Structural and chemical evolution of FeCoO based ceramics under reduction/oxidation—an in situ neutron diffraction study

Materials Science and Engineering B106 (2004) 6–26

Structural and chemical evolution of Fe–Co–O based ceramics underreduction/oxidation—an in situ neutron diffraction study

Yaping Lia, Evan R. Maxeya, James W. Richardson, Jr.a,∗, Beihai Mab

a Intense Pulsed Neutron Source, Argonne National Laboratory, Argonne, IL 60439, USAb Energy Technology Division, Argonne National Laboratory, Argonne, IL 60439, USA

Received 10 March 2003; accepted 7 July 2003

Abstract

Air-sintered ceramic samples in composition CoO·nFe2O3 (n = 1 and 2) were prepared in solid-state reactions, resulting in a single spinelphase with composition CoFe2O4, and a two-phase mixture of identical spinel with�-Fe2O3 for n = 1 and 2, respectively. Their structural andchemical evolution over pO2 range of 10−0.9 to 10−19 atm was investigated using in situ neutron diffraction at isothermal condition (∼900◦C).Neutron diffraction data were analyzed through Rietveld refinements. The following sequences of structural transformation from�-Fe2O3

hematite→ (Fe, Co)-spinel→ (Fe, Co)1−xO wustite→ (Fe, Co)O rocksalt→ �-(Fe, Co) → �-(Fe, Co) → (�-(Fe, Co)) were observedon the reduction of Fe–Co–O based ceramics. With the development of reduction at pO2 down to 10−15 atm, mixed valence (Fe2+ and Fe3+)spinel was first formed in Fe-excess (Fe, Co) spinel phases. The intermediate phases were usually Co-rich compared with their parent mixedoxide phases. Particularly, the initial metallic precipitate is Co-rich�, independent of initial stoichiometry. Reduction kinetics at pO2 of∼10−19 atm is extremely fast, but crystalline form and structural integrity are maintained. As crystal structures of the various involved phasesare very similar, few structural blocks were disturbed as oxygen was released from the samples. In addition, re-oxidation behavior of reducedproducts was also studied, and phase composition and microstructure of post-neutron experiments were characterized by X-ray diffractionand scanning electron microscopy.© 2003 Published by Elsevier B.V.

Keywords: Iron oxide; Cobalt oxide; Ceramics; Metals; Neutron scattering; Phase transitions

1. Introduction

Dense ceramic membranes in nominal composition ofSrFeCo0.5Ox (designated SFC2) have potential applicationfor oxygen separation due to their mixed conducting behav-iors at elevated temperature. Since such membranes weresuccessfully used for partial oxidation of methane to syn-gas (CO+ H2) [1], SFC2 and related materials have re-ceived considerable attention[2–7]. We are interested inthe quasi-binary Fe–Co–O system under low oxygen partialpressure for several reasons.

First, it has been clear that SFC2 is not a single phase ma-terial under operating conditions. At least four phases havebeen identified in SFC2, depending upon ceramic processingand pO2, including Sr2(Fe, Co)3O6.5−x (236 with structure

∗ Corresponding author. Tel.:+1-630-252-3554;fax: +1-630-252-4163.

E-mail address: [email protected] (J.W. Richardson Jr.).

of Sr4Fe6O13 [8]), Sr(Fe, Co)O3−y (perovskite), (Fe, Co)O(rocksalt), and (Fe, Co)3O4 (spinel) [9,10]. Mitchell et al.[11] have investigated the structural behavior of single phaseperovskite in composition SrFe0.8Co0.2O3−δ with a decreaseof pO2 value down to 10−13.3 atm. In their studies, cubicperovskite phase is stable over the entire pO2 range althoughoxygen vacancy concentration is increased with decreasingpO2. However, membrane tubes made of SrCo0.8Fe0.2Ox

perovskite exhibited a different behavior both in an experi-ment where exposed on one side to air and the other side tomethane[12] and in our in situ neutron diffraction study un-der single pO2 mode (unpublished). Therefore, the stabilityof perovskite not only depends upon pO2 level but also pO2gradients and composition. Up to now, structural and chem-ical pathways of other potentially vulnerable components ofSFC2 under a wide range of pO2 values are still lacking.But this information is important since one side of the SFC2ceramic tube will be subjected to methane streams with pO2about 10−19 atm, and the structural stability of SFC2 under

0921-5107/$ – see front matter © 2003 Published by Elsevier B.V.doi:10.1016/j.mseb.2003.07.004

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 7

strong reducing environment is dependent upon its compo-nents, and interaction of these phases under pO2 gradients.

Secondly, transition metal (Fe and Co) oxides with spinelstructures are of interest in controlling the electronic andcatalytic properties of the materials by modification ofsite population and valent state of cations. In addition,non-stoichiometric variation in spinel and rocksalt typestructures in Fe–Co–O system is well documented (e.g.wustite Fe1−xO [13], (Co, Fe)1−xO [14])) and these struc-tural defects may play important roles in electronic conduc-tion attributed to SFC2. The separation of its componentsfrom the multi-phase host in SFC2 may allow us to betterunderstand the nature of mixed conducting properties ofSFC2.

In addition, in contrast to the intensive investigation ofoxidation of metals and ceramics systems, few studies areperformed on the reduction of oxides, with major interests inactive catalysts. Recently, it was found that the microstruc-ture of metal-ceramic composites can be tailored by partialreduction of metal oxides through careful choice of reduc-ing atmosphere along with other processing parameters. Agood example is spinel oxide NiAl2O4, in which Ni metalphases were dispersed in Al2O3 ceramic-matrix by properreduction[15].

This contribution focuses on a detailed understanding ofstructural and chemical pathways in Fe–Co–O based ceram-ics through changing oxygen partial pressure at constanttemperature and total pressure by neutron diffraction. We areparticularly interested in determining the response of phasecomposition and crystal lattice to pO2 change, and Co and Fepartitioning in the reduced products to gain insights into thebehavior of SFC2 under reduction. Due to a large contrastin scattering lengths of Fe and Co from a neutron source,their distribution in oxides can be precisely determined. Fur-thermore, with higher penetrating power, neutrons are anideal source to probe structures of bulk materials. The insitu neutron data were collected at 900◦C and subsequentlyanalyzed by Rietveld methods. The temperature 900◦C ischosen because this is the expected operating temperaturefor SFC2 membranes. In addition, possible magnetic contri-butions to diffraction patterns can be avoided (e.g. the Curietemperature is 790 K for CoFe2O4).

