Varkolu Mohan,
Chakali Raghavendra,
Chodimella Venkata Pramod,
Burri David Raju and
Kamaraju Seetha Rama Rao*
Indian Institute of Chemical Technology, Hyderabad, India. E-mail: ksramarao@iict.res.in
First published on 17th December 2013
Among Ni supported on H-ZSM-5 catalysts with various loadings of Ni, a catalyst with 30 weight% Ni has been identified as an effective catalyst for the hydrogenation of levulinic acid to γ-valerolactone in vapour phase at atmospheric pressure. The catalysts have been characterized by different techniques such as XRD, BET surface area, pore size distribution, TPR, AAS, pulse chemisorption, FE-SEM-EDS, TEM, XPS and pyridine-adsorbed IR. XRD patterns suggest that the structure of H-ZSM-5 is intact even after incorporation of Ni. Pyridine-adsorbed IR patterns reveal the presence of both Lewis and Brønsted acid sites, which are responsible for the dehydration and cleavage of γ-valerolactone. The superior activity exhibited by the 30 wt% Ni/H-ZSM-5 catalyst (demonstrating the highest productivity of 0.9090 kgGVL kgcatalyst−1 h−1 at 250 °C) is due to the presence of a greater number of surface Ni species. Non-noble metal Ni-based catalysts offer advantages over the present processes involving Ru-based catalysts, which have certain constraints, some of which are unavoidable such as cost ineffectiveness, the use of volatile organic solvents and high pressure operation. The 30 wt% Ni/H-ZSM-5 catalyst can be seen as an alternative and promising catalyst which could be of great importance to the chemical industry.
Cellulosic biomass (a non-food renewable source) normally consists of biopolymeric species such as cellulose (35–50%), hemicellulose (15–25%) and lignin (15–30%). Acidic hydrolysis of cellulosic biomass gives levulinic acid, formic acid, furfural and char.7–11 Therefore, biomass-derived levulinic acid (LA) is an effective feedstock for the production of γ-valerolactone (GVL). GVL has several uses in solvents, insecticides, adhesives etc.12–14 Furthermore, GVL is a potential candidate for the production of fine chemicals, fuels and fuel additives.12,15–21 A key step in the production of levulinic acid is the acid hydrolysis of furfuryl alcohol, which is obtained from an abundant source of biomass and also from the extraction of black liquor in the paper-making process.22 Few researchers have studied the direct conversion of biomass and biomass-derived platform molecules to GVL.23,24 GVL serves as an intermediate for the production of diesel fuel and the production of branched alkanes with molecular weights appropriate for jet fuel. Approximately 50 kg of diesel/fuel can be formed per 100 kg of GVL. Alternatively, GVL can undergo catalytic conversion to produce butane and CO2, combining with butene oligomerization to form C8–C20 alkenes. It can also undergo ring-opening and reduction steps to form pentanoic acid, followed by ketonization to form 5-nonanone, and finally completed by hydrodeoxygenation to form nonane.
Christian et al. studied the liquid phase hydrogenation of LA under batch conditions using a Raney® Ni catalyst at 220 °C and 48 bar hydrogen pressure and over copper chromite at 250 °C and 220 bar hydrogen pressure.25 Schutte et al. reported LA hydrogenation over a platinum oxide catalyst in various organic solvents at 3 bar hydrogen pressure at room temperature.26 A patent filed by Manzer discloses the hydrogenation of LA in supercritical CO2 over Pt, Pd, Ru and Re catalysts supported over various oxides at 200 °C and 200 bar hydrogen pressure in 1,4-dioxane solvent.27 Similarly, Bourne et al. utilised supercritical CO2 for LA hydrogenation over Ru/SiO2 at 200 °C in 100 bar hydrogen pressure.28 Yan et al. used Ru/C in a batch reactor at 130 °C and 12 bar hydrogen pressure to yield GVL.29 Pt/TiO2 or Pt/ZrO2 catalysts have also been reported for the hydrogenation of levulinic acid at 200 °C, 40 bar H2 with a H2/LA molar ratio of 5:1, to obtain a greater amount of GVL.8 Dumesic et al. investigated LA hydrogenation over Ru/C and Ru–Sn/C at 180 °C and 35 bar hydrogen pressure.30,31 Recently, Galletti et al. reported a higher yield of GVL at 70 °C and 30 bar hydrogen pressure over a Ru/C catalyst using an ion exchange resin (Amberlyst A70) as a co-catalyst.32 Hengue et al. reported using Cu on various supports to obtain GVL at 200 °C and 35 bar pressure.33
It is interesting to note that most of them reported the use of batch conditions at very high H2 pressures. Only a few reports have been published describing experiments in the vapour phase at atmospheric pressure.34,35 However, 1,4-dioxane is used as a feed additive in all of these works.
