Electronic and optical properties of Zn-doped β-Ga₂O₃ Czochralski single crystals

β-Ga₂O₃ is a monoclinic tunable ultra-wide bandgap (4.5–4.9 eV) transparent semiconducting oxide single crystalline material with applications including gas sensors, Schottky diodes, high breakdown voltage switching devices, and photodetectors for the deep ultraviolet.^1–7 Rapid development has been possible due to single crystalline growth methods such as Czochralski (CZ), edge-defined film-fed growth (EFG), vertical gradient freeze (VGF), float zone (FZ), and vertical Bridgman (VB).^8–12 β-Ga₂O₃ demonstrates tunable conductivity from insulating to n-type conduction depending on the dopant or impurities.⁸ p-type β-Ga₂O₃ is difficult to realize as-grown due to low mobility self-trapped holes; however, recent work has demonstrated p-type doping via hydrogen incorporation.^13–18 Many possible p-type dopants have also been shown to be too deep for hole conduction in β-Ga₂O₃.^{14,15,19–21} Dopants such as Zr, Hf, Nb, Si, and Sn cause n-type behavior, with free carrier concentrations up to ∼1 × 10¹⁹ cm⁻³.^11,22–25 Considerable difficulties have been observed when growing β-Ga₂O₃ via CZ with a dopant that causes a high free carrier concentration, resulting in crystals with a higher tendency to grow spirally.^8,26,27

Semi-insulating and insulating behavior can be obtained by doping with Fe, Mg, Co, Ni, or Al.^28,29 Impurity-driven electrical behavior typically results in “unintentionally doped” (UID) β-Ga₂O₃ with Si contributing to n, and Fe slightly electrically compensating, typically with n on the order of ∼10¹⁷ cm⁻³.^16,30–32 Fe impurities exist in the Ga₂O₃ stock precursor powder and are introduced from crucibles used during the crystal growth, as Fe is a common impurity in the manufacturing process of Ir crucibles.

Another possible acceptor for β-Ga₂O₃ that has thus far been unreported in single crystals is zinc. Zn:β-Ga₂O₃ has been grown in thin films, however, and in various nano- and microstructures for optoelectronic applications.^33–35 The formation of the Zn dopant on Ga sites in Zn:β-Ga₂O₃ has been computationally studied, suggesting the formation of Zn on the tetrahedral site.³⁶ This is opposed to the common deep acceptor dopant, Mg, which has been shown to reside on the octahedral site.³⁷ Compounds of ZnO and Ga₂O₃ have also been reported. (100)-oriented ZnGa₂O₄ single crystal spinels have been grown via CZ resulting in a free carrier concentration of 3 × 10¹⁸–9 × 10¹⁹ cm⁻³ and are tunable to insulating through annealing in oxygen similar to β-Ga₂O₃; these crystals also have an ultra-wide bandgap of about 4.6 eV.³⁸ During ZnGa₂O₄ growths, Zn and Ga evaporation causes non-stoichiometry in the melt and CZ pulled crystal, leading toward a Ga-rich composition under an O₂ environment as Zn losses are higher than Ga losses.³⁸ 

In this work, we report the high quality bulk cylindrical growth of Zn:β-Ga₂O₃ (0.25 at. % Zn in the batch) with minimal Zn losses, resulting in an insulating n-type material due to Zn deep acceptors. We also report computational results suggesting Zn forms on both tetrahedral and octahedral sites, but slightly prefer the octahedral site.

Zn-doped β-Ga₂O₃ non-spiral single crystals (0.25 at. % Zn) were grown from the melt by similar methods to those previously published.^11,22 High purity (5N = 99.999%) Ga₂O₃ powder (ABSCO Limited, Haverhill, Suffolk, UK) was used as the source material. ZnO powder (3N purity, Sigma Aldrich) was added to achieve 0.25 at. % cation doping, and the powder was mixed for 18 h in a tungsten carbide ball mill at 50 rpm. Two charges of 400–450 g each were prepared, cold pressed at 140 MPa, and calcined in an alumina crucible with Pt foil lining at 1500 °C in air for 15 h. The first charge was melted and cooled, and the second charge was added to increase molten material volume. Crystals were subsequently grown in a 70 mm height × 70 mm diameter iridium crucible heated by a 25 kHz radio frequency inductive heating coil. A mixed Ar + O₂ gas was used during the melting and growth with varying O₂ depending on the growth stage.³⁹ Below 1200 °C, O₂ was <0.2% to reduce solid IrO₂ formation. Between 1200 °C and melting of Ga₂O₃, the fraction of O₂ in the furnace was ∼2.5%–3.5% to reduce evaporation of Ga and Zn, and during the growth of Zn:β-Ga₂O₃ crystals O₂, it was ∼10–11%. This gas flow scheme was also maintained during cooling of the crystal after growth.

