A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses

Corrosion Science xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Corrosion Science journal homepage: www.elsevier.com/locate/corsci

A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses Miriam Eisenbart a, Ulrich E. Klotz a, Ralf Busch b, Isabella Gallino b,⇑ a b

Fem Research Institute for Precious Metals & Metals Chemistry, Katharinenstrasse 17, 73525 Schwäbisch Gmünd, Germany Chair for Metallic Materials, Saarland University, Campus C6.3, 66123 Saarbrücken, Germany

a r t i c l e

i n f o

Article history: Received 21 September 2013 Accepted 20 April 2014 Available online xxxx Keywords: A. Alloy A. Glass B. XPS B. TEM C. Oxidation C. Amorphous structures

a b s t r a c t The tarnishing of the Au–Ag–(Pd)–Cu–Si bulk metallic glass system is studied based on the Yellowness Index detection and on microstructural investigations during various and prolonged low-temperature exposures. Polished surfaces tarnish faster than as-cast surfaces due to the removal of the native SiO2 film. The rate decreases with decreasing nominal Cu/Si and Si/Au ratios, with removing Pd, and with microalloying of Al. The tarnishing mechanism is controlled by the internal oxidation of the glassy matrix forming amorphous SiO2 dendrites. This triggers the partitioning of metallic elements that enhances the rate of out-diffusion and surface oxidation of Cu. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Duwez and co-workers [1] reported in 1960 that the Au–Si system can form amorphous alloys at around the eutectic composition of Au–18 at.% Si by rapid quenching. Even if the critical thickness of the amorphous sample was only 50 lm, this was the first time that crystallization of a metallic melt was avoided altogether during solidification. Since then, numerous metallic glass forming alloy systems have been explored. It was only in 2005 that the Au–Si system has been revisited by Schroers et al. who drastically enhanced the glass-forming ability by adding Pd, Ag and Cu [2]. The compositions reported by Schroers et al. have enough gold content to be considered as 18 carat gold alloys (i.e. at least 75 wt.% Au, corresponding to approximately 50 at.% Au) and are massive enough to be considered as bulk metallic glasses (hereafter called BMG), whose prerequisite is a critical thickness of at least 1 mm. Among those, the best Au-based glass-former is the Au49Ag5.5Pd2.3Cu26.9Si16.3 composition with a critical thickness of 5 mm and a glass transition temperature, Tg, of around 128 °C [2]. In 2009, Zhang and co-workers investigated the Pd-free Au–Ag– Cu–Si glass system and observed a sharp decrease in the glass transition with increasing gold content. For example the lowest reported Tg is 66 °C for the composition Au70Cu5.5Ag7.5Si17 BMG alloy [3]. Other researchers modified the ternary Au–Cu–Si systems by alloying with other elements such as Sn [4], Ti or Y [5], in order ⇑ Corresponding author.

to improve certain properties, like the glass forming ability, or simply to reduce the costs by reducing the noble metal content. The Au49Ag5.5Pd2.3Cu26.9Si16.3 BMG possesses a premium white gold colour [6,7], and a Vickers hardness value of 360 HV1 [2], that is approximately 1.5–2 times higher than the hardness of cold worked or age hardened conventional gold alloys (e.g. Au54.9Ag16.7Cu28.4 at.%) [8]. This value is much higher than the minimum hardness value required for jewellery items, which is set to be as least 100 HV1 for reasonable wear resistance and service performance [8]. Moreover, the low melting temperature of these BMG of approximately 370 °C, minimal solidification shrinkage (i.e. less than 0.5%) [9] and a good processing ability [2] are highly desired properties for jewellery and dentistry applications. However, any significant change in the colour of a jewellery or dentistry alloy due to exposure to atmosphere or contact with skin or saliva is obviously undesirable. The body temperature of 37 °C is 90° lower than the Tg for the Au49Ag5.5Pd2.3Cu26.9Si16.3. At this temperature, which is approximately 0.77Tg (in K), the glassy structure is stable against crystallization for hundreds of years according to the time-temperature-transformation diagram of Ref. [10]. During isothermal aging at 37 °C, structural relaxation processes (or a relaxation) will occur very slowly, taking decades to be completed due to its relative sluggish bulk diffusivity. At 0.7–0.8Tg (in K), most BMGs do not show fast oxidation or tarnishing effects. For example, Zr-based and Ni-based BMGs have extremely low parabolic oxidation rates at temperatures, which are 60–90° below Tg [11–13].

E-mail address: [email protected] (I. Gallino). http://dx.doi.org/10.1016/j.corsci.2014.04.024 0010-938X/Ó 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

2

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

It is intuitively reasonable to expect that an 18 carat gold alloy would not be distinctly affected by undesirable tarnishing. Nevertheless, after a few hours of exposure to a non-protective atmosphere, the Au49Ag5.5Pd2.3Cu26.9Si16.3 glassy alloy assumes dull yellowish gleams. The colour gradually deepens to become orange and later shows locally orange-brown corrosion products forming on the surface. The tarnishing is even more pronounced, as soon as the alloy is worn as jewellery and comes in contact with sweat or saliva. Therefore, recent studies have been focused on the wet corrosion behaviour of gold-based BMGs in artificial sweat or artificial saliva baths [6,7,14–16]. Some of these investigations have been performed on melt spun ribbons, which were consumed by the corrosion process [14,15]. In the present article a systematic colourimetric study of the tarnishing process is applied to the Au–Ag–(Pd) –Cu–Si BMG system. The study is based on detection of the Yellowness Index and the determination of the colour change. Various BMG compositions were cast into plates of 2 mm thickness and polished specimens were stored under different atmospheric conditions in order to detect the rate of the surface tarnishing as a function of the corresponding environmental condition. For the first time the colour change rate was quantified for any BMG system during exposure to air. A standard sulphide immersion test was also performed. Microstructural examinations of the tarnished surfaces were performed by STEM, TEM, FIB and XPS, revealing important information towards the understanding of the tarnishing mechanism.

2. Experimental methods Bulk metallic glass specimens of compositions listed in Table 1 were produced starting from the pure elements by using a Topcast TCE10 centrifugal casting device. The pure elements, with purity greater than 99.995%, were inductively heated in a quartz crucible coated with zirconia and the melt was cast into a split copper mould with plate geometry. The dimension of the mould cavity was 2  12  47 mm. The copper mould was not water cooled and the temperature of the mould in the vicinity of the cavity was monitored during the casting procedure. The Cu mould was massive and only a slight increase of the temperature of 3–5 °C has been detected during each casting. Each sample was mechanically ground with SiC papers down to 1200 grits and polished with suspensions of diamond particles with size from 6, down to 3, and to 1 lm. This procedure is hereafter called the standard polishing procedure. Prior to all experiments, the specimens were shown to be x-ray amorphous by X-ray diffraction (XRD). The onset temperature of the glass transition, Tg, was detected using small specimens representing the tips of the plates by differential scanning calorimetry (DSC) by scanning from 0 °C at a heating rate of 20 °C/min with a Perkin Elmer Hyper DSC 8500. Standard polished specimens of approximately 1 cm2 surface area of different compositions were stored together under controlled and stable storage conditions of temperature and air atmosphere. The controlled room temperature (RT) condition was 23 °C ± 2 °C with 50% ± 5%

