Influence of Ta insertions on the magnetic properties of MgO/CoFeB/MgO films probed by ferromagnetic resonance

Transcription

1 Typeset with apex.cls <ver > Applied Physics Express Influence of Ta insertions on the magnetic properties of MgO/CoFeB/MgO films probed by ferromagnetic resonance Maria Patricia Rouelli Sabino, Sze Ter Lim, Michael Tran Data Storage Institute, A*STAR (Agency for Science, Technology and Research), 5 Engineering Drive 1, Singapore We show by vector network analyzer ferromagnetic resonance measurements that low Gilbert damping α down to can be achieved in perpendicularly magnetized MgO/CoFeB/MgO thin films with ultra-thin insertions of Ta in the CoFeB layer. While increasing the number of Ta insertions allows thicker CoFeB layers to remain perpendicular, the effective areal magnetic anisotropy does not improve with more insertions, and also comes with an increase in α. Perpendicular magnetic anisotropy (PMA) is the key to further scale down spin transfer torque magnetoresistive random memory (STT-MRAM) devices, as it allows satisfying two key requirements: low critical current I c0 and high thermal stability, the latter being proportional to the energy barrier E b between the two stable magnetic states. A metric commonly used to account for both requirements is the spin torque switching efficiency defined as E b /I c0, which for a Stoner- Wohlfarth model is given by?) ( /4e) (η/α), whereαis the Gilbert damping parameter andη is the spin polarization factor which is related to the tunnel magnetoresistance ratio (TMR) by η=[tmr(tmr+2)] 1/2 /[2(TMR+1)]. It thus becomes evident that for high switching efficiency, one has to decreaseαwhile keeping TMR high. Magnetic tunnel junctions (MTJs) based on CoFeB/MgO systems are well-known to provide high TMR?) and has recently been shown to possess PMA attributed to the CoFeB/MgO interface. A Ta layer is usually placed adjacent to CoFeB in order to induce proper crystallization necessary for PMA and high TMR.?) In Ta/CoFeB/MgO systems however, spin pumping to Ta increasesα.?) Moreover, the CoFeB layer also needs to be ultra-thin (typically less than 1.5nm) in order to exhibit PMA.?) In order to improve thermal stability without increasing critical current, increasing the effective anisotropy energy density K e f f is desired. One approach to address these issues is the use of double MgO structures, i.e. both the barrier layer and capping layer straddling the free layer are made of MgO. Sato et. al?) demonstrated that a perpendicular easy axis can be attained in MgO/CoFeB/Ta/CoFeB/MgO stacks and also found that I c0 is comparable for single and double MgO MTJs, alluding to lowerα. Indeed, Konoto et al.?) reported α down to in MgO/FeB/MgO stacks; however, the stacks investigated by Konoto et al. were in-plane and did not have the Ta layer commonly used in practical free layers with perpendicular anisotropy. 1/??

2 In this work, we investigate the influence of Ta insertions within the CoFeB layer of MgO/CoFeB/MgO films by magnetometry and vector network analyzer ferromagnetic resonance (VNA-FMR) measurements. In these structures, spin pumping is suppressed by MgO layers while PMA may be achieved by having two CoFeB/MgO interfaces providing interface anisotropy and the insertion of n extremely thin Ta layers (0.3 nm) inside the CoFeB layer to aid with crystallization, thus also allowing larger total CoFeB thickness.?) Two sample series were deposited by magnetron sputtering on SiO 2 substrates with seed layers of Ta 5/TaN 20/Ta 5 in an ultrahigh vacuum environment (all thicknesses in nm). The stack configuration of the two sample series are: (1) MgO 3/ CoFeB 1.0/Ta 0.3/CoFeB /MgO 3 ( single insertion ) and (2) MgO 3/CoFeB 1.0/Ta 0.3/CoFeB /Ta 0.3/CoFeB 1.0/MgO 3 ( double insertion ), where the CoFeB composition is Co 40 Fe 40 B 20 (at%). Two other sample series were grown as reference: (a) MgO 3/CoFeB /MgO 3 ( zero insertion ), and (b) seed/ CoFeB /MgO 3 ( single MgO ). For all double MgO samples, an ultrathin CoFeB layer below the bottom MgO is also deposited for good growth of MgO. We have confirmed from separate measurements that this layer does not contribute to the magnetic signal. All samples are capped with 15 nm of Ta for protection and are annealed post-growth at 300 C for 1h in vacuum. While 3nm MgO is too thick for practical use in MTJs, it is chosen to ensure continuity of the MgO layers and lessen the influence of layers beyond it.?,?,?) The Ta insertion layer thickness is in the regime allowing strong ferromagnetic coupling between the CoFeB layers.?) Magnetization measurements were performed using an alternating gradient magnetometer (AGM). PMA improves with the doubling of the MgO/CoFeB interface and with increasing number of Ta insertions n, as shown in Fig.??(a) for samples with the same total nominal CoFeB thickness t nom = 2.5nm. We also confirm that we cannot obtain perpendicular easy axis in double MgO structures without Ta insertions.?) The double insertion sample, on the other hand, exhibits large out-of-plane remanence as shown in the inset of Fig.??(a). A coercive field less than 0.01mT also seen in the inset is typical of CoFeB films with PMA.?,?) It is known that Ta can create a magnetically dead layer (MDL) when in the proximity of a magnetic layer.?) In order to obtain the thickness of the MDL, we plot the magnetic moment per area of the double MgO samples versus the nominal thickness as shown in Fig.??(b). We obtain an MDL thickness t MDL of 0.24±0.09nm and 0.7±0.1nm for single and double insertion samples, respectively. These thicknesses are similar to the total Ta insertion thickness in the respective series, and is consistent with the picture of CoFeB intermixing with Ta to produce a magnetically dead volume. From the slopes, we obtain M S values of 1.12±0.05MA/m for the single insertion series and 1.13±0.05MA/m for the double insertion series. Vector network analyzer ferromagnetic resonance (VNA-FMR) was used to measure the effective anisotropy field and damping parameter of the samples. In the VNA-FMR setup, the samples were placed face down on a coplanar waveguide and situated in a dc magnetic field up to 1.2T applied 2/??

