Thermochimica Acta 590 (2014) 84–90

Characterization of bulk metallic glasses via fast differential scanningcalorimetry

S. Pogatscher *, D. Leutenegger, A. Hagmann, P.J. Uggowitzer, J.F. LöfflerLaboratory of Metal Physics and Technology, Department of Materials, ETH Zurich, 8093 Zurich, Switzerland

A R T I C L E I N F O

Article history:Received 29 April 2014Received in revised form 10 June 2014Accepted 11 June 2014Available online 12 June 2014

Keywords:Bulk metallic glassesCrystallizationNucleationKineticsFast differential scanning calorimetry

A B S T R A C T

This study explores the thermophysical properties of Au-based bulk metallic glasses (BMGs) via fastdifferential scanning calorimetry (FDSC). Using this technique, the glass formation of the alloys Au60+xCu15.5�xAg7.5Si17 (x = 0, 5 and 10) was investigated in situ. The critical cooling rate (Fc) and heating rate(Fh) required to avoid crystallization were analyzed for various sample masses and chip sensor surfacematerials. The results show that the alloy with the highest Au-content exhibits the lowest resistanceagainst crystallization. Silicon nitride, silicon oxide and graphite used as chip sensor surface material wereproven not to influence the measurements. In general, a dependence of crystallization on sample mass wasobserved for all compositions. Both the critical cooling and critical heating rates increase until a certainmass is reached. This phenomenon is explained via a size-dependent nucleation effect.

ã 2014 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Thermochimica Acta

journal homepa ge: www.elsev ier .com/locate / tca

1. Introduction

Bulk metallic glasses are non-crystalline metallic solids whichcan be produced by rapid cooling of metallic melts to temperaturesbelow their glass transition [1]. Compared to all other classes ofmaterials BMGs possess unique properties such as high strengthand elastic strain limit, good soft-magnetic properties, excellentcorrosion resistance and high hardness [2–5]. Their good viscousflow workability in the supercooled liquid and homogeneity andisotropy on a small scale are great advantages, especially in theproduction of small-scale devices (e.g. micro-electro-mechanicalsystems, micro-robotics and micro-manipulators) via imprinting,embossing, micro-replication or micro-molding [6–8]. Au-basedBMGs [9–12] in particular have been shown to be suitablematerials for this emerging field [8].

For metallic systems, BMGs demonstrate extraordinary stabilityagainst crystallization, i.e. they exhibit a low critical cooling rate forreaching the glass transition without crystallization during coolingfrom the equilibrium liquid. Nevertheless, crystallization still occursrapidly and thus limits many experimental studies in the super-cooled liquid region [3–5]. Using conventional thermo-analytical

* Corresponding author. Present address: Laboratory of Metal Physics andTechnology, Department of Materials, ETH Zurich, Vladimir-Prelog-Weg 4, 8093Zurich, Switzerland. Tel.: +41 44 633 64 65; fax: +41 44 633 14 21.

E-mail addresses: stefan.pogatscher@mat.ethz.ch (S. Pogatscher),danieleu@student.ethz.ch (D. Leutenegger), hagmanna@student.ethz.ch(A. Hagmann), peter.uggowitzer@mat.ethz.ch (P.J. Uggowitzer),joerg.loeffler@mat.ethz.ch (J.F. Löffler).

http://dx.doi.org/10.1016/j.tca.2014.06.0070040-6031/ã 2014 Elsevier B.V. All rights reserved.

methods (e.g. differential scanning calorimetry, DSC) it is notpossible to reach constant cooling rates higher than a few K s�1, andin situ probing of the glass formation from an equilibrium metallicmelt is not feasible. Recent chip-based fast differential scanningcalorimeters [13,14] enable thermo-analytical measurements atorders of magnitude higher rates. Heating and cooling with several104K s�1 and 103K s�1, respectively, can be realized with a recentlyavailable commercial instrument (Mettler Toledo Flash DSC 1 [15]).This instrument has generally been used to study polymers [15] andphase-change materials [16,17], but in recent studies it has also beensuccessfully applied to a Au49Ag5.5Pd2.3Cu26.9Si16.3 BMG [18]. Au-based BMGs are ideal candidates for investigation via FDSC becausehere in situ exploration of the glass formation and crystallizationbehavior in the whole supercooled liquid region is possible [18].Compared to most other known BMGs, Au-based BMGs have a lowliquidus temperature, which is accessible by FDSC, and are notsensitive to oxidation. However the characterization of BMGs viaFDSC is still a new procedure and no work on the measurementconditions and the influence of measurement parameters has so farbeen published. In this study we explore the crystallization and glassformation of Au60+xCu15.5�xAg7.5Si17 (x = 0, 5 and 10) in situ and focusalso on the effects of sensor surface material and sample mass.

