Prior to the corrosion studies, the microstructure of the ribbon pieces prepared from the specimens by single-roller melt-spinning has been checked in detail. Fig. 1 shows the XRD patterns of the as-spun Al_94-xNixGd₆ (x = 6 and 10, in at.%) samples with R_c of 29.3 and 25.6 m/s. For the ribbons spun with R_c of 29.3 m/s, a typical broad diffraction peak can be observed in the XRD patterns, which suggests that fully amorphous structure is formed under this circumferential speed. In the XRD patterns of samples with R_c of 25.6 m/s, several crystalline peaks can be found in the ribbon with the composition of Ni, c_Ni equal to 6 at.%, which are identified as α -Al phase, but are absent in the ribbon with c_Ni of 10 at.%. Apparently, those results confirm the argument that increasing c_Ni is helpful to improve the GFA of Al-based alloys ^[9] and the high cooling rate is beneficial to the increase in the amount of amorphous matrix. In addition, a shoulder diffraction peak is observed at about 2θ = 44° , which is often ascribed to the combined effects of strong interaction of Al-Ni bond ^{[1] and [8]} and pre-existed α -Al nuclei ^[26] in the glassy matrix.

Fig. 1

	Figure Option ViewDownloadNew Window
	Fig. 1 XRD patterns of the as-spun Al94-xNixGd6 (x = 6 and 10) samples with Rc of: (a) 29.3 m/s, (b) 25.6 m/s.

The TEM images and the corresponding selected area electron diffraction (SAED) patterns of the as-spun amorphous samples with R_c of 29.3 m/s are shown in Fig. 2. For the sample with c_Ni of 6 and 10 at.% just after electrochemical thinning, a layer of black precipitates is formed on the hole edge, and the image of the sample with c_Ni of 10 at% can be found in our earlier paper ^[8]. The compositions of the hole edge and internal matrix are also shown in the Fig. 2. For the sample with c_Ni of 10 at.%, the composition of the edge crystallites includes: c_Al = 67.7 at.%, c_Ni = 11.2 at.%, c_Gd = 5.5 at.% and c_O = 15.6 at.%, and the composition of the internal matrix is c_Al = 77.5 at.%, c_Ni = 11.9 at.%, c_Gd = 7.9 at.% and c_O = 2.7 at.%. The high ratios of c_Ni/c_Al and c_Gd/c_Al as well as high c_O on the edge indicate that the crystalline phases are composed of intermetallics and oxides. The patterns in the corresponding SAED of Fig. 2(b) include Al₄Ni₃ and NiO (the others are not given in this paper). The types and percentages of the precipitates identified from 104 sets of measured SAED patterns are listed in Table 1. The results show that the oxides fraction decreases and the content of intermetallics increases with the addition of nickel, which indicates that increasing nickel is beneficial to improving the inoxidizability for the amorphous alloys and increasing the cohesion of the atoms.

Fig. 2

	Figure Option ViewDownloadNew Window
	Fig. 2 (a) Bright field TEM images of Al94exNixGd6 (x=6) ribbons with Rc of 29.3 m/s; (b) the corresponding selected area electron diffraction (SAED) patterns and the composition informations of the dashed boxes; (c) and (d) bright field TEM images of ribbons (Rc =9.3 m/s) with x of6 and 10, respectively.

Table 1

Table 1

Table 1 Types and percentages of oxides and intermetallic phases identified from the SAED patterns for Al94-xNixGd6 (x = 6 and 10) ribbons with Rc of 29.3 m/s Sample Intermetallics Oxides Alx1Gdy1 Alx2Niy2 Gdx3Niy3 Al2O3 Gd2O3 Ni(1, 2)O x = 6 22.7% 11.4% 2.3% 13.6% 10.0% 29.5% x = 10 31.7% 21.7% 11.6% 25.0% 20.5% 0 x1 = 2, y1 = 1; x1 = 2, y1 = 3; x1 = 3, y1 = 1; x2 = 1, y2 = 1; x2 = 1, y2 = 3; x2 = 3, y2 = 1; x2 = 3, y2 = 5; x2 = 4, y2 = 3; x3 = 1, y3 = 5; x3 = 2, y3 = 15.

