Carbon Nanotube Flow Sensors

Shankar Ghosh, et al.

Science 299, 1042 (2003);
DOI: 10.1126/science.1079080

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of August 8, 2008 ):

A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/cgi/content/full/299/5609/1042#related-content
This article cites 9 articles, 3 of which can be accessed for free:
http://www.sciencemag.org/cgi/content/full/299/5609/1042#otherarticles
This article has been cited by 106 article(s) on the ISI Web of Science.
This article has been cited by 2 articles hosted by HighWire Press; see:
http://www.sciencemag.org/cgi/content/full/299/5609/1042#otherarticles
This article appears in the following subject collections:
Physics, Applied
http://www.sciencemag.org/cgi/collection/app_physics
Information about obtaining reprints of this article or about obtaining permission to reproduce
this article in whole or in part can be found at:
http://www.sciencemag.org/about/permissions.dtl

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2003 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.

Downloaded from www.sciencemag.org on August 8, 2008

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/cgi/content/full/299/5609/1042

REPORTS

References and Notes

1. I. Schlichting et al., Science 287, 1615 (2000).

2. T. L. Poulos, Curr. Opin. Struct. Biol. 5, 767 (1995).
, B. C. Finzel, I. C. Gunsalus, G. C. Wagner, J.
3.
Kraut, J. Biol. Chem. 260, 6122 (1985).
4. T. L. Poulos, B. C. Finzel, A. J. Howard, Biochemistry
25, 5314 (1986).
5. T. L. Poulos, R. Raag, FASEB J. 6, 674 (1992).
6. H. Y. Li, S. Narasimhulu, L. M. Havran, J. D. Winkler,
T. L. Poulos, J. Am. Chem. Soc. 117, 6297 (1995).
7. B. J. Wallar, J. D. Lipscomb, Chem. Rev. 96, 2625
(1996).
8. A. C. Rosenzweig, C. A. Frederick, S. J. Lippard, P.
Nordlund, Nature 366, 537 (1993).
9. A. C. Rosenzweig, P. Nordlund, P. M. Takahara, C. A.
Frederick, S. J. Lippard, Chem. Biol. 2, 409 (1995).
10. J. D. Lipscomb, Annu. Rev. Microbiol. 48, 371 (1994).
11. D. T. Gibson, R. E. Parales, Curr. Opin. Biotechnol. 11,
236 (2000).
12. T. Hudlicky, D. Gonzalez, D. T. Gibson, Aldrichim. Acta
32, 35 (1999).
13. B. Kauppi et al., Structure 6, 571 (1998).
14. A. M. Jeffrey et al., Biochemistry 14, 575 (1975).
15. B. E. Haigler, D. T. Gibson, J. Bacteriol. 172, 457
(1990).
, J. Bacteriol. 172, 465 (1990).
16.
17. M. D. Wolfe, J. V. Parales, D. T. Gibson, J. D. Lipscomb,
J. Biol. Chem. 276, 1945 (2001).
18. M. D. Wolfe, J. D. Lipscomb, J. Biol. Chem. 278, 829
(2003).

1042

19. M. D. Wolfe et al., Biochemistry 41, 9611 (2002).

20. K. Chen, L. Que, Angew. Chem. Int. Ed. 38, 2227 (1999).
21. K. Chen, M. Costas, J. Kim, A. K. Tipton, L. Que Jr.,
J. Am. Chem. Soc. 124, 3026 (2002).
22. E. Carredano et al., J. Mol. Biol. 296, 701 (2000).
23. See supporting data on Science Online.
24. D. R. Boyd, G. N. Sheldrake, Nat. Prod. Rep. 15, 309
(1998).
25. D. L. Harris, Curr. Opin. Chem. Biol. 5, 724 (2001).
26. S. Ahmad et al., Inorg. Chem. 27, 2230 (1988).
27. F. Neese, E. I. Solomon, J. Am. Chem. Soc. 120, 12829
(1998).
28. F. Mather et al., J. Am. Chem. Soc. 124, 4966 (2002).
29. S. M. Resnick, K. Lee, D. T. Gibson, J. Ind. Microbiol.
17, 438 (1996).
30. Z. Otwinowski, Data Collection and Processing
(Daresburry Laboratories, Warrington, UK, 1993), pp.
56 62.
31. J. W. Pugrath, Acta Crystallogr. D55, 1718 (1999).
32. We thank J. Hajdu and his group at Uppsala for help

