M.
Keppeler‡
a,
G.
Bräuning
a,
S. G.
Radhakrishnan
b,
X.
Liu§
a,
C.
Jensen¶
a and
E.
Roduner
*ab
aInstitut für Physikalische Chemie, Universität Stuttgart, Pfaffenwaldring 55, D-70569, Stuttgart, Germany. E-mail: e.roduner@ipc.uni-stuttgart.de
bDepartment of Chemistry, University of Pretoria, Pretoria 0002, Republic of South Africa
First published on 8th March 2016
Monodisperse Pt clusters of 13 ± 2 atoms, supported on the zeolites NaY and KL and saturated with chemisorbed hydrogen, are investigated as well-defined model catalysts for reactions of CO, NO, O2 and ethene. CO reacts within <10 min, leading to the formation of dinuclear Pt carbonyl molecular clusters. A similar behaviour of NO suggests an analogous reaction. In stark contrast, O2 reveals a very sluggish reaction on a timescale of days although the reaction with chemisorbed hydrogen to form H2O is thermodynamically still favoured. This is ascribed to the inability of O2 to adsorb atop of Pt when all neighbouring sites are blocked by chemisorbed hydrogen. The hydrogenation reaction of ethene yields ethane as the only product. The turnover frequency at room temperature is somewhat lower than the one reported for the same reaction on Pt(111) single crystal surfaces or on Pt nanoparticles, but its activation energy is double that typically found in other systems. This means that the reaction which has been known to be structure-insensitive becomes structure-sensitive for catalyst clusters as small as 13 atoms. This fact is ascribed to a significantly larger binding energy of H on Pt as a consequence of the small cluster size and the influence of the support.
The present work focuses on uniform clusters of 13 ± 2 Pt atoms supported on NaY and KL zeolite. The synthesis and characterization of these close to icosahedral or cuboctahedral clusters were reported in detail previously.2–7 A fraction of these clusters are EPR active and show a regular multiplet, split by Pt hyperfine interaction, that is indicative of a highly symmetric structure with 12 equivalent Pt atoms as they are found in icosahedral or cuboctahedral 13-atom clusters.5 Of particular interest in the present context is the amount of chemisorbed hydrogen, which reaches up to 3 H per Pt atom, while for larger Pt nanoparticles one assumes only 1 H per surface Pt atom.3 Furthermore, the D2 desorption energy is 131 kJ mol−1 (1.36 eV) in NaY and 203 kJ mol−1 (2.1 eV) in KL zeolite, significantly more than the 77 kJ mol−1 (0.8 eV) from single crystal Pt(111) surfaces.4,5,8 These values give evidence of a size as well as a support effect.
Molecular hydrogen adsorbs side-on and dissociates on the surface of platinum. Based on IR vibrational spectroscopy of Pt13 clusters it was concluded that H atoms are bridge-bound over-edge (η2) to two neighbouring Pt atoms, while on a Pt(111) surface H binds to the three-fold hollow site (η3). In this way, icosahedral Pt13 can bind 30 H atoms.7 Molecular oxygen can also dissociate on a Pt(111) surface, forming atomic oxygen which is preferentially η3 bound to the three-fold hollow site.9 Carbon monoxide is known to bind well to most transition metals, and IR vibrational spectroscopy is well established and shows that binding is head-on via the carbon atom, primarily atop Pt, but also bridge-bonded (η2), or even η3-bonded.10 Adsorption of CO on Pt13 clusters leads to disintegration to form Pt2(CO)m species with m ≈ 5.11 NO adsorbs on clean Pt(111) initially in two-fold bridge-bonded sites and at higher coverages atop Pt, but oxygen preadsorption suppresses the bridge-bonded site up to high NO coverages.12
Hydrogenation of alkenes on noble metal catalysts is of major industrial importance. Therefore, the hydrogenation of ethene, the simplest alkene, has been used as a prototype reaction in studies of the mechanism of hydrogenation reactions for several decades and for various modifications of the catalyst. Hydrogenation on Pt(111) single crystal surfaces in the absence of preadsorbed hydrogen is a stepwise process, starting below 52 K through physisorption via its π-orbital and proceeding to ethane via a chemisorbed ethyl intermediate, CH3CH2-Pt.13 As a competing reaction, the π-bond breaks at above 52 K along with the formation of two σ-bonds from the carbon atoms to the Pt substrate. At 240 K, the di-σ-bonded ethene converts to ethylidyne (PtC–CH3) by splitting an H atom off from one of the carbons and shifting the second one to the other carbon. Neither the π-bonded ethene nor the ethylidyne are involved as direct intermediates in ethene hydrogenation.13 This hydrogenation mechanism was confirmed via IR spectroscopy which revealed a direct correlation of the disappearance of the signal of CH3CH2-Pt with the appearance of C2H6.14 The reaction kinetics is of zero order with respect to ethylene partial pressure and of first order with respect to hydrogen partial pressure, as confirmed in a recent operando study.15
