Академический Документы
Профессиональный Документы
Культура Документы
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
a r t i c l e
i n f o
Article history:
Received 12 November 2013
Received in revised form 10 March 2014
Accepted 16 March 2014
Available online 2 April 2014
Keywords:
Hydrogen peroxide
Direct synthesis
Palladium catalyst
Kinetics
Product distribution analysis
Batch reactor modelling
a b s t r a c t
The direct synthesis of hydrogen peroxide from molecular hydrogen and oxygen on a supported palladium catalyst was studied at 258297 K in a laboratory-scale batch reactor. The catalyst was in the form
of nely dispersed slurry in methanol/CO2 to suppress the internal and external mass transfer resistances. Experiments carried out under kinetic control revealed that hydrogen peroxide was successfully
formed on the catalyst surface, but it was hydrogenated as the reaction time was prolonged. The mass
balances of the components were considered in detail and a reaction mechanism was proposed, based
on the competitive adsorption of hydrogen and oxygen on the palladium surface. The surface reactions
leading to the formation of hydrogen peroxide and water were assumed to be rate determining, and
the rate equations describing direct synthesis, water formation as well as peroxide hydrogenation and
decomposition were derived. A special kind of product distribution analysis was used to interpret the
kinetic phenomena and to make the estimation of the kinetic parameters very robust. The parameters
were estimated by nonlinear regression analysis and the model gave a good t to the experimental data.
The usefulness of the product distribution analysis was clearly demonstrated.
2014 Elsevier B.V. All rights reserved.
1. Introduction
The currently applied industrial process for the production of
hydrogen peroxide, the anthraquinone process, is based on successive hydrogenation and oxidation of quinonic components. Despite
the high yield of hydrogen peroxide per cycle, the process major
disadvantages are the side reactions, requiring regeneration of both
the working solution and the hydrogenation catalyst, and the several steps necessary for the purication and concentration of the
peroxide [1]. For both these reasons, the cost of the hydrogen peroxide is nowadays relatively high, limiting the industrial large scale
use of H2 O2 . Moreover, the current process has high capital and
operation costs, and is suitable for real large-scale operation only.
In future, the interest towards on-site production of chemicals in a
smaller scale is predicted to increase. This would require a simpler
process for the synthesis of hydrogen peroxide.
Notation
A
A
c
c*
D
Ea
K
k
k
N
n
nexp
R
r
ri
T
t
V
interfacial area
frequency factor
concentration
concentration of a surface species
denominator in rate expression
activation energy
equilibrium constant
reaction rate constant
merged rate constant, product of rate constant,
adsorption equilibrium constant and total concentration of active sites
ux
amount of substance
number of experimental data (Eqs. (44) and (48))
gas constant
reaction rate
component (i) generation rate
temperature
time
volume
gas-to-liquid volume ratio
parameter in product distribution analysis
stoichiometric number
parameter in product distribution analysis
Deguchi et al. [7] were the only one taking into account the
adsorption of the promoters in the kinetic study. Therefore, we
constructed a special batch reactor system to determine the very
precise kinetics of hydrogen peroxide synthesis and decomposition. The reaction rates and product distribution were measured at
different temperatures and partial pressures of hydrogen and oxygen to reveal the kinetic phenomena [8]. The goal of this work is to
improve our previous kinetic analysis [8] and obtain a more rapid
and reliable estimation of the kinetic parameters, neglecting mass
transfer limitations and negligible reactions. The time dependence
of the concentrations was also eliminated with the aid of product
distribution analysis. In particular, the product distribution analysis was applied in order to improve the reliability of the parameter
estimation in this multicomponent system of composite reactions.
2. Experimental
The batch reactor was a 600 ml unbafed autoclave (Bchi)
equipped with a self-sucking six-blade impeller. Typically, 0.15 of
a commercial 5 wt% Pd/C catalyst was loaded in the reactor vessel. Methanol expanded with carbon dioxide was used as solvent
in all experiments. Carbon dioxide and oxygen were introduced
to the vessel, after which 400 ml of methanol was injected and
109
(DS)
H2 + (1/2)O2 = H2 O (WF)
H2 O2 + H2 = 2H2 O (H)
H2 O2 = H2 O + (1/2)O2
(DE)
110
Table 1
Reaction steps, reaction routes and stoichiometric numbers.
