prevention of si-contaminated nanocone formation during plasma enhanced cvd growth of carbon...
TRANSCRIPT
Carbon 43 (2005) 835–840
www.elsevier.com/locate/carbon
Prevention of Si-contaminated nanocone formation duringplasma enhanced CVD growth of carbon nanotubes
Dong-Wook Kim a,b,*, L.-H. Chen b, J.F. AuBuchon b, I.-C. Chen b, Soo-Hwan Jeong a,In K. Yoo a, S. Jin b
a Samsung Advanced Institute of Technology, P.O. Box 111, Suwon, Kyongki 440-600, South Koreab University of California at San Diego, La Jolla, CA 92093-0411, USA
Received 15 September 2004; accepted 8 November 2004
Available online 30 December 2004
Abstract
We investigated the growth behavior and morphology of vertically aligned carbon nanotubes (CNTs) on silicon (Si) substrates by
direct current (DC) plasma enhanced chemical vapor deposition (PECVD). We found that plasma etching and precipitation of the Si
substrate material significantly modified the morphology and chemistry of the synthesized CNTs, often resulting in the formation of
tapered-diameter nanocones containing Si. Either low bias voltage (�500 V) or deposition of a protective layer (tungsten or titaniumfilm with 10–200 nm thickness) on the Si surface suppressed the unwanted Si etching during growth and enabled us to obtain cylin-
drical CNTs with minimal Si-related defects. We also demonstrated that a gate electrode, surrounding a CNT in a traditional field
emitter structure, could be utilized as a protection layer to allow growth of a CNT with desirable high aspect ratio by preventing the
nanocone formation.
� 2004 Elsevier Ltd. All rights reserved.
Keywords: A. Carbon nanotubes; B. Chemical vapor deposition; C. Scanning electron microscopy, Transmission electron microscopy
1. Introduction
Carbon nanotubes (CNTs) have been shown to be
useful for a variety of applications such as field emission
displays (FEDs) [1–3], nanoscale electromechanical actu-
ators [4], field-effect transistors (FETs) [5,6], nanointer-
connects [7], and scanning probe microscope (SPM)tips [8]. It is advantageous to grow CNTs directly on sil-
icon (Si) to realize future large-scale integrated nano-
electronic devices. Among several growth techniques,
plasma enhanced chemical vapor deposition (PECVD)
has attracted great attention owing to its ability to
grow vertically aligned CNTs at relatively low tempera-
0008-6223/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.carbon.2004.11.018
* Corresponding author. Tel.: +82 31 280 9398; fax: +82 31 280
9349.
E-mail address: [email protected] (D.-W. Kim).
tures, which makes the synthesis more compatible with
conventional semiconductor fabrication processes [9–
16].
The low temperature growth and complicated plas-
ma-related behaviors may cause a large concentration
of defects and some difficulty in the control of the pro-
cess [11,12]. Under certain growth conditions, a CNTno longer maintains the intended cylindrical shape and
changes into a defective, cone-like structure, which often
exhibits larger diameters and much Si contamination
[11,12]. The high aspect ratio and the excellent electrical
conductance properties, inherent in the cylindrical tube
geometry (as compared with the cone-geometry often
associated with a large-diameter base), are essential
characteristics for efficient operation of CNT field emit-ter devices and other electrical devices, such as FETs
and interconnects. Therefore, the ability to control the
836 D.-W. Kim et al. / Carbon 43 (2005) 835–840
morphology and reduce the defect concentration of
synthesized CNTs is important for ensuring satisfactory
device performance.
In this paper, we report how plasma etching of an Si
substrate influences the morphology and composition of
CNTs during a PECVD growth process. We reveal thatappropriate bias voltage or selective deposition of a pro-
tective layer prevents the unwanted Si etching and the
subsequent formation of Si-contaminated nanocone
structures.
Fig. 1. SEM images and schematic cross-sectional diagrams of (a)nanocones grown on a bare Si substrate and (b) nanotubes grown on a
10 nm thick Ti-coated Si substrate.
2. ExperimentalA direct current (DC) PECVD system was used to
grow vertically aligned CNTs on an Si substrate at a
temperature of about 700 �C [15,16]. A mixed gas of
acetylene (C2H2) and ammonia (NH3) at a ratio of 1:5
was used for the process. The gas pressure was held at
3 Torr during the growth. A DC bias voltage of 500–
600 V was applied between the anode and the cathode,
which maintained the plasma during the growth.For catalyst fabrication, a 10 nm thick nickel (Ni)
catalyst layer was sputter deposited onto an n-type
Si(100) substrate. Catalyst islands of �200 nm diameter
were fabricated on the Si substrates using resist (poly-
methyl methacrylate, PMMA) coating and e-beam
lithography followed by a lift-off process. We chose
the 200 nm diameter catalyst size so as to nucleate and
grow only a single CNT from each catalyst island [9–14].
