This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNANO.2019.2943476, IEEE
Transactions on Nanotechnology

Realizing Controllable Carbon Nanotube Arrays for

Soft Devices by Asymmetric Langmuir-Blodgett and
Hydrophilization Process

Qinghua Wang, Student Member, Min Zhang*, Member, IEEE, Qiuyue Huang, Student Member,
Zhiqiang Liao, and Chunhui Du

 Well-aligned CNT arrays are more promising for realizing high

Abstract—Well-aligned Carbon nanotube (CNT) arrays are
expected by flexible and stretchable devices. In this paper, an
asymmetric Langmuir-Blodgett (LB) method and inorganic
hydrophilization method have been proposed, with which
well-aligned CNT arrays on soft substrates like polyethylene
naphthalate, polyimide and polydimethylsiloxane have been
realized. After depositing 50 nm SiO2 for hydrophilization, the
water contact angle has been significantly reduced for different
substrates. The surface roughness and flexibility have also been
characterized to analyze their effect on device performance. The
nanotube density in the arrays can be controlled to 10-25
tubes/µm2, 25-35 tubes/µm2 or 35-45 tubes/µm2 by tuning the
surface pressure during the LB process. With aligned
semiconducting CNT arrays as channel and anisotropic metallic
CNT network as gate/source/drain electrodes, all-carbon
nanotube transistors on PI substrate have been fabricated for
demonstration. The fabricated devices show a subthreshold swing
of 92.9 mV/dec, a carrier mobility of 15.32 cm2/Vs, and an on/off
current ratio of 4.51×102. More importantly, with 50 nm thick
gate dielectrics, the changes of threshold voltage and carrier
mobility are within 16.62% and 28.78%, respectively, even after
bending 1000 times under a curvature radius of 3.66 mm.
Index Terms—Carbon nanotubes, Flexible
Langmuir-Blodgett method, Hydrophilization
I. INTRODUCTION
performance TFTs and accurate device mechanism analysis.
Langmuir–Blodgett (LB) method [5] is commonly used for its
convenience and controllability. However, the CNT arrays by
the traditional LB method with direct compression and dipping
still suffer from misalignments and low tube density. Some
improved methods were reported. X. Li et al. proposed a
method of compression-expansion cycles to achieve high
quality CNT arrays by reducing hysteresis [6]. Q. Cao et al.
proposed Langmuir-Schaefer method to obtain high density
CNT arrays on hard substrate [7]. Even though, controlling the
carbon nanotube density remains challenging, which is critical
for analyzing the device performance. Besides, realizing CNT
arrays on flexible or stretchable substrates is seldom reported.
The main reason is that soft substrates like Polyethylene
naphthalate (PEN), polyimide (PI) and polydimethylsiloxane
(PDMS) normally suffer from poor hydrophilicity. Therefore,
hydrophilization of the substrates is essential to enable CNTs to
be attached to the substrate tightly during the alignment process.
Some hydrophilization methods like oxygen plasma treatment
[8], UV/ozone exposure [9] and depositing an adhesion layer
transistors, [10] were reported. Oxygen plasma treatment and UV/ozone
exposure treatment can improve the hydrophilicity a lot, but the
hydrophobicity will be recovered after a short period of time.
When adopting an organic adhesion layer, the layer becomes a