humidity. The storage of specimens at 37 °C ± 1 °C and 75 °C ± 1 °C was performed in furnaces at laboratory atmosphere (not air-conditioned). Additionally, for the Au49Ag5.5Pd2.3Cu26.9Si16.3 composition, one specimen was stored at 75 °C ± 1 °C under constant Ar flux, one in vacuum at RT, one in a desiccator at RT filled with laboratory air, and one specimen in a freezer at 18 °C ± 1 °C. The standard sulphide immersion testing was performed with specimens with surface areas of approximately 1 cm2 and followed the standard test procedure: DIN EN ISO 1562:2004. Prior to the immersion test, the specimens were embedded in an epoxy resin material and standard polished as described above. After an initial colour detection, the samples were installed together in a testing device that automatically immersed the surface of the specimens in the solution every 60 s for a period of 13 ± 2 s of submersion time. The testing solution was a sodium sulphide aqueous solution of 0.1 mol/l of Na2S. The solution was prepared by dissolving 22.3 g of sodium sulphide hydrate, Na2S  H2O (about 35 wt.% Na2S) in twice deionized water until a final volume of 1000 ± 3 ml was achieved. During the first 72 h of testing, the test was interrupted every 24 h, the solution was renewed and the samples were removed and cleaned with ethanol and subjected to the colourimetric detection. The time out of the immersion testing apparatus was minimized and the samples were promptly re-installed in the automatic device for further immersion testing. After 72 h the immersion was automatically cycled every 60 s as described above for an additional 96 h without exchanging the solution, and the colourimetric detection was performed only at the end, after an overall testing time of 168 h (7 days). The colourimetric analyses were performed on 1 cm2 surface areas using 10  10  2 mm specimens. The initial colour was detected shortly after the top surface was standard polished and, afterwards, the colour change was detected regularly during the storage in controlled and stable conditions and during the sulphide immersion testing. The change in colour was determined according to the standard DIN 6174. Applying the standard DIN 5033-3, the colour is represented by three coordinates: L* for luminescence, a* for colours in the range from red to green, and b* for colours in the range from yellow to blue. Using these colour coordinates, the colour change DE was calculated as:

h i1=2  2 2 DE ¼ ðDL Þ þ ðDa Þ2 þ ðDb Þ

ð1Þ

The Yellowness Index (YI) was calculated according to the standard ASTM D1925. This is a numerical grading system to categorize the whiteness of white gold alloys [17,18]. The YI index is used to describe the aberration from a perfect white towards a yellowish colour, and quantifies the quality of commercial white gold alloys, as white gold tends to have a yellow component in its colour. A YI value between 0 and 19 is attributed to a premium white colour which is most desirable, while a YI between 19 and 24.5 describes a standard white. Off-white colours possess a YI value of 24.5–32 and colours with a YI exceeding 32 are called non-white. The upper

Table 1 Studied bulk metallic glass compositions with corresponding glass transition temperatures, Tg, and Cu/Si and Au/Si nominal atomic ratio and Yellowness Index values, YI. The Tg are the onset temperatures of the calorimetric glass transition signals during up-scans with 20 °C/min. The YI characterizes the whiteness of white gold alloys and are detected in this work soon after grinding and polishing the surfaces of freshly cast materials. A value of Yellowness Index below 19 is considered a premium white gold colour. BMG composition in at.%

Tg(onset) in (°C)

Atomic ratio Cu/Si

Atomic ratio Au/Si

Yellowness Index after polishing

Au49Ag5.5Pd2.3Cu26.9Si16.3 Au49Ag5.5Pd2.3Cu25.9Si16.3Al1 Au50Ag7.5Cu25.5Si17 Au60Ag7.5Cu15.5Si17 Au60Ag5.5Pd2Cu15.5Si17 Au65Ag7.5Cu10.5Si17

128 120 104 86 95 69

1.65 1.59 1.50 0.91 0.91 0.62

3.01 3.01 2.94 3.53 3.53 3.82

17.80 15.01 14.74 16.09 17.28 18.01

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

limit for the colour change detection is approximately DE = 50, which corresponds to an off-scale value of Yellowness Index (YI > 90). Table 1 lists each studied alloy composition, its corresponding nominal Cu/Si and Au/Si atomic ratios, and its Yellowness Index value in as-polished condition. The measurement of the YI was performed with a Konika Minolta CM5 spectrophotometer and had an error of ±1, which was estimated from reproducibility tests performed on the Au49Ag5.5Pd2.3Cu26.9Si16.3 composition for the exposure to air at 37 °C and 75 °C. X-ray diffraction patterns were measured on surfaces and cross sections with Cu Ka radiation on a Bruker D8 Discover high resolution X-ray diffractometer, equipped with 1 mm collimator, parallel beam optics and a Bruker Vantec-500 detector system. Microstructural analyses were performed in cross section. Thus, thin lamellae with rectangular shape were cut normal to the tarnished surface with a focused ion beam (FIB) in a Zeiss Auriga 60 and analysed in the STEM mode with 30 kV. A semi quantified analysis of the chemical composition was carried out with an Oxford EDX system and an operating voltage of 10–20 kV. X-ray photoemission spectra (XPS) were collected by using a Thermo Fisher Scientific 1 XPS system equipped with a hemispherical Alpha 110 Analyser. The analysis was carried out using Al Ka excitation with a resolution of 1 eV. The energy of the Ar+ ion beam has been set to 2.0 keV and the sample current density was 0.2 A/m2. The sputtering rate was estimated to be approximately 4.56 nm/min in Si-rich regions and 18 nm/min in Au-rich region. 3. Results 3.1. Colourimetric analyses Fig. 1 shows the Yellowness Index (YI) of amorphous Au49Ag5.5 Pd2.3Cu26.9Si16.3 measured for various temperatures as a function of exposure time to a laboratory air atmosphere. The freshly polished surface has an initial premium white gold YI average value of 17.8. This value increases with time of exposure, reaching offwhite colours in relatively short times. At ambient temperature it takes 2 months to reach off-white colours, whereas at body temperature only few days are necessary. At the end of the exposure the specimens were still X-ray amorphous. Fig. 2 shows how the environment affects the tarnishing rate of the glassy Au49Ag5.5Pd2.3Cu26.9Si16.3. The tarnishing rate decreases as the atmosphere gets less aggressive. For example, at room temperature, three storage conditions were applied (open symbols): vacuum conditions (with pressure below 105 hPa), dry air

Fig. 1. The Yellowness Index plot for the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 versus time of exposure to air at various temperatures.

3

Fig. 2. The Yellowness Index plot for the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 versus time of exposure to different atmospheres at various temperatures.

