5542 Chem. Mater.

2004, 16, 5542-5548

Optimized Synthesis of the Elusive E-Fe2O3 Phase via

Sol-Gel Chemistry
Mihaela Popovici,† Martı́ Gich,*,‡ Daniel Nižňanský,§ Anna Roig,‡ Cecilia Savii,†
Lluı́s Casas,‡ Elies Molins,‡ Karel Zaveta,| Corina Enache,† Jordi Sort,⊥
Sophie de Brion,# Gerard Chouteau,# and Josep Nogués+
Institute of Chemistry Timisoara of Romanian Academy, 24 Mihai Viteazul Blvd.,
RO-300223 Timisoara, Romania, Institut de Ciència de Materials de Barcelona
(ICMAB-CSIC), Esfera UAB, 08193 Bellaterra, Catalunya, Spain, Institute of Inorganic
Chemistry, Academy of Sciences of Czech Republic, 250 68, Řež near Prague, Czech Republic,
Joint Laboratory for Mössbauer Spectroscopy, Faculty of Mathematics and Physics, Charles
University, V Holesovickach 2, 180 00 Prague 8, Czech Republic, SPINTEC (URA 2512
CNRS/CEA), CEA/Grenoble, 17 Av. Martyrs, 38054 Grenoble Cedex 9, France, Grenoble High
Magnetic Field Laboratory, CNRS B.P. 166, Grenoble 38042 Cedex 9, France, and Institució
Catalana de Recerca i Estudis Avançats (ICREA) and Departament de Fı́sica,
Universitat Autònoma de Barcelona, 08193 Bellaterra, Catalunya, Spain
Received August 17, 2004

ǫ-Fe2O3 nanoparticles embedded in a SiO2 matrix have been synthesized by sol-gel

chemistry and high temperature heat treatments. Virtually pure ǫ-Fe2O3 (in excess of 93%)
is obtained, although a two-phase mixture, ǫ-Fe2O3 + R-Fe2O3, is observed for Fe2O3/SiO2
ratios greater than 37 wt %. The ǫ-Fe2O3 nanoparticles are stable up to ∼1600 K. Optimized
ǫ-Fe2O3 nanoparticles are ferrimagnetic, with a Curie temperature TC ≈ 510 K, and
remarkably high values of room-temperature coercivity, HC ) 20 kOe.

1. Introduction proposed to synthesize ǫ-Fe2O3. Pioneering works were

based on thermal decompositions in air of Fe precur-
Iron(III) oxides are among the most common iron sors,9 or on the oxidation of Fe-containing species
compounds found in nature, and they are readily promoted by fast energy deposition techniques. In the
synthesized. Hematite, R-Fe2O3, is the most stable of thermal decomposition experiments, Fe2O3 mixed ox-
the iron oxides in ambient conditions. However, there ides,9 basic ferric salts,10,11 and other precipitates,12
are a number of polymorphs of the iron(III) oxide obtained from ferric iron salts in basic solutions, were
system: R-Fe2O3 (hematite); γ-Fe2O3 (maghemite); thermally treated. Concerning high-energy deposition
ǫ-Fe2O3; â-Fe2O3; and amorphous Fe2O3.1,2 Because of syntheses, techniques such as electric discharge,13
its technological applications in magnetic recording, gamma irradiation,14 or laser-assisted pyrolysis15 were
γ-Fe2O3 is perhaps the most extensively studied poly- used to oxidize vaporized Fe, Fe(II) formate, and an
morph,3 and ǫ-Fe2O3 is a rare polymorph that is difficult Fe(CO)5-N2O gas mixture, respectively. More recently,
to synthesize as a single-phase sample. This phase is the sol-gel approach has opened new scenarios for the
particularly interesting because, in fine particle form, synthesis of the ǫ-Fe2O3 polymorph.16-18 Silicon alkox-
it exhibits very large coercive field values.4 Natural
occurrences of ǫ-Fe2O3 have been reported in some (6) Barcova, K.; Mashlan, M.; Zboril, R.; Martinec, P.; Kula, P.
plants, as biogenic nanoparticles5 and as thermal de- Czech. J. Phys. 2001, 51, 749.
(7) Barcova, K.; Mashlan, M.; Martinec, P. Hyperfine Interact. 2002,
composition products of almandine garnets6,7 and iron- 139, 463.
rich clays.8 A number of different techniques have been (8) Van Wonterghem, J.; Mørup, S.; Villadsen, J.; Koch, C. J. H. J.
Mater. Sci. 1987, 22, 438.
(9) Forrestier, H.; Guiot-Guillain, G. C. C. R. Acad. Sci. (Paris) 1934,
* Corresponding author. E-mail: [email protected]. 199, 720.
† Institute of Chemistry Timisoara of Romanian Academy. (10) Walter-Levy, L.; Quemeneur, M. E. C. R. Acad. Sci. (Paris)
‡ Institut de Ciència de Materials de Barcelona, Esfera UAB. 1963, 257, 3410.
§ Academy of Sciences of Czech Republic. (11) Kato, A.; Matuya, S. Denki Kagaku 1977, 45, 285.
| Charles University. (12) Trautmann, J.-M.; Forestier, H. C. R. Acad. Sci. (Paris) 1965,
⊥ SPINTEC. 261, 4423.
# Grenoble High Magnetic Field Laboratory, CNRS. (13) Schrader, R.; Buttner, G. Z. Anorg. Allg. Chem. 1963, 320, 220.
+ Institució Catalana de Recerca i Estudis Avançats (ICREA) and (14) Doroshenko V. N.; Kaurkovskaya, V. N.; Yakubenko, E. P.;
Departament de Fı́sica, Universitat Autònoma de Barcelona. Pobokin, D. I.; Entinzon, I. R.; Ogenko, V. M. High Energy Chem. 2002,
(1) Cornel, R. M.; Schwertmann, U. The Iron Oxides. Structure, 36, 157.
Properties, Reactions Occurrence and Uses; VCH: Weinheim, 1996. (15) Martelli, S.; Mancini, A.; Giorgi, R.; Alexandrescu, R.; Cojocaru,
(2) Zboril, R.; Mashlan, M.; Petridis, D. Chem. Mater. 2002, 14, 969. S.; Crunteanu, A.; Voicu, I.; Balu, M.; Morjan, I. Appl. Surf. Sci. 2000,
(3) Dorman, J. L.; Fiorani, D.; Tronc, E. Adv. Phys. Chem. 1997, 154-155, 353.
98, 283. (16) Viart, N. Thesis, University Strasbourg, 1996.
(4) Jin, J.; Ohkoshi, S.; Hashimoto, K. Adv. Mater. 2004, 16, 48. (17) Chanéac, C.; Tronc E.; Jolivet, J. P. Nanostruct. Mater. 1995,
(5) McClean, R. G.; Schofield, M. A.; Kean, W. F.; Sommer, C. V.; 6, 715.
Robertson, D. P.; Toth, D.; Gajdardziska-Josifovska, M. Eur. J. Mineral. (18) Dormann, J. L.; Viart N.; Rehspringer, J. L.; Ezzir, A.;
2001, 13, 1235. Niznansky, D. Hyperfine Interact. 1998, 112, 89.

