Buschow K.H.J. (Ed.) Concise Encyclopedia of Magnetic and Superconducting Materials
Подождите немного. Документ загружается.
Reliability will be used here to include error rate,
archivability, and environmental resilience. For all
technologies, corrected error rates of p1 10
15
and
a 10-year storage life are required. Resilience, e.g.,
shock resistance, will define market size. Hard drives
have shown considerable improvement in this area
and are now being considered for demanding appli-
cations such as portable cameras.
2. Patterned M agnetic Media
In conventional continuous polycrystalline thin film
magnetic media, transition noise is reduced by re-
ducing the grain size and the coupling between grains
(see Magnetic Recording Technologies: Overview,
Micromagnetics: Basic Principles). To circumvent su-
perparamagnetic effects which limit the minimum
grain size, it has been proposed to pattern media
composed of highly coupled grains into discrete sin-
gle domain-size bits. Then, the transition noise would
be determined by the accuracy with which the pattern
edges could be defined. In various embodiments, this
concept is critical to several potential technologies
discussed below. In this section, we address it in its
simplest implementation as a direct replacement for
continuous media in an otherwise conventional disk
system.
Using direct write e-beam lithography (DWEL),
magnetic circular dot arrays of 65 Gdots in
2
(Chou
et al. 1994) and holes at 4300 Gholes in
2
(Xu et al.
1995), close to the e-beam lithography limit, have
been prepared. The highest density array in which
individual elements have been characterized magnet-
ically had only 29 Gdot in
2
produced by DWEL and
showed a surprisingly large 758% variation in co-
ercivity (Haginoya et al. 1999). This is a serious
problem which must be investigated further. There
are, however, no convenient techniques to character-
ize individual elements with dimensions o50 nm, the
present practical limit of magnetic force microscopy.
It is also prohibitively expensive to routinely make
arrays with DWEL large enough to be characterized
on standard magnetometers.
In fact, the highest density array which has been
examined magnetically had 65 Gdots in
2
(Fig. 1)
over B10 mm
2
and was produced by interference li-
thography (Ross et al. 1999). This is an interesting
technique for producing 2D periodic patterns over
large areas with a periodicity somewhat greater than
half the wavelength of the illumination. Although
available sources and photoresists have limited the
periodicity to 100 nm, experiments are in progress to
reduce this by a factor of 2–4. If a periodocity of
25 nm could be achieved, it would seem to offer a
cost-effective process for producing patterned media
with densities of 1000 Gbit in
2
. However, it is not
clear how this can be done in a circular array com-
patible with conventional disk technology.
To minimize the lithography cost, another imagi-
native approach is to write the desired pattern on a
large area die and then use the die repeatedly to
transfer the pattern by printing a masking material
(Wang et al. 1997) or by imprinting the pattern into a
masking material. Using the latter technique to form
a template, Cui et al. (1999) were able to prepare
arrays of Ni particles 70 nm in diameter at a density
of 18 Gbits in
2
over an area of 16 cm
2
. Recalling that
contact lithography was plagued historically by resist
pull-outs, mask damage, and nonuniform contact,
considerable ingenuity will be required to obtain high
yield at nanometer dimensions especially with pro-
cesses which require high temperature or pressure.
However, recent results (Colburn et al. 1999) using a
photopolymerizable liquid between the substrate and
the template to form the etch mask demonstrate re-
liable replication of features as small as 60 nm. A key
feature is the release coating on the template formed
from a self-assembled monolayer.
Still another clever idea is the possible use of pat-
terned structures which can act as electron emitters in
the presence of large electric fields. An additional
magnetic field in combination with the electric field is
used to focus the electrons onto a closely spaced resist
film, thus forming an image of the pattern in the re-
sist. This has the potential for rapidly reproducing
high-resolution images over large areas. Resolution
better than 50 nm has been demonstrated (Tromp
2000).
The development of a reliable head/disk interface,
which allows heads to fly less than 25 nm above the
media surface, is a major factor in the continuing
increases in disk storage density. Providing a suitably
robust surface is a challenge for patterned media
which significantly complicates additive or subtrac-
tive fabrication processes. A more attractive ap-
proach in which the pattern is formed magnetically
rather than geometrically has recently been demon-
strated (Devolder et al. 1999). Ion implantation of
perpendicularly oriented Co–Pt multilayers with
Figure 1
Electron micrograph of evaporated Ni pyramids with
base 35 nm and height 40 nm. The pyramids form a
100 nm period square array, which appears
foreshortened in the micrograph (from Ross et al. 1999).
540
Magnetic Reco rding Devices: Future Technologies
30 keV He
þ
ions through a mask significantly re-
duced the coercivity and allowed regions as small as
50 nm to be defined in a planar film. In other exper-
iments, the Curie temperature was also reduced, sug-
gesting the possibility of leaving magnetic islands in a
nonmagnetic matrix.
