文部科学省科研費新学術領域 「コンピューティクスによる物質デザイン：複合相関と非平衡ダイナミクス」 2014.3.10　東京大学 本郷キャンパス 電界による磁気異方性制御：実験 Electric-field control of magnetic anisotropy: Experiment 野﨑隆行 産総研　ナノスピントロニクス研究センター Thank you for the introduction. I’m Takayuki Nozaki of AIST in Japan. First of all, I’d like to thank the organizers for being invited to this nice workshop. The main topic of today’s talk is the spin dynamics induced by the voltage control of magnetic anisotropy in an ultrathin ferromagnetic metal layer. This is the collaboration work with Osaka University. 計画研究：スピンエレクトロニクス材料の探索 研究代表者： 佐藤和則（阪大） 研究分担者： 小田竜樹 (金沢大), 小倉昌子(阪大), 野﨑隆行(産総研) 連携研究者： 黒田眞司(筑波大), 吉田博(阪大), 朝日一(阪大), 鈴木義茂(阪大), 　　　　　　　　 赤井久純(阪大), 下司雅章(阪大)

Introduction 1 スピントロニクス e- e- e- 駆動電力の低減 電子工学 磁気工学 電荷制御 (伝導・光学特性) 電子スピンの巨視的制御 e- e- 不揮発性固体磁気メモリ 生体用高感度磁界センサー トンネル磁気抵抗効果 それでは研究内容に移らせていただきます。 私が取り組んでいる研究はスピントロニクスと呼ばれる分野であり、従来電子工学分野と磁気工学分野においてそれぞれ独立に制御されてきた電子が持つ自由度である電荷とスピンを同時に、融合的に操作することによって、新しい機能性を有するデバイスの創製を目指しております。このスピントロニクスにおいて見出されたの重要な基礎物理現象の１つがトンネル磁気抵抗効果であり、数nmオーダーの絶縁層を強磁性層2層で挟んだ構造において、トンネル抵抗が上下の強磁性層の磁化の相対角度に依存して変化するという特徴を持ちます。特に絶縁層に酸化マグネシウムを用いると大きな抵抗変化が得られるというのが産総研発の技術であり、現在のスピントロニクスの発展を支えております。 この素子は電力を与え続けなくても磁化の向きとして情報を半永久的に保存することができるため、これを利用した、低待機電力な不揮発性固体磁気メモリや不揮発性SRAMなどの開発が進められています。また、最近では共振器がいらない高周波発振器や生体用の高感度磁界センサーなどへの展開も進められています。 これらの技術を実現していく上で重要な課題となっているのが駆動電力をいかに低減していくかという点です。 Yuasa et al. Nature Mater.(2004) 不揮発性SRAM 高周波発振器 課題 駆動電力の低減

Introduction 2 低消費電力化 バイポーラトランジスタ FET 真空管 駆動電力の低減に向けて 電界制御が必須！ 低消費電力なスピン状態（方向やダイナミクス）制御技術が必要不可欠 電流磁界 スピントルク (電流) 1996年～ Slonczewski, Berger 1820年～ Oersted　 電界制御が必須！ <1fJ 1bitの書き込み 消費電力 ~100 pJ ~100fJ 低消費電力化 n p バイポーラトランジスタ FET 真空管 エレクトロニクスとの対応 この点をもう少し詳しくご説明いたします。スピントロニクスデバイスにおいて情報の操作はスピン、つまり磁化の方向やダイナミクスを制御することによって行われます。現在このスピン制御は電流を用いて行われています。最も単純に思いつく方法は配線に電流を流した時に発生する電流磁界で、実際に長い歴史の中でこれが唯一の方法だったわけですが、磁石が小さくなるほど大きな電流が必要となる点が問題となっていました。最近になって、スピン角運動量を電流にのせて直接ナノスケールの磁石に注入した際に発生する“スピントルク”と呼ばれる量子力学的な現象を利用した制御法が開発され、数十nmの素子1bitの記録に必要なエネルギーは100fJ程度にまで下がってきています。しかしながら、電流による制御を使う限りオーミック損失による不要な電力消費が本質的に避けられない壁として存在します。この問題を抜本的に解決するためには電界によるスピン制御技術の確立が重要であると考えています。これは、エレクトロニクス分野の歴史において、電流駆動型の真空管やバイポーラトランジスタが淘汰されて、電界駆動型のFETが現在の中心技術となっているのと同様な技術革新をスピントロニクスにもたらす可能性があり、次のエポックメイキングな技術になると考えております。

