2.2. LaFeO3 and Ca2Fe2O5-Based Nanocomposites
For LaFeO
3 as a starting material and hexagonal BN as a reducing compound, the A-site cation AR (ACAR)-promoted exsolution can be written as reaction (5) (for clarity purposes, along with the nominal pseudo-reactions (6) and (7)):
Herein, the processes involving reaction (5) were carried out at elevated temperatures of 700–750 °C, in a reducing atmosphere, provided by a 10% H
2/Ar flow. Although hydrogen is not formally involved in the reaction, its role here is to maintain low pO
2 conditions, essential for the existence of metal Fe nanoparticles. The formation of lanthanum borate LaBO
3 provides an additional driving force for the reduction of LaFeO
3 and decreases the reaction temperature compared to a reduction with hydrogen. The specified temperatures of reaction (5) are below the decomposition temperature of LaFeO
3 in a hydrogen-containing atmosphere, which is reported to be >850 °C [
27,
28]. Thus, its decomposition, described by reaction (3), does not occur under these conditions.
It is important to maintain reaction (5) as a main process. It should be noted that the higher temperature stability under reducing conditions makes ferrites more preferable starting materials compared to other perovskite-like compounds of reducible transition metals (cobaltites, nickelates, cuprates, etc.) [
27].
Moreover, h-BN has a number of properties making it suitable for reactions of type (1, 5). First of all, it has decent reactivity in reactions of type (1, 4) with rare-earth (La, Y, etc.) and alkali-earth (Ca, Sr, etc.) perovskite-related ferrites. At the same time, it is exceptionally stable in inert and reducing atmospheres, and is also quite resistant to oxidation at moderate temperatures in air.
Figure 1 shows the powder XRD patterns of the LaFeO
3 ferrite-based samples after interaction with h-BN at ~750 °C with different amounts of exsolved Fe, depending on the LaFeO
3/h-BN ratio. As the reaction proceeds, the peaks corresponding to LaBO
3 and metal Fe appear in patterns, which are consistent with reaction (5) (
Figure 1(a1)). The metal Fe peaks are distinguishable, but they strongly overlap with the LaBO
3 ones. The corresponding Mössbauer spectra are shown in
Figure 1(b1,b2). Each of the spectra consist of two magnetically split sextet components and a paramagnetic singlet component. All the components are well resolved and do not broaden. According to their hyperfine parameters (
Table 1), the first of two sextets with an isomer shift (δ) of 0.37 mm s
−1 and a hyperfine magnetic field (H
hf) of 52.4 T correspond to LaFeO
3 and the second, with δ ~0 mm s
−1 and H
hf = 33 T, correspond to αFe. The paramagnetic singlet, according to its hyperfine parameters (δ~−0.1 mm s
−1 at 298 K), corresponds to metal γFe. Moreover, γFe is a high-temperature paramagnetic form, with a close-packed fcc crystal structure. It is important to note that γFe is metastable at temperatures below 910 °C [
29], while the synthesis temperature was ~750 °C. The metal γFe nanoparticles, however, may be undetectable in the XRD patterns (
Figure 1(a2,b3)). The SEM images of LaFeO
3 after the ACAR exsolution display smooth surfaces, without distinguishable metal Fe nanoparticles distributed on them (
Figure 2a).
Figure 2b,c shows typical TEM images of such samples. For the samples at the low or moderate extent of reaction (
Figure 2b,c), the images reveal agglomerated grains of different contrast, probably due to variations in thickness, but exsolved metal Fe particles are not clearly distinguishable. For the samples at the high extent of reaction >40%, obtained after prolonged heating for more than 15 h, the metal Fe particles are also largely undistinguishable, but some whisker-like metal Fe agglomerates are visible (
Figure 2d) in small amounts with respect to the reaction extent. Additionally, the EDX analysis revealed numerous La-rich oxide grains, which, according to the XRD analysis, are actually LaBO
3, since boron is undetectable. Therefore, the metal Fe particles are embedded in the oxide matrix after the ACAR-promoted exsolution, which makes them hard to distinguish, except for the formation of whiskers, if any.
Figure 1.
(a) Powder XRD patterns of the LaFeO3/h-BN-derived nanocomposites: (a1) 29% of the total Fe amount exsolved, where 11% is in γ form; (a2) 54% of Fe exsolved (8% γFe); (a3) sample (a1) oxidized at 300 °C and subsequently reduced at 700 °C, all 29% of Fe exsolved is in α form (all the Fe contributions were evaluated by Mössbauer spectroscopy). (b) Corresponding Mössbauer spectra: (b1,b2) of sample (a1) at RT and 78 K, respectively; (b3) of sample (a2); (b4) of sample (a3) at RT.
Figure 1.
(a) Powder XRD patterns of the LaFeO3/h-BN-derived nanocomposites: (a1) 29% of the total Fe amount exsolved, where 11% is in γ form; (a2) 54% of Fe exsolved (8% γFe); (a3) sample (a1) oxidized at 300 °C and subsequently reduced at 700 °C, all 29% of Fe exsolved is in α form (all the Fe contributions were evaluated by Mössbauer spectroscopy). (b) Corresponding Mössbauer spectra: (b1,b2) of sample (a1) at RT and 78 K, respectively; (b3) of sample (a2); (b4) of sample (a3) at RT.
Table 1.
