You are on page 1of 12

J Pharm Innov (2009) 4:5162

DOI 10.1007/s12247-009-9057-3

RESEARCH PAPER

The Engineering of Hydrogen Peroxide


Decontamination Systems
Stefan Radl & Stefanie Ortner & Radompon Sungkorn &
Johannes G. Khinast

Published online: 16 June 2009


# International Society for Pharmaceutical Engineering 2009

Abstract In this article, the latest developments for


designing hydrogen peroxide decontamination systems are
analyzed. Specifically, focus is given to the accurate
calculation of hydrogen peroxide condensation phenomena
and discussion of a new correlation for its accurate
prediction. A procedure for calculating the condensate
composition or the dew point out of this correlation is
detailed, and an hx diagram for moist, hydrogen peroxideladen air, which is of fundamental importance for the
rational design of hydrogen peroxide decontamination
systems, is proposed. Also presented are theoretical results
that illustrate the effect of condensation and evaporation in
these systems. Finally, some perspectives for improving
hydrogen peroxide systems, and the role computational
fluid dynamics (CFD) may have in this field, are provided.
Keywords Hydrogen peroxide . Decontamination .
Pharmaceutical engineering . Condensation .
Mollier hx diagram
Notation
Latin letters
ADHP
Total inner surface area of the DHP
chamber [m]
S. Radl : R. Sungkorn : J. G. Khinast (*)
Institute for Process and Particle Engineering,
Graz University of Technology,
8010 Graz, Austria
e-mail: khinast@TUGraz.at
URL: http://ipt.tugraz.at
S. Ortner
Ortner Cleanrooms Unlimited,
Uferweg 7,
9500 Villach, Austria

Bj
ci
cisat
cp,i
cp,chamber
C
C, C1,
C2
Di
f
g
h
h1+x
Hv,i
k
MWi
mchamber

N cond;i ci
Nl,i

Qloss
p
pi
pisat
ptot
~
R
R
Rgas
Ra

Parameters of the RedlichKister equation


[J/kmol]
Concentration of species i in the gas phase
[mg/l]
Saturation concentration of species i over the
liquid film [mg/l]
Heat capacity of species i in the gas phase
[kJ/kmol.K]
Heat capacity of the chamber wall material
[kJ/kg.K]
Dimensionless concentration of inlet gas
Constants for the turbulence model
Diffusion coefficient of species i in air [m2/s]
Target function for the dew point iteration [Pa]
Gravitational acceleration [m/s2]
Specific enthalpy [kJ/kmol]
Enthalpy [kJ/kgdry air]
Heat of vaporization of species i [kJ/kmol]
Turbulent kinetic energy [m2/s2]
Molecular weight of species i [g/mol]
Mass of the DHP chamber walls [kg]
Molar condensation rate of species i [kmol/s]
Molar amount of species i in the liquid phase
[kmol]
Heat loss [W]
Pressure [Pa]
Vapor pressure for species i in a liquid mixture
[Pa]
Vapor pressure of pure species i [Pa]
Total pressure [Pa]
Reynolds stress tensor [m2/s2]
Molar gas constant, 8.314472 [J/mol.K]
Gas constant for air, 287.05 [J/kg.K]
Rayleigh number

52

Sct
T
~
U
Vc

Vj
wi
X
xi
yi
Z
Greek
eff

8
i

air
t

k ,

J Pharm Innov (2009) 4:5162

turbulent Schmidt number


temperature [K]
Velocity vector [m/s]
Chamber volume [m3]
Volumetric flow rate [m3/s]
Mass fraction of species i
Absolute moisture content of the air
[g/kgdry air]
Molar fraction of species i in the liquid phase
Molar fraction of species i in the gas phase
Height level [m]
letters
Effective energy diffusion coefficient [W/m.K]
Heat transfer coefficient [W/m2.K]
Mass transfer coefficient [m2/s]
Energy dissipation rate [m2/s3]
Mass flux vector [kg/m2.s]
Activity coefficient for species i
Turbulent diffusion coefficient [kg/m.s]
Heat conductivity of air [W/m.K]
Turbulence viscosity [Pa.s]
Density [kg/m3]
Constants for the turbulence model

Introduction
In the (bio-)pharmaceutical industry, the efficient decontamination of surfaces in lock lines, rooms, and laboratories
is of central importance. The main challenge in these
applications is finding a reliable decontamination medium.
Currently, sterilization approaches based on gas-phase
decontamination are drawing significant interest. Use of
hydrogen peroxide as a decontamination medium since the
late 1980s has been attributed to its simplicity and its
environmentally friendly decontamination products of
water vapor and gaseous oxygen [1, 2].
In the past decade, decontamination by hydrogen
peroxide (DHP) has been used frequently for sterilization
purposes in a wide range of applications [38]. In these
processes, an aqueous hydrogen peroxide solution, typically 35% (w/w) H2O2, is evaporated, brought into contact
with a hot gas stream, and fed into the containment to be
sterilized. This process is often called gassing. Afterwards, the chamber (here, referred to as DHP chamber) is
purged with air until the hydrogen peroxide level is below
the OSHA safety level of 1 ppm. This process is referred to
as aeration.
There has been much interest in the theoretical analysis
of the evaporation of aqueous hydrogen peroxide solutions
due to its relevance for DHP systems [912]. However,
there are only a few practical guidelines for the proper

design and development of hydrogen peroxide decontamination systems. In this paper, we provide a guide on how
to accurately calculate the gasliquid equilibrium and
diagrams that may help engineers to successfully implement
DHP in cleanrooms. Furthermore, perspectives for improving DHP technology towards a more controllable and
reliable method for surface decontamination are presented.