The present investigation highlights the application of zeolite (H-ZSM-5) supported Ni catalysts to obtain GVL with high productivity via LA hydrogenation in the vapour phase without any feed additive at atmospheric pressure.
Field Emission Scanning Electron Microscopy (FE-SEM) (S-4800, Hitachi Company, Japan) along with Energy Dispersive X-ray analysis (EDS) (LINK ISIS-300, Oxford Instruments) was used to study the morphology as well as the composition of H-ZSM-5 supported Ni catalysts, respectively. The Ni content of the catalysts was also estimated by dissolving a known amount of catalyst in aqua regia followed by dilution with water and by using atomic absorption spectroscopy (AAS) (A-300, PerkinElmer, Germany). For the quantification of Ni, a 10 ppm Ni solution was used. FTIR spectra were recorded on a GX spectrometer (PerkinElmer, Germany) with a scan range of 4000–400 cm−1. The samples were finely ground in the ratio 1:10 (sample:KBr) and pelletized then measured in the spectral range of 400–4000 cm−1 with a resolution of 4 cm−1, and 10 scans were recorded for each spectrum. For pyridine-adsorbed FTIR analysis, the sample was oven dried at 100 °C for 1 h. To the oven dried sample (50 mg), 0.1 ml of pyridine was admixed directly. To remove the physisorbed pyridine present in the sample, the admixed sample tube was then kept in a vacuum oven at 120 °C for 1 h. After ice cooling to room temperature, the spectrum was recorded with a nominal resolution of 4 cm−1 in the spectral range of 1400–1700 cm−1 using a KBr background and 10 scans were recorded for each spectrum.36
The textural characteristics of the catalysts were obtained using a JEOL JEM 2000EXII transmission electron microscope, operating between 160 and 180 kV. The specimens were prepared by dispersing the samples in methanol using an ultrasonic bath and evaporating a drop of the resultant suspension onto a lacey carbon support grid.
Temperature programmed reduction (TPR) of the catalysts was performed on an AUTOSORB-iQ automated gas sorption analyser (Quantacrome instruments, USA). About 50 mg of the catalyst was placed in a quartz reactor and pre-treated in a He flow at 393 K for 1 h. Later, the catalyst was treated with 5% H2–Ar mixture gas (3600 cm3 h−1) with a temperature ramp of 20 °C min−1. The hydrogen consumption was monitored using a thermal conductivity detector (TCD).
H2-chemisorption using a pulse (100 μL) titration procedure was carried out at 40 °C on an Autosorb-iQ automated gas sorption analyser (Quantachrome Instruments, USA) to determine the dispersion and metal particle size, and metal surface area of the catalyst. Prior to the experiment, the catalyst was reduced at 500 °C for 2 h followed by evacuation for 2 h. The monolayer uptake, active metal surface area, metal dispersion, and active crystallite size were calculated.37–39
Monolayer uptake of hydrogen in micromoles per gram was calculated by
Nm = 44.61Vm |
AMSA = NmSAm/166 |
Metal dispersion (D) was calculated by:
D = NmSM/100L |
Active crystallite size was obtained by:
t (nm) = 100Lf/(AMSA × Z) |
The XPS analysis was performed using a KRATOS AXIS 165 apparatus equipped with a dual anode (Mg and Al) using a Mg Kα source. The non-monochromatized Al Kα X-ray source (hν = 1486.6 eV) was operated at 12.5 kV and 16 mA. Analysis was done at room temperature and prior to analysis the samples were maintained under a rigorous vacuum, typically in the order of 10−8 Pa to avoid the presence of contaminates. All binding energies measured were within an accuracy of ± 0.2 eV. For energy calibration, the carbon 1s binding energy was taken as a reference value of 284.6 eV .
