Molybdenum and its alloys are widely used in the steel industry, as high-temperature catalysts, and for their biomedical applications[1-4]. Hence, it is imperative to produce oxygen-free molybdenum. As molybdenum metal is not available naturally, the reduction of molybdenum oxides to molybdenum metal is necessary in a hydrogen gas environment at elevated temperatures[5,6], approximately 1000 °C, to allow the reaction between dissociated hydrogen and oxygen[6,7]. This reaction mainly occurs in MoO₂ as MoO₃ is readily reduced to MoO₂ during heating in reducing conditions.

Understanding the microscopic origin and reaction kinetics of the chemical reaction is important. Several reports[5,6,8] focused on chemical analysis to produce oxygen-free molybdenum. The reaction mentioned above is a thermodynamic equilibrium process. It is essential to know what happens when the reaction occurs in epitaxial thin films, where thermodynamic non-equilibrium states are non-negligible. For example, epitaxial films are under strained states[9] as they are typically grown on single crystalline substrates with a similar, but not same, lattice constant. This strain effect often plays a key role in unconventional phenomena, which cannot be explained in the frame of thermodynamic equilibrium theories[10].

In this study, the reduction of MoO₂ to Mo while maintaining their epitaxial forms is demonstrated by high-temperature reductive annealing processes. X-ray diffraction showed that (100) MoO₂ is transformed to (110) Mo without creating other orientations, as shown in Fig. 1. X-ray absorption spectra (XAS) at the O K-edge do not show the formation of metal-oxygen bonds at the sample surface. Optical spectroscopic ellipsometry data confirmed that the transformation of epitaxial MoO₂ to Mo happens at 900 °C. Compared with the optical conductivity spectrum of an as-grown epitaxial Mo film, the transformed Mo films were oxygen-free but possessed some defects.

Figure 1. (Color online) Schematic diagram to describe the epitaxial reduction of MoO₂ thin films. The left shows 100 plane of MoO₂, where the epitaxial MoO₂ film is grown on 0001 Al₂O₃. The right shows 110 plane of Mo, which is the product of high temperature reductive annealing.

Epitaxial MoO₂ thin films (~44 nm) were grown on (0001) Al₂O₃ substrates (Crystal bank, Pusan National University) using RF magnetron sputtering. The MoO₃ target was made by sintering pelletized MoO₃ powder (Alfa Aesar, purity: 99.998%). While sputtering at 500 °C, pure Ar was filled in the chamber to effectively reduce MoO₃ to MoO₂. The RF power, flow rate of Ar, and pressure were 50 W, 30 sccm, and 7 mTorr, respectively. As-grown MoO₂ was characterized in detail and the result can be found in Ref.[11]. Subsequent reduction experiments were performed in a tube furnace. Annealing was performed in Ar gas containing H₂ to induce a reduction reaction of O₂ in MoO₂[12-14]. Additionally, 3% H₂ was used to prevent the risk of explosion according to the H₂ gas ratio[15-17]. The heating rate was 200 °C/h. Each sample was annealed in 20 sccm of the forming gas (3% of H₂ and 97% of Ar) at temperatures of 800, 900, and 1000 °C.

After the reaction, each sample was characterized with X-ray diffraction for structural analysis (Bruker D8 Discover), X-ray absorption spectroscopy (2A in Pohang Accelerator Laboratory) for chemical bonding characteristics, and spectroscopic ellipsometry (VASE ellipsometer, J. A. Woollam Co.) for optical measurements. Atomic force microscopy was used to check the surface morphology of each sample after the reaction.

Figure 2 shows the X-ray diffraction data from the as-grown MoO₂ epitaxial film to the reduced MoO₂ epitaxial films annealed at different temperatures: 800, 900, and 1000 °C. The annealing time was limited to 20 min. By focusing on the (400) MoO₂ diffraction peaks, we observed slight lattice expansion along the c-axis when the film was annealed at high temperatures. This is likely due to the start of oxygen desorption from the lattice. The sample annealed at 1000 °C was completely transformed into epitaxial molybdenum metal. The peak position was consistent with that of epitaxial molybdenum grown by sputtering[18]. No other orientations were observed although other mixed phases were often found as evidence of the intermediate phases. Figures 2(b) and (c) show the surface morphology of a fully reacted MoO₂ film. The surface roughness was approximately 2.6 nm, indicating that the reduction process does not significantly change the surface morphology. This result is consistent with the result from a crystalline specimen[6]. Typically, the surface roughness of as-grown MoO₂ is less than 1 nm. The slight surface roughening is presumably due to the sublimation of surface MoO₃ and/or the difference in reduction reactions[19]. Note that the roughening is different from the surface roughening of other transition metals, such as cobalt, during thermal annealing[20].

Figure 2. (Color online) (a) X-ray diffraction results of the as-grown MoO₂ thin film and the reductively annealed films. (b) 10 μm by 10 μm and (c) 2 μm by 2 μm atomic force microscopy images of the annealed Mo film, which was annealed at 1000oC. No significant changes in surface roughness are observed during the chemical reaction.

Figure 3 shows O K-edge XAS of as-grown MoO₂ and reduced MoO₂ epitaxial thin films at two different temperatures: 300 and 80 K. The XAS of the as-grown MoO₂ is analogous to the reported spectrum[21]. The pre-edge peaks at 529.8 and 531.2 eV correspond to hybridizations of Mo 4d t2g–O 2p. In contrast to the oxidation case in MoO₂, the peak at 529.8 eV decreased compared to the peak intensity at 531.2 eV. When oxidized, the number of electrons in the t2g level decreased, leading to stronger hybridization. However, as shown in Fig. 3(a), there is a decrease in the peak intensity. This means the number of electrons in the t2g level increased by filling electrons owing to the removal of oxygen ions. This trend is systematically seen when annealing temperatures decrease. When the sample is reduced at 1000 °C, we obtain no signals at the O K-edge, i.e., the film is transformed to oxygen-free molybdenum.

