Report

The Roles of Subsurface Carbon and Hydrogen in Palladium-Catalyzed Alkyne Hydrogenation

See allHide authors and affiliations

Science  04 Apr 2008:
Vol. 320, Issue 5872, pp. 86-89
DOI: 10.1126/science.1155200

Abstract

Alkynes can be selectively hydrogenated into alkenes on solid palladium catalysts. This process requires a strong modification of the near-surface region of palladium, in which carbon (from fragmented feed molecules) occupies interstitial lattice sites. In situ x-ray photoelectron spectroscopic measurements under reaction conditions indicated that much less carbon was dissolved in palladium during unselective, total hydrogenation. Additional studies of hydrogen content using in situ prompt gamma activation analysis, which allowed us to follow the hydrogen content of palladium during catalysis, indicated that unselective hydrogenation proceeds on hydrogen-saturated β-hydride, whereas selective hydrogenation was only possible after decoupling bulk properties from the surface events. Thus, the population of subsurface sites of palladium, by either hydrogen or carbon, governs the hydrogenation events on the surface.

A fundamental understanding of catalytic reactions requires a “bottom-up” approach using surface-science experiments and theoretical calculations to provide insights into surface dynamics and adsorbed species, their coverage, and possible reaction paths. For simple heterogeneous catalytic systems (such as ammonia synthesis and CO oxidation), the level of characterization approaches the point where rational insights into the design of the catalyst become feasible (1). Experimental studies have been aided by instrumental modifications that allow many surface-science techniques that typically are used under high-vacuum conditions to be used at modest pressures of reactants over the catalysts.

The presumed simplification of most studies is that the bulk of the catalyst, the part of the active sample below the surface, is “frozen” and can be neglected. Although heterogeneous catalytic conversion is a surface process, there is accumulating evidence, particularly from experiments applying in situ functional analysis, that the bulk and especially the subsurface region (the few layers below the surface) can play a key role in surface events. Reaction conditions (such as temperature and the ambient reactive gas) may not only reconstruct the top surface layer, but also may create added rows and valleys of atoms (2, 3) or even massively change the whole morphology of the catalytic particles (4). Atoms that are part of the catalytic feed can dissolve in metallic particles, and can change the electronic structure of the surface (5, 6), and dissolved species can even participate in the reaction (7). Here, we present a still relatively simple but industrially relevant case, alkyne hydrogenation on palladium, in which different subsurface species define which of the possible reaction paths dominates the overall reaction.

Many studies have addressed the question of how to selectively hydrogenate a certain functional group. A hydrocarbon with a carbon-carbon triple bond can be hydrogenated partially to the alkene or fully to the alkane. We have explored why palladium can selectively hydrogenate alkynes to alkenes, when Pd itself is usually even more active in hydrogenating the corresponding alkene to the alkane. The typical explanation (8) is that the difference in the heat of adsorption of the feed alkyne and of the partial hydrogenation product alkene forces the intermediate product alkene to desorb and become replaced by the incoming alkyne of the feed. However, 30 years ago, Al-Ammar et al. (9) demonstrated that, contrary to thermodynamic factors, ethylene could be adsorbed on a catalyst of Pd supported on silica while acetylene was present in the gas phase. This is possible because the surface of catalysts is usually heterogeneous and can have discrete sites that facilitate selective adsorption. Furthermore, many research groups have suggested that carbonaceous deposits formed during reaction might substantially affect selectivity [(10, 11) and references therein]. Alkyne hydrogenation usually goes through an activation period (fig. S1), which strongly suggests that the catalyst is not identical to its “as-introduced” form.

We recently found that, under conditions that selectively hydrogenate 1-pentyne, the active state of Pd is a Pd-C surface phase (PdC), approximately three Pd layers thick (12). The amount of C incorporated within the top layers was 35 to 45 atomic %. This identification was mainly based on x-ray photoelectron spectroscopic (XPS) investigation under hydrogenation conditions (high-pressure XPS). Although XPS is typically used in ultrahigh-vacuum conditions, recent developments in instrumentation have made it possible to investigate catalysts under reaction conditions (13), provided that their reaction proceeds in the millibar range (10–4 to 10–2 atm). Selective 1-pentyne hydrogenation, and alkyne hydrogenation in general, follows the same formal kinetics at these reduced pressures as it does at higher-pressure conditions (fig. S2); thus, in situ XPS will detect the catalytically relevant surface state. Because PdC was found to be stable only in the reaction ambient (12), its investigation requires in situ characterization techniques. To relate C incorporation into Pd to hydrogenation selectivity, we performed 1-pentyne hydrogenation experiments in the millibar pressure range, varying the experimental conditions to induce modification in the product pattern.

