Effects of vacuum annealing on the electron mobility of epitaxial La-doped BaSnO3 films

Wide bandgap (Eg ∼ 3.1 eV) La-doped BaSnO3 (LBSO) has attracted increasing attention as one of the transparent oxide semiconductors since its bulk single crystal shows a high carrier mobility (∼320 cm2 V−1 s−1) with a high carrier concentration (∼1020 cm−3). For this reason, many researchers have fabricated LBSO epitaxial films thus far, but the obtainable carrier mobility is substantially low compared to that of single crystals due to the formation of the lattice/structural defects. Here we report that the mobility suppression in LBSO films can be lifted by a simple vacuum annealing process. The oxygen vacancies generated from vacuum annealing reduced the thermal stability of LBSO films on MgO substrates, which increased their carrier concentrations and lateral grain sizes at elevated temperatures. As a result, the carrier mobilities were greatly improved, which does not occur after heat treatment in air. We report a factorial design experiment for the vacuum annealing of LBSO films on MgO substrates and discuss the implications of the results. Our findings expand our current knowledge on the point defect formation in epitaxial LBSO films and show that vacuum annealing is a powerful tool for enhancing the mobility values of LBSO films. © 2018 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/1.5054154 Transparent oxide semiconductors (TOSs) are promising candidates for various future electronic devices such as transistors, solar cells, and display panels.1,2 For such applications, high electron mobility (μ) is essential, and this has been a great disadvantage for TOS materials since their mobility values are low compared to classical semiconductors. In this regard, perovskite La-doped BaSnO3 (LBSO) is gaining significant interest since its bulk single crystal exhibits a wide bandgap (∼3.1 eV) and a very high mobility of 320 cm2 V−1 s−1,3,4 which is comparable to that of doped single crystal Si (∼350 cm2 V−1 s−1).5 For this reason, there have been many attempts to utilize LBSO in thin films transistors. However, crystalline defects prevent the μ in LBSO films from reaching the single crystal value, and many studies were devoted to improving the crystal quality of epitaxial LBSO films.6–12 For example, since misfit dislocations occur at the film/substrate interface due to the lattice mismatch, buffer layers are commonly used to reduce the dislocations.13,14 To completely eliminate the film/substrate mismatch, Lee et al. grew LBSO films on single crystal BaSnO3 substrate.15 In addition, in our recent study, we fabricated LBSO films under an ozone atmosphere to reduce the amount of point defects.16 Unfortunately, while these approaches were successful in improving the mobility values in LBSO films (up to ∼120 cm2 V−1 s−1), they can significantly increase the fabrication cost, which may be a crucial issue in mass production systems at industrial scales. In large scale production facilities, designing a clever post treating process is often more economical than improving the quality of as-deposited samples. In this regard, very interesting experimental results were released in 2015 and 2018. In one study, N2 environment at 1000 ◦C was used to create oxygen vacancies in the LBSO films on SrTiO3 substrates, and the μ increased from 41 cm2 V−1 s−1 to 78 cm2 V−1 s−1.17 In the other study (same research group), oxygen vacancies were generated in the LBSO films on SrTiO3 substrates using H2 APL Mater. 7, 022507 (2019); doi: 10.1063/1.5054154 7, 022507-1

However, crystalline defects prevent the µ in LBSO films from reaching the single crystal value, and many studies were devoted to improving the crystal quality of epitaxial LBSO films. [6][7][8][9][10][11][12] For example, since misfit dislocations occur at the film/substrate interface due to the lattice mismatch, buffer layers are commonly used to reduce the dislocations. 13,14 To completely eliminate the film/substrate mismatch, Lee et al. grew LBSO films on single crystal BaSnO 3 substrate. 15 In addition, in our recent study, we fabricated LBSO films under an ozone atmosphere to reduce the amount of point defects. 16 Unfortunately, while these approaches were successful in improving the mobility values in LBSO films (up to ∼120 cm 2 V −1 s −1 ), they can significantly increase the fabrication cost, which may be a crucial issue in mass production systems at industrial scales.
