Palladium with interstitial carbon atoms as a catalyst for ultraselective hydrogenation in the liquid phase

Abstract

We report a green route to prepare Pd with interstitial carbon atoms as a new solid catalyst for fine chemical catalysis in the liquid phase. First, glucose was used as a reducing and capping agent under hydrothermal conditions for the controlled reduction of the Pd precursor and after thermal treatment, carbon atoms were found to occupy the Pd subsurface interstitial sites, as confirmed by PXRD and TPR. A simple hydrogen peroxide treatment was required to partially remove surface carbon to increase catalytic activity. Catalytic properties of the final material were investigated in the hydrogenation of 3-hexyn-1-ol and 4-octyne to the corresponding alkene products and compared with a classical Lindlar type catalyst and other commercial Pd/C. It was found that the carbonised glucose Pd nanocatalyst with subsurface carbon substantially reduced undesirable isomerisations and over-hydrogenations. This was achieved by the increase in desorption rates of alkene species in solution, an effect believed to be the result of hybridisation of the Pd d-state with the C sp-state, thereby increasing the overall cis-alkene selectivity.

---

Introduction

Fine chemicals manufacture is an area of major importance within the chemical industry with an estimated market value of $75 billion in 2007.¹ The applied areas include production of speciality products such as active ingredients in pharmaceuticals, dyes, fragrances and agro chemicals. The target compounds are often highly substituted, with high levels of complexity and therefore have limited stability and volatility, hence reactions are typically carried out in the liquid phase. Fine chemicals manufacturing generally requires multiple batch steps, using a variety of production equipment, as compared to bulk chemicals that typically use a continuous process. Due to the production time and the added value, they are produced in limited quantities (<10 000 t) and sold at a high price (>10 USD kg⁻¹).

Recently, increasing attention has been paid to monitor the quantity of waste produced in this field. The E-factor is typically used to quantify the level of waste generated, with each type of reagent being assigned an environmental quotient (EQ).² As seen in Table 1, the E-factor increases from bulk to fine chemical synthesis. This is largely through the use of stoichiometric reducing and oxidising agents such as NaBH₄ and KMnO₄ that also have the disadvantage of being toxic. This poses limitations from both economical and environmental viewpoints regarding both product purification and waste management. Therefore, applications that use green chemistry are favoured, benefiting from more efficient protocols by employing non-toxic and non-waste producing reagents. For example, the use of molecular hydrogen for chemical reductions and hydrogen peroxide for oxidations is typically preferred. The ease use of solid catalytic systems to facilitate separation of product solutions must also be considered in the development of novel environmentally aware protocols.

Table 1 E factors in various chemical industries²

Industry	Annual production/tonne	E (kg of waste/kg of product)
Bulk chemicals	Less than 10^4–10^6	Less than 1 to 5
Fine chemicals	10^2–10^4	5 to more than 50
Pharmaceuticals	10–10^3	25 to more than 100

---

Heterogeneous catalytic hydrogenation of alkynes to alkenes represents one of the most important steps in fine chemicals manufacture.³ It has been widely established that heterogeneous palladium-based catalysts are highly effective for this hydrogenation reaction, affording high alkene selectivities. Furthermore, the catalytic activity and selectivity can be enhanced by using amines/polymers,^4,5 shape control,⁶ support effects⁷ and bimetallic⁸ systems. Recent work by Teschner et al. demonstrated that the formation of subsurface C in Pd during catalysis greatly promotes the selectivity towards alkenes, in the gas phase.⁹ However, it has not been possible to isolate this active species with interstitial subsurface carbon atoms only being formed under transient conditions at temperatures under an atmosphere of reactive gas. Therefore, it is highly desirable to develop a Pd analogue modified with subsurface carbon atoms, which is stable under atmospheric conditions. It is anticipated that such a material would prove highly beneficial for liquid phase hydrogenations of alkyne to alkene, one of the key steps in the pharmaceuticals, fragrances and fine chemical catalysis.

Thus, we report a green route using glucose, which acts as a reducing and capping agent as well as a supporting matrix for the Pd nanoparticles that are formed. After heat treatment, the glucose derived carbon atoms encapsulating Pd particles reside in the octahedral interstitial sites of the metal, hence blocking the formation of β-hydride under reducing conditions. The Pd particles that are geometrically and electronically modified by the interstitial carbon atoms demonstrate the ability to substantially reduce undesirable processes such as alkene over-hydrogenations and isomerisations in the liquid phase. The alkyne reductions are stereoselective, occurring via syn addition affording mainly cis-alkenes with this new class of stable ‘Pd–C’ catalysts.

