ZrO2 catalysts applied in the hydrogenation of ethyl acetate

Journal of Catalysis 352 (2017) 120–129

Contents lists available at ScienceDirect

Journal of Catalysis journal homepage: www.elsevier.com/locate/jcat

On the bifunctional nature of Cu/ZrO2 catalysts applied in the hydrogenation of ethyl acetate J. Schittkowski, K. Tölle, S. Anke, S. Stürmer, M. Muhler ⇑ Laboratory of Industrial Chemistry, Ruhr-University Bochum, D-44780 Bochum, Germany

a r t i c l e

i n f o

Article history: Received 22 March 2017 Revised 10 May 2017 Accepted 12 May 2017

Keywords: Ester hydrogenation Copper nanoparticles Zirconia matrix Kinetics TPD Spillover

a b s t r a c t The catalytic hydrogenation of ethyl acetate to ethanol was studied at ambient pressure in the temperature range from 463 K to 513 K using Cu/ZrO2 catalysts obtained by co-precipitation as a function of the Cu loading. The hydrogenation was established as a reproducible probe reaction by determining optimal reaction parameters without deactivation or thermodynamic limitations. Power-law kinetics were determined yielding an apparent activation energy of 74 kJ mol
1 and reaction orders of 0.1–0.3 for H2 and
0.4 to 0.1 for ethyl acetate in the temperature range from 473 K to 503 K. Metallic Cu was found to be essential for the hydrogenation, but the catalytic activity was not proportional to the Cu surface area derived from N2O decomposition and temperature-programmed H2 desorption experiments identifying Cu/ZrO2 as bifunctional catalyst. The acidic sites of the ZrO2 matrix were probed by temperatureprogrammed experiments with ethyl acetate and NH3. Cu0 is assumed to provide atomic hydrogen by dissociative adsorption and spillover, but the reaction rate is more affected by the tight contact between the embedded Cu nanoparticles and the X-ray amorphous ZrO2 matrix. Ó 2017 Elsevier Inc. All rights reserved.

1. Introduction Zirconium dioxide has numerous applications in heterogeneous catalysis due to its high mechanical and thermal stability [1], associated with suitable acid and base properties and high specific surface area [2,3]. Correspondingly, zirconia is frequently applied in acid-base catalysis or as catalyst support. Zirconia is able to prevent the sintering of supported metal nanoparticles under reaction conditions especially in case of strong interactions with an amorphous zirconia matrix [3–6]. Zirconia was investigated for redox reactions doped with other metal oxides, which changed its phase composition [3], and its utilization as active material was investigated in detail [7,8]. In particular, zirconia was used in hydrogenation reactions [9,10], hydrogen generation by direct oxidation of hydrocarbons [11], and CO oxidation [12]. Although there were many different doping materials tested, the focus was often on catalysts containing Cu as active component. Numerous investigations of Cu/ZrO2 catalysts have dealt with the influence of the zirconia structure on the catalytic activity [3,9,10,13–16]. Cu supported on monoclinic zirconia was found to be active in methanol synthesis due to the presence of oxygen vacancies. It was reported that for materials with an equal

⇑ Corresponding author. E-mail address: [email protected] (M. Muhler). http://dx.doi.org/10.1016/j.jcat.2017.05.009 0021-9517/Ó 2017 Elsevier Inc. All rights reserved.

Cu surface area the catalyst with the monoclinic zirconia phase has a higher adsorption capacity for CO2 and CO [15]. Furthermore, Bell and co-workers [14,15,17,18] demonstrated the bifunctional nature of this catalyst system in a detailed analysis of the synthesis of methanol from CO or CO2 as well as the strong influence of the zirconia phase on the catalytic activity. According to the proposed reaction mechanism for methanol synthesis, CO or CO2 was adsorbed on the zirconia surface, and metallic Cu was needed to adsorb H2 dissociatively. The hydrogenation of the formate species occurred on the zirconia surface via hydrogen spillover. Similar observations were also reported by Otroshchenko et al. [19], who investigated propane dehydrogenation over Ru/ZrO2 catalysts. The reaction was assumed to take place at zirconium cations (Zrcus) located near anion vacancies. This hypothesis was supported by EPR and temporal analysis of products (TAP) reactor measurements. Metallic Ru is claimed to provide atomic hydrogen, which is comparable to the results of Bell and co-workers obtained with the Cu/ZrO2 system [19]. Additionally, Wu et al. [10] reported that the catalytic activity of Cu/ZrO2 cannot simply be attributed to the Cu surface area or the particle size of the metal, because it depends on the interaction between Cu and ZrO2 and the interface between the two components. This conjecture was also supported by Copéret and co-workers [13], who combined kinetic investigations of CO2 hydrogenation with IR and NMR spectroscopic and isotopic labeling studies. They postulated that the zirconia/copper interface

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

is crucial for the conversion of the formate species as reaction intermediate in methanol synthesis. In addition, Sato et al. [4,20] related the high activity in the dehydrogenation of ethanol to ethyl acetate over Cu/ZrO2 catalysts to their electronic properties. The zirconia phase was found to influence the ratio of Cu0/Cu+ due to the presence of oxygen vacancies in the monoclinic phase. The adsorption of ethanol took place either as ethoxy species on Cu+ and Zrd+ sites and as acetyl species on metallic Cu. Accordingly, the formation of ethyl acetate was only possible by a combination of both adsorption sites [4,20]. The high activity in ethanol conversion as well as in methanol synthesis from CO2 was also reported by Dumesic and co-workers [21], who observed an increase in turnover frequency by an order of magnitude due to the formation of Cu-ZrO2 interfacial sites. In recent years Cu has been frequently investigated in the hydrogenation of organic compounds. Cu was applied in several studies [22–25] because of its ability to hydrogenate C@O bonds selectively while leaving C@C bonds intact. First investigations of ester hydrogenation over Cu-containing catalysts were performed by Adkins et al. [22] at high pressure in the liquid phase with various esters. With Cu chromite as catalyst the reaction was carried out with an alcohol yield in the range from 80% to 98% requiring high pressures and temperatures [22]. In further studies, it was shown that several Cu-based catalysts have a high selectivity in the gas-phase conversion of simple esters to their corresponding alcohols such as ethyl acetate (EtOAc) to ethanol (EtOH) according to Eq. (1) [23–26]. The formation of a broader product distribution due to transesterification was reported [23,27,28].

