Home Search Collections Journals About Contact us My IOPscience

A high-bandwidth spintronic position sensor

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 Nanotechnology 25 375501

(http://iopscience.iop.org/0957-4484/25/37/375501)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 134.151.40.2
This content was downloaded on 05/09/2014 at 05:51

Please note that terms and conditions apply.

 Nanotechnology

Nanotechnology 25 (2014) 375501 (10pp) doi:10.1088/0957-4484/25/37/375501

A high-bandwidth spintronic position sensor

Tomas Tuma1,2, Angeliki Pantazi1, Deepak R Sahoo1,3, Philipp Eib1,
Gian Salis1, Haris Pozidis1 and Abu Sebastian1
1
IBM Research—Zurich, Säumerstrasse 4, 8803 Rüschlikon, Switzerland
2
Automatic Control Laboratory, Swiss Federal Institute of Technology, Physikstrasse 3, 8092 Zurich,
Switzerland

E-mail: [email protected], [email protected], [email protected], [email protected],

[email protected] and [email protected]

Received 11 December 2013, revised 15 May 2014

Accepted for publication 16 May 2014
Published 22 August 2014

Abstract
Position sensing with resolution down to the scale of a single atom is of key importance in
nanoscale science and engineering. However, only optical-sensing methods are currently capable
of non-contact sensing at such resolution over a high bandwidth. Here, we report a new non-
contact, non-optical position-sensing concept based on detecting changes in a high-gradient
magnetic field of a microscale magnetic dipole by means of spintronic sensors. Experimental
measurements show a sensitivity of up to 40 Ω /μm, a linear range greater than 10 μm and a noise
floor of 0.5 pm/ Hz . Also shown is the use of the sensor for position measurements for closed-
loop control of a high-speed atomic force microscope with a frame rate of more than 1 frame/s.

Keywords: position sensing, spintronics, scanning probe microscopy

(Some figures may appear in colour only in the online journal)

1. Introduction such as the change in capacitance due to the relative motion

of two parallel plates (capacitive sensors), the change in the
Nanoscale position sensing is a key enabling technology for length of the thermal conduction pathway (electrothermal
nanoscale science and engineering. For position sensing at sensors), or the piezoelectric voltage arising from the posi-
high resolution and high bandwidth, optical-sensing techni- tional-change-induced strain (piezoelectric sensors). In most
ques are routinely used in applications, such as wafer posi- of these cases, however, the only way to significantly increase
tioners for semiconductor lithography [1, 2]. However, the
cost and form factor of optical-sensing techniques are prohi-
the sensitivity of transduction ( ) is by increasing the form
∂Sx
∂x

bitive for numerous other applications, such as scanning factor of the sensor. Moreover, the measurement of the
probe microscopy [3–5], data storage [6] and nanofabrica- transduced signal, Sx , is often limited in bandwidth.
tion [7]. In this paper, we introduce a new concept for non-con-
Conventional non-optical position sensors, such as tact, non-optical position sensing that is based on using
capacitive sensors, can reach moderately high bandwidths spintronic sensing elements in combination with a low-den-
with a nanometer-scale resolution, but have a relatively large sity, high-gradient magnetic field created by a micromagnetic
footprint and high cost [8]. Piezoresistive and piezoelectric dipole. First, we present the sensing concept along with a
sensors [9], on the other hand, are not amenable to non- semi-analytical model for magnetoresistance-based sensing
contact sensing. Alternative sensing techniques, such as linear under non-uniform magnetic fields. Next, a position sensor
variable displacement transformers (LVDT) and electro- fabricated using micromagnets and giant magnetoresistance
thermal sensors [10], have limited bandwidth. sensors is presented. We characterize the key sensor proper-
In all these position sensors, the position, x, is transduced ties, such as the sensitivity, range, resolution and bandwidth.
into a measurable change in an alternate physical quantity, Sx , Finally, we present experiments in which the position sensor
is used for position control and high-speed imaging in an
3
Currently with the HH Wills Physics Laboratory, University of Bristol, UK atomic force microscope.

0957-4484/14/375501+10$33.00 1 © 2014 IOP Publishing Ltd Printed in the UK

Nanotechnology 25 (2014) 375501 T Tuma et al

x
Direction of sensing
Magnetic field at the sensor

N 1 Diple axis

Dipole - sensor vertical distance

Magnetic 0
Magnetic 0
field dipole FLUX DENSITY (a.u.)
-1
Dipole axis -1 -0.5 0 0.5 1

S Dipole-sensor
2
Sensing current vertical distance
0

Magnetoresistive
sensor GRADIENT (a.u.) ∞
-4
-1 -0.5 0 0.5 1
x position (a.u.)
Ω
Electrical resistance measurement

Figure 1. Schematic illustration of the position-sensing concept. The position sensor consists of a magnetic dipole positioned close by an
MR-based magnetic field sensor. When the dipole is displaced relative to the MR sensor, the change in the magnetic flux density changes the
electrical resistance of the MR sensor. In the inset panel, the magnetic flux density and gradient in the proximity of the pole of the magnetic
dipole are shown. The gradient increases with increasing proximity to the pole, resulting in increased displacement sensitivity; however, the
flux density remains low, which prevents saturation of the MR sensor.

