Abstract
Tailoring the degree of structural disorder in Ge-Sb-Te alloys is important for the development of non-volatile phase-change memory and neuro-inspired computing. Upon crystallization from the amorphous phase, these alloys form a cubic rocksalt-like structure with a high content of intrinsic vacancies. Further thermal annealing results in a gradual structural transition towards a layered structure and an insulator-to-metal transition. In this work, we elucidate the atomic-level details of the structural transition in crystalline GeSb2Te4 by in situ high-resolution transmission electron microscopy experiments and ab initio density functional theory calculations, providing a comprehensive real-time and real-space view of the vacancy ordering process. We also discuss the impact of vacancy ordering on altering the electronic and optical properties of GeSb2Te4, which is relevant to multilevel storage applications. The phase evolution paths in Ge-Sb-Te alloys and Sb2Te3 are illustrated using a summary diagram, which serves as a guide for designing phase-change memory devices.
Keywords: Phase-change materials, Vacancy ordering, Structural transition, Metavalent bonding, in situ TEM
Graphical abstract
1. Introduction
Chalcogenide phase-change materials (PCMs) are widely exploited for non-volatile memory (NVM) and in-memory computing (IMC) applications [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]. The basic principle is to utilize their capability of rapid phase transitions between amorphous and crystalline states, as well as the associated change in electrical resistance or optical reflectivity, to encode information. Ge-Sb-Te alloys along the GeTe-Sb2Te3 pseudo-binary line are the most widely investigated and used PCMs [11]. The atomic structure and defects of Ge-Sb-Te alloys have been extensively studied by X-ray diffraction (XRD) [11], [12], [13], transmission electron microscopy (TEM) experiments [14], [15], [16], [17], [18], [19], [20] and density functional theory (DFT) calculations [21], [22], [23], [24], [25], [26], [27]. Upon rapid crystallization of the amorphous (amor-) phase in devices [14], Ge-Sb-Te alloys form a metastable cubic rocksalt-like (cub-) crystalline phase with the anion-like sublattice occupied by Te atoms and the cation-like sublattice occupied by Ge, Sb and atomic vacancies. Further thermal annealing drives a gradual phase transformation from the cub-phase towards a stable trigonal phase (also termed as hexagonal, hex‑, in literature) with alternately stacked Ge-Sb-Te blocks and van der Waals (vdW)-like gaps.
An unconventionally high amount of atomic vacancies, e.g. 12.5% in GeSb2Te4 (abbreviated as “GST” in the following), is found in the cub-phase. These vacancies reduce the unfavorable antibonding interactions that would be present for a non-stoichiometric phase with fully occupied Ge/Sb sublattice [28]. The resulting electronic configuration yields on average 3 p electrons per site in the cubic rocksalt lattice [29]. Since the atomic arrangement is octahedral-like, the σ–bonds formed between adjacent atoms are only occupied by one electron (half an electron pair). This unconventional bonding configuration differs significantly from ordinary covalent bonding. It is better described as a distinct bonding mechanism, recently coined as metavalent bonding (MVB), which is characterized by the competition between electron delocalization as in metallic bonding and electron localization as in ionic and covalent bonding [30], [31], [32], [33], [34]. Vacancies can have a profound impact on this competition. The presence of atomic vacancies was unambiguously demonstrated by element-resolved atomic imaging experiments [18]. The statistical distribution of the intrinsic atomic vacancies induces Anderson-like localization of electrons at the Fermi level. Ordering of these vacancies drives a cub-to‑hex structural transition and an insulator-to-metal transition [35], [36], [37].
The vacancy ordering process was first predicted on the basis of DFT calculations [36], and then observed in ex situ TEM experiments upon thermal annealing [18], [19], [20]. In situ heating TEM experiments were also reported in the literature [38], [39], [40], [41]. However, the microscopic details of the vacancy migration and ordering are still lacking. In this work, we carry out in situ high-resolution TEM experiments in combination with electron beam irradiation, which enable a direct observation of the structural transition process locally. Together with insight provided via DFT calculations, our experimental analysis offers a comprehensive atomic picture of the cub-to‑hex structural transitions in GST.
2. Experimental and simulation methods
The amor-GeSb2Te4 thin films of ∼80 nm thickness were prepared on carbon-supported TEM grids via magnetron sputtering deposition from stoichiometric targets under 20 W power with a base pressure of 2 × 10−6 mbar and an argon gas flow of 20 sccm. Then the films were covered by an electron-transparent ZnS-SiO2 layer. The as-deposited films were annealed in a tube furnace under constant argon-flow of 1200 sccm at 150 °C for 1 hour, forming the cub-phase. The in situ electron beam irradiation experiments were performed for the cubic samples in a JEOL JEM-2100F TEM operated under 200 keV and in a JEM-200 CX TEM under 120 keV. For ex situ experiments, an amor-GeSb2Te4 thin film of ∼450 nm-thick was deposited with magnetron sputtering on a silicon substrate, and was annealed in a tube furnace with an argon flow rate of 1200 sccm at different temperatures, namely, 175, 225, 250 and 350 °C, respectively, for 1 hour. After each annealing step, the thin film was used to prepare TEM sample specimens (∼80 nm thickness) using a Helios Nano 600 dual-beam FIB system (FEI, Hillsboro, OR, USA) under 30 keV with beam current of Ga ions sequentially decreasing from 6.5 nA for coarse cutting to 28 pA for fine thinning. The energy dispersive X-ray (EDX) spectroscopy analysis confirms no Ga signal in the sample. The microstructures of all samples were characterized by bright-field and high-resolution TEM (HRTEM) imaging, selected area electron diffraction (SAED) analysis and fast Fourier transform (FFT) of HRTEM images in JEOL JEM-2100F TEM. Aberration corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) experiments were conducted at 200 keV on an JEOL-ARM200F TEM equipped with an EDX system.