ð1Þ Structure-activity correlations as well as kinetic parameters have been derived for Raney Cu [23] or silica-supported Cu [24] at temperatures from 483 K to 553 K and different partial pressures. The hydrogenation mechanism for acetates was suggested by Yan et al. [29] and Evans et al. [23] to proceed via dissociatively adsorbed acetate yielding acyl and alkoxy species:

RCOOR0 þ 2 ! RCO  þR0 O

ð2Þ

where  represents an active surface site. In further isotopic labeling studies, it was suggested that the alkoxy fragment is rapidly hydrogenated to its alcohol. In contrast, the acyl fragment is less reactive and can be partially hydrogenated to the aldehyde or fully to the alcohol. However, the hydrogenation of acetaldehyde was determined to be more than three orders of magnitude faster than that of ethyl acetate [26,30]. Thus, the amount of acetaldehyde being present is determined by the thermodynamic equilibrium constant of reaction (3):

RCH2 OH RCHO þ H2

ð3Þ

Here, the hydrogenation of ethyl acetate over Cu/ZrO2 catalysts was performed to establish a fast and reproducible probe reaction. Differently prepared Cu/ZrO2 catalysts can be compared based on the conversion of ethyl acetate and the yields of ethanol.

ð4Þ This probe reaction is high suitable for Cu/ZrO2 catalysts to investigate their bifunctional properties. The reaction kinetics of the hydrogenation was studied to evaluate the activity and stability in the used temperature range as well as to compare experimental data with literature data. Furthermore, this probe reaction was conducted to obtain structure-activity correlations for Cu/ZrO2 catalysts with varied Cu content. The catalysts were

121

characterized by temperature-programmed reduction (TPR), Xray diffraction (XRD), elemental analysis (ICP-OES), NH3 temperature-programmed desorption (TPD) and by the determination of the Cu surface area using H2 TPD in addition to N2O frontal chromatography. 2. Experimental 2.1. Sample preparation The Cu/ZrO2 catalyst precursor was synthesized by coprecipitation at pH 10.5 using copper nitrate (Cu(NO3)23H2O, Sigma Aldrich 99–104%), zirconium oxynitrate (ZrO(NO3)23H2O, Sigma Aldrich 99%) and 25% NaOH as precipitating agent. The precursor was aged for 15 min, filtered and washed with HPLC water until the filtrate was free of nitrates. Subsequently, the precursor was dried over night and calcined at 763 K for 3 h in synthetic air. Samples with 0 wt%, 1 wt%, 5 wt%, 10 wt%, 18 wt%, 22 wt%, 31 wt%, CuO were synthesized. 2.2. Characterization The metal content was determined by optical emission spectroscopy combined with inductively coupled plasma (ICP-OES) using an UNICAM PU 7000. XRD patterns were recorded in reflection geometry with an Empyrean Theta-Theta diffractometer (Panalytical, Almelo) using Cu Ka (k ¼ 1:5406 Å) radiation. The data was recorded from 5° to 80° in 2h with 0.25° divergence slit, 0.5° anti-scatter slit, 0.04 rad incident, diffracted beam soller slits, and a position sensitive PIXcel-1d detector. TPR experiments were performed using 100 mg of the catalyst sample with a sieve fraction of 255 lm to 350 lm placed in a glass-lined stainless steel U-tube reactor under plug-flow conditions. A thermocouple was placed directly in the catalyst bed to monitor the temperature. The catalyst was heated in diluted H2 (2%, purity 99.9999%, 10 mL min
1) to a maximum temperature of 573 K with 1 K min
1 including an isothermal plateau at 398 K for 14 h. To achieve an adsorbate-free Cu surface after the reducing pretreatment, the catalyst was flushed with He at 573 K and cooled to room temperature in He. N2 physisorption measurements were performed with 200 mg calcined catalyst in a sieve fraction of 250–355 lm using a BELSORP- max (BEL Japan, Inc.). Prior to the measurement, the catalyst was heated to 473 K for 2 h under vacuum to remove adsorbed water. To determine the specific Cu surface area of the catalysts two methods were used. First, the catalyst was treated with 1% N2O at 296 K assuming an adsorption stoichiometry according to N2 O þ 2 Cus Cus AOACus þ N2 , that is, one chemisorbed oxygen atom requires two metallic Cu surface atoms. A detailed description can be found elsewhere [31]. Surface determination was accomplished based on the total amount of N2O consumed. Furthermore, the specific surface area was determined by H2 TPD experiments. Pure H2 was passed at room temperature through the catalyst bed with a 30 mL min
1 flow rate and also during cooling with liquid nitrogen to 77 K. After holding this temperature for additional 20 min, weakly bound H2 was flushed out in a He flow of 50 mL min
1. Using this flow rate the reactor was heated to 573 K with a heating rate of b = 6 K min
1. NH3 TPD curves were obtained after 2 h adsorption of 4000 ppm NH3 at 323 K using a flow rate of 100 mL min
1 and 100 mL min
1 of sample. After flushing in He for 2 h the sample was heated in 100 mL min
1 He at a rate of 5 K min
1 up to 763 K. Additionally, the Cu/ZrO2 catalyst with 18 wt% Cu was compared with pure ZrO2 concerning the desorption and hydrogenation of adsorbed

122

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

ethyl acetate (temperature-programmed surface reaction, TPSR). For the TPD experiment, a flow of 3.2% of ethyl acetate was passed through the reduced catalyst at room temperature for 20 min. After weakly bound species had been flushed out with a He flow of 20 ml min
1, the temperature was increased to 513 K with a heating rate of 3 K min
1. For the TPSR experiments, ethyl acetate was adsorbed on the catalyst surface in the same way, and the catalyst was heated to 513 K with a heating rate of 3 K min
1 using 68% H2 in He and a flow rate of 20 ml min
1.

rupole mass spectrometer (QMS, GAM 422 Balzers). Degrees of conversion were calculated according to Eq. (6):

2.3. Catalytic activity measurements

Accordingly, the yield of ethanol was derived from Eq. (8).

