VOCl 2 ( CH 2 ( COOEt ) 2 ) } 4 ] as a molecular precursor for thermochromic monoclinic VO 2 thin films and nanoparticles †

The synthesis of thermochromic monoclinic vanadium(IV) oxide (VO2 (M)) thin films and vanadium oxide nanocrystals from a molecular precursor, [{VOCl2(CH2(COOEt)2)}4] is described. Thin films were synthesised using aerosol assisted chemical vapour deposition (AACVD) onto glass substrates at high temperatures and were subsequently characterised and tested for thermochromic efficiency. The film’s suitability as smart, energy efficient window coatings was investigated by calculating their solar modulation potential. Thin films were also doped with tungsten to lower their metal to semiconductor transition temperature (MST) through the addition of tungsten(VI) phenoxide during the AACVD process, lowering the MST from a typical B70 1C for an undoped VO2 (M) thin film to B50 1C for a typical W-doped example. The optimum deposition temperature of 550 1C produced films with thermochromic properties (solar modulation of 15.9%) comparable to the highest values reported. In addition, vanadium oxide nanostructures were synthesised using the thermal decomposition of [{VOCl2(CH2(COOEt)2)}4] and their shape controllably tailored by varying surfactant concentration and type. Thin films and nanostructures were characterised using X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS) and energy-dispersive X-ray spectroscopy (EDS). Thermochromic measurements were measured using UV-Vis spectroscopy with a variable temperature stage.


Introduction
The synthesis of monoclinic vanadium dioxide thin films and nanocrystals for energy saving, infrared radiation reflecting windows is currently an area of intense research due in no small part to the ever-increasing pressure placed on world energy feedstocks. [1][2][3][4] Therefore the need to reduce energy consumption or switch to sustainable energy sources have never been greater. Buildings contribute approximately 1/3 of greenhouse gas emissions, with the majority of these being due to heating and cooling loads to maintain a narrow range of temperatures. 5 The employment of smart, energy saving materials such as thermochromic VO 2 window coatings is one such initiative to moderate this vast energy consumption.
The monoclinic phase vanadium(IV) oxide (VO 2 (M)) undergoes a structural transition to the rutile phase (VO 2 (R)) at a metal to semiconductor transition temperature (MST) of ca. 68 1C, a phenomenon known as thermochromism. 6,7 VO 2 (M) is of particular interest as its MST is relatively close to that of room temperature, unlike other thermochromic V x O y species, such as V 2 O 3 . 8 Below the MST, VO 2 (M) is a semiconductor that transmits infrared and ultraviolet radiation consistently, whereas above the MST, VO 2 (R) exhibits metallic conduction, reflecting in the same region. 1,9 This property is crucial in the design of smart window coatings, that are transmissive of near IR wavelengths at low temperatures, but reflective at high temperatures, thus alleviating the use of expensive and energy-intensive air conditioning units in hot countries. 10 In order to decrease the MST, it is common to incorporate dopant atoms into the VO 2 matrix. Tungsten is by far the most common and most effective dopant in lowering the MST, and has the added advantage of giving the VO 2 a more aesthetically acceptable blue tint rather than the dirty yellow of VO 2 (M). 11 Other dopants such as: niobium, 12 fluorine, 13 magnesium 14 and molybdenum 15 have also been shown to reduce the MST.
The tungsten dopant in W-VO 2 has tetragonal symmetry, which distorts the monoclinic VO 2 towards the tetragonal form. Atoms with large atomic radii such as tungsten introduces distortions into the VO 2 lattice around the W centre, 16 as well as introducing areas of greater electron density into the antibonding orbitals of the V-V bond, and so destabilising the monoclinic phase. 17 These factors are thought to lower the energy barrier to the MST.
