3. Results and discussionTo provide an intuitionist comparison of the β-relaxations in the same and different systems, we fit the whole relaxation curves by coupling two Kohlrausch–Williams–Watts (KWW) equations in the Fourier transforms of
.[19] Here,
, where
and
represent instant elastic modulus and static elastic modulus, and f and τ are frequency and average relaxation time. β
is a nonexponential parameter related to the dynamic heterogeneity. The T dependent τ
for β-relaxation has the form of the Arrhenius process
, and
is described by the Vogel–Fulcher–Tammann (VFT) equation
. The pre-factors for the six metallic supercooled liquids are
s and
s, respectively, which are consistent with the previous works.[2]
Figure 1 shows the T dependent
of the three systems measured by DMA at 1 Hz, where T and
are scaled by the temperature of the α-relaxation peak
and the loss modulus
at
, respectively.
and
are both obtained from fittings. The circles represent the experimental data and the red and blue regions represent the scaled KWW fits of the β-relaxations for M–Cu and M–Ni (
, Ti, and Y), respectively. Considering the scaled fits of the α-relaxations nearly coincide for M–Cu/Ni, they are represented by a single grey region. The fitting of storage modulus data, which is not presented here, was carried on as an adjunct to the fitting position and intensity of the α-relaxation peaks. The β-relaxation activation energy
and
obtained from the fitting are listed in Table 1. As shown in Fig. 1(a), a β-relaxation hump can be observed at the low temperature flank of the α-relaxation peak for Zr
Ni
, while Zr
Cu
has only an excess wing, indicating that the substitution of Cu for Ni influences the β-relaxation behavior significantly. Similar results are observed in Ti-based MGs as shown in Fig. 1(b), however, the weak hump is further suppressed into an excess wing by the substitution of Cu for Ni. Figure 1(c) shows the effect of Cu/Ni substitution for Y-based MGs, which has the same tendency as that for Zr- and Ti-based MGs. Meanwhile, in sharp contrast to Zr- and Ti-based MGs, Y-based MGs exhibit the significant pronounced β-relaxation peaks at the temperature about
(the glass transition temperature) in the deep glassy state. It is so far the most obvious β-relaxation behavior found in MGs. The β-relaxation in Y-based MG is much stronger than that of La-based MG which was the strongest one in the previous works. Another characteristic is that the β- and α-relaxations of
Ni
MG are more separated compared with those of the other systems. It is also noteworthy that
for Y
Ni
(about 0.69) is even lower than that for La-based MGs (about 0.78) which is well known for the obvious β-relaxation.[7, 12]
All three systems have large negative
[20, 21] in a wide range as shown in Table 1. Here we define the reduced activation energy of β-relaxation
, and the smaller
indicates the separation between α- and β-peaks is more significant, which means the β-relaxation peak is more pronounced. The β-relaxation changes from an excess wing to a hump, finally to a peak with decreasing
.
was evaluated and listed in Table 1. For each system, the more negative
between two elements is, the less
will be, shown in Table 1 and Fig. 2. The
does not change significantly for Zr- and Ti-based systems, but the β-relaxation strength changes about 20% for both systems, which leads to the β-relaxation changing from excess wings to humps. It indicates that
is a sensitive parameter to the appearances of β-relaxations. This is further verified by Y–based MGs. With the smallest
for Y–Ni, the β-relaxation is farthest from the α-relaxation and the most pronounced.
Now the questions arise what the connections between
,
and local structures are and how these factors affect β-relaxation. As we known, mainly three factors affect local atomic structures in binary MGs, which are atomic radius, electronic configuration, and chemical interaction between atomic pairs. In order to study the chemical effect without interference of other factors, we carried out the substitution of Cu and Ni which have similar atomic radii and electronic configurations in each system. When considering different systems, the observation and analysis of the chemical effect are interfered by different principal elements. For instance,
of Y-based MG are only about 20–30 kJ/mol, which are not the most negative ones in the three systems, however, the β-relaxations are significantly pronounced, consistent with their small
. Such results further indicate that β-relaxations depend not only on the chemical effect between elements, but also strongly on the characteristics of major elements which might lead to different local structures. It is noticed that compared with transition elements, rare earth elements have some special properties so that the pronounced β-relaxations have only been observed in rare earth based MGs, such as La-based,[7] specially Y-based found in this work. Despite the complexity of the three factors, some regularity still could be figured out. Comparing M–Ni with M–Cu in each system, the MGs with larger negative
have smaller
. Thus, the chemical effect rule is valid for each system, but not applicable for the comparison between different systems. To compare different systems, besides the chemical effect, atomic radius and electronic configuration should also be considered.
The findings above demonstrate that the chemical effect on β-relaxations can only be compared in the same system. Meanwhile a question arises around how to understand this chemical effect of
. According to the coupling model proposed by Ngai,[15] an equation could be obtained as follows:
, where
and
are α- and β-relaxation time respectively,
is a constant about 2×10
s, and the coupling parameter n is associated with the interaction between particles and increases with stronger interaction. Larger negative
means that n and therefore (
are larger, thus the separation between α- and β-peaks is more significant. Furthermore,
provides an effective way to understand many chemical-related issues, such as the local structure. Historically, hard sphere dense random packing has been widely cited as a structural model for MGs.[22] It is now understood that the model fails to describe many binary MGs with significant chemical short-range order (CSRO) which is associated with atomic size ratio and
.[23] The short-range order (SRO) in Zr60Cu20Pd10Al10 liquid was investigated by the small-angle x-ray scattering,[24] and revealed that larger negative
is related to the stronger attractive interaction between solute and solvent. The coordination number (CN) of Zr–Pd pairs increases significantly with lowering liquid temperature, while that of Zr–Cu pairs increases slightly, indicating that the large negative
between Zr and Pd (−91 kJ/mol) leads to the formation of (Zr, Pd)-rich domains of SRO in the supercooled liquid.
