Rainer Waser JARA-FIT @ FZJ Forschungszentrum Jlich & RWTH
Aachen University Outline Forschungszentrum Jlich Center of
Nanoelectronic Systems for Information Technology Scaling
Projections for Resistive Switching Memories 1Introduction -
Features and Classification of Resistive RAM 2Electronic switching
effects - including FTJ (brief) 3Fuse-Antifuse switching effect
(brief) 4Ionic switching effects - cation migration redox systems
(electrochem. metallization cells) 5Ionic switching effects anion
migration redox systems Slide 2 - 2 Slide 3 Evolution of the memory
density geometry aspects - 3 R. Waser (ed.), Nanoelectronics and
Information Technology, 2nd ed. Wiley, 2005 Slide 4 Limit:
Dielectric Area Challenges Fundamentals: Physics & Chemistry
Resistive Superparaelectric Limit Interface & Scaling Effects
Ferroelectric DRAM Mega-Bit EraGiga-Bit EraTera-Bit Era Phase
Change Electrochem. Metallization Redox-based Oxides Molecular
Switches Random Access Memories - 4 Slide 5 write-Operation by
large voltage pulses (typically with current compliance) read
operation by small (sensing) voltage pulses Operation Electrical
switching between ON(LRS) and OFF(HRS) state Polarity modes of RRAM
I V Read RESET SET I V Read RESET SET Unipolar (symmetrical) - URS
Bipolar (antisymmetrical) - BRS Memory cell Two-terminal element
between electrodes Active Matrix: 6...10 F 2, 1 cell / 1 T Passive
Matrix: array size e.g. 8 x 8, approx 4 F 2 Basic Definitions of
Resistive RAM History many reports since the 1960s mainly binary
oxides, mainly unipolar switching CC - 5 Slide 6 homogeneously
distributed effect? effect confined to filaments? Location of the
switching event - In the electrode area along the entire path / in
the middle area? at one of the interfaces? Location of the
switching event - Between the electrodes asymmetry of the system
(e.g. different electrode materials) Requirement of bipolar
switching electrical stress as a precondition for switching
sometimes: formation during first cycle(s) Forming process - 6
Slide 7 Classification of the switching mechanisms Resistive
Switching Thermal effect Electronic effect Ionic effect Phase
Change Effect (well known) Fuse-Antifuse effect Charge trap Coulomb
Barrier Charge injection IMT transition effect Ferroelectric
Tunneling barrier Cation migration - electrometallization Anion
migration - redox effect R. Waser and M. Aono, Nature Materials,
2007 - 7 Slide 8 - 8 Slide 9 1.Charge trap effects / Coulomb
repulsion Trapping at an interface e. g. Taguchi Sensor effect or
at internal traps (compare: Flash) e.g. polymer - embedded metal
nanoclusters 2.Insulator-Metal Transition effects Charge doping
change of band structure 3.Ferroelectric tunneling barrier effects
tunneling parallel / antiparallel FE polarization modification of
the tunnel barrier (thickness or height) Variants General Effect
Electronic Switching Mechanisms Charge injection or movement of
displacement charges modification of the electrostatic barrier,
bipolar switching Geometry Homogeneous current density scaling
limits (?) - 9 Slide 10 Ferroelectric Tunnel Junction Basic Effect
Different Tunneling currents parallel and antiparallel to the
ferroelectric polarization (no lateral confinement required)
References: Esaki (1968), Tsymbal, Kohlstedt et al, Science (2006)
No reliable experimental evidence yet ! Results Slide 11
State-of-knowledge Projections Ferroelectric Tunnel Junction Memory
Scaling limits Speed Energy dissipation Challenges No experimental
verification yet; competition by redox-based processes. worse than
MTJ, because the ferroelectricity fades below approx. 2 nm
thickness (ab-initio theory & HRTEM experiment) no inherent
speed limits because polarization reversal in ferroelectrics <
1ns Recharging of a tiny FE capacitor (much smaller than in FeRAM)
low energy switching expected 1. realization of a pronounced effect
unlikely 2. expected scaling limits not favorable Slide 12 - 12
Slide 13 Materials Fuse-Antifuse Switching Mechanism MIM thin film
stack with I = transition metal oxide showing a slight conductivity
e. g. Pt/NiO/Pt SET process Controlled dielectric breakdown e. g.
by thermal runaway formation of a conducting filament RESET process
Thermal dissolution of the filament (fuse blow) disconnected
filament I. G. Baek et al. (Samsung Electronics), IEDM 2004 - 13
Slide 14 Temperature profile - Thermal effect assisting other
switching types? FEM simulation (Ansys ) of metallic TiO filament
(3 nm) in TiO2 matrix Pt TE Pt BE TiO 2 27 nm 54 nm 2 nm 5 nm Pt TE
Pt BE TiO 2 27 nm 54 nm 2 nm 1 filament 3 filaments 390 K 1100 K
Toggle between bipolar and unipolar switching has been possible by
adjusting the current compliance; demonstrated for TiO2 thin films
(Jeong et al. 2006) and Cu:TCNQ (Kever et al. 2006) High current
compliance unipolar fuse/antifuse switching Relationship to other
switching effects - 14 Slide 15 FEM Simulation of the RESET process
160 nm thick NiO film on n-Si with Au top electrodes U. Russo et
al., IEDM 2007 - 15 Slide 16 Current compliance Higher current
compliance (~ 3 mA) leads to transition of the switching mode (BRS
URS) This transition to URS is irreversible: (URS BRS) Details: see
talk H. Schroeder & D.-S. Jeong (F6) and Jeong, Schroeder,
Waser, APL89, 082909 (2006) Transition from the BRS to the URS mode
Pt/TiO2(27nm)/Pt stack, sputter deposited, electroformated at 1mA
for BRS operation LRS of a stable URS Slide 17 State-of-knowledge
Projections Fuse-antifuse effect Scaling limits Speed Energy
dissipation Challenges - thermo-chemical effect; - filamentary
nature confirmed; - nature of ON-state filament still unknown (Ni?
