Few properties have provided such a wealth of information on solids as measurements of charge carrier transport. The electrical resistivity in particular is highly valuable to characterize solids, since it varies by more than 32 orders of magnitude. Two different types of solids can be distinguished based upon the temperature dependence of their resistivity, or the reciprocal quantity, the electrical conductivity. While the conductivity of insulators vanishes as T goes to 0 K, metals reveal a finite conductivity. This immediately raises the questions on the nature of the transition from the metallic to the insulating state and the existence of a minimum conductivity for metals. Several concepts to explain a metal-insulator transition (MIT) have been established. According to Mott, a metal-insulator transition can be achieved if the electron correlation exceeds a critical value. A second route to insulating behavior has been identified by Anderson, who showed that increasing disorder turns a metal with delocalized electronic states at the Fermi energy into an insulator with localized states. Nevertheless, in three-dimensional samples it is often difficult to separate the role of electron correlation and disorder. Hence in this study we have focused on thin film samples, where the two-dimensional nature of the samples helps to unravel the origin of the metal-insulator transition. Our investigation of these thin films provides new evidence that an MIT is observed in some phase change materials that can solely be accounted to a varying degree of disorder (Anderson localization). The experimental data enabling this conclusion will be discussed in detail. Implications regarding both the application potential as well as related scientific opportunities will be presented.

In the last few years atomistic simulations based on density functional theory have provided useful insights on the properties of phase change materials (see ref. [1] for a review). However, several key issues such as the crystallization dynamics, the properties of the crystalline/amorphous interface and the thermal conductivity at the nanoscale, just to name a few, are presently beyond the reach of fully ab-initio simulations. A route to overcome the limitations in system size and time scale of ab-initio molecular dynamics is the development of classical interatomic potentials. Traditional approaches based on the fitting of simple functional forms turned out to be unfeasible due to the complexity and variability of the chemical bonding in the crystal and amorphous phases revealed by the ab-initio simulations. A possible solution has been demonstrated recently by Behler and Parrinello [2] who developed empirical interatomic potentials with close to ab-initio accuracy for elemental carbon, silicon and sodium by fitting large ab-initio databases within a neural network (NN) scheme. In general, a NN method is a non-linear technique that allows fitting any function to arbitrary accuracy and does not require any knowledge about the functional form of the underlying problem. By means of this technique, we have developed a classical interatomic potential for GeTe which is one of the compounds under scrutiny for applications in phase change memories. Simulations with the NN potential are from four to five order of magnitude faster than ab-initio ones for 4000-atom models. We will present results of NN simulations on the properties of liquid and amorphous GeTe including thermal conductivity and the homogeneous and heterogeneous crystallization of the amorphous. [1] D. Lencer, M. Salinga, and M. Wuttig, Advan. Mat. 23, 2030 (2011). [2] J. Behler and M.Parrinello, Phys. Rev. Lett. 14, 146401 (2007); J. Behler, R. Martonak, D. Donadio, M. Parrinello, Phys. Rev. Lett. 100, 185501 (2008); H. Eshet, R. Z. Khaliullin, T. D. Kuehne, J. Behler, and M. Parrinello, Phys. Rev. B 81, 184107 (2010); ibidem 81, 100103 (2010); N. Artrith, T. Morawietz, and J. Behler, Phys. Rev. B 83, 153101 (2011).

Phase change memories based on GeSbTe alloys act by the difference in electrical resistivity between their crystalline and amorphous phases. This depends on the difference in bonding between these phases. Shportko et al [1] and Huang [2] proposed that the fundamental difference in bonding is that the crystalline phase has resonant bonding due to secondary bonding between their primary covalent bonds, which must be aligned, whereas the amorphous phase has molecular bonding with no secondary bonding. Tominaga et al [3] recently proposed the Interfacial Phase Change Memory, in which the Sb2Te3 layers act as templates on which a GeTe layer switches between two local configurations. The lower entropy of this lower dimensionality transition than the crystal-amorphous transition leads to a smaller transition energy. The electronic conductivity of the two phases depends on their defects. We have calculated the defects in each phase. We find that the lowest cost defect in the crystal phase is the Ge vacancy, because rebonding of the adjacent Te sites occurs along the secondary bonding directions of its bulk. This pins Ef to the valance band edge. In contrast, secondary bonding is lost in the amorphous and non-resonant interfacial phases. The Ge vacancy is no longer the lowest cost defect, because it cannot re-bond. The lowest cost defect is now the Te interstitial, which along with band tail overlap will pin Ef at midgap. Thus defects also depend directly on the bulk secondary bonding. 1. K Shportko, S Kremers, M Woda, D Lencer, J Robertson, M Wuttig, Nature Materials 7 (2008) 653 2. B Huang, J Robertson, Phys Rev B 81 (2010) 081204 3. R E Simpson, P Fons, A V Kolobov, T Fukaya, T Yagi, J Tominaga, Nature Nanotech 6 501 (2011)

Phase-change (PC) memory materials display a large contrast in optical reflectivity and/or electrical resistivity between their amorphous and crystalline phases. The structural differences between these two phases are important in understanding the underlying physics, and most research on PC materials has investigated structural or electronic properties of these initial amorphous or final crystalline products of the phase transition. Although such information can provide indications, a basic knowledge of the atomic rearrangements occurring during optical or electrical excitation is necessary unambiguously to understand the mechanism of the fast amorphous-crystalline transition in PC materials. By using ab initio molecular-dynamics simulations, we successfully reproduce the early stages of crystallization in PC materials [Ge2Sb2Te5 (GST) and GeTe] with 180- or 300-atoms models. The crystallization onset time shows a stochastic distribution, as generally assumed in classical nucleation theory, and the critical crystal nucleus is estimated to comprise 5â?"10 (Ge, Sb)4Te4 cubes for GST. Simulated growth rates of crystalline clusters in amorphous GST are consistent with extrapolated experimental measurements. It was also found that the formation of ordered planar structures in the amorphous phase plays a critical role in facilitating the crystallization process by lowering the interfacial energy between crystalline clusters and the amorphous phase. We also investigated the role of vacancies during crystallization. Vacancies in a-GST facilitate crystal nucleation and growth by providing room for atomic rearrangement at the crystal-glass interface, which leads to the formation of stable crystal clusters in the amorphous matrix that potentially grow as nuclei for crystallization. Such selective vacancy diffusion with its particular redistribution facilitates the crystal-nucleation process, thereby significantly contributing to the fast speed of crystallization in this material.

The rapid reversible switching between the amorphous and crystalline phases of Ge2Sb2Te5 is investigated with ab initio molecular dynamics. The complete switching cycle involving crystal melting, liquid quenching, and amorphous annealing generating the crystal phase is simulated and the time scale is accounted for. Alternatively, the crystal phase is also generated by slow cooling from the liquid phase. Results revealed that in both crystallization methods a time of about one nanosecond is needed to complete the crystallization process. The time scale involved in the crystallization agree with the experimental time scale for crystallization of ultra small PCRAM cells. The crystallization of Ge2Sb2Te5 supercells with different number of atoms, from 63 to 207, indicate a clear tendency for separation of Ge and Sb species in layers. The crystallized models also indicate a clear tendency of segregation of vacancies, indicating that the vacancy layering plays a key role in the crystallization.

Different applications require different phase change material properties. For performance oriented applications, fast switching speed, low reset current and high cycling endurance are the most critical requirements. For storage oriented applications low reset current is also important, but long data retention and high density, namely, the ability to store multiple levels per cell (MLC), become more critical. This talk will review the connections between the phase change memory cellâ?Ts electrical characteristics and the phase change material physical properties. Switching speed is dominated by the set (amorphous-to-crystalline) operation. While an Sb-rich phase change material (e.g., Ge2Sb2Te5 + Sb) may have a faster set speed, it may also need a higher reset current due to a lower electrical resistivity. Conversely, a higher dynamic resistance phase change material may require less reset current but may also be more difficult to set to a low resistance. Cycling endurance can be improved with dense phase change materials (process optimization) and proper operation power (programming optimization). Ge-rich phase change materials have better data retention but also have high set resistance. Resistance drift and random set state resistance is the major obstacle for MLC operation. Other material requirements for all applications include proper bottom heater materials to reduce the reset current and maintain good reliability, and robust encapsulation materials to provide reliable properties of the phase change memory cell.

The phase-change-memory (PCM) relies on the ability of a chalcogenide material, usually Ge2Sb2Te5 (GST), to switch from the amorphous phase with a high electrical resistance (~MΩ) to a crystalline phase with a low electrical resistance (~kΩ). The structural stability of the amorphous phase is critically affected by temperature-activated crystallization and resistance drift due to structural relaxation (SR) [1]. While amorphous chalcogenides are relatively stable with respect to crystallization, thanks to a high activation energy above 2 eV [2], the lower activation energy of SR can strongly affect the PCM electrical properties of the amorphous phase, including resistance, activation energy for conduction and, most importantly, the threshold voltage VT. The latter marks the boundary between read and programming operations in PCM, thus VT instability must be carefully predicted to minimize read disturbs within the array. While drift models for resistance and activation energies have been reported [3,4], no model is available to describe the evolution of threshold voltage with time due to SR. In this work, the threshold voltage VT in PCM devices is studied from both experimental and modeling viewpoints. First, VT is measured for PCM devices as a function of the programmed state, highlighting its scaling dependence in sub-deca-nanometer sized amorphous volumes. Experimental results are discussed in terms of a constant-power model for threshold switching based on the local heating of carriers [5]. Then, drift measurements have been carried out to evaluate the time evolution of resistance, subthreshold slope and VT in a broad timescale, from about 1 microsecond to few days. After verifying the constant power model for switching [5], we show a new model describing the evolution of VT due to drift. The model is based on (i) Poole Frenkel (PF) modeling for electrical conduction, (ii) constant power model for threshold switching and (iii) SR kinetics for the drift kinetics. The model allows to predict the VT evolution with time based on the evolution of the activation energy for conduction in the amorphous chalcogenide material. A logarithmic increase of VT with time is evidenced within the model in agreement with experimental data over more than 10 decades of time. The results are finally discussed in terms of the evolution of the energy landscape for conduction in the amorphous chalcogenide [4]. [1] A. Pirovano, et al., IEEE Trans. Electron Devices 51, 714 (2004). [2] D. Ielmini and M. Boniardi, Appl. Phys. Lett. 94, 091906 (2009). [3] D. Ielmini, et al., IEEE Trans. Electron Devices 56, 1070 (2009). [4] D. Fugazza, et al., IEDM Tech. Dig. 652-655 (2010). [5] D. Ielmini, Phys. Rev. B 78, 035308 (2008).

The amorphous phase-change material shows time-dependent properties due to their metastable nature. This nature affects the reliability of the phase-change random access memory (PCRAM), such as resistance drift. Resistance drift is increase in the resistance of amorphous phase-change materials according to a power-law relationship with time (R ~ tν), results in the overlap of resistance between phases. The origin of the resistance drift has been explained using several models; the stress relaxation model, band broadening due to compressive stress relaxation leads to the increase of resistance. Another model is the defect annihilation model, the annihilation of defects as a self-stabilization leads to an increase of resistance. Because the stress relaxation is known as a result of defect annihilation in a microscopic viewpoint, the defect concentration can be estimated from the mechanical stress relaxation measurement. In this study, mechanical stress relaxation and electrical resistance measurements for the amorphous (as-dep) Ge2Sb2Te5 films were conducted, and calculations of the resistance according to the two models were performed in order to verify the relationship between the stress and the resistance drift. Mechanical stress relaxation was determined by the wafer curvature measurement and electrical measurements were utilized to measure the resistance drift for the same sample conditions during isothermal annealing at 80 oC. The electrical resistance drift was observed to follow a power-law relationship with time, and the power-law exponent (ν) was 0.076 which is in a similar range as values in real PCRAM devices (0.03 ~ 0.1), even though the sample is a thin film structure. In the mechanical stress data, relaxation of compressive stress was observed and initial stress was compressive stress amount of 15 MPa approximately. For the changes in the resistance with time according to the defect annihilation model and stress relaxation model to be calculated, the changes in the defect concentration and the strain with time were obtained by the fitting of stress relaxation curve to the bimolecular relaxation kinetics [1]. Obtained defect concentration and strain were substituted into the equation for the conduction mechanism according to the defect annihilation and stress relaxation model, respectively. In the comparison of calculated resistance with the measured resistance drift data, the defect annihilation model shows better matching where the initial defect concentration is estimated to be 2.0 Ã- 1021 m-3 approximately. This result shows that the defect annihilation in amorphous Ge2Sb2Te5 explains the resistance drift phenomenon successfully. This study will provide new insight of comparative analysis on the electrical properties and mechanical properties for understanding the nature of amorphous Ge2Sb2Te5 which determines the reliability of PCRAM devices. [1] J. Kalb et al., J. Appl. Phys. (2003).

This paper reports on the progress in technology and characteristics of 20 nm node PRAM. Of particular, diode electrical characteristics including on current, off current and swing is found to be strongly dependent on the doping technique, diode contact size, and height, etc. A novel switching cell structure, which is self-aligned confined one, offering the required reset current (< 100uA) and set resistance is introduced. By optimizing these key processes, highly reliable 20 nm 8 Gb density PRAM with 4F2 cell size has been successfully developed.

Flexible electronics has been highlighted because of its advantages such as low cost, light weight, environmentally friendly low temperature processing and various applications in bending situations. Although fast development of logic integrated circuits on flexible substrate has been progressed, the lack of high performance non-volatile memory becomes one fundamental challenge for flexible electronics. Phase change random access memory (PRAM), one type of nonvolatile memories using the resistance change property of phase-change materials (Ge2Sb2Te5), in which data is recorded by switching the material into either amorphous or crystalline state, has attracted attention as an alternative to flash memory due to its fast program and access times, large cycling endurance and extended scalability. Cell size of the MOS-switch PRAM has been limited due to large amount of reset current to change from crystalline to amorphous phase. To reduce the cell size without loss of current driving capability, a diode-switch PRAM has been presented. In terms of the cell size, the diode-switch PRAM is the most promising of the PRAM cells. Herein, we describe the first development of one diode-one resistor (1D-1R) PRAM on flexible substrates. Using microstructured semiconductor (μs-Sc) technology, the high performance flexible single crystal silicon diodes are integrated with bottom electrode (TiW) / phase change material (GST) / heating layer (TiN) / top electrode (TiW) to control logic state of memory. To form the contact area between GST and TiN, insulator (oxide) is deposited by PECVD (Plasma Enhanced Chemical Vapor Deposition), which is progressed at 300 degrees Celsius. It doesnâ?Tt make problems in plastic substrates, polyimide (PI) film. Contact holes in insulators, which size is 500nm x 500nm square, are generated by dry etching. For the active operation of the memory cell, the 1D-1R PRAM unit cells are interconnected through word, bit, and source lines in 8 x 8 arrays to control each memory unit cell independently. From the results, we demonstrate the PRAM on flexible substrates that has reliable memory performance in terms of retention and endurance.

