Decomposition of bulk metallic glasses

Decomposition of bulk metallic glasses

Materials Science and Engineering A250 (1998) 133 – 140 Decomposition of bulk metallic glasses M.K. Miller * Microscopy and Microanalytical Sciences ...

861KB Sizes 3 Downloads 161 Views

Materials Science and Engineering A250 (1998) 133 – 140

Decomposition of bulk metallic glasses M.K. Miller * Microscopy and Microanalytical Sciences Group, Metals and Ceramics Di6ision, Oak Ridge National Laboratory, PO Box 2008, Oak Ridge, TN 37831 -6376, USA

Abstract An atom probe field ion microscopy (APFIM) characterization has been performed on the decomposition of several bulk metallic glasses. Evidence of short range order for aluminum has been detected in as-cast nickel-containing Zr60Al15Ni25 and Zr55Al10Cu5Ni30 metallic glasses. Titanium-enriched, zirconium-enriched and Be2Zr phases have been analyzed in annealed Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 bulk metallic glass. © 1998 Elsevier Science S.A. All rights reserved. Keywords: Bulk metallic glass; Crystallization; APFIM; Short range order

1. Introduction Since the first amorphous deposit of nickel formed by Wurtz in 1845 [1], many different metallic glasses have been produced. Initially, amorphous metals were formed by condensation onto a cold surface or by electrolysis. In the 1960s, Duwez and his colleagues pioneered the use of splat-cooling techniques [2]. This approach was adopted to prevent crystallization by effectively freezing in the structure of the liquid with the use of cooling rates on the order of 106 K s − 1. In order to achieve this extremely high rate of heat transfer, at least one of the dimensions of the material produced was restricted in magnitude. Therefore, metallic glasses could only be produced in the form of thin films, foils, ribbons, wires, or powders. Recently, metallic glasses have been produced in bulk form from several families of multicomponent alloys with the use of relatively low cooling rates of :1 K s − 1 [3–5]. In these systems, the critical condition for glass formation is not the cooling rate but the level of undercooling that can be achieved in the metastable liquid state [6]. These slow cooling rates and the ability to form components such as gears and compressor blades in the viscous state should enable conventional casting methods such as injection molding and die casting to be used. Since these complex components may be produced in the near finished form, the machin* Tel.: +1 423 5744719; fax: [email protected]

+1 423 5740641; e-mail:

0921-5093/98/$19.00 © 1998 Elsevier Science S.A. All rights reserved. PII S0921-5093(98)00549-8

ing and fabrication costs may be significantly reduced. In addition, some of the unique properties of these bulk metallic glasses, including high strength (: 2 GPa), low density, high strength-to-weight ratio, excellent corrosion resistance, extremely low friction, and good wear resistance, may be used. Several of these new families of bulk metallic glasses have been investigated with the atom probe field ion microscope (APFIM) [7–13]. This high spatial resolution technique enables the presence of ultrafine scale inhomogeneities that may occur in the material due to phase separation, precipitation, clustering, and chemical short range ordering, to be detected and characterized. In addition, the compositions of the crystalline phases that are formed above the onset of crystallization temperature may be determined. The main thrust of the atom probe investigations has been the characterization of the distribution of the alloying elements, both in the as-cast condition and after various annealing treatments.

2. Experimental Several zirconium-based bulk metallic glasses including Zr65Cu27.5Al7.5, Zr60Al15Ni25, Zr55Al10Cu5Ni30, Zr41.2Ti13.8Cu12.5Ni10Be22.5 alloys were examined in the as-cast state. The ternary materials were prepared with a rapid solidification technique. The primary material used in the annealing investigation was a Zr41.2Ti13.8Cu12.5Ni10Be22.5 glass. This mate-


