Introduction

The search for unusual oxidation states fuels explorative inorganic chemistry for materials discovery1. In solid-state materials of 3d transition metals, the highest oxidation states are attained in oxides and fluorides, for example, M2FeVIO4 with M = Li, Na and Cs2CuIVF6, while solid-state nitrides feature substantially lower states, especially with the later 3d metals as in Li3FeIIIN3 or BaCuIN2,3,4,5,6. This disparity in observed oxidation states illustrates severe synthetic difficulties stemming from the physical properties of nitrogen. The low chemical potential of N2 reflects the very stable N ≡ N triple bond (944.8 kJ/mol) and nitrogen’s relatively low electronegativity and positive first electron affinity (EA(N) = +0.07 eV compared to EA(O) = −1.46 eV), so nitrides generally have lower free energies of formation, less ionic bonding, and a tendency to lose N2 at elevated temperatures7,8,9,10. As a result, the number of known 3d metal nitrides is small compared to that for oxides, despite much interest in their properties and applications9. Hence, high-throughput and data mining studies have been carried out to help discover unknown nitrides9,11,12,13, and synthetic advances have been made by exploiting the increase in chemical potential of N2 under pressure14, as exemplified in the recent diamond anvil cell (DAC) synthesis of MN2 (M = TiIV, FeIII to CuI/II) pernitrides and FeIIN4 featuring polymeric nitrogen chains15,16,17,18,19.

The large group of ternary transition metal nitrides AxMyNz, in which A is an electropositive metal, also have a rich variety of electronic, physical, and magnetic properties20,21,22,23,24,25,26,27,28, but their synthesis suffers from the limitation that the starting materials, e.g. binary transition metal nitrides or pure metals, usually feature lower oxidation states than the targeted compounds, for example, Mn4N vs. Ca6MnIIIN529. As recently described in a review by Salamat et al.30, the reactivity of N2 at ambient to moderate pressures restricts access to high oxidation states, in particular for nitrides of the later 3d metals Mn to Cu25,27,31,32.

In this work, we devise an adaptable synthesis route to highly oxidised nitridometallates by employing a standard large-volume press to generate high-pressure and high-temperature conditions with sodium azide NaN3 as a powerful solid-state nitriding agent for use in a sealed reaction crucible. The method is illustrated by a straightforward preparation of the previously-unreported calcium nitridoferrate(IV) Ca4FeN4 starting from accessible binary reagents Fe2N and Ca3N2. While in molecular chemistry, high valent iron(IV, V, VI) nitrido complexes have already been investigated owed to their relation to biocatalytic and environmental catalytic processes33,34,35,36,37,38,39, solid-state nitridoferrates prepared by direct combination reactions were reported with lower oxidation states +I to +III5,40,41,42. In the here presented reaction, sodium may act as a flux facilitating single-crystal growth and the large-volume-press enables the preparation of sufficient quantities of material (yield ca. 50 mg per experiment) for characterisation with 57Fe-Mössbauer spectroscopy, magnetometry, and neutron diffraction.

Results

Synthesis and chemical analysis

Ca4FeN4 was prepared according to Eq. (1) at 6 GPa and ca. 1200 °C in a multianvil large-volume press and crystallises as a black microcrystalline powder containing plate-like crystals up to 30 μm in size (Supplementary Fig. 5):

$$8{\rm{Ca}}_3{\rm{N}}_2 + 3{\rm{Fe}}_2{\rm{N}} + 3{\rm{NaN}}_3 \to 6{\rm{Ca}}_4{\rm{FeN}}_4 + 3{\rm{Na}} + 2{\rm{N}}_2.$$
(1)

As expected for nitridometallates, Ca4FeN4 is very sensitive to moisture and quickly hydrolyses forming the respective metal hydroxides and ammonia. To ensure a complete oxidation of Fe2N, the reaction was carried out with a surplus of NaN3. Moreover, heating ramps of minimum 1 h were optimal for synthesis of pure samples, as faster heating led to large amounts of impurities, and dwell times of 5 h were required for single-crystal growth (see Supplementary Methods). The equation suggests Na formation, however, powder X-ray diffraction revealed unidentified byproduct(s) (Supplementary Fig. 3) but no Na metal. Energy dispersive X-ray (EDX) spectroscopy of the sample (15 datapoints, normalised on Ca, Supplementary Table 9) resulted in an experimental composition of Ca4.0(4)Fe1.1(1)N4.1(6), which fits the sum formula determined by single-crystal diffraction without detectable Na substitution. The incorporation of Na into the crystal structure by substitution of Ca would necessitate either ~50% of FeV or ~25% FeVI or additional incorporation of equal amounts of O to retain FeIV. These can be ruled out as 57Fe-Mössbauer data show only one isomer shift at large negative velocity and neutron powder diffraction refinement of the N occupancy ruled out additional incorporation of O (see later sections). EDX revealed Na with a fraction of up to 5 at-%, close to the theoretical Na content from the reaction equation (5.5 at-%). This might either be finely dispersed sodium metal, undetected in the PXRD data, or a constituent of a Na–Ca/Fe–N byproduct.

