Embed Size (px)
Citation preview
Logo of Your organization
PAST, PRESENT, and FUTURE of MRAM
Shehzaad KakaNIST Magnetic Technology
325 Broadway, Boulder CO 80305Phone: +1-303-497-7365 FAX: +1-303-497-7364
E-mail: [email protected]
Presented at the THIC Meeting at the STK Bldg 8 Auditorium, 1 Storage Tek Dr, Louisville CO 80027-
9451July 22 - 23, 2003
Outline
•Introduction
•Basics of MRAM operation
•Prototype MRAM design
•Possible Future Directions •Magnetization Dynamics•Spin Momentum Transfer
-120 -80 -40 0 40 80 120
114
116
118
120
122
R (Ω
)
Happl (Oe)
Beginnings
Non-Volatile Magnetic Memory c.a. 1950(but a destructive read!)
ferrite toroidSpintronics:GMR, TMR
Why Now?Memory market dominated by pure semiconductortechnology
DRAM $16 billion LEAKYSRAM $ 4 billion BIGFLASH $ 9 billion (Non-Volatile) SLOW
MRAM (spintronics technology)
May be able to compete!
Universal Memory*but maybe first embedded memory
Desired non-volatile, fast, low power memory,low cost, technology widely being sought.
MRAM : IBM, Motorola, Infineon, Toshiba, Cypress, Micron, . . .
FRAM : Ramtron, Texas Instruments, Fujitsu, Samsung,NEC, …
Other Options – Ovonics, Polymer Ferroelectric memory,Nanotubes, PMC, …???FLASH???
MRAM (Attractive) Properties
• Non-Volatile
• High Density
• High Speed
• Non-Destructive Read
• Low Voltage Operation
• Unlimited Endurance
• Scalable?
• Reliable?
!!! Instant-ON computers !!!
MRAM Elements (Bits) Magnetic Tunnel Junctions
FM/Insulator/FM/AFM Spin-Dependent Tunnelingε
g(ε)
ε
g(ε)
Low Resistance
High ResistanceLarge Resistance Change
1
0
Cross Point Architecture
IBM
Programming Method u M Hs K app= − •12 0
20µ φ µsin M H
φHx
Hy
-50 0 50 100 150 200
Ener
gy (a
rb. u
nits
)
Magnetization Angle (degrees)
Single Domain
Hx
Hy
program
half-select
Astroid
Bit Sense1 Transistor and 1 MTJ element per unit cellAllows for a non-destructive read process
Bit cell size < 1 µm2 using 0.18 µm technology
ComplicationsTunnel Barrier
-very thin (< 2 nm); R, ∆R depends on barrier thickness
-∆R depends on interfaces to the barrier-Pinholes and roughness are problems
Micromagnetic Effects- magnetization not exactly single domain- switching fields sensitive to device shape
Layer Coupling Effects- edge charges on FM layers cause AP ground state- film roughness leads to Neel coupling – Parallel state
Thermal Stability Issues for small bits- caution for half-selected bits switching via thermal activation
+-
++
- ----
+++
---
State of the artMotorola* 1Mbit MRAM chip in 2002
0.6 µm process technology, 7.2 µm cell size, 50 ns read access time, 24 mW at 3V 20 MHz
IBM/Infineon 128 kbit MRAM0.18 µm technology, write/read access ~2ns
bitline
Cypress 64 kb and 256 kb product data sheets on website70 ns access time
Field
Innovations-SAF pinning structure
-Flux concentrating-cladded Cu lines MTJ
* IEEE J. Solid State Circ. 38 769 (2003)
High Bandwidth Magnetic MeasurementsMF HpTest structure with CPW and
µ-strip lines; sub-µm device
MP
Device Easy-axis field switching data