conditions (in a desiccator filled with laboratory air), and air exposure conditions (exposed simply to the air-conditioned room atmosphere). Two samples have been air treated in a furnace located in a not-air conditioned laboratory at 37 °C (open crossed symbols). During one of these isothermal tests (crossed circles labelled ‘humid’) the furnace hosted at the same time a sweat analysis immersion testing apparatus and therefore the atmosphere in the furnace was more ‘humid’ and corrosive than the specimen that was air-treated alone. Two heat treatments at 75 °C, shown in Fig. 2 with closed symbols, were performed one in air furnace placed in a not-air conditioned laboratory and the other under a constant argon flux of approximately 50 ml/min. In Fig. 3a the colour change is plotted against the exposure time to air at 75 °C to compare different compositions. The Au60Ag7.5 Cu15.5Si17 and the Au65Ag7.5Cu10.5Si17 are not included in this analysis because their Tg is too close to or lower than 75 °C (86 °C and 69 °C, respectively [3]). Each individual symbol in Figs. 1–3 corresponds to the average value of three individual colour detections. After circa one month of exposure the colour change for the Au49Ag5.5Pd2.3Cu26.9Si16.3 (solid square) reaches the detection limit (where YI > 90) and after that the solid square symbols appear to have a slower rate of colour change which is nonrepresentative. The cross section of each specimen was X-rayed after 321 days of air treatment at 75 °C. The XRD patterns are shown in Fig. 3b. The Au49Ag5.5Pd2.3Cu26.9Si16.3 and Au49Ag5.5Pd2.3Cu25.9Si16.3Al1 are still amorphous whereas the Au60Ag5.5Pd2Cu15.5Si17 and the Au50Ag7.5Cu25.5Si17 have crystallized during the treatment. The tarnishing rate of Au49Ag5.5Pd2.3Cu26.9Si16.3 (solid squares in Fig. 3a) is drastically slowed down by either reducing the Cu/Si ratio or by microalloying some selected elements like Al. For example for the BMG with composition Au60Ag5.5Pd2Cu15.5Si17 (solid circles in Fig. 3a), after 3 weeks of air exposure the colour change is only 50% of that for the Au49Ag5.5Pd2.3Cu26.9Si16.3. The colour change rate of Au49Ag5.5Pd2.3Cu25.9Si16.3Al1 (star symbols) is remarkably low. It is reduced by at least 75% over the same time of exposure with respect to the Au49Ag5.5Pd2.3Cu26.9Si16.3 (square symbols). The matrix of the Au49Ag5.5Pd2.3Cu25.9Si16.3Al1 has retained its amorphous structure similarly to the mother-alloy (see Fig. 3b). The Pd-free Au50Ag7.5Cu25.5Si17 (open triangles) has Cu/Si and Au/Si ratio very similar to those of the Au49Ag5.5Pd2.3 Cu26.9Si16.3 (see Table 1). Nevertheless the removal of Pd seems to improve the tarnishing resistance, however the Au50Ag7.5Cu25.5 Si17 has crystallized during the 1 year heat treatment at 75 °C (see Fig. 3b). A similar trend was observed at room temperature in the sulphide-immersion investigation. Fig. 4 shows the plot of

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

4

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

3.2. Microstructural analyses

Fig. 3. (a) The colour change for the Au49Ag5.5Pd2.3Cu26.9Si16.3, the Au49Ag5.5Pd2.3 Cu25.9Si16.3Al1, the Au60Ag5.5Cu15.5Si17 Pd2 and Au50Ag7.5Cu25.5Si17 BMG compositions detected during a 75 °C heat treatment in air. (b) The XRD patterns corresponding to each specimen shown in (a) at the end of the exposure (327 days).

Fig. 5 shows a photograph of the surface of an amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen that was standard polished leaving one macroscopic surface cast porosity on one side. This surface defect was caused by an air pocket trapped in the mould during form filling. The image was recorded after only 4 days of air exposure at 75 °C, and the polished surface appears already heavily tarnished whereas the as-cast surface at the pore appears uncompromised and maintains a white gold colour. After 131 days of storage at 75 °C, two FIB cross-section lamellae representing the as-cast surface at the pore (Fig. 6a) and the as-polished surface (Fig. 6b), respectively, were cut normal to the surface. The specimen was furthermore sectioned and its cross section was found (after 131 days at 75 °C) to be amorphous by XRD detection. In STEM mode, the as-polished surface (shown in Fig. 6b) revealed to have formed a continuous surface scale of approximately 20 nm thickness. This scale has a dark contrast in Fig. 6b and EDX analyses found high concentrations of Cu and O. Dendrite-like features were found to originate in the sub-scale matrix and grow inwards up to a depth of approximately 120 nm. EDX chemical analyses showed that the dendrites are rich in Si and O and depleted in the other constituents. This phenomenon was exclusively found to occur on as-polished surfaces as in Fig. 6b, and never on cast surfaces that solidified at the air pockets, as in Fig. 6a. The original as-cast surface of the sample shown in Fig. 5 appears unaffected from any surface attack. The STEM microphotograph of Fig. 6a presents a homogeneous glassy matrix that (in contrast to the as-polished surface of Fig. 6b) lacks the dark contrast surface scale and, even more importantly, the sub-surface dendrite-like features are absent. Instead, only a very thin continuous surface layer is present. This layer is less than 5 nm thick and has a blunt grey contrast in the microphotograph of Fig. 6a. However, this layer proved by EDX-mapping to be enriched in Si and O and seems to be passivating the matrix underneath, since Cu-rich corrosion products and dendrite-like internal oxidation products were not found anywhere in the analyses. Fig. 7a and b show the copper, oxygen and silicon distribution measured by EDX-mapping of the sample showed in Fig. 6 on an as-cast location and an as-polished location, respectively. According to the analysis of Figs. 6a and 7a, the thin scale (with grey contrast) detected on the as-cast surface corresponds to a native SiO2 film. In Figs. 6b and 7b the dark-contrast scale is rich in Cu and O and the dendrite-like

Fig. 4. The colour change plot during a week-long sulphide standard test as a function of immersion time for each studied Au-based BMG composition.

the colour change during the seven days sulphide immersion test for the studied gold-based BMGs. The colour change was detected every 24 h. The specimen were X-rayed at the end of the test and shown to be X-ray amorphous. The colour change rate decreases as the nominal Cu/Si ratio decreases.

Fig. 5. Photograph of an Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen after 4 days of air annealing at 75 °C. Only the as-polished surface appears tarnished, whereas the surface of an as-cast pore appears untarnished.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

5

327 days. On the surface Cu2O products were detected everywhere. At certain locations, cone-like corrosion products made of Cu2O outgrow with an average height of approximately 1 lm. In Fig. 10 two of them are visible. One of them was sectioned with the FIB to expose the cross section underneath. The SiO2 dendrite-like features appear to have nucleated and to have grown faster underneath the cone-like Cu2O outgrows, than underneath the adjacent thinner Cu2O surface scale. In Fig. 10 the internal oxidation of Si reaches a depth of approximately 1.15 lm, under conelike Cu2O surface products, whereas under ‘cone-like free’ surface locations the oxidation affected depth is much more restricted and ranges between 0.10 and 0.34 lm. 4. Discussion 4.1. The colourimetric study

Fig. 6. STEM microphotographs of FIB-lamellae cut normally to the surface of the Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen of Fig. 5 after 131 days of air annealing at 75 °C. (a) Cross section of the as-cast surface (at the pore). (b) Cross-section of the aspolished surface.

features are rich in O and Si in which Cu and Au are depleted (the Au x-ray map is not shown here). Fig. 8 shows an XPS concentration profile, detected by sputtering the surface of an amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen after 23 days of air exposure at 75 °C. The surface was standard polished prior to exposure. The first product sputtered away from the surface corresponds to a few-nm thick Cu2O phase. More internally, when a dendrite-like feature is sputtered away, SiO2 is detected. At the interface between the SiO2 dendrite and the metallic matrix a slight copper enrichment is detected. The TEM micrographs of Fig. 9 show the cross section surface of a polished amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 plate after air annealing for 23 days at 75 °C. The dendrite-features were found to be amorphous (see bright-contrast regions in Fig. 9b). The TEM analysis did not find detectable traces of sulphur in agreement with the dry conditions in the furnace, due to the lack of a constant flux of air. In addition, in the analyses of Fig. 9, the Cu-oxidation products (detected by XPS in Fig. 8) were not detected with certainty. In Fig. 9b, the interface between the metallic matrix (dark contrast) and the dendrite-branches (bright contrast) shows fringes that indicate long range order, namely crystallinity. Beside the amorphous rings, the electron diffractions of the matrix and the dendrite, in Fig. 9c and d, respectively, show some crystalline reflections that can be assigned to nanocrystalline Au. Some other reflections are present which are difficult to assign. One possible structure that could give these reflections is the Cu3Au. The FIB image of Fig. 10 shows the surface topography of the Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen after air exposure at 75 °C for