DFT calculations were carried out for GeSb2Te4 models of different size. Large models with 1008 atoms were built for electron localization calculations, which were performed using the CP2K package [42]. Goedecker pseudopotentials [43] and the Perdew–Burke–Ernzerhof (PBE) functional [44] were used, and the Brillouin zone was sampled at the Γ point. A triple-zeta plus polarization Gaussian-type basis set was employed to expand the Kohn–Sham orbitals, and plane waves with a cutoff of 300 Ry were used to expand the charge density. The VASP code [45] was employed to calculate the optical properties of smaller models. The PBE functional [44] and projector augmented-wave (PAW) pseudopotentials [46] were used with an energy cutoff value of 500 eV. We considered a unit cell model for the hex‑phase, and smaller supercell models for the cub-phase. The atomic structure of the amorphous phase was taken from our previous work [47]. The respective atomic models are shown in Fig. S1. Dense k-point meshes were used for the optical calculations, i.e., 16 × 16 × 1, 3 × 3 × 1 and 3 × 3 × 3 for the hex‑, cub- and amorphous phase, respectively.
3. Results and discussion
3.1. In situ TEM characterization
Firstly, we carried out structural characterization experiments using a JEOL JEM-2100F TEM with a moderate accelerating voltage Va of 200 keV, and confirmed that the sample was in the cub-phase. Within a limited recording time of seconds, no obvious damage to the specimen was detected. Permanent structural changes become evident by focusing the electron beam to small exposure areas and extending the irradiation time to tens of minutes, as shown in Fig. 1. The focused electron beams have multiple effects, including specimen heating, knock-on collisions and radiolysis effects. It is generally agreed that the kinetic knock-on effects get stronger at higher accelerating voltages, while the heating effects become dominant at lower accelerating voltages [48]. For in situ electron beam irradiation experiments with properly tuned size of focus areas, including line scanning, operated under 200–300 keV, the formation and motion of swapped bilayer defects [49], [50], [51] and vacancy layers [52, 53], vacancy-disordering induced structural transition [54, 55], and even non-thermal amorphization [56] were observed in Ge-Sb-Te and Sb2Te3 alloys and superlattices.
Fig. 1.
Structural phase transition in crystalline GST under electron beam irradiation. (a) Sketch of the effects of electron beams in inducing reversible structural transition between the cub- and hex‑phase of GST. (b) Bright-field TEM images and the FFT pattern corresponding to the area marked by the green box recorded at different irradiation time using the JEOL JEM-2100F TEM operated at 200 keV. (c) Bright-field TEM images and corresponding SAED patterns recorded at different irradiation time using the JEM-200CX TEM operated at 120 keV.
Using the JEOL JEM-2100F TEM operated at 200 keV, we managed to capture the cub-to‑hex transition process within 70 min by tuning the beam radius for irradiating an area of slightly larger than 200 nm, corresponding to a beam intensity of ∼1.0 × 1024 e m − 2 s − 1. We identified a cub-phase grain with the [011]c crystallographic orientation within the irradiated area, presented in Fig. 1b. In such view direction the cation-like and anion-like sublattices of the cubic rocksalt structure are well separated, and thus it is very suitable to observe vacancy ordering [18]. After extensive irradiation, planar stripes were observed in the bright-field images. The Fast Fourier Transform (FFT) analysis (insets) of the areas marked by green square shows a clear change in diffraction pattern from the cub- to hex‑phase. To enhance the heating effects, we repeated the irradiation experiments on the same thin film by using a JEM-200CX TEM operated at 120 keV. As shown in Fig. 1c, the cub-to‑hex transition took place faster and the majority of the irradiated area was transformed within ∼10 min, as evidenced by the corresponding SAED patterns. The typical size of the irradiation-induced hex‑phase grains is around a few hundreds of nanometers. Further irradiation will not induce visible changes inside the grains.
