-
ARTICLESPUBLISHED ONLINE: 21 NOVEMBER 2016 | DOI:
10.1038/NMAT4812
Current-induced switching in a magnetic insulatorCan Onur Avci,
Andy Quindeau, Chi-Feng Pai†, Maxwell Mann, Lucas Caretta, Astera
S. Tang,Mehmet C. Onbasli, Caroline A. Ross* and Geo�rey S. D.
Beach*
The spin Hall e�ect in heavy metals converts charge current into
pure spin current, which can be injected into an
adjacentferromagnet to exert a torque. This spin–orbit torque (SOT)
has been widely used to manipulate the magnetization in
metallicferromagnets. In the case of magnetic insulators (MIs),
although charge currents cannot flow, spin currents can propagate,
butcurrent-induced control of the magnetization in a MI has so far
remained elusive. Here we demonstrate spin-current-inducedswitching
of a perpendicularly magnetized thulium iron garnet film driven by
charge current in a Pt overlayer. We estimatea relatively large
spin-mixing conductance and damping-like SOT through spin Hall
magnetoresistance and harmonic Hallmeasurements, respectively,
indicating considerable spin transparency at the Pt/MI interface.
We show that spin currentsinjected across this interface lead to
deterministic magnetization reversal at low current densities,
paving the road towardsultralow-dissipation spintronic devices
based on MIs.
Ferromagnetic thin-film heterostructures provide the basisfor
hard disk data storage, magnetic random access memory,magnetic
sensor technology based on giant- and tunnelmagnetoresistance, and,
more recently, prototypes of logicand memory using domain walls.
The ability to manipulatethe magnetic state of thin-film
heterostructures is an essentialattribute of these devices, and it
is especially attractive if it canbe accomplished electrically,
rather than by the applicationof a magnetic field. Compared to
conventional spin transfertorque effects demonstrated in
spin-valves and magnetic tunneljunctions, the current-induced
spin–orbit torque (SOT) in non-magnetic heavy metal/metallic
ferromagnet heterostructures1,2and magnetically doped topological
insulator heterostructuresat low temperatures3,4 provides a highly
efficient way to controlthe magnetization of the ferromagnetic
material. With the SOTmechanism it is possible to realize
magnetization switching5,6,magnetic oscillations7,8, and ultrafast
chiral domain wall motion9,10.Spin currents generated by the spin
Hall effect (SHE)11 in heavymetals (HM) have been shown to be a
dominant source of thedamping-like SOT responsible for the observed
switching anddomain wall motion in such heterostructures.
In the case of magnetic insulators (MIs) such as yttrium
irongarnet (YIG), charge currents cannot flow, but spin
currentsgenerated by the SHE in an adjacent HM layer can be
transmittedacross the HM/MI interface. This has been shown to
giverise to novel spin transport phenomena such as spin
Hallmagnetoresistance (SMR) in HM/MI structures12–15, and used
toexcite high-frequency magnetization dynamics in YIG16,17 drivenby
current-induced spin torques. The low damping and long
spintransmission length (∼1mm) in magnetic insulators such as
YIGmake thempromising candidates for spin-wave communication
andultralow-power-dissipation applications16,18, as well as
insulatingcomponents of magnetic logic or memory devices. However,
SHE-induced control of the magnetization switching in a
magneticinsulator has yet to be reported.
In this work, we demonstrate deterministic
current-inducedswitching in a perpendicularly magnetized MI using
spin torquefrom the SHE in an adjacent Pt layer. The MI used here
consists
of an 8-nm-thick thulium iron garnet (Tm3Fe5O12, TmIG)film with
perpendicular magnetic anisotropy. We organize thisarticle as
follows: first, we report on the structural and
magneticcharacterization of TmIG continuous films with X-ray
diffraction(XRD), vibrating sample magnetometry (VSM),
magneto-opticalKerr effect (MOKE) and magnetic force microscopy
(MFM).Next, we characterize SMR by means of transverse (Hall
effect)measurements in TmIG/Pt bilayers, fromwhich we extract the
spin-mixing conductance G↑↓ that characterizes spin transport
acrossthe interface. Then, we perform harmonic Hall effect
measurementsto accurately quantify the damping-like SOT driven by
the spin Halleffect in Pt. Finally, we show that the
perpendicularmagnetization ofTmIG can be reversed by the
damping-like SOT stemming from theSHE in an adjacent Pt layer with
current densities comparable to orlower than that used to switch
all-metallic perpendicular magneticanisotropy (PMA)
heterostructures5,6, even though the magneticfilm thickness is one
order of magnitude larger. These results are animportant step
towards novelmemory or logic devices based onMIs.
