Selective hydrogenation of acetylene in ethylene rich feed streams at high pressure over ligand modified Pd/TiO₂

Abstract

The selective hydrogenation of acetylene from ethylene rich streams was conducted at high pressure and in the presence of CO over two 1 wt% loaded Pd/TiO₂ catalysts with differing dispersions. Although, the more poorly dispersed sample did not result in high acetylene conversion only a small proportion of the total available ethylene was hydrogenated to ethane. The more highly dispersed sample was able to remove acetylene to a level below the detection limit but this was at the expense of significant proportion (ca. 30%) of the available ethylene. Modification of the catalysts by exposure to triphenyl phosphine or diphenyl sulfide and subsequent reduction at 393 K led to improved performance with increased conversion of acetylene and decreased propensity to hydrogenate ethylene resulting in an overall net gain in ethylene. The higher dispersed sample which had been ligand modified provided the best results overall and in particular for the diphenyl sulfide treated sample which was able to completely eliminate acetylene and still obtain a net gain in ethylene. The differences observed are thought to be due to the creation of appropriate active ensembles of Pd atoms which are able to accommodate acetylene but have limited ability to adsorb ethylene. Sub-surface hydrogen formation was suppressed, but not eliminated, by exposure to modifier.

---

Introduction

The removal of trace acetylene from ethylene streams is a key step in obtaining commercially valuable stocks of the ethylene for polymerization, in particular the production of polyethylene.^1,2 Industrially, carbon monoxide is employed as a competitive adsorbate and is believed to enhance selectivity by lowering the surface availability of ethylene² (and hydrogen³). In addition, CO acts as moderator to ensure that the reaction temperature may be controlled. The presence of CO together with a limited hydrogen supply are, however, linked to increased oligomer or ‘‘green oil’’ formation and accelerated catalyst deactivation.⁴ Palladium catalysts used in this reaction have been modified by alloying, in particular with silver⁵ but other non-precious metal additives have been used.⁶ Non-pgm mixed metal catalysts are also being studied for this reaction.⁷ The origin of selectivity enhancement is attributed to either reduced active ensemble size effects or modified electronic characteristics of the active metal or more favorable hydrogenation to a particular intermediate. Additionally, modification of the Pd may influence the relationship between surface and bulk hydrogen availability with the latter often related to non-selective or over hydrogenation process.^8,9 One key objective of alloying, or surface modification, is to render the Pd catalysts less sensitive to CO, given that changes to surface coverage of this adsorbate can have dramatic consequences on the extent to which acetylene and ethylene are hydrogenated. This may occur either though competitive adsorption with ethylene,¹ or by lowering the availability of hydrogen,³ or by suppressing the formation of subsurface hydride and carbide phases.⁴ Maintaining appropriate, controlled levels of CO is challenging given that the catalysts continuously undergo some form of deactivation by formation of carbonaceous overlayers.^10,11

In the present study, acetylene hydrogenation from ethylene rich steams under conditions of excess hydrogen at high pressure and in the presence of CO is described for two Pd/TiO₂ catalysts. These catalysts have similar loadings but differ in their metal dispersion and have been modified by exposure to either diphenyl sulfide or triphenyl phosphine. It has previously been shown that such ligand modified Pd catalysts lead to enhancements in selectivity for acetylene hydrogenation in the presence of excess ethylene.¹² However, these earlier tests were conducted in a water cooled reactor at atmospheric pressure. Industrial acetylene hydrogenation reactors are generally operated at elevated pressures² and so, to investigate the efficiency of the ligand modified catalysts under more realistic conditions, the catalysts were tested in a reactor operating at a total pressure of 10 bar. In contrast to previous reported data for these catalysts,^12,13 no cooling of the reactor was employed to limit the effect of heat generated during the reaction and, therefore, CO was used in all experiments to ensure the temperature in all cases was controlled at 323 K. Catalysts data is presented here for loadings of 1%Pd although it is recognised that commercial Pd catalysts for purification of alkene streams contain less than 0.05 wt.%Pd, use alpha-alumina as the support, and the active metal exhibits an egg-shell distributions on the carrier. Such low loadings however, lead to extreme challenges in terms of metal particle characterisation and limit attempts to relate catalyst performance and behaviour with characteristics.

