In situ TEM study of crystallization and chemical changes in an oxidized uncapped Ge₂Sb₂Te₅ film

Phase-change materials are an important class of materials for their applications in optical and electronic memories driven by fast and reversible, temperature-induced amorphous to crystalline transformation.^1–5 Phase change memory (PCM) devices utilize the large contrast in optical reflectivity or electrical conductivity between the amorphous (reset) and crystalline (set) states.^6–8 The phase transformations have been studied extensively by temperature-dependent optical and electrical characterization. Ge–Sb–Te alloys have been extensively investigated by phase-change materials owing to their fast phase transformation, good thermal stability, and high cyclic endurance.^9–12 Ge₂Sb₂Te₅ (GST-225), in particular, has been the most widely used composition for electronic memories due to suitable crystallization temperature, resistance ratio between crystalline and amorphous states (c-GST and a-GST), and stability of the amorphous phase.¹³ Amorphous GST225 undergoes a first phase transformation to a metastable face-centered cubic (fcc) phase at ∼150 °C. With further heating, it transforms polymorphically to a stable hexagonal phase at ∼300 °C and it melts at ∼600 °C.¹⁴ Amorphization is obtained by melt-quenching of a small volume of material.

The crystallization temperature is a critical parameter that, together with others, determines the speed, power consumption, and retention of the PCM devices. The crystallization temperature (T_x) can be changed by incorporating dopants into GST-225 as demonstrated by several groups.^15–19 Light element dopants such as carbon, nitrogen, and oxygen were shown to influence T_x significantly. The change in T_x has been reported from 150.6 to 300.6 °C in carbon doped GST-225.²⁰ Song et al.¹⁹ have shown that a GST-225 film doped with 3.0 wt. % nitrogen crystallized at ∼200 °C, whereas the undoped counterpart crystallized at ∼130 °C. Jang et al.¹⁸ reported the suppression of the crystallization from 150 °C to 200 °C when 16.7 and 21.7 wt. % oxygen were incorporated into GST-225. Golovchak et al.¹⁷ showed that in the presence of oxygen, the initiation of the transformation to the cubic atomic arrangement is significantly hindered and consequently the crystallization is delayed. In contrast, crystallization is promoted when the films are oxidized. Kooi et al.²¹ studied oxidized GST-225 films and observed a significantly reduced crystallization temperature from 130 °C to 35 °C. Agati et al.²² reported that for an uncapped Ge-rich GST thin film, the crystallization temperature decreases by 50–60 °C as an effect of the air exposure. Berthier et al.²³ also showed that capping of GeTe films resulted in an ∼50 °C increase in crystallization temperature, compared to uncapped, oxidized films. It is apparent that when O is incorporated in the GST and forms a solid solution, this hinders the crystallization and subsequently delays the T_x. In contrast to this, oxidation promotes the chemical segregation and may lead to entirely different crystallization kinetics.

An aberration-corrected TEM equipped with a heating stage has been employed to investigate the effects of heating on crystallization in both imaging and selected-area diffraction (SAD) modes. Crystallization of phase-change materials has been studied by in situ observation of their microstructure and morphology in a TEM with a heating stage.^{19,21,24–27} However, the temporal resolution in these earlier studies was usually not sufficient to monitor the amorphous-to-crystalline phase transformation. The reason for these limitations is twofold. First, drift in the sample during heating in conventional heating holders leads to an inability to capture the images even at low magnifications. This may be partly circumvented using a MEMS-based heating holder such as that provided by Protochips (Aduro 300 fusion). Second, an acquisition camera with high temporal resolution and a large field of view is desired to capture the morphological and structural changes at the atomic scale. Of course, samples for TEM experiments must be very thin.

The present investigation makes use of uncapped GST-225 films (20 nm) directly deposited onto the SiN_x heating chip giving a unique island morphology. The as-deposited films have an amorphous island morphology, different from those in most other reports. Moreover, the present work used a high-speed, direct single-electron-detection camera with a large field of view and optimized for low-dose imaging. This enables the crystallization events to be recorded at their initial stages due to relatively higher temporal resolution. In addition, the low electron dose significantly minimizes the effect of beam-induced phase transformations. The present work deals with the early stages of crystallization and growth kinetics of the uncapped GST-225 film and the effect of oxygen on its microchemistry.

