Broadband Ferromagnetic Resonance of Magnetic Insulator Thin Films and Bilayers: Effect of Overlayer on Spin Dynamics

Transcription

1 Broadband Ferromagnetic Resonance of Magnetic Insulator Thin Films and Bilayers: Jimmy Shi, Riverside STEM Academy, California, USA Dr. Igor Barsukov (Advisor, University of California, Riverside, USA)

2 Abstract Magnetic resonance is a powerful technique to probe fast dynamics of spins such as nuclear spins or electron spins. In ferromagnetic materials, the dynamics of electron spins occurs in the gigahertz (GHz) frequency range or on the sub-nanosecond time scale. In this project, we measure ferromagnetic resonance (FMR) using a versatile broadband microwave apparatus to study microwave absorption in yttrium iron garnet (YIG) thin films and bilayers of YIG with a non-magnetic overlayer called topological insulator containing heavy elements such as bismuth (Bi). We employ an alternating magnetic field modulation and lock-in detection technique to enhance the sensitivity of the microwave signal, which is absorbed by thin film samples. The modulation amplitude dependence of the FMR linewidth is studied, and the real FMR linewidth is obtained by taking the zero-modulation amplitude limit. After careful analysis of the FMR data on the bare YIG thin film and bilayer, we find a very large effect of the overlayer on resonance linewidth and the Gilbert damping parameter, which reveals a new spin relaxation mechanism on the sub-nanosecond time scale. 2

3 Innovation Statement This submitted paper is the research results under the guidance of the advisor. To the best of my knowledge, the paper does not contain research results that have been published or written by other people or teams, except as specifically noted and acknowledged in the text. If there is any dishonesty, I am willing to bear all the relevant responsibilities. Participant:Jimmy Shi Advisor: Igor Barsukov 3

4 Contents Abstract.2 1. Introduction Resonance Magnetic resonance Ferromagnetic resonance Questions to address in this project Experimental Technique and samples Experimental Results and Data Analysis FMR spectra and Lorentzian curve fitting Modulation amplitude effect FMR resonance field H r vs. frequency f: the Kittel equation FMR linewidth vs. frequency f: Gilbert damping parameter 3.5 Discussion Conclusion Acknowledgements References..18 4

5 1. Introduction 1.1 Resonance Resonance is a ubiquitous physical phenomenon in nature. It occurs when an external oscillating driving force acting on an oscillator system causes its amplitude to increase at some specific oscillation frequency. The resonance phenomenon exists in a variety of periodic systems including mechanical resonance, acoustic resonance, electromagnetic resonance, nuclear spin resonance (NMR), and ferromagnetic resonance (FMR). The last one is the subject of this project. Figure 1. A periodic push force to a swing can make the amplitude of the swing increase over time, i.e. resonance. The frequency of the periodic push force has to match the natural frequency of the swing in order to reach mechanical resonance. A simple example in mechanics is a swing or pendulum (Fig. 1) subjected to a periodic force. Figure 2. Solution of driven damped harmonic oscillator with different damping. The vertical axis represents oscillation amplitude and horizontal axis is the frequency of the driving force with respect to the natural frequency of the oscillator. Smaller damping forces give rise to sharper resonance peaks. The frequency of the push must match the frequency of the swing, called the natural frequency, in order to have resonance. In mechanics, this can be modeled by a forced harmonic oscillator described by Newton s second law. The tension of the rope and gravity together act as a restoring force. In a real situation, the swing must encounter a frictional force due to air drag. In the absence of any push, the friction will cause the swing to stop over some time. Therefore, the 5