2. Material and methods

2.1. Alloy production

To obtain thin and chemically homogenous samples Au-basedglassy ribbons were produced by melt spinning. The elements Au

Fig. 1. DSC traces of Au60+xCu15.5�xAg7.5Si17 (x = 0, 5 and 10) metallic glassesmeasured with a heating rate of 0.33 K s�1 and corresponding DHm values.

S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90 85

(purity 99.99 wt.%), Ag (99.99 wt.%), Si (99.999 wt.%) and Cu(99.995 wt.%) were weighed according to the atomic composi-tions Au60Cu15.5Ag7.5Si17, Au65Cu10.5Ag7.5Si17 andAu70Cu5.5Ag7.5Si17 and inserted into quartz glass tubes with a

Fig. 2. Micrographs of pre-molten samples with various masses on the sample platformsquare of the chip sensor (a, b), technical artifacts may occur and can reduce the maxi

diameter of 5 mm. The tubes were purged several times with Ar(5 N purity) and closed under 200 mbar Ar pressure by meltingthe tube ends. To produce homogenous pre-alloys the elementswere mixed well in the tube, subjected to induction melting at1273 K [19], and finally quenched in water. The pre-alloys werepolished and broken up into small, manageable parts for ribbonproduction via melt spinning under a 500 mbar He atmosphere(5 N purity). The rotating frequency of the copper wheel used formelt spinning was 25 Hz and its distance from the hole of thegraphite crucible containing the melt was 0.2 mm. About 1 g ofthe pre-alloy was heated to 923 K within 7 min and held for 2 minat this temperature before the ribbons were produced. The over-pressure applied to push the melt out of the crucible onto therotating copper wheel was 150 mbar. The thickness of the ribbonsproduced ranged from 20 to 30 mm; 20 mm thick ribbons weredeployed for the FDSC investigations.

2.2. Thermo-analytical measurements

Conventional thermal analysis was performed in a differentialscanning calorimeter (Mettler-Toledo DSC1) to determine the massof the small-scale FDSC samples. The DSC measurements wereconducted at a heating rate of 0.33 K s�1 under Ar atmosphere (5 Npurity) at a flow rate of 30 ml min�1 and using aluminum pans onthe sample and reference platforms. The enthalpy of fusion (DHm)measured by conventional DSC was used as a reference valueaccording to Eq. (1) [20]:

of the sensor. For large sample masses (>10 mg) and samples larger than the smallmum rates accessible.

Fig. 3. FDSC scans when heating (a) Au60Cu15.5Ag7.5Si17, (b) Au65Cu10.5Ag7.5Si17, and(c) Au70Cu5.5Ag7.5Si17 metallic glasses on a standard silicon nitride, silicon oxide, orgraphite sensor membrane surface. The inserts illustrate the applied temperature–time programs and Tg, Tx and Tm represent the onsets of glass transition,crystallization and melting.

86 S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90

mFDSC ¼ DHm;FDSC

DHm;DSC� mDSC (1)

Fig. 1 shows DSC traces of the alloys investigated and thecorresponding DHm values. The glass transition (Tg), onset ofcrystallization (Tx) and onset of melting (Tm) are indicated asexamples for Au60Cu15.5Ag7.5Si17. Note that more than onecrystallization peak is visible for all BMGs investigated in Fig. 1.

