Sie sind auf Seite 1von 9

Article

pubs.acs.org/IECR

Modeling of Hydrogenation of Nitrate in Water on Pd−Sn/Al2O3


Catalyst: Estimation of Microkinetic Parameters and Transport
Phenomena Properties
Elém Patrícia Alves Rocha, Fabio Barboza Passos, and Fernando Cunha Peixoto*
Departamento de Engenharia Química e de Petróleo, Universidade Federal Fluminense, Niterói, Rio de Janerio 24210-900, Brazil

ABSTRACT: Excess nitrate in water is a known environmental problem, the remediation of which can be accomplished by
catalytic reduction of nitrate to N2 and NH4+. This work presents a model for the microkinetic modeling of a system that uses a
Pd−Sn/γ-Al2O3 catalyst taking into account the inherent transport phenomena. The pH control, which was carried out by
flowing CO2, was also modeled, leading to a considerably large (and stiff) system of ordinary differential equations, which was
dependent on a set of empirical parameters to be fitted. This regression was conducted using a maximum statistical likelihood
criterion, employing tailor-made optimization techniques. The results indicated mass-transfer effects should be considered to
obtain a complete description of the reaction system, especially regarding the pH profile.

1. INTRODUCTION reaction conditions, which indicates that special attention must


Underground water usage is a crucial factor for the subsistence be given to operational variables to avoid secondary pollutants
and food safety of 1.2 to 1.5 billion families living in rural and/ in the liquid phase (NO2− and NH4+) or in the gaseous phase
or poor areas of Asia and Africa, as well as for the internal (N2O). Therefore, a reliable model for the reaction mechanism
supply for a great part of the world’s population.1 Brazilian is important for predictive purposes.12,14
authorities estimate that 51% of the potable water in Brazil is Different mechanisms for this system can be found in the
obtained from this source.2 However, a gradual deterioration of literature. Basically, for palladium bimetallic catalysts, two
water quality from these sources is being observed throughout consecutive steps are usually proposed on the basis of
the world as a consequence of increasing nitrate concentration.3 experimental evidence: the reduction of nitrate, adsorbed on
Underground water nitrate concentrations can naturally bimetallic sites, to nitrite and the further reduction of nitrite to
range from 0.1 to 10 mg/L, and contaminated sources can nitrogen and/or ammonium on palladium monometallic
exhibit levels as high as 1000 mg/L.4 Nitrate intake leads to sites.15−17
nitrite, which can combine with blood hemoglobin and reduce The first step is usually modeled by a redox mechanism in
hemoglobin’s ability to carry oxygen to body cells, a disease which an interaction between the nitrate and the bimetallic site
known as methemoglobinemia. Children, especially those takes place;18 nitrate is converted to an intermediate, and the
under 6 years old, are very susceptible to this disease because promoter metal is oxidized and then regenerated by adsorbed
of some bacteria commonly found in their digestive systems hydrogen. The second step is believed to proceed easily over
that are able to convert nitrate to nitrite.5,6 Nitrate can also palladium monometallic sites, with the formation of adsorbed
interact with secondary amines to form N-nitrosamines, which intermediates (NO*, NH*). The NO* species is believed to be
are carcinogenic.6 important in the formation of hydrogen, while NH* would lead
The main causes of the increase in nitrate contamination are to the formation of NH4+.12
related to human activities, especially the use of nitrogen-based Prusse et al.12 reported the presence of N2O in the gaseous
fertilizers and inadequate sewage storage and/or transportation. phase, and its formation is considered, by some authors, to be
One of the technologies under investigation to solve this the main source of N2 formation.19,20 Ebbesen et al.21 did not
problem is the catalytic reduction of nitrate,7,8 which exhibits find any experimental evidence of N2O generation as a
environmental and economic advantages when compared to byproduct but indicated that this species could be an
other methods.9 intermediate that reduces quickly to N2. Wärna et al.,16 on
Briefly, in the catalytic removal of nitrate, contaminated the basis of kinetic studies, proposed that N2 formation takes
water is treated by a reductant, such as formic acid or hydrogen place through the decomposition of the NH* intermediate,
in the presence of a metallic catalyst, converted to nitrogen and generated through a reaction of NO* and H*. Ilinith et al.12
ammonium ion, which in turn is an undesirable side also stated that nitrogen is formed from NH* decomposition
product.10,11 Several works have pointed out that bimetallic but claimed that the reaction suggested by Wärna et al.16 is
catalysts, such as Pd−Cu10 or Pd−Sn11−13 are more effective, as unlikely. For Ilinith et al.,12 the interaction between NO* and
they are more active and selective for the reduction reaction.
The presence of bimetallic ensembles is essential for the Received: February 25, 2014
occurrence of the nitrate reduction. Revised: May 5, 2014
It is already established that the distribution of products Accepted: May 7, 2014
depends on not only the catalyst employed but also the Published: May 7, 2014

© 2014 American Chemical Society 8726 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