Next, we assess the detailed transition path of the structural transition process. Fig. 2 shows a local area within the green box presented in Fig. 1b with higher magnification. The initial phase looked homogenous, while after 15 min irradiation, some dark patches appeared in the {111}c planes, as marked by red arrows, which result in the reflection in the FFT pattern, ()c corresponding to (0003)h, due to vacancy ordering. The intersection angle between the two dark patches is measured to ∼71° in the HRTEM image. This observation of intersected dark patches with an angle of ∼71° is consistent with the atomic-scale imaging experiments done on a Ge2Sb2Te5 thin film annealed at 180 °C using HAADF-STEM [18]. As indicated by red arrows in the FFT pattern, these vacancy-ordered planes along the two different crystallographic planes are ()c and ()c planes, both of which are equivalent to the (111)c planes. More dark patches along the two directions were observed after 35 min irradiation. Upon further irradiation to 49 min, vacancy ordering in the ()c planes became more evident but vanished in the ()c planes. Later on, the atomic blocks were shifted locally when the concentration of vacancies reached a threshold value, inducing a change from the cub-phase stacking to the hexagonal one, which is consistent with our previous DFT prediction [36]. After 60 min irradiation, a hexagonal lattice with regular atomic blocks separated by structural gaps was observed. The corresponding FFT pattern shows the []h view direction of the hexagonal lattice. Although it is not possible to determine the critical concentration of vacancies that induces the stacking change from the HRTEM experiments, the overall transition process is consistently observed, see Fig. S2. The process can be summarized into four stages, namely, (I) uncorrelated vacancy diffusion, (II) correlated vacancy ordering, (III) change in atomic stacking sequence and (IV) further vacancy and compositional ordering.
Fig. 2.
HRTEM images and corresponding FFT patterns of the local area shown inFig. 1b. Vacancy ordering is represented by appearance of the lattice periodicity of the hex‑phase, as indicated by red and yellow arrows, respectively.
3.2. Ex situ TEM characterization
To confirm that this irradiation-induced phenomenon is mainly due to thermal effects, we also performed cross-sectional TEM characterization on the thin film samples annealed in oven. As shown in Fig. 3, the specimen annealed at 175 °C formed a standard cub-phase without any visible vacancy ordering, while the vacancy ordering process became evident in the specimen annealed at 225 °C. As indicated by the red arrows, the ordering of vacancies proceeded along the {111}c planes, consistent with the in situ observation. At this annealing temperature, both cub- and hex‑phases coexist [35]. At higher annealing temperatures, large hex‑phase grains with clear GeSb2Te4 blocks and structural gaps were observed. This set ex situ heating experiments was performed using thin films of ∼450 nm, which shows consistent transition behavior as compared with the experiments carried out using thinner film of ∼100 nm [35].
Fig. 3.
Theex situTEM experiments upon thermal annealing in oven. Vacancy ordering appears in the cub-phase grains in the thin film sample annealed at 225 °C (red arrows).
To obtain atomic-scale details on the transformed state, we carried out high-resolution HAADF-STEM imaging experiments on three sets of samples obtained by in situ irradiation experiments under different accelerating voltages and by ex situ heating experiments. As shown in Fig. S3, clear septuple-layer (SL) GeSb2Te4 blocks are revealed under ex situ heating (annealed at 300 °C) and under in situ irradiation using JEM-200 CX TEM, where heating effects dominated owing to the relatively lower accelerating voltage at 120 keV. However, quintuple-layer (QL) Sb2Te3 blocks were mostly found in the sampled irradiated in the JEOL JEM-2100F TEM operated at 200 keV, as shown in Fig. S3c. Due to the competing kinetic effect brought by the higher accelerating voltage, the surrounding matrix turned into amorphous phase [56] with Ge-richer composition. Inside the crystalline grain, the composition is deficient in Ge, accounting around 2.6 at% according to the EDX analysis. Therefore, the major structural fragment is SL, instead of QL. Nevertheless, the atomic details of vacancy ordering are the same for both Ge-Sb-Te alloys and the parent phase, Sb2Te3.
3.3. Electronic and optical properties
Next, we performed DFT calculations to gain further understanding of vacancy ordering in Ge-Sb-Te alloys. Following our previous work, we built three sets of orthorhombic supercell models as shown in Fig. 4a, where the vertical direction corresponds to the [111]c direction of the cub-phase [36]. Each model contains 144 Ge, 288 Sb, 576 Te and 144 vacancies. The first set of models contains randomly distributed vacancies occupying 25% of each cation-like lattice plane (denoted as cub-25%), and the second set includes 3 vacancy-rich (111)c planes containing 50% vacancies (denoted as cub-50%). In the third set of models, vacancies are distributed over the (111)c and its equivalent planes, and all the planes contain 50% vacancies (denoted as cub’−50%). For each configuration, five independent models with statistical distribution of vacancies, Ge and Sb atoms were generated and optimized by DFT relaxation calculations using CP2K. The obtained intersection angle between the two vacancy-rich planes is ∼71°, in perfect agreement with the HRTEM observation. The average energy of the cub-50% model is ∼5.5 meV/atom lower than that of cub-25%, while cub’−50% is slightly lower in energy than cub-50% by ∼1.8 meV/atom. This energy trend provides an explanation why vacancy ordering into both (111)c and its equivalent planes takes place in parallel at the early stage of the transition. Although from HRTEM images the specific concentration of vacancy cannot be determined for these vacancy-rich planes this uncorrelated vacancy ordering could proceed before the cub’−50% configuration is reached. In other words, it is feasible to build, e.g., cub’−30% or cub'−40% models. However, further ordering of vacancies must proceed via one set of {111}c planes, as it is no longer possible to increase the vacancy concentration in both sets of planes in parallel while keeping the correct stoichiometry. For instance, a hypothetical cub’−75% configuration in this orthorhombic supercell would request 216 vacancies, which is larger than the actual number of vacancies in our stoichiometric GeSb2Te4 model (144 vacancies).
