Voltage control of magnetization via strain in piezoelectric/magnetostrictive systems is a promising mechanism to implement energy-efficient straintronic memory devices. Here, we demonstrate giant voltage manipulation of MgO magnetic tunnel junctions (MTJ) on a Pb(Mg1/3Nb2/3)0.7Ti0.3O3 piezoelectric substrate with (001) orientation. It is found that the magnetic easy axis, switching field, and the tunnel magnetoresistance (TMR) of the MTJ can be efficiently controlled by strain from the underlying piezoelectric layer upon the application of a gate voltage. Repeatable voltage controlled MTJ toggling between high/low-resistance states is demonstrated. More importantly, instead of relying on the intrinsic anisotropy of the piezoelectric substrate to generate the required strain, we utilize anisotropic strain produced using a local gating scheme, which is scalable and amenable to practical memory applications. Additionally, the adoption of crystalline MgO-based MTJ on piezoelectric layer lends itself to high TMR in the strain-mediated MRAM devices.

Information storage technology is constantly challenged by an increasing demand for storage units that are small, retain information for the longest time, and dissipate miniscule amount of energy to store (write) and retrieve (read) information. Magnetic random access memory (MRAM) meets these requirements to a large extent and has been proposed as a universal storage device for computer memory.1–3 In MRAM technology, magnetic tunneling junctions (MTJ) comprise the main storage cells. Low-energy writing of bits requires an electrically tunable mechanism to reorient the magnetization of the MTJ. However, the widely studied switching mechanisms based on utilizing current induced spin-transfer-torques (STT)4,5 or spin-orbit-torques (SOT)6–8 incur high energy dissipation because of the relatively large writing current density.9,10 In recent years, several mechanisms based on using voltage to control magnetization have emerged as promising routes for ultra-low power writing of data.11–15 Among these approaches, the strain induced control of the magnetic anisotropy in multiferroic heterostructures (a magnetostrictive layer elastically coupled with an underlying piezoelectric layer) stands out as a remarkably energy-efficient switching mechanism.16–21 It has been widely investigated in various piezoelectric/ferromagnetic bilayer thin films22–26 or nano-structures.27–30 There are also several theoretical predications31–33 that such a method will dissipate only a few atto-Joules (aJ) of energy to write data. This establishes the promise of using strain to control the resistance of an MTJ for ultra-energy-efficient memory applications.

The key for strain control of the in-plane magnetization is that the in-plane strain should be anisotropic. In most of the previous reports,24–27,34 single crystalline piezoelectric substrates Pb(Mg1/3Nb2/3)0.7Ti0.3O3 (PMN-PT) with (011) orientation were utilized to generate an intrinsic anisotropic strain. However, for realistic strain-mediated MRAM, MTJ would be grown on top of a layer of polycrystalline piezoelectric thin film deposited on a traditional Si substrate for compatibility with silicon technology.19,32,35,36 In that case, one can no longer rely on the intrinsic anisotropy of the piezoelectric material to generate the required anisotropic strain. Moreover, the integration of piezoelectric layer with MTJ stack requires a practical gating scheme to achieve high scalability, low energy dissipation, and individual control, which is lacking so far.

In this paper, we demonstrate giant voltage manipulation of an MgO MTJ on PMN-PT substrate with (001) orientation. Two local gating configurations are applied to produce strong anisotropic strain from the isotropic piezoelectric layer for MTJ control. It is found that the magnetic easy axis, as well as the switching field (Hc) and the tunnel magnetoresistance (TMR) of the MTJ, can be efficiently controlled by strain from the underlying PMN-PT substrate generated by a gate voltage. Magnetic anisotropy can be induced either along the easy axis of the MTJ, resulting in an increase of Hc by more than 4 times, or along the hard axis of the MTJ, leading to a 90° rotation of the magnetization. Moreover, we demonstrate voltage controlled toggling of MTJ resistance between high- and low-resistance states. Our work is fundamentally different from the previous one by Li et al.34 Instead of relying on the intrinsic anisotropy of the piezoelectric substrate (which is not practical), our device utilizes the anisotropic strain generated via local gating schemes and is more amenable to practical memory applications.35,36 Moreover, the localized strain allows the control of an individual MTJ with a relatively small voltage, thus enabling scalability and overcoming the substrate clamping issue.36 The adoption of crystalline MgO as the spacer layer results in high-TMR straintronic MRAM devices. Most importantly, the side-gated MTJ prototype paves the way to realizing complete magnetization “reversal”, i.e., 180° rotation of the magnetization with a voltage,19 which is the ultimate goal of strain-based magnetization manipulation.34,37

