H
des
NART
v
D
des
_
10
dP
water
t
dt

NART
v
M
H
2
O
VR
p
exp
A1
T
i
t
B1
_ _
P
w
t
_ _
D
des
11
Heat Flux Expression
The literature data about heat transfer modeling during freeze-drying
expressed the total heat transferred from the surroundings and the shelf
to the vial as the result of three basic mechanisms: conductive heat
transfer by contact between the vial and the shelf, radiative flux from the
surroundings to the vial and the gas-conduction heat transfer at low
pressure through the gas layer between the bottom of the vial and the
shelf. Consequently, the overall heat transfer coefficient, noted K
v
, was
expressed, in a first approach, by the following relationship under the
hypothesis of additivity of these mechanisms
[7]
:
K
v
K
cont
K
rad
K
gas-cond
12
where the heat transfer coefficients K
cont
and K
rad
were constant for a
given type of vial and a given geometrical pattern of the vials on the
shelves.
[7]
However, the gas-conduction heat transfer coefficient was dependent
on the gas nature and on the total pressure inside the freeze-drying
chamber and described by the following equation
[7]
:
K
gas
-
cond


0
P
1
0
l
sep
P=
gas
0

13
where is an adimensional coefficient characterizing the heat transfer
efficiency, l
sep
is the equivalent mean distance between the shelf and the
bottom of the vial,
0
is the molecular thermal conductivity of the gas
and
0,gas
is the gas thermal conductivity at atmospheric pressure, these
two last parameters being intrinsic values characteristic of the gas nature.
It is worth noting that the vial heat transfer coefficient, K
v
, could be
1584 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
also calculated from the mass transfer resistance value, R
p
, and from the
total water vapor partial pressure driving force by the following equation:
K
v0

1
T
shelf
T
bottom

P
i0
P
c
:H
s
R
p

VM
H
2
O
H
des
NART
v
D
des
_ _
14
The temperature at the bottom of the vial, noted T
bottom
, was
assumed to be constant during the PRA run and calculated from the
steady state heat transfer hypothesis through the frozen layer by the
following relationship:
T
bottom

e
vial

vial

c
_ _
:P
i0
P
c
:
H
s
R
p
T
i0
15
The variation of K
v
as a function of the total pressure was obtained
by differentiating Eq. (13), namely:
dK
v
P
dP


0
1
0
l
sep
=
gaz
0
P
_ _
2
16
The final equation set, Eqs. (6)(8), (11), and (14)(16), subject to the
following initial conditions:
T
i
j
t0
T
i0
; P
i
j
t0
P
i0
; Pj
t0
P
c
; K
v
j
t0
K
v0
;
P
water
j
t0
P
c
; P
inert
j
t0
0
was integrated using Matlab

software. A nonlinear regression analysis


based on Marquardt-Levenberg algorithm was used to minimize the
objective function F
s
defined as the sum of the error function. This error
function was expressed as the difference between the experimental total
pressure values and the corresponding calculated ones.
Errort P
experimental
t Pt 17
F
S

jErrortj 18
This new advanced model, set up from constitutive and balance
equations of the whole system, allowed the identification of the parameters
values which characterized the desiccation kinetics: the sublimation front
temperature, T
i
, the resistance to mass transfer of the dried layer, R
p
, all
along the freeze-drying cycles by implementing different PRA runs. The
overall heat transfer coefficient K
v
can also be calculated from the steady
state equations as previously suggested by Obert.
[3]
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1585
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
RESULTS AND DISCUSSION
Total Pressure Rise Dynamics
Figures 1A and B depict a comparison of the different model
predictions for a typical set of experimental data obtained after 5 H of
primary drying (Fig. 1A) and after 11 H (Fig. 1B). Estimated thicknesses
of the dried layer were respectively 3 and 9 mm. Experimental pressure
rise curves showed two distinct parts as shown on this same figure. The
first part was observed for times smaller than 8 s, during which the total
pressure increased rapidly and resulted mainly from ice sublimation at
the interface. Then, a second part was observed when the total pressure
rose more slowly, mainly resulting from the product temperature increase
due to continuous vial heating as the total chamber pressure approached
the water vapor pressure in equilibrium with ice at the sublimation
interface.
At the beginning of primary drying, all the three previously described
models, with identified parameter values gathered on Table 1, fitted
very well to the experimental data points so that no noticeable dif-
ference between them can be observed on Fig. 1A. A poorer fit for
higher temperature values was observed for Miltons model as shown
on Fig. 1B. However, these results supported the apparent validity of these
mathematical models.
The three models led to similar values of T
i
, significantly higher than
the mean experimental product temperatures by 1 to 2

