IBM T J Watson Research Center Device Technology
IBM T. J. Watson Research Center Device Technology and the Future of Computing Thomas N. Theis, Director, Physical Sciences, IBM Research Frontiers of Extreme Computing, October 25 2005 © 2005 IBM Corporation
IBM Research My view… § Moore’s Law is an observation about the number of devices that can be placed on a chip at a given time. § There is no fundamental physical reason why device densities cannot follow the Moore’s Law trend for many more decades. – Lithographic dimensions will eventually be shrunk to atomic scale and lithographic processes can seamlessly combine with synthesis. – New devices can circumvent the scaling limits of FETs. Frontiers of Extreme Computing, October 25, 2005 2 © 2005 IBM Corporation
IBM Research Topics § Making devices smaller is not the problem! § Silicon CMOS Technology will be extended for at least a decade § The “ultimate” FET may not be made of silicon § Non-silicon memory devices will scale to very small dimensions. § New devices may enable adiabatic computing and reversible logic. Frontiers of Extreme Computing, October 25, 2005 3 © 2005 IBM Corporation
IBM Research A Brief History of Miniaturization Moore’s Law (1965) 90 nm Manufacturing (2004) 193 nm immersion EUV 2004 commercial niche lithographies industry roadmap 2004 best lab practice Frontiers of Extreme Computing, October 25, 2005 4 © 2005 IBM Corporation
IBM Research The silicon transistor in manufacturing … 35 nm Gate Length Frontiers of Extreme Computing, October 25, 2005 90 nm technology generation 5 © 2005 IBM Corporation
IBM Research … and in the lab. B. Doris et al. , IEDM , 2002 Frontiers of Extreme Computing, October 25, 2005 6 © 2005 IBM Corporation
IBM Research Power Density (log scale) Still, we are approaching some limits. 1. 0 Frontiers of Extreme Computing, October 25, 2005 0. 1 Gate Length (microns) 0. 01 7 © 2005 IBM Corporation
IBM Research Active Power Can Better Managed! Thermal Mapping of Fully Operating IBM Microprocessor (2 GHz, 1. 4 V) in a System Environment Filmstrip during bootup: Thermal images during bootup: max min H. Hamann (IBM), ISSCC 2005 Frontiers of Extreme Computing, October 25, 2005 IBM Confidential 0. 4 s per frame Total ~ 80 s 8 © 2005 IBM Corporation
IBM Research Better Information for Layout and High-level Design Access to data cache Access to memory controller Power map Temperatures max data cache memory controller min H. Hamann (IBM), ISSCC 2005 Frontiers of Extreme Computing, October 25, 2005 9 © 2005 IBM Corporation
IBM Research The Problem with Passive Power Dissipation: You Can’t Scale Atoms! Gate Source Drain 1. 2 nm oxynitride Field effect transistor § Direct tunneling through the gate insulator will be the dominant cause of static power dissipation. § Single atom defects can cause local leakage currents 10 – 100 x higher than the average current, impacting reliability and generating unwanted variation between devices. Frontiers of Extreme Computing, October 25, 2005 10 © 2005 IBM Corporation
IBM Research The Work-Around: High-k Insulator / Metal Gate Stack High-k/Metal Gate Stack 90 nm Si. ON Metal gate electrode High-k material Oxide interlayer 90 nm Gate Dielectric: Tinv = 19 A Tox. GL = 11 A Frontiers of Extreme Computing, October 25, 2005 High-k/Metal Gate Stack: Tinv = 14. 5 A Tox. GL = 16 A 11 © 2005 IBM Corporation
