Full text of DOI:10.1557/JMR.2002.0308

An in situ transmission electron microscope investigation into
grain growth and ordering of sputter-deposited
nanocrystalline Ni3Al thin films
H.P. Nga) and A.H.W. Ngan
Department of Mechanical Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong,
People’s Republic of China
(Received 10 March 2002; accepted 28 May 2002)
The grain growth kinetics and ordering behavior of direct-current magnetron
sputter-deposited Ni75at.%Al25at.% alloy films were investigated using in situ isothermal
annealing in a transmission electron microscope. Both normal and abnormal grain
growth modes were observed. The normal grain growth kinetics under isothermal
.
heating from 300 to 700 °C were found to comply with the Burke law d ס K/dn−1
where d is grain size and K and n are constants with respect to time. The grain
boundary mobility parameter K was found to obey an Arrehnius rate law with an
,
apparent activation energy of 1.6 eV, and n was found to increase gradually from 5.2
at 300 °C to 8.7 at 700 °C. Abnormal grain growth occurred at 500 °C or higher, and
grain coalescence was identified as an important operative mechanism. It was also
observed that the initially as-deposited state of the films was crystalline with a
disordered face-centered-cubic structure, but ordering into the equilibrium
L12 intermetallic structure followed from annealing at temperatures above
approximately 500 °C.
I. INTRODUCTION
The transition effect is found to be reversible with respect
to temperature changes, implying that the transition hap-
pens at constant microstructure. This is to be distin-
guished from another effect of electrical resistance
variation due to annealing of the film’s microstructures.
The as-deposited state of the Ni75at.%Al25at.% films is
known to be nanocrystalline with a typical grain size of
the order of a few nanometers.12,13 Such a refined
nanocrystalline configuration, together with a significant
oxygen content measured in the Ni75at.%Al25at.% films, is
believed to be a factor leading to the observed electrical
transition effect.12,14
The reversible transition is found to occur over a tem-
perature range from room temperature to about 200 °C,
within which grain growth is suppressed. One possible
application of the reversible transition effect in nanocrys-
talline Ni75at.%Al25at.% thin films would be as thermal
switches in microelectromechanical systems, but the ap-
plicability hinges on the thermal stability of the film with
respect to grain growth and other microstructural
changes. There are at least two reasons why grain growth
in nanocrystalline materials can be potentially very dif-
ferent from the situation in conventional large grained
materials. The first is the absence of dislocations in the
nanocrystalline state, and hence one would not expect
recovery or recrystallization to precede grain growth.
This point is of crucial importance to processing routes
involving ball milling in the powder metallurgy.15 Sec-
ondly, the physical dimension of the grain boundaries in
Intermetallic films are being actively investigated for
potential use as functional coatings for engineering ap-
plications ranging from aeroengines to microelectronic
devices. Numerous studies have been dedicated to inter-
metallic aluminide coatings in the last few years, with
substantial attention paid to the mechanical properties
and high temperature oxidation resistance of Fe–Al,
Ni–Al, and Ti–Al-based coatings.1–7 For instance, the
present authors and Xu et al. have reported that Ni3Al
coatings can provide Ni-based alloys with significantly
improved surface strength,3,4 as well as effective protec-
tion from thermal oxidation.3 Similarly, Leyens et al.
showed that Ti–Al-based coatings exhibit a variety of
beneficial properties on the oxidation, creep, and fatigue
of TIMETAL 1100 high temperature alloys.5,6,8 On the
other hand, the metallization of semiconductors by inter-
metallic aluminides like Ni–Al and Ti–Al as intercon-
nects represents a growing field of interest in the
microelectronic industry.9–11
In a recent paper, we demonstrated that direct-current
(dc) magnetron sputter-deposited Ni75at.%Al25at.% thin
films exhibit a phenomenal electrical transition from an
insulating state to a conducting state upon mild heating.12
a)Present address: Laboratoire des Mate´riaux et du Ge´nie Phy-
sique, ENSPG-INPG UM 5628, BP 46, 38402, St. Martin
d’He`res, France.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
© 2002 Materials Research Society
2085
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
a nanocrystalline material would limit the formation of
conventional second phase particles in large grained ma-
terials, and hence the conventional Zener drag mecha-
nism of grain boundary pinning would no longer be
applicable. The present paper therefore represents an at-
tempt to study the kinetics of grain growth and ordering
in nanocrystalline Ni75at.%Al25at.% thin films.
