Hydride Formation Diminishes CO2 Reduction Rate on Palladium.

The catalytic hydrogenation of CO2 includes the dissociation of hydrogen and further reaction with CO2 and intermediates. We investigate how the amount of hydrogen in the bulk of the catalyst affects the hydrogenation reaction taking place at the surface. For this, we developed an experimental setup described herein, based on a magnetic suspension balance and an infrared spectrometer, and measured pressure-composition isotherms of the Pd-H system under conditions relevant for CO2 reduction. The addition of CO2 has no influence on the measured hydrogen absorption isotherms. The pressure dependence of the CO formation rate changes suddenly upon formation of the β-PdH phase. This effect is attributed to a smaller surface coverage of hydrogen due to repulsive electronic interactions affecting both bulk and surface hydrogen.

Introduction

Catalytic hydrogenations are one of the most important processes in chemical industry. Supported metallic palladium is used for hydrogenation of organic compounds such as acetylene and diene impurities in ethylene and propylene feedstock on a Mt scale, and in various other fine chemical syntheses.1,2 Palladium has a unique combination of properties that can partially explain its catalytic activity in hydrogenation reactions. It is chemically stable, there is no significant barrier for hydrogen dissociation, and it has suitable enthalpies of chemisorption for hydrogen.

In addition to its unique surface properties Pd can take up large amounts of hydrogen.3,4 The amount of bulk hydrogen depends on pressure and temperature, which is usually summarized in pressure‐composition isotherms as shown in Figure 2.5,6 Whether its peculiar property of reversibly absorbing large amounts of hydrogen near ambient conditions is relevant for catalysis, has long since been a controversial debate: In general, the surface properties of palladium are similar to the ones of platinum, which hardly absorbs hydrogen in the bulk.7 A difference in catalytic activity may thus be attributed to the different bulk hydrogen sorption behaviour.8 However, catalytic reactions are contingent on details of the surface properties, and these are different between Pd and Pt rendering the comparison inconclusive. Early experiments have shown that bulk hydrogen in Pd can serve as a source of surface hydrogen, which becomes relevant when the external hydrogen supply is switched off.9 This effect is straightforward, as the possibility of atomic hydrogen absorption must include the kinetic exchange of hydrogen between gas, surface and bulk, and vice versa:

Figure 2

Hydrogen pressure‐composition isotherms using the Pd black catalyst measured under catalysis conditions at three different temperatures. The light blue curve, measured under pure hydrogen at 180 °C demonstrates that there is no influence of the CO2 on the pcT measurement. The black curves represent reference measurements of Frieske and Wicke [6] on Pd‐foil.

The asterisk denotes species adsorbed on a surface. The key question arises, whether this exchange influences the catalytic properties in general. Recent studies suggest that subsurface hydrogen plays an important role in H/D exchange and selective alkyne hydrogenation.10, 11, 12, 13, 14, 15 However, these experiments have been performed on well defined Pd‐substrates at low pressure (  mbar) and/or low temperature (  K). To answer the question experimentally, we developed a setup, which is able to vary and determine the bulk hydrogen content and simultaneously measure the conversion yield of the reaction of choice under catalytically relevant conditions:

It is important to note that the objective of the experiments described herein are not intended to develop a particularly good CO2 reduction catalyst but to show the direct correlation between bulk hydrogen concentration c and surface coverage describing the number of adsorbed hydrogens . The determination of hydrogen in the bulk is usually performed by gas analyzers based on volumetric changes. This is not possible in combination with catalysis experiments, as the latter influences the volume of the gas, too. Alternatives are indirect methods such as X‐ray diffraction and X‐ray absorption spectroscopy (see, e. g., Ref. [16] and references therein). In this paper, we describe a setup based on the gravimetric hydrogen analysis using a magnetic suspension balance probing the amount of hydrogen in the bulk at various pressures, temperatures and gas compositions, while an infrared spectrometer is used for gas analysis ideally suited to quantify CO2 and CO without being sensitive to the hydrogen background.

We show that upon total pressure increase the catalytic activity of Pd continuously increases, but changes its slope when entering the hydride phase. The steep increase of hydrogen uptake in the two phase regime has practically no influence on the conversion rate. We postulate that the main effect of hydrogen below the surface is its negative influence on the amount of surface hydrogen due to repulsive electronic interactions at high hydrogen concentrations.

Results and Discussion

By measuring the mass change of the sample and exhaust gas composition simultaneously, we are able to quantify the amount of hydrogen in the sample under reaction conditions (p up to 15 bar), as well as the conversion rate of CO2 to CO (for details see section 5).

