1 Introduction

Magnetic refrigeration is a technology utilizing the magnetic material as a refrigerant based on its intrinsic magnetocaloric effect (MCE), i.e., the alignment of magnetic moments as well as the magnetic entropy increased (or decreased) when the magnetic field is applied (or removed), and the material releases (or absorbs) heat during this process [1,2,3]. The magnetic refrigeration is a prospective replacement for the traditional gas compression refrigeration for its advantages of high efficiency, small occupied area and environmental friendliness [1,2,3,4,5,6]. According to the type of magnetic transition, magnetic refrigeration materials can be roughly divided into the first-order magnetic phase transition (FOMT) and the second-order magnetic phase transition (SOMT) materials [7]. The FOMT materials, e.g., LaFeSi, Mn(Fe,Co)PAs, Ni–Mn–Ga and GdSiGe alloys [5, 8,9,10], have a large MCE due to the simultaneous magnetic-structural transition, which is beneficial for the magnetic refrigerant. However, they usually have a narrow working temperature range. Besides, the accompanied large thermal and magnetic hysteresis as well as the weakened mechanical properties after frequent magnetic cycling impede their practical application [11]. Though the magnetic entropy change value (|ΔSM|) of SOMT materials is much lower than that of FOMT materials, their negligible thermal and magnetic hysteresis indicate better suitability for magnetic refrigeration [5, 12]. The Gd metal is a representative SOMT material and known as the prototype material [13], with the Curie temperature (TC) of 293 K and the maximum magnetic entropy change (|ΔSMpk|) of 5.5 J·kg−1·K−1 (ΔH = 2 T) [14]. Nevertheless, its high cost and poor chemical stability restrict the commercial utilization.

Among SOMT materials, amorphous alloys have the advantages of adjustable Curie temperature (TC), low magnetic hysteresis, low thermal hysteresis, low eddy current loss, high corrosion resistance, high strength, high thermal stability and easy processing and forming due to their unique long-range disordered and short-range ordered atomic structure, attracting more attention as a magnetic refrigerant candidate [6, 15,16,17,18]. Moreover, the amorphous alloy has wide magnetic transition temperature range and large relative cooling power, which is more suitable for the Ericsson thermodynamic cycle working at near room temperature [19]. To achieve ideal Ericsson cycle, the |ΔSM| has to be constant during the hot and cold reservoirs, which means that the table-like |ΔSM|–T curve needs to be configurated. Consequently, the composites consisting of two or more amorphous alloys with similar TC were proposed due to the usually asymmetrical λ-like shape of |ΔSM|–T curve of single-phase amorphous alloy. The different preparation processes, such as bonding, sintering and in situ precipitating methods, have been reported [6, 20,21,22]. For the bonding and sintering method, the heat transfer and the solid phase reaction between the different materials reduce the refrigeration efficiency [23], while it can be avoided through the in situ precipitation method. Mo et al. [24] reported the plateau of |ΔSM| values (~ 2.79 J·kg−1·K−1) in a wide temperature range from 137 to 180 K under the magnetic field change of 0–2 T in the Gd55Co35Mn10 amorphous-nanocrystalline composite manufactured by annealing the fully amorphous precursor at 600 K for 30 min. Wang et al. [25] reported the table-like |ΔSM|–T curve over a wide temperature range as well as a decreased peak value of magnetic entropy change in the Gd48Co52 amorphous-nanocrystalline composite obtained by controlling the cooling rate with copper wheel surface speed of 30 m·s−1.

In this study, in order to investigate the influence of microstructure on magnetocaloric properties of the amorphous-nanocrystalline composite produced by in situ precipitation method, the Gd60Co30Fe10 alloy ribbons with different cooling rates were produced by adjusting the melt-spinning speed. Their phase composition, microstructure and magnetocaloric properties were studied.

2 Experimental

The alloy ingots with nominal composition of Gd60Co30Fe10 were prepared by arc-melting the mixtures of pure Gd (99.9 wt%), Co (99.9 wt%) and Fe (99.8 wt%) under high purity argon atmosphere. The alloy ingots were remelted four times to make sure the homogeneity. As-spun ribbons with 2 mm in width and 20–30 μm in thickness were produced in a high-purified argon atmosphere by melt-spinning method using a single copper wheel with the surface linear rolling speeds of 6.0, 12.5, 25.0 and 50.0 m·s−1.

