Chemical vapour deposition of antimony chalcogenides with positional and orientational control: precursor design and substrate selectivity

A series of alkylchalcogenostibines, Me2SbSenBu, MeSb(SenBu)2, Sb(SenBu)3 and MeSb(TenBu)2, have been designed and synthesised as potential precursors for chemical vapour deposition (CVD) by reaction of nBuELi (E = Se, Te) with the appropriate halostibine, Me3−nSbCln (n = 1, 2, 3), and characterised by 1H, 13C{1H} and 77Se{1H} or 125Te{1H} NMR spectroscopy as appropriate. MeSb(SenBu)2 and MeSb(TenBu)2 are very effective single source precursors for the low pressure CVD of high quality crystalline thin films of Sb2Se3 and Sb2Te3, respectively, confirmed by scanning electron microscopy, energy dispersive X-ray spectroscopy, Raman spectroscopy and thin film X-ray diffraction. Hall conductivity, carrier mobility, carrier density and, in the case of Sb2Te3, Seebeck coefficient measurements reveal electronic characteristics comparable with Sb2E3 deposited by atomic layer deposition or molecular beam epitaxy, suggesting materials quality and performance suitable for incorporation into electronic device structures. Choice of substrate and deposition conditions were found to significantly affect the morphology and preferred orientation of Sb2Te3 crystallites, enabling deposition of films with either 〈1 1 0〉 or 〈0 0 1〉 alignment. Use of micro-patterned substrates allowed selective deposition of crystalline 2D micro-arrays of Sb2Te3 onto exposed TiN surfaces only.


Introduction
Antimony chalcogenides Sb 2 E 3 (E ¼ Se, Te) are versatile semiconductor materials with narrow bandgaps, which are increasingly demonstrating potential in a diversity of microelectronic applications, especially in low dimensional forms. The most established of these is as thermoelectric materials with moderate gures of merit (ZT), oen combined with Bi 2 E 3 in thermoelectric devices for near room temperature applications. 1 Thermoelectrics have the potential to provide a source of sustainable, emission-free energy or cooling, but advances are required to improve current material efficiencies. Controlling the dimensionality in such materials can lead to signicant improvements in ZT, 2 and superlattices of Sb 2 Te 3 and Bi 2 Te 3 have been claimed to have among the highest ZT values ever recorded. 3 Anisotropy of the thermal and electrical parameters in the layered lattice structures of these materials means that control over crystallite orientation offers another route to maximise ZT. [4][5][6] In addition, Sb 2 Te 3 is a topological insulator, with potential applications in quantum computing. 7 Exploitation of this effect requires single crystals of the binary material; connement to the nanoscale maximises the surface to volume ratio, preventing topological surface states from being dominated by bulk conduction states. 8 Finally, Sb 2 Se 3 has received considerable recent attention as an efficient light sensitiser, which can be incorporated as a thin lm into multilayer photovoltaic devices. 9,10 Chemical vapour deposition (CVD) is a technique widely used industrially to deposit thin lms of a range of materials, including binary semiconductors, due to the scalable and costeffective nature of the process. The potential of employing CVD as an effective route to low dimensional Sb 2 E 3 has been little explored. One very recent approach employs i Pr 3 Sb and Et 2 Te 2 as dual source precursors for the metal-organic (MO) CVD of at lms of Sb 2 Te 3 . 11 Single source precursors (SSPs) are a promising alternative for these systems, as they improve ease of handling and atom efficiency while allowing properties such as stoichiometry and morphology to be directed by precursor design. 12,13 There are very few examples of the use of SSPs for the deposition of Sb 2 E 3 . Individual Sb 2 Te 3 nanoplates and Sb 2 Se 3 nanorods have been synthesised by aerosol assisted (AA) CVD from [Sb{(TeP i Pr 2 ) 2 N} 3 ] and [Sb{SeC 5 H 3 (Me-3)N} 3 ] respectively, the latter being a very rare example of the preparation of Sb 2 Se 3 by CVD. 14,15 Preliminary investigations into the use of the organometallic species Et 2 SbTeEt and Te(SbEt 2 ) 2 as SSPs in AACVD did not yield stoichiometric Sb 2 Te 3 . 