Floating Gate Devices Kyle Craig
Flash Memory Cells – An overview Paolo Pavan, Roberto Bez, Piero Olivo and Enrico Zanoni
Motivation! Predicted Worldwide Memory Market Flash prediction, 6% of total memory market…
According to Gartner Research in 2006 flash consisted of 33% of the market
FGMOSFET If a charge can be forced onto the floating gate, it will remain there. Charge on the floating gate shifts the VT of the device Two VT device
By having two VTs depending on the charge of the floating gate, device can be used as a memory. No charge on floating gate = logic 1 Charge on floating gate = logic 0
Hot Electron Injection Electrons travel laterally from source to drain with applied voltage. Voltage on gate gives enough energy to inject through the thin oxide onto the floating gate Three principles “lucky” enough to gain enough energy No collisions in substrate No collisions in oxide
Fowler Nordheim Tunneling With an applied electric field, electrons are able to tunnel through the oxide. The thicker the oxide, the greater the applied voltage needs to be 10nm oxide is considered standard Variation in this oxide will lead to wide distribution of VT values
Side effects HEI and FN Tunneling can lead to charge being trapped in the oxide Change in the VTs of the device Inability to add or remove charge from floating gate
Flash Memory: Programming Assumed device starts with no charge on the floating gate i.e., storing a “1” Use HEI to put charge onto floating gate Shifts VT of device Depends on: Channel Length, Time, Temp, drain voltage
Flash Memory: Erasing Use Fowler-Nordheim Tunneling to pull electrons off of floating gate to the source Depends on: Oxide thickness, applied voltage Need to worry about breakdown of the source/substrate junction Limits Scaling!
Program, Erase, Read Source Control Gate (WL) Drain (BL) Read GND Vcc Vread Program Vpp Vdd Erase Floating Typical Values: Vcc = 5V Vpp = 12V Vdd = 5V Vread = 1V
Programming Disturbs Gate Disturbs – Cells not selected with active WL DC Erasing If cell has charge store on it, electrons can tunnel from the FG to the Control Gate DC Programming If the cell has no charge on it, electrons can tunnel from the substrate to the FG
Programming Disturbs Drain Disturbs Lowers the high VT value Electrons can tunnel from the FG to the drain Holes generated by impact-ionization in substrate then injected into FG Lowers the high VT value
Retention/Endurance Retention: Change in charge on FG Intrinsic: field-assisted electron emission, thermionic emission Extrinsic: Oxide defects, Ionic contamination Endurance: Change in threshold values based on number of cycles
Scaling Issues with scaling Decreasing L will increase performance but also increase number of disturbs Physical voltage constraints 3.2 eV energy barrier and 8-9MV/cm for FN Oxide thickness limit
NOR vs NAND NOR similar in structure to SRAM NAND more comparable to Harddrives
**From Micron NAND vs NOR Comparison
A 6V Embedded 90nm Silicon Nanocrystal Nonvolatile Memory R. Muralidhar, R.F. Steimle, M. Sadd, R.Rao, C.T. Swift, E.J. Prinz, J. Yater, L. Grieve, K. Harber, B. Hradsky, S. Straub, B. Acred, W. Paulson, W. Chen, L. Parker, S.G.H. Anderson, M. Rossow, T. Merchant, M. Paransky, T. Huynh, D. Hadad, KO-Min Chang, and B.E. White Jr.
Motivation Scaling of conventional floating gate memories is limited due to high voltages that are needed Use of Silicon Nanocrystals has some benefits over floating gate Immune to oxide defects during program/erase Reduction of oxide thickness Reduction of operating voltage
Replaces floating gate of traditional cell with discrete Nanocrystal particles Produced using a conventional CMOS process flow with only 4 additional masks compared to logic Control Gate
Current Characteristics Two Bits/cell operation is possible with proper nanocyrstal isolation Without needed isolation functions like a normal FG
Mask Adder
Endurance and Retention
Can be erased from the top oxide (between FG and control gate) or the bottom oxide (between nanocrystals and substrate) Erasing through the top oxide produces lower VT
Cycling on VT After cycling 1000 cycles Erase and Program VTs maintain tight distribution.
Summary Produced in 90nm .25um technology Operation voltages 6V 90% yield on 4 MB array