Carbon 42 (2004) 2263–2267

www.elsevier.com/locate/carbon

Fabrication of carbon nanotube field effect transistors by ACdielectrophoresis method

Jingqi Li *, Qing Zhang, Dajiang Yang, Jingze Tian

Microelectronics Center, School of Electrical and Electronic Engineering, Nanyang Technological University, S1-B2c-20, Singapore 639798, Singapore

Received 26 December 2003; accepted 4 May 2004

Available online 9 June 2004

Abstract

Single wall carbon nanotubes (SWNTs) suspended in isopropyl alcohol have been placed between two electrodes by AC di-

electrophoresis method. The number of SWNTs bridging the two electrodes is controlled by SWNT concentration of the suspension

and deposition time. Through selectively burning off the metallic SWNTs by current induced oxidation, the back-gate carbon

nanotube field effect transistors (CNTFETs) with a channel current on–off ratio of up to 7· 105 have been successfully fabricated.

The success rate of the CNTFETs in 20 samples is 60%. These results suggest that AC dielectrophoresis placement method is an

efficient technique to fabricate CNTFETs with some flexibilities of controlling CNT reconnection, length and orientation.

� 2004 Elsevier Ltd. All rights reserved.

Keywords: A. Carbon nanotubes; C. Atomic force microscopy; D. Electrical (electronic) properties

1. Introduction

Single wall carbon nanotubes (SWNTs) are an ideal

one-dimensional (1D) system because of their small

diameter (of the order of 1 nm) and great length (of the

order of micrometer). The 1D structure allows electrons

to move in only two directions, leading to a reduced‘‘phase space’’ for scattering processes [1]. In the absence

of scattering, the transport is ballistic, which makes the

carbon nanotube an ideal microelectronic device mate-

rial, especially for field effect transistors (FETs). Since

the first batch of CNT field effect transistors (CNT-

FETs) [2,3] were fabricated in 1998, their performance

has been significantly improved in aspects of CNT

channel current on–off ratio [4,5], hole mobility [6] andthe CNT–metal electrode contacts [5]. However, how to

selectively place semiconducting SWNTs in desirable

locations is not solved. At present, two methods are

generally used for CNTFET device fabrication. The first

is to spin-coat SWNT suspension onto structured wafers

[2,3,7,8]. However, the random distribution of SWNTs

over the wafers is the major drawback of this method.

The second method is to grow SWNTs along wafer

*Corresponding author. Tel.: +65-679-05855; fax: +65-679-33318.

E-mail address: p144703764@ntu.edu.sg (J. Li).

0008-6223/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.carbon.2004.05.002

surface to connect desirable electrodes [9,10]. The major

disadvantages of this technique are the catalyst con-

tamination and poor selectivity of SWNTs. Recently,

DC and AC electrophoresis methods have attracted

numerous interest. It has been demonstrated to be an

efficient method to deposit large number of CNTs [11–

13], SWNT matted sheet [14], and a single bundle withdefinite orientation [15]. Most recently, studies reported

depositing single bundle simultaneously onto an array of

electrodes [16]. However, except for Krupke et al. results

obtained in vacuum [16], there has been no further re-

port on the fabrication of CNTFETs using AC dielec-

trophoresis method.

This paper reports the CNTFETs fabricated by AC

dielectrophoresis method at an ambient environment.SWNTs were connected between the source and drain

electrodes along the direction of the electric field. The

number of SWNTs is controlled by the SWNT con-

centration in the suspension and the deposition time.

2. Experimental

2.1. Carbon nanotube deposition

CNTFETs were fabricated on p-type silicon wafers

which were thermally coated with a 500 nm thick silicon

Fig. 2. AFM images of SWNTs bridging the adjacent electrodes. The

deposition time is (a) 50 s and (b) 300 s, respectively. The black arrows

2264 J. Li et al. / Carbon 42 (2004) 2263–2267

dioxide layer. Source and drain electrodes were made of

20 nm thick Ti and 40 nm thick Au. The distance be-

tween the electrodes is in the range of 1–4 lm.

SWNTs with very high purity were deposited betweentwo electrodes using AC dielectrophoresis method. The

diameter of the SWNTs was around 1.4 nm. To prepare

the CNT suspension, 1mg of SWCNTs were dispersed in

500 ml isopropyl alcohol (IPA) and ultrasonically trea-

ted for 10 min. The suspension was transparent after

further dilution. A drop of the suspension was intro-

duced onto the structured wafers on which an AC bias

was applied, as shown in Fig. 1. The peak-to-peak ACvoltage was between 4 and 8 V, depending on the elec-

trode distance, and the frequency was fixed at 10 MHz.

