Making a Carbon Nanotube Ring Oscillator

In March 2006, IBM (White Plains, N.Y.) announced that its researchers had built the first complete IC around a single carbon nanotube (CNT) molecule.^1,2 The ring oscillator circuit was built using standard semiconductor processes and a single molecule as the base for all components in the circuit, rather than linking together individually constructed components. This approach simplifies manufacturing and provides a consistency needed to more thoroughly test and adjust the nanotube material. By integrating the complete circuit around a single nanotube (Figure ), the IBM team observed circuit speeds nearly a million times faster than previously demonstrated circuits with multiple nanotubes. While this is still slower than the speeds achieved with today's silicon chips, the team believes that the new nanofabrication processes will eventually lead to superior performance with carbon nanotube electronics.

We discussed the carbon nanotube work with Zhihong Chen, research staff member at IBM's TJ Watson Research Center (Yorktown Heights, N.Y.). What follows are excerpts of that conversation.

SI: How significant is this development relative to other announcements that have been made with respect to CNT research (inside and outside IBM)?

Chen: Owing to their small body thickness and unique electronic properties, carbon nanotubes have attracted widespread attention and interest. The promise is that nanotubes will enable a more aggressive channel length scaling without the introduction of short channel effects and mobility deterioration. This may enable future nanoelectronics with higher device speed, higher device density, and lower power consumption. Carbon nanotube-based field-effect transistors have shown to outperform state-of-the-art silicon MOSFETs at the device level.

The five-stage CMOS type nanotube ring oscillator using palladium p-type gates and aluminum n-type gates. The upper right inset shows the nanotube itself with a diameter of ~2 nm.

The next important step is to implement circuits on individual nanotubes to take best advantage of the smallness of the same and to evaluate whether the aforementioned device advantages translate into a better circuit performance as well. Another important question is whether we need a new concept to build completely different electronics, or can we still take advantage of the well-developed silicon technology and simply replace the channel material with nanotubes? In this context, we demonstrated a complete nanotube circuit based on a conventional CMOS architecture. Known as the first ring oscillator built on an individual carbon nanotube, it demonstrates a complete circuitry functioned by a single molecule. This breakthrough represents the first step to demonstrate the possibility of incorporating nanotubes into nanoelectronics and conventional circuit architectures. It also enables the detailed study of the performance-limiting aspects in carbon nanotubes and offers a way to evaluate their potential as a platform for future nanoelectronics technology.

SI: You stated that the circuit speeds with the ring oscillator are nearly a million times faster than previously demonstrated circuits with multiple nanotubes. What are the reasons for the difference in speed? How will speed be further increased?

Chen: Compared with previous circuits built by connecting separate nanotube transistors, our single nanotube ring oscillator works at a frequency of 72 MHz, approximately five to six orders of magnitude faster than previous demonstrations. This improvement is a result of our compact design that eliminates, to a large extent, external parasitic capacitance contributions.

In order to increase the speed further, more compact design will be needed to eliminate more parasitics. Our goal is to ultimately benefit from the intrinsically superior transport properties of carbon nanotubes to enable the predicted Terahertz switching speed of carbon nanotube transistors, and translate this individual device performance into a high-performance circuit design.

SI: What is the greatest barrier to making CNT transistors and circuits manufacturable?

Chen: There are several challenges remaining in the field that still require more research and development. One of them is a synthesis scheme that reproducibly provides nanotubes with desired diameters and electronic properties. The other important requirement is to place and align nanotubes at the desired positions so that they can be used as a platform for large-scale nanoelectronics.

SI: What type of circuit layout is used for carbon nanotubes?

Chen: We adopted the conventional CMOS architecture in our nanotube ring oscillator circuit. In order to obtain n-type and p-type transistors on the same nanotube, we developed a novel gate workfunction scheme to control the threshold voltages of our carbon nanotube transistors. Without the use of complex chemical doping schemes, this new approach allowed us to realize both p- and n- type transistors, the necessary components for a high-performance CMOS-type architecture.

SI: What types of approaches are used to align carbon nanotubes?

Chen: There is only a single carbon nanotube involved in the entire circuit. The circuit layout is designed to fit the shape of the nanotube. There are no alignment needs in this prototype demonstration.

SI: Describe the manufacturing process.

Chen: The nanotube was grown on 100 nm SiO₂ substrate by chemical vapor deposition, with an ~2 nm diameter and 18 µm length. Palladium source/drain contacts were defined on top of the nanotube, followed by the deposition of an Al₂O₃ gate dielectric and metal gate on top of each transistor channel.

The five-stage CMOS ring oscillator consists of five p-type FETs and five n-type FETs. The p-type FETs use palladium gates, and the n-type ones use aluminum gates. The circuit layout is designed such that the same type FETs from contiguous inverter stages share the same source/drain contacts, which makes the circuit rather compact. To avoid any interference from the measurement setup, we added an identical CMOS inverter stage right next to the ring oscillator, and its output is directed to a spectrum analyzer. The complete circuit, including the aforementioned six inverter stages, is 9 µm in width along the nanotube length. One more inverter stage is fabricated on the same nanotube to investigate the electrical properties of the nanotube, such that the ideal parameter set for the ring oscillator measurement can be determined beforehand.

SI: Describe the challenge of measuring the currents in CNTs.

Chen: Due to their small bodies, carbon nanotubes carry microampere currents (despite the current density being much higher compared with other materials such as silicon), which is the main challenge for measurements using available instruments designed for semiconductors carrying milliampere currents. As a direct result of this small current, our ring oscillator has output impedance of a few megohms. This causes a severe mismatch with the 50 V input impedance of the spectrum analyzer and results in rather small signals. A proper measurement setup to improve the signal to noise ratio is critical for the measurements.

SI: What are the next steps in your research?

Chen: This is the first demonstration of a compact single carbon nanotube circuit using a CMOS-type architecture. In the future, we are planning to optimize the circuit layout to further reduce the parasitic contributions, and will eventually be able to probe the intrinsic AC performance of carbon nanotubes. Further work is also needed to optimize the nanotube source/drain metal contacts and gate dielectrics to improve the individual device performance enabling a better circuit function.