The rates of certain reactions such as alkane hydrogenolysis or carbon–nitrogen ring opening reactions were found to depend on the Pt nanocrystal size.16 Furthermore, the catalytic activity and selectivity were shown to be sensitive to the shape of catalyst nanocrystals. The hydrogenation of benzene, for example, led to both cyclohexane and cyclohexene on cuboctahedral nanocrystals, whereas only cyclohexane was formed on cubic nanocrystals.5 However, ethene hydrogenation rates were independent of both shape and size and comparable to those on Pt single crystal surfaces, which means that the reaction is not structure sensitive down to 1.7 nm size.17 It is therefore of fundamental interest to check whether this behaviour pertains to even smaller catalyst species. Interestingly, a recent study revealed that size-controlled 12-atom Pt clusters which are metastable and have irregular coordination exhibit double the activity compared with the more regular near-icosahedral Pt13 cluster in the oxygen reduction reaction.18 It is generally thought that the additional inherent fluxionality of the metastable amorphous clusters is essential for maximum catalytic activity.19
The first part of the present work will contrast the different reactivities of the diatomic molecules CO, NO and O2 with fully hydrogenated Pt13 by directly observing the evolution of the EPR spectrum of the potentially catalytic centre. The second part will study the influence of these nanosize effects on the hydrogenation of ethene by in situ FT-IR spectroscopic investigation and by GC-MS product analysis in a flow reactor.
For hydrogen isotope exchange, deuterium gas was filled into an evacuated EPR quartz tube of 4 mm outer diameter containing about 200 mg of H-reduced Pt-loaded zeolite. For NaY, the gas was kept for 30 min at a D2 partial pressure of 500 mbar at room temperature; for KL the complete exchange takes ca. 24 h.
Isotopically enriched Pt salt was obtained from the Oak Ridge National Laboratory (USA).
Calculations of IR intensities along with the geometry and energy optimization were performed using Gaussian 09 (ref. 20) with the density functional B3LYP/6-311++G(d,p) and the results were analysed using the GaussView 05 program.21
The experiment was repeated with a sample consisting of a 194Pt (I = 0) isotopically enriched sample (Fig. 1b). Clearly, the splitting pattern of the multiplet and also much of the structure of the new signal disappear, confirming that it is due to Pt.
The reaction was investigated in detail based mainly on extended X-ray absorption fine structure (EXAFS) and vibrational FT-IR spectroscopy as reported elsewhere.11 It was concluded that the small cluster breaks up under the influence of binding energy to CO under formation of a binuclear Pt2(CO)m molecular cluster with m ≈ 5. In contrast, Pt nanoparticles of ≈2 nm diameter do not disintegrate on CO adsorption since they are sufficiently stabilised due to the higher Pt coordination number.
In contrast to the experiments with CO, NO addition leads to an additional new signal at half the magnetic field near g = 4.0046. This is a single quantum transition with ΔmS = 2 that is forbidden in isotropic systems but allowed for electron triplet states in anisotropic environments. Since the hyperfine shift is proportional to mS this contribution cancels out in the transition between the levels with mS = ±1. These transitions are therefore generally much narrower than the single quantum transitions. In the absence of transitions with ΔmS > 2, observation of this line is an unequivocal indicator of the presence of a triplet state. More detailed interpretation would require additional experimental information as e.g. EXAFS and electron nuclear double resonance (ENDOR) spectra, which are unavailable at this point.
