WorksPhysics and Astronomy
X-ray photo-induced atomic motion in Phase Change Materials and conventional covalent chalcogenide glasses
Abstract
X-ray Photon Correlation Spectroscopy (XPCS) enables direct access to atomic-scale dynamics in disordered materials, revealing both spontaneous and X-ray-induced relaxation processes. Here, we study two compositionally similar alloy glasses near their glass transition temperatures: the phase change material (PCM) Ge15Sb85 and the non-PCM alloy Ge15Te85. Both exhibit X-ray induced atomic motion, yet with markedly different responses. Ge15Sb85 undergoes an immediate transition to a photo-induced yielding state, characterised by stationary dynamics governed solely by the absorbed dose. In contrast, Ge15Te85 shows a progressive slowing-down of the relaxation process, accompanied by a crossover from compressed to stretched exponential decay in the density autocorrelation functions. This behaviour is consistent with the emergence of liquid-like collective motion as supported by de Gennes narrowing in the wave-vector dependence of the dynamics at length scales comparable with the first sharp diffraction peak. Unlike Ge15Sb85, this alloy does not reach a stationary regime within experimental timescales, implying that the yielding transition occurs only after thousands of seconds with the available dose rate. Its response is also temperature dependent: at lower temperatures, the dynamics reflects intrinsic stress relaxation processes, whereas at higher temperatures becomes dose-controlled. These findings demonstrate that the dynamical response to X-ray excitation is not determined solely by chemical composition or bonding character, but results from the interplay between irradiation effects and structural relaxation pathways.
Introduction
X-rays are a fundamental tool for studying the intrinsic properties of a material [1]. Over the past decades, improvements in the brilliance and coherence of X-ray beams delivered in modern synchrotron sources have led to the development of powerful techniques to investigate the dynamical properties of matter, one of the most significant being X-ray Photon Correlation Spectroscopy (XPCS) [2]. This method provides information on the collective microscopic rearrangements that occur in disordered materials at nanometric and atomic scales [2]. In XPCS, the dynamics is investigated by analysing the temporal fluctuations in the intensity of speckles, which are interference patterns generated by coherent X-rays scattered from the material. By recording series of speckle patterns, XPCS captures these fluctuations, offering a direct measure of the particle dynamics. Studies on the atomic motion in hard materials have shown that the response of these systems to the high fluxes of X-ray beams employed in XPCS can vary significantly depending on the material family [3, 4, 5, 6, 7, 8, 9, 10].
While in crystalline [3] and glassy [4] alloys XPCS provides information on the atomic diffusion and collective motion in the material, unexpected results have been reported in vitreous silica and germania [4]. In these materials, the incident flux was found to induce atomic dynamics within the samples, likely due to radiolysis [4]. This process occurs when an X-ray photon is absorbed, creating an electron-hole pair that leads to atomic displacement and the formation of a long-lived exciton. The latter gives rise to bond breaking and further atomic displacement. Subsequent studies have expanded this investigation to a variety of materials, including borate glasses [5, 6, 7, 10], sodium silicates [10], lithium-borates [6], and chalcogenide glasses [8, 11]. In particular, the studies on the oxide glass [12] and the chalcogenide glass [11] show that after a certain absorbed dose, the system reaches a stationary state of yielding, which is photo-induced by the X-rays. The specific absorbed dose depends on the physicochemical characteristics of the material.
In general, these studies collectively demonstrate that X-rays can be employed not only to probe spontaneous electron density fluctuations but also to drive them, granting access to material properties that would otherwise be inaccessible using conventional techniques. The response of a glass to X-ray irradiation contains information on the structure and the microscopic relaxation processes active in the material [4]. Hence, X-rays have emerged as a powerful probe to study spontaneous and induced dynamical fluctuations in complex systems in response to other physical parameters, such as temperature and pressure [11].
Despite extensive studies on metallic and oxide glasses, no XPCS studies have been performed on glassy phase change materials (PCM). These materials (such as , , GeTe, and Ag-In-doped ) are a class of functional materials relevant for electronic and optical switching devices [13, 14]. PCMs can be rapidly and reversibly switched between amorphous and crystalline phases by Joule heating or optical pulses. Their strong property contrast between the two phases can be used for encoding data or manipulating optical beams for functional devices. Although both PCMs and conventional chalcogenide-based non-PCM alloys (e.g. , , and ) are commonly considered as covalent glasses, they differ remarkably in physical properties such as liquid fragility, structural relaxations, crystallisation kinetics, and glass forming abilities [13, 14].
In this work, we characterize the evolution of the collective atomic motion induced by intense X-ray radiation in two compositionally similar glassy alloys, a PCM GeSb and a non-PCM GeTe, investigated in the vicinity of their glass transition temperature Tg. GeSb is a metallic-like alloy classified as a phase change material (PCM) due to its rapid and reversible glass-crystal switching properties, which are fundamental for non-volatile memory applications [15, 16, 17]. GeTe, on the other hand, is a conventional chalcogenide glass with high amorphous stability and slow crystallization kinetics, extensively studied in both liquid and glassy phases [16, 18, 19, 20, 21, 22, 23, 24, 25, 26]. The strong interest in this composition is driven by its favourable properties for selector devices, such as a high crystallisation temperature that ensures a wider operating range, a fast threshold switching speed enabling rapid off-to-on transitions, and a low threshold voltage for efficient switching [22].
From the dynamical point of view, recent studies performed by some of the authors showed that PCMs, including GeSb, shows a pronounced secondary β-relaxation in the glassy state, faster that the main α-relaxation process, implying the presence of local fast atomic motion and loosely bonded regions in the material [27]. By contrast, when Sb is replaced by Te, the similar composition GeTe shows virtually no β-relaxation, implying a rigid bonding network without local fast dynamics. The striking difference in relaxation behaviours is demonstrated to correlate with their markedly different glass forming abilities, thermal stabilities, and crystallisation kinetics [27]. While GeTe is a good glass former with a relatively stable supercooled liquid region, GeSb can easily crystallize on approaching the glass transition temperature.