Hyperfine parameters of Mössbauer spectra of the LaFeO
3/h-BN-derived nanocomposites measured at RT (except
Figure 1(b2) at 78 K): δ—isomer shift, ΔEQ—quadrupole splitting, H—hyperfine magnetic field, A—relative area, Γ—linewidth.
Table 1.
Hyperfine parameters of Mössbauer spectra of the LaFeO
3/h-BN-derived nanocomposites measured at RT (except
Figure 1(b2) at 78 K): δ—isomer shift, ΔEQ—quadrupole splitting, H—hyperfine magnetic field, A—relative area, Γ—linewidth.
Sample | Component | δ (mm s−1) ±0.01 | ΔEQ (mm s−1) ±0.01 | H (T) ±0.1 | A (%) ±0.5 | Γ (mm s−1) ±0.01 | Comments |
---|
In Figure 1(b1) | s11 | 0.37 | −0.07 | 52.4 | 71 | 0.26 | Fe3+ oct. in LaFeO3 |
s21 | −0.01 | −0.01 | 33.0 | 18 | 0.24 | αFe |
d11 | −0.10 | 0.00 | - | 11 | 0.25 | γFe |
In Figure 1(b2) 78 K | s12 | 0.48 | −0.07 | 56.2 | 71 | 0.25 | Fe3+ oct. in LaFeO3 |
s22 | 0.11 | 0.01 | 33.8 | 18 | 0.25 | αFe |
d12 | 0.01 | 0.00 | - | 11 | 0.25 | γFe |
In Figure 1(b3) | s13 | 0.37 | −0.07 | 52.4 | 46 | 0.27 | Fe3+ oct. in LaFeO3 |
s23 | 0.00 | 0.00 | 33.0 | 45 | 0.24 | αFe |
d13 | −0.10 | 0.00 | - | 8 | 0.25 | γFe |
In Figure 1(b4) | s15 | 0.37 | −0.07 | 52.3 | 72 | 0.25 | Fe3+ oct. in LaFeO3 |
s25 | 0.00 | 0.00 | 33.0 | 28 | 0.24 | αFe |
In Figure 3(1) | s123 | 0.37 | −0.06 | 52.4 | 68 | 0.27 | Fe3+ oct. in LaFeO3 |
s223 | 0.00 | 0.01 | 33.1 | 23 | 0.23 | αFe |
d123 | −0.10 | 0.00 | - | 9 | 0.26 | γFe |
In Figure 3(2) | s124 | 0.37 | −0.07 | 52.2 | 70 | 0.26 | Fe3+ oct. in LaFeO3 |
s224 | −0.01 | 0.01 | 33.0 | 17 | 0.21 | αFe |
d124 | −0.10 | 0.00 | - | 13 | 0.23 | γFe |
In Figure 3(3) | s125 | 0.37 | −0.07 | 52.2 | 71 | 0.25 | Fe3+ oct. in LaFeO3 |
s225 | 0.35 | −0.17 | 51.0 | 5 | 0.33 | Fe3+ oct. in αFe2O3 |
s325 | 0.00 | 0.01 | 33.0 | 23 | 0.23 | αFe |
d125 | −0.10 | 0.00 | - | 1 | 0.26 | γFe |
In Figure 3(4) | s126 | 0.37 | −0.07 | 52.2 | 69 | 0.26 | Fe3+ oct. in LaFeO3 |
s226 | 0.00 | 0.00 | 33.0 | 30 | 0.22 | αFe |
d126 | −0.12 | 0.00 | - | 1 | 0.25 | γFe |
In Figure 3(5) | s127 | 0.37 | −0.08 | 52.3 | 70 | 0.25 | Fe3+ oct. in LaFeO3 |
s227 | 0.38 | −0.18 | 51.4 | 30 | 0.25 | Fe3+ oct. in αFe2O3 |
In Figure 3(6) | s128 | 0.37 | −0.06 | 52.2 | 70 | 0.26 | Fe3+ oct. in LaFeO3 |
s228 | 0.00 | 0.00 | 33.0 | 14 | 0.21 | αFe |
d128 | −0.10 | 0.00 | - | 16 | 0.24 | γFe |
Figure 2.
SEM and TEM images of the LaFeO3/h-BN-derived nanocomposites: (a,b) SEM and bright field (BF) TEM images of the nanocomposites with ~45% of Fe total exsolved; (c,d) high-angle annular dark-field scanning TEM (HAADF-STEM) image and energy-dispersive X-ray (EDX) elemental analysis in selected locations of the nanocomposites with ~12% of Fe total (~10% γFe); (d) formation of the Fe whisker (w) in sample (a,b).
Figure 2.
SEM and TEM images of the LaFeO3/h-BN-derived nanocomposites: (a,b) SEM and bright field (BF) TEM images of the nanocomposites with ~45% of Fe total exsolved; (c,d) high-angle annular dark-field scanning TEM (HAADF-STEM) image and energy-dispersive X-ray (EDX) elemental analysis in selected locations of the nanocomposites with ~12% of Fe total (~10% γFe); (d) formation of the Fe whisker (w) in sample (a,b).
Figure 3.