epoxy (Epoxy Technology EPO-TEK 353ND, Billerica,
MA), which is specifically for high-temperature trans-
mission electron microscopy (TEM) experiments. The
cylindrical assembly was then sectioned by a low-speed
diamond saw into roughly 200-␮m-thick discs, followed
by gently polishing on the cut surfaces to a thickness of
less than 80 ␮m. The C-ring here played an important
role to protect the films from delamination during slicing
or polishing. Prior to thinning, the specimens were
soaked in acetone to remove the layer of Aron-Alpha௡
glue between the films, which would otherwise melt un-
der in situ heating in the TEM. The glue served a tem-
porary purpose to protect the film surface from the
intrusion of the G1 epoxy by capillary effect in its fluid
state, which could otherwise severely affect the imaging
and structural characterization of the cross-sectional
films under TEM. Finally, thinning of the discs was per-
formed by twin-jet electropolishing using a mixture of
90% methanol + 10% perchloric acid at 5 V and −50 °C
to obtain electron transparency near the film–substrate
interfaces. Precaution was taken to protect the film from
excessive impulse of the jets by sandwiching the speci-
men in aluminium foils before activating the jet flow.
The aluminium foils serve additionally as electrical con-
tacts to bridge the C-ring and the central core, which are
otherwise electrically insulated by the epoxy.
Atomic force microscopy (AFM) was applied to char-
acterize the planar surface morphology of the as-
deposited Ni75at.%Al25at.% films. AFM scanning was
performed in the contact mode using a Digital Instru-
ments Nanoscope IIIA controlled microscope (Santa
Barbara, CA). In situ TEM heat treatments of the speci-
mens were carried out on an Oxford resistive hot-stage in
a JOEL 2000FX TEM (Tokyo, Japan) operated at
200 kV. To study the grain growth kinetics as a function
of time and temperature, dynamic microstructural obser-
vations were carried out on the XTEM specimens during
a series of isothermal in situ heat treatments, each last-
ing for approximately 50 min and at different tempera-
tures up to the nominal instrumental limit of 800 °C. A
constant heating rate of 20 °C/min was used for all tem-
perature increments throughout.
II. EXPERIMENTAL PROCEDURES
The Ni–Al thin films were deposited by a dc planar
magnetron sputtering device (Bal-Tec MED-020) (BAL-
TEC AG, Balzers, Liechtenstein) with a water-cooled
Ni–27.4% Al alloy target, which is capable of yielding
the stoichiometric Ni3Al composition through control-
ling a number of parameters reported elsewhere.3 In
this work, the base sputtering pressure used was
1 × 10−6 mbar or better, compared to the 1 × 10−5 mbar
or higher in our previous study.12 Commercially pure
(>99.5%) nickel substrates with a thickness of 1 mm was
homogenized at 1000 °C for 24 h prior to surface treat-
ments and deposition for all experiments involving post-
deposition annealing so that any possible microstructural
changes in the substrates that may influence the films
could be avoided. The substrates were mechanically
ground and polished down to 1 ␮m followed by solvent
cleaning in ethyl alcohol and acetone to remove organic
contaminations. The in-line target-to-substrate distance
was fixed at 80 mm. To ensure the purity of the deposit,
the target was presputtered for more than 30 min with the
substrate protected by a shutter before actual deposition
on the substrate. The sputtering gas was ultrahigh purity
argon (>99.9999%) at a pressure of 5 × 10−2 mbar. All
deposition was performed at a sputtering current of
90–100 mA, which is equivalent to a sputtering power
of 70 W or a deposition rate of 2–3 Å/s. No intentional
heating was applied to the substrates in the course of
deposition, although the plasma discharge would be able
to raise the substrate temperature to about 100 °C. The
Ni75at.%Al25at.% films prepared for the present grain
growth analysis were all 2 ␮m in thickness unless oth-
erwise stated.
The film specimens for in situ cross-sectional trans-
mission electron microscopy (XTEM) were prepared us-
ing a special “C-Ring” method. In this method, nickel
substrates with a semihexagonal cross-section (2 mm
wide and 30 mm long) were prepared by electric dis-
charge wire cutting and deposited with Ni75at.%Al25at.%
films on their longitudinal surfaces. A pair of such coated
substrates were glued together with their coated surfaces
facing each other using the Aron-Alpha௡ instant-glue
(Tokyo, Japan). The resultant hexagonal substrate col-
umn was then firmly inserted and glued into a tailor-
made 3-mm-diameter metallic C-ring cylinder bored
with a matching hexagonal empty core using the G1
The determination of grain size was performed on
digitized XTEM micrographs with the aid of an image
analyzer software. The experimental grain size distribu-
tions were fitted with Gaussian distributions, and the
Gaussian mean and standard deviation were computed.
