IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 18 (2007) 445203 (5pp) doi:10.1088/0957-4484/18/44/445203

Operation of single-walled carbon

nanotube as a radio-frequency
single-electron transistor
Yong Tang1 , Islamshah Amlani2 , Alexei O Orlov1 ,
Gregory L Snider1 and Patrick J Fay1
1
Department of Electrical Engineering, University of Notre Dame, Notre Dame,
IN 46556, USA
2
Microelectronics and Physical Sciences Laboratory, Motorola Labs, Tempe, AZ 85284, USA

E-mail: [email protected]

Received 8 August 2007, in final form 1 September 2007

Published 9 October 2007
Online at stacks.iop.org/Nano/18/445203
Abstract
We demonstrate the operation of a radio-frequency single-electron transistor
(RF-SET) using single-wall carbon nanotubes (SWNTs). The device is
embedded in a resonant tank circuit and operates in reflection mode at
temperatures as high as 5 K. Both frequency domain and time domain results
are presented. With a gate modulation frequency of 1 MHz, a charge
sensitivity of ∼4.78 × 10−4 e/Hz−1/2 is obtained, which is limited by the
large parasitic capacitance between the gate and the source/drain pads.
Through the use of a top-gated configuration with reduced parasitic
capacitance, both the RF performance and the charge sensitivity can be
significantly improved.

1. Introduction form tunnel barriers with small (1 fF) junction capacitances.
Since the first observation of single-electron transport through
The radio-frequency single-electron transistor (RF-SET) is ropes of carbon nanotubes by Bockrath et al [6] and later in
the most sensitive high-speed electrometer available for sub- individual CNTs by Tans et al [7], CNT-based SETs have
electron charge detection. The operation of this device is shown several advantages over Al-based SETs. In particular,
based on loading a resonant RF tank ( LC ) circuit with an for CNT SETs, the area of the junctions and the volume of
SET; the readout is performed by monitoring the change in the quantum dot participating in Coulomb blockade are orders
the reflected [1] or transmitted [2] microwave power caused by of magnitude smaller than in Al-based SETs, where the area
changes in the charge on the gate electrode instead of directly of the junctions and the size of the quantum dot island are
measuring the change in the current or voltage in the SET. The defined by the resolution of fabrication processing (typically
major advantages of RF-SETs in comparison with conventional electron-beam lithography). This improved scaling results in a
SETs for charge detection are the ability to measure sub- high charging energy E C (typically, E C > 2 meV) and raises
electron charges in a much wider bandwidth, and higher the operation temperature for CNT-based devices to ∼4 K
charge sensitivity than that of conventional SETs due to the or higher [8–10]. Based on the experimental configuration
reduction in 1/ f noise in the detection bandwidth [1]. Since reported in [1], Roschier et al demonstrated the operation of
the first demonstration of an Al/AlOx /Al RF-SET [1], many RF-SETs based on multi-wall carbon nanotubes [11, 12] in a
applications of ultra-sensitive electrometers, such as read- dilution refrigerator with a base temperature of 10 mK.
out of quantum bits [3] and micromechanical displacement Here we demonstrate a single-walled carbon nanotube
detectors [4], have been shown. (SWNT) RF-SET that operates above liquid He temperatures
Carbon nanotube field-effect transistors (CNT FETs) are (5 K) and therefore does not require the use of sophisticated
promising candidates for future electronic systems [5]. At cryogenic equipment such as a dilution refrigerator for
low temperatures, CNT FETs typically exhibit single-electron operation. Moreover, due to the higher operating temperature
tunneling behavior because the metal/carbon nanotube contacts of these devices, the constraints on power dissipation are

