III. RESULTS AND DISCUSSION

Ga_xZn_1−xO_yN_1−y powders were synthesized by a solution combustion method using urea as the fuel and nitrates as the oxidizer at a wide range of temperatures (350–800 °C). For simplicity, these powders are referred to throughout the text by the precursor molar ratio (e.g., 1:1:10 refers to 1 part Ga, 1 part Zn, and 10 parts urea). Urea decomposes, beginning at temperatures of approximately 150–200 °C, into a number of reactive products, including biuret, cyanic acid, and ammonia [as NH₄⁺(aq) or NH₃(g)].^{Reference Schaber, Colson, Higgins, Thielen, Anspach and Brauer44} The reaction was allowed to proceed for 30 min, after which the reaction is considered complete. As synthesized, the powders had a foam-like texture that became more granulated as the reaction temperature was raised to 800 °C. Zinc-rich powders typically produced a deeper orange color while gallium-rich powders tended to a yellow-white hue. In general, a 5:1 fuel-to-metal ratio was found to produce powders with the largest visible-region range of absorption. Increasing or decreasing the fuel-to-oxidizer ratio typically favored the formation of only oxides, although there was evidence of slight N-doping in some samples (see Fig. S2).

The particle size of the powders and morphology of the as-synthesized Ga_xZn_1−xO_yN_1−y samples were estimated by scanning electron microscopy (SEM) imaging (Fig. 1) of painted films. It is clear from the images that the powders consist of particles ranging from a few tens of nanometers to microns in diameter. The porous structure of the particles is due to the rapid formation of gas bubbles during the synthesis procedure.

FIG. 1.

SEM image of the 1:1:10 Ga_xZn_1−xO_yN_1−y sample prepared at 500 °C for 30 min. The morphology seen here is a characteristic of all Ga_xZn_1−xO_yN_1−y samples imaged with the SEM. The accelerating voltage was 10 kV.

Figure 2 shows the absorption onset for several Ga_xZn_1−xO_yN_1−y samples. Direct band gaps^{Reference Fujishima, Kato, Maekawa and Honda45} were determined using Eq. (1) with diffuse reflectance measurements:

(1)$${\left( {h\nu \alpha } \right)^2} = A\left( {h\nu - {E_{\rm{g}}}} \right)\quad ,$$

where hν is the energy of the incident photon, E _g is the band gap of the material, A is a proportional constant, and α is the Kubelka–Munk absorption coefficient.

FIG. 2.

Kubelka–Munk absorbance of synthesized powders from different precursor molar ratios of Ga(NO₃)₃, Zn(NO₃)₂, and urea synthesized at 500 °C. Spectra were gathered from 850 to 350 nm at a 1 nm/s scan rate. BaSO₄ was used as a 100% reflectance reference material.

ZnO (>3.0 eV),^{Reference Fujishima, Kato, Maekawa and Honda45} GaN (3.4 eV),^{Reference Trapalis, Heffernan, Farrer, Sharman and Kean46} and ZnGa₂O₄ (>4.0 eV)^{Reference Ikarashi, Sato, Kobayashi, Saito, Nishiyama and Inoue4,Reference Oshima, Niwa, Mukai, Nagami, Suyama and Ohtomo47} are all wide band gap semiconductors, so absorbance in the visible region is due to the incorporation of nitrogen into the crystal lattice, leading to an upward, more negative shift of the valence band of the semiconductor.^{Reference Maeda, Takata, Hara, Saito, Inoue, Kobayashi and Domen16,Reference Tessier, Maillard, Cheviré, Domen and Kikkawa48}

The narrowest band gaps were observed for the 1:1:10, 1:2:15, and 1:3:20 samples, all of which have direct band gaps of approximately 2.2 eV. That a narrow band gap is observed across multiple samples with different ratios of Ga:Zn suggests that the narrowing is not strongly dependent on the actual nitrogen content but is tied to the final product being Zn-rich. This is supported by previous work on Ga_xZn_1−xO_yN_1−y prepared by the ammonolysis method.^{Reference Maeda, Teramura, Takata, Hara, Saito, Toda, Inoue, Kobayashi and Domen49} In Ga_xZn_1−xO_yN_1−y samples that have either more or less than a 5:1 fuel:oxidizer ratio, the Tauc plots reveal band gaps corresponding to the metal oxides (i.e., E _g ≥ ∼3.0 eV). A tail, though, was noticeable in the 0:1:5, 1:2:10, and 2:1:10 samples (Fig. S2). These tails are attributed to the incorporation of nitrogen at dopant levels during the combustion synthesis as either ZnO_xN_1−x or Zn₃N₂.^{Reference Lu, Zhang, Wang, Saito and Uchida50} Stable formation of a pure phase of Zn₃N₂, which can have an optical band gap of 1.3–1.5 eV, is unlikely, however, due to its tendency to quickly oxidize.⁵¹

We also note that band gaps are smaller than 3.0 eV for all precursor ratios used, except 1:1:0 (synthesis without urea), for temperatures between 400 and 700 °C (Fig. 3). Furthermore, these band gaps are mostly constant throughout this region, with the exception of the 3:1:20 molar ratio samples. For the 3:1:20 ratio, a slight increase in band gap is observed starting at a synthesis temperature of 550 °C and above.

