Author's personal copy

Transcription

1 Solid State Sciences 11 (2009) Contents lists available at ScienceDirect Solid State Sciences journal homepage: Synthesis and structural mechanisms of the 2201-type ferrites and polytypes: Fe 2 (Sr 2 x A x )FeO 6.5 d/2 (A ¼ Ba, La, Tl, Pb and Bi) Christophe Lepoittevin a,b, Sylvie Malo a, *, Gustaaf Van Tendeloo b, Maryvonne Hervieu a a Laboratoire CRISMAT-ENSICAEN. Bd du Maréchal Juin CAEN cedex. France b EMAT, University of Antwerp, Groenenborgerlaan 171, B-2020 Antwerp, Belgium article info abstract Article history: Received 28 October 2008 Received in revised form 8 December 2008 Accepted 9 December 2008 Available online 25 December 2008 Keywords: Doped strontium ferrites Non-stoichiometry Transmission electron microscopy 2201 Iron rich layered structure Transport properties The Fe 2 (Sr 2 x A x )FeO 6.5 d/2 systems have been investigated, by doping the iron rich 2201-type parent structure with Ba 2þ,La 3þ and 5d 10 post-transition cations. The syntheses have been carried out up to the limit of the 2201-type solid solutions, in order to test the role of the double iron layer Fe 2 O 2.5 d/2. The localisation of the charge carriers in these compounds is consistent with their strong antiferro-magnetism. The investigation was then carried out in the transition part of the diagram up to the formation of stable phases. The study of structural mechanisms was carried using high resolution electron microscopy (transmission and scanning transmission), electron diffraction and energy dispersive spectroscopy. Different non-stoichiometry mechanisms are observed, depending on the electronic structure and chemical properties of the doping elements. The specific behavior of the modulated double iron layer is discussed. Ó 2008 Elsevier Masson SAS. All rights reserved. 1. Introduction Transition metal oxides with perovskite-related structures have been intensively studied in recent years because of their unusual electric and magnetic properties. Among the atoms occupying the B site, iron is the source of a rich chemistry because of its aptitude to adopt different valence states and coordination numbers, able to generate complex structures. The Ae Fe O systems (Ae ¼ Sr and Ca) are exceptionally rich, as evidenced by the existence of the oxygen deficient perovskites and numerous families of perovskiterelated families, where iron atoms adopt the IV, V and VI coordinations. Among these interesting ferrites, Sr 4 Fe 6 O 13 [1] has been shown to be an oxygen deficient material, with a mixed-conducting behaviour and potential applications to high temperature electrochemical processes and the study of the related solid solution (Sr 4 y Ca y )(Fe 6 x Co x ) 13 d obtained by doping the two cation sites has revealed the stability of the materials and thermoelectric power [2]. However, as evidenced by recent studies [3 6], Sr 4 Fe 6 O 13 d phases exhibit modulated structures, commensurate or incommensurate, where the modulation vector is clearly correlated to the oxygen stoichiometry. Moreover, the analysis of the structure highlighted that its average structure can be directly * Corresponding author. address: sylvie.malo@ensicaen.fr (S. Malo). compared to the RP s phases Sr m þ 1 Fe m O 3m þ 1 [7 9] and derivatives of the high Tc superconducting cuprates [10]. All these oxides correspond to the intergrowth of single or multiple perovskite layers with single or double or even triple distorted rock salt (RS)- type layers. Numerous members of these RP-derivatives have been discovered in the pseudo-ternary metal oxides diagram Bi Sr Fe O, demonstrating the exceptional ability of the ferrites to form layered structures [11 15]. These layered compounds Bi 2 (Sr 2 x A x )Sr m 1 Fe m O 3m þ 3 þ w are denoted as 22(m 1)m, m indicating the number of perovskite layers. The rock salt-type block is built up from one double [BiO][BiO] layer sandwiched between two [SrO] layers, to form one triple (RS)- type block. The description of Sr 4 Fe 6 O 13 d as the intergrowth of one perovskite layer [SrFeO 3 y ](y z 0) with three distorted rock salt-type layers [Fe 2 SrO 3.5 d/2 ] shows that it belongs to the 2201-type members with a highly original character: the intermediate layers are double [FeO][FeO] layers [5] instead of [BiO][BiO] layers. This specificity, together with the presence of extra oxygen located in between the two [FeO] layers, leads to the developed formula Fe 2 (Sr 2 )FeO (3 y)þ(3.5 d/2). This description of the Sr 4 Fe 6 O 13 d compound as a 2201-type structure provided clues to head towards a large family of new ferrites, i.e. to increase the number of perovskite layers according to the general formula Fe 2 (A 2 )Sr m 1 Fe m O (3 y)m þ (3.5 d/2).atpresent,themembers m ¼ 2 Fe 2 (Bi 0.7 Sr 1.3 )SrFe 2 O 9.33 and m ¼ 4 Fe 2 (Sr 1.9 Tl 0.1 )Sr 3 Fe 4 O have been isolated [16,17]. The latter ferrites are denoted as Fe Bi and Fe Tl to highlight the structural characteristics of the /$ see front matter Ó 2008 Elsevier Masson SAS. All rights reserved. doi: /j.solidstatesciences

2 596 C. Lepoittevin et al. / Solid State Sciences 11 (2009) intermediate double layer of the RS block, with regard to the Bi-based compounds, denoted Bi-2212 for example [12,13]. In a more general way, any member belonging to the present ferrite family is denoted as Fe A -(n 1)2(m 1)m, the suffix A indicating the nature of the cation associatedwiththealkalineearth. The structural mechanisms of the RP-derivatives are highly complex [10]; they have been intensively investigated, especially in the systems of the Bi- and Tl-based cuprates, ferrites and manganites. The complex non-stoichiometry mechanisms which take place in these systems associated to the modulated character of the structures increase the difficulty to determine the limits of solid solutions in order to understand the properties. As a consequence, it is straightforward that a large amount of structural work, down to the nano-structural scale, would be necessary for understanding perfectly their mechanisms. The goal of the present paper is to present the structural behaviour of the Fe-2201-type phases, Fe 2 (Sr 2 x A x )FeO 6.5 d/2 when they are doped with La 3þ, Ba 2þ and 5d 10 post-transition cations, Tl 3þ,Pb 2þ and Bi 3þ. The syntheses have been first carried out up to the limit of the different solid solutions, in order to test the role of the double iron layer Fe 2 O 2.5 d/2. The investigation was then carried out beyond these limits, in the transition part of the diagrams, up to the formation of stable phases. 