The synthesis of VO 2 (M) is exacting due to the large number of stable oxide polymorphs of vanadium. 8,18 Vanadium oxides will easily form as both phase pure oxides (V 2 O 3 , VO 2 and V 2 O 5 ) and mixed oxidation states (V 2 O 7 , V 3 O 7 etc.) and thus isolating phase pure VO 2 (M) requires precise control of the partial pressure of oxygen in the process. 8 Thin films of VO 2 have been very well studied, with highly effective thermochromic VO 2 (M) and doped VO 2 (M) films deposited on a variety of substrates. [19][20][21] The paucity of low molecular weight, volatile precursors containing vanadium(IV) has led to the use of airsensitive VCl 4 22,23 and stable VO(acac) 2 24 precursors. It is also noteworthy that VO 2 (M) thin films have been deposited using VOCl 3 , a vanadium(V) source. 23,25,26 Thin films of VO 2 (M) have been deposited using both atmospheric pressure chemical vapour deposition (APCVD) [27][28][29][30] and aerosol assisted chemical vapour deposition (AACVD) methods, 11,31,32 both of which are highly effective methods for coating large substrates with thin films of VO 2 . The ability to tailor the vanadium precursor could lead to improved properties in the deposited films. This could be achieved by the synthesis of a single molecular precursor, which would have benefits such as the ability to precisely control dopant levels to achieve a given reduction in the phase transition temperature. To this end, AACVD is the preferred method for the CVD of single molecular precursors, due to the typical low volatility of such species. [33][34][35] The majority of synthetic methods for the synthesis of nanoparticles and nanostructures have focussed on high temperature methods. [36][37][38] Vanadium oxide nanoparticles have been synthesised using a variety of methods, including: pyrolysis, 39 sol-gel, 40 hydrothermal methods 41 and ball milling 42 for use in coatings. Hydrothermal (autoclaving) and sol-gel methods tend to produce large, highly crystalline VO 2 (M) nanostructures, from vanadium precursors in the presence of reducing agents such as acids to prevent over-oxidation to V 2 O 5 . 37,43,44 Autoclave methods have also shown to be effective in the growth of large nanowires, and remain the most widely used bottom-up method for the synthesis of VO 2 (M) nanocrystals. It is noteworthy that the manipulation of the gaseous environment to contain partial amounts of oxygen in order to prevent the reduction/oxidation of V (IV) species has met with considerable success in aforementioned high temperature syntheses. [45][46][47] However, to date, the use of single molecular precursors for vanadium oxide nanocrystals and thin films has been underexplored. Our work describes the synthesis of a single molecular precursor; dichloro(oxo) vanadium(IV) diethyl malonate [{VOCl 2 -(CH 2 (COOEt) 2 )} 4 ] [1] for VO 2 (Section 1), fabrication of doped and undoped VO 2 (M) thin films using AACVD (Section 2), and anisotropic vanadium oxide nanomaterials whose growth is controlled by surfactants (Section 3). Compound [1] has two key advantages over typical vanadium compounds used for the formation of VO 2 (M) thin films; superior solubility in preferred solvents for AACVD such as hexane, dichloromethane, chloroform, toluene etc. whereas compounds such as vanadyl acetylacetonate are sparingly soluble, and higher reactivity with oxygen due to the presence of volatile halide ligands. The films were doped with tungsten by adding small amounts of tungsten(VI) hexaphenoxide [W(OPh) 6 ] to the precursor solution, and a lowering of the MST observed. Vanadium oxide nanomaterials were synthesised by the thermal decomposition of [1] in the presence of a high boiling point solvent (1-octadecene) and structure directing surfactants (oleic acid and oleylamine). The structure of dichloro(oxo) vanadium(IV) diethyl malonate [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ] [1] was determined by X-ray crystallography, elemental analysis, 1 H and 13 C { 1 H} NMR spectroscopy. Thin films and nanostructures were characterised using X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), X-ray photoelectron spectroscopy (XPS) and energydispersive X-ray spectroscopy (EDS). Thermochromic measurements were measured using UV-Vis spectroscopy with a variable temperature stage.
