You are on page 1of 4

REPORTS

which vary from one crystal face to the next (3).


The total dipole is expected to vanish with
increasing size, because contributions from opposing faces tend to cancel each other. The
large permanent dipoles in V, Ta, and Nb clusters and their unusual size and temperature dependence point to a different mechanism.
Moreover, the fact that this extraordinary ferroelectric state is enhanced for even clusters and
suppressed for odd ones is important (35). Electron pairing in the free electron model should
result in the opposite effect: The unpaired and
hence more loosely bound electron should enhance the polarizability (1).
Metal cluster properties are characteristically closely related to those of their bulk
counterparts, allowing their evolution to be
traced from the atom to the bulk (1). The fact
that room-temperature polarizabilities of Nb
clusters are well approximated using bulk
metal properties even for the smallest sizes
exemplifies this principle. By inference, the
observed transition to a ferroelectric state at
low temperatures is likely to be related to a
corresponding transition in bulk Nb, and superconductivity is an obvious candidate. The
assumption is strengthened by the observation that electron pairing correlations in niobium lead to superconductivity, whereas they
appear to be related to the ferroelectric state
in small clusters. This, combined with the
correspondence of the superconducting transition temperatures of Ta, V, and Nb and the
critical temperatures associated with the ferroelectric state (Fig. 5), point to a common
physical origin for both properties.
References and Notes

1. W. A. de Heer, Rev. Mod. Phys. 65, 611 (1993).


2. V. V. Kresin, K. D. Bonin, Electric-Dipole Polarizabilities of Atoms, Molecules and Clusters (World Scientic, River Edge, NJ, 1997).
3. J. M. Ziman, Principles of Solids (Cambridge Univ.
Press, Cambridge, 1972).
4. B. Matthias, in Ferroelectricity, E. F. Weller, Ed.
(Elsevier, Amsterdam, 1967), pp. 176 182.
5. Whenever a symmetry is spontaneously broken, according to the Goldstone theorem there will then be
a gapless excitation that tends to restore the symmetry. Well-known examples are phonons for the
crystalline state and spin waves for the ferromagnetic
state [see, for example (6)].
6. P. W. Anderson, Basic Notions of Condensed Matter
Physics (Benjamin, Menlo Park, CA, 1984).
7. P. Ring, P. Schuck, The Nuclear Many-Body Problem
(Springer, New York, 1980).
8. H. Stern, Phys. Rev. 147, 94 (1966).
9. W. D. Knight, K. Clemenger, W. A. de Heer, W. A.
Saunders, Phys. Rev. B 31, 2539 (1985).
10. R. Schafer, S. Schlecht, J. Woenckhaus, J. A. Becker,
Phys. Rev. Lett. 76, 471 (1996).
11. E. Benichou et al., Phys. Rev. A 59, R1-R4 (1999).
12. M. B. Knickelbein, J. Chem. Phys. 115, 5957 (2001).
13. G. Scoles, Atomic and Molecular Beam Methods (Oxford Univ. Press, Oxford, 1988); W. A. de Heer, P.
Milani, Rev. Sci. Instrum. 62, 670 (1991).
14. See supporting material on Science Online.
15. S. A. Blundell, C. Guet, R. R. Zope, Phys. Rev. Lett. 84,
4826 (2000).
16. G. Herzberg, Molecular Spectra and Molecular Structure, III Electronic Spectra and Electronic Structure of
Polyatomic Molecules (Van Nostrand, Princeton, NJ,
1966).

