Advertisement

Phase-Change Memory Materials

  • Alexander V. Kolobov
  • Junji Tominaga
  • Paul Fons
Part of the Springer Handbooks book series (SPRINGERHAND)

Abstract

Phase-change materials are Te-containing alloys, typically lying along the GeTe-Sb2Te3 quasibinary tie line. Their ability to switch, reversibly and extremely quickly, between the crystalline and amorphous phases, combined with the high stability of both phases, makes them ideally suitable for memory applications. They have been long used in optical data storage in the form of DVD and Blu-Ray disks and recently have also emerged as a leading candidate for electronic nonvolatile memory devices. In this chapter, a detailed description of these materials is provided starting with the global and local structures of the two phases, which were extensively studied both experimentally and using ab initio computer simulations, and followed by the discussion of possible atomistic mechanisms of the phase-change process, with special accent on the role of electronic excitation. The chapter is concluded by a brief description of the present and emerging applications of this class of chalcogenide materials.

Nonvolatile memory devices are key elements of a wide variety of electronic and portable systems such as digital cameras, solid-state disks, smartphones, computers, e-books, tablets, etc., and their market has been increasing exponentially over the last decade. Even though flash memory represents today the leading technology, to allow its scalability down to the 16 nm technology node and beyond, new architectures are necessary. Therefore, new emerging nonvolatile memory concepts are under investigation and one of the leading candidates is phase-change memory (PCM ). PCM has been successfully used in optical memory devices such as DVD-RAM since the 1990s and recently commercial production of electronic nonvolatile phase-change random access memory (PRAM ) has been launched by two of the world’s leading memory makers Samsung and Micron. PCM is considered to be a storage class memory in which unlike flash its elements are individually addressable, have a demonstrated cyclability of over 1011 cycles, and have access times on the tens to hundreds of nanoseconds scale.

The basic idea of PCM of utilizing the property contrast between the amorphous and crystalline phases (the SET and RESET states) of some materials belongs to Ovshinsky and dates back to the 1960s [46.1] and is based on the following. When a liquid is cooled, it can solidify either discontinuously to form a crystal or in a continuous way to form a glass. The process is schematically illustrated in Fig. 46.1.
Fig. 46.1

A volume–temperature phase diagram demonstrating the formation of either a crystalline or amorphous phase from the melt

A very simplified picture of the phase-change process is based on the idea that as the temperature of the melt decreases, the viscosity becomes larger, and at a certain point the structure can no longer follow the change in temperature. As a result, if the melt is cooled slowly, the equilibrium crystalline phase is formed. If, on the other hand, the melt is cooled fast enough so that the structure cannot follow the change in temperature, the disordered phase is frozen; one obtains a supercooled liquid and subsequently a glass. When the obtained glass is annealed at temperatures between the glass-transition temperature and the melting point, it gradually transforms into the crystalline state. One can thus cycle a material between the ordered crystalline and disordered amorphous phases. In the device-relevant phase-change process, a short intense laser or current pulse melts the material that is subsequently quenched into the amorphous phase. A longer pulse of lower intensity reverts the material to the crystalline phase.

The research and development of phase-change optical storage has a long history starting with the discovery of the switching effect in some chalcogenide alloys [46.1, 46.2, 46.3]. In particular, in amorphous Si12Te48As30Ge10 it was found that when a voltage that exceeded a critical value, called the threshold voltage, was applied, the chalcogenide alloy switched from the low-conductivity state to the high-conductivity state [46.1]. As the current in the high-conductivity state was reduced below the characteristic value termed the holding current, the system switched back to the original low-conductivity state. The observed switching process was reversible and could be repeated many times. The observed increase in the conductivity was attributed to the formation of a current filament, growing in diameter with increasing current flow. In the same paper [46.1], it was reported that by tuning the composition a memory effect was observed that was attributed to a change in the structure of the material, namely:

[…] after switching from a highly resistive state, structural changes result in the preservation of a conductive state even when the current is totally removed. The material can be reversibly switched back to the highly resistive state by application of a current pulse of either polarity exceeding a threshold value.

This finding marked the beginning of intensive – and extensive – research on phase-change materials. In order for a material to become commercially relevant, it has to satisfy simultaneously several requirements such as sufficiently large property contrast between the two states and high stability of both phases at operating temperatures alongside high switching speed in both directions. Less obvious but equally important are the need for low thermal conductivity of the crystalline phase (to ensure low-energy switching), the relative softness of the material needed to withstand stresses generated at the amorphous-crystalline boundaries allowing for high cyclability, and good scalability. While there are many materials that satisfy some of these requirements, very few satisfy them all.

Years of research have singled out GeTe-based PCM alloys . In practical applications, GeTe is usually alloyed with Sb or other additives such as C or N to tune desirable properties, such as thermal stability, switching speed, or optical contrast (e. g., Ge2Sb2Te5 is used in DVDs but Ge8Sb2Te11 is used in Blu-ray disks). What makes the GeTe alloys special? Since the properties of a material are determined by its structure, we start by discussing the structure of GeTe-based alloys.

46.1 Structure of Ge-Sb-Te Phase-Change Alloys

46.1.1 Crystalline Phase

Binary GeTe

It is natural to start the description with the binary compound GeTe, which is the simplest material in the Ge-Sb-Te (GST) system. In the low-temperature ferroelectric phase, GeTe possesses a rhombohedral structure with the space group R3m. This structure can be viewed as a rock-salt structure slightly distorted along the ⟨ 111 ⟩  direction with a subsequent shear relaxation. The driving force for the formation of the rhombohedral phase has been a subject of several studies in the past [46.4]. In this phase, Ge and Te atoms are sixfold coordinated to each other with subsets of three shorter (2.83 Å) and three longer (3.15 Å) bonds often described as a Peierls distortion [46.5] due to the reduced coupling between the orthogonal p-type orbitals that constitute the basis for bonding in GeTe.

Based on diffraction studies, mainly neutron diffraction [46.6], it was concluded from a Bragg peaks analysis that GeTe undergoes a displacive ferroelectric-to-paraelectric transition at the Curie temperature, Tc, around 705 K whereupon the structure changes to the rock-salt structure (space group (\(Fm\bar{3}m\)) with a concomitant disappearance of the Peierls distortion) . Subsequent use of local probes such as extended x-ray absorption fine structure (EXAFS ) or pair-distribution function analysis of total scattering unambiguously demonstrated that the structure locally remains distorted above Tc in essentially the same manner as it is distorted at lower temperatures but the distribution of shorter and longer bonds becomes random, which manifests itself as a rock-salt average structure from the perspective on long-range averaging probes such as Bragg diffraction [46.7, 46.8].

