Low-temperature synthesis of cation-ordered bulk Zn3WN4 semiconductor via heterovalent solid-state metathesis

Metathesis reactions are widely used in synthetic chemistry. While state-of-the-art organic metathesis involves highly controlled processes where specific bonds are broken and formed, inorganic metathesis reactions are often extremely exothermic and, consequently, poorly controlled. Ternary nitrides offer a technologically relevant platform for expanding synthetic control of inorganic metathesis reactions. Here, we show that energy-controlled metathesis reactions involving a heterovalent exchange are possible in inorganic nitrides. We synthesized Zn3WN4 by swapping Zn2+ and Li+ between Li6WN4 and ZnX2 (X = Br, Cl, F) precursors. The in situ synchrotron powder X-ray diffraction and differential scanning calorimetry show that the reaction onset is correlated with the ZnX2 melting point and that product purity is inversely correlated with the reaction's exothermicity. Therefore, careful choice of the halide counterion (i.e., ZnBr2) allows the synthesis to proceed in a swift but controlled manner at a surprisingly low temperature for an inorganic nitride (300 °C). High resolution synchrotron powder X-ray diffraction and diffuse reflectance spectroscopy confirm the synthesis of a cation-ordered Zn3WN4 semiconducting material. We hypothesize that this synthesis strategy is generalizable because many Li–M–N phases are known (where M is a metal) and could therefore serve as precursors for metathesis reactions targeting new ternary nitrides. This work expands the synthetic control of inorganic metathesis reactions in a way that will accelerate the discovery of novel functional ternary nitrides and other currently inaccessible materials.


Introduction
Ternary nitrides are a promising class of semiconducting materials, [1] yet relatively few are known.[3][4] Molecular (di)nitrogen, N2, is highly stable, and high temperatures are needed to break the strong N≡N triple bond (945 kJ/mol). [5][8] Moreover, entropic penalties disfavor nitride incorporation in solids (i.e., gaseous N2 is favored).Finding a synthesis temperature that is hot enough for reactivity but cool enough to avoid decomposition is therefore challenging.Adding to the difficulty, O2 is more reactive towards most metals than N2, so syntheses must be conducted in rigorously air-free conditions to avoid the formation of oxide impurities.[3][4] Developing new synthesis methods will help narrow this disparity, and in doing so, discover new materials upon which improved technologies can be built.
Zn-containing ternary nitrides epitomize the promising applications and synthetic challenges of this class of materials.Fully nitridized compounds like ZnSnN2 and Zn3WN4 (with metals in the highest oxidation state) are of interest as semiconductors for their high earth abundance and tunable bandgaps (spanning ca. 1 eV for ZnSnN2 to 4 eV for Zn3WN4). [9,10]However, the bulk synthesis techniques that have been reported for Zn-M-N phases are limited to traditional ceramic methods (i.e., metals + N2 or NH3 at high temperatures) or high-pressure solid state metathesis reactions (e.g., 2 Li3N + ZnF2 + SnF4 à ZnSnN2 + 6 LiF). [11,12][22][23][24][25] The nitrogen-poor nature of these materials stems from the challenges described above (i.e., N2 stability, slow diffusion).Synthesizing fully-nitridized Zn-M-N (where M is a transition metal) in bulk would advance technologies in which thin film nitrides have already shown promise, like photoelectrochemical energy conversion (ZnTiN2), [26] transparent conducting oxides (ZnZrN2), [27] and non-linear optics (Zn3WN4). [28]However, no bulk synthesis methods have been reported for fully nitridized Zn-M-N ternaries.
Synthesizing Zn-M-N ternary nitrides via traditional methods is difficult.Many transition metals are highly refractory, meaning high temperatures would likely be needed for interdiffusion of reactants.However, Zn has a low melting point (419 °C) and a relatively low boiling point (907 °C), [29] meaning that high temperatures would volatilize Zn away from the other metal unless special measures were taken (e.g., high pressure, closed vessels).Forming binary nitrides to use as precursors instead of metals is also challenging: Zn (like other late-transition metals) does not react with N2 at elevated temperatures, so Zn3N2 must be synthesized under ammonia. [30]And as noted in thin film work, fully nitridized transition metal Zn-M-N phases have low decomposition temperatures on the order of 600-700 °C. [3,26,27,31]These challenges mean that bulk synthesis of Zn-M-N from the elements or binaries would likely proceed only at low temperatures and extremely slowly, unless special high-pressure methods were employed (e.g., ammonothermal synthesis, [12] diamond anvil cell synthesis, [32] etc.).
Metathesis reactions (i.e., ion exchange reactions) are one promising way to circumvent the challenge of diffusion in the solid state. [33]To synthesize nitrides, this strategy starts with one nitrogen-containing precursor and one halide precursor, rather than elements or binary nitrides.The balanced reaction targets the desired phase along with a byproduct (often a halide salt).The formation of this byproduct provides a large thermodynamic driving force for the reaction and (ideally) can be washed away post-reaction.For example, Kaner et al. showed that mixing Li3N with metal chlorides would produce LiCl in explosively exothermic metathesis reactions that yielded a range of binary nitrides [34][35][36][37][38][39][40][41][42][43][44] and some ternary nitrides. [45,46]Alternatively, less exothermic reactions can be conducted with greater synthetic control, [47][48][49][50][51] including lowtemperature topotactic reactions (Trxn ca.[54] As for Zn-M-N compounds, ZnSnN2 and ZnSiN2 have been made using high pressure metathesis reactions, where the pressure is necessary to avoid gaseous N2 loss. [11,19]Metathesis is well known for "turning down the heat" in solid state synthesis [55] but is underutilized for synthesizing nitrides. Here, we synthesize Zn3WN4 via a near-topotactic metathesis reaction between Li6WN4 and ZnX2 (X = Br, Cl, F) at 300 °C and ambient pressure.In situ synchrotron powder X-ray diffraction (SPXRD) paired with differential scanning calorimetry (DSC) measurements reveal the reaction pathways and show that using a ZnBr2 precursor is preferable over the fluoride or chloride analogs.High resolution SPXRD measurements indicate that the Zn3WN4 product is a mostly cationordered structure in space group Pmn21.We report some preliminary properties characterizations for Zn3WN4, revealing optical absorption onsets near 2.5 eV and 4.0 eV, as well as paramagnetism consistent with some degree of disorder and off-stoichiometry.The reaction is near-topotactic, in that the structures of the Li6WN4 precursor and the Zn3WN4 product are related by a shift in anion layers but the [WN4] tetrahedral unit is preserved.Using this synthesis approach, we also synthesized Zn3MoN4, albeit with lower levels of purity in our un-optimized reactions.This work demonstrates the viability of Li-M-N phases as metathesis precursors to synthesize other ternary nitride compounds, expanding the toolkit for materials discovery.