0957-4484/07/445203+05$30.00 1 © 2007 IOP Publishing Ltd Printed in the UK

Nanotechnology 18 (2007) 445203 Y Tang et al

much less strict. Consequently, larger RF carrier signal

levels can be used, resulting in improved signal-to-noise ratio
(SNR). With a carrier frequency of 336.737 MHz, a 1 MHz
modulation signal is applied to the SET gate. The response
of the CNT SET is detected in both the frequency and time
domains. A charge sensitivity of δq ≈ 4.8 × 10−4 e Hz−1/2
has been achieved. The level of high-frequency sensitivity
is limited in this device by parasitic capacitance between the
back gate and the source–drain bonding pads. Despite this
limitation, the device demonstrates promising RF response
at a comparatively high working temperature and shows the
potential for high charge sensitivity. These devices have
potential uses as highly sensitive electrometers for the readout
of charge configurations in advanced device architectures
including quantum computing [3, 13, 14], quantum-dot
cellular automata [15, 16], and also for real-time charge
detection [17–19]. The high speed and high operational
temperature of these devices promise to enable unprecedented
performance in these systems without the complications
associated with extremely low-temperature operation.

2. Experimental details Figure 1. (a) Cross-section of the device showing parasitic

capacitance between the back gate and the source/drain bonding
pads. The thickness of SiO2 is approximately 100 nm. (b) Schematic
The SWNT SET was fabricated by the Embedded Systems of the SWNT RF-SET measurement setup. The RF-SET is operated
Research group at Motorola Laboratories; the details of the in reflection mode and both frequency domain and time domain
fabrication process have been published previously [10]. In setups are shown.
brief, Al/Ni bilayer catalyst islands are patterned on an
oxidized silicon substrate using standard lithography and
evaporation techniques followed by the growth of SWNTs in 50 k), indicating that the device is composed of a p-type
a chemical vapor deposition (CVD) chamber. The nanotube semiconducting nanotube.
growth was performed at 800 ◦ C and a pressure of 100 mTorr The charging diagram obtained at a temperature of 5 K,
using methane (60 sccm) and hydrogen (40 sccm) source shown in figure 2(a), demonstrates single-electron transport
gases. Following nanotube growth, source and drain contacts with a charging energy of approximately 2 meV. This charging
are defined using standard photolithograpy and evaporation energy sets the upper limit on the operational temperature, T 
of Ti/Au metallization. A 2 μm separation between the two E C /3kB , to be ∼7 K. We observe quasi-periodic oscillations
lithographically defined contacts defines the effective tube with peaks of different heights, as shown in figure 2(b), which
length. All of the processes used in the fabrication are can result from either multiple quantum dots within the single
compatible with conventional planar fabrication processing CNT [10, 19–26], or possibly from a bundle of nanotubes
techniques. In operation, a highly doped Si substrate is used between the source and drain contacts, resulting in parallel
as a back-gate for the device. A schematic diagram of the quantum dots.
completed device structure is shown in figure 1(a). To use this device as an RF-SET, the total capacitance
The measurement setup is shown in figure 1(b). The RF from the drain to ground, Cd , is resonated with an external
signal from a signal generator is split equally in two: one half chip inductor to form a resonant tank circuit. A transmission
is used to provide power to the local oscillator port of the line couples the incoming RF signal to the tank circuit. At
mixer for homodyne detection, and the other half is attenuated resonance, the tank circuit transforms the parallel-connected
and fed to the RF-SET through a directional coupler and bias nanotube impedance Rd from ∼100 k to a much smaller
tee. The signal reflected by the RF-SET is separated from the impedance value Z T ≈ L/(Cd Rd ). Since the voltage reflection
incident signal by the directional coupler and is amplified using coefficient can be expressed as  = ZZ TT −Z 0
+Z 0
, where Z 0 = 50 
a cryogenic (4.2 K) high electron mobility transistor (HEMT) is the characteristic impedance of the transmission lines, a
pre-amplifier with a gain of 38 dB. The measurements are change of Rd due to charge modulation will cause a noticeable
performed in a He cryostat in the temperature range 4.5–5 K. change in  due to the impedance transformation by the tank
The low-frequency conductance of the device is monitored circuit.
using a lock-in technique, using a transimpedance amplifier For the device considered here, the parasitic capacitance
connected to the SET’s source contact to monitor the device between the drain and source bonding pads and the back
current. To ensure the grounding of the device at high gate, Cdg and Csg , respectively, are treated as parallel-plate
frequencies, the SET source is also shunted to ground through capacitors. Since the source and drain pads have identical
a 2000 pF chip bypass capacitor. At room temperature, geometries, we estimate Cdg = Csg = ε0 ε dS ≈ 9 pF,
the device shows decreasing conductance with increasing where S is the drain bonding pad area, d = 100 nm is the
positive gate bias (with on-resistance of approximately dielectric thickness and ε = 3.9 is the dielectric constant of