M.K. Miller / Materials Science and Engineering A250 (1998) 133–140

rial was examined in the as-cast state and after several different sets of isothermal heat treatments at 350, 370, 430 and 500°C. Samples of this glass were prepared from a mixture of the elements of purity ranging from 99.5 to 99.95% by induction melting and subsequent water quenching with a cooling rate of : 10 K s − 1 in a 6.35-mm diameter silica tube. Wet chemical analysis revealed that the bulk composition of the as-cast material was 43.3 at.% Zr, 13.1% Ti, 12.2% Cu, 9.9% Ni and 21.4% Be. These results indicated that some of the beryllium was lost during processing. The measured levels of hydrogen and oxygen in this alloy were B 0.1 and B0.01 at.%, respectively. Suitable blanks of these metallic glasses were cut from the interior of the bulk material to minimize the possibility of surface effects. These blanks were electropolished to form needle-shaped specimens with a standard two stage procedure [14] with a mixture of 25% perchloric acid in glacial acetic acid for the first stage and a mixture of 2% perchloric acid in 2-butoxyethanol in the second stage. Analyses were performed in the Oak Ridge National Laboratory’s energy-compensated atom probe field ion microscope. Neon was used as the imaging gas. Specimen temperatures of between 50 and 70 K were used. The pulse fraction was 20% and the pulse repetition rate was 1500 Hz.

characteristic ring structures, as shown in Fig. 2. It should be noted that the absence of the characteristic rings in the field ion image does not necessarily indicate that the alloy is fully amorphous since the image quality in high solute solid solutions is such that the characteristic rings may only be evident from the lowest index crystallographic poles. In addition, the size of a crys-

3. Methods of analysis

3.1. Field ion microscopy Field ion specimens have been successfully electropolished from many multicomponent bulk metallic glass systems including: ZrTiCuNiBe, ZrAlNi, ZrAlNiCu, ZrNbAlCuNi, ZrTiAlCuNi, TiZrCuNi, HfCuAl, PdNiP, PdNiCuP, PdNiFeP, and MgCuY [8,11,13]. A comparison of the field ion micrographs of Zr65Cu27.5Al7.5, Zr60Al15Ni25, Zr55Al10Cu5Ni30 bulk metallic glasses and a crystalline zirconium specimen is shown in Fig. 1. Field ion micrographs of most regions of these metallic glasses exhibit the random distribution of spots that is characteristic of an amorphous structure. Unfortunately, little information about the atomic arrangement of the atoms can be determined from these field ion images due to surface rearrangements that occur as a result of the high field applied to the specimen and the uncertainty of identity of the individual atoms [15]. FIM may be used to detect fine scale phase separation and crystallization. Phase separation is usually accompanied by a change in composition which will generally produce contrast in the field ion image. Small crystalline regions are evident in the field ion images, both from a change in local contrast and from their

Fig. 1. Comparison of field ion micrographs of as-cast (a) Zr65Cu27.5Al7.5, (b) Zr65Al15Ni25, (c) Zr55Al10Ni5Cu30 metallic glasses and (d) a pure Zr specimen.

M.K. Miller / Materials Science and Engineering A250 (1998) 133–140


3.2. Statistical analysis of atom probe data

Fig. 2. Field ion micrograph of an as-cast Zr55Al10Ni5Cu30 metallic glass showing a crystalline region.

talline region has to be sufficiently large that there is a high probability that a major low index pole is visible in the field ion image. A more effective method of determining if small crystalline regions are present in the material is to examine the field ion needle in the TEM. For example, the small brightly imaging regions in the field ion micrograph of a Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 bulk metallic glass in Fig. 3(a) are the same size and shape as those in the dark field transmission electron micrograph of the field ion specimen in Fig. 3(b).

Fig. 3. (a) Field ion micrograph, (b) transmission electron micrograph and (c) diffraction pattern of :2 nm crystalline regions in a Zr41.2Ti13.8Cu12.5Ni10Be22.5 metallic glass aged for 5 h at 370°C. (b) Courtesy of K.B. Alexander.