3 Magnetization (a.u.) (a) t nom = 2.5nm Magnetic Field (mt) M/A (A m 2 /m 2 ) x nm nm single MgO double MgO, n=0 double MgO, n=1 double MgO, n=2 (b) t nom (nm) Fig. 1. (Color online) (a) Out-of-plane AGM loops for samples with single MgO (red squares), zero insertion (purple inverted triangles), single insertion (blue circles), and double insertion (green triangles), all with total nominal CoFeB thickness t nom = 2.5nm. Inset shows a low field out-of-plane loop for the same double insertion sample. (b) Magnetic moment per unit area (M/A) as a function of the total nominal CoFeB thickness for single and double insertion samples. Linear fits are shown as solid lines. The intercept values corresponding to the dead layer are labeled for each series. perpendicular to the film plane. The transmission scattering parameter S 21 was measured at a specific frequency while the dc field was swept. For each sweep, the real and imaginary parts of the resonance response were fitted simultaneously using the complex susceptibility equation χ(h)= M e f f (H M e f f + i H 2 ) (H M e f f ) 2 ( 2π f gµ B ) 2+ i H(H Me f f ) where f is the frequency of the ac field, M e f f = M S HK, H is the full width at half maximum, H K is the anisotropy field perpendicular to the plane, g is the spectroscopic splitting factor,µ B is the Bohr magneton, and is the reduced Planck s constant. Non-magnetic contributions to the S 21 parameter and a linear time-dependent drift of the instruments were taken into account during the fit. We note that only one resonance peak is observed within the range studied. A representative fit of the susceptibility data is shown in Fig.??(a) for a double insertion sample with t nom = 2.5nm. In using Eq.??, a value of g=2 is first assumed to obtain values for M e f f and H, which does not affect the final result. For each frequency, a resonance field µ 0 H res ( f )= 2π gµ B f+µ 0 M e f f (2) according to the Kittel equation is calculated and plotted against frequency as in Fig.??(b). A linear (1) 3/??

4 S 21 Real (a.u.) S 21 Imag. (a.u.) (a) Data Fit (b) Applied Field (mt) Resonance Field (mt) Data Fit (c) Frequency (GHz) Linewidth (mt) Fig. 2. (a) Real and (b) imaginary parts of the S 21 parameter obtained from VNA-FMR measurements for a double insertion sample with t nom = 2.5nm at 12GHz while a perpendicular dc magnetic field is swept. The lines are fits to an expression using Eq.??, taking non-magnetic contributions to S 21 and a linear drift into account. (c) Field-swept linewidth and resonance fields for the same sample as a function of frequency. The linear fits described in the text are used to extract H Ke f f (= M e f f ) andα. fit, now with g and M e f f as fitting parameters is then performed. K e f f can be calculated from the effective anisotropy field H Ke f f (= M e f f ) as K e f f constant corresponds to a perpendicular easy axis. = H Ke f f M S /2, noting that a positive anisotropy To obtainα, we perform a linear fit of the measured FMR linewidth as a function of frequency to µ 0 H( f )= 4π α gµ B f+µ 0 H 0 (3) where H 0 is the inhomogeneous linewidth broadening and the value of g is used is the fitted value from Eq.??. We note that two-magnon scattering contributions to the linewidth are eliminated owing to the perpendicular measurement configuration.?) Such a fit is shown in Fig.??(b). Only data points taken well beyond the saturation field for each sample are used in the fit, and asymptotic analysis as described in Ref.?) for the accessible frequency range was further performed. We define an effective thickness t e f f = t nom t MDL and show the calculated K e f f t e f f (to which the E b is proportional) for both sample series in Fig.??(a). We find that for t e f f > 2nm, double insertion samples have higher K e f f t e f f than single insertion samples for the same t e f f. However, the maximum K e f f t e f f achieved for both single and double insertion series is only comparable to the maximum K e f f t e f f measured in our thinnest single MgO sample (t e f f = 1.0 nm). 4/??

