Probing the A1 to L1(0) transformation in FeCuPt using the first order reversal curve method

The A 1- L 1 0 phase transformation has been investigated in (001) FeCuPt thin ﬁlms prepared by atomic-scale multilayer sputtering and rapid thermal annealing (RTA). Traditional x-ray diffraction is not always applicable in generating a true order parameter, due to non-ideal crystallinity of the A 1 phase. Using the ﬁrst-order reversal curve (FORC) method, the A 1 and L 1 0 phases are deconvoluted into two distinct features in the FORC distribution, whose relative intensities change with the RTA temperature. The L 1 0 ordering takes place via a nucleation-and-growth mode. A magnetization-based phase fraction is extracted, providing a quantitative measure of the L 1 0 phase homogeneity. © 2014 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 Unported License. [http://dx.doi.org/10.1063/1.4894197] High magnetic are critical to key technologies such as ultrahigh density magnetic 1 and 2 ordered alloys in the L 1 0 are particularly 2 heat-assisted 3,4 bit-patterned

The A1-L1 0 phase transformation has been investigated in (001) FeCuPt thin films prepared by atomic-scale multilayer sputtering and rapid thermal annealing (RTA). Traditional x-ray diffraction is not always applicable in generating a true order parameter, due to non-ideal crystallinity of the A1 phase. Using the first-order reversal curve (FORC) method, the A1 and L1 0 phases are deconvoluted into two distinct features in the FORC distribution, whose relative intensities change with the RTA temperature. The L1 0 ordering takes place via a nucleation-and-growth mode. A magnetization-based phase fraction is extracted, providing a quantitative measure of the L1 0 phase homogeneity. © 2014 Author(s). All  High magnetic anisotropy materials are critical to key technologies such as ultrahigh density magnetic recording 1 and permanent magnets. 2 Among them, ordered FePt alloys in the L1 0 phase are particularly sought after. They are ideal candidate materials for the emerging >1 Terabit/in. 2 heatassisted magnetic recording (HAMR) media 3,4 or bit-patterned media, [5][6][7] as well as high energyproduct permanent magnets, 6,8,9 as they possess some of the largest anisotropy values known. 10 However, the highly desirable properties are associated with the tetragonal L1 0 phase. A critical challenge has been the elevated annealing temperature (typically 600 • C-650 • C) necessary to transform the as-deposited cubic A1 phase into the ordered L1 0 phase, 11 which is often incompatible with the rest of the manufacturing processes. Another key issue is the stringent requirement on switching field distribution (SFD) for ultrahigh density HAMR media (<5%), 10,12 where any residual A1 phase could be detrimental. Therefore, a complete understanding of the low anisotropy A1 to high anisotropy L1 0 phase transformation is critical. In prior studies, an order parameter S is often used to characterize the degree of the L1 0 ordering, using properly corrected, integrated intensities of the x-ray diffraction (XRD) peaks, since differences in lattice occupation are manifested in the structure factor. [13][14][15] This is an elaborate, time consuming process, yet its applicability still has significant limitations, e.g., for textured samples or materials with poor crystallinity, and for samples with tiny amount of minority phases.
We have recently demonstrated a convenient synthesis route to achieve (001) oriented FeCuPt films, using atomic-scale multilayer sputtering (AMS) and rapid thermal annealing (RTA) at 400 • C for 10 s, which is significantly more benign compared to earlier studies. 16 Magnetic properties of these films can be conveniently tuned with Cu content to achieve high anisotropy, large saturation magnetization, and moderate Curie temperature, a combination highly desirable for HAMR media. In this work, we have employed the first order reversal curve (FORC) method [17][18][19][20] to qualitatively and quantitatively investigate the A1-L1 0 phase transformation, particularly the RTA temperature dependence. We show that the L1 0 ordering takes place via a nucleation-and-growth mode during the non-equilibrium synthesis, and the magnetic phase fraction extracted by FORC is a convenient and complementary measure of the L1 0 ordering, which directly captures distributions in the relevant magnetic characteristics.