C. Besides,
identified R
p
values were comparable with a maximum variation of 20%
from one to another. K
v
values, calculated from the identified T
i
and R
p
values instead of identifying it from experimental data, were in the same
order of magnitude but Miltons model seemed to slightly overestimate
this parameter in comparison with the other two models. Finally, the
desorption rate values identified from the two models that take this
phenomena into account in their mathematical expression, differed from a
large factor of 400. Indeed, the PRA model identified D
des
values leading
to a negligible pressure rise as compared to all the other sources. Whereas
Oberts model could be contested as it did not take this parameter
into account in the heat balance, the PRA model provided values in
agreement with previous studies which reported that the amount of
bound water during primary drying was negligible with respect to the
amount of water removed by sublimation
[4,8]
. Therefore, the PRA model
was rewritten waving the desorption term from Eqs. (5), (9)(11), and
(14) and compared to the original one. A comparison of the two analyses
is presented on Table 2.
1586 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
Figure 1. (A) Experimental data () fit by the PRA model (solid line), Oberts
model (dash line), and Miltons model (dot dash black). T
shelf
0

C, P38 Pa.
Data fit for a mean experimental product temperature equal to 249.5K after
280 min, (B) Experimental data () fit by the PRA model (solid line), Oberts
model (dash line), and Miltons model (dot dash black, pointed by an arrow).
T
shelf
0

C, P38 Pa. Data fit for a mean experimental product temperature


equal to 256K after 680 min. (View image in color online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1587
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
It proved that no significant differences could be observed for the
identified parameter values, that the calculation times decreased and that
the numerical stability of the model increased (data not shown).
Consequently, the PRA model which did not take into account the
desorption phenomena, was chosen for the final study as it offered the
advantage to identify only two parameters T
i
and R
p
, instead of three for
all the other models.
Temperature Measurements
The main advantage of this rapid transient method was the
estimation of the mean temperature of the sublimation interface, T
i
,
without any intrusive sensor. From this estimation, it seems possible to
adjust the operating parameters (shelf heating rate) in order to maintain
the mean product temperature, estimated by the ice front temperature, as
high as possible and below the collapse temperature all along the primary
and the secondary drying periods. Experimental T
i
values were calculated
from the mean values given by thermocouple probes, assuming that
T
thermocouple
T
bottom
i.e., that the probe was in close contact with the
vial bottom and that the thermocouple introduction all along the central
Table 1. Comparison between the identified parameters values from the
three models for a 5% w/v mannitol solution after 280 min of primary drying.
Freeze-drying conditions: T
shelf
0

C, P38 Pa.
Model T
i
(K)
R
p
(kPa m
2
s kg
1
)
K
v
(Wm
2
K
1
)
D
des
(Pa s
1
)
PRA
251.9 312 22.8 6.26 10
4
Milton et al.
[1]
250.6 256 28.7 N.A.
Obert
[2]
251.3 276 22.7 0.260
Table 2. Comparison between PRA models with and without desorption term.
Model T
i
(K)
R
p
(kPa m
2
s kg
1
)
K
v
(Wm
2
K
1
)
D
des
(Pa s
1
)
PRA
251.9 312 22.8 6.26 10
4
PRA (without desorption)
251.7 319 18.4 N.A.
1588 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
axis of the vial had no significant effect on the recorded experimental
temperature:
T
i Thermocouple
T
Thermocouple

