Low Leakage in High‐k Perovskite Gate Oxide SrHfO3

Reducing the leakage current through the gate oxide is becoming increasingly important for power consumption reduction as well as reliability in integrated circuits as the semiconducting devices continue to scale down. Here, this work reports on the high‐k dielectric SrHfO3 (SHO) based devices with ultralow leakage current density via pulsed laser deposition (PLD). The ultralow current density is achieved by optimizing the growth conditions and the associated structural properties. In the optimized conditions, the dielectric properties of the 50‐nm‐thick SHO capacitors are measured: high dielectric constant (κ = 32), low leakage current density (<10−8 A cm−2 at 2 MV cm−1), and large breakdown field (EBD > 4 MV cm−1). The surprisingly low leakage current density of SHO is ascribed to the large bandgap (≈6 eV), the large conduction band offset (CB offset > 3 eV) with respect to the semiconductor, and the low density of defect states inside the bandgap. The optimized SHO dielectric with high dielectric constant and ultralow leakage current density is proposed for future low‐power consumption devices based on Si as well as perovskite oxide semiconductors.


Introduction
Along with the continuing progress of complementary metal-oxide-semiconductor (CMOS) integrated circuit technology, the increase in integration density inevitably escalates the power consumption. In order to reduce the power consumption, the operating voltage has to be reduced, which requires thinner gate oxide. As the gate oxide thickness approaches a few nm, the inevitable tunneling leakage current starts to appear, which in turn requires the development memory, and flash memory applications. [7] There have been some reports on the investigation of Hf-based high-k dielectrics in CMOS devices. [10,11] However, as the semiconductor devices are continuously scaling down, the need for high-k dielectrics with higher dielectric constant and lower leakage current keeps increasing for the future device applications.

Dielectric Properties
To measure the dielectric properties of SrHfO 3 (SHO) dielectric, 150-nm-thick conducting 4% La-doped BaSnO 3 (4% BLSO) [55] layer was epitaxially deposited on single crystal SrTiO 3 (001) substrate as a bottom electrode. Subsequently 50-nm-thick epitaxial SHO dielectric layer was grown and the circular 4% BLSO electrodes were deposited on top of the SHO dielectric layer, which forms metal-insulator-metal (MIM) capacitors. Schematic of the devices is shown in inset of Figure 2a. The area of the top electrode is about 53100 µm 2 (diameter = 260 µm). The capacitance-frequency and current density-electric field characteristic curves of a 50-nm-thick SHO (grown in 30 mTorr) capacitor are shown in Figure 2a,b. The dielectric constants (κ) of SHO capacitors were calculated from the measured parallel capacitance (C p ). DC field dependence of dielectric constant of SHO is shown in Figure S1 (Supporting Information), which shows very little dependence. The behavior of the leakage current density in the SrHfO 3 capacitor, optimally grown in 30 mTorr, is remarkable considering that the leakage density level remains below 10 −6 A cm −2 even at 4 MV cm −1 , which is close to its dielectric breakdown field. When we used a slow and quiet mode to better measure the ultrasmall leakage current quantitatively in the low field region, as shown in Figure S2 (Supporting Information), the leakage current was less than 4 pA, converting to less than 10 −8 A cm −2 , at 2 MV cm −1 . When considering the geometry of our capacitors, hundreds of µm which is quite large compared to the length scale of the current semiconducting devices, this low leakage current level is more remarkable and could be reduced further if the geometry of the devices is scaled down. To verify the effect of growth conditions on the dielectric properties of SHO capacitors, we fabricated other capacitors in different oxygen partial pressures. The representative cur- www.advelectronicmat.de rent density-electric field characteristic curves of the devices are shown in Figure S3 (Supporting Information). Comparison of dielectric constant, breakdown field and leakage current density at 2 MV cm −1 of the SHO capacitors deposited in different oxygen partial pressures is shown in Figure 2c. Considering that the dielectric breakdown field depends on the extrinsic

