The effects of Sb/Te ratio on crystallization kinetics in Ge-rich GeSbTe phase-change materials

Phase-Change Memory (PCM) has proven its high maturity in terms of scalability, fast programming speed, and high endurance compared to other emerging non-volatile memory technologies.¹ Ge $2$Sb $2$Te $5$ (GST225) is the most commonly referenced and widely studied alloy for PCM devices.^2–7 However, GST225 has a low crystallization temperature of about 160  $°$C, making it unsuitable for embedded applications like automotive ones. The enrichment in the Ge content and the doping with nitrogen (N) in GeSbTe alloys (GST) have shown the potential for high thermal stability, meeting the data retention requirements for high-temperature applications.^8–11 

However, Ge-rich GeSbTe (GGST) alloys are known to undergo phase segregation^9,12,13 and drift phenomena caused by amorphous residuals and Ge–Ge bond reorganization.¹⁴ This opens the quest for further developments in order to improve even more the unique properties of these alloys. It has been reported that Sb-rich GST phase-change materials can enhance the writing speed compared to Ge₂Sb₂Te₅ (GST225), featuring a rapid crystalline growth resulting in a more homogeneous layer in the PCM cell.^15–17 However, such Sb-rich alloys crystallize at low temperature, which is not compatible with applications demanding high thermal stability. Moreover, previous investigations of the structural evolution of GGST alloys with Sb-over-Te ratio (Sb/Te) lower than one emphasized the stability of Sb–Te structural units, with a rearrangement of Ge–Te bonds around them, suggesting that Sb–Te features can be considered the driver of the segregation kinetic of the system.^9,12 The same studies also demonstrated that introducing nitrogen enhances the thermal stability without affecting the crystallization dynamics of the system.

In this paper, we report on the in-depth investigation of the key role of Sb/Te in GGST and nitrogen-doped GGST alloys. We combine the results obtained with complementary techniques and measurements, i.e., resistivity vs temperature measurements (R vs T), Raman spectroscopy, in situ and ex situ x-ray diffraction (XRD). We study the evolution in the temperature of the resistivity of the GGST samples with Sb/Te higher (⁠ $Sb / Te>1$⁠) and lower than one (⁠ $Sb / Te<1$⁠), extracting the activation energy (E_a) of the crystallization for each sample. We characterize by Raman spectroscopy the structural features evolution starting from the amorphous phase up to the crystallization of Ge and GST phases at high temperature. We investigate by XRD the crystallization kinetics of GGST alloys highlighting the phases appearing during the annealing up to 550  $°$C together with the crystallites’ size. The in situ XRD analyses allow us to access to the E_a of the nucleation of both Ge and GST phases, showing an increased thermal stability in GGST alloys with a $Sb / Te>1$⁠. Finally, we demonstrate the key role of the Sb/Te ratio in driving the GST phase segregation, becoming an important knob for the development of GGST alloys for PCM applications demanding high-temperature stability.

We deposited amorphous 100 nm-thick layers of GGST on Si (100) wafers by magnetron co-sputtering from GST225 and Ge targets for $Sb / Te<1$ alloys (⁠ $Sb / Te∼0.4$⁠), and from GST225, Ge and Sb targets for $Sb / Te>1$ (⁠ $Sb / Te∼2.5$⁠) materials. Nitrogen was introduced by reactive co-sputtering from the same targets with Ar/N₂ gas mixture in the deposition chamber. The layers were protected by a thin carbon capping layer deposited in situ, in order to prevent surface oxidation.

Temperature-dependent sheet resistance experiments (R vs T) were conducted on GGST materials deposited on SiO₂ substrates and encapsulated by a GeN layer. We determined the E_a related to amorphous-to-crystalline transition by virtue of R vs T measurements at different heating rates (5, 10, 20, and 30  $°$C/min). We used an ex situ annealing fixed heating rate equal to 10  $°$C/min, to prepare samples with controlled thermal budget for Raman spectroscopy and XRD analyses.

Raman spectra were acquired at room temperature using a Renishaw inVia Raman spectrometer. A 532 nm laser diode served as the excitation source, using a focusing lens with a magnification of 100 $×$ and a numerical aperture of 0.85. The measurements were performed in the Raman shift range of 100–1000 cm $− 1$⁠. To prevent heating and unwanted crystallization of the samples, the laser power was maintained at 0.12 mW.