6 initial mechanical energy is dissipated. This damping force, opposite to the direction of the motion and proportional to the velocity, can be added to the equation of motion of the forced harmonic oscillator. The solution shows the resonance behavior of driven damped harmonic oscillators 1 (Fig. 2). The vertical axis represents the amplitude of the oscillator, and the horizontal axis is the frequency of the driving force normalized by the natural frequency of the oscillator 0, and represents different degrees of damping. Several features can be seen in this figure. First, a peak occurs in the amplitude for a range of damping parameters at the same frequency, 0 (true for weak damping). Second, the greater the damping, the smaller the maximum amplitude, indicating more energy dissipation. Third, the greater the damping, the broader the resonance peak. This means that resonance can occur over a certain range of frequency (lightly damped regime) near the natural frequency. Universal resonance curve. The resonance intensity can be generally represented by a universal resonance curve (Fig. 3): This is the Cauchy distribution function, or the Lorentzian function. is the full width at half maximum. As shown in Fig. 3, it is a Lorentzian with the full width at half maximum of 2 ( =1). If x represents frequency, then the function represents a resonance peak centered at x 0, the natural frequency, and with the damping related to the full with at half maximum,. The stronger the damping is, the smaller the value is, or the narrower the resonance linewidth becomes. Quality factor Q. Q is defined as: Figure 3. Lorentzian function (up to a factor of ) with =2.. A large Q indicates a narrow resonance width and a long decay time compared with the period of the oscillations. 1.2 Magnetic resonance According to electricity and magnetism, an electric current is just moving charge. When the charge is circulating around a center like an electron around the nucleus in an atom, the current forms a closed loop which has a corresponding magnetic moment, IA, with I being the current and A the area of the current loop. The magnetic moment is a vector and its direction is Figure 4. Pauli and Bohr, two great physicists, watch a spinning top. 6

7 defined to be perpendicular to the current loop by the right hand rule. When the current loop is placed in an external magnetic field but is not strictly perpendicular to it, there is a torque acting on the current loop that causes the normal of the current loop to precess around the external magnetic field. The precession of the current loop or the magnetic moment has its natural frequency called its Larmor frequency that is directly proportional to the magnetic field. The stronger the magnetic field is, the faster it precesses. The precession motion of the magnetic moment in an applied magnetic field is like a spinning top in a gravitational field (Fig. 4). The precession cannot go forever because of damping. To sustain the precession motion, a periodic driving force is needed just as in the mechanical cases. However, here the driving force should be of electromagnetic nature. Under an electromagnetic radiation field with the same frequency as the Larmor frequency 2, magnetic resonance occurs. Microscopic particles such as electrons and nuclei have magnetic moments. For nuclei, the magnetic resonance is called nuclear magnetic resonance (NMR) 3, a powerful tool for detecting protons which is widely used in chemistry and biology. Fig. 5 shows a giant magnet in Birmingham, UK for NMR experiments. For protons, the resonance frequency is 900 MHz under a 21.2 tesla field. Just as spinning protons, spinning electrons also have a magnetic moment. Due to the much smaller mass of electrons, the magnetic resonance for Figure 5. A superconducting magnet that supplies strong magnetic field for 900 MHz NMR. electron spin magnetic moment occurs at a much higher frequency for the same magnetic field strength. For electrons, the resonance frequency is in the microwave range (~ GHz range). 1.3 Ferromagnetic resonance (FMR) In ferromagnetic substances such as the refrigerator magnets and bar magnets, electron spin magnetic moments are locked together to form a giant permanent magnetic moment, or a large current loop, which has fixed south and north poles. Similar to single electron spin magnetic moment, this collective giant moment also precesses around an effective magnetic field direction with a natural frequency similar to but not equal to the Larmor frequency of single electrons. When microwave radiation is present, the precessing moment absorbs the energy from the microwave and is driven to resonance if the frequency matches the natural frequency. Similar to driven damped oscillators, there is also a damping action. In ferromagnetic materials, this damping is called Gilbert damping. The equation of motion is then a dynamic one 4 for magnetization M which is defined as the magnetic moment per unit volume, driven by mechanical torques,, and, as shown in Fig. 6. The first torque pushes the 7

