SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:

SPIE Photonics West 2019 | Call for Papers

SPIE Defense + Commercial Sensing 2019 | Call for Papers

2018 SPIE Optics + Photonics | Register Today



Print PageEmail Page


High-frequency enhanced multichannel carbon nanotube transistors

Experiment validates the theoretically predicted high-frequency performance of carbon nanotube field-effect transistors.
31 February 2007, SPIE Newsroom. DOI: 10.1117/2.1200612.0502

Carbon nanotubes (CNTs) exhibit a number of remarkable electronic and mechanical properties that make them most attractive for micro- and nanoelectronic applications. Using a semiconducting single-walled nanotube (SWNT) to model the channel of a field-effect transistor (FET), a recent theoretical study1 predicted that this device family may be faster than conventional field-effect devices and suitable for high-frequency (HF) operation. Experimental verification, however, remains very challenging. An important issue is the low current density in CNTs due to their small diameter (a few nanometers). This results in poor accuracy when measuring the scattering parameters that provide information about input/output impedances.

The problem occurs because accurate HF characterization of an active device requires that the mismatch between its input impedance and that of the 50Ω reference impedance of most HF measuring devices be minimized so as to ensure good power transfer. To improve the sensitivity of the measurement, we are then faced with two major problems: the high value of the SWNT input impedance and its low driving current. According to quantum mechanics, the resistance of an ideal SWNT is greater than 6kΩ. The driving current is then of the order of a few microamps for a few applied volts. This inherent mismatch between a single channel CNTFET and typical radio-frequency equipment prevents the direct measurement of HF electrical parameters.

Methods have been developed to explore the HF properties of CNTFETs.2,3 In some, the CNT device operates as part of a circuit, and the circuit response gives an indication of its HF properties. In others, the impedance mismatch is an important factor in frequency limitation. To circumvent the problem, we chose to fabricate CNTFET devices using parallel arrays of nanotubes, shown to improve impedance matching.4,5 We found the approach particularly well suited for HF figure-of-merit determinations (S-parameters, cutoff frequencies, and electrical circuit modeling). The resulting process technology and device topology allows direct measurement of HF properties.5

Our improvements in device topology are illustrated in Figure 1, which shows a CNTFET structure designed to reduce losses in the passive part of the device. We used a high-resistivity silicon substrate to reduce dielectric loss and a coplanar transmission line with a very short access line. A dual-gate device topology was also adopted for two main reasons: first, to lower the gate access resistance, and second, to increase the device current.

Figure 1. (a) The structure of our carbon nanotube (CNT) field-effect transistor is shown with its coplanar transmission line design for HF characterization. (b) In this cross-sectional view, the implanted silicon area acts as a gate.

CNTs were deposited on the gate area using a low-temperature solution process6 with evaporation of an organic layer on the gate region that can act as a sticky patch to improve the selective deposition of the nanotubes. Their density depends on process parameters, and AFM (atomic force microscope) images of the gate region were acquired to ensure that the CNTs were randomly oriented.

We first fabricated devices with an implanted doping semiconductor region acting as the gate. We then proceeded with direct measurements of S-parameters, which for the first time yielded the active properties of a FET up to 1GHz. An electrical model of the active device also performed up to 10MHz. The model, which took into account the device topology and the physical properties of the CNTs, proved very useful in bettering device performance. For example, the measured DC characteristics showed the coexistence of metal and semiconductor species in the device channel, linked to the high conductance value of the device in the electrical model.

Based on the model, a number of improvements could be made in the process.7 We replaced the semiconductor gate with a metallic gate, and the parasitic capacitances of the structure were reduced as shown in Figure 2. We also increased the nanotube density, with good uniformity on a number of devices (see Figure 3). As a consequence, a current gain cutoff frequency of 8GHz was obtained.

Figure 2. Shown is a scanning electron microscope image of the two-finger metallic back gate of the CNT field-effect transistor active region (a) and its cross-sectional view (b). Aluminum is used as a back gate, and the Al2O3 layer is obtained by thermal oxidation. SWCNT: Single-walled CNT.

Figure 3. (a) An atomic force microscope image shows the random arrangement of the CNTs during deposition. (b) A graph of the transconductance measured on several devices illustrates process uniformity. Bias conditions: VDS = 1V, VGS = 0V.

HF characterization of our CNTFET is progressing, and we are now tackling the challenge of demonstrating operation in the terahertz range. The multichannel device represents a move forward in design, and it is now possible to explore circuit potentialities in the gigahertz range. Nevertheless, the effectiveness of CNTFETs is limited by parasitic capacitances related to the device structure, by random orientation of the tubes, and by metallic CNTs in the device channel (current methods of synthesizing CNTs produce a mix of metallic tubes and semiconductor tubes). Our next steps will focus on these issues to improve HF performance of CNTs over a few tens of gigahertz.

Henri Happy
University of Lille 1
Villeneuve d'Ascq, FRANCE