A tunable carbon nanotube electromechanical oscillator
Vera Sazonova*, Yuval Yaish*, Hande Üstünel, David Roundy, Tomás A. Arias & Paul L. McEuen
Laboratory of Atomic and Solid-State Physics, Cornell University, Ithaca, New York 14853, USA
*These authors contributed equally to this work
Nanoelectromechanical systems (NEMs) hold promise for a number of scientific and
technological applications. In particular, NEMs oscillators have been proposed for
use in ultrasensitive mass detection1,2, radio-frequency signal processing3,4, and as a
model system for exploring quantum phenomena in macroscopic systems5,6. Perhaps
the ultimate material for these applications is a carbon nanotube. They are the
stiffest material known, have low density, ultrasmall cross-sections and can be
defect-free. Equally important, a nanotube can act as a transistor7 and thus may be
able to sense its own motion. In spite of this great promise, a room-temperature,
self-detecting nanotube oscillator has not been realized, although some progress has
been made8–12. Here we report the electrical actuation and detection of the guitarstring-like oscillation modes of doubly clamped nanotube oscillators. We show that
the resonance frequency can be widely tuned and that the devices can be used to
transduce very small forces.
Figure 1a shows a diagram of the measurement geometry and a scanning electron
microscope (SEM) image of a device. The fabrication steps have been described
elsewhere13; briefly, nanotubes (typically single- or few-walled, 1–4 nm in diameter and
grown by chemical vapour deposition14) are suspended over a trench (typically 1.2–
1.5 µm wide, 500 nm deep) between two metal (Au/Cr) electrodes. A small section of the
tube resides on the oxide on both sides of the trench; the adhesion of the nanotube to the
oxide15 provides clamping at the suspension points.
The measurement is done in a vacuum chamber at pressures below 10−4 torr. We
actuate and detect the nanotube motion using the electrostatic interaction with the gate
electrode underneath the tube. A gate voltage, Vg, induces an additional charge on the
nanotube given by q=CgVg, where Cg is the capacitance between the gate and the tube.
The attraction between the charge q and its opposite charge −q on the gate causes an
electrostatic force downward on the NT. If Cg′ = dCg / dz is the derivative of the gate
capacitance with respect to the distance between the tube and the gate, the total
electrostatic force on the tube is given by:
Page 1 of 12
Fel =
1
2
C g′Vg2
≅
1
2
C g′VgDC (VgDC + 2δ Vg )
(1)
where we have assumed that the gate voltage has both a static (DC) component and a
small time-varying (AC) component. The DC voltage VgDC at the gate produces a static
force on the nanotube that can be used to control its tension. The AC voltage δVg
produces a periodic electric force, which sets the nanotube into motion. As the driving
frequency ω approaches the resonance frequency ωo of the tube, the displacement
becomes large.
To detect the vibrational motion of the nanotube, we employ the transistor
properties of semiconducting16 and small-bandgap semiconducting carbon nanotubes17,18,
that is, that the conductance change is proportional to the change in the induced charge q
on the tube.
δ q = δ (CgVg ) = Cgδ Vg + Vgδ Cg
(2)
The first term is the standard transistor gating effect—the modulation of conductance due
to the modulation of the gate at the driving frequency—and it is observed at any driving
frequency. The second term is non-zero only if the tube moves (when the driving
frequency approaches the resonance); the distance to the gate changes, resulting in a
variation δCg in its capacitance.
To detect this conductance change we use the nanotube as a mixer19 (Fig. 1b).
This method helps avoid unnecessary complications due to capacitive currents between
the gate and the drain electrodes. The magnitude of the current is given by the product of
the AC voltage on the source electrode δVsd, and the modulated nanotube conductance
δG. Using equation (2) we derive the result that the expected current is:
δ I lock −in = δ Gδ Vsd =
1
dG
2 2 dVg
⎛
δ Cg
DC
⎜⎜ δ Vg + Vg
Cg
⎝
⎞
⎟⎟ δ Vsd
⎠
(3)
where δVg is the AC voltage applied to the gate electrode.
Figure 2a shows the measured current as a function of driving frequency at room
temperature. We see a distinctive feature in the current on top of a slowly changing
background. We attribute this feature to the resonant motion of the nanotube, modulating
the capacitance, while the background is due to modulating gate voltage. The response
fits well to a lorentzian function with a normalized linewidth Q−1=∆f/fo=1/80, a resonant
frequency fo=55 MHz, and an appropriate phase difference between the actuation voltage
and the force on the nanotube19.