C ARBON nanotubes (CNTs) have been considered as one of

the most promising candidates for flexible and stretchable
devices because of the excellent mechanical and electrical
properties [1-3]. All-carbon thin-film transistors (TFTs) with
metallic CNTs (MCNTs) as gate/source/drain electrode and
semiconducting CNTs (SCNTs) as channel have wide
applications on wearable system and display fields due to the
high transparency and mechanical flexibility. The methods
widely used for fabricating CNT devices such as spin-coating,
ink-jet printing, dip-coating etc. usually yield CNT networks,
which induce large tube-to-tube contact resistance [4].
Qinghua Wang, Min Zhang, Qiuyue Huang, Zhiqing liao and Chunhui Du
are with the Department of School of Electronic and Computer Engineering,
Peking University Shenzhen Graduate School, Shenzhen, China. (*e-mail:
zhangm@ece.pku.edu.cn )
This work was supported by Shenzhen Science and Technology Innovation
Grants JCYJ20180507181702150 and ZDSYS201802061805105, and the
National Natural Science Foundation of China 61504004.
part of the device, which introduces influence to device
performance and increases complexity when analyzing device
mechanism. Therefore, a fast, permanent, and effective
hydrophilization method with less effect on device is required
by flexible and stretchable devices.
In this work, an asymmetric LB method with compression-
expansion cycles and an inorganic hydrophilization method by
depositing a thin SiO2 layer have been proposed. By combining
the two methods, well-aligned CNT arrays have been realized
on flexible and stretchable substrates, such as PEN, PI, and
PDMS. The tube density of CNT arrays can also be controlled
in a specific range. The effects of the SiO2 thickness on
flexibility of the CNT array films and the surface roughness are
also investigated. Besides, all-carbon TFTs on a PI substrate
have also been realized with 99.9% purity SCNT arrays as
channel, MCNTs as source/drain/gate electrodes, and Al2O3 as
gate dielectrics. The devices show excellent electrical

1536-125X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1536-125X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNANO.2019.2943476, IEEE
Transactions on Nanotechnology

performance, with carrier mobility of 15.32 cm2/Vs and the

maximum transconductance (gm) of -7.05×10-9 S. The devices
can endure bending of 1000 times under a curvature radius of
3.66 mm.

II. EXPERIMENTS FOR ARRAY ALIGNMENT

At the beginning, three substrates, including PEN, PI, and
PDMS were cleaned by deionized water and blown dried by
Nitrogen. Then a SiO2 layer was deposited by Plasma
Enhanced Chemical Vapor Deposition (PECVD) with gas flow
of 100 sccm SiH4 and 400 sccm N2O at 150 ℃.
The single-walled carbon nanotube powders with different
purity were purchased from NanoIntegris Inc. The CNTs were
dispersed into 1,2-dichloroethane with poly (m-phenylene Fig. 2. AFM images after depositing SiO2 layer on PEN substrate, (a) 20 nm,
vinylene-co-2,5-dioctoxy-p-phenylenevinylene) as dispersant. (b) 30 nm, (c) 40 nm, (d) 50 nm.
After sonicating for 7 hours, the CNTs were dispersed in the The water contact angle has changed a lot after depositing a
solution uniformly with a concentration of 0.02 mg/mL. A SiO2 layer, and the hydrophilicity is different for different SiO2
KSV- Minimicro 2000 LB trough was used for LB process. The thickness. Atomic Force Microscope (AFM) was applied to
surface pressure was increased as the two barriers moved characterize the roughness of SiO2 surface with 20 nm, 30 nm,
towards each other, while it would be decreased with two 40 nm, and 50 nm SiO2 thickness, as shown in Fig. 2. From Fig.
barriers moving backwards. Fig. 1 (a) illustrates the change of 2, their average roughness of SiO2 surface is 0.78 nm, 0.86 nm,
surface pressure during the asymmetric compression LB 0.82 nm, 1.32 nm, respectively. With the thickness of SiO2
process. It can be seen that the surface pressure increased from increasing, the surface roughness of the substrates tends to
12 mN/m to 30 mN/m with a step of 2 mN/m. The compression increase and then the hydrophilicity becomes better.
force made the CNTs self-oriented as the two barriers moving The scanning electron microscopy (SEM) images in Fig. 3
towards each other. Then the two barriers moved backwards compare morphology of CNT arrays obtained by two different
and the surface pressure started to be decreased after being LB methods, direct compression and asymmetric
increased by 5 mN/m. After the surface pressure was decreased compression-expansion. Comparing the two SEM images, it
by 3 mN/m, the two barriers would move towards each other can be seen that both of the alignment and tube density have
again and began a new cycle. From the experiments, the cycles been improved a lot with asymmetric compression-expansion
of the compression-expansion can reduce misalignments
effectively and increase the tube density. The LB process in Fig.
1 (a) went through eight compression- expansion cycles in total.
When the surface pressure reached to 30 mN/m, the CNT array
was transferred to the substrate with a speed of 1 mm/min.
Finally, the CNT array was transferred to the target substrate
and baked for 1 hour at 150 ℃. Fig. 3. (a) SEM image of CNT array obtained by direct compression and (b)
asymmetric LB method.