The selectivity response of unsupported Pd black (mean particle size 230 nm) to the variation of hydrogen partial pressure, as well as the XP spectra of Pd 3d core levels (Pd 3d5/2), are shown in Fig. 1. Increasing H2 pressure accelerated selective hydrogenation, but above a certain pressure, total hydrogenation occurred. Pd, before contact with the hydrogenation feed, was in the metallic state (the Pd 3d component at 335 eV).

Fig. 1.

Catalytic 1-pentyne gas-phase hydrogenation as a function of pH2. (A) Selectivity for the two main reaction paths on Pd black; experiments were carried out in a closed-loop circulation setup. Hydrogenation selectivity is a strong function of pH2. Solid and open stars mark pressures at which XPS experiments were carried out. P, pressure; T, temperature. (B and C) Corresponding Pd 3d5/2 XP spectra recorded under hydrogenation conditions for Pd foil and black, respectively, using 720-eV excitation energy. The Pd component at 335 eV corresponds to bulk, metallic Pd, whereas the higher binding-energy peak (dashed line) represents the sum of adsorbate-induced surface core-level shift components and PdC. The reaction selectivity correlates with PdC: Selective hydrogenation occurs when the Pd peak is dominated by PdC, whereas total hydrogenation prevails on Pd containing much less C incorporated in the top few atomic layers. a.u., arbitrary units.

The regime of selective 1-pentyne hydrogenation was characterized by the strong peak at ∼335.6 eV, which included any adsorbate-induced surface core-level shift components and the contribution from PdC. At the applied 720-eV excitation energy, the adsorbate-induced surface component should contribute ∼20% to the total intensity of the peak (14); thus, the dominant part of the newly formed state corresponded to PdC. We used nondestructive depth-profile analysis (varying the energy of excitation and hence varying the depth corresponding to the detected photoelectrons) to verify that the location of the 335.6-eV component was on top, above the metallic Pd (12). However, when hydrogenation became unselective [such as at high partial pressure of H2 (pH2)], much less C was incorporated into the top few layers of Pd, which decreased considerably the Pd 3d component associated with PdC. The state of palladium depended strongly on reaction temperature as well (fig. S3), when the concentration of H2 in the gas phase was near the limit of the phase transition of Pd to Pd β-hydride (15). PdC appeared substantially above the decomposition temperature of β-hydride (fig. S3).

To discover whether PdC formation is a general phenomenon during any alkyne hydrogenation reaction, we carried out further investigation with other alkynes. The compilation in Fig. 2A shows that, with lower-chain alkynes, a similar dissolution of C in the near-surface region of palladium occurred and was indeed a general process of alkynes. Because it is likely that only atoms, rather than molecules, will penetrate into the metal lattice, many alkyne molecules must fragment in the early stage of the reaction, in agreement with numerous studies indicating massive irreversible C uptake at the beginning of any selective alkyne hydrogenation processes [(11) and references therein], and also in agreement with the activation period observed in such systems. The model of Pd during alkyne hydrogenation is summarized in Fig. 2B.

Fig. 2.

In situ Pd 3d5/2 spectra of Pd foil under acetylene, propyne, and 1-pentyne selective hydrogenation (A). The reaction mixture contained ∼0.1 mbar alkyne and 0.9 mbar H2; temperature was in the range of 343 to 353 K. The red curve corresponds to Pd metal and the blue curve represents the sum of adsorbate-induced surface core-level shift components and of PdC. The dashed curve indicates the recorded spectrum. PdC forms with all three alkynes on Pd foil (but also on other Pd catalysts), rendering the Pd samples selective in the gas-phase alkyne hydrogenation processes. (B) Schematic representation of Pd catalysts operating in the selective and unselective alkyne hydrogenation regimes. Blue and orange balls indicate C in PdC and β-hydride in Pd, respectively. Green balls indicate an alkyl group or H. Blue patches symbolize carbonaceous deposits. (For clarity, no C5 adsorbates are depicted on the surface.)