17. C. W. Townes, A. L. Shawlow, Microwave Spectroscopy (Dover, New York, 1975).


18. The classical xed dipole on a symmetric top model
by P. Dugourd et al. (19) predicts a symmetric broadening of the beam at low elds and asymmetric
broadening at high elds. The model fails both qualitatively and quantitatively to describe the observed
deections, which show essentially undeected
peaks superimposed on relatively at, extended, single-sided tails as shown in Fig. 1. The quantum mechanical model suggested here emphasizes the importance of very weak interactions between quantum
mechanical levels (leading to a dense system of small
avoided crossings) for which these deection measurements are extremely sensitive: Gaps as small as 1
MHz already produce observable effects. These are in
principle absent in the classical model. Also note that
the observed depletions are not due to spontaneous
ionization effects in the deection elds. This possibility (as well as others) was experimentally ruled out
by applying uniform electric elds of similar magnitude to the beam, which did not cause depletion.
19. P. Dugourd et al., Chem. Phys. Lett. 336, 511 (2001).
20. This simplied rotational spectrum is intended to
demonstrate the principle that the general features
are preserved when a symmetric rotor spectrum (involving J and K quantum numbers) is used (17).
21. In general, one expects that the dipole is coupled to
a symmetry axis in the cluster [so the P* P0KMJ/
J( J 1)]; however, the model appears to t the data
better when P0 is coupled to J. Also, it is probably
more realistic to assume that for E 0, the symmetry is not broken, so that P0 should be replaced by
[(P0E)2 G 2], where G is the tunneling splitting
between the aligned and antialigned states.
22. The avoided crossing model also explains anomalous
magnetic deections of paramagnetic alkali clusters
(23, 24).
23. W. A. de Heer, W. D. Knight, in Proceedings of the
13th International School, in Erice, G. Benedek, T. P.
Martin, G. Pacchioni, Eds. (Springer-Verlag, Berlin,
1988), pp. 45 63.
24. W. A. de Heer, thesis, University of California, Berkeley (1985).
25. For NbN, similar TGs are obtained using only T 20
K and E 80 kV/cm, because these elds sufce to
saturate R so that TG Tlog(1 R). This procedure

26.
27.
28.
29.
30.
31.
32.

33.
34.

35.

36.

was applied to VN and TaN data, because the noise


levels were too high to warrant a two-parameter t..
These properties already occur in trimers (but not in
dimers), making them accessible for rst-principles
calculations.
R. L. Whetten, M. R. Zakin, D. M. Cox, D. J. Trevor, A.
Kaldor, J. Chem. Phys. 85, 1697 (1986).
A. Berces, P. A. Hackett, L. Lan, S. A. Mitchell, D. M.
Rayner, J. Chem. Phys. 108, 5476 (1998).
V. Kumar, Y. Kawazoe, Phys. Rev. B 65, 125403
(2002).
M. B. Knickelbein, S. Yang, J. Chem. Phys., 93, 1476
(1990).
R. Resta, Rev. Mod. Phys. 66, 899 (1994).
The Lorentz polarization catastrophe results from
self-polarization of a system of mutually interacting
polarizable objects ( polarizability , density n) for
which the susceptibilty is n/(1 4n/3). In
small systems, the effect also occurs and is geometry-dependent (P. B. Allen, in preparation).
A. Solovyov et al., Phys. Rev. A 65, 053203 (2002).
Our low-temperature measurements on Co, Mn, Bi,
and AlCo do not exhibit any evidence for permanent
dipoles (W. de Heer et al., in preparation); alkali
cluster measurements at high temperatures also have
not presented evidence for permanent dipoles.
Magnetic deection measurements of Nb, V, and Ta
clusters at low temperatures indicate S 1/2 for odd
clusters and S 0 for even clusters [except for Nb2,
which is a triplet (W. de Heer, in preparation)], indicating a nondegenerate energy level structure (1),
with spacings greater than kT.
The authors gratefully acknowledge P. Poncharal and
P. Keghelian for the development of the apparatus
and R. W. Whetten and U. Landman for stimulating
discussions. Financial support was provided by the
U.S. Department of Defense (grant no. DAAG55-970133).

