Atomic‐Scale Study of Dead Layers in Epitaxial Perovskite Dielectric Thin Films with Oxide and Metal Top Electrodes

Perovskite‐oxide‐based capacitors, which exhibit high charge storage capacity, have attracted considerable attention as a potential candidate for overcoming the limitations of nanoscale integration. Unfortunately, a dead layer forms in these capacitors at the interface between the electrode and the dielectric, which degrades the charge storage capacity; thus, this layer has been extensively investigated. The dead layer in perovskite‐oxide‐based capacitors exhibits different characteristics depending on the electrode materials; however, a method for minimizing this layer is lacking. In this study, the charge storage capacity of a perovskite‐oxide‐based capacitor is evaluated considering the effect of the Ru and SrRuO3 top electrodes on the SrRuO3/Ba0.5Sr0.5TiO3 stack. Dead layers at the interface between each top electrode material and the dielectric are studied on the atomic scale. The results indicate that the Ru metal electrode causes oxygen to diffuse from the dielectric to the electrode, forming elongated perovskite oxide at the interface, which acts as a dead layer. However, minimizing the dead layer at the top interface increases the dielectric permittivity from 667 to 953. Consequently, the phenomenon and mechanism of the dead layer are intuitively identified. This study proposes a method to overcome the limitations of next‐generation dynamic random access memory (DRAM).


Introduction
Dynamic random access memory (DRAM) requires a high charge storage capacity to ensure small-scale integration.The charge storage capacity of a capacitor is related to its capacitance, which in turn is proportional to the dielectric constant, an inherent material property, of its dielectric layer.The overall capacitance is expressed by where C, ɛ, ɛ 0 , A, and d represent the capacitance of the capacitor, the dielectric permittivity of the dielectric layer, the dielectric permittivity of vacuum, the surface area of the capacitor, and the thickness of the dielectric layer, respectively.According to Equation (1), a dielectric layer with a high dielectric constant is crucial for developing a capacitor with a high charge storage capacity.ZrO 2 and HfO 2 , which are used in DRAM capacitors, have dielectric constants of 40 and 60, respectively.Meanwhile, the perovskite structure of SrTiO 3 shows a high dielectric constant of ≈150, and therefore, it has attracted attention as a material for next-generation DRAM capacitors.3][4] A capacitor with an electrode-dielectric-electrode stacked structure has an interface between each electrode and the dielectric.The overall charge storage capacity of the capacitor is determined by the capacitance of not only the dielectric but also that of each interface, as given by Near the interfaces, however, the dielectric material exhibits different properties from those of the bulk dielectric material because it is in contact with different materials at each interface.Specifically, these interfacial regions have lower dielectric constants than the bulk dielectric material and are thus referred to as "dead layers."Therefore, investigating the dead layer is important for enhancing the charge storage capacity of a capacitor with an electrode-dielectric-electrode stacked structure; to this end, several studies have been conducted. [5,6]erovskite dielectric materials show different dielectric properties in their dead layers based on what type of electrode material they are in contact with.Cases where the perovskite oxide is in contact with metal and oxide electrodes have been studied; however, there is no agreement regarding which electrode can effectively suppress the dead layer effect.Stengel and Spaldin [7] calculated the inverse permittivity profiles of the electrodedielectric interface of SrRuO 3 /SrTiO 3 /SrRuO 3 and Pt/SrTiO 3 /Pt capacitors using an ab initio method.The results indicated that a metal electrode (e.g., a Pt electrode, which has a short electronic screening length) is more effective for reducing the dead layer effect than an oxide electrode (e.g., SrRuO 3 ).Stengel et al. [8] performed an in-depth analysis of the interface between ferroelectric materials and the SrRuO 3 or Pt electrode by employing firstprinciples calculations.They reported that the polarization behavior of dielectric films at the interface was affected not only by the electronic screening property of the electrode but also by the chemical environment created by the metal oxide bonds of the electrode.Unlike SrRuO 3 , which has the same crystal structure as the ferroelectric film, metal electrodes such as Pt have enhanced interfacial dielectric properties because the metal forms weak atomic bonds with the ferroelectric film.Sai et al. [9] simulated the polarization at the interface between PbTiO 3 and Pt or SrRuO 3 through density functional theory (DFT) calculations.Their study focused on the charge-screening effect and the depolarization fields for different electrode materials.A larger depolarization field was formed with the SrRuO 3 electrode than with the Pt electrode because the former has a lower chargescreening effect.Therefore, less polarization was observed at the interface between SrRuO 3 and PbTiO 3 than between Pt and PbTiO 3 .
In contrast with the results of the abovementioned theoretical studies, however, many experimental studies reported that metal electrodes are more effective at minimizing the dead layer of the perovskite oxide than oxide electrodes.Plonka et al. [10] studied the difference in the dielectric property with different top electrodes (Pt or SrRuO 3 ) on an epitaxially grown (100)-oriented SrRuO 3 /BaSrTiO 3 stack.In their study, the capacitor with the SrRuO 3 top electrode showed higher dielectric permittivity than that with the Pt electrode.For each capacitor electrode, the interfacial dielectric permittivity was extracted from the y-intercept of the plot of the thickness of the dielectric layer versus the inverse of the capacitor, and the SrRuO 3 electrode showed higher values than that of the Pt electrode.Hwang [11] reported differences in the dielectric properties when Pt and IrO 2 were applied as the top electrodes of the BaSrTiO 3 dielectric layer.The capacitor with the IrO 2 electrode showed a higher interfacial capacitance density than that with the Pt electrode, indicating that the oxide electrode, which did not have the same crystal structure as the SrRuO 3 dielectric layer, enhances the interfacial dielectric permittivity.Xu et al. [12] studied the anomalous electrical properties at the interface formed by Au or SrRuO 3 with epitaxially grown SrTiO 3 on a Nb-SrTiO 3 substrate.Interfacial layers with low dielectric permittivity were formed at both Au/SrTiO 3 and Au/Nb-SrTiO 3 interfaces.The interfacial properties of perovskite oxide dielectrics as determined by the type of contact electrode are yet to be established.Therefore, an accurate understanding of the interface between the dielectric and the electrode is necessary to minimize the dead layer, thereby maximizing the charge storage capacity of the capacitor.
In this study, we studied the capacitor characteristics of epitaxially grown BaSrTiO 3 on an epitaxial SrRuO 3 bottom electrode with Ru and SrRuO 3 top electrodes.We focused on phenomena occurring in the dielectric region adjacent to each electrode.In addition, we defined the dead layer formed at the interface between the top electrode and dielectric layer through an experimental, analytical, and simulation approach.