Results and Discussion
In situ SPXRD measurements Zn3WN4 was successfully synthesized via metathesis (ion exchange) reactions.The net reaction is: Li6WN4 + 3 ZnX2 à Zn3WN4 + 6 LiX (X = Br, Cl, F) In situ variable temperature SPXRD measurements reveal that Li6WN4 directly converts to Zn3WN4 without intermediate crystalline phases or solid solution behavior as a function of temperature (Figure 1, Figures S2, S3).However, the halide precursor exerts an influence on the reaction kinetics and thermodynamics, which ultimately impact the reaction pathway and final product purity.[58] Analogous plots of the ZnCl2 and ZnF2 reactions are in Figures S2  and S3.
In situ SPXRD measurements reveal that the reaction pathway proceeds without intermediate nitrides between Li6WN4 and Zn3WN4.Figure 1 shows a heatmap for X = Br as an example; subsequent examination revealed that it leads to the most phase pure product.The reaction of Li6WN4 + 3 ZnBr2 à Zn3WN4 + 6 LiBr initiates near 170 °C and proceeds to completion within the 14 minutes of ramp time up to 310 °C.Near 170 °C, the Bragg peaks arising from crystalline Li6WN4 and ZnBr2 begin to gradually decline in intensity.Shortly thereafter, new sets of Bragg peaks that can be indexed to LiBr, Li2ZnBr4, and Zn3WN4 begin growing in intensity in the patterns.The Bragg peaks corresponding to Li2ZnBr4 gradually decrease in intensity between 210 °C and 270 °C, increase dramatically in intensity at 275 °C, and then disappear entirely at 305 °C.Such fluctuations may stem from crystal nucleation and growth within the capillary, especially given the small spot size of the synchrotron X-ray beam, possibly combined with crystallite motion in a liquid-like medium.Diffraction images show spotty diffraction patterns, consistent with crystallite growth.These data indicate that the synthesis proceeds directly via Li6WN4 + 3 ZnBr2 à Zn3WN4 + 6 LiBr.While this process occurs, an incidental reaction between the metal halides also occurs: 2 LiBr + ZnBr2 à Li2ZnBr4.We do not observe signs of a crystalline theoreticallypredicted LiZn4W2N7 structure, [59] although this does not rule out the presence or synthesizability of such a phase.Similar trends are noted with the ZnCl2 and ZnF2 reactions (Figures S2, S3), as shown by sequential Rietveld analysis (Figure 2).Sequential Rietveld analysis of in situ variable temperature SPXRD measurements of the Li6WN4 + 3 ZnX2 reactions shows that the ZnBr2 and ZnCl2 reactions initiate at much lower temperatures than the ZnF2 reaction (Figure 2).For both the ZnBr2 and ZnCl2 reactions (Figure 2a,b), the concentrations of the precursor phases start decreasing near 170 °C, followed shortly thereafter by Zn3WN4 and LiX formation and growth.Ternary halides Li2ZnBr4 and Li2ZnCl4 are short-lived, incidental intermediates.In contrast, in the fluoride reaction, the concentration of ZnF2 does not begin declining until approximately 300 °C (Figure 2c).The concomitant decrease in Li6WN4 concentration suggests reactivity, but neither Zn3WN4 nor LiF are detected in our data.Instead, very weak reflections for an unknown phase appear in the data (labeled as Unk*).This phase may be a Li-Zn-F ternary, but it does not index to any known ternary fluoride unit cells, including the reported Li2ZnF4 phase. [60]An amorphous phase is likely present in the 400 to 570 °C region, given Figure 2. Sequential Rietveld analysis of in situ SPXRD patterns yields the weighted scale factors (W.S.F.), which are plotted as a function of temperature.The W.S.F.indicates the relative phase concentrations for the crystalline phases during reactions between Li6WN4 and a) ZnBr2, b) ZnCl2, and c) ZnF2.One unknown phase (Unk*, likely a Li-Zn-F phase) is modeled using a peak fit to the most intense reflection, rather than full-pattern Rietveld analysis.the decrease in precursor peaks and lack of new intermediate peaks.Zn3WN4 and LiF crystallize above 570 °C, along with a rocksalt phase (fit with WN, but the material may be a (Zn,W)Nx phase as observed with the Mo-based system, Figure S11).We did not study the ZnF2 reactions further, given that phase-pure Zn3WN4 did not crystallize and given the challenge associated with washing away LiF from the product.Instead, we focus on ZnBr2 and ZnCl2 reactions.The lower exothermicity of the ZnBr2-based reaction leads to a more controlled release of heat and greater product purity, compared to the ZnCl2-based reaction.DSC measurements show that the ZnBr2 reaction has three small exotherms (Figure 3a).A gradual exotherm starts near 190 °C (ai), followed by two exotherms near 305 °C (a-ii) and 334 °C (a-iii).This third event is proceeded by a very minor endotherm, possibly consistent with Li2ZnBr4 melting.This reactivity occurs well below the 392 °C melting point of ZnBr2, suggesting the process is mostly a solid-state reaction.The ZnCl2 reaction (Figure 3b) starts similarly, with a gradual exotherm between 190 °C and 280 °C (b-i).Then at ~280 °C, a massive exotherm (b-ii) initiates just below the melting point of ZnCl2 (325 °C).This event likely corresponds to the formation of a liquid phase, such as a LiCl-ZnCl2 eutectic (287 °C at 91% ZnCl2). [61]Peak b-ii has curvature because this event releases heat so quickly that the DSC stage increases in temperature by approximately 15 °C, after which the DSC pan cools slightly.This kind of rapid exothermic event is common in metathesis reactions; once a liquid phase forms, reaction kinetics accelerate and the heat release self-propagates. [38]Lastly, a small endotherm is observed at 336 °C (b-iii), consistent with the melting of Li2ZnCl4.These results are broadly consistent with the in situ SPXRD results, albeit shifted slightly in temperature owing to differences in experimental configuration.

