First-principles study of interface doping in ferroelectric junctions

Effect of atomic monolayer insertion on the performance of ferroelectric tunneling junction is investigated in SrRuO3/BaTiO3/SrRuO3 heterostrucutures. Based on first-principles calculations, the atomic displacement, orbital occupancy, and ferroelectric polarization are studied. It is found that the ferroelectricity is enhanced when a (AlO2)− monolayer is inserted between the electrode SRO and the barrier BTO, where the relatively high mobility of doped holes effectively screen ferroelectric polarization. On the other hand, for the case of (LaO)+ inserted layer, the doped electrons resides at the both sides of middle ferroelectric barrier, making the ferroelectricity unfavorable. Our findings provide an alternative avenue to improve the performance of ferroelectric tunneling junctions.

Ferroelectric (FE) materials have attracted significant interests due to their technological application in electronic devices, such as field-effect transistors (FET) and nonvolatile random access memories [1][2][3] . The inherent spontaneous electric polarization can be switched between two (or more) stable polarization states and thus can be used to modulate the screening charge at the interface [4][5][6] , or can be used as a memory state variable. Furthermore, due to the existence of ferroelectricity in nanometer scale which has been demonstrated in experiments and theory [7][8][9][10][11][12][13] . FE heterostructures, such as ferroelectric tunneling junction (FTJ), have become a very promising candidate for application in FTJ-based nanoscale transducers and future non-volatile memories with high storage density, high speed, and low power consumption 14,15 . FTJ is a FE film sandwiched between two metallic electrodes, the surface charges in the ferroelectric are not completely screened by the adjacent metals and the depolarization field in the barrier is not zero 16 . In general, the interface inevitably exists between the metal and the FE barrier in FTJs, and it will bring great influence on the ferroelectricity of the barrier and the transportation property of FTJs 17 . The formation of intrinsic dipole moments at the interface has been confirmed. For three types of heterostructures, i.e., vacuum/LaO/BTO/LaO, LaO/BTO, and SRO/LaO/BTO/LaO, it was found that the polar interfaces create an intrinsic electric field which is screened by electron charges leaking into the BTO barrier 18 . This made a FE dead layer near the interface, which is nonswitchable and thus is detrimental to ferroelectricity. Different terminal atomic structure of the FE barrier will also influence the performance of FTJs. It was proved that the Pt/BTO/Pt junction with TiO 2 -terminated layer is more conductive than the BaO-terminated one 19 . In addition, it was found that due to an build-in interface dipole, BaO/RuO 2 interface in the SrRuO 3 /BaTiO 3 /SrRuO 3 (SRO/BTO/SRO) junction is unfavorable to the switchable FE polarization. Replacing one or two unit cells of BaTiO 3 with SrTiO 3 at this interface will alleviate this effect 20,21 . Therefore, interface engineering is a practical way to improve the performance of FE nanodevices.
Due to the atomic-layer control of the growth and atomic-scale measurement of composition and electronic structure at buried interfaces are possible, the atomic layer insertion becomes one of the effective interface engineering in multiferroic tunneling junction and FE heterostructures. It was proposed that when a Ni monolayer inserted at one interface in the epitaxial Fe/PbTiO 3 /Fe junction, large robust ME effects and good tunneling performances (TER and TMR) are obtained 22 . In the meantime, it was demonstrated that the insertion of the conducting layer LaNiO 3 between the Bi 6 FeCoTi 3 O 18 epitaxial film and the substrate is a powerful method in achieving high quality layered oxide thin films 23 . Also it was found that the existence of an additional FeO monolayer in the interface of Fe/BaTiO 3 /Fe multiferroelectric junction could lead to the vanishing critical thickness for ferroelectricity and the enhancement of ME coupling 24 .
LaAlO 3 (LAO) is a polar perovskite oxide which consists of the alternative stacked positively charged (LaO) + layer and negatively charged (AlO 2 ) − layer. As such, LAO can directly support electron (LaO termination) or hole (AlO 2 termination) doping at the interface when it is deposited on non-polar oxide via electronic reconstruction 25 . This motivates us to explore the effects of polar interface on the ferroelectricity of barrier in FTJs. In this report, using the typical SRO/BTO/SRO junction as a prototype, the (AlO 2 ) − monolayer and (LaO) + monolayer Scientific RepoRts | 6:24209 | DOI: 10.1038/srep24209 are inserted between the SRO electrode and BTO barrier. The change of crystal structure may have some influence on the results when the magnetic degree of freedom is considered 26 . Here, we focus on the polar distortion along z direction (both from ferroelectric BTO and the charged insertion layer (LaO) + or (AlO 2 ) − in these perovskite oxides. Following the discussions of SRO/STO/LaO/STO/SRO junction and LAO/PTO heterostructures 27,28 , the variation of crystal structure away from P4mm does not affect the main results. Since our junction system is assumed to be deposited on the STO substrate, only the lattice constants in z-direction are optimized. We perform ab initial simulation on the junctions containing different types of atomic layer insertion at the interface. The FE properties and the electronic structures of the junctions are discussed, and the mechanism that causes the enhancement of ferroelectricity under the case of (AlO 2 ) − monolayer insertion is revealed.

