WorksPhysics and Astronomy
On the interplay of liquid-like and stress-driven dynamics in a metallic glass former observed by temperature scanning XPCS
Abstract
Modern detector technology and highly brilliant fourth-generation synchrotrons allow to improve the temporal resolution in time-resolved diffraction studies. Profiting from this, we applied temperature scanning X-ray photon correlation spectroscopy (XPCS) to probe the dynamics of a Pt-based metallic glass former in the glass, glass transition region, and supercooled liquid, covering up to six orders of magnitude in time scales. Our data demonstrates that the structural alpha-relaxation process is still observable in the glass, although it is partially masked by a faster source of decorrelation observed at atomic scale. We present an approach that interprets these findings as the superposition of heterogeneous liquid-like and stress-driven ballistic-like atomic motions. This work not only extends the dynamical range probed by standard isothermal XPCS, but also clarifies the fate of the alpha-relaxation across the glass transition and provides a new perception on the anomalous, compressed temporal decay of the density-density correlation functions observed in metallic glasses and many out-of-equilibrium soft materials.
Teaser
For the first time, we observe and model the complex relaxation behavior of an in-situ vitrifying metallic glass former via XPCS.
Introduction
When a liquid is undercooled, its structural dynamics undergo a drastic slowdown over several orders of magnitude. The temperature dependence of this slowdown can differ significantly among different systems, as described by Austen Angell’s famous fragility concept [1] that categorizes liquid dynamics by their deviation from an exponential, Arrhenius-like temperature dependence. The dynamic slowdown of undercooled liquids is the precondition that eventually results in vitrification, where the liquid ‘falls out of equilibrium’ [1] and forms a non-ergodic, non-equilibrium glass. In other words, liquid dynamics predetermine vitrification kinetics. Hence, understanding the nature of the dynamic slowdown and fragility is of fundamental relevance for glass science and industry-relevant aspects like e.g. the glass forming ability (GFA) of a given system.
A mighty tool to determine the dynamics of disordered materials is X-ray photon correlation spectroscopy (XPCS), a time-resolved experimental approach [2] that uses monochromatic and coherent synchrotron radiation. Coherent diffraction on amorphous matter creates a speckle pattern that reflects the configuration of the microscopic ensemble in the probed sample volume. Temporal changes in the microscopic configuration induce changes in the speckle pattern and accordingly, the correlation between two speckle patterns will decay with growing temporal distance. Analyzing the intensity autocorrelation function thus allows to determine microscopic relaxation times. XPCS studies that resolute the atomic-scale dynamics of metallic glass formers require highly brilliant radiation to compensate the weak scattering signal of these amorphous materials and are therefore only possible since about one decade [3]. While the signal-to-noise ratio and detector technology limited the exposure time to several seconds in earlier studies [3, 4, 5, 6, 7, 8, 9, 10], recent advances allowed exposure times in the sub-second regime [11, 12, 13, 14, 15].
In the present study, we use the high flux, high photon energy, and advanced detector technology available at the ID10 beamline of the fourth-generation synchrotron radiation facility ESRF [16] to perform temperature scanning XPCS. This non-isothermal approach can only be sparsely found in literature [15, 17, 18], and we apply it for the first time on a metallic glass former, namely the alloy [19]. Its excellent thermal stability against crystallization allowed our groups to investigate the supercooled liquid (SCL) dynamics in earlier isothermal XPCS experiments [13]. The present work extends these studies by slowly heating and cooling the sample through the glass, the glass transition, and the SCL state.
In the equilibrium SCL, the decay of the obtained intensity autocorrelation functions can be modelled accurately by a conventional Kohlrausch-Williams-Watts (KWW) equation with stretched shape, which reflects the heterogeneous, liquid-like atomic motions of the -relaxation process, as previously observed by isothermal XPCS [13]. Furthermore, the temperature dependence of the determined relaxation times is found to mimic the alloy’s fragility measured by macroscopic viscosity measurements [20]. A different picture arises in the non-equilibrium state, i.e., in the glass and the glass transition region. Here, the conventional KWW fit fails to describe the decay of the intensity autocorrelation functions. Instead, we successfully model the complexly shaped decorrelation through a multiplication of two KWW functions. One of these features a stretched shape (KWW shape exponent < 1), while the other one features a compressed shape (KWW shape exponent > 1).
Based on the long-standing concept of spatially heterogeneous dynamics, we offer a scenario that explains these findings through two different, but simultaneously occurring types of atomic motion. More precisely, we interpret the vitrification process as the interlocking of nm-scale regions with slower dynamics, so termed ‘rigid domains’. This jamming-like effect results in volumetric frustration and internal stress gradients, which induce ballistic-like ‘microscopic drift’ movements with a typical compressed decorrelation footprint. On the other hand, we attribute the stretched decorrelation component to atomic motions related to the liquid-like -relaxation, which remain partially active in the non-equilibrium. The superposition of these two types of atomic motion finally creates the complex decorrelation shape observed in the non-equilibrium states.
Results
The thermal program applied during the XPCS experiments consists of a heating scan and a subsequent cooling scan, both performed with a rate of 0.0167 K/s (1 K/min). It is retraced ex-situ by differential scanning calorimetry (DSC) and the obtained heat flow (HF) signal is shown in Figure 1A. The initial heating of the as-spun ribbon sample starts in the glassy state, passes the glass transition region between 500 K and 525 K, and enters the SCL state until the maximum temperature of 548 K is reached. In the cooling scan, the sample is cooled until vitrification occurs between 521 K and 490 K, finally reentering the glassy state. Accordingly, the thermal ranges of the glass, glass transition, and SCL regions are distinguished by gray, light gray, and white backgrounds to guide the eyes.

Figure 1. Applied thermal program, representative curves, and different fitting approaches. (A) Ex-situ DSC scans retrace the 0.0167 K/s heating-cooling protocol applied for the XPCS measurements. The three thermal regions, namely glass, glass transition, and SCL, are distinguished through shaded backgrounds. Furthermore, circles highlight the temperatures of six representative evaluation batches, numbered from B1 to B6. The intensity autocorrelation functions obtained from these batches are given in (B) and (C), together with the KWW fits (black lines) and fits (green lines). KWW and fits both provide satisfying and similar results in the SCL state (B3 and B4), but while the KWW approach fails to describe the shape of the decorrelation in the glass and especially in the glass transition region, provides meaningful results (B1, B2, B5, and B6).
Intensity autocorrelation functions, , are evaluated in subsequent batches of 240 s duration, as further explained in the Materials and Methods section. To provide an overview, we focus in the following on six of these evaluation batches, numbered from B1 to B6. These are located at representative temperatures in the glass (480 K), glass transition (510 K), and SCL (541 K) upon heating and cooling, as indicated in Figure 1A. The corresponding data is given in Figure 1B and Figure 1C. The curves decorrelate at shorter times with rising temperature, thereby reflecting the expectable acceleration of the dynamics with increasing thermal motion. In liquids, the temporal decay of the intermediate scattering function (ISF) and the related function is usually described by a KWW function [21]
and since and are connected through the Siegert relation [22, 23],
it follows that the intensity autocorrelation function can be modelled as
where is the non-ergodicity parameter, is the relaxation time, is the shape exponent, is the experimental contrast, and is the baseline value. The resulting KWW fits of the data are displayed by black solid curves in Figure 1B and Figure 1C. In the SCL state (B3 and B4), the KWW function models the stretched decay of quite satisfactorily, in agreement with previous findings [6, 12, 13, 24]. A different picture arises in the non-equilibrium, i.e. in the glass transition and glass region (B1, B2, B5, and B6), where the KWW fits clearly fail to describe the complex shape of . More precisely, the initial parts of the data, roughly the first three decades in timescale, from 0.01 to 10 s, behave more stretched than the overall KWW fit curve, while the later parts appear more compressed. Hence, we observe a kind of ‘cut-off’ behavior, where an initially stretched decay abruptly changes into a more compressed appearance.
