Enhanced crystallization resistance and thermal stability via suppressing the metastable superlattice phase in Ni-(Pd)-P metallic glasses

1. Introduction

Metallic glasses (MGs) have attracted considerable attention due to their unique atomic structure and exceptional properties, such as high elastic limit [1], high yield strength [2], high hardness [3], large corrosion resistance [4], excellent soft magnetic properties [5] and good thermoplastic deformability [6]. These unique properties make MGs promising for potential applications; however, the limited glass-forming ability (GFA) and thermal stability hinder their widespread industrial applications [7]. Over the years, considerable efforts have been devoted to understanding the glass formation and improving GFA. It is generally accepted that glass formation is a consequence of the competition between the amorphous phase and crystalline counterparts of glass-forming liquids. The enhanced GFA is considered to be attributed to the influence of kinetic and thermodynamic factors [8]. An alloy with better GFA is expected to have a lower thermodynamic driving force for crystallization as well as slower crystallization kinetics. In practice, alloying is found to be an effective approach to enhancing GFA [3,4,[9], [10], [11]]. As an example, addition of a small amount of Al in the Cu-Zr binary alloy make its GFA increase by several times [12]. Relevant simulations and experiments indicate that alloying effects are often associated with the increase of icosahedral orders, which leads to the slow dynamics of supercooled liquids and enhancement of GFA [13,14]. However, the underlying mechanism of improving GFA is far from being fully understood.

In addition, solid MGs tend to crystallize upon reheating due to their thermodynamically metastable feature. The occurrence of crystallization destroys the unique properties of amorphous alloys, such as the embrittlement induced by crystallization [15]. From an application point of view, therefore, it is also crucial to improve the thermal stability of MGs. Indeed, developing MGs concurrently with decent GFA and high thermal stability has always been a focus of research activities since the first MG was discovered [16].

As a classical glass-forming alloy system, the binary Ni-P MG has great potential in engineering applications due to its excellent corrosion resistance and high hardness [17,18]. However, the marginal GFA of Ni-P MGs which are often synthesized as the micrometer-scale amorphous coating by electrochemical deposition seriously restricts its usage. To improve the GFA of Ni-P alloys, alloying effects of Pd has been investigated, which leads to the first reported bulk metallic glass (BMG), Pd₄₀Ni₄₀P₂₀ [19]. Various theoretical studies have been conducted to understand the reasons for the improved formability by the introduction of Pd from the electronic and atomic structures perspectives [20,21]. Guan et al. suggested that the hybrid packing scheme is the structural origin of the excellent GFA of Pd₄₀Ni₄₀P₂₀ [20]. Takeuchi et al. proposed that both the presence of the covalent bonds and their random networks contribute to the stabilization of amorphous phase [21]. Although these theoretical attempts provide helpful clues to reveal the effect of Pd in the Ni-P alloy, underlying physical mechanisms responsible for the striking GFA enhancement are still missing.

In this paper, we studied the Pd alloying effects on GFA and thermal stability of the Ni-P MG by systematically comparing the solidification and crystallization process of Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs. We reveal that the successful suppression of the competing metastable superlattice-like phase is the key to the enhanced thermal stability and GFA resulted from alloying addition of Pd.

2. Materials and methods

2.1. Materials

Ni₈₀P₂₀ master alloys were prepared by induction melting of pure nickel and phosphorus, while Pd₄₀Ni₄₀P₂₀ master alloys were fabricated by arc melting the mixture of pure Pd and the Ni₂P pre-alloy in an argon atmosphere. The alloy ingots were fluxed with B₂O₃ at 1273 K for 4 h. Ribbons of the Ni₈₀P₂₀ alloy with a cross-section of 0.02 × 2 mm² were produced via a single copper roller melt-spinning at roller tangential velocities of 20, 40 and 50 m/s in an argon atmosphere. Cylindrical rods of the Pd₄₀Ni₄₀P₂₀ alloy with a diameter of 2 mm were fabricated by water-cooled copper mold casting.