Monitoring the variation of the resistance between two

electrodes is an easy way to confirm the connection of

CNTs and the electrodes. The resistance was typically

decreased to several hundred kX after CNTs bridged the

electrodes. Once the resistance drop was detected, the

AC bias was turned off immediately. After the CNTshad been deposited successfully, the wafer was rinsed

using IPA and deionized water to clean the wafer sur-

face. In this way, CNTs could be placed across any two

of the four electrodes in Fig. 1. In addition, all the four

electrodes could be connected pair by pair without

influencing the CNT connection formed earlier. Fig. 2

shows CNT bundles between two electrodes. The dia-

meters of the bundles are typically in the range of 40–80nm. It is clear that the SWNTs bridge the electrodes

along the direction of the electric field. If the two elec-

trodes are parallel, the SWNT orientation is perpen-

dicular to the parallel electrodes, as observed in Refs.

[14–16]. Therefore, the SWNT length between them is

controllably close to the distance between the parallel

electrodes. The number of SWNTs between the elec-

trodes is determined by the deposition time and SWNT

Fig. 1. SEM image of four Ti/Au electrodes (white parts) on p-type Si

wafer with 500 nm thick SiO2. AC bias voltage is applied between two

adjacent electrodes, as indicated in the figure.

indicate the direction of electric field.

concentration of the suspension. The longer the depo-

sition time, the more the SWNTs are deposited.

Two forces can drive CNTs in the suspension to the

electrodes [14,16]. The first is an electrophoretic force

due to charge, the second is the dielectrophoretic force

due to the dielectric constant difference between CNTsep and the solvent medium em. Since a high frequency

AC bias was used here, the electrophoretic force did not

affect the motion of the CNTs because of time averaging

effect. The time-averaged dielectrophoretic force can be

expressed approximately as follows:

F!/ em

ep � emep þ 2em

rE2rms ð1Þ

where Erms is the average electric field strength [17].

Because the permittivity of the SWNTs is larger than the

other contaminants (for example, amorphous carbon) in

90

120

Curve 1

J. Li et al. / Carbon 42 (2004) 2263–2267 2265

the suspension [14], contaminations were rarely found

between electrodes under the high frequency bias, as

shown in Fig. 2. This phenomenon is consistent with

other group’s observation [11,12,14].

0 5 10 150

30

60

I d(x

10-6

A)

Vds(V)

Curve 3

Curve 2

Fig. 4. Id as a function of Vds during the burn-off process. Curve 1 and

2 are two consecutive sweepings from 0 to 10.5 V. Curve 3 is the final

sweeping from 0 to 11 V.

-7

1x10-6

-1.0 -0.5 0.0 0.5 1.0 1.5-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

5~202-1

-4

-7

-10

I d(x

10-6

A)

Vds(V)

Vg= -10~20VStep: 3V

SiO2

TubeS D

P - Si

(a)

2.2. Metallic tube breakdown process

After CNTs bridged the source and drain electrodes,

the drain current Id as a function of drain voltage Vds wasmeasured using a HP-4156B semiconductor analyzer at

ambient atmosphere. A typical Id–Vds characteristic of

as-prepared CNTFETs is shown in Fig. 3. It is clearly

that the conductance decreases with increasing the gate

voltage Vg in the range of )10 to 14 V, and Id almostsaturates for Vg > 14 V. When �10 < Vg < 14 V, the

semiconducting SWNTs dominate the total conduc-

tance. In contrast, the metallic SWNTs start to domi-

nate the channel conductance for Vg > 14 V because the

semiconducting SWNTs are switched off at the gate

voltage. In order to improve the performance of CNT-

FETs, the metallic SWNTs (m-SWNTs) must be re-

moved from the conduction channel. We use theelectrical breakdown method [18] to destroy the metallic

SWNTs. When Vg > 14 V, semiconducting CNTs are

completely depleted. With a sufficiently high Vds, a large

Id passed through the m-SWNTs and generated enough

heat to burn off them, leaving the semiconducting

SWNTs (s-SWNTs) essentially intact. Fig. 4 shows the

Id–Vds curves during the breakdown process. The

apparent current drops in Curve 1 and 3 indicatebreakdown of the metallic SWNTs.

Breakdown of the metallic SWNTs results from CNT

oxidation under the current-induced heat [18,19].

