Annealing-Induced High Ordering and Coercivity in Novel L1 0 CoPt-Based Nanocomposite Magnets

: A novel class of quaternary intermetallic alloys based on CoPt is investigated in view of their interesting magnetic properties induced by the presence of hard magnetic L1 0 phase. A Co 48 Pt 28 Ag 6 B 18 alloy has been prepared by rapid solidiﬁcation from the melt and subjected to various isothermal annealing procedures. The structure and magnetism of both as-cast and annealed samples as well as the phase evolution with temperature are investigated by means of thermal analysis, X-ray, and selected area electron diffraction, scanning and high-resolution electron microscopy, and magnetic measurements. The X-ray diffraction (XRD) analysis shows that both the as-cast alloy and the sample annealed at 400 ◦ C (673 K) have a nanocrystalline structure where fcc CoPt phase predominates. Annealing at 473 ◦ C promotes the formation of L1 0 phase triggered by the disorder-order phase transformation as documented in the differential scanning calorimetry results. The sample annealed at 670 ◦ C (943 K) shows full formation of L1 0 CoPt as revealed by XRD. Magnetic measurements showed coercivity values ten times increased compared to the as-cast state. This conﬁrms the full formation of L1 0 CoPt in the annealed samples. Moreover, detailed atomic resolution HREM images and SAED patterns show the occurrence of the rarely seen (003) superlattice peaks, which translated into a high ordering of the L1 0 CoPt superlattice. Such results spur more interest in ﬁnding novel classes of nanocomposite magnets based on L1 0 phase.


Introduction
There has been a recent surge of interest in L1 0 CoPt-based alloys, mainly due to the hard magnetic properties of the tetragonal L1 0 phase: high magnetocrystalline anisotropy (10 MJ/m 3 ), high coercivity, high Curie temperature (450 • C) and high chemical stability combined with the lower temperature of L1 0 -phase formation compared to the case of FePt L1 0 phase. These alloys were recently suggested as alternatives to the rare-earth permanent magnets and as potential candidates for patterned media in heat-assisted or microwave-assisted magnetic recording [1][2][3][4][5]. The ordered L1 0 materials such as FePt and CoPt feature magnetization stability against thermal fluctuations due to their high magnetocrystalline anisotropy. However, there are many challenges involved in obtaining the desired microstructure to achieve these good magnetic properties.
It is worth noting that the equilibrium phase of the stoichiometric CoPt bulk alloys at room temperature is the face-centered cubic disorder A1 phase. Only after appropriate annealing can the fcc phase be transformed into the face-centered tetragonal ordered L1 0 phase, which is magnetically hard.
The disorder-order transformation and its effect on remanence and coercivity enhancement in bulk, isotropic CoPt magnets has been studied by Xiao et al. [6]. They found that only after annealing molten alloy has been solidified and re-melted 3 times more. A total amount of 5 g has been used for each sample. The rapid solidification of the melt was performed on a Bühler Melt Spinner SC from Edmund Bühler GmbH, Bodelshausen, Germany. The obtained melt was then purged onto the surface of a Cu wheel. The wheel has 40 cm in diameter and rotates during the synthesis at 2000 r/min. The obtained ribbons were about 30 µm thick, 2-3 mm wide and several decimeters long.
Differential scanning calorimetry (DSC) measurement was performed for the rapidly quenched alloys in the range of 300-950 K with a heating rate of 10 K/min. The as-cast samples were annealed at 400 • C (673 K), 473 • C (746 K) and 670 • C (943 K) for 1 h in vacuum. The crystallization processes were monitored using the differential scanning calorimetry (DSC). DSC thermo-analysis has been performed with a Setaram DSC 111 system, from SETARAM Instrumentation, Calure, France, with a range of heating rates between 5 and 50 K/min in temperatures up to 900 K in argon atmosphere.