As we show below, this distinction also emerges from a different dynamical response to the X-ray beams. Specifically, GeSb exhibits an immediate transition to a yielding state, similar to GeSe [11], whereas GeTe is more resistant, requiring two to three orders of magnitude higher absorbed doses to reach the same state.
Experimental methods
Samples
Glassy samples of GeSb and GeTe were prepared through magnetron sputtering deposition of stoichiometric targets synthesised following the same procedure as described in a previous study [16]. A thick uniform layer of samples was deposited onto a 4-inch Si (100) wafer using DC 10 W, 10 sccm Ar flow, and mbar Ar pressure. A steep-edge profilometer determines the sputtering rates of the target compositions to estimate the sputtering time needed for the desired layer thickness (3-8 micrometers) to match the absorption length of the samples. The amorphous layers were separated from the Si substrate in the form of pieces or flakes, where single relatively large pieces (with a diameter of several millimetres) are selected for XPCS measurements. The nominal sample compositions were confirmed (within instrumental error ∼ 3%) twice using x-ray fluorescence (XRF) with a Rigaku NEX CG, first for the sputtering target and then for the as-deposited layers. The main properties of the two samples are summarised in (Table I) [13, 21, 27].
| [K] | FSDP [nm] | [g/cm] | [m] | [m] | |
| GeSb | 500 | 20.5 | 5.323 | 3 | 8.5 |
| GeTe | 412 | 19.6 | 5.53 | 8 | 6.9 |
Table I.
Table 1. Calorimetric glass transition temperature , wavevector of the first sharp diffraction peak (FSDP) maximum, mass density , sample thickness, and X-ray absorption length at the energies used in the XPCS experiments for GeTe and GeSb [13, 21, 27].
X-ray Photon Correlation Spectroscopy
XPCS experiments were carried out during two separate beamtimes at the beamline P10 of PETRA III synchrotron in Hamburg, Germany. In the first experiment, we measured the GeTe composition. An X-ray beam of 8 keV was focused onto a spot size of full width at half maximum (FWHM) of (HV) with a photon flux of ph/s, whereas in the second experiment, we measured the GeSb using a beam energy of 8.1 keV, with a spot size of and a photon flux of ph/s. The samples were placed under vacuum, to prevent oxidation, on a Cu sample holder, whose temperature was controlled by a resistive heater and a Lakeshore temperature controller. Measurements were performed at a wave vector corresponding to the position of the maximum of the FSDP for both compositions. GeTe was studied at temperatures K (corresponding to , see values in Table I), while GeSb was measured at K (). Additionally, the GeTe composition was investigated as a function of the scattering vector in a range at 295 K.
To explore the effect of the X-rays on the dynamics, the full beam intensity was attenuated through Si attenuators (att). The fractions of the transmitted beam intensity (i.e. the transmission ) and the relative beam fluxes used in the experiments are summarised in Table II.
| GeSb | GeTe | ||||
| att | [ ph/s] | att | [ ph/s] | ||
| att = 0 | 100% | 8 | att = 0 | 100% | 4.5 |
| att = 2 | 45% | 3.58 | att = 2 | 45% | 1.96 |
| att = 4 | 22% | 1.73 | att = 4 | 22% | 1.26 |
| att = 6 | 10% | 0.77 | att = 6 | 10% | 0.91 |
| att = 8 | 5% | 0.4 | att = 8 | 5% | 0.39 |
Table II. Transmission percentages for the silicon attenuators used in the XPCS experiments for each material.
Each XPCS measurement was performed on a fresh sample spot to avoid any cumulative radiation effects. Different series of speckle patterns were collected in transmission geometry using an EIGER 4M detector positioned 1.8 m from the sample stage. Fluctuating speckle patterns are obtained when an amorphous system scatters a coherent X-ray, and they reflect the internal dynamics of the sample. Thus, the autocorrelation of the scattered intensity allows for the calculation of the microscopic dynamics, being
This function is called two-times correlation function (TTCF). is the intensity measured at pixel and time , and denotes an average over detector pixels at a given momentum transfer . Examples of TTCFs are presented in Figure 2c-Figure 2d). The main diagonal of the TTCFs corresponds to and represents the correlation of a frame at a given time with itself, while all the off-diagonals are the correlation of a frame at time (or ) with a subsequent frame at time (or ).

Figure 2. Temporal evolution of the dynamics at for a fixed flux of the X-ray beam of ph/s in GeSb a) and of ph/s in GeTe b). The dynamics remains stationary within the probed time window for the GeSb a), while physical aging in the GeTe is signalled by the shift of the curves toward longer times with increasingly b). Two-times correlation functions measured in GeSb c) and GeTe d) showing stable dynamics and aging, respectively.
From , the autocorrelation function of the intensity fluctuations can be extracted:
where , , and represents the average over time. This means that the TTCFs are a map of the time evolution of the microscopic dynamics. The provides information on the microscopic dynamics through the determination of the intermediate scattering function (ISF) via the Siegert relation , where is the experimental contrast related to the degree of coherence of the X-rays [28]. The ISF is the time-dependent density-density correlation function and quantifies the extent to which the atomic configuration at a given moment resembles its configuration at a later time .
Results
We model the with a Kohlrausch-Williams-Watts (KWW) function [29]
where , with being the non-ergodicity factor and the experimental contrast, is the relaxation time, the degree of exponentiality of the decay, and accounts for the baseline [28]. All the parameters depend on the exchanged wavevector , the temperature , and the absorbed dose . The total radiation dose absorbed by the sample during an exposure time is calculated as:
where is the X-ray energy and is the fraction of the incident beam, , absorbed by the sample given by , with being the sample transmission. As anticipated in the experimental section, the value of the incident flux was varied by attenuating the full beam through Si attenuators with transmissions reported in (Table II). The incident fluxes employed in the experiments result in . The dose also depends on the mass within the scattering volume : , where (with as the focal spot area and the sample thickness) and is the mass density.
