www.sciencemag.

org/cgi/content/full/333/6045/1003/DC1



Supporting Online Material for

Guided Growth of Millimeter-Long Horizontal Nanowires with Controlled
Orientations
David Tsivion, Mark Schvartzman, Ronit Popovitz-Biro, Palle von Huth, Ernesto Joselevich*


*To whom correspondence should be addressed. E-mail: ernesto.joselevich@weizmann.ac.il


Published 19 August 2011, Science 333, 1003 (2011)
DOI: 10.1126/science.1208455



This PDF file includes:

Materials and Methods
SOM Text
Tables S1 and S2
Figs. S1 to S20
References (3237)
3

1. Materials and Methods
1.1 Substrate preparation
Sapphire (-Al
2
O
3
) wafers with eight different orientations were used. Singular
(i.e., well-cut) sapphire wafers, including C-plane (0001), M-plane ( 0 1 10 ), A-plane
) 0 2 11 ( and R-plane ( 02 1 1 ), were purchased from Roditi International. Miscut C-plane
(0001) sapphire wafers with different miscut directions were custom-produced as
previously described (16,32). The manufacturer (Gavish Ltd., Omer, Israel), produced the
wafers as directed, and the exact miscut tilt and azimuth angles were determined by our
own developed procedure of asymmetric double exposure back-reflection XRD (17).
Miscut C-plane and M-plane sapphire substrates were annealed at 1100 C and 1400 C,
respectively, in air, using a high-temperature muffle furnace (Carbolite RHF 16/8). Prior
to use, the substrates were sonicated for 10 minutes in acetone, then rinsed in acetone,
isopropyl alcohol (IPA), and distilled H
2
O, and blow-dried in N
2
.
1.2 Catalyst deposition and patterning
The Ni catalyst was deposited by electron-beam evaporation of a thin (5-20)
metal layer. First, the areas for catalyst deposition were defined using standard
photolithography with negative or positive photoresists. After pattern development and
surface cleaning by oxygen plasma (March Plasmod GCM 200, 2 min, 1 sccm of O
2
, 100
W) thin films (5-20 ) of Ni was deposited by electron-beam evaporation. These thin Ni
films undergo dewetting upon heating, thus generating the nanoparticles that serve as
catalyst for the VLS growth of nanowires (Fig. S18-20). We also performed experiments
using Ni, Fe and Au catalysts from colloidal suspensions and salts, obtaining similar
4

results. However, all the results presented in this report were performed with evaporated
Ni.
1.3 Nanowire synthesis
Nanowires were grown in a home-built hot-wall CVD with fast heating capability.
The gallium source was Ga
2
O
3
powder placed inside the oven. The nitrogen source is dry
NH
3
(99.999%, Maxima). H
2,
Ar or N
2
are used as carrier gases (all gases 99.999%,
Gordon Gas). In a typical experiment, a sample is placed on a fused-silica carrier plate
and inserted to a 25 mm diameter quartz tube. The tube is inserted into a split oven and
purged of oxygen by 5 cycles of pumping to 10 mbar and purging with N
2
or Ar at
elevated temperature. Once the tube is purged, dry NH
3
(2-30 sccm) and H
2
gas (30-120
sccm) are streamed into the tube, and pressure is maintained between 50 torr to
atmospheric pressure. Once the desired temperature (930-1000 C) is achieved, the oven
is moved over the Ga
2
O
3
powder, and turned off at the end of the growth time (10 min to
8 hours).
1.4 Microscopic characterization
The grown nanowires were characterized using field-emission SEM (Supra 55VP
FEG LEO Zeiss) at low working voltages of 1-5kV. Atomic force microscopy (AFM,
Veeco, Multimode Nanoscope IV). Images were acquired in air tapping mode using 70
kHz (FESP1) and 300 kHz (RTESP7) silicon tips (Nanoprobes).Thin lamellae for TEM
characterization were made using a FEI Helios DualBeam microscope, and inspected
using a FEI Tecnai F30-UT field emission TEM, equipped with a parallel EELS (Gatan
imaging filter) operating at 300 kV. TEM digital images were recorded using a Gatan
Ultrascan1000 CCD camera. TEM images were analyzed to determine crystallographic
5