Fig. 4.
Atomic and electronic structure of one of the models with equivalent vacancy-rich planes. (a) Atomic structures of representative cub-25%, cub-50% and cub’−50% models, respectively. The green, yellow, blue and red spheres represent Ge, Sb, Te and vacancies, respectively. (b) Calculated density of states (DOS) and inverse participation ratio (IPR) of the cub’−50% model. (c) Highest occupied molecular orbital (HOMO) of the cub’−50% model. The isosurface rendered in purple corresponds to an isovalue of 0.012 a.u.
We have shown that partial vacancy ordering at the stage of cub-50% is not sufficient to trigger the insulator-to-metal transition [36]. Here, we show that the cub’−50% configuration features Anderson localization of electrons, similar to that of the cub-50% case. Electron localization is consistently observed in all the five cub’−50% models as evidenced by DFT calculations. The calculated density of states (DOS) and inverse participation ratio (IPR) values of one such model is shown in Fig. 4b. The IPR is used to determine the degree of localization of a specific electronic state, i.e., the larger the IPR value, the more localized the state. Specifically, the IPR of a Kohn-Sham state is defined as , where the ’s indicate the expansion coefficients of with respect to the basis set. The DOS and IPR profiles shown in Fig. 4b reveal Anderson insulating behavior, in which the band tails contain localized states, i.e. states with a significantly larger IPR value, while the states away from the Fermi energy are delocalized. Fig. 4c shows that the highest occupied molecular orbital (HOMO) is localized around a vacancy-rich region, consistent with our previous studies [26, 36]. In short, the formation of equivalent vacancy-rich (111)c planes at the early stage does not yet suppress Anderson localization in the band tails in the cub-phase. We repeated the same set of calculations for Sb2Te3 (Fig. S4), and observed the same trend in total energy and electron localization upon vacancy ordering as in GeSb2Te4.
We also evaluated the impact of vacancy ordering on the optical reflectivity. For comparison, we also considered the optical response of the amor- and hex‑ phases. Smaller models were built for the calculation of the optical response using VASP. We calculated the frequency-dependent dielectric matrix within the independent-particle approximation without considering local field effects and many body effects, which have been shown not to affect the results significantly in similar compounds [57, 58]. Fig. 5 shows obvious contrast in the calculated optical reflectivity between the three phases. Conventional phase-change photonic devices mainly utilize the contrast between the amor-phase and the cub-phase, labeled as state “0″ and “1″. Given the similar DOS profile of the two phases, the significant contrast in reflectivity is attributed to the alignment (cub-phase) and misalignment (amor-phase) of the p orbitals of the Ge, Sb and Te atoms [59]. In particular, in cub-GST, the average number of p electrons is three per site, forming an extended network of highly aligned p orbitals, which leads to MVB [30], [31], [32], [33], [34]. Recently, the hex‑GST alloyed with Nitrogen impurities was used as third logic state in non-volatile displays [60], in addition to the amor- and cub-phase. Our calculations indicate that the differences in reflectivity between the cub-25%, cub-50% and cub’−50% models are small. However, the hex‑phase exhibits a much larger reflectivity, consistent with experimental observations [61]. As shown in the DOS profile, a notable peak is present at ∼1 eV near the edge of the conduction band for the hex‑phase, enhancing the optical inter-band excitations and, thus, the reflectivity.
Fig. 5.
The calculated optical reflectivity and DOS for the hexagonal, cub-25%, cub-50%, cub’−50% and amorphous phase. The atomic models shown in the left panel represent the hexagonal, cubic and amorphous phase, corresponding to the logic states “2″, “1″ and “0″, respectively. The full atomic structures are shown in Fig. S1.
3.4. Multifold phase transitions
At last, we provide a summary diagram of all the phase transition paths for Ge-Sb-Te alloys and Sb2Te3 in Fig. 6. Four phases are involved, namely the stable hexagonal phase, the metastable cubic and amorphous phases, and a transient liquid phase. Conventional phase-change memory makes use of Joule heating (thermal effect), induced by either electrical or optical pulses, to trigger the transition between the amorphous and cubic phase. Thermal effects directly drive the crystallization of the amorphous state into the cubic phase, while the reverse, “melt-quench” process has to go through the liquid state. By tuning the volume ratio between the amorphous and cubic phase using multiple weak pulses, multilevel states can be achieved via either partial amorphization (cubic → liquid → amorphous) or cumulative crystallization (amorphous → cubic), as shown in Fig. 6b. The former process defines multilevel states during programming in a more accurate manner, but suffers from the spontaneous structural relaxation (aging) of the amorphous phase [7]. The latter process allows a continuous and non-linear change in conductance and reflectivity, but is subjected to the high randomness associated with nucleation and grain growth into polycrystalline forms [2]. Both device operations are very important for multilevel storage and neuro-inspired and in-memory computing applications [62], [63], [64], and have been systematically improved via both materials optimization [65], [66], [67] and device engineering [68].
Fig. 6.
