lable at ScienceDirect

Acta Materialia 103 (2016) 735e745

Contents lists avai

Acta Materialia

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

Alloy design strategies for sustained ductility in Mg-based amorphousalloys e Tackling structural relaxation

K.J. Laws a, b, D. Granata a, J.F. L€offler a, *

a Laboratory of Metal Physics and Technology, Department of Materials, ETH Zurich, 8093 Zurich, Switzerlandb School of Materials Science and Engineering, UNSW, Sydney, NSW 2052, Australia

a r t i c l e i n f o

Article history:Received 14 July 2015Accepted 31 August 2015Available online xxx

Keywords:Metallic glassesAmorphous alloysMagnesiumStructural relaxationCrystallizationThermodynamicsMechanical propertiesDifferential scanning calorimetry

* Corresponding author.E-mail address: joerg.loeffler@mat.ethz.ch (J.F. L€of

http://dx.doi.org/10.1016/j.actamat.2015.08.0771359-6454/© 2015 Acta Materialia Inc. Published by E

a b s t r a c t

Mg-based metallic glasses are promising candidates for use in light-weight, high-specific-strength ap-plications, but their low fracture toughness and rapid room-temperature structural relaxation have so farprevented their deployment. In this study we present various design strategies for tackling structuralrelaxation via studying several Mg-based amorphous alloy systems during room-temperature aging. Ourinvestigations show that relaxation kinetics is strongly affected by solute content, solute type and glasstransition temperature Tg, and an increase in solvent content generally results in lower relaxationenthalpy, faster relaxation rates, prolonged ductility and a reduction of Tg, which increases the alloys'susceptibility to room-temperature crystallization. The results also show that chemical bonding andspecific topologies contribute to relaxation kinetics, where alloys with greater mixing enthalpies andgreater differences in atomic radii between the constituents exhibit greater relaxation enthalpies andrelaxation rates. In conclusion, we present design concepts towards enhancing long-term ductility in Mg-based glasses via a delicate balancing of their chemistry and topology.

© 2015 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Due to their low density, high specific strength and high elasticlimit, magnesium-based alloys are of emerging interest for use inaerospace, automotive and consumer electronics-based applica-tions. Bulk amorphous alloys or bulk metallic glasses are a rela-tively new class of advanced metallic alloys which exhibitextraordinary properties for structural, mechanical and functionalapplications. They exhibit high strength (three times that of crys-talline alloys) which approach the theoretical maximum strength[1], the highest elastic limits of all metallic materials (at least twicethat of regular metals), and improved corrosion resistance. Amor-phous alloys based on platinum and palladium, in particular,appear to possess the highest damage-tolerance of any materialknown to date [2,3].

Since the discovery of the Mg70Zn30 metallic glass in 1977 [4],numerous Mg-based amorphous alloys or bulk metallic glasses havebeen developed. So far the majority of these glass-forming compo-sitions have been based on MgeTMeRE or MgeTMeCa ternarysystems (where TM¼ transitionmetals Cu, Ni, Zn, Ag and RE¼ rare-earthmetals Y, Gd, La, Nd, Ce, etc.) [5e11]. Some of these alloys have

fler).

lsevier Ltd. All rights reserved.

exceptional glass-forming ability with critical casting sizes of up to27mm[11]. Essentially, the key issue precluding the use ofMg-basedglasses for at least small-scale structural applications is their inherentlow fracture toughness or brittle nature. This shortfall in mechanicalperformance is due to (i) the low activation energy of shear banding(which is directlyassociatedwith the chemistryof these glasses), and(ii) the rapid structural relaxation effects theyexhibit [12e15]. Due tothese glasses' low glass transition temperatures (Tg) severe embrit-tlement can occur even at room temperature (corresponding to 0.7Tg) and on relatively short timescales, strongly compromising theirpotential usefulness and commercial viability.

However, there are some examples which indicate a morepositive design direction for these high-potential materials. Ingeneral, the ductility of Mg-based metallic glasses can be improvedsignificantly by the incorporation of ductile crystalline phases[16,17]. Indeed, it has been found that the evolution of specificcrystallites in an amorphous matrix can alleviate embrittlementinduced by structural relaxation [18]. However, a significant vol-ume fraction of crystallites in such a composite structure oftenresults in the deterioration of the superior strength and corrosionproperties of metallic glass; further, composite microstructures areoften difficult to produce, requiring specific cooling rates orspecialized processing techniques [16e20]. Therefore this researchfocuses on design strategies for improving the ductility of fullyamorphous specimens.

Fig. 1. Typical image of an amorphous Mg-based ribbon sample being bent betweentwo micrometer platen faces. The variable d denotes the fracture or buckling distance,which is utilized to evaluate the fracture strain by bending.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745736

Considerable ductility in amorphous ribbons has been reportedby Gu et al. in MgeZneCa alloys with Mg concentrations from 68 to85 at% [5]; by Guo et al. in the MgeCueCa, MgeCueY andMgeNieY systems for alloys with a Mg content >85 at.% [21]; andby Yu et al. in MgeZneCaeYb alloys [22]. Modest ‘bulk’ glassductility has also been observed in Mg-rich MgeNieGd-based[10,23] and MgeNieCa [24] alloy systems, pointing to higher Mg-content as a first step in improving ductility. However, few re-ports offer the details of ductile performance over a prolongedperiod of time. Given that some of these ‘ductile’ alloys cancompletely embrittle at room temperature after 3e4 days[14,25,26], these details and a design approach for alleviatingmetallic glass embrittlement in long-term applications are critical ifthese materials are to be brought to the forefront of applied tech-nology [9].

In this study we present experimental results for a broad rangeof Mg-based amorphous alloy compositions, focusing specificallyon bending ductility or flexural performance over time at roomtemperature. The mechanical performance of these alloys isincrementally correlatedwith their thermal properties and changesin structural characteristics at regular time intervals. Based on theseresults, we devise a strategic direction for the future developmentof Mg-based glasses with extended toughness and service life.

2. Experimental procedure

Mg-based alloys were produced using high-purity elements Mg(99.99%, Cerac), Zn (99.99%, Alfa Aesar), Al (99.999%, Cerac), Cu(99.995%, Praxair), Ni (99.95%, Alfa Aesar), Ca (99.5%, Alfa Aesar),and Y (99.99%, Treibacher Industry). Alloys were prepared by in-duction melting in graphite crucibles. In the case of MgeCueY andMgeNieY elemental Mg was alloyed with eutectic Cu5Y10 andNi5Y10 pre-alloys, which were prepared by arc-melting under high-purity (6N) Ar atmosphere. Amorphous Mg-based ribbons withthicknesses of 40e45 mm were produced using the melt spinningtechnique (Melt spinner HV, Edmund Buehler). Differential scan-ning calorimetry (DSC) measurements were performed on samplesto determine thermal materials parameters such as glass transitiontemperature Tg, primary crystallization temperature Tx, and theenthalpy related to structural relaxation. Measurements were car-ried out in graphite pans using a Seiko DSC 220 CU instrument at aheating rate of 20 Kmin�1 under a flow of high-purity argon. X-raydiffraction (XRD) data were acquired using a STOE diffractometer(STADI P) and a monochromatic CoKa radiation source for MgeZn-based alloys and a CuKa source for Mge[Cu,Ni]-based alloys. Themeasurements were performed in Bragg-Brentano geometry with afixed incoming angle of 20� and using a 140� image plate detector(STOE, IP-PSD). Vickers hardness values were obtained using amicro-hardness testing device (MXT-a, Wolpert).

