Bi₂Fe₄O₉: Structural changes from nano- to micro-crystalline state

2.1 Synthesis

Samples were prepared by a polyol-mediated synthesis. In detail, 5 mmol Bi(NO₃)₃·5H₂O (Sigma-Aldrich, ≥98 %), 10 mmol Fe(NO₃)₃·9H₂O (Sigma-Aldrich, ≥98 %) and 40 mmol NaOH (VWR Chemicals, 99.2 %) were dissolved in 42 ml diethylene glycol (DEG) (AppliChem, 99 %). The mixture was heated in a three-necked flask under reflux using a metal bath at 423 K. After 2 h 5 mmol stearic acid was added. After cooling down to ambient conditions, the solid product was separated by centrifugation and washed several times with acetone with intermediate centrifugation steps. For comparison samples were also synthesized hydrothermally at 473 K for 24 h in Teflon coated steel autoclaves. Two series of samples were produced. First the molar ratio of Bi(NO₃)₃·5H₂O to Fe(NO₃)₃·9H₂O was varied from 1:2 to 2:2 in steps of 0.2 using always 10 mmol Fe(NO₃)₃·9H₂O in 20 mL of a 2 mol L^–1 NaOH solution. Thereafter a molar 1:1 ratio was used increasing the NaOH concentration up to 12 mol L^–1 in steps of 2 mol L^–1. After the heat treatment the samples where intensively washed with deionized water. This process was repeated three times before the samples were dried at 393 K for about 12 h.

2.2 Spectroscopy

The FTIR spectra were measured on a Bruker IFS 66v/S spectrometer using the standard KBr method (1 mg sample in 200 mg KBr) between 370 and 4000 cm^–1. Background as well as sample spectra were obtained from 128 scans each with a spectral resolution of approx. 1 cm^–1. The mode positions were determined by taking the point of the maximum intensity. The temperature-dependent ⁵⁷Fe Mössbauer spectra were taken between 293 and 973 K in an atmosphere of flowing synthetic air using a standard transmission Mössbauer spectrometer (Halder) in the sinusoidal driving mode employing a ⁵⁷Co/Rh γ-radiation source with a maximum activity of 1.91 GBq. The velocity scale was calibrated with an α-Fe absorber at room temperature and the isomer shifts (IS) are stated relative to the center of this calibration.

2.3 X-ray diffraction

XRPD data were collected on a X’Pert MPD PRO diffraction system (PANalytical GmbH, Almelo, The Netherlands) equipped with Ni-filtered CuK_α1,2 radiation (λ_α1,2 = 0.15406 nm, 0.15444 nm), a 0.25° divergence slit, a 0.5° antiscatter slit, a 0.04 rad soller slit in the primary beam and a X’Celerator detector system in the secondary beam in Bragg-Brentano geometry. Room-temperature data were measured between 5 and 85° 2θ with a step width of 0.0167° 2θ and a measurement time of 20 s per step. A HTK 1200N heating chamber (Anton Paar, Graz, Austria) was used for temperature-dependent measurements. Data were collected from 5 to 100° 2θ with a step width of 0.0167° 2θ and a measurement time of 75 s per step. The temperature was increased stepwise from 300 to 1120 K in 20 K steps with an equilibration time of 5 min. The obtained data were refined using the Rietveld method (topas V4.2, Bruker AXS, Karlsruhe, Germany). For the profile description the fundamental parameter approach was used, where the fundamental parameters were fitted against a LaB₆ standard material.

2.4 Scanning electron microscopy

Scanning electron microscopy (SEM) was carried out on a JSM-6510 (JEOL GmbH, Munich, Germany) equipped with energy dispersive X-ray (EDX) analysis facilities and an XFlash Detector 410-M (Bruker AXS GmbH, Karlsruhe, Germany). To obtain quality data the samples were sputtered with a thin film of gold having a thickness of approx. 10 nm. EDX spectra were collected using an excitation voltage of 20 kV.

