PY - PAP - B.Jankovic B.Adnade - 2008 - The Kinetic Study of Temperature-Programmed Reduction of Nickel Oxide in Hy
PY - PAP - B.Jankovic B.Adnade - 2008 - The Kinetic Study of Temperature-Programmed Reduction of Nickel Oxide in Hy
www.elsevier.com/locate/ces
Received 19 January 2007; received in revised form 9 July 2007; accepted 20 September 2007
Available online 1 October 2007
Abstract
The powder samples of nickel oxide were synthesized by sol–gel procedure. The non-isothermal reduction of nickel oxide using hydro-
gen was investigated by temperature-programmed reduction (TPR) at four different heating rates (2.5, 5, 10 and 20 ◦ C min−1 ). Reaction
kinetics parameters (Ea , ln A) were determined using a few methods: Stationary point (SP), Kissinger, Friedman, Flynn–Wall–Ozawa and
Kissinger–Akahira–Sunose methods. The Málek’s kinetic procedure was used for the determination of the most suitable kinetic model and
calculation of a complete set of kinetic parameters. The kinetic model of the investigated reduction process was determined, and this model
corresponds to the empirical two-parameter Šesták–Berggren equation which gives a more quantitative description. The kinetic triplet obtained
for the investigated process is Ea = 90.8 kJ mol−1 ; ln A = 19.50 (A expressed in min−1 ) and f () = 0.63 (1 − )1.39 . The reliability of kinetics
parameters calculated from TPR data is tested by comparing calculated and experimental non-isothermal data. Comparing both experimental
and calculated rate curves at constant heating rate assessed the adequate consistency of the kinetic triplet. In agreement with the obtained
results, the mechanism of nickel oxide reduction by hydrogen was considered.
䉷 2007 Elsevier Ltd. All rights reserved.
is the reaction at the metal–oxide interface between NiO and methods. In addition, the Málek’s (1992) criteria were used to
hydrogen atoms adsorbed on the previously reduced NiO. The confirm that the single-step model can be applied to fit the ex-
above models emphasized the chemical mechanism and kinet- perimental data in the case of NiO reduction. Also, the most
ics of the reduction process, but there was a little uniformity in accurate kinetic triplet (the conversion model, f () and Arrhe-
the physical properties of NiO grains and pellets used. Never- nius parameters) of the investigated process was established,
theless, it was suspected that morphological factors were just as involving the empirical two-parameter Šesták–Berggren (SB)
important as topological properties in determining the course of (Šesták and Berggren, 1971) reaction model.
the reduction. Generally, some authors (Richardson et al., 1994,
2003; Tikekar et al., 2003; Waldbillig et al., 2005) believed
2. Experimental procedure
that NiO reduction is the first-order and the rate has a linear
dependence with the partial pressure of hydrogen. Richardson
2.1. Materials and methods
et al. (1994) conducted a series of experiments using isothermal
H2 consumption and magnetization measurements to determine
The NiO samples were obtained by gel-combustion method
the Ni–O bond rupture (NiO conversion) and the growth of
described elsewhere (Mentus et al., 2005). A green-colored
nucleated Ni atoms, respectively. They found the growth pro-
transparent gel was obtained by drying an aqueous solution of
cess lagged NiO conversion by a time interval that increased
nickel nitrate hexahydrate (Fluka, 99.5%) and citric acid (Fluka,
with decreasing temperature, lower gas flow rates and the pres-
99.5%) dissolved in a mole ratio 1.8:1. This gel was made to
ence of H2 O added to the reducing gas. Rodriguez et al. (2002)
undergo a self-ignition by heating in air up to 300 ◦ C, and an
showed that in experiments with an NiO (1 0 0) crystal and
additional heating of up to 500 ◦ C produced a very fine nickel
NiO powders, oxide reduction is observed at atmospheric pres-
oxide powders. The mean size of nickel oxide particles deter-
sures and elevated temperatures (250–350 ◦ C), but only after
mined from XRD data is dm = 30 nm, and this value overlaps
an induction period. These authors showed that the presence
very well with the data published by Wu et al. (2007).
