Microwave-assisted deep eutectic-solvothermal preparation of iron oxide nanoparticles for photoelectrochemical solar water splitting

Here, we present a new microwave-solvothermal method for the preparation of iron oxide nanostructures using deep eutectic solvents as a more sustainable reaction medium. By varying the synthesis temperature and solvent water fraction, the methodology offers control over iron oxide phase, size, and morphology, using efficient, rapid (10 minute) microwave heating. Synthesis with pure DES gives small (<5 nm) superparamagnetic samples of g-Fe2O3 or a-Fe2O3, whereas hydrated DES yielded either nanoshards or large rhombohedral nanoparticles without the superparamagnetic response. Nanostructures were solution-cast onto F : SnO2 films. The photoelectrochemical response of the prepared photoanodes was assessed, with a maximum measured photocurrent response of 0.7 mA cm 2 at 1.23 V vs. RHE. We measured the solvent structure using synchrotron WAXS, demonstrating the differences between the dry and hydrated solvent before and after heat-treatment, and showing that the hydrated solvent is remarkably resilient to extensive degradation.


Introduction
Solution processes, whereby liquid solvents are used to facilitate chemical reactions and processes, dominate chemistry because of the unmatched convenience that they offer. 1 This fact is problematic in the face of increasing levels of environmental regulation that curtail the application of volatile organic solvents, which make up the majority of solvents that are used today. 2 Therefore, there is an increasing drive towards replacing traditional solvent systems with more environmentally-friendly ones. Important developments have included systems such as supercritical uids, 3 greener alternative molecular solvents, 4 and ionic liquids (ILs). 5 Of these alternative solvents, ionic liquids are particularly fascinating because of their intrinsically 'designer' nature that allows them to be optimised and tuned to suit certain processes. 6 Deep Eutectic Solvents (DESs) represent another, newer, class of alternative mixed solvent system, formed upon the complexation of various hydrogen-bonding salts and charge-neutral species, to create a low-melting liquid from the ensuing melting point depression. 7 Originally described as a sub-category of ILs, an increasing bank of evidence suggests that they are more akin to an ionic mixture, 8 with high entropy arising from a relatively disordered nanostructure 9 comprising hundreds of intermolecular bonding interactions with similar strengths. 10 Regardless, this broad denition allows DES to be made from a wide array of components, providing even more ways to tune and tailor the solvent around the matter of interest. 11 As a result, DES offer new and unprecedented opportunities to improve upon the sustainability of important and industrially-relevant chemical processes. 12 Research interest is therefore increasing rapidly in these solvents, 7,13 particularly for the preparation of nanomaterials, 14 where new structure-directing effects are becoming apparent. 15 Iron is the most common element on Earth. Because of their inherent sustainability and combination of interesting physicochemical properties, iron oxides are used for a wide variety of applications. For example, magnetic iron oxide nanoparticles such as Fe 3 O 4 (magnetite) have useful applications in medicine (i.e. magnetic resonance imaging), 16 ferrouids, and biosensors. 17 Haematite (a-Fe 2 O 3 ) nanomaterials have attracted significant research interest as a vector for the sustainable storage of solar energy, 18 much like many other metal oxides, because it can split water into H 2 and O 2 in a photoelectrochemical (PEC) process. [19][20][21][22][23][24][25][26][27][28] This is facilitated by the 2.2 eV bandgap of haematite, allowing it to absorb a meaningful component of solar radiation. However, the PEC performance of haematite has been limited because of various factors; charge transport is oen poor, the photon penetration depth produces carriers far from the active liquid junction, and the carriers that are produced subsequently recombine on the picosecond timescale. 22 Therefore, a signicant level of effort has gone towards nanostructuring haematite lms to improve their PEC performance, 21 and has resulted in structures such as nanowires, 18 nanorods, 29 and extremely active highly fractal structures, 23 all offering various levels of activity. A common theme amongst newlydeveloped nanostructuring techniques for haematite thin lm photoelectrodes is spiralling levels of complexity. Many nanostructures require increasingly exotic reagents, and experimental techniques with expensive instruments such as chemical vapour deposition, 23 or spray pyrolysis and plasma-based methods. 18 Here, we aim to use DESs as an alternative green solvent medium towards the goal of active haematite photoelectrodes. We present a synthesis of nanostructured iron oxides using a highly simple, fast, efficient and low-toxicity solvothermal process based on the biodegradable choline chloride-urea DES and its aqueous mixtures. 30 To do this, we utilise a simple, energy-efficient microwave-solvothermal methodology. 31

DES preparation
The DES components choline chloride (Acros, 99%), and urea (Sigma-Aldrich, $99.5%) were used as provided. The deep eutectic solvent reline (1 choline chloride : 2 urea) was prepared by mixing the two components in a sealed container and heating to 60 C with regular agitation. Aer the formation of a homogeneous phase, the mixture was dried using a lyophiliser for one week, yielding the pure DES with a water content of 100.6 ppm, as assessed by 1 H NMR. Hydrated DESs were subsequently prepared from dry reline by the addition of several mole equivalents of water, giving a choline chloride : urea : water ratio of 1 : 2 : 10, referred to here as reline-10w.

