Tuning magnetic properties of magnetoelectric BiFeO3–NiFe2O4 nanostructures

ARTICLE IN PRESS Journal of Magnetism and Magnetic Materials 321 (2009) L5–L9

Contents lists available at ScienceDirect

Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm

Letter to the Editor

Tuning magnetic properties of magnetoelectric BiFeO3–NiFe2O4 nanostructures S.P. Crane a,, C. Bihler b, M.S. Brandt b, S.T.B. Goennenwein c, M. Gajek d, R. Ramesh a,d a

Department of Materials Science and Engineering, University of California, Berkeley, CA 94720, USA ¨t Mu ¨ nchen, D-85748 Garching, Germany Walter Schottky Institut, Technische Universita Walther-Meissner-Institut, Bayerische Akademie der Wissenschaften, D-85748 Garching, Germany d Department of Physics, University of California, Berkeley, CA 94720, USA b c

a r t i c l e in fo

abstract

Article history: Received 4 August 2008 Available online 4 October 2008

Multifunctional thin film nanostructures containing soft magnetic materials such as nickel ferrite are interesting for potential applications in microwave signal processing because of the possibility to shrink the size of device architecture and limit device power consumption. An essential prerequisite to future applications of such a system is a firm understanding of its magnetic properties. We show that nanostructures composed of ferrimagnetic NiFe2O4 pillars in a multiferroic BiFeO3 matrix can be tuned magnetically by altering the aspect ratio of the pillars by depositing films of varying thickness. Magnetic anisotropy is studied using ferromagnetic resonance, which shows that the uniaxial magnetic anisotropy in the growth direction changes sign upon increasing the film thickness. The magnitude of this anisotropy contribution can be explained via a combination of shape and magnetostatic effects, using the object-oriented micromagnetic framework (OOMMF). The key factors determining the magnetic properties of the films are shown to be the aspect ratio of individual pillars and magnetostatic interactions between neighboring pillars. & 2008 Elsevier B.V. All rights reserved.

Keywords: Magnetoelectric Ferroelectric Ferrimagnetic Anisotropy Nanostructure Perovskite Spinel Ferrite

1. Introduction Research into magnetoelectric complex oxides has intensified over the last decade as investigators seek to study and develop these materials for applications in a wide variety of electronic devices as materials that enable electric control of magnetic properties [1]. Specifically, nanocomposite thin films consisting of a hard-magnetic spinel ferrite (CoFe2O4) and a perovskite ferroelectric (BaTiO3, BiFeO3 (BFO)) are one interesting approach for developing such materials and have shown well-behaved phase separation and both ferroelectric and ferrimagnetic properties [2–5]. The use of such a nanostructured system in particular provides a new approach for improving device properties, allowing ideal electric tunability of the magnetization orientation due to strain-induced magnetoelectric coupling between the ferroelectric and ferrimagnetic phases [6–9]. One limitation of previously studied systems has been the focus on CoFe2O4, a material with a large magnetocrystalline anisotropy component. This aspect prevents the ability to tune the magnetic properties by changing thickness, a valued capability in such thin film systems. A nanostructured multiferroic thin film with a soft-magnetic

 Corresponding author.

E-mail address: [email protected] (S.P. Crane). 0304-8853/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jmmm.2008.08.114

ferrite can be obtained using NiFe2O4 (NFO) as the magnetic component, which permits use in high-frequency communications as tunable signal filters, phase shifters, and resonators. Such a material provides a new approach to tunable microwave devices that enable both smaller devices and lower power consumption over current architectures [10]. Additionally, the use of multiferroic BFO, which is both ferroelectric and antiferromagnetic [11], might provide innovative device interaction mechanisms through both magnetoelectric and exchange interactions with NFO. Here we discuss the structural and magnetic properties of these nanostructured multiferroic thin films and show how the uniaxial contribution to the magnetic anisotropy can be systematically modified as a function of film thickness.