FDSC was performed in power compensation mode using theMettler-Toledo Flash-DSC 1. The sample support temperature ofthe FDSC was set at 183 K using a Huber intracooler TC90. Thefurnace was purged with Ar of 5 N purity at a flow rate of10 ml min�1. FDSC samples were prepared by cutting the melt-spun ribbons under a stereomicroscope to small pieces withweights of 30 ng to 20 mg and then transferred by an electrostaticmanipulator hair onto a conditioned and temperature-correctedMultiSTAR UFS1 sensor (according to the instrument provider’sspecification). Fig. 2 illustrates samples of various masses on thesensor.

To protect the samples from bouncing due to strains in thematerial they were pre-melted with a heating rate of 1 K s�1 fromroom temperature to 748 K, which is a temperature accessible formost sensors. For all experiments the samples were heated orcooled between 298 K and 748 K. The exact time–temperatureregimes used are displayed in the corresponding heat-flow figures.Reproducibility was always found to be very high, as was judgedfrom comparing the same thermal cycles at the start and end ofeach measurement series.

To explore sensor materials other than the silicon nitridesurface provided, the reverse side of the sensors made of siliconoxide was used, and the silicon nitride surface was also coated witha graphite layer of approximately 10 nm thickness.

3. Results

3.1. Sensor surface material

Fig. 3 shows FDSC scans at a heating rate of 100 K s�1 forAu60+xCu15.5�xAg7.5Si17 (x = 0, 5 and 10) on a standard siliconnitride sensor surface, on the reverse silicon oxide side of thesensor, and on a graphite-coated sensor membrane. The sampleswere amorphized in situ by quenching from 748 K to RT with acooling rate of 5000 K s�1 prior to the measurements. The inserts toFig. 3 illustrate the time–temperature regime. Clear glasstransitions followed by exothermic crystallization peaks andmelting of the samples can be observed for all Au-based glasses,and the sensor surface material does not influence the measure-ments. The curves are not normalized to the mass, whichintroduces some differences in the size of the peaks only.

3.2. Critical cooling rate

Fig. 4 shows typical FDSC scans of Au60+xCu15.5�xAg7.5Si17 BMGswith x = 0, 5 and 10 when cooling the equilibrium liquid at differentrates. The inserts illustrate the applied temperature–time pro-grams. The exothermic crystallization peaks (400–500 K) shift tolower temperatures and their enthalpy of crystallization decreaseswith increasing cooling rate until the crystallization peak vanishes.This means that for this and higher cooling rates no crystallizationoccurs and the critical cooling rate Fc is reached. All alloys alsodemonstrate a clear glass transition which depends on the coolingrate. Fig. 4(a) shows the curves of the heat flow measured duringcooling of a Au60Cu15.5Ag7.5Si17 melt at rates of 425 K s�1 up to675 K s�1 and a sample mass of 1.5 mg on the standard siliconnitride sensor membrane. The transition from crystallization in thesupercooled liquid region to in situ amorphization is observed at

Fig. 4. Typical FDSC scans when cooling the melt in the region of Fc for (a)Au60Cu15.5Ag7.5Si17, (b) Au65Cu10.5Ag7.5Si17, and (c) Au70Cu5.5Ag7.5Si17. The samplemasses are 1.5, 3.3, and 1.3 mg, respectively. In all cases a standard silicon nitridesensor membrane was used. The inserts illustrate the applied temperature–timeprograms, and Tg and Tx represent the onsets of glass transition and crystallization.

Fig. 5. Typical FDSC scans when heating the glass in the region of Fh for (a)Au60Cu15.5Ag7.5Si17, (b) Au65Cu10.5Ag7.5Si17, and (c) Au70Cu5.5Ag7.5Si17. The samplemasses are 1.5, 3.3, and 1.3 mg, respectively. In all cases a standard silicon nitridesensor membrane was used. The inserts illustrate the applied temperature–timeprograms, and Tg and Tx represent the onsets of glass transition and crystallization.

S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90 87

Fig. 6. Mass dependence of the critical cooling rates for (a) Au60Cu15.5Ag7.5Si17, (b)Au65Cu10.5Ag7.5Si17, and (c) Au70Cu5.5Ag7.5Si17measured on a standard silicon nitridesensor membrane. Results for samples on silicon oxide and graphite are also shown(marked by arrows). The dashed lines are guides for the eye.