An elegant way to avoid the lithography problem
would be to have the patterning done by nature
through a process of self-assembly. This is now the
subject of intense study across many disciplines well
beyond the focus of this review. However, one very
relevant scheme relates to the self-ordering of pores in
alumite. It is well known that the anodization of al-
uminum can produce a disordered porous structure.
For many years, workers used the structure as a
template to study electroplated magnetic nanowires
with diameters as small as 10 nm. Extended multiple
anodization has been shown to produce beautifully
ordered hexagonal arrays of pores (Fig. 2). This sug-
gested immediately the possibility of self-assembled
perpendicular media obtained by electroplating mag-
netic wires into the ordered pores.
Despite considerable efforts, however, it has not
been possible to increase the size of the ordered do-
mains to more than 100 mm
2
and that only after
lengthy anodization of up to 87 hours. In addition,
conditions for ordering constrain the periodicity to
B100 nm (64 Gpores in
2
). Remarkable progress in
circumventing the first of the two problems has been
demonstrated, however. Ordered pores over 20 mm
2
were produced by anodization of an Al surface in-
itially decorated by imprinting with a SiC die. The
highest density achieved was 64 Gpores in
2
, limited
by the resolution of the die, patterned with DWEL. If
this can be extended to higher densities and circular
arrays, it would offer another route to a practical
media process.
A chemical method for preparing self-assembled
thin film arrays with either cubic or hexagonal
symmetry of highly uniform spherical FePt single
crystal particles has been described by Sun et al.
(2000). After annealing above 500 1C to obtain the
high anisotropy FCT phase, randomly oriented 4 nm
particles on 6 nm centers with coercivities in excess
of 4000 Oe were obtained. This exciting develop-
ment has enormous potential both in the near term
where a single bit in a conventional rotating disk
system could include many particles and in the future
where particles might be addressed individually. The
latter would translate to densities in excess of
10 Tbits in
2
.
In summary, lithography limits are a major obsta-
cle to producing cost-effective patterned media with
densities 4100 Gbits in
2
. However, new develop-
ments in interference lithography, imprint technology
and patterned implantation could extend this to
1000 Gbits in
2
. Self-assembly offers the potential for
even higher densities. It is, of course, assumed that
processing technology will allow heads with com-
mensurate dimensions to be fabricated.
Figure 2
Atomic force micrograph (1.5 1.5 mm
2
) of ordered aluminum Al (plus Al
2
O
3
formed in air) after anodizing in
H
2
C
2
O
4
for 25 hours, then stripping Al
2
O
3
porous film with H
3
PO
4
/H
2
CrCo
4
. The darker spots are the pore bottoms.
541
Magnetic Reco rding Devices: Future Technologies
3. Probe-based Storage
The development in the last 15 years of scanning
probe microscopy (SPM) based on precisely posi-
tioned micromechanical cantilevers has resulted in a
variety of techniques for characterizing, imaging, and
manipulating surface layers with resolutions and po-
sitioning accuracies ranging from 1 10
1
–1 10
2
nanometers (Wiesendanger 1998). The potential ap-
plication to high-density storage has been widely rec-
ognized although inherent speed limitations posed
what initially appeared to be a fatal weakness. How-
ever, equally rapid developments in microelectro-
mechanical system (MEMS) devices suggest the
possibility of operating multiple probe heads in
parallel to achieve acceptable data rates.
The ability to fabricate probes with tips a few
nanometers in size is a key element to SPM. Fur-
thermore, since the probes and their supporting can-
tilevers are usually formed by etching from a silicon
wafer, it is possible to integrate active devices into the
probe structure allowing a wide range of functional-
ity based on electrical, magnetic, mechanical, optical,
and thermal phenomena. Four implementations are
of particular interest for storage: scanning tunneling
microscopy (STM), atomic force microscopy (AFM),
magnetic force microscopy (MFM), and scanning
near-field optical microscopy (SNOM). In STM, the
tunneling current between the probe tip and the sur-
face is used to measure the topography and electronic
characteristics of the surface with atomic resolution.
In AFM, the probe tip is kept close to or in contact
with the surface and the force on the tip used to de-
tect topography. Special AFM probes have also been
used to mark the surface mechanically or thermally
or to make electrical contact.
