Single pulse time-resolved X-ray scattering with the ultrabright and ultrashort X-ray pulses from the LCLS free electron laser has been used to study the phase transition dynamics in laser-excited phase change materials (GST225, AIST, and GeSb). For the frist time a complete picture of the structural changes during switching at an atomic level and on the relevant time-scales could be obtained. The response was found to be very similar for all investigated materials involving fast melting on sub-ps to ps time-scales, resolidification into an intermediate amorphous state within a few ns, eventually followed by the formation of the final amorphous or crystalline state on a tens of ns to µs-scale. These time-scales imply purely thermal mechanisms for reaching the final state despite the high degree of electronic excitation initially created by fs laser irradiation and are most likely determined by the nucleation and growth kinetics for the given material and sample geometry. Based on these results concepts will be discussed to use time-resolved X-ray scattering for a quantitative determination of nucleation and growth rates as a function temperature, parameters which are critical to understand and optimize the switching behavior in real devices.

Phase change memory devices are conventionally switched between a crystalline and amorphous phase by current-driven heating of the material. Numerous models attempt to explain the field-driven threshold switching process, whereby the conductivity of the amorphous material increases considerably under high electric fields. Knowledge of the exact nature of this threshold switching effect will help to understand the fundamental mechanisms of the switching process itself. We have applied single-cycle terahertz pump pulses of 300 femtosecond duration, with electric fields on the order of the DC threshold fields, and probed the time-resolved near-bandgap optical transmission of thin films of GeSbTe, GeTe, and GeSnTe alloys, in a repetitive non-switching regime. The amorphous films exhibited a transient decrease in transmission scaling with the square of the instantaneous terahertz field, consistent with a field-induced electrical breakdown process. This was followed by a weaker long-lived reduction in transmission, lasting hundreds of picoseconds, consistent with field-induced heating studied in optical pump-probe measurements. On the other hand, the crystalline sample had a much weaker time-zero response followed by a stronger long-lived reduction in transmission. These studies help to unravel the first steps in the electric field driven switching process in these materials.

Phase-change materials as symbolized by the prototypical phase-change alloy Ge₂Sb₂Te₅ (GST), experience large changes in physical properties in response to structural change. Typically this has utilized for memory applications via the switching of GST between the metastable crystalline and amorphous phases. The dramatic property differences that occur from the transition are due to a large change in bonding properties. The harnessing of such changes has enabled the development of optical storage (DVD-RAM) as well as the next generation of nonvolatile storage, phase-change random access memory (PC-RAM). Information storage via structure as opposed to charge is predicted to allow the future scaling of devices far beyond the limits of charge-based devices. The presence of a bonding hierarchy of long and short bonds and p-orbital based bonding in GST has been suggested theoretically to offer an ultrafast pathway to structural rearrangement via selective excitation of particular phonon modes[1]. In particular, such studies suggest that there may be energetic pathways between the two bonding states that do not require quenching from the melt, but are attainable due to excited state effects and strictly local rearrangement of bonds. Experimentally, it has been shown that light can enhance structural switching in GST [2,3] and that the use of ultrafast lasers to selectively generate coherent phonons can give rise to non-thermal structure changes. While most experimental results on coherent phonon generation have been based upon all optical pump-probe experiments, the advent of the free electron laser allows the direct probing of atomic rearrangement due to ultrafast optical pulses. Using the free-electron laser SACLA at SPring-8, we have observed the atomic rearrangement due to coherent phonon generation in GST in response to a 800 nm 30 fs CPA laser pump using a 10 fs, 8 keV ultrafast hard x-ray free-electron laser probe pulse. Here we use ultrafast diffraction to observe changes in lattice dynamics due to the generation of coherent acoustic and optical phonons by the ultrafast optical pump beam. While acoustic lattice dynamics have been modeled by solving the dynamical equations of motion for the lattice in one dimension and using the subsequent strain fields to compute x-ray diffraction intensity as a function of time using dynamical x-ray theory, there are clear indications of excited state effects that are not present for longer excitation pulses. Here we explore the possibilities of using excited state effects in combination with coherent phonon movement to switch atomic configurations on ultrafast time scales.
[1] A. V. Kolobov et al., Nature Chem., 3, 311 (2011).
[2] P. Fons et al., Phys. Rev. B, 82, 041203 (2010).
[3] K. Makino et al., Opt. Express, 19, 1260 (2011).
ACKNOWLEDGEMENT: The authors would like to acknowledge the support from the X-ray Free Electron Laser Priority Strategy Program of MEXT, Japan.

The rapid growth of the scientific interest in phase-change memory (PCM) is in large part due to its desirable characteristics such as scalability, fast writing speed and data retention. An exciting possibility for PCM is the use of non-equilibrium switching techniques. To date there have been several reports in the literature that underscore this possibility. One possible way implement non-thermal phase switching is the usage of coherent phonons dynamics. In particular by selective excitation of particular phonon modes that can give rise to a desired local change in coordination of particular atoms, it may be possible to switch between the two bonding states of PCM directly without an intermediate phase. The effective usage of this technique requires careful study that allows the following of atom dynamics on extremely short time scales.
We have carried out time-resolved ultrafast X-ray diffraction measurements at the SPring-8 Angstrom compact free electron laser (SACLA). The main goal of the current study is to explore atomic dynamics in epitaxial films of Ge2Sb2Te5 at the sub-picosecond time scale in a response to irradiation with an ultra-short (40 fs) laser pump pulse. Ten femtosecond x-ray pulses were then employed to probe collective atomic motion by monitoring transient changes in the diffraction signal. The effective time probed was varied by delaying the optical pulses with respect to the x-ray probe pulses using an optical delay line. Diffraction data was monitored by a fast readout CCD system was read out for each pump-probe pair and individually saved. In the course of data processing, the signal intensities from each diffraction pattern for specified time delays were integrated over a region of interest and plotted as a function of time. Every measurement was repeated 250 times to improve the statistics of the result. During the experiment there were two main complications making appropriate data acquisition a rather challenging task. The main challenge was signal to noise ratio optimization. In addition, data conversion, subsequent optimization of input parameters (CCD and I0 detector signals thresholds and offsets), useful data sampling and averaging required considerable computing to reduce the multiple terabyte data sets to usable data.
Among the observed transient diffraction patterns features a large drop in diffraction peak intensity occurred within a few picoseconds of the pump laser hitting the sample, while peak intensity fluctuations starting after the drop and an additional peak at lower Qz appeared after a few ps. Although the current talk will focus upon picosecond order response, in future work we will use femtosecond laser irradiation on shorter time scales as a tool for ultrafast switching back and forth between the “set” and “reset” states in Ge2Sb2Te5. These efforts will require sophisticated data process algorithms which will take into account time jitter correction.

Phase-change electronic memory utilizes the electric field-induced reversible structural change in chalcogenide materials to switch between crystalline and amorphous phases to store information in a rapid and non-volatile manner. In spite of extensive investigations of the phase-change switching phenomena, the underlying mechanisms involved in the relationship between structural and electrical properties in phase change materials are quite complex and their elucidation requires the continued development of new experimental and theoretical tools.
My group while working on the switching studies in phase change nanowires in the past several years realized that conventional understanding of melt-quench based amorphization process needed to be revisited. Nanowires, due to their long lengths (few micrometers) do not typically reach high enough temperatures required for conventional melting of the material. Furthermore, due to their cylindrical geometry, conventional melting would lead to the formation of an amorphous shell around the crystalline core, which cannot explain the rapid and abrupt resistance switching as always observed. By utilizing in situ transmission electron microscopy (TEM) technique to observe the nanowire devices with high spatial resolution while being electrically switched, we have shown that the application of electrical pulses introduces disorder in the form of dislocations, which migrate along the direction of hole-wind force and jam at a region of a local inhomogeneity, thus creating a defect-template. This defect template aids the low power switching of the device with the amorphization domain cutting through the entire cross-section of the nanowire (as against core-shell), which explains the abrupt resistance switching. We have now shown the generality of this concept on Ge2Sb2Te5 and GeTe nanowire systems. Using a simpler GeTe system, and utilizing both real and fourier space TEM imaging techniques, we have shown the evolution of intrinsic point defects to extended planar defects, and their jamming process. Through a detailed structure-property correlation, we also show that the accumulation of disorder changes the electronic properties of GeTe while still maintaining the single-crystalline long-range order, prior to solid-state amorphization- suggesting the role of electronic instabilities in the phase change process. The implication of these studies towards the design of low-power memory devices will be discussed.

Electrical wind force is an important element in switching behaviors of phase-change materials. It is hypothesized that the electrical wind force influences the degree of order-disorder existing in phase-change materials, by coupling with the motions of one-dimensional defect structures, namely, dislocations. We discuss electrical wind force-driven behaviors occurring in phase-change materials. First, we report mass transport behaviors when DC voltage biases are applied in line-shape Ge2Sb2Te5 (GST) devices. As the electrical current density reached 3-4 MA/cm2 by DC voltage bias, a directional mass transport was identified by forming asymmetric surface morphology. Joule-heating by the current density raised the temperature up to ~300 oC, implying that the mass transport of GST occurs in hexagonal phase (solid state) regime. Second, we extend the understanding about the presence of electrical wind force to the electrical switching behaviors of GST. We studied the effect of electrical voltage pulses on crystalline-to-amorphous phase transition of GST by in situ transmission electron microscopy (TEM). The electrical voltage pulse plays a critical role by creating dislocations through heat shock process: The rising edge of the pulse produces vacancies by heating, whereas during rapid cooling, atomic vacancies are condensed into dislocation loops. As the dislocations feel the electrical wind force, they become mobile and glide in the direction of hole-carrier motion. The continuous increase in the density of dislocations moving unidirectionally leads to dislocation jamming, which eventually induces the crystalline-to-amorphous phase transition. We interpreted it through one-dimensional traffic model in which an increase of the dislocation density exceeding a certain threshold point induces a catastrophic jamming of dislocations. Density functional theory (DFT) calculations of generalized-stacking-fault (GSF) energy showed that the basal plane of GST hexagonal phase provides favorable pathways of dislocation motions. Our understanding about the dislocation-templated amorphization suggests that outstanding capability of incorporating dislocations attributed to GST-layered structure is one of the origins of the fast switching behaviors.

Recently, Nam et al., [1] have demonstrated a novel low-power defect-template based solid-state amorphization route in Ge2Sb2Te5 single-crystalline nanowires. Using in situ transmission electron microscopy (TEM), it was shown that the application of electrical pulses introduces disorder in the form of dislocations, which migrate along the direction of hole-wind force and jam at a region of a local inhomogeneity, thus creating a defect-template, which aids the low power amorphization of the device. Here, we first demonstrate a defect-templated amorphization mechanism in GeTe nanowire system using similar in situ TEM techniques, proving the generality of this mechanism in nanowire phase-change memory. Furthermore, through a structure-property analysis we show that electrical pulses introduce disorder in the form of anti-phase boundaries (APB) in single-crystalline GeTe, and the migration and onset of jamming of these APBs at a local region aids efficient amorphization. To further reduce the power required for amorphization in this defect-templated approach, it would be desired to pre-disorder the single-crystalline GeTe nanowires and study their switching properties.
We used high-energy (2.2 MeV) He2+ ion-bombardment with varying dosages on single-crystalline, defect-free GeTe nanowire devices using a tandem ion-accelerator but ensuring to not amorphize the devices. The nature of introduced disorder was revealed to be vacancy clusters and APBs using TEM, whose density increases with increasing dosage; and the device resistivities at different dosages were measured to quantify the introduced disorder. We show that during subsequent switching events, the recrystallized phases are at least as disordered as the starting state at different dosages. The switching currents and voltages were observed to scale with the size of the nanowire device at all the dosages, following similar scaling laws as observed for conventional thin-film devices. We further demonstrate that with increasing pre-disorder, the switching voltages remarkably reduce for devices of similar dimensions. By pre-disordering the nanowires, the energy cost incurred for amorphization is just from the defect migration process and not defect creation, thus reducing the activation barrier and power required for the phase-change process. Therefore, the low-power defect templating strategy for amorphization is demonstrated to be a scalable strategy, whose power-efficiency can further be remarkably improved via pre-disordering the crystalline phase.
1. Nam, S.-W., et al., Science 336, 1561 (2012)

Unlike conventional solid-state devices, phase change memory (PCM) devices experience large variations in temperature and the material dynamically changes between the amorphous, crystalline and liquid phases. The gradual transitions from the amorphous to crystalline phase and between the different crystalline phases lead to local variations in material states that depend on the thermal history. Hence, modeling of PCM operation requires a component that dynamically tracks the local crystallinity state of the material.
The changes in the phase-change material can be modeled using a finite element tool by generating crystalline domains and growing them in time based on the theoretically predicted nucleation-growth dynamics [1]. However, this is computationally intensive approach as the boundaries of the nucleated domains need high-density meshing for the calculations. Alternatively, the crytallinity state of individual mesh points can be tracked and get switched between the amorphous and crystalline states in a binary fashion. However, this mesh-based approach requires a uniform mesh as variations in mesh sizes in the structure would result in variations in grain sizes. Hence, this is also computationally intensive.
We have developed a model that uses a crystal density approach where each point is tracked with a variable that represents a crystal density, rather than a binary switch between amorphous and crystalline states. Hence, the approach is immune to variations in mesh and allows non-uniform meshing of a device geometry. The model is calibrated to increase the crystal density based on the temperature-dependent nucleation-growth parameters of Ge2Sb2Te5 [1] as the material crystallizes, and rapidly decrease the crystal density upon melting. The reset-set cycling behavior of the phase change memory cells simulated for several cycles using this model is very much in-line with the experimental findings.
This model is built upon the electro-thermal models we have developed in our group which accounts for temperature-dependent material parameters, thermoelectric effects, thermal boundary resistances and field dependent electrical conductivity (dielectric breakdown) leading to threshold switching [2,3].
[1] G. W. Burr, et al. “Observation and modeling of polycrystalline grain formation in Ge2Sb2Te5” Journal of Applied Physics 111, 104308 (2012); doi: 10.1063/1.4718574
[2] A Faraclas, N Williams, A Gokirmak, H Silva, “Modeling of Set and Reset Operations of Phase-Change Memory Cells”, IEEE Electron Device Letters, 1-3
[3] N Kan'an, A Faraclas, N Williams, H Silva, A Gokirmak “Computational Analysis of Rupture-Oxide Phase-Change Memory Cells” IEEE Trans. Electr. Dev., 60 (5), 1649-1655

The rapid amorphous-to-crystalline transition is the key feature of phase change materials (PCM) which enables their function in commercial rewriteable optical disks (DVDs, Blu-ray Disc) and nonvolatile computer memory (PC-RAM). The structural phase transition is rapid, reversible, and accompanied by changes in the optical and electrical properties. The amorphous structures of PCMs are difficult to determine, and theory has played an important part in their characterization [1]. The nature of the ultra-fast crystallization mechanism remains the subject of much study and speculation, and density functional (DF) / molecular dynamics (MD) simulations have been performed recently to provide more insight. However, DF methods face severe limitations in terms of system size and time scale which call for vast computational resources and effective methods in order to improve the description of the crystallization process.
We have performed large scale density functional simulations (460 atoms for several nanoseconds) of crystallization in the prototype "nucleation dominated" phase change material Ge2Sb2Te5 (GST-225) and can provide details of the changes in order as crystallization proceeds. Our first DF/MD simulations [2] were on a 460-atom sample at 500, 600, and 700 K for up to 600 picoseconds, where crystallization was promoted by fixing the structure of a crystalline "seed" (58 atoms, 10% vacancies). The second simulation was also on a 460-atom sample at 600 K for 1.2 ns, but with no constraints on the geometry. Most recently, a further set of three recrystallization simulations were carried out at 600 K up to 3 ns, each, in order to collect more data on the stochastic process, and in particular nucleation. The density was adjusted during the simulation to allow for the difference between the amorphous and crystalline forms. Crystallization occurs in stages and was complete within 0.3-3.0 ns for successful attemps, depending on the simulation parameters (constraints, temperature) and sample history.
The structural changes were monitored by calculating the pair distribution functions, appropriate order parameters, the number of "wrong bonds" (bond pairs that do not occur in the ordered form), the variation in the cavities, and the electronic density of states. The simulations reveal the course of the crystallization process at the atomistic level, and the important stages include pre-structuring, nucleation, percolation, and final (rapid) collapse to the crystalline state. The orientation of the crystalline nucleus plays a role when the system starts to interact with its replica (other nuclei). Interestingly, the behavior of the individual elements (Ge, Sb, Te) is temperature-dependent.
[1] J. Akola and R.O. Jones, phys. stat. sol. (b) 249, 1851 (2012).
[2] J. Kalikka, J. Akola, J. Larrucea, R. O. Jones, Phys. Rev. B 86, 144113 (2012).