The important fact in the present context is the immediate reaction of CO and NO with the deuterium saturated cluster which reveals full access of the diatomic molecules on the fully deuterated cluster surface.
![]() | ||
Fig. 5 FT-IR difference spectra recorded with freshly reduced Pt13Hx clusters supported on KL zeolite following adsorption of ethene at 86 K, and in the presence of H2 gas at 153 K and 273 K. |
On heating beyond 273 K the bands of the surface-adsorbed species disappear or are masked by the prominent vibrational bands of the evolving ethane (Fig. 6). There is a superposition of C–H stretch vibrations between 2804 cm−1 and 3006 cm−1 which on expansion of the scale show P, Q, and R branches with partly resolved rotational fine structure, revealing that the observed species is desorbed gas phase ethane. The bending vibrations are observed below 1500 cm−1, superimposed on two unresolved bands near 1600 cm−1 and 1400 cm−1 due to Pt–H vibrations.7 For the deuterated Pt13Dx cluster these latter two bands are shifted to lower wavenumbers outside the transparent window of the zeolite. On the deuterated Pt cluster samples the C–D stretching vibrations of the formed C2H4D2 are seen as expected near 2200 cm−1. Neither ethene nor any product other than ethane is observed in these spectra, suggesting complete conversion with high selectivity.
![]() | ||
Fig. 6 FT-IR difference spectra recorded at room temperature for gaseous ethane formed upon adsorption of C2H4 on Pt13HX (red, offset by 0.1 absorption units) and Pt13DX (black) on KL zeolite. |
There is not sufficient resolution in the spectra to analyse for the different isotopologues and their conformers. We can nevertheless try to get further information from the band intensities. The ratio of the integrated intensities of the C–H and the C–D-stretching regime in the experiment carried out with deuterium gas (black curve in Fig. 6) could naively be expected to be 2:
1, instead it is found to be 3.84
:
1. Indeed, there is an intrinsic isotope effect due to the mass dependence of the root mean square displacement of the atoms which enters in the transition dipole moment as μ−1/2, where μ is the reduced mass of the oscillator. The intensity of a transition is proportional to the oscillator strength f, the square of the transition dipole moment. For a simple diatomic C–H(D) oscillator, fH/fD is thus expected to scale with μ(C–D)/μ(C–H) = 1.87. This brings the intrinsic intensity ratio of the expected C2H4D2 product from 2
:
1 to 3.71
:
1, which is close to the experimental value. However, the issue could be more complex, since any IR-forbidden transition in C2H6 becomes allowed when the molecular D3d symmetry of the staggered conformation is broken by deuteration. We have therefore calculated the theoretical band intensities for the C–H and C–D stretching modes of the expected isotopologues. Interestingly, it turns out that the intensity loss due to forbidden modes is compensated for by other modes so that the sum of intensities scales well with the number of C–H(D) oscillators (Table 1). The complete information is given in Fig. S1 and Table S1 of the ESI.† On the average, the predicted C–H/C–D band intensity ratio for the expected C2H4D2 stoichiometry is 4.0(1), close enough to the observed value of 3.84. The intensity ratio does not provide any information about the specific nature of the isotopologues.
Isotopologue | Conformation | IR intensity, Σ over stretching vibrations (per mode) | ||
---|---|---|---|---|
C–H | C–D | C–H/C–D ratio | ||
CH2D–CH3 | Staggered | 154.3 (30.9) | 15.0 (15.0) | 10.3 (2.06) |
CH2D–CH2D | D–D staggered | 123.6 (30.9) | 30.0 (15.0) | 4.1 (2.06) |
D–D anti | 122.9 (30.7) | 30.8 (15.4) | 4.0 (2.00) | |
CHD2-CH3 | Any | 123.8 (31.0) | 29.9 (15.0) | 4.1 (2.07) |
CH2D-CHD2 | H–D anti | 93.1 (31.0) | 44.8 (14.9) | 2.1 (2.08) |
H–D staggered | 92.4 (30.8) | 45.7 (15.2) | 2.0 (2.02) |
Mass spectra of the product of the reaction of C2H4 with D2 were consistent with a C2H4D2 stoichiometry. However, detailed analysis of the mass distribution revealed evidence of isotopic scrambling with about 50% C2H4D2, significant amounts of C2H5D and C2H3D3 and probably a small amount of C2H6.