The kinetics of the hydrogenation of ethyl acetate was studied at ambient pressure in a temperature range from 293 K to 513 K using the same set-up applied for the TPR and TPD experiments. High purity H2 (99.9999%) and He (99.9999%) (Air Liquide) were used, and ethyl acetate was added by passing a He flow through a EtOAc-filled saturator cooled to 273 K. The total flow rates of these gases were adjusted by three mass flow controllers (Bronkhorst). The regeneration of the Cu/ZrO2 catalyst was studied using the following measurement sequence: After the hydrogenation reaction a H2 TPR experiment as well as a temperature-programmed oxidation (TPO) experiment was performed combined with catalytic tests in between to obtain the degree of conversion. The catalytic activity was tested by increasing the temperature linearly with 1 K min
1 to 573 K using a volumetric feed rate of 50 ml min
1 containing 1% ethyl acetate and 20% H2. After its completion the reactor was cooled to room temperature and flushed with 2% H2 at a flow rate of 10 ml min
1 up to a temperature of 573 K with a heating rate of 1 K min
1. Subsequent to holding the maximum temperature for 30 min the catalyst was cooled to room temperature in He. Then, an activity measurement was performed. After its completion the reactor was cooled to room temperature and the TPO experiment was performed using 10% O2, a flow rate of 10 ml min
1 up to T max = 573 K with 1 K min
1. Prior to the kinetic investigations the catalyst was aged for 72 h to exclude effects of deactivation during the measurements. Reaction conditions were set to 1% of ethyl acetate and 68% of H2 with a volume flow of 50 ml min
1 at 513 K. Afterwards, a power-law rate expression was determined. The contact time of ethyl acetate was varied by setting the volumetric feed rate of the reactant steam to values of 25 ml min
1, 50 ml min
1, 100 ml min
1 and 150 ml min
1. This volume flow was held for 30 min to reach steady-state conditions while the effluent mole fractions were analyzed. To gain insight in the influence of the reaction temperature five different temperatures were used. To determine the orders of reaction with respect to the different reactants, the molar ratios of H2 (20%; 40% and 68.8%) and ethyl acetate (0.5%, 1.0% and 2.0%) were varied balanced with He to obtain a total volumetric feed rate of 50 ml min
1. The specific reaction rates were derived from the consumed amount of ethyl acetate at low degrees of conversion correlated to the catalyst weight:

Y EtOH ¼

r¼

n_ in
n_ out mcat

ð5Þ

For the activity measurements of the catalysts with varied Cu loading 1% ethyl acetate and 68% H2 were fed to the reactor adjusting a total flow rate of 50 ml min
1 with He. The temperature was increased with 0.5 K min
1 to a maximum temperature of 513 K with several isothermal steps lasting 90 min to ensure steadystate conditions. During cooling to room temperature the same temperature profile was used. The effluent gases during all experiments were passed through heated glass-lined tubings and analyzed quantitatively in a quad-

X EtOAc ¼

n_ EtOAc
n_ EtOAc;0  100 n_ EtOAc;0

ð6Þ

The selectivity of the product ethanol was calculated according to Eq. (7):

n_ EtOH
n_ EtOH;0 n_ EtOAc;0
n_ EtOAc

SEtOH ¼

n_ EtOH
n_ EtOH;0 n_ EtOAc;0

v EtOAc  100 v EtOH v EtOAc  100 v EtOH

ð7Þ

ð8Þ

To exclude pore diffusion limitations, moderate temperatures as well a sufficient small sieve fraction were chosen. Diffusion limitations can be assessed based on the Weisz-Prater module (Eq. (9)). For UWP < 1 pore diffusion can be neglected.

UWP ¼

reff  R20 Deff  c0

ð9Þ

with r eff as effective rate and R0 as radius of the spherical catalyst particle. To calculate the effective diffusion coefficient Deff , the molecular diffusion coefficient and the ratio of porosity and labyrinth factor can be used (Eq. (10)). Normally, for porous catalysts values of  ¼ 0:7 and s ¼ 3:5 are used.

Deff ¼ D12

 s

ð10Þ

According to Eq. (11) the diffusion coefficient can be calculated using the radius of the pores r, the universal gas constant R, and the molar mass M assuming Knudsen diffusion due to small pore diameters.

D12 ¼

2 rP 3

rffiffiffiffiffiffiffiffiffi 8RT pM

ð11Þ

3. Results 3.1. Characterization The degrees of reduction were derived from the TPR experiments based on the amount of H2 consumed and compared to the ICP-OES results. Except for 1 wt% CuO loading degrees of reduction close to 100% were obtained (Table 1). In Fig. 1a the XRD patterns of all catalysts after calcination are shown. A low Cu loading results in the tetragonal zirconia phase with main characteristic reflections at 30.2°, 35.3°, 50.3° and 60.2° 2H, whereas increasing Cu amounts lead to a X-ray amorphous phase for all Cu-rich samples (>5 wt%). For the catalyst with 10 wt% CuO the broad reflections of the amorphous phase and the sharper reflections of the tetragonal phase are overlapping, which are no longer detected for higher Cu loadings. At the highest loading of 31 wt% CuO, reflections of CuO at 35.6° and 38.8° 2h are observed. In Fig. 1b a change in phase composition for samples with low Cu loading is found after the hydrogenation reaction. For the samples with 0 and 1 wt% CuO the tetragonal phase is partly converted to monoclinc ZrO2. For the X-ray amorphous samples minor changes were observed. At Cu loadings >10 wt% reflections of metallic Cu were monitored at 43.44° and 50.55°. The particle size of these metallic Cu nanoparticles was calculated using the Scherrer equation resulting in 5 nm for 18 wt% and 22 wt% and 14 nm for 31 wt%. The Cu surface area was determined with two different methods using N2O decomposition and H2 TPD experiments. In Fig. 2

123

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129 Table 1 CuO loadings, degrees of reduction, specific surface areas and specific Cu surface areas of the Cu/ZrO2 catalysts derived from N2O chemisorption and H2 TPD experiments. Nominal CuO loading wt%

CuO loading by ICP-OES wt%

Degree of reduction %

SBET m2 g
1 cat

SCu N2O m2 g
1 cat

SCu H2TPD m2 g
1 cat

0 1 5 10 18 22 31

0 1 5 10 17 21 26

0 60 95 84 98 97 98

126 134 132 156 155 155 137

0 0.4 1.8 1.6 2.9 2.5 3.8

0.1 0.3 – 1.6 2.8 2.6 3.6

Fig. 1. XRD diffractions patterns for CuO loadings between 0 wt% and 31 wt% of (a) CuO/ZrO2 after calcination and (b) Cu/ZrO2 catalysts after hydrogenation of ethyl acetate and reference patterns of CuO, Cu, t-ZrO2 and m-ZrO2.