Carbon nanotubes attached to probes are partic-
ularly attractive because of their size, flexibility, and
electrical and magnetic characteristics. Although
atomic resolution is also possible with AFM, it is
typically not less than a few nanometers in the con-
tact mode, limited by the size of the probe tip. In the
most common form of MFM, the probe tip is coated
with magnetic material and the force (or force gra-
dient) measured between the tip and the stray field
above the surface of the sample arising from a diver-
gence of the magnetization in the sample (see
Magnetic Force Microscopy). Since all the magnetic
material on the probe contributes to the force, the
resolution of MFM is 30–50 nm, the poorest of the
four techniques. Instruments using other magnetic
interactions such as magnetic resonance or torque
have been developed but their applications to storage
have not been discussed. In SNOM, special lenses or
apertures in close proximity to a surface are used to
provide resolution in the near field several times less
than the wavelength of the illumination. In one con-
figuration using interference, the resolution observed
was an extraordinary l/800 (Zenhausern et al. 1995).
While the use of STM for surface analysis has be-
come ubiquitous, the realization of a practical storage
scheme still faces enormous challenges. Although
STM has the highest resolution, its extreme sensitiv-
ity to tip-to-surface spacing and the likelihood of
required operation in UHV to avoid surface contam-
ination, makes STM the least explored of the four
methods at this time. One recent report (Hou et al.
1999) however, claims the record for the smallest
observable induced features, 1.4 nm on 9 nm centers
(10 T features in
2
). They were obtained in ambient
conditions by applying a voltage pulse to an organic
thin film which caused a local transition from insu-
lating to metallic for reasons not yet understood. In
another very beautiful piece of work, Heinze et al.
(2000) used spin-polarized STM to image the anti-
ferromagnetic superstructure on the surface of a
Mn film with a periodicity o0.5 nm. Although these
demonstrations are far from a storage system, they
do highlight the advantages STM has in resolution
and in electronic access.
In contrast, considerable work on the utilization of
AFM and MFM has been reported. Much of it has
focused only on the formation of some observable
array of features particularly with contact AFM. For
example, Cooper et al. (1999) formed 8 nm titanium
oxide dots on 20 nm centers by anodizing a titanium
film through the tip of a carbon nanotube. While this
remarkable achievement corresponds to 1.6 Tdots
in
2
, it would appear to be limited to write-once ap-
plications and at the reported writing speed of 5 kHz,
would require massive parallel operation. The diffi-
culties in reducing an array of nanodots to a practical
AFM storage system have been considered most
completely by Mamin et al. (1999). They developed a
cantilever probe (Fig. 3) which could be used both to
write features thermally in a polymetric film on a
rotating disk and also to read the features with an
integrated piezoresistive sensor.
Successful operation was obtained with 120 nm
marks written at 50 kbits s
1
. (It is ironic to note that
thermal effects which limit magnetic storage can eas-
ily be detected in the noise spectrum at the resonant
frequency of a few megahertz in these sensitive
cantilevers.) The authors extended the study to a
read-only system in which 100 nm marks preformed
in a disk were read at 10 Mbits s
1
with a probe/disk
velocity of 1 ms
1
. In another test at only 30 mms
1
,
sufficient probe wear occurred after 148 hours to
cause an observable reduction in output. The critical
problem of tracking was also considered and the
authors concluded that pre-written timing marks
smaller than the present limit of DWEL would be
required for track pitches below 100 nm.
Improvements in writing thermally have been
reported by Binnig et al. (1999) who produced
40 nm pits on 40 nm centers (400 Gbits in
2
) (Fig. 4)
in a 40 nm thick film of polymethylmethacrylate
(PMMA) with 2 ms pulses (500 kbits s
1
). Using the
542
Magnetic Reco rding Devices: Future Technologies
temperature-dependent resistance of the cantilever,
they were also able to read with the same probe at
100 kbit s
1
limited by the 10 ms thermal response
time of the 1 mm thick cantilever. It was estimated
that to reduce the response time to 1 ms, levers with
thicknesses measured in nanometers would be re-
quired. The possibility that repeated block erasure of
the pits could be achieved by heating local regions to
150 1C for a few seconds was also demonstrated.
This concept was combined with a remarkable ad-
vancement in MEMS technology, the 1000 probe
‘Millipede’ (Vettiger et al. 2000) to test feasibility
(Lutwyche et al. 2000). The ‘‘Millipede’’ (Fig. 5) is a
32 32 array of AFM probes on 92 mm centers with
integrated read/write capability. Rows and columns
of probes can be accessed to allow 32 probes to write
and read in parallel. The array was positioned glo-
bally with respect to the polymer film medium with-
out separately controlling the height of each probe.
Data was written and read back at 1 kbits
1
per
probe giving a total data rate of only 32 kbits s
1
.
However, this limit was set by the test system and it is
believed that higher data rates are possible. The den-
sity was varied from 15 Gbits in
2
with 480% of
the probes operational to 300 Gbits in
2
with a lower
probe yield. No information on erasure or reliability
was given.
