Carrier doping and consequent control of carrier density are essential requirements for semiconductors. Therefore, carrier doping to a known insulating material has been a guiding principle in developing new semiconductors. However, it is difficult to dope carriers to a wide bandgap material, as explained, for example, by the intrinsic doping limit.1 Developing new amorphous semiconductors is another area of interest because amorphous semiconductors can be fabricated even at room temperature on plastic and are applicable to flexible devices. However, obtaining electron conduction in an amorphous material is much more difficult than obtaining electron conduction in a crystalline material. Only hydrogenated amorphous silicon and amorphous oxide semiconductors (AOSs) have been utilized as active layers in practical semiconductor devices.2

Developing an ultra-wide bandgap amorphous semiconductor is very challenging and also interesting from the viewpoint of materials science. It is important also for practical applications because such materials will allow us to integrate power devices with large-size glass and flexible substrates. β-Ga2O3 is a crystalline semiconductor with an ultra-wide bandgap of ~4.9 eV3 and has recently been examined for applications in deep-ultraviolet devices4 and power devices.5 Although carrier doping in β-Ga2O3 has been difficult, as recently explained by theoretical calculations,6 impurity doping using Sn or Si has been discovered to attain electronic conduction,1, 7 which has widened the range of applications of β-Ga2O3, for example, to power devices. However, although a large number of AOS materials with bandgaps of ~3.0 eV, such as a-In-Ga-Zn-Sn-O, have been developed to date,2, 8, 9 semiconducting behavior has never been observed for amorphous Ga2Ox (a-Ga2Ox); only ionic conduction and an insulator-metal transition at a high temperature have been reported.10, 11 A reason for this difficulty is that aliovalent ion doping does not work for amorphous semiconductors except for hydrogenated amorphous silicon; therefore, only metal/oxide ion off-stoichiometry and hydrogen doping would be effective doping methods in AOSs, and consequently carrier control of the conventional AOSs has been conducted mainly by controlling oxygen partial pressure during deposition and thermal annealing (PO2) as well as by hydrogen doping.12 Even these conventional doping methods could not produce semiconducting a-Ga2Ox.

Here we have succeeded in fabricating semiconducting a-Ga2Ox thin films on glass at room temperature. A key strategy is opposite to the strategy for conventional oxide semiconductors where a lower PO2 is chosen to obtain a higher electron density and better conductivity. Another important key is to increase the film growth rate to obtain a higher film density so that formation of electron traps is suppressed. Furthermore, incorporation of weakly bonded or excess oxygen should also be suppressed. Thus semiconducting a-Ga2Ox thin films were obtained with an electron Hall mobility of ~8 cm2 V−1 s−1, a carrier density of ~2 × 1014 cm−3 and a bandgap of ~4.12 eV, which produced operating semiconductor devices including thin-film transistors (TFTs) and Schottky diodes.

Materials and methods

a-Ga2Ox films were fabricated by pulsed laser deposition (PLD) using a KrF excimer laser (wavelength: 248 nm) in an O2 gas flow on silica glass substrates at room temperature. The back pressure for PLD was ~2 × 10−6 Pa. We synthesized polycrystalline targets of β-Ga2O3, (Ga0.35Zn0.65)2Ox and (Ga0.7Zn0.3)2Ox from powdered reagents of ZnO (purity 99.999%) and Ga2O3 (purity 99.99%) by sintering at 1400 °C for 5 h in air. PO2 during deposition and laser power were varied from 0 to 10 Pa and from 30 to 80 mJ, respectively. Some films were subjected to postdeposition thermal annealing at Ta=200–600 °C in vacuum, dry O2 and H2 gas flows.

Film structures and thicknesses were characterized and determined by high-resolution transmission electron microscopy, X-ray diffraction and grazing-incidence X-ray reflectivity spectroscopy, which confirmed that all of the films examined in this study were amorphous (see X-ray diffraction, high-resolution transmission electron microscopy and grazing-incidence X-ray reflectivity in Supplementary Figures S1–S3, respectively). Optical bandgap values (Eg) were estimated by Tauc’ plots for the amorphous films.