Results and Discussion
Capacitors were fabricated with SrRuO 3 and Ru as top electrodes on epitaxially grown SrRuO 3 /Ba 0.5 Sr 0.5 TiO 3 to focus on the polarization behavior at the interface according to the type of electrode in contact with the dielectric layer.The bottom SrRuO 3 electrode and the Ba 0.5 Sr 0.5 TiO 3 dielectric layer were grown epitaxially on a film with the same crystallinity to exclude the effect of variables other than the interface between the top electrode and dielectric.The crystallinity based on the stacking state of each layer was analyzed with X-ray diffraction (XRD), as shown in Figure 1a.The capacitors with the SrRuO 3 and Ru top electrodes demonstrated different fringe patterns because of differences in the crystallinity of the top electrode; however, all XRD patterns demonstrated peaks for the epitaxially grown bottom SrRuO 3 electrode and the Ba 0.5 Sr 0.5 TiO 3 dielectric layer.The cross-sectional morphology of both capacitors was analyzed by scanning transmission electron microscopy (STEM) and fast Fourier transform (FFT) methods, as shown in Figure 1b,c.Similar to the XRD results in Figure 1a, both capacitors exhibited the perovskite structure for the epitaxially grown SrRuO 3 bottom electrode and the Ba 0.5 Sr 0.5 TiO 3 dielectric layer.Meanwhile, the SrRuO 3 and Ru top electrodes were grown as a single crystal and polycrystal, respectively.These results confirm that there are no other structural factors affecting the capacitor characteristics besides the interface between the top electrode and the dielectric layer.
An atomic-scale analysis was performed on the top interfaces of the capacitors having Ru or SrRuO 3 as the top electrode to understand the polarization behavior at the interface between the dielectric and the two top electrodes according to the electrode material.The binding state of the Ti atoms in Ba 0.5 Sr 0.5 TiO 3 at the interface with the top electrode was analyzed using the Ti-L edge as recorded by electron energy loss spectroscopy (EELS) (Figure 2a,d).For Ba 0.5 Sr 0.5 TiO 3 in contact with the SrRuO 3 top electrode, all Ti atoms in Ba 0.5 Sr 0.5 TiO 3 showed t 2g and e g peaks, implying that the Ti atoms were in a Ti 4+ binding state and formed a perovskite structure.However, for the Ru top electrode, the Ba 0.5 Sr 0.5 TiO 3 in the three unit cells closest to the top interface demonstrated L2 and L2 peaks, which are commonly seen in the Ti 3+ binding state of the oxygen-deficient perovskite oxide. [13]hen Ru metal is in contact with Ba 0.5 Sr 0.5 TiO 3 , oxygen diffusion occurs from Ba 0.5 Sr 0.5 TiO 3 to Ru, unlike in the SrRuO 3 top electrode.The oxygen diffusion from Ba 0.5 Sr 0.5 TiO 3 is confirmed by the EELS profile at the Ru top electrode (Figure S1, Supporting Information).Approximately 1.6 nm from the interface, the EELS spectrum of Ru demonstrated an O-K edge feature, confirming oxygen diffusion to the Ru top electrode.Next, high-angle annular dark field (HAADF) intensity profiling of the A-site and B-site of perovskite Ba 0.5 Sr 0.5 TiO 3 was performed at the top interface, and the unit cell length and Ti displacement of Ba 0.5 Sr 0.5 TiO 3 were derived from the distance between HAADF peaks (Figure 2c,f).Ba 0.5 Sr 0.5 TiO 3 in contact with SrRuO 3 demonstrated a unit cell length of 0.4 nm, and the Ti atom was located at the center of the unit cell, similar to that in the bulk Ba 0.5 Sr 0.5 TiO 3 , regardless of the distance from the interface.However, for Ba 0.5 Sr 0.5 TiO 3 in contact with Ru, the first unit cell from the interface was elongated compared with that of the bulk Ba 0.5 Sr 0.5 TiO 3 , and the Ti atom was displaced above the center of the unit cell.If an epitaxially grown perovskite oxide, such as SrTiO 3 contains oxygen vacancies, its unit cell is elongated along the c-axis by the attractive force of atoms near these vacancies. [14,15]These results confirm that oxygen vacancies form in Ba 0.5 Sr 0.5 TiO 3 adjacent to the metal Ru electrode via oxygen diffusion, and these vacancies elongate the Ba 0.5 Sr 0.5 TiO 3 unit cells closest to the interface along the c-axis.
The structural changes and polarization behavior based on the presence of oxygen vacancies at the dielectric layer in contact with the top electrode were simulated through DFT calculations (Figure 3).Four different configurations of the supercell were modeled, namely, the SrRuO 3 -SrTiO 3 interface without an oxygen vacancy, the SrRuO 3 -SrTiO 3 interface with an oxygen vacancy, the Ru-SrTiO 3 interface without an oxygen vacancy, and the Ru-SrTiO 3 interface with an oxygen vacancy (Figure 3a).The final energy of the structure was calculated for the oxygen vacancies in various positions in the SrTiO 3 unit cell, and based on this result, the model in which the oxygen vacancy is located at the most stable position was adopted (Figure S2, Supporting Information).For each modeled supercell, the unit cell length and Ti displacement in the SrTiO 3 unit cell adjacent to the interface are plotted in Figure 3b.SrTiO 3 in contact with SrRuO 3 with or without an oxygen vacancy and SrTiO 3 in contact with Ru but without an oxygen vacancy all demonstrated a unit cell length of 0.39 nm, which is similar to that of the bulk SrTiO 3 , regardless of the distance to the interface.However, the first SrTiO 3 unit cell from the interface in contact with Ru containing the oxygen vacancy was elongated by 10% (≈2 nm), and the Ti atom was located above the center of the unit cell.These results are well matched to those illustrated in Figure 2.These phenomena are expected to occur in the dielectric layer at the interface with the top electrode, as shown in Figure 3c.Unlike the SrRuO 3 -SrTiO 3 interface, where oxygen does not diffuse from the dielectric to the electrode, oxygen diffuses from SrTiO 3 to Ru in the Ru-SrTiO 3 interface because of the oxygen gradient.A nonequilibrium state of charge forms because of the oxygen vacancies created by diffusion.Nearby oxygen and Ti atoms shift toward the oxygen va-cancies to compensate for this state of charge, which elongates the SrTiO 3 unit cell near the interface.The cubic structure of SrTiO 3 and the SrTiO 3 elongated 10% along the c-axis were modeled to predict the polarization behavior of the dielectric layer in contact with each top electrode (Figure 4a,b); the phonon dispersion (Figure 4c) and dielectric permittivity of each cell were also calculated.The cubic SrTiO 3 demonstrated an isotropic dielectric permittivity of 251.3 in the c-axes.However, 10% elongated SrTiO 3 showed a lower dielectric permittivity than that of the cubic SrTiO 3 in all axes.The lowered dielectric permittivity (19.5) of the elongated unit cell was observed along the c-axis; these results confirm that the crystal structure of the dielectric layer can be deformed depending on the adjacent electrode, which in turn changes the dielectric property.
The capacitance-voltage (C-V) and current-voltage (I-V) characteristics of the capacitors with each top electrode on the  SrRuO 3 /Ba 0.5 Sr 0.5 TiO 3 stack were evaluated (Figure 5).In the C-V plot (Figure 5a), capacitors with the SrRuO 3 and Ru top electrodes demonstrated maximum dielectric permittivities (k max ) of 953 and 667, respectively.This difference in the k max value is attributed to the effects of the c-axis-elongated layer formed by oxygen diffusion from the Ba 0.5 Sr 0.5 TiO 3 to the metal Ru electrode, as described in Figure 3 and Table 1.This layer demonstrated low dielectric permittivity and acted as a dead layer, thus depressing the dielectric property of the entire capacitor.Therefore, the capacitor with the Ru top electrode showed a lower k max than that with the SrRuO 3 electrode.Meanwhile, the capacitor with the SrRuO 3 top electrode reached its k max value at an applied bias of 0 V, whereas the k max with the Ru top electrode occurred at -0.68 V.These results can be explained by the built-in field between the Ru top and SrRuO 3 bottom electrodes (Figure S3, Supporting Information), which formed because the SrRuO 3 bottom electrode and Ru top electrode have a gap between their work functions of approximately 0.5 eV.The Ti atoms in Ba 0.5 Sr 0.5 TiO 3 shifted toward the SrRuO 3 bottom electrode because of this field.Therefore, when a negative voltage was applied to the top electrode, the Ti atom was located at the center of the octahedron of oxygen, where the Ti atom exhibited the maximum polarization behavior, i.e., k max .
Next, the I-V characteristics were examined, as shown in Figure 5b, which reveals the different behaviors of the capacitors with the two different top electrodes.The capacitor with the SrRuO 3 top electrode showed a symmetric leakage current density with applied bias, whereas it was asymmetric with the Ru top electrode.Specifically, its leakage current density in the positive bias was similar to that with the SrRuO 3 electrode; however, in the negative bias below −0.85 V, its leakage current density increased by two orders of magnitude.As this measurement was performed with a grounded bottom electrode and a biased top electrode, the region of negative bias reflects the characteristics of the top interface of the capacitor.Therefore, this asymmetric leakage current density is caused by the effect of oxygen vacancies at the top interface, as shown in Figure 2.
To understand the defect behavior in the conduction mechanism at the interface based on the top electrode, the negative voltage area was fitted in I−V plots at various temperatures (Figure 6a,b).The parameters used for the fitting are listed in Table 1.Both capacitors with SrRuO 3 and Ru top electrodes demonstrated conduction via Poole-Frenkel (P-F) emission, which can be expressed as follows: where J, q, E, q T , μ, and N c represent the leakage current density, charge of an electron, applied electrical field, trap energy level (i.e., the activation energy of trapped electrons), electronic drift mobility, and density of states (DOS) in the conduction band, respectively. [16]In a previous study, we determined that P-F emission can occur at the SrRuO 3 /Ba 0.5 Sr 0.5 TiO 3 stack in the capacitor.If Ru diffuses from the bottom SrRuO 3 electrode to the Ba 0.5 Sr 0.5 TiO 3 dielectric, the diffused Ru can form shallow-level  defects of 0.07-0.12eV and act as a trap site for P-F emission. [17]oth capacitors in this study demonstrated P-F emissions with trap sites that have an energy level of 0.1 eV; thus, this P-F emission can be attributed to Ru diffusion from the bottom electrode.Meanwhile, the dominant conduction mechanism of the capacitor with the Ru top electrode changed from the P-F emission to the hopping conduction in the high voltage area, unlike the capacitor with the SrRuO 3 top electrode.Hopping conduction can be expressed as follows: where a, n, and  represent the hopping distance, electron concentration in the conduction band, and thermal vibration frequency of electrons at the trap sites, respectively. [18]This hopping conduction was caused by the trap site with an energy level of 0.98 eV.The oxygen vacancies in BaSrTiO 3 demonstrated an energy level of approximately 1.0 eV. [19]Therefore, the 0.98 eV trap sites considered to have caused the hopping conduction were oxygen vacancies, which were formed by oxygen diffusion from the Ba 0.5 Sr 0.5 TiO 3 dielectric to the Ru top electrode at the top interface.Next, to analyze the defect distribution at the interfaces with the different top electrodes, the interface defect state densities (D it ) versus the AC frequency of the C-V measurement were determined at an applied DC bias of −1 V (Figure 6c).D it with the AC frequency term can be expressed as follows: where G p represents the parallel conductance, and  represents the AC frequency of the C-V measurement. [20]In the D it versus frequency relationship, all defects in the dielectric layer can respond to a low-frequency AC bias and are reflected in D it .However, as the frequency of the AC bias increases, the defects far from the electrode applying the AC bias do not respond; instead, only defects near the interface respond to the AC bias and are reflected in D it .The capacitor with the SrRuO 3 top electrode demonstrated a D it that decreased linearly with the increasing AC frequency.By contrast, the D it of the capacitor with the Ru top electrode decreased at first but then became constant above the AC frequency of 5 × 10 4 Hz.This result can be attributed to the oxygen vacancies in Ba 0.5 Sr 0.5 TiO 3 at the interface with the Ru top electrode; these results agree with the abovementioned results.

Conclusion
An atomic-scale study on the interface between electrode and dielectric was performed to investigate the formation mechanism of the dead layer and its effects on the charge storage capacity.The interface where the dead layer was formed was engineered using SrRuO 3 oxide or Ru metal as the top electrodes on the SrRuO 3 /Ba 0.5 Sr 0.5 TiO 3 stack.At the interface between the Ru metal top electrode and the Ba 0.5 Sr 0.5 TiO 3 dielectric, oxygen diffused from the dielectric to the electrode, resulting in the caxis elongation of Ba 0.5 Sr 0.5 TiO 3 at the interface, unlike that with the SrRuO 3 top electrode.This elongated Ba 0.5 Sr 0.5 TiO 3 acted as a dead layer and degraded the dielectric permittivity of the capacitor.The existence of oxygen vacancies in the Ba 0.5 Sr 0.5 TiO 3 dielectric near the Ru interface was confirmed by a conduction mechanism fitting and the measurement of interface defect state densities.When the interface was engineered using SrRuO 3 oxide instead of Ru metal as the top electrode, the dead layer was suppressed, which enhanced the value of k max from 667 to 956.Through this study, we determined the formation mechanism of the dead layer in a perovskite-oxide-based capacitor and its effect on the charge storage capacity.Our work is expected to provide insights into how to overcome the limitations of the nextgeneration DRAM.

Experimental Section
Film Growth and Sample Preparation: The pulsed laser deposition (PLD,  = 248 nm of the KrF excimer laser) method was used to epitaxially grow the SrRuO 3 bottom electrode and Ba 0.5 Sr 0.5 TiO 3 dielectric layer.Prior to growing the epitaxial film, single-crystal SrTiO 3 (100) substrates were treated to form a Ti-terminated surface.The substrates were etched with a buffered hydrofluoric acid etchant and annealed at 1273 K for 1 h under air.The epitaxial SrRuO 3 bottom electrode and Ba 0.5 Sr 0.5 TiO 3 dielectric layer were grown under a 100 mTorr oxygen atmosphere at a substrate temperature of 973 K.During the PLD process, a fluence of 2 J cm −2 and a 10 Hz repetition frequency were maintained for the excimer laser.The thickness of the SrRuO 3 bottom electrode and Ba 0.5 Sr 0.5 TiO 3 dielectric layer were fixed at 30 nm and 10 nm, respectively.Next, 5 nm of Ru or SrRuO 3 was grown as the top electrode via sputtering or PLD, respectively, to identify differences in the interface and dead layer based on the contacted electrode.Both top electrodes were defined through the patterning process with a Ti/Pt hard mask.
Structural and Electrical Analysis: The quality of each capacitor was confirmed using XRD and STEM.The interface with each top electrode was characterized by EELS and HAADF as tools of STEM.The temperatureand frequency-dependent C-V and I-V characteristics were analyzed using a probe station with a grounded bottom electrode and a biased top electrode.The dielectric leakage currents with temperature changes (298-423 K) were used to analyze the conduction mechanism.The defect positions in the dielectric film were estimated with the capacitance data measured with varying AC frequency (1 kHz to 1 MHz).
[23][24] An energy cut-off of 600 eV was used to truncate the plane-wave basis.A SrTiO 3 unit cell was relaxed with a force convergence criterion of 0.001 eV Å −1 .Further, 6.5 × Ru/9 × SrTiO 3 and 6.5 × SrRuO 3 /9 × SrTiO 3 superlattices with a SrO layer as the interface were prepared using the relaxed SrTiO 3 structure to examine the structural effects of interfacial oxygen vacancies based on the type of electrode.Considering the consistency of the crystal structure with perovskite SrTiO 3 , the face-centered cubic and perovskite crystal structure were applied for Ru and SrRuO 3 , respectively.The slab structures were created by applying a vacuum region with a length of more than 25 Å.An oxygen vacancy was formed at the interface, and structural relaxation was performed under the conditions of fixing the cell shape with a force convergence criterion of 0.01 eV Å −1 .To calculate the dielectric constants of the cubic and elongated SrTiO 3 , density-functional perturbation theory [25,26] implemented in VASP was used.The phonon band diagram was obtained from the PHONOPY package. [27]The local density approximation function was employed to determine the dielectric constants and perform the phonon calculations.