Crystallographic information of the nanocrystalline
Ni75at.%Al25at.% films were obtained by selected area dif-
fraction (SAD). The lattice constant a is deduced from
the observable diffraction rings or diffraction spots on a
SAD pattern according to the relationship
2
2
2
͌
h + k + l
xhkl
=
,
(1)
a
L␭
2086
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
B. Grain growth and microstructure evolution
during in situ annealing in TEM
where (hkl) is the diffraction plane indices, xhkl the radii
of the diffraction rings, L the camera length, and ␭ the
electron wavelength. The values of xhkl were mea-
sured using the software ProcessDiffraction16 (Research
Institute for Technical Physics and Materials Science,
Budapest, Hungary) on SAD patterns digitized at a
resolution of 300 dpi, which offers an equivalent
instrumental resolution of better than 85 ␮m on a pho-
tographic print.
Figure 3 shows the grain size distributions of film at
different stages of isothermal heat-treatments at tempera-
tures ranging from 300 to 700 °C. The annealing time
given in Fig. 3 was measured from the onset of attain-
ment of the set-point temperature and is used here to
indicate the grain growth rate after the set-point tempera-
ture was reached. In heat-treatments at high temperatures
such as 700 °C, significant grain growth had already oc-
curred before the set point temperature was reached, and
so the time used in Fig. 3 does not correspond to the
actual time lapsed from the onset of grain growth. The
grain size distributions at 300 °C [Fig. 3(a)] follow a
nearly symmetrical bell shape and can be satisfactorily
fitted by Gaussian distributions. Normal grain growth,
represented by a steady shift of the overall population to
the right side of the graphs as time increases, can be seen
to be the dominating growth behavior of the Ni–Al film
in the lower temperature regime. However, abnormal
grain growth became operative at 500 °C [Fig. 3(b)],
where a minor population of large grains started to
emerge after 30 min of annealing. The coexistence of
normal and abnormal grain growth is well demonstrated
in the histograms at 700 °C [Fig. 3(c)], where the distri-
butions are clearly bimodal. The fast rate of abnormal
grain growth at 700 °C is evidenced by the speedy propa-
gation of the second peak from initially at approximately
55 nm to finally at approximately 90 nm in 25 min of
annealing. The normal grain growth behavior, on the
other hand, showed a much milder grain size increment
during a comparable period of time.
III. RESULTS
A. Nanocrystalline structure and composition of
as-deposited Ni75at.%Al25at.% films
The surface topography of an as-deposited
Ni75at.%Al25at.% thin film deposited on nickel substrate as
revealed by AFM is show in Fig. 1(a). The film appears
to be composed of nanograins with grain diameters of the
order of 10 nm. Figure 1(b) shows a typical XTEM mi-
crograph of the Ni75at.%Al25at.% film near the film and
substrate interface, indicating the actual size of the con-
stituent nanocrystallites to be within 5 to 10 mm in the
as-deposited state. The relative compositions of Ni and
Al in the films were determined by energy dispersive
x-ray (EDX) analysis to be fluctuating about the stoichio-
metric ratio of 3:1 within 0.5 at.%. A typical EDX spec-
trum illustrating the overall chemical composition of
a Ni75at.%Al25at.% film is given in Fig. 2. In addition to
the Ni and Al peaks, a prominent oxygen peak at
0.523 keV, with an intensity comparable to those of the
Ni and Al counterparts in the spectrum, can be observed.
Such a significant oxygen content has been consistently
observed in our Ni75at.%Al25at.% films despite the use of
a high base vacuum of approximately 10−6 mbar in con-
junction with ultrahigh purity argon as the sputtering gas
in our film deposition process.
To characterize the normal grain growth behavior, the
Gaussian mean grain size d and the standard deviation for
grains growing normally were measured from the fitted
Gaussian distributions of the grain growth histograms
after the abnormal growth peaks were ignored when they
occurred, and these data are summarized in Table I. The
FIG. 1. Typical nanocrystalline structure of the as-sputtered Ni75at.%Al25at.% film revealed by (a) AFM and (b) XTEM imaging.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
2087
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
small difference in the mean grain size between the pre-
annealed state and the early stages of annealing at 300 °C
(both 7.2–7.3 nm) suggests that grain growth is insignifi-
cant below and up to 300 °C. Upon heat treatment at a
higher temperature, on the other hand, steady normal
grain growth occurred, corresponding to a continuous
increase in the mean grain size from slightly larger than
7 nm in the as-deposited state to about 19 nm towards the
end of the 700 °C heat treatment.
Figure 4 illustrates the evolution of the nanocrystalline
structure of the Ni75at.%Al25at.% film at different stages
of the isothermal annealing process. It is evident in
the figure that the overall grain size remained fairly
uniformly distributed up to a temperature of 400 °C.
Beyond 600 °C, abnormal grain growth became signi-
ficant at a few locations where grains several times
larger than the average size had developed quite sud-
denly out of the matrix. Examples can be found in
Fig. 4(f) at locations pointed at by arrows, where several
abnormal grains (over 100 nm in diameter) emerged after
25 min of annealing at 700 °C. Figure 5 illustrates the
microstructure of a Ni75at.%Al25at.% film undergoing ex-
tensive, abnormal grain growth after prolonged annealing
at 700 °C. A highly nonuniform grain size distribution
resulted.
