1.

INTRODUCTION

Hydrogen is ubiquitous in the technical as well natural environments and is thus involved in many reactions crucial for life and technology. Despite its relevance as well as abundance, its detection and chemical characterisation remains challenging particular, when studying surfaces, of, e.g., catalysts. The standard characterisation is X-ray photoelectron spectroscopy, capable of reliably detecting and quantifying all elements on a surface apart from hydrogen (and helium).¹ In this paper, we show that although photoemission spectroscopy reveals important information of hydrogen in the technically important catalyst support TiO2. This output relies on the use of a special operando technique (Fig. 1), and the use of tunable soft X-rays from a synchrotron enabling resonant photoemission measurements.^{2, 3} The study showcases a methodology for hydrogen induced changes in matter with high potential, considering the progress in the development of table top tunable X-ray sources.⁴

Figure 1.

Sketch of the Pd-membrane approach for high pressure hydrogen photoemission spectroscopy.^18–20

A typical catalyst consists of metal nano-particles on oxide supports. Unlike the name may suggest, the oxide support is involved in the reaction. Since long, the so-called strong-metal support interaction is used to describe the affinity of a metal-support system for hydrogen⁵ with Pt-TiO₂ as an archetypal example.⁶ A recent publication highlights that the ability to chemisorb hydrogen and the hydrogen mobility on the oxide surface is a crucial property of the support,⁷ which is defined by the reducibility of the corresponding oxide.^{8, 9} A similar behavior is found in corrosion, where the inertness of an oxide layer on a metal against hydrogen is the key parameter for achieving high corrosion resistance.¹⁰ Finally, the efficiency of the production of renewable hydrogen depends on the electro-catalytic properties of the electrodes used, and in case of photocatalytic water splitting on the light absorbance of the photoanodes/cathodes. The photocathode is unavoidably in contact with highly reactive hydrogen. This hydrogen interaction may change materials properties markedly, which has been found to be detrimental in WO₃, but beneficial in NiO¹¹ and TiO₂.^12-14

The modification of photocatalytically active materials such as TiO₂ by hydrogen is particularly interesting,¹⁵ because both the optical properties as well as chemical properties change. The corresponding scientific questions are: with hydrogen exposure, are the changes confined to the surface, or is hydrogen migrating into the bulk of material? Does hydrogen react with the oxide forming water, and leave vacancies behind, or does it form hydroxyl groups? Although the questions are long known, and a multitude of studies were published tackling them, see, e.g., the excellent review of Rahimi et al.,¹⁶ a concise picture is missing. This is partly due to the different preparation techniques. In this paper, we account for the latter uncertainty by following the changes induced by hydrogen operando with state-of-the-art resonant photoemission.¹⁷ For this, we employ a membrane approach to high pressure surface science, which consists of a hydrogen permeable membrane (Pd), onto which a thin TiO₂ film is deposited (Fig. 1). Applying a hydrogen pressure up to one bar at the feed side results in the nearly equal chemical hydrogen potential at the TiO2 thin film while keeping UHV conditions necessary for photoelectron spectroscopy experiments.¹⁸ The initial defects such as vacancies and hydroxyl groups are unavoidable, but intercalating hydrogen in-situ reduces the number of undetermined parameters potentially introduced by subsequent ex-situ hydrogenation. Furthermore, TiO₂ does not dissociate hydrogen very well, which would pose a high barrier for hydrogen interaction, if the surface was simply exposed to molecular hydrogen at UHV-compatible pressures. In our experiment, hydrogen is from atomic hydrogen diffusion through the Pd membrane into the thin TiO₂ layer (Fig. 1).

2.

EXPERIMENTALS

The high pressure X-ray photoelectron spectroscopy (XPS) study is based on the Pd membrane approach sketched in Fig. 1.^18–20 To maintain the pressure in the analysis chamber in the UHV-compatible range, the surface under investigation is not exposed directly to high pressure hydrogen, but hydrogen diffuses into the corresponding thin film from the backside. For this, the thin film is deposited onto a hydrogen selective Pd-membrane. If hydrogen permeation is rate-limiting, which is likely due to the low dissociation probability of hydrogen on TiO2, the hydrogen chemical potential at the surface (in UHV) complies that of the feed side.²⁰

The Pd membranes were coated with a TiO₂ layer by atomic layer deposition (ALD) at a substrate temperature of 100°C with argon as carrier gas at a base pressure of 13 Pa in a Fiji G2 system (Veeco Instruments Inc.). The precursors were tetrakis(dimethylamino)titanium(IV) (TDMAT) (Sigma Aldrich) and H₂O. TDMAT was kept at 75°C while H₂O was unheated. The growth rate was determined by ellipsometry on Si(100) reference substrates and compared to SEM micrographs which showed a similar thickness with a larger uncertainty. For TiO₂ linear growth was observed with a growth rate of 0.053 nm/cycle.