1 2 3 4 5
-1.0
-0.5
0.0
0.5
1.0
Mx/M
s
Time (ns)
24.4 mT 21.5 18.9 16.6 14.5 12.6 11.0 9.5
3 mm X 3 mmSEM Micrograph
Precessional SwitchingMF
MPHL
Hp
APL 80 2958 (2002)
ddt eff effM M H
MM M H= − × − × ×µ γ α µ γ
00 ( )
Landau-Lifshitz equation - dynamics
0 1 2 3 4 5
-1.0
-0.5
0.0
0.5
1.0
Mx/M
Time (ns)
Pulse ∆t (FWHM) 230 ps 280 ps 325 ps
Pulse Response
-1.0 -0.5 0.0 0.5 1.0
-0.05
0.00
0.05
-0.50.0
0.51.0
Mz/M
s
My/M
s
Mx /M
s
Simulation
Hp = 21.5 mT
Precessional Switching – Resulting Remanent States
0.0 0.5 1.0 1.5 2.0 2.5 3.0
0
20
40
60
80
100-1.0
-0.5
0.0
0.5Pr
obab
ility
%
Pulse Duration (ns)
Metastable Full Switch
Mx/M
Hp = 17 mT
The Spin Momentum Transfer Effectτ2 ∝ I M2 X (M1 X M2) ; I ~ 107A/cm2
(-) Current Stable Parallel Alignment Torque
e-
M1 M2(+) CurrentStable Anti-parallel Alignment Torque
e-
M1 M2
Slonczewski, JMMM 159 (1996)
The SMT Effect Nanopillar
-150 -100 -50 0 50 100 150 200
16.35
16.40
16.45
16.50
16.55
16.60
200 mΩ
R (Ω
)
µ0Hx (mT)
-4 -2 0 2 4
7.35
7.40
7.45
7.50
7.55
7.60
7.65
200 mΩdV
/dI (Ω
)
Current (mA)
50-100 nm
SMT ~ 1/r2
Oersted ~ 1/r
Hx = 65 mT
Magnetization control withouta magnetic field!
6.5 7.0 7.5 8.0 8.5
0.0
0.5
1.0
1.5
5 6 7 8
6.9
7.2
7.5
7.8
0 2 4 6 8 10 127.6
7.7
7.8
7.9
8.0
Current (mA)
8.5 mA
7.5 mA
6.5 mA
5.5 mA
4.5 mA
4.0 mA
Ampl
itude
(nV/
Hz1/
2 )
Frequency (GHz)
-0.23 GHz/mA
Pea
k f (
GH
z)
Current (mA)
= spectra
dV/
dI (Ω
)
Magnetization dynamics at in-plane fixed field (0.1T)•Excitation frequency is a
linear function of current•Only on for limited range
of currentpoint contact
Narrow Linewidths !∆f (6.5 mA) ~ 30 MHz
Q = f/∆f ~ 250α < 0.001 ??
Frequency decrease in current – Red Shift
Out of plane precession for out of plane field
5 10 15 2012
14
16
18
20
22
24 m8000sm
X Axis Title
Y Ax
is T
itle
-2E-10-1.12E-10-2.4E-116.4E-111.52E-102.4E-103.28E-104.16E-105.04E-105.92E-106.8E-107.68E-108.56E-109.44E-101.032E-91.12E-91.208E-91.296E-91.384E-91.472E-91.56E-91.648E-91.736E-91.824E-91.912E-92E-9
2 12 2 Bias Current (mA)
Freq
uenc
y (G
Hz)
Experimental Geometry
Currentflow
Mfixed
MInitial
CoFe
NiFe
Applied Field 0.8 T
12 25Frequency (GHz)
0
2
Am
pl. (
nV/H
z1/2 )
V/(Hz)0.5Point Contact Sample
Comparison ChartSRAM DRAM FLASH FRAM MRAM
read time Fast Moderate Fast Moderate Moderate/Fast
write time Fast Moderate Slow Moderate Moderate/Fast
nonvolatile? No No Yes** Yes* Yes
refresh No Yes No No No
cell size Large Small Small Medium Small
low voltage Yes Limited No Limited Yes
*Destructive read** Limited endurance
Motorola
Conclusions
MRAM product will utilize MTJ elements in a cross pointarchitecture
MRAM has attractive properties – maybe capable of surpassing conventional semiconductor memories
Fabrication challenges are being met as MRAM approachesproduction
Increases in magnetic memory speed performance willrequire understanding and control of dynamicsand damping
Spin Momentum Transfer effect may be utilized in advancedMRAM designs eliminating half-select problem
MRAM
