Microstructural evolution in a nanocrystalline Cu-Ta alloy: A combinedin-situ TEM and atomistic study

M. Rajagopalan a, K. Darling b, S. Turnage a, R.K. Koju c, B. Hornbuckle b, Y. Mishin c, K.N. Solanki a,⁎a School for Engineering of Matter, Transport, and Energy, Arizona State University, Tempe, AZ 85287, USAb Army Research Laboratory, Weapons and Materials Research Directorate, APG, MD 21014, USAc Department of Physics and Astronomy, George Mason University, MSN 3F3, Fairfax, VA 22030, USA

H I G H L I G H T S

• Nanocrystalline Cu-10at.%Ta shows ex-ceptional thermal stability unlike anyother NC and ultra-fine grained metals

• Local structural changes at the interfacebetween nanoclusters and matrix influ-ence the thermo-mechanical properties

• The evolution of such fine structures iscritical for developing nanocrystallinealloys with extreme properties

G R A P H I C A L A B S T R A C T

a b s t r a c ta r t i c l e i n f o

Article history:Received 18 June 2016Received in revised form 7 August 2016Accepted 9 October 2016Available online 11 October 2016

Under intense heating and/or deformation, pure nanocrystalline (NC)metals exhibit significant grain coarsening,thus preventing the study of length scale effects on their physical response under such conditions. Hence, in thisstudy, we use in-situ TEM heating experiments, atomistic modeling along with elevated temperature compres-sion tests on a thermally stabilized nanostructured Cu–10 at.% Ta alloy to assess the microstructural manifesta-tions caused by changes in temperature. Results reveal the thermal stability attained in NC Cu-10 at.% Tadiverges from those observed for conventional coarse-grained metals and other NC metals. Macroscopically,the microstructure, such as Cu grain and Ta based cluster size resists evolving with temperature. However,local structural changes at the interface between the Ta based clusters and the Cu matrix have a profound effecton thermo-mechanical properties. The lattice misfit between the Ta clusters and the matrix tends to decrease athigh temperatures, promoting better coherency. In other words, the misfit strain was found to decrease mono-tonically from 12.9% to 4.0% with increase in temperature, leading to a significant change in flow stress, despitewhich (strength) remains greater than all known NCmetals. Overall, the evolution of such fine structures is crit-ical for developing NC alloys with exceptional thermo-mechanical properties.

© 2016 Elsevier Ltd. All rights reserved.

Keywords:In situ TEMNanocrystallineAtomisticMisfit strain

1. Introduction

Metals with amean grain size (d) below 100 nm, i.e., nanocrystalline(NC) materials, have garnered significant interest due to their superiormechanical properties as compared to coarse-grained materials [1] A

Materials and Design 113 (2017) 178–185

⁎ Corresponding author.E-mail address: kiran.solanki@asu.edu (K.N. Solanki).

http://dx.doi.org/10.1016/j.matdes.2016.10.0200264-1275/© 2016 Elsevier Ltd. All rights reserved.

Contents lists available at ScienceDirect

Materials and Design

j ourna l homepage: www.e lsev ie r .com/ locate /matdes

large number of experimental and computational studies have exploredhow grain boundarymediated plasticity andmicrostructural size effectsimpact themechanical behavior of NCmaterials [2–4]. For example, theHall-Petch [5,6] relationship describes the experimentally-observed in-crease in yield strength with decreasing grain size down to grain diam-eters as small as 20 nm [1,2]; this behavior is generally followed by aplateau/negative slope region for grain sizes below a critical size (e.g.,8–15 nm for Cu [7]). This inverse Hall-Petch effect has been directly at-tributed to changes in the governing deformation mechanisms awayfrom traditional dislocation glide and pile-up processes [8]. Fundamen-tal changes in deformation mechanisms are also known to cause manyother intriguing and unexpected physical responses of NC metals, in-cluding altered strain rate and pressure dependencies of deformation[9], superplasticity [10], and low temperature creep [11], to name afew (see [12,13]). Generally, these unique deviations in behavior aresolely attributed to a continual reduction in grain size and an increasein the fraction of grain boundaries and triple junctions, which leads tothe experimentally reported mechanisms of deformation twinning,grain boundary (GB) rotation/sliding and viscous flow [14–16].

Despite significant gains in knowledge related to the developmentand engineering of plasticity in nanomaterials, there still persists a size-able gap in the fundamental understanding of the mechanical behaviorof thesematerials, especially under extreme conditions such as at ultra-high temperatures. This gap is critical as the material performance athigh temperatures is significantly different from the room temperature(RT) behavior. Limited studies have explored the mechanical behaviorof NC materials at room or moderately low temperatures where a dras-tic change in the microstructure (grain growth) has been reported. Forexample, Farrokh and Khan investigated the uniaxial compressive be-havior of NC Cu and Al (grain size 32 nm and 82 nm, respectively) pre-pared through mechanical alloying at a strain rate of 0.01/s (Cu) and asignificant effect of temperature on the strength was observed [17]. Atthe homologous temperatures of 0.4 Tm and 0.56 Tm (Tm being themelt-ing temperature), the strengthwas reported to be 40% and 72% of the RTstrength for NC Cu andAl, respectively [17]. In case of alloy systems suchas NC multiphase Al alloys (Fe, Cr and Ti minor additions), a similartrend in loss of strength with temperature was reported for sampleswith the grain sizes of 50 and 80 nm tested at quasistatic rates. This sig-nificant loss in strength in the NCmetal/alloy systemswas attributed tothe loss in nanocrystallinity (grain coarsening) [18–22]which led to thedislocation-based mechanism to be a dominant deformation mecha-nism at elevated temperatures [23].