5 To understand this further, we consider the different contributions to K e f f t e f f which reads K e f f t e f f = n int K i + (K v µ 0M 2 S 2 )t e f f (4) where n int is the number of CoFeB/MgO interfaces, K i and K v are the interfacial and volume anisotropy constant, respectively, and the demagnetizing energy is given by the MS 2 term. We assume that any interfacial anisotropy from the Ta/CoFeB interface is negligible.?) n int K i is commonly derived from the y-intercept of a K e f f t e f f plot. As it is possible that for CoFeB thickness below 1.0nm, K i is degraded due to Ta reaching the CoFeB/MgO interface,?) we only consider the linear region of the curve during the fit. We obtain 1.06±0.04mJ/m 2 and 1.3±0.1mJ/m 2 for single and double insertion samples, respectively. The larger K i of double insertion samples may be tied to the degree of crystallization at the CoFeB/MgO interface.?) The amount of Ta relative to CoFeB is of a higher percentage in double insertion samples, so that Boron may be absorbed more effectively than in single insertion samples. The proximity of the Ta insertion layer to the MgO interface may have also influenced the crystallization process. In the double insertion case, the Ta insertion layer is always separated from the MgO layer by only 1nm of CoFeB, while for single insertion samples, one MgO interface can be further away. A detailed study on the amount, proximity and profile of Ta would be necessary to clarify this effect. Turning our attention toα, we identify a single MgO sample (t e f f 0.8nm) and a single insertion sample (t e f f 1.3nm) with comparable K e f f t e f f 0.2mJ/m 2. We immediately notice thatαof the single insertion sample is around two times lower than the single MgO sample. This drastic decrease inαmay be attributed to the suppression of spin pumping by the MgO layers straddling both sides of the precessing magnet.?,?) Indeed, measurement of zero insertion samples show a low value ofα = ± comparable to the bulk value of Co 40 Fe 40 B 20?,?) and no thickness dependence (represented by the purple dashed line in Fig.??(b)). However, a decrease inαcan still be seen in both single and double insertion series. One possible reason is alloying of CoFeB and Ta, as Ta is known to readily intermix with CoFeB,?) and higher damping may be expected from CoFeBTa alloys.?) The relative percentage of CoFeBTa alloy decreases with increasing CoFeB thickness, coinciding with the α decrease. This picture is also consistent with the jump in α from single to double insertion samples, i.e. there is more CoFeBTa alloy due to more Ta insertions. It may also be possible that spin pumping to the Ta insertion layer occurs, as in the case of the Pd interlayer in CoFe/Pd multilayers.?) The complexity of our system, however, prevents us from utilizing a simple multilayer model. For one, the middle CoFeB layer (in the double insertion case) may have different properties from the CoFeB layers adjacent to MgO, since crystallization of CoFeB occurs from the MgO interface?) which the middle CoFeB has no contact with. The degree of Ta intermixing also depends on the deposition order and will be different across the structure.?) At this point, we are unable to discriminate the mechanism behind the damping behavior. It may be worthwhile to study the use of CoFeBTa alloys as interlayers to possibly have more control over the amount and distribution of Ta in the stack.?) 5/??

6 K eff t eff (mj/m 2 ) x single MgO double MgO, n=1 double MgO, n= MgO / CoFeB / MgO t eff (nm) Fig. 3. (a)k e f f t e f f and (b)αversus effective CoFeB thickness t e f f obtained from field-swept VNA-FMR measurements of single insertion samples (blue circles) and double insertion samples (green triangles). The solid lines in (a) are linear fits. The purple dashed line in (b) corresponds to the meanαvalue averaged over all zero insertion samples, which was found to be constant within error across the whole thickness range studied. In conclusion, we have demonstrated PMA and low damping in perpendicular double MgO structures. Although the maximum K e f f t e f f in double MgO samples does not exceed that of the single MgO sample, damping can be twice lower than a single MgO stack of comparable K e f f t e f f. Increasing the number of insertions allows larger thicknesses to remain perpendicular, but does not increase the maximum K e f f t e f f. Moreover, increasing the number of insertions does not provide a direct way of improving the spin torque efficiency, as it is accompanied by an increase inα. Acknowledgement 6/??

