ARTICLES

PUBLISHED ONLINE: 4 APRIL 2016 | DOI: 10.1038/NNANO.2016.44

Wafer-scale monodomain films of spontaneously

aligned single-walled carbon nanotubes
Xiaowei He1‡, Weilu Gao1‡, Lijuan Xie2, Bo Li3, Qi Zhang1, Sidong Lei3, John M. Robinson1†,
Erik H. Hároz4, Stephen K. Doorn4, Weipeng Wang3, Robert Vajtai3, Pulickel M. Ajayan3,
W. Wade Adams3, Robert H. Hauge3,5 and Junichiro Kono1,3,6*

The one-dimensional character of electrons, phonons and excitons in individual single-walled carbon nanotubes leads to
extremely anisotropic electronic, thermal and optical properties. However, despite significant efforts to develop ways to
produce large-scale architectures of aligned nanotubes, macroscopic manifestations of such properties remain limited.
Here, we show that large (>cm2) monodomain films of aligned single-walled carbon nanotubes can be prepared using slow
vacuum filtration. The produced films are globally aligned within ±1.5° (a nematic order parameter of ∼1) and are highly
packed, containing 1 × 106 nanotubes in a cross-sectional area of 1 μm2. The method works for nanotubes synthesized by
various methods, and film thickness is controllable from a few nanometres to ∼100 nm. We use the approach to create
ideal polarizers in the terahertz frequency range and, by combining the method with recently developed sorting
techniques, highly aligned and chirality-enriched nanotube thin-film devices. Semiconductor-enriched devices exhibit
polarized light emission and polarization-dependent photocurrent, as well as anisotropic conductivities and transistor
action with high on/off ratios.

the degree of alignment was low and the film thickness was small

O
ne of the grand challenges in nanoscience and nanotechnol-
ogy is how to create macroscopic devices by assembling and not controllable.
nano-objects while preserving their extraordinary proper- This means that the current state of this field is that there is still
ties. For example, individual single-walled carbon nanotubes no method available for producing large-area single-domain films of
(SWCNTs) possess unique one-dimensional properties that have highly aligned, densely packed and chirality-enriched SWCNTs,
stimulated much interest in diverse disciplines1, and worldwide despite many years of effort. The method we describe in the
efforts are in progress to produce large-scale architectures of present article provides a uniform, wafer-scale (>cm2) SWCNT
aligned SWCNTs2,3. Various methods have been proposed and/or film of an arbitrarily and precisely controllable thickness (from a
demonstrated, including both postgrowth and direct-growth few nanometres to ∼100 nm) with a high degree of alignment
schemes. An early study made use of a filtration method to align (S ≈ 1) and packing (∼1 × 106 nanotubes in a cross-sectional area
multiwalled CNTs, which showed strongly anisotropic properties4. of 1 μm2) in a well-controlled, simple and reproducible manner,
Vertically aligned SWCNTs can be grown directly by chemical regardless of the synthesis method, metallicity or chirality of the
vapour deposition (CVD)5,6, but the CNT densities in the resulting SWCNTs used. Furthermore, the produced films are compatible
films are low, making them incompatible with standard microfabri- with standard microfabrication processes used to fabricate various
cation technology. Furthermore, no type or chirality selectivity can electronic and photonic devices.
be implemented. Some alignment has been observed by dispersing
CNTs in liquid-crystal solvents that can be nematically ordered7–9. Global spontaneous alignment of CNTs
Stronger alignment, with a nematic order parameter S as high as The process starts with the preparation of a well-dispersed CNT
∼0.8, has been achieved by mechanically stretching a polymer film suspension. CNT powder is dispersed with surfactant in water to
containing CNTs10,11; however, the existence of the matrix material create an aqueous CNT suspension using tip sonication and ultra-
as well as low CNT densities in such films severely limit their range centrifugation (see Methods). In a second step, a vacuum filtration
of applications. As far as spontaneous alignment of SWCNTs is method (Fig. 1a,b) is used, which is a well-established technique for
concerned, large S values have been realized only locally12–14, and forming wafer-scale films of randomly oriented CNTs with control-
global ordering comes at the expense of S (<0.4)15,16. Magnetic align- lable thickness21. The CNT suspension prepared in the first step is
ment of SWCNTs has been demonstrated17,18, but impractically poured into a filtration funnel with a small-pore-size filter mem-
large magnetic fields are required to obtain meaningful alignment. brane, and a differential pressure across the filter membrane
Inkjet printing has proven useful for creating local alignment19, pushes the suspension slowly through the pores, leaving CNTs on
but no large-scale alignment has been reported. Recently, the the filter membrane. To obtain spontaneous CNT alignment, one
Langmuir–Schaefer method was used to prepare globally aligned, has to satisfy three critical conditions: (1) the surfactant concen-
semiconductor-enriched CNTs with high surface coverage20, but tration must be below the critical micelle concentration (CMC);

1
Department of Electrical and Computer Engineering, Rice University, Houston, Texas 77005, USA. 2 College of Biosystems Engineering and Food Science,
Zhejiang University, Hangzhou, Zhejiang 310058, China. 3 Department of Materials Science and NanoEngineering, Rice University, Houston, Texas 77005,
USA. 4 Center for Integrated Nanotechnologies, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, USA. 5 Department of Chemistry, Rice
University, Houston, Texas 77005, USA. 6 Department of Physics and Astronomy, Rice University, Houston, Texas 77005, USA. †Present address:
Department of Physics, University of Colorado, Boulder, Colorado 80302, USA. ‡These authors contributed equally to this work. *e-mail: kono@rice.edu

NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology 633

© 2016 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44

a b Filter d e f g
CNT film membrane

1 inch
1 µm 50 nm 4 nm 2 nm
CNT
suspension i
Transfer

h
25
c 20
1 cm

Counts
15
2σ ~ 3° j
10

0
Vacuum −10 −5 0 5 10
Deviation (°) 50 µm

Figure 1 | Fabrication and characterization of wafer-scale monodomain films of aligned CNTs. a, A CNT suspension goes through a standard vacuum
filtration system. For spontaneous CNT alignment to occur, the filtration speed must be kept low and the CNTs must be well dispersed in the suspension.
b, A wafer-scale, uniform CNT film is formed on the filter membrane. c, Optical image of the produced film after being transferred to a transparent substrate
by dissolving the filter membrane. d–f, SEM image (d), a high-resolution SEM image (e) and a top-view TEM image (f) of the film, showing strong
alignment and high density. g, A high-resolution cross-sectional TEM image, showing a high cross-sectional areal density of ∼1 × 106 μm–2. h, Angular
distribution of CNTs within a 1 cm2 area of the film, with a standard deviation of 1.5°, determined by SEM image analysis. i,j, The film is opaque to light
polarized parallel to the CNT alignment direction and transparent to light polarized perpendicular to the alignment direction on a macroscopic scale (i) and
a microscopic scale (j). Note also that the film can be easily patterned using conventional photolithography techniques (j).

(2) the CNT concentration must be below a threshold value; and to the visible, as shown in Fig. 2b (the energy axis is on a logarithmic
(3) the filtration process must be well controlled at a low speed. scale). In particular, there is no detectable attenuation within exper-
When these conditions are met, a wafer-scale, uniform and imental error for the perpendicular polarization in the entire tera-
aligned CNT film is formed on the filter membrane (Fig. 1b). hertz/infrared range (<1 eV), whereas there is a prominent, broad
The film can then be transferred to any substrate in a straight- peak at ∼0.02 eV in the parallel case due to plasmon resonance22.
forward manner after dissolving the filter membrane in an appropri- Figure 2c plots the same spectra with the energy axis on a linear
ate solvent (see Methods). The result is a large-area, semi-transparent scale, to more clearly show the interband absorption; that is, the first
film of aligned CNTs, as shown in optical (Fig. 1c), scanning electron two interband transitions for semiconducting nanotubes (E11 S
and
microscopy (SEM, Figs. 1d,e) and transmission electron microscopy E22 ) and the first interband transition in metallic nanotubes (E11
S M
).
(TEM, Fig. 1f) images. As the cross-sectional TEM image in Due to the exceptionally high degree of CNT alignment, these peaks
Fig. 1g shows, the film is densely packed, with ∼1 × 106 CNTs are completely absent for perpendicular polarization, and a broad
found in a cross-sectional area of 1 μm2. Individual CNTs within absorption feature is instead observed in an intermediate energy
the film are all aligned with one another, forming a globally region between the E11 S S
and E22 peaks. We attribute this feature to
ordered structure with an angle standard deviation of ∼1.5° across the cross-polarized and depolarization-suppressed E12 S S
/E21 absorption
the entire film (Fig. 1h). The film is optically polarized, that is, peak23, previously detected in polarized photoluminescence excitation
linearly dichroic (Fig. 1i,j), being opaque to light polarized parallel spectroscopy experiments in individualized CNTs24,25.
to the CNT alignment direction and transparent to light polarized As noted previously26, the exceptionally strong polarization depen-
perpendicular to the alignment direction. Using cross-polarized dence of terahertz transmission through aligned CNT films can be
microscopy, strong optical anisotropy can be demonstrated both on used to form an ideal terahertz polarizer with extremely large extinc-
a macroscopic (centimetre) scale (Fig. 1i) and a microscopic (micro- tion ratios (ER). ER = T∥ /T⊥ , where T∥ (T⊥) is the transmittance for
metre) scale (Fig. 1j), reflecting the global and local CNT alignment, parallel (perpendicular) polarization. Figure 2d shows time-domain
respectively. Finally, the film can be patterned easily using waveforms of terahertz radiation transmitted through an aligned
conventional photolithography techniques (Fig. 1j). arc-discharge SWCNT film on an intrinsic silicon substrate for polar-
izations parallel and perpendicular to the alignment direction,
Spectroscopy of spontaneously aligned CNTs together with a reference waveform obtained for the substrate alone.
Figure 2 summarizes the results of spectroscopic characterization The data for the perpendicular case completely coincide with the
measurements of the aligned CNT films. Figure 2a presents polarized reference trace; that is, no attenuation occurs within the SWCNT
Raman spectra for a 15-nm-thick aligned film of arc-discharge film. On the other hand, there is significant attenuation for the parallel
SWCNTs with an average tube diameter of 1.4 nm, taken with case. Note that the terahterz beam had mm2 dimensions, thus probing
an excitation wavelength of 514 nm in two polarization configur- a macroscopic area. Figure 2e shows a more detailed polarization-
ations. The data were analysed using standard equations for the angle dependence of terahertz attenuation, plotted as a function of
angular dependence of SWCNT Raman spectra9 to deduce the the angle between the terahertz polarization and the nanotube align-
value of S, which was 0.96 for this particular film (Supplementary ment direction. The attenuation anisotropy allows us to calculate the
Section 3). The electromagnetic response of this film was strongly value of S in a straightforward manner26, which also agrees with the
polarization-dependent in the whole spectral range, from the terahertz value obtained by Raman spectroscopy. The value of ER