The process of hydrogenation and dehydrogenation is fully reversible as evident from the raw data in Figure 1. There is also no measurable mass change of the catalyst after hydrogen/CO2 exposure indicating no potential irreversibly adsorbed products/intermediates such as carbon (coke). The spikes in the CO (black) and CO2 (light blue) concentrations correspond to the abrupt pressure changes. Because the back pressure controller (BPC) is placed after the reactor chamber, the gas flow will be decreased during pressure increase. This leads to longer residence times of CO2 in the reactor chamber and therefore a higher conversion rate to CO. The opposite effect is observed during pressure decrease. These effects are more distinct with larger pressure changes. There is a prolonged phase of increased conversion during the phase transformation from α‐PdH to β‐PdH (plateau region). This is attributed to the uptake of hydrogen by palladium which results in smaller gas flow and therefore longer residence times. The opposite effect is seen during hydrogen desorption although it is not as pronounced. For the equilibrium effects discussed below, only data during steady‐state conditions has been considered.

Figure 1

Left: Scheme of the combined gravimetric and spectroscopic measurement setup. Right graph: Raw data collected showing the applied gas pressure (orange), the mass change of the catalyst (green), the IR peak intensities of CO (black) and CO2 (light blue).

Pressure‐Composition Isotherms of Hydrogen in Palladium

The characteristic pcT curves correspond to three different phases of the metal‐hydrogen system: the solid solution of hydrogen in the metal (alloy); the two phase regime (the plateau), in which the metal phase and the hydride phase co‐exist; and the defective hydride phase (see Figure 1). At small hydrogen to metal ratios (H/M 0.1) hydrogen is dissolved (solid‐solution, α‐phase) in the Pd. The metal lattice expands proportionally to the hydrogen concentration by approximately 2 to 3 Å3 per hydrogen atom.3 At greater hydrogen concentrations (H/M>0.1, slightly depending on temperature) the hydride phase (β‐phase) nucleates and grows. The hydrogen concentration in the Pd hydride phase is around H/M=0.7. The β‐phase starts with a high defect concentration; and the defects are eventually occupied. While the solid solution and hydride phase co‐exist, the isotherms show a flat plateau, in which the hydride phase grows at the expense of the solid solution phase. The chemical potentials of all phases are equal:

A change in the chemical potential μ results from either a change in pressure p or temperature T:

This means that in equilibrium, surface coverage and bulk concentration depend exclusively on the gas pressure. In this paper, we simplify the system by neglecting that in addition to chemisorbed hydrogen ( ) and bulk hydrogen ( ) additional hydrogen states at subsurface sites exist.4,8,16 As the gas pressure is constant in the plateau, the surface coverage does not change, despite the marked changes of the hydrogen concentration from approximately to . However, it is worth repeating the physical origin of the existence of a hydride phase:3,17 the terms describe the ideal solubility at low hydrogen contents (Sieverts law). The non‐ideal or excess term is . It is zero at and becomes negative due to attractive elastic hydrogen‐hydrogen interactions (see Figure 4). As the lattice sites are becoming occupied, the decreasing Pd−H interaction due to filling of the d‐band outweighs the elastic interaction and increases.18 This is also the reason the plateau ends at , although there is space for more hydrogen atoms on interstitial sites.

Figure 4

Concentration dependence of the excess enthalpy μ in PdH . At low concentrations, the elastic interaction dominates, at high concentration (in the hydride phase) the repulsive electronic interaction increases. Adapted from [18].

It is reasonable to postulate that via the electronic interaction bulk hydrogen affects the surface hydrogen (compare Figure 4, see also discussion in 2.3). To account for this effect, we assume . is the concentration independent enthalpy of chemisorption. Analogous to the bulk, we further assume that is an increasing attractive potential for the solid solution, and a decreasing one for the hydride phase. It follows that the bulk H‐surface H interaction strengthens hydrogen chemisorption on α‐Pd, and weakens the chemisorption of hydrogen on β‐PdH. Indeed, experiments using H−D mass spectrometry found a smaller binding energy of H on the surface of β‐PdH than the one on α‐Pd.8,19

This effect may change the amount of bulk hydrogen too: in small particles the ratio of surface to bulk hydrogen is larger than the one in bulk systems. Thus the contribution of the surface energy to the total energy is enhanced in nano‐systems. A number of papers exist demonstrating the narrowing of the two‐phase regime (i. e. plateau‐length) with nano‐structuring.20 As hydrogen chemisorption is much stronger than bulk absorption, the presence of surface hydrogen suppresses the absorption on bulk sites. The here studied Pd black may already be considered as a nano‐system (particle size as estimated from BET‐surface is around 10 nm), as small deviations of the pcT from the bulk system are present (see Figure 2, compare Refs. [4, 20, 21]). Additionally, equation 4 predicts a plateau only if the Maxwell construction is applied. In small particles the absorption plateau might be as high as the pressure corresponding to the lower spinodal concentration.4

