Contents lists available at ScienceDirect

Materials Science & Engineering A

journal homepage: www.elsevier.com/locate/msea

Enhanced hardness via interface alloying in nanoscale Cu/Al multilayers

X.Z. Weia, Q. Zhoub, K.W. Xuc, P. Huangc,⁎, F. Wanga,d,⁎⁎, T.J. Lua,e

a State Key Laboratory for Strength and Vibration of Mechanical Structures, Xi’an Jiaotong University, Xi’an 710049, Chinab State Key Laboratory of Solidification Processing, Center of Advanced Lubrication and Seal Materials, Northwestern Polytechnical University, Chinac State Key Laboratory for Mechanical Behavior of Material, Xi’an Jiaotong University, Xi’an 710049, Chinad Shaanxi Key Laboratory of Environment and Control for Flight Vehicle, Xi’an Jiaotong University, Xi’an 710049, ChinaeMOE Key Laboratory for Multifunctional Materials and Structures, Xi’an Jiaotong University, Xi’an 710049, China

A R T I C L E I N F O

Keywords:Cu/Al intermetallic compoundsNanoscale multilayerMicrostructure characterizationHardness

A B S T R A C T

Ultrahigh hardness (yield strength) was achieved in magnetron sputtering nanoscale Cu/Al multilayers uponannealing. The microstructure and mechanical properties of the multilayers were systematically investigated byX-ray diffraction, transmission electron microscopy, energy dispersive X-ray spectroscopy and nanoindentation.Annealing promoted diffusion of Cu and Al atoms in the interfaces and the sharp interface turned to mix,resulting in the formation of Cu/Al intermetallic compounds and its deformation at nanoscale. The Cu/Al in-termetallic compounds mainly including Al2Cu grew toward to Al layers and would reducing the effective lengthbetween the reduced adjacent layers. As the annealing temperature was increased from 100 °C to 500 °C, variouskinds and larger size Cu/Al intermetallic compounds emerged, causing the hardness to first increase, reaching anunusually high peak (never reached before in other thin metallic multilayer systems), and then remain nearlyunchanged. The physical mechanisms underlying such remarkable enhancement were explored in terms of in-terface alloying, reduced layer thickness and grain size effects.

1. Introduction

Nanoscale metallic multilayers exhibit unique ultrahigh strength (orhardness) due mainly to the presence of abundant heterogeneous in-terfaces [1–3]. Typically, the yield strength could approach 1/2–1/3 oftheoretical strength, i.e., on the order of E/50, E being the Young'smodulus [4]. The hardness and yield strength of a thin metallic mul-tilayer are in general influenced by 1ayer thickness, epitaxy orientation,interfacial structure and fabrication method. It had been reported thatthe strength and hardness of nano-multilayered systems could increaseremarkably as the layer thickness dropped from micrometer- to nan-ometer-scale. For example, multilayers composed of Cu and Ni in thebulk possessed a yield strength on the order of 10–50MPa, whereas thesame multilayers of only a few nanometers in thickness exhibited apeak strength in excess of 2 GPa [5–8]. The enhancement in strengthstopped (peaked) when the thickness of the Cu/Ni multilayer droppedfurther to 1–2 nm.

Three different kinds of models had been proposed to characterizethe yield strength of nanoscale metallic multilayers. The first was theHall-Petch model based on the Hall-Petch scaling law, ∝ −σ h 1/2, whereh is the individual layer thickness [9–12]. This model was built uponthe assumption that dislocation pileups could be treated as a

continuum, which is valid at relatively large scales (50–200 nm). Thesecond was the confined layer slip (CLS) model based on the glide ofsingle Orowan-type loops bounded by two interfaces [4,13–15]. Thismodel was developed to explain the increase in strength with de-creasing layer thickness at smaller length scales (10–50 nm) where theHall–Petch model failed. The third is the interface crossing model basedon the interface crossing mechanism. When the layer thickness wasdecreased further to 1–2 nm, the strength approached a peak implying achange in deformation mechanism from CLS to interface crossing, whilethe CLS model tended to overestimate the strength. For thin metallicmultilayer systems, the interface barrier to slip transmission is depen-dent on interfacial structure rather than layer-thickness. Also, interfaceorientation affects the strength significantly. For example, a {111}multilayer with a [111]Cu/[111]Ni orientation relationship normal tothe interfaces [16] exhibited a higher/lower strength than that of a{001} multilayer having a strong cube-on-cube or [001]Cu/[001]Niorientation relationship [17], despite the two multilayers had similarlayer thickness and interface sharpness [7,17].

In addition to adjusting individual layer thickness and interfacemicrostructure, nanoscale metallic multilayers could also be hardenedby alloying via annealing [18,19], laser treatment [20], ion implanta-tion [21], ion irradiation [22] and explosive joining [23]. For instance,

https://doi.org/10.1016/j.msea.2018.04.065Received 11 February 2018; Received in revised form 13 April 2018; Accepted 16 April 2018

⁎ Corresponding author.⁎⁎ Corresponding author at: State Key Laboratory for Strength and Vibration of Mechanical Structures, Xi’an Jiaotong University, Xi’an 710049, China.E-mail addresses: huangping@mail.xjtu.edu.cn (P. Huang), wangfei@mail.xjtu.edu.cn (F. Wang).