Previous XPCS studies revealed that metallic glass formers show an anomalous crossover in their microscopic dynamics. While the metastable equilibrium SCL phase typically features stretched decorrelation with values below 1 [6, 12, 13, 24], compressed decorrelation with values above 1 is found in the non-equilibrium glass [3, 5, 9, 12, 14, 24, 25, 26, 27]. Yet, a failure of the KWW model due to the complex superposition of stretched and compressed components as observed in the present work is, to our best knowledge, a novel feature that has not been reported so far. We attribute this observation to the fast exposure time of 0.01 s, which allows us to explore almost 5 orders of magnitude in timescale and to observe the glass transition upon temperature scanning as a gradual crossover between liquid-typical stretched and glassy-typical compressed decay. Earlier XPCS studies at third generation synchrotrons were technically limited to exposure times in the order of seconds [3, 5, 7, 9, 10, 25, 27, 28, 29, 30, 31], therefore lacking the temporal resolution to detect such features.
The complex decorrelation behavior in the non-equilibrium asks for an adapted fitting approach. In general, the temporal decorrelation of the ISF results from all the relative positional changes between all scattering particles N that occur within the time coordinate t [32, 33, 34, 35]:
where and are the scattering factors of the respective particles. The liquid-like stretched exponential shape can be interpreted in terms of heterogeneous dynamics, , consisting of the contribution of exponential relaxations with a distribution extending over several decades. If spatial regions of the probed volume become extremely different in dynamics, with well separated timescale distributions, the heterogenous dynamics evolve from a stretched to a step-like decay. Such clearly separated relaxation phenomena can be modelled by a sum of two (or more) KWW expressions as given in Eq. (1) [15, 36, 37, 38]. Yet, compressed decorrelation cannot be explained by a distribution of exponential contributions. In this case, some models have demonstrated that stress-driven particle rearrangements can produce a compressed decay of the ISF [39]. Our ansatz will be the following: In the non-equilibrium state, the scattering particles can be subjected to two different types of motion, one liquid-like leading to a stretched decorrelation component (index S) and another one resulting in a compressed decorrelation component (index C), which is activated once the system is rigid enough to build the necessary stresses.
The relative positional changes of particles, therefore, result from a sum of these two types of particle movement, . Hence, it follows from Eq. (4) that the global function can be approached by a factorization [34, 40] as
Regarding the Siegert relation in Eq. (2), a multiplicative KWW fit function unfolds for the observed decay, termed in the following:
The fits of the six representative evaluation batches are shown in Figure 1B and C as green solid lines. In the SCL state (B3 and B4), no significant increase in fit quality is gained by changing from conventional KWW to , since both approaches create practically overlapping curves. This changes drastically in the glass transition region (B2 and B5), where provides adequate fitting while the conventional KWW fails to do so. In the glassy state (B1 and B6), the misfit between data and KWW fit might be less distinct than in the glass transition region, but still, allows a way better description.
Figure 2 focuses on batch B5 located in the glass transition region upon cooling to provide an in-depth comparison of the suboptimal KWW fit (black curve) and the well-functioning (green curve) approach. The orange and purple dashed curves depict the two components of the fit, as well as . Here, the rather fast ( s), but also highly compressed () provides almost no decorrelation in the first three decades below 10 s. This leads to the apparently counterintuitive result that the short-time domain is rather dominated by the slow ( s), but quite stretched () . It is only after the initial 10 s that the compressed gains momentum and starts to dominate, allowing to describe the cut-off appearance of the decay. Here, it shall be mentioned that the factorization in Eq. (5) is strictly valid only if the timescales of the two motion types are sufficiently different from each other [40] and, therefore, one of the two processes clearly dominates the decay. Hence the model will be an accurate approach in the short time region, the first three decades in Figure 2, where only the fast time portion of the broad relaxation time distribution underlying the stretched decay contributes to the heterogeneous ISF. In the rather narrow timescale region dominated by the compressed cut-off decorrelation, the two sources of motion cannot be precisely disentangled, and the multiplicative approach should be interpreted as a functional shape that can adapt to model the structural dynamics participated by both sources of motion.

Figure 2. The fitting approach in detail. Intensity autocorrelation function of the representative batch B5, located in the glass transition region upon cooling as indicated in Figure 1. The black solid curve represents the conventional KWW fit, which fails to describe the cut-off shape of the decorrelation. Instead, the fit (green solid line) describes the decay exceptionally well. It results from a multiplication of a stretched () and a compressed () component (orange and purple dashed lines).
Introducing additional fit parameters, and, therefore, additional degrees of freedom, to a fit function easily improves its adaptivity towards a given data set. To validate the physical meaning behind the fit results, Figure 3 compares and values obtained from conventional KWW and during both, heating and cooling. Again, the ex-situ DSC scans are provided, see Figure 3A and D, and the thermal ranges of the glass, glass transition, and SCL regions are separated by gray, light gray, and white backgrounds. For reasons of clarity, fit results are only displayed in the temperature regions where they appear meaningful. Hence, the KWW results are only shown in the SCL state, while the data is limited to the glass and glass transition region. Nevertheless, the complete data sets over the full temperature range can be found in Figure S2 in the Supplementary Materials (SM).

Figure 3. KWW and KWW fit results. Reference DSC scans for heating and cooling are provided in (A) and (D). The temperature range is accordingly separated into three regions, glass (dark gray background), glass transition (light grey background), and SCL (white background). (B), (C), (E), and (F) compare the relaxation time and shape exponent data resulting from conventional KWW and KWW fitting. As demonstrated in Figure 1, KWW only provides meaningful results in the SCL state, while KWW excels in the glass and glass transition region but appears redundant in the SCL. Accordingly, the data sets are reported only for the temperature regions where the different models provide reliable and more accurate fits (see also Figure S2 in the SM). The KWW fit parameters in the SCL show typical liquid behavior, namely a steep temperature dependence of and values below 1. In the glass transition region, the KWW approach reveals large differences between its two components KWW and KWW: While the KWW parameters follow glass-typical trends, in particular a relatively temperature-insensitive course of and a highly compressed shape, the KWW parameters show liquid-like behavior in form of a steep temperature dependence of and a stretched shape.

Figure S2. This figure is identical to Fig. 3 in the main text, only providing the complete data sets over the whole temperature range. In the glass and the glass transition region, the KWW fits fail to describe the decorrelation curves properly, the data should therefore be handled with care. Within the SCL state, the fit approach provides no significant improvement over the KWW fit and therefore appears redundant. Accordingly, the and parameters show signs of overdetermination in form of increased scatter and fitting errors.