Bending tests were performed using a digital micrometer setupat room temperature similar to that used in Refs. [13,14,26]. Theribbon samples were clamped between two micrometer platenfaces as shown in Fig. 1. The micrometer screw was turned incre-mentally to smaller distances, resulting in a gradual increase inribbon bending strain. Three modes of ribbon response wereconsidered: ductile, buckling and brittle fracture. Here we definebrittle as a ribbon that fractured completely during bend testing.Buckling ribbons did not completely fracture, but had clearly failedand could not sustain further plastic deformation, and ductile rib-bons exhibited no fracture or buckling behavior and remainedcompletely intact and plastic.

The fracture or buckling distances, d, were determined for eachalloy at specific times after manufacturing. The bending strain, εf =b,for fracture or buckling was evaluated by

εf =b ¼ t=ðd� tÞ; (1)

where t is the thickness of the ribbon. Here, εf =b ¼ 1 or 100% whend ¼ 2t, i.e. when the ribbon has bent completely and doubled itsthickness. Each test was repeated 5 times per alloy at specific timeincrements after manufacture.

Ribbons were held at room temperature for the duration of thestudy and characterized structurally (using XRD), thermo-physically (using DSC) and mechanically (using the bend testingmethod) at specific time intervals after manufacture. Typicaltesting time increments were <1 day, 1 week, 1 month and 3months, but time intervals were adjusted for more relaxation-prone compositions. It was noted that there was some fluctuationin strain-to-failure data in the buckling regime because of diffi-culties in determining the precise buckling point, given that thesesamples did not completely fracture.

3. Results

Table 1 gives the nominal alloy compositions, characteristicthermo-physical properties, relaxation-related parameters, time toembrittlement, and Vickers hardness of the alloys examined in thisstudy. The time to embrittlement is defined as the time atwhich theribbon bending response departs from fully ductile behavior withinthe bounds of the testing schedule. Due to the release of enthalpyassociated with structural relaxation, the glass transition tempera-tures reported here are determined in the later stages of aging.

3.1. Time-dependent bending ductility and hardness

Fig. 2 shows the bending performance of the compositionsstudied as a function of room-temperature aging, where respectivebending strains have been calculated using Equation (1). It illus-trates the influence of different alloy constituents (chemistry) andvarious topologies.

Fig. 2(a) focuses on binary MgeZn metallic glasses showingabrupt embrittlement to the buckling regime for Mg70Zn30 after 2days, whereas Mg75Zn25 remains ductile for 20 days beforebecoming brittle. Fig. 2(b) shows results for MgeZneCa alloys for afixed Ca concentration of 5 at.%. Here the alloy with the highest Mg-content, Mg74Zn21Ca5, remains ductile for the full duration oftesting, whereas those with reduced Mg-content, Mg70Zn25Ca5 andMg67Zn28Ca5, embrittle after 14 and 2 days, respectively. Fig. 2(c)gives data for the MgeZneAl system with a fixed Al-content of

Table 1Nominal compositions and selected thermo-physical properties of amorphous alloy ribbons examined in this study including the relaxation onset temperature (Tr,onset), theglass transition and crystallization onset temperatures (Tg and Tx), the equilibrium change in enthalpy (DHeq), the average enthalpy relaxation time (t), and the Kohlrauschexponent (b) determined at a heating rate of 20 Kmin�1. Also listed are the time to embrittlement tB and the Vickers Hardness (HV).

Composition Tr,onset [K] Tga [K] Tx

a [K] DHeq [J/g] t [days] b tB [days] Hardness [HV]

As-cast 90 days

Mg75Zn25 333 344 359 e e e 20 191± 2.5 263 ± 3.7b

Mg70Zn30 335 352 374 e e e 2 220± 5.4 218± 3.4Mg74Zn21Ca5 336 351 369 7.64 14.106 0.479 >90 206± 6.1 200± 3.0Mg70Zn25Ca5 340 353 370 e e e 14 210± 3.9 218± 3.6Mg67Zn28Ca5 338 359 374 10.07 8.07 0.466 2 223± 1.2 227± 3.3Mg74Zn21Al5 337 348 365 5.52 14.43 0.405 >90 191± 3.2 199± 2.7Mg70Zn25Al5 338 350 377 5.90 14.96 0.375 4 209± 6.2 218± 4.8Mg77Zn20La3c 335 338 366 3.24 11.26 0.434 >90 206± 4.5 203± 3.4Mg72Zn25La3 337 342 374 3.94 5.93 0.286 <1 218± 1.5 232± 3.7Mg85Cu5Y10 359 430 450 10.65 4.56 1.260 1 205 [27] e

Mg65Cu25Y10 e 420 [14] 473 [14] e e e <1 280 [27] e

Mg85Ni5Y10 378 450 [27] 467 [27] 10.99 12.08 0.960 >90 210 [27] e

Mg60Ni30Y10 e 466 [27] 482 [27] e e e <1 e e

a Tg and Tx data are taken at the stage of relaxation of 90 days for all alloys, with the exception of Mg75Zn25, for which the thermal properties were evaluated after 30 days.b This sample was measured at 30 days, at which point it had substantially crystallized.c This sample showed minor evidence of crystallites in the as-cast state.

Fig. 2. Bending strain of Mg-based amorphous ribbons over time and specific failure mode in (a) MgeZn binary alloys; (b) MgeZneCa alloys; (c) MgeZneAl and MgeZneLa alloys;and (d) MgeNieY and MgeCueY alloys.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745 737

5 at.% and the MgeZneLa system with a fixed La-content of 3 at.%.Interestingly, the MgeZneAl alloys do not completely embrittleover the testing duration, with the higher Mg-content alloyMg74Zn21Al5 remaining completely ductile and the Mg70Zn25Al5maintaining some structural integrity in the ductile and bucklingregime thereafter. MgeZneLa alloys appear ductile at considerablyhigher Mg-content (e.g. Mg77Zn20La3), whereas the Mg72Zn25La3alloy exhibits no ductile behavior after 1 day of production. Fig. 2(d)shows bending data for Mg85Cu5Y10 and Mg85Ni5Y10 alloys withconsiderably higher Mg-content. Here, the Cu-containing alloyembrittles within 2 days of production and the bending strain

which the sample can accommodate continues to degrade over theperiod of a week (see also Ref. [14]). This is a notable improvementon data reported for the Mg-lean Mg65Cu25Y10 composition, whichembrittles completely within 2 h of casting [14]. In contrast to theMgeCueY system, the Mg85Ni5Y10 alloy remains completelyductile for the test duration, whereas the Mg-lean Mg60Ni30Y10alloy is brittle immediately after production [27].