The ex situ XRD data were acquired using a PANalytical Empyrean two-circle diffractometer with monochromatized Cu K-alpha radiation (⁠ $λ=1.5406$ Å), in Bragg–Brentano geometry in the $20 °$– $60 °$ range in $2θ$⁠.

The in situ XRD was performed at the BM2-D2AM beamline at the European Synchrotron Radiation Facility (ESRF). The temperature-dependent XRD data were recorded using a heating plate, at different heating rates (5, 10, 20, and 30  $°$C/min). The incoming x-ray energy was 10.9 keV (⁠ $λ=1.1374$ Å). The x-ray beam was focused to a spot size of about $40×40$  $μ$m². The incidence angle was adjusted to $5 °$⁠. The XRD patterns were recorded with a 2D detector (XPAD^TM) located 220 mm from the sample, allowing to span a 2-theta range (⁠ $10 °$– $34 °$⁠). A reference sample (LaB₆) was used to calibrate the scattering angle. One single XRD image was acquired every 2 s and then corrected for baseline and spikes and processed to 1D diffraction patterns.

In Fig. 1, we present R vs T measurements performed on GGST alloys at a 10  $°$C/min heating rate. Highly resistive amorphous films undergo amorphous-to-crystalline phase transition at the crystallization temperature, characterized by a decrease in resistivity. This phenomenon is attributed to the segregation and crystallization of a low-resistive GST phase.

GGST with $Sb / Te<1$ features a double step transition with a first drop at 300  $°$C and a second one at 350  $°$C. On the contrary, GGST with $Sb / Te>1$ layer shows a single sharp transition at 365  $°$C. This behavior reveals the growth dominated nature of the segregated phase, as it is demonstrated in the following by comparing the grain sizes evolution in Sec. III C and supplementary material (Fig. S4). In situ and ex situ XRD analyses (see Fig. S3 in the supplementary material and Sec. III C) confirm that alloys with $Sb / Te<1$ segregate to cubic Ge and cubic GST225 or a close GST phase, as it was previously shown that it may exist different cubic GST alloys presenting diffraction peaks at the same angular positions.^13,18–21 On the contrary, alloys with $Sb / Te>1$ segregate to cubic Ge and a GST phase featuring a crystalline structure owing to the hexagonal crystal family, which we could address as Ge $1$1Sb $9$Te $4$ (GST194) combining in situ and ex situ XRD analyses. In the following, for the sake of simplicity, we will address hexagonal phase as the “hex” phase. N introduction delays the crystallization at higher temperatures, preserving the crystallization kinetics of the two systems with different Sb/Te ratios. By Kissinger analysis,²² we extracted the E_a of the crystallization corresponding to the first crystalline transition in Fig. 2. The plot highlights the increase of E_a with Sb/Te ratio. E_a increases even more with N introduction, as N-doping in as-deposited alloys leads to the formation of Ge–N bonds. These bonds form in as-deposited alloys and undergoes a rearrangement, which liberate some Ge atoms involved in crystallization of the Ge phase, delayed at higher temperature.⁹ 

We performed Raman spectroscopy and XRD analyses to investigate the structural evolution in GGST alloys with $Sb / Te<1$ and $Sb / Te>1$⁠. The Raman spectra of GGST alloys as-deposited and annealed at 450  $°$C are shown in Fig. 3. The spectra exhibit three main parts:⁹ 

• below 150 cm $− 1$⁠, the features are attributed to the vibrational modes of Ge–Te bonds;²³ 

• the peak at 154 cm $− 1$ is attributed to the stretching mode of SbTe₃ pyramidal units (Sb–Te bonds) present in the GeSbTe system;²⁴ 

• the third part of the spectra, covering the wavenumber range 190–300 cm $− 1$⁠, corresponds to Ge–Ge vibrational modes. The peaks at 230 and 270 cm $− 1$ are attributed, respectively, to the longitudinal optical (LO) and transverse optical (TO) vibrational modes of amorphous germanium (a-Ge). The position and the broadness of these peaks suggest that the Ge atoms are arranged in an amorphous disordered structure. The sharp peak at about 300 cm $− 1$ is attributed to the vibrational modes of crystalline germanium.^25,26

Annealing induces changes in the Sb–Te peak. In $Sb / Te<1$ alloys, the Sb–Te peak shifts toward higher frequencies, which is in agreement with the formation of a GST225 cubic phase and its gradual evolution toward the hex-GST225 as previously reported.⁹ In $Sb / Te>1$ alloys, the peak position does not change while we observe an important sharpening between the amorphous and crystalline phase, suggesting the higher order in the formed hex-GST phase with respect to the GST225 phase. We think that the well-defined peak at 120 cm $− 1$ may be associated with Sb–Sb vibrations, compatibly with a crystalline structure implying Sb planes instead of van der Waals gaps, as it was already suggested in previous studies.^27,28 For comparison, this spectrum is compared with the one of Sb in Fig. S2 in the supplementary material.