Structure discussion

The crystal structure of Ca4FeN4 (space group Ibca, a = 6.903(2), b = 6.919(3), and c = 22.552(8) Å) was solved and refined from single-crystal X-ray diffraction data. Details are in the Supplementary Discussion. The structure contains three Ca and one Fe position and can be described by a relatively dense packing of face-, edge-, and corner-sharing distorted CaN6 octahedra (Fig. 1a,b), in which trigonal-planar [FeIVN3]5− units (Fig. 1c) are embedded via edge-sharing. Fe forms a double-layered quasi-2D tetragonal sublattice (Supplementary Fig. 1) with short Fe–Fe distances (4.89 and 4.93 Å) within the layers and long ones (8.49 Å) between. The crystal field exerted on the Fe4+ ions (qualitatively displayed in Fig. 1d) is expected to result in low-spin configuration with S = 1 owing to strong dπ–pπ Fe–N bonding in accordance with existing studies on monoatomic double-faced π-donors in general and on nitrido-ligands particularly43,44,45. A detailed discussion of the structure, d-orbital splitting and spin-state of Fe are enclosed in the Supplementary Discussion, while experimental verification follows in the later sections through magnetisation, 57Fe-Mössbauer, and neutron diffraction measurements.

Fig. 1: Structure of Ca4FeN4.
figure 1

a Formal construction of the repeat unit, which can be subdivided into three layers. Layer 1 is formed by Ca1N6 octahedra and [FeN3]5− polygons, layer 2 is formed by Ca2N6 octahedra, and layer 3 by Ca3N6 octahedra. The repeat unit is created by stacking of the layers above and below layer 1, with an inversion centre in the middle of layer 1. b Structure of Ca4FeN4 with inequivalent Ca positions coloured differently, highlighting the repeat unit. c [FeN3]5− complex anion showing bond lengths (in Å) and N–Fe–N bond angles (in deg.). Ellipsoids at 95 % probability. d Qualitative d-orbital splitting vs. energy for the [FeN3]5− complex anion with double-faced π-donor N-ligands and d4 electron configuration with S = 1 (adapted from literature)43.

Full size image

As no electronic instability exists, the slight distortion towards Y-conformation in the [FeN3]5− complex anion (Fig. 1c, dFe–N = 1.731(6), 1.731(6), 1.758(7) Å, Fe–N–Fe angles 121.1(2)°, 121.1(2)°, 117.8(4)°) probably stems from the surrounding crystal structure rather than a Jahn-Teller distortion44. Ca4FeN4 exhibits shorter Fe–N bond lengths than the known Ca nitridoferrate(III) Ca6FeIIIN5 with dFe–N = 1.769(15) Å but similar bond lengths are observed in nitridoferrates(III) of the higher homologues, Sr3FeN3 and Ba3FeN3 with dFe–N = 1.73(1) Å, which feature a larger inductive effect, shortening the Fe−N bonds42,46,47. The range of observed Ca–N distances (2.43 < dCa–N < 2.82 Å, Supplementary Fig. 2 and Supplementary Table 4) is in good agreement with values found in other Ca nitridoferrates like Ca2FeN2 and Ca6FeN5 (2.33 < dCa–N < 2.74 Å) and Ca nitridometallates like CaTiN2, Ca4TiN4, and Ca3CrN3 (2.37 < dCa–N < 2.85 Å)20,23,41,47,48.

Bond valence sum (BVS) calculations (Supplementary Table 10) led to calculated valences of VFe1 = 4.15, VCa1 = 2.08, VCa2 = 1.85, VCa3 = 1.65 corroborating a high valence for the iron atom but also showing that Ca2 and Ca3 are underbonded, which might be related to metal-metal second nearest neighbour interactions between closely situated Ca2 and Ca3 atoms (dCa2–Ca2 = 2.997(3) Å, dCa2–Ca3 = 3.129(2), 3.272(2) Å)49,50. Further discussion of the bonding situation can be found in the Supplementary Discussion.