DSC measurements
These DSC results show why the ZnBr2 reaction yields the purest product while the ZnCl2 reaction exhibits a small Zn impurity.The rapid release of heat in the ZnCl2 reaction causes small portions of the material to decompose: Zn3WN4 à W + 3 Zn + 2 N2 or Zn3WN4 à WN + 3 Zn + 3/2 N2 (Figure S4).In contrast, the washed product of the ZnBr2 synthesis yielded a PXRD pattern with all Bragg peaks indexed to Pmn21 Zn3WN4.These differences can easily be seen in the color of the material (see insets, Figure 3), where Zn impurities in the ZnCl2 reaction led to a grey color.The Zn3WN4 sample produced by the ZnBr2 reaction is brown.These in situ SPXRD and DSC measurements guided our optimization of the synthesis for Zn3WN4.

Structural and composition analysis of Zn3WN4
The best conditions we found were to heat ZnBr2 with Li6WN4 at a ramp of +5°C/min to 300 °C for a 1 h dwell, followed by natural cooling in the furnace.This reaction was scaled up to ca. 1 g reactant mix for ex situ analysis.Washing with anhydrous methanol successfully removed byproduct LiBr and excess ZnBr2 while preserving the targeted phase.We used a slight excess of ZnBr2 (3.1 ZnBr2 + Li6WN4) to ensure complete conversion and minimize reaction temperature by acting as a heat sink.XRF measurements show a Zn:W ratio of 3.165(3):1, slightly higher than the expected 3:1 ratio of Zn3WN4 (representative XRF spectrum shown in Figure S5), which may be a result of the excess ZnBr2.PXRD techniques confirmed that this synthesis of Zn3WN4 proceeded without the formation of decomposition products (i.e., Zn, (Zn,W)Nx phases).High resolution SPXRD measurements confirm the successful synthesis of Zn3WN4 (Figure 4).Rietveld analysis of the SPXRD data (Figure 4a) shows that Zn3WN4 crystallizes in space group Pmn21 with lattice parameters a = 6.5602(8)Å, b = 5.6813(7) Å, and c = 5.3235(2) Å.The presence of intensity at the (010), ( 110), (101), and (011) Bragg positions indicates a substantial degree of cation ordering.The peaks for the (210), (002), and (211) reflections are characteristic of wurtzite-derived structures; these correspond to the (100), (002), and (101) reflections in the prototypical wurtzite structure (P63mc), respectively (Figure 4b).Rietveld-refined occupancies suggest a Zn:W ratio of 3.8:1, a higher ratio than that measured by XRF (3.2:1), with partial occupancy of Zn on the W site (Table S1), indicating a composition of Zn3.17W0.83N4(Figure 4d).The occupancies of the N atoms refined to 1 within error and so were fixed at unity.Alternative structural models were also considered (Table S2, Figures S6-S8), as discussed further in the section on cation ordering in Zn3WN4.

Property measurements
Diffuse reflectance spectroscopy measurements reveal two absorption onsets for Zn3WN4: one near 2.5 eV and another near 4.0 eV (Figure 5).This first feature is similar to the 2.0 eV to 2.4 eV absorption onset reported for cation disordered Zn3MoN4 and Zn3WN4 synthesized as thin films. [31,62]The second absorption feature occurs near 4.0 eV, which is consistent with the expected bandgap for fully cation-ordered Zn3WN4 in the Pmn21 space group.The GW-calculated indirect band gap is 3.96 eV, with a direct bandgap of 4.20 eV (NREL MatDB ID 287103; blue trace in Figure 5). [63,64]Other researchers using a hybrid functional, HSE06, calculated the bandgap to be 3.60 eV. [28]These two features therefore suggest that our Zn3WN4 powder sample is a mixture of cation-ordered and cation-disordered structures.Magnetic susceptibility measurements are consistent with the presence of an impurity, as the material does not exhibit purely diamagnetic behavior (Figure S9).

Cation ordering in Zn3WN4
Our metathesis approach yielded a different polytype for Zn3WN4 compared to prior thin film syntheses.We show here that metathesis between Li6WN4 and ZnBr2 successfully synthesized Zn3WN4 in space group Pmn21 (Figure 4).In contrast, prior thin film sputtering work produced cation-disordered P63mc structures. [2,31,62]While both the Pmn21 and P63mc structures are wurtzite-derived, the cation-ordered structure is expected to be the thermodynamic ground state. [2,27]In thin film sputtering, high-energy plasma precursors deposit onto a substrate and quench rapidly in a local energy minimum, thus locking in the disordered cation arrangement. [27]hile bulk syntheses can sometimes lead to cation-disordered structures, [7,11,47,48,51] the high charge on W (6+) likely encourages ordering to maximize the spacing between the hexavalent cations.The kinetics of our bulk metathesis reactions here proceed in a way that avoids the local energy minimum of the disordered structure, instead forming a (mostly) ordered structure.
The degree of cation ordering in our Zn3WN4 sample cannot be precisely determined from our current SPXRD data.While the SPXRD results suggest that Zn3WN4 was synthesized in a (mostly) cation ordered form (Figure 4), our optical spectroscopy results suggest that some degree of cation disorder may be present (Figure 5).Notably, the same batch of Zn3WN4 was used to create samples for the diffuse reflectance optical spectroscopy, magnetometry, XRF, and high resolution SPXRD measurements.The single-phase model includes some site disorder, leading to a refined composition of Zn3.17W0.83N4and a good statistical fit to the SPXRD pattern (Rwp = 3.989%).Attempts to model the pattern with two phases (a cation-ordered Pmn21 and a cation-disordered P63mc) yield similar-quality fits as the single-phase fit (Figure S6).The two-phase refinements suggest the powder may be approximately 10-20 mol% P63mc.These two-phase models are consistent with our optical data, as disordered P63mc Zn3WN4 likely exhibits lower-energy absorption (ca.2.5 eV) [31] compared to cation-ordered Pmn21 Zn3WN4, which has a predicted bandgap of ca.4.0 eV. [28]However, the small difference in statistical fit for the best two-phase model (Rwp = 3.742%) relative to the best single-phase model (Rwp = 3.989%) suggests that determining the precise cation order of the material is non-trivial (see the Supplemental Information for more discussion).In sum, these results indicate that our batch of cation-ordered Zn3WN4 exhibits some degree of disorder or inhomogeneity on the order of 10-20 mol%.Although our Zn3WN4 sample is not perfectly ordered, it exhibits a substantially larger degree of cation ordering than prior thin film work. [31,62]