3.1 Synthesis

The synthesis of Bi₂Fe₄O₉ using the hydrothermal method was described by Wang et al. [29]; the process strongly depends on the hydroxide concentration during the heat treatment at 373 K. The authors used a multi-step synthesis in which the metal nitrates were first dissolved in nitric acid, which was then brought, by adding slowly dropwise a KOH solution, to a pH = 8 leading to a precipitation of a brown solid. This solid was filtered, washed and transferred into a NaOH solution which was then treated hydrothermally. With a NaOH concentration of 2 mol L^–1 they observed the formation of a perovskite phase. Note that the authors [29] took this as the beginning of the crystallization of Bi₂Fe₄O₉, but from the reported diffraction pattern the formation of a perovskite is obvious. With increasing concentration the mullite-type phase was formed. To make the synthesis easier, we simply added the metal nitrates to an aqueous NaOH solution and kept it at 473 K for 24 h. For a first experiment a 2 mol L^–1 NaOH solution with a Bi to Fe ratio of 1.2:2 was used to form the Bi₂Fe₄O₉ mullite-type phase. Rietveld refinements of XRPD data confirmed that the product consists of 67 wt% BiFeO₃, 24 wt% Fe₂O₃ and 9 wt% Bi₂Fe₄O₉. However, using a ratio of 1.4:2 as much as 93 wt% Bi₂Fe₄O₉ was formed together with 7 wt% BiFeO₃. A further increase of the molar ratio to 1.6:2 and 1.8:2 leads to pure well crystallized Bi₂Fe₄O₉ phases as shown in Fig. 2. The latter sample was used for further investigations. Increasing the cation ratio further to 1:1 and increasing the NaOH concentration at this ratio to 6, 9 or 12 mol L^–1 leads to an increase of the average crystallite size, however, along with the formation of a few percent of the sillenite phase Bi₂₅FeO₃₉ [30].

Fig. 2:

Scanning electron micrographs of Bi₂Fe₄O₉ synthesized by the hydrothermal method (a), by the polyol method (b) and the polyol sample obtained after the XRD heating measurements, showing Bi₂Fe₄O₉ rods and needles (c and d).

The synthesis of nanoscale metal oxides using the polyol method was described by several authors [28, 31, 32]. The metal precursors were heated in a high-boiling alcohol such as most commonly used diethylene glycol (DEG) [31], ethylene glycol [22], triethylene glycol [33] and tetraethylene glycol [33] at elevated temperatures. Using this method various metal oxides can be produced showing particle sizes below ~100 nm [31, 33, 34]. However, to produce a crystalline sample a subsequent heat treatment is often required [19, 31, 35]. In this study, metal nitrates were chosen as metal precursors. They were dissolved in DEG in a three-necked flask followed by addition of 40 mmol NaOH. The temperature of the mixture was then ramped to the target reflux temperature of 423 K, concomitantly releasing NO_x gas and forming a brown precipitate at 388 K. Notably, the reflux (423 K) started far below the boiling point of DEG (519 K) owing to the high content of hydrate water in the metal nitrates. The as-synthesized particles are spherical in shape and possess a size distribution ranging from 200 to 700 nm as shown by SEM micrograph analysis (Fig. 2). XRPD gave no clear Bragg reflections. However, well crystalline mullite-type Bi₂Fe₄O₉ could be produced via a subsequent heat treatment. Whereas the hydrothermal synthesis nicely produced crystalline mullite-type material the structure of the polyol synthesized material could hardly be characterized due to the very small average crystallite size leading to very broad diffraction reflections as for an X-ray amorphous material.

