Steam Reforming of Acetic Acid over Co ... - ACS Publications

---

Research Article pubs.acs.org/journal/ascecg

Steam Reforming of Acetic Acid over Co-Supported Catalysts: Coupling Ketonization for Greater Stability Stephen D. Davidson, Kurt A. Spies, Donghai Mei, Libor Kovarik, Igor Kutnyakov, Xiaohong S. Li, Vanessa Lebarbier Dagle, Karl O. Albrecht, and Robert A. Dagle* Energy and Environmental Directorate, Institute for Integrated Catalysis, Pacific Northwest National Laboratory, 902 Battelle Boulevard, Richland, Washington 99352, United States S Supporting Information *

ABSTRACT: We report on the markedly improved stability of a novel 2-bed catalytic system, as compared to that of a conventional 1-bed steam reforming catalyst, for the production of H2 from acetic acid. The 2-bed catalytic system consists of (i) a basic oxide ketonization catalyst for the conversion of acetic acid to acetone, and (ii) a Co-based steam reforming catalyst, both catalytic beds placed in sequence within the same unit operation. Steam reforming catalysts are particularly prone to catalytic deactivation when steam reforming acetic acid, used here as a model compound for the aqueous fraction of biooil. Catalysts consisting of MgAl2O4, ZnO, CeO2, and activated carbon (AC) both with and without Co-addition were evaluated for conversion of acetic acid, and its ketonization product, acetone, in the presence of steam. It was found that over the bare oxide support only ketonization activity was observed, and coke deposition was minimal. With addition of Co to the oxide support steam reforming activity was facilitated, and coke deposition was significantly increased. Acetone steam reforming over the same Co-supported catalysts demonstrated more stable performance and with less coke deposition than with acetic acid feedstock. DFT analysis suggests that, over Co, surface CHxCOO species are more favorably formed from acetic acid versus acetone. These CHxCOO species are strongly bound to the Co catalyst surface and could explain the higher propensity for coke formation from acetic acid. On the basis of these findings, in order to enhance stability of the steam reforming catalyst, a dual-bed (2-bed) catalyst system was implemented. Upon comparison of the 2-bed and 1-bed (Cosupported catalyst only) systems under otherwise identical reaction conditions, the 2-bed demonstrated significantly improved stability, and coke deposition was decreased by a factor of 4. KEYWORDS: Biomass, Biofuel, Steam reforming, Ketonization, Acetic acid, Acetone, Hydrogen, Cobalt, Ceria, Hydrothermal liquefaction, Pyrolysis

■

INTRODUCTION Over the past several decades, worldwide energy demand and the environmental impact of fossil fuels have driven numerous advances in renewable energy technology. Of these, biomass has the unique potential to replace fossil fuels for transportation fuel production. Biomass-derived liquid fuels can be produced from direct liquefaction technologies that include fast pyrolysis, catalytic fast pyrolysis, and hydrothermal liquefaction.1,2 These conversion processes typically generate biphasic product streams that consist of an organic phase and an aqueous phase.3 The organic fraction can be further hydrotreated to remove oxygen in order to produce liquid fuels, while the aqueous stream is currently considered as a waste stream.4,5 Depending on the severity and extent of the deconstruction within the liquefaction process, the aqueous phase may contain a considerable amount of biogenic carbon. Thus, effective utilization of the aqueous phase oxygenates is highly desirable from an economic perspective.2,6 Analysis of the aqueous phase of bio-oil has identified acetic acid as one of the most consistent and abundant components, making it an excellent model compound.2,7 © 2017 American Chemical Society

While steam reforming of natural gas is the most common means of hydrogen production, steam reforming can also be applied to a wide range of compounds.8,9 Numerous bioderived oxygenated compounds have been studied as feedstocks including acetic acid. The reaction for complete steam reforming of acetic acid is shown below in eq 1. Preciousmetal-based catalysts (e.g., Rh) have been reported to be active for steam reforming, but they also can produce a significant amount of undesirable CH43,10 and are relatively expensive.11 Ni and Co have demonstrated comparable activity relative to platinum-group metals. Ni has similarly shown a high selectivity to CH4; however, Co does not demonstrate this high CH4 selectivity.12,13 In our prior study for ethylene glycol steam reforming, we rationalized the cause for the lower CH4 selectivity over Co catalysts relative to that of Ni and Rh catalysts via computational modeling.10 We found that over Co there is a relatively higher barrier, for CH3 hydrogenation exists. Received: June 23, 2017 Revised: August 19, 2017 Published: August 25, 2017 9136