Experimental

Catalyst preparation

The glucose encapsulated palladium nanocatalyst was prepared using a 300 mL Parr autoclave high pressure compact reactor series 5500 equipped with overhead mechanical stirring, gas inlet and outlet valves, a liquid sampling valve, a pressure gauge, a safety rupture disc, an internal cooling loop and an internal thermocouple connected to a 4836 series temperature controller (Fig. 1).

Fig. 1 Schematic diagram showing the setup for the in situ synthesis of glucose-coated (as encapsulating spheres) metal nanoparticles in the aqueous phase by hydrothermal treatment. (A) Temperature and stirrer controller, (B) heater, (C) pressure gauge, (D) overhead stirrer, (E) autoclave reactor, (F) flow-meter (exit to atmosphere), (G) nitrogen inlet, V₁,V₂: high-pressure 2-way valves, V₃: a high-pressure 3-way valve, (H) flow-meter (gas inlet).

A solution of palladium(II) nitrate, 15.11%, (1.006 mmol, Johnson Matthey) was added to a 90 mL deionised water solution of D-(+)-glucose (50.0 mmol, Aldrich) in a PTFE coated beaker. The pH was adjusted to the desired value by dropwise addition of 1 M NaOH solution. After assembly, the reactor was heated to 180 °C for 4 h and stirred at 800 rpm. The pressure reading from the autoclave at 180 °C was approximately 12 bar. The reactor was allowed to cool to room temperature where a black solid was isolated, confirming the growth of carbon materials during the hydrothermal treatment. This was filtered and washed with a copious amount of water followed by ethanol until the filtrate was colourless. The material was then dried for 18 h under a steady stream of nitrogen followed by calcination at 400 °C in a nitrogen atmosphere at a ramp rate of 10 °C per min with a dwell of 1 h. Oxidation to partially remove the amorphous carbon was achieved by stirring in a 50 : 50 v/v mixture of hydrogen peroxide (30 wt% in H₂O, Aldrich) and ammonium hydroxide solution (28.0–30.0% NH₃ basis, Aldrich) for 4 h at 20 °C. This oxidised catalyst was filtered and subsequently washed with water and ethanol until the filtrate was colourless and dried. This yields the final glucose encapsulated Pd nanocatalyst.

Catalyst characterisation

Powder X-ray diffraction (PXRD) patterns were obtained on a PANalytical X'Pert Pro diffractometer operating at 40 kV and 30 mA by pressing the sample onto an aluminium preparative slide. Fourier transform infrared (FTIR) spectra were acquired using a Nicolet 6700 ATR-IR spectrometer with a liquid-nitrogen-cooled MCT detector. The solid samples were pressed onto the smart golden gate-ZeSe/diamond crystal. The spectra were obtained by averaging 128 scans with a resolution of 4 cm⁻¹ over the wavenumbers ranging from 650 to 4000 cm⁻¹. Transmission electron microscopy (TEM) was undertaken on a Tecnai F20 microscope using a voltage of 200 kV, and a C₂ aperture of 30 and 50 m, in bright field (BF) STEM (HAADF) mode with in situ EDX analysis. The samples were crushed and dusted onto a holey carbon coated Cu grid. Thermogravimetric analysis (TGA) was performed using a TA instruments Q50 thermogravimetric analyser. Specimens were loaded onto a platinum sample pan and heated from RT to 1000 °C under nitrogen at a ramp rate of 10 °C min⁻¹ and holding at the final temperature for 1 h. Inductive coupled plasma emission spectroscopy (ICP-ES) was conducted using a Perkin Elmer Optima 3300RL instrument. Samples were prepared by Na₂O₂ infusion in a zirconium crucible and compared against standard gravimetrically determined solutions. Temperature programmed reduction (TPR) was performed using a CE instrument TPDRO 1100. Samples were pretreated in H₂ at 100 °C for 1 h, then cooled down to RT. Analyses were performed in a flow of 5% hydrogen in argon at 20 mL min⁻¹ with a linear ramp rate at 2 °C min⁻¹ from room temperature to 400 °C. X-Ray photoelectron spectroscopy (XPS) analyses were performed using a Thermo V. G. Scientific ESCALAB 250. Powders of the supported nanoparticles were dusted onto carbon tape and thereby mounted on a sample stub. The exciting radiation used in the studies was monochromatised aluminium Kα radiation in a 650 nm spot at 200 W power. Charge compensation was activated, provided by the in-lens flood gun at a 2 eV setting and the 401 argon ion flood source at a “zero energy” setting.

