A ring oscillator wires an odd number of inverters in series and then connects the last inverter's output back to the input. Given the odd number of inverters, the output state will be the opposite of the input state (1 or 0), thus creating an unstable circuit that will run at the fastest speed possible for the given design and process technology.
The inverters make up complementary n- and p-type transistors, gates wired together as the input, drains as the output; the sources of the n-type and p-type transistors, respectively, are wired to the supply's negative and positive voltages. When a positive voltage is input to the inverter, the p-type transistor turns off and the n-type turns on, inverting the input to a negative voltage, which is fed to the next inverter.
IBM's first design challenge was learning how to craft n-type and p-type nanotube transistors, a feat it touched on recently when it described the unusual operating regions possible with nanotubes (go to www.eetimes.com and search article ID: 181500304). IBM had previously reported doping nanotubes to make n-type and p-type devices. But its new method dispenses with doping, instead harnessing the nanotube transistor channels' unusual operating regions compared with those of silicon.
A silicon transistor has a sigmoid transfer function, staying off for voltages around zero but either ramping up for voltages above zero (p-types) or ramping down for negative voltages (n-types). And when turned on, a silicon transistor saturates at a fixed voltage above or below zero.
Nanotube transistors, by contrast, have a V-shaped transfer function and turn off for zero voltage and on for voltages either above or below zero.
So instead of doping the nanotubes to make p- and n-types, IBM used different metals for the gate electrodes--palladium for p-types and aluminum for n. That shifted the "zero" point of each transistor's V-shaped transfer function so that their "zero points" were about 0.7 volt apart, making the transfer functions look like a "W," with the center branches overlapping, when plotted. The researchers then used the region above where the two V's crossed to get traditional n-type and p-type transfer functions from the same underlying carbon nanotube channel.
"We used the p-branch of the W for one transistor and the n-branch for the other transistor," Appenzeller said. "That is the heart of what I am really proud of. It's how we avoided having to dope our nanotubes."
The second major obstacle IBM had to overcome was how to use a nanotube measuring only about 15 Å wide and 6 microns long as the channel for all five CMOS inverters of the ring oscillator. By using a single-molecule nanotube whose characteristics were uniform along its entire length, IBM was able to sidestep the problems that have foiled previous attempts to create nanotube-based ICs.
Collaborating with Avouris and Appenzeller were Zhihong Chen, Yu-Ming Lin and Paul Solomon, all members of the technical staff at IBM's T.J. Watson center.; Florida researchers Jennifer Sippel-Oakley and Andrew Rinzler; and Columbia's Jinyao Tang and Shalom Wind.
An animation of the nanotube ring oscillator is available online at www.research.ibm.com/nanotube_ringoscillator/.
See related image