in carrying out several experiments; M. Niebergall for

helpful tips; several synchrotron stations and beam
lines for help with data collection (European Synchrotron Radiation Facility, Grenoble, Max Lab, Lund, and
Industrial Macromolecular Crystallography AssociationCollaborative Access Team, Advanced Photon
Source); and J. Lipscomb for helpful discussions. Supported by the Swedish Research Council for Environment, Agricultural Science, and Spatial Planning (H.E.
and S.R.), the Swedish Research Council (H.E.), and
U.S. Public Health Service grants GM62904 (S.R.) and
GM29909 (D.T.G.).
Supporting Online Material
www.sciencemag.org/cgi/content/full/299/5609/1039/
DC1
Materials and Methods
References
Figs. S1 to S4
3 September 2002; accepted 19 December 2002

Carbon Nanotube Flow Sensors

Shankar Ghosh,1 A. K. Sood,1* N. Kumar2
We report that the ow of a liquid on single-walled carbon nanotube bundles
induces a voltage in the sample along the direction of the ow. The voltage that
was produced t a logarithmic velocity dependence over nearly six decades of
velocity. The magnitude of the voltage depended sensitively on the ionic
conductivity and on the polar nature of the liquid. Our measurements suggest
that the dominant mechanism responsible for this highly nonlinear response
involves a direct forcing of the free charge carriers in the nanotubes by the
uctuating Coulombic eld of the liquid owing past the nanotubes. We
propose an explanation based on pulsating asymmetric ratchets. Our work
highlights the device potential for nanotubes as sensitive ow sensors and for
energy conversion.
Carbon nanotubes are currently being studied
in an effort to understand their novel structural, electronic, and mechanical properties
and to explore their immense potential for
many applications in nanoelectronics (1), and
as actuators (2) and sensors (3). A metallic
single-walled carbon nanotube (SWNT),
which is a low-dimensional system, has electrons that form a strongly correlated liquid,
called a Luttinger liquid (4). This liquid
shows a separation of spin and charge and
demonstrates the associated characteristic
power-law dependence of the nanotube resistance on the bias voltage and temperature.
Kral and Shapiro (5) also showed theoretically the generation of an electric current in a
metallic carbon nanotube immersed in a
flowing liquid. One mechanism they proposed for the generation of the electric current and voltage is the transfer of momentum
from the flowing liquid molecules to the
acoustic phonons in the nanotube as the phonon quasi-momentum, which in turn drags
free charge carriers in the nanotube. Another
Department of Physics, Indian Institute of Science,
Bangalore 560 012, India. 2Raman Research Institute,
C.V. Raman Avenue, Bangalore 560 080, India.
1

*To whom correspondence should be addressed. Email: [email protected]

mechanism noted by these authors involves a

direct scattering of the free carriers from the
fluctuating Coulombic fields of the ions or
polar molecules in the flowing liquid. They
argued, however, that the latter mechanism
creates a current that is five orders of magnitude smaller than the current that results from
the phonon-induced electron drag.
We report the experimental observation of
the voltage generated by the flow of a polar
liquid over SWNT bundles. It was observed that
the induced voltage tends to saturate at flow
velocities as low as 105 m/s. The observed data
can be fitted empirically to a logarithmic dependence of the voltage on the flow velocity, in
sharp contrast to the linear dependence predicted by Kral and Shapiro (5). The magnitude of
the voltage induced along the nanotube by the
flow depends significantly on the ionic
strength of the flowing liquid, suggesting that
the mechanism involving fluctuating Coulombic fields dominates that of the phonon-mediated electron drag. This has led us to propose
a mechanism for the observed flow-induced
voltage based on a pulsating asymmetrical
thermal ratchet model. This mechanism involves the fluctuating Coulombic potentials
that result from the ionic liquid in flow, where
the asymmetry is provided by the velocity
gradient (shear) at the liquid-solid interface.