2.2. Methods

2.2.1. Microstructure observation

Structure characterization of the as-cast and annealed samples was conducted by X-ray diffraction (XRD) with Cu Kα radiation (MXP21VAHF) and transmission electron microscopy (TEM, Tecnai F30) techniques. The TEM samples were prepared using a focused ion beam (ZEISS AURIGA).

2.2.2. Thermal analyses

Differential scanning calorimetry (DSC) analysis was conducted using differential scanning calorimeter (Netzsch DSC214Polyma and DSC404F1). The instrument was calibrated with standard materials, i.e., indium and zinc. The annealed samples were prepared by cooling at 150 K/min from the target temperature to 293 K. Temperature modulated DSC (TMDSC) measurements were carried out in a TA DSC 25 under continuous heating rates of 2 K/min up to 720 K with an oscillation amplitude of 1 K and an oscillation period of 60 s.

3. Results

3.1. Glass formation of the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ metallic glasses

Fig. 1 shows XRD spectra of the as-cast Pd₄₀Ni₄₀P₂₀ rod with a diameter of 2 mm and Ni₈₀P₂₀ ribbons with a thickness of about 20 μm under different quenching rates (i.e., the roller tangential velocity of 20, 40 and 50 m/s). No sharp diffraction peaks corresponding to any crystalline phases are observed for the Pd₄₀Ni₄₀P₂₀ rod and the Ni₈₀P₂₀ ribbon obtained by the fastest cooling (i.e., 50 m/s), indicating amorphous nature of these specimens. The current experimental results show that the GFA is increased by at least two orders of magnitude with the addition of Pd. The XRD pattern of Pd₄₀Ni₄₀P₂₀ shows a board hump centered at 2θ = 41.6°, whilst that of the Ni₈₀P₂₀ ribbon also displays a board hump at 2θ = 44.8°. The distinct hump position implies that alloying of Pd varies atomic arrangements inside the Ni-P MG. However, for the Ni-P ribbon quenched at the roller velocity of 40 m/s, one can see several small crystalline peaks at 2θ = 37.6°, 40.9°, 47.4° and 48.3°, respectively, imposed on the board hump, implying that the competing primary phases formed due to the insufficient cooling rate applied. For the Ni-P ribbon quenched at an even slower cooling rate (e.g., the roller velocity of 20 m/s), the amorphous halo was disappeared and only sharp diffraction peaks corresponding to the Ni₃P and fcc-Ni phase was present, indicating that the sample is fully crystalline. Note that the diffraction peaks corresponding to the primary phase in the samples quenched at 40 m/s disappeared in this fully crystalline one, implying that the primary phase is metastable and decomposed into stable Ni₃P and fcc-Ni during insufficient cooling. In other words, it is the metastable phase that thermodynamically competes with formation of the glass structure in the Ni₈₀P₂₀ alloy.

View original graphic| Download| PPT slide

Fig. 1.   (Color online) X-ray diffraction patterns obtained from the as-cast Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ samples.

The thermal stability of these two MGs was investigated through DSC measurements. Figs. 2(a), 2(c) and 2(d) show the DSC scans of crystallization, melting and solidification processes, respectively, for the Pd₄₀Ni₄₀P₂₀ and Ni₈₀P₂₀ MGs at a constant heating rate of 10 K/min. Because the glass transition behavior is not obvious for the Ni₈₀P₂₀ MG as shown in Fig. 2(a), TMDSC measurements were conducted to probe its T_g [22,23]. Fig. 2(b) shows the TMDSC result for the Ni₈₀P₂₀ MG with a heating rate of 2 K/min. The glass transition was determined to be around 595 K as marked in the reversing heat flow (RHF) curve. As shown in Fig. 2(d), there is only one sharp exothermic peak on the solidification curve of the Pd₄₀Ni₄₀P₂₀ alloy, but two peaks are seen for the Ni₈₀P₂₀ alloy, indicating that the former alloy is very close to the eutectic composition while the latter is apparently off-eutectic. Table 1 lists the thermodynamic parameters including glass transition temperature (T_g), onset crystallization temperature (T_x), onset melting temperature (T_m), liquidus temperature (T_l), onset solidification temperature (T_s), width of the supercooled liquid region (ΔT_x), the reduced glass transition temperature (T_rg= T_g/T_m) and the γ parameter [γ = T_x/(T_g+T_l)] [24,25]. Compared with the Ni₈₀P₂₀ MG, the Pd₄₀Ni₄₀P₂₀ MG exhibits larger ΔT_x,T_rg and γ, suggesting the better GFA of the Pd₄₀Ni₄₀P₂₀ MG.