However, carbon nanotubes have strong carbon–carbon

bonds, and their current carrying capacities exceed 10

lA/nm2 [20,21]. Therefore, the oxidation begins from

the sites of defects and dangling bonds on the CNTs.

-1.0 -0.5 0.0 0.5 1.0 1.5

-3

-2

-1

0

1

2

3

I d(x

10-6

A)

Vds(V)

3022146

-2-10

Vg=-10V~30V

Step: 8V

Fig. 3. Id as a function of Vds measured at Vg ¼ �10 to 30 V in steps of

8 V. The Id � Vds curves are almost overlapped for Vg > 14 V.

-10 -5 0 5 10 15 201x10-14

1x10-13

1x10-12

1x10-11

1x10-10

1x10-9

1x10-8

1x10

I d(A

)

Vg(V)

Vds=-150mV

(b)

Fig. 5. Output and transfer characteristics of a CNTFET measured at

room temperature and ambient environment. P-Si acts as the back gate

and the thickness of the oxide layer is 500 nm: (a) Id � Vds curves

measured for Vg ¼ �10 to 20 V in steps of 3 V. Inset: Schematic cross

section of a back-gate CNTFET, (b) Id � Vg curves measured at

Vds ¼ �150 mV.

2266 J. Li et al. / Carbon 42 (2004) 2263–2267

The metallic SWNTs may be not burn off together at

one time. We suggest that the outmost metallic SWNTs

in the bundle are burnt off first due to being fully ex-

posed to oxygen, while the inner m-SWNTs are morestable and need a higher current to burn off. Some de-

fects must be created first by the large Id for the defect-

free m-SWNTs before they can be burnt off [18].

3. Results and discussion

Fig. 5 shows the output and transfer characteristics of

a CNTFET device fabricated by AC dielectrophoresis

placement method. The device was annealed at 300 �Cfor 15 min at an ambient environment to improve the

contact between nanotubes and electrodes. The Id � Vgcurve shows that the CNT channel is ON when Vg < 0.

The channel conductance decreases with increasing Vgpositively and becomes very small (in OFF state) for

Vg P þ 5 V, indicating a p-type transistor. The on–off

ratio at Vds ¼ �150 mV is 7 · 105, as shown in Fig. 5b.

This on–off ratio is very close to the highest value of 106

reported so far [4,5].

Analogous to the analysis for MOSFETs in the linear

Id � Vds region, The CNT channel current can be ex-

pressed as,

Ids ¼ lCg Vgs

�� Vth �

1

2Vds

�Vds=L2 ð2Þ

where Vth is the threshold voltage, L the length of the

CNT between the source and drain electrodes and Cg

the gate-channel capacitor. For CNTFETs, Cg ¼2pLee0= lnð2h=rÞ, where e, h and r are the dielectric

constant and thickness of silicon dioxide, and the radius

of the carbon nanotube bundle, respectively.

Differentiating Eq. (2), we have the transconductance

dIdsdVg

¼ lhCgVds=L2 ð3Þ

The maximum transconductance deduced from Fig. 5 is

0.19 ls at Vds ¼ 1 V, consistent with other group’s cor-responding values of their back-gate CNTFETs [3,4,7],

but smaller than that obtained from top-gate CNTFETs

[7]. Using e ¼ 3:9, h ¼ 500 nm, L ¼ 3:6 lm and r ¼ 35

nm (L and r are measured from AFM image), we obtain

lh � 106 cm2/Vs from Eq. (3). This value is five times

that of Martel et al. result [3]. However, since Schottky

barriers were found to play an important role in the

performance of CNTFETs [22,23], their effect should betaken into account in a more accurate derivation.

In our experiment, 20 pairs of electrodes have been

bridged by SWNTs and 60% of them were successfully

burned to CNTFETs with on/off current ratio of 3–6

orders of magnitude. Those that failed were destroyed

due to the high voltage applied in the breakdown pro-

cess. This result suggests that AC dielectrophoresis is an

efficient technique to fabricate CNTFETs. We also

found that the 40% failed electrodes could be reloaded

with CNTs by repeating the same process without

affecting the 60% prepared CNTFETs. This is a bigadvantage over the random spin-coating [2,3,7,8] and

direct growth [9,10] technique. In addition, the tube

length and orientation of tubes placed by AC dielec-

trophoresis method [14–16] are controllable, resulting in

uniform device characteristics. All these advantages

make this technique a potential application in large scale

fabrication of CNTFETs.

