Advanced Materials and Structures for Nanoscale CMOS Prof

  • Slides: 43
Download presentation
Advanced Materials and Structures for Nanoscale CMOS Prof. Tsu-Jae King Department of Electrical Engineering

Advanced Materials and Structures for Nanoscale CMOS Prof. Tsu-Jae King Department of Electrical Engineering and Computer Sciences University of California, Berkeley, CA 94720 -1770 USA March 29, 2004 FLCC Seminar

Outline • Introduction – MOSFET scaling • Advanced Gate-Stack Materials • Thin-Body MOSFETs •

Outline • Introduction – MOSFET scaling • Advanced Gate-Stack Materials • Thin-Body MOSFETs • Summary T. -J. King -2 - March 29, 2004

IC Technology Advancement Rapid advances in IC technology have been achieved primarily by scaling

IC Technology Advancement Rapid advances in IC technology have been achieved primarily by scaling down transistor lateral dimensions Technology Scaling ITRS 2003 Projection Investment Better Performance/Cost Market Growth T. -J. King -3 - March 29, 2004

Intel’s 90 nm CMOS Technology Used for volume manufacture of Pentium® and Intel®Centrino. TM

Intel’s 90 nm CMOS Technology Used for volume manufacture of Pentium® and Intel®Centrino. TM processors on 300 mm wafers • Lg = 50 nm • Tox = 1. 2 nm • Strained Si channel T. -J. King -4 - March 29, 2004

Bulk-Si MOSFET • Leakage current is the primary barrier to scaling • To suppress

Bulk-Si MOSFET • Leakage current is the primary barrier to scaling • To suppress leakage, we need to employ: – Higher body doping lower carrier mobility, higher junction capacitance, increased junction leakage – Thinner gate dielectric higher gate leakage – Ultra-shallow S/D junctions higher Rsd Lgate Metal-Oxide-Semiconductor Field-Effect Transistor: Desired characteristics: • High ON current (Idsat) • Low OFF current Gate Source Substrate T. -J. King Tox -5 - Leff Xj Drain Nsub March 29, 2004

SIA Int’l Technology Roadmap for Semiconductors (2003) Year Technolog y Node Tox 2004 200

SIA Int’l Technology Roadmap for Semiconductors (2003) Year Technolog y Node Tox 2004 200 5 90 nm 80 nm 200 6 200 7 201 0 201 3 201 6 70 nm 65 nm 45 nm 32 nm 22 nm Solutions Being No Known Solutions Pursued Idsat T. -J. King -6 - March 29, 2004

Approaching 1. 4 nm Tox, inv • Use high-k gate dielectric (by 2007) Thicker

Approaching 1. 4 nm Tox, inv • Use high-k gate dielectric (by 2007) Thicker physical thickness for given Cgate (F/cm 2) -> lower gate leakage current Hf. O 2 is a promising candidate Issues: Thermal stability Interfacial Si. O 2 layer -> increased Tox, inv - may be needed for good mobilities • Reduce/eliminate gate depletion effect Poly-Si. Ge or metal (by 2007? ) Issues: Process compatibility Work function T. -J. King -7 - March 29, 2004

Achieving Low Rsd • Increase source/drain dopant concentration Issue: Ultra-shallow junction formation • Use

Achieving Low Rsd • Increase source/drain dopant concentration Issue: Ultra-shallow junction formation • Use elevated source/drain structure Issue: Process complexity • Lower source/drain contact resistance Issue: New materials / process complexity T. -J. King -8 - March 29, 2004

Outline • Introduction • Advanced Gate-Stack Materials – Poly-Si. Ge gate electrode – Mo

Outline • Introduction • Advanced Gate-Stack Materials – Poly-Si. Ge gate electrode – Mo gate electrode • Thin-Body MOSFETs • Summary T. -J. King -9 - March 29, 2004

Poly-Si Gate Technology Issues No GDE • Gate depletion effect ® reduced Idsat~ Cox(Vdd-Vt)

Poly-Si Gate Technology Issues No GDE • Gate depletion effect ® reduced Idsat~ Cox(Vdd-Vt) CG • Boron penetration w/ GDE DEP ACC ® Vt variation ® degraded gate-dielectric reliability INV VG • Gate resistance • Remote charge scattering ® reduced carrier mobilities (M. Krishnan et al. , IEDM 1998) • Stability on high-k gate dielectrics T. -J. King - 10 - March 29, 2004