FIG. 3.

Calculated band gaps for various Ga:Zn:urea ratios at different reaction temperatures. In all cases, the reaction time was 30 min, and the atmosphere was static air. The inset picture is of the powders produced at 500 °C.

XRD data for several different precursor molar ratios prepared at 500 °C are compared in Fig. 4. ZnO is the primary crystalline phase in the 0:1:5 sample, as expected. As the amount of gallium present in the powders increases (0:1:5 to 2:1:15), however, the ZnO peaks shift to higher 2θ. This shift to higher 2θ angles suggests the co-formation of GaN (see blowup in Fig. 5) resulting in a GaN:ZnO complex. The Domen group^{Reference Maeda and Domen22,Reference Maeda, Teramura, Takata, Hara, Saito, Toda, Inoue, Kobayashi and Domen49} observed a similar shift in these peaks for gallium zinc oxynitrides synthesized by the solid-state ammonolysis method. The peaks corresponding to ZnO in the 2:1:15 and 3:1:20 samples broaden, indicating shrinking crystalline domains, and become smaller in intensity. At the same time, peaks corresponding to ZnGa₂O₄ become more prominent. In particular, peaks corresponding to the 002 and 102 planes of ZnO are gone in the 3:1:20 sample.

FIG. 4.

XRD patterns for samples with different precursor ratios synthesized at 500 °C. Scans were collected from 10 to 70° with a step size of 0.04° and scan rate of 1°/min.

FIG. 5.

Blow-up of Fig. 4, highlighting the shift from ZnO to GaN as the Zn/Ga precursor ratio changes. The red dotted line corresponds to the [100] crystallographic peak of ZnO.

Figure S3 shows XRD spectra for 1:1:10 samples with increasing reaction temperature. As expected, the crystallinity of the powders increases with increasing temperatures. We also note, however, slight shifts in the peak positions. From 350 °C up to 600 °C, peaks corresponding to ZnO and ZnGa₂O₄ shift to higher 2-theta positions, indicating the incorporation of nitrogen into the powders. Above 600 °C, however, the peaks shift back to lower 2-theta values corresponding to peak positions of ZnO and ZnGa₂O₄, indicating that oxide formation is favorable at higher temperatures.

XPS analysis confirmed the incorporation of nitrogen and was used to determine the atomic composition of and elemental valence states in the as-synthesized powders. Figure S1 shows typical Ga 2p and Ga 3d XPS spectra, which confirm the formation of a GaN phase.^{Reference He, Chikyow, Chen, Chen, Liu and Sun52} Figure 6 has N 1s spectra for Ga_xZn_1−xO_yN_1−y powders synthesized at different temperatures with a 1:1:10 precursor ratio. All the peaks are within a range that can be reasonably assigned to the formation of an N³⁻ metal nitride species, with a Ga Auger peak inconsistently observed.^{Reference Rignanese, Pasquarello, Charlier, Gonze and Car53} Table SII summarizes the measured atomic ratios of Ga_xZn_1−xO_yN_1−y samples synthesized at 500 °C with their corresponding starting molar ratios of precursors. It can be seen that the final ratio of Ga to Zn generally follows the ratio of starting materials. Of particular note, however, is that the apparent atomic compositions, as derived from XPS analysis, show a zinc-rich surface compared to the initial ratio of the precursors. Based on the reaction temperatures used, vaporization of Ga is not reasonable, suggesting that the zinc segregates preferentially to the surface during the synthesis. This was confirmed by the EDX analysis of some samples. For example, EDX analysis (not shown) of the 1:2:15 powder gave a Ga:Zn ratio of ∼1:2.5, much closer to the expected 1:2 ratio than that observed at the surface of the powder by XPS, which was ∼1:3.5.

FIG. 6.

N 1s spectra of 1:1:10 Ga_xZn_1−xO_yN_1−y powders synthesized at different reaction temperatures. All reaction times of all samples were held at 30 min.