3.2 Spectroscopy

3.2.1 Infrared spectroscopy

The infrared spectrum of the hydrothermally synthesized sample corresponds to that reported earlier for Bi₂Fe₄O₉ [36], whereas numerous different absorption bands are visible in the FTIR spectrum of the Bi₂Fe₄O₉ nano-material synthesized by the polyol method as shown in Fig. 3. Most of them are resulting from the (mainly organic) compounds used in the synthesis. The broad absorption bands at 3200 and 1650 cm^–1 can be assigned to water in the sample [37]. The NO₃^– anions are evidenced by the sharp absorption feature at 1385 cm^–1. The C–H stretching bands of stearic acid are represented by modes between 2850 and 2920 cm^–1 [38], whereas the respective carboxyl groups could be identified by their modes at 1670 cm^–1. Remaining DEG (after washing the samples with acetone) leads to the observation of C–O–C, C–OH and C–O vibrations with peak maxima at 1125(3), 1060(3) and 900(3) cm^–1, respectively [39]. Octahedral Fe–O stretching vibrations in bismuth ferrates are located in the region between 445 and 548 cm^–1 (BiFeO₃ [40]) and between 437 and 471 cm^–1 (Bi₂Fe₄O₉ [36]). The appearance of absorption peaks around 500 cm^–1 of the Bi₂Fe₄O₉ nano-material could therefore be assigned to the Fe–O stretching of FeO₆ octahedra. The assignments of the band positions of the vibrational modes of the hydrothermally synthesized mullite-type phase (Fig. 3) agree well with those of Voll et al. [36]. In the spectrum of this sample an additional strong mode with the maximum intensity at 812(1) cm^–1 belongs to the Fe–O–Fe stretching mode of the Fe₂O₇ double tetrahedral unit interconnecting the octahedral chains (as calculated for the isotypic aluminum compound [41]), which is absent in the polyol Bi₂Fe₄O₉ nano-material material.

Fig. 3:

Infrared spectra of Bi₂Fe₄O₉ synthesized by the hydrothermal method (a) and precursor formed by the polyol method (b), diethylene glycol (DEG) (c), Bi(NO₃)₃·5H₂O (d), and stearic acid (e).

3.2.2 ⁵⁷Fe Mössbauer spectroscopy

To get more insight into the coordination of the iron atoms in the synthesized samples, ⁵⁷Fe Mössbauer data were collected at different temperatures. The fit of the room-temperature ⁵⁷Fe Mössbauer spectrum of the polyol-synthesized sample (only polyol from now on) reveals the existence of two Fe³⁺ species in two different octahedrally coordinated sites, Table 1. The majority component shows exactly the same isomer shift (IS) as the octahedrally coordinated Fe site in Bi₂Fe₄O₉ [15]; thus identical coordination and similar bond lengths are expected. The minority component shows a slightly higher IS, indicating a lower electron charge density at the nucleus, pointing to slightly larger bond lengths. The quadrupole splitting (QS) of the two octahedrally coordinated Fe sites in the polyol material is significantly larger than that of the octahedrally coordinated site in Bi₂Fe₄O₉ [15]. In this context, it is interesting to compare the present results with those of nano-crystalline BiFeO₃. Park et al. [4] have studied the Mössbauer spectra of BiFeO₃ perovskites with crystallite sizes between <14 nm and about 100 nm. The present Mössbauer spectroscopic parameters exhibit some similarity with those of the sample with <14 nm (Table 1), for example in respect to the relative signal area, indicating a similar distribution of iron over two inequivalent sites in the structure. As seen, also quadrupolar interactions take similarly large values indicating significant coordinative distortion in both materials. However, IS values of the nano-perovskite are significantly lower, indicating a higher electron charge density at the nucleus than in the polyol material. The large room-temperature linewidth (Γ = 0.4 mm s^–1, Table 1) of the polyol material provides evidence of some kind of disorder around the two Fe-occupied sites. In conclusion, the polyol material appears to be of disordered nature and to possess an unknown local structure with two non-equivalent octahedral sites for Fe³⁺ occurring in an approximate 1:3 ratio. According to these signal intensities the material could, for instance, represent a (Bi₂Fe)Fe₃O₉ perovskite.

