self aligned ingaas finfets with 5 nm fin width and 5 nm
play

Self-Aligned InGaAs FinFETs with 5-nm Fin-Width and 5-nm - PowerPoint PPT Presentation

Self-Aligned InGaAs FinFETs with 5-nm Fin-Width and 5-nm Gate-Contact Separation Alon Vardi, Lisa Kong, Wenjie Lu, Xiaowei Cai, Xin Zhao, Jess Grajal* and Jess A. del Alamo Microsystems Technology Laboratories, Massachusetts Institute of


  1. Self-Aligned InGaAs FinFETs with 5-nm Fin-Width and 5-nm Gate-Contact Separation Alon Vardi, Lisa Kong, Wenjie Lu, Xiaowei Cai, Xin Zhao, Jesús Grajal* and Jesús A. del Alamo Microsystems Technology Laboratories, Massachusetts Institute of Technology, Cambridge, MA, U.S.A *ETSI Telecomunicación, Universidad Politécnica de Madrid, Madrid, Spain Dec. 5, 2017 Sponsors: DTRA (HDTRA 1‐14‐1‐0057) NSF E3S STC (grant #0939514) Lam Research Korea Institute of Science and Technology 1

  2. FinFETs Intel Si Trigate MOSFETs Double gate Tri gate FinFETs used in state‐of‐the‐art Si CMOS • improved short‐channel effects • smaller footprint • but… higher parasitics 2

  3. Si and InGaAs FinFETs normalized by gate periphery 3.5 3.0 Planar Si 2.5 32 nm g m [mS/  m] Si FinFETs 2.0 14 nm 22 nm 1.5 1.0 0.5 0.0 // 0 10 20 30 Planar W f [nm] • Si planar  FinFET: performance ↓ 3

  4. Si and InGaAs FinFETs normalized by gate periphery normalized by fin footprint AR=H c /W f 3.5 20 5.3 3.0 W f Si FinFETs (V DD =0.8 V) Planar Si 15 2.5 32 nm 4.3 g m /W f [mS/  m] g m [mS/  m] Si FinFETs 2.0 H c 14 nm 10 22 nm 1.5 1.0 5 0.5 0.0 // 0 // 0 10 20 30 Planar 0 10 20 30 Planar W f [nm] W f [nm] • Si planar  FinFET: performance ↓, performance per footprint ↑ • Key challenge for FinFETs  efficient transport on sidewalls 4

  5. Si and InGaAs FinFETs normalized by gate periphery normalized by fin footprint Planar InGaAs (V DD =0.5 V) Lin, EDL2016 AR=H c /W f 3.5 20 5.3 3.0 W f Si FinFETs (V DD =0.8 V) Planar Si 15 2.5 32 nm 4.3 g m /W f [mS/  m] g m [mS/  m] Si FinFETs 2.0 H c 14 nm 10 22 nm 1.5 1.0 5 0.5 0.0 // 0 // 0 10 20 30 Planar 0 10 20 30 Planar W f [nm] W f [nm] • III‐V planar ~ Si planar 5

  6. Si and InGaAs FinFETs normalized by gate periphery normalized by fin footprint Planar InGaAs (V DD =0.5 V) Lin, EDL2016 Target: W F =5 nm 3.5 3.5 20 5.3 3.0 3.0 W f Si FinFETs (V DD =0.8 V) Planar Si 15 2.5 2.5 32 nm 4.3 g m /W f [mS/  m] g m [mS/  m] g m [mS/  m] Si FinFETs Si FinFETs 5.7 2.0 2.0 H c 14 nm 10 22 nm InGaAs FinFETs 1.5 1.5 1.0 1.0 5 0.5 0.5 0.0 0.0 // 0 // 0 0 10 10 20 20 30 30 Planar Planar 0 10 20 30 Planar W f [nm] W f [nm] W f [nm] • g m (III‐V FinFETs) < g m (Si) • Target of W f =5 nm yet to be demonstrated 6

  7. Si and InGaAs FinFETs normalized by gate periphery normalized by fin footprint Planar InGaAs (V DD =0.5 V) Lin, EDL2016 Target: W F =5 nm 3.5 20 5.3 3.0 W f Si FinFETs (V DD =0.8 V) Planar Si 15 2.5 32 nm 4.3 g m /W f [mS/  m] g m [mS/  m] Si FinFETs 5.7 2.0 H c 14 nm InGaAs FinFETs 10 22 nm InGaAs FinFETs 1.5 1.0 5 0.5 0.0 // 0 // 0 10 20 30 Planar 0 10 20 30 Planar W f [nm] W f [nm] • III‐V FinFET: W f < 20 nm  g m ↓ • Challenge: Improve III‐V sidewall conductivity 7

