The commercially available titanium nanopowder (Nanostructured & Amorphous Materials, Inc. Ti, 99%) is an air-sensitive material and it was stored in the argon-filled glove box prior use. Stable Ti@TiO₂ NPs were prepared by SHT treatment of Ti nanopowder in pure water (Milli-Q 18.2 MΩ·cm at 25°C). The SHT reactor is shown in Supplementary Figure 1 and described previously (Cau et al., 2013; Nikitenko et al., 2015, 2018). In a typical experiment, 2 g of air passivated Ti nanopowder was dispersed in 50 mL of water using ultrasonic bath, placed in SHT reactor and heated at selected temperatures in the range of T = 100–214°C (autogenic pressure P = 1.0–19.0 bar) under simultaneous ultrasonic treatment (f = 20 kHz, P_ac = 17 W) for 3 h. After cooling, the treated NPs were recovered by centrifugation, washed with pure water and dried at room temperature under reduced pressure.

Powder X-Ray Diffraction (XRD) diagrams were recorded with the use of a Bruker D8 Advance X-ray diffractometer equipped with a linear Lynx-eye detector (Cu Kα_1,2 radiation, λ = 1.54184 Å). XRD patterns were collected between 10 and 90° (θ − 2θ mode) at room temperature, with a step size of Δ(2θ) = 0.02° and a counting time of 1.8 s·step⁻¹. Quantitative phase analysis was performed by Rietveld refinement with the phase detection limit about 5% (León-Reina et al., 2016). High-resolution transmission electron microscopic (HRTEM) and scanning transmission electron microscopic (STEM) measurements coupled with EDX mapping (SDD Oxford detector) were performed using a Jeol 2200FS (200 kV) microscope. XPS analysis was achieved with an ESCALAB 250 Thermo Electron device operated under ultra-high vacuum. The excitation source was a monochromatic source (Al-K_α, ε = 1486.6 eV) with a ca. 0.4 mm² X-ray spot. The pass energy was fixed at 20 eV. The photoelectron spectra were calibrated using the Au 4f 7/2 (83.9 ± 0.1 eV) and Cu 2p 3/2 (932.8 ± 0.1 eV) photoelectron lines. The XPS spectra were treated using AVANTAGE software. The binding energy scale was established by referencing the adventitious C 1 s peak at 284.8 eV. The reflectance spectra were recorded in BaSO₄ pellets with a Shimadzu UV-3600 spectrophotometer. Thermogravimetric analysis (TGA) was performed by means of TGA-DTA/DSC Setsys Evolution (Setaram Instrumentation) device in air flow (900°C, 10°C·min⁻¹). The concentration of titanium species in solution after photolysis was measured using SPECTRO ARCOS ICP-OES instrument (detection limit ~ 0.1 ppm).

Electrochemical Impedance Spectroscopy (EIS) was studied at dark conditions in the frequency range from 0.1 to 100 kHz with an AC amplitude 10 mV. The electrolytic cell was filled with 1 M KOH and bubbled with Ar for 20 min prior the measurements. Silver/silver chloride (Ag/AgCl) and platinum (Pt) were used as the reference electrode and counter electrode, respectively. The working electrode was prepared by ultrasonic dispersion of the synthesized photocatalyst (5 mg) in the mixture of 1 mL isopropanol (VWR, ≥99.7%) and 40 μL Nafion (Aldrich, 5 wt% lower aliphatic alcohols, 15–20% water). Then 5 μL of the photocatalyst suspension was deposited onto glassy carbon electrode.

The photocatalytic study was performed in aqueous glycerol (99% Sigma-Aldrich) solutions using a thermostated gas-flow cell made from a borosilicate glass and adapted to mass spectrometric analysis of the outlet gases. The image of the photocatalytic cell is shown in Supplementary Figure 2. In a typical run, 7.8 mg of photocatalyst was ultrasonically dispersed in 65 mL of aqueous glycerol solution and placed into the photoreactor. Photolysis was carried out using a white light of ASB-XE-175 W xenon lamp equipped with ozone blocking coatings. The lamp was placed at 8 cm away from the reactor and the light power at this distance was measured by X1-1 Optometer (Gigahertz-Optik) using UV-3710-4 (300–420 nm) and RW-3705-4 (400–1,100 nm) calibrated detectors. The obtained values of light power were equal to 8.9 and 0.6 W for vis/NIR and UV spectral ranges respectively, which provides the close spectral match to solar spectra. The solutions inside the reactor were stirred continuously and the temperature was gradually increased up to 95°C during photolysis. The Ar gas flow through the reactor was kept constant at 58 mL·min⁻¹ and controlled by a volumetric flowmeter. The gaseous products in the outlet gas were analyzed using a Thermo Scientific PRIMA BT mass spectrometer. The H₂ formation rate was quantified using external calibration curves prepared with standard gas mixtures in argon (Messer). The water vapors were trapped with molecular sieve (Sigma-Aldrich, 5 Å) prior to mass spectrometric analysis.