Figure 6 shows the microstructural evolution in the
XTEM specimen after 15 and 45 min of isothermal treat-
ment at 500 °C, during which abnormal grain growth was
about to begin. A tendency can be observed in the figures
for the nanograins to coalesce to form larger grains of a
few tens of nanometers. Grain coalescence was indeed
observed to be the main mechanism responsible for the
accelerated grain growth at high temperatures.
Figures 7(a)–7(l) capture the morphological changes dur-
ing coalescence between a pair of adjacent nanocrystals.
Interestingly, periodic and alternative changes in the im-
age contrast were found to occur between both grains,
i.e., at first the upper crystal appeared darker than the
lower one in Figs. 7(b)–7(d), but then the lower one be-
came darker than the upper one in Figs. 7(f)–7(j). The
contrast variation of the grains was also accompanied by
slow flickering of the corresponding diffraction patterns,
which are not shown here, and this indicates rotations of
crystallographic orientations of the grains within a small
angular range. The final amalgamation into a single
larger grain is shown in Figs. 7(k) and 7(l), in which the
combined grain exhibited a well-defined border with no
observable interior interface.
FIG. 2. Energy dispersive x-ray (EDX) spectrum showing the typical
chemical composition of a Ni75at.%Al25at.% film as-deposited by mag-
netron sputtering. The prominent OK peak observed in the spectrum is
indicative of a significant oxygen inclusion in the Ni–Al film during
the deposition process.
FIG. 3. Grain size distribution change of Ni75at.%Al25at.% films during isothermal heat treatments respectively at (a) 300 °C, (b) 500 °C, and (c)
700 °C, indicating an abnormal grain growth associated with the latter.
2088
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
TABLE I. The grain size measured for a Ni75at.%Al25at.% film during a series of isothermal annealing at
various temperatures.
300 °C
400 °C
500 °C
700 °C
Time
(min)
d
(nm)
s.d.
(nm)
Time
(min)
d
(nm)
s.d.
(nm)
Time
(min)
d
(nm)
s.d.
(nm)
Time
(min)
d
(nm)
s.d.
(nm)
15
30
45
7.301
8.487
8.981
1.014
0.870
1.621
20
35
50
9.149
9.526
10.530
1.367
1.489
1.541
15
30
45
10.576
11.313
12.153
1.428
2.093
2.200
15
25
40
17.095
17.584
19.156
4.271
5.565
6.910
As-deposited mean grain size ס 7.235 nm (s.d. ס 0.881).
FIG. 4. XTEM micrographs showing the progressive development of the overall nanogranular structure of Ni75at.%Al25at.% films at different stages
of the heat treatment process: (a) before annealing, (b) at 300 °C for 30 min, (c) at 400 °C for 35 min, (d) at 500 °C for 35 min, (e) at 600 °C for
40 min, and (f) at 700 °C for 25 min.
FIG. 6. XTEM micrographs showing the microstructural changes of a
FIG. 5. XTEM of film microstructure after abnormal grain growth at
700 °C.
Ni75at.%Al25at.% film from an annealing time of (a) 15 min to (b)
45 min at 500 °C, during which coalescence of nanocrystals occurred.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
2089
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
C. Growth kinetics analysis
increase gradually from 5.2 at 300°C to 8.7 at 700°C.
Burke17 further proposed that the constant K contains the
Boltzmann factor, i.e.,
Figure 8 shows the log–log plot of the Gaussian mean
grain size during normal grain growth (d) versus time
(t) during isothermal annealing at different tempera-
tures. It can be seen that the isothermal grain growth
approximately obeys the law d ϰ t1/n, where n is a con-
stant. This relationship is consistent with the conven-
tional grain growth law due to Burke17:
Q
K = K exp −
,
(4)
ͩ ͪ
o
kT
where Ko is constant, Q is the activation energy for grain
growth, and k and T have their usual meanings. In Fig. 9
are shown the Arrehnius plot of K versus 1/T, from
which it can be seen that Eq. (4) is verified. The ac-
tivation energy estimated from the slope in Fig. 8 is
about 1.6 eV.
.
d ס K/dn−1
,
(2)
where K is the grain-boundary mobility parameter. Inte-
grating Eq. (2) gives dn ס don ס Knt, where do is the
initial grain size, and if dno is negligible compared with dn,
this becomes
D. Ordering and crystallographic transformations
1
1
Despite the fact that the Ni75at.%Al25at.% films possess
a Ni-to-Al atomic ratio that would fall within the phase
field of the Ni3Al compound in the bulk condition, pre-
vious observations have shown that the low-temperature
as-deposited nanocrystalline Ni75at.%Al25at.% films have
a disordered face-centered-cubic (fcc) crystal structure
ln d͒ = ln t͒ + ln Kn͒
.
