Journal of Non-Crystalline Solids 419 (2015) 51–57

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids

journal homepage: www.elsevier.com/ locate/ jnoncrysol

An empirical criterion for predicting the glass-forming ability of

amorphous alloys based on electrical transport properties
Li-fang Wang a,b, Qi-dong Zhang a, Xiao Cui a, Fang-qiu Zu a,⁎
a
School of Materials Science & Engineering, Hefei University of Technology, Hefei 230009, China
b
School of Mechanical and Material Engineering, Jiujiang University, Jiujiang 332005, China

a r t i c l e i n f o a b s t r a c t

Article history: Based on the Anderson's theory and Mott–CFO model for the electrical transport properties of disordered mate-
Received 7 February 2015 rials, an empirical criterion has been proposed for predicting the glass-forming ability (GFA) of amorphous alloys,
Received in revised form 22 March 2015 which is termed as the relative electrical resistivity difference between the amorphous and fully crystallized
Accepted 30 March 2015
states at room temperature (RT), i.e., Δρ = (ρRT RT RT RT RT
amor − ρcrys)/ρcrys, ρamor and ρcrys denote the room-temperature
Available online 3 April 2015
electrical resistivities of an amorphous alloy and its corresponding crystal, respectively. A higher Δρ value of an
Keywords:
alloy always correlates with a better GFA, which has been unambiguously confirmed by a large number of results
Empirical criterion; in the Cu–Zr–Al–Ag, La–Ce–Al–Co, and Fe–Mo–Y–B alloy systems. Compared with the widely used criteria the re-
Glass-forming ability; duced glass transition temperature Trg, the supercooled liquid range ΔTx, and the parameter γ, in which the glass
Amorphous alloy; transition temperature Tg is necessary and its determination is sometimes arbitrary with inaccuracy for some
Electrical transport property amorphous alloys, Δρ can be quite easily and more exactly obtained, and is more consistent with the GFA. The
empirical criterion is very useful especially for optimizing the compositions and pinpointing the best glass former
with less cost in a given system.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction aforementioned criteria all have some limitations more or less, there-
fore it is quite essential to look for another relatively universal criterion
Since the discovery of the Au–Si binary metallic glass in 1960 [1], being able to be quantified easily to estimate the GFA of metallic glasses.
various empirical parameters have been proposed to assess the glass- Previous studies on understanding the GFA mainly focused on atom-
forming ability (GFA) of metallic glasses and were expected to serve ic structures and related properties. It was lack of effective efforts to un-
as guidelines for designing and developing new metallic glasses. derstand the GFA from electronic aspects and electronic transport
Among these GFA indicators, the reduced glass transition temperature properties. Recently, scientists have realized that there may be some re-
Trg (=Tg/Tl) [2], the supercooled liquid range ΔTx (=Tx − Tg) [3], and lationships between the electronic structures or electrical transport
the parameter γ (=Tx/Tg + Tl) [4] (Tg, Tx, and Tl denote the glass transi- properties and GFA of metallic glasses [8,9]. The excellent GFA of the
tion temperature, the onset crystallization temperature, and the Pd42.5Ni7.5Cu30P20 metallic glass is correlated with the decrease of the
liquidus temperature, respectively) are the most widely used. However, density of state (DOS) at the Fermi energy EF, N(EF) [8]. Dong et al.
accumulated data show that they fail to predict the GFA in some [10] investigated the influence of the free electrons per atom e/a on
cases [5–7]. Moreover, in the often-cited criteria Trg, ΔTx, and γ, Tg is in- the GFA of metallic glasses. Therefore, it is believed that exploring the
dispensable and its determination is sometimes difficult and arbitrary electronic structures and electrical transport properties of glassy alloys
without veracity for some metallic glasses, which reduces the feasibility will be helpful for better understanding the GFA.
and accuracy of these criteria. Based on the Anderson's theory and Mott–CFO model for the electri-
From the viewpoint of physical metallurgy, the critical cooling rate cal transport properties of disordered materials, the present paper will
Rc is the most direct GFA indicator. However, it is hard to be measured attempt to put forward an empirical criterion for predicting the GFA of
precisely in the actual cooling process. Another slightly less direct pa- metallic glasses. This criterion has been vindicated by lots of experimen-
rameter is the maximum diameter Dmax of a rod-like metallic glass tal results for the alloy systems of Cu–Zr–Al–Ag, La–Ce–Al–Co, and Fe–
that can be cast into fully amorphous structure. Although Dmax is easier Mo–Y–B.
to be obtained than Rc, tedious experiments are inevitable to determine
Dmax for different compositions in an alloy system. In a word, the 2. Derivation of the empirical criterion