10.1021/cm048628m CCC: $27.50 © 2004 American Chemical Society

Published on Web 11/17/2004
Synthesis of ǫ-Fe2O3 Phase via Sol-Gel Chemistry Chem. Mater., Vol. 16, No. 25, 2004 5543

ides with Fe nitrate precursors are an effective way to transformation of iron-nitrate to ǫ-Fe2O3 phase takes
synthesize ǫ-Fe2O3, but typically they yield mixtures of place. Yields in excess of 93% ǫ-Fe2O3 and particle sizes
ǫ-Fe2O3 plus R-Fe2O3 and/or γ-Fe2O3. Although some of up to 20 nm were achieved. Evidently, the supporting
the classical works in ǫ-Fe2O3 reported single-phase silica matrix stabilizes ǫ-Fe2O3 up to T ≈ 1600 K, at
material,9,12,19 subsequent studies indicated that yields which temperature ǫ-Fe2O3 starts to transform to
of greater than 70% ǫ-Fe2O3 are difficult to obtain.2 R-Fe2O3. Our results confirm that ǫ-Fe2O3 is ferri-
Nevertheless, it has been recently reported that the magnetic (Ms ≈ 25 emu/g) with a large coercivity, Hc ≈
addition of Ba2+ or Sr2+ ions in the synthesis appears 20 kOe, and TC ≈ 510 K.
to stabilize the ǫ-Fe2O3 phase.4 The formation of ǫ-Fe2O3
is very sensitive to synthesis conditions, for example, 2. Experimental Section
oxidizing power of the atmosphere (oxygen, air), dura-
tion of the oxidation, or the presence of hydroxyl groups 2.1. Synthesis. Sol-gel preparations followed a procedure
(excess water, high hydrolysis ratio). Interestingly, most similar to that described by Savii et al.25 Iron nitrate (Riedel-
de Haen, 96%) and tetraethoxysilane (TEOS) (Fluka, 98%)
studies cited herein seem to indicate that ǫ-Fe2O3 can were used as precursors of the ǫ-silica nanocomposite. Targeted
only be synthesized in nanoparticle form, which sug- compositions ranged from 13 to 43 wt % Fe2O3. Hydrolysis and
gests that surface effects may play an important role condensation processes occurred in acidic hydroethanolic
in the formation of this phase. medium at TEOS:H2O:EtOH ) 1:6:6 mole ratio, reaction pH
Reported properties of ǫ-Fe2O3 are controversial, in ≈ 0.9. Reactions are self-catalyzed by the nitric acid, which
part, because it is difficult to synthesize pure ǫ-Fe2O3. results from the hydrolysis of iron nitrate. Gelation at room
temperature took place after 20 days; wet gels were dried at
For example, the crystal structure of ǫ-Fe2O3 has been 60-80 °C for 14 h. The dried xerogels were rather hard and
reported as deformed rhombohedral,9 monoclinic,13 and had a red-brownish, translucent, glassy appearance. Samples
orthorhombic.18 This latter structure, which can be were crushed in an agate mortar, and the resulting powders
considered as an intermediate structure between rhom- were subjected to thermal treatments between 300 and 1100
bohedral R-Fe2O3 and cubic γ-Fe2O3, appears to be the °C in air. This was done in air with 3 h of annealing every
most probable.20 Another source of controversy is the 100 °C, and then the samples were slowly cooled to room
temperature. Samples are labeled as SX/T (where X represents
route by which ǫ-Fe2O3 forms; both γ-Fe2O3 f ǫ-Fe2O3
the nominal weight percentage of Fe2O3 and T denotes the
and â-FeO(OH) f ǫ-Fe2O3 have been suggested.21-23 In temperature of thermal treatment in °C).
unconfined samples, a wide range of temperatures, 2.2. Characterization. Elemental analysis of Fe and Si
∼700 K15 to ∼1300 K,9 have been reported for the was performed with an inductively coupled plasma-mass
transformation ǫ-Fe2O3 f R-Fe2O3; however, ǫ-Fe2O3 is spectroscopy (ICP-MS) system (Perkin-Elmer Optima 3200 RL)
reported to be stable up to ∼1700 K when confined in and by scanning electron microscopy with energy-dispersive
the pores of a silica matrix.17 X-ray analysis (EDX) in a JEOL JSM-6300 microscope equipped
with an Oxford Instruments LINK ISIS-200 EDX.
The magnetic properties of this material are far from
Transmission electron microscopy (TEM) observations were
being well understood. Ordered ǫ-Fe2O3 is ferrimagnetic
