Differences in \chSb2Te3 growth by pulsed laser and sputter deposition

Sb2Te3, \chBi2Te3, and \chBi2Se3 are typical van der Waals (vdW) layered chalcogenide crystals that can be used in phase storage memory devices, thermoelectrics, and spintronics[1, 2, 3, 4, 5, 6, 7]. \chSb2Te3 exists at one extreme of the \chSb2Te3-GeTe pseudobinary tie line, along which many well-known phase change materials (PCMs) exist that exhibit significant contrast between amorphous and crystalline states, low switching energy, and read-write repeatability[1, 2, 3]. They are also important topological insulators due to a stable Dirac cone at the gamma point in the electronic band structure[4, 5]. These materials are also thermoelectrics with high thermoelectric figures of merit[6, 7]. With so many technological applications, it is important to understand how to grow high quality \chSb2Te3 and the factors that may used to optimise its growth.

The \chSb2Te3 crystal has a R$\bar{3}m$ crystal structure with a unit cell consisting of three stacked quintuple layers (QLs) –Te₁–Sb–Te₂–Sb–Te₁–vdW– along its c-axis[8]. The QLs are bonded by a weak vdW interaction, whilst inter-layer bonding is strong because of the dominant covalent character within each block[9]. The crystal structure is shown in Figure 1(a).

The exotic properties of \chSb2Te3 depend on its crystal structure, thus growing high-quality single crystals or well-textured crystalline thin films is essential[10]. In PCMs, \chGeTe-Sb2Te3 superlattices switch at substantially lower energies than alloys of the same composition. This increase in switching efficiency is due atomic motions being limited to the superlattice layer interfaces, which reduces entropic losses[11]. \chSb2Te3 is often used as a seed layer for vdW epitaxy of phase change superlattices[11, 12], while also being useful for applying biaxial strain to enhance the performance of phase change memory devices[13]. Zhou et al. and Kalikka et al. used the \chGeTe-Sb2Te3 lattice mismatch to further improve the performance of these phase change superlattices. Biaxial stress influences diffusive atomic motions and the activation energy for Ge atoms to switch into the vdW interface[13, 14]. Clearly, for phase change memory applications it is important to grow \chSb2Te3 layers with high quality interfaces with the GeTe.

The thermoelectric properties of \chBi2Te3, \chSb2Te3, and their superlattices have the highest known ZT values in the research literature. The performance of a thermoelectric device is assessed by a figure of merit, the ZT value, where Z is the scale of a material’s thermal properties and T is the absolute temperature. Venkatasubramanian et al. reported a remarkable ZT$\sim$2.4 in a \chBi2Te3-Sb2Te3 superlattice device. Accurately growing ultra-short-period superlattice structures allowed the phonon and electron transport to be controlled and significantly enhance the ZT value[15]. Similarly, a high degree of ordering is effective at improving the thermal power. Tan et al. fabricated and integrated a highly oriented \chSb2Te3 thin film with a layered structure into a thermoelectric power generation device, which exhibited much better performance than randomly oriented or $(0\ 1\ 5)$ oriented \chSb2Te3 films[16, 17, 18].

These R$\bar{3}m$ chalcogenide crystals are also known as topological insulators. Moreover, the same superlattice structures that are applied to phase change memories are also Dirac semimetals[19]. It is, therefore, possible to form topological insulator–normal insulator memories or topological switching devices[20]. However these exotic properties strongly depend on the crystal structure and the number of atomic layers (thickness) of the structure. Hence, the structure must be fabricated with exquisite control. Highly textured \chSb2Te3 thin films were grown to investigate the electrical transport properties as a function of thickness and temperature; atomically flat single crystalline \chSb2Te3 exhibited the topologically insulating behaviour[21, 22]. Thus high quality crystal structures allowed the electronic properties to be engineered with topological states[23, 24].