Experimental

1 wt%Pd/TiO₂ was synthesized by wet impregnation. Pre-wetted TiO₂ (P25) was contacted with palladium nitrate solution (0.022 g of Pd(NO₃)₂ dissolved in 20 cm³ of de-ionized water) with the latter added using a dropping funnel into a continuously stirred aqueous suspension over a period of 1 h. Once complete, the solution was left stirring at room temperature for a further hour. The flask was then transferred to a furnace held at 393 K for 24 h to remove ca. 50% of the water. The solution was stirred and transferred to a round bottomed flask. This solution was then placed on a rotary evaporator and dried until a fine powder was obtained. The catalyst was stored in a glass vial until calcination was carried out. A second batch of similarly loaded catalyst was prepared by slight method modification which involved extending the period during which the solution and support remained in contact before initiating the drying process and this led to a sample which exhibited a higher dispersion. To distinguish these in text, the samples are denoted Pd(1.7)/TiO₂ and Pd(3.5)/TiO₂ where the numerical value refers to the Pd particle size in nm determined by CO chemisorption. Calcination of the catalysts was carried out in a quartz holder in a flow of air at 100 mL min⁻¹. The temperature was raised to 673 K at ca. 20 K min⁻¹. The temperature was held at 673 K for 3 h and subsequently lowered to room temperature. The calcined Pd/TiO₂ was then reduced at 423 K for 1 h in 50% H₂ in N₂ and then either stored for future use, or exposed to the modifier.

The deposition of the modifiers was carried out from a hexane solution. The amount of modifier was selected to produce a 2.5 to 1 (molar) modifier to (bulk) palladium metal ratio. The catalyst was stirred for 30 min and then dried on a rotary evaporator. The sample was dried under vacuum until a fine powder was obtained and the evaporated solvent collected and analyzed by GC-MS to confirm complete uptake of modifier. Modified samples are denoted using the prefixes, Ph₂S- and Ph₃P- to denote diphenyl sulfide and triphenyl phosphine modified samples, respectively.

Chemisorption measurements were carried out using a TPDRO 1100 (CE instruments). All catalysts were reduced in 5% H₂/N₂ for 1 h and cooled to room temperature in N₂ prior to the pulse chemisorption experiments with unmodified and modified catalysts reduced at 393 K. The pulsing experiments were carried out using a 285 μL sample loop filled with diluted CO (19.7% or 8% in He) being injected to the sample which was held at 293 K in a flow of He. The dispersion and particle size were calculated from the CO uptake assuming a Pd : CO stoichiometry of 2 : 1.¹⁴ This assumption, rather than a 1 : 1 ratio, was supported by FTIR spectra of adsorbed CO which showed¹³ that multiply bound CO made a significant contribution to the compliment of total CO adsorbed. A spherical particle shape was assumed in calculating particle diameter.

FTIR of adsorbed CO was performed using a PE Spectrum 100 FTIR using self-supporting discs of 16 mm diameter using approximately 45 mg of catalyst. The discs were then suspended in a quartz holder and placed in the IR cell fitted with CaF₂ windows. The catalysts were reduced in situ at 393 K for 1 h in a flow of 50% hydrogen in nitrogen. The cell was then evacuated to a pressure of approximately 3 × 10⁻³ mbar. A background spectrum (15 scans, resolution 4 cm⁻¹) was recorded prior to introduction of CO. Subsequent spectra were recorded following introduction of increasing pressures of CO with the samples at beam temperature (ca. 298 K).

Hydrogenation reactions were conducted at 323 K in a high pressure system with on-line GC for product analysis. 30 mg catalyst was pre-reduced for 1 h at 393 K in 50% H₂ in N₂ before introducing a reactant gas mixture comprising 6000 ppm C₂H₂, 30% C₂H₄, 15% H₂ (25 : 1 H₂ : C₂H₂), 100 ppm CO and N₂ to balance at 10 bar total pressure at GHSV ∼3000 h⁻¹. The catalyst temperature was measured by means of a K-type thermocouple placed in the middle of the catalyst bed. A second thermocouple was located on the outside of the reactor at the level of the catalyst bed. The reading difference between the two thermocouples was less than 2 °C, which ensured that real temperature of the catalyst bed was measured and not the exotherm determined by the reaction. The pre-prepared certified gas mixture was purchased from BOC and was controlled by a high pressure Aera PC-7700C mass flow controller connected to Aera ROD-4 controller box. Samples were taken at 40 min intervals and analysed using FID after separation of components on a 3 m long Hayesep N (80/100) column with nitrogen as a carrier.