2. Experimental section The syntheses have been carried out by increasing x up to the limit of the solid solutions and the formation of stable phases. According to our first attempts, the nominal A content Fe 2 (Sr 2 x A x )FeO 6.5 d/2 was varied from 0 to 0.5, by x ¼ 0.01 steps, for A ¼ Bi, Tl, La, Pb, Ba. The Bi-based samples have been prepared from mixtures of the starting materials Bi 2 O 3,Fe 2 O 3 and SrO in a glove box. The different oxide powders, in the stoichiometric ratio, were ground, pressed into bars and sealed in a silica tube. The mixtures were heated at 1100 C for 48 h, with a heating rate of 2 C/min and slow cooled at the same rate. The Tl-based samples were prepared, in the same way, from mixtures of the starting materials Tl 2 O 3, Fe 2 O 3 and SrO. The Pb and Ba based samples from PbO or BaO, Fe 2 O 3 and SrO, in a glove box. The silica tubes were introduced in an alumina tube in order to avoid contamination in case they should explode. The first energy dispersive spectroscopy (EDS) analyses carried out on the thallium based samples showed that the thallium content of the crystallites was way below the nominal one. Due to its volatility, a large part of the thallium oxide was deposited on the tube walls. The powder X-ray diffraction (PXRD) analyses were carried out at RT with a Philips diffractometer working with the Cu Ka (l ¼ Å) in the range 5 2q 105. The samples were studied by different transmission electron microscopy (TEM) techniques. For this work, a small piece of sample was crushed in a mortar containing alcohol, and then a droplet was deposited on a copper grid covered with holey carbon film. Electron diffraction (ED) was carried out with a JEOL 200CX electron microscope, equipped with an energy dispersive spectroscopy (EDS) analyzer. The microscopes used for high resolution TEM (HRTEM) were a JEOL 4000EX operating at 400 kv and a TOPCON 002B microscope operating at 200 kv. The Z-contrast images were obtained on a JEOL 3000F microscope equipped with a scanning transmission electron microscopy (STEM) unit and a high-angle annular dark field (HAADF) detector. The magnetic measurements were performed by SQUID magnetometry (T < 400 K) and the susceptibility measurements on a Faraday balance (0.3 T). The resistivity measurements were carried out by the four probes method on a Physical Properties measurements System (PPMS). 3. Results The structure of the double iron layer [Fe 2 O 2.5 d/2 ] is one of the key factors of the stabilisation of the present ferrites and the A 3þ for Sr 2þ substitution one way to increase the oxygen stoichiometry. This is a potential test of the upper limit of oxygen [Fe 2 O 2.5 d/2 ]. La and Bi are good candidates because their substitution for Sr has been carried out successfully in numerous perovskite systems, leading to complete solid solutions, as in La 1 x Sr x MO 3 x/2 (with M ¼ Fe, Mn, Co/) [18 20] or Bi 1 x Sr x FeO 3 x/2 [21,22]. Pb 2þ and Ba 2þ, isovalent with Sr 2þ, were selected for the size effect. The first step was to determine the limits of the Fe 2 (Sr 2 x A x )FeO 6.5 d/2 solid solutions. Another interesting point is the possible stabilisation of two and two 2212-type structures within one system. This point is illustrated by the Bi Sr Fe O diagram given in Fig. 1, with the relative distribution of the different phases in the ternary system. - The bismuth rich Bi-2212 and Bi-2201 compounds are represented by blue stars; (Bi 2 x Sr x )Sr 2 FeO 6.5 x/2 [11] is stabilized in the domain 0 x 0.9 (blue dotted line) (for interpretation of the references to colour in this figure, the reader is referred to the web version of this article). - The iron rich Fe-2212 [16] and Fe-2201 compounds are represented by red stars. The hatched zone is the theoretical existence domain of the Fe A -(n 1)2(m 1)m phases, from m ¼ 1 (pink line 2201 s) to m ¼ N (the perovskite limit); it is common to all the A cations (taking into account the possible mixed valences of Fe (Fe 2þ,Fe 3þ and Fe 4þ )). The second step was to understand the non-stoichiometry mechanisms of the A Sr Fe O systems beyond the limits of the solid solutions, depending on the doping cation The parent structure Fe 2 (Sr 2 )FeO 6.5 d/2 : brief recall Fig. 1. Bi Sr Fe O diagram showing the Bi- and Fe-based 2201 and 2212-type phases. The hatched zone represents the theoretical existence domain of the 22(m 1)m-type ferrites. In the first member of the family (m ¼ 1) Fe 2 (Sr 2 )FeO (3 y)þ(3.5 d/2), y has been determined to be close to 0, in the accuracy limit of our techniques [2 5], leading to an oxygen content O 6.5 d/2.itssubcellis

3 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 2. Structural model of one orthorhombic variant of the Fe-2201 phase for p ¼ orthorhombic (Fmmm-type) with azbza p 2 (ap is the parameter of the perovskite unit cell) and c z 19 Å. The modulation vector,!! q ¼ pa * þ rc!*,hasacomponent(p)whichvarieswiththeoxygen stoichiometry as p ¼ (1 d)/2 and its second component is r ¼ 1 (under our previous processing conditions 0.4 p < 0.5). Three types of iron coordination, determined from single crystal data [4], arelocated inthe RS block: trigonal bipyramids (TBP), tetragonal pyramids (TP) and monocaped tetrahedra (MT). These three polyhedra are arranged to form three different basic structural units, represented in Fig. 2: the ([TBP[][TBPY]) double block of trigonal bipyramids, the ([TP][TP]) double block of tetragonal pyramids, and the ([MT][TP][MT]) triple mixed block of two monocaped tetrahedra sandwiching one tetragonal pyramid. The oxygen excess of the [Fe 2 SrO 3.5 d/2 ] RS block is located in between the two trigonal bipyramids. The two limit commensurate structures of the [0.4 p < 0.5] range are the oxidized one, i.e. p ¼ 0.5 never stabilized at the present, which can be described as a regular sequence of ([TBP[][TBPY])([TP][TP]) along! a and the reduced one with p ¼ 0.4, which can be described as a regular sequence of ([TBP[][TBPY])([MT][TP][MT]) along! a.thestructuralmodelgivenin Fig. 2 is one of the variants. The incommensurate intermediate structures correspond to random sequences of these two limit structures. Whatever the m member, the general formula Fe 2 (A 2 )Sr m 1 Fe m O (3 y)m þ (3.5 d/2) accounts for the thickness of the perovskite slice (m) and the amplitude of the modulation vector along!