16 However, Te(SbEt 2 ) 2 was employed successfully as a SSP in the synthesis of Sb 2 Te 3 nanoplates by thermal decomposition methods, 17 and recent studies have shown that CVD from the same precursor gives poorly aggregated Sb 2 Te 3 lms at 200 C, with antimonyrich lms being obtained at higher temperatures. 18 We have previously developed systems for low pressure (LP) CVD of metal chalcogenide thin lms from coordination complexes with neutral chalcogenoether ligands. [19][20][21][22] In these systems we found that precursors containing -E n Bu groups (E ¼ Se, Te) out-perform those with shorter alkyl or aryl groups at the chalcogen, giving uniform thin lms of high quality. Coordination complexes of heavy p-block metals with chalcogen containing ligands have some potential disadvantages when considered as precursors for LPCVD, namely their relatively low volatilities, resulting in short deposition path lengths and reduced atom efficiency, and the presence of halides or other elements which in some circumstances can act as undesirable contaminants, severely diminishing electronic performance. Organometallic species containing bonds between p-block elements with only hydrocarbon substituents present a potential solution to these challenges.
As a result of these considerations, we identied compounds of the form R 3Àn Sb(E n Bu) n for development as improved precursors for LPCVD. Here we report the synthesis of a new series of these compounds and their employment for the CVD of high quality, uniform, polycrystalline Sb 2 E 3 thin lms. Modications in precursor design allow optimisation of stoichiometry and coverage, while variations in substrate and temperature give remarkable control over lm morphology and crystallite orientation, including fabrication of micro-scale arrays.

Synthesis of tailored single source precursors
The selenostibines, Me 2 SbSe n Bu, MeSb(Se n Bu) 2 and Sb(Se n Bu) 3 , were prepared by generating n BuSeLi in situ 23 and reacting this with the appropriate chlorostibine (Scheme 1).
Molecules of the type R 2 SbSeR 0 (R ¼ Me, Et; R 0 ¼ Me, Ph) have previously been synthesised by the comproportionation of R 2 SbSbR 2 with R 0 SeSeR 0 , 24,25 and one structure (R ¼ R 0 ¼ Me) has been determined by gas phase electron diffraction. 26 Heating of these compounds over a period of hours gives RSb(SeR 0 ) 2 and SbR 3 . 27 The only previously reported species of the form Sb(SeR) 3 3 , which was synthesised by reux of elemental Sb with Me 2 Se 2 in toluene. The crystal structure shows the presence of weak intermolecular Sb/Se contacts which form a 2D sheet. 28 The three new selenostibines with varying Sb : Se ratios (1 : 1, 1 : 2 and 1 : 3) were tested as precursors for the LPCVD of Sb 2 Se 3 under varying conditions. All three precursors deposited grey thin lms onto fused SiO 2 substrates at 500 C, which were observed by scanning electron microscopy (SEM) to be composed of rod-like crystallites. Energy dispersive X-ray spectroscopy (EDX) demonstrated that both Sb and Se were present in all lms, however those deposited from Me 2 SbSe n Bu had a variable Sb : Se ratio which was generally high in Sb (around 1 : 1), whereas those deposited from MeSb(Se n Bu) 2 and Sb(Se n Bu) 3 had Sb : Se ratios close to the desired 1 : 1.5. Of these, MeSb(Se n Bu) 2 gave lms of consistently superior uniformity and coverage, probably due to its higher volatility, and as a result this was the precursor selected for further study (vide infra). These results demonstrate how stoichiometry and morphology of thin lms deposited by CVD can be manipulated by the modication of single source precursor design.
Given the identication of MeSb(Se n Bu) 2 as the most effective precursor for the CVD of Sb 2 Se 3 , the analogous tellurostibine, MeSb(Te n Bu) 2 , was targeted as the precursor of choice for the CVD of Sb 2 Te 3 . It was prepared by the addition of 0.5 mol. equiv. of MeSbCl 2 to a freshly prepared solution of n BuTeLi 29 at À78 C (Scheme 2).
