Symposium S.EL09 : Phase-Change Materials for Electronic and Photonic Nonvolatile Memory and Neuro-Inspired Computing

The processing of an input dataset by a deep neural network, e.g. for classification tasks, is a series of vector-matrix multiplications (input data * synapse weight). The size and number of matrices that need to be processed are too large to be stored in a Static Random Access Memory (SRAM) unit. For this reason, data need to be shuffled constantly between memory and processing unit in a classical computing architecture, leading to inefficiency and performance mitigation. Dedicated neuromorphic computing hardware may help to overcome these issues and is a promising technology for the post–von Neumann computing era. Especially memristive devices are suitable for this dedicated hardware. A crossbar of memristive devices stores the synaptic weights in a physical matrix. The weight can be read non-destructively and updated e.g. during training. The matrix vector multiplication can be performed in-memory by coding the synaptic weight to the device conductance and the input data to the read voltage amplitude.
Several memristive technologies are under study for the implementation of neuromorphic hardware, and they rely on diverse physical mechanisms and materials [1]. However, improvements in the device characteristics are required for optimal hardware acceleration in both performance precision and energy consumption. Examples of challenges are the minimization of inter and intra device asymmetricity, non-linearity and stochasticity, and the increase of the dynamic range.
In this talk I will present Phase Change Memories (PCM) and filamentary-based Oxide Resistive RAM (OxReRAM), the two most promising candidates to represent artificial synapses in neuromorphic hardware [2,3]. In PCM-based memristors, the conductance change relies on a phase transition in the material, while for filamentary-based OxReRAM, the synaptic weight depends on the rupture/formation of oxygen vacancies based conductive paths. Strength and challenges of the two classes of memristors will be discussed. For each technology, an innovative design and material stack concepts will be presented, demonstrating enhanced operational characteristics. The material-stack characterization by means of X-ray reflectivity and diffraction, Hall-measurements as well as the electrical characterization of the devices will be covered. A discussion on the ideal class of applications for a more efficient use of PCM- or OxReRAM-based synapses will follow.

[1] Rajendran, B. & Alibart, IEEE J. Emerg. Sel. Top. Circuits Syst. 6, 198–211 (2016)
[2] Abu Sebastian et al., Jour. Of Phys. D, 52, 44 (2019)
[3] B. Govoreanu, et al., Proc. IEEE Int. Electron Devices Meeting, 31.6.1–31.6.4 (2011)

Crystallization of phase change materials has been studied by extensive density functional/molecular dynamics simulations (DF/MD). Four crystallization simulations of amorphous Ge₂Sb₂Te₅(460 atoms) have been completed at 600 K with simulation times up to 8.2 ns.^1-2 A sample with a history of order crystallizes completely in 1.2 ns, but ordering in others takes more time and is less complete. The amorphous starting structures without memory display phases (<1 ns) with subcritical nuclei (10–50 atoms) ranging from nearly cubical blocks to string-like configurations of ABAB squares and AB bonds extending across the cell. Percolation initiates the rapid phase of crystallization and is coupled to the directional p-type bonding. The results emphasize the stochastic nature of crystallization and the importance of sufficiently large samples. This is particularly evident in describing the role of crystallites that can merge to form larger units or hinder complete crystallization by the formation of grain boundaries.
Amorphous Sb is known to crystallize extremely rapidly already at room temperature. Crystallization of Sb has been studied at 600 K using six DF/MD simulations with up to 882 atoms.³ Crystallization proceeded layer-by-layer in most cases and was extremely rapid (∼36 m/s). As shown in Fig. 1, diffusion plays a minor role in the process as the crystallization proceeds from the crystalline rim, and the evolution of bond lengths and ring statistics supports the bond-interchange model of Sb-rich phase change materials.⁴
We have also carried out extensive DF/MD simulations (over 500 atoms, up to 100 ps) of liquid bismuth at four temperatures between 573 - 1023 K.⁵ These simulations provided details of the dynamical structure factors, the dispersion of longitudinal and transverse collective modes, and related properties (power spectrum, viscosity, and sound velocity). Agreement with available inelastic x-ray and neutron scattering data and with previous simulations is generally very good. The results show that DF/MD dynamics simulations can give dynamical information of good quality without the use of fitting functions, even at long wavelengths.

References:
1) J. Kalikka, J. Akola, J. Larrucea, and R. O. Jones, Phys. Rev. B 86, 144113 (2012).
2) J. Kalikka, J. Akola, and R.O. Jones, Phys. Rev. B 90, 184109 (2014); ibid. B 94, 134105 (2016).
3) M. Ropo, J. Akola and R.O. Jones, Phys. Rev. B 96, 184102 (2017).
4) Matsunaga, J. Akola, S. Kohara et al., Nature Materials 10, 129 (2011).
5) M. Ropo, J. Akola and R.O. Jones, J. Chem. Phys. 145, 184502 (2016).

Phase Change Materials (PCMs) represent an important materials class from both a fundamental and technological point of view. Although much effort is still required to determine all the physical mechanisms underlying their complex physical properties, memory devices based on PCMs, which exploit the resistance contrast between the amorphous and crystalline phases, are already at the production stage. The ternary (GeTe)_m(Sb₂Te₃)_n alloys are the prototypical phase change materials commonly used for optical and memory devices in the composition Ge₂Sb₂Te₅ (GST225). Among other fabrication methods, such as pulsed laser deposition and sputtering, which is the main technique used in industry, Molecular Beam Epitaxy (MBE) has recently gained in importance because of the advantage to obtain epitaxial quasi single-crystalline PCM layers. This achievement has made it possible the investigation of material properties for which structural perfection is of crucial importance. For example, we have been able to demonstrate that highly ordered crystalline phases can be deposited by van der Waals (vdW) epitaxy, where adjacent building blocks are weakly bonded among them and to the substrate. Furthermore, epitaxial quasi single-crystalline GST, with ordered stacking of intrinsic vacancies, exhibits a large resistivity range that is promising for the realization of memory cells with large programming windows [1].
In this presentation we will first review our advances in the growth of epitaxial GST alloys by MBE on both Si and InAs substrates. We have carried out a growth engineering of single-crystalline GST to achieve unprecedented control over structural order, phase, and composition. The combined use of X-ray diffraction and Raman spectroscopy allowed us to draw a growth phase diagram for GST [2].
The capability to retain the stored information during the memory lifetime is a fundamental property of PCM, and any spontaneous evolution of the amorphous phase toward the more stable crystalline one is undesired. This is especially true when the device working temperature may be higher than the crystallization temperature, as in automotive applications. In this regards we will present our first results in the search of suitable PCM alloys with improved physical properties in terms of crystallization temperature and crystallization speed. In particular we will show our first crystallization study of Ge-rich GST, aiming to identify thermally stable PCMs with high crystallization temperature, and our first results on the investigation, by X-ray Photoemission Spectroscopy (XPS), of the electronic properties of both epitaxial GST and Ge-rich GST. XPS data will be correlated with X-ray diffraction and Raman characterization.
[1] V.Bragaglia, F.Arciprete, W.Zhang, A.M.Mio, E.Zallo, K.Perumal, A.Giussani, S.Cecchi, J.E.Boschker, H.Riechert, S.Privitera, E.Rimini, R.Mazzarello, and R.Calarco, Sci. Rep. 6, 23843 (2016)
[2] V.Bragaglia, F.Arciprete , A.M.Mio , and R.Calarco, J. Appl. Phys. 123, 215304 (2018)

The working principle of conventional phase change memory device is based on ultrafast reversible phase changes between crystalline and amorphous phases of Ge-Sb-Te (GST) materials. For information storage, phase change memory device uses a large contrast either in electrical resistance between the amorphous phase (high-resistance state) and crystalline phase (low-resistances state) or in optical reflectivity between the amorphous phase (low reflectivity state) and crystalline phase (high reflectivity phase). The erase of GST-based memory cell is achieved by applying a high intensity either electrical or laser pulses. These processes lead to the amorphization via melting and subsequent fast quenching of the phase change alloy. However, GST alloys are poor glass formers. Thus, rapid cooling rates are required to suppress the recrystallization of the alloys during the erase process. On another hand, the write process is accomplished by applying either electrical or optical pulses with low intensity. This results in an amorphous-to-crystalline phase transition. Due to intrinsic feature of GST alloys, a main challenge in material science is the optimization of memory writing times, which are limited by the crystallization kinetics of the alloys. In order to speed up the crystallisation rates, several strategies were proposed, including doping of GST materials, precrystallization of an amorphous matrix or using GST based superlattices. Being a nucleation-dominant material, the crystallization of GST alloy in a nanosized phase-change memory cell might however proceed at the crystalline interfaces.
This contribution focuses on determination of crystallization dynamics in epitaxial Ge2Sb2Te5 (GST225) thin films on application relevant (nanosized) length and (nanosecond) time scales. We irradiate the thin films by a single UV laser pulse with a 20 ns pulse duration and with different laser fluences. First, we study the structural transitions in epitaxial GST225 thin films starting from the layered trigonal GST225 structures [1]. We compare the obtained results with the behaviour of GeTe-Sb2Te3 based superlattices (SLs) after the laser irradiation. In both cases, the results reveal the phase transition to the cubic GST225 structure. The cubic phase forms from a transient molten phase at the melt-crystalline interface upon cooling process and crystallizes with epitaxial relationship to the parent phase. Second, by introducing a method based on combination of high temporal and spatial resolution, we determine the crystallization rates of the cubic GST225 phase [2]. The rates are ranging from 0.4 m/s to 1.7 m/s. The values are well in agreement with the published experimental and theoretical data. Moreover, a variation of the laser fluence leads to different cooling rates, which result in different solidification rates and freezing of an amorphous state. Irradiation of amorphous GST225 phase by laser pulse with lower fluence leads to the re-crystallization of the phase and epitaxial formation of the cubic GST225 structure. Overall, our work shows an approach for the investigation of crystallisation kinetics in wide range of phase change materials on application relevant length and time scales. In addition, we demonstrate amorphization and crystallization of GST225 material by using UV laser with single pulse duration and wavelength only, where in the conventional amorphous-to-crystalline phase transitions the lasers with different pulse durations, number and wavelengths are usually applied.
[1] M. Behrens, A. Lotnyk et al., Nanoscale 10 (2018) 22946-22953
[2] M. Behrens, A. Lotnyk et al., ACS Appl. Mater. Interfaces (2019) https://doi.org/10.1021/acsami.9b16111

Phase change random access memory (PCRAM) has been successfully applied in the computer storage architecture, as storage class memory, to bridge the performance gap between DRAM and Flash-based solid-state drive due to its good scalability, 3D-integration ability, fast operation speed and compatible with CMOS technology. A good understanding of phase change mechanism is of great help to design new phase change materials with fast operation speed, low power consumption and long-lifetime. In this presentation, we firstly review the development of PCRAM and different understandings on phase change mechanisms in recent years, and then propose a new view on the mechanism, which is based on the octahedral structure motifs and vacancies. Octahedral structure motifs are the based units during phase transition. Robust octahedra and plenty of vacancies in both amorphous and crystalline phase, respectively avoiding large atomic rearrangement and providing necessary space, are of crucial to achieve the nanosecond or even sub-nanosecond operation of PCRAM. Based on this mechanism, we introduce Sc-based robust octahedra in Sb₂Te₃ phase change materials, called Sc-Sb-Te, and achieve 700 ps operation speed. Recently, the concentration of dopants has also been optimized by correlative atom probe tomography and transmission electron microscopy investigations. Focusing on phase change materials and PCRAM for decades, we have successfully developed 128Mb embedded PCRAM chips, which can meet the requirements of most embedded systems.

Previously, we have reported on a novel class of chalcogenide phase change materials, based on Ge-Sb-Se-Te (GSST) alloys, with excellent infrared transparency from 1.5 to > 15 micron wavelengths. Increasing Se substitution leads to a progressive lowering of extinction coefficient. A material in this family with a particular useful stoichiometry, Ge₂Sb₂Se₄Te₁ (GSS4T1), possesses broadband transparency through the infrared range while maintaining a large index contrast between its amorphous and crystalline phases. Defining a figure of merit as Δn/k (where n,k are the real, imaginary parts of the refractive index), our materials shows >100x figure of merit (FOM) improvement over conventional GST throughout the infrared range.

In this talk, we describe novel applications enabled by reversible switching of low loss GSS4T1 material. In the first application, we exploit the exceptional FOM of the Se-doped PCM to realize a nonvolatile photonic switch with unprecedented high performance. Our device, based on a racetrack SiN resonator with a PCM gate, exhibits a large switching contrast ratio of 42 dB and a low insertion loss of < 0.5 dB, far outperforming previous nonvolatile switches as well as devices based on the classical GST-225 material with a similar configuration.

Additional novel applications are enabled by our development of wafer-scale pixelated electrical switchable PCM devices. The pixelated devices are fabricated in a CMOS-compatible clean room on 8 inch wafers utilizing a full Si-based fabrication tool set. Devices of various pixel sizes, from 1 to 30 micron have been fabricated. Additionally, fabrication of more complicated sub-pixel metasurfaces was demonstrated. Electrical switching measurement is obtained by applying electrical pulse trains through a programmable voltage generator into the gate of a high powered transistor. Microsecond switching of 30-micron and 1-micron pixels in a wafer-scale device has been demonstrated. Current devices have demonstrated over 1,000 switching cycles and further durability improvements are underway.

As an application of pixelated electrical device, we have demonstrated a concept of compact infrared reflection spectrometer with no moving parts. Here, we leverage the ability to obtain gradually changing reflection signature of the pixel with voltage between fully-converted crystalline and amorphous phases. This wavelength shift is essentially a nonlinear filter which can be used to reconstruct an input spectrum. Since the reflection response of the pixel states is pre-determined, a spectral response of the scene can be extracted once the pixel spectrum is fully characterized with a known calibrated spectral source. We have carried out such characterization utilizing a supercontinuum laser in the short-wave infrared part of the spectrum and the results of spectral signature extraction for three different incident spectral scenarios.

DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited.

This material is based upon work supported by the Under Secretary of Defense for Research and Engineering under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Under Secretary of Defense for Research and Engineering.

Demands for applying silicon photonics to high-performance computing systems have grown significantly in recent years due to the breakdown of Dennard scaling and information transfer bottlenecks in the conventional von Neumann architecture. Data processing on photonic platforms is a promising approach because of the potential advantages over electrical approaches where large bandwidth, high efficiency, ultrafast modulation speed, and low crosstalk are crucial. Critical photonic components, such as lasers, modulators, switches, filters, multiplexers, photodetectors, and memory cells, have been developed on different material platforms, such as silicon nitride (SiN), GeSi, III-V semiconductors, and others, with silicon on insulator (SOI) being the dominant platform. Owing to its compatibility with mature CMOS processes and integration with electrical devices, fabless and large scale silicon photonic integrated circuits (PICs) are available and thus offer a promising route to future commercialization.

Recently fully integrated photonic memory devices have been demonstrated on silicon nitride and silicon. This approach not only enables photonic data storage on-chip, but also shows multilevel storage, improved SNR, and reduced switching energy over available optical storage technology. Previous work on silicon nitride waveguides demonstrated both multilevel storage (> 32 levels) and computation in a fully optical framework. This was despite the fact that silicon nitride does not benefit from a high refractive index contrast (~2/1.5 in Si₃N₄/SiO₂ vs 3.5/1.5 in Si/SiO₂) or the ability to integrate active photonics (such as modulators and photodetectors) directly in the waveguide. Silicon-On-Insulator (SOI), on the other hand, has both these advantages, leading to smaller footprint devices and easy integration with high-speed, CMOS-based electronics. In spite of rapid advances of demonstrations in this field on both silicon and silicon nitride platforms, a clear pathway towards choosing between the two has been lacking.

Here, we systematically evaluate and compare the computation performance of phase change photonics on a silicon platform and a silicon nitride platform. Our experimental results show that whilst Silicon platforms are superior to Silicon Nitride in terms of potential for integration, modulation speed, and device footprint, it requires trade-offs in terms of energy efficiency. We then successfully demonstrate single-pulse modulation using phase-change optical memory (PCOM) on silicon photonic waveguides and demonstrate efficient programming, memory retention, and readout of > 4bits of data per cell. This decreased the device footprint compared with silicon nitride photonics and reduced both the energy and time required for reaching arbitrary memory levels. We have characterized the reliability of this approach and outlined a comparison with other state-of-the-art programming methods. The use of silicon demonstrates a major step toward making phase-change photonic memory a viable and integrable technology. Our results have potential applications like deep learning based on vector-matrix multiplication and neuromorphic computing. Our approach paves the way for the in-memory computing on the silicon photonic platform.

Phase-Change Memory (PCM) is a promising candidate for next generation of non-volatile memory [1]. PCM operations rely on the reversible phase transition between the amorphous and the crystalline phase, which occurs upon current-induced joule heating. Therefore, the reliability of the programming operation depends also on the thermal efficiency of the device [2]. Heat retention is thus essential in PCM and in complete contradiction with usual conception of electronic devices where thermal dissipation is a key point.
In this context, optimized thermal encapsulation in Ge-rich Ge-Sb-Te based PCM devices enables programming current reduction and improved data retention [3]. Indeed, an SiC encapsulation layer provides a more uniform heating of the active volume of the PCM cell, with respect to a standard SiN layer. Thanks to TEM and EDX analyses performed on state-of-the-art PCM heater-based devices, we demonstrate the higher uniformity of the heating process in the PCM cell during the programming operation achieved by optimized SiC encapsulation. Moreover, nano-diffraction patterns analysis highlights the crystalline morphology of the programmed active volume and of the surrounding material. Finally, we show the agreement between experimental data and TCAD simulations.

[1] P. Zuliani et al., "Engineering of chalcogenide materials for embedded applications of Phase-Change Memory," Solid-State Electronics 111 (2015) pp. 27-31.
[2] F. Xiong et al., "Towards ultimate scaling limits of phase-change memory," 2016 IEEE International Electron Devices Meeting (IEDM), pp. 4.1.1-4.1.4.
[3] A. L. Serra et al., "Outstanding Improvement in 4Kb Phase-Change Memory of Programming and Retention Performances by Enhanced Thermal Confinement," 2019 IEEE 11th International Memory Workshop (IMW), 2019, pp. 1-4.

The design of non-volatile memory cells based on phase change materials (PCM) is the subject of many studies. And one of the most promising materials are the films of Ge2Sb2Te5 chalcogenide glassy semiconductor (CGS). The operation of memory cell is based on a reversible glass-crystal phase transition, which occurs when a current pulse passes through the film's nano-volume. The embedding of metal impurity in the films can improve the property of memory cells. However, the features of phase transition in the metal-modified Ge2Sb2Te5 films are poorly understood. This work presents the results of the study of the structure observed in nanoscale films of Ge2Sb2Te5 modified by Bi (Ge2Sb2Te5<Bi>). By the method of HR-TEM imaging it was found that the structure of Ge2Sb2Te5<Bi> films presents the amorphous matrix with isolated crystalline nanoregions (nanoclusters) of metallic bismuth with average size ~ 8 nm. Local atomic structure of the films was studied by Raman spectroscopy. It was found that under the laser irradiation there is the structure transition of nanoscale Ge2Sb2Te5 films from amorphous into polycrystalline hexagonal stable state through metastable polycrystalline cubic phase. In contrast, in the Ge2Sb2Te5<Bi> films the transition from amorphous to polycrystalline hexagonal stable state under the laser irradiation occurs without intermediate metastable cubic phase.

An investigation into the relationship between crystal fraction and electrical conductivity for GST materials was performed. Transition from amorphous to conduction state for homogeneous nucleation of conductive crystallites in the resistive amorphous layer and limiation of the JMAK equation to modeling this transition is analyzed. In particular, the analysis relate to the validity of the model in terms of time-temperature description of crystallization and dependencies between minimum amount of crystal during nucleation and percolative threshold
M. Avrami, J. Phys. Chem. 7, 1103 (1939);
M. Avrami, J. Phys. Chem. 8, 212
M. Avrami, J. Phys. Chem. 9, 177 1941
Andrea Redaelli Editor, Phase Change Memory Device Physics, Reliability and Applications, Springer (2018)

GeSbTe alloys, widely employed for phase change non volatile memory, are usually not suitable in embedded and automotive applications, because of the low crystallization temperature. Doping, in particular with Ge, has been shown to be a viable way to extend the thermal stability of GeSbTe alloys [1,2]. However, once an optimised composition has been selected, during device operation, the material stoichiometry may change, due to the atomic migration induced by the high temperature and/or by the electric field [3]. These variations may lead to programming and retention performance degradation or even to device failure. It is therefore crucial to understand the extent of the atomic migration and its impact on the stoichiometry, as well as on the crystallization properties. In order to decouple the effect of the atomic diffusion at high temperature from that of the field induced electromigration, we have irradiated crystalline Ge rich GeSbTe thin films by laser. The films were prepared by sputter deposition in the amorphous phase, and then converted into the crystalline structure by annealing at 400°C. The irradiation has been performed by using a Yb-YAG laser (515 nm) with 600 ns pulse, operating at different energy densities. By changing the energy density, complete or partial melting and quenching, with amorphization of the film, can be achieved. In this way we are able to study the thermal diffusion processes decoupled from the field induced electromigration. After melting and quenching, the atomic elements distribution has been studied by Transmission Electron Microscopy (TEM) and Electron Energy Loss Spectroscopy (EELS). We find that, even without the electric field, the melting give rises to a prominent diffusion of Ge atoms. By employing finite elements computational analysis, a diffusion coefficient of Ge on the order of 5x10^-5 cm² s^-1 in the molten phase has been estimated.
After amorphization under different irradiation conditions, we have followed the crystallization upon thermal annealing by in-situ time resolved reflectivity measurement. Such a crystallization process, after the first melting and quenching, is expected to be different from the crystallization of the as deposited amorphous film, since the melting produces not only sub-critical crystalline nuclei, but also a modification of the stoichiometry. The study of the crystallization in such a “primed” material is very similar to the situation occurring in a real memory device after the first erasing step and therefore it is relevant for reliability evaluations.

[1] Zuliani P et al 2013 Overcoming temperature limitations in phase change memories with optimized GexSbyTez IEEE Trans. Electron Devices 60 4020
[2] Navarro G et al 2016 N-doping impact in optimized Ge-rich materials based phase-change memory 8th IEEE Int. Non volatile Memory Workshop (IMW)
[3] Padilla A et al 2001 Voltage polarity effects in Ge2Sb2Te5- based phase change memory devices J. Appl. Phys. 110 054501

With the increasing demand of high-density storage and neuromorphic hardware driven by the rapidly growing of artificial intelligence, miniaturized phase-change memory device has been developed for energy-efficient data storage and computing[1]. Typical phase-change memory devices are based on complex compounds with precisely controlled stoichiometry. Therefore, monatomic phase-change materials using single element of antimony has been proposed as a promising candidate in ultra-small phase-change device without the need of compositional optimization[2]. On the other hand, the integration of phase-change materials with photonics has provided considerable opportunities of data storage[3], computing[4] and machine learning[5] in photonics, benefted from the fast speed, large bandwidth and low cross talking in optical domain[6] as well as the exceptional properties of phase-change materials[7]. Herein, we systematically studied the material properties, especially the optical properties, of thin-layer antimony films during the phase-changing process. With a huge contrast of optical constants between the amorphous and crystalline states, we demonstrated integrated non-volatile photonic memories using monatomic phase-change materials, resulting in a unique platform for phase-change photonic applications.

References
[1] Xiong, F.; Liao, A. D.; Estrada, D.; Pop, E., Low-power switching of phase-change materials with carbon nanotube electrodes. Science 332, 568-570 (2011).
[2] Salinga, M.; Kersting, B.; Ronneberger, I.; Jonnalagadda, V. P.; Vu, X. T.; Gallo, M. L.; Giannopoulos, I.; Cojocaru-Mirédin, O.; Mazzarello, R.; Sebastian, A., Monatomic phase change memory. Nat. Mater. 17, 681-685 (2018).
[3] Ríos, C.; Stegmeier, M.; Hosseini, P.; Wang, D.; Scherer, T.; Wright, C. D.; Bhaskaran, H.; Pernice, W. H. P., Integrated all-photonic non-volatile multi-level memory. Nat. Photon. 9, 725-732 (2015).
[4] Ríos, C.; Youngblood, N.; Cheng, Z.; Gallo, M. L.; Pernice, W. H. P.; Wright, C. D.; Bhaskaran, H., In-memory computing on a photonic platform. Sci. Adv. 5, eaau5759 (2019)
[5] Feldmann, J.; Youngblood, N.; Wright, C. D.; Bhaskaran, H.; Pernice, W. H. P., All-optical spiking neurosynaptic networks with self-learning capabilities. Nature 569, 208-214 (2019).
[6] Caulfield, H. J.; Dolev, S., Why future supercomputing requires optics. Nat. Photon. 4, 261-263 (2010).
[7] Zhang, W.; Mazzarello, R.; Wuttig, M.; Ma, E., Designing crystallization in phase-change materials for universal memory and neuro-inspired computing. Nat. Rev. Mater. 4, 150-168 (2019).

It has been a long-time dream of mankind to design materials with tailored properties. In recent years, the focus of our work has been the design of phase change materials for applications in data storage. In this application, the remarkable property portfolio of phase change materials (PCMs) is employed, which includes the ability to rapidly switch between the amorphous and crystalline state. Surprisingly, in PCMs both states differ significantly in their properties. This material combination makes them very attractive for data storage applications in rewriteable optical data storage, where the pronounced difference of optical properties between the amorphous and crystalline state is employed. This unconventional class of materials is also the basis of a storage concept to replace flash memory. Today’s talk will discuss the unique material properties, which characterize phase change materials. In particular, it will be shown that only a well-defined group of materials utilizes a unique bonding mechanism (‘Bond No. 6’), which can explain many of the characteristic features of crystalline phase change materials. Different pieces of evidence for the existence of this novel bonding mechanism, which we have coined metavalent bonding, will be presented. In particular, we will present a novel map, which separates the known strong bonding mechanisms of metallic, ionic and covalent bonding, which provides further evidence that metavalent bonding is a novel and fundamental bonding mechanism. This insight is subsequently employed to design phase change materials as well as thermoelectric materials.

Future Electronic Smart Systems (ESS) require energy efficient storage and fast processing of large amount of data. Phase-Change Materials, exhibiting large electrical and optical contrast between two phases also allowing multiple intermediate states, are highly desired for this application, because in principle they also have the ability to store and process data at the same physical place(Wuttig et al. Nat. Mater. 2007). Despite progresses which have been made to achieve the full potential of Ge₂Sb₂Te₅ based PCM devices, several reliability issues still hinder the realization of ultimate performance. Void formation due to density change upon switching, low crystallization temperature for automotive applications, and resistance drift over time are among the few. GaSb is considered as a promising candidate for future phase change random access memory (PCRAM) devices because of its fast crystallization, high crystallization temperature, and lack of density change upon switching at a specific composition of Ga₃₀Sb₇₀(Putero et al. APL Materials. 2013). Pulsed Laser Deposition (PLD) has become one of the most popular thin film deposition technique. Some of the reasons for this popularity are the flexibility, processing speed, and the wide range of materials that can be deposited. In this work we discuss PLD of single and alternating layers of Sb₂Te₃, GaSb, and Ge-rich GST thin films on various substrates. Amorphous, polycrystalline, and textured heterostructures of Sb₂Te₃/GaSb and Sb₂Te₃/Ge-rich GST are deposited. Reflection high-energy electron diffraction (RHEED) is used as in-situ growth monitoring tool during deposition to achieve desired film quality. Moreover, deposited films are characterized by Scanning Transmission Electron Microscope (S/TEM) for composition and quality analysis before producing single cell vehicles for electrical device testing.

