Room-Temperature Antiferromagnetic Memory Resistor: Articles

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

ARTICLES

PUBLISHED ONLINE: 26 JANUARY 2014 | DOI: 10.1038/NMAT3861

Room-temperature antiferromagnetic
memory resistor
X. Marti1,2,3*, I. Fina4,5, C. Frontera4, Jian Liu6, P. Wadley3,7, Q. He8, R. J. Paull1, J. D. Clarkson1,
J. Kudrnovsk9, I. Turek2,10, J. Kune3, D. Yi6, J-H. Chu6, C. T. Nelson11, L. You12, E. Arenholz8,
S. Salahuddin12, J. Fontcuberta4, T. Jungwirth3,7 and R. Ramesh1,6
The bistability of ordered spin states in ferromagnets provides the basis for magnetic memory functionality. The latest
generation of magnetic random access memories rely on an efficient approach in which magnetic fields are replaced by
electrical means for writing and reading the information in ferromagnets. This concept may eventually reduce the sensitivity
of ferromagnets to magnetic field perturbations to being a weakness for data retention and the ferromagnetic stray fields to an
obstacle for high-density memory integration. Here we report a room-temperature bistable antiferromagnetic (AFM) memory
that produces negligible stray fields and is insensitive to strong magnetic fields. We use a resistor made of a FeRh AFM, which
orders ferromagnetically roughly 100 K above room temperature, and therefore allows us to set different collective directions
for the Fe moments by applied magnetic field. On cooling to room temperature, AFM order sets in with the direction of the
AFM moments predetermined by the field and moment direction in the high-temperature ferromagnetic state. For electrical
reading, we use an AFM analogue of the anisotropic magnetoresistance. Our microscopic theory modelling confirms that
this archetypical spintronic effect, discovered more than 150 years ago in ferromagnets, is also present in AFMs. Our work
demonstrates the feasibility of fabricating room-temperature spintronic memories with AFMs, which in turn expands the base
of available magnetic materials for devices with properties that cannot be achieved with ferromagnets.

he energy barrier separating the two stable directions of


ordered spins in ferromagnets is partly due to the macroscopic
moment and the dipolar shape anisotropy fields present in the
sample. Another contribution is provided by the magnetocrystalline
anisotropy energy arising from spinorbit coupling. As the
relativistic spinorbit coupling does not distinguish between
ferromagnets and AFMs and magnetocrystalline anisotropies are
even in the microscopic moment direction, such a bistability can be
readily present even if the collective spin order is AFM (refs 1,2).
The magnetotransport counterparts of the relativistic energy anisotropies are the anisotropic magnetoresistance
(AMR) effects 3,4. Their presence in AFMs was predicted 5 and
experimentally confirmed 610 in tunnel junction devices with an
AFM IrMn electrode and a non-magnetic counter-electrode on
the other side of the tunnel barrier. The spin axis of AFM-aligned
moments can point in different directions and can be detected
electrically by the AMR as in ferromagnets.
Owing to the net moment in ferromagnetic materials, the
electron spins can be oriented by a magnetic field larger than the
coercive field, whose strength is limited by the magnetic anisotropy
fields and rarely exceeds 1 T. Compensated AFMs, in contrast,
have zero net magnetization and thus, unlike ferromagnets, produce
no stray magnetic fields. The zero magnetization also implies that
AFMs are much less sensitive to external magnetic fields than

ferromagnets. The sensitivity scale of AFMs is set by the 100 T


exchange fields 11.
Recently, the concept of an AFM memory was explored on
an atomic scale 12. Bistable AFM configurations of a few Fe
atom chains were detected using low-temperature, atomically
resolved measurement by a spin-polarized tip of a scanning
tunnelling microscope. Here we demonstrate an AFM-AMR
memory that has a simple ohmic resistor geometry and whose
magnetic state can be stored and detected at room temperature.
We show that it is insensitive to magnetic fields and produces
negligible stray magnetic fields. In the first part of the paper, we
summarize key experimental results demonstrating the writing,
reading and retention functionalities of our AFM memory and
present microscopic discussion and calculations of the AFMAMR phenomenon. In the second part we report structural
characterization of our FeRh samples and provide a more detailed
discussion of the magnetic and magnetotransport properties
measured over a broad field and temperature range covering both
the AFM and ferromagnetic states.

Demonstration of the AFM memory resistor concept


Our main results are summarized in Fig. 1. We have grown a 100nm-thick, single-crystalline film of the FeRh AFM (refs 1315) on a
cubic MgO substrate. The AFMferromagnetic transition occurs in

1Department of Materials Science and Engineering and Department of Physics, University of California, Berkeley, California 94720, USA, 2Department of
Condensed Matter Physics, Faculty of Mathematics and Physics, Charles University, 12116 Praha 2, Czech Republic, 3Institute of Physics ASCR, v.v.i.,
Cukrovarnick 10, 162 53 Praha 6, Czech Republic, 4Institut de Cincia de Materials de Barcelona, ICMAB-CSIC, Campus UAB, Bellaterra E-08193, Spain,
5Max Planck Institute of Microstructure Physics, Weinberg 2, Halle D-06120, Germany, 6Materials Science Division, Lawrence Berkeley National
Laboratory, Berkeley, California 94720, USA, 7School of Physics and Astronomy, University of Nottingham, Nottingham NG7 2RD, UK, 8Advanced Light
Source, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA, 9Institute of Physics ASCR, v.v.i., Na Slovance 2, 182 21 Praha 8, Czech
Republic, 10Institute of Physics of Materials ASCR, v.v.i., Zizkova 22, Brno 616 62, Czech Republic, 11National Center for Electron Microscopy, Lawrence
Berkeley National Laboratory, Berkeley, California 94720, USA, 12Department of Electrical Engineering and Computer Sciences, University of California,