Crystalline thin film vdW solids can be prepared by a variety of methods[25], such as mechanical exfoliation[26], chemical vapour deposition (CVD)[27], and physical vapour deposition (PVD)[28, 29, 30]. Exfoliation can obtain perfect single crystal layers but it is a very laborious and time-consuming process and can only cover small areas[31]; chemical vapour deposition (CVD) requires precursors and elevated temperatures, which causes impurities and mechanical instabilities due to thermal expansion and stress[27]. PVD methods are preferred because they can be highly automated, are industrially compatible, and allow in situ observation of the crystal growth. Recently, a number of publications have reported PVD growth of \chSb2Te3. Magnetron sputtering is commonly used to deposit phase change superlattice materials[28, 12, 32, 33], and \chSb2Te3-based alloys have been realised in industrial fabrication lines. However, it is hard to control the uniformity and accurate atomistic stacking of crystals. Molecular beam epitaxy (MBE) provides a greater degree of control of the deposited atoms. However, the deposition rate is slow and typically between 0.3 and 0.5 nm/min[29, 34], and it cannot be used to deposit directly from alloy sources. An alternative and somewhat straightforward growth method is pulsed laser deposition (PLD)[30, 35, 36]. A laser is used to ablate the surface of a target, forming a plume of plasma, from which the material is deposited on a substrate. It is capable of providing stoichiometric deposition from alloy targets with high deposition rates. Indeed, PLD has been used to successfully grow a wide gamut of materials, including Transition Metal Dichalcogenides (TMDCs) [35, 30, 36, 37, 38, 39, 40].

Anecdotally, the chalcogenide research community seems to accept that there may be differences in crystal quality for these different deposition methods, but an accurate comparison of their similarities and the factors that influence the crystal growth is currently lacking. Moreover, understanding these dependencies might help to explain why different research groups seem to grow different crystals despite their growth temperatures being similar. For example, there are some reports that the sputter grown \chSb2Te3 is actually Te deficient[32, 33], whilst other groups do not see this effect, and there are even reports that high quality \chSb2Te3 can be grown at temperatures as low as 140 ^∘C while other reports show that temperatures as high as 300 ^∘C are necessary[35, 13]. Boschker and Calarco discussed the differences between sputtering, PLD and MBE for \chGe2Sb2Te5 fabrication, but the origin of these differences is unclear[41]. Here, we quantify the differences in the growth conditions and provide insight into the origin of the conflicting observations in the literature.

It is clear that \chSb2Te3 crystal films are important but it remains unclear whether the deposition factors that influence the crystal quality are the same for different growth methods. Since a large number of factors are involved in the growth process, it is useful to employ statistical designs to find the factors that have a significant impact on the crystal quality. One-factor-at-a-time (OFAT) method is typically applied to study the effect of one experimental variable, but this is inefficient, time-consuming, and laborious, especially when there are more than 4 factors. In contrast, factorial design is a systematic method to determine the relationship between factors affecting a process and the output of that process. It has been applied in fields as diverse as industrial product design, materials development, chemical and medical research[42, 43]. If one is willing to decrease the sensitivity to interactions of multiple factors, then the number of measurements can be reduced to a half, quarter, or one-eighth by exploiting fractional factorial design (FFD) methodology.

In this paper, we present a systematic and quantitative study on the growth of highly $(0\ 0\ l)$ oriented \chSb2Te3 with vdW epitaxy. Herein, we use factorial designs and analysis of variance (ANOVA) to compare the factors that influence the crystal quality of PLD and sputter grown films. Non-adiabatic quantum molecular dynamics (NAQMD) simulations are also performed to study the effect of the PLD plasma on the atomic diffusivity of \chSb2Te3 at the temperatures typically used to grow \chSb2Te3 vdW growth experiments. We also used ground state density functional molecular dynamics (DF/MD) to illustrate and compare how ionised Sb and Te atoms crystallise with and without a seed layer. Our findings shine new light on the vdW epitaxy growth dependence on deposition methods, thus opening a way to grow high quality \chSb2Te3 vdWs crystals, its superlattices, and related materials such as TMDCs.

Sb2Te3 thin films were grown on silicon nitride TEM window grids (substrates) from a \chSb2Te3 alloy target by PLD. The composition measurements on different regions of different samples showed an average composition in at.$\%$ of $39.6\pm{1.6}$ Sb and $60.4\pm{2.0}$ Te, which indicates good stoichiometric transfer. This implies that the composition is relatively insensitive to the PLD growth temperature between 180 ^∘C to 240 ^∘C.

