The thermal stability of nanocrystalline copper cryogenically milled with tungsten

The thermal stability of nanocrystalline copper cryogenically milled with tungsten

Materials Science & Engineering A 558 (2012) 226–233 Contents lists available at SciVerse ScienceDirect Materials Science & Engineering A journal ho...

2MB Sizes 0 Downloads 14 Views

Materials Science & Engineering A 558 (2012) 226–233

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A journal homepage:

The thermal stability of nanocrystalline copper cryogenically milled with tungsten Mark A. Atwater a,b,n, Debdas Roy a,c, Kristopher A. Darling b, Brady G. Butler b, Ronald O. Scattergood a, Carl C. Koch a a

Department of Materials Science and Engineering, North Carolina State University, 911 Partner’s Way, EB I, Room 3002 Raleigh, NC 27606, USA US Army Research Laboratory, Weapons and Materials Research Directorate, RDRL-WMM-F, Aberdeen Proving Ground, MD 21005-5069, USA c Materials and Metallurgical Engineering Department, NIFFT, Ranchi 834003, India b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 19 July 2012 Accepted 31 July 2012 Available online 7 August 2012

Copper (Cu) was cryogenically milled with tungsten (W) in a high-energy ball mill. The process created W particles dispersed in a nanocrystalline Cu matrix. These ‘‘alloys’’ were then annealed to a maximum temperature of 800 1C. The addition of W stabilized the Cu at  40 nm during annealing to 400 1C for a 1 at% W composition and to 600 1C for 10 at% W. As evidenced through hardness measurement, the W provided a significant increase in strength over pure Cu, and the 10 at% W material maintained a 2.6 GPa hardness after annealing at 800 1C. The stabilization and strengthening mechanisms are compared against theoretical prediction and found to be in good agreement. Although the strength and stability are significantly improved over pure Cu, the maximum benefit was hindered by an extremely broad W particle size distribution ( 5–5000 nm). For the 10 at% W alloy, only half of the added W was reduced to nanoscale where kinetic pinning and strengthening become most effective. & 2012 Elsevier B.V. All rights reserved.

Keywords: Nanocrystalline Copper Tungsten Stabilization Grain boundary pinning Strengthening

1. Introduction The stabilization of nanocrystalline metals can be carried out in two principal manners: the reduction of grain boundary energy or grain boundary mobility. Eliminating the grain boundary energy, and therefore the driving force for grain growth, is considered the thermodynamic stabilization route. This has been accomplished experimentally by adding a solute which lowers grain boundary energy upon segregation (e.g. [1–4]). The second method is a form of kinetic stabilization that reduces the grain boundary mobility by methods such as solute drag and/or secondary particle pinning [5]. Because grain boundary mobility has an Arrhenius temperature dependence, the kinetic approach is thought to be less effective than the thermodynamic one in stabilizing nanograins [6]. Although solute drag has been found to be unable to completely prevent grain growth at elevated temperatures [7,8], secondary particle pinning, also known as Zener pinning [9], has been successful in increasing thermal stability to moderately high temperatures (e.g. [10–12]). To be successful at elevated temperatures, destabilization via particle coarsening [13] must be suppressed. n Corresponding author at: Department of Materials Science and Engineering, North Carolina State University, 911 Partner’s Way, EB I, Room 3002 Raleigh, NC 27606, USA. Tel.: þ1 505 918 2893; fax: þ1 919 515 7724. E-mail address: [email protected] (M.A. Atwater).

0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.