H* produces N* and oxygenated species, whereas NH* is “surface” path comprises all mass-transfer effects. As already
formed through interaction between N* and H*. stated, because of operational conditions, the concentration of
NH3* could be formed because of gradual addition of both H2 and CO2 in the gaseous phase is considered to be
hydrogen to NO*,16 but a hydrogenation of N* step was also constant; any gaseous product is immediately “swept” from the
proposed.22 system. In the “bulk” phase, several chemical equilibria are
It can be seen that despite the intense effort in studying the assumed to take place as well as mass transfer from the gas and
catalytic reduction of nitrate to nitrogen, the actual mechanism to the “surface”.
is not a consensus and efforts must still be made to develop a On the basis of the proposed redox mechanism, the following
reliable model for this reaction. In addition, it is important to assumptions were employed:14,18
establish a better understanding of the hydroxyl formation and (1) The adsorption of NO3−, H2, and NO2− on metallic sites
the role of the pH control in the reaction selectivity. This is due occurs in equilibrium with the liquid (“bulk”) phase.
to the fact that the hydrogenation leads to a stoichiometric (2) The reduction of nitrate occurs because of the interaction
production of OH−,10,23 causing a significant decrease in the with hydrogen, but each one adsorbed to a different site
conversion rate, once the OH− concentration in the catalyst of the catalyst.
pores inhibits the adsorption of NO2−.24 Therefore, the pH (3) The regeneration of the promoter metal from metal
affects the catalyst activity, increasing the selectivity to NH4+.15 oxide is fast and promoted by hydrogen transfer on active
A decrease in the pore size of the support caused an increase in bimetallic sites.
NH4+ concentration.25 (4) The adsorption and the Hspillover action on bimetallic sites
Addition of formic acid, carbon dioxide, or even hydrogen is very fast.
chloride is commonly applied to control the pH. Organic (5) The elementary hydrogenation steps are irreversible.
buffers showed nitrate removal higher than that of inorganic (6) The desorption of N2 and the desorption of NH3 are
buffers, but the chemical structure of the organic buffer both irreversible.
influenced nitrate reduction.26 In addition, the use of a cation (7) There is only one molecular layer of adsorbed species on
exchange resin improved the buffering properties near the the catalyst.
active site, with consequent improvement in N2 selectivity.27
However, the mentioned hydroxyl formation takes place inside On the basis of the above assumptions and other conclusions
the catalyst pore and the local pH within the porous media can found in the literature,11,12,14,22,30 the proposed reaction model
be significantly different from that of the solution.24 Because comprises the following steps:
any pH control method will be performed in the solution, a k1
(NO3−)s + (#) ↔ (NO3−)#
reliable model must also take into account the mass-transfer k2 (1)
effects. A higher selectivity to N2 was obtained because of
k3
enhanced mass transfer near the catalytic sites when a structure (H 2)s + 2( ∗) ↔ 2(H)*
catalyst was used.28 Isotopic labeling experiments showed the k4 (2)
selectivity for N2 increased with the concentration of adsorbed k5
NO on the catalytic sites.29 (NO3−)# + 2(H)* → (NO2−)* + (H 2O)s + (#) + ( ∗)
A previous work30 focused on establishing a reaction model (3)
without taking into account the mass-transfer effects (gas−
k6
liquid and liquid−catalyst porous media) and some equilibrium (NO2−)* ↔ (NOs−)s + ( ∗)
relations in the aqueous phase (ammonia, water, and the k7 (4)
carbonate−bicarbonate−CO2 system). Therefore, the present k8
work is devoted to coupling a reliable microkinetic mecha- (NO2−)* + (H)* → (NO)* + (OH−)* (5)
nism30 with transport phenomena and physical-chemical
k9
aspects;14 because this effort drastically increases the number (NO)* + (H)* → (N)* + (OH)* (6)
of equations, variables, and parameters, some mathematical
k10
strategies were also developed. 2(N)* → N2 + 2( ∗) (7)
2. METHODOLOGY k11
(N)* + (H)* → (NH)* + ( ∗) (8)
The main objective of the present work is devoted to the
modeling and parameter estimation of the catalytic reduction of k12
(NH)* + (H)* → (NH 2)* + ( ∗) (9)
nitrate over a Pd−Sn/γ-Al2O3 catalyst. The experimental
apparatus as well as all operational conditions employed can k13
be found in a previous work30 and will not be described here. (NH 2)* + (H)* → (NH3)* + ( ∗) (10)
Nevertheless, whenever an experimental condition is found to k14
be relevant for understanding an assumption, it will be clearly (NH3)* → (NH3)s + ( ∗) (11)
stated. Briefly, a Pd−Sn catalyst was admitted to an aqueous
k15
solution of nitrate under a gaseous atmosphere of H2 and CO2 (OH−)* → (OH−)s + (*) (12)
of constant concentration, supplied by a continuous feed of the
gaseous mixture. k16
(OH−)* + (H)* → (H 2O)s + 2( ∗) (13)
2.1. Model Formulation. 2.1.1. Microkinetic Mechanism.
For modeling purposes, the system was split into four “phases” The Arrhenius relation was used to model the dependence of
or subsystems: gas (g), bulk phase (b), surface (s), and catalyst each rate constant kn(s−1) to the temperature T (K):
(# for the bimetallic sites and * for the monometallic ones).
The last one denotes the adsorbed condition, and the “bulk”-to- kn = k 0, ne−Ea,n / RT (14)