Page 2 of 12
The DC voltage on the gate can be used to tune the tension in the nanotube and
therefore the oscillation frequency. Figure 2b and c show colour-scale plots of the
measured response as a function of the driving frequency and the static gate voltage. The
resonant frequency shifts upward as the magnitude of the DC gate voltage is increased.
Several distinct resonances are observed, corresponding to different vibrational modes of
the nanotube. We have found similar results in 11 comparable devices, with resonance
frequencies varying from 3 to 200 MHz for different samples and gate voltages.
To understand the frequency dependence of the nanotube oscillations with VgDC,
we have performed a series of simulations of the vibrational properties of nanotubes. We
model the nanotube as a slack beam suspended over a trench. Slack here means that the
tube is longer than the distance between the contacts, which is a result of the nanotube’s
curvature before suspension. Slack was observed for almost all imaged devices in an
SEM (Fig. 1a), and has also been inferred from atomic force microscope (AFM) force
measurements of similar samples13. A finite element model is then used to calculate the
vibrational frequencies for a nanotube with a typical geometry (length L=1.75 µm, radius
r=1 nm) and mechanical rigidities determined using the Tersoff–Brenner potential20.
The theoretical results for a representative device can be seen in Fig. 2d. For no
static electric force on the nanotube, VgDC≈0, the resonance frequency is determined by
the bending rigidity of the nanotube and is approximately that of an equivalent doubly
clamped beam with no tension. At small VgDC, there is a static electric force downward on
the nanotube (equation (1)), producing a tension T ∝ Vg2 which shifts the resonant
frequency of the nanotube, ∆ω0∝T∝Vg2 (ref. 21). At intermediate VgDC, the electrostatic
force overcomes the bending rigidity and the nanotube behaves as a hanging chain; the
profile of the tube forms a catenary. In this regime, the resonance frequency is given by
ω0∝√T∝Vg. In the large electrostatic force regime, the nanotube behaves as an elastic
string—the extensional rigidity becomes dominant. In this regime, the resonance
frequency is given by ω0∝√T∝Vg2/3 (ref. 21). The transition point from the bendingdominated regime to the catenary, and from the catenary to the stretching-dominated
regime, depends on the amount of slack in the nanotube. Either one, two or all three
described regimes may be relevant for a particular device. For the bending-dominated
and catenary regimes, the voltage dependence of the resonant frequencies scales as the
slack to the ¼ th power (Fig. 2d).
Comparing these predictions with Fig. 2b and c, we see a good qualitative
agreement with predicted dispersions. All of the resonances start dispersing parabolically,
Page 3 of 12
some continuing into a linear regime as the gate voltage is increased. For the lowest
resonance shown in Fig. 2c we can also observe the ω0∝Vg2/3 frequency dependence at
large gate voltages. The frequency dependence of the resonances are thus in good
qualitative agreement with theoretical expectations. We do, however, often find multiple
resonances lower in frequency than is predicted by the theoretical calculations. One lowfrequency mode is expected because any small asymmetric clamping will result in a nonzero frequency at zero gate voltage for the lowest branch in Fig. 2d. The additional lowfrequency modes could be caused by extra mass due to contaminants coating the
nanotube, or by a large asymmetry in the clamping conditions. Further studies are needed
to understand the exact nature of this frequency lowering.
To determine the other parameters of the nanotube oscillator, we have studied the
dependence of the measured resonance on the amplitude of the gate drive signal δVg.
Figure 3a shows results for one device. For low driving amplitudes, the response on
resonance is linear in δVg and Q is roughly constant. As the δVg is increased further, the
response saturates and the quality factor decreases. For some devices, there is also a
dramatic change in the signal shape observed at these high driving voltages (Fig. 3b).
Instead of a smooth lorentzian dip, the system develops a hysteretic transition between
low- and high-amplitude states of oscillation.