The equilibrium binary phase diagram of the Cu–W system [14] reveals no solid solubility and no compound formation on heating. This makes the Cu–W system an ideal system for studying the effects of Zener pinning as the systemic considerations are simplified. That is, the pinning phase (pure W) will not form intermetallic compounds with the Cu parent phase, thereby simplifying some assumptions about the particle matrix interface such as varying degrees of coherency. The high temperature kinetic pinning is bolstered by the negligible diffusivity of W in Cu, which suppresses coarsening of the secondary particles. The ability of W to stabilize nanocrystalline Cu by this kinetic mechanism has been demonstrated in deposited films  0.2– 1.2 mm thick [15,16] where a 10 at% W alloy was found to outperform Mo and Nb alloys of the same concentration. In these studies, Cu and W were co-deposited to create a solid solution. Upon annealing, the W particles which formed remained smaller than 10 nm up to 900 1C, and served to stabilize the Cu grain size below 30 nm to the same temperature. Nb and Mo coarsened at lower temperatures ( 400 1C and 600 1C, respectively), and as a result, grain growth proceeded more rapidly, though the exact grain sizes were not reported much above these coarsening onset points. While mechanical alloying via ball milling has been applied to the Cu–W system (e.g. [17–20]), truly effective stabilization, such as described in thin film structures, is determined by the ability to create extremely small particles which are well-dispersed.

M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

In an effort to generate similar results to the thin film work on a larger scale, Cu and W powders were processed via high-energy ball milling at cryogenic temperature. Alloying, as used here, must be loosely interpreted because of the immiscibility of the constituents, but it is important to note that some metastable solid solubility in the Cu–W system can be achieved by techniques such as ball milling [17–19], irradiation [21], or vapor deposition [22]. The thermal stability of Cu alloyed with 1, 5, and 10 at% W was investigated after milling for 8 h. Though the grain size retention after annealing was greatly improved over pure Cu, the effect was limited by the difficulty in reducing the W to nanoscale particles required to effectively pin the Cu nanograins.


3. Results XRD grain size estimation provides a quick and informative analysis technique, but the reasonable upper limit of accuracy in the reported data is only 40 nm due to instrumental broadening contributions masking grain growth. Grain size estimates greater than this limit are considered qualitative only. The results from XRD estimation are shown in Fig. 1 where it can be seen that as the percentage of W increases, the annealing temperature to which the microstructure remains nanoscale also increases. Although all of the samples have a comparable grain size at 400 1C, the 1 and 5 at% W samples show significant grain growth after annealing at 600 1C, whereas the 10 at% W sample remains qualitatively more stable even after annealing at 800 1C.

2. Experimental Elemental powders of Cu (Alfa Aesar, 99.9%) and W (Cerac, 99.95%,  200 þ325 mesh) were added in appropriate quantities to a 440 stainless steel vial (Spex SamplePrep, Metuchen, NJ) with grade 25, 440 stainless steel ball bearings (Salem Specialty Ball). The ball charge consisted of 17 balls 0.3125 in. in diameter and 16 balls 0.250 in. in diameter. The ball-to-powder weight ratio was maintained at 10:1. All materials were loaded into the vial in an argon atmosphere (O2 o1 ppm) and sealed before transferring to the mill. A modified Spex 8000M mixer/mill was used to mill the powders for 8 h in liquid nitrogen (  196 1C). W was added to the Cu in concentrations of 1, 5, and 10 at% (these alloys will be referred to as Cu1W, Cu5W, and Cu10W, respectively). The corresponding volume percentages for the W were 1.34%, 6.70%, and 13.40%, respectively. The milled powders were annealed under 2% H2 (balance Ar) for 1 h at 200, 400, 600, and 800 1C. After annealing the compacted powder, samples were polished, and at least 10 Vickers hardness measurements were collected for each annealing condition using a Buehler Micromet II hardness tester with a 25 g load applied for 12 s. Error bars for hardness denote the standard deviation. X-ray diffraction (XRD) analysis was conducted using a Rigaku DMax/A X-ray diffractometer using Cu Ka radiation. Proper alignment was verified before each use using a single crystal silicon standard. The patterns were processed by smoothing (function filter), removing the background, and stripping the Ka2 peaks (optimized Rachinger’s method) using Xpowder software ( The grain size was estimated using the Scherrer equation [23] applied to the Cu 111 peak of the pattern, as higher angle peaks were poorly resolved and convolution occurred between the Cu and W peaks making fitting less reliable. Scanning electron microscopy (SEM) analysis was performed with a JEOL JEM 6010LA using a 20 keV beam energy. The SEM is equipped with an energy dispersive X-ray spectrometer (EDS) which was used for X-ray mapping and to verify atomic concentrations. Focused ion beam (FIB) cross-sectioning and ion contrast channeling (FIBICC) were performed using an FEI Nano600 dual beam microscope. Transmission electron microscopy (TEM) was carried out using a JEOL JEM 2000FX using a beam energy of 200 keV. TEM samples were generated by uniaxially pressing the powder at room temperature to a pressure of 2.6 GPa in 3 mm tungsten carbide die and subsequently annealing the compact as described above. The compacts were mechanically thinned, electropolished using 30 vol% nitric acid in methanol at  20 1C in a Tenupol-2 electropolisher, and final thinning was performed with a Fischione Model 1010 ion mill. To determine the Cu grain size distribution at least 170 grains were measured. W particle sizes were determined using at least 200 particles.