Qualitatively, we found that depositing a 3 nm thick \chSb2Te3 seed layer increases the propensity for the PLD grown films to have a $(0\ 0\ l)$ crystallographic orientation. To test this we split the 16 XRD diffraction patterns that were used in the DoE study (Table 1) into two groups: those with a seed layer and those without a seed layer. These patterns are presented in Figure 2(a) (with an \chSb2Te3 seed layer) and 2(b) (no seed layer). Layered c-axis \chSb2Te3 crystals were only observed when the seed layer was deposited at room temperature prior to growing the thicker layer at a higher temperature. This is clear from Figure 2(a), which only shows $(0~{}0~{}l)$ peaks. This $(0\ 0\ l)$ orientation is especially strong for growth conditions #13 and #16, where sharp and intense peaks are observed. Electron microscopy analysis also showed that films #13 and #16 were particularly flat and clearly exhibited hexagonally-shaped grains; again indicating a high degree of $(0\ 0\ l)$ texture. In contrast, the group of samples that were grown without a seed layer exhibited broad and low intensity diffraction peaks, as is clear in Figure 2(b). Note, samples #2, #3, #9 and #12 were noncontinuous films with isolated islands, while sample #8 and #15 were continuous but exhibited lower and broader diffraction peaks. Also see Figure S1 in the Supporting Information.

To quantify the improvement in $(0~{}0~{}l)$ orientation and crystal quality, it is important to establish a reference sample against which the other samples can be compared. For this reference we used pressure, temperature, and fluence values midway between the extremes values used in the FFD study. We decided to use a 3 nm thick \chSb2Te3 seed layer and a 6 cm target-substrate distance to ensure that a continuous film was deposited. The reference film was grown on a silicon nitride membrane at room temperature. Subsequently we deposited \chSb2Te3 at a rate of 1.0 nm/min or 0.016 nm/pulse at $210^{\circ}$C. The resultant \chSb2Te3 film thickness was 34.5$\pm$0.5 nm. The growth condition is shown in Table 1 and the corresponding diffraction pattern is labelled R in Figure 2. The reference smaple $(0\ 0\ l)$ XRD peaks are clearly seen and the hexagonal morphology of the grains and degree of orientation can be seen by SEM, TEM and t-EBSD in Figure 3 and Figure 4, respectively. The STEM image also shows a relatively flat film with relatively large domains and clear grain boundaries (Figure 3(a)). The SAED pattern shows rings corresponding to $(h~{}k~{}0)$ planes, which are parallel with electron beam direction (Figure 3(b)). This indicates that the reference film has a layered crystal structure with random orientation in-plane and $(0\ 0\ l)$ texture out-of-plane. Additionally, the in-plane lattice constant, which was calculated from the position of these diffraction rings, is 4.266 Å. This is consistent with published measurements for bulk \chSb2Te3, which is 4.264 Å[8]. The layered structure of the reference sample is shown clearly in the cross-sectional TEM image (Figure 3(c)). The vdW gaps are clearly observed and are spaced 1 nm apart, as shown in the line profile (Figure 3(d)). This is in excellent agreement with the expected spacing between $(0~{}0~{}l)$ layers of a highly oriented \chSb2Te3 crystal structure. Thus the cross-sectional and SAED images both confirm the out-of-plane texture. The inverse pole figures of [0 0 1], [1 0 0] and [0 1 0] directions also show that the crystal grains are well-textured along the c-axis with random in-plane orientation. The EBSD crystal orientation maps for this reference film are shown in Figure 4(a). The centre of the $(0~{}0~{}0~{}1)$ crystallographic plane and the uniform annulus around the edge of the $(1~{}0~{}\bar{1}~{}0)$ pole figure is a signature of strong out-of-plane and random in-plane orientation, as shown in Figure 4(b). This agrees with the XRD and SAED measurements shown in Figure 2 and 3(b). The crystal grains have a narrow distribution around the $(0~{}0~{}0~{}1)$ plane, the misorientation distribution full width at half maximum (FWHM) is just 1.4^∘, as seen in Figure 4(c).

Note, the SAED images for all of the samples listed in Table 1 are shown in Supplementary Figure S2. Darker and more blurred diffraction rings indicate poorer crystal quality.

To quantify the importance of the seed layer relative to other growth factors, its statistical significance needs to be measured. We used ANOVA to assess the significance of the deposition factors on $Q_{(0\ 0\ l)}$, which was calculated using Eq. 1 from the XRD patterns shown in Figure 2 and compared against the reference sample. We found that the seed layer is the most and only significant factor in PLD \chSb2Te3 crystal growth. This is interesting because Behera et al. reported that the buffer type (W & Mo buffer layer), temperature, power and pressure significantly affect the quality of sputtered \chSb2Te3. Moreover, in Behera et al.’s study the influence of an amorphous \chSb2Te3 layer was not tested. Therefore, to directly compare sputtering and PLD, we further performed a temperature and \chSb2Te3 seed layer ANOVA test for sputtered \chSb2Te3 films.

