Tailoring of an unusual oxidation state in a lanthanum tantalum(IV) oxynitride via precursor microstructure design

Perovskite-type oxynitrides hold great potential for optical applications due to their excellent visible light absorption properties. However, only a limited number of such oxynitrides with modulated physical properties are available to date and therefore alternative fabrication strategies are needed to be developed. Here, we introduce such an alternative strategy involving a precursor microstructure controlled ammonolysis. This leads to the perovskite family member LaTa(IV)O2N containing unusual Ta4+ cations. The adjusted precursor microstructures as well as the ammonia concentration are the key parameters to precisely control the oxidation state and O:N ratio in LaTa(O,N)3. LaTa(IV)O2N has a bright red colour, an optical bandgap of 1.9 eV and a low (optically active) defect concentration. These unique characteristics make this material suitable for visible light-driven applications and the identified key parameters will set the terms for the targeted development of further promising perovskite family members. Perovskite-type oxynitrides hold great potential for optical applications due to their excellent visible light absorption properties. Here, a defined precursor microstructure is used to modulate the oxidation state of tantalum during ammonolysis leading to a bright red perovskite containing tantalum in a rare 4+ -state.

I n spite of the growing interest in perovskite materials, the number of oxynitride members is still limited. Most oxynitrides contain metal cations in d 0 or d 10 electronic configuration [1][2][3][4][5][6][7][8] , whereas alternative materials with B-site cations in e.g. d 1 configuration such as Ta 4+ are largely lacking. However, the latter could allow access to different electronic band structures, thus, expanding the applicability of perovskite-type oxynitrides in visible light-driven processes. The negligible number of known perovskite-type oxynitrides might be attributed to the reaction protocols normally used. Currently, the most widely used synthesis technique for this material class is the ammonolysis of oxides 9,10 . This procedure enables tuning of the electronic bandgap via (partial) substitution of oxygen by nitrogen 1 . Typically, the bandgap decreases, making the formed perovskite-type oxynitrides AB(O,N) 3 interesting for visible light-driven applications 1,4 . An important task in order to identify perovskite-type oxynitrides is the exact determination of the O:N ratio which often causes difficulties 11,12 .
The precursors are often mixtures of crystalline binary oxides (e.g. La 2 O 3 or Ta 2 O 5 ) 3,4,13 or ternary oxides (e.g. LaTaO 4 ) 14,15 , which are either prepared by solid state reactions (SSR) or via a Pechini method. The latter allows mild reaction conditions providing excellent product homogeneity 2 . In the case of LaTa (V)ON 2 , most studies involve a high-temperature treatment of the oxide precursors prior to ammonolysis, leading to high crystallinity of the precursor 13,14,16,17 . The formation of several perovskite-type oxynitrides such as LaTiO 2 N and SrTaO 2 N from crystalline oxide precursors has been described by a topotactic reaction scheme 18 . In contrast, the reaction behaviour of amorphous/nanocrystalline oxide precursors is still unclear and the detailed ammonolysis mechanism has yet to be clarified 1 . A better understanding and control of essential reaction steps might help to develop new synthesis strategies. Furthermore, a target-oriented electronic configuration of the B-site cation and the precise adjustment of the O:N ratio in oxynitrides are the key factors to generate various interesting physical properties.
In this in situ and ex situ experimental study, we override the above-mentioned strong topotactic relation between oxide precursor and resulting LaTa(O,N) 3 formation through a considered selection of well-characterised oxide precursors with different microstructures and an adjusted ammonia concentration. Furthermore, we demonstrate the formation of the LaTa(IV)O 2 N utilising nanocrystalline lanthanum tantalum oxide (n-LTO), which exhibits smaller primary particles (nm-range) and a higher specific surface area than microcrystalline LaTaO 4 (m-LaTaO 4 ). This adjusted precursor microstructure leads to a favoured Ta reduction in n-LTO. In contrast, ammonolysis of larger primary particles in the μm-range (m-LaTaO 4 ) results in conventional LaTa(V)ON 2 . Hence, we expand the experimental toolbox by an additional method to access further requested perovskite-type oxynitride family members.