Structural and magnetic characterizationFerrimagnetic TmIG was
chosen because its negativemagnetostriction and negative
magnetocrystalline anisotropycoefficient K1 leads to an
out-of-plane easy axis19 when grownepitaxially under tensile strain
on a (111)-oriented gadoliniumgallium garnet (Gd3Ga5O12, GGG)
substrate. Films with layerthicknesses of 8 nm and 30 nm TmIG were
grown by pulsed laserdeposition (PLD) onGGG (111). The 8 nm filmwas
used in the SOTswitching experiments, whereas the 30 nm film yields
a strongersignal for XRD analysis that gives insight into the
structure andstrain state of TmIG. The crystalline structure and
film thicknesswere investigated by high-resolution XRD and X-ray
reflectivity(XRR), respectively. For the 30-nm-thick film the
symmetric XRDscans around the (444) GGG-substrate and TmIG peaks
(Fig. 1a)demonstrate fully strained TmIG films with the anticipated
Lauediffraction peak shape and Laue fringes. Reciprocal space
mappingof the 30 nm TmIG film provides evidence that the in-plane
latticeparameter of TmIG was coherently strained to match that
ofGGG (111) (see Supplementary Information 1). High-resolution
Department of Materials Science and Engineering, Massachusetts
Institute of Technology, Cambridge, Massachusetts 02139, USA.
†Present address:Department of Materials Science and Engineering,
National Taiwan University, Taipei 10617, Taiwan. *e-mail:
[email protected]; [email protected]
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials 1
© Macmillan Publishers Limited . All rights reserved
http://dx.doi.org/10.1038/nmat4812mailto:[email protected]:[email protected]/naturematerials
-
ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4812
50.0 50.5 51.0 51.5 52.0 52.5 53.0100
101
102
103
104
105
106TmIG 30 nm film
8 nm film
30 nm film8 nm film
30 nm film8 nm film
I (c.
p.s.
)M
OKE
sig
nal (
a.u.
)
2 / (°)ωθ
GGG
−450 −300 −150 0H (Oe)
H (Oe)
150 300 450
−150
−100
−50
0
M (e
.m.u
. cm
−3)
50
100
150
−450 −300 −150 0 150 300 450
−1.0
−0.5
0.0
0.5
1.0
Phas
e sh
ift (°
)
0.5
0
a b
c d
2 μm
Figure 1 | Structural and magnetic properties of PMA TmIG. a,
X-ray di�raction patterns (2θ−ω scan) of a 30-nm-thick (orange) and
a 8-nm-thick(black) TmIG film grown on (111)-oriented GGG
substrates. b, Magnetization of both films as a function of
out-of-plane magnetic field Hz obtained by VSM.A paramagnetic
contribution from the GGG substrate has been subtracted. c, Polar
MOKE signal of films. d, MFM image of the as-grown 30-nm-thickTmIG
film showing up- and down-oriented domains as regions of light and
dark contrast.
transmission electron microscopy of a similar film (not
shownhere) revealed a fully coherent interface with no dislocations
overat least a 1.5 µm distance. Since the thicker TmIG film showed
nostrain relaxation, it is reasonable to assume that the thinner (8
nm)TmIG film is also fully coherent.
Figure 1b,c shows the out-of-plane hysteresis loop for the 8
nmand 30 nm TmIG films measured using VSM and polar
MOKE,respectively. Both films exhibit high out-of-plane
remanence,indicating strong PMA. The coercivity of the 30 nm film
islower than that of the 8 nm film. We tentatively attribute
thisbehaviour to the difference in the domain size, which is
expectedto be proportional to the square root of thickness, leading
to thethickness dependence of the field needed to nucleate and
propagatedomains. The saturation magnetization of both films was
about100 e.m.u. cm−3, approaching the room-temperature bulk value20
ofMs ∼ 110 e.m.u. cm−3. The magnetic domain configuration
wasvisualized in both the as-grown (demagnetized) and
saturatedstates via magnetic force microscopy (MFM). Figure 1d
shows anMFM image for the demagnetized state in the 30 nm film,
whichexhibits a labyrinthine domain structure typical of PMA thin
films.After out-of-plane saturation, the domain structure vanishes
(notshown), consistent with the high remanence of the
correspondinghysteresis loops.
Spin Hall magnetoresistanceSMRmeasurements have been widely used
to extract spin transportparameters in transition HM/MI
bilayers12–15, from which theinterfacial spin-mixing conductance
G↑↓ can be estimated. In suchmeasurements, an electrical current is
applied to a HM layer withsizeable SHE deposited on a MI. The
charge current j in the HMgenerates a transverse spin current
js=θSH(j×σ ) that will either betransmitted or reflected at the
HM/MI interface depending on the
relative orientations between the SHE spin polarization σ and
themagnetization direction m in the MI. Here θSH represents the
spinHall angle (SHA) of theHM. Interfacial spin
transmission/reflectionis characterized by the spin-mixing
conductanceG↑↓=G↑↓r + iG
↑↓
i .The real part G↑↓r is related to the damping-like torque
acting uponm, which is proportional tom× (σ×m). The imaginary
partG↑↓i isassociated with a field-like torque, which is
proportional to σ×m.The dependence of spin transmission/reflection
on m and σ willfurther modulate j in the HM layer due to the
inverse spin Halleffect21. This leads to magnetoresistance in the
HM, the so-calledSMR,which possesses a symmetry distinct from that
of, for example,anisotropic magnetoresistance.
