Non-Volatile Memory Based on Solid Electrolytes Michael Kozicki Chakku Gopalan Murali Balakrishnan Mira Park Maria Mitkova Center for Solid State Electronics Research
Introduction The electrochemical redistribution of nanoscale quantities of metal in structures containing solid electrolytes is the basis of Programmable Metallization Cell (PMC) technology. Applications in electronics, MEMS, microfluidics, optics We can use this effect to form and dissolve conducting pathways on command so that we have an electronic switch that can be closed and opened rapidly using small voltages. Such switches have potential applications in non-volatile memory and programmable logic devices
Solid electrolytes Solid electrolytes behave like liquid electrolytes e - e - M M Mobile ions M M OR M + M + M + M + M + M + M+ M + M + M + M + M M M + M + M + Liquid Lateral/coplanar Vertical M + e - Mobile ions Ions move under the influence of an electric field Electrochemical reactions are possible with electron supply: M + + e - M reduction with oxidizable metal: M M + + e - oxidation Electrochemistry occurs at a few 100 mv
Room temperature electrolytes Silver or copper can be added to a variety of glasses to form good solid electrolytes base glasses can be a variety of chalcogenides (e.g., compounds of S or Se) or oxides base glasses can incorporate many tens of atomic % of metal ternaries are nanostructured materials with high ion mobility (up to 10-3 cm 2 /Vs) but relatively high resistivity (> 10 2 Ω.cm) metals and base glasses are relatively easy to process at low temperatures
Electrolyte preparation λ < 500 nm, ~1 J/cm 2 Ag or Cu Ge x S(e) 1-x or WO 3 W Si 25 nm 50 nm Ge x S(e) 1-x, x < 0.33 or WO 3 thin films deposited by PVD (other formation options exist) Ag or Cu deposited by PVD Metal driven into base glass using UV light and/or heat
Electrolyte example: Resistivity of Ge-S vs. Ag content 16 14 12 T = 25 ºC Network modification Hi R >10 at.% Lo R Log ( ) (.cm ) 10 8 6 4 2 0 0.001 0.01 0.1 1 10 100-2 -4-6 (Bychkov) Concentration and distribution of Ag Percolation (as well as geometry) determines resistance of layer Reduction due to electrodeposition 0.5Ag 2 S-0.5GeS 2 (Belin) Ag saturated Ge-S (Elliot) Ag 2 S (Miyatani) Ag <10 at.% Ag (at %)
Inner Workings - Nanostructure Oxidizable electrode Inert electrode Glassy insulator <2 nm <10 nm 10 to 100 nm thick Superionic region
Inner Workings Switching Oxidizable electrode Inert electrode Glassy insulator Ions + I prog = na - ma Superionic region Electrodeposited metal V T = 0.25 V - Electrons
Information storage We store information as reduced metal in the solid electrolyte Write in forward bias to inject and reduce ions A few thousand electrodeposited atoms is fc charge range Ion mobility as high as 10-3 cm 2 /V.s and internal field around 10 5 V/cm leads to electrodeposit growth rates of 1 nm/nsec Erase in reverse bias to remove excess metal Decrease concentration by oxidation of electrodeposit Read options involve detection of amount of reduced metal in electrolyte Resistance change is large and easy to detect Other sensing options also exist
PMC test device layout Device array Cathode Via Anode Top electrode Anode layer Solid electrolyte Dielectric Inert electrode Device diameter, D
1.0E-06 Example of PMC device I-V 240 nm diameter, 50 nm Ag-Ge-Se on W cathode Current (A) 8.0E-07 6.0E-07 4.0E-07 2.0E-07 0.0E+00-2.0E-07 I prog = 1 µa R off = 10 10 Ω R on = 10 5 Ω Full erase possible Break in pathway New/erased threshold Threshold with electrodeposit present -4.0E-07-0.5-0.3-0.1 0.1 0.3 0.5 Voltage (V)
Example of PMC device I-V 240 nm diameter, 50 nm Ag-Ge-S on W cathode following 370 deg. C anneal 1.0E-06 Current(A) 8.0E-07 6.0E-07 4.0E-07 2.0E-07 0.0E+00 I prog = 1 µa R off = 10 10 Ω R on = 10 5 Ω -2.0E-07-4.0E-07-0.5-0.3-0.1 0.1 0.3 0.5 Voltage(V)