8727 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734


Industrial & Engineering Chemistry Research Article

where the pre-exponential factor k0,n (s−1) and the activation Kb = 10−[4.75595 − 2729.33(1/298.15 − 1/ T )] (36)
energy Ea,n (J/mol) are all subject to estimation.
On the basis of the framework described in eqs 1−13, rate The water dissociation constant is given by32,33
equations are given by
K w = exp[148.96502 − 13847.26/T − 23.6521 ln T ]
r1 = k1C(NO3−)Sθ# − k 2θ(NO3−)# (15) (37)
34
The carbonic acid dissociation constants are given by
r2 = k 3C(H2)Sθ∗2 − k4θ(H)
2
* (16)
K1 = 10−[−14.8435 + 34471.0/ T + 0.032786T ] (38)
2
r3 = k5θ (NO3−)# θ(H) * (17)
K 2 = 10−[−6.4980 + 2903.9/ T + 0.02379T ] (39)
r4 = k6θ(NO2−) * − k 7C(NO2−)Sθ (18)
*
leading to Ka = K1K2, which is the total dissociation constant.
r5 = k 8θ(NO2−) *θ(H) * (19) 2.1.3. Mass-Transfer Rates. For the transport phenomena
involved, it was assumed that each subsystem is homogeneous,
r6 = k 9θ(NO) *θ(H) * (20) the system is isothermal, and the active sites are uniformly
distributed over catalyst particles. Because the mass-transfer
2
r7 = k10θ(N) * (21) effects from the “bulk” phase to the “surface” were grouped, the
related rates were therefore written in terms of the
r8 = k11θ(N) *θ(H) * (22) corresponding driving forces and mass-transfer coefficients:35

r9 = k12θ(NH) *θ(H) * r14 = k17(C(NO3−)b − C(NO3−)S) (40)


(23)

r10 = k13θ(NH2) *θ(H) * (24) r15 = k18(C(H2)b − C(H2)S) (41)

r11 = k14θ(NH3) * (25) r16 = k19(C(NO2−)b − C(NO2−)S) (42)

r12 = k15θ(OH−) * (26) r17 = k 20(C(H2O)b − C(H2O)S) (43)


r13 = k16θ(OH−) *θ(H) * (27) r18 = k 21(C(N2)g − C(N2)S) (44)
where θ , θ#, θe, and Ce, are the fraction of void monometallic
* r19 = k 22(C(NH3)b − C(NH3)S)
sites, the fraction of void bimetallic sites, the fraction of sites (45)
occupied by species “e”, and the concentration of species “e”,
respectively. r20 = k 23(C(H2)g − C(H2)b) (46)
2.1.2. Chemical Equilibria. As already stated, the pH of the
liquid phase remains approximately constant, even though OH− r21 = k 24(C(CO2)g − C(CO2)b) (47)
is produced, because of the presence of CO2 and once the
following equilibrium reactions take place: r22 = k 25(C(OH−)b − C(OH−)S) (48)
b b Kb + b − b
(NH3) + (H 2O) ↔ (NH4 ) + (OH ) (28) These equations are able to predict an eventual OH−
Kw accumulation on the surface of the catalyst, as suggested by
(H 2O)b ↔ (H+)b + (OH−)b (29) experimental evidence.24 It must be noticed that the way the
model is formulated, no a priori assumption concerning the
K1
(CO2 )b + (H 2O)b ↔ (HCO3−)b + (H+)b (30) relative magnitude of kinetic and mass-transfer rates is made.
Even though the constants in the previous equations are
K2
(HCO3−)b ↔ (CO32 −)b + (H+)b subject to fitting, some relations can be used to establish initial
(31)
estimates for their values. Roughly, each constant can be
Therefore, the following equilibrium relations can be stated: expressed by
C(NH4+)bC(OH−)b hn a n
Kb = kn =
C(NH3)b (32) δnVn (49)

K w = C(H+)bC(OH−)b where an is the effective interfacial area, Vn the volume of the


(33)
“phase”, and δn the film thickness; hn refers to a diffusion and/
C(HCO3−)bC(H+)b or convective coefficient, depending on the scenario. It must be
K1 = emphasized that no accurate description of these entities is
C(CO2)b (34) needed at this point because they will be statistically fitted.
However, some order of magnitude analysis is useful for
C(CO32−)bC(H+)b establishing good initial guesses for the regression algorithm,
K2 =
C(HCO3−)b (35)
and this was made employing typical values found in the
literature35 and reasonable assumptions made on the basis of
where Kb, Kw, K1, and K2 are the corresponding equilibrium the experimental conditions.30 Therefore, when “n” is an
constants. Ammonia dissociation constant is calculated by31 electrolyte, the model to be fitted was set to be
8728 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

⎡A B ⎤
−1 dC(NH3)S
kn = ⎢ n + n ⎥ = r11 + r19
⎣T T⎦ (50) dt (68)

otherwise
dC(NO3−)b
⎡ An B ⎤−1 = −r14
(69)
k n = ⎢ 1.75 + n + Cn ⎥ dt
⎣T T ⎦ (51)
2.1.4. Material Balance. The mechanisms and rate dC(H2)b
equations described in the previous sections can be used to = −r15 + r20
dt (70)
express the material balance over all species as follows:
dC(NO3−)S dC(NO2−)b
= −r1 + r14 = −r16
dt (52) dt (71)
dθ(NO3−)# r1 − r3
= dC(N2)b
dt αC S (53) = −r18
dt (72)
dθ# −r + r3
= 1
dt αC S (54) dC(NH3)b
= −r19
dt (73)
dC(H2)S
= −r2 + r15
dt (55)
dC(CO2)b
dθ − 2r2 + 2r3 + r4 + 2r7 + r8 + r9 + r10 + r11 + r12 + 2r13 = r21
* = dt (74)
dt (1 − α)CS
(56)
dC(OH−)S
dθ(H)* 2r2 − 2r3 − r5 − r6 − r8 − r9 − r10 − r13 = r12 + r22
dt (75)
=
dt (1 − α)CS (57)
where α is the fraction of bimetallic sites. Considering that all
dθ(NO2−)* −r4 − r5 tin is associated with bimetallic sites (monometallic sites are
=
dt (1 − α)CS (58) palladium sites), it is easily shown that30
mSn
dC(H2O)S MSn
= r3 + r13 + r17 α=
dt (59)

