Ball milling nanocrystallization of arc-melted and melt-spun Fe78Co5Nb3Zr3B5Ge5Cu1 alloy: microstructure and magnetic properties

Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 1 Ball milling nanocrystallization of arc-melted and melt-spun Fe78Co5Nb3Zr3B5Ge5Cu1 alloy: Microstructure and magnetic properties J. S. BLÁZQUEZ †, V. FRANCO †, A. CONDE †*, S. ROTH ‡ † Departamento de Física de la Materia Condensada. ICMSE-CSIC. Universidad de Sevilla. P.O. Box 1065. 41080 Sevilla, Spain. ‡ IFW-Dresden, Institute for Metallic Materials, Helmholtzstrasse 20, 01069 Dresden, Germany. The viability of developing a nanocrystalline microstructure for the Fe78Co5Nb3Zr3B5Ge5Cu1 composition using a planetary ball mill was studied. Two starting samples were chosen: an arc-melted ingot and a partially crystallized melt-spun ribbon. The -Fe phase was detected in both initial samples, although different intermetallic phases were detected in each of them. Milling at 150 rpm, these intermetallics disappeared after 5 h for the melt-spun alloy (RQ) and did not disappear even after 70 h for the arc-melted sample (AM). However, the -Fe phase became nanosized in both cases. A good thermal stability was found for the RQ alloy after 10 hours of milling, although the nanocrystal size increased from ~8 to ~20 nm after heating up to 973 K. The AM samples exhibited a lower coercivity and a higher magnetization than RQ samples. Keywords: Nanocrystalline Materials; Mechanical Alloying; Magnetic Properties. *Corresponding author: Prof. A. Conde Departamento de Física de la Materia Condensada. Universidad de Sevilla. Apartado 1065, 41080 Sevilla (Spain). Phone: (34) 95 455 28 85 Fax: (34) 95 461 20 97 E-mail: [email protected] Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 2 1. Introduction Soft magnetic nanocrystalline alloys with a microstructure consisting of nano-sized -Fe type crystallites embedded in a residual amorphous matrix are among the softest magnetic materials known up to date [1]. Generally, the microstructure of these alloys is achieved in a procedure consisting of two steps: first, an amorphous alloy is obtained with the desired composition and, afterwards, this precursor is conveniently annealed to develop a nanocrystalline microstructure. In order to obtain an amorphous alloy by quenching a melted system with a suitable composition to develop a nanocrystalline microstructure, it is necessary to use very high quenching rates (~105-106 K/s). These very high cooling rates limit the geometry of the material to films or ribbons with a maximum thickness of few tens of microns. In order to overcome this limitation, consolidation of metallic powders has been proposed as a suitable technique to produce bulk shaped materials [2], widening their possible technological applications. In this sense, mechanical milling has been shown as an advantageous procedure to obtain nanocrystalline powder in many different composition systems [3,4]. However, the coercivity of the consolidated powder is 2-3 orders of magnitude larger than the values observed in melt-spun nanocrystalline ribbons, which has been associated to the absence of a residual amorphous matrix in the case of the mechanically milled nanocrystalline samples [2]. Therefore, the distribution of the atoms out of the -Fe nanocrystals, forming intermetallics or segregated at the grain boundaries and its thermal stability are interesting subjects to study. In this work, the preparation of a nanocrystalline Fe78Co5Nb3Zr3B5Ge5Cu1 alloy by milling has been studied. The addition of Ge and Co to Fe-based nanocrystalline alloys has been proposed in order to enhance the magnetic properties of the system at Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 3 high temperatures [5-7]. There are a lot of parameters influencing the milling process for a given composition: type of mill, time, speed and media of milling, ball-to-powder ratio, starting material, etc. This study is devoted to the dependency of the milling process on the microstructure of the starting system for the same composition and, therefore, the other milling variables were kept constant. Two initial samples were chosen: an arc-melted ingot and a partially crystallized ribbon obtained by meltspinning. Both systems are compositionally homogeneous at the scale of the powder particles (~m). This fact implies several advantages in comparison with the use of elemental powders as starting materials, especially in such complex compositions as the one studied in this work. The milling of elemental powders would need much more time to homogenize the composition and it would be quite hard to assure that the final composition will remain as the initial proportion of powder used due to preferential sticking of some elements on the milling media. On the other hand, a homogeneous composition is less reactive than pure elements, which will reduce the effect of oxygen contamination in the powder. 