Upon heating in the glass and through the full glass transition region, in Figure 3B features a glass-typical weak temperature dependence in agreement with previous works [3, 35]. While is roughly in the order of at lower temperatures in the glass, it departs from at about 480 K and starts to rise. We interpret this as aging behavior, the onset of major structural rearrangements in the proximity of the glass transition and the consequent relaxation of the system towards the SCL upon heating with a rate much slower than the fast quenching used to produce the as-spun glass [3, 35]. reaches a maximum in the initial part of the calorimetric glass transition at about 506 K and then adopts a steep negative temperature dependence, hence, approaching a liquid-like trendline. When entering the SCL, and merge into the conventional KWW curve, which continues the liquid-like temperature trend previously seen for to finally reach values as fast as 0.2 s at the maximum temperature. The shape exponents and depart drastically in the glass and glass transition region. With values in the order of 2, is extremely compressed, but abruptly drops to values around unity when reaching the SCL state. In contrast, features stretched values between 0.35 and 0.7 in the whole observed temperature range. In the glass, it shows a slight downwards trend with rising temperature. Yet, at 506 K, the temperature at which enters the liquid-like trendline, also starts to feature a faint upwards trend, which is in agreement with previous data obtained in the SCL by N. Neuber et al. [13]. This temperature behavior of also resembles the one observed in macroscopic stress relaxation dynamics [41].
Regarding the results of the cooling scan in Figure 4E and F, the and values obtained in the SCL state basically show the similar temperature trends as during heating. When entering the glass transition region, and again decouple drastically. While changes to a glass-typical, nearly temperature-insensitive course, follows the liquid-like trendline until the end of the glass transition at about 490 K, where it finally leaves the equilibrium course at high values in the order of s. The decoupling is simultaneously observed in the shape exponents. rises to highly compressed values of about 2, while keeps slightly decreasing with temperature until reaching values as low as 0.3 when the glass is approached. Comparably low values of the stretching exponent in well relaxed metallic glasses are found in macroscopic relaxation dynamics and simulations [42, 43].

Figure 4. Fragility comparison. and data from KWW and KWW fitting as well as data from isothermal XPCS [13] are converted into viscosity values and compared to equilibrium viscosity data determined by thermomechanical analysis (TMA) [20], which serve as a reference. All data sets are described by the same cooperative shear model (CSM) fit curve, demonstrating that the different methods observe the same fragility. The start and end temperatures of the glass transition in the XPCS heating and cooling scans are indicated by red and blue arrows, respectively.
Overall, the astonishingly good temperature agreement between calorimetric signals and changes in the KWW and fitting parameters shall be emphasized, especially concerning the thresholds between glassy state, glass transition region, and SCL. These findings speak in favor of the applied temperature correction procedure, which is described in the Materials and Methods section and in the SM.
Discussion
An option to validate the temperature scanning XPCS method is to evaluate if it arrives at the same supercooled liquid fragility as determined by other experimental approaches. For this purpose, Figure 4 compares the present KWW-fitted results to their equivalents from isothermal XPCS by N. Neuber et al. [13] as well as to equilibrium viscosities stemming from thermomechanical analysis (TMA) by O. Gross et al. [20]. To allow a direct comparison in the same physical quantity, the relaxation times are converted into viscosity values, , using the Maxwell relation [44] (G being a shear modulus) and the cooperative shear model (CSM) [45, 46], as explained in detail in the Materials and Methods section. The data sets show broad agreement and obey the same CSM fit curve (black line), demonstrating their basically identical temperature dependence, i.e., fragility. This demonstrates that temperature scanning XPCS can provide substantial results on par with isothermal XPCS [13] and macroscopic approaches like the here shown TMA [20] or also calorimetry [20, 47].
The data from the KWW fitting are also converted in viscosities, , and aligned in Figure 4. follows the CSM fit and the course of the TMA equilibrium viscosities throughout most of the glass transition region upon heating and cooling, as indicated by the arrows. This behavior confirms a trend already observed in Figure 3, which is the KWW component extrapolating the liquid-like dynamics from the equilibrium SCL state into the non-equilibrium glass transition region. During cooling, signs of vitrification can only be observed in the glass region, where leaves the equilibrium trendline. Upon heating, the KWW component reflects aging of the relatively unstable as-spun glass, as gradually evolves towards the equilibrium trendline. Hence, KWW can be seen as the KWW component that thaws first during heating and freezes last during cooling.
The shape exponents and in Figure 3C and F feature values distinctly below unity that reflect the stretched exponential decay typical for the heterogeneous nature of supercooled metallic glass forming liquids [6, 12, 13, 24]. Furthermore, and decrease with decreasing temperature, as previously observed in the isothermal XPCS studies by N. Neuber et al. [13]. This temperature trend can be attributed to the fact that liquid dynamics become more temporally [48, 49, 50, 51] and spatially [50, 52, 53] heterogeneous with undercooling. While the former describes a general, non-localized tendency towards a broadening relaxation time distribution, , the latter specifically refers to spatial fluctuations in the dynamics [53, 54]. P. Voyles et al. recently used novel electron correlation spectroscopy (ECS) [55] to image such spatio-temporally heterogeneous dynamics in supercooled PtCuNiP, an alloy similar to our present system [56, 57]. Large differences in relaxation time of up to two orders of magnitude were observed between neighboring nm-sized domains, thus, on a length scale that is typical for the medium range order (MRO). This implies a sub-diffusive [48, 58, 59] structural relaxation process that is governed by cooperative atomic rearrangements and caging effects.
The KWW component of KWW combines non-equilibrium-associated properties like highly compressed values and a weak temperature dependence of in the whole glass and glass transition region, see Figure 3. KWW therefore exhibits the inverse behavior of KWW, as it thaws last during heating and freezes first upon cooling. Similar compressed decays in photon correlation spectroscopy studies were first reported for non-equilibrium materials like jammed colloids, concentrated emulsions, and clays [39, 60, 61, 62, 63, 64]. Its appearance is mostly attributed to superdiffusive dynamics in form of ballistic motions, which manifest as collective particle movements promoted by gradients of internal stresses that arise during jamming or vitrification [60, 61, 65]. Later XPCS studies confirmed compressed decorrelation to be also a general feature of vitrified metallic glass formers [3, 5, 12, 14, 25, 26, 27], making a similar stress-based origin likely. Purely ballistic dynamics imply straight particle trajectories that create an archetypically compressed decay of the ISF with a value of 2 [34, 64, 66, 67, 68]. Furthermore, they are characterized by a -dependent relaxation time according to [60]. Recent XPCS results by A. Cornet et al. [14] suggest a rather ballistic-like -dependence also for glassy PtCuNiP.
To state an interim conclusion, the heterogeneous and likely sub-diffusive atomic motions of the -relaxation process seem to survive, to some degree, in the non-equilibrium, as indicated by the presence of the typical stretched exponential decay (KWW). Yet, they appear superimposed by a second type of atomic motion that is characteristic for the non-equilibrium state and can be described by the compressed decay component (KWW). This latter process arises from longer range elastic interactions leading to ballistic-like particle motions, probably related to intermittent cluster dynamics [14, 69, 70]. The question arises, how these stress-induced ballistic-like motions can be visualized. The present homodyne experiment observes dynamics at the value of the structure factor maximum and therefore probes both, the self and the distinct part of the ISF [35], thus reflecting the temporal decorrelation of the structure not only at the particle-particle distance but also on the MRO length scale [71]. Hence, specific traits of the stress fields that cause the compressed decay, such as the length and time scale of the stress fluctuations, or the intensity of stress gradients, cannot be straightforwardly quantified from the current observations. Publications exploring the relation between the particle dynamics and the homodyne function in multicomponent systems in the non-equilibrium are still scarce [69], and therefore, we can only propose here a qualitative picture.