dC(NO2−)S
mSn
MSn
+ ( mPd
MPd

mSn
MSn ) (76)
= r3 + r4 + r16
dt (60) where mi is the mass of i (g) and Mi is the molecular mass of i
dθ(NO) * (g/(g mol)).
r5 − r6
= The total “concentration” of catalytic sites Cs is given by
dt (1 − α)CS (61)
%Pd Ccat
dθ(OH−) * r + r6 − r12 − r13 CS =
= 5 100MPd (77)
dt (1 − α)CS (62)
where %Pd is the amount of palladium on the catalyst, mcat the
dθ(N) * r − 2r7 − r8
= 6 mass of catalyst (g), and Ccat the “concentration” of the catalyst
dt (1 − α)CS (63) in the reaction mixture (g/L).
The remaining degree of freedom in the mathematical model
dC(N2)S
= r7 + r18 is removed by the assumption of the electrical neutrality of the
dt (64)
solution:
dθ(NH) * r8 − r9
= C(Na+)b + C(H+)b + C(NH4+)b = C(NO3−)b + C(NO2−)b
dt (1 − α)CS (65)
+ C(OH−)b + 2C(CO32−)b + C(HCO3−)b (78)
dθ(NH2) * r9 − r10
=
dt (1 − α)CS (66) in which the concentration of Na+ is constant and equal to the
initial nitrate concentration.
dθ(NH3) * r10 − r11
= Therefore, combining the relevant equations, the rate of
dt (1 − α)CS (67) change of H+ concentration in the “bulk” phase, becomes
8729 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

⎡ C(H+)bKb dC(NH3)b K1 dC(CO


2)
b ⎤ Considering that the reaction system was previously exposed
dC(H+)b ⎢⎣ − K w dt
− C(H+)b dt ⎥⎦ to an excess of H2 for the catalyst activation, the initial
= concentration of H2 on the surface of the catalyst was equal to
dt ⎡ C(NH )bKb K w + K1C(CO 4KAC(CO ⎤
the one in the “bulk” phase and all monometallic sites were
b b
2) 2)
⎢⎣1 + ⎥⎦
3
Kw
+ 2 + 3
(C(H+)b) (C(H+)b)
occupied by H2; no other species was adsorbed, and the
⎡ dC(NO −)b dC(NO ⎤ fraction of void bimetallic sites was 100%.
2K a dC(CO2)b −b
2 )
⎢⎣ − (C + b)2 dt + dt
3
+ ⎥⎦ 2.2. Model Integration and Parameter Estimation
dt
+
(H ) Methods. Some attempts were made to solve the system of
⎡ C(NH )bKb K w + K1C(CO )b 4KAC(CO )b ⎤ differential-algebraic equations (DAE) composed by the
⎢⎣1 + K w + (C + b)2 + (C(H+)b)3 ⎥
3 2 2

(H ) ⎦ material balances and the chemical equilibrium equations


along with the restriction of electrical neutrality of the solution.
(79)
Several standard DAE-solving algorithms and some modified
which is a “key” equation for the formulation other rate ones were tested without success, probably because of the
equations: stiffness of the differential equations.30 Therefore, the insertion
dC(OH−)b −K w dC(H+)b of the algebraic equations into the differential, which lead to a
= 2 system of ordinary differential equations (ODEs), was crucial to
dt (C(H+)b) dt (80) the numerical strategy employed. It must be noted that the
integration of such a system of ODEs corresponds to the
dC(NH4+)b Kb dC(NH3)b C(NH3)b dC(OH−)b dynamic simulation of the reaction system and is completely
= −
dt C(OH−)b dt (C(OH−)b)2 dt dependent on the empirical parameters, which is adequate for a
(81) statistical fitting procedure (as far as experimental data were
available).
2.1.5. Initial Conditions. Once the concentrations of CO2 The numerical integration method used is described in detail
and H2 in the gas phase remain constant in each run in a previous work,30 but basically involves a backward
(depending only on the temperature), Henry’s law was used differentiation formulas (BDF) scheme in a vector−matrix
to calculate the initial concentration of these species in the format.
“bulk” phase:36 For the model regression, a maximum likelihood statistical
C(CO2)g criterion was employed, as also described in a previous work.30
C(CO2)0b = The set of estimated parameters ( β )̂ is, therefore, given by the
HCO2 (82)
following optimization:
C(H2)g
exp exp
C(H2)0b = β ̂ = min[C̲ (NO t
− b − C̲ (NO −)b( β̲ )] W[C̲
(NO −)b
− C̲ (NO3−)b( β̲ )]
HH 2 (83) { β} ) 3 3 3