DOI: 10.1021/acssuschemeng.7b02052 ACS Sustainable Chem. Eng. 2017, 5, 9136−9149

Research Article

ACS Sustainable Chemistry & Engineering This barrier coupled with relatively facile water activation favors an alternative reaction path, whereby adsorbed CH species undergo hydroxylation thus forming HCOH intermediate (then forming CO and H2) instead of CH hydrogenation.10 CH3COOH + 2H 2O → 2CO2 + 4H 2

Scheme 1. Primary Reaction Pathways for Acetic Acid Conversion over Co-Supported Catalysts

(1)

Steam reforming of oxygenates is more complex than that for hydrocarbons because of the CO bonds that introduce a large number of side reactions. The high tendency of thermally unstable oxygenates to decompose forming carbonaceous deposits, particularly acetic acid, has been recognized as a barrier to commercial implementation.14 Thus, more stable steam reforming catalysts are required. Takanabe et al. found that, for Pt supported on ZrO2, under acetic acid steam reforming conditions, the Pt sites quickly deactivated because of oligomerization reactions, and only the ZrO2 support showed any activity following deactivation.15,16 Lemonidou et al. found that both high oxygen mobility and excess steam were required for stable catalyst performance of acetic acid steam reforming.14 In addition, using isotopic labeling Lemonidou also reported that acetic acid adsorbs on the Rh crystallites forming acetates (CH3COO*) which then decarboxylate and form surface methyl groups that are subsequently converted to CO, CO2, and H2.14 It is has been suggested that these formed acetates can lead to coke formation.17 Relatively few studies have been done on the steam reforming of acetic acid with either Co or Ni. Those that are available have shown similar trends to those observed for the steam reforming of other oxygenates over Co and Ni.18−20 Zhang et al. performed a comparative study of acetic acid steam reforming over Co and Ni supported on Al2O3 and La2O3.18 Here, it was shown that not only is the active metal important, but also the nature of the support greatly affects the activity of the catalyst, particularly at lower temperatures where acetic acid steam reforming over La2O3-supported catalysts was significantly enhanced (both in terms of catalyst activity and stability) compared to catalysts supported on Al2O3.18 Similarly, Goicoechea et al. studied Co and Ni supported on Al2O3 and ZnO.20 Here again, the nature of the support was found to be critical, directing acetone ketonization and decomposition.20 Both studies also demonstrated the importance of steam-to-carbon ratio (S/C) in directing reaction selectivity toward CO2 over CO and mitigating coke formation.18,20 Wang et al. studied acetic acid steam reforming over coprecipitated Co−Fe catalysts.19 Here, it was demonstrated that Co alone is capable of steam reforming acetic acid; however, the high S/C needed for stable catalyst performance indicated the importance of support material to adsorb and activate water.21 Both acetic acid and ethanol feedstocks have also been observed to produce acetone, either as a side product or as a reaction intermediate.14,16,22 The reaction for ketonization of acetic acid is shown in eq 2. Acetic acid ketonization has been widely studied over various metal oxides.23,24 Many of the same catalyst supports used for steam reforming also facilitate ketonization reactions. For both steam reforming and ketonization catalysts surface basicity is often desired in order to avoid undesirable dehydration reactions. In addition, surface redox ability is beneficial for formation of hydroxyl groups and desorption of water.8,24 The two primary pathways for acetic acid conversion of Co-supported catalysts, steam reforming, and ketonization, are illustrated in Scheme 1. 2CH3COOH → CH3COCH3 + H 2O + CO2

