Flash Memory with Nanoparticle Floating Gates

Transcription

1 Flash Memory with Nanoparticle Floating Gates Sanjay Banerjee Director, Microelectronics Research Center Cockrell Chair Professor of Electrical & Computer Engineering University of Texas at Austin Why Nanoparticles in Flash Memory Self-assembly schemes Device Performance Acknowledgements: DARPA MARCO, NSF NIRT, Micron

2 The Scale of Things Nanometers and More Things Natural 10-2 m 1 cm 10 mm Things Manmade Head of a pin 1-2 mm The Challenge Ant ~ 5 mm 10-3 m 1,000,000 nanometers = 1 millimeter (mm) Dust mite 200 μm Human hair ~ μm wide Red blood cells with white cell ~ 2-5 μm ~10 nm diameter Fly ash ~ μm ATP synthase Nanoworld Microworld 10-4 m 10-5 m 10-6 m 10-7 m 10-8 m Ultraviolet Visible Infrared Microwave 0.1 mm 100 μm 0.01 mm 10 μm 1,000 nanometers = 1 micrometer (μm) 0.1 μm 100 nm Smaller is different 0.01 μm 10 nm Pollen grain Red blood cells Zone plate x-ray lens Outer ring spacing ~35 nm Self-assembled, Nature-inspired structure Many 10s of nm MicroElectroMechanical (MEMS) devices μm wide Nanotube electrode O O S O O O O O O O O O O O S O S O S O S P O O Fabricate and combine nanoscale building blocks to make useful devices, e.g., a photosynthetic reaction center with integral semiconductor storage. O O S O O S O S DNA ~2-1/2 nm diameter Atoms of silicon spacing ~tenths of nm 10-9 m m Soft x-ray 1 nanometer (nm) More is different 0.1 nm Quantum corral of 48 iron atoms on copper surface positioned one at a time with an STM tip Corral diameter 14 nm Carbon buckyball ~1 nm diameter Carbon nanotube ~1.3 nm diameter Office of Basic Energy Sciences Office of Science, U.S. DOE Version , pmd

3 Moore s Law No exponential is forever! But can we delay forever? source drain

4 Flash memory applications Markets Consumer Electronics Networking Wireless Industrial Control Automotive Floating gate Control Gate Tunnel dielectric

5 NAND Flash Market Segmentation Digital Settop Box 0% Others 5% *Others include GPS, games, emerging applications, etc MP3 players (Flash based) 22% $468 M 2005 NAND Demand TAM Percentage isupply: Q $2124 M USB Flash Drive 17% $1667 M $5230 M Removable Solid State Flash Cards 53% $244 M $70 M Mobile Handset (Embedded) 2% Digital Still Camera (Embedded) 1% NOR Flash had TAM of 7B$, mostly in wireless cell phones. Very high growth rates!!!

6 Flash Technology Scaling History 1986 / 1.5μm 1988 / 1.0μm 1991 / 0.8μm 1993 / 0.6μm 1996 / 0.4μm 1/~ / 0.25μm 2000 / 0.18μm 2002 / 0.13μm 2004 / 90nm 2006 / 65nm 10X 10X 10X Volume Production Year / Technology Generation Source: Intel Flash Invented in mid 1980 s NOR flash evolved from EPROM NAND started as poly-poly erase cell later evolving to present structure ~20 years & 10 Generations of ETOX (Intel NOR) High Volume Production 8+ years & 5 Generations of MLC: 2bit / cell

7 NVM- Long Term Requirements (ITRS 2006)

8 Nanoparticle Gate Flash Memory Gate Gate Source Drain Source Drain Conventional Flash Memory A defect totally discharges the floating gate Nano-floating Gate Memory A defect discharges only one dot Thick tunnel oxide High voltage/ power Low reliability/ speed High-k tunnel oxide Speed/ power/ density better Reliability improved New phenomena- self-assembly, Coulomb blockade, multi-level cells

9 SiGe Nanocrystals on High-K Dielectrics HfO 2 SiO 2 AFM scans (1 micron x 1micron) showing SiGe dots grown at ~ 500ºC for 90 s with 0.75 gas ratio of GeH 4 to Si 2 H 6. Kim., Banerjee, Growth of germanium quantum dots on different dielectric substrates by chemical-vapor deposition, J VAC SCI TECHNOL B 19 (4): 2001

10 Band diagram of HfO 2 and SiO 2 dielectric at low program voltage Nanocrystal Dots Si E c Channel SiGe E c Channel Dot gate Dot Interface oxide gate Tunneling oxide (HfO 2 ) (HfO 2 ) SiO 2 F-N tunneling (due to small φ b ) Tunneling oxide (SiO 2 ) (SiO 2 ) J g [A/cm 2 ] HfO 2 E.O.T 4nm Direct tunneling Direct tunneling V G [V]

11 Program & Erase Transient Characteristics Threshold Voltage shift (V) HfO 2 tunneling oixde =4.8nm SiO 2 tunneling oxide = 3.6nm Erase Program HfO 2 2.5V SiO 2 4V SiO 2 4V HfO 2 2.5V ΔV th (V) HfO 2 tunneling oxide SiO 2 tunneling oxide Without dots With SiGe dots Without dots Program/Erase Time (s) V CG (V) Kim DW, Kim T, Banerjee SK, ELECTRON DEV 50 (9): SEP 2003