We have checked by energy dispersion X-ray spectroscopy (EDX) the true chemical composition of all the as-cast samples, to observe the stoichiometry of the alloy and we found no noticeable difference between the nominal composition and the measured one.
Morphology and local atomic structure have been studied using scanning electron microscopy (SEM), high-resolution electron microscopy (HREM) and selected area electron diffraction (SAED). These studies have been carried out using a JEOL JEM ARM 200F electron microscope, from JEOL Ltd., Tokyo, Japan, operated at 200 kV. The specimens for microscopy have been prepared in plan-view orientation by ion milling at 7 • angle of incidence and 4 kV accelerating voltage with a Gatan PIPS installation.
The analyses of the crystal structure and the phase composition were carried out using XRD with Cu Kα radiation. The magnetic properties were measured using a with a Superconducting Quantum Interference Device (SQUID), from Quantum Design Europe GmbH, Darmstadt, Germany and the hysteresis loops were performed at room temperature up to 5 T applied field, parallel to the sample plane.

Results and Discussion
The phase evolution with temperature plays an important role in establishing the optimal properties of the nanocomposite magnets of L1 0 -type, since the L1 0 phase of interest is stabilized during high-temperature annealing. Therefore, to correctly identify the temperature of the disorder-order phase transformation, thermal analysis either by means of the Differential Scanning Calorimetry (DSC) or Differential Thermal Analysis (DTA) is required. We have performed DSC measurements in a temperature range from 300 K to 900 K with a heating rate of 10 K/min. on the as-cast CoPtAgB alloy. The obtained DSC curve is presented in Figure 1. molten alloy has been solidified and re-melted 3 times more. A total amount of 5 g has been used for each sample. The rapid solidification of the melt was performed on a Bühler Melt Spinner SC from Edmund Bühler GmbH, Bodelshausen, Germany. The obtained melt was then purged onto the surface of a Cu wheel. The wheel has 40 cm in diameter and rotates during the synthesis at 2000 r/min. The obtained ribbons were about 30 µm thick, 2-3 mm wide and several decimeters long. Differential scanning calorimetry (DSC) measurement was performed for the rapidly quenched alloys in the range of 300-950 K with a heating rate of 10 K/min. The as-cast samples were annealed at 400 °C (673 K), 473 °C (746 K) and 670 °C (943 K) for 1 h in vacuum. The crystallization processes were monitored using the differential scanning calorimetry (DSC). DSC thermo-analysis has been performed with a Setaram DSC 111 system, from SETARAM Instrumentation, Calure, France, with a range of heating rates between 5 and 50 K/min in temperatures up to 900 K in argon atmosphere.
We have checked by energy dispersion X-ray spectroscopy (EDX) the true chemical composition of all the as-cast samples, to observe the stoichiometry of the alloy and we found no noticeable difference between the nominal composition and the measured one.
Morphology and local atomic structure have been studied using scanning electron microscopy (SEM), high-resolution electron microscopy (HREM) and selected area electron diffraction (SAED). These studies have been carried out using a JEOL JEM ARM 200F electron microscope, from JEOL Ltd., Tokyo, Japan, operated at 200 kV. The specimens for microscopy have been prepared in planview orientation by ion milling at 7° angle of incidence and 4 kV accelerating voltage with a Gatan PIPS installation.
The analyses of the crystal structure and the phase composition were carried out using XRD with Cu Kα radiation. The magnetic properties were measured using a with a Superconducting Quantum Interference Device (SQUID), from Quantum Design Europe GmbH, Darmstadt, Germany and the hysteresis loops were performed at room temperature up to 5 T applied field, parallel to the sample plane.