In the SMR scenario, the longitudinal (R) and transverse
Hall(RH) resistance in a HM/MI bilayer can be expressed
as13,15,22
R=R0+1RSMR sin2 θ sin2 ϕ (1)
RH=RSMRH sin2θ sin2ϕ+RAHE,SMRH cosθ+R
OHEH Hz (2)
where R0 and 1RSMR represent the m-independent
longitudinalresistance and the modulation due to the SMR,
respectively. RSMRH ,RAHE,SMRH and R
OHEH represent the transverse manifestation of SMR,
the SMR-induced anomalous Hall effect (AHE) resistance, and
theordinary Hall resistance of the HM, respectively. The
magnetizationangles θ , ϕ are defined in Fig. 2a, as well as the
coordinate systemused throughout the article. We note that the
transverse SMR andthe SMR-induced AHE are identical to the planar
Hall resistanceand anomalous Hall resistance in conducting
ferromagnets bysymmetry.
Figure 2b–d summarizes measurements of transverse resistanceRH
on TmIG(8 nm)/Pt(5 nm) bilayers that reveal SMR in our films
2
© Macmillan Publishers Limited . All rights reserved
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials
http://dx.doi.org/10.1038/nmat4812www.nature.com/naturematerials
-
NATUREMATERIALS DOI: 10.1038/NMAT4812 ARTICLES
z
Pt (5 nm)TmIG (8 nm)
GGG substrate
VH
y
x
m
j
θ
ϕ
−200 −100 0 100 200
−0.6
−0.3
0.0
0.3
0.6
−4,000 −2,000 0 2,000 4,000
−8
−4
0
4
8
0 45 90 135 180
OHE
dc
b
R H (m
Ω)
R H (m
Ω)
−8
−4
0
4
8
R H (m
Ω)
Hz (Oe)
AHE
HIP (Oe)
Simulations
a
Data
FitH = 0°ϕ
H (°)ϕ
H = 45°ϕ
H = 135°ϕ
Figure 2 | SMRmeasurements on a TmIG/Pt bilayer device. a, Hall
device schematics, coordinate systems and the electrical
measurement set-up. b, Hallresistance RH is measured with an
out-of-plane-field Hz and an a.c. current of amplitude 1.5 mA
(r.m.s.), corresponding to j=4.3× 1010 A m−2 in the Ptlayer. RH
tracks the magnetization vector, due to the SMR-induced AHE, which
switches between up and down states at the coercive field ofHc=
110±5 Oe. The linear negative slope arises from the ordinary Hall
e�ect, scaling linearly with Hz and independent of the
magnetization direction.c, RH measured while sweeping an in-plane
field (HIP) at ϕH=0◦, 45◦, 135◦ in the presence of a constant
Hz=+200 Oe to ensure that the magnetizationremains single domain
during measurement. The curves at ϕH=45◦(135◦) are representative
of SMR behaviour where a large positive (negative) signalappears
when m is saturated in-plane. Orange curves are simulations based
on experimental parameters and an estimated perpendicular
magneticanisotropy field HK=2,700 Oe. d, Summary of RH measurements
performed at di�erent ϕH as shown in c. A fit to the data according
to equation (2)reveals RSMRH =6.5 m. Note that a constant
sample-dependent o�set is subtracted from the raw data shown in b
and c.
and allow the associated transport parameters to be
quantified(see Supplementary Information 2 for the longitudinal
SMRmeasurements). The bilayer was patterned intoHall cross devices,
asshown schematically in Fig. 2a, and all electricalmeasurements
wereperformed using a standard lock-in technique (see Methods).
RHversusHz (Fig. 2b) shows 100% remanence and
sharpmagnetizationreversal at a coercive field Hc = 110± 5 Oe. We
attribute thereduction in Hc as compared to the continuous film
(Fig. 1)to geometrical effects due to patterning. The linear
background(green dotted line) arises from the ordinary Hall effect
in Pt,consistent with equation (2). We find RAHE,SMRH =−0.46m
andROHEH =−5.8mT
−1, which yield ρAHE,SMRxy =−2.3× 10−4 µ cm
and ρOHExy =−2.9× 10−3 µ cmT−1, respectively, comparable to
recently reported values in Pt/YIG22. Next, we measure RH as
afunction of in-plane field (HIP) at different field angles ϕH.
Figure 2cshows measurements recorded at ϕH= 0◦, 45◦ and 135◦, where
theSMR contribution to RH is either zero at ϕH = 0◦ or
maximum(minimum) at ϕH=45◦(135◦) according to equation (2). Note
thatwe have applied a fixed out-of-plane field HOOP= 200Oe
duringthe HIP sweep to avoid magnetic domain formation and to
ensurecoherent rotation of the magnetization. By simulating the
SMRwith a macrospin model using experimental parameters (Fig.
