APPLIED PHYSICS LETTERS VOLUME 85, NUMBER 24 13 DECEMBER 2004

Zigzag-shaped magnetic sensors

F. C. S. da Silva
National Institute of Standards and Technology, Boulder, Colorado 80305
W. C. Uhlig
National Institute of Standards and Technology, Gaithersburg, Maryland 20899
A. B. Kos, S. Schima, and J. Aumentado
National Institute of Standards and Technology, Boulder, Colorado 80305
J. Unguris
National Institute of Standards and Technology, Gaithersburg, Maryland 20899
D. P. Pappasa)
National Institute of Standards and Technology, Boulder, Colorado 80305
(Received 18 May 2004; accepted 20 October 2004)
Magnetism in zigzag-shaped thin-film elements is investigated using scanning electron microscopy
with polarization analysis, magnetotransport measurements, and micromagnetic simulations. We
find that the angle of magnetization alternates along the length of the element, and is strongly
correlated to the corrugated edges. We show that this simple and unique geometry can be used as a
single-axis magnetic field sensor. In this configuration, the sensors are primarily sensitive to fields
parallel to the applied current. Our results can be interpreted in terms of a coherent rotation model
of the magnetization. These devices are scalable to nanometer dimensions.
[DOI: 10.1063/1.1834732]

Recently, images of nanometer-scale magnetic structures angle of the current flow is set at a = 45° and H is applied at
suggested that the magnetization follows the contour of the an angle f = 90° with respect to the uniaxial easy axis of the
edges.1 An important extension of this work on 250 nm wide film. The angle a is obtained with equally spaced electrical
elements with serrated edges (zigzag shaped)2 was demon- contacts at a 45° angle with respect to the magnetic easy axis
strated using Lorentz microscopy. In these elements, the of the element. In this configuration (a = 45° and f = 90°), the
magnetization was shown to exhibit an alternating pattern of magnetization rotates coherently with the applied field (that
easy axis orientations. To have control of the magnetization is, cos u ~ H) and, according to Eq. (1), the response of R
in this manner is of great importance in the development of with respect to H (the so-called transfer curve) will be linear
scalable magnetoresistive (MR) sensors because more com- around zero field. To obtain a uniaxial magnetic element, it is
plicated current- and field-biasing schemes are not required, common to choose a soft magnetic material and to use both
and nanoscale devices based on this technology have become shape and induced anisotropies to ensure the existence of
increasingly important. For example, these types of devices only one anisotropy axis. Therefore, elongated shapes are
have been instrumental in the data storage revolution that has typically used, with attention to the geometrical shape of the
occurred over the past decade.3 Similar MR devices have ends to avoid magnetization curl and vortices.11
also been used as high sensitivity sensors for current-4–6 and Therefore, in this current work, we fabricated single-
magnetization-induced7,8 field imaging. A critical aspect of layer, elongated devices that have a zigzag-shaped magnetic
the design of these devices is the biasing of the magnetiza- layer. The zigzag has a similar aspect ratios to those of Ref.
tion and current flow in order to obtain an antisymmetric, 2, as shown in Fig. 1. The structures were fabricated using
linear response. In this letter, we demonstrate a method of optical lithography and a lift-off process for both the mag-
maintaining a bias between the magnetization and the current netic and contact layers. A 30-nm-thick magnetic layer was
direction using the contour of the element.
In a single-film anisotropic magnetoresistive sensor, the
resistance can be described by9

R = Ro + DR cos2su − ad, s1d

where Ro and DR are material-dependent constants (DR / Ro
= 2 % −4% for Permalloy™), and u and a are, respectively,
the angles formed by the magnetization and the applied cur-
rent directions with respect to the easy axis of the material.
To obtain the desired response of the device versus the ex-
ternal magnetic field H, several schemes to bias the magne-
tization relative to the direction of the current have been
proposed. For example, in the barber-pole method,10 the FIG. 1. Diagram showing the dimensions and electrical contacts of the
zigzag structure. They are parameterized using a rectangular building block
with lateral dimensions 2Î2 mm 3 4Î2 mm. An eight-block zigzag pattern
a)
Electronic mail: david.pappas@boulder.nist.gov was fabricated. The structure is 35 µm long and 5 µm wide.