Results
In situ ammonolysis of lanthanum tantalum oxides. First, the reaction steps of the LaTaO 2 N (Fig. 1) synthesis were investigated by in situ ammonolysis (10 vol% Ar in NH 3 ) using thermogravimetric analysis (TGA). For a better comparability, similar measurement conditions were selected for both in situ and ex situ ammonolysis (see below). Nanocrystalline lanthanum tantalum oxide (n-LTO) and microcrystalline LaTaO 4 (m-LaTaO 4 ) were synthesised as precursors in order to investigate the effect of the microstructure on the reaction behaviour (synthesis and characterisation details in Supplementary Note 1, Supplementary  Fig. 1-5 and Supplementary Tables 1-4).
The TGA curve of n-LTO during in situ ammonolysis (Fig. 2) reveals several mass changes, the origin of which were determined by termination experiments. The initial mass change of −0.7% in the range of 25°C to 588°C (region I) can be assigned to the desorption of residual water or organic contaminants, which is why the precursor remains white (Fig. 2).
The subsequent larger mass change of −2.9% (region II) is accompanied by a local mass minimum at 844°C. At this temperature, the powder is black and the powder X-ray diffraction (PXRD) pattern shows the onset of crystallisation. In addition, hot gas extraction (HGE) reveals a composition of LaTaO 3.44(4) N 0.41(2) □ 0.14(6) (product of the termination experiment at 844°C) with an assumed ratio of La:Ta:O = 1:1:4 (HGE results are listed in Supplementary Table 5). Therefore, region II (between 588°C and 844°C) is characterised by an oxygen vacancy formation with simultaneous nitrogen incorporation. According to literature 19 , LaTaO 4 crystallises in space group A2 1 am up to around 800°C and exhibits octahedron chains with corresponding interspace (Fig. 1). Therefore, we assume that vacancy formation and nitrogen incorporation takes place in this interspatial region.
At the same time, the colour-change of the powder from white to black indicates a reduction of the contained Ta. As can be shown by X-ray photoelectron spectroscopy (XPS), heating to 844°C in ammonia atmosphere causes the two existing Ta(V)-O binding characters in n-LTO (Fig. 3a, n-LTO) to change their character. The evaluation of the Ta(V)-O binding characters in n-LTO and m-LaTaO 4 is described in Supplementary Note 4. The two new binding energies of the Ta 4f 7/2 orbitals in LaTaO 3.44 (4) N 0.41(2) □ 0.14(6) of E B (Ta 4f 7/2 ) = 25.2 eV and E B (Ta 4f 7/2 ) = 24.4 eV, respectively, can be derived from the applied peak fittings (Fig. 3a, n-LTO, NH 3 / 844°C).
Calculations using the equations estimated by Nordling 20 (Supplementary Note 5) suggest that substitution of oxygen with nitrogen at persisting Ta 5+ causes a chemical shift of 0.3 eV to lower binding energies. The in situ observed conversion of n-LTO to LaTaO 3.44(4) N 0.41(2) □ 0.14(6) between 588°C and 844°C is accompanied by a Ta 4f 7/2 binding energy shift from 25.5 eV to 25.2 eV for Ta(V)-O by nitrogen incorporation. Therefore, the binding energy E B (Ta 4f 7/2 ) = 25.2 eV can be allocated to a Ta(V)-(O,N) binding character since the original binding energy lowers by the expected ΔE B = 0.3 eV. The other Ta 4f 7/2 binding energy of Ta (V)-O, however, shifts from 26.7 eV in n-LTO to 24.4 eV in LaTaO 3.44(4) N 0.41(2) □ 0.14(6) , the difference being much higher than ΔE B = 0.3 eV. This might be explained by a change of the Ta oxidation state in addition to nitrogen substitution. Therefore, the binding energy at E B (Ta 4f 7/2 ) = 24.4 eV can be assigned to a Ta (IV)-(O,N) binding character. The Ta(IV)-(O,N) binding character can be determined by applying the point charge model 20,21 assuming that the atomic potential of Ta remains unaffected by a change of the oxidation state. Ta(0) at E B (Ta 4f 7/2 ) = 21.9 eV 22 and Ta(V) at E B (Ta 4f 7/2 ) = 25.3 eV in m-LaTaO 4 (P2 1 /c) (instead of E B (Ta 4f 7/2 ) = 25.9 eV in m-LaTaO 4 (Cmc2 1 )) were selected as references (Fig. 3b). The selection of the Ta(V) binding energy was based on the similar interatomic distances in and between the [Ta V O 6 ] 7octahedron together with the amount of neighbouring ions of Ta compared to the respective oxynitride ([Ta(O,N) 6 ] zoctahedron). Thus, due to the linear relationship between the oxidation state and the binding energy in the point charge model, a Ta oxidation state of 4 + (d 1 electronic configuration) at E B (Ta 4f 7/2 ) = 24.4 eV was determined (Fig. 3b).