The elastic properties of Geâ?"Sbâ?"Te (GST) alloys are important for phase-change devices (such as CD-RW, DVD-RW, Blu-ray, or phase-change random access memory) because the transition between the crystalline and amorphous phases involves a volume change accommodated by a strain estimated to be between 150 MPa and 10 GPa. However, the elastic properties of GST alloys are poorly characterized and the experimental and theoretical values show large discrepancies.We carry out a careful analysis of the elastic properties of a model system, crystalline Ge1Sb2Te4, using density functional theory and elastic anisotropy considerations. We show that Ge1Sb2Te4 exhibits significant anisotropy in its elastic properties.

Phase-change memory (PCM) is one of the most promising solutions for NVM technologies with major advantages in low voltages, fast read/write, and low cost due to reduced number of fabrication masks compared to standard Si-based NVMs. The active PCM material commonly used is GST (a ternary chalcogenide). This choice was mainly due to the in-depth knowledge previously developed on GST compound in the optical storage field. Despite rapid R&D advancement that already led to mass-production, thin-film GST-based PCMs have encountered multiple key challenges including scalability, energy efficiency, and embedded applications. Binary chalcogenides offer potential in addressing these issues. In addition, PRAM suffers from high programming currents which are necessary to thermally induced switching, and subsequent thermal crosstalk between nearby PRAM cells. Size effects at the nanoscale may overcome these challenges, but current lithographic technique limits the minimum size of the PRAM memory cell able to be fabricated using a top down process. Cylindrical, single crystal binary chalcogenide nanowires with diameters as low as 20nm have been synthesized for this study in a bottom up approach using the vapor-liquid-solid mechanism. The small size of the nanowire and the simplicity of the binary material allow for probing the scalability and energy efficiency of the PRAM memory technology, which has not yet been fully explored.

The scientific investigations in recent years show that the triple phase-change compounds (GeTe)_m(Sb_2Te_3)_n, abbreviated as GST, are the most accepted candidates for increasing the efficiency of optical storage devices like CD (compact disk), DVD (digital versatile disk), and BD (blue-ray disk). These compounds possess two phases, namely, amorphous and the crystalline. The electrical and optical excitations implemented with different strengths and frequencies cause fast and reversible phase transformation between these two states. This phase-exchange property of GSTs is the main reason for them to be the most popular research topic of recent years in advancing the rewritable multimedia technology. In this study, the atomic and electronic structure calculations of the GeSb_2Te_4 compound, which is one of the most well-studied materials among GST compounds, have been performed using an ab initio method based on pseudopotentials and density functional theory (DFT). In this connection, firstly, we have obtained the lattice parameters as a_0= 4.29 Ã. and c_0/3 = 9.51 Ã.. While the Teâ?"Ge bond lenght is 2.97 Ã., Sbâ?"Te bond lengths are found to be 3.0 Ã. and 3.16 Ã.. To investigate its electronic properties, we have determined the electronic band structure as well as the orbital nature of bonding in this material. We have found that GeSb_2Te_4 compound has a semiconducting character with a band gap of about 0.1 eV. Our result are seen to be in good agreement with the experimental data in the literature.

Thin films of Sn10Sb20Se70-X¬TeX (0 � X � 14) system were deposited by thermal evaporation technique and were found amorphous from X-ray diffraction studies. Resistivity measurement showed thermally activated intrinsic semiconducting behavior upto X = 8 composition while for X > 8 composition, a sharp decrease in resistivity was observed associated with phase transition. Further increase in temperature after transition showed thermally activated conduction with different activation energy. The observed transition was verified as amorphous to crystalline phase transformation by resistivity measurement of both heating and cooling run and for annealed films. The transition temperature showed a decreasing trend, while transition width increased with tellurium content in the samples. The resistivity ratio showed two order of magnitude improvements for sample exhibiting transition. The films exhibited more than 20% optical contrast in the whole visible range. The observed transition temperature in this system was found quite less than the already used Ge22Sb22Te55 system for optical and electronic memories and thus makes it potential candidate for energy efficient memory applications.

Phase change memory devices, based on chalcogenide films, are typically fabricated using sputter deposition techniques. Non-volatile memory structures use the resistance of the phase change material in its two distinct states as the memory element. High resistance in the amorphous phase is one memory state and the low resistance in the crystalline phase is the other memory state. This type of memory cell is targeted for use in non-volatile memory devices and is generally termed Phase Change RAM or PCRAM. An attractive alternative for the chalcogenide film deposition process is metal organic chemical vapor deposition (MOCVD). MOCVD provides for versatility in scaling and high film density, as well as excellent composition control, repeatability and overall economics. In this work, we report on the development and scaling of MOCVD process technology for a variety of chalcogenide materials through deposition on 200 mm wafers. Process development was started in a small vertical reactor on â?¤50 mm samples. The initial results were used to define the tool scaling parameters. The tool was scaled to have a 13.125 inch platter for 200 mm wafer processing and the capability to demonstrate deposition on 300 mm wafers. The MOCVD reactor is of stainless steel construction with a rotating disc susceptor. Computational flow modeling was used to evaluate the design parameters. MOCVD process investigations were carried out over the ranges of 0.3 to 5.0 Torr and 120 to 200 C, depending upon the specific chemistry addressed, and will be reviewed for the materials deposited. Deposition rates varied from 2.25 to 9.00 nm/minute, depending on film composition and other process parameters. The compositions evaluated included: Ge2Sb2Te5, GeSb2Te5, Ge1.5Sb2Te5, InSb4Te7, InSb6Te7, and alloys of In-Ge-Te, In-Si-Te, Si-Sb-Te, In-Si-Ge-Te and Ge-Si-Sb-Te. Samples were characterized using x-ray fluorescence, x-ray diffraction and composition analysis via EDS or WDS, and were processed into devices for resistance-current-voltage and set speed testing. Adhesion testing was performed on all films produced. Selected samples were also sent for radiation hardness total integrated dose testing at Sandia National Laboratories by Micro-RDC. Optical microscopy and SEM analysis were used to determine surface quality. While most compositions produced smooth continuous films, a few compositions did not produce films at all. Thermal analysis revealed a range of crystallization temperatures were achieved depending on composition. Several samples showed memory state cycling through 108 switching cycles.

Phase-change materials are known for their application for example as electronic devices for data storage. Furthermore, some phase-change materials are scalable down to very small nanoparticles with a diameter of about 2 nm like for germanium telluride. (cite M. Caldwell) As a result, the maximum energy, which is necessary during the process of writing, decreases with decreasing particle size (cite S. Raoux). Since the organic capping remains after the nanoparticles synthesis the characterization is hindered. Moreover completely inorganic structures allow a broad application of phase-change nanoparticles in the future. To allow the characterization and future application of isolated GeTe nanoparticles nanocomposites with different matrices are fabricated by an ex-situ and an in-situ method, whereas the ex-situ method offers a fast and reproducible way with a relatively high through-put to fabricate a nanocomposite solution which allows a material deposition by drop-casting. The structure and composition of the amorphous GeTe nanoparticles as well as the nanocomposites are characterized by EDAX, TEM and XRD. No negative influence of the matrices has been detected, which means the nanoparticles show the same structural behavior for crystallization as embedded in the inorganic matrices. In addition, we report quantitative DSC and TGA measurements elucidating the kinetic properties of the nanoparticles and the composites materials. With XRD and DSC an increase in the crystallization temperature for the nanoparticles compared to bulk GeTe has been found. For nanocomposites the crystallization temperature shows a marginal shift to higher temperatures relative to the temperature detected for the particles, which might be related to the bonding situation between the particles and the matrix material. We found a difference of factor six between the crystallization enthalpy for composites and particles that might be caused by the different bonding situation too.

Phase change memory (PCM) technology is considered to be among the most promising alternatives to conventional technologies in embedded memories [1]. The scalability of GeTe and GST is however limited by the high RESET current requirement [2], which can be partly improved by doping [3]. In this paper we focus on correlating electrical and material characterization results to understand the benefits obtained with Boron-doped GeTe (referred to as GeTeB) PCM devices. XRR and RBS were used to verify film thickness uniformity and 2%, 5%, 10% boron content in composition of 100nm thick films. The structural transition from amorphous to crystalline phase was studied by monitoring XRD, reflectivity and resistivity with temperature. B-doping increases crystallization temperature (Tc) by 61% over undoped GeTe, and is indicative of increased amorphous phase stability [4] and is reflected in Kissinger plot activation energies which predict significantly improved 10 year lifetimes. The amorphous as-fabricated devices are electrically â?omelt-quenchedâ? into amorphous state with relatively higher resistances compared to undoped GeTe. Electrical R-V data also indicates large memory window making GeTeB suitable for memory operations. Optical static-tester laser assisted amorphization cartographies show reduction in amorphization laser power with increasing B-doping. Ellipsometry measurements of crystalline optical refraction indices (n) and calculated optical extinction coefficient (k) reveal that amorphization power reduction is probably caused by modification of GeTe intrinsic properties (melting temperature, thermal conductivity, etc.) due to B-doping. This is well correlated by electrical I-V results which indicate higher holding voltages implying higher self-heating power generated within B-doped devices. Additionally, increasing B-doping by 5% decreases RESET current requirement by ~25% which can be beneficial for device scalability, thermal-disturb reduction and reliability. Regarding the crystallization dynamics, the results suggest the possibility of growth dominated crystallization mechanism [5], with growth rate slowed down by boron. This correlates with electrical SET characteristics with progressively slower SET speed accompanied by higher SET currents for higher B-doping. The reliability endurance shows that a stable two decade in resistance memory window can be maintained for 10E6 W/E cycles for GeTeB devices, with the window closure due to convergence of RESET state to low resistance state. The effect of cycling on device composition is investigated by STEM-EDX to provide a deeper insight into factors influencing cell performance and reliability. [1] R. Annunziata, p. 97, IEDM 2009; [2] L. Perniola et al., EDL, v.31, no. 5, p.488, 2010 ; [3] V.Sousa et al., EPCOS 2011 ; [4] Y.H. Shih et al., Proc. IEDM, p.1, 2008. ; [5] G. F. Zhou et al., Proc. SPIE 4090, p.108, 2000.

The scaling property of phase-change memory (PCM) is analyzed using both analytic and numerical methods [1]. The scaling scenarios of the three widely used PCM operation schemes (i.e., constant electric field, constant voltage, and constant current) are studied and compared. It is shown that if the device size is downscaled by a factor of 1/k (k > 1), the operation energy will be reduced by more than k^3 times, the operation current will be reduced by more than k times, and the operation speed will be increased by k^2 times. It is also shown that more than 90% of operation energy is wasted as thermal energy flux into the substrate and electrodes. We predict that, if the wasted thermal flux is effectively reduced by heat confinement technologies, the energy consumed per RESET operation can be decreased from about 1 pJ to less than 100 fJ. It is shown that reducing device aspect ratio helps decreasing PCM energy consumption. It is revealed that cross-cell thermal proximity disturbance is counter intuitively alleviated by scaling, leading to a desirable scaling scenario. [1] J. Liu, B. Yu, and M. P. Anantram, IEEE Electron Device Letters, vol. 32, page 1340-1342, 2011.

Phase-change memory materials (PCMMs) based on ternary Ge-Sb-Te system have captured full attention due to their excellent switching performance. Thin films of PCMMs are currently playing a dominant role in data storage, while phase-change nanowires (PCNWs) have the potential to overcome future challenges such as high data density and low power consumption. Among the existing methods to synthesize PCNWs, vapor-liquid-solid growth mechanism has been reported, but this method has it drawbacks such as large wire diameter and difficulty in forming periodic patterns. In this study, we report synthesis of Sb-Te PCNWs using a templated electrochemical method. Nanoporous anodic aluminum oxide (AAO) is used as a template for the nanowire growth, and silica is coated on the inner surface of the nanopores to control the pore diameter. Sb-Te PCNWs with different compositions, diameters, and aspect ratios are grown inside the silica coated AAO template by electrodeposition. Composition and structure of these NWs are characterized by energy dispersive x-ray spectroscopy, small angle x-ray scattering, nuclear magnetic resonance spectroscopy, and transmission electron microscopy. Moreover, the phase transition properties of the PCNWs are studied by in situ TEM. The structure and phase transition behavior of the nanowires due to the variation of composition, diameter and aspect ratio will be discussed. Our study sheds light on the geometric effect on the structural and thermodynamic properties of PCNWs. * Work has been supported by NSF DMR- 0906825.

Phase-change random access memory (PCRAM) exploits the resistance difference between the amorphous and crystalline phases of chalcogenide layers. Electromigration (EM) is hence a crucial issue for PCRAM since the phase transition is induced by the electrical Joule heating, causing the electrical stressing with high current density and thermal stress in the programming layer. This study presents the EM behaviors of pure Ge2Sb2Te5 (GST) and nitrogen (N)-doped GST layers under DC bias. Scanning electron microscopy (SEM) observed, regardless of the sample type, the failure always occurs at the cathode side of strip-shaped sample due the sufficiently high current density. With the aid of Blackâ?Ts formula, the mean-time-to-failure (MTTF) analysis of samples found the activation energy (Ea) of EM process are 1.07 eV for GST and 0.56 eV for N-doped GST. The decrease of Eaâ?Ts in N-doped GST was ascribed to the increase of grain boundaries due to the grain refinement which amplifies the short-circuit diffusion and consequently accelerates the EM failure. Energy dispersive spectroscopy (EDS) revealed the N-doping may alleviate the element segregation in GST in a moderate manner. The Blech structure samples were also prepared to evaluate the EM behaviors of chalcogenides. The samples comprised of a series of strips with a width of 5 μm and lengths ranging form 5 to 45 μm were prepared and then tested at 300°C with a stress current density of 1Ã-105 A/cm2. Analytical results yielded the critical product, i.e., the product of critical current density and critical length, for pristine and N-doped GST is separately equal to 200 and 80 A/cm. This is in agreement with previous analysis based on the Blackâ?Ts theory which illustrates the degradation of EM behaviors by doping under DC bias condition. SEM/EDS characterization indicated the electron wind force presented in all samples, causing the mass transport and accumulation of constituent elements toward the cathode end. However, the segregation of Ge and Sb atoms to the cathode end and Te atoms to the anode end was observed in pristine GST samples with lengths less than about 20 nm, implying the emergence of electrostatic force in short samples. As to the N-doped samples, the electrostatic force was more pronounced and consequently resulted in severe deviation of stoichiometry at the anode end of sample after EM test.