⁎ Corresponding author. Metallic glasses are disordered materials without long-range order
E-mail address: [email protected] (F. Zu). as compared to crystals. Anderson [11] divided the electronic states of

http://dx.doi.org/10.1016/j.jnoncrysol.2015.03.036
0022-3093/© 2015 Elsevier B.V. All rights reserved.
52 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57

a disordered system into the extended and localized states. The mean
free path of the electrons in the extended state is larger than interatomic
distance, while the electrons in the localized state are just limited within
certain local regions. Based on the Anderson's theory, Mott–Cohen–
Fritzsche–Ovshinsky (Mott–CFO) model [12,13] was developed. The en-
ergy eigenvalues are grouped in bands. Inside each band there exist crit-
ical energies Ec and E′c (called mobility edges) which separate electron
states into different natures. The states between Ec and E′c are extended,
while those in the tails of the density of states are localized. With the
disordered degree increasing, the mobility edges move towards the
center of the band, and more electrons are refined in the localized states.
Meanwhile, the activation energy for the hopping of electrons from the
localized states to the extended states increases. The extended-state
electrons give the main contribution to the electrical conductivity σ
(σ = 1/ρ, the reciprocal of the electrical resistivity ρ), whereas the con-
tribution to the electrical conductivity coming from the localized-state
electrons can be negligible unless activated to the extended states [14].
In a crystalline metal, the EF falls within a band. The electron wave Fig. 1. The sketch of the resistivity-versus-temperature characteristics of the amorphous
metal Pd0.8Si0.2 and its crystalline form.
functions are extended-state Bloch functions, and in a partly filled
band these plane-wave-like states give rise to the large electrical con-
ductivity. But the conduction of amorphous alloys occurs via the rela- the intrinsic properties of an amorphous solid and can be understood in
tively fewer extended-state electrons and is limited by disorder- terms of the disorder inherent in the structure of an amorphous solid
induced scattering processes. [20]. Further, there exists an interrelationship between the law of vari-
The electrical resistivity of amorphous alloys is often explained in ation in electrical resistivity and the ordering degree [21].
terms of the Ziman theory. The Ziman theory, originally used to describe From what has been discussed above, the value ρe,l/ρe,s reflects the dis-
the electrical transport properties of liquid simple metals [15], is a near- ordered degree of a liquid metallic system. However, it is hard to measure
ly free electron theory in which the current is carried solely by an the liquid and solid resistivities at Tm, so the widespread use of ρe,l/ρe,s for
unhybridized sp band and the scattering is calculated by use of an ap- evaluating the disordered degree of a liquid metallic system is constrained
propriate pseudopotential. Evans et al. [16] extended this to apply to liq- in real practices. Due to much higher resistivity and much lower sensitiv-
uid transition metals by replacing the pseudopotential by a single-site, ity to temperature of a metallic glass relative to its crystal, the ratio of the
on-shell t matrix t(q, EF). This leads to the so-called extended Ziman for- electrical resistivities of the amorphous and correspondingly fully
mula. Meisel and Cote [17] pointed out that this extension could be ap- crystallized states at room temperature (RT) ρRT RT
amor/ρcrys is larger than ρe,
plied to amorphous alloys as well. l/ρe,s. An alloy with a larger ρe,l/ρe,s will have an even larger
Based on the simple assumption that the atoms in a liquid metal vi- ρRT RT RT RT
amor/ρcrys. So, in a sense, ρamor/ρcrys can more pronouncedly and sensitive-
brate at about slowly-varying mean positions, Mott [18] has proposed ly disclose the disordered degree of a liquid metallic system than ρe,l/ρe,s,
an equation for the ratio of liquid/solid conductivity σe,l/σe,s at a metal's besides which, ρRT RT
amor/ρcrys reflects the structure difference between an
melting point Tm, which is expressed as: amorphous alloy and its corresponding crystal.
  For a given alloy, the room-temperature electrical resistivities of the
l
σ e;l =σ e;s ¼ exp −80Δs Hm =T m ð1Þ amorphous ρRT RT
amor and fully crystallized states ρcrys can be readily obtain-
ed. Here, Δρ is defined as the relative electrical resistivity difference at
room temperature between two states, as shown in the following
where ΔlsHm is the enthalpy of melting. Eq. (1) can be simplified to the
formula:
following expression:
   
RT RT RT
l
ρe;l =ρe;s ¼ exp 80Δs Sm ð2Þ Δρ ¼ ρamor −ρcrys =ρcrys : ð3Þ

where ΔlsSm is the entropy of melting. In spite of his simple treatment, According to the above discussion, Δρ can reliably and sensitively
the agreement obtained between theory and experiment for normal betoken the disordered degree of an original liquid and the structure
metals is surprisingly good. Based on the modern theory of electron difference between an amorphous alloy and its corresponding crystal,
transport, Ziman [15] has provided a reasonable interpretation for the in spite of yielding no specific structural information. It can be logically
ratio ρe,l/ρe,s (= σe,s/σe,l), which is not covered in details here. deduced that the higher the value Δρ for an alloy, the larger the struc-
Entropy is an important variable which describes the randomness of ture difference between an amorphous alloy and its corresponding crys-
a complex system. The larger the entropy of melting ΔlsSm, the more the tal, the more the disorder of its liquid, the harder it is to crystallize the
disorder of the liquid. According to Eq. (2), a higher degree of disorder molten alloy in its solidification process because of the nucleation and
means a larger ρe,l/ρe,s. Hence, the value ρe,l/ρe,s can be used as a measure growth difficulties, and the stronger the GFA. Thereby, Δρ is put forward
of the disordered degree of a liquid metallic system. The larger the value for evaluating the GFA of metallic glasses. The higher the value Δρ, the
ρe,l/ρe,s, the more disordered the melt, and it will be more difficult to stronger the GFA will be.
transform into a crystal during the liquid cooling process because of It is worth mentioning that there are other factors influencing the
the nucleation difficulty, and the disordered structure tends to be easily electrical transport properties besides the number of the extended-
retained, often referred to as the “confusion principle” [19]. state electrons and the disordered degree. For instance the amorphous
Metallic glasses are essentially “frozen liquids”, and to a large extent alloy state (such as structural relaxation and metallic glasses containing
an amorphous solid inherits the structure from its original liquid metal, few nano-crystalline particles) will affect the electrical resistivity of me-
which is demonstrated by the sketch of the resistivity-versus- tallic glasses. The electrical resistivity of the crystallized alloys will also
temperature characteristics of the amorphous and crystalline Pd0.8Si0.2 change with the crystallized alloy state (such as the grain size, crystal
[20]. As Fig. 1 shows, the amorphous Pd0.8Si0.2 has much higher resistiv- grain tropism, and defect). On the one hand, the amorphous and crystal-
ity and much lower sensitivity to temperature than its crystal, which is lized states are connected with the temperature–time history, on the
 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57 53