The phase structure of as-spun ribbons was identified by an X-ray diffractometer (XRD, Bruker D8 Advance) using Cu Kα radiation (λ = 0.15418 nm) at operation conditions of 40 keV and 120 mA in 2θ range of 20°–90°. The thermal properties of ribbons were characterized by a differential scanning calorimetry (DSC, NETZSCH STA 449F3) with a heating rate of 0.33 K·s−1 in the flow of argon. The microstructure analysis was performed using a high-resolution transmission electron microscope (HRTEM, JEOL JEM-2100) at an accelerating voltage of 200 kV, and the HRTEM specimens were prepared by the twin-jet electrochemical polishing with a solution of 4 vol% HClO4 and 96 vol% ethanol at about 250 K. The magnetic properties of as-spun ribbons were detected by the physical property measurement system (PPMS, Quantum Design PPMS-9T system). Concretely, the zero-field cooling (ZFC) temperature dependence of magnetization (M–T) curves was collected during heating process with the temperature range of 150–300 K under the applied magnetic field of 0.001, 0.005 and 0.020 T, while the isothermal initial magnetization (M-H) curves at different selected temperatures around the Curie temperature were measured under field change from 0 to 2 T.

3 Results and discussion

3.1 Characterization of microstructure

Figure 1a illustrates XRD patterns of Gd60Co30Fe10 alloy ribbons at different surface linear velocities of the copper roller. When the spinning speeds are 50.0 and 25.0 m·s−1, the typical amorphous broad diffuse scattering peak without any obvious diffraction peaks of crystalline phases can be obtained, which indicates the amorphous structure of the as-spun ribbons. However, sharp crystal phase diffraction peaks were detected in the samples with a rolling speed of 12.5 m·s−1, and the identified crystal phase of the peaks is Fe5Gd [PDF No. 00-025-1095]. The average size of the crystallite was calculated to be 23 nm using the Scherrer formula [26]:

$$ D = \frac{K\lambda }{{\beta \cos \theta }} $$
(1)

where D, K, λ, β and θ are the crystallite size, the Scherrer constant (generally 0.89), the wavelength of the radiation (0.15418 nm), the corrected peak width at half maximum of the intensity and the Bragg angle corresponding to the peak, respectively. In the ribbons spun with the surface linear rotating speed of 6.0 m·s−1, the amorphous phase was accompanied by the hcp-Gd and Fe5Gd type crystalline phases, and the grain size is roughly estimated to be ~ 25 nm. From the above phenomena, it can be observed that the amount of precipitated crystal phase and the crystallite size increases with cooling rate decreasing. In addition, (Co, Fe)5Gd crystals in the samples with spinning speeds of 12.5 and 6.0 m·s−1 present preferential orientation growth along the [101] direction during the rapid solidification. Furthermore, the relative strength of the diffraction peaks of hcp-Gd phase is higher than that of the Fe5Gd phase in the ribbons prepared at 6.0 m·s−1, indicating that the precipitation of hcp-Gd crystalline inhibited the preferential orientation growth of (Co, Fe)5Gd crystalline to some extent.

Fig. 1
figure 1

a XRD patterns and b DSC curves of Gd60Co30Fe10 alloy ribbons prepared at different spinning speeds

Full size image

As shown in Fig. 1b, obvious exothermic peaks can be found in DSC curves of Gd60Co30Fe10 alloy ribbons prepared at different cooling rates, which corresponds to the crystallization process of the amorphous phase in each sample, further demonstrating that the samples with spinning speeds of 12.5 and 6.0 m·s−1 were amorphous-nanocrystalline composites. The glass transition temperature (Tg) and the onset crystallization temperature (Tx) of the samples were determined by intersection of the baseline and extrapolation of the peak. The Tx values of the Gd60Co30Fe10 alloys solidified by cooling rates of 50.0, 25.0, 12.5 and 6.0 m·s−1 are 569, 581, 574 and 569 K, respectively. However, the Tg can only be obtained in the ribbons with spinning speeds of 12.5 and 6.0 m·s−1, and the values are 545 and 540 K. For the amorphous alloys solidified produced at cooling rate of 50.0 and 25.0 m·s−1, there is no evident glass transition in the curves, which is attributed to their lower thermal stability [27]. That is to say, in the Gd60Co30Fe10 alloy quenched at slower speeds of 12.5 and 6.0 m·s−1, the precipitation of the crystalline phases resulted in the composition change of the amorphous matrix with higher glass-forming ability. The Tx discrepancy in the amorphous alloys synthesized at 50.0 and 25.0 m·s−1 is ascribed to the component fluctuation, since the amounts of short-range ordered clusters are influenced by the cooling rate [28]. What should be pointed out is that the Tx of each ribbon is high enough to ensure the stability of the structure during the subsequent magnetic measurement.