Why Poly-Si. Ge Gate? • Advantages: – – Reduced gate depletion effect (GDE) Less

Why Poly-Si. Ge Gate? • Advantages: – – Reduced gate depletion effect (GDE) Less boron penetration through gate oxide Straightforward process integration Work functions are appropriate • N+ poly-Si. Ge gate for NMOS • P+ poly-Si. Ge gate for PMOS • Optimal Ge content ~20% (W. -C. Lee et al. , IEEE Electron Device Letters, Vol. 19, p. 247, 1998) • Poly-Si. Ge on high-k gate dielectric? T. -J. King - 11 - March 29, 2004

Poly-Si. Ge/Hf. O 2 MOSFET Q. Lu et al. , 2002 Symposium on VLSI

Poly-Si. Ge/Hf. O 2 MOSFET Q. Lu et al. , 2002 Symposium on VLSI Technology C-V comparison • PVD Hf. O 2 at UT Austin Prof. Jack Lee’s group • Conventional CMOS process flow – 800 o. C, 30 m furnace anneal + 1000 o. C, 10 sec RTA • Poly-Si. Ge gate yields lower Tox (EOT)! • Low gate leakage current maintained T. -J. King - 12 - March 29, 2004

Poly-Si. Ge/Hf. O 2 Gate Stack XTEM analysis results Si 0. 75 Ge 0.

Poly-Si. Ge/Hf. O 2 Gate Stack XTEM analysis results Si 0. 75 Ge 0. 25 gate Si gate Hf. O 2 Si T. -J. King Si • Interfacial layer eliminated! • EOT reduction similar to that achieved with surface nitridation (SN) – SN -> degraded mobilities • Promising for low EOT with low leakage, good mobilities (? ) - 13 - March 29, 2004

Metal Gate Technology Required properties: – – Appropriate work function High Tm (>1000 C)

Metal Gate Technology Required properties: – – Appropriate work function High Tm (>1000 C) Stable interface with gate dielectric Compatible with Si processing – deposition, etch Candidate materials: – – T. -J. King High-Tm metals (Ta, Mo, Ru, W) metal nitrides (Mo. N, WN, Ti. N, Ta. N) doped metal oxides (In 2 O 3, Sn. O 2, Ru. O 2) metal silicides (Cr. Si, WSix) - 14 - March 29, 2004

Dual-Work-Function Metal Gate Technology Possible approaches: 1) Deposit & etch 1 st metal; deposit

Dual-Work-Function Metal Gate Technology Possible approaches: 1) Deposit & etch 1 st metal; deposit & etch 2 nd metal (Q. Lu et al. , 2000 Symposium on VLSI Technology) û exposes gate-dielectric to etch process 2) Deposit 1 st & 2 nd metals; selectively remove 2 nd metal; anneal to interdiffuse or alloy metals (I. Polishchuk et al. , 2001 MRS Spring Meeting) 3) Deposit one metal; selectively adjust FM (Q. Lu et al. , 2001 Symposium on VLSI Technology) T. -J. King - 15 - March 29, 2004

Why Mo Gate? • Advantages (vs. other metals): – Compatible with conventional process flow

Why Mo Gate? • Advantages (vs. other metals): – Compatible with conventional process flow sequence • stable on high-k gate dielectrics (Q. Lu et al. , 2001 Symp. VLSI Tech. ) – Work function can be tuned by incorporating nitrogen • Impact on gate dielectric can be minimized – Damageless sputtering of Mo H. Takeuchi et al. , DRC 2003 – Diffusion of N into Mo from Ti. N 1+x cap R. J. P. Lander et al. , MRS 2002 Spring Meeting – Tilted N+ implantation into Mo T. Amada et al. , MRS 2002 Spring Meeting T. -J. King - 16 - March 29, 2004

FM Tuning by Ion Implantation P. Ranade et al. , IEDM 2002 § FM

FM Tuning by Ion Implantation P. Ranade et al. , IEDM 2002 § FM can be lowered by N+ implantation and thermal anneal DFM increases with § dose § energy (N segregates to Si. O 2 interface & forms Mo 2 N) Anneal time = 15 m except for 900 o. C (15 s) TMo = 15 nm T. -J. King - 17 - March 29, 2004