In general, the thermal and mechanical stability of NC microstruc-tures has been regarded low based on other experimental observations[19,24,25]. For instance, under thermal heating experiments, the pureNC Cu exhibits rapid grain growth to the micron-scale at just 100 °C[26–28]. Further, indentation studies at liquid nitrogen temperature re-vealed that NC Cu undergoes rapid grain growth, where themicrostruc-ture consists of grains as large as 700 nm (with sizeable volumefraction) after 30 min of dwell time in liquid nitrogen temperatures ascompared to an average grain size of 20 nm [24]. These examples illus-trate that, for nominally pureNCmetals, the nanoscalemicrostructure isan inherent barrier to experimental studies of their properties evenunder relatively low temperature conditions. It is expected that moreintense conditions, either thermal or mechanical, will result in a morerapid grain coarsening and a sudden loss of thematerial's intrinsic phys-ical and structural features. Recently, alternative methodologies havebeen successfully employed to impart greater stability to NC metals.These methodologies are based on thermodynamic considerations[29–33] coupled with classical kinetic mechanisms [34]. Such tech-niques allow for the retention of the as-processed fine grain size, espe-cially during high temperature consolidation. Hence, it may be statedthat such methodologies establish the fabrication pathway and, thusthe application of NC metals with desirable properties.

Recently, quasi-static and dynamic yield strengths of N1 GPa weremeasured in bulk samples of a NC Cu–Ta alloys, which could not be

explained by grain size strengthening alone [35]. The increase instrength was attributed to the thermal decomposition of a non-equilib-rium Cu rich Cu-Ta solid solution over a range of temperatures (700–900 °C), which led to the formation of a high density of small coherentTa-rich nanoclusters (~2 nm in diameter) [35,36]. The presence ofthese Ta nanoclusterswithin grains and along grain boundaries resultedin strength levels approximately twice as high as those predicted byHall–Petch hardening [37]. These studies suggest that the presence ofTa-based clusters play a commanding role in defining the deformationresponse as compared to the NC grain size alone. Therefore, in the pres-ent study, we examined themicrostructural evolution, i.e., the underly-ing changes in the Cu matrix grain size and the corresponding Tananoclusters by in-situ transmission electron microscopy (TEM)heating experiments and atomistic simulations.

Thiswork shows that the thermal stability achievedwith this NC Cu-10 at.% Ta alloy diverges from those observed for conventional coarse-grained metals as well as other NC metals. That is, macroscopically,the microstructure (Cu grain and Ta based cluster size) of NC Cu-10 at.% Ta was morphologically stable in shape and size. However,local structural changes at the interface between the Ta based clustersand the Cumatrix have a significant effect on thermo-mechanical prop-erties. Specifically, themisfit strain was found to decreasemonotonical-ly from 12.9% to 4.0% with increase in temperature, leading to asignificant change in flow stress, despite which remains greater thanall known NC metals.

2. Experimental details

High-energy cryogenic mechanical alloying was used to synthesizeNC powders with a composition of Cu-10 at.% Ta using very high purity(99.9%), ~325 mesh size elemental Cu and Ta powders. The ball millingwas carried out in a SPEX 8000M shaker mill at cryogenic temperatures(verified to be ~−196 °C) using liquid nitrogen with a milling time of4 h. The milling medium was 440C stainless steel balls and a ball-to-powder ratio of 5-to-1 by weight was maintained. Cryogenic mechani-cal milling resulted in an un-agglomerated powder mass with a partic-ulate size range of 20 μm to 100 μm. The as-milled powders were placedinto nickel cans and sealed inside the glove box. Prior to Equal ChannelAngular Extrusion (ECAE), the die assembly was heated to 350 °C. Thenickel cans loaded with as-milled powders were equilibrated (for40 min) in a box furnace purged with pure Ar cover gas at 700 °C. Theequilibrated canswere then quickly removed from the furnace, droppedinto the ECAE tooling, and extruded at an extrusion rate of 25.4 mm/susing 90° rotations for four passes. This process resulted in dense12mmdiameter rod sampleswith a stable initial grain size distribution.To identify the impurity content in these alloys, atom probe tomogra-phy was performed on the as-milled powder as well as the NC Cu −10 at.% Ta ECAE sample processed at 700 °Cwhere aminimal concentra-tion of 1.25 at.% was detected for O. Fe contamination was detectedusingAPT and varied between 0.05 at.% and1 at.%, indicating a relativelyimpurity-free alloy [9].

Quasistatic compression tests were carried out over a temperaturerange from 25 °C to 400 °C using an INSTRON load frame with a 50 kNload capacity and a furnace rated up to 1800 °C. The pushrods of theload framewere constructedwith precisionmachined ZrO2 rods tomin-imize heat losses. The specimens were cylinders with a 3 mm diameterand 3mm in length,machined using an electric dischargemachine fromthe same sample batch. Boron nitride lubricated polished WC-diskswere used as platens. The system was held at the testing temperaturefor about 30 min to attain equilibrium and thermocouples were at-tached on the specimen to monitor the temperature. Specimens wereloaded under strain control with a strain rate of 0.1/s.

In-situ TEM heating experiments were carried out in the aberrationcorrected FEI-TITAN TEM at 300 kV using an Inconel heating holder andthe temperature was monitored digitally. Aberration corrected ARM-200F operating at 200 kV was also used to acquire images, especially

179M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

of the nanoclusters illustrated in the forthcoming sections. Multiple im-ageswere acquired in the brightfield and the high resolution TEMmodeto analyze the microstructure and quantify the statistics such as grainsize distribution etc. Imaging was performed at the room temperature(25 °C) followed by imaging at temperatures 100 °C to 400 °C in100 °C intervals. Each temperature level was achieved in approximately20min and a holding time of 20min wasmaintained to reach equilibri-um temperature and minimize sample drift. Due to the effects of beamon a thin area, a relatively thick reference area was chosen for studyingthe microstructural evolution under temperature. The samples for in-situ heating were prepared through conventional thinning procedureswhere a 3 mm disk was punched from the bulk specimen and thinnedto about 100 μm followed by dimpling to about 5 μm thickness. Ionmill-ing was performed under liquid nitrogen temperatures using PrecisionIon Polishing System (PIPS) to obtain electron-transparent regions inthe specimens. The samples were then plasma cleaned in Ar prior toheating to minimize contamination.