Results

Nitrogen adsorption desorption isotherms were of type IV with type H1 hysteresis. BET surface areas of all samples, after metal loading and incorporation of modifiers and subsequent thermal treatment were all ca. 52 m² g⁻¹ indicating little or no structural modification to the parent P25 titania support material. The metal loadings of the samples after acid digestion were measured by ICP-MS using Au as internal standard. All batches of catalyst contained between 0.9 and 1wt%Pd. The metal surface areas and derived dispersion data for the metal component using the measured loadings are complied in Table 1. For the two parent 1wt%Pd loaded samples, estimated particle sizes of 3.5 and 1.7 nm were obtained from the CO surface area. Addition of the modifier and subsequent reduction at 393 K had a significant impact on the ability of Pd to adsorb CO under the pulse chemisorptions conditions with the most significant reduction occurring in the case of Ph₂S, irrespective of the Pd dispersion. For the 1.7 nm sample CO uptake values could not be determined suggesting either complete blockage by modifier or its decomposition products, or, more probably,¹³ a weakening of the Pd–CO bond strength due to electronic modification of the Pd atoms.

Table 1 Characteristics of Pd and ligand modified Pd catalysts as derived from CO chemisorption measurements on samples reduced at 393 K

Catalyst	CO uptake/μmol g^−1	Metal surface area/m^2 g^−1 (catalyst)	d/nm
Pd(3.5)/TiO2	13.6	1.28	3.5
Ph3P-Pd(3.5)/TiO2	2.4	0.23
Ph2S-Pd(3.5)/TiO2	0.2	0.02
Pd(1.7)/TiO2	30.5	2.87	1.7
Ph3P-Pd(1.7)/TiO2	7.9	0.74
Ph2S-Pd(1.7)/TiO2	—	—

---

The nature of the Pd surfaces in the presence and absence of ligands and after reduction at 393 K was investigated by using FTIR of adsorbed CO and comparing spectra obtained after exposure to the same pressure of gas. In agreement with the greater CO uptake measured by pulse chemisorption, the Pd(1.7)/TiO₂ showed the greater combined band envelope due to adsorbed CO (Fig. 1A(a)) and this was mainly as a result of greater population of bridging sites as the relative area of the band due to linearly bound CO was almost identical for both samples. The former feature became much narrower for samples which had been modified by triphenylphosphine (Fig. 1B) with significant loss of intensity between 1900 and 1850 cm⁻¹. There was also a notable loss in intensity of the feature at ca. 2080 cm⁻¹ due to adsorption in the on-top mode suggesting that the modifier had a preference to occupy these sites. The on-top mode was also less populated by CO after samples had been modified by diphenyl sulphide (Fig. 1A) as noted by the change in the absorbance scale employed. A second, higher frequency band at 2119 cm⁻¹ was apparent for samples which had been treated by diphenyl sulphide and this would suggest the presence of cationic sites in the Pd surface. This feature appeared irrespective of the Pd particle size. Although the on-top mode was significantly suppressed by the presence of the diphenyl sulfide modifier, the bridging mode was the most affected by treatment with this modifier, in contrast to the observation for triphenylphosphine.

Fig. 1 FTIR spectra of (A) unmodified Pd/TiO₂ (B) Ph₃P–Pd/TiO₂ and (C) Ph₂P–Pd/TiO₂ with particle sizes of (a) 1.7 nm and (b) 3.5 nm after exposure to 40 Torr.