  8. MIT InGaAs FinFET’s Gen. #2 vs. #1 Mo Gen #1: SiO 2 L f,g • Wet cap recess Mo/W 60 nm • 3 Digital etch cycles SiO 2 InGaAs channel Mo/W InP • 40 nm channel height InGaAs ~20 nm • δ‐doping InAlAs Wet recess Gen. #1: Vardi et al., VLSI 2016, EDL 2016 Gen #2: Mo • Dry cap recess HSQ SiO 2 L f SiO 2 • 5 Digital etch cycles Mo/W HfO 2 Mo/W L g • 50 nm channel cap cap 40nm ~5 nm Cap • Fin‐top passivation InGaAs channel • Remove δ‐doping (in 2 nd stage) Channel InAlAs Dry+DE recess Gen. #2: This work 8

  9. Process Technology: contact-first W direction L g direction SiO 2 Mo/W InP stopper InGaAs cap InGaAs channel  ‐Doping InAlAs Lin, IEDM 2013 Lu, EDL 2014 Contact deposition Vardi, EDL 2014 • Yield R C < 10 Ω∙µm 9

  10. Dry+Digital Etch cap recess W direction L g direction L f Mo/W L g InP stopper SiO 2 InGaAs cap Mo/W InGaAs channel 40nm ~5 nm Cap  ‐Doping InAlAs Channel Dry+DE recess Contact etch Reactive ion etching Digital etching • No metal pullback (Lin, IEDM 2013) • III‐V cap pullback only during digital etch 10

  11. Dry+Digital Etch fin definition W f HSQ W f direction Fin H c H f Zhao, EDL 2014, Vardi, VLSI 2016 L g direction HSQ SiO 2 L g 8 nm Mo/W cap 170 nm InGaAs channel InAlAs 100 nm Fin etch Cap recess fin • BCl 3 /SiCl 4 /Ar RIE + 5 DE cycles : smooth, vertical sidewalls and high aspect ratio (>10) 11

  12. Gate stack - Double gate FinFET W direction HSQ W f Top fin passivation HfO 2 H c H f Mo L g direction HSQ SiO 2 L g Mo/W InP stopper InGaAs cap InGaAs channel  ‐Doping InAlAs Cap recess Contact deposition Fin etch Gate stack fin • HSQ stays on top of fins  double‐gate FinFET • Gate oxide – 3 nm HfO 2 12 (vs. 2.3 nm in EDL2016)

  13. Device cross section FIB cross section in L g direction TEM of finished device in W f direction 5 nm • Fin pitch: 200 nm • 10‐200 fins/device • W f : 5‐25 nm • L g : 30 nm – 5 µm • Contact to channel separation set by DE : ~ 5 nm 13

  14. Electrical characteristics: W f =5 nm, L g =50 nm 700 250 1E-3 V GS =-0.2 to 0.5 V L g =50 nm g m,max =600 µS/µm V DS =500 mV V DS =0.5 V 600  V GS =0.1 V W f =5 nm L g =50 nm 1E-4 200 50 mV 500 W f =5 nm 1E-5 S sat =100 mV/dec 150 g m [  S/  m] 400 I d [  A/  m] I d [A/  m] S lin =73 mV/dec 1E-6 300 100 1E-7 200 50 1E-8 100 0 0 1E-9 0.0 0.1 0.2 0.3 0.4 0.5 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.2 0.0 0.2 0.4 0.6 0.8 V GS [V] V GS [V] V GS [V] Normalized by gate periphery • Well behaved devices with W f =5 nm 14

  15. On/Off performance: fin width scaling 2.0 V DS = 0.5 V V DS = 0.5 V 180 1.5 S sat [mV/dec] 120 g m [mS/  m] 1.0 60 0.5 L g =40‐60 nm L g =40‐60 nm 0 0.0 0 5 10 15 20 25 0 5 10 15 20 25 W f [nm] W f [nm] • W f ↓: S sat ↓ • W f ↓: g m ↓ 15