of O vacancies leads to an increase in the adsorption energy of
H2 and substantially lowers the energy barrier associated with
the cleavage of the H–H bond. Richardson et al. (2003) stud- 2.2. Temperature-programmed measurements
ied the hydrogen reduction of porous bulk NiO particles with
an in situ hot-stage X-ray diffraction (XRD) in the tempera- The experiments were carried out in a TA SDT 2960 device,
ture range 175–300 ◦ C. The results obtained by these authors capable of simultaneous TGA-DTA analysis in the temperature
indicated that reduction in the absence of water added to the range from 25 to 1500 ◦ C. The nickel oxide samples were re-
reducing gas followed several steps: (1) an induction period as- duced directly within the thermobalance, in korund pans, under
sociated with the initial reduction of NiO and the appearance (99.9995 vol%) hydrogen flowing at a rate of 100 mL min−1 ,
of Ni metal clusters; (2) acceleration of the reduction rate as and using various heating rates: 2.5, 5, 10 and 20 ◦ C min−1 ,
the size of the clusters increase and (3) a pseudo-first-order in the temperature range from an ambient one up to 500 ◦ C.
process in which NiO disappeared and Ni appeared in concert The sample mass used for thermogravimetric investigations was
until reduction slowed at a fractional conversion of about 0.8. about 25 ± 0.5 mg. The degree of conversion () of the reduc-
When 2.2 × 10−2 atm of H2 O was added to the reducing gas, tion process is expressed as
induction times increased by approximately a factor of 2 and
m0 − mt
reduction rates decreased, with an apparent activation energy = , (1)
of 126 ± 27 kJ mol−1 . Utigard et al. (2005) investigated the re- m0 − m f
duction kinetics of NiO granules formed by vapor deposition
where m0 , mt and mf refer to the initial, actual and final mass
from a chloride solution, using thermal gravimetry. They re-
of the investigated sample, respectively.
ported that in the temperature range from 400 to 600 ◦ C, the
rate of reduction increased with increasing temperature and in-
creasing hydrogen pressure. Microscopic analysis showed that 3. The calculating methods used to evaluate the kinetics
in this temperature range the reaction followed the shrinking parameters of the reduction process
core model (Utigard et al., 2005). The same authors found that
the activation energy for reduction process have a value of 3.1. Stationary point method
90 kJ mol−1 in the above temperature range.
Bearing in mind a considerable interest relating to the kinet- In the heterogeneous chemical reaction kinetics, function
ics and mechanism of NiO reduction, in this paper, we contin- d/dt = f (T ) (where d/dt is the velocity of process and
ued the investigation to find the best kinetic model describing T is the absolute temperature), which has zero-initial ve-
this process, using the recent theories of heterogeneous kinet- locity, can be observed as a function with local maximum.
ics. In our previous paper (Janković et al., 2007) using the only This maximum appears in the so-called stationary point (SP):
isoconversional Friedman (1963) (FR) method, the invariance S[(d/dt)max ; Tmax ] (Tmax is the temperature at the maximal
of the apparent activation energy of reduction process was es- velocity of process (d/dt)max ), where (d2 /dt 2 ) = 0 (Klarić
tablished in a limited conversion range (0.20 0.60). In et al., 1995). Consequently, SP is defined as a point in which
the present paper, this finding was confirmed by several other the reaction systems, under given conditions, have maximal
B. Janković et al. / Chemical Engineering Science 63 (2008) 567 – 575 569
reaction velocity. In that case, the following equation is valid: factor and n is the reaction order. Showing that the term
n(1 − m )n−1 ≈ 1 (for n = 1), Kissinger (1957) found the
d following expression:
= k(T , pj )f (max ), (2)
dt max
AR Ea
where (d/dt)max is maximal velocity of considered reaction ln 2
= ln − . (7)
or process, k(T , pj ) the temperature and partial pressure de- Tmax Ea RT max
pendent rate constant, f (max ) is a function of the reaction
2 ) versus 1/T
mechanism, max the degree of conversion at T = Tmax and pj From a plot of ln(/Tmax max and fitting to a straight
denotes the partial pressure of the gaseous products. For non- line, the pre-exponential factor A and the apparent activation
isothermal conditions, when the temperature varies with time energy Ea can be calculated from the intercept and slope,
with a constant heating rate (), = dT /dt, Eq. (2) can be respectively.
written in the following form:
3.3. Friedman method
d
= k(T , pj )f (max ). (3)
dT max Friedman (1963) method presents an isoconversional differ-
If Arrhenius dependence of (k(T , pj )) on temperature is ential method, based on the following equation:
supposed, Eq. (3) can be transformed in the following form:
d Ea,i
d Ea ln = ln[Af ()]i − , (8)
= A exp − f (max ), (4) dT i RT i
dT max RT max
where Ea is the apparent activation energy of the overall pro- where the subscript, i, denotes that the corresponding quan-
cess, whereas A and R are the pre-exponential factor and gas tity is evaluated at a specific value of the degree of conver-
constant, respectively. Since the logarithmic form of Eq. (4) is sion, i . The apparent activation energy, Ea,i , at this specific
degree of conversion, is evaluated from the slope of the plot
d Ea ln(d/dT )i versus 1/Ti , known in the literature as the
ln = ln[Af (max )] − (5)
dT max RT max isoconversional line. In this approach, Ea is expressed as a
function of degree of conversion (). Any change in Ea as the
and from the slope of the dependence between ln[(d/dT )max ] reaction proceeds indicates a change in the rate-limiting step.