Introduction 3 電界によるスピン制御の試み 実用デバイス化への要求 室温での安定な動作 高い繰り返し動作耐性 磁歪制御 キュリー点制御 マルチフェロイック ピエゾ素子 磁性薄膜 H. Ohno et al. Nature (2000). D. Chiba et al. Nature Mater. (2011) L. W. Martin et al. J. Phys. : Condens. Mater. (2008). V. Novosad et al. JAP (2000). Until now, so many approaches of electric-field control of magnetic properties have been suggested and experimentally demonstrated, for example, the control of magnetostriction, Currie temperature, multi-ferroic properties, structural phase transition, and so on. But at present, there is no technique which can satisfy all requirements for applications, that is, stable operation at room temperature, high speed operation, unlimited cycling endurance, and compatibility with magnetoresistance structures. 室温での安定な動作 高い繰り返し動作耐性 磁気抵抗素子との複合化 高速動作 実用デバイス化への要求

M. Weisheit et al. Science 315, 349 (2007) Introduction 4 電界磁気異方性制御 M. Weisheit et al.　Science 315, 349 (2007) Kerr rotation 0.4 V Pt Pt MgO FePt 1.0 V Electrolyte -0.12 -0.11 -0.1 0 0.1 Electric double layer H (T) H (T) 液体電界質による電気２重層の利用 4.5%の保磁力変化 室温において、3d遷移金属の垂直磁気異方性を電界で制御することが可能！

Introduction 5 Number of electrons 電界誘起磁気異方性変化の起源 -理論- Pt Fe M. Tsujikawa and T. Oda, Phys. Rev. Lett. 102, 247203 (2009). Pt Fe Number of electrons Electric field (V/Å)

Contents 全固体素子における電圧磁気異方性制御の実現 電圧磁気異方性変化を利用したスピンダイナイクス制御 実用デバイス化への要求を満たす電界スピン制御法の開発に向けて・・・ 室温で安定に動作する固体素子 高い繰り返し動作耐性 磁気抵抗素子との複合化 高速動作 全固体素子における電圧磁気異方性制御の実現 電圧磁気異方性変化を利用したスピンダイナイクス制御

全固体素子における電圧磁気異方性制御の実現 T. Maruyama, Y. Shiota, T.N. et al. Nature Nanotech. 4, 158 (2009) Y. Shiota, T. N. et al. Appl. Phys. Exp. 2, 063001 (2009) T. Nozaki et al. APL. 96, 022506 (2010) Y. Shiota, T.N. et al. APEX 4, 043005 (2011) 室温で安定に動作する固体素子 高い繰り返し動作耐性 磁気抵抗素子との複合化 高速動作

Experiment 1 Au (001) 50 nm 超薄膜磁性層における界面誘起の垂直磁気異方性 MgO(001) 10 nm 超薄膜Fe80Co20(001) tFeCo Au (001) 50 nm 界面磁気異方性エネルギー Ks = 650 mJ/m2 磁性層膜厚によって垂直磁気異方性の大きさを設計可能

Experiment 2 電圧印加による垂直磁気異方性制御 tFeCo = 0.58 nm 電圧印加により磁化容易軸が面内-面直間で遷移 極カー効果測定 tFeCo = 0.58 nm -200 V 200 V 3d遷移金属強磁性体/MgO接合は現在のスピントロニクスの基本構造！

CoCrPt-TiO2 nanocomposite 様々な材料系における電界磁気異方性制御の報告例 CoFeB FePd, FePt CoCrPt-TiO2 nanocomposite laser M. Endo et al. APL 96, 212503 (2010) K. Kita et al. APL 112, 033919 (2012) F. Bonell et al. APL 98, 232510 (2011) T. Seki et al. APL 98, 212505 (2011) T. Zhou et al. APL 96, 012506 (2010)