Electrical phenomena for probe-based storage such
as charge trapping (Terris and Barrett 1995) or non-
linear conduction in Langmuir-Blodgett films (Sakai
et al. 1988) have been investigated but more recent
reports are not available. One would also expect that
phase change or ferroelectric materials should be
candidates for probe-based media, but few results
have been presented.
The conspicuous advantages of magnetic materials
for re-writable storage applications makes MFM a
natural choice for investigation. Indeed, most of the
work on patterned magnetic media has used MFM to
characterize the elements with a significant effort
demonstrating that individual elements could be
switched with the field from the probe tip alone
and/or with an external field. Curiously, no results on
prototype MFM system comparable to those avail-
able for AFM have appeared. This may reflect a
combination of the difficulty described above in pre-
paring suitable patterned media, the relatively low
resolution of MFM probes, and the complexity of an
integrated active bipolar probe for writing.
However, Carley et al . (2000) presented a compre-
hensive and imaginative design paper on the imple-
mentation of a small capacity (2 GB) miniature
(2.5 cm
3
) re-writable system with a data rate of
20 MB. It relies heavily on MEMS technology to
position a media sled relative to a highly integrated
array of 1000 probes in which the height of each
probe is controlled individually relative to the media
(Fig. 6). An important feature is the assumption
of continuous perpendicular media for which it is
asserted that superparamagnetic effects will be less
severe than in conventional longitudinal media.
A bit size of 50 nm 50 nm is specified which cor-
responds to an intrinsic density of 4250 Gbits in
2
,
Figure 4
(a) Series of 40 nm data pits formed in a uniform array,
demonstrating reliable thermal writing. (b) Different
arrangement of data pits, with some pits written
virtually next to each other with no merging. This results
in bit areal densities of B400 Gbits in
2
. All images
obtained by the thermal readback technique (after
Binnig et al. 1999)
Figure 3
SEM images of a 29 mm long piezoresistive cantilever
with a separate resistive heater for writing (after Mamin
et al. 1999).
543
Magnetic Reco rding Devices: Future Technologies
although the effective density in the system is reduced
to 155 Gbits in
2
. The large number of probes oper-
ating in parallel increases the data rate to 420 MB. It
is envisioned that each probe would incorporate a
write coil and a recessed GMR sensor using the mag-
netic material on the pole tip to guide the flux from
the bit to the sensor. Calculations not presented in
detail suggest that the total write power would be
practical and that reasonable output signals could be
obtained. No discussion of the interconnect problem
which would seem quite formidable is presented.
In reflecting on the difficult technical challenges
inherent in a massively parallel MFM system, it is
sobering to recognize that an existing disk product
announced in 1999 had a capacity of 340 MB, a data
rate of 3 MB, a volume of 7.8 cm
3
and could be pur-
chased off the Internet for o$450. Given the evolu-
tionary improvements expected in disk technology, it
is clear that the target for future probe-based systems
must be even more aggressive than that set by Carley
et al. (2000). Furthermore, if continuous perpendic-
ular media can be used to extend the superparamag-
netic limit, it would seem that a single conventional
head with no greater level of processing complexity
than an array of MFM probe heads could be used in
a rotating disk system to achieve higher performance
at lower cost.
The use of SNOM was much heralded in several
commercial ventures to offer a factor of ten improve-
ment in density relative to conventional magnetic
storage. To date this has not been realized for reasons
that are best discussed elsewhere. However, the
unique capability of the microscope described by
Zenhausern et al. (1995) has been tested for a read-
only application by Martin et al. (1997) (Fig. 7). By
observing the modulation of the light scattered from
a sharp probe due to the dipole–dipole coupling be-
tween the probe and the surface being scanned, a
resolution of l/800 can be achieved with visible light.
The authors were able to detect 50 nm pits and es-
timated that it should be possible to detect 2 nmbits
(440 Tbits in
2
) at 100 MHz with 50 dB SNR.
In summary, probe-based technologies allow very
small features to be resolved and therefore offer the
potential for storage densities of 1 10
12
–1 10
14
bits in
2
. However, while present work suggests this
could be achieved in write-once systems, re-writable
technologies competitive with disk storage will be
Figure 5
SEM images of the cantilever array section of the millipede with approaching and thermal sensors in the corners,
array and single cantilever details, and tip apex (after Vettiger et al. 2000).
Figure 6
Conceptual diagram MEMS-actuated data storage
devices with x–y motion of the media and z motion of a
4 5 array of probe tips (after Carley et al. 2000).
544
Magnetic Reco rding Devices: Future Technologies
more difficult. In all cases, data rate remains a major
obstacle and the use of probe arrays as a solution is at
this time largely conceptual. Finally, probe systems
that require contact between the probe tip and
the media may be restricted to considerably lower
densities due to tip wear.