The Ge₂Sb₂Te₅ film of thickness 20 nm was deposited directly on a SiN_x membrane by magnetron sputtering from the Ge₂Sb₂Te₅ target. The details of the deposition conditions are given elsewhere.²⁸ In situ heating experiments have been carried out in a C_s-corrected FEI Titan ETEM operated at 300 kV employing an Aduro 300 Protochips holder (manufacturer's stated temperature accuracy 95%) with a heating rate of 5 °C/s from room temperature to 140 °C. Structural and morphological evolution were observed using a Gatan K3 IS camera at 20 frames per second with a low dose rate [6.7 e/(Ų/s)] to minimize the effect of the e-beam on the sample.²⁹ Microchemical characterization was carried out using FEI G2 80-200 chemiSTEM in probe-corrected mode with a monochromated beam. The chemistry of the film was investigated using a four-quadrant x-ray energy-dispersive spectroscopy (XEDS) detector. The chemistry of the as-deposited film was estimated to be Ge_2.7Sb_2.2Te₅, which shows a higher Ge concentration than expected. The Ge-rich GST-225 has already been reported when the film was sputter deposited using the Ge₂Sb₂Te₅ target.²⁸ The electron energy-loss spectrometry (EELS) was carried out using a GIF attached to a FEI Tecnai F30. The principles of these TEM-based techniques are well documented in the literature.³⁰ 

A python script run from within Digital Micrograph was used to process the in situ video dataset to produce a map of lattice-spacing orientation and visibility from individual frames in the dataset. These maps are based on FFTs of sub-regions and are displayed in color, where the hue corresponds to the direction of the strongest lattice-spacing peak in the diffractogram, while the brightness corresponds to the intensity of this peak. Since the Python script runs within Digital Micrograph, no data conversion was needed to process the raw data, and the result could be visualized in the same software as the raw data.

Figure 1 shows a BF-TEM micrograph and the corresponding SAD pattern as an inset of the as-deposited GST-225 films of thickness 20 nm acquired at room temperature. The microstructure shows an island morphology with an average size of ∼15 nm. The diffuse rings in the SAD pattern clearly show the amorphous nature of the film. The microchemical nature of the film, as deposited at room temperature, is shown in Fig. 2. Figure 2(a) shows an XEDS plot from a relatively large area where peaks of Ge, Sb, and Te are observed. Additional peaks corresponding to Si and N originate from the SiN_x support of the Protochips. For further confirmation of the presence of O and N, EELS was carried out from the same region which clearly shows N-K and O-K peaks at 401 eV and 532 eV, respectively [Fig. 2(b)]. The composition of the film was determined from XEDS to be Ge_2.7Sb_2.2Te₅. EELS confirms the presence of a significant amount of O in the as-deposited film. The oxygen concentration was estimated as ∼17.8 at. % from the EELS profile. However, the presence of N from the SiN_x support membrane is prevalent in the EELS spectrum and may introduce uncertainties in the quantification. The presence of O has been shown to influence the crystallization temperature in GST-225. The presence of O in the films can also affect its morphology as demonstrated previously. In the case of the uncapped GST-225 film, it was shown that Ge and O combine and segregate in the inter-island regions leaving behind Sb-Te rich islands at 200 °C.²⁸ 

The film was subjected to heat-treatment from room temperature to 140 °C to study the crystallization behavior and structural transformation inside the TEM. The representative microstructures at different temperatures (25 °C, 35 °C, 100 °C, and 140 °C) are depicted in Fig. 3. The corresponding power spectra are shown as an inset in each frame. The detailed video of the in situ transformation is given in the supplementary material. The in situ heating reaction was carried out with a ramp rate of 5 °C/s from room temperature (25 °C) to 140 °C. The initial morphology of the film before heating is shown in Fig. 3(a) and is considered as a reference for further comparison during heat-treatment. The film is initially amorphous in nature as evidenced by the corresponding power spectrum.