To understand these results, we first address the linear response regime. We
estimate the amplitude δz of the nanotube oscillation using the measured signal amplitude
relative to the background in conjunction with equation (3). From this, we can extract the
relative change in the capacitance δCg/Cg on resonance; for the data in Fig. 2a, where
δVg=7 mV, we obtain δCg/Cg=0.3% . Assuming a logarithmic model for capacitance
4πε 0 L
, where L is the suspended length of the NT, and z is the distance to the
Cg =
2 ln(2 z / r )
δ z δ Cg
=
ln(2 z / r ) . In Fig. 2a, we
gate, this can be translated into a distance change,
z
Cg
estimate the amplitude of motion to be δz≈10 nm. Calculating the driving force using
equation (1), we get F = C g′Vg DCδ Vg ≈ 60 fN . Thus, we estimate the effective spring
constant for this resonance to be keff =
F
Q ≈ 4 ×10−4 N m−1. Note that this effective
δz
spring constant is different for each resonance.
As the amplitude of the oscillation is increased, we can expect the nonlinear
effects due to the change in spring constant to become important. It is well known that
nonlinear oscillators have a bistable region in their response-frequency phase space
Page 4 of 12
which experimentally results in a hysteretic response22. The onset of nonlinear effects in
our case corresponds to driving voltages of 15 mV. Assuming the same parameters as
above yields an amplitude of motion of 30 nm.
Another important parameter characterizing the oscillator is the quality factor Q,
the ratio of the energy stored in the oscillator to the energy lost per cycle owing to
damping. Maximizing Q is important for most applications. It is in the range of 40–200
for our samples with no observed frequency dependence. Previous measurements on
larger multiwalled nanotubes at room temperature and ropes of single-walled nanotubes
at low temperatures yielded8–11 values of Q in the range 150–2,500.
Because one source of dissipation could be air drag, we have studied the
dependence of the resonator properties on the pressure in the vacuum chamber. Figure 3c
summarizes the results for one device. Q decreases with pressure, and the resonance is no
longer observed above pressures of 10 torr. This is in good agreement with calculations23.
At lower pressures, air losses should be minimal. Many other sources could be
contributing to the damping, including the motion of surface adsorbates and ohmic losses
due to the motion of electrons on and off the tube. The former is difficult to estimate, but
we have calculated the magnitude of the latter and find it to be insignificant. Another
important potential source of dissipation is clamping losses where the nanotube is
attached to the substrate; the tube may lose energy by sticking and unsticking from the
surface during oscillation. Experiments on devices with different clamping geometries
are necessary to investigate this issue.
The nanotube oscillator parameters presented above are representative of all of
our measured devices. Using these parameters, we can calculate the force sensitivity of
the device at room temperature. The smallest detected motion of the nanotube was at a
resonant driving voltage of δVg≈1 mV in the bandwidth of 10Hz. The sensitivity was
limited by the Johnson–Nyquist electronic noise from the nanotube. Using equations (1)
and (3) above, this corresponds to a motion of ~0.5 nm on resonance and a force
sensitivity of ~1 fN Hz−1/2. This is within a factor of ten of the highest force sensitivities
measured at room temperature24.
The ultimate limit on force sensitivity is set by the thermal vibrations of the
4k B kT
aN Hz−1/2
nanotube. The corresponding force sensitivity is δ Fmin =
= 20
ω 0Q
for typical parameters. The observed sensitivity is 50 times lower than this limit. This is
probably due to the relatively low values of transconductance for the measured nanotubes
Page 5 of 12
at room temperature. At low temperatures (~1 K), the sensitivity should increase by
orders of magnitude owing to high transconductance associated with Coulomb
oscillations19. Even without increasing Q, force sensitivities below 5 aN should
theoretically be attainable at low temperatures. This is comparable to the highest
sensitivities measured25–28. The combination of high sensitivity, tunability, and highfrequency operation make nanotube oscillators promising for a variety of scientific and
technological applications.
1.
Sidles, J. A. et al. Magnetic-resonance force microscopy. Rev. Mod. Phys. 67,
249–265 (1995).
2.
Roukes, M. Plenty of room indeed. Sci. Am. 285, 48 - 57 (2001).
3.
Nguyen, C. T. C. Frequency-Selective MEMS for Miniaturized Low-Power
Communication Devices. IEEE Trans. Microwave Theory Tech. 47, 1486-1503
(1999).
4.
Nguyen, C. T. C., Wong, A.-C. & Hao, D. in Solid-State Circuits Conference 7879, 1999. Digest of Technical Papers. ISSCC. 1999 IEEE International (San
Francisco, California)
5.
Cho, A. Physics — Researchers race to put the quantum into mechanics. Science
299, 36–37 (2003).