Sputtering MBE Junction size: 2 × 6 mm2 Free layer, tfree トンネル磁気抵抗素子における電界磁気異方性制御 Junction size: 2 × 6 mm2 Ru(7 nm) Sputtering Ta (5 nm) FeB (1.5 nm) Free layer, tfree Fe wedge (0-0.7 nm) MgO (2.5 nm) Fe (3 or 0.7 nm) Reference layer MBE Cr (30 nm) MgO (3 nm) MgO (001) substrate

Experiment 3 トンネル磁気抵抗素子における電界磁気異方性制御 Fe(3 nm)/MgO/Fe(0.3 nm)/FeB (1.5 nm) /Ta/Ru Hex Vbias = 30 mV Actually, we fabricated a fully epitaxial magnetic tunnel junction with the ultrathin FeCo layer and confirmed the voltage-induced anisotropy change through the bias voltage dependence of the TMR curves as shown in this example. An important progress in this experiment was that the applied voltage required to induce the anisotropy change was substantially reduced comparing with our first experiment, that was about several hundreds V, but in this case, several hundreds mV, thanks to the single MgO barrier. The slope of the surface anisotropy energy change was evaluated to be about 30 to 40 fJ/Vm in this structure.

Experiment 3 TMR曲線のバイアス電圧依存性 +V Positive bias: electron depletion Fe(3 nm)/MgO/Fe(0.3 nm)/FeB (1.5 nm) /Ta/Ru +V Positive bias: electron depletion Negative bias: electron accumulation

Experiment 3 Bias voltage dependence of Eperptfree 30% change in Eperp by 1V application

電圧磁気異方性変化を利用した スピンダイナイクス制御 T. Nozaki et al. Nature Phys. 8, 491 (2012) 室温で安定に動作する固体素子 高い繰り返し動作耐性 磁気抵抗素子との複合化 高速動作

Introduction 6 Basic research Applications 強磁性共鳴 (FMR) マグノニクス(スピン波) スピンポンピング マイクロ波アシスト磁化反転 スピントルク発振 スピントルク検波 ダンピング定数 磁気異方性 層間交換結合 飽和磁化 FMR signal スピンポンピング Let me start by introducing the background of my research. As you know, the ferromagnetic resonance, so-called FMR, is a collective spin excitation observed in ferromagnetic materials. This phenomenon has been widely used in basic research to characterize high frequency spin dynamics and also to evaluate important parameters of magnetic materials, like damping constant, magnetic anisotropy, interlayer exchange coupling and so on. And recently, an importance of FMR dynamics is increasingly growing even for applications due to its high efficiency, for example, for spin wave excitation in Magnonics, generation of pure spin current by spin pumping, MAMR effect, spin-transfer related phenomena and so on. マグノニクス Frequency S. Mizukami et al. PRB (2002) A. Brattas et al. PRB (2002) A. Serge et al. J. Phys. D: Appl. Phys. (2010)

Concept 電界磁気異方性制御による強磁性共鳴励起 Vrf LLG equation Hex DHd (Vrf) Concept is very simple. Now we got the way to control the effective demagnetization field Hd of the ultrathin ferromagnetic layer by the voltage application. So, the rf voltage application can induce the oscillatory change in the Hd, and if its frequency is tuned to the resonant frequency, the FMR dynamics should be excited around the static external magnetic field. 高周波電界による磁気異方性変化