Phase change memory is a potential candidate for the future of high-speed non-volatile memory, however significant improvements in cell design is crucial for its success in the mainstream market [1]. Due to the asymmetric geometry of phase change mushroom cells, high current density and the high temperature gradients generated, thermoelectric effects play a key role in determining energy consumption, cell performance, and reliability. In this study, rotationally symmetric 2D finite element simulations using COMSOL Multiphysics are implemented for GeSbTe (GST) phase change mushroom cells with TiN bottom (10 â?" 40 nm diameter) contact to examine thermoelectric effects in the active region with different voltage bias polarity applied. The TiN and GST are both electrically and thermally isolated from neighboring cells by ~400 nm of oxide, and the top and bottom contacts are both held at 300 K. Temperature dependent material parameters (electrical conductivity, thermal conductivity, heat capacity, and Seebeck coefficient), latent heat of fusion and thermoelectric Thomson heat are included in the model for accuracy. The conductivity of amorphous GST at room temperature is 2 orders of magnitude higher than in its crystalline state. Switching the direction of current shows a large change in peak molten volume within the cell, as well as current and power consumption due to thermoelectric effects. The thermal gradients forming in the cell also depend on the current polarity. Recent experimental results show that reliability of the cells have a strong dependence on the voltage polarity [1]. This interesting dependence on current polarity as well as model details and comparisons with experimental findings will be discussed. [1] A. Faraclas, N. Williams, A. Gokirmak and H. Silva, "Modeling of Set and Reset Operations of Phase-Change Memory Cells," Electron Device Letters, IEEE (accepted, future issue). [2] A. Padilla, G. Burr, K. Virwani, A. Debunne, C. Rettner, T. Topuria, P. Rice, B. Jackson, D. Dupouy and A. Kellock, "Voltage polarity effects in GST-based Phase Change Memory: Physical origins and implications," IEDM Technical Digest, pp. 656-659, 2010.

The development of phase change materials has, over the past 25 years, focussed on applications in data storage. Indeed, the active layer of re-writeable optical media (DVD-RW, CD-RW) is a phase change thin film. More recently this same class of material has been investigated for electrical phase change random access memory applications. It is common for the crystalline state of phase change materials to exhibit delocalized p- orbital bonding (resonant bonding) which results in an enhancement to the materialâ?Ts dielectric function. Disturbing these resonant bonds, which is usually achieved through heating, results in a large change to the materialâ?Ts optical and electrical properties. This property change is extremely scalable; Ge2Sb2Te5 phase change films just 2 nm thick have been shown to undergo a phase transition. The large change in optical properties combined with the scalability deems phase change materials highly suitable for application in active nanophotonics, where the large changes in the materialâ?Ts dialectic function are used to control light within nano scale structures. In this presentation, the potential of phase change nano photonic modulators in a variety of different optical structures will be discussed.

Recently, GeTe has been receiving much attention with regard to use in phase-change memory for high temperature applications such as automobiles due to its higher crystallization temperature than Ge2Sb2Te5 (GST) .[1] The physical properties of Ge1-xTex thin films such as crystalline structure, crystallization temperature and speed are known to depend strongly on composition.[1] In order to characterize the changes in local structure and bonding environment during amorphous-to-crystalline phase changes, the phonon modes of GeTe and Ge-rich Ge1-xTex thin films have been investigated in various studies [1-5]. Nevertheless, the phonon modes of Te-rich Ge1-xTex alloy thin films have not yet been reported. Furthermore, the dielectric function and optical gap energies of Ge1-xTex alloy have not been investigated for varying composition and we have examined these unreported characteristics in the present study. We grew Ge1-xTex (x= 0.273, 0.533, 0.667) thin films using thermal coevaporation of Ge and Te and annealed the resulting films for crystallization. From X-ray diffraction spectra of crystalline filmes, we found both GeTe4 and Te phases in GeTe matrix for Te-rich film whereas we found only Ge phase in GeTe matrix for Ge-rich film. We measured the dielectric functions and the phonon modes of Ge1-xTex films by using spectroscopic ellipsometry and Raman spectroscopy in order to investigate the electronic and the vibrational properties. Raman data showed that the phonon modes of Te-rich film were very different from those of Ge-rich and GeTe films in both amorphous and crystalline states. Various phonon modes in Te-rich films are assigned to Te, GeTe4, and GeTe phase. Note that GeTe4 phase was found in crystalline Te-rich films. The phonon spectra of amorphous (a-) and crystalline (c-) phase of Ge1-xTex thin films were very similar to each other for GeTe and Ge-rich films. This phenomenon is consistent with faster phase transition of GeTe and Ge-rich films than Te-rich films. For all Ge1-xTex thin films, the amplitude of the imaginary dielectric functions increased substantially between 1 and 3 eV after crystallization. At intermediate annealing temperature of 125 °C and 250 °C, the dielectric functions did not change from those of the amorphous phase for all the compositions. We estimated the optical gap energy by using linear extrapolation of the absorption coefficient. The optical gap energies of the amorphous (crystalline) phase are 0.89 (0.78), 0.75 (0.73), 0.89 (0.59) eV, respectively, for Ge-rich, GeTe, and Te-rich films. The indirect band gap energy decreased with increasing Te content. [1] S. Raoux et al., Appl. Phys. Lett. 95, 071910 (2009). [2] K. S. Andrikopoulos et al., J. Phys.: Condens. Mat. 18, 965 (2006). [3] E. Gourvest et al., Appl. Phys. Lett. 95, 031908 (2009). [4] R. De Bastiani et al., Phys. Rev. B 80, 245205 (2009). [5] R. Mazzarello et al., Phys. Rev. Lett. 104, 085503 (2010).

Two-dimensional electron systems offer enormous opportunities for science discoveries and technological innovations. Typically they are realized via electrostatic gating or charge transfer in semiconductors and oxides. Using electric-field tuning, conductive, superconducting, or magnetic surface layers have been created on materials that originally do not possess these properties. The sheet carrier density achieved is below ~ 10^14 /cm^2 limited by factors such as break-down field of the gate dielectric. Higher densities are much desired in exploration of correlated electron behaviour in low dimensions. Here we report the stabilization of an ultra-thin (< 3 nm) and ultra-dense (> 10^15 /cm^2) electron system on the surface of vanadium dioxide nanobeam via ionic liquid gating. The overall conductance of the nanobeam increases by nearly 100 times at a gate voltage of 3 V. The dependence of source-drain resistance on gate voltage has clear hysteretic and memory behaviour. An insulator-to-metal transition is electrostatically triggered on the surface, thereby unleashing an extremely high density of free electrons from the original valence band within a depth self-limited by the energetics of the system. The structure provides a new two-dimensional electron liquid system for investigation of correlated electron behavior.

We report AIMD simulations of the crystallization of GST-225, addressing the early stages of crystal nucleation and growth. We have determined the critical nucleus size for this material (at the annealing temperature of the simulations). Furthermore the simulated crystallization growth rate matches the experimental data, extrapolated to the simulated annealing temperature. Finally, we have demonstrated experimentally, in GST memory cells, sub-ns switching (reset and set) for the first time, using a particular voltage pulse sequence. We have confirmed this speed-up in crystallization rate by AIMD, and have thereby identified the atomistic origin of this effect.

In this talk we discuss the polyvalency of Ge-Te bonding. While Ge and Te can form covalent bonds forming Ge(4):Te(2) configuration when Ge atoms are sp3-hybridized, due to the unique combination of lone-pair electrons on Te atoms an empty Ge orbitals, dative bonds can be formed with pure p-orbitals generating Ge(3):Te(3) configurations. Both configurations satisfy the Mott rule whose formulation as 8 - N rule fails to work for dating bonds and should be avoided. We further discuss the role of polyvalency in the formation of negative-U centers that may be responsible for pinning the Fermi level in the amorphous phase. FInally, polyamorphism amorphous GeTe-based phase-change alloys is discussed.

Ge2Se3:Ag is an important material system in emerging memristor-based, reconfigurable electronics technologies. It is well known that Ag+1 can be introduced into a thin film from a metallic source by applying a positive bias to the silver-side of a two terminal device. Ag+1 ions are highly mobile and electroplate at the cathode. Columns of metallic material traverse the film setting up high-conductivity paths. Mitkova et al. have studied Raman spectra before and after photo-doping of GexSey . The spectra are consistent with the persistence of Ge-Ge bonds after doping, independent of selenium content. However ,Campbell has posited that Ag selectively replaces Ge in Ge-Ge dimers. Furthermore, the device performance relies on an intervening layer of SnSe between the metallic Ag and the Ge2Se3. We report density functional calculations on supercells and finite clusters based on a crystalline model of Ge2Se3 and on the incorporation of Sn and Ag into this model in both substitutional and interstitial sites. The calculations on pure Ge2Se3 show that electrons self-trap, preferentially in pairs, and that this self-trapping leads to either bond weakening (Ge-Ge and Ge-Se), for single self-trapped electrons (STE's), or Ge-Ge bond weakening and Ge-Se bond scission, for paired STE's. Using transition state theory, we have calculated the barrier to hopping for both the single and paired STE's. The calculated values are 0.26 eV and 0.6 eV , respectively. While there is strong evidence that STE's are -Ueff , because single STE's are absent in the periodic calculations on pure Ge2Se3, we cannot apply standard DFT to calculate defect levels from total energy. However, the densities of states for the -1 and -2 charge states show clearly that the defect level in the gap drops in energy by ~ 1 eV, far from the conduction band edge, when it is doubly occupied. We will argue that these results are crucial to understanding interstitial Sn and Ag because the physics is driven by STE's. Interstitial Ag autoionizes, dropping an electron into Ge2Se3 that becomes self-trapped. We will show that, while absent in pure Ge2Se3, single STE's are observed in a supercell with one Ag interstial. In the -1 charge state, paired STE's are the ground state. Thus, SnI donates two electrons to the host material that, in turn self-trap, breaking a Ge-Se bond. We also present results for substitutional Ag (AgS) and Sn (SnS), both isolated and paired with a Ge interstitial. We have calculated energies of formation as a function of eF for the interstitial, substitutional, and paired defects. AgI is -Ueff . Also AgI and AgS:GeI, compete as lowest energy defects. For eF below midgap, AgI dominates, above midgap, AgS:GeI dominates. Neutral Sn always prefers substitution for Ge to interstitial conformation. We will discuss how these results influence the physics of Ag incorporation for both photo-doping and bias-induced doping.

Emerging electrical memory technologies based on phase-change materials capitalize on a fast amorphous-to-crystalline transition. Recent evidence from measurements of relaxation oscillations and switching statistics in phase-change memory devices indicates the possibility that electric field induced crystal nucleation plays a key role in defining the characteristic electrical switching behavior. Here we present a detailed kinetics study of crystallization in the presence of an electric field for the phase-change material Ge2Sb2Te5. We derive quantitative crystallization maps to show the effects of both temperature and electric field on crystallization and we identify field ranges and parameter values where the electric field effects might play a significant role in both bulk materials and real devices.

Phase change magnetic materials enable to exploit the combination of the intrinsic spin of electrons and phase-change switching and its associated magnetic moment in solid state devices to develop new technology. In this paper, the phase-change magnetic materials (PCMM) exhibiting different electric, optical, and magnetic properties between amorphous and crystalline state were successfully synthesized. The intrinsic FM in phase change magnetic materials were studied by the first principle calculations and experiments. The density functional theory to examine the electronic structures in PCMM was performed. The tendency for magnetic elements to substitute at the sites of phase change materials was revealed. The experimental data of valance-band spectra, hole concentration and magnetic moment were measured, which are in good agreement with the calculation results and supports the proposed mechanism. The design principle that enable to develop new PCMM is proposed

As was first discovered for Se, Te [1] and later on proved for different films prepared by various methods [2] amorphous-crystalline transformation in thin films can be associated with rather general unusual phenomenon: strong (up to 300 degrees per micrometer) regular dislocation independent lattice bending round an axis (or axes) lying in the film plane of the growing crystal. Such "transrotational" structure has been revealed and studied mostly in situ, using transmission electron microscopy (TEM) bend-contour technique [2, 3] for the different amorphous films (primary chalcogenides and oxides), prepared by various methods and crystallized in different manner. HREM, AFM were used in due case to study some details of both initial amorphous and crystallized areas. In this paper we describe (on the level from nano- to microscale) different geometries, textures and gradients of lattice orientations, for this unusual microstructure named â?otransrotationalâ? [4]. Generally it can be considered as an intermediate structure between glassy and crystalline (similarly to the structure of liquid crystals regarded as intermediate between crystalline and liquid ones). It may be one of the reasons of good erasure properties of phase-change memory in chalcogenide-based films, which in fact tend to crystallize in such manner. Multiple reversible local transformations â?oamorphous â?" transrotational crystallineâ? observed in situ in TEM are presented for Se-Te films. Atomistic mathematical model for the atom positions in "transrotational" single crystal is proposed (based on mathematical instruments of conformal/quasi-conformal transformations of usual set of crystal lattice points to the transrotational one), as well as hypothetical mechanism of unusual crystal growth in an amorphous matrix based on the surface nucleation. Microcrystals and nanostructures with â?otransrotationâ? during last years have been eventually recognized/studied in a variety of thin film systems including well-known chalcogen/chalcogenide-based compositions (e.g. [5-6]). We suppose that the role of transrotational crystalline structures (recognized directly by TEM only) has been underestimated since corresponding microstructure parameters can strongly influence the phase switching, e.g. the time and energy needed for writing/rewriting in chalcogenide films for phase change information storage. [1] I. E. Bolotov, V. Yu. Kolosov and A.V. Kozhyn, Phys. Stat. Sol. 72a, 645 (1982). [2] V. Yu. Kolosov, Proc. XII ICEM, Seattle, 1, 574 (1990). [3] I. E. Bolotov and V. Yu. Kolosov, Phys. Stat. Sol. 69a, 85 (1982) [4] V. Yu. Kolosov and A. R. Thölen, Acta Mat. 48, 1829 (2000). [5] B. J. Kooi, and J. Th. M. De Hosson, J. Appl. Phys. 95, 4714 (2004). [6] E. Rimini, et. al., J. Appl. Phys. 105, 123502 (2009). Partially supported by RF Ministry of Higher Education and Science.