Phase transition paths for Ge-Sb-Te alloys and Sb2Te3. (a) Summary of the phase transitions between the amorphous (amor), cubic (cub), hexagonal (hex) and liquid phase. The red and blue arrows indicate the paths triggered by thermal effects and kinetic effects, respectively. (b) shows the cumulative crystallization and partial amorphization to realize multilevel states in phase-change devices, involving the transition paths triggered by thermal effects marked in the gray shaded area in a. (c) Atomic details of vacancy ordering/disordering involved in the phase transition between the cubic and hexagonal phase. The transition paths are highlighted in the shaded orange area in a. The big red dots and the thick red lines represent vacancies and vacancy-rich planes, whereas smaller dots represent the atoms.
The thermal effect can also trigger the transition from the metastable cubic phase to the stable hexagonal phase (cubic → hexagonal), while the reverse transition by this effect is rather complex, i.e., hexagonal → liquid → amorphous → cubic. In contrast, a direct hexagonal → cubic transition can proceed via the knock-on kinetic effect induced by electron irradiation [54, 55] or ion bombardment [69]. Stronger kinetic effects can even trigger a direct amorphization process, bypassing the liquid phase (cubic → amorphous) [56]. The crystal-crystal transitions are mediated by either vacancy ordering or vacancy disordering, as outlined in Fig. 6c. The ordering of vacancies in the cubic phase first proceeds through an uncorrelated ordering in the equivalent vacancy-rich planes, and is followed by correlated ordering with only one set of dominant vacancy-rich planes. Further vacancy ordering drives block shifts to form the hexagonal stacking. This process can be reversed in the presence of sufficient knock-on collisions. The structural evolution towards the hexagonal phase largely extends the programming window for both electrical and optical devices, and does not require any change in programming volume. Given the absence of aging effects in the crystalline states, the vacancy ordering process could be useful for advanced memory and computing applications, although more thermal power and/or wider pulse widths are needed to form, as well as to melt down, these more stable crystalline states.
4. Conclusion
In summary, we have shown a progressive vacancy ordering mediated structural transition from cub- to hex‑phase in GST and Sb2Te3 thin films under electron beam irradiation. The in situ HRTEM experiments present a comprehensive temporal and spatial view of the vacancy ordering process, and the DFT calculations unveil the energy origin and the associated changes in the electronic and optical properties along the transition path. Consistent with oven heating, the electron irradiation induced thermal effect triggers a progressive vacancy ordering and structural transformation, which proceeds via uncorrelated vacancy ordering within equivalent vacancy-rich planes, followed by correlated ordering with one set of vacancy-rich planes, and finally further ordering and block shifting. We also provide a comprehensive overview of all phase transition paths for Ge-Sb-Te alloys and Sb2Te3, induced by either thermal or kinetic agitation. The summary diagram can serve as a guide to tune the phase transition, and thereby, to tailor the properties of phase-change alloys for practical memory and computing applications.
Declaration of competing interest
The authors declare that they have no conflicts of interest in this work.
Acknowledgments
W.Z. thanks the support of National Natural Science Foundation of China (61774123). E.M. acknowledges the support of National Natural Science Foundation of China (52150710545). W.Z. and E.M. are grateful to XJTU for the support of their work at CAID. J.-J.W. and M.W. acknowledges financial support from Alexander von Humboldt Foundation. M.W. acknowledges funding from Deutsche Forschungsgemeinschaft within SFB 917 “Nanoswitches”. W.Z. thanks the support of 111 Project 2.0 (BP2018008) and the International Joint Laboratory for Micro/Nano Manufacturing and Measurement Technologies of Xi'an Jiaotong University. The authors also acknowledge the computational resources provided by the HPC platform of Xi'an Jiaotong University and the Hefei Advanced Computing Center, and the National Supercomputing Center in Xi'an. The authors thank Danli Zhang and Ruihua Zhu for their technical support on TEM experiments. W.Z. thanks Dr. Andriy Lotnyk for useful discussions.
Biographies

Ting-ting Jiang obtained her Bachelor's and Master's degrees from China University of Petroleum and China University of Geosciences, respectively. In 2022, she obtained the doctoral degree at School of Materials Science and Engineering, Xi'an Jiaotong University. Her research interests include phase-change materials for memory devices, electron microscopy characterization and molecular beam epitaxy.

Wei Zhang is a professor at School of Materials Science and Engineering of Xi'an Jiaotong University. He received his bachelor's and master's degrees in physics from Zhejiang University, China. He obtained his doctoral degree with distinction from RWTH Aachen University, Germany, in 2014. After his postdoc work at RWTH, he joined Xi'an Jiaotong University in 2015. His research interests include phase-change memory, neuro-inspired computing, amorphous alloys, and first-principles materials design.
Footnotes
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.fmre.2022.09.010.