Table 1 Types and percentages of oxides and intermetallic phases identified from the SAED patterns for Al94-xNixGd6 (x = 6 and 10) ribbons with Rc of 29.3 m/s

In addition, the crack of the sample with c_Ni = 6 at.% fabricated after electrochemical thinning has a bending behavior ( Fig. 2(c)), which is seldom found near the crack with c_Ni = 10 at.% sample ( Fig. 2(d)), indicating that the low nickel content ribbon has a higher stress difference between the free and wheel side. The cracks are initiated by stress concentrated regions, i.e. starting from the thinned transparent edge of the electropolished hole. Meantime, there exist some round bright zones embedded in the matrix (denoted by the arrows) which are proven to be amorphous structure in our earlier works ^[8], and the amount of the bright zone increases with increasing c_Ni ( Fig. 2(c, d)). In several Al-RE-TM glasses, the contrast in the bright field image suggests the existence of phase separation ^{[10] and [13]}.

Fig. 3 gives the potentiodynamic polarization curves of Al_94-xNixGd₆ (x = 6 and 10) samples in 7 wt% HNO₃ methanol solution with R_c of 29.3 and 25.6 m/s. The Al-Ni-Gd glasses exhibit an excellent spontaneous passivation behavior in good conductivity solution upon anodic polarization with a wide passive region. The involved electrochemical parameters such as corrosion potential (E_corr), passive current density (i_pass) and onset pitting potential (E_spit) as well as their differences between samples with R_c of 29.3 and 25.6 m/s are listed in Table 2. In general, the more negative the corrosion potential, the lower the corrosion resistance. As shown in Fig. 3, ribbons with c_Ni of 6 at.% get a more negative corrosion potential than ribbons with c_Ni of 10 at.%, indicating that addition of Ni can defer the anodic reaction and improve the corrosion resistance for the present glassy alloys. As shown in Table 2, the change of the i_pass and E_spit for sample with c_Ni of 10 at.% spun with different cooling rates is small, suggesting that the addition of Ni can decrease the corrosion sensibility of the cooling rates and increase the structure homogeneity.

Fig. 3

	Figure Option ViewDownloadNew Window
	Fig. 3 Potentiometric polarization curves of Al94-xNixGd6 (x = 6 and 10) amorphous ribbons with different Rc in 7 wt% HNO3 methanol solution.

Table 2

Table 2

Table 2 Corrosion potential (Ecorr), passive current density (ipass), onset pitting potential (Espit) and corresponding standard deviation of these parameters of the Al94-xNixGd6 (x = 6 and 10) samples after electrochemistry corroded in 7 wt% HNO3 methanol solution Rc (m/s) cNi (at.%) Ecorr (V) σ Ecorr (V) ( EcorrRS-EcorrSS) (V) log(ipass) (A/cm2) σ ipass (A/cm2) log(ipassRS)-log(ipassSS)(A/cm2) Espit (V) σ Epit (V) EspitRS-EspitSS(V) 29.3 x = 6 -0.125 0.004 0.035 -3.97 0.035 0.25 1.76 0.022 -0.14 x = 10 -0.098 0.032 -0.029 -4.19 0.067 -0.01 1.79 0.035 0.03 25.6 x = 6 -0.160 0.030 - -4.22 0.035 - 1.90 0.033 - x = 10 -0.069 0.009 - -4.18 0.048 - 1.76 0.005 -

Table 2 Corrosion potential (Ecorr), passive current density (ipass), onset pitting potential (Espit) and corresponding standard deviation of these parameters of the Al94-xNixGd6 (x = 6 and 10) samples after electrochemistry corroded in 7 wt% HNO3 methanol solution