For sputter-grown \chSb2Te3, we found that temperature significantly affects the \chSb2Te3 crystal quality and although the seed layer does have an effect, its statistical significance is lower. This is shown in Figure 5(a), which shows a large quality difference between the two temperature levels and small deviation in crystal-quality for samples prepared at the same temperature (small within-group variation). This dependence on temperature is in striking contrast to PLD, where the seed layer has a far more significant effect on the PLD-grown crystal quality, as shown in Figure 5(b). I.e. the red curve shows that the crystal quality shows a large positive improvement when a seed layer is included, but the effect of temperature (blue) is insignificant. Moreover, across the temperature range studied, a higher crystal quality of \chSb2Te3 is observed at higher temperatures for sputter grown films, but the quality decreases with temperature for the PLD-grown films. We will go on to show that this difference is caused by electronic excitation of the atoms in PLD.

The amorphous \chSb2Te3 seed layer facilitates highly-oriented $(0\ 0\ l)$ texture in both PLD and sputtering, as shown in Figure 2 and  5. Saito et al. discussed the advantages of a seed layer on the sputter depositing previously. We can now see that an amorphous \chSb2Te3 buffer is even more important for PLD grown films. Saito et al. argued that the Te layer terminates the dangling bonds on the substrate surface, which prevents good layered heteroepitaxial growth on it[62, 63]. The passivation effect of the Te-atomic layer was also studied in Hippert et al.’s work. They found that depositing a thin Te buffer layer allowed high quality and out-of-plane oriented \chSb2Te3 films to be grown on a range of substrates, including amorphous silicon, silicon oxide, \chTiN and \chWSi[64]. Hydrogen and Sb were also used to passivate dangling bonds and performed vdW epitaxial growth of \chGeTe-Sb2Te3 superlattices on Si(1 1 1) substrates by MBE[65]. Growing the seed layer at room temperature and then annealing at high temperatures prior to growing thicker \chSb2Te3 provides a highly-oriented template for subsequent epitaxial growth.[12, 66]. This seed layer may also reduce the lattice mismatch with the substrate such that a quasi-homoepitaxial growth is possible.

Figure 5 clearly shows that temperature plays a different role in PLD and sputtering. In equilibrium deposition methods, the adatom diffusivity is controlled by the substrate temperature. The surface diffusion coefficient $D_{s}$ is defined by

where $\upsilon$ is the jump frequency, $a$ the jump distance, $E_{a}$ the activation energy, $k_{B}$ the Boltzmann constant and $T$ the temperature. It is clear from Eq. 3 that increasing the deposition temperature increases the surface diffusivity, and this is important because the adatom must diffuse long distances to find a low energy and stable positions in the crystal structure[40]. We assumed that the sputtered films at high temperatures will form layered crystal structures, especially for films with \chSb2Te3 seed layers, as shown in Figure 6(a). However, the Te sticking coefficient is also related to the temperature and it is impractical to use temperature to control adatom diffusivity.

Tellurium has a high vapour pressure at modertately high temperatures. For example, the sticking coefficient is substantially reduced above 600 K [67], thus films become Te deficient. Therefore 600 K is the highest practical growth temperature for \chSb2Te3 using quasi-equilibrium deposition methods, such as MBE. This low sticking coefficient may also explain some of the non-stoichiometric films reported recently in the literature[32, 33] and the lower growth rates at temperatures above 533 K[35]. Figure 6(b) depicts the case where the substrate temperature is too high, and there is no seed layer. The deposited film is likely to be polycrystalline and Tellurium-light. This effect was observed in our PLD experiments where isolated islands or serpentine connected networks tended to occur at higher temperatures (see Figure S1, Supporting Information), which is likely due to surface desorption.

Low temperature growth is attractive because the Te vapour pressure is lower, and its sticking coefficient is higher, which enables stoichiometric films to be grown. However, lower temperatures imply lower diffusivity of adatoms which can hinder intralayer and interlayer mass transport and lower the crystal quality. This leads to amorphous or misoriented polycrystalline films with very small crystal grains, as shown in Figure 6(c).