2. Sensing concept intrinsically high bandwidth, high sensitivity and small form
factor, GMR- and TMR-based sensors have been employed in
The proposed position sensing concept relies on the use of the magnetic recording and non-volatile memory technologies
magnetic field close to the pole of a microscale magnetic [15], for ultra-weak magnetic field sensing [16], measurement
dipole (see figure 1). For an ideal magnetic dipole with of micro-scale biological forces [17], position sensing in
moment m, the total magnetic flux density is maximum at the industrial applications [18] and atomic force microscopy
pole, but along the lateral direction, x, at a vertical distance z (AFM) [19, 20].
from the pole, the flux density is If such an MR sensor is placed in close proximity to the
μ 3xz magnetic dipole, then a change in the position of the dipole
Bx (x , z) = 0 × 2.5
× m. along the x direction will translate to a significant change in Bx
4π (
x 2 + z2 ) as seen by the MR sensor, resulting in a substantial change of
Hence, as the proximity to the pole increases ( x → 0 , the electrical resistance of the sensor (denoted by R). The
z → 0), the gradient ∂Bx (x, z )/∂x tends to infinity, while sensitivity of this position-sensing scheme in the x direction,
Bx (x, z ) approaches zero with an increasing proximity to the denoted by ∂R /∂x , is governed by the sensitivity of the MR
sensor to the magnetic field (denoted by ∂R /∂B ) and the
pole. This attribute of the magnetic dipole field can be used
magnetic field gradient ∂Bx /∂x . The sensing noise is deter-
for position sensing if the position is translated to a change in
the displacement of the dipole axis along the x direction with mined by the Johnson-Nyquist thermal noise characteristics
respect to highly sensitive and high-bandwidth magnetic field of the MR sensor. The sensing resolution over bandwidth Δf
sensors, such as spintronic sensors. is then given by
In spintronics, the spin polarization of current carrying
2 kb · T · R · Δf
electrons can be exploited to control the current flow by Res = [m],
∂R ∂B
changing the magnetic moments of thin ferromagnetic films ∂B
· ∂x
·I
[11]. As a result, strong magnetoresistance (MR) effects can
be obtained, such as the giant magnetoresistance (GMR) [12] where kB is the Boltzmann constant, T is the ambient tem-
and the tunnel magnetoresistance (TMR) [13, 14]. MR sen- perature and I is the bias current. In a typical GMR sensor, the
sors based on these effects exhibit significant change in the measurement noise will be affected by additional phenomena,
electrical resistance as a function of the magnetic field. To such as random fluctuations in the magnetic domain orien-
yield a high-field sensitivity, these sensors are typically spin- tation and Barkhausen noise [21, 22]. However, the equili-
engineered for low to intermediate magnetic saturation brium fluctuations will not significantly affect the sensing
thresholds on the order of tens of millitesla [15]. At the same resolution over a high bandwidth because of their 1/f spec-
time, the sensitivity to the out-of-plane component of the trum, and the effect of Barkhausen noise can be significantly
magnetic field remains low because of shape anisotropy, reduced if the magnetic field change occurs only over a small
which makes the sensor highly directional. Because of their range.

2
Nanotechnology 25 (2014) 375501 T Tuma et al

103 independent nanoscale MR subelements,

ΔR = ∑ ΔRi ( Bx ( xi ) ),
102
Resolution over 1 MHz (nm)