Consistent with previous studies on other chalcogenide glasses [8, 11], our data show that the probed atomic motion is also triggered by the X-ray irradiation in the studied compositions. This effect is observed at all investigated temperatures and at the probed no appreciable structural damage is detected (see Figure (SM1)). However, within the limits of our experimental resolution, we cannot determine whether this apparent stability reflects a genuine resistance to structural damage, as reported for other glassy systems [4], or whether minor structural damage occurs below our detection threshold or in a different range.
An example of the probed dynamics is shown in Figure 1 where we report auto-correlation functions measured with XPCS at different values of the X-ray flux at , for both (1a) and (1b). The long time-decay of the curves corresponds to the characteristic time associated with the probed dynamics. In both cases, this decay shifts to larger times with decreasing flux (from blue to green curves), indicating that the dynamics can be tuned by varying the intensity of the X-ray beam. The black dashed lines are the best fit to the experimental data according to equation (3). To ensure consistency between data collected at different X-ray fluxes, the comparison of the curves was performed by evaluating the considering a fixed absorbed dose of 38 GGy for each composition.

Figure 1. Auto-correlation functions of the intensity fluctuations measured with XPCS at 0.7 in a) GeSb and b) GeTe at varying beam attenuations. The curves of each sample are obtained by ensuring a total absorbed dose of 38 GGy for both samples. The dashed black lines represent the best fits to the KWW model. Each curve has been measured on a fresh sample position.
Although the dynamics of both glasses clearly depends on the X-ray flux, the two materials exhibit distinct responses to X-ray exposure. This is particularly evident from the temporal evolution of their relaxation processes at a fixed beam intensity as shown in Figure 2. Here, the curves were evaluated by averaging the TTCFs according to Eq. (2) over time windows , centred at waiting time with , and s for the GeSb and s for the GeTe. In this context, the waiting time reflects the time elapsed since the onset of each measurement. As depicted in Figure 2a, curves acquired at different waiting times overlap each other, indicating that the dynamics of the GeSb remains stationary over time. This is also evident from the TTCF, shown in Figure 2c, where a constant intensity profile is observed along the main diagonal from the bottom left corner to the upper right corner. In contrast, GeTe exhibits a progressive slowing down of its dynamics over time. As shown in Figure 2b, the curves shift towards longer decay times as the waiting time progresses from 350 s to 2450 s, highlighting this effect. The presence of this process is also confirmed by the broadening of the intensity along the main diagonal of the TTCF of GeTe reported in Figure 2d.
The slowing down of the dynamics is a spontaneous process in glasses held at temperatures close to , known as physical aging [30]. As out-of-equilibrium systems, glasses naturally evolve over time towards more energetically stable configurations, leading to an increase in relaxation time. Under standard conditions, this aging process is thermally driven, as the system progressively explores its energy landscape. In our experiments, however, the observed slowing down is not solely spontaneous but is induced by the X-ray beam itself. The beam not only probes and stimulates atomic dynamics but also promotes a temporal evolution of the system. This effect may be linked to stress release occurring under irradiation or to local annealing processes initiated by the beam. These aspects will be discussed in greater detail later in this section.
The contrasting temporal evolutions of the dynamics in the two compositions imply a different response of the two systems to the accumulated dose . To investigate this aspect, we express the relaxation time , obtained from XPCS measurements and KWW fits (3), in terms of the absorbed dose required to induce the observed dynamics:
This formulation accounts for the interaction between the material and the X-rays. In the case of a process purely induced by irradiation [4], the relaxation time is expected to scale inversely with the incident X-ray flux , meaning that halving the beam flux () doubles . Consequently, the dynamics depend only on the absorbed dose, independent of flux variations. This expression also allows the comparison of the relaxation dynamics between different samples compositions and experimental parameters. To explore the distinct responses of GeSb and GeTe to X-ray exposure, Figure 3 and Figure 4 show the absorbed dose in relaxation time and the KWW exponent as functions of the accumulated dose , for different attenuator transmissions . This representation enables us to assess the dependence of the dynamical parameters on the absorbed dose and provides insight into the nature of the interaction, allowing us to distinguish between dynamics that are purely beam-induced and those influenced by additional mechanisms.

Figure 3. Evolution of the dynamics in GeSb as a function of the absorbed dose. a-b) At 0.7 and c-d) at 0.8, the and the tend to collapse into a master curve when plotted versus Dose. Both parameters are almost independent of the applied dose.

Figure 4. a-b) Evolution of the dynamics in GeTe as a function of the waiting time , at . and collapse into a master curve when plotted versus , suggesting the dynamics is primarily influenced by intrinsic relaxations active in the material. While the beam does not irradiate the sample spot, both and tend to recover the initial state (insets of a-b)). c-d) At , and collapse in a master curve by plotting them versus the Dose. a-c) As the waiting time or the deposited dose increases, the dynamics slows down, suggesting a final plateau. b-d) With increasing the absorbed dose, the shape parameter decreases from , signature of compressed correlation functions, to , typical of stretched exponential decays.
Let’s first focus on the dynamical behaviour of GeSb. The dependence of the XPCS parameters and on the accumulated dose is presented in Figure 3 at two temperatures, or . Panels a) and c) show the dose accumulated in a relaxation time at the two temperatures, while the dose evolution of the parameter is plotted in panels b) and d). Different symbols and colours refer to different beam fluxes. The probed dynamics is beam-induced, since the parameter is independent of the beam flux for almost all the datasets, implying that the decorrelation time is inversely proportional to the beam flux. Overall, a good collapse of points corresponding to different beam fluxes is observed in all panels. There is, however, one dataset that follows a different trend. The blue circles at , corresponding to full beam on the sample, exhibit a lower and a higher compared to the other points, likely due to local beam-induced heating that slightly increases the effective temperature of the probed spot. Apart from this dataset, the other points suggest that the system is in a yielded state at all the investigated doses, because the parameter is almost constant during the measurement and the stretching parameter is close to one, as in the case of LiBO [12] and GeSe [11].
