A. ZnO

ZnO has the wurtzite crystal structure and is relatively easy to grow in bulk or thin-film form.Reference Özgür, Alivov, Liu, Teke, Reshchikov, Doğan, Avrutin, Cho and Morkoç^33, Reference Pearton, Norton, Ip, Heo and Steiner³⁴ As noted in Sec. I, ZnO is an alternative to ITO for transparent contacts in solar cellsReference Nuruddin and Abelson³⁵ and transparent electronics.Reference Wager³⁶ It is also an efficient light emitter, with a band gap of ∼3.4 eV at room temperature, making it a candidate for solid-state lighting. The past decade of research has uncovered many interesting defect properties, most of which are also relevant to other oxides.

The status of available information on defects in ZnO was reviewed in 2009 by McCluskey and Jokela.Reference McCluskey and Jokela³⁷ As discussed in their article under reference, numerous calculations have revealed that oxygen vacancies are deep double donors (e.g., Refs. Reference Kohan, Ceder, Morgan and Van de Walle38–Reference Paudel and Lambrecht41). The oxygen vacancy is an example of a “negative U” defect, in which the neutral or +2 charge state is always energetically preferred over the +1 charge state. When the Fermi level is above the (0/2+) donor level, the oxygen vacancy is occupied and neutral. When the Fermi level is below the (0/2+) level, the donor is empty and has a charge of +2. The (0/2+) level is estimated to lie 1–2 eV below the conduction band minimum. At thermal equilibrium, such a deep level contributes essentially zero free electrons at room temperature. Electron paramagnetic resonance (EPR) experiments,Reference Vlasenko and Watkins⁴² combined with illumination by visible light,Reference Evans, Giles, Halliburton and Kappers⁴³ have supported the theoretical claim that oxygen vacancies are deep donors.

Another native defect, the zinc interstitial, has also been proposed as a source of n-type conductivity in ZnO. Whereas it is probably indeed a shallow donor, calculations have shown that the zinc interstitial has a low migration barrier.Reference Janotti and Van de Walle⁴⁴ At room temperature, it will simply migrate to the surface or become trapped by a defect. This is in agreement with EPR experiments that show zinc interstitials to be unstable at room temperature.Reference Vlasenko and Watkins⁴²

With native defects out of the picture, hydrogen is considered to be an important contributor to n-type conductivity. There are two forms in which hydrogen acts as a shallow donor, interstitial and substitutional hydrogen. Interstitial hydrogen forms an O–H bond that gives rise to a bond-stretching peak in the IR absorption spectrum. There are two different interstitial hydrogen donors, with peaks at 3326 cm⁻¹ (McCluskey et al.Reference McCluskey, Jokela, Zhuravlev, Simpson and Lynn⁴⁵) and 3611 cm⁻¹ (Lavrov et al.Reference Lavrov, Weber, Börrnert, Van de Walle and Helbig⁴⁶). Whereas it has been proposed that the 3326 cm⁻¹ line may be an O–H pair adjacent to a calcium impurity,Reference McCluskey and Jokela^47–Reference Bang, Choi and Chang⁴⁹ this model has not been established experimentally. Both interstitial hydrogen donors are unstable at room temperature, decaying away after a few weeks (Fig. 2).Reference Jokela and McCluskey⁵⁰ Some of the hydrogen atoms form H₂ molecules, which are electrically neutral and “hidden” from IR spectroscopy.Reference Shi, Saboktakin, Stavola and Pearton^51, Reference Lavrov, Herklotz and Weber⁵² Since interstitial donors are unstable, they are not responsible for the prevailing n-type conductivity in ZnO samples.

FIG. 2.

Free-electron density and O–H peak area for interstitial hydrogen donors in ZnO, at room temperature. The decay over time shows that interstitial donors are unstable.Reference Jokela and McCluskey⁵⁰ Inset: O–H absorption peak at 8 K and 300 K.