Results and Discussion
The phase evolution with temperature plays an important role in establishing the optimal properties of the nanocomposite magnets of L10-type, since the L10 phase of interest is stabilized during high-temperature annealing. Therefore, to correctly identify the temperature of the disorderorder phase transformation, thermal analysis either by means of the Differential Scanning Calorimetry (DSC) or Differential Thermal Analysis (DTA) is required. We have performed DSC measurements in a temperature range from 300 K to 900 K with a heating rate of 10 K/min. on the ascast CoPtAgB alloy. The obtained DSC curve is presented in Figure 1.   For the sample Co 48 Pt 28 Ag 6 B 18 one can observe two separated exothermic peaks well defined at 670 K and 746 K. These peaks represent the thermal signature of the energy exchanged during structural changes of the microstructure of the alloy such as disorder-order transition L1 0 -A1 and crystallization processes. Using the data obtained through DSC, we have performed isothermal annealing at various temperatures. Parts of CoPtAgB as-cast ribbons have been annealed at different temperatures (400 • C or 673 K, 473 • C or 746 K and 670 • C or 943 K for 1 h each). These values were chosen taking into account the peaks observed in DSC curves. The duration of the annealing has been set for 1 h as a compromise between the need of achieving proper complete phase transformation (as we have seen in other cases that shorter annealing treatments did not lead to complete disorder-order transformation) and the need to restrict as much as possible the grain growth (occurring for longer annealing treatments, e.g., several hours) in order to preserve a refined microstructure with magnetic regions/grains in the nanometer range.
The morphology of the ribbons as observed by Scanning Electron Microscopy (SEM) is presented in Figure 2. SEM images were obtained onto the wheel side of the ribbon surface using a field emission SEM (FE-SEM) with a resolution of 2 nm, under a 30 kV acceleration voltage. It has been documented [26] that there are significant morphological differences between the free side (solidified in Ar) and the wheel side (solidified in contact with the Cu wheel); however, these differences become important for Fe-based melt-spun ribbons in the case of particular effects such as the magneto-impedance observed in soft magnetic ribbons [27].
Both as-cast and annealed samples of Co 48 Pt 28 Ag 6 B 18 present a quasi-homogeneous surface with preferred direction of polymorphic growth as indicated by horizontal longitudinal terraces seen in the images. Please note that the surface morphology depends on the initial stoichiometry of the ribbons and on the annealing conditions. For the sample Co48Pt28Ag6B18 one can observe two separated exothermic peaks well defined at 670 K and 746 K. These peaks represent the thermal signature of the energy exchanged during structural changes of the microstructure of the alloy such as disorder-order transition L10-A1 and crystallization processes. Using the data obtained through DSC, we have performed isothermal annealing at various temperatures. Parts of CoPtAgB as-cast ribbons have been annealed at different temperatures (400 °C or 673 K, 473 °C or 746 K and 670 °C or 943 K for 1 h each). These values were chosen taking into account the peaks observed in DSC curves. The duration of the annealing has been set for 1 h as a compromise between the need of achieving proper complete phase transformation (as we have seen in other cases that shorter annealing treatments did not lead to complete disorder-order transformation) and the need to restrict as much as possible the grain growth (occurring for longer annealing treatments, e.g., several hours) in order to preserve a refined microstructure with magnetic regions/grains in the nanometer range.
The morphology of the ribbons as observed by Scanning Electron Microscopy (SEM) is presented in Figure 2. SEM images were obtained onto the wheel side of the ribbon surface using a field emission SEM (FE-SEM) with a resolution of 2 nm, under a 30 kV acceleration voltage. It has been documented [26] that there are significant morphological differences between the free side (solidified in Ar) and the wheel side (solidified in contact with the Cu wheel); however, these differences become important for Fe-based melt-spun ribbons in the case of particular effects such as the magnetoimpedance observed in soft magnetic ribbons [27].