The 1 H NMR spectrum shows a singlet corresponding to the MeSb group, with four multiplets corresponding to the n BuTe groups, in the expected ratios, and a singlet evident in the 125 Te { 1 H} NMR spectrum at 107 ppm. A minor peak at 230 ppm was identied as a small amount of Te n Bu 2 , peaks corresponding to which could also be discerned as minor features in the 1 H NMR spectrum. Attempted purication by distillation resulted in decomposition of the majority of the compound. The crude product was used for all CVD experiments described below, with no obvious detriment arising from the presence of the minor telluroether impurity. Though sensitive to both atmosphere and temperature (an insoluble dark brown substance forms upon air exposure or storage over days at room temperature), MeSb(Te n Bu) 2 is stable over periods of months Scheme 1 Synthesis of alkylselenostibines.
when stored under N 2 at À18 C. Related compounds, RSb(TeR 0 ) 2 , have been prepared previously by oxidative addition of cyclic or polymeric (RSb) n across diorganoditellurides, R 0 TeTeR 0 . 25,30,31 In some cases this reaction was reported to be reversible in solution, though we found no evidence for the formation of n BuTeTe n Bu from solutions of MeSb(Te n Bu) 2 . 30 The compound MeSb(Te n Bu) 2 is somewhat thermochromic, appearing as a bright red oil at room temperature, but reversibly darkening to a black-brown solid below $0 C. Thermochromism has also been reported for related tellurostibines, R 2 SbTeMe (R ¼ Me, Et). 32 Materials deposition and characterisation Sb 2 Te 3 thin lms. MeSb(Te n Bu) 2 is an efficient precursor for the deposition of Sb 2 Te 3 by LPCVD. Thermogravimetric analysis (TGA) under a owing argon atmosphere showed onset of a sudden mass loss (64%) at around 100 C, thought to correspond to evaporation, with no further loss of mass until 600 C where a gradual mass decrease begins, probably corresponding to a slow decomposition at high temperature ( Fig. S1, ESI †). During the CVD process MeSb(Te n Bu) 2 is seen to evaporate cleanly upon heating under vacuum, and thin, silver coloured lms are deposited across substrates heated at temperatures between 300 and 450 C. SEM images demonstrate that these lms are comprised of hexagonal crystallites ( Fig. 1 and S2 ESI †), while EDX conrms that Sb and Te are present ( Fig. S3 ESI †).
The peaks for these elements overlap in the EDX spectrum; quantitative WDX calibrated against elemental standards gives Sb : Te ¼ 41.9 : 58.1, which is in close agreement with the expected value of 40 : 60. No peak corresponding to C was identied in the EDX analysis, and combustion analysis performed on Sb 2 Te 3 powder separated from the substrate gave a C content of 0.74%. When deposition occurs onto fused silica substrates, good coverage is observed over up to 7 tiles (placed in a heated zone between 1 cm and 14 cm from the precursor). In our system, proximity to the precursor and precise substrate temperature are inseparably correlated, with tiles furthest from the precursor occupying the hottest zone of the tube furnace ( Fig. 2).
Films deposited onto fused silica substrates positioned in this hotter zone (450 C) (a) are attest, with SEM images showing the hexagonal faces of many of the Sb 2 Te 3 crystallites lying parallel to the substrate surface (Fig. 1a). In contrast, lms deposited in the cooler zone (around 300 C) closer to the source of precursor (b) are roughest, with SEM showing interpenetrating crystallites standing on end with their hexagonal faces approximately perpendicular to the surface (Fig. 1b). This disparity is observed regardless of quantity of precursor or deposition time.
Symmetric (q-2q) X-ray diffraction (XRD) patterns of these lms show the same single phase of Sb 2 Te 3 (R 3mh), with peak positions that correspond well with literature values (Fig. 3, see ESI † for lattice parameters). 33 However, there are marked differences in the intensities of the peaks for lms (a) and (b) in comparison with one another and with the literature pattern for bulk Sb 2 Te 3 .