Substitutional hydrogen, in contrast, is believed to be a shallow donor that is stable to ∼500 °C.Reference Shi, Saboktakin, Stavola and Pearton⁵¹ First-principles calculations showed that the hydrogen atom sits in an oxygen vacancy and forms weak bonds with the neighboring zinc atoms.Reference Janotti and Van de Walle⁵³ Because the bonds are weak, its predicted vibrational frequency is low (∼800 cm⁻¹), lying in a spectral region with strong IR two-phonon and free-carrier absorption. For this reason, detecting substitutional hydrogen by IR spectroscopy presents a technical challenge. To our knowledge, substitutional hydrogen has not been observed with IR spectroscopy in other oxides either, presumably for the same reason. Photoluminescence spectra at low temperatures show several sharp lines corresponding to substitutional donors in ZnO.Reference Meyer, Alves, Hoffman, Kriegseis, Forster, Bertram, Christen, Hoffman, Strassburg, Dworzak, Haboeck and Rodina⁵⁴ The “I ₄” line has been assigned to substitutional hydrogen donors.Reference Janotti and Van de Walle^53, Reference Lavrov, Herklotz and Weber⁵⁵

Along with its role as a shallow donor, hydrogen also passivates acceptors. In ZnO, the zinc vacancy is a double acceptor and is fully passivated by two hydrogen atoms. In essence, the hydrogen atoms provide the electrons that were taken away by the missing zinc atom. The protons then bind to oxygen atoms, resulting in two IR absorption peaks.Reference Lavrov, Weber, Börrnert, Van de Walle and Helbig⁴⁶ Hydrogen also has been known to passivate the deep acceptors copperReference Gärtner and Mollwo⁵⁶ and lithium.Reference Halliburton, Wang, Bai, Garces, Giles, Callahan and Wang^57, Reference Shi, Stavola and Beall Fowler⁵⁸ In both complexes, hydrogen attaches to an oxygen atom, forming an O–H bond.

Nitrogen acceptors are passivated by hydrogen in bulk ZnO grown in ammonia by chemical vapor transportReference Jokela and McCluskey⁵⁹ and in thin films grown by metalorganic chemical vapor deposition.Reference Kumar, Watanabe, Nakai, Nishimoto and Fujita⁶⁰ Through isotopic substitution, Jokela and McCluskey showed that the IR absorption peak at 3151 cm⁻¹ corresponds to the bond-stretching mode of a N–H pair (Fig. 3).Reference Jokela and McCluskey⁶¹ The local vibrational mode frequency agreed with a theoretical model, in which the hydrogen attaches to the nitrogen acceptor in an “antibonding” orientation.Reference Li, Keyes, Asher, Zhang, Wei, Coutts, Limpijumnong and Van de Walle⁶²

FIG. 3.

IR spectra (8 K) of N–H pairs in ZnO, for samples grown in Reference King, Veal, Payne, Bourlange, Egdell and McConville¹⁴NH₃, Reference Lany, Zakutayev, Mason, Wager, Poeppelmeier, Perkins, Berry, Ginley and Zunger¹⁵NH₃, or a mixture of Reference King, Veal, Payne, Bourlange, Egdell and McConville¹⁴ND₃ and Reference Lany, Zakutayev, Mason, Wager, Poeppelmeier, Perkins, Berry, Ginley and Zunger¹⁵NH₃.Reference Jokela and McCluskey⁶¹ The higher mass of the Reference Lany, Zakutayev, Mason, Wager, Poeppelmeier, Perkins, Berry, Ginley and Zunger¹⁵N isotope results in a slightly lower vibrational frequency. N-D peaks are not shown. Inset: ball-and-stick model.Reference Li, Keyes, Asher, Zhang, Wei, Coutts, Limpijumnong and Van de Walle⁶²

