IBM Research News

Engineering carbon nanotubes and nanotube circuits using electrical breakdown Phaedon Avouris, Philip G. Collins and Michael S. Arnold IBM T.J. Watson Research Center April 27, 2001 What was achieved? We developed a technique based on nanotube modification by electrical breakdown that allows the fabrication of carbon nanotube field-effect transistors (NT-FETs) from ropes of single-walled nanotubes (SWNTs) containing both metallic and semiconducting nanotubes without the need to separate the two types of tubes. Using the above breakdown technique we demonstrated an approach that allows the fabrication of dense arrays of NT-FETs using ropes of SWNTs without the need for any pre-alignment or orientation of the nanotubes on the electrodes. We demonstrated the shell-by-shell breakdown of multi-walled nanotubes (MWNTs), learned to control it, and used it to determine the nature of the individual shells of a MWNT. By taking advantage of the dependence of the band-gap of semiconducting shells on their diameter, we produced FETs with desired band-gaps. Despite the stunning Moore's Law exponential development of silicon-based microelectronics over the past several decades, such growth cannot be sustained forever. In fact, many scientists expect that within 10 years a number of factors -- the fundamental physics of materials, fabrication limitations and economic factors -- will halt the miniaturization of silicon devices. Additional microelectronic developments would then require the use of different materials, fabrication principles and/or device concepts. Among the possible new directions, molecular electronics is thought to be particularly promising. Carbon nanotubes (NTs) have unique properties that make them the most promising of several potential candidates for a new nanoelectronic technology. Specifically, they are very thin (as little as 1 nanometers in diameter), but their length can be controlled up to many microns. Nanotubes are also extremely strong mechanically, very stable thermally and chemically, are excellent heat conductors, and most importantly, depending on the orientation of the carbon hexagons with respect to the tube's axis, can be either metals or semiconductors. The semiconducting tubes have band-gaps that are inversely proportional to their diameter. There are two basic types of nanotubes: single-walled nanotubes (SWNTs), which have one shell of carbon atoms, and multi-walled nanotubes (MWNTs), which consist of multiple, nested carbon tubes. Moreover, nanotubes tend to adhere strongly to each other, forming bundles or ropes composed of both metallic and semiconducting tubes. Metallic nanotubes can carry extremely large current densities, while semiconducting nanotubes can be electrically switched on and off by the field generated by a gate electrode to produce field-effect transistors (FETs). Unfortunately, no methods exist for reliably preparing only metallic or semiconducting nanotubes either by selective synthesis or through post-synthesis separation. This lack of control, compounded by the single-walled nanotubes' tendency to bundle together to form ropes, has been the primary roadblock toward any nanotube-based electronic technology. Our work demonstrates a simple and reliable method for permanently modifying individual MWNTs and SWNT ropes to tailor their properties. Carbon nanotubes can withstand remarkable current densities (102-103 times higher than normal metals). At high enough currents, however, nanotubes ultimately fail. We found that we can control this breakdown to remove individual shells one at a time from MWNTs, or selectively destroy metallic tubes in a SWNT bundle. The different shells of a MWNT can have different electrical properties. As a result, the electronic structure and electron transport mechanisms of MWNTs have remained controversial. On the other hand, such complexity opens the possibility of selecting and using only the one MWNT shell that has the desired properties. The breakdown of MWNTs occurs in air at a certain threshold power, through the rapid oxidation of the outermost carbon shell. During breakdown the current flowing through the MWNT shows a step-like behavior with steps of surprising constancy corresponding to the breakdown of individual shells. By controlling the shell-by-shell removal process we can generate tubes with the desired outer shell characteristics, metallic or semiconducting, and most importantly, by selecting the diameter of the outer shell, we can obtain the desired band-gap Eg (note that Eg}1/diameter). If a SWNT rope is used to fabricate an FET, the metallic tubes in it cannot be switched by the gate field and will dominate the transport properties of the device. We have solved this problem by again using selective breakdown. Unlike in MWNTs, in a thin rope each SWNT can connect independently to the external electrodes. Thus, a MWNT rope may be modeled as independent, parallel conductors with total conductance G(Vg) = Gm + Gs(Vg), where Gm is the contribution of the metallic nanotubes and Gs is the gate-dependent conductance of the semiconducting nanotubes. In addition, multiple SWNTs within a rope are in contact with air, a potentially oxidizing environment, so many nanotubes can fail at once, rather than the uniform, shell-by-shell failure observed in MWNTs. Finally, the SWNTs within a small rope do not electrostatically shield each other as effectively as the concentric shells of a MWNT. As a result, a gate electrode can be used to effectively deplete the electrical carriers (electrons or holes) in the semiconducting SWNTs within the rope; in effect turning the semiconducting tubes into insulators. In this state, current-induced oxidation can be directed solely at the metallic SWNTs within the rope. We have taken advantage of these properties to selectively destroy the metallic nanotubes in SWNT ropes, while preserving the semiconducting SWNTs. Furthermore, since rope formation does not inhibit the FET fabrication, concentrated solutions of nanotubes can be used which allow dense arrays of FETs to be fabricated. Fabrication of these arrays is accomplished simply by depositing SWNT ropes on an oxidized Si wafer, and then an array of source, drain, and side-gate electrodes is fabricated lithographically on top of the ropes. The concentration of the tubes is pre-adjusted so that on the average there is one rope bridging the source and drain. No special arrangement/orientation of the nanotubes is required. The back gate (the Si wafer itself) is used to deplete the semiconducting tubes, followed by the application of a stress voltage to destroy the metallic tubes in the ropes, thus producing the NT-FETs. This new capability sidesteps the need for selective nanotube synthesis and assembly and has allowed us to fabricate dense arrays of nanotube FETs. As a result of our findings, we believe that we have made an important step toward the realization of an electronic technology using carbon nanotubes as its key elements, such as switches and interconnects.