Ge2Sb2Te5 and Related Materials

The first report on the structure of Ge-Sb-Te dates back to the late 1960s when Petrov et al. [46.9] investigated the structures of Ge2Sb2Te5 and GeSb4Te7. Analysis of electron-diffraction results led the authors to conclude that the studied crystals possessed a hexagonal cell with space group \(P\overline{3}m1\) and the lattice constants \(a={\mathrm{4.20}}\pm{\mathrm{0.02}}\) and \(c={\mathrm{16.96}}\pm{\mathrm{0.06}}\,{\mathrm{\AA}}\) and \(a={\mathrm{4.21}}\pm{\mathrm{0.02}}\) and \(c={\mathrm{23.65}}\pm{\mathrm{0.08}}\,{\mathrm{\AA}}\) for Ge2Sb2Te5 and GeSb4Te7, respectively. The structure of Ge2Sb2Te5 was described as a layered structure with the stacking sequence along the c-axis of Te-Sb-Te-Ge-Te-Te-Ge-Te-Sb-Te. This sequence is sometimes referred to as the Petrov sequence. The stable phase of Ge2Sb2Te5 was subsequently re-examined by Kooi et al. [46.10]. Very similar lattice parameters were obtained (a = 4.2 Å and c = 17.2 Å) but the stacking sequence was found to be different, namely, Te-Ge-Te-Sb-Te-Te-Sb-Te-Ge-Te, that is, with the Ge and Sb positions exchanged. A later study using synchrotron radiation found that the Ge/Sb layers were randomly occupied by these two elements [46.11].

Yamada [46.12] was the first to study the structure of a thin layer, crystallized from the amorphous phase. It was found that it possesses a structure that was different from that of the stable trigonal phase. This issue was further explored using x-ray diffraction (XRD ) by Nonaka et al. [46.13] and by Yamada and Matsunaga [46.14].

Based on these results, the crystal structure of Ge2Sb2Te5 was identified as the rock-salt (NaCl) structure (\(Fm\overline{3}m\)). From Rietveld refinement, it was concluded that the anion face-centered cubic (fcc ) sublattice was fully occupied by Te atoms with Ge and Sb atoms randomly located on the cation sublattice. The stoichiometry of the structure requires that there are vacancies on the Ge/Sb sites and their presence has been confirmed by structure refinement. The proposed structure is shown in Fig. 46.2 . It was subsequently shown that the rock-salt structure is characteristic of a wide range of GST alloys [46.15, 46.16, 46.17, 46.18].
Fig. 46.2

Metastable cubic structure of Ge2Sb2Te5

EXAFS studies performed by various groups [46.19, 46.20] demonstrated that the Ge-Te bond length in metastable Ge2Sb2Te5 is 2.83 ± 0.01 Å, that is, significantly shorter than might be expected based on the rock-salt structure as determined by XRD and the experimental lattice parameter of slightly over 6.00 Å. Based on these results, it was argued [46.20] that the structure of metastable Ge2Sb2Te5 does not possess rock-salt symmetry but is locally distorted. The local structure around the Ge species is very similar to that of the binary GeTe, which possesses a rhombohedral structure with subsets of three shorter and three longer bonds. The distortion of the Ge atom location was subsequently confirmed experimentally using scattering measurements [46.21, 46.22]. The Sb-Te bond length was also found to be shorter than half the lattice parameter (2.94 Å [46.20]). The distortion present in GST alloys is usually referred to as Peierls distortion [46.5].

While the obtained bond lengths are significantly shorter than those expected from the experimental lattice parameters and the rock-salt symmetry, they are at the same time significantly longer that the sum of the corresponding covalent radii, which suggests that the bonds are not purely covalent.

The rock-salt like arrangement of atoms in the crystalline phase requires the formation of six bonds by each participating atom while at the same time, the number of valence electrons located on Ge and Sb atoms is lower. A way to ensure six bonds is by virtue of sharing the valence electrons among several bonds with less than two electrons per bond on average through resonance bonding, a concept that is described as follows. Lucovsky and White [46.23] were the first to discuss this possibility and its consequences for IV-VI crystals such as GeTe.

In particular, in the cited work [46.23] resonance bonding in IV-VI and chalcogen (Se and Te) crystals was considered and it was suggested that we might expect materials that exhibit resonance bonding in the crystalline phase to have different properties in the amorphous phase. It was also stated that long-range order is crucial for resonance bonding to exist. If long-range order is lost, the system will be unable to achieve resonant bonding [46.23]. This approach has been subsequently elaborated in [46.24, 46.25, 46.26] to describe the unusually large property contrast in GST alloys and the fast switching rate.

How does the bond length asymmetry effect the electron density distribution? A plot of the charge density difference (CDD ) between a simulated relaxed GeTe model at 0 K and isolated pseudo-atoms shown in Fig. 46.3 demonstrates that the electron density is only significant along the short bonds. In contrast, the electron density pile-up along the long bonds is significantly lower demonstrating that there is a pronounced bonding energy hierarchy between short and long bonds [46.27]. The existing bonding energy hierarchy should manifest itself in the different response of the short and long bonds to thermal and/or electronic excitations. The fact that long-range order is associated with weak secondary bonds suggests that the crystalline phase is intrinsically fragile.
Fig. 46.3

Difference in electron charge density for GeTe at 0 K and isolated pseudo-atoms. A significant bonding charge can be seen midway along the shorter bonds which corresponds to covalent bonding. At the same time, the bonding localized along the longer bonds is significantly lower clear demonstrating a pronounced bonding energy difference between the short and long bonds. (Reprinted from [46.27] by permission from Macmillan Publishers Ltd., copyright (2011))

Vacancies present in the structure deserve a dedicated paragraph. Kolobov et al. [46.20] suggested that vacancies serve to stabilize the structure by ensuring the appropriate charge balance. The role of vacancies was a subject of several consequent studies, where it was also concluded that vacancies, whose concentration varies as \(x/(1+2x)\) for the GeTe1−x-( Sb2Te3) x system [46.15], are not electronic defects but are an intrinsic feature of the GST rock-salt structure [46.15, 46.24, 46.28]. Vacancies were also argued to account for the p-type conductivity of the crystalline phase of GeTe [46.29]. Wuttig et al. [46.28] demonstrated that introduction of vacancies into Ge/Sb sublattice decreased the total energy of the system.

46.1.2 Amorphous Phase

Experimental Studies

Of special interest may be the structure of amorphous binary GeTe (a-GeTe) that is the simplest phase-change material and an end point of the quasibinary GeTe-Sb2Te3 tie-line. Despite years of studies by various groups, there is still controversy regarding the structure of amorphous GeTe. X-ray diffraction structural studies made by Betts et al. [46.30] back in the 1970s revealed that the interatomic distance and the coordination number of a-GeTe are in poor agreement with those of c-GeTe. From a detailed analysis of the diffraction data, these authors concluded that a random covalent model with a 4(Ge):2(Te) local coordination was the most appropriate as a local coordination model for a-GeTe. Subsequent electron diffraction studies [46.31, 46.32] obtained similar radial distribution functions. The 4(Ge):2(Te)-coordinated structure was also suggested on the basis of EXAFS measurements around both Ge and Te K-edges [46.33, 46.34, 46.35]. The Ge-Te bond length (2.59 Å) was also found to be consistent with the 4(Ge):2(Te) bonding geometry [46.35]. Raman scattering and far-infrared absorption spectra for a-GeTe [46.36] were equally interpreted in terms of the presence of GeTe4 tetrahedra. These results strongly supported a random covalent network model of the 4(Ge):2(Te)-coordinated atoms, and excluded the possibility of a c-GeTe microcrystalline structure. Combined photoemission and inverse photoemission studies [46.37] also favored the 4(Ge):2(Te) coordination.