Improved Prediction of the GasLiquid Equilibrium


For the description of the gasliquid equilibrium between
an aqueous hydrogen peroxide solution and its vapor, a
large amount of data exists [13, 14]. Often, the description
is based on the concept of activity coefficients for the
description of non-ideal liquid mixtures. Using this concept, the vapor pressure for water and hydrogen peroxide
over an aqueous solution of hydrogen peroxide can be
calculated following Eq. (1).
pi psat
i  g i  xi

Here, pi, i, and xi are the vapor pressure, activity


coefficient, and mole fraction, respectively, for species i
(being either water or hydrogen peroxide) in the mixture,
and pisat is the vapor pressure of the pure liquid of species i
at a given temperature. Values for pisat for water and
hydrogen peroxide can be obtained from various sources,
e.g., in Manatt and Manatt [15, 16].
Having the vapor pressure for species i calculated using
Eq. (1), the molar fraction of the species in the gas phase,
yi, can be calculated using Daltons law:
yi

pi
ptot

Here, ptot is the measurable total ambient pressure, which


is generally accepted as 1.013105 Pa. The assumption of
non-interacting species required for use of Daltons law is
justified by the fact that in hydrogen peroxide applications
the pressure is usually atmospheric or near atmospheric
conditions. Using the ideal gas law, the concentration of
species i in mg/L can be calculated as:
ci

pi  MWi
RT

Here, MWi is the molecular weight of species i (g/mol),


R is the molar gas constant (8.314472 J/mol.K), and T is the
absolute temperature in K.
For a given aqueous solution at a certain temperature,
pisat and xi are known and the activity coefficients i must
be calculated; this is most conveniently done by fitting
parameters of a theoretically derived function to experimental data. Recently, Manatt and Manatt [16] reviewed the

J Pharm Innov (2009) 4:5162

53

available correlations for the activity coefficients of


aqueous hydrogen peroxide solutions and discussed their
accuracy. They concluded that the initial parameter set of
Scatchard et al. [13], frequently used today, is inaccurate,
especially at low temperatures. For example, the deviation
from the exact vapor pressure at 25C and can be as high as
14%. This would translate into a 14% error in the hydrogen

peroxide concentration in the gas phase. Hence, for a more


accurate analysis of phenomena in DHP systems, it may be
necessary to use improved correlations.
A four-parametric activity coefficient model presented
by Manatt and Manatt [16] had been found to fit the data
more accurately. It is based on the RedlichKister equation
and can be written as:




 x2
B0 T B1 T  3  4  xH2 O
ln g H2 O2 H2 O 
B2 T  1  2  xH2 O  5  6  xH2 O B3 T  1  2  xH2 O 2 7  8  xH2 O
RT

ln g H2 O



1  xH2 O 2
B0 T B1 T  1  4  xH2 O


B2 T  1  2  xH2 O  1  6  xH2 O B3 T  1  2  xH2 O 2 1  8  xH2 O
RT

The four parameters B0 to B3 are temperature dependent


and details can be found in Manatt and Manatt [16]. It
should also be mentioned that the fitted parameters of the
activity coefficient model depend slightly on the correlation
used for determining the vapor pressure pisat. Therefore, one
should use the same vapor pressure correlations as given in
Manatt and Manatt [15].
To illustrate the accuracy of this set of equations, we
compare experimental data to the calculated values. The
experimental data date back to the work of Scatchard et al.
[13]. These researchers used a special adapted equilibrium
still to account for the characteristics of the hydrogen
peroxide solutions. The comparison of the total vapor
pressure and the molar fraction of hydrogen peroxide in the
gas phase is shown in Table 1. For this example, we have
kept the mole fraction of hydrogen peroxide in the liquid
phase constant (xH2O2 =0.7, equal to a mass fraction of
wH2O2 =0.815). The error between the calculation and the
experimental data is very small, typically below 0.2%.

Table 1 Comparison of experimental data from Scatchard et al. and


calculated values for the total pressure and the gas-phase composition
yH2O2 for a mole fraction of xH2O2 =0.7
Total pressure
[bar] (exp.)

Total pressure
[bar] (calc.)

yH2O2
(exp.)

yH2O2
(calc.)

10
20
30

0.002768
0.005457
0.010215

0.002768
0.005458
0.010222

0.2356
0.2489
0.2616

0.2355
0.2487
0.2613

40

0.018184

0.018204

0.2749

0.2744

Temperature
[C]

The correlations of Manatt and Manatt for aqueous hydrogen peroxide


solutions has been used for the calculations

A comparison of the calculated data, the experimental


data, and the data computed by Hultman et al. is given in
Table 2. The comparison is shown for a constant temperature of 25C and different compositions of the liquid
phase. Clearly, the correlation used by Hultman et al.
predicts values significantly below those of the experimental data. This is especially true for low liquid mass fractions
of hydrogen peroxide. Conversely, the correlation of
Manatt and Manatt, used in this work, shows much better
agreement with the experimental data.

Calculation of Condensate Composition and Dew Point


Using Eqs. (1), (2), (3), (4), and (5), one is able to calculate
the molar fraction of hydrogen peroxide and water vapor in
an inert gas using a suitable computer program. Furthermore,
it is also possible to calculate the composition of the
condensate in equilibrium with gas having a certain
composition. However, this can be only done by an iterative
procedure, as Eqs. (4) and (5) are non-linear with respect to
the liquid-phase composition. For example, starting with
activity coefficients equal to unity, one can estimate the
liquid-phase composition using Eqs. (1) and (2). Then, the
activity coefficients can be evaluated and the procedure is
repeated. The exact values are obtained after a few iterations.
Calculating the dew point, i.e., the temperature where a
given composition of the moist, H2O2-laden air starts to
condense, is a little more demanding. In this case, the
following procedure is recommended:
1. The dew point for water vapor should be calculated
based on the given molar fraction of water vapor. This

54

J Pharm Innov (2009) 4:5162

Table 2 Comparison of data on the gas-phase composition over an


aqueous hydrogen peroxide solution at 25C for different liquid phase
compositions
H2O2 mass
fraction
liquid phase

H2O2 mass
fraction gas phase
(Hultman et al.)

