View Article Online / Journal Homepage / Table of Contents for this issue

PAPER www.rsc.org/pccp | Physical Chemistry Chemical Physics

Oxidation investigation of nickel nanoparticles

Pengxiang Song,a Dongsheng Wen,*a Z. X. Guob and Theodosios Korakianitisa
Received 14th January 2008, Accepted 6th May 2008
First published as an Advance Article on the web 27th June 2008
DOI: 10.1039/b800672e

This work reported an experimental investigation of complete oxidation of nickel nanoparticles

using simultaneous thermogravimetry analysis (TGA) and differential scanning calorimetry
(DSC). Nickel nanoparticles and their elemental compositions were characterized by
Brunauer–Emmett–Teller (BET) analysis, transmission electron microscopy (TEM) and energy
dispersive X-ray spectroscopy (EDS). The oxidation experiments were performed under
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E

isoconversion conditions for seven heating rates, varying from 2 to 20 K min1, with
temperatures up to 1000 1C. The experiments revealed unique oxidation behaviour of nickel at
Downloaded by Texas A & M University on 25 March 2013

the nanometre scale, such as early oxidation and melting phenomena, variable activation energies
and different oxidation kinetics between low and high conversion ratios. Unlike its bulk
counterpart where the activation energy is a constant, the activation energy of nickel
nanoparticles depended on the conversion ratio, ranging between 1.4 and 1.8 eV. The oxidation
kinetics of nickel nanoparticles changed from the classical diffusion controlled mechanism to a
pseudo-homogeneous reaction as conversion ratios were over 50%. The oxidation mechanisms of
nickel nanoparticles were further discussed and future studies to enhance understanding were
identified.

1. Introduction essential to understand the oxidation behavior of nickel

Nanometre-sized metal particles, from several to a few hun-
dred nanometres, have attracted intensive attention in recent
years due to their wide application in batteries, hard alloys,
catalysts, and electronics. Nickel particles have been conven-
tionally used in many industrial and consumer products
including magnets, stainless steel, special alloys, coinage,
plating and glasses. Due to the increased specific surface area
and reactivity, nanometre sized nickel particles could signifi-
cantly improve the properties of conventional products and
promote new applications, such as catalysis for hydrogena-
tion1 and for controlled growth of carbon nanotubes,2 appli-
cation of nickel colloids as printing inks in microelectronic
devices to manufacturing conductive multilayer electrical con-
tacts and interconnections,3 development of high-performance
solar absorbing coatings for increased solar energy conver-
sion,4 and being a potential ferromagnetic candidate as sus-
ceptor materials and Curie temperature based control for
processing applications in composite materials.5 Among all
these applications, nickel particles are typically exposed to an
oxidative environment and their performance is greatly af-
fected by the subsequent oxidation process. The formation of
an oxide layer at the nickel surface could degrade the perfor-
mance under catalytic and ferromagnetic conversion condi-
tions. With an increasing use of nickel nanomaterials, this
problem and cost of corrosion are increasingly of concern. It is
a
School of Engineering and Materials Science, Queen Mary,
University of London, London, UK E1 4NS.
E-mail: d.wen@qmul.ac.uk; Tel: +44-20-78823232
b
Department of Chemistry, University College London, London,
UK WC1H 0AJ
nanoparticles.
The growth kinetics of nickel oxide layers at the bulk level
have been widely investigated and the results have been
summarized in several books and review articles.6–9 Only
limited studies, however, have been carried out on the oxida-
tion of particles, and even fewer at the nanometre scale. For
large nickel particles, theories based on diffusion10 and on the
coupled current approach8,11 have been proposed. Within the
limited studies at the nanometre scale, the oxidation behaviour
of nickel nanoparticles has been found to be remarkably
different from that of large particles and bulk materials. These
experimental studies were generally based on the thermogravi-
metric analysis and investigated under isothermal conversion
conditions,4,11,12 where the conversion ratio was limited to low
values i.e. o20% in weight. It is unknown whether the
mechanisms at low oxidation ratios could be extended to high
ones, and the influence of the heating rate on oxidation
kinetics is also unclear. More profoundly at the nanometre
scale, metal particles are typically covered by passivated films
to protect them from further oxidation, the composition and
thickness of these layers play significant roles in the oxidation
process. The characterization and elemental analysis of these
oxide layers, however, are apparently neglected in most of the
open studies, which might be one of the primary reasons for
the different results reported.13 On the mechanistic under-
standing of oxidation at the nanometre scale, large variation
has been found in the values of activation energies if based
on the classical Arrhenius type mechanism.14 There are still
no generally accepted theories on the oxidation of metallic
nanoparticles.
In order to advance the understanding of oxidation
behavior at the nanometre scale, this study will focus on an