On the other hand, neutron scattering [46.38] and Mossbauer spectrometry of 125Te nuclei [46.39] studies suggested a 3(Ge):3(Te)-coordinated local structure, demonstrating that the local structure of a GeTe is still far from being resolved.

The first report on the local structure of the amorphous phase of GST alloys belongs to the LETI group in France [46.19]. These authors performed EXAFS studies on as-deposited and thermally crystallized layers of GeTe and GST and obtained the bond lengths characteristic of the two states. In particular, they found the Ge-Te bond to be 2.62 Å and the Sb-Te bond to be 2.84 Å (as compared to 2.83 and 2.94 Å, respectively in the crystalline phase), that is, the bonds are shorter in the amorphous phase. This work, however, only reported the obtained numerical values and stayed short of drawing further conclusions.

Kolobov et al. [46.20] were the first to go beyond simple measurements and proposed structural models. Their results are described as follows in more detail. It should be noted that in this work laser-crystallized and laser-amorphized states of GST were studied, that is, exactly the structures that are used in optical memory devices (the measurements were performed on real-device structures).

For the amorphous phase, it was found that Ge-Te and Sb-Te bonds become shorter and the structure possessed more local order than the crystalline phase as evidenced by the more intense and narrower peaks in the spectra corresponding to the amorphous phase (Fig. 46.4). It is informative to note that the measured Ge-Te and Sb-Te bond lengths are very close to the sum of the corresponding covalent radii for the elements (rGe = 1.21 Å, rSb = 1.40 Å, rTe = 1.36 Å [46.40]). (We note here that the covalent radius for Ge is given assuming sp3-hybridization).
Fig. 46.4a–c

Fourier transformed EXAFS spectra for Ge (a), Sb (b), and Te (c) K-edges for laser crystallized and laser amorphized Ge2Sb2Te5. (Reprinted from [46.20] by permission from Macmillan Publishers Ltd, copyright (2004))

The observed bond shortening and increased local order are highly unusual for covalent solids when, due to anharmonicity of the interatomic potential, disorder usually results in longer and weaker bonds and suggests that the local structures in the two cases are significantly different. It should also be noted that despite the bond shortening, the density of the amorphous phase is about 5% lower than that of the crystalline phase.

The conclusion that a significant change occurs in the local structure finds more support in the experimentally measured x-ray absorption near-edge structure (XANES ) spectra [46.20], where the overall observed changes in XANES upon the phase transition could be reproduced by first-principles multiple-scattering (FEFF) simulations [46.41] when the Ge atoms were placed into tetrahedral symmetry sites [46.20].

Amorphous Ge2Sb2Te5 has subsequently been studied using EXAFS by other groups [46.42, 46.43, 46.44] and similar results were obtained although in addition the presence of Ge-Ge bonds was also reported. It is interesting to note that the obtained values of the fraction of Ge-Ge bonds were similar to the concentration of tetrahedral Ge sites, suggesting a correlation between the two.

Bond shortening in the as-deposited amorphous phase of Ge2Sb2Te5 has also been observed by high-energy x-ray scattering experiments [46.45].

A gradual upwards drift in the electrical resistance of Ge-Sb-Te alloys has been observed after switching to the amorphous state that potentially can cause problems for multilevel storage applications in PRAM. This phenomena, often referred to as drift of the amorphous state, is usually attributed to the gradual [46.46, 46.47] relaxation of bonds toward a lower energy state. In particular, the loss of Ge-Ge bonds in the amorphous state leads to a widening of the bandgap and a gradual increase in resistivity.

Computer Simulations

Welnic et al. [46.48] were the first to apply ab-initio simulations based on density functional theory for phase-change materials, using the spinel structure (in which Ge atoms are tetrahedrally coordinated) to model the amorphous phase. Calculations of the band structure showed a band opening at the Γ point upon transformation from the rock-salt to the spinel phase in agreement with experiment.

A significant step forward was the in-silico generation of the melt-quenched amorphous phase. The first simulation of the melt-quenched amorphous Ge2Sb2Te5 was reported by Caravati et al. [46.49]. The liquid structure (270 atom) generated at 2300 K was equilibrated for 6 ps and then quenched in 16 ps and further equilibrated for 18 ps at 990 K. In order to generate a model of a-Ge2Sb2Te5, the liquid was then brought to 300 K in 18 ps. Subsequently these studies were extended [46.50]. The obtained results can be summarized as follows. Ge and Sb atoms are mostly fourfold coordinated and form bonds preferentially with Te atoms. A large fraction of Ge atoms (≈ 33%) were found on tetrahedral sites. It was additionally found that the presence of bonds with Ge or Sb favors a tetrahedral geometry.

Akola and Jones [46.51, 46.52] performed similar simulations using significantly larger cells (460 atom as opposed to the 270 atom of [46.49]) and carried out molecular dynamics simulations over hundreds of picoseconds, times that are comparable to the experimentally observed (ca. 1 ns) amorphization times.

The Ge-Sb (2.78 Å) and Sb-Te (2.93 Å) bond lengths of a GST were found to be shorter than in the crystalline form, but longer than those obtained from experimental studies. The most prominent topologies reported were Ge-Te4, Sb-Te3, and Te-Ge ( Sb)3. The total coordination numbers using a distance cut-off of 3.2 Å were 4.2 (Ge), 3.7 (Sb), and 2.9 (Te). About 60% of Ge atoms were found to be fourfold coordinated but only 34% were tetrahedrally bonded.

The study found square fragments dominated the structure of a GST. Denoting Te atoms as A and Ge/Sb atoms as B, the authors introduced ABAB squares as the building blocks of the a GST structure (Fig. 46.5). The rapid amorphous-to-crystalline transition was viewed by the authors as a vacancy-supported reorientation of ABAB squares. The same authors subsequently studied the Ge8Sb2Te11 composition used in Blu-ray disks [46.53].
Fig. 46.5

A fragment of the amorphous structure showing an ABAB cube [46.51]. (Copyright 2007 by the American Physical Society. Reprinted with permission)

Hegedüs and Elliott [46.54] performed a comprehensive study of phase-change atomistics in Ge2Sb2Te5 throughout the phase-change cycle. Starting with the liquid phase, they found the existence of four-membered rings (4-rings), analogous to the ABAB blocks of [46.51], even at temperatures as high as 1073 K. They found that the concentration of 4-rings increased as the temperature decreased and this increase was found to be correlated with the size of the maximal cluster of connected square rings.