Materials Science & Engineering A 726 (2018) 274–281

Available online 18 April 20180921-5093/ © 2018 Elsevier B.V. All rights reserved.

T

http://www.sciencedirect.com/science/journal/09215093https://www.elsevier.com/locate/mseahttps://doi.org/10.1016/j.msea.2018.04.065https://doi.org/10.1016/j.msea.2018.04.065mailto:huangping@mail.xjtu.edu.cnmailto:wangfei@mail.xjtu.edu.cnhttps://doi.org/10.1016/j.msea.2018.04.065http://crossmark.crossref.org/dialog/?doi=10.1016/j.msea.2018.04.065&domain=pdf

annealed multilayers with negative mixing of enthalpy had a higherhardness than co-sputtering alloy films [24,25]. However, at present,solid solution and intermetallic induced hardening mechanisms remainelusive. To explore the physical mechanisms underlying the strengthenhancement, characterizing the microstructures of nano-multilayeredfilms and their evolutions during annealing is of primary importance.

Length scale also affected the thermal stability of a metallic multi-layer [26]. Previous works on Ti/Ni multilayers suggested scientificmodulation design was important to acquire desired phase composi-tions [26] and optimal alloying [27]. However, there is yet a compre-hensive study devoted to investigating the effect of annealing tem-perature on the alloying and mechanical properties of metallicmultilayers. Besides composition, the complicated and convertible in-terface structure between metal and intermetallic played a key role indetermining the alloying degree, microstructure, strength and ductilityof the multilayers. For instance, our group [25] investigated the effectof layer thickness and alloy interface on the mechanical properties andmicrostructural evolution of Cu/Al multilayers. It was demonstratedthat chemical composition could significantly modulate the mechanicalbehavior of the alloy.

Existing studies about Cu/Al multilayers focused on one or moreaspects of microstructure, chemical composition, and mechanicalproperties. However, the effect of annealing on interfacial micro-structures and mechanical properties is yet examined, especially theinfluence of interfacial microstructural evolution on mechanical prop-erties after annealing.

In our experience, the hardness of a Cu/Al multilayer annealed atdifferent temperatures was higher than that of the as-deposited multi-layer, which is similar to the results mentioned above. In the presentwork, the diffusion phenomenon of Cu/Al bonding interface was ob-served. Two key issues of annealing induced hardness enhancement inCu/Al multilayers were also explored. First, annealing accelerate thediffusion of Cu and Al atoms，resulting to the consequent solid solutionand intermetallic precipitation strengthening the Cu/Al multilayers,would it further lead to the formation of Cu/Al intermetallic com-pounds with tailored composition, thus contributing to the hardness ofCu/Al multilayers? Second, how would the occurrence of Cu/Al reac-tions depend on the modulation length scale of Cu/Al multilayers?Finally, based on the measured hardness of the nanoscale metallicmultilayers, the effects of grain boundary (GB) on multilayer mechan-ical properties were analyzed.

2. Experimental procedures

2.1. Synthesis and microstructural characterization

Multilayers of Cu and Al were magnetron sputtered on Si (100)substrates at room temperature with a total thickness of 1200 nm. Thethickness of individual Cu layer was set as 20 nm while that of Al as40 nm. The base pressure prior to sputtering was 6.3× 10−5 Pa. Thechamber was evacuated to a base pressure of< 5×10−6 Torr prior todeposition and maintained at a working pressure of 4.6× 10−3 Torrwith argon pumped in during sputtering. The DC (direct current) powerof sputtering was set at 100W for Al, providing a deposition rate of4.6 nm/min. The AC (alternating current) power of sputtering was set at100W for Cu, providing a deposition rate of 11 nm/min. Subsequently,the as-sputtered sample was annealed in encapsulated vacuum quartztube with a pressure lower than 1× 10−5 Torr at 100–500 °C for30min. After annealing, samples were cooled down to room tempera-ture. The microstructure of annealed Cu/Al multilayers were in-vestigated by X-ray diffraction (XRD) and transmission electron mi-croscopy (TEM). The XRD experiments were performed on Rigaku D/max 7000 diffractometer with Cu K radiation (λ=0.154056 nm) atroom temperature. The modulation and cross-sectional microstructuresof Cu/Al multilayers both as-deposited and annealed (at 100 °C，200 °C，400 °C) were evaluated by high resolution transmission

electron microscopy (HRTEM) analysis via JEM-2100F operating at200 kV. Upon manual polishing, cross-sectional TEM samples wereprecisely prepared via Gatan Precision Ion Polishing System 691 byusing Ar ion.