carried out using a Philips CM 30 microscope operating at 300
with a Curie temperature of about TC ) 490 K,12,13 but kV and a Hitachi H800MT microscope operating at 200 kV.
disordered ǫ-Fe2O3 has been reported to be antiferro- Before the TEM observations, the samples were crushed and
magnetic with TN ) 490 K.24 The material has been long ultrasonically dispersed in ethanol. Drops of the solution were
known to exhibit rather large magnetic anisotropy.24 subsequently deposited onto Cu TEM-grids that were coated
Moreover, it has been recently shown that coercivities with a conductive polymer, and the ethanol was allowed to
in excess of HC ) 20 kOe can be obtained in acicular evaporate. Particle size distributions were performed by
sampling more than 150 particles from TEM micrographs and
ǫ-Fe2O3 nanoparticles;4 the corresponding value for the fitting data to log-normal distribution functions. Particle sizes
widely used hexagonal magnetoplumbite BaFe12O19 is are given as the diameter of the distribution maxima.
only HC ) 7.5 kOe, thus conferring to this material an The formation of the crystalline phases was studied by X-ray
important potential for technological applications. Con- diffraction (XRD) in a θ-2θ Bragg-Brentano geometry using
sistent with the orthorhombic crystal structure, Möss- a Siemens D5005 powder diffractometer, with diffracted beam
bauer spectra20 are interpreted as indicating four cation- monochromator and Cu KR radiation (λ ) 1.5406 Å). Diffrac-
sites: three octahedral and one tetrahedral. tion patterns were recorded from 10° to 90° with a step size of
0.1° and a scanning rate of 15 s per step. To quantitatively
We present a systematic optimization of the synthesis analyze the structural parameters, for example, phase per-
of ǫ-Fe2O3 in a silica matrix by sol-gel chemistry from centages, crystallite sizes, or cell parameters, Rietveld refine-
tetraethoxysilane (TEOS) and iron nitrate precursors. ments were performed using the MAUD program.26 The
Silica gel pores serve as nanovessels in which the orthorhombic (space group Pna21) structure, described by
Tronc et al.,20 and the usual trigonal structure (space group
R-3c) were adopted for ǫ-Fe2O3 and R-Fe2O3, respectively.
(19) Note that because no Mössbauer spectroscopy studies were
available in early studies such as in refs 9, 12, the purity of these Modeling of the silica glass is based on the assumption that
materials is only assessed from X-ray diffraction. Due to the broad from the X-ray diffraction pattern it is not possible to distin-
and overlapping peaks of this type of iron(III) oxide polymorphs, the guish if the structure is completely amorphous or nanocrys-
phase percentage as obtained from XRD may have limited accuracy. talline27 SiO2. This method approximates the SiO2 amorphous
Actually, in these systems, Mössbauer spectroscopy is perhaps the most phase as a nanocrystalline solid (cubic structure with space
accurate tool for phase quantification.
(20) Tronc, E.; Chanéac, C.; Jolivet, J. P. J. Solid State Chem. 1998, group P213) in which crystallite size is taken to be of the same
139, 93. order of magnitude as the cell parameters and the disorder is
(21) Zboril, R.; Mashlan, M.; Barcova, K.; Vujtek, K. Hyperfine statistically introduced by the microstrain effect.28
Interact. 2002, 139, 597.
(22) Chanéac, C.; Tronc E.; Jolivet, J. P. J. Mater. Chem. 1994, 6,
1905. (25) Savii, C.; Popovici, M.; Enache, C.; Subrt, J.; Niznansky, D.;
(23) Ponce-Castañeda, S.; Martı́nez, J. R.; Palomares-Sánchez, S.; Bakardzieva, S.; Caizer, C.; Hrianca, I. Solid State Ionics 2002, 151,
Ruiz, F.; Ayala-Valenzuela, O.; Matutes-Aquino, J. A. J. Sol.-Gel Sci. 219.
Technol. 2003, 27, 247. (26) Lutterrotti, L.; Gialanella, S. Acta Mater. 1997, 46, 101.
(24) Dézsi, I.; Coey, J. M. D. Phys. Status Solidi A 1973, 15, 681. (27) Le Bail, A. J. Non-Cryst. Solids 1995, 183, 32.
5544 Chem. Mater., Vol. 16, No. 25, 2004 Popovici et al.