Electrical properties were measured by the Hall effect with the van der Pauw configuration. Desorption of the film constituents and impurity-related species was measured by thermal desorption spectroscopy (TDS). For the TDS measurement, ~140-nm-thick a-Ga2Ox films were deposited under different deposition conditions. Then the samples were taken out to the air once from the PLD deposition chamber and then introduced into the TDS measurement chamber. The back pressure of the TDS chamber was ~2 × 10−7 Pa.

Chemical composition (Ga:O ratio) was determined using an electron-probe microanalyzer equipped with a field-emission-type electron gun and a wavelength dispersive X-ray detector. Hard X-ray photoemission spectroscopy (HAXPES) measurements at room temperature were performed using the BL15XU undulator beamline (the excitation X-ray energy: =5953.4 eV)13, 14 of SPring-8. The binding energy was calibrated with the EF of an evaporated Au thin film, and the total energy resolution was set to 240 meV, which was confirmed by the Fermi cutoff of the Au film. The energy levels of the conduction band minimum (CBM, ECBM) and the valence band maximum (VBM, EVBM) from the vacuum level (EVAC) (EVACECBM and EVACEVBM are electron affinity, χ, and ionization potential, Ip, respectively) were measured by ultraviolet photoemission spectroscopy (UPS) (excited by He I and II light sources). To prepare chemically pure surfaces for the UPS measurements, Ar ion sputtering was conducted for 1 h at an acceleration voltage of 1 kV. Work function was determined from the cutoff energy of secondary electrons, and Ip was estimated by combining EVAC and the measured EVBM (see Supplementary Figure S4). χ was speculated using the measured optical bandgap by χ=Ip+Eg.

Bottom-gate, top-contact TFTs were fabricated using semiconducting a-Ga2Ox channels on SiO2/n+-Si substrates (see Figure 5a for the device structure). The a-Ga2Ox layers were deposited at room temperature, followed by thermal annealing at 200 °C under vacuum. Finally, aluminum source/drain contacts were deposited by thermal evaporation. A-Ga2Ox/Pt Schottky diodes were fabricated on Pt/silica glass substrates, finished by evaporating aluminum top Ohmic contacts (Figure 5c).

Results and discussion

Controlling factor to obtain semiconducting a-Ga2Ox

In the beginning of this work, we employed film growth conditions that were similar to device-quality a-In-Ga-Zn-O (a-IGZO) for PLD (that is, PO2 of 1–5 Pa, target-substrate (TS) distance of 50 mm and laser power of 50 mJ) because these a-IGZO films produce good TFTs and have low defect densities even when deposited at room temperature on glass.15, 16 However, all of the a-Ga2Ox films thus fabricated were insulators with electrical conductivities below the measurement limit of our apparatus (<10−8 S cm−1), although we had examined wider deposition conditions from PO2=0 Pa and performed several postdeposition treatments such as thermal annealing in vacuum, O2, and H2 because these are standard procedures to dope electrons onto oxide semiconductors.

By investigating further wider deposition conditions including the TS distance and the laser power, we finally found the deposition conditions to obtain electron conduction as shown in Figure 1. Electrical conductivities were obtained only if the film density was >5.2 g cm−3 (Figures 1b and c), where the electrical conductivity increased exponentially with the increase in the film density. Figure 1c, in which the films exhibiting electrical conductivity are shown by the orange shaded region, shows that the TS distance and laser power are not the essential controlling factors because the smallest TS distance of 30 mm produced conducting films independent of the laser power, but the larger TS distances failed to attain conductivity, particularly for low laser power. Replotting these data with respect to the deposition rate (Figure 1d) gives a single universal curve, indicating that the film density is the essential controlling factor. The deposition rates for device-quality a-IGZO films are ~5 nm min−1, and the deposition rates required for semiconducting a-Ga2Ox films are larger by one to several orders of magnitude.

Figure 1
figure 1

Relation between electronic conduction and film density. (a) Deposition parameters affecting the film structures and properties such as (i) PO2, (ii) laser power and (iii) TS distance. (b) Electrical conductivity vs film density, indicating that film densities 5.2 g cm−3 are required to obtain semiconducting films. (c) Film density vs laser power and TS distance. (d) Replot of panel (b) with respect to deposition rate, giving a single universal curve between deposition rate and film density. (e) PO2 dependence of conductivity, showing that too low (<1 Pa) and too high PO2 (>6 Pa) produced insulating films only.

