Appl. Phys. A 74, 329–332 (2002) / Digital Object Identifier (DOI) 10.1007/s003390201280 Applied Physics A Materials Science & Processing Multishell conduction in multiwalled carbon nanotubes P.G. Collins 1,2, * , Ph. Avouris 2 1 Covalent Materials, Emeryville, CA 94608, USA 2 IBM T.J.Watson Research Center, Yorktown Heights, NY 10598, USA Received: 31 August 2001/Accepted: 3 December 2001/Published online: 4 March 2002 –  Springer-Verlag 2002 Abstract. The full electronic complexity of multiwalled car- bon nanotubes may be explored by sequentially removing in- dividual carbon shells. This technique is employed to directly measure the number of shells contributing to conduction at room temperature, as well as the contribution of each shell to the overall conductance. By exploring the gate dependence of the conductance, the random alternation between semicon- ducting and metallic shells can also be observed. PACS: 73.50.Fq; 72.10.Di; 73.61.Wp In the nascent field of molecular nanoelectronics, few ma- terials show as much promise as carbon nanotubes. Nano- tubes have been used to fabricate fantastically small elec- tronic devices [1–4], as well as to probe novel low tempera- ture phenomena [5–8]. Probably the most unusual property of nanotubes is the sensitivity of their bandstructure to geom- etry [9–11]. Due to this sensitivity, a collection of nearly identical carbon nanotubes will actually be a composite of both metallic and semiconducting conductors. To resolve this complexity, experimental progress has focused on single- walled nanotubes (SWNTs), each of which has a well-defined electronic structure. The characterization and understanding of multiwalled nanotubes (MWNTs) has proceeded more slowly because these materials are extremely complex conductors. High pu- rity, arc-grown MWNTs tend to have between five and thirty concentrically nested carbon shells. Each shell in a MWNT is geometrically unique from the next, so in principle each shell is electronically distinguishable from its neighbors. As the in- dividual shells interact weakly (in the same way that graphene sheets are weakly coupled in graphite), a MWNT can be mod- eled as a complex bundle of parallel conductors, each having a different band structure. Recently, we have outlined an electrical procedure for the destructive removal of the outermost carbon shell from a MWNT [12, 13]. The method uses current-induced oxida- tion and is selective because only the outermost shell is in * Corresponding author. (Fax: +1-510/658-0425, E-mail: pg_collins@excite.com) contact with the oxidizing environment (air) at any given moment. With repeated use of the technique, all of the var- ious shells of a MWNT can be removed. Albeit inefficient for counting the shells of a MWNT, the method’s unique strength is an ability to electrically characterize each shell in a MWNT. MWNT electronic devices were fabricated by first dis- persing arc-grown MWNTs in dichloroethane. After cen- trifugation to remove the particulate matter, the dispersions were applied to Si substrates having pre-patterned arrays of Au electrodes. By controlling the dispersion density, devices could be reliably made having an individual MWNT in con- tact with multiple electrodes. Figure 1a is a scanning electron micrograph image of a typical device, though most devices were exclusively characterized by atomic force microscopy to avoid deposition of amorphous hydrocarbon films. After a brief low temperature anneal to remove residual solvents, the difference between four-probe and two-probe resistance measurements was typically between 200 and 2000 Ω. Such low values indicate that our contacts are extremely well coupled and that this standard difference technique is proba- bly not an accurate measure of the true contact resistance. The nanotube resistances ranged between 5 and 15 kΩ. Figure 1b exhibits various current–voltage ( I –V ) char- acteristics obtained from the same MWNT as carbon shells were sequentially removed. At the maximum current on each curve, a discrete event was recorded corresponding to the fail- ure of a single shell. By triggering on this event, multiple I –V s could be easily recorded as the MWNT grew thin- ner. In some cases, the rapid loss of two shells occurred, as indicated in the figure by the dashed line. The data set is sub- stantially the same as another previously published [13] and reproduced in Fig. 1c. The primary difference is that the sec- ond data set was obtained in vacuum, under which conditions the carbon shells can withstand significantly higher voltages before oxidizing [12]. The vacuum case exhibits the remark- able current saturation observed in nanotubes and discussed elsewhere [12–14]. Below, we focus on the characteristics common to both data sets, which are independent of the meas- urement environment. Because the MWNTs merely sit upon Au electrodes, most of the inner carbon shells are not in direct contact