Figure 1. XRD patterns of the SiO2-ǫ-Fe2O3 composites of

different compositions annealed at 1000 K.

Mössbauer spectra (MS) were acquired at room temperature

(and for some samples at 80 and 4.2 K) using a conventional
Mössbauer spectrometer with a 57Co/Rh source for which
velocity calibration was done using a 25 µm foil of metallic
iron, and the Mössbauer parameters are given relative to this
standard at room temperature.
Phase stability in the 300-1800 K range was studied by
differential thermal analysis (DTA) on a Perkin-Elmer DTA
7 apparatus at several heating rates under an Ar atmosphere.
Magnetic thermogravimetry (MTG) measurements were
performed in Ar atmosphere using a Perkin-Elmer TGA-7 Figure 2. (a) Iron oxide phase composition as a function of
instrument upon cooling at a 5 K/min rate from 1000 to 300 the total iron oxide content. The R-Fe2O3 (9) and ǫ-Fe2O3 (4)
K, under an applied field of ∼25 Oe. weight compositions obtained from Rietveld analysis are
Hysteresis loops were measured at room temperature in presented together with R-Fe2O3 content given by Mössbauer
fields up to 220 kOe using an extracting magnetometer at the measurements (/). (b) Dependence of crystallite sizes for
Grenoble High Magnetic Fields Laboratory facility. R-Fe2O3 (9) and ǫ-Fe2O3 (4) as obtained from Rietveld analysis
on the total iron oxide content. The lines are guides to the eye.
3. Results and Discussion
with a total iron oxide content greater than 37 wt %.
The ICP-MS elemental analysis for Fe and Si quan- Figure 2a shows the phase percentages obtained from
tification performed on a sample of the S29 series was quantitative Rietveld fits. Up to about 30 wt % iron
interpreted on the basis of SiO2/Fe2O3 composite samples oxide content, the amount of ǫ-Fe2O3 remains above
and revealed no deviations from the nominal composi- 93%, but a slight increase in R-Fe2O3 content is detected
tion (29 wt % Fe2O3), thereby excluding phenomena with increasing Fe content. This trend is drastically
such as iron-precursor leaching during synthesis. This altered for iron oxide compositions exceeding 37 wt %
result was also confirmed, with reduced resolution, by in which hematite rapidly becomes the major poly-
performing quantitative elemental analyses on several morph.
samples by EDX spectroscopy. From these studies, the Figure 2b displays the evolution of ǫ-Fe2O3 and
presence of less than a 5 wt % C and O in excess of the R-Fe2O3 crystallite sizes versus the iron oxide concen-
expected amount was established. The excess of oxygen tration, as obtained from Rietveld refinement. Because
probably originates from adsorbed water (also detected amorphous SiO2 is the major phase and the concentra-
by DTA and MTG experiments) coordinated to hydroxyl tions of the nanocrystalline phases (especially hematite)
groups. This conclusion is supported by 57Fe Mössbauer are low, there is a considerable uncertainty associated
studies, which detected only ferric iron. with the crystallite size determination. That said, the
Figure 1 is a set of XRD spectra corresponding to crystallite size, 〈D〉, is estimated as ∼10 nm for ǫ-Fe2O3
samples with different iron oxide contents, subjected to and ∼25-40 nm for R-Fe2O3. Both ǫ-Fe2O3 and R-Fe2O3
the same thermal treatment at 1000 °C. All of the crystallite sizes are weakly dependent on the iron oxide
patterns exhibit the characteristic silica glass hump concentration.
centered around 2θ ) 22°, plus several crystalline peaks. Mössbauer spectra of some of the samples in the SX/
All peaks could be assigned to either R-Fe2O3 or ǫ-Fe2O3. 1000 series are shown in Figure 3. Spectra were fitted
As expected, an increase of relative intensity of the by assuming the presence of both ǫ-Fe2O3 (based on the
crystalline peaks with respect to the amorphous hump parameters given by Tronc20) and R-Fe2O3. However,
is observed as the sample iron content is raised. It is because two of the sites for ǫ-Fe2O3 have very similar
also apparent that ǫ-Fe2O3 is the major phase when the parameters,20 they have been fitted with only one single
total iron oxide percentage is less than 37 wt %, but sextet. Therefore, the spectra were fitted to a total of
that R-Fe2O3 becomes the dominant phase in samples three sextets with relative area ratios 1:1:2 correspond-
ing to the four Fe-sites in ǫ-Fe2O3 plus an additional
(28) Lutterotti, L.; Ceccato, R.; Dal Maschio, R.; Pagani, E. Mater. sextet for the Fe-site in R-Fe2O3. Table 1 shows the
Sci. Forum 1998, 278-281, 87. parameters that were used to fit the spectra for sample
Synthesis of ǫ-Fe2O3 Phase via Sol-Gel Chemistry Chem. Mater., Vol. 16, No. 25, 2004 5545