4. Optical Storage
The complex optical properties of many materials
offer a rich opportunity for exploitation in data stor-
age. To circumvent the constraint on areal density
imposed by the wavelength of light in conventional
optical storage, it has been proposed either to store
multiple bits in a single location using spectral hole-
burning or to store bits in three dimensions using
holography or two-photon absorption.
The possibility of using selective spectral absorp-
tion to store data has been considered for more than
20 years but after an intense but relatively unsuc-
cessful international effort to find appropriate mate-
rials, interest has declined. The basic concept, present
status and challenges have been reviewed recently by
Maniloff et al. (1999). In a variety of materials, so-
called inhomogeneous absorption occurs over a rela-
tively broad band of optical frequencies due to local
nonuniformities (Fig. 8(a)). However, selective ab-
sorption can also occur if the material is illuminated
with narrow band light (Fig. 8(b)) which can be de-
tected later as reduced absorption referred to as a
hole at that frequency in the inhomogeneous absorp-
tion spectrum (Fig. 8(c)). This raises the possibility of
storing many bits represented by different frequencies
inside a single volume of material since the ratio
of the two bandwidths could be several orders of
magnitude.
Bulk erasure would presumably be required. Un-
fortunately, the lifetimes of the excited states are lim-
ited, ranging from milliseconds to a few hours even in
the best rare-earth doped materials and then only at
cryogenic temperatures. Also, interactions between
the excited states significantly reduces the ratio of the
bandwidths and the fast tunable lasers which would
be required are not available. The best result that
has been achieved is a few thousand bits in a
30 mm 30mm spot (B3 Gbits in
2
). Finally, the rate
at which data can be recorded or read is limited fun-
damentally to be less than or equal to the spectral
channel width. This has led to significantly more
complex systems using holographic spatial modula-
tion as well to write and read data blocks in parallel.
Until solutions to these fundamental problems are
found, spectral hole-burning will not pose a challenge
to disk storage.
Holography is another technology with a long
unrealized promise for data storage. The concept
(Fig. 9) is basically the same today as it was 30 years
ago. Instead of storing data as bits in a localized spot,
a 2D array of bits composed as a page is illuminated
with a laser and the holographic image caused by the
interference between the array image and a reference
beam captured in a photosensitive material. The data
Figure 7
Principle of scanning interferometric apertureless
microscope in reflection mode (after Martin et al. 1997).
Figure 8
An illustration of a material’s (a) inhomogeneous
absorption and (b) homogeneous absorption. If the
material is illuminated by narrow-band light, a spectral
hole is produced in the inhomogeneous distribution, as
shown in (c) (after Maniloff et al. 1999).
545
Magnetic Reco rding Devices: Future Technologies
is read by illuminating the hologram and capturing
the image of the reconstructed data on a 2D detector.
Because the data is written and read in parallel, very
high data rates should be possible. By varying the
angle of the reference beam, multiple holograms
can be stored in the same volume giving the potential
for very high volume storage densities. When holo-
graphic storage was first conceptualized, every com-
ponent in the system faced significant practical
limitations so interest shifted to other less demand-
ing optical schemes. However, significant improve-
ments in laser light sources, spatial light modulators
(SLMs) and charge-coupled device (CCD) light
detectors led, in the last few decades, to a re-exam-
ination of holographic storage particularly by a con-
sortium of industry and university researchers funded
in part by DARPA. A comprehensive review of some
of this work placed in historical context has been
presented recently by Ashley et al. (2000).
Of the several problems still preventing the devel-
opment of a practical holographic storage system, the
recording material continues to be the most serious.
Although other major components need further im-
provement, their usefulness in other applications has
paced their development. One can expect smaller,
higher power, lower cost solid state lasers; continued
progress in displays means higher capacity, faster
SLMs for page composers; newly developed CMOS
detectors promise lower cost and higher performance
than CCD devices. However, although new photo-
polymer materials with improved sensitivity are
attractive for write-once applications, no materials
are known which meet the stringent characteristics
required for high data rate, high-density rewritable
storage competitive with hard disks.
Holographic materials capable of storing many
holograms must have high optical quality in layers
several millimeters thick, a large refractive index
change per unit exposure measured by the sensitivity
S, and a large dynamic range measured by the
M-number, which is defined as the sum of the dif-
fraction efficiency over M holograms in one loca-
tion. Ashley et al. (2000) estimate to achieve data
rates of B1Gbs
1
and densities of approximately
100 Gb in
2
under practical conditions, values of
SD20 cm
2
J
1
and MD5 are required. They also pro-
vide values of S and M for several potential media
candidates. The classic material has been bulk single
crystals of photorefractive Fe-doped lithium niobate.