A schematic of the strained-MTJ is shown in Fig. 1(a), illustrating the locations of the side and back gates for generation of the localized anisotropic strains. The structure of the MTJ stack is, from bottom to top, Ta(8)/Co20Fe60B20(10)/MgO(1.8)/Co20Fe60B20(4)/Ta(8) (all thicknesses are in nm) grown on a PMN-PT (001) substrate. The MTJ pillar is elliptical in shape (8 μm × 3 μm) with its easy axis (major axis) along the y-direction. It is located between a pair of side gates on the top side of the substrate (as shown in the optical image in Fig. 1(b)). The back side of the substrate is contacted to form a common back gate. Using the side and back gates, one can apply an electric field (E-field) across the PMN-PT substrate to generate a broad range of strain profile. The separation between the two side gates is 40 μm to ensure their electrical isolation from the MTJ.

FIG. 1.

(a) Schematic of the strained-MTJ device. (b) An optical micrograph image of the actual fabricated device. (c) In-plane strains in a bare PMN-PT (001) substrate as a function of applied electric field (voltage). The vertical jumps at the maximum E-fields result from the settling of the sample at those points for 10 min. (d) Magnetic hysteresis loop of patterned MTJ films and the MR loop of MTJ device on PMN-PT without gate voltage application. The magnetic field is along the major axis of the pillars in the y-direction.

FIG. 1.

(a) Schematic of the strained-MTJ device. (b) An optical micrograph image of the actual fabricated device. (c) In-plane strains in a bare PMN-PT (001) substrate as a function of applied electric field (voltage). The vertical jumps at the maximum E-fields result from the settling of the sample at those points for 10 min. (d) Magnetic hysteresis loop of patterned MTJ films and the MR loop of MTJ device on PMN-PT without gate voltage application. The magnetic field is along the major axis of the pillars in the y-direction.

Close modal

The piezoelectric behavior of a bare PMN-PT (001) substrate is shown in Fig. 1(c), where the in-plane strain is plotted as a function of the out-of-plane electric field, measured with a general purpose 120 Ω Constantan linear foil strain gauge (EA-06-062ED-120, Vishay Precision Group, Micro-Measurements). The strain curve under bipolar E-field poling from −8 kV/cm to +8 kV/cm (solid line) exhibits typical butterfly-like behavior and the curve under E-field with a smaller range (dashed line) exhibits almost linear behavior with a very small hysteresis.38 The magnetic hysteresis (M-H) loop of patterned MTJ films is obtained using vibrating sample magnetometry (VSM), as shown in Fig. 1(d), indicating that the thicker CoFeB layer is magnetically harder (with a larger coercivity), while the thinner layer is softer (with a smaller coercivity). The magnetoresistance (MR) loop of the MTJ device is also presented in Fig. 1(d), with zero gate voltage applied and the magnetic field swept along the major axis (y-axis). A post-annealing process at 250 °C was performed for 1 h to increase the TMR ratio of the MTJ. Since neither of the CoFeB layers in the MTJ is pinned by an anti-ferromagnet, the magnetic anisotropy of both layers can be affected by the strain. We can assume that the strains exerted on the soft layer and those exerted on the hard layer are very close to each other, since the strain relaxation between the layers is negligible in our devices. This is confirmed by making a second sample with the positions of the hard layer and the soft layer interchanged (so that the soft layer is closer to the piezoelectric substrate); the results of which is shown in the supplementary material.