Step #
H2 + 2* = 2H*
O2 + * = O2 *
O2 + 2* = 2O*
O2 * + H* = OOH* + *
OOH* + H* = HOOH* + *
HOOH* = H2 O2 + *
O* + H* = OH* + *
OH* + H* = HOH* + *
HOH* = H2 O + *
HOOH* + * = HOH* + O*
HOOH* + H* = OH* + HOH*
Step #
DS
WF
H
DE
I
II
III
IV
V
VI
VII
VIII
IX
X
XI
1
1
0
1
1
1
0
0
0
0
0
1
0
1/2
0
0
0
1
1
1
0
0
1
0
0
0
0
1
0
1
2
0
1
0
0
1/2
0
0
1
0
0
1
1
0
Some essential features can, however, be extracted from the previous studies. Hydrogen and oxygen are known to adsorb on Pd
surfaces. Oxygen co-exists on Pd surfaces in molecularly adsorbed
and atomic forms; the rst one might being active in the direct synthesis and second one in the water formation. We cannot exclude
that the uppermost atomic layers of the solid catalyst surface
change during the reaction, since oxygen is typically present in
excess compared to hydrogen in the reaction system. This could
be conrmed by XPS analysis of fresh and used Pd catalysts.
Still, we are in the situation that a water-proof evidence on
the true mechanism on the catalyst surface does not exist, but the
derivation of plausible rate equations has to be based on reasonable
hypotheses about the adsorption, surface reaction and desorption processes. Several basic assumptions are introduced here to
describe the rate equations for the overall reactions (DS) to (DE) in
as simple as possible manner. Hydrogen is assumed to adsorb dissociatively on the metal surface, while oxygen co-exists in molecular
and dissociated form on the surface. Surface hydroxyl groups are
formed and they play a key role in the formation of both hydrogen peroxide and water; hydrogen peroxide and water adsorb on
the metal surface. The decomposition of hydrogen peroxide in the
absence of the catalyst was neglected, since it did not play any
role under the current experimental conditions. Based on these
assumptions, the reaction mechanism the adsorption, desorption
and surface reaction steps along with the stoichiometric numbers
() are summarized in Table 1. By combining each reaction step
with the corresponding stoichiometric number along the reaction
routes, the four overall reactions are obtained (DS to DE).
The addition of the reaction steps gives the overall reactions
described above (DS to DE).
The table illustrates the complexity of the reaction mechanism:
11 steps are needed in total to explain the processes on the catalyst
surface. Note that in principle also other steps are possible to give
the same intermediate. However, though not explicitly given, they
can be obtained by combinations of the reactions in Table 1. For
instance, the breaking of the O O bond in the OOH* intermediate
is obtained by combination of step V and X: OOH* + H* = HOOH* + *,
HOOH* + * = HOH* + O*. The complete mechanism is difcult to
apply in practice, because it comprises so many adsorption and
kinetic parameters, which cannot be determined separately. Therefore, a further step is taken and some of the reaction steps are
merged to obtain a simplied mechanism, which is displayed in
Table 2. Specically, the hydrogenation steps were assumed to be
very fast, because they are known to be very favourable over heterogeneous Pd catalysts. Hence, they were assumed to occur in a
single step.
The adsorption and desorption steps are assumed to be rapid
enough to reach quasi-equilibria, while the surface reaction steps
are presumed to be slow steps, which limit the rates. It is in principle possible that the adsorption and/or desorption are the limiting
H2 + 2* = 2H*
O2 + * = O2 *
O2 + 2* = 2O*
O2 * + 2H* = HOOH* + 2*
HOOH* = H2 O2 + *
O* + 2H* = HOH* + 2*
HOH* = H2 O + *
HOOH* + 2H* =2HOH* + *
HOOH* + * = HOH* + O*
I
II
III
IV and V
VI
VII and VIII (combined step)
IX
VIII and XI (combined step)
X
steps of the process. Nonetheless, the quasi-equilibrium assumption has already proven to give reliable results using the same Pd
catalyst [8]. An investigation on the adsorption/desorption steps
is certainly desirable, and our research group is at the moment
investigating this possibility. However, this investigation goes well
beyond the scope of this work, mainly focused on a rapid and reliable estimation of the kinetic constants. Once a more accurate
reaction mechanism has been proven, the same technique presented in this work could be used to estimate the new reaction
parameters. The rate-limiting steps are assumed to be irreversible,
because the equilibria of the overall reactions (DS to DE) are
strongly shifted to the side of the products.