* a. Considering the oxygen content (3 y)m þ (3.5 d/2), the knowledge of the modulation vector allows to determine the excess of oxygen (0.5 d/2) located in between the two [FeO] layers whereas (3 y)m allows to determine the oxygen deficiency of the SrFeO 3 y perovskite layers. Fig. 3. [010] ED pattern of the limit compound (Fe 2 )(Sr 1.65 Pb 0.35 )FeO The Fe 2 (Sr 2 x A x )FeO 6.5 d/2 solid solutions The limits of the solid solutions can be hardly determined from the XRPD patterns, especially in the case of A ¼ Bi, Tl and Pb. As further shown, the difficulty lies in the nature of the secondary phases which present very close relationships with the 2201 s. For the different samples of each family, the combined EDS/ED analyses have been carried out on several tens of grains in order to check the cation distribution homogeneity and to detect the presence of impurity. The results are summarized in Table 1. For the trivalent cations, (La 3þ,Bi 3þ and Tl 3þ ) the solid solutions are considerably limited with upper x values of x La 3þ ¼ 0:025 and for x 3þ Bi ¼ 0:03. The EDS analyses confirm that the different A 3þ cations have really substituted the Sr atoms. Concomitantly, one observes that the modulation vector remains roughly constant, p z 0.46 and r ¼ 1, unchanged with regard to that of the undoped ferrite. The amplitude of the modulation vector, which is very sensitive to the oxygen distribution within the rock salt block, shows that the structure suffers only very minor (or none) changes. Table 1 Limits of the solid solutions Fe A (Fe 2 (Sr 2 x A x )FeO 6.5 d/2 ). A cation R VIII (Å) Maximum x value of the Fe A La 3þ p ¼ 0.46 Bi 3þ p ¼ 0.46 Tl 3þ 0.98 e a p ¼ 0.46 Pb 2þ p ¼ 0.45 Ba 2þ p ¼ 0.45 Modulation vector q ¼ pa!* þ c! * a For Tl-based compounds, no x value associated to a phase can be given, the p value is given for the Tl doped 2201-type grains. Table 2 Refined subcell parameters of the limits of the solid solutions for A ¼ Pb and Ba. Formulation Cell parameters (Å) (subcell) Volume (Å 3 ) (Fe 2 )(Sr 2 )FeO 6.46 a ¼ 5.552(1) b ¼ 5.574(2) c ¼ (9) (Fe 2 )(Sr 1.65 Pb 0.35 )FeO 6.46 a ¼ (2) b ¼ (2) c ¼ (8) (Fe 2 )(Sr 1.9 Ba 0.1 )FeO 6.46 a ¼ (5) b ¼ (3) Å c ¼ (2) Å

4 598 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 4. Inverse molar susceptibility vs. T of the parent structure Fe 2 (Sr 2 )FeO These results implied by increasing the oxygen content by a chemical way are consistent with the results observed using the oxygen pressure technique [2]. For the Tl-based compounds, no x value is given in Table 1 because rather large fluctuations in the Tl content of the grains is observed; the given p value corresponds to the maximum one observed for the 2201 grains containing significant amounts of Tl. On the opposite, the homogeneity range of the Pb 2þ and Ba 2þ solutions, which does not involve a variation of the oxygen content, are larger: (0 x 0.35) for Pb 2þ and (0 x 0.1) for Ba 2þ. The coupled EDS/ED analyses confirm the homogeneity of the samples. One typical [010] ED pattern of the lead rich limit composition (Fe 2 )(Sr 1.65 Pb 0.35 )FeO 6.46 is given in Fig. 3; the diffraction spots corresponding to the Fe-2201 subcell, as well the satellites are very sharp and well-defined. The components of the modulation vector are very close to that of the undoped compound, p z 0.45 and r ¼ 1. Using four indices hklm, the reflection conditions are also similar (hklm: k þ l þ m ¼ 2n, h0lm: h ¼ 2n, hk0m: h ¼ 2n) leading to the similar superspace group Amaa(a01) or Xmaa(a00) with X corresponding to the centering vector 0 1/2 1/2 1/2 [4]. The subcell parameters, refined from the XRPD data, are given in Table 2; the increase of the subcell volume Å 3 with regard to that of the undoped compound ( Å 3 ) is consistent with the larger radius of Pb 2þ (divalent cation 1.29 Å/1.26 Å for a cation in eightfold coordination) and the constant modulation vector. In order to investigate the role of the cation size of the doping element and that of the 6s 2 lone pair of Pb 2þ, we worked with Ba 2þ, an alkaline earth of the sixth period (1.42 Å). The homogeneity range is significantly smaller; the combined EDS/ED analyses show that the samples are homogeneous up to x ¼ 0.1. The components of the modulation vector remain very close to that of the undoped and Pb-doped compounds, p z 0.45 and r ¼ 1, in agreement with a similar valence for Sr, Ba and Pb. The subcell parameters, refined from the XRPD data, are given in Table 2, involving the increase of the subcell volume Å Magnetism and transport properties The evolution of the inverse molar susceptibility vs. temperature of the parent structure Fe 2 (Sr 2 )FeO 6.46 is given in Fig. 4. The Fig. 5. (a) Resistivity vs. T and 1/T of the Fe 2 (Sr 2 )FeO 6.46 and (b) corresponding activation energy.

5 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Table 3 Main phases observed beyond the limit of the Fe A (Fe 2 (Sr 2 x A x )FeO 6.5 d/2 ) solid solutions. A 0 cation x Main phases Modulation vector of the Fe A La 0.03 Fe La þ La 0.3 Sr 0.7 FeO 3 d p ¼ 0.46 Bi 0.04 Intergrowth mechanism Tl 0.04 Intergrowth mechanism Pb 0.4 Fe Pb þ terrace structure [20] p ¼ 0.45 Ba 0.1 Fe Ba þ BaSrFe 4 O 8 p ¼ 0.45 measurements during heating and cooling evidenced the existence of a reversible magnetic transition at about 730 K (the maximum temperature is limited up to 800 K in our experimental conditions, so that the linear paramagnetic domain is not reached). This value of the Neel temperature is of the same order than the ones observed for the ordered perovskites of the diagram, Bi 1/3 Sr 2/3- FeO 2.67 and Bi 1/2 Sr 1/2 FeO 2.75 [23] and the Fe Bi phase [16], all characterized by a trivalent state of iron. The high Neel temperature, T N z 730 K, is consistent with the super-exchange energy for the Fe 3þ O Fe 3þ magnetic interactions [24]. The resistivity (given as a function of T and 1/T in Fig. 5a) is of the order of 10 4 U cm at 300 K; this localisation of the charge carriers is consistent with a strong anti-ferromagnetism. The r(t 1 ) curve allows to extract a activation energy of 300 mev for this compound (Fig. 5b). The measurements carried out at 300 K on the Pb-( U cm), Ba-( U cm) and Bi-( U cm) doped compounds do not show any significant change in their magnetic and transport properties with regard to the undoped phase. p ¼ 0.45; the conditions of reflection remain similar (superspace group Amaa(a01)). For higher x values, elongated spots, diffuse streaks and additional weak nodes along!