88 S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90

Fc� 575 K s�1. Fig. 4(b) shows the cooling curves of aAu65Cu10.5Ag7.5Si17 melt (sample mass 3.3 mg) at rates of200–375 K s�1. The critical cooling rate is observed at around350 K s�1. Au70Cu5.5Ag7.5Si17 (sample mass 1.3 mg) was investigatedbetween 500 K s�1 and 1200 K s�1 and exhibits a higher Fc valueof around 1200 K s�1 (Fig. 4(c)). Note that Fig. 4 is also a goodexample of the instrument’s performance at different coolingrates. The signal-to-noise ratio is higher in Fig. 4(c) than in Fig. 4(a)and (b).

3.3. Critical heating rate

The samples used to determine Fc (Fig. 4) were alsoinvestigated for their critical heating rate Fh, at which the glass(obtained by in situ quenching from 748 K at a rate of 5000 K s�1)transforms on heating into the supercooled and finally equilibriumliquid without detectable crystallization (Fig. 5). Below Fh anexothermic crystallization peak and a melting peak can be seen.Fig. 5(a) shows curves for the heat flow measured during heating ofa glassy Au60Cu15.5Ag7.5Si17 sample at rates of 750–2500 K s�1 onthe standard silicon nitride sensor membrane. At rates higher thanaround 2300 K s�1 neither crystallization nor melting are visibleand a direct transition from the supercooled to the equilibriumliquid occurs with no detectable heat flow. Fig. 5(b) shows curvesfor glassy Au65Cu10.5Ag7.5Si17 at heating rates between 800 K s�1

and 2500 K s�1. The critical heating rate Fh is again observed ataround 2300 K s�1. Fig. 5(c) displays FDSC scans ofAu70Cu5.5Ag7.5Si17 at heating rates of 7000–12,000 K s�1. This alloyexhibits with a Fh of around 11,000 K s�1 the highest criticalheating rate of all samples investigated.

3.4. Mass dependency of Fc and Fh

To characterize the influence of different sample masses, themass dependence of the critical cooling (Fc) and heating (Fh)rates was investigated for Au60+xCu15.5�xAg7.5Si17 (x = 0, 5 and 10).Fc and Fh are defined as the rates at which no crystallizationpeak is detectable in the heat flow curves. Note that the errorin the corresponding Figs. 6 and 7 is generated by the step-sizein F.

A dependence of crystallization on sample mass was observedfor all compositions investigated. The critical cooling rate increasesuntil a certain mass is reached. This saturation starts at roughlyabove 1 mg. The critical cooling rates within this constant regimeare at around 600 K s�1 for Au60Cu15.5Ag7.5Si17 (Fig. 6(a)), 400 K s�1

for Au65Cu10.5Ag7.5Si17 (Fig. 6(b)) and 1700 K s�1 forAu70Cu5.5Ag7.5Si17 (Fig. 6(c)). No technical artifacts due to largesample sizes (see Fig. 2(a) and (b)) were observed within the rangeof cooling rates investigated. Results for samples on silicon oxideand graphite as sensor surface material are also shown (marked byarrows). No influence of the sensor material on Fc can be deducedfrom Fig. 6. However, although scattered, masses �1 mg generatedsignificantly lower Fc values.

The same trend was found for the critical heating rate to avoidcrystallization for all compositions investigated (Fig. 7), althoughthe values of Fh are much higher than those of Fc. The criticalheating rate Fh is also more difficult to deduce from FDSC curvesthan Fc, because the transition (crystallization of the supercooledliquid vs. no crystallization) is more blurred. Above 1 mg of samplemass, Fh was found to be around 2300 K s�1 for Au60Cu15.5Ag7.5Si17(Fig. 7(a)), 2400 K s�1 for Au65Cu10.5Ag7.5Si17 (Fig. 7(b)) and13,000 K s�1 for Au70Cu5.5Ag7.5Si17 (Fig. 7(c)). An investigation ofFh for large heating rates is restricted to masses smaller than 10 mg(see Fig. 7(c)) due to limitations of the setup. Again no effect of thesensor surface material was observed.

