Nano-imprinting potential of magnetic FeCo-based metallic glass

Figure 1 shows an XRD pattern of as-quenched Fe₄₀Co₃₅P₁₀C₁₀B₅ alloy ribbon. The presence of a single broad peak in the XRD suggests that the ribbons are structurally amorphous. The super-cooled liquid temperature range in which metallic glass deforms like a polymer was measured with DSC at different heating rates (figure 2). The DSC curves measured at high heating rates (>40 °C min⁻¹) are similar to a typical metallic glass alloy, i.e. a glass transition temperature (T_g) followed by a super-cooled liquid region, and then crystallization (T_x) of amorphous alloy (figure 2(a)). The single exothermic peak resulting from the crystallization of amorphous alloy was found to split into three overlapping peaks, when the heating rate during measurement was 20 °C min⁻¹ (figure 2(a)). Three overlapping peaks change to two with a further decrease in heating rate (≤10 °C min⁻¹). These results suggested that the crystallization of Fe₄₀Co₃₅P₁₀C₁₀B₅ amorphous ribbon is complex. At least three different types of phases can form on crystallization. Generally, the ability to detect an exothermic peak corresponding to a particular phase precipitated from an amorphous matrix depends on the activation energy, crystallization rate, and sensitivity of the DSC equipment. To our understanding, the activation energy for different types of phases in the present case is close to each other. The crystalline phase formed in the vicinity of the first peak (figure 2(a)) initiates precipitation of the second phase (i.e. second peak) quickly. That is why the first and second peaks in DSC tend to merge at a slower heating rate (figure 2(a)). We believe that the single peak at a high heating rate is due to the slower response of the DSC equipment. The crystallization fraction of different phases precipitated from the amorphous phase depends on the area of the exothermic peak. Thus, three overlapping peaks observed in the DSC curve for the glassy ribbon at a heating rate of 20 °C min⁻¹ were deconvoluted using Gaussian curve fitting (figure 2(b)). The areas of the three peaks estimated after fitting are ∼28, 24, and 44 J g⁻¹ for peaks 1, 2, and 3, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To understand the types of phases formed on heating, amorphous ribbons were heat treated at different temperatures (440–480 °C) for varying times (t_a). At first, the temperature was raised from room temperature to the required annealing temperature (T_a) at a heating rate of 10 °C min⁻¹. After attaining the required T_a, the ribbons were cooled naturally to room temperature, i.e. annealing time (t_a) is ∼0. Figure 3(a) shows XRD patterns of ribbons heat treated in such a way. The amorphous structure of the ribbons is stable up to rising temperature until 460 °C. Above it (T_a = 475 °C), minor precipitation of the Co_0.72Fe_0.28, and CoFe phases was observed. This is evident from the presence of two low intensity diffraction peaks superimposed on a broad amorphous peak (figure 3(a)). Raising temperature further to 480 °C, causes precipitation of Fe₃B_0.82P_0.18 phase. Similar results were obtained for ribbons annealed at a lower heating rate of 5 °C min⁻¹ (figure 3(b)). These results suggest that the types of phases (i.e. Co_0.72Fe_0.28, CoFe, and Fe₃B_0.82P_0.18) precipitated in glassy Fe₄₀Co₃₅P₁₀C₁₀B₅ alloy are independent of the heating rate used to reach the required annealing temperature (T_a).

Zoom In Zoom Out Reset image size

For nano-imprinting, it is necessary to understand the crystallization behaviour of ribbons with annealing time in a super-cooled liquid state, i.e. above T_g (>430 °C). As shown in figure 3, heating of ribbons up to 460 °C does not cause crystallization, but the situation may change when the ribbons are kept at these temperatures for longer times. Therefore, we annealed ribbons at 440 °C for 10–240 min, and the XRD patterns are shown in figure 4(a). The ribbons are amorphous up to an annealing time of 10 min. Minor precipitation of CoFe phases/grains in amorphous matrix take place within 30 min of annealing. These grains grow slowly, and their size reaches to ∼17 nm in 1 h of annealing. The phases containing boron and phosphorus, i.e. Fe₃(B_0.63P_0.37) start to appear within 2–3 h of annealing. By this time the grain growth of FeCo phases becomes sluggish, and the average grain size obtained from XRD is ∼50 nm. Further increase in annealing time (over 2 h) mainly results in the growth of boride and phosphide phases. This slow rate of crystallization and grain growth accelerated at higher temperatures (figure 4(b)). For example, at 480 °C, all the above mentioned phases precipitate within 5 min of annealing (figure 4(b)).