CO chemisorption confirmed that the ligand was, at least in part, located on the Pd surface. In order to determine the stability the ligands to pretreatment conditions, DRIFTS was performed on samples which had received no treatment other than solvent removal after exposure to the ligands and samples which had been ligand treated and then reduced at 393 K. Spectra of the as-prepared sample shows three sharp features at ca. 1585, 1480 and 1440 cm⁻¹ due to the ring stretching modes of the phenyl structures (Fig. 2a). Additional broader bands were found at 1630 (water deformation). Reduction in hydrogen at 393 K led to a material with a spectrum which was not significantly modified compared to the starting material for PPh₃–Pd(3.5)/TiO₂ consistent with the thermal stability of the ligands to temperatures in excess of 500 K.¹⁶ In contrast, this treatment led to modifications in the spectrum of the Ph₂S–Pd(3.5)/TiO₂ sample with considerable reduction in intensity of the 3 maxima due to the ring stretching modes of the phenyl groups suggesting that the treatment had led to the loss of a large proportion of the aromatic structures.

Fig. 2 DRIFTS spectra for (left) PPh₃–Pd(3.5)/TiO₂ and (right) Ph₂S–Pd(3.5)/TiO₂ catalyst (a) as prepared (b) after reduction at 393 K.

Fig. 3 shows the acetylene conversion as a function of time over the first 5 to 8 h on stream. The Pd(1.7)/TiO₂ and Ph₂S–Pd(1.7)/TiO₂ exhibited 100% acetylene conversion throughout this time period. All other catalysts tested showed an initial induction period during which the acetylene conversion increased, reached a maximum and then slowly decreased. The rate of loss of conversion was greatest for the Pd(3.5)/TiO₂ catalyst (Fig. 3) and, consequently, this sample exhibited a maximum conversion of the shortest duration. The difference between the parent catalysts of different dispersion may be attributed to differences in the rate of deactivation due to coke deposition although previous studies show that differences in morphology causes only small changes in the formation of retained hydrocarbons.¹¹ Due to the fact that a constant 100% acetylene conversion was shown by Pd(1.7)/TiO₂ catalyst, it was not possible to assess whether deactivation of this catalyst occurred or not. Other samples which failed to attain 100% conversion but which gave the best conversion levels were Ph₂S–Pd(3.5)/TiO₂ and Ph₃P–Pd(3.5)/TiO₂ although acetylene conversion levels at their maximum were in the range 87–89% at best.

Fig. 3 Acetylene conversion during the first 5 to 8 h on stream for catalysts (30 mg) pre-reduced at 393 K in 1 : 1 H₂ : N₂ for 1 h, then in reaction at 10 bar and 323 K with 6000 ppm C₂H₂, 30% C₂H₄, 15% H₂ (25 : 1 H₂ : C₂H₂), 100 ppm CO, N₂ balance.

Selectivity to the desired product, ethylene, as a function of time during the first 5 to 8 h on stream for each catalyst is displayed in Fig. 4. The lowest ethylene selectivity was exhibited by Pd(1.7)/TiO₂ (values from right hand axis). The catalyst did exhibit a step enhancement in selectivity between 200 and 240 min on stream; however, even after this step improvement, the selectivity, although constant was still very low. This was due to the large proportion of ethylene in the feed-stream being hydrogenated to ethane. The other unmodified sample, Pd(3.5)/TiO₂ performed better, but also exhibited very poor ethylene selectivity. Again this sample showed improved selectivity with time on stream, in particular during the first 2 h, and, thereafter, the selectivity levelled off. This is consistent with the general behaviour expected as carbonaceous deposits are formed with subsurface carbon thought to be responsible for high selectivity in Pd based catalysts.^10,11 All modified samples, irrespective of the parent catalyst dispersion, showed significantly better selectivity to ethylene. In each case, the modified catalysts showed initial selectivities of between 80 and 92% at 40 min on stream. However, all of the modified systems showed an almost linear decrease in selectivity to ethylene as a function of time on stream, albeit while still maintaining better selectivities than the unmodified parent Pd/TiO₂ samples (Fig. 4). Of the modified samples, Ph₂S–Pd(1.7)/TiO₂ showed the greatest decline in selectivity over 5 h on stream with Ph₂S–Pd(3.5)/TiO₂ showing the least.

Fig. 4 Ethylene selectivity during first 5 to 8 h on stream for catalysts (30 mg) pre-reduced at 393 K in 1 : 1 H₂ : N₂ for 1 h, then in reaction at 10 bar at 323 K with 6000 ppm C₂H₂, 30% C₂H₄, 15% H₂ (25 : 1 H₂ : C₂H₂), 100 ppm CO, N₂ balance.