2. Experimental

2.1. Material synthesis

Samples in composition CoO·nFe2O3 (n = 1 and 2)were prepared (designated C1F and C2F, respectively) bytraditional solid state reactions. Appropriate amounts ofCo(NO3)2·6H2O (Baker 99.5%) and Fe2O3 (Alfa 99.97%)were mixed with ZrO2 balls in isopropanol (IPA) media in aNALGENE bottle with a charge:ball:liquid ratio of 1:1:1.2.The dried powders were calcinated in 850◦C for 20 h inan alumina tray. The resulting powders were ground with a

marble mortar and pestle, and screened through a 150-meshscreen. The powder samples of C1F and C2F thus preparedwere uniaxilly die-pressed in 1/2 and 7/8 in. diameter die,respectively, to form disks with about 7 mm thickness un-der a pressure of 60 MPa. The green bodies of C1F andC2F were then pressurelessly sintered in a MoSi2 furnacein air at 1410 and 1440◦C, respectively, for 8 h. The den-sities of sintered ceramics were determined by Archimedesmethod using IPA as immersion media, which resulted in80 and 88%, respectively, based on theoretical density ofCoFe2O4 (5.29 g/cm3) and two-phase mixture (5.27 g/cm3)with 0.5CoFe2O4 and 0.5�-Fe2O3 (5.26 g/cm3) for C1Fand C2F, respectively.

2.2. Neutron diffraction

Time-of-flight neutron diffraction data were collected onas-sintered ceramic disks on the General Purpose PowderDiffractometer (GPPD) at Argonne’s Intense Pulsed Neu-tron Source (IPNS). A couple of disks of C1F and C2F weremounted on a “Miller” furnace, respectively, and variableoxygen partial pressure was realized by change of gas mix-ture involving Air, Ar, CO2, 1%CO/Ar and CO. The pre-setcontrol parameters for flow rate of gases of interest wereintegrated into data collection software. Gas flows werecontrolled with Brooks Series 5820 mass flow controllersbefore they entered a mixing loop and were delivered to thefurnace. The pO2 value was calculated by NASA computerprogram for calculation of chemical equilibrium[16] basedon the chosen mixtures, and monitored by a solid zirco-nia electrolyte oxygen sensor (Ceramic Oxide FabricatorsSIRO2). For the C2F sample, data were collected at 60 minintervals during progressive reduction and 30 min intervalsduring re-oxidation, with a change of pO2 levels, rangingfrom ∼10−0.9 to ∼10−19 atm, for every 10 runs; for theC1F sample, a constant gas mixture of CO2 and CO was in-troduced into the furnace, which gave a final oxygen partialpressure of about 10−19 atm, and data collection was basedon 30 min intervals. In both experiments, after samples weresufficiently reduced, air was switched back to exam the re-versibility of reaction. The data were recorded in the 2θ =±90◦ detector banks, and Rietveld profile analysis was per-formed with GSAS[17]. After the neutron experiments, lin-ear shrinkages of 18 and 7% were observed for C1F and C2F,respectively. The measured densities of interior samples (af-ter light polishing to remove oxide scales) of post-neutronexperiment are 73.7 and 55.5% for C1F and C2F, respec-tively, based on theoretical densities of 8.02 and 7.89 g/cm3,respectively, for metal composition of 67 and 80 at.%�-Fe.

2.3. X-ray diffraction

Room temperature X-ray powder diffraction patterns werecollected on both the surface and interior of the post-neutronsamples using a Scintag XRD 2000 diffractometer with CuK� radiation at 45 kV/40 mA. These measurements were

8 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

performed over 2� range of 5 to 90◦ and a scan rate of2◦/min.

2.4. Scanning electron microscopy

The microstructures and morphologies of samples of freshand post-neutron experiment were examined by scanningelectron microscopy (SEM) on a JEOL JSM-5400 scanningelectron microscope. The ratio of Fe:Co in ceramic andalloy-based composites (after neutron experiment) was de-termined by means of energy-dispersive X-ray spectroscopy(EDS). The samples for SEM/EDS analyses were embeddedinto a disc made of TRANSOPIC powder (Buehler) by HotMount, and were ground and polished to a 3�m diamondfinish. All the samples were coated with carbon to preventcharging, and an accelerating voltage of 20 kV was used inthe SEM/EDS studies.

3. Results and discussion

Fig. 1 is an ISAW[18] image showing diffraction data asa function of pO2 levels and time. The phases present in var-ious stages of reduction include�-Fe2O3 (hematite, R-3c),spinel (Fd-3m), rocksalt (Fm-3m),�(A2, BCC) and�(A1,FCC) for C2F, and spinel, rocksalt,� and� for C1F. In ad-dition, diffraction peaks from Al2O3 (corundum, R-3c), acomponent of the furnace, appear in the late stages of theexperiment on C1F when it was reduced and the samplevolume shrunk (We used Al2O3 as gravity plug to securethe samples in the furnace. Before the experiment, the di-mension and position of the sample in the instrument arecarefully measured, and furnace components that might beilluminated are masked by a Cd plate in beamline. However,once the sample shrunk significantly, the Al2O3 on the topof the sample could move into the incident beam). Sincethe complete diffraction data for the both experiments canbe divided into several regions with different characteristicphase compositions (Tables 1 and 2), a starting model thatrepresents phase composition for a specific region was firstrefined, and sequential structure refinements were carried

Table 1Microstructure and phase evolution in CoO·2Fe2O3 under reduction andoxidation

Region Microstructure Phases present

ReductionI Ceramic oxides Hematite+ spinel

Hematite+ spinel+ rocksaltII Metal–ceramic composites Spinel+ rocksalt+ �

Rocksalt+ �

III Metallic Alloys � + �

�

Oxidation � + � + rocksalt� + rocksalt?� + rocksalt?+ spinel?

Table 2Microstructure and phase evolution in CoO·Fe2O3 under reduction andoxidation

Region Microstructure Phases present

ReductionI Ceramic oxides Spinel

Spinel+ rocksaltII Metal–ceramic composites Spinel+ rocksalt+ �

Rocksalt+ � + �

Rocksalt+ �

III Metallic Alloys �

Oxidation �

� + rocksalt?� + rocksalt?+ spinel?

out to cover all diffraction data. In the structure refinements,both Fe and O sites for unreduced hematite were refined,but results did not reveal any significant departure from stoi-chiometric composition in the entire associated pO2 regime.Therefore, in subsequent refinements, refined parameters forhematite involve only atomic coordinates, lattice parame-ters, and phase fraction as well as non-structural parameters.The cation sites in structures of spinel and rocksalt were re-fined as Fe and Co, with total occupancy constrained to one,ignoring possible cation deficiency initially. This treatmentwill not greatly affect overall composition discussed sinceδ and x, if any, are very small for (Fe, Co)3−δO4 and (Fe,Co)1−xO. Difference-Fourier maps for structures of rocksaltwere examined later in order to look for cations in unoccu-pied cation sites since cation interstitials and cation vacan-cies are the most likely defect species in Fe and Co spineland rocksalt, such as (Co, Fe)3−δO4 spinel [19] and (Co,Fe)1−xO [14]. Because it is impossible to uniquely deter-mine the Fe and Co fractions in A1 and A2 structures with asingle crystallographic site, their occupancy was estimatedand fixed either through the known Fe:Co ratio in bulk ma-terials or by the relationship between lattice parameters andcomposition of Fe–Co alloys from the literature. For illus-tration, some typical plots of the profile fitting of neutrondata for materials under different stages of reduction areshown inFigs. 2–4. They show a good agreement betweenthe observed and calculated profiles.