Cu0. The broad peak at high temperatures is due to the dissociation of H2O on Cu0 resulting from the dehydroxlation of the support. For the catalysts with low Cu amounts no signal was detected. Based on the amount of desorbed H2 and an adsorption stoichiometry of 4 (Cu surface atoms to H2 molecules), the metallic Cu surface area can be calculated [33]. These values as well as the surface areas determined by the commonly applied decomposition of N2O at room temperature are summarized in Table 1. The results of both methods are in good agreement. Increasing Cu contents lead to slightly higher Cu surface areas, which are rather small amounting to 3:8 mg
1 cat for the catalyst with a Cu content of 31 wt%. Consequently, poor degrees of dispersion are obtained. The comparison of the surprisingly low specific Cu surface areas with the Cu particle sizes derived by applying the Scherrer equation indicates that the Cu nanoparticles are embedded in the Xray amorphous ZrO2 matrix.

Fig. 2. H2 TPD profile of Cu/ZrO2 catalysts with CuO loadings from 0 wt% to 31 wt%. V_ He = 50 ml min
1, b = 6 K min
1, mcat = 100 mg.

the H2 TPD profiles are shown. The H2 desorption peak at 300 K originates from the associative desorption of H2 from metallic Cu [32]. Accordingly, a symmetric peak with its maximum at 304 K ± 3 K is monitored for catalysts with sufficient amount of

3.2. Studies for parameter estimation The catalytic properties of Cu/ZrO2 were assessed by using the hydrogenation of ethyl acetate as probe reaction. The experimental parameters in literature [23,27] were in the temperature range from 493 to 573 K at ambient pressure using an excess of H2. In initial screening tests, first investigations were performed by varying the ethyl acetate to H2 ratio and monitoring the conversion of ethyl acetate with increasing temperature. It was found to be most suit-

124

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

able to limit the maximum temperature to 513 K. At higher temperatures a deficit in the carbon balance was detected pointing to coking. Correspondingly, CO and CO2 were detected during subsequent TPO experiments. Furthermore, it was found that a higher H2 partial pressure leads to a higher reactant conversion and also avoids coke formation. At 513 K acceptable conversion of ethyl acetate was observed without major deactivation processes. Ethanol was formed with high selectivity and acetaldehyde as major byproduct with low selectivity. Negligible amounts of CO and H2O were detected at maximum temperature. Prior to studying the reaction kinetics, the thermodynamic equilibrium conversion of ethyl acetate hydrogenation was investigated as a function of temperature using the software ChemCad. For these calculations the reactants ethyl acetate and H2 as well as the products ethanol and acetaldehyde were considered according to Eqs. (3) and (4). The degree of equilibrium conversion of ethyl acetate and the corresponding yields are shown in Fig. 3. The activity test using the chosen reaction parameters are far from equilibrium even at the maximum temperature of 573 K. The yield of ethanol is also clearly under kinetic control, but acetaldehyde reaches quickly thermodynamic equilibrium (Eq. (3)). Heating to the maximum temperature under the given concentrations of 1% ethyl acetate and 68% H2 were performed with different plateaus in between every 10 K holding for 90 min. Because of low conversion at low temperature the first plateau was set to 463 K. This stepwise increase/decrease of the temperature was repeated in a second cycle. In Fig. 4 the obtained concentration profiles of such a temperature profile are shown. It is clearly seen that every temperature increase resulted in a higher degree of ethyl acetate conversion as well as a higher yield of ethanol. All concentrations were essentially stable at every plateau, as confirmed in the second cycle. Additionally, the maximum temperature was reached without temperature plateaus by increasing the temperature with a slow heating rate of 0.5 K min
1. The monitored degrees of conversion showed minor differences to the stepwise increase, indicating that this transient mode of reactor operation is in quasi-steady state. 3.3. Reaction kinetics For varying the contact time of the reactant gas at different temperatures, a Cu/ZrO2 catalyst with 18 wt% CuO was used. In Fig. 5 the degree of conversion is shown as a function of residence time (weight to feed ratio, w/F). It is seen that conversion increased with higher residence times and higher temperatures. At short residence times conversion increased steeply, but then leveled off to higher degrees of conversion.

Rates of hydrogenation versus different H2-to-ester ratios are plotted in Fig. 6a and b. The H2 partial pressure has a relatively stronger influence on the conversion of ethyl acetate: the more H2 is provided in excess, the more ethyl acetate is consumed. A contrary effect is seen with higher concentrations of ethyl acetate, which lower the rate. The reaction orders were derived in the temperature range from 473 K to 503 K yielding 0.1–0.3 for H2 and
0.4 to 0.1 for ethyl acetate, which are rather low. The determined orders of reaction as well as the apparent activation energies Ea of this reaction are summarized in Table 2. The obtained values are in good agreement with literature data obtained with Cu/SiO2 applied in ethyl acetate hydrogenation [23,24,26]. The chosen temperature range has a great influence on the obtained data. Compared with Cu/SiO2 investigated at the same temperature, the reaction order with respect to H2 was lower and the reaction order of EtOAc was also estimated to roughly zero [26]. In case of Raney Cu a first-order dependence for H2 and
0.5 order with respect to ester were reported [23]. Additionally, an estimate of the overall activation energy was derived from an Arrhenius plot yielding 74 kJ mol
1. The overall activation energy estimated by Evans et al. [23] for Raney Cu and for Cu/SiO2 [24,30] were in the same order of magnitude. With activation energies in this region diffusion limitation can be neglected, which was confirmed by calculating the Weisz-Prater module (WP). Due to the low pore diameter of 3 nm, occurrence of Knudsen diffusion is assumed. Under these reaction parameters and given temperature UWP was calculated to have a maximum value of 0.32 with highest values of rreff , which also confirms the absence of diffusion limitations. 3.4. Activity test under steady-state conditions The activity tests were performed with isothermal plateaus stages to achieve steady state (Fig. 4). For all catalysts the overall trend was the same: ethanol was formed as the main product, whereas the formation of acetaldehyde was in equilibrium resulting in a low yield of maximum 6%. Thus, the decisive number to compare the different catalysts is the conversion as a function of temperature In Fig. 7 the degrees of conversion at maximum temperature as well as at the lowest temperature of 463 K are plotted versus the Cu content. Here, the key role of Cu in the catalytic hydrogenation is clearly identified. Catalysts with no or low Cu loading show poor activity, whereas a content of more than 5% CuO is sufficient to ensure adequate degrees of conversion. Additionally, conversion increases with increasing Cu loading with its maximum at 31 wt%. Deactivation during the isothermal stages was essentially negligible for all catalysts. When operating the catalyst with a CuO loading of 18 wt% for 4 days at 513 K, a relative deactivation of only 4% was observed. Considering the Cu surface area after reaction, the first trend is the same (Fig. 7): increasing Cu loadings yield higher Cu surface areas, again in a small range, but the Cu surface areas are only about half the Cu surface areas after reduction (Table 1). In contrast, the relative degree of conversion during the activity tests decreased just by maximum 1.4% for the Cu-rich samples. The results summarized in Table 3 clearly indicate that the catalytic activity of the Cu/ZrO2 catalysts depends to some extent on the exposed Cu surface area, but it is not all linearly correlated. 3.5. Temperature-programmed desorption and surface reaction of adsorbed ethyl acetate