K
v
:LT
shelf
T
Thermocouple

c
19
where L represents the thickness of the frozen layer and,
c
, its apparent
thermal conductivity.
Figure 2 shows an example of the data collected during the primary
drying of a 5% w/v mannitol solution.
Unlike the results reported by Milton et al.
[1]
for a 5% mannitol
solution, the interface temperatures identified by the three models were
systematically 1 to 3

C higher than experimental ones during the primary


drying. Such differences were greater than the numerical model
uncertainties and the experimental errors. Several hypotheses could
explain this discrepancy. Vials used for this study were smaller than the
ones used for previously reported MTM studies so that the introduction
of a thermocouple became a more important relative perturbation than
in the case of larger vials, which could considerably modify the freeze-
drying behavior of monitored vials. Indeed, the tip of the thermocouple
in contact with the bottom of the vial acted as a heterogeneous nucleation
site and decreased the degree of supercooling during the freezing stage.
242
244
246
248
250
252
254
256
258
0 100 200 300 400 500 600 700
Time (min)
T
i

(
K
)
Figure 2. Ice sublimation front temperature. Comparison between PRA
model (diamonds), Miltons model (triangles), Oberts model (circle), and mean
product temperatures (squares) measured by thermocouples. (View image in
color online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1589
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
The mean ice crystal size of such vials had the tendency to increase,
lowering the mass transfer resistance of the dried layer.
[2]
Moreover, the
body of the probe itself created a preferential pathway for the water
vapor flux from the ice front to the top of the dried layer, which also
decreased the mass transfer resistance. These artefacts on the dried layer
structure for monitored vials led to faster sublimation kinetics and to a
lower product temperature at steady state. Therefore, considering these
possible artifacts and the great sensitivity of the PRA model to parameter
T
i
value, the corresponding identified values were expected to be more
representative of the overall behavior of the whole batch in such
experimental conditions.
The difference between experimental and identified T
i
values from all
models was almost constant for L0.2 cm. However, when reaching the
end of primary drying, experimental values became even higher than
the identified ones as shown on Fig. 2. This behavior could be partly
explained by the fact that the thermocouples were not exactly in contact
with the vial bottom so that they lost contact at some point with the
remaining frozen layer. In addition, side vials on the tray completed their
sublimation significantly earlier. In order to compensate for the reduction
of the effective sublimation area, which was assumed to be constant,
all models tended to underestimate parameter T
i
. Furthermore, this
imprecise evaluation of T
i
appeared to be correlated to a poor fit of
experimental data by these models as shown Fig. 1B. Therefore, the PRA
model allowed a more appropriate determination of T
i
values on a larger
product temperature range, while identifying only two parameters instead
of three needed for Miltons and Oberts models.
Overall Heat Transfer Coefficient
The heat flux transferred from the chamber walls and from the
shelves to the vials during the PRA measurement period provided the
sublimation latent heat and was also responsible of the frozen layer
temperature increase. This temperature increase, which contributed to
maintain a water vapor driving force for the sublimation, and the leak
rate toward the chamber, mainly determined the total pressure rise during
the second half period of the PRA experiment. Nevertheless, Milton et al.
reported K
v
values from regression analysis 23 times higher than the
values calculated by direct measurements.
[9]
This tendency has also been
observed in this study by identifying their model to our experimental data
as shown Fig. 3.
1590 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
It may be observed that the identified K
v
values from Miltons model
are very scattered and varied from the lower bound imposed, namely
K
v
0, up to K
v
40 W/m
2
/K, which is twice the value obtained by the
two other models. Moreover, when product temperature got over 20