634 NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology

© 2016 Macmillan Publishers Limited. All rights reserved

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44 ARTICLES
a 2.0 b c 0.25
G 0.6 S
Parallel 0.20 E22
1.5 λexc = 514 nm Perpendicular E11M
Intensity (a.u.)
S
0.15 E11
0.4

−log10T

−log10T
1.0 IVV Parallel
0.10
IHH 0.2 S
E12/21
0.5
0.05
D Perpendicular
0.0 0.0
0.00

1,300 1,400 1,500 1,600 1,700 0.001 0.01 0.1 1 0.5 1.0 1.5 2.0
Raman shift (cm−1) Energy (eV) Energy (eV)

d e 1.0 f
6 Parallel
1.0 16
Perpendicular 0.8
Electric field (mV)

4 0.8 14
Reference

0.6 12

ER (dB)
−log10T
2 0.6

S
10
0.4 0.4
0
8
0.2 0.2
−2 6
0.0 0.0 4
2 4 6 8 0 90 180 270 360 40 60 80 100
Time (ps) Angle (°) Thickness (nm)

Figure 2 | Characterization of aligned CNT films through polarization-dependent optical spectroscopy. a, Raman spectra for an aligned arc-discharge CNT
film for two configurations. In the VV (HH) configuration, both the incident and scattered beams are polarized parallel (perpendicular) to the nanotube
alignment direction. b, Polarization-dependent attenuation spectra in a wide spectral range, from the terahertz/far-infrared to the visible. c, Expanded view of
b, showing interband transitions. d, Time-domain terahertz waveforms of transmitted terahertz radiation for parallel and perpendicular polarizations. The data
for the perpendicular case completely overlap with the reference trace; that is, no attenuation is detectable. The terahertz beam had a beam size on the scale
of mm2, thus probing a macroscopic area. e, Attenuation as a function of the angle between the terahertz polarization and the nanotube alignment direction.
f, Nematic order parameter (S, left axis), deduced from the terahertz attenuation data, and the extinction ratio (ER, right axis), as a function of film thickness.

monotonically increases with film thickness at ∼12 dB/100 nm, while ordering of rigid rods, for which Onsager’s theory predicts an
high values of S are maintained even for relatively thick films (Fig. 2f). upper limit of S = 0.79 (refs 15,27). The formation of a three-
dimensional nematic liquid-crystal phase of CNTs would require a
Universal applicability of vacuum filtration high CNT concentration8, with a threshold value of ∼5 mg ml–1
Our method of making aligned films is universally applicable to differ- for CNTs with lt/dt ≈ 1 × 103. This condition was not met in our
ent types of SWCNT. Table 1 lists seven representative suspensions, case (typical CNT concentration ∼15 μg ml–1 and lt/dt = 150–550),
which contained SWCNTs synthesized by the arc-discharge, CVD, suggesting that a different alignment mechanism is at work. A clue
CoMoCAT and HiPco methods. Here, dt is the average nanotube diam- to the nature of this mechanism comes from our observation that
eter and lt is the average nanotube length measured by atomic force the degree of alignment is sensitive to the hydrophobicity of the
microscopy (Supplementary Fig. 8). To disperse CNTs we used filter membrane surface, similar to an earlier report on two-dimen-
sodium deoxycholate (DOC) for suspensions 1 and 4–7 and sodium sional nematic ordering of DNA-wrapped CNTs12. Alignment was
dodecylbenzenesulfonate (SDBS) for suspension 2, as surfactant. achieved only when the filter membrane had a hydrophilic coating
Pre-functionalized, water-soluble SWCNTs were used in suspension layer—poly(vinlpyrrolidone) (PVP). Additionally, control of the
3 (Supplementary Section 1). The successful formation of aligned flow rate, CNT concentration and surfactant concentration were
films made from this last suspension indicates that surfactant is not crucial. Based on these observations, we propose that CNT alignment
a crucial element for spontaneous CNT alignment, as long as the occurs in a two-dimensional manner. A hydrophilic PVP coating
CNTs are well dispersed in the suspension. Note, however, that the makes the filter membrane surface negatively charged during fil-
CNTs in all these suspensions were negatively charged and so it is cur- tration and, as a result, negatively charged CNTs in the suspension
rently not clear whether CNT alignment can be achieved with posi- are repelled from the surface. At the same time, CNTs feel van der
tively charged or neutral surfactants. Also listed in Table 1 are the Waals attraction from uncoated regions of the membrane surface.
nematic order parameters STHz and SRaman , determined through tera- The competition between these two forces creates a potential
hertz and Raman measurements, respectively. Only SRaman is shown minimum near the surface, where CNTs accumulate, interact with
for suspensions 5, 6 and 7 because the films from these suspensions each other, and form an ordered two-dimensional phase12. Because
did not have sufficiently high carrier densities to show strong the formation of an ordered structure requires horizontal (that is,
enough terahertz attenuation to determine STHz. No obvious relation- in-plane) rotation of CNTs in a finite time period to arrange them-
ship is observed between the structure parameters of SWCNTs and the selves within the two-dimensional layer, an appropriate filtration
achieved values of STHz and SRaman. Arc-discharge SWCNTs tend to speed and CNT concentration are important. Furthermore, the sur-
align more strongly than other types of nanotube, but further systema- factant concentration affects the charge density on the PVP layer,
tic studies using diameter- and length-sorted samples are needed to which in turn influences the electrostatic repulsion potential, while
clarify whether this difference comes from differences in dt , lt or dt/lt. at the same time the suspension viscosity depends on the surfactant
concentration, which influences the rotational motion of CNTs in
Alignment mechanism the two-dimensional layer. A more detailed discussion on the influ-
The strikingly high values of S achieved preclude the possibility that ences of all critical factors on CNT alignment is provided in
the alignment mechanism is based on three-dimensional nematic Supplementary Section 4.

NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology 635

© 2016 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44

Table 1 | Different types of CNT suspension used for making aligned films.
Suspension Synthesis Surfactant d t (nm) l t (nm) S THz S Raman
1 Arc-discharge DOC 1.4 227 ∼1 0.96
2 Arc-discharge SDBS 1.4 246 ∼1 0.94
3 Arc-discharge None* 1.4 295 0.77 0.72
4 CVD† DOC 1.8 307 0.9 0.85
5 CoMoCAT DOC ∼1 166 N/A‡ 0.73
6 CoMoCAT§ DOC 0.73 420 N/A‡ 0.75
7 HiPco|| DOC ∼0.9 298 N/A‡ 0.72
dt , average tube diameter; lt , average tube length; STHz (SRaman), nematic order parameter obtained from terahertz (Raman) measurements; DOC, sodium deoxycholate; SDBS, sodium dodecylbenzenesulfonate.
*Suspension 3 contained arc-discharge-synthesized CNTs that were functionalized by polyaminobenzene sulfonic (PABS) acid; †suspension 4 contained TUBALL nanotubes from OCSiAl (http://ocsial.com/en/
product/tuball/); ‡suspensions 5, 6 and 7 did not have high enough carrier densities to show strong enough terahertz attenuation to determine STHz; §suspension 6 was enriched in (6,5) SWCNTs; ||suspension 7
was enriched in (7,6), (8,6) and (10,5) SWCNTs. N/A, not applicable.

Two-dimensional surface confinement is a novel route towards the surfactant concentration is low. Second, the cross-sectional areal
fabricating ordered nanostructures12,20,28,29. In the Langmuir– density is as high as 1 × 106 nanotubes μm–2, orders of magnitude
Schaefer method20, for example, nano-objects initially float on the larger than any previous reports12,20,28,29. These two advantages indi-
surface of water, confined at the interface between air and the cate that a transition from two-dimensional-like to three-dimen-
water, and eventually form an ordered phase. However, the CNT con- sional-like ordering occurs as CNTs gradually accumulate on the
centrations used in previous studies were extremely low (for example, surface. Once there is an aligned layer, the CNTs that follow tend to
0.1–1 μg ml–1; ref. 12). The obtained films of aligned CNTs were align along the already existing alignment direction. Finally, there is,
small12 and thin (limited to a monolayer12 or several layers20), and in principle, no limitation to the area of aligned films that is achievable.
the degree of alignment was low20 (for example, with a Raman Practically, the only limitation is the size of the filter membrane.
G-peak intensity anisotropy of ∼10, compared to ∼160 in our
case). In addition to having stronger alignment, our films have other Devices based on aligned CNT films
distinct advantages. First, unlike the case in previous two-dimensional With the ultimate aim of developing optoelectronic technologies
ordering studies, we can continue accumulating aligned CNTs to based on aligned single-chirality SWCNTs, we used CNT suspen-
achieve optically thick films (∼100 nm), and the fact that we obtain sions enriched in specific types and chiralities when making
S values higher than the rigid-rod limit to such thicknesses suggests aligned CNT films (for example, suspension 6 in Table 1) and fabri-
that some attractive interaction is present between the tubes when cated light emission and detection devices (Fig. 3). Recent advances

a b 6,000 c 0.7
Excitation photon energy (eV)

1.8 (9,1)
0.6 E11(6,5)
2.0 0.5
Absorbance

4,000
0.4
2.2
E11(9,1)
E11(8,4)

(8,4) 0.3
E22(6,5)
2.4 2,000
(6,5) 0.2
2.6 0.1
1 cm 0.0
0
0.9 1.0 1.1 1.2 1.3 1.2 1.6 2.0 2.4
Emission photon energy (eV) Energy (eV)

f
d 5 e
λexc = 570 nm 90°
Parallel 120° 60°
4
Perpendicular
Vol
Intensity (a.u.)

(6,5) 150° 30° tme

3 ter
5
3 4
(8,4) 180° 1 2 0° g 0.35
2
Voltage (mV)

0.30
1 210° 330°
0.25
240° 300° 0.20
0.9 1.0 1.1 1.2 1.3 270°
0.15
Energy (eV)
0 90 180 270 360
Angle (°)

Figure 3 | Optoelectronic devices made from aligned and (6,5)-enriched CNT films. a, Photographs of a (6,5)-enriched suspension (left) and a fabricated
large-area aligned film (right). b,c, Photoluminescence excitation (b) and optical absorption (c) spectra of the (6,5)-enriched suspension. d, Strongly
polarized photoluminescence spectra for an aligned and (6,5)-enriched CNT film, showing stronger emission polarized parallel to the CNT alignment
direction. e, Polar plot of the intensity of the emitted photoluminescence from the aligned and (6,5)-enriched CNT film. f, Schematic diagram showing the
photodetector device fabricated from an aligned and (6,5)-enriched CNT film. g, Polarization-dependent photovoltage observed for the device in f, with an
improved responsivity of ∼0.1 V W–1 compared with a structurally similar device made of unsorted SWCNTs33.

