Structure and Catalytic Activity of Iron Oxide andMagnesium Oxide Solid SolutionsPart 3.-E.S.R. CharacterizationB Y DANTE CORDISCHI, FRANCO PEPE, MARIO SCHIAVELLO AND MARIO VALIGICentro di Studio su “ Struttura ed attivitii catalitica di sistemi di ossidi ”,Istituto Chimico, University of Rome, Rome, ItalyandIstituto di Chimica Generale, University of Rome, Rome, ItalyReceived 3rd June, 1976Magnesium oxide + iron oxide, fired both in air and in a reducing atmosphere, and magnesiumoxide+iron oxideflithium oxide (up to 1 % atomic Fe) were investigated by e.s.r. spectroscopy.The effects of outgassing at various temperatures and of NzO decomposition on the e.s.r. spectraare discussed in terms of surface redox processes. The incipient formation of the spinel phase,MgFe204, and its precipitation, at the highest outgassing temperatures adopted, were readily studiedby the e.s.r.technique.The hypotheses previously proposed on the modification of the catalyst solid state, occurringduring N20 decomposition andlor in the vacuum treatment, are confirmed by the present study.Further details are also given.Electron spin resonance (e.s.r.) spectroscopy, which is commonly used for charac-terizing polycrystalline materials,l is also, due to its great sensitivity, a useful tool inthe field of heterogeneous catalysis, particularly when the catalyst is chemically alteredduring the course of the catalytic reaction. An example is the CuO + MgO systemin which e.s.r. results have been correlated with the catalytic activity for N20 de-corn posit ion.In this paper we report the results of an investigation on the system magnesiumoxide+iron oxide, fired in air or in a reducing atmosphere, and on the system mag-nesium oxide + iron oxide + lithium oxide.The structure and the catalytic propertiesfor N20 decomposition of such systems have been investigated previously,3* asolid state process being observed during the conditioning and catalytic procedures.The aim of this study was, therefore, to obtain a more detailed description of thesystem and to test the validity of some hypotheses proposed for the course of thecatalytic process. For these reasons the e.s.r. measurements were performed on the“ as prepared ” specimens as well as in conditions simulating the catalytic ones.EXPERIMENTALMATERIALSThe samples containing only iron were prepared by calcination of “ Specpure ” MgO(Johnson-Matthey) impregnated with Fe(NO,), solution at 1273 K for 5 h, in an oxidizing(air) or reducing (COz+ CO mixture giving an oxygen partial pressure of 1 .O x lo-’ N m-2)atmosphere, For samples containing both iron and lithium, a LiN0, solution was also6D.CORDISCHI, F. PEPE, M. SCHIAVELLO AND M. VALIGI 63added in the impregnation step ; they were fired at 1273 K for 5 h in air. Details of samplepreparation have been reported previ~usly.~Samples containing only iron are labelled MF and MF-R if prepared in air or in a reduc-ing atmosphere, respectively. Those also containing lithium are designated MFL. Thenumbers after the letters give the nominal concentration of iron and lithium with respect to100 Mg atoms.MgFe204 was prepared from MgO (from the carbonate decomposed in air at 873 K)impregnated with Fe(NO& solution in the correct stoichiometric ratio. The product wasfired in air at 1273 K for 5 h and cooled rapidly to room temperature.The formation and purity of the ferrite phase was tested by X-ray analysis.TREATMENTSA portion of the sample (0.1-0.3 g) was enclosed in a silica ampoule, having a standarde.s.r.tube as a side arm, connected to a vacuum system via a ground joint and isolated bymeans of a stop-cock. Treatments consisted of cycles of outgassing and subsequent N20addition. The N20 was added at 8000Nm-2 (60Torr) with the sample maintained at723 K ; outgassing was performed at various temperatures starting from 753 K.After each treatment the powder was transferred to the e.s.r.tube for measurements.E . S . R . PROCEDUREThe e.s.r. spectra were recorded at X-band frequencies on a Varian E-9 spectrometer atroom temperature or, occasionally, at 77 K.The absolute concentration of Fe3f ions was obtained by measuring the area under spectraintegrated electronically over 4000 G, using a single crystal of CuS04 5H20 as a standard.RESULTS AND DISCUSSIONSPECTROSCOPIC CHARACTERIZATION OF UNTREATED