Fig. 1. Scherrer grain size estimation for Cu and Cu–W samples (bold, red line at 40 nm indicates quantitative upper limit) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Fig. 2. XRD pattern of Cu10W milled for 8 h after annealing at labeled temperatures (in 1C except for ‘‘A-M’’ which is the as-milled condition). J—indicates W 110, 200, 211, and 220 reflections with increasing angle. K—indicates Cu 111, 200, 220, 311 and 222 reflections with increasing angle. The top-right inset highlights the W 110 and Cu 111 peaks.


M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

The evolution of the X-ray patterns with annealing for Cu10W is shown in Fig. 2. The W 110 fundamental peak shows a nearly imperceptible change in width throughout the annealing range. When the Scherrer equation is applied to the W peaks, the estimated particle size is between  17–30 nm for all alloys and at all temperatures. The peak broadening is not necessarily in direct relation to the average W particle size, as there may be large W particles which possess a nanoscale grain structure. The results do indicate that the W does not undergo any significant

Fig. 3. Hardness results for Cu and Cu–W samples.

structural changes during the annealing. The size and distribution of W particles are more precisely treated in the electron microscopy results. Hardness values (shown in Fig. 3) were obtained for each composition and annealing temperature reported for XRD in Fig. 1. The milled Cu is somewhat harder than the typical for high-purity Cu [24] and has a higher thermal stability than commonly reported for nanocrystalline Cu obtained by methods such as inert gas condensation [25] or electrolytic deposition [26]. Some highly pure nanocrystalline materials can undergo grain coarsening even at room temperature [27], but due to the nature of the ball milling process, where contaminants such as oxygen and iron can come from multiple sources [28], some minor impurity level may be introduced which enhances grain size stability [29]. Nonetheless, for a similar grain size (i.e. 18 nm), the as-milled hardness is within 10% of the as-deposited, highpurity results reported by Fougere et al. [30] and Sanders et al. [31] as compared in [24], as well as for results summarized in [32]. The annealing results demonstrate that even a 1 at% W addition appreciably decreases grain growth and increases residual hardness over the pure condition, so the impurity level does not significantly aid or negate the W effects. All of the Cu–W samples were nearly identical in the as-milled hardness. The variation, based on composition, does not become appreciable until 600 1C, a trend also appearing in the XRD grain size. Again, the Cu10W alloy showed the best property retention, but the hardness result cannot be solely attributed to grain size (see Section 4). When the samples were polished and examined using SEM (see Fig. 4) the broad W particle size distribution and the error of