2. Experimental The nanostructured BFO–NFO thin films were grown via pulsed laser deposition (PLD) on a variety of oxide substrates, such as (0 0 1)-oriented SrTiO3 (STO) single crystals and (0 0 1)-oriented (LaAlO3)0.3(Sr2AlTaO6)0.7 (LSAT) single crystals. For PLD, a 248 nm KrF excimer laser was pulsed at a rate of 10 Hz with a fluence of 1.5 J/cm2 leading to a growth rate of 2.5 nm/min at an oxygen pressure of 100 mTorr. The molar composition of the single oxide target was 70 mol% BiFeO3 and 30 mol% NiFe2O4, which produced

ARTICLE IN PRESS 6

S.P. Crane et al. / Journal of Magnetism and Magnetic Materials 321 (2009) L5–L9

thin films with a composition of approximately 65 vol% BiFeO3 and 35 vol% NiFe2O4. The films were characterized structurally using X-ray diffraction (XRD) y–2y scans with a PANalytical X’Pert PRO PW3040/60 Diffractometer system. Topographical data were obtained using a Veeco Multimode scanning probe microscope and Nanoscope IV in tapping mode with ultrasharp Si cantilevers (Mikromasch NSC15/AIBS). Magnetic measurements were carried out using a Lakeshore Model 7402 vibrating sample magnetometer, with corresponding simulations calculated using the object-oriented micromagnetic simulator (OOMMF) [12], assuming bulk NFO magnetic properties and a 40 nm macrospin approximation, which was found to produce identical magnetic properties to simulations run down to 5 nm macrospins. Angledependent ferromagnetic resonance (FMR) studies at o/2p ¼ 9.3 GHz were carried out at room temperature in a custom built magnetic resonance spectrometer using a TE102 resonance tube and magnetic field modulation.

3. Results and discussion We have grown BFO–NFO films on STO crystals at substrate temperatures of 500, 600, 700, and 800 1C. XRD results for these samples are shown in Fig. 1(a). At a growth temperature of 500 and 600 1C, the y2y scans exhibit reflections consistent with multiple orientations, which is not desirable. Above 600 1C, the intensity from the corresponding non-(0 0 l) diffraction peaks is suppressed. However, films deposited at 800 1C exhibit a reduction in intensity of the pseudo-cubic 002 BFO diffraction peak, possibly due to the volatility of the Bi ion. As a result, all further experiments were performed on films deposited at 700 1C. Fig. 1(b) shows XRD data for films grown at this temperature on both substrates. The 004 NFO and 002 BFO diffraction peaks in these films are at identical positions demonstrating the capability to grow high-quality BFO–NFO composite thin films on both substrates. LSAT was selected as the substrate for further studies due to its well-matched lattice parameter to the deposited materials and lower dielectric constant and loss tangent at microwave frequencies, as compared to STO [13]. Fig. 1(c) shows a typical image obtained from atomic force microscopy (AFM) of BFO–NFO thin films on LSAT. The image shows rectangular NFO pillars with an average edge length l ¼ 80 nm. This edge length is found for all samples regardless of the film thickness. This aspect is used to alter the average pillar aspect ratio for a given sample by depositing films of varying thickness. In this way, we produce films with aspect ratios ranging from 1:2 length-to-width ratio for a 40 nm film to 10:1 ratio for an 800 nm film. The effect of the shape of the nanostructures on the magnetic properties will be discussed below. As described in detail in Ref. [14], the relative orientation of the NFO pillars in the BFO matrix is cube-on-cube /1 0 0SNFOJ/1 0 0SBFO, with the interfaces lying in the {11 0} planes. AFM furthermore shows that the NFO pillars protrude out beyond the BFO matrix by about 10 nm (on average), a feature that is also not affected by film thickness. This roughness is known to stem from the initial stages of the growth process [5]. Cross-section TEM confirmed that the pillars are quite uniform throughout the thickness of the film and that the BFO–NFO interface is fully strain relaxed [14]. Reciprocal space maps of the 103 peaks (not shown) for both 40 and 800 nm thick films show the same relative position of the NFO peak with respect to the substrate peak, which matches with the predicted position of the bulk lattice peak, confirming full relaxation for all thicknesses studied. Magnetization loops were measured in order to determine the thickness dependence of the magnetic properties. All films have a saturation magnetization MSE2.8  105 A/m, which corresponds