2. Experimental A master alloy with the nominal composition Fe78Co5Nb3Zr3B5Ge5Cu1 was obtained by arc-melting several times in an argon atmosphere and after purifying the atmosphere by melting titanium (AM sample). A portion of this alloy was melted again and rapidly quenched by using a single-roller Bühler melt-spinner (RQ sample). The ball milling was performed under argon atmosphere in a Retsch PM4000 planetary ball mill using hardened steel balls and bowls. The rotation speed was 150 rpm for all the experiments and the ball to powder ratio was 10:1. Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 4 The RQ ribbon, ~100 m thick, was very brittle and thus, it was easy to make flakes which were used as the starting material for the milling procedure. After 2 h of milling, these flakes are still present and thus it is not possible to consider this RQ sample as a real powder at this stage. In order to obtain pieces of the AM sample small enough to start the milling in the planetary mill, big pieces (several mm) were pulverized using a SPEX shaker mill for short times (< 0.5 h). The XRD pattern of the obtained AM powder after milling in the SPEX did not show significant differences to the bulk AM alloy, except for a broadening of the diffraction maxima. The thermal stability of the samples was studied by annealing the powder in an argon atmosphere, using a differential scanning calorimeter (DSC) Netzsch DSC 404. The melting interval of the studied composition starts about Tm ~ 1450 K, as shown in Figure 1. The microstructure of the different samples was studied using a Philips PW 1050 diffractometer with Co K radiation. Room temperature magnetic properties of the samples were characterized using a vibrating sample magnetometer (VSM) for the saturation magnetization, MS, and a Förster Koerzimat for the coercivity, HC. The morphology of the powder was studied with a scanning electron microscope (SEM) Philips XL-30 using secondary electrons mode. ‘[Insert figure 1 around here]’ Some of the resultant powder was pressed at 973 K ~ (2/3)Tm during 5 min in vacuum, applying 575 MPa in an uniaxial hot press, in order to obtain bulk disk-shaped samples of 10 mm diameter and several mm high. The consolidation was also possible at 773 K ~ (1/2)Tm in the same conditions but unsuccessful at 573 K ~ (2/5)Tm. These bulk samples were prepared to check the reliability of the coercivity results obtained for powder samples using the Koerzimat. Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 5 3. Results and discussions 3.1 Microstructure Figures 2 and 3 show the XRD patterns of the RQ and AM samples after different times of milling. In the case of the AM sample labelled as-cast, the XRD pattern corresponds to the arc-melted ingot and, in the case of the RQ sample, it corresponds to the meltspun ribbon. As it can be observed for both as-cast samples XRD patterns present -Fe crystalline peaks but also other diffraction maxima of some crystalline phases can be detected. The XRD patterns of the RQ samples show peaks corresponding to metastable phase(s), which become negligible after 5 h of milling. However, the diffraction maxima of the AM samples can be indexed using the peaks of NbFeB and Fe2Zr phases. Unlike for RQ samples, these peaks do not disappear even after 70 h of milling. A tentative explanation of this fact must take into account the lower cooling rate of the melted composition in the case of the AM samples in comparison with that of the RQ samples. This provokes that the formation of stable crystalline phases different to -Fe is overcome in the case of the initial RQ ribbon (quenching rate of ~105-106 Ks-1), although some metastable crystalline phases appear. The metastability of these phases could be the responsible of the easiness for destroying them during milling, while the energy supplied by milling during our experiments was not enough to destroy the thermodynamically stable phases (-Fe, NbFeB or Fe2Zr in the initial AM sample). ‘[Insert figures 2 and 3 around here]’ The evolution of the -Fe phase with the milling time and the thermal treatment was studied for both sets of samples after fitting the XRD profile in the 2 range Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 6 between ~40-60 degrees. The main contribution corresponds to the (110) diffraction maximum of -Fe, fitted as a Pseudovoigt function. For a short milling time, tiny crystalline peaks of intermetallic