The rates of the rate-limiting steps can now be written as
2
r1 = k1 cO2 cH
2
r2 = k2 cO cH
(1)
(step IVV)
(2)
(steps VIIVIII)
2
r3 = k3 cHOOH cH
r4 = k4 cHOOH c
(3)
(steps VIIIIX)
(4)
(step X)
j = O2 , H2 O, H2 O2
(5)
k = O, H
(6)
(7)
After inserting the expressions (5) and (6) in Eq. (7) and solving the
concentration of vacant sites (c* ), we obtain
1
c
1/2
1/2
= (1 + (KH cH2 )
+ (KO cO )
+ KO2 cO2 + KH2 O cH2 O + KH2 O2 cH2 O2 ) = D1
cTOT
(8)
Eq. (8) is inserted into Eqs. (5) and (6) which are inserted in rate
equations (1)-(4). The nal forms of the rate equations become
r1 =
r2 =
r3 =
r4 =
(9)
(10)
(11)
(12)
1
2
r2 +
(13)
1
2
r4
(14)
rP = r1 r3 r4
(15)
rW = r2 + 2r3 + r4
(16)
(17)
rO2 = r1
(18)
1
2
r2
(19)
rW = r2 + 2r3
(20)
dnGi
dt
dnLi
B ri VL + Ni A =
dt
(21)
(22)
dcLi
= (i + 1)1 B ri
dt
This expression can be elaborated further, provided that the interfacial mass transfer is rapid compared to the chemical reaction rates.
The amounts of substance are dened by
(27)
where i is given by
i =
Ki VG
VL
(28)
1
dcGi
i
= Ki
+1
Ki
dt
B ri
(29)
= (O2 + 1)1 B r1
dcH2 O2
dt
dcH2 O
dt
1
2
r2
(30)
(31)
= B (r1 r3 )
(32)
= B (r2 + 2r3 )
(33)
By regarding at the above equations, some interesting stoichiometric relationships are immediately noticed, such as
(1 + H2 )dcH2
dt
dcH2 O2
dt
dcH2 O
dt
=0
(34)
(35)
(23)
(26)
where Ki is the equilibrium ratio, which depends on the composition, total pressure and temperature. The values of Ki were
estimated according to our previous study on vapourliquid equilibrium in this system [10]. Values of Ki for the key components
O2 and H2 were in the range 3.54.5 and 6.116, respectively. The
validity of the hypothesis of rapid gasliquid mass transfer was
conrmed in a previous paper of Gemo et al. [8]. The gas and liquid volumes can be assumed constant in the present case, since the
liquid volume is in fact determined by the solvent, methanol. After
inserting the relations (24)(26) in the balance equation (23) and
carrying out the differentiation, an explicit differential equation is
obtained for the liquid-phase concentrations:
dt
rP = r1 r3
111
dcH2 O2
dt
(1 + H2 )dcH2
dt
(36)
(37)
nGi = cGi VG
(24)
where c0O represents the initial concentration of dissolved oxygen. A comparison of Eqs. (35) and (37) reveals that the hydrogen
concentration can be easily eliminated:
nLi = cLi VL
(25)
(38)
112
r1 r3
r2 + 2r3
(39)
k2 cO
(40)
+ 2k3 cH2 O2
Eq. (40) also suggests that the momentaneous product distribution is independent of the hydrogen concentration, provided that
the changes in the state of the catalyst surface are negligible. This
is actually a direct consequence of the assumed reaction mechanism: all the reactions have the same dependence on cH2 , so that
the ratio between the variation of H2 O2 and H2 O is independent
of the hydrogen concentration. If a different reaction mechanism
is assumed, the momentaneous product distribution may become
dependent of cH2 . The stoichiometric relationship (38) gives the
concentration of dissolved oxygen:
cO2 = c0O2 (1 + O2 )1
c
H2 O2
+ cH2 O
(41)
(42)
cO2 cH2 O2
1/2
cO
(43)
+ 2cH2 O2
The estimation of the kinetic parameters k1 , k2 and k3 requires the
simultaneous solution of Eqs. (32), (33), (41)(43). Note that the
product distribution analysis (Eq. (43)) is a function of water concentration, whereas the mass balances (32) and (33) are functions
of time. Hence, the regression was necessarily carried out in two
subsequent steps: the parameters and were determined rst
(step 1), and their values were subsequently used to determine the
parameter k2 (step 2). Specically, the parameters and were
estimated solving the low index algebraic-differential equations
system given by Eqs. (41) and (43) (step 1). The hydrogen peroxide
concentration was determined as a function of water concentration by non-linear regression analysis using the experimental data
available from the batch reactor. The following error function was
used for each experiment (i.e. any given temperature) to t the
experimental data:
err =
nexp
P
i=1
exp
2 O2 ,i
|cH
exp
)
2 O2
(1/nH
cH2 O2 ,i |2
exp
n
P
i=1
(44)
exp
2 O2 ,i
cH
Note that the error between experimental and calculated concentrations has been rescaled.