In large scale production facilities, designing a clever post treating process is often more economical than improving the quality of as-deposited samples. In this regard, very interesting experimental results were released in 2015 and 2018. In one study, N 2 environment at 1000 • C was used to create oxygen vacancies in the LBSO films on SrTiO 3 substrates, and the µ increased from 41 cm 2 V −1 s −1 to 78 cm 2 V −1 s −1 . 17 In the other study (same research group), oxygen vacancies were generated in the LBSO films on SrTiO 3 substrates using H 2 ARTICLE scitation.org/journal/apm forming gas at 950 • C, which further improved the µ up to 122 cm 2 V −1 s −1 . 18 According to these studies, oxygen vacancies can neutralize the negative charges at threading dislocations [19][20][21] and induce lateral grain growth at elevated temperatures.
As removing oxygen ions (O 2− ) near threading dislocations decreases their thermal stability, oxygen vacancy doping creates a very strong driving force for lateral grain growths at high temperatures, which significantly increases the free propagation length of the carrier electrons. These results suggest that post treating of LBSO films can be just as effective as modifying the synthesis methods for improving the asdeposited crystal quality of LBSO films. In undoped BaSnO 3 films, vacuum annealing is commonly used to create oxygen vacancies and induce mobile charge carriers. 22,23 Since vacuum annealing process is much simpler than creating N 2 or H 2 forming gas environment, it can be an alternative method for inducing oxygen vacancy assisted grain growths in LBSO films. In addition, since the deposition of oxide takes place in a vacuum chamber, the film growth and post annealing can be combined into one process. There is one study that examined the effect of vacuum annealing on the electron transport properties of LBSO films, 24 but the reported µ values are too low (<4 cm 2 V −1 s −1 ) to validate vacuum annealing as an effective method for improving µ.
For optimizing the post annealing process, it is important to understand point defect formation in LBSO films because the µ-enhancement in LBSO films begins from generating oxygen vacancies. In stoichiometric BaSnO 3 , the oxidation states of its constituents are Ba 2+ = [Xe], Sn 4+ = [Kr] 4d 10 , and O 2− = [Ne]. Since two of the constituents (Ba 2+ , O 2− ) exhibit the same orbitals with inert gases (Xe, Ne), stoichiometric BaSnO 3 crystals are believed to be thermodynamically stable, 25 and stoichiometric BaSnO 3 films do not intrinsically conduct electricity as all electrons form firmly bound states. The formation energy of oxygen vacancy in BaSnO 3 is high and can only be lowered by reducing the chemical potential of oxygen, 26 which can be achieved by lowering oxygen pressure during the film growth 27,28 or vacuum annealing at high temperatures. 22,23 Therefore, as-deposited BaSnO 3 films do not have sufficient oxygen vacancies to conduct electricity unless they are intentionally created. By contrast, oxygen vacancies are much more common in LBSO films even if sufficient oxygen is provided during the film growth. 18 In one of our previous studies, Sn 2+ states were detected from LBSO films fabricated under 10 Pa of O 2 , which implies the presence of oxygen deficiency. 16,27 These results suggest that the La-dopants may be promoting oxygen vacancy formations in BaSnO 3 films. However, the relationship between La-dopants and oxygen vacancy in BaSnO 3 has not been investigated in detail.
In this study, we studied the effect of La-dopants on the formation of oxygen vacancy in LBSO films and investigated the feasibility of enhancing the µ of LBSO films using vacuum annealing. According to our results, vacuum annealing significantly increases the µ of LBSO films. We also found that La-dopants increased the oxygen vacancy vs. lattice oxygen (V O /L O ) ratio in as-deposited LBSO films and affected the vacuum annealing effect. The results of this study expand our current knowledge on the point defect formation in epitaxial LBSO films and show that vacuum annealing is a simple and effective method for enhancing the electron mobility of LBSO films.
To confirm that vacuum annealing can induce grain growth in LBSO films, a 2% LBSO film (∼64 nm) was prepared and cut it into two pieces. We annealed one piece in air and the other piece in vacuum (<10 −2 Pa) at 750 • C for 1 h. Then, we measured the lateral grain sizes of the films before and after the heat treatments using the RSM (Fig. 1). While the (204) diffraction spot of the air annealed film was almost the same with that of the as-deposited film, the (204) diffraction spot of the vacuum annealed was two times more intense compared to the other two films (as-deposited and air-annealed). The lateral grain sizes (D) of the as-deposited, air annealed, and vacuum annealed LBSO films were 9.1 nm, 10.2 nm, and 22.3 nm, respectively. The small grain size change after air annealing is not surprising since the film was deposited at 750 • C, and no significant change in the microstructure was expected. On the other hand, the vacuum annealing substantially increased the lateral grain size, which is consistent with the H 2 forming gas experiment. 18 This shows that vacuum annealing can indeed be an alternative method for triggering oxygen vacancy assisted grain growth in LBSO films.
In order to find the optimum vacuum annealing temperature, several 2% LBSO films with similar thicknesses (∼41 nm) and electrical properties were fabricated (supplementary Table S1), and they were annealed in vacuum (<10 −2 Pa) at different temperatures ranging from 650 • C to 800 • C for 30 min. With increasing annealing temperature, the D of the films increased gradually from ∼5 nm to ∼17 nm [ Fig. 2 The as-deposited film was cut into two pieces. One was annealed in air, while the other was annealed in vacuum. Almost no change in the RSM was observed from air-annealing. On the other hand, significant increase in the lateral grain size and peak intensity can be noticed from the vacuum annealed film. Unfortunately, electrical properties of the annealed samples could not be measured since they were cut from a 1 cm 2 square sample, and our probe station can only measure 1 cm 2 square coupons.
was annealed in the vacuum [ Fig. 2  Once the films became thicker than ∼120 nm, both the asdeposited and vacuum annealed mobility values saturated at ∼84 cm 2 V −1 s −1 and ∼100 cm 2 V −1 s −1 , respectively. Both as-deposited and vacuum annealed D exhibited strong thickness dependence (supplementary Fig. S2a). We believe the grain growth is hindered by the lattice strain, which decreases with increasing thickness. 29 The highest mobility observed was 101.6 cm 2 V −1 s −1 from vacuum annealed 117 nm LBSO film (as deposited: ∼74 cm 2 V −1 s −1 , supplementary Fig. S2d), which is comparable to that observed in LBSO films with buffer layers. This confirms that vacuum annealing is an effective method for enhancing the mobility of LBSO films.
To find the effect of La-dopants on the vacuum annealing process, we annealed LBSO films (average thickness: 43 nm) with varying [La 3+ ] at 725 • C in vacuum for 30 min. with [La 3+ ] while the vacuum annealed films exhibited higher n compared to the as-deposited films [ Fig. 3(a)]. The 0.1% LBSO film, which did not conduct electricity, also became electrically conductive. The absolute value of thermopower (S) decreased with [La 3+ ], which reflects the relationship between n and [La 3+ ] [ Fig. 3(b)]. 30 The µ Hall of the annealed LBSO films was dramatically improved compared with the as-deposited LBSO films [ Fig. 3(c)], but the lateral grain size enhancement exhibited a strong doping dependence. At low doping levels (≤0.55%), the grain sizes did not increase much after vacuum annealing. It is important to note that mobility improvements were observed from 0.1% to 0.55% LBSO films despite the small changes in their grain sizes. In epitaxial films, threading dislocations often exhibit negative charges and generate energy barriers (mobility edge). [19][20][21] Therefore, mobilities in epitaxial films can depend on the Fermi energy (E F ), which increases with the carrier concentration. Since vacuum annealing increases both n and D, the µ Hall improvements observed from 0.1% to 0.55% are mainly attributed to the E F shift from additional charge carriers because they did not show significant grain size changes. This also implies that mobility improvements observed at other doping levels could have been affected the increase in n (i.e., higher E F ). To confirm the effect of E F shift on µ Hall , we vacuum annealed a 2% LBSO (45.7 nm) film from 100 • C to 800 • C in sequence in steps of 100 • C (i.e., vac. anneal 100 • C → 200 • C →· · · → 800 • C) to generate additional charge carriers while minimizing the grain growth. During the sequence annealing, the carrier concentration of the film gradually increased, but its electron mobility remained almost unchanged (supplementary Fig. S5). The drastic drop in µ Hall at 800 • C is likely from structural damages due to high V O generation rate. The sequence annealing shows that the E F at n ∼ 0.5 × 10 20 cm −3 exceeds the mobility edge of the LBSO films on MgO substrates, and the vacuum annealed mobilities are not strongly affected by the Fermi energy shift if n > 0.5 × 10 20 cm −3 . This number is consistent with the H 2 forming gas experiment, where two posttreated LBSO films with the same carrier concentration of 1.1 × 10 20 cm −3 exhibited different mobility values due to structural differences. 18 According to our results, µ at low doping levels (0.1% and 0.55%) is dominated by the Fermi level. If n > 0.5 × 10 20 cm −3 , the Fermi level seems to be above the mobility edge, and µ is mainly affected by electron scatterings (supplementary Fig. S5). The sources of electron scattering are point defects (impurity, V O ) and grain boundary (threading dislocation). The vacuum annealing effects observed from 2% to 5% LBSO films suggest that grain boundary scattering dominates up to D ∼ 12 nm and point defect scattering dominates from D > 12 nm

ARTICLE scitation.org/journal/apm
we performed the X-ray photoelectron spectroscopy (XPS) measurement of the LBSO films [ Fig. 4(a): as-deposited, Fig. 4(b): vacuum annealed]. In case of oxygen in perovskite oxides, the lattice oxygen peak (L O ) is located at ∼529 eV. If oxygen vacancies are present, this peak shifts to ∼531 eV (V O ). 18,31,32 Another oxygen peak around 532 ∼ 533 eV (A O ) can emerge from chemically adsorbed oxygen from surface contamination by organic molecules. The source of chemically adsorbed oxygen is unknown, but we believe it is related to the status of the vacuum chambers (annealing, XPS) or storing conditions. For the XPS peak fitting, a convolution between Gaussian (70%) and Lorentzian (30%) was used. For each LBSO films, the full width at half maximum (FWHM) was constrained to be the same for all 3 oxygen peaks.
The V O peak energies of the LBSO films increased after vacuum annealing, especially for films with higher [La 3+ ] [ Fig. 4(c)]. The V O /L O area ratio from the XPS of the asdeposited LBSO films increased gradually when the [La 3+ ] exceeded 1% (supplementary Table S2). Upon vacuum annealing, this ratio increased further for all films except for 7% LBSO film. The largest change in V O /L O was observed from 2% doped LBSO film. Since the solubility limit of La in LBSO was reported to be ∼5%, the oxygen vacancy reduction in 7% LBSO is likely attributed to the formation of La 2 Sn 2 O 7 , 33 which could be observed from the ceramic targets used to deposit the films (supplementary Fig. S6). These results show that La-dopants in epitaxial LBSO films affect not only the oxygen stability but also the thermal stability of the film.
The XPS results show that vacuum annealing increases oxygen deficiency in the LBSO films, and oxygen vacancies in oxides normally provide additional conduction electrons. However, associating all changes in n to additional V O from vacuum annealing does not adequately explain our results. For example, according to the thickness dependence, the carrier concentration enhancement was greatly reduced with increasing thickness [Fig. 2(c)] although the oxygen vacancy generated from vacuum annealing would have been similar as the La 3+ doping levels were the same unless the chemical potential of oxygen depends on the lattice strain. In addition, in the [La 3+ ] dependence, the changes in n [ Fig. 3(a)] do not match the changes in V O /L O [Fig. 4(c)]. The implications of these phenomena can be very interesting as they suggest strain-dependent oxygen stability or oxygen vacancies not generating additional charge carriers. However, no firm conclusions can be drawn at this moment since the vacuum annealing effect on the activation of [La 3+ ] dopant is unclear.
Although we demonstrated that the mobility boost in the optimally doped films (2%) after vacuum annealing is strongly related to the lateral grain sizes, we would like to note that there were a couple unusual instances. In supplementary Figs. S2(a) and S2(d), µ Hall of vacuum annealed 280 nm film is slightly smaller than that of vacuum annealed 117 nm film although their grain sizes are both >12 nm and their electron densities are similar (2.2 × 10 20 cm −3 and 2.1 × 10 20 cm −3 ). This is another indicator that suggests strain-dependent oxygen vacancy generation, but this hypothesis requires more experimental evidence. Furthermore, in Fig. 2, the LBSO films

ARTICLE
scitation.org/journal/apm vacuum annealed at 650 • C and 700 • C exhibit different mobility values despite having similar grain sizes and electron densities (Fig. 2). We believe these phenomena are related to the combination of all processes induced by vacuum annealing: increase in the carrier concentration, increase in the oxygen vacancy, and lateral grain growth. While these data make our study not perfect, we believe these results are still valuable as they emphasize the necessity and importance of understanding oxygen vacancy in LBSO films.
Finally, the role of La-dopants in the oxygen vacancy formation mechanism in the as-deposited LBSO films is also vague [ Fig. 4(c)]. In this regard, we believe that the role of threading dislocation in point defect formation is important. For example, in case of unintentionally V O doped BaSnO 3−δ single crystal, vacuum annealing reduces the carrier concentration and therefore reduces the oxygen vacancy level. 34 This contradicts the behavior of epitaxial BaSnO 3 films, where vacuum annealing increases the carrier concentration. 22,23 Since the main structural difference between single crystals and epitaxial films is the presence of threading dislocations, it is plausible to think that they can promote point defect formation in epitaxial films. In the context of this research, since impurities often segregate between grains separated by dislocations, 35 one possibility is the segregation of La-dopants at threading dislocations, which is plausible since La 3+ ions can compensate the missing cationic charges at threading dislocations. [19][20][21] This scenario also explains the low dopant carrier activation rate observed in epitaxial LBSO films [ Fig. 2(a)]. In this case, La 3+ vacant sites in the grain interior may lose adjacent O 2− ions due to the lack of bonding electrons, but more work is required to confirm the role of threading dislocations as well as the activation of [La 3+ ] dopants. Furthermore, it will be very interesting to re-anneal the vacuum annealed LBSO films in ambient air to inject oxygen back and fill oxygen vacancies. Unfortunately, although MgO has excellent vacuum stability, its thermal stability in air is poor, [36][37][38] and such experiments could not be considered in this study. However, annihilation of oxygen vacancies in the vacuum annealed films could potentially reduce defect scattering and further enhance the electron mobility.
In summary, we examined the effect of vacuum annealing on the electron transport properties of the epitaxial LBSO films on (001) MgO substrates. Lateral grain sizes and carrier concentrations of the LBSO films substantially increased after vacuum annealing whereas it remained almost unchanged after air annealing. We also found the oxygen vacancy vs. lattice oxygen (V O /L O ) ratio in the XPS O 1s spectra increases with vacuum annealing, proving that oxygen vacancy generation indeed provides the driving force for this process. The results of this study clearly show that vacuum annealing improves electron mobility of LBSO films, where the mechanisms vary from Fermi energy shift to oxygen-vacancyassisted-grain-growth depending on the doping levels. Unfortunately, we were not able to explain all observed phenomena in detail, but our results do highlight the necessity for more studies on the thermodynamic processes involved with vacuum annealing of epitaxial LBSO films.
The vacuum annealing approach was very effective for films with small thicknesses. Therefore, it is a very good method for making LBSO film transistors since low thicknesses are desired for reducing the power consumption. We believe these results will be useful for designing low cost fabrication methods for high-mobility LBSO films or can be used to improve the carrier mobility of other perovskite stannates such as SrSnO 3 .
See supplementary material for detailed XPS characteristics and RSM patterns of the LBSO films.