Fig. 7. Mass dependence of the critical heating rate for (a) Au60Cu15.5Ag7.5Si17, (b)Au65Cu10.5Ag7.5Si17, and (c) Au70Cu5.5Ag7.5Si17measured on a standard silicon nitridesensor membrane. Results for samples on silicon oxide and graphite are also shown(marked by arrows). The dashed lines are guides for the eye.

S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90 89

4. Discussion

The crystallization and vitrification behavior of three differentAu-based BMGs was invested in situ by FDSC with a focus on theeffects of sensor surface material and sample mass.

Using different surface materials for the chip sensor membranedoes not influence the FDSC results. For silicon nitride, silicon oxideor graphite as surface material FDSC traces appear quite similar(Fig. 3) and are also comparable to conventional DSC traces of thealloys analyzed (Fig. 1). Note that the exothermic crystallizationpeaks in Fig. 3 are at higher temperatures than in Fig. 1 because ofthe higher heating rates used in the FDSC measurements.

The used sensor surface material also does not influence thecritical cooling and heating rates measured (see Figs. 6 and 7). Inprinciple these results confirm that the standard sensor surface issuitable for an investigation of Au-based BMGs, but we found thatseveral practical issues arise. Using the silicon oxide surface on thereverse side of the sensor allows rotation of the sensor duringsample transfer, which was found to be quite useful. However, wepropose a graphite-coated chip sensor as the best possibility.Samples are much easier to position on this sensor surface materialbecause they do not tend to jerk (as they do on silicon nitride oroxide) when pushed by the manipulator. In addition, it is mucheasier to remove a sample from a graphite-coated chip sensor. Thismeans that the sensors can be re-used for multiple samples.

A dependency of crystallization kinetics on sample mass wasobserved for all compositions (Figs. 6 and 7). Nevertheless, it isquite convenient that there is a broad range of masses (1–20 mg;1–10 mg samples showed the best handling conditions), which canbe used expediently.

The possibility to determine Fc in situ by using FDSC is veryimportant for future BMG development, because Fc is related tothe size to which BMG products can be made of. The observationthat the alloy with the highest Au-content clearly exhibits thehighest Fc is predictable because this trend has been also reportedfor the critical casting diameter at which an amorphous sample canstill be obtained [10,11]. However, Au65Cu10.5Ag7.5Si17 showed aslightly lower Fc than Au60Cu5.5Ag7.5Si17 in FDSC, which contrastswith the trend found in [10,11] for the critical casting diameter. Themethodology for determining the critical casting thickness in[10,11] was rather rough and parameters other than the criticalcooling rate (e.g. thermal conductivity of the alloy, casting setupetc. [21]) also influence the critical casting thickness. Thus,determining Fc by FDSC is believed to be a more sensitive methodthan casting rods of certain diameters (usually in mm steps) andchecking whether the samples are amorphous via X-ray diffrac-tion. This underlines the importance of FDSC for future BMGdevelopment.

The observation that the critical cooling rate for reaching theglassy state upon cooling from the melt without crystallizationdecreases at small sample masses is very interesting. During thisstudy, however, it mainly helped to determine an optimal FDSCworking range for Au-based BMGs. Nevertheless, one possibleexplanation for the decrease in Fc with decreasing sample massmay be the FDSC setup itself: in FDSC only the sensor platformserves as a heat source, while the ambient Ar atmosphere stays at183 K. This may introduce a temperature gradient in the sample,which could enhance crystallization on the colder sample side.Very small samples will not retain the thickness of the BMGribbons (20 mm) due to cutting issues, i.e. they may be thinner andtherefore exhibit a reduced temperature gradient, which wouldresult in a decrease of Fc. Fig. 8 shows the temperature gradient ina large Au60Cu15.5Ag7.5Si17 sample of mass 8.5 mg measured on thestandard silicon nitride sensor membrane. Indium as a referencematerial is located on the reference platform (melting apparentlyoccurs exothermically) for the first heating run of the sensor. In a

Fig. 8. Investigation of a potential temperature gradient in an Au60Cu15.5Ag7.5Si17sample of mass 8.5 mg measured on a standard silicon nitride sensor membrane. Anindium sample is located on the reference platform (melting then occursexothermically) in the first run. In a second run it is also located on top of theAu60Cu15.5Ag7.5Si17 sample on the sample platform. In both cases melting occurs atthe same temperature.