636 NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology

© 2016 Macmillan Publishers Limited. All rights reserved

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44 ARTICLES
a b 4

σ⊥~ 42 S cm−1
2

1 mm

V (V)
0

Parallel V
Perpendicular σ// ~ 2,500 S cm−1
−2
I

−4

−100 −50 0 50 100

I (μA)

c d e 10−5
10 μm 100 (6,5) Parallel
Parallel

CNT alignment
~1 μm2 (6,5) Perpendicular
50 10−7 Semi-arc parallel

Ids (pA)

Ids (A)
0 10−9

−50 Perpendicular
10−11
−100
10−13
−1.0 −0.5 0.0 0.5 1.0 −50 0 50
Vds (V) Vg (V)

Figure 4 | Electronic devices made from aligned CNT films. a, Photograph and schematic diagrams of Hall-bar devices with channel length of ∼5 mm
and channel width of ∼0.5 mm, used to characterize the macroscopic anisotropic transport of charge carriers in an aligned film consisting of unsorted
arc-discharge CNTs via four-terminal measurements. b, Voltage–current relationship of Hall-bar devices when the current flow is parallel and perpendicular
to the CNT alignment direction, showing extremely anisotropic conductivities. c, False-colour SEM images of a thin-film transistor with a channel width of
∼5 μm and channel length of ∼30 μm made from an aligned and (6,5)-enriched CNT film. d, Source–drain current versus source–drain voltage at zero gate
voltage of the transistor, showing anisotropic conductivities of the aligned (6,5)-enriched thin-film transistor. e, Source–drain current at a source–drain
voltage of 1 V versus gate voltage of the (6,5)-enriched transistor showing large and anisotropic transistor action. The on-current density is enhanced by a
factor of 50 in a transistor made from larger-diameter semiconductor-enriched arc-discharge SWCNTs.

in postgrowth separation and sorting techniques allow us to prepare proper thermal management through the selection and adjustment
suspensions of specific chirality, (n,m), SWCNTs in large enough of the substrates and other critical elements, the detector performance
quantities to make macroscopic films. Here, we used the aqueous can be further improved34.
two-phase extraction method30–32 to enrich (6,5) SWCNTs. Finally, we fabricated electronic devices with aligned CNT films
Figure 3a shows photographs of the prepared suspension and film, using standard microfabrication techniques and tested their conduc-
and Fig. 3b,c presents the optical absorption and photoluminescence tivities and transistor performance (Fig. 4). First, we observed strong
excitation spectra of the enriched suspension. There are trace conductivity anisotropy in Hall-bar devices made of aligned unsorted
amounts of (8,4) and (9,1) SWCNTs in the suspension, but it is arc-discharge CNTs at room temperature (Fig. 4a,b). The ratio of the
clear that the majority of the nanotubes are (6,5) species. The fabri- conductivity between the parallel and perpendicular directions was as
cated (6,5) film shows strongly polarized photoluminescence (Fig. 3d, high as 60 (Fig. 4b). The conductivity along the alignment direction
e). This macroscopic manifestation of light-emission anisotropy not was 2,500 S cm–1, a value that is much higher than the typical conduc-
only provides evidence that the CNTs in the film are highly aligned, tivities reported for aligned films made from CVD-grown CNTs, which
but also proves that there are very few residual metallic tubes, which show only ∼10 S cm–1 along the alignment direction35 because of their
are known to be efficient photoluminescence quenchers when in low densities. Furthermore, our films are more conducting than pre-
contact with semiconducting nanotubes. This result opens up new viously reported dense random-network films, whose conductivities
possibilities of developing CNT-based light sources for producing are typically on the order of hundreds of S cm–1 without intentional
polarized monochromatic radiation. doping36,37, while our films are not intentionally doped.
Our photodetector was a two-terminal device (see Methods and We also investigated the transistor behaviour of a (6,5)-enriched
Fig. 3f), which exhibited a strongly polarization-sensitive photosignal aligned film (Fig. 4c). As shown in Fig. 4d,e, the on-current density
(Fig. 3g). The polarization ratio of the photosignal magnitude was of the transistor in the parallel (perpendicular) direction is
∼2:1. The responsivity of the photodetector was estimated to be ∼2 nA μm–1 (∼80 pA μm–1), indicating that the on-current density
∼0.1 V W–1, after taking into account the actual power absorption can be improved by aligning the CNTs in one direction. Note that
of ∼3.1 mW when the light polarization was parallel to the CNT these nearly intrinsic semiconducting films are naturally much less
alignment direction. Compared to previous photodetectors with the conducting than purely metallic films or films with mixed electronic
same device architecture but consisting of a mixture of metallic types. As shown previously38, the on-current density can also be
and semiconducting SWCNTs33, the responsivity of the current enhanced by using larger-diameter nanotubes, which is also demon-
photodectector has been enhanced by at least three times, most prob- strated by our transistor based on semiconductor-enriched arc-dis-
ably due to the removal of metallic tubes. It should be noted, however, charge CNTs with an average diameter of 1.4 nm (Fig. 4e). The
that the current device design has not been fully optimized. With device shows an enhancement of on-current density by ∼50 times

NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology 637

© 2016 Macmillan Publishers Limited. All rights reserved

ARTICLES NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44

compared to the (6,5) CNT transistor at the same drain–source 19. Beyer, S. T. & Walus, K. Controlled orientation and alignment in films of
voltage. The on–off ratios of our transistors are around 1 × 103, single-walled carbon nanotubes using inkjet printing. Langmuir 28,
8753–8759 (2012).
which is comparable to previous results for transistors made of 20. Cao, Q. et al. Arrays of single-walled carbon nanotubes with full surface coverage
aligned CNT films. The relatively low on–off ratio is ascribed to the for high-performance electronics. Nature Nanotech. 8, 180–186 (2013).
charge screening effect caused by the high packing of CNTs in one 21. Wu, Z. et al. Transparent, conductive nanotube films. Science 305,
direction20. It can be potentially overcome by using a high-κ dielectric 1273–1276 (2004).
material or using a top-gate architecture. 22. Zhang, Q. et al. Plasmonic nature of the terahertz conductivity peak in single-
wall carbon nanotubes. Nano Lett. 13, 5991–5996 (2013).
23. Ajiki, H. & Ando, T. Aharonov–Bohm effect in carbon nanotubes. Physica B
Conclusions 201, 349–352 (1994).
24. Miyauchi, Y., Oba, M. & Maruyama, S. Cross-polarized optical absorption of
We have developed a new process to produce a wafer-scale (that is, single-walled nanotubes by polarized photoluminescence excitation
inch-size) film of aligned CNTs with adjustable thickness. This spectroscopy. Phys. Rev. B 74, 205440 (2006).
method works for CNTs synthesized by various methods. The 25. Lefebvre, J. & Finnie, P. Polarized photoluminescence excitation spectroscopy of
produced large-area films will enable a range of new opportunities single-walled carbon nanotubes. Phys. Rev. Lett. 98, 167406 (2007).
not only in fundamental research in physics, chemistry and 26. Ren, L. et al. Carbon nanotube terahertz polarizer. Nano Lett. 9,
2610–2613 (2009).
materials science, but also applications in electronics, optoelectro- 27. Onsager, L. The effects of shape on the interaction of colloidal particles. Ann. NY
nics, sensing, imaging and medicine. We have already demonstrated Acad. Sci. 51, 627–659 (1949).
terahertz/infrared polarizers using metal–semiconductor mixed 28. Yang, P. Nanotechnology: wires on water. Nature 425, 243–244 (2003).
CNTs, thin-film transistors, polarized light-emission devices and 29. Sharma, R., Lee, C. Y., Choi, J. H., Chen, K. & Strano, M. S. Nanometer
positioning, parallel alignment, and placement of single anisotropic
polarization-sensitive photodetectors using semiconducting CNTs.
nanoparticles using hydrodynamic forces in cylindrical droplets. Nano Lett. 7,
2693–2700 (2007).
Methods 30. Khripin, C. Y., Fagan, J. A. & Zheng, M. Spontaneous partition of carbon
nanotubes in polymer-modified aqueous phases. J. Am. Chem. Soc. 135,
Methods and any associated references are available in the online 6822–6825 (2013).
version of the paper. 31. Subbaiyan, N. K. et al. Role of surfactants and salt in aqueous two-phase
separation of carbon nanotubes toward simple chirality isolation. ACS Nano 8,
Received 25 July 2015; accepted 19 February 2016; 1619–1628 (2014).
32. Fagan, J. A. et al. Isolation of specific small-diameter single-wall carbon
published online 4 April 2016 nanotube species via aqueous two-phase extraction. Adv. Mater. 26,
2800–2804 (2014).
References 33. Nanot, S. et al. Broadband, polarization-sensitive photodetector based on
1. Jorio, A., Dresselhaus, G. & Dresselhaus, M. S. (eds) Carbon Nanotubes: optically-thick films of macroscopically long, dense, and aligned carbon
Advanced Topics in the Synthesis, Structure, Properties and Applications nanotubes. Sci. Rep. 3, 1335 (2013).
(Springer, 2008). 34. He, X. et al. Photothermoelectric p–n junction photodetectors with intrinsic
2. Ma, Y., Wang, B., Wu, Y., Huang, Y. & Chen, Y. The production of polarimetry based on macroscopic carbon nanotube films. ACS Nano 7,
horizontally aligned single-walled carbon nanotubes. Carbon 49, 7271–7277 (2013).
4098–4110 (2011). 35. Jakubinek, M. B. et al. Thermal and electrical conductivity of tall, vertically
3. Liu, B., Wang, C., Liu, J., Che, Y. & Zhou, C. Aligned carbon nanotubes: from aligned carbon nanotube arrays. Carbon 48, 3947–3952 (2010).
controlled synthesis to electronic applications. Nanoscale 5, 9483–9502 (2013). 36. Bekyarova, E. et al. Electronic properties of single-walled carbon nanotube