C,
Miltons model could not properly fit experimental data as shown by
Fig. 1B and the identified K
v
values were not reliable. This scattering
reflected the lack of precision of their model, due to terms 2 and 3 of
Eq. (1), which determined the product temperature increase as a func-
tion of time and mainly influenced the second part of the total pressure
dynamics curves. On the other hand, PRA model and Oberts model
calculated the initial overall heat transfer coefficient values, noted K
v0
,
assuming steady state transfer mechanisms between two runs.
[3]
Both
models led to stable values between 18 and 22 W/m
2
/K, independently of
the initial product temperature. These results were in agreement with
the experimental K
v
values obtained by gravimetric method, namely
K
v
20 W/m
2
/K, in the exact same experimental conditions.
[6]
Pikal
et al.
[9]
reported K
v
values equal to 27 W/m
2
/K for tubing vials and
18 W/m
2
/K for moulded vials. Furthermore, in order to develop a more
realistic and precise heat transfer model, the PRA model described
the evolution of the K
v
values during the PRA runs by using Eqs. (14)
and (16). Variations of K
v
values as a function of time after closing the
separating valve are presented on Fig. 4.
0
5
10
15
20
25
30
35
40
45
0 100 200 300 400 500 600 700 800
Time (min)
K
v

(
W
/
m

/
K
)
Figure 3. Overall heat transfer coefficient values identified by PRA model
(diamonds) by Miltons model (triangles) and by Oberts model (circles). (View
image in color online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1591
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
We observed that, during the 20 s time interval of a standard PRA
experimental run, K
v
values increased from 23 to 29 W/m
2
/K due to the
gas thermal conductivity increase. In order to verify the relevance of
introducing a variable K
v
parameter in the model, the identification was
also carried out with the PRA model considering that K
v
remained
constant and equal to the initial overall heat transfer coefficient, K
v 0
.
Results are presented on Table 3. Identified R
p
values varied of about
10% and a 0.2

C difference was observed on T


i 0
values. These variations
were not significant considering experimental errors so that, practically,
the PRA model could be more simplified by assuming constant K
v
values
all along the PRA run.
Figure 4. Evolution of the vial overall heat transfer coefficient during a PRA
measurement. T
shelf
0

C, P38 Pa, T
exp
249.5K, t 280 min. (View image in
color online.)
Table 3. Comparison between identified parameters by the new MTM model
for different configuration of K
v
T
shelf
0

C, P38 Pa, T
exp
249.5K,
t 280 min.
T
i 0
(K) R
p
(KPa ms kg
1
)
Variable K
v
251.9 312
Constant K
v
252.1 337
1592 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
Mass Transfer Resistance Measurements
Several techniques to estimate the mass transfer resistance values, R
p
,
were reported in the literature.
[9]
They were all based on gravimetric
method and used either a single or a limited number of vials. Estimation
of R
p
from the PRA model allowed the determination of the mean
product mass transfer resistance for the whole vial batch, averaging the
heterogeneity of the cake structures generated by the random freezing
phenomena inside each vial. In order to calculate R
p
as a function of the
dry layer thickness, the position of the sublimation interface and the
thickness of the dry layer were estimated as a function of time by using
the following relationship:
Lt

c
AL
0


it=t
acq

i1
K
v
A:T
shelf
i T
bottom
i=H
s
:t
acq

c
A
20
As shown on Fig. 5, a typical set of R
p
values identified by the PRA
model was compared with the two previous MTM models predictions.
As previously observed in the case of a 5%w/v mannitol solution,
[9]
the product resistance identified by the PRAmodel increased continuously
0
100000
200000
300000
400000
500000
600000
700000
800000
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ex10