Catalyst testing

Hydrogenations of 3-hexyn-1-ol were performed in a 100 mL Parr autoclave. To a PTFE lined beaker, the catalyst (0.028 mmol) was added followed by a pre-prepared standard solution of 3-hexyn-1-ol (30.0 mmol, Alfa Aesar) and the internal standard 1,4-dioxane (30.0 mmol, Alfa Aesar). The autoclave was flushed 5 times with hydrogen and charged up to 3 bar at 30 °C with a mechanical stirring speed of 400 rpm. Reaction progress was monitored using a hydrogen uptake apparatus. The sample was collected at the end of the reaction for GC analysis.

Hydrogenations of 4-octyne were performed in a standard 100 mL 3-neck round bottom flask. 50 mL of toluene was added followed by the catalyst (4.70 μmol) whilst maintaining a constant nitrogen purge. Agitation was induced using a glass coated stirrer bar. The suspension was sonicated to disperse the particles. Hydrogen was bubbled through the suspension via a needle value at a flow rate of ∼30 mL min⁻¹ at 50 °C. A hydrogen purge was maintained for a further 30 min before 4-octyne (9.38 mmol, Aldrich) and an internal standard p-xylene (2.51 mmol, Aldrich) were added to initiate the reaction. The molar ratio of the catalyst to the substrate was 1 : 2000. Samples were taken out at specific time intervals to probe the reaction kinetics. 0.5 mL aliquots were withdrawn and filtered through a cotton wool plug to separate the catalyst and diluted using 0.5 mL toluene to clean the filter. Samples collected were analysed by gas chromatography.

Results and discussion

Palladium encapsulated in the carbon spheres with tunable diameters were synthesised from aqueous glucose solutions at temperatures of 160–180 °C.¹⁰ Under these conditions, glucose readily undergoes glycosidation and carbonisation, with the growth of the carbon spheres confirming to the LaMer model.¹¹Fig. 2 shows FTIR spectra used to identify the functional groups present after hydrothermal treatment. The bands centred at around 3000–3500 cm⁻¹ are assigned to the O–H stretching, with glucose exhibiting greater intensity compared to the Pd encapsulated in the carbonised glucose. The bands at 1000–1300 cm⁻¹ are attributed to the C–O and O–H bending vibrations, implying a large number of hydroxyl groups for glucose. The peaks at 2950 cm⁻¹ for glucose are the C–H stretches and the bands at 1690 and 1600 cm⁻¹ are assigned to >C=O and >C=C<, respectively. The evolution of these peaks verified the aromatisation of glucose after the hydrothermal treatment.

Fig. 2 FTIR spectra in transmission mode of: (a) glucose, (b) Pd encapsulated in carbonised glucose.

Table 2 shows the particle size analysed by PXRD calculated using the Debye–Scherrer equation. The Pd particle size was large under the initial conditions. However, this can be tuned by varying the pH of the hydrothermal solution. This alters the binding strength of glucose to the palladium. Ikushima et al. reported on the formation of Au and Pt nanocrystals using NaBH₄ reduction which stabilised β-D-glucose into nanowire like structures. They found that nanoparticle dispersions can be stabilised for prolonged periods by varying the pH.^12,13 The underlying mechanism was attributed to the pH-dependent equilibrium between surface –OH and –O⁻ on the metal surface. At higher pH, relative to the pK value of the hydroxyl groups, the deprotonated form –O⁻ predominates and vice versa. This enhances the effective binding of the metal. Therefore, the effect of pH in the hydrothermal treatment was investigated. It was found that highest pH leads to dramatic decrease in Pd particle size from 25 to 5 nm, in agreement with the literature.