Fig. 1. (a) X-ray diffraction data for growth of BiFeO3–NiFe2O4 nanostructures on (0 0 1) SrTiO3 single crystal substrates at temperatures of 500–800 1C. The films grown at 700 1C show ideal structural properties, exhibiting crystalline heteroepitaxial BiFeO3 and NiFe2O4 phases with negligible secondary-phase formation. (b) X-ray diffraction data for BiFeO3–NiFe2O4 nanostructures on (0 0 1) SrTiO3, and (0 0 1) LSAT substrates grown at 700 1C. (c) Atomic force microscopy image of a typical film grown at 700 1C shows NiFe2O4 pillars (brighter) extending out of BiFeO3 matrix (darker). Note the in-plane BiFeO3–NiFe2O4 interface lies along {11 0}.

well with the room-temperature saturation magnetization for bulk NFO [15]. The out-of-plane [0 0 1] and in-plane [0 1 0] loops from 40 and 800 nm thick films are displayed in Fig. 2(a) and (b),

ARTICLE IN PRESS S.P. Crane et al. / Journal of Magnetism and Magnetic Materials 321 (2009) L5–L9

7

Fig. 2. Experimental and simulated magnetic hysteresis loops along the [0 0 1] (red) and [0 1 0] (black) directions for (a) 40 nm thick and (b) 800 nm thick films. The simulated loops are shown with a solid line, and the corresponding experimental loops are shown superimposed with (&). The shapes of the experimental and simulated curves correlate well, showing an accurate representation of the demagnetizing fields, but the simulations fail to accurately reproduce all the experimental coercive fields as discussed in the text.

normalized to the saturation magnetization. The most striking feature of these loops is the large effective demagnetizing field measured in each of them, given by the shearing of the hysteresis loop in a given direction. This shearing is found to be due to a combination of cubic and shape anisotropies, as well as magnetostatic interactions between neighboring pillars. We will investigate the magnitude of the different contributions through micromagnetic simulations. To identify the contributing factors determining the magnetization orientation in the NFO pillars, we now consider the following possible sources of anisotropy: magnetocrystalline, magnetostriction, exchange, shape, and magnetostatic effects. The hysteresis loops were taken along [0 1 0] and [0 0 1], two crystallographically equivalent directions, meaning that magnetocrystalline anisotropy should not play a role in comparing magnetic properties in these directions. As discussed earlier, the films are fully strain relaxed, implying the effect of any strain-induced inverse magnetostriction should be negligible [16]. We also investigated a possible exchange interaction between the NFO pillars and the surrounding BFO matrix. In field cooling experiments a sample was cooled through the antiferromagnetic ordering temperature TNE370 1C of BFO in an applied magnetic field m0H ¼ 1 T. If the BFO and NFO phases were interfacially exchange coupled, a shift of the hysteresis curve would have been expected [17], which was not observed, even at temperatures as low as 10 K. This indicates that in our system interfacial exchange does not contribute to the overall anisotropy. The remainder of this section will discuss the large role of shape and magnetostatic

Table 1 List of contributions to the uniaxial anisotropy field in the growth direction for 40 and 800 nm thick films Anisotropy source Shape m0Ms(NxNz) P Magnetostatic 2 (ExMSEzMS)/Ms 001 Sum 2Keff /M Experimental

40 nm 89 mT 60 mT 29 mT (HA) 26 mT (HA)

800 nm 155 mT 133 mT 22 mT (EA) 20 mT (EA)

Shape and demagnetizing magnetostatic effects are the only non-negligible effects, whose sum matches the values for the uniaxial anisotropy field in the growth direction found experimentally via FMR reasonably well. The 40 nm thick film exhibits a uniaxial anisotropy contribution with a magnetic hard axis (HA) in the growth direction, whereas the 800 nm thick film exhibits a uniaxial anisotropy contribution with a magnetic easy axis (EA) in the growth direction.