We performed ab initio molecular dynamics simulations of crystallization of the phase-change materials Sb2Te and Ag, In-doped Sb2Te (AIST).
Since crystallization of small amorphous bits of these compounds is dominated by crystal growth, we considered large amorphous and supercooled liquid models containing a crystallization seed and investigated the growth of the seeds at different temperatures. The models were generated by very fast quenching from the melt. Each model contained 810 atoms.
We show that, at temperatures below 550 K, the computed growth velocities for AIST are much higher than the experimental values obtained from time-resolved reflectivity measurements. This discrepancy mostly stems from the large differences between the calculated and experimental diffusivities at these temperatures. We suggest that these deviations are due to the different quenching rates, in combination with the very high fragility of the supercooled liquid phase of AIST.
At high temperatures close to 600 K, simulations are in much better agreement with experiments. These high-temperature simulations indicate that AIST and Sb2Te possess a sharp crystalline-liquid interface and that the presence of Ag and In impurities leads to a reduction in the crystal growth velocity, which is due to a decrease in the sticking coefficients at the interface.

The new generation of non volatile memories for data storage is based on a unique property of systems known as phase change materials, i.e. the super fast (ns) and reversible phase transition between the disordered and the crystalline phases. One of the reasons of the fast crystallization of GeTe-based phase change materials is the very high atomic mobility of the supercooled liquid phase, even close to the glass transition temperature. This feature is in turn a consequence of the fact that GeTe is a fragile liquid, i.e. it shows a breakdown of the Stokes-Einstein relation (SER) that relates viscosity and diffusion in the hydrodynamic regime.
In this work we investigate by large scale molecular dynamics simulations the microscopic origin of the breakdown of the SER in GeTe. To this end we employed an interatomic potential based onto a Neural Network framework that allows to overcome the limitations of conventional first principles calculations in terms of system size and timescale.
Our findings demonstrate that the breakdown of the SER is due to the presence of dynamical heterogeneities in the atomic motion. We quantified as a function of temperature the spatial extent of domains of slow/fast moving particles. The most mobile particles tend to cluster in domains that contain a significant number of chains of homopolar Ge-Ge bonds. The number and length of these chains increases with supercooling, boosting the atomic mobility and then the fast crystallization of GeTe-based phase change materials. We also found a certain degree of cooperative motion in this system, which is due to both first-shell correlated and string-like motion. Finally, we investigated the role of the domains of most immobile and mobile particles during crystallization, finding that mobile particles tend not to crystallize, but instead to flow around immobile (crystalline) nuclei facilitating the atomic rearrangement at the liquid-crystal interface.

Recently, chalcogenide-based phase change materials (PCMs) (GeTe,Ge2Sb2Te5, and etc.) have been used in memory devices such as optical memory disks and phase-change random-access memory [1]. These materials show a fast, reversible phase transition between crystalline and amorphous. To understand the crystallization process, Elliott [2] found that connected square rings are the basic building blocks in Ge2Sb2Te5 and their density can be used to detect the crystallization. It is also noticed that the umbrella-flip of Ge atoms between a tetrahedral and octahedral positions can play an important role in the transition [3].
Here we analyzed the crystallization of amorphous GeTe by an ab-initio density functional molecular dynamics simulation of atomic diffusion and the motion of square rings. A supercell of 128 atoms is used with different quenching rates from 100 K/ps to 1 K/ps and annealing temperature varied from 300K to 600K. The pair distribution function (PDF) becomes crystal-like. The sample crystallizes after annealing for 30-150ps from different runs.
First we explore the flipping of a Ge-Te bonding during the bulk crystallisation, where Ge and Te bonds exchange with each other. The transition state is studied using the nudged elastic band method. The energy barrier found to be 0.3-0.5eV depending on the neighbor environment. The difference between the two bonding states is 0.1-0.2eV.
The collective flipping of a layer of Ge sites as in an interfacial GST structure [4] is explored. The energy barrier is found to be 1.8 eV, much larger than the single flipping.
1. M Wuttig et al, Nature Materials. 6, 824 (2007)
2. J. Hegedus, S. R. Elliott, Nature Materials. 7, 399 (2008)
3. A. V. Kolobov et al, Nature materials. 3, 703 (2004)
4. R. E. Simpson et al, Nature Nanotech 6, 501 (2011)

Antimony plays a major role in many of the phase change materials (PCM) used in digital storage technology. Since their discovery [1], PCMs have been enormously successful as optical storage, including rewritable DVD and Blu-ray Disc, where GeSbTe alloys have been the materials of choice for many years. Recent years have also seen much interest in using PCM in nonvolatile memory devices[2] and the first products have been on the market for a few years. The applicability of these materials is based on the existence of two phases (amorphous and crystalline) with remarkably different optical and/ or electronic properties and the possibility of cycling rapidly (~10 ns) and repeatedly ( >1000000 times) between them. It is important for archival purposes that both phases be stable over extended time and temperature ranges[2]. While significant work has been performed on PCMs recently, there remain open questions about elemental Sb itself that deserve attention and may enable us to understand the function of PCMs better.
We have investigated previously both GeSbTe and AgInSbTe compounds and their crystallization properties[3,4] using combined density functional/ molecular dynamics (DF/MD) simulations. Furthermore, we have studied the properties of elemental Te, which like Sb, can exist in amorphous form but crystallizes readily around room temperature [5]. This work has now been extended to investigate the crystallization of amorphous antimony. The simulations have been performed using the CPMD program with the PBEsol approximation to the exchange-correlation energy. The simulation cell contained 588 Sb atoms in order to minimize effect of periodic boundary conditions and to improve statistics, and we discuss the structural properties of the amorphous and crystalline phases. These properties have allowed us to develop a model Hamiltonian to describe the system and perform kinetic Monte Carlo simulations of crystallization and amorphization on time and length scales presently inaccessible to DF/MD simulations.
[1] J. Feinleib et al. Applied Physics Letters 18, 254 (1971) and references therein.
[2] H. -S. P. Wong et al.Proceedings of the IEEE 98, 2201 (2010)
[3] J. Akola and R.O. Jones, physica status solidi (b) 249, 1851 (2012).
[4] T. Matsunaga et al. Nature Materials 10 ,129 (2011)
[5] J. Akola and R.O. Jones, Physical Review B 85, 134103 (2012).

Chalcogenide phase-change (PC) materials have been successfully used as the active layer in optical data storage media, and more recently in electronic PC memory (PC-RAM). The materials show many properties that are all critical to the performance of these memories such as reversible nanosecond timescale switching, phase stability at room temperature, and large optical and electrical contrast between the phases. The static properties of the material, as well as the phase-change kinetics are commonly tuned by adjusting the composition of the PC material.
The Ge2Sb2Te5 phase-change material lies on the GeTe-Sb2Te3 pseudobinary tie-line, and it is one of the commonly used materials in prototype PC memory devices. Pressure and stress have been shown to influence the crystallization temperature of various PC materials, including Ge2Sb2Te5. This indicates that it might be possible to tune the material properties by strain engineering, which would add a new "degree of freedom" that is independent of the material's composition to the design of PC materials with specific properties.
As PC-RAM devices are scaled to smaller dimensions to meet the demands of high density data storage, the influence of stress from electrode and capping materials will play an increasingly important role in the PC-RAM device performance. It is therefore critical to understand ways in which stress imposed from the electrode materials can be used to improve the performance of phase change materials. Herein, the effect of uniaxial stress on the Ge2Sb2Te5 recrystallization rate is studied by performing density functional (DF) / molecular dynamics (MD) simulations using the Vienna Ab-initio Simulation Package (VASP). The simulation model is a 108 atom amorphous structure that is under tension in one axis and compressed in the perpendicular axes. The atomistic evolution of the DF/MD simulation is monitored at 600 K, a temperature which permits crystallization of unstressed Ge2Sb2Te5 models. The crystallization is monitored with bond orientational order parameter, and changes can be seen in pair distribution functions and bonding topology.

Ge₂Sb₂Te₅ (GST) has been largely investigated since it shows a good combination of electrical/optical and phase changing characteristics for memory applications [1]. One of the more discussed issue about this material is the stability of the metastable amorphous state that influences data retention in phase-change memory. It has been previously observed that crystalline GST can be rendered amorphous by the application of hydrostatic pressure [2], suggesting that, in a certain regime, pressure can influence the retention of amorphous GST. Nevertheless, it was also noticed that pressures up to 5-6 GPa are expected upon melting of a confined bit due to its expansion caused by temperature increases and amorphization.
In this study, the crystallization of as sputtered amorphous GST has been investigated at several pressures and temperatures by in situ X-ray diffraction (XRD).
GST film, 100nm thick, has been deposited by RF sputtering on 100nm Poly(methyl methacrylate) (PMMA) covering a silicon wafer. GST film has been detached from the substrate dissolving PMMA in Acetone. The free standing film has been loaded on a Diamond Anvil Cell (DAC) using Argon as pressure transmitting medium. X-ray diffraction measurements were performed with a 50W Mo source (lambda;=0.71073A), collimated with a 100um pinhole, and a CCD camera. The system was also equipped with an in situ heater controlled by a PID to set an annealing temperature ranging from 25 to 400°C.
Analyses have been performed for three samples loaded in the DAC at ambient pressure (no pressure, NP), 5GPa (medium pressure, MP) and 12GPa (high pressure, HP). Each sample was annealed up to 300°C by 10divide;20°C temperature step and in situ analyzed by XRD measurements. At each intermediate temperature a XRD spectrum was recorded for 2 minutes to monitor the phase change transition. NP and MP samples crystallize in the fcc crystalline phase at about 130°C (within the experimental error of asymp;10°C) as non free-standing GST film [1] while HP sample does not crystallize up to 300°C. HP sample was then decompressed to 0GPa and annealed to 130°C, when it crystallizes in the fcc phase. Grain size measurements, performed by Transmission Electron Microscopy micrographs, have shown that crystalline grains are smaller compared to the NP annealed sample, where the grain size is about 200 nm. MP sample was also annealed up to 350°C. At this temperature the sample crystallizes in the hcp phase.
The results suggest that pressure can reduce the number of vacancies needed by the system to crystallize in the fcc phase [3] and, therefore, data retention in phase-change memories can be improved by pressure engineering, operating at a pressure ranging from 5 to 12 GPa. The measurements are then extended to cover in more details this range.
[1] Friedrich et al., J. Appl. Phys., 87, 214 4130 (2000)
[2] Kolobov et al., Phys. Rev. Lett., 97, 035701 (2006)
[3] Krbal et al., Mater. Res. Soc. Symp. Proc., 1251, 1251-H04-10 (2010)

Phase Change Random Access Memories (PCRAM) using chalcogenide Phase Change Materials (PCM), such as GeTe and Ge2Sb2Te5, are one of the most promising technologies for next generation non-volatile memories [1-2]. PCM exhibit the ability to switch reversibly between crystalline and amorphous phases with different optical and electrical properties [3]. Resistive memories based on PCM offer fast programming, good cyclability, good data retention and multi-level capability. Their expected very high scalability is one of their most promising properties. The phase transformation between the amorphous and crystalline phases is observed in very thin films (a few nm thick) [4-6] as well as in small clusters [7-8]. In devices, the PCM is in contact with insulators and metallic electrodes (Ta, TiN, Whellip;). Hence interface effects are of a great technological importance. No interface effects have been reported so far in Ge-Sb-Te films with thickness equal or above 100 nm. The measured crystallization temperature Tx in such films is then identified with the bulk material value. By contrast, strong interface effects have been reported when the film thickness is less than 10 nm.
We have investigated the effect of Ta cladding layers on the crystallization of GeTe and Ge2Sb2Te5 (GST). The effect of Ta is a relevant issue since Ta electrodes can be used in PCRAM devices. PCM films of various thickness (10, 30 and 100 nm) encapsulated by Ta have been studied by optical reflectivity, as well as similar films encapsulated by SiO2 or TiN for comparison purpose. We observe an unexpected increase of Tx in the case of Ta cladding layers with respect to the case of SiO2 layers, even for 100 nm thick films. X-Ray diffraction experiments performed on crystallized 100 nm thick GeTe films suggest different crystallization mechanism for Ta and SiO2 cladding layers. The proposed explanation differs from those most often discussed in literature that are based either on interface effects on bulk crystal nucleation [9] or strains exerted by the cladding layers [5].
[1] A. Fantini et al., 2010 IEEE International Electron Devices Meeting (IEDM), 2010, pp. 29.21.21-29.21.24.
[2] G. W. Burr et al., “Phase change memory technology”, J. Vac. Sci. Technol. B 28, 223 (2010).
[3] S. Raoux, M. Wuttig (eds.), “Phase change materials: Science and applications”, Springer, 2009.
[4] N. Ohshima, J. Appl. Phys. 79, 8357 (1996).
[5] R. E. Simpson, M. Krbal, P. Fons, A. V. Kolobov, J. Tominaga, T. Uruga and H. Tanida, Nano Lett. 10, 414419 (2010).
[6] S. Raoux, J. L Jordan-Sweet and A. J. Kellock, J. Appl.Phys. 103, 114310 (2008).
[7] G. E. Ghezzi, R. Morel, A. Brenac, N. Boudet, M. Audier, F. Fillot, S. Maitrejean, and F. Hippert, Appl. Phys. Lett. 101, 233113 (2012).
[8] M. A. Caldwell, S. Raoux, R. Y. Wang, H.-S. P. Wong and D. J. Milliron, J. Mater. Chem. 20, 1285 (2010).
[9] M. Zacharias and P. Streitenberger, PRB 62(12), 8391 (2000).

Phase change material is at the heart of phase-change memory (PCM) technology. Today almost all phase change memory IC&’s still use Ge2Sb2Te5 (GST-225) inherited from optical disk technology, even though poor high-temperature data retention due to its low crystallization temperature (Tx~150 oC) inhibit its use in some new applications such as automotive. Furthermore, many embedded system applications embrace a pre-coding procedure where system code data are pre-programmed into the non-volatile memory (NVM) before the chips are soldered onto printed circuit board. It is also impossible to pass the rigors of withstanding the 260 C soldering process using traditional GST-225 material.
Ga-Sb alloys have been studied as possible phase change materials for PCM because they show high thermal stability of the amorphous phase, high crystallization temperatures, and fast switching. In this study, we systemically explored the thin film properties of stoichiometric GaSb material in terms of usefulness for PCM applications, including crystallization times, crystallization temperatures, crystallinity and resistivity as a function of temperature. The switching and data retention behavior of GaSb was studied as well in prototype PCM devices.
Thin films were prepared by PVD sputtering from a compound Ga50Sb50 target. Crystallization temperature and electrical contrast were measured by in-situ resistivity measurements in van der Pauw (vdP) geometry during continuous heating in nitrogen atmosphere. The stoichiometric compositions of Ga50Sb50 is characterized with very high crystallization temperature indicating an excellent amorphous stability of this material. Ga50Sb50 shows two to five orders of magnitude resistance difference depending on the final heating temperature. Crystallization times were measured using a custom-made static laser tester. It was found that the stoichiometric alloy has an unusual inverse optical contrast compared to typical phase-change materials where the crystalline phase has lower reflectance compared to the amorphous phase. Moreover, Ga50Sb50 showed a very short crystallization time of around 20 ns.
The 30 nm GaSb and 30 nm TiN top electrode thin films were deposited into PCM devices with the bottom electrode size of 40 nm. A lift-off process was used to fabricate prototype mushroom PCM test devices. Almost 2 orders of magnitude SET/RESET resistance window was successfully achieved. 60 ns fast SET speed was demonstrated in the devices which is consistent with the fast switching performance from optical laser testing. PCM devices were programmed into the SET and RESET state, respectively, and exposed to the 260 oC solder bonding temperature cycle. The RESET and SET resistance both slightly increase but were still clearly separated after heating.
These results indicate that GaSb material with the stoichiometric composition is a promising candidate for phase-change memory by combining fast crystallization speed and good thermal stability.