Fig. 3. Calculated equilibrium concentrations as a function of temperature using ChemCad.

The temperature-programmed desorption of ethyl acetate adsorbed at room temperature was carried out to distinguish the different adsorption sites on Cu and ZrO2. In Fig. 8 the effluent mole

125

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

Fig. 4. Conversion of ethyl acetate and corresponding yields of ethanol and acetaldehyde as a function of time during a stepwise increase of the temperature. A 18 wt% Cu/ ZrO2 catalyst was used with a total flow of 50 ml min
1 and a reactant ratio of 1:68 (H2 in excess).

fractions of the desorbing species are shown as a function of temperature. Both for Cu/ZrO2 and ZrO2 ethanol starts to desorb immediately with a maximum desorption rate at approximately 360 K and 364 K, respectively. The amount of desorbing ethanol for pure ZrO2 is more than 4 times higher than for Cu/ZrO2 and does not decrease to zero at the maximum temperature. Also some traces of acetaldehyde are monitored. The major differences are seen when comparing the effluent mole fractions of H2 and H2O. For pure ZrO2 minor H2 concentrations were detected, whereas water increased to a constant level. In contrast, for Cu/ZrO2 a H2O desorption peak with a maximum at 405 K and a huge H2 peak starting at 373 K with its maximum at 447 K were observed. During the desorption experiment no ethyl acetate, CO2 and CO were detected. Additionally, the amount of ethyl acetate can roughly be estimated based on the adsorption process performed as frontal chromatography experiment. The amount of ethyl acetate adsorbed was determined on pure ZrO2 and Cu/ZrO2 to amount to 485 lmolgcat
1

Fig. 5. Conversion of ethyl acetate as a function of contact time and temperatures. V_ = 25 ml min
1, 50 ml min
1, 100 ml min
1 and 150 ml min
1 and reactant ratio of 1:68.

(a)

and 339 l respectively. The adsorption capacity of EtOAc is therefore higher for pure ZrO2 by a factor of 1.4.
1 molgcat ,

(b)

Fig. 6. Derivation of the reaction orders by varying the partial pressures of (a) H2 and (b) ethyl acetate.

126

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

Table 2 Comparison of determined orders of reaction and overall activation energies for the hydrogenation of ethyl acetate. Reaction orders

Cu/ZrO2 Cu/SiO2 [24] Cu/SiO2 [26] Raney Cu [23]

Ea

Temperature
1

H2

EtOAc

kJ mol

K

0.2 (493 K) 0.66 0.58 (493 K) 1

0 (493 K) 0.5 0.02 (493 K)
0.5

74.2 91 107.0 88.0

463–503 500–573 493–523 483–503

(a)

(b)

Fig. 7. (a) Degrees of conversion at 463 K and at 513 K initially and after 120 min at 513 K and (b) Cu surface areas of the catalysts with different Cu loadings before and after reaction. The fitted straight lines are intended to guide the eye.

Table 3 Conversion of EtOAc at 513 K, yields of ethanol and acetaldehyde (Ac), relative degree of deactivation after 120 min, Cu surface area (N2O) after reaction and relative decrease in Cu surface area for Cu/ZrO2 catalysts with different Cu content. CuO loading wt%

X max at 513 K %

Y EtOH at 513 K %

Y Ac at 513 K %

Relative deactivation %

SCu after reaction m2 g
1 cat

Decrease of SCu %

0 1 5 10 18 22 31

1.8 5.0 31.3 36.9 38.1 39.7 50.4

1.0 3.1 27.4 31.5 32.7 33.3 43.7

0.8 1.9 3.9 5.4 5.4 6.4 6.7

0.3 0.9 0.8 1.4 0.3 1.1 0.3

0.1 0.4 0.5 0.4 1.0 1.0 2.8

– 0 72 75 66 60 26

However, the two different samples showed similar behavior during the temperature-programmed surface reaction. During heating in H2 atmosphere just ethanol and no ethyl acetate were detected and also some traces of acetaldehyde. In both cases water is formed as the main product. Whereas pure ZrO2 shows one broad ethanol peak with a maximum at 410 K, Cu/ZrO2 has two maxima in ethanol formation at 373 K and 487 K. The amount of ethanol formed over pure ZrO2 is more than a factor of 2 higher. Due to the high H2 amount monitored during the desorption experiment, it is obvious that dehydrogenation and coking occur. The Cu/ZrO2 catalyst was investigated by a subsequent TPO experiment. Here, some rising amounts of CO, H2 as well as CO2 were detected up to the maximum temperature, which were rather low compared with the total adsorbed amount of ethyl acetate prior to the desorption experiment. Furthermore, the two samples were investigated by NH3 TPD experiments shown in Fig. 9. Both samples show a peak in the lower temperature region with a maximum at 405 K. The NH3 desorption peak of Cu/ZrO2 is higher, but desorption is completed roughly above 500 K, whereas desorption from pure ZrO2 can be

detected until 620 K. The desorbed amount for the pure support was 313 lmolgcat , for the Cu/ZrO2 catalyst 231 lmolgcat yielding a ratio of 1.4. The quantitative results of the TPD experiments are summarized in Table 4.
1


1

4. Discussion The steady-state experiments demonstrated the high activity of Cu/ZrO2 catalysts in the hydrogenation of ethyl acetate. This reaction was established successfully yielding a product distribution with just ethanol and acetaldehyde. It was possible to optimize the reaction parameters to obtain reproducible results with hardly any deactivation processes due to coke formation or sintering by limiting the maximum temperature to 513 K and using an excess of H2. It is assumed in literature [23,24,26] that the adsorption proceeds via dissociation of ethyl acetate to ethoxy and acetyl species. Next to the dissociative adsorption of H2, the proposed reaction mechanism by Dumesic and co-workers [24] consists of 4 steps.