(88)
where HCO2 and HH2 are the solubility constants of CO2 and H2,
exp
respectively, given by36 where C̲ (NO − b is the vector of all experimental points and W is
)3
1 a diagonal matrix of weights. Because small experimental values
= exp[1n( −159.8741 + 5528/T − 0.0011026T tend to be close to the measurement precision, each weight was
HCO2
set as the inverse of the value of the corresponding
+ 21.66941 ln(T )] (84) experimental point.
The final problem involves an optimization of an objective
1 function that demands the integration of all ODEs using the
= exp[1n(− 125.939 + 5528/T + 16.8893 ln T )]
HH 2 decision variables (the parameters to be estimated). Once
(85) again, the present work employed the previously developed
strategy, which consisted of a hybrid scheme, in which a
It must be said that this was performed only to generate a set simplex search preceded a quasi-Newton method with a
of adequate initial conditions; once equilibria 30 and 31 take Broyden−Fletcher−Goldfarb−Shanno approximation of the
place, Henry’s law would not be applicable. Hessian matrix.30 A total of 150 experimental points were
The initial concentrations of NO3−, NH4+, and Na+ were used to fit the 50 empirical parameters.
known, whereas NO2− and N2 were not initially present.
All routines were written with Scilab 5, a free and open
Therefore, the restriction of electrical neutrality of the solution
source software (distributed under CeCILL license, GPL
could be applied at the initial time, leading to
compatible) developed by Scilab Enterprises (http://www.
C(H+)0b + C(NH4+)0b = C(OH−)0b + 2C(CO32−)0b + C(HCO3−)0b scilab.org).
(86)
This was combined with chemical equilibrium equations to 3. RESULTS AND DISCUSSION
give: Once the regression was concluded, fitted parameters for the
(C(H+)0b)3 + C(NH4+)0b(C(H+)0b)2 − [K w + K1C(CO2)0b]C(H+)0b = 0
kinetic and mass-transfer models were available; these can be
found in Tables 1−3.
(87) The predictive model thus obtained can be compared to
which in turn was solved by Cardano’s method.36 experimental data. In Figure 1, a normalized scale for the nitrate
After (H+)b0 was caluculated, the initial concentration of both concentration was used, dividing each experimental value by
OH− and NH3 could be calculated by eqs 32 and 33. the initial concentration of the corresponding run.
8730 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

Table 1. Fitted Kinetic Parameters


k0,n Ea,n (kJ/mol)
k0,1 4.02 Ea,1 8.12
k0,2 8.69 × 10−12 Ea,2 2.69
k0,3 0.14 Ea,3 8.62
k0,4 0.000 057 Ea,4 1.71
k0,5 0.091 Ea,5 24.52
k0,6 0.0024 Ea,6 12.11
k0,7 0.33 Ea,7 1.67 × 10−16
k0,8 0.000 26 Ea,8 10.75
k0,9 0.000 26 Ea,9 10.75
k0,10 1097.02 Ea,10 25.65
k0,11 16.52 Ea,11 49.11
k0,12 0.61 Ea,12 31.60
k0,13 2.62 Ea,13 34.50
k0,14 0.51 Ea,14 34.35 Figure 2. Evolution of NO3−, NH4+, N2, and NO2− concentration in
k0,15 3648.78 Ea,15 62.92 the “bulk” phase (T = 35 °C).
k0,16 0.0080 Ea,16 1.26
It is important to notice that these results are in accordance
Table 2. Fitted Mass-Transfer Parameters for Nonelectrolyte with the literature.8,11,14 The NO2− concentration in the bulk
Species phase was quite low during the reaction time, with an
increasing profile reaching a maximum for a time lower than
An (s K) Bn (s K1/2) 100 min, followed by a decrease associated with a rapid
A18 410 843.46 B18 −22 496.44 adsorption on the surface of the catalysts and subsequent
A20 527 269.88 B20 −28 871.61 conversion to NO. The low concentration values were
A21 606 914.02 B21 −33 232.74 consistent with the fact that NO2− could not be observed
A22 459 904.04 B22 −25 182.81 experimentally.30 Figure 3 depicts the evolution of the nitrite
A23 2.58 × 109 B23 −1.43 × 108
A24 −7.42 × 108 B24 86 222 578.0

Table 3. Fitted Mass-Transfer Parameters for Electrolyte


Species
An (s K1.75) Bn (s K) Cn (s)
A17 4.67 × 1010 B17 −1.09 × 109 C17 1 499 880.6
A19 4.18 × 1010 B19 −9.72 × 108 C19 1 315 111.1
A25 1.18 × 1010 B25 −2.73 × 108 C25 366 220.94

Figure 3. Evolution of nitrite concentration in the “bulk” phase


(thicker lines) and on the “surface” (thinner lines).

concentration gradient between the “bulk” phase (thicker lines)


and the “surface” (thinner lines). The results shows that the
higher the temperatures, the sooner the nitrite reaches a
maximum both in the “bulk” phase and in the “surface”. This is
consistent with the effect of temperature on the conversion of
Figure 1. Fitted model and experimental data (normalized nitrate nitrate as described in the literature.28
concentration). The pH control was performed by a stream of CO2,30 and it
is useful to verify whether it was effective using the mass-
transfer effects evaluated by the present model. For that, in
Figure 4, the pH on both “bulk” phase (thicker lines) and
After the model is fitted, any other state variable could have “surface” (thinner lines) are represented.
its dynamics simulated, which is important for model validation. The prediction of a higher pH inside the porous media in
Figure 2 depicts the evolution of the most important N-based comparison with the one in the “bulk” phase is consistent with
entities present in the “bulk” phase (for simplicity, only the the literature.15 The addition of CO2 to control the pH has
results corresponding to 35 °C are shown). proven to be a better option than the addition of HCl.7,8,37
8731 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

Figure 4. Evolution of the pH in the “bulk” phase (thicker lines) and Figure 6. Evolution of the fraction of void bimetallic sites.
on the “surface” (thinner lines).