and 1/Tmax , it is possible to determine the value of the appar- Budrugeac et al. (2001) have presented some problems concern-
ent activation energy Ea of the investigated process. Method ing the evaluation of the activation energy from non-isothermal
of the SP does not directly enable the calculation of the pre- data for reactions with kinetic parameters dependent on the de-
exponential factor (A). The value of the pre-exponential factor gree of conversion. They have found out that if the activation
(A) can be determined only in the case when the exact math- energy depends on the degree of conversion, FR method, which
ematical form of the function f (max ) is known (Klarić et al., uses directly the equation of the reaction rate, can be used as
1995). Furthermore, by this method we obviously can obtain a recommended procedure. However, this method is known as
only one value of activation energy, the one that can be calcu- being noise sensitive.
lated comparing (d/dt)max at different temperatures and which
corresponds to max (the degree of conversion at T = Tmax ).
3.4. Flynn–Wall–Ozawa method
3.2. Kissinger’s method
The Flynn–Wall–Ozawa (FWO) method (Flynn and Wall,
1966; Ozawa, 1965) is an isoconversional integral method,
Kissinger (1957) method has been used in the literature to
which uses Doyle’s approximation for the temperature integral
determine the apparent activation energy of solid state reactions
p(x)(x = Ea /RT ) (Doyle, 1961a,b), and based on the follow-
from plots of the logarithm of the heating rate versus the inverse
ing equation:
of the temperature at the maximum reaction rate in constant
heating rate experiments. AE a Ea
The apparent activation energy can be determined by ln = ln − 5.331 − 1.052 , (9)
Rg() RT
Kissinger method without a precise knowledge of the reaction
mechanism, using the following equation:
where g() is the algebraic expression for integral form of the
AR Ea reaction model. For = const., ln versus 1/T obtained from
ln 2
= ln + ln[n(1 − max )n−1 ] − , (6) thermograms recorded at several heating rates yields a straight
Tmax Ea RT max
line whose slope allows evaluation of the apparent activation en-
where is the heating rate, Tmax and max are the absolute ergy. For x < 20, Doyle’s approximation leads to errors higher
temperature and the degree of conversion at the maximum than 10%. For such cases, Flynn (1983) suggested corrections
mass-loss rate (d/dt)max , respectively. A is the pre-exponential in order to obtain an accurate value for the activation energy.
570 B. Janković et al. / Chemical Engineering Science 63 (2008) 567 – 575
This is an isoconversional integral method (Akahira and The temperature-programmed reduction (TPR) curves of
Sunose, 1971) based on the following equation: nickel oxide in hydrogen atmosphere recorded at different
heating rates (2.5, 5, 10 and 20 ◦ C min−1 ) are shown in Fig. 1.
AR Ea
ln 2
= ln − . (10) The total mass-change of investigated nickel oxide samples
T Ea g() RT at all considered heating rates is about 21.4%.
For = const., ln(/T 2 ) versus 1/T obtained from thermo- The conversion (–T) curves of nickel oxide reduction by
grams recorded at several heating rates yields a straight line hydrogen at different heating rates are shown in Fig. 2.
whose slope allows evaluation of the apparent activation en- Table 1 shows the influence of heating rate of considered
ergy. This method is based on the Coats and Redfern (1964) system on the characteristic temperatures of reduction process:
approximation for the temperature integral (for 20 < x < 50). initial temperature (Ti ), inflection temperature (Tp ), final tem-
perature (Tf ) and temperature differences (T = Tf − Ti )
3.6. Málek’s method (M method) (Málek, 1992) (Fig. 1).
The increase in heating rate leads to the increase of Ti , Tp ,
Once the apparent activation energy has been determined, Tf and T values. Dependence of d/dt versus T at different
it is possible to find the kinetic model which best describes heating rates of the considered system is shown in Fig. 3.
the measured set of TA data. It can be shown that for this With increasing heating rate (), the rate of reduction process
purpose it is useful to define two special functions y() and of nickel oxide by hydrogen increased.
z(), which can easily be obtained by simple transformation of The values of temperatures Tmax (Fig. 3) at four different
experimental data. These functions can be expressed as follows heating rates for reduction of nickel oxide in hydrogen atmo-
(Málek, 1992): sphere are shown in Table 2.