Berkeley, Berkeley, California 94720, USA. *email: [email protected]


NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

Figure 1 | AFM-AMR memory functionality in a FeRh resistor. a, Schematic illustration of the AFM FeRh/MgO structure and of the memory writing and
reading set-up. For writing, the sample is cooled in a field HFC from a temperature above the AFMferromagnetic transition in FeRh (we used maximum
field of 9 T and temperature of 400 K allowed in our transport measurement set-up) to below the transition temperature (200 K). Black arrows denote the
orientation of the magnetic moments in the ferromagnetic phase whereas either red or blue arrows denote two distinct configurations of the magnetic
moments in the AFM phase. The resulting AFM spin axis in the low-temperature memory state depends on the direction of HFC , which is either along the
[100] or [010] crystal axis. For reading, electrical current j is driven between electrical contacts (yellow bars) along the [100] direction and the resistance
is detected. b, Resistance measured at 200 K and zero magnetic field after field-cooling the sample with HFC parallel (blue) and perpendicular (red) to the
current direction. The two resistance states are clearly distinct and many successive measurement steps demonstrate the stability of the distinct memory
states. c, The same as in b, but at room temperature. d, Stability of the two memory states at room temperature tested by measuring the resistance while
sweeping a magnetic field H between 1 T applied along the [100] direction. e, The same as in d, while rotating a 1 T magnetic field. f, AMR values
calculated for the Rh-rich (Fe1x Rhx )Rh random alloy using the Kubo formula CPA-TB-LMTO formalism. The AMR is defined as a relative difference
between the resistivity for the spin axis parallel and perpendicular to the current, AMR = (Rskj Rsj )/RskI . Results are shown for the AFM ground state
(filled symbols) and for a hypothetical zero-temperature ferromagnetic state (open symbols) of FeRh.

FeRh close to 400 K; at this temperature, we have applied a magnetic


field HFC to align its magnetization and the corresponding magnetic
moments of FM FeRh along the applied magnetic field. The sample
is then field-cooled below room temperature (200 K) and HFC is
then removed. In this AFM state with no applied magnetic field
we perform a series of four-probe resistance measurements with
the current j applied along the [100] substrate crystal direction.
The same protocol is repeated several times with HFC applied
during field-cooling either along the [100] or [010] substrate crystal
directions (Fig. 1a). The resulting resistances in the AFM state are
stable and fully reproducible in the successive writeread cycles,
and the two cooling-field directions define two distinct resistance
states of the AFM. They remain distinct not only on removing
the magnetic field but also when warming the AFM up to room
temperature, as shown in Fig. 1b,c.
In Fig. 1d,e we demonstrate that the two AFM memory states are
robust against strong magnetic field perturbations. After preparing
one of the states by the above cooling-in-field procedure, we rotate
the sample at room temperature in a magnetic field H of 1 T and
observe a negligible effect on the resistance in either of the two
AFM memory states (Fig. 1e). As in the rotation experiment, the
2

states are not disturbed by sweeping the magnitude of H at a fixed


applied field angle (Fig. 1d). In the detailed discussion below we
show that the retention in our AFM memory is not disturbed up to
the highest fields (9 T) available in our transport measurement setup. However, before resuming the detailed experimental analysis, we
focus in the following paragraphs on the microscopic physics behind
the observed distinct resistance states in our FeRh AFM.

Theory of the AFM-AMR


In this theoretical section, we first recall the fundamentals of the
AMR relevant to our experiments and then discuss our quantitative
modelling of the effect based on a relativistic density-functional
transport theory 1620. As already mentioned in the introduction,
conceptually the AMR phenomena are equally present in AFMs as
in ferromagnets. As AMR is an even function of the microscopic
magnetic moment vector, it is the direction of the spin axis (s) rather
than the direction of the macroscopic magnetization (M) relative
to the current direction that primarily determines the effect. In
collinear ferromagnets the two directions are equivalent. For the
staggered spin configuration of compensated AFMs only the spin
axis can be defined while the macroscopic magnetization is zero.

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

Figure 2 | Characterization of the magnetic and crystal structure of the FeRh film. a, Fe L-edge XMLD difference in absorption intensities in the AFM state
of FeRh for the cooling-field vector HFC k [100] and the X-ray linear polarization vectors E k [100] and E k [010]. b, The same as in a, for the cooling-field
vector HFC k [010] and the X-ray linear polarization vectors E k [100] and E k [010]. In a,b, the maximum field of 0.5 T allowed in the XMLD set-up was
applied when cooling the sample from 500 to 200 K and then removed before room-temperature XMLD measurements were performed. c, Fe L-edge
XMLD difference in absorption intensities in the ferromagnetic state of FeRh at 500 K for the X-ray linear polarization vector E k [100] and applied
magnetic fields H k [100] and H k [010]. The AFM and ferromagnetic XMLD signals are of comparable strength and the signs of the signals are consistent
with the AFM spin axis oriented perpendicular to HFC , as indicated in the insets. d, Cross-section image of the FeRh thin film at the interface with the MgO
substrate, viewed down the (112) and (100) zone axis of the two respective materials. Scale bar, 2 nm. e, 2 scan showing textured films; indicated are
the (00l) reflections of the substrate and the film. f, Compositional map of the Fe/Rh content of the film.