Both as-cast and annealed samples of Co48Pt28Ag6B18 present a quasi-homogeneous surface with preferred direction of polymorphic growth as indicated by horizontal longitudinal terraces seen in the images. Please note that the surface morphology depends on the initial stoichiometry of the ribbons and on the annealing conditions. The X-ray diffraction analysis was performed using a Bruker D8 Advance diffractometer, from Bruker Nederland BV, Leidendorp, The Netherlands, with Cu Kα radiation (λ = 0.15402 nm) in a Θ-2Θ geometry. The angular interval investigated was between 20 and 100 degrees in 2Θ). All spectra taken have been fitted using a full-profile, Rietveld-type algorithm MAUD (Materials Analysis Using Diffraction) software [28][29][30] version 2.75, developed by Luca Lutterotti, University of Trento, Italy. Figure 3 shows the XRD patterns of as-cast and annealed Co48Pt28Ag6B18 samples. For the as-cast and the 400 °C annealed samples, the diffractograms show rather similar features, with broad peaks of low intensity, which belongs to the CoPt face-centered cubic (fcc) phase, space group: Fm3m. The situation changes for the samples annealed at higher temperatures. Starting with 473°C annealing, one can observe additional Bragg peaks, for instance the superlattice peaks (001) and (110) belonging to the L10 tetragonal CoPt phase, space group: P4/mmm, accompanied by a structural refinement and grain growth illustrated by the narrowing of the peak's linewidth and by the evident splitting of some The X-ray diffraction analysis was performed using a Bruker D8 Advance diffractometer, from Bruker Nederland BV, Leidendorp, The Netherlands, with Cu Kα radiation (λ = 0.15402 nm) in a Θ-2Θ geometry. The angular interval investigated was between 20 and 100 degrees in 2Θ). All spectra taken have been fitted using a full-profile, Rietveld-type algorithm MAUD (Materials Analysis Using Diffraction) software [28][29][30] version 2.75, developed by Luca Lutterotti, University of Trento, Italy. Figure 3 shows the XRD patterns of as-cast and annealed Co 48 Pt 28 Ag 6 B 18 samples. For the as-cast and the 400 • C annealed samples, the diffractograms show rather similar features, with broad peaks of low intensity, which belongs to the CoPt face-centered cubic (fcc) phase, space group: Fm3m. The situation changes for the samples annealed at higher temperatures. Starting with 473 • C annealing, one can observe additional Bragg peaks, for instance the superlattice peaks (001) and (110) belonging to the L1 0 tetragonal CoPt phase, space group: P4/mmm, accompanied by a structural refinement and grain growth illustrated by the narrowing of the peak's linewidth and by the evident splitting of some of the peaks belonging to the cubic phase, observed in the previous patterns. More precisely, the (200) peak of the cubic phase splits into the (200) and (002) peaks of the tetragonal phases and the (220) peak of the cubic phase splits into the (220) and (202) peaks of the tetragonal phase. The pattern of the sample annealed at 670 • C shows also the occurrence of the well-formed superlattice peaks (001) and (110) that represents the signature of the L1 0 tetragonal CoPt phase. It is also observed that after annealing at 670 • C, the peaks become narrower and the grain size increases.
The average grain size and lattice parameters have been calculated from the fitting of the patterns with the full-profile MAUD algorithm [28][29][30]. Each Bragg peak has been fitted to a pseudo-Voigt line profile. The Voigt profile, which is represented by a combination of Gaussian and Lorentzian functions whose relative abundance is quantified by the so-called mixing parameter, accounts well for the broadening of the diffraction line by two mechanisms. The first mechanism is related to strain and gives rise to a Gaussian profile, while the second one is related to the size of the grains and would produce a Lorentzian profile.
We have calculated the average grain size, which is here understood as being the average crystallographically coherent domain size, by utilizing the lattice parameters, linewidths and mixing parameters obtained for each Bragg reflection. The method is based on the integral breadth algorithm.
The integral breadth β is defined as the peak area divided by the maximum height of the peak. The integral-breadth method summarized in [31], calculates the root-mean-square (RMS) strain and both surface-and volume-weighted domain sizes according to the "double-Voigt" method [32,33], which is equivalent to the Warren-Averbach approach [34].