Fig. 1 XRD profiles of reduced Ni/H-ZSM-5 catalysts (a) HZ, (b) 5NHZ, (c) 10NHZ, (d) 20NHZ, (e) 30NHZ, (f) 40NHZ and (g) 50NHZ. |
Catalyst | SBET (m2 g−1) | Vp (cm3 g−1) | Dp (Å) | d (nm) | EDS composition (wt%) | AAS composition (wt%) |
---|---|---|---|---|---|---|
HZ | 305.6 | 0.21 | 27.3 | — | — | — |
5NHZ | 292.8 | 0.20 | 26.8 | 9.82 | 4.8 | 4.1 |
10NHZ | 274.2 | 0.16 | 23.1 | 10.09 | 8.3 | 8.7 |
20NHZ | 215.9 | 0.16 | 28.7 | 10.39 | 15.5 | 16.7 |
30NHZ | 209.1 | 0.15 | 28.2 | 11.03 | 28.9 | 29.6 |
40NHZ | 173.5 | 0.11 | 25.8 | 11.39 | 33.0 | 32.4 |
50NHZ | 161.5 | 0.12 | 30.7 | 12.17 | 36.1 | 38.5 |
Fig. 2 depicts the N2 adsorption–desorption isotherms of H-ZSM-5 with various loadings of Ni supported on the H-ZSM-5. The nitrogen adsorption isotherms of the catalysts show the presence of a hysteresis loop, associated with capillary condensation, which indicates the formation of a mesoporous structure (secondary pores).40 The surface area (SBET), pore volume (Vp) and average pore size (Dp) of the samples are given in Table 1. The surface area of H-ZSM-5 is 305.6 m2 g−1 and the pore volume is 0.21 cm3 g−1. However, there is a marginal decrease in surface area from 292.8 to 161.5 m2 g−1 and in the pore volume from 0.20 to 0.11 cm3 g−1 as the Ni loading increases from 5 to 50 wt% due to pore blockage by Ni particles. The blockage of pores by Ni creates a strain on the structure of the zeolite that results in changes to the isotherms.41 In addition to the blockage of H-ZSM-5 pores by NiO particles, extra porosity due to NiO particles might have resulted. Because of this, the average pore diameter remains the same with loading. As can be seen from the XRD results, there is no significant change in the structural properties of the zeolite.
Fig. 2 N2 adsorption–desorption profiles of Ni/H-ZSM-5 catalysts (a) HZ, (b) 5NHZ, (c) 10NHZ, (d) 20NHZ, (e) 30NHZ, (f) 40NHZ and (g) 50NHZ. |
The TPR profiles of the catalysts are shown in Fig. 3. The peak observed at higher temperature for the 5NHZ sample indicates the presence of strong metal–support interactions, and the intensity of this phenomenon decreases as the metal loading increases for subsequent catalysts of the series. The peak at 500 °C can be ascribed to the reduction of smaller NiO particles and the other (low temperature signal at a Tmax of 400 °C) due to bulk NiO.42 Maia et al. compared the reduction patterns of Ni/H-ZSM-5 prepared by different methods such as impregnation and ion exchange method and concluded that if any H+ is exchanged by the incorporated Ni2+ in the zeolite, this could be reduced at a higher temperature than that located outside the zeolite.43 In addition, Pawelec et al. anticipated that nickel distributed in the zeolite was reduced above 630 °C, at the same time as the reduction of nickel oxide started around 500 °C.44,45 These results suggest that in all our prepared catalysts, NiO is located predominantly outside the zeolite, which was clearly indicated by the XRD patterns (preservation of zeolite structure in all the catalysts). To the best of our knowledge, in previous studies, various domains were suggested such as <500 °C, 500–630 °C and >630 °C. Among which, <500 °C could be assigned to the reduction of NiO, the 500–630 °C zone corresponds not only to smaller Ni particles but also to Ni oligomeric species and >630 °C might be due to Ni2+ in the zeolite.46 In the present case, no such peak was observed at higher temperature, which clearly demonstrates that Ni resides predominantly outside the zeolite.
Fig. 3 TPR profiles of Ni/H-ZSM-5 catalysts (a) 5NHZ, (b) 10NHZ, (c) 20NHZ, (d) 30NHZ, (e) 40NHZ and (f) 50NHZ. |
FE-SEM pictures (depicted in Fig. S1†) show a cubic shaped morphology. The composition of Ni obtained from EDS patterns (depicted in Fig. S2†) and AAS is very close to the theoretical value up to 30 wt% Ni (shown in Table 1). In the impregnation method, the deposition of metal/metal oxide on a support is generally limited beyond a certain amount. In the present case, 30 wt% appears to be a limiting value. It depends on the nature of the metal and the porosity of the support.