At 1830 °C, as measured by the two-channel Ircon Modline pyrometer, Zn:β-Ga₂O₃ crystals were grown both by the CZ method with a 2 mm/h pull rate and 2 rpm rotation and by VGF with no rotation upon cooling at 1–2 °C min⁻¹. More precise measurement of melt–surface temperature was accomplished with a sapphire light-pipe inside an alumina sheath located ∼2 cm from the melt and measured with a Sekidenko 1000F thermometer. No longer than 8 h was spent at growth temperatures of ∼1800 °C. Non-spiral CZ crystals were achieved, with a diameter of 36.5 mm and a cylindrical height of 17.6 mm or a height of 28.9 mm from tip to heel, on a 0.25 at. % Mg-doped β-Ga₂O₃ seed. Spiral growth was not observed, and the melt–crystal interface remained reliably stable throughout.

Crystals of size 0.25 × 0.25 mm² to 1 × 1 cm² of (100)-orientation were obtained from both the CZ pulled crystal and the VGF via cleaving along the two cleavage planes and cutting with a wire saw and inner diameter (ID) saw. Sample thickness was 0.1–8 mm depending on the needs of the characterization. Sample color, depending on thickness, ranged from colorless to yellow. Sample location within the VGF crystal affected sample color, possibly indicating a dopant segregation of Zn possibly due to evaporation. Surface material was clear to yellow in color; as depth into the crucible increased, the material was yellow to brown in color. The pulled crystal was homogenous with respect to color and sample morphology.

Sample orientation was probed with a Raman spectrometer (Thermo Fisher DXR2XI) with a 785 nm laser and a high resolution grating with a Raman shift range of 50–1800 cm⁻¹ and 1 cm⁻¹ resolution. Confirmation of crystal orientation was confirmed by an electron backscatter diffraction (EBSD) map. EBSD patterns were collected using an FEI SIRION 6400 scanning electron microscope (SEM) and indexed using orientation imaging microscopy (OIM) analysis software.

Sample purity and doping level were assessed with glow discharge mass spectrometry (GDMS) and secondary ion mass spectrometry (SIMS), respectively. GDMS was performed at EAG Laboratory (California, USA) on Zn:β-Ga₂O₃ VGF single crystals after crushing with a tungsten carbide mill. The results summarized in Table S1 in the supplementary material show elements that were present at concentrations above the detection limit. Fe concentration is as expected (1.6 × 10¹⁷ atoms/cm³) and along with Mg (1.2 × 10¹⁷ atoms/cm³) is too low to electrically compensate for Si impurities (5.9 × 10¹⁸ atoms/cm³). As-batched Zn incorporation is estimated ≈1.5 × 10¹⁸ atoms/cm³, which GDMS and SIMS both verify (1–4.8) × 10¹⁸ atoms/cm³. SIMS was performed on a single crystal of Zn:β-Ga₂O₃ from the CZ pulled crystal at EAG Laboratory (California, USA). The results shown in Fig. S1 in the supplementary material verify Zn concentration on the order of 10¹⁸ atoms/cm³.

Optical transmission measurements were performed with an ultraviolet–visible–near infrared spectrometer (Cary 5 UV-Vis-NIR) and a Fourier transform infrared spectrometer (Bruker Alpha FTIR) at room temperature for wavelengths 200–3000 nm (50 000–3333 cm⁻¹) and 1.3–25.0 μm (7500–400 cm⁻¹) on windows of Zn:β-Ga₂O₃ with thickness 0.1–7.9 mm. The spectra were analyzed with respect to absorptions due to the electronic band edge, free carriers, impurities, and multiphonon processes.

For all electrical measurements, ohmic contacts were placed on both sides of the (100) plane using 50–50 wt. % Ga–In in a two-point configuration. These samples were then annealed at 950 °C for 15 min and a small amount of Ga–In was placed on top of the contact again. These contacts have been shown to be ohmic at temperatures as low as ∼20 K.¹¹ Temperature dependent Hall effect measurements using 0.51 T magnetic field were attempted between 80 and 320 K with 10 K temperature step (Ecopia HMS 7000) with contacts in van der Pauw configuration; however, sample resistivity was so high that Hall effect could not obtain reliable data on free carrier concentration. Therefore, in order to assess resistivity, two-point through thickness I–V measurements and resistance measurements were obtained with a high-impedance picoammeter.