FM Variation with Gate Dielectric FM (e. V) N+ poly-Si 4. 05 Theory FMo,

FM Variation with Gate Dielectric FM (e. V) N+ poly-Si 4. 05 Theory FMo, vac= 4. 95 e. V 5. 05 4. 76 5. 17 Si. O 2 Si 3 N 4 Expt. Data 4. 94 Zr. O 2 4. 97 P+ poly-Si Hf. O 2 Gate Dielectric Q. Lu et al. , IEDM 2000 Y. -C. Yeo et al. , 2001 Symp. VLSI Technology T. -J. King - 18 - March 29, 2004

The Metal-Dielectric Interface Intrinsic States and Dipole Formation Interface State Character Energy-Band Diagram intrinsic

The Metal-Dielectric Interface Intrinsic States and Dipole Formation Interface State Character Energy-Band Diagram intrinsic interface states Electron Energy Evac Ec EF, m Charge-neutrality level +q’ _ ECNL, d – q” Ev 0% low-FM metal interface dielectric T. -J. King - 19 - 100% % acceptor-like (solid) vs. donor-like (dashed) March 29, 2004

Effective FM of Metal Gate Energy-Band Diagram Evac FM, eff FM, vac EF, m

Effective FM of Metal Gate Energy-Band Diagram Evac FM, eff FM, vac EF, m Charging of intrinsic interface states FCNL, d Interface dipole formation Ec ECNL, d Effective FM differs from vacuum value FM, vac Ev metal interface dielectric T. -J. King Bands line up toward zero dipole charge position - 20 - For low-FM gate: FM, eff > FM, vac March 29, 2004

Implications for FM, vac Requirements • Pinning effect is larger for high-k dielectrics Þ

Implications for FM, vac Requirements • Pinning effect is larger for high-k dielectrics Þ To get desired FM, eff on high-k gate dielectric: – even smaller FM, vac is needed for NMOS – even larger FM, vac is needed for PMOS T. -J. King - 21 - March 29, 2004

Outline • Introduction • Advanced Gate-Stack Materials • Thin-Body MOSFETs – Ultra-thin body –

Outline • Introduction • Advanced Gate-Stack Materials • Thin-Body MOSFETs – Ultra-thin body – Double-gate • Summary T. -J. King - 22 - March 29, 2004

Thin-Body SOI MOSFETs (by 2007? ) Ultra-Thin Body Double Gate Source SOI Si. O

Thin-Body SOI MOSFETs (by 2007? ) Ultra-Thin Body Double Gate Source SOI Si. O 2 Drain TSi TBOX Source Tox Silicon Substrate Gate 1 Vg SOI Drain TSi Gate 2 Common feature: A thin body, such that no conduction path is far from the gate T. -J. King - 23 - March 29, 2004

Ultra-Thin-Body MOSFET • Lg = 12 nm • UTB suppresses leakage • Thick S/D

Ultra-Thin-Body MOSFET • Lg = 12 nm • UTB suppresses leakage • Thick S/D => low Rseries Gate Source SOI Drain Si. O 2 Silicon Substrate M. Takamiya et al. , Proc. 1997 ISDRS, p. B. Yu 215 et al. , Proc. 1997 ISDRS, p. 623 T. -J. King - 24 - Drain Current [A/um] • Tox = 2 nm 1. E-02 Simulated Id -Vg Vds=1 V 1. E-04 1. E-06 1. E-08 1. E-10 1. E-12 Subthreshold swing S (units: m. V/dec) Tsi=8 nm Tsi=6 nm Tsi=4 nm 0 0. 2 0. 4 0. 6 0. 8 1 Gate Voltage [V] March 29, 2004

20 nm Lg UTB CMOSFETs Krivokapic et al. (AMD), presented at the Int’l Conference

20 nm Lg UTB CMOSFETs Krivokapic et al. (AMD), presented at the Int’l Conference on Solid-State Devices and Materials (Tokyo, Japan) Sep. ’ 03. • TSi = 6 nm • Tox = 1. 3 nm (Si. O 2/Si 3 N 4) • Strained Si channel • Excellent Idsat achieved with thick S/D structure T. -J. King - 25 - March 29, 2004