To follow the evolution trends of Ge–Ge features in both GGST alloys, with and without N introduction, we calculated the sum of the relative peak areas. We focused on the following contributions: 220 and 270 cm $− 1$ for amorphous Ge–Ge (a-Ge), and 297 cm $− 1$ for crystalline Ge (c-Ge). The results are shown in Fig. 4, at temperatures that were selected based on transitions observed in R vs T measurements.

For all the compositions, the amorphous Ge modes (a-Ge) gradually decrease, transforming into the sharp peak of crystalline Ge (c-Ge) (Fig. 3). Interestingly, for alloys without nitrogen content, crystallization is continuous when $Sb / Te<1$⁠, while Ge crystallization reaches a plateau at about 390  $°$C before continuing to increase for $Sb / Te>1$⁠. This plateau may be due to the high crystalline growth in alloys with $Sb / Te>1$⁠, which inhibits Ge crystallization until stability is achieved, after which Ge continues to grow.

For N-doped samples, an increase in the integrated intensity of the amorphous Ge modes is observed before crystallization, which is likely due to the liberation of Ge following the reorganization of Ge–N features.^9,11,29,30 Notably, in N-doped alloys, early ordering is observed in the Raman spectra (i.e., peak at about 280 cm $− 1$⁠) before the appearance of the Ge (111) Bragg peak in XRD patterns (see Fig. S3 in the supplementary material). This is likely due to Ge atoms rearrangement before crystallization, delayed at higher temperature by the presence of Ge–N bonds. The corresponding Raman spectra are provided in Fig. S1 in the supplementary material.

We present in Fig. 5 the in situ diffraction patterns acquired during heating rate equal to 5  $°$C/min, for alloys with $Sb / Te<1$ at 450 [Fig. 5(a)] and 550  $°$C [Fig. 5(b)], and at 450  $°$C for alloys with $Sb / Te>1$ [Fig. 5(c)]. The patterns reveal phase segregation toward GST and Ge phases in both cases. For $Sb / Te<1$⁠, segregation begins with cubic Ge and cubic GST225 phases [Fig. 5(a)], requiring higher temperatures to achieve the hex order. Indeed, only after annealing at 550  $°$C, we observe a mixture of cubic Ge, cubic GST225, and hex-GST225 phases [Fig. 5(b)]. On the contrary, for $Sb / Te>1$ GGST alloy, a direct crystallization toward cubic Ge and a stable hex phase occurs [Fig. 5(c)].

Figure 6 illustrates the evolution of the integrated intensity for Ge (111) and GST225 (200) in the case of $Sb / Te<1$⁠, and Ge (111) and GST194 (1013) in the case of $Sb / Te>1$⁠. In each alloy, nucleation is observed in both Ge and GST phases, until a given grain size is reached, where growth stops for the Ge phase in favor of the GST phase, which continues to develop its crystallinity. Once the GST phase is well-crystallized, the Ge grains continue to grow at higher temperatures. For alloys with $Sb / Te<1$ without N [Fig. 6(a)], there is a decrease in the integrated intensity of the GST (200) phase at about 420  $°$C, indicating the beginning of the ordering toward the hex phase.

Crystallites’ sizes were calculated using Scherrer’s law for GST225 (200), Ge (111), and GST194 (1013) at 400  $°$C. The results, shown in Table I, indicate larger grains in alloys with $Sb / Te>1$ compared to $Sb / Te<1$⁠, confirming the rapid crystalline growth of GST194 phase. Introducing N atoms limits the diffusivity of Ge atoms as Ge–N bonds form,⁹ allowing an increase in the crystallite size of GST phases in both $Sb / Te<1$ and $Sb / Te>1$ alloys. The evolution of crystallites’ size with temperature is presented in Fig. S4 in the supplementary material.