7 We express gratitude for support from the A*STAR Graduate Academy SINGA Program. 7/??

8 References 1) J. Z. Sun, S. L. Brown, W. Chen, E. A. Delenia, M. C. Gaidis, J. Harms, G. Hu, Xin Jiang, R. Kilaru, W. Kula, G. Lauer, L. Q. Liu, S. Murthy, J. Nowak, E. J. OSullivan, S. S. P. Parkin, R. P. Robertazzi, P. M. Rice, G. Sandhu, T. Topuria, and D. C. Worledge: Phys. Rev. B 88 (2013) ) S. Yuasa: J. Phys. Soc. Jpn. 77 (2008) ) D. C. Worledge, G. Hu, David W. Abraham, J. Z. Sun, P. L. Trouilloud, J. Nowak, S. Brown, M. C. Gaidis, E. J. OSullivan, and R. P. Robertazzi: Appl. Phys. Lett. 98 (2011) ) X. Liu, W.Zhang, M. J. Carter, and G. Xiao: J. Appl. Phys. 110 (2011) ) S. Ikeda, K. Miura, H. Yamamoto, K. Mizunuma, H. D. Gan, M. Endo, S. Kanai, J. Hayakawa, F. Matsukura, and H. Ohno: Nat. Mater. 9 (2010) ) H. Sato, M. Yamanouchi, S. Ikeda, S. Fukami, F. Matsukura, and H. Ohno: Appl. Phys. Lett. 101 (2012) ) M. Konoto, H. Imamura, T. Taniguchi, K. Yakushiji, H. Kubota, A. Fukushima, K. Ando, and S. Yuasa: Appl. Phys. Express 6 (2013) ) V. B. Naik, H. Meng, and R. Sbiaa: AIP Advances, 2 (2012) ) D. D. Lam, F. Bonell, S. Miwa, Y. Shiota, K. Yakushiji, H. Kubota, T. Nozaki, A. Fukushima, S. Yuasa, and Y. Suzuki: J. Kor. Phys. Soc. 62 (2013) ) M. P. R. G. Sabino, S. T. Lim, and M. Tran: (unpublished). 11) V. Sokalski, M. T. Moneck, E. Yang, and J.-G. Zhu: Appl. Phys. Lett 101 (2012) ) H. Sato, M. Yamanouchi, S. Ikeda, S. Fukami, F. Matsukura, and H. Ohno: IEEE Trans. Mag. 49 (2013) ) G. Malinowski, K. C. Kuiper, R. Lavrijsen, H. J. M. Swagten, and B. Koopmans: Appl. Phys. Lett. 94 (2009) ) S. Y. Jang, C.-Y. You, S. H. Lim, and S. R. Lee.: J. Appl. Phys. 109 (2011) ) N. Mo, J. Hohlfeld, M. ul Islam, C. S. Brown, E. Girt, P. Krivosik, W. Tong, A. Rebei, and C. E. Patton: Appl. Phys. Lett. 92 (2008) ) J. M. Shaw, H. T. Nembach, T. J. Silva, and C. T. Boone: J. Appl. Phys. 114 (2013) ) Y. Miura, M. Tsujikawa, and M. Shirai: J. Appl. Phys. 113 (2013) ) N. Miyakawa, D. C. Worledge, and K. Kita: IEEE Magn. Lett. 4 (2013) 2. 19) S. Ikeda, R. Koizumi, H. Sato, M. Yamanouchi, K. Miura, K. Mizunuma, H. Gan, F. Matsukura, and H. Ohno: IEEE Trans. Mag. 48 (2012) ) O. Mosendz, J. E. Pearson, F. Y. Fradin, S. D. Bader, and A. Hoffmann: Appl. Phys. Lett. 96 (2010) /??

9 21) M. Oogane, T. Wakitani, S. Yakata, R. Yilgin, Y. Ando, A. Sakuma, and T. Miyazaki: Jpn. J. Appl. Phys. 45 (2006) ) Y.-H. Wang, W.-C. Chen, S.-Y. Yang, K.-H Shen, C. Park, M.-J. Kao, and M.-J. Tsai: J. Appl. Phys. 99 (2006) 08M ) J. O. Rantschler, R. D. McMichael, A. Castillo, A. J. Shapiro, W. F. Egelhoff, B. B. Maranville, D. Pulugurtha, A. P. Chen, and L. M. Connors: J. Appl. Phys. 101 (2007) ) J. M. Shaw, H. T. Nembach, and T. J. Silva: Phys. Rev. B 85 (2012) ) S. Mukherjee, F. Bai, D. MacMahon, C.-L. Lee, S. Gupta, and S. Kurinec: J. Appl. Phys. 106 (2009) ) K. Tsunoda, H. Noshiro, C. Yoshida, Y. Yamazaki, A. Takahashi, Y. Iba, A. Hatada, M. Nakabayashi, T. Takenaga, M. Aoki, and T. Sugii. IEDM (2012) /??