Table 2 The effect of pH on the Pd particle size analysed by PXRD after hydrothermal treatment

pH	Pd particle size/nm
1.8	25.2
4.3	18.8
6.9	5.9
10.0	5.0

---

Fig. 3 and Table 3 show the heat treatment of the Pd glucose nanocatalysts and subsequent weight loss observed using TGA. Heat treatments at various temperatures were used to investigate the stability of the synthesised nanocatalysts and the binding ability of the carbonised glucose to Pd. Analysis showed that the carbonised Pd catalyst has increased stability as compared to glucose, which is characterised by the weight loss at higher temperatures. PXRD was used to investigate the Pd particle size of the heat treated samples and any evidence of carbon atom diffusion into the bulk of Pd (Fig. 4). A mild growth in the Pd crystallite was observed from 5 to 7 nm for the sample heated to 400 °C, but this was accompanied by a peak shift to lower 2θ, which became increasingly evident at higher angles. This is consistent with previous reports by Zlemecki and Jones that the lattice parameter, a₀, of the bulk Pd with carbon occupying interstitial octahedral holes has increased from 0.3890 to 0.3995 nm (Table 4).¹⁴ The amount of interstitial carbon occupied in Pd can be calculated using eqn (1):

a = a₀ + 6.9 × 10⁻² × x, (1)

where a₀ is the lattice parameter of pure Pd.^15,16 This represents carbon modified palladium with the formula Pd_1−xC_x (independent of size) where carbon saturation is reached at 15%. The occupation of C atoms in the tetrahedral sites may be ruled out as this was calculated to be too unstable and confirmed by Zlemecki et al. using neutron diffraction studies of differing amounts of doped carbon atoms.¹⁷ The occupation of the bulky O atoms can also be ruled out, as the ionic radii of O atoms are too large to occupy these interstitial octahedral or tetrahedral sites, which causes destabilisation to the Pd lattice.¹⁸ Pd_1−xC_x is a metastable phase and it is evident for the sample treated in H₂ at 400 °C, where the interstitial carbon atoms are more unstable than other surface carbons, that these can be expelled from the metal lattice. It was found that interstitial carbon can also be expelled when heating in N₂ at 500 °C or higher, forming the most thermodynamically stable pure Pd phase. PXRD also suggests that the carbonised glucose is amorphous in nature as no graphitic peaks were observed. The sample calcined to 400 °C was studied further due to the maximum subsurface carbon being achieved at this temperature. Furthermore, a severe metal sintering effect encountered at more elevated temperatures undermines the usefulness of such catalysts.

Fig. 3 TGA spectra showing percentage mass remaining versus temperature: (a) glucose, (b) 400 °C, (c) 500 °C, (d) 700 °C, (e) 900 °C, (f) 1000 °C.

Table 3 Mass remaining at various final temperatures by TGA

Final temperature/°C	Remaining mass (wt%)
400	70.3
500	60.4
700	53.8
900	46.0
1000	42.7

---

Fig. 4 XRD patterns illustrating the effect of thermal treatments on the phase and Pd particle size for samples heated to: (a) 400 °C, (b) 500 °C, (c) 600 °C, (d) 700 °C, (e) 900 °C, (f) 1000 °C. Star: aluminium preparative slide. Triangle: pure Pd peaks corresponding to the fcc structure. Circle: Pd–C peaks. Pd phases from left to right correspond with phases 〈111〉, 〈200〉, 〈220〉, 〈311〉 and 〈222〉, respectively.

Table 4 The effect of calcination temperature with respect to the variation in the average lattice parameter of Pd nanoparticles. The values in brackets represent the standard deviation of all measured Pd phases

Calcination of Pd glucose catalyst/°C	Lattice parameter a0^a	Approx. C at%
a Values in brackets represent the standard deviation of all observable Pd reflections.
As-synthesised	0.3923(4)	4.7
400	0.3994(3)	15.1
400 in H2	0.3892(3)	0.3
500	0.3887(2)	0
600	0.3887(0)	0
700	0.3892(1)	0.3
900	0.3889(1)	0
1000	0.3894(2)	0.6

---

Fig. 5 shows the TPR profile of the catalyst calcined at 400 °C in H₂ and N₂. It was evident that the sample pretreated in N₂ inhibits the formation of subsurface hydride. This is in contrast with the sample pretreated in H₂, where a strong negative peak characteristic of the decomposition of Pd–H is observed. It is well documented that the presence of subsurface carbon blocks the formation of β-hydride species by competing for the same octahedral interstices.^14,19 This provides clear evidence that carbon is clearly situated in the octahedral sites of the Pd.

Fig. 5 TPRs highlighting the decomposition of the β-Pd-H phase on the catalyst with: (a) interstitial carbon selectively removed, and (b) the blockage of β-hydride by the catalyst with the interstitial carbon.