The results obtained using the DFT simulations described earlier clearly demonstrated that not all Ge atoms acquire tetrahedral symmetry positions in the amorphous phase, although it was argued that the concentration of the tetrahedral Ge sites depends on the simulation details and can reach a value of 50% [46.55].

Krbal et al. [46.56] performed a detailed Ge K-edge XANES analysis on the MQ amorphous model of Hegedüs and Elliott [46.54] that contained 90 atom in the simulation cell, including 20 Ge atoms. The analysis of this amorphous model of GST [46.54] revealed that the local structure around Ge atoms can be grouped into three characteristic bonding configurations with respect to the first-nearest neighbor geometry:
  1. 1.

    Purely tetrahedral (T d ) configurations

     
  2. 2.

    Pyramids (P y ), with a Ge atom at the apex and with Te-Ge-Te angles very close to 90

     
  3. 3.

    Highly distorted octahedral (O h ) structures.

     
What makes the group-IV element Ge form P y configurations in the MQ amorphous phase? The answer to this question derives from the definition of a covalent bond, a concept that is characterized by the sharing of a pair of electrons between two bonded atoms [46.57]. Germanium has a s2p x 1 p y 1 p z 0 electronic configuration in its outer shell. In most materials, these orbitals hybridize and the electronic configuration changes to sp3, allowing for the formation of four equivalent covalent bonds.

A rather special situation is created when Ge bonds to a chalcogen. The electronic configuration of Te is ( s2 ) p x 1 p y 1 p z 2 , that is, there are two unpaired p electrons and a nonbonding lone-pair in the outer shell. The Te lone-pair p electrons have the option of interacting with the empty p-orbital of an unhybridized Ge atom, forming a two-electron dative bond, which, once formed, is indistinguishable from usual covalent bonds. A similar approach was used to explain the existence of Ge(3):Te(3) configurations in [46.58], where the authors also considered the formation of Ge-Te bonds that involved the lone-pair electrons of Te.

How does the incorporation of Sb atoms affect the structure and properties of this type of material? An Sb atom has one unpaired valence electron on each of its three p-orbitals and can thus form three conventional covalent bonds. On the other hand, Te atoms can form two conventional covalent bonds with Sb atoms; its lone-pair electrons cannot be used. As a result, two Sb atoms (shown in magenta in Fig. 46.6) are needed to replace three Ge atoms to ensure that all interatomic bonds are saturated, which determines the stability of the GeTe-Sb2Te3 tie-line. This substitution generates a vacancy on a Ge site. The three Te atoms (a Te triad) located next to the vacancy are twofold coordinated and possess unused lone-pair electrons. Because the Te atoms are low (twofold) coordinated, they can rather easily move in space in the directions show by the arrows in the figure. With 20% of Ge sites being vacant in a typical GST alloy Ge2Sb2Te5, the concentration of lone pairs is on the order of 1021 cm−3.
Fig. 46.6

Schematic of the formation of a vacancy and twofold coordinated Te atoms as a result of Sb substitution in GeTe [46.59]. Ge atoms – blue, Sb atoms – magenta, Te atoms – orange. (Copyright 2013 by the American Physical Society. Reprinted with permission)

The presence of lone-pair orbitals nearly aligned with covalent Ge-Te bonds leads to the formation of three-center four-electron (3c–4e) Te-Ge-Te bonds [46.59] as illustrated in Fig. 46.7 . The creation of such bonds with equal Te-Ge and Ge-Te distances on both sides of the central Ge atom has several important consequences. First, it leads to the strengthening of interlayer interactions reducing the bonding energy hierarchy characteristic of the ideal binary rhombohedral GeTe phase and makes the structure locally more cubic. Second, the similarity in the bond-lengths and energies lead to a more conducive environment for resonance bonding in the structure with 3c–4e bonds, making quasibinary GeTe-Sb2Te3 different from the ideal GeTe case that is characterized by a strong bond energy hierarchy.
Fig. 46.7a,b

Schematic of the formation of three-center four-electron Te-Ge-Te bonds (a) utilizing Te lone-pair electrons of the twofold coordinated Te atoms located around the Ge vacancy. Te atoms are shown in orange, Ge atoms in blue, Sb atoms in magenta, and the Ge-site vacancy is shown as an empty circle. (b) Shows the result of DFT simulation of the three-center Te-Ge-Te bond complexes. Two CDD slice – separated by a black dashed line – one for each of the two 3c–4e bonds, are shown on the right. The red spots of similar size and color midway between the Ge and Te atoms in the CDD map indicate covalent (like) interaction of similar strength on both sides of the central Ge atom along the Te-Ge-Te bond direction [46.59]. (Copyright 2013 by the American Physical Society. Reprinted with permission)

46.2 Mechanism of the Phase-Change Process

46.2.1 Structural Studies

In early work, it was tacitly assumed that upon exposure of the crystalline phase to an intense laser or current pulse, phase-change materials melt and are subsequently quenched into a (completely) disordered amorphous state. Recent years have witnessed significant progress in the understanding of the atomistic mechanism of phase change.

The first model that attempted to provide an atomistic description of the phase-change process was the umbrella-flip model [46.20]. Subsequent ab initio studies performed by different groups have not only confirmed the existence of Ge atoms on tetrahedral symmetry sites (about 30% of the total number of Ge atoms) but also found the presence of regular ABAB building blocks (A = Ge, Sb, B = Te) alternatively referred to as four-membered rings, square rings, or even-membered rings [46.45, 46.51, 46.52, 46.54] in amorphous GST . The crystallization was explained by the ordering of pre-existing four-membered rings.

The proposal that the bonding in the crystalline state of GST is resonant while in the amorphous state it is purely covalent [46.23, 46.24, 46.25] implicitly suggested that the underlying mechanism of the phase transition consisted of the establishment – or destruction – of resonance bonding between the covalently bonded fragments. In the crystalline phase, resonant bonding leads to the ordering and alignment of p orbitals on adjacent molecular units. This alignment is lost in the amorphous phase and this has a drastic effect on materials’s properties such as optical contrast and resistivity [46.26].