In the presence of preadsorbed hydrogen on the cluster the ethane product is observable in the IR spectra a few minutes after ethene adsorption at room temperature. On simultaneous adsorption of ethene and hydrogen in the absence of preadsorbed H the time delay until ethane appears is slightly longer, and when ethene is adsorbed alone in the absence of preadsorbed H the time to appearance of ethane is about a factor of 6 longer than for adsorption on Pt13Hm. The latter fact may be taken as qualitative evidence of an induction period during which chemisorbed H on Pt is made available for subsequent hydrogenation of ethene. This interpretation is in accord with Zaera who reported that hydrogenation can be enhanced by hydrogen coadsorption and that the activation energy of hydrogenation amounts to only about 25 kJ mol−1 in this case, while in the absence of preadsorbed hydrogen it is 40–45 kJ mol−1, determined by cleaving a C–H bond in π-bonded ethene.24
The turnover frequency decreases linearly by ≈40% over 60 minutes (Fig. 7). Since there is no obvious coloration which could be taken as evidence for coking, the origin of the deactivation is unclear. Furthermore, the TOF is lower by a factor of 2.1 when deuterium instead of hydrogen is preadsorbed. This hydrogen isotope effect demonstrates that the reaction is not transport-limited by ethene diffusion in the zeolite pores. Further aspects of it will be discussed below.
With H2 in the flow the TOF was monitored as a function of temperature. The results are displayed in the form of an Arrhenius plot in Fig. 8. A reliable determination of the TOF values requires the experiments to be performed under conditions of low turnover. Because of the high temperature dependence of the rate, experiments were limited to the range of 278–298 K over which the TOF varied by a factor of 12. For the three temperatures below 290 K the ethene to ethane conversion was less than 4%, but for 298 K it reached 29.7%. This data point was nevertheless used in the analysis, which has the consequence that the derived Arrhenius activation energy (Ea = 90 ± 10 kJ mol−1) and the corresponding apparent pre-exponential factor (log(A/s−1) = 15.2 ± 1.5) should perhaps be regarded as lower limits. The result is nevertheless highly interesting since the activation energy is significantly higher than most of the literature values which are in the range 40–45 kJ mol−1 when ethene is adsorbed on hydrogen-free surfaces17 and as low as 25 kJ mol−1 in the presence of preadsorbed hydrogen.24 According to Zaera, the rate-determining step in the absence of preadsorbed hydrogen is the dissociation of an ethene C–H bond which then provides surface chemisorbed H atoms for hydrogenation.23,24
![]() | ||
Fig. 8 Arrhenius plot of turnover frequencies for ethene hydrogenation on Pt13Hx supported on KL zeolite. |
Zaera analysed the reaction enthalpies of the gaseous (g) and adsorbed (a) species involved in the ethene hydrogenation reaction on Pt(111) and presented his results in terms of Born–Haber cycles (Fig. 9).23 We have amended this diagram to include the case of Pt13Hx clusters as a support. The important difference is the much higher chemisorption enthalpy of H2 on the clusters in KL zeolite (−202 kJ mol−1 compared with −80 kJ mol−1 on Pt(111)). In other words, the Pt–H bond strength amounts to 256 kJ mol−1 on Pt(111)24 but to 323 kJ mol−1 on Pt13/KL. The difference was ascribed in part to a nanosize effect and in part to a support effect.5 The consequence for ethene hydrogenation on Pt with preadsorbed hydrogen is that the first step of the reaction, the abstraction of a chemisorbed H atom, H(a), by adsorbed ethene to form the adsorbed ethyl radical is endothermic by 25 kJ mol−1 on Pt(111) but by 85 kJ mol−1 on Pt13/KL (reaction framed in red in Fig. 9). This reaction enthalpy is also the lower limit of the activation energy. It coincides with Ea when the back-reaction is not activated, a condition which seems to be met for the present system. The transfer of the second hydrogen requires less energy and is even exothermic on Pt(111) because the ethene double bond needs to be broken only in the first step.