TEM images depicted in Figure 1A reveal a quasi-spherical morphology of air passivated Ti particles with an average size of around 30–150 nm without any crystals at the surface. On the other hand, EDX mapping (Figure 2a) indicates the presence of oxygen-enriched layer at the surface of air-passivated Ti particles. This layer with a thickness about 11 nm is also clearly distinguishable in the HRTEM image of Ti NPs particles in Figure 1A. Sonohydrothermal treatment leads to the formation of nanocrystalline shell composed of 10–20 nm oxygen-enriched particles, however, formed core-shell particles preserve quasi-spherical morphology as it is displayed in Figures 1B,C, 2b. We noticed that at the SHT temperature of about 100°C only few nanocrystals are formed, nevertheless their amounts increase with the increase of SHT temperature. The average particle size increases on ca. 5–10% after the coating with TiO₂ compared to initial Ti particles.

XRD diagram of air passivated Ti NPs in Figure 3A displays the patterns of metallic α-phase titanium (JCPDS 00-044-1294) with admixtures of tetragonal non-stoichiometric titanium hydride TiH_x (x = 1.53 – 1.97, JCPDS 01-079-6209) as it often observed in commercial titanium powders (Ageev et al., 1976). However, more thoughtful analysis of XRD data using Rietveld refinement revealed the presence of scarce Ti₃O suboxide with P31c¯ space group symmetry (Figures 3B,C) (Yamaguchi, 1969). To the best of our knowledge, this is the first observation of Ti₃O phase at the nanoscale. It is worth noting that the XRD diagram of Ti NPs does not exhibit the presence of crystallized Ti(IV) oxides. The XRD data of the samples SHT treated at 164 and 214°C clearly point out drop of Ti₃O content and formation of TiO₂ anatase (JCPDS 00-021-1272), which is in line with HRTEM observations (Figure 1B).

High-resolution Ti 2p XPS spectra shown in Figure 4 were fitted using binding energies published in the NIST database (Wagner et al., 2003). The XPS spectrum of air passivated Ti NPs revealed the presence of Ti⁰ and lower oxidation states of titanium at the particle surface in agreement with the XRD analysis revealed the presence of titanium suboxide in Ti NPs. In contrast to XRD data, the XPS spectrum of Ti NPs also exhibits a signal typical for TiO₂ (Figure 4A). This discrepancy could be assigned to the formation of amorphous hydrated titanium oxide TiO₂·xH₂O at the surface of metallic titanium. SHT treatment causes disappearance of Ti⁰, TiO, Ti_xO_y, and Ti₂O₃ peaks from the XPS spectrum and the experimental XPS spectrum of Ti@TiO₂ NPs can be fitted perfectly well by defect-free TiO₂ spectrum indicating effective coating of metallic titanium core (Figure 4B).

Table 1 summarizes the phase compositions of initial Ti particles and prepared materials obtained from the Rietveld refinement of powder XRD data. One can conclude that air passivated Ti NPs are stable during the SHT treatment until ca. 100°C. Further heating until ca. 150°C causes oxidation of Ti₃O to TiO₂:

At higher temperature, metallic titanium is also oxidized yielding TiO₂:

According to the XPS data anatase nanocrystals at the surface of Ti NPs also can be formed by crystallization of TiO₂·xH₂O species. It is worth noting that SHT oxidation of Ti NPs is more effective than hydrothermal heating without ultrasound (Table 1), which was attributed to better heat and mass-transfer in the case of SHT process due to the acoustic cavitation, i.e., formation, non-linear oscillations, and implosion of microbubbles formed in the hydrothermal water under the effect of ultrasonic waves (Cau et al., 2013). Results of TGA analysis displayed in Supplementary Figure 4 and Table 2 point out the increase of titanium metallic core thermal stability with the increase of SHT temperature, which is explained by the formation of protective TiO₂ shell. It should also be emphasized that the hydrothermal oxidation of micrometric titanium powder begins at much higher temperature of about 450°C (Yoshimura et al., 1989), which is most likely related to the particle size effect.