The more common phase change materials (PCM) are tellurium-based ternary alloys belonging to the GeTe-Sb2Te3 pseudo-binary line of the ternary GeSbTe phase diagram [1,2]. However, for phase change random access memory application (PCRAM), tellurium was reported to disfavor the life time cycle of the devices [3] and the relatively low crystallization temperature makes impossible the use of such devices in automotive applications. In the quest for new materials, the chalcogen free Ge-Sb system was recently proposed. From the thermodynamics, the phase diagram of this system is quite simple as it only contains an eutectic transformation located at 15 at.% Ge and at 865 K. The eutectic composition, having the lowest melting point and an elevated crystallization temperature [4,5] was suggested for PCRAM applications since it could necessitate a lower current for programming the devices. However, the stability of this alloy is questioned since a two-step crystallization was systematically reported [5-7]. Several mechanisms are proposed to understand this behavior which led to a phase separation into a Ge-rich phase and a Sb-rich phase. The exact composition of the two phases is still under discussion. In this contribution, the crystallization process of Ge15Sb85 sputtered thin films is investigated by simultaneous EXAFS and DSC measurements. A new set-up particularly well adapted for material science was recently developed to address such issues. The advantage of combining these techniques is that on the one side calorimetry gives access to the energetics associated to a phase transition and on the other side, EXAFS is a spectroscopic tool sensitive to the local order of a given specie in the ordered as well in the disordered state. Owing to enthalpy calculations combined to EXAFS analysis of the local order around Ge and Sb atoms, the composition of the phases as well as the structural mechanisms leading to phase separation will be discussed. [1] S. Raoux et al. (2010,) Chemical reviews, 110(1), p.240 [2] D. Lencer et al. (2011), Adv. Materials, 23, p2030. [3] L. Krusin-Elbaum et al.(2007), Applied Physics Letters, 90,(14) p.141902. [4] S. Raoux et al. (2009), Journal of Applied Physics, 105 (6), p.064918. [5] C. Cabral et al., (2008),Applied Physics Letters, 93(7), p.1906. [6] S. Raoux et al., (2009), Applied Physics Letters, 95(14), p.143118. [7] P. Zalden et al. (2010),Journal of Applied Physics, 107(10), p.4312.

Phase change memory (PCM) device operation is based on electrical resistivity contrast between the amorphous (high resistivity) and the crystalline (low resistivity) phases of chalcogenide materials. A small volume of phase change material (active region) can be reversibly and rapidly switched between amorphous and crystalline phases by a suitable electrical pulse [1]. These operations require high temperatures, high current densities and result in high thermal gradients; which lead to significant thermoelectric effects. In this study, we have performed numerical modeling of the electro-thermal effects on nanoscale lateral Ge2Sb2Te5 (GST) phase change memory [2] cell structures such as line, dog bone, and T shapes using COMSOL Multiphysics in 3D. Thermoelectric Thomson heat is included in the physical models for current continuity and thermal transport. Physical parameters of GST, TiN and SiO2 such as resistivity, thermal conductivity, heat capacity and Seebeck coefficient (S) were obtained from the literature. Latent heat of fusion is modeled as increased heat capacity around melting point. Applied voltage is adjusted to reach just above the melting temperature for each GST structure. Strong asymmetry is observed in thermal profiles in all cases; the hottest spot appears closer to the higher potential end. Larger shifts are observed for the suspended structures. Simulation results show formation of a single liquid filament on the top surface of the structures resting on SiO2. Dynamics of melting of GST and the impact of thermoelectric effects on scaled devices with various lateral geometries will be discussed. References [1] H. Wong, S. Raoux, S. B. Kim, J. Liang, J. P. Reifenberg, B. Rajendran, M. Asheghi and K. E. Goodson, "Phase Change Memory," Proc IEEE, vol. 98, pp. 2201-2227, 2010. [2] S. Raoux, M. Wuttig and L. v. Pieterson, Phase Change Materials: Science and Applications. Springer Verlag, 2008, pp. 446.

Multi-component chalcogenides, such as Ge-Sb-Te and Ag-In-Sb-Te, are potentially used in optical data storage media in the forms of rewritable CDs, DVDs, and Blu-ray discs. Especially, Ge2Sb2Te5 (GST) has proven to be one of the highest-performance alloys among commercially available phase-change materials. A question arising from the dynamics of the phase change in Ge2Sb2Te5 is how fast the phase transformation between the amorphous and the crystalline phases occurs. Yet ultrafast (less than 1 picosecond) phase transition between the two phases has not been demonstrated, but it is indeed required to extend the application of chalcogenides to ultrafast non-volatile switching devices. There is a possibility of manipulating the rapid phase change in chalcogenide in ultrafast time spans if one uses ultrashort laser pulses, whose pulse duration is less than the time period of lattice motions, being typically 100 fs. This idea is based on the coherent control of local vibrations, whose atomic motion plays an important role in the rapid phase change. Here, we introduce experimental demonstration in Ge2Te2/Sb2Te3 superlattice that the phase change from amorphous into crystalline states can be manipulated within only a few cycles of local lattice vibration (â?^ 1 picosecond) by coherent excitation of the local lattice vibration using a pair of femtosecond laser pulses. To measure time-resolved reflectivity change of the sample as a function of the time delay, 20 fs-width optical pulses (λ = 850 nm) from a Ti:sapphire laser was utilized. A pair of pump-pulses was generated through a Michelson-type interferometer, in which the time interval of the double pump-pulses was precisely controlled by moving a piezo stage. Most importantly, as the results, the phase change occurs only when the time separation between the two pump-pulses is simultaneously resonant to the local vibration. In addition, the transient frequency of the local vibrational mode initially softened on a sub-picosecond time scale, followed by the quenching at different final states, depending on the fluence of the second pump-pulse. Furthermore, we have examined a possibility of bond-selective excitation in Ge2Te2/Sb2Te3 superlattice by tuning the pump polarization angle of the single pump pulse. In this new approach, p-polarized pump pulse is found to be more effective to induce the phase change, probably due to the displacement of Ge atoms (and surrounding Sb and Te atoms) along [111] direction due to the anisotropic excitation of a weaker bond with p-polarized pulses.

Phase Change Radom Access Memory (PCRAM) is a promising non-volatile memory (NVM) due to its near ideal properties. One important characteristics of PCRAM is multi-level capability which has been demonstrated [1, 2]. However, the resistance drift phenomenon, that resistance increases steady with time, hinders the development of multi-level PCRAM. Recently, many possible causes of the resistance drift have been reported with two widely discussed, structural relaxation and stress release [3, 4]. However, all these works mostly focus on investigating the causes by using physical models and simulations. Hence, in this work, experiments were designed to evaluate the possible resistance drift models and explore the possible solutions to minimize the drift. Different dopants were incorporated in GeSbTe to investigate the effect of dopant on the resistance drift. The experimental results show that the dopant significantly affects the resistance drift and the effect varies with the dopant type. Among the different dopants, it was found that the drift coefficient of metal doped GeSbTe was significantly reduced. With proper selection of metal dopant and concentration, the drift coefficient is nearly eliminated. This is possibly due to the density of localized states or conduction activation energy in amorphous GeSbTe was modulated by the dopant. Because the annihilation of defects by trap-fillings during structural relaxation significantly reduces the electronic transport properties of amorphous materials by increasing the electrical conduction activation energy and results in the resistance increase, making the conduction activation energy stable is possible to minimize the resistance drift. Different electrodes on the resistance drift were also investigated. It was found that smaller the mismatching of thermal expansion coefficient with electrode and GeSbTe would reduce the compression stress and thus reduce the drift effectively. [1] T. Nirshl, et al., â?oWrite strategies for 2 and 4 bit multi-level phase change memory.â? In IEDM Technical Digest, page 17.5, 2007 [2] F. Bedeschi, et al., â?oA multi-level cell bipolar selected phase change memoryâ?, In ISSCC Technical Digest, 2008 [3] D. Ielmini, et al., â?oPhysical interpretation, modeling and impact on phase change memory (PCM) reliability of resistance drift due to Chalcogenide structural relaxation,â? in 2007 IEDM Technical Digest, pp.939-42. [4] I. V. Karpov, et al., â?oFundamental drift of parameters in Chalcogenide phase change memory,â? Journal of Applied Physics, vol. 102, 2007

For applications to phase-change random access memory (PRAM) devices, Ge2Sb2Te5 (GST) faces serious reliability issues during cell operation. Among these, repeated volume changes, due to differences in density between the amorphous and crystalline phases, result in a local buildup of the mechanical stress inside the GST films or at the interfaces between GST and the surrounding materials. Changes in volume in Ge2Sb2Te5 (GST) and nitrogen-doped Ge2Sb2Te5 (NGST) films during phase transformations between amorphous and crystallized states were examined with the consequent void formations. Transmission electron microscopy (TEM) was used to measure changes in volume during laser-induced crystallization for both as-deposited and melt-quenched amorphous films for GST and NGST phases, and changes in density were estimated. We observed that the initiation of the void formation in the GST phase is different from that of the NGST phase. We also observed the formation of voids and their coalescence during repeated heat cycling by laser. A variation in chemical composition around the voids was confirmed using energy-dispersive spectroscopy analysis. The void formation and compositional changes can lead to degradation of reliability in phase-change random access memory devices.

Phase-change materials show remarkable nonlinear electrical behaviour in their amorphous state. At a critical threshold field of the order of 10 MV/m the amorphous state resistivity suddenly decreases by orders of magnitude. These threshold fields are material dependent [1]. The amorphous state resistivity below the threshold shows thermally activated behaviour and is observed to increase with time. Understanding the physical origins of threshold switching and resistance drift phenomena is crucial to improve non-volatile phase-change memories. Both phenomena are often attributed to localized defect states in the band gap [2-4]. However, little is known about the defect density in amorphous phase-change materials. This work presents an experimental study of defect states in a-GeTe combining Photothermal Deflection Spectroscopy and Modulated Photo Current Experiments. Based on these experimental findings a defect model for a-GeTe is developed, which is shown to consist of defect bands and band tail states. To get a better understanding of resistance drift phenomena we have studied the evolution of defect state density, optical band gap and resistivity in a-GeTe thin films with time at elevated temperatures. After heating the samples one hour at 140°C the activation energy for electric conduction increases by 30 meV, while the optical band gap increases by 60 meV. This finding demonstrates the impact of the band gap opening on resistance drift. Furthermore, the defect state densities of a-Ge15Te85 and a-Ge2Sb2Te5 are measured by Modulated Photo Current Experiments to investigate the connection between defect state densities and threshold switching phenomena. For the investigated alloys the measured density of midgap states is observed to decreases with decreasing threshold field known from literature. This is discussed within the frame-work of a generation-recombination model originally proposed by Adler for chalcogenide glasses [5]. [1] D. Krebs, et al., Appl. Phys. Lett., 95, 082101 (2009) [2] A. Pirovano et al., IEEE., 51, 714 (2004) [3] D. Ielmini, , Phys. Rev. B, 78, 035308 (2008) [4] D. Ielmini,, et al., IEEE , 56, 1070 (2009) [5] D. Adler, et al., J. Appl. Phy.s, 51, 3289 (1980)

We demonstrate control of phase-change memory (PCM) bits using individual carbon nanotube (CNT) electrodes with diameters of the order ~1 nm. This configuration achieves programming currents of ~1 uA, approximately two orders of magnitude lower than existing state-of-the-art PCM. Memory switching with pulsed measurements shows very low energy consumption (~fJ/bit), defining a path toward extremely scalable memory and programmable electronics [1,2]. We also describe recent progress on defining self-aligned nanowire-nanotube PCM bits [3], as well as PCM bits with graphene electrodes. [1] F. Xiong, A. Liao, E. Pop, Appl. Phys. Lett. 95, 243103 (2009). [2] F. Xiong, A. Liao, D. Estrada, E. Pop, Science 332, 568 (2011). [3] F. Xiong, M.-H. Bae, A. Liao, Y. Dai, E. Pop, E\PCOS (2011).