Table 1:

Hyperfine parameters of room-temperature ⁵⁷Fe Mössbauer spectra of Bi₂Fe₄O₉ powder prepared by the polyol method, as compared to Bi₂Fe₄O₉ synthesized by the glycerine method [15] and to bulk and nano BiFeO₃ [4].

Sample	IS 1/mm s–1	QS 1/mm s–1	A 1/%	IS2/mm s–1	QS 2/mm s–1	A 2/%	Γ/mm s–1
polyol	0.38(1)	1.09(1)	27(1)	0.36(1)	0.70(1)	73(1)	0.40(1)
hydrothermal	0.24(1)	0.95(1)	29(1)	0.36(1)	0.38(1)	29(1)	0.25(1)
Bi2Fe4O9 [15]	0.23(1)	0.95(1)	50(1)	0.35(1)	0.37(1)	50(1)	0.22(1)
BiFeO3 [4] bulk	0.39(3)	–0.10(5)a	47(2)	0.38(3)	0.34(5)	53(2)	0.35(2)
BiFeO3 [4] <14 nm	0.31(3)	1.29(5)	25(2)	0.33(3)	0.79(5)	73(2)	0.23(2)

^aData from the magnetic bulk material represent the quadrupolar perturbation parameter ε = QS·(3cos²θ–1)/2, where θ is the angle between the direction of the principal component of the electric field gradient and the direction of the local magnetic field.

Temperature-dependent Mössbauer spectra of the hydrothermal and polyol samples are shown in Fig. 4. Whereas the thermal evolution of the ⁵⁷Fe hyperfine parameters in Bi₂Fe₄O₉ prepared by the hydrothermal method is exactly the same as reported for bulk Bi₂Fe₄O₉ [15]_, the parameters for the polyol sample show a different behavior. During heating the spectral features change between 573 and 773 K, showing evidence for structural changes leading to the appearance of two differently coordinated iron sites. The two sites of octahedral coordination of Fe³⁺ in (Bi₂Fe)Fe₃O₉ transform into one tetrahedral and one octahedral site. Finally, at 973 K the spectrum is identical to that of Bi₂Fe₄O₉ prepared by the hydrothermal method. The temperature dependence of isomer shifts (IS) and quadrupole splittings (QS) of Bi₂Fe₄O₉ prepared by the polyol method is presented in Fig. 5.

Fig. 4:

Temperature-dependent ⁵⁷Fe Mössbauer spectra of Bi₂Fe₄O₉ powder materials prepared by the polyol (right) and hydrothermal method (left) in flowing synthetic air. The subspectrum shown in red is attributed to a small admixture of Fe₂O₃.

Fig. 5:

Temperature-dependent isomer shifts (IS) and quadrupole splittings (QS) of Bi₂Fe₄O₉ prepared by the hydrothermal (left) and polyol method (right).

3.3 Temperature-dependent X-ray powder diffraction

The in-situ heating XRPD experiment of the as-synthesized polyol material showed a transformation from an X-ray amorphous powder into a rhombohedral perovskite-type structure at 680 K followed by a second transformation into Bi₂Fe₄O₉ starting from 800 K on as shown in Fig. 6. Above 920 K the perovskite structure cannot be detected anymore. The sharp reflections which appear at diffraction angles greater than 57° 2θ above 620 K are attributed to the sample holder due to mechanical shrinking of the powder sample.

Fig. 6:

Temperature-dependent diffraction patterns of the polyol material showing the transformation from as-synthesized amorphous powder into rhombohedral BiFeO₃ and orthorhombic Bi₂Fe₄O₉ (different scale range above 900 K for better visibility).