The crystallization behavior of amorphous nano-regions (20-100 nm in diameter) embedded in a textured epitaxial Ge₂Sb₂Te₅ 25 nm thick film grown on Si (111) substrate has been investigated in situ by TEM analysis. The amorphous regions were obtained by irradiation with 30 keV Ge+ at a fluence of 1.5x10¹⁴ ions/cm² of masked samples. The adopted configuration simulates the GST structure of a device in the RESET state, it is then of relevance for understanding their data retention characteristics. The in situ TEM analysis indicates that the amorphous to crystal transition is characterized by an initial growth velocity of 3.6 pm/s at 75°C, probably related to the external partially damaged area. For the previous annealed sample, a velocity of about 2.6 pm/s was observed at 90°C and a growth speed of 170 pm/s at 110°C for the 50nm and 100nm diameter amorphous spots. The 20 nm diameter amorphous spots recrystallize after the annealing at 90°C. The transition is governed only by crystallization (nucleation is absent) of the atoms located at the boundary of the amorphous dot with the surrounding crystalline regions. In some case a preferential growth along surfaces normal to the [110] and to the [121] directions has been found. The crystallographic characterization of the regrowth crystal indicates a good matching with the zone axis of the surrounding material although the crystalline seed Si (111) at the bottom interface is missing for the damage.

In phase change memory cells, the majority of heat is lost through the electrodes during the programming process that leads to significant drops in the performance of the memory device. In this investigation, we report on the thermal properties of thin film carbon nitride with modest electrical resistivity of 5-10 mW cm, low thermal conductivity of 1.47W/m/K, and low interfacial thermal conductance between carbon nitride and phase change material for length scales below 40 nm. The thermally insulating property of carbon nitride makes it a suitable thermal barrier, allowing for less heat loss during Joule heating within the memory unit. We compare carbon nitride thermal properties against commonly used electrodes and insulators such as tungsten and silicon nitride, respectively, to demonstrate the promise of carbon nitride as a potential material candidate for electrode applications in phase change memory devices.

Phase-change materials (PCMs) allow for an ultrafast and reversible transition between their amorphous (a) and crystalline (c) state. This structural transition results in a pronounced change in their optical and electrical properties. The strong resistivity contrast of several orders of magnitude has been employed in electronic and neuromorphic memories, while the unity-scale refractive index change Δn= |n_c- n_a| has been applied in rewriteable optical-data-storage memories. Additionally, Δn is the subject of research in various photonic applications.^[1]
For tunable photonics, PCMs are very desirable, but expensive and laborious to produce. Thus, colloidal PCMs that can be easily synthesized could provide broader access to this material class and allow for deposition on a variety of substrates, including prepatterned holes and vertical interconnect accesses. However, due to the size reduction from GeTe thin films to small nanoparticles (NPs), new effects can arise. Examples are the occurrence of localized surface plasmon resonances in c-GeTe NPs,^[2] and size dependence of the crystallization temperature T_C.^[3] Thickness dependent temperature scaling was also reported for very thin sputtered PCM films, and related to their bonding characteristics.^[4] Recent results indicate that the spatial confinement of PCMs into small NPs results in an increase of the optical band gap E_g, while maintaining the pronounced contrast between the band gaps for a- and c-PCMs E_g,a and E_g,c respectively.^[5]
Insights into small-size PCM NPs can promote the application of these materials in different fields, and add value to fundamental scientific questions, especially regarding the current debate on the unique chemical bonding in PCMs.^[6]
Despite the bottom-up synthesis of PCM NPs, structural confinement can be realized by sub-wavelength patterning. While laser pulses do not allow for switching on such small length scales, tip-induced crystallization can lead to features below 100 nm. The resulting so-called metasurfaces feature extremely strong field confinement and nearly 2π phase shift.^[7]

[1] Hail, C.; Michel, A. U.; Poulikakos, D. and Eghlidi, H. Adv. Opt. Mater. 7 (2019).
[2] Polking, M. J.; Jain, P. K.; Bekenstein, Y.; Banin, U.; Millo, O.; Ramesh, R. and Alivisatos, A. P. Phys. Rev. Lett. 111 (2013).
[3] Yarema, O.; Perevedentsev, A.; Ovuka, V.; Baade, P.; Volk, S.; Wood, V. and Yarema, M. Chem. Mater. 30 (2018).
[4] Simpson, R. E.; Krbal, M.; Fons, P.; Kolobov, A. V.; Tominaga, J.; Uruga, T. and Tanida, H. Nano Lett. 10 (2010).
[5] Michel, A.; Sousa, M.; Yarema, M.; Yarema, O.; Ovuka, V.; Lassaline, N.; Wood, V. and Norris, D. J. In preparation.
[6] Wuttig, M.; Deringer, V. L.; Gonze, X.; Bichara, C. and Raty, J.-Y. Adv. Mater. 30 (2018).
[7] Michel, A.; Meyer, S.; Lassaline, N.; Lightner, C. R.; Bisig, S.; Norris, D. J. and Chigrin, D. N. In preparation.

Here we demonstrate heterogeneous integration of phase-change materials GST and GSST for memory functionality in silicon photonic circuits. Using these materials we show a) an electro-optic WRITE and optical READ as a dot-product multiplier that can be used as a a non-volatile ‘weight in photonic neural networks (NN), and b) an all-optical WRITE and optical READ performing the nonlinear activation function ‘thresholding’ of the NN.
An all-optical (AO) nonlinear (NL) material forms the ultimate basis for a new generation of photonic NNs. Here we show the response and integrate such material-based nonlinearity into a photonic perceptron. Besides passive weighting functionalities, which saves static power with respect to an implementation with EOMs, we will use a nonlinear activation function (NL AF, ‘threshold’ of NN) based on temporary optical response of GST when optically pumped. For the implementation of the optical nonlinearity, instead of an electro-optic mediated nonlinearity, as proposed by our group and others, we propose to integrate a thin film of GST in an all-optical thresholder in a pump-probe configuration. In this configuration there is no conversion between light and electronics until read out, that could be pushed at the very end of the network. Thus, the inference delay of this network is ps-short. This GST film acts as an optical intensity modulator of the weighted-addition on a second probe source. The optical transitions which modulates the nonlinear function are corroborated by a recent break-through in GST mediated optical nonlinearity which allows for optical modulation of light without requiring the material to recrystallize. This physical modulation mechanism relies on a femtosecond-short excitation, which directly detaches electrons from resonantly bonded states resulting in a non-equilibrium, immediate decrease in the dielectric function and the saturation of the imaginary part of the dielectric function at the same value as the amorphous-state. This femtosecond pulse induced change in the optical properties is not the same as those dictated by thermal process, used for the weighting functions, and hence does not result in the change in crystallinity. Without the complete depopulation of the resonantly bonded state, we can achieve up to 13% reversible modulation of the dielectric function, more than 10x larger than that observed in silicon photo-switches. This allows for a response time of about 100 fs of these all-optical NLAF (thresholds) for pure optical NNs. The neuron is then interconnected to others in a feedforward fully connected NN of L layers (multi-layered perceptron). The input signal is encoded in the laser intensity. Each layer consists of a logic block which performs pointwise matrix multiplication (i) and element-wise NLAF (ii). The weighed addition is performed using a similar approach seen in. In this case, the weights are not fixed thermally which requires a great expense of power budget and insulation strategies, but it is fixed using phase change materials, which have been optically written (in a similar fashion as a rewritable DVD) and kept at their respective state without any dissipation of static power. The NLAF is also performed using GST films in a pump-probe configuration; the optical pump, from free space, induces the temporary phase transition and the material alters its complex refractive index in a volatile way. The effective mode index of the waveguide, in which the probe signal travels, is altered as nonlinear function of the available power after the weighted addition.

Chalcogenide phase change materials (PCM) are widely-used in optical data storage technologies. Also, they are promising to become the next-generation mainstream electronic non-volatile memory technolgy. These have been discussed and simulated in our prior work [1-4]. The operation of PCM-based devices hinge on the repeated rupture and formation of atomistic-scale chemical bonds to reversibly switch between crystalline and amorphous phases, which require atomistic-scale simulation methodology to achieve in-depth optimization. Nowadays, there are two categories of such methodologies -- density functional theory (DFT) and artificial neural network (ANN). DFT-based molecular dynamics (MD) is accurate but inefficient -- it is practically applicable to only hundres of atoms. ANN-based MD is efficient but inaccurate -- its maximum error of atomic force prediction can be over 10 eV/Angstrom.

In our recent work [5], we present a method based on deep neural network (DNN) to model PCM, which is both accurate and efficient. The DNN-based MD has a computational complexity of O(N), where N is the number of atoms. So, DNN-based MD is orders of magnitude faster than DFT, whose computational complexity is around O(N³). Furthermore, the maximum error of atomic force prediction of DNN-based MD is around 0.1 to 1 eV/Angstrom, which is much more accurate than ANN in the existing literature. In our another recent work [6], we present a transfer learning (TL) method, to handle the needs to simulate multiple combinations (different PCM stoichiometry, doping, etc.). Compared to the state-of-art methods in PCM literature, the proposed TL method can significantly reduce the amout of DFT data used to train the neural network, which is beneficial to enable high-throughput simulations.

Combining the accuracy and efficiency [5], and the trasferrability between different materials [6], we aim at establishing a general-purpose PCM optimization method based on high-throughput simulations.

[1] Jie Liu, et. al., "Impact of doping on bonding energy hierarchy and melting of phase change materials", Journal of Applied Physics, 124, 094503, 2018. DOI: 10.1063/1.5039831
[2] Jie Liu, "Microscopic Origin of Electron Transport Properties and Ultrascalability of Amorphous Phase Change Material Germanium Telluride, IEEE Transactions on Electron Devices, 2017. DOI: 10.1109/TED.2017.2685341
[3] Jie Liu"A multi-scale analysis of the crystallization of amorphous germanium telluride using ab initio simulations and classical crystallization theory", Journal of Applied Physics, 2014. DOI: 10.1063/1.4861721
[4] Jie Liu, et. al., "A multi-scale analysis of the impact of pressure on melting of crystalline phase change material germanium telluride", Applied Physics Letters, 2014. DOI: 10.1063/1.4901044
[5] P.H. Mo, M.C. Shi, Jie Liu, "Deep Neural Network for Accurate and Efficient Atomistic Modeling of Phase Change Memory", 2019 (under review)
[6] P.H. Mo, M.C. Shi, W.Z. Yao, Jie Liu, "Multitask Inductive Transfer Learning of Potential Energy Surfaces", 2019 (under review)

Phase-change materials (PCM) are technologically attractive materials for non-volatile memory, neuromorphic computing, optoelectronics, and recently, radiofrequency (RF) applications. Chalcogenide PCMs present a dissimilar resistance between the crystalline (low-resistivity) and amorphous phase (high-resistivity). The phase transition is thermally activated, and its kinetics can span orders of magnitude in time, down to sub-nanosecond timescale. Hence, proper probing of the thermal actuation is required to understand the fundamental material properties of PCMs, particularly the physics of melt-quench processes in nanoscale devices, interface vs. bulk effects, drift, and threshold voltage phenomena. Research efforts have been made to characterize PCM using a microthermal stage (MTS) [1]. The MTS structure used either a top Pt heater for the thermal actuation of a lateral Ge₂Sb₂Te₅ (GST) device or a Pt heater surrounding the GST. Thus, the MTS allowed for probing with thermal time scales on the order of microseconds or longer.
Four-terminal, in-line, indirectly heated phase-change switches (IPCS) have been proposed for high-performance RF applications, thanks to their state-of-the-art figure-of-merit (FOM)=1/(2πR_ONC_OFF) [2, 3]. A typical IPCS consists of two RF ports in-line with the PCM (i.e., the RF path), separated by a small gap, and two terminals for thermal actuation using an embedded heater. The heater runs transversely to the direction of the RF path, under the PCM. The heater and the RF path are separated by a thin electrically isolating film (e.g., Si₃N₄). GeTe has been so far the material of choice for RF applications due to its low resistivity in the crystalline phase and relatively large resistivity contrast with the amorphous state, but recently an Sb₇Te₃ IPCS was demonstrated with similar FOM and improved energy efficiency [3].
In this study, we report on the use of IPCS device structure as a platform for electro-thermal characterization of phase transition properties of PCM in nanoscale films at nanosecond temporal resolution. The isolation between the heater and the PCM layer electrically decouples the heating pulse and PCM probing. Thus, the phase transition, as well as temperature-dependent properties up to temperatures of ~1100 K, can be measured during heating pulse application with ns resolution. Furthermore, since the heater is buried underneath the PCM, our setup allows for 1) steady-state PCM temperature validation using scanning thermal microscopy (SThM) from the top surface, 2) separation between contribution of contact (interface) and bulk effects by varying overlap and underlap between the heater and the PCM contacts, and 3) test different materials of interest without significant change in the fabrication process. We use GeTe as a prototype material with known melting temperature, crystallization kinetics, and surface treatment for low ohmic contacts [4]. Measurements are carried out using RF probes for the heater and the PCM to reduce the overshoot and ringing of short-pulses and produce clean waveforms for nanosecond probing. The experimental results are compared to a detailed finite element electro-thermal model to extract key material properties. Overall, our platform can be used to uncover the intriguing kinetics of crystallization and amorphization in chalcogenides and other amorphous semiconductors.
[1] J. Lee et al., IEEE EDL. 32, 7 (2011).
[2] N. El-Hinnawy et al., IEEE EDL. 34, 10 (2013).
[3] E. Yalon et al., IEEE EDL. 40, 3 (2018).
[4] H. M. Aldosari et al., J. Appl. Phys. 122, 175302 (2017)

Phase change materials (PCMs) are well-known for their crystalline-to-amorphous transitions that are both quick and reversible, offering a large contrast in electrical and optical properties of the two phases. Responsible for enabling the development of rewritable CD and DVD technology, this class of materials has greatly contributed to the development of modern data storage, entertainment, and computing; and due to so many contributions, research continues to point to applications in which chalcogenide PCMs provide novel solid-state devices. One such PCM, germanium telluride (GeTe), has been the focus of numerous studies to develop next generation non-volatile memory cells, photonic devices, and radio frequency (RF) switches.
The geometry of GeTe-based devices for radio frequency switches and non-volatile memory technologies often places GeTe thin films in contact with metal thin films. Despite the potential effect of metal/GeTe reactions on device performance, few studies have addressed the reactivity between elemental metals and GeTe or systematically approached the thermal stability of GeTe with metals. In response to this need, reactivity was determined by calculating ternary phase diagrams of metal-Ge-Te systems and performing transmission electron microscopy (TEM) both after metal deposition and after the samples were annealed for 12 h at 200 ^oC. GeTe is thermodynamically favored to react with many metals at room temperature. Nine of the 24 studied metals are not reactive with GeTe (Au, Ir, Mo, Os, Re, Ru, Ta, W, and Zn), while 15 metals have a thermodynamic driving force to react with GeTe at room temperature (Ag, Al, Cd, Co, Cu, Fe, Hf, Mn, Ni, Pd, Pt, Rh, Sc, Ti, and Y). Most of the unreactive metals, except Au and Zn, are not in thermodynamic equilibrium with GeTe at room temperature. These metals are refractory, and the lack of reactivity is ascribed to kinetic limitations.¹
Solid-state reactions between contact metals and GeTe produce an unexpected trend between metal work function and metal/GeTe contact resistance (R_c), which is actually the opposite to what is projected by the well-known Schottky-Mott Law. For a p-type semiconductor like GeTe, high work function metals, like Ni and Pt, would be expected to provide the lowest R_c values. However, comparing all contact metals (Au,² Ni,³ Mo,⁴ Sn, Ti⁵, Pt, and Cr), Mo-based contacts (with a work function of only 4.60 eV) offered the lowest contact resistance. From cross-sectional TEM analysis, only Au- and Mo-based contacts did not react with GeTe.^2,4 Thus, the work function of the contact metal of the reactive contacts is no longer the deciding factor in setting R_c, and we explore explanations for this trend in our ongoing work.