Instrumentation 1 H and 13 C{ 1 H} NMR spectroscopy was carried out on a Bruker A-600 MHz spectrometer, operating at 295 K and 600.13 MHz ( 1 H). Signals are reported relative to SiMe 4 (d = 0.00 ppm) and the following abbreviations are used s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), b (broad). Scanning electron microscope images were recorded on a Jeol JSM-6301F SEM at an acceleration voltage of 15 kV. Transmission electron microscopy (TEM) images and were obtained using a high resolution TEM Jeol 2100 with a LaB 6 source operating at an acceleration voltage of 200 kV. Images were recorded on a Gatan Orius Chargecoupled device (CCD). Samples were prepared by drop-casting a nanoparticle dispersion in n-hexane onto a 400 mesh gold grid with a thin holey carbon film (Agar Scientific). Energy dispersive X-ray spectra (EDS) were recorded on an Oxford Instruments XMax EDS detector running AZTEC software. X-Ray diffraction (XRD) patterns were recorded using a Bruker D8 Discover X-ray diffractometer using monochromatic Cu K a1 and Cu K a2 radiation of wavelengths 1.54056 and 1.54439 Å respectively, emitted in an intensity ratio of 2 : 1 with a voltage of 40 kV and a current of 40 mA. The incident beam angle was 11 and data was collected between 101 and 661 2y with a step size of 0.051 at 2 s per step. XRD studies on the nanostructures were carried out using a Stoe (Mo) StadiP diffractometer with a Mo X-ray source (Mo tube 50 kV 30 mA), monochromated (Pre-sample Ge (111) monochromator selects K a1 only) and a Dectris Mython 1K silicon strip detector covering 181 2y. Samples were run in transmission mode, with the sample under rotation in the X-ray beam. All diffraction patterns obtained were compared with database (ICSD) standards. X-ray photoelectron spectroscopy was conducted on a Thermo Scientific K-alpha spectrometer with monochromated Al Ka radiation, a dual beam charge compensation system and constant pass energy of 50 eV (spot size 400 mm). Survey scans were collected in the binding energy range 0-1200 eV. High-resolution peaks were used for the principal peaks of V (2p), O (1s), and C (1s). Data was fitted using CASA XPS software, with doublet separations of 7.5 eV for V (2p) and 2.1 eV for W (4f). 49 Thermochromic measurements were recorded using UV/Vis spectroscopy with a variable temperature stage on a Perkin Elmer Lambda 950 UV/Vis/NIR Spectrophotometer in diffuse reflectance mode. Heating of samples in the UV/Vis spectrometer was achieved by an aluminium temperature cell controlled by RS cartridge heaters, Eurotherm temperature controllers and k-type thermocouples. A Labsphere reflectance standard was used as a reference for the UV/Vis measurements.

Synthesis of [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ][1]
Synthesis of [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ] was carried out under nitrogen obtained from BOC Ltd. in anhydrous solvents using standard Schlenk techniques. VOCl 3 (2 ml, 21.1 mmol) was added to 30 ml of anhydrous n-hexane under an atmosphere of nitrogen using a cannula. Diethyl malonate (0.5 ml, 3.3 mmol) was added dropwise and the mixture stirred under nitrogen for 2 h. An excess of VOCl 3 was used to ensure complete malonate complexation. Unreacted VOCl 3 was removed along with the n-hexane via filtration, leaving behind a dark red/black solid precipitate. The precipitate was filtered and washed 3 times with ca. 20 ml hexane and dried in vacuo to give the target compound 1, in good yield (0.96 g, 83%). 1 H and 13 C NMR were carried out on this product under nitrogen. A 5 cm 3 concentrated solution of [1] in dichloromethane was made up and layered with 15 ml of n-hexane. Small crystals formed over approximately a week. 1 13

Deposition of thin films
All films were deposited using AACVD. Compound [1] (0.30 g, 0.25 mmol) was dissolved in anhydrous toluene (20 ml) and transferred to a glass bubbler under nitrogen using a cannula.
The resulting solution was atomized for 5 minutes to ensure complete solvation of the precursor. An aerosol was generated using a piezoelectric humidifier (Ultrasonic Liquids Atomizer LIQUIFOG s ). The depositions were carried out using a carrier gas of 2% O 2 in N 2 at a flow rate of 3 L min À1 . The glass substrate was standard float glass of 4 mm thickness with a 50 nm SiO 2 barrier layer, supplied by Pilkington/NSG. A top plate was suspended 0.5 cm above the substrate to ensure a laminar flow. A stream of air was introduced into the baffle via a second tube, as a means of introducing oxygen to oxidise a precursor. 51 Depositions were carried out at 540 1C, 550 1C and 560 1C, each taking approximately 30 minutes.

