Interface control of ferroelectricity in LaNiO3-BaTiO3 superlattices

LaNiO$_{3}$-BaTiO$_{3}$ superlattices with different types of interfaces are studied from first-principles density-functional theory. It is revealed that the ferroelectricity in the superlattice with (NiO$_2$)$^-$/(BaO)$^0$ interfaces is enhanced from that of the superlattice with (LaO)$^+$/(TiO$_2$)$^0$ interfaces. The origin lies at the polar discontinuity at the interface, which makes the holes localized within the (NiO$_2$)$^-$/(BaO)$^0$ interface, but drives a penetration of electrons into BaTiO$_3$ component near (LaO)$^+$/(TiO$_2$)$^0$ interface. Our calculations demonstrate an effective avenue to the robust ferroelectricity in BaTiO$_3$ ultrathin films.


I. INTRODUCTION
Ferroelectric ultrathin films and heterostructure have attracted great interest in recent years 1−14 . The ferroelectricity, in particular, the critical thickness, was found depend strongly on the proper choice of the interface. By controlling the chemical bonding at the interface, both the enhanced and decreased ferroelectricity have been found in BaTiO 3 and PbTiO 3 ultrathin films. In fact, polar discontinuity at the heterointerface widely exists at oxide based heterostructures 15−20 . Generally, it will lead to carrier doping in formerly insulating component and consequently emergent electronic phase, e.g., conducting twodimensional electron gas in LaAlO 3 -SrTiO 3 heterointerface. For ferroelectric materials, as studied in the BaTiO 3 bulk system 21 , such a kind of carrier doping will strongly decrease the ferroelectricity, equally for the electron doping and the hole doping. In the heterostructures or ferroelectric ultrathin films, however, the situation is yet not clear. In all the previous studies of the ferroelectric ultrathin films 1−12 , metallic electrode either of SrRuO 3 (with alternating (SrO) 0 and (RuO 2 ) 0 layers) or Pt shows no polar features.
Generally, as demonstrated in Fig. 1, the charge and potential diagrams in the metalferroelectrics superlattice can be simply described by a screening model. The charge distribution of symmetric metal-ferroelectrics interfaces with normal metal, such as Pt and Au, is illustrated in Fig. 1(a) with corresponding potential energy in Fig. 1(d). However, if a positively charged layer (e.g. the (LaO) + layer) is allowed at the interface, then, screening electrons will accumulate at adjacent ferroelectric layer (see Fig. 1(b), here spontaneous polarization is set zero for clarity), and an inhomogeneous electron doping will take place in the ferroelectric component. Taking the polarization into account, as illustrated in Fig. 1(e), the potential profile is the superposition of them. On the other hand, for a negatively charged layer, similarly asymmetric potential but with reversal sign is shown in Fig. 1(f). Such a picture suggests that the ferroelectric performance in the thin film geometry should be the same for electron doping case and hole doping one, the situation of which is similar to the bulk system 21 .
In order to check the validity of the above picture, we use LaNiO 3 as electrode with alternating charged layers of (LaO) + and (NiO 2 ) − . Perovskite oxide LaNiO 3 is structurally compatible with ferroelectric BaTiO 3 , and ferroelectric thin films with LaNiO 3 electrodes were found to possess better fatigue properties compared with other metallic oxides 22,23 .
2 As shown in Fig. 2, LaNiO 3 -BaTiO 3 superlattices are proposed. Here, two types of polar discontinuities are clearly distinguished, namely (LaO) + /(TiO 2 ) 0 with positive charge and (NiO 2 ) − /(BaO) 0 negative one, which correspond to the situation described in Fig. 1(b) and (c), respectively. In the following, we will provide a systematic first-principles study of atomic and electronic structures in these LaNiO 3 -BaTiO 3 superlattices. Our calculations provide solid evidence of contrast between electron doping and hole doping in nanoscale ferroelectrics.

II. SYSTEMS AND METHODS
Our ab initio calculations are performed using the full-potential projector-augmented wave (PAW) method 24 , as implemented in the Vienna ab initio simulation package (VASP) 25 .