In this study, we present our experimental results for two different gating scenarios: a gate voltage Vg is applied either between the back gate and the bottom electrode of the MTJ (Configuration I in Fig. 2) or between the back gate and a pair of side gates (Configuration II in Fig. 3). In both cases, an anisotropic strain is produced, which is highly localized in the MTJ region as illustrated in Figs. 2(b) and 3(b). However, the direction of the strain profile is opposite in the two gating scenarios. Note that in our experiments, positive Vg corresponds to the E-field being parallel to the piezoelectric polarization (poling) direction and negative Vg corresponds to the E-field being anti-parallel to the poling direction.

FIG. 2.

Results for strained-MTJ in Configuration I. (a) Schematic of Configuration I. (b) Simulation result showing the mapping of the in-plane anisotropic strain εxxεyy upon application of the gate voltage of Vg=+50V. The solid line ellipse at the origin denotes the MTJ pillar and the dashed lines denote the positions of electrodes and side gates. (c) Experimental magnetoresistance (MR) curves characterized under different gate voltages. (d) Measured variation of the switching field (square-line) and TMR ratio (circle-line) of the MTJ as a function of Vg.

FIG. 2.

Results for strained-MTJ in Configuration I. (a) Schematic of Configuration I. (b) Simulation result showing the mapping of the in-plane anisotropic strain εxxεyy upon application of the gate voltage of Vg=+50V. The solid line ellipse at the origin denotes the MTJ pillar and the dashed lines denote the positions of electrodes and side gates. (c) Experimental magnetoresistance (MR) curves characterized under different gate voltages. (d) Measured variation of the switching field (square-line) and TMR ratio (circle-line) of the MTJ as a function of Vg.

Close modal
FIG. 3.

Results for strained-MTJ in Configuration II. (a)–(d) are similar to those in Fig. 2.

FIG. 3.

Results for strained-MTJ in Configuration II. (a)–(d) are similar to those in Fig. 2.

Close modal

First, we study the gating effect on MTJ in Configuration I (Fig. 2(a)). With the magnetic field swept along the y-axis, the MR loops under three different gate voltages Vg=150V, 0, and +150V are presented in Fig. 2(c) (see supplementary material for more data). At Vg=0, a normal MR loop similar to that in Fig. 1(d) is obtained with sharp transitions between high- and low-resistance states. However, when a negative gate voltage Vg=150V is applied, the sharp transitions in the MR loop change to gradual slopes, indicating that the easy axes of both magnetic layers have rotated towards the transverse direction (x-direction). On the other hand, when Vg is positive, the switching field increases significantly upon increasing the gate voltage suggesting enhancement of the magnetic anisotropy along the major axis (y-axis). The variation of the switching field of the hard CoFeB layer versus gate voltage is plotted in Fig. 2(d). It can be seen that Hc increases almost linearly and becomes more than 4-fold larger when Vg is increased from 0 to +150 V. In order to have a better understanding of how the magnetization of the ferromagnetic layers is affected by the gate voltage in our devices, 3D piezoelectric finite element simulations were performed using the COMSOL Multiphysics package. The simulated piezoelectric strain mapping on the top surface of the substrate is presented in Fig. 2(b) with Vg=+50V. A positive E-field applied in the out-of-plane direction produces an out-of-plane expansion (d33) and in-plane contraction (d31) in the substrate. Therefore, upon application of a positive Vg, an in-plane bi-axial strain is generated in the region beneath the stripe-shaped electrode where the voltage is applied, and the strain is compressive (negative) in both x- and y-directions. Since the electrode is long in the y-direction and narrow in the x-direction, the strain component εxx along the x-direction is dominant, resulting in an anisotropic strain on the MTJ. We define the in-plane anisotropic strain as εxxεyy. From the simulation, a strain of εxxεyy=274ppm is produced on the MTJ at Vg=+50V. Such an anisotropic strain compresses the MTJ along the x-direction. The strain induced magnetic anisotropy in the ferromagnetic layers can be expressed as Kme=32λσ, where λ represents the magnetostriction coefficient and σ=(εxxεyy)Y represents the stress with Y being the Young's modulus.36,39 Considering λ>0 for CoFeB,40 the negative εxxεyy increases the magnetic anisotropy along the y-direction. As a result, the MR loops in Fig. 2(c) are significantly broadened with positive Vg. On the other hand, with negative Vg,εxx is dominant over εyy with a positive value (tensile), i.e., εxxεyy>0. In this case, magnetic anisotropy is induced along the x-direction and the easy axis of the MTJ rotates by 90°, indicated by the slanted MR loop in Fig. 2(c) at Vg=150V.