2c,orange curves) we obtain a perpendicular magnetic anisotropy
field
HK≈2,700Oe. Finally, by repeating the measurements at variousϕH
we find that RH ∝ sin 2ϕ, as shown in Fig. 2d, as expectedfrom the
SMR. A fit to the data yields RSMRH = 6.5m, which leadsto ρSMRxy
=3.25×10
−3 µ cm. This value is approximately 14 timeslarger than the
SMR-induced AHE, implying that the real partof the spin-mixing
conductance is significantly larger than theimaginary part.
Following the SMR model given in ref. 23 and by usingρSMRxy ,
ρ
AHE,SMRxy obtained above, we can calculate G
↑↓. By takingρPtxx=40.4 µ cm as measured on the same device,
assumingθSH=0.08 for Pt24 and taking the spin diffusion length to
beλPt = 1.4 nm (ref. 23), we estimate G↑↓r = 1.0 × 10
14 −1 m−2and G↑↓i =7.1×10
12−1 m−2, which compare well with commonlyaccepted values for
YIG/Pt13,14.
Determination of damping-like spin–orbit torqueThe large G↑↓r
suggests the possibility of achieving significantspin injection
across the TmIG/Pt interface, and hence exerting adamping-like
torque on the magnetization. To quantify the SHE-induced SOT, we
performed harmonic Hall effect measurementssimilar to those
commonly used in all-metallic HM/ferromagnetbilayers1,25. In such
measurements, an a.c. current inducesoscillations in the
magnetization due to alternating SOT, giving rise
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials
© Macmillan Publishers Limited . All rights reserved
3
http://dx.doi.org/10.1038/nmat4812www.nature.com/naturematerials
-
ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4812
−200 −100 0 100 200
−5.2
−5.0
−4.8
4.8
5.0
5.2
Hy (Oe)
Hy (Oe)
mup
mup
mdown
mdown
−200 −100 0 100 200−0.2
−0.1
0.0
0.1
0.2
−0.01 0.00 0.01−0.2
−0.1
0.0
0.1
0.2c d
b
mupmdown
Linear fit
a
xy
z
m
HFLHDL
j
Pt
TmIG
(2VSMR sin2 )/(Hsin H) (μV Oe−1)θ θH
V (μ
V)
ω
V 2 (μ
V)
ω
V 2 (μ
V)
ω
Figure 3 | Determination of the damping-like SOT in TmIG/Pt by
harmonic measurements. a, Measurement geometry for the harmonic
Hall voltagemeasurements. Blue and red arrows show the e�ective
fields associated with damping-like (HDL) and field-like (HFL) SOT,
respectively, for current flowalong the x-axis and magnetization
tilted along the y-axis. b, First-harmonic Hall voltage (Vω) mainly
driven by the SMR-induced AHE, acquired during anin-plane field
sweep of Hy=±200 Oe with an applied current of jr.m.s.=2.1× 1011 A
m−2. The curvature is di�erent for mup and mdown due to
misalignmentof the external field with respect to the y-axis,
causing a small SMR contribution which is even with respect to
magnetization reversal whereas SMR-inducedAHE is odd. c,
Second-harmonic voltage (V2ω) recorded under the same conditions as
in b. Di�erent slopes are a consequence of a current-induced
spinSeebeck e�ect voltage (see text and Supplementary Information 4
and 5 for more details). d, V2ω plotted versus (2VSMRH sin
2 θ)/(HsinθH) to quantify HDLby evaluating the slope given by
the linear fit (black curves). A constant sample-dependent o�set
has been subtracted from the data presented in b and c.
to a second-harmonic Hall voltage, V2ω. Usually, the
damping-likeand field-like torques are probed by measuring V2ω
while sweepingan in-plane magnetic field that tilts the
magnetization in the zxand zy planes, respectively. Under these
conditions the a.c. currentgenerates out-of-plane magnetization
oscillations in the respectiveplanes, and the resulting AHE
oscillations dominate the V2ω signal.In the present system,
however, the SMR is much larger than theSMR-induced AHE, and hence
V2ω mainly reflects the in-planecomponent of magnetization
oscillations. In this case, the damping-like SOT is expected to
give the dominant contribution to V2ωwhen the magnetization is
tilted in the zy plane, since the field-liketorque drives
out-of-plane oscillations, whose contribution to V2ωscales with the
much smaller SMR-induced AHE. At the same time,thermal
contributions to V2ω from the spin Seebeck effect (SSE)arising from
the Joule-heating-induced perpendicular temperaturegradient scale
as VSSE ∝∇Tz ×m, and are hence minimized inthis configuration26.
Therefore, harmonic Hall measurementsunder a transverse field are
well-suited to quantitatively extract thedamping-like SOT in this
system.
