Phase-change materials are some of the most promising materials for data-storage applications. They are already used in rewriteable optical data storage and offer great potential as an emerging non-volatile electronic memory. This review looks at the unique property combination that characterizes phase-change materials. The crystalline state often shows an octahedral-like atomic arrangement, frequently accompanied by pronounced lattice distortions and huge vacancy concentrations. This can be attributed to the chemical bonding in phase-change alloys, which is promoted by p-orbitals. From this insight, phase-change alloys with desired properties can be designed. This is demonstrated for the optical properties of phase-change alloys, in particular the contrast between the amorphous and crystalline states. The origin of the fast crystallization kinetics is also discussed.

The global demand for data storage and processing has increased exponentially in recent decades. To respond to this demand, research efforts have been devoted to the development of non-volatile memory and neuro-inspired computing technologies. Chalcogenide phase-change materials (PCMs) are leading candidates for such applications, and they have become technologically mature with recently released competitive products. In this Review, we focus on the mechanisms of the crystallization dynamics of PCMs by discussing structural and kinetic experiments, as well as ab initio atomistic modelling and materials design. Based on the knowledge at the atomistic level, we depict routes to improve the parameters of phase-change devices for universal memory. Moreover, we discuss the role of crystallization in enabling neuro-inspired computing using PCMs. Finally, we present an outlook for future opportunities of PCMs, including all-photonic memories and processors, flexible displays with nanopixel resolution and nanoscale switches and controllers.

Memristors are continuously tunable resistors that emulate biological synapses. Conceptualized in the 1970s, they traditionally operate by voltage-induced displacements of matter, although the details of the mechanism remain under debate. Purely electronic memristors based on well-established physical phenomena with albeit modest resistance changes have also emerged. Here we demonstrate that voltage-controlled domain configurations in ferroelectric tunnel barriers yield memristive behaviour with resistance variations exceeding two orders of magnitude and a 10 ns operation speed. Using models of ferroelectric-domain nucleation and growth, we explain the quasi-continuous resistance variations and derive a simple analytical expression for the memristive effect. Our results suggest new opportunities for ferroelectrics as the hardware basis of future neuromorphic computational architectures.

Memristive devices are electrical resistance switches that can retain a state of internal resistance based on the history of applied voltage and current. These devices can store and process information, and offer several key performance characteristics that exceed conventional integrated circuit technology. An important class of memristive devices are two-terminal resistance switches based on ionic motion, which are built from a simple conductor/insulator/conductor thin-film stack. These devices were originally conceived in the late 1960s and recent progress has led to fast, low-energy, high-endurance devices that can be scaled down to less than 10 nm and stacked in three dimensions. However, the underlying device mechanisms remain unclear, which is a significant barrier to their widespread application. Here, we review recent progress in the development and understanding of memristive devices. We also examine the performance requirements for computing with memristive devices and detail how the outstanding challenges could be met.

Traditional von Neumann computing systems involve separate processing and memory units. However, data movement is costly in terms of time and energy and this problem is aggravated by the recent explosive growth in highly data-centric applications related to artificial intelligence. This calls for a radical departure from the traditional systems and one such non-von Neumann computational approach is in-memory computing. Hereby certain computational tasks are performed in place in the memory itself by exploiting the physical attributes of the memory devices. Both charge-based and resistance-based memory devices are being explored for in-memory computing. In this Review, we provide a broad overview of the key computational primitives enabled by these memory devices as well as their applications spanning scientific computing, signal processing, optimization, machine learning, deep learning and stochastic computing.

Rios, Carlos; Hosseini, Peiman; Bhaskaran, Harish Univ Oxford, Dept Mat, Oxford OX1 3PH, England. Stegmaier, Matthias; Wang, Di; Scherer, Torsten; Pernice, Wolfram H. P. Karlsruhe Inst Technol, Inst Nanotechnol, D-76344 Eggenstein Leopoldshafen, Germany. Wang, Di; Scherer, Torsten Karlsruhe Inst Technol, Karlsruhe Nano Micro Facil, D-76344 Eggenstein Leopoldshafen, Germany. Wright, C. David Univ Exeter, Dept Engn, Exeter EX4 4QF, Devon, England. Pernice, Wolfram H. P. Univ Munster, Inst Phys, D-48149 Munster, Germany.