* c, whose intensity increased with x, are superposed on the Fe-2201 sub-system of intense spots. The corresponding EDS analyses evidence a small but significant increase of the deviation with regard to the ideal ratio (Bi þ Sr)/Fe ¼ 2/3 of a Fe Bi member. (Bi þ Sr)/Fe ¼ 0.725: the results are obtained for the actual composition of the crystallites [0.15/2.02/3]; this ratio (Bi þ Sr)/ Fe ¼ remains inferior to the one expected for the Fe-2212 members (0.75). Two systems of reflections are observed in the [010] ED pattern (Fig. 6a). The system of intense nodes is similar to that of Fe Bi solid solution but it differs by the symmetry and the satellites associated with the modulated structure. The reconstruction of the reciprocal space allowed to conclude that the F-type symmetry of the subcell is violated, as shown especially by the intensity of the 110 reflection in Fig. 6b; the reflection conditions involve Cccm or Ccc2 as possible space groups. The satellites are intense and the component (p) of the modulation vector along! a * is p ¼ 0.4 (and r ¼ 1). The satellites are indexed using four hklm indices in Fig. 6. The peculiar value of p ¼ 0.4 ¼ 2/5 involves that the modulation is commensurate. The second system is associated to a reinforcement (small arrow) in the diffuse scattering along!* c, but without sharp extra maxima, except the one indicated by the 3.3. Beyond the upper limits of the 2201 phase: stabilisation of Bi- and Tl-polytypes As mentioned above, beyond the limit of the solid solutions, the evolution of the La and Ba systems have been easily observed from the XRPD patterns, through the formation of one secondary phase. The perovskite La 0.3 Sr 0.7 FeO 3 d is observed for x La and the hexagonal phase BaSrFe 4 O 8 [25] for x Ba > 0.1. For the Bi, and Tl-systems, the XRPD patterns exhibit broadened and/or very small extra peaks, which cannot be simply interpreted. These samples have been systematically characterized by electron diffraction in a first step and by high resolution electron microscopy in a second step. The results are summarized in Table 3. The modification observed in the XRPD patterns mainly result from order disorder phenomena, in the form of the stabilisation of polytypes for A ¼ Bi, Tl; they are necessarily associated to a small deviation of the [A/Sr/Fe] cation ratio with regard to that of the 2201-type compound Fe 2 (Sr 2 x A x )FeO 6.5 d/2, i.e. [x/(2 x)/3] or (A þ Sr)/ Fe ¼ 2/3. The results of the TEM study are therefore described as a function of this deviation Electron diffraction The Bi-based ferrites 2/3 < (Bi þ Sr)/Fe < 0.7: For a nominal x content slightly superior to 0.03, the system of intense spots in the ED patterns is that of the Fe-2201 sub-system. However, weak diffuse streaks are observed along c!*, which result from disorder phenomena associated to a small deviation of the cationic ratio: these additional phenomena show that the upper limit of the solid solution has been crossed. As an example, for the nominal x ¼ 0.04 (Bi/Sr/Fe ¼ [0.04/1.96/3]) sample, the actual cationic ratio of the Fe-2201 type crystallites is [0.04/2.02/3] (calculated for 3 iron atoms per unit). The component of the modulation vector along a!* of the Fe-2201 subcell is Fig. 6. Typical (a) [010] and (b) [001] ED patterns recorded for a cation ratio (Bi þ Sr)/ Fe ¼ (the hklm indices in italic are those of the modulated cell).

6 600 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 7. [010] ED patterns recorded (a) for a cation ratio (Bi þ Sr)/Fe ¼ 0.75 and (b) for a cation ratio (Tl þ Sr)/Fe ¼ The red dots indicate the reflections of the Fe-2201 sub-system (for interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article). larger arrow, which corresponds to an inter-planar spacing d z 3.8 Å (za p ). They are the signature of order disorder phenomena. (Bi þ Sr)/Fe ¼ The case of the crystallites Fe 2 (Bi 0.15 Sr 2.1 )Sr- FeO 6.46 [0.15/2.1/3], which exhibit a (Bi þ Sr)/Fe ¼ 0.75 ratio similar to that of the Fe Bi member, is of interest because it allows to highlight another factor. The crystallites [0.15/2.1/3] are characterized by the ratio (Bi/(Bi þ Sr) ¼ 0.067), which is considerably smaller than 0.23, the experimental value of the stable Fe-2212 member Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O 9.33 [16]. The [010] ED pattern is given in Fig. 7a, where red dots are superposed on the reflections of the Fe-2201 type sub-system. One observes that the 004 and 006 nodes of the Fe-2201 system are blurred in elongated groups of reflections. The large and smaller arrows in the left part of the pattern indicate two of the nodes, which appear in the diffuse streaks along c!*. The first elongated node (larger arrows) is associated to an inter-planar distance d z 28 Å and the second one (smaller arrows) to d z 15 Å, likely correlated to the ordering phenomena (further discussed) The Tl-based ferrites In the Tl-based ferrites, a drastic decrease of the actual Tl content is observed with regard to the nominal composition. For example, for a nominal composition Tl/Sr/Fe ¼ [0.5/1.5/3], the average actual composition is [0.015/2/3] calculated for 3 Fe atoms per unit, however this average composition has no real sense since the thallium content varies from 0 to 0.03 and consequently the variations of Sr reach up to 20%. In fact, a multi-nanophased sample is obtained, the majority of the crystallites belonging to the Fe Tl -22(m 1)m system. A typical example of [010] ED pattern is given in Fig. 7b and can be compared to that of the Bi-based compounds (Fig. 7a). Referring to the system of red dots, it appears that the Fe-2201 is no longer the major phase but the most intense nodes correspond to d z a p z 3.8 Å (yellow arrow), suggesting that high m members with thicker perovskite slices, are formed. For example, in Fig. 7b, the main system corresponds to d z 37.5 Å (calculated as the 002 of the Fe-2278 member) The Pb-based ferrites The evolution of the system is totally different from the Bi- and Tl-based ones. Beyond the limit (x 0.4), the increasing presence