127

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

Fig. 8. Effluent mole fractions during ethyl acetate desorption from (a) Cu/ZrO2 with a CuO loading of 18 wt% and from (b) pure ZrO2 as well as surface reaction of adsorbed ethyl acetate of (c) Cu/ZrO2 and (d) pure ZrO2.

Table 4 Adsorbed amounts of ethyl acetate and NH3 derived from the frontal chromatography adsorption and TPD experiments, respectively. The Cu/ZrO2 catalyst had a Cu loading of 18 wt%.

lmolg
1 cat

ZrO2

Cu/ZrO2

Ratio

Ethyl acetate NH3

485 313

339 231

1.4 1.4

ð12Þ ð13Þ ð14Þ Fig. 9. NH3 TPD curves obtained with Cu/ZrO2 (black) with a CuO loading of 18 wt% and ZrO2 (blue) after 2 h adsorption of NH3 at 323 K. 100 mg of the sample was heated in 100 mL min
1 He flow at a rate of 5 K min
1.

They suggested a kinetic model consisting of the dissociative adsorption of EtOAc to acetyl and ethoxy species (Eq. (12)), and the subsequent formation and desorption of acetaldehyde (Eq. (13)) and ethanol (Eq. (14)). Additionally, the readsorption of acetaldehyde as ethoxy species is taken into account (Eq. (15)).

ð15Þ The hydrogenation/dehydrogenation equilibrium between acetaldehyde and ethanol is assumed to be established. Further byproducts by consecutive reactions of the adsorbed species were not observed. The splitting of the CAO ester bond is one important step in this mechanism and also the first hydrogenation step yielding acetaldehyde. The further hydrogenation takes place quickly,

128

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

which is also seen in the desorption experiments, in which no ethyl acetate is monitored during the temperature ramp. After adsorption just acetyl and ethoxy species are present on the surface of ZrO2. When comparing pure ZrO2 with Cu/ZrO2 it was observed that the adsorption capacity of EtOAc is higher by a factor of 1.4 in the absence of Cu derived from the monitored adsorption experiment prior to the TPD experiment. Additionally, the formation of ethanol proceeds in the absence of H2 during the TPD experiments. For Cu/ZrO2, Cu catalyzed the dehydrogenation leading to some coking already at low temperatures generating a large amount of H2. Sato et al. [20] assumed that the adsorption sites responsible for the dehydrogenation of ethanol in the Cu/ZrO2 catalyst are Cu0 sites, which are required for the adsorption of the acetyl species, and Zrb+ sites, which adsorb the ethoxy species. The acetyl species was assumed to be on Cu0 due to the high barrier of CAH scission [20]. For the formation of ethyl acetate, these actives sites have to be adjacent, leading to the conclusion that the Cu-ZrO2 interface is crucial for catalytic activity [4]. This has also been recently suggested by Dumesic and co-workers [21] observing a higher turnover frequency due to the formation of Cu-ZrO2 sites. However, the quantitative TPD experiments with ethyl acetate and NH3 suggest that the dissociative adsorption of ethyl acetate occurs predominantly on the acidic ZrO2 sites, as the amount of adsorbed ethyl acetate was found to correlate with the number of surface acidic sites determined by NH3 desorption (Table 4). For both probe molecules pure ZrO2 showed a higher amount by a factor of 1.4 compared with Cu/ZrO2 with a CuO loading of 18 wt%. Obviously, both the dissociative adsorption of ethyl acetate and the adsorption of ammonia occur on acidic zirconia sites. Furthermore, ZrO2 is kinetically identified as suitable support because of the lower overall activation energy in the hydrogenation of 74 kJ mol
1 over Cu/ZrO2 compared with other Cu-containing catalysts (Table 2), which can be attributed to the beneficial bifunctional properties. Nevertheless, when testing the Cu/ZrO2 catalysts with different Cu loading in the hydrogenation of ethyl acetate, a large influence of Cu on the catalytic activity is found. Whereas pure ZrO2 showed minor activity, just small amounts of Cu increased the conversion of ethyl acetate dramatically, but a further increase of the Cu loading resulted in a slightly higher increase in conversion only. In methanol synthesis, a linear relationship between the catalytic activity and the Cu surface area is generally accepted [34]. Moreover, Chorkendorff and co-workers [35] showed on a Cu(1 0 0) single crystal surface that metallic Cu0 is sufficient for catalytic activity, but this hypothesis cannot be easily applied to Cu/ZrO2. The increasing Cu loading led to a higher Cu surface area, but the correlation to conversion is weak. Even with 31 wt% of CuO loading, a small Cu surface area of about 4 m2 g
1 cat was obtained, and, consequently, a poor degree of dispersion. Presumably most of the Cu nanoparticles are located within the ZrO2 matrix or are encapsulated [10] due to the preparation by co-precipitation. This conjecture is supported by the Cu particle sizes in the range from 5 nm to 14 nm as derived from the Scherrer equation, which should lead to substantially larger specific Cu surface areas. Obviously, just the presence of a small amount of Cu (>5 wt%) is needed to reach a higher degree of conversion underlining the key