www.advelectronicmat.de
factors such as local defects, its deviation is larger than the dielectric constant, which should be treated statistically. [44] The error bars in Figure 2c represent the maximum and minimum value of breakdown fields. The dependence of the dielectric constants of the SHO capacitors on the oxygen partial pressure is relatively small, though the dielectric constant of SHO grown in 30 mTorr carries a maximum value of 32. However, the leakage current density of the device shows a clear minimum at 30 mTorr, suggesting that the defect density of state inside the bandgap of SHO is minimal in the 30 mTorr growth since the leakage current in the field range away from the dielectric breakdown field is dominated by the conduction through the defect states.
To estimate the remarkably low leakage current density, Fowler-Nordheim (FN) tunneling analysis was applied to the current density-electric field characteristic curve of the SHO near the breakdown field and compared to other high-k perovskite dielectrics such as BaHfO 3 (BHO), [26] LaScO 3 (LSO), [19] and LaInO 3 (LIO), [53] as shown in Figure 3a. The good linear fitting in the ln(J E −2 ) versus E −1 curve of SHO in the high field regime validates that FN tunneling is the dominant conduction mechanism in the high field regime near the breakdown field. In the case of the degenerately doped 4% BLSO, the fermi level of 4% BLSO is 0.47 eV above its conduction band (CB) minimum, estimated from the following relation: where E F,BLSO , E CBM,BSO , n and BLSO * m are the fermi level of 4% BLSO, the CB minimum energy level of BSO, the carrier concentration of 4 × 10 20 cm −3 , and the effective mass of 4% BLSO of 0.42m 0 , [56] respectively. Taking this into account, from the FN tunneling analysis, the CB offset between BaSnO 3 (BSO) and SHO was calculated to be 3.3 eV, which is a very large value. According to the theoretical prediction by Bjaalie et al., the CB offset between SHO and BSO is 3.27 eV, [57] which matches closely with our value. This CB offset is as large as the value between Si and SiO 2 , which is 3.2 eV. [7] Figure 3b compares the band alignment of some relevant materials. Compared to other dielectrics, the CB offset of SHO to BSO is 2-3 times larger than other materials, which is believed to be the main cause for the low leakage current density. As mentioned in our previous paper, [44] the electron affinity of Si (4.05 eV) [7] is similar to the work function of 4% BLSO (≈4 eV). Therefore, large CB offset of about 2.8 eV between silicon and SHO is expected, which would result in very low leakage current, as experimentally reported. [28][29][30] In order for the defect-assisted tunneling current to remain very small, concomitant with the large CB offset, there must be very little defect states inside the bandgap which can make a conducting path. Optical bandgap of SHO thin films grown on MgO (001) substrate is shown in Figure S4 (Supporting Information). Tauc's plot was used to estimate both direct and indirect bandgap of SHO. The large optical band of SHO around 6 eV is also the origin of the low leakage current density.

Field-Effect Transistor
To confirm the low leakage current in a field effect device, we fabricated an n-type FET using SHO as the gate oxide and lightly doped BSO as the channel and degenerately doped BSO as the electrodes. The schematic of the device is shown in Figure 4a. To reduce the threading dislocations from the lattice mismatch between the channel and the substrate, a 150-nm-thick BSO Comparison of the band levels of high-k perovskite oxides, BHO, LSO, LIO, and SHO, with respect to BSO and Si.
www.advelectronicmat.de buffer layer was deposited before the channel deposition. On top of it, a 20-nm-thick 0.3% BLSO channel layer was grown using a line patterned silicon stencil mask. To define channel length of 60 µm, 4% BLSO was deposited as the source and drain electrodes using a stainless stencil mask. Finally, 100-nmthick SHO gate oxide and 4% BLSO gate electrodes of length 70 µm were added. Figure 4b shows the optical microscope image of the device. The epitaxial growth of FET was confirmed by high-resolution X-ray diffraction (HRXRD) in Figure S5 (Supporting Information). The output characteristics of the device are shown in Figure 4c, which shows typical behavior of n-type FETs. Source-drain voltage (V DS ) was swept from 0 to 15 V and gate-source voltage (V GS ) was applied from 0 to 18 V with an interval of 2 V. Figure 4d is the transfer characteristics of the device in a saturation region, which is confirmed by the output characteristics. The on-off ratio of the device is about 10 8 , and the maximum mobility of the device is about 80 cm 2 V −1 s −1 . The leakage current of the device is 10 −11 A at 2 MV cm −1 , corresponding to 10 −7 A cm −2 which is a similar value to those measured in the capacitor structures although it is likely the small 10 pA leakage current measured in a normal mode will decrease when measured more carefully in a quiet mode. The clockwise hysteresis in the transfer curve is very small, indicating that the charge trap densities responsible for the hysteresis is indeed small in the SHO gate oxide. Again, this is consistent with the low leakage current and the low defect density inside the bandgap of SHO gate oxide. The I D 0.5 versus V GS plot of the device to extrapolate threshold voltage is shown in Figure S6 (Supporting Information).

Structural Properties
In order to elucidate the excellent dielectric properties of SHO grown in 30 mTorr oxygen partial pressure by PLD, reciprocal space mapping (RSM) measurement around (013) plane was performed for SHO capacitors deposited in 1, 10, 30, and 100 mTorr, which is shown in Figure 5a. The values of Q x corresponding to the diffraction peaks for SHO and 4% BLSO are almost identical, indicating that SHO dielectrics are grown on 4% BLSO coherently. Figure 5b shows the in-plane, out-of-plane lattice constant, and unit cell volume for SHO. As oxygen partial pressure increases, the in-plane lattice constant increases and the out-of-plane lattice constant decreases, whereas the unit cell volume (V unit cell = (a ip ) 2 × a op ) barely changes. The unit cell volume of SHO dielectrics deposited in different oxygen pressures is in the range of 68.6-69.0 Å 3 , which is slightly smaller than that of the Pm3m cubic phase (V unit cell, bulk = 69.6 Å 3 ). [37] To obtain the insights into the phase of SHO dielectrics grown in 30 mTorr, we perform the RSM (013) by rotating the substrate clockwise 0°, 90°, 180°, and 270° (see the Supporting Information, Figure S7, Supporting Information). The Q z values of SHO peak positions in Figure S6a (Supporting Information) are identical, suggesting that the in-plane symmetry is fourfold for the SHO dielectric grown in 30 mTorr. Therefore, we can infer that the polymorph of SHO dielectric grown in 30 mTorr is tetragonal or cubic under tensile strain, resulting in the slightly larger in-plane lattice constant (a ip = 4.100 Å) than out-of-plane lattice constant (a op = 4.092 Å). Since the peak positions of SHO and 4% BLSO were very close, it is difficult to study the dependence of structural properties of SHO on oxygen partial pressures. Subsequently, we prepared SHO thin films deposited in various oxygen partial pressures directly on SrTiO 3 (STO) (001) substrate. The thickness of the films is in the range of 45-55 nm, which was confirmed by X-ray reflectivity (XRR) measurement (see Figure S8, Supporting Information). RSM measurement around (013) plane was performed by SHO thin films deposited in 1, 10, 30, and 100 mTorr, which is shown in Figure 5c. Unlike SHO www.advelectronicmat.de dielectrics grown on 4% BLSO, SHO thin films are not grown coherently on STO substrate due to the large lattice mismatch between SHO and STO substrate (≈4.5%). The peak intensity and shape of SHO film deposited in 30 mTorr are well defined compared to other films despite all films having similar thickness. The in-plane, out-of-plane lattice constants, and unit cell volume with increasing oxygen partial pressure, which is summarized in Figure 5d, change similarly to the SHO grown on 4% BLSO. Figure 6a shows the θ-2θ scans and rocking curves for SHO films fabricated under different oxygen partial pressures. The 2θ peak position of SHO moves to the left as the oxygen partial pressure decreases. The out-of-plane lattice constants calculated from the (002) Bragg peaks of SHO films are summarized in Figure 6b. The out-of-plane lattice constants decrease as the oxygen partial pressure increases. The lattice constant of all films is larger than that of bulk SHO with Pnma orthorhombic phase (a pc = 4.087 Å). [37] Considering the lattice mismatch between SHO and STO substrate (≈4.5%), the out-of-plane lattice constant of SHO films should be larger than that of bulk SHO in light of the compressive in-plane strain. As the oxygen partial pressure increases, the relatively lighter Sr ions with molecular mass of 87.62 are scattered more than the Hf ions with molecular mass of 178.486, resulting in suppressed stoichiometry of Sr relative to Hf. Therefore, as the oxygen partial pressure increases, the Hf-ratio in the SHO films is enriched and the lattice constant decreases with increasing Hf 4+ ions having an ionic radius smaller than Sr 2+ ions.
Such changing trend of the cation stoichiometry which depends on the oxygen partial pressures was confirmed by X-ray photoelectron spectroscopy (XPS) and inductively coupled plasma atomic emission spectroscopy (ICP-AES) shown in Figure S9 (Supporting Information), which also includes the XPS spectra. We also prepared the SHO thin films while changing the growth temperature and the energy fluence in fixed oxygen partial pressure of 30 mTorr. There is no significant dependence of out-of-plane lattice constant on growth temperature and energy fluence, whereas the crystallinity of films deteriorates below the growth temperature of 750 °C (see Figure S10, Supporting Information).
Rocking curves of the SHO films deposited in 20, 30, 50, and 100 mTorr consist of narrow (yellow) and broad (black) components, whereas other SHO films grown in lower pressure consist of only a broad component as shown in Figure S11 (Supporting Information). Rocking curves composed of two components are commonly seen in epitaxial films. [58] Generally, narrow components associated with long-range correlation emerge in the films grown partially strain-relaxed or coherently on substrate, whereas broad components attributed to defect-related short-range disorder exist in the films grown partially or fully strain-relaxed on substrate. [59,60] The intensity ratios of the narrow to the broad component (I narrow /I broad ) are plotted in Figure 6c. The ratio decreases at the oxygen partial pressures higher or lower than 30 mTorr. Especially, for SHO films deposited in 1 and 10 mTorr, there is only a broad compo-nent, suggesting that there exist significant short-range structural disorders. In the case of SHO film deposited in 30 mTorr, the intensity ratio of the narrow component is the largest,

www.advelectronicmat.de
indicating that there exist the lowest defects, consistent with its excellent dielectric properties. Figure 6d shows the result of deconvolution of the 2θ peak for SHO films grown in 1, 10, 30 and 100 mTorr. The 2θ peak of SHO films grown in oxygen partial pressure lower than 30 mTorr becomes asymmetric and can be deconvoluted into two Voigt functions, suggesting that there exist disordered domains with slightly different out-of-plane lattice constant. Considering the results of XRD measurement for SHO films grown under various conditions, oxygen partial pressure is a crucial factor in determining the lattice constant and crystallinity of the film. Additionally, we confirmed that the SHO film deposited in 30 mTorr have a symmetric 2θ peak shape, the Kiessig fringes on both sides of the peak, a proper lattice constant, and the highest crystallinity, suggesting that it is grown in the optimum growth conditions.
Microstructures of the SHO film grown in 30 mTorr were characterized by scanning transmission electron microscopy (STEM) analysis. Cross-sectional STEM images of the film show uniform thickness of the SHO film of ≈50 nm (Figure 7a). In the high-angle annular dark-field (HAADF)-STEM image, vertical contrasts propagating from the STO-BLSO interface through the film are seen, which originate from threading dislocations. Some of the dislocations extending up to the SHO region are noted. The mid-angle annular dark-field (MAADF)-STEM image of the region shows bright contrast at the STO-BLSO interface as well as the threading dislocations due to the local strain fields around them. Meanwhile, the BLSO-SHO interface does not show visible strain contrast. Atomic-resolution HAADF-STEM image of the BLSO-SHO interface shows the epitaxial growth of SHO on BLSO with the interface roughness of 1-2 nm.
The SHO region shows nonuniform contrasts with clearly visible dark patches in the STEM images. Energy dispersive X-ray (EDX) elemental maps obtained from the SHO region are presented in Figure 7b, showing local nonstoichiometry with less Sr content in the dark patches (yellow arrows). The dark patches are attributed to nonstoichiometric and amorphous regions in the SHO film as witnessed by their amorphous textures in the HAADF-and bright-field (BF)-STEM images (Figure 7c). Additionally, presence of small (less than 20 nm diameter) grains misoriented from the matrix crystal is observed as well (yellow arrows). The presence of such defects is consistent with the nonvanishing broad component in the X-ray rocking curves in Figure 6a even in our optimum deposition condition of 30 mTorr pressure. We speculate that the SHO films grown at different conditions will exhibit dissimilar degree of disorder, e.g. size and density of the grains and dislocations, which will impact the material's dielectric properties and the device performance. Whereas comparative microstructure analysis is needed to elucidate the structure-property relationship, the study requires careful approach with a reliable quantification method, and thus we leave this for future research. www.advelectronicmat.de

Conclusion
In conclusion, we studied on high-k SHO dielectric with ultralow leakage current density. The remarkably low leakage current density of SHO dielectric is attributed to the large optical bandgap of SHO and the large CB offset between BSO and SHO through the analysis of optical spectra and FN tunneling. In addition, we investigate the origin of the superior low leakage current by analyzing the structural properties of SHO deposited in various oxygen pressures through high-resolution XRD measurements. The optimized SHO dielectric has low defect states inside the large bandgap, which is also responsible for the superior low leakage current density. Finally, we demonstrated the n-type accumulation mode FET using SrHfO 3 gate oxide with high mobility (µ FE = 80 cm 2 V −1 s −1 ) and low leakage current density (10 −7 A cm −2 at 2 MV cm −1 ). The excellent dielectric properties of the SHO, especially in terms of its leakage current density, suggest that its use in future electronics looks promising; the epitaxial SHO can play an important role like SiO 2 in the Si-based devices for integration of other multifunctional perovskite oxides with Si.

Experimental Section
Heterostructure Samples Fabrication: All SrHfO 3 (SHO) samples were grown on TiO 2 -terminated STO (001) single crystal substrate by PLD method (KrF excimer laser, λ = 248 nm, Coherent). One side polished STO (001) substrates were provided by MTI Korea. All targets of SrHfO 3 , 0.3%, 4% La-doped BaSnO 3 , and BaSnO 3 were provided by Toshima manufacturing Co., in Japan. (a) SHO capacitors: First, a 100-nm-thick 4% La-doped BaSnO 3 (4% BLSO) and 50-nm-thick SHO layer as bottom electrode and dielectric were grown on the entire area of the 5 × 5 mm TiO 2 -terminated STO (001) substrate. Next, a 150-nm-thick 4% BLSO layer as the top electrodes were grown using a stainless-steel mask with a total of 21 (7 × 3) circular holes, which builds 21 (7 × 3) mesalike capacitors on a single chip. (b) FET: First, a 150-nm-thick undoped BSO buffer layer was grown on the entire area of the 5 × 5 mm TiO 2terminated STO (001) substrate. Next, a 20-nm-thick 0.3% BLSO channel layer was deposited using a Si stencil line mask with a channel width of 140 µm. Subsequently, 50-nm-thick 4% BLSO source-drain contact layers were deposited using a butterfly-shaped stainless-steel mask making the channel length 60 µm. After the growth of contact layers, a 100-nm-thick SHO dielectric layer was grown using a rectangular-shaped Si stencil mask. In the final step, a 4% BLSO gate contact layer was grown on the top of the SHO dielectric layer using a Si line mask with a width of 70 µm, which covers the entire channel length with some overlap with the source and drain electrodes.
XPS Measurement: XPS analysis of SHO thin films were conducted by using AXIS Supra (Kratos, United Kingdom) equipped with a monochromatic Al-Kα source (1486.6 eV) under ultrahigh vacuum environment (10 −9 Torr). The X-ray spot size is 700 × 400 µm. Samples were attached to the mu-metal holder, by using vacuum-compatible, high-purity double-sided carbon tape. Neutralizer gun was used to minimize the impact of charging. The spectra were collected by tilting the incident beam at 54.7° with respect to the normal to the sample plane and measuring at 0.1 eV energy step. The binding energies of the components were calibrated by using the reference of C 1s binding energy (248.5 eV), removing the influence of sample charging. All peaks are fitted by using a Voigt function with L/G = 30%, and quantification was performed based on the area of peaks for Sr 3d, Hf 4f, and O 1s after a Shirley type background subtraction by using the ESCApe software.
Structural Properties Measurement: The structural properties were analyzed by θ-2θ scan, rocking curves, and RSM and STEM. θ-2θ scan and rocking curves were conducted by using SmartLab with a Cu Kα 1 source (λ = 1.5406 Å; Rigaku, Japan) at room temperature. An X-ray cross beam optics system, a Ge (220) 2-bounce monochromator, and a 1-dimensional semiconductor array detector (hybrid photon counting detector; HyPix-3000) were used for the high-resolution crystalline qualities. STEM samples were prepared by focused ion beam (FIB) lift-out method by using Zeiss Crossbeam 550 with the Gemini2 column. STEM analyses were performed by using FEI Talos F200X (S)TEM equipped with a super-X EDX detector. STEM imaging was carried out at 200 kV with a screen current of ≈35 pA. The probe convergence angle was 10.5 mrad, and the detector angles were as follows: HAADF inner angle of 64 mrad, MAADF inner angle of 25 mrad, BF outer angle of 21 mrad.
Electrical Properties Measurement: The electrical properties were measured using the Keithley 4200 semiconductor characterization system. The parallel capacitance (C p ) and dissipation factor (tan δ) were obtained from the admittance measurement with AC voltage of 30 mV root-mean square amplitude applied. For the breakdown field (E BD ) measurement, leakage current through the capacitors was measured with voltage sweep measurement.
Optical Bandgap Measurement: The optical absorption of SHO films was measured by a grating spectrometer (Cary 5000, Bruker) over 200-2000 nm (0.6-6.2 eV). The spectrometer had a quartz Iodine lamp light source for 2000-350 nm and a deuterium UV lamp light source for below 350 nm. Samples were mounted on a holder with a 3 mm diameter hole. Absorbance was calculated as the minus logarithm of the transmittance and absorption coefficient (α) was calculated by accounting for the sample thickness. A total of 50 nm thick SHO layer was grown on MgO substrate (E g = 7.8 eV) to prevent absorption by the substrate. The absorption of the substrate was removed by measuring MgO substrate's optical absorption separately and subtracting absorbance of the two samples. From the optical absorption measurement, the Tauc's plot of (αhν) n versus photon energy (for n = 0.5, 2) was plotted, where α denotes absorption coefficient.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.