Computation details
To explore the influence of the inserted atomic monolayer on the FE properties of the SRO/BTO/SRO junction, we construct three types of supercells with symmetric electrodes and interfaces, namely, (SrRuO 3 ) 7 -SrO-TiO 2 -(BaTiO 3 ) 8 , (SrRuO 3 ) 7 -SrO-AlO 2 -BaO-(TiO 2 -BaO) 8 -AlO 2 , and RuO 2 -(SrRuO 3 ) 6 -LaO-TiO 2 -(BaTiO 3 ) 8 -LaO, as shown in Fig. 1. One can see that the above three structures possess the same number of FE layers (8.5 L), and they correspond to the case of no interfacial insertion, (AlO 2 ) − atomic insertion, and LaO atomic insertion at the interface, respectively. In order to conveniently express the above supercells, the symbols S I , S II and S III are hereafter referred to as the FTJ structure without interface insertion, with (AlO 2 ) − atomic insertion, and with (LaO) + atomic insertion, respectively.

Results and Discussions
We start with an out of plane FE displacement obtained from the bulk BTO in the junction, then the out-of-plane lattice constant of each supercell is relaxed, together with the ionic relaxation. The relative cation-anion displacements within the electrode SRO and the barrier BTO are obtained based on the optimal structures and shown in Fig. 2(a), where the polarization of barrier points to the right. Squares correspond to the case of no atomic insertion, where the displacements are nearly symmetric with respect to the middle layer of the barrier, which is consistent with the result of ref. 20. As shown in Fig. 2a, without the inserted polar layer, screening of bound charges in such conventional FTJs arises mainly from the electronic screen from metallic electrode, and the rumpling in electrode SRO is always positive. However, when we insert (LaO) + or (AlO 2 ) − within the interface, rumpling in the electrode changes significantly. For (LaO) + in S III , the rumpling in the left electrode becomes negative. For (AlO 2 ) − in S II , the rumpling in the right electrode becomes negative. This grantees an effective screening of the inserted positive charged (LaO) + or negatively charged (AlO 2 ) − . The rumpling within the barrier is nearly symmetric with respect to middle layer in the conventional FTJ (in S I ). When (LaO) + is inserted, the rumpling in the barrier changes greatly, accompanying with the decrease of polarization and the appearance of domain wall. The local polarization of each unit cell in BTO is estimated using a model based on Born effective charge 30 as follows, where N is the number of atoms in the primitive unit cell, δ Z m is the displacement of the mth atom away its position in the symmetric structure, and Ω the volume of unit cell. Although the method based on the Born effective charges calculated for bulk BTO cannot provide a quantitative accurate description of the local polarization in heterostructures. Nevertheless, it can give an estimation on the local polarization of FE barrier in tunneling junctions 31 . From Fig. 2(b), one can see that the polarization of the middle layer of the barrier is increased when the atomic monolayer is inserted, and the polarization of the middle region of the junction S I approaches the bulk polarization of BTO. Due to the appearance of a FE domain wall near the right interface in structure S III , this leads to the decrease of the average polarization of the barrier. After a simple summation on the local polarizations, the average barrier's polarizations are 0.26 C/m 2 , 0.28 C/m 2 and 0.25 C/m 2 for structures S I , S II and S III , respectively. Therefore, the interfacial (AlO 2 ) − inserted-layer in structure S II will raise the average polarization, while the (LaO) + inserted-layer in structure S III will lower the average polarization. To give a deeper understanding on the variation of the barrier's polarization with the introduce of an atomic insertion at the interface, the electronic structure and charge transfer across the interface are given and discussed in latter paragraphs.