A further mass change of + 1.0% occurs in the temperature range of 844°C to 891°C (region III). Simultaneously, the colour of the powder changes from black to ochre. The weight fractions at 891°C determined by HGE amount to 13.7 wt% O and 6.7 wt % N (compared to 14.4 wt% O and 1.5 wt% N at 844°C).
However, the total mass increase due to the strong nitrogen enrichment accompanied by only a small oxygen loss cannot be explained by a simple refill of the previously generated oxygen vacancies in "LaTaO 4 ", since the total anionic weight fraction substantially exceeds the calculated maximum value of 16.7 wt% for "LaTaO 4 " (Supplementary Table 5). Evaluation of the respective powder pattern shows that the ochre-coloured phase is not fully crystallised (Supplementary Note 6 and Supplementary A further temperature increase to 950°C (region IV) leads to an abrupt mass change of −6.9%. During the following 10 h ammonolysis prior to cooling to 25°C the mass remains near-constant. The respective termination experiment indicates a colour change from ochre to red after the 10 h ammonolysis. HGE of the red phase reveals a composition of LaTaO 1.44(1) N 1.63 (9) . This phase is further transformed to red LaTaO 1.26(9) N 1.83(4) by a second heating cycle under ammonia at 1000°C for 14 h. Such intermediate compositions during the synthesis of LaTa(O,N) 3 are often reported in literature when large amounts of oxide precursors are used or the applied precursor has been crystallised at around 1000°C before the ammonolysis 24 . Additionally, a large sample amount can cause inhomogeneous exposure to the reducing species during ammonolysis. The positive and negative mass changes observed during the heating and cooling steps seem to be caused by desorption and adsorption of gaseous species (e.g. H 2 O and/or NH 3 ). The nitrogen content of both red phases falls short of that of LaTaON 2 , revealing that n-LTO is susceptible to the reduction of tantalum (Ta 5+ to Ta 4+ ). The usage of m-LaTaO 4 as a   Fig. 2 In situ TGA ammonolysis of nanocrystalline lanthanum tantalum oxide (n-LTO). In situ TGA of the ammonolysis (10 vol% Ar in NH 3 ) of n-LTO, including respective powder X-ray diffraction (PXRD) patterns and anionic compositions determined via hot gas extraction (HGE) after selected termination experiments. The illustrative coloured powders at specific temperatures complement the analysis. The reduction of tantalum at 844°C is highlighted with a red frame. Additionally, the first four mass changes (region I-IV) during in situ ammonolysis of n-LTO are magnified in the inset. The in situ ammonolysis results of the microcrystalline LaTaO 4 (cyan curve (micro) in the same inset) can be found below.
precursor for in situ ammonolysis results in LaTaON 2 , neither involving intermediate phases nor reduction of Ta(V) to Ta(IV) (no black powder indicating a reduction). Instead, the initial formation of oxygen vacancies (yellow LaTaO 3.87(7) □ 0.12(3) ) is followed by a one-step mass change of −5.5% (onset at 820°C) indicating the conversion to LaTaON 2 (Fig. 4).
To be more precise, the in situ ammonolysis of m-LaTaO 4 starts with a mass change of -0.5% between 25°C and 820°C (Fig. 4). According to the PXRD results of termination products, the crystal structure of m-LaTaO 4 remained unchanged in this temperature range. However, a colour change from white (25°C) to yellow (820°C) indicated a change of the chemical composition including oxygen vacancy formation, which was confirmed via hot gas extraction (Supplementary Table 5). Further heating to 950°C led to a change of the powder colour from yellow to purple accompanied by a massive mass change of −5.5%. The detected mass change is in accordance with the expected mass change of Δm calc. = −5.2% calculated for the formation of LaTaON 2 from LaTaO 4 . Termination experiments at 820°C and at 950°C together with hot gas extraction measurements revealed the conversion of microcrystalline LaTaO 4 to LaTaON 2 above 820°C by nitrogen incorporation. An isothermal step with a near-constant mass was followed by a second ammonolysis cycle at 1000°C for 14 h. The PXRD pattern of the final product clearly showed the presence of a perovskite-type phase (insets, Fig. 4). The mass changes during these heating and cooling steps might be attributed to the reversible adsorption and desorption of ammonia and water molecules as previously mentioned for n-LTO. We assume that a strong topotactical relation between crystalline ABO 4 and crystalline ABON 2 often described in literature 18   During oxygen vacancy formation and subsequent nitrogen incorporation they connect to the neighbouring chains in a "zipper-type" manner 18 . However, the microstructure of n-LTO is different and exhibits therewith a "soft" topotactic relation to the resulting oxynitride along the interspace of the octahedron chains ( Fig. 1) inducing a faster reduction of Ta(V) to Ta(IV). Afterwards, a "soft" topotactic reaction with a simultaneous "zipper-type" octahedral connection 18 enables the afore mentioned filling of oxygen vacancies by nitrogen.