To compare and contrast the performances of all catalysts, including the generation of other by-products, the data at a fixed time on stream (5 h) were assessed and are presented in Fig. 5 and 6. The data are presented in a format where the residual acetylene in ppm is shown together with the net impact on the ethylene present (this may appear as either a net gain or a net loss). Additionally, the total FID signals due to undesired by-products, other than ethane are shown. These by-products were mainly C4 hydrocarbons including butadiene consistent with previously reported results.¹⁵Fig. 5 shows results obtained after 5 h on stream over the 3.5 nm Pd/TiO₂ based catalysts. In contrast to the data obtained under conditions of low pressure and in the absence of CO,¹² the conversion of both acetylene and ethylene were relatively low with 4572 ppm of acetylene remaining of the original 6000 ppm in the reagent mixture when using the unmodified Pd catalyst. On the other hand, the influence of the modifiers was still observed as found with atmospheric pressure tests,^12,13 namely the modifications led to enhanced conversion of acetylene with respect to unmodified catalyst and suppressed the conversion of ethylene. The ethylene gain and acetylene conversion was superior for the phenyl sulfide modified catalysts compared with the phosphine modified materials.

Fig. 5 Remaining acetylene, loss/gain of ethylene and by-product formation after 5 h on stream over a) Pd(3.5)/TiO₂ b) PPh₃–Pd(3.5)/TiO₂ c) Ph₂S–Pd(3.5)/TiO₂. Reaction conditions: 30 mg catalyst, pre-reduced at 393 K 1 : 1 H₂ : N₂ 1 h, P = 10 bar, 323 K, reactant gas contained 6000 ppm C₂H₂, 30% C₂H₄, 15% H₂ (25 : 1 H₂ : C₂H₂), 100 ppm CO, N₂ balance.

Fig. 6 Remaining acetylene, loss/gain of ethylene and by-product formation after 5 h on stream over (a) Pd(1.7)/TiO₂, (b) PPh₃–Pd(1.7)/TiO₂ (c) Ph₂S–Pd(1.7)/TiO₂. Reaction conditions: 30 mg catalyst, pre-reduced at 393 K 1 : 1 H₂ : N₂ 1 h, P = 10 bar, 323 K reactant gas contained 6000 ppm C₂H₂, 30% C₂H₄, 15% H₂ (25 : 1 H₂ : C₂H₂), 100 ppm CO, N₂ balance.

Results for the 1.7 nm Pd/TiO₂ catalysts are shown in Fig. 6 and show a greater variation within the series compared with the Pd(3.5)/TiO₂ based catalysts. The conversion of both acetylene and ethylene was significantly enhanced on the unmodified Pd(1.7)/TiO₂ catalyst compared with the Pd(3.5)/TiO₂ catalyst. Whilst this led to the complete removal of acetylene (Fig. 5), in addition, significant (ca. 30%) loss in the ethylene was also found. This may be compared with only 1.3% loss of ethylene found for the Pd(3.5)/TiO₂ catalyst. Conversely, both of the ligand modified catalysts produced a net gain of ethylene with triphenyl phosphine modification providing the greater benefit. However, this modified catalyst failed to reduce acetylene levels to values which would be industrially acceptable.^1,2 In contrast, diphenyl sulfide modified catalyst not only led to a net gain in ethylene but also led to conversion of acetylene to below detectable limits. In this case, by-product formation was significantly higher than found on the unmodified or triphenyl phosphine modified samples. Although by-product formation over Ph₂S–Pd(1.7)/TiO₂ catalysts was the greatest over all the tested samples, it was noteworthy that the levels of undesired by products were also high over Ph₂S–Pd(3.5)/TiO₂ suggesting that the propensity to form by-product was more closely related to the presence of this modifier, or its residues,¹³ than being directly related to Pd dispersion.