Tungsten doping
Tungsten doping was affected by the addition of tungsten(VI) hexaphenoxide [W(OPh) 6 ] into the precursor solution prior to an AACVD deposition. W(OPh) 6 was synthesised using a literature protocol by Cross et al. 52

Synthesis of nanoparticles
Nanoparticle samples were prepared using the thermal decomposition of [1] in the presence of a high boiling point solvent and alkyl surfactants. Briefly, compound [1] (0.5 g, 0.41 mmol) was weighed out into a nitrogen-purged three-necked 250 ml flask with a condensor. At this stage, 1-octadecene (dried over sodium) (20 ml), oleylamine (Table 1), oleic acid (Table 1) and 1,4-tetradecanediol were added. The flask was subjected to 3 vacuum/back fill cycles with nitrogen and heated to the required synthesis temperature (Table 1) at a heating rate of 3.3 1C min À1 under dynamic nitrogen. Once the desired temperature was reached, the temperature was maintained for 1 hour before cooling to room temperature naturally. The reaction remained under nitrogen throughout the cooling period. The nanoparticles were precipitated with ethanol (ca. 100 ml) and centrifuged at 3000 Â g. The precipitate was subjected to three further ethanol washes (3 Â 100 ml)/centrifugation cycles before storage in a vacuum desiccator. Due to the presence of the alkyl surfactants, the nanomaterials exhibited a high level of dispersibility in organic solvents such as chloroform and n-hexane.

Annealing
Nanoparticle samples were annealed in a tube furnace under a constant argon flow of 0.6 L min À1 for 10 hours at 550 1C at a heating rate of 10 1C min À1 and allowed to cool to room temperature naturally under argon flow before analysis.
Results and discussion , was synthesised under an atmosphere of nitrogen as per literature procedure. 53 It is widely reported that the Achilles heel of vanadium(V) precursors for the formation of VO 2 is the 5+ oxidation state (these precursors preferring the formation of V 2 O 5 ) and as such vanadium(IV) precursors were sought for this work. Recently our group reported a synthetic study on the reactivity of vanadium(V) oxytrichloride with a range of ligands in order to isolate and fully characterise a library of novel vanadium precursors. 53 Interestingly, striking ligand dependant structural and chemical variations were exhibited, altering the functionality of these compounds. This recent work described the reaction of vanadium(V) oxytrichloride with the smaller b-diketonate: acetylacetonate (acac) which resulted in the formation of HCl and a vanadium(V) monomeric species. Additionally, reaction with the larger diethyl succinate ligand was reported to proceed with elimination of Cl 2 and reduction of the vanadium centre, which yielded the more reactive (although polymeric) vanadium(IV) species. It was shown that the use of a ligand slightly smaller than the succinate: diethyl malonate, prevented the unwanted polymerisation, whilst still reducing the vanadium(V) oxytrichloride to vanadium(IV) which was desirable, forming the tetrameric precursor: [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ].