To date, studies of AMR phenomena in AFMs have focused


on tunnelling devices 510. In analogy to the tunnelling AMR in
ferromagnets 2126, the origin of the phenomenon is ascribed to the
changes in the equilibrium relativistic electronic structure (density
of states) induced by rotating the spin axis with respect to crystal
axes 5. A recently reported AMR in an ohmic device fabricated from
an AFM Sr2 IrO4 film shares the anisotropic electronic structure
origin of the tunnelling AMR (ref. 27). In the Sr2 IrO4 AMR
experiment, the current was driven through the film in the direction
perpendicular to the sample plane while the AFM spin axis was
rotated in the plane. This experimental geometry in which the
angle between the spin axis and current is fixed and the resistance
depends on the angle between the spin axis and crystal axes is
referred to as the crystalline ohmic AMR (refs 4,28). For both the
tunnelling AMR and the crystalline ohmic AMR, the quantitative
relativistic transport theory would require the combination of
the calculated density of states anisotropy with the tunnelling or
scattering matrix elements, respectively. As a proper modelling of
these matrix elements for realistic sample parameters is in general a
difficult problem, the theories have focused primarily on assessing
the qualitative origin of these AMR phenomena based on the density
of states anisotropy calculations.
In our ohmic FeRh resistor the different directions of the AFM
spin axis are set by a field-cooling procedure in which the magnetic
field is applied either along the [100] or [010] crystal axis of our
MgO/FeRh sample. Owing to the in-plane cubic symmetry of the
structure, the equilibrium density of states of the FeRh film is
identical for the AFM spin axis set by the cooling field applied along

equivalent cubic axes. Unlike the above tunnelling or crystalline


ohmic AMR, the equilibrium density of states anisotropy is not the
origin of the resistance anisotropy we observe in our FeRh resistor.
Instead, the effect originates from the change of the angle of the spin
axis with respect to the electrical current direction. This is akin to
the common non-crystalline AMR in ferromagnets 4,28, which is the
only component contributing to the AMR in polycrystalline samples
or when the spin axis alternates between equivalent symmetry axes
of the crystal. The dependence on the relative angle of the spin axis
and current is the phenomenon that Kelvin detected in his more
than 150-year-old seminal work on the AMR in Fe and Ni (ref. 3).
As the equilibrium density of states anisotropy of a clean
crystal does not provide even a qualitative proxy to the noncrystalline AMR, its modelling remains a challenge despite the
long history of the effect and the simple experimental geometry
used for its detection. To first order, the non-crystalline AMR
reflects the dependence of the scattering matrix elements of
electrons, with momentum parallel to current, on the angle between
the spin axis and the momentum. The spin dependence of the
momentum scattering is due to the relativistic spinorbit coupling.
A quantitative modelling approach requires the identification of
the dominant scattering mechanism and the use of a full quantum
transport theory framework. In ferromagnets, such a quantitative
description has been successfully applied to random metal alloys
in which scattering is dominated by the alloy disorder and the
relativistic band structure is reliably obtained using the densityfunctional theory 1620. Here we apply an analogous formalism to
calculate the non-crystalline ohmic AMR in our FeRh AFM.

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

Figure 3 | Magnetic properties. a, Temperature dependence of the magnetization at 1, 5 and 9 T applied magnetic fields. b, Magnetic field dependence of
the magnetization at 200 K, room-temperature and 400 K. c, Comparison of 400 K magnetization loops for magnetic field applied along the [100] and
[010] MgO crystallographic directions. d, Magnetic field dependence of the resistance at 200 K, room-temperature, 320 and 340 K. Black arrows indicate
the direction of the temperature sweep.

Results of our microscopic calculations are presented in Fig. 1f.


We employed the relativistic, tight-binding linear muffin-tin orbital
(TB-LMTO) density-functional theory of the band structure, the
Kubo formula description of transport, and the alloy disorder
is accounted for within the coherent potential approximation 19,20
(CPA). The AMR plotted in Fig. 1f is defined as a relative difference
between the resistivity for the spin axis parallel and perpendicular
to the current, AMR = (Rskj Rsj )/Rskj . This applies to both the
AFM ground state of FeRh and a hypothetical zero-temperature
ferromagnetic state that we also consider in our calculations for
comparison. Following the chemical composition measurements
shown below in Fig. 2, we assume in the calculations a Rh-rich
(Fe1x Rhx )Rh random alloy and calculate the AMR as a function
of x. The experimentally determined non-stoichiometry in our
samples corresponds to x 0.020.03 (Fe/Rh 0.940.96). For
these values, the calculated AMR 0.11% in the AFM state is
in good agreement with the measured difference between the two
resistance states (see also the temperature-dependent AFM-AMR
measurements below). To confirm the robustness of the theoretically
predicted AMR in AFMs we also performed control calculations
for FeRh in which disorder- and temperature-dependent scattering
(through phonons or magnons) is effectively modelled by a finite
quasiparticle broadening of the Fermi level states. For resistivities
similar to the experimental values, we consistently obtain AFMAMR values close to a half of a per cent. The comparison of
our calculated FeRh AMRs in the AFM ground state and the
hypothetical zero-temperature ferromagnetic state highlights the
expected equal presence and comparable magnitude of the AMR
phenomena in these two distinct ordered spin states. Note that
the calculated AFM-AMR even exceeds the ferromagnetic-AMR at
stronger alloy disorder (larger x), as seen in Fig. 1f.
Important evidence is drawn from the calculated sign of the
AFM-AMR in our FeRh sample. Similar to the calculated sign of
4