Supporting Online Material


www.sciencemag.org/cgi/content/full/300/5623/1265/
DC1
SOM Text
Fig. S1
10 February 2003; accepted 22 April 2003

Thin-Film Transistor Fabricated in


Single-Crystalline Transparent
Oxide Semiconductor
Kenji Nomura,1,2* Hiromichi Ohta,1 Kazushige Ueda,2
Toshio Kamiya,1,2 Masahiro Hirano,1 Hideo Hosono1,2
We report the fabrication of transparent eld-effect transistors using a singlecrystalline thin-lm transparent oxide semiconductor, InGaO3(ZnO)5, as an
electron channel and amorphous hafnium oxide as a gate insulator. The device
exhibits an on-to-off current ratio of 106 and a eld-effect mobility of 80
square centimeters per volt per second at room temperature, with operation
insensitive to visible light irradiation. The result provides a step toward the
realization of transparent electronics for next-generation optoelectronics.
Transparent electronic circuits (1) are expected to serve as the basis for new optoelectronic
devices. A key device for realizing transparent circuits is the transparent field-effect transistor (TFET). TFETs have been developed
on the basis of compound wide band gap
semiconductors such as GaN (2) and SiC (3).
These exhibit good performance [e.g., a fieldeffect mobility of 140 cm2 V1 s1] and

durability in high-temperature and highpower operation. Oxide semiconductors


present an alternative opportunity for discovering new transparent electronics applications with added functionality, because
oxides display many properties in their
magnetic and electronic behavior that originate from a variety of crystal structure and
constituent elements.

www.sciencemag.org SCIENCE VOL 300 23 MAY 2003

1269

REPORTS
Transparent conductive oxides such as
indium-tin oxide (ITO) and ZnO have found
applications as electrical interconnections and
as window electrodes in flat panel displays and
solar cells. The discovery of a p-type transparent oxide semiconductor (TOS), CuAlO2 (4),
and the development of key device components
fabricated from TOSs, such as pn-junction
rectifiers (5, 6) and ultraviolet light emitting
diodes (7), have led to the emergence of TOSs
as viable materials for further development of
transparent electronics.
However, the TFETs fabricated to date
using conventional TOSs such as SnO2 and
ZnO (8, 9) exhibit poor performance. For
instance, their on-to-off current ratios and
field-effect mobilities are on the order of 103
and as low as 10 cm2 V1 s1, respectively,
and the device exhibits normally-on characteristics. Although a polycrystalline ZnO
TFET has previously been found to have
normally-off characteristics with an on-tooff current ratio of 107, its field-effect mobility is only 3 cm2 V1 s1. Grain-boundary
potential barriers are thought to limit the
performance (10). The large off-current and
the normally-on characteristics may originate
from the fact that these conventional TOSs
contain many carriers in the as-prepared state
(as the result of a somewhat large nonstoichiometry in the chemical composition),
making it difficult to control the carrier density down to less than 1017 cm3 without
counterdoping of acceptors. Hence, it is imperative to choose a material that can control
the carrier concentration down to the intrinsic
level and to develop a method to grow a
high-quality single-crystalline thin film.
We report the fabrication and performance of TFETs that use a single-crystalline
film of a TOS, InGaO3(ZnO)5, for the active
channel layer. This material has advantages
over conventional TOSs, including easy
growth of a high-quality single-crystalline
film and good controllability of carrier concentration. The use of a gate insulator with a
high dielectric constant (high-k dielectric),
amorphous HfO2, was found to improve the
FET performance.
The structure of InGaO3(ZnO)5 is characterized by its layered superlattice structure
(Fig. 1A)in which InO2 layers and
GaO(ZnO)5 blocks are alternately stacked
along the 0001 axis (11, 12)and is
thought to be the origin of its superior
electronic properties. The layers are similar
to those of ITO and Ga-doped ZnO, in