TEM was also used to probe the morphology and size of the glucose carbonised Pd catalyst (Fig. 6). The Pd is clearly seen inside the amorphous carbon derived from carbonised glucose. Furthermore, the metal appears to be clustered together rather than being spherical. The mean particle size was calculated to be 7.0 nm with the mode at 6 nm, after calcination at 400 °C, which complements the calculation using the Scherrer equation by PXRD. HAADF and EDX confirmed the presence of Pd in the specimen (Fig. 6). The d spacing of the lattice fringes along the 〈111〉 direction was calculated to be 2.35 Å, also in agreement with our data from PXRD and the literature value.

Fig. 6 TEM images of the glucose carbonised Pd catalyst: (a) after 400 °C calcination, (b) after peroxide treatment, (c) close-up of (b) showing the Pd lattice fringe in the 〈111〉 direction, (d) HAADF image of the Pd catalyst, (e) EDX analysis of region 1 on image (d) showing the presence of Pd. (f) Relative particle size distribution after peroxide treatment.

It is evident that the Pd is covered by the carbonised glucose, therefore, it was hypothesised that the carbon will need to be partially removed in order to expose more metal to increase the catalytic activity. This was later confirmed by catalytic testing. Thus, the catalyst was treated in a mixture of hydrogen peroxide and ammonium hydroxide. This mixture is known to oxidatively remove some amorphous carbon.^20,21 Moreover, it was observed that there were no such leaching effects under the mild oxidising conditions. This contrasts with a harsher treatment in nitric acid or piranha solution, which is susceptible to Pd leaching. Another advantage is that the mixture is known to reduce surface functionalisation of carboxylic groups created under typical oxidative conditions. TEM analysis of the partially oxidised sample (Fig. 6) demonstrated that there was no increase in the Pd particle size as a result of the oxidative treatment.

XPS was utilised to study the electronic structure of the calcined Pd nanocatalyst and a sample pre-oxidised with peroxide solution (Table 5). The carbon 1s signals are typical of elemental carbon with some carbon–oxygen functions present. More carbon–oxygen functions are present on the sample surfaces, which include alcohols/ethers at ∼286.5 eV and carboxyl groups at ∼288.5 eV (FTIR for samples after oxidation displayed an increased C=O stretch intensity). Oxygen 1s signals were present mainly from carbon–oxygen functions such as alcohol (∼533.5 eV) and carboxyl groups (∼532 eV). An increase in surface oxygen concentration confirmed the oxidative treatment. The Pd 3d signals indicated the presence of the metallic Pd phase for both samples, illustrating that the relatively mild oxidative nature of the said peroxide treatment does not detrimentally affect the palladium. The information here, however, is limited on determining any peak shift in the Pd 3d binding energies that arise from the carbon in the interstitials, due to the weakness of the Pd signals. XPS is a surface-sensitive technique, probing the outermost atomic layers of the material. This is in contrast to the metal loading at 9% determined from ICP. The comparatively low Pd at% observed here suggests that the Pd was saturated with the external modifications, agreeing with TEM analysis.

Table 5 Results from the XPS analysis

Catalyst	Peak BE			At%
	C 1s	O 1s	Pd 3d	C	O	Pd	Others
PdGluC400	284.61	533.24	335.58	84.2	11.7	0.05	4.05
PdGluC400[O]H2O2/NH4OH	284.63	532.42	335.61	76.8	19.4	0.03	3.8

---

The ability of the carbonised Pd nanocatalyst after oxidative treatment was probed in the stereoselective hydrogenation of 3-hexyn-1-ol to corresponding alkenol (Scheme 1). This is a highly valuable reaction widely used in the fragrance industry for the production of leaf fragrance alcohol.²² The new catalyst also showed high stability with no apparent deactivation or loss in selectivity upon repeated testing. Results were compared to the classical Lindlar catalyst and a commercial Pd/C in order to compare their activity and selectivity in low temperature, liquid phase conditions. Fig. 7 displays typical H₂ uptakes of the catalysts for hydrogenation. Strikingly, the glucose derived catalyst was significantly different from Pd/C, and maintained an impressive selectivity to alkene (Fig. 8). Only one equivalent of H₂ was taken up even after prolonged exposure to excess H₂ after complete alkyne conversion, while two equivalents of H₂ were consumed over Pd/C, which led to total alkane formation. While the more active classical Lindlar catalyst only results in the uptake of one mole of hydrogen, this suffers rapid isomerisation processes leading to reduced alkene yields compared to the glucose derived catalyst. This therefore demonstrates the superior performance of this new class of Pd–C catalysts in minimising the undesirable alkene isomerisations and over-hydrogenations, despite prolonged exposure to a H₂ atmosphere.

Scheme 1 Stereoselective hydrogenation of 3-hexyn-1-ol.