To obtain a quantitative estimation for the CSRO in binary supercooled liquids, the CSRO degree was estimated using the Warren parameter[25]
where
is the concentration of species
j,
the total CN around species
i, and
. Negative values of
α
indicate a preference for unlike-neighbor bonds, otherwise homopolar pairs are preferred.
[25, 26] Honeycutt and Anderson pair analysis is able to characterize the local environment surrounding each atomic pair, and has been widely used to characterize the local cluster structure.
[27] In this method, a set of four integers
i,
j,
k,
j is designed to describe different local configurations. For instance, 1551 index is the characteristic of icosahedral short range order (ISRO), while 1541 and 1431 indices are the characteristic of distorted ISRO.
[26, 28] Usually, 1551, 1541, and 1431 pairs are referred to the common local atomic configuration characteristic of supercooled liquid. Table
2 summarizes the data of
,
α
, and the relative abundance
w of the selected main bonded pairs
for different MGs and liquid alloys. Figure
3 shows the dependence of
α
and
w on
. Both
α
and
w decrease with increasing negative
, which indicates the chemical effect on the local structure in MGs. Especially for Zr–Cu (
kJ/mol) and Zr–Ni (∼−53 kJ/mol) systems, it has been widely accepted that the dominant SRO in Zr–Cu is of topology with icosahedral symmetry (Voronoi index
),
[26, 31, 32] in contrast, the ISRO degree is much lower in Zr–Ni MG.
[27, 33–35] The interaction is much stronger for Zr–Ni with large negative
, and the polyhedra around Ni have lower CNs less than 10,
[34] which increases the energy of Ni-centered icosahedral clusters with CN of 12. Therefore, large negative
leads to a tendency for unlike-neighbor bonds and relatively fewer icosahedral clusters. Nevertheless, the population of icosahedral clusters is also controlled by the radius ratio of component elements.
[23, 28] The center atom of a perfect icosahedral cluster is slightly smaller, and the radius ratio is 0.902. When the atomic radius ratio deviates from 0.902, the population of other polyhedra types increases, such as the Frank–Kasper type, the BSAP type, and the TTP type.
[23] Furthermore, the electronic configuration of component elements has an important influence on the local structure, such as the covalent-like bonding in the systems containing metalloids
[36] and bond shortening due to the special interaction between the outer electrons of the component elements.
[28, 37] Thus to predict the relative population of icosahedral clusters, the multiple factors should be considered but not the single factor of
for different MG systems.
β-relaxation and SRO in MGs both have a close relation with crystallization and mechanical properties,[11, 12, 38, 39] which implies that β-relaxation may be related with SRO. Plenty of researches confirm the correlation between the nonaffine displacement and the local structures. It was indicated that in Cu–Zr MG the non-pentagonal regions, which are sources for the generation of high free volume structures, preferentially undergo irreversible atomic rearrangements in the early stage of the deformation process. Non-pentagonal regions are easily deformed plastically because of high energy and low dynamic stability.[40] The degree of local fivefold symmetry (LFFS) was used as the structural indicator to predict plastic deformation of local structures and the plastic events propagate in regions with a lower degree of LFFS.[41] Therefore, the perfect and distorted icosahedral clusters with high fivefold symmetry have relative high dynamic stability. Recently, it was found that β-relaxation is related to the cooperative motion of particles chains by the MD simulations.[42] In the simulations, they pinned randomly a small fraction of the particles to the coordinate frame, as a result the β-relaxation was suppressed more sufficiently with increasing the pinned particle fraction. These pinned particles were only allowed to participate in the affine transformation, which is similar to the icosahedral clusters with high dynamic stability in MGs. These findings indicate that icosahedral clusters could suppress the motion of β-relaxation. Our MD simulations have revealed the atomic fraction associated with icosahedral configuration for the Zr-based MGs. The fraction of the 20 most abundant Voronoi indices that appear in the Zr-based MGs is presented in Fig. 4(a), where the Voronoi index
associated with icosahedral clusters is 1.09% and 0.76% for Zr
Cu
and Zr
Ni
, respectively. The population of the Voronoi indices related to distorted icosahedral structures, such as
and
, for Zr
Cu
are also larger than those for Zr
Ni
, which basically agrees with the previous results.[28, 43] Figure 4(b) and 4(c) show the distribution of the icosahedral clusters for Zr
Cu
and Zr
Ni
respectively, which indicates that Zr
Cu
has more icosahedral clusters and the distribution is homogenous without significant aggregation for both compositions. Besides, we also performed MD simulations for Ti-based MGs, and obtained similar results. Correspondingly, Zr
Cu
and Ti60Cu40 MGs have less pronounced β-relaxations and higher
compared with that of Zr
Ni
and Ti60Ni40, respectively. These experimental and simulation results suggest the pinning effect of dynamic stable clusters on the motion of β-relaxation.