16 Device Applications Using Spin Dependent

Tunneling and Nanostructured Materials

Mark Tondra, Dexin Wang, and Zhenghong Qian

NVE Corporation, 11409 Valley View Road, Eden Prairie, MN 55344, USA

Abstract. This paper describes recent developments in magnetoresistive materials

and their applications. Also, it will focus on a few magnetoresistive materials and ap-
plications that are under active development. Two applications are chosen, in part,
due to their connection to nano-scale structures. These are 1) magnetoresistive iso-
lators whose speed will be enhanced by nano-crystalline ferromagnetic layers, and 2)
detection and manipulation of nano-magnetic particles using magnetoresistive sensors
and integrated straps.

16.1 Introduction
16.1.1 Scope
This paper is intended to be a useful reference for a graduate student who is
looking for information about possible research projects in the field of nanos-
tructured magnetic materials and their applications and in understanding the
broader technological implications of their work. More experienced researchers
may also find it useful as a presentation of a few particular applications that are
relatively new in the field of magnetoresistive sensors. No attempt will be made
to cover all applications of magnetoresistive devices. Rather, the focus will be
on materials and devices that are under active development at NVE. Two ma-
jor applications will also not be covered, though they are under development at
NVE. Specifically, read heads for disk drives are discussed in other papers from
the workshop, and Magnetoresistive Random Access Memory (MRAM) requires
a separate discussion altogether.
NVE is actively working on two kinds of magnetoresistive materials, Giant
Magnetoresistive (GMR) and Spin Dependent Tunneling (SDT). Products using
GMR are available, while those using SDT are still under development. Products
using GMR include simple resistor bridge field sensors and gradiometers, and
digital signal isolators.
Discussion will be limited to magnetoresistive effects related to thin ferro-
magnetic films of Ni, Fe, Co, and alloys. Specifically excluded are Hall effect
devices and Colossal Magnetoresistance effect devices.

16.1.2 Why Magnetoresistance?

Magnetoresistance is the change in the electrical resistance of materials as a
function of an externally applied magnetic field. There are many forms of mag-
netoresistance, but the most useful ones are observed in ferromagnetic thin films,

D. Shi et al. (Eds.): LNP 593, pp. 278–289, 2002.

c Springer-Verlag Berlin Heidelberg 2002


 16 Device Applications Using SDT 279

usually alloys of Ni, Fe, and Co. These effects are useful because magnetoresis-
tive thin films can be manufacturable in an integrated circuit process. They can
be deposited on semiconductor wafers (with or without integrated circuits) and
patterned into many forms of devices using standard photolithography and etch-
ing techniques. Resulting magnetoresistive products, then, benefit from the same
commercial forces that have made integrated circuits so valuable and successful.
They are extremely small, very low cost, and are easily combined with many
electronic functions.
The easy combination with electronics is the key to what makes magnetore-
sistive thin films so important. The same underlying technology can be applied
to many different applications. Examples are: generic linear magnetic field sen-
sors; special function magnetic field sensors; read heads for computer disk drives,
magnetic tapes, credit cards, currency, etc.; current sensing, digital data trans-
mission isolation, and digital data storage. There are many details that must be
addressed in order to have a successful device in any of these categories. The
size of the magnetotransducer will vary, as will the circuit complexity, packaging,
power requirements, cost, etc. But the basic magnetoresistive effect is common
to all of them.