This journal is 
c the Owner Societies 2008 Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 | 5057

experimental investigation of complete oxidation of nickel
nanoparticles based on the isoconversion method. The sample
size profile and elemental composition will be characterized
through transmission electron microscopy (TEM) and energy
dispersive X-ray spectroscopy (EDS), and the experiments will
be performed by simultaneous thermogravimetric analysis
(TGA) and differential scanning calorimetry (DSC) technique,
as detailed below.
2. Experiments
2.1 Sample characterization
The nickel particles, synthesized from the electric explosion of
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E
wire (EEW) method, were purchased from the Sigma-Aldrich
Company. The particles before and after oxidation were
characterized under a TEM (JEOL JEM 2010) to determine
Downloaded by Texas A & M University on 25 March 2013
their size and shape profile. Sample TEM pictures are shown
in Fig. 1. The as-received nickel particles are approximately
spherical with a wide size distribution between 10 and 100 nm.
The averaging particle size estimated from TEM images is as
B30 nm, which is similar to the nominal size from the
provider. A similar average size value is also obtained from
the Brunauer–Emmett–Teller (BET) multipoint method. The
specific surface area of original nickel samples is measured as
4.405 m2 g1, which is equivalent to an average particle size of
Fig. 1 Sample TEM pictures of nickel nanoparticles, (a) before
oxidation and (b) after oxidation.
5058 | Phys. Chem. Chem. Phys., 2008, 10, 5057–5065
View Article Online
B28 nm. For oxidized particles, the shape becomes irregular,
Fig. 1b. The particles are in the form of aggregation or even
sintering, an implication of melting occurred during the
oxidation experiments, which will be discussed further by the
measured DSC curves in section 3.
In order to quantitatively investigate the oxidation beha-
viour, the elemental composition of as-received nickel nano-
particles and particles after oxidation experiments were
analyzed by an energy dispersive X-ray spectroscopy (EDS)
from Oxford Instruments, equipped with INCA Energy 300
system. Unlike nickel particles investigated by other research-
ers, i.e. Karmhag et al.,9,15 which contained some metal
impurities such as Fe, Co, Cr, Cu, the EDS analysis confirmed
the high purity of nickel particles, B99.9% of the metallic
components are nickel, which is consistent with the data from
the manufacturer. The EDS analysis also identified the pre-
sence of oxygen element in nickel particles, 0.7 to 4.9% by
weight at different sampled areas prior to the experiments,
Table 1. This implies that different levels of initial oxidation of
nickel particles occurred during the production or storage
period, which might be caused by the non-uniform distribu-
tion of defects of nickel crystal structures, i.e. areas with large
defects could result in high oxidation levels and vice versa.
Statistical oxygen weight concentration of these seven areas
was 1.64% by weight, which equals a nickel oxide concentra-
tion of 7.65% by weight. This represented a mean oxidation
layer thickness of 1.0 nm for an average nickel particle size of
30 nm. This calculated average oxide thickness is smaller than
some reported values, i.e. 2B3 nm, in some early studies.9,16
However the thickness of oxide layers is affected by different
synthesis methods and the duration of storage period, as well
as different calculation methods. If taking into consideration
the large non-uniformity of the oxygen content, the oxide
thickness could vary between 0.5 and 3.2 nm, which falls
within the general range of prior studies. During the experi-
ments, attempts have been made to observe initial oxide layers
by running TEM at the high resolution mode, but with limited
success, possibly due to the ultra thin oxide layers as indicated
by the calculations. Though limited by the accuracy of the
equipment, this simple EDS analysis clearly reveals that the
initial oxidation layer is highly non-homogenous. One should
therefore be cautious to model oxidation kinetics without
considering this initial oxidization or simply assuming a uni-
form oxide layer thickness.
2.2 Oxidation experiments
Simultaneous Thermal Analyser (STA1500) from Rheno-
metric Scientific was selected to carry on the oxidation experi-
ments. Both TGA and DSC analysis of STA1500 were carried
out at same time, assuring identical conditions. The tempera-
ture uncertainty of the measurement is small as STA1500 has a
highly accurate and internationally accepted temperature
Table 1 Oxygen concentration of as-received nickel nanopowders
S1 S2 S3 S4 S5 S6 S7 Average
Oxygen/% 1.29 0.94 2.23 1.44 4.9 1.6 0.72 1.64
Initial 0.607 0.439 1.041 0.672 2.287 0.752 0.336 0.766
oxidized/mg
This journal is 
c the Owner Societies 2008
 View Article Online

calibration. The weight of nickel nanoparticles was controlled

to small amounts, B10 mg, to minimize the temperature
gradient within the sample. The samples were loosely depos-
ited on an alumina cylindrical crucible with a diameter of 5.77
mm. All oxidation experiments in TGA/DSC were performed
under atmospheric pressure conditions using air as the oxi-
dant. The gas flow rates were set as 20 ml min1. Caution was
taken to avoid the air floating and convection effect that could
induce data fluctuation. The experiments were started at room
temperature and performed under isoconversion conditions,
with heating rates varying from 1 to 20 K min1. The experi-
ments were stopped after a few hours, depending on the
heating rate, when the sample weight did not change with
time. Complete oxidation was achieved in the experiments,
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E

which is confirmed by EDS and XRD studies. Simultaneous

TGA and DSC data were recorded with good repeatability for Fig. 3 Example of TGA (dashed line) and DTG (circle marks) data at
further processing. a heating rate of 12 K min1.
Downloaded by Texas A & M University on 25 March 2013