PO2 is also an essential controlling parameter to attain conducting films. We should, however, emphasize that the optimum PO2 obtained for a-Ga2Ox is very different from the value for conventional oxide semiconductors for which a lower PO2 is required to obtain a higher electronic conduction and a higher free electron density. That is, low PO2 including zero (that is, in vacuum) produced insulating a-Ga2Ox films only, and the best conductivity of ~2 × 10−4 S cm−1 was obtained at a rather high PO2=5 Pa. This value can be explained from the subgap defects observed in optical absorption spectra (Figure 2a) and HAXPES spectra around the bandgap region (Figure 2b). Similar to previous reports,10, 17 the low PO2 films seem to have small apparent Eg due to subgap absorption, but the actual Eg is estimated to be 4.12 eV from the Tauc’ plot of the PO2=6 Pa film because it has the smallest subgap absorption. Strong subgap absorption extends from Eg=4.12 to 0.5 eV for the PO2=0 film. The subgap absorption originates from the subgap defects just above the VBM (near-VBM states) as seen in the HAXPES spectra in Figure 2b, where the VBM level is estimated to be ~4.1 eV from the Fermi level (EF) by extrapolating the straight VB tail to zero. The film deposited at a low PO2=1 Pa (the film density of 5.27 g cm−3) shows high-density peak-shape near-VBM states (denoted ‘VO defect’ for the black curve), while the film deposited at the optimum PO2=6 Pa (5.39 g cm−3) has much smaller near-VBM states (the red curve). These results are similar to previously reported a-IGZO cases and explain that a low PO2 condition produces electron traps, probably due to oxygen deficiency with free space (void) structures.18 This result explains the PO2 vs conductivity result in Figure 1e because lower PO2 may generate free electrons due to oxygen deficiency as usual for oxide semiconductors, while it also produces the near-VBM defects and these defects trap all the generated electrons, leading to the stronger charge compensation at lower PO2 and consequently to the insulating films.

Figure 2
figure 2

Subgap defects in a-Ga2Ox. (a) Optical absorption spectra as a function of PO2 from 0 to 6 Pa. (b) HAXPES spectra around the bandgap region for conducting and insulating films deposited at PO2=6 Pa (the red curve) and 1 Pa (the black curve), respectively. The binding energy is measured from the Fermi level, and the VBMs are located at ~4.1 eV. (c) Transmittance+reflectance spectra and photo of the conducting a-Ga2Ox film of ~120 nm thickness.

The next question is why the film deposition rate determines the film density as seen in Figure 1d. TDS spectra in Figures 3a and b show that the film density is altered largely by the TS distance and PO2. The low-density film deposited at a large TS distance of 50 mm exhibited high-density desorption of H2O molecules even from the very low temperature of ~80 °C, implying that lower density films incorporate more impurities such as H2O, OH and H-related molecules. This result suggests that the low density would come from incorporation of residual H2O and H-related molecules in the PLD chamber because a lower growth rate requires a longer deposition time and incorporates more impurity from the deposition atmosphere into the growing film as illustrated in Figure 3. There is another important factor to determine the film density; a higher deposition rate condition would produce deposition precursors with higher kinetic energies, which enhances structural relaxation in the growing surface region and assists in forming a denser structure.

Figure 3
figure 3

Schematic illustration of the effects of deposition parameters (TS distance and PO2) on incorporation of impurity species. (a, b) TDS spectra of (a) H2O (TS distances are 50 mm for blue and 35 mm for purple) and (b) O2 (PO2 are 10 Pa for green and 5 Pa for purple). The TS distance of 35 mm and the PO2=5 Pa are the optimum conditions and produced the high-density semiconducting a-Ga2Ox film.