IBM Research Improving Performance § No longer possible by scaling alone – New Device Structures – New Device Design point – New Materials Since the 90’s hydrogen 1 H Device Elements: helium Before the 90’s 2 He boron 5 B Beyond 2005 carbon 6 C silicon 14 Si calcium 20 Ca scandium titanium vanadium chromium 22 22 23 21 24 Ti Ti V Cr Sc strontium 38 Sr yttrium 39 Y barium 56 Ba barium 71 Lu lanthanium 57 La zirconium niobium molybdenum 42 40 41 Mo Zr Nb tantalum Tungsten tungsten Rhenium hafnium Tantalum 73 74 75 72 Ta W Re Hf cerium 58 Ce Praeseo dymium 59 Pr neodymium 60 Nd Frontiers of Extreme Computing, October 25, 2005 cobalt 27 Co nickel 28 Ni zinc 30 Zn nitrogen oxygen fluorine 7 8 O F N 9 phosphorus 15 P germanium arsenic 32 Ge 33 Bromine 35 35 Br Br As Ruthenium rhodium palladium 43 45 45 Ru Rh Pd iridium 77 Ir bismuth 83 Bi platinum 78 Pt samarium europlum gadollinium terbium dysprosium holmium 64 63 65 66 62 67 Gd Eu Tb Dy Sm Ho erbium 68 Er thullium ytterbium 69 70 Tm Yb 12 © 2005 IBM Corporation
IBM Research Innovation Will Continue: Transistor Roadmap Options 2004 2007 2010 2013 2016 2020 Physical Gate Ultrathin SOI 37 nm 25 nm 18 nm 13 nm 9 nm High k gate dielectric 6 nm Fin. FET Double-Gate CMOS In general, growing power dissipation and increasing process variability will be addressed by introduction of new materials and device structures, and by design innovations in circuits and system architecture. Frontiers of Extreme Computing, October 25, 2005 13 © 2005 IBM Corporation
IBM Research Memory may be easier to shrink than logic. SL SL SL m+1 m m-1 WL n+1 § Everyone is looking for a dense (cheap) cross-point memory. § It is relatively easy to identify materials that show bistable hysteretic behavior (easily distinguishable, stable on/off states). Frontiers of Extreme Computing, October 25, 2005 14 © 2005 IBM Corporation
IBM Research Relative Maturity of Nonvolatile Memory Technologies Frontiers of Extreme Computing, October 25, 2005 15 © 2005 IBM Corporation
IBM Research 16 Mbit Magnetic Random Access Memory (MRAM) Demo Cross-section: Chip image: MTJ Materials Advances 1995 TMR (%) 2005 Field (Oe) Frontiers of Extreme Computing, October 25, 2005 Field (Oe) 16 © 2005 IBM Corporation
IBM Research Post-Silicon CMOS: The Quest for the Ultimate FET W 200 n m Self-Aligned Carbon Nanotube FET: Extension Contacts Based on Charge-Transfer Chemical Doping Frontiers of Extreme Computing, October 25, 2005 Vertical Transistor Based on Semiconductor Nanowires 17 © 2005 IBM Corporation
IBM Research Intrinsic Performance of Cabon Nanotube FETs Simple back-gated CNTFET Output Characteristics Pd Pd nanotube Subthreshold Characteristics Temperature dependence Yu-Ming Lin et al. (IBM), EDL 2005 Frontiers of Extreme Computing, October 25, 2005 18 © 2005 IBM Corporation
IBM Research Intrinsic Switching Speed of CNFETs Cut-off Frequency Cg : gate capacitance Lin et al. (IBM) Javey et al. (Stanford) Seidel et al. (Infineon) ~ 1. 8 nm ~ 1. 7 nm ~ 1. 1 nm 10 -nm Si. O 2 8 -nm Hf. O 2 12 -nm Si. O 2 Maximum gm 12. 5 S 27 S 3. 5 S Cg/L 38 p. F/m 120 p. F/m 32 p. F/m f. T @ Lg = 65 nm 800 GHz 550 GHz 260 GHz Diameter Gate Dielectric Yu-Ming Lin et al. (IBM), EDL 2005 Frontiers of Extreme Computing, October 25, 2005 19 © 2005 IBM Corporation
IBM Research Steep Sub-threshold Slopes for Low-Voltage Operation Bulk switching S=63 m. V/dec Dual-Gate CNTFET Sub-thermal S through band-to-band tunneling Frontiers of Extreme Computing, October 25, 2005 20 © 2005 IBM Corporation
IBM Research Adiabatic Computing and Reversible Logic Can we operate FETs at or below the “k. T” limit? Frontiers of Extreme Computing, October 25, 2005 21 © 2005 IBM Corporation
IBM Research Adiabatic Computing How much energy must be dissipated to charge a capacitor? Trade-off depends strongly on device physics! Frontiers of Extreme Computing, October 25, 2005 22 © 2005 IBM Corporation
IBM Research Frontiers of Extreme Computing, October 25, 2005 23 © 2005 IBM Corporation
IBM Research Applications of Adiabatic Charging § Drive specific capacitances which cause large dissipation. – Power supplies – Energy conserving data bus drivers § Broadly implement reversible logic. – Retractile cascade, reversible pipelines – High-efficiency regenerative power supply (very complex) Frontiers of Extreme Computing, October 25, 2005 24 © 2005 IBM Corporation