It can be seen here that amorphous alloys with higher Mg-content generally exhibit improved ductility within a given alloysystem. In particular cases, some alloys remain fully ductile afterthree months of room-temperature aging, namely the

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745738

Mg74Zn21Ca5, Mg74Zn21Al5, Mg77Zn20La3 and Mg85Ni5Y10 metallicglasses. It is also interesting to note that the Mg74Zn21Al5 andMg85Ni5Y10 amorphous alloys display prolonged ductility wellbeyond the timeframe of this study. It is also evident from Table 1that in the as-cast state, the hardness of thesemetallic glasses tendsto decrease with increasing Mg-content, and hardness generallyincreases over time, which is likely the result of structural relaxa-tion/densification [25,28].

3.2. Calorimetric investigation and XRD analysis

In order to capture important relaxation and crystallizationevents that identify ductile-to-brittle transitions over time, com-plementary DSC and XRD data were also gathered. Both theexothermic relaxation below Tg and the enthalpic height of theglass transition peak are related to the excess structural free volumeof the material. Changes in enthalpy can be linearly correlated withfree volume changes associated with structural relaxation[25,29e35]. Figs. 3e7 show time-incremented DSC and XRD tracesfor the Mg-based amorphous alloys investigated.

Fig. 3(a) shows data for the Mg75Zn25 metallic glass. Here theDSC trace taken at <1 day shows a distinct structural relaxationenthalpy release, between 333 K and 370 K, which quickly di-minishes during 30 days of room-temperature aging, revealing Tg.Within this time frame the alloy undergoes embrittlement, and

Fig. 3. (a) and (c) DSC traces of the Mg75Zn25 and Mg70Zn30 amorphous ribbons, and (b) anMagnification of the region prior to the glass transition and crystallization events associate

after 60 days the glass transition and the first crystallization peakdisappear, indicating the completion of the primary crystallizationreactions. The secondary crystallization reactions at temperaturesabove 500 K have also been altered substantially due to the room-temperature crystallization events of the primary phases at 60 and90 days. Fig. 3(b) shows that the Mg75Zn25 alloy has started crys-tallizing after 7 days of aging and the alloy continues to crystallizeinto a range of equilibrium and metastable phases, namely Mg,Mg51Zn20 and MgZn [36]. It seems that initially this small volumefraction of crystallites does not affect bending performance, andnor does it significantly affect the DSC response up to 30 days.

Fig. 3(c) shows DSC data for the Mg70Zn30 metallic glass. Here asubtle enthalpy release beginning at around 335 K is evident.Following this, distinct differences in the enthalpy release prior toTx between 1 and 90 days are prominent, decreasing in intensityover time. This enthalpy release is not present after 90 days, wherethe inflection related to Tg is now observable. However, these datashow no distinct correlation with the mechanical embrittlementobserved in Fig. 2(a). Fig. 3(d) indicates that the Mg70Zn30 alloyremains amorphous within the resolution of XRD for 90 days.

Fig. 4 shows DSC and XRD data for three MgeZneCa alloys. The<1 day traces highlight a large enthalpy release associated with arelatively high degree of structural relaxation. Here a specific trendin structural relaxation and chemistry emerges, where alloys with ahigher Mg-content have lower Tg and Tx onset temperatures.

d (d) corresponding XRD data taken at specific time increments. Insets in (a) and (c):d with structural relaxation.

Fig. 4. (a)e(c) DSC traces of the Mg74Zn21Ca5, Mg70Zn25Ca5, and Mg67Zn28Ca5 amorphous ribbons, and (d) XRD data of the Mg74Zn21Ca5 ribbon taken at specific time increments.Insets in (a)e(c): Magnification of the region prior to the glass transition and crystallization events associated with structural relaxation.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745 739

Similarly the onset temperature of relaxation prior to Tg in the as-cast state also tends to shift to lower temperatures withincreasing Mg-content. All MgeZneCa metallic glasses remainedamorphous throughout the duration of this study.

Fig. 5(a)e(d) shows DSC and XRD results for Mg74Zn21Al5 andMg70Zn25Al5 amorphous alloys. These alloys exhibit a considerablerelaxation enthalpy release prior to Tg within 7 days of aging. Thiscorrelates well with the mechanical data for Mg70Zn25Al5, whichembrittles substantially after 4 days. However, this is not the casefor Mg74Zn21Al5, which maintains complete ductility over 90 daysand beyond. The total enthalpy release associated with relaxationof these alloys over 90 days is considerably less than that of theMgeZneCa alloys, yet slightly greater than that of the binaryMgeZn alloys with similar Mg-content. Despite the clear shift incomposition and Tx, the onset of the relaxation event in the as-caststate begins at 337 K and 338 K for Mg74Zn21Al5 and Mg70Zn25Al5,respectively, which correlates well with their similar Tg values of348 K and 350 K. Fig. 5(b) reveals that the Mg74Zn21Al5 amorphousalloy devitrified partially between 30 and 90 days of aging, where ashoulder is evident on the amorphous halo at 40e41� 2q after 30days which develops into crystal peaks after 90 days. The majorityof these peaks can be indexed as a-Mg. There is also evidence oftrace amounts of the metastable Mg51Zn20 crystalline phase in the

later stages of aging, which correlates well with similar devitrifi-cation studies of this system [37]. This suggests that the presence ofa-Mg in this alloy may promote ductility.

Fig. 6 shows DSC and XRD results for the Mg77Zn20La3 andMg72Zn25La3 amorphous alloys. These show a very low enthalpyrelease associated with relaxation over 90 days, similar to the bi-nary MgeZn glasses. This is likely a result of the relatively high Mg-content and low (only 3 at.%) solute addition of La. This slowenthalpy release/relaxation corresponds well to the mechanicaltesting results, where the Mg77Zn20La3 alloy remains ductile for theentire 90 days and the Mg72Zn25La3 alloy's ability to accommodatestrain diminishes very slowly over time. The Tg of these alloys is notso evident. For the Mg77Zn20La3 alloy a shallow, reproducibleexothermic reaction attributed to the precipitation of a-Mg (similarto that observed in Mg75Zn25) is evident. Fig. 6(b) indicates thepresence of a-Mg in the as-cast microstructure of the ribbon, whichwas also reported in Ref. [18] for this alloy. This figure also indicatesthat the ribbon contains a trace amount of themetastable Mg51Zn20

phase after 90 days of aging, similar to theMg74Zn21Al5 alloy, whichmay assist in maintaining this alloy's bending ductility. Given therelatively weak Tg signal and low relaxation enthalpy for theseMgeZneLa alloys and the presence of crystallites in Mg77Zn20La3, itcan be assumed that their glass-forming ability is fairly low.

