PORTLAND, Ore.--IBM Research Zurich, Infineon Technologies AG and Globalfoundries Inc. have joined forces with a half dozen European universities and research centers on a mission to cut the power consumption of electronic devices by10-fold and reduce standby power to zero.
Called Steep—because of the steep cutoff slope leading to zero standby power—the European Union-funded project aims to perfect the tunneling field-effect transistor (T-FET) using nanowire channels of silicon, silicon germanium and III-V-on-silicon. The three-year $5.5 million effort aims to create processes that can be run on CMOS lines to facilitate a smooth switchover from today's complementary metal-oxide semiconductors (CMOS) to T-FETs.
"The tunneling FET [field effect transistor] has been known since 1994, but we want to give the T-FET a performance boost," said scientist Heike Riel at IBM Research Zurich. "A nanowire geometry gives optimal electrostatic performance for its gates, and the use of novel materials such as III-V-on-silicon, along with contemporary high-k dielectrics and strain engineering, will improve the tunnel FET's performance as well as significantly lower the voltage needed to operate it."
In a traditional CMOS chip, complementary p-n-p type (PNP) and n-p-n type (NPN) transistors are used. T-FETs, on the other hand, use a p-intrinsic-n (PIN) architecture, where the transistor channel is the undoped intrinsic channel between a p-doped source and the n-doped drain, controlled by a gate that enhances or shuts off tunneling across the reverse-biased channel.
To get high performance with a .5 volt power supply, IBM plans to cast the nanowire channels into III-V materials like indium-arsenide, which is not possible using traditional mask lithography, due to the lattice mismatch between III-V materials and silicon. IBM, however, has already demonstrated epitaxial growth of III-V-on-silicon with a recent tunnel diode named after its inventor, Japanese physicist Leo Esaki.
"We have recently demonstrated our III-V-on-silicon growth technique for a Esaki tunnel diode using indium-arscenide-on-silicon," said Riel.
Dr. Heike Riel, who leads the nanoscale electronics group at IBM Research Zurich, is part of Project Steep.
To fabricate the indium-arsenide on silicon, instead of using catalysts or "seeds" to initiate grow, as is often done, IBM used an oxide mask.
"We use a catalyst free method," said Riel. "We start with an oxide mask on top of the silicon that is patterned with holes and then grow with selective epitaxy the III-V material directly onto the silicon in these holes."
Those vertical columns grow a transistor channel just a few nanometers in diameter at every location on the silicon wafer where you want to build a T-FET.
IBM told me that the turn-on slope of TFETs could be as steep as 10mV per decade, compared to a 60mV per decade limit for MOS-FETs. And the negative resistance at zero volts should reduce leakage to "almost" zero.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.