Direct correlation and strong reduction of native point defects and microwave dielectric loss in air-annealed (Ba,Sr)TiO₃

Ba_xSr_1−xTiO₃ (BST) has long been studied for microwave devices due to their electric field-dependent dielectric properties, suitable for tunable high frequency oscillators, delay lines, and phase shifters.^1–4 Besides their field-dependent tunability, interdigitated dielectric capacitors (IDCs) must have low dielectric loss to minimize heat dissipation and maximize sensitivity. For high frequency microwave applications, an ideal BST film should have a high tunability, reasonable dielectric constant (>100), and low loss (tan δ < 0.02) at microwave frequencies. BST film growth conditions, dopants, post-processing, and stoichiometry all can have significant effects on microwave properties.^5–10 

So far, BST has been grown mainly by chemical solution deposition (CSD), pulsed laser deposition (PLD), metal organic chemical vapor deposition (MOCVD), and sputtering from complex oxide targets.^11–15 With the advent of oxide molecular beam epitaxy (MBE), especially hybrid oxide MBE, the ability to grow high quality stoichiometric BST films has increased, allowing greater ability to determine the role of defects and dislocations in BST films.⁶ Very recently, Lee et al. succeeded in making highly tunable and low loss Ruddlesden-Popper (RP) phase Sr_n+1Ti_nO_3n+1 (n = 6) microwave dielectric ultrathin films with a high quality factor comparable to bulk BST through oxide MBE interface engineering.¹⁶ These successes can be attributed to good control of dielectric film stoichiometry and low defect densities, e.g., oxygen, V_O, and cation, V_C, vacancies. Thin films grown by MBE, due to the high quality nature of the film and the losses being comparable to those of the substrate, make it difficult to extract dielectric properties from the film, thereby requiring thicker films to extract the properties accurately. Properties of these thicker films are similar to bulk-like properties in films grown by sintering from powder.¹⁷ However, it remains very difficult to grow practical, thick film BST, especially >500 nm, by MBE due to the inherent variations in beam fluxes and oxygen deficiencies always present in long-running growth process. Up to now, however, there has been no direct correlation established between such native defects in BST films and their corresponding microwave properties.

Here, we grew 1 μm BST films on MgO (100) substrates and annealed them in air for 12 h at varying temperatures. We used depth-resolved cathodoluminescence spectroscopy (DRCLS) to measure the native point defects with states in the band gap of these films, their density variation with process conditions, and their impact on IDC microwave properties. The DRCLS technique can optically identify point defects and their spatial distributions in complex oxides.¹⁸ The correlation of defect densities with MBE growth and process conditions provides a useful avenue to optimize IDC performance for BST without the need to fabricate metal-insulator-metal devices and maintaining epitaxy.

A Veeco GEN 930 MBE system was used to grow BST films and an ultra-high vacuum (UHV) connected PHI 5000 VersaProbe™ X-ray photoemission spectroscopy (XPS) system to characterize them.¹⁹ Sr and Ba molecular beams were evaporated from two Veeco low temperature effusion cells with Ti crucibles, whereas Ti was evaporated by a Veeco high temperature effusion cell with Ta crucible. XPS standards included SrTiO₃ (STO) single crystals and BaTiO₃ (BTO) powder. We calibrated the beam flux using a shuttered reflection high energy electron diffraction (RHEED) oscillation method to determine the 1 monolayer (ML) dose shutter time for STO and BTO growth. RHEED oscillations displayed uniform layer-by-layer stoichiometric growth consistent with maintaining BST stoichiometry to within 1% of stoichiometry.²⁰ Besides reciprocal space mapping (RSM) maps presented below, high resolution X-ray diffraction (HRXRD) 2θ-ω scan exhibits phase pure diffraction features (supplementary material²¹). Similarly, DRCLS spectra of the as-deposited BST presented below display uniform features over the outer 200 nm, consistent with uniform stoichiometry.