We now turn to the dynamical properties of GeTe. As discussed before, a purely beam-induced dynamics is expected to scale with the accumulated absorbed dose : if the dynamics is entirely governed by the X-rays, parameters such as the absorbed dose in XPCS relaxation times, , and the shape parameter should collapse onto universal curves when plotted as a function of . However, for GeTe at 0.7, this universal scaling is not observed (see Supplementary Figure (S2)). Although the beam induces the dynamics, and exhibit significant variations with the incident flux when plotted against , suggesting that the absorbed dose alone does not fully describe the system’s evolution.
To further clarify this behaviour, we plot and as functions of the experimental waiting time , defined as the time elapsed since the start of exposure. Interestingly, when expressed in terms of , the data collapse onto master curves across all fluxes (Figure 4a-Figure 4b)), revealing a universal, time-dependent evolution of the dynamics. To our knowledge, such a collapse with respect to the waiting time, rather than the absorbed dose, has not been previously reported in studies of beam-induced dynamics. This observation indicates that, while X-ray irradiation initiates atomic motion, the evolution at remains governed primarily by intrinsic relaxation processes with timescales dictated by the material itself, rather than solely by the accumulated dose. This phenomenon can be rationalised by considering that for GeTe corresponds to room temperature. It is therefore highly likely that the sample retains residual stresses from its preparation, which are progressively released under X-ray exposure. The stress relaxation drives the evolution of the dynamics, dominating the beam-induced effect.
Further evidence supporting this interpretation comes from “beam-of” experiments (Figure 4 insets), where the X-ray beam was switched off for extended periods (see SM and Figure S3 for further details). For short interruptions (10 minutes), stress relaxation continues, reflecting intrinsic structural relaxation mechanisms triggered by the X-rays. Differently, for longer interruptions (1 hour), both and reverted toward their initial values, indicating a recovery of the original state in the absence of irradiation. Such a reversible response upon switching off the beam has never been observed in non-stationary beam-induced dynamics so far, and supports the absence of structural damage within the experimental doses.

Figure S3. Dynamics evolution at of GeTe switching off the beam for 600 s and for 3600 s. a) The relaxation time increases in time, but while the beam is off it tends to go back to the initial values. b) the shape parameter decreases in time, but while the beam is off it tends to go back to the initial values.
This hypothesis of room temperature stress-release is confirmed by the fact that at , both and exhibit a universal scaling with the absorbed dose , regardless of the incident flux (Figure 4c-Figure 4d)). This indicates that, at , the dynamics depends solely on the X-ray flux and is thus purely beam-induced. Remarkably, a modest 10% increase in temperature is sufficient to eliminate the stress-driven contributions observed at , placing the system entirely under the control of irradiation-driven dynamics.
Next, we examine how the dynamical parameters of the two compositions depend on the absorbed dose (or on the waiting time ) at both and . To do that, we refer to Figure 3 and (4) for and , respectively. As anticipated, the glass exhibits stationary dynamics and a KWW exponent at both investigated temperatures. These features correspond to a local state of photo-induced yielding by the X-rays, resembling the previous studies on other chalcogenide and oxide glasses [8, 11, 12]. In materials science, yielding refers to the transition from an elastic response, where deformation is reversible, to a plastic state, where the material undergoes irreversible atomic rearrangements. In glasses, yielding is particularly subtle since plastic events appear at very small scales and are associated with localised atomic rearrangements known as shear transformation zones [12]. Martinelli et. al. [12] showed that X-ray irradiation acts like an external force, generating tiny defects that trigger these hidden structural rearrangements, effectively pushing the material into a plastic, irreversible state at the microscopic scale. In the case of the chalcogenide glass , the yielding is induced within 100 s of exposure at ambient temperature [11], while the composition reaches this stationary state within the exposure time of 1 s at the probed temperatures of 350 K and 400 K ( and , respectively). This difference in time scales is likely the consequence of the larger thermal fluctuations at the probed temperatures, which facilitate the occurrence of local yielding at a lower dose. In fact, in agreement with previous works, we also observe a transitory state with a weak aging at ambient temperature () in accompanied by a transition in the KWW parameter from compressed () to stretched () and the reaching of a stationary regime around 40 GGy (see S4). As discussed in detail in Ref. [12], these are the typical characteristics of the induced yielding transition.
On the other hand, the exhibits a strong dependence on the waiting time and absorbed dose at both temperatures, as evident from Figure 4. We see that neither the absorbed dose associated with the relaxation time nor the parameter reaches a complete plateau. Only the measured at and the at at suggest the proximity to a stationary regime. To understand whether this behaviour corresponds to the onset of the photo-induced yielding transition, we measured the wave-vector dependence of the dynamics (and thus of the parameters and ) as a function of increasing absorbed dose (see Figure 5). Following Ref. [12], during the transition to local yielding, one should also observe a crossover from a wave-vector dependence of and an almost q-independent evolution of , to a liquid-like behaviour.

Figure 5. Wave vector dependence of a) and b) at of GeTe for three selected doses: GGy (blue circles), GGy (red squares) and GGy (yellow triangles). The vertical dashed and dashed-dotted black lines represent, respectively, the position of the pre-peak and the main peak. a) The inset shows the dependence on the dose of the power-law exponent describing the linear dependence of at low -values.