of the terrace phase Pb 4 Sr 13 Fe 24 O 53 [26] besides the Fe Pb phase Fe 2 (Sr 1.65 Pb 0.35 )FeO This terrace structure can be described from the intergrowth Fe Pb through the formation of periodic crystallographic shear planes. This study shows that the Fe 2 (Sr 2 x A x )FeO 6.5 d/2 phases behave according to two types of mechanisms. Biphasic samples are formed by La and Ba doping with the coexistence of the Fe A limit compounds and a stable phase of the diagram. Beyond the limit of the solid solution, which is the higher x value (0.35), the Pb doping is accommodated through the formation of a CS phase. The phenomena detected on the ED patterns of the Biand Tl-based ferrites are the signature of order disorder mechanisms and show that a non-stoichiometry mechanism takes place in the matrix as the amount of doping element increases. HRTEM will be used to unravel the nature of the order disorder mechanism High resolution transmission electron microscopy The EDS analyses coupled to the ED study showed that, beyond the limits of the solid solutions of the Bi- and Tl-systems, the order Table 4 Major phases stabilised by increasing x in Fe 2 (Sr 2 x A x )FeO 6.5 d/2. (Bi þ Sr)/Fe [Bi/Sr/Fe] Variants/polytypes a b c b ( ) p [0.04/1.96/3] Classical incommensurate m ¼ 1 single phase [5] Subcell a p p 2 ffiffiffi a p Å [0.04/2.02/3] Classical m ¼ 1 major phase [5] and m ¼ 2,3, 4 as defective members Subcell a p 2 a p p 2 ffiffiffi 18.9 Å [0.15/2.02/3] New commensurate m [ 1 (p ¼ 0.4) and m ¼ 3 and 4 as defective members 5a p 2 a p p 2 ffiffiffi z23.8 Å z127.5 (0.4) 0.75 and Bi/Bi þ Sr ¼ Two major polytypes [2201] 1/2 [2234] 1/2 subcell a p 2 a p p 2 ffiffiffi z23.0 Å 0.43 [0.15/2.1/3] [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2 a p 2 a p 2 z89 Å and Bi/Bi þ Sr ¼ 0.10 [0.15/2.1/3] Transition state commensurate m ¼ 1 (p ¼ 0.4) þ Fe Bi [16] 0.75 and Bi/Bi þ Sr > 0.14 Commensurate Fe Bi [16] 3a p 2 a p Å (0.33) (Tl þ Sr)/Fe [Tl/Sr/Fe] polytype a b c b( ) p [0.03/1.97/3] [2201] 1/2 [2234] 1/2 5a p 2 a p Å z107 (0.4)

7 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 8. (Bi þ Sr)/Fe ¼ 0.725: (a) typical [010] HREM image and (b) FFT of the m ¼ 1 matrix. disorder phenomena are clearly associated to deviations of the cationic ratio and take place, in more or less ordered sequences along c!, in the Fe-2201 matrix Bi-based variants and polytypes as a function of x In the Bi-based ferrites, beyond but close to the limit of the solid solution (0.667 < [(Bi þ Sr)/Fe] < 0.7), the progressive formation of defective members, essentially the lower m values (m ¼ 2, 3 and 4) is observed in the Fe Bi matrix. The density of defective members m ¼ 4 increases with x, whereas the p value tends to decrease. Increasing x, new microphases are stabilized. The results are summarized in Table (Bi þ Sr)/Fe ¼ 0.725: a new commensurate variant of the Fe Bi For a cation ratio Bi/Sr/Fe ¼ [0.15/2.02/3] (ED in Fig. 6), the typical [010] HRTEM image is presented in Fig. 8. The major phase is afe Bi type whereas the m ¼ 2, 3 and 4 defective members are locally inserted in this matrix. The defective members are easily identified through the number of [SrO] layers, i.e. (m 1), belonging to the perovskite blocks, which appear as the rows of brighter dots for the selected focus value. The RS blocks exhibit the typical modulated contrast; note that only one exception is observed in this image, where the m ¼ 2 member is rotated over 90 (labeled m t ). In the present area, the very regular contrast at the level of the RS blocks is associated to a monoclinic variant of the p ¼ 0.4 modulated structure; the FFT is given in Fig. 8b, showing that the modulation vector (in yellow)! q ¼ pa!* þ rc!* exhibits components p ¼ 0.4 but r is locally s1 (for interpretation of the references to colour in this figure, the reader is referred to the web version of this article). Using the imaging code determined from the crystal data and the HRTEM study for the Fe-2201 compounds [5], a model can be proposed for the monoclinic variant stabilized for this Bicontent. The p ¼ 0.4 value of the component along!* a involves a commensurate modulation leading to a m ¼ 5a p 2, bm ¼ a p 2 (the subscript m referring to the monoclinic supercell) as described above. The contrast of the enlarged [010] HRTEM images at the level of the rock salt block (Fig. 9a) can be described by the presence of one bright double stick separated by three greyer ones. This contrast has been associated to a regular sequence of two structural units built up from five iron polyhedra ([TBP[][TBPY])([MT][TP][MT]) along! a [5], which is doubled to respect the C centering, leading to the 5a p 2 periodicity. In the upper and lower adjacent slices, this sequence is periodically translated by! a sub. It results in a monoclinic supercell a m ¼ 5a p 2, bm ¼ a p 2, cm z 23.8 Å and b z The corresponding model is drawn in Fig. 9b. A sequence built up from five iron polyhedra, likely ([TBP[][TBPY])([MT][TP][MT]), has been previously observed in the ferrites Fe Tl [17] with a thick perovskite slice (m ¼ 4) and the Fe-2201 phase, obtained under reducing atmosphere [4,5]. This point is of interest because the fact that it could be realised either by reducing the parent structure (in the Fe-2201) or by doping it by Tl 3þ (2234) or Bi 3þ (this work) could appear in contradiction. The differences stay in the roles played by the Fe-2201 matrix and the inserted defective members. In the reduced Fe 2 Sr 2 FeO type sample, the loss of oxygen is likely accommodated by the diminution of the extra oxygen in the double iron layer of the RS block rather than in the single (SrFeO 3 ) layer, as a way to prevent strains effects due to the presence of oxygen vacancies. In the A 3þ doped ferrite with m 2, the excess of oxygen supplied by the A 3þ for Sr 2þ substitution would likely be located in the thicker oxygen deficient (SrFeO 3 y ) perovskite layers, in order to decrease the strain effects due to the oxygen/vacancy ordering [27 33]. This observation is consistent with the decrease of the p ¼ 0.33 value of the Fe Bi -2212, Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O According to this hypothesis, the oxygen excess supplied by the Bi in the compounds Fig. 9. (a) Enlarged [010] HREM image of the commensurate variant (p ¼ 0.4) of the Fe Bi matrix and (b) idealized structural model.