Discussion

Previous studies have indicated that surface modification of Pd/TiO₂ catalysts by P and S based ligands leads to improved performance of these catalysts in the ambient pressure hydrogenation of acetylene from ethylene rich feedstreams.^12,13 The present work suggests that these earlier promising findings can be extended to more practical conditions of high pressure as employed industrially and with similar beneficial effects. The key advantage exhibited by modified samples over unmodified Pd/TiO₂ is the high ethylene selectivity from the initial contact with the reactant feed stream (Fig. 4). This is in contrast with the unmodified samples which showed a progressive improvement in selectivity with time on stream, probably as a result of accumulating levels of carbonaceous deposits being formed.^10,11 The other prerequisite for an effective catalyst, that residual acetylene levels are reduced to concentrations <5 ppm^1,2 was met only by the Ph₂S–Pd(1.7)/TiO₂ and the unmodified Pd(1.7)/TiO₂ (Fig. 3) although the latter did so at the expense of very poor ethylene selectivity. The Ph₂S–Pd(1.7)/TiO₂ met both criteria for an effective catalysts.

It should be noted that the effects of the ligand modified catalysts are observed following a pre-treatment, namely reduction for 1 h at 393 K in 50% H₂ in N₂. The triphenyl phosphine was thermally stable as evidenced in the DRIFT spectra (Fig. 2) and did not undergo decomposition or degradation at these temperatures.¹⁶ This modifier mainly reduced access to sites where CO was adsorbed in the on-top-mode and in the 1900–1850 cm⁻¹ region which might be attributed to the blocking of 3-fold hollow sites. In the case of diphenyl sulphide modification, the ligand was partially decomposed following reduction at 393 K as evidenced by the DRIFT spectra (Fig. 2). This decomposition/reduction process is expected to create a surface partially covered with adsorbed atomic S or HS species,^12,13 It has been previously noted that at 300 K, a chemisorbed layer of S on Pd(111) forms a bulk sulfide phase.¹⁷ Therefore some of the S atoms penetrating the top layer of the metal to form a sulfide like phase would lead to a partial oxidation of the Pd surface and this was evidenced by the appearance of the carbonyl feature at 2119 cm⁻¹ after exposure to CO (Fig. 1C). The ratio of bridged to linear species was different for the surfaces exposed to the different modifiers with values of 2.27 and 2.84 for the 3.5 and 1.7 nm particles modified by triphenyl phosphine but 1.07 and 1.22, respectively for the diphenyl sulfide modified samples.

Given the significant differences in the nature of the surface moieties, it is striking to note the degree of similarity in terms of the impact on catalytic behaviour exhibited by both series of modified samples with respect to the parent Pd/TiO₂ samples. In particular, there are remarkable similarities between Ph₂S–Pd(3.5)/TiO₂ and PPh₃–Pd(1.7)/TiO₂ in terms of the profiles with time on stream for acetylene conversion (Fig. 3) and ethylene selectivity (Fig. 4). Given the notable differences between the nature of the surface modifiers in terms of steric and electronic factors, an interpretation of the origin of the enhanced selectivity based on addition of the modifier might reasonably be made on the basis of creation of ensembles of Pd atoms of appropriate dimensions which are able to readily accommodate acetylene but limit access of ethylene (and amount of hydrogen) to the metal surface. Such a scenario would be analogous to the creation of selective ensembles resulting from carbonaceous deposits,^10,18 or the presence of a second metallic component at the surface.⁶ This postulation has already been made for the modified Pd/TiO₂,^12,13 however, these studies, in addition to having been performed at 1 bar pressure, were conducted in the absence of CO. Consequently, in addition to any selectivity enhancements based on creation of appropriate Pd ensembles, the present selectivities will have been partly determined by the competitive adsorption characteristics of CO,^1,3,4 which were absent in studies conducted at 1 bar. The steric/site blocking effects of the modifiers on the surface may also lead to a reduction in the ability of the catalyst to form subsurface hydrogen. The latter has been linked with over reduction of ethylene to ethane^8,9 and its formation will be reduced by decreasing the overall surface coverage of hydrogen due to the presence of the surface modifiers.