16.1.3 Physics Background

AMR. The understanding of the physics of magnetoresistance in ferromagnetic
materials goes back as far as 1857 when the Anisotropic Magnetoresistance
(AMR) effect was first discovered in macroscopic ferromagnetic wires [1]. The
AMR effect is observed as a relationship between the resistance of the material
and the angle between the magnetization vector and the electrical current. The
resistance is lowest when the current and magnetization are perpendicular and
maximum when they are parallel/antiparallel. This effect is observable in bulk
materials and in thin films. The magnitiude of the effect can be as large as 8% at
room temperature. However, the AMR effect in the most commonly used mate-
rials is between 2 % and 4 %. The main tradeoffs involved are the desire to have
low saturation fields, low hysteresis, low magnetostriction, and high linearity;
while still maintaining high magnetoresistance. The most popular material for
this compromise is the Ni65 Fe15 Co20 (atomic percentage).
The physical origins of the AMR effect are “spin-orbit” interactions, and the
imbalance of spin-up and spin-down conduction (Fermi energy) electrons. The
spin orbit interactions cause different directional scattering of spin up and spin
down electrons. The net spin imbalance at the Fermi surface creates a net im-
balance of the scattered charge. These two effects together result in a resistivity
that varies with the orientation of magnetization with respect to electrical cur-
rent. Detailed explanations of observed resistivity variations have been largely
worked out by Smit [2] and others.
GMR. The development of thin film deposition technology led to AMR sen-
sors being the dominant technology for the read heads in hard disk drives. This
dominance remained until the mid 1990’s when Giant Magnetoresistive (GMR)
devices took over. The GMR effect is fundamentally a thin film phenomenon. It
280 M. Tondra, D. Wang, and Z. Qian

is empirically observed to be a change in resistance as a function of the angle

between the magnetizations of adjacent thin ferromagnetic thin films. Consider
the simplest case of two thin NiFeCo ferromagnetic films separated by a thin
Cu “spacer” layer (this structure is referred to as a “sandwich”). These films
must not be in direct contact with each other or else their magnetizations will
be directly coupled, and will not be able to rotate with respect to each other.
However, they may not be separated by too much distance, or the electrons
flowing in the sandwich will not be able to pass from one ferromagnetic layer
to the other without loosing their “spin information”. Typical dimensions for
such a GMR sandwich are NiFeCo 5 / Cu 3.5 / NiFeCo 5 (thicknesses in nm).
The thicknesses must be small compared to the effective scattering lengths of
conduction electrons. The GMR effect can be observed with current flowing ei-
ther in the plane of the thin films or perpendicular to it. Practically speaking,
however, devices are made using the Current In the Plane (CIP) configuration
because the Current Perpendicular to the Plane (CPP) configuration has a very
low resistance unless the lateral dimensions can be made extremely small (∼10
nm).
From a developmental standpoint, the discovery of GMR was not possible be-
fore vacuum deposition technology advanced to the point where layer thicknesses
and roughnesses could be controlled to sub-nanometer dimensions. The usually
cited discoverers of GMR were pushing on this technology in their search for
magnetoresistive effects [3] and exchange coupling in magnetic multilayers [4].
Since these discoveries in the late 1980’s, GMR has rapidly evolved to surpass
AMR as the dominant magnetoresistive effect used in hard drive read heads in
the mid 1990’s. The basic read head structure is called a “spin valve.” It is a
sandwich as described above with one of the two magnetic layers “pinned” by an
antiferromagnetic layer such as FeMn, or CrPtMn. A simple example of a spin
valve structure is NiFeCo 5 / Cu 3.5 / CoFe 5 / CrPtMn 30.
The thin layer structure is critical because the physical effect that is being
exploited is that the spin up and spin down electrons in the sandwich have
different mean free paths that depend on the orientation of the magnetizations.
These mean free paths are on the order of 10 nm.
The magnetic operation of a spin valve is such that the resistance is maxi-
mized when the magnetic layers are antiparallel and minimized when they are
parallel.
The total magnitude of the spin valve GMR effect is 5 % to 25 % depending
upon layer thickness, quality, and configuration.
SDT. Spin Dependent Tunneling (SDT) devices are again related to the rel-
ative orientations of two adjacent ferromagnetic films. However, the tunneling
effect takes place when electrons pass across a nominally insulating barrier sep-
arating the two magnetic layers. Thus it is strictly a CPP effect. The first useful
SDT devices were demonstrated in 1995 [5,6], and the first one using lithograph-
ical fabrication techniques in 1996 [7]. The tunnel barrier must be extremely
thin in order to have a measurable tunneling current. A typical SDT structure is
NiFeCo 12.5 / Al2 O3 1.5 / CoFe 50 / CrPtMn 30. Here, the Al2 O3 tunnel barrier
 16 Device Applications Using SDT 281

