Supporting Information

Transcription

1 Supporting Information Wiley-VCH Weinheim, Germany

2 A pure bismuth A site polar perovskite synthesised at ambient pressure Craig A. Bridges, Mathieu Allix, Matthew R. Suchomel, Xiaojun Kuang, Iasmi Sterianou, Derek C. Sinclair, Matthew J. Rosseinsky* Experimental details Solid state reactions: Stoichiometric amounts of Bi 2 O 3 (Alfa Aesar %), TiO 2 (Alfa Aesar %), Fe 2 O 3 (Alfa Aesar %), and either NiO (Alfa Aesar %) or MgO (Sigma Aldrich %) were mixed under acetone in an agate mortar and pestle. The resulting powder was pelletised and placed on Pt foil in an alumina crucible. All ramp rates were typically 5 C min -1 for both heating and cooling. Syntheses were performed to examine the influence of cooling rate at both 1 C min -1 and 10 C min -1 - this did not significantly alter the final product, as determined by powder X- ray diffraction. Reactions were typically carried out on the 0.8g scale; attempts to scale up the reaction were hampered in general by an increase in the level of impurities. However, it was found that by ball milling the reagents for 16 hours under acetone with Mg-stabilised zirconia milling media at the initial stage, as well as during the intermediate regrinding stages, up to 20g of material could be successfully mixed at one time to afford material equivalent to that prepared on a smaller scale. The initial target composition was BiTi 1/3 Fe 1/3 Ni 1/3 O 3 : subsequent compositions investigated are shown in Figure S1. Initial syntheses were carried out in air at 800 C, 850 C, and 900 C for 12 hours each, with intermediate regrinding. Final firings were

3 performed at 925 C for the BiTi 3/8 Fe 2/8 Mg 3/8 O 3 and 975 C for BiTi 3/8 Fe 2/8 Ni 3/8 O 3 systems. The synthesis of BiTi 3/8 Fe 2/8 Mg 3/8 O 3 is best achieved by firing in the following sequences, each followed by regrinding and pelletising: 750 C 24h / 800 C 12h; 800 C 3 h / 850 C 12h; 850 C 3 h / 900 C 12 h; final firing 875 C 12h. The synthesis of BiTi 3/8 Fe 2/8 Ni 3/8 O 3 is best achieved by firing in the following sequences, each followed by regrinding and pelletising; 750 C 24h; 850 C 6h / 900 C 6h; 950 C 36h; 975 C 18h. Reaction above 925 C for the Mg case or above 975 C in the Ni case, or prolonged heating at the maximum temperature for either compound, eventually led to sample decomposition. Ceramics: Disks for impedance and polarization measurements were prepared using the synthesis protocol detailed above for each compound. However, for the polarization measurements, before the final firing ~0.3g of powder were mixed with a polyvinyl acetate based binder and pressed in a 10mm die. These pellets were then sealed within an evacuated latex bag and pressed using a cold isostatic press at 2000 bar. After the final firing step the pellets of BiTi 3/8 Fe 2/8 Mg 3/8 O 3 were found to be 90(3)% of theoretical density on average. Diffraction: X-ray powder diffraction data were collected with Co Kα 1 radiation using a Panalytical X Pert Pro diffractometer equipped with an X Celerator detector. Neutron data were collected for BiTi 3/8 Fe 2/8 Ni 3/8 O 3 on the Polaris instrument (resolution d/d ) at the ISIS spallation source, Rutherford Appleton Laboratory at room temperature. Room temperature neutron diffraction data were collected for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 on the HRPD instrument (resolution d/d ) at the ISIS spallation source. Data from the backscattering bank were used in the refinements.