Fig. 4. (a) SEM image of Cu10W as-milled for 8 h and (b) corresponding elemental map of W (yellow) by EDS. Particle cross-section shows Cu10W after annealing at 800 1C by secondary electron imaging in (c) and by ion channeling contrast (d) taken in different areas. Arrows indicate tungsten particles. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

the W particle size estimate using XRD became evident. Though it was found that there is a reduction in overall particle size as milling time increases, even after 8 h there are still many large particles in the as-milled material. EDS mapping of the alloys (such as in Fig. 4b) showed a significant amount of tungsten dispersed throughout the bulk, possibly as unresolved particles. By using a focused ion beam to cross-section a particle annealed at 800 1C (Fig. 4c), it is evident that many small, highly-aligned, elliptical W particles still exist. Using ion channeling contrast (see Fig. 4d), the grains can be seen to align with the particles. This behavior is elaborated on in the TEM results. TEM was used for a more accurate determination of unresolved, dispersed W particle diameters and Cu matrix grain diameters. The Cu10W sample was chosen as a focus case since it exhibited the best stability. Fig. 5 shows the Cu10W sample after annealing at 600 1C. This annealing temperature was the limit of stability as defined by the limitations of the reported XRD grain size estimates (i.e. o40 nm). Dark field (Fig. 5b) confirms nanoscale grain size retention (47.4 nm and 32.6 nm volume and number averages, respectively). Using the W 110 ring in the SAD pattern, the particle distribution was highlighted (Fig. 5c), and there is a considerable distribution of nanoscale W particles


which appear in bright contrast (white). The volume and number averages for the W particles are 14.5 nm and 9.1 nm respectively. The distributions of grain and particle sizes are shown in Fig. 5d. The maximum particle size included in this analysis was 40 nm, as those any larger were rarely observed at the given magnification, and the larger particles significantly skewed the volume average but had little effect on the number average. SEM analysis more accurately accounts for the large W particles neglected in the TEM analysis. According to XRD, after annealing at 800 1C the grains should grow significantly, and based on previous experience [33], the grains are expected to be well above the Scherrer estimate of 59 nm. This is indeed the case as the SAD pattern (Fig. 6a) is now quite discontinuous and the dark field image of the Cu grains (Fig. 6b) reveals a larger size (volume average of 130.3 nm and number average of 110.3 nm). The W particles, in bright contrast (white) in Fig. 6(c), have volume and number averages of 13.4 nm and 7.1 nm, respectively. Although the grains have grown significantly, the average particle size has remained approximately the same suggesting that the W particles have not coarsened. The SAD pattern in Fig. 6(a) indicates strong texturing which was observed in many of the diffraction patterns at 600 1C and

Fig. 5. TEM of Cu10W sample milled for 8 h and annealed at 600 1C for 1 h. (a) bright field image with SAD pattern inset top-left (b) dark field image using Cu 111 and 200 rings (c) dark field image using W 110 ring and (d) Cu grain size and W particle size distributions.


M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

Fig. 6. TEM of Cu10W sample milled for 8 h and annealed at 800 1C for 1 h. (a) Bright field image with SAD pattern inset top-left (b) dark field image using Cu 111 and 200 rings (c) dark field image using W 110 ring and (d) Cu grain size and W particle size distributions.

800 1C. Closer microstructural examination provides clues into why this may occur. As shown in Fig. 7, the W particles (some labeled with arrows) are elongated and have a preferred orientation. When this area is examined more closely (Fig. 7b), there is evidence that the grains also follow this preferred orientation with longitudinal axes in the same general direction as the W particles. This may be attributed to the orientation dependent pinning efficiency of elliptical particles [9,34,35]. This particle orientation preference can also be seen in SEM (Fig. 4) where the W particles follow local ‘‘flow paths’’ so that, although randomized over a large scale, they are highly oriented on a finer scale.