2.2. Nanoindentation test

Nanoindentation was performed using a Nano Indenter®XPInstrument with a diamond Berkovich tip (load resolution: 50 nN; dis-placement resolution: 0.01 nm) via CSM (Continuous StiffnessMeasurement). For hardness measurement, every sample was testedsixteen times at a fixed strain rate of 0.05 s−1 to ensure the reliability oftest data. The indentation depth was 200 nm, limited to 20% of the totalfilm thickness to avoid substrate effects. The hardness was evaluatedfrom load versus indentation depth curves using the Oliver and Pharrmethod, assuming a Poisson ratio of 0.3 for both as-deposited and an-nealed multilayers.

3. Results

3.1. Microstructure characterization

3.1.1. XRD characterizationPhase constitutions of as-deposited and annealed Cu/Al multilayers

were examined via XRD analysis as shown in Fig. 1. For as-depositedCu/Al multilayers, Cu peaks at (111), (200), (220) and Al peaks at(111), (311), (220) were identified in Fig. 1(a), wherein the peak in-tensities of (111) for both Cu and Al were much higher than the otherpeaks. After annealing at 100 °C, strong Al2Cu (110) peak emergedwhile Cu (111) peak separated into two peaks, i.e., peaks of Al2Cu(112) and Al2Cu (310), as shown in Fig. 1(b). This result clearly re-vealed the formation of Cu/Al intermetallic compounds upon an-nealing. By increasing the annealing temperature to 200 °C, the resultsof Fig. 1(c) demonstrate that the peak of Al2Cu (310) in Fig. 1(b) dis-appeared while the peak of Al4Cu9 (330) appeared; and the intensity ofAl (111) peak decreased dramatically compared with that shown inFig. 1(b). As shown in Fig. 1(d), increasing further the annealing tem-perature to 350 °C led to enhanced intensity of Al2Cu (110) peaks andnumerous types of Cu/Al intermetallic compounds, such as AlCu3 andthe like. When increasing the annealing temperature to 400 °C and450 °C as shown in Fig. 1(e) and (f), the XRD patterns indicated moreAl-Cu peaks while the intensity of Al2Cu (110) peak exhibited thestrongest texture.

3.1.2. TEM characterizationFig. 2 presented the cross-sectional transmission electron micro-

scopy (TEM) micrographs and selected area diffraction (SAD) patternsof the Cu/Al multilayers. The lamellar structure was obvious and theinterfaces between the layers were sharp in Fig. 2(a). SAD patterns onthe right showed a polycrystalline ordered structure composed of Cuand Al. The cross-sectional microstructure and SAD ring of 100 °C an-nealed Cu/Al multilayer was shown in Fig. 2(b). Obvious lamellarstructure and straight interfaces were observed. Some grains with anaverage grain size 15 nm (Table 1) were located at the interface be-tween Cu layer and Al layer and extended to the Al layer. Also, from theSAD ring shown in Fig. 2(b), Al2Cu (110) was detected, revealing Cuand Al atoms diffused across the interface upon annealing. As shown inFig. 2(c), more grains had a grain size larger than 28 nm (Table 1), andmore kinds of Al2Cu appeared when the annual temperature was in-creased to 200 °C. With regard to 400 °C annealed Cu/Al multilayer, asshown in Fig. 2(d), the lamellar structure was less clear compared withthat at lower annealing temperature, as more grains with an averagegrain size of 67 nm (Table 1) were detected. SAD pattern on the right ofFig. 2(d) showed that more Cu/Al intermetallic compounds emergedinstead of Cu and Al. In detail, Fig. 3(a) with fast Fourier transform(FFT) analysis displayed the structure of Al2Cu when the Cu/Al

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

275

multilayer was annealed at 100 °C. As the annual temperature was in-creased to 200 °C, more Al2Cu emerged, such as (111) as shown inFig. 3(b). At 400 °C, more kinds of Cu/Al intermetallic compounds wereformed, e.g., Al2Cu, Al4Cu9, Al3Cu and AlCu, as verified by FFT ana-lysis in Fig. 3(c). From the EDX elemental maps shown in Fig. 4, theinterfaces in a Cu/Al multilayer annealed at 100 °C remained relativelyclear. Nonetheless, when the annealing temperature was increased to200 °C, the interfaces changed to larger mixing, and could no longerremain clear. The grain sizes of the Cu/Al intermetallic compounds andsingle crystal grain (Cu and Al) size are summarized in Table 1.

3.2. Hardness

Fig. 5 plotted the hardness of Cu/Al multilayers as a function ofannealing temperature up to 500 °C. Relative to as-deposited multi-layers, the hardness increased dramatically as the annealing tempera-ture was lower than ~ 300 °C, and leveled off at higher annealingtemperatures. Specifically, the hardness of as-deposited Cu/Al multi-layers was 5.8 GPa, while a maximum hardness of ~ 9.0 GPa wasachieved in Cu/Al multilayers annealed at 300 °C or even higher tem-peratures. As shown in Fig. 6, this extremely high hardness is higherthan all the CuAl alloys [28–34] and the Cu/Al multilayers[14,16,25,35–37] reported in the open literature.