View original graphic| Download| PPT slide

Fig. 2.   (Color online) DSC curves for (a) crystallization,(c) melting and (d) solidification processes of the Pd₄₀Ni₄₀P₂₀ and Ni₈₀P₂₀ MGs; (b) TMDSC curves for the Ni₈₀P₂₀ MG.

3.2. Crystallization behavior of the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ metallic glasses

As shown in Fig. 2(a), there are two exothermic peaks, referred as peak 1 and peak 2, in the DSC curves for both MGs. The onset (T_x) and peak (T_P1 and T_P2) crystallization temperatures of the crystallization event for the Pd₄₀Ni₄₀P₂₀ MG are higher than those for the binary Ni₈₀P₂₀ MG (Table 1), implying that the Pd₄₀Ni₄₀P₂₀ MG has the higher thermal stability. Detailed structural investigation on the crystallization products was conducted to figure out the structural evolution during crystallization of these two MGs. Fig. 3 shows XRD patterns of the samples heated to the peak and end temperatures for the two exothermic peaks. When the Ni₈₀P₂₀ MG was annealed above the onset crystallization temperature (T_x), a primary crystallization phase emerged [620 K in Fig. 3(a)]. At higher temperatures, the body-centered tetragonal Ni₃P appeared in addition to the primary phase [627 K and 650 K in Fig. 3(a)]. It is found that the peak positions of diffraction peaks for the primary crystalized phase are identical with those for the primary phase formed during the melt quenching [see Fig. 1, Fig. 3 (a)], indicating that the competing phase upon reheating of the Ni₈₀P₂₀ solid glass is the same with that upon cooling of the Ni₈₀P₂₀ liquid. When the sample was annealed to the second crystallization peak temperature, the relative intensities of the primary phase decreases significantly and the fcc-Ni phase starts to appear [695 K in Fig. 3(a)]. Meanwhile, the relative intensities for the Ni₃P phase increases. After being annealed over the second crystallization peak, the final crystalized product is consisted of Ni₃P and fcc-Ni phases [720 K in Fig. 3(a)] which agrees with the previous results [26], and the primary phase fully decomposed. These results also confirm that the primary phase is metastable and decomposes into the Ni₃P and fcc-Ni phases during the late crystallization stage.

View original graphic| Download| PPT slide

Fig. 3.   (Color online) X-ray diffraction patterns for the (a) Ni₈₀P₂₀ and (b) Pd₄₀Ni₄₀P₂₀ MGs during crystallization at different temperatures.

With the addition of Pd, however, the crystallization process became significantly different, as shown in Fig. 3(b). When the Pd₄₀Ni₄₀P₂₀ MG was heated to the first exothermic peak temperature [i.e., 670 and 684 K in Fig. 3(b)], the amorphous phase crystallized into multiple crystalline phases simultaneously. The crystalline phases can be indexed as five crystalline compounds, i.e., tetragonal Ni₁₂P₅-like structure, fcc-(Ni, Pd) solid solution, orthorhombic Pd₃P-like structure, body-centered tetragonal Ni₃P-like structure and monoclinic Pd₅P₂-like structure, which is in agreement with previous studies [27]. When the sample was heated to the second exothermic peak [707 and 730 K in Fig. 3(b)], no new crystalline phases emerged while the relative intensities of all the crystalline peaks have slightly varied, implying the difference in growth behavior among these crystallized phases.