3.1. Phase composition and microstructure of CoO·nFe2O3(n = 1 and 2) before reduction

Rietveld refinement of the neutron data collected for C1Fand C2F before reduction showed that C1F contains a sin-gle spinel phase (Fig. 2a), while C2F contains two phases:spinel and unreacted�-Fe2O3 (Fig. 2b), consistent with thatpredicted by phase diagram in system CoO–Fe2O3 [20]. Thestructure of spinel has both tetrahedral (A) and octahedral(B) sites, residing in Wycoff positions 8b and 16c, respec-tively, in space group of Fd-3m, with a general formulaof AB2X4. The structure refinement on spinel in C1F re-sulted in structure formula of Fe0.73Co0.27(Fe0.65Co0.35)2O4

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 9

Fig. 1. (a) Neutron diffraction data for CoO·2Fe2O3 (C2F), showing evolution of diffraction peaks from hematite (H), spinel (S), rocksalt (R), and metallicalloys � and �. (b) Neutron diffraction data for CoO·Fe2O3 (C1F), showing evolution of diffraction peaks from spinel (S), rocksalt (R), and metallicalloys � and �. In addition, peaks from corundum (C), a furnace component, also shown. Diffraction intensities are expressed as darkness. Vertical axisrepresents pO2 and time evolution (from top to bottom), and horizontal axis representsd-spacing.

(content of octahedral sites in bracket), e.g. close to idealstoichiometric compound CoFe2O4, which is in line withthe results obtained from Co ferrite CoFe2O4 nanoparticlesat 523 K, in which Fe0.72Co0.28(Fe0.63Co0.37)2O4 was sug-gested from neutron diffraction[21]. The quantitative phaseanalysis suggested that the C2F contains 64 wt.% spinel and36 wt.% hematite, with spinel in C2F having identical com-position and cation configuration to that in C1F. Consider-

ing electro-neutrality, it is expected that all the Fe and Co inspinel are in Fe3+ and Co2+ oxidation state. This indicatedthat before reduction the spinel phase in Fe–Co–O ceramicshas mixed spinel structure with inversion degreeδ = 0.73,slightly higher than the random number of 2/3 expected athigh temperature.

SEM micrograph of sintered ceramic samples (Fig. 5)indicated that both C1F and C2F samples contain large

10 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Fig. 2. Observed (+) and calculated (solid lines) intensity at 900◦C in air. The difference between the observed and calculated profiles is shown in thebottom of the figures. (a) For CoO·Fe2O3 (C1F), fitted with spinel with vertical bars of allowed reflections. (b) For CoO·2Fe2O3 (C2F), the Rietveldrefinements included spinel and hematite, with vertical bars of corresponding allowed reflection positions shown from bottom to up, respectively.

amount of porosity, which is consistent with the measureddensity and shown in dark area. Backscatter electron im-age (Fig. 5a) basically showed a uniform distribution ofFe and Co in C1F sample. However, slight color contrastswere observed in C2F sample (Fig. 5b). EDS analysis re-vealed that grey grains in C2F have a Fe:Co ratio of 97:1,and bright grains have a Fe:Co ratio of 2.3:1, thus cor-responding to�-Fe2O3 (hematite) and CoFe2O4 (spinel)two phases, respectively, as determined by neutron diffrac-tion data. SEM examination revealed that C2F sample con-sisted of aggregated hematite grains, which were randomlydistributed in spinel matrix. Small amount of Co detectedin grey area could be due to electron beam penetration

into underneath adjacent spinel grains. EDS analysis basedon area average (on 100× magnification) gave a ratio ofFe:Co= 1.9:1 and Fe:Co= 3.9:1 for C1F and C2F, respec-tively.

3.2. General trend of reduction/oxidation of C1Fand C2F

In general, the following sequence of phase evolution wasobserved in C2F as pO2 decrease:�-Fe2O3 → spinel →rocksalt → �(A1) → �(A2) → �(A1), and �(A1) →�(A2) → rocksalt → spinel on oxidation (Fig. 6 andTable 1). However, transformation of spinel→ rocksalt→

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 11

Fig. 3. Typical observed (+) and calculated (solid lines) intensity for CoO·2Fe2O3 (C2F) at 900◦C in various stages of reduction. The differencebetween the observed and calculated profiles is shown in the bottom of the figures. Vertical bars indicted the phases involved in Rietveld refinement,and corresponded to allowed reflection positions for each structure, and shown from bottom to top, respectively, for (a) spinel, hematite and rocksalt, (b)spinel,�(A1) and rocksalt, (c)�(A2) and rocksalt, (d)�(A2) and �(A1) and (e)�(A1). Each diffraction pattern was collected in 1 h.

12 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Fig. 3. (Continued ).

�(A1) → �(A2) was observed on the reduction of C1F,and little sign of oxidation occurred when gas input wasswitched back to air (Fig. 7 andTable 2). This transforma-tion sequence was basically consistent with the results ofprevious studies on the reduction of iron oxides[22,23]. Asshown inFigs. 6 and 7, as a function of pO2 observed phasestability fields in Fe–Co–O system could be divided intothree regions at 900◦C, and they correspond to ceramic ox-ides (I), metal-ceramic composites (II), and metallic alloys(III), respectively, on the basis of structure characterization.Since we are mainly interested in the behavior of metal ox-ides in reduction process, the following discussion is con-centrated on the detailed structural and chemical evolutionof these phases based on C2F sample where the neutrondata for this sample are more directly related to our objec-tive. Useful information retrieved from C1F will be used tomake a comparison and to give a more complete picture onthe Fe–Co–O system.

3.3. Hematite → spinel transformation

C2F samples initially contain two phases (spinel and�-Fe2O3), as identified by structure refinement. It has beenknown that hematite is thermodynamically unstable, andcan be transformed to magnetite with spinel structure atcertain pO2 through the following reaction:

6Fe2O3(hem) = 4Fe3O4(mag) + O2 (1)

Thermodynamic equilibrium pO2 at H/M boundary at900◦C is 10−6.5 atm, calculated using data in[24]. It wasfound by quantitative phase analysis in Rietveld refinementthat the weight fraction of spinel gradually increased with adecrease of pO2 (Fig. 6), and the transformation of hematiteto spinel progressed at a significant rate in the range of pO2∼10−12 atm. The maximum amount of spinel that the sam-ple contained is∼85 wt.%. There are at least two possi-ble underlying mechanisms that could explain the observed

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 13

Fig. 4. Typical observed (+) and calculated (solid lines) intensity for CoO·Fe2O3 (C1F) at 900◦C in a constant pO2 level in about 10−19 atm. Thedifference between the observed and calculated profiles is shown in the bottom of the figures. Vertical bars indicted the phases involved in Rietveldrefinement, and corresponded to allowed reflection positions for each structure, and shown from bottom to top, respectively, for (a) spinel and rocksalt,(b) spinel, rocksalt and�(A1), (c) �(A2), rocksalt and�(A1), (d) �(A2) and rocksalt, and (e)�(A2) and corundum from furnace. Each diffraction patternwas collected in 30 min.