From TEM bright-field image of Gd60Co30Fe10 with a cooling rate of 50.0 m·s−1 displayed in Fig. 2, it can be seen the uniform contrast without any spots and streaks, indicating that there is no obvious precipitation of crystal phase in the sample. The inset of Fig. 2b exhibits the fast Fourier transform (FFT) pattern of the selected area, and only a broad diffuse diffraction halo corresponding to the amorphous structure can be observed, which is consistent with the XRD result in Fig. 1 and demonstrates the fully amorphous microstructure in the alloy.

Fig. 2
figure 2

TEM bright-field image and Fourier transform diffraction pattern of Gd60Co30Fe10 with a cooling rate of 50.0 m·s−1

Full size image

For comparison, HRTEM bright-field image and FFT patterns of the selected areas in Gd60Co30Fe10 with cooling rate of 6.0 m·s−1 are presented in Fig. 3. It is evident that there are crystallites (corresponding to the streak region and FFT patterns with a set of spots) embedded in the amorphous matrix with FFT patterns of a halo in Fig. 3d. The interplanar spacing (d) of the different grains shown in Fig. 3b, c was calculated to be 0.315 and 0.298 nm, respectively, which is coherent with the (100) plane of hcp-Gd (0.3144 nm) and the (101) plane of Fe5Gd (0.2977 nm) structure. Typically, both the morphology and FFT pattern of (Fe,Co)5Gd grains revealed their preferential orientation growth. Besides, it can be seen from Fig. 3a that the crystal grain size is about 5–20 nm. All these results confirm the analysis of XRD pattern and the amorphous-nanocrystalline composite structure of the sample.

Fig. 3
figure 3

a HRTEM bright-field image and b-d fast Fourier transform patterns of different selected areas in Gd60Co30Fe10 alloy with a cooling rate of 6.0 m·s−1

Full size image

3.2 Magnetic and magnetocaloric properties

In order to investigate the influence of partially crystallization on the magnetic and magnetocaloric properties of the Gd60Co30Fe10 alloys, the ribbons prepared at spinning speeds of 50.0, 12.5 and 6.0 m·s−1 with different microstructures were chosen as the representative samples. Figure 4a exhibits the temperature dependence of magnetization (M-T) curves under different magnetic fields of each ribbon. With temperature raising, the magnetization decreases gradually and single distinct transition from ferrimagnetism to paramagnetism occurs in every curve. This magnetic transition correlates with the amorphous matrix phase in the samples, demonstrating that the amount of crystalline phase in the composite samples is slight and their magnetic contribution is neglectable. In this work, the Curie temperature (TC) of the material is determined according to the inflection point method, namely picking the temperature corresponding to the minimum of the dM/dT–T curves, which are plotted in Fig. 4b. Under the applied field of 0.001 T, the TC values of three samples fabricated at cooling rates of 50.0 m·s−1 (243 K), 12.5 m·s−1 (244 K) and 6.0 m·s−1 (245 K) are almost the same. For the ribbon with a cooling rate of 50.0 m·s−1, the TC values detected under different applied magnetic fields are also similar. However, the TC values of the ribbons with cooling rates of 12.5 and 6.0 m·s−1 decrease gradually when the field increases. In addition, the degree of variability is larger for the sample with slower cooling rate. This phenomenon can be ascribed to that the composition change of the amorphous matrix enhances the local composition fluctuation, which results from the precipitation of crystals [29].

Fig. 4
figure 4

a ZFC M–T curves and b dM/dT–T curves under different magnetic fields of Gd60Co30Fe10 as-spun alloys with different cooling rates

Full size image

The isothermal initial magnetization (M–H) curves of Gd60Co30Fe10 alloys with different spinning speeds under magnetic field from 0 to 2 T are displayed in Fig. 5. Each sample experienced the ferrimagnetic-paramagnetic transition during heating up, and the composite alloy reached magnetic saturation more difficultly than the fully amorphous counterpart because of the complicated microstructure although their magnetizations are similar. To determine the type of magnetic transition in a material, there are several methods supposed [30,31,32,33]. In this work, the nature of magnetic transition in the alloys was researched by the Arrott plot (M2 vs. H/M, converted from the M–H curve), as illustrated in Fig. 6. According to the Banerjee criterion [32, 33], the positive slopes of the curves correlate with the SOMT, and in contrary, the negative slope implies the FOMT. It can be seen that all the slopes of the Arrott plots in Fig. 6 are positive, verifying that the magnetic transition of the Gd60Co30Fe10 alloys is SOMT [34].