Magnetometry

Susceptibility measurements on Ca4FeN4 were carried out in a field of 30 kOe with zero-field-cooling for the range from 2 to 300 K and also in field-cooled mode from 2 to 50 K (Fig. 2). The susceptibility curve shows paramagnetic behaviour down to approximately 50 K and a sudden decrease at 25 K consistent with an antiferromagnetic transition. The upturn observed below 7 K is likely a Curie tail from a paramagnetic impurity, with effective moment <0.3 μB (see Supplementary Discussion). A Curie–Weiss fit to the paramagnetic regime of the susceptibility from 80 to 300 K resulted in an effective paramagnetic moment of μeff = 3.08(1) μB and a Weiss-constant of Θ = −123(1) K. The negative value of the Weiss-constant is consistent with the observed antiferromagnetic ordering. However, the observed transition temperature is much lower than |Θ|, which is possibly due to low magnetic dimensionality originating from the quasi-2D tetragonal Fe sublattice (Supplementary Fig. S1). The observed paramagnetic moment is slightly greater than the ideal spin-only value of 2.83 μB for a spin S = 1 system, in keeping with the 2.95 μB value observed for a hexahydrazide cage iron(IV) complex, and hence corroborates the low-spin state in the trigonal-planar [FeIVN3]5− units33.

Fig. 2: Temperature dependent molar susceptibility χmol of Ca4FeN4.
figure 2

Zero-field cooled and field cooled data are shown. The Curie–Weiss fit was performed for the temperature range of 80 to 300 K. The inset shows that inverse molar susceptibility χmol−1 varies linearly above 75 K, highlighting that a single dominant Curie–Weiss paramagnetic phase is present.

Full size image

Isothermal magnetisation curves (Supplementary Fig. 6) show only a small hysteresis due to impurities, equivalent to ~0.01% Fe metal, and confirm that the ground state of Ca4FeN4 is antiferromagnetic.

Mössbauer spectroscopy

To further corroborate the FeIV oxidation state, 57Fe-Mössbauer spectra of Ca4FeN4 (Fig. 3) were collected at 297, 50, and 10 K. The RT and 50 K spectra, which are above the antiferromagnetic transition temperature, show doublets at large negative velocity of δ297K = −0.65 mm/s and δ50K = −0.55 mm/s. Similar isomer shifts have been observed in FeV containing oxides like La2LiFeO6 (δ=−0.41 mm/s) and Fe-doped K3MnO4 (δ=−0.55 mm/s)51,52. The isomer shift is, however, strongly related to covalency of the Fe–ligand bond as well as the coordination environment, both influencing the s-electron density at the nucleus53. Ab initio calculations on Ba3[FeIIIN3] with δ263K = −0.55 mm/s of FeIII showed that trigonal-planar coordination leads to strong covalent σ-type Fe–N bonding filling the 4s-orbital while dπ−pπ backbonding depletes the electron density in the d-orbitals resulting in a reduced shielding effect45. The observed isomer shift in Ca4FeN4 of δ297K = −0.65 mm/s can thus be rationalised from the higher oxidation state FeIV as well as a similar bonding situation as in Ba3[FeIIIN3]. The oxidation state FeIV in nitrides has only previously been reported in mixed FeIII/FeIV materials Li3−x[FeN2] observed during topotactic electrochemical redox deintercalation of Li3[FeIIIN2] for which in operando Mössbauer spectroscopy revealed isomer shifts of up to δ = −0.33 mm/s54.

Fig. 3: 57Fe Mössbauer spectra of Ca4FeN4.
figure 3

Data (blue circles) were collected at 297, 50 and 10 K, orange curve belongs to FeIV of Ca4FeN4, green curve belongs to an unidentified byproduct, grey dashed line is the sum of the individual curves. Green byproduct doublet appears at isomer shifts δ297K = 0.22, δ50K = 0.08, δ10K = −0.6 mm/s (intensity ratio Fe4+/Feimpurity = 85.8/14.2 was fixed in the low temperature fits using the 297 K value).

Full size image

The quadrupole splitting observed in the 297 K and 50 K spectra of Ca4FeN4 (Δ297 K = 0.98 mm/s, Δ50K = 1.00 mm/s) is expected for the non-cubic coordination environment. At 10 K, the Mössbauer spectrum shows sextet peaks in line with the antiferromagnetic ordering at 25 K observed in the susceptibility data. The fit of the sextet resulted in δ10K = −0.52 K, Δ10 K = 0.31 mm/s and a hyperfine splitting of 16.7 T. The increase in isomer shift with decreasing temperature is probably owed to the second-order Doppler effect53.