The analysis of Zn dopant solubility, preferential site occupation, and electronic behavior was evaluated from their defect formation energies (E^f) as calculated within a supercell approach.⁴⁰ This approach enables a quantitative assessment of the expected dopant incorporation as a function of growth conditions and the energy positions of possible defect levels introduced within the bandgap. All E^f were calculated using the Heyd–Scuseria–Ernzhof screened hybrid functional (HSE06)⁴¹ and projector-augmented wave (PAW) approach⁴² as implemented in the VASP code.^43,44 We adopt supercells with 160-atoms (a 3 × 4 × 1 repetition of the 20-atom unit cell) with the same computational approach and as detailed in other previous publications.^11,22,45 We additionally account for the effects of limiting phases in the calculated chemical potential of Zn dopants as a function of conditions, finding ZnGa₂O₄ (calculated ΔH[ZnGa₂O₄] = –13.92 eV/formula unit), the solubility-limiting phase in both the O-rich and Ga-rich limits.

Zn:β-Ga₂O₃ crystals grown via CZ were cylindrical (Fig. 1) and displayed excellent stability during growth, with no tendency to spiral or grow offset. A pull rate of 2 mm/h was used for the majority of the growth except for the last hour, where a 4–5 mm/h pull rate was used. Even when the growth rate increased, the melt–surface interface remained stable, evident by good control over the growth rate at a given power and constant cylindrical growth. In previous attempts with UID or high free carrier concentration β-Ga₂O₃, increasing the growth rate drastically increased the likelihood of spiral growth; however, due to low free carrier concentration here, faster pull rates appear feasible, which reduces Ir losses from the crucible and reduces residence time at high temperature, which tends to leach impurities from the crucible into the melt during growth.

Raman microscopy and EBSD were applied to study the orientation of the samples, due to the atypical diagonal orientation of the cleaved surface shown in Fig. 1(b). Raman microscopy shows orientation related anisotropy in several peak intensities in β-Ga₂O₃ regardless of dopant, especially noticeable in the A_g(10) mode.⁴⁶ Shown in Fig. 2(a), Raman spectra were collected on two transparent cleavable surfaces. Utilizing the A_g(10) peak intensity, the (100) and (001) surfaces can be identified. The orientations of our sample surfaces measured in this work were confirmed to be (100) through evaluating A_g(10) intensity and by EBSD shown in Fig. 2(b). Due to charging issues with the highly insulating sample, obtaining a clear diffraction pattern was sometimes difficult, resulting in poor indexing evident by noise in the EBSD map.

Shown in Fig. 3, optical absorption measurements show no appreciable change in the optical bandgap of the material as compared to UID material, as expected with doping on the order of 0.25 at. %. Free carrier absorption is suppressed in the near infrared region of the material compared to UID β-Ga₂O₃ grown by the same methods. A shoulder near the band edge is observed in both UID and Zn:β-Ga₂O₃. This anisotropic^47–49 shoulder is only visible in the (100) direction due to dipole forbidden transitions in other directions and is indicative of a sub-bandgap absorption, with evidence pointing toward free carrier related absorption located 0.23 eV below the conduction band.^50–52 Similar to Mg:β-Ga₂O₃, samples with thickness >1–2 mm have enough absorption to detect several more features (Fig. 4). Absorption at 1940 nm (5155 cm⁻¹) may be assigned to Ir⁴⁺ on octahedral Ga sites, Ir⁴⁺–O stretches (present in this oxidation state due to the Zn acceptor),⁵³ and sidebands are likely due to transitions perturbed by nearby Zn atoms, similar to that reported for Mg-doped crystals.⁵⁴ Several near-edge band features [Fig. 4(b)] are attributed to Ir³⁺ and Ir^4+/3+ transitions, including those observed ∼424 nm (2.9 eV) and ∼352 nm (3.5 eV), though the feature ∼280 nm (4.4 eV) remains unidentified, while the band edge for the thick sample is ∼267 nm (4.64 eV).⁵³ Other absorptions 2800–3500 cm⁻¹ are thought to be due to hydrogen-related complexes and will be the subject of a later study. After the transparency window, absorption features 4.5–6.5 μm (2200–1500 cm⁻¹) are likely due to multiphonon absorption.⁵⁵ 

Bulk Zn:β-Ga₂O₃ shows no promise for p-type conductivity due to the nature of the Zn dopant as a deep acceptor. There is no appreciable hole mobility that can be measured by the Hall effect. Room temperature high-impedance picoammeter measurements of CZ Zn-doped β-Ga₂O₃ samples show a resistivity of 4–8 × 10¹¹ Ω cm. I–V measurements (Fig. S2 in the supplementary material) of CZ Zn-doped β-Ga₂O₃ show a resistivity of 4 × 10¹¹–6 × 10¹³ Ω cm, while those of VGF Zn-doped β-Ga₂O₃ result in a resistivity of 4 × 10¹¹–1 × 10¹⁴ Ω cm. The variation of resistivity reported is not ascribed wholly to a dopant and impurity (Si,Fe) segregation within the material, but rather uncertainty within the I–V measurement, which are large for high resistivity samples, especially with two-point contacts.^56,57 In fact, the resistivity within in the VGF or CZ boules were all within the 10¹¹–10¹⁴ range, indicating consistent Zn concentration above that of impurities imparting n-type conduction. Table I shows measured values from multiple samples.

