Supporting Information. Designing a Lower Band Gap Bulk Ferroelectric Material with a Sizable. Polarization at Room Temperature

Transcription

1 Supporting Information Designing a Lower Band Gap Bulk Ferroelectric Material with a Sizable Polarization at Room Temperature Shyamashis Das, Somnath Ghara, # Priya Mahadevan, A. Sundaresan, # J. Gopalakrishnan, and D. D. Sarma *, Solid State and Structural Chemistry Unit, Indian Institute of Science, Bengaluru # Chemistry and Physics of Materials Unit, Jawaharlal Nehru Centre for Advanced Scientific Research, Jakkur, Bengaluru Department of Condensed Matter Physics and Material Sciences, S.N. Bose National Centre for Basic Sciences, Block JD, Sector-III, Saltlake, Kolkata Corresponding Author s ID: sarma@iisc.ac.in

2 Experimental Section Synthesis: All samples were synthesized in the conventional solid state method. For synthesis of BaTi1-x(TM1/2Nb1/2)xO3 with x 0.15, for TM = Mn, and 0.02 x for TM = Fe, first BaCO3 was preheated at 500 C for 5 hr. to remove any absorbed moisture and then stoichiometrically weighed amounts of BaCO3 (Sigma Aldrich, 99.9%), TiO2 (Sigma Aldrich, 99.9%), Mn2O3/Fe2O3 (Sigma Aldrich, 99.99%) and Nb2O5 (Sigma Aldrich, 99.99%) were ground thoroughly for 2 hr. in an agate mortar and pestle. The resulting mixture was heated in air at 1000 C for 12 hr. for decarboxylation of BaCO3. The powder obtained after this step was again ground for 1 hr. and pressed into pellets of 6 mm diameter and sintered in air at a temperature of 1300 C for 2 hr. Measurement Methods: Powder X-ray diffraction patterns were recorded in a PANalyitcal diffractometer using Cu K radiation for the angular range with step size. Dielectric measurements were performed using an impedance analyzer (HP4194A) over a frequency range of 1 khz - 1 MHz within the temperature range of K. Polarization vs electric field (P-E) hysteresis loops were recorded within frequency range Hz using a Radiant s work station. Pyrroelectric current was measured using a Keithley 6517A electrometer which was attached to a multifunctional probe inside the physical property measurement system (PPMS, Quantum Design). Pyroelectric current measurements were made by first cooling the sample in an electric field from 380 K to 220 K, and, then after removing the field, the sample was shorted for a long time to remove stray charges; following these steps the pyrocurrent was recorded as the sample was warmed up with a rate of 3 K per minute. For all electrical measurements (dielectric, polarization and pyrroelectric), electrical contacts were made by painting conducting silver paste on the two opposite sides of a flat pellet of the sample. Two

3 conducting electrical cables were then attached to the silver paste on the two faces of the pellet using small amount of silver paste as the adhesive. These electrical wires serve as leads to attach the pellet to the respective instruments for specific measurements. Electrical contacts were cured by heating the pellet with contacts at 120 C for about half an hour. The area and thickness of each pellet were measured prior to making electrical contacts. Diffused reflectance spectra for powder samples were recorded using Perkin-Elmer UV-Visible spectrometer with a range from near IR to ultraviolet ( nm) wavelength. Powder X-ray diffraction and Rietveld Refinement All the powder XRD patterns of the bulk samples were recorded with a PAnalytical diffractometer using Cu K radiation ( = Å). The data was recorded with very slow scan. A scan speed of 0.5 per minute with step-size is required to get a good signal to noise ratio. Firstly, to check the presence of any impurity phase the diffraction patterns were analyzed with respect to the standard diffraction pattern of the parent composition available at ICSD XRD database. In the next step the diffraction patterns of pure phase materials were subjected to Rietveld refinement using Fullprof software. The starting structure for the refinement is tetragonal phase of BaTiO3. For the compositions BaTi1-x(Mn1/2Nb1/2)xO3 with x = 0.1 and 0.15 room temperature PXRD patterns do not show any tetragonal splitting in the diffraction peaks. So, these two compositions are refined in cubic phase. For refinement all the occupancies are set to full and Ti 4+, Mn 3+ and Nb 5+ cations are constrained to the same atomic co-ordinate. Fractional co-ordinate of B-site cations and of oxygen ions are refined independently. For all the samples investigated here, lattice parameters, atomic positions, isotropic displacement parameters, scale factors and background parameters are refined independently.