However, our results show that although the presence of CO2


improves the pH control, a pH gradient within the pores still
remains. Thus, the present model allowed to us to follow the
intrapore pH gradient, what was not possible for the model
used before.30 This indicates the need to couple mass-transfer
and kinetic steps to get a complete picture of the system. This
model explains why better selectivities were obtained for large
pore supports25 and when an acidic cation exchange resin was
used as support.27
The simulation of the dynamics of occupation of catalyst
active sites is also of great interest. As expected, it was found
that an almost instantaneous complete occupation of bimetallic
sites takes place, as can be seen in Figures 5 and 6. The

Figure 7. Evolution of the fraction of monometallic sites occupied by


NO.

combination with H to form NH species. The fraction of the


monometallic sites occupied by hydrogen is high at initial times
(Figure 8), with a steep decrease due to the hydrogenation of
nitrate and consequent formation of other intermediates. For
longer times, as hydrogen is continuously fed to the system,
and because of desorption of products, there is an increase in
the fraction of the occupied sites by hydrogen. The selectivity
for N2 is influenced by the relative ratio of sites occupied by
NO and hydrogen.12,21,28 Also, a higher occupation of NO
leads to an increase in selectivity for N2.29 In fact, when the
Figure 5. Evolution of the fraction of bimetallic sites occupied by fraction of sites occupied by NO increases, there is an increase
nitrate. in the formation of NH4+ (Figure 9).
The proposed model showed good predictive power (Figure
9) as shown by the comparison of the experimental formation
decrease in the fraction of bimetallic sites occupied by nitrate is of NH4+ at 25 °C and the curve predicted by the model. It must
associated with the reaction progress. The occupation is shorter be stressed that the experimental data for NH4+ concentration
for higher temperatures, and this may be related to the higher was not included in the regression procedure.
conversion to nitrite. As the nitrite leaves the bimetallic sites, A model variance of 3.8 × 10−9 (mol/L)2 was estimated,
these are able to adsorb new nitrate species. denoting low uncertainties of predicted values, taking into
Figure 7 depicts the evolution of the fraction of account that experimental nitrate concentrations were around
monometallic sites occupied by NO. The fraction of NO 0.0007 mol/L, which denotes an improvement from a previous
increases initially and reaches a maximum because of its work.30
8732 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734
Industrial & Engineering Chemistry Research Article

involved 50 parameters; it is not even possible to use the same


initial guess for comparison purposes.
Despite the good adherence to experimental data and low
variance achieved, considerably large variations in the fitted
parameters set tended to have little impact in the objective
function, as in other works devoted to kinetics parameters
regression.38,39 This apparent paradox can be explained by the
peculiar topology of the likelihood function when dealing with
kinetic models, which seems to exhibit a large flat region
around its minimum (which leads to low parameter sensitivity
and large parameter variances and covariances) but at relatively
low values of the likelihood function itself (which leads to low
model fundamental variance). Of course, this jeopardizes
establishing the set of the most relevant parameters to the
complete model and decreases the reliability of each one but
also guarantees the fluctuations in the fitted parameters (which
could be caused by experimental errors, for instance) will have
little effect on the predictive capabilities of the model. Thus, the
Figure 8. Evolution of the fraction of monometallic sites occupied by
H. good predictive power and the complete statistical description
of the model were benefits of the employed methodology.

4. CONCLUSION
The addition of the mass-transfer effects to the microkinetic
model of the catalytic reduction of nitrates improved the
understanding of the catalytic reduction of nitrate. The
proposed phenomenological model comprises a system of
differential equations related to the molar balances of all
reactants and products in the reactor. Coded in an open source
software, the model allowed a detailed analysis of the
composition in each phase of the nitrate hydrogenation system,
showing a good predictive capability.
The main assumptions of the complete kinetic model were
mass transfer between the gas, liquid (“bulk”), and surface
phases. In the “bulk” phase, several chemical equilibria are
assumed to take place as well as mass transfer from the gas and
to the “surface”. In addition, a redox mechanism was employed,
with two different sites being responsible for the reaction. A
Figure 9. Fitted model and experimental data (NH4+ concentration).
complete description of the adsorbed species was obtained.
This model allowed following the gradient of the pH within the
pores, which affects the selectivity for N2, and it was able to
When it is compared to the previous work,30 which was able predict NH4+ profiles.
to simulate the evolution of chemical entities in only two
“phases” (“bulk phase” and catalyst), the introduction of mass-
transfer effects made it possible to simulate relevant state
■ AUTHOR INFORMATION
Corresponding Author
variables in four “phases” (gas, “bulk phase”, surface, and *Rua Passo da Pátria 156, 24210-240, Niterói, Brazil. Tel.: +55-
catalyst). However, the computational time for system 21-2629-5562. E-mail: fpeixoto@vm.uff.br.
simulation increased 60% when the number of variables Notes
increased from 17 to 53. Therefore, when choosing between The authors declare no competing financial interest.