The lower nitrogen content of LaTa(O,N) 3 in comparison to LaTaON 2 when using n-LTO for in situ ammonolysis can be explained by the specific surface area and the primary particle size and, thus, the microstructure of the oxide precursors. The specific surface area of n-LTO is S BET = 7 m²/g (primary particle size: nmrange) and that of m-LaTaO 4 S BET = 2 m²/g (primary particle size: μm-range). Further details about the oxide precursors can be found in Supplementary Note 1. The higher surface area and the smaller primary particle size of the nanocrystalline precursor compared to m-LaTaO 4 makes it more susceptible to Ta reduction, since the diffusion of reducing agents is simplified (at higher temperatures NH 3 decomposes into nitrogen-and hydrogen-containing species and molecular hydrogen 9 . A detailed explanation of the interaction of the reducing species with n-LTO can be found in the Supplementary Note 7: Reduction of Ta. This is confirmed by the finding that in situ ammonolysis of n-LTO that has been previously converted to m-LaTaO 4 also leads to LaTaON 2 ( Fig. 5 and Supplementary Note 2).
Moreover, scanning electron microscopy (SEM) images of the termination products of the annealing process of n-LTO in syn. The first mass change of -1.8% is caused by desorption of water. The second negative mass change between 803°C and 923°C is due to pyrolysis of organic residues from the preceded soft chemistry synthesis. Thermogravimetric measurements coupled with mass spectrometry (TGA-MS) revealed the release of CO 2 , NO 2 and organic fragments. The measurement results are displayed in Supplementary Fig. 4 and discussed in Supplementary Note 3. The measurement was first carried out in synthetic (syn.) air prior to the ammonolysis step. air prior to in situ ammonolysis (Fig. 6a) show a continuous microstructural change with increasing temperature leading to a morphology very similar to m-LaTaO 4 (Fig. 6c).
In parallel, the onset temperature of the ammonolysis reaction steadily increases from 588°C to 820°C with increasing primary particle size and crystallinity (n-LTO to m-LaTaO 4 , Figs 2, 5 and 6). Hence, a well-designed microstructure allows to enable or supress significant Ta reduction.
Ex situ ammonolysis of lanthanum tantalum oxides. Eventually, in situ ammonolysis of n-LTO did not deliver LaTaO 2 N which might be attributed to the less reductive atmosphere caused by the 10 vol% Ar in NH 3 needed to protect the device. The application of nearly 100 vol% instead of 90 vol% NH 3 in a classical thermal gas flow ammonolysis setup turned the in situ into an ex situ ammonolysis. The ex situ ammonolysis only slightly differs from the in situ setup with respect to the sample environment. Using the same temperature profile as in the in situ ammonolysis and an NH 3 flow rate of 300 mL/min we finally converted n-LTO into bright red LaTaO 1.98(7) N 0.98(5) (S BET = 25 m 2 /g) after 10 h at 950°C via ex situ ammonolysis (anionic composition obtained by HGE, Supplementary Table 6). However, when using the microcrystalline instead of the nanocrystalline precursor for ex situ ammonolysis with the same temperature profile and ammonia flow as in in situ ammonolysis, purple-coloured LaTaO 1.05(3) N 2.02(3) is obtained (very similar to the in situ ammonolysis product). This indicates a very similar parameter selection in ex situ and in situ ammonolysis. Considering the results of all in situ and ex situ ammonolysis   experiments, we have identified the reduction sensitivity due to microstructural differences and the ammonia concentration as main parameters to adjust the O:N ratio in LaTa(O,N) 3 . The aforementioned red LaTaO 1.98(7) N 0.98 (5) and the purple LaTaO 1.05(3) N 2.02(3) phases prepared by ex situ ammonolysis are subjected to a second ex situ ammonolysis cycle (same temperature profile as in in situ ammonolysis) with KCl flux addition in order to heal possible defects 25 . After the second cycle the compositions slightly change to a bright red LaTaO 1.97(3) N 1.02(5) (S BET = 8 m 2 /g) and to purple LaTaO 0.99(1) N 2.00(9) (S BET = 3 m 2 /g), respectively.  (5) ) also confirmed the absence of Ta 5+ (observed, however, in LaTaON 2 (LaTaO 0.99(1) N 2.00 (9) ) eliminating the possibility of a 1:1 mixture of Ta 3+ and Ta 5+ in LaTaO 2 N.