Appendix. Supplementary materials
References
- 1.Wuttig M., Yamada N. Phase-change materials for rewriteable data storage. Nat. Mater. 2007;6:824–832. doi: 10.1038/nmat2009. [DOI] [PubMed] [Google Scholar]
- 2.Zhang W., Mazzarello R., Wuttig M., et al. Designing crystallization in phase-change materials for universal memory and neuro-inspired computing. Nat. Rev. Mater. 2019;4:150–168. [Google Scholar]
- 3.Sebastian A., Gallo M.Le, Khaddam-Aljameh R., et al. Memory devices and applications for in-memory computing. Nat. Nanotechnol. 2020;15:529–544. doi: 10.1038/s41565-020-0655-z. [DOI] [PubMed] [Google Scholar]
- 4.Rao F., Ding K., Zhou Y., et al. Reducing the stochasticity of crystal nucleation to enable subnanosecond memory writing. Science. 2017;358(6369):1423–1427. doi: 10.1126/science.aao3212. [DOI] [PubMed] [Google Scholar]
- 5.Feldmann J., Youngblood N., Wright C.D., et al. All-optical spiking neurosynaptic networks with self-learning capabilities. Nature. 2019;569:208–214. doi: 10.1038/s41586-019-1157-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Zhang Y., Chou J.B., Li J., et al. Broadband transparent optical phase change materials for high-performance nonvolatile photonics. Nat. Commun. 2019;10:4279. doi: 10.1038/s41467-019-12196-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Zhang W., Ma E. Unveiling the structural origin to control resistance drift in phase-change memory materials. Mater. Today. 2020;41:156–176. [Google Scholar]
- 8.Shen J., Jia S., Shi N., et al. Elemental electrical switch enabling phase segregation–free operation. Science. 2021;374(6573):1390–1394. doi: 10.1126/science.abi6332. [DOI] [PubMed] [Google Scholar]
- 9.Wang X.-.D., Zhang W., Ma E. Monatomic phase-change switch. Sci. Bull. 2022;67(9):888–890. doi: 10.1016/j.scib.2022.01.020. [DOI] [PubMed] [Google Scholar]
- 10.Xu M., Mai X., Lin J., et al. Recent advances on neuromorphic devices based on chalcogenide phase-change materials. Adv. Funct. Mater. 2020;30(50) [Google Scholar]
- 11.Yamada N., Ohno E., Nishiuchi K., et al. Rapid-phase transitions of GeTe-Sb2Te3 pseudobinary amorphous thin films for an optical disk memory. J. Appl. Phys. 1991;69(5):2849–2856. [Google Scholar]
- 12.Matsunaga T., Yamada N. Structural investigation of GeSb2Te4: A high-speed phase-change material. Phys. Rev. B. 2004;69(10) [Google Scholar]
- 13.Yamada N., Matsunaga T. Structure of laser-crystallized Ge2Sb2+xTe5 sputtered thin films for use in optical memory. J. Appl. Phys. 2000;88(12):7020–7028. [Google Scholar]
- 14.Park J.-.B., Park G.-.S., Baik H.-.S., et al. Phase-change behavior of stoichiometric Ge2Sb2Te5 in phase-change random access memory. J. Electrochem. Soc. 2007;154(3):H139. [Google Scholar]
- 15.Ross U., Lotnyk A., Thelander E., et al. Direct imaging of crystal structure and defects in metastable Ge2Sb2Te5 by quantitative aberration-corrected scanning transmission electron microscopy. Appl. Phys. Lett. 2014;104(2) [Google Scholar]
- 16.Behrens M., Lotnyk A., Gerlach J.W., et al. Ultrafast interfacial transformation from 2D- to 3D-bonded structures in layered Ge-Sb-Te thin films and heterostructures. Nanoscale. 2018;10(48):22946–22953. doi: 10.1039/c8nr06567e. [DOI] [PubMed] [Google Scholar]
- 17.Momand J., Wang R., Boschker J.E., et al. Interface formation of two- and three-dimensionally bonded materials in the case of GeTe-Sb2Te3 superlattices. Nanoscale. 2015;7(45):19136–19143. doi: 10.1039/c5nr04530d. [DOI] [PubMed] [Google Scholar]
- 18.Zhang B., Zhang W., Shen Z.-.J., et al. Element-resolved atomic structure imaging of rocksalt Ge2Sb2Te5 phase-change material. Appl. Phys. Lett. 2016;108(19) [Google Scholar]
- 19.Zheng Y., Wang Y., Xin T., et al. Direct atomic identification of cation migration induced gradual cubic-to-hexagonal phase transition in Ge2Sb2Te5. Commun. Chem. 2019;2(1):13. [Google Scholar]
- 20.Zhu M., Ren K., Liu L., et al. Direct observation of partial disorder and zipperlike transition in crystalline phase change materials. Phys. Rev. Mater. 2019;3(3) [Google Scholar]
- 21.Sun Z., Zhou J., Ahuja R. Structure of phase change materials for data storage. Phys. Rev. Lett. 2006;96(5) doi: 10.1103/PhysRevLett.96.055507. [DOI] [PubMed] [Google Scholar]
- 22.Zhang W., Wuttig M., Mazzarello R. Effects of stoichiometry on the transport properties of crystalline phase-change materials. Sci. Rep. 2015;5:13496. doi: 10.1038/srep13496. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Jones R.O. Phase change memory materials: Rationalizing the dominance of Ge/Sb/Te alloys. Phys. Rev. B. 2020;101(2) [Google Scholar]