In melt-spun ribbons, accounting for different diffusion mechanisms as well as due to the heat loss gradient during rapid quenching, strain is expected to be induced during casting of the ribbons. Consequently, the lattice strain between the atomic planes should be significant. In this respect, RMS strain is defined as the amount (change in size or volume) by which a crystal lattice deforms under stress or force and is given as a ratio of the deformation to the initial size of the lattice. It is only a qualitative indication about the strain induced in the ribbons, though. Detailed, quantitative strain analysis using pole figures and stress tensor measurements as thoroughly done by Kurlyandskaya et al. in [35], though very useful for estimation of magnetoelastic properties, is however outside the scope of our paper.
The average volume-weighted grain size, lattice parameters and RMS strain, calculated from the fitting of the patterns with the full-profile MAUD algorithm, followed using integral breadth method, are presented in Table 1. It is seen that the average grain size increases with the annealing temperatures, ranging from around 20 nm for the as-cast state to 190 nm for the sample annealed at 670 • C, due to the structural refinement imposed by the further crystallization and the disorder-order transition with formation of L1 0 tetragonal CoPt from the precursor fcc CoPt phase occurring in the as-cast state. We have also observed an evident RMS strain evolution with the annealing temperature. In the as-cast state, the lattice RMS strain is large (around 0.78%), as expected due to the preparation conditions and different quenching rates on the two sides (wheel and free) of the ribbons. This strain however diminishes strongly (to around 0.33%) after annealing at 400 • C and this reduces even more for higher annealing temperatures (down to 0.16% after annealing at 670 • C). Moreover, it is possible from XRD measurements to determine the texture coefficient which can be calculated using the following formula [36,37]: Thkl being the texture coefficient of the corresponding Bragg reflection, Ihkl its integral intensity as determined by the full-profile fitting of XRD patterns, I 0 hkl corresponds to the bulk L10 CoPt Bragg peak intensity (from the ICDD file) and n is the number of Bragg peaks observed.
The texture results for the L10 CoPt superlattice peaks (001), (110) and (200) are presented in Table 2. One may notice that for the sample annealed at lower temperature, the texture coefficient for all the 3 L10 superlattice peaks are less than 0.5, indicating that the L10 CoPt is less ordered. The samples annealed at 670 °C have well-formed and quite intense L10 superlattice peaks with texture coefficients close to 1. Highest texture coefficient is observed for the (100) Bragg line therefore it may be inferred that there is a slight preferential ordering of the L10 superlattice in the (100) direction. Other studies on soft magnetic alloy texture have been done by checking the transformation of deformation texture into recrystallization texture during cooling of the alloys [38]. Moreover, it is possible from XRD measurements to determine the texture coefficient which can be calculated using the following formula [36,37]: T hkl being the texture coefficient of the corresponding Bragg reflection, I hkl its integral intensity as determined by the full-profile fitting of XRD patterns, I 0 hkl corresponds to the bulk L1 0 CoPt Bragg peak intensity (from the ICDD file) and n is the number of Bragg peaks observed.