(m)
R
p

(
P
a
.
m

.
s
/
K
g
)
Figure 5. Product resistance as a function of the thickness of the dry layer
identified by PRA model (diamonds), Miltons model (triangles), Oberts model
(circle), and calculated from published data
[9]
(solid line). (View image in color
online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1593
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
and linearly as a function of the dry layer thickness. Mass transfer
resistance of the dried layer was empirically correlated to our data as
follows (S.I. units):
Rp 5:528 10
7
e 4:8 10
4
21
Experimental data correlation lines are passing close to the axis
origin, so that we did not consider any significant crust effect with this
system and with these freezing and freeze-drying conditions. Besides, R
p
values derived from the three models were similar for e <0.8 cm and
slightly lower than values found in the literature. However, R
p
values
from Oberts model were underestimated for e >0.8 cm and were 50%
lower than literature values while PRA and Milton models continued
providing consistent values. Therefore, the PRA model provided more
consistent Rp values on a wider product temperature range than other
literature models.
Mean Product Temperature Increase During PRA Runs
The temperature of the sublimation interface was calculated at each
time increment from the identified parameters values as shown on Fig. 6.
It was observed that during the first 3 s of the PRA run duration
time, the product temperature increase was low because most of the
heat supplied to the vial was used for the sublimation of ice. Then, the
estimated product temperature increased linearly as a function of time.
Therefore, it was possible to compare calculated and experimental mean
product temperature increase during a 20 s measurement time interval,
as shown on Fig. 7.
Experimental data were obtained by averaging the temperature
increases given by thermocouples in 2 monitored vials located at the tray
centre. We observed that experimental and calculated values were in very
good agreement during the first 9 h of the primary drying. Temperature
increase was quite constant all along this period, ranging from 0.4 to
0.8

C. However, when reaching the end of primary drying, the product


temperature increase reached a value of 1.6

C whereas the PRA model


predicted a temperature increase of only 0.8

C. As the model under-


estimated the product temperature increase when very little ice remained,
this last observation could confirm the hypothesis that only the glass wall
of the vial in contact with the frozen product has to be taken into account
in the vial heat balance
Furthermore, according to Milton et al., product temperature increase
during the 20 s time interval of a MTM experiment was less than 1

C.
1594 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
Figure 6. Modeling of the evolution of the ice sublimation interface temperature
during PRA measurement. T
shelf
0

C, P38 Pa, T
exp
249.5K, t 280 min.
(View image in color online.)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
0 200 400 600 800
Time (min)
T
e
m
p
e
r
a
t
u
r
e

i
n
c
r
e
a
s
e

(
K
)
Figure 7. Overall product temperature increase during a standard PRA
experiment. Observed experimentally (squares) and simulated by PRA model
(diamonds). (View image in color online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1595
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
We observed that this happened to be true only during the early stage
whereas this value was 23 times higher at the end of the primary drying.
Consequently, it seemed that using this method to determine the end point
of the primary drying
[3]
or to optimize the shelf temperature during the
process by monitoring with the material collapse temperature was very
efficient when freeze-drying robust products. Nevertheless, it could be
quite hazardous when freeze-drying very sensitive pharmaceutical
products such as vaccines, which cannot generally suffer temperature
increase as high as 2

C, even for a short time period.