Furthermore, SEM was employed to investigate the morphologies of the Al_94-xNixGd₆ (x = 6 and 10) glassy ribbons after polarization tests ( Fig. 4). After electrochemical corrosion some corrosion pitting holes appear on the surfaces of all four ribbons. For the sample with c_Ni of 6 at.%, the pit size of the sample with R_c of 25.6 m/s is much larger than that with R_c of 29.3 m/s. The pit size difference is more drastic than the sample with c_Ni of 10 at.%, indicating that the R_c sensitivity of the samples is weakened with increasing c_Ni. To understand the difference between the heavier corrosion pitting region and the film matrix, the EDS results are shown in Table 3. Here, the matrix film is denoted with F and the particles are denoted with P. The ratio of c_Ni/c_Al is higher in the particle (near the pit region), which indicates that Ni atoms are more noble and have a better corrosion resistance. Meantime, the oxygen content of the film decreases with increasing c_Ni, which is consistent with the improvement of the inoxidizability of sample during the electrochemical thinning ( Fig. 2). However, the c_O of particles in the ribbons with c_Ni of 10 at.% is slightly higher than that with c_Ni of 6 at.%.

Fig. 4

	Figure Option ViewDownloadNew Window
	Fig. 4 SEM micrographs of free surfaces of the Al94-xNixGd6 (x = 6 and 10) glassy ribbons by electrochemical corrosion in 7 wt% HNO3 methanol solution: (a) Rc = 29.3 m/s and x = 6; (b) Rc = 29.3 m/s and x = 10; (c) Rc = 25.6 m/s and x = 6; (d) Rc = 25.6 m/s and x = 10. The film matrix is labeled as F and the particles are labeled as P for the EDS region.

Table 3

Table 3

Table 3 Atomic fraction of elements on the surface of Al94-xNixGd6 (x = 6 and 10) glassy ribbons by EDS after electrochemistry corrosion Rc (m/s) Composition Position cAl (at.%) cNi (at.%) cNi/cAl cGd (at.%) cGd/cAl cO (at.%) 29.3 x = 6 Film 85.1 5.5 0.06 5.8 0.07 2.6 Particle 28.1 2.9 0.10 0.9 0.03 68.9 x = 10 Film 88.8 5.7 0.06 5.5 0.06 - Particle 23.9 3.4 0.14 1.3 0.05 71.8 25.6 x = 6 Film 86.9 5.3 0.06 5.3 0.06 2.5 Particle 34.2 1.9 0.05 0.6 0.01 63.3 x = 10 Film 88.0 6.4 0.07 5.6 0.06 - Particle 26.8 2.6 0.09 0.8 0.02 69.8

Table 3 Atomic fraction of elements on the surface of Al94-xNixGd6 (x = 6 and 10) glassy ribbons by EDS after electrochemistry corrosion

To further investigate the corrosion behavior of Al_94-xNixGd₆ (x = 6 and 10) glassy alloys in the good conductivity solution without electrochemical potential, immersion experiment was performed. Fig. 5 shows the pictures and SEM images of the as-spun and immersed Al_94-xNixGd₆ (x = 6 and 10) glassy alloy specimens. The samples are immersed in 27 wt% HNO₃ methanol solution for 120 h at room temperature, Fig. 5(a, c, e) shows the images of samples with R_c of 29.3 m/s and Fig. 5(b, d, f) displays the ones with R_c of 25.6 m/s. In the optical pictures ( Fig. 5(a, b)), the upper and lower ribbons are as-spun and immersed samples, respectively. In Fig. 5(a), c_Ni of left ribbons (1 and 2) and right ribbons (3 and 4) are 6 and 10 at.%, respectively. The bending degree of immersed ribbon with c_Ni of 6 at.% (ribbon 2) is severer than that with c_Ni of 10 at.% (ribbon 4). In Fig. 5(b), c_Ni of left ribbons (5 and 6) and right ones (7 and 8) are 6 and 10 at.%, respectively. Similarly, immersed ribbon 6 (c_Ni = 6 at.%) has a larger bending degree than ribbon 8 (c_Ni = 10 at.%). The bending behavior of ribbons may be explained by the different contraction degree induced by the internal corrosion stress between two sides of samples immersed in 27 wt% HNO₃ methanol solution.