H2O2 mass
fraction gas
phase (exp.)

H2O2 mass
fraction gas
phase (calc.)

0.0187
0.0800
0.2410
0.3500
0.5640

0.0216
0.0964
0.2638
0.3307a
0.5684

0.0216
0.0961
0.2635
0.3231
0.5679

0.321
0.557
0.739
0.778
0.883
a

This value has been linearly interpolated from the experimental data

can be done using Eq. (2) and an appropriate vapor


pressure correlation for water. The obtained dew point
temperature is the starting value for the following
iteration.
2. The composition of the condensate should be calculated with the previously calculated dew point temperature from step 1. The activity coefficient for this first
calculation should be chosen to be unity.
3. The dew point temperature should be calculated using a
Newton iteration, i.e., the corrected temperature Tn+1 at
iteration n is
T n1 T n 

f
f0

The target function f for this iteration is:


X
X
f T ; xi
pi T ; x i 
yi  ptot
i

If Eq. (7) is equal to zero, the dew point has been found.
For evaluating the corrected temperature using the Newton
algorithm, we need the derivative f of the target function f
with respect to the temperature (see Eq. (6)). As ptot and yi
are constant, this involves the calculation of derivatives of
the pi(T, xi) terms. These derivatives are complex functions
of the temperature and the molar fractions xi which cannot
be evaluated easily. As a suitable approximation for the
derivative of the target function f, the derivative of the
vapor pressure correlation for water can be taken instead.
This converges quickly, although use of an appropriate
under-relaxation factor is suggested, as the algorithm can
get unstable in some cases.

hx Diagram for Hydrogen Peroxide in Air


As has been shown in the previous section, the evaluation
of equilibrium data is a complex task. It is therefore

instructive to use computer programs for this purpose.


However, it is also convenient to use appropriate diagrams
that can be used quickly for solving practical problems.
One of the most useful diagrams for solving such problems
in the heating, ventilating, and air conditioning (HVAC)
industry is Molliers hx diagram for moist air. This
diagram is commonly used for calculating processes
involving mixing, cooling, heating, or humidification of
airall processes encountered in DHP applications. Furthermore, bio-decontamination by hydrogen peroxide and
water vapor is of interest [11]. Hence, an hx diagram for
moist, H2O2-laden air would be beneficial for engineering
and developing DHP systems.
An hx diagram provides information on condensation
phenomena (which can be described with equilibrium data
as presented in the section entitled Improved Prediction of
the GasLiquid Equilibrium) and the enthalpy h1+x of
moist air in kJ/kgdry air. Here 1+x indicates that enthalpy
is calculated for both dry air and the moisture content X, but
on the basis of 1 kg of dry air, where X is the absolute
humidity in gH2O/kgdry air. The underlying assumptions of
an hx diagram are that (a) a possible condensate (i.e., fog)
is in thermal and thermodynamic equilibrium with the
gaseous phase, and (b) all phases, i.e., the gas and a
possible condensate, are well mixed.
For describing the enthalpy of the mixture of air, water,
and hydrogen peroxide vapor, we used data on the heat
capacity cp. In this work, cp was accurately modeled using
temperature-dependent functions. These functions have
been taken from Liley et al. [17, Table 2-198]. The specific
enthalpy can be found from these functions by simple
integration and subsequent summation over all involved
species (hydrogen peroxide, water, nitrogen, and oxygen).
The standard conditions for the enthalpy, i.e., the state were
the enthalpy is zero, was chosen to be 0C for both water
and hydrogen peroxide in the liquid state. For the heats of
vaporization at 0C (denoted as Hv,i), values of 45.025
and 53.200 MJ/kmol were used for water and hydrogen
peroxide, respectively. Hence, the specific enthalpy in kJ/
kmol of the moist air is given by
hT

X
i

0
@yi 

ZT

1
cp;i T dTA Hv;w  yw Hv;H2 O2  yH2 O2

8
Finally, this specific enthalpy was multiplied with the
number of moles in 1+X (kg) of moist air to find the value
for the enthalpy h1+x used in the diagram. It is noted here
that, in the case of fog formation, additional terms for the
precipitating liquid phase have to be considered. However,
this is not discussed here in detail as it is of little relevance
for typical DHP applications.