Fig. 4(a) shows TEM bright-field image of the crystallized samples for the Ni₈₀P₂₀ MG heated to the first peak temperature (650 K). It is seen that the crystallized products consist of two phases, i.e., the larger plate-shaped phase with a size of ∼200 nm embedded with the dispersed nano-scaled particles. The high resolution TEM (HRTEM) image corresponding to the plate-shaped phase is illustrated in Fig. 4(b), which reveals that the larger plate-shaped phase has a periodic modulated structure with a periodicity of ∼1.3 nm. The inset in Fig. 4(b) which is the fast Fourier transform (FFT) pattern from the embedded nanoparticle [marked as region A in Fig. 4(b)] confirms that the dispersed nanoparticle is the Ni₃P phase. Therefore, it is clear that the superlattice phase corresponds to the primary phase. Fig. 4(c) shows TEM bright field image of the Ni₈₀P₂₀ MG heated to the second peak temperature (720 K). Similar to that observed in Fig. 4(a), there are also two phases with different sizes in the crystallized sample [Fig. 4(c)], i.e., dispersed nanoparticles embedded in the larger main phase. The selected area electron diffraction (SEAD) pattern of the major phase [inset in Fig. 4(c)] identifies that it is the Ni₃P phase. The dispersed nanocrystalline shown in Fig. 4(c) is determined to be fcc-Ni based on its HREM image [Fig. 4(d)] and the corresponding FFT pattern [inset in Fig. 4(d)]. These results indicate that the final crystallization products of the Ni₈₀P₂₀ MG include Ni₃P and fcc-Ni, which is consistent with the XRD results [Fig. 3(a)]. Combining the DSC measurements with XRD and TEM analyses, it is clear that the crystallization process of the Ni₈₀P₂₀ MG proceeds by two steps; first, a metastable periodic superlattice phase accompanied with a small amount of dispersed Ni₃P nanocrystals precipitated from the amorphous matrix during the first crystallization event; Subsequently, the metastable periodic superlattice phase decomposed into the stable Ni₃P and fcc-Ni phases during the second crystallization stage.

View original graphic| Download| PPT slide

Fig. 4.   (Color online) TEM and HRTEM observations of crystallized Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs. (a), (c) TEM bright-field images of Ni₈₀P₂₀ MG after Peak 1and Peak 2; (b), (d)HREM images of the corresponding regions in (a) and (c); (e), (f) TEM images of Pd₄₀Ni₄₀P₂₀ MG after Peak 1and Peak 2.

Compared with the binary Ni₈₀P₂₀ MG, the crystallization product of the ternary Pd₄₀Ni₄₀P₂₀ MG is more complicated. According to the XRD results [Fig. 3(b)], there are five different phases simultaneously formed from the amorphous phase. Fig. 4(e) shows the TEM bright field image of these five crystallized phases in the sample heated up to the first peak temperature (684 K). Distinct from that observed in the partially crystallized Ni₈₀P₂₀ sample where the primary superlattice phase is significantly larger than the subsequently precipitated Ni₃P nanocrystal, the crystallized phases in the Pd₄₀Ni₄₀P₂₀ MG have similar shapes and sizes, implying that the BMG crystallized via the so-called “eutectic crystallization” mode. In other words, the five crystallized phases are competing for each other and finally have comparable driving forces for crystallization. When the sample heated up to the second peak temperature (730 K), the crystallized phases grow from an average grain size of ∼100 to ∼200 nm [Fig. 4(f)], but no new phase emerges. Compared with the grain growth in the Ni₈₀P₂₀ MG during the crystallization process [from Fig. 4(a) to (c)], the growth rate of crystallized phases in the Pd₄₀Ni₄₀P₂₀ MG [from Fig. 4(e) to (f)] is much slower due to the complicated crystallization products.