phases appear in this 2 range (fitted as Lorentzian functions for simplification). In the case of RQ samples, an amorphous halo (simulated as a Gaussian peak) was also necessary to fit the XRD profile. In the case of the AM samples, the Lorentzian contribution to the Pseudovoigt function increases from 0.3 up to 0.6 with the milling time and, after heating up to 973 K, decreases to 0.3, independently of the milling time. This indicates that the relative effect of the small grain size on the broadening of the diffraction maxima increases with the milling time in comparison with the microstrain contribution and decreases with the annealing process. In the case of the RQ samples, the Lorentzian contribution is higher (decreasing from ~1 to 0.7 as the milling time increases, being ~1 after annealing). This indicates that the effect on the broadening of the (110) maximum for the RQ samples due to microstrain relative to small crystal size increases with the milling time, unlike for AM samples. From the broadening of the (110) deconvoluted maximum, it was possible to estimate the grain size, <D>, and the microstrain,  . On the other hand, the lattice parameter was measured using the (110) and (200) diffraction maxima of the -Fe phase. These results appear in table 1. ‘[Insert table 1 around here]’ As it is expected, the microstrain of the samples increases with the milling time for both RQ and AM systems from 0.006 up to 0.020 (after 15 h of milling) and up to 0.008 (after 70 h of milling), respectively. This indicates a more significant and much faster increase of the microstrain for RQ samples than for AM samples. After annealing, the microstrain of the RQ samples could not be measured due to the negligible Gaussian Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 7 contribution to the broadening of the (110) diffraction maximum, whereas AM samples do not show significant change. The decrease in grain size of the -Fe phase saturates in both types of samples at 8 nm, which is in good agreement with the value observed for pure Fe [8]. However, this value is obtained after 5 h of milling for the RQ alloy and only after 50 h of milling for the AM alloy. On the other hand, only the RQ samples milled during 10 and 15 h and heated up to 975 K remain with a value of D below 20 nm. For AM samples, the annealing process increases the grain size to 20 nm or higher, even for milling times of 70 h. This fact indicates that a certain microstructural stability can be achieved for the RQ sample after 10 h of milling but the microstructure of the AM samples is less thermally stable even after 70 h of milling. In fact, experiments performed on RQ sample heated up to 573 K show that the nanocrystal size increases from 8 to 15 nm for the alloy milled during 5 h, but the samples milled 10 and 15 h show the same crystal size after that thermal treatment. Figure 4 shows the XRD pattern of those RQ samples heated up to 973 K, for which -Fe was the only crystalline phase detected after milling. ‘[Insert figure 4 around here]’ The lattice parameter, a, does not differ for both as-cast samples (0.2874 and 0.2873 nm for the RQ and the AM alloys, respectively), but presents a different evolution during the milling and annealing processes. In the case of RQ samples, a is almost constant throughout the milling process and after annealing the sample. However, for AM samples, the lattice parameter increases for long milling times up to 0.2884 nm and a clear reduction of a down to ~0.2860 nm is detected after annealing, even for samples milled during short times. Figure 5 shows the a values as a function of Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 8 the milling time for both set of samples. The value of pure -Fe phase is also shown for comparison. ‘[Insert figure 5 around here]’ The constant value of a for RQ samples is clearly larger than that of pure -Fe (0.28663 nm) indicating the presence of some stable impurities in the crystal lattice. Although long milling time of AM samples yields an increase of the lattice parameter of the -Fe phase, it is worth noting that RQ samples were not submitted to such long milling times and, therefore, this increase could be ascribed to the higher energy accumulated for AM samples. The decrease of a observed after annealing AM samples occurs even for short milling times, comparable to those of RQ samples, for which no change in a is detected. This fact might be ascribed to the presence of stable intermetallic phases in the AM samples unlike for the RQ samples. The elements of the studied composition with better solubility in the -Fe phase are Co and Ge. Although Co is highly soluble in -Fe [9], the solubility of Ge