1. Introduction

Semiconducting single-walled carbon nanotubes (SWNTs) have attracted much attention because of their unique electronic properties and promising applications for nanoscale electron devices such as carbon nanotube fieldeffect transistors (NTFETs) and quantum-effect devices. The performance of the NTFETs is expected to exceed conventional Si-based transistors [1, 2]. Carbon nanotubebased transistors have shown dramatic progress since the first fabrication of NTFETs [3]. Recently, 50 GHz mixer operation [4] and a 1.9 ns/gate delay time [5] have been reported using NTFETs. In order to fully develop the potential of NTFETs, there are still some issues that should be addressed. The control of the conduction type is one of the most important issues. From the application point of view, both n- and p-type NTFETs are necessary in order to

implement CMOS-like logic gates and integrated circuits [5]. Doping techniques have also been proposed for controlling conduction characteristics such as the adsorption of potassium on the nanotube surface [6, 7], the partial exposure to the gas [8], covering the channel region by polymer [9, 10], and the encapsulation of metals [11] or organic molecules [12] into the nanotube. Non-chemical doping such as electrostatic doping also makes it possible to control the conduction characteristics [13, 14]. Recently, we have reported n-type NTFETs fabricated using Ca with a small work function (φm ) as the contact electrodes without any doping into the nanotube [15]. This simple technique for implementing n- type NTFETs is based on the Schottky barrier transistor model, in which the transistor action in NTFETs is realized by modulation of the Schottky barrier formed at the interface between the nanotube and the contact metal [16, 17]. The Schottky barrier height for electrons, which is approximately

Figure 1. Layer structures of the contact electrodes and the work function alignment of the contact metals.

10-6

VGS (V)

Figure 2. |ID|–VGS characteristics of the fabricated devices: Pd-NTFET (broken line), Ti-NTFET (dot–dash line), Mg-NTFET (solid line), and Ca-NTFET (dotted line). The transfer ID–VGS characteristics varied from p-type through ambipolar to n-type with decreasing φm .

(This figure is in colour only in the electronic version)

given by the difference between the work function of the contact metal and the electron affinity of the nanotube in the ideal case, is lowered by using the contact metals with small φm [18], and hence electron injection from the electrode to the nanotube becomes easy. However, we have found that the Schottky barrier for electrons at the interface exists even though the work function of Ca (2.9 eV) is much smaller than that of the nanotubes ( 4.8 eV [19]). It is still necessary to investigate the interface properties at the metal/nanotube contacts.

In this paper, we have investigated the relation between the conduction property and the work function of the contact metals in the NTFETs, using four kinds of contact metals such as Pd (φm = 5.1 eV), Ti (4.4 eV), Mg (3.6 eV), and Ca (2.9 eV). The work functions of these metals range from below the valence-band edge (EV) to above the conduction-band edge (EC) of SWNTs, as shown in figure 1. We have also fabricated the nanotube quasi-pn diode, which consists of Ca and Pd for the cathode and anode electrodes, respectively.

h

e

e

h Source SWNT

Figure 3. Schematic band diagrams at the source contacts for the ON state in (a) Pdand Ti-NTFETs, (b) Mg-NTFET, and (c) Ca-NTFET.

2. Device structure and fabrication

NTFETs fabricated using contact metal M (Pd, Ti, Mg and Ca) are referred to as M-NTFETs in this paper (Pd-NTFETs and Mg-NTFETs, for example). The layer structures of each contact electrode are shown in figure 1. The upper layers of Au/Ti and Al were used as cover metals to prevent the oxidation or the degradation of the thin contact metals. The fabricated devices are back-gate-type NTFETs, with its 3 µm channel. A heavily doped p+-Si substrate with a 100 nm-thick SiO2 layer was used as the back gate. Following patterning of the catalysts, consisting of a double layer of Co (2 nm) on Pt (10 nm), the SWNTs were grown on the substrate by alcohol catalytic chemical vapour deposition [20, 21]. A mixture of ethanol (10 sccm) and argon (100 sccm) was used as a carbon source. The total pressure in the furnace was 1.3 Torr. The growth temperature and time were 800 ◦ C and 1 h, respectively. Source and drain contact metals were deposited by the electron-beam evaporation. Because Mg and Ca are active materials, oxidation of contact metal is expected. In order to avoid the oxidation and the degradation of thin contact layers, the deposition of contact metals was carried out in a high vacuum, followed by successive evaporation of the upper cover metals of Au/Ti and Al. Even though the characteristics of Ca-NTFETs change gradually in ambient air, they are stable for 2–3 months by keeping them in vacuum. In the present experiments, Mg-NTFETs and Ca-NTFETs are kept in a vacuum chamber after device fabrication. The electrical measurements of the NTFETs were performed at room temperature in vacuum.