orientation and epitaxial relationships using Fourier transform (FFT) from selected areas
in the nanowire cross-sections. Indexing of the FFT peaks was done according to
crystallographic tables for bulk GaN, as shown in Fig. S11. Atomic models were verified
with TEM simulations, using EMS software (33) for nanowires grown on C, A, and M-
plane sapphire.
1.5 Optical characterization
Photoluminescence spectra from single GaN nanowires were acquired with a Jobin-
Yvon LabRam HR 800 micro-Raman system, equipped with a liquid-N
2
-cooled detector.
For the excitation, we used a He-Cd laser at 325 nm. The laser power on the sample was
5 mW. The measurements were taken with a 2400 g mm
-1
grating for the UV (spectral
resolution 4 cm
-1
) and a confocal microscope with a 100 m aperture and a 10 m laser
spot. Care was taken to measure spectra from isolated nanowires. Photoluminescence
spectra were measured up to 730 nm for (i) horizontally grown nanowires on C(0001),
R( 02 1 1 ), M( 0 1 10 ), annealed M( 0 1 10 ) and miscut C(0001) toward ] 0 1 10 [ , (ii)
nanowires vertically grown as forests under the same conditions as (i) and dispersed on
A( 0 2 11 ) sapphire. Spectra from more than 20 single nanowires were measured, yielding
quite similar results. No yellow photoluminescence was observed in any of the
nanowires. Occasionally, a few nanowires, both horizontally and vertically grown,
displayed a small shoulder around 420 nm. No other peaks were observed at longer
wavelengths up to 730 nm.
6

3.6 Electrical characterization
GaN nanowire field effect transistors (FETs) were fabricated by defining source
and drain electrodes using electron-beam lithography (JEOL 6400 SEM equipped with
JC Nabity Nanopattern Generator System), followed by electron-beam evaporation of Ti
(20 nm) and Au (30 nm) (34), and liftoff in acetone. Rapid thermal annealing (500
o
C,
3min) was done to ensure low-resistance ohmic characteristics of the fabricated contacts.
The typical electrodes were of 2 m width, separated by 2-5m from each other. For
gate dielectric, 400 nm thick SiO
2
was deposited by plasma enhanced chemical vapor
deposition (PECVD, Nextral 890). A top-gate electrode was patterned by electron-beam
lithography, electron-beam evaporation (Cr 5 nm/Au 50 nm) and liftoff.
Two-terminal electrical characterizations were done by applying source-drain DC
bias, and measuring I-V curves as a function of gate voltage. To characterize the possible
influence of the contact resistance on our measurements, two additional external
electrodes were added to allow 4 point resistance measurement, giving the value of 183
kO vs. 197 kO obtained by direct two point measurement. The slight difference between
the two values should be attributed to the contact resistance, which was found to be two
orders of magnitude lower than that of the characterized nanowires, and can then be
neglected.
The charge carrier mobility () was extracted from the transconductance g
m
, which
is defined as the slope of I
sd
-V
g
curve in the linear region and is given by equation (S2),
where V
g
is the gate voltage, I
dc
is the mobility, L is the nanowire channel length, and C is
the capacitance (35). The capacitance is calculated using quasi-circular cross-section
approximation given by equation (S3), where h is the dielectric thickness and a is the
7

width of the horizontal GaN nanowire (36). The carrier concentration n
e
was extracted
from the threshold voltage V
th
using equation (S4), where r is the nanowire radius and the
total charge Q =
th
CV .
Overall, about 20 nanowire transistors with channel lengths in the range of 3-5
m were fabricated and characterized, giving the mobility values ranging from 60 cm/V
.
s
to 250 cm/V
.
s, and the carrier density values ranging from 1x10
18
cm
-3
to 2x10
19
cm
-3
.


sd
const V
g
dc
m
V
L
C
dV
dI
g
sd
) (
2
.
= =
=
(S2)