RT Mössbauer spectra of the LaFeO3/h-BN-derived nanocomposites during redox cycling: (1) as prepared with 32% of Fe exsolved (~9% γFe); (2) sample (1) after consecutive oxidation (air, 500 °C) and reduction (10% H2/Ar, 700 °C) (~30% of Fe total, ~13% γFe); (3) sample (2) after oxidation at 300 °C in air (~23% αFe, ~1% γFe, ~5% αFe2O3); (4) after reduction of sample (3) (10% H2/Ar, 700 °C) (~30% αFe); (5) after oxidation of sample (4) (air, 500 °C) (~30% αFe2O3); (6) after reduction (10% H2/Ar, 700 °C) of sample (5) (~30% of Fe total, ~16% γFe).
Figure 3.
RT Mössbauer spectra of the LaFeO3/h-BN-derived nanocomposites during redox cycling: (1) as prepared with 32% of Fe exsolved (~9% γFe); (2) sample (1) after consecutive oxidation (air, 500 °C) and reduction (10% H2/Ar, 700 °C) (~30% of Fe total, ~13% γFe); (3) sample (2) after oxidation at 300 °C in air (~23% αFe, ~1% γFe, ~5% αFe2O3); (4) after reduction of sample (3) (10% H2/Ar, 700 °C) (~30% αFe); (5) after oxidation of sample (4) (air, 500 °C) (~30% αFe2O3); (6) after reduction (10% H2/Ar, 700 °C) of sample (5) (~30% of Fe total, ~16% γFe).
This is rather different from the metal exsolution from the A-site-deficient perovskites or the high-temperature hydrogen reduction of stoichiometric perovskite oxides reported previously, where the exsolved metal nanoparticles are typically clearly visible on the surfaces of oxide matrixes [
10,
11,
12,
13]. In these processes, the reducible metal cations migrate toward the outer and inner grain surfaces, which are exposed to the reducing environment, and reduce to metal, with subsequent agglomeration, grain growth, etc. The oxide matrices, in turn, undergo shrinkage because of a loss of oxygen and metal constituents, promoting socketing of the exsolved nanoparticles [
5,
10,
12,
23]. In our ACAR-promoted exsolution process, metal reduction/exsolution is accompanied by the formation of other complex oxides of two or more elements instead of the initial perovskite oxide, even those of lower density, i.e., LaBO
3 vs. LaFeO
3. Their formation begins in the contact areas between the initial ferrite grains and the reducing reagents. These newly formed Fe-free oxides (viz. LaBO
3) on the top of the initial perovskite grains presumably create a diffusion barrier for the Fe species. Consequently, the exsolved metal Fe nanoparticles do not appear on top of the oxide grain surfaces, but instead localize underneath their surfaces in generated in situ voids.
Perovskite-like ferrites of alkaline-earth elements, such as Ca
2Fe
2O
5 and Sr
2Fe
2O
5, can also be used in reaction (1, 4) to produce metal nanoparticle-bearing composites. Because Ca is lighter than La and Sr, it may be more favorable for microscopic investigations in terms of the nanoparticle–matrix contrast. As stated above, the starting ferrite is the main source of oxygen for the formation of the resulting oxides in ACAR-promoted exsolution processes involving reaction (1, 4). Consequently, in the case of La ferrite, other La borates, such as LaB
3O
6, cannot be formed from the LaFeO
3 precursor due to a lack of oxygen. For Ca
2Fe
2O
5 and h-BN as starting materials, the formation of several Ca borates is possible, viz. Ca
2B
2O
5 and Ca
3B
2O
6 (without/with CaB
2O
4). The powder XRD patterns of Ca
2Fe
2O
5 after the ACAR-promoted exsolution reactions with h-BN and the corresponding Mössbauer spectra are shown in
Figure 4. According to
Figure 4(a2), the main complex oxide product is Ca
3B
2O
6 and, therefore, the primary reaction can be written as follow:
Hydrogen, in addition to maintaining the low pO
2, reacts with excessive oxygen in this case. It is worth noting that no diffraction peaks corresponding to Ca
3B
2O
6 and/or other possible resulting oxides were observed at the low extent of reaction, indicating that the oxide products were X-ray amorphous (
Figure 4(a1)). Ca
2Fe
2O
5 is more active in this type of reaction compared to LaFeO
3 and is less stable in reducing conditions as well, so minor additional peaks of CaO can be observed in the patterns of some samples depending on the preparation and reduction conditions (
Figure 4(a3)). The Mössbauer spectra of the Ca
2Fe
2O
5-derived samples are shown in
Figure 4b. The spectra are comprised of three magnetically split sextet components and a paramagnetic singlet component. The Ca
2Fe
2O
5 (=CaFeO
2.5) has a brownmillerite structure, which is oxygen deficient compared to ABO
3 perovskites with fully ordered oxygen vacancies at room temperature, where Fe
3+ cations equally occupy distorted octahedral and tetrahedral oxygen polyhedra. Consequently, two of the sextets of nearly equal spectral contributions, according to their hyperfine parameters, correspond to the brownmillerite subspectrum, i.e., Fe
3+tet (δ~0.19 mm s
−1 and H
hf = 43.4 T) and Fe
3+oct (δ~0.37 mm s
−1 and H
hf = 51.2 T) in Ca
2Fe
2O
5. The third sextet (δ~0 mm s
−1 and H
hf = 33 T) and a singlet (δ~−0.1 mm s
−1) correspond to metal Fe in α and γ forms, respectively, i.e., to the metal Fe subspectrum (
Table S2). Like in the LaFeO
3 case, at ~650–730 °C, the temperatures during the ACAR-promoted exsolution process were much lower than 910 °C. Both the Fe forms can be distinguished in the powder XRD patterns (
Figure 4a). Similar to the previous case, the SEM images of the Ca
2Fe
2O
5-derived samples display smooth surfaces, without distinguishable metal Fe particles (
Figure 5a). The TEM images (
Figure 5b,c) show agglomerates of different contrasts, without metal Fe particles being clearly visible on the surfaces. At the same time, a small number of Fe whiskers can be observed in some samples (
Figure 5d). However, as it follows from
Figure 5c, the EDX mapping in several locations shows areas of Fe segregation, which, taking into account the distributions of other elements, can be identified as Fe-embedded nanoparticles. These exsolved, embedded Fe nanoparticles, identified by the EDX analysis, were about 15–25 nm in size. The EDX analysis also displays Ca-rich O-containing agglomerates (i.e., Ca
3B
2O
6, etc.).