IBM Research Reversible Logic: Implementation with FETs § It is conceptually possible to build general purpose reversible computers with energy dissipation per operation going asymptotically to zero as frequency goes to zero. § But, frequency must be reduced by about 1/1000 to achieve benefits with respect to conventional approaches to CMOS logic. Dissipation of 4 bit ripple counter (D. J. Frank, 1995) Frontiers of Extreme Computing, October 25, 2005 25 © 2005 IBM Corporation
IBM Research Recent Optimized Power vs Frequency Assessment DJ Frank, MIT Workshop on Reversible Computation, February 14, 2005 Frontiers of Extreme Computing, October 25, 2005 26 © 2005 IBM Corporation
IBM Research Qualitative Overall Comparison Conventional CMOS with fixed supply TSPC (a fast non-energy recovering dynamic CMOS logic family) Scaled static CMOS (estimate) 2 N-2 N 2 P (an energy-recovering logic family) Highly reversible DJ Frank, MIT Workshop on Reversible Computation, February 14, 2005 Frontiers of Extreme Computing, October 25, 2005 27 © 2005 IBM Corporation
IBM Research Are there devices that are better suited than FETs for the implementation of reversible logic? Frontiers of Extreme Computing, October 25, 2005 28 © 2005 IBM Corporation
IBM Research Beyond Charged-Based Logic? § Spins § Nanomechanics § Photons § DNA Chemistry Frontiers of Extreme Computing, October 25, 2005 29 © 2005 IBM Corporation
IBM Research Nanoelectronics Research Initiative (NRI) § AMD, Freescale, Micron, TI, IBM, Intel Joint Industry funding of University Research § Leveraging existing NSF and new (joint NSF/NRI) funding § Promoting both – Invention / Discovery (distributed research, “let many flowers bloom”) – Proof of Concept (focused university consortia with outstanding facilities) § “Extend the historical cost/function reduction, along with increased performance and density … orders of magnitude beyond the limits of CMOS” – Non-charge-based logic – Non-equilibrium systems – Novel energy transfer mechanisms for device interconnection – Phonon engineering – Directed self-assembly of complex structures Frontiers of Extreme Computing, October 25, 2005 30 © 2005 IBM Corporation
IBM Research Conclusions § Silicon CMOS logic will be extended at least another 10 years. – New materials and transistor structures – Cooperative circuit and device technology co-design § The “ultimate FET” may not contain silicon. § New, dense, cheap memory devices are on the way. § Reversible logic is possible, but may be practical only with devices that are “beyond the FET”. § The search for logic devices “beyond the FET” is underway, (but there should be more focus on devices suitable for adiabatic computation and reversible logic) Frontiers of Extreme Computing, October 25, 2005 31 © 2005 IBM Corporation
IBM Research Thanks! Frontiers of Extreme Computing, October 25, 2005 32 © 2005 IBM Corporation
IBM Research Non-Si FET § The potential of carbon nanotubes (CNT) – Metallic & semiconducting nanotubes: interconnects (up to ~109 A/cm 2) & switches. – 1 D transport (low elastic and inelastic scattering low energies, ballistic or quasi-ballistic transport, f. T≤ 1 THz. – Low energy dissipation (power density problem, no electromigration) – Good control of electrostatics (“thin body devices”, no mobility degradation). – Inert (no dangling bonds to passivate, wide choice of gate insulators – high k materials) – Direct band-gap materials (integrate electronic & opto-electronic devices using the same material). Frontiers of Extreme Computing, October 25, 2005 33 © 2005 IBM Corporation
- Slides: 33