EUROPEAN “NANOMETRIC” CMOS PROJECTS
Transcription
EUROPEAN “NANOMETRIC” CMOS PROJECTS
EUROPEAN “NANOMETRIC” CMOS PROJECTS Σ! 2365 Guillermo Bomchil STMicroelectronics 21/11/05. G.Bomchil. MEDEA+ Forum 2005. Barcelona 2 CONTENTS European projects: from 90 nm to 32-22 nm MEDEA+ T201 90 nm: industrial exploitation MEDEA+ T207 65 nm: present status EC NANOCMOS 45 nm: present status Future projects MEDEA+ 2T 103: 45 nm Full CMOS Integration EC PULLNANO: Pushing the limits of CMOS Added value of the collaboration and conclusions G.Bomchil. MEDEA+ Forum 2005 Barcelona 3 NANOELECTRONICS PROJECTS 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 100 nm FEOL HUNT ITRS 90 MEDEA+ T201 90 nm full CMOS integration ARTEMIS 65 nm BEOL 65 nm FEOL ITRS 65 ULISSE 65 nm full CMOS integration MEDEA+ T207 dev. architecture 45 – 32 – 22 nm NESTOR MEDEA+ HYMNE 65/45 nm manufacturing sciences 45 nm FEOL/BEOL + SRAM ITRS 45 NANOCMOS MEDEA+ 2T103 Strained Silicon on SOI 45 nm full CMOS integr. MEDEA+ SILONIS 32 nm FEOL/BEOL + SRAM + 22 nm PULLNANO 32 nm full CMOS integration G.Bomchil. MEDEA+ Forum 2005 Barcelona MEDEA+ T 32nm 4 T201 90 nm Strategic Objectives Development of worldwide competitive 90 nm CMOS logic technology Demonstrator: 54 M tr, 300 M contacts, 9 Km of interconnects (Cu-low k ) Start- end: 01/2001 – 12/2002 Consortium: ST MicroelectronicsPHILIPS Semiconductors, BULL, LEICA, IMEC, LETI, PHILIPS Research, AIR LIQUIDE, AIXTRON, EPICHEM, JOBIN YVON, INPG/CNRS, LTM/CNRS Objectives were achieved as schedule The project obtained the JPN award 2004 Bulk silicon, 1.6 nm nitrided gate oxide, cobalt salicide, G.Bomchil. MEDEA+ Forum 2005 Barcelona 5 CMOS 90 nm industrial exploitation The 90 nm products were introduced in Crolles 2, the joint FREESCALE, PHILIPS Semiconductors and ST Microelectronics 300 mm pilot manufacturing facility. 90 nm products include digital, analog, RF and embedded memory devices for diverse applications, such as wireless handsets, printers, networking components,… In many cases the products had 100 % first-pass success rate demonstrating design efficiency and optimum use of technology and manufacturing capabilities. Altogether the three partners, starting from October 2003, report more than 25 processed circuits for their customers. The partners consider that they have addressed many of the industry’s widely reported challenges with 90 nm and are applying the expertise to the next generations of technology. “The achievements at 90 nm over the past year set the stage for successful 65 nm prototyping from end 2005”! G.Bomchil. MEDEA+ Forum 2005 Barcelona 6 Some examples of 90nm circuits Code Name Option Market Status / comments A Olympus2b GP. 9ML uProcessor PG Oct’03. 120mm2 B Crystal LP. 8ML Hardware emulator PG April’04. 240mm2 PG cut2 Feb’05 C PIAGET GP . 7ML TV PG May’04 ~70mm² D 7100 GP . 7ML Set Top Box PG Oct’04 ~80mm². PG cut2 E ADV1 GP . 7ML Set Top Box PG Sept’04 ~30mm² H 8010 GP. 7ML DVD recorder PG Dec’04 ~45mm² G GSP3 LP. 6ML GPS Mobile PG Dec’04. 9mm² I 288 GP. 7ML Demodulator PG Apr’05. 3mm² J DARS3G GP. 7ML Decoder PG Apr’05. 20mm² K Bigoot LP. 7ML. eDRAM Decoder PG May’05. 45mm² L 5525 GP. 7ML Set Top Box PG June’05. 43mm² M 7109 GP. 7ML Set Top Box PG July’05. 72mm² N COST3 LP. 7ML WLAN PG Aug’05. 