Then, the rate equations (9)(11) were rewritten as a function
of the parameters and :
r1 = k2 cH2 cO2
1/2
r2 = k2 cH2 cO
(45)
(46)
113
Fig. 2. Arrhenius plots of the rate constants (data from Fig. 1).
(47)
Note that the denominator D was neglected, as previously mentioned. Eqs. (45)(47) were introduced in the mass balances (32)
and (33), that were solved with the Eqs. (41) and (42) simultaneously (step 2). Eqs. (32), (33), (41) and (42) lead to an
algebraic-differential equations system, whose solution yield the
evolution in time of the concentration of the reactants (i.e. H2 and
O2 ) and the products (i.e. H2 O2 and H2 O). Note that in this system
only the parameter k2 has to be determined, since the parameters
and were calculated in step 1. Hence, the value of k2 was estimated in each experiment tting the experimental data with the
following error function:
err =
nexp
P
exp
|cH O ,i
i=1
2 2
exp
)
2 O2
(1/nH
cH2 O2 ,i
exp
n
P
i=1
exp
2 O2 ,i
cH
|2
exp
nP
i=1
exp
2 O,i
|cH
exp
)
2O
(1/nH
cH2 O,i |2
exp
n
P
i=1
exp
2 O,i
cH
(48)
Once again, the errors between experimental and calculated concentrations have been rescaled. Finally, the rate parameters k1 =
k2 and k3 = k2 are calculated. All the integrations were efciently carried out using the ode15s ADEs solver in Matlab, also
suitable for stiff equations, being based on a multistep variable
order method based on the numerical differentiation formulae.
The regressions were independently carried out with the experimental data [9] collected at 258, 268, 273, 283 and 297 K, so that
the temperature dependence of the parameters was checked with
Arrhenius plots.
5. Estimation results and discussion
The parameter estimation was carried out successfully by
applying the methodology which was described above. All the
experimental data are taken from our previous work [9]. The errors
of the parameters were relatively small and the overall degree of
explanation was high (exceeding 95% in all cases).
The parameter estimation results are presented in Fig. 1ae,
along with simulated concentrations of hydrogen and oxygen in the
liquid phase. As the gure reveals, the proposed model very truly
describes the experimentally observed concentrations of hydrogen
peroxide, water and hydrogen in the liquid phase. The simulated
liquid-phase concentrations of dissolved hydrogen and oxygen
show a declining trend; this is expected, since the experiments
were done batchwise. The product distribution (hydrogen peroxide versus water) is predicted very correctly in most cases, which
114
[16] G. Blanco-Brieva, E. Cano-Serrano, J. Campos-Martin, J.L.G. Fierro, Chem. Commun. (2004) 1184.
[17] S. Melada, R. Rioda, F. Menegazzo, F. Pinna, G. Strukul, J. Catal. 239 (2006)
422.
[18] Q. Liu, K. Gath, J. Bauer, R. Schaak, J. Lunsford, Catal. Lett. 132 (2009) 342.
[19] U. Rossi, S. Zancanella, L. Artiglia, G. Granozzi, P. Canu, Chem. Eng. J. 207208
(2012) 845.
[20] C. Samanta, Appl. Catal. A: Gen. 350 (2008) 133.