4 Variable temperature data were collected on HRPD with the BiTi 3/8 Fe 2/8 Mg 3/8 O 3 sample contained within a silica tube open to air, using the sequence 22 C, 650 C, 740 C, then C in 5 C steps followed by a final temperature of 650 C. High temperature data were collected to 20µA with the exception of 740 C (125µA). Absorption was refined for using a linear absorption function suitable for time of flight data. Powder neutron and X-ray diffraction data were analysed by Rietveld refinement using the programmes GSAS and Fullprof. Electron microscopy: Energy dispersive spectroscopy (EDS) data were collected with a JEOL 2000FX electron microscope. Samples were prepared by crushing the powder in n- butanol and the small crystallites in suspension were deposited onto a holey carbon film, supported by a copper grid. Data were recorded at the Bi L edge and at the K edge for Ti, Fe, Ni, and Mg. Electron diffraction data indicated well crystallized particles and were consistent with the derived R3c space group symmetry. Electrical measurements: Dielectric and impedance properties were investigated as a function of frequency and temperature using a Solatron 1255B Frequency Response Analyzer and a Solatron 1296 dielectric interface over the frequency range of Hz. Electrodes made of Pt paste, fired at 800 C, were used with Pt lead wires to provide electrical contacts for sintered pellets. Samples of BiTi 3/8 Fe 2/8 Ni 3/8 O 3 were measured between 25 C and 400 C; above this temperature the samples became too conductive to measure. Samples of BiTi 3/8 Fe 2/8 Mg 3/8 O 3 were measured on both heating and cooling between 25 C and 850 C. Ferroelectric switching measurements were performed at room temperature on thin (t = 0.30 mm) dense sintered pellets (ø = 6.5 mm) using a Radient RT66A testing system operating in virtual ground mode.

5 Phase stability and composition Figure S1: (a) Investigated compositions in the Bi-Ni-Ti-Fe-O phase field at 850 C. (b) Investigated compositions in the Bi-Mg-Ti-Fe-O phase field. The dashed line indicates an

6 average B site oxidation state of 3+. For the target composition reported here (shown in red), an Mg phase was prepared with 1-3% of a Sillenite type impurity and negligible quantities of Aurivillius type and spinel impurities. A Ni phase containing approximately 0.3% NiFeO 4 as the only impurity was prepared. A solid solution of BiTi 1/3+x Fe 1/3-2x Mg 1/3+x O 3 (estimated to be in the range of < x < 0.084) was observed by EDS around the composition BiTi 3/8 Fe 2/8 Mg 3/8 O 3. BiTi 3/8 Fe 2/8 Mg 3/8 O 3 was isolated as a single phase and discussed in the paper. The associated unit cell parameters are given in Table S1. Table S1: Unit cell parameters for reaction compositions along the line BiTi 1/3+x Fe 1/3-2xMg 1/3+x O 3. [a] The existence of a solid solution was not immediately evident from the lattice parameters, which vary little over a wide composition range. Although the relatively low impurity level for the compositions BiTi Fe Mg O 3 and BiTi Fe Mg O 3 was not alone considered sufficient evidence for a solid solution, the existence of a solid solution was confirmed through the use of EDS studies on the composition BiTi Fe Mg O 3. Though higher impurity levels were observed for the BiTi Fe Mg O 3 reaction, this was chosen as it presented the best opportunity to observe a composition different from that of BiTi 3/8 Fe 2/8 Mg 3/8 O 3. The EDS data indicated a composition for the perovskite phase which differed significantly from BiTi Fe 0.25 Mg O 3 ; given that some uncertainty exists in absolute composition when utilizing the EDS method alone, no attempt is made here to define a precise composition range. The range of this solid solution will be defined in greater detail in future work involving a combination of detailed Rietveld refinement and EDS studies. The apparent lack of variation in lattice parameters may explained simply on the basis of ionic radii:

7 the similar size of the Fe 3+ cation (0.645Å) as compared to the average size of the Ti 4+ (0.605Å) and Mg 2+ (0.72 Å) cations, which is 0.66Å. The existence of solid solution is significant in the context of controlling the electroceramic behaviour of these materials. Reaction Composition a (Å) c (Å) V(Å 3 ) BiTi Fe 0.29 Mg O (7) (3) (9) BiTi Fe 0.25 Mg O (4) (1) (4) BiTi Fe Mg O (3) (1) (3) BiTi Fe Mg O (7) (2) (7) [a] All lattice parameters were calculated by first extracting peak positions, followed by least-squares refinement in the space group R3c. A sample height displacement term was included.

