Continuously graded anisotropy in single „Fe53Pt47...100−xCux films

C. L. Zha, R. K. Dumas, Y. Y. Fang, V. Bonanni, J. Nogués, and Johan Åkerman Material Physics, Royal Institute of Technology (KTH), Electrum 229, 16440 Kista, Sweden NTNU NanoLab, Norwegian University of Science and Technology (NTNU), 7491 Trondheim, Norway Department of Physics, University of Gothenburg, 41296 Gothenburg, Sweden Institució Catalana de Recerca i Estudis Avançats (ICREA) and Centre d’Investigació en Nanociència i Nanotecnologia (ICN-CSIC), Campus Univ. Autònoma de Barcelona, 08193 Bellaterra, Spain

The roadmap to magnetic recording media with areal bit densities beyond 1 Tbit/in. 2 is plagued by the so-called magnetic recording "trilemma" where the thermal stability and signal-to-noise ratio ͑SNR͒ must be simultaneously balanced with the writabilty of a given bit. 1,2 To maintain both a sufficient SNR and high bit density, the volume of the individual grains that constitute a given bit must approach the superparamagnetic limit. While high anisotropy ͑K u ͒ materials, such as L1 0 FePt alloys, 3 would dramatically improve thermal stability, the magnetic fields required for switching such a high K u bit far exceed the capabilities of the write head. Tilted, 4,5 and exchange coupled composite ͑ECC͒ media, 6 as well as heat-assisted magnetic recording, 7 and microwave-assisted magnetic recording techniques, 8 have been proposed to lower the necessary writing field. An extension of ECC media is termed graded media where the anisotropy is varied, either discretely or continuously, 9 over many layers and is predicted to provide additional gains in writabilty over conventional bi-layer hard/soft ECC media. In graded media, the low K u layers act to reduce the switching field while the high K u layers preserve thermal stability.
Graded K u materials are also potentially useful in devices which rely on spin transfer torque ͑STT͒ phenomena, such as spin torque oscillators and STT magnetoresistive random access memory. 10,11 Recent work has shown a considerable reduction in the critical current densities necessary for the switching of a composite hard/soft free layer. 12,13 A free layer with a graded K u is the next step in further reducing the current density. Fabrication of graded media, in particular the realization of a continuous gradient, is challenging and has, until now, been based on multilayered structures. [14][15][16][17][18] In this letter, we report on a simple approach to fabricate continuously graded-K u single films based on the strong dependence of the A1 ͑low K u ͒ → L1 0 ͑high K u ͒ ordering temperature on the Cu content in ͑Fe 53 Pt 47 ͒ 100−x Cu x ͑x =0-30͒ films. 19 First, a compositional gradient is achieved by varying the Cu concentration from Cu-rich ͑Fe 53 Pt 47 ͒ 70 Cu 30 at the bottom to Cu-free Fe 53 Pt 47 at the top during film deposition, as schematically represented in Fig. 1 ͑left panel͒. The K u gradient is then realized after proper annealing. The Cu-rich regions transform from the as-deposited low K u cubic A1 phase into the high K u tetragonal L1 0 phase at a lower annealing temperature than the Cu-poor regions, thus establishing an K u gradient through the thickness of the film. Magnetic measurements support the creation of an K u gradient with reduced coercivity and moderate thermal stability.
