Optimization of indirect magnetoelectric effect in thin-film/substrate/piezoelectric-actuator heterostructure using polymer substrate
Mouhamadou Gueye, Fatih Zighem, Damien Faurie, Mohamed Belmeguenai, Silvana Mercone
OOptimization of indirect magnetoelectric effect inthin-film/substrate/piezoelectric-actuator heterostructure using polymer substrate
Mouhamadou Gueye, ∗ Fatih Zighem, † and Damien Faurie, Mohamed Belmeguenai and Silvana Mercone LSPM-CNRS, Université Paris XIII, Sorbonne Paris Cité, 93430 Villetaneuse, France (Dated: July 17 th a priori undesirable and huge uniaxial anisotropy which seems tobe related to a non equibiaxial residual stress inside the magnetostrictive film. The “strong” ampli-tude of this non equibiaxiallity is due to the large Young’s modulus mismatch between the polymerand the magnetostrictive film which leads to a slight curvature along a given direction during theelaboration process and thus to a large magnetoelastic anisotropy. Prospect of local magnetization control using an elec-tric field or a pure voltage, for low power and ultrafastnew electronics has resulted in an amount of new re-search domains mainly focused on artificial engineeredmaterials[1–4]. Artificial magnetoelectric systems seemsto be a promising route for such control. The easiest ar-tificial architecture for a magnetization voltage controlin those kinds of materials appears to be the piezoelec-tric/magnetostrictive bilayers presenting a good strain-mediated coupling at the interface [4–12]. However, inthese systems, clamping effects due to the substrate lim-its the strain applicable by the piezoelectric environmentto the magnetization and therefore reduces perspectiveon applications. Indeed, in this kind of bilayers, a sig-nificant magnetoelectric coupling is obtained only in thepresence of non negligible in-plane stresses in the mag-netic media. Therefore, more the elastic strains are trans-ferred at the interfaces of such systems, more this indirectmagnetoelectric effect is optimized. Generally, the mag-netic thin films are first deposited on a thick substrate( ∼ hundreds of microns), which is then cemented on apiezoelectric actuator [9, 10, 13, 14]. This process limitsthe desired phenomenon, especially when the substrate isstiff such as for commonly used wafer (Si, GaAs, ...). Thisoften reported limitation[14] (a few ten percents of lossesin best cases) can be avoided by depositing the magneticthin film on a compliant substrate such as polyimides [10]that are more and more used in flexible spintronics[16].The present study presents a quantitative comparisonof the obtained effective MagnetoElectric (ME) couplingin two different artificial magnetoelectric heterostruc-tures characterized by the presence of a flexible or a rigidsubstrate between a ferromagnetic film and a piezoelec-tric actuator. The effective ME coupling will be deducedfrom in situ MicroStrip FerroMagnetic Resonance (MS-FMR). The artificial ME heterostructures are composed ∗ Electronic address: [email protected] † Electronic address: [email protected] V Piezoelectric actuatorNi (200 nm) V Piezoelectric actuatorSi (500 µm)Ni (200 nm) a) b) x y
Kapton® (125 µm)
Figure 1: Schematics of the two studied heterostructures. Theonly difference comes from the substrate (either flexible (a)or rigid (b)) on which the Ni thin film has been deposited. of a 200 nm Nickel film deposited by radio frequencysputtering either on a rigid substrate (Silicon of thick-ness 500 µ m) or on a flexible one (Kapton® of thickness125 µ m) and then glued onto a piezoelectric actuator aspresented on Figure 1. It is worth mentioning that Nimaterial has been chosen because of its well-known neg-ative effective magnetostriction coefficients at saturationeven in a polycrystalline film with no preferred orien-tations, which is closely the situation for both systems.In this condition, only one magnetostriction coefficientat saturation ( λ ) is sufficient to characterize the magne-toelastic anisotropy. Moreover, the two substrates havebeen chosen because of their Young’s modulus ( ∼ ∼
180 GPa for Si).MS-FMR experiments have been performed at a fixedfrequency by sweeping the applied magnetic field from0 to 3 kOe [15]. The microwave driven frequency wasfixed at 8 GHz in all the presented experiments, wherethe resonance fields are sufficiently higher than the in-plane magnetic anisotropies fields present in the struc-tures, avoiding undesirable magnetic and magnetoelastichysteresis effects. The in situ applied voltage experi-ments have been done by varying the external voltagefrom 0V to 100V and back to 0V with step of around5 V. Figure 2 shows typical MS-FMR spectra for twodifferent applied voltages (0 and 100V). The magneticfield is applied along y direction which corresponds to a a r X i v : . [ c ond - m a t . m t r l - s c i ] A ug