Atomic-scale multilayer films with nominal structure of [Fe(0.9 Å)/Cu(x)/Pt(1.4 Å)] 16 were grown by DC magnetron sputtering from elemental targets on amorphous SiO 2 (200 nm)/Si substrates in a vacuum chamber with base pressure of 7 × 10 −7 Torr. The Cu thickness (x) is varied to adjust the stoichiometric ratio (e.g., 0.4 Å for Fe 39 Cu 16 Pt 45 ). The introduction of Cu has been shown to selectively replace Fe atoms and help to tune the material properties to be desirable for HAMR applications. 16 Subsequently the films were annealed by RTA at various temperatures (10 s rise + 10 s dwell time) in 1 × 10 −5 Torr vacuum using infrared (IR) heating lamp with wavelength of 400-1100nm. 21 Due to its small band gap of 1.1 eV, the Si substrate readily absorbed the IR light, in contrast to the amorphous SiO 2 , which has a much larger 8.9 eV band gap. The thin FePt was mainly heated through thermal conduction across the SiO 2 barrier. As a result, the quick thermal expansion of the Si substrate exerts significant tensile stress on the atomic-scale multilayers of FePt across the SiO 2 and transfers the thermal energy to assist the L1 0 ordering as well as (001) orientation of the film. 16,21,22 After RTA the films were capped with a thin layer of Ti (20 Å) to prevent oxidation. For simplicity the rest of the paper will mainly focus on a series of Fe 39 Cu 16 Pt 45 samples (referred to as FeCuPt hereafter), while results from other samples are provided in the supplementary material. 23 X-ray diffraction was performed using a Bruker D8 thin film x-ray diffractometer with Cu K α radiation, and used for phase identification following procedures outlined previously. 15 Film microstructure and morphology were characterized by electron diffraction, transmission electron microscopy, and atomic force microscopy. Magnetic measurements were performed using vibrating sample magnetometry (VSM) at room temperature with the field applied perpendicular to the films, unless otherwise noted. To investigate detailed magnetization reversal, FORC measurements were performed as follows: [17][18][19][20] From positive saturation the magnetic field was swept to a reversal field H R , where the magnetization, M(H,H R ), was measured under increasing applied field, H, back to positive saturation, tracing out a FORC. The process was repeated for decreasing H R until negative saturation was reached. The normalized FORC-distribution was then extracted, 24 where M S is the saturation magnetization. Alternatively the FORC distribution could be represented in another coordinate system defined by local coercivity H C = (H − H R )/2 and bias field H B = (H + H R )/2. XRD patterns for the Fe 39 Cu 16 Pt 45 films are shown in Fig. 1. For RTA temperatures T RTA < 350 • C there are no appreciable FeCuPt peaks, indicating poor initial crystalline ordering. At T RTA ≥ 350 • C, both (001) and (002) peaks are clearly observed. The presence of the (001) peak indicates the establishment of the L1 0 phase, as it is forbidden in the A1 phase. The order parameter S is calculated from the I (001) /I (002) ratio to quantify the degree of L1 0 ordering, where I (001) and I (002) are the integrated intensity of the (001) and (002) peak, respectively, corrected for absorption, Lorentz factor, Debye-Waller factor, and angular dependent atomic scattering factors. [13][14][15] The order parameter is normalized to the calculated maximum value, S Max = 1 − 2 , where = 0.05 is the variation in the (FeCu):Pt atomic ratio from 50:50. The S/S Max is determined to be 1.0, 0.97, and 0.97 for T RTA of 350 • C, 375 • C, and 400 • C, respectively, seemingly suggesting a high degree of L1 0 ordering at these temperatures.
However, at T RTA < 350 • C, the absence of (001) or (002) peak makes it impractical to extract the L1 0 ordering parameter using this approach. The XRD order parameter is dependent on (I L1o 001 ) + (I A1 001 )/(I L1o 002 + I A1 (002) ); the (001) peak is forbidden in the A1 phase, (I A1 001 ) = 0, thus S/S Max will scale between 0 for the fully A1 phase to unity for the fully L1 0 phase. In the present case, no (002) peak is observed for the A1 phase, thus (I A1 002 ) = 0, and S is only dependent on I L1o (001) /I L1o (002) , which is identical to S Max . Therefore S does not vary with the phase transformation and cannot be used in its usual context to measure the L1 0 ordering. More generally, when the sample's initial crystalline ordering is non-ideal (e.g., the (002) intensity changes substantially during annealing), the order parameter S is no longer applicable in gauging the L1 0 ordering extent.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://aplmaterials.aip.org/about/rights_and_permissions Downloaded to Transmission electron microscopy (TEM) and electron diffraction show that all samples have crystalline grains (see the supplementary material 23 ); at T RTA = 325 • C, both rounded and elongated grains are observed, with an average size of ∼6 nm; at T RTA = 400 • C, only elongated grains are observed, with an average size of ∼12 nm, which are consistent with the effects of the tensile stress during the RTA. It is the crystalline nature of the films that allows us to use the conventional S parameter to gauge the degree of ordering. Cross-sectional TEM studies on a similar series of Fe 52 Pt 48 samples (i.e., Cu content is zero) illustrate that during the phase transformation, residual disordered A1 regions exist only in certain parts of the otherwise (001) oriented L1 0 films (Fig. 2), rather than homogeneous dispersion over numerous sites throughout the films. However, it is difficult to quantify the L1 0 phase fraction from TEM alone.