Figure 4. XRD patterns of the SiO2-ǫ-Fe2O3 composites with

29 wt % Fe2O3 annealed at different temperatures.

particles are magnetically blocked at room temperature,

which is consistent with the results of Bødker et al.30
for crystallite sizes in the 25-40 nm range that are
relevant to this study. The determination of the R-Fe2O3
relative concentration is readily obtained from the
evaluation of the R-Fe2O3 sextet relative area, assuming
that the recoilless fractions are the same for both
polymorphs. Indeed, the increase of the most external
hematite sextet observed in the room-temperature MS
of Figure 3 is interpreted as an increase of the R-Fe2O3
content at higher iron oxide compositions. Note that for
the smallest Fe content no R-Fe2O3 peak is observed,
which suggests that pure ǫ-Fe2O3 was obtained. A plot
Figure 3. Room-temperature Mössbauer spectra from SiO2-
of the phase percentages versus total iron oxide con-
ǫ-Fe2O3 composites of different compositions annealed at 1000
K. At the top of the figure, the contributions of the different centration, from both MS and XRD Rietveld refinement
Fe sites are schematically presented. The continuous lines are (Figure 2a), indicates good agreement between the two
fits to the parameters described in the text. techniques.
The effect of the annealing temperature has been
Table 1. Hyperfine Parameters Deduced from the 300 K
Mo1 ssbauer Spectrum on Sample S29/1100a
studied on the S29 series (29 wt % iron oxide) that has
the highest Fe2O3/SiO2 ratio with ǫ-Fe2O3 ≈ 93%. The
δFe (mm/s) 2ǫ (mm/s) Γ (mm/s) Bhf (T) A (%) XRD spectra obtained from some of these samples,
((0.02) ((0.02) ((0.02) ((0.4) ((2)
annealed at different temperatures, are presented in
ǫ-Fe2O3 I-II 0.37 -0.27 0.45 44.5 40 Figure 4. A comparison of the different XRD patterns
III 0.39 -0.04 0.47 38.7 25
IV 0.23 -0.15 0.70 25.6 28 reveals that increasing the annealing temperature
R-Fe2O3 0.39 -0.30 0.40 51.0 7 correlates with a sharpening of the ǫ-Fe2O3 diffraction
a δ
peaks, which indicates that larger crystallite size is
Fe is the isomer shift quoted relative to metallic Fe, 2ǫ is the
quadrupolar shift, Γ is the fwhm of the band shift, all in mm/s, obtained by annealing at higher temperatures; for
and Bhf is the hyperfine field in tesla. A is the percentage of the example, ǫ-Fe2O3 crystallite size ranges from 〈D〉XRD ≈
total resonant area associated with each subspectrum. 5 nm for S29/900 to 〈D〉XRD ≈ 19 nm for S29/1100.
Although the peak intensities increase with increasing
S29/1100. To fit most spectra, an additional doublet, for annealing temperature, the area of the peaks, and thus
Fe magnetically unblocked superparamagnetic ǫ-Fe2O3 the ǫ-Fe2O3 content, remain constant. The TEM image
particles, was required. These particles are superpara- of sample S29/1100 (see Figure 5) shows the presence
magnetic because their magnetization-relaxation times of roughly spherical nanosized particles, well isolated
are faster than the Mössbauer measurement time scale. inside the silica matrix. The particle size obtained from
The relaxation time is τ ) τ0 exp[KV/kBT], where τ0 is the maximum of the log-normal distribution, 25 nm, is
the time constant characteristic of the material, kB is in good agreement with the XRD crystallite size for this
Boltzmans’ constant, K is the magnetic anisotropy, V sample. The same conclusion, that is, the correlation of
is the particle volume, and T is the temperature. These larger crystallite size with higher annealing tempera-
particles yield doublet spectra rather than a set of ture, can be inferred from the RT Mössbauer spectros-
sextets.29 This hypothesis was confirmed by low-tem- copy results of this series (see Figure 6). The spectra
perature measurements in which the doublet component reveal superparamagnetic behavior in the ǫ-phase which
disappeared and the corresponding ǫ-Fe2O3 area sub- strongly depends on the annealing temperature and can
spectra increased, while the area of the hematite sextet be explained by differences in the volumes of the ǫ-Fe2O3
remained constant (as shown later). Thus, hematite particles among these samples (see the equation for the