other hand, are related with innate material characteristics. In a certain 4. Results
sense, the innate properties of material likewise reflect the ease of vitri-
fication and the stability of an amorphous alloy. Therefore, when the Fig. 2 shows the DSC and DTA curves of the CuxZr84 − xAl8Ag8 (x =
GFA of an alloy system is explored, in order to eliminate the probable 42, 40, 38, and 36) BMGs at a heating rate of 10 K/min under an argon
influence of external conditions on the electrical resistivity, it is essential atmosphere, respectively. Table 1 lists the characteristic parameters of
to adopt the same producing technique such as the same thermal these BMGs including Tg, Tx, Tm and Tl, and the GFA indicators ΔTx, Trg,
history and the same raw materials during the sample preparation and γ. The Dmax of four BMGs [23] is also listed in Table 1. Among
process. three GFA indicators, only ΔTx is suitable for characterizing the GFA of
Well then, whether the above theoretical analysis is proper and the Cu–Zr–Al–Ag BMGs.
empirical criterion is valid in practice? Taking the doubt into account, Fig. 3(a) shows the temperature dependence of the electrical resis-
the following experiments were carried out. tivity for the CuxZr84 − xAl8Ag8 (x = 42, 40, 38, and 36) BMGs and
their crystals from RT to 1050 K at a heating rate of 0.167 K/s. It can be
seen that the resistivity of the BMGs is much larger than that of the crys-
3. Experimental procedures talline counterparts. Fig. 3(b) plots the magnification of the electrical re-
sistivity of four BMGs below Tx. Both Tg and Tx obtained from the DSC
Ingots with nominal compositions of CuxZr84 − xAl8Ag8 (x = 42, 40, curves were marked in Fig. 3(b).
38, and 36), (LaxCe1 − x)65Al10Co25 (x = 0.3, 0.4, 0.5, 0.6, 0.7, and 0.8), The values of ρ RT RT
amor , ρ crys , and Δρ as well as D max for the
and Fe72 − xMoxY6B22 (x = 1, 2, 3, 4, 5, and 6) were prepared by arc- Cu x Zr 84 − x Al 8 Ag 8 (x = 42, 40, 38, and 36) alloys are listed in
melting mixtures of the elements, each with 99.5 wt.% purity or better, Table 2. Fig. 4 shows the compositional dependences of ΔT x , T rg ,
under a Ti-gettered high-purity argon atmosphere. Each ingot was γ, Δρ, and D max for the Cu x Zr 84 − x Al 8 Ag 8 BMGs, respectively. As
inverted and remelted four times to ensure a chemical homogeneity shown in Fig. 4(a)–(c), only ΔT x can adequately reflect the GFA.
followed by sucking the melt into a copper mold with an internal cavity The change of Δρ with the compositions (Fig. 4(d)) is similar to
of 0.01 × 0.01 × 0.08 m3. The amorphous nature of the samples was ver- that of Dmax (Fig. 4(e)), with a higher Δρ corresponding to a larger
ified by X-ray diffractometry on a D/MAX-2500V diffractometer using Dmax, implying that Δρ is a good indicator in evaluating the GFA in
CuKα radiation. The thermal analysis of the CuxZr84 − xAl8Ag8 bulk the Cu–Zr–Al–Ag alloys.
metallic glasses (BMGs) was carried out by differential scanning Table 3 collects the maximum diameters and characteristic parame-
calorimetry (DSC) on a PerkinElmer DSC-8000 differential scanning cal- ters of the (LaxCe1 − x)65Al10Co25 (x = 0.3, 0.4, 0.5, 0.6, 0.7, and 0.8) and
orimeter and differential thermal analysis (DTA) on a Netzsch STA449F3 Fe72 − xMoxY6B22 (x = 1, 2, 3, 4, 5, and 6) BMGs, which are taken from
differential thermal analyzer at a heating rate of 0.167 K/s under a Refs. [24] and [25], respectively. The ρRT RT
amor, ρcrys, and Δρ values of these
flowing high-purity argon atmosphere. The values of the characteristic alloys are listed in Table 3 as well.
temperatures were determined from the DSC and DTA traces with an Fig. 5 exhibits the changes of ΔTx, Trg, γ, Δρ, and Dmax with the com-
accuracy of ± 1 K. Rectangular sheet samples with dimensions of positions for the (LaxCe1 − x)65Al10Co25 BMGs, respectively. There are no
0.001 × 0.04 × 0.05 m3 were cut from the as-cast BMGs by spark evident correlations between the often-cited criterion ΔTx, Trg, or γ and
machining for the electrical resistivity measurement. The electrical re- GFA as indicated in Fig. 5(a)–(c). However, as presented in Fig. 5(d)–(e),
sistivity was detected by the direct current 4-probe method with an the same changing tends of Δρ and Dmax with the compositions show
error of ±1 × 10−10 Ω·m, for which details have already been reported that the alloy with a better GFA has a higher value of Δρ. Thus, among
[22]. The dimensional precision of the specimens for the electrical the ΔTx, Trg, γ, and Δρ criteria, Δρ is the best in judging the GFA for the
resistivity measurement was about ± 1% on thickness and ± 2% for (LaxCe1 − x)65Al10Co25 alloys.
the other dimensions, yielding about ±5% error on the absolute value Fig. 6 displays the relationships between the composition and ΔTx,
of the resistivity. To achieve reliable results, all the measurements Trg, γ, Δρ, as well as Dmax for the Fe72 − xMoxY6B22 (x = 1, 2, 3, 4, 5,
were five times repeatedly performed and the average values were and 6) BMGs, respectively. Because the Tg of the Fe66Mo6Y6B22 BMG
taken. To ensure the same thermal history, the fully crystallized alloys was not observed during heating process, it is impossible to evaluate
were prepared by annealing the amorphous samples at a temperature the GFA of the alloy according to the often-cited parameter ΔTx, Trg, or
of 100 K below the melting temperature Tm for 600 s in a vacuum γ. Moreover, from Fig. 6(a)–(c), it can be seen that the values of ΔTx,
furnace (10−3 Pa). Trg, and γ parameters of the other BMGs are not in good agreement