Structural relation of precursor and product
The synthesis reported here is distinct from literature on prior ion exchange syntheses of nitrides in that the ions undergoing exchange have different formal charges.All prior reports on nitrides exchanged ions of the same charge (e.g., displacing Na + with Cu + in ATaN2, or Ca 2+ with Mg 2+ in A2Si5N8; where A represents the exchangeable cations). [49,52,53]Here, we replace a monovalent cation (Li + ) with a divalent cation (Zn 2+ ).While such exchange has been conducted in oxides [65,66] and sulfides (e.g., 2 NaCrS2 + MgCl2 à MgCr2S4 + 2 NaCl), [67] to the best of our knowledge this is the first report of such an exchange in nitrides.The resulting decrease in the cation:anion ratio (from 7:4 to 4:4) means that a truly topotactic replacement is unlikely to occur.However, the transformation appears to be near-topotactic.The transformation from Li6WN4 to Zn3WN4 involves slight structural rearrangements (Figure 6).The W 6+ retains its tetrahedral coordination and (for the most part) its oxidation state through the process, but the orientations of the polyhedra change.The fcc anion lattice of Li6WN4 converts to the hcp anion lattice of Zn3WN4.During this change, half of the W 6+ ions in Li6WN4 migrate through an octahedral site to a new tetrahedral site in Zn3WN4 (red annotations).The anion packing layers also decrease in spacing from 2.767(1) Å in Li6WN4 to 2.662(1) Å in Zn3WN4.The shortest W-W distance decreases from 4.927(1) Å to 4.646(1) Å. Lastly, the centrosymmetric structure of Li6WN4 (P42/nmc) converts to a polar structure of Zn3WN4 (Pmn21).We did not observe any signs of solid solution behavior (i.e., Li6-xZnx/2WN4) in the in situ SPXRD studies.However, solid solution behavior may be present but undetected by the in situ SPXRD data if it occurs on short timescales (<30 s) or small length scales (ca 10 nm).

Reaction pathway
Our results demonstrate that the halide anion X in ZnX2 exerts a powerful influence over the metathesis reaction thermodynamics and kinetics of Zn-containing ternary nitrides.In situ SPXRD and DSC measurements reveal that the phase purity of the product is correlated with the melting points of the phases and the reaction energy (Figure 7).The ZnF2-based reaction is highly exothermic, and ZnF2 has a high melting point.These two factors result in high-temperature reactivity and partial Zn3WN4 phase decomposition.In contrast, ZnCl2 and ZnBr2 react at much lower temperatures (owing to the formation of liquid phases) and release less energy during the reaction.However, the slightly lower melting point of ZnCl2 compared to ZnBr2 combined with the slightly higher ∆Hrxn of the respective reaction leads to a runaway exothermic event (Figure 3) and slight Zn3WN4 decomposition (Figure S4).Overall, the ZnBr2-based reaction is optimal for the synthesis of Zn3WN4 because of the low ∆Hrxn and the low melting point of ZnBr2, but further optimization for phase purity may be needed in future research on other materials.
Figure 7.The initial formation of Zn3WN4 (as detected by in situ SPXRD) is correlated with the melting point of the halide salt but is not correlated with calculated reaction enthalpy (∆Hrxn, color).The dashed line shows where melting point would equal onset temperature.
This type of reaction control has been explored in oxides but has not previously been detailed for ternary nitrides.][70][71] In particular, work on "co-metathesis" identified that when eutectic halide mixtures form in situ, these liquids decrease reactant onset temperatures relative to systems without eutectics. [65,71]Similar eutectics are likely forming between ZnX2 and LiX in our syntheses of Zn3WN4.

Generalizability to other materials
There are numerous Zn-M-N phases that have been demonstrated to be synthesizable via thin film sputtering but that have not yet been made in bulk.In addition to Zn3WN4, [2,62] sputtering has been used to synthesize fully nitridized transition metal ternaries: ZnTiN2, ZnZrN2, Zn2VN3, Zn2NbN3, Zn2TaN3, and Zn3MoN4. [2,3,26,27,31]Although computational predictions for these thin film materials find that cation-ordered structures are the thermodynamic ground state (Figure 4f), [72] these sputtered films tend to form in cation-disordered structure variants (Figure 4e). [2,3]This disorder tends to decrease the bandgap of the material by creating localized electronic states. [8,73]ulk syntheses of these materials could advance the development of these new semiconductors by studying the effect of structure (e.g., ordering) on optoelectronic properties of these new materials.
The synthesis of Zn3WN4 from Li6WN4 and ZnX2 suggests a promising strategy for future materials discovery of cation-ordered heterovalent ternary nitrides via metathesis from lithiumbased ternary nitride precursors.Lithium-based ternary nitrides are the most well-studied subset of ternary nitrides, [2] suggesting that many Li-M-N phases exist that could be used to synthesize additional A-M-N phases via exchange with AXn (where A and M are metals and X is a halide).Following our results here, X should be selected to minimize reaction energy and thus minimize the risk of decomposing the target phase via gaseous N2 loss.To demonstrate this point, we also synthesized Zn3MoN4 from Li6MoN4 and ZnBr2 (Figure S11).Zn3MoN4 was the main product, but some decomposition products were also observed, indicating that additional reaction optimization is needed.Unlike in Zn3WN4, the ZnBr2-based reaction was not sufficiently lowenergy to avoid this decomposition for Zn3MoN4.While we were not able to synthesize phasepure Zn3MoN4 here, further reaction engineering, like adding NH4Cl to manage heat flow, [42,43,45,46] may be able to produce phase-pure Zn3MoN4.As we found that the reaction onset temperature is correlated with AX2 melting point, phases with high-melting temperature precursors may be difficult to synthesize below the decomposition point of the targeted ternary.Therefore, future work should consider ways to decouple the reaction onset from the AX2 melting point.In sum, this work shows how Li-M-N phases can be promising precursors for accelerating the discovery of new ternary nitrides.