where Bx ( xi ) is the magnetic flux density at the position of the

101
ith subelement (see figure 3(b)) and ΔRi denote the resistance
change of each MR subelement which is characteristic to the
100
specific MR sensor that is deployed. Here, we consider a class
of GMR-based sensors in which ΔRi ( Bx ) is linear and directly
10-1
proportional to the magnitude of the applied field |Bx | ⩽ Bsat
Relative MR 10%
Relative MR 100% below the magnetic saturation threshold, Bsat , and constant
10-2 Relative MR 500%
above the magnetic saturation threshold.
The position sensor characteristics as a function of the
10-3 dimensions of the MR sensor, as well as the
100 200 300 400 500 600 700 800 900 1000
Dipole - sensor vertical distance (μm) micromagnet along the sensing direction, are illustrated in
figures 3(c) and (d), respectively. For purposes of the simu-
Figure 2. Sensing resolution achievable with the spintronic position- lation, we assume a micromagnet with remanent magnetiza-
sensing scheme as a function of the magnetoresistance of the MR tion of 1 T at the pole, magnetic saturation threshold of the
sensor and the dipole-sensor distance. Subnanometer resolution over
1 MHz is predicted for sensors with magnetoresistance of more than MR sensor of 0.1 T and magnet-sensor distance of 50 μm. In
10% and dipole-sensor distances of less than 200 μm. Further scaling figure 3(c), it can be seen that increasing the sensor width
in resolution and range is possible with more sensitive MR sensors results in reduced linearity owing to the non-uniform mag-
and by decreasing the dipole–sensor distance. netic fields. In figure 3(d), it can be seen that increasing the
width of the micromagnet increases the sensing range while
decreasing the width results in increased sensitivity and,
To illustrate the resolution achievable with this sensing hence, better resolution. In the limiting case, with ultra-thin
scheme, we present calculations for a hypothetical scenario micromagnets, the sensing characteristics of a position sensor
in which an infinitesimally small MR sensor is operated in with an ideal magnetic dipole are approached. This gives the
close proximity to a magnetic dipole with a magnetic MR position sensor a unique property: its resolution increases
moment of approximately 8 × 10−7 Am 2 (corresponding, for as its dimensions are scaled down, setting it apart from most
example, to a bulk cube-shaped neodymium (NdFeB) of the existing position sensors.
magnet of 100 μm side). In figure 2, the resolution over To further increase the sensor resolution without affect-
bandwidth of Δf = 1 MHz is shown as a function of the ing the linearity, the micromagnet can be extended perpen-
sensor-relative magnetoresistance, defined as the maximum dicular to the sensing direction as shown in figure 3(e). The
MR-sensing elements are placed coaxially and interconnected
relative change in resistance of the sensor, and the dipole-
in a zig-zag-shaped structure. Here, by connecting N such
sensor distance. We can see that the resolution achievable
sensing elements, the sensing resolution can be improved by a
even with moderately sensitive MR sensors and dipole-
sensor distances on the order of hundreds of micrometers factor of N assuming independent sensing noise.
reaches levels as low as 1 nm over 1 MHz. The sensing
resolution can be further enhanced by increasing the field
3. Experimental characterization
sensitivity of the MR sensors and by increasing the magnetic
moment of the dipole.
A prototype of the position sensor was realized using a GMR
For practical realization of the sensing concept, ultra-thin
magnetic field sensor and a block-shaped rare-earth NdFeB
permanent magnets can serve as approximations to the
micromagnet. Figure 4(a) shows a micrograph of the
magnetic dipole. This is also needed to achieve a linear sensor experimental sensor realization. The GMR sensor is a single-
response over an acceptable range of displacement. Practical axis analog magnetometer that detects the magnetic field
limits also dictate that the MR sensor should have finite strength in one direction in the sensing chip plane and has a
dimensions. A schematic illustration of such a realization is cosine-like fall-off for fields out of the sensing direction
shown in figure 3(a). Due to the deviations from an ideal (NVE Corporation, Eden Prairie, MN, USA). The sensor has
magnetic dipole and an infinitesimally small MR sensor, a multiple conductive and soft ferromagnetic layers in a sand-
semi-analytical model was developed to predict the sensing wich-like structure, with no pinning layer and an omnipolar
characteristics. response characteristic. The nominal relative magnetoresis-
In the model, the magnetic flux density is computed for a tance is approx. 13–16%, the active sensing area is 35 μm
rectangular micromagnet using an analytic expression [23]. wide along the sensing direction and the sensor has a zig-zag-
To capture the non-uniform field distribution over the sensor, like sensing element geometry as is evident from the top-view
the resistance change of the MR sensor, ΔR , is modeled as a micrograph in the inset. The magnetic-field-sensing band-
summation of the resistance change of a serial network of width is greater than 1 MHz. The GMR sensor chip is also

3
Nanotechnology 25 (2014) 375501 T Tuma et al

a Sensing direction
b
x
Bx(x)
Low conductivity
Micromagnet
Width
R1(Bx (x1)) R2(Bx (x2)) RN(Bx (xN))
RHIGH
Ferromagnetic layer
High conductivity
Sensing current Conduction layer c d
Ferromagnetic layer 0 % Sensor width 0%
Magnet width

Change in resistance (Ω)