In our prior study, we reported on the steam reforming of a synthetic fast-pyrolysis-derived aqueous bio-oil over Ni-, Rh-, and Co-based catalysts; here again, lower CH4 selectivity was observed with the Co-based catalyst as well as more stable catalytic performance.3 We also demonstrated an optimal catalyst stability at 500 °C, as compared to the higher temperatures investigated (i.e., 700 and 800 °C), as a result of decreased carbon formation.3 Thus, a concept analogous to the petroleum industries’ use of a prereformer, operated at approximately 500 °C for steam reforming of the heavier naphtha components, could be applied in the case of steam reforming biomass-derived oxygenates.3 However, while carbon formation was reduced when operating at 500 °C, it remained problematic. In this work, we investigate the catalytic performance of Co metal supported on different reducible (e.g., CeO2) and nonreducible [e.g., MgAl2O4, ZnO, and AC (activated carbon)] supports for the steam reforming of acetic acid and its ketonization product, acetone. A dual-bed (2-bed) system of ketonization followed by steam reforming was also evaluated, primarily explored as a potential means to suppress carbon formation. While adding excess steam or oxidant to the feedstock is one option for coke reduction, it also adds significant operating cost. Thus, our aim is to develop a catalytic process whereby acetic acid steam reforming is enabled without the requirement of excessive oxidant in the feed but still with minimal coke deposition.

■

EXPERIMENTAL METHODS

Catalyst Preparation. MgAl2O4 is a commercial Sasol Puralox 30/ 150 and was calcined under air at 500 °C for 3 h prior to impregnation. CeO2 was made by mixing ethanol and ceria oxide nanoparticles stabilized in nitrate (NYACOL CeO2(NO3), particle size 10−20 nm), with biopolymer Pluronic F-127, drying at 110 °C for 8 h, and then calcination at 525 °C for 4 h (5 °C/min ramp rate). Activated carbon (AC) was supplied from PICA (coconut-shellderived, 60−100 mesh) and was used as received. The Co/CeO2, Co/ MgAl2O4, and Co/AC catalysts were prepared by incipient wetness impregnation of their respective supports with solutions of cobalt nitrate, Co(NO3)2. After impregnation, the catalysts were dried at 110 °C for 8 h and then calcined under air at 500 °C for 3 h; for the Co/ AC catalyst, calcination was performed under N2. The Co/ZnO was prepared by coprecipitation of Co(NO3)2 and Zn(NO3)2 aqueous solutions as described elsewhere.11 Commercially supplied ZnO typically has a relatively low surface area (e.g., ∼5 m2/g), and thus Co/ZnO was prepared using coprecipitation in order to obtain a higher surface area (i.e., 24 m2/g, as shown in Table 1) and to remain consistent with prior steam reforming studies.11 The nominal cobalt loadings were 15 wt %. In this work, the catalysts are referenced by the weight percent of Co and the support material used. For example, 15 wt % Co supported on ZnO is designated as 15%Co/ZnO.

(2) 9137

DOI: 10.1021/acssuschemeng.7b02052 ACS Sustainable Chem. Eng. 2017, 5, 9136−9149

Research Article

ACS Sustainable Chemistry & Engineering

catalysts were reduced at 500 °C for 8 h under a 10% H2/N2 gas mixture. The MgAl2O4-supported catalysts were reduced at 850 °C. Carbon balance for all experiments was found to be 85% or higher. The catalysts were tested at atmospheric pressure and at a reaction temperature of 500 °C. Nitrogen gas (typically ∼20 mol % in the feed) was introduced into the system by a Brooks mass flow controller (5890 E series) to serve as the carrier gas and internal reference standard. The liquid reactant feed was introduced using an HPLC pump (Chrom Tech series 1500) through a 1/16 in. stainless-steel line to a microchannel vaporizer set at 150 °C. For the evaluations with acetic acid, a 32.3 wt % acetic acid in water feed was used. This corresponds to a molar steam-to-carbon ratio (S/C) of 3.5. For the evaluations with acetone, a 23.5 wt % acetone in water feed was used. This corresponds to a molar steam-to-carbon ratio (S/C) of 3.5. The flow rate of dry gas products was measured by a digital flow meter (DryCal). The gas composition was determined by a fourchannel Agilent Micro GC equipped with MS-5A, PPU, alumina, and OV-1 columns and a TCD detector for each column. Liquid products were trapped in a 1 L condenser cooled by circulating ethylene glycol for postrun analysis by LC. The liquid product from the reforming experiments was injected with and without dilution into an Agilent 1100 LC system with an Aminex HPX-87H ion exclusion column (Bio-Rad), 300 mm long, 7.8 mm inner diameter guard column. The column was eluted isocratically with 0.005 M sulfuric acid through a refractive index detector with an optical temperature of 35 °C. The amount of carbon deposited on the spent catalyst was measured by a Vario EL cube elemental analyzer. Spent samples were weighed on a Mettler XP6 microbalance into tin foil boats for analysis. Conversion was calculated from a simple mass balance on either acetic acid or acetone.