At low , a linear trend emerges in the log-log plot of versus (Figure 5a), indicating a power-law relationship between the relaxation time and the wavevector. By analysing this dependence in the range [2, 3, 4, 5, 6, 7] nm, we can access the nature of the particle dynamics and how it evolves with absorbed dose. Fitting the data with the expression , where the exponent characterizes the type of motion, we obtain , independently of dose (see inset of Figure 5a). This value is larger than the one reported for LiBO glass, where decreases from ∼ 1.1 to ∼ 0.3 with increasing dose. This trend was interpreted as a transition in the nature of collective motion, from ballistic-like dynamics to a regime where defect creation and annihilation are balanced, similar to a steady-state condition observed in liquids [12]. Differently, in the GeTe, the larger value of suggests the presence of diffusive motion at low -s, with a negligible dose dependence. Additional data at lower values, well separated from structural contributions, are required to confirm this behaviour.
At high (greater than 7 nm), we observe an overall increase in and a concurrent decrease in with increasing dose . At all doses, presents a peak at ∼ 10.5 nm, which mirrors the corresponding prepeak of the static structure factor [21]. At high doses, a peak arises at ∼ 19.6 nm, corresponding to the main first sharp diffraction peak (FSDP). Even the KWW parameter exhibits features resembling the evolution of the , less evident at low doses but pronounced at higher doses. The resemblance between the -dependence of and , and that of the is reminiscent of the phenomenon known as de Gennes narrowing, which takes the name from the narrowing at -values corresponding to the FSDP of the width of the quasi-elastic contribution in neutron scattering experiments on liquids [31]. Given that the half-width at half maximum of the quasi-elastic peak is inversely proportional to structural relaxation time, , this results in an increase in near structure factor peaks, implying that the most probable structural configuration is also the most stable. Therefore, the observed -dependence in GeTe suggests that the sample volume irradiated by the X-rays displays a liquid-like behaviour and thus a local yielding as in the LiBO. This interpretation is supported by the transition of from values greater than 1, typical of a stress-driven dynamics of glasses well below [32], to values lower than 1, fingerprint of liquid-like heterogeneous relaxation dynamics [33, 34].
We note also that at low doses (below 0.5 GGy), the de Gennes narrowing is entirely absent for LiBO, in both and [12]. In our case, instead, the de Gennes narrowing effect is already present at the 10.5 nm pre-peak in (), even in the lowest probed dose regime (below 7 GGy). In particular, at low doses only the signature of the pre-peak around 10.5 nm is observed in the evolution of with , while at high doses, the signature of the main peak around 19.6 nm arises as well. Such a strong dynamical signature corresponding to the weak pre-peaks in structure factors is a general feature of glass-forming liquids, which has been associated with the influence of dynamic heterogeneity phenomena at the single and many particle level [35, 36].
Discussion and conclusion
Our results reveal a strong correlation between the local atomic structure and the dynamical response to intense X-ray fluxes in two compositionally similar glasses, GeSb and GeTe. Consistent with previous findings in other oxides and chalcogenides [8, 11, 12], the X-rays can induce a local state of yielding in both compositions, where the amorphous solids present liquid-like features in the irradiated zone. The propensity of this phenomenon, however, strongly depends on the chemical and physical details of the material.
In the PCM GeSb, our data exhibit an almost instantaneous photo-induced yielding transition at 0.7 and 0.8. This rapid response is favoured by thermal motion, since at room temperature the transition requires 400 s of continuous irradiation to occur (see SM). This is in agreement with what has been observed in GeSe [11] at ambient temperature, where yielding occurs in 100 s. The faster transition at room temperature in GeSe is likely due to the Se being lighter than Sb, a fact supported by a faster relaxation time of s in GeSe in comparison to s in GeSb for similar doses. Increasing the temperature to 350 K (0.7) enhances atomic mobility within GeSb, enabling the system to reach the yielding state almost immediately.
Unlike GeSb and GeSe, GeTe does not completely reach the yielding state within the experimental timeframe, neither at 0.7 (which corresponds to room temperature) nor 0.8. This can be explained by the fact that Te atoms are heavier than Se or Sb atoms, reducing atomic mobility and resulting in a greater resistance to structural change upon irradiation with X-ray beams. This is consistent with the absence of visible structural changes on the FSDP, which are instead clear in GeSe. Moreover, GeTe exhibits recovery effects upon beam switching off: once the X-ray beam is turned off, the system begins to recover toward its initial dynamical state and the longer the absence of irradiation, the closer the dynamical parameters return to their original values (see insets of Figure 4a-Figure 4b)). In contrast, the GeSe, once it has reached the yielding state, shows a stable dynamics: during the absence of the beam, the system remains in that state. It is important to emphasise that these differing responses cannot be explained solely by considering the type of chemical bonding. Both GeSe and GeTe are covalent glasses with good glass-forming ability and slow crystallisation kinetics. Nevertheless, they show distinct X-ray-induced dynamics, suggesting that factors beyond bonding, such as atomic mass, network topology, and local structural heterogeneity, play a crucial role in determining the onset and nature of yielding under irradiation.
The faster photo-induced atomic motion in the PCM GeSb and the higher sensitivity to the X-rays induced yielding is likely the consequence of its locally loosely bonded microscopic structure and the presence of a pronounced secondary -relaxation which promotes local mobility at the atomic scale [27]. This mobility is also confirmed by the fact that the dynamics does not scale completely with the dose , strengthening the idea of the presence of an intrinsic dynamic process already active at the probed temperatures. By contrast, the rigid covalently bonded network in is less prone to changes of the local atomic environment, which reflects in longer relaxation times at the atomic scale and a pure beam-induced dynamics but at ambient temperature. At this temperature, the scaling law is not fulfilled, probably due to the presence of some additional stresses related to the sample preparation, which sum up with the irradiation-induced yielding transition.
Interestingly, we note that the exhibits a dependence on the absorbed dose similar to that reported in the oxide glass [12]. In both systems, increasing the absorbed dose leads to a transition in the KWW shape parameter from a compressed () to a stretched () regime, which is accompanied by the emergence of a liquid-like wave-vector dependence of the dynamics at large doses. As previously discussed, this behaviour can be attributed to the transition from a purely beam-induced stress-driven motion to a local yielding transition of the material under X-ray irradiation [12].