Phase change materials are at the core of phase change memory (PCM) technology. They possess a unique combination of physical properties that allow storing and retaining data.
Ga-Sb alloys have recently be proposed as phase change materials for PCM because of their very fast switching speeds but several other material parameters are not optimized for PCM such as the low resistances in the amorphous and crystalline phases and for some alloys the low crystallization temperature.
Here we report on a new class of phase change materials which are designed by starting from the stoichiometric Ga:Sb=50:50 material and adding various materials to it. Stoichiometric Ga:Sb=50:50 has unusual properties because its electrical properties are similar to other phase change materials, but its optical contrast and mass density change behavior is opposite to most phase change materials [1-3]. The crystalline phase has a lower reflectivity than the amorphous phase and the mass density of the crystalline phase is also lower than that of the amorphous phase, as opposed to most other phase change materials. Here we report on the properties of alloys which were deposited by mixing materials that show typical phase change behavior including Sb, Ge and Si into Ga:Sb=50:50.
Addition of Sb over a wide range produces materials with with very fast recrystallization times on the order of 15 ns, but reduced crystallization temperature with increased Sb content. In particular, the Ga:Sb=30:70 material is interesting because it shows no mass density change upon crystallization but still has high electrical contrast [4]. This makes it a great candidate for improved cyclability because void formation upon cycling is one of the main failure mechanisms of PCM and believed to be caused by mass density change upon switching. The optical contrast changes from negative (Ga:Sb=50:50) to three distinct levels of reflectivity for Ga:Sb=30:70 to positive contrast for the materials with high Sb content.
Adding Si or Ge increases the crystallization temperature but also re-crystallization times. Ge has the strongest effect leading to a crystallization temperature that can be as high as 450 °C (about 220 °C for Ga:Sb=50:50), but this is accompanied by an increase in the re-crystallization time from 20 ns of Ga:Sb=50:50 to 80 ns of materials with about 50 atomic % Ge added. The alloys with Ge and Si added show the negative optical contrast observed for Ga:Sb=50:50, indicating that with the right amount of Ge and Si added also alloys with no mass density change might be found. These experiments show that by modifying phase change material composition the properties can be tuned over a wide range and optimized for specific applications.
[1] Cheng et al., EPCOS Proc. p. 103, 2011
[2] Raoux et a., Physica Status Solidi 249 (2012) 1999
[3] Putero et al., Appl. Phys. Lett. (submitted 2013)
[4] Putero et al., Appl. Phys. Lett. Mater. (submitted 2013)

Materials design and optimization are at the core of the Phase Change Random Access Memory (PCRAM) technology and often seemingly conflicting materials requirements need to be met for a successful integration. For a prototypical phase change material such as Ge2Sb2Te5 which enabled the development of CD and DVD optical storage [1], the amorphous phase exhibits low optical reflectivity, low mass density and very high electrical resistivity. In contrast, the crystalline phase shows high optical reflectivity, high mass density and low electrical resistivity. Unfortunately, large optical contrast is also associated with a large mass density change [2] leading to void formation in PCRAM cells that represents one of the main failure mechanisms [3].
In order to optimize phase change materials for PCRAM application Ga-Sb binary system was investigated by tuning the composition between Ga:Sb=9:91 (in atomic %) and Ga:Sb=45:55. Combined in situ sheet resistance measurements and synchrotron X-ray scattering techniques performed during heating enabled demonstrating a reduced crystallization temperature while Sb content increases. Besides, the electrical contrast increases with increasing Sb content and the resistivity in both the amorphous and crystalline phase decreases.
Whatever the composition, X-ray diffraction showed an elemental segregation evidenced by the detection of two crystalline phases, the rhombohedral Sb phase and the cubic GaSb phase. For alloys close to the stoichiometric composition a stepwise crystallization was observed [4] during a temperature ramp with the GaSb phase crystallizing first and the Sb one crystallizing at higher temperature. In contrast, the reverse sequence was observed for higher Sb concentrations with Sb phase crystallizing first. In between, it was shown that the Ga:Sb=30:70 alloy undergoes a simultaneous crystallization of both phases. Finally, in situ X-ray reflectivity revealed a very interesting change as Sb concentration decreases: while the mass density increases upon crystallization in Ga:Sb=9:91 (as usually observed in most of the phase change materials), it decreases in GaSb close to stoichiometric composition. At the crossroad, the composition Ga:Sb=30:70 exhibits upon crystallization neither change in mass density nor change in film thickness. This result is of primary importance for memory applications since the lack of density change may considerably reduce the mechanical stresses as the PCM programmable volume is cycled between amorphous and crystalline states.
[1] M. Wuttig and N. Yamada, Nat. Mater. 6, 824-32 (2007).
[2] Y. Saito, Y. Sutou, and J. Koike, Appl. Phys. Lett. 102, 051910 (2013).
[3] C.-F. Chen et al., in Int. Mem. Work. Monterey, CA, IEEE (2009).
[4] M. Putero et al., Appl. Phys. Lett. submitted, (2013).

A large number of potential phase-change materials have been have produced promising results experimentally; however, only Ge2Sb2Te5-based PCMs have been significantly explored so far using ab initio molecular-dynamics (AIMD) simulations. We present the first AIMD study of the full melt/quench/anneal cycle for Ga -Sb PC materials, for compositions ranging from the near-eutectic alloy Ga16Sb84, to stoichiometric GaSb. The primary changes in local environment associated with crystallisation are demonstrated, and correlated with increasing Ga content. For Ga16Sb84, it is shown that crystallisation is characterised by the transition of Ga atoms from a tetrahedral to octahedral-like coordination. In GaSb, the opposite transition occurs for Sb atoms, from octahedral-like to tetrahedral coordination. The electronic density of states and the optical reflectivity are calculated for each phase and demonstrate good agreement with experimental results.

In order to continue with the PCM roadmap, after entering into the mass production phase since last year, new advanced technology development must be faced. Considering alternative non volatile memory technologies, PCM is still one big player that is competitive through different memory segment. Nevertheless in order to be effective through different application scenarios, new phase change materials need to be evaluated on top of the well know GST225. Embedded segment and in particular automotive market is continuously demanding for larger retention capabilities without losing other electrical performances (low programming current, good read window, fast setability, high endurancehellip;) still preserving the scalability aspect.
Aim of this talk is to describe the starting point of actual PCM product inside general memory trend, highlight main issues for future technology nodes and show the main strategy to physically and electrically screen alternative phase change materials to face new markets, in particular embedded one, maintaining focus on wall architecture.
Finally some highlight on different concept like chalcogenide superlattice will be discussed as alternative way to overcome specific PCM limitations that would enable PCM as a very low power memory.

Phase change memories (PCM) are one of the most promising technology to fulfill the requirements of non-volatile memories, both for stand-alone and embedded applications. However, some challenges still remain to enlarge the application spectrum, and one of them is the low resistance SET state drift observed under soldering reflow conditions. Recently, the material GST-Ge45%-N4% has attracted lots of attention thanks to the excellent trade-off it offers between data retention and SET-RESET performances. In this study, we present a thorough investigation of both the electrical performances and physico-chemical characteristics of germanium-rich GST devices doped with nitrogen, providing for the first time a full insight of the devices integrating this specific material.
We first proceed to several material characterizations on GST thin films with variation of Ge content and nitrogen. The resistivity measurements as a function of temperature reveal the benefits of Ge enrichment on the thermal stability of the amorphous phase while the kissinger plots of GST-Ge45%-N4% also shows a boost of more than 60% of the activation energy. In addition, DRX studies show the segregation of cubic Ge at 400°C and formation of hexagonal GST at 550°C.
The GST devices with different Ge and N contents are then electrically characterized. The R-I curves obtained show a decrease of the RESET current in GST-Ge45%-N4% devices up to 33%, and an increase of the SET resistance. The data retention measurements display a drastic increase of the high resistance RESET state stability with Ge and N doping, but also highlight the loss of stability of the low resistance SET state at high temperatures, strongly worsen by doping. The influence of pulses fall times is also investigated and reveals that a resistance lower than the SET state can be reached by applying longer fall times than standard square pulses. The drift curves of this so-called SETMIN state exhibit a smaller drift than the SET state, and has been proven to be stable under soldering reflow conditions without affecting the functionality of the devices, which can still be re-programmed in the RESET & SET state afterwards.
TEM studies focus on the GST-Ge45%-N4% material : three devices are electrically programmed in the SETMIN, SET & RESET states and prepared for TEM observations by FIB etching. HR-TEM images enable shows the amorphous volume on top of the heater resulting from the RESET operation. The crystalline nature of both the SET and SETMIN states is also confirmed. EELS cartographies are then performed, providing information on the elements distribution. Those cartographies point out the segregation of elements in the SETMIN state, showing a rejection of germanium above the heater. Further analysis of diffraction patterns gathered on multiple points of interest confirm the formation of a crystalline phase above the heater which differs from the crystalline structures identified in the rest of the PCM layer.

Phase change memory (PCM) is considered as one of the most promising candidate for future non-volatile memory (NVM) technology [1]. Among the emerging NVMs, PCM is also the first which has reached the industrial maturity [2]. PCM devices are based on the reversible crystalline-amorphous transition in a chalcogenide material such as Ge2Sb2Te5 (GST). While the crystalline (set) state is stable, the amorphous (reset) state is metastable, showing spontaneous thermally-activated crystallization. Crystallization can be obtained either by increasing the ambient temperature or by electrical pulses, where the T increase due to Joule heating leads to phase transition in the 10 ns timescale. For this reason, crystallization plays an important role not only in data retention [3] and programming speed [4], but also in reset transition [5], read disturb [6] and program disturb [7].
In this work, we propose a unified finite element model able to predict electrically induced crystallization in PCM devices. The simulated structure is a mushroom cell with confined bottom electrode contact (heater). Literature values for electrical and thermal conductivities were used. To correctly describe Joule-heating in PCM from ambient temperature to melting point (around 900 K), also the T-dependence of GST electrical and thermal conductivities were taken into account [8]. The model relies on continuity and Fourier equations for electrical and thermal transport, while a first order differential equation describes the thermal activated evolution of the crystalline fraction. The fragile nature of GST glass was modeled by considering the non-Arrhenius-activated kinetic constant driving crystallization equation [9]. To model the reset state during set transition, we have described threshold switching by the formation of a highly conducting filament within the amorphous cap [9]. The model is able to predict the measured resistance R evolution in set experiments even at extremely low currents near the hold current Ih (around 40 mu;A), which is the minimum current necessary for the self-sustaining threshold switching mechanism. In addition, the model also describes crystallization in the subthreshold regime (I < 5 mu;A), where T is below 500 K and R decay in the 1000 s time-scale. This unified approach is important to predict read disturb effect in PCM devices, which must be thoroughly understood to avoid data loss under repeated read operations [6].
[1] H.-S. P. Wong, et al., Proc. IEEE 98 2201 (2010)
[2] G. Servalli, et al., IEDM Tech. Dig., 113-116 (2009)
[3] D. Mantegazza, et al., IEDM Tech. Dig. 311 (2007)
[4] D. Loke, et al., Science 336, 1566 (2012)
[5] D. H. Kang, et al., Symp. VLSI Tech. Dig., 96 (2007)
[6] A. Pirovano, et al., IEEE TDMR, 4 (2004)
[7] M. Boniardi, et al., IEDM Tech. Dig. (2013)
[8] A. Faraclas, et al., IEEE VLSI, 78-83 (2012)
[9] N. Ciocchini, et al., Trans. Electron Dev. 60, 3767-3774 (2013)

Multi-level-cell (MLC) operations of phase change memory (PRAM) require stable intermediate resistance states to minimize data errors during the lifetime of the device. It has been investigated that the amorphous state of phase change materials (e.g., Ge2Sb2Te5) possesses an inherent meta-stability, which results in resistance drift phenomena with the elapsed time after the onset of programming. Intermediate resistance states, a mixture of amorphous and crystalline phases, also exhibit the resistance drift, whose power law exponent is approximately proportional to the modeled volume fraction of the amorphous phase. Moreover, the volume fraction itself is not the only parameter that affects the resistance drift but the microstructural morphology is another control parameter for the resistance drift.
To investigate the microstructural morphology of the phase change material during multi-level resistance switching, we have analyzed line-shapes and their temperature dependences of current-voltage characteristics. These analyses led to a model of microstructural evolution during multi-level switching, which states filamentary nucleation, coalescence of nuclei in one dimensional shapes, and cross-sectional areal growth. To confirm this modeling, we have compared resistance drift characteristics in each regime of microstructures. Finally, we suggest some switching strategies to reduce resistance drift characteristics in intermediate resistance levels, and suggest some cell geometries for a potential improvement in resistance drift characteristics. We expect that our proposed modeling and suggestions would provide a promising route to reliable MLC PRAMs.

Even though phase change materials have found their way into commercial products for memory applications, some fundamental problems linked to the relaxation and crystallization kinetics remain to be solved. The inherent relaxation of the amorphous structure for example causes an increase of resistance over time, called resistance drift. This prompts a major challenge for the implementation of multilevel storage which is necessary to achieve high storage densities. On the other hand a precise knowledge about the crystallization kinetics is necessary to develop materials which allow the balancing act of ultra fast write operations but long data retention at the same time.
Previous studies[1,2] have linked changes in the density of states with the resistance increase and found an empirical relation for the time evolution of the activation energy of conduction for a constant annealing temperature Ta: Ea(t,Ta) = E0 + κTa ln((t+t0)/t1). The resistance R = R0 exp(Ea/(kBT)) increase follows with this time evolution a power law as experimentally observed. The models that have been proposed to explain this power law dependence by structural relaxation[3,4,5] are all based on a uniform distribution of activation energies for structural relaxation. This however seems unphysical since structural relaxation processes in glasses are typically attributed to α-relaxations following a single activation.[6]
In this work, we present temperature dependent resistance measurements and demonstrate how temperature and time dependence of drift can be decoupled. By describing the relaxation process with an order parameter and a single activation energy for structural relaxation we find a differential equation that can describe experimental data with arbitrary temperature and time evolution of the resistance. Furthermore we find that the empirical relation for the activation energy of conduction is a limiting case for constant temperature of this differential equation.
In order to find potential candidates to represent the order parameter in a microscopic picture, molecular dynamics simulations have been carried. They show that there is a correlation between an increase of the optical band gap and a tendency towards a local structure that reassembles the local order in the crystal. This result suggests that the processes occurring during relaxation are related to the crystallization kinetics.
[1] Krebs et al., Journal of Non-Crystalline Solids 358, 2412 (2012)
[2] Oosthoek et al., Journal of Applied Physics 112, 084506 (2012)
[3] Karpov et al., Journal of Applied Physics 102, 124503 (2007)
[4] Ielmini et al., Applied Physics Letters 92, 193511 (2008)
[5] Fantini et al., Applied Physics Letters 102, 253505 (2013)

Phase change memory (PCM) is the most recent non-volatile memory technology in the marketplace as a flash memory alternative and has the potential to become a non-volatile DRAM replacement. PCM devices work based on electrical resistivity contrast between highly resistive amorphous and highly conductive crystalline phases of phase change materials. A small volume of a phase change material (active region) switches between amorphous and crystalline phases by suitable electrical pulses. These devices experience melting and resolidification in nanoseconds time-scale and their active region reaches ~ 1000 K during the operation. Hence, high-speed and high-temperature characterization of these materials is crucial.
In this study a set of high-speed high-resolution and long duration electrical measurements were performed on nanoscale Ge₂Sb₂Te₅ (GST) line cells in a 125 K- 673 K temperature range. Electrical resistivities of metastable amorphous (above ~400 K) and metastable fcc (face centered cubic) (above ~550 K) GST were extracted. The resistance drift in amorphous phase in 250 K - 500 K temperature range and crystallization dynamics immediately after amorphization at elevated temperatures are characterized. Details of the measurement technique and results will be presented.

Ge2Sb2Te5 is the most studied phase change material for phase-change memory applications. However, temperature dependent electrical characterization of the material is not complete. In this work, we have developed a high-temperature thin-film electrical characterization setup that allows simultaneous measurement of Seebeck coefficient and electrical resistivity up to melting temperature. Our results in repeated annealing and cool-down cycles show a consistent Seebeck and electrical resistivity behavior in temperature. The correlations between these two parameters for Ge2Sb2Te5 in a wide temperature range will be presented and possible explanations of carrier transport in GeSbTe compouds will be discussed.