Fig. 2. (a) DSC and (b) DTA curves of the CuxZr84 − xAl8Ag8 BMGs (x = 42, 40, 38, and 36).
54 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57

Table 1 Table 2
Maximum diameters and characteristic parameters of the CuxZr84− xAl8Ag8 (x = 42, 40, Dmax, ρRT RT
amor, ρcrys, and Δρ for the CuxZr84− xAl8Ag8 (x = 42, 40, 38, and 36) alloys.
38, and 36) BMGs.
Alloys Dmax (mm) ρRT
amor (μΩ·cm) ρRT
crys (μΩ·cm) Δρ (%)
Alloys Dmax Tg Tx Tm Tl ΔTx Trg γ
Cu42Zr42Al8Ag8 12 181.36 ± 9.12 101.38 ± 6.56 78.89 ± 15.47
(mm) (K) (K) (K) (K) (K)
Cu44Zr40Al8Ag8 15 183.58 ± 9.18 87.88 ± 5.17 108.90 ± 16.33
Cu42Zr42Al8Ag8 12 710 762 1078 1234 52 0.5754 0.3920 Cu38Zr38Al8Ag8 20 186.59 ± 9.98 71.55 ± 4.94 160.78 ± 20.85
Cu44Zr40Al8Ag8 15 702 772 1108 1166 70 0.6021 0.4133 Cu36Zr36Al8Ag8 25 189.67 ± 9.57 57.23 ± 4.49 231.42 ± 24.57
Cu38Zr38Al8Ag8 20 690 784 1102 1148 94 0.6010 0.4266
Cu36Zr36Al8Ag8 25 670 771 1097 1146 101 0.5846 0.4246