According to Henderson [15], values of Yellowness Index as low as the ones reported in this work (in Table 1), are to be considered among the lowest values ever detected on ‘premium white’ gold alloys. It means that, according to the definition given in the experimental methods section, the studied glassy compositions possess, among white gold alloys, the smallest aberrations from a perfect white colour towards a yellow colour. The YI of the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 alloy was previously reported as 25.3, an intermediate value between standard white to off-white gold colour [6]. The value of Ref. [6] was, however, not immediately measured after polishing, but rather after a non-protective storage period. In Fig. 1, a value of 25.3 indeed corresponds to a tarnishing after air exposure time of circa 60 days at 23 °C. Only when kept in the freezer (circle symbols in Fig. 1) the rate of the YI variation is very slow. In Fig. 11 the time is plotted that is necessary to reach off-white gold colours (24.5 YI) as a function of temperature when the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 polished specimen is simply exposed to air atmospheres. This figure shows that for this BMG alloy the white-gold regime exists only for a short time. For example, at room temperature, the polished surface of this composition tarnishes to off-white gold values after only 2 months of air exposure. The samples kept at higher temperatures are so intensively tarnished that the YI reaches off-scale values after 2 months of storage. Rate constants for the initial tarnishing reaction were obtained by applying a linear rate dependency (valid in first approximation) to the data of Fig. 1. The rate constants are plotted in the inset of Fig. 11 against the inverse temperature. The corresponding thermal activation energy is of approximately 0.45 eV. This is a typical average value for a corrosion mechanism, which includes more than an oxidation mechanism. In comparison, the thermal activation energy for Cu+ diffusing into Cu2S is 0.3 eV and that for Cu+ diffusing throughout Cu2O is 1.8 eV. For Si2O the thermal activation energy of formation is 0.70.8 eV [19]. 4.2. The role of the glassy matrix The most important result of this study is that the large amount of Cu (25 at.%) and Si (16 at.%) alloyed to gold to form a glassy structure is the major contributing factor to the fast tarnishing kinetics observed at low temperatures on the studied Au–Ag– (Pd)–Cu–Si BMGs. In the glassy structure, despite of short- or middle-range ordering phenomena, atomic positions are random. The lack of grain boundaries and the chemical homogeneity are often advocated for an expected better corrosion and oxidation resistance of the glassy structure with respect to crystalline alloys. However, it was proved that amorphous alloys can have worse oxidation resistance than their crystalline counterparts [20]. In crystalline alloys,

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

6

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

Fig. 7. EDX-maps for Cu, O and Si of the cross section of the as-polished surface of a Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen after 131 days of air annealing at 75 °C. The scanned region is indicated in the back scattered electron microphotograph. Si-map the dark points reflect concentration of Si between 7.5 and 16 at.%.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

Fig. 8. XPS concentration profile of a polished surface of a specimen of composition Au49Ag5.5Pd2.3Cu26.9Si16.3 after air annealing at 75 °C for 23 days as a function of sputtering time. An average sputtering rate of 4.56 nm/min was used for the plotting.

7

grain boundaries, interphase boundaries and other defects such as dislocations are known to act as fast diffusion paths and as reactive heterogeneous sites on the surface. The assumption that the lack of these defects would alone slow down diffusion and consequently oxidation or corrosion processes is an oversimplification. It is based on the assumption that diffusion in metallic glasses is comparable in nature and kinetics to volume-diffusion in crystals. However, diffusion in metallic glasses is sharply different from vacancy-mediated diffusion in most crystalline metals. During the diffusion process a certain amount of frozen-in ‘free volume’, or ‘quasivacancies’ according to Faupel et al. in Ref. [21], serve as diffusion vehicles until they have annealed out. The diffusion process, in the order of one atomic displacement, involves thermally activated, highly collective atomic motions of 10–20 atoms in a chainlike manner [21]. To complete the picture, the interatomic interactions between the constituents have to be taken into consideration. In the metal-metalloid glass system of the present study, the interactions between the metallic elements and the metalloid (Si) are weaker than the (already weak) metal–metalloid interactions in the crystal. In general, the interatomic energy potential in the glass is much

Fig. 9. TEM analyses of a cross section of the Au49Ag5.5Pd2.3Cu26.9Si16.3 after 23 days at 75 °C. (a) Bright field image of the as-polished tarnished surface. The rings indicate the excitation zone for the diffractions of c and d. The square indicates the location of the photo of b. (b) Bright field image at high magnification showing that the core locations of a silica dendrite-like feature (bright contrast areas) are amorphous. The matrix (dark contrast) is not transmitting well due to the high present of Au. However, in the matrix adjacent the glassy silica formations crystalline fringes are visible. (c) Electron diffraction of the bulk metallic matrix. Only at the edge of the specimen, where the part of the lamella burned away by the FIB bombardment, nanocrystalline reflections of Au has been detected. Otherwise the matrix is amorphous. (d) Electron diffraction of the silica dendrite-like feature. Beside the amorphous rings they are visible the reflections of nanocrystalline Au and probably Cu3Au.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

8

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

Fig. 10. FIB-back scattered electrons microphotograph of the surface of Au49Ag5.5Pd2.3Cu26.9Si16.3 after air exposure at 75 °C for 327 days. A Cu2O-rich cone-like protrusion was cross-sectioned revealing underneath an extensive internal oxidation of Si in the metallic matrix as dendrite-like products. The matrix appears homogenous and amorphous.

Fig. 11. The time necessary to reach off-white gold colours (corresponding to a value of 24.5 YI) as a function of temperature in the case the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 polished specimen is exposed to air atmospheres. In the inset, the logarithmic values of the tarnishing rates are plotted against the inverse temperature. The thermal activation energy, Q, is determined to be approximately 0.45 eV.

shallower in shape than in the crystal and therefore the bond strength is weaker in the glass than in the crystalline structure. This is also reflected in the smaller elastic modulus of the glass compared to the crystalline counterparts. It implies a certain facility of the metastable glassy matrix to local atomic rearrangements, which can lead to easy partitioning and dealloying processes. At the studied concentrations (i.e. close to the Au–Si eutectic concentration), a homogeneous single phase crystal structure cannot thermodynamically exist in equilibrium. Due to its extremely low solubility in gold, the silicon would precipitate out at this overall concentration (intermetallic compounds do not thermodynamically exist in the Au–Si system). In the solidification process, the random atomic structure of the liquid is frozen during the glass formation and a metastable homogenous phase is formed. In contrast, the crystalline counterpart is not homogeneous. In fact, the equilibrium crystalline state at the nominal composition of the alloy is a multiphase structure made of pure Si, Au-rich solid solution and other three intermetallic phases rich in Au, in Pd and in the less-noble Cu. [10]. This crystalline mixture is therefore already decomposed and cannot be compared to the homogeneous glassy matrix in terms of diffusivity and surface reactivity.

Fig. 12. XPS Au/Si, Cu/O and Si/O concentration profiles of a polished surface of a Au49Ag5.5Pd2.3Cu26.9Si16.3 specimen after air annealing at 75 °C for 23 days as a function of sputtering time. The sputtering rate used for the plotting is 4.56 nm/ min.

Fig. 13. Double-logarithmic plot of the colour change DE versus time of air exposure of various Au-based BMGs at 75 °C.