In addition to tuning the switching field Hc, Vg also changes the TMR ratio, as shown in Figs. 2(c) and 2(d). It increases from 90% at Vg=0 to 95% at Vg=+150V. We believe that there are two main factors that contribute to it. One is that the strain makes the magnetizations in the soft layer and hard layer align better along the easy axis (y-axis) due to the enhancement of the magnetic anisotropy when the gate voltage is positive. The other is the modification of the MgO tunnel barrier by the strain since the quantum transport properties of the MTJ could be significantly changed by even a small stretching/squeezing of the crystalline lattice of MgO.41 

Next, the voltage controlling is investigated in Configuration II, where the MTJ is flanked by a pair of side gates (Fig. 3(a)). In this scheme, the E-field is generated directly underneath the two side gates. The simulated strain mapping is presented in Fig. 3(b). As one can see, when a positive Vg=+50V is applied, the strain fields are formed due to the out-of-plane expansion and in-plane contraction of the region underneath the side gates. In the central gap between the pair of side gates, a strong anisotropic strain (εxxεyy>0) is produced with a tensile component εxx and a compressive component εyy, resulting from the interaction of the strain fields under the side gates.36 In this case, the sign of εxxεyy exerted on the MTJ is opposite to that of Configuration I. Hence, the modification of the behavior of MTJ by the gate voltage (Figs. 3(c) and 3(d)) is opposite to that of Configuration I (Figs. 2(c) and 2(d)), as expected. A gate voltage of Vg<0 results in εxxεyy<0; therefore, the magnetic anisotropy of CoFeB layers is enhanced along the y-axis and the switching field is increased by ∼4-fold from 25 Oe (Vg=0) to 95 Oe (Vg=150V). Similarly, Vg>0 leads to εxxεyy>0 and consequently the MR loop becomes slanted, due to the induced magnetic anisotropy along the x-axis. Moreover, the TMR ratio slightly increases upon the application of a negative gate voltage as shown in Fig. 3(d).

Finally, we have demonstrated strain-induced toggling of the MTJ resistance by applying gate voltage pulses of ±80 V (Fig. 4(d)). This experiment is performed with Configuration II, and the variation of MR loop from Vg=80V to Vg=+80V (as shown in Fig. 4(c)) is consistent with the result in Fig. 3(c). A micromagnetic simulation has been performed (Figs. 4(a) and 4(b)) utilizing the Object Oriented Micro Magnetic Framework (OOMMF)42 to help understand the MTJ toggling. At Vg=80V (Fig. 4(a)), the magnetizations of both hard and soft CoFeB layers in the MTJ become parallel along the y-axis (with a small bias field H = 30 Oe applied along +y-direction to overcome any dipole interaction), leading to the low-resistance state denoted by the blue arrow in Fig. 4(c). Once Vg changes to +80 V, magnetizations of the soft layer and the hard layer rotate towards the ±x-directions (i.e., opposite directions) because of the strain generated (which overcomes both shape anisotropy and the bias magnetic field). They rotate in opposite directions because of the magnetostatic dipole coupling between the layers, which favors their anti-parallel alignment (Fig. 4(b)). This increase in the angular separation between the magnetizations of the two layers results in a high-resistance state for the MTJ. When the voltage is switched back to −80 V, the magnetizations of the two layers again become parallel along the +y-direction because of the bias magnetic field, and the MTJ resistance drops. Therefore, by alternative application of the gate voltages of +80 V and −80 V, the MTJ cell can be toggled between high (anti-parallel) and low (parallel) resistance states.