It is compelling that below Tx the BMG with a better GFA has a higher
with their GFA. Nevertheless, the criterion Δρ and Dmax exhibit a similar electrical resistivity, as shown in Fig. 3. It is generally believed that the
variation tendency with the compositions, showing the alloy with a better the GFA, the more the disorder and the denser the atomic struc-
larger Δρ corresponding to a larger maximum diameter (and therefore ture of an amorphous alloy [27–29]. Because the atomic structure is
to easier glass formation). Therefore, Δρ is the most appropriate for closely related to the electronic structure, it is reasonable to speculate
predicting the GFA of the Fe72 − xMoxY6B22 alloys. that the valence band electrons have less space in a close-packed metal-
lic glass than those in an incompact metallic glass [30]. The decrease in
the size of the valence electron cloud will strengthen the attractive Cou-
5. Discussion lomb force between the atomic nucleus and the valence electrons, and
then increase the binding energy of the valence electrons [31]. The
Owing to the sensitivity of the electrical resistivity to the atomic stronger binding energy will refine more valence electrons in the local-
structure, the resistivity measurement can offer an effective method ized states, correspondingly the extended-state electrons decrease. The
for characterizing the structures of amorphous and crystalline materials higher disordered degree and the fewer extended-state electrons slow
[26]. As shown in Fig. 3, in the lower temperature range the electrical re- down the electron flow and increase the electrical resistivity.
sistivity of BMGs decreases very little and approximately linearly with The electrical conductivity σ (σ = 1/ρ) is generally written as [32]
increasing temperature. With further increasing temperature, the elec-
2 2 2
trical resistivity deviates to a smaller slope from the linear behavior, σ ¼ e υ F τN ðE F Þ=3 ¼ e υ F Λ F NðE F Þ=3 ð4Þ
which is mainly due to the relaxation mechanism transformation. The
resistivity curve has a hump near Tg. As temperature increases, because where e is the electronic charge, υF is the Fermi velocity, τ is the relaxa-
of the formation of primary phases the electrical resistivity exhibits a tion time, ΛF is the mean free path given by ΛF = υFτ and N(EF) is the
sharply drop at Tx. There are two sudden drops suggesting that the crys- density of states at the Fermi energy EF. If υFΛF is kept constant, the resis-
tallization will complete in two stages. After finishing the first-stage tivity should then be inversely proportional to N(EF).
crystallization, some crystals form, which act as scattering centers In the metal–metal Cux Zr 84 − x Al 8Ag8 or rare-earth-metal
resulting in the transient increase of the electrical resistivity. As temper- (La x Ce1 − x )65 Al10 Co 25 alloys, the atoms are mainly linked by
ature continues to increase, the second sudden drop of the electrical metal–metal bonds. The band structure of glassy and crystalline states in-
resistivity may be induced by the secondary crystallization of the re- fluences the concentration of conduction electrons in a complicated man-
maining supercooled liquid or the transformation of the primary meta- ner, which leads to the change of N(EF). Copper and zirconium or
stable phases. For the full crystallized alloys, the electrical resistivity lanthanum and cerium have different semi-filled d-states or d- and f-
increases with increasing temperature. When heat is applied to metallic states. The DOS of the crystalline and amorphous phases for all of the
crystals, the atoms gain thermal energy and vibrate at a particular am- studied alloys is predominantly dominated by the d-states or s-d hybrid-
plitude and frequency. Thus, the increase in the temperature would sig- ization or d- and f-states in the vicinity of the Fermi surface [33]. When the
nificantly increase the probability of scattering electrons within the Cu/Zr or La/Ce ratio is varied, the concentration of conduction electrons
crystal. Therefore more collisions will take place, slowing down the and the degree of disorder will change accordingly, which results in the
electron flow. This decreases the transfer rate of electrons and then in- shifting of EF and is responsible for the resistivity behavior. The change
creases the electrical resistivity. of the concentration of conduction electrons also corresponds to the

Fig. 3. (a) The temperature dependence of the electrical resistivity of the CuxZr84 − xAl8Ag8 (x = 42, 40, 38, and 36) BMGs and their crystalline counterparts from RT to 1050 K.
(b) Magnification of the electrical resistivity of four BMGs below Tx.
 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57 55

Fig. 4. The compositional dependences of (a) ΔTx, (b) Trg, (c) γ, (d) Δρ, and (e) Dmax for the
Fig. 5. The changes of (a) ΔTx, (b) Trg, (c) γ, (d) Δρ, and (e) Dmax with the compositions of
CuxZr84 − xAl8Ag8 (x = 42, 40, 38, and 36) BMGs, respectively. The solid lines are guides to
the (LaxCe1 − x)65Al10Co25 (x = 0.3, 0.4, 0.5, 0.6, 0.7, and 0.8) BMGs, respectively. The solid
the eyes.
lines are guides to eyes.

variation of compositional short-range ordering. So the CuxZr84 − xAl8Ag8 of the Fe/Mo ratio, which causes the different electrical transport
or (LaxCe1 − x)65Al10Co25 amorphous and crystalline alloys exhibit the properties.
different electrical transport properties with varying composition. To understand the glass formation and GFA from electronic aspects,
In the metal–metalloid Fe72 − xMoxY6B22 alloy system, B is a metal- a nearly free electron model [35] considering the interaction of conduc-
loid element. Besides the metal–metal bonds, there also exists metal– tion electrons with the ionic assemblages was proposed and it conclud-
metalloid and metalloid–metalloid bonding types in the alloy system. ed that the stability of amorphous structure occurs when the Fermi
The metal–metalloid bonds are much stronger than the metal–metal surfaces of the electrons coincide with the quasi-Brillouin boundaries.
bonds due to the strong attractive interaction and are primarily respon- Similar to liquid metals and quasicrystalline alloys, a pseudo gap is ex-
sible for the degree of atomic packing in the glasses and local strong or- pected to occur in the energy bands for metallic glasses, and a corre-
dering structures. In the meanwhile, the covalent-like metal–metalloid sponding minimum of N(EF) [35–37]. Nagel and Tauc suggested that
bonds play the most important role in connecting atoms [34]. The an excellent GFA is closely related to the decrease of N(EF) [35]. Howev-
strong metal–metalloid bonds can closely tie the neighboring atoms to- er, the minimum N(EF) is not always corresponding to the best of GFA
gether and effectively hinder the atomic rearrangement, which would [8,38]. Shen et al. [39] concluded and authenticated that the GFA of an
effectively prevent the crystallization of the Fe-based BMGs. The locali- alloy is associated not only with the DOS of the glassy state, but also
zation of electrons among atoms, connected with the shared electron with the DOS of the crystalline state. The room-temperature electrical
pairs, determines the formation of the covalent-like metal–metalloid resistivities of an amorphous alloy and its corresponding crystal are
bonds. Moreover, the bond energy of the covalent-like metal–metalloid closely related to the DOS of the glassy and crystalline states, respective-
bonds will be enhanced by the increased localized electrons. In the Fe– ly. That is to say, the new criterion Δρ takes into account the DOS of both
Mo–Y–B alloy system, the concentration of localized electrons and glassy and crystalline alloys and exhibits the DOS difference between
metal–metalloid atomic pairs, the bond energy of metal–metalloid the amorphous and crystalline alloys, which may be the physical origin
bonds and the structural packing density will change with the variation of the GFA and manipulate the GFA of an alloy.