Room temperature hysteresis loops in the out-of-plane (OOP) and in-plane (IP) geometries for the Fe 39 Cu 16 Pt 45 samples annealed by RTA at 300 • C-400 • C are shown in Fig. 3. After RTA at 300 • C and 325 • C, the OOP major loops are nearly closed with negligible remanence, while the IP loops are sharp with large remanence and small coercivity, indicating an out-of-plane hard axis and suggesting that the films are primarily in the low anisotropy A1 phase. As the RTA temperature T RTA is increased to 350 • C, the OOP loop shows a gradual increase in remanence and coercivity, still showing largely hard-axis behavior, while the IP loop shows a sudden increase in coercivity and decrease in remanence. A more distinct change is observed for RTA at 375 • C and 400 • C, where the OOP loop exhibits easy-axis behavior with remanence approaching unity and a much enhanced coercivity, 25 and the IP loop has the smallest remanence, suggesting the film is primarily L1 0 ordered with out-of-plane anisotropy. The T RTA -dependent remanence, coercivity, and saturation This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://aplmaterials.aip.org/about/rights_and_permissions Downloaded to The two distinct FORC features directly correspond to the low and high anisotropy A1 and L1 0 phase, respectively. Their coexistence at intermediate T RTA of 325 • C and 350 • C, while the major loop only shows the hard-axis reversal, demonstrates the capability of FORC in deconvoluting magnetic phases. Furthermore, the evolution of these two FORC features under increasing T RTA reveals details of the A1-L1 0 phase transformation mechanism. The fact that these two features remain well separated in the FORC distribution (in H C − H B coordinates), and their relative intensity changes under increasing T RTA , is indicative of a nucleation-and-growth mechanism of the L1 0 phase. 26 Under this mechanism, annealing changes the relative ratios of the magnetically hard and soft phases, primarily modifying the intensity of the corresponding FORC features. This is consistent with the TEM image shown in Figure 2, which illustrates certain residual A1 "pockets" in the matrix of L1 0 films. Alternatively, had the phase transformation with T RTA been realized through a uniform growth mode, we would have expected a gradual shift of a single FORC feature as a whole, from low to high coercivity, and more extensive A1 regions in the TEM view. Thus the FORC technique is able to uniquely identify and separate contributions from the hard and soft phases, despite the single phase appearance of the major loop.
Furthermore  27 Thus, it can be used to quantitatively determine the amount of magnetic phases. 28,29 By selectively integrating the normalized FORC distribution (Eq. (1)) over the horizontal feature corresponding to the L1 0 phase, compared to the saturation magnetization of the entire sample, we can extract a magnetizationbased L1 0 phase fraction. Since the saturation magnetization of the A1 (818 emu/cm 3 ) and L1 0 (790 emu/cm 3 ) phases are nearly identical, the magnetic phase fraction can be directly correlated to the structural phase fraction. 30,31 In addition, it is possible to reconstruct the major hysteresis loops of each phase.
The evolution of the L1 0 phase fraction in Fe 39 Cu 16 Pt 45 determined by FORC with T RTA is shown in Fig. 5, indicating a gradual increase in the L1 0 ordering up to 350 • C, followed by an abrupt ordering near 375 • C. This is consistent with the kinetic ordering temperature reported earlier. 26,32 For comparison, the XRD order parameters are also included in Figure 5, which show good agreement only for highly L1 0 ordered 400 • C sample. Similar trends are also found in another series of Fe 28 Cu 27 Pt 45 films (see the supplementary material 23 ). While the structural order parameter is not always readily available (such as the case here for the 300 • C and 325 • C samples) or reliable (depending on capturing the proper crystal planes of the A1 phase), the magnetization-based phase fraction provides a complementary and direct measure of the L1 0 This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://aplmaterials.aip.org/about/rights_and_permissions Downloaded to phase homogeneity, which is critical to SFD and eventual application of such materials in HAMR media.
In conclusion, we have investigated the A1-L1 0 phase transformation in (001) FeCuPt thin films prepared by a non-equilibrium strain-assisted synthesis approach. The A1 and L1 0 phases manifest themselves as two distinct features in the FORC distribution, whose relative intensities change as the RTA temperature increases from 300 • C to 400 • C. The L1 0 ordering takes place via a nucleationand-growth mode. Traditional x-ray diffraction suggests an unrealistic order parameter due to poor initial crystalline ordering of the A1 phase. An alternative magnetization-based L1 0 phase fraction is extracted. The FORC method not only sheds insight into the L1 0 ordering mechanism in these ultrathin films, but also provides a quantitative measure of the L1 0 phase homogeneity. This approach is applicable to other L1 0 materials such as MnAl 33 and MnGa 34 that are being keenly pursued for rare-earth-free permanent magnets, as well as magnetic phase separation studies in general. This work has been supported by the NSF (Grant No. DMR-1008791). Work at NTHU has been supported in part by the Hsinchu Science Park of Republic of China under Grant No. 101A16.