3.3. Crystallization thermodynamics and kinetics of the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ metallic glasses

In order to evaluate the difference in thermal stability and GFA between these two MGs, we calculated the activation energy E for crystallization by measuring DSC curves at different heating rates. As shown in Fig. 5, the characteristic temperatures (T_x, T_P1 and T_P2) for crystallization are shifted to higher temperatures with increase of the heating rate, indicating the thermally activated kinetics in nature. Following the Kissinger model [28], the rate-dependence of the onset temperature (T_x) and two peak temperatures (T_P1 and T_P2) for both samples was used to calculate the activation energy E for crystallization:

$ln(\frac{B}{T_{2}})=-\frac{E}{RT}+C$ (1)

where B is the heating rate, R the gas constant, C a constant, and T the characteristic temperature. Fig. 6 shows the typical Kissinger plots corresponding to the crystallization processes for both samples. The activation energy for the different characteristic crystallization temperatures were derived by linearly fitting the Kissinger plots. As listed in Table 2, the activation energies for the T_x, T_P1 and T_P2 of the Pd₄₀Ni₄₀P₂₀ MG are all larger than those of the Ni₈₀P₂₀ MG. The crystallization reaction of MGs usually consists of the nucleation and growth process. In general, the activation energy for the onset temperature (E_x) and peak temperature (E_P1 and E_P2) corresponds to the nucleation and growth process of crystallization, respectively. Clearly, enhanced thermal stability is realized with the MG changing form the binary to ternary system with alloying of Pd.

View original graphic| Download| PPT slide

Fig. 5.   (Color online) DSC curves of the (a) Ni₈₀P₂₀ and (b) Pd₄₀Ni₄₀P₂₀ MGs under non-isothermal conditions at different heating rates.

View original graphic| Download| PPT slide

Fig. 6.   (Color online) Kissinger plots for the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs, (a) T_x, (b) T_P1 and (c) T_P2.

When following the rule of additivity [29], it was found that the Johnson-Mehl-Avrami (JMA) model could also be applied for isochronal heating experiments. Combining the Kissinger method [28] and the equations proposed by Gao et al. [30,31], the relationships are defined as follows:

$d[ln(\frac{dx}{dt})_{p}]/d(1/T_{p})=-E/R$ (2)

$K_{p}=(BE)/(RT^{2}_{p})=K_{0}exp(-E/RT_{p})$ (3)

$(dx/dt)_{p}=0.37nK_{p}$ (4)

where n is the Avrami exponent, K₀ the frequency factor, T_p the peak temperature and (dx/dt)_p the maximum crystallization rate.

Fig. 7 shows the volume fraction x as a function of temperature estimated at different heating rates for these two MGs. The sigmoid type curves are observed, which is common in amorphous materials under a non-isothermal crystallization process. The crystallization rate at the beginning and end of the crystallization is slow. Since the heating rate B is constant, the temperature and time are actually linearly related. Then the crystallization rate as a function of temperature can be calculated from the volume fraction curves as shown in Fig. 8. For each heating rate, the (dx/dt)_p means the maximum value at the peak temperatures, which is used for calculating the Avrami exponent n. It is worth noting that obvious differences in shape was seen for the second DSC peak of these two MGs, implying different types of crystallization reactions in the late stage. Based on Eqs. (2) - (4), the values of Avrami exponent n were calculated, which can be used to describe the nucleation and crystal growth mechanism during crystallization.

View original graphic| Download| PPT slide

Fig. 7.   (Color online) Crystallized volume fraction for the (a, c) Peak 1 and (b, d) Peak 2 of the Ni₈₀P₂₀ MG and Pd₄₀Ni₄₀P₂₀ MG, respectively, as function of temperature at different heating rates.