Pole gures were collected on key reections for both lms, conrming that the disparities in intensity are due to preferred orientation of the crystallites. The symmetric XRD pattern (Fig. 3a)     the substrate and the hexagonal (a and b) faces of each plate lie at against the surface (calculated values for a in h0 0 1i orientation: 0 0 15 a ¼ 90 ; 1 0 10 a ¼ 50.5 ).
In contrast, the symmetric XRD pattern for lm (b) shows a signicant enhancement of the 1 1 0 reection and a corresponding suppression of the 0 0 l reections, with a pole gure on the 1 1 0 reection (2q ¼ 42.3 ) this time showing a peak at 90 (Fig. 4.3), whereas the ring in the pole gure for the 1 0 10 refection is now seen at around 29 ( Fig. 4.4). This is evidence that the crystallites in lm (b) have h1 1 0i preferred orientation, opposite to that seen in lm (a), with the c-axis parallel to the surface of the substrate (calculated values for a in h1 1 0i orientation: 1 1 0, a ¼ 90 ; 1 0 10, a ¼ 29.5 ), in accord with observations from the SEM images of hexagonal platelets standing on end. The broader peak shapes in the pole gures of lm (b) suggest that the degree of orientation is not as high as indicated by the very sharp peaks in those of lm (a).
In order to provide a direct comparison for patterned substrates (vide infra), lms were also deposited onto substrates manufactured by coating a 1 mm thick layer of silica onto a silicon wafer by physical vapour deposition (henceforth referred to as 'PVD silica' substrates). Under identical deposition conditions these lms had somewhat patchier coverage than those deposited onto fused silica, and SEM images show that they have yet a different morphology, now comprising stacks of thin hexagonal plates (Fig. 5).
The XRD pattern demonstrates the same crystalline phase as those deposited onto fused silica (Fig. S4 ESI †), with an enhancement of the 0 0 l peaks, suggesting that the majority of the crystals have their c-axes oriented at 90 to the substrate, including those substrates deposited in zone (a). It would appear that variation in substrate texture or surface termination can appreciably affect the morphology of these deposits onto substrates which are otherwise chemically almost identical. This ability to control crystallite orientation could be highly advantageous for thermoelectric applications, providing an effective way to optimise anisotropic thermal and electrical parameters.
Cross-sectional images were obtained by snapping the tiles aer deposition. The lm thickness was found to be around 2.2 mm, with layered stacks of crystallites visible side on (Fig. 5).
Raman measurements on these thin lms reveal three main peaks positioned at 120, 138, and 165 cm À1 , respectively (Fig. 6). These are in good agreement with the E 2g , A 2u and A 1g vibration modes of Sb 2 Te 3 previously reported. 34 The electronic and thermal properties of the Sb 2 Te 3 thin lms were also investigated. Hall effect measurements show the deposited Sb 2 Te 3 to be p-type semiconductor with a charge carrier density of p ¼ 4.73 Â 10 19 cm À3 . A charge carrier mobility of m ¼ 138 cm 2 V À1 S À1 was measured, yielding an electrical conductivity of s ¼ 1010 S cm À1 . The Seebeck coefficient of the material (deposited in zone (b)) was measured to be 90 mV K À1 . These values are comparable with Sb 2 Te 3 deposited from other techniques such as molecular beam epitaxy 35 and atomic layer deposition, 36 although lower than the values obtained for perfectly stoichiometric Sb 2 Te 3 deposited at low temperature. 37 These results suggest that the material deposited by CVD is of quality suitable for thermoelectric and microelectronic applications. Sb 2 Te 3 arrays. The substrate dependence of these depositions was further investigated competitively by using patterned substrates manufactured by photolithography, consisting of a silicon wafer, surface coated by PVD with a layer of TiN and  subsequently a layer of silica which is back-etched aer lithography to reveal TiN areas (henceforth referred to as 'patterned substrates'). Using patterned substrates has previously allowed substrate selective CVD of patterned arrays of a number of metal chalcogenides using chalcogenoether coordination complexes of metal halides as single source precursors. [19][20][21] When MeSb(Te n Bu) 2 is used in CVD in conjunction with patterned substrates, highly selective deposition is observed for substrates in zone (a). SEM images show that deposition occurs exclusively onto the exposed TiN areas of the substrates, with the silica surface remaining bare (Fig. 7).