Nevertheless, many works interpret the glass transition in terms of heterogeneities in dynamics, structure, and density [52, 53, 72, 73, 74, 75, 76, 77] and considering the earlier mentioned spatio-temporally heterogeneous dynamics imaged via ECS in a compositionally similar Pt-based alloy [56, 57], such a scenario suggests itself in case of the present PtCuNiP system. We thus want to offer a picture that identifies spatial domains with slower relaxation times as the carriers of the stressdriven, ballistic-like motions. To illustrate our considerations in detail, we will focus on the cooling scan, where the SCL gently vitrifies without interfering aging effects as observed during heating. For this purpose, Figure 5 connects the representative batches B4, B5, and B6 from Figure 1 with simplified visualizations of the atomic dynamics at the given temperatures.

Figure 5. _Spatial heterogeneities, stress-driven dynamics, and the glass transition._ We present a possible scenario that illustrates the micro-scale dynamics during the cooling scan and connects it to the experimentally determined interplay of stretched and compressed decorrelation components. The shown curves correspond to the representative batches B4, B5, and B6 introduced in Figure 1. Individual liquid-like atomic motions are symbolized by orange arrows. _(A)_ In the equilibrium SCL state, spatio-temporally heterogeneous dynamics already have evolved. Regions with slower dynamics, termed rigid domains, are indicated by underlying gray shadows. The broad spectrum of relaxation times creates a stretched decay in that can be described by a conventional KWW function. _(B)_ Further cooling towards vitrification leads to increased spatial heterogeneity as well as volume shrinkage of the system. Accordingly, the rigid domains start to gradually interlock and jam to form a rigid dynamic backbone. Further volumetric adaptations to temperature changes are frustrated, the system falls out of equilibrium and stress gradients arise. This causes the rigid domains to push, drift, and rotate against each other (underlying purple arrows), introducing collective ballistic-like atomic motions and a typical compressed decorrelation in that competes with the stretched liquid-like decorrelation component. Hence, the conventional KWW fit fails and the KWW approach is needed to model the complex shape of the decay. _(C)_ In the glass, the rigid dynamic backbone is broadly established, and the compressed decay caused by the ballistic-like motions dominates the decorrelation.
Figure 5A sketches the spatially heterogeneous dynamics of the equilibrium SCL. Here, the liquid-like and likely sub-diffusive motions of individual atoms are illustrated by orange arrows. Heterogeneity manifests spatially through nm-sized [56, 57] domains of diverse mobility, likely corresponding to fluctuations in density or atomic packing [76, 78]. Domains with slower dynamics, which will be referred to as ‘rigid’ in the following, are highlighted by grey shadows to guide the eyes. In the SCL, the overall fast atomic mobility still enables the system to compensate perturbations immediately without ‘falling out of equilibrium’. The measured decorrelation is of stretched exponential shape, indicating a broad distribution of relaxation timescales spanning over several decades, and can be fitted by a conventional KWW model.
With ongoing cooling, the dynamic heterogeneity further increases [56, 57] while the overall free volume of the system decreases. Eventually, this must lead to the glass transition, which is a gradual, rate-dependent kinetic arrest [79, 80] that manifests calorimetrically as a step in the system’s heat capacity beginning at about 521 K, see Figure 1A and Figure 3D. It can be assumed that the more rigid domains play a decisive role in this process, as their slow dynamics likely lend them rather viscoelastic than liquid-like behavior [74], therefore enabling them to bear certain stresses in a solid-like manner. Hence, we pick up an approach earlier formulated for polymeric melts [77] and interpret vitrification as an interlocking and jamming process of the more rigid domains. The interlocked domains then act as a gradually formed backbone structure that frustrates further volume shrinkage with ongoing cooling, and so, the system starts falling out of equilibrium. The volumetric frustration causes stress gradients among the rigid domains, forcing them to perform (small-scale) ballistic-like collective motions that could be imagined as drifts or rotations, as illustrated by the purple underlying arrows in Figure 5B. A macroscale analogy can be seen in the dynamics of drift ice, where interactions within densely packed ice floe agglomerations (representing the backbone structure formed by jammed rigid domains) create tectonic motions [81]. Eventually, the present homodyne XPCS experiment detects the velocity gradients [34] that go hand in hand with these ballistic-like ‘microscopic drift’ motions in form of the typical compressed decay modelled by KWW. Nevertheless, the gradual nature of the glass transition implies that stress-relaxation also still occurs through liquid-like -relaxation processes that remained partially active, and therefore, the KWWs parameters still follow their liquid-like trends. Accordingly, vitrification is characterized by rivaling types of atomic motions. This is mirrored in the ISF by different decorrelation components with roughly the same timescale, finally creating the typical cut-off appearance described by the KWW model.
At about 490 K, the end of the glass transition is reached, where the heat flow signal approaches the nearly constant glassy level, as shown in Figure 1A and Figure 3D. Here, the last portions of liquid-like -relaxation processes become arrested and, therefore, and finally start departing from their equilibrium trendlines as seen in Figure 3 and Figure 4. Regarding our qualitative picture in Figure 5C, the rigid dynamic backbone has broadly evolved, and stresses mainly relax through ballistic-like drift motions. Accordingly, the observed decorrelation is dominated by the typical compressed decay.
The vitrification in the cooling scan is a rather ‘gentle’ process [82, 83, 84]. The system leaves the equilibrium at a low fictive temperature thanks to the slow cooling rate. In contrast, the as-spun ribbon features a distinctly higher fictive temperature, since it has been vitrified with a rate in the order of K/s [30]. The resulting configurational differences between the as-spun and the slowly cooled glass are observed in the particle motion component corresponding to liquid-like dynamics, KWWs, since and values significantly differ between the heating and the cooling scan (compare Figure 3C and F). In view of our scenario, a higher fictive temperature configuration corresponds to a less rigid dynamic backbone and a faster liquid-like component. The higher atomic mobility in the as-spun glass allows for structural relaxation upon heating, as it is visible in Figure 3B and C, where and age towards equilibrium. When these parameters have reestablished their equilibrium trendlines, further devitrification with increasing temperature basically occurs as the inverse vitrification process, namely via a gradual dissolving of the interlocked rigid regions until a full loss of compressed decorrelation characteristics is reached when approaching the equilibrium SCL.
In summary, we used slow temperature scanning XPCS to study the micro-scale dynamics of a Pt-based metallic glass former in the glass, glass transition, and SCL state upon heating and cooling. By assuming two different types of particle motion, we describe a scenario that reduces the difference between equilibrium and non-equilibrium again to a question of crossing timescales. The system remains in equilibrium as long as liquid-like atomic motions are able to relax perturbations within the timeframe determined by the underlying thermal protocol, creating a purely stretched decorrelation of the ISF. The glass transition upon cooling initiates as soon as rigid spatial domains start to interlock, causing volumetric frustration with further decreasing temperature. The resulting stress gradients induce collective particle movements in form of ballistic-like ‘microscopic drift’ motions, eventually creating a compressed decorrelation component that competes with the stretched decay component stemming from the liquid-like atomic motions. The present KWW approach enables us for the first time to model this interplay and clearly separate these coexisting decorrelation components. It shall be highlighted again that the fast exposure time of 0.01 s is the crucial technical condition to resolve the presence of such different dynamic processes, as exemplarily shown in Figure 2. Thus, we address the fact that superimposed stretched and compressed decay was never reported in earlier XPCS studies on metallic glasses to limitations of the temporal resolution. Accordingly, we expect this superposition to become a regularly observed effect with future technological advances. Furthermore, we can imagine that both approaches, temperature scanning XPCS as well as multiplicative KWW fitting, will become mighty tools to study countless cases of non-equilibrium states or transition effects in amorphous matter, e.g. regarding phase separations, liquid-liquid transitions, or secondary relaxations.