1. K.A. Cooley and S. E. Mohney. J. Vac. Sci. Technol. A, in press.
2. H. M. Aldosari, K. A. Cooley, S.-Y. Yu, K. C. Kragh-Buetow, and S. E. Mohney. Thin Solid Films 621, 145 (2017).
3. H. M. Aldosari, H. Simchi, Z. Ding, K. A. Cooley, S.-Y. Yu, and S. E. Mohney. ACS Appl. Mater. Interfaces 8, 34802 (2016).
4. H. M. Aldosari*, K. A. Cooley*, S.-Y. Yu, H. Simchi, and S. E. Mohney. J. Appl. Phys. 122, 175302 (2017).
*Both co-authors contributed equally to the work.
5. H. Simchi, K. A. Cooley, Z. Ding, A. Molina, and S. E. Mohney. ACS Appl. Mater. Interfaces 10, 16623 (2018).

Phase-Change Materials (PCMs) are the active switching element in novel electronic memory devices. Their data storage is based on a crystallization and glass formation mechanism that offers (i) fast crystallization over a wide range of elevated temperatures and (ii) glass formation at ambient conditions. Fundamentally, these regimes are characterized by a low and high activation energy of viscosity, respectively. The transition between these regimes has been previously associated with a fragile-to-strong crossover[1] and/or with the glass transition[2], but its microscopic origin has been elusive - mostly due to the experimental limitation imposed by the fast crystallization of this disordered state. Recent measurements using femtosecond X-ray diffraction at an X-ray free electron laser (XFEL) resolved the atomic structure during this transition[3]. An optical laser pulse was employed to melt the PCM, whose structure was probed at different time delays after the excitation, i.e., during the quenching process. The resulting data show that the increase of kinetic activation energy coincides with the onset of a Peierls distortion in the PCMs Ag₄In₃Sb₆₇Te₂₆ (AIST) and Ge₁₅Sb₈₅. Based on the weak dependence of this onset temperature on the cooling-rate, we attribute this onset to a liquid-liquid phase transition. Ab-initio computer simulations confirm this trend in the atomic structure and provide further insight into the underlying mechanism: The distortion breaks the local symmetry, opens an electronic band gap and localizes charges, giving the bonds more covalent character, which is commonly associated with a decrease in the flexibility of bond angles. This insight reveals a relation between the structure and kinetics of PCMs. In this contribution, we will also discuss similar measurements on the most common PCM Ge₂Sb₂Te₅ and contrast its behavior to a similar but distinct transition mechanism observed in supercooled liquid Ge.
[1] J. Orava, A. L. Greer, B. Gholipour, D. W. Hewak, and C. E. Smith, “Characterization of supercooled liquid Ge₂Sb₂Te₅ and its crystallization by ultrafast-heating calorimetry,” Nature Mater., vol. 11, no. 4, pp. 279–283, Mar. 2012.
[2] M. Salinga et al., “Measurement of crystal growth velocity in a melt-quenched phase-change material.,” Nature Commun., vol. 4, p. 2371, 2013.
[3] P. Zalden et al., “Femtosecond x-ray diffraction reveals a liquid–liquid phase transition in phase-change materials,” Science, vol. 364, no. 6445, pp. 1062–1067, 2019.

Relaxation processes are decisive for many relevant physical properties of amorphous materials. For amorphous phase-change materials (PCMs) employed in non-volatile memories, relaxation processes are, however, difficult to characterize due to the lack of bulk samples. Here, instead of bulk samples, we use powder mechanical spectroscopy for powder samples to detect the prominent excess wings – a characteristic feature of β-relaxations – in a series of amorphous PCMs at temperatures below the glass transition. By contrast, β-relaxations are vanishingly small in amorphous chalcogenides of similar composition, which lack the characteristic features of phase-change materials. This conclusion is corroborated upon crossing the border from PCMs to non-PCMs, where β-relaxations drop significantly. Such a distinction implies that amorphous PCMs belong to a special kind of covalent glasses whose locally fast atomic motions are preserved even below the glass transitions. These findings also suggest a correlation between β-relaxation and crystallization kinetics of PCMs, which may have technological implications for phase-change memory functionalities.

In order to model phase change memory (PCM) cells, dynamic materials models have to be coupled with charge and heat transport models [1]-[7]. Since PCM cells experience high temperatures, very large thermal gradients (~ 50 K/nm) and high current densities during their normal reset and set operations, thermoelectric contributions are significant for both charge and heat transport in these cells. Furthermore, some of the assumptions made for large scale thermoelectric devices do not hold under these extreme conditions.
Charge transport in PCM cells can be modeled using a drift- diffusion approximation. However, electronic material properties such as carrier concentrations, mobilities and diffusivities are not well characterized for phase change materials. In this study, we construct (i) a drift-diffusion model using thermoelectric characteristics of the material, measured in a large temperature range, and (ii) an electrothermal heat transport model that does not assume mild temperature gradients only.
Our simulation results show that both Thomson heat (thermoelectric effects in the bulk materials) and Peltier heat (thermoelectric effects at the contact regions) play a significant role, and give rise to significantly asymmetric thermal profiles in symmetric PCM cells.

Acknowledgments: This work is partially supported by NSF under award DMR-1710468.

References:
[1] Z. Woods and A. Gokirmak, IEEE Trans. Electron Devices 64, 4466 (2017).
[2] Z. Woods et al., IEEE Trans. Electron Devices 64, 4472 (2017).
[3] J. Scoggin et al., Applied Physics Letters 112, 193502 (2018).
[4] J. Scoggin et al., Applied Physics Letters 114, 043502 (2019).
[5] A. Faraclas et al., IEEE Trans. Electron Devices 61, 372 (2014)
[6] A. Faraclas et al., IEEE Trans. Electron Devices 32, 1131 (2011)
[7] G. Bakan et al., Scientific reports 3, 2724 (2013)

Phase change materials including Ge-Sb-Te (GST) alloys provide useful changes in the electronic and optical properties thanks to the reversible a structural phase transition between amorphous and crystalline phases. The phase change materials have been used for commercially-available optical and electrical memory technologies by exploiting the phase change property. In this context, GST is expected to be utilized also for terahertz (THz) wave engineering since the frequency of THz wave ranges between infrared light and radio wave. So far, we revealed that a THz pulse excitation is capable of structural control thorough unique nonlinear field-induced phenomena [1,2]. In addition, we performed a THz emission spectroscopy as well as a THz time-domain spectroscopy and found that these spectroscopic measurements can be used for evaluation of the film quality and electronic characteristics of the GST. Based on the result of THz time-domain spectroscopy, we propose that the reversible and nonvolatile changes in the THz properties of GST are promising for a variety of applications e.g. THz photonics and THz amplitude modulator [3].

[1] K. Makino et al, Sci. Rep. 8, 2914 (2018).
[2] Y. Sanari et al, Phys. Rev. Lett. 121, 165702 (2018)
[3] K. Makino et al, J. Mater. Chem. C 7, 8209 (2019).

Phase change memory (PCM) technology utilizes the large resistivity contrast between the crystalline and amorphous phases of chalcogenide materials. However, the resistance of the amorphous phase deviates from its originally programmed value over time, which is commonly referred to as resistance drift. This resistance drift may lead to inaccurate read operations and thus poses a major challenge to the realization of multilevel cell (MLC) PCM.

We have performed electrical characterization of Ge₂Sb₂Te₅ (GST) line cells at low temperatures expecting to see very small or no drift since resistance drift has been understood to be caused by structural relaxation which is a thermally activated defect annihilation process [1]-[3]. However, we observed resistance drift at cryogenic temperatures with drift coefficients comparable to the room temperature value of ~0.1 [4]. Furthermore, we observed a significant acceleration in resistance drift with high electric field stress in 80 K - 200 K temperature range. Resistance of the cells stabilizes after the electrical stress allowing stable field and temperature-dependent electrical characterization.

These results suggest that charge traps and charge injection to these traps play a significant role in the resistance drift of amorphous phase change materials.


Acknowledgments: This work is partially supported by NSF under award ECCS 1711626.

References:
[1] M. Boniardi and D. Ielmini, “Physical origin of the resistance drift exponent in amorphous phase change materials,” Appl. Phys. Lett., vol. 98, no. 24, pp. 1–4, 2011.
[2] A. Sebastian, D. Krebs, M. Le Gallo, H. Pozidis, and E. Eleftheriou, “A collective relaxation model for resistance drift in phase change memory cells,” IEEE Int. Reliab. Phys. Symp. Proc., vol. 2015-May, no. 1, pp. MY51–MY56, 2015.
[3] D. Ielmini, D. Sharma, S. Lavizzari, and A. L. Lacaita, “Physical mechanism and temperature acceleration of relaxation effects in phase-change memory cells,” IEEE Int. Reliab. Phys. Symp. Proc., pp. 597–603, 2008.
[4] R. S. Khan, S. Muneer, N. Noor, H. Silva, and A. Gokirmak, “Evidence of Charge Trapping Giving Rise to Resistance Drift of Metastable Amorphous Ge₂Sb₂Te₅,” in APS March Meeting, 2019.

The contact resistance between TiN and GST 225 was measured using the transfer length method (TLM) and the end resistance method for the crystalline and amorphous phases of GST 225. We found a 3000x increase in contact resistance when GST 225 over the TiN electrodes changed from crystalline to amorphous. When considering common phase change material (PCM) memory cell geometries, our measurements suggests that the contact resistance is a large component of the change in the resistance when the device is switched between the crystalline and the amorphous phases. In other words, the increase of the device resistance when the device is reset is substantially due to the increase in the contact resistance and not solely due the change in the bulk resistance of the PCM.

The functioning of PCM materials in the context of involatile digital memory technology involves a complex interplay of liquid, glass and crystal phase behaviors. Here we focus attention on the peculiarities of the liquid state as judged by the temperatures at which transitions from metallic to semiconducting states occur, relative to melting points. In some cases like As₂Se₃ and As₂Te₃ this transition occurs far above the melting point. In others, like pure Te, it is observed to occur quite deep in the supercooled liquid state. Data are sparse for elements and binary compounds, but we can include other compositions by introduction of a metallicity parameter, M_P, which is the inverse of the composition-averaged Pauling electronegativity, χPi , _,MP = 1/(x1●χP1 + x2●χP2 +..+xi●χP3). When we plot the temperatures of at which the M-SC transition occurs (scaled by the melting points), against M_P, we find, by extrapolation, that all of the preferred PCM compositions have their M-SC transitions at similar depth in the supercooled liquid state, where the crystallization rate is exceptionally high. The implications, and applications, of this finding will be discussed.

The large resistivity contrast between the highly resistive amorphous phase and low resistive crystalline phase of phase change memory (PCM) provides opportunities for multi bit per cell memory and neuromorphic applications. However, the drift of resistance of amorphous phase limits the potential of PCM as multiple intermediate states can overlap over time. Although resistance drift is generally attributed to structural relaxation [1], we have observed significant resistance drift in Ge₂Sb₂Te₅ line cells at cryogenic temperatures and also being affected by optical illumination, indicating charge trapping [2]. A detailed understanding of the mechanisms behind resistance drift may allow for mitigating techniques. In this work, we study the effect of optical illumination between 125 K and 300 K and also the effect of different optical illumination conditions on the drift coefficient.
References:
[1] D. Ielmini, S. Lavizzari, D. Sharma, and A. L. Lacaita, “Physical interpretation, modeling and impact on phase change memory (PCM) reliability of resistance drift due to chalcogenide structural relaxation,” Tech. Dig. - Int. Electron Devices Meet. IEDM, pp. 939–942, 2007.
[2] R. S. Khan, S. Muneer, N. Noor, H. Silva, and A. Gokirmak, “Evidence of Charge Trapping Giving Rise to Resistance Drift of Metastable Amorphous Ge₂Sb₂Te₅,” in APS March Meeting, 2019.