Neuromorphic photonics has recently emerged as a promising hardware accelerator, with significant potential speed and energy advantages over digital electronics for machine learning algorithms, such as neural networks of various types. Integrated photonic networks are particularly powerful in performing analog computing of matrix-vector multiplication (MVM) as they afford unparalleled speed and bandwidth density for data transmission. Incorporating nonvolatile phase-change materials in integrated photonic devices enables indispensable programming and in-memory computing capabilities for on-chip optical computing. Here, we demonstrate a multimode photonic computing core consisting of an array of programable mode converters based on on-waveguide metasurfaces made of phase-change materials. The programmable converters utilize the refractive index change of the phase-change material GeSbTe during phase transition to control the waveguide spatial modes with a very high precision of up to 64 levels in modal contrast. This contrast is used to represent the matrix elements, with 6-bit resolution and both positive and negative values, to perform MVM computation in neural network algorithms. We demonstrate a prototypical optical convolutional neural network that can perform image processing and recognition tasks with high accuracy. With a broad operation bandwidth and a compact device footprint, the demonstrated multimode photonic core is promising toward large-scale photonic neural networks with ultrahigh computation throughputs.

The rapidly growing demand for data storage and processing, driven by artificial intelligence (AI) and other data-intensive applications, is posing a serious challenge for current computing devices based on the von Neumann architecture. For every calculation, data sets need to be shuffled sequentially between the processor, and multiple memory and storage units through bandwidth-limited and energy-inefficient interconnects, typically causing 40% power wastage. Phase-change materials (PCMs) show great promise to break this bottleneck by enabling nonvolatile memory devices that can optimize the complex memory hierarchy, and neuro-inspired computing devices that can unify computing with storage in memory cells. The articles in this issue of MRS Bulletin highlight recent breakthroughs in the fundamental materials science, as well as electronic and photonic implementations of these novel devices based on PCMs.

High-throughput computational materials design is an emerging area of materials science. By combining advanced thermodynamic and electronic-structure methods with intelligent data mining and database construction, and exploiting the power of current supercomputer architectures, scientists generate, manage and analyse enormous data repositories for the discovery of novel materials. In this Review we provide a current snapshot of this rapidly evolving field, and highlight the challenges and opportunities that lie ahead.

The rapid advancement of big data and artificial intelligence has resulted in new data-driven materials research and development (R&D), which has achieved substantial progress. This fourth paradigm is believed to improve materials design efficiency and industrialized application and stimulate the discovery of new materials. The focus of this work is on the emerging field of machine learning-assisted material R&D, with an emphasis on machine learning predictions and optimization design. Following a brief description of feature construction and selection, recent developments in material predictions on phases/structures, processing-structure-property relationships, microstructure, and material performance are reviewed. This paper also summarizes the research progress on optimization algorithms with machine learning models, which is expected to overcome the bottlenecks such as the small size and high noise level of material data samples and huge space for exploration. The challenges and future opportunities for machine learning applications in materials R&D are discussed and prospected.

大数据和人工智能技术的快速发展推动数据驱动的材料研发快速发展成为变革传统试错法的新模式，即所谓的材料研发第四范式。新模式将大幅度提升材料研发效率和工程化应用水平，推动新材料快速发展。本文聚焦机器学习辅助材料研发这一新兴领域，以材料预测和优化设计为主线，在简述材料特征构建与筛选的基础上，综述了机器学习在材料相结构、显微组织、成分-工艺-性能、服役行为预测等方面的研究进展；针对材料数据样本量少、噪音高、质量差，以及新材料探索空间巨大的特点，综述了机器学习模型与优化算法和策略融合，在新材料优化设计中的研究进展和典型应用。最后，讨论了机器学习在材料领域的发展机遇和挑战，展望了发展前景。