It is, however, important to underline the main differences between and . One crucial factor influencing beam-induced dynamics is the X-ray absorption. consists of heavy atoms, which absorb X-rays more efficiently than the lighter elements present in . This difference is reflected in the experimental conditions: the thickness of the sample was comparable to its attenuation length at 8 keV (), ensuring substantial absorption, whereas was significantly thinner (125 ) relative to its attenuation length at 8.4 keV (680 ). Consequently, the dose range explored for reached tens of GGy, while for , it remained at a few GGy. Despite this large difference in the absorbed dose, both materials exhibit a similar evolution of the KWW parameter and as a function of the absorbed dose. This suggests that atomic rearrangements in require a significantly higher number of electron-hole pairs, likely due to its heavier atomic composition. This interpretation aligns with the slower dynamics observed in (on the order of hundreds of seconds) compared to (), despite its higher absorbed dose.
In conclusion, our XPCS study reveals distinct beam induced dynamical responses in glassy and near their glass transition temperatures. Both systems are influenced by X-ray irradiation, yet their responses differ markedly despite their similar composition. undergoes a rapid transition to a stationary yielding regime within the exposure time, characterized by a KWW parameter . By contrast, exhibits strong dose-dependent evolution, including aging and a progressive shift of the shape parameter from compressed () to stretched (). This transition is consistent with the emerging of liquid-like collective motion, as supported by de Gennes narrowing observed in the wavevector -dependence of the dynamics. This glass does not reach a stationary regime within experimental timescales, indicating that local yielding requires thousands of seconds. This resistance to photoinduced yielding in is consistent with its heavier atomic constituents, which enhance structural rigidity and demand much higher deposited doses to induce local transformations. These findings align with its higher amorphous thermal stability and slow crystallisation kinetics [22], in contrast to which exhibits rapid crystallisation kinetics and less covalent bonding [16]. Overall, neither chemical composition alone nor bonding character alone dictates the X-ray induced response of the non-PCM and PCM ; instead, it emerges from the interplay between irradiation effects and intrinsic relaxation pathways. Finally, we note that such beam-induced dynamics may provide a novel means of probing -relaxations in PCMs below . Previous studies have shown that PCMs such as , GeTe, and AIST exhibit pronounced excess wings in the loss modulus associated with -relaxations, in stark contrast to those chalcogenide non-PCMs whose excess wings are vanishingly small [37].
Acknowledgements
We gratefully acknowledge PETRA III synchrotron (Hamburg, Germany) for providing beamtime, including experiments carried out at P10 under the LTP project II-20230020 EC. This project received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (Grant Agreement No 948780).
Author Contributions
G.B. and B.R. conceived and designed the project. G.B., A.C., I.F., T.F., J.M., A.R., B.R., M.S., S.W., and F.W. performed the experiments. J.M., R.E., A.P. and R.E. carried out sample fabrication. I.F. analysed the data. G.B., B.R. and I.F. wrote the manuscript with contributions from all the authors. All the authors contributed to the discussion of the results, reviewed the data and provided feedback on the manuscript.
Supplementary information to: X-ray photo-induced atomic motion in Phase Change Materials and conventional covalent chalcogenide glasses
I. Festi; A. Cornet; T. Fujita; J. Moesggard; A. Ronca; J. Shen; M. Sprung; S. Wei; F. Westermeier; R. Escalier; A. Piarristeguy; G. Baldi; B. Ruta
Methods
Absorbed dose calculation
The dose rate of X-ray absorption from a material is
where and are respectively the energy and the flux of the X-rays, is the electron charge, and is the transmission of the Si attenuators employed to attenuate the beam intensity. The dose rate also depends on the absorbed fraction of the beam
where is the sample thickness, while is the absorption length of the material at the energy . Finally, the dose rate depends on the mass within the scattering volume , according to , being the mass density of the sample. The volume is found considering an elliptical beam cross section, whose semiaxes correspond to the FWHM/ values.
The absorbed dose is found considering the exposure amount of time of the sample spot to the beam
Data treatment of and vs
The -dependence of the dynamics of shown in Figure 5 of the main article was measured at with attenuator 2. Due to the lower statistics at high -values, we applied a two-point averaging to the parameter to improve data clarity and reduce statistical fluctuations. This averaging was not applied to the relaxation time, as it is a significantly more robust parameter and did not require smoothing.
Beam damage
The scattered intensity from the samples showed no significant variation over the duration of the XPCS measurements (Figure S1), suggesting that either the underlying structure remained largely unchanged by the X-ray beam, indicating no appreciable structural damage, or that any structural modifications occurring were not detectable through the analysis of the mean scattered intensity alone. This observation implies that the observed dynamical changes are likely not a direct consequence of gross structural degradation of the amorphous network within the probed timeframe.

Figure S1. The average scattered intensity collected by the detector for each frame, normalized to the relative beam intensity at various attenuation levels. As time passes, no appreciable structural damage is occurring.
versus absorbed dose
At , if we plot and versus the absorbed dose we observe a failure in the collapse into a universal master curve, differently from what was observed at .

Figure S2. Dynamics evolution at of GeTe. a) and b) fail to collapse into a master curve when plotted versus .
Recovery of dynamics after beam interruption
To further investigate the nature of the beam-induced dynamics in GeTe, we conducted experiments where the X-ray beam was switched off after an initial measurement on a single sample spot, and then turned back on for a subsequent measurement on the same location. Our findings revealed that during the period when the beam was off, the dynamics of the material tended to revert towards its initial state (Figure S3).
GeSb
T=0.6T
While GeSb generally exhibits stationary dynamics at 0.7T and 0.8T, measurements at 0.6T reveal some hints of aging, specifically at the highest beam intensities. At lower fluxes, the relaxation time remains stable. At 0.6T, we also observe a decrease in the stretching exponent from approximately 1.1 to 0.9, a trend reminiscent of GeTe but significantly less pronounced (where shifts from to ). This transient behaviour in GeSb at room temperature () is likely due to residual stresses introduced during sample preparation. These stresses may make the material more susceptible to the X-ray beam, a phenomenon far more pronounced in GeTe, before dissipating at higher temperatures where the intrinsic stability of the glass dominates.

