1 Introduction

Sodium bismuth titanate (Na0.5Bi0.5TiO3, abbreviated as NBT) belongs to the perovskite structure with a general formula ABO3, where Bi and Na occupy the A site. It shows a relaxor behavior due to the cation disorder. The NBT is known to exhibit strong ferroelectric property with large polarization (P r  = 38 μC/cm2) [1]. During last decade, the related performances including piezoelectric, pyroelectric, high leakage and so on, have been studied widely [25]. For the case of thin film form, it is suitable for integration on substrates used in the microelectronic industry. Therefore, many researches about NBT-based films have performed on improving the electrical performances [68], and demonstrated that site engineering is an available way for improving the electrical properties [9, 10]. Also, our recent work suggests that the 2 at.% Fe substitution is effective not only in reducing the leakage current but also in improving the ferroelectric properties of NBT thin film [11].

Previous researches have demonstrated that the microstructure and electrical properties of the ferroelectric thin films tend to be strongly influenced by the used substrates. PZT thin films are epitaxial, highly oriented and polycrystalline by using various substrates such as SrTiO3(100), MgO(100) and γ-plane sapphire [12]. BiFeO3 thin film on SrTiO3 has a rhombohedral-like structure, while that on LaAlO3 has a pure tetragonal-like phase [13]. Compared with the Pt/Ti/SiO2/Si substrate, the (Pb,Ba,Sr)TiO3 film shows higher dielectric permittivity and tenability on LaNiO3/LaAlO3 [14]. BiFeO3 films exhibits better ferroelectric properties on sapphire substrates than that on Pt/Ti/SiO2/Si due to the improved interface of ferroelectric/bottom electrode [15]. Furthermore, when BiFeO3 thin films were grown on different substrates, the differences in domain structure have been found tested by the piezoelectric force microscope [16]. For NBT-based films, various electrode compositions and single crystals have been chosen, such as Pt coated-(110)MgO [17], LaNiO3(100)/Si [18] and SrTiO3/Pt/Ti/SiO2/Si [19]. However, the study of NBT-based thin films simultaneously deposited on different substrates at the same preparing conditions has rarely been reported.

In this work, the Na0.5Bi0.5(Ti0.98Fe0.02)O3 (NBTFe) thin film was selected as the research object. The NBTFe film has been prepared on the monocrystalline silicon, Al-doped ZnO (AZO)/glass and Pt/TiO2/SiO2/Si at the same preparing condition. We aim at investigating the structural, ferroelectric, and dielectric properties of NBTFe thin films on different substrates.

2 Experimental methods

The NBTFe films were prepared by metal organic decomposition together with rapid thermal annealing. An appropriate amount of bismuth acetate [Bi(CH3COO)3], sodium acetate (CH3COONa) and iron nitrate [Fe(NO3)3·9H2O] were dissolved in glacial acetic acid at 60 °C. Tetrabutyl titanate [Ti(OC4H9)4] was reacted with the chelating agent of acetylacetone (CH3COCH2COCH3). Then, two solutions were mixed and diluted with 2-methoxyethanol to adjust the viscosity at room temperature. Dusts and other suspended impurities were removed by filtering through a 0.2 μm syringe filter. Finally, the transparent and stable yellow precursor solution was obtained. For the case of the NBTFe thin film on Si (100), the precursor solution was deposited on substrate by spin-coating at 3000 rpm for 25 s. The wet film was first pyrolyzed on a hot plate at 300 °C for 15 min to remove organic solvent. Subsequently, it was heated to 400 °C at a rate of 15 °C/s for 20 s and then ramped up at 10 °C/s to the crystallized temperatures ranging from 500 to 600 °C for 10 min by rapid thermal processor. This process was repeated many times to obtain a desired thickness. For NBTFe on AZO/glass and Pt/TiO2/SiO2/Si substrates, the preparing procedures are similar to that on Si except that the final annealing temperature of 550 °C is selected. The thickness of each NBTFe film was estimated to be all about 400 nm by a step (step obtained by chemical etching) profilometer made by the Ambios Technology Company of USA.

The crystallization was studied by X-ray diffraction using a Rigaku D/MAX-γA X-ray diffractometer. For the measurements of electrical properties, several Au dots were evaporated through a mask on the films, and an Au layer was deposited on the backside of Si substrate to form metal–ferroelectric–metal (MFM) and metal–ferroelectric–semiconductor (MFS) configurations. The schematic diagrams are shown in Fig. 1. The insulating leakage was conducted using a Keithley 4200 semiconductor characterization system. The ferroelectric loop was conducted using a standard ferroelectric tester (Precision Pro. Radiant Tech., Albuquerque, NM). The dielectric property was measured by an impedance analyzer (HP4294A).