The photoactivity of select samples was evaluated with a series of photoelectrochemical measurements. SO₃²⁻ was used in some experiments to scavenge holes at the photoelectrode/electrolyte interface.^{Reference Stefik54} The observed photocurrents (Fig. 7) are in line with that observed previously for dropcast films of Ga_xZn_1−xO_yN_1−y powders.^{Reference Hashiguchi, Maeda, Abe, Ishikawa, Kubota and Domen55} For the 1:1:10 powder synthesized at 350 °C, no photocurrent is observed. For all other temperatures, photocurrent is observed, with the highest activity coming from the powders synthesized at 500 and 600 °C. The low photocurrent is attributable primarily to the large particle size of the oxynitride powders. The large particle size increases the likelihood of charge recombination within the particle before extraction occurs and also hinders interparticle charge transfer.

FIG. 7.

Current–time curves for different Ga_xZn_1−xO_yN_1−y powders (denoted by Ga:Zn:urea ratio). Potential was held at 0 V versus SCE. Counter electrode was a Pt wire. Electrolyte was 0.2 M Na₂SO₃/0.1 M Na₂SO₄.

The 1:1:10 film gave clear and repeatable photoresponse upon illumination with visible light (λ > 435 nm) both in the presence (Fig. S4) and in the absence (Fig. S5) of a hole scavenger, indicating that overall splitting of water is feasible even in the absence of hydrogen or oxygen evolution catalysts. Photocurrent was also observed in the 0.5 M Na₂SO₄ solution without any externally applied bias (Fig. S6).

Ga_xZn_1−xO_yN_1−y powder suspended in water produced no detectable hydrogen or oxygen gas evolution under either AM1.5 or visible-light-only irradiation after 6 h, so sacrificial half-reactions were used to evaluate the potential for gas evolution. Oxygen evolution experiments were carried out in 10 mM AgNO₃ aqueous solution under simulated AM1.5 irradiation at ∼100 mW/cm². Without La₂O₃ present as a pH buffer, 3.4 μmol of O₂ and 28 μmol of N₂ were evolved after 6 h of irradiation, matching previous work that attributed the low rate of evolution to a significant acidification of the unbuffered solution as the silver nitrate is reduced.^{Reference Kasahara, Nukumizu, Takata, Kondo, Hara, Kobayashi and Domen56} In the presence of 200 mg of La₂O₃, 14.1 μmol of O₂ and 28 μmol of N₂ were evolved after 6 h (Fig. S8). Addition of a CoO_x oxygen evolution catalyst did not increase the rate of oxygen evolution, suggesting that the intraparticle charge recombination dominates. Hydrogen evolution was carried out using a 50 vol% ethanol solution and 1:1:10 Ga_xZn_1−xO_yN_1−y powder loaded with a Rh/CrO_x catalyst. After an initial induction period of approximately 1 h, hydrogen gas was evolved steadily over the next 3 h at a rate of 273 μmol/(h g) photocatalyst before plateauing (Fig. S7). The plateau and slight decline is due to the build-up of surface-adsorbed acetate that oxidizes at a significantly slower rate.^{Reference Vayssieres57}

Based on the observed results, we have identified the relatively large particle size as a first point of improvement. Work is ongoing to reduce particle size in two different ways: (i) by extending the reaction time under nitrogen atmosphere and (ii) utilizing the flame spray pyrolysis technique to nebulize the precursor solutions and produce smaller reaction pools. There are also a number of post-treatements that should enable further improvements upon the method demonstrated in this work, even with larger particles. Previous work has demonstrated that necking treatments can enhance the photocurrent generated by facilitating better charge collection via reducing interparticle recombination.^{Reference Abe, Higashi and Domen58} Creating an intimate back contact via sputtering a microns-thick metal electrode directly onto the powder has also been shown to enhance the photoelectrochemical output by effectively eliminating the charge collection bottleneck.^{Reference Minegishi, Nishimura, Kubota and Domen59}

Another key to generalizing this synthesis method is understanding the role of the metal salt anions. As an example, substitution of ZnCl₂ for Zn(NO₃)₂ results in a different final product when all other reaction conditions are held constant. We are currently working on elucidating the role of the salt anion in the overall reaction and developing a system that will allow for analysis of the particles at various stages of the reaction while controlling the synthesis temperature and atmosphere. In addition, it is not yet clear that simply having NO₃⁻ present is sufficient for the reaction to proceed successfully—concomitant substitution of the metal salt anion (i.e., Cl⁻ for NO₃⁻) and addition of ammonium nitrate does not give the same final product. Both of these issues are being explored to gain insight into the reaction mechanism.