Solvothermal synthesis
In a polytetrauoroethylene (PTFE) autoclave liner (CEM OMNI XP-1500), 5 g of Fe(NO 3 ) 3 $9H 2 O (Acros, $99%) was added to 50 g of the chosen DES composition (either pure reline or reline-10w) to obtain a constant molal concentration of 0.25 mol kg À1 . A standard (solvent only) sample was simultaneously prepared in a separate vessel, using the same quantity of the same DES composition but with no inorganic precursor. The two vessels were sealed, jacketed, and placed into a CEM Corporation MARS-5 microwave reactor equipped with temperature and pressure sensors. Using a maximum of 300 W of microwave power, the vessels were heated, and regulated to a maximum pressure of 5.5 MPa up to the desired temperature (100 C, 150 C or 200 C) for a duration of 10 minutes. Following reaction, the samples were le to cool to below 60 C before isolation. The reacted mixture containing inorganic materials was decanted into cellulose dialysis tubing (Sigma-Aldrich, 99.99% retention, MWCO 12.4 kDa) and dialysed against puri-ed water (Elga, 18.2 MU) to remove ionic impurities. The products were then washed with ethanol to remove organic impurities prior to drying in an oven, yielding a series of ne powders of varying hue, from amber to crimson.
Synchrotron small-and wide-angle X-ray scattering Synchrotron SWAXS measurements of the as-prepared and heat-treated DES were made using the I22 instrument at Diamond Light Source, Harwell, UK. The incoming beam was monochromated at 18 keV, giving effective q-ranges for the two Dectris Pilatus-2M detectors of 0.007 A À1 # q # 0.6 A À1 at smallangle, and 0.4 A À1 # q # 8 A À1 for the wide-angle array. Samples were placed into soda-glass X-ray capillaries with a path-length of 1.5 mm and wall diameter of 10 mm. The capillaries were measured in a water-recirculating brass block sample holder with the temperature regulated to 30 C. The instrumental background was removed and the data were reduced using the standard procedures of the DAWN soware suite. 32

Nanomaterial characterisation
Powder X-ray diffraction of the prepared inorganic samples was carried out using a Bruker D8-ADVANCE instrument equipped with a Bruker VÅNTEC-1 CCD X-ray detector. The monochromated Cu-K a radiation from the X-ray source has a wavelength of 1.5418 A, and is slit-collimated to coincide upon the powder sample under Bragg-Brentano geometry. Diffraction data was collected across a scattering vector range of 1.42 A À1 # q # 5.76 A À1 , corresponding with an angular range of 20 -90 2q in this conguration. Laboratory small-and wide-angle X-ray scattering (SWAXS) measurements of the prepared powders, lms, and colloids were made using an Anton Paar SAXSess instrument, using a slit-collimated beam of Cu-K a radiation and a static phosphorescent image plate as a SWAXS detector, to achieve a q-range of 0.01 A À1 # q # 2.80 A À1 . The desmearing procedure for the slit geometry and subsequent SWAXS data processing was completed using Anton Paar SAXSQuant soware. 1D 1 H NMR experiments were carried out using d 6 -DMSO as a solvent, with data collected using an Agilent ProPulse 500 MHz NMR system, and processed using Bruker TopSpin soware. Preliminary SEM and EDX measurements were made using a JEOL SEM6480LV instrument. FE-SEM images of the prepared photoanodes were collected using a JEOL JSM6301F instrument using 10 nm of conductive chromium coating. TEM imaging of the prepared nanoparticles was done using a JEOL JEM-2100 Plus instrument operating with 200 kV of accelerating potential.
The magnetic properties of the prepared nanoparticles were determined using instrumentation at the Materials Characterisation Lab of the STFC ISIS Pulsed Neutron and Muon Source. Analogous procedures were carried out using either a Quantum Design PPMS Vibrating Sample Magnetometer, or a Quantum Design MPMS-XL SQUID Magnetometer. Samples ($100 mg) were weighed into a gelatin capsule and packed with PTFE tape. The M(T) measurement of the magnetic moment as a function of temperature was carried out by rst cooling the sample under zero-eld conditions. Aer charging the instrument to a eld of 100 oersted, measurements were periodically made as the sample chamber was heated to 300 K at a ramp rate of 10 K min À1 . The magnetic moment of the samples was measured as a function of applied eld at 300 K. The M(H) scan began under zero-eld, before ramping the applied eld to 5000 Oe at a rate of 100 Oe min À1 . Aer settling at the desired eld strength, the hysteresis behaviour was determined by ramping the eld to À5000 Oe and back again to 5000 Oe at the same rate. Finally, the measured magnetic moment data were normalised to absolute units of emu g À1 .