The phase-change (amorphization) process can be viewed as the destruction of three-centre bonds with the transformation of symmetrical three-center Te-Ge-Te bonds in the crystalline phase with equal Ge-Te distances on both sides of the central Ge atom into asymmetric structures with different Ge-Te distances in the amorphous phase [46.59] as illustrated in Fig. 46.8. This process requires very small atomic displacements and may alternatively be viewed as the formation/destruction of connected four-membered rings [46.54] or ABAB blocks [46.51], provided the bond cut-off distances are properly chosen to account for the formation and/or destruction of the 3c–4e bonds.
Fig. 46.8a,b

Fragments of defective octahedral Ge sites in the crystalline (a) and amorphous (b) phases. Despite very similar atomic structures, indicating minute atomic motion, the character of interatomic interaction changes drastically as illustrated by the shown CDD isosurfaces (gray) obtained through DFT simulations. While in the crystalline phase the central Ge atom has covalent like interactions with four Te neighbors, in the amorphous phase, it is covalently bonded to only three Te atoms [46.59]. (Copyright 2013 by the American Physical Society. Reprinted with permission)

Subsequently, it was argued that Ge-Ge bonds in the amorphous phase are formed as a result of the formation of dynamic three-center bonds [46.60], which becomes possible due to the presence of lone-pair electrons subtended at partially hybridized threefold coordinated Ge atoms [46.61]. The Ge-Ge bonds serve to stabilize the amorphous phase. The process is illustrated in Fig. 46.9.
Fig. 46.9

(a) Schematic of the formation of a tetrahedral Ge configuration. Ge atoms are shown in green and Te atoms are shown in orange. When a Ge atom with an LP-orbital (marked A) comes close to another Ge atom (B) and is aligned with the neighboring Ge-Te bond (between atoms B–C) (left panel), a three-center A–B–C bond is established (middle), whose subsequent rupture at the opposite arm results in the formation of a GeTd–GePy configuration (between atoms A–B), leaving behind a twofold coordinated Te atom (C) (right). (b) Evolution of CDD clouds during the in-silico amorphization process using DFT simulations substantiating the schematic shown in the upper panel. CDD clouds corresponding to the LP-electrons of an sp3 hybridized Ge orbital and a Te lone-pair p-orbital can be seen in the left and right panels, respectively, in addition to increased CDD midway between Ge and Te atoms that are signatures of covalent bonds. The presence of CDD clouds on both sides of the Ge atom (marked B in the figure) in the central panel is evidence of the formation of a transient three-center Ge-Ge-Te bond. (c) Zooms into vicinities of the atoms that participate in the formation of three-center bonds. (Reproduced from [46.60])

Finally, it is important that the destabilization of the subsystem of the weaker bonds with the preserved covalent backbone not only allows for the low-temperature amorphization of the material, however, the preserved covalent backbone also ensures the memory of the initial crystalline structure making the (reverse) crystallization process fast and ensuring high cyclability of the crystallization–amorphization process.

46.2.2 Photo-Assisted Amorphization of Ge2Sb2Te5

While it is generally believed that the role of light in the phase-change process is to heat the material above the melting point – which is likely to be correct for longer pulses – exposure to shorter pulses is different. The photo-induced amorphization of Ge2Sb2Te5 upon exposure to 600 ps pulses was first studied using in-situ x-ray absorption spectroscopy [46.62].

It was found that following exposure to a short laser pulse the white-line intensity (the spectra maximum after the absorption edge) first monotonically decreases and reaches a minimum value in about 1 ns. This fast initial decrease in the white-line intensity is followed by its partial recovery and within ≈ 2 ns a new saturation value is reached. Most interestingly, the minimum value of the white-line intensity during the amorphization process is significantly higher than that corresponding to the static liquid state [46.63] suggesting that Ge2Sb2Te5 does not melt in the conventional sense upon its transformation from the crystalline to the amorphous phase [46.62].

The role of electronic excitation in the phase-change process was also investigated theoretically [46.64]. The electronic excitation was modeled by removing electrons from the high lying valence band states according to the strength of the excitation, where the authors used a jellium background charge to compensate for the loss of charged carriers. From these studies it was concluded that electronic excitation considerably lowers the critical amorphization temperature and reduces the atomic diffusion coefficient with respect to that of the corresponding liquid phase by at least one order of magnitude.

Experimental evidence for the important role of nonthermal processes is gradually growing. Thus, rapid nonthermal control of resonant bonding was demonstrated in [46.65] using a combination of single-shot femtosecond electron diffraction and optical spectroscopy. In line with this, in [46.66], ultrafast time-resolved electron diffraction and single-shot optical pump-probe measurements were used to study lattice dynamics in polycrystalline Ge2Sb2Te5 initiated by a femtosecond near-ultraviolet pulse and it was concluded that a nonthermal crystal-to-amorphous phase transition was initiated by the displacement of Ge atoms.

46.3 Present Applications and Future Trends

Phase-change memory alloys have been used since the 1990s in rewritable optical disks such as compact disks (CD), and later also in digital versatile disks (DVD ) and BluRay disks (BD). The structure of these disks is essentially the same and is shown in Fig. 46.10. What varies is the thickness of individual layers, which is determined by the optics used (the length of the numerical aperture and laser wavelength), and the exact composition of the PCM material. Thus, in DVDs Ge2Sb2Te5 is used but in order to optimize the optical contrast at shorter wavelengths Ge8Sb2Te11 is used in BDs. Figure 46.11 shows the different phase change alloys used for rewritable optical disks on a Ge-Sb-Te ternary phase diagram.
Fig. 46.10

Comparison of optics, recording densities, recording capacities and disk structures used for CD, DVD, and BD. Changes in the lens system are shown in the upper portion; the disk structure is depicted beneath. The change in beam cross-section is displayed at the bottom. From CD to DVD and DVD to BD, the effective laser beam cross section has been reduced by a factor of 2 and 5, respectively. The recording density increased by a factor of 7 going from CD to DVD. Going from DVD to the dual-layer BD another increase by a factor of 11 has been accomplished. (Reprinted from [46.67] by permission from Macmillan Publishers Ltd., copyright (2007))

Fig. 46.11

Ternary phase diagram depicting different phase-change alloys , their year of discovery as a phase-change alloy and their use in different optical storage products. (Reprinted from [46.67] by permission from Macmillan Publishers Ltd., copyright (2007))

Electrical cell designs for the large-scale integration of phase-change devices have reached commercial development. While lateral device designs such as that proposed by NXP Semiconductors (Fig. 46.12) showed early initial promise [46.68], industry has focused on variations of the mushroom cell design (Fig. 46.13 ) with a layer of phase-change material (typically GST) sandwiched between two conducting electrodes due to the greater areal density achievable as well as the possibility for three-dimensional stacking of cells. Energy requirements for mushroom like cells have been found to scale roughly linearly with contact area down to effective contact diameters of 3 nm [46.69, 46.70]. In 2010, Samsung reported on the commercial production of 512 Mbit PRAM chips for use its cellular phones. In 2012, a 20 nm sized mushroom-type device was also announced in a 8 Gb chip by Samsung [46.71]. In 2012, Micron also reported full-scale commercial production of 45 nm node, 1 Gbit PRAM cells. Micron and Intel in July 2015 jointly announced a new generation of what is believed to be a new stackable form of PRAM called 3-D cross-point memory with the initial chips produced reportedly having a capacity of 128 Gbit. These chips were slated to reach the market in 2017. Considering the widespread use of memory devices, for example, in computers or mobile phones, it may be fair to say that most households in the world have phase-change memory devices. Few other electronic materials can compete with this.
Fig. 46.12

A schematic of a lateral type line cell along with a SEM image of an actual cell. (Reprinted from [46.68] by permission from Macmillan Publishers Ltd., copyright (2005))

Fig. 46.13

A schematic of a mushroom-type cell in both RESET and SET states

A major development in energy efficiency was realized by the spatial separation of GeTe and Sb2Te3 into atomically thin layers. Devices based upon this structure were recently shown to result in a significant reduction (by 90%) in energy consumption, which was interpreted as being due to significantly decreased entropic losses [46.72] arising from the one-dimensional motion of Ge atoms at interfaces [46.72, 46.73]. As a consequence, this kind of memory materials was named interfacial phase-change memory (iPCM ).