Figure 3b and c shows the first- and second-harmonic
signalswhenm is tilted along±y with a swept field Hy and an a.c.
currentdensity jr.m.s.∼= 2.1× 1011 Am−2. Ideally, the
first-harmonic signalVω arises solely from the SMR-induced AHE in
this geometry(see equation (2)), which decreases with increasing Hy
as m tiltstowards the plane (Vω = V AHE,SMRH cos θ). However,
unintentionalmisalignment of H with respect to the y axis creates
non-negligible
SMR (∝ sin2 θ) and OHE (∝ Hz ) contributions that distort
theexpected quadratic AHE signal, as depicted in Fig. 3b.
We now focus on V2ω which, as shown in Fig. 3c, varies
linearlywith Hy , having opposite slopes for up (mup) and down
(mdown)magnetization states. This symmetry is typical for a SOT
signal andcan be driven by both HFL and HDL according to ref.
1:
V ϕ=90◦
2ω =(V AHE,SMRH −2V
SMRH cosθ sin2ϕ
) dcosθdH
HFLsin (θH−θ)
+2V SMRH sin2θ cos2ϕ
HDLH sinθH
(3)
where θH is the out-of-plane angle of the external field. SinceV
SMRH � V
AHE,SMRH , and assuming HDL≈HFL, the signal in Fig. 3c
must be dominated by the second term on the right-hand
side.Therefore, to estimate the damping-like SOT we plot V2ω asa
function of (2V SMRH sin
2 θ)/(H sinθH) in Fig. 3d (note thatcos2ϕ∼= sinθH∼=1 and sin
2ϕ∼=0 when Hy tilts the samplemagnetization in the zy plane) and
take the slope, which directlyyields HDL in units of Oe. We obtain
−9.4Oe and +15.1Oe formup and mdown, respectively, giving on
average HDL=12.3±2.8 Oeper jr.m.s. ∼= 2.1× 1011 Am−2, with the sign
compatible with theSHE in Pt. The difference in slopes between mup
and mdown is dueto a small x-component creating a non-negligible
VSSE, adding apositive slope for both up and down orientation (see
Supplementary
4
© Macmillan Publishers Limited . All rights reserved
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials
http://dx.doi.org/10.1038/nmat4812www.nature.com/naturematerials
-
NATUREMATERIALS DOI: 10.1038/NMAT4812 ARTICLES
−200 −100 0 100 200
−0.35
0.00
0.35
c d
ba
Hz (Oe)100 200 300 400 500
−0.35
0.00
0.35
−2.34
0.00
2.34
0 300 600 900 1,2000.5
1.0
1.5
2.0
2.5
0.35
0.70
0.00
Hx (Oe)
ΔRH (m
Ω)
Time (ms)
−2 0 2
−1
0
1
j (×1011 A m−2)
j (×1
011 A
m−2
)
R H (m
Ω)
R H (m
Ω)
Hx = +500 Oe
Hx = −500 Oe
j (×1011 A
m−2)
R H/|R Hs
at|
Hx = +500 Oe
Figure 4 | Current-induced magnetization switching of TmIG/Pt.
a, Hall resistance measurements serving as reference for mup and
mdown states(a constant o�set and planar Hall e�ect (PHE)
contribution are subtracted). b, Magnetization switching with
5-ms-long current pulses ofj=±2.34× 1011 A m−2 amplitude in the
presence of an external field of Hx=±500 Oe. The switching polarity
changes on reversing either the currentpolarity or field direction.
c, Switching behaviour of the device as a function of the injected
current density for a given field of Hx=+500 Oe. Above thecritical
current of jc=±1.7× 1011 A m−2 the magnetization switches
systematically and reproducibly. d, Magnetization switching phase
diagram showingthe switching behaviour as a function of applied
current and field. It is constructed with the help of measurements
shown in c performed withHx=0→ 1,200 Oe.
Information 4). By assuming that HDL is entirely driven by
theSHE, we can estimate the SHA via27 θSH= (2e/})(MstPtHDL/j)
usingappropriate material parameters, where e is the elementary
charge,} is the reduced Planck constant,Ms is the
saturationmagnetizationof TmIG and tPt is the Pt layer thickness.
We find θSH≈ 1%, whichconstitutes the lower bound due to current
spread to theHall voltagebranches, effectively reducing the current
density of the central areaof the Hall cross1,28. When this
correction is taken into account weestimate θSH∼1.5–2%, close to
the reported values for Pt in contactwith YIG13.
Current-induced magnetization switchingNext, we show that the
large damping-like SOT revealed byharmonic measurements can be used
to switch the magnetizationof TmIG, as summarized in Fig. 4. First,
we recorded the Hallresistance during a Hz sweep, as shown in Fig.