which carrier doping is controlled by the


amount of Ga. However, the Ga3 ion incorporated in the GaO( ZnO)5 block does
not generate carriers in InGaO3(ZnO)5 because the Ga3 ion does not substitute the
Zn2 tetrahedral sites only, but also takes
trigonal-bipyramidal coordination sites,
which keeps the local electroneutrality.
Moreover, the In2O3 layer may work as a
blocking barrier for oxygen outdiffusion, and
thereby may suppress the formation of oxygen vacancy; this idea is supported by our
transmission electron microscope observation
that mass transport proceeds much faster in
the GaO( ZnO)5 blocks than across the
InO2 and GaO( ZnO)5 layers. It is therefore
easier to maintain the material in stoichiometry and control the carrier concentration
down to the intrinsic level in a single crystal.

1
Hosono Transparent ElectroActive Materials, Exploratory Research for Advanced Technology (ERATO),
Japan Science and Technology ( JST ), 3-2-1 Sakado,
Takatsu, Kawasaki 213-0012, Japan. 2Materials and
Structures Laboratory, Tokyo Institute of Technology,
4259 Nagatsuta, Midori, Yokohama 226-8503, Japan.

Fig. 1. Structure of InGaO3(ZnO)5. (A) Schematic of the crystal structure. A HRTEM lattice image is
shown for comparison. The InO2 layer (In3 ion locates at an octahedral site coordinated by oxygens)
and the GaO(ZnO)5 block (Ga3 and Zn2 ions share trigonal-bipyramidal and tetrahedral
sites) are alternately stacked along the 0001 direction at a period of 1.9 nm (d0003). (B and
C) Cross-sectional HRTEM images of a InGaO3(ZnO)5 thin lm grown on YSZ(111) by reactive
solid-phase epitaxy. Periodic stacking of the InO2 layer and the GaO(ZnO)5 block is clearly
visible, which is also conrmed in the electron diffraction image [(C), inset]. Single-crystalline
lm is formed over the entire observation area. The topmost layer of the lm is the InO2 layer.

*To whom correspondence should be addressed. Email: nomura@lucid.msl.titech.ac.jp

1270

However, because of the complex structure and composition, it is difficult to obtain


single-crystalline films of such oxides with
the use of a conventional vapor-phase growth
technique alone. Complex oxides, in general,
require high temperatures to grow in singlecrystalline phase, and some chemical components may evaporate at lower temperatures in
the vacuum deposition chamber. We have
developed a reactive solid-phase epitaxy (RSPE) technique that can be used to grow
single-crystalline TOS thin films with an
atomically flat surface (13, 14). The selection
of the film formation technique is not crucial
as long as a suitable template layer is formed.
We have shown that this concept is applicable to a variety of materials (15).
Pulsed laser deposition (PLD) (16, 17)
was used to deposit a 2-nm-thick ZnO epi-

23 MAY 2003 VOL 300 SCIENCE www.sciencemag.org

REPORTS
taxial layer at 700C on a (111) single-crystal
yttriastabilized zirconia ( YSZ ) substrate as
the template, followed by a 120-nm-thick
InGaO3( ZnO)5 layer at room temperature.
The resulting bilayer structure was covered
with a YSZ plate to suppress the evaporation
of film components and was then subjected to
thermal annealing at 1400C for 30 min in an
atmospheric electric furnace, resulting in the
growth of its single-crystalline phase (18).
Cross-sectional high-resolution transmission electron microscopy (HRTEM) images
of the InGaO3(ZnO)5 film (Fig. 1B) showed
the distinct layered lattice structure composed
of the periodic stacking of the InO2 layers
and GaO(ZnO)5 blocks. The film-substrate
interface is atomically flat without a reaction
layer despite the high-temperature annealing,
as confirmed by TEM energy-dispersive xray spectrum. The lattice mismatch is relaxed
in a few atomic layers at the film-substrate
interface, and field-emission scanning microscopic observation revealed no defect structure such as grain boundary and dislocation
over the entire area.
Figure 2 shows an atomic force microscope image of a single-crystalline
InGaO3(ZnO)5 thin film as fabricated, showing an atomically flat terraces-and-steps
structure. The step height (1.9 nm) corresponds to the separation between adjacent
InO2 layers in the InGaO3(ZnO)5 crystal.
The topmost layer is made of InO2, as observed by HRTEM. The film conductivity is
less than 105 S cm1. The carrier concentration is estimated to be 1013 cm3, as
derived from an electron mobility value of
80 cm2 V1 s1 (which is obtained as a
field-effect mobility).
We fabricated top-gate TFETs with the
use of a single-crystalline film grown on a 10
mm by 10 mm YSZ chip (Fig. 3A). The