Neutron diffraction

Neutron diffraction patterns were collected at 50 and 1.5 K, above and below the 25 K antiferromagnetic transition temperature of Ca4FeN4. A low temperature nuclear structural model was established by Rietveld refinement of the 50 K data and then transferred to the 1.5 K data (Supplementary Fig. 4, Supplementary Table 7). Atom positions were freely refined, and displacement parameters of equal atom types were constrained. Tentative refinement of possible N/O mixed anion positions indicated complete N-occupation. Results of the refinements are summarised in the Supplementary Discussion.

Three magnetic reflections are observed in the diffraction pattern obtained at 1.5 K (Fig. 4a). The difference 1.5–50 K diffractogram was used for magnetic structure refinement as one magnetic reflection overlapped with a small reflection from a byproduct. The magnetic reflections were indexed on a unit cell of the same dimensions as the nuclear one but with space group Pbca (space group no. 61), which is a maximal subgroup of the nuclear model’s space group Ibca (no. 73). The best fit was achieved with refinement in magnetic group Pbc’a (a non-standard setting of Pb’ca) of the crystallographic Bravais-class (see Supplementary Discussion). The spins of the resulting antiferromagnetic structure (Fig. 4b) are oriented parallel to the crystallographic bc-plane and the observed magnetic moment μ = 1.92(10) μB (components μy = 1.36(10) μB, and μz = 1.36(3) μB) is in good agreement with the ideal saturated moment of 2S = 2 μB.

Fig. 4: Magnetic structure of Ca4FeN4.
figure 4

a Magnetic structure refinement carried out on the difference of data collected at 1.5 and 50 K of detector bank 1. Blue crosses are datapoints, orange is the fit of the magnetic structure, green is the difference curve. Positions of reflections are marked by vertical drop lines, their hkl indices displayed above the reflection. b Magnetic structure shown in two different directions. [FeN3]5− polygons in orange, CaN6 polyhedra in blue.

Full size image

We report the synthesis of Ca4FeN4 with Fe in high oxidation state +IV via high pressure and high-temperature azide-mediated oxidation and its characterisation through 57Fe-Mössbauer, magnetisation, and neutron diffraction measurements. High pressures may be a necessity for the synthesis of Ca4FeN4 as bond valence sum calculations indicated metal-metal repulsion between adjacent Ca atoms, which might hinder the formation of this structure at lower pressures. At ambient pressure, the structure might be stabilised through the electron inductive effect of Ca. The crystal field splitting in the [FeN3]5− anions do not result in an electronic instability in the low-spin Fe4+ d-orbitals and thus probably are stabilised with respect to the disproportionation into Fe3+ and Fe5+, which is observed in CaFeO3 and other high valent iron perovskites55. Susceptibility and neutron diffraction measurements do not suggest the presence of ligand holes (d5L) as often observed for FeIV compounds56, which might be destabilised through lower lying p-orbital levels of N when compared to O.

The formation of Ca4FeN4 seemingly resembles a Na-flux reaction as crystallisation is facilitated through longer reaction times but the oxidation mechanism to Fe4+ is in question32. Achieving this high oxidation state certainly requires a strong nitriding agent, however, the behaviour of NaN3 under high pressure and temperature conditions has not been elucidated. NaN3 could decompose into Na and N2 or the pressure could prevent the decomposition and lead to the stabilisation of the highly reactive azide itself or its partial decomposition product the diazenide, as reported for decomposition of M(N3)2 (M = Ca, Sr, Ba) under pressure57.

The investigation of the here described oxidation process will probably require in situ experiments. However, the already applicable advantages of this route for the synthesis of nitrides are a highly nitriding environment in a large volume press, Na probably acting as a flux facilitating crystal growth, and large sample quantities available for property characterisation. Moreover, it does not rely on specific properties like the ability to deintercalate Li, which gave the only previous report of a Fe4+ nitride in the Li3−x[FeN2] system54. The NaN3-route overcomes the challenging thermodynamics of nitride chemistry and enables the direct synthesis of a highly oxidised transition metal nitride, which was not obtained from standard ambient and medium pressure methods nor with DACs at extreme N2 pressures (>30 GPa) and temperatures (>2000 K)19,27,32. This route is thus expected to significantly advance nitride research by accessing new materials in high transition metal oxidation states and consequent new chemical and physical properties.