the AMR in the ferromagnetic state of FeRh and to the AMR sign in
common transition-metal ferromagnets 4,1620, our theory predicts a
lower resistivity state for the AFM spin axis aligned perpendicular
to the current. This confirms the scenario, anticipated in Fig. 1a,
by which the field-cooling procedure sets in the spin-axis direction
in the FeRh AFM state. In agreement with the theoretical sign of
the AFM-AMR, the ferromagnetically aligned moments along the
applied magnetic field HFC at the high temperature rotate clockwise
in one spin sublattice and anticlockwise in the other sublattice of
FeRh until they reach the AFM configuration at the low temperature
with the spin axis oriented perpendicular to HFC .

X-ray magnetic linear dichroism


Further evidence for the above field-cooling writing scenario
is provided by our Fe L-edge X-ray magnetic linear dichroism
(XMLD) experiments. Using X-rays offers the elemental specificity
while the magnetic linear dichroism, characterized by different
absorption of light with its linear polarization vector E oriented
parallel or perpendicular to the spin axis, represents the optical
analogue to the AMR (ref. 29). XMLD is also an even function of
the microscopic magnetic moment vector and is equally present
in AFMs as in ferromagnets 29. After field-cooling the sample
from above the AFMferromagnetic transition down to 200 K and
removing HFC , we warmed the sample to room temperature and
performed the XMLD measurements in the AFM state of FeRh. In
Fig. 2a we plot the XMLD spectrum for the cooling-field vector
HFC k [100] and the X-ray polarization vectors E k [100] and
E k [010](IEk[100] IEk[010] ). In Fig. 2b the X-ray polarization vector
directions were set in the same order as in Fig. 2a; however, HFC
was applied along the [010] axis. The opposite sign of the XMLD
spectra in Fig. 2a,b confirms the magnetic origin of the linear
dichroism signal and confirms that the AFM spin axes in the two
field-cooling experiments are rotated by 90 with respect to each

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

Figure 4 | Stability of the memory states at high magnetic fields. a, Resistance of the two memory states at 200 K and in a magnetic field swept between
9 T along the [100] direction. The insets show two distinct configurations of the AFM moments in the device. b, The same as in a, in a 9 T rotating
magnetic field. In a,b, the red square and blue circle indicate the zero-field resistances of the two memory states after the cooling-in-field writing
procedure. c, The same as in a, measured at room temperature. d, The difference between resistances of the two memory states plotted in c. In c,d, the
arrows indicate the direction of sweeping of the magnetic field.

other. To determine the spin-axis orientation, in Fig. 2c we plot the


Fe L-edge XMLD spectrum in the high-temperature ferromagnetic
state of FeRh. The measurement was performed with E k [100]
and in applied magnetic fields H k [100] and H k [010](IHk[100]
IHk[010] ). The ferromagnetic XMLD signal has a similar shape but an
opposite sign to the XMLD (IEk[100] IEk[010] ) in the AFM state with
HFC k [100] (Fig. 2a). As the moment/spin axis is parallel to the
field in the ferromagnetic state, this sign change indicates that the
AFM spin axis is along [010], confirming the expected AFM spinaxis orientation perpendicular to HFC , in agreement with the above
discussed sign of the AMR. Note that the comparison of measured
AFM and ferromagnetic XMLD signals allowed us to confirm the
spin-axis direction for the given direction of HFC provided that the
sign of the XMLD signal is independent of whether the spin-axis
direction corresponds to the AFM or ferromagnetically aligned Fe
moments. This can be verified by the full-potential ab initio theory
of XMLD spectra in transition-metal magnets 30. We performed the
calculations and found, as expected, that the sign of the XMLD
signal is the same for the AFM and ferromagnetic states of FeRh
and that the overall XMLD spectra are almost identical for the two
types of magnetic order.

Detailed structural and magnetic characterization


Having established the microscopic physical description of key
functionalities of our AFM memory resistor we resume in
the remaining paragraphs the more detailed description of our
experimental findings. The high epitaxial quality of our FeRh films
is illustrated in Fig. 2df. The cross-section of the FeRh thin film
at the interface with the MgO substrate (Fig. 2d), viewed down the
(112) and (100) zone axis of the two respective materials, shows
that the FeRh epilayer is fully coherent with the MgO substrate. The
micrograph demonstrates nearly perfect ordering of the atoms in
the FeRh unit cell. X-ray diffraction patterns (Fig. 2e) show only the