J Pharm Innov (2009) 4:5162

In an hx diagram for air and water vapor, lines of


constant temperature and constant relative humidity are
plotted. The enthalpy in hx diagrams for moist air is in kJ/
kgdry air and the absolute humidity X is denoted in gH2O/
kgdry air. These notations enable one to perform simple
calculations, e.g., those for a mixing process, directly in the
diagram. To expand on this concept, we define the absolute
moisture (including water and hydrogen peroxide) of the air
in g(H2O + H2O2)/kgdry air. As in the hx diagram for air and
water vapor, the lines of constant enthalpy have been
designed to have a constant negative slope such that the
isotherm at 0C is horizontal.
An hx diagram for a total pressure of 1105 Pa and
injection of an aqueous hydrogen peroxide solution of 35%
(w/w) is shown in Fig. 1. We have assumed that the air is
absolutely free of water before the hydrogen peroxide
Fig. 1 hx Diagram for moist,
hydrogen peroxide-laden air
(1105 Pa, 35 wt.% H2O2 in the
injected aqueous solution)

55

solution is injected and that no hydrogen peroxide is


decomposed during evaporation. Only under these assumptions is the ratio between water and hydrogen peroxide
vapor constant, which enables a simple construction of the
hx diagram. To facilitate the use of the diagram, the
hydrogen peroxide volume fraction in ppm has been added
as a second x-axis.
An example that illustrates the use of the diagram is
shown in Fig. 2. Starting with dry, H2O2-free air at state (1),
aqueous hydrogen peroxide solution is injected into this air
stream (12). This process is modeled as isenthalpic (i.e.,
constant enthalpy), because the enthalpy of the injected
liquid is typically very small compared to the energy
content of the air. Due to evaporation of the solution, the
temperature of the moist air after the injection (2) is
significantly lower than before (1) and is cooled down to

56

J Pharm Innov (2009) 4:5162


saturation

5 [%]

10 [%]

15 [%]

20 [%]

80
70 [C]

70

110

30 [%]

100

60 [C]

40 [%]
50 [%]
60 [%]

h1+x [kJ / kgdry air]

60
90

50 [C]

50

80 [%]
100 [%]

40 [C]
80

40

30 [C]

30

70

50

60

40

20

30

20 [C]

Fog

20

10
10

0
0

10

12

14

16

18

20

X [g(H2O + H2O2) / kgdry air]

Fig. 2 Injection of H2O2 (12), cooling (23), and mixing (345) in


the hx diagram for moist, hydrogen peroxide-laden air

ambient temperature (3). This process is observed in


practice when the sterilizing medium (the moist, H2O2laden air) is blown into the containment to be sterilized. In
such a case, rapid cooling of the sterilizing medium is
achieved by the walls of the containment to be sterilized. In
most cases, the process is designed to allow a maximum
relative humidity of 90% to ensure that no condensation
can occur; state (3) indicates a case close to 100%
saturation. Consequently, the injected amount of hydrogen
peroxide solution must be controlled with respect to the
temperature of the walls.
To illustrate a mixing process, we now focus on the
mixing of a high-moisture air stream (4) with air of state
(3). The rule for determining the mixing point (5) is that we
first have to connect the two initial states with a straight
line. This rule follows from the energy balance and is not
discussed here in detail. Second, the absolute moisture
content Xmix at the mixing point can be calculated from a
simple mass balance:
Xmix

air in a compartment that is already filled with sterilant.


Hence, care has to be taken in such cases.



X1  mdry air;1 X2  mdry air;2


mdry air;1 mdry air;2

Here, m dry air;j denotes the mass flow of dry air in the
stream j. In our case, we assumed to have equal amounts of
dry air in each of the two streams to be mixed. Here, state
(5) has an absolute moisture content of 12 g(H2O+H2O2)/
kgdry air and is fully defined. In this example, the mixing
point is in the fog region and condensate will form at the
walls of the containment. Thus, this example shows that it
is possible to obtain condensing conditions even if the two
initial gas streams are below saturation. This situation may
occur if gas with very high moisture content is mixed with

Mass and Heat Balances for a Well-Mixed Chamber


Despite the fact that the hx diagram presented in the
previous section may be beneficial for everyday practice, it
has its limitations. For example, it is valid only for a
constant ratio between water and hydrogen peroxide vapor.
As a consequence, the effect of pure water injection cannot
be studied. Also, changes of the gas composition during
condensation cannot be predicted, because more hydrogen
peroxide will condense compared to water vapor due to the
chemical equilibrium described in previous sections.
Hence, the ratio of these two species in the gas phase will
also change during condensation. In these cases, it is
necessary to use computer programs based on exact heat
and mass balances. Such programs have been developed inhouse and validated with experimental data. The novelty of
our computer program is that we take into account heat and
mass transfer to surrounding objects, e.g., walls.
In order to predict the conditions in the chamber, mass
and energy balances have to be solved in addition to the
chemical equilibrium discussed before. In DHP applications, the accumulation of hydrogen peroxide and water
vapor in the containment is important. Unsteady mass and
energy balances have to be used, and it is necessary to
model gas mixing in the containment, as mixing will
influence the concentration of the gas leaving the chamber.
Furthermore, heat and mass transfer has to be modeled. In
the current work, we have decided to follow two routes for
modeling these phenomena:
1. A simplified analysis in which we assume a perfectly
mixed gas phase, a uniform composition of a possible
condensate that forms, as well as constant heat and
mass transfer coefficients.
2. A refined analysis where we use computational fluid
dynamics (CFD) to predict gas-phase mixing, as well as
heat transfer to walls. In this analysis, we exclude mass
transfer, i.e., condensation, from our analysis.
In this section, we expand on the simplified analysis (the
refined analysis is described in the next section of our
paper) by assuming the simplest case of mixing, i.e., a
perfectly mixed chamber. In this case, the decontamination
chamber behaves similar to a continuously stirred tank
reactor (CSTR). In this case, the time-dependent material
balance for species i can be written as:

Vc 

 X 

dci X  


V j  ci;j 
V k  ci;k  N cond;i ci
dt
j
k

10

J Pharm Innov (2009) 4:5162

57

Here, Vc stands for the gas volume inside the chamber


and dci/dt is the time derivative of the molar concentration


of species i in the gas phase. V j and V k denote the
volumetric flow rate of the gas feed j and the effluent k,
respectively. ci,j and ci,k denote the molar concentrations in
those gas streams. We have included condensation effects

via N cond;i ci , which is the molar condensation rate for
species i. This is relevant for some DHP applications where
condensation is allowed. The condensation rate of hydrogen peroxide is



