The kinetic concept of glass formation was developed by Gustav Tammann1 85 years ago. According to his assumption, a glass can be formed by cooling if the curves of the temperature dependency of nucleation and growth are significantly separated. The tendency of a metallic alloy to form a glass is usually characterized by the glass-forming ability (GFA), which corresponds to a critical cooling rate, βc,cr, at which no crystallization occurs during cooling from the melt2,3,4,5,6. After cooling at βc,cr to temperatures below the glass transition region the material is expected to form a monolithic glass, i.e., a glass whose atomic structure is completely amorphous. The structure of the glass can be changed by annealing it in the glassy state7,8,9,10,11,12 or by varying the cooling rate during glass formation13,14,15,16.
Another kinetic phenomenon is the critical heating rate, βh,cr, at which a glass does not crystallize upon heating14,15,17. It may be expected that βh,cr ≫ - βc,cr, because nucleation is much more dominant at deep undercooling, i.e., the maximum of the nucleation rate is at lower temperatures than the maximum of the growth rate14,15, and the critical size of nuclei may decrease at lower temperatures15.
The competition between glass formation and nucleation is a general phenomenon and occurs not only in rapidly quenched metallic alloys, but in various classes of metastable materials with different types of bonding. While glass formation is relatively rare in metallic alloys, polymers usually form a semicrystalline structure where the macromolecule is part of the crystalline and amorphous phase. Complete crystallization from the melt is thus basically impossible in such systems, so that glass formation in polymers can be considered a universal phenomenon. In polymeric materials the terms amorphous glass and semicrystalline glass are commonly used and distinguished from one another. For metals, a somewhat different terminology is used and the equivalent terms are monolithic glass and metallic glass composite, respectively.
Since the introduction of non-adiabatic chip calorimetry18,19 and its commercialization20,21, calorimetric measurements at defined cooling and heating conditions have been possible using fast differential scanning calorimetry (FDSC). This technique is frequently used to study glass transition phenomena and nanostructure formation in polymers22,23,24, molecular glass formers25, and chalcogenides26,27,28,29. However, recently FDSC has also been identified by the metals community as a suitable method for studying glass formation, nucleation, and phase transitions in metastable metallic alloys14,30,31,32,33,34,35,36. In14, one of the authors of this paper reported on the crystallization kinetics of an Au-based bulk metallic glass (Au49Ag5.5Pd2.3Cu26.9Si16.337), and constructed, via isothermal measurements in the millisecond range, complete time–temperature–transformation (TTT) diagrams of crystallization in the undercooled/supercooled liquid range upon cooling and heating.
In this study we use a newly developed Flash DSC2+ instrument (see Methods) which allows us to perform calorimetric experiments at ultrafast cooling rates of 40,000 Ks−1 and heating rates greater than 60,000 Ks−1. The latter also allows us to up-quench a certain phase, where up-quenching denotes a heating process that is so rapid that no structural changes occur before melting of the previously frozen phase36.
In this work we performed FDSC investigations on the bulk metallic glass (BMG) Au49Ag5.5Pd2.3Cu26.9Si16.3 at ultrafast rates with linear heating and cooling. We illustrate here that in the case of metals, amorphous (monolithic) glasses need to be classified into two categories. We term these self-doped glass (SDG) and chemically homogeneous glass (CHG). We show that quenched-in nuclei or nucleation precursors form in the SDG upon cooling at medium rates, which generates significant differences in the critical cooling and heating rates (βc,SDG ≈ 500 K s−1 ≪ βh,SDG ≈ 20,000 K s−1). In contrast, such nuclei no longer form upon rapid cooling, which leads to the fact that the rates of critical cooling (βc,CHG ≈ 4000 K s−1) and critical heating (βh,CHG = 6000 K s−1) are very similar for the CHG. This similarity has not been reported before because the critical cooling rate measured has always been that of an SDG38,39. The ramifications for the understanding of metallic glass processing are, however, significant because the conjectured and/or measured great differences between critical cooling and heating rates have often been explained by a pronounced asymmetry in crystallization behavior14,15,17. By analyzing the FDSC results and discussing the commonalities and differences involved in nucleation and glass formation for metallic and polymeric glass formers, we also intend to contribute to the development of a more generalized glass theory.