The ͑Fe 53 Pt 47 ͒ 100−x Cu x ͑x =0-30͒ ͑called graded FePtCu hereafter͒ films ͑20 and 50 nm thick͒ were deposited at room temperature on thermally oxidized Si substrates by magnetron sputtering in a chamber with a base pressure of better than 5 ϫ 10 −8 Torr. During the co-sputtering of high purity Fe, Pt, and Cu targets the compositionally graded films were realized by gradually decreasing the sputtering power of the Cu gun from 35 to 0 W. Therefore, the bottom of the film is Cu-rich ͑Fe 53 Pt 47 ͒ 70 Cu 30 while the top is Cu-free Fe 53 Pt 47 . A compositionally uniform sample with the same total Cu content as the graded sample, ͑Fe 53 Pt 47 ͒ 85 Cu 15 ͑called uniform FePtCu hereafter͒, was deposited as a reference. To improve surface roughness, ͑111͒-texture, and lower the chemical ordering temperature of the L1 0 phase, the FePtCu films were a͒ Electronic mail: akerman1@kth.se. deposited on Ta͑6 nm͒/Pt͑3 nm͒. 20 A capping layer of Ta ͑5 nm͒ was deposited to prevent oxidation. The as-deposited uniform and graded samples were simultaneously annealed at T A ϭ300, 400, and 500°C for 35 min in vacuum. The composition of the films was determined by energy dispersive x-ray spectrometry and the crystallographic structure was investigated by x-ray diffraction ͑XRD͒ with Cu K ␣ radiation. Magnetization measurements were performed at room temperature with in-plane fields on an alternatinggradient magnetometer ͑AGM͒ and a home-built longitudinal magneto-optical Kerr effect ͑MOKE͒ setup.
The structural properties of the uniform and graded FePtCu samples are shown in Fig. 1͑a͒-1͑c͒ for T A ϭ300, 400, and 500°C. We clearly observe a strong ͑111͒ texturing for all annealing temperatures ͓Figs. 1͑a͒ and 1͑b͔͒. For both the uniform and graded samples, as the annealing temperature is raised, the intensity of the ͑111͒ diffraction peak progressively increases, indicating an improvement in the ͑111͒textured crystallization. Importantly, for both uniform and graded samples the ͑111͒ peak shifts to higher angles, Fig.  1͑c͒, with increasing annealing temperature. This, in turn, implies an enhancement of the chemical ordering degree of the L1 0 phase for higher annealing temperatures. 21 Note that the ͑111͒ peak is shifted to angles higher than that of fully ordered FePt ͑41.08°͒ due to the Cu dopant which has been shown to decrease the c lattice parameter. 19 While an increase in L1 0 order is found with increasing annealing temperature, how the structural A1-L1 0 ordering gradient ͑and thus the magnetic K u gradient͒ is manifested through the thickness of the film is not readily accessible from the XRD measurements.
Magnetic measurements, on the other hand, not only show enhanced L1 0 chemical ordering with T A but also provide unambiguous evidence of the graded nature of the properly annealed films. First, the major loop coercivity ͑H C ͒, Fig. 1͑d͒, is found to increase with T A , consistent with the XRD data and the improved high K u L1 0 order. Additional insight into the graded nature of the samples is found by analysis of the major hysteresis loops. Major loops measured using an AGM, Figs. 2͑a͒, 2͑c͒, and 2͑e͒, along with the derivatives of the descending branches of the major loops, Figs. 2͑b͒, 2͑d͒, and 2͑f͒, highlight the important role of proper T A in establishing the K u gradient.
At a low T A of 300°C, Figs. 2͑a͒ and 2͑b͒, both the uniform and graded samples remain relatively soft ͑note the field scale͒, consistent with the as-deposited, low K u A1 phase. The graded sample shows a small increase in H C compared to the uniform sample annealed at the same temperature indicating the onset of L1 0 ordering of the Cu-rich fractions of the film. The loop derivative is also significantly broader for the graded sample, which is also in agreement with the existence of a high K u phase and the incipient K u gradient in this film.