FIG. 4.

Demonstration of voltage manipulation of MTJ toggling (in Configuration II). (a) and (b) Micromagnetic simulation results demonstrating the magnetization configuration of hard and soft CoFeB layers after application of (a) Vg=80V and (b) Vg=+80V. The dimension of the magnet is 3 μm × 6 μm. Black arrows indicate the direction of magnetic moments. (c) MR loops for Vg=80V and Vg=+80V. The blue arrow indicates the switchable high- and low-resistance states. (d) Toggling of the MTJ between high- and low-resistance states with the application of ±80 V gate voltage pulsing. A small bias magnetic field of 30 Oe is applied along the +y-axis to overcome the dipole interaction between the two magnetic layers.

FIG. 4.

Demonstration of voltage manipulation of MTJ toggling (in Configuration II). (a) and (b) Micromagnetic simulation results demonstrating the magnetization configuration of hard and soft CoFeB layers after application of (a) Vg=80V and (b) Vg=+80V. The dimension of the magnet is 3 μm × 6 μm. Black arrows indicate the direction of magnetic moments. (c) MR loops for Vg=80V and Vg=+80V. The blue arrow indicates the switchable high- and low-resistance states. (d) Toggling of the MTJ between high- and low-resistance states with the application of ±80 V gate voltage pulsing. A small bias magnetic field of 30 Oe is applied along the +y-axis to overcome the dipole interaction between the two magnetic layers.

Close modal

It should be noticed that although both the soft and hard layer get altered with Vg=+80V in Fig. 4, the rotation of the soft layer is easier because of its smaller shape anisotropy. The soft layer is thinner and thus has larger out-of-plane demagnetization factor Ndz. Hence, the shape anisotropy field (which is proportional to the difference in the two in-plane demagnetization factors, NdxNdy) is smaller in the soft layer than in the hard layer. Consequently, if the two layers see comparable levels of stress (in other words, the same stress anisotropy field), the stress anisotropy is able to beat the shape anisotropy more effectively in the soft layer than in the hard layer.

There are several advantages of using our local gating scheme compared to previous approaches.34,39 A tunable anisotropic in-plane strain can be generated in an isotropic piezoelectric material with this local gating scheme. Hence, having a piezoelectric single crystalline substrate with a specific orientation (like PMN-PT (011)) is no longer a necessity to provide an anisotropic strain profile. Moreover, having a highly localized strain in the MTJ region helps to overcome the substrate clamping issue for next generation strain-MRAM devices using a piezoelectric thin film (like PZT) grown on a Si substrate.35,36 Additionally, it has been predicted that with one more pair of side gates, a deterministic 180° reversal of the magnetization can be achieved.19,32 The side-gated MTJ demonstrated in our work paves the way towards this fully strain-induced MTJ switching in a double gating device.