Bulk Hydrogen in Pd Affects CO2 Reduction on Pd

The catalytic reduction of CO2 proceeds via various reaction steps. Each of these steps may be rate limiting, e. g., the dissociation of the corresponding molecules. In traditional catalysis research, one tries to overcome these barriers. In this paper, we use a model system suited to study the influence of bulk hydrogen on surface reactions, because of the reversible and measurable amount of bulk hydrogen, as well as the low dissociation barriers for CO2 and H2 on its surface.22 However, Pd is not perfectly suited for CO2 reduction, because the residual times of CO2 and CO are too short to allow for high turnover. The situation can be improved by depositing Pd particles on a support. Here, Pd acts as the active element, and the support acts as the reservoir, referred to as spillover (Ref. [23] and refs. therein). For example, Pd supported on SiO2 was reported to catalyse CO2 reduction to methanol at moderate rates.24 Interestingly, the authors observed a change of reaction rate at a specific pressure, which coincides with the corresponding plateau pressure of PdH at that temperature. The study of such systems may be of interest to applied catalysis research, but the identification and quantification of surface and bulk species, in particular hydrogen, is difficult due to the various chemically different surfaces. In our model system Pd black, any chemical change such as ad/absorption can be related to the Pd particles. Hydrogen absorption by Pd is very well studied (compare discussion above), and there is some knowledge on surface hydrogen, although studies on coverage and dynamics of surface hydrogen at pressures above 1 bar are scarce. Our experiments show that the hydrogen pressure‐composition isotherms depend only on the partial pressure of hydrogen: pcTs in CO2/H2 mixtures fall on top of the ones measured in pure H2 (see Figure 2). As the pcTs were derived from gravimetric measurements, the amount of adsorbed CO2, and potential reaction intermediates/products is very small. Furthermore, the measurements demonstrate the perfect reversibility of the system. This is not implicit: we performed similar measurements on alloy hydrides such as LaNi5H which decomposes and oxidises during CO2 hydrogenation (to be published) in agreement with various literature studies.25, 26, 27 The pressure‐composition isotherms of Pd exhibit a small hysteresis (Figure 2), which might also affect the catalysis. However, within the pressure step sizes, no hysteresis effects could be observed, and thus we only discuss the effect of increasing pressure on hydrogen content and CO2 reduction.

The amount of CO2 on the surface is negligible. We thus assume an Eley‐Rideal mechanism, in which only hydrogen is chemisorbed at the surface, and CO2 reacts from the gas phase or physisorbed state.28 As the desorption of products is not rate‐limiting, we obtain the following pressure dependence of the CO yield according to equation (5):

The impingement rate of CO2, is directly related to the partial pressure .29,30 The hydrogen coverage, though, is high, and thus the exponent l is small, although it depends on the hydrogen content (see discussion above). During catalysis experiments, we vary the total pressure giving:

The measured pressure dependence of the CO formation rate shows two different regimes (Figure 3). At small pressures l is larger than at high pressure. The change coincides with the phase transition from α‐PdH to β‐PdH. We measure a linear pressure dependence in both regimes at all temperatures studied. We cannot derive the change of the exponent in the pressure dependence (eq. 6), due to limited data points (Figure 3). However, a change in the slope of the pressure dependence is indicative of a change of the exponent l. Assuming an Eley‐Rideal mechanism, we can relate the reduced slope to a decreased number of chemisorbed hydrogen in good agreement with Refs. [8, 19] measured at UHV and ambient conditions, respectively.

Figure 3

The CO evolution rate (circles) shows a change in pressure dependence upon the formation of the hydride phase indicated by the pressure‐composition isotherm (triangles)