Figure 3. (Color online) O K-edge x-ray absorption spectra of the as-grown MoO₂ thin film and reductively annealed films at different temperatures. The x-ray absorption data were taken at (a) 300 K and (b) 80 K. The data clearly shows the evolution of pre-edge spectra near 530 eV. Note that complete removal of O K-edge is consistently seen from the reductively annealed MoO₂ film at 1000 °C.

Figure 4 shows spectroscopic ellipsometry data of MoO₂ and reduced MoO₂ epitaxial thin films. For comparison, data from as-grown epitaxial Mo film is also displayed. The spectrum of the as-grown MoO₂ shows the four distinct features in the spectral range[11]. The optical absorptions at approximately 1.0 and 3.2 eV correspond to the inter-site d-d transitions of Mo. The absorption feature at lower energies (α) is related to the intra-band transition within the e_g band, whereas the feature at higher energies (β) is related to the interband transition (t2g→eg). Additionally, two additional absorption peaks (γs) can be observed between 4 and 5 eV. These are known to be the interband transitions from the filled oxygen 2p band to the partially filled Mo 3d band.

Figure 4. (Color online) Optical conductivity spectra as a function of photon energy from (a) the as-grown MoO₂, reductively annealed MoO₂ films at 800 °C (b), 900 °C (c), and 1000 °C (d), and (e) the epitaxial Mo film directly grown from a Mo target.

There were several spectral changes in the MoO_2-x epitaxial films resulting from annealing of as-grown films at 800 and 900 °C. The α peak is shifted toward the lower energy by filling electrons in the 3d band of molybdenum. This is consistent with the X-ray absorption data discussed above. Additionally, the β peak is shifted toward the lower energy as the excitation comes from the electrons near the Fermi level. While the evidence of electron filling in the e_g band is observable in the d-d transitions, the feature in the p-d transition is unclear. However, this is discussed by comparing with the data of pure Mo films[22,23]. Figure 4(d) shows the spectrum of the MoO₂ epitaxial thin film reduced at 1000 °C, where it is converted to epitaxial Mo film by the reaction. The feature related to the α peak is completely absent. Additionally, new features emerge at 1.7 and 2.4 eV. The feature of the so-called β peak is not visible. This suggests that the electronic structure of the sample is different from the epitaxial MoO₂ films.

To unveil the unknown features in the p-d transition, an epitaxial Mo film was directly grown on 0001 Al₂O₃ by sputtering a Mo target. A custom-made magnetron sputtering was used at 700 °C[18]. Spectroscopic ellipsometry was performed at room temperature. The fitted optical conductivity spectrum is very similar to the data from Ref.[24]. The multiple peaks represent the interband transitions of the Mo band structure[24]. The broad peak near 4.1 eV in Fig. 4(c) is likely the appearance of surface molybdenum. The appearance of peaks at 1.7 and 2.4 eV represents the conversion to crystalline Mo during the high-temperature thermal annealing.

We demonstrated epitaxial conversion of MoO₂ films to Mo by reductive thermal annealing. The chemical phase transition may seem insignificant as it is an established technique to produce molybdenum metal[5]. However, the conversion involves an epitaxial transformation. The epitaxy of (100) MoO₂ on (0001) Al₂O₃ is known as the anisotropic strain may contribute to stabilizing the epitaxial relationship. If we consider a rectangular lattice as the case of the (100)-plane of MoO₂, the lattice constants of MoO₂ are 5.63 and 4.86 Å, while the corresponding lattice constants of Al₂O₃ are 4.76 and 5.72 Å. This creates 1.58% of the tensile strain along the [001]-direction of MoO₂, while -2.1% of the compressive strain along the [101]-direction of MoO₂. Such a small lattice mismatch can create epitaxial MoO₂ with high crystallinity. In the case of the (110)-plane of Mo, lattice constants are 4.16 Å along [001]_Mo and 5.88 Å along [1-10]_Mo. Considering the lattice mismatch between Mo and Al₂O₃, we can expect -2.79% of compressive strain along [1-10]_Mo and 12.6% of tensile strain along [001]_Mo. One can hardly claim that coherent strain will be held along [001]_Mo with such a large mismatch. However, a partial contribution of the strain effect along [1-10]_Mo and thermal energy at high temperatures may result in the epitaxial conversion. Such a large lattice mismatch can create high defect densities in samples. The overall reduction in the optical conductivity spectrum of the converted Mo film compared to the directly grown Mo film, as shown in Figs. 4(d) and (e), might be relevant to the formation of defects.

In conclusion, we investigated the reduction process of epitaxial MoO₂ to Mo using high-temperature annealing with forming gas. The conversion does create epitaxial Mo film with relatively smooth surface. X-ray absorption spectroscopy indicates that the converted film is oxygen-free. Spectroscopic ellipsometry data shows clear evolution from MoO₂ to Mo. Even if the anisotropic strain and thermodynamics are two reasons to create epitaxial Mo from epitaxial MoO₂, naturally inherited lattice mismatch between Mo and Al₂O₃ may create a large fraction of defects in the converted film. Thus, it suppresses the optical conductivity in the converted Mo films.

This work is supported by PNU-RENovation (2021–2022).