90 S. Pogatscher et al. / Thermochimica Acta 590 (2014) 84–90

second run a larger amount of In is additionally located on top ofthe Au60Cu15.5Ag7.5Si17 glass on the sample platform. The onset ofboth melting peaks is similar (within 1 K) and temperaturegradients in 20 mm thick Au-based BMG samples are thereforenegligible. Thus temperature gradients cannot be made responsi-ble for the decrease of Fc at small sample masses.

A decreasing Fc with decreasing sample size (mostly indispersed systems) of BMGs [22], or similarly an increasingundercooling in pure metals with decreasing sample size [23,24],has also been observed previously and discussed in literature.Schroers et al. [22] reported that small particles of Pd-based BMGsexhibit a lower critical cooling rate and related this observation tothe lower probability of impurities acting as nucleation sites. Inprinciple this is also possible in our case. In addition to thisheterogeneous nucleation effect, Wilde et al. [25] reported anincreasing undercooling for a reduced sample size of pure Ni anddiscussed this effect by homogenous nucleation within theframework of classical nucleation theory. While it was not withinthe scope of this work to clarify the mechanisms behind thedecrease in Fc with decreasing sample size, it is likely that it can belinked to a decrease in the probability of nucleation. The generalobservation that Fc is always lower than Fh is expected becausefor metallic glasses the maximum in the nucleation rate alwaysappears at a lower temperature than the maximum in the growthrate (see, e.g., Ref. [18,22]).

5. Conclusions

In summary, we have shown that crystallization and vitrifica-tion of bulk metallic glasses can be characterized well by fastdifferential scanning calorimetry. The most important findingsare:

� The sensor surface material used (silicon nitride, silicon oxide orgraphite) does not influence the FDSC measurements of Au-based bulk metallic glasses.

� Critical cooling and heating rates to avoid crystallization dependon sample mass. Both values decrease below 1 mg, which may beexplained by size-dependent nucleation.

� Optimal practical working conditions for Au-based bulk metallicglasses are obtained with sample masses of 1–10 mg on sensorscoated with graphite.

Most importantly, the demonstrated capability of fast differen-tial scanning calorimetry to determine the critical cooling rate ofBMGs in situ is expected to be a great advantage in future BMGdevelopment.

Acknowledgments

The authors thank Fabio Krogh at LMPT and Jürgen Schawe atMettler Toledo AG for fruitful discussions. Support by the SwissNational Science Foundation (SNF Grant No. 200020-135100) isgratefully acknowledged.

References

[1] C. Suryanarayana, A. Inoue, Bulk Metallic Glasses, Taylor & Francis, CRC Press,Boca Raton, London, New York, 2010, p. 1.

[2] A.L. Greer, Metallic glasses, Science 267 (1995) 1947–1953.[3] W.L. Johnson, Bulk glass-forming metallic alloys: science and technology, MRS

Bull. 24 (1999) 42–56.[4] A. Inoue, Stabilization of metallic supercooled liquid and bulk amorphous

alloys, Acta Mater. 48 (2000) 279–306.[5] J.F. Löffler, Bulk metallic glasses, Intermetallics 11 (2003) 529–540.[6] G. Kumar, H.X. Tang, J. Schroers, Nanomoulding with amorphous metals,

Nature 457 (2009) 868–872.[7] G. Kumar, A. Desai, J. Schroers, Bulk metallic glass: the smaller the better, Adv.

Mater. 23 (2011) 461–476.[8] Y.C. Chen, J.P. Chu, J.S.C. Jang, C.W. Wu, Thermoplastic deformation and micro/

nano-replication of an Au-based bulk metallic glass in the supercooled liquidregion, Mater. Sci. Eng. A 556 (2012) 488–493.