The study of N–H complexes was essential for understanding the nature of nitrogen acceptors in ZnO. As reviewed in Ref. Reference McCluskey and Jokela37, numerous theoretical and experimental reports claimed that nitrogen is a shallow acceptor dopant that could produce p-type material. However, more recent calculations have indicated that nitrogen is a deep acceptor with a (0/-) level over 1 eV above the valence-band maximum.Reference Lyons, Janotti and Van de Walle^63, Reference Lany and Zunger⁶⁴ To test the theory, experiments were performed on bulk, single-crystal ZnO:N samples that were grown in an ammonia ambience.Reference Tarun, Zafar Iqbal and McCluskey⁶⁵ In these samples, the majority of the nitrogen acceptors were passivated. To activate the nitrogen acceptors, the samples were annealed, causing the N–H pairs to break up. This led to a decrease in the N–H peak intensity and an increase in “red” luminescence centered at ∼730 nm (1.70 eV). The red luminescence is in agreement with the predictions of the deep-acceptor model.Reference Lyons, Janotti and Van de Walle⁶³ These results provided evidence that nitrogen is a deep acceptor in ZnO and cannot lead to p-type conductivity.

Regarding hydrogen diffusion, experiment and theory are at odds. Experimentally, a hydrogen migration barrier of ∼1 eV has been found by Mollwo,Reference Mollwo⁶⁶ Thomas and Lander,Reference Thomas and Lander⁶⁷ and Nickel.Reference Nickel⁶⁸ This value concurs with the temperature-dependent decay of interstitial hydrogen donors, which has activation energy of 0.96 eV.Reference Jokela and McCluskey⁵⁰ However, local density approximation (LDA) calculations predict a much lower migration barrier of 0.4–0.5 eV.Reference Wardle, Goss and Briddon^69, Reference Bang and Chang⁷⁰ The reason for this major discrepancy is not currently known.

B. SnO₂

Tin oxide, also known as stannic oxide or cassiterite in its mineral form, has the rutile crystal structure and a direct band gap of 3.6 eV.Reference Batzill and Diebold⁷¹ Fluorine-doped SnO₂ is used as a transparent electrode for thin-film amorphous silicon– and cadmium telluride (CdTe)-based solar cellsReference Delahoy, Guo, Luque and Hegedus⁷² and dye-sensitized solar cells.Reference Anscombe⁷³ Transparent thin-film transistors have also been fabricated.Reference Presley, Munsee, Park, Hong, Wager and Keszler⁷⁴ As discussed in Sec. I, SnO₂ coatings are used to improve the energy efficiency of windows, by transmitting visible light while reflecting IR light.Reference Lewis and Paine⁷⁵ They are also used as transparent electrodes in electrochromic windows, in which the window transparency is controlled by an applied voltage.Reference Gordon⁷⁶ Finally, SnO₂ is used in gas-sensing applications, in which the conductivity changes in the presence of specific gas molecules.Reference Göpel and Schierbaum⁷⁷ The sensitivity to molecules such as carbon monoxide increases significantly when SnO₂ particles have a diameter smaller than 10 nm.Reference Lu, Liu, Dong and Wang⁷⁸

As with many other oxides, the prevailing assumption that oxygen vacancies are responsible for n-type conductivity was challenged by DFT calculations.Reference Singh, Janotti, Scheffler and Van de Walle⁷⁹ These calculations estimated that the (0/2+) donor level is ∼1.8 eV above the valence band maximum, which is in the middle of the band gap. Such a deep level cannot contribute an appreciable concentration of free electrons to the conduction band. The tin interstitial was calculated to have high formation energy in n-type material, resulting in a negligible concentration. The authors proposed instead that substitutional and interstitial hydrogen could act as shallow donors in SnO₂.

The theoretical predictions were tested experimentally by annealing bulk, single-crystal SnO₂ in H₂ gas at 600 °C.Reference Hlaing Oo, Tabatabaei, McCluskey, Varley, Janotti and Van de Walle⁸⁰ An IR peak observed at 3156.1 cm⁻¹ was attributed to interstitial hydrogen, with the O–H bond aligned perpendicular to the c-axis. The hydrogenated samples showed free-carrier absorption and an increase in electrical conductivity, consistent with the shallow-donor nature of interstitial hydrogen. As in ZnO, the O–H peak is unstable, decaying away after a few weeks at room temperature (Fig. 4).

FIG. 4.