Fig. 5. (a) and (c) DSC traces of the Mg74Zn21Al5 and Mg70Zn25Al5 amorphous ribbons, and (b) and (d) corresponding XRD data taken at specific time increments. Insets in (a) and (c):Magnification of the region prior to the glass transition and crystallization events associated with structural relaxation.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745740

Fig. 7 shows data for the high Mg-content Mg85Cu5Y10 andMg85Ni5Y10 amorphous alloys measured in the as-cast state, andafter several days and >2 years of room-temperature relaxation. Inboth alloys most relaxation occurs rapidly within several days ofproduction, in agreement with previous studies [14,25,26].Despite their relatively high Mg-content, it is expected that thethermal stability of these alloys over time is quite high given theirrelatively high Tg. This is probably a result of the high Y content.DSC traces shown in Fig. 7(b) for the Mg85Ni5Y10 alloy show thatthe relaxation event is almost completely detached from Tg,occurring and completing at a substantially lower temperature.Furthermore, XRD traces of the Mg85Cu5Y10 and Mg85Ni5Y10 alloysshow that these alloys do not crystallize within 2 years of room-temperature aging.

In most cases considerable enthalpy release associated withstructural relaxation prior to Tg and Tx has been observed here,particularly in alloys relaxed for less than one day (generally within1 h of casting). Differences in enthalpy release over time can belinearly correlated with free volume changes associated withstructural relaxation [25,29e35],

DH ¼ aDv; (2)

where DH is the difference in enthalpy released between the as-cast (unrelaxed) and relaxed states, a is a relaxation rate con-stant, and Dv is the free volume concentration change due torelaxation. It has been shown that the kinetics of free-volume

related relaxation in the sub-Tg region can be accuratelydescribed by a stretched exponential relaxation function [31e35],known as the Kohlrausch-Williams-Watts equation,

4ðtÞ ¼ exp

"��tat

�b#; (3)

where ta is the annealing time, t is the average enthalpy relaxationtime, and b is the Kohlrausch exponent. Hence, the enthalpy ofrelaxation for a given time increment and relaxation temperature,DH(ta), is given by Refs. [31e35]:

DHðtaÞ ¼ DHeq

n1� exp

h� ðta=tÞb

io; (4)

where DHeq is the equilibrium value of DH as ta approaches infinityfor a given temperature. Fig. 8 plots the relaxation enthalpy changesobtained from DSC with respect to the as-cast condition over theavailable time periods for all compositions investigated. A stretchedexponential curve was fitted to each data set containing four ormore data points in accordance with Equation (4), allowing thedetermination of characteristic relaxation-related parameters, suchas the equilibrium relaxation enthalpy DHeq, the average enthalpyrelaxation time t, and the Kohlrausch exponent b. These parame-ters allow for a direct quantitative comparison of relaxation phe-nomena among these alloys. The corresponding values are reportedin Table 1.

Fig. 6. (a) and (c) DSC traces of the Mg77Zn20La3 and Mg72Zn25La3 amorphous ribbons, and (b) and (d) corresponding XRD data taken at specific time increments. Insets in (a) and (c):Magnification of the region prior to the glass transition and crystallization events associated with structural relaxation.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745 741

It is found that for a given alloy system the equilibrium enthalpyrelease, DHeq, tends to increase with decreasing Mg-content, whichprobably results from the rearrangement of atomic bonds betweendissimilar atoms (which have a higher bond enthalpy than similaratom bonds) during relaxation in higher solute-content alloys. Theaverage enthalpy relaxation time, t, and Kohlrausch exponent, b,increase with increasing Mg-content for a given alloy system.Despite having considerably higher Tg and Tx values, theMg85Cu5Y10and Mg85Ni5Y10 alloys relax considerably faster than the MgeZn-based alloy systems. These trends are related to the ease of structuralrearrangement due to chemical and topological effects. The relativeamount of heat (enthalpy) released upon relaxation also correlateswith the heat of mixing between constituents and the relativeamount of these constituents present, particularly for the third atomadditions to the MgeZn-based alloys.

4. Discussion

The proliferation and dynamics of shear bands and hence theability of a metallic glass to accommodate strain are essentially aresult of the inherent physical and chemical properties of the glass[28,38,39]. In the following we discuss the critical factors affectingthe mechanical response of theMg-based glasses studied. We focuson the effects of chemistry, topology, structural relaxation andthermal stability. Finally, we present a design concept forenhancing long-term ductility.

4.1. Effects of chemistry on mechanical response

Bonding in metallic glasses is essentially metallic in nature.During deformation, bonds are broken and re-formed, and the easewith which this occurs depends on the specific bond strength, ri-gidity or preferred angularity of these bonds [28,40]. A material'sresponse to deformation can be approached by considering its bulk(K) and shear (G) moduli, which account for atomic bond stretchingor compressing and bond distortion, respectively, and are stronglydependent on the underlying bonding state of the constituents[40]. A correlation between fracture energy and these elasticproperties exists, where amorphous alloys with G/K < 0.41 orPoisson's ratio n> 0.34 are tough and exhibit large plastic strain tofailure, while those with G/K > 0.43 or n< 0.32 are brittle[2,3,26,40e43].

A molecular dynamics study of MgeZn and MgeZneCa metallicglasses by Mahjoub et al. [44] showed how elastic constants aredramatically affected by composition. An increase in Mg-contentdecreased the alloys' shear modulus (and thus increased the Pois-son's ratio), and the shear modulus was also found to be highlysensitive to the Ca content. This was correlated with the electronicdensity of states of the species within the amorphous structure,where minimizing solute elements that exhibit large electroniccharge transfer with the solvent species reduced the alloys' shearmodulus (and thus increased the Poisson's ratio). This effect wasdemonstrated through simulations and experimentally for otheralloy systems and has been qualitatively linked to mixing enthalpy

Fig. 7. (a) and (b) DSC traces of Mg85Cu5Y10 and Mg85Ni5Y10 amorphous ribbons, and(c) corresponding XRD data taken at specific time increments. Insets in (a) and (c):Magnification of the region prior to the glass transition and crystallization eventsassociated with structural relaxation.

Fig. 8. Time dependence of enthalpy release due to room-temperature structuralrelaxation relative to the as-cast state for all alloys examined in this study. Fittingaccording to Equation (4) was performed for all data sets with four or more datapoints, and the resulting fitting parameters are listed in Table 1. The lines forMg70Zn25Ca5 and the binary alloys are guides for the eye.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745742

and electronegativity among the constituents [7,10,21,24,45e49].This phenomenon is reflected in the findings of this study, wherealloys with higher Mg-content and low amounts of high charge-transferring elements such as Ca, La and Y (which have high mixingenthalpies with Mg [50] compared to e.g. Zn or Al) exhibit higherductility in the as-cast and progressively aging states.

Despite having the same Mg-concentration, Mg85Cu5Y10 andMg85Ni5Y10 amorphous alloys show completely differing mechan-ical behavior. Ni has a slightly higher mixing enthalpy with Mgcompared to Cu [48], but, as with other group 10 elements(including Pd and Pt), it can exhibit a reduced oxidation/valencestate of zero (i.e. a filled d-shell of electrons). In terms of chemicalbonding, these elements have the potential to provide the ‘mostmetallic-type’ bonds in terms of charge transfer and bond-anglefreedom, which potentially explains why Pd- and Pt-basedglasses exhibit very high Poisson's ratios and the highest tough-ness of all glassy materials. It is also apparent from this work, andother reports [5,6,18,21], that adding divalent Ca at the expense oftrivalent rare-earth elements (such as Y or La) promotes improve-ments in ductility. Our results also show that completely amor-phous alloys become harder (stronger) with increasing solutecontent, in particular in alloys containing Y and La (see Table 1).