2
Nanotechnology 18 (2007) 445203 Y Tang et al

Both of these problems can be eliminated by using a top-

gate configuration, with much reduced coupling between the
source/drain and the gate. In a top-gate device, a larger
resonator quality factor and increased resonant frequency could
be achieved, resulting in higher charge sensitivity and wider
detection bandwidth.

3. Results and discussion

The sensitivity of the SWNT RF-SET was determined using

a high-frequency lock-in technique. The gate voltage is
modulated with a sinusoidal signal Vm cos(ωm t), so
Vbg = Vbg0 + Vm cos(ωm t). (1)
Assuming that the conductance of the device has a linear
response at small modulation amplitudes Vm , both the
conductance and reflection will be modulated at the same
frequency:
dG
G = G0 + × Vm cos(ωm t) (2)
dVbg
d
 = 0 + × Vm cos(ωm t), (3)
dVbg
where G 0 and 0 = (LG 0 − Cd Z 0 )/(LG 0 + Cd Z 0 ) are the
conductance and the reflection coefficient at Vbg0 . Since
the microwave carrier frequency is much larger than the
modulation frequency, the reflected microwave signal can be
written as 
d
Figure 2. (a) Charging diagram of the CNT SET at T = 5 K, Vrefl = Vc  cos(ωc t) = Vc 0 +
showing a charging energy E c ∼ 2 meV. (b) Gate dependence of dVbg
Coulomb blockade oscillations at T = 5 K. Different peak heights 
and periods are likely due to either multiple quantum dots within the × Vm cos(ωm t) cos(ωc t), (4)
single CNT, or possibly from a bundle of nanotubes between the
source and drain contacts, resulting in parallel quantum dots formed where Vc is the carrier amplitude. With a non-zero 0 , the
between the source and drain contacts. The smallest Vm between reflected signal is amplitude modulated by d/dVbg , which
adjacent conductance peaks is estimated to be 80.6 mV. is proportional to the derivative of conductance dG/dVbg or
(This figure is in colour only in the electronic version) the slope of the Coulomb blockade oscillations. Mixing the
reflected signal with the carrier signal, and using a lock-in
amplifier to detect the demodulated signal at the output of
SiO2 . We included an RF choke in series with the gate as the mixer, the dependence of dG/dVbg versus Vbg can be
shown in figure 1(b) to prevent the RF carrier signal applied recovered. Figure 3 shows a comparison of dG/dVbg inferred
to the drain from coupling into the external gate circuit. In from numerical derivative of the low-frequency conductance
this case the capacitance associated with the resonant tank and that extracted from the microwave reflection measurement.
circuit arises from the series interconnection of the drain- Measurement of the derivative of the conductance allows
to-gate and gate-to-source capacitance. Thus the total tank the minimum detectable signal, Vm , to be determined as a
capacitance is Cd = (Csg Cdg )/(Csg + Cdg ) = Cdg /2 ≈ 4.5 pF. function of Vbg . Once the point of maximal charge sensitivity
An external chip inductor of L ∼ 50 nH is used to form is determined, the SET is biased at that point (indicated by
the resonant √tank circuit with designed resonant frequency of the arrow in figure 2(b)). An RF carrier power level of
f 0 = 1/2π LCd = 337.4 MHz. The measured resonant −60 dBm was chosen to maximize the derivative signal, which
corresponds to a 632.5 μVpp source–drain carrier signal.
frequency is f0 = 336.737 MHz, close to the calculated value.
With the SET biased at the operational point shown in
Unfortunately, the high value of pad capacitance (compared
figure 2(b), the smallest gate modulation signal that resulted
to a typical values of 0.3–0.5 pF for implementations with Al-
in a detectable output was Vm = 10.75 μVpp . The
based RF-SETs [1–3, 27]) limits the gate modulation frequency
minimum period of Coulomb blockade oscillations observed
response of the RF-SET, and therefore imposes a limit on in figure 2(b) is Vbg
min
≈ 80.6 mV. If we take this value as the
the smallest resolvable signal at high frequencies. Another oscillation period corresponding to one electron being added
drawback caused by the large drain-to-gate capacitance is the or removed from the dots formed within the CNTs, then the
signal coupling between the gate and the carrier signal path, smallest detectable charge is given by [5]
which leads to the direct propagation of the gate signal into the
Vm e
drain circuit. This results in interference with the modulated δqrms = √ ≈ 4.78 × 10−4 e/Hz−1/2 , (5)
carrier signal if it appears within the detection bandwidth. Vbg
min
B