(3)
͑
͑
͑
n
n
Thus from the slopes and intercepts of the best straight
lines in Fig. 8, the constants n and K can be obtained at
each annealing temperature. The value of n is found to
FIG. 7. The sequence of grain coalescence recorded for a Ni75at.%Al25at.% film at 800 °C during a constant time interval of 30 s. Note the periodic
changes in image contrast, which corresponds to grain rotations prior to the merging.
FIG. 8. Log–log plot of the Gaussian mean grain size of the
Ni75at.%Al25at.% nanocrystals against the time of isothermal annealing
performed at various temperatures.
FIG. 9. Arrehnius plot of the grain-boundary mobility parameter.
2090
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
instead.12 The minimum temperature required to initiate
ordering into the equilibrium L12 structure was previ-
ously found to be around 500 °C–700 °C,12,18 depending
on the fabrication techniques and processing conditions.
In the present work, 500 °C was found to be an adequate
temperature to trigger the ordering of the Ni75at.%Al25at.%
nanocrystallites inside the TEM.
The crystallographic transformation of the Ni–Al film
during isothermal annealing at 500 °C is indicated by
the changes of the SAD patterns shown in Fig. 10. The
initial crystal structure of the Ni75at.%Al25at.% film is
evidently disordered fcc, as suggested by the ringlike
SAD pattern and the absence of superlattice reflections.
By assuming a cubic crystal structure, the linear fit of
FIG. 10. Indexed SAD patterns (left) and the fits according to Eq. (1) corresponding to the crystallographic structures of a Ni75at.%Al25at.% film
during isothermal annealing at 500 °C for (a) 0 min, (b) 20 min, (c) 65 min, and (d) 165 min.
J. Mater. Res., Vol. 17, No. 8, Aug 2002
2091
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
TABLE II. Values of xhkl measured on photographic prints of SADP for a Ni75at.%Al25at.% film during
different stages of isothermal annealing at 500 °C.
xhkl/cm
h
k
l
ꢀh2 + k2 + l2
t ס 0 min
t ס 20 min
t ס 65 min
t ס 165 min
1
0
1
1
0
1
1
2
2
1
1
2
0
0
1
0
0
1
0
1
0
1
2
1.000
1.414
1.732
2.000
2.236
2.449
2.828
3.000
3.162
3.317
3.464
—
—
0.961
1.113
—
—
1.584
—
0.553
0.813
0.959
1.109
—
1.329
1.561
—
0.553
0.787
0.957
1.112
—
1.322
1.576
—
0.548
0.763
0.940
1.096
1.219
1.344
1.559
1.666
1.731
1.813
1.888
1
1
2
2
2
2
2
3
3
2
—
1.853
—
—
1.857
1.946
—
1.835
1.924
2
2
2
ꢀ
h + k + l against the corresponding xhkl values for the
observable low-order reflections, as tabulated in Table II,
verifies a single disordered fcc phase [Fig. 10(a)]. Order-
ing into the L12 superlattice structure is found to take
place after 20 min of annealing at 500 °C. This is evi-
denced by the formation of an intensified (100) superlat-
tice reflection ring, which appeared hitherto as a diffused
halo in the center of the diffraction pattern at the onset
of the 500 °C heat treatment, as well as the emergence of
additional superlattice reflections such as (110) and
(211) in the form of discrete spots [Figs. 10(b)
and 10(c)]. The transformation from an initially ringed
SAD pattern into a spotty appearance also suggests sig-
nificant grain growth accompanying the ordering pro-
gress. Figure 10(d) shows a SAD pattern of the
Ni75at.%Al25at.% film at t ס 165 min. In spite of an excep-
tional ambiguous spot observed between the (100) and
(110) reflections, the satisfactory matching of all the other
reflections into the L12 structure as evidenced by the linear
ball-milled nanocrystalline metallic materials with a
similar initial grain size and at comparable temperatures.
Varin et al.19 reported that ball-milled L12 TiAl3
nanograins initially grew from several nanometers to 30–
100 nm on annealing at 600 °C for 1 h, depending on the
milling conditions. They used a fast heating rate of
200 °C/min in their annealing experiments, and so the
heating time to the set-point temperature is negligible
compared to the isothermal annealing time they used.