Patterns of squares, circles and trenches with dimensions between 100 mm and 1 mm were lled in this way. EDX mapping of 70-100 mm arrays conrms that Sb and Te are conned to the TiN areas of the pattern, surrounded by uncoated silica surface (Fig. 8). By controlling the quantity of precursor, the thickness of these deposits can be varied from those which protrude signicantly from the surrounding silica (Fig. 7a), to very thin layers of crystals which lie at on the base of the TiN wells (Fig. 7b). This type of substrate selectivity is very unusual for CVD methods, having previously only been observed using coordination complexes of chalcogenoethers as SSPs. The ability to achieve these levels of control in this system highlights the power of using SSPs in an LPCVD framework. Sb 2 Se 3 thin lms. As discussed above, MeSb(Se n Bu) 2 was found to be an effective precursor for the deposition of Sb 2 Se 3 by CVD. SEM images show that lms deposited onto PVD silica substrates are composed of rod-like crystallites (Fig. 9).
EDX gives a mean Se/Sb ratio of 1.46, close to the expected value of 1.50 (Fig. S3 ESI †). Grazing incidence powder X-ray diffraction matches the literature pattern for orthorhombic Sb 2 Se 3 with small variations in peak intensity likely due to a limited degree of preferred orientation (Fig. 10). 38 The Raman spectrum of the Sb 2 Se 3 thin lms was collected (Fig. 11); this is in good agreement with previously reported data. 39 The main feature at about 190 cm À1 is the characteristic of the Sb-Se stretching mode of the SbSe 3/2 -pyramids. 40 The peak at $150 cm À1 can be associated to the Sb-Sb bonds and the vibration at around 125 cm À1 is related to the Se-Se bonds. These three peaks can be assigned as the A 2 1g , A 2 2u , and E 2 g vibration mode, respectively. 41 Hall effect measurements conducted on the Sb 2 Se 3 lms reveal a conductivity of s ¼ 1.14 Â 10 À6 S cm À1 . The low value of conductivity matches well with the wide band gap (1.1 eV) reported for Sb 2 Se 3 . 42 These measurements also conrmed the Sb 2 Se 3 to be a p-type semiconductor with carrier density and carrier mobility of 5.75 Â 10 11 cm À3 and 14.2 cm 2 V À1 S À1 , respectively. However, these lms were found to be quite brittle and poorly adhered to the substrate, and hence reliable measurement of the Seebeck coefficient was not possible. It is clear from cross-sectional SEM images (Fig. 9b) that the lms are composed of loosely packed large ($5 mm long) individual rod-shaped crystallites, leading to a lm of low density and with non-uniform thickness.    Substrate effects were once again probed by the use of patterned substrates. Marked differences were observed in the morphology of Sb 2 Se 3 deposited onto TiN areas of the substrates compared to silica areas (Fig. 12), though substrate selectivity comparable to that seen for Sb 2 Te 3 was not observed. Instead, even under precursor limited conditions, some deposition occurs on both surfaces; on SiO 2 this consists of separated clusters of Sb 2 Se 3 crystallites, whereas on TiN a continuous thin lm of smaller crystallites is observed. This suggests that TiN is a promising material on which to deposit more uniform and well adhered thin lms of Sb 2 Se 3 .