1 μm 100 μm
10 μm 500 μm
RLOW Spintronic sensor 50 μm 1000 μm

100 % 100 %
Magnetic flux density Flux density gradient -100 0 100 -100 0 100
X position (μm) X position (μm)

e Moving body
Direction of sensing
Micromagnet
Sensing element

S
Magnet-sensor gap

Spintronic sensor Frame of reference

Figure 3. Practical realization of the position sensor. (a) Schematic of the position sensor, which consists of a micromagnet positioned near an
MR-based magnetic field sensor. (b) Model of the position sensor response based on a network of nanoscale MR subelements. (c) Effect of
the MR sensor width on the position sensor response; a fixed micromagnet width of 500 μm is assumed. (d) Effect of the micromagnet width
on the position sensor response; a fixed MR sensor width of 10 μm is assumed. (e) Schematic of a possible realization of the position sensor
in three dimensions, including a zig-zag-shaped MR sensor for high sensitivity and resolution.

equipped with two flux concentrators to increase the sensi- For the flux concentrators, we assumed saturation magnetiza-
tivity. The width of the micromagnet (Magtech Industrial, tion of 735 kA m −1 and exchange constant of 1.3 × 10−11 Jm−1.
Hong Kong) in the sensing direction was chosen to be 800 Figure 4(b) shows the magnetic field flux in the x-
μm, i.e. approximately 20 times the width of the sensing area direction, Bx , as computed by the finite difference method.
of the GMR sensor, to ensure a high magnetic field gradient Due to the proximity of the magnet pole, the magnetic flux is
and, at the same time, to preserve the sensing linearity over a highly non-uniform with a zero density at x = 0, the center of
relatively large sensing range. the sensing element, with a flux gradient of more than 5 mT
Micromagnetic simulations were performed to analyze the over 10 μm. The effect of the flux concentrators can be
magnetic flux density associated with the prototype sensor. We observed on the magnetic field lines (shown in gray). In this
used the Object Oriented MicroMagnetic Framework case, the flux concentrators increase the gradient by more than
(OOMMF, available online at http://math.nist.gov/oommf) for a factor of two.
finite-difference micromagnetic simulations of the permanent A calibrated nanopositioner (Physik Instrumente P-517)
magnet and the flux concentrators. The simulations were con- with a three-dimensional (3D) displacement capability over a
ducted in a quasi-2D geometry (xz-plane) with a cell size of range of 100 μm (in plane) and 20 μm (out of plane) was used
2 × 2 μm . The remanent magnetization of the micromagnet to characterize the response of the position sensor. The per-
was estimated from experimental magnetometric measurements manent micromagnet was mounted on a fixed support, and the
obtained using an independent calibrated magnetometer GMR sensor was positioned relative to the micromagnet by
(Metrolab, Switzerland), and was assumed to be 0.15 T at the means of the nanopositioner. The nanopositioner was under
feedback control using the built-in capacitive sensors to
pole. For the micromagnet, we assumed saturation magnetiza-
ensure high motion precision and accuracy. The GMR sensor
tion of 256 kA m −1 and exchange constant of 1.3 × 10−11 Jm−1 . was biased with a constant current of 1 mA; a custom-built

4
Nanotechnology 25 (2014) 375501 T Tuma et al

a
Micormagnet Sensing element top view

60 μm
Magnet-
Sensing sensor 115 μm
Flux element distance Flux concentrator
concentrator 305 μm

Chip substrate
Flux concentrator Flux concentrator

Chip substrate
Sensing direction

100 μm
Zig-zag-shaped
sensing element

b
Micromagnet

30
Magnet-
sensor
distance
Position z (μm)

GMR sensor chip

Flux concentrator Flux concentrator

Sensing element

Substrate

Sensing direction 5 mT 0 -5 mT

-30
-90 Position x (μm) 90
Figure 4. Experimental realization of the position sensor. (a) A micrograph of the sensor, with the micromagnet positioned near the GMR
sensor. The inset image shows a top view of the sensing element, which in this case is the NVE AA002-01 magnetic field sensor. (b) Model
of the sensor experimental realization and the magnetic flux density obtained through micromagnetic simulations. The magnetic field is
highly non-uniform over the area of the sensing element; the additional flux concentrators increase the field gradient by more than a factor
of two.

low-noise instrumentation amplifier was used to read out the captures the key sensor characteristics, such as the saturation
electrical resistance of the sensing element. effect, sensitivity and linearity, remarkably well over a large
Figure 5(a) shows the experimentally obtained position- range of operating conditions.
sensor response for various magnet–sensor (vertical) dis- The sensitivity as a function of the magnet–sensor dis-
tances, and figure 5(b) shows a simulated response based on tance is plotted in figure 6(a). For large and intermediate
the semi-analytical model developed earlier. In the simula- vertical distances, the response is linear over a range of sev-
tion, the sensor response was modeled as a serial inter- eral tens to hundreds of micrometers, with sensitivities of
connection of N = 20 independent non-linear resistive approximately 5–10 Ω /μm. The gradient of the magnetic field
elements, with each element having a linear response with and hence the magnetic sensitivity increases with shorter
gain of ΔR /N over the linear region, ΔR = 700 Ω , and a vertical distances, resulting in sensitivities of up to 40 Ω /μm.
magnetic saturation threshold of approximately 0.015 T. The
The maximum resistivity change of approximately 700 Ω
effective width of the sensing element was 15 μm and the
amounts to approximately 12% of the nominal resistance of
magnetic field amplification factor due to the flux con-
the sensing element used (R = 5500 Ω ). The linear range in
centrators was 2.5. It can be seen that, despite the complex
the high-sensitivity regime exceeds 10 μm.
underlying physical phenomena, the semi-analytical model