effects on the overall anisotropy of the system. The shape of the pillars themselves determines the demagnetizing fields and hence is expected to contribute significantly to the anisotropy. As discussed in Ref. [18], the demagnetizing factors for rectangular prisms are determined by comparing the demagnetizing effect of the free surfaces normal to the Cartesian axes, which is found to scale with the aspect ratio of the prism. The shape anisotropy can be estimated as the difference between the demagnetizing factors for two orthogonal directions, NX and NZ, for rectangular prisms with a 1:2 aspect ratio (40 nm thick film) and a 10:1 aspect ratio (800 nm thick film), which are listed for both cases in Table 1. We can gauge the expected shape

ARTICLE IN PRESS 8

S.P. Crane et al. / Journal of Magnetism and Magnetic Materials 321 (2009) L5–L9

contribution from the pillar aspect ratio alone, which is given as NXNZ ¼ 0.490.26 ¼ 0.24 for the 40 nm thick film and NXNZ ¼ 0.060.47 ¼ 0.41 for the 800 nm thick film. The magnitude of this contribution to the overall anisotropy is significant for both film thicknesses. The change in sign indicates that it should be possible to evolve from a hard axis (HA) in the growth direction to an easy axis (EA) by adjusting film thickness alone. However, another important contribution to the anisotropy of the system is the dipole interactions between neighboring pillars, which act to increase magnetic frustration and to demagnetize the overall structure, as shown in similar magnetic nanostructure arrays by Farhoud et al. [19]. Due to the complexity of our nanostructure array, the micromagnetic simulator OOMMF was used to estimate the overall strength of this demagnetizing effect. Monochromatic images are loaded into the program that detail the strength of the magnetic response which are taken by converting AFM images taken of each respective film, restricting the ferromagnetic response to that from the NFO pillars. Simulation parameters are set, such as a 40 nm macrospin approximation and bulk NFO anisotropy constants, and an applied field is swept between 71 T, producing theoretical hysteresis curves. Superimposed on the experimental plots shown in Fig. 2 are the hysteresis loops simulated using OOMMF with the corresponding thicknesses and orientations discussed. A good agreement with the experimental data is observed with regard to overall shape of the loops, implying a good estimate of the demagnetizing fields. Another observable aspect is the poor estimate of the coercive field for the magnetically hard directions in each case. While the reason for this poor estimate is not well understood, it is not expected to affect the estimates of the demagnetizing fields because they were taken at saturated fields. The magnetostatic interaction is taken as the difference between the simulated demagnetizing fields found for two orthogonal directions (i.e. x and z), taken parallel and perpendicular to the film plane; this process is similar to the one used to estimate the shape anisotropy discussed earlier. The estimates of the relative magnetostatic fields for each film are also shown in Table 1, and in both cases this estimate is in the order of the shape contribution and acts as a counteracting force. When the total contributions are summed, one can see the anisotropy is expected to switch from an HA to an EA with increasing thickness but interpillar interactions act to lower the overall magnitude from that expected from shape effects alone. In order to confirm the predicted anisotropy values, the films were studied using FMR, which allows for the quantification of the anisotropy coefficients in magnetically frustrated systems. Fig. 3(a) and (b) shows the angular dependence of the FMR spectra for two BFO–NFO films with a film thickness of 40 and 800 nm, respectively. The spectra were each taken at different magnetic field orientations in the (1 0 0) plane as shown by the inset in Fig. 3(a). The red curves show the result of the simulation of the angular dependence of the FMR resonance field obtained from a free energy approach, which is thoroughly discussed in Ref. [20]. We note that these simulations also consistently explain the FMR spectra obtained for rotations of H in the (0 0 1) plane (not shown). For the simulation of the resonance fields for the 40 nm thick sample (Fig. 3(a)) we obtained a g-factor g ¼ 2.34, a uniaxial 001 anisotropy field in the growth direction 2Keff /M ¼ 26 mT, and a cubic anisotropy field 2KC1/M ¼ 50 mT. For the 800 nm thick 001 sample we obtained g ¼ 2.62, 2Keff /M ¼ 20 mT, and 2KC1/ M ¼ 50 mT. The peak-to-peak linewidth of the FMR spectra of the 40 nm thick sample is about m0DHpp55 mT, while the 800 nm thick sample exhibits m0DHpp80 mT, irrespective of the magnetic field orientation. The measured g-factors should be equal to the g-factor of the divalent magnetic ion on the octahedral site (in this case Ni2+) [21] due to the antiferromagnetic coupling of the