IR spectra (10 K) of O–H peaks from interstitial hydrogen donors and defect-hydrogen complexes in SnO₂.Reference Hlaing Oo, Tabatabaei, McCluskey, Varley, Janotti and Van de Walle⁸⁰ The top spectrum was taken after an 11 day “anneal” at room temperature.

Along with interstitial hydrogen donors, hydrogenated SnO₂ also showed a series of O–H peaks in the 3250–3350 cm⁻¹ range. From first-principles calculations, it was proposed that two of these peaks arise from tin vacancies decorated with one hydrogen atom.Reference Hlaing Oo, Tabatabaei, McCluskey, Varley, Janotti and Van de Walle⁸⁰ Two different orientations of the O–H bond result in two different vibrational frequencies (3261.7 and 3282.0 cm⁻¹). The hydrogen atom partially passivates the tin vacancy, converting it into a triple acceptor. The calculated formation energies for these (V _Sn-H)⁻³ defects is low for n-type material.

Experiments by Bekisli et al.Reference Bekisli, Stavola, Beall Fowler, Boatner, Spahr and Lüpke⁸¹ confirmed the assignment of the 3156.1 cm⁻¹ peak to hydrogen donors. By correlating the strength of this peak with that of free-carrier absorption, the authors determined that a second donor, probably substitutional hydrogen, was stable to 600 °C. The properties of interstitial and substitutional hydrogen donors are remarkably similar to those in ZnO. Through systematic IR polarization spectroscopy, they assigned the peaks in the 3250–3350 cm⁻¹ range to defects containing one or two hydrogen atoms. All but two lines were polarized perpendicular to the c-axis.

First-principles calculations by Varley et al.Reference Varley, Janotti, Singh and Van de Walle⁸² showed that the acceptors aluminum, gallium, and indium readily substitute for tin, perhaps providing a way toward p-type doping in SnO₂. They are all passivated by hydrogen, resulting in O–H bonds that should give rise to well-defined vibrational peaks in the IR spectrum. To our knowledge, however, these acceptor-hydrogen complexes have not been observed experimentally.

C. TiO₂

Titanium dioxide or titania has an indirect band gap of 3.0 eV in the rutile structure and 3.2 eV in the anatase structure.Reference Kavan, Grätzel, Gilbert, Klemenz and Scheel⁸³ It is widely used as a white pigment in paint. TiO₂ surfaces, especially the well-characterized rutile (110) surface, are used in photocatalytic processes such as water splitting (Sec. I) and photodegradation of organic molecules.Reference Diebold⁸⁴ The surface adsorption of various organic species has been studied extensively with FTIR spectroscopy.Reference Ojamäe, Aulin, Pedersen and Käll⁸⁵ TiO₂ is a preferred material for photoelectrochemical solar cells.Reference Grätzel⁸⁶ In these devices, sunlight creates an electron and hole in the TiO₂. The electron travels through an external circuit whereas the hole is “scavenged” by a molecule in the electrolyte. Dye-sensitized solar cells use a monolayer of dye molecules to push the absorption profile into the visible.Reference O’Regan and Grätzel⁸⁷

Rutile TiO₂ shows various hydrogen vibrational modes, including one with a frequency 3290 cm⁻¹ that corresponds to an O–H bond aligned perpendicular to the c-axis.Reference Soffer⁸⁸ Herklotz et al.Reference Herklotz, Lavrov and Weber⁸⁹ investigated the fine structure of this mode and its temperature dependence. They attributed two IR absorption peaks, at 3288.3 and 3292.0 cm⁻¹, to the neutral charge state, and a mode at 3287.4 cm⁻¹ to the +1 charge state. As the sample was warmed up, the increase in the 3287.4 cm⁻¹ line was correlated with an increase in free-carrier absorption. These results suggest a shallow donor with ionization energy of 10 meV. The experimental work is in contrast to the calculations of Filippone et al., which predicted that interstitial hydrogen should be a deep donor.Reference Filippone, Mattioli, Alippi and Amore Bonapasta⁹⁰ Other calculations, however, found that interstitial hydrogen is a shallow donor in TiO₂.Reference Kiliç and Zunger^27, Reference Peacock and Robertson^91, Reference Bjørheim, Stølen and Norby⁹²