8 602 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 10. HREM image showing the formation of [010] and [100] oriented domains in a crystallite characterized by the intergrowth of Fe Bi and Fe Bi members. Fig. 12. [010] Typical HREM image of a Tl doped Fe-2201 ferrite showing the formation of high m members. for the cationic ratios close to (Bi þ Sr)/Fe ¼ would be located in the defective higher m members (Bi þ Sr)/Fe ¼ 0.75 and Bi/(Bi þ Sr) ¼ 0.067: stabilisation of polytypes The crystallites of [0.15/2.1/3] composition are characterized by a cationic ratio of (Bi þ Sr)/Fe ¼ 0.75 similar that of the Fe-2212 s (the corresponding composition belonging to the m ¼ 2 (green) line on the diagram) but its Bi content, Bi/(Bi þ Sr) ¼ 0.067, is considerably smaller with regard to the actual value 0.23 [16]. As described above, extra nodes are observed in the ED patterns. The HRTEM images reveal that the number of defective members has increased, but, more interesting, it clearly appears that locally ordering phenomena occur over several nanometers. Two regular sequences of m ¼ 4 and m ¼ 1 members are observed. Fig. 11. (a) [010] HREM image of a [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2 polytype and (b)fft of the [010] HREM image The Bi-based polytype [2201] 1/2 [2234] 1/2 One typical HREM image is given in Fig. 10. At the level of the triple RS block, the alternation of bright and grey areas is associated to the modulation along! a, whereas the straight rows of bright dots are correlated to the [SrO] layers of the perovskite blocks. In the present case, the numbers of [SrO] layers in the perovskite blocks are 0 (belonging to the Fe-2201) and 3 (belonging to the Fe-2234). The two members alternate rather regularly, forming locally short sequences. Along! c, the local periodicity is! c pol ¼½1=2! c 2201 þ 1=2! c 2234Š, close to 23 Å; the polytype is denoted [2201] 1/2 [2234] 1/2 (the exponents refer to the c parameter of the corresponding structural subunit). The FFT of the corresponding area shows the p component remains close to One of the consequences of the insertion of high m members is a decrease of the strain effects generated by the

9 C. Lepoittevin et al. / Solid State Sciences 11 (2009) Fig. 13. [010] HREM images of the polytype Tl-[2201] 1/2 [2223] 1/2 for two focus values, with high electron density zones appearing as bright (a) and dark (b) dots. (c) Structural model of the polytype Tl-[2201] 1/2 [2223] 1/2. modulated structure of the RS blocks. This relaxation leads to the growth of [010] oriented (upper part of Fig. 10) and [100] oriented (bottom part of Fig. 10) domains along c!. The modulated contrast at the level of the RS block, characteristic of the [010] areas has disappeared when viewing the crystallites along [100]. The perovskite layers exhibit the same contrast along both orientations, as a proof of the absence of long (LRO) or short range (SRO) ordering commonly associated to the [SrFe 3þ O 3 y ] perovskites layers. The twin boundaries (TB) are always coherent and located at the level of high member rich areas. The formation of these twin domains contributes to the relative low intensity of the satellites in certain ED patterns The Bi-based polytype [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2 For the same composition ((Bi þ Sr)/Fe ¼ 0.75 and Bi/ (Bi þ Sr) ¼ 0.067), long range ordering, built up from the same m ¼ 4 and m ¼ 1 members, is stabilized (Fig.11a). Along! c, regular sequences of Fe-2201 and Fe-2234 members are observed with the periodicity ð! c pol ¼½3=2! c 2201 þ 1=2! c 2234 þ 1=2! c 2201 þ 1=2! c 2234Š Þ, so that the polytype is denoted [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2. The calculated parameter along! c is z89 Å. The additional nodes detected in the ED patterns (Fig. 7a) can be explained in this way: the first elongated node, d z 28 Å, is close to d 001 of the Fe-2212 (27.1 Å) and d 003 of the Fe-2234 (29.6 Å) whereas the second one, d z 15 Å, can be associated to d 006 (14.8 Å) of the polytype. The FFT of the corresponding area (Fig. 11b) shows that the amplitude of the modulation vector along a!* remains close to 0.43, similar to that of the overall crystallite. Note that the number of RS and perovskite layers, which have been observed for this nominal composition [0.15/2.1/3] (above section and Fig. 7), lead to an average value m ¼ 2. This is actually three times this unit [(2201) 2 þ (2234)¼(6636)] and could therefore be written as Fe 2 (Bi 0.2 Sr 1.8 )SrFe 2 O w. The regular m ¼ 2 Fe Bi member of the system, Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O 9.33 [16], is in fact stabilized for a higher Bi content. One of the origins of these structural differences is likely correlated to the oxygen content and the oxidation state of iron. The average oxidation state of iron is trivalent in Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O 9.33 [16] and the relationship between the oxygen content 9.33 and the amplitude of the modulation vector along a! (p ¼ 0.33) shows that the amount of oxygen vacancies, i.e. y, is really tiny in the perovskite layers (SrFeO 3 y ). For the m ¼ 4 member Fe 2 (Sr 1.9 Tl 0.1 )Sr 3 Fe 4 O [17], the p ¼ 0.4 value and the mixed valence of iron (zfe þ3.2 ) have allowed us to show that the y value is close to 0.2, with a random distribution of oxygen atoms and vacancies. Assuming that the potential oxygen vacancies are likely located in the quadruple perovskite block and taking into account p ¼ 0.43, a simple calculation would lead for the polytype [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2 to a y value close to 0.25 in the SrFe 3þ O 3 y perovskite layers of

10 604 C. Lepoittevin et al. / Solid State Sciences 11 (2009) the m ¼ 4 slices, y decreasing down to 0 in the hypothesis of a mixed valence of iron, with a maximum oxidation state of Fe þ The Tl-based polytype [2201] 1/2 [2223] 1/2 A similar mechanism of SRO-LRO is observed in the Tl doped ferrites, with the appearance of defective members in a Fe-2201 matrix, up to the stabilisation of polytypes. However, they present two major differences. The first one is the possible insertion of very high m members in the crystallites. In the particular example of Fig. 12, the local formation of Fe-2256 and Fe members is observed. These thick perovskite slices are systematically separated by one single Fe-2201 slice, leading to the formation of [2201] 1/2 [22(m 1)m] 1/2 members. No real long range ordering is observed, but locally [2201] 1/2 [2256] 1/2 and [2201] 1/2 [ ] 1/2 nanophases are formed. Such an observation raises the problem of the oxygen content in the (SrFe 3þ O 3 y ) perovskite layers. The very regular contrast shows that there is no signature of strain effect or ordering at the level of these layers, as mentioned above for the Bi-doped compounds and the m ¼ 4Fe Tl phase [17], which exhibits a mixed valence Fe III/IV. One can assume that a similar mechanism takes place in these compounds, thanks to a thallium oxide pressure in the silica tube during the synthesis, which implies a decrease of the Tl content with regard to the nominal composition in certain crystallites but which favors in return an increase of the oxygen content in the perovskite layers. The second point deals with the nature of the members stabilized in the polytypes. In Fig. 13 an example is imaged under two typical focus values. In Fig. 13a, the high electron density zones are highlighted and the [Sr 1 x Tl x ] positions appear as bright dots (see yellow arrows); in Fig. 13b, the dark dots are associated with the high electron density zones, i.e. the strontium and iron positions in the perovskite layers (for interpretation of the references to colour in this figure, the reader is referred to the web version of this article). These images allow an easy interpretation of the intergrowth: the sequence along! c is one RS block, one perovskite layer, one RS block and three perovskite layers, i.e. a polytype [Fe 2 (Tl e Sr 1 e )FeO 6.4 ] 1/2 Fe 2 (Tl e Sr 1 e )Sr 2 Fe 3 O 12.4 ] 1/2 denoted Tl- [2201] 1/2 [2223] 1/2. The FFT, inserted in the upper left corner of the image, shows that the component p along!