9mm² G.Bomchil. MEDEA+ Forum 2005 Barcelona 7 T207 CMOS 65 nm design rules Design Rule pitch (L/S) in nm OD PO CO M1 Vx Mx SRAM cell size (um2) Gate density (kgt/mm2) 9 tracks 12 tracks 14 tracks CMOS12 360 310 340 340 410 410 (210/150) (180/130) (180/160) (180/160) (220/190) (210/200) 2.5 2.07 200 CMOS090 250 240 260 240 280 280 (110/140) (100/140) (120/140) (120/120) (130/150) (140/140) 1.27 1.15 0.999 CMOS065 190 (80/110) 180 (60/120) 200 (90/110) 180 (90/90) 200 (100/100) 200 (100/100) shrink % 76 75 77 75 71 71 0.525 (LP) 0.499 (GP) 53 50 >800 [>720] >650 ~2.0x 430 [380] 350 G.Bomchil. MEDEA+ Forum 2005 Barcelona 8 ~2.0x Technology migration 90nm(T201) Æ 65nm(T207) (T207>>directly in 300 mm wafers) substrate lithography STI Gate dielectric Gate electrode Junction RTA Silicide Strain Contact Low-k Barrier Interconnect CMOS090 bulk 193nm scale 120nm STI furnace-nitrided oxide 1500A poly-Si predope offset spacer + implant spike anneal CoSi2 W CVD BDI (keff~3.1) PVD Cu (6-9 ML) CMOS065 bulk, rotated <100> 193nm (high-NA) scale 90nm STI plasma-nitrided oxide 1000A poly-Si predope a-C hard mask no offset spacer (co-)implants aggressive spike low thermal budget NiSi CESL-induced W CVD keff 2.7 PVD Cu (6-9 ML) G.Bomchil. MEDEA+ Forum 2005 Barcelona 9 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 1.E-10 1.E-11 -1.2 Drain current [µA/µm] Drain current [A/µm] Selected examples: GP Transistor characteristics 1200 1000 1.2V 800 600 1V W=1µm 0.8V 400 0.6V 200 -0.7 -0.2 0.3 Vgs [Volt] 0.8 0 -1.2 -0.9 -0.6 -0.3 0 Sub-Threshold regime G.Bomchil. MEDEA+ Forum 2005 Barcelona 0.3 0.6 0.9 1.2 Vds [Volt] Output characteristics 10 GP NMOS digital performance -4 Vdd=1V Log Ioff (A/µm) -5 Vdd=1,2V -6 -7 -8 1300 -9 810 500 700 990 900 1100 Ion (µA/µm) 1300 G.Bomchil. MEDEA+ Forum 2005 Barcelona 1500 11 Core LP transistor short channel control 0.6 NMOS Vt sat [Volt] 0.4 0.2 Vds=1.2V 0 -0.2 -0.4 PMOS -0.6 30 50 70 90 L silicon [µm] G.Bomchil. MEDEA+ Forum 2005 Barcelona 12 LP-6transistor digital performance NMOS PMOS -8 -9 -10 -11 300 710 -12 Ion [µA/µm] G.Bomchil. MEDEA+ Forum 2005 Barcelona 13 10 00 0 90 80 0 70 0 60 0 50 0 40 0 30 0 20 0 10 0 -13 0 Ioff [A/µm] -7 100 90 80 70 60 50 40 30 20 10 0 Cumulative plot [%] Cumulative plot [%] Selected examples: MLM electrical characteristics 12M Vias yield 0 1 2 3 4 5 6 7 8 100 90 80 70 60 50 40 30 20 10 0 M1 -11 -10 -9 -8 -7 Leakage [Log A] Rsq [Ohm/via] M2 105nm V1 M1 6 metal levels ULK G.Bomchil. MEDEA+ Forum 2005 Barcelona 14 -6 -5 0.5µm² SRAM bit-cell SRAM after NiSi process step Internal node Shared contact Word line (Copper) NiSi 45nm 45nm NiSi NiSi Pull Up transistor G.Bomchil. MEDEA+ Forum 2005 Barcelona 15 0.5µm² SRAM bit-cell Static Noise Margin 1.4 SNM=240mV 300 SNM (mV) 1.2 VR, Vin (V) 1.0 0.8 250 200 150 100 0.7 0.6 0.9 1.1 1.3 Vdd (V) 0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 Vin, VL (V) G.Bomchil. MEDEA+ Forum 2005 Barcelona 16 1.4 T 207 CMOS 65 nm statuts: October 2005 Process freeze achieved Significant yield improvements in FEOL and BEOL Reliability meeting the specifications The process is open for prototyping customers chips Risky production is expected from Q1 06 The Crolles 2 partners will be sharing their 65nm cell libraries and IP blocks. They are confident about the success of its 65nm process after getting all its 90nm designs in the last twelve months right first time G.Bomchil. MEDEA+ Forum 2005 Barcelona 17 EC NANOCMOS 45 nm Achieve a first “demonstration of feasibility” of a 45 nm CMOS logic process based on a selection of the most appropriate technology among several options Bulk SOI SON DG Strain DevicePerformance Performance •• Device Devicesuitability suitabilityto tolow lowcost costprocess process ••Device Devicesuitability suitabilityto tolow lowpower power ••Device Feasibilityin in300mm 300mmfab fab ••Feasibility Timing ••Timing Competitor/literaturetrend trend ••Competitor/literature SRAM SRAM Feasibility Feasibility Demonstration Demonstration Transfertto toaa Transfert MEDEA+Project Project MEDEA+ HK Metal Gate R§D activities for 32 nm nodes in cooperation with the network of