Electrodes for carbon nanotube devices by focused electron beam induced

deposition of gold
T. Brintlinger and M. S. Fuhrera兲
Department of Physics and Center for Superconductivity Research, University of Maryland, College Park,
Maryland 20742
J. Melngailis
Department of Electrical and Computer Engineering and Institute for Research in Electronics and Applied
Physics, University of Maryland, College Park, Maryland 20742
I. Utke
Nanomechanics and Nanopatterning Group, EMPA Materials Science and Technology, Feuerwerkerstrasse
39, CH-3602 Thun, Switzerland
T. Bret, A. Perentes, and P. Hoffmann
Advanced Photonics Laboratory, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne,
Switzerland
M. Abourida and P. Doppelt
Centre d’Etude de Chimie Métallurgique (CNRS), 15 rue Georges Urbain, 94407 Vitry sur Seine, France
共Received 2 June 2005; accepted 3 October 2005; published 7 December 2005兲
Individual carbon nanotubes 共CNTs兲 often occur in randomly dispersed two-dimensional as well as
three-dimensional configurations that make device fabrication difficult. Making electrical contact to
such CNTs is of practical interest. To this end, we make contact to individual metallic single-walled
carbon nanotubes 共SWNTs兲 using the focused electron-beam-induced deposition 共FEBID兲 of pure
gold. The SWNTs are grown by chemical vapor deposition on a flat substrate, and the gold leads are
made through FEBID using inorganic metallic precursor gas, chloro共trifluorophosphine兲gold共I兲, or
AuClPF3, in a high vacuum scanning electron microscope. The same scanning electron microscope
is also used to image carbon nanotubes, allowing for simultaneous alignment. We find equivalent
one-dimensional resistivities for the SWNTs of 10– 15 k⍀ / ␮m for both FEBID gold leads and leads
deposited using conventional electron-beam lithography 共EBL兲 and thermal evaporation of gold,
suggesting similarly low contact resistances. We use electrostatic force microscopy to verify
quantitatively similar contact resistances for one nanotube sample, 10 共±6兲 and 54 共±6兲 k⍀, for
FEBID and EBL leads, respectively, with most voltage dropping across the long metallic SWNT.
© 2005 American Vacuum Society. 关DOI: 10.1116/1.2130355兴

I. INTRODUCTION damages CNTs and affects the substrate/CNT/contact inter-

face through momentum transfer and ion implantation.10,11
Carbon nanotube 共CNT兲 devices1–3 typically rely on con- Environmental electron-beam-induced deposition 共EEBD兲 of
ventional electron-beam or optical lithography for their gold–carbon composites to multi-walled nanotubes
manufacture. This requires nominally flat substrates for resist 共MWNTs兲 has also been demonstrated in the making of in
spinning, constraining many CNT devices to two- situ, 3D CNT devices with good contact resistances.12,13
dimensional 共2D兲 templates; etchants can free the CNT-metal
However, EEBD requires high 共⬎0.8 Torr兲 water vapor
devices, but the initial template remains essentially 2D.
pressure, undesirable for pristine CNT devices, and has been
However, interesting new devices based on the mechanical
studied using only large-diameter MWNTs, not the more
or electromechanical properties of nanotubes4–7 or nanotubes
as electrically active scanned-probe elements, will require technologically interesting single-walled carbon nanotubes
new methods of making electrical and mechanical connec- 共SWNTs兲.
tions on three-dimensional 共3D兲 structures. Of similar impor- Here we employ focused electron-beam-induced deposi-
tance, exposure to air, water, and other gases, as well as tion 共FEBID兲 of gold metal lines to make contact to
solvents and resists, affects the substrate/CNT/contact inter- individual metallic single-walled carbon nanotubes SWNTs
face and device performance8 making in situ, pristine con- using an inorganic metallic precursor, chloro共trifluorophos-
tacts desirable to study inherent CNT behavior absent impu- phine兲gold共I兲 or AuClPF3, in high vacuum.14,15 FEBID is a
rity effects. Focused ion beam 共FIB兲 induced deposition9 has direct-writing method of making electrical contacts using a
been used for in situ contacts to CNT devices,10 but FIB scanning electron microscope and a variety of metal-
containing precursor gases. Some current uses include in situ
a兲
Author to whom correspondence should be addressed; electronic mail: fabrication of magnetoelectrical devices16 and nanoscopic
[email protected] four-point thermal measurements.17 We find SWNT/FEBID