4. De Heer, W. A. et al. Aligned carbon nanotube films: production and optical and networks. J. Am. Chem. Soc. 127, 5990–5995 (2005).
electronic properties. Science 268, 845–847 (1995). 37. Zhang, D. et al. Transparent, conductive, and flexible carbon nanotube
5. Murakami, Y. et al. Growth of vertically aligned single-walled carbon nanotube films and their application in organic light-emitting diodes. Nano Lett. 6,
films on quartz substrates and their optical anisotropy. Chem. Phys. Lett. 385, 1880–1886 (2006).
298–303 (2004). 38. Chen, Z., Appenzeller, J., Knoch, J., Lin, Y. & Avouris, P. The role of metal–
6. Pint, C. L., Xu, Y., Pasquali, M. & Hauge, R. H. Formation of highly dense, nanotube contact in the performance of carbon nanotube field-effect transistors.
aligned ribbons and ultra-thin films of single-walled carbon nanotubes from Nano Lett. 5, 1497–1502 (2005).
carpets. ACS Nano 2, 1871–1878 (2008).
7. Lynch, M. D. & Patrick, D. L. Organizing carbon nanotubes with liquid crystals. Acknowledgements
Nano Lett. 2, 1197–1201 (2002). This work was supported by the Basic Energy Sciences (BES) programme of the US
8. Lagerwall, J. P. F. & Scalia, G. Carbon nanotubes in liquid crystals. J. Mater. Department of Energy through grant no. DE-FG02-06ER46308 (for the preparation and
Chem. 18, 2890–2898 (2008). characterization of aligned carbon nanotube films) and the Robert A. Welch Foundation
9. Zamora-Ledezma, C. et al. Anisotropic thin films of single-wall carbon through grant no. C-1509 (for terahertz and infrared characterization). S.K.D. and E.H.H.
nanotubes from aligned lyotropic nematic suspensions. Nano Lett. 8, acknowledge support from the LANL LDRD programme. Portions of this work were
4103–4107 (2008). performed at the Center for Integrated Nanotechnologies, a US Department of Energy,
10. Kim, Y., Minami, N. & Kazaoui, S. Highly polarized absorption and Office of Science user facility. The authors thank H. Kasai, A. Zubair, C. Sewell, S. Peters and
photoluminescence of stretch-aligned single-wall carbon nanotubes dispersed in T. Higashira for their assistance with terahertz characterization measurements and
gelatin films. Appl. Phys. Lett. 86, 073103 (2005). I. Kurganskaya, A. Lüttge, R. Headrick and M. Pasquali for discussions.
11. Shaver, J. et al. Magnetic brightening of carbon nanotube photoluminescence
through symmetry breaking. Nano Lett. 7, 1851–1855 (2007). Author contributions
12. Mclean, R. S., Huang, X., Khripin, C., Jagota, A. & Zheng, M. Controlled X.H. and W.G. developed the process of making aligned CNT films, performed most
two-dimensional pattern of spontaneously aligned carbon nanotubes. Nano of the characterization experiments, and analysed the data obtained, under the supervision
Lett. 6, 55–60 (2006). and guidance of W.W.A., R.H.H. and J.K. L.X. prepared chirality-enriched SWCNT
13. Dan, B., Ma, A. W. K., Hároz, E. H., Kono, J. & Pasquali, M. Nematic-like suspensions, in collaboration with E.H.H. and S.K.D. B.L. made the AFM measurements
alignment in SWNT thin films from aqueous colloidal suspensions. Ind. Eng. and B.L. and W.W. performed TEM imaging, under the advisement of R.V. and
Chem. Res. 51, 10232–10237 (2012). P.M.A. Q.Z. and J.M.R. participated in the terahertz and infrared spectroscopy
14. Oh, J. Y. et al. Easy preparation of self-assembled high-density buckypaper with measurements. S.L. helped characterize the FET devices. X.H., W.G. and J.K. wrote
enhanced mechanical properties. Nano Lett. 15, 190–197 (2015). the manuscript. All authors discussed the results and commented on the manuscript.
15. Puech, N. et al. Highly ordered carbon nanotube nematic liquid crystals. J. Phys.
Chem. C 115, 3272–3278 (2011). Additional information
16. Zakri, C. et al. Liquid crystals of carbon nanotubes and graphene. Phil. Trans. R. Supplementary information is available in the online version of the paper. Reprints and
Soc. A 371, 20120499 (2013). permissions information is available online at www.nature.com/reprints. Correspondence and
17. Walters, D. A. et al. In-plane-aligned membranes of carbon nanotubes. Chem. requests for materials should be addressed to J.K.
Phys. Lett. 338, 14–20 (2001).
18. Zaric, S. et al. Estimation of magnetic susceptibility anisotropy of carbon Competing financial interests
nanotubes using magneto-photoluminescence. Nano Lett. 4, 2219–2221 (2004). The authors declare no competing financial interests.