4. Discussion 4.1. Grain size stability and morphology

spherical grain boundary of radius, R, and energy, g, is given by [9] P¼


As the radius of a grain decreases into the nano regime (10  9 m), there exists a substantial pressure for expansion by grain growth. When considering second phase drag by Zener pinning, a volume fraction, f, of randomly distributed spherical particles of radius, r, will produce a pressure, Pz, on a grain boundary with an energy, g, of [5,9,34] Pz ¼

3f g 2r


This indicates that as the volume fraction of the pinning particles increases and their size decreases, their ability to stabilize a given grain size will improve. Using this approach, the limiting grain diameter, D, can be given by equating the two opposing pressures [36]: D¼

To determine the effect of the W dispersoids on grain size retention in the Cu matrix, a simplified analysis can be applied. For curvature driven grain growth, the driving pressure on a

2g R

4r 3f


In [36], there is a proportionality term, a, included in Eq. 3 which is set equal to 1 here. Strictly speaking, in Cu10W f¼0.134,

M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233


Fig. 7. TEM images of Cu10W milled for 8 h and annealed at 600 1C for 1 h. (a) Arrows indicate some of the tungsten particles (b) magnified view of boxed area in (a) with longitudinal axes of some oblong grains labeled with black arrows.

but from TEM a lower and more reasonable value for contributing particles is given by the effective volume fraction, f*, which will be elaborated on in the section on strengthening mechanisms. A limiting grain size of 111 nm is predicted from Eq. (3) using values of r ¼5 nm and f* ¼0.06 which are derived from experimental observation. For the sample annealed at 800 1C, D E120 nm which agrees well with the calculated value. There has been a great deal of work on refining the equations for critical grain size [37], and a more recent analysis [38] has found a coefficient of 0.4 more favorable for predicting critical grain diameter than 1.33 as used in Eq. (3). Using 0.4, and experimental values for the sample annealed at 800 1C, the limiting grain diameter is predicted to be 33 nm (closer to the 600 1C result). Using the modified analysis [38], in this instance, is complicated by the considerable W content in the 10 at% sample. This complication is because at large volume fractions (f 40.05) the limiting grain size of a given system is predicted to approach the original (larger) Zener size limit [37]. Many of the W particles are elliptical, and the effect of the deviation from sphericity will be two-fold as described in [34]. For a boundary approaching the particle normal to its longitudinal axis (Case 2 in Fig. 8), the drag force will be higher than for a spherical particle of equivalent volume. If approaching parallel to that axis (Case 1 in Fig. 8), the force will be lower, and the intensity of the effect is sensitive to the aspect ratio of the particle. Many of the smallest W particles had an aspect ratio of 3:1 which gives an effectiveness ratio for the elliptical particles to their spherical equivalents of approximately 1.5 and 0.3 for the respective geometries. This anisotropy yields a factor of 5 preference for growth in the parallel (Case 1) direction over the transverse (Case 2) direction. The variation in elliptical pinning efficiency with angle is treated specifically in [35], and the texture arising from precipitate distribution and particle shape is discussed in more detail in [34]. Although Eq. (3) predicts that the smallest particles are most effective at stabilizing grain size, the smaller cross-section of the parallel particles (Case 1) is unfavorable toward stabilization based on the associated volume. That is, spherical particles with the same interaction diameter would have a much smaller volume, or conversely, the number of spherical particles for the same volume would be greater. As such, the elliptical particles are not beneficial over spherical particles in a general way, but rather, they exhibit anisotropy which makes

Fig. 8. Depiction of pinning anisotropy by elliptical particles (after [34]).

the Case 2 orientation more effective when considering particles of a given volume. The combination of particle shape and local alignment gives cause for elongated grain growth/texturing after annealing since migrating grain boundaries can be more effectively pinned along one axis.

4.2. Strengthening mechanisms Typically, the strength of an alloy is also affected by a second phase presence, and can be very significant [39]. Using the approach presented in [40], the hardness has contributions from solid solution (Hss), Orowan (HOro), and grain size (HH–P) strengthening. Here, we also include rule-of-mixture hardening (HROM) and assume negligible solid solution strengthening. The hardness is then given by the grain size through the Hall–Petch (H–P) relation [41], and precipitate strengthening through an Orowan mechanism for a random distribution of particles [42] as well as the hardness of the second phase (modified from [43]). Using Tabor’s relation (H¼3sy) and the Von Mises flow rule (t ¼ sy/O3) [44],