SAMPLESSAMPLES CONTAINING IRON ONLY FIRED I N AIR (MF)As reported previously this system contains two phases, MgO and the ferromag-netic MgFe,O,.Ferromagnetic resonance in powdered samples has been studied in only a fewcases,1 mainly due to the fact that the resonance field depends on crystal orientationas well as on the shape and porosity of the specimen.The purpose of the present investigation is not the study of the ferromagneticresonance of the ferrite, but rather the use of e.s.r.to detect the presence of thisphase.The e.s.r. spectra of the MF samples are very similar to those of the pure MgFe,O,[fig. l(a), (b), (c)] and to the calculated 0nes.l The main spectral feature of thesesamples is a very strong and broad signal whose line-shape varies slightly with sample,the amount of powder and the microwave power. Nevertheless, the overall linewidth (AH,,) at room temperature is fairly constant (800-1000 G). At 77 K abroadening by a factor of two occurs.The intensity of this signal is roughly pro-portional to total iron content but is so high that, with the exception of the mostdiluted samples (MF O.l), it obscures all other signals. Thus, due to the very strongintensity and in spite of the very broad linewidth, the ferrite phase can be detectedeasily by e.s.r. in very dilute samples (MF 0.1) in which X-ray investigation failed todetect it or in samples (with lithia, see below) in which this phase is present as a minorcomponent.The spectrum of the most dilute sample (MF 0.1) also shows a signal due toFe3+ ions in solid solution in MgO. The spectral feature of this signal will be describedbelow64 E.S.R. OF2000 3000 4000 5000HIGFIG. 1IRON OXIDE' 2000 3000 4000 500dHIGFIG.2FIG. 1.-X-band e.s.r. spectra at room temperature of untreated samples of (a) MF 0.5 ; (b) MF 1 ;(c) MgFez04 ; (d) MFL 1 : 0.5 ; (e) MFL 0.05 : 0.5 ; (f) MFL 0.5 : 0.25 ; (9) MFL 1 : 5. The num-ber under each spectrum indicates the relative gain.FIG. 2.-A'-band e.s.r. spectra at room temperature of MF-R 0.5 after various treatments : (a) un-treated sample; (b) after the first vacuum treatment at 753 K (treatment no. 1 of fig. 3 and 4);(c) after vacuum at 823 K (treatment no. 13) ; (d) after vacuum at 873 K (treatment no. 15) ; (e) aftervacuum at 973 K (treatment no. 19). In spectra (a)-@) the signals from impurities MnZ+ and Cr3+are aIso present (narrow peaks). The number under each spectrum indicates the relative gain.SAMPLES CONTAINING BOTH IRON A N D LITHIUM FIRED I N AIR (MFL)In these samples Fe3+ and Li+ are in solid solution in MgO when the ratio [Li]/[Fe] 3 1 ; MgFe,O, is also present when [Li]/[Fe] < l.3 In all the samples with[Li)/[Fe] < 1 the strong band of the ferrite which dominates the e.s.r.spectrum ispresent. In the samples at [Li]/[Fe] = 0.5 [fig. l(d)] it is also present with an inten-sity of -20 % of that of the sample without lithia having the same iron content.As [Li]/[Fe) increases the other signal, observed only in the dilute MF samples(MF O.l), progressively increases in intensity. This is the same signal generallyobserved in commercial undoped MgO samples of normal purity, fired in air, in whichiron is the most common paramagnetic impurity.This signal can be assigned toFe3+ ions in solid solution in MgO. At low iron contents (0.05-0.1) and with [Li]/[Fe] > 1, this signal shows all its details [fig. l(e)]. It consists of a central line ofunresolved structure at g = 2.00 with several shoulders on each side. These shouldersare not normally observable in undoped samples containing iron at impurity level,because their intensity is lower by a factor of - 10 with respect to the central lineD. CORDISCHI, F . PEPE, M. SCHIAVELLO AND M . VALIGI 65The complex structure of the e.s.r. signal of the Fe3+ ion in powder samples ofMgO can be completely interpreted from the known parameters of the spin Hamilton-iaii of Fe3+ in MgO, obtained from a single crystal study (see Appendi~).