III. RESULTS AND DISCUSSIONS
First, we study the inherent ferroelectric displacement and its evolution with the thickness of ferroelectric layer. In Fig. 3(a)-(f) and 3(g)-(l), the relative cation-anion displacements within both LaNiO 3 and BaTiO 3 are plotted. In BaTiO 3 , the spontaneous polarization is proportional to the relative cation-anion displacements 3 . Here, the existence of ferroelec-tricity in ferroelectric barrier is determined by the non-zero total polarization of the whole BaTiO 3 barrier 11 . Clearly, for (LaO) + /(TiO 2 ) 0 interfaces, one can see that the BaTiO 3 film is paraelectric (PE) when the thickness of BaTiO 3 is smaller than six unit cells. On the other hand, for the case of (NiO 2 ) − /(BaO) 0 interfaces, the ferroelectricity of BaTiO 3 film will be preserved down to the thickness of four unit cells. The above results differ greatly from the SrRuO 3 -BaTiO 3 -SrRuO 3 capacitor, where the critical thickness of BaTiO 3 film with SrO/TiO 2 interfaces is thinner than that of BaTiO 3 film with RuO 2 /BaO interfaces 9 . The polarity-discontinuity at metal-ferroelectric interfaces is therefore dominating. In addition, the relative cation-anion displacements for m = 8 approach the bulk value of BaTiO 3 .
To gain a deeper insight into the above differences, electronic structure of local density of state is plotted in Here, the doping is inhomogeneous among the interface and the interior of BaTiO3 film.
Things will be different for (NiO 2 ) − /(BaO) 0 polarity-discontinuity interfaces in Fig. 4 where Fermi energy shifts to the top of valence band of BaTiO 3 , and the interfacial BaO layer becomes obviously metallic, which is induced by the hole doping in the interfacial BaO layer. If the DOS of the interfacial BaTiO 3 layer is conductive, then we say the interface is metallic, i.e., the contact between the electrode and BaTiO 3 barrier is Ohmic contact 13 .
Similarly, when the DOS of the interfacial BaTiO 3 layer is insulative, then a tunneling barrier is formed at the interface, which has also been defined in Ref. 13 for SrRuO 3 /n-BaTiO 3 interface.
With the presence of polarization in Fig. 5 for the superlattice (LaNiO 3 ) 10.5 -(BaTiO 3 ) 8.5 with the ferroelectric polarization pointing to the right, an asymmetric distribution of electronic states is found, with the valence band near the left interface upward and the conduction band near the right interface downward. For the superlattice with (LaO) + /(TiO 2 ) 0 interfaces in Fig. 5(a), ferroelectric polarization controls the accumulation or depletion of electrons near the left and right interfaces, and it is found that the right interface is ohmic contact, while the left interface turns to Schottky tunnel barrier with the development of ferroelectricity in the BaTiO 3 component. The left and right interfaces are both ohmic contact for PE state as shown in Fig. 4(a). For the case of (NiO 2 ) − /(TiO 2 ) 0 interfaces in Fig. 5(b With the above analysis, the contrast between electron doping and hole doping is evident in nanoscale ferroelectrics. The electrostatic potential as demonstrated in Fig. 1(c) is therefore improper for the LaNiO 3 -BaTiO 3 superlattice with (NiO 2 ) − /(BaO) 0 interfaces, where a more symmetric situation is expected. The polar discontinuity can be neutralized via interface electronic reconstruction at oxide heterointerface, however, for the metalferroelectric heterostructure only electrons can penetrate into BaTiO 3 component. Such differences also lead to variation of ferroelectric critical thickness, which is 4 for hole-doped case and 6 for electron-doped case.
In summary, ferroelectricity in LaNiO 3 -BaTiO 3 superlattices with polar discontinuity at the interfaces are investigated from first-principles. It is found that the polar discontinuity in the (NiO 2 ) − /(TiO 2 ) 0 interface is neutralized within the interface, in contrast to the case of (LaO) + /(TiO 2 ) 0 interface, where the screening electron can penetrate into the 5 BaTiO 3 layer of four unit cells. Such differences make the ferroelectricity in the ultrathin films with (NiO 2 ) − /(BaO) 0 interface much stronger than that in the ultrathin films with (LaO) + /(TiO 2 ) 0 interfaces, therefore, providing an avenue to realize the interface control of ferroelectricity as well as functionalities in ferroelectric nanodevices.