4. Discussions

The ultrahigh hardness derived in the present Cu/Al multilayers wasmainly achieved via interface alloying, which effectively enhanced thestrength of the multilayers. Physical mechanisms underlying suchstrengthening were discussed next.

4.1. Microstructure

Cu/Al multilayers consisting of alternating nanocrystalline Culayers and nanocrystalline Al layers had been synthesized, and it wasdemonstrated that the structure and phase composition of the interfacebetween Cu and Al layers could be tailored by changing the annealingtemperature.

As-deposited Cu/Al multilayers retained sharp layers structure afterthe magnetron sputtering process. From the X-ray diffractograms (XRD)(as shown in Fig. 1(a)) and SAD patterns (as shown in Fig. 2(a)), onlythe presence of diffraction peaks of Cu and Al were revealed, thus in-dicating that no Cu/Al intermetallic compounds emerged during mag-netron sputtering. In contrast, upon annealing at 100 °C, XRD(Fig. 1(b)) revealed the presence of diffraction peaks of Al2Cu(110), asverified by the SAD patterns of Fig. 2(b). This implied that Cu/Al in-termetallic compounds were formed during annealing, accompanied bypronounced intermixing along layer interfaces and recrystallizationprocess between Cu and Al. Increasing further the annealing

Fig. 1. XRD spectra of Cu20/Al40 multilayers with different annealing temperature: (a) As-deposited, (b) annealed at 100 °C, (c) annealed at 200 °C, (d) annealed at350 °C, (e) annealed at 400 °C, (f) annealed at 450 °C.

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

276

Fig. 2. Bright-field cross-sectional TEM micrographs showing microstructure of the Cu/Al multilayers and the corresponding selected area diffraction patterns(SADPs) with different annealing temperatures (a) as-deposition, (b) 100 °C, (c) 200 °C and (d) 400 °C.

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

277

temperature led to more kinds of Cu/Al intermetallic compounds, suchas Al2Cu (112)/(310) and Al4Cu9 (330), which was verified by XRDand SAD patterns, respectively. Careful examination of Fig. 1(c) re-vealed that Al2Cu was the first to emerge, suggesting that the formationof Al2Cu was easier than the other Cu/Al intermetallic compounds [38]as shown in Fig. 1. In general, high symmetry alloys with small unitcells will easily crystallize, while low symmetry phases are less likely toform due to long-range diffusion. Therefore, Al2Cu dimorphs in tetra-gonal crystal system have smaller unit cells (P4/mmm, three atoms perunit cell, 44.06 Å3; 4/mcm, 12 atoms per unit cell, 179.52 Å3) thanAl4Cu9 (P4-3m, 52 atoms per unit cell, 660.05 Å3) and AlCu (C12/ml,20 atoms per unit cell, 280.62 Å3), and were anticipated to appear firstduring annealing [39].

From the cross-sectional TEM characterization of Fig. 3(a), it wasshown that the alloys were nucleated in the interface and subsequentlyextended to the Al layer. Firstly, grain boundary mediated diffusion isthe dominated mechanism for diffusion in nanocrystalline materials.The grain boundary is a high-energy structure and easily to change, thevacancy formation energy is lower than other structure in Cu/Al mul-tilayers. In the early process of inter-diffusion, the Al(Cu) solid solutioncan be first formed when Cu diffused to Al layer and the Cu(Al) solidsolution can be formed when Al diffused to Cu layer. The solubilitylimit of Cu in Al is ~ 0.15 at% in the temperature of ~ 723 K, whereasthe maximum solubility limit of Al in Cu in the identical temperaturerange is ~ 18 at%. Since the solubility limit of Cu in Al is almost twoorders of magnitude less than that of Al in Cu, the Al(Cu) solid solution