In the recorded video, the onset of the crystallization was observed after just 2 s, which corresponds to a temperature of 35 °C and is presented in Fig. 3(b). The appearance of the long-range frequencies is manifested in the power spectrum with a prominent pair of spots. One such crystallized domain is marked in the micrograph and a magnified image is presented in Fig. 3(e). The lattice spacing of the initially transformed crystallite is measured to be 0.34 nm, which corresponds to the (111) plane spacing of fcc-GST. The observed amorphous-to-crystalline transformation temperature (35 °C) in the present investigation is about 100 °C lower than those of reported values for GST-225. This observation of a lower crystallization temperature can be attributed to (1) earlier crystallization detection capability due to the large field of view of the K3 camera, (2) advanced nucleation stage in the aged film, (3) heterogeneous nucleation and significant amount of O present in the uncapped film of island morphology, and (4) the electron beam effects on the crystallization, notwithstanding the very low electron dose used to acquire images. The present investigation uses a large field of view and a low-dose single-electron-detection camera where the effect of the beam damage was significantly minimized. However, this reduces the contrast of the image due to the low signal. The similar chemistry of the crystallized grains and the amorphous matrix further reduces the contrast of the recorded images in the film under investigation. Figure 3(c) depicts the morphology and power spectrum of the film heated at 100 °C. The power spectrum clearly shows the increase in the number density of the nucleated crystallites, which is corroborated by the appearance of additional long-range frequencies as indexed. The magnified view of one such crystallite is shown in Fig. 3(f). The crystalline domains at 140 °C are shown in Fig. 3(d), and the magnified version of one such crystallite is given in Fig. 3(g).

The contrast-enhanced image along with the originally recorded one at 140 °C with a low electron dose [6.7 e/(Ų/s)] after processing with Python scripting is shown in Fig. 4. The three selected regions and their corresponding processed counterparts are depicted in Figs. 4(a) and 4(b), respectively.

The dynamics of the early stages of crystallization is deduced from the in situ heating process as illustrated in the plot of density of crystallites as a function of temperature [cf. Figure 5(a)]. The number density of crystallites was determined from the processed images recorded using a low dose and large field of view direct detection camera. While processing, a power spectrum was generated from a ROI of 128 × 128 pixels and the entire image was scanned. Only the regions that exhibit distinct periodic spots in the power spectrum are considered crystalline. In the recorded micrographs under the present investigation, 128 pixels correspond to ∼4.7 nm and this is therefore considered to be the threshold dimension. The possibility of the crystalline regions with size less than 4.7 nm, however, is not ruled out and fluctuation electron microscopy is commonly used to probe the medium range order below 3 nm.³¹ The quantification of the crystallites in the present work has been exercised for crystals which are relatively larger in size compared to barely supercritical nuclei of size ranging between 2 and 4.7 nm. This is a typical nucleation curve showing incubation temperature below 35 °C followed by nucleation and growth regime from 35 °C to 90 °C. The average size of the crystallites measured at each temperature is presented in Fig. 5(b). The average size of the crystallites measured at 40 °C, 65 °C, and 140 °C is ∼8 nm, 13 nm, and 15 nm, respectively. The increase in the average size is quite fast until 75 °C, and beyond that the average size of the crystallites is observed to increase slowly with a rise in temperature. The maximum density of crystallites has been observed at 90 °C and nucleated crystallites coalesce with a further increase in the temperature. A lowering of the crystallization temperature in the presence of oxygen has previously been reported by Kooi et al.²¹ They have shown that a 10 nm oxidized GST-225 film is fully crystallized at 35 °C. However, this phenomenon does not appear to have been reported for a sample with this type of island morphology. We have previously observed the initiation of the long-range order for similar morphology of the 30 nm GST-225 film in identical deposition and oxidation conditions at higher temperature (∼130 °C).²⁸ Hence, we believe that the main difference here behind the observation of significantly lower crystallization temperature is the earlier crystallization detection capability offered by the Gatan K3 camera and the further aging of the film.