Transmission electron microscopy (TEM) and selected area electron diffraction (SAED) images of the reduced and used catalysts of 30NHZ (depicted in Fig. S3†) indicate that the average particle size of the reduced and used catalysts calculated by Image-J software were 10.98 nm and 11.74 nm, respectively, which is close to the particle size obtained from pulse chemisorption (9.22 nm) and also from the XRD (11.03 nm). The SAED pattern of Ni indicates the presence of regular spots and also fringes attributed to the crystalline Ni with fcc structure as reported by Wu et al. (confirmed from the XRD).47
The XPS spectra of Ni species in the reduced and used catalysts are shown in Fig. 4. In the case of the reduced catalyst, two major peaks are observed with binding energies (BE) of 852.5 eV (69.4%) and 869.7 eV (30.6%), corresponding to the core level Ni 2p3/2 and Ni 2p1/2 transitions, while in the case of the used catalyst, two major peaks are observed with BEs of 852.3 eV (70.7%) and 870.3 eV (29.3%), respectively, indicating the presence of nickel in the metallic state.48 The BE values are consistent with the literature reported by Xiao et al.49 No satellite peak was observed with either of the catalysts. The difference between the two peaks (17.2 eV for the reduced and 18 eV for the used) is very close to that reported by Xiao et al.49
Fig. 5 illustrates the FTIR patterns when various loadings of Ni were impregnated on the HZ. The significant broad bands at 1230 cm−1, 1100 cm−1 and 796 cm−1 are attributed to the Si–O–Si asymmetric and symmetric stretching frequencies as well as internal stretching frequencies of the zeolite. The band at 1637 cm−1 is associated with H–O–H for all the catalysts. The absorption bands around 543 and 450 cm−1 are characteristic of the ZSM-5 crystalline structure.50
Fig. 5 FT-IR profiles of Ni/H-ZSM-5 catalysts (a) HZ, (b) 5NHZ, (c) 10NHZ, (d) 20NHZ, (e) 30NHZ, (f) 40NHZ and (g) 50NHZ. |
Fig. 6 shows Py-IR spectra of the catalysts at room temperature, used to investigate the acidity of the catalysts. The bands at 1445, 1595 and 1612 cm−1 are attributed to the adsorption of pyridine coordinated on Lewis acid sites created by metallic Ni as well as the support.51–54 The band appearing at 1490 cm−1 is ascribed to the vibration frequency of the pyridine ring on Brønsted and Lewis acid sites51,52 created by both Ni and the support. In addition, IR bands near 1545 and 1637 cm−1 are due to pyridinium ions formed by the abstraction of a proton from the catalyst support.51,52 The above observations reveal that both Brønsted and Lewis acid sites are present on these catalysts in which one or both may be responsible for the formation of GVL.
Fig. 6 Py-IR profiles of Ni/H-ZSM-5 catalysts (a) HZ, (b) 5NHZ, (c) 10NHZ, (d) 20NHZ, (e) 30NHZ, (f) 40NHZ and (g) 50NHZ. |
It is also a well-known fact that Ni favours the hydrogenation reaction. In previous reports, Ni (111) was demonstrated to be active for the hydrogenation reaction.55 To further clarify the formation of GVL, we carried out the reaction of angelica lactone (obtained from the dehydration of LA over HZM-5) over Ni and observed the formation of GVL and minor quantities of valeric acid. Hence, one can assume that Ni is responsible for GVL formation by the hydrogenation of an intermediate, angelica lactone.
Pulse chemisorption studies were conducted for all the catalysts and the data compiled in Table 2. We observed that the hydrogen uptake increases correspondingly with the Ni loading up to 30NHZ. Thereafter, the uptake value decreased due to a decrease in the number of surface Ni species which form bulk Ni. The average particle size and dispersion are in good agreement with each other.
Catalyst | Nm (μmol g−1) | AMSA (m2 gNi−1) | t (nm) | D (%) |
---|---|---|---|---|
5NHZ | 74.79 | 117.01 | 5.76 | 17.56 |
10NHZ | 137.80 | 107.79 | 6.25 | 16.17 |
20NHZ | 218.00 | 85.27 | 7.90 | 12.79 |
30NHZ | 280.30 | 73.09 | 9.22 | 10.97 |
40NHZ | 231.76 | 45.32 | 14.87 | 6.80 |
50NHZ | 190.68 | 29.83 | 22.59 | 4.48 |
Fig. 7 Reported LA hydrogenation reaction pathways.56–59 |
To gain insight into the reaction pathway, we have carried out the reaction over H-ZSM-5 (51% conversion of LA and selectivity to alpha angelica lactone (A) is 71% and to beta angelica lactone (B) is 29%) that contains Brønsted as well as Lewis acid sites confirmed by pyridine adsorption studies, obtained angelica lactone (A) which isomerised to angelica lactone (B), which on further hydrogenation over Ni gave the target molecule. It is reported that the formation of angelica lactone has been observed over acid catalysts.60,61 Hence, we believe that the reaction proceeds via path I as shown in the scheme. In a similar manner, Galletti et al. studied the role of the acid as a co-catalyst to understand the reaction pathway.32 In our case, we monitored the reaction with GC-MS to confirm the reaction intermediates by carrying out the reaction on the support only, as well as on a Ni-incorporated support.