6.
LaHaye, M. D., Buu, O., Camarota, B. & Schwab, K. C. Approaching the
quantum limit of a nanomechanical resonator. Science 304, 74–77 (2004).
7.
Dresselhaus, M. S., Dresselhaus, G. & Avouris, P. Carbon Nanotubes (SpringerVerlag: Berlin, 2001).
8.
Poncharal, P., Wang, Z. L., Ugarte, D. & de Heer, W. A. Electrostatic deflections
and electromechanical resonances of carbon nanotubes. Science 283, 1513–1516
(1999).
9.
Gao, R. P. et al. Nanomechanics of individual carbon nanotubes from
pyrolytically grown arrays. Phys. Rev. Lett. 85, 622–625 (2000).
10.
Reulet, B. et al. Acoustoelectric effects in carbon nanotubes. Phys. Rev. Lett. 85,
2829–2832 (2000).
Page 6 of 12
11.
Purcell, S. T., Vincent, P., Journet, C. & Binh, V. T. Tuning of nanotube
mechanical resonances by electric field pulling. Phys. Rev. Lett. 89, 276103
(2002).
12.
Babic, B., Furer, J., Sahoo, S., Farhangfar, S. & Schonenberger, C. Intrinsic
thermal vibrations of suspended doubly clamped single-wall carbon nanotubes.
Nano Lett. 3, 1577–1580 (2003).
13.
Minot, E. D. et al. Tuning carbon nanotube band gaps with strain. Phys. Rev.
Lett. 90, 156401 (2003).
14.
Kong, J., Soh, H. T., Cassell, A. M., Quate, C. F. & Dai, H. J. Synthesis of
individual single-walled carbon nanotubes on patterned silicon wafers. Nature
395, 878–881 (1998).
15.
Hertel, T., Walkup, R. E. & Avouris, P. Deformation of carbon nanotubes by
surface van der Waals forces. Phys. Rev. B 58, 13870–13873 (1998).
16.
Tans, S. J., Verschueren, A. R. M. & Dekker, C. Room-temperature transistor
based on a single carbon nanotube. Nature 393, 49–52 (1998).
17.
Zhou, C. W., Kong, J. & Dai, H. J. Intrinsic electrical properties of individual
single-walled carbon nanotubes with small band gaps. Phys. Rev. Lett. 84, 5604–
5607 (2000).
18.
Minot, E. D., Yaish, Y., Sazonova, V. & McEuen, P. L. Determination of
electron orbital magnetic moments in carbon nanotubes. Nature 428, 536–539
(2004).
19.
Knobel, R. G. & Cleland, A. N. Nanometre-scale displacement sensing using a
single electron transistor. Nature 424, 291–293 (2003).
20.
Brenner, D. W. Empirical potential for hydrocarbons for use in simulating the
chemical vapor-deposition of diamond films. Phys. Rev. B 42, 9458–9471
(1990).
21.
Sapmaz, S., Blanter, Y. M., Gurevich, L. & van der Zant, H. S. J. Carbon
nanotubes as nanoelectromechanical systems. Phys. Rev. B 67, 235414 (2003).
22.
Yurke, B., Greywall, D. S., Pargellis, A. N. & Busch, P. A. Theory of amplifiernoise evasion in an oscillator employing a nonlinear resonator. Phys. Rev. A. 51,
4211–4229 (1995).
Page 7 of 12
23.
Bhiladvala, R.B. & Wang, Z. J. Effect of fluids on the Q-factors and resonance
requency of oscillating beams. Phys. Rev. E. 69, 036307 (2004)
24.
Jenkins, N. E. et al. Batch fabrication and characterization of ultrasensitive
cantilevers with submicron magnetic tips. J. Vac. Sci. Technol. B 22, 909–915
(2004).
25.
Stowe, T. D. et al. Attonewton force detection using ultrathin silicon cantilevers.
Appl. Phys. Lett. 71, 288–290 (1997).
26.
Mohanty, P., Harrington, D. A. & Roukes, M. L. Measurement of small forces in
micron-sized resonators. Physica B 284, 2143–2144 (2000).
27.
Stipe, B. C., Mamin, H. J., Stowe, T. D., Kenny, T. W. & Rugar, D. Magnetic
dissipation and fluctuations in individual nanomagnets measured by
ultrasensitive cantilever magnetometry. Phys. Rev. Lett. 86, 2874–2877 (2001).