academic teams SINANO G.Bomchil. MEDEA+ Forum 2005 Barcelona 18 45nm BULK General Purpose applications (L=30nm, Tox=1.1nm, Vdd=1V) Low Power applications (L=45nm, Tox=1.7nm, Vdd=1.2V) Conventional Gate stack Poly-Silicon Gate on SiON Strained Si as Technological Booster: – Low-cost Strained Liners (CESL,SMT), Rotated Channels -6 Tensile CESL NiSi 30nm 50nm 30nm Log(Ioff) (A/µm) GP nMOS device Vdd=1V std CESL -7 tensile CESL -8 +20% SMT Layer (100) Channel -9 0.8 0.9 1 1.1 1.2 1.3 normalized Ion G.Bomchil. MEDEA+ Forum 2005 Barcelona 19 1.4 1.5 NANOCMOS first functional bulk based SRAM simulation 1000 190mV PG 800 Vout (mV) 0.248µm² 30nm 600 400 200 Vdd=1. 0 V 0. 248µm² 800 0 0 PD 500 1000 Vin (m V) PU Shared V dd 100nm contact Vout (mV) 600 400 200 Vdd=1.0 V 0 30nm 0 pM O S 50nm Pull-up pM O S P ull-up 500 Vin (mV) 1000 Functional 0.25µm² SRAM-bit-cell in BULK technology! June 2005 fabricated at Crolles in 300 mm G.Bomchil. MEDEA+ Forum 2005 Barcelona 20 Alternative trigate option Specific design, all Optical S=0.314µm² VDD = 0.4 V-1.2 V (0.2 V incre me nts ) Functional 0.314µm² SRAM Cell ! fabricated at IMEC. June 2005 1 in te r n a l n o d e 2 (V ) 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 1.2 inte rna l node 1 (V) G.Bomchil. MEDEA+ Forum 2005 Barcelona 21 2T 103 MEDEA+ FOREMOST Objectives FOREMOST main goal : demonstration of a full CMOS 45 process technology in European 300 mm manufacturing industrial facilities. Starts from NANOCMOS results. The proposal targets both CMOS logic and DRAM/FLASH process technologies . Promotes synergy between the competences of Crolles 2 Alliance and Infineon in Dresden. Demonstration of feasibility of a complex test vehicle representative of 45 nm design rules( 3x more complex than today most complex 90nm design in PHILIPS). Project start January 2006.First validation is proposed for Q4 2007 in Crolles 2. Reliability data in Q2 /08. G.Bomchil. MEDEA+ Forum 2005 Barcelona 22 PULLNANO 32 nm/22 nm FinFET LOP compatible Devices Ioff ~ 1nA/µm – 10pA/µm Planar : SOI or SON major changes in materials and NiSi architectures. 55nm NiSi channel BOX 32 nm Target:0.14µm² SRAM cell !! Advanced R§D activities for the 22 nm node primarily done three clusters of academic teams and institutes (19 academic partners representing more than 35 laboratories) G.Bomchil. MEDEA+ Forum 2005 Barcelona 23 NiSi CONCLUSIONS Several European Commission and MEDEA + projects contribute to the development of nanometer CMOS generations. They place the European IC industry at the top front of the technology The 90 nm is in industrial production. The 65 nm starts product prototyping. First choices for a 45 nm technology are available MEDEA+ 2T103 will continue towards full process integration R§D of 32 and the 22 nm generation is the objective of PULLNANO. The project gathers an unprecedented number of academic teams The Crolles2 Alliance organization triggers a tight collaboration between a multitude of teams and experts. The achievements of these projects are possible thanks to an excellent cooperation: IC manufacturers, equipment suppliers, institutes, academic teams.. It is expected that cooperation between the European Commission and National Public Authorities, MEDEA+ organization, .… would allow to structure the appropriate support to continue with these very successful chain of projects. G.Bomchil. MEDEA+ Forum 2005 Barcelona 24 Acknowledgements Danielle Thomas T207 and 2T 103 FOREMOST Project Leader Peter Stolk T207 WP4 leader The NANOCMOS Sub-Project Leaders The PULLNANO Sub-Project Leaders G.Bomchil. MEDEA+ Forum 2005 Barcelona 25 MEDEA+ Forum 2005 21-22 November Barcelona 21/11/05. G.Bomchil. MEDEA+ Forum 2005. Barcelona 26