) / 4 ln(
2
0
a h
L
C
tcc
~

(S3)


L r e n CV Q
e th
2
t = =

(S4)

2. Further discussion on the guided growth mechanism
As mentioned in the main text, the mechanism of guided growth is not yet fully
understood. Ref (31) presents a model that could be relevant to our results. However, as a
continuum model, it does not refer to the epitaxial relationship between the nanowire
material and the substrate. To explain and predict the orientation control in our results, a
more detailed model, beyond the scope of this Report, will be required. To this end, a few
important observations need to be taken into account. First, the epitaxial relationships of
most of our nanowires are consistent with those reported in literature for thin films
produced by different methods (22). This suggests a significant degree of thermodynamic
control. On the other hand, the most common facets observed in our nanowires are not
necessarily the most thermodynamically stable ones. For instance, the { 2 2 11 } facets,
which are quite common in our results (Fig. S1 and Table S1), are not part of the
equilibrium Wulff shape of GaN, but are known as fast-growing planes in the kinetic
8

Wulff plot (37). This indicates a significant degree of kinetic control. Moreover, in terms
of symmetry, the unidirectional growth of nanowires with nonpolar orientations on the R
and M planes of sapphire cannot be explained by the thermodynamics of the nanowire-
substrate interaction alone, so it is clear that kinetics and the catalysts play a major role in
the mechanism of guided VLS growth. This also suggests that by changing the growth
conditions, the orientation control could be further improved.
The mechanism of termination of nanowire growth is also yet to be understood. We
may speculate that growth termination can be caused by pinning or poisoning of the
catalysts by defects or impurities on the substrate. Since we pattern the catalyst in stripes
and the nanowires grow onto the clean sapphire areas (Figs. S18-S20), a nanowire can
grow very long before the catalyst gets pinned or poisoned. The guided growth of the
nanowire should exert on the catalyst a large and highly anisotropic force, which can
overcome pinning forces, thus allowing the nanowire to keep growing for very large
distances exceeding 1 mm. The force exerted on the catalyst by graphoepitaxial growth
on a faceted surface is probably more highly anisotropic than that exerted by epitaxial
growth on an atomically flat surface. This could explain why graphoepitaxial nanowires
are generally better aligned than epitaxial nanowires.

9

3. Complete Nanowire Characterization

10

Fig. S1. Growth directions, crystallographic orientations and faceting of horizontally
grown GaN nanowires on different sapphire substrates. The growth directions are
schematically represented in relation with the morphological symmetry of the substrate.
The SEM images show the observed growth direction from patterned catalyst islands.
The cross-sectional TEM images show the most common epitaxial relationship,
crystallographic orientation, faceting and size of the GaN nanowires for each sapphire
plane. The model is based on these observations. Comprehensive data are shown in Table
S1 and Figs. S2-S16.

11

Table S1. Comprehensive crystallographic data for horizontally grown GaN nanowires.
Nanowire axial orientations are highlighted in yellow.
*
These nanowires have the same crystallographic axis orientation as the previous case, but are rotated with respect
to their axis.

These nanowires have all the same crystallographic axis orientation, but are rotated with different rotations with
respect to their axis.

XXXX indicates variable directions or planes.

Some of these nanowires have the same crystallographic axis orientation, but are rotated with different rotations
with respect to their axis.

UUUU Indicates undefined planes.

Axial
GaNAl
2
O
3

Transversal
GaNAl
2
O
3

C (uuu1) 9 (uuu1)(uuu1) |1u1

u]|112

u] |12

1u]|11

uu] (12

12),(uuu1

),(12

12

)
A (112

u) 7 (uuu1)(112

u) |1u1

u]|1u1

u] |12

1u]|uuu1] (12

12),(uuu1

),(12

12

)
R (11

u2)
1u ( 12

16

)(11

u2) |1u1

u]|1

1u1] |12

12]|1

2u] (12

12),(uuu1

),(12

12

)
8
*
(uuu1)(11

u2) |1u1

u]|1

1u1] |12

1u]|1

2u] (12

12),(uuu1

),(12

12

)
2 (S2

u)(11

u2) |14S

S]|1

1u1] |1u1

S]|1

2u] (1

u12),(u1

1S)
N (1u1

u)