of the specimen. Notably negative DSC values are observed at

3. Experimental results
3.1 Example of TGA/DSC curves at a fixed heating rate
Fig. 2 shows a typical example of measured TGA and DSC
data at a fixed heating rate of 12 K min1. The weight of nickel
nanopowders stays constant, or shows a very slight decrease
(r0.37% by weight) for some samples at temperatures below
200 1C. This is probably caused by the absorbed residues such
as water and/or carbon dioxide as the samples were exposed to
the atmosphere.13 The residue weights of all samples are
deducted from the original weight for data analysis. The
nanoparticles begin to show weight gain at B250 1C in the
thermogravimetric curve where a slight increase in the DSC
curve is observed, an implication of oxidation occurs at this
temperature. Comparing to the typical initial oxidation tem-
perature of bulk nickel materials, B600 1C, the oxidation of
nanoparticles occurs at much lower temperature. The fastest
reaction occurs at 398 1C where the DSC peak is detected due
to the exothermic oxidation reaction. A pronounced change in
the oxidation kinetics, as observed from the slope decrease of
the TGA curve, occurs at B450 1C. The decreased oxidation
rate extends towards higher temperatures before approaching
asymptotically a constant value, an indication of full oxidation
temperatures over B600 1C. As the DSC curve can only
become negative when the endothermic melting rate exceeds
the exothermic oxidation reaction, an early partial melting of
nickel particles must have been experienced, far below its bulk
value, B1450 1C. The early melting phenomenon is also
supported by TEM pictures of oxidized samples where large
aggregates with irregular shapes are formed, Fig. 1b. Similar
characteristic points are also observed from the differential
thermogravimetric curve (DTG), as shown in Fig. 3.
The final weight increase of this sample material is 25.4%,
which is slightly lower than the ideal weight increase of 27.3%
for a complete pure-nickel oxidation. As observed from the
EDS analysis, typical nanoparticle samples contain some
initial oxides, presumably a thin layer of amorphous oxide,
before experiments. By assuming a completed oxidation at the
end of the experiment, the initial oxygen content can be re-
calculated as 1.5%, which is consistent with the EDS analysis.
3.2 The effects of heating rates on the DSC and TGA curve
Fig. 4 illustrates different thermogravimetric curves under
different heating rates. The TGA curves shift towards higher
temperatures as heating rates increase but all have similar
patterns as shown in Fig. 2. Both the initial oxidation

Fig. 2 Example of DSC (solid line) and TGA (dashed line) data at a
heating rate of 12 K min1. Fig. 4 TGA curves at different heating rates.

This journal is 
c the Owner Societies 2008 Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 | 5059
 View Article Online
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E

Fig. 5 DSC curves at different heating rates.

Fig. 6 DTG curves at different heating rates.
Downloaded by Texas A & M University on 25 March 2013

temperature and the characteristic temperature of oxidation rate constant k(T) is typically expressed by the Arrhenius
slope change increase with increasing heating rates. There are, relation:
however, some variations, from 24.7 to 25.8%, on the final
k(T) = k0 exp[Ea/kBT] (2)
weight increase ratio, which correspond to initial oxygen
concentrations between 1.2 and 2%. where k0, the pre-exponential factor, is assumed to be inde-
Similar DSC patterns are also observed under different pendent of temperature, Ea is the activation energy, T is the
heating rates, Fig. 5. The positive sign of the DSC curve absolute temperature, and kB is the Boltzmann constant. A
implies a strong exothermic reaction due to oxidation. Both combination of eqns (1) and (2) gives
peak heat fluxes and their correspondent temperatures in-
da
crease with increasing heating rates, Table 2, similar to the ¼ k0 f ðaÞ expEa =kB T ð3Þ
oxidation experiments reported by Trunov et al.17,38 At high dt
temperatures, most of the baselines of DSC curves overlap For an isoconversion process where the heating rate, b = dT/
within the instrumental uncertainty for different heating rates. dt, is a constant, the rate of conversion can be written as
It is worthy of note that there is a base heat flux B3 mW
initially for the highest heating rate 20 K min1, which is da k0
¼ f ðaÞ expEa =kB T ð4Þ
considered for further data processing. The differential TG dT b
curves (DTG) have similar patterns with DSC curves, which
Integration of eqn (4) from an initial temperature, T0, corre-
again reflect the oxidation rate, Fig. 6.
sponding to a degree of conversion, a0, to a temperature, Tp,
where a = ap, gives
4. Oxidation kinetics of nickel particles Z ap Z
da k0 Tp
¼ expEz =kB T dT ð5Þ
4.1 The kinetic theory a0 f ðaÞ b T0