3174 J. Vac. Sci. Technol. B 23„6…, Nov/Dec 2005 0734-211X/2005/23„6…/3174/4/$22.00 ©2005 American Vacuum Society 3174
3175 Brintlinger et al.: Electrodes for carbon nanotube devices 3175

gold devices have equivalent one-dimensional 共1D兲 resistiv-

ities to SWNT devices with gold contacts deposited using
conventional electron-beam lithography 共EBL兲. Electrostatic
force microscopy is used to directly measure the contact re-
sistance between FEBID gold deposits and SWNTs, which is
found to be comparable to or less than the contact resistance
of electron-beam lithography 共EBL兲 defined electrodes on
the same SWNT.

II. EXPERIMENT FIG. 1. Carbon nanotube with focused electron beam induced deposition
共FEBID兲 of gold lines on SiO2. Large leads 共⬃1 ␮m wide; labeled “EBL”兲
A. Carbon nanotube growth at left and right of image were pre-deposited using conventional electron-
beam lithography 共EBL兲 and thermal evaporation. 共An array of gold dots
Using a technique following Ref. 18, we first deposit car- defined by EBL can also be seen; it is unrelated to this experiment.兲 Carbon
bon nanotubes using chemical vapor deposition. The starting nanotube 共SWNT兲 runs from top to bottom of image. Focused electron-
beam-induced deposition 共FEBID兲 of ⬃150 nm, solid gold lines 共labeled
substrate is 500 nm of wet-oxidized SiO2 on degenerately “FEBID”兲 was done in a scanning electron microscope to connect the
doped silicon where the silicon acts as a universal back gate SWNT to the EBL leads. Inset shows region of overlap between FEBID and
allowing determination of metallic or semiconducting EBL lead.
SWNTs after device manufacture. We disperse catalyst par-
ticles by dipping chips in a solution of ferric nitrate and
isopropyl alcohol followed by a dip in hexane. This precipi- heated stages and in situ annealing.17,20–22 The precursor flux
tates the ferric nitrate and leaves a random distribution of at the syringe exit was about 1 ⫻ 1017 molecules/ cm2 s as
iron nitrate particles on the surface. The chips are then measured from weight loss, and the backpressure during
heated under argon to 900 ° C in a tube furnace. Following deposition was maintained at 6 ⫻ 10−6 mbar with a cryogenic
Ref. 19, growth gases 共ethylene/methane/hydrogen at cooling system. We then image SWNTs on the insulating
⬃100/ 1300/ 2000 sccm respectively兲 are then turned on for surface. In Fig. 1, the labeled CNT is visible at the top,
10 min after which the gases are turned off, and the chips are bottom, and between the FEBID electrodes. It shows varying
allowed to cool under argon. This chemical vapor deposition contrast due to the tilt of the SEM and the history of the
growth tends to deposit long 共⬃1 – 300 ␮m兲, randomly posi- beam scan; the CNT is charging and discharging the nearby
tioned SWNTs 共diameter: ⬃1 – 3 nm兲.18,19 substrate which causes a local dynamic voltage contrast
effect.23 After finding locations for gold deposition on top of