Materials and Methods
Sample preparation
The sample preparation of PtCuNiP consisted of inductively alloying the raw elements to obtain the master alloy, which was further purified through a fluxing treatment. An amorphous plate-shaped bulk specimen was produced from the master alloy by induction melting and tilt-casting. A detailed description of these processes can be found in [13, 20]. For XPCS, amorphous ribbons with approximately 30 m thickness were produced from the master alloy by melt-spinning as reported in [13].
XPCS experimental procedure
XPCS was carried out at the ID10 beamline at ESRF. The measured as-spun ribbon piece was sanded to a thickness of 10 m to achieve a regular cross section and an optimized signal-to-noise ratio. The partially coherent beam with a cross section of featured an energy of 21.0 keV and a photon flux of about 1/s. The speckle patterns were obtained in frames of 0.01 s exposure time using an Eiger 4M2 CdTe detector with a sample-detector-distance of 5100 mm. All measurements were performed at a fixed wave vector of Å, directly at the first sharp diffraction peak of the structure factor [13, 85]. The as-spun ribbon sample was installed in a custom-build, PID-controlled furnace to apply a temperature scan program under high vacuum. In a first step, the sample was heated with a rate of 0.0167 K/s from the glassy state into the supercooled liquid to 548 K. In the second step, the sample was subsequently cooled back into the glassy state, again with 0.0167 K/s. The two times correlation functions (TTCFs), , were obtained experimentally as
where is the pixel intensity of the detector at the (absolute) experimental time and is the average over all pixels of the detector [22, 28]. Considering that the intensity fluctuations are well described by Gaussian statistics and using the Siegert relationship [86, 87], the TTCFs were averaged over batches of 24000 frames to obtain the intensity autocorrelation functions, . This batch size corresponds to a time span of s or, respectively, a temperature span of K and was found to be adequate for further KWW and KWW fitting as it provides enough temporal range to capture most of the decorrelation process while still retaining a reasonable error in temperature within each batch ( K). To achieve a sliding average effect, consecutive batches overlapped halfway, hence by 12000 frames (120 s, 2 K).
Temperature correction procedure
To achieve a precise temperature calibration, further XPCS heating scans with the same experimental conditions as described earlier were performed on two other metallic glass forming systems, which show first signs of crystallization at about 390 K and 630 K, respectively. As soon as crystals appeared in the probed sample volume, they created quasi-stationary Bragg reflections in the speckle pattern. This leads to an abrupt and harsh increase of the baseline as shown in Figure S1 in the Supplementary Materials (SM), allowing for a precise definition of the crystallization temperature . Differential scanning calorimetry (DSC) scans of these systems were performed with the same heating rate of 0.0167 K/s using a Mettler Toledo DSC3 with Al crucibles under high-purity Ar flow. Here, was identified as starting point of the exothermal crystallization event in the heat flow signal, see also Figure S1 in the SM. Comparing the values defined by calorimetry and XPCS scans allowed to define a linear two-point temperature calibration that is applied for all XPCS data shown in this article. Figure S1C in the SM provides the correction formula.

Figure S1. XPCS temperature calibration procedure used to correct the temperatures published in the main article. In (A), a low-temperature metallic glass former is heated with 1 K/min in an ex-situ DSC scan as well as in a XPCS temperature scan (here, the uncorrected raw temperature is used for the abscissa). Crystallization manifests as a sharp exothermal heat flow event, or respectively, as a distinct increase in the detected minimum of each batch. The offset between both events is 18 K. The same procedure is applied for a high-temperature metallic glass former in (B), allowing to establish a linear correction function as shown in (C).
Ex-situ DSC scans
To provide heat flow signals for comparison and orientation, the temperature corrected heating-cooling program applied for the XPCS studies on the PtCuNiP ribbon was retraced ex-situ by DSC, again using the Mettler-Toledo DSC3 with Al crucibles under high-purity Ar flow. The as-spun PtCuNiP ribbons were heated from 463 K to 548 K and subsequently cooled back to the initial temperature of 463 K, using a rate of 0.0167 K/s.
KWW and KWW fitting procedure
The first approach to fit the data using the KWW and KWW models, see Eq. (3) and Eq. (6), would be to leave all parameters free to obtain fitting curves that describe each individual data set to the best extent. Yet, at elevated temperatures in the SCL, the fast dynamics cause significant decorrelation even within the first time increment of 0.01 s, resulting in curves with too low initial heights, as demonstrated for example in Figure 1B and C. Here, KWW fitting with free parameters leads to an underestimation of c. In contrast, the curves at low temperatures in the glassy state may not reach full decorrelation within the correlation window as can be seen in Figure 1B and C, resulting in a misestimation of the baseline b in case of free parameterization. To solve these problems, we define fixed values for b and c, analogous to an XPCS analysis previously described in [12]. For b, an average value of 1.00675 is determined from those high-temperature batches that show full decorrelation. An average c value of 0.02517 is derived from low-temperature batches that show a full initial plateau. Hence, only and remain as free fitting parameters.
Aligning XPCS timescale data and equilibrium viscosity data
Equilibrium viscosities from thermomechanical analysis (TMA) published in [20] are shown in Figure 4 and serve us as a reference in terms of the SCL state’s fragility. To allow a direct comparison, the present and data from KWW and KWW fitting are transferred into the viscosity domain and aligned with the TMA data. To do so, the equilibrium viscosities are fitted in a first step using the cooperative shear model (CSM) [45, 46]:
Here, is the minimum viscosity reached at high temperatures of Pa s [20], is the viscosity value of Pa s commonly defined by convention as the glass transition value [1] and is the corresponding temperature value, which is 498 K in the present case. The only remaining free fit parameter is a measure of the apparent fragility and is determined as . This corresponds to an m-fragility of roughly 57 [20, 47] and a Vogel-Fulcher-Tammann (VFT) fragility parameter of 15.3 [13, 20, 47]. In a second step, the CSM model is combined with the Maxwell relation [44], , to be applied in the timescale domain as
Fitting the equilibrium relaxation times (hence excluding data points that indicate vitrification due to deviation from the liquid behavior) from XPCS with this equation using the fixed from the viscosity fit leaves only the shear modulus as a free fit parameter. The thereby determined values are Pa (, heating), Pa (, cooling), Pa (, heating), and Pa (, cooling). Now, the relaxation times can be transformed into corresponding viscosity data by means of the Maxwell relation and each respective value. The resulting and values are depicted in Figure 4. Their steepnesses agree well with the TMA equilibrium viscosities, indicating the same fragility among the data sets.
Acknowledgments
We acknowledge ESRF (Grenoble, France), for providing beamtime. We want to thank the technicians and staff at ESRF for their support in terms of the XPCS studies. We further thank Dr. Frank Aubertin for fruitful discussions. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant Agreement No 948780).