Conclusions
Here, we report the bulk synthesis of cation-ordered Zn3WN4, through ion exchange reactions beginning from Li-based ternary nitride precursors: Li6WN4 + ZnX2 à Zn3WN4 + 6 LiX (X = Br, Cl, F).These reactions proceed directly (i.e., without intermediates), as measured by in situ synchrotron powder X-ray diffraction and differential scanning calorimetry.The reaction onset temperature correlates with the melting point of the ZnX2 precursor, allowing ZnCl2-and ZnBr2based reactions to proceed at ≤300 °C.The more exothermic reactions lead to greater degrees of Zn3WN4 decomposition, meaning that the least exothermic reaction (with ZnBr2) is the most favorable for synthesis.High resolution synchrotron powder X-ray diffraction data is consistent with cation-ordered Zn3WN4 (Pmn21).This finding is distinct from prior thin film syntheses, which yielded cation-disordered P63mc Zn3WN4.Diffuse reflectance spectroscopy shows that Zn3WN4 powders exhibit absorption onsets near 2.5 eV and 4.0 eV, suggesting that a small amount of a cation-disordered P63mc phase of Zn3WN4 may be mixed in with the cation-ordered Pmn21 phase.Preliminary work targeting Zn3MoN4 from Li6MoN4 and ZnBr2 suggests this synthesis approach may readily extend to other systems.These findings indicates that Li-M-N compounds may serve as precursors for synthesizing numerous other ternary nitrides.
Li6WN4 was synthesized in a method modified from that of Yuan et al. [1] Solid precursors (2.1 Li3N + W, ca. 5 mol% excess Li3N to account for loss by evaporation) were ground with a mortar and pestle and loaded into Zr crucibles with Zr lids (ca. 1 g loose powder).The Zr crucibles were then loaded into sacrificial quartz tubes (open on one end), which were loaded into quartz process tubes and heated in a tube furnace.Custom endcaps with quick disconnects enabled air-free transfer from the glovebox to the tube furnace (under Ar or N2).The samples were reacted under flowing N2 (50 sccm, 99.999% purity) with a +5 °C/min ramp followed by a 12 h dwell at 850 °C and then natural cooling after turning off the furnace.Samples were recovered into the glovebox for subsequent analysis and use.The beige-colored products were confirmed to be phase pure by powder X-ray diffraction (Figure S1).
Syntheses for Zn3WN4 were conducted by grinding together Li6WN4 with ZnX2 (X = Cl, Br) in a ratio of approximately 1:3.The powders were pelletized with 6 mm diameter dies in an arbor press (ca. 100 mg per pellet), loaded into quartz ampules (10 mm OD, 10 mm ID, ca 10 cm length), sealed under vacuum (<0.03 Torr), and heated in a muffle furnace.The optimized synthesis for Zn3WN4 used a ratio of Li6WN4 + 3.1 ZnBr2 and was scaled up to a 3 g batch, sealed in a quartz ampule under vacuum (< 0.03 Torr) and heated at +5 °C/min to 300 °C for a 1 h dwell, then allowed to cool naturally.Samples were recovered into the glovebox.Reaction byproduct LiX was washed away using anhydrous and degassed methanol that was dried over molecular sieves.For washing, centrifuge tubes were loaded with approximately 500 mg of product powder and 1.5 mL methanol.The tube was agitated with a vortex, centrifuged, and the supernatant was decanted.This wash was repeated for a total of 3 cycles.Recovered powders were dried overnight under vacuum.However, Zn3WN4 ultimately proved to be stable against air and water, and we note that the anhydrous washing may not be necessary.

Synthesis of Zn3MoN4
Just like Li6WN4, Li6MoN4 was synthesized using Li3N and Mo (≥99.9%,1-5 micron powder, Sigma Aldrich), following a method modified from that of Yuan et al. [1] Heating the powders at 850 °C for 12 h (with slight Li3N excess) resulted in a phase pure Li6MoN4 (Figure S1b).This Li6MoN4 was then mixed with ZnBr2, pelletized, sealed in an ampule under vacuum, and heated at +5 °C/min to 300 °C for a 1 h dwell, followed by natural cooling.The product was then washed with anhydrous methanol.

In situ synchrotron powder X-ray diffraction analysis
In situ synchrotron powder X-ray diffraction (SPXRD) measurements were conducted at beamline 17-BM-B of the Advanced Photon Source at Argonne National Laboratory.For these experiments (λ = 0.24101 Å), the PerkinElmer plate detector was positioned 700 mm away from the sample.Homogenized precursors were packed into quartz capillaries in an Ar glovebox and flame-sealed under vacuum (<30 mTorr).Capillaries were loaded into a flow-cell apparatus [3] and heated at 5 °C/min to the specified temperature.A thermocouple was placed against the tip of the sample capillary, approximately 2 mm horizontally from the position of the X-ray beam.Diffraction pattern images were collected every 30 s by summing 20 exposures of 0.5 s each (10 s of summed exposure), followed by 20 s of deadtime.Images collected from the plate detector were radially integrated using GSAS-II and calibrated using a silicon standard.
Sequential Rietveld refinements were conducted on in situ SXPRD datasets using TOPAS Professional v6. [4]Lattice parameters, background terms, and scale factors were refined for each phase as a function of temperature, while atomic coordinates and occupancies were held constant at the initial values of the reference structure.A weighted scale factor (W.S.F.) Q was calculated for each phase p as a product of scale factor S, cell volume V, and cell mass M: Qp = Sp•Vp•Wp. [5]e note that amorphous and liquid phases are inherently not observed in powder diffraction measurements and therefore cannot be accurately included in this analysis.A Lorentzian size broadening term was refined for each phase to model the peak shape using the pattern showing the greatest intensity of the relevant phase; this term was then fixed for the sequential refinements to better account for changes in intensity.To help stabilize the sequential refinement, isotropic displacement parameters (Biso) were fixed at 1 Å 2 for all atoms, but we note that this is likely not physical for a variable temperature investigation.