would expect to saturate first. So the Cu/Al intermetallic compoundsfirst formed in the Al side at the interface, and the main components areAl2Cu for the concentration of Cu in the content of 0–34.8 at%. With thediffusion further, the Cu (Al) solid solution was saturated, and the Cu/Al intermetallic compounds form at the Cu(Al) solid solution–Al2Cuinterface, and the main components are Al4Cu9 for the concentration ofCu in the content of 59.2–100 at% [40]. The diffusion of Al-Cu system iscontrolled by vacancy mechanism. Cu and Al have different meltingpoints of 1083 °C and 660 °C, respectively. As diffusion is more rapid inlower melting point metals, the diffusion in Al is more pronounced thanCu. In other words, Al atoms are easily to move and the vacancy for-mation energy of Al is lower [41,42]. The inter-diffusion coefficient andthe self-diffusion coefficient of Cu in the aluminum solid solution wereboth determined as about 5× 10−10 cm2/s at 500 °C. The inter-diffu-sion coefficient and the self-diffusion coefficient of Al in copper solidsolution were about 1× 10–13cm2/s at 500 °C [43]. The diffusion rateof Cu in aluminum solid solution is higher than that of Al in coppersolid solution though the self-diffusion of Al is faster, resulting to moreCu/Al intermetallic compounds forming in Al layer. So the lower so-lubility limit Cu in Al can promote the formation of Cu/Al intermetalliccompounds in the Al layer first, and the higher diffusion rate of Cu inaluminum solid solution can promote more formation of Cu/Al inter-metallic compounds in Al layer. Secondly, as Al atom stacking sequencein Al(111) and Al2Cu(110) is very similar, the atomic stacking of eight(110) planes of Al2Cu in a unit cell is … Cu AlA AlB Al A Cu AlC AlDAlC …[44]. And the atomic stacking of (111) planes of Al can be ob-tained by stacking the two layers -AlA AlB (or, equivalently, AlC AlD)into one layer with almost no extra vertical movement, requiring onlysmall energy to form the new interface. However, the atomic stacking ofCu is much different from that of Cu in Al2Cu. In this case, large energyis needed to change Cu atomic stacking in Cu structure to Cu atomicstacking in Al2Cu structure, implying more energy is needed in theformation of interface between Cu and Al2Cu. Consequently, the in-terface between Al and Al2Cu is easier to form compared with that

Table 1Average grain size and alloy average grain size obtained from Dark-field cross-sectional TEM micrographs.

Annealing temperature (°C) As-deposited 100 200 400

Pure Al and Cu grain size (nm) 17 20 29 50Intermetallic grain size (nm) 0 15 28 67

Fig. 3. Bright-field cross-sectional TEM micrographs show the microstructure of the Cu/Al multilayers with different annealing temperature (a) 100 °C, (b) 200 °Cand (c) 400 °C. Insets are lattice image and the Fourier reconstructed pattern of the area marked with yellow rectangles. (For interpretation of the references to colorin this figure legend, the reader is referred to the web version of this article.)

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

278

between Cu and Al2Cu for less interface formation energy.

4.2. Attribution from interface alloying

Super-hard nanoscale multilayer systems with a positive enthalpy ofmixing (H) have been extensively investigated [3,45,46]. Multilayerswith a positive enthalpy of mixing can sustain sharp interfaces, whichact as a strong barrier to dislocation activities. On the other side,multilayers with negative enthalpy of mixing form intermixing inter-faces, leading to inferior hardness. Since the mixing interface relaxeshigh coherent stresses, the interface could hardly act as an effective

barrier for dislocation motion. However, a high hardness was achievedin Cu/Al multilayers with negative H when the individual layer thick-ness (h) was reduced to as small as 1 nm [25]. The negative H of 1 kJ/mol for Cu and Al could drive the Cu and Al atoms to intermix at theinterface, resulting in interfacial alloying via inter-diffusion of metalatoms [47]. The negative interface energies for Al/Cu would certainlyaffect the interface structure, forcing the constituent atoms to mix, evenforming intermetallic compound at higher annealing temperatures.CuAl alloy had been widely used to tailor the mechanical properties ofstructural alloys due to its low density and high strength. As a result,the formation of Cu/Al intermetallic compounds (hard phases) com-pared to Cu/Al (soft phases) multilayers by annealing is responsible toenhanced strength or hardness. Specifically, the much larger strength ofAl-Cu intermetallic in Al/Cu interface can make a positive contributionto the anti-plastic deformation, and hence increase the hardness of Al/Cu thin films. On the other hand, the formation of Cu/Al intermetalliccompounds in the interface changes the structure of multilayers fromsharp metal/metal interface to alloy/metal system. The alloy structurewould significantly reduce the stacking fault energy landscape of metallayer due to strong Cu–Al affinity at the interface [48]. In this case,stacking faults are easily formed in Cu/Al intermetallic compounds,which can block dislocation transmission in a way similar to grainboundaries, thus enhancing the hardness. During the process of an-nealing, alloys form continuously as Cu and Al atoms diffuse in theinterface, enhancing the strength of the multilayer. Then, as the an-nealing temperature is increased, the thermal driving force also in-creases, which contributes to accelerate the diffusion of Cu and Alatoms in the interface between Cu and Al layers. The nucleation rate ofalloys thence increases with increasing annealing temperature, which

Fig. 4. STEM-EDS elemental maps of the Cu/Al multilayer annealed at (a) 100 °C and (b) 200 °C. (The red color represents Al atom, and the green color represents Cuatom). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Hardness vs. annealing temperature for the Cu/Al multilayer filmssubjected to annealing for a time of 30min.

Fig. 6. (a) Comparison between Cu/Al multi-layers with Al-Cu alloy and other Al-Cu alloys[28–34] with the yield strength versus an-nealing temperature. Data for the other mate-rials are taken from the literature. (b) Com-parison between the peak strength of Cu/Almultilayers and the strength of other multi-layers with Cu or Al elements. Data for theother materials are taken from the literature[14,16,25,35–37].