8 Table S2: Parameters characterizing the tilting and displacements of rhombohedral perovskites from the ideal Pm 3 m parameters. All phases given are in the R3c space group for data collected at room temperature, and include BiTi 3/8 Fe 2/8 Mg 3/8 O 3 (BTFM), BiTi 3/8 Fe 2/8 Ni 3/8 O 3 (BTFN), BiFeO 3 (BFO) and PbZr 0.53 Ti 0.47 O 3 (PZT). [a] The parameters s and t indicate displacements of the A site and B site respectively, d the displacement of the oxygen y parameter, ω is the mean tilt angle along the [111] p direction of the pseudocubic perovskite cell, l is the mean octahedral edge length, l is the deviation from the mean edge length, K is computed using a = 2Kl cosω, and V A and V B are the polyhedral volumes around the A and B sites. The meaning of these parameters is described in further detail in references [a] and [b] below. The value of s given in the table for BTFM and BTFN are for the Bi high symmetry position on the threefold axis (000) in the models used here; if the distortion perpendicular to the hexagonal c axis is considered, the displacements are Å and 0.618Å respectively. BTFM BTFN BFO [c] PZT-F [a] R (x Zr =0.53) s (2) (0.651Å) (5) (0.606Å) (5) (0.676Å) 0.029(2) (0.412Å) t (3) (0.244Å) (9) (0.222Å) (5) (0.272Å) (0.118Å) d (2) (9) (3) (5) ω ( o ) 11.94(5) 12.21(3) 12.2(2) 1.4(9) l (Å) 2.864(2) 2.857(2) (4) l (Å) 0.137(2) 0.131(2) (4) K V A (Å 3 ) 52.15(3) 51.35(3) V B (Å 3 ) 11.00(3) 10.85(3) V A /V B 4.74(3) 4.73(3) [a] J. Frantti, S. Ivanov, S. Eriksson, H. Rundlöf, V. Lantto, J. Lappalainen, and M. Kakihana, Physical Review B 2002, 66, [b] N.W. Thomas and A. Beitollahi, Acta Crystallographica B.1994, 50, 549. [c] P. Fischer, M. Polomska, I. Sosnowska, M. Szymanski, Journal of Physics C: Solid State Physics 1980, 13, 1931.

9 Diffraction data analysis Figure S2: Rietveld refinement for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 from (a) powder X-ray diffraction (R wp = 18.85%) and (b) powder neutron diffraction data (R wp = 28.8%) at room temperature in the centrosymmetric non-polar R 3 c space group with 3 m1 microstructure parameters (χ 2 overall = Powder neutron data were collected on the HRPD beamline at the ISIS facility, and powder X-ray data on a Panalytical X-pert Pro diffractometer.

10 The microstrain in BiTi 3/8 Fe 2/8 Mg 3/8 O 3 is apparent in both laboratory powder X-ray and high resolution powder neutron data which cannot be fitted in R3c symmetry with either isotropic or anisotropic 3 m1 symmetry strain broadening appropriate to the R3c spacegroup (Figure S3) I /10 4 counts θ / o I /10 3 counts d / Å Figure S3: Rietveld refinement for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 from (a) powder X-ray diffraction (χ 2 cumul = 7.79, R wp = 12.0%) and (b) powder neutron diffraction data (χ 2 cumul = 3.45, R wp = 11.7%) at room temperature in the R3c space group with 3 m1 microstructure parameters appropriate for the Laue symmetry of R3c. Powder neutron data were collected on the HRPD beamline at the ISIS facility, and powder X-ray data on a Panalytical X-pert Pro diffractometer. The fits here show that lower symmetry strain broadening is required to fit the powder profiles Figure 2 shows the improvement associated with this.