4 Computational Methods To compare ferroelectric stabilization energy, we performed first principle density functional theory (DFT) calculation using projector augmented wave (PAW) method as implemented in the Vienna ab initio Simulation Package (VASP) 1-2 for one representative composition (x 0.11) from each series of compounds BaTi1-x(Mn1/2Nb1/2)xO3 and BaTi1- x(fe1/2nb1/2)xo3, and for one specific realization of a more dilute dopant composition (x 0.06) in order to keep the computational effort tractable, as explained later in this section. Generalized Gradient Approximation (GGA) potential was used for the exchange correlation part. The plane wave energy cutoff was 550 ev and a k-mesh dimension of has been used for Brillouin zone sampling. GGA+U calculation was performed with a U on the dopant transition metal site equal to 4 and 5 ev for Mn and Fe, respectively, consistent with earlier estimates for doped compositions. 3-4 We compared experimentally measured optical band gaps with gaps obtained from electronic structure calculations. However, density functional calculations that use GGA and local density approximation (LDA) for the exchange correlation functional have a limitation that it underestimates the optical band gap of most materials. Hybrid functional calculation is known to overcome this drawback. 5 Accordingly, we have used hybrid functional HSE06 to compute electronic density of states and to obtain theoretical estimates of bandgaps. Other details of these calculations are same as GGA+U calculations that are discussed above. We note that it is necessary to perform such calculations for substituted systems using large supercells; on the other hand, the calculation complexity increases with number of atoms in such a supercell, thereby effectively limiting the size of the supercell that can be employed. For most of the calculations presented here, we use a supercell starting with experimental cell parameters of the tetragonal BTO with two Ti atoms replaced separately by Mn-Nb pair and Fe-

5 Nb pair to represent the composition BaTi1-x(TM1/2Nb1/2)xO3, x 0.11 with TM = Mn and Fe, respectively. Within this supercell, this substitution of two Ti atoms can be achieved in six symmetry inequivalent ways as shown in Figure S2 (Supporting Information) and we performed calculations for each of these symmetry-distinct realizations of specific substituted systems. In order to probe the ferroelectric stability of the doped composition, we identified the ground state of the tetragonal structure with the lowest total energy; for the same realization of the dopant distribution, we carried out an independent geometry optimized calculation of the paraelectric state in the cubic structure. The energy difference between the ground state tetragonal structure and the paraelectric cubic structure provides the estimate of the ferroelectric stabilization for that composition. We note that x 0.11 is slightly larger than the most relevant compositions investigated by us. This prompted us to perform the ferroelectric stability calculation for an even larger sized (4 3 3) supercell, thereby allowing us to achieve a lower composition of x However, this proved to be very computation intensive, making it impossible to explore all possible realizations of dopant distributions and we restrict calculations for x 0.06 with this higher supercell only for the ground state tetragonal structure and the corresponding paraelectric cubic structure, ensuring proper geometry optimization in both cases. Local geometry around dopant ions In order to understand the origin of better ferroelectric properties of the samples BTMNO series compared to those of BTFNO series, we have carried out large-scale firstprinciple calculations based on the density functional theory on supercells of sizes and basic units with Mn/Fe and Nb replacing two Ti sites in each supercell, as discussed in the Computational Methods section. It is stated there that there are six symmetry distinct ways where