14 FEBRUARY 2003 VOL 299 SCIENCE www.sciencemag.org

Downloaded from www.sciencemag.org on August 8, 2008

NDO with oxygen or hydrogen peroxide, as

proposed by Wolfe and Lipscomb (17, 18).
The unique feature of NDO that sets it apart
from P450 is its ability to catalyze stereospecific
addition of dioxygen to aromatic compounds,
resulting in the formation of cis-dihydrodiols
(12, 24). For cytochrome P450 and MMO, transient high iron oxidation states are required for
the necessary oxygen cleavage (25); this is not
the case for NDO, where both atoms of dioxygen react in concert with the substrate. The four
electrons required for catalysis are one each
from the active-site ferrous iron and the Rieske
center, and two from the double bond of the
substrate. An important difference in the geometry of the active sites of cytochrome P450 and
NDO is that the active-site iron is highly accessible in NDO, with a large part of its surface
available for the binding of both oxygen atoms
to iron. The favorable position of the substrate
and Asn201 further favors side-on binding. It is
possible that for cytochrome P450 and other
heme proteins, the presence of the planar heme
group reduces the accessible surface of iron,
resulting in an end-on binding of oxygen.
We have directly observed a side-on dioxygen species bound to iron. Computational
studies (26) have predicted the existence of
this species, and there are spectroscopic data
consistent with side-on binding (27, 28). We
suggest that side-on binding of oxygen as a
starting point for oxygen activation by NDO
has direct implications for the mechanism of
dihydroxylation of NDO in terms of the stereo- and regiospecific reactions catalyzed by
the enzyme (29). The results presented here
provide a new basis for further computational
and structural investigations of oxygen activation by biological systems and may facilitate the design of chemical catalysts capable
of cis-dihydroxylation.

The SWNT bundles were prepared by the

electric-arc method, followed by purification
processes (6 ). These nanotube bundles,
which have an average tube diameter of 1.5
nm, were densely packed between two metal
electrodes. The dimensions of the sensor
were 1 103 m along the flow, 2 104
m thick, and 2 103 m wide. The resistivity of the material was about 0.02 ohmm, and
the sensor had a linear current-voltage characteristic (up to 20 mV). To avoid any turbulent flow that may have been caused by the
expansion of the flow at the inlet of the flow
chamber, we placed the nanotube sensor at
the center of a glass tube with an inner diameter of 0.03 m and a length of 0.9 m. The
schematic of the experimental setup is shown
in Fig. 1. The velocity u of the liquid flowing
past the sample was measured from the bulk
flow rate, which was varied by adjusting the
height of the reservoir and the opening of the
valve. The liquid enters the flow chamber
against gravity such that the formation of air
pockets is avoided. The flow at the center of
the cell is expected to be laminar because the
Reynolds number (Re) (300 for a velocity
of 102 m/s) was much less than the critical
value of 2000 for the onset of turbulent flow
in a pipe. The measurements shown here
were taken after the transients had subsided.
When the sample was immersed in water
at rest, a voltage (1 mV) developed along
the sample as an electrochemical potential
difference at the interface of the SWNT bundle with the metal electrodes. This offset
voltage was subtracted from the voltage measured at finite velocities of the liquid to obtain the flow-induced voltage. The voltage
developed only along the flow direction; no
voltage was generated in the direction perpendicular to the flow (i.e., along the width).
Figure 1 shows the observed dependence
of the flow-induced voltage on the flow velocity of water. Even at the very low flow
velocity of 5 106 m/s, a voltage of 0.65
mV is generated. The solid line in Fig. 1 is a
fit to the empirical relation V log(u
1), where and are parameters, with
0.6 mV and 6.5 106 s/m for best fit.
The induced voltage for a fluid flow velocity
of 1.8 103 m/s was 2.67 mV for a sample
length of 1 103 m and was 1.64 mV for a
sample length of 0.5 103 m, suggesting a
one-dimensional bulk property rather than
any local tip effects. Figure 2 compares the
voltage developed along the nanotubes as a
function of the flow velocity for a series of
fluid mixtures. A notable effect is the strong
dependence of the flow-induced voltage on
the ionic strength of the flowing liquid. It can
be seen that for u 104 m/s, the voltage for
1.2 M HCl is about five times that for water.
We compare our experimental results with
those expected from the calculations of Kra l and
Shapiro (5). They considered metallic nanotubes