14 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Fig. 4. (Continued ).

transformation from hematite to spinel. The first is that re-action (1) is initiatized from the sample surface. As a result,Fe-spinel (Fe3O4) would mainly be formed by reduction of�-Fe2O3 on the surface, followed by a moving reaction frontin the interface Fe3O4/hematite toward the interior of thesample, and finally reaching the equilibrium composition ofspinel. The second mechanism involves spontaneously pre-cipitated Fe-spinel at the interface between hematite andspinel throughout the sample volume, and the final result-ing composition of spinel being reached by the site-mixingreaction of Co-bearing spinel with Fe-spinel on adjacentgrains. If there are significant chemical variations of spinelgrains in the cross-section of the sample, as suggested inthe first mechanism, peak broadening is expected in neutrondiffraction patterns because the scattering comes from theentire sample. We have examined the diffraction peaks fromspinel under reduction, and no significant peak broadeningwas observed compared with the degree of lattice parameter

variation, suggesting that the second mechanism is favored.The transformation of hematite to spinel in C2F sample islargely therefore a homogenous internal process. The appar-ent reaction would be described as follows:

(1 − x)Co2+Fe3+2

O4(s) + 32xFe3+

2 O3(h)

→ Co2+1−x

Fe2+x

Fe3+2

O4(s) + 14xO2 (x ≤ 0.4) (2)

Long-range cation diffusion is not necessary to completethe reaction, and the slow step is reduction of hematite, asevidenced by pO2-dependent reaction rate.

Refinements of site occupancy for the structures of unre-duced spinel gave an unambiguous cation distribution intheir structures. As can be seen inFig. 8, the ratio of Fe:Coin spinel departed from the original 2:1 in reduced environ-ments, corresponding to the point where extensive transfor-mation of hematite to spinel took place and approached 4:1when hematite was almost fully reduced, consistent with

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 15

Fig. 5. Backscatter electron image of polished surface of fresh ceramicsamples before neutron experiment for (a) CoO·Fe2O3 (C1F), and (b)CoO·2Fe2O3 (C2F).

bulk composition of a ratio of Fe:Co= 4:1. The structure re-finement further indicated that Fe-enrichment in spinel wasrealized by concurrent increase of Fe population in bothtetrahedral and octahedral sites (Fig. 9).

The relationship of spinel lattice constant versus time andpO2 was plotted inFig. 10, which showed a monotonic in-crease until spinel disappeared. Lattice expansion of spinelunder reduction can be attributed to composition and struc-ture change or lattice strains. Since there is no apprecia-ble change of lattice parameters for the coexisting hematite(Fig. 11), it is reasonable to believe that lattice change ofunreduced spinel is due mainly to chemical and structural

variation. Since Fe2+ is larger than Co2+ cations with highspin state, which is generally true for oxides (Fe2+: 0.63 Åfor IV HS and 0.78 Å for VI HS; Co2+: 0.58 Å for IV HS and0.745 Å for VI HS[25]), it is plausible that the increase inlattice parameters of unreduced spinel is due to apparent sub-stitution of Co2+ by Fe2+. Continuous expansion of spinellattice was observed in hematite-spinel two-phase region.Note that lattice discontinuity at 53 h was observed whenresidual�-Fe2O3 was completely consumed and metallicphases started to precipitate (Figs. 6 and 10). A correlateddiscontinuity in the ratio of Fe:Co in spinel (Fig. 8) im-plies the composition origin of abrupt lattice expansion. Itis expected that the lattice constant for this spinel with thehighest content of Fe is close to that of Fe3O4 (magnetite)although these data are not available.

It is important to note that the structure of Fe-excessspinel has both Fe2+ and Fe3+ cations on tetrahedral (Td)and octahedral (Oh) sites. Tendency to have any cation or-dering is very small at high temperature although the ex-act extent of charge ordering of Fe2+–Fe3+ is unknown.Increase of bond lengths of both T–O and O–O with in-crease of Fe population indicated there is no preferred pop-ulation of Fe2+ cations in the two cation sites. Generally,significant increase in electronic conduction is expected ifthe multi-valent cations reside in crystallographically equiv-alent sites. The spinel phase with multi-valent Fe cationsoccupying the same crystallographic sites is helpful for en-hanced electronic conduction of SFC2 at modest pO2 levelas d electrons can hop between Fe2+/Fe3+ in Oh and Tdsites.

If the trend in growth of spinel phase inFig. 6 can befollowed at pO2 of 10−15 atm, and�-Fe2O3 was com-pletely reduced and transformed into spinel, the limitingcomposition of this spinel is Co2+

0.6Fe2+0.4Fe3+

2O4 basedon stoichiometry. Using observed approximate linear re-lationship between lattice constant and weight fraction ofspinel that coexisted with only hematite in reducing envi-ronment, the lattice parameter of spinel with compositionCo2+

0.6Fe2+0.4Fe3+

2O4 can be extrapolated, and found tobe 8.4918 Å (Fig. 12), consistent with 8.4900(2) Å (shownon Fig. 12) and 8.4947(3) Å (not shown inFig. 12 butshown inFig. 10) refined on last two diffraction patternsthat contained spinel, with corresponding compositionCo0.61Fe0.39Fe2O4 and Co0.36Fe0.64Fe2O4, respectively.

3.4. Spinel → rocksalt transformation

As pO2 further decreases to∼10−15 atm, a new phase withrocksalt type structure begins to grow with a growth rate sim-ilar to that of spinel under pO2 of 10−12 atm, co-existing asthree non-equilibrium phases of spinel, hematite and rocksalt(Fig. 6). At the same time, weight fraction of hematite con-tinuously decreased at the same rate, while weight fractionof spinel, after reaching a maximum, undergoes a plateauwhen rocksalt started to grow (during 44–52 h inFig. 6).It seems that the rocksalt is precipitated from parent spinel

16 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

0

0.2

0.4

0.6

0.8

1

1.2

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66

Time (H)

Wei

gh

t frc

atio

n

Spinel

alpha-Fe2O3

Rocksalt

alpha

gamma

I II III

Fig. 6. Phase evolution and weight fraction of phases present during reduction and oxidation of CoO·2Fe2O3 (C2F) at 900◦C. Vertical lines representthe points where gas mixture was changed, and corresponding measured “equilibrium” pO2 levels for each segment may be referred toFig. 10.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36

Time (H)

Wei

gh

tfr

acti

on

-20

-18

-16

-14

-12

-10

-8

-6

-4

-2

0

log

(pO

2 )

spinelrocksaltgammaalphaAl2O3 from furnace

I II III

Fig. 7. Phase evolution and weight fraction of phases present during reduction and oxidation of CoO·Fe2O3 (C1F) at 900◦C. The superimposed curveis the corresponding measured pO2 values with axis on the right. Note that additional phase of corundum from the furnace is also included.