4. Conclusion

Single wall carbon nanotubes (SWNTs) are control-lably placed between electrodes by AC dielectrophoresis

method. Using the current-induced burn-off process,

metallic SWNTs are easily broken down. The residual

semiconducting SWNTs act as p-type channels of the

CNTFETs, which have an on/off current ratio as high as

7 · 105. This technique provides a high success rate of

fabricating CNTFETs with a flexibility of controlling

the CNT alignment direction and the CNT length.

References

[1] Avouris P. Carbon nanotube electronics. Chem Phys 2002;281:

429–45.

[2] Tan S, Verschueren A, Dekker C. Room temperature transistor

based on a single carbon nanotube. Nature (London) 1998;393:

49–52.

[3] Martel R, Schmidt T, Shea HR, Hertel T, Avouris Ph. Single-and

multi-wall carbon nanotube field-effect transistors. Appl Phys Lett

1998;73(17):2447–9.

[4] Martel R, Wong H-S, Chan K, Avouris Ph. Carbon nanotube

field-effect transistors for logic applications. Proc IEDM 2001:

159–61.

[5] Javey A, Guo J, Wang Q, Mand L, Dai H. Ballistic carbon

nanotube field-effect transistors. Nature 2003;424:654–7.

[6] Durkop T, Getty S, Cobas E, Fuhrer M. Extraordinary mobility

in semiconducting carbon nanotubes. Nano Lett 2004;4(1):35–9.

[7] Wind SJ, Appenzeller J, Martel R, Derycke V, Avouris Ph.

Vertical scaling of carbon nanotube field-effect transistors using

top gate electrodes. Appl Phys Lett 2002;80(20):3817–9.

[8] Bachtold A, Hadley P, Nakanishi T, Dekker C. Logic circuits with

carbon nanotube transistors. Science 2001;294:1317–20.

[9] Javey A, Shim M, Dai H. Electrical properties and device of large-

diameter single-walled carbon nanotubes. Appl Phys Lett

2002;80(6):1064–7.

[10] Liu X, Lee C, Zhou C, Han J. Carbon nanotube field-effect

inverters. Appl Phys Lett 2001;79(20):3329–31.

[11] Chen XQ, Saito T, Yamada H, Matsushige K. Aligning single-

wall carbon nanotubes with an alternating-current electric field.

Appl Phys Lett 2001;78(23):3714–6.

[12] Yamamoto K, Akita S, Nakayama Y. Orientation and purifica-

tion of carbon nanotubes using ac electrophoresis. J Phys D: Appl

Phys 1998;31:L34–6.

[13] Krupke R, Hennrich F, Lohneysen H, Kappes M. Separation of

metallic from semiconducting single-walled carbon nanotubes.

Science 2003;301:344–7.

J. Li et al. / Carbon 42 (2004) 2263–2267 2267

[14] Nagahara L, Amlani I, Lewenstein J, Tsui RK. Directed

placement of suspended carbon nanotubes for nanometer-scale

assembly. Appl Phys Lett 2002;80(20):3826–8.

[15] Krupke R, Hennrich F, Weber HB, Beckmann D, Hampe O,

Malik S, et al. Contacting single bundles of carbon nanotubes

with alternating electric fields. Appl Phys A 2003;76:397–440.

[16] Krupke R, Hennrich F, Weber HB, Kappes MM, Lohneysen HV.

Simultaneous deposition of metallic bundles of single-walled

carbon nanotubes using Ac-dielectrophoresis. Nano Lett 2003;

3(8):1019–23.

[17] Jones TB. Electromechanics of particles. Cambridge, UK: Cam-

bridge University Press; 1995. p. 39–48.

[18] Collins PG, Arnold MS, Avouris Ph. Engineering carbon

nanotubes using electrical breakdown. Science 2001;292:706–9.

[19] Collions PG, Hersam M, Arnold M, Martel R, Avouris Ph.

Current saturation and electrical breakdown in multiwalled

carbon nanoutbes. Phys Rev Lett 2001;86(14):3128–31.

[20] Dai H, Wong EW, Lieber CM. Probing electrical transport in

nanomaterials: conductivity of individual carbon nanotubes.

Science 1996;272:523–6.

[21] Frank S, Poncharal P, Wang ZL, deHeer WA. Carbon nanotube

quantum resistors. Science 1998;280:1744–6.

[22] Wind S, Appenzeller J, Avouris Ph. Lateral scaling in carbon-

nanotube field-effect transistors. Phys Rev Lett 2003;91:058301-1–

058301-4.

[23] Li J, Zhang Q. Schottky barriers in carbon nanotube field effect

transistors fabricated by AC dielectrophoresis method (in

press).