Figure 6. Dynamics evolution at T = of GeSb. a) the relaxation time shows a constant behaviour, except for very high beam intensities. b) decreases from to .
References
- [1]José Baruchel, Jean-Louis Hodeau, and Mogens S. Lehmann. Neutron and synchrotron radiation for condensed matter studies. Volume I, Theory, instruments and methods. Springer, 1993.
- [2]Anders Madsen, Andrei Fluераsu, and Beatrice Ruta. Structural dynamics of materials probed by x-ray photon correlation spectroscopy. In Eberhard Jaeschke, Shaukat Khan, Jochen R. Schneider, and Jerome B. Hastings, editors, Synchrotron Light Sources and Free-Electron Lasers: Accelerator Physics, Instrumentation and Science Applications, pages 1–21. Springer International Publishing, 2015.
- [3]Michael Leitner, Markus Stana, Manuel Ross, and Bogdan Sepiol. Acceleration of atomic dynamics due to localized energy depositions under x-ray irradiation, arXiv, arXiv:1510.01918 2015.arxiv.org/abs/1510.01918
- [4]B. Ruta, F. Zontone, Y. Chushkin, G. Baldi, G. Pintori, G. Monaco, B. Rufflé, and W. Kob. Hard x-rays as pump and probe of atomic motion in oxide glasses. Scientific Reports, 7(1):3962, Jun 2017.DOI
- [5]G. Pintori, G. Baldi, B. Ruta, and G. Monaco. Relaxation dynamics induced in glasses by absorption of hard x-ray photons. Phys. Rev. B, 99:224206, Jun 2019.DOI
- [6]Dallari, Pintori, Baldi, Martinelli, Ruta, Sprung, and Monaco. X-rays induced atomic dynamics in a lithium borate glass. Condensed Matter Physics, 22(4):43606, November 2019.
- [7]Katharina Holzweber, Christoph Tietz, Tobias Michael Fritz, Bogdan Sepiol, and Michael Leitner. Beam-induced atomic motion in alkali borate glasses. Phys. Rev. B, 100:214305, Dec 2019.
- [8]Jianheng Li, Meera Madhavi, Spencer Jeppson, Louie Zhong, Eric M. Dufresne, Bruce Aitken, Sabyasachi Sen, and Roopali Kukreja. Observation of collective molecular dynamics in a chalcogenide glass: Results from x-ray photon correlation spectroscopy. The Journal of Physical Chemistry B, 126(28):5320–5325, 2022. PMID: 35730616.DOI
- [9]Francesco Dallari, Alessandro Martinelli, Federico Caporaletti, Michael Sprung, Giacomo Baldi, and Giulio Monaco. Stochastic atomic acceleration during the x ray-induced fluidization of a silica glass. Proceedings of the National Academy of Sciences, 120(2):e2213182120, 2023.
- [10]E. Alfinelli, F. Caporaletti, F. Dallari, A. Martinelli, G. Monaco, B. Ruta, M. Sprung, M. Zanatta, and G. Baldi. Amorphous-amorphous transformation induced in glasses by intense x-ray beams. Phys. Rev. B, 107:054202, Feb 2023.DOI
- [11]Jacopo Baglioni, Alessandro Martinelli, Peihao Sun, Francesco Dallari, Fabian Westermeier, Michael Sprung, Gerhard Grübel, and Giulio Monaco. Uniqueness of glasses prepared via x-ray induced yielding. Reports on Progress in Physics, 87(12):120503, nov 2024.
- [12]Alessandro Martinelli, Federico Caporaletti, Francesco Dallari, Michael Sprung, Fabian Westermeier, Giacomo Baldi, and Giulio Monaco. Reaching the yield point of a glass during x-ray irradiation. Phys. Rev. X, 13:041031, Nov 2023.DOI
- [13]Shuai Wei, Pierre Lucas, and Charles Angell. Phase change materials: The view from the liquid phase and the metallicity parameter. MRS Bulletin, 44:691–698, 09 2019.DOI
- [14]Piarristeguy, Andrea, Noé, Pierre, and Hippert, Françoise. Verres de chalcogénures pour le stockage de l’information. Reflets phys., (74):58–63, 2022.DOI
- [15]Shuai Wei, Pierre Lucas, and C. Austen Angell. Phase change materials: The view from the liquid phase and the metallicity parameter. MRS Bulletin, 44(9):691–698, 2019.DOI
- [16]Jens Moesgaard, Tomoki Fujita, and Shuai Wei. Unveiling the boson peaks in amorphous phase-change materials. Journal of Physics: Condensed Matter, 37(2):025101, oct 2024.arxiv.org/abs/2404.17226
- [17]Jean Yves Raty, Wei Zhang, Jennifer Luckas, Chao Chen, Riccardo Mazzarello, Christophe Bichara, and Matthias Wuttig. Aging mechanisms in amorphous phase-change materials. Nature Communications, 6(1):7467, Jun 2015.DOI
- [18]M. Chen, K. A. Rubin, and R. W. Barton. Compound materials for reversible, phase-change optical data storage. Applied Physics Letters, 49(9):502–504, 09 1986.DOI
- [19]Shuai Wei, Pierre Lucas, and Charles Angell. Phase change alloy viscosities down to T₍g₎ using Adam-Gibbs equation fittings to excess entropy data: A fragile-to-strong transition. Journal of Applied Physics, 118:034903, 07 2015.DOI
- [20]Hans Weber, Jiri Orava, Ivan Kaban, Julian Pries, and A. Lindsay Greer. Correlating ultrafast calorimetry, viscosity, and structural measurements in liquid GeTe and Ge₁₅Te₈₅. Phys. Rev. Mater., 2:093405, Sep 2018.