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 17

Fig. 8. The ratio of Fe:Co in spinel, rocksalt and bulk materials in CoO·2Fe2O3 (C2F), obtained from Rietveld refinements.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56

Time (H)

Sit

eo

ccu

pan

cy Fe(Oct)

Fe(Tetr)

Co(Oct)

Co(Tetr)

Fig. 9. Cation distribution in octahedral and tetrahedral sites of spinel in CoO·2Fe2O3 (C2F) vs. time and pO2.

18 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

8.47

8.475

8.48

8.485

8.49

8.495

8.5

8.505

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66

Time (H)

a(Å

)

-25

-20

-15

-10

-5

0lo

g(P

O2 )

Fig. 10. Variation of lattice constant of spinel in CoO·2Fe2O3 (C2F) vs. time and pO2. Vertical lines represent the points where gas mixtures werechanged, and corresponding pO2 levels are shown by horizontal lines with axis on the right.

-5.00E-04

0.00E+00

5.00E-04

1.00E-03

1.50E-03

2.00E-03

2.50E-03

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56

Time (H)

Rel

ativ

ech

ang

eo

fla

ttic

eco

nst

ant

spinel

a, hematite

Fig. 11. The relative change of lattice parameters of spinel and coexisting hematite vs. time and pO2 during reduction.

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 19

8.473

8.475

8.477

8.479

8.481

8.483

8.485

8.487

8.489

8.491

8.493

0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Weight fraction

a(Å

)

in air

with rocksalt

Fig. 12. Correlation of lattice parameter and weight fraction of spinel in the reduction of CoO·2Fe2O3 (C2F).

phase in the reaction:

(Fe, Co)3O4(s) → (Fe, Co)O(r) + O2 (3)

rather than from�-Fe2O3, since the structure refinement in-dicated that newly formed rocksalt is a Co-bearing phasewith initial composition of (Fe0.40Co0.60)O, and growth rateof spinel phase slowed down due to partial decomposition. Itis concluded that in this three-phase region the two reactions(2) and (3) occur simultaneously: hematite was continuouslyreduced to spinel while some spinel was reduced to rock-salt. Initially the two reactions progressed at the same rateso that the weight fraction of spinel remained unchanged.When pO2 reached∼10−19 atm, rapid growth of rocksaltby reduction of spinel was apparent by structure refinement,and spinel disappeared in 2 h at such pO2 level (Fig. 6).Phase composition of C2F at pO2 ∼10−19 atm was betterpresented inTable 3. The composition of spinel at the startof reduction to rocksalt is Co2+

0.77Fe2+0.23Fe3+

2O4, whichsuggest that the spinel stability field was limited by pO2level instead of composition.

A similar reduction process from spinel to rocksaltwas more clearly observed in C1F sample under pO2 of∼10−19 atm (Fig. 7 andTable 4). Within 1.5 h, spinel wascompletely reduced to rocksalt, followed by the precip-itation of �. Rietveld refinement showed that the initialcomposition of rocksalt phase is roughly (Fe0.6Co0.4)O,suggesting again Co-rich compared with its host. Similar tothe behavior of C2F at pO2 of ∼10−19 atm, reaction kinet-ics of C1F is very fast with respect to the 30 min interval

of recording the diffraction data (Fig. 7). In fact, there areonly six diffraction patterns containing intermediate phases(rocksalt and�) before the C1F sample was completelyreduced to�-(Fe, Co) alloy, which took place in about 4 h(Table 4).

Examination of lattice parameter of rocksalt in C2F, il-lustrated inFig. 13, is reasonably flat in the oxide region,consistent with stable phase chemistry of rocksalt from Ri-etveld refinement (Fig. 8). Significant increase of lattice pa-rameters of rocksalt was observed when metallic phases areprecipitated (see below,Fig. 13), in agreement with an in-crease in Fe concentration in the rocksalt structure (Fig. 8).Note that, with a decrease of pO2, Rietveld refinement basedon ideal structure models resulted in a lower ratio of Fe:Cothan the expected value of 4:1 (Fig. 8), which may suggestthat structure defects do exist in spinel and rocksalt struc-tures under low pO2 atmosphere.

3.5. Precipitation of transient γ phase

Metallic phases (� and �) based on simple fcc and bccstructures began to appear at pO2 of ∼10−19 atm (Figs. 6 and7, Tables 3 and 4). It is unfortunate that Rietveld refinementcannot uniquely determine the Fe and Co compositions inthese metal structures; phase fraction and unit cell contentare correlated as well. To overcome this problem, scale fac-tors of oxide and metallic phases were refined “manually”by adjusting the ratios of Fe:Co in metallic phases until theknown overall Fe/Co stoichiometry was obtained. In many

20 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Table 3Phase distribution (wt.%) and corresponding lattice parameters during reduction (pO2 = 10−19 atm) for CoO·Fe2O3 at 900◦C

Time (h) Spinel Rocksalt � �

↑ 100%, 8.4559(2) Å0.5 60%, 8.4778(3) Å 40%, 4.3244(2) Å1 13%, 8.4834(9) Å 74%, 4.3326(2) Å 12%, 3.5990(8) Å1.5 72%, 4.3416(1) Å 21%, 3.6042(7) Å 7%, 2.8789(4) Å (Fe0.52Co0.48)a

2 53%, 4.3475(2) Å 46%, 2.8834(1) Å (Fe0.59Co0.41)a

2.5 31%, 4.3525(2) Å 69%, 2.8878(1) Å (Fe0.66Co0.34)a

3 14%, 4.3558(3) Å 86%, 2.8909(1) Å (Fe0.70Co0.30)a

↓ 100%, 2.8880(1) Å (Fe0.67Co0.33)

a Alloy composition estimated from composition-dependent lattice parameter from ref.[26].