Photoanode fabrication
Photoanodes were prepared by adapting a previously-described solution-processed colloidal method from Sivula et al. 22 A paste was prepared from 100 mg of the desired iron oxide powder and 0.1 mL of a 10% solution of acetylacetone (acac) in octanol. This paste was subsequently diluted by the addition of aliquots of a 10% solution of acac in isopropanol (IPA), until 2.5 mL of the acac/IPA solution had been added. The dispersions were then sonicated using an ultrasonic bath for 10 minutes. At this stage, a small ($200 mL) aliquot of each colloid was sampled in case further (SWAXS) analysis was necessary. 1 mL of a 10% solution of hydroxypropylcellulose in IPA was then added to the colloidal dispersions as a porogen and viscosity-modier, before a nal sonication step of ten minutes. Aluminoborosilicate glass slides (Solaronix) coated with F : SnO 2 (FTO) as a transparent conducting layer were used as a substrate, with a spacer of 40 mm invisible Scotch tape (3M). The nal colloidal iron solution was doctor-bladed onto the substrate and allowed to air-dry. The dried lms were pre-treated to ensure the removal of organics by heating in a tubular furnace up to 400 C, with a temperature ramp rate of 1.5 C min À1 and holding at temperature for 12 hours. Aer allowing to cool, the nal treatment step was performed by placing samples directly into a tubular furnace at 800 C for 20 minutes.

Photoelectrochemical testing
The solar photoelectrochemical water splitting performance of the prepared lms was estimated using in-house photocurrent measurements. The Fe 2 O 3 -coated substrate was connected as the working electrode of a three-electrode cell, with an electrolyte solution of 1 M NaOH, a platinum counter electrode and a Ag/AgCl (3.5 M KCl) reference electrode. A PTFE circular mask (radius 0.3 cm) was placed on the reverse (uncoated) side of the sample, and this slide side was subsequently illuminated. Back illumination was used because the entire surface area of the lm is wetted due to the high porosity, and therefore photogenerated hole sites are able to reach the liquid junction at any point within the lm structure, and an optimal photocurrent value is subsequently obtained. A representative comparison of front and back illumination is given in the ESI. † Linear sweep voltammetry (LSV) measurements were made of the prepared photoanodes, under dark conditions and under simulated sunlight, and nally using a chopped shutter oscillating between light and dark conditions with a periodicity of 0.5 s À1 . The prepared Fe 2 O 3 electrodes were scanned between À300 mV and 800 mV against the Ag/AgCl reference using a scan rate of 20 mV s À1 , and data reported herein are stated against the potential of the reversible hydrogen electrode (RHE). Simulated sunlight was calibrated to standard solar conditions (100 mW cm À2 ) at the sample position, derived from a 300 W Xe lamp and an AM1.5 lter. The electrode stability was assessed by holding the electrode at a constant potential of 0.22 V vs. Ag/ AgCl (1.25 V vs. RHE), cycling between illuminated and dark conditions every 3 minutes for one hour.

Results and discussion
Deep eutectic-solvothermal microwave synthesis As far as was possible, elements of the synthesis described herein were specically designed with the principles of green chemistry in mind. 33 For example, reactive mixtures were prepared using a relatively high constant molal concentration of the iron precursor (0.25 mol kg À1 ), which takes advantage of the high mutual compatibility of many metal salts in DESs, 34 thereby maximising the atom efficiency of the process whilst minimising solvent requirements. The choline chloride-urea DES (reline) was chosen for various positive attributes; it is simple to prepare, it is derived from cheap, abundant and natural precursors, and is biodegradable. 35 Furthermore, we have shown previously that in inorganic syntheses based around metal nitrate precursors containing a highly-charged cation, the reline solvent effectively acts as a supramolecular catalyst for the reaction, with a prestructuring effect bringing the reactants together to enable milder synthesis conditions. 15 A microwave-solvothermal methodology was chosen because of obvious rate and efficiency improvements from microwave heating, which has already been shown to work synergistically with the benets offered by using DESs as reaction media in the synthesis of organic molecules. 36 Anhydrous procedures using reline and FeCl 3 were discarded because HCl was liberated during the strongly exothermic mixing, which formed an intractable and corrosive brown mixture. Hydrated iron (III) nitrate (Fe(NO 3 ) 3 $9H 2 O) was favoured, due to mixing safely, spontaneously and endothermically with the DES, and with the additional benet of increasing tractability relative to the pure DES due to the presence of small hydrogen-bonding molecules (water and nitrate). A further viscosity improvement was achieved by dosing the DES with a known quantity of water to produce a DES-water mixture with hydration level of 10w ($41 wt% H 2 O). 37 This additional water eases the processing of the products, 38 and simultaneously modies the solvent environment, whilst remaining below the hydration level where DESs become aqueous solutions. 30 Although different reaction times were trialled the results presented here are derived solely from the shortest reaction time (10 minutes) in the interests of energy efficiency; although in some cases relatively high (100, 150, and 200 C) temperatures are used for the solvothermal reaction, in every case the power input was limited to a maximum of 300 W of microwave irradiation. Products are puried by dialysis against deionised water, such that a nal ethanol rinse is the only time that a 'traditional' molecular volatile organic solvent is used. Hereaer, inorganic products are described as Fe-x-y, where x refers to the synthesis temperature and y refers to the chosen DES-water molar hydration ratio (w). 15,30 Characterisation of nanomaterials Characterisation of the as-prepared iron oxide nanoparticles using powder X-ray diffraction ( Efforts to characterise the different phases using Raman spectroscopy (shown in the ESI †) were confounded by their tendency to oxidise to a-Fe 2 O 3 under even weak and brief laser illumination. 39 However, as these conditions do not contain a Fe 2+ ion source and are performed under air at a reasonably high temperature, g-Fe 2 O 3 is the most likely product. 17 The XRD data also highlighted differences in the particle sizes between the products synthesised in hydrated and pure DESs, which is evident from the greater Bragg peak height and width for the hydrated syntheses. For the Fe-150-0 product no scattering features are observed, and the sample is essentially X-ray amorphous, due to peak broadening arising from small particle size. Scherrer analysis of the peak FWHM for the dominant (104) and (110) reections of the haematite-type Fe-200-y products shows a mean crystallite size of 15.9 nm for Fe-200-0, as compared to 49.1 nm for the Fe-200-10 product. This suggests a greater degree of crystallite growth in the hydrated case, likely to reect the faster kinetics of the less viscous aqueous mixture.