B. Conventional lithography and deposition of gold SWNTs, a lithography system controls the electron beam
during FEBID. At an accelerating voltage of 25 kV and
Following SWNT growth, we use conventional EBL and 100 pA beam current, the beam diameter is about 80 nm at
lift-off methods to deposit chromium/gold bonding pads and the focus. The beam is scanned at 11.6 nm/ s 共corresponding
leads on top of and near to SWNTs. The pads and leads are to a line dose about 86 ␮C / cm or 5 ⫻ 107 electrons/ nm兲
thermally evaporated from 99.99% pure gold to typical across regions containing SWNTs attached to the surface,
thickness of 50– 70 nm after deposition of a 2 – 3 nm chro- depositing solid gold leads on top of the SWNTs by stripping
mium adhesion layer. The geometry of the pads and leads off all the ligands from the chloro共trifluorophosphine兲gold共I兲
together tends to incorporate 3–20 SWNTs per 1 mm molecule releasing pure gold.14,15 In Fig. 1, these appear as
⫻ 1 mm area that also includes 16 different leads attached to ⬃150-nm-wide and ⬃20-nm-thick lines connecting the
large bonding pads. We then typically image the SWNTs larger EBL leads with a SWNT. The inset is a higher mag-
using both scanning electron microscopy 共SEM; Philips nification view of the region where the FEBID gold attaches
XL30 or Zeiss DSM982兲 to determine SWNT locations, and to the EBL contact. While it appears that it may be discon-
atomic force microscopy 共Digital Instruments Dimension nected, it is still electrically conducting. Additional FEBID-
5000 or JEOL JSPM-4210兲 for SWNT locations and diam- deposited gold lines were also deposited between EBL con-
eters, before FEBID. Because FEBID entails simultaneous tact pads without nanotubes, in order to investigate the
alignment with imaging and deposition happening concur- resistivity of the FEBID gold deposit. After FEBID, the
rently, these steps may be either omitted or completed after samples are removed from the SEM for further imaging and
SWNT device manufacture using FEBID. electrical transport measurements.
C. Focused electron beam induced deposition of gold
III. RESULTS AND DISCUSSION
We continue with the FEBID of gold leads to connect
A. Electrical resistivity of FEBID deposit
SWNTs to the EBL-defined gold leads, shown in Fig. 1.
Following Ref. 14, the sample is inserted into a scanning Figure 1 also allows the resolution of individual grains of
electron microscope dedicated for FEBID 共Cambridge gold within the FEBID deposit, more easily seen in the inset
S100兲. A small syringe-like reservoir of chloro共trifluorophos- of Fig. 1. The ability to resolve individual grains with
phine兲gold共I兲, AuClPF3, is inserted into the SEM chamber. ⬃20– 50 nm separation indicates lack of a carbonaceous ma-
Deposition is done at room temperature, avoiding usage of trix typical with organometallic precursor deposition, which