However, its values of S ¼0.02 and M ¼1 fall well
below the target values. In addition, since the refrac-
tive index change is linear in light intensity, data can
be erased during the read operation. Photopolymers
with S ¼20 and M ¼1.5 have recently been reported.
However, their thicknesses are limited to o1 mm and
they are not rewritable.
In summary, the realization of practical holo-
graphic systems will require continued improvement
in all major components. The most serious problem is
the storage material which, at the present time, is
limited to write-once applications.
Another scheme to achieve volumetric optical stor-
age is the so-call two photon process. The idea is
conceptually simple and based on a well-documented
effect in many materials in which the absorption of
energy in the form of photons produces a measurable
chemical change from an initial state A to a final state
B. Then, A can be used to represent a ‘‘one’’ and B to
represent a ‘‘zero.’’ If the energy E
w
required to write,
i.e., cause the transition from A to B, is provided by
two photons, each with energy oE
w
then writing will
occur only at those locations where both photons are
coincident in time. Thus, writing can be made to
occur at any point in a 3D matrix by steering two
optical beams to the desired location. In order to
read optically, the optical characteristic of A and B
must be distinguishable in some way which does not
disturb the stored data.
The reduction to practice of the two-photon con-
cept has been discussed in several papers. One pro-
posed material system was photochromic spiropyran
for A, which after absorption at 350 nm (E
w
) converts
to merocyanine (B). For reading, the latter conven-
iently has an absorption peak of B550 nm. After ab-
sorption of an incident ‘‘read’’ photon, it fluoresces
immediately back to the ground state of B. Detection
of the fluorescence identifies the site as a ‘zero’. The
reverse conversion from B to A is less straightfor-
ward. Since (apparently) A is a lower energy state
than B, the addition of sufficient energy to B will
Figure 9
Basic holographic data storage system. Data are
imprinted onto the object beam with a pixelated input
device called a SLM. A pair of lenses image the data
through the storage material onto a pixelated detector
array such as a CCD. A reference beam intersects the
object beam in the storage material, allowing the storage
and later retrieval of holograms (after Ashley et al.
2000).
546
Magnetic Reco rding Devices: Future Technologies
result in its transformation back to A. Bulk erasure
by heating the storage medium may be necessary.
Very little information about the practical operat-
ing characteristics for a two-photon re-writable sys-
tem have been given. The asymmetry between the
A-B and B-A transitions suggests that these ma-
terials may be more suited to write-once applications.
Certainly, the reliability of rewriting has not been
demonstrated. No information has been given on the
speed of either the write or read process or the timing
margin for writing.
The problem of achieving spatial and temporal co-
incidences is not trivial so more recent work (Zhang
et al. 2000) has used a single beam with variable focus
depth to achieve vertical spatial resolution on a ro-
tating disk. Data bits 3.5 3.5 20 mm
3
were record-
ed in an undefined material using a Ti:sapphire laser
on 25 layers with a 75mm layer spacing. Using a con-
tinuous 532 nm laser, a CNR of 42 dB in a bandwidth
of 1 kHz was achieved at a laser power of 5 kW cm
2
.
At 20 kW cm
2
, significant data erasure was observed
after multiple read cycles. Erasure or re-writability
were not discussed.
Clearly, while the two-photon concept is elegant,
the performance is severely hampered by available
materials. Systems may be limited to write-once
applications and are not competitive with hard disks.
5. Concluding Remarks
The continued improvement in cost and performance
of magnetic hard drives poses an extremely difficult
challenge for new technologies. Lithographically pat-
terned media may offer only an incremental improve-
ment over continuous media. 2D optical technologies
face fundamental wavelength limitations and 3D con-
cepts require materials which do not yet exist. At the
present time, practical storage systems using self-
assembled media addressed with multiple probe heads
cannot be realized. However, enormous progress is
being made both in understanding the fundamental
aspects of self-assembly and in the fabrication of
complex MEMS devices. Together, they may provide
the next revolution in storage especially suited for
the miniaturization of personal computing devices
expected in the 21st century.
Acknowledgment
This work was supported by the NSF Materials
Research Science and Engineering Center (DMR-
9809423). The author would also like to acknowledge
valuable information on disk storage provided by T.
Howell, Quantum Corporation.
See also: Magnetic Recording: Patterned Media;
Magnetic Recording Heads: Historical Perspective
and Background; Magneto-optic Recording: Over-
write and Associated Problems; Magneto-optical
Effects, Enhancement of
Bibliography
Ashley J, Bernal M P, Burr G W, Coufal H, Guenther H,
Hoffnagle J A, Jefferson C M, Marcus B, MacFrlane R M,
Shelby R M, Sincerbox G T 2000 Holographic data storage.