Precursor preparation and characterisation
Reactions were conducted using Schlenk, vacuum line and glove-box techniques under a dry nitrogen atmosphere. The reagents were stored and manipulated using a glove box. Hexane was distilled from Na wire; THF was distilled from Na/ benzophenone ketyl. n BuLi (1.6 M solution in hexanes) was obtained from Acros and used as received. SbCl 3 was obtained from Sigma-Aldrich and sublimed in vacuo prior to use. Me 2 -SbCl and MeSbCl 2 were prepared by treatment of Me 2 SbPh and MeSbPh 2 with HCl, according to the literature method. 43 The 1 H, 13 C{ 1 H}, 77 Se{ 1 H} and 125 Te{ 1 H} NMR spectra were recorded at 298 K in CDCl 3 using a Bruker DPX400 spectrometer and referenced to the residual protio-solvent resonance, neat SeMe 2 and neat TeMe 2 respectively. Microanalyses were undertaken by Medac Ltd and London Metropoliton University. Thermogravimetric analyses (TGA) used a Mettler Toledo TGA/SDTA851e analyser under a ow of Ar at 65 mL min À1 , contained within a dry, N 2 -purged glove box. The temperature was increased at a rate of 10 C min À1 .
Me 2 SbSe n Bu. A suspension of nely ground selenium (0.37 g, 4.8 mmol) in THF (40 mL) was cooled in a bath of liquid N 2 until frozen. n BuLi (3.0 mL of 1.6 M in hexanes, 4.8 mmol) was added, and the mixture allowed to warm to room temperature, resulting in a clear, pale yellow solution. A solution of SbMe 2 Cl (0.9 g, 4.8 mmol) in THF (2 mL) was added causing an immediate intensication of the solution's colour to bright yellow. The reaction mixture was stirred for one hour, then the volatiles removed in vacuo to yield some pale solid and yellow oil, which were extracted with hexane (20 mL). The solution was ltered and the remaining solids washed with hexane (10 mL). The combined organics were concentrated in vacuo to give the product as a yellow oil (0.9 g, 65%). 1  MeSb(Se n Bu) 2 . A suspension of nely ground selenium (0.99 g, 12.8 mmol) in THF (80 mL) was cooled in a bath of liquid N 2 until frozen. n BuLi (8.0 mL of 1.6 M in hexanes, 12.8 mmol) was added, and the mixture allowed to warm to room temperature, resulting in a clear, pale yellow solution. A solution of SbMeCl 2 (1.33 g, 6.4 mmol) in THF (8 mL) was added causing an immediate orange colour which then faded to yellow. The reaction mixture was stirred for one hour, then the volatiles removed in vacuo to yield some pale solid and yellow oil, which were extracted with hexane (15 mL). The solution was ltered and concentrated in vacuo to give the product as a yellow oil (1.5 g, 57%). 1     analytical measurements very challenging. The 77 Se NMR spectrum did not reveal any other Se-containing species, and no other species were evident in the 1 H NMR spectrum.) Sb(Se n Bu) 3 . A suspension of nely ground selenium (2.0 g, 26.3 mmol) in THF (100 mL) was cooled in a bath of liquid N 2 until frozen. n BuLi (16.4 mL of 1.6 M in hexanes, 26.3 mmol) was added, and the mixture allowed to warm to room temperature, resulting in a clear, pale yellow solution. A solution of SbCl 3 (2.0 g, 8.8 mmol) in THF (20 mL) was added causing an immediate dark orange colour which gradually changed to dark yellow. The reaction mixture was stirred for one hour, then the volatiles removed in vacuo to yield some pale solid a yellow oil, which were extracted with hexane (60 mL). The solution was ltered and the remaining solids washed with hexane (10 mL). The combined organics were taken down in vacuo to give the product as a yellow oil (2.8 g, 60%). 1  MeSb(Te n Bu) 2 . A suspension of nely ground tellurium (1.46 g, 11.4 mmol) in THF (80 mL) was cooled in a bath of liquid N 2 until frozen. n BuLi (7.13 mL of 1.6 M solution in hexanes, 11.4 mmol) was added and the mixture allowed to warm to 10 C, resulting in a clear, pale yellow solution, which was cooled again to À78 C. A solution of SbMeCl 2 (1.19 g, 5.7 mmol) in THF (5 mL) was added dropwise, giving a deep orange/red solution which was stirred for 30 min. at À78 C, then allowed to warm to room temperature. The volatiles were removed in vacuo leaving a dark brown paste, which was extracted with hexane (60 mL). The resulting deep orange solution was removed from the dark solid residues by ltration and the volatiles removed in vacuo yielding an orange/red oil (2.1 g, 72%). The NMR spectra showed a small amount of n Bu 2 Te in addition to the product, and was unchanged aer storage under N 2 at À18 C, at which temperature it is a brown solid. The colour change seen between the solid and liquid states is signicant, but fully reversible. 1

