Oxygen vacancy region formation in BaTiO3 adjacent to the interface between the internal electrode and the dielectric layer in Ni–Sn internal electrode multilayer ceramic capacitor exhibiting leakage current suppression

This study underlines the ceramic BaTiO3 dielectric layer adjacent to the electrode of long-term reliability-improved Ni–Sn alloy internal electrode BaTiO3-based multilayer ceramic capacitor to clarify the cause of electric barrier formation. Electron energy loss spectroscopy measurements of the Ti L 3,2 near the edges and the O K near the edge structure changes to characterize the existence of an oxygen vacancy region, approximately 60 nm in width, and generated in BaTiO3 adjacent to the interface. Accordingly, the n-type semiconductor layer of BaTiO3 that originated from the oxygen vacancies, led to the formation of a rigid Schottky barrier at the interface.

N i base-metal-internal-electrode BaTiO 3 -based multilayer ceramic capacitors (MLCCs) are indispensable components in the manufacture of a various modern electronic devices, such as consumer and industrial electronics. The computerization, electronification, and automation of automobiles have been accelerated by the introduction of advanced driver assistant systems 1) and self-driving. Thus, ensuring the total reliability of the systems is crucial, considering the safety of such fast applications, including the long-term reliability of MLCCs. However, MLCCs with a few hundred layers and dielectric thickness in the order of a few micrometers have already been commercialized. 2) The thinner dielectric layer inevitably generates a stronger electric field in the layer under same rated voltage; that is, the electric field strength doubled with half the thickness of the dielectric layer under the same rated voltage. Generally, an increased electric field strength increases the leakage current as the MLCC degrades. A highly accelerated life test (HALT) is employed to evaluate the life time of an MLCC in a practical time period. [3][4][5] Presently, almost all commercialized MLCCs use Ni metal as the internal electrode, instead of noble metals, such as Pd, Ag, and Pt. 6) Instead of Ni, a Cu internal electrode MLCC was successfully fabricated. Its electrical properties were then compared with those of a normal Ni internal electrode MLCC. Based on the I-V characteristics, the leakage current density of the Cu internal electrode MLCC was suppressed compared with that of the Ni internal electrode MLCC at a high DC electric field. 7) The improvement in leakage current characteristics can be attributed to the double Schottky barrier enhancement at the interface of the dielectric layer/internal electrode. 8,9) Another approach to inducing a material effect on the internal electrode to improve the long-term reliability involves alloying the Ni internal electrode. 10) The suppression of leakage current and improvement of long-term reliability of Ni-Sn MLCC was derived from the change in the Schottky barrier height owing to alloying, as shown in the barrier height change in Ni-BaTiO 3 and NiO-BaTiO 3 interfaces. 11) Ni and Ni-Sn internal electrodes were alternately stacked in the fabricated MLCC samples to clarify and characterize the contribution of the electrical barrier at the electrode and dielectrics interface. HALT was performed on the MLCC with the Ni-Sn and Ni electrodes as the anode and cathode, respectively, compared with the Ni MLCC and Ni-Sn MLCC. 12) Based on the results, the MLCC with the Ni-Sn internal electrode at the anode drastically suppressed the leakage current degradation for a longer time compared with the MLCC with the Ni-Sn internal electrode at the cathode. The leakage current characteristic of the MLCC with the Ni-Sn internal electrode at the cathode was comparable to the leakage current characteristics of the entire Ni MLCC, whereas the leakage current characteristic of the MLCC with the Ni-Sn internal electrode at the anode was comparable to Ni-Sn MLCC leakage current characteristics. The Ni-Sn internal electrode at the anode case completely coincided with the Schottky barrier under a reverse bias voltage, whereas the Ni-Sn internal electrode at the cathode case coincided with Schottky barrier under a forward bias voltage. Furthermore, energy dispersive X-ray spectrometry (EDS) chemical mapping of the Ni-Sn alloy electrode obtained using a Scanning Transmission Electron Microscope (STEM) indicated that approximately 2 nm (width) Sn segregation in the Ni-Sn electrode adjacent to the electrode and dielectric layer interface. Sn supposedly trapped oxygen from the dielectric layer BaTiO 3 by oxidation, resulting in its accumulation at the interface. 12) In addition, the oxidation of accumulated Sn inferred oxygen diffusion to Sn from the contacted region of the dielectric layer BaTiO 3 associated with the generation of oxygen vacancies. In other words, the contact between the electrode and the semiconductor BaTiO 3 is resulted from the generation of oxygen vacancies. Thus, a metal-semiconductor or metal-insulator-semiconductor structure with an electrical barrier can be considered.
To identify the oxygen vacancies in the BaTiO 3 -based MLCC, Ti L 3,2 electron energy-loss near-edge structures (ELNES) in electron energy loss spectroscopy (EELS) have been utilized for the analysis. 13,14) Non-defected BaTiO 3 and related perovskite-type compounds indicate the splitting of each edge, L 3 and L 2 , into two main peaks in the Ti L 3,2 ELNES. The density of unoccupied states of the Ti atoms was sensitive to short-range coordination and the four well split peaks in the Ti L 3,2 edge could be attributed to excitations of 2p 3/2 and 2p 1/2 subshells to unoccupied states, t 2g and e g . 15) Following the introduction of oxygen vacancies, the splitting between the two peaks at the Ti L 3 edge and the Ti L 2 edge is noted as a disappearance. These spectral changes have been interpreted as the valency change of Ti 4+ to Ti 3+ following the introduction of oxygen vacancies based on comparisons with the experimental spectra of reference compounds. 16,17) Based on ab initio calculation, it contributed the less appearance of peak splitting; thus, an oxygen vacancy concentration can be detected at 1%. 18) Additionally, the O K ELNES of perovskite materials indicate the existence of oxygen vacancies based on the shape of the EELS edge, which reflects the underlying electronic structure. 16) The O K edge fine structure is naturally sensitive to O-O ordering, and an increase in the oxygen vacancy concentration results in a decrease in the intelligibility of the peaks, similar to Ti L 3,2 ELNES. 19) In the case of SrTiO 3 , 4% oxygen vacancies could be detected from the spectrum shape change as the smallest value. 20) The cluster simulations of the grain boundaries of SrTiO 3 21) and the carrier density measurement by Hall effect assuming two free carriers for each oxygen vacancy, 22) support the detection limit of oxygen vacancy concentration. In the case of previously investigated quantitative oxygen vacancy analysis using Ti L 3,2 ELNES and O K ELNES, a few percent of the vacancies could be detectable using EELS measurements. Spatially resolved EELS (SR-EELS) 23,24) can be applied to cross-sectional TEM specimen to obtain a depth-resolved EELS. 25,26) In this study, after validating the improvement of longterm reliability of an an Ni-Sn MLCC, we characterize the depth profile of Ti L edges and O K edge of the BaTiO 3 dielectric layer adjacent to the interface of the electrode and the dielectric layer to clarify the origin of electric barrier formation at the interface.
First, we prepared 2012 case sized (1.2 × 1.2 × 2.0 mm) Ni MLCC and Ni + Sn MLCC with Electronic Industries Alliance standard X5R characteristics (capacitance changing rate <15% from −55°C to +85°C) and rated voltage of 10 V. The dielectric layer number was 273 with 1.5 μm in thickness. The fabrication process is as follows. BaTiO 3 -based perovskite powder was ball-milled with additives, such as Dy 2 O 3 , Y 2 O 3 , MgO, MnO 2 , and SiO 2 . Once the powders were uniformly mixed and dried, the raw material powder was milled with an organic solvent to obtain a raw material slurry. The slurry was cast on plastic film and a ceramic green sheet was obtained after drying. Ni paste was printed on the green sheet as the internal electrodes. SnO 2 of 1 weight % was added into Ni paste and homogeneously mixed by a 3-roler mixer. Similarly, SnO 2 added into the Ni paste was printed onto the green sheet. The Ni printed sheets were stacked and cut as a MLCC green chip of Ni MLCC, and the Ni + SnO 2 printed sheets were stacked and cut as a green chip of Ni-Sn MLCC. After an organic binder burn out at 300°C, they were fired at 1200°C for 2 h at an oxygen partial pressure of 1 × 10 −11 Pa, and then re-oxidized at low temperature. Finally, the Cu terminal electrode was formed by baking at 820°C for 30 min. In the reducing atmosphere sintering condition, SnO 2 changed into Sn metal and form an Ni-Sn alloy. Subsequently, the leakage current degradation of both MLCCs were evaluated at 140°C and 40 V.
As fabricated specimens were investigated using STEM (Hitachi HF3300) fitted energy filter (Gatan Tridium 863) for EELS measurement and silicon drift EDS detector (Bruker XFlash 5060 T) with 300 kV operation voltage. SR-EELS images were obtained using a SR-EELS slit, from which the ELNES were extracted from selected positions.
The leakage current degradation time dependence of both MLCCs are shown in Fig. 1. Leakage current degradation of Ni MLCC started from the onset; in contrast, notably Ni-Sn MLCC effectively suppressed the leakage current, as previously reported. 10,12) Cross-sectional TEM bright field images of Ni MLCC and Ni-Sn MLCC are shown in Figs. 2(a) and 2(b). Their differences are insignificant, showing a single-grain internal electrode layer with a thickness of approximately 1 μm sandwiched by dielectric layers comprising less than 0.2 μm in the diameter of BaTiO 3 grains. However, their EDS Sn chemical mapping differed considerably. Sn could not be detected in the inner electrode of Ni MLCC but in Ni-Sn MLCC. Therefore, Sn solid solution into Ni electrode occurred in Ni-Sn MLCC [ Fig. 2(d)]. Furthermore, as previously reported, Sn segregation at the electrode and the  dielectric layer interface using STEM 12) was validated in high magnification high-angle annular dark field (HAADF) STEM images and EDS Sn mapping [ Figs. 3(a) and 3(b)]. A bright contrast region could be observed in the electrode interface in the HAADF STEM image [ Fig. 3(a)]. This can be attributed to the higher atomic number of Sn compared with Ni. Corresponding to the bright contrast, a higher concentration of Sn can be clearly observed in the Sn EDS chemical mapping. Considering the tilted interface it could be less than 10 nm in width [ Fig. 3(b)].
