The carbon nanotube (CNT) is a hollow tube composed of carbon atoms with its diameter ranging from a few nanometers to tens of nanometers. CNTs have been extensively studied and analyzed since their discovery [1] in 1991 by Dr. Sumio Iijima of the NEC Fundamental Research Laboratory, as their excellent electrical and mechanical properties appear to be promising for many micro- and nanoelectronic applications, such as field-emission displays [2], sensors [3,4] and field-effect transistors (FET) [5]. The reason why the CNTs possess particular mechanical and thermal properties is strongly related to their structure. Colbert et al. indicates that these properties are closely related to the hexagonal array of graphene, which is the densest possible packing of atoms in two dimensions, together with the extraordinary strength of C-C sp2 bonds [6]. The CNT is about 100 times stronger than steel, yet one-sixth the weight because of its hollow center and chicken-wire-like structure. As for thermal conduction, the CNT surpasses even that of diamond. All of the mechanical, thermal and electrical characteristics make the CNTs (both single- and multi-walled) become an attracting candidate for the future micro- and nano-electronics.
CNTs fall under the category of fullerenes, which in turn falls under the category of all carbonic substances. There are three major forms of carbon, i.e., graphite, diamond and fullerene.
First of all, graphite is composed of stacked sheets (i.e., graphene sheets) of carbon. Each graphene sheet is composed of hexagonally arranged carbon atoms and
the structure is analogous to honeycomb. Figure 1.1(a) shows three graphene sheets stacked together. It is worth noting that a piece of graphite can consist of millions of layers of stacked graphene sheets. Each dot in Figure 1.1 represents a carbon atom, and each line between two dots represents the covalent bond between two carbon atoms.
Second, Figure 1.1(b) shows the atomic structure of diamond. Diamond is a kind of ores that is hard and highly thermally conductive and the carbon atoms are arranged tetrahedrally.
Fullerene, the third form of carbon, exhibits hollow and cage-like structures.
The Bucky ball is one of the well-known structures (Figure 1.1(c)). The soccer-ball-like sphere is constructed hexagonally or pentagonally by carbon atoms, and can range in diameter.
The CNT is basically a fullerene and is an extended Bucky ball (Figure 1.2). It can be viewed as a graphene sheet rolled up to form a seamless cylinder of variable length. In Figure 1.2 we can see that the CNT is capped by half of a Bucky ball at both ends. CNTs include both single-and multi-walled structures. Single-walled carbon nanotubes (SWNTs) comprise just only one cylindrical graphite sheet. Their diameters are typically ~1nm. The multi-walled nanotubes (MWNTs) comprise several to tens of concentric cylinders, and the space between two successive graphitic shells is 3.4
A0 . Typically, the MWNTs tend to have diameters in the range 2-100nm.
Figure 1.3 shows the size distribution of the major carbonic substances.
Theoretically, the smallest SWNT diameter is 0.4nm[7], and was demonstrated by Wang et al. in 2000 [8]. It is also noted that the length to width ratio (i.e., aspect ratio) of a CNT can be up to one thousand or even higher. This characteristic also prolongs
the application field of the CNTs.
There are three major methods to manufacture CNTs. They are laser ablation, arc-discharge and catalytic chemical vapor deposition (CCVD).
In laser ablation process, a graphite target loaded with a catalyst is positioned in a tube furnace and irradiated by a laser. This method can manufacture both SWNTs and MWNTs with high yield. The nanostructures are deposited at the cooler zone near the end of the furnace tube in the direction of the gas flow. The method has several advantages, such as high-quality SWNT production, diameter control and the investigation of growth dynamics.
In arc-discharge method, an arc discharge between two graphite rods is ignited and results in the consumption of one electrode, forming different carbon nanostructures that can be collected at different positions in the reactor chamber.
SWNTs, MWNTs and fullerenes can be grown with different yields. Usually, a considerable amount of soot and carbon nanoparticles are also formed concurrently which must be removed by complex after-treatments. Large-scale synthesis of MWNTs by arc discharge has been achieved by using He-gas, and their thermal purification has also been successful. When a graphite rod with metal catalyst (Fe, Co, Ni, etc.) is used as the anode with a pure graphite cathode, SWNTs are generated in the form of soot. The crystallinity and perfection of arc-produced CNTs are generally high.
All kinds of carbon nanostructures (i.e., filaments) can be synthesized by applying catalytic chemical vapor deposition (CCVD) method. In this process, a furnace is loaded with a metal catalyst and fed with a carbon-containing gas (carbon source). Then the CNTs are synthesized on the catalyst surfaces at moderate temperatures in the range of 600 to 1000°C (and different pressures).