Since the development of dynamic random access memory (DRAM) and FLASH, there have been huge advances in memory devices that have been pioneering a new era of information and technology in society. However, the development of new non-volatile memories is needed to overcome the limitations of the existing DRAM and NAND flash.1, 2 Phase change random access memory (PCRAM) has attracted many attention as the candidates for next generation memory3. However, several critical issues related to operational power consumption need to be resolved in order for PCRAM to become a universal memory. To solve this issue, many studies have reported with regard to the electrical resistivity and melting temperature of phase change materials4. In addition, trench-structure cells have been investigated to minimize thermal energy loss and thermal cross-talk among adjacent memory cells5,6. In this research, a new screen remote plasma-enhanced atomic vapor deposition (SPEAVD) technique was studied for depositing Sb-Te phase change materials inside trench structures with high aspect ratios. We have investigated the new deposition process to increase step coverage, which is a critical problem in the deposition on small trench sizes. In our work, The theoretical model of the screen mechanism including the concepts of a plasma sheath and sticking coefficient was considered. Plasma sheath was introduced to control the flow of ions and electrons for redistribution the energy of the plasma in the reaction chamber. Our research team confirmed the screen effect of the plasma by observing the gap-filling characteristics of the screen remote-plasma enhanced AVD and the direct remote plasma-enhanced AVD. We expect this research to provide a new deposition method that allows for the fine control of step coverage and other characteristics.
[1] A.L. Lacaita, Solid-state Electron. 50, 24 (2006)
[2] S. Hudgens and B. Johnson, MRS Bull. 29, 829 (2004)
[3] M. H. R. Lankhorst, B. W. S. M. M. Ketelaars, and R. A. M. Wolters, Nat. Mater. 4, 347 (2005)
[4] L. Wu, M. Zhu, Z. Song, S. Lv, X. Zhou, C. Peng, F. Rao, S. Song, B. Liu, S. Feng, J. Non-Cryst. Solids 358, 2409 (2012)
[5] W. Wang, D. Loke, L. Shi, R. Zhao, H. Yang, L.T. Law, L. T. Ng, K. G. Lim, Y. C. Yeo, T. C. Chong, and A. L. Lacaita, Sci. Rep-UK. 2, 360 (2012)
[6] F. Pellizzer and R. Bez, IEEE. ICICDT (2012) 6232857

Amorphous state in chalcogenides is influenced by the sample preparation and processing (e.g. sputtering, melt-quenching, ion implantation, priming, pre-annealing). In particular, amorphous samples, during annealing at temperature well below the crystallization and even at room temperature, exhibit structural relaxation and defect annealing.
The relaxation strongly influences the electrical conductivity and it has been recently shown that the mobility gap for conduction has a correlation with the incubation time for crystallization.
However, it is still not clear if these relaxation effects are determined by local bond rearrangements, defect annealing or local atomic mobility changes.
In order to better investigate the structural relaxation processes, we have studied the diffusion of Au in amorphous GeTe, comparing as deposited and pre-annealed samples (annealed for several hours at 120°C in Ar atmosfere). Such a pre-annealing treatment produces the amorphous relaxation and it has been recently reported to be very effective in reducing the nucleation rate, therefore increasing the crystallization temperature.
Amorphous GeTe 100 nm thick films were deposited at room temperature (RT) by DC sputtering on a SiO₂/Si substrate. Then, Au film, 25 nm thick, was sputtered on both as deposited and pre-annealed GeTe layers. To induce the diffusion of gold in GeTe, annealings were performed at 110°C for 3 and 7 hours and at 115°C for 1 hour. X-ray Diffraction confirmed that all the thermal-treated GeTe films retain the amorphous phase. The samples have been analyzed with 2.0 MeV Rutherford backscattering spectrometry (RBS), by cross-sectional STEM HAADF micrographs and EDX microanalysis.
Gold is found to be a very fast diffuser in as deposited amorphous GeTe, even at room temperature, with diffusion coefficient extimated to be larger than 10^-16 cm²s^-1 at 25°C and 10^-14 cm²s^-1 at 110°C.
The situation is completely different in the pre-annealed GeTe film, where a concentration of 1-2% of Au into the film was observed after maintaining the samples for few months at room temperature (asymp;25°C), but no further Au migration was observed after subsequent annealing at 110°C and 115°C.
Such a result is a clear indication that the structural reordering induced by the pre-thermal treatment inhibits the Au transport in the layer. Quite interestingly, this behavior differs from that found in amorphous silicon, where thermal treatment enhances the subsequent Au diffusion. The difference should be associated probably to a different diffusion mechanism. Indeed, in amorphous silicon the defects act as traps for Au and therefore the defect reduction enhances the diffusion. In the case of GeTe it seems that Au diffusion is mediated by a large amount of void defects, available in the as deposited film. This diffusion is almost completely suppressed after relaxation, indicating that the defects mediating the Au transport have been eliminated.

We present a model for the crystallization of Ge₂Sb₂Te₅ (GST) thin films and nanostructures by simulating nucleation and growth of crystal grains. Typically, dedicated crystallization models only examine a 2D rectangular area of material in which the material is being heated uniformly throughout the sample [1-3]. Our model is developed to simulate crystallization of GST in any structure in 2D or 3D under arbitrary heating conditions where a thermal gradient and/or a transient may be present. Simulations are performed in COMSOL Multiphysics with the integration of a MATLAB function to handle the logistics of nucleation and grain growth. This approach will allow for the modeling of dynamic crystallization during device operation, allowing for updates of the electrical and thermal conductivities of the material as it crystallizes.
This model calculates a spatial map of nucleation rate in the GST for each time-step in the simulation using temperature dependent nucleation rate [3]. Crystal nuclei are generated via a probability function based on the nucleation rate map, and are then grown into crystal grains according to temperature dependent growth rate data obtained from the literature [1,3]. Nucleation is most significant at lower temperatures (~ 500 K) and diminishes at higher temperatures where growth is most significant (~ 870 K) for GST. This model can be used to predict the crystallization of GST structures during device operation, capturing the stochastic nature of nucleation and growth processes and distribution of the grain sizes.
[1] S. Senkader and C. D. Wright. "Models for phase-change of GeSbTe in optical and electrical memory devices". Journal of applied physics 95 (2004): 504.
[2] K. B. Blyuss, et al. "Master-equation approach to the study of phase-change processes in data storage media". Physical Review E 72.1 (2005): 011607.
[3] G. W. Burr, et al. "Observation and modeling of polycrystalline grain formation in Ge₂Sb₂Te₅”. Journal of Applied Physics 111.10 (2012): 104308-104308.

The GST system has been of interest since its structural transformations are associated with memory switching properties. In phase change memories (PCM) and rewritable optical disks data are recorded by switching the material from the amorphous to the crystalline phase. The heating of the material via laser or electrical current pulses of proper intensity and duration induces the reversible phase transitions. This working principle causes electrical stress due to the high current density and thermal cycling and it produces variations in the cell electrical characteristics. The aging effects are observable in the stoichiometry changes [1] and in voids formation [2]. The stoichiometry variations, produced by the high electric field, are localized near to the anode and cathode regions of the cell. In particular from recent studies, it was observed an increase of Germanium and Antinomy concentration in proximity of the cathode and a concentration increase of Tellurium in proximity to the anode areas. There is then a quite relevant interest in the investigation of new alloys with better characteristics. In this work we investigated the morphological and electrical characteristics of two chalcogenide compounds: Ge15Sb49Te37 [3 ] and Ge14Sb35Te51. Thin amorphous films, about 50 nm thick, were deposited at room temperature on thermally grown Silicon Nitride using a rf magnetron co-sputtering from elemental targets. The concentration of Ge, Sb and Te in the alloy was varied by changing the applied rf powers. The temperature resolved reflectivity curves recorded during the annealing of the two non stoichiometric alloy and a constant heating ramp of 6 °C/min was used. The enhancement of reflectivity observed during the annealing process is correlated to the transition from the amorphous to the crystalline phase. The crystallization temperature of the two non stoichiometric alloy Ge15Sb49Te37 was 190 °C, and Ge14Sb35Te51 was 165 °C. The crystallization of two alloys, Ge15Sb49Te37 and Ge14Sb35Te51, films has been studied by transmission electron microscopy (TEM) and X-ray diffraction (XRD) and was correlated to the optical reflectivity and to electrical sheet resistance. Comparison of the XRD patterns for GST, Ge15Sb49Te37 and Ge14Sb35Te51 thin films revealed that the crystallization behavior of the non stoichiometric films were quite different from that of GST. In the case of the Ge15Sb49Te37 film, no metastable FCC phase was observed, there was a phase transition from the amorphous phase directly to the HCP phase. Instead the Ge14Sb35Te51 composition films show at 200° C the coexistence of FCC and HCP phase and at 400° C the presence of a X-peak compatible with a rhombohedral symmetry.
REFERENCES
1 K. Kim and S. J. Ahn, IEEE Reliability Physics Symposium Proceedings, 2005, 157-162.
2 G. W. Burr et al. J. Vac. Sci. Technol. 28 (2), 223-262.
3 SANG-OUK RYU Journal of ELECTRONIC MATERIALS, Vol. 37, No. 4, 2008

Antimony Trisulfide (Sb2S3) has a band gap, which due to quantum confinement effects, is tunable from 1.5-2.2 eV. Consequently this material has been predominantly investigated for use in photovoltaics. Herein, the phase change properties of Sb2S3 films are investigated for new applications in data storage and nano photonic switches.
Sb2S3 films with thickness ranging from 20 nm to 1000 nm were fabricated by pulsed laser deposition and RF sputtering. The crystal nucleation and crystal growth phase transition kinetics in Sb2S3 thin films is studied as a function of film thickness, deposition parameters, interfacial materials and other external stimuli.
The presentation will describe a few experiments that demonstrate control the crystallization of Sb2S3 films using a variety of different external parameters.

Phase change memory (PCM), having just recently been commercialized in the past few years [1], has been attracting considerable attention from the scientific community. Investigations on thermoelectric phenomena, both within the phase change layer and at the junctions with the electrical contacts, have shown that asymmetric PCM cells have one operation polarity that is more favorable than the other for heating the active region [2, 3]. In this work, we have used our established thermoelectric PCM model to evaluate the merit of n-type and p-type contacts with a large contrast of Seebeck coefficients. The analysis is performed on Ge2Sb2Te5 mushroom cells. The material properties of the bottom electrode determine the thermoelectric heat release at the junction as well as thermal losses. Hence, the heater material has a significant impact on the device operation dynamics and dependence on the voltage polarity.
References
[1] J. Rice, Micron Announces Availability of Phase Change Memory for Mobile Devices: First PCM Solution in the World in Volume Production. 2012. Available:http://investors.micron.com/releasedetail.cfm?ReleaseID=692563.
[2] A. Faraclas, G. Bakan, N. Williams, A. Gokirmak and H. Silva, "Modeling of Thermoelectric Effects in Phase Change Memory Cells," Transactions on Electron Devices, Submitted with Revisions, 2013.
[3] A. Faraclas, G. Bakan, N. Williams, A. Gokirmak and H. Silva, "Thermoelectric Effects in Phase-change Memory Cells," Mat. Res. Soc. Spring Meeting, vol. H1.03, 2013.

Phase change materials (PCM) based memory device is considered as one of the most promising candidates for next-generation non-volatile solid-state memory. The set and reset states in this device correspond to a low resistance and a high resistance of the cell, which in-turn correspond to the crystalline and amorphous states of the phase change material, respectively. The total resistance of a phase change memory cell, however, consists of the resistance from the PCM and the interfacial contact resistance of the PCM to the electrodes. Although a large amount research has been done on characterization of PCM resistance, little attention is paid to study the contact resistance. Here in this work, the contact resistance of Ge2Sb2Te5 to titanium nitride (TiN) electrode has been characterized in both set and reset states using a nanowire structure obtained from spacer etch. This spacer etch is a novel technique and can be used as a low-cost alternative to E-beam lithography for sub-hundred nanometre nanowire fabrication. Unlike bottom-up technology, it is compatible with current CMOS process and the geometry and location of the nanowires can be precisely controlled. In this case it allows us make long structures with small contact area to separate the resistive contribution of bulk and interface.
A high-insulating silicon dioxide (SiO2) layer was first patterned by photolithography and etched to form a step with a depth of 100 nanometers. A 100 nm layer of Ge2Sb2Te5 was deposited by sputtering and anisotropically etched using an ion beam, leaving a spacer of Ge2Sb2Te5 next to the oxide structure.
Three different lengths (20 mu;m, 25 mu;m and 30 mu;m) of Ge2Sb2Te5 nanowires with same cross-section area (50 nm × 100 nm) were fabricated by space etching process. TiN electrodes with a thickness of 200 nm were then patterned on both sides of the nanowire by lift-off. The electrical characterization reveals the resistivity of the as-deposited Ge2Sb2Te5 nanowire material to be 0.6 Omega;m. The specific contact resistance between the TiN electrode and amorphous Ge2Sb2Te5 was extracted to be 3.59×10-6 Omega;m^2. Then nanowires were then thermally switched to crystalline state with resistivity of 3.37×10^(-4) Omega;m and specific contact resistance of 7.07×10-9 Omega;m^2. Even for these very long wires, the Roff/Ron ratio of 1.78x10^3 is partially determined by the contact resistance. These results indicate that for real memory cell layout, the contact resistance is the dominant factor in Ge2Sb2Te5 phase change memory devices.

The ever increasing demand for a universal memory which combines rapid read and write speeds, high storage density and non-volatility is driving the development of new memory concepts and materials. Phase change materials based random access memory (Phase Change RAM) has emerged as a leading candidate for the next generation of non-volatile memory. However, the critical issue of thermal cross-talk between adjacent cells when scaling down the cell size is yet to be solved. Growth of the entire device inside a contact hole may be more favourable as it could reduce the thermal cross-talk and simultaneously reduce the power required for the switching operations. The conventional deposition of phase change materials by sputtering does not allow selective deposition and is unable to uniformly fill small holes. Hence, alternative deposition approaches need to be investigated.
Here we report the selective deposition of phase change materials using chemical vapor deposition (CVD) with new, custom-made single source precursors. CVD is well established as a deposition process with most processes using dual or multiple sources, Being able to deposit alloys using single source reagents can be advantageous as it can offer improved stoichiometry and fewer defects, as well as often being safer and easier to handle. More importantly, through these single source reagents area selective deposition behavior by CVD was discovered. This selectivity is observed when depositing materials onto patterned substrates which contain TiN “holes” in a SiO2 film. A set of binary phase change materials, SnSe2, Ga2Te3, Bi2Te3 and Sb2Te3 have all demonstrated highly selective deposition behavior on these substrates.
The characterization of phase change materials will be presented. All as-deposited materials are crystalline and of a high purity. The properties of all materials are studied using scanning electron microscopy (SEM), energy dispersive X-ray spectroscopy (EDX), X-ray diffraction (XRD), Raman spectroscopy and Hall measurements. When depositing the materials onto patterned substrates, the deposition occurs preferentially onto the exposed TiN (conductive) area inside the holes while leaving the outside SiO2 (insulator) bare. This selective deposition behavior can be observed in micro-patterned (2 µm to 100 µm) or even nano-patterned (100 nm to 1000 nm) holes. The possible reasons for this selective behavior will be discussed.

Thermal conductivities of chalcogenide thin films, Ge₂Sb₂Te₅(GST) and Ce-doped GST (Ce-GST), were measured by the 3-omaga (3omega;) method and their thermal boundary resistances (TBR) properties at the interface of chalcogenide and TiN contact layer were analyzed in terms of the plywood sample structures. Analytical results were consequently implanted in a finite-element simulation in order to analyze the thermal and electrical characteristics of phase-change memory (PCM) devices. The simulation based on a three-dimensional electric and thermal coupling model indicated that the TBR property significantly affects the temperature profile and the heating efficiency of PCM cells subjected to a pulse heating operation.
In PCM cell utilizing doped GST, i.e., Ce-GST, as the programming layer, a better thermal confinement effect was observed when the same amount of heat was generated during programming. This is ascribed to the increment of resistivity in Ce-GST which results in the decrease of programming current in PCM cells and thermal conductivity of Ce-GST layer. In the analysis regarding of TBR effect, dissipation of programming power caused by the inhibition of heat propagation from bottom electrode to programming layer of devices was observed. Simulation indicated the presence of TBR_GST/TiN = 7.6×10^minus;8 m²°K/W in PCM cell containing pristine GST leads to a 30% decrement of programming power whereas the presence of TBR_Ce-GST/TiN = 9.6×10^minus;8 m²°K/W in PCM cell containing Ce-GST leads to a 27% power decrement in comparison with the cases without the TBR effect. The influences of device scale-down and TBR properties on the programming conditions of PCM cells were also examined. It found that the smaller TiN contact width benefits the increment of cell temperature. The TBR effects caused the 16% and 19% increments of set and reset currents in PCM containing GST whereas the increment of set and reset current are 19% and 18% for PCM containing Ce-GST.