Since the reaction of vanadium(V) oxytrichloride with diethyl malonate does not chemically alter the ligand, and instead, it coordinates via both carbonyl oxygen atoms to the vanadium centre, we propose that this coordination then facilitates a bond between V and a Cl atom to break, thus reducing the metal centre. Four of these units then come together, the vanadium(IV) centre being stabilised through the coordinating VQO from another identical unit, resulting in the tetramer shown below (Scheme 1). This mechanism is confirmed through spectroscopic analysis: of particular interest the 1 H NMR which indicates no loss of proton from the diethylmalonate backbone as well as the longer C-O bond lengths reported in the single crystal X-ray structure, 1.492(7) and 1.498(7) Å (compared with 1.398(4) Å and 1.383(4) in the [VOCl 2 (acac)] compound). 53

Film deposition
All films were deposited using AACVD. This differs from traditional CVD insofar as AACVD involves delivery of the precursor to the substrate in the liquid phase at room temperature rather than the deposition of films from high temperature vapours, requiring volatile precursors. In AACVD, the precursor is dissolved in a suitable solvent, from which an aerosol is generated. The precursor-containing aerosol droplets are transported into the reaction vessel via a carrier gas, at which point the high temperature inside the reactor causes the solvent to evaporate bringing the precursor in contact with the substrate. The precursor then undergoes decomposition, nucleation and crystal growth to form a thin film. As the transport of the precursor to the substrate is a liquid phase process, the volatility of the precursor becomes unimportant, the suitability of the precursor instead being dependant on its solubility in a suitable solvent system. This enables a range of new precursor molecules to be employed. Additionally, the cost and complexity of the delivery process is reduced considerably, thus making AACVD suitable for industrial scale-up.
In all depositions, toluene was used as the solvent. The precursor for both the VO 2 film and the tungsten dopant were both dissolved in the same solution and transported to the reactor using a carrier gas consisting of 98% nitrogen and 2% oxygen. It is noteworthy that in order to attain VO 2 (M), a separate oxygen source was required in order to counteract the reducing nature of an almost total azotic atmosphere. Due to the air sensitive nature of [1], a dual stream of air and nitrogen would be highly detrimental as the precursor would oxidise before entering the reaction chamber. Therefore, a separate gas stream has to be introduced into the chamber itself. This is shown in the schematic in Scheme S1 (ESI †).
Thin films of VO 2 were deposited via AACVD of [1] in toluene over a temperature range of 540-560 1C using a carrier gas of 2% O 2 in N 2 at a flow rate of 3 L min À1 in order to encourage oxidation of the precursor. The deposited films were of good coverage and pale yellow in colour, with some darker, metallic patches formed at higher temperatures. It is noteworthy that AACVD of films at 500 1C did not deposit, whilst raising the temperature to 600 1C yielded poorly crystalline films (see Fig. S4 and S5, ESI †). Tungsten was doped into the films over a range of concentrations from (0.4-0.8%) by adding small amounts of tungsten(VI) hexaphenoxide [W(OPh) 6 ] to the precursor solution. [W(OPh) 6 ] was chosen as the tungsten source as it is better suited to lower temperature applications due to its ease of handling in air and good solubility in the chosen AACVD solvent (toluene) when compared to other tungsten sources such and WF 6 and WCl 6 . The molecular mass of [W(OPh) 6 ] as a percentage of [1] present in the precursor solution ranged from 5% to 1%. The W doped films each had greater optical transparency than the undoped films deposited at the same temperature, with colours ranging from pale yellow to azure, becoming more blue with increased concentration. All films were fairly adherent, resisting the ''Scotch Tape'' test. 54,55 SEM images were obtained for films of VO 2 deposited at 550 1C, both with and without the incorporation of tungsten into VO 2 film. The undoped film consisted of highly disordered rod-like parties of width approximately 50 nm and length 300-500 nm (Fig. 1a). Incorporation of tungsten into the film appears to greatly alter the morphology, with the film consisting of densely packed, rounded, vertical protrusions with a width of around 100 nm, typical of doped VO 2 thin films obtained by CVD (Fig. 1b). 11,26,29 XRD patterns were referenced against those of VO 2 (M) (ICSD: 34033), VO 2 (A) (ICSD: 51213), VO 2 (B) (ICSD: 199) V 2 O 5 (ICSD: 15798) and V 2 O 3 (ICSD: 1473). Films deposited in the temperature range 540-560 1C were indexed as VO 2 (M) as seen in Fig. 1c. There were, however, additional peaks in the XRD patterns, which indicated low levels of an impurity phase, such as a Magneli. 56 There was no evidence of either V 2 O 3 or V 2 O 5 in the XRD patterns. Tungsten incorporation induces a slight shift in the XRD pattern (27.81 (011) for undoped VO 2 versus 27.91 (011) for W-doped) and a loss of crystallinity. This, coupled with the lowering of the MST is highly indicative of successful doping. Film thickness, measured by side-on SEM (see Fig. S7, ESI †), gave increasing thicknesses ranging between 1.8 AE 0.2 mm for the VO 2 grown at 540 1C, 2.1 AE 0.6 mm for 550 1C and 4.1 AE 1.6 mm for 560 1C. Tungsten doping at the same temperatures gave thicker films with W-VO 2 grown at 550 1C giving a thickness of 2.8 AE 0.5 mm.