Carbon below the topmost layer of Pd has been observed with scanning tunneling microscopy (16), and allowing C to occupy octahedral subsurface sites was also theoretically verified to be energetically favorable (17). These studies have indicated that, in general, subsurface species can trap surface adsorbates; but, according to Yudanov et al. (17), subsurface C will, for example, weaken the binding energy of CO to Pd. Thus, C dissolved in the top layers modifies the surface electronic structure of Pd that apparently favors partial hydrogenation. Additionally, catalytic properties of Pd complexes in homogeneous hydrogenation are strongly affected by the coordinating ligands (18).

The rate-limiting step of alkyne hydrogenation is usually assumed to be the first H-addition step, deduced mainly from surface-science experiments in C2 chemistry, but this cannot explain why selectivity should be a strong function of subsurface C content. However, the type of hydrogen involved in hydrogenation might be critically influenced by C incorporation. It was experimentally demonstrated (7), and later theoretically validated (19, 20), that bulk dissolved H, being much more energetic than adsorbed surface H, can hydrogenate surface adsorbates upon emerging to the surface. Temperature programmed desorption experiments by Khan et al. (21), however, provided clear evidence that subsurface H strongly enhanced total hydrogenation of acetylene, whereas surface H alone (without any subsurface population) was much more selective toward ethylene.

Provided that the conclusion of these surface-science experiments and density functional theory calculations can be transferred to real catalytic conditions, C incorporated in the top Pd layers will strongly affect the transport of hydrogen, and thus disturb the equilibrium of H between surface and deeper layers and hinder the participation of subsurface H in the catalytic process. To verify this hypothesis, the amount of H dissolved in Pd had to be quantified during the hydrogenation event. Because most of the spectroscopic methods are not sensitive for hydrogen (or cannot be applied in situ), we developed existing prompt gamma activation analysis (PGAA) (22) into an in situ technique by placing a continuous-flow reactor instead of a normal specimen into the neutron beam.

The nuclei of the chemical elements can capture neutrons, and during the de-excitation, they emit characteristic prompt gamma radiation. Hence, the H atoms dissolved in Pd could be analyzed with PGAA. When a fresh Pd black sample was introduced into flowing H2 at room temperature, the H content was 0.75 H per Pd atom (Table 1), in perfect agreement with the Pd/H phase diagram at 1 bar of H2 (15). The phase diagram indicates that β-hydride should contain 0.73 H (PdH0.73). However, when the H content was recorded after alkyne hydrogenation events, the ratio was slightly higher, 0.87 on average, which may indicate that carbonaceous deposits simply contain additional hydrogen. When running unselective 1-pentyne hydrogenation (H2/C5 >7), the H content is slightly higher, which means (i) that the reaction proceeds on saturated βhydride and (ii) that additional deposits and adsorbates should carry even more H. This finding validates the idea that bulk-dissolved and subsurface H are very reactive but unselective species and, furthermore, that equilibrium between surface and bulk is maintained during total hydrogenation.

Table 1.

Steady-state atomic H/Pd ratios of a 7-mg Pd black sample under H2 flow or under 1-pentyne hydrogenation, according to the in situ PGAA. 1-pentyne flow was kept constant at 1.6 cm3 min–1. Temperature was near room temperature, except during high conversion, when the temperature rise of the adiabatic reactor was up to 10 K. Because the sample had low dispersion (2%, the fraction of surface atoms), and the gas-phase contribution of H has been subtracted after careful background experiments, the value in pure H2 approaches the bulk H/Pd ratio corresponding to β-hydride.

View this table:

However, when 1-pentyne was hydrogenated at low H2/C5 ratios (<5), after a room-temperature H2 pretreatment, hydrogenation was selective and the H/Pd ratio was low (0.15 on average). Thereafter, further selective 1-pentyne hydrogenation experiments always gave similar low H/Pd values. This value should arise as a sum of H in the adsorbates, hydrocarbonaceous deposits, and the low-concentration α-hydride phase. However, if selective hydrogenation was achieved directly after total hydrogenation by decreasing the H2 flow (at constant 1-pentyne flow), the H/Pd ratio—measured not directly in the transient but after at least half an hour on stream—was always high (0.72, averaged from nine experiments). These higher values were not accompanied by any substantial changes in the activity or selectivity of the sample (fig. S4), which indicates that the reaction rate in the regime of selective hydrogenation is not affected by the amount of dissolved H or by the presence of bulk β-hydride.