TEM imaging of the iron oxide nanoparticles showed a variety of nanostructures, with size and morphology determined by the synthesis conditions. Representative images are displayed in Fig. 2. In all cases, the prepared nanoparticles showed a strong tendency to aggregate, which is not surprising since no stabilising species were used. We observe that in the case of the reline-0w synthesis at 150 C, aggregated spongelike composite spheres are formed with an average diameter of 50 nm. Interestingly, these spheres are composed of many extremely small spherical nanoparticles, which have an individual size of 2-3 nm and are relatively homogeneous. Upon adding water and reacting at the same temperature, a completely different 1D nanoshard morphology is formed, rather than the extremely small nanoparticles. These nanostructures are not as homogeneous as the particles prepared in the water-free case, and accurately determining a mean size is difficult because of the bundle formation. However, the nanoshards generally display a width ranging from 10-20 nm, with lengths of 80-100 nm. Interestingly, the lattice planes of the g-Fe 2 O 3 nanoshards can be seen in the TEM imaging without extremely high levels of magnication, and were found universally to run parallel with the major axis of extension. The TEM data therefore suggest that the main reason for the diffuse X-ray scattering of the Fe-150-0 product is small particle size and the accompanying peak broadening rather than any amorphous nature, and we believe that this is the same g-Fe 2 O 3 phase as the Fe-150-10 product, which is clearly larger and evidently crystalline. The nanoshard morphology may be driven by the selective capping of certain crystal planes by the DES components during crystal growth.
Increasing the synthesis temperature, the Fe-200-0 product also yields a system of relatively monodisperse spheroid nanoparticles with a tendency to aggregate, much like Fe-150-0. In this case, the morphology is similar to the Fe-150-0 particles but with slightly larger sub-particles, with an apparent size of around 5 nm. This suggests once again that the diffuse a-Fe 2 O 3 Bragg peaks in the XRD data are indicative not of an amorphous nature, but a small particle size. Increasing the water content for the 200 C synthesis yields again a completely different nanostructure. Very large nanoparticles are formed, displaying a rhombohedral morphology that is evocative of the crystal structure of a-Fe 2 O 3 . The prepared particles are not homogeneous, and range from 100-200 nm in length, with a width of Fig. 1 Powder X-ray diffraction data for the iron oxide nanoparticles prepared using pure reline and hydrated reline. Pseudo-Voigt fits calculated using Fityk software are shown for the haematite-type iron oxide products. 40 Bragg peaks suspected to correspond with the g-Fe 2 O 3 phase are denoted with an asterisk. around 100 nm. These data therefore make it clear that the reaction temperature and reaction water content are two independent variables that both have an impact upon the nanoparticle growth rate. Additionally, there appears to be a subtle structure-directing effect exerted by the DES, because a similar morphological relation of the prepared nanoparticles was observed for a DES-solvothermal preparation of ceria, with smaller, less crystalline materials when the pure DES is used relative to the hydrated system, and 1D nanostructures formed in hydrated DES at low reaction temperatures. 15 This is particularly interesting because ceria has a cubic uorite structure that is entirely different from either the a-Fe 2 O 3 or g-Fe 2 O 3 unit cell. The Fe-200-10 synthesis combines the most extreme reaction conditions of the highest water content and highest temperature, to give the highest in situ autoclave pressure and the most rapid kinetics of reaction and growth. The DESsolvothermal methodology therefore offers tunability of the size, shape, and phase of the prepared iron oxide nanoparticles, and does so whilst being rapid, simple, and environmentally friendly.