IBM J. Res. Develop. 44, 341–66
Bertram H N, Williams M 2000 SNR and density limit esti-
mates: A comparison of longitudinal and perpendicular
recording. IEEE Trans. Magn. 36, 4–9
Binnig G, Despont M, Drechsler U, Ha
¨
berle W, Lutwyche M,
Vettiger P, Mamin H J, Chui S W, Kenny T W 1999 Ultra-
high-density atomic force microscopy data storage with erase
capability. Appl. Phys. Lett. 74, 1329–31
Carley L R, Bain J A, Fedder G K, Greve D W, Guillow D F,
Lu M S C, Mukherjee T, Santhanam S, Abelnann L, Mim S
2000 Single chip computers with microelectromechanical sys-
tems-based magnetic memory. J. Appl. Phys. 87, 6680–5
Chou S Y, Wei M S, Krauss P R, Fischer P B 1994 Single
domain magnetic pillar array of 35 nm diameter and 65
Gbits/in
2
density for ultrahigh density quantum magnetic
storage. J. Appl. Phys. 75, 6673–5
Colburn M, Johnson S, Stewart M, Damle S, Bailey T, Choi B,
Wedluke M, Michaelson T, Sreenivasan S V, Willson C G
1999 Step and flash imprint lithography: A new approach to
high-resolution patterning procedures. SPIE 3676, 379–89
Cooper E B, Mamalis S R, Fang H, Dai H, Matsumoto K,
Minne S C, Hunt T, Quate C F 1999 Terabit-per-square-inch
data storage with the atomic force microscope. Appl. Phys.
Lett. 75, 3566–7
Cui B, Wu W, Kong L, Sun X, Chou S Y 1999 Perpendicular
quantized magnetic disks with 45 Gbits on a 4 4cm
2
area.
J. Appl. Phys. 85, 5534–6
Despont M, Brugger J, Dreschler U, Durig U, Haberle W,
Lutwyche M, Roithuizen H, Stutz R, Widmer R, Roher H,
Binning G K, Vettiger P 1999 VLSI-MEMS chip for AFM
data storage. IEEE Int. Micro Electro Mechanical Sys. 564–9
Devolder T, Chappert C, Chen Y, Cambril E, Bernas H, Jamet
J P, Ferr E 1999 Sub-50nm planar magnetic nanostructures
fabricated by ion irradiation. Appl. Phys. Lett. 74, 3383–5
Haginoya C, Heike S, Ishibashi M, Nakamura K, Koike K,
Yoshimura T, Yamamoto J, Hirayama Y 1999 Magnetic
nanoparticle array with perpendicular crystal magnetic
anisotropy. J. Appl. Phys. 85, 8327–31
Heinze S, Bode M, Kubetzka A, Pietzsch O, Nie X, Blu
¨
gel S,
Wiesendanger R 2000 Real-space imaging of two-dimension-
al antiferromagnetism on the atomic scale. Science 288,
1805–8
Hou S M, Zhao X Y, Yang C, Xue Z Q, Yang W J, Chen H Y
1999 Ultrahigh density data storage in an organic film with a
scanning tunneling microscope. J. Vac. Sci. Technol. B 17,
2467–70
Lutwyche M I, Despont M, Dreschler U, Du
¨
rig U, Ha
¨
berle w,
Rothuizen H, Stutz R, Widmer R, Binig G K, Vettiger P 2000
Highly parallel data storage system based on scanning probe
arrays. Appl. Phys. Lett. 77, 3299–301
Mamin H J, Ried R P, Terris B O, Rugar D 1999 High-density
data storage based on the atomic force microscope. Proc.
IEEE 87, 1014–27
Maniloff E S, Johnson A E, Mossberg T W 1999 Spectral data
storage using rare-earth doped crystals. MRS Bull. 46–50
547
Magnetic Reco rding Devices: Future Technologies
Martin Y, Rishton S, Wickramasinghe H K 1997 Optical data
storage read out at 2.56 Gbits/in
2
. Appl. Phys. Lett. 71, 1–3
Ross C A, Savas T A, Smith H I, Hwang M 1999 Modelling of
hysteresis loops of arrays of 100 nm period nanomagnets.
IEEE Trans Magn. 35, 3781–3
Sakai K, Matsuda H, Kawada H, Eguchi K, Nakagiri 1988
Switching and memory phenomena in Langmuir-Blodgett
films. Appl. Phys. Lett. 53, 1274–6
Sun S, Murray C B, Weller D, Folks L, Moser A 2000 Mono-
disperse FePt Nanoparticles and ferromagnetic FePt nano-
crystal superlattices. Science 287, 1989–92
Terris B D, Barrett R C 1995 Data storage in NOS: Lifetime
and carrier-to-noise measurements. IEEE Trans. Elec. Dev.