Experiment 4 ホモダイン検波法 電界励起FMR信号例 Prf = -15 dBm qH: 65deg. DC voltage RA: 420 kWmm2 Junction size: 2 × 6 mm2 Prf = -15 dBm qH: 65deg. ホモダイン検波法 To demonstrate this effect, we used the similar MTJ sample with the ultrathin FeCo layer, but for the special specification of the FMR excitation, the FeCo thickness was tuned to the critical thickness of the interface-induced perpendicular anisotropy, where the transition of the magnetic easy axis between in-plane and out-of-plane occur. To detect the FMR dynamics electrically, we used a homodyne detection technique through the TMR effect. This technique was first used to investigate the spin-transfer torque induced FMR excitation in this paper. But to exclude the influence of current induced torques, we designed the RA value large enough in our device, about 9 kWmicro2, which is about 3000 times larger than that used in this spin-torque induced experiment. An rf voltage with the frequency omega was applied from a signal generator to the MTJ through an rf port of a bias tee. If the FMR dynamics is excited, it causes the resistance oscillation. The product of the oscillating resistance and very small tunneling current flowing in the element produces signals of dc voltage and rf voltage with the frequency of 2 omega component. So, by measuring the output dc voltage from the MTJ using a voltmeter, we can detect the FMR dynamics. External magnetic field was applied with the elevation angle of thetaH. Because of the designed very small effective demagnetization field, magnetization of the ultrathin FeCo layer is easily aligned with the external magnetic field, while that of the top thick Fe layer always lie in the film plane. A. A. Tulapurkar et al. Nature (2005) (Spin-torque induced FMR: RA~3 Wmm2) DC voltage

電界による磁化反転制御 静電界印加 短パルス電界印加 歳差運動を利用したダイナミック磁化反転 磁化反転できない Y. Shiota, T.N. et al. Nature Mater. 11, 39 (2012). Y. Shiota, S. Miwa, T.N. et al. Appl. Phys. Lett. 101, 102406 (2012) 静電界印加 短パルス電界印加 歳差運動を利用したダイナミック磁化反転 Next, I’d like to introduce the control of bi-stable magnetization switching by using the voltage-induced dynamics. As I showed you in the introduction, the static voltage application can induce the switching of the magnetic easy axis between the in-plane and out-of-plane directions, but when we switch of the voltage, magnetization is turned back to the initial state, that means, we cannot control the bi-stable switching only by the static voltage application. But as you know, the bi-stable switching is the most important technique for the binary information processing in memory devices. Our question is how can we realize it by using the voltage effect. 磁化反転できない

パルス電界によるダイナミック磁化反転(シミュレーション) 0.4 ns 立ち上り&立下り: 70 ps V パルス電圧印加下における磁気エネルギー変化 エネルギー障壁 Hbias Hbias Hbias Voltage OFF： Hext,z = 700 Oe Ha, ⊥ = 1400 Oe x y Voltage ON： Hext,z = 700 Oe Ha,⊥ = 600 Oe x y Heff Voltage OFF： Hext,z = 700 Oe Ha,⊥ = 1400 Oe x y

Experiment 5 Hex Minor loop ( 84°) AP P 84° Vpulse=-1.35 V tFeCo: 0.7 nm (in-plane) tMgO: 1.5 nm Junction size: 0.2×0.8 mm2 AP P 84° Hex Vpulse=-1.35 V tpulse = 0.65 ns Lock-in amplifier 2 kW sign out 50 mV, 333 Hz Pulse Generator Au 50 nm or MgO Au SiO2 反平行磁化状態 平行磁化状態 This is the measurement setup for the demonstration of the voltage-induced dynamic switching. We used the MTJ with the FeCo of 0.63nm, which is the in-plane magnetized film with the relatively small effective demagnetization field of about1400 Oe. The external magnetic field was applied to the elevation angle of 87 degree from the in-plane. The reason of this small tilt angle is to compensate the stray field from the top thick Fe layer with keeping the perpendicular field component. This is the minor MR loop measured under this external field condition. The observed hysteresis comes from the in-plane uniaxial anisotropy of the ultrathin FeCo layer. If the precessional motion is excited by the short pulse voltage application, bi-stable switching between these parallel and anti-parallel states can be induced. The pulse signal with the fast rise and fall time of 70ps was applied through an RF port of the bias tee, and the switching event was monitored by measuring a resistance using a lock-in amplifier.