The data show that the Ca
2Fe
2O
5/h-BN system demonstrates similar behavior to the LaFeO
3/h-BN system in the exsolution reactions facilitated by ACARs. In both the systems, the exsolved metal Fe nanoparticles, which can be determined by EDX analysis, are mostly located within oxide matrices, consisting of initial ferrites and newly formed iron-free borates, in the generated in situ voids. Presumably, the reaction zones and, hence, the adjacent voids are somehow connected to the low pO
2 environment, otherwise, in most cases, the metal reduction reaction will not proceed to any significant extent. Metal Fe in the exsolved nanoparticles can exist in two forms, namely metastable at temperatures <910 °C γFe and stable αFe, the latter can also form whiskers in some cases. The interior particle growth in a confined space in voids appears to be a key factor, along with a nanosized dimension, for the γFe formation at temperatures well below 910 °C. The fcc close-packed crystal structure of γFe is denser than the bcc structure of α(δ)Fe [
30,
31], so that the compressive strain developed when the particles of nano-scale dimensions grow under confined conditions is conducive to the formation of γFe nanoparticles and their subsequent stabilization upon cooling. The strain-induced formation of γFe nanoparticles, smaller than ~20 Å, under confined conditions in oxide matrices (Al
2O
3, MgO), which are stable at an ambient temperature, has been reported previously [
32,
33]. The stabilization of γFe was also observed in iron coatings produced by arc plasma deposition on porous alumina substrates, when Fe was localized inside pores with a diameter <160 nm [
34]. Note that in this case, the deposited Fe layers only covered the inner walls of the pores and did not completely fill their interior space, leaving central gaps.
The γFe/αFe ratio depends on the reaction conditions and extent: the contributions of γFe are larger in the initial stages and at the low extent of reactions. However, the nanocomposites containing only γFe have not been obtained using individual Ca2Fe2O5 or LaFeO3 ferrites. Ca2Fe2O5 is more active during the described process than LaFeO3 and reacts at lower temperatures, and the reaction can proceed almost completely, whereas for LaFeO3, its extent is limited to ~50–60% (based on the Fe content in coexisting phases obtained from the Mössbauer spectra).
2.3. ACAR-Promoted Metal Exsolution Using the Substituted Ferrites
The properties of Ca
2Fe
2O
5 or LaFeO
3 ferrites can be significantly modified by substitutions. Accordingly, the effect of cation substitution on ACAR-promoted metal exsolution and γFe formation was investigated. Rare-earth and alkaline-earth ABO
3−γ perovskite ferrites allow substitution in both A- and B-sites and a wide variation in oxygen content of 0 ≤ γ ≤ 0.5 [
14,
15].
Figure 6b shows the powder XRD patterns of A and B double-substituted compounds, Ca
1.4Y
0.6Fe
1.8Zn
0.2O
5.2 (=Ca
0.7Y
0.3Fe
0.9Zn
0.1O
2.6), after ACAR-promoted exsolution using h-BN. Y was chosen as the lightest rare-earth 3+ cation. The Ca
1-xY
xFeO
3−γ solid solutions have not been studied in detail in the literature, but for our process it is important that they belong to a pseudobinary system, i.e., no additional phases other than the perovskite solid solutions coexist. The solubility ranges from both the Ca side (the brownmillerite type solid solutions with oxygen excess) and the Y side (the LaFeO
3−γ type ones with mainly disordered oxygen vacancies) are not well defined and, presumably, depend on the temperature [
35]. Similar to the case of unsubstituted Ca ferrite, the resulting borates can only be correctly identified at relatively large reaction extents, such as in the sample in
Figure 6(b2), with ~29% of Fe exsolved, according to the Mössbauer spectrum shown in
Figure 6(a2). The XRD phase analysis revealed that substitution with Y increases the number of oxide products formed, and the ACAR-promoted exsolution process in this case can be written in a simplified form as follows:
The main oxide products are Ca
3B
2O
6 and Y
2O
3 (Zn
5B
4O
11 could be a minor phase). The Mössbauer spectra (
Figure 6(a2,a3)) show that 29% of the Fe total content was exsolved in this sample, where γFe accounted for ~9% and αFe for 20%. The substitution manifests itself in a significant broadening of lines in the Mössbauer spectra, due to the introduction of local distortions and the disruption of the magnetic superexchange interactions. Consequently, the components of the brownmillerite subspectrum that correspond to the tetrahedral and octahedral plus pentagonal positions in the ferrite structures were fitted with combinations of several Zeeman sextets. Sextets with δ~0.34–0.37 mm s
−1 and H
hf~49–51 T correspond to Fe
3+ cations in octahedral and pentagonal positions (poorly resolved at RT), while those with δ~0.18–0.20 mm s
−1 and H
hf ~40–43 T correspond to the tetrahedral ones. The two other components with narrow lines correspond to the metal Fe subspectrum, comprising of the αFe sextet with δ~0 mm s
−1 and H
hf ~33 T and the γFe singlet with δ~−0.1 mm s
−1. The Mössbauer measurements at 78 K result in the narrower lines of the components, but the spectral contributions of the subspectra remain about the same. At a lower exsolution level of ~3–6%, all metal Fe was in the γ form (
Figure 6(a1),
Table S3). The TEM images of the latter samples are shown in
Figure 7. Similar to the unsubstituted Ca
2Fe
2O
5-derived samples, it was difficult to differentiate the Fe particles among the agglomerates with different contrasts. However, the Fe nanoparticles <50 nm in size embedded in the oxide matrix were clearly identified by the EDX analysis and mapping (
Figure 7a,c). The HAADF-STEM images, together with the Fourier transform imaging, confirmed that the exsolved nanoparticles are metallic γFe (
Figure 7b). The EDX mapping also shows some degree of Y and O segregation on the scale of tens of nanometers, which is consistent with the XRD phase analysis.