4.2. Effects of structure/topology on mechanical response

The local distribution and ease of re-distribution of free volumegoverns metallic glass deformation characteristics. The macro-scopic plastic deformation of a metallic glass (perceived asductility) is essentially the sum of local strains accommodated via

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745 743

the operation of multiple shear transformation zones (STZs) andthe redistribution of free volume [28,51,52]. Here a structure con-sisting of higher free volume will more readily accommodate localshear events, and free volume is inherent in glass processing(cooling rate/degree of relaxation) and its composition-specifictopology. Atomistic simulations have shown that short-range to-pological and chemical ordering affects shear [53,54]. Shear-bandinitiation and propagation are also found to differ due to differ-ences in local atomic structure and chemistry [55e57]. Glasses thatcontain more ‘liquid-like’ atomic structures (highly distortedpolyhedra or high free-volume configurations) are known tofeature improved ductility and typically have high solvent content[2,48,57e61].

Within the results presented, despite their distinct differencesin chemistry there are also distinct structural/topological differ-ences in the systems investigated due to the varying atomic radii oftheir constituents (Mg ¼ 160 pm, Cu ¼ 127 pm, Ni ¼ 126 pm,Zn ¼ 139 pm, Al ¼ 141 pm, La ¼ 187 pm, Ca ¼ 197 pm [62]). Out-lined by the results of this and other work [6e8,10,13e15,27], Cu-and Ni-containing Mg-alloys (where Cu and Ni have lower coor-dination numbers than Zn or Al) require a considerably greater Mg-content to exhibit ductility. There may also be a similar topologicaleffect when small amounts of larger elements are added, where Caatoms promote ductility at lower Mg-concentrations compared toslightly smaller rare-earth elements [18,21].

4.3. Effects of structural relaxation on mechanical response

The major drawback of structural relaxation in metallic glassesis the progressive annihilation of structural free volume. Theamount and distribution of free volume governs metallic glassplasticity, and hence structural relaxation has a pronounced effecton mechanical response. Structural relaxation is known to promotean increase in K due to a reduction in interatomic spacing[28,41,63]. However, this also leads to an even greater increase in Ghindering STZ operation and thus generating an overall reductionin toughness [26,28,41,63e67]. A reduction in free volumemay alsosuppress shear-banding susceptibility (the number density of shearbands produced by strain). Hence the size of the plastic zone at thecrack tip is reduced, further contributing to embrittlement andlimiting the actual size/thickness at which ductility can be sus-tained in a sample [65e67].

In this work no direct connection was observed betweenrelaxation time, relaxation enthalpy and time to embrittlement.This is an expected outcome, as with progressing structural relax-ation the G/K or n of an amorphous alloy gradually advances to-wards a state less favorable for metallic glass ductility andtoughness. The initial properties of the alloy (which are chemistry-and structure-dependent) and the rate at which relaxation occursdictate the time needed for an alloy to cross the ductile-to-brittletransition. Hence if an alloy does not cross this transition aftercomplete structural relaxation it will theoretically remain ductile.

In extremecases,where the aging temperature is highenough foratomic diffusivity, i.e. relatively close to Tg and Tx, the ongoingrelaxation of the metastable amorphous structure can lead to crys-tallization below Tg [37,68]. The density of an ordered crystalstructure will differ from that of the vitreous solid, and thus when acrystal precipitates within the glass there is a local volumetricchange inducing a localized stress within the amorphous matrixwhich can generatemechanical instability. The type of precipitatingcrystal can also affect the mechanical response of a partially devit-rified glass, whereby a ductile solid-solution phase (e.g. a-Mg) ismuch preferred over a brittle intermetallic phase because of itsability to accommodate strain and effectively ‘absorb’ shear bands.This is illustrated in this study by the progressive bending response

of the Mg75Zn25 alloy, which crystallizes significantly during aging.In this case the precipitation of primarilyMg51Zn20 quickly degradesbending ductility, whereas for Mg74Zn21Al5 and Mg77Zn20La3 thepresence of a-Mg supports ductility in a way similar to other Mg-based composites [16e18].

4.4. Effects of thermal stability on structural relaxation rates andshear-band dynamics

Higher Tg and Tx appear to slow relaxationprocesses and lower analloy's susceptibility to ambient temperature crystallization. Resultsshow that for a given alloy system, increasing Mg-content corre-sponds to a reduction inTg; and forMg75Zn25,Mg70Zn30,Mg74Zn21Al5andMg77Zn20La3 alloys, which exhibit low Tg values, relative degreesof crystallization are observed. Similar trends have been reported forMgeZn [36] and MgeAleCa alloys [69], where crystallite volumefraction after room-temperature aging increaseswith increasingMg-content. Based on the results presented here and those found inliterature, alloys with a Tg < 350 K and Tx < 369 K have a great ten-dency to crystallize at room temperature within less than 90 days.

The rate of structural relaxation is related to the diffusion andannihilation of structural free volume and can be clearly deducedfrom the data in Fig. 8 for the alloys studied here. The relaxationrate is mathematically represented by the Kohlrausch exponent, b,given in Table 1. It has been recognized that b is the consequence ofthe degree of cooperativity in the rearrangement of structural unitswithin a glass [34,70], and b < 1 reflects a broad distribution ofrelaxation times and structural rearrangements rather than a single(Debye-like) relaxation time/event. From a chemistry standpoint,alloys which contain large quantities of rare-earth elements, suchas Mg85Cu5Y10 and Mg85Ni5Y10, and also exhibit high mixing en-thalpies between the components, relax quite rapidly with a b closeto 1. Even for the MgeZn-based alloys, where b is around 0.4, alloyswhich contain 5 at.% Ca and 3 at.% La tend to relax faster than theMgeZn or MgeZneAl alloys with similar Mg-content. Here theevolution of specific chemical short-range order (between ele-ments with more negative heat of mixing, for example) maycontribute to rapid or more cooperative rearrangement sequencesof atoms during relaxation.

From a structural/topological perspective, the Mg85Cu5Y10 andMg85Ni5Y10 alloys share a similar topological structure (both pos-sessing constituents of similar atomic radii), which differs greatlyfrom that of the MgeZn-based glasses. Fig. 8 reveals a strong cor-relation between relaxation rate (or b) and the radius ratios of thesmallest and largest alloy constituents, i.e. the smaller the radiusratio between the largest and smallest elements in the system, thefaster the relaxation rate. Correspondingly, Mg85[Cu,Ni]5Y10 alloysrelax fastest, followed by MgeZneCa, MgeZneLa, MgeZneAl, andthe MgeZn binary. Individual diffusion constants aside, this in-dicates that the specific topologies generated by such atomic sizedistributions may have an inherent geometry-based structural freevolume which affects atomic mobility during relaxation.