Subsequent studies demonstrated that iPCM, as well as hexagonal Ge2Sb2Te5 with certain stacking order, can be a topological insulator or a Dirac semimetal [46.74, 46.75]. In addition, interesting magnetic responses from iPCM were observed [46.76, 46.77] that were not present in composite material of the same average composition, making iPCM-based structures promising for spintronic and topotronic applications.

In another development, researchers have grown PRAM on flexible (Kapton) substrates with bottom electrode dimensions of 150 nm. Problems with conventional submicron lithography and flexible substrates were avoided by using a Si-based block co-polymer for patterning in lieu of conventional optical lithography [46.78]. Repeated cycling even after flexing of the substrate demonstrated the potential application of such cell designs to future wearable electronics.

Although the ideas behind neuron-based computing date back to the 1940s with the advent the theory of threshold computing [46.79], the development of electrical devices capable of carrying out such operations in a physical system came much later. In the 1980s with the pioneering work of Hopfield et al. [46.80], the concept of neural networks came into vogue. Ovshinksky was the first person to discuss the cognitive computing and neural networks in the context of phase change materials [46.81]. An example of Ge2Sb2Te5 based synapses exhibiting the spike-timing-dependent plasticity required for neuromorphic computing was reported in 2012 [46.82]. More recently, a large-scale 165000 synapse neural network based upon phase-change materials was fabricated and in a prototypical use applied to the recognition of handwriting [46.83].

With the commercial realization of phase-change memory, the field of phase-change materials is primed for further progress. With parallel developments occurring in diverse areas such as neuromorphic computing, iPCM structures, as well as manipulation of the topological properties of chalcogenide-based superlattices , the next 10 year of research will bring a rich array of different devices to the forefront of technology.