The layer-resolved density of states (LDOS) for FTJs with different interface configurations are plotted in Fig. 3. In FE films with either interface structures and for the polarization pointing to the right, the alignment of conduction-band minimum (CBM) and the valence-band maximum (VBM) across each layer of the barrier become curving. The bended alignment of CBM (or VBM) is triggered by the occurrence of depolarization field within the barrier, and the slope is proportional to the magnitude of internal depolarization field. Compared with the LDOS of S I in Fig. 3(a), the Fermi level shifts to the top of valence band for the FTJ with (AlO 2 ) − interface insertion in Fig. 3(b), while in Fig. 3(c) for the case of (LaO) + insertion, the Fermi level will shift to the bottom of conduction band. And the interfacial layer on the side of BTO becomes more conductive with respect to the case without atomic insertion, it indicates that there is a net charge transfer across the interface. As the (AlO 2 ) − monolayer is inserted in the interface, one can easily find that the electrons transfer from the barrier to electrode in Fig. 3(b), and this leads to the hole-doping at the interfacial layers of BTO. However, for the case of (LaO) + insertion, the electron doping at the interfacial layers of BTO occurs, which results in the electrons shifting from electrode to the barrier. The projected density of states (PDOS) of the interfacial Ba, Ti, and O atoms are shown in Fig. 4. To see the occupation state of the transferred charges perceptibly, the projected DOS of left interfacial atoms and the right interfacial atoms within barrier BTO are shown for (AlO 2 ) − and (LaO) + interfacial insertion, respectively. Bear in mind that the spontaneous polarization is chosen and fixed pointing to the right, unless otherwise specified. From Fig. 4(a), it is found that the holes mainly occupy at O 2p orbitals both at BaO layer and TiO 2 layer. On the counterpart of (LaO) + insertion, the transferred electrons almost site on Ti 3d orbitals, as shown in Fig. 4(b). Up to here, we obtain that the interfacial hole doping raises the barrier's polarization, while the interfacial electron doping will suppress the polarization of barrier of FE heterostructures. However, it is well known thing will be different for the bulk BTO 32 , where the uniform carrier doping is always against the stability of ferroelectricity.
The distributions of transferred charges within the barriers for the case of interfacial (AlO 2 ) − insertion and (LaO) + insertion are given in Fig. 5(a,b), respectively, where the number of charge doping in the leftmost layer is reduced to unit, and the amount of charge doping in other layer is a relative value with reference to that in the leftmost layer. As has also been illustrated in Fig. 4, the electrons are extracted from the barrier to electrode, and the holes are almost site on oxygen atoms for the case of (AlO 2 ) − insertion. From Fig. 5(a), one can see that the holes are asymmetrically distributed on the left side of barrier, which is resulted by the mediation of the spontaneous polarization. If the barrier stays at paraelectric state and with (AlO 2 ) − insertion, then the holes will symmetrically distributed at both the left and right side due to the mirror symmetry of the system. Under the action of an intrinsic electric field, the holes are easily to hop between O p orbitals, while the electron hopping between Ti d orbitals is difficult owing to the long distance between neighboring Ti atoms compared with the distance between neighboring O atoms along the direction of field. As reported in ref. 25, the hopping matrix elements between neighboring O p orbitals show no significant discontinuity. Then the theoretical high mobility of hole in the FTJ and electrons' transferring to the electrode will help to screen the bound charges of the barrier, and further enhance the barrier's polarization. In experiment, holes will easily be trapped in oxygen vacancies in the heterointerface which will reduce their mobility 33 . Therefore, the discrepancy between experiment and theory needs to be intensive studied by additional experiments and theories 25 . From Fig. 5(b), it is found that the electron inhabits the Ti d orbitals and there is no doping charge in the BaO layer. The depth of carrier penetrating into the barrier is large as compared with that in case of (AlO 2 ) − insertion in Fig. 5(a). Furthermore, the concentration of electron doping in Fig. 5(b) shows abnormality between the 2nd and 3rd TiO 2 layer from the right, which is caused by the occurrence of FE domain wall within the barrier, and is consistent with the distribution of local polarization in Fig. 2(b). Consequently, the average barrier's polarization will be suppressed in the FTJ with interfacial (LaO) + insertion.
In summary, the atomic monolayer insertion at the interface of a typical SRO/BTO/SRO FTJ was investigated by the use of the first-principles calculations. The local polarizations are calculated based on the Born effective charge method, and it is found that the interfacial (AlO 2 ) − insertion is in favor of the enhancement of FE polarization. Through the analysis of electronic structures and the carrier doping within the barrier, the increase of barrier's polarization for (AlO 2 ) − insertion is attributable to the hole doping near the interfaces, and the comparatively short penetrating length of the doped holes. Theoretically, the electron doping and hole doping have the opposite effect on the polarization of FE heterostructures. This is different from the doping effects in the bulk, where the carrier doping is always detrimental to the ferroelectricity. Therefore, the atomic monolayer insertion at the interface may be an effective way to enhance the polarization in FE heterostructures, and then improve the performance of FTJ-based nano-transducer. Further experimental studies should be carried out on the mechanism of the transportation of charges within the barrier, and a more practical FTJ with asymmetric electrodes or with asymmetric interfacial structures should be adopted for achieving a large TER.