Figure 1e shows that the optimum PO2 were 5–6 Pa, but the conductivity dropped sharply by further increasing PO2, and no conducting film was obtained at >6 Pa. The O2 TDS spectra in Figure 3b show that the film deposited at the optimum condition of PO2=5 Pa did not show O2 desorption up to 350 °C (note that this film is of high density), while the film deposited at a high PO2=10 Pa (a low-density film) exhibited a large amount of O2 desorption even from the low temperature of 80 °C. A similar result is observed in a-IGZO films fabricated under strong oxidation conditions, and similar low-temperature O2 desorption is attributed to weakly bonded oxygen that forms an electron trap.19, 20, 21 Similar to the a-IGZO case, the insulating behavior of the high PO2 a-Ga2Ox films can be explained by the charge compensation due to the electron traps originating from the weakly bonded oxygen.

Electronic structure of a-Ga2Ox

Next, we investigated the electronic structure of a-Ga2Ox. To systematically investigate the variation of electronic structure, we compared the electronic structure of a-Ga2Ox with the electronic structure of its solid-solution system a-(Ga1-yZny)2Ox (y=0, 0.30, 0.65). Their electrical properties, the temperature dependence of electrical conductivity (σe), electron Hall mobility (μe) and free electron density (ne) are summarized in Supplementary Figure S5 and Supplementary Table S1, which substantiated that all of the films were n-type semiconductors and that their donor levels measured from the CBM deepen with the decreasing Zn content from ECBMED=0.08 to 0.32 and to 0.68 eV for a-(Ga0.35Zn0.65)2Ox to a-(Ga0.70Zn0.30)2Ox and to a-Ga2Ox, respectively. The total donor densities n0 were also estimated from the equation ne=n0 exp(−(ECBMED)/(kBT)), providing n0=1.0 × 1018, 1.5 × 1018 and 7.5 × 1019 cm−3 for a-(Ga0.35Zn0.65)2Ox, a-(Ga0.70Zn0.30)2Ox and a-Ga2Ox, respectively. These densities indicate that although n0 is the largest for a-Ga2Ox, ne is limited due to the deep donor level, ECBMED. All of the a-(Ga1-yZny)Ox films exhibited rather large electron mobility for amorphous semiconductors, 6–8 cm2 V−1 s−1 at room temperature, similar to a-IGZO.15, 16 This result substantiates that the low value for σe of a-Ga2Ox of ~2 × 10−4 S cm−1 at most originates from the low ne, not from μe. This origin would be reasonable because both a-IGZO and a-GaOx have similar CBM structures made mainly of spherical metal s orbitals, which can form a highly dispersed conduction band with small electron effective masses even in a disordered amorphous structure.22

The Eg values, which were determined from Tauc’ plots (Supplementary Figure S4d), were 4.12, 3.78 and 3.47 eV for a-(Ga0.35Zn0.65)2Ox, a-(Ga0.70Zn0.30)2Ox and a-Ga2Ox, respectively, and the ionization potential (IP=EVBM) values were obtained by UPS measurements (Supplementary Figure S2). Consequently, the band alignment diagram is built as shown in Figure 4a. The diagram shows that EVBM deepens as the Zn content decreases while ECBM remains almost unchanged, indicating that the difference in Eg comes mostly from the deepening of EVBM. This result would be unexpected because one may expect that EVBM values of oxide semiconductors are similar because their VBM levels are mainly made of the common chemical states, O 2p–O 2p antibonding states, and consequently, it would be natural to expect that the Ga3+ 4s level, which constitutes the CBM in Ga2Ox, lies at a shallower energy than Zn2+ 4s as speculated from the difference in Eg. However, this result indicates that the energy levels of the unoccupied s orbitals in Ga3+ and Zn2+ are similar, while Eg is determined mainly by EVBM. A similar result is also reported for crystalline β-Ga2O3.24 This trend is explained by an O 2p–Zn 3d interaction as reported for II–VI semiconductors;25 that is, the Zn 3d level is shallower than Ga 3d and has a larger antibonding coupling with O 2p states, which raises EVBM as the Zn content increases. The doping limit in many n-type oxides is determined by their ECBM measured from the vacuum level.26, 27 Thus β-Ga2O3 and a-Ga2Ox can be doped to n-type because their ECBM values are close to that of ZnO and within this doping limit. We also see an interesting relation between ED and EVBM in Figure 4a. In the a-(Ga1-yZny)Ox case, the ED level deepens as the Zn content increases (that is, ECBMED=0.08, 0.32 and 0.68 eV for y=0.65, 0.30 and 0, respectively) and appears to have a stronger correlation with EVBM than with ECBM; that is, the EDEVBM are almost the same at 7.45, 7.49 and 7.13 eV.