Zoom In Zoom Out Reset image size

The local microstructure of the annealed ribbons was examined by using a TEM. The low density of big sized grains (∼200 nm) embedded in an amorphous matrix were observed for the ribbons annealed at 440 °C for 1 h (figure 5(a)). Based on the TEM image contrast it appears that the grain is composed of more than one crystal. Nano-beam electron diffraction (NBD) patterns at different locations of the grain were recorded. Two different types of NBD patterns were detected (figures 5(b) and (c)). Indexing of the diffraction spots confirmed the presence of CoFe and Co_0.72Fe_0.28 crystals. The density and size of these grains increases with an increase in annealing time up to 2 h (figure 5(d)). The microstructure of 4 h annealed ribbons (figure 5(e)) looks similar to 2 h of annealing. However, there are differences in electron diffraction patterns (figure 5(e)). In addition to CoFe and Co_0.72Fe_0.28 phases, diffraction spots corresponding to Fe₃(B_0.63P_0.37) phase were detected (figure 5(e)). These results are consistent with the x-ray results shown in figure 4(a).

Zoom In Zoom Out Reset image size

The DSC measurements [25] can shed some light on the relative amount of amorphous phase left after annealing of ribbons at 440 °C for different times (figure 6). Three distinct peaks observed in the as-quenched ribbon tends to merge into one peak after annealing for 1 h. A careful examination of the DSC curve (figure 6) reveals that the peak is asymmetrical, and it may be due to two overlapping peaks. The total area of the peak (ΔH ∼ 69 J g⁻¹) is larger than the ΔH remaining after precipitation of the CoFe and Co_0.72Fe_0.28 phases (figure 2(b)). It suggests that the crystallization of these two phases is not complete yet. This is consistent with the TEM and x-ray measurements, which showed precipitation of boride and phosphide phases in about 2 h of annealing. That is why the DSC peak tends to be symmetrical after 2 h of annealing, and its area decreases to 55 J g⁻¹. The peak area decreases to ∼38 J g⁻¹ in 3 h, which is lower than the area (44 J g⁻¹) remaining after precipitation of the CoFe phases (figure 2(b)). This clearly suggests that the precipitation of the Fe₃(B_0.63P_0.37) phase started within 2–3 h of annealing. Again, this is consistent with the TEM and x-ray results shown in figures 4 and 5.

Zoom In Zoom Out Reset image size

Figure 7 shows the variation in coercivity (H_c) and saturation magnetic flux density (B_s) with annealing time at 440 °C. Corresponding XRD curves are in figure 4(a). The as-quenched glassy ribbons are typically soft with B_s ∼ 1.24 T. A slight decrease in B_s was noticed after annealing for 10–30 min. The exact cause of the reduction in B_s is unknown because the structure does not change after annealing (i.e. remained amorphous). A possible explanation could be dissolved medium range FeCo clusters on heating and/or relaxation of amorphous structure. After 30 min of annealing B_s increases rapidly, and reaches to a maximum of ∼1.35 T in 2 h. After that it starts to decrease. This trend of B_s with annealing time seems to be consistent with the formation crystalline phases (figure 4(a)). Within 2 h of annealing, a high B_s FeCo phase (B_s ∼ 2.4 T) precipitates and grows in size, which leads to the enrichment of B and P in the remaining amorphous matrix. The depletion of magnetic elements from the amorphous matrix can affect magnetic properties strongly. A gradual decrease in B_s after 2 h of annealing (figure 7(a)) could be the result of decreasing B_s of the remaining amorphous matrix.

Zoom In Zoom Out Reset image size

The changes in H_c with annealing time is quite different from the B_s. In the amorphous state H_c is below the VSM measurement range. Once the crystalline phases start to precipitate, H_c increases. The relatively high magnetocrystalline anisotropy of the crystalline phase as compared to that of the amorphous is responsible for the increase in H_c. A higher volume fraction and bigger grains with increasing annealing time make H_c larger. The remaining amorphous matrix plays an important role in mediating magnetic exchange coupling between the precipitated crystalline phases. Therefore, the magnetic properties of the remaining amorphous matrix are important for the overall H_c of the ribbons. The amorphous matrix with higher B_s can mediate the stronger exchange coupling which leads to a strong decrease in H_c. The B_s of the remaining amorphous matrix decreases with the increase in annealing time due to the depletion of magnetic elements. This causes weakening of intergranular exchange. That is why H_c keeps on increasing even after annealing for more than 2 h. Eventually, it will saturate when there is no further changes in the remaining amorphous matrix. After 4 h of annealing H_c is ∼7 kA m⁻¹. Atomic diffusion is much stronger at higher temperatures, thus a similar value of H_c was obtained within 5–10 min of annealing at ∼480 °C.