3. Electrical characteristics depending on the work function of the contact metal

Figure 2 shows the drain current (| ID|) as a function of the gate voltage (VGS) of the four NTFETs with Pd, Ti, Mg, and Ca contact electrodes. Here, the drain voltage |VDS| was 0.1 V. The transfer ID–VGS characteristics were p-type in the Pdand TiNTFETs, ambipolar in the Mg-NTFET, and n-type in the CaNTFET. This result suggests a change in the metal Fermi level (EF) at the contact/nanotube interface depending on φm . This is illustrated schematically in figure 3, where band alignments at the source contacts are shown for the FETs in the ON state. The p-type conductions probably result from the large φm of

10-9

Work function (eV)

Figure 4. |ID| of various NTFETs as a function of φm for (a) hole current and (b) electron current. Here, |VDS | and |VGS| were 1 and 10 V, respectively.

the Pd and Ti electrodes. In these cases, the EF of the source electrodes would be positioned close to the valence band (EV) of the nanotube, as shown in figure 3(a), in which the holes are easily injected into the nanotube. The ambipolar characteristics observed in the Mg-NTFET suggests that the EF is positioned at around the middle of the band gap of the nanotube, as shown in figure 3(b). In this case, it is possible to inject both electrons and holes into the nanotube, depending on the polarity of VGS. In the case of the Ca-NTFET, in which the n-type conduction property is realized, the EF of the Ca contact is thought to be positioned close to the conduction band (EC) of the nanotube.

These EF alignments at the metal/nanotube interfaces are also supported by the magnitude of the drain current, which also depends on the φm of the contacts. | ID| is plotted as a function of φm in figure 4(a) for the p-channel NTFETs and in figure 4(b) for the n-channel NTFETs. Here, |VDS|

VAC (V)

Figure 5. I –V characteristics of the fabricated nanotube quasi-pn diode. The insets show schematic band diagrams of the nanotube diode at forward and reverse biases, respectively.

and |VGS| were 1 and 10 V, respectively. Even though it is not frequent, some Ca-NTFETs show weak ambipolar characteristics with small p-channel ID. This behaviour is consistent with the Schottky barrier transistor model in which the barrier height for holes is very large. The scatter in ID is thought to be due mainly to the effect of the contact resistance. It seems reasonable to discuss the φm dependence of ID, focusing on devices with large ID because these devices are less affected by the contact resistance. The hole current increased exponentially with increasing φm . In contrast, the electron current seems to decrease with increasing φm , and could not be measured in Tiand Pd-NTFETs. In the case of NTEFTs with large work function contact metals, the barrier height against electrons is high, leading to the non-detectable ID. The exponential dependence of the drain current on φm suggests that the Schottky barrier height which determines the ID is dependent on φm , because the transmission coefficient of carriers tunnelling through a potential barrier via thermionic field emission decays exponentially with the barrier height [16]. This suggests that the Fermi-level pinning is weak in the case of NTFETs. This contrasts with conventional semiconductors such as Si and GaAs, in which the EF of the contact metals is usually independent of φm and pinned at a level in the band gap of the semiconductor. This would be one of the advantages of carbon nanotubes as a material for semiconductor nano-devices.