In addition to geometric effects, the adsorbed species may provide an electronic contribution to the Pd surface. Triphenylphosphine can be regarded as an electron donating ligand^19,20 while the high polarisability of the sulfur atom in sulfur moieties determines its nucleophilic activity.¹ Yu and Spencer have postulated that increasing the nucleophilicity of adsorbed hydrogen should increase the selective reduction of alkynes to alkenes²⁰ while Hu et al.¹⁹ have shown that charge transfer from triphenyl phosphine to the metal should promote activation of molecular hydrogen and thus promote the hydrogenation reaction. However, they also argue that this should promote activation of the olefin through donation of electrons to the π* orbital of the C=C.¹⁹ One would envisage that this effect should also enhance the strength of adsorption of other adsorbates where metal to π* back bonding occurs such as in CO. However, experiments using FTIR demonstrated that the presence of either modifier led to a weakening of the Pd–CO bond.¹³ Other arguments based on electronic effects of additives include a weakening of the chemisorption strengths of the acetylene and ethylene,²¹ and it has been suggested that interaction with nucleophilic additives weakens the Pd affinity to an alkene to a greater extent than to an alkyne.¹ Such effects would facilitate the desorption rather than the reaction of ethylene.^21–23 Decreasing the residence time of the ethylene on the surface of the catalysts will also result in a reduction in the ethane formation. Finally, high surface coverage of the modifiers would not only limit the amount of available surface hydrogen but might also have reduced the extent to which sub-surface hydrogen was formed. The latter is often related to non-selective or over hydrogenation process.^8,9 Sub-surface hydrogen formation was suppressed, but not eliminated, by exposure to modifier¹³ and the presence of CO here would also be expected to make a contribution to further suppression of this form of hydrogen.^3,4

As well as the selectivity-activity relationship for these catalysts, the importance of the propensity of the catalyst to form oligomers or “green oil” is important. This leads to catalyst deactivation and ultimate failure of the catalyst bed under certain conditions.¹ Interestingly, although the phosphine and sulfur modified catalyst samples showed similarities for the production of ethylene compared with ethane, dissimilar behaviour was found when they were compared on the basis of the higher molecular weight by-products formed (Fig. 5 column c and Fig. 6 column b). Compared with the parent Pd/TiO₂ catalyst, sulfide modifier increased the gas phase C4 compounds whilst a reduction was found for the phosphine based systems (Fig. 4 and 5). Both sulfide and phosphine modifiers reduced the amount of non-C2 based products when the reaction was conducted at atmospheric pressure and in the absence of CO.^12,13 The presence of C4 compounds has been linked with the formation of “green oil”; however, the presence in the gas phase may suggest two scenarios:

(i) Increased surface concentration of butadiene

(ii) Decreased adsorption energy of butadiene

If the extent of catalyst deactivation and the amount of C4 species in the gas phase are considered together, it can be seen that the sulfide based catalysts, in general, may suppress the formation of C4+ oligomers. As a result, although C4 is increased in the gas phase, the site blocking is reduced and hence the deactivation of these catalysts is the slowest observed. This is consistent with previous reports which have shown that the use of sulfur containing molecules is known to suppress coke formation, possibly by blocking sites where multiple binding of the alkyne molecules takes place.¹ As an impact on the extent of oligomerisation may arise from geometric modifications due to limiting the amount of end on alkyne–alkyne reaction,²⁴ the combination of these adsorbed sulfur moieties and the CO, is likely to have an impact on the extent to which oligomers are formed.¹ It is this combination of surface species which leads to the displacement of butadiene into the gas phase as in the absence of CO, albeit at 1 bar, sulfide modification reduced the amounts of butadiene observed.¹³ It should be noted that CO has an important but probably minor role in this displacement from a comparison of the differences between the parent Pd/TiO₂ and the sulfide and phosphine modified samples.

Overall, the presence of the modifier and the CO in the feed both result in significant changes to the activity-selectivity profile as well as to the extent of deactivation and by-product formation. Performance could also be optimised by tuning the CO feed to map the deactivation profile rather than the constant feed employed here. This complex interaction between modifier, CO and deactivation results in changes in adsorption energy as well as surface coverage which, at this stage, is not possible to provide a general elucidation of the predominant factors which determine the catalytic behaviour. However, it is clear that modification of the palladium surface with sulphides and phosphines provides a method by which the selectivity and deactivation of the catalysts may be controlled under realistic industrial reaction conditions.