Our experiments show that the PLD-grown films are much less sensitive to temperature than sputter-grown films. Moreover, others have shown that high quality PLD-grown \chSb2Te3 vdW layered films are possible at extremely low temperatures, just 413 K, whereas sputtering produces polycrystalline films at this temperature, as shown in Figure 6(c) and (d)[35, 28]. We hypothesise that this means that PLD adatoms have a higher diffusivity at lower temperatures than sputtered adatoms. Both PLD and sputtering are plasma-based deposition methods, but the plasma excitation energies are very different[69]. Moreover, the plasma intensity in PLD is orders of magnitude higher than in sputtering. So the proportion of the ions in an excited state is much higher for PLD than sputtering. These excited state adatoms are likely to diffuse further than ground state atoms because the valence electrons, which are normally used to form bonds, are now lacking[51]. We propose that this explains why the crystal quality of PLD grown films is less sensitive to the substrate temperature and more sensitive to the seed layer.

It is also worth noting that neutral atoms separated by a distance $R$ attract each other via the weak $1/R^{7}$ vdW force. This is likely to be the case for sputtering where the majority of the atoms are not excited. However, in PLD where a larger proportion of the atoms are likely to be in an excited state, the force between atoms should scale as $1/R^{4}$, and this can be large enough to bind atoms into long range structures. In fact, provided atoms are in different excited states this relationship should still remain true[70]. With this force scaling argument in mind, we should expect substantial differences in the structure resulting from sputtering and PLD.

To demonstrate the effect of the excited state on Sb and Te diffusivity during PLD growth, we performed non-equilibrium (excited state) DFT computations. Figure 7(a) shows the mean square displacement (MSD) for different proportions of valence electrons excited. We can see that the diffusivity of both Sb and Te atoms radically increases with the proportion of electrons excited. The resultant radial distribution function (RDF) shows significant smearing when more than 10% of the electrons have been excited. This is indicative of athermal melting at temperatures substantially below the equilibrium melting temperature. We now can affirm that PLD provides a means for atomic diffusivity to be high at the low temperatures where the Te sticking coefficient is also high; thus explaining how PLD can be used to grow high quality and stoichiometric \chSb2Te3 crystals at temperatures as low as 413 K [35]. It also explains our observation that PLD is much less dependent on the growth temperature than sputtering – i.e. the electronic excitation dictates the adatom diffusivity in PLD whereas the diffusivity is determined by the substrate temperature in sputtering. Moreover, these results suggest that excited state deposition methods are generally more likely to produce higher quality crystals when one or more of the adatoms has a low sticking coefficient at relatively low temperatures and hints that PLD is more suitable than MBE for growing similar Tellurium-based crystals.

These excited state computations also allow us to follow the charge transfer between atoms as a function of time. A charge transfer process occurs for all the excitation levels studied (See Figure 7(c)–(f)). Although the excited state lifetime is picosecond duration, long lived cascaded excitations can increase the effective lifetime of the excited state[71]. However, for weak excitation levels (n = 2.6%) Sb remains positively charged and Te remains negatively charged after the transfer. In contrast, at high excitation levels, which we assume occurs in PLD, the charges are redistributed equally resulting in neutral Sb and Te atoms. This takes between 1 and 2 ps and depends on the proportion of electrons excited. We expect that these neutral atoms can move through a neutral \chSb2Te3 network with minimal Coulomb interaction, resulting in a larger diffusivity and MSD, as seen in Figure 7(a) and (b). This enables adatoms to find low energy and stable positions required for high quality crystal growth and explains why PLD-growth depends more strongly on the seed layer rather than the growth temperature.

DF/MD equilibrium simulations were also performed to study how the seed layer and ionisation influence the crystallisation of Sb and Te atoms. The effect of the buffer was studied using a fixed monolayer of Te atoms, whilst the excited state atoms were modelled by removing the 10% of the highest energy valence electrons from the band structure. Ideally, we would like to apply NAQMD to study crystallisation. However, for our model, the NAQMD crystallisation computation time is unfeasibly long. As a compromise, we decided to run long ground state DF/MD calculations with artificially excited electronic levels, and then reconfirm the general trends with shorter NAQMD simulations.

To model sputtered crystal film growth on a \chSb2Te3 seed layer, where the majority of the atoms are not in an excited state, recrystallisation molecular dynamics simulations were run from a melt-quenched disordered structure (see Figure 8(a)-1) with a fixed layer of Te atoms (green area) at a temperature of 800 K, as shown in Figure 8(b). Surprisingly, the recrystallised structure did not exhibit a layered structure after 180 ps, but recrystallised into a cubic layered crystal structure, as seen in Figure 8(a)-2.