Electrical properties of conductive-bridge random access memory (CBRAM) devices containing Ge₂Sb₂Te₅ (GST) chalcogenide thin films as the solid-state electrolyte, silver (Ag) as the active electrode (AE) and tungsten-titanium as the counter electrode were investigated. The correlations of electrical properties with the microstructures and compositions of CBRAM samples analyzed by conductive-atomic force microscopy, Auger electron nanoscope and transmission electron microscopy are also presented.
Electrical measurement observed no resistive switching behaviors in the samples containing amorphous GST. This implies the crystallinity of GST is essential to the resistive switching behaviors of CBRAM samples and the grain boundaries in polycrystalline GST should play a key role in the evolution of conduction channels since they might serve as the fast transport paths for Ag elements to achieve the resistive switching behaviors. The sample with the best electrical performance (V_Set = minus;0.14 V; V_Reset = 0.24 V; R-ratio asymp; 511) and satisfactory retention and cycleability properties was achieved by the insertion of ZnS-SiO₂ dielectric layer at the Ag/GST interface and a post-annealing at 250°C for 30 min. In addition to the resistive memory characteristics of GST, electrical measurement clearly illustrated the advantage of low operational power of GST-CBRAM devices. Moreover, the polarity reversal phenomenon was observed in the sample containing the ZnS-SiO₂ layer. Microstructure and composition analyses indicated that ZnS-SiO₂ insertion likely reverses the role of electrodes in the sample when electrical bias is applied, consequently altering the Ag distribution in GST and the polarity of such a CBRAM device. Analytical results also revealed the gradient distribution of Ag in GST and, hence, the formation and rapture of Ag conductive channels in the vicinity of AE/GST interface are responsible to the resistive switching behaviors of CBRAM samples.

The speed at which phase change memory devices can operate depends strongly on the crystallization kinetics of the amorphous phase. To better understand factors that affect the crystallization rate, we have carried out basic studies of crystallization of Ge2Sb2Te5 and GeTe films as a function of how they are made. We found that the crystallization kinetics varies strongly with the temperature of the substrate on which the film is deposited. We argue that this reveals characteristics of the atomic scale structure of the film, allowing correlation of the atomic scale structure with crystallization kinetics of films with the same composition.
1 µm-thick Ge2Sb2Te5 and GeTe films were deposited by sputter deposition on Si substrates with a thin layer of native SiO2. Different samples were deposited at room temperature, 60oC, 80oC and 100oC.
Crystallization of these films was monitored using x-ray synchrotron radiation (lambda;=0.124 nm) while samples were annealed at a constant rate of 2oC/min, 5oC/min or 10oC/min from room temperature to temperatures at which they were fully crystalline. The fraction of the film that had transformed was determined as a function of time for different heating rates to provide transformation curves. This allowed calculation of the effective activation energy for crystallization using Kissinger&’s method. The apparent crystallization temperature was defined as the temperature at which half of the sample was crystallized. The microstructures of the samples were characterized using scanning electron microscopy, energy dispersive x-ray spectroscopy and transmission electron microscopy.
For Ge2Sb2Te5 samples, the apparent crystallization temperature and the effective activation energy for crystallization were found to decrease with increasing deposition temperatures. Microscopy revealed that as-deposited samples were composed of small crystallites (<10nm) embedded in an amorphous matrix for all deposition conditions, with the number of crystallites increasing with increasing deposition temperature. The shapes of the transformation curves suggest that crystallization occurred though growth of these pre-existing crystallites. The increased number of pre-existing crystallites in films deposited at high temperature causes the apparent crystallization temperature to decrease with increasing deposition temperature.
The apparent crystallization temperature and effective activation energy of GeTe films were also found to decrease with increasing deposition temperature. However, the GeTe samples were found to be amorphous in the as-deposited state. The transformation curves also indicate that nucleation was required for crystallization. We suggest that the decrease in the apparent crystallization temperature with increasing deposition temperature correlates with different degrees of ordering to form sub-critical nuclei in the as-deposited films. This is supported by analysis of the measured transformation curves.

PRAM devices have demonstrated great advantages for next generation memories owing to the fast speed, high retention and high resistance margins between amorphous and crystalline states. Recently, many researches have been carried out to realize PRAM as Flash memory due to its merits on low fabrication costs compared with other resistance based memory, and also its scalability below sub 20 nm design cells. In order to achieve high density of PRAMs for Flash memory applications, MLC (multi-level cells) characteristics should be realized.
In this study, we investigate the Se-doped GeSb as the phase change material for MLC since it shows larger ratio of Rreset/Rset than GeSbTe material. We observed the rapid decrease of sheet resistance from amorphous to crystalline and the higher sheet resistance ratio between amorphous and crystalline states than conventional GST. When the cell structures are fabricated with Se-doped GeSb films, shorter crystallization time and lower SET voltages are demonstrated compared with the one with conventional GST. Based on our results, we conclude that the Se-doped GeSb has the feasibilities as MLC phase change materials for next generation PRAM applications.

Phase change random access memory (PCRAM) is a great candidate for the next generation computing because of its great scalability, fast speed, and low power consumption. Especially, one-dimensional (1-D) chalcogenide nanostructures (e.g., GeTe, Sb2Te3, AgSbTe) are commercially utilized with the expectation of high performance in virtue of device architecture and excellent phase-transition properties of the materials. So far, extensive researches have fulfilled the field by enhancement of switching speed, endurance, and scalability with conventional techniques. However, none of studies systemically investigates effect of composition whereas the composition change during the switching can additionally affect the PCM performance in addition to amorphous-crystal structure transformation. Moreover successful applications of chalcogenides rely on synthesis techniques that can reproducibly fabricate hierarchical and dimensionally uniform nanostructures with controlled crystallinity and composition in a cost effective manner. Electrodeposition is a high-yield, cost-effective and versatile process that operates near room temperature, has low energy requirements, and is capable of handling complex geometries at a variety of length scales. Furthermore, it is able to synthesize solid solutions and non-stoichiometric alloys based on its non-equilibrium reaction that is difficult using other processes.
In the work, template-directed electrodeposition was utilized to synthesize antimony telluride (SbxTe1-x) nanostructures with well-controlled dimensions, morphology and composition. In order for study on crystal structure and phase transition of synthesized SbxTe1-x nanowires, XRD and TEM analysis were conducted under thermal treatment. In addition, electrical transport properties of individual SbxTey nanowires such as temperature-dependent resistivity, temperature coefficient of resistivity(TCR), activation energy(Ea) and FET mobiltiy were correlated with observed solid-state structural transformation to investigate effects of Sb content and phase.

Optical recording discs are used extensively as rewritable nonvolatile high-density memory storage media (DVD-RAM, DVD-RW, Blu-ray Disc). The most widely used recording materials are the GeTe-Sb2Te3 (GST) alloys, particularly Ge2Sb2Te5.[1] This material is used also for the phase change RAM (PC-RAM), which is a contender of Flash memory in computers and electronic devices.
In DVDs, laser light induces a transition between the amorphous and crystalline states by heating the recording layer, and the phase change in PC-RAMs is caused by heating that follows an applied voltage. The state is monitored by measuring changes in the reflectivity or resistivity. Low-power operation of phase change memories (PCM) has been difficult to achieve, but one possible solution in a PC-RAM cell is to lower the heat loss by increasing the thermal resistance between the GST layer and the heater electrode. Extremely small dimensions and low set and reset currents have been achieved recently by using carbon nanotubes (CNT) as memory electrodes.[2] The use of CNTs or graphene as the interface material should then reduce the heat loss significantly.
We have studied thin films of GeTe and GST in graphene sandwich structures by simulating the amorphous and crystalline phases of PCM layers. The systems comprised an approximately 17 Å thick layer of GeTe, a 21 Å thick layer of GST, and a single sheet of graphene, where both PCM layers were in rocksalt lattice (111 face). Geometries and orientations were optimized with density functional (DF) (CP2K software) and molecular dynamics (MD) (CPMD software) methods. A partial charge analysis of the optimized structures shows no charge transfer between the PCM and graphene. In addition, the sandwich structure with amorphous GeTe shows graphene undulation, and the PCM experiences the graphene as a physical "wall".
Electronic structures and phonon dispersion curves of the systems were computed with DF method (Quantum Espresso software). The electronic band structure, density of states (DOS) and the partial DOS in crystalline GeTe showed that the Dirac Cone of the graphene is unaltered - no band gap or shifting in relation to the Fermi energy. This is further evidence that there is no charge transfer between the layers. We plan to compare the features of GeTe and GST and also the effect of the phase change.
[1] J. Akola, R. O. Jones, Phys. Rev. B 76, 235201 (2007)
[2] J. Liang, R.G.D. Jeyasingh, H.-Y. Chen, and H.-S.P. Wong, 2011 Symposium on VLSI Tech., 100 (2011).

The amorphous chalcogenide phase change materials (PCM) have a lot of interesting electron transport properties, e.g. the sub-threshold linear (exponential) current-voltage curve when the bias is low (large), the threshold switching, and the current-voltage curve snap-back. Interestingly, the stat-of-art PCM scaling research has found that these peculiar electron transport properties exist not only in the bulk PCM (tens of nm or larger) measurements, but also in the ultra-scaled PCM nanostructure (sub-10 nm) measurements [1]. However, though the electron transport of the bulk PCM has been well studied [2-3], the understanding of the electron transport of ultra-scaled PCM nanostructures is still missing. It is important to study the electron transport properties of the ultra-scaled PCM nanostructures, which are crucial to enable ultra-dense PCM device technologies.
In this study [4-5], we investigate the sub-threshold electron transport properties of the prototypical PCM GeTe ultrathin film in the amorphous phase, by using the ab initio molecular dynamics, density functional theory, and non-equilibrium Green&’s function simulations. Our purely ab initio simulations reproduce the linear (exponential) shape of the measured current-voltage curve for low (large) bias in the sub-threshold region. We find that, when the bias is small, the transmission coefficients are not significantly dependent on the bias. And, the linear shape of the current-voltage curve is a manifestation of the bias window enlarging. In contrast, when the bias is large, the transmission coefficients is significantly bias-dependent. The exponential shape of the current-voltage curve is jointly determined by the bias-induced change of the transmission coefficients and the bias window enlarging. Our simulation results also successfully reproduce other measured physical properties, like the bandgap, intra-bandgap donor-like and acceptor-like defect states, and the p-type conductivity of the amorphous GeTe. We hope our simulation results and conclusions will help better understand the electron transport behavior in ultra-scaled PCM nanostructures, and improve the design of ultra-scaled PCM devices.
Reference:
[1]. Kim, S, Bae, B.J., Zhang, Y., et al, IEEE Trans on Electr Dev, 58, 1483-1489, 2011.
[2]. Ielmini, D, Phys. Rev. B, 78, 035308, 2008.
[3]. Ielmini, D, and Zhang, Y, J. Appl. Phys., 102, 054517, 2007.
[4]. Liu, J., Xu, X., and Anantram, M.P., “Ab Initio Simulation of Sub-threshold Electron Transport Properties of Amorphous Germanium Telluride Phase-Change Ultrathin Film”, (in preparation).
[5]. Liu, J., and Anantram, MP, J. Appl. Phys., 113, 063711, 2013.

The state-of-art experiments have found that the crystallization of phase change material (PCM) can be accomplished within about 0.5 ns [1]; and the PCM crystallization capabilities can be kept in sub-2 nm PCM nanostructures [2]. These unique crystallization properties (ultra-fast phase switching and superb scalability) of PCM are indispensable to enable ultra-fast and ultra-dense nonvolatile PCM memory devices. Also, they have spurred intensive research interests, to understand the governing physics of the PCM crystallization [1,3,4,5]. Though a lot of physical insights have been obtained in previous studies [1-5], the quantitative analysis of the various physical quantities (Gibbs free energy density, interfacial energy density, specific heat capacity, critical formation energy, critical nuclei radius, etc.) that determine the crystallization properties of PCM and the temperature dependence of these physical quantities are still missing.
In this study [6], these physical quantities and their temperature dependence are calculated and analyzed in a multi-scale way, using density functional theory, ab initio molecular dynamics, and classical thermodynamic theory. Here, we focus on prototypical PCM GeTe. We show that, as temperature rises, the Gibbs free energy density difference between amorphous phase and crystalline phase monotonously decreases; and the amorphous/crystalline interface energy density monotonously increases. We quantitatively unveil the importance of the elastic energy in PCM crystallization, which is largely ignored in the existing PCM crystallization studies. Our analysis reveals that the elastic energy plays an important role in determining various crystallization properties and the ultimate scaling limit of PCM. By omitting the elastic energy, the critical formation energy (critical nuclei radius) will be underestimated by 41.7% (22.4%) and the nucleation rate will be overestimated by 74.2% when the annealing temperature is 600 K. We find that the PCM crystallization properties and the PCM scalability are jointly determined by the Gibbs free energy density difference, interfacial energy density, and elastic energy density. Our results show that the critical nuclei radius of the crystalline cluster is smaller than 1.4 nm when the annealing temperature is lower than 600 K, indicating extremely promising scaling scenario of the PCM technology.
Reference:
[1]. Loke, D. et al, Science, 336, 1566-1569, 2012.
[2]. Raoux, et al, J. Appl. Phys., 103, 114310, 2008.
[3]. Hegedus, J. et. al, Nature Mater., 7, 399-405, 2008.
[4]. Kalikka, J. et al, Phys. Rev. B, 86, 144113, 2012.
[5]. Matsunaga, T, et al Nature Mater., 10, 129-134, 2011.
[6]. Liu, J. and Xu, X. and Brush, L. and Anantram, M.P., “A Multi-scale Analysis of the Crystallization of Amorphous Germanium Telluride Using Ab-Initio Simulations and Classical Crystallization Theory”, J. Appl. Phys. (submitted).

Based on the unique threshold switching phenomenon of amorphous chalcogenide glasses [1], the Ovonic Threshold Switching (OTS) is being studied for the potential application in cell selector in memory devices. For the development of a reliable OTS device, the chalcogenide glass must have properties such as high crystallization temperature, low threshold voltage, and high on/off ratio, most of which are closely related to the nature of bonding between constituent atoms. In this regard, understanding the correlation between electronic structures and the bonding nature in chalcogenide glasses is very important in material design for implementing the high-performance OTS devices.
We investigated the electronic and vibrational structure of the Ge1-xSex (xle;0.67) and Ge1-x-ySexAsy amorphous thin films by using spectroscopic ellipsometry and Raman spectroscopy. Amorphous thin films of Ge1-xSex and Ge1-x-ySexAsy were prepared on Si substrate by cosputtering method using two or three targets of Ge, Ge0.4Se 0.6, and Ge0.4As 0.6 at room temperature.[2,3] Using spectroscopic ellipsometry, we measured the ellipsometric angles at room temperature in the spectral range of 0.8 - 6 eV, from which we determined the dielectric functions of the amorphous thin films. The Raman data were obtained using defocused 488 nm with 15 mW laser power to prevent photo-induced crystallization. Resonance Raman phenomenon at 514 nm (laser power 0.2 mW) was compared to other laser excitation wavelengths. We determined the optical gap energies and Urbach energies from the absorption coefficients, and found a strong positive bowing effect for the optical gap energy for Ge1-x-ySexAsy where the endpoint binaries were Ge0.50Se0.50 and Ge0.31As0.69. We attributed the strong bowing to the mixing of lone pair electron states in valence bands associated with As and Se atoms. We found that As-As mode in Ge0.31As0.69 was transformed to As(Se1/2)3 mode in Ge1-x-ySexAsy, and finally Se cluster mode in Ge0.5Se0.5 as Se content increased. Particularly, we observed resonant Raman phenomenon from Ge0.38Se0.62 at the laser excitation of 514 nm (2.41 eV). We verified that this laser energy corresponded to the transition energy of Ge0.38Se0.62 using the second derivative of the dielectric function of Ge0.38Se0.62. The Raman peak 169 cm-1 (ethane-like mode of Ge2(Se1/2)6) and 283 cm-1 (stretching modes of Se chains and rings) at off-resonance shifted to 172 cm-1 and 292 cm-1 at resonance, respectively, whereas 200 cm-1 peak (corner-sharing mode of Ge(Se1/2)4) did not shift. The finding of resonant Raman scattering in amorphous films suggests that the electronic interband transitions associated with short range order can cause the resonant Raman phenomenon.
References
[1] S. R. Ovshinsky, Phys. Rev. Lett. 21, 1450 (1968).
[2] H.-W. Ahn et al., Appl. Phys. Lett. 103, 042908 (2013).
[3] S.-D. Kim et al., ECS Solid State Letters 2, Q75 (2013).