Tetragonal FeCo alloys are predicted to exhibit hard magnetic characteristics [23, 26]. Theoretical simulations also suggested that the boron in FeCo alloy is effective for enhancing tetragonal distortion, and the resulting alloy may have high magnetocrystalline anisotropy of 8 × 10⁵ J m⁻³ [21]. Experimentally, there are some reports on the development of body centered tetragonal FeCo alloy [20, 27, 28]. However, tetragonal distortions relax with increasing thickness, and strain induced anisotropy decreases significantly. In the present study, alloy composition includes boron element, and crystallized ribbons also show the presence of an FeCo phase. Therefore, it is interesting to find out the presence/absence of a hard magnetic FeCo phase. At a first glance, the low inplane H_c shown in figure 7(a) suggests that the ribbons are soft magnetic. Generally, it is difficult to detect a low volume fraction of a hard magnetic phase in a soft magnetic alloy. The H_c of a soft magnetic alloy is low due to the easy propagation of the domain wall [29]. It increases in the presence of defects (such as pinholes, roughness, grain boundaries, non-magnetic impurities, etc.), which act as as pinning centers for domain wall motion. Generally, non-magnetic pinning centers or grain boundaries in a soft magnetic alloy do not increase H_c to more than a few hundred Oe. The presence of high anisotropy grains can significantly alter the domain wall propagation. Due to the magnetic nature, hard magnetic grains embedded in a soft magnetic alloy are strongly exchange coupled. This results in a lowering of domain wall pinning strength, when the magnetic field is applied along the easy axis, i.e. inplane of the ribbon. However, along the hard axis, i.e. out of plane, magnetization rotates with the field. In the absence of hard magnetic grains (pinning centers) magnetization varies linearly with magnetic field until saturation is attained [30]. The remanence magnetization (M_r) and H_c are zero. However, the existence of hard magnetic grains in a soft magnetic alloy can result in finite M_r and H_c [31]. Figure 7(b) shows the inplane and out of plane hysteresis curves measured at room temperature for the ribbon annealed at 440 °C for 240 min. As mentioned above the inplane hysteresis curve has low H_c, which is typical for a soft magnetic alloy. But in out of plane, there is a finite H_c and M_r, which are different when compared with the behavior of a soft magnetic alloy. The H_c depends on the saturation magnetization of the soft magnetic matrix, magnetocrystalline anisotropy of precipitated hard magnetic grains, and their volume fraction. Recoil remanence curves in the out of plane direction are useful to understand the behavior of the low density of hard magnetic grains in soft magnetic matrix [30]. Figure 7(c) shows the recoil remanence [M_d(H)] curve. The first derivative of it (shown in the inset) suggests that there are some grains whose magnetization is quite stable up to few kilo-orested of reverse magnetic field. The average magnetic field required to switch the magnetization of these grains is ∼2600 Oe.

Ribbons annealed at higher temperatures for shorter time (i.e. 480 °C for 5 min) exhibit a rapid decrease in demagnetizing remanence M_d(H) at lower fields. At higher fields it decreases slowly, similar to the ribbon annealed at 440 °C for 240 min. These results suggest that there exist two type of grains, soft and hard magnetic [30, 31]. In view of the structural characterization, the hard magnetic grains could be tetragonal FeCo. However, exact confirmation requires further investigation.