Fig. 5

Figure Option ViewDownloadNew Window

Fig. 5 Optical pictures and SEM images of immersed Al94-xNixGd6 (x = 6 and 10) glassy ribbons in 27 wt% HNO3 methanol solution for 120 h. The cooling rates for the samples in left and right columns are 29.3 and 25.6 m/s, respectively. In optical picture with Rc of 29.3 m/s, ribbon 1, as-spun sample with x of 6; ribbon 2, immersed sample with x of 6; ribbon 3, as-spun sample with x of 10; and ribbon 4, immersed sample with x of 10. In optical picture with Rc of 25.6 m/s, ribbon 5, as-spun sample with x = 6; ribbon 6, immersed sample with x of 6; ribbon 7, as-spun sample with x of 10; and ribbon 8, immersed sample with x of 10. (c) and (d) SEM images of immersed samples with x of 6, (e) and (f) SEM images of samples with x of 10. Film denotes the film matrix, fragment denotes the heavy corrosion region.

In SEM images, the morphologies of the immersed samples with c_Ni of 6 and 10 at.% are investigated ( Fig. 5(c-f)). With c_Ni of 6 at.%, cracks propagate preferentially along with the line directions A and B in Fig. 5(c, d), respectively. However, with c_Ni of 10 at.%, cracks propagate isotropically, indicating that increasing c_Ni can improve the homogeneity of the present glass alloys. To understand the effect of the heavier local corrosion, the EDS results of the fragment area and film are shown in Table 4, in which the local heavy corrosion region is labeled as “ Fragment” and the matrix film is labeled as “ Film” . Apparently, the oxygen content c_O of fragment area of ribbons is higher than that of the matrix film, which is consistent with the electrochemical test ( Fig. 4 and Table 3). Moreover, the c_O of the fragment area in the ribbon with c_Ni = 10 at.% is higher than that with c_Ni of 6 at.%, which is also consistent with the electrochemical test ( Table 3).

Table 4

Table 4

Table 4 Atomic fraction of elements on the surface of Al94-xNixGd6 (x = 6 and 10) with Rc of 29.3 m/s glassy ribbons by EDS after immersion corrosion Composition Position cAl (at.%) cNi (at.%) cNi/cAl cGd (at.%) cGd/cAl cO(at.%) x = 6 Film 74.5 7.2 0.09 6.9 0.09 11.4 Fragment 59.0 5.8 0.09 5.8 0.09 29.4 x = 10 Film 76.3 9.6 0.12 5.7 0.07 8.4 Fragment 41.1 6.2 0.15 3.8 0.09 48.9

Table 4 Atomic fraction of elements on the surface of Al94-xNixGd6 (x = 6 and 10) with Rc of 29.3 m/s glassy ribbons by EDS after immersion corrosion

After being immersed in the nitric acid methanol solution for 120 h, the XRD patterns of the immersed samples are shown in Fig. 6. In contrast with the patterns of Fig. 1, the α -Al phase is corroded into the solution and all four samples appear in a fully amorphous structure. The broad main peak, generally corresponding to Al clusters, of the immersed ribbons is lower, and the shoulder peak, corresponding to Ni/Gd-rich clusters, is higher compared with the as-spun ribbons, implying that the aluminum clusters dissolve more quickly than Ni/Gd-rich clusters during the immersion test. The selective dissolution of elements/clusters can introduce the local stress to the sample and has a relationship with the cracking and bending in Fig. 5. Meanwhile, with increasing c_Ni, the shoulder peak increases after immersion test, showing a more typical phase separation phenomenon.

Fig. 6

	Figure Option ViewDownloadNew Window
	Fig. 6 XRD patterns of the immersed Al94-xNixGd6 (x = 6 and 10) sample with: (a) Rc = 29.3 m/s, (b) 25.6 m/s.