8 magnetization to precess around the effective field with a certain natural frequency. This motion does not dissipate the energy of the precession motion. The second torque pulls the magnetization toward the effective field and does negative work, therefore taking energy away. The second torque is the damping action. Because of damping, the magnetization spirals towards the direction of the effective field. Once the magnetization is fully aligned with, precession motion stops and both torques vanish. This only occurs when there is no resonance and the microwave radiation is not absorbed. However, if the microwave frequency matches the precession frequency, absorption takes place. The absorbed energy overcomes the damping and the magnetization can settle in a constant precession motion. Like in other resonances, the damping is related to the magnetic resonance linewidth and the quality factor as described earlier, but the physics behind the Gilbert damping is very different from mechanical oscillators or NMR. It is associated with particular types of microscopic processes that act like some friction to the magnetic moment of electrons. 1.4 Questions to address in this project Figure 6. Two torques acting on magnetization M: M H eff In this project, we address the following scientific questions. First, how do we accurately determine the Gilbert damping parameter through measuring the FMR linewidth? Can we manipulate the Gilbert damping parameter by controlling the microscopic processes? We choose yttrium iron garnet (YIG) as our sample. Strictly speaking, YIG is a ferrimagnet. Similar to a ferromagnet, there is a net magnetic moment, but microscopically, it is more complicated. For our study, we treat it as a ferromagnet. It has well-defined FMR. We then deposit an overlayer containing heavy elements such as Bi, and study the effect of the overlayer on the Gilbert damping. dm and M. The dt former drives precession and the latter pulls M toward H eff. 2. Experimental Technique and Samples The relevant frequency range for the ferromagnetic resonance of ferromagnetic materials is 1-10 GHz, or 3-30 cm in wavelength, in the laboratory accessible magnetic field range. A conventional way of measuring FMR in many laboratories is to use rigid rectangular waveguides and cavity to create a certain microwave standing-wave mode. Because the waveguides of certain dimensions work for a fixed wavelength or a fixed frequency, this technique is only limited to a single frequency and the resonance is achieved by sweeping the magnetic field that varies the natural frequency of the magnetic moment precession in the samples. 8

9 A versatile FMR technique which has been widely adopted for small samples is the broadband technique that uses strip lines called co-planar waveguide (CPW) to propagate microwaves of a broad range of frequencies. As shown in the block diagram (Fig. 7), the Microwave Source outputs microwaves of a selected frequency, that reaches the sample through the strip line CPW. The electromagnet, i.e. Magnet, produces both a direct current (DC) and an alternating current (AC) magnetic fields. The AC magnetic field has a very small oscillating amplitude compared to the strength of the DC field and its frequency is much smaller than that of the microwave. This oscillating field is used for AC modulation. What the AC modulation does is to generate an AC response from the sample which is measured by the Microwave Diode. The response from the sample is a combined one to both DC and AC magnetic fields. The amplitude of the AC response, therefore, depends on Figure the sensitivity 8. Photograph of the resonance curve at a given of the DC experimental field. The AC setup for this project detection is accomplished by using the Lock-in in Barsukov s lab Amplifier which rejects any noises in 5. the The sample (dark response signal with frequencies other than the square) sits on top of reference frequency of the the modulation. waveguide This lock-in detection technique (yellow) provides mounted a in much more sensitive detection the of small gap of signals the than DC detection without using electromagnet. the field modulation and is a powerful technique frequently used in physics experiments. The key component of the actual experimental setup in Dr. Barsukov s laboratory at the University of California, Riverside is shown in Fig. 8. The measurement system is fully automated. With this broadband FMR setup, spin dynamics of a variety of magnetic materials can be measured with extremely high efficiency. Since the AC amplitude is very small, for a small fixed AC amplitude, the AC response approximately records the derivative of the resonance with respect to the magnetic field. For FMR, the AC modulation technique outputs the derivative of the Lorentzian function, sometimes derivative of multiple Lorentzian functions at different frequencies, instead of the Lorentzian function itself or sum of multiple Lorentzian functions. The samples we choose for this project include: Figure 7. Schematic diagram of broadband FMR measurement setup. 9