{00l} family of peaks of both FeRh and MgO, confirming that the
samples are textured with no spurious phases. In Fig. 2f we show the
wavelength dispersive X-ray spectroscopy analysis of the chemical
composition of our FeRh film. We observe a small Fe deficit from
the targets nominal 1:1 stoichiometry with Fe/Rh 0.940.96.
Figure 3 presents magnetic characterization of our FeRh film
by the superconducting quantum interference device and vibrating
sample magnetometer, which provides further insight into the
behaviour our AFM memory. In Fig. 3a we show temperaturedependent magnetization data recorded at different magnetic fields.
When warming up, the magnetization exhibits a clear signature
of the transition from the AFM to the ferromagnetic phase of
FeRh. The thermal hysteresis reflects the first-order nature of the
transition 4. When increasing the magnetic field, the ferromagnetic
phase is stabilized and the transition occurs at lower temperatures.
Note that at 400 K, a field of 9 T is strong enough to fully stabilize
the ferromagnetic state. In our transport measurement set-up, the
highest allowed temperature was 400 K, so we applied HFC of 9 T
to safely enter the ferromagnetic state at this temperature. At lower
temperatures, the system gradually turns into the AFM state. The
XMLD experimental set-up allowed us to perform the field-cooling
procedure from 500 K. In this case, HFC of 0.5 T was sufficient to
stabilize the two orthogonal metastable spin axes configurations in
the room-temperature AFM memory state.
Complementary magnetic-field-dependent magnetization traces
are plotted in Fig. 3b,c. At 400 K, the data are consistent with
the ferromagnetic alignment of Fe spins and identical signal is
detected while sweeping the magnetic field along the [100] or
[010] crystal directions (Fig. 3c). This confirms the expected inplane cubic symmetry of our FeRh film deposited on the cubic
MgO substrate. We note that the loops were obtained by sweeping
the magnetic field between 9 and 400 K. Magnetization data
measured at room temperature and 200 K (Fig. 3b) confirm the

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

Figure 5 | Temperature dependence of the AMR. a, Resistance measured while field-cooling from 400 K in HFC of 9 T k [100] and while subsequently
warming up in zero magnetic field. b, Relative resistance difference, (R(HFC k [010]) R(HFC k [100]))/R(HFC k [100]), for the field-cool and
zero-field-warm measurements. c, Low-temperature zero-field-warm relative resistance differences, (R(HFC k [010]) R(HFC k [100]))/R(HFC k [100])
(solid line) and (R(HFC k [110]) R(HFC k [100]))/R(HFC k [100]) (dashed line). d, High-temperature field-cool relative resistance differences,
(R(HFC k [010]) R(HFC k [100]))/R(HFC k [100]) (solid line) and (R(HFC k [110]) R(HFC k [100]))/R(HFC k [100]) (dashed line).

AFM order. Only a very weak residual moment remains in the


system, as seen from the minor hysteretic contribution to the
signal at low (100 mT) fields. Its origin can be associated with
the small departure of our FeRh film from the nominal 1:1
stoichiometry of the fully compensated FeRh AFM (Fig. 2b) or
with the presence of residual ferromagnetic regions in the AFM
matrix as is common in first-order transitions. We ascribe the
weak variation of the resistance seen in Fig. 1e in the fieldrotation measurement to this residual moment. We point out,
however, that the residual moment that is reversed at the low,
100 mT field (Fig. 3b) cannot be responsible for the AMR
signal distinguishing the two memory states because these distinct
resistance states remain stable up to the largest field (9 T) applied
in our experiments. The residual moment has also no marked
effect on our AFM memory states as the difference between the
corresponding resistances is much larger than the weak variations
of the individual angular dependencies of each of the two memory
states.
Magnetization data in Fig. 3ac are further complemented
by the temperature-dependent magnetotransport measurements
shown in Fig. 3d. At 340 K, the moments are aligned in the
large applied magnetic field, resulting in a negative and hysteretic
magnetoresistance trace in the field-sweep measurement. Similar
data are obtained at 320 K. However, the hysteretic field-dependent
part of the resistance trace is shifted to higher magnetic fields as we
move further away in temperature from the AFMferromagnetic
transition. We note that these measurements are consistent with
a recent study of similar FeRh/MgO samples that focused on
magnetotransport effects near the AFMferromagnetic transition 14.
The aim of our paper is, however, to demonstrate the memory
functionality and AMR in the AFM state below the transition. In
particular, at room temperature and at 200 K we do not see any field
6

dependence of the resistance on the xy-scale of Fig. 3d, confirming


that the FeRh film behaves basically as a compensated AFM with a
negligible residual moment at these low temperatures and up to the
highest applied magnetic fields.
In Fig. 4 we inspect the 200 K and room-temperature resistance
traces in the large field range and on a fine resistance scale, for
the two memory states set up by the cooling-in-field procedures
described in Fig. 1. At 200 K we observe a parabolic positive
magnetoresistance that reaches 1% at 9 T (Fig. 3a). We ascribe it to
the ordinary, Lorentz force magnetoresistance that is independent
of spin. Indeed, the two traces have just a constant offset; that is,
cycling our AFM memory between 9 and 200 K has no effect
on the stability of the two distinct AFM states. This is further
confirmed in Fig. 4b where instead of sweeping the magnetic field
we rotate the sample in a 9 T applied field. We observe small
variations of the resistance, which we ascribed above to the residual
uncompensated moments. However, as seen in Fig. 4b, the two
distinct AFM memory states remain clearly separated even in the
9 T rotating field.
In Fig. 4c we plot room-temperature magnetoresistance traces on
a similar fine scale and we observe a different behaviour in the strong
magnetic fields than at 200 K. The ordinary spin-independent
positive magnetoresistance is not visible because of a stronger
hysteretic negative magnetoresistance. Compared with the traces at
320 and 340 K plotted in Fig. 3d, it is clear that this, relatively small
hysteretic magnetoresistance seen at room temperature reflects that
with increasing temperature and large field the system approaches
the transition to the ferromagnetic coupling of spins. However, at
room temperature and up to the highest applied magnetic field
of 9 T the system remains far from the complete transition to the
ferromagnetic order. As further highlighted in Fig. 4d, this implies
that we can cycle our AFM FeRh memory between 9 T fields