4a, from whichwe subtracted a sample-dependent constant offset and
the OHEcontribution. This serves as reference for mup and mdown, to
whichwe compare the action of the current pulses. We then
successivelyapplied ±5-ms-long pulses of amplitude j= 2.34× 1011
Am−2 inthe presence of an in-plane field of Hx = ±500 Oe along
thecurrent injection axis. We note that for deterministic
SOT-inducedswitching in PMA materials it is necessary to apply an
in-planeexternal field to break the rotational symmetry of HDL
(ref. 5,6).In Fig. 4b we show the measured Hall resistance after
each currentpulse in the presence of a constant Hx . We observe
that RH changesbetween 80–100%with respect to its full
amplitude—that is, 0.7m,after every pulse. The switching polarity
reverses when reversing
the applied field direction +H x→−H x , as expected from
theSOT-induced switching. Reference measurements performed withthe
identical configuration showed that the switching polarity
iscompatible with that of Pt/Co when the structural inversion
factoris taken into account (Supplementary Information 6).
Figure 4c shows the switching behaviour of the same sample asa
function of j for Hx =+500Oe. Each data point represents theoutcome
of averaging ten measurement sequences where, duringeach sequence,
we measured RH after a given pulse was appliedsynchronously with Hx
, and compared with the reference RHrecorded after positive and
negative out-of-plane saturation fieldpulses (for details of the
measurement protocol see SupplementaryInformation 6). We see that
the switching events are systematicabove the critical current
density of jc= 1.8× 1011 Am−2 for thisspecific Hx . We note that
the switching events above jc saturatearound 90%,meaning that the
probed area of theHall cross does notfully reverse. This behaviour
was reproducible for several devicesand tentatively explained by
partial switching of areas in the vicinityof the Hall cross where j
is relatively lower but still contributes to RH.
To examine the switching behaviour in more detail, we
repeatedthe measurements shown in Fig. 4c for fields between
0–1,200Oe,and constructed the switching probability phase diagram
shown inFig. 4d. Red (blue) colour shows wherem can (cannot) be
switchedbetween up and down states with the combined action of j
andHx , as determined by averaging ten switching events. We
observeseveral interesting features. First, the minimum Hx required
forswitching is as low as several tens of Oe. Second, the
switchingphase boundary exhibits two regimes above and
belowHx=900Oe.
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials
© Macmillan Publishers Limited . All rights reserved
5
http://dx.doi.org/10.1038/nmat4812www.nature.com/naturematerials
-
ARTICLES NATUREMATERIALS DOI: 10.1038/NMAT4812We speculate that
for Hx > 900Oe it becomes comparable to aneffective decrease of
the anisotropy field due to Joule heating,and thus a significant
decrease of the energy necessary to switchm. On the other hand, for
Hx < 900Oe the behaviour is quasi-linear, similar to the data
reported for, for example, Ta/CoFeB29.Finally, we find that
switching is possible for current densities aslow as
j=0.5×1011Am−2, much lower than in all-metallic systems,despite the
relatively small θSH found with harmonic measurementsin the
single-domain state. Since magnetization reversal is mediatedby a
domain nucleation and propagation30, this behaviour suggeststhat
the damping-like SOT acting on domain walls is far moreefficient
than in metallic systems. We also note that Joule heating(see
Supplementary Information 7) probably reduces the coercivity,which
may play a role in reducing the switching current thresholdat
higher current densities.
In summary, we show highly efficient magnetization switchingof
an insulating ferrimagnet with PMA utilizing the spin Halleffect of
an adjacent Pt layer. Spin Hall magnetoresistanceand harmonic
measurements show considerable spin-mixingconductance and
damping-like torque, explaining the observedbehaviour. These
results open up a range of possibilities from bothfundamental and
applied viewpoints. Detection of perpendicularmagnetization in an
insulator via the anomalous Hall effectwith unprecedented precision
makes characterization of magneticdomain wall dynamics electrically
accessible. Furthermore, theharmonic measurement scheme introduced
here makes the SHEcharacterization of metal/insulating systems
feasible and fast.Finally, the deterministic magnetization
switching and its detectionrepresent a significant advance towards
the development of room-temperature spintronic devices based on
magnetic insulators.
MethodsMethods and any associated references are available in
the onlineversion of the paper.
Received 12 April 2016; accepted 12 September 2016;published
online 21 November 2016
References1. Garello, K. et al . Symmetry and magnitude of
spin–orbit torques in
ferromagnetic heterostructures. Nat. Nanotech. 8, 587–593
(2013).2. Brataas, A. & Hals, K. M. D. Spin–orbit torques in
action. Nat. Nanotech. 9,
86–88 (2014).3. Fan, Y. et al . Magnetization switching through
giant spin–orbit torque in a
magnetically doped topological insulator heterostructure. Nat.
Mater. 13,699–704 (2014).
4. Fan, Y. et al . Electric-field control of spin–orbit torque
in a magnetically dopedtopological insulator. Nat. Nanotech. 11,
352–360 (2016).
5. Miron, I. M. et al . Perpendicular switching of a single
ferromagnetic layerinduced by in-plane current injection. Nature
476, 189–193 (2011).
6. Liu, L. Q. et al . Spin-torque switching with the giant spin
Hall effect oftantalum. Science 336, 555–558 (2012).