Funding: European Research Council (ERC), European Union’s Horizon 2020 research and innovation programme, Grant Agreement No 948780 (AC, JS, BR)
Author contributions: Conceptualization: EP, MF Methodology: EP, NN, MF Validation: FW, MS Formal analysis: EP, MF, NN Investigation: EP, NN, AC, YC, FZ, BR, MF Resources: NN, MF Visualization: MF, EP Supervision: EP Writing - original draft: MF, EP, BR Writing - review & editing: NN, SSR, AC, YC, FZ, LR, BA, MN, FY, JS, FW, MS, DC, VDL, IG, RB
Competing interests: The authors declare that they have no competing interests.
Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this work may be requested from the authors.
This PDF file includes:
Figs. S1 to S2
References
- [1]C. A. Angell, Formation of glasses from liquids and biopolymers. Science 267, 1924–1935 (1995).
- [2]G. Grübel, F. Zontone, Correlation spectroscopy with coherent X-rays. J. Alloys Compd. 362, 3–11 (2004).
- [3]B. Ruta, Y. Chushkin, G. Monaco, L. Cipelletti, E. Pineda, P. Bruna, V. M. Giordano, M. Gonzalez-Silveira, Atomic-scale relaxation dynamics and aging in a metallic glass probed by X-ray photon correlation spectroscopy. Phys. Rev. Lett. 109, 1–5 (2012).
- [4]H. Zhou, S. Hilke, E. Pineda, M. Peterlechner, Y. Chushkin, S. Shanmugam, G. Wilde, X-ray photon correlation spectroscopy revealing the change of relaxation dynamics of a severely deformed Pd-based bulk metallic glass. Acta Mater. 195, 446–453 (2020).DOI
- [5]Z. Evenson, A. Payes-Playa, Y. Chushkin, M. Di Michiel, E. Pineda, B. Ruta, Comparing the atomic and macroscopic aging dynamics in an amorphous and partially crystalline Zr44Ti11Ni10Cu10Be25 bulk metallic glass. J. Mater. Res. 32, 2014–2021 (2017).DOI
- [6]S. Hechler, B. Ruta, M. Stolpe, E. Pineda, Z. Evenson, O. Gross, A. Bernasconi, R. Busch, I. Gallino, Microscopic evidence of the connection between liquid-liquid transition and dynamical crossover in an ultraviscous metallic glass former. Phys. Rev. Mater. 2, 1–6 (2018).
- [7]B. Ruta, V. M. Giordano, L. Erra, C. Liu, E. Pineda, Structural and dynamical properties of Mg₆₅Cu₂₅Y₁₀ metallic glasses studied by in situ high energy X-ray diffraction and time resolved X-ray photon correlation spectroscopy. J. Alloys Compd. 615, 45–50 (2014).
- [8]D. Soriano, H. Zhou, S. Hilke, E. Pineda, B. Ruta, G. Wilde, Relaxation dynamics of Pd-Ni-P metallic glass: Decoupling of anelastic and viscous processes. J. Phys. Condens. Matter 33 (2021).DOI
- [9]A. Das, P. M. Derlet, C. Liu, E. M. Dufresne, R. Maaß, Stress breaks universal aging behavior in a metallic glass. Nat. Commun. 10, 1–9 (2019).DOI
- [10]A. Das, E. M. Dufresne, R. Maaß, Structural dynamics and rejuvenation during cryogenic cycling in a Zr-based metallic glass. Acta Mater. 196, 723–732 (2020).
- [11]B. Ruta, S. Hechler, N. Neuber, D. Orsi, L. Cristofolini, O. Gross, B. Bochtler, Wave vector dependence of the dynamics in supercooled metallic liquids. Phys. Rev. Lett. 125, 1–14 (2020).DOI
- [12]N. Amini, F. Yang, E. Pineda, B. Ruta, M. Sprung, A. Meyer, Intrinsic relaxation in a supercooled ZrTiNiCuBe glass forming liquid. Phys. Rev. Mater. 5, 1–8 (2021).
- [13]N. Neuber, O. Gross, M. Frey, B. Bochtler, A. Kuball, S. Hechler, F. Yang, E. Pineda, F. Westernmeier, M. Sprung, F. Schäfer, I. Gallino, R. Busch, B. Ruta, Disentangling structural and kinetic components of the α-relaxation in supercooled metallic liquids. Commun. Phys. 5, 1–10 (2022).arxiv.org/abs/2209.09569
- [14]A. Cornet, G. Garbarino, F. Zontone, Y. Chushkin, J. Jacobs, E. Pineda, T. Deschamps, S. Li, A. Ronca, J. Shen, G. Morard, N. Neuber, M. Frey, R. Busch, I. Gallino, M. Mezouar, G. Vaughan, B. Ruta, Denser glasses relax faster : Enhanced atomic mobility and anomalous particle displacement under in-situ high pressure compression of metallic glasses. Acta Mater. 255, 119065 (2023).
- [15]A. Jain, F. Schulz, F. Dallari, V. Markmann, F. Westermeier, Y. Zhang, G. Grübel, F. Lehmkühler, Three-step colloidal gelation revealed by time-resolved x-ray photon correlation spectroscopy. J. Chem. Phys. 157 (2022).DOI
- [16]M. Jankowski, V. Belova, Y. Chushkin, F. Zontone, M. Levantino, T. Narayanan, O. Konovalov, A. Pastore, The complex systems and biomedical sciences group at the ESRF: Current status and new opportunities after extremely brilliant source upgrade. Nucl. Instruments Methods Phys. Res. Sect. B Beam Interact. with Mater. Atoms 538, 164–172 (2023).DOI
- [17]X. Lu, S. G. J. Mochrie, S. Narayanan, A. R. Sandy, M. Sprung, How a liquid becomes a glass both on cooling and on heating. Phys. Rev. Lett. 100, 5–8 (2008).
- [18]A. Ricci, G. Campi, B. Joseph, N. Poccia, D. Innocenti, C. Gutt, M. Tanaka, H. Takeya, Y. Takano, T. Mizokawa, M. Sprung, N. L. Saini, Intermittent dynamics of antiferromagnetic phase in inhomogeneous iron-based chalcogenide superconductor. Phys. Rev. B 101, 1–6 (2020).DOI
- [19]J. Schroers, W. L. Johnson, Highly processable bulk metallic glass-forming alloys in the Pt-Co-Ni-Cu-P system. Appl. Phys. Lett. 84, 3666–3668 (2004).DOI
- [20]O. Gross, B. Bochtler, M. Stolpe, S. Hechler, W. Hembree, R. Busch, I. Gallino, The kinetic fragility of Pt-P and Ni-P-based bulk glass-forming liquids and its thermodynamic and structural signature. Acta Mater. 132, 118–127 (2017).
- [21]R. Kohlrausch, Theorie des elektrischen Rückstandes in der Leidener Flasche. Ann. Phys. 167, 179–214 (1854).
- [22]Y. Chushkin, C. Caronna, A. Madsen, A novel event correlation scheme for X-ray photon correlation spectroscopy. J. Appl. Crystallogr. 45, 807–813 (2012).
- [23]E. J. Jaeschke, Synchrotron Light Sources and Free-Electron Lasers (2016).
- [24]B. Ruta, Y. Chushkin, G. Monaco, L. Cipelletti, V. M. Giordano, E. Pineda, P. Bruna, Relaxation dynamics and aging in structural glasses. AIP Conf. Proc. 1518, 181–188 (2013).