10 (1) Yttrium iron garnet (YIG) thin film (~ 30 nm in thickness) grown by pulsed laser deposition. The sample ID is: YIG (2) Bilayer of YIG and a (Bi x Sb 1-x ) 2 Te 3 overlayer. The (Bi x Sb 1-x ) 2 Te 3 (BST) overlayer is 5 nm in thickness YIG/BST. Figure 9. Schematic drawing of atomic arrangement of neighboring Y, Fe and O atoms in Y 3 Fe 5 O 12. R 3+ refers to Y 3+ which is a rare earth element in YIG. Red arrows represent Fe 3+ magnetic moments. Arrows for Fe 3+ ions on a-site and d-site point opposite to each other. YIG is a ferrimagnetic insulator. Its chemical formula is Y 3 Fe 5 O 12, and each formula unit contains three Y 3+ ions, five Fe 3+ ions, and 12 O 2- ions. Solid garnet structure is quite complicated. As shown in Fig. 9, the five Fe 3+ ions occupy two different types of sites in the crystal: two on one type (a-site) and three on the other type (d-site). Fe 3+ carries a magnetic moment. It is known that the magnetic moment of Fe 3+ ions on the a-site points opposite to that on the d-site and four out of five Fe 3+ magnetic moments are just cancelled out. There is a net magnetic moment from one Fe 3+ ion in each formula unit. The net magnetic moments in all formula units in a solid are locked together and the macroscopic YIG sample shows a permanent magnetic moment like a refrigerator magnet. From many years of studies, it is known that the YIG crystal has very unique magnetic properties compared to any other magnets, i.e., it has an extremely low Gilbert damping parameter. In FMR experiments, high-quality YIG crystals show linewidth as narrow as a fraction of oersted (Oe), comparable with the earth magnetic field strength. This is two orders of magnitude narrower than that of most conducting ferromagnets. In the thin film form, however, the FMR linewidth is broader than that in the bulk crystal form of the same material. The high-quality YIG films for this project were grown by Y.W. Liu and V. Ortiz, graduate students at the University of California, Riverside (UCR) using pulsed laser deposition, a powerful technique for growing oxide thin films. These films show fairly narrow FMR linewidth (~ 20 Oe at 9.6 GHz) and low damping ( ~ 10-3 ) 6,7. In recent studies, researchers have found that the FMR linewidth of YIG films is broadened by having another material such as Pt adjacent to it 8. The broadening happens only if the adjacent layer is a material containing heavy elements. To study the effect of FMR linewidth broadening, it is imperative to first accurately determine the FMR linewidth of the same magnetic material before and after the deposition of the second material. YIG is an excellent choice because of its narrow FMR linewidth, as small changes in its linewidth can be more unambiguously detected. 10

11 On the other hand, it demands high-precision measurements of the linewidth. For standard DC measurements, the DC field steps need to be much smaller than the FMR linewidth so that the apparent linewidth from the measurements do not depend on the field step-size. Similarly, for AC modulation measurements, the amplitude of the AC magnetic field needs to be smaller so the apparent linewidth does not depend on the AC amplitude. It is the first objective of this project to study the AC modulation effect on the FMR linewidth of high-quality YIG films which have narrow linewidth. In the YIG/BST composite sample, the BST thin film (5 nm thick) overlayer was prepared by Y.W. Liu (UCR) and Y.F. Zhao at Pennsylvania State University (PSU), respectively. BST is a topological insulator (TI) that it contains a heavy element, Bi (Z=83). Therefore, the BST overlayer is expected to cause significant FMR linewidth broadening. The detailed microscopic picture is not fully established due to lacking experimental data. A main reason is that highquality YIG/BST samples are not readily available. More systematic experimental data are clearly needed at this point. It is the second objective of this project to study the effect of the overlayer containing heavy elements. 3. Experimental Results and Data Analysis FMR spectra of all samples have been taken at room temperature. For each frequency, a DC magnetic field is swept from 0 to 5000 Oe while the derivative signal is measured by the microwave diode. After one field sweep is finished, another discrete frequency is set, and the field sweep is repeated. The microwave frequency varies from 1 GHz to 15 GHz. To study the AC modulation amplitude dependence, the procedure is repeated for each modulation amplitude. For all measurements, the frequency of the AC modulation is fixed at khz. In addition, the FMR measurements are performed over a range of microwave output power, from 10 to 20 dbm. A suitable microwave output power is chosen to maximize the FMR signals without causing changes due to heating of the waveguide or the sample. 3.1 FMR spectra and Lorentzian curve fitting When the microwave frequency f is fixed and the external DC magnetic field is swept, the resonance intensity is also represented by a Lorentzian function 9 which is centered at H r with the full width at half maximum, H, i.e., ( ) ( ) ( ). With AC modulation, the measured output signal by the microwave diode is the derivative of the Lorentzian with respect to the magnetic field, or dp/dh, which is anti-symmetric about the resonance field H r. In practice, the derivative signal often contains a small symmetric component which can be isolated with curve fitting. Fig. 10 shows a representative FMR derivative spectrum taken on YIG with the AC modulation amplitude of ~ 0.5 Vrms (~ 6.5 Oe), microwave power of 10 dbm, and microwave frequency of 6 GHz. The symbols represent experimental data, and the dashed line is the Lorentzian fit, which does not include any other contributions. This simple Lorentzian 11