All Ti-based materials studied in this work have intense black color indicating extented photoresponse with nearly full solar spectrum. The solid state reflectance spectra of both Ti and Ti@TiO₂ NPs (Figure 5) exhibit a broad band spanning from UV to NIR spectral range, which was attributed to intraband and interband transitions of metallic Ti (Gawande et al., 2015; Nikitenko et al., 2015). In addition, the spectra of Ti@TiO₂ NPs reveal an absorption band centered at 220–350 nm typical for the bandgap transition of crystallized TiO₂. Kubelka-Munk treatment of the spectral data using (F(R)hv)^1/2 vs E function (López and Gómez, 2012) shown in Supplementary Figure 5 gives a bandgap energy value of 3.52 and 3.51 eV for SHT 164°C and SHT 214°C respectively, which is slightly greater than the bandgap energy reported for commercial anatase TiO₂ powder (E = 3.26 eV) (López and Gómez, 2012). This difference may originate from the overlap of Ti⁰ interband/intraband transitions and TiO₂ bandgap in the UV spectral range (Figure 5). Kubelka-Munk treatment could lead to the overestimated value of the bandgap energy in this case.

The electrochemical impedance spectroscopy (EIS) provides a valuable information about the charge transfer and charge recombination processes at the interface of the catalysts and electrolytes (Barsoukov and Macdonald, 2005). Figure 6 shows the Nyquist plots and the proposed equivalent circuits of the passivated Ti NPs and Ti@TiO₂ NPs obtained after SHT at 101, 164, and 214°C. The reduced size of semicircles indicates a lower charge transfer resistance at the particle/electrolyte interface. Surprisingly, Ti particles coated with TiO₂ nanocrystals exhibit more effective electron transfer than air-passivated Ti NPs or Ti NPs treated at 101°C with very low content of TiO₂ NPs at the Ti metal surface. The resistances calculated for equivalent circuits are summarized in the Table 3. The R₁ refers to the bulk resistance of electrodes and electrolytes, and R₂ represents the resistance at the interface of the particles and electrolytes. A third resistance R₃ is only observed for Ti@TiO₂ NPs and is most likely related to the second interface formed between TiO₂ nanoparticles from the shell and Ti⁰ core. Interesting that R₃ increases with the temperature of SHT treatment, or, in other words, with the increase of TiO₂ anatase content. High R₂ value (9.05 kΩ) for Ti⁰ NPs most probably is attributed to the passivating layer at the metal surface leading to the hindering of charge transfer. Among Ti@TiO₂ NPs, the sample obtained at 214°C displays the lowest R₂ value indicating the fastest charge transfer rate for this material. This conclusion is in an agreement with an effective charge separation in Ti@TiO₂ NPs recently demonstrated using photoluminescence spectroscopy (Nikitenko et al., 2018).

Typical hydrogen emission profiles and calculated H₂ yields for studied photocatalysts shown in Figure 7 highlight the crucial role of TiO₂ nanocrystalline shell in the reaction of photocatalytic glycerol reforming. Kinetic data reveal a clear relationship between the temperature of SHT treatment, or, in other words, between anatase content in Ti@TiO₂ NPs, and H₂ yield (Figure 7B). In addition, photocatalytic activity correlates with EIS data exhibiting more effective charge transfer for SHT 214°C NPs. The role of TiO₂ in the reaction mechanism can be understood in terms of the charge separation between semiconducting TiO₂ shell and metallic Ti⁰ core. The optical spectra (Figure 5) show that the incident light of the Xe lamp in UV/vis/NIR spectral range is mainly absorbed by Ti⁰ core. In non-plasmonic metal, such as titanium, electron-hole pairs can be created by interband transitions via non-radiative Landau damping mechanism (Kale et al., 2014). However, in a highly conducting, metallic material, the electron-hole recombination would be extremely rapid in the absence of hole and/or electron scavengers. Therefore, one might suggest that in Ti@TiO₂ core-shell NPs TiO₂ provides effective charge separation, what was confirmed by fluorescence spectroscopy in previous study (Nikitenko et al., 2018). In addition, TiO₂ can absorb at least part of the UV incident light followed by electron-hole pairs generation in semiconducting particle. In this case, charge separation would be provided by electron migration to Ti⁰ core, as it occurs in TiO₂ photocatalyst loaded with noble metal nanoparticles (Ma et al., 2014; Ghosh, 2018).