The atom probe data may be examined for phase separation, crystallization, clustering and ordering on the atom-by-atom scale and also at longer distance scales. It should be noted that decomposed regions may not be intersected in the analysis if they are present in a low number density and therefore there may be some significant analysis-to-analysis variations. Several statistical tests based on Markov chain techniques have been applied to the atom probe data to detect deviations from a random distribution of the alloying elements in these bulk metallic glasses. In these tests, the arrangement of the adjacent atoms in the atom probe ion-by-ion data chain is examined. The atom-by-atom data chain is a reconstruction of the atoms that originate in a cylinder in the specimen. It is therefore important that the smallest possible probe aperture that defines the lateral extent of this cylinder is used in order to increase the probability that adjacent atoms collected in the mass spectrometer were nearest neighbors in the specimen. In cases in which more than one ion was collected on a field evaporation pulse, the order of the ions on that pulse should be randomized to prevent bias. The atom-by-atom data chain may be examined for ABA, ABBA, ABBBA, etc., sequences, where ABA represents the solute atom of interest (i.e. B) in the data chain flanked with any other element (i.e. A). The probability of detecting a chain containing n B atoms is given by [16] P(n)= p nq 2 or D(n)= NP(n), where p is the probability of collecting a B atom, q= 1−p, and N is the number of atoms in the chain. The significance of the experimental value is given by (Dap(n)−D(n))/s, where s is taken as Np nq 2. The Johnson and Klotz ordering parameter method [17] may also be applied to the atom probe data. The Johnson and Klotz ordering parameter, u, is determined from the number of AB and BB pairs in the data chain [17]. If the ordering parameter is statistically greater than 1, the data contains more BB atom pairs than expected in a random solid solution. This result suggests that the material contains solute clusters. If the ordering parameter is statistically less than 1, the number of BB atoms in the data chain is significantly less and the number of AB pairs is greater than that expected in a random alloy. This result suggests the presence of chemical short range order. The significance of the result is given by (u− 1)/s, where the s is the S.E. Significances greater than 2 or less than −2 indicate non-random behavior. Since these statistical techniques examine the atomby-atom data, there are several other factors that need to be considered to evaluate their reliability. These techniques assume that adjacent atoms in the data chain are nearest atoms in the specimen. However, the


M.K. Miller / Materials Science and Engineering A250 (1998) 133–140

less than ideal detection efficiency of the single atom detector (62% in this study), the finite width of the probe aperture, the problem with multiple ions collected on one field evaporation pulse, and the ambiguity introduced by the overlap of some of isotopes of the alloying components (e.g. 90Zr3 + and 60Ni2 + ) should be considered. Larger scale composition fluctuations are characterized by dividing the atom-by-atom data into small blocks and calculating the composition of each block, thereby creating a composition profile. These composition profiles can be visually examined for concentration variations or used as the input to a variety of statistical tests. The simplest of these statistical tests is to compare the frequency of the concentrations with those derived from a binomial distribution for the same average concentration. Phase separation is usually evident by a broadening or a narrowing of the frequency distribution and can be quantified with the standard x 2 test [15]. However, this test is relatively insensitive to low volume fraction phase separation. Co- and anti-segregation of the different solutes can be tested for by constructing pairwise contingency tables from the composition profiles [15].

4. Characterization of bulk metallic glasses

4.1. As-cast bulk metallic glasses A number of bulk metallic glasses have been examined in the as-cast condition. The aim of these characterizations was to determine the distribution of the solutes and to detect any fine scale decomposition by applying the techniques described previously. Most of the specimens examined by FIM exhibited an amorphous structure, as shown in Fig. 1. However, some crystalline regions were found, Fig. 2. The results of the Markov ABA chain and the Johnson and Klotz analyses from ternary and quaternary Zr60Al15Ni25, metallic glasses, Zr65Cu27.5Al7.5, Zr55Al10Cu5Ni30, are summarized in Fig. 4 and Table 1, respectively. Results from some other related materials have been presented previously [11]. The data used in these statistical analyses were scanned to ensure that no second phases were present. With the exception of the aluminum in both nickel-containing materials, most of these results do not exhibit any evidence of clustering or short range order. The results from the Zr60Al15Ni25 and Zr55Al10Cu5Ni30 glasses both show large negative significances for aluminum – aluminum sequences in both the Markov chain data and the Johnson and Klotz ordering parameters indicating that aluminum prefers non-aluminum neighbors. Similar results were found in a previous investigation of two other aluand minum-containing Zr52.5Ti5Al10Cu17.9Ni14.6 Zr57Nb5Al10Cu15.4Ni12.6 metallic glasses.

Fig. 4. Summary of the Markov chain results from as-cast (a) Zr65Al15Ni25, (b) Zr65Cu27.5Al7.5 and (c) Zr55Al10Ni5Cu30 metallic glasses. Missing data points indicate no chains of that length were encountered.

The results from these zirconium-based alloys are distinctly different from those obtained from an as-cast Pd40Ni40P20 alloy [13]. In the Pd40Ni40P20 alloy, statistically significant evidence of short range order for all elements was found. Representative examples of composition profiles through these alloys are shown in Fig. 5. Some local fluctuations are apparent. Although a comparison of the frequency distribution of the experimental data and that from a binomial distribution with the same mean concentration, shown in Fig. 6, revealed no significant

M.K. Miller / Materials Science and Engineering A250 (1998) 133–140


deviations from a random solid solution, the number of experimental observations in the tails of the distributions was often slightly higher than that observed in the binomial distribution.

4.2. Comparison of rapidly solidified and bulk Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 metallic glass Due to the large differences in cooling rates, the microstructure and solute distributions between rapidly solidified and slow-cooled bulk materials may be significantly different. Therefore, a microstructural comparison has been performed between bulk and rapidly solidified ribbon forms of a commercially available Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 metallic glass. The material used in this study was a zirconiumbased metallic glass with a nominal composition of Zr41.2Ti13.8Cu12.5Ni10.0Be22.5. The specimens were fabricated from blanks cut from the interior of 15 mm high by 30 mm diameter button or the central portion of the : 10-mm-thick ribbon. The materials were characterized in the as-cast condition. Field ion images of both materials exhibited the random distribution of dots characteristic of an amorphous alloy, as shown in Fig. 7. No evidence of any crystalline regions was apparent in the field ion images. Atom probe analyses were performed and the resulting data was examined for clustering and ordering of each of the solutes with statistical methods described above. The statistical analysis of the atom probe data is summarized in Fig. 8 and Table 2. The results for the bulk material indicate little deviation from a random solute distribution. In contrast, the results from the rapidly solidified material reveal some significant deviations from a random solute distribution. The rapidly solidified material was Table 1 Johnson and Klotz results for the as-cast Zr60Al15Ni25, Zr55Al10Cu5Ni30 metallic glasses

Zr60Al15Ni25 u s Significance Zr65Cu27.5Al7.5 u s Significance Zr55Al10Cu5Ni30 u s Significance



0.993 0.005 −1.36

0.899 0.051 −1.98

1.011 0.005 2.18 0.991 0.014 −0.62

1.143 0.82 0.60




0.734 0.066 −4.06 0.997 0.034 −0.09

0.934 0.170 −0.39

1.009 0.030 0.28

0.823 0.072 −2.47

Fig. 5. Composition profiles from as-cast (a) Zr65Al15Ni25, (b) Zr65Cu27.5Al7.5 and (c) Zr55Al10Ni5Cu30 metallic glasses.

found to be significantly higher in dissolved oxygen (1.19 0.08 at.% O) compared to the bulk material (0.39 0.04 at.% O). This higher oxygen level may lead to heterogeneous nucleation during annealing treatments.


M.K. Miller / Materials Science and Engineering A250 (1998) 133–140

4.3. Annealed bulk metallic glasses In a continuation of a previous study [11], a series of different annealing treatments above and below the crystallization temperature in a Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 glass have been characterized. The aim of this study was to determine the decomposition products that form during ageing. It should be noted that the mechanical properties of this material deteriorate rapidly as the glass crystallizes and this has a significant impact on the success of atom probe experiments. Brightly imaging crystalline regions that exhibited well-defined poles were evident in field ion images of material annealed for 11 min at 430°C as shown in Fig. 9. Atom probe selected area analysis of these crystalline regions revealed that they were Be-rich with a composition of 63.292.0 at.% Be, 35.392.0% Zr and 1.59 0.5% Ti. This composition is consistent with the Be2Zr (hexagonal C32) phase. Fine scale decomposition was also evident in this material aged for 1 h at 500°C, as shown in the composition profile in Fig. 10. An ultrafine titaniumand nickel-enriched and beryllium-depleted region is evident. The composition of this region was determined to be 41.4 9 3.7% Zr, 30.993.5% Ti, 8.892.1% Cu, 17.19 2.8% Ni and 2.991.3% Be. This titanium level represents an enrichment of 220% over the bulk level. There was also some indication that the local beryllium content in the surrounding region was increased, indicating that the beryllium was rejected from the titanium-enriched region. A similar titanium-enriched precipitate is evident in the composition profile, Fig. 11, from a specimen aged for 10 h at 350°C. The average composition of this precipitate was determined to be

Fig. 6. Comparison of the experimental and binomial frequency distributions for the Zr65Al15Ni25 bulk metallic glass.

Fig. 7. Field ion micrographs of (a) as-cast bulk and (b) RS Zr41.2Ti13.8Cu12.5Ni10Be22.5 metallic glasses.

47.99 2.7% Zr, 29.89 2.5% Ti, 5.49 1.2% Ni, 3.09 0.9% Cu and 13.99 1.9% Be. Although the titanium and zirconium levels are approximately the same in both precipitates, the levels of copper, nickel and beryllium vary significantly. The typical size of these regions was estimated to be : 2–4 nm from their extents in the field ion micrographs. Although these regions are likely to be crystalline from the annealing treatment, this could not be conclusively confirmed from the field ion micrographs due to their small size and poor imaging behavior. In addition, an extremely small beryllium-enriched region is also evident in Fig. 11. These highberyllium regions may be the precursors to the Be2Zr precipitates. Zirconium- and titanium-enriched regions were also observed in material aged for 5 h at 370°C. The composition of one such crystalline region was determined to be 75.99 1.9% Zr, 21.69 1.9% Ti, 0.49 0.3% Cu, 2.19 0.6% Be. In addition to these zirconium-based materials, several studies have been performed on the Pd–Ni–P system and its derivatives [7,8,13]. Fine-scale phosphorus enrichments were observed by Oehring and Hassen

M.K. Miller / Materials Science and Engineering A250 (1998) 133–140


Fig. 9. Field ion micrograph of Zr41.2Ti13.8Cu12.5Ni10Be22.5 bulk metallic glass aged for 11 min at 430°C showing a Be2Zr precipitate.

5. Summary

Fig. 8. Markov chain results of (a) as-cast bulk and (b) RS Zr41.2Ti13.8Cu12.5Ni10Be22.5 metallic glass.

in Pd35Ni45P20 metallic glass that was annealed at 330°C for 2–6 h [7]. Read et al. characterized Pd46Ni36P18 and Pd48Ni32P20 glasses annealed in the supercooled liquid region and indicated that no compositionally modulated microstructures were formed prior to the onset of crystallization [8]. Short range ordering was measured in an as-cast Pd40Ni40P20 glass by Miller at al. [13]. Phase separation at the nanometer level was observed in glassy samples after annealing above the glass-transition temperature and crystallization was found to proceed by phase separation into three distinct crystalline phases [13].

This study of bulk metallic glasses has shown that the as-cast microstructure and the decomposition that occurs during annealing may be characterized with APFIM. Evidence of short range order for aluminum has been detected in as-cast nickel-containing Zr60Al15Ni25 and Zr55Al10Cu5Ni30 glasses. Three different phases have been detected and analyzed in annealed Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 glass.

Acknowledgements The author would like to thank Dr C.T. Liu, D.S. Easton, K.B. Alexander and K.F. Russell of Oak Ridge National Laboratory, Dr. R.B. Schwarz of Los Alamos National Laboratory, and Professor W.L. Johnson and Dr R. Busch of the California Institute of Technology for providing the materials used in this study and for many helpful discussions. This research was sponsored by the Division of Materials Sciences, US Department of Energy, under contract DE-AC0596OR22464 with Lockheed Martin Energy Research

Table 2 Johnson and Klotz results for the as-cast bulk and rapidly solidified (RS) Zr41.2Ti13.8Cu12.5Ni10.0Be22.5 metallic glasses Zr


Bulk u s Significance

0.995 0.009 −0.57

0.955 0.050 −0.89

RS u s Significance

0.984 0.007 −2.45

0.892 0.034 −3.17




1.035 0.057 0.61

0.792 0.083 −2.51

0.997 0.034 −0.10

0.912 0.033 −2.68

1.100 0.071 1.40

0.979 0.032 −0.67


M.K. Miller / Materials Science and Engineering A250 (1998) 133–140

Fig. 10. Composition profile through a specimen of Zr41.2Ti13.8Cu12.5Ni10Be22.5 bulk metallic glass aged for 1 h at 500°C showing a Ti-enriched and Be-depleted region.

Corp. This research was conducted utilizing the Shared Research Equipment (SHaRE) User Program facilities at Oak Ridge National Laboratory.

References [1] P. Duwez, Metallic Glasses, ASM, Metals Park, OH, 1978, p. iii. [2] P Duwez, R.H. Willens, W. Klement, Appl. Phys. Lett. 31 (1960) 1136. [3] A. Inoue, T. Zhang, T. Masumoto, Mater. Trans. Jpn. Inst. Met. 31 (1991) 425. [4] T. Zhang, A. Inoue, T. Masumoto, Mater. Trans. Jpn. Inst. Met. 32 (1991) 1005. [5] A. Peker, W.L. Johnson, Appl. Phys. Lett. 63 (1993) 2342. [6] J.H. Perepezko, J.S. Smith, J. Non-Cryst. Solids 44 (1981) 65. [7] M. Oehring, P. Hassen, J. Phys. 47 (C7) (1986) 275. [8] H.G. Read, K. Hono, A.P. Tsai, A. Inoue, J. Phys. IV 6 (C5) (1996) 211.


Fig. 11. Composition profile through a specimen of Zr41.2Ti13.8Cu12.5Ni10Be22.5 bulk metallic glass aged for 10 h at 350°C showing Be-enriched and Ti-enriched regions.

[9] R. Busch, S. Schneider, A. Peker, W.L. Johnson, Appl. Phys. Lett. 67 (1995) 1544. [10] R. Busch, Y.J. Kim, S. Schneider, W.L. Johnson, Mater. Sci. Forum 225 – 227 (1996) 77. [11] M.K. Miller, K.F. Russell, P.M. Martin, R. Busch, W.L. Johnson, J. Phys. IV 6 (C5) (1996) 217. [12] M.P. Macht, N. Wanderka, A. Wiedenmann, H. Wollenberger, Q. Wei, H.J. Fecht, S.G. Klose, Mater. Sci. Forum 225–227 (1996) 65. [13] M.K. Miller, D.J. Larson, R.B. Schwarz, Y. He, Mater. Sci. Eng. A, 250 (1998) 138. [14] M.K. Miller, G.D.W. Smith, Atom Probe Microanalysis: Principles and Applications to Materials Problems, MRS, Pittsburgh, PA, 1989. [15] M.K. Miller, A. Cerezo, M.G. Hetherington, G.D.W. Smith, Atom Probe Field Ion Microscopy, Oxford University Press, Oxford, UK, 1996. [16] T.T. Tsong, S.B. McLane, M. Ahmad, C.S. Wu, J. Appl. Phys. 53 (1982) 4180. [17] C.A. Johnson, J.H. Klotz, Technometrics 16 (1974) 483.