The photoelectron spectroscopy measurements were carried out at the GasPhase beamline of the Elettra Synchrotron light source in Trieste (Italy). The beamline is equipped with a dedicated differential pumping system.²¹ The combination of a high resolution monochromator with the high transmission of the electron analyzer allowed to collect spectra with a minimal energy resolution of 50 meV.²²

The synchrotron measurements were recorded using 650 eV photon energy for the survey spectra and 160 eV for the valence band. During hydrogenation experiments, the photoemission spectra were recorded continuously in single sweep mode, with each spectrum taking approximately 120 s. The applied hydrogen pressure was measured by an external pressure gauge. As soon as the partial hydrogen pressure in the chamber was constant, the external hydrogen pressure was increased. This lead to approximately ten spectra being recorded for each external hydrogen pressure. These spectra were averaged for the subsequent data evaluation to improve signal to noise ratios. Data analysis was performed using the CasaXPS software. Due to the significant workfunction W_F shift during hydrogenation, the binding energy (and intial energy) were recalibrated using the Ti 2p and 3p peaks as given by Goepel et al.²³ Photoionization cross sections to estimate the oxygen titanium elemental concentrations assuming similar spectrometer transmission functions for the Ti2p and O1s transitions were taken from Yeh and Lindau. ²⁴

3.

RESULTS

Figure 2 shows synchrotron XPS spectra (ħω = 650 eV) of TiO₂ on a Pd membrane at various conditions. The spectra are typical for ALD grown TiO₂ with a small amount of Ti³+,¹⁷ as derived from fitting of the shoulder of the Ti 2p^3/2 peak. The shoulder of the O 1s peak at 533.5 eV is attributed to OH species.²⁵ Minimum temperature was 120°C to remove most contaminants from air, in particular water. However, carbon contaminants could not fully be removed. The valence band spectra have a relatively high noise, but all main features such as the gap states are visible.

Figure 2.

Synchrotron XPS spectra (ħω = 650 eV) O 1s, Ti 2p and valence band from TiO₂ on Pd membrane as grown (blue, T = 120°C, рн₂ < 10^–7 mbar), exposed to hydrogen (green, T = 120°C, рн₂ = 1000 mbar), and hydrogen removed at slightly higher temperature (red, T = 210°C, рн₂ < 10^–7 mbar). The color code is retained in the following figures. From the fitted intensities, we derive the Ti³⁺/Ti⁴⁺ ratio, the ratio of OH to lattice oxygen, and the oxygen to titanium elemental ratio shown in Fig. 6.

Both photoionization cross section and the photon flux of the beam line are higher at lower excitation energies, which give higher resolutions of the valence band as shown in Fig. 3. The valence bands of the ALD grown TiO2 film correspond to TiO2 with many defects. The character of these defects may be derived by comparison with literature. The reference spectra in Fig. 3 were measured on heavily Ar-sputtered crystalline TiO₂, i.e., TiO₂ with oxygen vacancies. However, in ALD-grown TiO₂, OH defects are more likely.^{17, 26} Strong indication of this assumption is the occurrence of band gap states. The corresponding shallow trap states (STS) near E_F increase with applying hydrogen, but decrease with annealing at 200°C in vacuum. Annealing in vacuum and beam damage leads to deep trap states (DTS, see Fig. 3).²⁶

Figure 3.

Synchrotron valence band spectra (ħω = 98 eV). Sample conditions (color code) are the same as in Fig. 2. The small inset shows literature valence spectra.²³ The sketch visualizes the electronic structure of TiO₂ with hydrogen induced (OH) shallow trap states and oxygen vacancy induced deep trap states. The intensities are compared to Ti³⁺/Ti⁴⁺ ratio and the oxygen to titanium elemental ratio in Fig. 6.

