Next Generation Memory Devices

1
Next Generation Memory Devices
Sakhrat Khizroev
Center for Nanoscale Magnetic Devices
Florida International University Miami, Florida,
U.S.A.
2
Outline

• Background
• Perpendicular Magnetic Recording
• Three-dimensional Magnetic Recording
• Protein-based memory
• Summary

3
Background
Traditionally, Scaling Laws were followed to
advance data storage technologies
Scaling

At 1 Gbit/in2 information density, bit sizes are
400 x 1600 nm2 At 100 Gbit/in2 40 x 160 nm2 At
1 Tbit/in2 13 x 52 nm2
4
Scaling Smaller Transducers and Media
At 1 Tbit/in2 information density, Bit Sizes are
13 x 52 nm2
5
Superparamagnetic Limit
Magnetic grains
Bit transition
SNR log(N), N - number of grains per bit While
scaling, need to preserve number of grains per
bit to preserve SNR Grain size is reduced for
higher areal densities
6
Media Stability
Probability of magnetization reversal due to
thermal fluctuations
H
Thermally stable media
Relaxation time ? 72 sec for KuV/kT40
? 3.6x109 years for KuV/kT 60
7
Superparamagnetism
If altaminimum, medium becomes thermally unstable
leading to severe deterioration of recorded data
over time.

• Approaches to avoid superparamagnetic
  instabilities
• Decrease aminimum by increasing KU
• Increase a by decreasing the number of grains per
  bit
• Demagnetization fields in transitions shorten the
  relaxation time

HAMR
Patterned Media
Perpendicular Recording
S. Khizroev and D. Litvinov, Perpendicular
Magnetic Recording, Kluwer Academic Publishers,
2004 ISBN 1-4020-2662-5.
8
Perpendicular Recording Well-defined Anisotropy
Magnetic grains
In a typical longitudinal recording layer the
magnetic anisotropy axes of individual grains are
randomly oriented in the plane of the film
2D random medium
In perpendicular recording layer the anisotropy
axis is relatively well aligned (lt2-4 degrees)
perpendicular to the plane of the film
oriented medium
Substantially relaxes the requirements for write
field gradients Can use thicker recording layer -
better thermal stability !!! (increased V in
KUV/kBT ratio)
S. Khizroev and D. Litvinov, Perpendicular
Magnetic Recording, Kluwer Academic Publishers,
2004 ISBN 1-4020-2662-5.
9
Nanoscale Device Tbit/in2 Recording Transducer
Perpendicular
Longitudinal
Bit Sizes 13 x 52 nm2
S. Khizroev, D. Litvinov, Physics of
perpendicular recording writing process, Appl.
Phys. Reviews Focused Review, JAP 95 (9), 4521
(2004).
10
Gap Versus Fringing Field Writing
Higher areal density media requires higher write
fields !!!
In perpendicular recording the write process
effectively occurs in the gap (Write Field lt
4pMS) In longitudinal recording the write process
is done with the fringing fields (Write Field lt
2pMS)
S. Khizroev and D. Litvinov, Perpendicular
Magnetic Recording, Kluwer Academic Publishers,
2004 ISBN 1-4020-2662-5.
11
FIB to Trim Regular Transducers into Nanoscale
Devices
The most critical step is to make
a probe with Nanoscale dimensions
FIB Etch to Define a Nanoprobe
FIB Deposition
S. Khizroev, D. Litvinov, FIB Review in
Nanotechnology 14, R7-15 (2004).
12
Numerical Calculations
Modeled Fields (Quantum-mechanical)
Gallium Ion Implantation
Longitudinal
Perpendicular
Jointly with Integrated Inc. ,a group at Durham
University, UK, and groups at Carnegie Mellon
University
13
FIB-fabricated Nanoscale Transducers
Longitudinal Transducer with a 30 nm Width
Perpendicular Transducer with a 60 nm Width
Note It takes 10 minutes to make one such
device in the University environment
S. Khizroev, D. Litvinov, FIB Review in
Nanotechnology 14, R7-15 (2004).
14
Control of Gallium Diffusion
Ion Dose
AFM
MFM
2 x 106 Ions/cm2
dose increase
3 x 105 Ions/cm2
NOTE Although NO texture change is observed
through AFM, substantial magnetic
grain change is seen through MFM
D. Litvinov, E. Svedberg, T. Ambrose, F. Chen,
E. Schlesinger, J. Bain, and S. Khizroev, Ion
implantation of magnetic thin-films and
nanostructures, JMMM 277 (3-4), xxx (2004).
15
Nanoscale FIB Process
The process how to make Nano-precision patterns
with FIB was shared with a few companies and
successfully implemented by Carnegie Mellon
University, IBM, Seagate, and others
A Part of a Device made in the Industry before
the process was implemented
Same Device made with the process implemented
Side wall
500 nm
500 nm
S. Khizroev, D. Litvinov, FIB Review in
Nanotechnology 14, R7-15 (2004).
16
Characterization
Dynamic Kerr Measurement of the Field from a
Nanoscale Transducer
Kerr-Image Snap-Shots for a SPH Transducer
(Near-field Kerr Microscopy)
These experiments were repeated at Seagate, CMU,
and IBM
D. Litvinov, J. Wolfson, J. Bain, R. White, R.
Chomko, R. Chantrell, and S. Khizroev, Dynamics
of perpendicular recording, IEEE Trans. Magn. 37
(4), 1376-8 ( 2001).
17
Characterization
Ion image of a FIB-fabricated and magnetically
active 3-nm-long feature
MFM image of recorded nanoscale magnetic "dots"
18
Perpendicular Recording with Bit Widths of less
than 65 nm
MFM Images of Nanoscale Size Information
190 ktpi
CoCrPtTa alloy
400 nm
CoB/Pd multilayer
400 ktpi
130 nm
The FIB-made transducer
Current state-of-the-art longitudinal recording
is lt100ktpi
S. Khizroev, D. A. Thompson, M. H. Kryder, and
D. Litvinov, Appl. Phys. Lett. 81 (12), 2256
(2002) Editor's choice for the Virtual Journal
of Nanoscale Science Technology, Sep 23rd 2002.
19
Three-dimensional Magnetic Recording