As little is known of the mechanisms of oxidation at the If T0 is low, it may be reasonably assumed that a0 = 0 and
nanometre scale, conventional methods will be initially used considering that there is no reaction between T = 0 and
for data processing. In general, kinetic studies assume that the T = T0:
rate of conversion, da/dt, is a linear function of a temperature- Z ap Z
dependent rate constant, k(T), and a temperature-independent da k0 T p
¼ expEa =kB T dT
function of the conversion ratio, f(a), as below.4 0 f ðaÞ b 0

da Z 1
¼ kðTÞf ðaÞ ð1Þ k0 E a expðyÞ
dt ¼ dy ð6Þ
bkB yp y2

where a is the oxidation ratio, defined as the ratio of oxidized

fraction, x, to the maximum possible conversion fraction, k0 E a
¼ pðyp Þ ¼ gðaÞ
xmax, a = x(t)/xmax, where xmax = x(t - N). The reaction bkB

Table 2 DSC peak heat flux and correspondent temperature

Heating rate/K min1 2 5 8 10 12 15 20

Peak heat flux/mW 1.95 11.23 20.76 25.90 32.86 43.48 57.43
Correspondent peak temperature/K 635.33 651.61 660.55 662.99 671.31 674.54 676.74

5060 | Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 This journal is 
c the Owner Societies 2008
 View Article Online

where y = Ea/kBT, yp = Ea/kBTp and , which is the The Starink method. Ref. 23 adopts a more general expres-
integral function of conversion ratio that has different formats sion of the conversion function, f(a), as
depending on different approximations.18–23 As eqn (6) is  
1 q
not analytically soluble, some approximations of p(y) f ðaÞ ¼ ð1  aÞp ln ð10Þ
1a
have to be used. Some established approaches based on both
integral and differential methods of approximation are where p and q are constants that account for different reaction
adopted in this work to calculate kinetics constants, as kinetics, i.e. q = 0 for any homogenous pth order reaction
described below. kinetics. Eqn (11) is derived as a more general form for
calculating the activation energy.
4.2 Determination of the activation energy
b AEa 1
The calculation of activation energy varies according to the ln ¼ þ C3 ð11Þ
Tp1:8 kB T p
characteristics of different reactions. Different methods have
been proposed to calculate kinetic parameters depending on where C3 is the integral constant, A = 1.007  1.2  105Ea
both experimental conditions and mathematical treatment with Ea in units of kJ mol1. The activation energy can be
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E

of data. obtained through the slope of ln b/Tp1.8 versus 1/Tp curves.

The ASTM method. A reference method for activation

Downloaded by Texas A & M University on 25 March 2013

4.2.1 The macroscopic models

energy calculation from isoconversion data is provided by
The Kissinger method. The Kissinger integral method24 is
ASTM standard E698-79,25 which is used in this work for
based on the assumption of y c 1 involving no assumption of
comparison. The method is based on the Ozawa method by
the reaction mechanisms. This can be reasonably justified for
assuming Arrhenius-type homogeneous reaction kinetics,
most of the solid state reactions as the typical activation
energy is in the range of 1B2 eV, which is much larger than da/dt = k(T)(1  a)n. The activation energy is obtained
the thermal energy of the carriers. The integral function can through the slope of log b versus 1/T curve.
R1
then be simplified as pðyÞ ¼ y expðyÞ
y2
dy. By inserting this pure The Friedman method. While all above approaches are based
mathematical approximation into eqn (6), the Kissinger on the integration method, the differential method has also
method can be expressed as: been widely used. The Friedman method26 is based on the
Z ap intercomparison of the rates of conversion, da/dT, for a given
da k0 Ea 1
¼ ln þ ln 2  yp ð7Þ degree of conversion, a, at different heating rates. By differ-
0 f ðaÞ kB byp
entiating eqn (4), the equation below is obtained:
 
At a constant fraction transformed, a, this leads to da Ea 1
ln b ¼ þ ln½k0 f ðaÞ ð12Þ
dT kB T
b Ea 1
ln ¼ þ C1 ð8Þ For the isoconversion method where a = constant, a straight
Tp2 kB T p
line should be found between and 1/T for different heating
rates. The activation energy and the product, k0f(a), can be
where C1 is the integral constant
obtained from the slope and the intercept of the straight line.
Z ap The Friedman method does not involve any approximations
k0 kB da
C1 ¼ ln  ln and is mathematically exact. However it requires both
Ea 0 f ðaÞ
temperatures and conversion rates, da/dT, at different
The activation energy can be calculated by plotting a curve of convection ratios.
ln Tp2/b vs. 1/Tp, which should result in a straight line with a
4.2.2 Determination of activation energies. The basic para-
slope equal to Ea/kB.
meters for the calculation of activation energies, conversion
ratios at given temperatures, are determined from TGA data
The Flynn–Wall–Ozawa method. The Ozawa method21 is
by assuming a stoichiometric oxidation reaction, i.e., the
based on the Doyle approximation,18 which assumes the
weight gain is assumed to be equal to the consumed oxygen.
integral function as log p(y) = 2.315 + 0.4567y. The result
Although some stoichiometric oxide such as Ni2O3, Ni2O2 and
of the integration form of eqn (6) is:
NiOx were reported during the oxidation of some nano-nickel
0:457Ea 1 structures,27–30 a homogeneous transition from Ni to NiO is
log b ¼  þ C2 ð9Þ assumed in this study, which has also been confirmed by a
kB Tp
separated XRD experiment. As discussed in section 2.1, there
where C2 is the integral constant, are initial oxide layers for all specimens before the experi-
ments. This initial oxidation has to be taken into consideration
Z ap
k0 kB da in the data processing. Attempts have been made to calculate
C2 ¼ log  ln  2:315
Ea 0 f ðaÞ the activation energy without considering this oxide layer,
which results in large reduced values of activation energies.
The activation energy can be derived by plotting log b vs. 1/T The calculated values could differ in B30% or more if talking
curves at different conversion ratios, without requiring knowl- all the sample weight as pure nickels. This could help explain
edge of the reaction order. some data scattering reported in the literature on the