The character of the band gap states are further analyzed using resonant photoemission (Fig. 4). A resonant photoemission experiment is the measurement of the kinetic energy of photoelectrons at a set of various excitation energies.^{2, 3} These are then plotted in a 2D map as shown in Fig. 4 for excitation energies around the Ti-L edge. Pioneering work with this techniques was performed by Das et al.¹⁷ The authors demonstrated that the integration over all kinetic energies as a function of the excitation energy gives an X-ray absorption spectrum (electron yield XAS). Employing this on the 2D maps in Fig. 4 gives spectra (Fig. 5), which are very similar to XAS-spectra of anatase derived from total electron yield X-ray absorption spectroscopy.²⁷ The differences - in particular the broadening is due to the amorphous crystalline state in good agreement with literature.¹⁷ In a simple view, the XAS-spectra represent the conduction band of TiO₂, which consists of (even more simplified) only unoccupied Ti 3d states. The formation of Ti³+ states will move some of these states to the valence band. This is visible particularly clearly at 458 eV, and may used to quantify the ratio of Ti³+ to Ti⁴+ ratio.²⁸ The changes of the ratio observed by XAS are in good agreement with ones determined by XPS, i.e., both techniques evidence a Ti³+ / Ti⁴+ ratio, which depends on the applied pressure and temperature.

Figure 4.

The color map (left panel) displays the number of photoelectrons excited from a TiO₂ sample as function of their kinetic energy (recalculated as initial state energy) and excitation energy. Although the initial state energy is defined as given in the figure, in practice the energy was aligned as described in the experimental section. The 2D map can be simplified by integrating over the full kinetic energy range yielding X-ray absorption spectrum (top right panel). More details are found plotting the initial state energies at constant excitation energy (right bottom and middle panels). The excitation energies used are indicated in the 2D map by colors. The sketch visualizes the electronic structure of TiO₂ and possible electronic transitions near the Ti L-edge. The shallow trap states are strongly enhanced in resonance conditions, i.e., Ti2p electrons are emitted to the conduction band and occupy subsequently trap states, in good agreement with the experiment.

Figure 5.

Left panel: XAS-spectra from integrating the resonant photoemission over the whole kinetic energy range. The anastase reference is from Ref.²⁷ Right panel: Ti-resonance valence band spectra of TiO₂ at different conditions. The sample conditions (color code) are the same as in Fig. 2. The intensity of the shallow trap states follow the same trend as found by XPS and VB spectroscopy.

However, the resonance spectra contain additional information. An extensive review on of the various effects is given by Das et al.,¹⁷ but goes beyond the scope of this paper. We would like to focus on the band gap states, which can be assigned to the shallow trap states already found by XPS- and VB spectroscopy. These state vary strongly with excitation energy (Fig. 4), are either in or out-off resonance with Ti2p (L-edge) excitation.³ This is very convincing evidence that the shallow trap states are Ti³+ states. Furthermore, the intensity of these states depend on the sample state also under resonance conditions, i.e., follow the same trend as found by XPS and VB spectroscopy: increasing with hydrogen exposure, but vanishing in vacuum at higher temperature (Fig. 5).

4.

DISCUSSION

Before in-depth discussion of the photoemission results, it is worth reviewing some general facts about the hydrogen - oxygen - titanium system. The enthalpy of formation of solid TiO₂ and TiH₂ are -938.72 kJ/mol and -160 kJ/mol, respectively.^{29, 30} TiO₂ may thus considered to be a non-reducible oxide, as also titanium hydroxides do not exist in solid form. The surface may be covered with hydroxyl groups,^{29, 31} though. However, hydrogen causes defects in the bulk associated with marked materials changes.⁸ In this respect it is more reducible than, e.g., Al₂ O₃.^{7, 9} Some potential defects in TiO₂ are sketched in Fig. 6, such as oxygen vacancies, intercalated hydrogen on an oxygen site, and OH in the bulk and at the surface of TiO₂. The number and character of these defects is debated, partly due to the very low concentration. From IR-absorption experiments,³² a maximum hydrogen defect concentration of the order of 10¹⁵ to 10¹⁶ cm^–3 was detected in plasma hydrogenated ТiO₂, which corresponds to ppm concentrations (H per ТiO₂).³² Herklotz et al. found mainly OH defects, although this might be due to the limitation of the IR technique. However, calculations support this finding.³³

Figure 6.