From integrated intensities calculated during annealing at different heating rates, we can extract the E_a of the nucleation for each crystalline phase using the Kissinger law, similarly to what performed for R vs T measurements (Fig. 7). We observe that increasing the Sb/Te ratio raises the E_a of GST phase nucleation, suggesting that more energy is required to crystallize directly into a hex phase rather than passing through a cubic metastable phase. Additionally, the E_a of the GST194 phase with $Sb / Te>1$ is lower than the one of Ge phase, indicating that the GST phase is easier to nucleate compared to Ge. In contrast, for $Sb / Te<1$ (GST225), the E_a of Ge and GST phases is equivalent.

The introduction of N increases E_a for the nucleation of both Ge and GST phases. For alloys with $Sb / Te>1$⁠, the difference in activation energies between the GST194 and Ge phases is more pronounced, likely because Ge–N bonds affect mainly the Ge nucleation.

We demonstrated that in Ge-rich alloys with $Sb / Te>1$ ratio, the phase segregation at high temperature leads to the appearance of a stable hex-GST phase, together with the expected cubic Ge phase, that can lead to a more homogeneous GST crystallization, compared to alloys with $Sb / Te<1$⁠, similarly to what observed in previous studies of Sb-rich alloys.¹⁵ By properly tuning the Sb/Te ratio of as-deposited GGST alloys, we did not observe any Sb segregation together with the hex-GST194 phase (as reported in Fig. 5). Based on the recent results published on Sb-rich SbTe systems,³¹ we can speculate on the formation of a hex phase implying Sb planes instead of van der Waals gaps.^27,28 These results should be certainly confirmed by TEM analyses in the future; however, the hex nature of the segregated phase is well evidenced. The observed hex-GST phase exhibits a rapid crystalline growth, resulting in larger grains (as shown in Table I). This can be advantageous for limiting the SET resistance drift caused by amorphous residuals at the GST grain boundaries. The Sb/Te ratio also influences the crystalline structure of GST, as observed in both Raman spectra (Fig. 3) and XRD results (Fig. 5). For $Sb / Te<1$⁠, a Raman shift toward higher wavenumbers is observed indicating a decrease in Sb–Te bond length as crystallization progresses (i.e., appearance of the hex-GST225 phase). It has been shown that the formation of van der Waals gaps in hexagonal phases located on the Sb₂Te₃–GeTe tie line is the consequence of the ordering of the high number of vacancies present in their cubic phase,³² with a subsequent decrease in the Sb–Te bond length (i.e., compact structure) as deduced for The hex-GST225 phase from Fig. 3. In alloys with $Sb / Te>1$⁠, the direct transition toward a hex phase, likely means that the system does not undergo vacancies reordering, leading to the absence of van der Waals gaps in the hex structure. This results in a less compact structure with longer Sb–Te bonds, as confirmed by the lower frequency of the Sb–Te vibrations in Raman spectra (Fig. 3).

There is a correspondence between the E_a trends extracted from R vs T measurements (Fig. 2) and in situ XRD (Fig. 7). In both cases, E_a for GGST alloys with $Sb / Te>1$ is higher than that of alloys with $Sb / Te<1$⁠, indicating that more energy is required to directly achieve the stable hex phase. The difference in the extracted absolute E_a values, arises from the different aspects of the crystallization process they represent. E_a extracted from resistivity measurements reflects the energy required when the crystallization process is well advanced. In contrast, the E_a obtained from XRD analyses corresponds to the nucleation energy (energy barrier to be overcome for the appearing of the first nuclei). This distinction is important as it underscores that while both measurements indicate higher energy requirements for alloys with $Sb / Te>1$⁠, the absolute values are not directly comparable due to their different stages in the crystallization process. Given the limitations of the Kissinger analysis in systems with phase separation and the need for cautiousness in interpreting E_a values, we acknowledge more the trends observed comparing the different systems than the absolute values. Moreover, in situ XRD allows to determine the E_a of the nucleation for each crystalline phase. For alloys with $Sb / Te<1$⁠, the E_a of the nucleation is the same for both GST225 and Ge phases. On the contrary, for $Sb / Te>1$ alloys, hex-GST194 shows a lower nucleation energy barrier than Ge. This explains the more homogeneous crystallization of such a hex phase (sharper XRD peaks in Fig. 5).

