Investigation of Ga doping for non-stoichiometric sodium bismuth titanate ceramics

The electrical properties of Ga3+ doping Na0.5Bi0.5TiO3-based oxygen ionic conductors were studied. The Na0.52Bi0.47Ti1−xGaxO3−δ (x = 0, 0.01, 0.015, 0.02) samples were fabricated by the means of traditional solid-state reaction. The results of AC impedance measurements showed that the bulk conductivity of Na0.52Bi0.47Ti1−xGaxO3−δ samples decreased monotonously with the increase of Ga3+ doping concentration. At 673 K, the bulk conductivity of the Na0.52Bi0.47Ti0.98Ga0.02O3−δ sample is 7.19 × 10–4 S/cm, which is lower than that of the Na0.52Bi0.47TiO3−δ sample under the identical test temperature. The highest total conductivity emerged in the Na0.52Bi0.47Ti0.99Ga0.01O3−δ (x = 0.01) sample with 1.387 × 10–4 S/cm at 623 K, which demonstrated that the slight Ga3+ doping supported the enhancement of the total conductivity. A relaxation peak was observed in Na0.52Bi0.47Ti1−x GaxO3−δ compounds. As the Ga3+ ions were introduced into the Na0.52Bi0.47TiO3−δ compound, there was an increasing trend of the relaxation activation energy educed by the internal friction test. In addition, the oxygen relaxation height decreased with Ga3+ doping, which suggested that the introduction of Ga3+ ions resulted in the decrease of mobile oxygen vacancy.


Introduction
Oxygen ionic conductors have been quite broadly used, such as oxygen pumps, oxygen separation membranes, and solid oxide fuel cells (SOFC) [1][2][3][4][5][6]. In the past several decades, there are multitudinous research groups devoting to studying oxygen ionic conductors [1][2][3]. Recently, Na 0.5 Bi 0.5 TiO 3 (NBT), a new family of ferroelectric material with perovskite structure [7,8], can be found that there is a large leakage conductivity due to the migration of oxygen defect [6]. The oxygen defect mainly comes from the loss of the low melting point elements during preparation [6,9]. It is worth mentioning that the bulk conductivity of the Na 0.5 Bi 0.49 Ti 0.98 Mg 0.02 O 2.965 sample designed by bismuth deficiency and Mg 2? doping can reach 1 9 10 -2 S/cm at the temperature of 873 K [1,7]. The experimental result offers a new option for exploring intermediate-temperature oxygen ionic conductors.
In order to get the higher electrical properties in NBT-based oxygen ionic conductors, higher oxygen vacancy content is necessary. There are two ways to introduce oxygen vacancies into NBT compounds: Bi deficiency and acceptor doping [10]. The ionic conductivity can be greatly improved by introducing a low-level non-stoichiometric defect (\ 1 at.%) in NBT [1,6,7]. Especially, the compositions containing Bi deficiency or Na excess exhibit 3-4 orders of magnitude higher than the compositions with Na deficiency or Bi excess [11]. The other way of introducing oxygen vacancies mainly focuses on the acceptor doping for A-or B-sites. A-site acceptor doping mainly focuses on the trivalent Bi 3? ions replaced by the monovalent ions (Li ? , Na ? , K ? ) or divalent ions (Ca 2? , Sr 2? , Ba 2? ) [1,7,12,13]. Yang et al. [14] have reported that Sr 2? doped Bi-deficient NBT-based compound (Na 0.5 Bi 0.47 Sr 0.02 TiO 2.975 ) is a profitable means to improve the electrical properties of oxygen ionic conductors. The B-site doping mainly concentrates on low valent ions, such as Mg 2? , Ga 3? , or Ti 4? [1][2][3]15]. The bulk conductivity of the Mg 2? doped bismuth-deficient NBT-based compounds (Na 0.5 Bi 0.49 Mg 0.02 Ti 0.98 O 2.965 ) which were designed by the above two methods is higher than that of the stable ZrO 2 doped with 8 mol% Y 2 O 3 [1,6]. Xu et al. have investigated that the impact of K ? and Ga 3? codoped NBT-based oxygen ionic conductors on the electric properties [2]. In our previous work, we have introduced excess Na ? in NBT-based oxygen ion conductor, the bulk conductivity of the Na 0.54 Bi 0.46-TiO 2.96 sample is 1.6 9 10 -3 S/cm at 673 K. Considering the ionic radius of Ga 3? (0.062 nm) is very close to that of Ti 4? (0.061 nm), the matching ion radius is helpful to reduce the elastic strain energy which further supports the formation of stable solid solution [1,2]. Hence, Ga 3? ion was selected as an acceptor ion to substitute the B-site Ti 4? ion included in the Bideficient Na 0.52 Bi 0.47 TiO 3-d compound to gain the higher electrical properties. The Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples were prepared by conventional solid-phase reaction method. The crystalline phase of the Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d compounds were detected by powder X-ray diffraction patterns. The electrical properties of the ceramic samples were studied by impedance spectrum technique. The internal friction spectrum was applied to explore oxygen ion diffusion in the Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples.