(29) Mørup, S. Hyperfine Interact. 1990, 60, 959. (30) Bødker, F.; Mørup, S. Europhys. Lett. 2000, 52, 217.
5546 Chem. Mater., Vol. 16, No. 25, 2004 Popovici et al.

Figure 7. Temperature-dependent Mössbauer spectra corre-

Figure 5. TEM image of the SiO2-ǫ-Fe2O3 composite corre- sponding to sample S29/900.
sponding to sample S29/1100.

Figure 8. Influence of the annealing temperature influence

on the ǫ-Fe2O3 crystallite size (4) and on the Mössbauer
superparamagnetic fraction (9). The lines are guides to the
eye.

samples are treated at progressively higher tempera-

tures, completely disappearing for S29/1100. This im-
plies that the relaxation time at room temperature, and
consequently the particle size, becomes progressively
larger for higher annealing temperatures. This is evi-
denced in Figure 7 where Mössbauer spectra at 4.2, 80,
and 300 K for S29/900 are shown. A progressive increase
of the blocked Mössbauer sextet subspectrum can be
observed as the temperature is lowered. At 4.2 K, the
spectrum is completely blocked. Even though for samples
annealed at lower annealing temperature (700 and 800
°C) the blocking process is also observed by Mössbauer
spectroscopy, the presence of some poorly crystalline
iron oxide/oxyhydroxide phases precursors of the ǫ-phase
Figure 6. Room-temperature Mössbauer spectra from SiO2- cannot be excluded. However, most of such possible
ǫ-Fe2O3 composites with 29 wt % Fe2O3 annealed at different precursor phases are already not stable above 500 °C.1
temperatures. At the top of the figure, the contributions of the As expected, Figure 8 shows a good correlation between
different Fe sites are schematically presented. The continuous the increase of the ǫ-Fe2O3 XRD crystallite size and the
lines are guides to the eye.
decrease of the superparamagnetic fraction of the
relaxation time of the magnetization). Note the different ǫ-Fe2O3 phase obtained from the MS fittings. Crystallite
annealing temperatures seem to have no effect on sizes in samples annealed at temperatures below 900
R-Fe2O3 (most external sextet). However, the large °C are not presented because the low crystallinity of
doublet observed for sample S29/700 gradually trans- these samples makes Rietveld refinements unreliable.
forms into the ǫ-Fe2O3 characteristic sextets as the Rietveld refinements also revealed no significant varia-
Synthesis of ǫ-Fe2O3 Phase via Sol-Gel Chemistry Chem. Mater., Vol. 16, No. 25, 2004 5547