4.3. The role of Cu and Au partitioning during Si oxidation This work shows the experimental evidence that the tarnishing mechanism is controlled by the internal oxidation of Si, which is linked in some way to an enhanced out diffusion of Cu and Au and an enhanced oxidation of Cu on the surface. It is important to notice that this happens rapidly even at ambient temperature. Amorphous SiO2 was found to form on polished surfaces as well (see Fig. 9b). Just underneath, a Si-depleted (Au-enriched) zone is found (see Fig. 12), and more internally glassy silica dendrite-like features nucleate at low partial pressure of oxygen within the sub-surface matrix. Kelton and Spaepen have reported that Pd–Si metallic glasses have a native surface layer of SiO2 and have seen the evidence of a release of elemental Pd into the matrix [22]. Similarly, in the analysed Au–Ag–(Pd)–Cu–Si BMGs, during the oxidation of Si the adjacent matrix gets depleted in Si and enriched in the other elements. The fast colour change of polished surfaces from white to yellow-gold colours can be attributed to the surface partitioning of Au and Cu as amorphous silica underneath is colourless and transparent to light. When elemental Au and Cu are

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

9

Fig. 14. Schematics of the proposed tarnishing mechanism for Au-based BMGs depending on the surface conditions: (a) as-cast surface which solidified in contact with air, and (b) surface which was polished prior to exposure.

released out of the matrix during the formation of SiO2, they can diffuse out extremely fast. The diffusivity of Au in bulk Au–Si eutectic is, indeed, one order of magnitude faster than in pure Au at the same temperature [23,24]. Gold is also well known to have a strong tendency for surface granulation, and this has been detected in bulk metallic glasses as well. The nanocrystallization of Au particles as a result of air-oxidation of Zr-based BMGs has been reported by Köster and co-workers [25]. Recent work performed on melt spun ribbons by Battezzati and co-workers on the dealloying of Au49Ag5.5Pd2.3Cu26.9Si16.3 in artificial sweat [26] and on the Au–Cu–Ti–Si BMG system [15] have found experimental evidence of Au-nanocrystalline particles left on top of dealloyed ribbons. In the present work, gold nanoparticles were not detected directly on the surface of air-exposed specimens (the bright-contrast nanoparticles dispersed on the surface of Fig. 10 could be attributed only to Cu2O). The only evidence that they might form during the tarnishing in air was found in the TEM electron diffraction patterns (Fig. 9d) in the matrix adjacent to the SiO2 features. Since we cannot exclude nanocrystallization processes taking place during the FIB-lamellae sample preparation, as in Fig. 9c, we cannot conclude with certainty that the gold nanocrystallization is an oxidation-induced process. At longer times of exposure, the tarnishing to orange and brown colours may be discussed in terms of Cu partitioning, out diffusion and its consequent corrosion on the surface. An abnormal surface oxidation of Si at room temperature has been reported to occur in the presence of copper [27–29]. First, Harper et al. proposed the possibility of the following reactions [27]:

3Cu þ Si ¼ Cu3 Si;

ð2Þ

and

Cu3 Si þ 2O ¼ SiO2 þ 3Cu:

ð3Þ

After copper-ion implantation, on pure silicon, the copper was found diffusing through the Si dioxide surface layer even at room temperature and the Si oxide growth was found to depend on the amount of Cu present at the moving Si/SiO2 interface [28]. It was proposed that Cu is essential for the dissociation of oxygen on the Si surface, by forming Cu2O according to the reaction:

4Cu þ 2O ¼ 2Cu2 O

ð4Þ

Moreover, Cu oxide is known to be reduced by the presence of Si to form SiO2 and elemental copper, following the reaction:

2Cu2 O þ Si ¼ 4Cu þ SiO2

ð5Þ

This is because SiO2 has a very negative driving force for formation in comparison to Cu2O [30].

The result of the proposed chemical reactions (Eqs. (3 and 5)) is that elemental copper is released (partitioned) and SiO2 is formed by oxidation. The partition of elemental copper in the surroundings is even enhanced by the glassy nature of the matrix, due to weak atomic interactions, as discussed in Section 4.2. Once Cu is partitioned, or dealloyed from the glassy matrix, it can become highly mobile even at room temperature and can diffuse to the surface very rapidly. In fact, as a consequence of its small ionic radius, copper is known to have relatively weak interactions with the other atoms (especially with the metalloid Si as reported in Ref. [31]). The internal oxidation of Si was observed in this work to occur underneath polished surfaces only (see Figs. 6b, 7b, 9b, and 10). The formation of similar SiO2 dendrite-like features was reported in Cu–Si alloys only for high temperatures (650–1000 °C) [32–34]. They are formed as a result of the internal oxidation of Si and are observed in various Cu–Si crystalline structures: in polycrystalline alloys [32], in bulk of single crystals [33], and along the grain boundary of bicrystals [34]. In this system, dendrite-like SiO2 structures form in a glassy Au–Cu–Si-based matrix at much lower temperatures (ambient to 75 °C). At these low temperatures one can expect fast atomic diffusivities in the glassy matrix because, in addition to the arguments discussed in the Section 4.2, they scale with the low Tg (128 °C) rather than with the high melting point of the Cu–Si crystalline alloys (800 °C or greater). The nucleation and growth of the glassy silica features appears to be controlled by the supply of oxygen and the partitioning of copper. The largest dendrite-like formations appear to grow faster at locations underneath porous cone-like Cu2O products (see Figs. 6b, 7b, and 10). Similar observations were reported by Hinode and coworkers in Ref. [29]. The oxidation of Si occurs abnormally at room temperature in the presence of copper. Ref. [29] shows that areas of a Si substrate covered with very thin or cone-like copper oxidized fast, forming 100 nm thick SiO2 films. Areas covered with a thick and non-cone-like Cu-rich film, did not. The supply of oxygen through the Cu oxide layer controls the Si oxidation process. The presence of a Cu silicide, such as Cu3Si, was found in Ref. [29] to be not necessarily needed for oxidation, which was reported earlier in Ref. [27]. Copper atoms were detected at the SiO2/Si interface and identified not as silicide but as body-centred-cubic-structured Cu several atom layers thick. In the amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 alloy oxidized at 75 °C for almost a year (as in Fig. 10) the internal SiO2 formation reaches a depth of approximately 1.15 lm under cone-shape Cu2O products and only 0.1–0.3 lm at other locations. Furthermore, in Fig. 12 the detected XPS concentration values are plotted by using the ratio of Cu and Si to O, as well as the ratio Au/Si. This way the Au-enrichment (and Si-depletion) underneath the Cu2O scale is very evident (at 100 s) as well as the presence of a Cu enrichment at the front tip