11 Le Bail fits to the high resolution powder neutron diffraction data from BiTi 3/8 Fe 2/8 Mg 3/8 O 3 converge to χ2 = 1.24 (Cc, 2/m strain) and 1.32 (R3c, 1 strain). Comparison of Rietveld refinements gave χ2 = 1.43 (Cc, 2/m strain, 31 parameters) and 1.40 (R3c, 1 strain, 29 parameters, anisotropic displacement parameters). This leads to selection of the higher symmetry model discussed in the main text. It should be noted that χ2 = 1.47 when isotropic temperature factors are used in the R3c fit, supporting the use of anisotropic displacement parameters on O and Bi. Combined refinement of the X-ray and neutron powder diffraction data for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 was carried out in lower-symmetry centric (I2/c) and non-centric (Pc and C1) space groups with lower quality fits in all cases. The centric space group I2/c has the same systematic absences as non-centric Cc; the restrictions imposed by the inversion centre on the crystal structure resulted in a very poor fit (χ 2 = 8.92). Figure S4(a) View along threefold axis (c direction) showing elongation of the Bi anisotropic displacements perpendicular to this axis in BiTi 3/8 Fe 2/8 Mg 3/8 O 3 (R3c, refined with -1 symmetry microstrain). Bi-O bonds are shown as red 2.29 Å blue 2.52 Å yellow 3.20Å white 3.44 Å

12 Figure S4(b): Equivalent view of the anisotropic displacement parameters at the Bi site in BiFeO 3 Kubel, F.;Schmid, H. Acta Crystallographica B (1990) 46, (structure refined by single crystal X-ray diffraction). The three short Bi-O bonds due to the [111] p displacement are shown as thick lines. The unphysical shape of the Bi ADP in BiTi 3/8 Fe 2/8 Mg 3/8 O 3 lead to the final model described in the main text with disordered static local displacements perpendicular to the threefold axis within the xy plane.

13 I /10 4 counts d / Å I /10 4 counts d / Å Figure S5: Rietveld refinement for BiTi 3/8 Fe 2/8 Ni 3/8 O 3 from powder neutron diffraction data at room temperature in space group R3c with microstructure parameters appropriate for (a) 2/m (χ 2 cumul = 2.20, R wp = 1.08%), and (b) 3 m1 (χ 2 cumul = 2.52, R wp = 1.16%) Laue class. The refinement with 3 m1 microstructure parameters has been reported due to the relatively minor improvement in fit with the additional 2/m parameters. Powder neutron data were collected on the Polaris beamline at the ISIS facility. A weak impurity (~0.3%) of NiFe 2 O 4 was observed in this case, indicated by the lower set of reflection markers. A weak peak at 2.14Å from the vanadium sample can was not included in the refinement.

14 Fig S6: Arrhenius plot showing temperature dependence (T -1 ) of the electrical conductivity (ln σ) of BiTi 3/8 Fe 2/8 Mg 3/8 O 3 and BiTi 3/8 Fe 2/8 Ni 3/8 O 3 (upper and lower, respectively). In BiTi 3/8 Fe 2/8 Mg 3/8 O 3, a reversible hysteretic transition is evident at between 730 C (on heating) and 680 C (on cooling). Activation energies calculated for the BiTi 3/8 Fe 2/8 Mg 3/8 O 3 sample from a fit to the data are 1.41 ev above the transition (HT) and 1.29 ev below the transition (LT). In the more conductive BiTi 3/8 Fe 2/8 Ni 3/8 O 3 sample, no transition in observed and the calculated energy is 0.41 ev.

15 Figure S7: Rietveld refinement for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 from powder neutron diffraction data for the (a) backscattering bank (R wp = 14.16%), and (b) 90 o bank at 740 o C (R wp = 6.72%) in the Pm 3 m space group (χ 2 overall = 3.105). Powder neutron data were collected on the HRPD beamline at the ISIS facility. A weak peak due to the vanadium sample can at ~2.17Å has not been modeled. Displacement of the oxygen site onto a site with higher multiplicity was used to obtain reasonable bond valence sums around each ion, but due to the relatively poor quality of the data difficulty arose in obtaining sensible atomic displacement parameters. For that reason, only variations in lattice parameter and phase fraction are reported.