Investigations of the magnetic properties of LaTaO 2 N and LaTaON 2 by superconducting quantum interference device (SQUID) measurements were carried out in order to confirm the presence of Ta 4+ although challenging due to the presence of magnetic impurities such as Ta 3+ as observed from XPS analysis (Fig. 3) and the in general limited knowledge about the magnetism of 5d transition metal compounds (compared to their 3d counterparts) 28,29 . The respective zero field cooled (ZFC) curves at 500 Oe down to 2 K are shown in Supplementary Note 10 and Supplementary Fig. 9. The low temperature regions (below 70 K) can be described by a paramagnetic Curie-Weiss-like behaviour with very small effective moments (1.3 × 10 −4 µ B /Ta at 2 K for the LaTaO 2 N sample and 9.8 × 10 −5 µ B /Ta for the LaTaON 2 sample, respectively) pointing to an activation of magnetic impurity states rather than an intrinsic materials property. The effective magnetic moment is further drastically reduced at increasing temperature. Overall the observed magnetisation is much lower than 1 µ B /Ta as expected from the simplest paramagnetic model. The small magnetic moment and the fact that Ta is a 5d element suggests that the largely extended 5d orbitals in LaTaO 2 N are strongly hybridised and overlapping with the O/N 2p orbitals 30 leading to weakly localised electrons hampering the up-built of a significant magnetisation. It is wellknown for 3d transition metal containing perovskites ABO 3 and ABF 3 that with decreasing electronegativity of the anion the hybridisation (or in other words the covalency of the B-X bond) is enhanced resulting in the aforementioned stronger delocalisation of the electrons and hence the spin 30 . We believe the same is valid for 5d transition metal containing perovskites. Besides, two further factors could contribute diminishing the effective magnetic moment: (i) the spin-orbit coupling being expected to be much stronger in 5d than in 3d materials, (ii) the experimentally observed presence of stretched octahedra (Supplementary Note 11 and Supplementary Table 9e) results in a splitting of the initially degenerated t 2g levels. This splitting leads to the formation of four-fold degenerated 5d xz and 5d yz states lower in energy than the initial state only occupied by one electron, making it difficult to develop magnetic ordering 31 .
Phase purity of LaTaO 2 N is proven by high resolution (HR)-PXRD. LaTaON 2 has already been reported in the space groups Imma and C2/m 17,32,33 and both are considered for LaTaO 2 N as well. Since we have observed no clear evidence for a monoclinic distortion and the Imma space group gives a slightly lower χ 2 residual than C2/m, we propose Imma as space group for LaTaO 2 N. The same applies to LaTaON 2 (Supplementary Fig. 10 and Supplementary Table 9 for complete crystal structure analysis). The unit cell volumes of LaTaO 2 N (V cell = 264.78(3) Å 3 ) and LaTaON 2 (V cell = 264.68(2) Å 3 ) are very similar. The slightly larger unit cell of LaTaO 2 N can be explained by the increase of the effective ionic radius of Ta from 0.64 Å (Ta 5+ ) 34 to 0.68 Å (Ta 4+ ) 34 after ammonolysis. This expansion is mostly compensated by the different O:N ratio. The partial replacement of the larger nitrogen (1.46 Å) 34 by the smaller oxygen (1.40 Å) 34 leads to a contraction of the unit cell. Furthermore, LaTaO 2 N is phase pure, while in LaTaON 2 a small amount of Ta 3 N 5 (Cmcm) was detected through HR-PXRD ( Supplementary Fig. 10b and Supplementary Table 9d). Since XPS reveals TaN on the surface and HR-PXRD Ta 3 N 5 in the bulk, a nitrogen gradient combined with a decreased susceptibility for the reduction of Ta (from Ta 5+ to Ta 3+ ) can be assumed. However, TaN can also be amorphous or the amount below the HR-PXRD detection limit.