Table 4Phase distribution (wt.%) and corresponding lattice parameters during pO2 of 10−19 atm in CoO·2Fe2O3 at 900◦C

Time (h) after pO2toward 10−19 atm

Spinel Rocksalt � �

1 22.1%, 8.4947(3) Å 74.7%, 4.3349(1) Å 7.5%, 3.6039(9) Å2 74.3%, 4.3510(1) Å 25.7%, 2.8874(1) Å3 46.2%, 4.3582(1) Å 53.9%, 2.8946(1) Å4 19.2%, 4.3606(1) Å 80.8%, 2.89763(4) Å5 25.6%, 3.6536(1) Å 74.4%, 2.89865(4) Å6 91.5%, 3.6588(1) Å 8.5%, 2.8977(2) Å7 100%, 3.6670(1) Å

2.86

2.88

2.9

2.92

2.94

2.96

2.98

3

3.02

3.04

3.06

3.08

3.1

44 46 48 50 52 54 56 58 60 62 64 66 68 70

Time (H)

a(Å

)

3.6

3.61

3.62

3.63

3.64

3.65

3.66

3.67

3.68

a(Å

)

rocksalt

alphagamma

S+H

oxidation zonereduction zone

Fig. 13. Variations of lattice parameters of rocksalt,� and � vs. time and pO2 during reduction and oxidation in CoO·2Fe2O3 (C2F). The latticeparameters of rocksalt have been scaled according toFig. 16. Other coexisting phases are labeled as S for spinel, and H for hematite.

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 21

cases, metallic phase composition is just fixed to overall sto-ichiometry when it is the predominant phase.

Once metallic phases are formed, elastic strain-inducedlattice variations have to be considered, as in Ni and�-Al2O3system from the reduction of NiAl2O4 [26]. The correla-tion of lattice parameters of rocksalt,� and� is shown inFigs. 13 and 14andTables 3 and 4for C2F and C1F, and fullinterpretation of these experimental findings are difficult be-cause of limited time resolution and possible complex straineffects in metal-ceramic composites as prepared in this wayand discussed by several workers[15,27]. We do not intendto present strain analysis of the system, but several observedfeatures from this study may be useful to better understandunderlying mechanisms since line shifts and broadening ofassociated phases in neutron diffraction patterns would po-tentially provide information on in situ stress and defects.

Exact Fe and Co contents in these alloys are unknown, andalloy composition estimated from composition-dependentlattice parameters alone cannot always give satisfactoryanswers. Particularly, lattice parameters of rocksalt and�shifted in the same direction in the metal-ceramic two-phaseregion in both C1F and C2F (Fig. 13, Tables 3 and 4),which is hard to attribute to composition variation alone. Itimplied that the observed lattice parameters are dependenton a number of factors, including composition, state ofstrain or registrated growth of secondary phase. Examina-tion of FWHM of Fe(1 1 0) showed that all peaks are welldefined, and sharp (all�d/d around 5× 10−3), suggestingstrain was self-relieved in the system.

2.878

2.88

2.882

2.884

2.886

2.888

2.89

2.892

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36

Time (H)

a(Å

)

R+γ

R

R

R

oxidation zonereduction zone

Fig. 14. Variation of lattice parameters of� during reduction and oxidation in CoO·Fe2O3 (C1F). Coexisting phases are labeled as R for rocksalt, and�.

Careful scrutiny of one of the diffraction patterns in C2Frevealed that a tiny amount of� phase was nucleated fromspinel and rocksalt ceramic matrix at pO2 of ∼10−19 atm,and quickly gave way to� form (Fig. 6). Lattice parameterof this � phase was refined to be 3.6039(9) Å, and this alloycomposition was estimated to be close to Fe0.19Co0.81 basedon composition-dependent lattice parameters from[28] withreference lattice constant of 3.6669 Å in alloy compositionFe0.8Co0.2, which was observed in a late stage of the sameexperiment. Using this composition, Rietveld refinement,however resulted in a lower overall Fe:Co ratio (∼2.7:1)than expected from stoichiometry of 4:1. As a matter of fact,a lower bulk Fe:Co ratio was seen before� appeared. Weare aware that exact composition of this� phase is difficultto know because it accounts for only a small fraction oftotal mass and elastic strain-induced lattice change may haveto be considered even though Rietveld refinement indicatedthat this� phase has composition close to Fe0.5Co0.5 basedon the overall stoichiometry. That� phase is Co-rich couldbe supported by the fact that the intensity of this phaseis very weak due to smaller coherent scattering power ofCo.

Similar to C2F, the initial metallic phase reduced fromoxide is� in C1F. Rietveld refinement further indicated thatthe� phase precipitated in spinel and rocksalt matrix in C1F,with lattice constant 3.5990(8) Å, has a composition thatis close to Fe0.5Co0.5 in order to satisfy stoichiometry re-quirement, implying again that Co is preferentially rejectedfrom oxides. The� phase disappeared before it grew to full

22 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Fig. 15. Phase diagram of the Fe–Co binary system[29].

capacity, as was the case in C2F. Note that about 20 wt.% ofthe� transformed to� before rocksalt was fully reduced tometal. Based on equilibrium phase diagram of Fe–Co alloy(Fig. 15), at 900◦C any equilibrium� phase formed mustbe Co-rich in alloy composition.

The epitaxial relationships of metal/oxide and ox-ide/oxide in various systems have been studied includingFe3O4/�-Fe2O3 [30]. Epitaxial growth of FeO during ox-idation of Fe is recognized[31], and it is assumed thatthere are similar nucleation phenomena in reduction of (Fe,Co)O. Once metal phase began to form, it is expected thatthe crystal lattices of unreduced oxides and metal phasewere distorted due to volume shrinkage caused by the re-duction and mismatch of lattice parameters in the interfaceof rocksalt/metal. Whatever metallic phase formed is tominimize total free energy of the system. As can be noticedfrom Fig. 16, showing the lattice relationships among�,� and underlying rocksalt structures,� and � have latticemisfit (�) of 6 and 17%, respectively, in respect to rocksaltlattice. Although�, � and rocksalt have identical atomicconfiguration on various crystallographic planes, consid-ering incoherent lattice strain alone, bcc would provide asmaller metal/rocksalt interface energy than fcc. The forma-tion of � in ceramic matrix will be associated with relativelyhigher elastic strain energy. Since this term is proportionalto the volume of� phase in the matrix, and it would beinsignificant in initial stages.

Metal cations in rocksalt structure and reduced metal sharesimilar structural units, and only slight shifts in metal–metaldistances are required to match the two structures (Fig. 16).The layered structure of SFC2 contains both perovskite layerand rocksalt layers. Much of the behaviors of rocksalt lay-ers could be derived from current study on the reduction ofrocksalt structures. That is, the framework of metal cationsis very robust, thus absorbing and releasing oxygen basi-cally do not result in significant rearrangement of cationpositions but only valence transformation. Violent changeof phase composition of the materials under study in pO2of ∼10−19 atm, with predominant phase of spinel, rocksalt,and metals, respectively, over a relatively short time period(Figs. 6 and 7, Tables 3 and 4) does not necessarily meansignificant change in the skeleton of the underlying struc-tures because of similarities in atomic arrangements. How-ever, one may consider the effect of formation of metallicphases on SFC2 composites due to different physical prop-erties of oxides and metals.