16 (a) Reduced Cell Parameter (Å) BTFM - cubic BTFM - a r BTFM - c r Temperature (K) (B) BTFM - cubic BTFM - R3c Volume (Å 3 ) (C) Phase Fraction Temperature (K) BTFM - Cubic BTFM - R3c Temperature (K) Figure S8: Variation in (a) lattice parameter, (b) volume and (c) phase fraction for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 from powder neutron diffraction data. Data shown were collected during cooling from 740 o C to 650 o C. Reduced hexagonal lattice parameters corresponding to a/ 2 and c/ 12 are used for ease of comparison in (a). Powder neutron data were collected on the HRPD beamline at the ISIS facility. A transition near 685 o C from the high temperature cubic phase to the lower temperature R3c phase is evident. The mixed phase nature of the sample in the shaded region on the plot, as well as the rapid volume change, provide further evidence for the first order nature of the transition. Lattice parameters for the R3c phase were not included at 285 o C; the peaks from this phase were too weak to be refined at this temperature. The observed phase coexistence region matches the hysteresis observed in the conductivity (Figure S6) and permittivity (Figure 3).

17 Fig S9: Dielectric permittivity (solid data points) and dielectric loss (open data points) of BiTi 3/8 Fe 2/8 Ni 3/8 O 3 plotted vs temperature for three selected frequencies.

18 Figure S10: The mean octahedral site position (cyan) is displaced towards one triangular face of the octahedron to give three short (blue) and three long (grey) bonds and away from the face towards which the neighbouring A site Bi cation is displaced (indicated by the blue short Bi-O bonds). The displacement from the centroid is 0.24Å for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 (similar to 0.27Å in BiFeO 3 ) while that of BiTi 3/8 Fe 2/8 Ni 3/8 O 3 (0.22Å) is slightly smaller.

19 Table S3: Strain Parameters from the Stephens Model. Parameters were obtained using the program Fullprof. The BiTi 3/8 Fe 2/8 Ni 3/8 O 3 Polaris data were refined using the 3 m1 microstructure parameters, while for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 both data required 1 microstructure parameters. Parameter BTFN (Polaris) BTFM (HRPD) BTFM (XRD) S (5) 1.5(1) 1.86(5) S (3) 2.4(2) S (1) (1) (7) S S (1) 2.17(3) S (1) 2.83(6) S (1) -2.1(7) S 121-7(2) -0.2(8) S (3) 2.8(4) 2.0(2) S (8) 4.4(5) S (2) -2.99(7) S 130-5(1) 11.2(6) S S S Lorentzian mixing (X) 0.26(2) 0.421(9) 0.520(3)

20 0.8 BiTi Fe 0.5 Ni O 3 BiTi Fe 0.5 Mg O FWHM / o 2θ θ / o Figure S11: Comparison of the variation in FWHM vs 2θ for BiTi 3/8 Fe 2/8 Ni 3/8 O 3 (square) and BiTi 3/8 Fe 2/8 Mg 3/8 O 3 (circle) taken from the Panalytical XRD data. The greater dispersion of FWHM for the Mg 2+ case is evident, which relates to a greater degree of microstrain.

21 Figure S12: Rietveld refinement for BiTi 3/8 Fe 2/8 Mg 3/8 O 3 from powder X-ray diffraction data. (χ 2 cumul = 3.20, Rwp = 7.67%) at room temperature in the R3c space group with 1 microstructure parameters. Powder X-ray data were collected on a Panalytical X-pert Pro diffractometer. A weak impurity of approximate composition Bi 25 Ti 2.3 Fe 2.0 Mg 2.1 O x (Sillenite type from EDX data) corresponds to the upper set of reflection markers. The weak impurity was not observed in the neutron diffraction data due to the higher noise level of these data.