3. Computational details

Interatomic interactions in the Cu-10 at.% Ta systemwere describedby a semi-empirical angular-dependent atomistic potential developedby Pun et al. [38]. This potential was parameterized using an extensivedatabase of energies and configurations from density functional theory(DFT) calculations of energy differences between various crystal struc-tures of pure Cu and pure Ta, the formation energies of coherent Cu-Ta interfaces, and the binding energy of several ordered compounds,such as L12-Cu3Ta, L10-CuTa, L11-CuTa, B2-CuTa, and L12-Ta3Cu [38].More details on the validation of the potential at different temperaturescan be found in [38]. This potential was recently applied to study struc-tural stability of Cu-Ta alloys and the Zener pinning of Cu grain bound-aries by nano-scale Ta clusters [39]. The simulations were performedusing the large-scale atomic/molecular massively parallel simulator(LAMMPS) [40].

To estimate the lattice misfit and coherency as functions of the clus-ter size and temperature, spherical Ta clusters were created inside cubicsimulation blocks of pure FCC Cu with several different sizes up to 40 ×40× 40 nmwith periodic boundary conditions. The clusters were creat-ed by replacing Cu atoms by Ta atoms within a spherical region at thecenter of the block. Five different cluster radii ranging from 1.5 to3.5 nm were studied in this work. Note that in this simulation scheme,the Ta atomswere initially arranged in FCC structure perfectly coherentwith the surrounding Cu. The structure was then relaxed by molecularstatics (total energy minimization), which resulted in a partial loss ofcoherency and transformation of certain regions inside the clusterfrom FCC to BCC. The cluster structure was examined using the visuali-zation tool OVITO [41]. Besides FCC and BCC, a small fraction of otherstructures or unstructured regions was usually found in the cluster, es-pecially near the Cu-Ta interface. The systemwas then slowly heated upto the temperature of 927 °C (1200 K) by molecular dynamics simula-tions and the structure evolution was monitored by periodicallyquenching the simulation block to −273 °C (0 K) and examining itsstructure with OVITO.

4. Results and discussion

The primary TEMmicrostructural characterization and XRD analysis(Figs. 1 and S1) of the as-received Cu-10 at.% Ta ECAE sample processedat 700 °C revealed the presence of Cu and Ta based phases, that is FCC Cuand BCC Ta. The orientations indexed through precession diffractiondata for the FCCCu andBCCTaphase of the as-received sample indicatesa random texture (Fig. S1 [36]). The TEM characterization along withgrain size distributions is illustrated in Fig. 1. The selected area diffrac-tion pattern confirms the nanocrystallinity of the sample at RT. Surpris-ingly, even though our NC material was consolidated to bulk throughsever plastic deformation at 700 °C with a total accumulated strain of

4.6 (i.e., 460%), the as received averaged grain size was 52 ± 14.3 nm,still in theNC regime (Fig. 1). In comparison, NC-materials have been re-ported to exhibit dramatic grain-coarsening even under low homolo-gous temperatures, such as NC Cu which was found to coarsen to themicron regime at just 100 °C [26–28]. Therefore, this NC Cu-10 at.% Taalloy exhibits a very stable microstructure. The processing conditionsproduced a wide range of Ta particle sizes, ranging from atomicnanoclusters to much larger precipitates (see Fig. 1A–D). Further, Fig.

noitcarfreb

muN

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

Ta nanoclusters

Diameter (nm)

Cunoitcarfreb

muN

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 10050 60

000

011

111

111011

000

011

111

111011

Ta (200)Cu (200)Cu (111)Ta (110)

Ta (211)Cu (211)

Diameter (nm)

Taparticles

noitcarfreb

muN

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 00150 60

Fig. 1. TEM characterization of as received NC Cu-10 at.% Ta (A) bright field TEM imageshowing the microstructure consisting of a distinct Cu(FCC) and Ta(BCC) phase (inset:SAED pattern), size distributions of (B) Cu and (C) larger Ta grains based on TEM images(average 200 grains), (D) size distribution of Ta nanoclusters with an average size of3.18 nm distributed along the grains and grain boundaries, and inverse Fourier filteredimages (IFFT) of (E–F) an coherent (d = 3.47 nm) and (G–H) semicoherent nanocluster(d = 4.11 nm). The interface between the Ta nanocluster and the Cu matrix arehighlighted using dashed green lines in (E–H). Inset: Indexed SAED patterns confirmingnanocrystallinity.

180 M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

Diameter (nm)

AVG. = 56.1 nmSTDEV = 13.7 nm

200 °C

noitcarf rebmu

N

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 00150 60

Diameter (nm)

AVG. = 61.5 nmSTDEV = 14.5 nm

400 °C

noitcarf rebmu

N

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 00150 60

Diameter (nm)

AVG. = 59 nmSTDEV = 13.6 nm

300 °C

noitcarf rebmu

N

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 00150 60

Diameter (nm)

AVG. = 52 nmSTDEV = 14.3 nm

25 °C

noitcarf rebmu

N

0.25

0.2

0.16

0.12

0.08

0.04

00 10 20 30 40 07 08 09 00150 60

noitcarf rebmu

N

0.08

0.04

0

0.25

0.2

0.16

0.12

Diameter (nm)0 10 20 30 40 07 08 09 00150 60

AVG. = 55 nmSTDEV = 15.5 nm

100 °C

noitcarf rebmu

N

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

400 °C

AVG. = 4.5 nmSTDEV = 0.88 nm

noitcarf rebmu

N

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

300 °C

AVG. = 3.65 nmSTDEV = 0.84 nm

noitcarf rebmu

N

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

200 °C

AVG. = 3.24 nmSTDEV = 0.83 nm

noitcarf rebmu

N

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

100 °C

AVG. = 3.18 nmSTDEV = 0.83 nm

noitcarf rebmu

N

0.23

0.19

0.15

0.11

0.08

0.04

0

Diameter (nm)0 1 2 3 4 65 7

25 °C

AVG. = 3.18 nmSTDEV = 0.86 nm

Ta (200)

Cu (200)

Cu (111)

Ta (110)

Ta (211)

Cu (211)

Fig. 2. Microstructural evolution of NC Cu-10 at.% Ta subjected to in-situ heating and corresponding grain size distributions at each temperature level. The grain sizes for Cu and Tananoclusters do not change significantly indicating the stability of the alloy.