The magnetic properties of iron oxide nanoparticles are known to have a strong dependence upon the morphology, phase, and size. This has given rise to a number of important medical applications, such as their usage in bioseparations or as MRI contrast agents. 16 The magnetisation of the prepared materials was therefore measured rstly as a function of temperature under 100 Oe of applied eld, having been cooled under zero-eld conditions (ZFC). Following this, M(H) measurements were made on the samples at 300 K to determine the hysteresis behaviour as the eld was varied from 5000 Oe to À5000 Oe. The results of the M(T) scans are shown in Fig. 3. The observed ZFC magnetisation curves show completely different behaviours that can be related to the nanoparticle phase and size, which are themselves a function of the synthesis temperature and water content. The Fe-150-0 and Fe-200-0 nanoparticles show analogous behaviour, with a sharp rise in magnetisation as the temperature is increased, with a relatively strong peak of 0.16 emu g À1 for Fe-150-0 and 0.23 emu g À1 for Fe-200-0, before the magnetisation falls with increasing temperature. g-Fe 2 O 3 generally displays ferrimagnetic behaviour in the bulk phase, whereas a-Fe 2 O 3 is typically a weak ferromagnet or canted antiferromagnet. 41 This kind of magnetisation response is characteristic of superparamagnetic iron oxide particles, which are formed with hyperne nanoparticles that are below 10 nm, such that each nanoparticle acts as a single-domain paramagnet, not large enough to have multi-domain ordering. 41 This is conrmed by the observations made during TEM experiments. The superparamagnetic blocking temperatures for the prepared Fe-150-0 and Fe-200-0, respectively, are 69 K and 62 K.
Conversely, the nanoparticles synthesised in hydrated DES show an entirely different and far weaker magnetic response. The Fe-150-10 product shows a minor uctuation in the degree of magnetisation as a function of temperature, with two 'ne' transitions occurring at 110 K and 250 K, which is a more typical response for a nanoparticulate maghemite phase. The Fe-200-10 product shows classic a-Fe 2 O 3 characteristics, with minimal magnetic response until a temperature of 250 K, at which point it undergoes the Morin transition to either a canted antiferromagnetic or ferromagnetic state. 17 This is in accordance with the large, highly crystalline nanoparticles that can be observed using TEM. The suspected magnetic phases are conrmed by M(H) measurements at room temperature, which are shown in Fig. 4. The Fe-150-0 and Fe-200-0 products both show a completely straight line with negligible hysteresis, characteristic of a paramagnetic, or superparamagnetic state, in this case. The Fe-200-10 product shows a classic ferromagnetic behaviour, with a weak hysteresis of approximately 0.06 emu g À1 . The Fe-150-10 material shows a weak ferrimagnetic response at room temperature with very minor hysteresis and curvature, and so this sample was also measured at 220 K, below the ne transition observed at 250 K. An increase in the overall magnetic moment and a slightly stronger hysteresis behaviour was observed, consistent with an increase in magnetic ordering occurring as the nanoparticles are cooled. 42 Additionally, these magnetisation measurements offer denitive proof that the Fe-150-y nanoparticles are composed of a g-Fe 2 O 3 phase rather than Fe 3 O 4 , because the measured magnitude of the magnetisation is signicantly lower than would be seen for magnetite. 16 Solvothermal reaction mechanism and solvent degradation structural studies Efforts were made to determine the mechanism of solvothermal reaction. We note that this is not the rst time that DES have been used for the preparation of iron oxides; Chen et al. used a co-precipitation route to prepare magnetic Fe 3 O 4 nanoparticles, 43 and Xiong et al. prepared haematite nanospindles using a precipitation method from heated, dry DES upon the addition of water. 44 Despite the promising properties of the prepared materials, these precipitation methods cannot truly be described as a solvothermal process, but there are obvious similarities in the conditions because they propose the same urea-hydrolytic pathway that has been seen in numerous other preparations. 45 The materials characterisation detailed above provided no evidence that the novel methodology reported here represents a signicant deviation from other solvothermal protocols in terms of the urea hydrolysis and subsequent ironoxide forming mechanisms, which have been previously and exhaustively addressed. 15,[46][47][48] The reaction temperature and water fraction have an important effect upon the urea hydrolysis rate in DESs. The decomposition of DES at elevated temperature was noted previously by Parnham et al. in their studies of DESs as alternative solvent media for the synthesis of hybrid inorganic materials. 