Beside structural and magnetic properties, viscosity is an important parameter that determines the nucleation and growth of crystalline phases, when glassy alloy is heated close to the super-cooled liquid temperature range. It is also the most important parameter for fabrication of three dimensional nano/micron scaled patterns by imprinting. Figure 8(a) shows the temperature dependence of dynamic viscosity (η_a) and steady-state viscosity (η_s) of Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glasses. The dynamic viscosity (solid lines) depends strongly on the heating rate. Its minimum decreases from 1.2 × 10⁹ to 1.7 × 10⁸ as the heating rate changes from 5–40 °C min⁻¹. Simultaneously, it deviates far from the steady-state viscosity (figure 8(a), dotted curve). When the temperature is higher than T_g, the equilibrium viscosity (η_eq) of the alloy can be determined directly by η_a under an isochronal heating scan. Otherwise, η_s can be used for η_eq under an isothermal creep experiment. The temperature dependence of η_eq is usually fitted by the Vogel–Fulcher–Tammann (VFT) formula, $\eta ={\eta }_{0}\exp \left[{D}^{* }{T}_{0}/(T-{T}_{0})\right],$ where η₀, D*, and T₀ are constants [24]. And their fitting result is shown as the dashed line in figure 8(a), $\eta =1.02\times {10}^{-6}\exp \left[{31.9}^{* }386.9/(T-386.9)\right].$ In addition, the parameter of 'fragility (m)' is also introduced to estimate whether the metallic glass is strong or fragile, where $m=\partial (\mathrm{log}\,\eta )/{\left.\partial \left({T}_{g}^{* }/T\right)\right|}_{{T}_{g}^{* }=T}.$ In this formula, T_g* is the temperature at which equilibrium viscosity is $1.0\times {10}^{12}\,{\rm{Pa}}\cdot {\rm{s}}.$ Based on the fitting result, the fragility is 40.4 for the Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass. In figure 8(b), we compared a fragility-normalized Angell's plot of Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass with non-magnetic bulk metallic glasses (BMGs), which showed good viscoelastic workability. The behavior of the present alloy is similar to other BMGs. Table 1 summarizes the various parameters for example deformability $\left[{d}^{* }(\alpha )=\,\mathrm{log}\left[\eta \left({T}_{g}^{* }\right)/\eta \left({T}_{x}\left(\alpha \right)\right)\right]\right],$ where α is the heating rate and η(T_x(α)) = η_eq,min], thermal stability [based on ${\rm{\Delta }}{T}_{rx}^{* }\,=({T}_{x}-{T}_{g}^{* })/{T}_{g}^{* }$], etc., and their comparison with other BMGs. These parameters are useful in deciding the usefulness of a material for viscoelastic workability [24]. The workability and thermal stability of magnetic Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass is comparable to Au_7.8Ge_13.8Si_8.4 metallic glass, and not far from well-known Pt-, Pd-, and Zr- based BMGs [24].

Zoom In Zoom Out Reset image size

To further understand the effect of crystallization on viscosity in the super-cooled liquid state, isothermal tensile creep experiments were performed on Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass alloy at different temperatures (figure 9(a)). At ∼415 °C, i.e. much lower than T_g, the initial viscosity is high (∼10¹⁰ to more than 10¹² Pa · s), and the alloy is difficult to deform. The initial viscosity decreases below ∼5 × 10⁸ Pa · s at 440 °C. However, after 1 min, it increases rapidly to ∼5 × 10⁹ Pa.s. Then, a slow increase continued for more than 1 h. Again a jump of nearly two orders of magnitude in viscosity was observed, and a steady state (>10¹² Pa.s) was reached after 2 h of annealing. The isothermal DSC measurement at 440 °C (figure 9(b), upper curve) exhibits a broad peak at the same time when the second jump in viscosity takes place (figure 9(b), bottom curve). It means, more than two orders of magnitude increase in viscosity is due to crystallization of the alloy. The evolution of crystalline phases with annealing time at 440 °C has already been shown in figure 4(a). The isothermal creep results are well consistent with the crystallization of alloy. Lowest viscosity in the beginning is due to glassy state. The first jump is caused by precipitation of the FeCo phases. The slow increase in viscosity corresponds to grain growth of FeCo. The second jump in viscosity is related to the crystallization of boride and phosphide phases. The initial viscosity decreases further at higher temperatures (figure 9(a)), but the time interval between the first and second jump in viscosity reduces significantly (e.g. figure 9(a) at 460 °C) because of accelerated diffusion resulting in the rapid grain growth of the FeCo phases. A typical viscous flow behavior was noticed at 480 °C. The alloy maintained a low initial viscosity of ∼10⁸ Pa.s for 1 min. Afterward rapid precipitation and growth of crystalline phases increases the viscosity to above 10¹² Pa.s.

Zoom In Zoom Out Reset image size

The creep experiments clearly suggested that the working temperature for imprinting is from 435 °C–480 °C. It is possible to obtain high aspect ratio nanostructures at higher working temperatures, but the nanostructures may have crystalline phases. To obtain amorphous nanostructures, the working temperature should be lower than 460 °C .