One of the key advantages of our proposed device is the reduction of the operating voltage. The typical value of E-field required for 90° rotation of magnetization is ∼8 kV/cm,23,25,26,34 corresponding to a 400 V voltage applied across the 0.5 mm-thick substrate. In our experiments, however, the switching voltage Vg reduces to around 100 V, owing to the concentration of the E-field in the local gating scheme (see supplementary material). Had we used a piezoelectric thin film of ∼100 nm thickness deposited on a Si substrate as opposed to a 0.5 mm thick substrate used here, the gate voltages Vg would have been reduced by a factor of roughly 5000 to about 20 mV. Although there are no reports of switching the magnetization of nanomagnets on a 100 nm piezoelectric film, there is a recent report of controlling the states of nanomagnets on a 1000 nm thick piezoelectric film deposited on Si substrate.30 In such a clamped thin film, the piezoelectric coefficient dropped by 40%. If we assume an 80% drop in the piezoelectric coefficient in a 100 nm thin film, then the gate voltage will increase five-fold to 100 mV. The gate capacitance C has been estimated in previous works to be about 2 fF depending on the dimensions of the electrodes.19,32 Hence, the energy dissipated to toggle the MTJ resistance would have been CVg2 = 20 aJ, which would make our proposal the lowest energy writing scheme existent.

In summary, we have demonstrated a giant voltage manipulation of CoFeB/MgO/CoFeB MTJ deposited on PMN-PT (001) substrate by using local gating scheme for strain generation. The generated strain is anisotropic and highly localized in the MTJ region which is also confirmed by simulation results. Application of tensile strain along the easy axis (and/or compressive strain along the hard axis) increases the magnetic anisotropy resulting in a significant increase (by a factor of 4) in the switching field of the MTJ. Application of strains of the opposite sign decreases, and ultimately overcomes, the magnetic shape anisotropy, causing a 90° rotation of the magnetization away from the easy axis. Thus, by applying a voltage of alternating sign, which generates strains of alternating signs, we were able to toggle an MTJ between high/low-resistance states. The demonstration of highly effective voltage manipulation of MTJ via localized strains paves the way towards deterministic 180° MTJ switching and represents a key step towards realizing realistic strain-based MRAM with write energy of few tens of aJ/bit provided the piezoelectric properties scale to 100 nm thickness.

See supplementary material for the details of the piezoelectric finite element simulations, additional experimental data, and discussions about the scalability of the device.

This work was partially supported by the Center for Spintronic Materials, Interfaces and Novel Architectures (C-SPIN), one of six SRC STARnet Centers. N. D'Souza and J. Atulasimha were partially supported by NSF CAREER Grant No. CCF-1253370. S. Bandyopadhyay was supported by the NSF Grant No. ECCS 1124714.