5
Nanotechnology 25 (2014) 375501 T Tuma et al

Magnetic saturation Zero position

a. Experimental sensor response

Direction of micromagnet motion

x 0

Resistance change (Ω)

Micromagnet -200
Vertical
distance
-400
Ferromagnetic layer
-600 1000
Conduction layer
800 μm)
Current Ferromagnetic layer Linear sensing 600 (
-800 region ce
-50 400 a n
GMR sensor 200 ist
0
0 a ld
x position (μm) 50 rti
c
b. Simulated sensor response Ve

Direction of micromagnet motion Single element

x

Resistance change (Ω)

response 0

Micromagnet R -200
Vertical
distance
-400

B -600 1000 m)
Current N elements 800 (μ
e
-800 600
a nc
-50 400 ist
0 200 a ld
c
x position (μm) 50 0 rti
Ve
Figure 5. Experimental response of the position sensor and a qualitative comparison to the semi-analytical model. (a) The experimentally
measured response of the sensor is shown in the right panel as a function of the magnet x-position for varying magnet–sensor vertical
distances. As the left panel shows, the change in the GMR sensor resistance occurs due to complex spintronic phenomena induced by the
micromagnet motion, namely due to the scattering effects between the soft ferromagnetic layers of the GMR sensor. (b) The right panel
shows the sensor response obtained by using the semi-analytical model developed in this paper. The response is obtained as the electrical
resistance of a series of nonlinear resistive elements affected by the magnetic field computed from the micromagnet model. Despite the
conceptual simplicity of the model, the key characteristics of the experimental sensor response are captured remarkably well.

Figure 6(b) shows the experimentally obtained frequency consists of an aluminum cantilever 2 × 5 × 5 mm in size that
spectrum of the measurement noise and compares this to the is actuated in the sensing direction of the position sensor by a
Johnson-Nyquist noise level computed for the nominal thin-plate piezoelectric actuator (Noliac, Denmark) with a
resistance of the sensing element (R = 5.5 kΩ ) at temperature bandwidth of approx. 500 kHz. The cantilever was equipped
T = 300 K. To estimate the measurement noise, the GMR- with a micromagnet for position sensing in the actuation
sensing noise was recorded using a high-bandwidth data direction. By means of a custom-modified high-speed, high-
acquisition system (Spectrum, Germany) with a sampling current amplifier (FCE Electronics, Sweden), the actuator was
frequency of 10 MHz. Frequencies above 1 MHz were low- excited by a chirp waveform of 1 V amplitude in the fre-
pass filtered in the custom-built instrumentation amplifier. quency range of 1 kHz to 1 MHz. The cantilever motion over
As expected, the measurement noise spectrum has a 1/f the entire frequency spectrum was recorded using the MR
characteristic at frequencies below 10 kHz and is then flat up position sensor, as well as with a commercial optical position
sensor (Polytec OFV-5000, Germany) with a sensing band-
to 1 MHz at the level of 0.5 pm/ Hz . The spurious peaks at
width of 350 kHz, which served as a reference measurement
frequency of 50 Hz and its harmonics correspond to the of the oscillator frequency response.
ambient electromagnetic noise due to the power line. Figure 6(c) shows the frequency response obtained by the
Figure 6(b) also shows reference measurements conducted in MR sensor and by the reference sensor over the bandwidth of
a noise-free laboratory of the Binnig and Rohrer Nano- 350 kHz. Multiple mechanical resonances can be observed
technology Center, Rüschlikon, Switzerland. They show that across the measured frequency range, with low-frequency
the influence of the ambient electrical and magnetic noise can mechanical vibration modes corresponding to the cantilever
indeed be further reduced by shielding the sensor and the resonant modes, and high-frequency resonances above 30
read-out electronics. kHz corresponding to the piezo actuator modes. At fre-
To further demonstrate the resolution, bandwidth and quencies above approximately 200 kHz, displacements of less
range of the position sensor, we present a frequency response than 1 nm are recorded. We can see that the response of the
analysis of a miniature harmonic oscillator and compare it to a MR sensor matches remarkably well with that of the reference
reference optical interferometric measurement. The oscillator interferometric sensor for low frequencies of up to