Phase change materials such as Ge2Sb2Te5 are characterized by large volume change upon crystallization, which could negatively affect the performance and reliability of phase change memory devices. Moreover, the phase change memory cell is confined between metals and dielectric layers, making crystallization-induced structural change more severe. Hence, it is essential to understand the effects of interfacial layers on the structural volume change of the film. In this study, structural change as a result of crystallization of amorphous Ge2Sb2Te5 films by thermal annealing was characterized by atomic force microscopy (AFM). A cross sectional area analysis method is used for evaluation of the structural volume change. Ge2Sb2Te5 film with and without dielectric cap layer were considered. Ge2Sb2Te5 film was sputtered deposited on a four-point test structure for evaluation of phase change through electrical measurement. Reduction in cross sectional area at 130oC and 200oC for SiO2 capped and non-cap films were observed. This reduction corresponds to the phase transitions from amorphous to FCC phase structure and FCC phase to HCP-phase, respectively, as revealed by the X-ray diffraction (XRD). Electrical resistance measurements were also correlated to the phase transitions. The thickness and cross sectional area reduction for SiO2 capped film for phase transition from amorphous to FCC phase structure and FCC phase to HCP phase is relatively higher. This is believed to be associated with impact of the SiO2 film on the crystallization temperature and also possible due to the large thermal expansion coefficient (TEC) difference between SiO2 and Ge2Sb2Te5 where SiO2 has the lowest thermal expansion coefficient (TEC), which is about 1/20 of the Ge2Sb2Te5. AFM allows precise and non-destructive determination of the phase transformation via observing the physical change of the film not only in one dimensional but rather in three dimensions. The present setup allows the understanding of the effect of dielectric cap layer on the volume change of the phase change material by the cross sectional area analysis method. Moreover, the results show the importance of the dielectric layer for the design of phase-change memory devices.

Recently, amorphous material has been receiving much attention both theoretically and practically because of varied material properties that are particular distinguishing features unlike crystalline materials. In order to generate the amorphous structure theoretically, various methods have been used; melt-quench method based on molecular dynamics (MD) is the most favored method since it resembles experimental melt-quench procedure. Because the interatomic force fields are not known in many cases, the ab initio approach based on the density functional theory (DFT) is a popular method of choice. However, the computational cost of DFT-MD is too expensive and typical supercells comprise only about 100-200 atoms. For more realistic modeling of amorphous structures, it is critical to simulate on the large cell.
In this work, we propose a new method to generate the amorphous structure using the local geometry obtained from DFT-MD. Although amorphous structures do not have long-range order, they have short-range order at the atomic length scale due to the chemical bonding. By considering the local order such as coordination number, bond length, and the type of the bonded atom, we were able to generate a reasonable amorphous structure with much smaller computational cost than conventional melt-quench method. To improve the medium-range order, an annealing simulation is carried out below the melting point. We compare the final structures obtained by this procedure with melt-quenched structures for various types of materials, and the agreements are reasonably good.

PRAM( Phase-change Random Access Memory ) has been spotlighted due to its fantastic characteristics such as good shrinkability, high endurance, long-time retention and so on. Finally PRAM has been commercialized for mobile phone. Naturally, main interest for PRAM is moving from possibility of production to its usage and aplication-driven technology. In this presentation, I will share my idea about the evolution of PRAM applications which will be evolved into cost effective interface, innovative system and brain-inspired computation with respect to efficiency of energy-saving. And the advanced technologies related to the evolution of applications will discussed too.
Effective Interface: PRAM can be operated by bit-by-bit and at low voltage compared to conventional Flash memories. These advantages make PRAM possible to share the interface of central processing unit (CPU) and DRAM with system&’s simplicity and effectiveness. This kind of application is being commercialized in some IT systems. Furthermore, global researchers and developers are focusing on other advantages such as fast latency and better reliability in order to find out novel applications [1].
Innovative System: Legacy computer architectures consist of CPU, DRAM as main memory, and NAND or HDD as storage. Unfortunately there has been a big latency gap between memory (<100ns) and storage (~100us), which masks the system be complicated and inefficient. When introducing the higher speed and better endurance of PRAM to reduce the latency gap, a new fusion technology so called memorage (a combination word of memory and storage) is emerging. This innovative fusion system can make computing system more cost-effective, more energy-saving, and much faster.
Brain-Inspired Computation: PRAM has very recently started to search new application for brain-inspired computing, which belongs to the kind of fusion technologies between electronics and neuroscience. This may cause paradigm shift in computation method from von Neumann type to cognitive-based algorithm [2]. I will show that phase change memory cell array is able to emulate synaptic functions by introducing novel core algorithms, which can realize more energy-saving IT systems as way of making PRAM to be adaptable in neuromorphic applications.
References
[1] Youngdon Choi et. al., ISSCC digest of technical papers (2012) 46.
[2] M. Suri et. al., IEDM digest of technical papers (2012) 235.

Phase Change Memory (PCM) utilizes the large electrical resistivity contrast between the amorphous (high resistance) and the crystalline (low resistance) phases of chalcogenides. These materials can be reversibly and rapidly switched between the two phases by self-heating via electric pulses [1]. Understanding the crystallization dynamics during set operation is critically important. A baseline (offset) voltage plays an important role in set operation achieved by melting the amorphous region and utilizing growth-from-melt templated from the crystalline regions [2][3].
In this work, we experimentally investigate the effect of baseline voltage on crystallization behavior of nanoscale Ge₂Sb₂Te₅ (GST) line cells by applying a voltage pulse (melting pulse) with varying baseline voltages. After the melting pulse, the GST wires stay partially molten (retention of a current carrying filament) and recrystallize, or they resolidify as amorphous, depending on the applied baseline. Our electro-thermal models using finite element simulations also show the same current-time characteristics and capture the filament formation and crystallization dynamics. Details of the electrical measurements, simulation results, and analysis will be presented.
References
[1] H. -. P. 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] D. Loke, T. Lee, W. Wang, L. Shi, R. Zhao, Y. Yeo, T. Chong and S. Elliott, "Breaking the speed limits of phase-change memory," Science, vol. 336, pp. 1566-1569, 2012.
[3] M. Cassinerio, N. Ciocchini and D. Ielmini, "Logic Computation in Phase Change Materials by Threshold and Memory Switching," Adv Mater, 2013.

Nanoelectronic devices - and their constituent materials and interfaces - present some of the most promising and challenging opportunities for the study of thermal transport in the solid state. Phase change memory devices provide a particularly compelling example owing to the importance of both electrons and phonons in one of the crystalline phases of the functional chalcogenide material (e.g., 1). Additional complications (and opportunities) arise from the impacts of thermoelectric transport and thermal boundary resistances on the key figures of merit of memory devices, such as the time and energy required for bit writing and the bit stability (e.g., 2).
This presentation will describe a multi-year investigation of thermal transport in GeSbTe compounds using thermoreflectance and Joule-heating methodologies as well as transmission electron microscopy, x-ray diffraction, and a novel “micro thermal stage” that facilitates incomplete transitions between the crystalline and amorphous phases. While the findings have been primarily helpful with the design of phase change memory, they are also enabling more exploratory research on brain-inspired computing, field-programmable gate arrays, and a solid-state thermal switch.
1Wong, Raoux, Kim, Liang, Reifenberg, Rajendran, Asheghi, Goodson, "Phase Change Memory," Proceedings of the IEEE 98, 2201-2227 (2010).
2Lee, Asheghi, Goodson, "Impact of Thermoelectric Phenomena on Phase-Change Memory Performance Metrics and Scaling," Nanotechnology 23, 205201 (2012).

The presentation will focus on two main topics: 1. Crystallization kinetics of GexSb1-x films (50 and 200 nm thick with x = 6, 7, 8, 9 and 10) studied using mainly high speed optical microscopy; 2. Electrical and structural characterization of PRAM line cells using a 2-point probe technique and scanning and transmission electron microscopy.
In part 1 Alloys ranging from Ge6Sb94 to Ge10Sb90 were studied and it was found that crystallization properties change dramatically with small changes in Ge concentration. Crystal growth rates were measured over five orders of magnitude (100 nm/s to over 1 cm/s). Using a viscosity model a fragility m=60 for Ge8Sb92 and Ge9Sb91 was found and based on this model extrapolation of the experimental data led to a maximum crystal growth rate of ~15 m/s at a temperature of ~90% of the melting temperature. Moreover, the large influence of moderate stresses on crystal growth was directly observed by using a four point bending stage in combination with high speed optical microscopy. Finally, two competing growth modes were found in Ge8Sb92, Ge9Sb91 and Ge10Sb90 alloys with growth rates that differ more than an order of magnitude while external conditions are kept constant.
In part 2 the evolution of many characteristics of single line cells during their cycling up to 100 million cycles is shown. We also present extensive work done on the intricate time and temperature dependence of the amorphous resistance of the line cells as influenced by drift. Finally, the correlation between electrical properties and the nanostructure of the line cells is shown.

Chalcogenide materials undergoing reversible crystal-to-amorphous phase transitions continue to play a key role in information technology. In particular, phase-change memory (PCM) has emerged as the most promising new nonvolatile solid-state memory technology. A key property that makes these materials attractive for such applications is the crystallization process that occurs at the nanometer length scale and the nanosecond time scale. Moreover, in spite of this fast crystallization rate at elevated temperatures, the melt-quenched amorphous state is stable at lower temperatures for several years. Given the length and time scales associated with this unique crystallization process, experimental measurements are extremely challenging.
Most experimental investigations have been limited to low temperatures, at which the crystallization rate is small. A recent effort to measure the crystal growth rate using ultrafast differential scanning calorimetry was reported by Orava et al. [1]. These measurements were conducted on as-deposited phase-change materials. A more recent effort involved the use of laser-based time-resolved reflectivity measurements [2]. But both techniques were limited in terms of the temperature range over which crystallization rate was measured.
In this talk, we present the measurement of crystallization rate with a PCM cell using electrical means. This approach is capable of measuring the crystallization rate down to nanometer dimensions as well as on the nanosecond time scale thanks to, respectively, the dimensions of the PCM cell and the fast thermal dynamics associated with such cells. We present a consistent description of the temperature dependence of the crystal growth velocity in the glass as well as the super-cooled liquid from room temperature up to the melting temperature. Isothermal measurements are conducted to measure the growth velocity in the glass state, which is shown to follow an Arrhenius relationship. The melting temperature, where the crystallization rate is zero, and the temperature at which this rate is maximum are estimated by exploiting the spatial and temporal evolution of temperature in the PCM cell created by the application of suitable voltage pulses. A cryogenic probe station to vary the ambient temperature and finite-element models to support the experimental findings are used. The overall temperature dependence of the growth velocity is quantified using these experiments in combination with an estimate of the evolution of the effective amorphous thickness as a function of the application of voltage pulses.
[1] J. Orava et al., Nature Materials 11, 279 (2012)
[2] M. Salinga et al., Nature Communications 4, 2371 (2013)

We report on a new platform for studying the scaling of a phase change material (PCM) memory bit, enabling electrical as well as structural characterization of a PCM cell formed between point electrodes with gaps as small as 2 nm. In this work the platform was used to study memory bits made of Ge₂Sb₂Te₅ (GST) with dimensions as small as 10×10×2 nm. A record low I_SET current of 0.4 mu;A was obtained.
Conductive 10 nm wide and 10 nm thick Pd nano-wire electrodes were fabricated over a thin silicon nitride membrane using e-beam lithography and a lift-off process. The 50 nm thick silicon nitride film served as a TEM membrane to allow the imaging of the PCM bit before and after setting of the bit by passing electrical current. Optical lithography and metallization was used to define probe pads with connections to the nano-wires. The nano-wires were coated with a thin Al₂O₃ layer and gaps cut with dimensions as small as 2 nm were made in each of the nano-wires using a focused high-energy electron beam. An amorphous GST film was blanket deposited over the structure by physical vapor deposition (PVD) from a GST 225 target. The deposited GST was verified by TEM to fill the gap that was cut in the nano-wires. Since the nano-wire electrodes were coated with Al₂O₃ prior to the gap formation, electrical contact to the deposited PCM material took place only at the exposed metal in the cut gap. This allowed the fabrication of PCM memory bits with dimensions as small as 10×10×2 nm.
Current-Voltage (I-V) sweeps forcing current and measuring voltage were taken to set the PCM bits. A record low set current (I_SET) of 0.4 mu;A for switching the device to the low resistance, crystalline state, was achieved. The formation of a crystalline phase change material that bridges the gap between the wires was verified in the switched devices by TEM. Additional DC current sweeps verified that the bit maintained the SET state.

One of the most attractive merits of PCM technology is its superb scalability. The state-of-art scaling experiments have shown that the sub-2 nm PCM nanostructures still can keep phase change properties [1]. Simulations and experiments have shown that the PCM devices, downscaled to 3.8 nm [2] and 6 nm [3] respectively, can offer enough ON/OFF ratio. The promising scaling scenario unveiled in these studies have spurred intensive interests to develop ultra-scaled PCM devices, to offer ultra-dense data storage solutions. It is known that inelastic electron-phonon scattering plays an important role in determining electron transport properties of bulk PCM (tens of nm or larger) [4-5]. But the understanding of its role in ultra-scaled PCM (sub-10 nm) is still missing. It is important to investigate this problem, in order to offer deeper physical insight for designing ultra-scaled PCM devices.
In this work [6], we study the electron transport in ultra-scaled amorphous GeTe sandwiched by TiN electrodes. The electronic structure of the TiN-GeTe-TiN model is obtained using density functional theory; the electronic transport is simulated using non-equilibrium Green&’s function (NEGF) [2]; and the inelastic electron-phonon scattering is accounted for using Born approximation. Our results show that, though the measured transport properties (e.g. current-voltage curve, negative differential resistance, and threshold switching) of ultra-scaled PCM highly resemble those of bulk PCM, their governing physical mechanisms are totally different. In bulk PCM devices, it is known that the transport is largely diffusive (Ohm&’s Law), partially due to inelastic electron-phonon scattering. Furthermore, the inelastic electron-phonon scattering plays a crucial role to excite the trapped electrons so that they can participate in the transport (Poole-Frenkel Law) [3-4]. In contrast, we find that the electron transport in the ultra-scaled PCM nanostructures is largely elastic. For the ultra-scaled 6 nm GeTe ultrathin film used in our simulation, less than 5% of the energy carried by the incident electrons is transferred to GeTe lattice when they are transported from source to drain. This indicates that, in the ultra-scaled PCM nanostructures, the inelastic electron-phonon scattering exerts limited influence on the electron transport process and the total current. Therefore, the Poole-Frenkel effect does not play a role in ultra-scaled PCM.
Reference:
[1]. Raoux, S et. al, J. Appl. Phys., 103, 114310, 2008.
[2]. Liu, J. et. al, J. Appl. Phys., 113, 063711, 2013.
[3]. Kim, S. et. al, IEEE Trans. on Electr. Dev, 58, 1483, 2011.
[4]. Ielmini, D, Phys. Rev. B, 78, 035308, 2008.
[5]. Ielmini, D, et. al, J. Appl. Phys., 102, 054517, 2007.
[6]. Liu, J., Xu, X., and Anantram, M.P., “Analysis of the role of inelastic electron-phonon scattering in ultra-scaled GeTe ultrathin film”, (in preparation).