N cond;H2 O2 ci ADHP  bH2 O2  cH2 O2  csat
11
H2 O2
The condensation and evaporation rates of water can be
calculated in a similar way. Hence, the model described
here is applicable to each phase of the DHP process, i.e.,
gassing as well as aeration.
In Eq. (11), ADHP denotes the total inner surface area to
be decontaminated. The underlying assumption for reevaporation of possible condensation is that this total inner
surface is completely wetted with condensate. In Eq. (11),
H2O2 is the mass transfer coefficient, and cH2O2sat is the
saturation concentration of hydrogen peroxide above the
condensate film. Note that cH2O2sat can be calculated based
on the chemical equilibrium and the actual composition of
the condensate film. For the calculation of H2O2, however,
further approximations have to be made. In the current
work, we have calculated the mass transfer coefficient
using an analogy relation between heat and mass transfer
[18]. Specifically, we assume that the Nusselt and Sherwood numbers are equal for all species. This is a reasonable
approximation for the gases and conditions typically
observed in DHP chambers. Under this assumption we
can calculate H2O2 as:
bH2 O2

a  DH2 O2
lair

12

where denotes the heat transfer coefficient, DH2O2 is the


diffusion coefficient of hydrogen peroxide in air, and lair is
the heat conductivity of air. While DH2O2 and lair can be
taken from numerous chemical engineering textbooks or
correlations [17, 19], the heat transfer coefficient depends
on the specific flow conditions within the chamber.
However, the flow conditions are not known and can be
predicted only with the use of specialized techniques, e.g.,
CFD. In the current work, we have used a value for mixed
convection of =10 W/m2K.
To account for the accumulation of water and hydrogen
peroxide in the condensate film, we define a mass balance
in the liquid film for each species:
dNl;i

N cond;i ci
dt

13

Here Nl,i stands for the molar amount of species i in the


liquid film. Using Eq. (13), the composition of the
condensate film xi is:
Nl;i
xi P
Nl;i

14

When using Eq. (14), we assume a uniform composition


of the condensate film, i.e., an ideally mixed liquid phase.
Using the composition of the condensate film, the saturation concentration of hydrogen peroxide cH2O2sat can be
calculated and used in Eq. (11). By numerically integrating
Eqs. (10), (11), (12), (13), and (14), the concentration time
profiles in the gas phase, as well as in the condensate film,
can be determined.
Finally, the energy balance should be considered to
account for heating of surrounding walls and the gas in the
containment. This is important, as the chemical equilibrium
depends strongly on the temperature. For this purpose, we
have made the assumption that the gas in containment has
the same temperature as the surrounding walls. This is a
reasonable assumption, as the convective heat transfer to
the walls is relatively fast. Also, CFD simulations (to be
detailed in the final section of this paper) show that the
mean temperature is only negligibly higher than the wall
temperature. Hence, we can write an integral energy
balance over the control volume and the walls:

dT X 

mchamber  cp;chamber 

hT j  V j  ci;j
dt
j
X



hT k  V k  ci;k
15
k

Here, mchamber denotes the total mass of the surrounding


walls in thermal equilibrium with the gas, whereas the mass
of the contained gas is negligible. In Eq. (15), cp is the
known heat capacity of the surrounding walls (usually
made of steel). Furthermore, h(T)j and h(T)k denote the
specific enthalpy of the streams j and k. Also, Eq. (15) has
to be integrated over time to calculate the temperature
during the decontamination cycle. Furthermore, this integral energy balance accounts for the heat of condensation,
if any, as the latter is already considered in the calculation
of the specific enthalpies h(T)j and h(T)k.
Figure 3 shows example time profiles of condensate and
gas composition during aeration of a DHP chamber.
Figure 3a is a result of the calculation with the simplified
approach presented in this paper, whereas Fig. 3b refers to
measurements in an experimental setup involving a lab-scale
DHP chamber. The lab-scale DHP chamber consisted of a
temperature-controlled box made of stainless steel (0.5
0.50.5 m in dimension) connected to a H2O2 generator
(Geschko MLT 07, PEA GmbH, Calw, Germany). The

58

J Pharm Innov (2009) 4:5162

(a)

should be noted that our model was neither calibrated with


nor fitted to any experimental data. Hence, there is the
possibility to further improve the reliability of our model,
which is currently ongoing.

Perspectives for Improved DHP System Design

Gas Composition [ppm(v)]

1,400
1,200

(b)

H2O2(g)

1,000
800
600
400
200
0
11:33

11:36

11:39

11:42

11:45

11:48

11:51

Time [hh:min]

Fig. 3 Time profiles for gas and liquid phase composition during
aeration (condensation has occurred during the gassing phase). a
Calculation, b experimental data for the hydrogen peroxide concentration in the gas

chamber was designed and operated such that it resembled


the situation in a real industrial application of a lock line.
The H2O2 concentration was measured using an electrochemical sensor (an ATI C16 PortaSens II Portable Gas
Detector from Analytical Technology Inc., Collegeville, PA,
USA). The results shown in Fig. 3 refer to a situation where
condensation of hydrogen peroxide and water vapor has
already occurred in the DHP chamber. Consequently, the
condensate has to re-evaporate during aeration. As can be
seen from Fig. 3a, the predicted hydrogen peroxide and
water content of the gas phase reach their maximum after
about 1.5 min due to evaporation at the wall interface. A
similar phenomenon, i.e., an increase of the hydrogen
peroxide concentration during the aeration phase, has been
observed in the experimental setup (see Fig. 3b). However,
the time scale for evaporation in the experimental setup is
approximately twice that of the simulation, possibly attributed to only partial wetting of the walls and, consequently, a
smaller interfacial area for evaporation in the experiment. As
a result, the mass transfer rate to the air is smaller (refer to
Eq. (11)), and it takes longer for the condensate to reevaporate. Furthermore, as was illustrated with our calculations, as water evaporates more rapidly from the liquid
phase, the hydrogen peroxide content therein increases
abruptly and theoretically approaches 100% (see Fig. 3a). It