At an intermediate T A of 400°C, Figs. 2͑c͒ and 2͑d͒, the H C of both the graded and uniform films dramatically increases as a larger fraction of the films transforms into the high K u L1 0 phase. Interestingly, after annealing at 400°C, the graded film exhibits a fundamentally different reversal path compared to the uniform film. As we can clearly see from the derivatives, Fig. 2͑d͒, the graded sample exhibits an asymmetric peak with an extended switching shoulder. This shoulder corresponds to the softer portions of the film beginning to reverse first. In fact, the overall shape of the loop is reminiscent of simulated hysteresis curves of films with a graded K u that show an initial linear decrease in the magnetization ͓see arrow in Fig. 2͑c͔͒. 22 The initial linear decay of the magnetization can be associated with the formation of a partial domain wall through the film thickness that precedes switching. Actually the plateau in the major loop derivative ͓see arrow in Fig. 2͑d͔͒ directly indicates this initial linear decrease in the magnetization. Most importantly, the coercivity of the compositionally graded film, H C = 5.67 kOe, is smaller than the uniform film, H C = 7.21 kOe, which is also consistent with the existence of an K u gradient. Additionally, the remanence ͑M r / M s , where M r and M s are the remnant and saturation magnetizations, respectively͒ of the graded sample, M r / M s = 0.91, is larger than the uniform sample, M r/ M s = 0.79. This implies that the graded sample has a substantial fraction of soft phase.
Finally, after annealing at 500°C, Figs. 2͑e͒ and 2͑f͒, the uniform and graded samples essentially become identical. We find that the H C , the location of the ͑111͒ peak, the loop derivatives, and even the overall shape of the loops become nearly indistinguishable for the uniform and graded samples. We, therefore, conclude that a significant amount of interdiffusion occurs at this elevated T A . It is also not surprising that the compositionally uniform and graded samples become virtually identical in that they both contain the same total Cu concentration.
To assess the thermal stability of our graded samples we have carried out dynamic H C measurements. The dynamic H C , H C,dyn , is the field required to bring the magnetization to zero after a characteristic time, t o , and functionally obeys Sharrock's equation. 23 In our calculations, an attempt frequency f o =5ϫ 10 9 Hz is used and we can presume n =1. 22,24,25 From this analysis, the thermal stability factor, defined as ␤ ϵ ⌬E o / k B T, and time independent intrinsic H C , H o , can then be extracted. Here ⌬E o is the energy barrier and k B T is the thermal activation energy. The H C,dyn measure-  The AGM, which measures the total magnetic moment, is not able to directly determine how the K u varies through the thickness of the film. To probe the graded K u we have measured surface magnetic properties of a 50 nm thick graded FePtCu film using the MOKE setup where the finite penetration depth of light ͑ϳ30 nm in FePt͒ can be exploited. The MOKE and AGM in-plane loops are shown in Fig. 3͑a͒ for a sample annealed at 300°C. Clearly the MOKE loop has a smaller coercivity, H C,MOKE = 1.31 kOe, than the AGM loop, H C,AGM = 1.92 kOe and exhibits a different reversal path. As expected, the MOKE loop, which only probes the upper portions of the film, is significantly softer because the Cu-poor regions near the top of the film have not fully transformed into the high K u L1 0 phase at this particular T A. These depth sensitive results, along with the observed in-plane easy axis ͑induced by the thin film shape anisotropy͒, confirm that the anisotropy varies through the film thickness instead of forming, e.g., random L1 0 clusters embedded in an A1 matrix. However, after annealing at 400°C both the AGM and MOKE loops, Fig. 3͑b͒, have the same H C , H C,AGM = H C,MOKE = 7.03 kOe, and nearly identical loop shape. This indicates that after annealing at 400°C the 50 nm graded FePtCu film has achieved a uniform K u distribution, similar to the 20 nm thick sample annealed at 500°C. As reported for uniform FePtCu films, 26 the L1 0 chemical ordering process also depends on film thickness and is actually promoted in thicker films.
In summary, we propose and demonstrate a simple procedure to establish a continuous K u gradient through the thickness of FePtCu films. The K u gradient is realized after proper annealing of compositionally graded films. In addition, similar L1 0 FePtCu ͑111͒ films have been reported to have tilted anisotropies. 27,28 This fact, combined with the advantages of graded K u may result in a very versatile magnetic structure, which may not only be appealing for magnetic recording but also for STT applications. 29,30 We anticipate that the experimental procedure presented here can be further improved and extended to other, e.g., perpendicular and/or granular, material systems where a graded anisotropy is desired.