Typical XPS data from VO 2 (M) and W-doped VO 2 (M) thin films are shown in Fig. 2. The V2p spectra are split into two regions which were then fitted with two separate environments as it was clear that more than one oxidation state was present. The visible shoulders of the main V2p 3/2 peak indicate the presence of V (IV) in the films. XPS is highly surface sensitive, and as such a higher proportion of V 2 O 5 was present at the surface than in the film, hence a large V (V) peak. Surface vanadium species will always be oxidised and this is commonly observed for XPS spectra of thin films of VO 2 . 57, 58 The fitted peak positions of 516.5 eV and 516.7 eV for V2p 3/2 in the undoped and doped films respectively are indicative of V (IV) in VO 2 . 49 The second feature at higher binding energy in both spectra (518.1 eV and 518.2 eV respectively) was assigned to V 2 O 5 . 49 The W 4f environment showed the presence of only one oxidation state. This gave a value of 35.28 eV (W 4f 7/2 ) and 37.38 eV (W 4f 5/2 ) which is consistent with W 6+ . This has been previously seen for W-doped VO 2 in both thin film and nanoparticle synthesis. 1,29,59 Variable temperature transmission UV/Vis spectroscopy was carried out on the films deposited from [1] in order to determine whether the films underwent the structural transition associated with VO 2 (M). A spectrum was taken of the film at room temperature. Following this the films were rapidly heated to 90 1C and the measurement repeated. The films were then cooled to room temperature and the initial measurement repeated to ensure that the sample has remained in the same position throughout the heating (Fig. 3).
Films deposited at 550 1C showed the highest MST by a considerable margin, with the transmittance at 2500 nm falling from 60% to around 2% when the sample was heated from room temperature to 90 1C. Films deposited at 560 1C demonstrated similar characteristics at room temperature to those deposited at 550 1C. Testing the samples for thermochromic activity revealed that the monoclinic phase of the VO 2 is far less pure, roughly halving the transmission to 30%. The VO 2 film deposited at 540 1C was by far the least effective, seeing a reduction in transmission from 50% to 40% at 2500 nm. These measurements reveal that the optimum temperature for the formation of monoclinic VO 2 from [1] was 550 1C. All undoped films displayed a phase transition temperature of ca. 68 1C, which is typical of undoped VO 2 (M). 60 Tungsten doped films exhibited a reduction in MST, to ca. 50 1C. This was determined from hysteresis data obtained via UV/Vis spectroscopy, giving a narrow temperature window for the transition of ca. 5 1C (see Fig. S6, ESI †).