In order to make good use of the promising material, CNTs have been intensively studied by many research groups in many different fields in the paste decade. Among them, we pay our attention to the electrical properties mainly in this dissertation.
Again, CNTs have several impressive properties, including ultrahigh mobility, high current density capacity, and a suitable on/off ratio for FET purposes [9].
Generally, CNTs depict two different types of electrical characteristics (i.e., metallic- or semiconducting-type), depending mainly on the CNTs’ chirality. The word
“chirality” refers to the angle in which the hexagonal network of the nanotube is formed with respect to the tube axis. Therefore, not only the length and diameter of the CNT can be varied, but also helicity of the hexagonal network.
For clarity, the CNTs discussed in the following sections are single-walled carbon nanotubes (SWNTs). SWNTs are made from a single graphene sheet (Figure 1.2). The SWNT consists of a single CNT, typically on the order of 1.4nm in diameter.
Figure 1.4 shows a TEM picture of a SWNT. The two dark lines in the TEM picture correspond to two sides of the SWNT’s wall.
Figure 1.5 illustrates the concept of chirality. The STM picture (Figure 1.5(a)) shows five CNTs with different chiralities for demonstration [10]. The dotted line (vector T) is drawn along each CNT axis. Then a vector, H, is drawn from the same starting point with T. It is worth noting that vector H is parallel to the rows of consecutive hexagons in the carbon atom network and the vector is chosen to be parallel to what are called nearest-neighbor (with respect to the tube axis) hexagon rows. The angle between vectors T and H is defined as Φ. The chiral angle (i.e., chirality), θ, is defined as:
θ = 30° -Φ
Depending on θ value, there are three types of chiralities. A CNT with 30˚chiral angle is categorized as armchair CNT. The CNT is called zigzag CNT (No.7 CNT in Figure 1.5(a)) if the chiral angle is zero (Φ = 30˚), while those with the chiral angles ranging from 0˚ to 30˚ are categorized as chiral CNTs. For examples, Number 10, 11 and 1 in Figure 1.5(a) are chiral CNTs with chiral angles of 23˚ (Φ = 7˚), 16˚ (Φ = 14˚) and 5˚ (Φ = 25˚) respectively. The electrical property of CNTs with different chiralities will be briefly described later.
After giving the definition of chiral angle, we must also define another parameter, the chiral vector, which can indicate the CNT structure. In other words, the chiral vector can represent the chiral angle and diameter of the CNT simultaneously.
Basically, the chiral vector is a line that traces the CNT along its circumference from one carbon atom (i.e., the reference atom) back to itself. Imaginably, if a CNT is cut open along the tube axis and through the reference atom, the CNT can be spread out and become a graphene sheet (Figure 1.5(b)). The dotted lines at both left and right sides of the figure represent the cut made along the CNT. Although the chiral vector begins and ends on the same reference atom, the end is represented by position (11,7) in the graphene sheet. It is worth noting that the locations (0,0) and (11,7) coincide on the same reference atom when the graphene sheet is wrapped to form a cylinder.
Again, vector H is drawn parallel to the nearest neighboring row. Whereas the chiral vector is perpendicular to the tube axis, the armchair (dotted line) is perpendicular to the H vector (The row of large dots in Figure 1.5(b) indicates the nearest neighboring row of hexagons). The resulting rolled-up carbon nanotube would be an armchair CNT, if the chiral vector lines up with the armchair line. In other words, the armchair line bisects every hexagon it passes through.
The unit vectors, a1 and a2, the unit vectors, both begin at one corner of a single
hexagon and end two corners away in the same hexagon. Since a1 and a2 each traverse one whole hexagon, the coordinates (n, m) represent atoms that are n and m hexagons away from the reference atom in the a1 and a2 directions respectively. The chiral vector is therefore represented as following:
C = na1 + ma2
A CNT can be characterized by the notation (n,m) which refers to the chiral vector, where n and m are positive integers.
According to the chiral angle equation mentioned earlier, the angles θ and Φ always combine to form 30°. A (n,n) configuration will result in an armchair CNT, while (n,0) and (0,m) configurations will result in zigzag CNTs. Finally, the (n,m) indicates the chiral CNT when both n and m are non-zero integers and n≠m[10].
After introducing the fundamental concepts of the CNTs, we will discuss the metallic and semiconducting conduction of CNTs with different chiralities.
It has been proved that metallic tubes have conductivities higher than copper and can carry a current density that meets or exceeds the best metals (Table 1.1). The excellent metallic behavior makes CNTs a potential candidates for nanoscale wires [11]. It is also worth noting that semiconducting tubes have mobilities and transconductances that meet or exceed the best semiconductors [12]. The following sections will discuss how to distinguish metallic CNTts from semiconducting CNTs by using the chiral vector mentioned earlier.