4. Nanotube quasi-pn diodes

As described above, it is possible to select the type of carriers injected into the nanotube by choosing the contact metal. By utilizing this property of metal/nanotube contacts, we have fabricated nanotube quasi-pn diodes using Pd and Ca for the anode and cathode contact electrodes, respectively. Figure 5 shows the I –V characteristics of the fabricated nanotube diode. Here, the gate-cathode voltage VGC is 0 V. The rectification I –V characteristics were confirmed. The insets show the

schematic band diagrams of the nanotube diode at positive and negative anode–cathode voltages (VAC), respectively. The Schottky barrier height formed at the Pd and Ca contacts are low for holes and electrons, respectively. At VAC > 0, the Schottky barrier at the Pd contact is small, and holes are injected into the nanotube. On the other hand, the current does not flow at VAC < 0 because the Schottky barrier for electrons is high and thick at the Pd anode. The carriers are not injected at the Ca contact, because EF is located in the gap. The ON/OFF ratio of the diode current was 103 at VAC = ±1 V.

5. Summary

In summary, we have investigated the relation between the conduction property and the φm of the contact metal in NTFETs. In contrast to NTFETs with Ti and Pd contact electrodes, which showed p-type conduction characteristics, devices with Mg contact electrodes showed ambipolar characteristics and most devices with Ca contact electrodes showed n-type characteristics. This behaviour can be explained by the φm -dependent Schottky barrier height, which suggests that the Fermi-level pinning at the interface is weak. We also fabricated a nanotube quasi-pn diode utilizing different contact metals with different φm s: Pd for the anode and Ca for the cathode. Rectification I –V characteristics with an ON/OFF ratio of 103 were obtained.

References

[1]Dekker C 1999 Phys. Today 52 22

[2]Dresselhaus M S, Dresselhaus G and Avouris Ph 2001 Carbon Nanotubes: Synthesis, Structures, Properties, and Applications (Berlin: Springer)

[3]Tans S J, Verschueren A R M and Dekker C 1998 Nature

393 49

[4]Rosenblatt S, Lin H, Sazonova V, Tiwari S and McEuen L 2005 Appl. Phys. Lett. 87 153111

[5]Chen Z, Appenzeller J, Lin Y M, Sippel-Oakley J, Rinzler A G, Tang J, Wind S J, Solomon P M and Avouris Ph 2006

Science 311 1735

[6]Bockrath M, Hone J, Zettl A, McEuen P L, Rinzler A G and Smalley R E 2000 Phys. Rev. B 61 R10606

[7]Zhou C, Kong J, Yenilmez E and Dai H 2000 Science 290 1552

[8]Liu X, Luo Z, Han S, Tang T, Zhang D and Zhou C 2005 Appl. Phys. Lett. 86 243501

[9]Lu C, Fu Q, Huang S and Liu J 2004 Nano Lett. 4 623

[10]Klinke C, Chen J, Afzali A and Avouris Ph 2005 Nano Lett.

5555

[11]Izumida T, Jeong G H, Neo Y, Hirata T, Hatakeyama R, Mimira H, Omote K and Kasama Y 2005 Japan. J. Appl. Phys. 44 1606

[12]Shiraishi M, Nakamura S, Fukao T, Takenobu T, Kataura H and Iwasa Y 2005 Appl. Phys. Lett. 87 093107

[13]Lee J U, Gipp P P and Heller C M 2004 Appl. Phys. Lett.

85145

[14]Lin Y M, Appenzeller J, Knoch J and Avouris Ph 2005 IEEE Trans. Nanotechnol. 4 481

[15]Nosho Y, Ohno Y, Kishimoto S and Mizutani T 2005 Appl. Phys. Lett. 86 073105

[16]Heinze S, Tersoff J, Martel R, Derycke V, Appenzeller J and Avouris Ph 2002 Phys. Rev. Lett. 89 106801

[17]Mizutani T, Iwatsuki S, Ohno Y and Kishimoto S 2005 Japan. J. Appl. Phys. 44 1599

[18]Chen Z, Appenzeller J, Knoch J, Lin Y M and Avouris Ph 2005

Nano Lett. 5 1497

[19]Suzuki S, Bower C, Watanabe Y and Zhou O 2000 Appl. Phys. Lett. 76 4007

[20]Murakami Y, Miyauchi Y, Chiashi S and Maruyama S 2003

Chem. Phys. Lett. 374 53

[21]Ohno Y, Iwatsuki S, Hiraoka T, Okazaki T, Kishimoto S, Maezawa K, Shinohara H and Mizutani T 2003 Japan. J. Appl. Phys. 42 4116