The segregation and crystallization kinetics of GGST materials leading to the formation of the previously described phases i.e., cubic GST225 (or close) or hexagonal GST194 (or close) and cubic Ge, which depend on the Sb/Te ratio introduced at the deposition, are schematically described in the ternary phase diagram represented in Fig. 8. The Sb/Te ratio drives the layer morphology after crystallization: $Sb / Te<1$ GGST alloys lead to small crystallites sizes based on a cubic GST phase, while in GGST alloys with $Sb / Te>1$⁠, we observe larger grains (see Table I), highlighting the impact of a rapid crystalline growth in these materials. Based on previous results on Ge-rich GST systems, we think that the Ge content should play a role more in retarding the crystallization process than in defining the segregated GST phase.

As reported by previous studies,^9,11,29,30 nitrogen introduced in GGST alloys primarily bonds with Ge atoms delays the crystallization to higher temperatures and, therefore, enhances the thermal stability of the amorphous phase. This translates into the observation of a higher E_a compared to GGST without N. This stability is preserved even though the Sb/Te ratio is increased, while also retaining the benefits of these alloys, such as fast crystalline growth. Moreover, the difference between the E_a of the nucleation for the hex-GST194 phase and Ge (with $Sb / Te>1$⁠) becomes more pronounced, confirming that N limits the diffusivity of Ge atoms by forming Ge–N bonds and increases the energy necessary to its crystallization. On the contrary, the GST hex phase presents a reduced energy barrier to the nucleation.

In summary, we studied Ge-rich GeSbTe (GGST) alloys with different Sb/Te ratios with and without the N content. We demonstrated that GGST with high Sb/Te ratio promotes the segregation at high temperature of a GST phase featuring a crystalline structure that belongs to the hexagonal crystal family, resulting in larger and more uniform grains with respect to GGST alloys with low Sb/Te ratio. The latter is highly suitable to reduce the possible amorphous residuals at the grain boundaries, which can enhance the drift phenomena in PCM devices. We showed that a high Sb/Te ratio leads to an increase in the activation energy of the nucleation of the GST phase that is beneficial for a high thermal stability of the amorphous phase. Moreover, Ge nucleation requires a higher energy with respect to the GST phase in $Sb / Te>1$ GGST alloys, allowing a favored crystallization of the GST hex phase, responsible for the programming operations in PCM devices. Finally, the benefits on thermal stability from N introduction remain unchanged in GGST with $Sb / Te>1$ alloys, delaying the crystallization of the layer at higher temperature. Sb/Te tuning in GGST alloys represents an important knob to control the structure and the stoichiometry of the segregated GST phase, responsible for the performance of PCM devices targeting applications with strict temperature stability requirements.

See the supplementary material for Raman spectra (Fig. S1 and Fig. S2), ex situ XRD patterns (Fig. S3), and the evolution of crystallites’ size with temperature (Fig. S4).

This work was partially supported by the European Commission and French State through ECSEL-IA 101007321 project StorAIge and partially supported by the French Public Authorities within the frame of France 2030 as part of the IPCEI Microelectronics and Connectivity. Part of this work, carried out on the Platform for Nanocharacterisation (PFNC), was supported by the “Recherche Technologique de Base” and “France 2030 - ANR-22-PEEL-0014” programs of the French National Research Agency (ANR). The experiments at BM2-D2AM beamline at the European Synchrotron Radiation Facility (ESRF) benefited from the beam time allocations N $°$ 20220680 and N $°$ 20230838.

The authors have no conflicts to disclose.

O. Daoudi: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (equal); Methodology (equal); Software (equal); Visualization (equal); Writing – original draft (lead); Writing – review & editing (equal). E. Nolot: Conceptualization (lead); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Y. Mazel: Data curation (equal); Formal analysis (equal); Investigation (equal); Resources (equal); Software (equal); Writing – review & editing (supporting). M. Dupraz: Data curation (equal); Formal analysis (equal); Investigation (equal); Resources (equal); Software (equal). H. Roussel: Investigation (equal); Resources (supporting); Writing – review & editing (supporting). F. Fillot: Resources (supporting). V.-H. Le: Resources (supporting). M. Dartois: Resources (equal). M. Tessaire: Resources (equal). H. Renevier: Conceptualization (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). G. Navarro: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (lead); Supervision (equal); Validation (lead); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

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