Preparation of iron(IV) nitridoferrate Ca4FeN4 through azide-mediated oxidation under high-pressure conditions

Transition metal nitrides are an important class of materials with applications as abrasives, semiconductors, superconductors, Li-ion conductors, and thermoelectrics. However, high oxidation states are difficult to attain as the oxidative potential of dinitrogen is limited by its high thermodynamic stability and chemical inertness. Here we present a versatile synthesis route using azide-mediated oxidation under pressure that is used to prepare the highly oxidised ternary nitride Ca4FeN4 containing Fe4+ ions. This nitridometallate features trigonal-planar [FeN3]5− anions with low-spin Fe4+ and antiferromagnetic ordering below a Neel temperature of 25 K, which are characterised by neutron diffraction, 57Fe-Mössbauer and magnetisation measurements. Azide-mediated high-pressure synthesis opens a way to the discovery of highly oxidised nitrides.


Synthesis
Ca4FeN4. The title compound was prepared following the stoichiometry of equation (1) from calcium nitride Ca3N2 (99.9 %, AlfaAesar), iron nitride Fe2N, and sodium azide NaN3 (99.99%, Sigma-Aldrich). Reactions were carried out between 4 and 8 GPa and between 800 and 1200 °C, and optimum synthesis conditions were found to be 6 GPa and ca. 1200 °C with 60/300/120 minutes of ramp up, dwell, and ramp down, respectively. Under the inert conditions of an Ar-filled glovebox (c(O2/H2O) < 1 ppm), all compounds were ground and transferred into a copper capsule (0.025 mm thickness, 99.999 %, Puratronic®, AlfaAesar), which was then placed in a h-BN crucible that was subsequently placed inside the sample octahedron. The reaction conditions were achieved by a Voggenreither 1000 t large volume press (Mainleus, Germany) with multianvil technique employing an 18/11 octahedron-within-cubes payload. The pressure medium consisted of a Cr2O3-doped MgO-octahedron (Ceramic Subtrates & Components, Isle of Wight, U.K) with 18mm edgelength. Heating was enabled by resistance heating using two graphite sleeves (Schunk, Heuchelheim, Germany) to minimize the temperature gradient. Additional information about the setup can be found in literature. 1 Caution: NaN3 is highly toxic and can undergo rapid decomposition when exposed to high temperatures.
Approximately 50 mg samples were obtained from each experiment, and these were assessed by powder X-ray diffractogram as shown in Supplementary Fig. 3. One sample was used for single crystal diffraction, EDX, and magnetisation measurements, another sample was used for neutron and powder X-ray diffraction, and several samples were combined for 57 Fe-Mössbauer measurements. This and possible part-decomposition due to moisture sensitivity, accounts for the varying impurity contribution seen in different measurements, as discussed later.
Fe2N. The starting material Fe2N has been prepared by reaction of Fe metal powder with a constant flow of dried ammonia (5.0, Air Liquide) at 500 °C. The temperature ramps were 5 °C/min and dwell 20 h. The ammonolysis was repeated twice with intermittent grinding of the obtained grey-metallic powder until the X-ray diffraction pattern showed only single-phase Fe2N.

Diffraction
Single-crystal X-ray diffraction data of Ca4FeN4 were obtained with a Bruker D8 Venture diffractometer on single crystals mounted in glass capillaries (Hilgenberg, Malsfeld, Germany) under dried paraffin oil to prevent hydrolysis. Data collection, indexing, data reduction, and absorption correction were carried out with the APEX3 software. 2 Analysis of systematic absent reflections and space group determination was supported by the XPREP software. 3 Structure solution and refinement was carried out with the SHELX software implemented in WINGX. 4,5 Structures were visualized with VESTA. 6 Further information regarding the crystal structure can be obtained through the joint CCDC/FIZ Karlsruhe inorganic crystal structure database by quoting number CSD 2015297.
Powder diffraction was carried out with a STOE StadiP diffractometer equipped with a DECTRIS MYTHEN 1K Si-strip detector in modified Debye-Scherrer geometry. Samples were sealed in glass capillaries with an inner diameter of 0.5 mm (Hilgenberg, Malsfeld, Germany). Owing to the Fe-content of the sample, a Mo X-ray source was used with a Ge(111) monochromator singling out the Mo-Kα1 radiation. Data were collected in the range from 2 < 2θ < 71° and Rietveld refinement was carried out with the Topas Academic V4.1 software. 7 The background was handled with a shifted-Chebychev function, peak profiles with a fundamental parameter approach, and preferred orientation with spherical harmonics. Positions of the heavier atoms Fe and Ca were refined as well as the displacement parameters.
Neutron powder diffraction was carried out at the WISH beamline of the ISIS Neutron and Muon Source at temperatures of 1.5 and 50 K. 8 For the refinement, data collected between 12.1 and 1.4 Å on four detector banks were used. Rietveld refinement and magnetic structure solution and refinement were performed with GSAS-II. 9 The background was handled with shifted-Chebychev functions fitted to manually set points owing to strongly curved background and peak overlap. The background was subtracted before visualization of the fits in the figures. The magnetic structure was displayed with VESTA. 6

Magnetization curves M(H,T) were recorded with a Quantum Design Physical Properties
Measurement System (PPMS). Powdered samples of Ca4FeN4 were packed into polyethylene capsules and sealed with glue to prevent decomposition of the sample owing to moisture. Isothermal magnetization curves in the range of ±50 kOe were recorded at temperatures of 300, 40, 19, and 2 K, while the temperature dependent susceptibility was obtained at 30 kOe in the range of 2 to 300 K. The obtained magnetizations were corrected for the diamagnetic contribution of the capsule.