![]() | ||
Fig. 9 Born–Haber cycle for C2 hydrocarbon fragments adsorbed on Pt(111) (black numbers) and on Pt13/KL zeolite (red numbers). The transformation from physisorbed to chemisorbed hydrogen is exothermic by 80 kJ mol−1 in the case of H2 on Pt(100) but by 202 kJ mol−1 in the case of D2 on Pt13. In the presence of preadsorbed hydrogen the addition of the first chemisorbed hydrogen to ethene (red frame) determines the observed activation energy of the reaction. The numbers are reaction enthalpies in kJ mol−1 (figure amended from ref. 24, with direction of reaction arrows and sign of reaction enthalpies inverted). |
Since the activation of the back-reaction of the first hydrogen transfer (frame in Fig. 9) is small or negligible this step leads to isotopic scrambling when different isotopes are provided in ethene and on the surface. Scrambling was reported by Zaera,24 and it was also observed in the present experiments (see the context of Fig. 6).
In the present context we can also discuss the isotope effect on the TOF (Fig. 7). Under the assumption of isotope-independent preexponential factors the observed TOF ratio of 2.1 translates into a difference in activation energies of only 1.8 kJ mol−1. Based on the observed vibrational frequencies of Pt13Hx the difference in zero-point energies of chemisorbed H and D is estimated to be at least 6 kJ mol−1. However, one also has to consider the difference in zero-point energy in the reaction product, the isotopic ethyl radical. It is clear that this leads to a partial compensation, but because of the uncertainty of the exact structure of the transition state and the presence of isotopic scrambling a further interpretation of the effect appears too speculative.
The binding energy per Pt atom is 557 kJ mol−1 (5.77 eV) in the bulk with a coordination number of 12. The surface atoms of a Pt13 cluster have coordination number 6 in the icosahedron so that the binding energy reduces to 434 kJ mol−1.25 The cluster will break up on CO adsorption when the energy gained on adsorption exceeds the energy spent to break up the cluster. The (CO)(x−1)Pt+−CO (1 ≤ x ≤ 4) bond energy depends strongly on x and amounts to 212 ± 10 kJ mol−1 (x = 1), 193 ± 10 kJ mol−1 (x = 2), 98 ± 5 kJ mol−1 (x = 3), and 53 ± 5 kJ mol−1 (x = 4).26 The first CO molecule can adsorb atop of each surface atom of the intact cluster. Two CO per Pt atom can provide 405 kJ mol−1, which is slightly less than the required 434 kJ mol−1 for cluster break-up. However, the reported stoichiometry of the disintegrated carbonyl complex is Pt2(CO)5,11 giving a stabilisation energy per Pt of 454 kJ mol−1 and thus exceeding the threshold value to break up the cluster, which is in excellent agreement with experimental data and supports the proposed stoichiometry of the carbonyl complex.
The binding energy of NO to Pt was predicted to be slightly more than that of CO. This supports the suggestion based on the EPR spectra (Fig. 2 and 3) that the cluster may break up into similar fragments as with CO, but confirmation by EXAFS would be highly desirable.