0003-6951/2004/85(24)/6022/3/$22.00 6022
Downloaded 09 Dec 2004 to 132.163.130.151. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
Appl. Phys. Lett., Vol. 85, No. 24, 13 December 2004 da Silva et al. 6023

FIG. 2. (Color) (a) Experimental SEMPA image of the eight-block zigzag

structure. (b) Simulation performed on the same geometry using OOMMF.
The magnetization direction maps onto the angle color map ring.

sputtered from a Ni80Fe20 target onto a SiO2-coated Si wafer

with an external field of 15 kA m−1 applied along the long
axis of the structure. Two 5-nm-thick Ta layers were depos-
ited before and after the Permalloy™ layer. The Ta serves to
texture and protect the magnetic layer. The Al contacts at the
ends were electron-beam evaporated onto the magnetic ele-
ment after a brief ion mill to remove the top Ta layer in that
region.
Finite element analysis of the current flow in these de-
vices shows that the current will primarily flow down the
center of the element along the x̂ axis. The magnetic state in FIG. 3. Magnetization line scan from the SEMPA data and OOMMF simula-
zero magnetic field was imaged using scanning electron mi- tion along (a) x̂ and (b) ŷ directions. The uncertainty in the SEMPA data is
croscopy with polarization analysis (SEMPA). SEMPA is a 5°, as seen by scatter in the data.
noninvasive, high spatial resolution s,15 nmd technique that
measures the magnetization direction by analyzing the spin-
polarization of the secondary electrons emitted from the zation within each parallelogram alternates between ±25°, in
sample in a scanning electron microscope.12,13 a zigzag fashion as shown in the scan across the center of the
The micromagnetic simulations were performed using element in the x̂ direction [Fig. 3(a)]. This remanent configu-
the Object Oriented MicroMagnetic Framework ration indicates a minimization of the dipolar energy between
(OOMMF).14–16 OOMMF is a public domain program for mi- contiguous regions and the existence of an effective aniso-
cromagnetics that has been developed at NIST. Micromag- tropy axis along the length of the zigzag pattern (x̂ axis). In
netics is a continuum model of three-dimensional vector fact, the average aspect ratio of the structure together with
magnetization in ferromagnetic materials. The simulations in the magnetocrystalline anisotropy texture induced during
this letter compute equilibrium magnetization configurations sample fabrication can be responsible for this effective an-
that correspond to minima in the energy landscape. The en- isotropy.
ergy here consists of magnetostatic, exchange, and magneto- Line scans perpendicular to the axis, along the ŷ axis
crystalline anisotropy components. The electrical transport [Fig. 3(b)], show the magnetization profile in a domain wall
measurements were obtained using a four-probe technique and at the centers of two contiguous parallelograms. The
with the external magnetic field applied in the plane of the domain wall magnetization vector is relatively constant
film (Fig. 1).
along the ŷ direction. However, in the parallelograms the
Figure 2(a) shows a SEMPA image of the magnetization
angle of magnetization decays from ±45° at the edges to
of the zigzag structure for the remanent state obtained after
about ±25° in the middle of the structure. In order to model
the sample was saturated in the x̂ direction. Along the axis,
this numerically, the SEM topographic image of the structure
we find parallelogram-shaped regions where the magnetiza-
tion alternates direction along the axis, with domain walls was used as a template for a magnetization simulation with
oommf, shown in Fig. 2(b). The simulation was obtained us-
spanning the element from each pair of opposing acute and
obtuse angles. This pattern matches that observed in Ref. 2 ing a two-dimensional model with cell size of 10 nm
on nanoscale elements, demonstrating the scaling of this con- 3 10 nm and 30 nm thickness. Typical magnetic parameters
figuration. We also observe vortex structures near the ends of for Permalloy™ films were used. Specifically, the magneti-
our devices. The presence of vortices depends on the satura- zation was set to 8003 103 A / m, the exchange interaction
tion history of the sample. We find that they can be elimi- was set to 133 10−12 J / m, and the magnetocrystalline aniso-
nated with a strong enough saturation field. However, the tropy was varied from 0 to 1200 J / m3. We found that the
magnetization in the parallelograms tends to return, revers- remanent state magnetization was relatively insensitive to the
ibly, back to the zero field values after saturation. This is value of the anisotropy. The simulated data agree both quali-
presumably because the domain walls are pinned at the cor- tatively and quantitatively with the SEMPA data, as shown
ners of the structure. The average direction of the magneti- by the solid lines in Fig. 3.
Downloaded 09 Dec 2004 to 132.163.130.151. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
6024 Appl. Phys. Lett., Vol. 85, No. 24, 13 December 2004 da Silva et al.