This journal is 
c the Owner Societies 2008 Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 | 5061

activation energy of nickel particles.9,14,15 As shown in Fig. 4
and also confirmed by XRD studies, complete oxidation can
be assumed at the maximum weight. The initial oxide weight,
Wiox, is calculated by solving a mass balance equation
Wiox + (Wi  Wiox)(NNiO/NNi) = Wf (13)
where Wi and Wf are the initial and final weight of the
specimen, NNiO and NNi are the molecular weight of NiO
and Ni, respectively. The initial oxide weight is calculated
from the following equation:
Wiox = (WiNNiO/NNi  Wf)/(NNiO/NNi  1) (14)
and the initial nickel weight, WiNi, is
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E
WiNi = Wi  Wiox (15)
The conversion ratio at a particular temperature T, a(T), is
Downloaded by Texas A & M University on 25 March 2013
then determined from:
ðWðTÞ  Wiox Þ  WiNi
aðTÞ ¼ ð16Þ
ðWf  Wiox Þ  WiNi
Both differential and integral methods are used to calculate
activation energies at different conversion ratios and the
results are shown in Fig. 7. It clearly shows that all these
methods produce similar profiles of the activation energy.
Unlike its counterpart at the bulk level where the activation
energy is a fixed value, the activation energy of nickel nano-
particles exhibits a strong function of the conversion ratio. It
increases first with increasing conversion ratios, reaching a
maximum value at a conversion ratio of B50% and decreas-
ing there afterwards. The calculation from all integral methods
is pretty similar: all activation energy data fall within a
boundary line of 10% around the mean values. There is an
early peaking of activation energy and slightly larger scatter-
ing of the activation energy based on the differential method.
This is due to the approximation of differential values, da/dT,
whose value is difficult to obtain accurately because of its high
sensitivity to the baseline stability.23
Fig. 7 also shows that the activation energy of nickel
nanoparticles generally falls between 1.4 and 1.8 eV. The value
is smaller than its bulk value, which is in the range of 2.0B2.6
eV,31 and similar to those derived from the isothermal method
Fig. 7 Activation energies as a function of conversion ratios.
5062 | Phys. Chem. Chem. Phys., 2008, 10, 5057–5065
View Article Online
for nickel particles at micro/nanometre scale, i.e., 1.55 eV for
79 nm nickel particles,5 1.5 eV for 5 mm particles,14 and
1.34 eV for 14 nm particles.9 All these isothermal studies,
however, derived only a single value of activation energy for
nanoparticles. Some early studies (e.g., by Sales and Maple),32
however did observe a departure from the fixed activation
energy value at temperatures over the Curie point (yc =
631 K) for microsized nickel particles around 60B125 mm.
In theory, a fixed activation energy should be held for homo-
geneous reactions, where all freely moving reactant molecules
are identical and unaffected by the product formation. How-
ever for solid state reactions, especially oxidation at the
nanoparticle level, the reacting entities in a solid are not
isolated as the reactions proceeds in the rigid structure, they
interact with neighbours to which they are bonded. There is a
possibility that during the reaction, the reactants may undergo
progressive modification of their reactivity by factors such as
crystal defect formation, particle disintegration and develop-
ment of intra-crystalline strain. Furthermore, the initial reac-
tivity of individual particles that constitute the original
reactant may be appreciably different, due to variations of
particle sizes, boundary faces of different indexes exposed,
crystal imperfections and damages. As a consequence, the
average reactivity of the assemblage of reactant particles
may not remain constant as the reaction progresses. It appears
plausible that variable activation energies should be used in
order to accurately model the oxidation kinetics of nanopar-
ticles. More discussion on the implication of variable activa-
tion energy can be found from a review by Galwey.33
4.3 Determination and discussion of reaction kinetics
The conventional master plot method is used to determine the
reaction kinetics. By using a reference at a half conversion
ratio, a = 0.5, eqn (6) is converted into:
k0 Ea
að0:5Þ ¼ pðy0:5 Þ ð17Þ
bkB
where y0.5 = Ea/kBT0.5, dividing eqn (6) by eqn (17) gives:
gðaÞ pðyÞ
¼ ð18Þ
gða0:5 Þ pðy0 Þ
Theoretical master plots can be obtained by plotting, g(a)/
g(a0.5), for various g(a) functions as listed in Table 3. Once the
activation energy has been determined, an appropriate kinetic
model can be found by comparing the experimental master
plot with the theoretical master plots for various reaction
models. The plotting of experimental curves requires an
approximation of p(y) as it is analytically insoluble and
various approximation methods have been proposed as de-
scribed in section 4.2.1. As the variation of these approxima-
tions is generally small, Kissinger’s approximation is used
here, p(y) E exp(y)/y2.
The comparison of experiments with a number of reaction
models is shown in Fig. 8. The variable activation energies
obtained from the model-free Kissinger method are used in the
calculation. The figures show that at low conversion ratios,
Fig. 8a, the difference between different reaction models is
small. A number of models including the diffusion-based
Jander equation, the Guintling equation and the 2-D diffusion
This journal is 
c the Owner Societies 2008
 View Article Online