Fig. 5
figure 5

M–H curves of Gd60Co30Fe10 ribbons with different spinning speeds at selected temperatures near TC

Full size image
Fig. 6
figure 6

Arrott plots of Gd60Co30Fe10 alloy samples prepared at cooling rates of a 50.0 m·s−1, b 12.5 m·s−1 and c 6.0 m·s−1

Full size image

The magnetic entropy change (ΔSM) was utilized to characterize the MCE of Gd60Co30Fe10 alloys with different cooling rates, and it can be calculated by the data of M–H isotherms, according to Maxwell relation [33, 35],

$$\begin{aligned} \Delta S_{\text{M}} (T,H) & = S_{\text{M}} (T,H) - S_{\text{M}} (T,0) \\ & = \int_{0}^{H} {\left\{ {\frac{\partial M(T,H)}{{\partial T}}} \right\}_{H} {\text{d}}H} \\ \end{aligned}$$
(2)

where M, T and H are the magnetization, temperature and magnetic field, respectively. The correlation between |ΔSM| and temperature of the ribbons under the applied field of 0–2 T is depicted in Fig. 7, and all the curves are table-like in vicinity of TC (inset of Fig. 7), e.g., the |ΔSM| of the ribbon with 6.0 m·s−1 is almost constant in the temperature range of 200–230 K, which is more suitable for the Ericsson cycle in comparison to the typical λ-type curves of the materials with SOMT [18, 22]. As analyzed above, the magnetic properties are contributed by the amorphous matrix in the Gd60Co30Fe10 alloys with different spinning speeds, resulting in the similar shape of their |ΔSM|–T curve. The maximum values of magnetic entropy change (|ΔSMpk|) of the ribbons with cooling rates of 50.0, 12.5 and 6.0 m·s−1 are 2.08, 1.75 and 2.19 J·kg−1·K−1, respectively. The discrepancy of |ΔSMpk| could be ascribed to the different composition of the amorphous matrix as well as the localized composition fluctuation and heterogeneous microstructure (e.g., cluster and free volume) in the different samples [36, 37].

Fig. 7
figure 7

|ΔSM|–T curves of Gd60Co30Fe10 ribbons synthesized at different spinning speeds under magnetic field change of 0–2 T

Full size image

Table 1 lists the magnetocaloric properties of several representative amorphous and crystalline alloys whose TC is similar to the samples in this study [14, 16, 24, 38,39,40]. Usually, the broadened magnetocaloric effect relates to a higher relative cooling power (RCP, obtained by multiply the value of |ΔSMpk| by full width of half maximum (δTFWHM) in the |ΔSM|–T curve). Despite the |ΔSMpk| values of the Gd60Co30Fe10 alloys with different spinning speeds are smaller than that of the Gd60Co35Fe5 amorphous alloy, their RCP are comparable. Additionally, compared with the Fe89Co1Sc10 amorphous alloy, all the Gd60Co30Fe10 alloys exhibit higher |ΔSMpk| and RCP. Among the three Gd60Co30Fe10 samples, the one with cooling rate of 6.0 m·s−1 possesses the highest |ΔSMpk|. Furthermore, it shows higher RCP than Gd58Co42 amorphous alloy and the Gd55Co35Mn10 (600 K, 20 min) amorphous-nanocrystalline composite. Although the influence of quenching rate on the MCE of the Gd60Co30Fe10 ribbons changes irregularly, it has been demonstrated that the moderate decrease of spinning speed would not deteriorate the magnetocaloric performance of the samples seriously.

Table 1 Magnetocaloric properties of Gd60Co30Fe10 as-spun alloys with different cooling rates and some other representative materials (A and C indicating amorphous and crystalline structures, respectively)
Full size table

4 Conclusion

In summary, the Gd60Co30Fe10 alloy ribbons with different microstructures were fabricated by adjusting the cooling rate at various surface linear velocities of the copper wheel. Fully amorphous alloy can be obtained at spinning speeds of 25.0 and 50.0 m·s−1. However, the (Fe,Co)5Gd nanocrystals were precipitated and embedded in the amorphous matrix with the quenching rate of 12.5 m·s−1, and hcp-Gd nanocrystalline was separated out from the amorphous matrix if the velocity is furtherly reduced to 6.0 m·s−1. With applied magnetic field increasing, the magnetic transition temperatures of all the ribbons shifted to the lower value, and the degree of variability was larger for the sample with slower cooling rate. All the |ΔSM|–T curves of the alloys are table-like in vicinity of TC, which is more suitable for the Ericsson cycle. The |ΔSMpk| of Gd60Co30Fe10 with cooling rates of 6 m·s−1 under the magnetic field change of 0–2 T is 2.19 J·kg−1·K−1 at 217 K and the |ΔSM| is approximately constant in the temperature range of 200–230 K. The discrepancy among the magnetic performance of different ribbons could be attributed to the composition change of the amorphous matrix and local composition fluctuation in the composites due to the slight amount of crystalline phase. In addition, it was revealed that the moderate reduction of quenching speed will not deteriorate the magnetocaloric performance of the Gd60Co30Fe10 ribbons seriously.