The vibrational lifetime of the O–H bond-stretching mode provides information that is relevant to hydrogen diffusion. IR pump-probe experiments measured the temperature dependence of the ground-state recovery lifetime.Reference Spahr, Wen, Stavola, Boatner, Feldman, Tolk and Lüpke⁹³ The shape of the lifetime-versus-temperature curves suggest that the O–H stretch mode quickly couples to a wag mode, which enables it to migrate to a neighboring oxygen site. This model implies that IR light can lead to a giant enhancement of hydrogen migration in TiO₂.

The theoretical situation for oxygen vacancies is an area of some disagreement. DFT calculations using the generalized gradient approximation (GGA) show that the +2 charge state of the oxygen vacancy is energetically preferred for all values of the Fermi level within the band gap; i.e., it acts as a shallow double donor.Reference Bjørheim, Stølen and Norby^92, Reference Cho, Han, Ahn, Lee, Kim and Hwang⁹⁴ However, DFT calculations, using a local spin density + U method, find that the oxygen vacancy is a deep donor. We note that this latter technique was the same one that also found hydrogen to be a deep donor.Reference Filippone, Mattioli, Alippi and Amore Bonapasta⁹⁰

D. In₂O₃ and Ga₂O₃

As discussed in Sec. I, In₂O₃ is the preferred TCO for flat-panel displays and other applications. Despite its popularity and success, relatively little is known about the properties of hydrogen impurities. In₂O₃ has the bixbyite crystal structure, with 80 atoms in the conventional cell. Although it has a direct band gap of 2.7 eV,Reference Bourlange, Payne, Egdell, Foord, Edwards, Jones, Schertel, Dobson and Hutchison⁹⁵ strong dipole-allowed absorption occurs ∼1 eV above this value, making it quite transparent up to ∼3.7 eV. Koida et al.Reference Koida, Fujiwara and Kondo⁹⁶ demonstrated that hydrogen doping of sputtered films resulted in free-electron concentrations of 3 × 10²⁰ cm⁻³ and mobilities as high as 130 cm²/V·s. Calculations by Limpijumnong et al.Reference Limpijumnong, Reunchan, Janotti and Van de Walle⁹⁷ indicated that substitutional and interstitial hydrogen act as shallow donors, in agreement with experiment. Researchers have exploited the effect of hydrogen on conductivity, using nanocrystalline In₂O₃ thin films as hydrogen gas sensors.Reference Shukla, Seal, Ludwig and Parish⁹⁸ The (0/2+) donor level of the oxygen vacancy was calculated to lie close to the experimental conduction-band minimum, suggesting that it may act as a shallow donor.Reference Reunchan, Zhou, Limpijumnong, Janotti and Van de Walle⁹⁹

β-Ga₂O₃ has a monoclinic crystal structure and a direct band gap of 4.9 eV, giving it transparency well into the UV.Reference Orita, Ohta, Hirano and Hosono¹⁰⁰ DFT calculations indicate that oxygen vacancies are deep donors, whereas interstitial and substitutional hydrogen are shallow donors.Reference Varley, Weber, Janotti and Van de Walle¹⁰¹ Whereas isolated cation vacancies are energetically unfavorable in SnO₂ and In₂O₃, gallium vacancies have low formation energy in Ga₂O₃.Reference Varley, Peelaers, Janotti and Van de Walle¹³ In the presence of hydrogen, however, the formation energies are reduced and vacancy-hydrogen complexes may form in all three materials.