Among the chalcogenide alloys, some of the fastest crystallizing phase change materials have compositions along the GeTe-Sb2Te3 tie line, Ge2Sb2Te5 (acronym GST) being the most extensively investigated material already integrated in phase change memories (PCM) [1,2]. Besides, to overcome retention and power consumption issues of PCMs, the stoichiometric GeTe compound attracts much attention due to its higher crystallization temperature [3] and its short recrystallization time [4]. Intense efforts have been undertaken for a better understanding of the crystallization mechanisms by employing separated in situ characterization techniques to follow the amorphous-to-crystalline phase transition [5,6,7]. The present contribution brings a new insight into the chalcogenide amorphous-to-crystalline phase transition by studying temperature-induced crystallization of amorphous GST and GeTe films from a unique combination of in situ synchrotron techniques and sheet resistance measurements performed simultaneously during thermal annealing. Ge2Sb2Te5 and GeTe thin films (thickness in the range 20 to 120 nm) were deposited by radiofrequency magnetron sputtering from respective high-purity stoichiometric targets on 200 nm thick SiO2/Si(100) substrates. Simultaneous in situ X-ray diffraction, X-ray reflectivity and sheet resistance measurements were performed on the BM05 beamline at ESRF (Grenoble, France) using an incident energy of 12 keV. To complement X-ray structural analysis, transmission electron microscopy (TEM) imaging was performed on samples before and after the in situ thermal annealing. At (425+-3) K and (453+-4 K) for the GST and GeTe films respectively, the appearance of diffraction peaks associated with a metastable crystalline phase is unambiguously correlated to a density increase in combination with a layer thickness reduction and a resistivity switch towards a lower-resistance state. In the case of GST films, the thermal annealing conditions led to a film consisting of a polycrystalline layer capped by an amorphous layer [8]. This peculiar two-layer microstructure strongly lowers the electrical conductivity by impairing the percolation of charge carriers through the top amorphous GST layer. The persistence of an amorphous top layer will be discussed regarding various hypotheses such as the lack of pre-existing nucleation sites at the free surface or a phase separation that may occur during crystallization. The influence of GST or GeTe layer thickness on the crystallization kinetics will also be discussed. [1] N. Yamada et al., J. Appl. Phys. 69, 2849 (1991) [2] S. Raoux et al., Chem. Rev. 110, 240â?"267 (2010) [3] S. Raoux et al., Appl. Phys. Lett. 95, 143118 (2009) [4] S. Raoux et al., Appl. Phys. Lett. 95, 071910 (2009) [5] N. Kato et al., Appl. Surf. Sci. 244, 281â?"284 (2005) [6] P. La Fata et al., J. Appl. Phys. 105, 083546 (2009) [7] J. Kalb et al., J. Appl. Phys. 98, 054910 (2005) [8] M. Putero et al., J. Appl. Cryst. 44, 858â?"864 (2011)

We report single-crystal growth of GST on Si (111) by sputtering GST on a hot Si (111) substrate. The epitaxial film growth was studied as a function of the Si substrate temperature in the range of 110°C to 200°C. We also report solid phase epitaxy of GST. The top portion of a single-crystal GST film was amorphized by ion implantation and then the amorphous film was re-grown as a single-crystal GST film by annealing. To confirm epitaxial growth the films were characterized by high resolution transmission electron microscope (HRTEM) and high-resolution x-ray diffraction (HRXRD). Energy dispersive x-ray spectroscopy (EDX) profiles taken with scanning TEM (STEM) were used to study the epitaxial films composition. We also confirmed the film composition by secondary ion mass spectrometry (SIMS). To achieve epitaxy the native oxide was etched from the Si substrate surface by hydrofluoric acid just prior to loading the substrate to the growth chamber. The Si substrate was set to a selected temperature in the range of 110°C to 200°C and sputtering was carried out from a GST target. HRTEM showed that epitaxy was obtained for all films with the exception of the film deposited at 110°C, which appeared poly-crystalline. HRXRD measurements confirmed the HRTEM measurements, although a very weak GST (111) peak was observed for the film grown at 110°C (~250x smaller than the peak measured for the 175°C grown film). EDX data showed that incorporation of Ge in the film decreased with higher deposition temperatures, while Te incorporation increases. SIMS profiles also indicated that silicon is present in the epitaxial film. The low growth temperature rules out Si diffusion form the substrate, and raises an interesting question as to what is the growth mechanism? The epitaxial films grown by sputtering were used to study solid phase epitaxy of GST. The top portion (~ 12 nm) of the epitaxial layer was amorphized by implanting Ge into the film. The top portion of the film was verified to be amorphous by TEM. The film was then annealed (250°C/30s in nirtogen) to re-crystallize the amorphous GST layer. TEM measurements show that the amorphous film re-crystallizes as a single-crystal GST, and no apparent interface was observed following solid phase epitaxy.

Our study focusses on the crystallization kinetics of fast-growing Sb-rich GeSb films. The crystal growth of these films was studied using a high speed optical camera. Films of 200 nm Ge7Sb93 and Ge9Sb91 were deposited on poorly heat-conducting glass or polycarbonate substrates, and capped with a 5nm SiO-ZnS layer. During nucleation and crystallization interesting crystal growth front instabilities were observed. In Ge9Sb91 films on glass substrates two competing crystal growth modes were observed in the temperature range of 181 °C to 187 °C. Below 181 °C isotropic growth is favored; while above 187 °C much faster dendritic growth is favored. Between 181 °C and 187 °C the crystal growth is initially isotropic but after certain incubation time some crystals start to grow in a dendritic manner and overtake the slower growing isotropic crystals. The crystallization fronts of Ge7Sb93 become unstable when the sample is heated above the glass transition temperature of the polycarbonate substrate. Additionally we see wrinkling of the film caused by the softening of the substrate creating a stress field parallel to the tip in the amorphous material, promoting crystal growth parallel to the stress field. After crystallization, wrinkles are formed in the crystalline tip perpendicular to the growth direction of the tip. The huge effect of stresses in the phase-change film on the crystal growth is unequivocally demonstrated by changing 2-dimensional crystal growth into 1-dimensional growth by applying compressive bending stresses. Furthermore, on a polycarbonate substrate the crystallization temperature of Ge7Sb93 is 20 °C higher than that on the glass substrate. This is caused by tensile stresses present in the film due to a difference in the thermal expansion coefficients α between the polycarbonate substrate and the phase change film. Currently we are addressing the effects of different film thicknesses on the two growth modes in Ge9Sb91. In addition, we are combining the optical measurements with electrical characterization to further investigate and quantify the influence of stresses on the crystallization process.

The local atomic order of phase-change materials is well known to deviate significantly between the amorphous and the crystalline phase. Unlike sp3-bonded semiconductors, where the bond lengths are similar in both phases, the amorphous phase of phase-change materials is stabilized by shorter bonds [1]. This can be explained by differences in the bonding mechanism: The amorphous phase employs purely covalent bonds while the crystalline phase is dominated by resonance bonding [2]. This local transition can even explain a macroscopic property: The large optical contrast. This immediately raises the question if changes in bonding mechanism should also lead to a change in the vibrational properties of the amorphous and crystalline state. Hence we have investigated the full spectrum of vibrational modes of amorphous and crystalline Ge2Sb2Te5 by inelastic neutron scattering. Additional MD-DFT calculations were performed on amorphous Ge2Sb2Te5. These calculations help to interpret the observed density of phonon states and to determine the phonon energies of specific local configurations. In addition, elastic neutron scattering data have been obtained to derive average coordination numbers and bond lengths of both phases, amorphous and crystalline. The resulting data are in perfect agreement with EXAFS data, confirming the change in bonding mechanism upon crystallization. References: [1] Kolobov, A. Understanding the phase-change mechanism of rewritable optical media. Nature Materials 3, p. 703-708 (2004). [2] Shportko, K. et al. Resonant bonding in crystalline phase-change materials. Nature Materials 7, p. 653â?"658 (2008).

We present an x-ray absorption spectroscopy based study of the electrically switched fully device integrated phase change memory material Ge2Sb2Te5 (GST225). For the first time measurements on single devices have been successfully conducted, employing synchrotron radiation based x-ray absorption (fluorescence) spectroscopy experiments utilizing an x-ray beam of about 200 nm dimension spot size at beamline BL39XU of SPring-8. The devices under investigation were either of Lance type or line cell devices with device sizes in the order of 100 nm to 500 nm. The ultra small spot size enabled us to contain the x-ray beam entirely within the dimensions of a single device allowing data acquisition exclusively from the devices without interference from the surrounding material. Employing the change in fluorescence response between the amorphous and crystalline phase of GST225 when excited with a photon energy of 11.105 keV we have recorded spatial 2D arrays of the germanium and tungsten fluorescence response investigating the state of the material and the device structure in and around the device. These maps clearly show the contrast in the state of the phase change material with regard to position. EXAFS and XANES data were collected on the crystalline and amorphous regions for structural information of the two phases and to allow characterization of the differences - if any - between the states of laser and electrically modified phase change material phases. In order to gain a more complete understanding of the spatial properties of the changed phase material in the device spatially overlapping XANES scans have been recorded along a line crossing the device. These scans track the transition from the amorphous to the crystalline phase. Up to this point we have carried out static measurements on devices previously switched into the desired state at our home laboratory. In the future (december) we will perform in situ switching of devices at the synchrotron. Synchronization of x-ray beam pulses with the switching event will allow for dynamic measurements of the switching event while adjusting the timing of the x-ray coincidence with the electrical pulse will yield information on the temporal development of the material structure during the switching process.

Phase change materials are the key ingredient for Phase Change Random Access Memory (PCRAM). Their properties influence critically the functionality and endurance of PCRAM. Depending on the application of the memory, required functionality and endurance can vary widely. PCRAM specifications for different applications include switching speed, data retention at operating temperature, power consumption for switching, threshold field, and number of switching cycles. Most successful phase change materials are based on Ge-Sb-Te alloys with variable composition, and often with dopants to modify and optimize their properties further. As much as optical phase change data storage has seen a change in the phase change storage media for each product generation a similar optimization of phase change materials for various PCRAM applications can be expected. For example, very fast materials with poor data retention but extremely high cycle numbers will be selected for a possible DRAM replacement, or moderately fast materials with very good data retention at elevated temperature and high cycle numbers are required for automotive applications. Here we report on switching properties of phase change alloys based also on the Ge-Sb-Te system with compositions GeSbxTe and GexSb2Te as well as GeTe and Ge2Sb2Te5 doped with various dopants including N, Si, Ti and Al2O3. Alloys were prepared by sputtering from elemental or compound targets, and N doping was achieved by reactive sputtering in an N2/Ar mixture. Samples were characterized by resistivity vs. temperature measurements, time-resolved x-ray diffraction and static laser testing. In addition, pore-type PCRAM devices were fabricated and the threshold field was determined. For alloys of the GeSbxTe composition it was observed that an increase in Sb fraction increases crystallization times but also leads to a reduced crystallization temperature. A compromise needs to be found for a given application so that data retention and data rate requirements can both be met. For the GexSb2Te alloys it was found that alloys richer in Ge have higher crystallization temperatures. Ge-rich alloys also have higher resistances in the amorphous phase and a larger electrical contrast. Doping of GeTe and Ge2Sb2Te5 increased the crystallization temperatures for all dopants applied. Ti doping however reduced strongly the electrical contrast and is thus not beneficial for PCRAM. Doping also modified the threshold field, leading for some materials to an increase and for some material to a decrease of the threshold field. A high threshold filed is desirable for ultra-scaled PCRAM devices because it enables reading of the device state without switching it. The highest threshold field was found for N-doped GeTe. These experiments demonstrate that variation in phase change material composition and doping are effective means to optimize the materials for different PCRAM applications.

Interfacial phase-change memory (iPCM) have been developed for greatly reducing entropic energy loss accompanied with phase transition [1]. Replacing a Ge2Sb2Te5 alloy into a multilayer of [(GeTe)2(Sb2Te3)1]n fabricated with a highly crystalline growth orientation of the two sub-layers, the switching energy could be saved by 90% compared with that in a PCRAM using the alloy. In our iPCM model, as input energy is used to transite Ge atoms in between tetrahedral and octahedral (or trigonal) states limited in a low dimension, the Sb2Te3 blocks basically do not change the structure nor exchange the atomic positions each other before and after switching. Therefore, iPCMs may provide the best platform to study topological insulators because Sb2Te3 and a special crystalline phase of Ge2Sb2Te5 were recently predicted as topological insulators theoretically [2, 3]. As reported [3], Sb2Te3 is a topological insulator, while GeTe is not. According to our computer simulations, however, the band structures of [(GeTe)2(Sb2Te3)y]z (y and z are integer) are similar to that of Sb2Te3, holding topological features, except for two spin bands attributed to GeTe crossing the Fermi level at around M or K points. More interestingly, the spin bands are lifted from degeneracy in the vicinity of Î" point with a Rashba energy, which is changed by the stacking parameter y and z. For example, [(GeTe)2(Sb2Te3)4]8 showed a giant magnetoresistivity ( >2000%) at room temperature under magnetic field (0.1T) [4], while [(GeTe)2(Sb2Te3)1]16 showed a large magneto-reflection change ( >1.0%) between the two different polarities of a magnet (0.2T), which was placed at the edge of the iPCM film, over wide visible wavelengths between 400 nm and 800 nm at room temperature. The unusual magnetic properties were never observed using control films made of the poly-cystalline phases with the same compositions. We report the details of the magnetic properties of iPCMs in the symposium. [1] R. Simpson et al. Nature Nano. 6, 501 (2011). [2] H. Zhang et al., Nature Phys. 5, 438 (2009). [3] J. Kim et al., Phys. Rev. B 82, 201312 (2010). [4] J. Tominaga et al. Appl. Phys. Lett. 99, 152105 (2011).

The rapid expansion of the Internet has lead to the consumption of vast amounts of electricity in huge server farms. Significant savings in energy and the corresponding environmental benefits could be achieved by faster and denser non-volatile memories. Very recently it was reported that superlattices of GeTe and Sb2Te3 layers show impressive switching characteristics substantially surpassing those of the composite material with the same average composition [1]. The structure switches at least twice as fast, the switching current is an order of magnitude lower, and the cyclability increases by several orders of magnitude. Those samples were grown using helicon sputtering in the amorphous form. After crystallization the samples exhibited a pronounced preferred orientation along the 111 axis of the cubic cell. The interface quality of such heterostructure might offer the possibility to tune device physical properties. However, the sputtering technique cannot reach perfect interfaces and often does not host dedicated in-situ characterization techniques. Therefore we employ molecular beam epitaxy (MBE), which combines superior thickness control with ultrahigh purity and the possibility of using a variety of in-situ characterization tools. In the following, we present the MBE growth of GeTe, Sb2Te3 and Sb2Te3/GeTe heterostructures on Si(111) substrates. GeTe and Sb2Te3 layers have been grown at several substrate temperatures. Streaky reflection high-energy electron diffraction (RHEED) patterns, which are an indication of a smooth surface, are observed for both materials. The average surface roughness was precisely assessed by atomic force microscopy (AFM) and X-ray reflectivity. Triangular shaped islands, which might indicate the presence of rotational twin domains, are observed in case of GeTe whereas for Sb2Te3 the triangular shape of the grown island is less pronounced or almost absent and the samples present lower average roughness with respect to GeTe. GeTe thin films exhibit distorted rock salt structure and grow (111)-oriented, whereas Sb2Te3 grows hexagonal with its c-plane parallel to the Si(111) surface, as shown by X-ray diffraction (XRD). Interestingly all the nominal GeTe epitaxially grown MBE samples display a composition of GeS0.46Te0.54, which is slightly off from the composition usually obtained for sputtered samples. GeTe/Sb2Te3 heterostructures were fabricated growing as first Sb2Te3 due to its overall good surface morphology, aiming at achieving a smooth interface. Several growth temperatures were used to investigate the effect of intermixing during growth. All samples were investigated by means of RHEED, quadrupole mass spectrometry, XRD and AFM. [1] R. Simpson et al., Nat. Nanotechnol. 6, 501 (2011)

GeTe is an important binary-alloy phase change material. Its high crystallization temperature relative to other commonly used chalcogenide-base phase change alloys [1,2] and rapid switching speed [2] make it attractive for random access memory applications. The crystallization of amorphous films has been studied with x-ray diffraction [1], but it is difficult to perform such experiments when the transformation occurs on the nanosecond scale, as during laser- or current-driven crystallization. In this study, 15-ns time-resolved transmission electron microscopy (TEM) during laser crystallization was performed. The dynamic transmission electron microscope (DTEM) is a unique instrument capable of nanosecond-scale time-resolved electron imaging and diffraction of transient phenomena and may be used for the study of rapidly driven phase transformations [3]. Amorphous GeTe films were laser crystallized in situ in the DTEM. Diffraction patterns with 15-ns time resolution were recorded at various times (from 0 ns â?" 3 μs) after initiation of crystallization with a laser. Additionally, time-resolved TEM images were recorded with the same time delays as used for the diffraction experiments and the crystallographic development was correlated with the transient microstructures. While the occurrence of phase transformations, such as crystallization and melting, may be used to estimate temperatures attained during in situ electron microscopy experiments, the direct measurement of the rapidly changing temperature profiles during laser-driven in situ TEM experiments is experimentally inaccessible. Hence, we will describe simulation methods used to model the spatial and temporal temperature profiles during laser heating. This work performed under the auspices of the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. This work was funded in part by the Laboratory Directed Research and Development Program at LLNL under project tracking code 08-ERD-032. References [1] S. Raoux et al., Appl. Phys. Lett.95 (2009) 143118. [2] G. Bruns et al., Appl. Phys. Lett.95 (2009) 043108. [3] T. LaGrange et al., Ultramicroscopy108 (2008) 1441-9.