References

  1. 46.1
    S. Ovshinsky: Phys. Rev. Lett. 21, 1450 (1968)CrossRefGoogle Scholar
  2. 46.2
    A.D. Pearson, W. Northover, J.F. Dewald, W. Peck: Chemical, physical, and electrical properties of some unusual inorganic glasses in Glass Tech (Plenum, New York 1962) p. 357Google Scholar
  3. 46.3
    B.T. Kolomiets, E.A. Lebedev: Radio Eng. Electron. 8, 1941 (1963)Google Scholar
  4. 46.4
    K. Rabe, J. Joannopoulos: Phys. Rev. B 36, 6631 (1987)CrossRefGoogle Scholar
  5. 46.5
    J.-P. Gaspard, A. Pellegatti, F. Marinelli, C. Bichara: Philos. Mag. B 77, 727 (1998)CrossRefGoogle Scholar
  6. 46.6
    T. Chattopadhyay, J. Boucherle, H. Von Schnering: J. Phys. C 20, 1431 (1987)CrossRefGoogle Scholar
  7. 46.7
    P. Fons, A.V. Kolobov, M. Krbal, J. Tominaga, K. Andrikopoulos, S. Yannopoulos, G. Voyiatzis, T. Uruga: Phys. Rev. B 82, 155209 (2010)CrossRefGoogle Scholar
  8. 46.8
    T. Matsunaga, P. Fons, A.V. Kolobov, J. Tominaga, N. Yamada: Appl. Phys. Lett. 99, 231907 (2011)CrossRefGoogle Scholar
  9. 46.9
    I. Petrov, R. Imamov, Z. Pinsker: Sov. Phys. Cryst. 13, 339 (1968)Google Scholar
  10. 46.10
    B. Kooi, J.T.M.D. Hosson: J. Appl. Phys. 92, 3584 (2002)CrossRefGoogle Scholar
  11. 46.11
    T. Matsunaga, N. Yamada, Y. Kubota: Acta Cryst. B 60, 685 (2004)CrossRefGoogle Scholar
  12. 46.12
    N. Yamada: MRS Bull. 21, 48 (1996)CrossRefGoogle Scholar
  13. 46.13
    T. Nonaka, G. Ohbayashi, Y. Toriumi, Y. Mori, H. Hashimoto: Thin Solid Films 370, 258 (2000)CrossRefGoogle Scholar
  14. 46.14
    N. Yamada, T. Matsunaga: J. Appl. Phys. 88, 7020 (2000)CrossRefGoogle Scholar
  15. 46.15
    T. Matsunaga, R. Kojima, N. Yamada, K. Kifune, Y. Kubota, Y. Tabata, M. Takata: Inorg. Chem. 45, 2235 (2006)CrossRefGoogle Scholar
  16. 46.16
    T. Matsunaga, N. Yamada: Phys. Rev. B 69, 104111 (2004)CrossRefGoogle Scholar
  17. 46.17
    T. Matsunaga, R. Kojima, N. Yamada, K. Kifune, Y. Kubota, M. Takata: Appl. Phys. Lett. 90, 161919 (2007)CrossRefGoogle Scholar
  18. 46.18
    T. Matsunaga, H. Morita, R. Kojima, N. Yamada, K. Kifune, Y. Kubota, Y. Tabata, J.-J. Kim, M. Kobata, E. Ikenaga: J. Appl. Phys. 103, 093511 (2008)CrossRefGoogle Scholar
  19. 46.19
    B. Hyot, X. Biquard, L. Poupinet: Local structure of amorphous and crystalline GeTe and GeSbTe, Proc. Eur. Phase Change Symp. (2001)Google Scholar
  20. 46.20
    A. Kolobov, P. Fons, A. Frenkel, A. Ankudinov, J. Tominaga, T. Uruga: Nat. Mater. 3, 703 (2004)CrossRefGoogle Scholar
  21. 46.21
    S. Shamoto, N. Yamada, T. Matsunaga, T. Proffen, J.W. Richardson Jr., J.-H. Chung, T. Egami: Appl. Phys. Lett. 86, 081904 (2005)CrossRefGoogle Scholar
  22. 46.22
    S.I. Shamoto, K. Kodama, S. Iikubo, T.U. Taguchi, N. Yamada, T. Proffen: Jpn. J. Appl. Phys. 45, 8789 (2006)CrossRefGoogle Scholar
  23. 46.23
    G. Lucovsky, R. White: Phys. Rev. B 8, 660 (1973)CrossRefGoogle Scholar
  24. 46.24
    J. Robertson, K. Xiong, P.W. Peacock: Thin Solid Films 515, 7538 (2007)CrossRefGoogle Scholar
  25. 46.25
    K. Shportko, S. Kremers, M. Woda, D. Lencer, J. Robertson, M. Wuttig: Nat. Mater. 7, 653 (2008)CrossRefGoogle Scholar
  26. 46.26
    B. Huang, J. Robertson: Phys. Rev. B 81, 081204R (2010)CrossRefGoogle Scholar
  27. 46.27
    A.V. Kolobov, M. Krbal, P. Fons, J. Tominaga, T. Uruga: Nat. Chem. 3, 311 (2011)CrossRefGoogle Scholar
  28. 46.28
    M. Wuttig, D. Lüsebrink, D. Wamwangi, W. Welnic, M. Gillessen, R. Dronskowski: Nat. Mater. 6, 122 (2007)CrossRefGoogle Scholar
  29. 46.29
    A.H. Edwards, A.C. Pineda, P.A. Schultz, M.G. Martin, A.P. Thompson, H.P. Hjalmarson, C.J. Umrigar: Phys. Rev. B 73, 045210 (2006)CrossRefGoogle Scholar
  30. 46.30
    F. Betts, A. Bienenstock, S.R. Ovshinsky: J. Non-Cryst. Solids 4, 554 (1970)CrossRefGoogle Scholar
  31. 46.31
    D. Dove, M. Heritage, K. Chopra, S. Bahl: Appl. Phys. Lett. 16, 138 (1970)CrossRefGoogle Scholar
  32. 46.32
    O. Uemura, Y. Sagara, M. Tsushima, T. Kamikawa, T. Satoh: J. Non-Cryst. Solids 33, 71 (1979)CrossRefGoogle Scholar
  33. 46.33
    M. Wakagi, Y. Maeda: Phys. Rev. B 50, 14090 (1994)CrossRefGoogle Scholar
  34. 46.34
    A. Kolobov, P. Fons, J. Tominaga, A. Ankudinov, S. Yannopoulos, K. Andrikopoulos: J. Phys. Cond. Mat. 16, S5103 (2004)CrossRefGoogle Scholar
  35. 46.35
    K. Hirota, K. Nagino, G. Ohbayashi: J. Appl. Phys. 82, 65 (1997)CrossRefGoogle Scholar
  36. 46.36
    G. Fisher, J. Tauc, Y. Verhelle: Vibrational modes and bonding in the amorphous GexTe1-x system. In: Amorphous and Liquid Semiconductors, ed. by J. Stuke (Taylor and Francis, London 1974) p. 1259Google Scholar
  37. 46.37
    S. Hosokawa, Y. Hari, T. Kouchi, I. Ono, H. Sato, M. Taniguchi, A. Hiraya, Y. Takata, N. Kosugi, M. Watanabe: J. Phys.-Cond. Mat. 10, 1931 (1998)CrossRefGoogle Scholar
  38. 46.38
    S.J. Pickart, Y.P. Sharma, J.P. de NeufvilleL: J. Non-Cryst. Solids 34, 183 (1979)CrossRefGoogle Scholar
  39. 46.39
    P. Boolchand, B.B. Triplett, S.S. Hanna, J.P. de Neufville: Mossbauer Effect Methodology (Plenum, New York 1974)Google Scholar
  40. 46.40
    P. Pykkö, M. Atsumi: Chem. Eur. J. 15, 186 (2009)CrossRefGoogle Scholar
  41. 46.41
    A. Ankudinov, B. Ravel, J. Rehr, S. Conradson: Phys. Rev. B 58, 7565 (1998)CrossRefGoogle Scholar
  42. 46.42
    P. Jovari, I. Kaban, J. Steiner, B. Beuneu, A. Schops, M.A. Webb: Phys. Rev. B 77, 035202 (2008)CrossRefGoogle Scholar
  43. 46.43
    D. Baker, M. Paesler, G. Lucovsky, P. Taylor: J. Non-Cryst. Solids 352, 1621 (2006)CrossRefGoogle Scholar
  44. 46.44
    M. Paesler, D. Baker, G. Lucovsky, A. Edwards, P. Taylor: J. Phys. Chem. Solids 68, 873 (2007)CrossRefGoogle Scholar
  45. 46.45
    S. Kohara, K. Kato, S. Kimura, H. Tanaka, T. Usuki, K. Suzuya, H. Tanaka, Y. Moritomo, T. Matsunaga, N. Yamada: Appl. Phys. Lett. 89, 201910 (2006)CrossRefGoogle Scholar