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

10

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

of the growing SiO2 particle (at 1250 s) and between its branches (e.g. at 400 s). Other evidence that the formation of internal SiO2 involves a considerable partitioning of elemental Au and elemental Cu is that the surrounding matrix can form Au-nanoparticles and Cu-rich intermetallic (see TEM Images of Fig. 9). 4.4. Oxidation of Si during solidification producing a coating film In the present study we have observed faster tarnishing kinetics on mechanically polished surfaces, similarly to scratched plate surfaces or HF-etched granules surfaces of Ref. [16]. This may be attributed to the lack or removal of a protective Si-oxide surface layer. Distinctly slower tarnishing kinetics was observed onto ascast surfaces which oxidized during the solidification- namely surfaces at air pockets as in Fig. 6a or surfaces of granules like in Ref. [16]. It is important to notice that the Si oxide layer, which has been formed at high temperatures during the solidification, acts as a protective coating, differently from the amorphous SiO2 that forms onto solid surfaces at low temperatures on polished glassy specimens. It has been found that dense SiO2-like coatings possess excellent barrier effects against tarnishing of silver artefacts [35]. The protective effectiveness of the coating acts against the diffusion of water and other gaseous aggressive agents present in the environment and prevents them from coming into contact with the metal surface. The formation of amorphous silica or silicides is also the probable reason advocated for the good oxidation resistance of alloys containing high amounts of Si. It is also known that vitreous well-adhering scales of Silica or of silicides (e.g. MoSi2, TiSi2) are highly protective. Ionic diffusion through those scales is very slow. Fukumoto et al. [36] reported that also the hot corrosion resistance of the stainless steels has been greatly improved by the surface alloying of Si, due to the formation of protective SiO2 which would resist basic or acidic dissolution in molten salts during electrolysis in fused Na2SO4. The formation of the protective SiO2 film is thought to be the result of oxidation of the Si-enriched surface layer of the molten material during the solidification process. Indeed, in our work we have detected SiO2 films on the surface of cast specimens at air-pockets (few nm thick) or on the surface of granules (from few nm to 100 nm thick in some locations) which have been exposed to air during free dropping [16], and not on the surface of specimens that have been solidified in contact with copper cavities and/or under high vacuum or high-purity argon flux. This effect has been observed also by Battezzati and co-workers on the surface of amorphous Au49Ag5.5Pd2.3Cu26.9Si16.3 melt spun ribbons. There, XPS analysis proved that the formation of SiO2 occurs only on the air-side surface and not on the copper wheel side surface [14,15]. 4.5. Enrichment of Si on the surface of molten material The enrichment of Si, and therefore its oxidation, on the surface of molten Au49Ag5.5Pd2.3Cu26.9Si16.3 may be understood considering that liquid Si has a lower surface tension than Au. Recently the evidence of a Si-enriched surface layer 2.5 nm thick has been reported by Shpyrko et al. on the Au81Si19 eutectic formed at the eutectic temperature of 363 °C [37,38]. The composition of the surface layer was found to be approximately 70 at.% Si [37]. Similarly, surface crystallization of presumably Si-enriched phases has been detected in liquid eutectic Au-Si, in liquid Au77Ge14Si9, and in liquid Au49Ag5.5Pd2.3Cu26.9Si16.3 [37,39,40]. The surface crystallization has been proven by the evidence of melting of ordered surface-structures detected for these systems at temperatures much higher than the liquidus temperature (about 800–900 °C) [39,41]. It has also been reported that the Au49Ag5.5Pd2.3Cu26.9Si16.3 metallic glass requires significant pressure for filling cavities below 100 nm at temperatures between 500 and 700 °C due to an anti-wetting

behaviour on quartz [42], which can be considered an indication of its surface crystallization according to Ref. [40].

4.6. Formation of Cu2O versus Cu2S and the role of the Cu/Si and Au/Si ratios The present work reports that the aggressiveness of the environment plays an important role in the tarnishing effect (see Fig. 2). A previous work [16], showed that an as-cast BMG plate with composition Au49Ag5.5Pd2.3Cu26.9Si16.3 after two years of being exposed to air, wet environments and to human pollutions, has built an Cu2S (Cu(I) sulphide) corrosion scale of the surface. In the XPS concentration profile the external corrosion scale is Cu2O and not Cu2S (ratio Cu/O ffi 2 in Fig. 12). This is attributed to the fact that in the furnace the air has was not highly polluted with H2S gases and the contact to sweat was avoided altogether by handling the specimens with gloves. In terms of Cu2O formation versus Cu2S the relative humidity of the atmosphere has to be taken into account. In Ref. [43], the nucleation rate of Cu2O on a pure copper surface is reported to be stimulated at high relative humidity and the average Cu2O thickness is increased by a factor of 48, when relative air humidity increases from 40% to 80%. In the case of the Cu2S, the mechanism of formation involves besides a gas-period also an aqueous-period, which could have been the case with Au49Ag5.5Pd2.3Cu26.9Si16.3 in the contact to human sweat [16]. In the present work a standard immersion 7-day sulphide test has been performed on all of the studied Au-based BMG alloy compositions. The colour change caused by the formation of copper sulphides was detected every day of the test and plotted in Fig. 4. The rate of the colour change of the Au49Ag5.5Pd2.3Cu26.9Si16.3 is reduced by reducing the Cu/Si content in the alloy. For example for the BMG with composition Au60Ag5.5Pd2Cu15.5Si17 (circles in Fig. 3a), after 3 weeks of air exposure the colour change is only 50% of that for the Au49Ag5.5Pd2.3Cu26.9Si16.3. It is remarkable that the rate can be drastically slowed down by reducing the Cu/Si ratio independently of the exposure, as shown in Figs. 3a and 4. When the data of Fig. 3a are plotted as double logarithmic (see Fig. 13) the tarnishing rate appears to change behaviour as the Cu/Si ratio is decreased and the Au/Si ratio is increased. For example, the Au49Ag5.5Pd2.3Cu26.9Si16.3 (squares) and the Au50Ag7.5Cu25.5Si17 (open triangles) follows a fast pure parabolic rate, and the Au60Ag5.5Pd2Cu15.5Si17 (circles) shows a more passivizing pure logarithmic behaviour. The logarithmic law was proved by a linear fitting of the DE values versus log(time). Based on this observation, it is expected that the tarnishing of Au49Ag5.5Pd2.3Cu26.9Si16.3 BMGs can be brought to acceptable rates and the long term white gold stability region can be prolonged by several orders of magnitude, if the nominal Cu/Si and/or the Si/Au ratios in the alloy are in some way drastically reduced. In Fig. 13, the Au49Ag5.5Pd2.3Cu26.9Si16.3 and the Pd-free Au50Ag7.5Cu25.5Si17 change mode of tarnishing behaviour from a pure-parabolic to a slower rate at long exposure times. In the case of Au49Ag5.5Pd2.3Cu26.9Si16.3 (squares) this is an apparent effect due to the reaching of the detection limit of log(DE) = 1.7 for DE = 50. For the Au50Ag7.5Cu25.5Si17 (open triangles) this is an indication that the alloy crystallized during the heat treatment (as shown in Fig. 3b). This alloy has a lower Tg (104 °C) and therefore expected faster bulk kinetics than the Au49Ag5.5Pd2.3 Cu26.9Si16.3. Despite that, its parabolic tarnishing behaviour (before it crystallizes) is slower than the parabolic tarnishing rate of Au49Ag5.5Pd2.3Cu26.9Si16.3. The Al-containing alloy Au49Ag5.5Pd2.3 Cu25.9Si16.3Al1 (stars) follows a cubic rate law which could be attributed to a mixture between logarithmic and parabolic behaviour. It is very stable and it is still X-ray amorphous after almost a year at 75 °C (Fig. 3b).