From a mechanical standpoint, relationships between thermalstability (or Tg), the activation of shear, and shear-band dynamicsare somewhat ambiguous. Shear-band initiation sites (zones ofincreased free volume) require a high Tg to thermally stabilize themagainst annihilation at room temperature [71]. However, an in-crease in Tg may be accompanied by an unfavorable change inelastic properties (such as that observed when increasing solutecontent in MgeZneCa alloys [44]), making successful shear acti-vation less probable [72]. Stable shearing (slow shear-band dy-namics) is promoted by strong bonds and small shear activationstresses [57]. Although an increase in Tg cannot guarantee an in-crease in bond strength, it is likely to promote an increase in thecritical stress for shear activation, which is unfavorable for stable

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745744

shear-band sliding as it lowers the activation barrier for shear-banddynamics [57,73,74]. This said, the most ductile Mg-based bulkmetallic glasses are MgeNieGd-based (again, notably containingNi) which exhibit significant plastic, serrated flow deformationcharacteristics and the highest reported Tg values [6,10].

Although this study focuses primarily on samples in ribbonform, we recognize that there is a significant trade-off betweenglass-forming ability and ductility. It is well known that significantsolute additions are essential to maintain the high chemical andtopological entropy contributions associated with low criticalcooling rates and high glass-forming ability, which is the antithesisof the requirements for ductility found in this work.

4.5. Design strategy for the development of ductile Mg-basedmetallic glasses

In order to maintain high ductility, even after complete struc-tural relaxation, it becomes clear from this work and from theliterature that a high Mg-content is essential. It is also clear thatsolute elements which have a higher tendency for transferringelectrons or contribute strongly to bond hybridization should beminimized or avoided, whereas those with potentially full electronshells should be utilized where possible. These results also inferthat alloy topologies which contain elements with larger soluteradii similar to Zn or Al (~140 pm) as opposed to Cu or Ni (~126 pm)may have an inherent topological advantage, resulting in signifi-cantly slower relaxation rates and longer relaxation times. Giventhat Tg and thermal stability tend to decrease with increasing Mg-content, alloying of elements with higher melting points is alsopreferred in the attempt to increase thermal stability and thus toreduce structural relaxation rates. Here these underlying charac-teristics have provided sustained ductility in MgeZneCa,MgeZneAl, MgeZneLa, MgeNieY, MgeZneCaeYb [21], andMgeCueCa [22] metallic glass ribbons; and in MgeNieCa [7] andMgeNieGd [6,10] bulkmetallic glasses. However, compositionally adelicate balance between elastic properties, strength and hardness,thermal stability, and glass-forming ability needs to be considered.

Based on these critical aspects, high-potential alloying candi-dates for the development of next-generation Mg-based (bulk)metallic glasses with improved ductility, high thermal stability, andfracture toughness should include combinations of Pd, Pt, Ni, Ca,and/or Yb. A thorough investigation of the glass-forming range, andthermal and mechanical characterization of these novel alloys, iscurrently underway and will be reported elsewhere.

5. Conclusions

Although the room-temperature structural relaxation of Mg-based amorphous alloys may be inevitable, we have identifiedseveral alloy compositions that overcome the embrittling effects ofstructural relaxationover a prolongedperiodof time (90days). TheseincludeMg74Zn21Ca5,Mg74Zn21Al5,Mg77Zn20La3 andMg85Ni5Y10.Wehave seen that several key factors contribute to theextendedductilitylifetime of these and other alloys found in literature. In summary, thesolution to the challenge of ductility versus embrittlement due tostructural relaxation in Mg-based glasses appears to be a matter ofchemistry and topology. Tailoring the elasticmodulus of the alloy forimproving ductility is achieved in two ways: first, through mini-mizing chemical bond potential (i.e. maximizing Mg-content andminimizing solute additions that exhibit a large bond potential/charge transferwithMg); and second throughmaximizing structuralfree volumeby selecting specific compositions andalloying elementsthat provide ‘liquid-like’ structures and inherently high free-volumetopologies. To arrest or dramatically hinder the rate of structuralrelaxation and possible crystallization, thermal stability must be

increased via alloy chemistry, e.g. via reducing the Mg-content andalloying with higher melting point solutes that have higher bondpotentials. These aspects are, however, mutually exclusive and adelicate balance between these and other aspects such as glass-forming ability need to be maintained to find future applicationsfor these exceptional materials.

Acknowledgments

The authors thank Bruno Zberg (now at Straumann AG) andDaniel Ehrler for providing various DSC data taken during theirtenure at ETH Zurich.

References

[1] M.F. Ashby, A.L. Greer, Metallic glasses as structural materials, Scr. Mater. 54(2006) 321e326.

[2] M.D. Demetriou, M.E. Launey, G. Garrett, J.P. Schramm, D.C. Hofmann,W.L. Johnson, R.O. Ritchie, A damage-tolerant glass, Nat. Mater. 10 (2011)123e128.

[3] J. Schroers, W.L. Johnson, Ductile bulk metallic glasses, Phys. Rev. Lett. 93(2004) 255506.

[4] A. Calka, M. Madhava, D.E. Polk, B.C. Giessen, H. Matyja, J. Vander Sande,A transition-metal-free amorphous alloy: Mg.70Zn.30, Scr. Metall. Mater. 1(1977) 65e70.

[5] X. Gu, G.F. Shiflet, F.Q. Guo, S.J. Poon, MgeZneCa bulk metallic glasses withstrength and significant ductility, J. Mater. Res. 20 (2005) 1935e1938.

[6] G.Y. Yuan, K. Amiya, A. Inoue, Structural relaxation, glass-forming ability andmechanical properties of MgeCueNieGd alloys, J. Non Cryst. Solids 351(2005) 729e735.

[7] K.J. Laws, K.F. Shamlaye, J.D. Cao, J.P. Scicluna, M. Ferry, Locating new Mg-based bulk metallic glasses free of rare-earth elements, J. Alloy Compd. 542(2012) 105e110.

[8] A. Inoue, T. Masumoto, Mg-based amorphous alloys, Mater. Sci. Eng. A 173(1993) 1e8.

[9] B. Zberg, P.J. Uggowitzer, J.F. L€offler, MgZnCa glasses without clinicallyobservable hydrogen evolution for biodegradable implants, Nat. Mater. 8(2009) 887e891.

[10] E.S. Park, H.J. Chang, D.H. Kim, Mg-rich MgeNieGd ternary bulk metallicglasses with high compressive specific strength and ductility, J. Mater. Res. 22(2007) 334e338.

[11] Q. Zheng, J. Xu, E. Ma, High glass-forming ability correlated with fragility ofMgeCu(Ag)eGd alloy, J. Appl. Phys. 102 (2007) 113519.

[12] X.K. Xi, D.Q. Zhao, M.X. Pan, W.H. Wang, Y. Wu, J.J. Lewandowski, Fracture ofbrittle metallic glasses: brittleness or plasticity, Phys. Rev. Lett. 94 (2005)125510.