  46. 46.46
    K.V. Mitrofanov, A.V. Kolobov, P. Fons, X. Wang, J. Tominaga, Y. Tamenori, T. Uruga, N. Ciocchini, D. Ielmini: J. Appl. Phys. 115, 173501 (2014)CrossRefGoogle Scholar
  47. 46.47
    S. Gabardi, S. Caravati, G.C. Sosso, J. Behler, M. Bernasconi: Phys. Rev. B 92, 054201 (2015)CrossRefGoogle Scholar
  48. 46.48
    W. Welnic, A. Pamungkas, R. Detemple, C. Steimer, S. Blugel, M. Wuttig: Nat. Mater. 5, 56 (2005)CrossRefGoogle Scholar
  49. 46.49
    S. Caravati, M. Bernasconi, T. Kühne, M. Krack, M. Parrinello: Appl. Phys. Lett. 91, 171906 (2007)CrossRefGoogle Scholar
  50. 46.50
    S. Caravati, M. Bernasconi, T.D. Kuehne, M. Krack, M. Parrinello: J. Phys. Cond. Mat. 21, 499803 (2009)CrossRefGoogle Scholar
  51. 46.51
    J. Akola, R. Jones: Phys. Rev. B 76, 235201 (2007)CrossRefGoogle Scholar
  52. 46.52
    J. Akola, R. Jones: J. Phys. Cond. Mat. 20, 465103 (2008)CrossRefGoogle Scholar
  53. 46.53
    J. Akola, R. Jones: Phys. Rev. B 79, 134118 (2009)CrossRefGoogle Scholar
  54. 46.54
    J. Hegedüs, S. Elliott: Nat. Mater. 7, 399 (2008)CrossRefGoogle Scholar
  55. 46.55
    M. Micoulaut, K. Gunasekera, S. Ravindren, P. Boolchand: Phys. Rev. B 90, 094207 (2014)CrossRefGoogle Scholar
  56. 46.56
    M. Krbal, A.V. Kolobov, P. Fons, J. Tominaga, S.R. Elliott, J. Hegedus, T. Uruga: Phys. Rev. B 83, 054203 (2011)CrossRefGoogle Scholar
  57. 46.57
    L. Pauling: The Nature of the Chemical Bond and the Structure of Molecules and Crystals: An Introduction to Modern Structural Chemistry (Cornell Univ. Press, Ithaca 1960)Google Scholar
  58. 46.58
    M. Xu, Y. Cheng, H. Sheng, E. Ma: Phys. Rev. Lett. 103, 195502 (2009)CrossRefGoogle Scholar
  59. 46.59
    A.V. Kolobov, P. Fons, J. Tominaga, S.R. Ovshinsky: Phys. Rev. B 87, 165206 (2013)CrossRefGoogle Scholar
  60. 46.60
    A.V. Kolobov, P. Fons, J. Tominaga: Sci. Rep. (2015), doi: 10.1038/srep13698
  61. 46.61
    A.V. Kolobov, P. Fons, J. Tominaga: Phys. Rev. B 87, 155204 (2013)CrossRefGoogle Scholar
  62. 46.62
    P. Fons, H. Osawa, A.V. Kolobov, T. Fukaya, M. Suzuki, T. Uruga, N. Kawamura, H. Tanida, J. Tominaga: Phys. Rev. B 82, 041203 (2010)CrossRefGoogle Scholar
  63. 46.63
    A. Kolobov, P. Fons, M. Krbal, R. Simpson, S. Hosokawa, T. Uruga, H. Tanida, J. Tominaga: Appl. Phys. Lett. 95, 241902 (2009)CrossRefGoogle Scholar
  64. 46.64
    X.-B. Li, X.Q. Liu, X. Liu, D. Han, Z. Zhang, X.D. Han, H.-B. Sun, S.B. Zhang: Phys. Rev. Lett. 107, 015501 (2011)CrossRefGoogle Scholar
  65. 46.65
    K.V. Mitrofanov, P. Fons, K. Makino, R. Terashima, T. Shimada, A.V. Kolobov, J. Tominaga, V. Bragaglia, A. Giussani: Sci. Rep. 6(20), 633 (2016)Google Scholar
  66. 46.66
    M. Hada, W. Oba, M. Kuwahara, I. Katayama, T. Saiki, J. Takeda, K.G. Nakamura: Sci. Rep. 5, 13530 (2015)CrossRefGoogle Scholar
  67. 46.67
    M. Wuttig, N. Yamada: Nat. Mater. 6, 824 (2007)CrossRefGoogle Scholar
  68. 46.68
    M. Lankhorst, B. Ketelaars, R. Wolters: Nat. Mater. 4, 347 (2005)CrossRefGoogle Scholar
  69. 46.69
    H.-S. Wong, S. Raoux, S. Kim, J. Liang, J.P. Reifenberg, B. Rajendran, M. Asheghi, K.E. Goodson: Proc. IEEE 98, 2201 (2010)CrossRefGoogle Scholar
  70. 46.70
    F. Xiong, A.D. Liao, D. Estrada, E. Pop: Science 332, 568 (2011)CrossRefGoogle Scholar
  71. 46.71
    Y. Choi, I. Song, M.-H. Park, H. Chung, S. Chang, B. Cho, J. Kim, Y. Oh, D. Kwon, J. Sunwoo, J. Shin, Y. Rho, C. Lee, M.G. Kang, J. Lee, Y. Kwon, S. Kim, J. Kim, Y.-J. Lee, Q. Wang, S. Cha, S. Ahn, H. Horii, J. Lee, K. Kim, H. Joo, K. Lee, Y.-T. Lee, J. Yoo, G. Jeong: A 20 nm 1.8 V 8 Gb PRAM with 40 MB/s program bandwidth, Proc. IEEE Int. Solid-State Circuits Conf. Dig. Tech. Pap. (ISSCC) (2012) p. 46Google Scholar
  72. 46.72
    R.E. Simpson, P. Fons, A.V. Kolobov, T. Fukaya, M. Krbal, T. Yagi, J. Tominaga: Nat. Nanotech. 6, 501 (2011)CrossRefGoogle Scholar
  73. 46.73
    X. Yu, J. Robertson: Sci. Rep. (2015), doi: 10.1038/srep12612
  74. 46.74
    A.A. Burkov, L. Balents: Phys. Rev. Lett. 107, 127205 (2011)CrossRefGoogle Scholar
  75. 46.75
    J. Tominaga, A.V. Kolobov, P. Fons, T. Nakano, S. Murakami: Adv. Mat. Interf. 1, 1300027 (2014)CrossRefGoogle Scholar
  76. 46.76
    J. Tominaga, R.E. Simpson, P. Fons, A.V. Kolobov: Appl. Phys. Lett. 99, 152105 (2011)CrossRefGoogle Scholar
  77. 46.77
    D. Bang, H. Awano, J. Tominaga, A.V. Kolobov, P. Fons, Y. Saito, K. Makino, T. Nakano, M. Hase, Y. Takagaki: Sci. Rep. (2014), doi: 10.1038/srep05727
  78. 46.78
    B.H. Mun, B.K. You, S.R. Yang, H.G. Yoo, J.M. Kim, W.I. Park, Y. Yin, M. Byun, Y.S. Jung, K.J. Lee: ACS Nano 9, 4120 (2015)CrossRefGoogle Scholar
  79. 46.79
    W. McCulloch, W. Pitts: Bull. Math. Biophys. 5, 115 (1943)CrossRefGoogle Scholar
  80. 46.80
    J.J. Hopfield: Proc. Nat. Acad. Sci. 79, 2554 (1982)CrossRefGoogle Scholar
  81. 46.81
    S.R. Ovshinsky, B. Pashmakov: Innovation providing new mulitiple functions in phase-change materials to achieve cognitive computing, Proc. Symp. GG/HH – Adv. Charact. Tech. Data Storage Mater. – Phase Change Nonmagnetic Mater. Data Storage (2003)Google Scholar
  82. 46.82
    D. Kuzum, R.G.D. Jeyasingh, B. Lee, H.-S.P. Wong: Nano Lett. 12, 2179 (2012)CrossRefGoogle Scholar
  83. 46.83
    G. Burr, R. Shelby, S. Sidler, C. di Nolfo, J. Jang, I. Boybat, R. Shenoy, P. Narayanan, K. Virwani, E. Giacometti: IEEE T. Electron Dev. 62, 3498 (2015)CrossRefGoogle Scholar

Copyright information

© Springer International Publishing AG 2017

Authors and Affiliations

  • Alexander V. Kolobov
    • 1
  • Junji Tominaga
    • 2
  • Paul Fons
    • 2
  1. 1.National Institute of Advanced Industrial Science and TechnologyTsukubaJapan
  2. 2.National Institute of Advanced Industrial Science and TechnologyTsukubaJapan

Personalised recommendations