Fig. 2. AFM image of the InGaO3(ZnO)5 lm.


Atomically at terraces and steps are observed.
The step height is 1.9 nm, corresponding to the
space between the adjacent InO2 layers. The
upper gure shows a cross section measured
along the line A-B.

source, drain, gate contacts, and gate insulator were defined by standard photolithography and lift-off techniques. An 80-nm-thick
amorphous HfO2 (a-HfO2 ) layer was used for
the gate insulator, and ITO (10% Sn) was
used for source, drain, and gate electrodes.
The ITO and a-HfO2 layers were deposited
by PLD at room temperature. The dielectric
constant of a-HfO2 films was measured to be
18, which is a reasonable value compared
with those reported previously (19). The
channel length and gate width were 50 m
and 200 m, respectively, corresponding to a
width-to-length ratio of 4 :1 (Fig. 3B). The
chip is optically transparent in the whole
visible-light region (Fig. 3B, inset). The optical transmittance is 80% in the wavelength range between 390 nm and 3200 nm
[including the effects of the YSZ(111) substrate], which indicates that transmission
losses due to the film and the TFETs are
negligible. The reproducibility of the device
characteristics was confirmed by measuring
more than 100 fabricated TFETs.

Typical TFET characteristics (Fig. 4)


show that source-to-drain (IDS) current increases markedly as source-to-drain voltage
(VDS) increases at a positive gate bias VGS
(Fig. 4A), hence the channel is n-type and
electron carriers are generated by positive
VGS. A large IDS (1 mA) is obtained at
VGS 10 V and VDS 15 V. IDS exhibits a
clear pinch-off and current saturation, which
indicates that the operation of this TFET
conforms to the standard field-effect transistor theory and that the Fermi level in the
channel is fully controlled by the gate and
drain bias. A field-effect mobility eff 80
cm2 V1 s1 is obtained both from the
transconductance value and from the saturation current. The large eff value obtained is
thought to result from high-quality singlecrystalline InGaO3(ZnO)5 thin film and the
improved channel-insulator interface. Note
that TFETs using amorphous aluminum oxide for the gate insulator gave at most a eff
value of only 2 cm2 V1 s1, indicating that
the choice of a gate insulator material is also

Fig. 3. (A) Illustration of the TFET device structure. The InGaO3(ZnO)5 channel layer and the
a-HfO2 gate insulator layer are 120 nm and 80 nm in thickness, respectively. Channel length and
gate width are 50 m and 200 m, respectively. (B) Optical transmission spectrum of a TFET chip.
Those of YSZ substrates with and without an InGaO3(ZnO)5 lm are given for comparison, showing
that the TFET is fully transparent to visible light. Inset: Photograph of a TFET chip placed on a
background text (left) and a magnied photograph of a TFET device (right). The light illumination
condition was tuned to make the TFET device structure visible.