Fig. 16.1. The resistance of a spin valve as a function of the angle between magneti-
zations of top and bottom magnetic layers. The high resistance values shown are more
likely to be observed in a tunnel junction. The typical sheet resistance of a spin valve
is about 25 ohms/square, so this data would be representative of a GMR spin valve
with 2000. I.e. its length to width ratio is 2000.

has replaced the Cu as the separating film in the spin valve. Magnetically, the
operation of a pinned SDT device is identical to that of a GMR spin valve. The
magnetoresistance, however, is larger. Total resistance changes on the order of
45 % are common.
SDT construction. Tunnel junction devices are more difficult to fabricate
than GMR devices. Two main challenges are the creation of the very thin tunnel
barrier with no pinholes, and the subsequent formation of usable structures out
of the initial blank tunnel sandwich layers.
At NVE, the SDT devices are made in a Perkin Elmer 2400 with a baseline
pressure in the 10−8 torr range. The layers are deposited using RF diode sput-
tering in an Ar plasma using an Ar pressure of 10’s of mTorr. The Al2 O3 barrier
is formed by first depositing a 1.2 nm layer of Al, and subsequently oxidizing
it by introducing some oxygen into the chamber while a plasma is still present
over the sample wafer. As it oxidizes, the Al film expands about 30 % to its final
thickness of about 1.5 nm. One or more annealing steps are required to optimize
the SDT devices’ resistive and magnetic properties.
Patterning of the junctions takes place in two steps. The first step is to
etch away portions of the top electrode material while leaving the entire bottom
electrode intact. The second step is to etch away undesired bottom electrode
material. Both etches are performed using an ion mill for material removal, and
a photolithographically defined etch mask. The finished result is a stack where
the edges of top electrodes are entirely within the edges of the bottom electrode.
The bottom electrode can support one or two tunnel junctions depending upon
the intended finished structure.
The two junction SDT structure allows a tighter packing of junctions in a
given area, and makes fabrication slightly easier as all connections to junctions
are to the top electrodes.
282 M. Tondra, D. Wang, and Z. Qian

Fig. 16.2. SDT material stack and magnetic biasing mode.

16.2 Applications

16.2.1 Generic Magnetic Field Sensor

In order for the junctions to be useful as magnetic field sensors, more work is
required. Such details are addressed here in the context of the SDT sensors.
However, similar technology is also applied to GMR and AMR sensors.
Bridge. Typically, a resistance bridge is desirable rather than a single resis-
tor or string of resistors. A bridge allows for null readings, helps compensate for
temperature effects, and makes amplifier design much easier. A typical bridge
starts with four identical resistors on each of the four bridge legs. For the re-
sistance bridge to do its job, however, features must be added to make some
legs behave differently than others. If this is not done, all the legs’ resistances
will increase and decrease at the same rate, and no net output will be observed.
The tool NVE uses most commonly to achieve a non-zero bridge output are flux
concentrators.
Field biasing. The SDT device response must be linearized. That is, the
“square loop” output of an unbiased device must be turned into a linear, non-
hysteretic output. The square loop shape occurs when the free layer magneti-
zation of the SDT junction flips back and forth to be parallel and antiparallel
to the pinned layer. This output becomes much smoother when a perpendicular
bias field is applied so that the soft layer magnetization rotates rather than flips.
The magnitude of the perpendicular bias field is just over the anisotropy field
of the soft layer. This is about 18 Oe in NiFeCo, and less than 5 Oe in regular
permalloy (Ni80 Fe20 ). In the finished devices, the perpendicular field is generated
by an on-chip planar coil. A second coil is also provided as a sensitive axis bias.
The difference in biased and unbiased SDT response can be seen by comparing
Figs. 16.3 and 16.4.
Flux concentrators. These are usually 15 µm thick plated NiFe. Two flux
concentrators on a sensor chip are arranged so that they are separated by a long
narrow “gap”. The sensitive axis of the sensor is parallel to the short dimension
of the gap. Two opposite resistor legs (gap legs) are placed in the gap, while
the other two are placed underneath the flux concentrators (shield legs). The
field in the gap is “amplified” by the flux concentrators such that the gap field
is [flux concentrator length] / [gap] x [external field]. The field under the flux
 16 Device Applications Using SDT 283