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

4.7. The role of Pd and Ag The Pd-free Au50Ag7.5Cu25.5Si17 has Cu/Si and Au/Si atomic ratios very similar to those of the Au49Ag5.5Pd2.3Cu26.9Si16.3 (see Table 1). The removal of Pd increases the Ag content. This Pd-free BMG has a lower Tg and therefore expected faster bulk kinetics in the glassy matrix. Despite that, the tarnishing rate is halved (even before the Au50Ag7.5Cu25.5Si17 crystallizes) with respect to the Au49Ag5.5Pd2.3 Cu26.9Si16.3 (see Figs. 3a and 4). Crystalline Au–Cu–Ag alloys show the opposite trend. For example, Suoninen et al. [44] showed experimental evidences that small additions of Pd to dental crystalline Au–Ag–Cu alloys played an important role in essentially reducing the thickness of the sulphide scale on surfaces of samples treated in aqueous Na2S solutions. The Pd did not enrich the sulphide phase, in contrast to Ag, but was found enriched below the sulphide scale. The mechanism for the impeding tarnishing effects was assumed to be a decrease in diffusion from the bulk alloy to the surface due to the enrichment layer. Similarly Lang et al. have found that when Pd is added to dental Au–Ag–Cu a drastic reduction in the degree of tarnishing is observed [45]. There, an atomic ratio of Pd/Au of 1/12 (which is the ratio in Au49Ag5.5Pd2.3Cu26.9Si16.3) is often enough to cause the largest reduction of the degree of tarnishing. The opposite trend is observed here, and could indicate that the tarnishing mechanism in the case of bulk metallic glass alloys is surface controlled, rather than bulk diffusion controlled like in the case of crystalline alloys. 4.8. The role of Al The addition of 1 at.% aluminium to Au49Ag5.5Pd2.3Cu26.9Si16.3 seems to be able to drastically reduce (by 2/3) the tarnishing rate when it is exposed to air and therefore dry conditions (see Fig. 3a) and to follow a slower cubic law in comparison to the mother-alloy as it is indicated in the double logarithmic plot of Fig. 13. It is nevertheless interesting to notice that Al in as-cast Zr–Ni–Cu–Al–Nb based BMG has been detected in its oxidized state, even in the bulk amorphous matrix by XPS analyses [46]. The other elements were found in bulk in their metallic state. Therefore, we cannot argue that the much higher affinity to oxygen of Al can compete with Si to form a passivizing oxide on the as-polished surfaces, since the Al is probably already in its oxidized state after casting. The Al role in the apparent slower tarnishing rate is, therefore, not yet fully understood. The alloy with 1 at.% Al did not show any improvement with respect to the mother-alloy in terms of colour change reduction during the sulphide test (Fig. 4) when Cu2S products are involved. This could also be an apparent effect caused by dealloying effects and the occurrence of loss of some corrosion products in the liquid test solution.

11

of exposure towards yellow and red, respectively. At longer times of exposure copper-rich corrosion scales were experimentally observed. When the surface of the samples is polished prior to exposure, the glassy matrix underneath the Cu-rich scale appears destabilised and triggers the internal nucleation of amorphous SiO2 structures which grow inwards in a dendritic fashion. This internal oxidation occurs at low temperatures (RT–75 °C) at low partial pressure of oxygen and it is linked to the partitioning of the other elements in the surrounding matrix and, therefore, enhances the out-diffusion of Au and Cu. Au is expected to form nanocrystalline particles and may be left behind on the surface during dealloying in wet environments (as in Ref. [15]). Cu can react on the surface and form Cu2O and Cu2S depending on the exposure. The removal of the Pd from the alloy and the increase of Ag content in the glass seems impeding the tarnishing, opposite to crystalline alloys. The microaddition of Al slows down the tarnishing of the glassy alloy when the product is in contact to the air. The addition of Al does not help when the material is kept in wet conditions especially because the passivizing products may be etched away. 6. Conclusions The tarnishing behaviour of the Au49Ag5.5Pd2.3Cu26.9Si16.3 BMG is very detrimental to its widespread application in jewellery or dentistry casting. The native SiO2 layer produced during casting can be scratched away by wearing or simply by polishing or etching the final products. The tarnishing process is driven by partitioning of the metallic atoms such as the highly mobile and less noble Cu. The atomic partitioning is linked with the internal oxidation of the metalloid Si. The key to understand the mechanism lies in the abnormal room temperature formation of amorphous SiO2 dendrite-like features in the subsurface glassy matrix which is a result of a destabilization of the metastable glassy phase. Aiming at slower tarnishing rates, the role of each atomic constituent as well as the role of minor additions should be taken into consideration in future works. Acknowledgments This research was supported the German Federation of Industrial Research Associations (AiF/IGF) through Grant No. 16843N. The authors wish to thank A. Zielonka, J. Schmauch (TEM analyses), and J. Schroers for useful discussions, and to express their gratitude to C. Hafner Edemetall Technologie and the World Gold Council for the noble metals supply. I. Gallino gratefully acknowledges the support of the Deutsche Forschungsgemeinschaft (German Research Foundation) through Grant No. GA 1721/2-1. References

5. Summary of the proposed tarnishing mechanism As shown in the schematics of Fig. 14, a native SiO2 layer has formed on as-cast surfaces (Fig. 14a) during the solidification in contact to air, like at air-pockets (as in Fig. 5), or surfaces of granules (as shown in Ref. [16]). This layer seems to be so dense that it acts as a diffusion barrier limiting the supply of oxygen and in general those surfaces experience minimal tarnishing. When this silica coating-like layer is removed (Fig. 14b) or damaged, the thin amorphous silica layer that forms on the surface of the bulk metallic glass at room temperature does not act as a diffusion barrier or coating as much as the native SiO2 layer that forms during solidification. Au and Cu are released in the matrix as elements during the Si oxidation and diffuse out to the surface through the thin SiO2 surface layer. They cause a change in the colour at the beginning

[1] W. Klement, R.H. Willens, P. Duwez, Non-crystalline structure in solidified gold–silicon alloys, Nature 187 (1960) 869–870. [2] J. Schroers, B. Lohwongwatana, W.L. Johnson, A. Peker, Gold based bulk metallic glass, Appl. Phys. Lett. 87 (2005) 061912. [3] W. Zhang, H. Guo, M.W. Chen, Y. Saotome, C.L. Qin, A. Inoue, New Au-based bulk glassy alloys with ultralow glass transition temperature, Scripta Mater. 61 (7) (2009) 744–747. [4] S.-H. Wang, T.-S. Chin, Tin-modified gold-based bulk metallic glasses, Gold Bull. 45 (2012) 3–8. [5] G. Fiore, L. Battezzati, Developing Au-based amorphous alloys, Rev. Adv. Mater. Sci. 18 (2) (2008) 190–192. [6] S. Mozgovoy, J. Heinrich, U.E. Klotz, R. Busch, Investigation of mechanical, corrosion and optical properties of an 18 carat Au–Cu–Si–Ag–Pd bulk metallic glass, Intermetallics 18 (12) (2010) 2289–2291. [7] U.E. Klotz, M. Eisenbart, Gold based bulk metallic glasses: hard like steel, moldable like plastics, in: Eddie Bell (Ed.), Proc. 27th Santa Fe Symp., Albuquerque, New Mexico, 2013, pp. 221–235. [8] C.W. Corti, The role of hardness in jewelry alloys, in: Eddie Bell (Ed.), Proc. 22th Santa Fe Symp., Albuquerque, New Mexico, 2008, pp. 103–120.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024