View original graphic| Download| PPT slide

Fig. 8.   (Color online) Variations of the crystallization rate for the (a, c) Peak 1 and (b, d) Peak 2 of the Ni₈₀P₂₀ MG and Pd₄₀Ni₄₀P₂₀ MG, respectively, as function of temperature at different heating rates.

Table 3 shows the n value and other corresponding parameters including K_p and K₀ in the two crystallization peaks for Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs. For the Ni₈₀P₂₀ MG, the n values for the two crystallization peaks are largely varied; however, for the Pd₄₀Ni₄₀P₂₀ MG, the n values for the two crystallization peaks are close, i.e. 2.4 ± 0.3 for peak 1 and 2.2 ± 0.1 for peak 2. According to the classical theory of phase transition [32], the n value above 2.5 means that the crystallization takes place through a three-dimensional interface-controlled growth, whilst that below 2.5 suggests a three-dimensional interface-controlled growth with a constant nucleation rate. Note that the n value for peak 1 of the Ni₈₀P₂₀ MG is above 2.5, indicating that the first stage of crystallization in this MG takes place through a three-dimensional interface-controlled growth with an increasing nucleation rate, i.e., the Ni₈₀P₂₀ MG has experienced a more copious nucleation process than the Pd₄₀Ni₄₀P₂₀ MG. The n value of peak 2 for both the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs is between 1.5 and 2.5, indicating that the later crystallization event is dominated by a three-dimensional interface-controlled growth mode with a constant nucleation rate.

4. Discussion

As shown in Fig. 1, the critical size for forming a fully amorphous structure is significantly increased by 100 times at least, from glassy ribbons to bulk samples. The atomic radius of Pd, Ni, and P is 0.138, 0.125, and 0.110 nm, respectively. Obviously, the alloying of Pd widens the atomic size distribution and increase its mismatch, which is beneficial to enhance the efficiency of atomic packing. It has been suggested that the increased packing density leads to the increase of viscosity and thus decreases the atomic mobility [[33], [34], [35]]. The higher viscosity is beneficial to suppress atomic rearrangement and diffusion, and thus can stabilize the melt by reducing the possibility of nucleation and growth of crystalline phases. In addition, the heat of mixing among these elements is 0 kJ mol^-1 for Pd-Ni, -36.5 kJ mol^-1 for Pd-P, and -34.5 kJ mol^-1 for Ni-P, indicating that introduction of Pd can increase the negative heat of mixing among the constituents. Moreover, note that the enthalpy of mixing for the Pd-P and Ni-P is close to each other, implying that the atomic diffusion and local rearrangement are difficult to take place due to the complicated interactions among the atoms. As a result, the synergistic topological and chemical effects induced by the introduction of Pd can contribute to the enhanced GFA in the Pd-Ni-P system from a standpoint of thermodynamics.

It is well known that the glass formation is a competition process between the primary crystalline phase and amorphous phase. Therefore, it is helpful to understand the enhanced GFA of Pd-Ni-P alloys in terms of the crystallization process of MGs. As identified by the XRD [Fig. 1, Fig. 3(a)] and TEM [Fig. 4(a)] results, the competing crystalline phase in the Ni₈₀P₂₀ MG is the superlattice phase, which is a metastable and finally decomposes into the stable Ni₃P and fcc-Ni phases under higher temperatures [Fig. 1, Fig. 3(a)]. As shown earlier, the activation energy for the nucleation (E_x) and growth (E_P1) of this competing metastable phase is much smaller than that in the Pd₄₀Ni₄₀P₂₀ glass. In particular, the E_P1 in the Ni₈₀P₂₀ MG is 235.9 kJ/mol, only two thirds of that (348.2 kJ/mol) for all the crystallized phases in the Pd₄₀Ni₄₀P₂₀ MG, indicating that formation of the metastable superlattice phase is much easier. Moreover, the calculated n value for the first peak of the Ni₈₀P₂₀ MG reaches 3.1 ± 0.2, also indicating that the superlattice phase has strong potency to nucleate upon crystallization, as evidenced by the larger grain size of the crystallized superlattice phase in Fig. 4(a). Due to the strong formation tendency the metastable phase, therefore, glass formation in this binary alloy needs extremely high cooling rates to avoid occurrence of crystallization.