The Ti L 3,2 edges ELNES depth profiles of Ni MLCC and Ni-Sn MLCC are shown in Fig. 4. An SR-EELS slit was placed at the interface. The TEM image through SR-EELS slit and the SR-EELS image of Ni MLCC and Ni-Sn MLCC are shown in (a) and (b), respectively. The clear splitting between t 2g and e g peaks at the Ti L 3,2 edges can be observed from the interface to the bulk in the SR-EELS image of Ni MLCC [ Fig. 4(a)]. Meanwhile, the clear splitting recognized at the bulk disappeared when approaching the interface in the SR-EELS image of Ni-Sn MLCC [ Fig. 4(b)]. ELNES extracted from the colored boxes in Figs. 4(a) and 4(b) are shown in Figs. 4(c) and 4(d). Corresponding to the SR-EELS image, the ELNES extracted from box number 1 to 5 showed clear peak splitting of t 2g and e g peaks at the Ti L 3,2 edges for the same figure in Fig. 4(c). This might be attributed to the absence of detectable oxygen vacancies in the BaTiO 3 layer adjacent to the interface in Ni-MLCC. In contrast, in Fig. 4(d), the ELNES extracted from box number 1 and 2 show peak splitting of t 2g and e g peaks at the Ti L 3,2 edges; the splitting disappeared in ELNES from box numbers 3 and 4 and completely disappeared at the ELNES from box number 5, which is nearest to the interface. The disappearance depth length from the interface was approximately 60 nm. Oxygen vacancies were generated in this region.
The O K edge ELNES depth profile of Ni MLCC and Ni-Sn MLCC are shown in Fig. 5. The O K edge ELNES from the interface to the bulk of Ni-MLCC is shown in Fig. 5(c). The differences in the ELNES figures are insignificant, the same as the the Ti L 3,2 edges ELNES of Ni MLCC. They could be compared with the O K edge experimentally obtained from bulk BaTiO 3 . 19) The O K edge ELNES depth profile of Ni-Sn MLCC is shown in Fig. 5(d). The ELNES extracted from box numbers 1 and 2 are similar to the ELNES figures that appeared in Ni MLCC ELNES depth profile; however, the flattening of high energy peaks could be recognized at the ELNES from box number 3; it flattens more at the ELNES from box numbers 4 and 5. This coincides with the reduced SrTiO 3 O K edge ELNES. 16) Thus, the Ti L 3,2 edges and O K edge ELNES depth profiles of Ni-Sn MLCC indicate the formation of oxygen vacancies in the region, approximately 60 nm in width, and adjacent to the electrode and BaTiO 3 interface. The n-type semiconductor region resulting from the oxygen vacancies and the contact to metal electrodes, accordingly, formed a Schottky barrier at the interface. Moreover, the oxygen vacancy region suggested the depletion layer of the Schottky barrier. The formation of Schottky barrier suppressed the leakage current   and improved the life-time. The electrical structure can be interpreted as a metal-semiconductor contact form; moreover, the metal-insulator-semiconductor structure can be considered when a segregated Sn layer was continuously oxidized at the interface as an insulator. The formation of the heavily doped n-type region (n + ) in the depletion layer of the Schottky barrier results in electron tunneling and resultant ohmic contact. 27) The number of oxygen vacancies detected by SR-EELS could result from a donor that generates active oxygen vacancies and neutral defect dipoles of oxygen vacancy and doped rare Earth. 28) The considerably small total addition of Sn to Ni and the formation of thin Sn segregation layer in the order of nanometers at the interface could inhibit the formation of excess oxygen vacancies which forms to n+ region. In addition, in the case of the detectability of SR-EELS, the n-type semiconductor region would be wider than 60 nm. This phenomenon would be better understood by conducting future studies on operating real band structures and oxygen vacancy distribution.
In conclusion, Schottky barrier formation improved the reliability of Ni-Sn MLCC, which was validated by SR-EELS analysis. The previously reported forward and reverse barrier properties assumed because of the formation of a Schottky barrier were validated by the existence of an oxygen vacancy layer with a width of approximately 60 nm in the BaTiO 3 dielectric layer adjacent to the interface between the dielectric layer and Ni-Sn internal electrode.