12 Figure 10. FMR derivative signal of YIG taken with the microwave frequency of 6 GHz and Lorentzian fit. The fitting only includes the derivative of the Lorentzian function. The linewidth and resonance field extracted from the fitting are 23.7 Oe and Oe, respectively. fit works extremely well. By fitting the equation shown in Fig. 10 (lower left corner) to the experimental data points taken with f=6 GHz, we know that resonance occurs at about H r = Oe, with a linewidth (full width at half maximum) of H=23.7 Oe. These are the two most important parameters for the driven damped precession motion of the YIG magnetization. We perform the same fitting procedure to all frequencies over a range of AC modulation amplitude and obtain fitting parameters for all conditions, some of which will be summarized and presented in the following sub-sections. We note that in some cases, the Lorentzian derivative function alone cannot fit the spectra well, which can be easily seen in the asymmetry between the peak and dip intensities at the resonance. In those cases, a symmetric function needs to be added to obtain a good fit. The full fitting function including the Lorentzian derivative can be found in ref. xx. 12

13 The same procedures are applied to measurement and analysis of other YIG and YIG/BST samples. Direct comparison between YIG and YIG/BST samples with exactly the same YIG film will be presented later. 3.2 Modulation amplitude effect As discussed earlier, the AC modulation technique measures the derivative response, but strictly speaking, it is only true under the limit of zero amplitude. For any finite AC amplitude, the derivative is averaged over the field range of the AC magnetic field. In other words, the response is smeared within the AC amplitude field window and therefore the resolution of the AC response is limited by the AC amplitude. Thus, the AC amplitude matters, especially when the Figure 11. Measured FMR linewidth as a function of the root-mean-square (rms) amplitude of AC modulation in volts. With increasing of the modulation amplitude, the apparent FMR linewidth becomes larger. The true FMR linewidth is the extrapolated value when the amplitude approaches zero. The dashed line is a polynomial fit, from which the zero-amplitude linewidth is extracted. FMR linewidth is very narrow. The smaller the AC amplitude, the more accurate the derivative signal at each DC field is, but the AC response detected by the lock-in amplifier is smaller so that the advantage of using AC modulation is then lost. On the contrary, if the AC amplitude is larger than the FMR linewidth, the average of the AC response would be zero and the lock-in signal is lost as well. A sensible AC amplitude range for good detection is as large as possible to 13