- 24.He S., Zhu L., Zhou J., et al. Metastable stacking-polymorphism in Ge2Sb2Te5. Inorg. Chem. 2017;56(19):11990–11997. doi: 10.1021/acs.inorgchem.7b01970. [DOI] [PubMed] [Google Scholar]
- 25.Wang X.-.P., Li X.-.B., Chen N.-.K., et al. Element-specific amorphization of vacancy-ordered GeSbTe for ternary-state phase change memory. Acta Mater. 2017;136:242–248. [Google Scholar]
- 26.Xu Y., Wang X., Zhang W., et al. Materials screening for disorder-controlled chalcogenide crystals for phase-change memory applications. Adv. Mater. 2021;33(9) doi: 10.1002/adma.202006221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Caravati S., Bernasconi M., Kühne T.D., et al. First principles study of crystalline and amorphous Ge2Sb2Te5 and the effects of stoichiometric defects. J. Phys. Condens. Matter. 2009;21 doi: 10.1088/0953-8984/21/25/255501. [DOI] [PubMed] [Google Scholar]
- 28.Wuttig M., Lusebrink D., Wamwangi D., et al. The role of vacancies and local distortions in the design of new phase-change materials. Nat. Mater. 2007;6(2):122–128. doi: 10.1038/nmat1807. [DOI] [PubMed] [Google Scholar]
- 29.Lencer D., Salinga M., Wuttig M. Design rules for phase-change materials in data storage applications. Adv. Mater. 2011;23(18):2030–2058. doi: 10.1002/adma.201004255. [DOI] [PubMed] [Google Scholar]
- 30.Wuttig M., Deringer V.L., Gonze X., et al. Incipient metals: Functional materials with a unique bonding mechanism. Adv. Mater. 2018;30(51) doi: 10.1002/adma.201803777. [DOI] [PubMed] [Google Scholar]
- 31.Zhu M., Cojocaru-Mirédin O., Mio A.M., et al. Unique bond breaking in crystalline phase change materials and the quest for metavalent bonding. Adv. Mater. 2018;30(18) doi: 10.1002/adma.201706735. [DOI] [PubMed] [Google Scholar]
- 32.Raty J.Y., Schumacher M., Golub P., et al. A quantum-mechanical map for bonding and properties in solids. Adv. Mater. 2019;31(3) doi: 10.1002/adma.201806280. [DOI] [PubMed] [Google Scholar]
- 33.Maier S., Steinberg S., Cheng Y., et al. Discovering electron-transfer-driven changes in chemical bonding in lead chalcogenides (PbX, where X = Te, Se, S, O) Adv. Mater. 2020;32(49) doi: 10.1002/adma.202005533. [DOI] [PubMed] [Google Scholar]
- 34.Guarneri L., Jakobs S., von Hoegen A., et al. Metavalent bonding in crystalline solids: how does it collapse? Adv. Mater. 2021;33(39) doi: 10.1002/adma.202102356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Siegrist T., Jost P., Volker H., et al. Disorder-induced localization in crystalline phase-change materials. Nat. Mater. 2011;10:202–208. doi: 10.1038/nmat2934. [DOI] [PubMed] [Google Scholar]
- 36.Zhang W., Thiess A., Zalden P., et al. Role of vacancies in metal-insulator transitions of crystalline phase-change materials. Nat. Mater. 2012;11:952–956. doi: 10.1038/nmat3456. [DOI] [PubMed] [Google Scholar]
- 37.Li K., Peng L., Zhu L., et al. Vacancy-mediated electronic localization and phase transition in cubic Sb2Te3. Mater. Sci. Semicond. Proc. 2021;135 [Google Scholar]
- 38.Park Y.J., Lee J.Y., Kim Y.T. In situ transmission electron microscopy study of the nucleation and grain growth of Ge2Sb2Te5 thin films. Appl. Surf. Sci. 2006;252(23):8102–8106. [Google Scholar]
- 39.Zhang B., Wang X.P., Shen Z.J., et al. Vacancy structures and melting behavior in rock-salt GeSbTe. Sci. Rep. 2016;6:25453. doi: 10.1038/srep25453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Zheng Y., Cheng Y., Huang R., et al. Surface energy driven cubic-to-hexagonal grain growth of Ge2Sb2Te5 thin film. Sci. Rep. 2017;7:5915. doi: 10.1038/s41598-017-06426-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Aryana K., Gaskins J.T., Nag J., et al. Interface controlled thermal resistances of ultra-thin chalcogenide-based phase change memory devices. Nat. Commun. 2021;12:774. doi: 10.1038/s41467-020-20661-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Hutter J., Iannuzzi M., Schiffmann F. cp2k: Atomistic simulations of condensed matter systems. WIREs Comput. Mol. Sci. 2014;4:15–25. [Google Scholar]
- 43.Goedecker S., Teter M., Hutter J. Separable dual-space Gaussian pseudopotentials. Phys. Rev. B. 1996;54:1703. doi: 10.1103/physrevb.54.1703. [DOI] [PubMed] [Google Scholar]
- 44.Perdew J.P., Burke K., Ernzerhof M. Generalized gradient approximation made simple. Phys. Rev. Lett. 1996;77:3865–3868. doi: 10.1103/PhysRevLett.77.3865. [DOI] [PubMed] [Google Scholar]
- 45.Kresse G., Hafner J. Ab initio molecular dynamics for liquid metals. Phys. Rev. B. 1993;47:558–561. doi: 10.1103/physrevb.47.558. [DOI] [PubMed] [Google Scholar]