The miniaturization of memory devices is one of the major driving forces in the development of ever faster, more efficient and more compact electronic devices. However, pushing towards smaller, more densely packed memory bits can cause the overlap of neighboring data points in the classical thin-film mushroom cell geometry of phase change random access memory (PCRAM) devices. Therefore, different architectures, such as incorporating the phase change material into a nanostructured insulator, have to be considered. The preparation of materials inside nanostructured substrates proves challenging for the widely used sputtering techniques and new ways of preparing the phase change materials must therefore be investigated.
We present an alternative materials preparation approach based on the electrochemical deposition of phase change materials, such as SbTe and GeSbTe from soluble precursors that provide sources for the individual elements. Electrodeposition as a preparation technique is already well established in the electronics industry through the Damascene process which is used for the preparation of copper vias in microchips. In the case of phase change materials the electrochemical approach allows the direct deposition of the binary or ternary alloy materials into a nanopatterned device substrate consisting of a TiN bottom electrode covered by a patterned SiO2 insulator. The bottom-up nature of electrodeposition prevents the issue of void formation during material deposition. Besides the advantages provided by direct materials preparation within nanopatterned substrates the electrochemical approach also allows tuning of the material composition, for example, by adjusting the amount of the Ge, Sb and Te precursors in the deposition bath.
We will present an electrochemical system used for the preparation of the SbTe and GeSbTe phase change materials and characterize the prepared materials using scanning electron microscopy (SEM), energy/wavelength dispersive X-ray spectroscopy (EDX / WDX) and X-ray diffraction (XRD). The materials are electrodeposited either as a thin film for materials characterization or within nanopatterned substrates prepared by electron-beam lithography. The control over the material composition within the GeSbTe composition triangle will be discussed. We will also present methods for the accurate filling of nanopatterned substrates with hole sizes ranging from a few micrometers down to less than 50 nanometers. The electrical switching and the related quality of the obtained materials within the nanopatterned substrates will be discussed.

The needs for storing and utilizing ‘big data&’ are growing, and resistive switching non-volatile memory (NVM) has been widely investigated for replacing flash memory. A charge-injection GeTe/Sb2Te3 superlattice (SL) phase change memory (PCM) is proposed as a candidate NVM. Charge-injection PCM is “non-melting”, and its SET and RESET conditions differ from those of conventional Joule-heating (“melting”) PCM. Non-melting PCM operates on the basis of Ge atomic movement. Ge&’s 4-fold state is the RESET state, a high resistance state (HRS), and the 6-fold state is the SET state, a low resistance state (LRS). Non-melting PCM&’s theoretical programming energy was calculated to be less than that of melting PCM. First principle calculations verified that the difference in resistance was caused by the differences in local DOS and the transmission coefficients of HRS and LRS. In the RESET state, electrical conduction channels are disrupted by the isolation of GeTe layers from Sb2Te3 layers, resulting in HRS. First principle calculations also showed that the movement of Ge was driven by charge injection. The transition from LRS to HRS must be driven by electron injection, and the transition from HRS to LRS requires electron removal or hole generation. Previous works showed experimental evidence of non-melting resistance changes. Analysis by TEM showed the structural stability of GeTe/Sb2Te3 SL after an endurance of 1M cycles. A thermal analysis showed that the transition of GeTe/Sb2Te3 SL was different from the fcc to hcp transition of GeSbTe. Charge injection enhancement was supported by the observation of negative resistance states (NRS) (electron removal or hole injection) in SET operations. A DC sweep RESET also proved the charge injection enhancement since the RESET of melting PCM was only possible with an AC pulse.
The stacking films of GeTe/Sb2Te3 must be carefully deposited for acquiring true performance of an accurate SL. Previous work with the optimized deposition technique showed that RESET current density was less than 10 MA/cm2, and the RESET voltage applied to the SL film is less than 0.5 V. Furthermore, a high quality GeTe/Sb2Te3 SL PCM film exhibited a RESET current of 70 uA and a SET speed of 10 ns when the contact electrode was 50 nm in diameter. These devices exhibited an endurance of 100 M cycles and multi-level cell (MLC) capabilities. The MLC operations were investigated by observing the NRS.
In conclusion, this paper describes the current status of theoretical and experimental understanding of charge-injection GeTe/Sb2Te3 PCM. First principle calculations demonstrate the charge-injection enhanced Ge atom movement. Experimental results prove that non-melting resistance change and high-quality GeTe/Sb2Te3 films enable low-voltage (RESET < 0.5 V) and fast (SET = 10 ns) operations. An endurance of 100M cycles and MLC capabilities are also achieved.

Interfacial phase change memory (iPCM) has recently attracted many attentions on the point of view of green nanoelectronics and spintoronics. Although the original objective for the invention was to save switching energy based on the entropic energy loss in PCM, several unusual properties were discovered later, for example, electrical-induced giant magneto-resistance. Recently, it is speculated that the magnetism is attributed to a topological state of iPCM structures. iPCM is consisted of a multilayer, which has building blocks of Sb2Te3 and GeTe sub-layers, and the former is a 3D-topological insulator (TI). When thin TI layers are alternatively stacked with thin normal insulating layers alternatively, a Dirac gapless state on one TI layer&’s top-surface and another Dirac gapless state on the adjacent TI layer&’s bottom-surface are coupled each other. If the insulating layer is thick enough, the coupling effect is weak and negligible, resulting in all TI layers play as a TI independently, while it plays as a bulk insulator if the thickness is thin enough because the helical spin states from both surfaces rotate oppositely, resulting in an open gap. Therefore, the Dirac gapless edge of the bulk in iPCM is fragile to each layer thickness and external field such as electrical and magnetic field. In addition, the RESET phase of iPCM has both spatial inversion symmetry and time reversal symmetry, which the SET phase only has time reversal symmetry. Therefore, the former may have a reversible Rashba effect to the external electrical field, while the latter may have a permanent spin-splitting band structure. It is interesting that in iPCM these two states can be switched at a threshold voltage applying an external electrical field. Based on these theoretical predictions, one may design a new spintronics device completely dissimilar to a spin-torque MRAM.
In presentation, we explain the details of the mechanism using ab-initio simulations and provide several experimental proofs and applications.

Recently, interfacial phase change memory (iPCM) has been designed and fabricated using a GeTe/Sb2Te3 superlattice structure and superior performances such as extremely lower power switching with a much faster switching speed were demonstrated compared with a PCM using a conventional Ge-Sb-Te alloy film. Besides such the excellent performances for the memory application, it was discovered that iPCM showed electrical-induced magneto-resistance effect more than 2000% in spite of using only non-magnetic elements, which have been speculated to a relation to topological insulators. Although the iPCM has been attracting considerable attention as both electrical non-volatile memory and magnetic device, the material research has not yet been carried out sufficiently. In this work, as an alternative intercalated layer holding the topological properties of Sb2Te3 blocks, we select a SiTe layer instead of a GeTe layer and report the structural and electrical properties of [(SiTe)x/(Sb2Te3)y]z (x, y and z are integer) structure films theoretically and experimentally. In order to investigate the structural stability and corresponding band structure, two kinds of ab-initio density functional theory codes, called CASTEP and WIEN2K, were used. Ab initio calculations predicted that (SiTe)2/(Sb2Te3)n (n: 1,2,4 and 6) structures could stably exist with a Dirac semimetal-like band structure. Based on the simulations, we made (SiTe)2/(Sb2Te3)4 structure film on Si single crystal by RF-magnetron sputtering at elevated temperature using SiTe and Sb2Te3 alloy targets. It was found that all the X-ray diffraction (XRD) peaks were attributed to the 00X planes of Sb2Te3 and superlattice, indicating the highly orientation of the film normal to the substrate. The lattice spacing obtained from XRD analyzes are in good agreement with the simulated results. Transmission electron microscopy (TEM) also revealed the highly oriented superlattice characterized with the lattice fringes parallel to the Si substrate. The superlattice composed of 2 SiTe layers and 4 Sb2Te3 layers are clearly identified by high angle annular dark-field (HAADF) image and are in good agreement with the simulated structure. We will also discuss the relationship between atomic alignment of superlattice and corresponding band structure, and electrical properties of [(SiTe)x/(Sb2Te3)y]z superlattice film.

Bio-inspired computer is one of the best approaches to solve the problem of scaling limitation and to further explore the function of the computer. Neuromorphic circuit is the key component for development of bio-inspired computer. Phase change materials are very suitable for building plastic-connectable neuromorphic circuit (PCNC). In this talk, the recent progress in development of PCNC is presented. The main challenges and the possible solutions for the development of PCNC will be discussed.

Research in the field of neuromorphic and cognitive- computing has generated a lot of interest in recent years. With potential application in field such as large-scale data driven computing, robotics, intelligent autonomous systems to name a few, bio-inspired computing paradigms are being investigated as the next generation (post-Moore, non-Von Neumann) ultra-low power computing solutions.
In this work we discuss the role that different emerging non-volatile resistive memory technologies (RRAM), specifically (i) Phase Change Memory (PCM), (ii) Conductive-Bridge Memory (CBRAM) and Metal-Oxide based Memory (OXRAM) can play in dedicated neuromorphic hardware. We focus on the emulation of synaptic plasticity effects such as long-term potentiation (LTP), long term depression (LTD) and spike-timing dependent plasticity (STDP) with RRAM synapses.
We developed novel low-power architectures, programming methodologies, and simplified STDP-like learning rules, optimized specifically for some RRAM technologies.
We show the implementation of large-scale energy efficient neuromorphic systems with two different approaches (i) deterministic multi-level synapses and (ii) stochastic-binary synapses. Prototype applications such as complex visual- and auditory- pattern extraction are also shown using feed-forward spiking neural networks (SNN).

Complementary metal-oxide-semiconductor technology (CMOS) has allowed following Moore&’s law in the last four decades. However, scaling is near to some challenging issues, like heat dissipation and mobility limitations [1]. To further increase the computing capability of logic circuits, alternative approaches are being explored, such as spin-based logic gates, [2] magnetic domain-walls, [3] molecular devices [4] and memristors. In memristors the resistance R can be tuned in response to an applied electrical pulse. Phase change memory (PCM) can be included in this category, thanks to the large number of states with different R and threshold voltage Vt that can be obtained by changing the crystalline fraction fx of the active layer. PCM based memristor is certainly a good candidate for logic application, thanks to its fast switching [5], low programming current [6] and large R window.
In this work, we demonstrate boolean logic operations with PCM memristors as logic gates [7]. To achieve this result, we exploited two fundamental properties of the chalcogenide material Ge2Sb2Te5 (GST), namely conditional switching and additive crystallization. The first property refers to the amorphous GST, displaying threshold switching at threshold voltage Vt. Above Vt, the amorphous phase is highly conductive thanks to the strong electronic and thermal excitation of carriers [8]. Threshold switching allows to compare an externally applied voltage V to Vt, i.e., for V < Vt no threshold switching takes place and the cell remains in its previous state, while for V > Vt the large current flowing in the amorphous phase induces crystallization. Since the Vt is dictated by the state (i.e. the resistance) of the memristor, it is possible to conditionally switch the cell depending on its initial state. The second property we exploited is additive crystallization, which means that the same memristive state can be obtained with a single or multiple pulses, provided that the total pulse duration is the same.
By means of this two functionalities, we were able to perform a logically complete set of boolean operation, such as negation (NOT), negated sum (NOR) and negated product (NAND). In addition, memristive logic offers totally new possibility for logic circuits. First, the non-volatile nature of PCM enables logic-in-memory application. Second, while CMOS logic relies on the gate topology to yield a certain functionality, PCM logic is simply based on memristive properties. This allows to develop reconfigurable circuits where each PCM can provide any logic functionality depending on the initial configuration.
[1] ITRS 2011 release
[2] A. A. Khajetoorians , et al., Science, 332 (2011)
[3] D. A. Allwood, et al., Science, 309 ( 2005)
[4] W. Liang , et al., Nature, 417 ( 2002)
[5] D. Loke, et al., Science 336, 1566 (2012)
[6] F. Xiong, et al., Science 332, 568 (2011)
[7] M. Cassinerio, et al., Adv. Mat, DOI: 10.1002 (2013)
[8] D. Ielmini, Phys. Rev. B 78, 035308 (2008)

Phase change materials are good candidates for reconfigurable electronic devices, thanks to their rapid modifications upon electric pulses. Such a property may be also employed to obtain resistors with high resistivity, stable over a large temperature range, re-adjustable using current or laser pulses.
To obtain high precision resistors, very thin films (usually a few nm of Si:Cr alloys) are commonly employed, since they are characterized by high sheet resistance and low temperature coefficient of resistance (TCR<100 ppm/K). However, even with very well controlled processes, the as-fabricated resistance value typically falls within a 3-15% range of tolerance. Therefore to get more accurate values a subsequent trimming operation can be required. This is usually done by laser trimming or by electrical methods, but it is not a reversible operation.
Ge₂Sb₂Te₅ (GST), thanks to its polymorphism and to the high resistivity compared to common metals (few mOmega; cm in the hcp), could be an interesting material to manufacture resistors with very high resistivity, in which both the resistance and the TCR can be reversibly adjusted by electric pulses.
GST resistors have been manufactured by depositing 40 nm thick amorphous film on oxidized Si wafers and by pattering through electron beam lithography. After patterning, the stable hcp phase was formed by annealing at 320°C. The resistors have an optimized layout, containing a narrower region in the center (width W=1-3 um), in order to favor current crowding and localized melting.
The electrical trimming has been achieved by two steps: applying high voltage pulses (15-20V) of short duration, to melt and quench into the amorphous phase the narrower region of the resistor; crystallizing the amorphous region in the metastable fcc phase.
For the amorphization a pulse generator and an oscilloscope in series with the resistor have been employed. The resistance has been measured by the parameter analyzer HP4156B, by using a switching matrix. The crystallization has been obtained either by annealing at 170 °C or electrically, by current sweep.
After trimming a resistor containing the two crystalline phases is obtained. The hcp phase has a metallic behavior. The fcc is a semiconductor, with resistivity one order of magnitude higher. By changing the amount of material melt-quenched and converted into the fcc phase, the final resistance and its temperature dependence can be finely tuned. Resistors with positive, negative or zero TCR have been obtained.
By describing the final resistance as a series of resistors with different resistivity and TCR, the trimming condition to simultaneously obtain both a target resistance R and TCR=0 can be evaluated. It is found theoretically that the ratio between the number of squares (N=L/W) in the fcc phase and that in the hcp should be 6.4e-3. According to this description, resistors with low TCR have been experimentally obtained for fcc/hcp ratios very close to the expected value.

Understanding charge transport in phase change materials is crucial to extend the application range of these exciting materials. With this goal in mind, we have studied the resistivity and the thermal transport of crystalline phase change materials. A pronounced dependence of the room temperature resistivity upon annealing temperature is observed for crystalline phase change materials such as Ge1Sb2Te4 or Ge2Sb2Te5. This finding is corroborated by low temperature measurements as well as FTIR data, which confirm that a metal - insulator transition is observed without a change in crystallographic state. This is indicative for an electronically driven MIT. A similar transition is also observed for the thermal conductivity.
Anderson has shown that increasing disorder turns a metal with delocalized electronic states at the Fermi energy into an insulator with localized states. In this talk, arguments for a disorder induced localization of charge carriers will be presented. The observations are compared with doped semiconductors such as Si:P, where both disorder and correlations are crucial to describe the charge transport. Experimental and theoretical attempts to unravel the origin of disorder induced localization will be presented. These calculations reveal that it is the ordering of vacancies into vacancy layers which drives the transition to the metallic state [2]. This vacancy ordering also has a pronounced impact on the thermal conductivity. The potential of this remarkable impact of disorder for applications as well as our fundamental understanding of solids is discussed.
[1] T. Siegrist et al., Nature Materials 10, 202, (2011)
[2] W. Zhang et al., Nature Materials, 11, 952 (2012).