Fig. 3. Ferromagnetic resonance response for (a) the 40 nm thick (1:2 pillar aspect ratio) and (b) the 800 nm thick (10:1 pillar aspect ratio) BiFeO3–NiFe2O4 thin films. The angle-dependent measurements were performed varying the out-of-plane angle y. The difference of the resonance field for HJ[0 0 1] and HJ[0 1 0] or HJ[0 1¯ 0] in the two samples (indicated by the black arrow, with the thin vertical line representing the resonance field for HJ[0 1 0]) directly visualizes the change in the sign of the uniaxial anisotropy fields in the growth direction. The red curves show the result of the simulation of the angular dependence of the FMR resonance field obtained from a free-energy approach. Further details are discussed in the text.

tetrahedral and octrahedral sites in an ideal inverse spinel. This would result in g ¼ g(Ni2+ oct) ¼ 2.3 [22], which is in good agreement with the g-factor of the 40 nm thick sample. For the 800 nm thick sample, the larger g-factor might be indicative of a different ion distribution (Ni ions on tetrahedral sites would contribute a g-value of g(Ni2+ tetr) ¼ 3.5), but more careful studies would be required to adequately quantify the ion distribution using this method. The cubic anisotropy field 2KC1/M ¼ 50 mT in both samples is characteristic for bulk NFO [23,24], which verifies the quality of the spinel but is not expected to contribute to the uniaxial anisotropy, which is a measure of the preferential magnetization in such thin film systems. The difference of the resonance field for HJ[0 0 1] and HJ[0 1 0] or HJ[0 1¯ 0] in the two samples—indicated in Fig. 3(a) and (b) by the black arrow— directly visualizes the change in the sign of the uniaxial anisotropy fields in the growth direction. These results indicate that for an increasing pillar aspect ratio the nanostructures are

ARTICLE IN PRESS S.P. Crane et al. / Journal of Magnetism and Magnetic Materials 321 (2009) L5–L9

progressively acquiring an out-of-plane easy magnetic axis, consistent with the magnetization results shown in Fig. 2(a and b). This confirms that shape indeed plays a strong role in determining the anisotropy properties. The overall magnitude and sign of the 001 effective anisotropy in each case, 2Keff /M ¼ 26 mT for the 40 nm 001 thick sample and 2Keff /M ¼ 20 mT for the 800 nm thick sample, also agree well with the predicted anisotropy, implying an accurate estimate for the demagnetizing magnetostatic interactions between pillars.

4. Conclusion We have grown nanostructured multiferroic BiFeO3–NiFe2O4 thin films on different crystalline substrates using PLD. XRD spectra show optimal crystallinity for all samples studied, and a combination of AFM and TEM shows a clear phase separation into ferromagnetic NFO pillars embedded in a BFO matrix. Magnetic simulations estimate that the overall anisotropy is due to competing effects between shape and magnetostatic interactions. FMR analyses accurately confirm the predicted magnetic anisotropy of the two samples with prolate and oblate NFO pillars in terms of both sign and magnitude. The effective out-of-plane uniaxial anisotropy changes sign with increasing film thickness, demonstrating that film thickness can be used as a tool to control magnetic properties in magnetoelectric nanostructured thin films.

Acknowledgments This project is funded by an ONR MURI program under contract no. E-21-6RU-G4 at the University of California, Berkeley. The authors at the Walter Schottky Institut and the Walther-MeissnerInsitut are funded by Deutsche Forschungsgemeinschaft (DFG) through SFB 631 and SPP 1157, respectively. The first author would like to acknowledge the National Science Foundation for support via the graduate research fellowship program. The authors would also like to acknowledge the assistance of M.A. Philippine,