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861


at room temperature without disturbing either of the two distinct
memory states.
Finally, in Fig. 5, we provide a more detailed investigation of
the temperature dependence of the FeRh AMR spanning a broad
temperature range from 400 to 77 K. To relate these measurements
to the writing and reading methods demonstrated above for our
FeRh AFM memory we show transport data acquired while cooling
the sample in a magnetic field and, subsequently, by warming it
in zero magnetic field. The temperature-dependent resistances are
plotted in Fig. 5a. The resistance trace measured while cooling in
an applied field (HFC of 9 T k [100]) shows the step corresponding
to the ferromagneticAFM transition at a lower temperature than
the resistance trace measured while subsequently warming the
sample in a zero field, consistent with the measurements shown in
Fig. 4. The relative difference between resistance traces for HFC k
[010] and [100] is plotted in Fig. 5b. On the low-temperature
side below the ferromagneticAFM transition, the data correspond
to the temperature-dependent AFM-AMR as defined in Fig. 1.
As expected, the magnitude of the AMR increases when moving
from the transition towards low temperatures and reaches the
1% scale in agreement with our microscopic calculations. On the
high-temperature side, we can use the measurements in applied
magnetic field to assess the AMR in the ferromagnetic state. A
detailed comparison of the AFM-AMR on the low-temperature
side and the ferromagnetic-AMR on the high-temperature side is
presented in Fig. 5c,d. Considering the relative angle of the spin
axes and current, the measurements confirm that the AFM and
ferromagnetic AMRs have the same sign because in the former case
the spin axis aligns perpendicular to HFC while in the latter case the
spin axis (magnetization) is parallel to HFC . In Fig. 5c,d we also show
the relative difference between resistance traces for HFC k [110]
and [100]. Consistent with the AMR phenomenology, HFC k [110]
(45 from the current direction) produces an intermediate resistance
state between the two extreme resistance states prepared by HFC k
[100] and HFC k [010].

Discussion
We have demonstrated a room-temperature AFM memory and
explained the AFM-AMR effect we employed for a simple ohmic
resistance read-out. Our magnetic memory shows the unique
features of AFMs. In particular, it is insensitive to strong magnetic
field perturbations and produces negligible stray fields. The
magnitude of the AFM-AMR effect in our FeRh sample is sufficient
to reproducibly detect the two distinct memory states at room
temperature.
Our work represents the proof of principle demonstration of
an AFM memory resistor, yet there is much room for improving
all of the functionalities of such a device. We note, however, that
the magnetic memory concept we introduce is generic to a broad
class of spinorbit-coupled AFMs. Our microscopic calculations
and temperature-dependent AFM-AMR measurements indicate a
route for increasing the magnitude of the AFM-AMR in materials
with higher transition temperatures. The same theory framework
or measurements can systematically map ohmic AMRs in a variety
of metallic AFMs. To improve our cooling-in-field procedure for
writing, future studies of AFM memories can benefit from the
extensive research and development of techniques in ferromagnets
for heat-assisted magnetic recording 31. In our FeRh samples, the
temperature and field strength can still be optimized to achieve the
most efficient writing in this material. Other systems that show the
high-temperature AFMferromagnetic transition can be considered
or, as recently demonstrated in a tunnelling IrMn device 10, heating
close to the AFM to paramagnetic transition in materials without
the AFMferromagnetic transition point may by a viable alternative.
A highly attractive approach for AFMs is to use spinorbit coupling
not only for reading but also for writing the state by electrical means.

This is a concept that in ferromagnets has recently drawn much


attention from both fundamental physics and memory application
perspectives 32,33 but in AFMs is yet to be explored. As the effect
of the spin-axis reorientation on electrical transport has been
demonstrated in our work, the commonly observed reciprocity
between spin and charge dynamics implies the possibility to
generalize the ferromagnetic spin-torque phenomena 34 to AFMs.
Beyond the electrical means, spins in magnetic materials can be
reoriented optically, which in AFMs may provide for ultrafast,
precession-free writing schemes 35.