Rietveld refinements of the temperature-dependent data of the polyol material were carried out using the structure model of Moreau et al. [42] for BiFeO₃ with R3c symmetry and lattice parameters of a = 558.76(3) pm and c = 1386.7(1) pm in the hexagonal setting. Taking the results of the ⁵⁷Fe Mössbauer study into account, Fe³⁺ and Bi³⁺ were both refined on the same 6a position (0, 0, 0(fixed)) whereas the second 6a position (0, 0, 0.2215(18) at 780 K) was taken as solely occupied by iron ions. The oxygen atoms were found on the 18b position (0.421(11), 0.011(8), 0.952(5) at 780 K). With increasing temperature the (Bi_1–xFe_x)FeO₃ crystal structure was observed to be rapidly changing as depicted in Fig. 7. The lattice parameter a increases from 561.4(4) pm (at 740 K) to 562.0(2) pm at 780 K and then decreases to 561.62(6) pm at 860 K as shown in Fig. 8. The lattice parameter c of (Bi_1–xFe_x)FeO₃ is significantly smaller than the one reported for BiFeO₃ [42], supporting the site co-shared by Fe³⁺ and Bi³⁺. Accordingly, the shrinking of the lattice parameter can be explained in terms of the smaller ionic radius of Fe³⁺ [43]. Starting from 740 K, the lattice parameter c increases linearly with temperature from 1375(2) pm and approaches the value reported for BiFeO₃ [42] at about 800 K (c = 1383(1) pm). Notably, this is the temperature at which the lattice parameter a as well as the strain, as shown in Fig. 9, decrease rapidly. The unit cell volume increases linearly with temperature. At 860 K, reflections appear (marked with crosses in Fig. 7) that can be attributed to Fe-bearing sillenite Bi₂₅FeO₃₉ in space group I23 [30]. Above 920 K, neither reflections of sillenite nor of the perovskite structure can be detected.

Fig. 7:

Left panel: Temperature-dependent XRPD patterns from 680 to 920 K in 20 K steps showing the evolution of (Bi_1–xFe_x)FeO₃ and Bi₂Fe₄O₉ phases. Some intense reflections of the BiFeO₃ (104) (110) and Bi₂Fe₄O₉ phase (121) (211) are indicated by diamonds and stars, respectively. Reflections marked by a cross can be attributed to Bi₂₅FeO₃₉. Right panel: Magnified feature of a part of the left panel, showing the appearance and disappearance of reflections marked by the cross.

Fig. 8:

Temperature-dependent lattice parameter a (left), c (middle) and cell volume (right) of (Bi_1–xFe_x)FeO₃.

Fig. 9:

Temperature-dependent strain (left) and occupancy of Bi³⁺ (right) of the (Bi_1–xFe_x)FeO₃ phase.

Substitution of Fe³⁺ with an effective ionic radius of 64.5 pm [43] on the Bi position causes a high strain in the unit cell due to the much larger ionic radius of Bi³⁺ 103 pm [43]. The micro-strain at 740 K could be refined to as much as 0.41(1) % giving further evidence that Bi³⁺ is partially substituted by Fe³⁺. The micro-strain is reduced to about 0.09(1) % at 860 K. The Rietveld refinement results in an occupancy of 69(2) % for Bi³⁺ at 740 K, the rest is occupied by Fe³⁺ according to the constrained occupancy parameter refinements assuming a fully occupied position. This value fits well to the Mössbauer spectroscopic results, which have suggested 1/3 of the occupancy probability. With increasing temperature the amount of Bi³⁺ on the position is increasing (Fig. 9).

After completion of the temperature-dependent XRD experiments the samples were investigated by scanning electron microscopy. Needle-like rods and plates were identified as shown in the SEM micrograph of Fig. 2. It is assumed that the long chains of the stearic acid present in the precursor caused the pronounced unidirectional growth. The sample contains larger crystals as well as very thin needle-like plates (Fig. 2). The length of the thin needles ranges from 5 to 10 μm, the width takes values of about 250 nm. Due to their transparency in SEM micrographs these needles are assumed to be only a few nm thick. EDX analyses revealed a molar ratio of Bi:Fe of 1:2.1(2) as expected for Bi₂Fe₄O₉. Nevertheless, a remaining content of 3.3(3) mol% Na was found, but we assume, based on the substitution experience with this structure type (e.g. [44]), that sodium is not incorporated in the mullite-type structure.