The program described is beneficial for quick estimates of


the conditions in a decontamination chamber. However, due
to the underlying assumption of an ideally mixed gas inside
the chamber, it cannot provide answers regarding the spatial
distribution of the decontamination medium and effects
such as local condensation. However, the local concentration and the existence of localized liquid phases determine
the sterilization kinetics of the microorganisms and,
consequently, the cycle time needed for a reliable decontamination. Clearly, for optimizing the decontamination
cycle, this spatial distribution is critical. For example, if the
gas distribution inside the containment to be decontaminated is unsatisfactory (e.g., due to dead zones), some areas
may not be treated with the decontamination medium
sufficiently. As a result, cycle times may need to increase
significantly based on the worst case (i.e., the location
where the concentration is lowest).
Gas distribution and mixing of gases is difficult to
predict, especially for complex geometries. Furthermore,
these processes are dominated by turbulence and, most
importantly, by buoyancy forces due to temperature
gradients. The only way to safely predict mixing in locks
and rooms is the use of sophisticated computational fluid
dynamics (CFD) simulations in combination with scalar
transport codes. To underline the effects of buoyancy in
DHP applications, we detail here on our studies involving
CFD simulations with an accurate modeling of buoyancy
effects. Our code is based on the open-source software
OpenFOAM [20] augmented by the meshing tool Cubit
[21]. In addition, a modified solution strategy has been
developed to handle buoyancy-driven flows more efficiently. We have used the steady-state Reynolds-Averaged
NavierStokes (RANS) equations to model the multicomponent gas flow in a typical decontamination chamber. The
model equations consist of the continuity equation:
~0
r  rU

16

as well as the steady-state momentum conservation


equation:
~ r  r~
r  U
R rp  rrgz
*

17

Here, p, g, , and R refer to the pressure, the


gravitational acceleration, the mass flux, and the Reynolds
stress tensor, respectively, and z denotes the local height

J Pharm Innov (2009) 4:5162

59

above an arbitrary reference level. The term gz models


the buoyancy force, i.e., the force due to differences in the
density . The mass flux vector is the product of the
~ and the density. The Reynolds stress
velocity vector U
~
tensor R mimics the turbulent fluid motion, which is not
directly calculated. Hence, the Reynolds stress term has to
be modeled. In our work, the k turbulence model has
been used. The selection of this turbulence model is
motivated by its economy and reasonable accuracy and
robustness for a wide range of turbulent flows. For the k
turbulence model, two additional transport equations for the
turbulent kinetic energy k, as well as for its dissipation rate
, have to be solved:



 * 
 *T 
*
mt
 r"
r  rU k r 
rk 2mt r U r U
sk
18



 * 
 * T 
*
mt
"
"2
 C2" r
r  rU " r 
r" C1" 2mt rU r U
k
s"
k

19
Here, the turbulence viscosity t is defined as:
mt r  Cm 

k2
"

20

The Reynolds stresses can be calculated using the


extended Boussinesq relationship:

 * T 
*
2
m
~
R  k  I  2  t rU rU
21
3
r
where I denotes the unity tensor. The model constants for
the k turbulence model were C =0.09, k =1.0, =1.3,
C1 =1.44, and C2 =1.92.
We have used wall functions to approximate the velocity
profile near the wall and to set the boundary conditions for
the turbulence model. With this approach, the wall shear
stress is calculated from the assumption of a logarithmic
velocity profile near the wall. This approach is suitable for
most industrial applications of CFD and is widely used in
that field.
In order to model buoyancy effects, the local density
has to be known. In our work, we have used the ideal gas
law to calculate :
p r  Rair  T

22

Here, Rair denotes the specific gas constant for air,


which has a value of 287.05 J/kgK. As pressure differences in the chamber will be very small (usually in the
order of several Pa), the influence of the pressure can be
safely neglected. Hence, the air in the chamber will be
treated as incompressible. Furthermore, the concentration

of water and hydrogen peroxide vapor is very small in


typical DHP applications (in the order of 0.1% to 0.3%)
and the local concentration will have only a small
influence on the density. This can also be seen from the
hx diagram (Fig. 1), in which lines of constant density are
almost parallel to lines with constant temperature. Thus, the
composition of the gas does not significantly alter the density
of the multicomponent gas mixture. The major influence on
the density will be from temperature differences, which is
modeled directly via the ideal gas law.
The information on the local temperature T is provided
from the steady-state energy conservation equation:
cp  r  T  r  aeff rT 0

23

Here, cp is the specific heat capacity under constant


pressure for air, which is assumed to be constant in our
CFD studies. The effective energy diffusion coefficient eff
is defined as:
aeff cp mt lair