181M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

1B, C and D shows histograms, taken from multiple images similar toFig. 1A, indicating that the larger Ta particle size distribution has an av-erage diameter of 39 ± 19 nmwhile the smaller Ta based nanoclusterspossess an average diameter of 3.18±0.86 nm. Small aggregations of Taatoms (e.g., ~10–50 atoms) are referred as nanoclusters and sizes largerthan 14 nm (diameter) are considered as particles, although we recog-nize that there is no sharp boundary between the two. The energy of theinterface [42] between the nanocluster and the Cumatrix can be used toquantify the type of coherency of these nanoclusters, and the range ofthe grain sizes which correspond to them. Characterizing the coherencyat RT has indicated that thismaterial has coherent, semicoherent and in-coherent nanoclusters (d b 3.898 nm, 3.898 to 15.592 nm, and N15.592 nm, respectively) [42].

Next, in-situ heating experiments were performed to probe and as-sess the microstructural manifestations caused by heat and/or changesin temperature as a fraction of the absolute melting point of Cu. Fig. 2shows a series of TEM brightfield (BF) images captured at periodic tem-perature intervals along with the grain size distributions of Cu grainsand Ta based nanoclusters. Each temperature level was achieved in20 min with 20 min of hold time and 20 mins of imaging. Thus, at400 °C, the time since the start of the test is approximately 240min. No-tably, the average grain size at 100 °C, 200 °C, 300 °C and 400 °C wasevaluated based on statistical analysis as 55 nm (60 min), 56.1 nm(120 min), 59 nm (180 min) and 61.5 nm (240 min), respectivelywhich corresponds to a change in grain size of 12% at 0.5 Tm. The grainsize distributions indicate limited grain growth in NC Cu-10 at.% Tawhen subjected to annealing (Figs. 2 and S2). While the average grainsize is estimated to increase only by 12% for Cu phase, the full widthhalf maximum values indicate an increase only about 4% for Cu grainand 10% for Ta nanoclusters at 0.5 Tm. However, the grain size of thelarger Ta particles (Fig. S2) virtually remains unaffected indicating thattheir evolution does not play a major role in the behavior of this mate-rial. This suggests that the alloy is resistant to grain growth and thereis no significant shift in the size distribution. From a practical perspec-tive the small increase in grain size does not manifest a perceivablechange in the mechanical properties as given by current Hall-Petch as-sessments of NC Cu [12]. Furthermore, the limited growth observedmay have more to do with enhanced surface diffusion kinetics relatedto the direct heating of the prepared TEM foil. The embeddednanoclusters also exhibit a strong resistance to coarsening, with the cor-responding size distributions for these nanoclusters provided in Fig. 2.The high stability is not completely unexpected as the initial consolida-tion temperature is 300 °C higher than the maximum temperature ofthe in-situ heating experiments. Therefore, from a macro perspectiveunder limited re-heating, the stability of the grain size and nanoclustersis very high in this alloy and diverges from those observed for conven-tional coarse-grained metals as well as other NC metals [27]. While ingeneral, the microstructure is morphologically stable in shape andscale, it is the fine changes in interfacial structure that dictate mechan-ical properties at elevated temperatures, owing to the dispersions ini-tially coherent/semi-coherent nature with the Cu matrix at roomtemperature.

To understand the stability of this alloy at a finer length scale, we ex-amined the effect of temperature on the high density of coherent andsemi-coherent nanoclusters (density = 6.5 × 1023/m3 [43]) distributedin the Cu matrix and along the grain boundaries. In general, the coher-ency of the particle also determines the misfit strain at the interface.Themisfit strain can be viewed as an indirectmeasurement of the inter-facial bonding strength and is correlated with the material's thermalstability and mechanical properties. Coherent particles have a lower in-terfacial energy as compared to incoherent particles, therefore coherentparticles aremore energetically favorable to form below a critical diam-eter. Fig. 1E and G illustrate the high resolution TEM (HRTEM) images ofa semi/coherent nanoclusters in the as-receivedmicrostructurewith di-ameters of 3.47 and 4.11 nm, respectively. The corresponding inversefast Fourier transform (IFFT) images appear in 1F and 1H, where the

yellow lines along the Cu-Ta interface in these images are indicative ofwhere the lattice deviates from the bulk (i.e.) the presence of misfit dis-locations. These misfit dislocations are introduced to minimize thestrain at the interface between the clusters and matrix. The frequencyof which can be seen to increase with the diameter of the clusters, seeFig. 1F and H.

In order to characterize the change inmisfit quantitatively, the aver-age misfit strain from in-situ heating experiments (ε⁎) was evaluatedusing the equation [44].

ε! ¼ 2 d2−d1ð Þ= d2 þ d1ð Þ ð1Þ

where d1 and d2 are the interplanar lattice spacings of the Ta based par-ticles and Cumatrix computed from the fast Fourier transform (FFTs) ata particular temperature. The strain values were averaged over the ref-erence area (illustrated in Fig. 1) and the average misfit value as a func-tion of temperature is represented in Fig. 3. In case of ODS steel, themisfit strain between the nano-sized oxide particle and matrix wasalso quantified to be relatively high, (i.e.) ~12.6% [45]. Although themis-fit strain in the as-received condition is high, i.e., 12.9%± 2%, themicro-structure appears to be relatively stable. With the increase intemperature, the average misfit strain was estimated to reduce to avalue about 4% ± 0.49%, at 400 °C. Partially, this can be explained bystructural relaxation of the clusters, resulting in increased coherencyand transformation to a more thermodynamically stable structure. An-other factor is the difference in thermal expansion factors of Cu and Tabased clusters, which may reduce the lattice parameter difference athigh temperatures.