12

M. Eisenbart et al. / Corrosion Science xxx (2014) xxx–xxx

[9] B. Lohwongwatana, J. Schroers, W.L Johnson, Liquidmetal-hard 18K and.850 Pt alloys that can be processes like plastics or blown like glass, in: Eddie Bell (Ed.), Proc. 21th Santa Fe Symp., Albuquerque, New Mexico, 2007, pp. 289–303. [10] G. Fiore, P. Rizzi, L. Battezzati, Phase constitution and glass formation in an Aubased alloy, J. Alloys Compd. 509 (S1) (2011) S166–S169. [11] Triwikantoro, D. Toma, M. Meuris, U. Köster, Oxidation of Zr-based metallic glasses in air, J. Non-Cryst. Solids 250 (1999) 719–723. [12] U. Köster, Triwikantoro, Oxidation of amorphous and nanocrystalline Zr–Cu– Ni–Al alloys, Mater. Sci. Forum 360–363 (2001) 29–36. [13] I. Gallino, R. Busch, H. Choi Yim, L. Jastrow, U. Köster, High temperature oxidation of the refractory alloy glass Nb35 Ni60 Sn5, J. Alloys Compd. 434 (2007) 225–228. [14] P. Rizzi, I. Corazzari, I. Fenoglio, B. Fubini, S. Kaciulis, L. Battezzati, Ion release and tarnishing behavior of Au and Pd based amorphous alloys in artificial sweat, Corros. Sci. 77 (2013) 135–142. [15] F. Scaglione, A. Gebert, L. Battezzati, Dealloying of an Au-based amorphous alloy, Intermetallics 18 (2010) 2338–2342. [16] M. Eisenbart, U.E. Klotz, R. Busch, I. Gallino, On the abnormal room temperature tarnishing of an 18 karat gold bulk metallic glass alloy, J. Alloys Compd. S13 (2013), http://dx.doi.org/10.1016/j.jallcom.2013.11.167. [17] S. Henderson, D. Manchanda, White gold alloys: colour measurement and grading, Gold Bull. 38 (2) (2005) 55–67. [18] C.W. Corti, What is a white gold? Progress on the issues!, in: Eddie Bell (Ed.), Proc. 19th Santa Fe Symp., Albuquerque, New Mexico, 2005, pp. 103–119. [19] B.E. Deal, A.S. Grove, General relationship for the thermal oxidation of silicon, J. Appl. Phys. 36 (12) (1965) 3770–3778. [20] U. Köster, D. Zander, Triwikantoro, A. Rudiger, J. Jastrow, Environmental properties of Zr-based metallic glasses and nanocrystalline alloys, Scripta Mater 44 (8–9) (2001) 1649–1654. [21] F. Faupel, W. Frank, M.P. Macht, H. Mehrer, V. Naundorf, K. Ratzke, H.R. Schober, S.H. Sharma, H. Teichler, Diffusion in metallic glasses and supercooled melts, Rev. Mod. Phys. 75 (1) (2003) 237–280. [22] K.F. Kelton, F. Spaepen, A study of the devitrification of Pd82Si18 over a wide temperature-range, Acta Metall. 33 (3) (1985) 455–464. [23] A. Pasturel, E.S. Tasci, M.H.F. Sluiter, N. Jakse, Structural and dynamic evolution in liquid Au–Si eutectic alloy by ab initio molecular dynamics, Phys. Rev. B 81/ 14 (2010) 140202. [24] A. Bruson, M. Gerl, Diffusion-coefficient of Au-195 in the liquid Au0.81Si0.19 alloy, J. Appl. Phys. 53 (5) (1982) 3616–3619. [25] L. Jastrow, U. Köster, M. Meuris, Catastrophic oxidation of Zr-TM (noble metals) glasses, Mater. Sci. Eng. A 375–377 SI (2004) 440–443. [26] S. Kaciulis, A. Mezzi, G. Fiore, I. Ichim, L. Battezzati, P. Rizzi, XPS study of goldbased metallic glass, Surf. Interface Anal. 42 (2010) 597–600. [27] J.M.E. Harper, A. Chari, L. Stolt, F.M. d’Heule, P.M. Fryer, Room temperature of silicon catalyzed by Cu3Si, Appl. Phys. Lett. 56 (25) (1990) 2519–2521. [28] T.L. Alford, E.J. Jaquez, N.D. Theodore, S.W. Russel, M. Diale, D. Adams, S. Anders, Influence of interfacial copper on the room temperature oxidation of silicon, J. Appl. Phys. 79 (4) (1996) 2074–2078.

[29] K. Hinode, K. Takeda, S. Kondo, Abnormal room-temperature oxidation of silicon in the presence of copper, J. Vac. Sci. Technol. A 20 (5) (2002) 1653– 1658. [30] I. Gallino, M.E. Kassner, R. Busch, Oxidation and corrosion of highly alloyed Cu– Fe–Ni as inert anode material for aluminum electrowinning in as-cast and homogenized conditions, Corros. Sci. 63 (2012) 293–303. [31] A.A. Istratov, E.R. Weber, Physics of copper in silicon, J. Electrochem. Soc. 149 (1) (2002) G21–G30. [32] P. Bolsaitis, M. Kahlweit, The internal oxidation of Cu–Si alloys, Acta Metall. 15 (1967) 765–772. [33] S.B. Bhattacharyya, K.C. Russel, Coarsening kinetics of silica in copper, Metall. Trans. 7A (1976) 453–462. [34] M. Sato, I. Ota, T. Fujii, S. Onaka, M. Kato, Morphological evolution of grain boundary SiO2 in internal oxidizes Cu–Si bicrystals, Metall. Trans. 45 (3) (2004) 818–823. [35] E. Angelini, S. Grassini, G.M. Ingo, D. Mombello, F. Fracassi, F. Palumbo, Surface analysis of SiO2-like high-barrier thin films for protection of silver artifacts, Surf. Interface Anal. 42 (2010) 666–670. [36] M. Fukumoto, C. Tachikawame, Y. Matzuzaka, M. Hara, Formation of Si diffusion layer on stainless steels and their high temperature corrosion resistance in molten salt, Corros. Sci. 56 (2012) 105–113. [37] O.G. Shpyrko, R. Streitel, V.S.K. Balagurusamy, A.Y. Grigoriev, M. Deutsch, B.M. Ocko, M. Meron, B. Lin, P.S. Pershan, Surface crystallization in a liquid AuSi Alloy, Science 313 (5783) (2006) 77–80. [38] O.G. Shpyrko, R. Streitel, V.S.K. Balagurusamy, A.Y. Grigoriev, M. Deutsch, B.M. Ocko, M. Meron, B. Lin, P.S. Pershan, Crystalline surface phases of the liquid Au–Si eutectic alloy, Phys. Rev. B 76 (2007) 245436. [39] L. Bischoff, J. Teichert, T. Ganetsos, G.L.R. Mair, Temperature dependence of the electric characteristics of liquid metal alloy ion sources, J. Phys. D: Appl. Phys. 33 (2000) 692–695. [40] S. Mechler, E. Yahel, P.S. Pershan, M. Meron, B. Lin, Crystalline monolayer surface of liquid Au–Cu–Si–Ag–Pd: metallic glass former, Appl. Phys. Lett. 98 (2011) 251915. [41] L. Bischoff, G.L.R. Mair, C.J. Aidinis, C.A. Londos, C. Akhmadaliev, Th. Ganetsos, A Au82Si18 liquid metal field-ion emitter for the production of Si ions: fundamental properties and mechanisms, Ultramicroscopy 100 (2004) 1–7. [42] S. Ding, J. Kong, J. Schroers, Wetting of bulk metallic glass forming liquids on metals and ceramics, J. Appl. Phys. 110 (2011) 043508. [43] T.T.M. Tran, C. Fiaud, E.M.M. Sutter, Oxide and sulphide layers on copper exposed to H2S containing moist air, Corros. Sci. 47 (2005) 1724–1737. [44] E. Suoninen, H. Hero, E. Minni, Effect of palladium of sulphide tarnishing of noble metal alloys, J. Biomed. Mater. Res. 19 (1985) 917–934. [45] B.R. Lang, S.H. Bernier, Z. Giday, K. Asgar, Tarnish and corrosion of noble metal alloys, J. Prosthet. Dent. 48 (3) (1982) 245–252. [46] J. Heinrich, R. Busch, F. Müller, S. Gradthyll, S. Hüfner, Role of aluminum as an oxygen-scavenger in zirconium based bulk metallic glasses, Appl. Phys. Lett. 100 (2012) 071909.

Please cite this article in press as: M. Eisenbart et al., A colourimetric and microstructural study of the tarnishing of gold-based bulk metallic glasses, Corros. Sci. (2014), http://dx.doi.org/10.1016/j.corsci.2014.04.024