Solar modulation calculations (DT sol ) were performed as described by Taylor et al. 61 Solar modulation values take into account the effect that gases, such as CO 2 and H 2 O, have on the intensity of solar radiation at ground level. This gives a value which can be used to give a better representation of the absolute change in the optical properties of the material. The majority of the energy in the solar spectrum, at ground level, is accounted for by the UV/Visible region (380-780 nm); this leaves a maximum change of ca. 20% for the near IR region (780-2500 nm). VO 2 (M) shows negligible changes in visible wavelength absorption/transmission when it passes through the MST. When comparing the solar modulation values for the VO 2 films, it can be seen that the sample deposited at 550 1C shows a solar modulation of 15.9% which is close to this maximum allowed solar modulation for single layer VO 2 (M) and is comparable to the best results seen for VO 2 thin films by other research groups. 60

Nanostructure synthesis and characterisation
Nanostructures were also synthesised from [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ] [1] using thermal decomposition in the presence of a high boiling point solvent and structure-directing surfactants under an inert atmosphere. Thermal decomposition of metal-organic precursors has been shown to be extremely effective at producing nanoparticles with low polydispersity and shape anisotropy which are subsequently dispersible in a wide range of [organic] solvents. [62][63][64][65] A large number of the metal oxide nanoparticle syntheses are based on the in or ex situ formation of metal-fatty acid complexes and their subsequent decomposition to yield the metal oxide nanoparticles forested by surfactants (fatty acids) rendering them readily dispersible in organic solvents. [66][67][68][69][70] Examples using this methodology for VO 2 nanocrystals are few and far between, however it is noteworthy that Paik et al. utilised a similar methodology to synthesise VO x nanocrystals from VOCl 3 in the presence of oleylamine and octadecanol, resulting in colloidally stable, monodisperse mixed-phase VO x nanocrystals. 71 The VO x nanocrystals obtained were flashannealed at 500 1C for 5 minutes in air at 1 mTorr in order to give thermochromic VO 2 (M), as the formation of phase-pure VO 2 (M) at low (o450 1C) temperature and atmospheric pressure is thermodynamically unfavourable. 18 The authors make reference to the fact that no reaction occurs from VOCl 3 without 1-octadecanol as vanadium-oxygen bonds are formed from the non-hydrolitic reaction between a primary alcohol and a metal halide (sic 1-octadecanol and VOCl 3 ). [71][72][73] Although the synthesis described herein is conspicuous by the absence of a primary alcohol, the composition of the precursor with co-ordinated oxygen species is the de facto product of the non-hydrolitic reaction described above. This, coupled with the high temperature of the synthesis, allows the nucleation and growth of VO x nanoparticles to proceed.
The investigation on the decomposition of [1] in the presence of surfactants in a high boiling point solvent was designed to investigate the effect of differing surfactant ratios, temperature and the inclusion of a fatty alcohol, 1,4-tetradecanediol. The sample preparation conditions are summarised in Table 1. All reactions were undertaken in 20 ml of dry 1-octadecene (dried over sodium) and after 3 vacuum/re-fill cycles of nitrogen. The reaction was heated from room temperature to the target temperature at 3.3 1C min À1 under a dynamic flow of nitrogen, during which the reaction mixture darkened in colour from green to a final black. The addition of oleylamine also induced a colour change of the precursor to blue, presumably due to complexation with the vanadium centre, most likely a dd transition due to the formation of an octahedral complex. On precipitation with ethanol and centrifugation all samples (bar sample 7) gave a black precipitate which could be readily dispersed in organic solvents.
From TEM analysis of the different samples, it was clear that different surfactant blends affected the morphology of the final product. Oleylamine in particular promoted the growth of spine-like structures as well as nanoparticles which were not seen in its absence, whereas oleic acid purely promoted particle growth. Mixtures of oleic acid and oleylamine were found to be the most effective at generating readily dispersible products, with the optimum blend of equimolar amounts of oleic acid and oleylamine [used by Sun and Zeng in their seminal paper on iron oxide nanoparticle synthesis] 74 in samples 1, 6, 7 and 8. The ''6,6'' blend produced a mixture of some small particles but predominantly large, spine-like structures of 50-100 nm in length. These features were preserved on annealing in nitrogen for 8 hours, but were not as pronounced in sample 6, in which the synthesis temperature is lower (280 1C), leading us to deduce that the formation of the vanadium oxide spines is temperature as well as oleylamine dependent. However, the features emerged with annealing treatment at 550 1C under nitrogen for 10 hours (Fig. 4d and e). Sample 7, in which the synthesis temperature was reduced to 240 1C was evidently lower than the nucleation temperature for this system as no nanoparticles were produced. The addition of an oxygen source, 1,4-tetradecanediol in sample 9 with the ''6,6'' blend of oleic acid and oleylamine had no effect on the phase and seemed to the 1,4-tetradecanediol counteracted the structuredirecting effects of oleylamine to form almost exclusively particulate material (Fig. S2d, ESI †).