9

N. Balke, Q. He, and Y.-H. Chu for various characterization measurements and D. Carleton and M. Donahue for valuable discussions regarding OOMMF. References [1] V.E. Wood, A.E. Austin, Int. J. Magn. 5 (1974) 303. [2] H. Zheng, J. Wang, S.E. Lofland, Z. Ma, L. Mohaddes-Ardabili, T. Zhao, L. Salamanca-Riba, S.R. Shinde, S.B. Ogale, F. Bai, D. Viehland, Y. Jia, D.G. Schlom, M. Wuttig, A. Roytburd, R. Ramesh, Science 303 (2004) 661. [3] H. Zheng, J. Wang, L. Mohaddes-Ardabili, M. Wuttig, L. Salamanca-Riba, D.G. Schlom, R. Ramesh, App. Phys. Lett. 85 (2004) 2035. [4] H. Zheng, Q. Zhan, F. Zavaliche, M. Sherburne, F. Straub, M.P. Cruz, L.-Q. Chen, U. Dahmen, R. Ramesh, Nanoletters 6 (2006) 1401. [5] H. Zheng, F. Straub, Q. Zhan, P.-L. Yang, W.-K. Hsieh, F. Zavaliche, Y.-H. Chu, U. Dahmen, R. Ramesh, Adv. Mater. 18 (2006) 2747. [6] G. Srinivasan, E.T. Rasmussen, J. Gallegos, Yu.I. Bokhan, V.M. Laletin, R. Srinivasan, Phys. Rev. B 64 (2001) 214408. [7] M.I. Bichurin, D.A. Filippov, V.M. Laletsin, N. Paddubnaya, V.M. Petrov, Phys. Rev. B 68 (2003) 132408. [8] M.I. Bichurin, V.M. Petrov, G. Srinivasan, Yu.V. Kiliba, Phys. Rev. B (2002) 134404. [9] F. Zavaliche, H. Zheng, L. Mohaddes-Ardabili, S.Y. Yang, Q. Zhan, P. Shafer, E. Reilly, R. Chopdekar, Y. Jia, P. Wright, D.G. Schlom, Y. Suzuki, R. Ramesh, Nanoletters 5 (2005) 1793. [10] J.D. Adam, L.E. Davis, G.F. Dionne, E.F. Schloemann, S.N. Stitzer, IEEE Trans. Micro. Theory Tech. 50 (2002) 721. [11] G.A. Smolenskii, I.E. Chupis, Sov. Phys. Usp. 25 (1982) 475. [12] M.J. Donahue, D.G. Porter, OOMMF Users Guide, 1.0, 1999, NISTIR6367. [13] C. Zuccaro, M. Winter, S. Penn, N.M. Alford, P. Filhol, G. Forterre, N. Klein, Appl. Superconductivity 158 (1997) 69. [14] Q. Zhan, R. Yu, S.P. Crane, H. Zheng, C. Kisielowski, R. Ramesh, Appl. Phys. Lett. 89 (2006) 172902. [15] A.S. Boxer, Handbook of Microwave Ferrite Materials, Academic Press, NJ, 1965. [16] S. Chikazumi, Physics of Ferromagnetism, Oxford University Press, Oxford, UK, 1997. [17] J. Nogues, I.K. Schuller, J. Magn. Magn. Mater. 192 (1999) 203. [18] A. Aharoni, J. Appl. Phys. 83 (1998) 3432. [19] M. Farhoud, H.I. Smith, M. Hwang, C.A. Ross, J. Appl. Phys. 87 (2000) 5120. [20] C. Bihler, M. Kraus, H. Huebl, M.S. Brandt, S.T.B. Goennenwein, M. Opel, M.A. Scarpulla, P.R. Stone, R. Farshchi, O.D. Dubon, Phys. Rev. B 75 (2007) 214419. [21] J. Smit, H.P.J. Wijn, Ferrites, Philips Gloeilampenfabrieken, Eindhoven, 1959. [22] E.W. Gorter, Nature 173 (1954) 123. [23] J.K. Galt, B.T. Matthias, J.P. Remeika, Phys. Rev. 79 (1950) 391. [24] W.A. Yager, J.K. Galt, F.R. Merritt, E.A. Wood, Phys. Rev. 80 (1950) 744.