NiOx 0nm 25nm 5 m NC-AFM of etched (100) surface of strontium
titanate (estimated density of dislocations of 4*10 9 /cm 2 ) c-AFM
mapping of local conductivity of (100) surface of thermally reduced
strontium titanate -> hot spots density of strong hot spots
~5*10 10 /cm 2 density of weak hot spots ~5*10 11 /cm 2 8nA 4nA 2nA
8nA 7pA U=3.5V Electrical characteristics of dislocations:
Preferential conductivity paths in SrTiO 3 single crystals K. Szot,
1999, 2004 - 35 Slide 36 I/V 1 I/V 2 -0.5 0.0 0.5 Applied bias (V)
0.01 0.001 100 10 1 1000 0.1 metallic semicond. insulating I/V 1
I/V 2 1.0 0.1 10 0.01 50nm 1nm I ~ 1.2 nA I ~0.009nA Current (nA)
(nA) Thermal preformation by reduction annealing: conductive Tip
AFM Mapping types of I-V Characteristics SrTiO 3 s.c. thermally
reduced at 850 C, pO2 ~ 10 -20 bar K. Szot, W.Speier, G. Bihlmeyer,
R. Waser, Nature Materials, 2006 - 36 Slide 37 a b 50nm 0nm 3nm
50nm 0nm 3nm Applied bias (V) 0 1234 5 n1n1 n2n2 n3n3 n4n4 n5n5
n6n6 n7n7 800 600 400 200 0 n 15 Current (nA) 1000 Resistance ()
1.4 10 10 10 6 10 10 8 3.2 10 6 80 0 40 Distance (nm) 10 12 c d
non-metallic metallic on off K. Szot et al., Nature Materials, 2006
- 37 Slide 38 Edge dislocations in SrTiO 3 crystal (stacking fault)
Jia et al PRL.(2006) TiO 2 SrO/SrO SrO TiO 2 SrO TiO 2 /TiO 2 Slide
39 Formation of localized metallically conducting sub-oxides by
electroreduction N V O 10 22 cm -3 A-B Ti +2 Ti +3 Ti +4 Ti +2 Ti
+3 Ti +4 AB Extended defects after reduction - 39 Slide 40 Redox
Reactions at the Electrodes - Interconnected network of extended
defects - Switching ON oxygen vacancy accumulation near the
surface; conduction through the Ti (4-x)+ sublattice - 40 Slide 41
Learning from lateral cells SrTiO 3 K. Szot, etc. Nature Materials,
2006 M. Janousch et al. Adv. Mater. 19, 2232 (2007) Optical
micrograph and CAFM (above) and Cr K-edge XANES (right) mapping
Slide 42 I(mA) 11 22 33 44 5 & 6 V(V) Switching of SrTiO 3
(100), Potential distribution RT, p=10 -8 mbar Interface I
Interface II I(mA) V(V) 1 2 2 3 3 4 4 5 5 6 E max >10 4 V/cm E
max ~30V/cm E max ~20V/cm E max >10 4 V/cm 11 22 33 44 66 55
I-Source Generator Electrometer SrTiO 3 crystal 3mm K. Szot, to be
published Slide 43 K. Szot, R. Dittmann, R. Waser, rrl-pss, 2007
LC-AFM characterization of epi-STO (10nm) / SRO / STO - 43 Slide 44
Write / erase patterning LC-AFM write/erase processes on epi-STO
(10nm) / SRO / STO K. Szot et al., rrl-pss, 2007 - 44 Slide 45 K.
Szot, R. Dittmann, et al., rrl-pss, 2007 Temperature dependence -
45 Slide 46 Local redox equilibria Continuity equation Poisson
equation Concentration profiles, field profiles, space charge
profiles N V O 10 22 cm -3 A-B Ti +2 Ti +3 Ti +4 Ti +2 Ti +3 Ti +4
Switching model Formation and resistive switching of redox-active
dislocations 3-D network of extended defects? Nanoscale effects ?
Phase formation or frozen kinetics? Electronic charge injection
mode (Schottky emission, FN ?) Fast transport tracks ? Current
questions K. Szot, 2004 - 46 Slide 47 xx i = 1 i = ni = 2 Switching
model former results T. Baiatu, K. H. Hrdtl, R. Waser, J. Am. Cer.
Soc. (1990) Slide 48 Switching model recent results D.-S. Jeong et
al., to be published Pt/TiO2(27nm)/Pt stack, sputter deposited,
electroformated at 1mA for BRS operation Slide 49
State-of-knowledge Projections redox-type memory effect Scaling
limits Speed Energy dissipation Challenges - redox mechanism
suggested during formation; - details of the switching mechanism
need to be studied (thermal assisted?) - nature of ON-state
filament: conducting (sub-)oxide? estimated filament diameter: 2 10
nm ? perhaps similar to ECM; possibly thermal assisted, i.e.
additional dissipation Much better understanding required in order
to optimize material and processing (and to make more precise
projections) < 30 ns ? (perhaps even faster; compare to
fuse-antifuse); literature reports of values of approx. 10 ns Slide
50