- [25]V. M. Giordano, B. Ruta, Unveiling the structural arrangements responsible for the atomic dynamics in metallic glasses during physical aging. Nat. Commun. 7, 1–8 (2016).DOI
- [26]Z. Evenson, B. Ruta, S. Hechler, M. Stolpe, E. Pineda, I. Gallino, R. Busch, X-ray photon correlation spectroscopy reveals intermittent aging dynamics in a metallic glass. Phys. Rev. Lett. 115 (2015).
- [27]X. D. Wang, B. Ruta, L. H. Xiong, D. W. Zhang, Y. Chushkin, H. W. Sheng, H. B. Lou, Q. P. Cao, J. Z. Jiang, Free-volume dependent atomic dynamics in beta relaxation pronounced La-based metallic glasses. Acta Mater. 99, 290–296 (2015).DOI
- [28]B. Ruta, Y. Chushkin, G. Monaco, L. Cipelletti, V. M. Giordano, E. Pineda, P. Bruna, Relaxation dynamics and aging in structural glasses. AIP Conf. Proc. 1518, 181–188 (2013).
- [29]T. Xu, X. D. Wang, E. M. Dufresne, Y. Ren, Q. Cao, D. Zhang, J. Z. Jiang, Anomalous fast atomic dynamics in bulk metallic glasses. Mater. Today Phys. 17, 100351 (2021).DOI
- [30]B. Ruta, G. Baldi, G. Monaco, Y. Chushkin, Compressed correlation functions and fast aging dynamics in metallic glasses. J. Chem. Phys. 138 (2013).DOI
- [31]M. Leitner, B. Sepiol, L. M. Stadler, B. Pfau, Time-resolved study of the crystallization dynamics in a metallic glass. Phys. Rev. B - Condens. Matter Mater. Phys. 86, 1–7 (2012).
- [32]P. N. Pusey, “Liquids, Freezing and Glass Transition” in Les Houches Session 51. (1991; https://books.google.co.uk/books?id=YC30nwEACAAJ).
- [33]R. Pecora, Dynamic Light Scattering: Applications of Photon Correlation Spectroscopy (2008; http://dx.doi.org/10.1007/978-1-4020-4465-6_13).
- [34]S. Busch, T. H. Jensen, Y. Chushkin, A. Fluerasu, Dynamics in shear flow studied by X-ray Photon Correlation Spectroscopy. Eur. Phys. J. E 26, 55–62 (2008).
- [35]B. Ruta, E. Pineda, Z. Evenson, Topical Review: Relaxation processes and physical aging in metallic glasses. J. Phys. Condens. Matter 29, 503002 (2017).DOI
- [36]F. Liénard, É. Freyssingeas, P. Borgnat, A multiscale time-Laplace method to extract relaxation times from non-stationary dynamic light scattering signals. J. Chem. Phys. 156 (2022).DOI
- [37]D. Parisi, J. Ruiz-Franco, Y. Ruan, C. Y. Liu, B. Loppinet, E. Zaccarelli, D. Vlassopoulos, Static and dynamic properties of block copolymer based grafted nanoparticles across the non-ergodicity transition. Phys. Fluids 32 (2020).DOI
- [38]F. A. D. M. Marques, R. Angelini, E. Zaccarelli, B. Farago, B. Ruta, G. Ruocco, B. Ruzicka, Structural and microscopic relaxations in a colloidal glass. Soft Matter 11, 466–471 (2015).
- [39]E. E. Ferrero, K. Martens, J. L. Barrat, Relaxation in yield stress systems through elastically interacting activated events. Phys. Rev. Lett. 113, 248301 (2014).
- [40]G. G. Fuller, J. M. Rallison, R. L. Schmidt, L. G. Leal, The measurement of velocity gradients in laminar flow by homodyne light-scattering spectroscopy. J. Fluid Mech. 100, 555–575 (1980).DOI
- [41]Y. J. Duan, L. T. Zhang, J. Qiao, Y.-J. Wang, Y. Yang, T. Wada, H. Kato, J. M. Pelletier, E. Pineda, D. Crespo, Intrinsic Correlation between the Fraction of Liquidiikde Zones and the beta Relaxation in High-Entropy Metallic Glasses. Phys. Rev. Lett. 129, 175501 (2022).DOI
- [42]Y. J. Duan, M. Nabahat, Y. Tong, L. Ortiz-Membrado, E. Jiménez-Piqué, K. Zhao, Y. J. Wang, Y. Yang, T. Wada, H. Kato, J. M. Pelletier, J. C. Qiao, E. Pineda, Connection between Mechanical Relaxation and Equilibration Kinetics in a High-Entropy Metallic Glass. Phys. Rev. Lett. 132, 1–7 (2024).DOI
- [43]Y. T. Sun, R. Zhao, D. W. Ding, Y. H. Liu, H. Y. Bai, M. Z. Li, W. H. Wang, Distinct relaxation mechanism at room temperature in metallic glass. Nat. Commun. 14, 1–8 (2023).
- [44]C. Maxwell, On the dynamical theory of gases. Philos. Trans. 157, 49–88 (1867).
- [45]M. D. Demetriou, J. S. Harmon, M. Tao, G. Duan, K. Samwer, W. L. Johnson, Cooperative shear model for the rheology of glass-forming metallic liquids. Phys. Rev. Lett. 97, 41–44 (2006).
- [46]W. L. Johnson, M. D. Demetriou, J. S. Harmon, M. L. Lind, K. Samwer, Rheology and ultrasonic properties of metallic glassforming liquids: A potential energy landscape perspective. MRS Bull. 32, 644–650 (2007).DOI
- [47]M. Frey, N. Neuber, O. Gross, B. Zimmer, W. Possart, R. Busch, Determining the fragility of bulk metallic glass forming liquids via modulated DSC. J. Phys. Condens. Matter 32 (2020).DOI
- [48]A. Arbe, J. Colmenero, M. Monkenbusch, D. Richter, Dynamics of glass-forming polymers: “Homogeneous” versus “heterogeneous” scenario. Phys. Rev. Lett. 81, 590–593 (1998).DOI
- [49]F. Faupel, W. Frank, M.-P. Macht, H. Mehrer, V. Naundorf, K. Rätzke, H. R. Schober, S. K. Sharma, H. Teichler, Diffusion in metallic glasses and supercooled melts. Rev. Mod. Phys. 75, 237–280 (2003).
- [50]P. Ballesta, A. Duri, L. Cipelletti, Unexpected drop of dynamical heterogeneities in colloidal suspensions approaching the jamming transition. Nat. Phys. 4, 550–554 (2008).DOI
- [51]D. Cangialosi, A. Alegría, J. Colmenero, On the temperature dependence of the nonexponentiality in glass-forming liquids. J. Chem. Phys. 130 (2009).
- [52]M. D. Ediger, Spatially heterogeneous dynamics in supercooled liquids. Annu. Rev. Phys. Chem. 51, 99–128 (2000).
- [53]R. Richert, Heterogeneous dynamics in liquids: Fluctuations in space and time. J. Phys. Condens. Matter 14 (2002).
- [54]R. Richert, Perspective: Nonlinear approaches to structure and dynamics of soft materials. J. Chem. Phys. 149 (2018).DOI
- [55]L. He, P. Zhang, M. F. Besser, M. J. Kramer, P. M. Voyles, Electron Correlation Microscopy: A New Technique for Studying Local Atom Dynamics Applied to a Supercooled Liquid. Microsc. Microanal. 21, 1026–1033 (2015).