Kelton et al.^[10] and Poon et al.^[27] have proposed the backbone argument that the Ni/Gd-rich backbone clusters as well as Al inclusion clusters exist in the Al-Ni-Gd glassy alloys. They suggested that the bulk metallic glasses (BMGs) can be divided into MSL (majority atom-small atom-large atom) class and LS (large atom-small atom) class^[27]. The atomic sizes of Al, Ni and Gd in the present Al-Ni-Gd alloy are 0.143, 0.124 and 0.180 nm, considered as M (majority), S (small) and L (large) size, respectively^[28]. The interaction of the stress fields can lead to the formation of stable short-ranged ordered clusters of small (Ni) and large (Gd) atoms. The structure-reinforced network is called the backbone clusters. The small inclusion clusters are separated by the network structure^[29]. It should be mentioned that the basic structural units of backbone clusters are solute centered quasi-equivalent clusters surrounded by Al atoms^[30]. The backbone clusters are more stable and stronger than Al inclusion clusters^[31]. Accordingly, it is expected that a higher c_Ni/c_Al present in the severe corrosion zone after electrochemical thinning, polarization and immersion tests ( Fig. 2 and Table 3).

The percolation theory has been adopted to describe the mechanical and structural properties of several metallic glasses^{[31] and [32]}. It is inferred that the mechanical properties may be changed discontinuously when the microstructural percolation is achieved^[32]. Introducing particular solute species may promote the percolating backbone of short range orders in some glassy systems^[31]. It is understood that, with R_c of 25.6 m/s, the Al precipitates appear in the sample with c_Ni of 6 at.%, but are absent in the sample with c_Ni of 10 at.% ( Fig. 1(b)). The sensitivity to the composition of the alloys could be dramatic due to the phase transition nature of a percolation, which has possibly mixed the discontinuous and continuous characters ^[33]. Increasing the cooling rate is another method to promote the percolation of the backbone clusters in the glass alloys ^[34]. It is expected that, with c_Ni of 6 at.%, the Al precipitates appear in the sample with R_c of 25.6 m/s, but absent in the sample with R_c of 29.3 m/s ( Fig. 1).

When a glassy alloy is electrochemically decomposed in the electrolyte, the noble metal atoms diffuse through the alloy surface to form new nuclei and grow to crystals^[35]. During dealloying Al-Cu-Mg amorphous precursor alloys the selective dissolution of elements occurs and nanoporous crystal Cu is synthesized^[36]. Our earlier ab initio molecular dynamic simulation on Al_94-xNixGd₆ (x = 6-12) amorphous alloys indicated that the Al-Ni and Al-Gd bonds are strong bonds ^[8]. Hence, it is expected that crystalline intermetallics Al₄Ni₃ and Al₂Gd₃ have remained in the sample after electrochemical thinning ( Fig. 2 and Table 1). The solvent aluminum is active in the Al-Ni-Gd glassy alloys in the annealing process and Al clusters in the melt have a lower stability than the Ni/Gd-rich clusters in the cooling process ^[4]. Therefore, it is expected that Al phase precipitates from the melt with c_Ni of 6 at.% at a low cooling rate of 25.6 m/s ( Fig. 1(b)). According to the resemblance between the glass state and the liquid state ^[37] and the fact that the aluminum is easier to be dissolved into the solution ^[8], it is understood that the α -Al peaks in XRD pattern of as-spun sample with c_Ni of 6 at.% and R_c of 25.6 m/s disappear after immersion test ( Fig. 1 and Fig. 6).