that were densely packed in two layers separated

by a layer of water of thickness d (10 nm). In
their mechanism, the force exerted by the liquid
with a flow velocity u on the surface of the
nanotubes is F (u/d) interface periphery interface length. This force results in the
steady-state quasi-momentum (p) transfer to the
phonon bath, with p F, where is the shear
viscosity of the liquid and is the phonon
umklapp time in the nanotubes. Using this value
of p, we calculated the steady-state electron
distribution from the Boltzmann equation containing the electron-phonon scattering term,
which, in turn, gave the hot phononinduced
electronic current ( Jep). In this model, the calculated current Jep is directly proportional to
the flow velocity u and to the liquid viscosity .
However, as is shown in Fig. 1, the experimentally observed voltage induced by the flow is
nearly logarithmic in the flow velocity u. To
check for the dependence of the voltage on ,
we repeated the experiments using water-glycerol mixtures. The mixtures used were water:
glycerol 88:12 ( 113 mPa s) and water:
glycerol 75:25 ( 234 mPa s). These
mixtures have viscosities about two orders of
magnitude higher than that of water ( 0.9
mPa s). We observed that the flow-induced

voltage fell with increasing fluid viscosity. We

also found that the voltage developed in the case
of the polar liquids (such as water) was greater
than the voltage developed in the case of the less
polar liquids (such as methanol). Thus, for u
6 104 m/s, Vwater 2.1 mV, whereas
Vmethanol 0.2 mV. Thus, our data set reveals
three features: the sublinear (almost logarithmic)
voltage response to flow velocity, its strong
dependence on the ionic and polar nature of the
flowing liquid, and its relatively weak dependence on the liquid viscositythe latter, in fact,
acting to reduce the flow-induced voltage.
The one-dimensional nature of the SWNT is
crucial for the generation of a net electrical
signal in the sample. Experiments on graphite
did not produce any measurable electrical signal. However, experiments with multiwalled
carbon nanotubes did generate a voltage, but it
was about 10 times smaller than the voltage
produced by the SWNT for similar dimensions
of the sensing element and for the same flow
velocities of the liquid.
The most notable feature of our experimental data is the slow, nearly saturating growth of
the induced electric voltage (current) with the
increasing flow velocity. This strongly sublinear, almost logarithmic dependence on the flow

www.sciencemag.org SCIENCE VOL 299 14 FEBRUARY 2003

Fig. 1. Variation of voltage (V )

developed as function of velocity
(u) of water. Solid line is a t to
functional form as V log
(u 1), where and are
constants. Inset shows experimental setup where R is the reservoir, L is the valve controlling
the liquid ow, S is the cylindrical glass ow chamber, and G is
the voltmeter.

Downloaded from www.sciencemag.org on August 8, 2008

REPORTS

Fig. 2. Induced Voltage (V ) as

function of liquid velocities, for
water (F); methanol (E), mixture of water:glycerol (75:25)
(); Water:glycerol (88 :12) ();
0.6 M HCl (), and 1.2 M HCl
(). The ratios for the waterglycerol mixtures are by volume. Lines show the best t of
the data to Eq. 2. Inset shows
a schematic of asymmetric
ratchet potential.