The left graph shows the dependence of the gap state intensity on the OH to lattice oxygen ratio, to the oxygen to titanium elemental ratio, and to Ti³⁺/Ti⁴⁺ ratio, as derived from Figs. 2 and 3. The right sketch illustrate possible vacancies and defects in TiO₂ and Pd induced by hydrogen. See text for details.

Particular interest sparked so-called black ТiO₂, usually nano-sized titania, prepared by exposing to hydrogen at pressures of several bar and elevated temperature (> 200°C).³⁴ Defects in ТiO₂ are known to color titania crystals,²⁹ but black ТiO₂ absorbs light particularly strong in the visible and infrared wavelength region.³⁴ Superior photocatalytic properties have been demonstrated on the material.¹⁵ Mo et al. determined the hydrogen content of black titania as a function of the applied pressure, and found a particularly high hydrogen content of 100 ppm at 0.95 bar of hydrogen and an annealing temperature of 600°C using NMR, but simultaneously an oxygen loss of 300 ppm.³⁵ Similar to low hydrogen content ТiO₂, the character of hydrogen induced defects/vacancies is debated. However, Chen et al. claim that Тi³⁺ defects or impurities are not responsible for the absorption in black ТiO₂, in particular they found a contrary behavior of Тi³⁺ states as derived from Ti2p peaks and gap states being responsible for the absorption.³⁴ This is a verifiable observation. Figure 6 shows the dependence of the gap state intensity on the OH to lattice oxygen ratio, to the oxygen to titanium elemental ratio, and to Ті³⁺/Ті⁴⁺ ratio. Clearly, the behavior is in sharp contrast to that described by Chen et al.³⁴ Furthermore, the on/off switching with and without Ті-resonance is evidence for the involvement of Ті-electrons.

Chen et al. put another idea into play: hydrogen induces a disordered phase on the surface (region) of titania.³⁴ This idea has charm, because the gap state intensity should not be regarded as the number of defects, but as the number of electrons present in the trap states, which depend more on their excitation and lifetime than on the intrinsic number of defects.¹⁴ The ALD grown films in this publication (and others) have initially high disorder,²⁶ which is increased by hydrogenation, but cured upon annealing in vacuum. The lifetime of electrons in STS depends thus directly on the hydrogenation state. Furthermore, the formation of other trap states (DTS) will open additional decay channels further shortening the lifetime (recombination centers). The additional parameters are then only spectators indirectly indicating a change of the crystallinity / disorder of the TiO₂. The unexpected behavior of Ti³+ amount and O/Ti ratio can be understood that the removal of hydrogen defects by lowering pressure and/or annealing is a complex process, in which curing the crystal (reducing disorder) by hydrogen desorption without oxygen loss competes with the generation of oxygen vacancies by removing hydrogen and oxygen (water formation). The latter is enhanced by charge damage under UV/X-ray illumination.²⁶

The results are representative for this topic in materials science: difficult to detect hydrogen at often low concentrations causes drastic materials changes, and the response of the systems depend on the specific materials preparation. Such research must rely on various at best simultaneously applied characterisation methods, and the repetitive and defined materials preparation. This is not limited to research associated with hydrogen,³⁶ but equally important in battery research,³⁷ and many other topics in materials science. The success of XPS, also coined ESCA (electron spectroscopy for chemical analysis)¹ is based on this demand. However, the paper at hand is one example that the standard XPS alone is not sufficient to detect hydrogen. The extension to resonant photoemission has potential to become a standard characterisation techniques, if reliable, small and cheap table top tunable X-ray sources are available. Various principles of such table top sources have been demonstrated,⁴ and the advent of commercial system combined with photoemission is in sight. Furthermore, the utilisation of X-ray lasers may enable new techniques giving deeper insights into the materials science of hydrogen.

5.

CONCLUSIONS

The use of the operando membrane approach allowed us to follow the complex behavior of hydrogen interaction with TiO₂ by resonant photoemission. Striking effect upon hydrogenation is the formation of defects in materials, which alter optical as well catalytic properties. We showed that the shallow trap states in ALD grown titania on Pd-membranes are Ti³⁺ electrons. The intensity of these states is related to the hydrogen treatment as well as Ti³⁺/Ti⁴⁺ ratio evidence that STS are caused by hydrogen intercalation into TiO₂. However, the main effect of hydrogen seems not to alter the number of defects, but the disorder it creates in the material. This study offeres a perspective for future use of tabletop X-ray lasers (see also Sects. 1-4 [38])