Ex situ powder X-ray diffraction analysis of Zn3WN4
The products of all reactions were characterized by powder X-ray diffraction (PXRD).Laboratory X-ray diffraction patterns were collected on a Rigaku Ultima IV diffractometer with Cu Kα X-ray radiation at room temperature.All samples were initially prepared for PXRD measurements inside the glovebox; powder was placed on off-axis cut silicon single crystal wafers to reduce background scattering and then covered with polyimide tape to impede exposure to atmosphere.After Zn3WN4 was determined to be moderately air stable, PXRD patterns were collected without polyimide tape to decrease the background signal.
High resolution synchrotron powder X-ray diffraction (SPXRD) measurements were conducted at beamline 28-ID-2 of the National Synchrotron Light Source II (λ = 0.1821 Å) at Brookhaven National Laboratory.Samples were sealed under vacuum in quartz capillaries, which were then nested in Kapton capillaries.Data were collected for 60 seconds at T = 25 °C while spinning.Scattered photon intensity was measured using a Perkin-Elmer XRD 1621 Digital Imaging Detector.The data were reduced using Dioptas. [6]Pawley fits and subsequent Rietveld refinements were conducted using TOPAS Academic. [4]etveld refinements were conducted for the laboratory PXRD and SPXRD patterns using TOPAS and TOPAS Academic, v6 (Bruker AXS). [4]Reference structures were sourced from the Inorganic Crystal Structure Database (ICSD).The Pmn21 Zn3MoN4 structure (ICSD Col. Code 255744) was used as a starting model for Pmn21 Zn3WN4, with the Mo replaced by W. [7] For the cation disordered P63mc Zn3WN4 structure, P63mc ZnO was used as a starting model, with atomic occupancies adjusted to match the stoichiometry of Zn3WN4, and lattice parameters adjusted to match the SPXRD pattern.For each structure, lattice parameters, isotropic displacement parameters, and general atomic coordinates were refined.For some models, Zn and W occupancy were refined as detailed in the Discussion section and the Supplemental (Figure S7).Structural visualizations and reference PXRD patterns were generated using VESTA. [8]mpositional, thermodynamic, and property measurements The composition of nominal Zn3WN4 was measured by X-ray Fluorescence spectroscopy (XRF) and combustion analysis.Cation composition was quantified by XRF using a Bruker M4 Tornado with a Rh X-ray source.Samples were pelletized and XRF spectra were collected at 4 points across the pellet.Zn and W ratios were quantified from each spectra using the Bruker M4 software.
Differential scanning calorimetry (DSC) experiments were conducted using a Q20 system from TA Instruments.Samples were prepared in an Ar-filled glovebox.Samples (ca. 10 mg) were loaded into aluminum pans, which were crimped closed with an aluminum lid.The reference pan was also crimped closed under argon.Pans were then transferred out of the glovebox for measurement, and data were collected upon ramping up to 400 °C at a rate of 10 °C/min.Thermodynamic calculations for reaction enthalpies (∆Hrxn) were conducted using formation enthalpy (∆Hf) values reported in the Materials Project database. [9,10]-vis measurements were conducted on a Cary 6000 UV-Vis-NIR spectrometer.PTFE was used as a white reflectance standard.Absorbance was calculated with the Kubelka-Munk transformation, k/s = (1− R) 2 / 2R (where R is the reflectance, k is the apparent absorption coefficient, and s is the apparent scattering coefficient).
DC susceptibility data were measured on a Quantum Design Physical Properties Measurement System (PPMS) from T = 2 to 305 K in applied fields up to μ0H = 14 T.

In situ variable temperature SPXRD measurements
Figures S2 and S3 show the in situ variable temperature SPXRD heatmaps for the ZnCl2 and ZnF2 reactions, respectively.The analogous ZnBr2 reaction is shown in the main text (Figure 1).These data were used for sequential Rietveld analysis, which is presented in Figure 4 in the main text.
In situ SPXRD of the 3 ZnCl2 + Li6WN4 reaction (Figure S2) proceeds similarly to the ZnBr2based reaction shown in the main text (Figure 1).Bragg peaks arising from the precursors (3 ZnCl2 + Li6WN4) stay steady up to 200 °C, where they begin to decrease in intensity.Simultaneously, Bragg peaks for Zn3WN4, LiCl, and Li2ZnCl4 begin to increase in intensity.The set of Bragg peaks corresponding to the Li2ZnCl4 phase fade out by 310 °C.We suspect that Li2ZnCl4 is not an essential intermediate but rather the product of a transient side reaction between the precursor ZnCl2 and product LiCl.The intensity of the Bragg peaks corresponding to the LiCl phase reaches a maximum near ca.340 °C, and then slowly decreases in intensity up to 550 °C (LiCl melting point is 605 °C). [11]Zn3WN4's Bragg peaks remain approximately constant in intensity above 300 °C, persisting through the duration of the heating process.These processes are consistent with the following reactions: Figure S3 shows that the 3 ZnF2 + Li6WN4 reaction does not yield Zn3WN4 until 565 °C, a much higher temperature than the ZnCl2 and ZnBr2 reactions.Initial reactivity begins near 310 °C, indicated by a decrease in intensities for the set of Bragg peaks corresponding to Li6WN4 and ZnF2.This initial reactivity is well below the melting point of ZnF2 (872 °C). [11]Concurrent with this initial reaction, an unknown phase briefly grows in (between 310 °C and 404 °C).Extrapolating from the ZnBr2 and ZnCl2 reactions, the phase is likely a Li-Zn-F intermediate, but ternary fluorides in this space are poorly characterized.Li2ZnF4 has been reported, but the structure is not well described and the unit cell does not match the unknown phase.This intensity of the Bragg peaks arising from the intermediate phase decreases to zero by 404 °C, above which Li6WN4 is the only crystalline phase up to 565 °C.At this point, Li6WN4 fades out and Zn3WN4, LiF, and a rocksalt phase fit with WN grow in.The presence of this rocksalt phase indicates that the higher reaction temperature and greater exothermicity of the ZnF2 reaction (compared with the Cl and Br versions) leads to a substantial degree of decomposition of the Zn3WN4 phase.Therefore, this reaction was not explored further.Reference patterns for the reactants and products/intermediates are simulated at the bottom and top, respectively (ICSD Col. Codes 2459 for ZnCl2, 66096 for Li6WN4, 402399 for Li2ZnCl4, and 27981 for LiCl). [1]igure S3.Heatmap of in situ SPXRD data upon heating 3 ZnF2 + Li6WN4 at +10 °C/min.Reference patterns for the reactants and products/intermediates are simulated at the bottom and top, respectively.(ICSD Col. Codes 9169 for ZnF2, 66096 for Li6WN4, 18012 for LiF). [1]vidence of decomposition in the 3 ZnCl2 + Li6WN4 reaction Figure S4 shows that a trace Zn impurity can be detected in reactions between Li6WN4 + 3.1 ZnCl2, even when heated at low temperatures (ca.250 °C).This impurity makes the powder appear grey in color.The Zn is likely produced via decomposition of Zn3WN4 during the highly exothermic reaction (See DSC measurements, Figure 5).Surprisingly, we do not observe W or WN.This may indicate that tungsten remains in the Zn3WN4 phase (which would then be Zn-poor), or that the W (or WN) is amorphous.