The choice of simulation methods in computational materials science is driven by a fundamental trade-off: bridging large time- and length-scales with highly accurate simulations at an affordable computational cost. Venturing the investigation of complex phenomena on large scales requires fast yet accurate computational methods. We review the emerging field of machine-learned potentials, which promises to reach the accuracy of quantum mechanical computations at a substantially reduced computational cost. This Review will summarize the basic principles of the underlying machine learning methods, the data acquisition process and active learning procedures. We highlight multiple recent applications of machine-learned potentials in various fields, ranging from organic chemistry and biomolecules to inorganic crystal structure predictions and surface science. We furthermore discuss the developments required to promote a broader use of ML potentials, and the possibility of using them to help solve open questions in materials science and facilitate fully computational materials design.

Materials genome engineering (MGE) is a frontier technology in the field of material science and engineering, which is well capable to revolutionize the research and development (R&D) mode of new materials, greatly improve the R&D efficiency, shorten the R&D time, and reduce the cost. This paper reviews the progress of MGE in China from the aspects of the fundamental theory and methods, key technology and equipment, the R&D of new materials and related engineering application, talents training, formation and promotion of new concept of material genetic engineering. The paper also looks forward to the future development of MGE in China.

材料基因工程是材料领域的颠覆性前沿技术，将对材料研发模式产生革命性的变革，全面加速材料从设计到工程化应用的进程，大幅度提升新材料的研发效率，缩短研发周期，降低研发成本，促进工程化应用。本文从基础理论与方法、关键技术与装备、新材料研发与工程化应用、人才培养以及材料基因工程新理念的形成和推广等方面，综述了中国材料基因工程的研究进展，并提出了未来发展方向建议。

Developing affordable and light high-temperature materials alternative to Ni-base superalloys has significantly increased the efforts in designing advanced ferritic superalloys. However, currently developed ferritic superalloys still exhibit low high-temperature strengths, which limits their usage. Here we use a CALPHAD-based high-throughput computational method to design light, strong, and low-cost high-entropy alloys for elevated-temperature applications. Through the high-throughput screening, precipitation-strengthened lightweight high-entropy alloys are discovered from thousands of initial compositions, which exhibit enhanced strengths compared to other counterparts at room and elevated temperatures. The experimental and theoretical understanding of both successful and failed cases in their strengthening mechanisms and order-disorder transitions further improves the accuracy of the thermodynamic database of the discovered alloy system. This study shows that integrating high-throughput screening, multiscale modeling, and experimental validation proves to be efficient and useful in accelerating the discovery of advanced precipitation-strengthened structural materials tuned by the high-entropy alloy concept.© 2021. The Author(s).

High-entropy alloys are solid solutions of multiple principal elements that are capable of reaching composition and property regimes inaccessible for dilute materials. Discovering those with valuable properties, however, too often relies on serendipity, because thermodynamic alloy design rules alone often fail in high-dimensional composition spaces. We propose an active learning strategy to accelerate the design of high-entropy Invar alloys in a practically infinite compositional space based on very sparse data. Our approach works as a closed-loop, integrating machine learning with density-functional theory, thermodynamic calculations, and experiments. After processing and characterizing 17 new alloys out of millions of possible compositions, we identified two high-entropy Invar alloys with extremely low thermal expansion coefficients around 2 × 10 per degree kelvin at 300 kelvin. We believe this to be a suitable pathway for the fast and automated discovery of high-entropy alloys with optimal thermal, magnetic, and electrical properties.

Inorganic semiconductors exhibit multifarious physical properties, but they are prevailingly brittle, impeding their application in flexible and hetero-shaped electronics. The exceptional plasticity discovered in InSe crystal indicates the existence of abundant plastically deformable two-dimensional van der Waals (2D vdW) materials, but the conventional trial-and-error method is too time-consuming and costly. Here we report on the discovery of tens of potential 2D chalcogenide crystals with plastic deformability using a nearly automated and efficient high-throughput screening methodology. Seven candidates e.g., famous MoS, GaSe, and SnSe 2D materials are carefully verified to show largely anisotropic plastic deformations, which are contributed by both interlayer and cross-layer slips involving continuous breaking and reconstruction of chemical interactions. The plasticity becomes a new facet of 2D materials for deformable or flexible electronics.© 2022. The Author(s).