[9] J. Schroers, B. Lohwongwatana, W.L. Johnson, A. Peker, Gold based bulkmetallic glass, Appl. Phys. Lett. 87 (2005).

[10] W. Zhang, H. Guo, M.W. Chen, Y. Saotome, C.L. Qin, A. Inoue, New Au-basedbulk glassy alloys with ultralow glass transition temperature, Scripta Mater. 61(2009) 744–747.

[11] H. Guo, W. Zhang, M.W. Chen, Y. Saotome, M. Fukuhara, A. Inoue, Effect of Aucontent on thermal stability and mechanical properties of Au-Cu-Ag-Si bulkmetallic glasses, Metall. Mater. Trans. A 42 (2011) 1486–1490.

[12] G. Fiore, P. Rizzi, L. Battezzati, Phase constitution and glass formation in an Au-based alloy, J. Alloys Compd. 509 (2011) 166–169.

[13] E. Zhuravlev, C. Schick, Fast scanning power compensated differential scanningnano-calorimeter: 1. The device, Thermochim. Acta 505 (2010) 1–13.

[14] E. Zhuravlev, C. Schick, Fast scanning power compensated differential scanningnano-calorimeter:2.Heat capacityanalysis,Thermochim. Acta 505 (2010)14–21.

[15] V. Mathot, M. Pyda, T. Pijpers, G. Vanden Poel, E. Van De Kerkhof, S. VanHerwaarden, F. Van Herwaarden, A. Leenaers, The Flash DSC 1, a powercompensation twin-type, chip-based fast scanning calorimeter (FSC): firstfindings on polymers, Thermochim. Acta 522 (2011) 36–45.

[16] J. Orava, A.L. Greer, B. Gholipour, D.W. Hewak, C.E. Smith, Characterization ofsupercooled liquid Ge2Sb2Te5 and its crystallization by ultrafast-heatingcalorimetry, Nature Mater. 11 (2012) 279–283.

[17] J. Orava, A.L. Greer, B. Gholipour, D.W. Hewak, C.E. Smith, Ultra-fast calorimetrystudy of Ge2Sb2Te5 crystallization between dielectric layers, Appl. Phys. Lett.101 (2012).

[18] S. Pogatscher, P.J. Uggowitzer, J.F. Löffler, In situ probing of metallic glassformation and crystallization upon heating and cooling via fast differentialscanning calorimetry, Appl. Phys. Lett. (2014), doi:http://dx.doi.org/10.1063/1.4884940.

[19] S. Mozgovoy, J. Heinrich, U.E. Klotz, R. Busch, Investigation of mechanical,corrosion and optical properties of an 18 carat Au-Cu-Si-Ag-Pd bulk metallicglass, Intermetallics 18 (2010) 2289–2291.

[20] G.V. Poel, D. Istrate, A. Magon, V. Mathot, Performance and calibration of theFlash DSC 1, a new, MEMS-based fast scanning calorimeter, J. Therm. Anal.Calorim. 110 (2012) 1533–1546.

[21] K.J. Laws, B. Gun, M. Ferry, Influence of casting parameters on the criticalcasting size of bulk metallic glass, Metall. Mater. Trans. A 40 (2009) 2377–2387.

[22] J. Schroers, W.L. Johnson, Extremely low critical cooling rate measured ondispersed Pd43Ni10Cu 27P20, Appl. Phys. Lett. 80 (2002) 2069–2071.

[23] B. Yang, A.S. Abyzov, E. Zhuravlev, Y. Gao, J.W.P. Schmelzer, C. Schick, Size andrate dependence of crystal nucleation in single tin drops by fast scanningcalorimetry, J. Chem. Phys. 138 (2013).

[24] J.H. Perepezko, J.L. Sebright, P.G. Höckel, G. Wilde, Undercooling andsolidification of atomized liquid droplets, Mater. Sci. Eng. A 326 (2002)144–153.

[25] J. Bokeloh, R.E. Rozas, J. Horbach, G. Wilde, Nucleation barriers for the liquid-to-crystal transition in Ni: experiment and simulation, Phys. Rev. Lett. 107(2011) 145701.