Continuous Hydrothermal Decarboxylation of Fatty Acids and Their Derivatives into Liquid Hydrocarbons Using Mo/Al2O3 Catalyst.

In this study, we report a single-step continuous production of straight-chain liquid hydrocarbons from oleic acid and other fatty acid derivatives of interest including castor oil, frying oil, and palm oil using Mo, MgO, and Ni on Al2O3 as catalysts in subcritical water. Straight-chain hydrocarbons were obtained via decarboxylation and hydrogenation reactions with no added hydrogen. Mo/Al2O3 catalyst was found to exhibit a higher degree of decarboxylation (92%) and liquid yield (71%) compared to the other two examined catalysts (MgO/Al2O3, Ni/Al2O3) at the maximized conditions of 375 °C, 4 h of space time, and a volume ratio of 5:1 of water to oleic acid. The obtained liquid product has a similar density (0.85 kg/m3 at 15.6 °C) and high heating value (44.7 MJ/kg) as commercial fuels including kerosene (0.78-0.82 kg/m3 and 46.2 MJ/kg), jet fuel (0.78-0.84 kg/m3 and 43.5 MJ/kg), and diesel fuel (0.80-0.96 kg/m3 and 44.8 MJ/kg). The reaction conditions including temperature, volume ratio of water-to-feed, and space time were maximized for the Mo/Al2O3 catalyst. Characterization of the spent catalysts showed that a significant amount of amorphous carbon deposited on the catalyst could be removed by simple carbon burning in air with the catalyst recycled and reused.

Introduction

Renewable resources are of increasing interest to produce fuel range hydrocarbons, especially liquid transportation fuels because of the limited reserves of petroleum fuel. Fats & oil (which mainly contain triglycerides and fatty acids) have been used as renewable feedstocks for producing transportation fuels such as biodiesel[1−4] or green diesel.[5−8] Nonedible resources (jatropha oil, algae, waste cooking oil, animal fats, tallow, etc.) are preferred as feedstocks over edible resources to avoid the food versus fuel issue. However, higher oxygen content and acidity of these feedstocks prevent them from being used directly as fuel because of their corrosive properties and higher viscosity compared to fossil fuels.[9] To lower the oxygen content, deoxygenation is an efficient upgrading pathway to produce diesel range liquid hydrocarbons to minimize any issues related to engine compatibility. Liquid hydrocarbon formed via the deoxygenation process can obtain larger cetane numbers (85 to 99) compared to petroleum diesel (45 to 55).[10]

Decarboxylation is a cost-effective deoxygenation process for liquid hydrocarbon production as it requires less/no hydrogen compared to other processes such as hydrodeoxygenation. The decarboxylation process requires a catalyst to produce higher yield/selectivity of liquid hydrocarbons. Catalytic decarboxylation of lipid-based feedstocks in water, that is, hydrothermal processing, is a promising pathway for producing liquid hydrocarbons.[6,8,11] Water has several unique properties at its subcritical conditions (200–375 °C, 5–20 MPa), including a lower dielectric constant, which enhances solubility of the lipid-based feedstocks.[12] The relatively large ionization constant of subcritical water makes a highly reactive reaction media by minimizing mass-transfer limitations.[12] Water is also a good solvent for lipid feedstocks containing a large moisture content (algal biomass, sewage sludge, or fish and animal fat processing residues).[13−15]

Most catalytic processes for decarboxylation have used noble metal catalysts and high-pressure and external sources of hydrogen or carbon dioxide to enhance the feedstock conversion and liquid product yield/selectivity.[4,6,16−23] Yang et al.[19] studied decarboxylation of oleic acid (OA) using Pt/zeolite 5A and Pt/ZIF-67/zeolite 5A catalysts in the presence of high-pressure (20 bar) hydrogen to obtain 98.7% conversion of OA with 72.6 and 81.5% selectivity of heptadecane, respectively. In the presence of a CO2 (20 bar) atmosphere, decarboxylation of palmitic and lauric acids using this catalyst system gave 95% conversion of both fatty acids with 91.7 and 93.5% selectivity of pentadecane and undecane, respectively.[18] Mäki-Arvela et al.[21] conducted tall oil fatty acid deoxygenation using a Pd/C catalyst with 1% H2 balanced with argon (17 bar) to enhance the selectivity and conversion of heptadecane. The main shortcoming of noble metal catalysts is their limited availability and high cost. Fu et al.[8] and Hossain et al.[5,24] showed that hydrothermal decarboxylation did not require external hydrogen, with the required hydrogen to saturate the product being produced in situ by the water–gas shift reaction. Water is well known to participate as a coreactant with CO in the water–gas shift reaction under hydrothermal conditions, with CO being produced via decarbonylation or thermal cracking of the fatty acids.[5,24]

On the other hand, sulfided catalysts (e.g., sulfided CoMo or NiMo oxides) are also used for decarboxylation,[25,26] which are comparatively cheaper than noble metal catalysts, but leaching of sulfur may deactivate the catalytic activity and contaminate the products, requiring an additional step for sulfur recovery.[27,28] Besides the noble or sulfided catalysts, researchers reported non-sulfided catalysts for decarboxylation. Morgan et al.[29] reported the production of C5 to C17 hydrocarbons from triolein and soybean oil using a nickel-supported catalyst. Wu et al.[30] obtained over 80% selectivity of heptadecane using a nickel-supported catalyst during stearic acid decarboxylation. Na et al.[31] reported more than 98% conversion and less than 1 wt % of oxygen content in the product mixture during decarboxylation of OA using a supported MgO catalyst. Content of hydrocarbon (81.1%) and a low moisture content of biofuel from carinata oil were reported by Zhao et al.[32] using an Mo–Zn/Al2O3 catalyst. Most of these reported studies were conducted in batch, whereas the reaction chemistry in continuous operation (important for commercialization) is not well-known. Moreover, Hengst et al.[33] reported that acidic catalysts are preferable for enhancing the decarboxylation chemistry. However, they used noble metal (Pd) on different supports. In this study, we introduced Mo as a low-cost acidic catalyst for decarboxylation for the first time. For comparison purposes, Ni and MgO were used as reference catalysts.