M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

the total hardness can be given by " # pffiffiffi Gbln l ln 2r 3=2 H ¼ HOro þHHP þHROM ¼ 3 3 ln l 2pl h pffiffiffii   þ 3 so þ k= d þ f ðHW HHP Þ

useful in high stress and wear applications such as electrical discharge machining [51]. Based on the improvements in thermal performance, W can be considered a useful addition for stabilizing and strengthening nanocrystalline Cu even at high temperature. ð4Þ

where G and b are the shear modulus (48 GPa) and Burgers vector (0.25 nm) of Cu respectively, d is the grain size, HW is the hardness of the tungsten taken as 5.9 GPa [45], so is the frictional stress required to move dislocations and k is the H–P slope (25.5 MPa and 0.11 MN/ m3/2 respectively [41]). l is the interparticle spacing as used in [46] rffiffiffiffiffiffiffiffi  p l ¼ 2r ð5Þ n 1 4f where the effective volume fraction is used since large particles do not contribute significantly to HOro [45]. All of the length parameters in the natural log terms are normalized by the dislocation core radius (taken as 4b). Based on this approach, the hardness contributions for Cu10W at 600 1C by grain size, Orowan, and ROM will be 1.73, 0.75, and 0.56 GPa, respectively, for a total calculated hardness of 3.04 GPa (d¼40 nm, r¼ 5 nm, f* ¼0.06, and f¼0.134). At 800 1C, the Orowan hardening does not change, but the grain size contribution will drop to 1.03 GPa and the ROM contribution will now be 0.65 GPa for a total expected hardness of 2.43 GPa (d¼120 nm, r¼ 5 nm, f* ¼0.06, and f¼0.134). Experimentally, it was found that the hardness of the Cu10W sample annealed at 600 1C was 2.7 GPa and 2.6 GPa when annealed at 800 1C. The respective calculated values of 3.04 GPa and 2.43 GPa are in good agreement with the experimental results given the simplicity of the analysis. The grain size-to-Orowan contributions are calculated to be 3:1 (HH–P to HOro) at 600 1C and 3:2 at 800 1C. Because the Orowan contribution is considerable, the values chosen for the second phase are quite important. Examination using TEM such as that in Fig. 7 revealed that the area fraction of W particles under 40 nm is  4%–6%, so even though the volume fraction for the 10 at% W sample should be 13.4%, the effective volume fraction was only 6%. The nanoscale size range will be the most effective at Zener pinning and Orowan strengthening and is believed to be more representative of the amount of W most active in stabilizing and strengthening the Cu. The reduced volume fraction incurred by using this method of sample preparation is likely unavoidable due to the tendency of nanocomposite materials to reach a lower limit for dispersoid size [47], though the size is expected to decrease as a whole and increase in uniformity upon further milling. The poor comminution of W in Cu has been observed elsewhere [20] and attributed to the large, positive heat of mixing and not solely based on hardness. Due to a balance of fracture and cold welding, the use of cryogenic milling here optimizes second phase size reduction since W particles saturate at a larger size in Cu when milling temperature increases [48]. At saturation, the W is suggested to simply ‘‘swim’’ in the Cu matrix similar to oxide or carbide particles at saturation [49]. Despite the low volume fraction of nanoscale W, the stability is certainly an improvement over pure copper. The grain size could be maintained near 100 nm after annealing at 800 1C, although the total solute content (13.4 vol%) is rather high compared to the effective amount ( 4-6 vol%). As indicated in Eq. (3), if the pinning phase was entirely nanoscale, the amount required to retain a nanoscale grain size ( o100 nm) would be greatly reduced. The thin film work on this system [15,16] was found to be highly effective, and this can be attributed to the W starting in solution and particles remained below 10 nm even after annealing to 900 1C, which is highly advantageous to grain pinning. Typically, it is unfavorable to use large alloying additions if interested in maintaining the properties of pure Cu (e.g. high electrical conductivity [50]), but higher W concentrations are