Conclusions

Triphenylphosphione and diphenyl sulfide have been shown to be effective modifiers of Pd/TiO₂ catalysts in the selective hydrogenation of acetylene from ethylene rich feedstreams at high pressures and in the presence of CO. Initial ethylene selectivities of between 85 and 92% were attained while in one case 100% acetylene conversion was possible. The combined effect of modifier and CO may have enhanced the ability of the surface to liberate light oligomers and avoid rapid deposition of surface carbonaceous layers.

Acknowledgements

We thank SABIC for financial support of this project, and for permission to publish the data obtained and Dr A. M. Ward (SABIC) for helpful discussions. LM acknowledges funding from the EPSRC through the CASTech programme.

References

1. S. A. Nikolaev, IL. N. Zanaveskin, V. V. Smirnov, V. A. Averyanov and K. L. Zanaveskin, Russ. Chem. Rev., 2009, 78, 231.
2. D. B. Tiedtke, T. T. P. Cheung, J. Leger, S. A. Zisman, J. J. Bergmeister and G. A. Delzer, 13th Ethylene Producers Conference, 2001, 10, 1.
3. D. L. Trimm, I. O. Y. Liu and N. W. Cant, Appl. Catal., A, 2010, 374, 58.
4. M. García-Mota, B. Bridier, J. Perez-Ramírez and N. López, J. Catal., 2010, 273, 92.
5. H. Zea, K. Lester, A. K. Dayte, E. Rightor, R. Gulotty, W. Waterman and M. Smith, Appl. Catal., A, 2005, 282, 237.
6. J. Osswald, K. Kovnir, M. Armbrüster, R. Giedigkeit, R. E. Jentoft, U. Wild Y. Grin and R. Schlögl, J. Catal., 2008, 258, 219.
7. B. Bridier, J. Perez-Ramırez, A. Knop-Gericke, R. Schlogl and D. Teschner, Chem. Sci., 2011, 2, 1379.
8. N. A. Khan, S. Shaikhutdinov and H.-J. Freund, Catal. Lett., 2006, 108, 159.
9. D. Teschner, J. Borsodi, A. Wootsch, Z. Revay, M. Havecker, A. Knop-Gericke, S. D. Jackson and R. Schlögl, Science, 2008, 320, 86.
10. A. Borodziski and G. C. Bond, Catal. Rev. Sci. Eng., 2008, 50, 379.
11. A. Sárkány, A. Horváth and A. Beck, Appl. Catal., A, 2002, 229, 117.
12. F.-M. McKenna, R. P. K. Wells and J. A. Anderson, Chem. Commun., 2011, 47, 2351.
13. F.-M. McKenna and J. A. Anderson, J. Catal., 2011, 281, 231.
14. J. A. Anderson, M. Fernandez-Garcia and A. Martinez Arias, in Supported Metals in Catalysis, ed. J. A. Anderson and M. Fernandez-Garcia, ICP, London, 2004, p. 123.
15. I. Y. Ahn, J. H. Lee, S. S. Kum and S. H. Moon, Catal. Today, 2007, 123, 151.
16. F.-M. McKenna, R. P. K. Wells and J. A. Anderson, manuscript in preparation, 2011.
17. C. H. Paterson and R. M. Lambert, Surf. Sci., 1987, 187, 339.
18. A. S. McLeod and R. Blackwell, Chem. Eng. Sci., 2004, 59, 4715.
19. G. Hu, Z. Feng, D. Han, J. Li, G. Jia, J. Shi and C. Li, J. Phys. Chem. C, 2007, 111, 8632.
20. J. Yu and J. B. Spencer, Chem. Commun., 1998, 1103.
21. F. Studt, F. Abild-Pedersen, T. Bligaard, R. Z. Sørensen, C. H. Christensen and J. K. Nørskov, Angew. Chem., Int. Ed., 2008, 47, 9299.
22. J. H. Kang, E. W. Shin, W. J. Kim, J. D. Park and S. H. Moon, Catal. Today, 2000, 63, 183.
23. S. Tracy, A. Palermo, J. P. H. Vazquez and R. M. Lambert, J. Catal., 1999, 179, 231.
24. B. Bridier, N. López and J. Pérez-Ramírez, Dalton Trans., 2010, 39, 8412.

---