the present model or the previous one,30 there is a compromise
between the level of mathematical detail and the computational ACKNOWLEDGMENTS
time required.
This new model allowed a better prediction of the selectivity CAPES is acknowledged for sponsoring the scholarship for
E.P.A.R.


for the several products, specially NH4+ (Figure 9), which is
highly dependent on local pH. The model developed in this
work confirmed that a pH gradient remains in the reaction
REFERENCES
system even when CO2 is used to control the pH. This was also (1) United Nations Educational, Scientific and Cultural Organization.
an improvement; in the previous work,30 the pH was Water in a Changing World; UNESCO: Paris, 2009
(2) MMA − Ministério do Meio Ambiente, Plano Nacional de
considered to be uniform throughout the system. ́
Recursos Hidricos − Iniciando um processo de debate nacional, SRH −
It must be pointed out that no comparison can be made with ́
Secretaria de Recursos Hidricos, ́ - DF, 2005 (in portuguese).
Brasilia
the same previous work30 in terms of the computational time (3) World Health Organization. Nitrates and Nitrites in Drinking-
devoted exclusively to parameter estimation, which is highly Water; WHO: Geneva, Switzerland, 2004
dependent on the initial guess. The mentioned previous work30 (4) Feitosa, F.; Manoel Filho, J. Hidrogeologia: Conceitos e Aplicaçoẽ s,
involved the estimation of 14 parameters, and the present work 2nd ed.; CPRM: Brazil, 2000 (in portuguese).