~ Theresolution of the spectrum depends on several factors, such as total iron concentrationand [Li]/[Fe] ratio.In the less resolved spectra only the most intense shoulders areobservable [see fig. l(f), (g)]. However, the characteristic shape of the central peak,its linewidth and the presence of the shoulders are a clear indication of the presenceof Fe3+ ions in solid solution in MgO in sites of cubic symmetry. Since these featuresare visible in all MFL samples up to 1 % iron content (and in the most dilute MF),we conclude that isolated Fe3+ ions in sites of cubic symmetry are present in MgO.Table 1 reports the absolute Fe3+ concentrations, evaluated from integrated spec-tra, for all samples fired in air. In the same table the linewidths of the ‘‘ cubic ”central peak and of the ferrite signal (when present) are also reported.TABLE 1.-E.S.R.DATA OF MF AND MFL SAMPLES FIRED IN AIRsampleMF 0.1MF 0.5MF 1MgFeaO4MFL 0.05 : 0.05MFL 0.1 : 0.05MFL 0.1 : 0.1MFL 0.1 : 1MFL 0.5 : 0.25MFL 0.5: 0.5MFL 0.5 : 2.5MFL 1 : 0.5MFL 1 : 1MFL 1 : 5Fe3+ in solid solution Fe3+ as MgFezO4AHPP Fe+3 * AH,, %4545454545454748464760Yes 1100 100950 1001050 1001100 1000.05 - -Yes 880 200.080.10 - -Yes 940 200.33 - -0.55 -Yes 1040 200.52 - -0.91 - -I --a Fe*j ions per 100 M g atoms.The following points are relevant to these data :(a) When the ferrite signal is present no reliable quantitative evaluation of theFe3+ concentration is possible, only the presence of the cubic signal being indicated.In all MF samples it has been assumed that all Fe3+ is present as ferrite.In theMFL samples in which the ferrite signal is present, the ferrite content has been roughlyestimated from the intensity of its signal relative to that of the corresponding MFsample.(b) In the MFL samples, the ferrite signal is absent only when the ratio [Li]/[Fe] 2 1. Only in the samples with [Li]/[Fe] > 1 does the Fe3+ concentration,(estimated error -20 %) agree with the nominal values. Since from analytical andmagnetic data,3 all iron is present as Fe3+ in the air-fired samples, when the [Li]/[Fe] = 1 some Fe3+ ions escape e.s.r. detection. Probably these ions are near to, or on, thesurface, giving a very broad signal because of the large anisotropic interactions.These results show that firing at 1273 K in air induces a uniform distribution ofFe3+ in the MgO matrix only when a large excess of Li+ is present.1-66 E.S.R.OF IRON OXIDESAMPLES CONTAINING IRON FIRED I N REDUCING ATMOSPHERE (MF-R)Previous results have shown that samples without lithia, prepared in a reducingatmosphere (CO+CO,) at 1273 K, (MF-R), were solid solutions with the iron inthe Fe2+ oxidation ~ t a t e . ~ Their e.s.r. spectra show the signal of “cubic” Fe3+ions, which is of low intensity. In addition, signals due to Mn2+ and Cr3+, presentas impurities in the MgO used, are observed [fig. 2(a)]. The intensity of the “ cubic ”Fe3+ signal is independent of the total iron content and corresponds to a smallfraction of the total iron present.The previous results are thus ~onfirrned,~ the Fe3+ content being well under thelimit of sensitivity of the analytical method.Also, at room temperature only Fe3+can be detected by e.s.r., whereas Fez+ (a d6 ion) can be seen in MgO only at liquidhelium temperatures, because of its short relaxation time.6EFFECTS OF TREATMENTS I N VACUO AND/OR IN N,QM F AND MFL SAMPLESThe treatments in vacuo and in N20 carried out on MF and MFL samples, pre-pared in oxidizing conditions, were as follows. The MF 0.5 specimen was treatedin vacuo at 753 K for 5 h, then in an N,O atmosphere (723 K ; P = 8000 N rn-,).The cycle was repeated and finally the specimen was outgassed at 923 K.The specimens MFL 1 : 0.5, MFL 0.1 : 0.05, MFL 0.1 : 0.1 and MFL 0.1 : 1 wereoutgassed at 873 K, exposed to NzO (723 K, P = 8000 N m-2) and finally evacuatedat 973 K.In general all these treatments have a small effect on the e.s.r.spectrum of oxidizedsamples. In particular vacuum treatment at the highest temperatures tested (923 K)causes small variations in lineshape of the ferrite signal (when present) without affect-ing its intensity. The effect of the treatments on the “ cubic ” signal is negligible inspecimens with [Li]/[Fe] 2 1 and is small and rather erratic in those with [Li]/[Fe] < 1.MF-R SAMPLESThe effect of treatments is quite large on reduced samples (MF-R). In generalthe vacuum treatment causes a decrease in the intensity of the “ cubic ” Fe3+ signal,while the N,O treatment causes an increase.The vacuum-N,O cycle was repeatedseveral times on the same sample, progressively increasing the temperature of thevacuum treatment but maintaining a constant temperature (723 K) for the N 2 0treatment. Fig. 3 shows the