Ge-Sb-Te (GST) alloys lying along the pseudobinary tie line between GeTe and Sb2Te3 are considered to be the most promising materials for phase change non-volatile memories. Conventional films are fabricated by sputtering, and thus possess a polycrystalline nature. This interferes with a detailed analysis of the switching process between the amorphous and crystalline states. Molecular beam epitaxy (MBE) of GST films offers the possibility to grow single-crystalline films, which possibly help in better understanding of the GST physical properties. The epitaxial growth of GST alloys is a challenging topic because the temperature window available for growth, i.e., from 170 to 250 °C, is low, making accurate and reproducible temperature control hard to achieve. In particular, the growth rate decreases dramatically with increasing temperature, which means that only a fraction of the supplied flux incorporates into the film. We hence monitor the kinetics of desorption in the course of growth, which allows to achieve equivalent surface conditions (temperature) on different substrates even in the presence of varying thermal coupling between sample holders and thermocouple. We use a line of sight quadrupole mass spectrometer (QMS) to monitor the desorbed Ge/Sb/Te species from the substrate. The QMS data indicate that the amount of desorbed species changes with growth time, suggesting that the temperature of the sample surface changes during the course of deposition. To achieve a homogeneous composition throughout the specimen and good reproducibility of the nominal alloy composition, the substrate temperature is therefore controlled so that the QMS signal remains constant during the entire growth run. X-ray fluoroscence measurements of GST films grown at different substrate temperatures actually confirm that a slight change in temperature of a few degrees results in a drastic change in the overall composition of the GST epilayers, with the Te concentration being independent of the substrate temperature; Sb concentration decreasing, and Ge increasing with temperature. Synchrotron radiation x-ray diffraction confirms the single-crystalline nature of the grown layers which are exclusively (111)-oriented aligning their 1-10 and 11-2 azimuthal directions with the 1-10 and 11-2 azimuths of the Si(111) substrate, respectively.

Phase change materials (PCM) are technologically relevant because the amorphous-crystalline transition, which is accompanied by changes in the optical and electrical properties, is reversible and rapid in *both* directions. While much information about the structure of crystalline phases is provided by x-ray diffraction, amorphous structures continue to provide experimental and theoretical problems. Generally accepted models for the amorphous phases of PCM have only been developed in the past few years, which has meant that the crucial crystallization mechanism has remained the subject of speculation. We have now performed the first large-scale (several hundred atoms over hundreds of picoseconds) simulations of this process in the prototype "nucleation dominated" phase change material Ge2Sb2Te5 (GST-225) and can provide details of the changes in order as crystallization proceeds. Close interplay between theory (density functional/ molecular dynamics simulations) and experiment (particularly x-ray diffraction and EXAFS) has shown that the amorphous structure of GST-225 and other alloys of Ge, Sb, and Te can be characterized by "ABAB alternation" (A: Ge,Sb, B: Te) with four-membered ABAB rings being a dominant motif. This pattern is also prevalent in the metastable (rock salt) crystalline structure, so that it is plausible that the rapid amorphous-to-crystalline transition be viewed as a re-orientation (nucleation) of disordered ABAB squares supported by the space provided by cavities. To study the process in detail, we have performed DF/MD simulations on a sample of 460 and 648 atoms at 500, 600, and 700 K for up to 600 picoseconds in each case. Crystallization has been promoted by fixing the structure of a crystalline seed" (4x4x4 sites, 58 atoms, 10% vacancies). The densities were adjusted during the simulation to allow for the difference between the amorphous and crystalline forms. Crystallization occurs at 600 and 700 K during these times. The structural changes are monitored by calculating the (partial) pair distribution functions, appropriate order parameters, the number of "wrong bonds" (bond pairs that do not occur in the ordered form), and the electronic density of states. The use of periodic boundary conditions in such calculations means that the crystallization mechanism is influenced by the presence of "seeds" in neighboring cells. This means that simulations must be performed on larger samples than those used in earlier crystallization studies.

Phase-change materials (PCMs) are among the most promising candidates for the next generation of ultra-fast, high-density digital storage technologies. An ever-increasing demand for smaller, higher-capacity electronic memory has led to the recent widespread interest in PCMs which can act as multi-state systems. In such materials, the application of different SET or RESET voltages allows various resistance levels between the amorphous and crystalline states to be obtained, enabling memory cells to store more than one bit of data. However, while it is possible to achieve a large number of levels during amorphisation, is has proven challenging to reliably obtain as many during the SET process. We have employed ab initio molecular-dynamics simulations to investigate multilevel crystallisation in the prototypical PCM, Ge2Sb2Te5. We have studied the temperature-dependent crystallisation behaviour of amorphous Ge2Sb2Te5, focussing in detail on the structural ordering during annealing, and unravelling the multilevel phase-change mechanism at the atomistic level. Intermediate states, both kinetically stable and unstable, are observed at annealing temperatures of 500 K and above. Kinetically-stable intermediate states formed at low temperatures are found to consist of crystalline-like GST structural units embedded in the bulk amorphous phase, and correspond to intermediate states obtained experimentally in GST memory devices in a narrow voltage window. At higher annealing temperatures, crystallisation proceeds via a series of kinetically-unstable intermediate states rich in planar structures, which would be difficult to isolate reliably in a practical device. Additionally, it is observed that an increase in annealing temperature leads to the progressive elimination of unfavourable chemical bonds, removing bonding defects from the material. By gaining a greater understanding of the multilevel crystallisation process, it should be possible to understand the limitations that it poses to achieving multi-state programming. This may aid the modifications of existing multilevel PCMs, and the design of novel ones, to improve the bit density of future PCM devices.

Topological insulating phase refers to a state of semiconductors that have the energy gap at bulk phase but contain helical conducting surface states. These surface states are protected from external perturbation by time-reversal symmetry. We report a theoretical finding, using first-principles methods and model Hamiltonian, of topological phase transition in the superlattice of phase change materials (PCMs). Our first-principles calculations showed that GST and GBT compounds show topological insulating phase or band insulating phase depdending on their layer sequences and thickness. We developed a model Hamiltonian that treats the superlattice of topological insulator (TI) and band insulator (BI) comprized of chalcogenide heterostructures as 1D array of interacting Dirac fermions. We showed that this interacting 1D Dirac-cone superlattice undergoes the band insulator â?" topological insulator transition depending on the TI-BI coupling strength, the mass gap, and the band gap of BI layer. These parameters are, in turn, controlled by external strain and/or layer thickness. Our finding of topological insulating phase transition in GST and GBT compounds provides not only a new approach that helps explain the electrical properties associated with the phase change but also a fundamental framework to explore many other interesting properties of PCMs and related materials.

Ge2Sb2Te5 (GST) is a popular phase-change material that is applied to the rewritable phase change optical memories and the phase change random access memory (PRAM). The fundamental understanding of phase change materials is important to resolve many technological issues. Among them, the atomic structures of liquid and amorphous GST should be the baseline in any discussion of the material properties. Several studies have been carried in this direction, but unfortunately there are serious discrepancy between first-principles results and the extant experimental observations. In particular, Ge-Te distance is longer than experimental values by 0.1~0.2 Ã. in both liquid and amorphous structures. This is closely related to the fact that semilocal functionals such as GGA and LDA tend to favor octahedral Ge atoms. Being similar to the crystalline bonding geometry, the octahedral Ge atoms lead to more delocalized electrons. However, in disordered systems such as liquid and amorphous phase, the electrons at band edges are inherently localized and chemistry-based functionals such as BLYP or Hatree-Fock can be better choices. In this presentation, we discuss on the liquid and amorphous structures of GST studied with molecular dynamics simulations using various exchange-correlation functionals such as LDA, GGA, BLYP, and hybrid functionals (HSE06 and PBE0). We find that the conventional LDA or GGA functionals produce results that are at variance with the experimental data as mentioned above. On the other hand, the hybrid functionals significantly increase the population of tetrahedral Ge atoms and the liquid structures are in much better agreement with experiment compared to results with semilocal functionals. This implies that the chemistry-based functionals such as hybrid functionals might better describe the disordered systems than LDA or GGA. The amorphous structures are obtained through the melt-quench process and hybrid functionals produce amorphous structures that satisfy 8-N rule more closely than semilocal functionals. A detailed analysis on the electronic structure will also be provided.

The unusual behaviour of thermodynamic properties (specific heat, compressibility and thermal expansion coefficients) of the eutectic Ge15Te85 alloy have led to numerous studies of this model phase-change material. We have studied its amorphous structure by combining the results of density functional (DF) / molecular dynamics (MD) simulations and high-energy x-ray (XRD) and neutron diffraction (ND) measurements. Our goal is to find a structural model that is the optimum compromise between the diffraction measurements and DF simulations. Three different atomic models of 560 atoms have been constructed: (a) Structure based on DF/MD simulations and reverse Monte Carlo (RMC) refinement without Ge-Ge bonds, (b) RMC-generated geometry with DF relaxation and without Ge-Ge bonds, and (c) as in (a) but *with* Ge-Ge bonds. For the MD simulations, the amorphous starting structure was heated to viscous regime at 500 K (below melting point, 680 K) and cooled to 300 K in 5 subsequent simulation steps of 20 ps each (simulated annealing). The RMC refinements were aimed to fit the experimental ND and XRD structure factors S(Q), while keeping the total energy near the DF minimum. All structures show semiconducting properties near the Fermi energy, and the electronic spectra are consistent with the starting structures. Structure (c) satisfied the criteria best: The agreement of S(Q) is good for both XRD and ND, total energy of the DF annealed structure is low, and there is a small band gap at the Fermi energy. Our results show that Ge-Ge bonds, even in small numbers, should be allowed. Approximately one half of Te atoms are in contact with Ge and form an extensive GeTe network, which is interlocked with Te domains (Te atoms not in contact with Ge). These domains form a similar network and are unable to grow. The high Te content means that cavities are present (22-24% of total volume). The GeTe network comprises ABAB squares, which are the basic structural units of Ge50Te50, and these units combine occasionally to form larger fragments, ABAB clusters. The Ge coordination displays a coexistence between tetrahedral fourfold coordination and defective octahedral 3+3 coordination also found in the Ge50Te50 and Ge2Sb2Te5 alloys, and both situations are equally likely to occur. The calculated bond orders confirm that there is a difference in chemical bonding between the two types of Ge. Ge-Te and Te-Te bonds are nearly covalent.

The one dimensional nano-scale building blocks have advanced in the past decade to the point where a vast range of semiconducting, metal, and oxidative materials are available for use in electronic, optical, optoelectronic, and piezoelectric nano-devices due to the structural versatility and unique chemical and physical properties. In order to realize these applications, the precise manipulation and understanding of mechanical behavior is of great importance since these define physical properties and integrated device at nano-metre scales. However, a persistent challenge has been the development of a general strategy for the manipulation of individual nanowires with arbitrary composition. The Mott transition between a low-temperature insulating phase and a high-temperature metallic phase usually occurs at 68oC in vanadium dioxide, but the compressive strain allows us to reduce this transition temperature to room temperature. Here, we show that we can manipulate ordered arrays of metallic and insulating domains along single-crystal vanadium dioxide nanowires with diameter below 100 nm by continuously tuning the strain over a wide angle of curvature.Single-crystal vanadium dioxide nanowires were fabricated by using a vapor-phase transport process. Also, through the strain dependence resistance profile, we have shown that the metal and insulator phase of vanadium dioxide nanowires are strongly affected by interplay between strain and domain nucleation. These insights open the door toward more systematic approaches to synthesis for vanadium dioxide nanostructures in desired phase and to use for applications including ultrafast optical switching, smart window, meta-material, resistance RAM and synapse devices.

Phase-change memory utilizes the electric field-induced reversible structural change in chalcogenide materials to switch between crystalline (low resistance) and amorphous (high resistance) phases to store information in a rapid and non-volatile manner. In spite of extensive investigations of the phase-change and electronic switching, the underlying mechanisms involved in the relationship between structural and electrical properties in phase change materials is quite complex and remains poorly understood. Despite some success in theoretical approaches to explain the atomic motions involved at the atomic-scale, direct visualization of electrically-driven structural transition has been experimentally challenging. This limitation is mainly because the active phase-change material in thin-film PCM devices are embedded within multiple layers, which prohibits direct probing of structural transformations on a working device with high spatial resolution while under electrical biasing. Detailed understanding of electrically-driven switching process in phase change materials is essential to achieve their successful integration into memory devices. We have performed real-time monitoring of electric-field induced switching process in phase change Ge2Sb2Te5 nanowire (NW) devices assembled on electron transparent membranes by insitu transmission electron microscopy (TEM). It is observed that short electrical pulses produce dislocations in the crystalline GST material, which become mobile above a certain applied voltage and move in the direction of hole current due to electric wind force. The continuous increase in dislocation density in the material due to the application of electrical pulses leads to dislocation jamming, which eventually leads to switching to the amorphous phase at the jammed region, where the disorder is expected to be extremely high. Our observation of solid-state, electric wind force assisted amorphization behavior sheds now insights into the origin of fast switching behavior in phase change materials, which are excellent candidates for low-power nonvolatile memory devices.