- [21]Shuai Wei, Moritz Stolpe, Oliver Gross, William Hembree, Simon Hechler, Jozef Bednarčik, Ralf Busch, and Pierre Lucas. Structural evolution on medium-range order during the fragile-strong transition in Ge₁₅Te₈₅. Acta Materialia, 129:259–267, 2017.DOI
- [22]Rathinavelu Sengottaiyan, Nishant Saxena, Krishna Dayal Shukla, and Anbarasu Manivannan. Temperature dependent structural evolution and crystallization properties of thin Ge₁₅Te₈₅ film revealed by in situ resistance, x-ray diffraction and scanning electron microscopic studies. Journal of Physics D: Applied Physics, 53(2):025108, oct 2019.
- [23]Pierre Lucas, Shuai Wei, and Charles Angell. Liquid liquid phase transitions in glass-forming systems and their implications for memory technology. International Journal of Applied Glass Science, 11, 02 2020.DOI
- [24]Julian Pries, Shuai Wei, Felix Hoff, Pierre Lucas, and Matthias Wuttig. Control of effective cooling rate upon magnetron sputter deposition of glassy Ge₁₅Te₈₅. Scripta Materialia, 178:223–226, 2020.
- [25]Julian Pries, Yuan Yu, Peter Kerres, Maria Häser, Simon Steinberg, Fabian Gladisch, Shuai Wei, Pierre Lucas, and Matthias Wuttig. Approaching the glass transition temperature of GeTe by crystallizing Ge₁₅Te₈₅. physica status solidi (RRL) – Rapid Research Letters, 15(3):2000478, 2021.
- [26]Tomoki Fujita, Yuhan Chen, Yoshio Kono, Seiya Takahashi, Hidetaka Kasai, Davide Campi, Marco Bernasconi, Koji Ohara, Hirokatsu Yumoto, Takahisa Koyama, Hiroshi Yamazaki, Yasunori Senba, Haruhiko Ohashi, Ichiro Inoue, Yujiro Hayashi, Makina Yabashi, Eiji Nishibori, Riccardo Mazzarello, and Shuai Wei. Pressure induced reversal of Peierls-like distortions elicits the polymorphic transition in GeTe and GeSe. Nature Communications, 14, 12 2023.DOI
- [27]Yudong Cheng, Qun Yang, Jiangjing Wang, Theodoros Dimitriadis, Matthias Schumacher, Huiru Zhang, Maximilian J. Müller, Narges Amini, Fan Yang, Alexander Schoekel, Julian Pries, Riccardo Mazzarello, Matthias Wuttig, Hai-Bin Yu, and Shuai Wei. Highly tunable β-relaxation enables the tailoring of crystallization in phase-change materials. Nature Communications, 13(1):7352, Nov 2022.DOI
- [28]Gerhard Grübel and Federico Zontone. Correlation spectroscopy with coherent x-rays. Journal of Alloys and Compounds, 362(1):3–11, 2004. Proceedings of the Sixth International School and Symposium on Synchrotron Radiation in Natural Science (ISSRNS).
- [29]J C Phillips. Stretched exponential relaxation in molecular and electronic glasses. Reports on Progress in Physics, 59(9):1133, sep 1996.DOI
- [30]Zach Evenson, Beatrice Ruta, Simon Hechler, M. Stolpe, Eloi Pineda, Isabella Gallino, and Ralf Busch. X-ray photon correlation spectroscopy reveals intermittent aging dynamics in a metallic glass. Physical Review Letters, 115:175701, 10 2015.DOI
- [31]P.G. De Gennes. Liquid dynamics and inelastic scattering of neutrons. Physica, 25(7):825–839, 1959.DOI
- [32]Maximilian Frey, Nico Neuber, Sascha Riegler, Antoine Cornet, Yuriy Chushkin, Federico Zontone, Lucas Ruschel, Bastian Adam, Mehran Nabahat, Fan Yang, Jie Shen, Fabian Westermeier, Michael Sprung, Daniele Cangialosi, Valerio Di Lisio, Isabella Gallino, Ralf Busch, Beatrice Ruta, and Eloi Pineda. Liquid-like versus stress-driven dynamics in a metallic glass former observed by temperature scanning x-ray photon correlation spectroscopy. Nature Communications, 16, 05 2025.
- [33]B. Ruta, S. Hechler, N. Neuber, D. Orsi, L. Cristofolini, O. Gross, B. Bochtler, M. Frey, A. Kuball, S. S. Riegler, M. Stolpe, Z. Evenson, C. Gutt, F. Westermeier, R. Busch, and I. Gallino. Wave-vector dependence of the dynamics in supercooled metallic liquids. Phys. Rev. Lett., 125:055701, Jul 2020.
- [34]Nico Neuber, Oliver Gross, Maximilian Frey, Benedikt Bochtler, Alexander Kuball, Simon Hechler, Fan Yang, Eloi Pineda, Fabian Westermeier, Michael Sprung, Isabella Gallino, Ralf Busch, and Beatrice Ruta. Disentangling structural and kinetic components of the α-relaxation in supercooled metallic liquids, Communications Physics, 5:316, 2022.
- [35]Vladimir Novikov, K Schweizer, and Alexei Sokolov. Coherent neutron scattering and collective dynamics on mesoscale. The Journal of chemical physics, 138:164508, 04 2013.
- [36]Beatrice Ruta, Giacomo Baldi, Y Chushkin, Benoit Rufflé, Luigi Cristofolini, A Fontana, Marco Zanatta, and Francesco Nazzani. Revealing the fast atomic motion of network glasses. Nature communications, 5:3939, 05 2014.
- [37]Si-Xu Peng, Yudong Cheng, Julian Pries, Shuai Wei, Hai-Bin Yu, and Matthias Wuttig. Uncovering β-relaxations in amorphous phase-change materials. Science Advances, 6(2):eaay6726, 2020.DOI