Fig. 4. Typical TFET characteristics fabricated in a single-crystalline InGaO3(ZnO)5 lm. (A) Output
characteristics; (B) transfer characteristics. The TFET operates in the enhanced mode with a
threshold voltage of 3 V. A eld-effect mobility of 80 cm2 V1 s1 and an on-to-off current
ratio of 106 are obtained. The gate leak current is orders of magnitude less than the source-todrain current, which guarantees that the FET characteristics are not affected by the gate leak.

www.sciencemag.org SCIENCE VOL 300 23 MAY 2003

1271

REPORTS
an important factor for achieving good performance (20).
The off-current is very low, on the order of
109 A, and an on-to-off current ratio of 106
is obtained (Fig. 4B). The threshold gate voltage is 3 V, showing that the TFET operates in
the enhancement mode. These characteristics
are much improved over those reported for
TOS TFETs fabricated using SnO2 (9).
We examined the photoresponse against
the light illumination from a commercially
available 30-W fluorescent tube. A photoresponse of the off-current weaker than the
dark level (109 A) was observed under
typical room illumination conditions (1.6
W m2). At six times this intensity of illumination (10 W m2), the off-current increased only to 3 109 A.

References and Notes

1. G. Thomas, Nature 389, 907 (1997).


2. S. Arulkumaran et al., Appl. Phys. Lett. 81, 1131 (2002).
3. S. Harada et al., Mater. Sci. Forum 389 393, 1069
(2002).
4. H. Kawazoe et al., Nature 389, 939 (1997).
5. A. Kudo, H. Yanagi, K. Ueda, H. Hosono, H. Kawazoe,
Appl. Phys. Lett. 75, 2851 (1999).
6. H. Yanagi et al., Solid State Commun. 121, 615 (2001).
7. H. Ohta et al., Appl. Phys. Lett. 77, 475 (2000).
8. M. W. J. Prins et al., Appl. Phys. Lett. 68, 3650 (1996).
9. M. W. J. Prins, S. E. Zinnemers, J. F. M. Cillessen, J. B.
Giesbers, Appl. Phys. Lett. 70, 458 (1997).
10. R. L. Hoffman, B. J. Norris, J. F. Wager, Appl. Phys.
Lett. 82, 733 (2003).
11. N. Kimizuka, M. Isobe, M. Nakamura, J. Solid State
Chem. 116, 170 (1995).
12. C. Li, Y. Bando, M. Nakamura, M. Onoda, N. Kimizuka,
J. Solid State Chem. 139, 347 (1998).
13. H. Ohta et al., Adv. Funct. Mater. 13, 1398 (2003).
14. K. Nomura et al., Thin Solid Films 411, 147 (2000).
15. H. Hiramatsu et al., Appl. Phys. Lett. 81, 598 (2002).

Geometric Origin of Hexagonal


Close Packing at a Grain
Boundary in Gold
G. Lucadamo and D. L. Medlin*
Using electron microscopy, we identify local, intergranular regions of hexagonal
close-packing at a grain boundary in gold. By analyzing the topological defects
that connect this layer to the adjacent face-centered cubic grains, we explain
the geometric origin of this interfacial reconstruction. We extend this analysis
to predict the stacking arrangements found over a range of intergranular
misorientations. These results help to unify our understanding of the defects
that control the behavior of polycrystalline materials by showing how line
defects that are already well understood in the bulk also can determine the
atomic arrangements at grain boundaries.
Grain boundariesthe solid-state interfaces
that separate neighboring crystals of differing
orientationcontrol much of the behavior of
polycrystalline solids (1). For instance, grain
boundaries play a critical role in the plastic
deformation of nanocrystalline metals
through intergranular sliding and by providing nucleation sites for dislocations and extended, planar defects (27). Although a
grain boundary often is conceived as a twodimensional dividing surface, it is now
known that grain boundary cores can spread
into complex, three-dimensional configurations of nanometer-scale width. In metals that
possess a low stacking fault energy (SFE),
such an interfacial reconstruction can occur
by the emission of dense arrays of stacking
faults that extend from the boundary plane
(8). Although there are now many observations and calculations of such boundary reDepartment of Thin Film and Interface Science, Sandia National Laboratories, Post Ofce Box 969, Livermore, CA 94551, USA.
*To whom correspondence should be addressed. Email: dlmedli@sandia.gov