In order to gathermore insight about themisfit dislocations and var-iation of misfit strain, atomistic simulations were performed. Fig. 3shows the lattice misfit strain calculated from atomistic simulations atroom temperature which agrees quite well with experimental findings.However, because of the complex two-phase structures of thenanoclusters observed in atomistic simulations (see Fig. 4A and B) andthe thermal noise at high temperatures, it was not possible to accuratelydetermine the interplanar spacing between Ta atoms from the atomisticsimulations in this work. Hence, this precluded direct calculations usingatomistic simulations of the temperature dependence of the latticemis-fit defined by Eq.(1). Nevertheless, the misfit strain calculated from at-omistic simulations at RT agrees well with in situ TEM experimentalmeasurement.

Despite this limitation, atomistic simulations were able to revealcritical structural changes within the Ta based clusters which remainedmostly (although not fully) coherentwith the Cumatrix, see Fig. 4. Since

Temperature ( °C)

0

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

0 200 300 400100

(niartstifsi

mecitta

Lε*

)

ExperimentsSimulations

Fig. 3.Averagemisfit strain evolution as a function of temperature for the Ta nanoclusters.With the increase in temperature the misfit strain decreases from 12.9% to 4% fromexperiments.

182 M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

the clusters/particles were initially created and statically relaxed atT = −273 °C / 0 K when the lattice misfit was large, they were neverperfectly coherent. Due to the partial loss of coherency, part of the initialFCC structure of the particles always transformed to BCC and otherstructural forms (Fig. 4A). In small clusters (e.g., with the radius of1.5 nm), the fraction of the FCC structure was found to increase withtemperaturewhile the fraction of the BCC structure decreased (see Sup-plementary Fig. S3). By contrast, in larger Ta clusters (particles), most ofthe initial FCC structures transformed to BCC already during the staticrelaxation (Fig. 4B). This resulted in nearly complete loss of coherencyand during the subsequent high-temperature anneals the fraction ofthe BCC structure in the cluster increased. The effects of the phasechange on the behavior of the alloy and detailed analysis will be includ-ed in a follow up article. However, one important observation was thatthe loss of coherencywas accompanied by emission of dislocation loopsinto the Cu matrix as illustrated in Fig. 4C. This figure shows a single Taparticle with an initially coherent boundary. The loss of coherency re-sulted in the formation ofmisfit dislocations at the particle/matrix inter-face. These dislocations then separated from the particle and formedloops gliding into the Cu matrix along (111) planes. For clarify, the per-fect FCC Cu atoms are not shown in Fig. 4c and the particle was cut inhalves during the visualization to reveal the its cross-section and a bet-ter view of the dislocation loops. Note the dislocation splitting intoShockley partials separated by a stacking fault (shown in red). It iswell known that a coherent nanocluster can lose coherency once itgrows above a critical diameter. This happens when the elastic energywith the coherent nanocluster becomes large and it is energetically fa-vorable for a dislocation to form at the matrix-nanocluster interface,see experimental Fig. 4D and E. The generation of such dislocations iseasier for an incoherent nanocluster in comparison to a coherentnanocluster. This dislocation raises the free energy of the system to anamountwhich equals to its formation energy but this increase in energyis compensated by the decrease in the self-energy of the nanocluster.Due to the presence of the misfit strain around nanocluster, the pin-ning-depinning of propagating dislocations during the deformationwith in the matrix will be hindered [46]. This hindrance will have a di-rect consequence on themechanical behavior. Hence, the change in co-herency can be linked to the change in mechanical properties [47,48].

Strength and stability of an alloy at elevated temperatures is con-trolled by not only the global microstructural changes but also thelocal changes in the microstructure. The high strength in alloys is influ-enced by the retention of the grain size as well as the nature of the par-ticle-matrix interface (in our case, Ta nanocluster - Cu matrix) with the

increase in temperature. To examine the effect, we inspected the com-pressive flow stress of the alloy at various temperatures. The flow stresswas extracted at 10% strain from the stress-strain response of samplessubjected to quasistatic compression tests at 0.1/s strain rate. The flowstress was found to decrease with the increase in temperature. It is crit-ical to note that even the high-temperature of 400 °C (0.5 Tm) the flowstrength is on par with pure NC Cu composed of 20 nm grain tested atroom temperature [30] and the strength at 400 °C for NC Cu-10 at.%Ta is 24% higher in comparison to the data reported for NC Cu (d =32 nm) sample tested at 250 °C [17]. This is understandable becausethe Ta nano-clusters distributed along the GBs and the matrix is be-lieved to be responsible for the exceptional thermal stability and

Fig. 4. Cross sections of Ta clusters with a diameter of (A) 3 nm and (B) 7 nm at 426 °C (700 K). Ta atoms in FCC and BCC environments are shown in green and blue. The grey atomsrepresent other structural environments; (C) dislocation loops emitted by the Ta particle with d = 7 nm at 426 °C. Note the dislocation splitting into partials separated by a stackingfault (red color). The FCC Cu atoms were removed for clarity and half of the Ta particle was cut out to reveal its spherical cross-section. (D) HRTEM and (E) IFFT image of a semi-coherent particle with d = 5.0 ± 0.035 nm at 400 °C with misfit dislocations highlighted in yellow (The scale bar is identical for d and e and corresponds to 2 nm).

Fig. 5. TEM bright field images in the (A) as-received condition and (B) 400 °C for NC Cu-10 at.% Ta alloy. The distances between large Ta particles, Ta nanocluster (P) and GB ishighlighted for both the conditions. There is a significant Ta nanoclusters pinning the GBat both the conditions validating the Zener pinning theory [39].

183M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

mechanical strength. That is, the stability arises from the nanoclustersthat pin the grain boundaries by the Zener mechanism, therebypreventing significant grain coarsening and the strength arises fromthe endurance of the nanoclusters to dislocation motion and activationof twinning basedmechanism [39,49,50]. Fig. 5 highlights once such re-gion in the sample at RT and at 0.5 Tm where a dense distribution ofnanoclusters can be identified along the matrix and the GB. This obser-vation confirms the presence and the interaction between the grainboundaries and the Ta nanoclusters that is responsible for the uniquestability and strength exhibited by the alloy, also see [39].