8733 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734


Industrial & Engineering Chemistry Research Article

(5) Zublena, J. P.; Cook, M. G.; Clair, M. B. Pollutants in Catalysts for Nitrate Removal in Water. J. Mol. Catal. A: Chem. 2013,
Groundwater: Health effects; North Carolina Cooperative Extension 366, 294−302.
Service, 1993. (28) Soares, O. S. G. P.; Fan, X.; Ó rfão, J. J. M.; Alexei A. Lapkin, A.
(6) Mahler, R. L.; Colter, A.; Hirnyck, R. Nitrate and Groundwater; A.; Pereira, M. F. R. Kinetic Modeling of Nitrate Reduction Catalyzed
University of Idaho Extension, 2007. by Pd−Cu Supported on Carbon Nanotubes. Ind. Eng. Chem. Res.
(7) Prüsse, U.; Hahnlein, M.; Daum, J.; Vorlop, K. D. Improving the 2012, 51, 4854−4860.
Catalytic Nitrate Reduction. Catal. Today 2000, 55, 79−90. (29) Zhang, R.; Shuai, D.; Guy, K.; Shapley, J. R.; Strathmann, T. J.;
(8) Pintar, A.; Setinc, M.; Levec, J. Hardness and Salt Effects on Werth, C. J. Elucidation of Nitrate Reduction Mechanisms on a Pd-In
Catalytic Hydrogenation of Aqueous Nitrate Solutions. J. Catal. 1998, Bimetallic Catalyst using Isotope Labeled Nitrogen Species.
174, 72−87. ChemCatChem 2013, 5, 313−321.
(9) Centi, S.; Perathoner, G. Remediation of Water Contamination (30) Costa, A. O.; Ferreira, L. S.; Passos, F. B.; Maia, M. P.; Peixoto,
using Catalytic Technologies. Appl. Catal., B 2003, 41, 15−29. F. C. Microkinetic Modeling of the Hydrogenation of Nitrate in Water
(10) Prusse, U.; Vorlop, K. Supported Bimetallic Palladium Catalysts on Pd−Sn/Al2O3 Catalyst. Appl. Catal., A 2012, 445, 26−34.
for Water-phase Nitrate Reduction. J. Mol. Catal. A: Chem. 2001, 173, (31) Clegg, S. L.; Whitfield, M. A. Chemical Model of Seawater
313−328. Including Dissolved Ammonia and the Stoichiometric Dissociation
(11) Hörold, S.; Vorlop, K. D.; Tacke, T.; Sell, M. Development of Constant of Ammonia in Estuarine Water and Seawater from −2 to 40
Catalysts for a Selective Nitrate and Nitrite Removal from Drinking °C. Geochim. Cosmochim. Acta 1995, 59, 2403−2421.
Water. Catal. Today 1993, 17, 21−30. (32) Zeebe, R. E.; Wolf-Gladrow, D. CO2 in Seawater: Equilibrium,
(12) Ilinitch, O. M.; Nosova, L. V.; Gorodetskii, V. V.; Ivnov, V. P.; Kinetics, Isotopes. Elsevier Oceanogr. Ser. (Amsterdam) 2001, 65, 1−
Trukhan, S. N.; Gribov, E. N.; Bogdanov, S. V.; Cuperus, F. P. 346.
(33) Handbook of Methods for the Analysis of the Various Parameters of
Catalytic Reduction of Nitrate and Nitrite Ions by Hydrogen:
the Carbon Dioxide System in Sea Water, version 2; ORNL/CDIAC-74;
Investigation of the Reaction Mechanism over Pd and Pd-Cu
Dickson, A., Goyet, C., Eds.; DOE: Washington, DC, 1994.
Catalysts. J. Mol. Catal. A: Chem. 2000, 158, 237−249.
(34) Mehrbach, C.; Culberson, C. H.; Hawley, J. E.; Pytkowicz, R. M.
(13) Sá, J.; Vinek, H. Catalytic hydrogenation of nitrates in water
Measurement of the Apparent Dissociation Constants of Carbonic
over a bimetallic catalyst. Appl. Catal., B 2005, 57, 247−256. Acid in Seawater at Atmospheric Pressure. Limnol. Oceanogr. 1973, 18,
(14) Fan, X.; Franch, C.; Palomares, E.; Lapkin, A. A. Simulation of 897−907.
catalytic reduction of nitrates based on a mechanistic model. Chem. (35) Welty, J. R.; Wicks, C. E.; Wilson, R. E. Fundamentals of
Eng. J. (Amsterdam, Neth.) 2011, 175, 458−467. Momentum, Heat and Mass Transfer, 3rd ed.; John Wiley & Sons: New
(15) Gavagnin, R.; Biasetto, L.; Pinna, F.; Strukul, G. Nitrate York, 1984.
Removal in Drinking Water: The Effect of Tin Oxides in the Catalytic (36) Perry, R. H.; Green, D. W. Perry’s Chemical Engineer’s Handbook,
Hydrogenation of Nitrate by Pd/SnO2 Catalysts. Appl. Catal., B 2002, 8th ed.; McGraw-Hill, New York, 2008.
38, 91−99. (37) Maia, M. P.; Rodrigues, M. A.; Passos, F. B. Nitrate catalytic
(16) Warna, J.; Turunen, I.; Salmi, T.; Maunula, T. Kinetics of reduction in water using niobia supported palladium−copper catalysts.
Nitrate Reduction in Monolith Reactor. Chem. Eng. Sci. 1994, 49, Catal. Today 2007, 123, 171−176.
5763−5773. (38) Salazar, J. B.; Ferreira, L. S.; Peixoto, F. C.; Maia, M. P.; Passos,
(17) Strukul, G.; Pinna, F.; Marella, M.; Meregalli, L.; Tomaselli, M. F. B. Kinetics of Nitrate Hydrogenation in Water on Alumina and
Sol−gel Palladium Catalysts for Nitrate and Nitrite Removal from Niobia Supported Palladium-Copper Catalysts. Int. J. Chem. React. Eng.
Drinking Water. Catal. Today 1996, 27, 209−214. 2012, 10, 1−21.
(18) Epron, F.; Gauthard, F.; Pinéda, C.; Barbier, J. Catalytic (39) Kobolakis, I.; Wojciechowski, B. W. The Catalytic Cracking of a
Reduction of Nitrate and Nitrite on Pt-Cu/Al2O3 Catalysts in Aqueous Fischer-Tropsch Synthesis Product. Can. J. Chem. Eng. 1985, 63, 269−
Solution: Role of the Interaction between Copper and Platinum in the 287.
Reaction. J. Catal. 2001, 198, 309−318.
(19) Mikami, I.; Sakamoto, Y.; Yoshinaga, Y.; Okuhara, T. Kinetic
and adsorption studies on the hydrogenation of nitrate and nitrite in
water using Pd−Cu on active carbon support. Appl. Catal., B 2003, 44,
79−86.
(20) Tanaka, K.; Ikai, M. Adsorbed Atoms and Molecules Destined
for a Reaction. Top. Catal. 2002, 20, 25−33.
(21) Ebbesen, S. D.; Mojet, B. L.; Lefferts, L. In Situ ATR-IR Study
Of Nitrite Hydrogenation Over Pd/Al2O3. J. Catal. 2008, 256, 15−23.
(22) Rahkamaa, K.; Salmi, T.; Keiski, R.; Wärnå, J.; Zhou, Y.
Transient Reduction Kinetics of NO over Pd-Based Metallic
Monoliths. Chem. Eng. Sci. 2001, 56, 1395−1401.
(23) Deganello, F.; Liotta, L. F.; Macaluso, A.; Venezia, A. M.;
Deganello, G. Catalytic Reduction of Nitrate and Nitrite in Water
Solution on Pumice-Supported Pd-Cu Catalysts. Appl. Catal., B 2000,
24, 265−273.
(24) Sakamoto, Y.; Kamiya, Y.; Okurara, T. Selective Hydrogenation
of Nitrate to Nitrite in Water over Pd-Cu Bimetallic Cluster Supported
on Active Carbon. J. Mol. Catal. A: Chem. 2006, 250, 80−86.
(25) Krawczyk, N.; Karski, S.; Witońska, I. The Effect of Support
Porosity on the Selectivity of Pd−In/Support Catalysts in Nitrate
Reduction. Reac. Kinet., Mech. Catal. 2011, 103, 311−323.
(26) Bae, S.; Jung, J.; Lee, W. The Effect of pH and Zwitterionic
Buffers on Catalytic Nitrate Reduction by TiO2-supported Bimetallic
Catalyst. Chem. Eng. J. (Amsterdam, Neth.) 2013, 232, 327−337.
(27) Barbosa, D. P.; Tchiéta, P.; Rangel, M. C.; Epron, F. The Use of
a Cation Exchange Resin for Palladium−Tin and Palladium−Indium

8734 dx.doi.org/10.1021/ie500820a | Ind. Eng. Chem. Res. 2014, 53, 8726−8734

Das könnte Ihnen auch gefallen