Scanning electron microscopy gives insight into the morphology of the oxynitrides synthesised by ex situ ammonolysis. LaTaO 2 N shows primary particles in the nm-range, while LaTaON 2 exhibits porous and sintered particles in the μm-range (Fig. 7a).
In addition to the determination of the specific surface area (Supplementary Table 10), an increased pore formation in LaTaON 2 during ammonolysis can also be confirmed by comparing the SEM images of the oxide with its respective oxynitride. Furthermore, the particle size increases during ammonolysis of n-LTO, whereas using m-LaTaO 4 the particle size remains constant (Fig. 6c). The respective SEM images of the oxides can be found in the SI (Supplementary Fig. 2c).
As the colours of the obtained oxynitrides differ from each other, diffuse reflectance spectroscopy (DRS) measurements were performed. The dataconverted to Kubelka-Munk 35 curves (Fig. 7b) show a difference of ΔE g = 0.1 eV between the optical bandgaps of the LaTaO 2 N (1.9 eV) and LaTaON 2 (1.8 eV). Detailed investigation shows that the Kubelka-Munk curve of LaTaO 2 N converges to zero in contrast to that of LaTaON 2 , indicating a much lower optically active defect concentration. In addition, after the optical bandgap of E g = 1.8 eV an intensity increase as observed for LaTiO 2 N 36 , which exhibited undesired defects was not detected for LaTaO 2 N. LaTaO 2 N possess an unusual colour (bright red), which has been expected to be darker (bluish 37 ) due to the reduction of Ta 5+ to Ta 4+ . Additionally, the larger optical bandgap of LaTa(IV)O 2 N compared to LaTa(V)ON 2 is noticeable. Based on both observations, we assume a larger crystal field splitting for LaTaO 2 N caused by the d 1 state of Ta 4+ , since Ta is coordinated in a distorted octahedral environment (Jahn-Teller effect 31 ) indicated by bond length analysis via Rietveld refinements. Specific visible light-driven applications require materials with appropriate defect concentrations [38][39][40] and suitable bandgaps [41][42][43] . Therefore, LaTaO 2 N could be an interesting candidate for further investigations in this application field.

Discussion
In this in situ and ex situ ammonolysis study the key parameters to tailor the oxidation state and to synthesise a perovskite family member LaTa(IV)O 2 N containing unusual Ta 4+ were identified. The oxynitride has an optical bandgap of 1.9 eV, a bright red colour, and a low optically active defect concentration providing promising physical properties for light-driven applications. The formation of LaTa(IV)O 2 N is boosted by the oxide precursor's microstructure (primary particle size, specific surface area and crystallinity) and the ammonia concentration (nearly 100%). In previous studies, the ammonolysis mechanism was already investigated and mostly clarified for the topotactic case. The findings shown here make a substantial contribution to the elucidation of the ammonolysis mechanism in general and to that of amorphous/nanocrystalline oxide precursors in particular. This opens up new perspectives and possibilities for the synthesis of further perovskite-type oxynitrides.

Methods
Synthesis of LaTaO 2 N and LaTaON 2 . The oxynitrides LaTaO 2 N and LaTaON 2 were prepared from the respective nanocrystalline and microcrystalline oxide precursors (cf. Supplementary Methods) via thermal ammonolysis (synthesis of the respective oxides is described in the SI). The precursor oxides (200 mg) were transferred to an alumina boat and ammonolysed at 950°C for 10 h with a NH 3 gas flow rate of 300 mL/min (Westfalen AG, > 99.98%). A second ammonolysis step together with KCl flux (Roth, ≥ 99%, Ph. Eur.) was carried out in a 1:1 weight ratio of sample and flux at 1000°C for 14 h.
Sample characterisation. In order to clarify the phase purity and crystal structure of the produced oxides and oxynitrides, powder X-ray diffraction (PXRD) measurements were carried out on a Rigaku Smartlab powder X-ray diffractometer using Cu-K α 1,2 radiation. To avoid a contribution of Cu-K β radiation a thin nickel foil as filter was used (efficiency is~90%). The continuous scan covered an angular range of 10°≤ 2θ ≤ 90°with an angular step interval of 0.025°. For selected oxide and oxynitride samples additional high-resolution synchrotron radiation PXRD measurements were performed at the beam line ID22 of the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. The powders were filled into 0.7 mm diameter Kapton® capillaries (wavelength, see specific refinements). The collected diffraction data were evaluated via Rietveld refinements 44,45 using Full-Prof. 2k 46 . A pseudo-Voigt function was selected to describe the reflection profile and the background was linearly interpolated between a set of background points with refinable heights. The anionic composition of the oxynitrides were fixed according to the respective compositions determined by HGE because the virtually equal form factors make it impossible to discriminate between O 2− and N 3− by means of X-rays. O and N were statistically assigned to the two independent crystallographic sites.