LPCVD
In a typical experiment, 5-50 mg of the reagent, followed by the substrate tiles (0.5 Â 8 Â 20 mm), were loaded into a closed-end silica tube in a glove box. The substrates were positioned end-toend through the heated zone. The tube was set in the furnace such that the precursor was 1.5 cm away from the edge of the heated zone. The tube was evacuated ($0.05 mm Hg), and then the furnace set to 500 C (Sb 2 Se 3 ) or 450 C (Sb 2 Te 3 ). Temperature proling demonstrates that there is a temperature gradient along the furnace with the central zone reaching the set temperature and each end being cooler by up to 150 C. Aer 40 min. the tube was cooled to room temperature under vacuum, the substrates subsequently being unloaded and handled in air. Thin lms were deposited onto substrates in all temperature zones, and were silvery grey in appearance.
Thin lm characterisation XRD measurements were carried out using a Rigaku Smartlab diffractometer with a 9 kW Cu-K a source, parallel line focus incident beam and a DTex250 1D detector. Raman scattering spectra of the deposited lms were measured at room temperature on a Renishaw InVia Micro Raman Spectrometer using a helium-neon laser with a wavelength of 633 nm. The incident laser power was adjusted to $1 mW for all samples. SEM was performed using a Zeiss EVO LS 25 with an accelerating voltage of 10 kV, and EDX data were obtained with an Oxford INCAx-act X-ray detector. WDX was obtained with a ThermoFisher Mag-naRay probe using NiC80 and PET X-ray diffracting crystals. High resolution SEM measurements were carried out with a eld emission SEM (Jeol JSM 7500F) at an accelerating voltage of 2 kV. Microanalyses were undertaken by Medac Ltd. Hall effect measurements were performed at room temperature on a Nanometrics HL5500PC with a current of 1 mA under a eld of 0.5 Tesla at 300 K. The Seebeck coefficient was determined using a custom-made Seebeck measurement unit, which was calibrated against a polycrystalline Bi foil reference standard. The measurement accuracy was found to be within 5% and the system was calibrated using copper-constantan thermocouples and a high precision Keithley DMM 2000/E digital multimeter with 0.1% accuracy.

Conclusions
Several new seleno-and tellurostibines have been prepared and screened as single source precursors for LPCVD of Sb 2 E 3 (E ¼ Se, Te); MeSb(E n Bu) 2 were identied to be particularly effective for the deposition of highly crystalline thin lms of p-type Sb 2 E 3 , leading to very good coverage. Electrical characterisation of the lms has revealed their conductivities, carrier densities and mobilities are comparable with those for Sb 2 E 3 lms deposited by MBE or ALD, conrming there are of the quality required for incorporation into functional electronic devices.
Subtle substrate and temperature effects were found to affect the preferred orientation of the crystallites, and were investigated using detailed XRD analysis and SEM imaging. Adjusting these parameters allows control over the orientation of deposited Sb 2 Te 3 crystallites either parallel or perpendicular to the SiO 2 surface. This is an important factor for applications of these materials, for example in thermoelectric devices, where the h1 1 0i alignment is favoured, as it maximises the anisotropy of the thermoelectric parameters, but is oen very difficult to achieve in thin lms.
Positional control of Sb 2 Te 3 growth on the mm scale has also been demonstrated via LPCVD using MeSb(Te n Bu) 2 . Exploiting the subtle substrate discrimination made possible using this single source precursor, highly regular arrays of Sb 2 Te 3 crystallites have been deposited into patterned templates.
Work is ongoing to extend this capability to nano-patterned templates for fabrication of electronic devices.