UTB MOSFET Scaling • Issues for bulk-Si MOSFET scaling obviated – Body does not

UTB MOSFET Scaling • Issues for bulk-Si MOSFET scaling obviated – Body does not need to be heavily doped – Tox does not need to be scaled as aggressively – Ultra-shallow S/D junction formation is not an issue • Body thickness must be less than ~1/3 x Lg – Formation of uniformly thin body is primary challenge – For TSi < 4 nm, quantum confinement & interface roughness VT variation and degraded gm K. Uchida et al. , IEDM 2002 T. -J. King - 26 - March 29, 2004

Double-Gate “Fin. FET” • Self-aligned gates straddle thin silicon fin • Current flows parallel

Double-Gate “Fin. FET” • Self-aligned gates straddle thin silicon fin • Current flows parallel to wafer surface Source Gate Length = Lg Gate 2 Gate 1 Drain Current Flow Source Drain Fin Height = Hfin = W Fin Width Wfin = TSi T. -J. King - 27 - March 29, 2004

Lg TSi DRAIN GATE SOURCE 15 nm Lg Fin. FETs 10 nm SOURCE Output

Lg TSi DRAIN GATE SOURCE 15 nm Lg Fin. FETs 10 nm SOURCE Output Characteristics Y. -K. Choi et al. , IEDM Technical Digest, pp. 421 -424, 2001 - 28 - 20 nm DRAIN Lg < 20 nm; TSi = 10 nm; Tox = 2. 1 nm Sub-Threshold Characteristics T. -J. King GATE March 29, 2004

10 nm Lg Fin. FETs B. Yu et al. , IEDM Technical Digest, pp.

10 nm Lg Fin. FETs B. Yu et al. , IEDM Technical Digest, pp. 251 -254, 2002 T. -J. King - 29 - March 29, 2004

Fin. FET Scaling • Compared with UTB-MOSFET: – Reduced short-channel effects => more scalable

Fin. FET Scaling • Compared with UTB-MOSFET: – Reduced short-channel effects => more scalable – Higher current drive due to • steeper subthreshold swing (60 m. V/dec) • lower channel electric field => higher carrier mobilities • Fin width must be less than 2/3 x Lgate Formation of narrow fin is primary challenge • sub-lithographic process needed T. -J. King - 30 - March 29, 2004

Reducing Gate Leakage • DG-MOSFET provides lower Ig è Tox can be more aggressively

Reducing Gate Leakage • DG-MOSFET provides lower Ig è Tox can be more aggressively scaled (improved Idsat) L. Chang et al. , IEDM 2001 T. -J. King - 31 - March 29, 2004

Selective Deposition of Ge • Conventional LPCVD tool – Ge. H 4 gas, 340

Selective Deposition of Ge • Conventional LPCVD tool – Ge. H 4 gas, 340 o. C, 300 m. T • Ge deposits selectively onto Si XTEM of UTB MOSFET w/ raised Ge S/D Ge Si T. -J. King Si. O 2 Si 3 N 4 XTEM of selectively grown Ge Gate Ge Source Si. O 2 - 32 - Ge Drain Si TSi = 3 nm March 29, 2004

Fin. FET with Raised Ge S/D • Ge can be selectively deposited on top

Fin. FET with Raised Ge S/D • Ge can be selectively deposited on top of Si fin(s) Plan-view SEM of sub-50 nm Lgate multi-fin device after selective Ge XTEM of Si fin with selectively deposited Ge Gate Selective Ge Si fin Source Si. O 2 Si Fins T. -J. King Spacer Drain 50 nm - 33 - March 29, 2004

Why Si. Ge for S/D Contacts? • Lower resistivity than Si (for p+ doping)

Why Si. Ge for S/D Contacts? • Lower resistivity than Si (for p+ doping) • Selective deposition onto Si is easy – thicker S/D contact regions lower Rsd • Low specific contact resistivity – smaller bandgap smaller Schottky barrier rc ~10 -8 W-cm 2 for germanosilicides on Si. Ge ITRS requirement for rc Prof. Ozturk’s group at NCSU => meets ITRS requirement! T. -J. King - 34 - March 29, 2004