5. Conclusion The kinetic route to stabilization by way of Zener pinning relies on a large volume fraction of small, second phase particles distributed in the matrix. This is a difficult proposition when starting with large particles of a hard metal (W) distributed in a soft metal (Cu), even when using cryogenic milling. The W particles ranged widely in size with some beingo10 nm and others were41 mm. Despite the size distribution, a nanocrystalline Cu grain size could be stabilized below 100 nm up to 400 1C by 1 at% W and to 600 1C by adding 10 at% W. The fraction of nanoscale W was found to be approximately half of the total W concentration after milling the 10 at% W sample for 8 h. The 10 at% W addition acted to maintain a small grain size (  120 nm) and high hardness (2.6 GPa) even after annealing at 800 1C.

Acknowledgments The authors wish to acknowledge the support of this research by the Office of Naval Research under Grant no. N00014-10-1-0168. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

R. Kirchheim, Acta Mater. 55 (2007) 5129–5138. P.C. Millett, R.P. Selvam, A. Saxena, Acta Mater. 54 (2006) 297–303. J. Weissmuller, Nanostruct. Mater. 3 (1993) 261–272. P. Wynblatt, R.C. Ku, Surf. Sci. 65 (1977) 511–531. C.C. Koch, R.O. Scattergood, K.A. Darling, J.E. Semones, J. Mater. Sci. 43 (2008) 7264–7272. F. Liu, R. Kirchheim, J. Cryst. Growth 264 (2004) 385–391. Z. Chen, F. Liu, H.F. Wang, W. Yang, G.C. Yang, Y.H. Zhou, Acta Mater. 57 (2009) 1466–1475. J. Li, J. Wang, G. Yang, Scripta Mater. 60 (2009) 945–948. F.J. Humphreys, M. Hatherly, Recrystallization and Related Annealing Phenomena, 2nd ed., Elsevier, Amsterdam, 2004. D.G. Morris, M.A. Morris, Acta Metall. Mater. 39 (1991) 1763–1770. R.J. Perez, H.G. Jiang, E.J. Lavernia, C.P. Dogan, Metall. Mater. Trans. A 29 (1998) 2469–2475. L. Shaw, H. Luo, J. Villegas, D. Miracle, Acta Mater. 51 (2003) 2647–2663. D.A. Porter, K.E. Easterling, Phase Transformations in Metals and Alloys, 2nd ed., Chapman Hall, London, 1992. P.R. Subramanian, D.E. Laughlin, in: S.V.N. Naidu, P.R. Rao (Eds.), Phase Diagrams of Binary Tungsten Alloys, Indian Institute of Metals, Calcutta, 1991. X. Zhang, N.Q. Vo, P. Bellon, R.S. Averback, Acta Mater. 59 (2011) 5332–5341. N.Q. Vo, S.W. Chee, D. Schwen, X. Zhang, P. Bellon, R.S. Averback, Scripta Mater. 63 (2010) 929–932. C.-S. Xiong, Y.-H. Xiong, H. Zhu, T.-F. Sun, E. Dong, G.-X. Liu, Nanostruct. Mater. 5 (1995) 425–432. T. Raghu, R. Sundaresan, P. Ramakrishnan, T.R.R. Mohan, Mater. Sci. Eng. A 304–306 (2001) 438–441. E. Gaffet, C. Louison, M. Harmelin, F. Faudot, Mater. Sci. Eng. A 134 (1991) 1380–1384. D.G. Morris, M.A. Morris, Scripta Metall. Mater. 24 (1990) 1701–1706. W. Hiller, M. Buchgeister, P. Eitner, K. Kopitzki, V. Lilienthal, E. Peiner, Mater. Sci. Eng. A 115 (1989) 151–154. A.G. Dirks, J.J v.d. Broek, J. Vac. Sci. Technol. 3 (1985) 2618–2622. B.D. Cullity, S.R. Stock, Elements of X-Ray Diffraction, Prentice Hall, Upper Saddle River, 2001, p. 170. S. Takeuchi, Scripta Mater. 44 (2001) 1483–1487. S. Okuda, M. Kobiyama, T. Inami, S. Takamura, Scripta Mater. 44 (2001) 2009–2012. L. Lu, N.R. Tao, L.B. Wang, B.Z. Ding, K. Lu, J. Appl. Phys. 89 (2001) 3408–3414. M. Ames, J. Markmann, R. Karos, A. Michels, A. Tschope, R. Birringer, Acta Mater. 56 (2008) 4255–4266. C. Suryanarayana, E. Ivanob, V.V. Boldyrev, Mater. Sci. Eng. A 304–306 (2001) 151–158. J.M. Tao, X.K. Zhu, P.Z. Wong, R.O. Scattergood, C.C. Koch, Proceedings of the 3rd International Nanoelectronics Conference (INEC), IEEE, 2010, pp. 957– 958.