Compositional characterization
X-ray Fluorescence (XRF) spectroscopy identified a Zn:W ratio of 3.8:1, in excess of the expected 3:1 ratio for Zn3WN4.XRF was conducted on pelletized powder after washing away the bromide byproduct.The excess zinc may be incorporated in the wurtzite-derived lattice, as suggested by Rietveld analysis (discussed below).A representative raw spectrum is shown in Figure S5.Spectra were collected at 4 different points across the pellet, and each spectrum was fit using the Bruker software to quantify Zn and W atomic ratios.The Zn:W ratio of 3.8:1 was calculated by averaging the Zn:W values from the 4 spectra.
Figure S5.Representative XRF spectrum from the surface of a Zn3WN4 pellet.The inset shows the characteristic X-ray lines used in the fitting software.

Structural models for the high resolution SPXRD measurements
The refined lattice parameters for Zn3WN4 are shown in Tables S1.We considered several structural models of Zn3WN4 when conducting Rietveld analysis against our high resolution SPXRD patterns (Table S2).Several terms were allowed to vary for each approach: sample displacement, lattice parameters, size broadening (Lorentzian), strain broadening (Lorentzian), isotropic displacement parameters, and a 15-term background Chebyshev polynomial.Our most robust model was a single-phase model that allowed for a small degree of cation disorder for Pmn21 Zn3WN4, but with each cation site fixed to full occupancy (e.g., the Zn1 site was refined with Zn occupancy set to 1-x and W occupancy x).This model resulted in an Rwp of 3.989 % and is shown in Figure 4, Table S1, and Figure S6a.For comparison, a simpler model of Pmn21 Zn3WN4 with fixed cation occupancies (e.g., the Zn1 site fixed with 1.0 Zn occupancy) gave a significantly worse fit to the data (Rwp = 4.638 %).Atomic positions were allowed to refine for both these single phase models.However, our diffuse reflectance spectroscopy measurements suggest that the material is not a single homogeneous phase.
Given the two distinct absorption onsets shown in the diffuse reflectance spectrum (Figure 5), we also considered a two-phase model in our Rietveld refinements.For the first two-phase model, we started with the Pmn21 Zn3WN4 from the fixed cation occupancy models.We then fixed atomic positions.Next, we created a model for cation disordered Zn3WN4 in a P63mc structure (i.e., the wurtzite structure type), and set the lattice parameters to a = 3.280 Å and c = 5.324 Å such that the (100), (002), and (101) reflections of the P63mc structure matched the (210), (002), and (211) reflections, respectively, of the Pmn21 structure (i.e., adis = 0.5aord and cdis = cord; where "dis" and "ord" indicate the disordered P63mc and the ordered Pmn21 structures, respectively).This structure is consistent with the cation-disordered Zn3WN4 synthesized via thin film sputtering. [13]We then refined the size and strain broadening for both phases.This refinement resulted in 78 mol% phase fraction of Pmn21, 22 mol% for P63mc, and an Rwp value of 3.953 % (Figure S7a), comparable to the single-phase model.
The best fit was obtained via a two-phase model, but with non-physical lattice parameters.
Allowing the P63mc lattice parameters to freely refine results in the best fit we obtained by the Rietveld method (Rwp = 3.742 %).However, the model is likely non-physical.In this model, the refined c lattice parameter for this P63mc phase increases substantially (5.451(2) Å) compared to the Pmn21 phase (c = 5.3228(2) Å), which is not consistent with prior studies of order-disorder transitions in wurtzite derived structures (e.g., ZnGeN2). [14]The c lattice parameters of ZnGeN2 are identical in the Pna21 (the ordered structure) and P63mc (the disordered structure), because cation disorder does not affect the layer spacing of the hcp anions along the (00l) direction.This analysis reveals ambiguities in these two-phase models.
Given the limitations of the two-phase models, we posit that the single phase model provides the most reliable fit to the SPXRD data without over-fitting the pattern. [15]Yet, the SPXRD measurements probe the long range average ordered structure.Local ordering-which we do not probe here-may influence the optical absorption properties shown in Figure 5.The impact of local ordering on optical properties has been characterized in the halide perovskite CsSnBr3, [16] in Fe doped SrTiO3, [17] and in carbon coated FeF3. [18]ble S2 4), and b) fixed metal site occupancy at cation-ordered Zn3WN4.Qualitative inspection of the 2D diffraction images show that the diffraction rings for the Zn3WN4 powder are homogeneous (Figure S8).If rings with two different morphologies were present, this would suggest the presence of two distinct phases with different crystallinity, size, and strain.That we see only one morphology of ring in the 2D detector image supports either a single phase or multiple phases with nearly identical crystallinity, size, and strain.In our Rietveld refinement that modeled the data using two phases, the fit to the data is significantly worse when we constrain the size and strain broadening terms to be the same for both phases.These findings support our use of the single-phase model.