Fig. 16.3. Unbiased SDT response. The change in resistance occurs when the mag-
netization of the free layer flips 180 degrees. The left side of the plot shows higher
resistance where the two magnetizations are antiparallel. The lower resistance on the
right side is where the magnetizations are parallel. This tunnel junction is 120 x 320
microns and has a resistance of about 50 Ohms.

concentrators reduced by a similar fraction. The net result is that the gap legs
see a much greater field than the shield legs, and a net bridge offset is realized.
The flux concentrators also have the effect of attenuating off-axis fields, so that
the sensor becomes nearly a vector sensor (outputting the magnitude of the
sense-axis field).
Sensor output. Once all of the pieces are together, the sensor output can be
observed. Typical sensor output is shown in Fig. 16.5.

16.2.2 Isolators
Digital signal isolators are a relatively new application for magnetoresistive ma-
terials. This section describes how they are used, and how adapting a nano-
crystalline magnetic material to the SDT stack will enhance the performance of
future versions of the magnetoresistive isolator.
The purpose of signal isolators is to allow information to pass from one
electrical circuit to another without having a direct conduction path between
the two. This helps reduce noise in data transmission, and protects sensitive
284 M. Tondra, D. Wang, and Z. Qian

Fig. 16.4. When the SDT device has a proper magnetic bias, the magnetization of the
free layer rotates smoothly.

Fig. 16.5. Actual SDT sensor output vs. magnetic field. This bridge is being supplied
with 2 Volts, so the output range of 200 mV represents about 10% of the total supply
voltage. The most sensitive part of the output is shifted about 1 Oe due to coupling
between the pinned and free layers. The shift in voltage (vertical) is due to imperfectly
matched SDT resistor legs.

circuitry from potentially damaging voltages. In practice, it is challenging to

do this data transmission at a high rate of speed without propagation delays.
Isolators are used to protect the integrity of both digital and analog data. This
paper will focus on the digital applications because that is what NVE has on
the market.
As a way of describing basic isolator functions while also explaining why
magnetorestive materials are so well suited for the application, the operation of
optical and magnetic based isolators will be contrasted.
 16 Device Applications Using SDT 285

Optical. The dominant form of low cost integrated digital signal isolator is
optically based. Current flowing into the input side of the isolation barrier goes
through a Light Emitting Diode (LED). The light from this LED is directed
across the isolation barrier to a semiconductor photodetector. The photodetector
registers a signal once sufficient photons have been collected from the beam from
the LED.
Magnetoresistive. The newest and fastest form of integrated digital signal
isolator is based on integrated GMR devices. Current flowing into the input
side of the isolation barrier goes through an on-chip planar coil that generates
a magnetic field. The coil is situated so that the GMR spin valve devices are
on the other side of the isolation barrier but in the field generated by the coil.
The GMR is configured in a bridge so that the voltage out switches high or
low depending upon which direction the coil current is flowing. This contrast is
shown below in Figure 16.6.
Delays. At first, it is not obvious why the GMR isolator is faster than the
optical isolator. Photons travelling from the LED to the photodetector should
be just as fast as the magnetic field propogation from the coil to the GMR
detector. The difference, though, is in the generation and detection of the signal
rather than its propogation. The LED requires some time, typically 10’s of ns,
to generate enough photons to drive the photodetector beyond the switching
threshhold of the detection electronics. This speed can be increased in proportion
to the amount of current passing through the LED, but is limited by practical
power requirements. The current in the GMR isolator, however, generates a
magnetic field instantaneaously. The speed limit for the GMR, then, is how fast
the magnetic sensing film’s magnetization can rotate in the applied field. Thus,
the underlying physics questions that must be asked are: 1) how fast can the
sensing layer rotate, and 2) how much current (how large a field) is needed to
cause rotation. Mathematically, this boils down to solving the Gilbert equation.
Empirically, it is found that for reasonable amounts of current, the switching
takes place in 1 ns or so.