2.
J.
DeBrosse
,
D.
Gogl
,
A.
Bette
,
H.
Hoenigschmid
,
R.
Robertazzi
,
C.
Arndt
,
D.
Braun
,
D.
Casarotto
,
R.
Havreluk
,
S.
Lammers
,
W.
Obermaier
,
W. R.
Reohr
,
H.
Viehmann
,
W. J.
Gallagher
, and
G.
Müller
,
IEEE J. Solid-State Circuits
39
,
678
(
2004
).
3.
A. D.
Kent
and
D. C.
Worledge
,
Nat. Nanotechnol.
10
,
187
(
2015
).
4.
D. C.
Ralph
and
M. D.
Stiles
,
J. Magn. Magn. Mater.
320
,
1190
(
2008
).
5.
A.
Brataas
,
A. D.
Kent
, and
H.
Ohno
,
Nat. Mater.
11
,
372
(
2012
).
6.
L.
Liu
,
C.-F.
Pai
,
Y.
Li
,
H. W.
Tseng
,
D. C.
Ralph
, and
R. A.
Buhrman
,
Science
336
,
555
(
2012
).
7.
Z.
Zhao
,
M.
Jamali
,
A. K.
Smith
, and
J.-P.
Wang
,
Appl. Phys. Lett.
106
,
132404
(
2015
).
8.
A. R.
Mellnik
,
J. S.
Lee
,
A.
Richardella
,
J. L.
Grab
,
P. J.
Mintun
,
M. H.
Fischer
,
A.
Vaezi
,
A.
Manchon
,
E.-A.
Kim
,
N.
Samarth
, and
D. C.
Ralph
,
Nature
511
,
449
(
2014
).
9.
M.
Gajek
,
J. J.
Nowak
,
J. Z.
Sun
,
P. L.
Trouilloud
,
E. J.
O'Sullivan
,
D. W.
Abraham
,
M. C.
Gaidis
,
G.
Hu
,
S.
Brown
,
Y.
Zhu
,
R. P.
Robertazzi
,
W. J.
Gallagher
, and
D. C.
Worledge
,
Appl. Phys. Lett.
100
,
132408
(
2012
).
10.
M.
Cubukcu
,
O.
Boulle
,
M.
Drouard
,
K.
Garello
,
C.
Onur Avci
,
I. M.
Miron
,
J.
Langer
,
B.
Ocker
,
P.
Gambardella
, and
G.
Gaudin
,
Appl. Phys. Lett.
104
,
042406
(
2014
).
11.
J.
Allibe
,
S.
Fusil
,
K.
Bouzehouane
,
C.
Daumont
,
D.
Sando
,
E.
Jacquet
,
C.
Deranlot
,
M.
Bibes
, and
A.
Barthélémy
,
Nano Lett.
12
,
1141
(
2012
).
12.
X.
He
,
Y.
Wang
,
N.
Wu
,
A. N.
Caruso
,
E.
Vescovo
,
K. D.
Belashchenko
,
P. A.
Dowben
, and
C.
Binek
,
Nat. Mater.
9
,
579
(
2010
).
13.
W.-G.
Wang
,
M.
Li
,
S.
Hageman
, and
C. L.
Chien
,
Nat. Mater.
11
,
64
(
2012
).
14.
P. K.
Amiri
and
K. L.
Wang
,
Spin
02
,
1240002
(
2012
).
15.
F.
Matsukura
,
Y.
Tokura
, and
H.
Ohno
,
Nat. Nanotechnol.
10
,
209
(
2015
).
16.
J.-M.
Hu
,
Z.
Li
,
L.-Q.
Chen
, and
C.-W.
Nan
,
Nat. Commun.
2
,
553
(
2011
).
17.
M.
Ghidini
,
R.
Pellicelli
,
J. L.
Prieto
,
X.
Moya
,
J.
Soussi
,
J.
Briscoe
,
S.
Dunn
, and
N. D.
Mathur
,
Nat. Commun.
4
,
1453
(
2013
).
18.
A. K.
Biswas
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
Appl. Phys. Lett.
103
,
232401
(
2013
).
19.
A. K.
Biswas
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
Appl. Phys. Lett.
105
,
072408
(
2014
).
20.
A.
Khan
,
D. E.
Nikonov
,
S.
Manipatruni
,
T.
Ghani
, and
I. A.
Young
,
Appl. Phys. Lett.
104
,
262407
(
2014
).
21.
N. A.
Pertsev
and
H.
Kohlstedt
,
Appl. Phys. Lett.
95
,
163503
(
2009
).
22.
T.
Brintlinger
,
S.-H.
Lim
,
K. H.
Baloch
,
P.
Alexander
,
Y.
Qi
,
J.
Barry