Table 1.1 and Table 1.2 show the armchair, zigzag, and chiral tubes and their corresponding electrical properties. The chirality and diameter of a CNT is extremely important because they determine the properties of the CNTs, especially the electrical characteristics [10]. In short, both the diameter and chirality determine whether the CNT will be metallic or semiconducting.
Armchair CNTs (i.e., chiral angle is 30°) have been demonstrated theoretically and experimentally to be metallic in conduction (Table 1.2). Similarly, zigzag and chiral tubes have been shown to be metallic- or semiconducting-type given the appropriate diameter. Jeroen W. G. Wildooer et al. indicates that the energy gap (Eg) is dependend on the diameter [10], that is, Eg is proportional to 1/diameter (i.e.,
diameter
Eg =1 ).
For a given (n, m) nanotube, if 2n + m = 3q (where q is an integer), then the nanotube is metallic, otherwise the nanotube is semiconducting. Thus all armchair nanotubes (n = m) are metallic.
However, CNTs can consist of multiple layers of grapheme sheets concentrically, resulting in the well-known multi-walled carbon nanotubes (MWNTs). The diameter of the outermost tube in a MWNT ranges from 10-20nm typically. Figure 1.6 shows TEM pictures of three MWNTs with different quantities of tubes concentrically. The conductivity of each tube in a MWNT is different, just like the SWNT. In short, there is high possibility that both metallic and semiconducting-type nanotubes (layers) exist in the same MWNT.
Currently, both metallic- and semiconducting-type CNTs are produced simultaneously in the three CNT synthesis methods mentioned earlier. For further applications of CNTs in micro- or nano-electronics, it is crucial to be able to distinguish metallic-type CNTs from semiconducting-type CNTs. Some research groups has developed a technique successfully for sorting SWNTs by their electronic properties and diameters. They found out that a certain sequence of single-stranded DNA could be formed as a helical structure around individual SWNTs. The most important discovery was that the electrical characteristics of the DNA-CNT hybrid strongly depend on the diameter and chirality of SWNT strongly. Later, a technique
called anion exchange[13] was used to filter out the hybrids and the mixture of metallic and semiconducting type CNTs could be sorted out.
Collins et al. demonstrated a method [14] for selectively removing single carbon shells from multi-walled CNTs (MWNTs) stepwise and individually characterizing the different shells using the partial electrical breakdown of a MWNT at constant voltage stress. By choosing among the shells, Collins et al could convert a MWNT into either metallic or semiconducting conductor. This approach takes advantage of current-induced electrical breakdown to eliminate individual shells one at a time, and the outer shells are more likely to breakdown. However, the applied current needs to be controlled precisely, otherwise, both metallic and semiconducting CNTs would fail.
Moreover, this method is time-consuming.
Balasubramanian et al. have disclosed a selective electrochemical approach to fabricate CNT-FETs [15]. They used electrochemistry for selective covalent modification of metallic CNTs, resulting in exclusive electrical transport through the unmodified semiconducting CNTs. The semiconducting CNTs were rendered nonconductive by application of an appropriate gate voltage prior to the electrochemical modification. The FETs fabricated in this manner display good hole mobilities and a ratio approaching 106 between the current in the ON and OFF state.
However, this approach is problematic. For example, when there are much more metallic nano-tubes than semiconducting nano-tubes in the deposited CNT-based material, this electrochemical approach can only improve the electrical characteristics of the few semiconducting CNT-FETs and still fails to increase the percentage of semiconducting CNT-FETs. On the other hand, this approach requires the chip to be immersed in a chemical solution, which reduces the yield and throughput. Moreover, the phenyl group in the solution may react with semiconducting CNTs to form
covalent bonds and adversely affects the electrical characteristics of the chip, which makes it unsuitable for use in sensors.
Since CNTs exhibit two different types of electrical properties, they can be employed as FETs as well as interconnects/vias/contact holes. However, there are challenges that need to be tackled before their adoption to many practical applications, especially the process compatibility with the existing silicon-based semiconductor technologies, controlling the placement and manipulation of massive numbers of CNTs at precise locations, the chirality of the CNTs, and the manufacturing of both n- and p-type CNTs simultaneously on the same substrate. Among these challenges, an effective method must be developed for efficiently controlling the placement of massive numbers of CNTs because conventional silicon-based micro- or nanoelectronics often consist of millions or billions of devices.