Since Ca is more active in the exsolution reactions than Y, the remaining Ca–Y ferrite can also be enriched in Y despite the formation of Y
2O
3. This can be seen from the spectra of the samples with high degrees of exsolution, as shown in
Figure S1 and Table S4. They reveal that the contributions of spectral components corresponding to Fe
3+ in tetrahedral coordination decrease significantly because of the transformation of Ca ferrite-based Ca
1−xY
xFeO
3−γ solid solutions with x = 0.3 to Y ferrite-based solid solutions (presumably with x > 0.6 [
35]).
Preferable formation of Ca-rich borates was similarly observed for the La-substituted Ca
1−xLa
xFeO
3−γ starting ferrites. This system is also pseudobinary. Within this system, the Ca
2LaFe
3O
8 Grenier phase, with a crystal structure intermediate between brownmillerite and perovskite types, was reported. In addition, the formation of microdomains of close compositions is possible at different values of x and γ [
36,
37,
38]. The main oxide product in ACAR-promoted exsolution reactions of Ca–La ferrites with h-BN is Ca
3B
2O
6, like for Ca
0.7La
0.3FeO
2.65 in
Figure 8(a2,a3)), while the formation of La
2O
3 was not observed. At higher exsolution degrees, double borate Ca
3La
3(BO
3)
5 was additionally formed (
Figure 8(a3)). Since La is retained more in the Ca
0.7-zLa
0.3+zFeO
2.65+δ solid solution, its content increases, as does the oxygen index value. At some level of La content, the resulting ferrite solid solution loses oxygen vacancy ordering, i.e., it becomes the La
0.3+zCa
0.7−zFeO
3−γ type perovskite-like solid solution. This is reflected in the Mössbauer spectra as the disappearance of Fe
3+ tetrahedral spectral components (
Figure 8(b2),
Table S5). A simplified reaction is shown in (11):
In the case of LaFeO
3-based compounds, the substitution of Fe
3+ cations by non-reducible cations can also lead to the formation of γFe exclusively.
Figure 9 shows that the Zn-substituted solid solution, LaFe
0.8Zn
0.2O
2.9, reacts with h-BN, according to the simplified reaction:
At ~8% of the total metal Fe content, all of the exsolved Fe is in the γ form (
Figure 9(b1),
Table S5). The TEM EDX mapping of this LaFe
0.8Zn
0.2O
2.9-derived sample with ~8% of γFe exsolved, allowed for the identification of the embedded Fe metal nanoparticles (
Figure 10a). The HAADF-STEM images, together with the Fourier transform imaging, confirmed that the exsolved nanoparticles are highly twinned metal γFe nanocrystals (
Figure 10b,c).
In the case of double substitution by Ca and Zn, the ACAR exsolution with h-BN will proceed as follows (
Figure S2(a2,a3,b2,b3)):
At low reaction extents only γFe will be exsolved (
Figure S2(b2)). In both cases, Zn-containing phases may precipitate at high reaction extents, presumably in oxide (ZnO) or borate forms, but in most of our samples they did not appear in the XRD patterns.
The lattice parameters of the cubic fcc cell of the exsolved γFe nanocrystals can be estimated from the room-temperature powder XRD patterns, in the range of 0.357–0.358 nm. These values correspond to those calculated for austenite solid solutions at room temperature when extrapolated to pure iron [
39,
40]. Since the γFe nanoparticles synthesized here at 650–750 °C are crystalline and are confined in oxide/borate matrices, they are not pyrophoric in air at room temperature. Moreover, they persist during heating/cooling cycles in a reducing atmosphere. However, a mild mechanical impact on the nanocomposites with high γFe content, by gently grinding in a mortar and pestle for 1–2 min, leads to the transformation of most of the γFe into αFe, and irreversibly so (
Figure 9(a2,a3,b2,b3) and
Figure S2(a3,a4,b3,b4), Table S6). This transition is, to some extent, analogous to transformations of retained austenite caused by mechanical deformation at low temperatures [
41,
42].