1S (11

uS)(1u1

u) |112

u]|uuu1

] |1

1u1]|12

1u] (uuu1

),(1

1u1)
S
*
(11

uu)(1u1

u) |112

u]|uuu1

] |uuu1]|12

1u] (uuu1

),(11

u1)
4 (S2

u)(1u1

u) |14S

S]|uuu1

] |1u1

S]|12

1u] (1

u12),(u1

1S)
S (112

2)(1u1

u) |112

]|uuu1

] |11

uu]|12

1u] (u11

1),(112

2),(1u1

1)
Annealeu N
(1u1

u)

9

(XXXX)

(1u1

u) |14S

S]|12

1u] |XXXX]

|uuu1] (1

u12),(11

u1)
1 (1u1

u)(1u1

u) |uuu1]|12

1u] |12

1u]|uuu1] (u11

u),(1u1

u),(11

uu)
1 (12

12)(1u1

u) |1u1

u]|12

1u] |1

21

S]|uuu1] (uuu1

),(uuu1),(12

12)
Annealeu C
(uuu1) miscut
towaiu |11

uu]
4

(11

u1)(uuu1) |14S

S]|112

u] |1

u12]|11

uu] (1

u12),(11

u1)
1 (uuu1)(uuu1) |112

u]|112

u] |1u1

u]|11

uu] (11

u2

),(uuu1

),(11

u2)
Annealeu C
(uuu1) miscut
towaiu |1

1uu]
4

(11

u1)(uuu1) |14S

S]|112

u] |1

u12]|11

uu] (1

u12),(11

u1)
Annealeu C
(uuu1) miscut
towaiu |112

u]
S (1

1u1)(uuu1) |11

u2]|11

uu] |112

u]|112

u] (S

122),(1

S2

2),(1

1u1)
2 (uuu1) (1

21

4) |224

]|11

uu] |1u1

u] |112

u] (112

2),(S

122)
1 (u11

)(uuu1) |1

2S]|11

uu] |112

u] |S1

1] (1u1

u),(0000)

12

Table S2. Mismatch between GaN and Al
2
O
3
along and across the nanowires
o
a : bulk lattice parameter

The mismatch is calculated using equation (S1), where a
0
is the lattice spacing for the
substrate and the nanowire in the relevant direction. The mismatch is calculated for 1:1
ratio, and also for other ratios, where such a ratio improves the matching. The table
shows no clear trend favoring a lower mismatch along than across the nanowire.
Therefore, we are not able to claim that strain minimization along the nanowire is the
major factor determining the growth direction. It is thus likely that other factors, such as
the kinetics of nanowire growth could also play a role in determining the growth
directions.
) (
) ( ) (
3 2
3 2
O Al a
GaN a O Al a
mismatch
o
o o

=
(S1)



Epitaxy
GaNAl
2
O
3

Longitudinal Transversal
o
a ()
GaN : Al
2
O
3

Ratio

GaN : Al
2
O
3
Mismatch
o
a ()
GaN : Al
2
O
3
Ratio

GaN : Al
2
O
3

Mismatch
(uuu1)(uuu1)
2.762 : 2.380 1 : 1 -0.16 1.595 : 4.123 1 : 1 0.60322
7.975 : 8.246 5 : 2 0.033
(uuu1)(112

u)
2.762 : 4.123 1 : 1 0.33 1.595 : 2.165 1 : 1 0.263
8.286 : 8.246 3 : 2 -0.0048 7.975 : 8.660 5 : 4 0.079
(uuu1)(11

u2)
2.762 : 5.189 1 : 1 0.461 1.595 : 2.380 1 : 1 0.330
5.524 : 5.189 2 : 1 -0.077 4.785 : 4.760 3 : 2 -0.0049
(11

uu)(1u1

u)