消費電力比較 310 fJ 1.4 fJ 90 fJ 電流駆動型と比較して約２桁の低消費電力化の可能性 本研究 (200×800 nm2) ＊Toshiba Co., IEDM2012 本研究 (200×800 nm2) 見込み値 （φ30 nm） スピントルク型* (φ30 nm) 310 fJ 1.4 fJ 90 fJ 電流駆動型と比較して約２桁の低消費電力化の可能性 Here, I would like to discuss about power consumption in writing process. Consuming power can be expressed by sum of leakage current energy and charge/discharge energy to the capacitor. In our device, the power from leakage current is 260 fJ, and electrostatic energy is 50 fJ. And in totally we needs 310 fJ. And if we assume the junction size to 30 nm-diameter MTJs, it becomes 1.4 fJ. This value is much lower than spin-transfer-torque MRAM. But still leakage current energy is dominant When the switching was done only thorough the voltage effect in ideal devices, writing energy consumes only sub-fJ. So the switching using voltage is very desirable method for writing process in memory devices.

今後の課題 磁性層の超薄膜化による熱安定性の低下 ⇒ 微細化（大容量化）に対応できない 磁性層の超薄膜化による熱安定性の低下　 　⇒　微細化（大容量化）に対応できない エネルギー障壁 D 磁気異方性 体積 不揮発性メモリなどの応用には D ~ 40-60kBTが必要 磁化の向き 目標 Au/FeCo/MgO*1 MgO/FeB/MgO*2 超Gbit級 素子サイズ Φ30nm Φ10nm D 9 51 50 垂直磁気異方性 (Merg/cc) 1 2 20 電界効果 (fJ/Vm) 30 100 1000 1* T. Nozaki et al. APL. 96, 022506 (2010), 2* T. Nozaki et al. Appl. Phys. Exp. 6, 073005 (2013) 高結晶磁気異方性材料 Pt/Fe(1ML)/Pt(1ML) ~80×er (fJ/Vm) High-k誘電体の導入　　e.g. SrTiO3 er > 200 (er, MgO ~ 10)

Summary 全固体素子(トンネル磁気抵抗素子)における電界磁気異方性を実現 電界による高速スピンダイナミクス（強磁性共鳴）励起を実証 電界パルスを用いたダイナミック磁化反転を実証 Voltage control of magnetic anisotropy Voltage-driven MRAM Highly-sensitive detector We hope these developed techniques can open up the new route to the electric-field based spin technologies. Voltage-driven three terminal device (Spin transistor) Voltage-induced spin wave excitation

Phase diagram of the Pswitch Results 5 Phase diagram of the Pswitch measurement range Vpulse = -0.75 V AP ⇒ P P ⇒ AP This is the main result, phase diagrams of the switching probability as functions of pulse duration time and external magnetic fields both for AP to P switching and P to AP switching induced by the negative pulse voltage application. Here the switching probability, Pswitch, which was defined as the switching event number divided by the 100 trials. For this experiment, first magnetization state was initialized every time before the pulse voltage application In both cases, very beautiful oscillatory behaviors of the switching probability were clearly observed. For example, this is the cross section of this line, and you can see clear oscillation with the period of about 1ns and it attenuates in the longer pulse duration time.

Macro-spin mode simulation z ＜LLG equation＞ x y Magnetic energy： From experiment; Ms = 1.54 T Hc(0K) = 25 Oe Hshift = 75 Oe Hperp (0 V) = 1400 Oe Hperp (-0.75V) = 600 Oe T = 300K Parameter; a = 0.01 1600 1400 1200 1000 800 600 400 To discuss the observed tendency, we made a simple macro-spin model simulation, it’s based on the conventional LLG equation, but the voltage effect is included as the change in the effective field of out-of-plane direction. Parameters used in the calculation were obtained from the experiment, except for the damping constant. The most important parameter is the perpendicular anisotropy field under the voltage application, and it was estimated from our previous experimental results of quantitative evaluation of perpendicular anisotropy under the bias applications. Hperp (Oe) -0.75 -0.5 -0.25 0 0.25 DC bias voltage (V)