• Perpendicular Recording promises to defer the
  superparamagnetic
  limit to 1 Terabit/in2
• Heat-Assisted and Patterned Media are still 2-D
  limited and relatively slow

It is expected that Moores law will inevitably
reach its limit between 2010 and 2020 ?
Time to stack multiple active layers on top of
each other
3-D Magnetic Recording is a data storage form of
3-D integration
Conventional and 3-D
Recording Media
Note Each cell is 50 x 50 nm2
20
3-D Magnetic Recording
Lead Ph.D. Graduate Student Yazan Hijazi,
Sakhrat Khizroev

• The development of 3-D magnetic recording is
  divided into two phases
• Multi-level Recording not optimally utilized 3-D
  space
• Note Effective areal density increase is
  by a factor of Log2L (where L is the number of
  signal levels)
• 3-D Recording each magnetic layer is separately
  addressed
• Note Effective areal density increase is by a
  factor of N (where N is the number of recording
  layers)

Note These are not active layers
Note Each cell is 50 x 50 nm2
21
Recording Head
The current in the single pole head is varied to
vary the recording field Each recording is
performed via two pulses 1) a cell is saturated
and 2) the information is recorded
Simulated Recording Field
Schematics of a Transducer
22
Playback Head
The playback head is designed to preferably read
the vertical field component which is dominant in
this case
Stray Field from 3D Medium
Differential Reader Configuration
Electronic Images of FIB-fabricated Transducer
23
Multi-level Recording on a Continuous Medium
Recording Step 2
Recording Step 1

• Major Disadvantages
• Every time a track is recorded into the bottom
  layer, there are side regions in the top layer in
  which the earlier recorded information is lost
  because of the overlapping side region
• The superparamagnetic limit

Recording Step 3
24
Multi-level Recording on a Patterned Medium
FIB-etched Patterned Medium
Patterned Media by Toh-Ming Lu
Note 1 The tilt angle can be controlled via
deposition condition
Note 2 The inter-layer separation should be
sufficient to break the quantum-mechanical
exchange coupling
Note Each cell is 50 x 50 nm2
25
Multi-level Recording on a Patterned Medium
Writing
Note 2 The inter-layer separation should be
sufficient to break the quantum-mechanical
exchange coupling
Micromagnetic Simulation Illustrating Two Cases
of Interlayer Separation
a) lt 1 nm and
b) gt 2 nm
e.a.
M up
M down
Recording Field Profile
26
Multi-level Recording on a Patterned Medium
Writing
H -?
H1gtHc gtH2
Recording Field Profiles in Individual Layers at
a Given Current Value
1
1
2
2
3
3
4
4
Hc
5
5
1
1
1
2
2
2
3
3
3
4
4
4
5
5
5
H4gtHc gtH5
H3gtHc gtH4
H2gtHc gtH3
27
Multi-level Recording on a Patterned Medium
Playback
Magnetic Charge Representation of the Playback
Process
10
9
5
Simulated Stray Field from a 3-D Medium at
different levels of recording
28
MFM Images of Two Types of Media
Each cell is 60 x 60 nm2
29
SNR Limitations