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

279

enhances the strength (Fig. 7). That is, the attribution from interfacealloying will increase with increasing annealing temperature, with theharder intermetallic making positive contribution to the elevatedoverall hardness. According to the mean field theory [49], the hardnessof a multilayer thin film increases with increasing volume fraction ofthe intermetallic.

4.3. Reduced layer thickness

The present experimental results revealed that Cu/Al intermetalliccompounds formed in the interface between Cu and Al layers and grewup as the annealing temperature was increased. First, interfacial dif-fusion by annealing would lead to the formation of particles of Cu/Alintermetallic compounds in the interface between Cu and Al layers.When the annealing temperature was increased, the growth direction ofCu/Al intermetallic compounds could be divided into two directions:perpendicular to the interfaces and parallel to the interfaces. On oneside, the growth of Cu/Al intermetallic compounds perpendicular to theinterfaces would change the structure of Cu/Al multilayers to Cu/al-loys/Al systems, thus reducing the effective length leff between thereduced adjacent layers, as shown in Fig. 7. According to the Hall-Petchand CLS models, the strength increases significantly as the layerthickness h is decreased when h exceeds 10 nm. It was then not sur-prising that the structure of Cu/alloys/Al with a lower individualthickness upon annealing exhibited a higher strength.

4.4. Grain size effects

Annealing at higher temperatures resulted in obvious grain-coar-sening accompanied with a hardness drop, as shown in Fig. 7. Grain sizewas also found to play an important role in determining the hardness ofmultilayers [50]. As is known, three different kinds of models had beenproposed to predict the yield strength of thin metallic multilayers: theHall-Petch model, the CLS (confined layer slip) model and the InterfaceCrossing model. With the first two models, the predicted strength in-creases significantly as the layer thickness is decreased when the grainsize is larger than a critical value (about several nanometers). However,when the layer thickness drops below the critical value, the yieldstrength would either remain constant or decrease with decreasinggrains size. A peak strength is reached since the deformation me-chanism is changed to interface cutting by single dislocation. Conse-quently, when the grain size of Cu/Al intermetallic compounds in-creases from 0 to larger dimensions, the multilayer strength canincrease first and then decrease.

As shown in the present TEM pictures, the size of Cu/Al

intermetallic compounds increased with annealing temperature. As aresult, the horizontal axis in hardness-annealing temperature plots canbe replaced by the corresponding alloy size. The hardness increasedfirst with alloys nucleation and grain size increasing to a tiny size, thendropped when the grain size increased further.

4.5. Grain boundary stability

Apart from alloys, grain boundary (GB) stability upon annealingshould inevitably contribute to the strengthening process of poly-crystalline multilayers. The inherent GB stability in annealed sampleswas enhanced due to structural relaxation and element segregation,which substantially lowered the GB energy [51]. Stabilized GBs withlesser steps required a very high applied stress for nucleation of ex-tended dislocations [52], as shown in Fig. 7. With increasing annealingtemperature, the inherent GB stability increased and the GB-mediatedprocesses were suppressed, and hence plastic deformation was carriedby extended partial dislocations, resulting in substantial hardening.

When the annealing temperature was relatively low, only a smallamount of alloys was nucleated and the contribution from interfacealloying was minor. The main strengthening mechanism was thereforecontrolled by grain size. When the annealing temperature was in-creased, more alloys appeared, indicating that the contribution frominterface alloying was increasing. As the same time, the grain size of thealloys increased, thus decreasing the hardness. In Fig. 7, let T0 denotethe critical temperature. When T < T0 (a and b), the combined con-tribution from interface alloying, reduced layer thickness and grainsize, and grain boundary stability helped to enhance the hardness.However, when T > T0 (c and d), grain size coarsening would decreasethe hardness (strength) of multilayers. At the same time, the con-tribution from interface alloying, reduced layer thickness and grainboundary stability slightly increased due to a near-finished relaxingprocess. As a result, on the whole, the hardness remained nearly con-stant.

To the best of our knowledge, among all the metallic multilayersreported to date [14,16,25,35–37], the Cu/Al multilayers exhibited thehighest strength at both room and high temperatures, as summarized inFig. 6(b).

5. Conclusions

The microstructure and mechanical properties of Cu/Al inter-metallic compounds introduced by the multilayer method was sys-tematically investigated. Annealing was found to promote the diffusionof Cu and Al atoms in thin Cu/Al multilayers, resulting in nanoscale

Fig. 7. Schematic illustration of the strengthening mechanism of Cu/Al multilayers annealed at different temperature from 100 °C to 500 °C.