The texture results for the L1 0 CoPt superlattice peaks (001), (110) and (200) are presented in Table 2. One may notice that for the sample annealed at lower temperature, the texture coefficient for all the 3 L1 0 superlattice peaks are less than 0.5, indicating that the L1 0 CoPt is less ordered. The samples annealed at 670 • C have well-formed and quite intense L1 0 superlattice peaks with texture coefficients close to 1. Highest texture coefficient is observed for the (100) Bragg line therefore it may be inferred that there is a slight preferential ordering of the L1 0 superlattice in the (100) direction. Other studies on soft magnetic alloy texture have been done by checking the transformation of deformation texture into recrystallization texture during cooling of the alloys [38]. The findings obtained by XRD have also been confirmed by direct imaging of the microstructure of the as-cast and annealed at 670 • C samples, by means of high-resolution transmission electron microscopy (HREM) imaging and selected area electron diffraction (SAED) of the areas imaged in HREM. A representative image of the as-cast Co 48 Pt 28 Ag 6 B 18 sample presented in Figure 4 confirms the fact that the disordered fcc phase (Fm3m) is predominant in this sample. Large grain areas (between 10 nm and 25 nm lateral size) with planes separated by reticular distances that reveal the occurrence of (200) and (222) reflections are shown in Figure 1. In the associated SAED pattern (Figure 4, inset), the most intense (111) and the (200), (220) and (311) diffraction rings of fcc CoPt phase can be identified. From the observed diffraction rings a value of lattice constant of a = 3.72 ± 0.005 Å was calculated, in good agreement with the value a = 3.729 Å determined from XRD for this sample. In some diffraction patterns, an additional diffraction ring, observable at d = 2.12 Å, could be attributed to the (111) plane of ordered rhombohedral phase L1 1 , reported by previous authors [1,21] as being a metastable phase that precedes the formation of the tetragonal L1 0 phase during the disorder-order phase transformation, and having a = 5.34 Å and α = 61.5 • as lattice parameters.
The HREM image of the Co 48 Pt 28 Ag 6 B 18 sample annealed at 670 • C is shown in Figure 5. In the image, areas of L1 0 phase (between 5-20 nm lateral size) are shown and the (101) and (110) lattice planes are clearly visualized, in good agreement with XRD patterns obtained for the sample annealed at 670 • C. SAED patterns have a polycrystalline nature as revealed by the most intense ring (101) which could be continuous (as in the inset of Figure 5) in the case of large ordered areas or consisting of many spots in the case of smaller ordered regions. In the first case one can also observe the (001) ring typical for the L1 0 superstructure whereas only a weak (110) ring is apparent in the second case. Figure 6 shows a SAED pattern of the same Co 48 Pt 28 Ag 6 B 18 sample annealed at 670 • C which confirms that there are large ordered regions of L1 0 phase, as confirmed in some cases in SAED patterns which show, besides the (002) spot, the (001) and the rarely observed and indexed (003) superlattice peak, characteristic to the high level of ordering of the L1 0 phase, whereas in XRD patterns only (001) and (002) superlattice peaks are observed. It is of interest to note that the higher the order of the superlattice peaks observed, i.e., higher l in the (00l) notation of the observed peaks indexation, the higher the degree of ordering of the L1 0 phase. The hysteresis loops of the samples Co 48 Pt 28 Ag 6 B 18 as-cast and annealed at 670 • C (Figure 7) have been recorded using SQUID magnetometry with the applied field parallel to the ribbons plane. The magnetic field was applied up to 50,000 Oe and measurements were taken at 300 K (or 27 • C). For the as-cast sample, it is observed that the hysteresis loop shows strong approach to the saturation at relatively low applied magnetic field (few hundreds of Oe). A large value of maximum magnetization (82 emu/g for a maximum applied field of 50,000 Oe) is then reached. The loop also exhibits very low coercivity. The exact value of the coercive field is estimated between 270 and 370 Oe. The lack of more experimental points in that region impedes furnishing of a more accurate value for the coercivity; however, the magnetic behavior is consistent with the presence of the disordered fcc A1 CoPt phase in the sample. Also, a detailed theoretical approach to saturation of the magnetization requires modeling, similar to the one presented in [39].