2 ) versus (1/T
Fig. 4 shows the dependence of ln(/Tmax max )
d for the reduction of nickel oxide by hydrogen.
y() = exp(x), (11)
dt The values of kinetics parameters of nickel oxide reduction
using hydrogen which were determined by SP and Kissinger
d T
z() = (x) , (12) methods are shown in Table 2 (Columns 3,4 and 5,6, respec-
dt
tively). The values of Ea obtained by the SP and Kissinger
where x =Ea /RT , (x) is an approximation of the temperature methods (Table 2) are higher than the value of Ea obtained by
integral. It was suggested that (x) may be accurately estimated Richardson et al. (2003) (Ea = 85.0 kJ mol−1 (non-isothermal
by means of the fourth rational expression of Senum and Yang
(1977):
(x 3 + 18x 2 + 88x + 96)
(x) = . (13)
(x 4 + 20x 3 + 120x 2 + 240x + 120)
The y() function is proportional to the f () function. Thus
by plotting the y() dependence, normalized within the (0,1)
interval, the shape of the function f () is obtained. The y()
function is therefore characteristic for a given kinetic model
(Málek, 1992), and it can be used as a diagnostic tool for kinetic
model determination. It has been verified that the shape of the
y() function, as well as the maximum m of the y() function
and ∞p of the function z(), can help to choose the best kinetic
model. Knowing the value of the apparent activation energy and
the kinetic model, the pre-exponential factor can be calculated
using the following equation:
xp
A=− exp(xp ), (14)
Tp f (p )
where f () is the differential form of the kinetic model
[df ()/d], p is the degree of conversion corresponding to
the maximum on differential rate curve and p is the maximum
of differential rate curve.
The comparison between y() and z() functions may be
used as the criterium of a reaction model in the following sense:
if y() and z() functions notably differ mutually in the shape, Fig. 1. Temperature-programmed curves for nickel oxide reduction process
then one may exclude the assumption of a single-step model. in hydrogen atmosphere at different heating rates.
B. Janković et al. / Chemical Engineering Science 63 (2008) 567 – 575 571
Fig. 2. The degree of conversion () as a function of temperature (T) for Fig. 3. Experimental (lines) reaction rate curves (d/dt versus T) for the
the reduction of nickel oxide under linearly rising temperature, in hydrogen reduction of nickel oxide by hydrogen at different heating rates.
atmosphere, at the following heating rates: 2.5, 5, 10 and 20 ◦ C min−1 (from
left to right).
Table 2
The values of Tmax and values of kinetics parameters obtained by Stationary point (SP) and Kissinger methods (K), for reduction process of nickel oxide
using hydrogen
(◦ C min−1 ) Tmax (◦ C) EaSP (kJ mol−1 ) ln[ASP f (max )], ASP (min−1 ) EaK (kJ mol−1 ) ln AK , AK (min−1 )
ones (Koga et al., 1993). Thus, the most reliable kinetic model
can be determined by comparing the y() and z() functions, that the values of m and ∞ p parameters conform to the two-
according to Eqs. (11) and (12). The normalized y() and z() parameter autocatalytic kinetic model of Šesták and Berggren
functions are presented in Fig. 6. (1971).
The shapes of both the y() and (less expressed) the z() plot In some cases, it is convenient to use more flexible empir-
vary somewhat with the variations in heating rate. Namely, as ical kinetic model functions, f (), for the description of the
already noted, the data in Table 3, extracted from Fig. 6, show crystallization processes. The advantage of such an approach
that m (for y() function) and ∞ p (for z() function) depend
is that it is possible to describe experimental TA data even in
on the heating rate. the case when some of the basic assumptions for the validity of
The maxima of y() and z() functions fall into the range the Avrami–Erofe’ev model are no longer held. These models
of 0.27m 0.34 and 0.52 ∞ p 0.54, respectively. These
can also be understood in terms of the accommodation function
variances correspond to the variances in the calculated val- introduced by Šesták (1991), expressing a deviation from the
ues of the apparent activation energies. Consequently, one can ideal case due to a more complex processes. In this respect, it is
conclude that the kinetics of the process under investigation very convenient to use the SB function (Šesták and Berggren,
may not be exactly described by a one single-step reaction 1971):
model, and the calculated values of activation energy are ap- f () = m (1 − )n , (15)
parent values and differ from the actual values. The m val-
ues are lower than ∞ ∞
p values and p exhibits values lower and this is the most popular form of f () for examination of
than 0.632 (a characteristic value for the kinetic model deter- auto-catalyzed chemical reactions. This flexible two-parameter
mination) (Montserrat et al., 1998). From Table 3, it is evident (m, n) SB function, in fact, also includes the Avrami’s (1939,
B. Janković et al. / Chemical Engineering Science 63 (2008) 567 – 575 573
Table 4
Kinetic parameters evaluated for the reduction of nickel oxide by hydrogen
(◦ C min−1 ) s m n ln Aa (min−1 )
Fig. 6. The y() and z() functions calculated by applying Eqs. (11) and (12)
for reduction process of nickel oxide using hydrogen at various heating rates.