Methods
Thin films were prepared by d.c. sputtering. MgO(001) substrates were heated
up to 550 K in a base pressure of 108 torr. Subsequently, Ar gas was introduced
(3 mtorr) and films were grown using 50 W power at a rate of 1 nm min1 using a
stoichiometric FeRh target.
X-ray diffraction experiments were carried out using a Panalytical XPert Pro
diffractometer by Philips. The composition and thickness of the thin films were
obtained by variable-voltage electron-probe microanalysis, using a CAMECA SX-50
electron microprobe equipped with four wavelength-dispersive X-ray spectrometers.
X-ray intensities were measured at 10, 12, 15 and 20 keV electron incident energy
and they were analysed with the help of the program STRATAGEM (SAMX), which
calculates the thickness and composition of a multilayer target by least-squares
fitting of an analytical X-ray emission model to the experimental data.
The cross-sectional sample for transmission electron microscopy was prepared
by mechanical polishing followed by Ar ion milling. High-angle annular dark-field
scanning transmission electron microscopy imaging was performed on an FEI titan,
the TEAM0.5 at the National Center for Electron Microscopy. The image shown in
Fig. 2a is an averaged region from a scanning transmission electron micrograph
taken with a high-angle annular dark-field detector. This mode highlights the atom
nucleus with intensity corresponding to the atomic number; here Rh (Z = 45)
appears brightest followed by Fe (Z = 26), and Mg (Z = 12) in the substrate. The
image was averaged laterally over approximately 30 adjacent unit cells.
Magnetic characterization with magnetic fields up to 5 T was carried out in a
superconducting quantum interference device by Quantum Design using the
Reciprocating Sample Option. Measurements up to 9 T were performed in a
vibrating sample magnetometer by Quantum Design.
Fe L-edge XMLD measurements were performed at Beamline 4.0.2 of the
Advanced Light Source. Field-cooling was performed from 500 K with the largest
magnetic field of 0.5 T allowed by the XMLD measurement set-up.
Transport experiments were four-probe resistance measurements carried
out using the Physical Property Measurement System by Quantum Design in
resistivity mode and equipped with a horizontal rotator module. For parallel
current (bi-polar constant current ranging from 1 to 10 A) and magnetic field
position is defined as 0, along the substrate [100] direction shown. Maximum
magnetic field and temperature allowed by the transport measurement set-up are
9 T and 400 K, respectively.
Kubo formula CPA-TB-LMTO calculations were performed for the cubic, CsCl
crystal structure of FeRh. The structure of the AFM state was described as a
stacking of four interpenetrating face-centred-cubic sublattices along the
[111]-direction occupied respectively by Fe[up], Rh, Fe[down] and Rh atoms.
Substitutional RhFe disorder was placed on both Fe sublattices in the considered
Rh-rich (Fe1x Rhx )Rh random alloy. The model of quasiparticle broadening was
treated by adding a constant, spin-independent imaginary part of the self-energy
(of order 103 102 Ry). A large number (107 ) of points in the full
face-centred-cubic Brillouin zone was used to obtain converged AMR values.

Received 7 August 2013; accepted 10 December 2013;


published online 26 January 2014

References
1. Umetsu, R. Y., Sakuma, A. & Fukamichi, K. Magnetic anisotropy energy of
antiferromagnetic L10-type equiatomic Mn alloys. Appl. Phys. Lett. 89, 052504
(2006).
2. Szunyogh, L., Lazarovits, B., Udvardi, L., Jackson, J. & Nowak, U. Giant
magnetic anisotropy of the bulk antiferromagnets IrMn and IrMn3 from first
principles. Phys. Rev. B 79, 020403 (2009).
3. Thomson, W. On the electro-dynamic qualities of metals: Effects of
magnetization on the electric conductivity of nickel and of iron. Proc. R. Soc.
Lond. 8, 546550 (1857).
4. McGuire, T. & Potter, R. Anisotropic magnetoresistance in ferromagnetic 3d
alloys. IEEE Trans. Magn. 11, 10181038 (1975).
5. Shick, A. B., Khmelevskyi, S., Mryasov, O. N., Wunderlich, J. & Jungwirth, T.
Spin-orbit coupling induced anisotropy effects in bimetallic antiferromagnets:
A route towards antiferro-magnetic spintronics. Phys. Rev. B 81, 212409
(2010).

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

ARTICLES

NATURE MATERIALS DOI: 10.1038/NMAT3861

6. Park, B. G. et al. A spin-valve-like magnetoresistance of an


antiferromagnet-based tunnel junction. Nature Mater. 10, 347351 (2011).
7. Duine, R. Spintronics: An alternating alternative. Nature Mater. 10, 345 (2011).
8. Marti, X. et al. Electrical measurement of antiferromagnetic moments in
exchange-coupled IrMn/NiFe stacks. Phys. Rev. Lett. 108, 017201 (2012).
9. Wang, Y. Y. et al. Room-temperature perpendicular exchange coupling and
tunneling anisotropic magnetoresistance in an antiferromagnet-based tunnel
junction. Phys. Rev. Lett. 109, 137201 (2012).
10. Petti, D. et al. Storing magnetic information in IrMn/MgO/Ta tunnel junctions
via field-cooling. Appl. Phys. Lett. 102, 192404 (2013).
11. Marder, M. P. Condensed Matter Physics (Wiley, 2000).
12. Loth, S., Baumann, S., Lutz, C. P., Eigler, D. M. & Heinrich, A. J. Bistability in
atomic-scale antiferromagnets. Science 335, 196199 (2012).
13. Shirane, G., Chen, C. W., Flinn, P. A. & Nathans, R. Mossbauer study of
hyperfine fields and isomer shifts in the FeRh alloys. Phys. Rev. 131, 183190
(1963).
14. Sharma, M. et al. Magnetotransport properties of epitaxial MgO(001)/FeRh
films across the antiferromagnet to ferromagnet transition. J. Appl. Phys. 109,
083913 (2011).
15. Mariager, S. O., Guyader, L. L., Buzzi, M., Ingold, G. & Quitmann, C. Imaging
the antiferro-magnetic to ferromagnetic first order phase transition of FeRh.
Preprint at http://arxiv.org/abs/1301.4164.
16. Banhart, J. & Ebert, H. First-principles theory of spontaneous-resistance
anisotropy and spontaneous Hall effect in disordered ferromagnetic alloys.
Europhys. Lett. 32, 517522 (1995).
17. Ebert, H., Vernes, A. & Banhart, J. Anisotropic electrical resistivity of
ferromagnetic CoPd and CoPt alloys. Phys. Rev. B 54, 84798486 (1996).
18. Vernes, A, Ebert, H. & Banhart, J. Electronic conductivity in NiCr and NiCu
fcc alloy systems. Phys. Rev. B 68, 134404 (2003).
19. Turek, I. & Zalezak, T. Residual resistivity and its anisotropy in random CoNi
and CuNi ferromagnetic alloys. J. Phys. Conf. Ser. 200, 052029 (2010).
20. Turek, I., Kudrnovsk, J. & Drchal, V. Ab initio theory of galvanomagnetic
phenomena in ferromagnetic metals and disordered alloys. Phys. Rev. B 86,
014405 (2012).
21. Gould, C. et al. Tunneling anisotropic magnetoresistance: A spin-valve like
tunnel magnetoresistance using a single magnetic layer. Phys. Rev. Lett. 93,
117203 (2004).
22. Brey, L., Tejedor, C. & Fernndez-Rossier, J. Tunnel magneto-resistance in
GaMnAs: Going beyond Jullire formula. Appl. Phys. Lett. 85, 19961998
(2004).
23. Shick, A. B., Mca, F., Maek, J. & Jungwirth, T. Prospect for room
temperature tunnelling anisotropic magnetoresistance effect: Density of states
anisotropies in CoPt systems. Phys. Rev. B 73, 024418 (2006).
24. Gao, L. et al. Bias voltage dependence of tunneling anisotropic
magnetoresistance in magnetic tunnel junctions with MgO and Al2 O3 tunnel
barriers. Phys. Rev. Lett. 99, 226602 (2007).
25. Moser, J. et al. Tunneling anisotropic magnetoresistance and spinorbit
coupling in Fe/GaAs/Au tunnel junctions. Phys. Rev. Lett. 100, 056601 (2007).
26. Park, B. G. et al. Tunneling anisotropic magnetoresistance in
multilayer-(Co/Pt)/AlOx /Pt structures. Phys. Rev. Lett. 100, 087204 (2008).
27. Marti, X. et al. Anisotropic magnetoresistance in antiferromagnetic
semiconductor Sr2 IrO4 epitaxial heterostructure. Preprint at
http://arxiv.org/abs/1303.4704.