Furthermore, photocatalytic process with Ti@TiO₂ NPs exhibits strong photothermal effect in an agreement with our previous results (Nikitenko et al., 2018). Figure 7B demonstrates that the yield of H₂ increases in ca. 4 times when the bulk temperature increases from 37 to 95°C. On the other hand, in dark conditions formation of H₂ is not observed even at 95°C indicating the photonic origin of studied process. Kinetics of H₂ formation obeys an Arrhenius law (Supplementary Figure 6) and the calculated apparent activation energy, E_act, for found to be equal to 31 ± 5 and 25 ± 5 kJ·mol⁻¹ for SHT 164°C and SHT 214°C, respectively. It is worth noting that these values are much lower than the typical activation energy of chemical bonds and has been previously assigned to the diffusion of intermediates at the catalyst surface (Nikitenko et al., 2018). Alternatively, the temperature dependence of photocatalytic reactions can be also related to the dynamics of water adsorbed at the surface of the catalyst (E_act = 16–21 kJ·mol⁻¹, Parrino et al., 2017).

Figure 8 shows the influence of glycerol concentration on H₂ formation with the most active SHT 214°C Ti@TiO₂ photocatalyst. The yield of H₂ increases with the glycerol concentration and then levels off at [Glycerol] ≥ 0.5 M in an agreement with Langmuir-type function indicating that the mechanism of photocatalytic process involves glycerol adsorption at the active sites of the catalyst. Similar behavior has been reported for photocatalytic glycerol reforming over Pt/TiO₂ (Jiang et al., 2015) and NiO/TiO₂ (Fujita et al., 2016) NPs. On the other hand, kinetics with Ti@TiO₂ NPs reveals H₂ emission even from pure water at elevated temperature (Supplementary Figure 7). This observation is largely in agreement with the above conclusion that both water and glycerol can contribute to the observed photothermal effect.

Mass spectrometric measurements indicate the absence of CO₂ emissions (Supplementary Figure 8) during the photothermal process with glycerol in consonance with our previous study (Nikitenko et al., 2018). It is worth noting that photocatalytic glycerol reforming in the presence of nanocrystalline P25 TiO₂ (83% of anatase and 17% of rutile) leads to CO₂ and CO formation (Liu et al., 2014), which implies a certain difference in reaction mechanisms with Ti@TiO₂ and TiO₂ photocatalysts. We found that in studied system H₂ formation is accompanied by acidification of photolyte from pH = 6 to pH = 4 in a ca. 3 h of photolysis indicating formation of some acidic products without decarboxylation. It has recently been suggested that glycerol is oxidized to glyceric acid in the studied process (Nikitenko et al., 2018). In addition, it is known that the wet catalytic oxidation of glycerol can also lead to the formation of other carboxylic acids without decarboxylation, such as tartronic acid, mesooxalic acid, and hydroxypyruvic acid (Worz et al., 2009).

On the basis of the above data, it can be concluded that the mechanism of H₂ formation in studied system involves two reaction pathways illustrated in Figure 9: splitting of water molecules (Equations 3–5 presuming four-electron process) and glycerol reforming (Equations 6–12):H₂O splitting:

Glycerol reforming:

where RCOOH in the Equation (11) could represent a mixture of mentioned above carboxylic acids. It is worth mentioning that the direct water splitting is practically observable at the sufficiently high temperature only. In general, higher yield of hydrogen for glycerol compared to water was attributed to more effective hole scavenging by glycerol than by water (Jiang et al., 2015, Tang et al., 2008). The link between surface charge carrier dynamics and photocatalytic activity of Ti@TiO₂ NPs is in line with this conclusion.

Finally, the stability of Ti@TiO₂ photocatalysts at studied conditions was tested using HRTEM and ICP-OES techniques. Modification of the particle's morphology was not detected by HRTEM after photothermal experiments as shown in Supplementary Figure 9. ICP-OES analysis has been performed with SPECTRO ARCOS instrument (detection limit of titanium was ca. 0.1 ppm). Particles of catalyst have been removed from the reaction mixture by filtration through 0.2 μm PTFE filter prior analysis. The ICP-OES analysis after 8 h of photocatalytic experiments in hot 0.5 M glycerol solutions revealed a leak of about 0.09 and 0.29% of titanium for SHT 214°C and SHT 164°C Ti@TiO₂ NPs, respectively.