3
Nanotechnology 18 (2007) 445203 Y Tang et al

Figure 3. Derivative measurements demonstrating high-frequency

operation of an SWNT SET. (a) Back-gate dependence of
conductance. (b) Numerical derivative of dG/dVbg . (c) Detected
dG/dVbg using lock-in detection of the demodulated microwave
reflection signal. The carrier frequency is 336.737 MHz and the gate
modulation frequency is 1 MHz.

where B = 0.078 Hz is the detection bandwidth of the lock-in

Figure 4. (a) Frequency spectrum of reflected signal with 4.52 mVpp
amplifier and e is the electron charge. sine-wave gate modulation. The carrier frequency is 336.737 MHz
The spectrum of the reflected signal shown in figure 4(a) and the gate modulation frequency is 1 MHz. The resolution
was obtained by inspecting the reflected signal directly using bandwidth is 30 Hz. The tall central peak corresponds to the reflected
an RF spectrum analyzer. A 4.52 mVpp modulation signal was carrier signal and the two side peaks correspond to the gate
modulation. The SNR is the ratio between the side-peak height and
applied to the gate, resulting in two sidebands at f c ± f m (insets
the noise floor, approximately 20 dB. (b) Time domain response of
of figure 4(a)). Here the carrier frequency, fc , is 336.737 MHz an SWNT RF-SET with 4.48 mVpp (0.055epp ) sine-gate modulation
and the modulation frequency, fm , is 1 MHz. Similarly, the at 1 MHz (top) and output curve (1648 trace average) (bottom).
rapid response of the RF SET in the time domain can be seen
by examining the amplified output signal of the mixer on a
digital oscilloscope. The average of 1648 individual traces, can be eliminated by using a top-gate device structure with
taken with a 1 MHz sine wave of amplitude 4.48 mVpp applied reduced parasitic gate–drain capacitance. The high working
to the gate, is shown in figure 4(b). The output shows a sine- temperature and the high charge sensitivity of CNT-based
wave response, demonstrating the high-frequency capability of SET devices will make them promising candidates in future
the CNT SET. nanotechnology.

4. Conclusions
Acknowledgments
We have demonstrated an RF-SET using single-walled carbon
nanotube quantum dots. The SET operates at 5 K with We gratefully acknowledge support by the Keck Foundation,
a resonant frequency of 336.737 MHz and a modulation ONR, and NSF grant ECS-0100075.
frequency of 1 MHz. Both frequency domain and time domain
results are presented, and a charge sensitivity of δq ≈ 4.78 ×
10−4 e/Hz−1/2 is estimated. References
The large parasitic capacitance between the back gate and [1] Schoelkopf R J, Wahlgren P, Kozhevnikov A A, Delsing P and
bonding pads due to the device geometry leads to compromised Prober D E 1998 Science 280 1238
charge sensitivity. Furthermore, the direct propagation of [2] Fujisawa T and Hirayama Y 2000 Japan. J. Appl. Phys.
the gate modulation signal into the RF carrier signal path 39 2338
limits the device performance for detection of non-sinusoidal [3] Aassime A, Johansson G, Wendin G, Schoelkopf R J and
Delsing P 2001 Phys. Rev. Lett. 86 3376
modulating signals, as harmonic components of the signal [4] Blencowe M P and Wybourne M N 2000 Appl. Phys. Lett.
can appear within the signal bandwidth. Such limitations 77 3845