Table 3 Reaction models for comparison in Fig. 8

Symbol Model f(a) g(a)

n=3 3rd homogeneous reaction (1  a) 3 1
2[(1  a)2  1]
n=4 4th homogeneous reaction (1  a) 4 1
3[(1  a)3  1]
n=5 5th homogeneous reaction (1  a) 5 1
4[(1  a)4  1]
n=6 6th homogeneous reaction (1  a) 6 1
5[(1  a)5  1]
1 2
1-D One-dimension symmetry 2a a
2-D Two-dimension symmetry [ln(1  a)]1 (1  a)[ln(1  a)] + a
Jander Three-dimension symmetry with constant diameter 3 2/3
2(1  a) [1  (1  a)1/3]1 [1  (1  a)1/3]2
Guintling Three-dimension symmetry with changing diameter 3
2[(1  a)
1/3
 1]1 (1  2a/3)  (1  a)2/3
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E

model, and a 4th order homogeneous model can simulate the The diffusion-based mechanism with the consideration of a
oxidation mechanism reasonably well, with the Jander equa- built-in electric field has been accounted for the observed
tion having the best fit. However as the conversion ratio oxidation phenomena of nickel at various particle sizes, as
Downloaded by Texas A & M University on 25 March 2013

increases to over 50%, Fig. 8b, the diffusion-based mechan-
isms become unreliable and there is a large difference between
the experimental results and predictions. The difference in-
creases with increasing conversion ratios. However the experi-
mental data can be predicted approximately by a pseudo-
homogeneous model at an order between 4 and 6.
Fig. 8 Kinetic model fitting for nickel nanoparticles, (a) conversion
ratio o0.5 and (b) conversion ratio 40.5.
outlined in the Introduction. Such conclusion was arrived at
mostly through isothermal and electrochemical experi-
ments,32,34,35 where the conversion ratio is typically small,
i.e. less than 20% in weight. The diffusion mechanism can be
expected within this low conversion ratio regime. However as
shown in the TGA/DSC curves in Fig. 4 and 5, there is a
salient slope change at the conversion ratio B0.5. This has
also been reflected in the activation energy curve: the calcu-
lated activation energies from both differential and integral
methods decrease at conversion ratios larger than 0.5. It is
therefore expected that there is a change in the oxidation
mechanism, which supports the kinetic model fitting. Such a
mechanism change at different oxidation stages has been
mentioned also in other macroscopic studies. For instance,
Peraldi et al.36 observed a crystallographic orientation depen-
dent growth of bulk nickel oxidation and found that at high
temperatures (41000 1C) the diffusion could not be the only
mechanism; Sales and Maple32 observed different activation
energies for nickel oxidation in the vicinity of its Curie point.
These possible influences might be linked to a mechanism
change from diffusion-controlled to pseudo-homogeneous re-
action models. In this study, the experiment shows that the
oxidation rate was slow at a conversion ratios larger than 0.5,
whereas the activation energy was small in this regime. This
might be caused by the early sintering or melting of nickel
nanoparticles that hinders the transportation of reactive
species. Further investigation of this point is clearly needed.
Although the diffusion-based Jander equation has been
found to model the oxidation very well at low conversion
ratios, it still remains unknown if this well-fitting is a reactive
character reflecting size effect or discrepancies between the
theory and experiments, because of the large agglomeration
after oxidation, Fig. 1b, and a lack of detailed reaction
information. The Jander equation assumes a constant particle
size during the reaction, and it is expected that it is valid only
at low conversion ratios even at the macroscopic level. As the
oxidation progresses to high conversion ratios, there will be a
large change in particle size due to the bonding of oxygen
atoms. The Guintling equation, which modifies the Jander
equation by considering the particle size variation during the
oxidation process, should be more suitable in this aspect.37
However, as shown in Fig. 8, this modification could not