• Patterned Media (ideally, fabrication technique
  limited)
• Electronic noise sources are 10 Ohm GMR Sensor
  and 0.2 nV/sqrt(Hz) preamp noise over a 500 MHz
  CTF bandwidth at 1 Gbit/sec

Note 1 Special encoding channels should be used
to reduce BER
Note 2 The demagnetization field could be fairly
large for some configuration. Special bit
encoding should be considered to avoid the
unfavorable bit configuration.
Hdemag gtgt 4?Ms
30
Three-dimensional Recording
Schematic Diagrams of a 3-D Memory Device
Biasing Conductor for Layer Identification during
Writing
2-D Recording/Reading Grid (similar to MRAM)
Soft Underlayer
31
Magnetically-induced Writing
K-th layer is identified
(K-1)-th layer is identified
Note The current in the biasing conductor is
continuously decreased from the maximum to zero
to identify individual layers starting from the
top to the bottom
32
Thermally-induced Writing
(jointly with Seagate Research)
Simulation by Roman Chomko
33
3-D Reading
Different Implementations

• Active layers MRAM devices stacked together

• Magnetic Resonance FM

• Magnetic Resonance FM

CoCrPtTa alloy
34
3-D Reading Magnetic Resonance Force Microscopy
Comparative MFM Images of Atomic-size Information
obtained by the Conventional State-of-the-art MFM
(left) and the FIU-developed Smart Nano-probe
rf-coil
CoCrPtTa alloy
Electron Image of Smart Nano-probe (made via
FIB)
35
3-D Reading Magnetically-induced Reading
Note 1 Through the variation of the softness
of the SUL, one can vary the sensitivity field of
each cell
Sensitivity Field with a Free SUL (red) and
Saturated SUL (black)
Note 2 Effective physical scanning in the
vertical direction is produced via the variation
of the softness of the SUL. Thus, each layer
could be independently addressed
According to the Reciprocity principle, the
signal in each cell is given by Expression
Provisional patent filed with US PTO on August
4th 2004
36
Parallel Set of Signals at Ibias 0 (A turn)
Recorded Pattern in Layer 6
37
Parallel Set of Signals at Ibias 1.56 (A turn)
Recorded Pattern in Layer 4
38
Parallel Set of Signals at Ibias 5.85 (A turn)
Recorded Pattern in Layer 2
39
Summary on 3-D Magnetic Recording

• The study of 3-D magnetic recording has been
  initiated
• During the last year, the PIs have authored 8
  peer-review papers on the underlying physics of
  magnetic and magneto-thermal recording
• Specific designs of 3-D magnetic devices have
  been proposed
• The university is in the process of filing a
  patent on the proposed mechanism.

Commitment
Within two years, demonstrate an experimental
prototype of a stable (for at least 50 years at
room temperature) 3-D magnetic memory with at
least ten recording layers with an effective
areal density of at least 1 Terabit/in2 and a
data rate faster than 2 Gbit/sec
40
Protein-based Memory

• Why Protein?
• Naturally occurring residues of proteins
  (Bacteriorhodopsin (bR) mutants) in the form of
• molecules with a diameter of less than 3 nm
  (more than 100 times smaller than polymeric
• material used to DVDs) demonstrate
  unprecedented thermal stability at room
  temperature
• (critical advantage over magnetic storage,
  correspond to areal densities of much beyond 10
• Terabit/in2
• Unprecedented recyclablity of protein medium it
  can be rewritten more than 10 million times (more
  than 1000 times better than CD/DVD)
• The light-sensitive properties of proteins
  integrated with the modern semiconductor laser
• technology provide a relatively straightforward
  control of recording and retrieving information
• from the protein media.
• Much faster time response of protein media (as
  compared to magnetic media) the time
• response in the protein media is in the
  picosecond region (as compared to the nanosecond
• region in magnetic media)
• Economical
• Non-volatile

41
Wild-life Bacteriorhodopsine (bR) produced by
Halobacteria
Salinos del Rio on Lanzarote Island
Schematics of a halobacterial cell and its
functional devices
R. R. Birge, Scientific American, 90-95, March
1995
42
Protein-based Memory
Goal is to demonstrate the feasibility of
recording/storing/retrieving information on/from
photochromic proteins at areal densities of above
1 Terabit/in2 and data rates of above 10
Gigahertz.