- 46.Kresse G., Joubert D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B. 1999;59:1758. [Google Scholar]
- 47.Sun L., Zhou Y., Wang X., et al. Ab initio molecular dynamics and materials design for embedded phase-change memory. npj Comput. Mater. 2021;7:29. [Google Scholar]
- 48.Williams D.B., Carter C.B. Springer Press; New York: 2009. Transmission Electron Microscopy: A Textbook For Material Science. [Google Scholar]
- 49.Lotnyk A., Ross U., Dankwort T., et al. Atomic structure and dynamic reconfiguration of layered defects in van der Waals layered Ge-Sb-Te based materials. Acta Mater. 2017;141:92–96. [Google Scholar]
- 50.Kolobov A.V., Fons P., Saito Y., et al. Atomic reconfiguration of van der Waals gaps as the key to switching in GeTe/Sb2Te3 superlattices. ACS Omega. 2017;2(9):6223–6232. doi: 10.1021/acsomega.7b00812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Wang J.-.J., Wang J., Xu Y., et al. Layer-switching mechanisms in Sb2Te3. Phys. Status Solidi RRL. 2019;13(10) [Google Scholar]
- 52.Lotnyk A., Bernütz S., Sun X., et al. Real-space imaging of atomic arrangement and vacancy layers ordering in laser crystallised Ge2Sb2Te5 phase change thin films. Acta Mater. 2016;105:1–8. [Google Scholar]
- 53.Lotnyk A., Dankwort T., Hilmi I., et al. In situ observations of reversible vacancy ordering process in van der Waals bonded Ge-Sb-Te thin films and GeTe-Sb2Te3 superlattices. Nanoscale. 2019;11(22):10838–10845. doi: 10.1039/c9nr02112d. [DOI] [PubMed] [Google Scholar]
- 54.Jiang T.-.T., Wang X.-.D., Wang J.-.J., et al. In situ study of vacancy disordering in crystalline phase-change materials under electron beam irradiation. Acta Mater. 2020;187:103–111. [Google Scholar]
- 55.Wang X., Zhang H., Wang X., et al. Designing Sb2Te3 heterophase homostructure. Chin. Sci. Bull. 2022;67(22):2662–2671. [Google Scholar]
- 56.Jiang T.-.T., Wang J.-.J., Lu L., et al. Zhang, Progressive amorphization of GeSbTe phase-change material under electron beam irradiation. APL Mater. 2019;7(8) [Google Scholar]
- 57.Wełnic W., Wuttig M., Botti S., et al. Local atomic order and optical properties in amorphous and laser-crystallized GeTe. Compt. Rendus Phys. 2009;10(6):514–527. [Google Scholar]
- 58.Caravati S., Bernasconi M., Parrinello M. First principles study of the optical contrast in phase change materials. J. Phys. Condens. Matter. 2010;22(31) doi: 10.1088/0953-8984/22/31/315801. [DOI] [PubMed] [Google Scholar]
- 59.Huang B., Robertson J. Bonding origin of optical contrast in phase-change memory materials. Phys. Rev. B. 2010;81(8) 081204(R) [Google Scholar]
- 60.Bi C., Xu K., Hu C., et al. Three distinct optical-switching states in phase-change materials containing impurities: from physical origin to material design. J. Mater. Sci. Technol. 2021;75:118–125. [Google Scholar]
- 61.Hu C., Yang Z., Bi C., et al. All-crystalline” phase transition in nonmetal doped germanium–antimony–tellurium films for high-temperature non-volatile photonic applications. Acta Mater. 2020;188:121–130. [Google Scholar]
- 62.Sebastian A., Gallo M.Le, Burr G.W., et al. Tutorial: Brain-inspired computing using phase-change memory devices. J. Appl. Phys. 2018;124(11) [Google Scholar]
- 63.Zhang W., Mazzarello R., Ma E. Phase-change materials in electronics and photonics. MRS Bulletin. 2019;44(9):686–690. [Google Scholar]
- 64.Wright D., Bhaskaran H., Pernice W. Integrated phase-change photonic devices and systems. MRS Bulletin. 2019;44(9):721–725. [Google Scholar]
- 65.Ding K., Wang J., Zhou Y., et al. Phase-change heterostructure enables ultralow noise and drift for memory operation. Science. 2019;366(6462):210–215. doi: 10.1126/science.aay0291. [DOI] [PubMed] [Google Scholar]
- 66.Liu B., Li K., Liu W., et al. Multi-level phase-change memory with ultralow power consumption and resistance drift. Sci. Bull. 2021;66(21):2217. doi: 10.1016/j.scib.2021.07.018. [DOI] [PubMed] [Google Scholar]
- 67.Ma P., Tong H., Xu M., et al. Suppressed resistance drift from short range order of amorphous GeTe ultrathin films. Appl. Phys. Lett. 2020;117(2) [Google Scholar]
- 68.Koelmans W.W., Sebastian A., Jonnalagadda V.P., et al. Projected phase-change memory devices. Nat. Commun. 2015;6:8181. doi: 10.1038/ncomms9181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69.Privitera S., Mio A.M., Smecca E., et al. Structural and electronic transitions in Ge2Sb2Te5 induced by ion irradiation damage. Phys. Rev. B. 2016;94 [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.