Table 3
Maximum diameters [24,25], characteristic parameters [24,25], ρRT RT
amor, ρcrys, and Δρ of the (LaxCe1− x)65Al10Co25 (x = 0.3, 0.4, 0.5, 0.6, 0.7, and 0.8) and Fe72− xMoxY6B22 (x = 1, 2, 3, 4, 5, and
6) BMGs.

Based metal Alloys Dmax (mm) Tg (K) Tx (K) Tm (K) Tl (K) ΔTx (K) Trg γ ρRT
amor (μΩ·cm) ρRT
crys (μΩ·cm) Δρ (%)

LaCe-based (La0.3Ce0.7)65Al10Co25 9 416 436 675 778 20 0.54 0.365 168.97 ± 8.55 112.90 ± 5.74 49.66 ± 12.67
(La0.4Ce0.6)65Al10Co25 12 425 448 689 789 23 0.54 0.369 169.02 ± 8.23 99.27 ± 4.98 70.26 ± 13.31
(La0.5Ce0.5)65Al10Co25 15 427 453 696 776 26 0.55 0.376 171.32 ± 8.67 90.92 ± 4.73 88.43 ± 14.74
(La0.6Ce0.4)65Al10Co25 20 437 467 705 835 30 0.52 0.367 172.37 ± 8.49 87.48 ± 4.47 97.04 ± 14.81
(La0.7Ce0.3)65Al10Co25 25 437 472 710 850 35 0.51 0.367 173.76 ± 8.63 86.53 ± 4.52 100.81 ± 15.20
(La0.8Ce0.2)65Al10Co25 12 439 476 712 869 37 0.50 0.364 169.33 ± 8.42 99.82 ± 5.03 69.64 ± 13.47
Fe-based Fe71Mo1Y6B22 2.5 902 960 1376 1517 58 0.595 0.397 162.43 ± 8.33 130.84 ± 6.86 24.14 ± 11.61
Fe70Mo2Y6B22 3.5 907 969 1378 1508 62 0.601 0.401 164.27 ± 8.37 128.75 ± 6.57 27.59 ± 11.60
Fe69Mo3Y6B22 6 908 958 1379 1488 50 0.610 0.399 167.45 ± 8.42 124.48 ± 6.35 34.52 ± 11.87
Fe68Mo4Y6B22 6.5 915 944 1385 1488 29 0.615 0.393 168.15 ± 8.41 120.72 ± 6.15 39.29 ± 12.06
Fe67Mo5Y6B22 3.5 920 941 1380 1483 21 0.620 0.392 164.54 ± 8.24 128.36 ± 6.44 28.19 ± 11.44
Fe66Mo6Y6B22 2 – 937 1384 1473 – – – 162.28 ± 8.18 131.17 ± 6.63 23.72 ± 11.29
56 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57

to the large differences in the electronegativity, the mean free path, the
band structure and the atomic size between the non-neighboring ele-
ments, both the electronic structure (DOS near EF, chemical bonding,
and valence band electron concentration) and the atomic structure
(packing efficiency and local atomic arrangements) of the glassy and
crystalline states of the alloy system will be drastically altered. The al-
loys would belong to a new alloy system. Therefore, the new empirical
criterion is no longer valid.