M.A. Atwater et al. / Materials Science & Engineering A 558 (2012) 226–233

[30] G.E. Fougere, J.R. Weertman, R.W. Siegel, S. Kim, Scripta Metall. Mater. 26 (1992) 1879–1883. [31] P.G. Sanders, J.A. Eastman, J.R. Weertman, Acta Mater. 45 (1997) 4019–4025. [32] S. Ozerinc, K. Tai, N.Q. Vo, P. Bellon, R.S. Averback, W.P. King, Scripta Mater., in press. [33] K.A. Darling, B.K. VanLeeuwen, C.C. Koch, R.O. Scattergood, Mater. Sci. Eng. A 527 (2010) 3572–3580. [34] E. Nes, N. Ryum, Hunderi, Acta Metall. 33 (1985) 11–22. [35] W.-B. Li, K.E. Easterling, Acta Metall. Mater. 38 (1990) 1045–1052. [36] P. Bate, Acta Mater. 49 (2001) 1453–1461. [37] P.A. Manohar, M. Ferry, T. Chandra, ISIJ Int. 38 (1998) 913–924. [38] R.D. Doherty, Mater. Sci. Forum 715–716 (2012) 1–12. [39] H. Kotan, K.A. Darling, M. Saber, J.C. Ledford, C.C. Koch, R.O. Scattergood, Acta Mater., to be submitted. [40] D. Srinivasan, R. Corderman, P.R. Subramanian, Mater. Sci. Eng. A 416 (2006) 211–218.


[41] M. Meyers, K. Chawla, Mechanical Behavior of Materials, Cambridge University Press, New York, 2009, pp. 346–347. [42] D.J. Bacon, U.F. Kocks, R.O. Scattergood, Philos. Mag. 28 (1973) 1241–1263. [43] J.B. Ferguson, H. Lopez, D. Kongshaug, B. Schultz, P. Rohatgi, Metall. Mater. Trans. A 43 (2012) 2110–2115. [44] J. Moon, S. Kim, J.-I. Jang, J. Lee, C. Lee, Mater. Sci. Eng. A 487 (2008) 552–557. [45] K.V. Rajulapati, R.O. Scattergood, K.L. Murty, Z. Horita, T.G. Langdon, C.C. Koch, Metall. Mater. Trans. A 39 (2008) 2528–2534. [46] A.M. Redsten, E.M. Klier, A.M. Brown, D.C. Dunand, Mater. Sci. Eng. A 201 (1995) 88–102. [47] D.L. Zhang, Prog. Mater Sci. 49 (2004) 537–560. [48] I. Sabirov, R. Pippan, Scripta Mater. 52 (2005) 1293–1298. [49] R. Pippan, S. Scheriau, A. Taylor, M. Hafok, A. Hohenwarter, A. Bachmaier, Annu. Rev. Mater. Res. 40 (2010) 319–343. [50] S.-H. Hong, B.-K. Kim, Z.A. Munir, Mater. Sci. Eng. A 405 (2005) 325–332. [51] S.H. Lee, X. Li, J. Mater, Process Tech. 139 (2003) 315–321.