6 the two dopants, Mn/Fe and Nb, can be distributed in a supercell, as shown in Fig. S2. The relative total energies and various bond distances obtained by geometry optimization of each of these six structures are listed in Table S2 and S3 for BTMNO and BTFNO, respectively. For both BTMNO (Table S2) and BTFNO (Table S3), the configuration III with the two dopant ions situated along the c-axis (see Fig. S2) comes out to be the lowest energy or the ground state and therefore, the ferroelectric stabilization energy was calculated for this configuration with the two dopant ions situated along the c-axis. Depending on the local geometry we have classified two types of Ti atoms in these configurations. The Ti atoms residing adjacent to dopant atoms (Mn/Fe and Nb) are labeled as 1 Ti and the others which are not immediate neighbors of dopants are labelled as 2 Ti in Table S2 and S3. Analysis of local geometry of Ti atoms from the optimized structure of the supercell under consideration indicates that both the types of Ti atoms ( 1 Ti and 2 Ti) possess significant off-centering displacement to contribute ferroelectricity in theses doped samples. Among the dopant transition metal ions, Nb 5+ also resides inside a distorted NbO6 octahedron though the extent of distortion is less compared to Ti 4+ ions. This is qualitatively understood from the fact that both Ti 4+ and Nb 5+ due to their d 0 -electronic configuration undergo second order Jahn-Teller distortion which plays an important role in determining the local geometry of the corresponding ions. On the other hand, Mn 3+ because of 3d 4 electronic configuration shows first order Jahn-Teller distortion in MnO6 octahedra which is reflected in the longer Mn-O bond-lengths along c-direction in lowest energy or ground state configuration (configuration III in Table S2). This expansion of Mn-O bonds of MnO6 octahedra results a compensating compression in adjacent Nb-O bonds of NbO6 octahedra. This breathing type distortion is however absent in case of substitution with Fe-Nb pair as reflected in almost

7 equal Fe-O bond-lengths of FeO6 octahedra in its ground state configuration (configuration III in Table S3). Comparison of ferroelectric stabilization energy between BTMNO and BTFNO series The ferroelectric stabilization energies, defined as energy stability of the ferroelectric configuration compared to the corresponding paraelectric state and shown in Table S4, predict the experimentally observed trends very well, thereby helping us to understand microscopic origin of our experimental findings. For example, the Mn-Nb and Fe-Nb doped compounds with the lower concentration of dopants (x 0.06) are shown to have ferroelectric ground states in Table S4, with ferroelectric state being stable compared to paraelectric state by about 11 and 4 mev/f.u., respectively, in agreement with our experimental finding of the nearest composition, x = 0.05, for both series being ferroelectric. Moreover, the enhanced stability (~11 mev/f.u.) of the ferroelectric state for Mn-Nb substitution compared to that (~ 4 mev/f.u.) of the Fe-Nb substitution is reflected in the higher transition temperature, TC, for Mn-Nb (378.5 K) as compared to that (315.8 K) for the Fe-Nb substituted compound at x = 0.05; also, a larger polarization of the Mn-Nb compared to the Fe-Nb compounds reflects a more robust ferroelectric state in the former. We have already discussed the experimental trends of a rapid suppression of the ferroelectric state with an increasing substitution, the trend being more pronounced in Fe-Nb substituted series. This tends are also accurately reflected in calculated results. Table S4 clearly reflects the rapid suppression of the ferroelectric state in favor of the paraelectric state with an increase of x from 0.06 to 0.11 in terms of a strong reduction of the ferroelectric stabilization energy, making the ferro- and para-electric states nearly degenerate for Mn-Nb compound and the paraelectric state more stable for the Fe-Nb compound with x 0.11.

8 Table S1. Rietveld refinement results of BaTi1-x(Mn1/2Nb1/2)xO3. The compositions 0 x are refined in tetragonal P4mm space group and the compositions 0.1 x 0.15 are refined in cubic Pm-3m space group. Composition Space Group a (Å) c (Å) 2 R p R wp x = 0 P4mm x = P4mm x = 0.05 P4mm x = P4mm x = 0.1 Pm-3m x = 0.15 Pm-3m

9 Polarization ( C/cm 2 ) Polarization ( C/cm 2 ) Hz 100 Hz T=77 K (a) BTMNO (x = 0.075) Electric Field (kv/cm) 10 Hz 100 Hz T=77 K (b) BTFNO (x = 0.075) Electric Field (kv/cm) Figure S1. Effect of electric field frequency on the shape of hysteresis loops. Frequency dependent polarization vs electric field ferroelectric hysteresis loops of composition (a) BaTi1- x(mn1/2nb1/2)xo3 with x = and (b) BaTi1-x(Fe1/2Nb1/2)xO3 with x = at T=77 K.