6. Conclusions

In this paper, based on the Anderson's theory and Mott–CFO model

for the electrical transport properties of disordered materials, an
empirical criterion for predicting the GFA of metallic glasses is devel-
oped, i.e., the relative electrical resistivity difference between the amor-
phous and corresponding fully crystallized states at room temperature
Δρ (= (ρRT RT RT
amor − ρcrys)/ρcrys). The alloy with a larger Δρ has a better
GFA. This newly defined criterion has been successfully applied to the
Cu–Zr–Al–Ag, La–Ce–Al–Co, and Fe–Mo–Y–B alloy systems. The new
criterion reveals that the DOS difference between the amorphous and
corresponding crystalline alloys may be the physical origin of the GFA
and manipulate the GFA of an alloy. In contrast to the widely used
criteria Trg, ΔTx, and γ, Δρ can be quite easily and more exactly quanti-
fied, and shows a better correlation with the GFA. With the empirical
Fig. 6. The relationships between the composition and (a) ΔTx, (b) Trg, (c) γ, (d) Δρ, as well criterion, the effect and rule of micro-alloying, element substitution, or
as (e) Dmax for the Fe72 − xMoxY6B22 (x = 1, 2, 3, 4, 5, and 6) BMGs, respectively. The solid changing composition ratios on the GFA can be easily detected in ad-
lines are guides to the eyes.
vance by casting or even spinning samples with the same size and with-
out X-ray inspections for a complex system, and the best glass former in
Glass formation is a competing growth between liquid and the most the given system can be located promptly with less cost.
stable crystalline phases. The GFA of an alloy is determined by the out-
come of such competition. Factors that stabilize the liquid structure
clearly enhance the GFA. Factors which facilitate the formation of the Acknowledgments
competing crystalline phases, affected by the difficulties during the nu-
cleation and growth processes, must sharply reduce the GFA [40]. Dur- The authors are grateful to Professor B. Zhang for extending help
ing the melt solidification the atomic rearrangement for ordering during the experimental part of the work. This work was financially
requires a highly correlated set of atomic motions. If the atomic motions supported by the National Natural Science Foundation of China (Grant
are retarded within an experimental time scale, the sufficient disor- Nos. 50971053 and 11304073) and the National Basic Research Pro-
dered structure will be captured to form a glass. As elaborated earlier, gram of China (973 Program) (Grant No. 2012CB825702).
Δρ reflects the disordered degree of a liquid metallic system and the
structure difference between an amorphous alloy and its corresponding References
crystal. In essence, the Δρ criterion has considered two aspects, the sta-
[1] W. Klement, R.H. Willens, P. Duwez, Non-crystalline structure in solidified gold–silicon
bility of liquid phase and the resistance to crystallization. Therefore, the alloys, Nature 187 (1960) 869–870.
criterion Δρ can predict the GFA of an alloy. [2] W.L. Johnson, Bulk glass-forming metallic alloys: science and technology, MRS Bull.
The above extensive experimental results confirm the validity of Δρ 24 (1999) 42–56.
[3] A. Inoue, Stabilization of metallic supercooled liquid and bulk amorphous alloys,
for predicting the GFA not only in the metal–metal amorphous alloys
Acta Mater. 48 (2000) 279–306.
but also in the metal–metalloid type of glassy alloys. Compared with [4] Z.P. Lu, C.T. Liu, A new glass-forming ability criterion for bulk metallic glasses, Acta
the widely used criteria Trg, ΔTx, and γ, Δρ is more consistent with the Mater. 50 (2002) 3501–3512.
GFA. In addition, the value Δρ can be easily and accurately quantified [5] T.D. Shen, R.B. Schwarz, Bulk ferromagnetic glasses prepared by flux melting and
water quenching, Appl. Phys. Lett. 75 (1999) 49–51.
by casting or even spinning samples with the same size. Hence, Δρ is ap- [6] A.H. Cai, X. Xiong, Y. Liu, W.K. An, J.Y. Tan, Y. Pan, Design of new Zr–Al–Ni–Cu bulk
plicable for investigating the GFA of metallic glasses. metallic glasses, J. Alloys Compd. 468 (2009) 432–437.
The merit of the empirical criterion Δρ is obvious. In exploring and [7] L. Ma, A. Inoue, On glass-forming ability of Fe-based amorphous alloys, Mater. Lett.
38 (1999) 58–61.
developing new BMGs, it is a common resort to increase the GFA by [8] S. Hosokawa, H. Sato, N. Happo, K. Mimura, Y. Tezuka, T. Ichitsubo, E. Matsubara, N.
micro-alloying, element substitution, or changing composition ratios Nishiyama, Electronic structure of Pd42.5Ni7.5Cu30P20, an excellent bulk metallic glass
step by step for a given alloy system, and it is inevitable to tediously former: comparison to the Pd40Ni40P20 reference glass, Acta Mater. 55 (2007)
3413–3419.
melt and cast various diameter samples and thereafter with lots of X- [9] C. Dong, Y.M. Wang, J.B. Qiang, D.H. Wang, W.R. Chen, C.H. Shek, Composition rules
ray inspections for each composition. However, with the relationship from electron concentration and atomic size factors in Zr–Al–Cu–Ni bulk metallic
between GFA and Δρ, the effect and rule of micro-alloying, element sub- glasses, Mater. Trans. 45 (2004) 1177–1179.
[10] Y.M. Wang, C.H. Shek, J.B. Qiang, C.H. Wong, W.R. Chen, C. Dong, The e/a factor
stitution, or changing composition ratios on the GFA can be easily de- governing the formation and stability of (Zr76Ni24)1− xAlx bulk metallic glasses,
tected in advance by casting or even spinning samples with the same Scr. Mater. 48 (2003) 1525–1529.
size and without X-ray inspections so that the best glass former in a [11] P.W. Anderson, Absence of diffusion in certain random lattices, Phys. Rev. 109
(1958) 1492–1505.
given system can be pinpointed quickly. Therefore, lots of tedious and
[12] N.F. Mott, Electrons in disordered structures, Adv. Phys. 16 (1967) 49–114.
costly works can be avoided. [13] M.H. Cohen, H. Fritzsche, S.R. Ovshinsky, Simple band model for amorphous semi-
Finally, it should be pointed out that the empirical criterion works conducting alloys, Phys. Rev. Lett. 22 (1969) 1065–1068.
for a given alloy system in which one constituent element is replaced [14] N.F. Mott, Conduction in glasses containing transition metal ions, J. Non-Cryst. Solids
1 (1968) 1–17.
by a neighboring element in the periodic table. If a non-neighboring el- [15] J.M. Ziman, A theory of the electrical properties of liquid metals. I: the monovalent
ement is substituted for one major constituent element in quantity, due metals, Philos. Mag. 6 (1961) 1013–1034.
 L. Wang et al. / Journal of Non-Crystalline Solids 419 (2015) 51–57 57