Phase Change Random Access Memories (PCRAMs) are nonvolatile memories that employ the large electrical conductivity contrast between the amorphous and the crystalline phases of so-called phase-change materials (PCMs). Phase transition is induced by proper heating of the PCM through controlled electrical pulses. Such a transition is reversible, rapid, and radiation resistant. The reset operation entails heating the PCM above its melting temperature using a relatively intense but short electrical pulse, followed by rapid quenching, resulting a highly resistive amorphous phase. The set (recrystallization) operation is accomplished by applying an electrical pulse with a smaller amplitude but longer duration to heat up the material to a temperature between the glass transition and melting points. The state of the memory can then be read by measuring its electrical resistance using nondestructive currents/voltages. In this work we present preliminary calculations and simulations to demonstrate feasibility of using a single silicon nanowire (SiNW) ballistic field-effect transistor (FET) as a heating element to program a nanoscale PCRAM cell. Memory cells based on ballistic transistors bear the advantage of having a small size and high-speed operation with low power requirements. A one-dimensional single-band Si FET model (FET toy) was used to estimate the output current of the nanowire as a function of its diameter. The gate oxide thickness was 1.5 nm, and the Fermi level at source was set to -0.32 eV. For the case of VDS = VGS = 1 V, when the SiNW diameter was increased from 1 to 60 nm, the output power density dropped from 109 to 106 W cm-2 , while the current increased from 20 to 90 μA. Thermal analysis were carried out on a cylindrical PCM cell made of Ge2Sb2Te5 (GST) chalcogenide, connected in series to the SiNW. The length and diameter ranges of the PCM cell were 5l20 nm, and 50d100 nm respectively. The electrical resistances of the crystalline and amorphous phases were calculated at room temperature and glass transition points using published data. With the assumption that the bulk thermal conductivity of GST does not change significantly at nanoscale, the required thermal powers for phase transform were calculated at glass transition temperature (380 C). For instance, the reset to set resistance ratios was Rreset/Rset = 6700 for l =13 nm and d = 100 nm, and the corresponding reset/set current ratio was Ireset/Iset = 1.44. The results presented herein can help in the design of low cost, high speed, and radiation tolerant nanoscale PCRAM cells.

The recently demonstrated scalability below 20 nm of PC-RAM [1] is a strong motivation for investigating the effect of size reduction on the Phase Change Material (PCM) structure and properties. One important property under scrutiny is the crystallization temperature and, more generally, the phase transition mechanism. It has been shown for instance that the crystallization temperature of Ge2Sb2Te5 (GST) thin films increases when the thickness is reduced below 30 nm [2]. But if studies with thin films offer the possibility to investigate confinement effects in one dimension, it is nevertheless obvious that nanometric clusters are the ideal system in order to fully assess the scalability of PCM unique properties. In this study we present results on the crystallization upon thermal annealing of GST clusters with diameter below 6 nanometers. These particles are among the smallest PCM systems to have been studied [3]. The GST clusters were grown using a sputtering gas-aggregation cluster source [4]. This technique rests on the magnetron sputtering of a solid target using high pressure argon plasma, in a liquid nitrogen-cooled growth chamber. Supersaturation in the gas of sputtered atoms results in the nucleation and growth of clusters, due to collisions and thermalization by the rare gas. This apparatus produces a narrow beam of free-flying clusters, expelled from the cluster source into a UHV chamber equipped with a deposition stage and an additional RF sputtering-head. For the present study GST clusters with sizes in the range 2 â?" 6 nm (measured in-situ by time-of-flight spectrometry) were deposited on Si substrates along with alumina. GST thin films were also deposited using the same apparatus for comparison. TEM analysis reveals that the clusters are spherical in shape and, when deposited on alumina, do no aggregate or coalesce. In order to assess the change in the structure, GST clusters were annealed under vacuum at different temperatures, from 200°C up to 400°C. Following annealing, x-ray diffraction experiments were performed at CRG-D2AM beamline (ESRF Grenoble, France). GST clusters annealed at 200°C are crystallized, while no Bragg peaks were detected in the as-deposited clusters. The Bragg peaks observed in the crystalline clusters correspond to those observed in the GST thin film annealed at 200°C and are characteristic of the cubic GST phase. In situ annealing experiments have also been performed in order to determine the crystallization temperature of GST clusters. [1] I.S. Kim et al., 2010 Symposium on VLSI Technology (VLSI), 2010; pp 203-204. [2] R. E. Simpson, M. Krbal, P. Fons, A. V. Kolobov, J. Tominaga, T. Uruga, and H. Tanida, Nano Lett. 2010, 10, 414-419. [3] H.S. Choi, K.S. Seol, K. Takeuchi, J. Fukita, and Y. Ohki, Jpn. J. Appl. Phys. 2005, 44, 7720-7722. [4] R. Morel, A. Brenac, P. Bayle-Guillemaud, C. Portemont, and F. La Rizza, Eur. Phys. J. D 2003, 24, 287-290.

Doping phase change materials with metals is known to alter their atomic structure and thermodynamic properties. Among various metal dopants, silver is particularly intriguing due to its large ionic mobility in the amorphous network. We synthesize Ag-doped Ge1Sb2Te4 films by radio-frequency sputtering with dopant concentrations up to ~ 50 at.%. The Ag-doped films before and after crystallization were analyzed by EDXS, XRD, SAXS, and EXAFS. The phase-transition properties of the as-sputtered films were characterized by temperature-dependent electric conductivity measurement. The effect of silver dopant on the structure and phase-transition properties of the Ge1Sb2Te4 films will be discussed. Molecular dynamic simulations were also performed on Ag-doped Ge1Sb2Te4 with similar compositions to the experiment. The atomic and electronic structures were determined from the simulations and a link between the experiment and theory was established. Our combined experimental and theoretical study provides structural insights into the phase-transition behavior of Ag-doped phase change materials. *Work was supported by NSF under Grant No. DMR-0906825.

Phase change material based memory (PCM) has become a promising versatile memory candidate since it is fast and non-volatile and exhibits a much higher endurance than Flash. Thus, in the memory hierarchy it can act as a hybrid between DRAM-like memory and Flash-like storage type application.[1] In order to reach high storage density for storage type application it is desirable to store multiple bits in one cell. This requires to write as many levels of resistance in a cell as still distinguishable with a reasonable bit error rate considering the noise level at a given read out speed. The drift of resistance occurring in the amorphous phase of phase change materials, thus, naturally complicates multilevel storage.[2] For memory type application good performance in terms of fast read and write speed, low reset power and high endurance is most important. Consequently, a search for new materials with low drift coefficients and fast switching properties is expedited. In this work, we present switching properties of GaSb lines bridging TiN electrodes. We show that GaSb can be switched between a high and a low conductive state with more than 2 orders of magnitude contrast. Typical performance metrics as switching speed, reset current and threshold switching field are evaluated. Special attention is drawn to the drift performance of the line cells. Our results show that GaSb exhibits very similar performance like typical phase change materials as e.g. Ge2Sb2Te5. However, GaSb is rather a conventional III-V compound with tetrahedral coordination, than a typical phase change material which have a distorted octahedral coordination with 3 short and 3 long bonds in the crystalline phase and often contain chalcogenide elements of the VIs group of the periodic table.[3] This demonstrates, that threshold and memory switching is not limited to the â?otypicalâ?[4] phase change materials and investigation should also incorporate more exotic alloys. [1] Burr et al., J. Vac. Sci. Technol. B 28, 223 (2010). [2] Krebs et al., submitted to J. of Non-cryst. Solids special issue for ICANS24. [3] Lencer et al., Adv. Mater. 23, 2030 (2011) [4] Lencer et al., Nature Mater. 7, 972 (2008).

In recent years, phase change random access memory (PCRAM) has attracted considerable research interests owing to its excellent scalability, fast switching speed and long cycle lifetime. Our group has proposed and demonstrated a new category of PCRAM based on superlattice phase change material which promises reduced switching energy and faster switching speed compared to conventional composite phase change material [1]. The details of the phase change process, however, remain to be elucidated. In order to facilitate experimental investigation of PCRAM materials using advanced structure analysis tools such as x-ray absorption fine structure (XAFS), a planar type device structure is preferable. Additionally the planar structure can provide a straightforward platform to compare the switching behaviors of superlattice and composite material devices. Therefore, in this work, we designed and fabricated a sub-micron planar type PCRAM structure by electron beam lithography and reactive ion etching, employing both composite material and superlattice material based on Ge2Sb2Te5. The device, which is similar in configuration to that proposed in ref [2], consists of a nanoscale line-cell made of phase change material and a pair of contact pads. This design allows for direct access to the switching area of the device with no interference from the non active areas, when employing a sub micron x-ray probe, such as beamline BL39XU at the SPring-8 synchrotron facility, where the phase states before and after switching were observed and analyzed. The contact pad material has been chosen to not interfere with XAFS analysis. The electric-thermal process of the device was numerically simulated using the finite element software COMSOL Multiphysics. The switching properties and cycle performance of the cells have been measured and compared. Some preliminary examinations on the results will be discussed. [1] R. E. Simpson, P. Fons, A. V. Kolobov, T. Fukaya, M. Krbal, T. Yagi&J. Tominaga, Nature Nanotechnology 6, 501 (2011). [2] Martijn H. R. Lankhorst, Bas W. S. M. M. Ketelaars and R. A. M. Wolters, Nature Materials 4, 347 - 352 (2005).

Thin films of alloys of Germanium (Ge), Antimony (Sb) and Tellurium (Te), generally referred to as GST, with various compositions and GST225 being the most notable composition have recently drawn a lot of attention as a material for novel phase change memory (PCM) application. PCM is expected to become more and more important due to its cost effectiveness based on scalability, non-volatility, fast switching speed and high write-rewrite endurance especially as DRAM and NAND-Flash run into scaling issues in advanced nodes. First GST devices based on physical vapor deposition (PVD) are already on the market, however, as we move to more advanced nodes the PVD devices will be replaced by chemical vapor deposition (CVD) or atomic layer deposition (ALD) devices, which will require planarization. Therefore, there is a need to develop advanced CMP slurries to planarize GST films before a top electrode can be deposited. The performance requirements for such CMP slurries will get more and more stringent as increasingly deeper and smaller hole or line structures have to be filled with GST alloy. Hence, CMP slurries with better defect control will have to be developed in order to protect the relatively soft and weekly adhering GST films. In this paper we will show how polishing GST films is very different from traditional metal CMP, since none of the three components of the GST film actually form a passivating oxide layer as is part of the mechanism of metal CMP for example tungsten or aluminum. This different chemical behavior will be shown particularly by electrochemical data and the first part of the presentation will focus on the basis polishing mechanism. The use of a strong corrosion inhibitor is needed in order to prevent simple chemical dissolution (corrosion) of the GST film during polishing. However, defect control in advanced hole or line structures of the GST films has become one of the foremost tasks in slurry design. We have developed CMP slurries for advanced GST applications and we have demonstrated good defect control on deep hole structures filled with GST225. We will present polishing data along with SEM images of the polished GST devices to show improvements regardidng the defect control for different slurry formulations.

Our previos studies successfully control the film composition including stoichiometric Ge2Sb2Te5 by regulating deposition condition. The deposited films had a smooth surface and it was possible to fill a high aspect hole. The mechanism to control the film composition in a high aspect hole is expounded in detail based on the precursor interferences experiment in the cavity which is macro model of a high aspect hole. We first selected the precursors having potentials for low deposition temperature, high vapor pressure, safty and low carbon impurity. Tertiarybutylgerman (t-C4H9GeH3) was selected as a Ge precursor. Triisopropylantimony ((i-C3H7)3Sb) and diisopropyltellurium ((i-C3H7)2Te) were selected as Sb and Te precursor, respectively. The precursor vapors were controlled by the H2 carrier gas flow rate, bottle temperature, and pressure. After the substrate was heated up to the deposition temperature, these precursor vapors were injected into the chamber. The deposition temperature and pressure were varied from 200 to 300C and from 1 to 50 Torr, respectively. GST materials were deposited on the SiO2/Si, TiN/Si, high aspect hole, or the cavity which was macro model of a high aspect hole. The height and width of the opening cavity were 700 μm and 20 mm, respectively. The atomic concentrations of Ge, Sb, Te and other impurities such as C in the film were measured using energy dispersive X-ray spectroscopy (EDX). The deposited films were examined by X-ray diffraction (XRD). On considering the various deposition conditions which were the amount of precursor supply, deposition temperature and pressure, we succeeded in the wide range of composition control including a stoichiometric Ge2Sb2Te5 film. Additionally, we were successful in filling a high aspect hole with GST materials in the condition which was possible to deposit a stoichiometric Ge2Sb2Te5. However, the film composition in the high aspect hole might be different from that on the surface. Therefore, we conducted the cavity experiment and measured the film composition inside the cavity along the aspect ratio. The Ge concentration increased with increasing the aspect raito, while Sb decreased. In order to evaluate the interference between Ge and Sb precursors in the atmosphere, we investigated the deposited films without Te precursor supply. The Ge and Sb compositions were hardly changed without Te supply. As well, we investigated the interference between Sb and Te, Ge and Te, and found that no composition change along the cavity was observed in the both case. Furthermore, the film composition without Ge supply resulted in the no film deposition at the same temperature. So, we confirmed that Sb and Te did not react in the atmosphere. Consequently, there were some interference between Ge and Sb precursor in the atmosphere only when Te was supply. We believe this kind of study is important to achieve filling a high aspect hole with a composition controlled GeSbTe materials.