x̂ directions, these sensors would have a rejection ratio r

i
; So / S'
o , for H y greater than 10. In addition, the inset Fig.
4(a) shows that Ro is the same for both branches of the trans-
fer curve, but the derivative with respect to H has different
signs depending on which branch of the transfer curve the
system is prepared.17 This memory effect, combined with the
fact that the device switches without forming minor loops, is
scalable, and can be probed with very little power, shows
that this device could also be useful as a cell in a magnetic
random access memory device.
These results show that it is possible to naturally bias the
magnetization in MR sensors by choosing the appropriate
contour of the magnetic element. In the particular case of a
single-layer MR sensor with a zigzag shape, the geometrical
biasing mechanism works over a wide range of size scales,
from the nanoscale region (width 250 nm) to the microscale
(width 5 µm), where this study was conducted. This scaling
should extend to small sizes until the distance between cor-
ners becomes comparable to the domain wall width in the
magnetic material. We also find that modern computational
models, such as oommf, can be used to understand and accu-
FIG. 4. Transfer curves of the eight-block zigzag structure measured for rately predict the magnetic properties of devices at the
applied field in (a) the x̂ direction and (b) the ŷ direction. The insets show nanoscale level. Based on the scalability and the fact that
both branches of the transfer curve at higher fields.
these elements can be described with a simple coherent ro-
tation model, we expect this type of nanoengineering of the
The magnetoresistance of the structures was then mea- shape of the magnetic layer will have a significant impact in
sured at room temperature. The contacts, shown in Fig. 1, the areas of magnetic field sensors and memory applications.
were designed to cover the ends of the element in order to
eliminate the effects of the vortices. Figure 4 shows the ex- We acknowledge support from NATO grant
perimental transfer curves of the structure measured for two PST.CLG.979374. In addition, we would like to thank Alex-
orientations of the applied field H. The dependence of the ander P. Popov, John N. Chapman, and Robert McMichael
voltage V across the structure on H was measured for an for useful discussions. Contribution of the U.S. government,
applied current I = 3 mA. In the remanent state the zero-field not subject to copyright.
voltage across the structure Vo = 0.446 V corresponds to a
1
resistance Ro = 149 V. M. Hermann, S. McVitie, and J. N. Chapman, J. Appl. Phys. 87, 2994
For fields applied in the x̂ direction an antisymmetric (2000).
2
M. Hermann, PhD thesis, University of Glasgow, 2000.
transfer curve is observed. Figure 4(a) shows one branch of 3
J. C. Mallinson, Magneto-resistive and Spin-valve Heads, 2nd ed. (Aca-
the hysteresis curve at low field. Following this branch down demic, San Diego, 2002).
from the positive saturated state, the voltage across the struc- 4
P. I. Nikitin, S. I. Kasatkin, A. M. Muravjov, P. M. Vetoshko, M. V.
ture decreases monotonically and reversibly to 0.45% of its Valeiko, V. I. Konov, and T. Meydan, Sens. Actuators, A 106, 26 (2003).
5
saturation value as the magnetization angles increase relative S. Bae, K. Scheiman, W. Mertin, E. Kubalek, and M. Maywald, Micro-
electron. Reliab. 39, 975 (1999).
to the current direction. This is followed by an irreversible 6
B. D. Schrag and G. Xiao, Appl. Phys. Lett. 82, 3272 (2003).
jump at uHu < 1 kA m−1, where the domain walls change di- 7
S. Tumański and A. Liszka, J. Magn. Magn. Mater. 242–245, 1253
rection. For fields less than this value, we find that the trans- (2002).
8
fer curves are completely reversible, with no evidence of P. L. Rossiter, The Electrical Resistivity of Metals and Alloys (Cambridge
minor loops even after many cycles. After the jump, the volt- University Press, Cambridge, UK, 1987).
9
age follows the opposite branch in the same fashion (see See, for example, U. Dibbern, in Sensors a Comprehensive Survey, edited
by R. Boll and K. J. Overshott (VCH Publishers, Inc., Weinheim, 1989),
inset). The average sensitivity of the structure for a field Chap. 9.
range DH = 2 kA m−1 around zero is defined as So 10
S. Tumański and M Stabrowski, Sens. Actuators 7, 285 (1985).
= sDV / DHds1 / Vod.9 Therefore, for the parallel field configu- 11
K. J. Kirk, J. N. Chapman, and C. D. W. Wilkinson, Appl. Phys. Lett. 71,
i
ration, So = 3.54 mV/ sV kA m−1d. 12
539 (1997).
M. R. Scheinfein, J. Unguris, M. H. Kelley, D. T. Pierce, and R. J. Celotta,
For fields applied in the ŷ direction, the transfer curve is
Rev. Sci. Instrum. 61, 2501 (1990).
comprised of two nearly symmetric flat branches near zero 13
J. Unguris, Magnetic Imaging and its Applications to Materials, edited by
field. In the reversible region, shown in Fig. 4(b), where M. de Graef and Y. Zhu (Academic, Burlington, MA, 2000), pp. 167–303.
uHu ø 1 kA m−1, the voltage changes by only 0.05% of the 14
M. J. Donahue and D. J. Porter, OOMMF’s User Guide, Version 1.0,
remanent value. This corresponds to a sensitivity S' o
NISTIR 6376 (National Institute of Standards and Technology, Gaithers-
= 0.31 mV/ sV kA m−1d for a field range of DH = 2 kA m−1 burg, MD, 1999). For more information, see website at http://
math.nist.gov/oommf/
around zero. 15
W. F. Brown, Jr., Micromagnetics (Interscience, New York, 1963).
Therefore, these devices make highly directional low- 16
J. Fidler and T. Schrefl, J. Phys. D 33, R135 (2000).
17
field sensors. Using the sensitivity values for H in the ŷ and J. Aumentado and V. Chandrasekhar, Appl. Phys. Lett. 74, 1898 (1999).

Downloaded 09 Dec 2004 to 132.163.130.151. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp