Thermal stability of nanocrystalline Pd81Zr19

Thermal stability of nanocrystalline Pd81Zr19

Available online at Acta Materialia 58 (2010) 4292–4297 Thermal stability of nanocrystalline P...

599KB Sizes 0 Downloads 52 Views

Available online at

Acta Materialia 58 (2010) 4292–4297

Thermal stability of nanocrystalline Pd81Zr19 Brian K. VanLeeuwen a,*, Kristopher A. Darling b, Carl C. Koch a, Ron O. Scattergood a, Brady G. Butler c a

Room 3000, Department of Materials Science and Engineering, North Carolina State University, 911 Partners Way, Raleigh, NC 27695-7907, USA b Battelle, RDRL-WMM-B, Aberdeen Proving Ground, MD 21005-5069, USA c Booz Allen Hamilton, RDRL-WMM-B, Aberdeen Proving Ground, MD 21005-5069, USA Received 16 March 2010; received in revised form 5 April 2010; accepted 10 April 2010 Available online 10 May 2010

Abstract Grain growth stability in mechanically alloyed nanocrystalline Pd81Zr19 was investigated. Previous research suggested that the alloy is thermodynamically stable to very high temperatures. When X-ray diffraction (XRD) is used to estimate the grain size of annealed samples the alloy appears to have remarkable resistance to growth. Microscopy done here on the same alloy indicated that the XRD estimates are not accurate for samples annealed above 600 °C. It appears that when this alloy is annealed at high temperatures XRD peak broadening is retained for reasons that are unrelated to the grain size. The alloy still has much improved grain growth stability compared with pure Pd, but not as significant as suggested by the XRD results. A similar phenomenon was observed in Fe–Zr alloys. Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Nanocrystalline materials; Grain growth; Grain boundary energy; Grain boundary segregation; Solute segregation

1. Introduction Nanocrystalline materials (grain size <100 nm) have unique properties compared with their coarse grained counterparts. However, the nanoscale grain size is unstable in many metals and alloys [1–3]. It has been observed that some alloys retain a nanoscale grain size at high temperatures [1,4– 18]. The stabilization of nanoscale grain sizes in these alloys has been attributed to kinetic or thermodynamic stabilization effects. Kinetic stabilization of grain size slows the rate at which grain growth progresses [8,9]. For example, reducing thermally activated grain boundary mobility by solute drag or precipitated secondary phases would produce kinetic stabilization. Large kinetic barriers can lead to a near zero rate of grain growth at low temperatures. Kinetic stabiliza-


Corresponding author. E-mail addresses: [email protected], [email protected] (B.K. VanLeeuwen).

tion is time and temperature dependant; if a kinetically stable nanocrystalline alloy is heated during processing or service, rapid grain growth has been shown to occur at higher homologous temperatures. Segregation of non-equilibrium solute atoms to grain boundaries can lower the effective grain boundary free energy and thus reduce the driving force for grain growth [1,5–8,10–12]. If the grain boundary energy is zero, then there exists a metastable equilibrium grain size that is not time dependent, and only weakly temperature dependant. Strategies to create nanocrystalline alloys with thermodynamically stable grain sizes less than 100 nm by the addition of solutes which segregate to the grain boundaries have been reported [4–7,10–17,19,20]. The effective grain boundary energy, c, is: c ¼ c0 þ DGseg C


where c0 is the intrinsic (non-segregated) grain boundary energy and C is the excess amount of segregated solute per unit area on the grain boundaries. If the entropy contribution to the free energy change DGseg for segregation

1359-6454/$36.00 Ó 2010 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2010.04.023

B.K. VanLeeuwen et al. / Acta Materialia 58 (2010) 4292–4297

is neglected, then this is equal to the change in enthalpy, i.e. DGseg = DHseg. The enthalpy change is often assumed to be the elastic energy change associated with segregation of non-equilibrium solutes. If solute segregation leads to c = 0, a metastable equilibrium state is reached and the system would increase its free energy by increasing its average grain size. Solute segregation can therefore produce a grain size that is thermodynamically stable. In a series of papers [4–6] several alloys of palladium and zirconium, including Pd81Zr19, were observed to be resistant to grain growth at very high temperatures. This resistance to grain growth was attributed to thermodynamic stabilization by Zr segregation to the grain boundaries. The enthalpy of segregation of Zr in Pd is 31 kJ mol–1 based on the elastic size misfit strain energy of zirconium solutes in pure palladium [21]. This assumes that the elastic size misfit strain energy is the only significant component of the enthalpy of segregation. The grain sizes reported in previous studies of a Pd81Zr19 alloy were estimated from XRD line broadening [4–6]. A single transmission electron microscopy (TEM) micrograph was presented in Krill et al. [4], which indicated a nanoscale grain size. However, no grain size data was obtained using TEM. It is well known that XRD estimates of grain size can be inaccurate when grains grow and so alternative methods of determining grain size must be used to accurately measure grain size after growth. The current work presents a more complete investigation of Pd81Zr19 using microscopy techniques to characterize the grain size. The motivation for re-examining the Pd81Zr19 alloy was due in part to recent results for FexZr(1–x), in which the XRD grain size after high temperature annealing was not confirmed by grain size results obtained using microscopy [1,7,13]. 2. Experimental procedure The procedure described by Krill et al. for milling Pd81Zr19 was initially attempted [4,5], but a satisfactory powder yield could not be obtained. Powder milled under the conditions described in that study cold welded to the vial and the balls to such an extent that the powder yield was minimal. The small amount of powder that was created in this way had significant iron contamination (10– 20 at.% Fe) from the vial and milling media. After a number of attempts to produce acceptable powder by this technique, the procedure was modified. Milling was performed at cryogenic temperatures instead of room temperature to lower the ductility of the powder, to increase the yield to practical amounts. Instead of using the vials as received, the vials and the milling media were pre-coated with the alloy to minimize iron contamination. This modified procedure produced nanocrystalline Pd81Zr19 powders with very little iron contamination (<1 at.%). The zirconium powder (–325 mesh, 99.7% pure) was purchased from Alfa Aesar. The palladium was purchased in 99.95% pure bullions from Northwest Territorial Mint


and was reduced to small shavings using an endmill. The composition selected was 19 at.% Zr, because it showed nearly as much stability as the 20 at.% Zr alloy in previous studies [4,5]. Beyond 20 at.% Zr secondary phases were observed, therefore 19 at.% Zr was used to leave room for a margin of error. The milling media used were obtained from SPEX: a ½ in. stainless steel ball and four 1= 4 in. balls. All milling was performed in an ultra high purity argon (<1.0 ppm O2) atmosphere. To coat the vials and the milling media, 7.5 g of the powder mixture was milled at room temperature in a SPEX 8000 High Energy Ball Mill for 24 h [22]. After the vials were coated, an additional 7.5 g of the powder mixture was added to each vial to generate the yield. Cryomilling was carried out using a modified SPEX 8000 mill which allows liquid nitrogen to flow around the vial during the milling process. Cryomilling the powder mixture for 10 h yielded alloyed powder with an as milled grain size of 6.3 nm, estimated by XRD using the Scherrer method. The yield did not contain enough contamination to resolve any additional peaks by scanning electron microscopy energy dispersive spectroscopy (SEMEDS). The as milled powder was compacted into 100 mg, 2.9 mm diameter compacts under an applied uniaxial pressure of 3.0 GPa in tungsten carbide dies. Compacted samples for hardness, microscopy and XRD were annealed under vacuum (<106 torr) using a standard tube furnace set-up with a diffusion pump. Zirconium metal was used as a getter during the annealing process to provide a low oxygen partial pressure. Annealing time was 1 h for all samples. When the annealing temperature exceeded 950 °C the compacted powder samples swelled and an increase in porosity was observed. Further investigation showed that this was due to the expansion of argon trapped in the metal during the cryomilling process. The pores do not affect the grain growth because they have an average spacing of the order of micrometers, compared with the nanometer grain sizes. A Rigaku DMax X-ray diffractometer with a Cu Ka source was used for XRD. The contributions of Cu Ka2 were subtracted out using XRD software and the instrumental broadening was removed. The Scherrer method for estimating grain size from XRD peak width was used. Conventional polishing and optical microscopy techniques were used to prepare and image the surface of the compacts. Vickers hardness testing was performed using a MicroMet II fitted with a higher magnification optic to allow more precise indentation measurement (allowing 1000 total magnification) [3]. 50 mg loads and 15 s load times were used. The samples were imaged using focused ion beam channeling contrast imaging (FIB-CCI). This technique can accurately resolve grain sizes greater than 100 nm. For smaller grain sizes the ion beam was used to prepare electron transparent samples for TEM. The TEM instrument used was a JEOL JEM 2100F with an accelerating voltage of 200 keV.


B.K. VanLeeuwen et al. / Acta Materialia 58 (2010) 4292–4297

3. Results and discussion The composition of the alloy was determined to be 81 at.% palladium and 19 at.% zirconium, with no detectable contamination, using SEM-EDS, transmission electron microscopy energy dispersive spectroscopy (TEM-EDS) and XRD. SEM-EDS, TEM-EDS and XRD also confirmed that the elemental powders had been fully alloyed and that no secondary phases were present in any of samples except for a sample which was annealed at 1250 °C. The 1250 °C sample had a small amount of the intermetallic phase Pd3Zr present. The Pd–Zr system has an equilibrium solubility limit of 17 at.% Zr at 1580 °C and at lower temperatures the equilibrium solubility is reduced further to about 12 at.% Zr [23,24]. It is suggested that kinetic limitations prevent the equilibrium Pd3Zr phase from forming at lower temperatures [4]. The formation of intermetallic Pd3Zr in the Pd81Zr19 alloy lowers the amount of zirconium solute available in the solid solution phase and is presumed to be detrimental to the thermal stability of the grain size. The grain sizes of annealed Pd81Zr19 samples estimated by XRD line broadening analysis were similar to those reported in the previous research [4,5]. Fig. 1 compares the grain size estimates by XRD in the present study with those of Krill et al. [4,5]. The estimates shown in Fig. 1 were made using the Scherrer method, although Wilson and Williamson–Hall approaches gave similar numbers for grain size. The strain estimated by the Williamson–Hall method decreased throughout the annealing temperature range and was undetectable for temperatures above 700 °C. The lattice parameters of each annealed sample were estimated by Cohen’s method [29] and the results were similar to those published by Krill et al. [4]. Fig. 2 shows the indentation hardness of the alloy after being annealed at each temperature. A noticeable feature of this plot is the large drop in hardness between 950 °C and 1000 °C, from 6.3 to 1.6 GPa. This dramatic decrease in hardness corresponded with a relatively small increase in the XRD grain size from 49 to 60 nm over the same temperature range. This small increase in grain size cannot

Fig. 1. Grain sizes of annealed Pd81Zr19 samples estimated from XRD line broadening analysis.

Fig. 2. Vickers hardness versus annealing temperature of Pd81Zr19. Error bars are omitted because they are insignificant relative to the magnitude (all were <0.1 GPa).

Fig. 3. Comparison of grain size estimates from XRD line broadening analysis to grain sizes observed by microscopy. Pure palladium data is from Wurschum et al. [28].

account for the drop in hardness using a Hall–Petch relationship [3]. Although the grain sizes estimated using XRD suggest that the alloy was nanocrystalline throughout the temperature range shown in Fig. 1, when these samples were analyzed using FIB-CCI or TEM it was apparent that the true grain size deviated significantly from the XRD estimates. This deviation began between 600 °C and 700 °C. The grain size observed by microscopy is plotted against the XRD grain size in Fig. 3. Figs. 4–6 show the microstructures of the samples annealed at 700 °C, 950 °C and 1000 °C, respectively. The microstructures were composed of high angle equiaxed grains. The average grain size at 700 °C was determined to be 52 nm and at 950 °C was 260 nm. Coarsening to the micron scale occurred when annealing the sample at 1000 °C. In all samples a significant fraction of the microstructure was twinned. Along with the increase in grain size, the spacing of the twin boundaries was also observed to increase with annealing temperature. The elastic enthalpy of zirconium solutes in pure palladium is 31 kJ mol–1 [21]. Enthalpy of segregation can be assumed to be approximately equal to the elastic enthalpy if it is assumed that the alloy is an ideal solution, i.e. the

B.K. VanLeeuwen et al. / Acta Materialia 58 (2010) 4292–4297


Fig. 6. FIB-CCI micrograph of Pd81Zr19 annealed at 1000 °C for 1 h.

following result [19] based on a surface segregation model discussed by Wynblatt and Ku [20,30]: DH seg  1=2ðcZr rZr  cPd rPd Þ  XPd Zr ½ðX Zr  X Zr Þ  1=2ðX Zr  1=2Þ þ 5=12ðX Zr  1=2Þ  DH elastic Fig. 4. TEM micrographs of annealed Pd81Zr19 samples at 700 °C for 1 h. (top) low magnification; (bottom) high magnification.

enthalpy of mixing is zero [25]. The earlier results on Pdx Zr (1–x) alloys made this assumption [4–6]. However, the Pd–Zr system has a large negative chemical bond–interaction parameter XPdZr = –364 kJ mol–1 [21], and therefore a negative enthalpy of mixing. Models to estimate enthalpy of segregation must therefore include the chemical effect in addition to the elastic strain energy. We derived the


cZr is the grain boundary energy of zirconium (630 mJ m–2) [26], cPd is the grain boundary energy of palladium (680 mJ m–2) [26], rZr is the area per mole of zirconium surface atoms (4.91  104 m2 mol–1) [21], rPd is the area per mole of palladium surface atoms (3.63  104 m2 mol–1) [21], X Zr is atomic fraction of solute on the grain boundary (X Zr  X Zr þ 1=2CrPd81Zr19 ), XZr is the bulk solute concentration (19 at.% Zr) and DHelastic = 31 kJ mol–1 is the elastic enthalpy of zirconium in palladium. High angle grain boundary energies were estimated as a third of surface energies [25,26]. Using Eqs. (1) and (2), the effective grain

Fig. 5. TEM micrographs of annealed Pd81Zr19 samples at 950 °C for 1 h. (left) low magnification; (right) high magnification.


B.K. VanLeeuwen et al. / Acta Materialia 58 (2010) 4292–4297

Fig. 7. Pd81Zr19 effective grain boundary energy versus grain boundary solute excess. Fig. 9. High resolution TEM micrograph of nanotwins (spacing 1 nm) in Pd81Zr19 sample annealed at 700 °C.

grains indicated that the boundary spacing of these twins was of the order of 1 nm, as seen from the image in Fig. 9. These nanotwins can form during annealing and could therefore contribute to increased hardness with increased annealing temperature. 4. Summary and conclusions

Fig. 8. TEM micrograph of twinned grains in Pd81Zr19 sample annealed at 700 °C. (inset) high magnification of nanotwins.

boundary energy (c) can be plotted as a function of segregated solute on the grain boundary (C), as shown in Fig. 7 for Pd81Zr19. The grain size of an alloy is thermodynamically stable when the effective grain boundary energy (c) is equal to zero. This plot indicates that c for Pd81Zr19 cannot be zero for any value of C, i.e. the thermodynamic mechanism for grain size stabilization is not effective when the chemical interaction is taken into account. In contrast, for the ideal solution model shown in Fig. 7 the segregation enthalpy is equal to the elastic strain energy enthalpy. This predicts that the Pd81Zr19 alloy will have zero effective grain boundary energy when the solute excess is about 2.2  105 mol Zr m–2 or approximately a monolayer of Zr on the grain boundary. The hardness as plotted in Fig. 2 shows an increase from 5.6 GPa in the as milled condition to 9.4 GPa in the Pd81Zr19 sample annealed at 600 °C. Over the same range the average grain size estimated by XRD, which is reliable at these sizes, increased from 6.3 to 14.5 nm. Figs. 8 and 9 are TEM micrographs of the Pd81Zr19 sample that was annealed at 700 °C. Fig. 8 shows the largest grain observed in the prepared foil approximately 200  400 nm. The inset of this image shows that this grain was heavily twinned. Higher magnification imaging of these heavily twinned

Pure nanocrystalline palladium will have significant grain growth at low temperature [27,28] and these results confirm that the addition of zirconium solute can stabilize the nanocrystalline microstructure. XRD peak broadening analysis indicated a nanocrystalline grain size throughout the entire annealing temperature range, in agreement with the findings of previous studies [4,5]. However, the microscopy observations showed that grain sizes estimated by XRD are not reliable for higher annealing temperatures and the alloy does not remain nanocrystalline when annealed above 700 °C. It must therefore be concluded that there is a significant source of XRD peak broadening other than the grain size in annealed Pd81Zr19. The observed nanotwins in larger grains (>100 nm) could be this source. Based on the predictions of a segregation model that includes chemical interaction effects in a Pd81Zr19 alloy, thermodynamic stabilization of a nanocrystalline grain size is not expected. It therefore appears that the stabilization observed, with respect to pure Pd, up to about 700 °C is due to a kinetic mechanism associated with the Zr solute. Acknowledgements The authors thank the National Science Foundation for support (DMR Grant No. 0504286). The authors also thank Jonathon Semones for his assistance with the experimental work. References [1] Darling K. Thermal stability of nanocrystalline microstructures. PhD thesis, North Carolina State University, 2009.

B.K. VanLeeuwen et al. / Acta Materialia 58 (2010) 4292–4297 [2] Koch CC. Scripta Mater 2003;49:657. [3] Dieter G. Mechanical metallurgy. New York: McGraw-Hill; 1986. [4] Krill CE, Ehrhardt H, Birringer R. Z Metallkd 2005;96:1134– 41. [5] Ehrhardt H. Thermische stabilitat nanostrukturierter materialien. Aachen, Germany: Shaker; 1998. [6] Krill CE, Klein R, Janes S, Birringer R. Mater Sci Forum 1995;179– 181:443–8. [7] Darling KA, VanLeeuwen BK, Koch CC, Scattergood RO. Mater Sci Eng A (in press), doi:10.1016/j.msea.2010.02.043. [8] Chen Z, Liu F, Wang H, Yang W, Yang G, Zhou Y. Acta Mater 2009;57:1466–75. [9] Li J, Wang J, Yang G. Scripta Mater 2009;60:945–8. [10] Detor AJ, Schuh CA. Acta Mater 2007;55:4221. [11] Millett PC, Selvam RP, Saxena A. Acta Mater 2007;55:2329–36. [12] Weissmuller J. J Mater Res 1994;9:4. [13] Darling KA, Chan RN, Wong PZ, Semones JE, Scattergood RO, Koch CC. Scripta Mater 2008;59:530. [14] Beke DL, Cserha´ti C, Szabo´ IA. J Appl Phys 2004;95:4996. [15] Kirchheim R. Acta Mater 2002;50:413. [16] Liu F, Kirchheim R. Scripta Mater 2004;51:521–5.


[17] Weissmuller J, Krauss W, Haubold T, Birringer R, Gleiter H. Nanostruct Mater 1992;1:439–47. [18] Detor A, Schuh C. J Mater Res 2007;22:3233–48. [19] VanLeeuwen BK, Darling KA, Koch CC, Scattergood RO. Scripta Mater (to be submitted for publication). [20] Wynblatt P, Chatain D. Metall Mater Trans A. Phys Metall Mater Sci 2006;37:2595. [21] Boer FRD. Cohesion in metals: transition metal alloys. Amsterdam: North-Holland; 1988. [22] Suryanarayana C. Progr Mater Sci 2001;46:1. [23] Waterstrat RM, Shapiro A, Jeremie A. J Alloys Compd 1999;290:63–70. [24] Guo C, Du Z, Li C. Calphad 2006;30:482. [25] Porter D, Easterling KE, Sherif MY. Phase transformations in metals and alloys. Boca Raton (FL): CRC Press; 2009. [26] Vitos L, Ruban AV, Skriver HL, Kollar J. Surf Sci 1998;411:186–202. [27] Weissmuller J, Loffler J, Kleber M. Nanostruct Mater 1995;6:105–14. [28] Wurschum R, Reimann K, Grub S, Kubler A, Scharwaechter P, Frank W, et al. Philos Mag B 1997;76:407–17. [29] Razik NA. J Appl Phys A 1985;37:187–9. [30] Wynblatt P, Ku RC. Surf Sci 1977;65:511–31.