Fig. 1. (a) Change of the surface pressure during the asymmetric compression
LB process. (b) Change of water contact angle of PEN, PI, PDMS substrates
after depositing SiO2 layer.

III. RESULTS AND DISCUSSION

Fig. 1 (b) compares the change of water contact angel after
depositing 20 nm, 30 nm, 40 nm, 50 nm SiO2 on PEN, PI,
PDMS substrates, respectively. It is obvious that the water
contact angle decreases as the thickness of SiO2 increases. From
the figure, the water contact angles of PEN, PI, and PDMS
substrate change from 85.45°, 104.11°, 111.94° to 31.5°, Fig. 4. SEM images of CNT arrays on (a) PEN (b) PI (c) PDMS (d) glass
29.05°, 33.96°, respectively, after depositing 50 nm SiO2. substrate.

1536-125X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNANO.2019.2943476, IEEE
Transactions on Nanotechnology

Fig. 5. SEM images of low tube density of CNT arrays. The surface pressure is
(a) 10 mN/m, (b) 20 mN/m.

cycles. By this novel LB process, well-aligned CNT arrays

have been prepared on PEN, PI, PDMS and glass, as shown in Fig. 7. Process flow of the all-carbon device on PI/PEN substrate.
Fig. 4, and the tube density is approximately 35-45 tubes/µm2.
flow on PI substrate are shown in Fig. 7. Firstly, a Mo layer was
The tube density can be controlled by simply modulating the
sputtered on the PEN substrate as a reference for lithography
target surface pressure. The SEM images of the CNT arrays
and the PI tape was stuck to the PEN/Mo surface. Then, a 40
with target surface pressure of 10 mN/m and 20 mN/m are
nm SiO2 layer by PECVD at 150 ℃ was deposited for
shown in Fig. 5 (a) and Fig. 5 (b), respectively. The tube
hydrophilization. Thirdly, MCNTs with metallic purity of 95%
densities of CNT arrays with target surface pressure of 10
were applied as source/drain electrodes. The MCNTs have an
mN/m and 20 mN/m are approximately 10-25 tubes/µm2 and
average length of 15 µm, with the inner diameter of 4 nm and
25-35 tubes/µm2, respectively.
outer diameter of less than 8 nm. Then, the SCNTs with a purity
of 99.9% were aligned by asymmetric LB method to work as
active channel. The SCNTs have a diameter of 1.4 nm and a
length of around 1 µm. Both the SCNTs and MCNTs were
purchased from NanoIntegris Inc. The source/drain and channel
were defined by photolithography, and patterned by O2/N2
plasma etching. Then, 33-nm-thick Al2O3 was formed by
Atomic Layer Deposition (ALD) as gate dielectrics. Finally,
Fig. 6. (a) The change of equivalent resistance with increasing bending cycles gate electrode was MCNTs and patterned by photolithography,
at a curvature radius of 1.7 mm of (a) PEN substrate (b) PI substrate (c) The and the useless CNTs were removed by O2/N2 plasma. In order
change of equivalent resistance with increasing stretching cycles after
stretching to 120% of PDMS substrate.
to expose the source/drain electrodes, the Al2O3 on the
source/drain electrodes was etched away with H3PO4.
Depositing SiO2 layer may induce influence on the flexibility
and stretchability of the aligned-CNT films. Fig. 6 (a) and Fig.
6 (b) show the changes of the equivalent film resistance for the
aligned-CNT films on PEN and PI substrates with the
increasing bending times under the curvature radius of 1.7 mm.
It can be found that the change of equivalent resistance shows
obvious downtrend when depositing 20 nm SiO2 to improve the
hydrophilicity. Besides, the flexibility is decreased with the Fig. 8. The contact between CNT array channel and source/drain electrodes.
increasing of SiO2 thickness. From Fig. 6 (a), the change of the
film resistance is less than 5% after bending 2000 times with 20 The SEM image in Fig. 8 shows the contact between channel
nm SiO2. From Fig. 6 (b), the change of equivalent resistance and electrode. It is clearly that the active region is filled with
with 40 nm SiO2 is within 4%. Fig. 6 (c) shows the change of well-aligned SCNTs and the source/drain electrodes are MCNT
equivalent resistance for CNT array on PDMS substrates with network. The image shows the good alignment of CNT arrays
increasing stretching cycles at a stretching strain of 120%, for in active region with a tube density of 25-35 tubes/µm2.
samples with various SiO2 thickness. From Fig. 6 (c), the Electrical characters were measured by B1500 semiconductor
changes of equivalent resistance increase with the increase of
SiO2 thickness. When depositing 20 nm SiO2 for
hydrophilization, the change of film resistance is within 10%
even after stretching 1000 times. Therefore, the thickness of
SiO2 has significant effect on the performance of the
aligned-CNT arrays. The best thickness is varied for different
substrates.