role of Cu in the hydrogenation reaction. Correlating the Cu surface area after reaction with conversion leads to the same conclusion. Again, the surface area increases with higher Cu loading, but the loss in Cu surface area during reaction is huge. After reaction the Cu surface areas are mostly lowered by a factor of 2 for all catalysts, while essentially no deactivation during reaction was recorded. This observation implies encapsulation of the Cu nanoparticles and confirms that the Cu surface area is an important, but not sufficient criterion. The catalytic activity seems to be more related to the tight contact between the metallic Cu nanoparticles and the X-ray amorphous ZrO2 matrix than to the specific Cu surface area. Further microstructural studies are in progress aiming at a profound characterization of the embedded state of the Cu nanoparticles. It is also demonstrated that different Cu loadings influence the structure of the catalyst. With low amounts of Cu the formation of the tetragonal phase is found, while increasing Cu contents lead to the formation of an essentially X-ray amorphous ZrO2 phase. It was shown by Köppel and Baiker [36] that the amorphous phase is stabilized by the presence of Cu2+ ions in the matrix, which is achieved by a higher loading than 5 wt% (Fig. 1). For loadings >5 wt%, the long-range order of ZrO2 is lost due to the presence of Cu2+ ions. In addition to the decreasing Cu surface area, some structural changes during hydrogenation were detected. The formation of the monoclinic phase was found to occur for catalysts with lower CuO content than 5 wt%, whereas no phase change can be seen for higher loadings. Due to the low conversion of these catalysts an influence of phase composition on activity cannot be identified here, but Jung and Bell [17] demonstrated that the phase composition plays an important role and affects the reaction rate in methanol synthesis due to the high acidity and basicity as well as the high adsorption capacity. The TPD results showed that the ratio of the adsorption capacity of EtOAc for pure ZrO2 and Cu/ZrO2 is correlated with the amount of acidic surface sites probed by ammonia. It is clearly seen that ZrO2 alone has the ability to adsorb ethyl acetate dissociatively, implying that surface sites such as ZrAOH and Zrb+ are present. The hydrogenation of the ethoxy species is very fast leading to a high amount of ethanol, whereas the acetyl species remains longer at the surface leading to acetaldehyde or, after full hydrogenation, to ethanol [26]. Additionally, Eq. (15) was found to be equilibrated on the product site [24] leading to a higher amount of ethanol formed. When no H2 is supplied during desorption, ethanol is formed as the main product with a maximum rate at 360 K (Fig. 8). These results lead to the conclusion that ethyl acetate adsorbs dissociatively on ZrO2 and is then further hydrogenated by hydrogen species originating from the dehydrogenation of the adsorbed organic species. Also, dehydroxylation occurs by desorption of terminal hydroxyl groups yielding water [37], which can lead to further hydrolysis H2 O þ Zr
O
CH2 CH3

Zr
OH þ EtOH. When using the Cu/ZrO2 catalyst, formed hydrogen can desorb from the Cu0 nanoparticles. In summary, during steady-state hydrogenation Cu has high relevance providing atomic hydrogen for the reaction as illustrated in Fig. 10. Furthermore, Cu effects ZrO2 positively with respect to the

Fig. 10. Schematic illustration of the proposed mechanism of ethyl acetate hydrogenation.

J. Schittkowski et al. / Journal of Catalysis 352 (2017) 120–129

phase present and its stability. Thus, Cu/ZrO2 acts in a bifunctional manner: Cu0 is crucial for the reaction providing atomic hydrogen by dissociative adsorption, but the reaction rate is more affected by the tight contact between Cu0 and ZrO2. This synergistic effect and the interaction between Cu and ZrO2 are essential and part of ongoing spectroscopic work. 5. Conclusions ZrO2-supported Cu nanoparticles catalysts obtained by coprecipitation achieved high conversion in the hydrogenation of ethyl acetate. For this probe reaction the experimental parameters were established to achieve kinetic control in the absence of deactivation in order to obtain structure-activity correlations. It was shown that temperatures higher than 513 K lead to coke formation, which can be suppressed by using a large excess of H2. Power-law kinetics were determined yielding an apparent activation energy of 74 kJ mol
1 and reaction orders of 0.1–0.3 for H2 and
0.4 to 0.1 for ethyl acetate in the temperature range from 473 K to 503 K. The catalyst showed high stability resulting in a minor degree of relative deactivation of 4% during a period of 4 d at 513 K using a ratio of ethyl acetate to H2 of 1:68. A series of Cu/ZrO2 catalysts with a different Cu loading was prepared, and the incorporation of Cu ions within the ZrO2 matrix was found to influence the phase composition. It was shown that Cu0 is essential for high conversion, but there was no linear correlation with the Cu surface area determined by reactive frontal chemisorption of N2O and H2 TPD experiments. The low specific Cu surface areas of less than 4 m2 g
1 cat indicate that the Cu nanoparticles with sizes in the range from 5 nm to 14 nm are embedded in the X-ray amorphous ZrO2 matrix. As shown by the TPD and H2 TPSR experiments with ethyl acetate, ZrO2 provides adsorption sites for ester dissociation, where also its hydrogenation takes place due to hydrogen spillover induced by dissociative H2 adsorption on the metallic Cu nanoparticles. Acknowledgement The authors gratefully acknowledge the fruitful cooperation with BASF SE. References [1] L. Castro, P. Reyes, C. Correa, J. Sol-Gel Sci. Techn. 25 (2002) 159–168, http://dx. doi.org/10.1023/A:1019920531309. [2] K. Tanabe, T. Yamaguchi, Catal. Today 20 (1994) 185–197, http://dx.doi.org/ 10.1016/0920-5861(94)80002-2. [3] V. Ramaswamy, M. Bhagwat, D. Srinivas, A.V. Ramaswamy, Catal. Today 97 (2004) 63–70, http://dx.doi.org/10.1016/j.cattod.2004.06.141. [4] A.G. Sato, D.P. Volanti, D.M. Meira, S. Damyanova, E. Longo, J.M.C. Bueno, J. Catal. 307 (2013) 1–17, http://dx.doi.org/10.1016/j.jcat.2013.06.022. [5] N. Takezawa, M. Shimokawabe, H. Hiramatsu, H. Sugiura, T. Asakawa, H. Kobayashi, React. Kinet. Catal. Lett. 33 (1987) 191–196, http://dx.doi.org/ 10.1007/BF02066722. [6] J. Agrell, H. Birgersson, M. Boutonnet, I. Melián-Cabrera, R.M. Navarro, J.L.G. Fierro, J. Catal. 219 (2003) 389–403, http://dx.doi.org/10.1016/S0021-9517(03) 00221-5.