6
Nanotechnology 25 (2014) 375501 T Tuma et al

a b
40 10-4 Measurement noise in ambient conditions
Magnet Measurement noise in a noise-free lab
Johnson noise
10-5
35

PSD (nm2/Hz)
Sensor

Short magnet-sensor 10-6

distance
30
10-7
Sensitivity (Ω/μm)

10-8
25
101 102 103 104 105 106
Frequency (Hz)
c 1000
Micrometer motion
20
100
Magnet

Amplitude (nm)
15 10

Sensor 1
10 0.1 Laser vibrometer
GMR displacement sensor
Large magnet-sensor
distance
Subnamometer motion
5 0.01
1000 800 600 400 200 103 104 105
Magnet-sensor vertical distance (μm) Frequency (Hz)

Figure 6. Sensitivity, resolution and bandwidth measurements. (a) Sensitivity of the sensor as a function of the magnet–sensor distance. (b)
Experimentally obtained measurement noise spectra over the bandwidth of 1 MHz. The 1/f noise corresponds to random fluctuations in the
orientation of the magnetic domains; the noise floor is determined by the thermal noise of the sensing element. (c) Frequency response of a
piezo-actuated harmonic oscillator obtained using the MR position sensor and a reference optical sensor over the bandwidth of 350 kHz.

approximately 60 kHz. At higher frequencies, the measure- 8000

ments are inevitably affected by the differences between both
7000
sensing modalities (e.g. the exact location of the laser spot
and the magnet dimensions play a role), and the dynamics can 6000
only be compared qualitatively.
Sensed position (nm)

MR position sensing is inherently affected by magnetic 5000

hysteresis. In the GMR-based sensor, the hysteresis comes
mainly from the magnetic hysteresis of the free ferromagnetic 4000

layers of the GMR-sensing element and of the flux con-

3000
centrators, if they are used. To measure the DC hysteresis
present in our GMR-based position sensor, we performed 2000
repetitive position measurements and compared them to the
independent measurements from the calibrated sensors in the 1000

nanopositioner. Figure 7 shows the position measured by the

0
MR sensor as a function of the nanopositioner position. 0 1000 2000 3000 4000 5000 6000 7000 8000
During the measurements, the micromagnet was moved Magnet position (nm)

relative to the sensing element repeatedly according to a tri- Figure 7. DC hysteresis of the MR position sensor over the range of
angular waveform of frequency 1 Hz over a range of 8 μm. 8 μm.
The sensitivity was 35 Ω /μm. The hysteresis measured in the
center of the sensing range was less than 0.2%. In experi- 4. Application to high-speed AFM
ments with larger range and lower sensitivity, the hysteresis
typically did not exceed 1%. We attribute the low amount of The high resolution and high bandwidth, non-contact mode of
hysteresis to the relatively small operating range and the fact operation, small form factor and low cost render MR-based
that the GMR sensor and the flux concentrators are operated position sensing highly suitable for applications in scanning
in a low flux density Bx without saturation. Also, it has been probe microscopy, and in particular in AFM. We present an
observed [24] that Barkhausen noise is strongly reduced if the application of the MR position sensor in a custom-built
sign of the magnetic field Bx is not changed, possibly because nanopositioner that enables high-speed AFM imaging with
of a lower size of the magnetic domains. subnanometer resolution.

7
Nanotechnology 25 (2014) 375501 T Tuma et al

a
Scanner Cantilever Sample

Micromagnet

Sample

Sensor
Scan table
Cantilever
Micromagnet support
GMR

b c

Cantilever

Sample
Image
reconstruction
Amplifier Scanner

Spintronic Position
Control
sensor
effort measurement

Feedback
controller
Reference signal

Figure 8. Application of the MR position sensor in a high-speed AFM. (a) Photograph of a high-speed AFM scanner equipped with a
prototype of the position sensor. (b) Feedback loop used for control of the scanner position using the position measurement signal from the
position sensor. (c) AFM topography reconstruction of a 2 × 2 μm scan area obtained in 0.8 s, yielding an image stream with frame rate of
more than 1 frame s−1. The sample consists of platinum nanostructures on a carbon surface.