IV. DEVICE DEMONSTRATION AND DISCUSSION Fig. 9. (a) The transfer characteristics and (b) output characteristics of the
all-carbon nanotube array transistor.
Structure of the proposed all-carbon TFT and the process

1536-125X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TNANO.2019.2943476, IEEE
Transactions on Nanotechnology
Fig. 10. (a) The transfer characteristics of eight TFT devices on the same wafer
with gate dielectric thickness of 33 nm. (b) The change of transfer
characteristics of the TFT with gate dielectric thickness of 33 nm after 1 and
1000 bending cycles, respectively, with a curvature radius of 8 mm. (c) The
transfer characteristics of the TFT device with gate dielectric of 50 nm after
bending 1000 times with a curvature radius of 3.66 mm.
analyzer. The transfer characteristics and output characteristics
of a typical TFT with channel length of 100 µm and channel
width of 50 µm are shown in Fig. 9 (a) and Fig. 9 (b). On/off
current ratio (Ion/Ioff) is 4.51×102. The low on/off current ratio
is due to the existing of MCNTs in active channel, and the
aligned-CNT array makes the MCNTs connected more easily.
Subthreshold swing (SS) is 92.9 mV/dec, and the maximal
transconductance (gm) is -7.05×10-9 S. Then, carrier mobility
(µ) is 15.32 cm2/Vs, calculated by the equation µ=
gmL/(CoxVdW), where Vd=-0.1 V, W=50 µm and L=100 µm.
Fig. 10 (a) shows the transfer characteristics of eight devices on
the same wafer. Threshold voltage (Vth) of eight devices
changes little and the standard deviation is 0.25 V. Fig. 10 (b)
shows the change of transfer characteristics for a device with
100 µm length and 50 µm width after bending 1 and 1000 times,
respectively, under a curvature radius of 8 mm. It can be seen
that the Vth change is larger after the first bending and then the
device performance tends to be stable. The change is induced
by the morphology change of the gate dielectrics during the
bending test. Increasing the gate dielectrics thickness can
improve the bending performance of the TFTs. As shown in Fig.
10 (c), after increasing the Al2O3 dielectric thickness to 50 nm,
the device can bear the bending of 1000 cycles under a
curvature radius as small as 3.66 mm. From Fig. 10 (c), the
change of Vth, SS, µ is within 16.62%, -8.85%, -28.78%,
respectively.
V. CONCLUSION
In conclusion, an inorganic hydrophilization method by
depositing a thin SiO2 layer on flexible/stretchable substrates
and an asymmetric Langmuir-Blodgett method by
compression-expansion cycles have been proposed and
combined to realize the well-aligned CNT arrays on PEN, PI,
and PDMS substrates. Furthermore, the tube density can be
controlled by simply adjusting the target surface pressure with
the asymmetric LB method. The tube density can be controlled
within 10-25 tubes/μm2, 25-35 tubes/μm2, and 35-45 tubes/μm2
by tuning the target surface pressure to 10 mN/m, 20 mN/m,
and 30 mN/m, respectively. The CNT arrays on PEN, PI and
PDMS substrates show excellent flexibility and stretchability.