Detailed TEM analysis of the as-synthesised VO x nanostructures (especially lattice plane images) were exacting due to the presence of high carbon contamination, predominantly caused by surfactant layers and decomposition products from the precursor itself. The latter was in evidence through the presence of chlorine by EDS as residue from initial VOCl 3 ( Fig. 4c and f). Samples 1 and 6 were both annealed at 550 1C under argon in order to increase crystallinity. Analysis of lattice planes in the annealed samples (Fig. 4e) gave a d-spacing of 0.326 nm corresponding to the h111i plane of vanadium oxide (V 4 O 9 , ICSD 15041). Analysis of lattice planes in Fig. S3 (ESI †) gave a d-spacing of 0.327 nm, again corresponding to the h111i plane of V 4 O 9 , confirming that heat treatment had an identical effect on both samples despite initial differences with synthesis temperatures.
EDS spectra showed the persistence of chlorine in the system as a consequence of the chlorine in the precursor. It also showed  Table S1 (ESI †).
Powder X-ray diffraction (PXRD) was performed on the as synthesised and the post-annealed nanoparticle samples as shown in Fig. 5. The as-synthesised nanoparticles, Fig. 5a and b,  have only a few diffraction peaks-it was not possible to confirm the phase of the material from these. The broad reflections between 5-121 2y also suggest that there was a large amorphous component to the samples.
The post-annealed samples, Fig. 5c and d, have a greater number of diffraction peaks in the data. The intensities of these diffraction peaks are also greater, with a reduction in the number of broad diffraction peaks-suggesting higher crystallinity for the annealed samples. Phase analysis of these samples suggested a mix of Magneli phases, with V 3 O 7 -22.21 h020i, V 4 O 9 -11.11 h102i, 16.51 h113i and V 5 O 9 -5.21 h001i, 24.11 h211i and 28.81 h203i lattice planes all being identified in the data. This is not surprising, as vanadium is known to readily form oxides with multiple oxidation states of vanadium present. 75,76 XPS was used as a comparison between annealed and nonannealed samples. XPS spectra for the V2p region in Sample 1 pre (a) and post (b) anneal are shown in Fig. 6. Both show similar profiles to the thin films in Fig. 2, with V (IV) peaks reported at 516.5 eV and 516.3 eV for non-annealed and annealed samples respectively (V2p 3/2 ) and assigned as VO 2 . The V (V) peaks, at 517.7 eV and 517.8 eV respectively were again assigned as V 2 O 5 . 49

Conclusions
The synthesis of [{VOCl 2 (CH 2 (COOEt) 2 )} 4 ] [1] was demonstrated, and its versatility as an AACVD precursor for vanadium oxide thin films and nanocrystals investigated. [1] was shown to be a highly effective precursor for the formation of thermochromic VO 2 (M) thin films on glass by AACVD. The AACVD process has numerous advantages over conventional CVD methods including: the wide range of precursors including those with low volatility, high control over dopant concentration and is applicable to industrial scale-up. 6 VO 2 (M) films produced at 550 1C (DT sol = 15.9%) were shown to have superior solar modulation when compared to films produced at 540 1C and 560 1C and compare favourably with literature values. The thermochromic switching temperature was reduced through the addition of [W(OPh) 6 ] to the precursor solution, reducing the MST to B50 1C. This demonstrates the viability of the AACVD method when using a designed molecular precursor to form thermochromic materials with tunable properties.
In terms of nanomaterial synthesis, [1] was decomposed in the presence of alkyl surfactants at elevated temperatures in an inert atmosphere. It was shown that different surfactants affected the shape of the final product, with oleylamine promoting ''spine-like'' growth, and oleic acid particulate material. The control over shape and size of the resultant nanomaterials by manipulation of surfactant type and concentration in the decomposition of [1] bodes well for the future for designed molecular vanadium oxide nanomaterial precursors. [1] represents the first example in a class of activated vanadium complexes for the synthesis of a variety of VO 2 thin films and shape tunable V x O y nanomaterials. The design of these molecules may lead to a single precursor for the low temperature, one-pot synthesis of colourless thermochromic VO 2 (M) nanocrystals -a key step towards the routine use of smart windows, potentially saving millions of kWh per annum worldwide.