3. Results and discussion
It has been realized that CNTs grown on bare Si
often result in a nanocone morphology while a buffer
layer-coated Si produces more cylindrical CNTs. Here-
after, we will call the cylindrical CNT �nanotubes� com-pared with the cone shaped �nanocones�. 1 Fig. 1(a) and(b) shows scanning electron microscope (SEM) images
and schematic cross-sectional diagrams of CNTs grown
at the bias voltage of 550 V on a bare Si substrates and a
10 nm thick titanium (Ti) coated Si substrate, respec-
tively. A Ni catalyst layer was patterned into islands
of 200 nm diameter and 2 lm spacing. Although the
deposition conditions were nearly identical for the twosamples, their geometry showed a dramatic difference.
1 In general, the CNTs synthesized by DC PECVD processing are
of �nanofiber� type multiwall nanotubes with the graphene layer wallspositioned at some angles rather than having the walls parallel to the
long axis. The nanotubes grown by microwave plasma CVD or
thermal CVD tend to have the carbon walls parallel to the long axis.
The carbon nanostructures with inclined walls are still denoted as
�CNT� in this paper in a broad sense, as many of them still have hollow
cores in the middle.
The CNTs grown on a bare Si substrate are cone-like,
but those on a Ti layer exhibit a more cylindrical mor-
phology. The nanocones were significantly contami-
nated with Si while the cylindrical CNTs were not, as
will be discussed later.
Fig. 2 shows an SEM image and schematic cross-sec-tional diagram of a CNT array grown on e-beam pat-
terned Ni-on-Ti catalyst islands. These islands were
prepared by e-beam patterning and subsequent sputter
deposition of a 10 nm thick Ti film on an Si substrate
followed by 10 nm thick Ni sputter deposition. Fig. 2
shows that the CNTs grown on patterned Ti islands
are cone-like instead of cylindrical, like the CNTs shown
in Fig. 1(b). From the results of Fig. 1, it could bethought that the Ti layer serves as a diffusion barrier
that prevents the diffusion of Si from the substrate to
the growing CNT during the high temperature growth
process at �700 �C, and as a result, the nanotubes hav-ing the desirable cylindrical morphology are formed in
the presence of the Ti buffer layer. The Ti islands in
Fig. 2. SEM image and schematic cross-sectional diagram of carbon
nanocones grown on Ni/Ti islands.
D.-W. Kim et al. / Carbon 43 (2005) 835–840 837
Fig. 2, which are originally directly underneath the Ni
catalyst islands, and separate the CNT base from the
Si substrate, should have prevented or minimized possi-
ble diffusion reactions between the CNTs and the Si sub-
strate [11]. However, the SEM images in Fig. 2 clearly
show that the nanocone formation still occurs implyingthat another factor, not the Si diffusion, should be influ-
encing the CNT morphology.
We also noted that morphology of CNTs could be
significantly altered by adjusting the applied bias voltage
during the PECVD process. Fig. 3(a) shows an example
of carbon nanocones grown on an Si substrate at an ap-
plied bias voltage of 600 V. In this case, the catalyst
layer was deposited as a uniform thin film (which breaksinto islands on heating to the CVD temperature), rather
than e-beam patterned islands. The base diameter of the
nanocones is as large as �400 nm, and the cone angle is�15–20�. In contrast, nanotubes are formed on the samebare Si substrate by PECVD at a lower bias voltage of
500 V, as shown in Fig. 3(b). The diameter of the nano-
Fig. 3. Effect of bias voltage on CNT morphology grown by DC
PECVD at 700 �C. (a) Nanocones obtained at 600 V and (b)
nanotubes at 500 V.
tubes does not significantly change from the bottom to
top.
To further investigate the growth mechanism of the
CNTs, a high-resolution transmission electron micro-
scope (TEM) was used to study the microstructure and
composition of our CNTs. Fig. 4(a) and (b) shows typi-cal TEM images of carbon nanocones and nanotubes,
respectively, with the two samples prepared under differ-
ent bias voltages. The nanocone has an inner cone,
which is covered by a thinner outer layer. The cone is
crystalline with internal inclusions. Energy dispersive
spectroscopy (EDS) analysis indicates that the cone ma-
trix is mostly silicon and carbon, with the Si content esti-
mated to be at least 40 at.%, and the tiny inclusions aremainly composed of Ni. Si is found in the cone every-
where, which indicates that the precipitation of Si into
the nanocone occurs throughout the duration of growth.
The nanotubes shown in Fig. 4(b) have an almost con-
stant diameter and Ni catalyst particles cap on their
tops. The EDS analysis on nanotube samples in Fig.
4(b) shows that no detectable amount of Si is found
for most parts of the sample. A trace of Si is only foundat the very bottom part of the nanotube implying that a
slight diffusion of Si into the base may have occurred.
Elemental composition distributions of the nano-
cones and nanotubes grown on e-beam patterned cata-
lyst islands, like the samples shown in Figs. 1 and 2,
were also acquired. Since these nanocones were grown
under a moderate bias voltage of 550 V and on pat-
terned Ti buffer islands, the Si concentration is less than20 at.% in the body. However, it is noticeable that a
Fig. 4. TEM micrographs for (a) nanocone, (b) nanotube.
Fig. 5. SEM image and schematic diagram of a CNT array grown on
an Si substrate with a patterned Ti layer. A 10 nm thick Ti film covers
most of Si surface and the only exposed Si area is a 60 · 60 lm2 square
minus a 20 · 20 lm2 Ti square at the center.
838 D.-W. Kim et al. / Carbon 43 (2005) 835–840
finite amount of Si can be found even at the very top of
the nanocone by the EDS analysis. The cylindrical
nanotubes grown on a uniform Ti layer do not have
detectable amounts of Si, except for at the bottoms
where it is presumably caused by some solid-state diffu-
sion of Si from the substrate. All these results imply thatthe CNT morphology is related to its Si defect concen-
tration. They also indicate that while some diffusion of
Si from the substrate may contribute to the CNT con-
tamination, the main cause for the contamination is
the precipitation of plasma etched Si onto the CNTs.
The CNTs are grown in both the vertical and lateral
directions: the vertical growth is related to the diffusion
of carbon (C) through the Ni catalyst particles, and thelateral growth results from the precipitation of both C
and Si. The DC discharge may induce etching of sub-
strate materials and some portion of the grown CNTs
[10,11]. A proper gas ratio of C2H2:NH3 and bias volt-
age can equilibrate growth of graphene sheets and etch-
ing of amorphous carbon, resulting in growth of
cylindrically shaped carbon nanotubes along the applied
electric field direction [11]. Also, it should be noted thatthe catalyst particles suffer from sputtering and frag-
mentation during growth [17]. Prolonged growth may
result in loss of the catalyst particle and termination of
the vertical growth; however, the lateral growth can con-
tinue if the C or Si supply is maintained. In our growth
conditions, a rather low C2H2:NH3 ratio of 1:5 was uti-
lized, which may allow a somewhat fast etching rate of
an Si substrate [12]. Moreover, either exposure of a bareSi surface or higher bias voltage is likely to enhance Si
etching and incorporation of Si atoms into the growing
CNTs in the plasma. The availability of such etched Si
in the plasma can promote the lateral growth, resulting
in nanocone formation as shown in Figs. 1–3. Optimiz-
ing the bias voltage or adding a protective metal thin
film on the Si surface could suppress the Si etching
and enable the growth of the nanotubes with minimaldefects. Regarding the protective layers, we tested sev-
eral kinds of thin films such as Ti, Ti nitride (TiN), tung-
sten (W), and silicon dioxide (SiO2). Most of these metal
thin films worked well as a protective layer that pre-
vented Si etching, except for SiO2. It seems that acti-
vated ions in the plasma are able to etch the SiO2 film
as well as the Si substrate during the DC discharge
[18]. The structural and chemical analyses of our carbonnanocones and nanotubes support the above interpre-
tations.
Fig. 5 shows an SEM image and a schematic diagram
of a CNT array grown on an Si substrate with a pat-
terned Ti layer. As illustrated, a 10 nm thick Ti film cov-
ers most of the Si surface, and the only exposed Si
surface is a picture frame shaped area with a
60 · 60 lm2 sized square minus a 20 · 20 lm2 sized Tisquare at the center. As shown in Fig. 5, cylindrical
nanotubes can be grown even on the bare Si surface with
the aid of the surrounding Ti protective layer. This SEM
image clearly confirms that etching and subsequent Siincorporation, rather than Si diffusion, dominantly
modifies the CNT morphology. This kind of surround-
ing protection layer might slightly increase the Si defect
concentration compared with the uniform protection
layer case, such as shown in Fig. 1(b). Physical proper-
ties of an underlying layer and a resulting interfacial
layer may be critical for device performance. A direct
growth of CNTs on an Si substrate without any metallicinterfacial layer is desirable in some cases [13]. For such
a case, surrounding area protection can suffice to allow
the direct growth of CNTs on Si, as the very limited Si
surface exposed to the plasma will minimize Si etching
and incorporation into the CNTs, thus allowing an opti-
mization of their electrical and structural properties
simultaneously.