To gain additional insight into the effects of Zn incorporation in the lattice, we performed hybrid functional calculations of Zn in β-Ga₂O₃. We include the calculated formation energies for substitutional Zn dopants on both the tetrahedral (Zn_GaI) and octahedral (Zn_GaII) sites in Fig. 5, shown for different growth extremes of the Ga-rich and O-rich limits, and including the effects of ZnGa₂O₄ as a solubility-limiting phase for Zn. Interestingly, we find that Zn exhibits a nearly equivalent preference for incorporation on the tetrahedral and octahedral sites, with the octahedral site favored for the acceptor state by only 0.06 eV. This is in contrast to previous reports that find the tetrahedral site the preferred site.¹⁵ We find that this is a consequence of the treatment of the Ga 3d states in the calculations, with their omission as valence states influencing the relative energetic preference by 0.15 eV. This effect is also found for similar-sized Ge dopants, which also exhibit a nearly equivalent preference to the octahedral and tetrahedral sites.³⁰ The low formation energies of Zn_Ga acceptors support that it may be readily incorporated on both sites for a range of growth conditions, particularly in n-type samples where it will drive the Fermi level from the conduction band toward midgap. This is consistent with both the high concentrations of incorporated Zn dopants in excess of 1 × 10¹⁸ atoms/cm³ and with the increased resistivity.

Previous reports have identified that Zn_Ga behaves as a deep acceptor with the possibility of trapping an additional hole to also form a donor state.^14,15,19 Our results in Fig. 5 support this, finding the deep acceptor (0/–) transition 1.24 eV (1.34 eV) above the VBM for the Zn_GaI (Zn_GaII) to be in good agreement with previous reports. We find the hole associated with the neutral state preferentially localizes on an adjacent OI oxygen site for both Zn_GaI and Zn_GaII, similar to Mg acceptors. However, the expected incorporation of Zn on both sites is in contrast with Mg that strongly prefers to occupy only the octahedral site.⁵³ The calculated second hole for the donor state is expected to trap for Fermi levels 0.85 eV (0.87 eV) above the VBM for the Zn_GaI (Zn_GaII), results, which are also in good agreement with previous reports of ∼0.9 eV.^14,15,19 We find that the second hole preferentially traps on the OI oxygen site for Zn_GaII, while it traps on an adjacent OII oxygen site for Zn_GaI, as shown in Figs. 5(c) and 5(d). The different environment of holes with Zn and Mg was previously detailed in Ref. 36 and may better facilitate the identification of the Zn site incorporation though measurements like photo-excitation electron paramagnetic resonance (photo-EPR) or optically detected magnetic resonance (ODMR), in comparison to other measurements like photoluminescence that would likely show broad overlapping peaks for both sites.⁵⁸ Furthermore, experimental EPR results have been recently conducted on Zn-doped material grown at WSU.⁵⁹ 

We have shown that Zn:β-Ga₂O₃ is a highly insulating material, with resistivity on the order of 10¹¹–10¹³ Ω cm. While Zn evaporation has been an issue in growth of ZnGa₂O₄ spinel, Zn in β-Ga₂O₃ is incorporated at a concentration (10¹⁸ atoms/cm⁻³) capable of compensating for impurities like silicon that typically cause n-type conduction. Computational results show Zn substitutes on both octahedral and tetrahedral sites but with a slight preference for the octahedral site. The results support previous conclusions that Zn acts as a deep acceptor on both sites and can exhibit multiple hole-trapping that may help facilitate experimental identification and validation of the local environment of incorporated Zn_Ga.

See the supplementary material for crystal purity data by GDMS and SIMS as well as the I–V test curve.

This material is based on work supported by the Air Force Office of Scientific Research under Award No. FA9550-18-1-0507 monitored by Dr. Ali Sayir. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the United States Air Force. We thank Colin Merriman for generating the diffraction library to complete the EBSD mapping and Parker Toews for crystal growth support. This work was partially performed under the auspices of the U.S. DOE by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344 and supported by the Critical Materials Institute, an Energy Innovation Hub funded by the U.S. DOE, Office of Energy Efficiency and Renewable Energy, Advanced Manufacturing Office.

The data that support the findings of this study are available from the corresponding author upon reasonable request.