XPS showed near-stoichiometry high quality BST films, consistent with the high resolution X-ray RSM results with high (001) orientation. In-situ XPS established precise Ba and Sr beam flux doses for Ba_0.50Sr_0.50TiO₃ films growth on MgO (100), i.e., to determine the correct shutter time to grow 0.5 ML of BaO and 0.5 ML SrO to match 1 ML TiO₂ for each shuttered growth cycle. For calibration and subsequent film growths, oxygen plasma power and pressure were maintained constant at 200 W and 1.0 × 10⁻⁶Torr; substrate temperature was kept at $≃$650 °C measured by an IRCON InfraRail system, as described in the detail previously.²² Relatively low oxygen partial pressure maintained with both turbo and cryopump, large Ti Knudsen cell, and XPS stoichiometry monitored over extended time to identify periods of stable beam flux all contributed to mitigate compositional changes. The 10⁻⁶Torr oxygen pressure was selected for maximum stable power of the Veeco plasma source and to minimize Knudsen source oxidation and MBE component degradation.

Four 5 × 5 mm² MgO (100) substrates were mounted together on a single sample holder to ensure identical growth conditions during the full MBE growth process. MgO substrates provided a close lattice match to BST as well as low (9–10.1 ε₀) permittivity for accurate microwave measurements. To avoid parasitic capacitance and to confine the field in dielectric film, BST films with 1-μm-thick were grown on MgO (100) substrates. Once growth was complete, these four BST samples were transferred quickly into the XPS chamber in vacuo to measure the stoichiometry using a peak area integration method with 1% elemental precision. After XPS measurement, these four samples were transferred in UHV into another UHV chamber for DRCLS measurements at 80 K with a constant power of 1 mW and electron beam energies E_B from 1 kV to 5 kV in 1 kV steps. Electron-hole excitation depths increase with increasing beam energy E_B, varying on a scale of tens of nm or less according to Monte Carlo simulations.^18,23 The DRCLS system used for this study is described elsewhere.^23,24 In order to reduce the defect densities, we annealed three of these four as-deposited BST films in air separately for 12 h each at 500 °C, 725 °C, and 950 °C. Besides measuring DRCLS variations of oxygen vacancies with temperature, previous Naval Research Laboratory studies of air-annealed BST for IDCs suggested similar times and temperature ranges for oxygen to diffuse into BST.²⁵ 

High resolution RSM on both symmetric and asymmetric planes was performed before and after annealing using a high resolution Rigaku Smartlab system. Stoichiometry was measured again after annealing. Finally, IDCs were fabricated across each surface using a structure and mask with eight fingers. The finger length, width, and the air gap length were 80, 10, and 10 μm, respectively. See also the supplementary material.²¹ S₁₁ measurements were made on these interdigitated capacitors, whose electric field is applied along with the [100] in-plane direction of the films, through an HP 8510C network analyzer. Samples were analyzed from 1 to 20 GHz and a voltage sweep of 40 V. Microwave measurement results were compared directly to DRCLS features to correlate microwave properties with native point defect intensities in as-deposited and post-annealed BST films.

Table I shows that the stoichiometry of these four BST films changes with anneal temperature. Based on multiple XPS measurements of all the MBE-grown and annealed samples, standard deviations for [Ba], [Sr], [Ti], and [O] percentages were 0.29%, 0.35%, 0.35%, and 0.61%, respectively. Table I shows a statistically significant [O] increase versus [Ba] and [Sr] decrease between the as-deposited versus the 725 °C and 950 °C films. These results indicated that the 725 °C and 950 °C air anneals induced significant oxidation and, together with RSM results below, improved crystallinity above the 650 °C growth temperature. The presence of a BaO-rich top layer for the as-deposited BTO²⁶ and the vapor pressure differences between BaO²⁷ and SrO²⁸ at the elevated annealing temperatures used can account for the Ba/Sr ratio changes shown. After air annealing at 950 °C, the Ba/Sr ratio is very close to 50:50. Considering the XPS standard deviations, Table I shows that 950 °C annealing in air produces a MBE-grown BST film with stoichiometry closest to Ba_0.50Sr_0.50TiO₃.

Figure 1 shows DRCLS measurements for as-deposited and 950 °C air-annealed BST versus maximum excitation depth (Bohr Bethe range R_B) for depths of 20–200 nm, measured at 80 K versus 300 K to increase emission intensities. Peak features correspond to band gap and various defect transitions diagrammed previously.¹⁸ In order to compare native defect densities in these four BST films, DRCLS intensities were normalized to the 3.60 eV direct bandgap of BST.^29,30 For as-deposited BST in Figure 1(a), broad sub-band gap emission between 2.0 and 3.0 eV dominates the spectra, corresponding to V_C transitions at 2.35 eV and V_O complex emissions at 2.1–2.2, 2.6, and 3.0 eV.¹⁸ Corresponding metal vacancy-V_O complex densities for the as-grown BST in Figure 1(a) are >4 × 10¹⁸cm⁻³ from previous calibration by positron annihilation spectroscopy.¹⁸ Such densities are orders of magnitude below XPS sensitivities. After 950 °C annealing, all the normalized defects in the BST band gap decrease by nearly two orders of magnitude within the BST bulk. The sole exception is a 1.9 eV V_C peak within a few nm of the surface.¹⁸ Figure 1(b) shows that nearly all sub-gap BST emission in the bulk (E_B = 5 keV) decreases to below the 3.6 eV band gap intensity within a depth of $≃$ 200 nm. With the overall BST defect decrease, a 1.73 eV defect re-emission from the underlying MgO substrate excited by the BST emission becomes more apparent in Figure 1(b).³¹ The 1 μm BST films annealed at intermediate temperatures of 500 °C and 725 °C show monotonic decreases intermediate between Figs. 1(a) and 1(b) (supplementary material²¹), indicating that increasing temperature systematically lowers defect densities with states in the band gap.

Figure 2 shows the high resolution RSM result of both as-deposited and 950 °C annealed BST films. The symmetric planes of BST (002) and MgO (002) appear in Figs. 2(a) and 2(b), whereas the asymmetric planes of BST (113) and MgO (113) appear in Figs. 2(c) and 2(d). Note that the center of the BST (002) peak lies directly along the vertical with the MgO (002) peak, indicating a (00 l) epitaxial orientation of BST films for both as-deposited and 950 °C annealed. No secondary peaks were observed indicating phase purity. Note that the map of 950 °C annealed BST (002) became narrow along Q_x direction, compared to the map of as-deposited BST (002). This indicates that 950 °C annealing improves the crystallinity of as-grown BST films, reducing the FWHM of BST (002) to 0.00473 from 0.00522 nm⁻¹ along the Q_x direction. The position of the (113) BST peaks, not along the same vertical as the substrate peak but rather along a line connecting the origin and substrate peak, which indicates that the BST film is relaxed with respect to the substrate. After 950 °C annealing, the FWHM of BST (113) was also reduced to 0.00685 from 0.00707 nm⁻¹ along the Q_x direction. Combining BST (002) and BST (113), we obtain the in-plane and out of plane lattice constants a and c of these BST films, i.e., a = 3.938 Å and c = 3.986 Å for as-deposited, and a = 3.953 Å and c = 3.991 Å for BST annealed at 950 °C. Although the crystal structure of Ba_0.5Sr_0.5TiO₃ should be cubic at room temperature, the as-deposited and 950 °C annealing BST films still show tetragonal structure, probably due to film defects,³² e.g., Vo, formed during MBE growth with low oxygen pressure. The change of c/a ratio, from 1.012 of as-deposited to 1.009 of 950 °C annealed, indicates that the BST film becomes less tetragonal and much closer to cubic after 950 °C annealing, which suggests that Vo-related defects cause a tetrahedral distortion in these BST films since the Vo density was greatly reduced after 950 °C annealing.

In order to explore the role of native defect in BST film microwave properties, microwave reflection measurements (S₁₁) were performed on the IDC capacitor devices of as-deposited and 950 °C annealed 1 μm BST films through a vector network analyzer. Before the measurement, calibration was performed on the commercial MgO (001) wafer used as the substrate of these BST films. The measured S₁₁ data were fitted to a parallel resistor-capacitor model shown in the inset of Figure 3 to determine capacitance and dielectric loss. The dielectric constants are extracted using a modified conformal mapping partial capacitance method from measured capacitance and dimensions of the interdigitated structure.³³ Figure 3(a) shows the plot of the reflection coefficient (S₁₁) on the Smith chart for as-deposited and 950 °C annealed BST films. From 1 to 20 GHz, 950 °C annealed BST shows pure capacitive response, whereas as-deposited BST shows leaky capacitor properties even at low frequency and can be modeled as a capacitor with a parasitic shunt resistor. This shows that native point defects play a major role in the conductivity of these BST films. In particular, the higher V_O and other defect densities in as-deposited BST cause large leakage currents, whereas the lower defect density in 950 °C annealed BST results in a high quality capacitor with low leakage current at higher GHz, consistent with the previous reports.^5,34,35

To further evaluate the microwave properties of 950 °C annealed BST, we measured dielectric permittivity at room temperature versus dc bias from −40 V to 40 V and back in 5 V steps to generate a maximum 40 kV/cm⁻¹ in-plane electric field. The measured dielectric permittivity of 115–155 and tunability of 13% appear in Figure 3(b). The slight offset from the zero field tunability arises from the persistent polarization due to residual ferroelectricity.^11,36 Fig. 3(c) shows the dielectric loss from 1 to 20 GHz at zero volt dc bias. From 1 to 5 GHz, tan δ varies between ∼0.01 and 0.03, then stabilizes at ∼0.02 from 5 to 15 GHz. Tan δ is typically <0.01 at 1 GHz and 0.02 at 10 GHz. Over 15 GHz, tan δ increases rapidly since electromagnetic wavelength in the BST film has nearly reached the dimension of the IDC capacitor geometry so that the lumped-element model is no longer applicable.³⁷ Indeed, dielectric permittivity shows a similar frequency-dependent behavior (supplementary material²¹), but its tunability is nearly independent of frequency. Figure 3(d) shows the dielectric loss at bias voltages from −40 to 40 V at 5, 10, and 20 GHz. For 10 GHz, tan δ is reduced to nearly 0.002 at −40 V dc bias field, corresponding to a Q of 500 with field gradient of only 40 kV/cm. This is comparable to values with higher field gradients for recently reported state-of-art microwave properties in Refs. 6 and 16. Our 13% tunability is lower but comparable to ∼18% obtained previously with higher (50 kV/cm) field gradient and tan δ < 0.01 in this frequency range.¹⁶ Besides our lower field gradient, the lower tunability is due, in part, to a tradeoff between tan δ and permittivity.³⁸ The dramatic improvement of capacitive response with reduction of features at 2.2–2.4 eV and 2.95 eV suggests that V_O-related^18,39 and possibly V_C¹⁸ defects, contribute to lower dielectric loss.

In summary, we have grown 1 μm-thick BST with low loss at GHz frequencies by oxygen plasma-assisted MBE and a subsequent air annealing process guided by DRCLS. MBE is well suited to tune the Ba/Sr ratio for specific device applications, and DRCLS provides a guide to identify annealing temperatures that reduce Vo and Vc-related densities to optimize microwave properties. This correlation between native point defect densities and microwave properties opens an avenue to optimize microwave performance of BST films by controlling growth conditions and post-processing. Further reductions of defects with growth and processing guided by DRCLS have the potential to achieve even higher microwave performance.

This work supported by NSF MRSEC Grant No. DMR-1420451 (Charles Ying) and NSF Grant No. DMR-1305193 (Charles Ying and Haiyan Wang). A. Podpirka acknowledges support as an NRC postdoctoral fellow at the Naval Research Laboratory.