42, 944–9
Tromp R M 2000 Low-energy electron microscopy. IBM
J. Res. Dev. 44, 503–16
Vettiger P, Despont M, Drechler U, Du
¨
rug U, Ha
¨
berle W,
Lutwyche M I, Rothuizen H E, Stutz R, Widmer R, Bining G
K 2000 The ‘Millipede’—more than one thousand tips for
future AFM data storage. IBM J. Res. Dev. 44, 323–40
Wang D, Thomas S G, Wang K L, Xia Y, Whitesides G M
1997 Nanometer scale patterning and pattern transfer on
amorphous Si, crystalline Si and SiO
2
surfaces using self-
assembled monolayers. Appl. Phys. Lett. 70, 1593–5
Weisendanger R 1998 Scanning Probe Microscopy. Springer-
Verlag, Berlin
Wong J, Sherer A, Todorovic M, Schultz S 1999 Fabrication
and characterization of high aspect ratio perpendicular pat-
terned information storage media in an Al
2
O
3
/GaAs subst-
rate. J. Appl. Phys. 85, 5489–91
Wood R 2000 The feasibility of magnetic recording at a tera-
byte per square inch. IEEE Trans. Magn. 35, 36–42
Xu W, Wong J, Cheng C C, Johnson R, Sherer A 1995 Fab-
rication of ultrasmall magnets by electroplating. J. Vac. Sci.
Technol. B 13, 2372–5
Zenhausern F, Martin Y, Wickramasinghe H K 1995 Scanning
interferometric apertureless microscopy: Optical imaging at
10 angstrom resolution. Science 269, 1083–5
Zhang H, McCormick F B, Dvornikov A S, Chapman C,
Walker E P, Kim N-H 2000 Single-beam two-photon-
recorded monolithic multi-layer optical disks. Proc. SPIE
4090, 174–8
W. D. Doyle
The University of Alabama, Alabama, USA
Magnetic Recording Devices:
Head/Medium Interface
Magnetic storage and retrieval devices include tape,
flexible disk, and rigid disk drives used for audio,
video, and data processing applications. The mag-
netic recording process involves relative motion be-
tween a magnetic medium (tape or disk) and a
stationary or rotating read/write magnetic head. For
construction and materials used in magnetic head and
medium components see Bhushan (1996a, 1999a,
2000). The need for ever-increasing recording densi-
ties requires that surfaces should be as smooth as
possible and flying height (physical separation or
clearance between a head and a medium) be as low as
possible. In modern tape and flexible and rigid disk
drives, flying height ranges from about 25 nm to
50 nm, and roughness of the head and medium sur-
faces ranges from about 1.5 nm to 5 nm r.m.s. As an
analogy, a magnetic head slider flying over a disk
surface with a flying height of 25 nm with a rela-
tive speed of 20 ms
1
is equivalent to an aircraft flying
0.2 mm above the ground at 900 kmh
1
(Fig. 1). This
is what a disk drive experiences during its operation.
The ultimate objective in magnetic storage devices
is to run two surfaces in contact (with practically zero
physical separation) if the tribological issues can be
resolved. Smooth surfaces lead to an increase in
adhesion, friction, and interface temperatures, and
closer flying heights lead to occasional rubbing of
high asperities and increased wear (Bhushan 1999b).
Friction and wear issues are resolved by an appro-
priate selection of interface materials and lubricants,
by controlling the dynamics of the head and medium,
and by the environment Bhushan (1996a, 1999a,
2000). A fundamental understanding of the tribology
(friction, wear, and lubrication) of the magnetic head/
medium interface, both on macro- and micro/nano-
scales, becomes crucial for the continued growth of
the more than US$80 billion a year magnetic storage
industry.
In this article magnetic storage devices and com-
ponents are introduced first. Results for selected tri-
bological problems important for the reliability of
magnetic storage device s are then presented. Fi nally,
results are presented of micro/nano tribological stu-
dies conducted using atomic force/f riction force
microscopy to better understand the fundamental
mechanisms of friction and wear.
1. Magnetic Storage Devices and Components
Schematics of tape and rigid disk drives and the con-
struction and materials used in magnetic heads and
media are shown in Figs. 2 and 3. In a linear data
processing IBM 3490/STK4490 tape drive as shown
in Fig. 2(a), after loading the cartridge into the drive,
Figure 1
Magnetic head slider flying over a disk surface
compared with an aircraft flying over ground with a
close physical spacing.
548
Magnetic Reco rding Devices: Head/Medium Interface
Figure 2
Schematics of (a) a tape path in an IBM 3490 data processing tape drive, (b) an inductive/MR thin-film head (with a
radius of cylindrical contour of 20 mm), and (c) sectional views of metal-particle (MP) and metal-evaporated (ME)
magnetic tapes.
549
Magnetic Reco rding Devices: Head/Medium Interface