57 Fe-Mössbauer
57 Fe-Mössbauer measurements were performed in transmission geometry with a constant-acceleration spectrometer using a 57 Co/Rh radiation source. The velocity scale and the isomer shift were determined with the relative values of α-Fe at room temperature. The spectra were fitted with Lorentzian functions by using the standard least-squares method. A small unidentified impurity was detected in the 57 Fe-Mössbauer spectra, which probably is related to the one detected by PXRD and magnetic measurements.

Scanning electron microscopy
A Zeiss EVO-Ma 10 scanning electron microscope was used for obtaining micrographs and electron energy dispersive (EDX) measurements. Selected crystallizes of the sample were placed on conductive and adhesive carbon tape and quickly inserted into the electron microscope, which was equipped with a field emission gun run at 15 keV and a Bruker X-Flash 410-M detector. Data were analysed with the QUANTAX 200 software package.
Oxygen content was not taken into account owing to the short exposure in air and consequential hydrolysis. Samples were not sputtered owing to the hydrolysis. 6

Single-crystal diffraction
The dataset of Ca4FeN4 was indexed in an I-centred orthorhombic unit cell with pseudotetragonal metric a = 6.903 (2) Table 6) with respect to a regular octahedron and thus are a contrast to the more ordered region of the first and second layer, which is also mirrored by relatively large displacement ellipsoids of Ca3. 12 Sandwiching layer 1 by layers 2 and 3 yields the smallest repeat unit of the structure (marked in Figure 1b). The automorphism group of this repeat unit contains an inversion centre by which layers 2 and 3 on top and bottom are related to each other. The repeat units themselves are stacked along c by inversion, or by the a-glide reflections perpendicular to the c-axis.

Additional discussion of the bonding situation
The  13,14,15,16,17,18,19 The qualitative d-orbital splitting is shown in Figure 1d. While weak π-donors like oxides and halides usually lead to high-spin configuration, it was shown by ab initio calculations on [M III N3] 6− complex anions with M III = V, Cr, Fe that the strong π-bonding character of N leads to high energies of the π-antibonding e'' and e' orbitals, thus enabling low-spin configuration. [17][18][19] Owing to an even lower electron repulsion and higher crystal field splitting in Fe IV than in Fe III , a low-spin state is also assumed for [FeN3] 5− complex anions leading to occupied a'1 and half-occupied e' orbitals and spin of S = 1 (Figure 1d). Ca133Mn216N260, which were prepared by gas-solid reaction at ambient pressure. 24

Details on neutron powder diffraction
Owing to a maximum resolution of ca 1.1 Å obtained with the four detector banks at 50 K, the atom positions and displacement parameters obtained from refinement might not be reliable. Moreover, preferred orientation and bad overall crystallinity of the small platelike crystallites complicated the refinement. Hence, the low temperature nuclear structure models were compared with the room temperature single-crystal structure with the COMPSTRU tool of the Bilbao crystallographic server, which indicated only slight structural changes (Supplementary Table 8 The determination of Bravais-class and subsequent Shubnikov-group was complicated by the pseudo-tetragonal metric of the unit cell, leading to overlapping reflections as indicated in Figure 4a. The (1 0 0) reflection gave the worse fits as its theoretical position did not perfectly coincide with the observed reflection. A magnetic ordering parallel to the (1 0 0) planes was thus ruled out in favour of the better fitting (0 1 0) reflection leading to magnetic ordering parallel to the bc-plane. Such an ordering could be realized in several Shubnikov-groups but the fit in group Pbc'a gave the best fit. Pbc'a is nonstandard setting of Pb'ca and was chosen to retain the nuclear structure setting.

Impurity effects
Impurity contributions are seen in the powder diffraction, magnetic, and Mössbauer data, and it is important to estimate their influence in particular on the reported magnetic properties.
Powder X-ray and neutron diffraction data contain unidentified impurity peaks up to 13% ( Fig. S3) and 12 % (Fig. S4) of the maximum Ca4FeN4 peak integrated intensity, which serve as crude estimates of impurity phase proportion. Impurity effects in the magnetization data (Figs. 2, S6 and S7) are, however, rather small. A linear Curie-Weiss fit ( Fig. 2) and constant Curie-Weiss paramagnetic moment (Fig. S7) close to the expected value for S = 1 Fe 4+ are observed over a wide temperature range (80-300 K) showing that no large amounts of paramagnetic impurity are present. An observed low temperature paramagnetic tail is consistent with no more than 3% of a S = ½ impurity (Fig. S7), and M-H loops show that only a trace of ferromagnetic impurity is present, equivalent to ~0.01% Fe metal (Fig. S6). Taken together, these results indicate that the secondary phases observed in the powder diffraction data are mainly non-magnetic, in keeping with the 4:1 Ca:Fe ratio of metals and also Na in the bulk sample.
The Mössbauer sample is inconsistent with the magnetisation results as 14 atom-% of an Fe impurity is observed (Fig. 3), and any Fe-based phase would be likely to be para-or ferro-magnetic. This may be due to mixing of products of varying purity to make the Mössbauer sample and perhaps also partial decomposition of the highly air-and moisture-sensitive Ca4FeN4 during sample transfers or transport from Europe to Japan for this measurement. Inclusion of a 14 atom-% impurity doublet in the fits to the low temperature Mössbauer data enabled good fits of the magnetic sextet from Ca4FeN4 to be obtained.
The major magnetic features in the susceptibility, magnetic neutron and Mössbauer data may thus be assigned to Ca4FeN4 with high confidence, despite the presence of unidentified secondary phases.