638 NATURE NANOTECHNOLOGY | VOL 11 | JULY 2016 | www.nature.com/naturenanotechnology

© 2016 Macmillan Publishers Limited. All rights reserved

NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2016.44 ARTICLES
Methods spectroscopy experiments did not detect any residual surfactants (Supplementary
CNT suspension preparation. Two batches of arc-discharge SWCNTs (P2-SWNT Fig. 10). The properties of the fabricated CNT films were characterized by various
and P8-SWNT) with an average diameter of 1.4 nm were purchased from Carbon methods. Anisotropic optical transmission on a large scale was characterized by
Solutions. Two types of CoMoCAT SWCNT (CG 200 and SG 65i) were purchased polarized optical microscopy in both co-polarized and cross-polarized
from Sigma-Aldrich. The former had a wide diameter distribution (0.7–1.4 nm), configurations. The alignment structure (on micro- and nanometre scales) was
whereas the latter was enriched in (6,5) SWCNTs with ∼90% purity. CVD-grown examined by scanning electron microscopy (JEOL 6500F scanning electron
TUBALL CNTs, which were 75% SWCNTs, were obtained from OCSiAl and had an microscope) and transmission electron microscopy (JEOL 2100 field emission gun
average diameter of ∼1.8 nm. HiPco SWCNTs (batch #195.5), with a diameter range transmission electron microscope). The quality of CNT alignment was characterized
of 0.9–1 nm, were used after purification and diameter sorting. Sodium in terms of local and global nematic order parameters based on polarized Raman
deoxycholate (DOC, Sigma-Aldrich) and sodium dodecylbenzenesulfonate (SDBS, spectroscopy (Renishaw inVia Raman microscope) and terahertz transmission
Sigma-Aldrich) were used as surfactants for CNT dispersion. All CNTs (except for spectroscopy (a home-made time-domain terahertz spectroscopy system). Film
P8-SWNT) were initially dispersed in either 1% (wt/vol) DOC or 0.4% (wt/vol) thicknesses were measured by atomic force microscopy (Bruker Multimode 8). The
SDBS by bath sonication (Cole-Parmer 60 W ultrasonic cleaner, model #08849-00) average lengths of CNTs in suspension were measured by atomic force microscopy
for 5–10 min at a starting concentration of 0.2–0.8 mg ml–1. The obtained (Supplementary Fig. 8). Photoluminescence excitation spectra of the single-chirality
suspension was further sonicated with a tip sonicator (XL-2000 Sonicator, Qsonica, (6,5) CNT suspension and the polarized photoluminescence of the single-chirality
1/4-inch probe, ∼30 W) for 45–60 min. The suspension was cooled in an ice-water (6,5) films were measured with a home-made photoluminescence excitation
bath during sonication. The suspension was then centrifuged for 1 h at 38,000 r.p.m. spectroscopy set-up.
(Sorvall Discovery 100SE Ultracentrifuge using a Beckman SW-41 Ti swing bucket
rotor) to remove large bundles of CNTs. After centrifugation, the upper 60% of the Device fabrication. For the photodetector, aligned (6,5) nanotube films were first
supernatant was collected and then diluted with Nanopure water. The concentration cut into ribbons and then transferred to glass substrates. Two electrodes were
of surfactant was diluted below its critical micelle concentration (CMC). Depending formed by sputtering 50 nm gold with a shadow mask at the two ends of a ribbon.
on the type of CNTs used, the final concentration of DOC before film-making varied A 660 nm laser beam with a maximum power of ∼50 mW was incident on and
from 0.02 to 0.1%, and the final concentration of SDBS was ∼0.02%. near a gold–CNT junction. A half-wave plate was used to rotate the polarization of
the incident light beam without changing the incident power. The generated
Vacuum filtration. The well-dispersed CNT suspension with dilute CNTs photovoltage was amplified by a Stanford Research SR560 voltage preamplifier and
(∼1–15 μg ml–1) and dilute surfactant (below the CMC) was filtered though a then fed into a Stanford Research SR830 lock-in amplifier. For the Hall-bar device
vacuum filtration system in a well-controlled manner. Polycarbonate filter and field-effect transistor, an aligned nanotube film was transferred onto a
membranes (Nuclepore track-etched polycarbonate hydrophilic membranes) with substrate composed of heavily doped silicon (acting as a global backgate for the
different pore sizes from 50 to 200 nm were used. The filtration speed was kept low field-effect transistors) and a 285-nm-thick layer of thermal silicon oxide (acting as
and controlled in the range of 1–2 ml h–1. The pressure applied to the system by the insulating layer). The first photolithography step was to define the electrode
vacuum pumping was monitored by sensitive pressure gauges. A speeding-up area with Shipley Microposit S1813 photoresist. The electrodes were defined and
procedure was performed at the end of the filtration process to quickly dry the film. fabricated by lifting off electron-beam-evaporated titanium (1 nm)/palladium
Before transferring the film onto another substrate, the film was pumped on for an (10 nm)/gold (20 nm). The next step was to define a Hall bar structure with one
additional 15–30 min. channel parallel and the others perpendicular to the CNT alignment direction, as
shown in Fig. 4a,c. Anisotropic transport measurements of the aligned film
Transfer and characterization. For characterization and device fabrication, films consisting of unsorted arc-discharge CNTs were performed under ambient
were transferred onto solid substrates (for example, SiO2/Si and quartz) using a wet conditions via four-terminal sensing using a Keithley 2400 source meter. Field-
transfer process13, in which the polycarbonate filter membrane was dissolved away effect transistor measurements were performed under vacuum (∼1 × 10–5 torr),
with organic solvent (N-methyl-2-pyrrolidone or chloroform). The films were then using a Keithley 2634B source meter (for the source–drain voltage) and a Keithley
thoroughly washed by acetone and Nanopure water; subsequent X-ray photoelectron 2400 source meter (for the gate voltage).

NATURE NANOTECHNOLOGY | www.nature.com/naturenanotechnology

© 2016 Macmillan Publishers Limited. All rights reserved

Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.