JVST B - Microelectronics and Nanometer Structures

3176 Brintlinger et al.: Electrodes for carbon nanotube devices 3176

FIG. 3. Scanning probe microscopy of single-walled carbon nanotube

共SWNT兲 device synthesized using both focused electron-beam-induced
deposition 共FEBID兲 and conventional electron-beam lithography with ther-
mal evaporation 共EBL兲 of chromium-gold leads. 共a兲 Topography image
showing FEBID lead on the bottom and EBL lead at top left, heights= 5 and
20 nm, respectively. Visible between the leads is a SWNT, height= 1.4 nm.
共b兲 Electrostatic potential map acquired simultaneously with 共a兲 using elec-
FIG. 2. Energy dispersive x-ray analysis of focused-electron-beam-induced trostatic force microscopy. Brightness corresponds to potential at a given
deposition 共FEBID兲 of gold from chloro共trifluorophosphine兲gold共I兲. Labeled point, with a driving voltage of 400 mV on the EBL lead.
peaks indicate the presence of gold 共M line兲, but little to no fluorine or
chlorine. The presence of phosphorous cannot be ruled out due to overlap
with the M line of gold. Inset is a high magnification scanning electron
micrograph of individual grains of the analyzed deposition material. Such much less than the device length and a constant intrinsic
resolution 共⬍50 nm兲 also suggests pure gold deposits. resistivity in the SWNT, this equivalence in 1D resistivities
indicates a rough equivalence in contact resistance discussed
further in Sec. III C. 共The EBL lead includes a chromium
prevents such resolution due to charging in the insulating adhesion layer in contact with the nanotube; however, other
carbon matrix.15 A separate deposition from that of Fig. 1 is researchers have shown that chromium/gold contacts are
featured in Fig. 2. A high-magnification image showing gold similar to pure gold.24兲 We note that our 1D resistivities cor-
grains formed in this separate deposition is in the inset. Fig- respond to mean free paths ᐉ = 共2Go␳兲−1 ⬇ 0.4– 0.7 ␮m,
ure 2 also shows energy dispersive x-ray analysis of the FE- which is on order, but somewhat lower than estimates in the
BID gold lines with strong gold M-line peaks at 2.20 keV for literature.24–27 The lowest 3D resistivity of electron-beam de-
both FEBID and EBL gold and negligible/nonexistent signa- posited leads reported to date is still 5 – 10⫻ that of conven-
tures for the nongold inorganic precursor elements, fluorine tional gold15 again making the equivalence of FEBID gold
and chlorine, at 0.7 and 2.7 keV, respectively. As a phos- and EBL chromium–gold displayed here on the same SWNT
phorus K peak at 2.05 keV would overlap the large Au band, worth noting.
the presence of some P impurities cannot be excluded, but
would correspond to low concentrations 共⬍10% 兲. This is C. Electrostatic force microscopy
further indication of solid gold FEBID leads contacting the
SWNT. Figure 3 displays both topography and electrostatic force
Two FEBID gold lines connecting EBL pads without microscopy images comparing FEBID contacts on a single
nanotubes were deposited at line doses of 86 ␮C / cm. These SWNT also contacted using EBL gold leads. Figure 3共a兲
lines had resistivities of 43 ␮⍀ cm 共measured by two-point shows a topographical image of a SWNT, diameter
contact兲. We note that the resistivity value obtained at low = 1.4± 0.5 nm, spanning an EBL lead at top left and a FEBID
dose is about three orders of magnitude lower than in Ref. 15 lead at the bottom of the image. Figure 3共b兲 illustrates the
due to an improved precursor supply used in the present simultaneously acquired electrostatic potential map using
experiments, but remains ⬃20⫻ larger than the resistivity of electrostatic force microscopy, a type of alternating current
bulk or thin film gold deposited by sputtering or evaporation. conducting scanning probe technique.26 Briefly, the conduct-
This is attributed to the percolated structure of the FEBID ing scanning cantilever responds to local changes in electro-
leads, see Fig. 2. static potential which are being driven at the cantilever reso-
nance by an ac bias voltage 共400 mV in this case兲. This
response is detected at the photodiode of the scanning probe
B. Nanotube two-terminal electrical transport
microscope. Drops in potential at any given point are propor-
After applying a bias voltage to such SWNT devices, we tional to the resistance at that point; thus, electrostatic force
find one-dimensional resistivities of ␳ ⬇ 10– 15 k⍀ / ␮m for microscopy noninvasively determines contact resistance.
both lithographically defined and FEBID gold leads on the Figure 4 plots the local electrostatic potential along the
same SWNT. Also, the current changes little with variations length of the tube, taken from Fig. 3共b兲. The red triangles and
in 共back兲 gate voltage, indicating metallic SWNTs, with no black squares represent a bias being applied to either the
significant change in behavior for FEBID, FEBID/ FEBID or EBL lead, respectively. The symmetry with re-
lithographic, or purely lithographic leads. These values rep- spect to reversal of the vertical axis in these data 共as ex-
resent measurements of several contact pairs fabricated on pected for an ac bias voltage兲 is a useful check on the lin-
two long carbon nanotubes. Assuming a scattering length earity of the EFM potential measurement. The total