tion of the hematite phase percentage as a consequence a key role in these processes. Actually, when the particle
of annealing in the 900-1100 °C range. size is decreased to the nanometer range, the surface-
To investigate the thermal stability of ǫ-Fe2O3, DTA to-volume ratio becomes more important and the surface
measurements of the S29/1000 sample were performed energy σi has to be considered. The free energy variation
between room temperature and ∼1600-1800 K, at per unit mass can be expressed as ∆Gmi ) (3σi/Fir) -
several heating rates. The DTA curves displayed no (∆GVi/Fi) for spherical crystallites of radius r of the ith
thermal process that could be assigned to the ǫ-Fe2O3 polymorph, where Fi is the phase density.32 Conse-
to R-Fe2O3 transformation. In fact, there are several quently, it can be inferred that even if ∆GVR > ∆GVǫ,
reasons that could make the detection of the ǫ to R the metastable ǫ-Fe2O3 can be stabilized over R-Fe2O3
transformation by DTA in our samples difficult. On one (i.e., ∆Gmǫ < ∆GmR), provided that σǫ < σR and the
hand, the content of ǫ-Fe2O3 never exceeds 30 wt % of particle size is small enough. The latter condition can
the composite and the crucible capacity is limited to be imposed either by fast reaction kinetics or, as in the
about 20 mg of sample. On the other hand, it is possible present case, by limiting the volume available for
to assume that the transformation does not take place particle growth. The stabilization of metastable nano-
at a fixed temperature but rather in a range of temper- crystalline polymorphs has been described in systems
atures. Indeed, XRD analysis of the sample after the such as BaTi2O3, ZrO2, or Al2O333-35 and in the case of
DTA experiment revealed that the transformation had iron oxide even an R-Fe2O3 to γ-Fe2O3 transformation
partially occurred: the pattern (not shown) of S29/1000 by size reduction has been reported.36 In our system,
after being heated twice to ∼1600 K at 40 K/min when Fe2O3 is formed by dehydroxylations of the FeO-
presented the diffraction peaks corresponding both to (OH) polymorphs, there is a replacement of hydroxy by
R-Fe2O3 and to ǫ-Fe2O3 phases, and a comparison with oxo bonds and the structure is densified. This is followed
the XRD spectrum of S29/1000 before the DTA runs by the development of microporosity due to expulsion
revealed that the content of the former phase increased of water and subsequent processes of micropore coales-
at the expense of the latter. This XRD spectrum also cence into macropores2 that favor the diffusion and
showed some transformation of the amorphous SiO2 into aggregation of the oxide particles. If particles grow large
the crystalline polymorph cristobalite. Moreover, the enough, the energy barrier impeding R-Fe2O3 formation
unexpectedly high intensity of the R-Fe2O3 (110) peak is overcome and ǫ-Fe2O3 is no longer favored. The
suggests the presence of Fe3O4, which has its most presence of the silica matrix prevents the process by
intense reflection, (311), located at the same 2θ value limiting the growth of the forming nanoparticles to
as R-Fe2O3 (110). To establish an upper limit for the within the silica pores, thereby avoiding particle coa-
thermal stability of ǫ-Fe2O3 and clarify which phases lescence. The above expression for ∆Gmi also allows a
are stabilized at high temperature, a subsequent DTA qualitative understanding of the increased thermal
analysis on S29/1000 sample was performed up to stability achieved with confinement in silica. Even if the
∼1800 K. In this case, the run showed superimposed temperature is high enough to supply the activation
exothermal peaks starting at temperatures above 1600 energy needed for the ǫ-Fe2O3 to R-Fe2O3 transforma-
K that after an XRD analysis of the heated sample were tion, the nucleated hematite cannot be stabilized be-
attributed to the transformation of amorphous silica into cause the growth is still impeded due to the confine-
the cristobalite and tridymite-o crystalline polymorphs. ment; hence, any fluctuation will turn the R-Fe2O3
This XRD spectrum revealed also no trace of ǫ-Fe2O3 nuclei back to ǫ-Fe2O3. However, this no longer holds
and the presence of Fe3O4, which is the stable phase at temperatures above 1300 K due to the enhanced
above 1700 K in an inert atmosphere.31 Thus, we can atomic diffusion. The fact that the pores in the SiO2
conclude that for temperatures below 1600 K only a matrix are actually interconnected can also explain why
fraction of the ǫ-Fe2O3 nanoparticles embedded in the R-Fe2O3 formation is favored above a “threshold”
amorphous SiO2 transform to the more stable R-Fe2O3. concentration of the Fe precursor: the increased iron
At higher temperatures, when the silica crystallization nitrate volume fraction leaves less free space between
takes place, the ferric oxides transform to Fe3O4. This the pores so that the distances between Fe2O3 growing
behavior contrasts with the results obtained by Traut- clusters are smaller and diffusion between neighboring
mann12 et al. in a system containing nonconfined clusters at the temperatures of the oxide formation is
ǫ-Fe2O3 particles; they reported a transformation by permitted.
DTA at ∼1000 K. Figure 9 is a plot of two representative room-temper-
The formation and stability of ǫ-Fe2O3 can be quali- ature hysteresis loops for the samples that contain
tatively understood on the basis of the confinement ǫ-Fe2O3 nanoparticles. Sample S29/700 exhibits a hys-
imposed by the porous silica matrix. In principle, there teresis loop with zero remanence and zero coercivity
is no general relation between the original iron precur- typical of superparamagnetic samples, in agreement
sor and the Fe2O3 crystalline structure obtained by with the doublet observed in the Mössbauer spectrum
thermal transformations.2 Two factors determine both of Figure 6. However, sample S29/1100 (with a com-
the formation of a given Fe2O3 phase from a precursor pletely blocked Mössbauer spectrum, as shown in Figure
cluster and the transformations between polymorphs: 6) exhibits a typical ferromagnetic (or ferrimagnetic)
the bulk free energies ∆GVi of the different i-Fe2O3
phases (i ) R, â, γ, ǫ) and the energy barrier that must (32) Rana, S.; Ram, S. Phys. Status Solidi A 2001, 201, 427.
be overcome for the transformation to take place. Both (33) Flaschen, S. S. J. Am. Chem. Soc. 1955, 77, 6194.
(34) Garvie, R. C. J. Phys. Chem. 1978, 82, 218.
factors depend on many parameters, and kinetics plays (35) McHale, J. M.; Auroux, A.; Perrotta, A. J.; Navrotsky, A. Nature
1997, 277, 788.
(36) Randrianantoandro, N.; Mercier, A. M.; Hervieu, M.; Grenèche,
(31) Darken, L. S.; Gurry, R. W. J. Am. Chem. Soc. 1946, 68, 798. J. M. Mater. Lett. 2001, 47, 150.
5548 Chem. Mater., Vol. 16, No. 25, 2004 Popovici et al.