Computers in which processing and memory functions are performed simultaneously and at the same location have long been a scientific â?~dreamâ?T, since they promise dramatic improvements in performance along with the opportunity to design and build 'brain-like' systems. This â?~dreamâ?T has moved a step closer following recent investigations of memristor (memory resistor) devices. However, as originally pointed out by Ovshinsky [1,2], phase-change materials also offer a promising route to the practical realisation of new forms of general-purpose and biologically-inspired computing. Here we provide an experimental proof-of-principle of such a phase-change material-based 'processor'. We demonstrate reliable experimental execution of the four basic arithmetic processes of addition, multiplication, division and subtraction, with simultaneous storage of the result [3]. This arithmetic functionality is possible because phase-change materials exhibit a natural accumulation property, a property that can also be exploited to implement an 'integrate and fire' neuron. The ability of phase-change devices to 'remember' previous excitations also imbues them with memristor-type functionality, meaning that they can also provide synaptic-like learning[4]. Our results demonstrate convincingly these remarkable computing capabilities of phase-change materials. Our experiments are performed in the optical domain, but equivalent processing capabilities are also inherent to electrical phase-change devices. Indeed, since phase-change materials are both optically and electrically active we can process information in either domain and even transfer information between domains. This 'extra degree of freedom' available with phase-change materials leads us to suggest a new type of mixed-mode phase-change 'memflector' (memory-reflector) device. [1] S R Ovshinsky, B. Pashmakov, Mat. Res. Soc, Symp. Proc., 803, 49 (2004) [2] S R Ovshinsky, Jpn. J. Appl. Phys., 43(7B), 4695 (2004)[3] C D Wright, Y Liu, K I Kohary, M M Aziz and R J Hicken, Adv. Mater. 23, 3408, (2011) [4] D Kuzum, R G D Jeyasingh, B Lee, H-S P Wong, NanoLetters, doi:10.1021/nl201040y (2011)

Due to their high optical contrast, chalcogenide-based phase change materials have found their major application in optical data storage in the last decade. The high electrical contrast, switching speed, scalability and lifetime promote them as one of the most promising materials for non-volatile electrical memories. We report here the investigation of the structural order of laser switched single crystal phase change materials investigated by nano-beam x-ray diffraction. The films with a Ge2Sb2Te5(GST) composition were grown on Si(111) substrates via molecular beam epitaxy. For comparison conventionally sputtered GST films grown on Silica were also investigated. Nano-beam x-ray diffraction available at beam line ID 13 at the European synchrotron radiation facility (ESRF) with a beam size of less than 180 nm vertically and 150 nm horizontally was employed. For the switching experiment a regeneratively amplified Titanium Sapphire laser with a pulse power of 5 mJ and spot size of about 500 μm was used. The film thickness was chosen to be comparable to the penetration depth of the 800 nm wavelength of the switching laser, which amounts to 20 nm. A Silicon Nitride cap was deposited on top of the GST film to avoid oxidation during switching . The diffraction signal and thus the grain size was evaluated on a nm-, as well as on the μm-scale in each spot. The evaluation of the structural properties was carried out for different spots, i.e., the as-deposited cubic phase, the switched disordered phase and the re-crystallized cubic phase. A detailed discussion on orientational preference in the different phases as well as residual order in the re-crystallized films will be given.

Phase-change random access memory (PcRAM) stands on the changes in the electrical properties between amorphous and crystalline phases of multicomponent chalcogenides. Phase-change is occurred by flowing high electrical current (or high electric field) sufficient to heat the material to a temperature above the crystallization and melting temperature. Because of partial heating by current crowding and nano-scaled device size, temperature gradient is very high in phase-change materials during the PcRAM operations. The phase-change that accompanies the change in atomic density and thermal cycling during the switching induces mechanical stress in confined volume of phase change material. Electric current (or electric field) and temperature gradient and mechanical stress can act as driving forces for atomic transport in materials. The various driving forces are simultaneously generated and mixed in PcRAM during the operation. These operational conditions lead to the atomic-transport-induced compositional change and void formation in phase change volume which are known as the origin of endurance failures of PcRAM. If the driving forces in PcRAM are understood and quantitatively evaluated, life-time will be improved by modulating the environmental conditions affect the driving forces. However, the atomic transport in PcRAM has not been systemically studied because it is hard to observe and quantify atomic transport in the device-level thereby analyzing each driving force is also difficult. Electromigration in phase-change material have been suggested as the origin of the mass transport, but these cannot fully describe the mass transport behavior. Therefore, this study considers more various driving forces including temperature gradient-, mechanical stress-, and concentration gradient-induced migration as well as electromigration. The driving forces are systemically understood and quantified by a model study using a large symmetric and asymmetric bridge device and finite difference method (FDM). We also show that the back driving force to reduce the chemical potential produced by the mass transport is critical to suppress the mass transport in phase change material. The retardation effect of the back driving forces will be discussed with the geometry of the PcRAM device and reset current. The critical condition to inhibit the mass transport is calculated by the FDM with consideration of the back forces. The systematic consideration of operational conditions with cell geometry to maximize the back forces is helpful to inhibit the mass transport-induced failure and to greatly enhance the cycling endurance of PcRAM.

The switching process in phase-change materials provides an excellent platform for the application of ab-initio techniques for the prediction of material properties due to the sub-nanosecond time scales and limited spatial extent of the switching process. To date, much effort has been dedicated to determining the details of the RESET state with the implicit assumption that the SET state is crystalline with little or no disorder. While x-ray absorption near edge scans (XANES) of the disordered RESET phase of GeTe agree quite well with theoretically calculated spectra based upon coordinates derived from a melt quench-phase structure calculated using ab-initio techniques, XANES structure from the crystalline SET phase do not. The findings suggest significantly more disorder is present in the SET phase than is current believed. Until a recent x-ray absorption fine structure (XAFS) study, it was also believed that the ferroelectric phase transition at Tc=550 K in GeTe was displace in nature with GeTe changing from a rhombohedral to a cubic structure at the Curie temperature Tc. The XAFS study demonstrated that the symmetry of the local structure retained its rhombohedral character up to and beyond Tc in apparent contradiction to Bragg (BD) diffraction measurements from which the idea of a displacive transition was proposed. The apparent conflict was resolved by the realization that BD shows average structure.In this paper, we report on a total scattering investigation of local order in GeTe and Ge8Sb2Te11 for which we measure pair correlations simultaneously over a range of length scales derived from both diffuse and coherent scattering. A simple model was fit to the measured reduced pair-distribution function that demonstrated a region as small as two rhombohedral reduced cells could account for the pair correlation for r values ranging from 2-30 Angstroms; these findings strongly suggest that disordering occurs on very short length scales. In addition to GeTe, we shall also discuss disordering in other Ge-rich alloys along the (GeTe)x(Sb2Te3)1-x pseudobinary tie-line based upon both total scattering and XAFS observations. We shall also discuss the implications of this disorder on property predictions by ab-initio techniques.

Ge2Sb2Te5 (GST) is widely used for data storage devices, in which the write operations are triggered by a thermal transition between the amorphous and crystalline phases. Therefore, to characterize thermal conductivity (κ) of GST in a wide temperature window becomes increasingly important to optimize the device performance. The 3Ï? method [1] was often used to deduce κ of GST films at room temperature while its temperature dependence has not been well investigated. Herein, we report a successful determination of absolute κ(T) of GST films in a temperature range of 30-200 deg. C by the 3Ï? method, which particularly employs the method of subtracting the substrate contribution. We observed a sharp increase κ(T) in the amorphous-cubic phase transitions around 130 deg. C, consistent with the transition temperatures in previous reports [2]. By simultaneously measuring resistivity (ρ(T)) and κ(T), we could extract the phonon and electron contribution at each temperature, and also found that both phonon and electron contributions increase with temperature increase. It was further revealed that the κ change follows the Wiedemann-Franz law upon cooling, which shows that the electron contribution dominates once the cubic structure becomes fixed after performing the high temperature treatment. [1] D Risk et al., Appl. Phys. Lett. 94, 101906 (2009) [2] J. P. Reifenberg et al., Appl. Phys. Lett. 91, 111904 (2007)

Chalcogenide alloys have specific properties that allow their use in data storage devices. The data storage concept is based on the huge change in the optical and electrical properties associated to the reversible amorphous to crystal transition induced by laser or electrical current pulses [1]. Ge2Sb2Te5 (GST) has been largely investigated since it shows a good combination of electrical/optical and phase changing characteristics for memory applications [2][3]. Crystallization of the amorphous regions will determine the maximum speed and the data retention capability of the devices. It is known that the degree of short range order in amorphous solids depends on the sample processing. Faster crystallization has been reported in ion and laser irradiated samples with respect to as deposited amorphous films [4][5]. It has been recently demonstrated by EXAFS and Raman spectroscopy that a local rearrangement of the amorphous network occurs during ion and laser irradiation promoting the system to a state closer to the crystalline phase by the breaking of homopolar â?owrong bondsâ? [6]. On the other hand, the formation of subcritical embryos, during laser irradiation, enhances the subsequent crystallization[7]. In this study we enlighten the role of subcritical embryos on the crystallization of amorphous GST in the low temperature regime (~130°C) by Transmission Electron Microscopy (TEM) and Time Resolved Reflectivity (TRR) measurements. Amorphous Ge2Sb2Te5 films, 50 nm thick, were deposited at room temperature on thermally grown SiO2 layer (85 nm) and then irradiated by pulsed (10 ns) Nd:YAG laser (λ = 532 nm second harmonic) at an energy density ranging from 0 to 180 mJ/cm2. The laser spot diameter was 3 mm. The crystallization of irradiated samples, during thermal annealing, has been investigated by TRR measurements. Another set of samples has been partially crystallized and analysed ex situ by TEM. The nucleation and growth rates as function of laser irradiation fluence have been measured at several annealing temperatures. An estimation of the sub critical nuclei abundance has been obtained. The results are of relevance for the data retention capability of phase change memories. REFERENCES [1] M. Wuttig and N. Yamada, Nature Mater., 6, 824 (2007) [2] I. Friedrich, V. Weidenhof, W. Njoroge, P. Franz, and M. Wuttig, J. Appl. Phys., 87, 214 4130 (2000). [3] N. Yamada, E. Ohno, K. Nishiuchi, N. Akahira, and M. Takao, J. Appl. Phys., 69, 216 2849 (1991). [4] P. K. Khulbe, E. M. Wright, and M. Mansurpir, J. Appl. Phys., 88, 3926 (2000). [5] R. De Bastiani, A. M. Piro, M. G. Grimaldi, E. Rimini, G. A. Baratta, and G. Strazzulla, Appl. Phys. Lett. 92, 241925 (2008). [6] E. Carria, A. M. Mio, M. Miritello, S. Gibilisco, F. dâ?TAcapito, M.G. Grimaldi, and E. Rimini, Electrochem. Solid-State Lett. (in press) [7] B.S. Lee, G.W. Burr, R.M. Shelby, S. Raoux, C.T. Rettner, S.N. Bogle, K. Darmawikarta, S.G. Bishop, J.R. Abelson, Science 326, 980 (2009)

The phase-change chalcogenide material, Ge2Sb2Te5 (GST) is technologically important for phase-change random access memory (PRAM) devices because it can be switched rapidly back and forth between the amorphous and crystalline phases by applying appropriate heat pulses. For applications in PRAM devices, the difference in switching behavior between the crystalline and amorphous phases under actual device operation conditions needs to be considered. Especially, transformations between the two phases should be examined in a time scale of nanoseconds and in dimensions of nanometers. In this study, we characterized the crystallization behaviors of as-deposited and melt-quenched GST and NGST films by measuring reflectance changes using a pulsed laser heating method. Different crystallization kinetics, such as the nucleation and growth mechanisms, were examined by TEM observations. In the as-deposited GST and NGST films, despite the longer incubation time, once abrupt and irregular nucleation occurred, the subsequent grain growth progressed rapidly, in contrast with the melt-quenched GST and NGST crystallized with many nucleation seeds in a shorter incubation time. Consequently, the grain growth progressed with a more uniform and finer grain distribution until the grains impinged. The microstructures of a melt-quenched NGST system showed the smaller grain size in the crystallized NGST sample with a higher nitrogen content.

Phase-change memory is an emerging technology and one of the promising properties of phase change memory, which could make it become competitive with Flash memory, is its very good scalability. However, the encapsulation material surrounding the GST has an increasingly dominating effect on the material's crystallization behavior. It greatly influences the limit to which GST can be scaled and as a result significantly affects the device performance. An increase in the crystallization temperature was found originating from the compressive stress exerted from the encapsulation material, while a decreased crystallization temperature was observed when tensile stress was produced by the encapsulation layer. In this study, we investigated the crystallization behavior of stressed GST films with decreasing GST film thickness. The crystallization behavior, including crystallization times, crystallization temperatures (Tx) and crystallinity were systemically measured. 150 nm W films were deposited by sputtering at 23 mTorr and 2 mTorr working pressure to induce a 790 MPa tensile stress and 1.7 GPa compressive stress, respectively, on the backside of 500 nm low stress SiN/Si substrates. 2.5 nm to 30 nm GST films were deposited on the front side of these stressed substrates and a protective 5 nm SiO2 capping layer was consecutively deposited on GST without breaking the vacuum. The W layer was removed by H2O2 chemical solvent to obtain stressed GST films. Time-resolved x-ray diffraction (XRD) was used to study the crystallization behavior of these stressed GST films. The GST film thickness does not have a large effect on the Tx for these stacks (SiN/GST/SiO2) with or without stress. However, stress has a large influence for the transition temperature from the rocksalt to the hexagonal phase (Trs-hex). Compressive stress and tensile stress inhibited the rocksalt to hexagonal phase transition when the GST film thickness was thinner than 10 nm. The Trs-hex of not stressed GST films was influenced by GST film thickness and occurred at lower temperatures when the film thickness was decreased. The 2.5 nm stressed GST still crystallized in the rocksalt phase but no hexagonal phase was observed after heating up to 450 oC. This is good news for GST scaling in PCRAM applications because the big columnar structure of the hexagonal phase that causes grain-size variation and/or void formation after high temperature (~ 400oC) back-end of line process (BEOL) is absent if the appropriate encapsulation material was chosen to induce stress into GST. Resistivity vs. temperature and crystallization times were systemically measured as well. These results provide good information about the crystallization behavior of GST under stressed conditions, especially when GST is scaled to ultimate physical limits. These insights also identify essential design criteria to enable GST phase-change memory cells to be scaled to the scaling limit.