To address the reasoning behind the change in flow strength, weprobed the change in flow stress as a function of average misfit strainat the interface of the high density of nanoclusters and the matrix.Fig. 6 indicates the flow stress values plotted against the average misfitstrain of the Ta nanoclusters-matrix interface in NC Cu-10 at.% Ta proc-essed at 700 °C for three levels of temperature. The variation in flowstresswas found to be strongly correlatedwith the averagemisfit strain,with high flow stress values observed at the RTwhich decreasewith thedecrease in misfit strain values. Other typical contributions to thermalsoftening including microstructural coarsening (Cu grain size and Tabased cluster) and changes in the elastic modulus may be excluded astheir contributions are negligible over the given testing temperatures(24–400 °C). The higher the coherency, the lower the misfit latticestrain and the lower the flow stress as the strain field surrounding theTa based clusters decreases and interactions with dislocations becomeless prominent., i.e., the pinning-depinning of propagating dislocationsduring the deformation within in the matrix will be less hindered asthe coherency increases [46]. Additionally, as temperature increases,normal particles undergo coarsening and transition from a coherent tosemi-coherent to incoherent state and strength is lost due to over-aging. In this case the particles gain coherency and at high enough tem-peratures may be more easily bypassed. However, in this class of mate-rials, a significant coherency strain is retained (~4%), thereby providingsignificant resistance to high temperature deformation. This is identicalwith the data reported for superalloys, where lattice mismatch and co-herency at the γ/γ′ interface determines the resistance for dislocationpropagation ultimately leading to superior thermal and mechanicalproperties for the superalloys [51]. Overall, the mechanical response ofthe NC Cu-10 at.% Ta alloy is still remains greater than all known NCmetals and alloys. Further, the NC Cu-10 at.% Ta alloy exhibits negligiblegrain growth, which deviates from the conventional trend, thusresulting in a kinetically stable material with a superior flow strengtheven at 0.5 Tm.

5. Conclusion

In this work, we use in-situ TEM heating experiments, atomisticmodeling alongwith elevated temperature compression tests on a ther-mally stabilized nanostructured Cu–Ta alloy (Ta=10 at.%) to probe andassess themicrostructural manifestations caused by heat and/or chang-es in temperature. The results show that the thermal stability achievedwith this NC Cu-10 at.% Ta diverges from those observed for convention-al coarse-grained metals as well as other NC metals. While the averagegrain size was estimated to increase only by 12% for Cu phase, the fullwidth half maximum values indicate an increase of only about 4% forCu grain and 10% for Ta nanoclusters at 0.5 Tm. This suggests that thealloy was resistant to coarsening and there was no significant shift inthe size distribution. From a practical perspective, the small increasein grain size does not manifest a perceivable change in the mechanicalproperties as given by current Hall-Petch assessments of NC Cu. There-fore, macroscopically, the microstructure (Cu grain and Ta based clustersize) of NC Cu-10 at.% Tawas morphologically stable in shape and scale.However, local structural changes at the interface between the Ta basedclusters and the Cu matrix have been observed. The lattice misfit be-tween the Ta clusters and the matrix tends to decrease at high temper-atures, promoting better coherency. In other words, the misfit strainwas found to decrease monotonically from 12.9% to 4.0% with increasein temperature, leading to a significant change in flow stress, which re-mains greater than for all known NC metals. Overall, the evolution ofsuch fine structures is for critical the thermal-mechanical properties ofthis alloy.

Acknowledgements

M. Rajagopalan, S. Turnage, and K. N. Solanki are grateful for the fi-nancial support for this work from the Army Research Laboratoryaward number W911NF-15-2-0038 and would also like to thank theLeRoy Eyring Center for Solid State Science at Arizona State University.R. K. Koju and Y. Mishin were supported by the U.S. Army Research Of-fice under contract number W911NF-15-1-007.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.matdes.2016.10.020.

References

[1] H. Gleiter, Nanostructuredmaterials: basic concepts andmicrostructure, ActaMater.48 (2000) 1–29, http://dx.doi.org/10.1016/S1359-6454(99)00285-2.

[2] M.A. Meyers, A. Mishra, D.J. Benson, Mechanical properties of nanocrystalline mate-rials, Prog. Mater. Sci. 51 (2006) 427–556, http://dx.doi.org/10.1016/j.pmatsci.2005.08.003.

[3] M. Dao, L. Lu, R.J. Asaro, J.T.M. De Hosson, E. Ma, Toward a quantitative understand-ing of mechanical behavior of nanocrystalline metals, Acta Mater. 55 (2007)4041–4065, http://dx.doi.org/10.1016/j.actamat.2007.01.038.

[4] Q. Wei, B.E. Schuster, S.N. Mathaudhu, K.T. Hartwig, L.J. Kecskes, R.J. Dowding, K.T.Ramesh, Dynamic behaviors of body-centered cubic metals with ultrafine grainedand nanocrystalline microstructures, Mater. Sci. Eng. A 493 (2008) 58–64, http://dx.doi.org/10.1016/j.msea.2007.05.126.

[5] E.O. Hall, Variation of hardness of metals with grain size, Nature 173 (1954)948–949, http://dx.doi.org/10.1038/173948b0.

[6] N.J. Petch, The cleavage strength of polycrystals, J. Iron Steel Inst. Lond. 173 (1953)25–28.

[7] J. Schiøtz, F.D. Di Tolla, K.W. Jacobsen, Softening of nanocrystalline metals at verysmall grain sizes, Nature 391 (1998) 561–563, http://dx.doi.org/10.1038/35328.

[8] A.H. Chokshi, A. Rosen, J. Karch, H. Gleiter, On the validity of the hall-petch relation-ship in nanocrystalline materials, Scr. Metall. 23 (1989) 1679–1683, http://dx.doi.org/10.1016/0036-9748(89)90342-6.