1272

constructions (917), what controls the atomic arrangements within the dissociated layers
has been unclear. Resolving this issue is important because the processes that occur at
grain boundaries, such as the accommodation
of lattice strain or the segregation and diffusion of point defects and impurities, are fundamentally governed by the atomic-scale details of the interfacial structure.
Here, we consider the specific question of
how the ordering of stacking fault arrays at
dissociated grain boundaries is related to the
crystallographic orientations of the two joined
crystals. We focus on the formation of hexagonal close-packing (HCP), which is an important limiting case because it is the close-packed
arrangement that possesses the highest possible
density of stacking faults. Through atomic resolution electron microscopic observations, we
demonstrate the formation of local, intergranular regions of HCP at a grain boundary in
face-centered cubic (FCC) gold. From a detailed analysis of the topological defects required to connect this intergranular layer to the
neighboring FCC crystals, we show how these
observations, as well as the dissociated bound-

16. H. Ohta, H. Tanji, M. Orita, H. Hosono, H. Kawazoe,


Mater. Res. Soc. Symp. Proc. 570, 309 (1999).
17. M. Orita, H. Ohta, M. Hirano, S. Narushima, H.
Hosono, Philos. Mag. B 81, 501 (2001).
18. The thermal annealing temperature can be reduced
to 1000C if the annealing duration is extended to 60
min.
19. Y. S. Lin, R. Puthenkovilakam, J. P. Chang, Appl. Phys.
Lett. 81, 9 (2002).
20. The device showed a hysteresis with 1 V width in
the transfer characteristics. The density of interface
trap state Dit was estimated to be 1012 eV1 cm2
from the subthreshold slope. This value is rather large
relative to that of optimized Si metal oxide semiconductor FETs (Dit 1010 eV1 cm2). These values
will be improved by further optimization of the insulator-channel interface, including the fabrication
condition of a-HfO2 and the reduction of interface
contamination due to air exposure before the formation of an a-HfO2 layer.
7 February 2003; accepted 14 April 2003

ary structures found at other misorientations,


can be understood and predicted from relatively
simple geometric considerations.
For reasons that we discuss below, we
anticipated that HCP stacking would form at
a 110 tilt boundary misorientation of
80.6. We set out to grow a boundary near
this misorientation, choosing gold because its
relatively low stacking fault energy (33
mJ/m2 ) (18, 19) favors grain boundary dissociation and because its tendency to form copious growth twins provides a convenient
route to generating boundaries of controlled
crystallographic orientation. We deposited
gold by thermal evaporation onto a 110oriented NaCl crystal substrate (T 300C).
The resulting thin film microstructure consists of grains that inherit either the orientation of the substrate or an orientation that is
related to the substrate through a series of
70.53 twinning rotations about the 110
axis. Specifically, the boundary that is produced by four successive twinning rotations,
which ideally would possess a misorientation
of 77.9, is considered in this paper. Specimens suitable for high-resolution electron microscopy (HRTEM) were prepared by dissolving the NaCl substrate in deionized water, supporting the free-standing gold film on
a fine-meshed grid, and then thinning the film
to electron transparency by Ar ion milling.
The specimens were observed in a JEOL
4000 EX HRTEM (Japan Electron Optics
Laboratory, Ltd., Tachikawa, Tokyo) operated at 400 kV.
Figure 1A, an atomic-resolution image
projected along a 110-type zone axis,
shows the intersection of five grain boundaries. Four of the boundaries are {111} twins
and possess compact, well-defined cores. The
fifth boundary, however, exhibits a broad,
dissociated core with a width of approximately 1 nm. The extent of this spreading can be
estimated from the set of nearly horizontal
{111} lattice fringes that intersect the two

23 MAY 2003 VOL 300 SCIENCE www.sciencemag.org

You might also like