Table 1. Catalyst Physiochemical Characterizations

a b

catalyst

BET surface area [m2/g]

pore volume [cm3/g]

MgAl2O4 ZnO CeO2 AC 15%Co/MgAl2O4 15%Co/ZnO 15%Co/CeO2 15%Co/AC

142 17 148 1430 107 24 93 666

0.57 0.31 0.40 0.25 0.44 0.30 0.22 0.17

Co0 crystallite sizea [nm] N/Ab N/A N/A N/A 15.7 ± 13.5 ± 11.9 ± 20.2 ±

1.7 1.5 1.6 2.6

Co0 crystallite size calculated using XRD by the Scherrer equation. N/A, not applicable.

Catalyst Characterization. BET surface area and BJH pore volume and pore size measurements were performed on a Micromeritics Tristar 3000 instrument using N2 adsorption at 77 K. Between 0.1 and 0.3 g of sample was used for each measurement. Prior to adsorption measurements, samples were degassed under vacuum at 150 °C for 12 h. A Micromeritics Autochem 2920 instrument was used to perform H2 temperature-programmed reduction (H2-TPR) experiments and temperature-programed oxidation (TPO) of spent catalyst. For H2TPR, 0.05 g of sample was loaded and pretreated under 50 sccm of He at 120 °C for 120 min. The sample was then cooled to ambient temperature and the gas flow changed to 5%H2/Ar at 50 sccm. Temperature was then ramped to 800 °C at 10 °C/min. For TPO, 0.10 g of spent catalyst with α-Al2O3 diluent was loaded and pretreated under 50 sccm of He at 120 °C for 120 min. The sample was then cooled to ambient temperature and the gas flow changed to 5% O2/He at 50 sccm. Temperature was then ramped to 800 °C at 5 °C/min. For both H2-TPR and TPO analysis, gases were monitored by TCD. X-ray diffraction (XRD) measurements were performed on a Philip’s X’Pert XRD with a Cu Kα source set to 50 kV and 40 mA. For fresh measurements, the samples were finely ground and packed into an amorphous glass holder. For reduced measurements, samples were finely ground then reduced ex situ under 50 sccm of 5% H2/Ar at 500 °C (15%Co/MgAl2O4 was the only sample reduced at 825 °C) for 2 h. The samples were then cooled to ambient temperature and passivated under 1% O2/N2 overnight. The passivated samples were then loaded into an amorphous glass holder. XRD patterns were collected from 10−80° 2θ at a rate of 2° 2θ/min and a step size of 0.06° 2θ/step. Peaks were analyzed and fit using MDI Jade 9 software, and a reference background to powder Si was used for the peak fitting process. Co0 particle size was calculated using the Scherrer equation based on the 44.3° 2θ for 15%Co/ZnO and 15%Co/CeO2 samples and the 51.5° 2θ for 15%Co/MgAl2O4. The STEM analysis was performed with an FEI Titan 80−300 microscope operated at 300 kV. The instrument is equipped with a CEOS GmbH double-hexapole aberration corrector for the probeforming lens, which allows for imaging with 0.1 nm resolution in scanning transmission electron microscopy (STEM) mode. The images were acquired with a high-angle annular dark field (HAADF) detector with inner collection angle set to 52 mrad. Elemental analysis was performed with Gatan’s electron energy loss spectrometer (EELS). Reactivity Measurements. Steam reforming and ketonization reactions were performed in a 1/2 in. o.d. fixed-bed alumina reactor. Alumina reactors were used to minimize potential side reactions that might occur with, for example, stainless steel reactors. Catalyst loadings of 0.20 g were used, except for acetic acid steam reforming over CeO2-based catalysts where 0.10 g of catalyst was used. The catalyst was diluted with α-Al2O3 (20:1 weight ratio of α-Al2O3-tocatalyst) in order to minimize the effect of heat transfer guises as discussed in our prior work.25 Catalyst and the α-Al2O3 mixture were mixed and held in place with quartz wool. A K-type thermocouple was placed in the center of the loaded catalyst bed to measure catalyst bed temperature. Prior to testing, ZnO-, AC-, and CeO2-supported