The magnetic behavior of the annealed sample is quite different. The approach to saturation of the magnetization is slower, eventually reaching 80 emu/g only at around 32,000 Oe applied field. More importantly, the hysteresis loop shows up to 10 times increase of the coercivity (about 4340 Oe). This dramatic increase of the coercivity corroborates the structural results, as such large values are typical for hard magnetic materials such as the ones exhibiting pure L1 0 phase (either CoPt or FePt). Indeed, such large values of coercivity at 300 K observed in the annealed sample confirm the results obtained in HREM and XRD where the highly ordered L1 0 CoPt phase is documented, inclusively by the presence of the rarely observed (003) superlattice ring in SAED patterns. as-cast sample, it is observed that the hysteresis loop shows strong approach to the saturation at relatively low applied magnetic field (few hundreds of Oe). A large value of maximum magnetization (82 emu/g for a maximum applied field of 50,000 Oe) is then reached. The loop also exhibits very low coercivity. The exact value of the coercive field is estimated between 270 and 370 Oe. The lack of more experimental points in that region impedes furnishing of a more accurate value for the coercivity; however, the magnetic behavior is consistent with the presence of the disordered fcc A1 CoPt phase in the sample. Also, a detailed theoretical approach to saturation of the magnetization requires modeling, similar to the one presented in [39]. The magnetic behavior of the annealed sample is quite different. The approach to saturation of the magnetization is slower, eventually reaching 80 emu/g only at around 32,000 Oe applied field. More importantly, the hysteresis loop shows up to 10 times increase of the coercivity (about 4340 Oe). This dramatic increase of the coercivity corroborates the structural results, as such large values are typical for hard magnetic materials such as the ones exhibiting pure L10 phase (either CoPt or FePt). Indeed, such large values of coercivity at 300 K observed in the annealed sample confirm the results obtained in HREM and XRD where the highly ordered L10 CoPt phase is documented, inclusively by the presence of the rarely observed (003) superlattice ring in SAED patterns. In fact, the material can be regarded as a composite that contains magnetic particles/regions of a nanometric scale embedded in a structurally and magnetically disordered residual matrix. The magnetic behavior is however consistent with results previously published in [40] and the magnetic behavior of L10 CoPt nanoparticles as described in [41]. In both papers, the evolution with annealing parameters (time, temperature) is studied and the results show that the magnetic behavior is typical for multiphase materials, with coercivity lightly reduced, as compared to bulk L10 CoPt, due to the interphase interactions between the hard magnetic L10 CoPt grains and the disordered magnetic intergrain regions (or embedding matrix).

Conclusions
In search for new nanocomposite magnets based on L10 phase, a detailed thermal, structural, and magnetic study of the Co48Pt28Ag6B18 alloy has been undertaken. Thermal analysis revealed the exothermic peaks attributable to the L10 ordering temperature and crystallization processes. The In fact, the material can be regarded as a composite that contains magnetic particles/regions of a nanometric scale embedded in a structurally and magnetically disordered residual matrix. The magnetic behavior is however consistent with results previously published in [40] and the magnetic behavior of L1 0 CoPt nanoparticles as described in [41]. In both papers, the evolution with annealing parameters (time, temperature) is studied and the results show that the magnetic behavior is typical for multiphase materials, with coercivity lightly reduced, as compared to bulk L1 0 CoPt, due to the interphase interactions between the hard magnetic L1 0 CoPt grains and the disordered magnetic intergrain regions (or embedding matrix).

Conclusions
In search for new nanocomposite magnets based on L1 0 phase, a detailed thermal, structural, and magnetic study of the Co 48 Pt 28 Ag 6 B 18 alloy has been undertaken. Thermal analysis revealed the exothermic peaks attributable to the L1 0 ordering temperature and crystallization processes. The evolution of the phase structure with temperature has been investigated. In that purpose, several annealings at 400 • C, 473 • C and 670 • C for 1 h each have been performed and the structure of the resulting samples have been investigated by XRD and HREM. It has been revealed that for lower annealing temperatures (400 • C) the microstructure consists of crystallized regions of soft magnetic fcc CoPt phase, but, starting with 473 • C annealings, the hard magnetic L1 0 phase becomes predominant. It is shown that at 670 • C the L1 0 phase achieves a high degree of ordering as proven by the occurrence of the rarely observed (003) superlattice peak. Furthermore, the hysteresis loop of the annealed sample shows up to 10 times increase of the coercivity (about 4340 Oe) compared to the as-cast one. These findings are promising in view of use of such materials as a future class of rare-earth free-permanent magnets.