Complementary Method for Determining the
End of Primary Drying
The previous critical analysis of the transient methods (PRA or
MTM) for the control of freeze-drying processes of sensitive pharmaceu-
ticals, in spite of their numerous advantages, has shown some limitations
due to the product temperature increase during the duration of the
pressure rise test, mainly at the end of the primary drying. Therefore, in
the perspective of a complete process control, it seemed necessary to set
up a complementary method to the PRA method in order to determine
precisely andnoninvasively the end of primary drying. Acontrol procedure
was tried by recording continuously the water vapor pressure in the
sublimation chamber by using a capacitance sensor (PANAMETRICS)
and by assessing the end of the sublimation period for the batch as a whole
to a sharp drop of the water vapor partial pressure in the chamber. The use
of this kind of sensor for such an application was first reported by Roy and
Pikal
[10]
who were able to detect the presence of a very small amount of ice,
down to 0.3% of vials in a batch having residual ice. More recent studies
reported by Bardat et al.
[11]
showed that this type of sensor had also a
greater sensitivity than other known methods such as Pirani gauge
determinations
[12]
and product temperature monitoring by thermo-
couples. Therefore, this technique was used during a cycle in order to
check its applicability to our pilot and for the investigated product. Figure
8 shows the graph obtained for the freeze-drying cycle of a 5%w/v
mannitol solution during which the product temperature and the water
vapor partial pressure were monitored.
PRA runs appeared as spikes in the total pressure and in the water
partial pressure curves but they did not introduce any significant
perturbation in the control of the process itself. Indeed, the time response
of used sensors was fast enough to follow the dynamics of such runs and
to get back to the nominal values before activating the alarms of the
1596 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
process. The end of sublimation in monitored vials by thermocouple
corresponded to a sharp temperature increase (1) and occurred about 5 H
before observing a noticeable change in the water vapor partial pressure
given by the moisture sensors (2). The end point of the sublimation
corresponded to the time when the partial pressure of water stabilized to
its lower value (3). The PRA method and the moisture sensor did not
interfere one with the other and could effectively be coupled in order to
get an optimal process control system, able to monitor in real time critical
drying parameters and to determine the end point of sublimation.
CONCLUSIONS
In this study, PRA method was evaluated as a method for
monitoring the freeze-drying cycles of vaccines in small vials. The main
advantages of this technique rely on its rapidity and on its noninvasive
principle. A new physical model for analyzing the dynamics of the
sublimation chamber total pressure increase was presented and validated.
This model allowed the identification of the main key-parameters of
-40
-30
-20
-10
0
10
20
30
40
13 15 16 18 20 21 22 24 25 26 29 31 33 34
Time (Hours)
T
e
m
p
e
r
a
t
u
r
e

(

C
)
0
100
200
300
400
500
600
700
800
P
r
e
s
s
u
r
e

(

b
a
r
)
1
2
3
Figure 8. Freeze-drying cycle of a 5% mannitol during a standard PRA
experiment. T
shelf
0

C, P38 Pa, T
exp
249.5K, t 280 min. Shelf temperature
(diamonds), product temperatures of two vials (solid lines), total chamber
pressure (X), and water partial pressure () are represented. (View image in color
online.)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1597
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
the freeze-drying process, namely the sublimation front interface
temperature, the mass transfer resistance of the dried layer and the
overall vial heat transfer coefficient on a wider product temperature range
than previous MTM models. Identified ice front temperature values were
close to the experimental data within a range of about 28C, while mass
transfer resistances and overall heat transfer coefficients were consistent
with values from the literature. Furthermore, we showed that the
desorption phenomena during primary drying was not significant enough
to be taken in account by the pressure rise analysis method and that it
could be practically neglected for the PRA modeling.
Besides, we showed experimentally and theoretically that the product
temperature increase during the time period of the PRA test was not
negligible at the end of the primary drying (up to 2