10 Figure S2. Possible distribution of dopant ions in a perovskite structure. Six possible configurations of Mn-Nb or Fe-Nb co-doped BaTiO3 considered for ab initio calculation in order to find out most stable configuration. The Ba atoms are not shown for clarity.

11 Table S2. Relative energies and M-O bond lengths in respective MO6 octahedra of six different configurations of BaTi1-x(Mn1/2Nb1/2)xO3 with x 0.11 which are shown in Fig. S2. The pair of bond lengths which are marked bold indicate distorted geometry along the respective axis. Configuration Relative Energy (mev/f.u.) Bond Length (Å) Axis 1 Ti-O* 2 Ti-O* Mn-O Nb-O a I. 0.6 b c a II. 0.6 b c a III. 0.0 b c a IV. 4.3 b c a V. 2.0 b c a VI. 5.2 b c * Ti atoms residing adjacent to dopant atoms (Mn and Nb) are labeled as 1 Ti and the others which are not immediate neighbors of dopants are labelled as 2 Ti.

12 Table S3. Relative energies and M-O bond lengths in respective MO6 octahedra of six different configurations of BaTi1-x(Fe1/2Nb1/2)xO3 with x 0.11 which are shown in Fig. S2. The pair of bond lengths which are marked bold indicate distorted geometry along the respective axis. Configuration Relative Energy (mev/f.u.) Bond Length (Å) Axis 1 Ti-O* 2 Ti-O* Fe-O Nb-O a I. 2.6 b c a II. 2.6 b c a III. 0.0 b c a IV. 4.6 b c a V. 4.8 b c a VI. 4.5 b c * Ti atoms residing adjacent to dopant atoms (Fe and Nb) are labeled as 1 Ti and the others which are not immediate neighbors of dopants are labelled as 2 Ti.

13 Table S4. Total energy of paraelectric and ferroelectric phase of two sets of composition of Mn- Nb co-doped and Fe-Nb co-doped samples. Ferroelectric stabilization energy which is defined as lowering of energy of ferroelectric phase with respect to paraelectric phase has also been listed. Composition Energy (ev/super Cell) Tetragonal (c/a=1.01) Cubic (c/a=1.0) F.E. Stabilization energy (mev/f.u.) BTMNO (x 0.06) BTFNO (x 0.06) BTMNO (x 0.11) BTFNO (x 0.11)

14 References (1) Kresse, G.; Furthmüller, J. Efficient Iterative Schemes for ab initio Total-energy Calculations Using a Plane-wave Basis Set. Phys. Rev. B 1996, 54, (2) Kresse, G.; Furthmüller, J. Efficiency of ab-initio Total Energy Calculations for Metals and Semiconductors Using a Plane-wave Basis Set. Comput. Mater. Sci 1996, 6, (3) Chainani, A.; Mathew, M.; Sarma, D. D. Electron Spectroscopic Investigation of the Semiconductor-metal Transition in La1-xSrxMnO3. Phys. Rev. B 1993, 47, (4) Chandra, H. K.; Gupta, K.; Nandy, A. K.; Mahadevan, P. Ferroelectric distortions in doped ferroelectrics: BaTiO3:M (M=V-Fe). Phys. Rev. B 2013, 87 (21), (5) Wahl, R.; Vogtenhuber, D.; Kresse, G. SrTiO3 and BaTiO3 Revisited Using the Projector Augmented Wave Method: The Performance of Hybrid and Semilocal Functionals. Phys. Rev. B 2008, 78,