Note that for all the substituted ferrites investigated, the formation of Fe whiskers, which is considered undesirable during the described processes, was not observed.
2.4. Redox Behavior of the Exsolved Nanoparticles
The exsolved metal nanoparticles can be oxidized in air to iron oxides at elevated temperatures. The oxidation of metal nanoparticles is generally associated with a decrease in density and an increase in volume. As it was shown in [
23], for CoNi exsolved, socketed nanoparticles, these volume changes lead to their migration from the initial sites under redox cycling conditions. In addition, redox cycling can cause the growth and coarsening of oxidized metal nanoparticles or their reintegration into perovskite matrices during oxidation [
10,
43,
44]. For all the nanocomposites obtained herein, the embedded nanoparticles are also completely oxidized to αFe
2O
3 at 500 °C and above, in air. The γFe nanoparticles are oxidized first, and at lower temperatures of 200–300 °C, and their oxidation is accompanied by the transition of residual γFe into the α form (
Figure 3(3,5),
Table 1). However, the nanocomposites produced by ACAR-promoted exsolution exhibited peculiar behavior during redox cycling, or at least at the temperatures investigated. In these nanocomposites, the initial metal nanoparticles, after complete oxidation into αFe
2O
3 at 500 °C and above, can revert back to the metallic state by subsequent reduction in 10% H
2/Ar at 650–750 °C and, remarkably, the metal γFe nanoparticles can be reinstated in the γ form through such a reduction of αFe
2O
3. Moreover, the γFe fractions may increase compared to the initial content, especially when Ca
2Fe
2O
5 and its solid solutions have been used (
Figure 3(1,2) and
Figure 4(a3,a4,b3,b4)). To our knowledge, the formation of γFe nanoparticles from Fe oxides by hydrogen reduction at temperatures below 910 °C has not been reported previously. Like the formation of the initial γFe nanoparticles, this is presumably a compressive strain-driven behavior, which can be attributed to the preservation of the localization of all the nanoparticles, i.e., initially exsolved metallic-derived oxide and restored metallic oxide, within the voids, along with their strong binding to the void walls during the redox cycles. The γFe formation can, thus, be considered as a kind of indicator of the localization of nanoparticles inside the voids during these processes. Note that at the same time, there should also be free space inside the voids close to nanoparticles, sufficient to compensate for the metal/oxide volume difference, otherwise the matrix may be destroyed.
Using the above assumption on the nanoparticle redox behavior, the reversible transformation of γFe into αFe, and vice versa, through redox cycling at different oxidation temperatures can be realized (
Figure 3,
Table 1). The sample of the LaFeO
3/h-BN based nanocomposite with ~30% of Fe exsolved was subjected to a preliminary oxidation (at 500 °C)/reduction cycle to maximize the γFe contribution (
Figure 3(1,2),
Table 1 (d123, d124)). To convert most of the Fe into the α form, the first stage involved oxidation at a low temperature of ~300 °C, where more chemically active γFe nanoparticles were partially oxidized/partially converted into αFe (
Figure 3(3)), but the resulting oxides did not sufficiently sinter with the matrix. In the second stage, subsequent reduction at 700 °C yields αFe (
Figure 3(4), see also
Figure 1(a3,b4)). To regenerate γFe, the obtained αFe nanoparticles were first completely oxidized at 500 °C (
Figure 3(5)), which provides sufficient sintering of the formed αFe
2O
3 particles and the matrix. Subsequent reduction, at the same temperature of 700 °C, regenerated the γFe nanoparticles (
Figure 3(6)). According to the Mössbauer spectra, the Fe
0 (metal)/Fe
3+ (in matrix) ratio remained approximately the same during redox cycling (
Table 1).
The TEM images of the reduced sample with ~11% of γFe and ~6% of αFe, obtained by the ACAR-promoted exsolution reaction of Ca
0.7Y
03(Fe
0.9Zn
0.1)O
2.6 with h-BN, after an oxidation–reduction cycle demonstrate features similar to uncycled (only exsolved) nanocomposites: using EDX analysis, the Fe nanoparticles can be identified in some locations, embedded in the oxide matrix, along with some Y segregation (
Figure S3).