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

280

alloys deformation in a large number of interfaces. As the annealingtemperature was increased, more Cu/Al intermetallic compoundsemerged, grew in size, and then extended from the interfaces to the Allayers. First, because the melting point of Al is lower than Cu, diffusionin Al layer is faster, and there are more voids in the Al layer than in theCu layer, it is easier for Cu atoms to enter the Al layer, causing theformation of Cu/Al intermetallic compounds formation in the Al layer.Second, as the Al atom distribution in Al(111) is very similar to that inAl2Cu(110), it is easy for Al (111) layer to turn into Al2Cu(110) whenCu atom are nearby. As the annealing temperature was increased, thehardness of nanoscale Cu/Al multilayers increased first, reaching a peakof about 9.0 GPa, and then remained nearly constant. Three factorscontributed to such strengthening: interface alloying, reduced layerthickness, and grain size effect. Interface alloying and reduced layerthickness prohibited the increasing of hardness upon annealing. Grainsize effects initially enhanced the hardness, but then led to decreasedhardness when the grain size of Cu/Al intermetallic compounds reacheda critical thickness, corresponding to a specific annealing temperature.The combined effect of the three factors caused the hardness to increasefirst, reaching a peak, and then remain constant as the annealingtemperature was increased.

Acknowledgments

This work was financially supported by the National Natural ScienceFoundation of China (51471131, 51271141), and the FundamentalResearch Funds for the Central Universities.

References

[1] A. Dhar, T.L. Alford, APL Mater. 1 (2013) 222107.[2] Michael Nastasi, NATO ASI 233 (1993).[3] M.Z. Wei, L.J. Xu, J. Shi, G.J. Pan, Z.H. Cao, X.K. Meng, Appl. Phys. Lett. 106 (2015)

304.[4] J.D. Embury, J.P. Hirth, Acta Metall. Mater. 42 (1994) 2051–2056.[5] A. Misra, H. Krug, Adv. Eng. Mater. 3 (2001) 217–222.[6] D. Tench, J. White, Metall. Trans. A 15 (1984) 2039–2040.[7] D.M. Tench, J. Electrochem. Soc. 138 (1991) 3757–3758.[8] S. Menezes, D.P. Anderson, ChemInform 21 (1990).[9] P.M. Anderson, C. Li, Nanostruct. Mater. 5 (1995) 349–362.

[10] L.H. Friedman, D.C. Chrzan, Phys. Rev. Lett. 81 (1998) 2715–2718.[11] A. Misra, M. Verdier, Y.C. Lu, H. Kung, T.E. Mitchell, M. Nastasi, J.D. Embury, Scr.

Mater. 39 (1998) 555–560.[12] H. Huang, F. Spaepen, Acta Mater. 48 (2000) 3261–3269.[13] M.A. Phillips, B.M. Clemens, W.D. Nix, Acta Mater. 51 (2003) 3157–3170.[14] A. Misra, J.P. Hirth, H. Kung, Philos. Mag. A 82 (2002) 2935–2951.[15] P.M. Anderson, T. Foecke, P.M. Hazzledine, MRS Bull. 24 (1999) 27–33.

[16] X.Y. Zhu, X.J. Liu, R.L. Zong, F. Zeng, F. Pan, Mater. Sci. Eng. A 527 (2010)1243–1248.

[17] J.S. Carpenter, A. Misra, P.M. Anderson, Acta Mater. 60 (2012) 2625–2636.[18] T. Lehnert, H. Grimmer, P. Böni, M. Horisberger, R. Gotthardt, Acta Mater. 48

(2000) 4065–4071.[19] X. Bing, X. Wang, L. Yun, Appl. Surf. Sci. 253 (2006) 2695–2701.[20] S. Petrović, B. Radak, D. Peruško, P. Pelicon, J. Kovač, M. Mitrić, B. Gaković,

M. Trtica, Appl. Surf. Sci. 264 (2013) 273–279.[21] S. Petrović, D. Peruško, M. Mitrić, J. Kovac, G. Dražić, B. Gaković, K.P. Homewood,

M. Milosavljević, Intermetallics 25 (2012) 27–33.[22] M. Milosavljević, D. Toprek, M. Obradović, A. Grce, D. Peruško, G. Dražič, J. Kovač,

K.P. Homewood, Appl. Surf. Sci. 268 (2013) 516–523.[23] K. Topolski, P. Wieciński, Z. Szulc, A. Gałka, H. Garbacz, Mater. Des. 63 (2014)

479–487.[24] J. Shi, Z.H. Cao, M.Z. Wei, G.J. Pan, L.J. Xu, X.K. Meng, Mater. Sci. Eng. A 618

(2014) 385–388.[25] Q. Zhou, S. Li, P. Huang, K.W. Xu, F. Wang, T.J. Lu, APL Mater. 4 (2016) 012102.[26] R. Gupta, M. Gupta, S.K. Kulkarni, S. Kharrazi, A. Gupta, S.M. Chaudhari, Thin Solid

Films 515 (2006) 2213–2219.[27] P. Bhatt, S.M. Chaudhari, M. Fahlman, J. Phys.-Condens. Matter 19 (2007) 376210.[28] C.S. Tiwary, S. Kashyap, K. Chattopadhyay, Scr. Mater. 93 (2014) 20–23.[29] J. Thangaraj, J. Eng. Appl. Sci. 11 (2016) 4471–4477.[30] L. Ceschini, A. Morri, S. Toschi, S. Johansson, S. Seifeddine, Mater. Sci. Eng. A 648