The mean grain growth rate in their study can therefore
be estimated to be about 30 to 100 nm/h at 600 °C. In our
Ni75at.%Al25at.% films, because of the instrumental limi-
tation of the heating rate to only 20 °C/min, the heating
time to a set-point temperature of 700 °C, for example,
was 35 min, and this is not negligible. The data given in
Table I suggests that after 25 min of isothermal anneal-
ing at 700 °C, the grain size reached approximately
18 nm from an initial value of about 7 nm. The mean
grain growth rate during the first hour of annealing,
including the 35 min of heating-up, is thus roughly
11 nm/h. Bearing in mind that the mean growth rate dur-
ing the first hour of heating calculated this way could
only increase with a faster heating rate, the grain growth
rate of our Ni75at.%Al25at.% nanocrystallites can be ex-
pected to have a similar order of magnitude as the
ball-milled TiAl3 nanograins in the study by Varin
et al.19 It is also worth noting that the grain size in
our Ni75at.%Al25at.% film, which was from about 7 nm
as-deposited to a maximum value of about 19 nm after
annealing at 700 °C, is far smaller than the mini-
mum thickness of the TEM disks, which was typically
100–150 nm. The grain growth of the Ni75at.%Al25at.%
nanocrystallites in our XTEM experiments can therefore
be expected to be predominantly three dimensional rather
than two dimensional. In other words, the surface effects
typical in TEM in situ experiments are not important in
the nanocrystalline situation.
2
2
2
ꢀ
plot of h + k + l against xhkl indicates that the nanocrys-
tals were fully ordered into the L12 phase after prolonged
annealing at 500 °C.
There is a clear trend of lattice expansion recorded for
the Ni75at.%Al25at.% nanocrystallites as ordering occurs.
The lattice parameter a deduced from the slopes of the
plots in Fig. 10 according to Eq. (1) had an initial value
of 3.59 Å at the onset of the 500 °C annealing, and it
eventually increased to 3.65 Å after an annealing period
of 165 min after which full ordering occurred. These are
in qualitative agreement with the expected lattice expan-
sion on ordering of the nickel-rich ␥ phase (a ס 3.52 Å)
into the ordered L12 ␥Ј phase (a ס 3.57 Å) in the bulk
condition.
IV. DISCUSSION
The grain growth of the Ni75at.%Al25at.% nanocrystal-
lites measured from our XTEM experiments is of a simi-
lar order of magnitude compared with those observed for
The data in Table I suggest that grain growth became
very sluggish at large grain sizes. The thermal stability of
2092
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
large grain sizes is traditionally ascribed to grain-
boundary pinning mechanisms that hinder the grain
boundary mobility essential for grain growth. As pro-
posed by Burke,17 stagnation of grain growth can be
attributed to various drag forces acting on grain bound-
aries due to segregation of second phase particles, impu-
rities, and pores. The pinning mechanism by second
phase particles, however, is unlikely to be operative in
the nanocrystalline state, leaving impurity or pore drag as
the viable mechanisms for the growth stagnation ob-
served in the present films. In fact, nanocrystalline me-
tallic materials are well known to be susceptible to
impurity contaminations, especially C, N, and O.20–24
The segregation of these interstitial contaminants to grain
boundaries, and/or the formation of oxides or carbides,
were postulated by Morris-Mun˜oz et al.22 and Lee
et al.24 to have contributed to the grain-size stabilization
observed in their nanocrystalline Fe–Al and Ni, respec-
tively. In Fig. 2 and in a previous paper,12 we reported
the observation of a significant amount of oxygen in our
nanocrystalline Ni75at.%Al25at.% thin films deposited by
magnetron sputtering. The absence of oxide phases ob-
servable in the Ni75at.%Al25at.% films during XRD and
TEM characterizations suggests that the oxygen could
either be trapped by means of gas-phase physical adsorp-
tion or, less likely, form amorphous oxides with the me-
tallic species. We have insufficient information at present
to interpret the mechanisms of such a significant oxygen
inclusion effect. However, because it is an impurity in
nature, the oxygen is likely to play an important role on
grain boundary pinning, especially by impurity drag,
against the grain growth during the heat treatments of the
nanograins.
Impurity drag is well-known to be a grain-size depen-
dent effect, considering the fact that the concentration of
impurity segregated at the grain boundaries should in-
crease with the reduction in overall boundary area
brought about by grain growth. This implies that grain
growth will be retarded as the grain size continues to
increase. The kinetics for normal grain growth in our
Ni75at.%Al25at.% films are found to be consistent with the
Burke law in Eq. (2). Although the Burke law was de-
rived based on the assumption of constant driving force,
Michels et al.25 have recently shown that even when a
linear dependence of grain-boundary pinning on grain
size is taken into consideration, the resultant grain-
growth curve can still be accurately described by the
Burke law, albeit a constant rescaling of the boundary
mobility parameter K in Eq. (2). The activation energy
calculated from the temperature dependence of the
boundary mobility parameter using the Burke law would
therefore not be affected by any grain-size dependence of
boundary pinning. The activation energy for grain
growth in the nanocrystalline Ni75at.%Al25at.% is
determined to be 1.6 eV, which is in good agreement
with the value obtained for bulk Ni3Al.26 This sug-
gests that the mechanism for normal grain growth in
nanocrystalline Ni75at.%Al25at.% is similar to that in the
bulk condition.