129

[7] C.F. Grain, J. Am. Ceram. Soc. 50 (1967) 288–290, http://dx.doi.org/10.1111/ j.1151-2916.1967.tb15111.x. [8] R.P. Ingel, D. Lewis III, J. Am. Ceram. Soc. 69 (1986) 325–332, http://dx.doi.org/ 10.1111/j.1151-2916.1986.tb04741.x. [9] Y. Sun, P.A. Sermon, J. Chem. Soc., Chem. Commun 16 (1993) 1242–1244, http://dx.doi.org/10.1039/C39930001242. [10] G. Wu, Y. Sun, Y.-W. Li, H. Jiao, H.-W. Xiang, Y. Xu, J. Mol. Struct. -TheoChem 626 (2003) 287–293, http://dx.doi.org/10.1016/S0166-1280(03)00102-7. [11] S. Velu, K. Suzuki, M. Okazaki, M.P. Kapoor, T. Osaki, F. Ohashi, J. Catal. 194 (2000) 373–384, http://dx.doi.org/10.1006/jcat.2000.2940. [12] M.K. Dongare, V. Ramaswamy, C.S. Gopinath, A.V. Ramaswamy, S. Scheurell, M. Brueckner, E. Kemnitz, J. Catal. 199 (2001) 209–216, http://dx.doi.org/10.1006/ jcat.2001.3173. [13] K. Larmier, W.-C. Liao, S. Tada, E. Lam, R. Verel, A. Bansode, A. Urakawa, A. Comas-Vives, C. Coperet, Angew. Chem. Int. Ed. Engl. 56 (2017) 2318–2323, http://dx.doi.org/10.1002/anie.201610166. [14] K.-T. Jung, A.T. Bell, Catal. Lett. 80 (2002) 63–68, http://dx.doi.org/10.1023/ A:1015326726898. [15] M.D. Rhodes, A.T. Bell, J. Catal. 233 (2005) 198–209, http://dx.doi.org/10.1016/ j.jcat.2005.04.026. [16] K. Samson, M. S´liwa, R.P. Socha, K. Góra-Marek, D. Mucha, D. Rutkowska-Zbik, J.-F. Paul, M. Ruggiero-Mikołajczyk, R. Grabowski, J. Słoczyn´ski, ACS Catal. 4 (2014) 3730–3741, http://dx.doi.org/10.1021/cs500979c. [17] K.-D. Jung, A.T. Bell, J. Catal. 193 (2000) 207–223, http://dx.doi.org/10.1006/ jcat.2000.2881. [18] K.-T. Jung, A.T. Bell, J. Catal. 204 (2001) 339–347, http://dx.doi.org/10.1006/ jcat.2001.3411. [19] T. Otroshchenko, S. Sokolov, M. Stoyanova, V.A. Kondratenko, U. Rodemerck, D. Linke, E.V. Kondratenko, Angew. Chem. 127 (2015) 16107–16111, http://dx. doi.org/10.1002/ange.201508731. [20] A.G. Sato, D.P. Volanti, J.V. Nicaci, E. Longo, J.M.C. Bueno, Congress Brasileiro de Engenharia Quimca (2016) 1–8. [21] I. Ro, Y. Liu, M.R. Ball, D.H.K. Jackson, J.P. Chada, C. Sener, T.F. Kuech, R.J. Madon, G.W. Huber, J.A. Dumesic, ACS Catal. 6 (2016) 7040–7050, http://dx.doi.org/ 10.1021/acscatal.6b01805. [22] H. Adkins, K. Folkers, Am. Chem. Soc (1931) 1095–1097. [23] J.W. Evans, M.S. Wainwright, N.W. Cant, D.L. Trimm, J. Catal. 88 (1984) 203– 213, http://dx.doi.org/10.1016/0021-9517(84)90064-2. [24] M.A. Natal Santiago, M.A. Sánchez-Castillo, R.D. Cortright, J.A. Dumesic, J. Catal. (2000) 16–28, http://dx.doi.org/10.1006/jcat.2000.2883. [25] P. Claus, M. Lucas, B. Lücke, T. Berndt, P. Birke, Appl. Catal., A 79 (1991) 1–18, http://dx.doi.org/10.1016/0926-860X(91)85002-F. [26] A.K. Agarwal, N.W. Cant, M. Wainwright, D.L. Trimm, J. Mol. Catal. (1987) 79– 92, http://dx.doi.org/10.1016/0304-5102(87)87023-2. [27] T. Turek, D.L. Trimm, N.W. Cant, Catal. Rev. - Sci. Eng. 36 (1994) 645–683, http://dx.doi.org/10.1080/01614949408013931. [28] X. Huang, N. Cant, J. Evans, M. Wainwright, Catal. Today 93-95 (2004) 113– 119, http://dx.doi.org/10.1016/j.cattod.2004.06.020. [29] T.Y. Yan, L.F. Albright, L.C. Case, From industrial & engineering chemistry product research and development (1965) 101–107. [30] A.K. Agarwal, N.W. Cant, M. Wainwright, D.L. Trimm, J. Mol. Catal. (1988) 247– 254, http://dx.doi.org/10.1016/0304-5102(88)80014-2. [31] O. Hinrichsen, T. Genger, M. Muhler, Chem. Eng. Technol. 23 (2000) 956–959, doi:10.1002/1521-4125(200011)23:11<956::AID-CEAT956>3.0.CO;2-L. [32] J. Schittkowski, D. Buesen, K. Tölle, M. Muhler, Catal. Lett. 5 (2016) 1011–1017, http://dx.doi.org/10.1007/s10562-016-1712-y. [33] T. Genger, O. Hinrichsen, M. Muhler, Catal. Lett. 59 (1999) 137–141, http://dx. doi.org/10.1023/A:1019076722708. [34] G.C. Chinchen, J. Catal. 103 (1987) 79–86, http://dx.doi.org/10.1016/00219517(87)90094-7. [35] P.B. Rasmussen, P.M. Holmblad, T. Askgaard, C.V. Ovesen, P. Stoltze, J.K. Nørskov, I. Chorkendorff, Catal. Lett. 26 (1994) 373–381, http://dx.doi.org/ 10.1007/BF00810611. [36] R.A. Köppel, A. Baiker, Appl. Catal., A 84 (1992) 77–102, http://dx.doi.org/ 10.1016/0926-860X(92)80340-I. [37] S. Kouva, K. Honkala, L. Lefferts, J.M. Kanervo, Catal. Sci. Technol. 5 (2015) 3473–3490, http://dx.doi.org/10.1039/C5CY00330J.