2.5. ACAR-Promoted Exsolution of FeNix Alloys
It is well-established that the existence of a region of the γ form can be significantly extended to lower temperatures by creating an Fe alloy with certain elements, such as Ni [
29]. The exsolution of NiFe alloy nanoparticles from Ni and Fe-containing perovskite-like oxides has also been reported [
10,
16,
21], so the formation of FeNi
x alloys was expected during our process as well. Ni is less chemically active and more reducible compared to Fe, so Ni additions to La, and especially Ca, ferrites significantly decrease their stability in a reducing H
2-containing atmosphere [
27]. For this reason, the starting ferrites with low Ni content were used herein, to avoid decomposition. According to the XRD patterns shown in
Figure 11a, the ACAR-promoted exsolution with h-BN from LaFe
0.8Ni
0.2O
2.9, La
0.8Ca
0.2Fe
0.8Ni
0.2O
2.8, and La
0.5Ca
0.5Fe
0.9Ni
0.1O
2.7 can be written as follows:
In these reactions, the resulting borates are the same as for the Ni-free ferrites described before (
Figure 11(a1,a2,a4)). Note that while it is difficult to achieve exsolution levels of more than ~50% using metal Fe for the Ni-free ferrites due to cation mobility limitations, the reaction extent can be significantly higher for Ni-containing ferrites. As follows from the Mössbauer spectra shown in
Figure 11(b1,b2,b4) and
Table S7, Ni additions effectively stabilize the fcc structure of the exsolved nanoparticles. Their spectra are mainly comprised of two subspectra. The first, magnetically split with broad lines, correspond to Fe
3+ cations in ferrites. It was fitted as a set of sextets, with δ~0.37 mm s
−1 and H
hf~49–53 T. The second subspectrum, which was fitted as a paramagnetic singlet with δ~−0.07 mm s
−1, corresponds to Fe in γFeNi
x alloys. According to the Fe–Ni phase diagram, at temperatures close to ambient, the αFe-based bcc phase αFeNi
x coexists with intermetallic compounds of Fe
3Ni and FeNi fcc types in the Fe-rich region; although, the phase boundaries at these temperatures are difficult to determine [
29]. The thermodynamically stable phases of FeNi
x alloys, viz. the Fe-rich bcc and the Ni-rich fcc alloys, are magnetically ordered at room temperature [
45]. At elevated temperatures, there is a continuous solid solution of γFeNi
x with the eutectoid temperature of ~345–400 °C at ~50 at% of Ni. At ~10–20 at% of Ni, which matches the initial ferrite stoichiometry, the transition temperature is about ~650–700 °C. Since Ni is more reducible than Fe, the alloys at low degrees of exsolution will be enriched in Ni and their transition temperatures will be even lower. These temperatures are below the temperatures of 700–750 °C at which reactions (14, 15, and 16) were carried out. When cooled to room temperature, the XRD patterns show that all the exsolved FeNi
x nanoparticles retained their γ structure at room temperature (
Figure 11(a1,a2,a4)). The Mössbauer subspectra corresponding to the FeNi
x nanoparticles consist of singlets with δ~−0.1 mm s
−1 (
Figure 11(b1,b2,b4)), evidencing that they are paramagnetic. It is consistent with their γ form, since paramagnetic behavior at room temperature has been reported for metastable γFeNix alloys with high Fe content [
46,
47]. The stabilization of the most exsolved γFeNi
x nanoparticles in γ form upon cooling, which can be explained by the strain developed due to their localization in voids, suggests that this is the main type of localization. Similar to γFe exsolved nanoparticles, a large part of γFeNi
x nanoparticles can be transformed into ferromagnetic αFeNi
x (
Figure 11(b3)) by gently grinding in a mortar and pestle for 1–2 min. This treatment significantly reduced the FeNi
x reflections visually, in the XRD pattern of those samples (
Figure 11(a3)). The mechanical stress-induced martensite γ to α transformation in Fe-rich Fe-Ni bulk alloys at room temperature has been previously reported in the literature [
48,
49].
The exsolved γFeNi
x nanoparticles can be completely oxidized at 500 °C and above, to the spinel solid solution Ni
1±xFe
2±xO
4 (
Figure 11(a5,b5)). Subsequent reduction in 10% H
2/Ar atmosphere at 650–700 °C will restore the γFeNi
x metal nanoparticles (
Figure 11(a6,b6)). It is noteworthy that the exsolution level for this sample was >80% (
Table S7).
The TEM images of Ni-containing samples after the ACAR exsolution from La
0.5Ca
0.5Fe
0.9Ni
0.1O
2.7 ferrite, with ~41% of Fe exsolved (
Figure 11(a1,b1)), are shown in
Figure 12a–d. Similar to other samples, the images show agglomerates of different contrast (
Figure 12a–c). The embedded FeNi
x nanoparticles <30 nm in size, however, can be identified by EDX analysis, e.g., like in locations 2, 3, and 4 in
Figure 12d. The alloy nanoparticles contained approximately about ~30 at% of Ni at this exsolution degree.
Figure 12e shows the TEM images of this sample after gently grinding in a mortar with a pestle. There is not much difference compared to the unground ones, except for some rounded agglomerates of <25–30 nm located on the grain surfaces enriched in Fe and Ni, which can be identified as alloy nanoparticles (
Figure 12e, loc. 2).
Figure 12f shows the TEM images of the same unground sample after the oxidation (at 500 °C, air)/reduction (700 °C, 10% H
2/Ar) cycle. Here again, it looks similar to the original sample with embedded FeNi
x nanoparticles, which were identified by EDX analysis (
Figure 12f, loc. 2).
2.6. Separation of the Individual Nanoparticles
The reversible oxidation/reduction of metallic nanoparticles, while maintaining their localization within the in situ-created individual voids, is a remarkable feature of the nanocomposites produced by ACAR-promoted exsolution. It makes it possible to transform the initial metallic nanoparticles into various oxides, viz. Fe1−xO, Fe3O4, α/γFe2O3 etc., while preventing their agglomeration. The components of matrices, viz. the starting and resulting ferrites, resulting borates, etc., in turn, have very diverse chemical properties depending on their composition. In particular, unsubstituted, and certain substituted, Ca ferrites are susceptible to hydrolysis and dissolve in dilute mineral acids like HCl. Ca3B2O6 is also soluble in dilute acids. Since Fe3O4 (and γFe2O3) particles can dissolve reasonably slowly in dilute acidic aqueous solutions at RT, the crystallized and sintered Fe oxide nanoparticles, produced by the oxidation of the exsolved metal nanoparticles, can be separated from such oxide matrices by acid hydrolysis.