[9] B.C. Hornbuckle, T. Rojhirunsakool, M. Rajagopalan, T. Alam, G.P.P. Pun, R. Banerjee,K.N. Solanki, Y. Mishin, L.J. Kecskes, K.A. Darling, Effect of Ta solute concentration onthe microstructural evolution in immiscible Cu-Ta alloys, JOM 67 (2015)2802–2809, http://dx.doi.org/10.1007/s11837-015-1643-x.

[10] O.D. Sherby, J. Wadsworth, Superplasticity—recent advances and future directions,Prog. Mater. Sci. 33 (1989) 169–221, http://dx.doi.org/10.1016/0079-6425(89)90004-2.

Lattice misfit strain (ε*)0.120.100.080.060.040.020

1400

1200

1000

800

600

200 °C400 °C

0.14

)aP

M(sserts

wolF 25 °C

Fig. 6. Flow stress as a function of average misfit strain of the Ta nanoclusters in NC Cu-10 at.% Ta processed at 700 °C. The flow stress was extracted at 10% strain from thestress-strain response of samples subjected to quasistatic compression tests at 0.1/sstrain rate.

184 M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185

[11] M.A. Bhatia, S.N. Mathaudhu, K.N. Solanki, Atomic-scale investigation of creep be-havior in nanocrystalline Mg and Mg–Y alloys, Acta Mater. 99 (2015) 382–391,http://dx.doi.org/10.1016/j.actamat.2015.07.068.

[12] M.A. Tschopp, H.A. Murdoch, L.J. Kecskes, K.A. Darling, “Bulk” nanocrystallinemetals: review of the current state of the art and future opportunities for copperand copper alloys, JOM 66 (2014) 1000–1019, http://dx.doi.org/10.1007/s11837-014-0978-z.

[13] J. Li, T. Suo, C. Huang, Y. Li, H. Wang, J. Liu, Adiabatic shear localization in nanostruc-tured face centered cubic metals under uniaxial compression, Mater. Des. 105(2016) 262–267.

[14] M. Chen, E. Ma, K.J. Hemker, H. Sheng, Y. Wang, X. Cheng, Deformation twinning innanocrystalline aluminum, Science 300 (2003) 1275–1277.

[15] X.Z. Liao, Y.H. Zhao, S.G. Srinivasan, Y.T. Zhu, R.Z. Valiev, D.V. Gunderov, Deformationtwinning in nanocrystalline copper at room temperature and low strain rate, Appl.Phys. Lett. 84 (2004) 592–594, http://dx.doi.org/10.1063/1.1644051.

[16] Y.T. Zhu, X.Z. Liao, X.L. Wu, Deformation twinning in nanocrystallinematerials, Prog.Mater. Sci. 57 (2012) 1–62, http://dx.doi.org/10.1016/j.pmatsci.2011.05.001.

[17] B. Farrokh, A.S. Khan, Grain size, strain rate, and temperature dependence of flowstress in ultra-fine grained and nanocrystalline Cu and Al: synthesis, experiment,and constitutive modeling, Int. J. Plast. 25 (2009) 715–732.

[18] T.R. Malow, C.C. Koch, Grain growth in nanocrystalline iron prepared by mechanicalattrition, Acta Mater. 45 (1997) 2177–2186, http://dx.doi.org/10.1016/S1359-6454(96)00300-X.

[19] D.S. Gianola, S. Van Petegem, M. Legros, S. Brandstetter, H. Van Swygenhoven, K.J.Hemker, Stress-assisted discontinuous grain growth and its effect on the deforma-tion behavior of nanocrystalline aluminum thin films, Acta Mater. 54 (2006)2253–2263, http://dx.doi.org/10.1016/j.actamat.2006.01.023.

[20] F.X. Zhao, X.C. Xu, H.Q. Liu, Y.L. Wang, Effect of annealing treatment on the micro-structure and mechanical properties of ultrafine-grained aluminum, Mater. Des.53 (2014) 262–268, http://dx.doi.org/10.1016/j.matdes.2013.06.075.

[21] L. Zhou, J. Guo, G. Li, L. Xiong, S. Wang, C. Li, Investigation of annealing behavior ofnanocrystalline NiAl, Mater. Des. 18 (1997) 373–377.

[22] M. Akbarpour, H. Kim, Microstructure, grain growth, and hardness during annealingof nanocrystalline Cu powders synthesized via high energy mechanical milling,Mater. Des. 83 (2015) 644–650.

[23] L.L. Shaw, H. Luo, Deformation behavior andmechanisms of a nanocrystallinemulti-phase aluminum alloy, J. Mater. Sci. 42 (2007) 1415–1426.

[24] K. Zhang, J.R. Weertman, J.A. Eastman, Rapid stress-driven grain coarsening in nano-crystalline Cu at ambient and cryogenic temperatures, Appl. Phys. Lett. 87 (2005)61921, http://dx.doi.org/10.1063/1.2008377.

[25] J.M. Tao, X.K. Zhu, R.O. Scattergood, C.C. Koch, The thermal stability of high-energyball-milled nanostructured Cu, Mater. Des. 50 (2013) 22–26, http://dx.doi.org/10.1016/j.matdes.2013.02.083.

[26] Y. Huang, A. Menovsky, F. De Boer, Calorimetric analysis of the grain growth innanocrystalline copper samples, Nanostruct. Mater. 2 (1993) 587–595.

[27] S. Simões, R. Calinas, P.J. Ferreira, M.T. Vieira, F. Viana, M.F. Vieira, Effect of annealingconditions on the grain size of nanocrystalline copper thin films, Mater. Sci. Forum587–588 (2008) 483–487, http://dx.doi.org/10.4028/www.scientific.net/MSF.587-588.483.

[28] S. Simões, R. Calinas, P.J. Ferreira, F. Viana, M.T. Vieira, M.F. Vieira, TEM and SEM in-situ annealing of nanocrystalline copper thin films, Microsc. Microanal. 14 (2008)49–52, http://dx.doi.org/10.1017/S1431927608089368.