E. SrTiO₃

SrTiO₃ is a perovskite oxide (Fig. 5) with an indirect band gap of 3.25 eVReference van Benthem, Elsässer and French¹⁰² and is commonly used as a substrate for oxide thin films, including high-T_c superconductors.Reference Kawai, Watanabe and Hanada¹⁰³ The observation of blue light emission in irradiated SrTiO₃ suggests potential optoelectronic applications.Reference Kan, Terashima, Kanda, Masuno, Tanaka, Chu, Kan, Ishizumi, Kanemitsu, Shimakawa and Takano¹⁰⁴ The cubic perovskite structure undergoes a symmetry-lowering transition to a tetragonal phase below 105 K, which involves a rotation of the oxygen octahedron with respect to the strontium cage.Reference Houde, Lépine, Pépin, Jandl and Brebner¹⁰⁵ Its dielectric constant depends strongly on temperature, increasing from a few hundred at room temperature to tens of thousands at low temperatures. The dielectric screening of ionized impurities at low temperature contributes to excellent free-electron mobilities.Reference Son, Moetakef, Jalan, Bierwagen, Wright, Engel-Herbert and Stemmer¹⁰⁶ Although p-type SrTiO₃ has been reported via doping with scandiumReference Higuchi, Tsukamoto, Sata, Ishigame, Tezuka and Shin¹⁰⁷ and indium,Reference Fix, Bali, Stelmashenko and Blamire¹⁰⁸ the evidence–as with claims of p-type ZnO–is far from conclusive.

FIG. 5.

Perovskite structure of SrTiO₃.

In as-grown SrTiO₃, hydrogen forms an O–H bond with a host oxygen atom, resulting in an IR absorption peak near 3500 cm⁻¹ due to the bond-stretching vibration.Reference Houde, Lépine, Pépin, Jandl and Brebner¹⁰⁵ In samples annealed in hydrogen gas, Tarun and McCluskeyReference Tarun and McCluskey¹⁰⁹ investigated this peak, labeled H_I, along with a pair of peaks at 3355 and 3384 cm⁻¹, labeled H_II. All the peaks showed an expected isotope shift to lower frequency when samples were annealed in deuterium. When samples were annealed in a mixture of hydrogen and deuterium, new H_II peaks appeared. These peaks corresponded to HD centers, showing unambiguously that the H_II center contains two hydrogen atoms.

The structures of the hydrogen complexes are not known definitively. Theoretical studies have suggested that the H_I line arises from an isolated hydrogen interstitialReference Villamagua, Barreto, Prócel and Stashans¹¹⁰ whereas H_II represents a pair of hydrogen interstitials that are attracted to each other via lattice strain.Reference Bork, Bonanos, Rossmeisl and Vegge¹¹¹ Tarun and McCluskeyReference Tarun and McCluskey¹⁰⁹ attributed H_II to a strontium vacancy that is fully passivated by two hydrogen atoms. However, the experimental results cannot exclude the possibility that H_II is a titanium vacancy decorated by two hydrogen atoms,Reference Thienprasert, Fongkaew, Singh, Du and Limpijumnong^112, Reference Varley, Janotti and Van de Walle¹¹³ or two neighboring hydrogen interstitials.Reference Bork, Bonanos, Rossmeisl and Vegge¹¹¹ Furthermore, one must consider the possibility that the H_I configuration is a defect-hydrogen complex rather than simply isolated hydrogen.Reference Thienprasert, Fongkaew, Singh, Du and Limpijumnong¹¹²

By monitoring the evolution of peaks in deuterated SrTiO₃, Tarun and McCluskeyReference Tarun and McCluskey¹⁰⁹ provided evidence for “hidden hydrogen,” possibly H₂ molecules, in the sample. As shown in Fig. 6, the deuterated sample contains O–D peaks corresponding to the H_I and H_II configurations. When the sample is annealed at 500 °C, one can see the appearance of an HD peak. This occurs because a hydrogen atom has come out of its reservoir and replaced a deuterium atom. At the same time, the H_I O–H peak increases, for the same reason. At 600 °C, the released hydrogen atoms have totally replaced the deuterium atoms. Future studies, perhaps involving Raman spectroscopy, will be performed to determine the identity of the hidden hydrogen.

FIG. 6.

IR absorption spectra (200 K) of O-D and O-H peaks in deuterated SrTiO₃.Reference Tarun and McCluskey¹⁰⁹ Two different complexes are labeled H_I and H_II. Above 500 °C, the D atoms are replaced by H, which come from a “hidden hydrogen reservoir” in the sample.