The phase separation prior to crystallization, forming Al-rich and solute-rich regions, has been found in Al₈₈Gd₆La₂Ni₄^[10], Al₈₅Ni₅Y₆Fe₂Co₂^[13], Al-Fe-Y^[3] and Al₈₉Ni₆La₅^[38] amorphous alloys. 3D-atomic probing technology (APT) investigation demonstrates the phase separation occurring in the Al-Fe-Y and Al-Ni-La metallic glasses, which have negative mixing enthalpies among their three binary subsystems^{[3] and [38]}. As suggested first by Hume-Rothery and Anderson^[39], in a region of composition beneficial to the development of short- and medium-range order involving five-fold symmetry, the G-curve dips, which makes the occurrence of eutectic be more likely. In metallic glasses, the eutectic composition is usually considered as the ideal composition to form the glass. According to Hume-Rothery and Anderson's hypothesis^[39], a narrow miscibility gap may open near the composition corresponding to maximal atomic packing efficiency due to rapid dip in the Gibbs free energy curve^[40]. The hypothetical form of the free energy curve and a metastable miscibility gap, including a spinodal region of amorphous phase separation, in Al-RE binary amorphous alloys can be presented by solid lines in Fig. 7^[41]. Here, the second derivative of the free energy of mixing defines the spinodal points (C and D), which locate between the binodal points (A and B, which is defined as the first derivative of the Gibbs free energy). Above the binodal line, two liquids are miscible and the system is stable(s); below this line there is a metastable region (m), in which the system is stable to small fluctuations but is unstable to large fluctuations; below the spinodal line, the system is unstable^[42]. In our earlier work^[8], the electropolished as-spun Al₈₄Ni₁₀Gd₆ amorphous ribbon formed white regions, which were still amorphous structure in the matrix; the corresponding SAED patterns showed an asymmetric halo of the amorphous structure, which may correspond to two overlapped diffraction peaks. The phase separation phenomenon is also confirmed by the XRD patterns of sample with c_Ni = 10 at.% after the immersion corrosion ( Fig. 6).

Increasing the content of nickel can widen the composition range of phase separation, which is shown schematically by the dashed line in Fig. 7. Considering the similar RE content of both alloys as well as Hume-Rothery and Anderson's hypothesis^[39], we suppose that the binodal curve for x equal to 10 is overlapped by that with x of 6. Here, with increasing c_Ni, the backbone clusters are strengthened, leading to a heavier dip in the Gibbs free energy ^{[41] and [43]}. Accordingly, the spinodal region is wider and it is easier to perform the spinodal decomposition with a higher content of nickel. Hence, it is expected that the sample with c_Ni of 10 at.% undergoes a heavier phase separation after corrosion than that with c_Ni = 6 at.% ( Fig. 2 and Fig. 6). The composition range of phase separation locates always near/inside the optimized glass formation ranges, i.e. the more easily the phase separation occurs in a glass, the more difficultly the nucleation and growth of crystallites occur ^[41]. Hence, it is understood that the sample with c_Ni of 10 at.% has a heavier phase separation and a higher crystallization temperature than the counterpart with c_Ni of 6 at.% ( Fig. 2 and Fig. 6 and Table 3 in Ref. ^[8]).

Fig.7

Figure Option '>ViewDownloadNew Window

'>

Fig.7 Fig7 Schematic presentation of a hypothetical form of the free energy curve and a metastable miscibility gap in Al94-xNixGd6 (x = 6 and 10) amorphous alloys. A and B define binodal points, and the composition between C (C') and D (D') belong to a spinodal region. Above the binodal line, the liquid is stable. Below this line, there is a metastable region (m). Within the metastable region the system is stable to small fluctuations but is unstable to large fluctuations. Below the spinodal line, the system is unstable and easy to turn from one phase to two-phase system.

Oxidation is an intrinsic phenomenon associated with the electropolishing method, especially for alloy systems containing highly reactive elements^{[8] and [44]}. During the electropolishing, a range of clusters, from embryo clusters to clusters approaching the critical condition, and possibly oxidized clusters are present in the alloy's surface layer^[45]. During the HCl/HF solution etching process, the relatively inert Cu atoms in Cu-Hf-Al glass are left and form Cu₂O^[16]. It is known that, the closer to the hole the zone is, the heavier etching degree it endures, and the larger stress it has. In our earlier work, the stress induced by autoclave and compression treatment can cause the crystallization of Al-Ni-La glass^[34]. Meanwhile, the composition near the hole is far away from the ideal composition with a good GFA^[46]. Hence, the crystallization and oxidation can emerge more easily on the edge, which has sustained heavier selective dissolution of elements, and it is understood that the ratios in c_Ni/c_Al and c_Gd/c_Al as well as c_O in the crystallites are higher than those in matrix ( Fig. 2). Generally, Ni is a noble element and can increase the inoxidizability of the alloy. It is expected that with increasing c_Ni, the c_O of the hole edge decreases, and the fractions and types of oxides also decrease ( Fig. 2 and Table 1).