In addition to normal grain growth, the abnormal
growth of certain Ni3Al grains has also been observed at
elevated annealing temperatures of 500 °C or higher. A
theoretical treatment on two-dimensional abnormal grain
growth performed by Rollett and Mullins27 yields a com-
plex dependence of abnormal grain growth on the origi-
nal grain size as well as on the competition between the
grain boundary mobility and the grain-matrix interfacial
energies. Yet, grain boundary mobility is regarded as the
most important factor for the abnormal growth. The oc-
currence of abnormal grain growth at high temperatures
may thus be associated with a restoration of grain-
boundary mobility for certain grains. This is presumably
possible if boundary-dragging species have a low ther-
mal stability and decompose at high temperatures, a case
of which has been reported previously.24
On the other hand, coalescence in Fig. 7 has been ob-
served to be an operative mechanism for abnormal grain
growth. An interesting feature associated with the coa-
lescence process is the rotation of the adjoining grains,
presumably required for crystallographic alignment be-
fore the subsequent merging of the two grains. Our ob-
servation is in good agreement with the work of Haslam
et al.,28 who performed numerical simulation on the
growth of nanocrystalline fcc metals and concluded that
rotation coalescence is crucial to the growth of
nanograins rather than the curvature-driven grain-
boundary migration alone. The driving force (or torque)
for the grain rotation might arise from the tendency of the
grains to reduce their overall grain boundary energy due
to misorientations. In Fig. 7, one cycle of the grain rota-
tion completes in a few minutes at 800 °C (note the time
scale in Fig. 7). To our knowledge, no such rapid grain
rotation has ever been observed in the coalescence of
large grains. The reason may be that inertia scales with
the volume while the driving force for rotation scales
with the surface area. Hence the speed of grain rotation
should have a significant size effect; namely, rapid rota-
tion can only occur at very small grain sizes.
Another interesting phenomenon in nanocrystalline
Ni75at.%Al25at.% thin films is the disorderness in the as-
deposited state. De Almeida et al.13 have recently studied
radio-frequency magnetron sputter deposited
Ni75at.%Al25at.% thin films, which were found to be or-
dered into the L12 equilibrium phase even in the as-
deposited state. However, using a low-power dc
magnetron sputtering condition as similar to the present
experiments, we have found previously that the as-
deposited state of similar films is disordered,3,12 but or-
dering into the equilibrium L12 phase follows on
annealing. The present results confirm our previous
J. Mater. Res., Vol. 17, No. 8, Aug 2002
2093
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252

H.P. Ng et al.: An in situ TEM investigation into grain growth and ordering of sputter-deposited nanocrystalline Ni3Al thin films
3. H.P. Ng, X.K. Meng, and A.H.W. Ngan, Scripta Mater. 39, 1737
finding that as-deposited Ni75at.%Al25at.% films made us-
ing low-power sputtering indeed have a disordered fcc
structure, and ordering follows on heating at 500 °C or
(1998).
4. W.H. Xu, X.K. Meng, A.H.W. Ngan, X.Y. Chen,and Z.G. Liu,
Mater. Lett. 44, 314 (2000).
above. The lack of thermal agitation under the low-power
sputtering condition is thought to be the reason for not
achieving the equilibrium L12 phase in the as-deposited
state. Another plausible reason may be that the thermo-
dynamic driving force for ordering may have a size ef-
fect; i.e., when the grain size is small enough, surface
effects become increasingly important, which may re-
duce the overall driving force for ordering. A detailed
analysis of the relative phase stability between ␥ and ␥Ј
in the nanometer scale would be required to understand
this potential effect.
5. C. Leyens, K.H. Trautmann, M. Peters, and W.A. Kaysser, Scripta
Mater. 36, 1309 (1997).
6. C. Leyens, J.W. van Liere, M. Peters, and W.A. Kaysser, Surf.
Coat. Technol. 108–109, 30 (1998).
7. H.N. Lee, Z.M. Park, M.H. Oh, K.Y. Kim, and D.M. Wee, Scripta
Mater. 41, 1073 (1999).
8. C. Leyens, M. Peters, and W.A. Kaysser, Adv. Eng. Mater. 2, 265
(2000).
9. J.M.E. Harper and K.P. Rodbell, J. Vac. Sci. Technol. B 15, 763
(1997).
10. X. Federspiel, F. Voiron, M. Ignat, T. Marieb, and H. Fujimoto, in
Advanced Interconnects and Contact Materials and Processes for
Future Integrated Circuits, edited by S.P. Murarka, M. Eizenberg,
D.B. Fraser, R. Madar, and R. Tung (Mater. Res. Soc. Symp.