W. –G. Wang et al. Nat. Mater. (2012) Structure Slope (fJ/Vm) Ref. Au / Fe80Co20 (0.4 nm) / MgO -37 T. Nozaki et al. APL 2010 Ta / Co40Fe40B20 (1.33 nm) / MgO -33 M. Endo et al. APL 2010 Au / Fe80Co20 (0.7 nm) / MgO -31 Y. Shiota et al. APEX 2011 Au/FePt (1.5 nm) /MgO 19 T. Seki et al. APL 2011 Ta / Co60Fe20B20 (1.2 nm) / MgO / Al2O3 -11 K. Kita et al. JAP (2012) Ta / CoFeB (1.3 nm) / MgO -50 W. –G. Wang et al. Nat. Mater. (2012) MgO / CoFeB (1.8 nm) / Ta S. Kanai et al. APL (2012) Ru / CoFeB (1.4 nm) / MgO 18 Y. Shiota et al. APL (2013) MgO / FeB (1.5 nm) / MgO -108 T. Nozaki et al. APEX (2013) MgO / Fe(0.3 nm) / FeB (1.5 nm) / Ta -105 This study Theory Vacuum / Fe (15 ML) / Vacuum -20 C.-G. Duan et al. PRL 2008 Vacuum / Fe (1 ML) / Vacuum -33 K. Nakamura et al. PRL 2009 Pt / Fe(1 ML) / Pt (1 ML) / Vacuum -72 M. Tsujikawa et al. PRL 2009 Cu / Fe (9 ML) / MgO 100 M. K. Niranjan et al. APL 2010 Au / Fe(2 ML) / MgO 11.6 M. Tsujikawa et al. JAP 2012

Comparison of experimental results and simulation AP ⇒ P P ⇒ AP Experiment Vpulse = -0.75 V Here is the comparison of experimental results and simulation for both from AP to P and P to AP switching. The observed tendency is reproduced excellently by this very simple calculation. So, we can conclude that the observed switching originates from the voltage-induced precessional motion of the magnetization. This is the first demonstration of the bi-stable magnetization switching only by using the voltage effect in the magnetic tunnel junction structure. Simulation

Introduction 5 垂直磁気異方性の起源 3d-白金族合金の結晶磁気異方性 L10-Fe(Co)Pt, Fe(Co)Pd 白金族系の強いスピン-軌道相互作用を活用 2. 界面誘起の垂直磁気異方性 MgO/3d遷移金属界面 Fe/MgO ; T. Shinjo et al. J. de Physique 40, C2-86-87 (1979). CoFeB / MgO ; Ikeda et.al. Mature Mater. 9, 721 (2010). L10 ordered FePt Ikeda et.al. Mature Mater. 9, 721 (2010) z2 (m=0) Fe O Mg MgO 弱いFe dz2- O pz混成 ⇒　垂直磁気異方性

Voltage-induced torque z x y

No influence from spin transfer torque?? Discussion 1 Estimation of flowing tunneling current (current density) Too small! Here, I’d like to discuss about the origin of the observed FMR signal, especially from a point of an influence of current. Thanks to the designed high RA value, maximum tunneling current flowing in the element is estimated to be about 10 micro A, which corresponds to the current density of 10^7 A/m2. An The Oersted field generated by this current is too small, about mOe order, so we can neglect its influence, and also the amplitude of the spin transfer torque estimated from this current density is about one thousandth of the expected voltage torque. So, in usual case, we can completely neglect the influence of current. But on the other hand, in this experiment, the thickness of the FeCo layer was designed to have very small effective demagnetization field, so it can be very sensitive to the current torque than the usual experimental condition. So, in order to make sure that the voltage torque is the dominant origin, we performed another examination using the homodyne detection technique, it is the elevation angle dependence of the signal amplitude. However…the ultrathin FeCo layer can be very sensitive to the current torques due to the very small HZ…

Spin transfer torque

Voltage-induced torque 103 times larger !