⎛ R fed − (R out,gas + R out,liquid) ⎞ ⎟⎟100 X = ⎜⎜ R fed ⎝ ⎠

(3)

Here, the terms are as follows: conversion is X, Rfed is the calculated feed rate of reactant (i.e., acetic acid or acetone) based on mass change during run period and feed composition, Rout,gas is reactant detected on the Agilent Micro GC integrated over the run, and Rout,liquid is reactant detected in LC analysis. Selectivity was calculated on a carbon basis using the following equation:

⎛ C productFproduct ⎞ ⎟⎟100 Sproduct = ⎜⎜ ⎝ ∑ C productFproduct ⎠

(4)

Here, the terms are as follows: Sproduct is selectivity to a product compound (e.g., CO2), Cproduct is the number of carbons in the product molecule, and Fproduct is the molar flow rate of the product. H2 yield was calculated based on the stoichiometric H2 value from eq 1 using the following equation:

⎛ FH2 ⎞ YH2 = ⎜ ⎟100 ⎝ 4Faceticacid ⎠

(5)

Here, FH2 is the outlet molar flow of H2, and Faceticacid is the molar feed rate of acetic acid. Computational Details. All calculations were performed using spin-polarized density functional theory (DFT) within the generalized gradient approximation (GGA) as implemented in the Vienna ab initio simulation package (VASP).26−28 The core and valence electrons were represented by the projector augmented wave (PAW) method29,30 with a kinetic cutoff energy of 400 eV. The exchange correlation functional was described by the Perdew−Burke−Ernzerhof (PBE) functional.31 The Co(0001) plane has been reported to be representative of bulk Co surfaces, and hence Co(0001) was used to mimic the supported Co catalyst for acetic acid conversion in this work.32,33 The ground-state atomic geometries of the clean and the adsorbed Co(0001) was obtained by minimizing the forces on each atom to below 0.03 eV/Å. A periodic p(2 × 2) supercell Co(0001) surface slab 9138

DOI: 10.1021/acssuschemeng.7b02052 ACS Sustainable Chem. Eng. 2017, 5, 9136−9149

Research Article

ACS Sustainable Chemistry & Engineering

Figure 1. XRD patterns for the 15 wt % cobalt catalysts supported on (A) MgAl2O4, (B) ZnO, (C) CeO2, and (D) AC. α = bare support; β = Cosupported catalyst prior to reduction; γ = Co-supported catalyst after reduction and passivation. Dotted lines mark features of the support material; φ = Co3O4; ρ = Co0. with four atomic layers was used in this work. During the geometric optimizations and the transition-state searching processes, the adsorbate(s) and the metal atoms in the top two atomic layers were allowed to relax while the metal atoms in the bottom two atomic layers of the surface slab were fixed. A 15 Å vacuum layer was inserted between the Co(0001) surface slab in the z direction to avoid nonphysical interaction artifacts between the periodic systems modeled. To ensure the accuracy of calculations, the effects of slab thickness (up to six atomic layers) and different Monkhorst−Pack (MP) mesh sampling were tested. A (3 × 3 × 1) MP sampling schedule was found to be accurate to reach the total energy convergence of