Figure 4
figure 4

(a) Band alignment diagram of a-(Ga1-yZny)2Ox in comparison with crystalline ZnO built from Eg and UPS results in Supplementary Figure S4 of the supporting information. The values for ZnO and β-Ga2O3 are taken from Klein23 and Mohamed and Schottky24, respectively. The ED value for β-Ga2O3 is taken from Varley et al.6 EVBM (ionization potential, IP) is shown by the top edges of the blue bars and ECBM (electron affinity, χ) by the bottom edges of the purple bars. ED levels are also shown by the horizontal bars with spheres. (b) The schematic diagram shows routes to convert an insulator to a semiconductor.

Here we discuss why the semiconductor conversion of a-Ga2Ox was so difficult. As schematically shown in Figure 4b, good transparent conducting oxides such as Al-doped ZnO (ZnO:Al) have a high free electron concentration, for example, at >1020 cm−3, generated from a shallow donor level of AlZn, while the electron trap density (Dt) would be negligible compared to the free electron concentration. However, the blue region in Figure 4b has high-density electron traps and/or low-density free electrons generated from the donors, producing electrical insulators. Ultra-wide bandgap amorphous materials can attain electronic conduction only with difficulty because the ultra-wide gap leads to low Ne and the amorphous structure leads to high Dt. As seen in Figure 4a, the donor level in a-Ga2Ox is very deep at 0.68 eV and gives Ne as small as 1015 cm−3; therefore, very small Dt>1015 cm−3 are critical, and only insulating a-Ga2Ox films had been obtained to date. Consequently, as shown by ‘Route(ii)’, the semiconducting a-Ga2Ox films here are obtained by suppressing Dt.

Applications to semiconductor devices

The potential of the semiconducting a-Ga2Ox films was demonstrated by fabricating TFTs (Figure 5a) and Schottky diodes (Figure 5c). The a-Ga2Ox TFT exhibited a reasonable performance such as a saturation mobility of ~1.5 cm2 V−1 s−1, an on/off ratio of >107 and a subthreshold swing voltage of ~0.4 V dec−1 (Figure 5b). The a-Ga2Ox/Pt Schottky diode also shows a good current–voltage characteristic with a very high on/off ratio (>109) with a negligible hysteresis (Figure 5d). Although the a-Ga2Ox/Pt Schottky junction was reported by Aoki,11 it operated via ion conduction in a-Ga2Ox.

Figure 5
figure 5

a-Ga2Ox devices. (a) TFT structure and (b) transfer characteristic. (c) Schottly diode structure and (d) current–voltage characteristic.


Semiconducting a-Ga2Ox thin films with a bandgap of 4.12 eV were successfully fabricated on glass at room temperature, and the electronic structure was clarified in comparison with its Zn solid solutions. The keys to obtaining electronic conduction in a-Ga2Ox are (i) a high film density and (ii) an appropriately high PO2 to suppress formation of electron traps. As the donor level is very deep, 0.68 eV from ECBM, for a-Ga2Ox, the free electron density is still low, 2 × 1014 cm−3 at best, although the total donor density is rather high, 7.5 × 1019 cm−3. The formation of a trace amount of electron traps at an order of 1014 cm−3 is therefore critical to fully compensate for the generated free carriers, making a sharp contrast with conventional oxide semiconductors. Those conventional oxides have shallow donor levels and high-density free electrons, and therefore, the effect of charge-compensating electron traps is invisible and low PO2 deposition conditions effectively produce high carrier density, high conductivity films. The present result shows that this strategy for conventional oxide semiconductors should be reconsidered for developing new semiconductor materials that have very wide bandgaps, deep donor/acceptor levels and/or other extreme conditions/requirements. Seeking a deposition/annealing condition/method to suppress charge compensating defects should also be considered.

This discovery, that is, the importance of controlling charge compensation, will lead to developing further new semiconductors as well as to improving carrier transport properties and device performance of known oxide semiconductors in which generating high-density carriers or obtaining high-performance devices is difficult, such as p-type oxide semiconductors.