The PGAA experiments show that, during selective alkyne hydrogenation, the H content of palladium is not a direct function of the actual reaction mixture but reflects the prehistory of the sample, and also show that the surface properties are necessarily decoupled from the bulk. This separation is made possible by the propensity of Pd to fragment part of the hydrocarbon feed in the early stage of the reaction and dissolve the product C atoms in the topmost few layers. The high concentration of dissolved carbon excludes H from populating the subsurface region and hence prevents total hydrogenation of alkyne.

Having established the relation between selectivity and subsurface chemistry, and considering the metastability of PdC (12), we considered whether the system could be brought to a state sensitive enough to induce selectivity fluctuations. Activity fluctuations in simple reactions (such as CO or H2 oxidation) on the nanoscale can be observed that arise from coverage fluctuations, hence, by pure surface origin, facilitating the formation of surface chemical waves and different spatiotemporal patterns (23, 24). If the kinetics of H subsurface diffusion, C dissolution, and the surface reaction comes to a period during which the population of the subsurface changes, spontaneous switching between the different hydrogenation regimes without modification of the experimental conditions is expected. Figure 3 indicates fluctuations observed during the PGAA experiments, in which the selectivity of 1-pentyne hydrogenation switched between pentane and 1-pentene formation at almost full conversion.

Fig. 3.

Spontaneous fluctuation of the reaction selectivity in 1-pentyne hydrogenation over 7 mg of Pd black during the in situ PGAA experiment with 16 cm3 min–1 of H2 and 1.6 cm3 min–1 of 1-pentyne flow. S, selectivity. Triangles indicate temperature. Conversion levels were always >95%. The reactor was operated adiabatically, at near room temperature. (The lines are only a guide for the eyes.)

According to PGAA results, the H content in this experiment was PdH0.86 (measured over the whole time on stream). The relatively high H2 (H2/C5 = 10) ratio of the feed and the quick H diffusion to the bulk ensured that unselective hydrogenation dominated the initial period. Because C dissolution during total hydrogenation proceeds at a low rate, it takes a relatively long time until a critical amount of subsurface C accumulates to turn the hydrogenation selective. (If the rate of C dissolution is too low, no PdC will build up, because H hydrogenates away the C, and the hydrogenation remains unselective.) Because hydrogenation is exothermic, and the reaction was run adiabatically, the surface temperature when selective hydrogenation activated decreased (the thermo element reading indicated a drop of temperature from 303 to 301 K), which changed slightly the kinetics of both C and H dissolution. Because C dissolution, requiring alkyne fragmentation, is expected to be more activated, H dissolution should be more favorable at slightly lower temperatures and turn the selectivity to favor alkanes again, which makes switching back and forth possible. Considering the positive effect of defects on decreasing the switching time between possible bistable regions in pure surface-related reactions (25) and the complex parameter field of hydrogenation (such as partial pressures, temperature, surface morphology, and support effects), the possible operational window of selectivity fluctuation is likely to be narrow and requires adiabatic operation. However, kinetic discontinuities [that is, switching between two distinct activity (and sometimes selectivity) regimes] observed in a few other hydrogenation systems (26, 27) can be related to an analogous origin.

Although gas-phase alkyne hydrogenation on palladium catalysts is a surface process, we have shown that the population of the subsurface region by either C or H will determine the surface events. This result suggests that not only the surface but also the subsurface region is affected by the chemical potential of the reaction mixture. Because both H and C are part of the hydrogenation feed, the chemical potential creates feedback circles, in the form of H or C dissolution, superimposed on the surface event and performs a major role in the selectivity of hydrogenation.

We are aware that many other factors, such as promoters in the form of a second metal or selective poison, can strongly modify the hydrogenation selectivity. Our aim was to shed some light on the importance of subsurface chemistry in hydrogenation processes. We believe that to take the next step toward rational catalyst design, a critical level of understanding of both surface and subsurface dynamics in these and other complex processes of heterogeneous catalysis is required.

Supporting Online Material

www.sciencemag.org/cgi/content/full/320/5872/86/DC1

Materials and Methods

Figs. S1 to S5

References

References and Notes

View Abstract

Navigate This Article