Proc. 514, Warrendale, PA, 1998), p. 547.
V. CONCLUSIONS
11. J.T. Benoit, S. Chin, R.R. Grzybowski, S.T. Lin, R. Jain,
P. McCluskey, and T. Bloom, in Fourth Int. High Temp. Electron.
Conf. (IEEE, New York, 1998) p. 109
12. H.P. Ng and A.H.W. Ngan, J. Appl. Phys. 88, 2609 (2000).
13. P. de Almeida, R. Scha¨ublin, A. Almazouzi, M. Victoria, and
F. Le`vy, Thin Solid Films 368, 26 (2000).
14. H.P. Ng and A.H.W. Ngan, in Nanophase and Nanocomposite
Materials III, edited by S. Komarneni, J.C. Parker, and H. Hahn
(Mater. Res. Soc. Symp. Proc. 581, Warrendale, PA, 2000), p.
571.
15. V.L. Tellkamp, S. Dallek, D. Cheng, and E.J. Lavernia, J. Mater.
Res. 16, 938 (2001).
16. J.L. La´ba´r, Proc. EUREM 12, edited by L. Frank and F. Ciampor
(Czechoslovak Society for Electron Microscopy, Brno, Czecho-
slovakia, 2000), p. I379.
17. J.E. Burke and D. Turnbull, Prog. Met. Phys. 3, 220 (1952).
18. X.K. Meng, H. Vehoff, and A.H.W. Ngan, J. Mater. Res. 15, 2595
(2000).
The normal grain growth kinetics under in situ isother-
mal heating from 300 to 700 °C in the TEM were found
to obey the Burke growth law. The grain-boundary mo-
bility was found to obey an Arrehnius rate law with an
apparent activation energy of 1.6 eV, which agreed well
with the value in bulk. In addition to the normal grain
growth, abnormal grain growth was also observed. Grain
coalescence was identified as a mechanism for abnormal
grain growth. Rapid grain rotations to achieve crystallo-
graphic alignment were observed as a prerequisite step
prior to coalescence. The initially as-deposited state of
the films was crystalline with a fcc disordered structure.
Ordering into the equilibrium L12 superlattice structure
followed from annealing at temperatures above approxi-
mately 500 °C.
19. R.A. Varin, J. Bystrzycki, and A. Calka, Intermetallics 7, 785
(1999).
20. A. Michels, C.E. Krill, H. Natter, and R. Birringer, in Grain
Growth in Polycrystalline Materials III, edited by H. Weiland,
B.L. Adams, and A.D. Rollet (TMS, Warrendale, PA, 1998), p.
449.
21. T.R. Marlow and C.C. Koch, Acta Mater. 45, 2177 (1997).
22. M.A. Morris-Mun˜oz, A. Dodge, and D.G. Morris, Nanostruct.
Mater. 11, 873 (1999).
ACKNOWLEDGMENT
The work described in this paper was supported by a
grant from the Research Grants Council of the Hong
Kong Special Administrative Region, People’s Republic
of China, China (Project No. HKU 7077/00E).
23. M. Jurczyk, K. Smardz, W. Rajewski, and L. Smardz, Mater. Sci.
Eng. A 303, 70 (2001).
24. J. Lee, F. Zhou, K.H. Chung, N.J. Kim, and E.J. Lavernia, Metall.
Mater. Trans. 32A, 3109 (2001).
25. A. Michels, C.E. Krill, H. Ehrhardt, R. Birringer, and D.T. Wu,
REFERENCES
1. R.N. Wright, J.R. Fincke, W.D. Swank, D.C. Haggard, and
C.R. Clark, in Elevated Temperature Coatings: Science and Tech-
nology I, edited by N.B. Dahotre, J.M. Hampikian, and
J.J. Stiglich (Proc. High Temperature Coatings I, TMS, Warren-
dale, PA, 1995), p. 157.
2. Z. Gonzalez, J.G. Rodriguez, M. Casales, M. Amaya, and
L. Martinez, Brit. Corr. J. 36, 65 (2001).
Acta Mater. 47, 2143 (1999).
26. M.D. Baro`, S. Surinach, J. Malagelada, M.T. Clavaguera-Mora,
S. Gialanella, and R.W. Cahn, Acta Metall. Mater. 41, 1065
(1993).
27. A.D. Rollett and W.W. Mullins, Scripta Mater. 36, 975 (1997).
28. J. Haslam, S.R. Phillpot, D. Wolf, D. Moldovan, and H. Gleiter,
Mater. Sci. Eng. A 318, 293 (2001).
2094
J. Mater. Res., Vol. 17, No. 8, Aug 2002
http://journals.cambridge.org
Downloaded: 22 Mar 2015
IP address: 169.230.243.252