,
J.
Melngailis
,
L.
Salamanca-Riba
,
I.
Takeuchi
, and
J.
Cumings
,
Nano Lett.
10
,
1219
(
2010
).
23.
M.
Liu
,
J.
Lou
,
S.
Li
, and
N. X.
Sun
,
Adv. Funct. Mater.
21
,
2593
(
2011
).
24.
Y.
Chen
,
T.
Fitchorov
,
C.
Vittoria
, and
V. G.
Harris
,
Appl. Phys. Lett.
97
,
052502
(
2010
).
25.
S.
Zhang
,
Y.
Zhao
,
X.
Xiao
,
Y.
Wu
,
S.
Rizwan
,
L.
Yang
,
P.
Li
,
J.
Wang
,
M.
Zhu
,
H.
Zhang
,
X.
Jin
, and
X.
Han
,
Sci. Rep.
4
,
3727
(
2014
).
26.
T.
Jin
,
L.
Hao
,
J.
Cao
,
M.
Liu
,
H.
Dang
,
Y.
Wang
,
D.
Wu
,
J.
Bai
, and
F.
Wei
,
Appl. Phys. Express
7
,
043002
(
2014
).
27.
M.
Buzzi
,
R. V.
Chopdekar
,
J. L.
Hockel
,
A.
Bur
,
T.
Wu
,
N.
Pilet
,
P.
Warnicke
,
G. P.
Carman
,
L. J.
Heyderman
, and
F.
Nolting
,
Phys. Rev. Lett.
111
,
027204
(
2013
).
28.
N.
D'Souza
,
M. S.
Fashami
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
Nano Lett.
16
,
1069
(
2016
).
29.
H.
Ahmad
,
J.
Atulasimha
, and
S.
Bandyopadhyay
,
Sci. Rep.
5
,
18264
(
2015
).
30.
J.
Cui
,
C. Y.
Liang
,
E. A.
Paisley
,
A.
Sepulveda
,
J. F.
Ihlefeld
,
G. P.
Carman
, and
C. S.
Lynch
,
Appl. Phys. Lett.
107
,
092903
(
2015
).
31.
K.
Roy
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
J. Appl. Phys.
112
,
023914
(
2012
).
32.
A. K.
Biswas
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
Appl. Phys. Lett.
104
,
232403
(
2014
).
33.
K.
Roy
,
S.
Bandyopadhyay
, and
J.
Atulasimha
,
Appl. Phys. Lett.
99
,
063108
(
2011
).
34.
P.
Li
,
A.
Chen
,
D.
Li
,
Y.
Zhao
,
S.
Zhang
,
L.
Yang
,
Y.
Liu
,
M.
Zhu
,
H.
Zhang
, and
X.
Han
,
Adv. Mater.
26
,
4320
(
2014
).
35.
C.-Y.
Liang
,
S. M.
Keller
,
A. E.
Sepulveda
,
W.-Y.
Sun
,
J.
Cui
,
C. S.
Lynch
, and
G. P.
Carman
,
J. Appl. Phys.
116
,
123909
(
2014
).
36.
J.
Cui
,
J. L.
Hockel
,
P. K.
Nordeen
,
D. M.
Pisani
,
C.
Liang
,
G. P.
Carman
, and
C. S.
Lynch
,
Appl. Phys. Lett.
103
,
232905
(
2013
).
37.
Z.
Wang
,
Y.
Wang
,
W.
Ge
,
J.
Li
, and
D.
Viehland
,
Appl. Phys. Lett.
103
,
132909
(
2013
).
38.
C.
Thiele
,
K.
Dörr
,
O.
Bilani
,
J.
Rödel
, and
L.
Schultz
,
Phys. Rev. B
75
,
054408
(
2007
).
39.
M.
Liu
,
S.
Li
,
O.
Obi
,
J.
Lou
,
S.
Rand
, and
N. X.
Sun
,
Appl. Phys. Lett.
98
,
222509
(
2011
).
40.
N.
Lei
,
T.
Devolder
,
G.
Agnus
,
P.
Aubert
,
L.
Daniel
,
J.-V.
Kim
,
W.
Zhao
,
T.
Trypiniotis
,
R. P.
Cowburn
,
C.
Chappert
,
D.
Ravelosona
, and
P.
Lecoeur
,
Nat. Commun.
4
,
1378
(
2013
).
41.
L. M.
Loong
,
X.
Qiu
,
Z. P.
Neo
,
P.
Deorani
,
Y.
Wu
,
C. S.
Bhatia
,
M.
Saeys
, and
H.
Yang
,
Sci. Rep.
4
,
6505
(
2014
).
42.
M.
Donahue
and
D.
Porter
,
OOMMF User's Guide
1999
.

Supplementary Material