Chemical bonding between atoms plays a key role in determining properties of phase-change materials (PCMs). To understand existing materials and to predict new, better ones by rational design, it is crucial to unravel how structures, chemical bonding, and physical properties are linked in solid PCMs. This explained, for example, the origin of intrinsic vacancies in the data-storage alloy GeSb₂Te₄: the material expels cations which strengthens the remaining bonds, and thus its bonding network is stabilized by forming structural defects [1]. Furthermore, a “treasure map” for PCMs has recently been proposed, again using iconic chemical-bonding concepts (such as orbital hybridization) to map the complex compositional space of those functional materials onto a simple chart [2]. Despite these exciting developments, it is clear that new theories and computational tools will be needed to cope with increasing materials complexity in the framework of realistic (that is, costly) simulations.
The aforementioned interplay of vacancies and chemical bonding [1] had been investigated using an orbital-pair resolved indicator for electronic-structure computations, namely, the COHP technique, within a tight-binding DFT framework. Here, we present a novel theoretical tool [3,4] that conveniently allows to perform similar analyses in the framework of state-of-the-art plane-wave DFT simulations. This makes COHP and related tools available for realistic models of PCMs in simulation cells that contain hundreds of atoms.
We then demonstrate the feasibility of the new method by deriving structure-property correlations for one of the most fundamental PCMs, germanium telluride (GeTe). The atomic motion in the crystalline phase is influenced by chemical substitutions and also vacancy formation [5], which we here rationalize in a unified chemical-bonding framework. Furthermore, we show how the conceptual approach of chemical-bonding analysis can be transferred to two-dimensional surface models (which are crucial to simulate nanoscale PCMs). This way, an explanation is offered for the intriguing stability differences at various GeTe surfaces [6], and directions for future applications are pointed out.
References
[1] M. Wuttig, D. Lüsebrink, D. Wamwangi, W. Welnic, M. Gillessen, R. Dronskowski, Nature Mater. 6, 122 (2007).
[2] D. Lencer, M. Salinga, B. Grabowski, T. Hickel, J. Neugebauer, M. Wuttig, Nature Mater. 7, 972 (2008).
[3] V. L. Deringer, A. L. Tchougréeff, R. Dronskowski, J. Phys. Chem. A 115, 5461 (2011).
[4] S. Maintz, V. L. Deringer, A. L. Tchougréeff, R. Dronskowski, J. Comput. Chem. 34, 2557 (2013).
[5] V. L. Deringer, M. Lumeij, R. P. Stoffel, R. Dronskowski, Chem. Mater. 25, 2220 (2013).
[6] V. L. Deringer, M. Lumeij, R. Dronskowski, J. Phys. Chem. C 116, 15801 (2012).

GeTe and GST amorphous materials are re-investigated using ab initio dft molecular dynamics to compare the effect of various functionals that include the treatment of dispersion (Van der Waals) forces on the structural and dynamical properties of the final amorphous structures. We show that the proportion of tetrahedral Ge to other types of environments as well as the ratio of 3-fold and 2-fold bonded Te atoms is much dependent on the choice of functional. The different functionals yield variable agreement with the available structural experimental data. Properties such as the diffusion coefficient and vibrational densities of states are computed, indicating that models that are extremely close in energy may have very different experimental signatures.

Phase-change materials on the pseudo-binary GeTe-Sb2Te3 tieline, such as GeSb2Te4 or Ge2Sb2Te5, exist in at least two phases, amorphous and crystalline, which significantly differ in their optical and electrical properties [1]. Switching of those materials normally occurs between the amorphous phase and the rock-salt like metastable crystalline phase. The thermodynamically stable, well-ordered crystalline phase with a layered hexagonal structure exhibits weak Te-Te bonds as they also appear in the crystal structure of one starting point of the pseudo-binary line, antimony telluride (Sb2Te3).
We present the results of carefully calibrated density-functional theory (DFT) based calculations on the vibrational properties of layered tellurides [2] using the ab initio force-constant method [3]. Because it is well known [4] that some common DFT parametrizations fail in describing the non-covalent Te-Te interactions within those layered structures, the influence of dispersion-corrected DFT [5] as well as different parametrizations on the predicted vibrational properties is investigated.
Outgoing from the knowledge of the lattice dynamics, other important properties are derived using a unified approach that we have described recently [6]. Among these properties are those most crucial for applications, such as Gibbs free energies, heat capacities, atomic thermal displacements and Grüneisen parameters. In order to prove the reliability of our theoretically calculated results, they are compared to recent experimentally measured ones, based, for example, on nuclear inelastic scattering (NIS) [7]. This allows to put confidence in the data obtained, and we not only show how our data help to interpret the experimental results, but we also give a more detailed insight into the quantum-mechanical background and practical machinery.

References
[1] D. Lencer, M. Salinga, M. Wuttig, Adv. Mater. 23, 2030 (2011).
[2] R. P. Stoffel, V. L. Deringer, R. Simon, R. P. Hermann, R. Dronskowski, in preparation.
[3] A. Togo, F. Oba, I. Tanaka, Phys. Rev. B 78, 134106 (2008).
[4] G. C. Sosso, S. Caravati, M. Bernasconi, J. Phys.: Condens. Matter 21, 095410 (2009).
[5] S. Grimme, J. Comput. Chem. 27, 1787 (2006).
[6] R. P. Stoffel, C. Wessel, M.-W. Lumey, R. Dronskowski, Angew. Chem. Int. Ed. 49, 5242 (2010).
[7] D. Bessas, I. Sergueev, H.-C. Wille, J. Persson, D. Ebling, R. P. Hermann, Phys. Rev. B 86, 224301 (2012).

Phase-change materials (PCMs) are commonly employed in confined spaces of memory devices, but undergo a change in density between 5% and 9%[1] during transformation. As a result, switching operations will induce pressures different from ambient conditions. When large pressures of several GPa are applied to the meta-stable crystalline phase of GeTe-Sb2Te3 (GST)-based materials, they irreversibly transform into a pressure-induced amorphous phase, which is formed without thermally quenching the material[2][3]. Both, the different atomic radii as well as the large concentration of vacancies have been proposed as origin for this transformation.
We have performed in-situ Raman spectroscopy experiments while applying various pressures to the PCMs using a diamond anvil cell. Since the amorphous and crystalline phases possess distinct vibrational spectra, Raman spectroscopy can resolve the transition process. We have studied several GST-based compounds and SnSb2Te4 - a material with atoms of similar size. We find that this material also transforms irreversibly into an amorphous phase and conclude that the atomic radii are not crucial to the amorphization process[4]. A further comparison of the Raman spectra of the as-deposited and pressure-induced amorphous phases suggests that the local atomic configurations are very similar - despite the fact that some of their properties differ.
[1] T. P. Leervad Pedersen, J. Kalb, W. K. Njoroge, D. Wamwangi, M. Wuttig, and F. Spaepen, “Mechanical stresses upon crystallization in phase change materials,” Appl. Phys., vol. 79, no. 22, p. 3597, 2001.
[2] A. V. Kolobov, J. Haines, A. Pradel, M. Ribes, P. Fons, J. Tominaga, Y. Katayama, T. Hammouda, and T. Uruga, “Pressure-Induced Site-Selective Disordering of Ge2Sb2Te5: A New Insight into Phase-Change Optical Recording,” Phys. Rev. Lett., vol. 97, no. 06, pp. 035701-1, 2006.
[3] S. Caravati, M. Bernasconi, T. Kühne, M. Krack, and M. Parrinello, “Unravelling the Mechanism of Pressure Induced Amorphization of Phase Change Materials,” Phys. Rev. Lett., vol. 102, no. 20, pp. 1-4, 2009.
[4] W.-P. Hsieh, P. Zalden, M. Wuttig, A. Lindenberg, and W. L. Mao, “High-pressure Raman spectroscopy of phase change materials,” accepted for publication in Appl. Phys. Lett., 2013.

The prototypical material choice for phase-change non-volatile memory is Ge₂Sb₂Te₅ (GST-225), which is a nucleation dominated phase-change material. It is well known that device endurance is hindered by failure mechanisms associated with phase segregation or compositional redistribution. In this study, we have applied micro-Raman spectroscopy to study the structural evolution and chemical bonding changes of PVD GST-225 as a function of cycle number consisting of alternating RESET and SET laser pulses. Our study is performed on SiO₂-capped 20nm thick GST-225 blanket films using a static laser tester. The effect of thermal or phase-change induced stress is therefore being examined independently of electric field induced changes in structure. The distinct changes in the vibrational modes associated with GeTe and Sb_xTe_y as a function of cycle number have been correlated to microstructural changes and provide insight into phase-change induced segregation. The laser tester was modified to apply alternating RESET / SET pulse cycles that span 6 orders of magnitude. The GeTe modes are exhibited in the as-deposited film and as we cycle the sample, they begin developing into stronger and sharper features at a quicker rate than Sb_xTe_y related modes. There is an overall decrease in linewidth for all Raman modes with increased cycling from 1 to 1.3E6 cycles. At 10K cycles, there is an onset of a GeTe mode at 140cm^-1, which continues to increase in intensity with cycle number. Furthermore, at 50K cycles, there is the appearance of the Sb₂Te₃ A²_1g mode at 168cm^-1, which also continues to increase in intensity toward higher cycle numbers. The main GeTe mode at 122cm^-1, is the only feature present in as-deposited amorphous, laser-densified, laser melt-quenched and as a set state following a single reset/set cycle. Additionally, microstructural and volumetric changes have been studied for each cycled condition with cross-sectional TEM. Our study shows that micro-Raman spectroscopy can be a powerful analysis tool, particularly when used simultaneously with laser reflectivity measurements, offering facile structural insights in a non-destructive manner.

Semiconductor spintronics - the branch of electronics aiming at exploiting the electron spin, in addition to its charge - has for long concentrated on devices, such as the Datta-Das spin-FET, where the semiconductor plays a passive role. In the compelling quest for multifunctionality and non-volatility, however, a breakthrough would come from turning the semiconductor into an active element. We recently proposed to radically change the perspective of current semiconductor spintronics by bringing in a novel functionality, through the coexistence and coupling between ferroelectricity (i.e. presence of a permanent and switchable polarization) and Rashba effects (i.e. presence of a k-dependent spin splitting in the band structure) in Ferroelectric Rashba SemiConductors (FERSC). The test-case material is represented by GeTe, where theory appealingly predicts a giant Rashba spin-splitting to be reversed upon switching ferroelectric polarization.[1] To exploit this peculiar property, we proposed a spin-FET device whose impedance can be electrically controlled, thanks to the interplay between Rashba effects, electronic spin precession and ferroelectricity. Noteworthy, non-volatility is ensured by the permanent polarization of the GeTe channel, so that a new generation spintronics devices with non-volatile logic functions associated with the remanent ferroelectric states can be envisaged.
In this work we present the surface and bulk band structure of GeTe(111) thin films [2] as determined by Angular Resolved PhotoEmission Spectroscopy (ARPES). A huge Rashba splitting of the valence band has been measured at low temperature, in nice agreement with calculations and thus confirming the potential of this material for applications. FM/insulator/GeTe tunnelling contacts have been then fabricated on top of GeTe films, and the spin diffusion length investigated via Hanle effect measurements.
This work paves the way to the demonstration of the basic functionalities of spin-FET devices with an active channel made of a FERSC material.
[1] D. Di Sante et al., Adv. Mat. 25, 509-513 (2013)
[2] A. Giussani et al., Phys. Status Solidi B 249, 1939-1944 (2012)

Germanium telluride (GeTe) possesses a large panel of very different properties that are worth studying, both from a fundamental and technological point of view. For its ability to be switched reversibly between an amorphous and a crystalline phase, changing drastically its optical and electrical properties, GeTe-Sb₂Te₃ compounds are already used in phase change data storage applications [1]. In recent works, Simpson et al. reported very promising results for GeTe/Sb₂Te₃ interfacial phase-change memories [2]. GeTe is also studied as a narrow-gap ferroelectric semiconductor using piezoresponse force microscopy [3]. Recently, Sante et al. also predicted theoretically a giant Rashba effect in bulk GeTe linked to its ferroelectric properties [4].
In all these cases, there is an obvious need for very high quality crystalline GeTe films for in-depth fundamental studies. The epitaxial growth of GeTe on a Si(111)-(7×7) surface using molecular beam epitaxy has been already demonstrated [5]. However, the presence of in-plane rotational domains spaced by sim;14° was observed, and the surface roughness was unsatisfactory. In the present work, the crystalline quality of the GeTe film was significantly improved by growing on a Si(111)-(radic;3×radic;3)R30°-Sb surface.
X-ray diffraction phi;-scan around the α-GeTe{0-114} Bragg peaks showed single peaks spaced by 60°, pointing to the presence of only one rotational domain plus its twinned domain. The intensity of the peaks associated with the twinned domains were up to 20 times lower than that of the main ones, indicating that only a very small volume of the film is twinned.
The amorphous transition at growth onset reported by Giussani et al. [5] was not observed by in-situ real-time reflection high-energy electron diffraction; the GeTe growth starts immediately crystalline, which enables the growth and study of thinner epitaxial GeTe films.
AFM surface profiles showed large flat triangular terraces with sub-nanometer RMS roughness. There are occasional gaps in between the terraces but their number and size could be minimized by optimizing the growth parameters.
[1] N. Yamada, E. Ohno, N. Akahira, and K. Nishiuchi, Japanese Journal of Applied Physics, vol. 26, no. 4, pp. 61-66, 1987.
[2] R. Simpson, P. Fons, A. Kolobov, T. Fukaya, M. Krbal, T. Yagi, and J. Tominaga, Nature nanotechnology, vol. 6, pp. 501-5, Aug. 2011.
[3] J. J. Gervacio-Arciniega, E. Prokhorov, F. J. Espinoza-Beltran, and G. Trapaga, Journal of Applied Physics, vol. 112, no. 5, p. 052018, 2012.
[4] D. Di Sante, P. Barone, R. Bertacco, and S. Picozzi, Advanced materials, vol. 25, pp. 509-13, Jan. 2013.
[5] A. Giussani, K. Perumal, M. Hanke, P. Rodenbach, H. Riechert, and R. Calarco, Physica Status Solidi (B), vol. 249, pp. 1939-1944, Oct. 2012.

We present the first angle-resolved photoemission study of a ternary phase change material, namely Ge₂Sb₂Te₅, epitaxially grown on Si(111) in the metastable cubic phase. The observed upper bulk valence band shows a minimum at Gamma-bar being 0.3 eV below the Fermi level E_F and a circular Fermi contour around Gamma-bar with a dispersing diameter of 0.24-0.36 #8491;^-1. Previous density functional theory calculations find the valence band maximum away from Gamma-bar only if the Z₂ topological invariant nu;₀ is non-trivial while for a trivial Z₂ invariant the valence band maximum is always found at Γ-bar. This implies that the measured band structure of Ge₂Sb₂Te₅ is topologically non-trivial. Scanning tunneling spectroscopy exhibits a band gap of 0.4 eV around E_F.

Epitaxy is a powerful method for controlling and improving the functional properties of materials. Here we study the growth of Sb2Te3, a building block of interfacial phase change memory [1], on Si(111). We especially focus on the influence of Si(111) surface reconstructions on the structural properties. We show that Sb2Te3 grown on Si(111)-7x7 has a single out-of-plane orientation, but consists of domains with different in-plane orientations and strain states. We link the in-plane rotations to the presence of dangling bonds on the Si(111)-7x7 surface. Furthermore, we demonstrate that an alignment of the Sb2Te3 lattice with the Si lattice can be achieved when Sb2Te3 is grown on Sb/Si(111)-radic;3xradic;3 or H/Si(111). This is attributed to the passivation of dangling bonds on the Si(111) surface by the Sb or H ad-atoms. These improvements in the epitaxy of Sb2Te3 on Si(111) are important for the growth of Sb2Te3/GeTe superlattices and thus for the optimization of interfacial phase change memory.
[1] R. E. Simpson, P. Fons, A. V. Kolobov, T. Fukaya, M. Krbal, T. Yagi, and J. Tominaga, “Interfacial phase-change memory,” Nat. Nanotechnol., vol. 6, pp. 501-5, Aug. 2011.

Processes involved in rewritable, non-volatile memory must be rapid and reversible, and the demands on speed and stability continue to increase. Commercial phase change memory (PCM) materials (DVD-RW, Blu-ray Disc, .. ) utilize the rapid change between the crystalline and amorphous phases induced by laser light, and the corresponding change in electronic properties, for this purpose. Understanding the mechanism behind the transition requires a detailed knowledge of the structure of both phases, and the combination of density functional (DF) calculations with molecular dynamics (MD) has played a vital role in the context of commercial PCM such as alloys of Ge/Sb/Te [1].
An alternative mechanism for non-volatile memory also involves the rapid switching of structure and properties, this time induced by an electric field, and we are applying the combined DF/MD approach to this case. Conductive bridge memory cells involve two electrodes (one usually copper or silver) and a solid electrolyte, often a chalcogenide glass. The structural changes here involve the electrochemical growth and dissolution of metallic filaments between the electrodes, with dramatic changes in the resistivity, and such devices promise lower power and increased endurance than possible with current non-volatile memory. As a first step towards simulating the mechanism, we have performed DF calculations on two alloys of As/S/Ag, and refined the structure using x-ray diffraction, neutron diffraction, and EXAFS data. The final structure reproduces all these data very well, and the electronic structure is consistent with other available information. In addition to the amorphous structures at 300 K, the dynamic properties and diffusive motion of Ag has been studied at temperatures up to 600 K.
[1] J. Akola and R. O. Jones, Phys. Status Solidi (b) 249, 1851 (2012)