FeRAM, MRAM, RRAM

1
FeRAM, MRAM, RRAM

• Possible successors of
• DRAM and SRAM

Stefano Bonetti, Johan Dahlbäck, Hanna Henricsson
and Jutta Müntjes
26th of October 2005
2B1750 Smart Electonic Materials, KTH
2
Static RAM, SRAM

• Stored data is unchanged as long as power is
  supplied.
• Fast, but expensive to produce (4-6 transistors/
  cell).
• In PCs mostly used for cache memory.

I
NAND
Steady currents
Input
Output
NAND
I
Simple SRAM cell
There are two stable configurations current in
the red or the blue curcuit. To change state,
one of the inputs is energized, the corresponding
NAND-gates output current will cease, thus
flipping to the other stable state. Reading is
conducted by sensing in which output there is a
current.
3
Dynamic RAM, DRAM

• Stored data needs to be refreshed. Hence
  dynamic.
• DRAM is cheaper, but slower, than static RAM.
  (One transistor and capacitor/cell)
• At least in PCs, DRAM constituates all RAM
  except CPU-caches.

A 2 x 2 bit DRAM-cell
To store a 1 the world line and the bit line
are energized simultaneously. If only the word
line is energized a 0 will be stored. Data is
read by sensing if there is a current in the bit
line, when the word line is energized.
4

• Development aims for RAM
• More memory
• Faster readouts
• Neither SRAM nor DRAM can fulfill both aims
  properly.
• Would another technology make it possible?

5
FeRAM - Theory
Example PZT (lead zirconate-titanate)

• Spontaneous polarization above the
  Curie-temperature TC is the structure cubic,
  below a dipole moment occurs (displacement)
• A different charge ?Q can be observed whether the
  material is switching or non-switching

6
FeRAM Failure mechanisms

• A decrease of the remanent polarization reduces
  the difference between switching- and
  non-switching charge
• Polarization fatigue (after repeated read-write
  cycles)
• Retention loss (with time)
• Imprint
• shift of the hysteresis loop leads to preference
  of one polarization state (write failure only
  critical at low voltage) or loss of polarization
  (read failure)
• Increase of temperature leads to worse material
  properties (i.e. defect distribution)

7
FeRAM - Requirements

• Small size
• High speed
• High lifetime
• Destructive reading (after every reading
  operation is a writing operation required)
• Low coercive field
• Low power memory devices
• Large hysteresis
• High remanent polarization

8
FeRAM - Technological Aspects

• Different cell designs
• Problem reduced thickness increases coercive
  field and reduces remanent polarization
• High quality semiconductor/ferroelectric material
• Using proper electrode material to obtain high
  remanent polarization and low coercive field
  (i.e. Pt electrodes for PZT)

9
FeRAM - 1T/1C-Cell

• Read
• WL adressed
• DL adressed with positive voltage Vcc
• BL capacitor divider between Cfe and Cbl, sense
  amplifier compares voltage with Vref
• VltVref Binary state 0
• VgtVref Binary state 1
• But reading operation is destructive,
  information needs to be restored

• Write
• WL adressed
• DL pulse VCC (half length)
• BL VCC 1, ground0

10
M-RAM physical principle

• Tunnel MagnetoResistance (TMR)

• different of states available for spin-up and
  spin-down
• currents in the two different configurations

11
M-RAM simple scheme
Anti-Ferromagnet

• Tipical thicknesses
• few nm for magnetic layers
• lt 1 nm for barriers

12
M-RAM real design operation /1
Read
bit line
different orientation of magnetization
digit line
Transistor On
13
M-RAM real design operation /2
Write
bit line
digit line
Transistor Off
14
M-RAM technological issues

• Accomplished
• sub-micrometric lithography
• uniform deposition of thin films (lt1nm for
  isolation and RKKY coupling barrier)
• integration of TMR material with CMOS
• Challenges in scaling M-RAM techonolgy
• reducing the resistance-area product value (RA)
  mantaining the MR ratio
• generating the switching magnetic fields using
  shrinking metal lines
• accomodating the increased magnetostatic fields
  generated by the reduced dimensions