* a is 0.4, i.e. the modulation is commensurate. In the HRTEM image, the contrasts at the level of the modulated rock salt block (Fig. 13a and b) confirm the alternation of groups of two and three different structural units, as previously described, and suggest the sequence ([TBP[][TBPY])([MT][TP][MT]) of the iron polyhedra. The Fe-2201 slice is similar to that observed for the Bi-doped 2201 (Fig. 9), it is intergrown with a Fe-2223 member and the arrangement of the iron polyhedra suffers a translation by! a sub. The polytype Tl-[2201] 1/2 [2223] 1/2 exhibits a monoclinic commensurate supercell with a ¼ 5a sub ¼ 5a p 2, b ¼ b sub ¼ a p 2, c z 28.2 Å and b z 107. An idealized model is proposed in Fig. 13c. Referring to the investigation of the thallium rich-part of the Tl Sr Fe O diagram which has demonstrated that double rock salt layers [SrO] and [TlO] can be intergrown with SrFeO 3 y perovskites layers [34], the stabilization of these polytypes raises the problem of the localization of Tl in the matrix, despite the very small amount of Tl. Our previous studies [17] have shown that high-angle annular dark field (HAADF) is the appropriate technique because the intensity related to a column of atoms is proportional to Z n (with 1 < n < 2) and Z the (average) atomic number of the projected column. The results are given in Fig. 14 with an image of the polytype Tl-[2201] 1/2 [2223] 1/2 (Fig. 14a) and that of a reference sample (Fig. 14b). As a reference we have the Fe Bi -2212, Fe 2 (Bi 0.69 Sr 1.31 )Sr- Fe 2 O 9.33 [16], because its commensurately modulated structure has been solved from single crystal X-ray diffraction data and the amount of Bi is sufficient to be detected in a regular structure. In a HAADF image, the higher Z the brighter the dot on the image is. In Fe Bi -2212, the Bi atoms are located in the [Bi 0.35 Sr 0.65 O] layers at the boundary between the RS and P blocks, whereas the third layer, [SrO], is fully occupied by Sr. In Fig. 14b, the [010] HAADF STEM image of Fe Bi clearly shows two rows of brighter dots associated to the projection of z1/3bi (Z ¼ 83) and z2/3sr (Z ¼ 38) columns, sandwiching a single row of weaker dots associated to the projection of 1 Sr column. In the Tl-[2201] 1/2 [2223] 1/2 polytype the amount of thallium is very small so that it cannot significantly affect the contrast in the hypothesis of a random distribution in the equivalent layers. For a ratio Tl/(Tl þ Sr) ¼ 0.02, the composition of the layer would be [Tl Sr O]. Z-contrast imaging however is very efficient in detecting any formation of defective [TlO] layers or Tl clusters. The [010] HAADF STEM image of Tl-[2201] 1/2 [2223] 1/2 (Fig. 14a) shows one group of four bright rows in the triple perovskite block and two rows in the single perovskite block. All the dots of these different rows exhibit a homogeneous brightness (note that the waviness of the layers is not a real effect, but induced by the recording). This Fig. 14. [010] HAADF STEM image of (a) polytype [2201] 1/2 [2223] 1/2 and (b) Fe Bi

11 C. Lepoittevin et al. / Solid State Sciences 11 (2009) observation allows us to discard the hypothesis of thallium defects to explain the stabilization of these polytypes and is in agreement with the results obtained for the Fe Tl phase [17] Transition state: the polytype Bi-[2201] n [2234] m toward the Fe Bi As mentioned before, for the ideal cationic ratio Bi/(Bi þ Sr) ¼ 0.067, the extra nodes observed in the ED patterns can be partly explained by the easy formation of the polytype Bi- [2201] 3/2 [2234] 1/2 [2201] 1/2 [2234] 1/2. The average composition of the latter ((2201) 2þ(2234) ¼ 6636) exhibits the ideal ratio expected for the 2212 s phases ((Bi þ Sr)/Fe ¼ 0.75); in spite of that, the formation of Fe Bi defective members has been rarely observed. We have increased the Bi content up to Bi/(Bi þ Sr) ¼ This small increase is sufficient to suppress the high m ¼ 3andespe- cially m ¼ 4 members and to form only the m ¼ 2 as defective member. Fig. 15 illustrates the typical [010] ED patterns and HRTEM images observed for the materials synthesized in this part of the diagram. The ED patterns exhibit almost sharp nodes, which result from the coexistence of two phases, and diffuse scattering phenomena, which can be interpreted by substitutional partial ordering and the associated complex atomic displacements. In Fig. 15a, red and blue dots have been superposed to the experimental pattern to highlight the two phases in presence (for interpretation of the references to colour in this figure, the reader is referred to the web version of this article). The blue system is that of the Fe Bi -2201, described above (Fig. 9), with a commensurate modulated structure p ¼ 0.4 leading to a monoclinic supercell with a m ¼ 5a p 2, bm ¼ a p 2, c m z 23.8 Å and b z The red system is that of the Fe Bi (Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O 9.33 ) [16], with a commensurate modulated structure p ¼ 1/3 leading to an orthorhombic supercell (Table 1). The diffuse scattering is associated to a transition state, which should be an intermediate between short range ordering and long range ordering; it is characterized by diffuse intensity along well-defined loci [35]. The loci are reinforced at positions in reciprocal space, corresponding to reflections of one of the long range ordered supercells. In the present materials, two mechanisms are engaged (Fig. 15b). The first one, which is one-dimensional, is related to the intergrowth of the Fe Bi and Fe Bi along! c ; it generates diffuse intensity along [001] *. One of the reinforcements is observed for a periodicity close to 40 Å, which could correspond to the local formation of [2201][2212] sequences. The second one is two-dimensional and related to the evolution of the modulation vector. The shape is a circular arc, centered on the 00lm satellites of the Fe Bi phase (0031 and in Fig. 15b) up to the next satellites of the Fe Bi phase (0031 and 0051 for one arc and and for the symmetric one); a reinforcement is observed for p z In Fig. 16, the [010] HAADF STEM image of a [2201] 1/2 [2212] 1/2 polytype is shown; the FFT (in insert) confirms the periodicity close to 40 Å and the p value (0.37) of the component of the modulation vector along a!*. The bright and less bright dots are associated to the (Sr,Bi) and Sr positions, running in rows perpendicular to c! ; the Fe-[2201] subunit is characterized by two rows and the Fe-[2212] subunit by three rows of bright dots. The important information displayed by this image deals with the distribution of the Bi atoms in the different layers of the polytype. The brightness of the dots is not uniform but that the brighter dots are only observed in the two outer rows of the Fe-[2212] subunit. It means that by increasing the Bi content (for intermediate Bi/ (Bi þ Sr) ratio), the Bi and Sr atoms are not randomly distributed in the different layers surrounding the two RS blocks of the [2201] 1/ 2 [2212] 1/2 polytype but are located in the two [(Sr,Bi)O] layers sandwiching the RS block of the Fe-[2212] subunit. For Bi/(Bi þ Sr) ratio ¼ 0.1, the detailed formulation would be close to [Fe 2 (Sr 2 )FeO 6.37 ][Fe 2 (Bi 0.5 Sr 1.5 )SrFe 2 O 9.37 ]. This observation outlines the important role of Bi in the stabilization of the Fe Bi -[2212] phase. For a higher Bi/(Bi þ Sr) ratio (>0.14), the amount of Fe Bi phase increases up to the formation of a single phase for the composition Fe 2 (Bi 0.69 Sr 1.31 )SrFe 2 O 9.33 [16]. 4. Discussion In this work, we have investigated the Fe 2 (Sr 2 x A x )FeO (3 y) þ (3.5 d/2) systems by increasing x and observed different non-stoichiometry mechanisms in the phases, depending on the nature of the A cations. The easy formation of mixed La/Sr and Bi/Sr perovskites relative compounds over large domains [18,21 23,36] attests the great ability of these cations to occupy the Sr sites. In that way, the very limited homogeneity range of the solid solutions Fe 2 (Sr 2 x A x 3þ )FeO 6.5 d/2 (A 3þ ¼ La 3þ,Bi 3þ )(Table 1), x 0.03, and the invariance of the component p ( ) along a!*, suggest that the important parameter lies in the nature of the double iron block. In our first approaches of the structure [3,5], it was shown that one of the ideal limits of the family is the formulation Fe 2 (A 2 )FeO 6.5, which is associated to the component p z 0.5 and Fig. 15. Formation of the Fe Bi through a small increasing of the Bi/Bi þ Sr content: (a) [010] ED pattern, (b) ED pattern enlargement around the three first 00l reflections and (c) HREM image.