[16] R. Evans, H.J. Güntherodt, H.U. Künzi, A. Zimmermann, An extension of the Faber– [30] L. Marot, R. Schoch, R. Steiner, E. Meyer, Rhodium and silicon system: I. Glassy me-
Ziman formula to liquid alloys of transition metals, Phys. Lett. A 38 (1972) 151–152. tallic alloy formation, Nanotechnology 21 (2010) (365706-1-365706-6).
[17] L.V. Meisel, P.J. Cote, Application of the extended Ziman theory to amorphous nickel– [31] Y.Z. Lu, Y.J. Huang, G. Wang, J. Shen, Insights into the ultrahigh glass-forming ability
phosphorus alloys, Phys. Rev. B 15 (1977) 2970–2973. of the Fe–Co–Cr–Mo–C–B–Y alloy system from the electronic-structure perspective,
[18] N.F. Mott, The resistance of liquid metals, Proc. R. Soc. Lond. A 146 (1934) 465–472. EPL 99 (2012) (37004-1-37004-5).
[19] A.L. Greer, Confusion by design, Nature 366 (1993) 303–304. [32] U. Mizutani, M. Sasaura, Y. Yamada, T. Matsuda, Electronic structure and electron
[20] P. Duwez, Structure and properties of alloys rapidly quenched from the liquid state, transport of Ca–Mg–Al metallic glasses, J. Phys. F Met. Phys. 17 (1987) 667–678.
ASM Trans. Quat. 60 (1967) 605–633. [33] U. Mizutani, M. Tanaka, H. Sato, Studies of negative TCR and electronic structure of
[21] P. Allia, D. Andreone, R.S. Turtelli, F. Vinai, G. Riontino, Structural relaxation and ir- nonmagnetic metallic glasses based on Y, J. Phys. F. Met. Phys. 17 (1987) 131–141.
reversible changes of electrical resistivity of Fe–Ni–Mo–B amorphous alloys, J. [34] K. Tanaka, T. Saito, K. Suzuki, R. Hasegawa, Role of atomic bonding for compound
Appl. Phys. 53 (1982) 8798–8804. and glass formation in Ni–Si, Pd–Si, and Ni–B systems, Phys. Rev. B 32 (1985)
[22] J. Guo, F.Q. Zu, Z.H. Chen, S.B. Zheng, Y. Yuan, Exploration of a new method in deter- 6853–6860.
mining the glass transition temperature of BMGs by electrical resistivity, Solid State [35] S.R. Nagel, J. Tauc, Nearly-free-electron approach to the theory of metallic glass al-
Commun. 135 (2005) 103–107. loys, Phys. Rev. Lett. 35 (1975) 380–383.
[23] W. Zhang, Q.S. Zhang, C.L. Qin, A. Inoue, Synthesis and properties of Cu–Zr–Al–Ag [36] P.F. Guan, T. Fujita, A. Hirata, Y.H. Liu, M.W. Chen, Structural origins of the excel-
glassy alloys with high glass-forming ability, Mater. Sci. Eng. B 148 (2008) 92–96. lent glass forming ability of Pd40Ni40P20, Phys. Rev. Lett. 108 (2012) (175501-1-
[24] R. Li, S.J. Pang, C.L. Ma, T. Zhang, Influence of similar atom substitution on glass for- 175501-5).
mation in (La–Ce)–Al–Co bulk metallic glasses, Acta Mater. 55 (2007) 3719–3726. [37] Q. Jiang, B.Q. Chi, J.C. Li, A valence electron concentration criterion for glass-
[25] X.M. Huang, C.T. Chang, Z.Y. Chang, X.D. Wang, Q.P. Cao, B.L. Shen, A. Inoue, J.Z. Jiang, formation ability of metallic liquids, Appl. Phys. Lett. 82 (2003) 2984–2986.
Formation of bulk metallic glasses in the Fe–M–Y–B (M = transition metal) system, [38] V.L. Moruzzi, P. Oelhafen, A.R. Williams, R. Lapka, H.J. Güntherodt, Theoretical and
J. Alloys Compd. 460 (2008) 708–713. experimental electronic structure of Zr-based transition-metal glasses containing
[26] M. Dikeakos, Z. Altounian, M. Fradkin, Electronic transport properties in amorphous Fe, Co, Ni, Cu, Rh, and Pd, Phys. Rev. B 27 (1983) 2049–2054.
and crystalline FeZr2 examined via the density of states, Phys. Rev. B 70 (2004) [39] W.M. Yang, H.S. Liu, C.C. Dun, J.W. Li, Y.C. Zhao, B.L. Shen, Nearly free electron model
(024209-1–024209-6). to glass-forming ability of multi-component metallic glasses, J. Non-Cryst. Solids 361
[27] O.N. Senkov, D.B. Miracle, Effect of the atomic size distribution on glass forming abil- (2013) 82–85.
ity of amorphous metallic alloys, Mater. Res. Bull. 36 (2001) 2183–2189. [40] Z.P. Lu, C.T. Liu, A new approach to understanding and measuring glass formation in
[28] Y. Li, Q. Guo, J.A. Kalb, C.V. Thompson, Matching glass-forming ability with the den- bulk amorphous materials, Intermetallics 12 (2004) 1035–1043.
sity of the amorphous phase, Science 322 (2008) 1816–1819.
[29] Z.J. Yan, J.F. Li, S.R. He, Y.H. Zhou, Evaluation of the optimum solute concentration for
good glass forming ability in multicomponent metallic glasses, Mater. Res. Bull. 38
(2003) 681–689.