The TEM and SEM images of the nanoparticles separated from the Ca
2Fe
2O
5/h-BN and Ca
2FeAlO
5/h-BN-derived nanocomposites (~30–35% of Fe exsolved) are shown in
Figure 13 and
Figure S5. Ca
2Fe
2−xAl
xO
5 ferrites of the brownmillerite type, including Ca
2FeAlO
5, obtained by the isovalent substitution of Fe
3+ cations with Al
3+, are also prone to hydrolysis as unsubstituted Ca
2Fe
2O
5. They react with h-BN, similar to Ca
2Fe
2O
5, according to the simplified reaction (17), where Al is mainly retained in the ferrite phase (
Figure S4):
The RT Mössbauer spectrum of the Ca
2FeAlO
5/h-BN-derived nanocomposite is shown in
Figure S4(1). The subspectrum corresponding to Ca
2Fe
2−xAl
xO
5 is significantly broadened and unresolved due to the magnetic dilution by diamagnetic Al
3+ cations. Therefore, the spectral contribution of exsolved Fe, which totaled ~35% (~5% γFe), was evaluated from the 78K Mössbauer spectra (
Figure S4(2)).
Figure 13(a1,b1) shows the powder XRD patterns and Mössbauer spectra (
Table S8) of the Fe
3O
4 nanoparticles separated from the Ca
2FeAlO
5/h-BN-derived nanocomposite. The transformation of metal Fe nanoparticles into Fe
3O
4 nanoparticles was carried out by their oxidation to αFe
2O
3 at 500 °C, followed by their reduction in 10% H
2/Ar flow at 460–480 °C. The matrices were dissolved in 0.5–1.5% HCl aqueous solutions at RT. The Fe
3O
4 nanoparticles were magnetically separated and washed thoroughly with distilled water. It is worth noting that the leaching process of Fe
3O
4 can be carried out with high yields. This is facilitated by the fact that the particles were produced at relatively high temperatures at both stages of synthesis, viz. at oxidation of the metallic particles to Fe
2O
3 and at their subsequent reduction to Fe
3O
4 and were, therefore, crystallized and sintered well, which should reduce their solubility in diluted acids. Similar acid leaching in diluted aqueous HCl solutions has previously been successfully utilized to extract SrFe
12O
19-based nanoparticles, during high-temperature glass–ceramic synthesis [
50].
The HAADF-STEM and SEM images of the separated nanoparticles obtained from the Ca
2Fe
2O
5/h-BN and Ca
2FeAlO
5/h-BN-derived nanocomposites (~30–35% of Fe exsolved) reveal that most of them are cap-shaped (hemispherical) hollow nanostructures with sizes below ~200 nm and ~100 nm, respectively (
Figure 13c,d and
Figure S5). According to
Figure 13 and
Figure S5, the wall thickness of the oxide nanoparticles is of the order of a few nm, thus the wall thickness of the parent metallic nanoparticles should be of the same scale.
The separated nanoparticles provide insight into the shape and size of the parent metal nanoparticles within the matrices. These hollow cap-shaped oxide nanoparticles indicate that the exsolved metal nanoparticles are formed as an inner layer in individual voids generated in situ within the matrices. This particle shape can potentially provide free space sufficient to accommodate the volume increase during oxidation, if the hollow particles are not completely filled with other reaction products. At the same time, these shapes and locations appear to play an important role in creating the conditions necessary for the development of compressive strain in metal nanoparticles sufficient to stabilize γFe, which is somewhat analogous to the γFe formation in cylindrical pores [
34].
The extracted nanoparticles were different when La-substituted Ca
1−xLa
xFeO
2.5+δ ferrite was used as the starting material in the ACAR-promoted exsolution process.
Figure 13e,f and
Figure S6 show the TEM images of the oxide nanoparticles extracted from the nanocomposite obtained by the reaction of Ca
0.8La
0.2FeO
2.6 with h-BN (~35% of Fe exsolved). Metallic Fe nanoparticles were converted into Fe
3O
4 nanoparticles in a similar manner as described above. According to reaction (14), the La content in the remaining perovskite oxide Ca
0.8−zLa
0.2+zFeO
2.6+δ increases and, thus, its solubility in the dilute HCl aqueous solution decreases.
Using Mössbauer spectroscopy and XRD analyses, it was found that the extracted nanoparticles consisted mainly of spinel Fe
3O
4 and perovskite La
0.2+zCa
0.8−zFeO
3−γ phases (
Figure 13(a2,b2),
Table S8). The TEM with EDX analysis revealed that the La
0.2+zCa
0.8−zFeO
3-γ crystallites were segregated around Fe
3O
4 nanoparticles (
Figure 13e,f and
Figure S6). Most of them were almost completely covered by the resulting La–Ca ferrite, forming open shells. The size of the Fe
3O
4 nanoparticles was estimated to be smaller (<40–50 nm) than in the previous case. The composition and microstructure of the extracted nanoparticles reflect the relative arrangement of the metal/resulting ferrite phases in the parent nanocomposites. They show that the growth of exsolved metal nanoparticles in matrices is accompanied by the formation of shells, consisting of the resulting La-rich ferrites.