28. Rushforth, A. W. et al. Anisotropic magnetoresistance components in


(Ga,Mn)As. Phys. Rev. Lett. 99, 147207 (2007).
29. Sthr, J., Padmore, H. A., Anders, S., Stammler, T. & Sheinfein, M. R.
Principles of X-ray magnetic dichroism spectromiscroscopy. Surf. Rev. Lett. 5,
12971308 (1998).
30. Kune, J. & Oppeneer, P. M. Anisotropic x-ray magnetic linear dichroism at
the L edges of cubic Fe, Co, and Ni: Ab initio calculations and model theory.
Phys. Rev. B 67, 024431 (2003).
31. Chappert, C., Fert, A. & Dau, F. N. V. The emergence of spin electronics in
data storage. Nature Mater. 6, 813823 (2007).
32. Miron, I. M. et al. Perpendicular switching of a single ferromagnetic layer
induced by in-plane current injection. Nature 476, 189 (2011).
33. Liu, L. et al. Spin-torque switching with the giant spin hall effect of tantalum.
Science 336, 555558 (2012).
34. Kurebayashi, H. et al. Observation of a Berry phase anti-damping spin-orbit
torque. Preprint at http://arxiv.org/abs/1306.1893.
35. Kimel, A. V., Kirilyuk, A., Tsvetkov, A., Pisarev, R. V. & Rasing, Th.
Laser-induced ultrafast spin reorientation in the antiferromagnet TmFeO3 .
Nature 429, 850853 (2004).

Acknowledgements
The authors acknowledge the support from the NSF (Nanosystems Engineering
Research Center for Translational Applications of Nanoscale Multiferroic Systems,
Cooperative Agreement Award EEC-1160504) and DOE. Transmission electron
microscopy characterization was performed at NCEM, which is supported by the
Office of Science, Office of Basic Energy Sciences of the US Department of Energy
under Contract No. DE-AC0205CH11231. J.F. acknowledges financial support from
the Spanish Government (Projects MAT2011-29269-C03, CSD2007-00041) and
Generalitat de Catalunya (2009 SGR 00376); C.F. acknowledges financial support from
the Spanish Government (Projects MAT2012-33207, CSD2007-00041). I.F.
acknowledges a Beatriu de Pins postdoctoral scholarship (2011 BP-A 00220) and the
Catalan Agency for Management of University and Research Grants
(AGAUR-Generalitat de Catalunya). X.M. acknowledges the Grant Agency of the
Czech Republic No. P204/11/P339. Research at the University of Nottingham was
funded by EPSRC grant EP/K027808/1. T.J. acknowledges support from the ERC
Advanced Grant 268066, Praemium Academiae of the Academy of Sciences of the
Czech Republic, and from the Ministry of Education of the Czech Republic Grant
LM2011026. S.S. acknowledges funding by STARnet FAME. J. Kune 83 and I.T.
acknowledge the Czech Science Foundation No. P204/11/1228.

Author contributions
Sample preparation, R.J.P., J.D.C., L.Y.; scanning transmission electron microscopy,
C.T.N.; magnetotransport and structural characterization, I.F. and C.F.; data analysis,
I.F., C.F., P.W., J-H.C. and D.Y.; X-ray linear dichroism, J.L., E.A. and Q.H.; theory, J.
Kudrnovsk, I.T. and J. Kune; writing and project planning, X.M., T.J., J.F., P.W., S.S.
and R.R.

Additional information
Reprints and permissions information is available online at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to X.M.

Competing financial interests


The authors declare no competing financial interests.

NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials


2014 Macmillan Publishers Limited. All rights reserved.

You might also like