3.6. Transformations of γ → α → γ in C2F and γ → α

in C1F

In both experiments,� can not be fully developed (formedonly in a transient sense), and transformed to� form (Figs. 6and 7) with the development of reduction. In the C2F sam-ple, shortly after� reached∼70 wt.%, it started to transform

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 23

4.33

3.602.88

∆ = 17%∆ = 6%

Å

Å Å

(a)

(b) (c)

Fig. 16. The structure of (a) rocksalt, (b)�, and (c)�. The filled black circles represent Fe/Co cations or atoms, and unfilled circles represent oxygen anions.

to �, and the final reduced product in C2F is�-Fe0.8Co0.2.In the C1F sample,� remained as the final reduced phasewith composition of�-Fe2/3Co1/3 based on the overall sto-ichiometry.

Since in binary alloy systems the equilibrium phase iscomposition-dependent, the observed� → � transforma-tion in both experiments could be due to an increase of Feconcentration in alloy, which will stabilize the low temper-ature� phase, as suggested in the phase diagram of Fe–Cobinary alloy. It is unclear whether the phase transformationis simply due to strain relaxation since the total energy ofsystem may be dominated by the lattice strain energy as thevolume of� phase increases, and stress would be reducedby the� → � phase transformation.

Again referring to the Fe–Co phase diagram, the trans-formation temperature of� ↔ � is 912◦C for pure iron,and this temperature is raised with increase of Co content.Since the experimental set temperature of 900◦C is nearthe �/� transformation boundary, several factors needed tobe considered. At 900◦C, the two alloy compositions bothshould fall into the region of low temperature� phase. Itis possible that our set temperature is∼15◦C lower thanthe real temperature (i.e. the sample temperature is as highas 915–930◦C). If this is the case, alloy compositions ofFe0.8Co0.2 and Fe2/3Co1/3 in final reduction products of C2Fand C1F should result in� and� single phase, respectively,consistent with trend in equilibrium phase diagram of Fe–Coalloy and with our observations. It is worth mentioning thatthe temperature in our experiments is well controlled within

±5◦C, and the possibility of phase transformation due totemperature fluctuation can basically be ruled out.

3.7. Structure defects in the structure of rocksalt

Oxygen occupancy in the rocksalt structure was re-fined, but the results do not support oxygen vacancy modelwithin the precision of the refinements. Since the mostlikely defects in rocksalt type structures are cation defi-ciencies and cation interstitials, once structure refinementswere completed based on ideal stoichiometric composition,difference-Fourier maps from rocksalt data in various stagesof reduction of C1F samples were systemically examinedin order to locate possible extra atoms, with particularattention on tetrahedral sites in the structures. In fact, a pre-viously postulated transition from cation-deficient wustitestructure to rocksalt[22] was supported by evidence thatdifference neutron density (�ρ) on (1/4, 1/4, 1/4) positionon difference Fourier maps decreased with the reduction(Fig. 17a–c). The occupation of this tetrahedral site, pre-sumably by Fe3+, requires three adjacent octahedral sitesto be vacant[32], resulting in a formula (Fe, Co)1−xO. It isinteresting to note that peak density appeared in a new po-sition (1/4, 1/4, 0) in rocksalt that is coexisting with�(bcc)phase (Fig. 17d–f), which might reflect some artifacts ofreduced metal lattice.

The oxygen atoms in both spinel and wustite are inface-centered cubic close packing, and the major differencebetween the two structures lies in the cation arrangement

24 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Fig. 17. Difference-Fourier map atz = 1/4 for the structures of rocksalt. Horizontal and vertical lines are the [1 0 0] and [0 1 0] direction, respectively,each with map size of 8 Å, and centered at (0 0 0). (a–f) correspond to rocksalt phases from top to bottom inTable 4, respectively. The numbers besidecontour lines in the maps are�ρ.

on available tetrahedral and octahedral sites. The structureof wustite Fe1−xO resembles that of spinel Fe0.75O, andtheir structure relationship was demonstrated by an elegantcomparison, in which lattice parameter of Fe0.75O wouldbe extrapolated by a plot of twice the lattice parameters ofFe1−xO versus vacancy concentration[13]. The structureof wustite/rocksalt is extremely important in maintainingcontinuous structural pathways over the entire pO2 rangeas in one side, it is related to oxidation phase spinel, andon the other side it is associated with reduction productmetal.

3.8. Oxidation behavior of C1F and C2F

C2F and C1F samples showed quite different behaviorswhen the resulting metallic alloys were exposed to air withpO2 of ∼10−0.9 atm. In C2F,� → � transformation wassoon recorded, followed by the formation of rocksalt, andthen spinel (Fig. 6). At isothermal condition, pO2-inducedphase transformation was not reported before in the Fe–Cosystem, and is peculiar. It could be related to the formationof rocksalt on the surface of the alloy. However,� phaseremained in C1F although the cell length of� was slightly

Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26 25

decreased within the precision of structure refinement (Fig.14), and further study is needed to determine if there is anydissolved C in reduced� phase.

In contrast to C2F, significant shrinkage was found in C1Fduring both reduction and oxidation periods, as evidencedby increasing Al2O3 (furnace component) mass abundancein diffraction pattern with time evolution (Fig. 7). Shrink-age in C1F could be associated with higher grain-boundarymobility due to purer phases. Interestingly, alloys in C1Fdid not show oxide phases even after∼15 h in air except forlimited oxidation on the surface. Based on oxidation modes

Fig. 18. SEM micrograph of polished surface of interior of the samplesafter neutron experiments for (a) CoO·Fe2O3 (C1F), and (b) CoO·2Fe2O3

(C2F).

established previously, it seems that denser oxide layers onthe surface of C1F prevent oxygen diffusion inward.

The phase composition on both the surface and interior ofC1F and C2F post-neutron experiment were confirmed byX-ray diffraction. As expected from different textures in sur-face and interior of samples and from neutron data,� is themajor phase in the interior of both samples although a smallamount of an unidentified secondary phase and spinel wasalso present in C1F and C2F samples, respectively, suggest-ing that internal oxidation occurred in C2F. On the surfaceof the C2F sample, spinel was identified as the major phasealong with an unidentified secondary phase. According to thephase diagram, an ordered phase�′ would be formed dur-ing the cooling of C1F sample, but this disorder to orderedphase transformation was not evident in X-ray diffractiondata. Internal strains could suppress the phase transforma-tion as discussed in the Fe–Co alloy system of Lee et al.[33].

Typical microstructures of C1F and C2F samples afterneutron experiments were shown inFig. 18(a) and (b),respectively. SEM micrograph indicated that C1F samplecontained only isolated pores separated by alloy grains, incontrast to C2F sample, where pores are largely intercon-nected. In both cases, irregular grain boundary of alloys wasnoted. The discrepancy between higher porosity calculatedbased on only alloy phase and SEM observation on theC2F sample could indicate that some of the alloy was oxi-dized in the interior of the C2F sample, which is supportedby both neutron and X-ray diffraction data. Average EDSanalysis on selected area resulted in an overall compositionof Fe:Co ratio of 1.9:1 for C1F and 4.0:1 for C2F, respec-tively, suggesting no significant loss of Co in reduction. Itis remarkable that oxide layers did not detached from theinside when cooling to room temperature.