28.
Mamin, H. J. & Rugar, D. Sub-attonewton force detection at millikelvin
temperatures. Appl. Phys. Lett. 79, 3358–3360 (2001).
Acknowledgements We thank E. Minot for discussions. This work was supported by the NSF through the
Cornell Center for Materials Research and the NIRT program, and by the MARCO Focused Research
Center on Materials, Structures, and Devices. Sample fabrication was performed at the Cornell Nano-Scale
Science and Technology Facility (a member of the National Nanofabrication Infrastructure Network),
funded by the NSF.
Correspondence and requests for materials should be addressed to L.M. (mceuen@ccmr.cornell.edu).
Figure 1 Device geometry and diagram of experimental set-up. a, A false-colour SEM
image of a suspended device (top) and a schematic of device geometry (bottom). Scale
bar, 300 nm. Metal electrodes (Au/Cr) are shown in yellow, and the silicon oxide surface
in grey. The sides of the trench, typically 1.2–1.5 µm wide and 500 nm deep, are marked
with dashed lines. A suspended nanotube can be seen bridging the trench. CVD growth is
known to produce predominantly single- and double-walled nanotubes, but we did not
perform detailed studies of the number of walls for the nanotubes on our samples. b, A
diagram of the experimental set-up. A local oscillator(LO) voltage δVsdω+∆ω (usually
around 7 mV) is applied to the source(S) electrode at a frequency offset from the high
frequency (HF) gate voltage signal δVgω by an intermediate frequency ∆ω of 10 kHz. The
current from the nanotube is detected by a lock-in amplifier through the drain electrode
(D), at ∆ω, with time constant of 100 ms.
Page 8 of 12
Figure 2 Measurements of the resonant response. The measurements were done on 11
devices, both semiconducting and small bandgap semiconducting nanotubes in a vacuum
chamber at pressures below 10−4 torr. The maximum conductance Gmax, and the
transconductance dG/dVgmax, are given below for the presented devices. a, Detected
current as a function of driving frequency taken at Vg=2.2 V, δVg=7 mV for device 1
(Gmax=12.5 µS, dG/dVgmax=7 µS V−1). The solid black line is a lorenzian fit to the data
with an appropriate phase shift between the driving voltage and the oscillation of the
tube. The fit yields the resonance frequency fo=55 MHz, and quality factor Q=80. b, c,
Detected current (plotted as a derivative in colour scale) as a function of gate voltage and
frequency for devices 1 and 2 (Gmax=10 µS, dG/dVgmax=0.3 µS V−1). Panel a is a vertical
slice through panel b at Vg=2.2 V (marked with a dashed black line). The insets to the
figures show the extracted positions of the peaks in the frequency–gate voltage space for
the respective colour plots. A parabolic and a Vg2/3 fit of the peak position are shown in
red and green, respectively. d, Theoretical predictions for the dependence of vibration
frequency on gate voltage for a typical device with length L=1.75 µm, and radius r=1 nm.
The calculations were performed for several different values of slack s (s=(L−W)/W,
where L is the tube’s length and W is the distance between clamping points). The
calculations for 0.5%, 1% and 2% slack are shown in blue, red and green, respectively.
Notice the appropriately rescaled x-axis.
Figure 3 Amplitude and pressure dependence of the resonance. a The measured quality
factor Q of the resonance and the height of the resonance peak for device 3 (Gmax=15 µS,
dG/dVgmax=4 µS V−1) are shown in red open squares and black solid squares,
respectively, as a function of driving voltage δVgω. Linear behaviour is observed at low
voltages, but Q decreases and the height of the peak saturates at higher driving voltages.
b, Trace of detected current versus frequency with the background signal subtracted for
device 2 at two different driving voltages δVg=8.8 mV and δVg=40 mV. The solid black
line is a lorenzian fit to the low bias data. The traces of the current as the frequency is
swept up and down are shown in blue and black, respectively. Hysteretic switching can
be observed. c, Pressure dependence of the resonance peak for device 4 (Gmax=7.7 µS,
dG/dVgmax=0.6 µS V−1). The Q of the resonance peak is shown in red open squares. The
peak was no longer observed above pressures of 10 torr.
Page 9 of 12
Figure 1.
Page 10 of 12
Figure 2.
Page 11 of 12
Figure 3.
Page 12 of 12