46,47 Interestingly, they observed that the DES predominantly plays a templating function, with the controlled degradation of the labile species such as urea and its functionalised analogues delivering structuring agents for the synthesis. In our studies of the more closely related metal oxide ceria, we directly observed a pre-structuring of the reactive components within the reline mixture, effectively decreasing the activation energy for the reaction, whilst also noting hydrolysis and subsequent reaction of urea with solvated cerium ions. 15 Based on the morphological dependence upon synthesis conditions, there may be some similar processes occurring in the iron oxide synthesis. Samples of the pure reline-0w and reline-10w solvents (50 g) without iron precursor were therefore placed into separate vessels and exposed to the microwave heating treatment alongside the reacting samples. The urea decomposition was monitored using 1 H NMR spectroscopy, the results of which are shown in Table 1. It was observed that the pure reline DES did not undergo any signicant urea hydrolysis aer heat treatment at either 100 C or 150 C. The pure DES had to be heated to 200 C before there was any quantiable degradation of urea, with the choline chloride-urea ratio diminished to around 1 : 1.8 aer this treatment. Interestingly, the iron-forming reactions for the pure DES were found to form a water-soluble crust around the autoclave lid. Analysis of this product using 1 H NMR and powder X-ray diffraction suggests it contained a highly crystalline combination of the various likely DES degradation products, such as biuret, cyanuric acid, and ammonium carbonate (see ESI †). For the hydrated DES, signicantly more urea hydrolysis was seen. Treatment at 100 C and 150 C again yielded almost identical results of a 1 : 1.8 choline chlorideurea nal ratio, and in the most extreme circumstances, the 200 C hydrated synthesis resulted in a 1 : 1, distinctly offeutectic mixture of choline chloride-urea following the thermal treatment. Unlike the study of Querejeta-Fernández et al., we do not observe any signicant signal corresponding with the formation of NH 4 OH in the solvent. 48 Generally, when contrasting with previous syntheses, we observe signicantly less solvent degradation, which is almost certainly a product of the rapid reaction times facilitated by the efficiency of microwave heating. This can be visualised by the relative simplicity of the NMR spectra collected aer treatment (see ESI †), which can be contrasted with the extensive degradation observed when a DES is heat-treated in a conventional oven for a week. 46  a It should be noted that the measured hydrolysis in the pure DES may not necessarily be representative of the reacting system, which contains additional low-level water from the iron precursor, and may experience some further effect from the paramagnetic iron content. Errors are stated assuming a standard 5% deviation in the veracity of the NMR integrals.
We subsequently aimed to determine the degradative effects of urea hydrolysis upon the nanostructure of the DES and the hydrated DES. It is unclear what effect that both the loss of urea, and the subsequent integration of the hydrolysis products themselves, have on the overall solvent structure in the case of both the pure and hydrated systems. 49 To understand this, measurements of the primary structure factor S(q) of the solvents were made using synchrotron wide-angle X-ray scattering (WAXS) at beamline I22 of Diamond Light Source, UK. Data were collected before and aer heat-treatment, with background-corrected and normalised scattering patterns shown in Fig. 5. In pure d 17 -choline chloride : urea, there are two primary constructive scattering interferences, giving peaks at 1.45 and 2.15 A À1 , which represent the two most common interaction lengths in the disordered liquid, and respectively describing real-space separations of 4.3 and 2.9 A (d ¼ 2p/q). The measured X-ray structure factor S(q) for reline therefore matches accurately with the scattering form factor and primary correlation lengths observed in the pure DES by wide q-range neutron diffraction. 9 Upon adding water, the data show that the system becomes more disordered, signied here by the additional peak broadening. 50 The intensity of the secondary 'shoulder' peak at 2.15 A À1 is increased noticeably in the hydrated system, signifying the partial loss of the DES-DES (i.e. the choline-choline, choline-chloride, choline-urea, urea-urea and urea-chloride) interactions. 8 Despite this, the position of the main scattering feature (at q ¼ 1.45 A À1 ; d ¼ 4.3 A) remains in the hydrated system. This demonstrates that the hydrated DES still contains a signicant portion of the pure DES hydrogen-bonding nanostructure, 30 and validates the approach of adding water as a processing enhancement option for DES. 37 In the case of the 100 C and 150 C heat treatments, there is very little variation in the structure between both the pure solvent and the hydrated solvent before and aer microwave treatments. The pure DES is conrmed to degrade slightly because of the minor peak broadening observed when the solvent is treated at 150 or 200 C, whereas the reline-10w DES sees some broadening alongside a signicant extension to the satellite correlation at 2.15 A À1 , signifying that the hydrated DES is somewhat more affected by the heat treatment, as was suggested by the NMR analysis. However, in both instances the DES display remarkable nanostructural resilience with regard to shiing to an off-eutectic composition. 8 This is likely to be a product of the short reaction time that is facilitated by the usage of microwave irradiation, the hydrogen-bonding contribution of certain likely degradation product molecules such as isocyanuric acid and biuret, and the hydrogen-bonding contribution from water in the hydrated system. These ndings raise the possibility that the DES could even be recovered and recycled aer such syntheses, further improving the efficiency.