Experimental procedure
The Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d compounds (x = 0, 0.01, 0.015, 0.02) were elaborated by conventional solid-phase reaction method using high-purity Na 2-CO 3 , Bi 2 O 3 , TiO 2 and Ga 2 O 3 [13]. In order to eliminate absorbed water and CO 2 , the raw materials mentioned above were dried at 573 K for 12 h. The detailed preparation procedure was shown in Ref. [12]. The initial reactive Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d powders were compacted into the cylindrical and bar samples, then the compacted samples were calcined at 1323 K for 12 h with the same compositions Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-x powder embedded around them.
The crystalline phases of Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d samples were detected by a laboratory X-ray diffractometer (Japanese Science Ultima IV diffractometer) using CuKa incident radiation in the range of 10°B 2h B 80°. The conductive silver paste was well distributed applied on the upper and lower surfaces of the cylinder sample and baked at 973 K for 2 h to serve as an electrode. In order to study the electrical performance of the samples, the impedance spectroscopy technique was applied using Impedance Analyzer instrument (Instrument type: IM 3536, 10-8 MHz) with a frequency range of 1 Hz to1 MHz from 473 to 723 K. The low-frequency internal friction (IF) spectroscopy which performed on an inverted torsion pendulum in the form of forced vibration controlled by a computer was used to study oxygen ion diffusion . Figure 1 shows the XRD patterns of Na 0.5 Bi 0.5 TiO 3 [16] and Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples. Comparing the diffraction pattern with Na 0.5 Bi 0.5 TiO 3 compound, there were no extra peaks of impurity phase in the compositions of Na ? excess and Ga 3? doping NBT samples, which suggested that excess Na ? and Ga 3? ions were dissolved into the perovskite lattice of Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d compounds (PDF #360340). According to Scherrer equation, D = Kk/bcosh, where D is the grain size, K is the Scherrer constant (generally 0.89), b is the full width at half maximum (FWHM) of the diffraction peak, and 2h is the diffraction angle, the grain size is inversely proportional to the FWHM [3]. From the XRD patterns, the FWHM of the diffraction peak For perovskite structure materials, Goldschmidt tolerance factor t is of great significance to the structural stability and properties [17]. By the definition of the Goldschmidt tolerance factor, the tolerance factor of the Na 0.52 Bi 0.47 Ga x Ti 1-x O 3-d (x = 0, 0.01, 0.015, 0.02) samples is 0.97662, 0.97658, 0.97654, and 0.97652, separately. The tolerance factor for the Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples was virtually unchanged, which indicated that Ga 3? doping had very little effects on the lattice distortion. And the depressed semicircles, locating at the intermediate-frequency range, can be ascribed to the grain boundary response [18]. An equivalent circuit formed by three R//CPE elements in series was used to fit the impedance spectrum [19,20]. The curve fitting results are presented in Table 1. The capacitance at high frequency and low frequency is about 10 -10 F and 10 -5 F, which is the typical grain response and electrode response of oxygen ion conductors [12,13,20].

Ionic conductivity
The conductivity of Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-x (x = 0, 0.01, 0.015, 0.02) samples can be obtained by the formula r = L/(SR), where L and S refer to the bottom area and the thickness of the cylinder sample, respectively. The bulk conductivity and total conductivity can be calculated by r b = L/(SR b ) and where R t is the total resistance of the sum of grain resistance and grain boundary resistance [3,20]. Figure 3 exhibits the Arrhenius plots of the bulk conductivity for Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d samples. The bulk conductivity of the samples increases with the rise of measuring temperature. The r b -1/T curve shown in Fig. 3 was split into two sections by the keen point around 563 K, which may be due to the structural phase transition (rhombohedral phase to orthorhombic phase) [15]. According to the Arrhenius law r ¼ r 0 exp ÀE = kT  [1]. When the Ga 3? doping concentration increased from 1 to 2 mol%, the bulk conductivity decreased from 1.01 9 10 -3 to 7.19 9 10 -4 S/cm at 673 K. Figure 5 shows the total conductivity of the Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-x samples dependent on the Ga 3? doping content at the different measuring temperatures [13,15]. With the introduction of Ga 3? in Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-x samples, the total conductivity showed a tendency to increase first and then decrease. When 1 mol% Ga 3? was introduced  Table 1 The fitting results of AC impedance spectra for Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples  [3]. Therefore, the grain boundary conductivity in Na 0.52 Bi 0.47 Ti 0.99 Ga 0.01 O 3-d (x = 0.01) sample is higher, which further leads to higher total conductivity.