Therefore, the following question arises: Why do CO and NO react immediately with the hydrogen-saturated Pt cluster, while O2 nearly does not? We are not aware of any high level calculations of oxygen adsorption on hydrogen-saturated Pt clusters, but on bare surfaces O2 and also HO2 are predicted to adsorb preferentially by bridging two Pt atoms parallel to the edge with an adsorption energy on the order of 100 kJ mol−1.9 End-on adsorption is not an energetic minimum. Atomic oxygen is much more strongly bound (by 360–390 kJ mol−1) with a clear preference to occupy the 3-fold hollow sites. It is this high binding energy which drives dissociation of O2 that has a bond dissociation energy of 498 kJ mol−1. On the hydrogen-covered clusters dissociation cannot occur because the required precursor state, the adsorbed molecular O2, cannot form due to blocking of the adsorption site by hydrogen. We therefore suggest that the correct explanation for the near absence of a reaction of O2 is the blockage of sufficiently strong adsorption sites. It is generally assumed that O2 needs two neighbouring adsorption sites for dissociative chemisorption. On a hydrogen-saturated cluster such a situation is difficult to find. This is consistent with Pt(111) for which it was also reported that O2 chemisorbs only on bare spots.27,28
Ethene is analogous to O2 in the sense that adsorption atop Pt leads to a single bond to Pt and a reduction of the CC double bond strength to a single bond. The C
C double bond in ethene is given as 728 kJ mol−1 and that of the single bond in ethane is 377 kJ mol−1,29 so that the Pt–C chemical bond would have to compete with the difference, 351 kJ mol−1. However, the bond of ethylidyne to Pt(111) is rated only at ca. 245 kJ mol−1,30 which is not sufficient to break the double bond. It may be inferred that the Pt–C bond may be stronger due to the size effect of Pt13 and the support effect of the KL zeolite, in the same way as for Pt–H where there is as much as 30% enhancement. Secondly, since this is still insufficient, we note that we have used the bond strength of ethylidyne instead of the ethyl radical for which we do not have a bond strength. Furthermore, abstraction of H from the Pt surface under formation of the adsorbed ethyl radical could drive the reaction. Indeed, the ethane C–H bond energy is 423 kJ mol−1, 100 kJ mol−1 more than for the Pt–D bond, so that this does provide a considerable driving force. We also remind that the experiment revealed an activation energy of 90 ± 10 kJ mol−1, which is sufficient to explain the mismatch of the above numbers for the initial chemisorption of ethene so that the overall picture is consistent despite the fact that some of the numbers used in the discussion may not be completely appropriate.
The thermodynamic argument of bond order reduction to explain reactivity is delicate since it holds only when the heat of reaction equals the activation energy and there is no further activation barrier. In the case of ethene, reducing the CC bond order cannot alone be the full explanation since Fig. 7 shows different TOFs for Pt–H and Pt–D for otherwise identical systems, proving that the Pt–hydrogen bond also plays a role. We therefore give preference to the explanation derived in Fig. 9. It is expected that high-level quantum chemical calculations will support and further clarify the details of the reactions studied here.
Owing to the significantly larger binding energy of H on the surface of the under-coordinated atoms of the Pt13 clusters the hydrogenation reaction shows roughly double the activation energy compared with the reaction on larger nanoparticles or single crystal surfaces. Thus, on clusters as small as Pt13 ethene hydrogenation finally becomes structure-sensitive. Nevertheless, as found previously, there is significant isotopic scrambling when deuterium is preadsorbed on the catalyst in place of H2. Together with IR spectroscopic evidence at low temperature this suggests that the mechanism of the reaction is the same as reported for the single crystal surface.
Note added in proof: Shortly after submission of this work measurements of the ethene hydrogenation reaction on soft-landed Pt clusters on magnesia as a function of Pt cluster size were reported by Crampton et al.31 A clear maximum TOF for ethane production of 0.0026 s−1 per Pt atom was found for a cluster size of 13 atoms, giving evidence of the structure sensitivity of the reaction at this small cluster size, in good agreement with the conclusion from the present work. The absolute TOF value at 300 K is lower than the present value of 1.0 s−1 by a factor of 385 and is amongst the lowest values reported from various experiments (0.0029–53.4 s−1 per Pt atom).17 We want to emphasise that it is not straightforward to compare the various sets of data despite the fact that we are reporting identical cluster size since the rates will depend significantly on support effects as demonstrated by the different Pt–H bond energies for different zeolite supports.5 Furthermore, the work by Campton was carried out with an equimolar dosage of ethene and hydrogen, whereas the present work relates to a saturation coverage of up to 3 H per Pt atom or 36 H per Pt13 cluster.3
Footnotes |
† Electronic supplementary information (ESI) available: Calculated IR vibrational spectra of deuterated ethane isotopologues. See DOI: 10.1039/c6cy00182c |
‡ Present address: Eichenweg 5c, D-89290 Buch, Germany |
§ Present address: Faculty of Materials Science and Engineering, Hubei University, No. 368 Youyi Avenue, Wuchang, Wuhan 430062, China |
¶ Present address: ThyssenKrupp Industrial Solutions AG, Business Unit Process Technologies, Neubeckumer Straße 127, D-59320 Ennigerloh, Germany |
This journal is © The Royal Society of Chemistry 2016 |