Based on the array alignment method, the well-aligned-array
all-CNT TFTs on PI substrate have also been demonstrated
with MCNTs as source/drain/gate electrodes, SCNTs array as
channel and Al2O3 as dielectrics. Furthermore, the CNT array
1536-125X (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
all-carbon devices show excellent electrical properties, with SS
of 92.9 mV/dec, gm of -7.05×10-9 S, µ of 15.32 cm2/Vs, and
Ion/Ioff of 4.51×102. Besides, the array all-carbon TFTs in the
same wafer show good uniformity, with a standard deviation of
Vth to be 0.25 V for eight devices. Increasing gate dielectrics
thickness can improve the bending performance of the devices.
With 50 nm gate dielectrics, the device shows excellent
flexibility performance even after bending 1000 cycles under
3.66 mm curvature radius.
REFERENCES
[1] M. Endo, M. S. Strano, and P. M. Ajayan, "Potential
Applications of Carbon Nanotubes," Topics in Applied
Physics, vol. 111, no. 5, pp. 13-61, 2008.
[2] H. Koo et al., "Scalability of carbon-nanotube-based thin
film transistors for flexible electronic devices
manufactured using an all roll-to-roll gravure printing
system," Scientific Reports, vol. 5, 2015, Art. no. 14459.
[3] N. Matsuhisa et al., "Printable elastic conductors with a
high conductivity for electronic textile applications,"
Nature Communications, vol. 6, 2015, Art. no. 7461.
[4] M. Tsukuda, K. Ishizeki, K. Takashima, and T. Yamamoto,
"Random stick network analysis of electronic transport in
carbon nanotube thin films," Applied Physics Express, vol.
12, no. 5, May 2019, Art. no. 055006.
[5] Y. Kim, N. Minami, W. H. Zhu, S. Kazaoui, R. Azumi, and
M. Matsumoto, "Langmuir-Blodgett films of single-wall
carbon nanotubes: Layer-by-layer deposition and in-plane
orientation of tubes," Japanese Journal of Applied Physics
Part 1-Regular Papers Short Notes & Review Papers, vol.
42, no. 12, pp. 7629-7634, Dec. 2003.
[6] X. Li et al., "Langmuir-Blodgett assembly of densely
aligned single-walled carbon nanotubes from bulk
materials," Journal of the American Chemical Society, vol.
129, no. 16, pp. 4890-4891, Apr. 2007.
[7] Q. Cao, S.-j. Han, G. S. Tulevski, Y. Zhu, D. D. Lu, and W.
Haensch, "Arrays of single-walled carbon nanotubes with
full surface coverage for high-performance electronics,"
Nature Nanotechnology, vol. 8, no. 3, pp. 180-186, Mar.
2013.
[8] S. Hemmila, J. V. Cauich-Rodriguez, J. Kreutzer, and P.
Kallio, "Rapid, simple, and cost-effective treatments to
achieve long-term hydrophilic PDMS surfaces," Applied
Surface Science, vol. 258, no. 24, pp. 9864-9875, Oct.
2012.
[9] S. Bilgin, M. Isik, E. Yilgor, and I. Yilgor,
"Hydrophilization of silicone-urea copolymer surfaces by
UV/ozone: Influence of PDMS molecular weight on
surface oxidation and hydrophobic recovery," Polymer,
vol. 54, no. 25, pp. 6665-6675, Nov. 2013.
[10] T. Jianshi et al., "Flexible CMOS integrated circuits based
on carbon nanotubes with sub-10 ns stage delays," Nature
Electronics, vol. 1, no. 3, pp. 191-196, Mar. 2018.