4
Nanotechnology 18 (2007) 445203 Y Tang et al

[5] Jing G, Hasan S, Javey A, Bosman G and Lundstrom M 2005 [16] Orlov A O, Kummamuru R K, Ramasubramaniam R, Lent C S,
IEEE Trans. Nanotechnol. 4 715 Bernstein G H and Snider G L 2003 Surf. Sci. 532 1193
[6] Bockrath M, Cobden D H, McEuen P L, Chopra N G, Zettl A, [17] Lu W, Ji Z, Pfeiffer L, West K W and Rimberg A J 2003 Nature
Thess A and Smalley R E 1997 Science 275 1922 423 422
[7] Tans S J, Devoret M H, Dai H, Thess A, Smalley R E, [18] Bylander J, Duty T and Delsing P 2005 Nature 434 361
Geerligs L J and Dekker C 1997 Nature 386 474 [19] Biercuk M J, Reilly D J, Buehler T M, Chan V C, Chow J M,
[8] Nygard J, Cobden D H, Bockrath M, McEuen P L and Clark R G and Marcus C M 2006 Phys. Rev. B 73 201402
Lindelof P E 1999 Appl. Phys. A 69 297 [20] Amlani I, Zhang R, Tresek J and Tsui R K 2003 J. Vac. Sci.
[9] Tans S J, Verschueren A R M and Dekker C 1998 Nature Technol. B 21 2848
393 49 [21] Li H, Zhang Q and Li J 2006 Appl. Phys. Lett. 88 013508
[10] Amlani I, Zhang R, Tresek J and Tsui R K 2004 IEEE Trans. [22] Matsumoto K, Kinoshita S, Gotoh Y, Kurachi K, Kamimura T,
Nanotechnol. 3 202 Maeda M, Sakamoto K, Kuwahara M, Atoda N and
[11] Roschier L, Sillanpaa M, Taihong W, Ahlskog M, Iijima S and Awano Y 2003 Japan. J. Appl. Phys. 42 2415
Hakonen P 2004 J. Low Temp. Phys. 136 465 [23] McEuen P L, Bockrath M, Cobden D H, Yoon Y-G and
[12] Roschier L, Hakonen P, Bladh K, Delsing P, Lehnert K W, Louie S G 1999 Phys. Rev. Lett. 83 5098
Spietz L and Schoelkopf R J 2004 J. Appl. Phys. [24] Postma H C, Teepen T, Yao Z, Grifoni M and Dekker C 2001
95 1274 Science 293 76
[13] Shnirman A and Schon G 1998 Phys. Rev. B 57 15400 [25] Tsuya D, Suzuki M, Aoyagi Y and Ishibashi K 2005 Japan. J.
[14] Fujisawa T, Hayashi T and Sasaki S 2006 Rep. Prog. Phys. Appl. Phys. 44 2596
69 759 [26] Sapmaz S, Jarillo-Herrero P, Kouwenhoven L P and
[15] Orlov A O, Kummamuru R K, Ramasubramaniam R, Toth G, van der Zant S J H 2006 Semicond. Sci. Technol. 21 52
Lent C S, Bernstein G H and Snider G L 2001 Appl. Phys. [27] Bladh K, Gunnarsson D, Aassime A, Taslakov M,
Lett. 78 1625 Schoelkopf R J and Delsing P 2003 Physica E 18 91