is much more reduced (up to 10 at % at room temperature) [9]. The mixing enthalpies of Ge with Nb and Zr (-24 and -25 kJ/mol, respectively) are much lower than those of Fe with the two other elements (-3 and -1 kJ/mol, for Nb and Zr respectively) [10]. Therefore, it is possible that during annealing Ge atoms diffuse to the intermetallics, rich in Nb and Zr, and thus the -Fe phase will reduce the concentration of this element, in agreement with the reduction of a observed for annealed AM samples. The value of a observed for RQ samples and not annealed AM samples after low milling times, ~0.2873 nm, would correspond to a Ge content of 4 at % [11], as Co does not influence the lattice parameter of -Fe below 10 at % and for higher content a decreases [12]. The other elements Nb, Zr and B are strongly insoluble in the -Fe phase. For AM samples, these elements might be found preferentially in the Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 9 intermetallics; however, in the case of RQ samples, the boundaries between nanocrystals are suitable regions to be enriched in these elements. 3.2 Magnetic properties In order to perform magnetic measurements, the powder samples were encapsulated in silver foil, which was folded in order to avoid the movement of the powder inside. Saturation magnetization, MS, was measured in the VSM (the maximum applied field was 1500 kA/m) and the Koerzimat was used to obtain a value of coercivity, HC, after magnetizing the sample with an applied field of 200 kA/m. The resultant values appear in table 2. ‘[Insert table 2 around here]’ It is worth noting that these values of HC were obtained in such a system consisting of a mixture of powder and air gaps. However, considering the repetitiveness of the measurement (low errors bar) and that all the samples were prepared in the same way, the reported values could be considered as self-comparable under this study. On the other hand, the RQ powder milled during 15 h and the AM powder milled during 70 h were pressed at 973 K during 5 min in order to form bulk disks (1 cm in diameter and several mm high) and the coercivity measured in these two bulk samples is in good agreement with those of the corresponding powder samples encapsulated in silver and heated up to 973 K (see table 2). Once this consideration is established, it is possible to observe lower values of coercivity for the AM samples than for the RQ samples, about 2500-3000 and 4500-5000 A/m, respectively. In both cases, annealing of the samples reduces the value of HC. Figure 6 shows the HC values for the different studied samples. Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 16 Figure captions Figure 1. DSC scans at 20 K/min showing the melting and solidifying behaviour of the composition Fe78Co5Nb3Zr3B5Ge5Cu1. Figure 2. XRD patterns of powder obtained from a melt-spun ribbon after milling for different times. Figure 3. XRD patterns of powder obtained from an arc-melted alloy after milling for different times. Figure 4. Room temperature XRD patterns of RQ samples annealed at 973 K. Figure 5. Lattice parameter, a, as a function of the milling time for both RQ and AM samples in the as-milled and annealed state. Figure 6. Coercivity, HC, as a function of the milling time for both RQ and AM samples in the as-milled and annealed state. Figure 7. Secondary electrons SEM images of AM sample milled for 50 h (upper) and RQ sample milled for 10 h (lower). The bar corresponds to 100 m. Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 17 Figure 1 1400 1450 1500 heating (20 K/min) cooling (20 K/min) (5 W/g) dH/dt T (K) Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 18 Figure 2 30 40 50 60 70 80 90 RQ as-cast xxxx x 2 (degrees) RQ 2 h RQ 5 h Intensity (a.u.) RQ 10 h o  -Fe x metastable phase RQ 15 h o o Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 19 Figure 3 30 40 50 60 70 80 90 I I Ix xx xxx MA as-cast 2  (degrees) MA 10 h Intensity (a.u.) MA 30 h MA 50 h o  -Fe x NbFeB | Fe 2 Zr o o MA 70 h Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 20 Figure 4 30 40 50 60 70 80 90 heated up to 973 K RQ 5 h 2  (degrees) RQ 10 h Intensity (a.u.) RQ 15 h Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 21 Figure 5 0 10 20 30 40 50 60 70 2.85 2.86 2.87 2.88 2.89 pure bcc-Fe MA as-milled MA after 973K RQ as-milled RQ after 973 K a (10 -10 m) Milling time (h) Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 22 Figure 6 0 10 20 30 40 50 60 70 2000 2500 3000 3500 4000 4500 5000 5500 Coercivity (A/m) Milling time (h) MA as-milled MA after 973 K RQ as-milled RQ after 973 K Philosophical Magazine. Vol. 86. Núm. 15. 2006. Pag. 2271-2282 http://dx.doi.org/10.1080/14786430500300074 23 Figure 7