As one of the most promising candidates for developing new neuromorphic architectures for non-von Neumann computing and information storage, Ge₂Sb₂Te₅(GST225) in photonic devices shows non-volatile, fast, and multi-step changes between the amorphous state and the meta-stable states. In order to search for other composition ratios with desired properties, e.g. large figure of merit, lower switching energy, etc., we have systematically investigated a broad composition range of the Ge-Sb-Te (GST) system. The GST combinatorial spread was fabricated by co-sputtering Ge, Sb and Te targets and was confirmed by wavelength dispersive spectroscopy that the composition variation across the Si wafer covers most part of the GST phase diagram. X-ray diffraction measurements (synchrotron radiation), Raman spectroscopy and resistance mapping clearly show evolution of the structure and phase-change temperature in the GST system. Refractive index and extinction coefficient were calculated at each composition spot using a scanning ellipsometer: some composition ratios were found to have a small extinction coefficient in the amorphous state, a high phase-change temperature, and a large figure of merit. Then, the nano-size photonic devices fabricated by these ratios were measured and showed good reversibility (more than 40000 cycles) and multi-level symmetric switching (at least 15 levels), suggesting that such compositions can be promising for photonic devices for neuromorphic control. Further, we confirm that compositions with promising properties are near the boundary between the Sb-Te and the Ge-Sb-Te phases, indicating that co-existence of the phases at microstructural level may be playing a role in giving rise to their properties. This work is funded by an ONR MURI (Award No. N00014-17-1-2661).

Phase change materials (PCMs) are an extremely attractive material platform for the realization of multi-functional and reconfigurable meta-optics. For example, PCMs can be exploited to synthesize metasurfaces and metamaterials to enable a variety of tunable devices such as beam-steerers, optical shutters, spectral filters, and adaptive focal length lenses [1]–[5]. However, the expanded degrees of design freedom that PCMs offer can make direct device design intractable for all but the most experienced engineers. This challenge is best overcome through the use of advanced inverse-design tools and state-of-the-art optimization algorithms. To this end, a number of successful meta-device inverse-design approaches have been demonstrated in the literature including those based on topology optimization, deep learning, and global optimization [6]. While each method has its pros and cons, one method stands out as an ideal candidate for reconfigurable meta-optic design: multi-objective optimization. In contrast to ubiquitous single-objective optimization algorithms, which require users to combine multiple goals into a single cost function usually via a weighted sum, true multi-objective optimization (MOO) algorithms allow designers to minimize multiple competing objectives simultaneously without the need for a priori information on how best to weight a single cost function [7], [8]. Thus, MOO algorithms are perfectly suited for reconfigurable meta-optic design as each independent functionality can be assigned a unique cost function and optimized. Moreover, MOO algorithms provide the user with a collection of designs called the Pareto set that can be analyzed to determine the inherent tradeoffs between competing design objectives. In our presentation, we will introduce an efficient multi-objective optimization enabled design framework for the generation of multi-functional unit cells based on phase change materials. Additionally, several reconfigurable meta-optic design examples will be presented, and future research directions discussed.

References
[1] A. Karvounis, B. Gholipour, K. F. MacDonald, and N. I. Zheludev, “All-dielectric phase-change reconfigurable metasurface,” Applied Physics Letters, vol. 109, no. 5, p. 051103, Aug. 2016.
[2] Q. Wang, E. T. F. Rogers, B. Gholipour, C.-M. Wang, G. Yuan, J. Teng, and N. I. Zheludev, “Optically reconfigurable metasurfaces and photonic devices based on phase change materials,” Nature Photonics, vol. 10, no. 1, pp. 60–65, Jan. 2016.
[3] L. Liu, L. Kang, T. S. Mayer, and D. H. Werner, “Hybrid metamaterials for electrically triggered multifunctional control,” Nature Communications, vol. 7, p. 13236, Oct. 2016.
[4] L. Kang, R. P. Jenkins, and D. H. Werner, “Recent progress in active optical metasurfaces,” Advanced Optical Materials, vol. 7, no. 14, p. 1801813, 2019.
[5] A. V. Pogrebnyakov, J. A. Bossard, J. P. Turpin, J. D. Musgraves, H. J. Shin, C. Rivero-Baleine, N. Podraza, K. A. Richardson, D. H. Werner, and T. S. Mayer, “Reconfigurable near-IR metasurface based on Ge₂Sb₂Te₅ phase-change material,” Opt. Mater. Express, vol. 8, no. 8, pp. 2264–2275, Aug. 2018.
[6] S. D. Campbell, D. Sell, R. P. Jenkins, E. B. Whiting, J. A. Fan, and D. H. Werner, “Review of numerical optimization techniques for meta-device design [Invited],” Opt. Mater. Express, vol. 9, no. 4, pp. 1842–1863, Apr. 2019.
[7] K. Deb, Multi-objective Optimization Using Evolutionary Algorithms, Paperback edition. Chichester: Wiley, 2008.
[8] J. Nagar and D. H. Werner, “A comparison of three uniquely different state of the art and two classical multiobjective optimization algorithms as applied to electromagnetics,” IEEE Transactions on Antennas and Propagation, vol. 65, no. 3, pp. 1267–1280, Mar. 2017.

The mechanical properties, Hardness and Young module, of crystalline micrometer stripes of Ge₂Sb₂Te₅ and of the surrounding 2.2 µm thick amorphous film have been measured simultaneously. Crystallization was performed by a Continuous Wave laser irradiation process in the range 1.25-3.2 mW. The density changes between the amorphous and the crystalline phase causes a relevant stress in the processed material. Usually the stress is measured by the wafer curvature. This methodology doesn’t allow a detailed investigation of the local mechanical modifications when the two phases amorphous and crystalline are contiguous, as in an optical or electrical storage device. The use of micrometer patterned structures and nano-indentation measurements allow us to determine the mechanical behavior of the two contiguous GST phases. The measured values of the mechanical property are more dispersed for the crystallized material than those for the contiguous amorphous phase. The Young module and the hardness are 51±8 GPa and 2880±450 MPa for the crystallized and 33±4 GPa and 1989±300 MPa for the amorphous phase respectively. Microcracks are also seen, together with fracture lines. The occurrence of delamination at laser power densities above 1.70 mW indicates that for these crystallized thicknesses the maximum tensile strength in the amorphous and crystalline phases was reached and exceeded. The phase transition has also been investigated by several analytical tools (TEM, SEM, Raman and XRD diffraction). A simulation code was adopted to compute the temperature-time-space profile as a function of the laser power. These results were used to compare the measured crystallized thickness to that calculated.

Transition metal dichalcogenides (TMDs) exist in various crystal structures with semiconducting, semi-metallic, and metallic properties. The dynamic control of these phases is of immediate interest for next generation electronics such as phase change memories and TMDS have the potential to outperform current phase-change materials such as GeSbTe on energy efficiency while also being fundamentally more scalable. Of the binary Mo and W-based TMDs, MoTe₂ is attractive for electronic applications because it has the lowest energy difference (40 meV) between the semiconducting (2H) and semi-metallic (1T’) phases, allowing for MoTe₂ phase change by electrostatic doping.

Here we report phase change between the 2H and 1T’ polymorphs of MoTe₂ in thicknesses ranging from the monolayer case to effective bulk (73nm) using an ionic liquid electrolyte at room temperature and in air. We find consistent evidence of a partially reversible 2H-1T’ transition using in-situ Raman spectroscopy where the phase change occurs in the top-most layers of the MoTe₂ flake. We empirically find that the transition voltage of the phase change increases with thickness which provides a direct avenue to transition-voltage control for phase change devices. We also show evidence of electrochemical activity during the gating process by observation of Te metal deposition. This finding suggests the formation of Te vacancies which have been reported to lower the energy difference between the 2H and 1T’ phase, potentially aiding the phase change process. This provides another avenue for transition-voltage control through defect engineering but also emphasizes the need for better understanding of electrolyte gating and chalcogenide vacancy formation. Our discovery that the phase change can be achieved on the surface layer of bulk materials reveals that this electrochemical mechanism does not require isolation of a single layer and the phase change effect may be more broadly applicable than previously thought. We will also discuss recent work on implementation of a solid state device with the use of ionic gels and porous dielectrics.

Intentionally introducing impurity atoms into a semiconductor has been a common method for modifying material properties that are critical to many electronic and optoelectronic devices. For GeTe, a phase change material (PCM), there has been interest in using metal incorporation for this approach to materials engineering. GeTe thin films doped and/or alloyed with metals have been reported to exhibit improved device performance, like improved crystallization speed, thermal stability, and power consumption. However, these films have often been fabricated using non-equilibrium methods with high metal concentrations (>10 at. %). Since switching between the low-resistance crystalline and high-resistance amorphous states requires a heating cycle, the stability of metal-incorporated GeTe (GeTe:metal) films is critical to practical implementation of these materials in electronic devices. Understanding the effect of metal-incorporation in GeTe films provides valuable insights for engineering future PCM devices, both in terms of doping and discovery of ternary PCMs.

In this work, we present first principles calculations concerning the stability of GeTe doped with select metals (Cu, Fe, Mn, Mo, and Ti), as well as the effect of increasing dopant atom concentration (2-6 at.%) on the crystal structure and electronic properties of GeTe. From density-functional theory calculations of the formation energy of the ternary solid (GeTe doped with 2-6 at. % of Cu²⁺, Fe²⁺, Mn²⁺, Mo²⁺, and Ti²⁺), all metals favored substitution into the Ge site over the Te site. The formation energy of the GeTe:metal structure increases (or becomes less stable) with increasing metal incorporation for all metals except Mn. Certain metals (Ti, Mo) clearly favor dopant atom clustering. In addition, different metal dopants have a varied distortion of the GeTe crystal structure and projected density of states. Computational results are compared to observed solubility trends in cross-sectional transmission electron microscopy (TEM) studies of metal/GeTe thin film systems (Cu, Mn, Mo, and Ti)¹ and TEM, spectroscopic ellipsometry, and transport data from newly characterized co-sputtered GeTe:Fe films.

1. K. A. Cooley and S. E. Mohney. J. Vac. Sci. Technol. A, in press.

We present our results on design, fabrication and characterization of a tunable metalens based on optical phase change materials (O-PCMs). We first introduce a phase change meta-atom design concept enabling switching of metasurface devices between two arbitrary optical states. We then implement this principle to demonstrate a bifocal switchable metalens made of the low-loss non-volatile O-PCM Ge2SbsSe4Te1 (GSST). The metalens can achieve large change in focal length, from 1.5 mm to 2 mm, which is caused by drastic change in meta-atoms refractive index. Additionally, the metalens exhibits focusing efficiencies above 20% in both amorphous and crystalline states. We further experimentally demonstrated aberration-free imaging using the lens and achieved diffraction-limited resolution in both states. This result represents the first experimental demonstration of a non-mechanical active metalens with diffraction-limited imaging capability.

Phase-change memory is the most promising candidate for the next generation memory technology. It utilizes the large property contrast between the amorphous and crystalline phases of phase change materials (PCMs), which switches to each other within nanoseconds. In the past decades, a lot of efforts have been devoted to explore how the structure defines the particular properties of PCMs, and it is discovered that the structure of the amorphous phase plays the key role in this case. For example, the fragility of the glass determines the crystallization speed (the programming speed in memory); the glass aging results in the resistance drift in the devices; the stability of the glass (data retention) is determined by the covalent bonds in this material. Since the amorphous phase is absent of long-range order and defects such as dislocation and grain boundary, the physical properties are usually determined by the short- and medium-range orders, and hence we call it the materials “gene” of glass.

We discovered in the our research that the local structure of PCMs could be described as “octahedral motifs”, which can be easily modified by adding different dopants. For example, the carbon, which forms tetrahedral clusters, can increases the stability of the glass to elongate the life of memory devices; the addition of Yi and Sc stabilizes the nuclei in the glass, remarkably accelerating the crystallization speed. The discovery of new PCMs enabled by the Materials Genome Engineering paves the way for the design of high-density phase change memory.

The nonvolatile reconfiguration capability of optical phase change materials (O-PCMs) makes them highly attractive for tunable dielectric metasurfaces. However, conventional switching methods based on either laser pulsing or electrical-current triggered transition require raster-scanned writing and dedicated off-chip switching instruments, therefore incompatible with large-scale integration. Here we report the design, fabrication and characterization of a scalable electrical switching platform for tunable dielectric metasurfaces based on O-PCMs.
We first introduce Ge₂Sb₂Se₄Te₁ (GSST), a novel O-PCMs we developed recently that is specifically targeted for high-performance photonic applications. We demonstrate that GSST exhibits low-loss at both phases over a broad wavelength range, while still possessing a huge optical contrast (Δn = 1.7). Its improved amorphous phase stability also gives rise to a larger critical thickness for complete switching. Comparing to a critical thickness of only 100 nm or less for traditional PCMs such as Ge₂Sb₂Te₅, GSST is a preferred material for providing desired phase coverage for metasurface applications.
Geometrically optimized on-chip micro-heaters are implemented to achieve uniform temperature distribution within the metasurface region. This eliminates damage to the metasurface due to overheating. Devices of various heater dimensions, from 30 to 200 microns, have been fabricated. The fabricated devices are wire-bonded onto a custom printed circuit board carrier to enable in-situ device characterizations.
We then implement the switching platform to demonstrate a tunable metasurface spectral filter. Multi-cycle bi-stable reversible switching is realized. Thanks to the huge optical contrast of GSST, the device provides a high reflectance modulation of 40%. In addition to bi-stable switching, broadband quasi-continuous resonance tuning over half an octave is also demonstrated. This is accomplished through controlling the fraction of crystallization of O-PCMs by varying pulse parameters.
We further demonstrate a nonvolatile beam deflection control method with a tunable metasurface reflectarray. The fabricated device deflects incident laser beam from +1 to 0 diffraction order upon phase transition, and it exhibits significant switching contrast between its two phases.
This result presents, to the best of our knowledge, the first experimental demonstration of a scalable switching platform for dielectric metasurfaces based on O-PCMs. We further demonstrate functional metasurface devices with bistable and quasi-continuous tuning capabilities. We believe that this platform will facilitate the development of numerous emerging applications capitalizing on its scalability and nonvolatile nature.

DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited.
This material is based upon work supported by the Under Secretary of Defense for Research and Engineering
under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations
expressed in this material are those of the author(s) and do not necessarily reflect the views of the Under
Secretary of Defense for Research and Engineering.

Fast and reversible phase transitions in chalcogenide phase-change materials (PCMs), in particular, Ge-Sb-Te compounds, are not only of fundamental interests but also make PCMs based random access memory a leading candidate for nonvolatile memory, and even neuromorphic computing devices. For instance, a recently designed phase-change heterostructure (PCH), which consists of alternately stacked phase-change and confinement nanolayers to suppress the noise and drift, has attracted increasing attention. However, to RESET the memory cell, crystalline PCMs has to undergo phase transitions first to a liquid state and then to an amorphous state, corresponding to an abrupt change in electrical resistance. In this work, we demonstrate a progressive amorphization process in GeSb2Te4 thin films under electron beam irradiation on a transmission electron microscope (TEM). Melting is shown to be completely absent by the in situ TEM experiments. The progressive amorphization process resembles closely the cumulative crystallization process that accompanies a continuous change in electrical resistance. The same nonthermal amorphization process was also verified in the phase-change nanolayers of PCH. The displacement forces induced by knock-on collision effect of E-beams drives this nonthermal amorphization process, thus our work suggests that if displacement forces can be implemented properly, it should be possible to emulate symmetric neuronal dynamics by using PCMs.
Key words: phase-change materials, phase change heterostructure, phase transition, electron beam irradiation, in situ TEM

In the reset process of phase change memories, the active material is brought rapidly above the melting temperature by Joule heating. Atomic migration in the liquid state due to the high electric field can lead to alloy demixing and eventually to device failure. The electromigration force F responsible for ionic migration is proportional to the electric field via the effective charge Z^* (F= Z^* E). Information on the effective charges is thus of great relevance for the electrothermal modeling of the device. However, experimental estimates of the effective charges in the liquid state suffer from large uncertainties. A possible route to estimate Z^* relies, in fact, on the modeling of the concentration profile of the different species in highly cycled memory cells which, however, depends on several mostly unknown parameters that have to be plug into phenomenological transport equations [1].

In this work, we show that a direct first principles calculation of the effective charges in metallic liquids is possible by computing the atomic forces in the presence of both an electric field and an electronic current within the Non-Equilibrium Green Function scheme implemented in the code Smeagol [2]. We will present results on the calculation of the effective charges, including the wind force, for the GeTe and Sb₂Te₃ phase change compounds in the liquid state.

Another feature of the liquid state that affects device operations is the opening of an electronic gap upon cooling from the melting point to the glass transition temperature during reset. The temperature at which the metallic liquid turns into a semiconductor is another important parameter for the electrothermal modeling of the device. By means of density functional molecular dynamics and the use of a hybrid exchange and correlation functional, we have estimated the temperature of gap opening in the supercooled liquid phase of GeTe and Ge₂Sb₂Te₅ compounds.

[1] L. Crespi et al., IEEE 7th International Memory Workshop (IWM), p. 25 (2015).
[2] R. Zhang, I. Rungger, S. Sanvito and S. Hou, Phys. Rev. B 84, 085445 (2011).

Fast and reversible phase transitions in chalcogenide phase-change materials (PCMs) play a key role in next-generation memory and computing chips. The Ge-Sb-Te compounds, such as Ge₂Sb₂Te₅ and GeSb₂Te₄, are under active investigations, not only because they serve as the key element in commercial products, but also because they provide a rich platform for fundamental research in materials science.

Disorder-driven metal insulator transition has been reported in crystalline Ge-Sb-Te phase-change materials (PCMs), where the high concentration and statistical distribution of atomic vacancies were identified as the key factor in shaping the localization properties of electrons and, thus, the electrical transport. Vacancy ordering has been consistently observed in crystalline Ge-Sb-Te thin films upon thermal annealing, triggering a structural transition from a cubic rocksalt structure to a layered hexagonal structure and an insulating to metallic transition.

In this work, we demonstrate an opposite vacancy disordering process upon extensive electron beam irradiation, which drives the reverse transition from the stable hexagonal phase to the metastable cubic phase. The combined in situ transmission electron microscopy experiments and density functional theory nudged elastic band calculations reveal three transition stages, including (I) the vacancy diffusion in the hexagonal phase, (II) the change in atomic stacking, and (III) the vanishing of vacancy-rich planes. The driving force of the vacancy disordering process is attributed to the kinetic knock-on collision effects of the high-energy focused electron beams, which prevail over the heating effects.

Our electron irradiation experiment not only provides an alternative approach to tune the vacancy distribution that is crucial for the disorder-driven metal-insulator transition of GST crystals, but also makes an instruction for the TEM measurement of structural details of a particular solid-state phase of PCMs that the three critical TEM parameters, namely, accelerating voltage, beam intensity and recording time, should be reduced as much as possible to avoid structural transitions during the measurement.

In the reset process of phase change memories, the active material is brought rapidly above the melting temperature by Joule heating. Atomic migration in the liquid state due to the high electric field can lead to alloy demixing and eventually to device failure. The electromigration force F responsible for ionic migration is proportional to the electric field via the effective charge Z^* (F= Z^* E). Information on the effective charges is thus of great relevance for the electrothermal modeling of the device. However, experimental estimates of the effective charges in the liquid state suffer from large uncertainties. A possible route to estimate Z^* relies, in fact, on the modeling of the concentration profile of the different species in highly cycled memory cells which, however, depends on several mostly unknown parameters that have to be plug into phenomenological transport equations [1].

In this work, we show that a direct first principles calculation of the effective charges in metallic liquids is possible by computing the atomic forces in the presence of both an electric field and an electronic current within the Non-Equilibrium Green Function scheme implemented in the code Smeagol [2]. We will present results on the calculation of the effective charges, including the wind force, for the GeTe and Sb₂Te₃ phase change compounds in the liquid state.

Another feature of the liquid state that affects device operations is the opening of an electronic gap upon cooling from the melting point to the glass transition temperature during reset. The temperature at which the metallic liquid turns into a semiconductor is another important parameter for the electrothermal modeling of the device. By means of density functional molecular dynamics and the use of a hybrid exchange and correlation functional, we have estimated the temperature of gap opening in the supercooled liquid phase of GeTe and Ge₂Sb₂Te₅ compounds.

[1] L. Crespi et al., IEEE 7th International Memory Workshop (IWM), p. 25 (2015).
[2] R. Zhang, I. Rungger, S. Sanvito and S. Hou, Phys. Rev. B 84, 085445 (2011).

It has been a long-time dream of mankind to design materials with tailored properties. In recent years, the focus of our work has been the design of phase change materials for applications in data storage. In this application, the remarkable property portfolio of phase change materials (PCMs) is employed, which includes the ability to rapidly switch between the amorphous and crystalline state. Surprisingly, in PCMs both states differ significantly in their properties. This material combination makes them very attractive for data storage applications in rewriteable optical data storage, where the pronounced difference of optical properties between the amorphous and crystalline state is employed. This unconventional class of materials is also the basis of a storage concept to replace flash memory. Today’s talk will discuss the unique material properties, which characterize phase change materials. In particular, it will be shown that only a well-defined group of materials utilizes a unique bonding mechanism (‘Bond No. 6’), which can explain many of the characteristic features of crystalline phase change materials. Different pieces of evidence for the existence of this novel bonding mechanism, which we have coined metavalent bonding, will be presented. In particular, we will present a novel map, which separates the known strong bonding mechanisms of metallic, ionic and covalent bonding, which provides further evidence that metavalent bonding is a novel and fundamental bonding mechanism. This insight is subsequently employed to design phase change materials as well as thermoelectric materials.

Fast and reversible phase transitions in chalcogenide phase-change materials (PCMs) play a key role in next-generation memory and computing chips. The Ge-Sb-Te compounds, such as Ge₂Sb₂Te₅ and GeSb₂Te₄, are under active investigations, not only because they serve as the key element in commercial products, but also because they provide a rich platform for fundamental research in materials science.

Disorder-driven metal insulator transition has been reported in crystalline Ge-Sb-Te phase-change materials (PCMs), where the high concentration and statistical distribution of atomic vacancies were identified as the key factor in shaping the localization properties of electrons and, thus, the electrical transport. Vacancy ordering has been consistently observed in crystalline Ge-Sb-Te thin films upon thermal annealing, triggering a structural transition from a cubic rocksalt structure to a layered hexagonal structure and an insulating to metallic transition.

In this work, we demonstrate an opposite vacancy disordering process upon extensive electron beam irradiation, which drives the reverse transition from the stable hexagonal phase to the metastable cubic phase. The combined in situ transmission electron microscopy experiments and density functional theory nudged elastic band calculations reveal three transition stages, including (I) the vacancy diffusion in the hexagonal phase, (II) the change in atomic stacking, and (III) the vanishing of vacancy-rich planes. The driving force of the vacancy disordering process is attributed to the kinetic knock-on collision effects of the high-energy focused electron beams, which prevail over the heating effects.

Our electron irradiation experiment not only provides an alternative approach to tune the vacancy distribution that is crucial for the disorder-driven metal-insulator transition of GST crystals, but also makes an instruction for the TEM measurement of structural details of a particular solid-state phase of PCMs that the three critical TEM parameters, namely, accelerating voltage, beam intensity and recording time, should be reduced as much as possible to avoid structural transitions during the measurement.

Previously, we have reported on a novel class of chalcogenide phase change materials, based on Ge-Sb-Se-Te (GSST) alloys, with excellent infrared transparency from 1.5 to > 15 micron wavelengths. Increasing Se substitution leads to a progressive lowering of extinction coefficient. A material in this family with a particular useful stoichiometry, Ge₂Sb₂Se₄Te₁ (GSS4T1), possesses broadband transparency through the infrared range while maintaining a large index contrast between its amorphous and crystalline phases. Defining a figure of merit as Δn/k (where n,k are the real, imaginary parts of the refractive index), our materials shows >100x figure of merit (FOM) improvement over conventional GST throughout the infrared range.

In this talk, we describe novel applications enabled by reversible switching of low loss GSS4T1 material. In the first application, we exploit the exceptional FOM of the Se-doped PCM to realize a nonvolatile photonic switch with unprecedented high performance. Our device, based on a racetrack SiN resonator with a PCM gate, exhibits a large switching contrast ratio of 42 dB and a low insertion loss of < 0.5 dB, far outperforming previous nonvolatile switches as well as devices based on the classical GST-225 material with a similar configuration.

Additional novel applications are enabled by our development of wafer-scale pixelated electrical switchable PCM devices. The pixelated devices are fabricated in a CMOS-compatible clean room on 8 inch wafers utilizing a full Si-based fabrication tool set. Devices of various pixel sizes, from 1 to 30 micron have been fabricated. Additionally, fabrication of more complicated sub-pixel metasurfaces was demonstrated. Electrical switching measurement is obtained by applying electrical pulse trains through a programmable voltage generator into the gate of a high powered transistor. Microsecond switching of 30-micron and 1-micron pixels in a wafer-scale device has been demonstrated. Current devices have demonstrated over 1,000 switching cycles and further durability improvements are underway.

As an application of pixelated electrical device, we have demonstrated a concept of compact infrared reflection spectrometer with no moving parts. Here, we leverage the ability to obtain gradually changing reflection signature of the pixel with voltage between fully-converted crystalline and amorphous phases. This wavelength shift is essentially a nonlinear filter which can be used to reconstruct an input spectrum. Since the reflection response of the pixel states is pre-determined, a spectral response of the scene can be extracted once the pixel spectrum is fully characterized with a known calibrated spectral source. We have carried out such characterization utilizing a supercontinuum laser in the short-wave infrared part of the spectrum and the results of spectral signature extraction for three different incident spectral scenarios.

DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited.

This material is based upon work supported by the Under Secretary of Defense for Research and Engineering under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Under Secretary of Defense for Research and Engineering.

Phase-change materials (PCMs) allow for an ultrafast and reversible transition between their amorphous (a) and crystalline (c) state. This structural transition results in a pronounced change in their optical and electrical properties. The strong resistivity contrast of several orders of magnitude has been employed in electronic and neuromorphic memories, while the unity-scale refractive index change Δn= |n_c- n_a| has been applied in rewriteable optical-data-storage memories. Additionally, Δn is the subject of research in various photonic applications.^[1]
For tunable photonics, PCMs are very desirable, but expensive and laborious to produce. Thus, colloidal PCMs that can be easily synthesized could provide broader access to this material class and allow for deposition on a variety of substrates, including prepatterned holes and vertical interconnect accesses. However, due to the size reduction from GeTe thin films to small nanoparticles (NPs), new effects can arise. Examples are the occurrence of localized surface plasmon resonances in c-GeTe NPs,^[2] and size dependence of the crystallization temperature T_C.^[3] Thickness dependent temperature scaling was also reported for very thin sputtered PCM films, and related to their bonding characteristics.^[4] Recent results indicate that the spatial confinement of PCMs into small NPs results in an increase of the optical band gap E_g, while maintaining the pronounced contrast between the band gaps for a- and c-PCMs E_g,a and E_g,c respectively.^[5]
Insights into small-size PCM NPs can promote the application of these materials in different fields, and add value to fundamental scientific questions, especially regarding the current debate on the unique chemical bonding in PCMs.^[6]
Despite the bottom-up synthesis of PCM NPs, structural confinement can be realized by sub-wavelength patterning. While laser pulses do not allow for switching on such small length scales, tip-induced crystallization can lead to features below 100 nm. The resulting so-called metasurfaces feature extremely strong field confinement and nearly 2π phase shift.^[7]

[1] Hail, C.; Michel, A. U.; Poulikakos, D. and Eghlidi, H. Adv. Opt. Mater. 7 (2019).
[2] Polking, M. J.; Jain, P. K.; Bekenstein, Y.; Banin, U.; Millo, O.; Ramesh, R. and Alivisatos, A. P. Phys. Rev. Lett. 111 (2013).
[3] Yarema, O.; Perevedentsev, A.; Ovuka, V.; Baade, P.; Volk, S.; Wood, V. and Yarema, M. Chem. Mater. 30 (2018).
[4] Simpson, R. E.; Krbal, M.; Fons, P.; Kolobov, A. V.; Tominaga, J.; Uruga, T. and Tanida, H. Nano Lett. 10 (2010).
[5] Michel, A.; Sousa, M.; Yarema, M.; Yarema, O.; Ovuka, V.; Lassaline, N.; Wood, V. and Norris, D. J. In preparation.
[6] Wuttig, M.; Deringer, V. L.; Gonze, X.; Bichara, C. and Raty, J.-Y. Adv. Mater. 30 (2018).
[7] Michel, A.; Meyer, S.; Lassaline, N.; Lightner, C. R.; Bisig, S.; Norris, D. J. and Chigrin, D. N. In preparation.