24

In our study, we have used constant temperature


boundary conditions on all walls and the outlet, as well as
constant temperature at the inlet. Steady-state equations for
gas flow and temperature distribution in the chamber have
been used. This enables very time-efficient simulations
needed in industrial practice. We have also quantified the
effect of the transient development of the velocity and
temperature field in a DHP chamber and found that the
flow is fully developed within a few minutes. This is much
shorter than usual gassing times that can be up to a few
hours for large rooms.
To study the concentration distribution of H2O2 vapor
injected into the chamber, the assumption that H2O2 vapor
is inert has been adopted. This limits our work to DHP
systems where condensation and re-evaporation in the
chamber is excluded, i.e., so-called dry systems. However,
from the local distribution of hydrogen peroxide vapor in
the gas phase predicted with our CFD simulations, critical
areas prone to condensation effects can be anticipated.
Consequently, we write the species transport equation for
an inert scalar C as follows:
@rC
r  C  r   rC 0
@t

25

Here, C and represent the dimensionless concentration,


i.e., the concentration of H2O2 normalized with the inlet
concentration, and the turbulent diffusion coefficient,
respectively. is represented by the following equation:

mt
Sct

26

where the value of the turbulent Schmidt number Sct is 0.7


[22]. We have used the zero gradient boundary condition on

60

all walls and the outlet, as well as a constant composition


boundary condition at the inlet. After successful implementation of the algorithm into the CFD software, validation of
the code was performed. This was done by comparison of
our results with experimental results, as well as with a
reference numerical solution of Corvaro and Paroncini [23]
for a buoyancy-driven flow. Details on this validation are
documented in the Appendix of this paper.
Selected snapshots of some simulations conducted by us
are shown in Fig. 4 and Fig. 5. For this purpose, we have
modeled a typical industrial DHP chamber with a nominal
volume of 3 m3 and a gassing flow rate of 20 m3/h. The
simulations were performed on a sufficiently fine numerical grid to give a grid-independent solution. In Fig. 4a,
the flow field for an isothermal inlet condition, i.e., the gas
flowing into the DHP chamber has the same temperature
as the chamber itself, is illustrated. A plot of the local
distribution of the magnitude of the velocity field
illustrates convective transport in the system. In contrast
to this isothermal flow, we show the flow field for an inlet
temperature of 75C in Fig. 4b. This temperature is typical
for most dry hydrogen peroxide decontamination systems where the air is preheated to avoid condensation in
the gas piping and the DHP chamber. Clearly, this change
in the inlet temperature causes a significant alteration of
the flow field. In Fig. 4b, the main flow direction is
towards the top region of the chamber due to the lower
density of the warm inlet air. In the bottom region of the
chamber, there is almost no gas motion; here, no
convective transport is possible.
The corresponding distribution of the inlet gas is
shown in Fig. 5. As expected, the distribution of the inlet
gas reflects also the low convective transport in the case of
a higher gas inlet temperature. In conclusion, our
simulations show that it is critical to take buoyancy
effects into account when dealing with gas flow in DHP
chambers.
Fig. 4 Comparison of the magnitude of the velocity vector in
an industrial lock line during
gassing using CFD. In the left
figure (a), the gas has been
injected with the same temperature as the air in the lock line
(no buoyancy effects), whereas
in the right figure (b), the inlet
temperature was 75C (strong
buoyancy effects)

J Pharm Innov (2009) 4:5162

Conclusion
The engineering of an optimal DHP system is a difficult
task. As has been shown in this work, buoyancy effects
play a significant role in DHP chambers operated at typical
process conditions, i.e., at inlet temperatures of approximately 75C. Such effects completely change the flow
pattern and, consequently, the distribution of the inlet gas.
Hence, we suggest paying attention to given process
parameters like temperature, flow rates, and the H2O2 inlet
concentration during the engineering phase of DHP
systems.
In the current work, we have focused on different
modeling strategies for DHP systems. In our models, we
have incorporated an advanced description of the chemical
equilibrium of aqueous hydrogen peroxide solutions. We
demonstrated how this improvement for previously published models leads to more reliable calculations for the
water/hydrogen peroxide system. Furthermore, we have
illustrated that even a relatively simple model assuming a
well-mixed gas phase in a DHP chamber can reproduce
experimental data qualitatively well, and that an hx
diagram can be constructed for moist, hydrogen peroxideladen air. This diagram will be of benefit for both engineers
and end users of DHP systems. However, future work will
have to consider partial wetting of the chambers internals
to quantitatively predict complex phenomena like reevaporation of condensate.
For an optimized design and inclusion of DHP systems
in clean rooms, it is clear that CFD must become a standard
tool to predict flow and mixing in those systems. As in
other industries, where CFD is already well established, the
complex interactions between fluid dynamics, heat transfer,
turbulence, and buoyancy forces will force engineers to use
more sophisticated tools to design DHP systems. However,
as the current work is one of the first systematic studies in
this area, we were not able to incorporate all aspects of flow

U [m/s]

(a)

U [m/s]

(b)

J Pharm Innov (2009) 4:5162

61

Fig. 5 Comparison of the


dimensionless concentration of
the inlet gas in an industrial lock
line during gassing using CFD.
a Isothermal inlet conditions
(no buoyancy effects), b inlet
temperature of 75C (strong
buoyancy effects)

(a)

as well as heat and mass transfer into our CFD simulations.


Clearly, the detailed modeling of the condensation of the
hydrogen peroxide/water mixture out of the inert gas (i.e.,
air) still remains a challenge. Furthermore, more elaborate
validation of CFD codes, as well as more advanced studies
on the effect of different turbulence models, will be
important to increase the reliability of such engineering
tools.
Acknowledgement The authors acknowledge financial support of
Ortner Cleanrooms Unlimited GmbH.

