Anomalous Schottky Barriers and Contact Band-to-Band Tunneling in Carbon Nanotube Transistors David J. Perello, † Seong ChuLim, ‡ Seung Jin Chae, ‡ Innam Lee, † Moon. J. Kim, § Young Hee Lee, ‡, * and Minhee Yun †, * † Department of Electrical Engineering, University of Pittsburgh, Pittsburgh, Pennsylvania 15219, ‡ Department of Physics, Department of Energy Science, Sungkyunkwan Advanced Institute of Nanotechnology, Suwon 440-746, Republic of Korea, and § Department of Materials Science and Engineering, University of TexasODallas, Richardson, Texas 75080 C arbon nanotube field effect transis- tors (CNT-FETs) have high current- carrying capability, 1 on/off ratio greater than 10 6 , 2-4 and switchable polar- ity dependent upon environment and chemical treatment. 5,6 Rapid optimization and performance enhancement in CNT- FETs has occurred. 7 Nevertheless, the phys- ics and the underlying mechanisms for transport are still incomplete. For instance, experimentally distinguishing energy band alignment at the metal-CNT contact has proven to be exceedingly difficult 8,9 and scarce experimental evidence exists in the form of Schottky barrier heights. Initially, it was believed that gas adsorbates in the CNT channel induce a charge transfer from CNTs to adsorbates, resulting in p-type conduc- tion in ambient measurement conditions. 10,11 However, more recently it was argued that the adsorbate-induced pla- nar dipole layers at the interface are the key factor for determining the majority car- rier. 12 The model of the induced dipole layer at the interface generally explains the CNT device operation. For Au contacts, conduc- tion variations are attributed to the induced dipole moment due to oxygen adsorbates at the interface. In the absence of oxygen, Au electrons are spilled over to CNT. In the presence of oxygen however, strong charge transfer occurs from Au to oxygen, forming a dipole opposed to the spillover. 13 This di- pole layer depletes the CNT, pulling the Fermi level (E F ) of CNT toward the valence band maximum. Similarly, in the case of Ti, a commonly used contact metal, oxygen- induced surface potential variations were introduced to explain the dominant p-type behavior. 12 However, contrary to these ear- lier reports, McClain et al. observed that un- der atmospheric O 2 exposure, the range of off-state gate bias was extended signifi- cantly into the n-region until the electron conducting on-state weakly appeared. 14 This occurrence cannot be explained by a planar dipole model, nor is it consistent with an earlier oxygen doping model. The formation of a dipole layer at the interface (due to oxygen in this case) should give rise to a threshold voltage change exclusively. Conversely, increases in dopant (oxygen) concentrations should be followed by a sig- nificant increase in off-current 15 in contrast to the CNT devices which often exhibit the reverse phenomenon. In a circuit with multiple current paths, carriers will always traverse the path with the lowest resistance, even if the physical length of such a path is longer. When con- tacted with low work function metals, the electron contribution of the subsurface metal-covered CNT is often neglected be- cause of the longer conduction pathway *Address correspondence to leeyoung@skku.edu, miy16@pitt.edu. Received for review February 17, 2010 and accepted May 19, 2010. Published online May 28, 2010. 10.1021/nn100328a © 2010 American Chemical Society ABSTRACT Devices incorporating nanoscale materials, particularly carbon nanotubes (CNTs), offer exceptional electrical performance. Absent, however, is an experimentally backed model explaining contact-metal work function, device layout, and environment effects. To fill the void, this report introduces a surface-inversion channel model based on low temperature and electrical measurements of a distinct single-walled semiconducting CNT contacted by Hf, Cr, Ti, and Pd electrodes. Anomalous barrier heights and metal-contact dependent band-to-band tunneling phenomena are utilized to show that, dependent upon contact work function and gate field, transport occurs either directly between the metal and CNT channel or indirectly via injection of carriers from the metal- covered CNT region to the CNT channel. The model is consistent with previously contradictory experimental results, and the methodology is simple enough to apply in other contact-dominant systems. KEYWORDS: carbon nanotubes · Schottky barrier · electrical transport · band to band tunneling · work function ARTICLE www.acsnano.org VOL. 4 ▪ NO. 6 ▪ 3103–3108 ▪ 2010 3103