[13] A. Niikura, A.P. Tsai, A. Inoue, T. Masumoto, Chemical structural relaxation-induced embrittlement in amorphous MgeCueY alloys, J. Non Cryst. Solids159 (1993) 229e234.

[14] A. Castellero, B. Moser, D.I. Uhlenhaut, F.H. Dalla Torre, J.F. L€offler, Room-temperature creep and structural relaxation of MgeCueY metallic glasses,Acta Mater. 56 (2008) 3777e3785.

[15] A. Inoue, T. Nakamura, N. Nishiyama, T. Masumoto, MgeCueY bulk amor-phous alloys with high tensile strength produced by a high-pressure diecasting method, Mater. Trans. JIM 33 (1992) 937e945.

[16] H. Ma, J. Xu, E. Ma, Mg-based bulk metallic glass composites with plasticityand high strength, Appl. Phys. Lett. 83 (2003) 2793e2795.

[17] O. Biletska, K.J. Laws, M.A. Gibson, M. Ferry, Fabrication of an in-situ bulkmetallic glass composite with high magnesium content, Metall. Mater. Trans.A 45 (2014) 2352e2356.

[18] A. Inoue, N. Nishiyama, S.G. Kim, T. Masumoto, Fabrication and mechanicalproperties of MgeZneLa amorphous alloys containing nanoscale hcp-Mgparticles, Mater. Trans. JIM 33 (1992) 360e365.

[19] X. Zhou, K.D. Ralston, K.J. Laws, J.D. Cao, R.K. Gupta, M. Ferry, N. Birbilis, Effectof the degree of crystallinity on the electrochemical behaviour of Mg65Cu25Y10and Mg70Zn25Ca5 bulk metallic glasses, Corrosion 69 (2013) 781e792.

[20] M. Ferry, K.J. Laws, C. White, D. Miskovic, K.F. Shamlaye, W. Xu, O. Biletska,Recent developments in ductile bulk metallic glass composites, MRS Comm. 3(2013) 1e12.

[21] F. Guo, S.J. Poon, X. Gu, G.J. Shiflet, Low-density Mg-rich metallic glasses withbending ductility, Scr. Mater. 56 (2007) 689e692.

[22] H.J. Yu, J.Q. Wang, X.T. Shi, D.V. Louzguine-Luzgin, H.K. Wu, J.H. Perepezko,Ductile biodegradable Mg-based metallic glasses with excellent biocompati-bility, Adv. Funct. Mater. 23 (2013) 4793e4800.

[23] J. Yin, G.Y Yuan, Z.H. Chu, J. Zhang, W.J. Ding, MgeNie(Gd, Nd) bulk metallicglasses with improved glass-forming ability and mechanical properties,J. Mater. Res. 24 (2009) 2130e2140.

[24] K.J. Laws, J.D. Cao, C. Reddy, K.F. Shamlaye, B. Gun, M. Ferry, Ultra magnesium-rich, low-density MgeNieCa bulk metallic glasses, Scr. Mater. 88 (2014)37e40.

K.J. Laws et al. / Acta Materialia 103 (2016) 735e745 745

[25] D.I. Uhlenhaut, F.H. Dalla Torre, A. Castellero, C.A.P. Gomez, N. Djourelov,G. Krauss, B. Schmitt, B. Patterson, J.F. L€offler, Structural analysis of rapidlysolidified MgeCueY glasses during room-temperature embrittlement, Philos.Mag. 89 (2009) 233e248.

[26] A. Castellero, D.I. Uhlenhaut, B. Moser, J.F. L€offler, Critical Poisson ratio forroom temperature embrittlement of amorphous Mg85Cu5Y10, Philos. Mag.Lett. 87 (2007) 383e392.

[27] S.G. Kim, A. Inoue, T. Masumoto, High mechanical strengths of MgeNieY andMgeCueY amorphous alloys with significant supercooled liquid region,Mater. Trans. JIM 31 (1990) 929e934.

[28] C.A. Schuh, T.C. Hufnagel, U. Ramamurty, Mechanical behavior of amorphousalloys, Acta Mater. 55 (2007) 4067e4109.

[29] A. van den Beukel, J. Sietsma, The glass transition as a free volume relatedkinetic phenomenon, Acta Metall. Mater. 38 (1990) 383e389.

[30] P. Tuinstra, P.A. Duine, J. Sietsma, A. van den Beukel, The calorimetric glasstransition of amorphous Pd40Ni40P20, Acta Metall. Mater. 43 (1995)2815e2823.

[31] A. Slipenyuk, J. Eckert, Correlation between enthalpy change and free volumereduction during structural relaxation of Zr55Cu30Al10Ni5 metallic glass, Scr.Mater. 50 (2004) 39e44.

[32] Y. Zhang, H. Hann, Study of the kinetics of free volume in Zr45.0Cu39.3Al7.0Ag8.7bulk metallic glasses during isothermal relaxation by enthalpy relaxationexperiments, J. Non Cryst. Solids 355 (2009) 2616e2621.

[33] O. Haruyama, Y. Nakatama, R. Wada, H. Tokunaga, J. Okada, T. Ishikawa,Y. Yokoyama, Volume and enthalpy relaxation in Zr55Cu30Ni5Al10 bulkmetallic glass, Acta Mater. 58 (2010) 1829e1836.

[34] G.J. Fan, J.F. L€offler, R.K. Wunderlich, H.J. Fecht, Thermodynamics, enthalpyrelaxation and fragility of the bulk metallic glass-forming liquid Pd43Ni10-Cu27P20, Acta Mater. 52 (2004) 667e674.

[35] V. Wessels, G. Le Men�e, S.F. Fischerauer, T. Kraus, A.M. Weinberg,P.J. Uggowitzer, J.F. L€offler, In-vivo performance and structural relaxation ofbiodegradable bone implants made from MgeZneCa bulk metallic glasses,Adv. Eng. Mater. 14 (2012) B357eB364.

[36] A. Calka, The room temperature stability of amorphous MgeZn alloys, J. Phys.F Metall. Phys. 16 (1986) 1577e1582.

[37] B. Wielage, T. B€ottger, G. Fritsche, S. Mücklich, Crystallization behaviour ofamorphous MgeZneAl-alloys, Adv. Eng. Mater. 10 (2008) 439e443.

[38] R. Maaß, J.F. L€offler, Shear-band dynamics in metallic glasses, Adv. Funct.Mater. 25 (2015) 2353e2368.

[39] Y. Zhang, A.L. Greer, Thickness of shear bands in metallic glasses, Appl. Phys.Lett. 89 (2006) 071907.

[40] G.N. Greaves, A.L. Greer, R.S. Lakes, T. Rouxel, Poisson's ratio and modernmaterials, Nat. Mater. 10 (2011) 823e837.

[41] J.J. Lewandowski, W.H. Wang, A.L. Greer, Intrinsic plasticity or brittleness ofmetallic glasses, Philos. Mag. Lett. 85 (2005) 77e87.