  16. To improve Off performance: remove δ-doping Gen. #2 ‐ Undoped Gen. #2 ‐ δ‐doped Gen. #1 Mo Mo HSQ HSQ SiO 2 SiO 2 Mo Mo cap cap cap cap InGaAs InGaAs  ‐Doping InAlAs Extrinsic area fin • δ‐doping  R sd ↓ • Dry gate recess allows to remove δ‐doping • Impact on the intrinsic fin transport 16

  17. W F =5 nm FinFET: Electrical characteristics: δ-doped vs. undoped 700 250 1E-3 V GS =-0.2 to 0.5 V L g =50 nm g m,max =600 µS/µm V DS =500 mV V DS =0.5 V 600  V GS =0.1 V W f =5 nm L g =50 nm g m,max =500 µS/µm 1E-4 200 50 mV 500 W f =5 nm 1E-5 S sat =100 mV/dec 150 g m [  S/  m] 400 I d [  A/  m] I d [A/  m] S lin =73 mV/dec 1E-6 300 100 1E-7 200 S sat =75 mV/dec 50  doped 1E-8 100 S lin =65 mV/dec undoped 0 0 1E-9 0.0 0.1 0.2 0.3 0.4 0.5 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.2 0.0 0.2 0.4 0.6 0.8 V GS [V] V GS [V] V GS [V] Normalized by gate periphery Undoped fins: • better OFF performance • Undoped Similar ON performance 17

  18. Electrical characteristics: δ-doped vs. undoped 2.0 V DS = 0.5 V 180 Undoped Undoped  -doped  -doped 1.5 S sat [mV/dec] 120 g m [mS/  m] 1.0 60 0.5 L g =40‐60 nm L g =40‐60 nm 0 0.0 0 5 10 15 20 25 0 5 10 15 20 25 W f [nm] W f [nm] • Undoped fin: improved electrostatics • For W f <20 nm undoped‐fin ON performance also better 18

  19. Electrical characteristics: V T rolloff δ‐doped Undoped W f =5 nm 0.5 0.5 W f =9 W F ↓ 0.4 0.4 W f =13 W f =17 0.3 0.3 W F ↓ W f =21 0.2 0.2 W f =25 V T,lin [V] V T,lin [V] 0.1 0.1 0.0 0.0 -0.1 -0.1 -0.2 -0.2 0 200 400 600 0 200 400 600 L g [nm] L g [nm] Undoped fins smaller variation of V T with W F  Improved V T rolloff 19

  20. Benchmarking normalized by gate periphery 3.5 Undoped 3.0  -doped 2.5 InGaAs FinFETs g m [mS/  m] Si FinFETs 2.0 H c H c 1.5 1.0 0.5 0.0 0 10 20 30 40 W f [nm] • Systematic g m degradation for W f < 15 nm for both δ‐doped and undoped structures • No improvement from increased #DE cycles • Higher EOT  lower g m w.r.t. to Gen. 1 20

  21. Benchmarking normalized by gate periphery Target: W F =5 nm 3.5 20 W f Undoped 5.3 3.0 Si FinFETs (V DD =0.8 V)  -doped 15 2.5 4.3 10 g m /W f [mS/  m] Undoped InGaAs FinFETs g m [mS/  m] Si FinFETs 2.0 H c H c  -doped 10 1.5 1.0 5 InGaAs FinFETs 0.5 0.0 0 0 10 20 30 40 0 5 10 20 30 40 W f [nm] W f [nm] • Record W f with good electrical performance • Approaching Si FinFETs even at V DD =0.5 V • Record AR=10 21

  22. Long-channel Mobility vs. W f Capacitance measured @ 1GHz δ‐doped undoped W f =25 nm W f =9 nm • Strong µ degradation as W f ↓ • W f < 10 nm  µ independent of n l 22

  23. Simulations – charge distribution ON state: n l =3x10 7 cm ‐1 W f =25 nm undoped δ‐doped y • Undoped fin: better use of sidewalls • δ‐doped fin: conduction close to lower facet of channel 23

  24. Simulations – charge distribution ON state: n l =3x10 7 cm ‐1 W f =9 nm W f =25 nm undoped undoped δ‐doped δ‐doped y • Narrow fin: volume inversion in both δ‐doped and undoped fins 24

  25. Simulations – capacitance ON state: n l =3x10 7 cm ‐1 (V GT ~ 0.4V) W f =9 nm W f =25 nm undoped undoped undoped undoped Simulation Simulation δ‐doped δ‐doped δ‐doped δ‐doped E s (y) y Simulation Simulation • Reasonable agreement between measurement and simulations  extract n l ‐ E s relation 25

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend