Acceptor and compensating donor doping of single crystalline SnO (001) films grown by molecular beam epitaxy and its perspectives for optoelectronics and gas-sensing

Reliable bipolar carrier transport remains a challenge in most transparent semiconducting oxides (TSOs), limiting the widespread adoption of oxides for optoelectronic devices.¹ Most widely applied TSOs can readily be doped n-type, while their p-type doping remains challenging, if not untenable. However, few TSOs show p-type conductivity² with tin (II) oxide (SnO) being among them. Compared to other binary p-type TSOs, its optical bandgap of ≈ 2.7 eV and hole mobility of ≈ 3–5 cm²/V s, for phase-pure single crystalline (001) layers, make it a candidate material for oxide complementary integrated circuits, p-channel thin film transistors, or transparent oxide heterojunctions with n-type materials.^3–7 Polycrystalline, Sn-rich SnO has even been shown to exhibit a room-temperature hole mobility of up to 30 cm²/V s.^8,9 Conductometric gas sensors are another large application domain of TSOs. While SnO has not yet been widely explored for this application, p-type oxides are generally considered to allow for high sensitivity and selectivity.¹⁰ Unintentionally doped (UID) p-type conductivity in SnO is believed to originate primarily in Sn-vacancies¹¹ or their complexes with hydrogen.¹² Theoretical studies have also suggested the possibility for bipolar doping in SnO^12,13 and enhancement of p-type properties by native defects.¹⁴ However, experimental studies exploring the effect of intentional impurities in SnO toward possible bipolar conductivity are still rare and partially conflicting. Hayashi et al.¹⁵ showed an electron density of 3.0 × 10¹⁵ cm⁻³ in UID SnO. Hosono et al. showed that the doping of SnO with >8 cation at. % of Sb (N_Sb > 2.4 × 10²¹ cm⁻³) can produce an electron density of 10¹⁷ cm⁻³,¹⁶ whereas Guo et al. demonstrated slight improvement in p-type properties by Sb-doping also.¹⁷ 

While acceptor doping of SnO should increase its hole concentration beyond the UID level, donor doping holds promises for obtaining n-type SnO conductivity or semi-insulating material, which can be useful for desirable oxide p–n homojunctions or isolating buffer layers, respectively, for SnO-based devices. Here, we revisit the possibility for reliable bipolar doping in single crystalline, phase pure SnO thin films. An experimental study, backed by theoretical verification, on the effects of La and Ga dopants in SnO is presented. Transport measurements evidence that Ga acts as acceptors in SnO, which we attribute to a preferential incorporation of Ga_Sn in SnO in the Ga¹⁺ oxidation state. La dopants, in contrast, show a clear compensating donor behavior as highly La-doped SnO thin films became semi-insulating. These measurement results are corroborated by density functional theory calculations, which predict that In_Sn and Ga_Sn are soluble acceptors in SnO, while La_Sn and other group III dopants like B_Sn and Al_Sn behave as donors.

Approximately 100–300 nm-thick UID and (La and Ga)-doped SnO (001) films were grown on YSZ (001) substrates by molecular beam epitaxy (MBE). We explored two growth techniques for SnO: plasma-assisted-MBE (PA-MBE) with a metallic Sn source and an activated oxygen source¹⁸ and suboxide-MBE (S-MBE)¹⁹ using a SnO₂+Sn mixture as an efficient and pure SnO source²⁰ without the use of additional oxygen. In both growth techniques, the dopant elements were incorporated using the metal effusion cells, and the dopant concentration was varied by changing the dopant cell temperature for different growths. Hence, the dopant flux reaching the substrate, being proportional to the metal vapor pressure, is controlled by the metal cell temperature. All samples were grown at similar substrate temperatures between 350 °C and 400 °C at which both SnO and dopant desorption for the growth surface are negligible. The thickness of the films was determined by in situ laser reflectometry. The grown layers were structurally investigated ex situ by x-ray diffraction (XRD) and atomic force microscopy (AFM). The dopant concentration was derived from energy dispersive x-ray spectroscopy (EDX) measurements as described in the supplementary material. Detailed electrical properties of the doped films were obtained from van der Pauw–Hall measurements. Hybrid functional calculations of dopant formation energies with projector-augmented wave (PAW) potentials were performed using the HSE-screened hybrid functionals implemented in the Vienna ab initio simulation package (VASP).^21–24 Defect calculations were performed using supercells adopting 192 atoms for the bulk SnO structure with the same calculation parameters as described in Ref. 12. For the bulk and dopant PAW potentials, we consider Sn:[5s²5p²], In:[5s²5p¹], Ga:[4s²4p¹], Al:[3s²3p¹], B:[2s²2p¹], and La:[5s²5p⁶5d¹6s²] electron configurations treated as valence states. Solubility limiting phases considered for the respective dopants were In₂O₃, Ga₂O₃, Al₂O₃, SnB₄O₇, and La₂Sn₂O₇.

Figure 1 shows the Ga- and La-concentrations as a function of the dopant cell temperature for doped samples grown using PAMBE, as well as the Ga-concentration for Ga-doped samples from S-MBE growth. An increasing cell temperature coincides with increasing dopant concentration for all dopants. These dopant concentrations follow an Arrhenius-like behavior indicating efficient incorporation of the dopants in the film. While an activation energy of ≈2.3 eV is obtained for Ga-doped samples grown using S-MBE, i.e., without additionally supplied oxygen, the Ga activation energy obtained for samples grown using the PA-MBE method is almost two times lower. This points to a possibility that during doping with the elemental Ga source, the oxygen background during PA-MBE growth leads to the formation and subsequent evaporation of suboxide Ga₂O in the source, in contrast to pure Ga evaporation from the source in our S-MBE growth.²⁵ 

The solubility limit of dopants in SnO was estimated from XRD. On-axis 2θ-ω scans in Fig. 2 show that phase-pure, single-crystalline SnO (001) films are obtained up to the solubility limit of the various dopants. XRD data in Fig. 2(a) indicate phase-pure SnO (001) samples at Ga concentrations up to 6.4 × 10²⁰ cm⁻³. At a higher Ga content of 1.2 × 10²¹ cm⁻³, a weak peak due to β-Ga₂O₃ (004) is present while the SnO related peak is absent. Similarly, in Fig. 2(b), at a La concentration of 9.2 × 10²⁰ cm⁻³, no SnO related peak is observed. Hence, in the concentration range studied, the incorporation of the extrinsic dopants was limited by the formation of secondary phases or amorphous layers. These XRD data indicate that the solubility limits of Ga and La in SnO are between 6.4 − 12 × 10²⁰ and 1.4 − 9.2 × 10²⁰ cm⁻³, respectively. In Fig. S1 (the supplementary material), the AFM micrographs of an ≈120 nm UID layer grown at 350 °C by PA-MBE show fine and dense surface morphology films with a root mean square roughness of ≈1.8 nm. Approximately 300 nm thick, PA-MBE grown, Ga-doped layers also maintained similar morphology as the UID layer, and an RMS roughness of ∼2.1 nm is obtained for Ga doping of 6.4 × 10²⁰ cm⁻³ while UID SnO has an RMS roughness of 1.8 nm. However for La doping at a similar film thickness, a different morphology with less coalesced grains is observed. This resulted in an order of magnitude higher RMS roughness compared to UID and Ga doped SnO layers.

Figures 3(a) and 3(b) show the Hall-measured hole concentrations and resistivities, respectively, of doped SnO thin films as a function of the dopant concentration. For comparison, red and black shaded areas correspond to the range of properties of UID reference samples grown for this study via PA-MBE and S-MBE, respectively. The UID p-type conductivities in SnO (001) layers obtained from S-MBE and PA-MBE are markedly different with higher hole concentrations for the PA-MBE grown films. This difference in transport properties for the different growth methods could be due to the enhanced formation of Sn vacancies resulting from the energetics of different growth processes with negligible dependence on the film growth rate. For S-MBE growth, SnO molecules formed in the crucible are reaching the substrate surface, which is expected to decrease the formation of Sn vacancies and complexes compared to the PA-MBE growth, where elemental Sn and activated oxygen are supplied to the substrate. Unintentional impurities in the MBE chamber may also play a role in the UID hole concentration as these values are higher than previously reported values for MBE-grown single crystalline SnO(001).^17,18,26

The reference UID SnO (001) layers grown for this study under different conditions by PA-MBE exhibit room temperature (RT) Hall hole concentrations (p_H) in the range of ≈0.6–1.0 × 10¹⁹ cm⁻³, Hall hole mobilities (μ_H) of ≈2.4–3.5 cm²/V s, and bulk resistivities (ρ) of ≈0.73–0.2 Ω cm. We find that p_H increases to 5.0 × 10¹⁹ cm⁻³ and ρ decreases to 0.063 Ω cm for films doped with increasing amounts of Ga. SnO (001) films grown for this study by S-MBE show lower UID p_H in the range of 0.62–2.0 × 10¹⁸ cm⁻³ (Ref. 19) and ρ = 2.6–5.0 Ω cm, p_H increases up to 3.0 × 10¹⁹ cm⁻³, and ρ decreases to 0.08 Ω cm with increasing Ga doping. In contrast, thin films doped with increasing concentrations of La show a reduction in p_H and a remarkable increase in ρ up to 580 Ω cm without transition to n-type conductivity. Samples with higher La doping became semi-insulating and were not measurable in our Hall setup. Compared to previous studies on extrinsic doping in SnO using Ga, Na, Y, Sb, and Ag and their corresponding UID films, summarized in Table I, our Ga-doped SnO shows the highest hole concentration.

Closer inspection of Fig. 3(a), however, suggests that dopant efficiency in SnO is limited for the range of dopants in our study. The reduction in the hole concentration is weaker than the increase in the La (compensating donor) concentration. Likewise, the increase in the hole concentration is well below that of the Ga (acceptor) concentration, and a strong saturation of the hole density is observed with increasing Ga concentration above 7.0 × 10¹⁹ cm⁻³ in both the S-MBE and PA-MBE growths. Below a Ga concentration of 1.2 × 10²¹ cm⁻³, where Ga begins to form a secondary phase, transport data suggest that Ga dopants are not particularly shallow acceptors in SnO and likely not all acceptors can be ionized at RT. The approximate acceptor activation energies are determined from the temperature-dependent Hall measurements as shown in Fig. 4 for S-MBE grown samples. The Arrhenius plot of the hole concentration for UID SnO and SnO with different Ga-doping levels indicates an apparent activation energy of 74 meV for UID SnO grown via S-MBE. The activation energy decreases from 43 to 14 meV with increasing Ga concentrations from 7.0 × 10¹⁹ to 1.1 × 10²⁰ cm⁻³, indicating a likely impurity band formation near the valence band with increasing acceptor doping. To estimate the UID acceptor concentration, the temperature dependent Hall data for the UID sample are extrapolated to T = ∞ (1/T = 0) indicating a UID acceptor concentration of 1.0 × 10¹⁹ cm⁻³.³⁰ The temperature dependent hole mobility in Fig. 4(b) shows strongly increasing mobility with decreasing temperature for UID SnO and Ga-doped SnO with an acceptor concentration of 7.0 × 10¹⁹ cm⁻³, indicating dominant phonon scattering. For samples with a higher Ga doping of 1.1 × 10²⁰ cm⁻³, the influence of scattering due to ionized donors, which is dominant at low temperatures, led to the decrease in the mobility at low temperatures.

To explain the acceptor (donor) doping of SnO (001) with Ga(La), it is proposed that the ionic radii of the elements in different charge states promote the solubility and substitution of Ga¹⁺ (La³⁺) in the Sn²⁺ lattice leading to the observed acceptor(donor) doping behavior.³¹ While Ga prefers the more stable 3+ state, 1+ states have been shown to be obtainable during molecular beam epitaxy by the formation of Ga₂O on the growth front or in the effusion cell.^25,32,33 Interestingly, as shown in Table S1 (supplementary material), the ionic radius of Ga¹⁺ (113 pm) is close to that of Sn²⁺ (118 pm), while Ga³⁺ has a drastically different ionic radius of 62 pm. From a structural perspective, the similarity in the ionic radii of Ga¹⁺ and Sn²⁺ could suggest that the substitutional incorporation of Ga¹⁺ will be favored over Ga³⁺, thus promoting acceptor doping of SnO (001) with Ga_Sn. Also, La³⁺ with an ionic radius of 103 pm should be structurally favored over La¹⁺ with an ionic radius of 139 pm, leading to the observed compensating donor doping in SnO:La.