1043

velocity rules out any explanation of our experimental results based on a linear response to the
fluid flow. We thus suggest a transport mechanism based on the known idea of a directed
voltage response resulting from the pulsating
asymmetric thermal ratchets (7, 8).
An idealized model envisages a one-dimensional periodic ratchet potential that is asymmetric, such as a saw-tooth potential (Fig. 2, inset),
and that is modulated in time by a multiplicative
process that can be, and usually is, stochastic.
Physically, the features that are essential for the
rectifying mechanism of this pulsating ratchet to
operate are the spatial bias (asymmetry) and the
(stochastic) temporal modulation. As for the
spatial order, it is physically essential for the
sense of the bias to be the same all along the
sample. This should enable us to invoke the
general Curie principle (7, 9) that ensures a
directed voltage and current response. These
minimal conditions are realized in our case, in
which the Coulombic potential resulting from,
for example, the fluctuating imbalances in the
local charge neutrality of the ionic fluid can
provide the stochastic potential, whereas the
Poiseuille shear-flow past the nanotube surface
provides the unidirectional bias. Thus, one can
imagine a transient accumulation of charge imbalance that is shear-deformed by the velocity
gradient near the surface, resulting in an asymmetric fluctuating potential. Indeed, a similar
idea of shear-induced dipolar (ellipsoidal) distortion of the ion-plus-polar atmosphere at the
liquid-solid interface was treated in detail by
Onsager and Fuoss (10) in their classic paper on
electrolytes (11). Although the exact mapping
would involve many details of the fluid-surface
interaction, the basic idea is generic in the sense
of the Curie principle (9). The one-dimensional
confinement of the charge carriers in the nanotube is essential to ensure that the uniformly
biased but randomly placed pulsating ratchets
along the sample length act additively in series
to induce the overall directed voltage. More
specifically, the electrostatic ratchet potential
would force the nanotube -electrons, whose
kinetics would resemble that of a classical diffusion in real space, which is consistent with the
observed high resistivity of the sample. Again, a
symmetrical, fluctuating potential moving with
the flow can, in principle, drag charge carriers to
induce a voltage. This, however, will give rise to

a linear dependence of the induced voltage and

current on flow velocity (5), which is in disagreement with our observations.
The inset of Fig. 2 shows a schematic of
the idealized asymmetric ratchet potential
acting on the charge carriers in the nanotubes.
For this idealized model of a pulsating asymmetric ratchet potential, the particle current
per particle is known to be given by (7, 12)

2
Q

Liquid

V0
(mV)

1.2 M HCl
0.6 M HCl
Water
Water:glycerol (88 :12)
Water:glycerol (75:25)
Methanol

9.2
5.3
2.2
0.6
0.4
0.16

1044

a (s/m) b (s/m)
5.4 104
4.8 104
7 104
15 104
25 104
53 104

471
191
16
14
14
3

e x
U x

e(y)
U( y)

d ydx

(1)

where (x) [()/(Q)]

1/U( y)]} d y;

prime denotes the derivative with respect to the

space coordinate; is the period (length) of the
pulsating ratchet ( 1 2); Q is the
strength of the fluctuating Coulombic potential
assumed to be a white noise; is the frictional
coefficient for the carriers in the nanotubes; Um
is the maximum of the ratchet potential; and
[(2 1)/()] is the dimensionless bias
(asymmetry) parameter. The above expression
for (x) assumes {[(kBT2 )/(QUm2 )]} 1,
where kBT is the thermal energy. Here, is the
all-important asymmetry parameter, which
should be proportional to the liquid flow velocity. Physically, can be identified with the
dimensionless (dipolar) ellipsoidal deformation
of the ionic atmosphere due to the shear-flow at
the interface (10) and is on the order of (/2
kBT)(u/), where is the boundary layer thickness near the nanotube, 1 is the Debye
screening length, and is the frictional force on
an ion moving with unit velocity. Typically,
(l/u), where is the kinematic viscosity
of the fluid and l is the interface length. For the
experimental parameter values, 1.
For small values of corresponding to
small flow velocities, Eq. 1 reduces to give
the induced voltage ( proportional to the current j) as
V