variation of the cubic Fe3+ signal intensity, evaluatedfrom the height of the central peak. When the temperature of the vacuum treatmentis relatively low (753 K) the main effect on the e.s.r. spectrum is a variation in intensity,without any significant alteration in the resolution [fig. 2(b)]. The concentration ofFe3+ evaluated from the integrated spectrum, remains rather low (a few percent oftotal iron) (fig. 4).Increasing the temperature of the vacuum treatment to 823 K provokes largervariations of the cubic signal intensity in the vacuum-N20 cycle (fig.3) and theappearance of another broad signal (AH,, - 120 G), which becomes progressivelymore intense [fig. 2(c)]. For vacuum treatment at 873 K this signal dominates thespectrum [fig. 2(d)]. The concentration of Fe3+ from the integrated spectrum nowcorresponds to a large fraction (up to 100 %) of the total ironD. CORDISCHI, F. PEPE, M. SCHIAVELLO AND M. VALIGII67+.-.- 753K -i823K+1 , 1 , , , 1 , 1 , , , , , , t0 4 8 12 16treatment numberFIG. 3.-Intensity of the central cubic peak (from first derivative spectra) of the samples MF-R 0.5(- - - -) and MFR 1 (-) after 0, 0 NzO treatment at 723 K, 4 h ; @, vacuum treatmentfor 12 h at the temperatures indicated.I7 4 II9 73 K-rnIIIIIII+ 753 K-----r-l 4823 K+ lw873 K H/8- - ./u,n-.- -&n-m- - - - -I 0-s0 4 8 12 16 20treatment numberFIG.4.-Integrated intensity of MF-R 0.5, after various treatments : the meaning of the symbols isthe same as for fig. 3.I l l l l r , l l l , l l , , , 68 E.S.R. OF IRON OXIDEThe largest variations in the intensity of this broad signal occur after the firstvacuum treatment at a higher temperature. In contrast to the behaviour of the*‘ cubic ” signal, the broad signal is insensitive to the subsequent N 2 0 treatment.By increasing the temperature of the vacuum treatment to 973 K the e.s.r. spn d r u mbecomes very intense and is similar to that of the MF samples ; it is, as discussed be-fore, assigned to the phase MgFe204.The ferrite phase has, in fact, been detccted’oy X-ray analysis of these samples at the end of the treatments.In a different set of treatments on the same samples (MF-R 0.5 and MF-R I),starting from 823 K (instead of 753 K as in the experiments of fig. 3) for the firstvacuum treatment, the broad signal develops only after the vacuum treatment at873 K and has an intensity an order of magnitude lower. Therefore, the repzatedvacuum-N,O cycles at 753 K not only cause reversible variations in intensity of thecubic Fe3+ sigiial (fig. 3), but also appear to assist the subsequent irreversiblc forma-tion of the broad signal.This broad signal can be assigned to associated Fe3+ ions in the ferrite phase. Theincrease in intensity of this signal does not indicate only the progressive oxidationof Fe2+ ions, but mainly the ordering of the Fe3+ ions into the ferrite phase.Infact, the intensity of the broad signal increases after the vacuum treatments and notafter the N20 treatments (fig. 4).When the temperature of the vacuum treatment is relatively low (up to 873 K),the mobility of ions and vacancies is very limited and the ferrite phase is rather dis-ordered and with extremely small particles. Only at 973 K is the ferrite well formed.These conclusions are in agreement with the results on diluted samples sintered in airat high temperature (1673 K) in which a large fraction of iron is in solid solution inMgO as Fe3+.’*CONCLUSIONThe results of the e.s.r. study confirm those previously obtained 3 9 and providefurther details.The MF system, in which the Fe3+ ions are almost completely present as MgFe204,was found to be rather inactive for N20 decomy~sition.~ Its inactivity was attributedto the difficulty of Fe2+ ion formation, the presence of which is essential for thecatalytic reaction to occur.The e.s.r. results confirm the resistance to reduction ofFe3+ ions in the ferrite phase.For the MF-R system, the highest activity was found for the same reactio!~.