Appendix
In order to validate the developed solver, a comparison of
numerical and experimental has been performed. Therefore,
Fig. 6 Numerical results for a
the temperature distribution and
b the magnitude of the velocity
field in a heated cavity

(b)

our results have been compared with the results from


Corvaro and Paroncini [23]. The setup consisted of a square
cavity (with height of 0.05 m and a total length of 0.41 m)
with a heated strip placed on the bottom wall. In the case
studied in our work, the heated strip was positioned at the
center of the bottom wall. The fluid in the cavity is air and
the heated strip was made of brass. The side walls of the
cavity were cooled and had a constant temperature of
285.5 K. The temperature of the heated strip was 302.1 K,
such that the resulting Rayleigh number was Ra=2.02105.
Figure 6 shows the temperature and velocity distribution
obtained with the simulations code detailed in this work.
These results agree very well qualitatively with the
experimental measurements, as well as the simulation
results of Corvaro and Paroncini [23]. The computed mean
dimensionless heat transfer coefficient (i.e., the mean
Nusselt number Num) averaged over the heated strip using

62

our calculations was determined as Num =6.16. From


simulations reported by Corvaro and Paroncini [23],
Num =6.28, whereas the experimental result was Num =
6.45. Hence, the differences are below 5%, which is
acceptable.

References
1. Klapes NA, Vesley D. Vapor-phase hydrogen-peroxide as a
surface decontaminant and sterilant. Appl Environ Microbiol.
1990;56:5036.
2. Wang J, Toledo RT. Sporicidal properties of mixtures of hydrogenperoxide vapor and hot air. Food Technol. 1986;40:627.
3. McAnoy AM, Sait M, Pantelidis S. Establishment of a vaporous
hydrogen peroxide bio-decontamination capability. Report DSTOTR-1994. Human Protection Performance Division DSTO. 2007.
4. Johnston MD, Lawson S, Otter JA. Evaluation of hydrogen
peroxide vapour as a method for the decontamination of surfaces
contaminated with Clostridium botulinum spores. J Microbiol
Methods. 2005;60:40311.
5. Rohtagi N, Schuber W, Koukol R, Foster TL, Stabekis PD.
Certification of vapor phase hydrogen peroxide sterilization
process for spacecraft application. Report 02ICES-57. Society of
Automotive Engineers, Inc. 2001.
6. Heckert RA, Best M, Jordan LT, Dulac GC, Eddington DL,
Sterritt WG. Efficacy of vaporized hydrogen peroxide against
exotic animal viruses. Appl Environ Microbiol. 1997;63:39168.
7. Hall L, Otter JA, Chewins J, Wengenack NL. Use of hydrogen
peroxide vapor for deactivation of Mycobacterium tuberculosis in
a biological safety cabinet and a room. J Clin Microbiol.
2007;45:8105.
8. Fischer J, Caputo RA. Barrier isolator decontamination systems.
Pharm Technol 2004; Issue Nov 2004:6882.
9. Hultman C, Hill A, McDonnel G. The physical chemistry of
decontamination with gaseous hydrogen peroxide. Pharm Eng.
2007;27:2232.

J Pharm Innov (2009) 4:5162


10. Watling D, Parks M. The relationship between saturated hydrogen
peroxide, water vapour and temperature. Pharm Technol Eur
2004; 01-03-2004
11. Unger-Bimczok B, Kottke V, Hertel C, Rauschnabel J. The influence
of humidity, hydrogen peroxide concentration, and condensation on
the inactivation of Geobacillus stearothermophilus spores with
hydrogen peroxide vapor. J Pharm Innov. 2008;3:12333.
12. Watling, D. Theory and Practice of Hydrogen Peroxide Vapour. In:
Pharmaceutical International. Copybook Solutions Ltd. 2007. http://
www.pharmaceutical-int.com/categories/biodecontamination-usinghydrogen-peroxide-vapour/theory-and-practice-of-hydrogen-peroxidevapour.asp
13. Scatchard G, Kavanagh GM, Ticknor LB. Vapor liquid equilibrium. 8. Hydrogen peroxide water mixtures. J Am Chem Soc.
1952;74:371520.
14. Schumb WC, Satterfield CN, Wentworth RL. Hydrogen peroxide.
New York: Reinhold; 1955.
15. Manatt SL, Manatt MRR. On the analyses of mixture vapor
pressure data: the hydrogen peroxide/water system and its excess
thermodynamic functionscorrigendum. Chemistry. 2006;12:3695.
16. Manatt SL, Manatt MRR. On the analyses of mixture vapor
pressure data: the hydrogen peroxide/water system and its excess
thermodynamic functions. Chemistry. 2004;10:654057.
17. Liley PE, Thomson GH, Friend DG, Daubert TE, Buck E. Physical
and chemical data. In: Perry RH, Green DW, Maloney JO, editors.
Perrys chemical engineers handbook. New York: McGraw-Hill;
1999.
18. Bird RB, Steward W, Lightfood EN. Transport phenomena. New
York: Wiley; 2002.
19. Reid RC, Prausnitz JM, Poling BE. The properties of gases and
liquids. 4th ed. New York: McGraw-Hill; 1987.
20. OpenCFD Ltd. OpenFOAMThe Open Source CFD toolbox,
version 1.4. Berkshire, UK: OpenCFD Ltd.; 2007.
21. Sandia National Laboratories. Cubit 10.2 User Documentation.
Albuquerque: Sandia Corporation; 2007.
22. Fox RO. Computational models for turbulent reacting flows.
Cambridge: Cambridge University Press; 2003.
23. Corvaro F, Paroncini M. A numerical and experimental analysis on
the natural convective heat transfer of a small heating strip located
on the floor of a square cavity. Appl Therm Eng. 2008;28:2535.

You might also like