To model sputtered-growth on a substrate without a seed layer, a model was used with no fixed atoms. This amorphous structure also recrystallised into a similar cubic layered structure (see Figure S3, Supporting Information). However, the recrystallisation time was slightly shorter and crystallinity marginally higher (see inset in Figure 8(d)). Due to the stochastic nature of crystallisation, this difference in crystallisation time between that with a Te layer and that without a fixed Te layer is unlikely to be significant.

Importantly, when 10% of the atoms were ionised and when a fixed Te atomic layer was included in the model, a highly crystalline and layered structure resulted. These conditions model that of PLD with a \chSb2Te3 seed layer, The structure gradually recrystallised to form the ideal hexagonal \chSb2Te3 layered structure. This was possible despite mixed Sb and Te atomic layers, as shown in Figure 8(a)-3. Figure 8(c) shows the evolution of atomic positions along the c-axis of the hexagonal simulation cell (z-coordinate). The atoms closest to the seed layer tend to stabilise first; this is clear by the lower variation in the atomic positions. After 600 ps most of atomic layers have formed and the variance in the atomic z-coordinate is reduced, thus indicating that the atoms have found a suitable low energy position in the crystal structure. The BOP curves reveal that the excited state structure has a longer nucleation time and crystal growth only proceeds after 100 ps. This is in contrast to the other curves where crystal growth proceeds immediately. This makes sense because ionisation depopulates valence band electrons, so it is more difficult for the atoms to form bonds, thus resulting in a higher diffusivity. So the atoms can wander around on the surface until they find a much lower energy position, which is much more likely to result in a more perfect crystal (see orange line in Figure 8(d)). A similar effect is seen for the MSD in non-equilibrium DFT computations–see Figure 7. Note that the excited structure without the seed layer showing a disordered state did not crystallise during the 750 ps molecular dynamics simulation. Its final structure is shown in Figure 8(a)-4. Generally, both the DF/MD with the 10% highest energy electron removal, and the NAQMD computations qualitatively agree and indicate that electronic excitation increases the diffusivity of Sb and Te. However, we refrain from computing absolute quantities, such as adatom diffusivity, because the electronic excitation level is larger than what we might expect in reality.

It is this diffusivity increase that allows PLD to grow high quality \chSb2Te3 vdW layered films at low temperatures, where the high Te sticking coefficient allows stoichiometric compositions Figure 5 & 8 show that the statistical analysis of the growth experiments, and the DF/MD models show that a substantial improvement in the hexagonal crystal quality of \chSb2Te3 can be achieved when both a seed layer and electronic excitation (PLD) is used to grow the films. This is the case depicted in Figure 6(d). Thus there is a strong interaction between the excited state and the seed layer that influences the crystal quality; a low quality crystalline film results when there is no seed layer for PLD (excited state) grown films but high quality films are possible with no seeds for sputtering (less excited). Moreover, these results provide a reasonable explanation for PLD being more suitable than sputtering for growing vdW layered crystals and their superlattices.

Previous growth studies on \chSb2Te3, as well as other vdW layered materials, tended to generalise the findings of a single deposition method to other PVD methods. However, we have used sputtering, PLD, statistical design of experiments, and non-equilibrium DFT to unequivocally show that the factors influencing the growth of simple binary chalcogenide crystals are very different for different growth methods, and therefore it is not possible to generalise the findings from one growth method to another. We found that the quality of PLD grown films are insensitive to the growth temperature, and instead a seed layer significantly improves the crystal quality. We showed that this is due to the PLD plasma exciting \chSb2Te3 valence electrons, which substantially increases the adatom diffusivity. This higher diffusivity allows crystal growth to occur at the low temperatures, which is necessary for Te atoms to stick to the substrate and achieve stoichiometric transfer. The \chSb2Te3 seed layer provides a template from which the \chSb2Te3 film can grow with a high degree of crystallographic orientation and this seems to be especially important for PLD-grown films. This work highlights important differences between two plasma-based deposition methods that we suspect are applicable to not only tellurium-based chalcogenides but other narrow band gap semiconductors.

The authors thank Dr. Václav Ocelik for the guidance of t-EBSD measurement. The SUTD research was funded by a Singapore MoE Project ”Electric-field induced transitions in chalcogenide monolayers and superlattices”, grant MoE 2017-T2-1-161. The USC research was supported as part of the Computational Materials Science Program funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award Number DE-SC0014607. Ms Jing Ning is grateful for her MoE PhD scholarship. The SUTD authors are grateful for space and facilities provided by the SUTD-MIT International Design Center (IDC).