~The activity was attributed to the presence of Fe2+ ions which, during N,O decorn-position, were found to oxidize to Fe3+. In the subsequent outgassing treatment,the Fez+ ions were restored on the surface while the structure of the ferrite phasebuilt up.At the highest temperatures used, precipitation of the ferrite phase wasobserved. All these findings are consistent with the results of the present work.For the MFL system with [Li]/[Fe] > 1, the low activity was attributed to the diffi-culty of reducing Fe3+ ions in MgO and in the presence of excess Li20.4 This isfully confirmed ir, this study since the “ cubic ” Fe3+ signal is not affected appreciablyby the treatments simulating the catalytic conditiom.The high activity of the MFL specimens with [Li]/[Fe] < 1 was ascribed to thereduction of Fe3+ ions in MgO, the outgassing process causing lithium oxide loss fromthe latticee4 However, the expected decrease in the cubic Fe3+ signal was not alwaysobserved in e.s.r. measurements. The erratic behaviour of the intensity variationin this case does not allow any definite conclusions to be drawnD.CORDISCHI, F. PEPE, M. SCHIAVELLO AND M. VALIGI 69APPENDIXTheiFe3+ ion in MgO has a high cubic zero field splitting (a = +205 x lo-' ~ m - ~ ) , 'which gives an angular dependence in cubic symmetry and a coiriplex powder spectrum.Being at X-band frequencies, la1 < gpH, the perturbation theory can be used to findthe resonance conditions for all the transitions. The formulae for allowed transitions(AM = & l), to second order, given by Abragam and Bl~aney,~ areU &3 c+ ++ liv = gPH+2pa+(2-9p+7p2)-15gPHwhere the parameter p is related to the cubic potential : *p = (l4+m4+n4-3) and2, m, n are direction cosines of the static magnetic field with respect to the cubic axes.The extreme values of p are + 1 along a (loo} axis and - 3 along <I 1 1> ; anotherstationary value is -$ along (01 1).As seen from formulae (l), the transition + 3 t-) - 3 depends on a only to secondorder, while in the other transitions a fbst order term is also present.The e.s.r.signals of a powder sample in the normal first derivative presentation appear at fieldsat which a discontinuity in absorption occurs. This happens generally at the extremeand stationary values of the field for each transition.Lineshape calculations of the " central peak " have been reported previously. *For this transition the second order term gives the extreme values of the field at40 ar2 Hmin = HO- --27 Hofor p = - 3wherea hva ' = - and Ho = -SP SP'With the known values of a' = 219 G, g = 2.0037 and Ho = 3370 G (under ourexperimental conditions) we obtain AH = H,,, - Hmin = 45.0 G which is the observedwidth of the '' central peak ".For the other transitions, as the first order term, linear in p , is dominant withrespect to the second order term, the observed signals outside the central line willcorrespond merely to the extreme and stationary values of p .Without second ordercorrection the spectrum would show a symmetrical pattern ; the observed asymmetryis due to the second order effect.In table 2 the fields for the particular values of p , calculated from formulae (I),are reported. The calculated values agree quite well with the observed field positionsof the shoulders. As can be seen in fig. l(e), those belonging to +$ c+ -+ f transitionsare more intense than the corresponding ones belonging to -k3 f-) +$ transitions atldthe pair belonging to + 4 f-) ++, and p = - 3 is the most intense70 E.S.R. OF IRON OXIDETABLE 2.rALCULATED AND OBSERVED VALUES OF FIELD POSITION OF THE SHOULDERS OF E.S.R.SPECTRUM OF Fe3+ IN MgOP 1 1 -3 -3 -a -t -a -a -3 -+ 1 1transition-+*-* +%*+a ++u++ -*--+ +$c*+& -+w-s +Q.H+* -+--a +;-u+% -+-+ -++--3 ++u.Hce.10. 2818.3 2931.7 3015.3 3067.3 3220.5 3256.0 3475.0 3494.5 3651.6 3724.7 3808.3 3913.5Hobs. 2823 2935 3016 3066 3215 - - 3498 3645 3730 3800 3905The authors thank Prof. A. Cimino for valuable discussions and Dr. M. Petrerafor help given in some experiments.P. C. Taylor, J. F. Baugher and H. M. Kriz, Chem. Rev., 1975,75,203.D. Cordischi, F. Pepe and M. Schiavello, J. Phys. Chem., 1973,77, 1240.M. Valigi, F. Pepe and M. Schiavello, J.C.S. Faraday I, 1975, 71, 1631.M. Schiavello, M. Valigi and F. Pepe, J.C.S. Faraday Z, 1975, 71, 1642.W. Low, Proc. Phys. SOC. B, 1956,69, 1169.J. W. Orton, Electron Paramagnetic Resonance (Ilife Books, London, 1968), p. 200.G. P. Wirtz and M. E. Fine, J. Amer. Ceram. Soc., 1968, 51, 402.K. N. Woods and M. E. Fine, J. Amer. Ceram. SOC., 1969, 52,186.A. Abragam and B. Bleaney, EIectron Paramagnetic Resonance of Transition Ions (Clarendon,Oxford, 1970), p, 147.lo J. H. Lunsford, J. Chem. Phys., 1965,42,2617.(PAPER 6/1056