The presence of angelica lactone (A & B) confirms the pathway of the reaction and GVL formation confirms the further hydrogenation of angelica lactone to yield the desired product. Very recently, W. Luo et al. reported the influence of the support by comparing the acidity of H-ZSM-5, H-β, TiO2 and ZrO2 for the levulinic acid hydrogenation.62
Fig. 8 Effect of temperature and Ni loading on levulinic acid hydrogenation with various NHZ catalysts. |
To gain further insight into the mechanism, the reaction was carried out by pumping a larger quantity of reactants (liquid feed (LA) = 5 cm3 h−1, H2/LA = 5.5) over 1 g of catalyst and 85% conversion of LA was observed along with a 54% yield of GVL, a 16.9% yield of angelica lactone (A), a 7.9% yield of angelica lactone (B) and minor quantities of valeric acid. Hence, we believe that the reaction is propagating in a sequence of steps. So, we proposed the mechanism in our own way as shown in Fig. 10.
Furthermore, we carried out the reaction over Ni/SiO2 and Ni/Al2O3 (prepared by a wet impregnation method) which resulted in lower yields than with the Ni/HZM-5 catalyst at a H2/LA molar ratio of 8. The yields over Ni/SiO2 and Ni/Al2O3 were 87 and 83% with complete conversion, respectively.
Fig. 11 Effect of time-on-stream on levulinic acid hydrogenation over 30NHZ catalyst at 250 °C and a ratio of H2/LA = 8. |
The productivity of other reported catalysts were also calculated and compared with our results and summarized in Table 3. Upare et al. reported a 98.6% yield of GVL over 5% Ru/C while the yield of GVL was 90 and 30% over 5%Pd/C and 5%Pt/C, respectively, at atmospheric pressure.34 Though the yields are good over Ru/C, the feed was diluted (10% LA in 1,4-dioxane) with a solvent (1,4-dioxane). As well as being a highly cost effective non-noble metal catalyst, the productivity of our catalyst is twice that of Ru/C catalyst, whereas the productivity of other catalysts follows the order 5% Cu/SiO2 > 5%Pd/C > 5%Pt/C. The same group observed a 99.9% yield of GVL over Cu/SiO2 at 10 bar pressure and also with the same feed ratio. Significantly, we are getting a 92.2% yield of GVL with a productivity of 0.9090 kgGVL kgcatalyst−1 h−1 without adding any solvent at atmospheric pressure.
Catalyst | Catalyst source/preparation method | Reaction conditions | Reaction temperature (°C) | LA conversion | GVL yield | Productivity [kgGVL kgcatalyst−1 h−1] | Ref. |
---|---|---|---|---|---|---|---|
30NHZ | Wet impregnation method | Vapour phase (1 bar) 100% LA | 250 | 100 | 92.2 | 0.9090 | Present study |
5%Ru/C | Sigma-Aldrich | Vapour phase (1 bar) 10% LA in 1,4-dioxane | 265 | 100 | 98.6 | 0.4545 | 34 |
5%Pd/C | Sigma-Aldrich | Vapour phase (1 bar) 10% LA in 1,4-dioxane | 265 | 100 | 90 | 0.4040 | 34 |
5%Pt/C | Sigma-Aldrich | Vapour phase (1 bar) 10% LA in 1,4-dioxane | 265 | 100 | 30 | 0.1515 | 34 |
5%Cu/SiO2 | Precipitation–deposition method | Vapour phase, H2 (10 bar), 10% LA in 1,4-dioxane | 265 | 100 | 99.9 | 0.4444 | 35 |
The turnover frequency (TOF) of LA on each catalyst is defined as the number moles of LA converted per one surface Ni atom per second. It is calculated by
Effect of TOF on the loading of Ni is displayed in Fig. 12. The number of surface Ni sites are assumed to be equal to half the number of H2 moles adsorbed on the catalyst per gram. The TOF was found to decrease up to a 20 wt% Ni loading, beyond which the TOF was consistent. This clearly indicates that LA hydrogenation is structure-sensitive65,66 up to 20 wt%, above which it is structure-insensitive.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3ra46485g |
This journal is © The Royal Society of Chemistry 2014 |