12 The Bio-nanometer Perspective Atom Water DNA Virus Protein Bacterium Cells Hair Fish egg Human eye nm μm Nanomaterials: Nanotubes Dendrimers Nanopores Quantum dots Nanoshells Nanoparticles mm Gate/ Transistor Interconnect Integrated circuit Microprocessor dielectrics/substrate Slide by C. Li

13 Yamashita, IEDM 2006, p. 447 (Ferritin) Ishii, Nature Vol. 423, 2003 (Chaperonin)

14 Protein assembly of Metal (Co, Au,..) and Semiconductor (PbSe, CdSe, Ge..) NCs Schematic structure of Chaperonin 60 (GroEL)

15 STEM images of PbSe nanocrystals on SiO 2 50 nm 50 nm Without chaperonin template With chaperonin template Various semiconductor and metal nanocrystals self-assembled, including Co Density~ cm -2

16 ΔV FB comparison of protein-mediated PbSe NC MOS capacitor and control samples TaN Gate Control Oxide Tunnel Oxide Si Substrate NCs SiO 2 10 nm PbSe NCs TaN Si A PbSe NCs Protein Template (Annealed) B C D ΔV FB (V) 0.8 Program/Erase time=1s A B C,D Program Voltage (V)

17 Endurance and Retention Characteristics V FB (V) PbSe -10V 200ms Co -8V 200ms PbSe 10V 200ms Co 8V 200ms Number of Cycles V FB (V) PbSe -10V 200ms Co -8V 200ms PbSe 10V 200ms Co 8V 200ms Retention Time (s) Devices survived 10 5 cycles of program/erase operation; no sign of window closure Devices have good retention.

18 Heat Shock Proteins (Trent, NASA)

19 Coulomb Blockade in SiGe dot on SiO 2 and HfO 2 2.0x10-5 SiGe on SiO 520 ºC Drain Current (A) 1.5x x x10-6 n=0 n=1 0.0 SiO 2 tunneling oxide = 3.6nm Gate Voltage (V)

20 Vertical Flash Memory Sidewall channels Nanocrystal floating gate Gate Source Drain contact Gate poly Source contact (outside view) Device mesa n + Source/Drain Schematic side view Scanning Electron Micrograph

21 Vertical Flash Retention at 300K ± 9 V, 100 ms tunneling P/E 10 nm E1237 PCL_ _VFR5W4_04_3DFlash_Nanocrystals.dm3 MAG: 295kX

22 Nanoparticle Self-Assembly Using PS-b-PMMA PMMA Copolymer Process Flow Employing a sandwich of organic/ inorganic/ organic layers (Polyimide/ SiO 2 / PS-b-PMMA) to engineer the aspect ratio of the patterns Material Characterization (a) (b) 100 nm 100 nm (c) (d) glue 6 Normalized Count 100 nm 10nm Si Substrate Nickel Nanocrystals SEM micrographs of (a) Copolymer template, (b) transferred polymer patterns into the underlying SiO2 layer, (c) ultimate array of Ni nanoparticles. (d) Cross-sectional image of the embedded nanoparticles within SiO 2. Histogram of the copolymer pore size distribution shown in Fig. (a)

23 Characteristics of MOSCAP Memory Devices ±10V, 100 msec ±8V, 100 msec Control Sample - w/ o Ni dots 10V 8V 6V Gate Voltage (V) ΔV FB (V) Write/ Erase time (s) High frequency C-V 1MHz Memory window for different program voltages Transient characteristics of the memory device

24 Molecular Memories, Misra, IEDM p ; IEDM p Variability, DOS? SONOS trap densities ~10 19 cm -3 ev -1 For ~3 nm layer, ~3X10 12 cm -2 traps spatially and energetically distributed

25 Non-MOS Memory Contenders (adapted from Fazio, Intel) MRAM FeRAM Phase Change Cell Size λ 2 Large ~40 20 Large ~25 Small ~6.5 CMOS <200C Post Magnetic Fe reduces in hydrogen Compatible with backend Integration Tight MJT control Etching difficult for Fe CMOS metal processing Read Write Cycling Endurance Scalability Application Color Code Poor OK Good Non-Destructive, Fast, Low Power Power constrained, Scales poorly; half select issue Theoretically Infinite Write current increases with scaling, materials engineering required at each scaling node; superparamagnetic limit? Embedded Working Memory Low Density Destructive: Endurance limited read Low power capacitive Theoretically good speed 1e-8 1e12; claims made, but limited data 3D cell required sub 90nm. Material engineering required at each scaling node Embedded Low Power Low Density Non-Destructive, Moderate speed, Scales poorly Power constrained, large drivers, Improves with scaling ~1e12; claims made, but limited data; Erratic failures No material changes, No physical limits known down to ~5nm Stand Alone or Embedded High Density Low Cost Data Storage Region Chalcogenide Phase Change Material Amorphous Poly Crystalline Heater Resistive Electrode

26 1R

27 Nanocrystal Dots Channel Dot gate Tunneling oxide End-of-roadmap (2020, L=6 nm) NAND F= 14 nm, Tox= 6 nm, Cell area ~ 1000 nm 2; (3X for NOR) NC spacing ~ tunnel oxide thickness ~ 4 nm Spacing & NC size/dos affects charge capture crosssection & Read Optimal spacing depends on dielectric & NC band diagram, retention (10 8 s), Erase/Write times (10-6 s) E barr = 1eV For NC densities of ~ cm -2, room for ~10 NC Variability is a challenge for NC & trap-based cells Self-assembly and non-planar structures?

28 There is plenty of room at the bottom! Feynman Will these devices make it? The answer is a very definite.. Maybe!!