15
R-RAM physical principle
16
R-RAM Set / Reset

• 1 High-resistance state
• 2 V Vset? Set transition
• 3 Low resistance state
• 4 - -
• 5 V Vreset ? Reset transition
• 6 High resistance state
• States stable

17
R-RAM simple scheme
Top electrode
Pr1-xCaxMnO3
100 nm 600 nm
Bottom electrode
silicon

• Size of the top electrode 100µm in diameter
• Typical size of the small domains 10nm

18
R-RAM Basic circuit diagram
High resistance state ? 1 Low resistance state
? 0 If 1 cell pass-through current
lower If 0 current higher Sense amplifier
senses the current ? determines if 1 or 0
stored Alt. Voltage sense amplifier
19
R-RAM technological aspects
RRAM - promising candidate for next generation
non-volatile memory. - TMO RRAM integrated
with CMOS technology. - Stable high
temperature programming possible up to 300 C.
- The memory resistor can be switched between
high- and low resistance state over a
large number of cycles without memory
degradation. (106 times of set/reset and
1012 times of reading cycles confirmed) -
Cell resistance can be read without affecting
stored data.
20
Performance of FeRAM, MRAM, RRAM

• All three technologies
• Already much faster than DRAM and uses less
  energy.
• Good possibilities to reach SRAM speeds.
• Non-volatile. Possibly replacing hard-drives and
  almost eliminating booting time.
• MRAM seems to be further ahead commercially than
  FeRAM.
• RRAM has size independent properties and
  performance is not degraded at higher
  temperatures.
• Failures and destructive reading proposes
  problems for FeRAM

21
Conclusion

• MRAM is a good candidate to replace DRAM on a few
  years sight.
• RRAM is far from commercial production, but will
  probably prevail over the others in due time.

22
References

• SRAM
• C.R. Nave, hyperphysics.phyastr.gsu.edu/hbase/elec
  tronic/ nandlatch.html, Georgia State University,
  2005
• DRAM
• A Cardon LJL Fransen, Dynamic Semiconductor RAM
  Structures, Pergamon, 1984
• Charles M. Kozierok, www.pcguide.com/ref/ram/,
  2004
• FeRAM
• Rainer Waser (Ed.), Nanoelectronics and
  Information Technology Advanced Electronic
  Materials and Novel Devices, Wiley-VCH, 2003
• Kenji Uchino, Ferroelectric Devices, Marcel
  Dekker, 2000

23

• Yuhuan Xu, Ferroelectric Materials and Their
  Applications, North-Holland, 1991
• www.fujitsu.com (pictures)
• MRAM
• Rainer Waser (Ed.), Nanoelectronics and
  Information Technology Advanced Electronic
  Materials and Novel Devices, Wiley-VCH, 2003
• V. Korenivski, Text reference for Spintronics,
  5A1379, KTH-Physics, Stockholm, 2005
• J. Slonczewski and V. Korenivski, Elements of
  Spintronic Theory for Magnetic Memory, IBM and
  KTH, 2005
• S. Parkin, Magnetic Tunneling Junctions and
  Transistors Magnetic Memory and Field Sensors,
  IBM, 2002

24
RRAM M.J.Sanchez et al, Strong electron
correlation effects in non-volatile electronic
memory devises, Centro Atómico Bariloche,
2005 W.W.Zhuang et al, Novell colossal
magnetoreistive thin film nonvolatile
resistance random accesss memory(RRAM), Sharp
laboraties of America, Sharp corporation,
2002 M.J.Rozenberg, Non-volatile electronic
memories with transition metal oxides, LPS
CNRS/Universite Paris.Sud, M.J.Rozenberg et al,
A model for non-volatile electronic memory
devices with strongly correlated materials,
Université Paris-Sud (France), 2005