correlates rather well with Mössbauer data. The samples

exhibiting large doublets are superparamagnetic, while
the ones showing clear sextets are ferrimagnetic with
rather large coercivities. In addition, shown in the inset
of Figure 9 is the MTG curve and its derivative for the
S29/1000 sample. A sharp decrease in magnetic weight
is evident at T ≈ 510 K, which is consistent with the
Curie temperature of ǫ-Fe2O3 reported by several au-
thors.12,13,24 The remanent magnetic force after this
transition disappears around 900 K, similar to the Néel
temperature of bulk R-Fe2O3 (TN ) 956 K1).
4. Conclusions
Sol-gel chemical synthesis of ǫ-Fe2O3 nanoparticles,
embedded in a silica matrix, was investigated. Samples
with ǫ-Fe2O3 phase percentages exceeding 93% were
Figure 9. Room-temperature hysteresis loops of samples S29/
700 (9) and S29/1100 (4). Shown in the inset is the MTG curve obtained with Fe2O3/SiO2 ratios up to 30% and heat
of sample S29/1000 and its derivative. treatments above T ) 700 °C. Transmission electron
microscopy observations reveal that ǫ-Fe2O3 particles
hysteresis loop with a coercivity of HC ) 20 kOe. These optimized for maximum content are roughly spherical
results are similar to what was reported for Ba-doped and have sizes of around 25 nm. The magnetic proper-
acicular ǫ-Fe2O3 nanoparticles which exhibited coerciv- ties of the samples range from superparamagnetism to
ites of up to HC ) 20 kOe.4 The saturation magnetiza- a ferrimagnetic behavior with a Curie temperature of
tion of sample S29/1100 was estimated using the law 510 K and a coercivity of 20 kOe at room temperature,
of approach to saturation,37 and a value MS ≈ 13 emu/g depending on the synthesis conditions. Interestingly, it
(mass of composite) was obtained. Taking into account is found that this phase remains stable up to very high
the ǫ-Fe2O3 phase percentage, the saturation magneti- temperatures, that is, T ≈ 1600 K. The fact that the
zation for the ǫ-Fe2O3 phase becomes MS ) 25 emu/g nanoparticles are confined inside the pores of a xerogel
(mass of ǫ-Fe2O3), which is similar to the value reported plays an important role in the formation and enhanced
by Jin et al. for Ba-doped ǫ-Fe2O34,38 and consistent with thermal stability of this phase, evidencing the interest
the values reported for isomorphic systems, such as of exploring chemical transformation in spatially re-
AlFeO3 or FeGaO3.39,40 The MS ) 25 emu/g value implies stricted fields.
an average magnetic moment of 0.27µB (Bohr magneton)
per Fe ion. Because the expected moment for Fe3+ ions Acknowledgment. Partial support from project
should be 5µB, this indicates that ǫ-Fe2O3 is ferrimag- 6/2001 of NATO program of the Czech Republic, Gen-
netic, as proposed by Tronc et al. on the basis of eralitat de Catalunya (Grant 2001SGR00335 and
Mössbauer measurements under applied fields.20 Actu- 2001SGR00189), Ministerio de Educación y Ciencia
ally, the magnetic behavior of the studied samples (Grant MAT-2003-01052 and MAT-2004-01679), and the
EU Human Potential Program (Grant HPMT-CT2000-
(37) Chikazumi, S. Physics of Magnetism; John Wiley & Sons: 0006 and HPRI-1999-CT00030) is acknowledged. We are
London, 1978. grateful to J. Oró and Dr. A. Fuertes for their assistance
(38) Note that in ref 4 the value of magnetization, M, at 70 kOe is
given rather than the saturation magnetization, MS. in thermal analysis measurements and to Dr. B. P.
(39) Schieber, M.; Frankel, R. B.; Blum, N. A.; Foner, S. J. Appl. Burton for the critical reading of the manuscript.
Phys. 1967, 38, 1282.
(40) Abrahams, S. C.; Reddy, J. M. Phys. Rev. Lett. 1964, 13, 688. CM048628M