Final resistance vs. programming current Resistance (Ohms) 1.00E+05 1.00E+04 1.00E+03 1.00E+02 1.00E+01 Binary coding of full range split into 8 bands R on = V w /i p V w is around 100 mv Band width R = ±40% 000 001 010 011 100 101 110 111 Range can be split into bands representing states/binary numbers 1.00E-06 1.00E-05 1.00E-04 1.00E-03 1.00E-02 Low power memory range Erasable antifuse range Current (Amps) 2 decades and R = ±25%would give 9 states (3+ bits) 4 decades and R = ±25%would give 18 states (4+ bits)
Example of R off /R on vs. operating T Resistance (Ohms) 1.0E+10 1.0E+09 1.0E+08 1.0E+07 1.0E+06 1.0E+05 1.0E+04 E a = 0.3 ev Device programmed at T E a < 0.1 ev I write = 1 µa 0 50 100 150 Temperature (deg.c)
Input Cycling example around 10 billion cycles 75 nm diameter, 50 nm Ag-Ge-Se on Ni cathode Start of test 8.9 x 10 9 cycles Test circuit current ( µa) 15 12 9 On 6 3 0 Off -3 1.0E+09 1.0E+10 1.0E+11 Cycles 1.7 x 10 10 cycles On transition Off transition 3.2 x 10 10 cycles
Result of 25 nsec write pulse Current(A) 1.0E-07 8.0E-08 6.0E-08 4.0E-08 2.0E-08 0.0E+00 After 25 nsec pulse, on resistance = 1 MΩ 0 0.02 0.04 0.06 0.08 0.1 Voltage(V) Before 25 nsec pulse, off resistance >13 MΩ Note: write current (and therefore on resistance) limited by test set-up parasitics
Low Programming Current (2 µa) Room Temperature Retention On resistance (ohms) 1.0E+08 R off /R on still >10x after 10 years 1.0E+07 <100 mv sense 1.0E+06 1.0E+05 1.0E+00 1.0E+02 1.0E+04 1.0E+06 1.0E+08 Time (seconds) 1 µm diameter, 50 nm Ag-Ge-Se on W cathode
Long Term Retention 1.00E+06 10 years @ 70 ºC E a = 0.2 ev On resistance ( Ω ) 1.00E+05 1.00E+04 10 years @ RT Initial resistance Sweet spot? 1.00E+03 15 20 25 30 35 40 Programming current (µa)
Retention example with elevated sense voltage 1.E+14 1.E+12 Out of range Resistance ( Ω ) 1.E+10 1.E+08 1.E+06 1.E+04 1.E+02 200 mv sense 1.E+00 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 Time (s) 2.5 µm diameter, 50 nm Ag-Ge-S on W cathode
Scalability to <100 nm Ag Ag PMMA Ni 80 nm Ag 33 Ge 20 Se 47 Ag 33 Ge 20 Se 47 PMMA Ni 40 nm Si0 2 Si0 2 Devices fabricated using electron-beam lithography (EBL) using PMMA as dielectric
Nanoscale device example (40 nm) 40 nm diameter, 50 nm Ag-Ge-Se on Ni cathode fabricated using EBL 1.2E-03 Current (A) 1.0E-03 8.0E-04 6.0E-04 4.0E-04 2.0E-04 0.0E+00-2.0E-04-4.0E-04-6.0E-04 I prog = 1 ma R off = 10 7 Ω R on = 10 2 Ω 6 consecutive writre/erase operations -8.0E-04-0.5-0.3-0.1 0.1 0.3 0.5 Voltage (V)
Alternative PMC device materials Electrolyte Although chalcogenide glasses such as Ge-S and Ge-Se have excellent ion mobilities, certain oxide glasses make good electrolytes and are very inexpensive to produce and integrate Considering their success in the field of electrochromics, transition metal oxides are a good choice Anode/mobile ions Silver is an ideal material for many reasons but copper has already found its way into the industry and hence is currently cheaper to implement
Oxide electrolyte/ag anode example: Tungsten oxide grown on W cathode 600 1 µm diameter, WOx on W cathode Ag anode Current ( na) 500 400 300 200 100 0-100-0.5-0.3-0.1 0.1 0.3 0.5-200 -300 I prog = 500 na R off = 10 10 Ω R on = 10 5 Ω 8 consecutive writre/erase operations Voltage (V)
Oxide electrolyte/cu anode example: Tungsten oxide grown on W cathode 1.E+11 1 µm diameter, WOx on W cathode Cu anode Resistance (Ohms) 1.E+10 1.E+09 1.E+08 1.E+07 1.E+06 1.E+05 I prog = 500 na, 3 consecutive writre/erase operations First sweep Second sweep Third sweep -1-0.5 0 0.5 1 Voltage (V)
Oxide electrolyte/cu anode example: Tungsten oxide deposited on W cathode 5 µm diameter, 50 nm WOx on W cathode Cu anode I prog = 10 µa, 3 consecutive writre/erase operations
Summary of PMC characteristics Scalable Nanostructured materials allow small physical device size Scalable voltage, current, power, and energy! Flexible functionality Non-volatile, fast, good endurance Manufacturable? Simple structure and simple process BEOL additive with copper protocol Devices do not consume silicon real estate Potentially removes the barrier between memory and logic Opens up EPLD applications space