Magnetic susceptibility measurements
Magnetic measurements were performed on Zn3WN4 using a Quantum Design Physical Property Measurement System (PPMS).A powder sample of Zn3WN4 was loaded into a small packet (0.001125 mg) and secured inside a plastic straw for the measurement.Magnetic susceptibility (χ) of Zn3WN4 was measured as a function of temperature shows largely diamagnetic behavior with a trace paramagnetic impurity (Figure S9a).Similarly, magnetization (M) as a function of applied field (H) at 2 K shows that diamagnetism dominates the field-dependent magnetization (Figure S9b).These findings are inconsistent with pure Zn3WN4, which should be purely diamagnetic.Zn impurities, if present, would also give a diamagnetic response.The paramagnetic component suggests the possibility of a reduced tungsten species (e.g., W 5+ ), possibly as a sub-nitride (e.g., Zn3WN4-δ), an oxynitride impurity (e.g., Zn3WN4-xOx), or a W-rich phase (e.g., Zn3-δW1+δN4).Full structure visualizations of Li6WN4 and Zn3WN4

Synthesis of Zn3MoN4
The synthesis strategy used for Zn3WN4 was also applied to synthesize Zn3MoN4, but the product exhibited partial decomposition.Analysis of the SPXRD pattern collected for the Mo analog of Li6WN4, synthesized via the reaction Li6MoN4 + 3 ZnBr2 à Zn3MoN4 + 6 LiBr, suggests phase decomposition (Figure S11).In addition to the desired Zn3MoN4 (56 mol%), Rietveld analysis of High resolution SPXRD data show that a rocksalt (RS) structure fit as ZnMoN0.5 also forms (23 mol%), along with a Zn impurity (20 mol%).This RS phase exhibits a substantially larger lattice parameter (a = 4.7106(3) Å) than the defect-RS phase Mo2N (a = 4.16 Å to 4.19 Å). [19,20] Therefore, we hypothesize it may be a (Zn,Mo)Nx material, as octahedra Zn 2+ has a substantially larger ionic radius (0.74 Å) than octahedral Mo x+ (0.65 Å for Mo 4+ , 0.69 Å for Mo 3+ ). [21]The rocksalt ZnMoN0.5 phase was created from a Fm3 # m Mo2N starting model.Rietveld analysis with the composition of ZnMoN0.5 provides a reasonable fit.Further analysis of this material is beyond the scope of this manuscript.Additional minor peaks that we have not indexed are present (possibly higher order oxides).These impurity phases suggest that Zn3MoN4 is less stable at elevated temperatures than Zn3WN4.This decomposition occurs despite the excess ZnBr2 which was intended to serve as a heat sink during the exothermic reaction.Despite the partial decomposition of the phase, Zn3MoN4 is still the major phase in the pattern.As with Zn3WN4, the SPXRD pattern for Zn3MoN4 shows evidence of cation-ordering in the form of the Pmn21 reflections at low angle: e.g., (010), ( 110), (101), (011).However, these reflections are weaker than in the W case, owing to the lower scattering factor of Mo compared to W. We focused our work on Zn3WN4 because W scatters Xrays more strongly than Mo (facilitating characterization) and because our Zn3WN4 products exhibited higher phase purity.

Figure 3 .
Figure 3. DSC measurements of Li6WN4 reacting with a) 3 ZnBr2 and b) 3 ZnCl2, with inset photos showing the products from reactions heated at 300 °C for 1 h.

Figure 4 .
Figure 4. a) High resolution SPXRD pattern and Rietveld refinement of Zn3WN4 powder.Simulated patterns are shown for reference: b) the cation-disordered P63mc model and c) the fully ordered Pmn21 model.Visualizations of d) the Pmn21 structure of Zn3WN4 refined from the SPXRD data, e) the cation-disordered P63mc model, and f) the fully ordered Pmn21 model.

Figure 6 .
Figure 6.Arrangement of [WN4] tetrahedral units when looking down a) on one layer of Li6WN4 (P42/nmc) in the (201) plane, and b) on a layer of Zn3WN4 (Pmn21) in the (001) plane.Red annotations in (a) show the displacement undergone by some W 6+ ions in the transition to Zn3WN4.Side views of three layers of c) Li6WN4 and d) Zn3WN4 stacked along the [201] and [001] directions, respectively.The shading of the [WN4] units in (c) indicates depth (fainter tetrahedra are farther away from the viewer).Li and Zn are omitted for clarity.Representations with Li and Zn are shown in the Supplemental Information (Figure S10).

Figure S4 .
Figure S4.Laboratory PXRD measurement of the product of Li6WN4 + 3.1 ZnCl2 heated at +5 °C/min to 250 °C for a 10 h dwell.

Figure S7 .
Figure S7.Rietveld refinement of the high-resolution SPXRD data of Zn3WN4 using a twocomponent model with a) the P63mc lattice parameters fixed relative to the Pmn21 values and b) the P63mc lattice parameters freely refined.

Figure
Figure S9.A) Magnetic susceptibility (χ) of Zn3WN4 as a function of temperature with an applied field (µ0H) of 1 T. b) Magnetization (M) as a function of µ0H at 2 K.The grey outline in (b) is shown in more detail via the inset.

Figure S10 .
Figure S10.Reaction scheme showing the structure of Li6WN4 and Zn3WN4, with Li and Zn atoms included to complement the main text visualization (Figure 6).

Li 6 WN 4 Figure
Figure S11.SPXRD pattern and Rietveld refinement of the washed products from the reaction between Li6MoN4 + 4.2 ZnBr2 (excess ZnBr2).

Table S1
. Summary of models considered for the high resolution SPXRD data for Zn3WN4.