This journal is 
c the Owner Societies 2008 Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 | 5063

account for the discrepancies between experiments and pre-
dictions. Attempts have been made to observe the local
phenomenon such as the increase of particle size and volume
expansion ratio of Ni to NiO after the oxidation experiments.
These phenomena are difficult to observe in situ due to the
large agglomeration and early melting. The good fitting by the
Jander equation at low conversion ratios is likely to be
associated with the average effect of early melting and agglom-
eration. It should be noted that all data analysis are based on
conventional oxidation kinetics, and the models developed at
bulk level may not be applicable to the oxidation of nanopar-
ticles. Further investigation on these aspects is ongoing.
5. Conclusions
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E
A complete oxidation experiment of nickel nanoparticles
based on the isoconversion method was investigated in this
Downloaded by Texas A & M University on 25 March 2013
work. Specific conclusions are drawn as follows:
 The surface of nickel nanoparticles was covered by a non-
uniform nickel oxide layer, whose average thickness was
estimated at B1 nm. Neglecting this oxide layer would result
in large difference in the calculated activation energy values.
 The initial oxidation and melting temperature of nickel
nanoparticles were observed to occur around 300 and 500 1C,
respectively, which are much smaller than the bulk values.
 Six different methods were investigated to calculate the
activation energy, and similar activation energy values were
obtained. The activation energy of nickel nanoparticles was
smaller than that of the bulk nickel, implying a decreased
energy barrier for oxidation.
 The activation energy of nickel nanoparticles was depen-
dent on the conversion ratio. It increased with conversion
ratios at a o 0.5, and then decreased afterwards, ranging
between 1.4 and 1.8 eV.
 The oxidation kinetics were found to be diffusion
dominated and can be well-modelled by the classical Jander
equation at a o 0.5. At higher conversion ratios, a
pseudo-homogeneous model with reaction orders between
4 and 6 could describe the experiment reasonably well.
Further studies on the variable activation energy and oxida-
tion mechanisms are being carried out.
Glossary
Latin
Ea Activation energy/eV
kB Boltzmann constant, 1.38  1023/J K1
k0 Pre-exponential factor in eqn (4)
k(T) Reaction rate constant in eqn (4)
T Temperature/1C
Wiox Initial oxide weight/kg
Wi Initial weight of the specimen/kg
Wf Final weight of the specimen/kg
NNiO Molecular weight of NiO/g mol1
NNi Molecular weight of Ni/g mol1
WiNi Initial nickel weight/kg
x Ratio of oxidized fraction/%
xmax Maximum conversion fraction/%
5064 | Phys. Chem. Chem. Phys., 2008, 10, 5057–5065
View Article Online
(continued )
Greek
a Fraction of oxidation/%
a0 Initial fraction of oxidation/%
b Heating rate/K min1
yc Curie point/1C
Acronym
BET Brunauer–Emmett–Teller theory
DSC Differential scanning calorimetry
DTG Differential thermogravimetric
EDS Energy-dispersive X-ray spectroscopy
PDF Powder diffraction file
TEM Transmission electron microscopy
TGA Thermogravimetric analysis
XRD X-Ray diffraction
Acknowledgements
The authors would like to extend their thanks to EPSRC for
financial support under Grant EP/F027281/1.
References
1. G. Pina, C. Louis and M. Keane, Nickel particle size effects in
catalytic hydrogenation and hydrodechlorination: phenolic trans-
formations over nickel/silica, Phys. Chem. Chem. Phys., 2003, 5,
1924–1931.
2. C. Ducati, I. Alexandrou, M. Chhowalla, J. Robertson and
G. Amaratunge, The role of the catalytic particle in the growth
of carbon nanotubes by plasma enhanced chemical vapor deposi-
tion, J. Appl. Phys., 2004, 95, 6387–6391.
3. W. Tseng and C. Chen, Dispersion and rheology of nickel
nanoparticle inks, J. Mater. Sci., 2006, 41, 1213–1219.
4. R. Karmhag, T. Tesfamichael, E. Wäckelgård, G. Niklasson and
M. Nygren, Oxidation kinetics of nickel particles: comparison
between free particles and particles in an oxide matrix, Sol.
Energy, 2000, 68, 329–333.
5. W. Suwanwatana, S. Yarlagadda and J. W. Gillespie, An
investigation of oxidation effects on hysteresis heating of nickel
particles, J. Mater. Sci., 2004, 38, 565–573.
6. A. T. Fromhold, Jr, Book: Theory of Metal Oxidation Volume I—
Fundamentals, Series in Defects in Crystalline Solids,
North-Holland, Oxford, 1976, vol. 9.
7. A. T. Fromhold, Jr, Book: Theory of Metal Oxidation, Volume
II—Space Charge, Series in Defects in Solids, North-Holland,
Oxford, 1980, vol. 12.
8. A. T. Fromhold, Jr, Growth rate of low-space-charge oxides on
spherical metal particles, J. Phys. Chem. Solids, 1988, 49,
1159–1166.
9. R. Karmhag, G. Niklasson and M. Nygren, Oxidation kinetics of
nickel nanoparticles, J. Appl. Phys., 2001, 89(5), 3012–3017.
10. R. E. Carter, Kinetic model for solid-state reactions, J. Chem.
Phys., 1931, 34, 2010.
11. G. Niklasson and R. Karmhag, Oxidation kinetics of metallic
nanoparticles, Surf. Sci., 2003, 532–535, 324–327.
12. W. Tseng, C. Hsu, C. Chi and K. Teng, Thermal and micro-
structural characterizations of nickel nanoparticles at elevated
temperatures, Mater. Lett., 2002, 52, 313–318.
13. T. Uchikoshi, Y. Sakka, M. Yoshitake and K. Yoshihara, A study
of the passivating oxide layer on fine nickel particles, Nanostruct.
Mater., 1994, 4, 199–206.
14. A. Ortega, The kinetics of solid-state reactions toward consen-
sus—Part I: Uncertainties, failures, and successes of conventional
methods, Int. J. Chem. Kinet., 2001, 33, 343–353.
This journal is 
c the Owner Societies 2008