12 606 C. Lepoittevin et al. / Solid State Sciences 11 (2009) the formation of multi-nanophased phase materials but have been shown to lead to a possible oxygenation, which favors the formation of rather thick perovskite slices. The result is the coexistence of m ¼ 1 associated to m 00 varying from 3 to 12 (and more) members. The m! value still evolves from 1 to 2, but with important Dm values, involving the growth of heterogeneous crystallites. The component p decreases from 0.46 to 0.4 (value of the m ¼ 4). The larger homogeneity ranges observed for the solid solutions Fe 2 (Sr 2 x A x )FeO 65 d/2 of the two divalent cations can be understood in terms of oxygen content, since these cations do not provide extra oxygen. The ionic radius of Pb 2þ, close to that of Sr 2þ, its isovalence and lone pair support the significant homogeneity range of the Pb-doped 2201 with regard to the Ba-2201 s. The formation of crystallographic shear planes is a mechanism easily engaged by varying the cation ratio or introducing Ba 2þ or Pb 2þ in the Bi cuprates and Bi ferrites relatives [11,37,38,39]. Increasing the lead content involves indeed the collapse of the Fe Pb framework, with the formation of the terrace phase Pb 4 Sr 13 Fe 24 O 53 [26]. The role of Ba 2þ is that of a divalent cation, but is assumed to be limited by its larger radius and its 6s 0 electronic structure. Fig. 16. [010] HAADF STEM image of [2201] 1/2 [2212] 1/2 polytype (FFT in insert). the sequence ([TBP[][TBPY])([TP][TP]) of the iron polyhedra. The previous works [4,5] showed that syntheses under oxygen pressure are not able to stabilize this limit (and even to go beyond with an oxygen content of O 6.5 þ d/2 [2]) and the present work shows that chemical pressure fails as well. The structure of the RS block is likely one of the key factors of the non-stoichiometry mechanisms in the Fe-2201 phases. A second interesting point deals with the non-stoichiometry mechanisms, which take place by polytypism in the Bi-phases. The polytypes are only observed in the Bi and Tl-based compounds. In the Bi-based ferrites, two factors are able to play an important role, the oxygen and the lone pair of Bi 3þ ; this electronic property has been reported as playing a major role in the stabilization of the family of the Bi-cuprates and relatives [10] and could, here again, be one of the important parameter of the Fe A -(n 1)2(m 1)m ferrites. Assuming that O 6.5 is the limit oxygen content of the Fe phase, the increase of the Bi content (i.e. extra oxygen) can be ensured either by the insertion of extra (SrFe 3þ O 3 y ) layers (i.e. inserting defective high m members) or by a partial reduction of iron (mixed valence Fe 3þ /Fe 2þ ). The chemical process and TEM study are in favour of the first hypothesis. The Table 4 and diagram (Fig. 1) allow to follow the mechanisms. The existence domain of the 2201 phase is the pink line (m ¼ 1): as x increases, the first higher members m ¼ 2 (green line), 3 and 4 (red line) are formed. The corresponding compositions are located inside the hatched domain, in agreement with the observed deviation of the actual composition. Their density increases as x, short range ordering and polytypes are then stabilized for (Bi þ Sr)/Fe ¼ 0.75, composition on the m ¼ 2 (green line). The oxygen content is indeed too low for stabilizing a Fe-2212 phase (Bi/(Bi þ Sr) ¼ 0.067), so that one observes a disproportionation between the m ¼ 1 and m ¼ 4 (red line) phases. Following the m ¼ 2 (green line) domain, the first m ¼ 2 members are observed for a minimum oxygen content obtained for a cation ratio close to Bi/(Bi þ Sr) ¼ 0.1 up to its stabilisation. In that way, the m value evolves from 1 to 2 without large variation of the perovskite block (Dm > 3) and the component p decreases from 0.46 to 0.33 (value of the m ¼ 2), keeping a trivalent iron state. The volatility of the thallium oxide is most likely at the origin of this behavior. In the Tl ferrites, the concomitant insertion of higher m members in the m ¼ 1 matrix and loss of thallium have a double effect, as shown in Fe 2 (Sr 2 e Tl e )Sr 3 Fe 4 O [17]. They often induce 5. Conclusion The investigation of the Fe 2 (Sr 2 x A x )FeO (3 y) þ (3.5 d/2) systems, A ¼ Bi, Tl, La, Pb, Ba, allowed to determine the limits of the solid solutions of the Fe A type phases. These observations show that one crucial factor is the oxygen content of the double iron layer. They provide information on the original triple RS blocks of these ferrites, which appear to be one of the key parameters of the non-stoichiometry mechanisms. They suggest that the sequence ([TBP[][TBPY])([TP][TP]) of the iron polyhedra, associated with the ideal p z 0.5 limit value, could be difficult to build. One of the reasons could originate from the intense stress created by the simple association of these two structural units, which cannot be released through a single perovskite slice. In the phases synthesized up to now, the relaxation of this strain effect by increasing the thickness of the perovskite slice involves systematically a decrease of the p values combined with an increase of the oxygen content in the oxygen deficient perovskite layers (SrFeO 3 y ). Bi- and Tl-based systems evolve by the intergrowth of different members. Several new variant and polytypes have been observed and characterized by TEM techniques. The Pb-based system evolves through a collapsing mechanism through the formation of a crystallographic shear phase. Two major factors have been observed to govern the non-stoichiometry mechanisms in the Fe 2 (Sr 2 x A x )FeO 6.5 1/2d phases. Whatever the A valence, the lone pair of the cation appears to be a favourable parameter. For the trivalent cations, which tend to increase the oxygen content, the phases mainly evolve by decreasing the strain effect that an excess of oxygen would involve in the RS block. One observes a double mechanism, which leads to increase concomitantly the number of polyhedra between the ([TBP[][TBPY]) structural units in the RS block and the oxygen content in the oxygen deficient perovskite slices (SrFeO 3 y ), which both act by releasing the strain effects. Acknowledgements The authors acknowledge financial support from the European Union under the Framework 6 program under a contract for an Integrated Infrastructure Initiative. Reference ESTEEM. References [1] F. Kanamuru, M. Shimada, M. Koizumi, J. Phys. Chem. Solids. 33 (1972) [2] S. Guggilla, T. Armstrong, A. Manthiram, J. Solid State Chem. 145 (1999) 260.