(2015) 340–349.[31] L. Ceschini, A. Morri, A. Morri, S. Toschi, S. Johansson, S. Seifeddine, Mater. Des. 83

(2015) 626–634.[32] A. Inoue, Prog. Mater. Sci. 43 (1998) 365–520.[33] J.M. Park, N. Mattern, U. Kühn, J. Eckert, K.B. Kim, W.T. Kim, K. Chattopadhyay,

D.H. Kim, J. Mater. Res. 24 (2009) 2605–2609.[34] J. Tao, G. Chen, W. Jian, J. Wang, Y. Zhu, X. Zhu, T.G. Langdon, Mater. Sci. Eng. A

628 (2015) 207–215.[35] E.G. Fu, N. Li, A. Misra, R.G. Hoagland, H. Wang, X. Zhang, Mater. Sci. Eng. A 493

(2008) 283–287.[36] Y. Chen, K.Y. Yu, H. Wang, J. Chen, X. Zhang, Int. J. Plast. 49 (2013) 152–163.[37] Y.P. Li, G.P. Zhang, W. Wang, J. Tan, S.J. Zhu, Scr. Mater. 57 (2007) 117–120.[38] K. Gao, S. Song, S. Li, H. Fu, J. Alloy. Compd. 674 (2016) (477–477).[39] H. Xu, C. Liu, V.V. Silberschmidt, S.S. Pramana, T.J. White, Z. Chen, V.L. Acoff, Acta

Mater. 59 (2011) 5661–5673.[40] Y. Wei, J.L. Li, J.T. Xiong, F.S. Zhang, Eng. Sci. Technol. Int. J. 19 (2015) 90–95.[41] M.J. Gillan, J. Phys.: Condens. Matter 1 (1989) 689–711.[42] Z.D. Popovic, J.P. Carbotte, G.R. Piercy, J. Phys. F: Met. Phys. 4 (1974) 351–360.[43] Y. Funamizu, K. Watanabe, Mater. Trans. JIM 12 (1971) 147–152.[44] Q. Zhou, J. Wang, A. Misra, P. Huang, F. Wang, K. Xu, Int. J. Plast. 87 (2016)

100–113.[45] Q. Zhou, J.Y. Xie, F. Wang, P. Huang, K.W. Xu, T.J. Lu, Acta Mech. Sin. 31 (2015)

319–337.[46] R.G. Hoagland, R.J. Kurtz, C.H.H. Jr, Scr. Mater. 50 (2004) 775–779.[47] C.T. Chan, K.P. Bohnen, K.M. Ho, Phys. Rev. Lett. 69 (1992) 1672–1675.[48] S.K. Yadav, S. Shao, J. Wang, X.Y. Liu, Sci. Rep. 5 (2015) 17380.[49] W.F. Smith, McGraw-Hill, 1986.[50] Q. Zhou, P. Huang, M. Liu, F. Wang, K. Xu, T. Lu, J. Alloy. Compd. 698 (2017)

906–912.[51] J. Hu, Y.N. Shi, X. Sauvage, G. Sha, K. Lu, Sci. Found. China 355 (2017) 1292.[52] V. Yamakov, D. Wolf, S.R. Phillpot, A.K. Mukherjee, H. Gleiter, Nat. Mater. 3 (2004)

43–47.

X.Z. Wei et al. Materials Science & Engineering A 726 (2018) 274–281

281

http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref1http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref2http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref3http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref3http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref4http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref5http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref6http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref7http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref8http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref9http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref10http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref11http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref11http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref12http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref13http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref14http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref15http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref16http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref16http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref17http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref18http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref18http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref19http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref20http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref20http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref21http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref21http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref22http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref22http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref23http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref23http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref24http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref24http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref25http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref26http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref26http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref27http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref28http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref29http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref30http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref30http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref31http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref31http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref32http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref33http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref33http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref34http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref34http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref35http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref35http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref36http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref37http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref38http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref39http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref39http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref40http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref41http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref42http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref43http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref44http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref44http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref45http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref45http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref46http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref47http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref48http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref49http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref49http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref50http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref51http://refhub.elsevier.com/S0921-5093(18)30563-X/sbref51

Enhanced hardness via interface alloying in nanoscale Cu/Al multilayersIntroductionExperimental proceduresSynthesis and microstructural characterizationNanoindentation test

ResultsMicrostructure characterizationXRD characterizationTEM characterization

Hardness

DiscussionsMicrostructureAttribution from interface alloyingReduced layer thicknessGrain size effectsGrain boundary stability

ConclusionsAcknowledgmentsReferences