Generally, the properties of a material are dependent on its microstructure. The physical and chemical properties of samples in the percolated state are apparently different from those in the unpercolated state^{[31] and [34]}. As we have mentioned, increasing the cooling rate is a method to promote the percolation of the backbone clusters in the glass alloys. Hence, the corrosion potential E_corr, passive current density i_pass and onset pitting potential E_spit of sample with c_Ni of 6 at.% under R_c of 29.3 m/s differ from those under R_c of 25.6 m/s; moreover, the difference degrees with c_Ni of 6 at.% are larger than those with c_Ni of 10 at.% ( Fig. 3 and Table 2).

Moreover, it is known that there is a cooling rate difference between the wheel and air side of as-spun ribbons, which can bring structure difference between both sides^{[47] and [48]}. As mentioned above, the R_c sensibility of the sample with c_Ni of 6 at.% is higher than the sample with c_Ni of 10 at.% ( Fig. 1 and Fig. 3). In other words, the percolation of the backbone cluster in the air side (low cooling rate side) of sample with c_Ni of 6 at.% and R_c of 25.6 m/s is more easily destroyed (the corresponding crystalline precipitates with R_c of 29.3 m/s can be ignored in the XRD analysis due to the small amount), and structure difference between two sides is larger than that with c_Ni of 10 at.%. The wheel side usually has a higher cooling rate than the free side, leading to a higher amorphous fraction and free volume in the wheel side for the as-spun ribbons. When the ribbons sustain heavy corrosion, the contraction of the wheel side is larger than the free side and the effect of the contraction can surpass the original stress, resulting in a tensile stress on the wheel side and a compressive stress on the free side. After electrochemical thinning, the corrosion rate difference between the two sides with c_Ni of 6 at.% is much higher, inducing a larger compressive stress on the free side and a higher tensile stress on the wheel side compared with the sample with c_Ni of 10 at.%. Therefore, it is understood that the bending degree of the hole edge of electrochemical thinned sample with c_Ni of 6 at.% as well as the immersed sample is higher than with c_Ni of 10 at.% ( Fig. 2 and Fig. 5).

As mentioned above, the minor Ni addition can not only enhance the structural and chemical homogeneity by promoting glass formability, but also enhance the percolation of backbone structures in the present glass. Meanwhile, Ni is more noble than Al and Gd^[8], and it is expected that the E_corr of the sample with c_Ni of 10 at.% is higher than that with c_Ni of 6 at.% ( Fig. 3 and Table 2) and that the oxygen content of the film region decreases with the addition of nickel ( Table 3 and Table 4).

In addition, the shear localization and strain rate independence appear to be robust features of the samples, in which a backbone of atoms percolates through the material^[49]. In the cracking position of sample with c_Ni of 10 at.%, a higher threshold stress should be surpassed to trigger the localized corrosion; in other words, a higher internal stress exists in the fragment area of sample with c_Ni of 10 at.% than that with c_Ni of 6 at.%. It should be noted that the internal stress near the fragment area is completely different from the tensile stress on the wheel side owing to a quicker cooling rate than that of the air side. The latter is global stress and the former is much localized. In addition, the phase separation can embrittle the sample ^[50], and the phase separation of sample with c_Ni of 10 at.% is heavier than that with c_Ni of 6 at.%. Accordingly, it is understood that the crack in the immersed sample with c_Ni of 10 at.% is more developed than that with c_Ni of 6 at.% ( Fig. 5). The accumulated highly localized stress is beneficial to the crystallization and coupled with oxidation ^[51]. It is explained that a higher oxygen content presents in the particles near the pits of the electrochemical polarized and in the fragment area of the immersed samples with c_Ni of 10 at.% compared with that with c_Ni of 6 at.% ( Table 3 and Table 4). We have to mention that the discrepancy of c_O between the TEM and SEM samples is mainly due to the checking area of TEM investigation (100 nm) is much smaller than that in SEM test (10 μ m), resulting in a larger error of the TEM sample.