The chemical composition of the produced samples was investigated via inductively coupled plasma optical emission spectroscopy (ICP-OES) using a Spectro Ciros CCD ICP-OES instrument for cations and hot gas extraction technique (HGE) using an Eltra ONH-2000 analyser for the anions.
The investigation of oxynitride formation from nanocrystalline and microcrystalline oxides and the reverse reaction of LaTaO 2 N to the corresponding oxide was performed by thermogravimetric analysis (TGA) using a Netzsch STA 449 F3 Jupiter. In situ ammonolysis experiments were carried out under flowing NH 3 (80 mL/min NH 3 + 8 mL/min Ar) on alumina plates with a heating rate of 10°C /min up to 1000°C. To protect the TGA device from corrosion the measurements were performed in 10 vol% Ar in NH 3 . A fast cooling rate of 40°C/ min was used to successfully quench the intermediates since the mass changes observed by TGA and HGE were in good agreement. Reoxidation was carried out on an alumina plate under flowing synthetic air (20.5 vol% O 2 in N 2 , Westfalen AG, 50 mL/min) to study the anionic composition of LaTaO 2 N. The sample was first heated up to 200°C and maintained at this temperature for half an hour to remove surface adsorbed water. Thereafter, heating was continued up to 1400°C at a heating rate of 10°C/min. The temperature was maintained for 2 h in order to achieve full conversion of oxynitride to respective single-phase oxide. TGA-MS to determine possible organic residues in n-LTO was carried out with a Netzsch STA 449 C Jupiter coupled with a GAM 200 (InProcess Instruments) mass spectrometer. The oxide was heated to 1200°C at a rate of 10°C/min in a crucible under syn. air (20.5 vol% O 2 in N 2 , Westfalen AG, 50 mL/min) and then cooled down to 25°C. X-ray photoelectron spectroscopy (XPS) on LaTaO 2 N, LaTaON 2 , n-LTO, m-LaTaO 4 and termination products was carried out to investigate the oxidation states using a Thermo VG Theta Probe 300 XPS system from Thermo Fisher Scientific. The incident beam provided monochromatic and micro-focused Al K α radiation and a spot size of 400 μm. The powders were fixed on a carbon tape and a flood gun was used to avoid charging effects. For background subtraction a Shirley-type inelastic background was utilised and the zero-shift correction was done by normalizing the measured C 1 s peak to 284.5 eV. The peak fitting was implemented by carefully considering quantum mechanical requirements for the intensity and energy relations of Ta 4f 7/2 and Ta 4f 5/2 orbitals with ΔE B = 1.91 eV 22 and identical full width at half maximum (FWHM) for both orbital contributions.
The morphology of the produced oxynitrides and oxides was analysed via scanning electron microscopy (SEM) (ZEISS GeminiSEM 500, 5 kV). The in-lens detector was used for imaging.
UV-visible diffuse reflectance spectra (DRS) were obtained using a Carry 5000 UV-VIS NIR spectrophotometer. The baseline was measured with BaSO 4 . The spectra were recorded in the range of 200-800 nm. The Kubelka-Munk 35 conversion was applied to the obtained reflectance spectra and the optical bandgap was estimated by extrapolating the onset of absorption to the abscissa.
The specific surface area was obtained via nitrogen sorption, first annealing the samples at 120°C to remove adsorbed water. Adsorption and desorption isotherms were collected at liquid nitrogen temperature using an Autosorb-1-MP (Detection limit: S BET > 1 m 2 /g) from Quantachrome Instruments. The specific surface area was determined via the Brunauer-Emmet-Teller 47 (BET) method.
SQUID measurements to investigate the magnetic behaviour of Ta in LaTaO 2 N and LaTaON 2 were carried out with a commercial VSM MPMS3 Superconducting Quantum Interference Device (SQUID) from Quantum Design. For zero field cooling (ZFC) measurements the magnetic field was set to 500 Oe.