500 1000 1500 2000 ‐ ‐ ‐ Ni/KaptonNi/Si I n t e n s i t y ( a . u . ) Applied field (Oe) R H R H Figure 2: Experimental spectra recorded at 8 GHz for anin-plane magnetic field at 90 degree with respect to the mainpositive strain axis of the actuator (along y -axis). The largestshift of the resonance is clearly put into evidence for the Nifilm deposited onto Kapton®. negative induced in-plane strains. Indeed, in the studiedrange of applied voltage [0-100V], the piezoelectric ac-tuator presents a main positive in-plane strain along x direction ( ε xx > ) and a smaller negative one along y direction ( ε yy < with ε xx ∼ − ε yy ) [10]. We observedthat in both structures the resonance field decreases asfunction of the applied voltage. This behavior is consis-tent with the negative magnetostriction coefficient at sat-uration of Ni material ( λ < ). Indeed, in first approxi-mation, the magnetoelastic anisotropy can be viewed as avoltage-induced uniaxial magnetoelastic anisotropy field (cid:126)H ME ( V ) along y direction. Thus, since the in situ MS-FMR experiments are performed along y , this y -axis willbe easier (when increasing the applied voltage) for themagnetization direction leading to lower values of theresonance field. Figure 2 shows a clear shift of the res-onance field δH R (defined as δH R = H R (0) − H R ( V ) )between 0 V and 100 V for both heterostructures: it isclose to 50 Oe for Si substrate and reaches a value around350 Oe when a flexible substrate is used.The complete δH R variations as function of the ap-plied voltage are reported in Figure 3 where circles (resp.squares) refer to results obtained for Si (resp. Kap-ton®) substrate. A non linear and hysteretic variationis put into evidence when using Kapton® as a substratewhereas an almost linear dependence is found in the sec-ond structure indicating a difference of the strain trans-mission efficiency. This non linear and hysteretic behav-ior is due to the intrinsic properties of the ferroelectricmaterial used during piezoelectric actuator fabrication[10]. However, in first approximation, if this non linearand hysteretic variation of δH R is neglected and adjustedby a linear fit, an effective magnetoelectric coupling α ME (in V.cm − .Oe − being given the width of the piezoelec-tric actuator ( . cm)) can be estimated. The linearadjustments to the experimental data are presented in Ni/Si
Ni/Kapton V o l t a g e ( V ) H R (Oe) u p d o w n Figure 3: Resonance field shift ( δH R ) variations as functionof the applied voltage for the two studied heterostructures.The lines correspond to linear fits of the experimental data.Arrows refer to up and down applied voltage sweeps (0 V to100 V and 100 V to 0 V, respectively). Figure 3. A strong magnetoelectric coupling is found forthe Ni/Kapton®/Piezoelectric system with an effectivemagnetoelectric coefficient α ME ∼ . V.cm − .Oe − . In-terestingly, in the case of the Ni/Si/Piezoelectric sys-tem, it can be conclude that the magnetoelectric cou-pling is roughly seven times less efficient because of itshigher magnetoelectric coefficient value ( α ME ∼ . V.cm − .Oe − ). The relatively weak coupling found inthe second system is due to a weaker transmission of thein-plane elastic stresses. The strain loss of a factor of al-most 7 cannot be explained only by the stiffness of Siliconbut also by the often reported imperfection of cementa-tion of the epoxy glue[14]. Obviously these imperfectionsare insignificant when the substrate is highly stretchablelike polymers.However, depositing a magnetic thin film on a flexiblesubstrate can lead to an a priori undesirable effects re-lated to its possible anisotropic non-flatness during elab-oration process. Moreover, because of the large contrastbetween the polymer and the Ni Young’s modulus ( ∼ ∼
200 GPa, respectively), a more or less pro-nounced additional curvature due to growth stress de-velops during film deposition. These phenomenona leadto a stress state in the magnetic film with a possiblein-plane non-equibiaxiallity[17]. This behavior is illus-trated in Figure 4 showing in-plane angular dependen-cies (thereafter ϕ H is the angle between the in-plane ap-plied magnetic field and x -axis) of the resonance field ofthe as-deposited structures. A huge uniaxial anisotropy(anisotropy field close to 300 Oe) characterized by a hor-izontal peanut shape of the resonance field variation isobserved when the Ni film grown on Kapton® whereasa very weak one (anisotropy field around 30 Oe) is foundwhen using the Si substrate. Since the present Ni filmsare polycrystalline with no preferential crystallographic R e s o n a n c e f i e l d ( O e ) Ni/Si