1.595 : 2.165 1 : 1 0.263 2.594 : 2.380 1 : 1 -0.0897
6.380 : 6.495 4 : 3 0.018
13


Fig. S2. Cross-sectional TEM of GaN nanowires on C(0001) sapphire. All the nanowires
have the same orientation and the same epitaxial relationship with the substrate, as
summarized in Table S1.
14


Fig. S3. Side-view TEM of a GaN nanowire on C(0001) sapphire. The epitaxial
relationships are in agreement with the cross-sectional TEM images (Fig. S1).
15



Fig. S4. EELS elemental map of GaN on C-plane sapphire (left: Al and Ga, right: N and
O). No interdiffusion between nanowire and substrate is observed.





16


Fig. S5. Cross-sectional TEM of GaN nanowires on A ) 0 2 11 ( sapphire. All the
nanowires have the same orientation and the same epitaxial relation with the substrate, as
summarized in Table S1.
17


Fig. S6. Cross-sectional TEM of GaN nanowires on R ) 02 1 1 ( sapphire. The vast
majority of nanowires grow along their [ 0 1 10 ]
GaN
direction. In about half of these
18

nanowires, the (0001)
GaN
plane is parallel to the substrate, as the nanowire shown in Fig.
2A. The other wires have similar axis orientation, but are rotated by 30 with respect to
their axis, such that their ( 6 1 2 1 )
GaN
plane is parallel to the substrate. The facets are
always ( 12 2 1 ), ( 2 1 2 1 ) anu ( 1 000 ). The nanowire growth axis is usually [ 0 1 10 ]
GaN
,
but sometime it can slightly deviate toward the [ 1 20 2 ] direction. This small deviation is
within the range of sample alignment at the TEM. An additional nanowire, growing along
the [ 3 5 14 ]
GaN
orientation is also shown, but this orientation was only seen twice, on non-
flat regions of the substrate. Orientations and epitaxial relationships are summarized in
Table S1.
19


Fig. S7. Cross-sectional TEM of GaN nanowires on M( 0 1 10 ) sapphire having a
| 0 2 11 ]
GaN
axis, and a ( 03 1 1 )
GaN
( 0 1 10 )
Al20S
interface, as summarized in Table S1.
20



Fig. S8 Cross-sectional TEM of GaN nanowires on M( 0 1 10 ) sapphire, having a
| 0 2 11 ]
GaN
axis, and a ( 00 1 1 )
GaN
( 0 1 10 )
Al20S
interface, as summarized in Table S1.

Fig. S9. Cross-sectional TEM of GaN nanowires on M(101

0) sapphire having a |14S

S].
axis, as summarized in Table S1.
21


Fig. S10. Cross-sectional TEM of GaN nanowires on M(1u1

u) sapphire having a |112

]
axis, as the nanowire shown in Fig 2B, and summarized in Table S1.


Fig. S11. Explanation of the variation in the rotation position of GaN nanowires on M-
plane sapphire, based on symmetry of the substrate material. The nanowire growth on M
plane (1u1

u) sapphire can be divided into 3 families, according to the growth axis: (i)
] 0 2 11 [ (by far, the most common), (ii) ] 3 2 11 [ , and (iii) |14S

S]. However, due to the

u) planes in three ways: (A) The GaN ) 0001 ( planes match the sapphire
) 0 2 11 ( planes. (B) The GaN ) 0001 ( planes match the sapphire ) 10 2 1 ( planes. (C) The
GaN ) 0001 ( planes are aligned with the sapphire ) 110 2 (

planes.

23


Fig. S12. Cross-sectional TEM of GaN nanowires on annealed M(1u1

u) sapphire. Three
such types of nanowires, labeled A, B and C can be identified, as summarized in table S1.
The nanowires with the most common orientation, labeled A, share a common grow axis.
24

However, these nanowires can be rotated in different angles with respect to their growth
axis, so they have different epitaxial relationships with the substrate. This could be due to
the competition between the interfaces of the GaN with the R(11

02) and S(11

01

) planes
of the sapphire simultaneously, where each plane has its own preferred orientation.