14 obtain strong signals, but much smaller than the FMR linewidth. When good spectra are obtained, reliable linewidth can be found as a function of the AC amplitude, as shown in Fig. 11. Then the limit of zero AC amplitude is taken to get the true FMR linewidth. A complete data set should include FMR spectra taken with different AC amplitude and the true linewidth is therefore obtained under all conditions. This correction becomes important for low frequency data. 3.3 FMR resonance field H r vs. frequency f: the Kittel equation FMR peak position is often called resonance field, denoted by H r. For magnetic thin films under an in-plane magnetic field, the resonance condition is given by the Kittel equation: ( ). (1) In this equation, is the angular frequency which is related to the frequency f by =2 f, is called the gyromagnetic ratio, and is the effective magnetization. Both and are treated as fitting parameters by fitting the Kittel equation to the frequency dependence of the resonance field. In magnetic thin films, the magnetization prefers to lie in the film plane to minimize the magnetostatic energy associated with magnetic poles. This is described by the socalled shape anisotropy energy, which makes the resonance frequency differ from the Larmor frequency, the precession frequency of electrons in free space, i.e.,. Figure 12. Microwave frequency vs. resonance field for YIG. Symbols are experimental data and line is the Kittel equation (equation 1) fit. The fitting works very well. An important fitting parameter here is πm eff. It is slightly smaller than πm s for YIG. 14

15 The above Kittel equation describes the magnetic resonance condition for an infinitely large thin film. In the absence of any other anisotropy than the shape anisotropy as discussed above, is simply equal to the saturation magnetization of the magnetic thin film,. If there is another anisotropy besides the shape anisotropy, out-of-plane or in-plane which makes the magnetization easier or harder point to the normal direction of the film plane, then also contains the effect of that anisotropy. In other words, FMR is capable of measuring magnetic anisotropy by fitting the microwave frequency vs. the resonance field. If and therefore the differs from (1760 G for YIG), then it indicates there is an addition anisotropy energy. For example, if >, there is an in-plane anisotropy which favors the magnetization of the YIG film to lie in the film plane. On the other hand, if <, it indicates that the anisotropy favors the magnetization orientation to lie out of the film plane. Fig. 12 shows the resonance field data for YIG and the Kittel equation fit. The resonance field data for each microwave frequency f was extracted from the FMR spectrum as discussed in 3.1. The frequency range of the experiments is from 1 GHz to 13 GHz, within which the resonance field falls in the range of the electro-magnet. The resonance field and frequency data are tabulated and plotted. A least-square fitting was performed using a computer program. It is obvious that the experimental data are described by the Kittel equation extremely well. From the fitting, we find = Oe, which is smaller than. The difference between the two quantities indicates some anisotropy that favors magnetization out-of-plane, which may rise from the strain in YIG. The detailed mechanism to account for this difference requires more studies. 3.4 FMR linewidth vs. frequency f: Gilbert damping parameter An important objective of this project is to study the effect of an overlayer on the spin dynamics of YIG film. In this sub-section, we will highlight the difference in FMR linewidth between the bare YIG film and the YIG film with a topological insulator overlayer, YIG/BST. Here BST is short for (Bi 0.26 Sb 0.74 ) 2 Te 3 thin film. The BST overlayer thickness is only 5 nm. The FMR linewidth itself is not an intrinsic property of the oscillator, but the quality factor is. For a given oscillator, its quality factor is fixed. In general, the higher the microwave frequency, the broader the resonance linewidth. However, in practical thin films with defects, even for the linewidth does not vanish, but remains finite at. The linewidth depends on the frequency f in the liner fashion:. (2) Here is the Gilbert damping parameter. Clearly if is zero, would be the quality factor. In general, both the Gilbert damping parameter and are important parameters. From the FMR 15

16 linewidth as a function of the microwave frequency f, we can extract both the Gilbert damping parameter and. Fig. 13 is the FMR linewidth vs. frequency f, for both YIG and YIG/BST. After the nonlinear correction which is attributed to the two-magnon effect 10, the linewidth in both samples shows good linearity up to ~13 GHz. From the linear region, we can extract the slope of the curve. From the fitting parameter, obtained from the Kittel equation fitting, we calculate the Gilbert damping parameter for this YIG sample. It is found to be. in YIG. With a 5 nm thick BST overlayer, the Gilbert damping parameter is increased to., by a factor of four! Roughly speaking, by adding a 5 nm thick BST layer, the spin relaxation speed (or the damping time is shortened) is quadrupled! The other parameter only increases by a small amount, from 3.01 Oe to 4.01 Oe, both are much smaller than that for ferromagnetic metals. 3.5 Discussion There are several questions raised from this project which will be addressed in future research. 16