Table 3
The values m and ∞ p of the y() and z() functions evaluated for the
reduction of nickel oxide using hydrogen
(◦ C min−1 ) m ∞
p
Málek, J., Criado, J.M., Šesták, J., Milittký, J., 1989. The boundary conditions Šesták, J., 1991. Diagnostic limits of phenomenological kinetic models when
for kinetic models. Thermochimica Acta 153, 429–432. introducing the accommodation function. Journal of Thermal Analysis 36,
Málek, J., Messaddeq, Y., Inoue, S., Mitsuhashi, T., 1995. The crystallization 1997–2007.
kinetics of InF3 -based glass. Journal of Materials Science 30, 3082–3086. Šesták, J., Berggren, G., 1971. Study of the kinetics of the mechanism of
Mentus, S., Majstorović, D., Tomić, B., Dimitrijević, R., 2005. Reduction solid-state reactions at increasing temperatures. Thermochimica Acta 3,
of NiO–WO3 oxide mixtures synthesized by gel-combustion technique: a 1–12.
route to Ni–W alloys. Materials Science Forum 494, 345–350. Šesták, J., Šatava, J., Wendlandt, W.W., 1973. The study of heterogeneous
Montserrat, S., Málek, J., Colomer, P., 1998. Thermal degradation kinetics processes by thermal analysis. Thermochimica Acta 7, 333–336.
of epoxy-anhydride resins: I. Influence of a silica filler. Thermochimica Tikekar, N., Armstrong, T., Virkar A., 2003. The electrochemical society
Acta 313, 83–95. proceedings series. In: Singhal, S.C., Dokiya, M. (Eds.), Proceeding of
Ozawa, T., 1965. A new method of analyzing thermogravimetric data. Bulletin the Solid Oxide Fuel Cells, vol. VIII, p. 670.
of Chemical Society of Japan 38, 1881–1886. Utigard, T.A., Wu, M., Plascencia, G., Marin, T., 2005. Reduction kinetics
Pacchioni, G., 2000. Ab initio theory of point defects in oxide materials: of Goro nickel oxide using hydrogen. Chemical Engineering Science 60,
structure, properties, chemical reactivity. Solid State Sciences 2, 161–179. 2061–2068.
Richardson, J.T., Lei, M., Turk, B., Forster, K., Twigg, M.V., 1994. Reduction Vyazovkin, S., 2001. Modification of the integral isoconversional method to
of model steam reforming catalysts: NiO/-Al2 O3 . Applied Catalysis A account for variation in the activation energy. Journal of Computational
110, 217–237. Chemistry 22, 178–183.
Richardson, J.T., Scates, R., Twigg, M.V., 2003. X-ray diffraction study of Vyazovkin, S.V., Goryachko, V., Lesnikovich, A.I., 1992. An approach to the
nickel oxide reduction by hydrogen. Applied Catalysis A 246, 137–150. solution of the inverse kinetic problem in the case of complex processes.
Rodriguez, J.A., Hanson, J.C., Frenkel, A.I., Kim, J.Y., Pérez, M., 2002. 3. Parallel independent reactions. Thermochimica Acta 197, 41–51.
Experimental and theoretical studies on the reaction of H2 with NiO: role Waldbillig, D., Wood, A., Ivey, D.G., 2005. Thermal analysis of the cyclic
of O vacancies and mechanism for oxide reduction. Journal of American reduction and oxidation behaviour of SOFC anodes. Solid State Ionics
Chemical Society 124, 346–354. 176, 847–859.
Scholz, J.J., Langell, M.A., 1985. Kinetic analysis of surface reduction in Wu, Y., He, Y., Wu, T., Chen, T., Weng, W., Wan, H., 2007. Influence of
transition metaloxide single crystals. Surface Science 164, 543–557. some parameters on the synthesis of nanosized NiO material by modified
Senum, G.I., Yang, R.T., 1977. Rational approximations of the integral of sol–gel method. Materials Letters 61, 3174–3178.
the Arrhenius function. Journal of Thermal Analysis 11, 445–447.