7. Demidov, V. E. et al . Magnetic nano-oscillator driven by
pure spin current.Nat. Mater. 11, 1028–1031 (2012).
8. Liu, L. Q., Pai, C-F., Ralph, D. C. & Buhrman, R. A.
Magnetic oscillationsdriven by the spin Hall effect in 3-terminal
magnetic tunnel junction devices.Phys. Rev. Lett. 109, 186602
(2012).
9. Emori, S., Bauer, U., Ahn, S. M., Martinez, E. & Beach,
G. S. D. Current-drivendynamics of chiral ferromagnetic domain
walls.Nat. Mater. 12, 611–616 (2013).
10. Ryu, K. S., Thomas, L., Yang, S. H. & Parkin, S. Chiral
spin torque at magneticdomain walls. Nat. Nanotech. 8, 527–533
(2013).
11. Sinova, J., Valenzuela, S. O., Wunderlich, J., Back, C. H.
& Jungwirth, T. SpinHall effects. Rev. Mod. Phys. 87, 1213–1259
(2015).
12. Nakayama, H. et al . Spin Hall magnetoresistance induced by
a nonequilibriumproximity effect. Phys. Rev. Lett. 110, 206601
(2013).
13. Hahn, C. et al . Comparative measurements of inverse spin
Hall effects andmagnetoresistance in YIG/Pt and YIG/Ta. Phys. Rev.
B 87, 174417 (2013).
14. Weiler, M. et al . Experimental test of the spin mixing
interface conductivityconcept. Phys. Rev. Lett. 111, 176601
(2013).
15. Vlietstra, N., Shan, J., Castel, V., van Wees, B. J. &
Ben Youssef, J. Spin-Hallmagnetoresistance in platinum on yttrium
iron garnet: dependence onplatinum thickness and
in-plane/out-of-plane magnetization. Phys. Rev. B 87,184421
(2013).
16. Kajiwara, Y. et al . Transmission of electrical signals by
spin-waveinterconversion in a magnetic insulator. Nature 464,
262–267 (2010).
17. Hamadeh, A. et al . Full control of the spin-wave damping in
a magneticinsulator using spin–orbit torque. Phys. Rev. Lett. 113,
197203 (2014).
18. Chumak, A. V., Vasyuchka, V. I. I., Serga, A. A. A. &
Hillebrands, B. Magnonspintronics. Nat. Phys. 11, 453–461
(2015).
19. Kubota, M. et al . Stress-induced perpendicular
magnetization in epitaxial irongarnet thin films. Appl. Phys. Exp.
5, 103002 (2012).
20. Paoletti, A. Physics of Magnetic Garnets (North-Holland
Publishing Company,1978).
21. Saitoh, E., Ueda, M., Miyajima, H. & Tatara, G.
Conversion of spin current intocharge current at room temperature:
inverse spin-Hall effect. Appl. Phys. Lett.88, 182509 (2006).
22. Meyer, S. et al . Anomalous Hall effect in YIG vertical bar
Pt bilayers.Appl. Phys. Lett. 106, 132402 (2015).
23. Chen, Y. T. et al . Theory of spin Hall magnetoresistance.
Phys. Rev. B 87,144411 (2013).
24. Liu, L. Q., Moriyama, T., Ralph, D. C. & Buhrman, R. A.
Spin-torqueferromagnetic resonance induced by the spin Hall effect.
Phys. Rev. Lett. 106,36601 (2011).
25. Kim, J. et al . Layer thickness dependence of the
current-induced effective fieldvector in Ta vertical bar CoFeB
vertical bar MgO. Nat. Mater. 12, 240–245(2013).
26. Schreier, M. et al . Current heating induced spin Seebeck
effect. Appl. Phys. Lett.103, 242404 (2013).
27. Liu, L. Q., Lee, O. J., Gudmundsen, T. J., Ralph, D. C.
& Buhrman, R. A.Current-induced switching of perpendicularly
magnetized magnetic layersusing spin torque from the spin Hall
effect. Phys. Rev. Lett. 109, 96602 (2012).
28. Ibrahim, I., Schweigert, V. & Peeters, F. Diffusive
transport in a Hall junctionwith a microinhomogeneous magnetic
field. Phys. Rev. B 57,15416–15427 (1998).
29. Avci, C. O. et al . Fieldlike and antidamping spin-orbit
torques in as-grown andannealed Ta/CoFeB/MgO layers. Phys. Rev. B
89, 214419 (2014).
30. Lee, O. J. et al . Central role of domain wall depinning for
perpendicularmagnetization switching driven by spin torque from the
spin Hall effect. Phys.Rev. B 89, 024418 (2013).
AcknowledgementsThe authors would like to acknowledge support
from C-SPIN, one of the six SRCSTARnet Centers, sponsored by MARCO
and DARPA. A.Q. acknowledges funding fromthe Deutsche
Forschungsgemeinschaft (DFG, German Research Foundation) and
fromthe Max-Planck-Institute of Microstructure Physics. C.O.A. and
C.-F.P. thank K. Uedaand A. J. Tan for fruitful discussions.