Comparison of the power consumption Spin torque induced FMR Cf. S. Ishibashi et al. Appl. Phys. Express, 3, 073001 (2010) CoFeB / MgO / CoFeB MTJ 100×150 nm2 RA: 2 Wmm2, MR: 100% Precession angle: 1 deg. Consumed power: 1 mW Voltage-field induced FMR The most fascinating feature of the electric-field control is its low power consumption. Typically, the spin-transfer effect requires a few microwatts to excite the linear FMR dynamics with a precession angle of 0.5 degree in deep submicron scale tunnel junctions. On the other hand, for the case of voltage-induced FMR, if we assume the same sample size and only consider the power consumed in the element, the same order of precession angle can be excited by applying only 2nW, thus, the power consumption is reduced to about one-thousandth. These results provide sufficient evidence of the effectiveness of the voltage control comparing with the current-based control. Anyway, we succeeded to excite the FMR dynamics by the direct voltage application, that is, we obtain a new approach to control the spin dynamics. Assuming the same sample size… Precession angle : 1 deg. Consumed power: 0.005 mW Power reduction of 1/200!

Result 5 Tilted field angle: 55deg. Hex = 500 Oe Input voltage dependence of the signal amplitude Tilted field angle: 55deg. Hex = 500 Oe From the amplitude of the homodyne signal, we can roughly estimate the precession angle, and it reaches about 1 degree under this experimental condition.

Results 1 TMR curves Difference in the saturation fields reflect the FeCo thickness dependence (normalized MR curve) Difference in the saturation fields reflect the surface magnetic anisotropy

Result 2 +V H -V Anisotropy change slope: 37 fJ/Vm Bias voltage dependence of mag-noise spectrum (tFeCo: 0.68 nm (Hperp=1500 Oe), tMgO: 1.5 nm, Hex = 2500 Oe) Kittel’s equation +V H -V Mag noise is an rf signal generated by the thermally excited spin dynamics observed in a magnetoresistance device under a dc bias current application, and from the noise peak we can evaluate the resonance frequency of the free layer. For this experiment, we used the MTJ with the FeCo thickness of 0.68 nm, it’s the in-plane magnetized film and has the perpendicular anisotropy field of 1500Oe. The experiment was performed under the perpendicular external field of 2500 Oe, it is enough larger than the perpendicular anisotropy field. These are examples of mag-noise spectra measured under various dc bias voltages. We can see clear shift of the peak frequency depending on the bias voltages, and here is the summary. The peak frequency changes linearly depending on the applied electric fields. From the peak frequency, we can calculate the perpendicular anisotropy field Hperp using this modified Kittel’s equation as shown here, and then it can be converted into the perpendicular magnetic anisotropy energy Eperp from this relation. The estimated value of Eperp times FeCo thickness is shown in this right axis. From this result, the anisotropy change slope was evaluated to be about 37fJ/Vm. This value is the same order with the recent theoretical expectation. Anisotropy change slope: 37 fJ/Vm Cf. Theory: Fe(1ML) / MgO (3 ML) 29 fJ/Vm R. Shimabukuro et al. Physica E 42, 1014(2010)

Macro-spin model simulation Hex = 700 Oe Tilted angle: 84 degree

Estimation of precession angle, Dq

<o|lz|u> <o|lx|u> Second order perturbation theory (D. S. Wang et al. PRB, 47, 14932 (1993)) ko : k vector of occupied state ku : k vector of unoccupied state la (a= x, z): angular momentum operators. <o|lz|u> Out-of-plane <o|lx|u> In-plane x2-y2(m=±2) xy (m=2) xz,yz (m=±1) EF z2 (m=0) xz,yz (m=±1) xy (m=2) Simplified band structure of the monolayer Fe

Electric field induced anisotropy change M. Tsujikawa and T. Oda, PRL 102, 247203 (2009). Pt Fe

Ms = 1.83×106 A/m （Experiment） g= -2.3×105 m/(A sec) （g=2.1） a = 0.01 （parameter） Hc(0K) = 25 Oe （Experiment） Hshift = 73.2 Oe （ Experiment ） Hs,perp (0 V/nm) = 1400 Oe （ Experiment ） Hs,perp (-1V/nm) = 600 Oe （ Experiment ）

Thickness dependence of the MS for Fe80Co20 layer 1.8 1.44

J. Stöhr et al. Appl. Phys. Lett. 94, 072504 (2009) Dynamic switching 2 Voltage-induced magnetization switching of perpendicularly magnetized film J. Stöhr et al. Appl. Phys. Lett. 94, 072504 (2009)