Internal friction spectroscopy
The temperature dependence of IF (Q -1 ) for Na 0.52-Bi 0.47 Ga 0.015 Ti 0.985 O 3-d sample at two frequency of 2 Hz and 4 Hz is exhibited in Fig. 6 [13,15]. Within the test temperature range, three obvious IF peaks (entitled by P 1 around 343 K, P 2 around 543 K, and P 3 around 785 K for 2 Hz) can be observed. With the increasing test frequency, the peak position of P 1 peak shifted to higher temperature, which suggested that P 1 peak has typical thermal activation relaxation characteristics [13,21]. For P 2 and P 3 peaks, the peak positions hardly moved when the test frequency changed, which is the representative phase transition peak characteristic. Based on the reported research results, the P 2 and P 3 peaks may result from the transition process, which correspond to the rhombohedral to orthorhombic phase and the orthorhombic to tetragonal phase, respectively [15,22].   [15]. For thermally activated relaxation process, the relation between the relaxation time s and activation energy E can be expressed by Arrhenius law: s ¼ s 0 exp ðE=K B TÞ [13,23]. By changing the tested frequency measurement, the relaxation parameters E and the relaxation time s can be got: E = 0.78 eV for the Na 0.52 Bi 0.47 TiO 3-d sample [13]. It is easy to find that the activation energies of the Ga 3? doping Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0.01, 0.015, 0.02) samples range from 0.83 to 0.86 eV, which indicated that Ga 3? doping is not beneficial to oxygen ion diffusion. Considering the internal friction P 1 peak temperature (around 350 K) and the internal friction characteristics, the internal friction P 1 peaks may be derived from oxygen ions via vacancies shortdistance diffusion in the rhombohedral phase  [12]. Considering the experimental errors, it can be found that the activation energy from the electrical conductivity test (low-temperature section) is almost identical to that of the internal friction measurement.
In order to further study the influence of Ga 3? doping NBT-based oxygen ion conductor on oxygen ion diffusion, the P 1 peak curves of the Na 0.52 Bi 0.47-Ga x Ti 1-x O 3-d samples measured at 4 Hz are shown in Fig. 8. Considering the clarity of the image, the curve of the Na 0.52 Bi 0.47 Ti 0.98 Ga 0.02 O 3-d sample was not given. Although the IF relaxation peak positions almost unchanged, the IF relaxation peak height decreased with the increase of Ga 3? doping concentration, which can be clearly observed in the inset of Fig. 8.

Discussion
Through bismuth deficiency and Na/Ga acceptor doping methods, oxygen vacancy can be introduced into NBT-based compounds just like the following Kroger-Vink equations [3,13]: ð1Þ According to the principle of electric neutrality, there is 3.5 mol% nominal oxygen vacancy in the Na 0.52 Bi 0.47 TiO 3-d sample, which results from Bi 3? non-stoichiometric defect and Na ? acceptor doping. Additionally, there are more 0.5 mol% oxygen vacancies that can be introduced into the Na 0.52-Bi 0.47 Ti 0.99 Ga 0.01 O 3-d compounds when the Ga 3? doping content is 1 mol%, which suggested that oxygen vacancy concentration can be increased with introducing Ga 3? ions in the Na 0.52 Bi 0.47 Ti 1-x Ga x-O 3-d samples.
According to the reported results [12], not all oxygen vacancies are involved in the oxygen ion diffusion migration and there are immobile oxygen vacancies in the oxygen ionic conductors. Normally, the mobile oxygen vacancies are of great significance to the electrical properties of oxygen ion conductors [3]. Based on the point defect relaxation theory, the height of the relaxation peak has a linear relationship with the mobile vacancy concentration [13]. The inset of Fig. 8 [1,2].
For the perovskite structural oxygen ion conductors, oxygen ion migration mainly passes through the Na-Bi-Ti saddle point as a rate-limiting step [7]. When Ga 3? ions were introduced into the Na 0.52-Bi 0.47 TiO 3-d compound, the polarizability of the donor ion (a Ga = 1.50 Å 3 ) is lower than that of the substituted ion (a Ti = 2.93 Å 3 ), which is unfavorable for oxygen vacancy transport [2]. The increase of the oxygen relaxation activation energy educed from the IF measurement also proved that Ga 3? doping is not beneficial to oxygen ion diffusion in NBT-based oxygen ion conductors.

Conclusion
Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d (x = 0, 0.01, 0.015, 0.02) samples with single perovskite phase were fabricated by the way of conventional solid-state reaction. Through the AC impedance test, a slight of Ga 3? doping can decrease the grain boundary resistivity and increase the total ionic conductivity. The highest total conductivity emerged in Na 0.52 Bi 0.47 Ti 0.99 Ga 0.01-O 3-d (x = 0.01) sample, which is 1.387 9 10 -4 S/cm at 623 K. The bulk conductivity of Na 0.52 Bi 0.47 Ti 1-x Ga x O 3-d samples exhibited a monotonous reduction trend with the increase of Ga 3? doping. The bulk conductivity of Na 0.52 Bi 0.47 Ti 0.99 Ga 0.01 O 3-d sample reached 1.01 9 10 -3 S/cm at 673 K, which is declined to 7.19 9 10 -4 S/cm for the Na 0.52 Bi 0.47 Ti 0.98 Ga 0.02-O 3-d sample at the same temperature. Through the internal friction measurement, there was an increasing trend of the oxygen ion relaxation activation energy along with the introduction of Ga 3? ions. In addition, the mobile oxygen vacancy concentration decreased with Ga 3? doping owing to the formation of local defect clusters, which is the possible reason that the bulk conductivity of Na 0.52 Bi 0.47 Ti 1-x Ga x-O 3-d (x = 0, 0.01, 0.015, 0.02) samples reduced with the Ga 3? doping concentration.