Author contributionsG.S.D.B. and C.A.R. proposed and supervised
the study. C.O.A., M.M., C.-F.P. andG.S.D.B. designed the transport
experiments. A.Q., A.S.T. and M.C.O. fabricated theTmIG samples.
A.Q. performed structural and magnetic analysis. M.M. carried
outphotolithography processing. C.O.A., M.M. and A.Q. carried out
transportmeasurements. M.M. and L.C. designed and established the
electrical measurementequipment.
Additional informationSupplementary information is available in
the online version of the paper. Reprints andpermissions
information is available online at
www.nature.com/reprints.Correspondence and requests for materials
should be addressed to C.A.R. or G.S.D.B.
Competing financial interestsThe authors declare no competing
financial interests.
6
© Macmillan Publishers Limited . All rights reserved
NATUREMATERIALS | ADVANCE ONLINE PUBLICATION |
www.nature.com/naturematerials
http://dx.doi.org/10.1038/nmat4812http://dx.doi.org/10.1038/nmat4812http://dx.doi.org/10.1038/nmat4812http://dx.doi.org/10.1038/nmat4812http://www.nature.com/reprintswww.nature.com/naturematerials
-
NATUREMATERIALS DOI: 10.1038/NMAT4812 ARTICLESMethodsMaterial
growth. A stoichiometric TmIG target was fabricated using
high-puritypowders of the respective binary oxides, which were
sintered and cold-pressed intoa one-inch diameter target. The
material was deposited onto 1 cm-by-1 cmGGG(111) substrates via PLD
at a laser wavelength of 248 nm, a fluence of∼1.5 J cm−2, and a
target-to-substrate distance of 5 cm. During deposition,
thesubstrate temperature was 810 ◦C and the oxygen pressure was
200mtorr. Nopost-deposition annealing process was performed, but
the cooling step to roomtemperature at 200mtorr oxygen pressure
occurred at−5 ◦Cmin−1.
Device fabrication. After the deposition of the TmIG(8 nm) thin
film via PLD, a5-nm-thick layer of the spin Hall metal Pt was
deposited by sputtering. The basepressure was 9×10−7 torr, and the
deposition rate was 1.5 nmmin−1. Hall crosseswith lateral
dimensions of 7 µm× 6 µm were patterned via a top-down ion
millingprocess after using a standard photolithography method to
define the mesastructures. Ta(4 nm)/Au(100 nm) were deposited as
contact pads for electricalmeasurements.
Magnetic characterization. In-plane and out-of-plane hysteresis
loopswere measured using VSM, and the out-of-plane loop shape was
verified using
a polar MOKE magnetometer. A paramagnetic background from
thesubstrate was subtracted from the VSM data shown in Fig. 1b.
TheMFMmeasurements were conducted within a scan area of 10 µm2
usinga magnetic Co/Cr-coated antimony-doped Si tip at a lift
heightof 20 nm.
Electrical measurements.Hall effect measurements reported in
Fig. 2 wereperformed by injecting an a.c. voltage of Vr.m.s.=0.5V
with frequencyω/2π=3,678Hz, giving rise to a current of amplitude
Ir.m.s.=1.5mA, using astandard lock-in amplifier. A typical device
resistance was Rdevice≈300, to whicha capacitor from a bias-T (1
µF) and a resistor (50) were connected in series. Themeasured a.c.
Hall voltage was averaged over several scans and was converted to
aHall resistance using RH= (Vω/I). To generate measurable torques
and perform theharmonic measurements shown in Fig. 3, we increased
the voltage to Vr.m.s.=3Vand recorded the second-harmonic Hall
voltage. Each curve shown in Fig. 3b,cwas averaged over 200 loops
with a field sweep rate of 0.4Hz. To perform theswitching
measurements reported in Fig. 4 we have used a bias-T (1 µF–1.05 k)
todecouple the output from the pulse generator and the lock-in
amplifier. Pulses weregenerated by amplifying a digital-to-analog
converter output with a Crown DC150audio amplifier.
NATUREMATERIALS | www.nature.com/naturematerials
© Macmillan Publishers Limited . All rights reserved
http://dx.doi.org/10.1038/nmat4812www.nature.com/naturematerials
Current-induced switching in a magnetic insulatorStructural and
magnetic characterizationSpin Hall magnetoresistanceDetermination
of damping-like spin–orbit torqueCurrent-induced magnetization
switchingMethodsFigure 1 Structural and magnetic properties of PMA
TmIG.Figure 2 SMR measurements on a TmIG/Pt bilayer device.Figure 3
Determination of the damping-like SOT in TmIG/Pt by harmonic
measurements.Figure 4 Current-induced magnetization switching of
TmIG/Pt.ReferencesAcknowledgementsAuthor contributionsAdditional
informationCompeting financial interestsMethodsMaterial
growth.Device fabrication.Magnetic characterization.Electrical
measurements.