Fig. 7 H₂ uptake curves in the stereoselective hydrogenation of 3-hexyn-1-ol using: (a) 5% Pd/C, (b) 5% PdPb/CaCO₃, (c) PdGluC400[O]H₂O₂/NH₄OH. Conditions: Catalyst to substrate ratio: 1 : 1000, 3 bar H₂, 30 °C, 400 rpm.

Fig. 8 GC analysis of 3-hexyn-1-ol hydrogenation at the end of the H₂ uptake (above). Conversion based on alkyne consumed and selectivity sum of cis and trans-hexan-1-ol. The numbers on the graph represent the cis/trans-alkene ratio.

Stereoselective hydrogenation of the non-polar substrate 4-octyne was also studied in order to investigate the scope and the general utility of this new class of subsurface modified Pd nanocatalysts and to observe any underlying effects due to subsurface C in Pd (Scheme 2). Thus, catalysts containing surface modification, subsurface interstitial modification or both were investigated and any attenuation on alkene adsorption was monitored. The hydrogenation of 4-octyne was examined using the glucose carbonised catalysts following a H₂ pretreatment, in order to selectively remove C_subsurface and compared to that without a H₂ pretreatment, therefore retaining the C_subsurface. These two catalysts were also compared to the commercial 5% Pd/C (provided by Johnson Matthey). The hydrogenation kinetic reaction profiles were fitted based on the Langmuir–Hinshelwood (L–H) kinetic model. This model assumes equilibrium establishment of hydrocarbons observed in the reaction and a weak H₂ adsorption, based on previous experiments.²³ The rate equations were written according to Scheme 2 with the assignment of an absorption equilibrium constant for each observed product:

where the species are defined as follows: A = 4-octyne; B = cis-4-octene; C = trans-4-octene; D = 3-octene; E = n-octane, respectively. K_i is defined as the adsorption constant of species i, P_(i) is the equilibrium concentration of species i and θ_i is probably a surface site occupied by species i. The defined equations used for the model fittings to the experimental data were excellent with good reproducibility, reinforcing the validity of our model (Fig. 9). Table 6 shows the result of the optimised kinetic model fittings. The initial unit from the optimised calculations was defined as per minute and this was then normalised against the initial amount of the substrate used per mole of the catalyst per hour.

Scheme 2 Proposed reaction pathways of the stereoselective hydrogenation of 4-octyne.

Fig. 9 Kinetic model fittings based on the L–H model of: (a) PdGlu400[O]H₂O₂/NH₄OH with contributions from C_surface and C_subsurface, (b) PdGlu400H₂[O]H₂O₂/NH₄OH with C_subsurface removed, (c) commercial 5% Pd/C. Points and lines are experimental and fitted data, respectively.

Table 6 Rates fitted using the L–H model of each hypothetical step of the catalysts tested in the stereoselective hydrogenation of 4-octyne

Dataset (molsub mol^−1cat h^−1)	PdGlu400[O]H2O2/NH4OH	PdGlu400H2[O]H2O2/NH4OH	5% Pd/C
k 1	494	397	3005
k 2	0	0	0
k 3	111	248	9406
k 4	10	9	3228
k 5	0	0	0
k 6	67	106	7311
k 7	69	254	0
k 8	10	9	158

---

It is evident, from Fig. 9 and Table 7, that the reaction is characterised by the strong adsorption of 4-octyne in all cases, with the adsorption constant approximately two orders of magnitude higher than all alkene species, in agreement with previous calculations.²⁴ The consequence of this effect was almost exclusive production of cis-4-octene (although a small degree of direct semi-hydrogenation to trans-4-octene was also observed), until 4-octyne was nearly consumed in the cases when Pd contains C_surface or C_surface and C_subsurface. This is due to the lower adsorption energy of the weaker adsorbate, in this case alkene over alkyne. The consequence of this is that the catalyst surface will favour re-adsorption of 4-octene after near complete alkyne consumption, only then it will undergo isomerisation and over-hydrogenation reactions.