Attempts to further reduce the reaction time and temperature met kinetic limitations; unlike the 150 C or 200 C preparations, syntheses performed for 10 minutes at 100 C had only fractional Fe 2 O 3 yields of 40% (Fe-100-0) or 73% (Fe-100-10). EDX measurements suggested that the prepared Fe-100-0 product had only a surface coating of the desired Fe 2 O 3 , with the particulate bulk composed of crystallised FeCl 3 . Because this salt was not used, this must represent the dominant dynamically-solvated iron species in the reline DES, which is necessarily chloride-rich. 51 In spite of any preferential solventreactant structuring, it seems likely that in this case the kinetic limitation of the pure DESs lies in their relatively high viscosity, which represents a diffusion-limited regime. In the case of the aqueous DES this kinetic limitation is mitigated, as the additional water has the effect of dramatically reducing the solvent viscosity and hence, increasing the solvent self-diffusion coef-cient relative to the pure DES. 49 Therefore, there are clearly some synthetic advantages to be had by using hydrated DESs over the pure form. The optimal conditions for a hydrated deep eutectic-solvothermal reaction can be found by tailoring the DES hydration level to obtain the desired combination of solvent diffusion and pre-structuring effects, whilst remaining below an aqueous regime. 37 Application as photoelectrodes Photoanodes were prepared by adapting a previously-described solution-processed colloidal methodology, rather than developing a DES-based method because of the likely introduction of impurities such as chloride and other organics from the incomplete calcination of the low vapour pressure ionic mixture. 52 In this process, a stable colloid of the iron oxide nanoparticles is prepared, 22 using acetylacetone to stabilise the nanoparticles by acting as a 'capping' hydrotrope within the isopropanol dispersant phase. A structure-directing agent (hydroxypropylcellulose) is then added to ensure the produced electrodes are porous aer the thermal treatment. 21 The lms were cast onto F : SnO 2 -coated aluminoborosilicate glass and heat-treated in two steps, with a preliminary 400 C, 12 hour treatment to remove the organic residues and convert the g-Fe 2 O 3 phase to the photoactive and stable a-Fe 2 O 3 phase. The nal step is a fast 20 minute treatment at 800 C. This latter treatment is unfortunately intensive, but is necessary to produce sufficiently active haematite photoelectrodes without precious metal dopants, because the Sn 4+ diffuses into the iron oxide during sintering, modifying the lattice parameters and resulting in electronic modication. 22 The optical properties were found to vary according to the nanoparticle size, with the most transparent lms prepared from the smallest nanoparticles (Fe-150-0) and the most opaque lms derived from the largest nanoparticles (Fe-200-10). FE-SEM measurements of the prepared photoanodes aer the dual heat treatments yielded a similar morphology to those prepared previously from the related casting method, with iron oxide from the thermal decomposition of Fe(CO) 5 , 22 and these are shown in Fig. 6. These images show the partial sintering and growth of the nanoparticles, which form necked aggregates reaching a diameter of around 50 nm in the Fe-150-0 lm and 100 nm in the Fe-150-10 lm. The larger feature size of the Fe-150-10 lm can be related to the larger size of the nanoshards that are used for the preparation, as compared to the miniscule Fe-150-0 nanoparticles. This is additionally reected in the clearly more tightly-packed Fe-150-0 lm, and the very large feature size of the Fe-200-10 lm. The porous nature of the photoelectrode is conrmed by the imaging, suggesting a high accessible surface area, ideal for the photoelectrochemical splitting of water. 21 The prepared photoanodes were measured under standard solar conditions (100 mW cm À2 at the sample position) in a three-electrode conguration, using a 1 M NaOH electrolyte, platinum counter-electrode and 3.5 M KCl reference electrode. The reverse (uncoated) side of the electrode was found to give the maximum photocurrent response, because despite the complete wetting of the electrode nanostructures by the electrolyte, more photoinduced electrons are generated closer to the FTO substrate than with front illumination, with more electrons then moving to the cathode. Additionally, there is inevitable absorption of light with a corresponding decay in intensity when lms are greater than a threshold thickness. 53 An example of front vs. back illumination performance is given in the ESI. † Linear sweep voltammetry data for the electrodes under both light and dark conditions are shown in Fig. 7, and the calculated values of the photocurrent density at 1.23 V versus the RHE are shown in Table 2. The majority of the prepared systems deliver a photocurrent density competitive with other examples in the literature, which can be related to the properties of the iron oxides used to prepare the respective photoelectrodes. The Fe-150-0 electrode has a photocurrent of 0.53 mA cm À2 at 1.23 V vs. RHE, identical to the value obtained for the Fe-200-0 electrode. This is representative of the very similar nanoparticle size and morphology of the two systems, with the minor differences between the two systems negated aer the high-temperature sintering treatment. The anode derived from the Fe-200-10 rhombohedral nanoparticles gave the weakest measured  photocurrent response of 0.19 mA cm À2 . This can be related to the particularly large and more difficult to coalesce, low-surface area nanoparticles that this anode is derived from. Conversely, the strongest observed photocurrent response was for the Fe-150-10 lm, derived from the nanoshards prepared in the hydrated DES at lower temperature. This anode gave a photocurrent response of 0.7 mA cm À2 at 1.23 V vs. RHE, which is competitive with other literature preparations, and notably higher than photoanodes prepared using the thermal decomposition method from Fe(CO) 5 , despite the possibility of our products passing through a hydroxyl-containing goethite phase due to the synthesis mechanism. 