Although IBM’s research group has demonstrated the repositioning of a single CNT on particular surface successfully, it is obvious that this technique is limited to manipulate one CNT at a time. If an IC chip uses nanotubes as channel/active layer of FETs (semiconducting-type CNTs) or interconnect (metallic-type CNTs), millions or billions of nanotubes would require accurate placement over the chip. The physical manipulation of nanotubes one at a time is absolutely inefficient for current IC technology.
Efforts have been made by many research groups to overcome the manipulation issue. There are two major approachs to manufacture CNT-FETs in the past few years.
One approach is to first create the source and drain electrodes throughout the wafer, and then disperse CNTs on the wafer. Undoubtedly, there exists a small probability for CNTs to bridge some of the electrodes to form functional CNT-FETs. The drawback is the yield is quite low and impractical [16].
Another popular technique is to spread large quantities of CNTs on a wafer.
Then EM and STM are applied to find the location of CNTs with desired chirality and dimensions. Electrodes can then be deposited on top of the CNT with the desired properties by e-beam lithography and lift-off method. It goes without saying that the yield of this technique is also very low [17].
It is obvious that the physical manipulation of numerous CNTs or spreading CNTs randomly one at a time is laborious and impractical for mass production, thus the ability to form massive numbers of CNTs in precise locations remains a key issue for CNTs in nanotechnology applications.
In order to overcome the manipulation problem, a number of techniques have been proposed to achieve a regular CNT network by controlling the gas flow direction [18], using porous templates [19], using electric-field-assisted assembly [20,21], utilizing chemically functionalized template [22], adopting fluidic alignment [23,24], or using electric-field-directed-growth of CNTs [25]. Although these methods all achieve acceptable results in both the growth direction and the length of CNTs, they require additional equipments [26]. It is obvious that the aligned-CNT-growth methods [27] are more promising than post-growth-assembly-of-CNT methods for CNT-FETs mentioned above. One of the aligned-CNT-growth methods, which has emerged as the most popular method, involves the catalytic disproportionation of carbon source (carbon monoxide usually) on bimetallic catalysts containing molybdenum/cobalt in chemical vapor deposition (CVD) system [28].
Although catalytic mixtures of cobalt (Co) and molybdenum (Mo) have been considered essential for the growth of single-walled carbon nanotubes (SWNTs) from carbon monoxide (CO) or hydrocarbons by the CVD method, we demonstrate the growth of bundled-CNTs with only Co particles as catalyst [29,30]and ethanol as
carbon source in this dissertation. Some previous reports also indicated that CNTs manufactured by CVD methods with Co catalyst usually resulted in predominantly multi-walled tubes [31,32]. In this thesis, our reiterative and systematic experiments show that the selective growth of bundled-CNTs produces mostly SWNTs.
For CNT-FETs and biosensors, it is necessary to employ single-walled carbon nanotubes (SWNTs) instead of multi-walled carbon nanotubes (MWNTs) because of the unique semiconducting property of the SWNTs. In order to obtain SWNTs, the catalyst size should be reduced to as small as possible [29]. In this dissertation, a method is proposed to synthesize SWNTs and form bridged-CNTs between two catalyst islands. The dominant parameters in the aligned growth of SWNTs are found to be the size and the location of catalyst nanoparticles. The characteristics of embedded Co nanoparticles in patterned cobalt-mix-tetraethoxysilane (CMT) islands for SWNT growth are discussed under different hydrogen reduction conditions, catalyst concentrations, and carbon ratios during CNT growth.
Since for the mainstream complementary metal oxide semiconductor (CMOS) circuit applications, both p- and n-type metal-oxide-semiconductor field effect transistors (MOSFET) are called for simultaneously on the same chip. It is thus necessary to fabricate n-type, in addition to p-type CNT-FETs, on the same chip for the complementary circuits. In general, the CNT-FET acts like a p-type conduction device when the CNT is exposed to air [33-37]. However, it is quite difficult to manufacture n-type CNT-FETs. Several approaches have been previously reported to form n-type CNT-FETs by employing complex doping processes (i.e., adopting alkali metals) [38-41] or thermal/electrical annealing processes [41]. These approaches, however, require extra processing and masking steps to convert generic p-type CNT-FETs in vacuum or in the inert gas. In contrast, no extra annealing steps are
needed to form air-stable n-type CNT-FETs [42-44] using the passivation method proposed in this dissertation
In conclusion, we will demonstrate that CNT is an appropriate material for FET applications. We will show the result of manufacturing both air-stable p- and n-type CNTFETs for CMOS without any complex ion doping process. Research for enhancing the CNTFET electrical properties and practicality is still on-going, and has made some progress.
1.2 Organization of the Dissertation