The BF-TEM micrograph and the corresponding SAD pattern of the heat-treated film at 140 °C are shown in Fig. 6. The island morphology persists even after the heat-treatment with no significant changes in their dimensions. The crystallites that are formed within the islands are indicated by changes in the contrast of the BF micrograph. The diffraction rings confirm the crystallized nature of the film and are indexed as (111), (220), and (311) planes of the fcc-GST-225 phase, as marked. The microchemical nature of the film after heat-treatment is presented in Fig. 7. The measured composition of Ge, Sb, and Te from XEDS was found to be Ge_2.8Sb_2.1Te₅, which does not show a significant deviation from the as-deposited composition (cf., Fig. 2). The corresponding EELS profile shows the presence of a significant amount of O after heat-treatment [Fig. 7(b)], which does not show any significant deviation from the as-deposited film. The STEM-XEDS elemental maps of Ge, Sb, and Te of the films at room temperature and also after heat-treatments at 100 °C and 140 °C are shown in Fig. 8(a). It is apparent from the room-temperature maps that inter-island regions are Ge-rich, whereas Sb and Te are uniformly distributed. This tendency of chemical partitioning increases with increasing temperature. The O distributions in the film at the same temperatures (RT, 100 °C, and 140 °C) are displayed in Fig. 8(b). These maps confirm that the islands are rich in Sb and Te whereas inter-island regions are comprised of Ge and O. Chemical partitioning of GST-225 thus appears to be energetically more favorable in the presence of O rather than crystallization into known crystalline phases. In the present condition, the formation of Ge–O and Sb₂Te₃ is thermodynamically more viable owing to a higher Ge–O bond dissociation energy than that of Ge–Te, Te–O, and Sb–O.³² The chemical maps of a similar film have been discussed previously.²⁸ 

A representative HRTEM micrograph acquired at 140 °C showing three different crystallites is displayed in Fig. 9(a). The corresponding power spectrum of each of the crystallites (marked in red, blue, and green color) is also shown along with the indexed planes in Figs. 9(b)–9(d). The crystallites marked in blue and green are indexed as fcc-GST-225 (ICDD # 00-054-0454) phase whereas those marked in red matches with rhombohedral Sb₂Te₃ (ICDD # 01-071-0393).

The following conclusions can be drawn from the present study:

1. The Ge-rich uncapped and aged GST-225 film with island morphology crystallizes at ∼35 °C, as observed using a low electron-dose camera on a high temporal and spatial resolution.

2. The phase transformation is nucleation-dominated until ∼90 °C and beyond this, growth supersedes nucleation.

3. The observation of a lower crystallization temperature is mainly attributed to the earlier crystallization detection capability offered by the K3 camera and to an advanced nucleation stage due to the aging of the film.

4. The exposure to atmosphere of the uncapped film promotes the chemical segregation of Ge, Sb, and Te. The islands are rich in Sb and Te whereas inter-island regions are comprised of Ge and O.

5. Crystallization kinetics in phase-change materials can be studied in detail by TEM with a high-speed low electron dose camera.

See the supplementary material for the video of the in situ transformation of GST-225 during heating from room temperature to 140 °C. The recorded video shows the transformation in real space as well as in reciprocal space (embedded FFT in the video).

M.K.S. and C.G. contributed equally to this work.

This research is supported by the National Science Foundation (NSF) under Award No. DMR-1710468. The TEM studies were carried out at the Center for Integrated Nanotechnologies, an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science, and in the Materials Characterization Department, Sandia National Laboratories. The GST films were deposited at the Institute of Materials Science and Nanotechnology (UNAM) at Bilkent University, Turkey. The authors thank Dr. Matthew T. Janish and Dr. Ali Gokirmak for helpful discussions. The use of the K3 IS camera was provided courtesy of Gatan. The Sandia National Laboratories are managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the U.S. DOE's NNSA under Contract No. DE-NA-0003525. The views expressed here do not necessarily represent the views of the U.S. DOE or the U.S. Government. Los Alamos National Laboratory, an affirmative action equal opportunity employer, is managed by Triad National Security, LLC for the U.S. Department of Energy's NNSA, under Contract No. 89233218CNA000001.

The data that support the findings of this study are available from the corresponding author upon reasonable request.