Figure 1 .
Figure 1.Structural characterization of each capacitor.a) XRD patterns of each stack, b) and c) cross-sectional HAADF-STEM images for the capacitors with the SrRuO 3 and Ru top electrodes, respectively; the insets show the FFT pattern of each layer.

Figure 2 .
Figure 2. Interfacial characterization of each capacitor.EELS profiles of Ba 0.5 Sr 0.5 TiO 3 adjacent to the a) SrRuO 3 and d) Ru top electrodes, HAADF intensity profile of Ba 0.5 Sr 0.5 TiO 3 adjacent to the b) SrRuO 3 and e) Ru top electrodes, and unit cell length and Ti displacement of Ba 0.5 Sr 0.5 TiO 3 adjacent to the c) SrRuO 3 and f) Ru top electrodes.

Figure 3 .
Figure 3. Structural behavior prediction of SrTiO 3 at the interface.a) Different supercell structures according to the adjacent electrode and the existence of an oxygen vacancy, b) unit cell length and Ti displacement for each supercell, and c) schematics of the structural behavior of SrTiO 3 according to the SrRuO 3 and Ru top electrodes.

Figure 4 .
Figure 4. Polarization behavior prediction of SrTiO 3 .Supercells of SrTiO 3 a) with a cubic structure and b) with a 10% elongation along the c-axis; c) phonon dispersion curves of each supercell.

Figure 5 .
Figure 5. Electrical properties of each capacitor.a) C-V and b) I-V characteristics; inset shows a schematic of the measured device.

Figure 6 .
Figure 6.Defect state characterization of each capacitor.C-V characteristics at different temperatures (298-423 K) and fitting for each conduction mechanism of the capacitor with the a) SrRuO 3 and b) Ru top electrodes; c) interface defect state densities according to the frequency of AC bias in the C-V measurements.

Table 1 .
Fitting parameters used for each conduction mechanism of capacitors with each type of top electrode (TE).