25


Fig. S13. Analysis of cross-sectional TEM images and identification of crystallographic
nanowire orientation, exemplified for two different GaN nanowires grown on annealed
M-plane sapphire. (A) TEM image. (B) Fourier transform (FFT) from a selected area in
26

the nanowire cross-section. Indexing of the FFT peaks is done according to
crystallographic tables for bulk GaN. The nanowire axis corresponds to the zone axis,
namely the cross product of the first linearly independent indices, (1

u12) (11

u1) =
|14S

S]. (C) Simulated electron diffraction pattern for the latter crystallographic
orientation, including d-spacings and o = 87.7, for comparison with the FFT (B). Here,
the R(1

u12) planes of GaN nanowire are aligned with the R(11

u2) planes of the

S], but is rotated around its axis with respect to that of the first
nanowire (A-C), so that the S(11

u1) planes of the GaN are roughly aligned with the

u]

direction, which is also an
intersection between R and S planes of GaN, although different from those whose
intersection defines the |14S

S] direction.








27


Fig. S14. Cross-sectional TEM of GaN nanowires on annealed vicinal miscut C(0001)
sapphire 2 toward ] 00 1 1 [ . Two types of such nanowires, labeled A and B can be
identified, as summarized in table S1.


28


Fig. S15. Cross-sectional of GaN nanowires on annealed miscut C(0001) sapphire 2
toward ] 100 1 [ . All wires have the same orientation and the same epitaxial relationships
with the substrate, as summarized in table S1.
29


Fig. S16. Cross-sectional TEM of GaN nanowires on annealed miscut C(0001) sapphire
miscut 2 toward ] 0 2 11 [ . Three types of nanowires, labeled A, B and C, can be
identified, as summarized in table S1.
30



Fig. S17. Demonstration of a thin-film transistor (TFT) built on an array of perfectly
aligned horizontal GaN nanowires. (A) Schematic of the device. (B) SEM image (before
31

deposition of the top gate). (C) I-V characteristics of the device for different gate
voltages. The device comprises ca. 40 nanowires spread across a width of 40 m, with
common source and drain electrodes separated by 5 m, and a top gate over a thin (400
nm) SiO
2
film. The transconductance at V
sd
=1.0 V

is g
m
= 4.5x10
-7

-1
, corresponding to
an electron mobility = 246 cm
2
/Vs per nanowire, in consistency with our single-
nanowire FET measurements.

32



Fig. S18. SEM images showing a catalyst nanoparticle (top) at the end of each epitaxial
GaN nanowire horizontally grown on R-plane sapphire from Ni islands (bottom).

33


Fig. S19. SEM images showing a catalyst nanoparticle (bottom) at the end of each
graphoepitaxial GaN nanowire horizontally grown on annealed M-plane sapphire from Ni
islands (top). The substrate nanogrooves are visible in the bottom images.
34


Fig. S20. SEM images showing catalyst nanoparticles at the ends of many
graphoepitaxial GaN nanowires horizontally grown on annealed M-plane sapphire from
Ni islands (bottom) after a short period of time, prior to removal of the vertically grown
nanowires. Each of the 17 nanowires visible in the bottom image shows a catalyst
nanoparticle at its end. The substrate nanogrooves are visible in the bottom images.

35

4. References

32. A. Ismach, E. Joselevich, Nano Lett. 6, 1706 (2006).
33. P.A. Stadelmann, Ultramicroscopy 21, 131 (1987).
34. J. S. Hwang et al., Appl. Phys. Lett. 85, 1636 (2004).
35. R. Martel, T. Schmidt, H. R. Shea, T. Hertel, P. Avouris, Appl. Phys. Lett. 73,
2447 (1998).
36. Y. Cui, X. Duan, J. Hu, C. M. Lieber, J. Phys.Chem. B 104, 5213 (2000).
37. V. Jindal, F. Shahedipour-Sandvik, J. Appl. Phys. 106, (2009).