Ni/Kapton
Figure 4: In-plane angular dependence of the resonancefield at zero applied voltage for the two studied systems.Symbols refer to experimental data while lines are bestfits to the experimental data by using equations 2 and3. These fits allow an evaluation of the in-plane non-equibiaxial stress: (cid:12)(cid:12) σ residualxx − σ residualyy (cid:12)(cid:12) = 200 MPa and (cid:12)(cid:12) σ residualxx − σ residualyy (cid:12)(cid:12) = 15 MPa for the structures with arigid and flexible substrate, respectively. orientations, no in-plane macroscopic magnetocrystallineeffects are expected in both cases. Thus this uniaxialanisotropy should have a magnetoelastic origin. The in-fluence of the residual stresses on the magnetic proper-ties of the as deposited films can be modeled using anisotropic magnetoelastic anisotropy energy term F ME : F ME = − λ (cid:18)(cid:18) γ x − (cid:19) σ residualxx + (cid:18) γ y − (cid:19) σ residualyy (cid:19) (1) σ residualxx and σ residualyy being the in-plane principalresidual stress tensor components while γ x and γ y cor-respond to the direction cosines of the in-plane magne-tization. In this condition the resonance field expressioncan be written as H R = H + H with: H = (cid:34)(cid:16) πM s + 3 λ M S (cid:0) σ residualxx sin ϕ H + σ residualyy cos ϕ H (cid:1)(cid:17) + (cid:16) πfγ (cid:17) (cid:35) . − πM S (2) H = − λ M S (cid:34) σ residualxx (1 + 3 cos 2 ϕ H )+ σ residualyy (1 − ϕ H ) (cid:35) (3)In the above expressions H essentially represents aconstant shift in the resonance field baseline because πM S ( M S being the saturation magnetization) and πfγ ( γ is the gyromagnetic factor) are found to be larger thanthe equivalent magnetoelastic anisotropy field. The influ-ence of in-plane residual stresses on the magnetic prop-erties is included in H term. It should be noted thatthis expression is obtained in the assumption of a uni-form magnetization (macrospin approximation) alignedalong the applied magnetic field. This assumption iswell fulfilled at the 8 GHz driven frequency chosen forthis study. It clearly appears that an equibiaxial resid-ual stress ( σ residualxx = σ residualyy ) leads to an isotropicvariation of the resonance field with a slight increaseor decrease of the mean value depending on the sign ofthe residual stress. However, a non equibiaxial resid-ual stress ( σ residualxx (cid:54) = σ residualyy ) leads to an anisotropicangular variation of the resonance field. Thanks tothis model, the experimental angular variations of theresonance field have been fitted (see continuous linesin Figure 4) by using usual Ni material magnetic pa-rameters: M S = 480 emu.cm − , γ = 1 . × Hz.Oe − and λ = − × − [10]. The deduced non-equibiaxiallity is (cid:12)(cid:12) σ residualxx − σ residualyy (cid:12)(cid:12) = 200 MPa and (cid:12)(cid:12) σ residualxx − σ residualyy (cid:12)(cid:12) = 15 MPa for the Ni film depositedonto flexible and rigid substrate, respectively. Theseresults show that MS-FMR experiments can also beused for the determination of the non-equibiaxial resid-ual stress in magnetostrictive films. It should be notedhere that this kind of knowledge is not straightforwardto get with standard techniques (sample deflection, X-raydiffraction), for which an equibiaxial stress state is gener-ally assumed. Indeed, this parameter cannot be neglectedsince it determines the initial magnetization direction inthe magnetic thin film being considered.In conclusion, an optimization of the indirect magne-toelectric coupling in thin-film/substrate/piezoelectric-actuator heterostructure has been performed by employ-ing a polymer as a substrate. This optimization is due toa better strain transmission between the piezoelectric ac-tuator and the Ni film which is due to the weak Young’smodulus of the Kapton®. Furthermore, at zero appliedvoltage, a huge magnetoelastic anisotropy is evidenced inthe Ni/Kapton®/Piezoelectric system. It is attributedto a “residual” non-equibiaxial stress state due to a slightcurvature along a given direction which generally appearswhen depositing a metallic film onto a flexible substrate.However, this effect could be a limitation for some ap-plications where weak in-plane magnetic anisotropies arerequired.