Idsat Improvement w/ Raised S/D Lgate = 90 nm; Wfin = 70 nm; W

Idsat Improvement w/ Raised S/D Lgate = 90 nm; Wfin = 70 nm; W = 2 x Hfin = 100 nm Additional process steps: w/ Ge w/o Ge • • • Remove Si. O 2 over S/D Selectively grow 70 nm Ge Implant dopants 750 o. C activation anneal* 400 o. C FGA ® 28% improvement in Idsat *Ge offers low thermal processing budget for S/D formation - useful for advanced gate-stack materials T. -J. King - 35 - March 29, 2004

Circuit Performance Comparison • No channel doping is needed in thin-body FETs higher Idsat

Circuit Performance Comparison • No channel doping is needed in thin-body FETs higher Idsat achieved, dopant fluctuation effects avoided Simulated circuit performance comparison Bulk SG-UTB DG VT must be adjusted by FM engineering in the range 4. 5 e. V to 5. 0 e. V L. Chang et al. , IEDM 2000 T. -J. King - 36 - March 29, 2004

Molybdenum-Gated Fin. FETs Y. -K. Choi et al. , 2002 IEDM N+ Poly. Si

Molybdenum-Gated Fin. FETs Y. -K. Choi et al. , 2002 IEDM N+ Poly. Si Mo S Lg=80 nm, TSi=10 nm Vds=0. 05 V N+ VT shift D W Tbody Lgate • Tilted N implantation (60 ) used for sidewall gates ü N implantation lowers gate work function T. -J. King - 37 - March 29, 2004

Impact of S/D Contact Structure H. Kam and T. -J. King, to be presented

Impact of S/D Contact Structure H. Kam and T. -J. King, to be presented at the 2004 Silicon Nanoelectronics Worksh Top Contact 3 D simulated Id-Vds curves Wrapped Contact Lg=18 nm, Wfin=10 nm, Tox=5 A End Contact resistance will limit Idsat maximize Acontact , minimize rc rc=10 -9 W-cm 2 ü “wrapped contact” structure ü (germano)silicided thick S/D T. -J. King - 38 - March 29, 2004

Minimizing Parasitic Resistance • Use thin-body structure to control leakage • Use metallic source/drain

Minimizing Parasitic Resistance • Use thin-body structure to control leakage • Use metallic source/drain to minimize Rseries e. g. fully silicided source/drain regions – Ideally, Schottky barrier height Fb 0 e. V Gate TSi SOI Si. O 2 TBOX Silicon Substrate T. -J. King - 39 - March 29, 2004

Thin-Body Metallic-S/D MOSFETs J. Kedzierski et al. , IEDM Technical Digest, pp. 57 -60,

Thin-Body Metallic-S/D MOSFETs J. Kedzierski et al. , IEDM Technical Digest, pp. 57 -60, 2000 • TSi 10 nm • Pt. Si (Fbop=0. 24 V) for PMOS; Er. Si 1. 7 (Fbon=0. 28 V) for NMOS ü Excellent IOFF ü High NMOS contact resistance due to oxidation of Er. Si 1. 7 T. -J. King - 40 - March 29, 2004

Outline • Introduction • Advanced Gate-Stack Materials • Thin-Body MOSFETs • Summary T. -J.

Outline • Introduction • Advanced Gate-Stack Materials • Thin-Body MOSFETs • Summary T. -J. King - 41 - March 29, 2004

Summary • Advanced materials are needed for bulk-Si MOSFETs to meet ITRS specifications –

Summary • Advanced materials are needed for bulk-Si MOSFETs to meet ITRS specifications – High-k gate dielectric – Poly-Si. Ge and eventually metallic gate electrodes – Raised S/D structure (Si. Ge for low Rsd) • Thin-body MOSFETs provide pathway to 22 nm (9 nm Lgate) technology node – High-k gate dielectric will not be necessary – Metal gate and low-Rsd materials will be needed T. -J. King - 42 - March 29, 2004

MOSFET Scaling Scenario high-k dielectric metal S/D metal gate Lg (nm): 65 50 40

MOSFET Scaling Scenario high-k dielectric metal S/D metal gate Lg (nm): 65 50 40 20 10 low channel doping G S 30 D classical Si G S ultra-thin body SOI D Si. O 2 Si substrate G double-gate S SOI G T. -J. King - 43 - March 29, 2004 D