Fig. 6(a) and (b) shows exemplary SEM images andcross-sectional diagrams of a gated CNT field emitter di-
rectly grown on Si substrates. The insulator and gate
materials were 1.5 lm thick SiO2 and 0.2 lm thick W,
respectively. We used a self-aligned fabrication process,
similar to that of Prior et al., except for the gate material
[9]. Thin films of W and SiO2 were etched by SF6-based
plasma and buffered oxide etch solution (6 parts 40%
NH4F and 1 part 49% HF), respectively. During theCNT PECVD, the W gate electrode covers most of the
SiO2 surface and protects it from etching. As a result,
we can grow cylindrical CNTs directly on Si substrates.
In this example, we find that a properly chosen gate
metal layer can be used simultaneously for the purpose
of Si surface protection during the PECVD process
and also for the device operation after the fabrication.
A CNT field emitter is a superior electron sourcecompared with conventional Schottky or field emission
Fig. 7. (a) SEM image of a single CNT grown at the center of SiO2hole and (b) a schematic cross-sectional diagram of an example vertical
transistor using a CNT.
Fig. 6. (a) SEM image and (b) cross-sectional diagram of a gated CNT
field emitter directly grown on Si substrates.
D.-W. Kim et al. / Carbon 43 (2005) 835–840 839
sources since it has high brightness electron beams
(109 A m�2 sr�1 V�1 s), a small energy spread (0.2–0.3 eV), and long-term stability [1]. However, poor emis-
sion uniformity due to the extreme sensitivity of the
Fowler–Nordheim tunneling process to small variations
in nanotube radii and heights, gate diameters, position
of the gate holes, and the work function, just like vari-
ous other field emission sources, limits commercial
device application of the CNT field emitter. An intro-
duction of a resistive layer between the field emitterand the emitter lines helps to mitigate the uniformity
problem [19]. A lateral resistor mesh is used to homog-
enize the emission current and prevent a short-circuit
effect by limiting the electrical current. In general,
doped-Si, e.g. having a resistivity in the level of
�105 X cm and a thickness of �0.1 lm, can be used asthe resistive layer and is compatible with the manufac-
turing process of the emitters. We expect that such astructure would show a better emission uniformity com-
pared with field emitter arrays grown on metal elec-
trodes (e.g., TiW/Mo/TiW) [9]. Moreover, the high
aspect ratio of the cylindrical CNTs is likely to show
lower turn-on voltages than carbon nanocones, which
generally exhibit lower aspect ratios [12].
Desirable single CNTs with minimal defects, not
nanocones contaminated with Si, can also be grown in-side a small aperture hole in SiO2, as shown in Fig. 7(a).
As mentioned earlier, SiO2 is also susceptible to etching
during the PECVD process, thus it can supply Si ele-
ment to the plasma, resulting in carbon nanocone for-
mation. Small holes, such as a gate cavity, might
significantly affect the nanotube growth process due to
the possible impediment of gas flow to the structure or
local variation of electric field strength due to the pres-
ence of the gate structure. In spite of such concerns, a
�single� and �cylindrical� CNT was successfully grown
at the center of 600 nm gate hole. This structure can also
be useful for vertical electronic components such as
diodes/transistors and nanointerconnects [6,12,13]. Fig.
7(b) shows a schematic cross-sectional diagram of anexemplary vertical transistor [6]. For such electrical
applications, minimal chemical and structural defects
(such as the formation of Si-contaminated nanocones)
are crucial for achieving low electrical resistance and
desirable field effect switching.
4. Conclusions
We investigated the plasma etching effects on the
morphology and chemistry of carbon nanotubes (CNTs)
grown by DC plasma CVD. The formation of Si-con-
taminated nanocone-type defective structure was pre-
vented by suppressing the etching of the Si substrate
material during the plasma processing by using either
a low bias voltage (�500 V) or deposition of a protectivemetal layer on the Si surface such as Ti and W, which
provided strong etching resistance to plasma. Proper
choice of materials also allowed us to use a device ele-
ment itself, such as a gate metal electrode in a field emit-
ter array as the protective layer during the PECVD
growth. Control of the CNT morphology demonstrated
840 D.-W. Kim et al. / Carbon 43 (2005) 835–840
by our Si-fab-compatible techniques can be very useful
for improving the device performance.