- [56]P. Zhang, J. J. Maldonis, Z. Liu, J. Schroers, P. M. Voyles, Spatially heterogeneous dynamics in a metallic glass forming liquid imaged by electron correlation microscopy. Nat. Commun. 9, 1–7 (2018).DOI
- [57]P. M. V. Debaditya Chatterjee, Ajay Annamareddy, Jittisa Ketkaew, Jan Schroers, Dane Morgan, Fast surface dynamics on a metallic glass nanowire. ACS Nano 15 (2021).
- [58]P. Chaudhuri, L. Berthier, W. Kob, Universal nature of particle displacements close to glass and jamming transitions. Phys. Rev. Lett. 99, 2–5 (2007).
- [59]E. R. Weeks, J. C. Crocker, A. C. Levitt, A. Schofield, D. A. Weitz, Three-dimensional direct imaging of structural relaxation near the colloidal glass transition. Science (80-. ). 287, 627–631 (2000).
- [60]L. Cipelletti, L. Ramos, S. Manley, E. Pitard, D. A. Weitz, E. E. Pashkovski, M. Johansson, Universal non-diffusive slow dynamics in aging soft matter. Faraday Discuss. 123, 237–251 (2003).
- [61]K. Trachenko, A. Zaccone, Slow stretched-exponential and fast compressed-exponential relaxation from local event dynamics. J. Phys. Condens. Matter 33 (2021).DOI
- [62]M. Bouzid, J. Colombo, L. V. Barbosa, E. Del Gado, Elastically driven intermittent microscopic dynamics in soft solids. Nat. Commun. 8, 1–8 (2017).DOI
- [63]R. Bandyopadhyay, D. Liang, H. Yardimci, D. A. Sessoms, M. A. Borthwick, S. G. J. Mochrie, J. L. Harden, R. L. Leheny, Evolution of particle-scale dynamics in an aging clay suspension. Phys. Rev. Lett. 93, 2–5 (2004).
- [64]J. Gabriel, T. Blochowicz, B. Stühn, Compressed exponential decays in correlation experiments: The influence of temperature gradients and convection. J. Chem. Phys. 142 (2015).DOI
- [65]N. Gnan, E. Zaccarelli, The microscopic role of deformation in the dynamics of soft colloids. Nat. Phys. 15, 683–688 (2019).DOI
- [66]F. Sciortino, L. Fabbian, S. H. Chen, P. Tartaglia, Supercooled water and the kinetic glass transition. II. Collective dynamics. Phys. Rev. E - Stat. Physics, Plasmas, Fluids, Relat. Interdiscip. Top. 56, 5397–5404 (1997).DOI
- [67]P. Gallo, F. Sciortino, P. Tartaglia, S. H. Chen, Slow Dynamics of Water Molecules in Supercooled States. Phys. Rev. Lett. 76, 2730–2733 (1996).
- [68]D. Orsi, B. Ruta, Y. Chushkin, A. Pucci, C. Ruggeri, G. Baldi, T. Rimoldi, L. Cristofolini, Controlling the dynamics of a bidimensional gel above and below its percolation transition. Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys. 89, 1–11 (2014).
- [69]Z. Wang, B. Riechers, P. M. Derlet, R. Maaß, Atomic cluster dynamics causes intermittent aging of metallic glasses. Acta Mater. 267, 119730 (2024).DOI
- [70]G. Rodriguez-Lopez, K. Martens, E. E. Ferrero, Temperature dependence of fast relaxation processes in amorphous materials. Phys. Rev. Mater. 7, 1–19 (2023).DOI
- [71]B. Wu, T. Iwashita, T. Egami, Atomic Dynamics in Simple Liquid: De Gennes Narrowing Revisited. Phys. Rev. Lett. 120, 135502 (2018).DOI
- [72]C. Chang, H. P. Zhang, R. Zhao, F. C. Li, P. Luo, M. Z. Li, H. Y. Bai, Liquid-like atoms in dense-packed solid glasses. Nat. Mater., doi: 10.1038/s41563-022-01327-w (2022).DOI
- [73]Q. Wang, Y. Shang, Y. Yang, Quenched-in liquid in glass. Mater. Futur. 2 (2023).
- [74]Z. Wang, B. A. Sun, H. Y. Bai, W. H. Wang, Evolution of hidden localized flow during glass-to-liquid transition in metallic glass. Nat. Commun. 5 (2014).DOI
- [75]J. C. Qiao, Q. Wang, J. M. Pelletier, H. Kato, R. Casalini, D. Crespo, E. Pineda, Y. Yao, Y. Yang, Structural heterogeneities and mechanical behavior of amorphous alloys. Prog. Mater. Sci. 104, 250–329 (2019).
- [76]H. Tanaka, Two-order-parameter description of liquids. I. A general model of glass transition covering its strong to fragile limit. J. Chem. Phys. 111, 3163–3174 (1999).
- [77]D. Long, F. Lequeux, Heterogeneous dynamics at the glass transition in van der Waals liquids, in the bulk and in thin films. Eur. Phys. J. E 4, 371–387 (2001).
- [78]P. M. Derlet, H. Bocquet, R. Maaß, Viscosity and transport in a model fragile metallic glass. Phys. Rev. Mater. 5, 1–7 (2021).
- [79]E. Donth, The Glass Transition: Relaxation Dynamics in Liquids and Disordered Materials (Springer Science & Business Media, 2013)vol. 48.
- [80]J. W. P. Schmelzer, I. S. Gutzow, Glasses and the Glass Transition (2011).
- [81]T. Martin, Arctic Sea Ice Dynamics: Drift and Ridging in Numerical Models and Observations. Univ. Bremen, 1–248 (2007).
- [82]C. Rodríguez-Tinoco, M. Gonzalez-Silveira, J. Ràfols-Ribé, A. Vila-Costa, J. C. Martinez-Garcia, J. Rodríguez-Viejo, Surface-Bulk Interplay in Vapor-Deposited Glasses: Crossover Length and the Origin of Front Transformation. Phys. Rev. Lett. 123, 155501 (2019).
- [83]M. Lulli, C. S. Lee, H. Y. Deng, C. T. Yip, C. H. Lam, Spatial Heterogeneities in Structural Temperature Cause Kovacs’ Expansion Gap Paradox in Aging of Glasses. Phys. Rev. Lett. 124, 095501 (2020).
- [84]M. Ruiz-Ruiz, A. Vila-Costa, T. Bar, C. Rodríguez-Tinoco, M. Gonzalez-Silveira, J. A. Plaza, J. Alcalá, J. Fraxedas, J. Rodriguez-Viejo, Real-time microscopy of the relaxation of a glass. Nat. Phys., doi: 10.1038/s41567-023-02125-0 (2023).
- [85]O. Gross, N. Neuber, A. Kuball, B. Bochtler, S. Hechler, M. Frey, R. Busch, Signatures of structural differences in Pt–P- and Pd–P-based bulk glass-forming liquids. Commun. Phys. 2, 83 (2019).
- [86]E. Bartsch, V. Frenz, J. Baschnagel, W. Schärtl, H. Sillescu, The glass transition dynamics of polymer micronetwork colloids. A mode coupling analysis. J. Chem. Phys. 106, 3743–3756 (1997).
- [87]L. Cipelletti, D. A. Weitz, Ultralow-angle dynamic light scattering with a charge coupled device camera based multispeckle, multitau correlator. Rev. Sci. Instrum. 70, 3214–3221 (1999).DOI