4. Conclusions

1. In situ neutron diffraction has been successfully per-formed at high temperature and controlled atmosphereto probe the structural and chemical pathways of Fe–Coceramic oxides, especially Fe and Co cation distributionin oxide phases.

2. The following sequences of structural transformationfrom �-Fe2O3 hematite→ (Fe, Co)3O4 spinel→ (Fe,Co)1−xO wustite→ (Fe, Co)O rocksalt→ �-(Fe, Co) →�-(Fe, Co) → (�-(Fe, Co)) were observed on the re-duction of Fe–Co–O based ceramic oxides; the finalproduct of reduced alloy is dependent on temperatureand composition as predicated by phase diagram.

3. Kinetics of reduction of spinel and rocksalt to metallicalloys are very fast at 900◦C and in pO2 of ∼10−19 atm,and metallic phases will be unavoidably formed at the re-ducing side of a membrane if there is no efficient oxygendiffusion from the oxidizing side of the SFC2 membrane.

4. By preferentially rejecting Co, each intermediate reduc-tion product in the Fe–Co–O system is usually initially

26 Y. Li et al. / Materials Science and Engineering B106 (2004) 6–26

Co-rich compared with the composition of parent mixedoxide phases, and then progressively increases in Fe con-centration.

5. Initially formed rocksalt has wustite-type structure, withcations in both octahedral and tetrahedral sites, a relicfeature of its precursor spinel.

Acknowledgements

This work has been benefited from the use of the In-tense Pulsed Neutron Source at Argonne National Labora-tory. This facility is funded by the US Department of Energyunder Contract W-31-109-ENG-38.

References

[1] U. Balachandran, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L.Mieville, P.S. Maiya, M.S. Kleefisch, S. Pei, T.P. Kobylinski, C.A.Udovich, A.C. Bose, Am. Ceram. Soc. Bull. 74 (1995) 71.

[2] S. Guggilla, A. Manthiram, J. Electrochem. Soc. 144 (1997) L120.[3] S. Kim, Y.L. Yang, R. Christoffersen, A.J. Jacobson, Solid State

Ionics 109 (1998) 187.[4] R. Bredesen, T. Norby, A. Bardal, V. Lynum, Solid State Ionics 135

(2000) 687.[5] B. Ma, U. Balachandran, J.H. Park, J. Electrochem. Soc. 143 (1996)

1736.[6] B. Ma, U. Balachandran, Mater. Res. Bull. 33 (1998) 223.[7] B. Ma, J.P. Hodges, J.D. Jorgensen, D.J. Miller, J.W. Richardson,

U. Balachandran, J. Solid State Chem. 141 (1998) 576.[8] A. Yoshiasa, K. Ueno, F. Kanamaru, H. Horiuchi, Mater. Res. Bull.

21 (1986) 175.[9] B.J. Mitchell, J.W. Richardson, C.D. Murphy, B. Ma, U. Balachan-

dran, J.P. Hodges, J.D. Jorgensen, Mater. Res. Bull. 35 (2000)491.

[10] B.J. Mitchell, J.W. Richardson, C.D. Murphy, B. Ma, U. Balachan-dran, J.P. Hodges, J.D. Jorgensen, J. Eur. Ceram. Soc. 22 (2002) 661.

[11] B.J. Mitchell, R.C. Rogan, J.W. Richardson, B. Ma, U. Balachandran,Solid State Ionics 146 (2002) 313.

[12] S. Pei, M.S. Kleefisch, T.P. Kobylinski, J. Faber, C.A. Udovich, V.Zhang-McCoy, B. Dabrowski, U. Balachandran, R.L. Mieville, R.B.Poeppel, Catalysis Lett. 30 (1995) 201.

[13] F. Koch, J.B. Cohen, Acta Cryst. B25 (1969) 275.[14] R. Subramanian, S. Tinkler, R. Dieckmann, J. Phys. Chem. Solids

55 (1994) 69.[15] E. Ustundag, P. Ret, R. Subramanian, R. Dieckmann, S.L. Sass,

Mater. Sci. Eng. A195 (1995) 39.[16] B.J. McBride, S. Gordon, Computer Program for Calculation of Com-

plex Chemical Equilibrium Compositions and Applications, NASAReference Publication 1311, 1996.

[17] A.C. Larson, R.B. Von Dreele, Report No. LA-UR-86-748, LosAlamos National Laboratory, Los Alamos, NM, 1987.

[18] T.G. Worlton, P.F. Peterson, A. Chatterjee, J.P. Hammonds, D.J.Mikkelson, R.L. Mikkelson, Integrated Spectral Analysis Work-bench software (ISAW), available fromhttp://www.pns.anl.gov/computing/isaw/.

[19] F.H. Lu, S. Tinkler, R. Dieckmann, Solid State Ionics 62 (1993) 39.[20] J. Robin, Ann. Chim. 10 (1955) 393.[21] A.J. Rondinone, A.C.S. Samia, Z.J. Zhang, J. Phys. Chem. B103

(1999) 6876.[22] I.V. Murin, V.M. Smirnov, G.P. Voronkov, V.G. Semenov, V.G. Po-

varov, B.M. Sinel’nikov, Solid State Ionics 133 (2000) 203.[23] L.C. Varanda, M. Jafelicci, P. Tartaj, K. O’Grady, T. Gonzalez-

Carreno, M.P. Morales, T. Munoz, C.J. Serna, J. App. Phys. 92(2002) 2079.

[24] L.B. Pankratz, Thermodynamic properties of elements and oxides,US Bureau of Mines, January 1983.

[25] F. D. Bloss, Crystallography and Crystal Chemistry, MineralogicalSociety of America, 1994, pp. 545.

[26] E. Ustundag, B. Clausen, M.A.M. Bourke, App. Phys. Lett. 76 (2000)694.

[27] R. Subramanian, E. Ustundag, S.L. Sass, R. Dieckmann, Mater. Sci.Eng. A195 (1995) 51.

[28] I. Ohnuma, H. Enoki, O. Ikeda, R. Kainuma, H. Ohtani, B. Sundman,K. Ishida, Acta Mater. 50 (2002) 379.

[29] T. Nishizawa, K. Ishida, in: T.B. Massalski (Ed.), Binary Alloy PhaseDiagrams, vol. 2, second ed., ASM International, 1990.

[30] R.A. Fellows, A.R. Lennie, H. Raza, C.L. Pang, G. Thornton, D.J.Vaughan, Surface Sci. 445 (2000) 11.

[31] J. Benard, Metall. Rev. 9 (1964) 473.[32] W.L. Roth, Acta Cryst. 13 (1960) 140.[33] B. Lee, B.S. Ahn, D. Kim, S. Oh, H. Jeon, J. Ahn, Y.D. Kim, Mater.

Lett. 57 (2003) 1103.