54 The FE-SEM data (Fig. 6) reect this, in the high solvent-accessible surface area of this electrode and the high quantity of elongated, necked arrays that are derived from the sintered shard nanostructures. We note that the average width of the features is commensurate with the photon penetration depth in haematite. 22 This level of photocurrent response does not match extremely high performance benchmarks such as the 2.2 mA cm À2 that can be achieved by chemical vapour deposition of high-surface-area, porous cauliower-type fractal haematite. 23 However, the simple solvothermal process reported here is comparatively rapid, facile, and more environmentally benign, whilst not requiring CVD equipment, or volatile and harmful reagents such as tetraethylorthosilicate (TEOS). 21 Some interesting differences were observed in the photocurrent response between those products isolated from dry and hydrated DESs. Firstly, the different DES resulted in a slight shi in the position of the dark current onset potential. The dark current occurs at approximately 1.6 V for the Fe-x-0 materials, but consistently occurs at a lower potential of 1.55 V for the Fe-x-10 products. Moreover, the Fe-150-0 measurements do not show any signicant transient in the dark current alongside the photocurrent onset potential, whereas there is a slight dark current at lower potentials for the Fe-x-10 lms, likely due to trace contamination. Finally, an interesting feature is noted in the insets of Fig. 7. The prepared samples were also measured using sweeping current-potentiometry scans with a chopped shutter in order to determine the presence of photocurrent transients as the incident light is periodically cycled between on and off. The Fe-x-10 photoanodes show sharper characteristic transient spikes upon cycling, corresponding with a high concentration of photo-generated hole sites being generated within 100 nm of the semiconductor liquid junction (SCLJ). 55 This suggests that the recombination step is likely to be limiting for both sets of materials, 29 which can occur either in the bulk or at grain boundaries. 18 Finally, the stability of the prepared photoanodes was measured in order to demonstrate the resilience of the prepared lms. This was done by measuring at a constant potential, cycling between illuminated and dark with a periodicity of 0.33 min À1 for one hour, and representative data for the Fe-200-y electrodes are shown in Fig. 8. In each case, the fabricated photoanodes show good resistance to the repeated cycling. For the Fe-200-0 thin lm, the average rst photocurrent response of 0.532 AE 0.005 mA cm À2 drops to 0.526 AE 0.004 mA cm À2 aer one hour of this treatment. This corresponds with an activity reduction of 5.3 mA cm À2 , or just 1% of the total activity. Similarly, the Fe-200-10 photoanode response falls from the rst average of 0.114 AE 0.001 mA cm À2 to the nal measurement of 0.108 AE 0.001 mA cm À2 , corresponding with a comparable absolute loss in photocurrent of 5.5 mA cm À2 , but a 5% loss in this case due to the lower total photocurrent density of this anode. Therefore, the prepared electrodes are remarkably stable, with only minimal loss in photocurrent density aer repetitive on-off cycling.

Conclusions
We have demonstrated that the DES reline and its hydrated mixtures can be used as viable sustainable alternative solvents in the manufacturing of green materials for the photoelectrochemical splitting of water to hydrogen. The microwaveassisted solvothermal methodology is rapid, malleable, and represents a signicantly more environmentally-friendly route towards this goal due to lower energy usage than traditional solvothermal methods and the avoidance of volatile organic solvents. We have found that the reaction mechanism is similar to previous preparations following a urea-hydrolysis pathway. The solvent degradation was monitored, showing that following the reaction the mixtures are in an off-eutectic composition.  Synchrotron WAXS measurements showed that this has surprisingly little effect on the major correlation structure in the solvent, signifying their resistance to this change. We nd additionally using WAXS that the hydrated solvent has a different structure from the pure system, with a more disordered structure, but one that retains some of the DES intermolecular bonding. The prepared iron oxide nanostructures vary in phase, size, and morphology as the synthesis conditions are varied. The DES was found to have some inherent structuring effect, in line with previous studies. Preparations using the pure DES always yielded very small nanoparticulates, whereas synthesis using hydrated DESs gave either 1D nanoshards at 150 C, or large rhombohedral nanoparticles at 200 C. Samples prepared at 150 C were the g-Fe 2 O 3 phase, whereas the syntheses conducted at 200 C yielded the a-Fe 2 O 3 phase. Investigations into the magnetic properties of these nanoparticles showed that the small g-Fe 2 O 3 and a-Fe 2 O 3 nanoparticles were sufficiently small that they exhibited superparamagnetic behaviour. The large, more crystalline samples synthesised with hydrated DESs showed ferrimagnetic or ferromagnetic hysteresis. Photoanodes were prepared from the nanoparticles using a previouslydeveloped solution-processed colloidal method, and photoelectrochemical measurements of these showed a competitive photocurrent density, with a maximum measured photocurrent of 0.7 mA cm À2 at 1.23 V vs. RHE. Whilst short of the most extreme reported values, this is a strong response when considering the environmental credentials of the process that was used to prepare them. We additionally demonstrate that the measured photocurrent is remarkably stable under repeated cycling.
We therefore present here a new route towards functional and highly active iron oxide nanomaterials to be used in photocatalytic water splitting applications, based around greener DESs as a structure-directing solvent medium. These new fundamental insights into the DES role in nanomaterials synthesis, and in particular the solvent structure information from synchrotron WAXS studies will aid with the development of future, greener processes towards other nanomaterials using DESs and hydrated DESs.