Acknowledgments
We acknowledge the support of the work by Univer-
sity of California Discovery Fund under Grant No.
ele02-10133/Jin, NSF NIRTs under Grant No. DMI-
0210559 and DMI-0303790. The authors would like to
thank Dr. Alexander Tikhonovsky for assistance in
TEM studies and helpful discussions.
References
[1] Jonge ND, Lamy Y, Schoots K, Oosterkamp TH. High brightness
electron beam from a multi-walled carbon nanotube. Nature 2002;
420:393–5.
[2] Zhu W, Bower C, Zhou O, Kochanski G, Jin S. Large current
density from carbon nanotube field emitters. Appl Phys Lett
1999;75:873–5.
[3] Jeong SH, Hwang HY, Lee KH, Jeong Y. Template-based carbon
nanotubes and their application to a field emitter. Appl Phys Lett
2001;78:2052–4.
[4] Kim P, Lieber CM. Nanotube nanotweezers. Science 1999;286:
2148–50.
[5] Tans S, Verschueren A, Dekker C. Room-temperature transistor
based on a single carbon nanotube. Nature 1998;393:49–52.
[6] Choi WB, Bae E, Kang D, Chae S, Cheong B, Ko J, et al.
Aligned carbon nanotubes for nanoelectronics. Nanotechnology
2004;15:S512–6.
[7] Wei BQ, Vajtai R, Ajayan PM. Reliability and current carrying
capacity of carbon nanotubes. Appl Phys Lett 2001;79:1172–4.
[8] Snow E, Campbell P, Novak J. Single-wall carbon nanotube
atomic force microscope probes. Appl Phys Lett 2002;80:2002–4.
[9] Pirio G, Legagneux P, Pribat D, Teo KBK, Chhowalla M,
Amaratunga GAJ, et al. Fabrication and electrical characteristics
of carbon nanotube field emission microcathodes with an
integrated gate electrode. Nanotechnology 2002;13:1–4.
[10] Wen JG, Huang ZP, Wang DZ, Chen JH, Yang SX, Ren ZF,
et al. Growth and characterization of aligned carbon nanotubes
from patterned nickel nanodots and uniform thin films. J Mater
Res 2001;16:3246–53.
[11] Chhowalla M, Teo KBK, Ducati C, Rupesinghe NL, Amaratunga
GAJ, Ferrari AC, et al. Growth process conditions of vertically
aligned carbon nanotubes using plasma enhanced chemical vapor
deposition. J Appl Phys 2001;90:5308–17.
[12] Melechko AV, McKnight TE, Hensley DK, Guillorn MA,
Borisevich AY, Merkulov VI, et al. Large-scale synthesis of
arrays of high-aspect-ratio rigid vertically aligned carbon nano-
fibers. Nanotechnology 2003;14:1029–35.
[13] Yang X, Guillorn MA, Austin D, Melechko AV, Cui H, Meyer
III HM, et al. Fabrication and characterization of carbon
nanofiber-based vertically integrated Schottky barrier junction
diodes. Nano Lett 2003;3:1751–5.
[14] Li J, Ye Q, Cassell A, Stevens GNR, Han J, Meyyappan M.
Bottom-up approach for carbon nanotube interconnects. Appl
Phys Lett 2003;82:2491–3.
[15] AuBuchon JF, Chen L, Gapin AI, Kim DW, Daraio C, Jin S.
Multiple sharp bendings of carbon nanotubes during growth to
produce zigzag morphology. Nano Lett 2004;4:1781–4.
[16] Chen L, AuBuchon JF, Gapin AI, Daraio C, Bandaru P, Jin S,
et al. Control of carbon nanotube morphology by change of
applied bias field during growth. Appl Phys Lett 2004;85:
5373–5.
[17] Han J, Yoo J, Park CY, Kim H, Park GS, Yang M, et al. Tip
growth model of carbon tubules grown on the glass substrate by
plasma enhanced chemical vapor deposition. J Appl Phys 2002;91:
483–6.
[18] Taschner C, Pacal F, Leonhardt A, Spatenka P, Bartsch K, Graff
A, et al. Synthesis of aligned carbon nanotubes by DC plasma-
enhanced hot filament CVD. Surf Coat Tech 2003;174–175:81–7.
[19] Meyer R. Electron source with microtip emissive cathodes. US
Patent 5194780; 1993.