Table 7 Adsorption equilibrium constants derived from the L–H model

Dataset (atm^−1)	PdGlu400[O]H2O2/NH4OH	PdGlu400H2[O]H2O2/NH4OH	5% Pd/C
K 1	9.5676 × 10^1	3.2691 × 10^1	2.8423 × 10^1
K 2	6.9165 × 10^−1	4.0043 × 10^−1	2.8594 × 10^−1
K 3	6.9562 × 10^−1	4.0126 × 10^−1	3.0599 × 10^−1

---

The kinetic plots and adsorption constant calculations clearly demonstrate that alkyne adsorption is much stronger than other surface species and therefore effects between surface modification and subsurface contributions cannot be separated. Thus, an attempt to differentiate these two effects was made using the subsequent isomerisation and over-hydrogenation processes. These two pathways were believed to be sensitive to the adsorption strength on the catalyst surface and therefore may be able to differentiate the said contributions from C_subsurface to those on the surface.

Fig. 10 shows a plot of rate ratios normalised using the semi-hydrogenation pathway (Σk_semi or Σk_1,8), against two parameters: (1) the sum of isomerisations (Σk_iso or Σk_3,4,7) and (2) the sum of over-hydrogenations (Σk_full or Σk_2,5,6). The rate ratios may give an indication on how the catalyst minimises these undesirable processes such as isomerisations and over-hydrogenations. On comparing the catalyst with C_surface and C_subsurface against C_surface only, the influence of subsurface carbon atoms was clearly evident on reducing these processes, despite the fact that they have no contact with substrate molecules. This was characterised by the high semi-hydrogenation rates against the suppression of isomerisations and over-hydrogenations, leading to the high rate ratios. The commercial Pd/C, despite having high activity, suffers from excessive isomerisations and over-hydrogenations, leading to the lowest rate ratios. Overall, it was found to be about 17.5 and 10.6 times less in full hydrogenations and isomerisations for samples with C_surface and C_subsurfacecf. 5% Pd/C, respectively. The selective removal of subsurface carbon by H₂ with the retention of only C_surface gave only 8.9 and 3.2 times in reduction in full hydrogenations and isomerisations c.f. 5% Pd/C, respectively. This provided clear evidence that subsurface C does indeed affect catalytic events in the surface for the first time in liquid phase reactions.

Fig. 10 A plot of rate ratios comparing to the catalysts tested in the hydrogenation of 4-octyne.

Thus, apart from surface carbon effects, this study had shown that interstitial carbon atoms in the subsurface of Pd play a significant role in reducing the rates of alkene intermediates from over-hydrogenation and isomerisation reactions. This could be related to the inhibition of the β-Pd-H phase, which is known to alter the hydrogen atom equilibrium as shown here by TPR. In addition, the palladium could also be electronically modified. For example, using density functional theory (DFT) calculations, Yudanov et al. noted that subsurface C weakens the binding energy of CO¹⁸ and this was later reinforced by Lim et al. by DFT calculations.²⁵ The effect is believed to be due to an increase in the Fermi level of the Pd through the hybridisation of the Pd d-band with C s–p bands, thus favouring desorption and overall reduction of alkene adsorption energy.^26,27 Although electronic effects due to the influence of interstitial C on Pd were not detected by XPS in this study, however, Teschner et al. enforced our theory as the authors observed changes in electronic character of Pd caused by interstitial C probed using a high energy XPS technique.⁹ On the basis of these accounts, we believe an electronic effect must play a more significant role in the dramatic change of catalytic surface events observed in our study. This was achieved by lowering adsorption energy of alkenes, leading to increase in desorption rates and therefore reducing undesirable processes, such as isomerisation and over-hydrogenation. On the other hand, the combination of kinetically stable interstitial carbon atoms from our preparation and additional effects in the liquid phase such as the solvent cage effect, stronger molecular adsorption, and molecular re-adsorption could render the ‘carbon effects’ more significant than other gas phase reactions. Nevertheless, we demonstrate that trapped interstitial carbon atoms in the Pd subsurface, created by our green methodology using glucose, were indeed effective to catalyse the hydrogenation of alkynes to cis-alkenes in the liquid phase. This new class of catalysts could open up rational ways to tune substrate adsorption with other subsurface interstitials for ultraselective catalysis.

Conclusions

In conclusion, we report a green route using glucose, as a reducing and capping agent as well as the supporting matrix, in order to stabilise Pd at a molecular level. Upon heat treatment, the carbonised glucose encapsulated the Pd nanoparticles with carbon atoms also taking residence in the octahedral holes, blocking the formation of β-Pd-H. Stereoselective hydrogenations of 3-hexyn-1-ol and 4-octyne in the liquid phase were tested and effects produced by the surface and the subsurface were separated using a simple H₂ pretreatment to remove subsurface C. It was shown for the first time that the geometrically and electronically modified Pd with interstitial carbon atoms reduced the adsorption energy of alkenes. This substantially reduced undesirable processes such as isomerisations and over-hydrogenation of alkenes, leading to higher overall cis-alkene selectivity.

