Imec, a nanoelectronics research center, announced today that it has successfully demonstrated the first III-V compound semiconductor FinFET devices integrated epitaxially on 300mm silicon wafers, through a unique silicon fin replacement process. The achievement illustrates progress toward 300mm and future 450mm high-volume wafer manufacturing of advanced heterogeneous CMOS devices, monolithically integrating high-density compound semiconductors on silicon. The breakthrough not only enables continual CMOS scaling down to 7nm and below, but also enables new heterogeneous system opportunities in hybrid CMOS-RF and CMOS-optoelectronics.
“To our knowledge, this is the world’s first functioning CMOS compatible IIIV FinFET device processed on 300mm wafers,” stated An Steegen, senior vice president core CMOS at imec. “This is an exciting accomplishment, demonstrating the technology as a viable next-generation alternative for the current state-of-the-art Si-based FinFET technology in high volume production.”
The proliferation of smart mobile devices and the ever growing user expectations for bandwidth and connectivity, will drive the continual need for software and hardware advancements that extend from networks to data servers and mobile gadgets. At the core of the hardware will be new process technologies that allow for more power-efficient CMOS transistors and increased integration, enabling a higher level of functionality. This prompts process technologies that enable heterogeneous devices spanning operating ranges for targeted circuits, maximizing the system performance.
Aaron Thean, director of the Logic R&D at imec commented: “During the last decade, transistor scaling has been marked by several leaps in process technologies to provide performance and power improvements. The replacement of poly-silicon gate by high-k metal-gate in 45nm CMOS technology in 2007 represented a major inflection in new material integration for the transistor. The ability to combine scaled non-silicon and silicon devices might be the next dramatic transistor face-lift, breaking almost 50 years of all-silicon reign over digital CMOS. This work represents an important enabling step towards this new paradigm.”
At the finest grain, co-integration of high-density heterogeneous transistors has been challenged by the ability to combine disparate materials and structures while maintaining low enough complexity and defectivity. Imec’s breakthrough process selectively replaces silicon fins with indium gallium arsenide (InGaAs) and indium phospide (InP), accommodating close to eight percent of atomic lattice mismatch. The new technique is based on aspect-ratio trapping of crystal defects, trench structure, and epitaxial process innovations. The resulting III-V integrated on silicon FinFET device shows an excellent performance.
Imec’s research into next-generation FinFETs is performed as part of imec’s core CMOS program, in cooperation with imec’s key partners including Intel, Samsung, TSMC, Globalfoundries, Micron, SK Hynix, Toshiba, Panasonic, Sony, Qualcomm, Altera, Fujitsu, nVidia, and Xilinx.
Imec is headquartered in Leuven, Belgium, and has offices in Belgium, the Netherlands, Taiwan, US, China, India and Japan. Its staff of more than 2,000 people includes more than 650 industrial residents and guest researchers. In 2012, imec’s revenue (P&L) totaled 320 million euro.
It seems too early to claim that selectively replacing silicon fins with III-V semiconductor materials such as InGaAs and InP , and demonstrating the functional test chips on 300mm wafers could break the Si based FinFETs. First, IMEC doesn’t show at what technology node the TEM image and corresponding Id-Vg device characteristics are based on. It appears to be at least 50nm node because the fin-width at the bottom of fin can be equal to but can’t be larger than Lg(gate length) or channel length in order to suppress short channel effects or FinFET leakage current. Therefore, the gate length is at least 50nm. The 50nm fin-width of the InGaAs/InP is estimated from TEM image. The linear and saturation electric currents of Imec’s InGaAs and InP are very poor with high leakage current for such a long channel device. I would like to see InGaAs/InP based FinFET at 22nm node so that it can be compared with Intel’s 22nm FinFETs that are in high volume manufacturing over two years. The 14nm will be manufactured in the first quarter of 2014. TSMC’s first 16nm FinFET will be presented at IEDM, 2013, and the volume manufacturing has started in 4th quarter, 2013. Major semiconductor companies will adopt si based FinFETs. There are a large number of papers published recently on very small quantum wells with InGaAs and InP that are built in a planar bulk silicon. Their electrical transfer characteristics are significantly superior to IMEC’s III-V FinFETs. I very doubt that the III-V semiconductors such as InGaAs and InP can break the Si based FinFETs. There is no viable alternative to the FinFET today. Sang KIM