15. R. Karmhag and G. Niklasson, Oxidation kinetics of small nickel
particles, J. Appl. Phys., 1999, 85, 1186–1191.
16. H. Cichy and E. Fromm, Oxidation Kinetics of Metal Films at 300
K Studied by the Piezoelectric Quartz Crystal Microbalance
Technique, Thin Solid Films, 1991, 195, 147–158.
17. M. A. Trunov, M. Schoenize, Zhu and E. L. Dreizin, Effect of
polymprphic phase transformations in Al2O3 film on oxidation
kinetics of aluminum powders, Combust. Flame, 2005, 140,
310–318.
18. C. D. Doyle, Series approximations to the equation of thermo-
gravimetric data, Nature, 1965, 207, 290.
19. P. G. Boswell, On the calculation of activation-energies using a
modified Kissinger method, J. Therm. Anal., 1980, 18, 353–358.
20. E. J. Mittemeijer, Analysis of the kinetics of phase-transiforma-
tions, J. Mater. Sci., 1992, 27, 3977–3987.
21. T. Ozawa, Estimation of activation-energy by isoconversion
methods, Thermochim. Acta, 1992, 203, 159–165.
22. J. W. Graydon, S. J. Thorpe and D. W. Kirk, Interpretation of
Published on 27 June 2008 on http://pubs.rsc.org | doi:10.1039/B800672E
activation energies calculated from non-isothermal transforma-
tions of amorphous metals, Acta Metallurg. Mater., 1994, 42,
3163–3166.
Downloaded by Texas A & M University on 25 March 2013
23. M. J. Starink, A new method for the derivation of activation
energies from experiments performed at constant heating rate,
Thermochim. Acta, 1996, 288, 97–104.
24. H. E. Kissinger, Reaction kinetics in differential thermal analysis,
Anal. Chem., 1957, 29, 1702–1706.
25. ASTM standard E698-79 (reapproved 1984).
26. H. L. Friedman, New methods for evaluating kinetic parameters
from thermal analysis data, J. Polym. Sci., Part B: Polym. Lett.,
1969, 7, 41–46.
27. S. Han, H. Chen, C. Chen, T. Yuan and H. C. Shih, Character-
ization of Ni nanowires after annealing, Mater. Lett., 2007, 61,
1105–1108.
This journal is 
c the Owner Societies 2008
View Article Online
28. B. Sasi and K. G. Gopchandran, Nanostructured mesoporous
nickel oxide thin films, Nanotechnology, 2007, 18, 115613, p. 9.
29. F. Allouti, L. Manceron and M. E. Alikhani, The Ni2 + O2
reaction: the IR spectrum and structure of Ni2O2. A combined IR
matrix isolation and theoretical study, Phys. Chem. Chem. Phys.,
2006, 8, 3715–3725.
30. Y. Zhou, D. Gu, Y. Geng and F. Gan, Thermal, structural and
optical properties of NiOx thin films deposited by reactive
dc-magnetron sputtering, Mater. Sci. Eng., B, 2006, 135,
125–128.
31. A. Atkinson and R. I. Taylor, The diffusion of 63Ni along grain
boundaries in nickel oxide, Philos. Mag. A, 1981, 43, 979–998.
32. B. C. Sales and M. B. Maple, Oxidation of nickel in the vicinity of
its Curie temperature, Phys. Rev. Lett., 1977, 39, 1636–1639.
33. A. K. Galwey, Is the science of thermal analysis kinetics based on
solid foundations? A literature appraisal, Thermochim. Acta, 2004,
413, 139–183.
34. A. Atkinson, Transport processes during the growth of oxide films
at elevated temperature, Rev. Mod. Phys., 1985, 57, 437–470.
35. M. J. Graham, D. Caplan and M. Cohen, Growth via leakage
paths of nickel oxide on nickel at high temperatures,
J. Electrochem. Soc., 1972, 119, 1265–1267.
36. R. Peraldi, D. Monceau and B. Pieraggi, Correlations between
Growth Kinetics and Microstructure for Scales Formed by
High-Temperature Oxidation of Pure Nickel, Oxidat. Met.,
2002, 58, 249–273.
37. K. Park, D. Lee, A. Rai, D. Mukherjee and M. R. Zachariah,
Size-resolved kinetic measurements of aluminum nanoparticle
oxidation with single particle mass spectrometry, J. Phys. Chem.
B, 2005, 109(15), 7290–7299.
38. M. A. Trunov, S. M. Umbrajkar, M. Schoenitz, J. T. Mang and
E. L. Dreizin, Oxidation and melting of aluminum nanopowders,
J. Phys. Chem. B, 2006, 110(26), 13094–13099.
Phys. Chem. Chem. Phys., 2008, 10, 5057–5065 | 5065