In the AFM, the sample is positioned relative to an an input to a feedback controller that generates a control effort
atomically sharp probe tip by means of a nanopositioner, or based on the difference between the position measurement
scanner [25]. For non-contact sensing of the scanner dis- and the position reference signal. The control effort is used as
placement, a micromagnet is mounted on the movable part input to an amplifier that drives the high-speed scanner.
of the scanner (scan table) and a GMR sensor is mounted on We present an experiment in which the topography of a
the stationary frame near the magnet. By optimizing the sample that consists of an array of platinum nanostructures on
magnet and the sensor placement, point-based displacement a carbon surface is reconstructed at speeds exceeding
sensing can be established precisely at the location where 1 frame s−1. We used a cantilever with thermo–electric sen-
position control is required with minimal spatial require- sing capability [27, 28], which was operated in constant-
ments, which enables high-bandwidth feedback control of
height contact mode. Figure 8(c) shows a, 3D topography
the scanner.
reconstruction of a 2 × 2 μm scan area that was obtained in
In the experiment presented, the MR position sensor is
0.8 s with the fast scan axis frequency of 125 Hz and closed-
integrated into the single-axis high-speed scanner of a cus-
tom-built dual-stage AFM setup [26]. In the dual-stage AFM, loop feedback bandwidth exceeding 1 kHz. The image quality
the high-speed scanner is used during the raster scan to move confirms that nonlinear effects, such as hysteresis, and
the cantilever relative to the probe along the fast scan axis external disturbances, such as the magnetic field from the
only. Actuation in the remaining axes is facilitated by a large- scanner actuator, have a negligible impact on the measure-
range, low-bandwidth scanner to which the high-speed ment and the closed-loop positioning accuracy. The nan-
scanner is mechanically coupled. Figure 8(a) shows a ometer-scale roughness of the carbon surface is attributed to
photograph of the high-speed scanner with the MR position the roughness of the carbon surface; from the image, we can
sensor mounted directly underneath the scan table. The sensor also infer the conformal deposition of the platinum nanos-
signal is used in a model-based feedback control architecture tructures. These imaging results demonstrate that the MR
to control the motion of the high-speed scanner along the position sensors can be readily used in applications as
lateral scan direction. Namely, as figure 8(b) illustrates, the demanding as high-speed AFM with a fraction of the cost of
position measurement signal from the MR sensor is used as comparable sensing techniques.