To further understand the behavior of these and other dopants, their ab initio-calculated formation energies and charge transition levels for Sn-rich and O-rich conditions are shown in Fig. 5. In addition to La and Ga dopants, we further explore the theoretical behavior of other group III species such as In, Al, and B, which can motivate further doping experiments. The results suggest that La_Sn acts as a shallow donor in SnO. However, experimental data indicate that semi-insulating rather than theoretically expected n-type conductive films are obtained as the La concentration exceeds the UID acceptor concentration. This observation suggests that the incorporated La_Sn is being compensated by an increasing acceptor concentration relative to UID samples such as V_Sn or related complexes. This could be a result of the Fermi level position during growth, where higher concentrations of La_Sn donors shift the Fermi level deeper into the bandgap, facilitating an increased incorporation of native acceptor compensators. Figure 5 also indicates that Ga_Sn and In_Sn act most favorably as acceptors. Relative to the valence band maximum (VBM), the calculated ϵ(0/–) transition levels are 0.25 eV for Ga_Sn and 0.13 eV for In_Sn, suggesting In_Sn to be a more effective acceptor than Ga_Sn. While uncertainties on the order of 0.1 eV can be expected for the calculations, Ga_Sn and In_Sn acceptors are likely to exhibit incomplete ionization at room temperature. An additional ϵ(+/0) donor state exists approximately 0.02 eV below the VBM for Ga_Sn and 0.04 eV below the VBM for In_Sn, which suggests that these dopants may be effectively pinning the Fermi level in the vicinity of the VBM through self-compensation. The results are consistent with the experimental observation of saturating hole concentration with increasing Ga doping owing to the fact that it exhibits a localized electronic characteristic. The stabilization of 1+ oxidation states for the Ga dopant is also in contrast to the other group III dopants like Al and B, which we theoretically find to preferentially act as shallow donors. Further measurements are needed to clarify the dominant electronic contributions and ionization energies of these dopants, as well as the role of compensator species that may be simultaneously incorporated to counteract the intended doping.

In conclusion, we have shown in this study that SnO (001) can be controllably doped using La and Ga atoms. X-ray diffraction data show that the solubility limits of Ga and La in SnO are between 6.4 and 12 × 10²⁰ cm⁻³ and 1.4–9.2 × 10²⁰ cm⁻³, respectively. The hole concentration p_H increases to 5.0 × 10¹⁹ cm⁻³, and ρ decreases to 0.063 Ω cm for PA-MBE-grown, Ga-acceptor-doped SnO(001) films. However, SnO (001) films grown via S-MBE show a lower unintentional p_H = 1.2 × 10¹⁸ cm⁻³ and, ρ = 4.9 Ω cm and p_H increases up to 3.0 × 10¹⁹ cm⁻³ and ρ decreases to 0.07 Ω cm with Ga doping. In contrast, thin films doped with higher concentrations of La show a reduction in p_H and a remarkable increase in ρ up to 580 Ω cm without transition to n-type conductivity. Computational results identify that Ga and In preferentially act as deep acceptors with modest ionization energies in SnO, while La, Al, and B act as donors. Our results reveal that p-type conductivity in SnO can be controlled by intentional Ga and La-doping over several orders of magnitude, whereas the successful n-type conduction of SnO remains challenging.

By extending the boundary of conductivity in SnO across highly conductive SnO by Ga doping and semi-insulating SnO by La doping, we demonstrated the material that can be applied in a wide variety of devices: SnO-based p-channel thin film transistors can, thus, be optimized and even tuned from normally on (high hole concentration) to normally off (semi-insulating SnO).^34,35 Highly conductive Ga-doped SnO can be applied in p-type low resistance transparent contacts as well as promote the formation of the depletion zones inside (ultra)wide bandgap n-type oxides in heterojunction p–n diodes and field effect transistor devices.⁵ Finally, La-doping of SnO can lead to highly sensitive thin film conductometric gas sensors by reducing the gas insensitive bulk conductivity of the film,³⁶ which is electrically in parallel to the gas-sensitive, conductive surface.³⁷ 

See the supplementary material for the details of the EDX measurement, AFM images of UID and doped SnO(001) thin films, and a table of the ionic radii of the involved cations.

The authors thank H.-P. Sch$o¨$nherr and C. Hermann for MBE support, G. Hoffmann and A. Ardenghi for useful discussions, and D. Dinh for critically reading the manuscript. This work was performed in the framework of GraFOx, a Leibniz-Science Campus partially funded by the Leibniz association. The work by J.V. was performed under the auspices of the U.S. DOE by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344.