Acknowledgments
The authors gratefully acknowledge the CNRS for hisfinancial support through the “PEPS INSIS” program(FERROFLEX project). This work has been also par-tially supported by the Université Paris XIII through a“Bonus Qualité Recherche” project. [1] Jing Ma , Jiamian Hu , Zheng Li and Ce-Wen Nan, Adv.Mater. , 1062 (2011)[2] Carlos A. F. Vaz , Jason Hoffman , Charles H. Ahn andRamamoorthy Ramesh, Adv. Mater., , 2900 (2010)[3] Pedro Martins and Senentxu Lanceros-Méndez, Adv.Funct. Mater., , 3371 (2013)[4] N. Lei, S. Park, P. Lecoeur, D. Ravelosona and ClaudeChappert, Phys. Rev B , 012404 (2011)[5] M. Liu, O. Obi, Z. Cai, J. Lou, G. Yang, K. S. ZiemerandN. X. Sun, J. Appl. Phys. , 073916 (2010)[6] J. Lou, M. Liu, D. Reed, Y. Ren, and N. X. Sun, Adv.Mater. , 4711 (2009)[7] Z. Wang, R. Viswan, B. Hu, J.-F. Li, V. G. Harris andD. Viehland, J. Appl. Phys. , 034108 (2012)[8] Z. Li, J. Hu, L. Shu, Y. Gao, Y. Shen, Y. Lin and C. W.Nan , J. Appl. Phys. , 033918 (2012)[9] C. Pettiford, J. Lou, L. Russell, and N. X. Sun, App.Phys. Lett. , 122506 (2008)[10] F. Zighem, D. Faurie, S. Mercone, M. Belmeguenai andH. Haddadi, J. Appl. Phys. , 073902 (2013)[11] M. Weiler, A. Brandlmaier, S. Geprägs, M. Altham-mer, M. Opel, C. Bihler, H. Huebl, M. S. Brandt, R. Grossand S. T. B. Goennenwein, New Journal of Physics , 043913(2011)[13] C. Bihler, M. Althammer, A. Brandlmaier, S. Geprägs,M. Weiler, M. Opel, W. Schoch, W. Limmer, R. Gross,M. S. Brandt and S. T. B. Goennenwein, Phys. Rev. B , 045203 (2008)[14] A. Brandlmaier, S. Geprägs, M. Weiler, A. Boger, M.Opel, H. Huebl, C. Bihler, M. S. Brandt, B. Botters, D.Grundler, R. Gross and S. T. B. Goennenwein, Phys.Rev. B , 104445 (2008)[15] M. Belmeguenai, H. Tuzcuoglu, M. S. Gabor, T. Petrisor,Jr., C. Tiusan, D. Berling, F. Zighem, T. Chauveau, S.M. Ch!erif, and P. Moch, Phys. Rev. B , 184431 (2013)[16] Amilcar Bedoya-Pinto, Marco Donolato, Marco Gobbi,Luis E. Hueso and Paolo Vavassori, App. Phys. Lett. ,062412 (2014)[17] X. Zhang, Q. Zhan, G. Dai, Y. Liu, Z. Zuo, H. Yang, B.Chen, and R.-W. Li, J. Appl. Phys.113