Fig. 16.6. Contrasting GMR isolator with optical isolator concepts.

286 M. Tondra, D. Wang, and Z. Qian

Since the fundamental limitation for the GMR isolator switching time is
related to the dynamics of the soft film’s magnetization, research has been done
to develop soft films with higher “speed limits”. This is where the nanocrystalline
ferromagnetic material becomes relevant.

16.2.3 Nanocrystalline Ferromagnetic Devices

CoFeHfO has been developed for use in high frequency transformers. Its high
anisotropy and saturation magnetization give it a very high ferromagnetic res-
onance frequency (FMR) of about 2.8 GHz, while its high and adjustable re-
sistivity reduces the energy lost due to eddy currents. It’s properties enable a
projection of GMR digital signal isolator operation above 100 GHz, assuming the
associated integrated circuitry can keep up. The magnetoresistive and magnetic
properties of CoFeHfO are shown below in Figs. 16.7 and 16.8.
The resisitivity has been adjusted over a range from 50 to 2000 micro-Ohm-
cm. The saturation magnetization is relatively constant over this range. By mak-
ing the resistivity as high as possible, one reduces the energy losses due to eddy
currents in the films. These induced eddy currents are a main limitation for high
frequency transformer applications.

Fig. 16.7. Magnetoresistance of an SDT device with CoFeHfO as a free layer.

Fig. 16.8. Magnetization vs. field of a CoFeHfO film. The two curves are hard and
easy axis magnetization loops.
 16 Device Applications Using SDT 287

16.2.4 Magnetic Nano-particle Detection

Magnetoresistive sensors can be used to detect magnetic nanoparticles. Such

nanoparticles find uses in a wide range of applications. On one end of the spec-
trum is as the magnetic material in ferrofluids. On the other is as “labels” in
biochemical assays. In between are other useful methods of sorting, heating and
cooling, sealing, etc.
This section will focus on the use of GMR sensors as the detection scheme
in biological assays. These assays become magnetic in nature when the “label”
scheme uses nano-magnetic particles. Labeling is the term used for the attach-
ment of a detectable object to the biochemical species of interest. As an example,
a magnetic particle treated with a certain protein becomes “sticky” to certain
other proteins but not others. This is because their physiochemical structure
is designed to allow these two proteins to bind specifically to each other, but
not to other proteins. Such “lock and key” interactions are omnipresent in the
biological world. Perhaps the best known example is the DNA chain, where the
left and right halves of the rungs of the “ladder” only connect if they match.
Other labels are more common. The most common are light-based labels. They
emit light due to excitation from a laser, a chemical interaction, or simply reflec-
tion. Optical detection, however, is not easy or convenient. Bulky and expensive
lenses, mirrors, and other gear are required to achieve high degrees of resolu-
tion. Detection of magnetic labels, however, can be done with fully integrated
magnetoresistive detectors. That is the basis of interest in this technique.
BARC. It is not the purpose of this paper to explain in detail how biochemical
assays work. Rather, it is assumed that such assays generate samples where the
number and / or concentration of nano-magnetic particles is an important value
to measure. One such assay coats the surface of a GMR chip with an array of dif-
ferent biochemically active materials. Some of these are “sticky” to the specially
prepared magnetic nano-particles while others are less so. The areas and degree
of magnetic particle adhesion is then a quantifiable assay result. The question is
then how to detect and quantify these particles. This idea came from the Naval
Research Laboratory (NRL) which has developed the technology into a Bead AR-
ray Counter (BARC) using custom GMR sensor arrays fabricated by NVE [8].
Magnetic Design. There are several magnetic modes that these assays can use.
The BARC uses an out-of-plane field to magnetize the particles for steady-state
detection. It is also possible to use an in-plane field. In all cases, the sensitive
axis of the GMR detector is in the plane. Nominally, the magnetic nano-particles
are paramagnetic or superparamagnetic, so their magnetizations and resulting
stray fields are linearly proportional to the applied field. The stray fields from
the magnetized particle are dipolar in shape. The field that the GMR sensor
actually sees from the particle are orthogonal to or in the opposite direction
from the applied field. So the total field at the GMR sensor is less than the
applied field by an amount determined by the size and separation from magnetic
particles. This difference in fields can be detected by comparing the output of
the sensing resistor with a reference resistor known to have no particles above
it. Signal to noise ratio is enhanced greatly by using a time varying field.
288 M. Tondra, D. Wang, and Z. Qian