CO2‐Reduction on Hydrides

The catalytic hydrogenation of CO2 at surfaces includes the dissociation of hydrogen and further reaction with CO2 and intermediates. As hydrogen dissociation is activated on many metals, and hydrogenation of organic molecules by hydrides such as LiAlH4 is standard practise, the potential use of metal hydrides for the hydrogenation of CO2 has received some attention.31,32 Alanates,33 metal borohydrides,34 and various intermetallic hydrides (MNi5H,27,35 Mg2NiH4,36 ZrCoH 37) were found to be efficient for the direct reduction of CO2 upon decomposition. However, in all systems, oxygen from the reactant and intermediates reacts with the metal surface to form passivating oxides, hydroxides, and even carbonates. This impedes reversibility required for potential uses. Furthermore, a gravimetric analysis as utilized here, and by others, is precluded. The amount of surface hydrogen potentially involved in the reaction is thus inaccessible. CO2 catalyzed on PdH is special: the chemical interaction between Pd and CO2 and oxygen containing molecules, respectively, is too weak to lead to a substantial amount of adsorbed species. This is a prerequisite for the measurements presented here, but is also the origin of a rather low catalytic activity. However, the main result of this paper is that underlying (bulk) hydrogen does not generally support the amount of surface hydrogen due to increasing repulsive electronic interactions. The origin of this interaction is long kown: Introducing hydrogen into palladium creates two new states in the DOS.38 A low lying bonding state and the corresponding anti‐bonding state located just above the metal d bands. During the phase transformation the Fermi level rises above the d bands. It has been shown, that the filling of the d bands decreases the metal hydrogen interaction strength relative to H2.39 This change will shift the equilibrium between H* and H2 towards the latter, resulting in a lower surface coverage . The concept that hydride formation diminishes catalytic performance of hydrogenation reactions was also found for intermetallic phases with palladium.40 In this manuscript, we provide evidence that the bulk hydrogen affects the surface states. The influence of bulk hydrogen on surface properties can be explained indirectly via the induced changes of the electronic structure of the underlying metal. This explanation is valid both for surface and subsurface states. In the alpha phase hydride diffusion of hydrogen atoms is much slower than in the beta phase hydride.41,42 This would indicate a behaviour reverse of what is observed here, which might be another indirect hint for the validity of the given explanation. Access to the corresponding electronic structure changes by hydrogen at high pressures providing direct evidence is possible using membrane‐XPS awaiting more interesting results.43

Conclusions

The pressure‐composition isotherms of the PdH system have been measured by a gravimetric method using a mixture of H2 and CO2. The addition of CO2 has no influence on the plateau pressure of the phase transformation. The pressure dependence of the CO2 hydrogenation exhibits a smaller slope after formation of the hydride phase. It is concluded that the same electronic hydrogen‐palladium interactions responsible for the maximum H/Pd ratio of 0.7 are causing a decrease in the hydrogen surface coverage.

Experimental Section

Instrument Setup

The pressure‐composition isotherms were recorded using a magnetic suspension balance (Rubotherm, Bochum) equipped with an online infrared gas analysis system (Bruker Alpha Spectrometer, 0.8 cm−1 resolution, 7 cm pathlength) under a total gas flow of 100 ml/min (see Figure 1). The balance raises and lowers the sample container, measuring the mass difference every 90 s. This leads to very robust sampling statistics. It is set up in an argon filled glovebox to avoid oxygen and water contamination of the catalysts during loading and unloading.

The IR‐spectra were collected as averages of 5 scans from 4500–1000 cm−1 at 0.8 cm−1 resolution. These parameters lead to approximately 30 s of measuring time per spectrum. The gas flows were set by two mass flow controllers (Bronkhorst EL‐Flow®Select) working in conjunction with a back pressure controller (Bronkhorst EL‐Press Select). This setup records the thermodynamic state of the catalyst and quantifies catalysis products simultaneously.

Experimental Procedure

Palladium black (99.95 %) with a surface area of 40–60 m2/g was purchased from Sigma Aldrich. This corresponds to a particles size of 10 nm assuming spherical particles. Approximately 1 g was loaded into the stainless steel sample holder of the balance. To condition the catalyst sample, it was hydrogenated three times under 10 bar hydrogen (H2 2.5, Messer) at 200 °C.

The sample was heated to the desired temperature under 100 ml/min pure hydrogen and kept under these conditions for 2–3 h. During this stage, the background for infrared spectroscopy was recorded. Then 10 ml/min of carbon dioxide (CO2 4.5, Messer) was added while keeping the total flow constant. Pressure increments were done rapidly. After each increment the system was left to equilibrate for 1 h.

Data Treatment

All data points recorded during non steady‐state conditions were disregarded. A buoyancy correction was performed on the obtained mass change data according to the ideal gas law:

where is the measured mass difference, the averaged molar mass of the gas, p the total pressure, V the volume of sample and sample holder respectively, R the universal gas constant and T the temperature.

The analysis of the infrared spectra was based on peak intensities rather than integrated peak areas due to overlapping signals.44 Due to the high CO2 concentration of approximately 10 vol %, the detector was saturated by the 12CO2 signal intensity. Therefore, the 13CO2 P‐branch line at 2272.1 cm−1 was used. CO was analysed by the R‐branch line of 12CO at 2172.8 cm−1 avoiding signal overlap from 13CO. Because the infrared spectrometer is always at atmospheric pressure, there is no need to correct for pressure differences.

Conflict of interest

The authors declare no conflict of interest.