[29] K.A. Darling, M.A. Tschopp, B.K. VanLeeuwen, M.A. Atwater, Z.K. Liu, Mitigating graingrowth in binary nanocrystalline alloys through solute selection based on thermo-dynamic stability maps, Comput. Mater. Sci. 84 (2014) 255–266, http://dx.doi.org/10.1016/j.commatsci.2013.10.018.

[30] M. Saber, H. Kotan, C.C. Koch, R.O. Scattergood, A predictive model for thermody-namic stability of grain size in nanocrystalline ternary alloys, J. Appl. Phys. 114(2013) 103510, http://dx.doi.org/10.1063/1.4821040.

[31] T. Chookajorn, H.A. Murdoch, C.A. Schuh, Design of stable nanocrystalline alloys, Sci-ence 337 (2012) 951–954, http://dx.doi.org/10.1126/science.1224737.

[32] H.A. Murdoch, C.A. Schuh, Estimation of grain boundary segregation enthalpy andits role in stable nanocrystalline alloy design, J. Mater. Res. 28 (2013) 2154–2163.

[33] T.J. Rupert, The role of complexions inmetallic nano-grain stability and deformation,Curr. Opin. Solid State Mater. Sci. (2016), http://dx.doi.org/10.1016/j.cossms.2016.05.005.

[34] C.C. Koch, R.O. Scattergood, M. Saber, H. Kotan, High temperature stabilization ofnanocrystalline grain size: thermodynamic versus kinetic strategies, J. Mater. Res.28 (2013) 1785–1791, http://dx.doi.org/10.1557/jmr.2012.429.

[35] K.A. Darling, M.A. Tschopp, R.K. Guduru, W.H. Yin, Q. Wei, L.J. Kecskes, Microstruc-ture and mechanical properties of bulk nanostructured Cu–Ta alloys consolidatedby equal channel angular extrusion, Acta Mater. 76 (2014) 168–185, http://dx.doi.org/10.1016/j.actamat.2014.04.074.

[36] K.A. Darling, M. Rajagopalan, M. Komarasamy, M.A. Bhatia, B.C. Hornbuckle, R.Mishra, K.N. Solanki, Extreme Creep Resistance in a Microstructurally Stable Nano-crystalline Alloy, 2016, http://dx.doi.org/10.1038/nature19313.

[37] K.A. Darling, M. Rajagopalan, M. Komarasamy, M.A. Bhatia, B.C. Hornbuckle, R.S.Mishra, K.N. Solanki, Extreme creep resistance in a microstructurally stable nano-crystalline alloy, Nature 537 (2016) 378–381, http://dx.doi.org/10.1038/nature19313.

[38] G.P. Pun, K. Darling, L. Kecskes, Y. Mishin, Angular-dependent interatomic potentialfor the Cu–Ta system and its application to structural stability of nano-crystalline al-loys, Acta Mater. 100 (2015) 377–391.

[39] R.K. Koju, K.A. Darling, L.J. Kecskesz, Y. Mishin, Zener pinning of grain boundariesand structural stability of immiscible alloys, JOM (2016) 1–9, http://dx.doi.org/10.1007/s11837-016-1899-9.

[40] S. Plimpton, Fast parallel algorithms for short-range molecular dynamics, J. Comput.Phys. 117 (1995) 1–19, http://dx.doi.org/10.1006/jcph.1995.1039.

[41] A. Stukowski, Visualization and analysis of atomistic simulation data withOVITO—the Open Visualization Tool, Model. Simul. Mater. Sci. Eng. 18 (2010)15012, http://dx.doi.org/10.1088/0965-0393/18/1/015012.

[42] J.D. Eshelby, The determination of the elastic field of an ellipsoidal inclusion, and re-lated problems, Proc. R. Soc. Lond. Math. Phys. Eng. Sci. 241 (1957) 376–396, http://dx.doi.org/10.1098/rspa.1957.0133.

[43] T. Rojhirunsakool, K.A. Darling, M.A. Tschopp, G.P.P. Pun, Y. Mishin, R. Banerjee, L.J.Kecskes, Structure and thermal decomposition of a nanocrystalline mechanicallyalloyed supersaturated Cu–Ta solid solution, MRS Commun. 5 (2015) 333–339.

[44] J.M. Howe, Interfaces in Materials: Atomic Structure, Thermodynamics and Kineticsof Solid-Vapor, Solid-Liquid and Solid-Solid Interfaces, Wiley-Interscience, 1997.

[45] J. Ribis, Y. de Carlan, Interfacial strained structure and orientation relationships ofthe nanosized oxide particles deduced from elasticity-driven morphology in oxidedispersion strengthened materials, Acta Mater. 60 (2012) 238–252, http://dx.doi.org/10.1016/j.actamat.2011.09.042.

[46] H. Van Swygenhoven, P.M. Derlet, A.G. Frøseth, Nucleation and propagation of dislo-cations in nanocrystalline fcc metals, Acta Mater. 54 (2006) 1975–1983, http://dx.doi.org/10.1016/j.actamat.2005.12.026.

[47] A. Alavudeen, N. Venkateshwaran, J.T.W. Jappes, A Textbook of Engineering Mate-rials and Metallurgy, Firewall Media, 2006.

[48] M.F. Ashby, R. Bullough, C.S. Hartley, DislocationModelling of Physical Systems: Pro-ceedings of the International Conference, Gainesville, Florida, USA, June 22–27,1980, Elsevier, 2013.

[49] P. Manohar, M. Ferry, T. Chandra, Five decades of the Zener equation, ISIJ Int. 38(1998) 913–924.

[50] M. Bhatia, M. Rajagopalan, K. Darling, M. Tschopp, K. Solanki, The role of Ta ontwinnability in nanocrystalline Cu–Ta alloys, Mater. Res. Lett. 1–7 (2016).

[51] J.X. Zhang, J.C. Wang, H. Harada, Y. Koizumi, The effect of lattice misfit on the dislo-cation motion in superalloys during high-temperature low-stress creep, Acta Mater.53 (2005) 4623–4633, http://dx.doi.org/10.1016/j.actamat.2005.06.013.

185M. Rajagopalan et al. / Materials and Design 113 (2017) 178–185