[42] X.J. Gu, A.G. McDermott, S.J. Poon, G.J. Shiflet, Critical Poisson's ratio forplasticity in FeeMoeCeBeLn bulk amorphous steel, Appl. Phys. Lett. 88(2006) 211905.

[43] G. Ravichandran, A. Molinari, Analysis of shear banding in metallic glassesunder bending, Acta Mater. 53 (2005) 4087e4095.

[44] R. Mahjoub, K.J. Laws, J.P. Scicluna, J.E. Daniels, M. Ferry, A first principlesmolecular dynamics study of the relationship between atomic structure andelastic properties of MgeZneCa amorphous alloys, Comput. Mater. Sci. 96(2015) 246e255.

[45] M. Widom, B. Sauerwine, A.M. Cheung, S.J. Poon, P. Tong, D. Louca, G.J. Shiflet,Elastic properties of Ca-based metallic glasses predicted by first-principlessimulations, Phys. Rev. B 84 (2011) 054206.

[46] X.J. Gu, S.J. Poon, G.J. Shiflet, M. Widom, Ductility improvement of amorphoussteels: roles of shear modulus and electronic structure, Acta Mater. 56 (2008)88e94.

[47] S.F. Guo, N. Li, C. Zhang, L. Liu, Enhancement of plasticity of Fe-based bulkmetallic glass by Ni substitution for Fe, J. Alloy Compd. 504 (2010) S78eS81.

[48] Y.Q. Cheng, E. Ma, Atomic-level structure and structureeproperty relationshipin metallic glasses, Prog. Mater. Sci. 56 (2011) 379e473.

[49] H. Horikiri, A. Kato, A. Inoue, T. Masumoto, New Mg-based amorphous alloys

in MgeYemisch metal systems, Mater. Sci. Eng. A 179/181 (1994) 702e706.[50] A. Takeuchi, A. Inoue, Classification of bulk metallic glasses by atomic size

difference, heat of mixing and period of constituent elements and its appli-cation to characterization of the main alloying element, Mater. Trans. 46(2005) 2817e2829.

[51] A.S. Argon, H.Y. Kuo, Plastic flow in a disordered bubble raft (an analog of ametallic glass), Mater. Sci. Eng. 39 (1979) 101e109.

[52] F. Spaepen, A microscopic mechanism for steady state inhomogeneous flow inmetallic glasses, Acta Metall. 25 (1977) 407e415.

[53] Y. Shi, M.L. Falk, Does metallic glass have a backbone? the role of percolatingshort range order in strength and failure, Scr. Mater. 54 (2006) 381e386.

[54] D. Srolovitz, V. Vitek, T. Egami, An atomistic study of deformation of amor-phous metals, Acta Metall. Mater. 31 (1983) 335e352.

[55] L.Q. Xing, Y. Li, K.T. Ramesh, J. Li, T.C. Hufnagel, Enhanced plastic strain in Zr-based bulk amorphous alloys, Phys. Rev. B 64 (2001) 180201.

[56] D.H. Bae, S.W. Lee, J.W. Kwon, S. Yi, J.S. Park, Deformation behavior ofZreAleCueNieSn metallic glasses, J. Mater. Res. 21 (2006) 1305e1311.

[57] P. Thurnheer, R. Maaß, K.J. Laws, S. Pogatscher, J.F. L€offler, Dynamic propertiesof major shear bands in ZreCueAl bulk metallic glasses, Acta Mater. 96 (2015)428e436.

[58] M.D. Demetriou, M. Floys, C. Crewdson, J.P. Schramm, G. Garrett, W.L. Johnson,Liquid-like platinum-rich glasses, Scr. Mater. 65 (2011) 799e802.

[59] D.B. Miracle, E.A. Lord, S. Ranganathan, Candidate atomic cluster configura-tions in metallic glass structures, Mater. Trans. JIM 47 (2006) 1737e1742.

[60] D. Granata, E. Fischer, V. Wessels, J.F. L€offler, Fluxing of PdeSieCu bulkmetallic glasses and the role of cooling rate and purification, Acta Mater. 71(2014) 145e152.

[61] Y. Yokoyama, H. Tokunaga, A.R. Yavari, T. Kawamata, T. Yamasaki, K. Fujita,K. Sugiyama, P.K. Liaw, A. Inoue, Tough hypoeutectic Zr-based bulk metallicglasses, Metall. Mater. Trans. A 42 (2011) 1468e1475.

[62] O.N. Senkov, D.B. Miracle, Effect of the atomic size distribution on glassforming ability of amorphous metallic alloys, Mater. Res. Bull. 36 (2001)2183e2198.

[63] H.S. Chen, The influence of structural relaxation on the density and Young'smodulus of metallic glasses, J. Appl. Phys. 49 (1978) 3289e3291.

[64] T. Egami, Structural relaxation in amorphous alloys-compositional short rangeordering, Mater. Res. Bull. 13 (1978) 557e562.

[65] N. Nagendra, U. Ramamurty, T.T. Goh, Y. Li, Effect of crystallinity on the impacttoughness of a La-based bulk metallic glass, Acta Mater. 48 (2000)2603e2615.

[66] P. Murali, U. Ramamurty, Embrittlement of a bulk metallic glass due to sub-Tgannealing, Acta Mater. 53 (2005) 1467e1478.

[67] M. Lind, G. Duan, W.L. Johnson, Isoconfigurational elastic constants and liquidfragility of a bulk metallic glass forming alloy, Phys. Rev. Lett. 97 (2006)015501.

[68] J.F. Li, D.Q. Zhao, M.L. Zhang, W.H. Wang, CaLi-based bulk metallic glasses withmultiple superior properties, Appl. Phys. Lett. 93 (2008) 171907.

[69] M.S. Ong, W.M. Rainforth, Y. Li, H. Jones, Effect of long-term room-tempera-ture storage on the structure and properties of glassy melt-spun MgeAleCaalloys, Metall. Mater. Trans. A 31 (2000) 2155e2162.

[70] I.M. Hodge, Effects of annealing and prior history on enthalpy relaxation inglassy polymers. 4. Comparison of five polymers, Macromolecules 16 (1983)898e902.

[71] J. Ding, S. Patineta, M.L. Falka, Y. Cheng, E. Ma, Soft spots and their structuralsignature in a metallic glass, Proc. Natl. Acad. Sci. U. S. A. 111 (2014)14052e14056.

[72] Z.Q. Liu, Z.F. Zhang, Strengthening and toughening metallic glasses: the elasticperspectives and opportunities, J. Appl. Phys. 115 (2014) 163505.

[73] W.L. Johnson, K. Samwer, A universal criterion for plastic yielding of metallicglasses with a (T/Tg)2/3 temperature dependence, Phys. Rev. Lett. 95 (2005)195501.

[74] D. Klaumünzer, R. Maaß, F.H. Dalla Torre, J.F. L€offler, Temperature-dependentshear band dynamics in a Zr-based bulk metallic glass, Appl. Phys. Lett. 96(2010) 061901.