The crystallization behavior of amorphous nano-regions (20-100 nm in diameter) embedded in a textured epitaxial Ge₂Sb₂Te₅ 25 nm thick film grown on Si (111) substrate has been investigated in situ by TEM analysis. The amorphous regions were obtained by irradiation with 30 keV Ge+ at a fluence of 1.5x10¹⁴ ions/cm² of masked samples. The adopted configuration simulates the GST structure of a device in the RESET state, it is then of relevance for understanding their data retention characteristics. The in situ TEM analysis indicates that the amorphous to crystal transition is characterized by an initial growth velocity of 3.6 pm/s at 75°C, probably related to the external partially damaged area. For the previous annealed sample, a velocity of about 2.6 pm/s was observed at 90°C and a growth speed of 170 pm/s at 110°C for the 50nm and 100nm diameter amorphous spots. The 20 nm diameter amorphous spots recrystallize after the annealing at 90°C. The transition is governed only by crystallization (nucleation is absent) of the atoms located at the boundary of the amorphous dot with the surrounding crystalline regions. In some case a preferential growth along surfaces normal to the [110] and to the [121] directions has been found. The crystallographic characterization of the regrowth crystal indicates a good matching with the zone axis of the surrounding material although the crystalline seed Si (111) at the bottom interface is missing for the damage.

Future Electronic Smart Systems (ESS) require energy efficient storage and fast processing of large amount of data. Phase-Change Materials, exhibiting large electrical and optical contrast between two phases also allowing multiple intermediate states, are highly desired for this application, because in principle they also have the ability to store and process data at the same physical place(Wuttig et al. Nat. Mater. 2007). Despite progresses which have been made to achieve the full potential of Ge₂Sb₂Te₅ based PCM devices, several reliability issues still hinder the realization of ultimate performance. Void formation due to density change upon switching, low crystallization temperature for automotive applications, and resistance drift over time are among the few. GaSb is considered as a promising candidate for future phase change random access memory (PCRAM) devices because of its fast crystallization, high crystallization temperature, and lack of density change upon switching at a specific composition of Ga₃₀Sb₇₀(Putero et al. APL Materials. 2013). Pulsed Laser Deposition (PLD) has become one of the most popular thin film deposition technique. Some of the reasons for this popularity are the flexibility, processing speed, and the wide range of materials that can be deposited. In this work we discuss PLD of single and alternating layers of Sb₂Te₃, GaSb, and Ge-rich GST thin films on various substrates. Amorphous, polycrystalline, and textured heterostructures of Sb₂Te₃/GaSb and Sb₂Te₃/Ge-rich GST are deposited. Reflection high-energy electron diffraction (RHEED) is used as in-situ growth monitoring tool during deposition to achieve desired film quality. Moreover, deposited films are characterized by Scanning Transmission Electron Microscope (S/TEM) for composition and quality analysis before producing single cell vehicles for electrical device testing.

Phase change random access memory (PCRAM) has been successfully applied in the computer storage architecture, as storage class memory, to bridge the performance gap between DRAM and Flash-based solid-state drive due to its good scalability, 3D-integration ability, fast operation speed and compatible with CMOS technology. A good understanding of phase change mechanism is of great help to design new phase change materials with fast operation speed, low power consumption and long-lifetime. In this presentation, we firstly review the development of PCRAM and different understandings on phase change mechanisms in recent years, and then propose a new view on the mechanism, which is based on the octahedral structure motifs and vacancies. Octahedral structure motifs are the based units during phase transition. Robust octahedra and plenty of vacancies in both amorphous and crystalline phase, respectively avoiding large atomic rearrangement and providing necessary space, are of crucial to achieve the nanosecond or even sub-nanosecond operation of PCRAM. Based on this mechanism, we introduce Sc-based robust octahedra in Sb₂Te₃ phase change materials, called Sc-Sb-Te, and achieve 700 ps operation speed. Recently, the concentration of dopants has also been optimized by correlative atom probe tomography and transmission electron microscopy investigations. Focusing on phase change materials and PCRAM for decades, we have successfully developed 128Mb embedded PCRAM chips, which can meet the requirements of most embedded systems.

The large resistivity contrast between the highly resistive amorphous phase and low resistive crystalline phase of phase change memory (PCM) provides opportunities for multi bit per cell memory and neuromorphic applications. However, the drift of resistance of amorphous phase limits the potential of PCM as multiple intermediate states can overlap over time. Although resistance drift is generally attributed to structural relaxation [1], we have observed significant resistance drift in Ge₂Sb₂Te₅ line cells at cryogenic temperatures and also being affected by optical illumination, indicating charge trapping [2]. A detailed understanding of the mechanisms behind resistance drift may allow for mitigating techniques. In this work, we study the effect of optical illumination between 125 K and 300 K and also the effect of different optical illumination conditions on the drift coefficient.
References:
[1] D. Ielmini, S. Lavizzari, D. Sharma, and A. L. Lacaita, “Physical interpretation, modeling and impact on phase change memory (PCM) reliability of resistance drift due to chalcogenide structural relaxation,” Tech. Dig. - Int. Electron Devices Meet. IEDM, pp. 939–942, 2007.
[2] R. S. Khan, S. Muneer, N. Noor, H. Silva, and A. Gokirmak, “Evidence of Charge Trapping Giving Rise to Resistance Drift of Metastable Amorphous Ge₂Sb₂Te₅,” in APS March Meeting, 2019.

In phase change memory cells, the majority of heat is lost through the electrodes during the programming process that leads to significant drops in the performance of the memory device. In this investigation, we report on the thermal properties of thin film carbon nitride with modest electrical resistivity of 5-10 mW cm, low thermal conductivity of 1.47W/m/K, and low interfacial thermal conductance between carbon nitride and phase change material for length scales below 40 nm. The thermally insulating property of carbon nitride makes it a suitable thermal barrier, allowing for less heat loss during Joule heating within the memory unit. We compare carbon nitride thermal properties against commonly used electrodes and insulators such as tungsten and silicon nitride, respectively, to demonstrate the promise of carbon nitride as a potential material candidate for electrode applications in phase change memory devices.

Crystallization of phase change materials has been studied by extensive density functional/molecular dynamics simulations (DF/MD). Four crystallization simulations of amorphous Ge₂Sb₂Te₅(460 atoms) have been completed at 600 K with simulation times up to 8.2 ns.^1-2 A sample with a history of order crystallizes completely in 1.2 ns, but ordering in others takes more time and is less complete. The amorphous starting structures without memory display phases (<1 ns) with subcritical nuclei (10–50 atoms) ranging from nearly cubical blocks to string-like configurations of ABAB squares and AB bonds extending across the cell. Percolation initiates the rapid phase of crystallization and is coupled to the directional p-type bonding. The results emphasize the stochastic nature of crystallization and the importance of sufficiently large samples. This is particularly evident in describing the role of crystallites that can merge to form larger units or hinder complete crystallization by the formation of grain boundaries.
Amorphous Sb is known to crystallize extremely rapidly already at room temperature. Crystallization of Sb has been studied at 600 K using six DF/MD simulations with up to 882 atoms.³ Crystallization proceeded layer-by-layer in most cases and was extremely rapid (∼36 m/s). As shown in Fig. 1, diffusion plays a minor role in the process as the crystallization proceeds from the crystalline rim, and the evolution of bond lengths and ring statistics supports the bond-interchange model of Sb-rich phase change materials.⁴
We have also carried out extensive DF/MD simulations (over 500 atoms, up to 100 ps) of liquid bismuth at four temperatures between 573 - 1023 K.⁵ These simulations provided details of the dynamical structure factors, the dispersion of longitudinal and transverse collective modes, and related properties (power spectrum, viscosity, and sound velocity). Agreement with available inelastic x-ray and neutron scattering data and with previous simulations is generally very good. The results show that DF/MD dynamics simulations can give dynamical information of good quality without the use of fitting functions, even at long wavelengths.

References:
1) J. Kalikka, J. Akola, J. Larrucea, and R. O. Jones, Phys. Rev. B 86, 144113 (2012).
2) J. Kalikka, J. Akola, and R.O. Jones, Phys. Rev. B 90, 184109 (2014); ibid. B 94, 134105 (2016).
3) M. Ropo, J. Akola and R.O. Jones, Phys. Rev. B 96, 184102 (2017).
4) Matsunaga, J. Akola, S. Kohara et al., Nature Materials 10, 129 (2011).
5) M. Ropo, J. Akola and R.O. Jones, J. Chem. Phys. 145, 184502 (2016).

Chalcogenide phase change materials (PCM) are widely-used in optical data storage technologies. Also, they are promising to become the next-generation mainstream electronic non-volatile memory technolgy. These have been discussed and simulated in our prior work [1-4]. The operation of PCM-based devices hinge on the repeated rupture and formation of atomistic-scale chemical bonds to reversibly switch between crystalline and amorphous phases, which require atomistic-scale simulation methodology to achieve in-depth optimization. Nowadays, there are two categories of such methodologies -- density functional theory (DFT) and artificial neural network (ANN). DFT-based molecular dynamics (MD) is accurate but inefficient -- it is practically applicable to only hundres of atoms. ANN-based MD is efficient but inaccurate -- its maximum error of atomic force prediction can be over 10 eV/Angstrom.

In our recent work [5], we present a method based on deep neural network (DNN) to model PCM, which is both accurate and efficient. The DNN-based MD has a computational complexity of O(N), where N is the number of atoms. So, DNN-based MD is orders of magnitude faster than DFT, whose computational complexity is around O(N³). Furthermore, the maximum error of atomic force prediction of DNN-based MD is around 0.1 to 1 eV/Angstrom, which is much more accurate than ANN in the existing literature. In our another recent work [6], we present a transfer learning (TL) method, to handle the needs to simulate multiple combinations (different PCM stoichiometry, doping, etc.). Compared to the state-of-art methods in PCM literature, the proposed TL method can significantly reduce the amout of DFT data used to train the neural network, which is beneficial to enable high-throughput simulations.

Combining the accuracy and efficiency [5], and the trasferrability between different materials [6], we aim at establishing a general-purpose PCM optimization method based on high-throughput simulations.

[1] Jie Liu, et. al., "Impact of doping on bonding energy hierarchy and melting of phase change materials", Journal of Applied Physics, 124, 094503, 2018. DOI: 10.1063/1.5039831
[2] Jie Liu, "Microscopic Origin of Electron Transport Properties and Ultrascalability of Amorphous Phase Change Material Germanium Telluride, IEEE Transactions on Electron Devices, 2017. DOI: 10.1109/TED.2017.2685341
[3] Jie Liu"A multi-scale analysis of the crystallization of amorphous germanium telluride using ab initio simulations and classical crystallization theory", Journal of Applied Physics, 2014. DOI: 10.1063/1.4861721
[4] Jie Liu, et. al., "A multi-scale analysis of the impact of pressure on melting of crystalline phase change material germanium telluride", Applied Physics Letters, 2014. DOI: 10.1063/1.4901044
[5] P.H. Mo, M.C. Shi, Jie Liu, "Deep Neural Network for Accurate and Efficient Atomistic Modeling of Phase Change Memory", 2019 (under review)
[6] P.H. Mo, M.C. Shi, W.Z. Yao, Jie Liu, "Multitask Inductive Transfer Learning of Potential Energy Surfaces", 2019 (under review)

The working principle of conventional phase change memory device is based on ultrafast reversible phase changes between crystalline and amorphous phases of Ge-Sb-Te (GST) materials. For information storage, phase change memory device uses a large contrast either in electrical resistance between the amorphous phase (high-resistance state) and crystalline phase (low-resistances state) or in optical reflectivity between the amorphous phase (low reflectivity state) and crystalline phase (high reflectivity phase). The erase of GST-based memory cell is achieved by applying a high intensity either electrical or laser pulses. These processes lead to the amorphization via melting and subsequent fast quenching of the phase change alloy. However, GST alloys are poor glass formers. Thus, rapid cooling rates are required to suppress the recrystallization of the alloys during the erase process. On another hand, the write process is accomplished by applying either electrical or optical pulses with low intensity. This results in an amorphous-to-crystalline phase transition. Due to intrinsic feature of GST alloys, a main challenge in material science is the optimization of memory writing times, which are limited by the crystallization kinetics of the alloys. In order to speed up the crystallisation rates, several strategies were proposed, including doping of GST materials, precrystallization of an amorphous matrix or using GST based superlattices. Being a nucleation-dominant material, the crystallization of GST alloy in a nanosized phase-change memory cell might however proceed at the crystalline interfaces.
This contribution focuses on determination of crystallization dynamics in epitaxial Ge2Sb2Te5 (GST225) thin films on application relevant (nanosized) length and (nanosecond) time scales. We irradiate the thin films by a single UV laser pulse with a 20 ns pulse duration and with different laser fluences. First, we study the structural transitions in epitaxial GST225 thin films starting from the layered trigonal GST225 structures [1]. We compare the obtained results with the behaviour of GeTe-Sb2Te3 based superlattices (SLs) after the laser irradiation. In both cases, the results reveal the phase transition to the cubic GST225 structure. The cubic phase forms from a transient molten phase at the melt-crystalline interface upon cooling process and crystallizes with epitaxial relationship to the parent phase. Second, by introducing a method based on combination of high temporal and spatial resolution, we determine the crystallization rates of the cubic GST225 phase [2]. The rates are ranging from 0.4 m/s to 1.7 m/s. The values are well in agreement with the published experimental and theoretical data. Moreover, a variation of the laser fluence leads to different cooling rates, which result in different solidification rates and freezing of an amorphous state. Irradiation of amorphous GST225 phase by laser pulse with lower fluence leads to the re-crystallization of the phase and epitaxial formation of the cubic GST225 structure. Overall, our work shows an approach for the investigation of crystallisation kinetics in wide range of phase change materials on application relevant length and time scales. In addition, we demonstrate amorphization and crystallization of GST225 material by using UV laser with single pulse duration and wavelength only, where in the conventional amorphous-to-crystalline phase transitions the lasers with different pulse durations, number and wavelengths are usually applied.
[1] M. Behrens, A. Lotnyk et al., Nanoscale 10 (2018) 22946-22953
[2] M. Behrens, A. Lotnyk et al., ACS Appl. Mater. Interfaces (2019) https://doi.org/10.1021/acsami.9b16111