17 (1) Heavy elements vs. atomic arrangements in BST. The striking increase in the Gilbert damping parameter by adding only a 5 nm thick BST reveals a very important role of BST in influencing spin dynamics in the underlying YIG. Although some modest increase was reported by adding a thin layer of heavy metal such as platinum, here we observe a 400% increase in the Gilbert damping parameter by replacing only 26% Sb atoms with Bi, suggesting an unusual effect of Bi atoms, the heaviest element in BST, or the BST special property such as the way different atoms are organized. In this YIG/BST sample, the BST overlayer was grown by molecular beam epitaxy and the structure was monitored by electron diffraction which indicates an expected crystalline structure. In order to address the question about heavy elements vs. special structure, it would be interesting to systematically prepare samples with a fixed concentration but a varying degree of crystalline order. Such samples may be designed and prepared by a simpler growth method. (2) Effect of Bi concentration. Similar to what was just discussed, samples with different Bi concentrations will provide more important information to address the previous question. (3) Comparison with other heavy metal overlayers. In this project, a YIG/platinum sample was prepared, but the sample quality was not satisfactory. For future research, more experiments will be carried out with different overlayers, such as platinum and gold. These experiments will provide further evidence about the effect of heavy elements. 4. Conclusion By applying broadband FMR technique with AC modulation, we have accurately measured the FMR linewidth by extrapolating the modulation amplitude to zero. Using this method, we have studied the effect of the BST topological insulator overlayer on the YIG magnetic insulator thin film. From accurately determined ferromagnetic linewidth and the resonance field over a wide range of microwave frequencies, we have found a significant FMR line broadening and corresponding enhancement in the Gilbert damping parameter by using only a 5 nm thick BST overlayer. The physical mechanism behind this enhancement is currently not clear. Much more experimental and theoretical studies are needed to understand the new mechanism. In addition to the interesting physics, the significantly increased spin relaxation process through nanoscale material design is important for potential technological applications. 5. Acknowledgements Jimmy Shi acknowledges Dr. Barsukov for providing the research opportunity in his laboratory and for his supervision of this project. Many thanks to Dr. Barsukov s graduate students, Bassim Arkook, Xian Wang, Arezoo Etesamirad, and Rodolfo Rodriguez for their training, guidance and technical assistance throughout the project; and Yawen Liu, Victor Ortiz, and Yifan Zhao for preparing YIG and YIG/BST samples for this project. 17

18 6. References Vonsovskii, S. V. (2013). Ferromagnetic Resonance: The Phenomenon of Resonant Absorption of a High-Frequency Magnetic Field in Ferromagnetic Substances. Elsevier. ISBN C. Tang, M. Aldosary, Z.L. Jiang, H.C. Chang, B. Madon, K. Chan, M.Z. Wu, J. E. Garay, and J. Shi, Applied Physics Letters 108, (2016). 7. Y. Sun, Y.-Y. Song, H. Chang, M. Kabatek, M. Jantz, W. Schneider, M. Wu, H. Schultheiss, and A. Hoffmann, Applied Physics Letters 101, (2012). 8. Y.Y. Sun, H.Y. Chang, M. Kabatek, Y.-Y. Song, Z.H. Wang, M. Jantz, W. Schneider, M.Z. Wu, E. Montoya, B. Kardasz, B. Heinrich, S. G. E. te Velthuis, H. Schultheiss, and A. Hoffmann, Physical Review Letters 111, (2013). 9. Z. Celinski l, K.B. Urquhart, B. Heinrich, Journal of Magnetism and Magnetic Materials (1997). 10. I. Barsukov, F. M. Romer, R. Meckenstock, K. Lenz, J. Lindner, S. Hemken to Krax, A. Banholzer, M. Korner, J. Grebing, J. Fassbender, and M. Farle, Physical Review B 84, (R) (2011). 18