J. Vac. Sci. Technol. B, Vol. 23, No. 6, Nov/Dec 2005

3177 Brintlinger et al.: Electrodes for carbon nanotube devices 3177

IV. SUMMARY
We have shown the conductance of SWNTs contacted
through FEBID gold leads to be equivalent to EBL gold
leads. Although not demonstrated here, this technique allows
for in situ contact to SWNTs on not only 2D substrates, but
in 3D architectures as well. FEBID can be performed with-
out the application of resists or solvents to the SWNT, and it
does not require a separate alignment step before contact is
made. While FEBID is an unlikely candidate for large-scale
production of SWNT devices 共due to line doses that exceed
conventional EBL doses by three or more orders of magni-
tude兲, we feel its equivalence to EBL gold leads for contact-
FIG. 4. Electrostatic potential along length of carbon nanotube in Fig. 3. The ing SWNTs introduces a new tool for the broad and continu-
red triangles and black squares are for biases applied to the FEBID and EBL ing research on the physics and applications of carbon
lead, respectively. The symmetry indicates uniform response to local poten-
nanotube devices.
tial, regardless of bias or ground lead. The drop at the leads indicates
10 共±6兲 and 54 共±6兲 k⍀ contact resistances, for FEBID and EBL leads,
respectively, with most voltage dropping across the long metallic SWNT, ACKNOWLEDGMENTS
total resistance= 430 k⍀.
This research was supported by the Laboratory for Physi-
cal Sciences and the U.S. Army Research Laboratory Power
resistance in this SWNT device was 430 k⍀. The drop in and Energy Electronics Research Program, and the Swiss
potential at each of the leads quantitatively illustrates the National Science Foundation 200021-103333.
contact resistance at that lead, 10 共±6兲 and 54 共±6兲 k⍀, for 1
P. G. Collins and P. Avouris, Sci. Am. 283, 38 共2000兲.
FEBID and EBL leads, respectively, with most voltage drop- 2
R. Martel et al., Appl. Phys. Lett. 73, 2447 共1998兲.
3
ping across the long metallic SWNT. This is further verifica- S. J. Tans et al., Nature 共London兲 393, 49 共1998兲.
4
tion of the two probe measurements already mentioned, J. Cumings and A. Zettl, Science 289, 602 共2000兲.
5
A. M. Fennimore et al., Nature 共London兲 424, 408 共2003兲.
which suggest that the contact resistance between SWNTs 6
Y.-L. Li et al., Science 304, 276 共2004兲.
and FEBID materials is similar to that of thermally evapo- 7
M. Yu et al., Nanotechnology 10, 244 共1999兲.
8
rated EBL material. Of note, the contact resistance of the W. Kim et al., Nano Lett. 3, 193 共2003兲.
9
FEBID/SWNT junction is close to the theoretical minimum J. Melngailis, J. Vac. Sci. Technol. B 5, 469 共1987兲.
10
R. Vajtai et al., Smart Mater. Struct. 11, 691 共2002兲.
for a single SWNT contact, 1 / 共4G0兲 ⬇ 3 k⍀, where G0 11
P. M. Ajayan et al., Phys. Rev. Lett. 81, 1437 共1998兲.
= h / 2e2 is the conductance quantum. Electrostatic force mi- 12
K. Mølhave et al., Nanotechnology 15, 1047 共2004兲.
croscopy on the device in Fig. 1 共not shown兲 showed some- 13
14
D. N. Madsen et al., Nano Lett. 3, 47 共2003兲.
what higher contact resistances 共⬃100 k⍀兲, but again the P. Hoffmann et al., Mater. Res. Soc. Symp. Proc. 624, 171 共2000兲.
15
I. Utke et al., J. Vac. Sci. Technol. B 18, 3168 共2000兲.
FEBID contact resistance was less than that of EBL leads on 16
G. Boero et al., Appl. Phys. Lett. 86, 042503 共2005兲.
the same SWNT. 17
K. Edinger et al., J. Vac. Sci. Technol. B 19, 2856 共2001兲.
18
Also of note, measurements of EEBD on MWNTs showed W. Kim et al., Nano Lett. 2, 703 共2002兲.
19
resistances of 9 – 29 k⍀ and no correlation with length,13 in- S. Huang et al., J. Am. Chem. Soc. 125, 5636 共2003兲.
20
A. J. DeMarco and J. Melngailis J. Vac. Sci. Technol. B 19, 2543 共2001兲.
dicating that the measured resistances were largely contact 21
H. W. P. Koops et al., J. Vac. Sci. Technol. B 14, 4105 共1996兲.
resistance to the 共more conducting兲 MWNT. Such contact 22
G. C. Gazzadi and S. Frabboni, J. Vac. Sci. Technol. B 23, L1 共2005兲.
23
resistances are on the same order as in our experiments, even T. Brintlinger et al., Appl. Phys. Lett. 81, 2454 共2002兲.
24
J.-Y. Park et al., Nano Lett. 4, 517 共2004兲.
though the contact resistance to large-diameter MWNTs may 25
B. Gao et al., Phys. Rev. Lett. 92, 216804 共2004兲.
in principle be much lower due to the larger number of 1D 26
A. Bachtold et al., Phys. Rev. Lett. 84, 6082 共2000兲.
27
subbands available for conduction. A. Javey et al., Phys. Rev. Lett. 92, 106804 共2004兲.

JVST B - Microelectronics and Nanometer Structures