Phase-change materials (PCM) are technologically attractive materials for non-volatile memory, neuromorphic computing, optoelectronics, and recently, radiofrequency (RF) applications. Chalcogenide PCMs present a dissimilar resistance between the crystalline (low-resistivity) and amorphous phase (high-resistivity). The phase transition is thermally activated, and its kinetics can span orders of magnitude in time, down to sub-nanosecond timescale. Hence, proper probing of the thermal actuation is required to understand the fundamental material properties of PCMs, particularly the physics of melt-quench processes in nanoscale devices, interface vs. bulk effects, drift, and threshold voltage phenomena. Research efforts have been made to characterize PCM using a microthermal stage (MTS) [1]. The MTS structure used either a top Pt heater for the thermal actuation of a lateral Ge₂Sb₂Te₅ (GST) device or a Pt heater surrounding the GST. Thus, the MTS allowed for probing with thermal time scales on the order of microseconds or longer.
Four-terminal, in-line, indirectly heated phase-change switches (IPCS) have been proposed for high-performance RF applications, thanks to their state-of-the-art figure-of-merit (FOM)=1/(2πR_ONC_OFF) [2, 3]. A typical IPCS consists of two RF ports in-line with the PCM (i.e., the RF path), separated by a small gap, and two terminals for thermal actuation using an embedded heater. The heater runs transversely to the direction of the RF path, under the PCM. The heater and the RF path are separated by a thin electrically isolating film (e.g., Si₃N₄). GeTe has been so far the material of choice for RF applications due to its low resistivity in the crystalline phase and relatively large resistivity contrast with the amorphous state, but recently an Sb₇Te₃ IPCS was demonstrated with similar FOM and improved energy efficiency [3].
In this study, we report on the use of IPCS device structure as a platform for electro-thermal characterization of phase transition properties of PCM in nanoscale films at nanosecond temporal resolution. The isolation between the heater and the PCM layer electrically decouples the heating pulse and PCM probing. Thus, the phase transition, as well as temperature-dependent properties up to temperatures of ~1100 K, can be measured during heating pulse application with ns resolution. Furthermore, since the heater is buried underneath the PCM, our setup allows for 1) steady-state PCM temperature validation using scanning thermal microscopy (SThM) from the top surface, 2) separation between contribution of contact (interface) and bulk effects by varying overlap and underlap between the heater and the PCM contacts, and 3) test different materials of interest without significant change in the fabrication process. We use GeTe as a prototype material with known melting temperature, crystallization kinetics, and surface treatment for low ohmic contacts [4]. Measurements are carried out using RF probes for the heater and the PCM to reduce the overshoot and ringing of short-pulses and produce clean waveforms for nanosecond probing. The experimental results are compared to a detailed finite element electro-thermal model to extract key material properties. Overall, our platform can be used to uncover the intriguing kinetics of crystallization and amorphization in chalcogenides and other amorphous semiconductors.
[1] J. Lee et al., IEEE EDL. 32, 7 (2011).
[2] N. El-Hinnawy et al., IEEE EDL. 34, 10 (2013).
[3] E. Yalon et al., IEEE EDL. 40, 3 (2018).
[4] H. M. Aldosari et al., J. Appl. Phys. 122, 175302 (2017)

As one of the most promising candidates for developing new neuromorphic architectures for non-von Neumann computing and information storage, Ge₂Sb₂Te₅(GST225) in photonic devices shows non-volatile, fast, and multi-step changes between the amorphous state and the meta-stable states. In order to search for other composition ratios with desired properties, e.g. large figure of merit, lower switching energy, etc., we have systematically investigated a broad composition range of the Ge-Sb-Te (GST) system. The GST combinatorial spread was fabricated by co-sputtering Ge, Sb and Te targets and was confirmed by wavelength dispersive spectroscopy that the composition variation across the Si wafer covers most part of the GST phase diagram. X-ray diffraction measurements (synchrotron radiation), Raman spectroscopy and resistance mapping clearly show evolution of the structure and phase-change temperature in the GST system. Refractive index and extinction coefficient were calculated at each composition spot using a scanning ellipsometer: some composition ratios were found to have a small extinction coefficient in the amorphous state, a high phase-change temperature, and a large figure of merit. Then, the nano-size photonic devices fabricated by these ratios were measured and showed good reversibility (more than 40000 cycles) and multi-level symmetric switching (at least 15 levels), suggesting that such compositions can be promising for photonic devices for neuromorphic control. Further, we confirm that compositions with promising properties are near the boundary between the Sb-Te and the Ge-Sb-Te phases, indicating that co-existence of the phases at microstructural level may be playing a role in giving rise to their properties. This work is funded by an ONR MURI (Award No. N00014-17-1-2661).

GeSbTe alloys, widely employed for phase change non volatile memory, are usually not suitable in embedded and automotive applications, because of the low crystallization temperature. Doping, in particular with Ge, has been shown to be a viable way to extend the thermal stability of GeSbTe alloys [1,2]. However, once an optimised composition has been selected, during device operation, the material stoichiometry may change, due to the atomic migration induced by the high temperature and/or by the electric field [3]. These variations may lead to programming and retention performance degradation or even to device failure. It is therefore crucial to understand the extent of the atomic migration and its impact on the stoichiometry, as well as on the crystallization properties. In order to decouple the effect of the atomic diffusion at high temperature from that of the field induced electromigration, we have irradiated crystalline Ge rich GeSbTe thin films by laser. The films were prepared by sputter deposition in the amorphous phase, and then converted into the crystalline structure by annealing at 400°C. The irradiation has been performed by using a Yb-YAG laser (515 nm) with 600 ns pulse, operating at different energy densities. By changing the energy density, complete or partial melting and quenching, with amorphization of the film, can be achieved. In this way we are able to study the thermal diffusion processes decoupled from the field induced electromigration. After melting and quenching, the atomic elements distribution has been studied by Transmission Electron Microscopy (TEM) and Electron Energy Loss Spectroscopy (EELS). We find that, even without the electric field, the melting give rises to a prominent diffusion of Ge atoms. By employing finite elements computational analysis, a diffusion coefficient of Ge on the order of 5x10^-5 cm² s^-1 in the molten phase has been estimated.
After amorphization under different irradiation conditions, we have followed the crystallization upon thermal annealing by in-situ time resolved reflectivity measurement. Such a crystallization process, after the first melting and quenching, is expected to be different from the crystallization of the as deposited amorphous film, since the melting produces not only sub-critical crystalline nuclei, but also a modification of the stoichiometry. The study of the crystallization in such a “primed” material is very similar to the situation occurring in a real memory device after the first erasing step and therefore it is relevant for reliability evaluations.

[1] Zuliani P et al 2013 Overcoming temperature limitations in phase change memories with optimized GexSbyTez IEEE Trans. Electron Devices 60 4020
[2] Navarro G et al 2016 N-doping impact in optimized Ge-rich materials based phase-change memory 8th IEEE Int. Non volatile Memory Workshop (IMW)
[3] Padilla A et al 2001 Voltage polarity effects in Ge2Sb2Te5- based phase change memory devices J. Appl. Phys. 110 054501

The mechanical properties, Hardness and Young module, of crystalline micrometer stripes of Ge₂Sb₂Te₅ and of the surrounding 2.2 µm thick amorphous film have been measured simultaneously. Crystallization was performed by a Continuous Wave laser irradiation process in the range 1.25-3.2 mW. The density changes between the amorphous and the crystalline phase causes a relevant stress in the processed material. Usually the stress is measured by the wafer curvature. This methodology doesn’t allow a detailed investigation of the local mechanical modifications when the two phases amorphous and crystalline are contiguous, as in an optical or electrical storage device. The use of micrometer patterned structures and nano-indentation measurements allow us to determine the mechanical behavior of the two contiguous GST phases. The measured values of the mechanical property are more dispersed for the crystallized material than those for the contiguous amorphous phase. The Young module and the hardness are 51±8 GPa and 2880±450 MPa for the crystallized and 33±4 GPa and 1989±300 MPa for the amorphous phase respectively. Microcracks are also seen, together with fracture lines. The occurrence of delamination at laser power densities above 1.70 mW indicates that for these crystallized thicknesses the maximum tensile strength in the amorphous and crystalline phases was reached and exceeded. The phase transition has also been investigated by several analytical tools (TEM, SEM, Raman and XRD diffraction). A simulation code was adopted to compute the temperature-time-space profile as a function of the laser power. These results were used to compare the measured crystallized thickness to that calculated.

Relaxation processes are decisive for many relevant physical properties of amorphous materials. For amorphous phase-change materials (PCMs) employed in non-volatile memories, relaxation processes are, however, difficult to characterize due to the lack of bulk samples. Here, instead of bulk samples, we use powder mechanical spectroscopy for powder samples to detect the prominent excess wings – a characteristic feature of β-relaxations – in a series of amorphous PCMs at temperatures below the glass transition. By contrast, β-relaxations are vanishingly small in amorphous chalcogenides of similar composition, which lack the characteristic features of phase-change materials. This conclusion is corroborated upon crossing the border from PCMs to non-PCMs, where β-relaxations drop significantly. Such a distinction implies that amorphous PCMs belong to a special kind of covalent glasses whose locally fast atomic motions are preserved even below the glass transitions. These findings also suggest a correlation between β-relaxation and crystallization kinetics of PCMs, which may have technological implications for phase-change memory functionalities.

In order to model phase change memory (PCM) cells, dynamic materials models have to be coupled with charge and heat transport models [1]-[7]. Since PCM cells experience high temperatures, very large thermal gradients (~ 50 K/nm) and high current densities during their normal reset and set operations, thermoelectric contributions are significant for both charge and heat transport in these cells. Furthermore, some of the assumptions made for large scale thermoelectric devices do not hold under these extreme conditions.
Charge transport in PCM cells can be modeled using a drift- diffusion approximation. However, electronic material properties such as carrier concentrations, mobilities and diffusivities are not well characterized for phase change materials. In this study, we construct (i) a drift-diffusion model using thermoelectric characteristics of the material, measured in a large temperature range, and (ii) an electrothermal heat transport model that does not assume mild temperature gradients only.
Our simulation results show that both Thomson heat (thermoelectric effects in the bulk materials) and Peltier heat (thermoelectric effects at the contact regions) play a significant role, and give rise to significantly asymmetric thermal profiles in symmetric PCM cells.

Acknowledgments: This work is partially supported by NSF under award DMR-1710468.

References:
[1] Z. Woods and A. Gokirmak, IEEE Trans. Electron Devices 64, 4466 (2017).
[2] Z. Woods et al., IEEE Trans. Electron Devices 64, 4472 (2017).
[3] J. Scoggin et al., Applied Physics Letters 112, 193502 (2018).
[4] J. Scoggin et al., Applied Physics Letters 114, 043502 (2019).
[5] A. Faraclas et al., IEEE Trans. Electron Devices 61, 372 (2014)
[6] A. Faraclas et al., IEEE Trans. Electron Devices 32, 1131 (2011)
[7] G. Bakan et al., Scientific reports 3, 2724 (2013)

Transition metal dichalcogenides (TMDs) exist in various crystal structures with semiconducting, semi-metallic, and metallic properties. The dynamic control of these phases is of immediate interest for next generation electronics such as phase change memories and TMDS have the potential to outperform current phase-change materials such as GeSbTe on energy efficiency while also being fundamentally more scalable. Of the binary Mo and W-based TMDs, MoTe₂ is attractive for electronic applications because it has the lowest energy difference (40 meV) between the semiconducting (2H) and semi-metallic (1T’) phases, allowing for MoTe₂ phase change by electrostatic doping.

Here we report phase change between the 2H and 1T’ polymorphs of MoTe₂ in thicknesses ranging from the monolayer case to effective bulk (73nm) using an ionic liquid electrolyte at room temperature and in air. We find consistent evidence of a partially reversible 2H-1T’ transition using in-situ Raman spectroscopy where the phase change occurs in the top-most layers of the MoTe₂ flake. We empirically find that the transition voltage of the phase change increases with thickness which provides a direct avenue to transition-voltage control for phase change devices. We also show evidence of electrochemical activity during the gating process by observation of Te metal deposition. This finding suggests the formation of Te vacancies which have been reported to lower the energy difference between the 2H and 1T’ phase, potentially aiding the phase change process. This provides another avenue for transition-voltage control through defect engineering but also emphasizes the need for better understanding of electrolyte gating and chalcogenide vacancy formation. Our discovery that the phase change can be achieved on the surface layer of bulk materials reveals that this electrochemical mechanism does not require isolation of a single layer and the phase change effect may be more broadly applicable than previously thought. We will also discuss recent work on implementation of a solid state device with the use of ionic gels and porous dielectrics.