However, there are five stable crystalline phases precipitated simultaneously in the ternary Pd₄₀Ni₄₀P₂₀ MG when the crystallization starts [Fig. 3(b)]. Obviously, the introduction of Pd has a significant effect on the phase competition upon cooling. For the Ni₈₀P₂₀ MG, the ferromagnetic attraction feature of Ni leads to the aggregation behavior of Ni atoms, which promotes the formation of superlattice structures [36,37]. Nevertheless, previous studies have demonstrated that Ni and P tend to diffuse into the superlattice phase during the growth of the superlattice nanostructure while the Pd undergoes opposite direction of diffusion [38], implying that introduction of Pd can effectively suppress the formation of such metastable phase. Moreover, the complicated short-range orders in the ternary Ni-Pd-P system is beneficial for the homogeneous nucleation of different crystals [27], as confirmed shown in Fig. 2(d) which illustrates a typical melting behavior for eutectic alloys. As a consequence, five crystalline phases instead of the superlattice phase simultaneously precipitate at the beginning of crystallization of the Pd₄₀Ni₄₀P₂₀ MG. The five crystallized phases have different crystallographic structures and distinct chemical compositions, which manifests that long-range diffusion is needed to meet the requirements of composition and atomic structure for nucleation and growth of these crystalline phases. As such, the crystallization of the Pd₄₀Ni₄₀P₂₀ MG becomes much sluggish, which is evidenced by its larger crystallization activation energies (Table 2) and the slower growth rate of crystallized phases [i.e., the finer grain size observed in Fig. 4(e)]. Therefore, the introduction of Pd not only successfully suppresses the formation of superlattice phase but also alters the competing crystalline phases. As a consequence, the change in crystallization mode from primary crystallization in the binary system to the eutectic one in the ternary system endows the Ni-Pd-P system superior GFA.

5. Conclusions

To sum up, effects of the Pd introduction on glass formation, crystallization and thermal stability of the Ni₈₀P₂₀ and Pd₄₀Ni₄₀P₂₀ MGs were systematically investigated. The main results are summarized as follows:

1) The GFA of the Ni-Pd-P MG is dramatically improved from the micrometer to millimeter scale with the introduction of Pd. On one hand, enhanced packing efficiency and increased negative heat of mixing induced by Pd is beneficial to the stabilization of the supercooled liquid. On the other hand, the alloying of Pd not only suppresses the formation of superlattice phase in the binary system but also alters the crystallization mechanism from primary to eutectic crystallization mode, which makes the crystallization sluggish in the ternary system and thus facilitates the glass formation.

2) The introduction of Pd enhances the thermal stability of the Pd-Ni-P metallic glass significantly. Thermodynamically, increased crystallization temperature and larger crystallization activation energy clearly suggest that crystallization process in the ternary Pd₄₀Ni₄₀P₂₀ metallic glass is more difficult than that in the binary Ni₈₀P₂₀ MG. Kinetically, compared with the complicated crystallization products in the ternary Pd₄₀Ni₄₀P₂₀ MG, the metastable superlattice phase primarily crystallized from the binary Ni₈₀P₂₀ MG is much easier to nucleation and growth, which is evidenced by the lower activation energy for crystallization and larger Avrami exponent n value.

Acknowledgments

This research was supported by the National Natural Science Foundation of China (Nos. 51871016, 51671021, 51671018, 51531001 and 11790293), 111 Project (B07003), the Program for Changjiang Scholars and Innovative Research Team in University (IRT_14R05) and the Projects of SKL-AMM-USTB (2018Z-01, 2018Z-19). H.W. would like to thank the financial support from the Fundamental Research Fund for the Central Universities of China (No.FRF-TP-18-004C1).