Based on the crystallization behavior and thermal stability of the Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass in the super-cooled liquid state, an imprinting temperature below 460 °C should allow enough working time to obtain amorphous patterns. Therefore, at first imprinting experiments were performed at 450 °C for different working times such as 180, 240, and 360 s by using a V-groove Si mold of width (W_d) ∼5 μm (figure 10(a)). The Si mold was fabricated by KOH etching (figure 10(a)). To avoid the diffusion/reaction of the Si mold with the Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass during imprinting, the mold surface was covered with a SiO₂ layer of thickness ∼0.85 μm. For all the experiments imprinting pressure was fixed to 50 MPa. Figures 10(b)–(d) show SEM images of the imprinted ribbons with working times of 180, 240, and 360 s, respectively. Figure 10(e) is an SEM image at a relatively lower magnification (imprinting time 360 s), which shows the excellent imprintability of Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass alloy on a large area. The deformation of alloy during imprinting strongly depends on imprinting time. We have measured the height (h) of patterns, and percentage of Si mold filling factor (R_f) for the imprinted ribbons (figure 10(f)). Both of them increase rapidly with increase in working time. A high R_f > 80% with a pattern height of ∼3.2 μm was obtained when the ribbons were imprinted at 450 °C for 360 s. It is interesting to note that the imprinted patterns are very smooth, and match with the surface characteristics of the Si mold. Usually, melt-spun ribbons are not so smooth, and contain some fine lines resulting from the roughness of the Cu wheel/flow of the alloy during fabrication. The smooth surface of the ribbons after imprinting suggests that the viscous flow of Fe₄₀Co₃₅P₁₀C₁₀B₅ metallic glass is strong enough to overcome the inherent surface roughness.

Zoom In Zoom Out Reset image size

The mold filling factor (R_f) was found to depend on Si mold size. The maximum value of R_f is ∼65% for the mold of width W_d = 10 μm. This is much smaller than the mold of W_d = 5 μm. Figure 11(a) shows the variations in %R_f with the W_d of the Si mold. It is interesting to note that %R_f starts to increase rapidly as the W_d decreases below 20 μm. One possible reason for this behavior is the effect of capillary forces on the viscous flow. A modified Hagen–Poiseuille equation was reported to determine this contribution in the metallic glass molding process [5].

$$\begin{eqnarray}&&P=\displaystyle \frac{32\eta }{t}{\left(\displaystyle \frac{l}{d}\right)}^{2}-\displaystyle \frac{4\gamma \,\cos \,\theta }{d}\end{eqnarray} \tag{ 1 }$$

where P is the required pressure to flow the viscous liquid into a channel, η is the viscosity of the viscous liquid, t is the filling time, l is the length of the formed pattern in the channel, d is the diameter of the channel, γ is the metallic glass-vacuum interfacial energy (∼1 N m⁻¹), and θ is the dynamic contact angle between the metallic glass and the mold. The two terms in the right-hand side of equation (1) represent viscous flow and capillary motion, respectively. The ratio of these two terms, i.e. β can be used to define the contribution of capillary effect. It is easy to see that the β is inversely proportional to the diameter, i.e. W_d. The capillary effect becomes comparable to the viscous effect when β ≈ 1. The present situation is a little bit different as the molds are V-grooved. The effective d of the mold decreases with the depth. Based on equation (1), the capillary effect (i.e. the second term) will increase as the alloy pushes down into the mold. According to he experimental results, the aspect ratio of the imprinted patterns is ∼0.2, and the contact angle lies between 20° and 30°. If we consider η = 10⁸ Pa · s, t =300 s, and γ = 1 N m⁻¹, the capillary effect becomes comparable to viscous (i.e. β ≈ 1) when W_d ∼ 10 μm. This seems reasonable as the measured %R_f starts to increase rapidly below W_d = 20 μm.

Zoom In Zoom Out Reset image size

After the imprinting experiments, we analyzed the structure of the ribbons by x-ray. Figure 11(b) shows the XRD patterns of the ribbons imprinted at 450 °C for the time of up to 360 s. A single broad peak similar to the as-quenched ribbon suggested that there are no major changes in the structure after imprinting, and the ribbons remained in the amorphous state. We have also performed the imprinting experiments at high temperatures. In general, height (or %R_f) of the patterns increases with increase in temperature for a fixed time of imprinting. However, imprinting at temperatures above 460 °C causes crystallization, which resulted in saturation in pattern height after some time. This is due to the rapid decrease in viscosity with the crystallization of the ribbon (as discussed in previous sections). The crystallization of ribbons after imprinting was confirmed by x-ray measurements. For example the ribbons imprinted at 460 °C for 480 s exhibited very low intensities of diffraction peaks matching with CoFe phases (similar to figure 4(b), 460 °C for 5 min).