8
Nanotechnology 25 (2014) 375501 T Tuma et al

5. Summary and future work of the manuscript. Special thanks go to Evangelos Eleftheriou
for his support of the work.
Over the last few years, significant progress has been made on
position sensing using optical-sensing techniques, such as laser
interferometry and optical encoders. It can be argued that this References
has helped to meet the stringent positioning requirements
associated with semiconductor device fabrication and metrol- [1] Munnig Schmidt R, Schitter G and van Eijk J 2011 The Design
ogy at current technology nodes. However, optical-sensing of High Performance Mechatronics (Amsterdam: IOS Press)
techniques have significant limitations in terms of their cost, [2] Butler H 2011 Position control in lithographic equipment
form factor and complex read-back electronics, making them applications of control IEEE Control Systems 31 28–47
[3] Ando T 2012 High-speed atomic force microscopy coming of
unsuitable for several emerging applications in nanotechnol- age Nanotechnology 23 062001
ogy. Compared to optical-sensing techniques, progress on non- [4] Kodera N, Yamamoto D, Ishikawa R and Ando T 2010 Video
optical-sensing techniques has been rather minimal. imaging of walking myosin V by high-speed atomic force
Our measurements suggest that the proposed spintronic microscopy Nature 468 72–76
position sensing scheme exploiting the magnetic field dis- [5] Hansma P K, Schitter G, Fantner G E and Prater C 2006 High-
speed atomic force microscopy Science 314 601–2
tribution close to a magnetic dipole has the potential of [6] Eleftheriou E et al 2003 Millipede-a MEMS-based scanning-
achieving subnanometer resolution at bandwidths that have probe data-storage system IEEE Trans. Magn. 39 938–45
been unreachable for non-optical position-sensing schemes [7] Pires D, Hedrick J L, de Silva A, Frommer J, Gotsmann B,
until now. An especially appealing attribute of this sensing Wolf H, Despont M, Duerig U and Knoll A W 2010
scheme is the extremely small form factor, which allows Nanoscale three-dimensional patterning of molecular resists
by scanning probes Science 328 732–5
point-based position sensing in microscale structures and [8] Fleming A J 2013 A review of nanometer resolution position
devices where this was not possible before. sensors: operation and performance Sensors and Actuators
The ease with which the resolution and range of opera- A: Physical 190 106–26
tion can be altered is a key attribute of this sensing scheme. [9] Li M, Tang H X and Roukes M L 2007 Ultra-sensitive NEMS-
On one hand, by reducing the dimensions of the magnet and based cantilevers for sensing, scanned probe and very high-
frequency applications Nat. Nanotechnology 2 114–20
MR sensors, the resolution can be increased at the expense of [10] Lantz M A, Binnig G K, Despont M and Drechsler U 2005 A
reduced sensing range. On the other hand, by increasing the micromechanical thermal displacement sensor with
distance between the magnet and the sensing element and/or nanometre resolution Nanotechnology 16 1089–94
by increasing the magnet dimensions, the range of operation [11] Prinz G A 1998 Magnetoelectronics Science 282 1660–3
can be increased significantly. It is also possible to extend the [12] Baibich M N, Broto J, Fert A, van Dau F N, Petroff F,
Etienne P, Creuzet G, Friederich A and Chazelas J 1988
sensor in the third dimension to achieve better resolution Giant magnetoresistance of (001) Fe/(001) Cr magnetic
without compromising the sensing range. Resolution can also superlattices Phys. Rev. Lett. 61 2472–5
be improved by using magnets with higher magnetization and [13] Yuasa S, Nagahama T, Fukushima A, Suzuki Y and Ando K
MR sensors with higher field sensitivity. 2004 Giant room-temperature magnetoresistance in single-
Another key advantage is the simplicity of the read-out crystal Fe/Mgo/Fe magnetic tunnel junctions Nat. Mater. 3
868–71
electronics compared to other sensing schemes. This attribute, [14] Daughton J 2003 Spin-dependent sensors Proc. IEEE 91 681–6
along with the small form-factor, could facilitate application [15] Parkin S, Jiang X, Kaiser C, Panchula A, Roche K and
of these sensors in integrated sensing applications, as well as Samant M 2003 Magnetically engineered spintronic sensors
large sensor networks. It also readily fosters configurations and memory Proc. IEEE 91 661–80
with multiple sensors for large-range position sensing, as well [16] Pannetier M, Fermon C, le Goff G, Simola J and Kerr E 2004
Femtotesla magnetic field measurement with
as enhanced sensing resolution. It can be seen that there are magnetoresistive sensors Science 304 1648–50
several key attributes that make this sensing concept a sig- [17] Baselt D R, Lee G U, Natesan M, Metzger S W,
nificant advancement in the field of nanotechnology, and it is Sheehan P E and Colton R J 1998 A biosensor based on
foreseen that it will play a central role in a multitude of sci- magnetoresistance technology Biosensors Bioelectronics 13
entific and industrial applications. 731–9
[18] Rieger G, Ludwig K, Hauch J and Clemens W 2001 GMR
sensors for contactless position detection Sensors Actuators
A: Phys. 91 7–11
Acknowledgments [19] Kartik V, Sebastian A, Tuma T, Pantazi A, Pozidis H and
Sahoo D R 2012 High-bandwidth nanopositioner with
We thank Urs Egger, Walter Häberle and Prasad Jonnala- magnetoresistance based position sensing Mechatronics 22
295–301
gadda for their technical assistance, Rolf Allenspach for [20] Sahoo D R, Sebastian A, Häberle W, Pozidis H and
comments on the manuscript, Emmanuel Lörtscher and Bernd Eleftheriou E 2011 Scanning probe microscopy based on
Gotsmann for their help with the measurements in the Binnig magnetoresistive sensing Nanotechnology 22 145501
and Rohrer Nanotechnology Center, and Mark Lantz for his [21] Hardner H, Weissman M, Salamon M and Parkin S 1993
support with the magnetic hardware, and Naveen Shamsudhin Fluctuation-dissipation relation for giant magnetoresistive 1/
f noise Phys. Rev. B 48 16156
for his help with the magnetometric measurements. We thank [22] Hardner H, Weissman M, Miller B, Loloee R and Parkin S
John Lygeros, Lukas Kull and Daniel Krebs for technical 1996 Resistance noise in uncoupled giant magnetoresistive
discussions, and Charlotte Bolliger for help with preparation multilayers J. Appl. Phys. 79 7751–6

9
Nanotechnology 25 (2014) 375501 T Tuma et al

[23] Xiao-fan G, Yong Y and Xiao-jing Z 2004 Analytic expression scanning probe microscopy IEEE Trans. Mechatronics 19
of magnetic field distribution of rectangular permanent 1035–45
magnets Appl. Math. Mech. 25 297–306 [27] King W P, Kenny T W, Goodson K E, Cross G, Despont M,
[24] Hardner H, Weissman M and Parkin S 1995 History dependent Dürig U, Rothuizen H, Binnig G and Vettiger P 2001
domain structures in giant-magnetoresistive multilayers Atomic force microscope cantilevers for combined
Appl. Phys. Lett. 67 1938–40 thermomechanical data writing and reading Appl. Phys. Lett.
[25] Salapaka S and Salapaka M 2008 Scanning probe microscopy 78 1300–2
IEEE Control Systems 28 65–83 [28] Sebastian A and Wiesmann D 2008 Modeling and
[26] Tuma T, Haeberle W, Rothuizen H, Lygeros J, Pantazi A and experimental identification of silicon microheater dynamics:
Sebastian A 2014 Dual-stage nanopositioning for high-speed a systems approach J. Microelectromech. Syst. 17 911–20

10