C). This is why the


industrial implementating of this PRA method could be hazardous
for very sensitive products like vaccines, etc. So, in these cases we
recommend the monitoring of industrial freeze-drying cycles by measur-
ing and following on-line the water vapor partial pressure in the
sublimation chamber with a capacitance type sensor.
NOMENCLATURE
A Ice sublimation front area of one vial (m
2
)
A1 Constant of Antoines law (Dimensionless)
B1 Constant of Antoines law (Dimensionless)
Cp Frozen layer specific heat (J/kg/K)
Cp
vial
Vial glass specific heat (J/kg/K)
D
des
Desorption constant during primary drying (Pa/s)
e Thickness of the dried layer (m)
Error(t) Cf. Eq. (17) (Pa)
F
leak
Freeze dryer leak rate (Pa/s)
l
sep
Mean separation distance between the vial and the
shelf (m)
L Thickness of the remaining frozen layer (m)
L
0
Initial thickness of the frozen layer (m)
kappa Miltons model constant (J/s/m
2
/K)
K
cont
Contact heat transfer coefficient for one vial (W/m
2
/K)
K
gas-cond
Gas conduction heat transfer coefficient for one vial
(W/m
2
/K)
K
rad
Radiative heat transfer coefficient for one vial (W/m
2
/K)
K
v
Overall heat transfer coefficient for one vial (W/m
2
/K)
1598 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
K
v0
Overall heat transfer coefficient before closing the
separation valve (W/m
2
/K)
M
H
2
O
Molecular weight of water (kg/kmol)
m
vial
Mass of one vial (kg)
N Number of vials for a whole batch
P Total pressure in the lyophilization chamber (Pa)
P
c
Initial total pressure in the lyophilization chamber (Pa)
P
i
Water vapor partial pressure at the interface (Pa)
P
i0
Water vapor partial pressure at the interface before
closing of the separation valve (Pa)
P
inert
Inert gas partial pressure in the lyophilization chamber
(Pa)
P
w
Water vapor partial pressure in the lyophilization
chamber (Pa)
Q Thermal energy (J)
R Universal gas constant (J/kmol/K)
R
p
Resistance to mass transfer of the dried layer
(Pa m
2
s kg
1
)
t Time (s)
T
bottom
Temperature of the frozen layer at the bottom of the
vial (K)
T
exp
Mean experimental product temperature (K)
T
i
(t) Temperature at the sublimation front (K)
T
i0
Initial temperature of the sublimation front before
closing the separation valve (K)
T
v
Temperature of gas in the lyophilization chamber (K)
T
shelf
Shelf temperature (K)
V Volume of the lyophilization chamber (m
3
)
Greek Letters
Efficiency coefficient for heat transfer (Dimensionless)
Hs Enthalpy of ice sublimation (J/kg)
t PRA measurements interval (s)
T Temperature gradient between the bottom of the frozen
product and the sublimation front (K)

c
Thermal conductivity of the frozen layer (W/m/K)

gas
0
Thermal conductivity of water vapor at atmospheric
pressure (W/m/K)

c
Ice density (kg/m
3
)

0
Water vapor molecular thermal conductivity (W/m
2
/Pa/K)
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1599
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
ACKNOWLEDGMENT
The authors wish to thank Aventis Pasteur and Re gion Rho ne-Alpes for
their financial support and their interest in this work.
REFERENCES
1. Milton, N.; Pikal, M.J.; Roy, M.L.; Nail, S.L. Evaluation of
manometric temperature measurement as a method of monitoring
product temperature during lyophilization. PDA Journal of
Pharmaceutical Science and Technology 1997, 51 (1), 716.
2. Pikal, M.J. Freeze-drying of proteins, part I: process design.
Biopharm 1990, 3, 1827.
3. Obert, J.P. Mode lisation, Optimisation et Suivi en Ligne du Proce de
de Lyophilisation. Application a` lAme lioration de la Productivite
et de la Qualite des Bacte ries Lactiques Lyophilise es. Ph.D. thesis,
INRA, Paris-Grignon, 2001; 186.
4. Liapis, A.I.; Sadikoglu, H. Dynamic pressure rise in the drying
chamber as a remote sensing method for monitoring the temperature
of the product during the primary drying stage of freeze-drying.
Drying Technology 1998, 16 (6), 11531171.
5. Perry, R.H.; Chilton, C.H. Chemical Engineers HandBook, 5th Ed.;
McGraw Hill, 1973.
6. Chouvenc, P. Optimisation du Proce de de Lyophilisation des
Liposomes. Ph.D. thesis, Universite Claude Bernard Lyon 1,
France, 2004.
7. Pikal, M.J. Heat and mass transfer in low pressure gases: applications
to freeze-drying. In Transport Processes in Pharmaceutical Systems;
Amidon, G.L., Lee, P.I., Topp, E.M., Eds.; Marcel Dekker: NY,
2000; 611686.
8. Sheehan, P.; Liapis, A.I. Modeling of the primary and secondary
drying stages of the freeze-drying of pharmaceutical products in
vials: numerical results obtained from the solution of a dynamic
and spatially multi-dimensional lyophilization model for different
operational policies. Biotechnology and Bioengineering 1998, 60,
712728.
9. Pikal, M.J. Use of laboratory data in freeze-drying process design:
heat and mass transfer coefficients and the computer simulation
of freeze-drying. Journal of Parenteral Science and Technology.
MayJune 1985, 39 (3), 115139.
1600 Chouvenc et al.
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

ORDER REPRINTS
10. Roy, M.L.; Pikal, M.J. Process control in freeze-drying: determina-
tion of the end point of sublimation drying by an electronic moisture
sensor. Journal of Parenteral Science and Technology Mar.Apr.
1989, 43 (2), 6066.
11. Bardat, A.; Biguet, J.; Chatenet, E.; Courteille, F. Moisture
measurement: a new method for monitoring freeze-drying cycles.
Journal of Parenteral Science and Technology Nov.Dec. 1993,
47 (6), 293299.
12. Nail, S.L. Methodology for in process determination of residual
water in freeze-drying products. International Symposium on
Biological Product Freeze-Drying and Formulation, Bethesda,
USA, 1990; Dev. Biol. Stand: Karger, Basel, 1991; 74, 137151.
13. Liapis, A.I.; Pikal, M.J.; Bruttini, R. Research and development
needs and opportunities in freeze drying. Drying Technology 1996,
14 (6), 12651300.
14. Sadikoglu, H.; Liapis, A.I. Mathematical modeling of primary and
secondary drying stages of bulk solution freeze-drying in trays:
parameter estimation and model discrimination by comparison
of theoretical results with experimental data. Drying Technology
1997, 15, 791810.
Freeze-Drying Cycle: A New Model for Pressure Rise Analysis 1601
D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4





















Request Permission/Order Reprints

Reprints of this article can also be ordered at
http://www.dekker.com/servlet/product/DOI/101081DRT200025605
Request Permission or Order Reprints Instantly!
Interested in copying and sharing this article? In most cases, U.S. Copyright
Law requires that you get permission from the articles rightsholder before
using copyrighted content.
All information and materials found in this article, including but not limited
to text, trademarks, patents, logos, graphics and images (the "Materials"), are
the copyrighted works and other forms of intellectual property of Marcel
Dekker, Inc., or its licensors. All rights not expressly granted are reserved.
Get permission to lawfully reproduce and distribute the Materials or order
reprints quickly and painlessly. Simply click on the "Request Permission/
Order Reprints" link below and follow the instructions. Visit the
U.S. Copyright Office for information on Fair Use limitations of U.S.
copyright law. Please refer to The Association of American Publishers
(AAP) website for guidelines on
Fair Use in the Classroom.
The Materials are for your personal use only and cannot be reformatted,
reposted, resold or distributed by electronic means or otherwise without
permission from Marcel Dekker, Inc. Marcel Dekker, Inc. grants you the
limited right to display the Materials only on your personal computer or
personal wireless device, and to copy and download single copies of such
Materials provided that any copyright, trademark or other notice appearing
on such Materials is also retained by, displayed, copied or downloaded as
part of the Materials and is not removed or obscured, and provided you do
not edit, modify, alter or enhance the Materials. Please refer to our
Website
User Agreement for more details.


D
o
w
n
l
o
a
d
e
d

b
y

[
U
n
i
v
e
r
s
i
d
a
d

D
e

C
o
n
c
e
p
c
i
o
n
]

a
t

0
8
:
2
9

1
7

J
u
n
e

2
0
1
4

Das könnte Ihnen auch gefallen