Fig. 1
figure 1

Schematic structures of a MFS and b MFM with NBTFe thin film as a ferroelectric layer

3 Results and discussion

As reported earlier, the impurity phase of pyrochlore can usually be observed for the NBT-based film, especially for the monocrystalline silicon substrate [19, 20]. So, we firstly deposited the precursor solution on the Si substrate and changed the annealing temperatures ranging from 500 to 600 °C, as shown in Fig. 2a. Fortunately, although magnitude of the lattice mismatch between the film and substrate is large to 39 % [19], it can be seen that all the prepared films have the phase-pure rhombohedral perovskite structures without any secondary phases. This better crystallization may be ascribed to the following two factors. (1) The high uniformity of our precursor solution favors the nucleation of perovskite NBTFe grains. (2) For the sequential layer annealing process in rapid thermal annealing furnace adopted in this work, each layer can serve as a seeding layer for the next layer [21]. For the XRD patterns of the NBTFe films, mainly (100), (110) and (200) peaks are observed. Significantly, with the increase of the annealing temperature, the (l00) diffraction peaks intensity increases gradually, whereas the (110) peaks intensity decreases slightly. The ratio of the (100) and (110) diffraction peaks intensity [I(100)/I(110), I(hkl) denoting the intensity of (hkl) peak] reaches the highest (~5) in NBTFe(Ta=600 °C) film. We ascribe this to the fact that the competitive growth between the (l00)- and (110)-oriented grains inhibits the growth of (110)-oriented grain because the minimization of surface energy can favor the (100)-oriented grain growth in the high temperature. Actually, such a competitive growth phenomenon or similar structure evolution can be frequently observed in CaBi4Ti4O15 and BiFeO3 films [22, 23].

Fig. 2
figure 2

XRD patterns of NBTFe thin films on different substrates: a monocrystalline silicon and annealed at various temperatures, b AZO/glass and Pt/TiO2/SiO2/Si

During integrating the ferroelectric thin film with the silicon semiconductor substrate, the processing temperature should not to be high in order to prevent the silicon semiconductor to be seriously damaged [24]. Also considering the crystallinity of the NBTFe thin film, we choose the 550 °C as the optimum annealing temperature and fabricate the NBTFe thin film on other two substrates. The XRD patterns of the NBTFe films deposited on Pt/TiO2/SiO2/Si and AZO/glass is shown in Fig. 2b. The thin films on both substrates are fully crystallized with high intensity of (110) peak. This similar (110) preferred orientation indicates that the formation of (110) orientation in NBT-based films should be an intrinsic growth mode. It may be related to the growth rate anisotropy of the NBT films along different orientations. At 550 °C, the degrees of (110) orientation for NBTFe on Si, Pt/TiO2/SiO2/Si and AZO/glass are 31, 36 and 57 %, respectively calculated by the Lotgering factor [25]. This strongest orientation degree of the (110) diffraction peak on AZO/glass indicates the high amount of (110)-oriented grain, which may be due to the fact that the oxide electrode of AZO can promote the nucleation and grain growth of the NBTFe films.

The influence of different substrates on the insulating properties of NBTFe films were measured and shown in Fig. 3. The leakage current density of each NBTFe film increases gradually with the increase of electric field. This can be ascribed to the release of (Fe Ti )′ by breaking the defect complexes in higher electric field regions. Additionally, the NBTFe thin film deposited on Pt/TiO2/SiO2/Si substrate has the lower leakage current density than the other two films. Since higher leakage currents can be frequently observed in epitaxial films [26], the higher leakage current in NBTFe on AZO/glass may be ascribed to the high (110)-orientation. For all samples, the current curves are asymmetry under the positive and negative electric fields. We consider that the phenomenon is mainly due to the asymmetric configuration of electrodes.

Fig. 3
figure 3

Leakage current density as a function of electric field for NBTFe thin films on Si, AZO/glass and Pt/TiO2/SiO2/Si substrates

The capacitance–voltage (CV) characteristic of Au/NBTFe/Si capacitor was measured at a frequency of 100 kHz with a small ac signal of 300 mV amplitude. Figure 4 shows the relationship between CV curve and sweeping voltage. The gate voltage was swept from accumulation to inversion region and then swept back. During the voltage sweeping, VFBf and VFBr are defined as the flatband voltage (VFB) from the negative to the positive bias and the VFB from the positive to the negative bias. The width of the memory window (Vm) is defined as the difference of the flatband voltage shift, namely VFBf–VFBr [27]. It can be found that each CV hysteresis loop is clockwise, which is referred to as polarization-type switching mainly controlled by the ferroelectric polarization. When the voltages sweep from ±4 to ±6 V, the hysteresis loops increase due to the increased ferroelectricity. The maximum memory window is 2 V obtained at ±6 V. However, with the voltages increasing from ±6 to ±10 V, the gradually enhanced positive charges injection occurs, the traps in the film will be filled up and then the space charges will be appear [28]. This results in the shift of the VFBf and VFBr toward the negative voltage and the significant decrease of the memory window.

Fig. 4
figure 4

The CV characteristics of Au/NBTFe/Si capacitor under different sweep voltages

Figure 5 shows the typical polarization–electric field (PE) hysteresis loops of the NBTFe films deposited on the AZO/glass and Pt/TiO2/SiO2/Si substrates, respectively. It can be seen that the NBTFe film deposited on AZO/glass exhibits a poor PE loop reflected by the round shape, which is clearly due to the large contribution from high leakage current, as displayed in Fig. 5a. In contrast, well defined PE hysteresis loops can be obtained in the NBTFe film deposited on Pt/TiO2/SiO2/Si substrate. With the increase of the electric field, the remanent polarization (P r ) increases monotonically. This may be ascribed to more ferroelectric domains can be switched with the increase of applied electric field. It is worth noting that at 600 kV/cm, the value of P r for NBTFe film deposited on Pt/TiO2/SiO2/Si substrate is about 18 μC/cm2, which is much larger than that of Nb5+-doped (Na0.85K0.15)0.5Bi0.5TiO3 thin film on the same substrate (P r  ~ 7 μC/cm2) [29] and NBT/Bi1.07Nd0.03FeO3/NBT sandwiched structure on LaNiO3/Si (P r  ~ 10 μC/cm2) [30].

Fig. 5
figure 5

PE hysteresis loops of NBTFe films deposited on a AZO/glass and b Pt/TiO2/SiO2/Si substrates

Also, the PE loops shift towards the negative electric field obviously, suggesting an internal field occurs in the film. For Fe-doped NBT films, when Ti4+ is substituted by low valence of Fe3+, defect of (Fe Ti )′ can be formed simultaneously. Then, defect dipole (Fe Ti )′ − (V O )·· is constituted between (Fe Ti )′ and (V O )··, where using Kröger and Vink notation [31]. Therefore, a local field (E loc ) is established between the defect dipoles in the film. Besides, the asymmetric top (Au) and bottom (Pt) electrodes should not be neglected because they have different work functions and electric affinities.

Figure 6 shows the variation of the relative dielectric constant (ε r ) and loss tangent (tan δ) with the frequency in the range of 1 kHz to 1 MHz. The NBTFe films are measured in the MFM under a dc bias voltage of 0 V. However, the bias voltage for NBTFe film in the MFIS capacitor is −6 V, which can ensure the capacitance lying in the accumulation state. At this bias voltage, the accumulation layer present in the Si semiconductor substrate can act as a near metallic electrode. Therefore, the capacitance in the accumulation regime should be due solely to the ferroelectric film. One can note that the ε r decreases and tan δ increase gradually with the increasing frequency. Among all samples, the NBTFe thin film deposited on Pt/TiO2/SiO2/Si shows the highest ε r of 313 and the lowest tan δ of 0.1 at 100 kHz. Obviously, the ε r of NBTFe film on Si is much small in the whole range of the measuring frequencies. This may be due to the fact that an interfacial layer of SiO2, which is in series with the NBTFe is formed unavoidably during the process of preparing the ferroelectric film.

Fig. 6
figure 6

Frequency dependence of the dielectric constant and dissipation factor for NBTFe deposited on different substrates

4 Conclusion

In this study, we investigated the structural, leakage current, ferroelectric and dielectric properties of the NBTFe films deposited on three kinds of substrates. The results show that the film on Si annealed at 600 °C has a distinctly preferential (100) orientation, while the films on AZO/glass and Pt/TiO2/SiO2/Si crystallized at 550 °C show high intensity of (110) peaks. A maximum memory window of 2 V in capacitance–voltage curve has been obtained for NBTFe film on Si. The NBTFe on Pt/TiO2/SiO2/Si shows a well defined PE hysteresis loop with a large P r value of 18 μC/cm2. The ε r and tan δ of NBTFe thin film on Pt/TiO2/SiO2/Si substrate are about 313 and 0.1 at 100 kHz, respectively. These findings suggest that the structure and properties of NBT-based films should depend strongly on substrate, which is reflected in our experimental observations.