• Problems with Protein Media
• Early proteins were unstable (Solved with
  discovery of bacteriorhodopsin)
• Polymers, on which protein structures are made,
  are less stable than proteins themselves
• It is not trivial to immobilize proteins in 3-D
• Holographic methods are not perfected for
  ultra-high densities (far from competing with
  magnetic)

• Approach (2-D Single Molecule Level instead of
  3-D) is
• to take advantage of the 2-D stability of BR
  media to record on one surface at a
  single-molecule level or/and use a stack of
  layers to record in 3-D and
• take advantage of the most advanced nanoscale
  recording system so called heat-assisted
  magnetic recording (HAMR) based on the near-field
  optical recording transducer

43
Data Recording/Retrieval in Protein-Based Storage
Thermal Cycle with Two Stable States
44
Recording Mechanism Two photon processes
Fig. Writing digital 1. Transition A ? B. Two
photon absorption causes transition to
intermediate state, which then relax to the
second stable state B.
Cascade two photon absorption.
Note Using two photon and other nonlinear
processes makes possible remote writing digital
information inside optical media volume. It is
applicable for nonvolatile multi-layered optical
memory.
45
Earlier Proposed Protein Memory
Parallel Data Access (page by page via
positioning of the green light)
Issues

• Optics never could record high densities
• 3-D media are not trivial to immobilize

R. R. Birge, Scientific American, 90-95, March
1995
46
The Proposed Solution to Demonstrate the
Feasibility of Protein Based Storage

• All the above-described methods of
  recording/retrieving data are quite complicated
  and it is hard to see whether they will be
  implemented and if yes, when. In fact, so far no
  physical demonstration of ultra-high density
  recording has been made!
• The PIs propose
• first, to use a bit-by-bit 2-D type of recording
  to demonstrate the feasibility
• of the protein-based storage (it is trivial to
  immobilize 2-D media)
• then, to apply one of the available parallel
  data recording/retrieving
• mechanism (e.g. holographic).

To accomplish this goal, the PIs use the
transducer design earlier developed for
heat-assisted magnetic recording (HAMR). HAMR is
the most advanced recording mechanism proposed so
far. The PIs have pioneered one of the most
efficient design of the transducer for HAMR
T. McDaniel, W. Challener, Issues in
heat-assisted perpendicular recording, IEEE
Trans. Magn. 39 (4), 1972-9 (2003).
47
Novel Recording Transducer for Areal Densities
Above 1 Terabit/in2
Note Focused ion beam (FIB) is used to fabricate
apertureless transducers (with aperture
dimensions of less than 100 nm ltlt than the
wavelength)
Electron Image of FIB-fabricated Apertureless
Transducer
Air-bearing-surface (ABS) view of laser diode
with a thin layer of Al with FIB-etched "C" shape
aperture
lt 90 nm
In-house made
F. Chen, A. Itagi, J. A. Bain, D. D. Stancil,
and T. E. Schlesinger, Imaging of optical field
confinement in ridge waveguides fabricated on
very-small-aperture laser, Appl. Phys. Lett. 83
(16), 3245 (2003).
48
Two-Dimensional Protein Media

• Easy to fabricate
• Naturally stable

Optical spectra of a gelatin-mixed BR film in
two states, the ground state and one of the
intermediate M states
AFM Image of a 2-D Pattern with a 2.4-nm Period
The decay absorption signal in the excited
M-state measured at a wavelength of 410 nm
Note Patented approach to immobilize proteins
Into stable thin-film recording media (H.
Arjomandi, V. Renugopalakrishnan)
A gelatin-mixed bR film under study was
fabricated by Lars Lindvold
The spectra were recorded with a Varian CARY 50
spectrophotometer.
49
Experimental setup to record and read information
on/from proteins
Custom-made Near-field System built around
Aurora-3 by DI
Schematic Diagram
Note The modular structure of the system allows
simultaneously using more than one (currently, up
to four) sources (red to blue lasers, UV lamps)
to conduct photons through a fiber to the sample
in the near-field regime. In addition, as
described below, the system will allow
implementing diode lasers assembled right at the
air bearing surfaces (ABS) of the recording
probes attached to the SPMs cantilever.
50
Early Results Reading Tracks from Photochromic
BR Media
Near-field Optical Readback Signal
Narrowest track is 100 nm
The signal is the absorbed power in the
detector system in the reflection mode
51
Summary
100 Gbit/in2
0. End to Longitudinal Recording

• 1. Perpendicular Recording
• 2. Use smaller GrainsDeal with Write Field
  Problem (10x gain)
• Heat Assisted Magnetic Recording (HAMR)
• E.g. high anisotropy 3 nm FePt grains
• 3. Single Grain per Bit Recording combined with
  HAMR (5x gain)
• Patterned Media
• 4. 3-D Magnetic Recording
• 5. Protein-based Memory (Single-Molecule
  Recording)

1 Tbit/in2
10 Tbit/in2
50 Tbit/in2
Ultimate Recording Density gt 50 Tbit/in2
conceivable