Nanometer-Scale I I I -V Electronics J. A. del Alamo Microsystems - PowerPoint PPT Presentation

Nanometer-Scale I I I -V Electronics J. A. del Alamo Microsystems Technology Laboratories, MIT The Age of Silicon Symposium MIT, July 25, 2014 Acknowledgements: • D. Antoniadis, A. Guo, D.-H. Kim, T.-W. Kim, D. Jin, J. Lin, W. Lu, A. Vardi, N. Waldron, L. Xia, X. Zhao • Sponsors: Intel, FCRP-MSD, ARL, SRC, NSF, Sematech, Samsung • Labs at MIT: MTL, NSL, SEBL 1

2 Moore’s Law = exponential increase in transistor density Intel microprocessors Moore’s Law

What if Moore’s Law had stopped in 1990? Cell phone circa 1990 GPS handheld device circa 1990 3

4 What if Moore’s Law had stopped in Laptop computer circa 1981 1980?

5 What if Moore’s Law had stopped in Desktop calculator 1970 1970?

What if Moore’s Law had never happened? Insulin pump circa 1960 “Personal calculator” circa 1960 1960 6

7 ? How far can Si support Moore’s Law? Moore’s Law

Transistor scaling  Voltage scaling Power management demands reduction in supply voltage. 6 5 Supply voltage (V) 4 3 2 1 Intel microprocessors 0 1980 1985 1990 1995 2000 2005 2010 2015 Year of introduction Supply voltage reduction saturating in recent years 8

Voltage scaling  Si transistor performance suffers Transistor current density: Transistor performance saturated in recent years 9

10

Options for post-Si CMOS NMOS:  GaAs, InGaAs, InP ~2x Si PMOS:  Ge, InGaSb ~2x Si Different lattice constant for n-FETs and p-FETs del Alamo, Nature 2011 11

Electron injection velocity: I nGaAs vs. Si Measurements in High Electron Mobility Transistors (HEMTs): del Alamo, Nature 2011 • v inj (InGaAs) increases with InAs fraction in channel • v inj (InGaAs) > 2v inj (Si) at less than half V DD • ~100% ballistic transport at L g ~30 nm 12

I nGaAs HEMT: high-frequency record vs. time f T =710 GHz (NCTU ) Teledyne/MIT: 800 f T =688 GHz 700 600 MIT devices 500 on InP f T (GHz) substrate 400 300 200 on GaAs 100 substrate 0 1980 1990 2000 2010 Kim, EDL 2010 Year Best high-frequency performance of any transistor on any material system 13

I nGaAs Electronics Today UMTS-LTE PA module Chow, MTT-S 2008 40 Gb/s modulator driver 77 GHz transceiver Carroll, MTT-S 2002 Tessmann, GaAs IC 1999 Bipolar/E-D PHEMT process Single-chip WLAN MMIC, Morkner, RFIC 2007 Henderson, Mantech 2007 14

I nGaAs HEMT vs. MOSFET HEMT not suitable for logic: too much gate leakage current MOSFET incorporates gate oxide  gate leakage suppressed 15

I nGaAs MOSFETs vs. HEMTs: historical evolution Lin, IEDM 2013 * *inversion-mode Progress reflects improvements in oxide/III-V interface 16

What made the difference? Atomic Layer Deposition (ALD) of oxide ALD eliminates surface oxides that pin Fermi level  “Self cleaning” Huang, APL 2005 Clean, smooth interface without surface oxides • First observed with Al 2 O 3 , then with other high-K dielectrics • First seen in GaAs, then in other III-Vs 17

I nGaAs MOSFET: possible designs Enhanced gate control  enhanced scalability 18

Self-Aligned I nGaAs Quantum-Well MOSFETs • Channel: In 0.7 Ga 0.3 As/InAs/In 0.7 Ga 0.3 As (1/2/5 nm) • Gate oxide: HfO 2 (2.5 nm, EOT~0.5 nm) 1.0 V gs -V t = 0.5 V L g =20 nm • Self-aligned contacts (L side ~5 nm) R on =224  m 0.8 I d (mA/  m) 0.4 V • Si compatible process (RIE, metals) 0.6 0.4 0.2 0.0 Lin, IEDM 2013 0.0 0.1 0.2 0.3 0.4 0.5 V ds (V) 5 nm 19

I nGaAs double-gate Fin-MOSFET Key enabling technologies: • BCl 3 /SiCl 4 /Ar RIE • digital etch 40 nm 30 nm Vardi, DRC 2014 Zhao, EDL 2014 20

Vertical nanowire I nGaAs MOSFETs 30 nm diameter InGaAs NW-MOSFET Zhao, IEDM 2013 Zhao, EDL 2014 • Nanowire MOSFET: ultimate scalable transistor • Vertical NW: uncouples footprint scaling from L g scaling • Top-down approach based on RIE + digital etch 21

Si integration: I nGaAs SOI MOSFETs III ‐ V bonded SOI process Mo n + cap of IBM Zurich: InGaAs channel Czornomaz, IEDM 2012 BOX p-Si InP donor wafer BOX: Al 2 O 3 InGaAs channel n + cap BOX BOX InP donor wafer InP donor wafer p-Si p-Si 1. MBE growth 2. ALD Al 2 O 3 3. Wafer bonding 4. InP etch back 22

Conclusions: exciting future for I nGaAs electronics on Silicon • Most promising material for ultra-high frequency and ultra-high speed applications  first THz transistor? • Most promising material for n-MOSFET in a post- Si CMOS logic technology  first sub-10 nm CMOS logic? • InGaAs + Si integration:  THz + CMOS + optics integrated systems? 23