16.3 Conclusions
Magnetoresistive technology is being used in many important and interesting
applications. This technology does best when the devices are compatible with
high volume semiconductor-type manufacturing. This manufacturing method
provides a way to make enormous numbers of sensors at a very low price per
part.
The most ubiquitous sensors are the GMR read heads in hard disk drives.
Many other sensing applications are possible depending upon the technical and
commercial constraints. The most common sensing applications are in automated
manufacturing and automotive environments where the position and speed of
many components must be monitored. Ultra high performance magnetoresistive
digital logic may become a commercial reality in the form of MRAM, currently
under development by NVE, Motorola, Honeywell, IBM, and others. NVE has
developed magnetoresistive isolators that will ultimately replace many optical
isolators.
The current technological thrust in magnetoresistive devices is toward smaller,
cheaper, and faster items. Research in nanomagnetics, then, is a clear enabler for
these trends. Of special interest are making magnetic devices with stable prop-
erties at smaller sizes. Also, improving the properties of nanomagnetic particles
(both in magnetic and dimensional terms) will enable many new applications in
biochemical sensing and detection.

Acknowledgements
The authors aknowledge the many people at NVE who have participated in
the development efforts described here. Jim Daughton, the founder of NVE,
has provided not only the ideas behind a lot of the development, but also the
energy and drive to bring ideas to reality. John Taylor and Seraphin Akou have
done excellent work testing devices. Erik Lange and Dan Reed have helped with
electronics development. Cathy Nordman, John Anderson, Bob Sinclair, Carl
Smith, and Bob Schnieder have all worked towards improving the underlying
technology through various research efforts. And, we must thank our sponsors
for their strong and continued financial support. They include NASA, Army,
Navy, Air Force, DARPA, NSF, and DOC.

References
1. W. Thomson, “On the electro-dynamic qualities of metals; Effects of magnetization
on the electric conductivity of nickel and iron,” Proc. Roy. Soc., 8, pp. 546-550,
(1857)
2. J. Smit, “Magnetoresistance of ferromagnetic metals and alloys at low tempera-
tures,” Physica 16, pp. 612-617, (1951)
3. M.N. Baibich, J.M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Eti-
enne, G. Creuzet, A. Freiderich, and J. Chazelax: Giant magnetoresistance of
(001)Fe/(001)Cr magnetic superlattices, Phys. Rev. Lett. 61, pp. 2472-2475, (1988)
 16 Device Applications Using SDT 289

4. P. Grunberg, R. Schreiber, Y. Pang, M.B. Brodsky, and H. Sowers: Layered mag-

netic structures: Evidence for antiferromagnetic coupling of Fe layers across Cr
interlayers, Phys. Rev. Lett. 57, pp. 2442-2445, (1986)
5. J.S. Moodera, L. R. Kinder, T.M. Wong, and R. Meservey, Phys. Rev. Lett. 74, p.
3273, (1995)
6. T. Miyazaki and N. Tezuka, J. Magn. Magn. Mat. 139, L231 (1995)
7. R.S. Beech, J. Anderson, J. Daughton, B.A. Everitt, and D. Wang, IEEE Trans.
Magn. 32, p. 4713 (1996)
8. D. R. Baselt, G. U. Lee, M. Natesan, S. W. Metzger, Paul E. Sheehan, and R.
J. Colton, “A biosensor based on magneotoresistance technology,” Biosensors &
Bioelectronics, 13, 731-739 (1998)
9. Jacob Fraden, AIP Handbook of Modern Sensors: Physics Designs and Applica-
tions, (American Institute of Physics, New York, 1993), 243-262