Table 1. Comparison of the parameters V0, a, and

b for a series of liquids.

j 0 1 e QU m

V 0 1 e au
e bu

(2)

where V0, a, and b are the parameters following

from Eq. 1. Figure 2 shows the fit of our data for
various liquids to the functional form in Eq. 2
with V0, a, and b as the fit parameters (Table 1).
The saturating effect is clearly seen and the fit is
reasonable over many decades of flow-velocity
variation. The addition of HCl increases the
concentration of H ions (along with their hydration shells), and hence, the observed increase
in the induced voltage.
An important point must be made regarding

the direction ( polarity) of the voltage and current in relation to the fluid flow direction. In
our experiments, we observed that the electric current and the fluid were flowing in
the same direction. In a linear response
theory based on the viscous drag (the phonon wind), the particle current has the same
direction as the fluid flow velocity, and,
therefore, the sign of the electric voltage
and current is determined by the sign of the
charge carriers (electron or hole). Thus, for
the holelike carriers, the induced voltage
and current have the same direction as the
flow velocity. This, however, is not the case
for a pulsating ratchet model in general (7).
For a given direction of the ratchet bias
(asymmetry), the direction of the voltage is
independent of the sign of the charge on the
carriers. It will, of course, depend on the
sense of the bias (asymmetry) of the ratchet
potential. In a polar fluid of a given ionic
strength, such as concentrations of H and
OH ions (along with their hydration shells),
we expect to have ratchet potentials of either
sign (bias). The net effect is then determined
by the dominant ratchet. What is important
here, however, is that there is no a priori
(symmetry) reason to expect the oppositely
biased ratchets to give an exact cancellation.
Finally, a flow sensor that is based on
SWNT directly produces an electrical signal in
response to a fluid flow. We believe that this
sensor can be scaled down to length dimensions
on the order of micrometersi.e., the length of
the individual nanotubesmaking it usable in
very small liquid volumes. The sensor also has
high sensitivity at low velocities and a fast response time (better than 1 ms). The nanotubes
also could be used to make a voltage and current
source in a flowing liquid environment, which
may have interesting biomedical applications.
References and Notes

1. P. G. Collins, M. S. Arnold, P. Avouris, Science 292,

706 (2001).
2. R. H. Baughman et al., Science 284, 1340 (1999).
3. J. Kong et al., Science 287, 622 (2000).
4. R. Egger, A. Gogolin, Phys. Rev. Lett. 79, 5082 (1997).
5. P. Kra
l, M. Shapiro, Phys. Rev. Lett. 86, 131 (2001).
6. P. V. Teredesai et al., Chem. Phys. Lett. 319, 296 (2000).
7. P. Reimann, Phys. Rep. 361, 57 (2002).
8. F. Julicher, A. Ajdari, J. Prost, Rev. Mod. Phys. 69, 1269
(1997).
9. P. Curie, J. Phys. (Paris) Ser. 3, 3, 393 (1894).
10. L. Onsager, R. M. Fuoss, J. Phys. Chem. 36, 2689 (1932).
11. More microscopically, one may consider the Coulomb
potential of an ion undergoing thermal diffusion,
uid advection, and reection from the uid-solid
interface with aberration caused by the uid velocity.
Such an aberration, much like the shear, provides the
unidirectional bias throughout the length of the
nanotube.
12. P. Reimann, Phys. Rep. 290, 149 (1997).
13. We thank C. N. R. Rao and A. Govindaraj for SWNT
samples and A. Das for his help in performing some of
the experiments. A.K.S. thanks the Department of
Science and Technology (Government of India) for
nancial assistance.
4 October 2002; accepted 24 December 2002
Published online 16 January 2003;
10.1126/science.1079080
Include this information when citing this paper.

14 FEBRUARY 2003 VOL 299 SCIENCE www.sciencemag.org

Downloaded from www.sciencemag.org on August 8, 2008

REPORTS