Acknowledgements

C.W.A.C gratefully acknowledges financial support from the EPSRC, Johnson Matthey for a CASE studentship. We also thank R.A.P. Smith for XPS, G. Goodlet for TEM and G. Nnorom-Junior for ICP-ES measurements at the Johnson Matthey Technology Centre.

References

1. P. Pollak, Fine Chemicals: The Industry and the Business, John Wiley & Sons Inc., 2007.
2. Fine Chemicals through Heterogeneous Catalysis, ed. R. A. Sheldon and B. H. Van and Wiley-VCH Verlag GmbH, 2001.
3. Á. Molnár, A. Sárkány and M. Varga, J. Mol. Catal. A: Chem., 2001, 173, 185–221.
4. S. Mori, T. Ohkubo, T. Ikawa, A. Kume, T. Maegawa, Y. Monguchi and H. Sajiki, J. Mol. Catal. A: Chem., 2009, 307, 77–87.
5. A. Drelinkiewicz, W. Stanuch, A. Knapik, A. Ghanem, R. Kosydar, A. Bukowska and W. Bukowski, J. Mol. Catal. A: Chem., 2009, 300, 8–18.
6. N. Semagina and L. Kiwi-Minsker, Catal. Lett., 2009, 127, 334–338.
7. P. Weerachawanasak, O. Mekasuwandumrong, M. Arai, S.-I. Fujita, P. Praserthdam and J. Panpranot, J. Catal., 2009, 262, 199–205.
8. J. A. Anderson, J. Mellor and R. P. K. Wells, J. Catal., 2009, 261, 208–216.
9. D. Teschner, J. Borsodi, A. Wootsch, Z. Revay, M. Havecker, A. Knop-Gericke, S. D. Jackson and R. Schlogl, Science, 2008, 320, 86–89.
10. X. Sun and Y. Li, Angew. Chem., Int. Ed., 2004, 43, 597–601.
11. V. K. L. Mer, Ind. Eng. Chem., 1952, 44, 1270–1277.
12. Y. Ikushima, J. Liu, G. Qin and P. Raveendran, Chem.–Eur. J., 2006, 12, 2131–2138.
13. J. Liu, P. Raveendran, G. Qin and Y. Ikushima, Chem. Commun., 2005, 2972–2974.
14. S. B. Zlemecki and G. A. Jones, J. Catal., 1985, 95, 621–622.
15. P. Canton, C. Meneghini, P. Riello, A. Balerna and A. Benedetti, J. Phys. Chem. B, 2001, 105, 8088–8091.
16. S. B. Ziemecki, G. A. Jones, D. G. Swartzfager, R. L. Harlow and J. Faber, J. Am. Chem. Soc., 1985, 107, 4547–4548.
17. S. B. Ziemecki, G. A. Jones and D. G. Swartzfager, J. Less Common Met., 1987, 131, 157–162.
18. I. V. Yudanov, K. M. Neyman and N. Rosch, Phys. Chem. Chem. Phys., 2004, 6, 116–123.
19. N. Krishnankutty and M. A. Vannice, J. Catal., 1995, 155, 312–326.
20. V. Datsyuk, M. Kalyva, K. Papagelis, J. Parthenios, D. Tasis, A. Siokou, I. Kallitsis and C. Galiotis, Carbon, 2008, 46, 833–840.
21. Y. J. Kim, T. S. Shin, H. D. Choi, J. H. Kwon, Y.-C. Chung and H. G. Yoon, Carbon, 2005, 43, 23–30.
22. J. C. A. A. Roelofs and P. H. Berben, Chem. Commun., 2004, 970–971.
23. N. Semagina, E. Joannet, S. Parra, E. Sulman, A. Renken and L. Kiwi-Minsker, Appl. Catal., A, 2005, 280, 141–147.
24. E. Joannet, C. Horny, L. Kiwi-Minsker and A. Renken, Chem. Eng. Sci., 2002, 57, 3453–3460.
25. K. H. Lim, K. M. Neyman and N. Rösch, Chem. Phys. Lett., 2006, 432, 184–189.
26. J. Haglund, A. F. Guillermet, G. Grimvall and M. Korling, Phys. Rev. B: Condens. Matter, 1993, 48, 11685.
27. C. D. Gelatt, A. R. Williams and V. L. Moruzzi, Phys. Rev. B, 1983, 27, 2005.

---