The development of novel materials has experienced three paradigms: purely empirical, theoretical models, and computational materials science. Currently, the huge amount of data generated by experiments and simulations has facilitated a shift in materials science to a data-driven fourth paradigm. Therefore, the development of high-throughput automatic integrated computations and data mining algorithms based on material databases and artificial intelligence algorithms is critical for accelerating the design of novel materials. This paper presents an open-source distributed computational platform called Artificial Learning and Knowledge Enhanced Materials Informatics Engineering 2.0 (ALKEMIE2.0) based on the AMDIV (automation-modular-database-intelligence-visualization) design concepts. The ALKEMIE2.0 platform includes five core components of automation, modular, materials database, artificial intelligence, and visualization, which are suitable for the computational design of novel materials. The overall characteristics of ALKEMIE2.0 are divided into five pillars. ALKEMIE-Core integrates multiscale calculations and simulation software using the ALKEMIE-Plugin application programming interface. Its high-throughput calculation workflows that support 104 magnitude concurrencies are implemented by integrating the automatic frameworks of model constructions, calculation workflows, and data analyses. Furthermore, the platform is based on the ALKEMIE-Server, which can easily and automatically open daemon services and realize information interactions in distributed supercomputers. With its strong portability and scalability, ALKEMIE has been deployed in the National Supercomputing Tianjin Center. In addition, the multitype materials database called the ALKEMIE-Data Vault contains structure, task, workflow, and material property databases, which combined with the power of supercomputing, enables the rapid application of artificial intelligence algorithms in the design of new materials. In particular, the many user-friendly interfaces, which were elaborately designed using the ALKEMIE-GUI and are suitable for scientists with broad backgrounds, make structural building, work flowcharts, data analysis, and machine learning models more transparent and maneuverable. Finally, the main features of ALKEMIE2.0 are demonstrated using two examples of multiplatform deployment and high-throughput screening of binary aluminum alloys.

材料研发模式经历了经验主导的第一范式、理论模型主导的第二范式和计算模拟主导的第三范式，如今正处于数据驱动的第四范式。为加速新材料的设计与研发，发展基于材料数据库和人工智能算法的高通量自动集成计算和数据挖掘算法变得至关重要。本文介绍了作者团队自主开发的分布式高通量自动流程集成计算和数据管理智能平台ALKEMIE2.0 (Artificial Learning and Knowledge Enhanced Materials Informatics Engineering 2.0)，该平台基于AMDIV设计理念，包含了自动化、模块化、数据库、人工智能和可视化流程等5个适用于数据驱动的材料研发模式核心要素。概括来说，ALKEMIE2.0以模块化的方式集成了多个不同尺度的计算模拟软件；其高通量自动纠错流程可实现从建模、运行到数据分析，全程自动无人工干预；支持单用户不低于10⁴量级的并发高通量自动计算模拟。进一步而言，ALKEMIE2.0具有强大的可移植性和可扩展性，目前已部署在国家超算天津中心，基于多类型材料数据库结合超算强大的计算能力使得人工智能算法在新材料设计与研发中得以快速的应用和实践。更重要的是，ALKEMIE2.0设计了用户友好的可视化操作界面，使得结构建模、工作流计算逻辑、数据分析和机器学习模型具有更高的透明性和更强的可操作性，且适用于对材料计算模拟掌握程度从初级到专业的所有材料研究人员。最后，通过多平台部署和高通量筛选二元铝合金2个算例详细展示了ALKEMIE2.0的主要特色及功能。

Operation speed is a key challenge in phase-change random-access memory (PCRAM) technology, especially for achieving subnanosecond high-speed cache memory. Commercialized PCRAM products are limited by the tens of nanoseconds writing speed, originating from the stochastic crystal nucleation during the crystallization of amorphous germanium antimony telluride (GeSbTe). Here, we demonstrate an alloying strategy to speed up the crystallization kinetics. The scandium antimony telluride (ScSbTe) compound that we designed allows a writing speed of only 700 picoseconds without preprogramming in a large conventional PCRAM device. This ultrafast crystallization stems from the reduced stochasticity of nucleation through geometrically matched and robust scandium telluride (ScTe) chemical bonds that stabilize crystal precursors in the amorphous state. Controlling nucleation through alloy design paves the way for the development of cache-type PCRAM technology to boost the working efficiency of computing systems.Copyright © 2017 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

Differential scanning calorimetry (DSC) is widely used to study the stability of amorphous solids, characterizing the kinetics of crystallization close to the glass-transition temperature T(g). We apply ultrafast DSC to the phase-change material Ge(2)Sb(2)Te(5) (GST) and show that if the range of heating rates is extended to more than 10(4) K s(-1), the analysis can cover a wider temperature range, up to the point where the crystal growth rate approaches its maximum. The growth rates that can be characterized are some four orders of magnitude higher than in conventional DSC, reaching values relevant for the application of GST as a data-storage medium. The kinetic coefficient for crystal growth has a strongly non-Arrhenius temperature dependence, revealing that supercooled liquid GST has a high fragility. Near T(g) there is evidence for decoupling of the crystal-growth kinetics from viscous flow, matching the behaviour for a fragile liquid suggested by studies on oxide and organic systems.

Phase change materials are widely considered for application in nonvolatile memories because of their ability to achieve phase transformation in the nanosecond time scale. However, the knowledge of fast crystallization dynamics in these materials is limited because of the lack of fast and accurate temperature control methods. In this work, we have developed an experimental methodology that enables ultrafast characterization of phase-change dynamics on a more technologically relevant melt-quenched amorphous phase using practical device structures. We have extracted the crystallization growth velocity (U) in a functional capped phase change memory (PCM) device over 8 orders of magnitude (10(-10) < U < 10(-1) m/s) spanning a wide temperature range (415 < T < 580 K). We also observed direct evidence of non-Arrhenius crystallization behavior in programmed PCM devices at very high heating rates (>10(8) K/s), which reveals the extreme fragility of Ge2Sb2Te5 in its supercooled liquid phase. Furthermore, these crystallization properties were studied as a function of device programming cycles, and the results show degradation in the cell retention properties due to elemental segregation. The above experiments are enabled by the use of an on-chip fast heater and thermometer called as microthermal stage (MTS) integrated with a vertical phase change memory (PCM) cell. The temperature at the PCM layer can be controlled up to 600 K using MTS and with a thermal time constant of 800 ns, leading to heating rates ∼10(8) K/s that are close to the typical device operating conditions during PCM programming. The MTS allows us to independently control the electrical and thermal aspects of phase transformation (inseparable in a conventional PCM cell) and extract the temperature dependence of key material properties in real PCM devices.

Raty, Jean Yves Univ Liege, Phys Solids Interfaces & Nanostruct, B-4000 Sart Tilman Par Liege, Belgium. Zhang, Wei; Luckas, Jennifer; Chen, Chao; Wuttig, Matthias Rhein Westfal TH Aachen, Inst Theoret Solid State Phys, D-52056 Aachen, Germany. Zhang, Wei; Mazzarello, Riccardo Univ Luxembourg, Fac Sci Technol & Commun, L-1511 Luxembourg, Luxembourg. Luckas, Jennifer Rhein Westfal TH Aachen, JARA FIT, D-52062 Aachen, Germany. Luckas, Jennifer Rhein Westfal TH Aachen, JARA HPC, D-52062 Aachen, Germany. Mazzarello, Riccardo; Wuttig, Matthias Aix Marseille Univ, Ctr Interdisciplinaire Nanosci Marseille CINaM, F-13288 Marseille, France. Bichara, Christophe CNRS, F-13288 Marseille, France.

In phase-change memory devices, a material is cycled between glassy and crystalline states. The highly temperature-dependent kinetics of its crystallization process enables application in memory technology, but the transition has not been resolved on an atomic scale. Using femtosecond x-ray diffraction and ab initio computer simulations, we determined the time-dependent pair-correlation function of phase-change materials throughout the melt-quenching and crystallization process. We found a liquid-liquid phase transition in the phase-change materials AgInSbTe and GeSb at 660 and 610 kelvin, respectively. The transition is predominantly caused by the onset of Peierls distortions, the amplitude of which correlates with an increase of the apparent activation energy of diffusivity. This reveals a relationship between atomic structure and kinetics, enabling a systematic optimization of the memory-switching kinetics.Copyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

Phase-change random-access memory (PCRAM) is one of the leading candidates for next-generation data-storage devices, but the trade-off between crystallization (writing) speed and amorphous-phase stability (data retention) presents a key challenge. We control the crystallization kinetics of a phase-change material by applying a constant low voltage via prestructural ordering (incubation) effects. A crystallization speed of 500 picoseconds was achieved, as well as high-speed reversible switching using 500-picosecond pulses. Ab initio molecular dynamics simulations reveal the phase-change kinetics in PCRAM devices and the structural origin of the incubation-assisted increase in crystallization speed. This paves the way for achieving a broadly applicable memory device, capable of nonvolatile operations beyond gigahertz data-transfer rates.

By controlling the amorphous-to-crystalline relative volume, chalcogenide phase-change memory materials can provide multi-level data storage (MLS), which offers great potential for high-density storage-class memory and neuro-inspired computing. However, this type of MLS system suffers from high power consumption and a severe time-dependent resistance increase ("drift") in the amorphous phase, which limits the number of attainable storage levels. Here, we report a new type of MLS system in yttrium-doped antimony telluride, utilizing reversible multi-level phase transitions between three states, i.e., amorphous, metastable cubic and stable hexagonal crystalline phases, with ultralow power consumption (0.6-4.3 pJ) and ultralow resistance drift for the lower two states (power-law exponent < 0.007). The metastable cubic phase is stabilized by yttrium, while the evident reversible cubic-to-hexagonal transition is attributed to the sequential and directional migration of Sb atoms. Finally, the decreased heat dissipation of the material and the increase in crystallinity contribute to the overall high performance. This study opens a new way to achieve advanced multi-level phase-change memory without the need for complicated manufacturing procedures or iterative programming operations.Copyright © 2021 Science China Press. Published by Elsevier B.V. All rights reserved.

Cache-type phase-change random-access memory is a remaining challenge on the path to universal memory. The recently designed Sc0.2Sb2Te3 (SST) alloy is one of the most promising phase-change materials (PCMs) to overcome this challenge, as it allows subnanosecond crystallization speed to reach the crystalline ("1") state at elevated temperatures (e.g., 600 K) but years of reliable retention of the amorphous ("0") state for data storage at room temperature. This contrast in kinetics behavior, upon a relatively small temperature excursion, is more dramatic than that in other PCMs. From the temperature dependence of the crystallization kinetics uncovered via ultrafast differential scanning calorimetry, here, we report an apparent fragile-to-strong crossover in the SST supercooled liquid. We illustrate that two factors are at work simultaneously. First, Sc-stabilized precursors serve as heterogeneous sites to catalyze nucleation, reducing the stochasticity and thereby accelerating the nucleation rate. Second, the SST exhibits an enlarged kinetic contrast between elevated and ambient temperatures. Together they constitute a recipe for the design of PCMs that meets the needs of cache-type nonvolatile memory.

Artificial intelligence and other data-intensive applications have escalated the demand for data storage and processing. New computing devices, such as phase-change random access memory (PCRAM)-based neuro-inspired devices, are promising options for breaking the von Neumann barrier by unifying storage with computing in memory cells. However, current PCRAM devices have considerable noise and drift in electrical resistance that erodes the precision and consistency of these devices. We designed a phase-change heterostructure (PCH) that consists of alternately stacked phase-change and confinement nanolayers to suppress the noise and drift, allowing reliable iterative RESET and cumulative SET operations for high-performance neuro-inspired computing. Our PCH architecture is amenable to industrial production as an intrinsic materials solution, without complex manufacturing procedure or much increased fabrication cost.Copyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works.

The study of metal-insulator transitions (MITs) in crystalline solids is a subject of paramount importance, both from the fundamental point of view and for its relevance to the transport properties of materials. Recently, a MIT governed by disorder was observed in crystalline phase-change materials. Here we report on calculations employing density functional theory, which identify the microscopic mechanism that localizes the wavefunctions and is driving this transition. We show that, in the insulating phase, the electronic states responsible for charge transport are localized inside regions having large vacancy concentrations. The transition to the metallic state is driven by the dissolution of these vacancy clusters and the formation of ordered vacancy layers. These results provide important insights on controlling the wavefunction localization, which should help to develop conceptually new devices based on multiple resistance states.

[Figure: see text].

Multimode silicon photonics is attracting more and more attention because the introduction of higher-order modes makes it possible to increase the channel number for data transmission in mode-division-multiplexed (MDM) systems as well as improve the flexibility of device designs. On the other hand, the design of multimode silicon photonic devices becomes very different compared with the traditional case with the fundamental mode only. Since not only the fundamental mode but also the higher-order modes are involved, one of the most important things for multimode silicon photonics is the realization of effective mode manipulation, which is not difficult, fortunately because the mode dispersion in multimode silicon optical waveguide is very strong. Great progresses have been achieved on multimode silicon photonics in the past years. In this paper, a review of the recent progresses of the representative multimode silicon photonic devices and circuits is given. The first part reviews multimode silicon photonics for MDM systems, including on-chip multichannel mode (de) multiplexers, multimode waveguide bends, multimode waveguide crossings, reconfigurable multimode silicon photonic integrated circuits, multimode chip-fiber couplers, etc. In the second part, we give a discussion about the higher-order mode-assisted silicon photonic devices, including on-chip polarization-handling devices with higher-order modes, add-drop optical filters based on multimode Bragg gratings, and some emerging applications.

Localization of charge carriers in crystalline solids has been the subject of numerous investigations over more than half a century. Materials that show a metal-insulator transition without a structural change are therefore of interest. Mechanisms leading to metal-insulator transition include electron correlation (Mott transition) or disorder (Anderson localization), but a clear distinction is difficult. Here we report on a metal-insulator transition on increasing annealing temperature for a group of crystalline phase-change materials, where the metal-insulator transition is due to strong disorder usually associated only with amorphous solids. With pronounced disorder but weak electron correlation, these phase-change materials form an unparalleled quantum state of matter. Their universal electronic behaviour seems to be at the origin of the remarkable reproducibility of the resistance switching that is crucial to their applications in non-volatile-memory devices. Controlling the degree of disorder in crystalline phase-change materials might enable multilevel resistance states in upcoming storage devices.

Phase-change optical memories are based on the astonishingly rapid nanosecond-scale crystallization of nanosized amorphous 'marks' in a polycrystalline layer. Models of crystallization exist for the commercially used phase-change alloy Ge(2)Sb(2)Te(5) (GST), but not for the equally important class of Sb-Te-based alloys. We have combined X-ray diffraction, extended X-ray absorption fine structure and hard X-ray photoelectron spectroscopy experiments with density functional simulations to determine the crystalline and amorphous structures of Ag(3.5)In(3.8)Sb(75.0)Te(17.7) (AIST) and how they differ from GST. The structure of amorphous (a-) AIST shows a range of atomic ring sizes, whereas a-GST shows mainly small rings and cavities. The local environment of Sb in both forms of AIST is a distorted 3+3 octahedron. These structures suggest a bond-interchange model, where a sequence of small displacements of Sb atoms accompanied by interchanges of short and long bonds is the origin of the rapid crystallization of a-AIST. It differs profoundly from crystallization in a-GST.

Phase-change materials are technologically important due to their manifold applications in data storage. Here we report on ab initio molecular dynamics simulations of crystallization of the phase change material Ag4In3Sb67Te26 (AIST). We show that, at high temperature, the observed crystal growth mechanisms and crystallization speed are in good agreement with experimental data. We provide an in-depth understanding of the crystallization mechanisms at the atomic level. At temperatures below 550 K, the computed growth velocities are much higher than those obtained from time-resolved reflectivity measurements, due to large deviations in the diffusion coefficients. As a consequence of the high fragility of AIST, experimental diffusivities display a dramatic increase in activation energies and prefactors at temperatures below 550 K. This property is essential to ensure fast crystallization at high temperature and a stable amorphous state at low temperature. On the other hand, no such change in the temperature dependence of the diffusivity is observed in our simulations, down to 450 K. We also attribute this different behavior to the fragility of the system, in combination with the very fast quenching times employed in the simulations.

Modern simulation techniques have reached a level of maturity which allows a wide range of problems in chemistry and materials science to be addressed. Unfortunately, the application of first principles methods with predictive power is still limited to rather small systems, and despite the rapid evolution of computer hardware no fundamental change in this situation can be expected. Consequently, the development of more efficient but equally reliable atomistic potentials to reach an atomic level understanding of complex systems has received considerable attention in recent years. A promising new development has been the introduction of machine learning (ML) methods to describe the atomic interactions. Once trained with electronic structure data, ML potentials can accelerate computer simulations by several orders of magnitude, while preserving quantum mechanical accuracy. This Review considers the methodology of an important class of ML potentials that employs artificial neural networks.© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

We provide an introduction to Gaussian process regression (GPR) machine-learning methods in computational materials science and chemistry. The focus of the present review is on the regression of atomistic properties: in particular, on the construction of interatomic potentials, or force fields, in the Gaussian Approximation Potential (GAP) framework; beyond this, we also discuss the fitting of arbitrary scalar, vectorial, and tensorial quantities. Methodological aspects of reference data generation, representation, and regression, as well as the question of how a data-driven model may be validated, are reviewed and critically discussed. A survey of applications to a variety of research questions in chemistry and materials science illustrates the rapid growth in the field. A vision is outlined for the development of the methodology in the years to come.

Phase change materials are of great interest as active layers in rewritable optical disks and novel electronic nonvolatile memories. These applications rest on a fast and reversible transformation between the amorphous and crystalline phases upon heating, taking place on the nanosecond time scale. In this work, we investigate the microscopic origin of the fast crystallization process by means of large-scale molecular dynamics simulations of the phase change compound GeTe. To this end, we use an interatomic potential generated from a Neural Network fitting of a large database of ab initio energies. We demonstrate that in the temperature range of the programming protocols of the electronic memories (500-700 K), nucleation of the crystal in the supercooled liquid is not rate-limiting. In this temperature range, the growth of supercritical nuclei is very fast because of a large atomic mobility, which is, in turn, the consequence of the high fragility of the supercooled liquid and the associated breakdown of the Stokes-Einstein relation between viscosity and diffusivity.

The combination of deep learning and ab initio calculation has shown great promise in revolutionizing future scientific research, but how to design neural network models incorporating a priori knowledge and symmetry requirements is a key challenging subject. Here we propose an E(3)-equivariant deep-learning framework to represent density functional theory (DFT) Hamiltonian as a function of material structure, which can naturally preserve the Euclidean symmetry even in the presence of spin-orbit coupling. Our DeepH-E3 method enables efficient electronic structure calculation at ab initio accuracy by learning from DFT data of small-sized structures, making the routine study of large-scale supercells (>10 atoms) feasible. The method can reach sub-meV prediction accuracy at high training efficiency, showing state-of-the-art performance in our experiments. The work is not only of general significance to deep-learning method development but also creates opportunities for materials research, such as building a Moiré-twisted material database.© 2023. The Author(s).