The scope of this work is toward the continuous hydrothermal decarboxylation of OA as a model compound, while also examining other fatty acid derivatives including castor oil, frying oil, and palm oil. A current challenge for the continuous decarboxylation process is to develop low-cost stable catalysts. Moreover, the effect on the catalyst due to the exposure of the harsh sub or near supercritical water is poorly reported.

Here, we synthesize molybdenum (Mo)-, MgO-, and Ni-loaded alumina (γ-Al2O3) catalysts and investigate them for OA decarboxylation. The best catalyst was chosen among these three based on the degree of decarboxylation of OA. With the best catalyst determined, the effects of process parameters including temperature, space time, and water-to-OA (v/v) ratio were studied in hydrothermal media (subcritical water) using a continuous fixed-bed catalytic reactor. The catalytic activities for decarboxylation of several fatty acid derivatives including castor oil, frying oil, palm oil were then investigated. Characterization of fresh and spent catalysts was conducted using several physicochemical techniques including N2-physisorption [Brunauer–Emmett–Teller (BET) surface area and Barrett–Joyner–Halenda (BJH) pore size distribution], NH3 temperature-programmed desorption (NH3-TPD), X-ray diffraction (XRD), H2 temperature-programmed reduction (H2-TPR), CO pulse chemisorption, X-ray photoelectron spectroscopy (XPS), transmission electron microscopy (TEM), scanning electron microscopy (SEM), and energy dispersive X-ray spectroscopy (EDXS).

Results and Discussion

Characterization of Fresh Catalysts

Mo, MgO, and Ni catalysts are extensively used for decarboxylation of lipid-based feedstocks[30−32,34−38] and of interest for this study. The first part of this study investigated the production of three metals on an alumina support using an incipient impregnation method, with a target of 10 wt % metal on alumina. EDXS analysis was performed to quantify the actual metal loading for catalyst synthesis, with the results shown in Table and Figure S1. The actual loading of Mo and Ni was found to be 10.1 and 10.0 wt %, respectively. The wt % of Mo (9.10%) was also confirmed by XPS analysis. Because EDXS cannot determine the oxidation state of any metal, the actual loading of MgO was calculated from the weight difference between the catalyst sample after calcination and before loading, that is, 9.76 wt %. However, no elemental S was detected for the 10 wt % MgO–Al2O3 catalyst during EDXS analysis.

Table 1

Elemental Compositions of Fresh 10 wt % Mo–Al2O3 and 10 wt % Ni–Al2O3 Catalysts

catalysts	Al (wt %)	O (wt %)	Mo (wt %)	Ni (wt %)
10 wt % Mo–Al2O3	24.7	65.2	10.1
10 wt % Ni–Al2O3	25.4	64.6		10

Textural properties of any catalyst are important parameters to measure its catalytic activity. N2 adsorption–desorption isotherms and pore size distributions are displayed in Figure S2 and their corresponding pore properties are shown in Table . The isotherms of Al2O3 in Figure S2(i)a, fresh 10 wt % Mo–Al2O3 in Figure S2(i)b, 10 wt % MgO–Al2O3 in Figure S2(i)c, and 10 wt % Ni–Al2O3 in Figure S2(i)d showed a typical type IV isotherm with a H1-type hysteresis loop. This indicates that all fresh catalysts possessed a mesoporous structure.[39,40] Mesoporosity of the prepared catalysts was confirmed by pore size distributions in Figure S2(ii). As presented in Table , the BET surface area and pore volume of all fresh catalysts are lower than those of the support, indicating that Mo, MgO, and Ni loadings partially blocked Al2O3 pores during the catalyst preparation step (impregnation method). BET surface area and pore volume of the three synthesized catalysts are slightly different from each other, which follows the order: 10 wt % MgOAl2O3 > 10 wt % Mo–Al2O3 > 10 wt % Ni–Al2O3. The average pore diameters of all fresh catalysts are analogous to each other and slightly lower than that of the Al2O3 support.

Table 2

Summary of BET Surface Area, Pore Volume, and Pore Size of Fresh and Spent Catalysts

sample name	fresh/spent	BET surface area (m2/g)	pore volume (cm3/g)	average pore size (nm)
Al2O3	fresh	179	0.50	11.1
10 wt % Mo–Al2O3	fresh	160	0.44	9.6
	spent	78	0.23	10.2
10 wt % MgO–Al2O3	fresh	163	0.45	10.5
	spent	99	0.22	18.4
10 wt % Ni–Al2O3	fresh	158	0.45	10.5
	spent	111	0.28	10.0

Decarboxylation of free fatty acids is known to largely depend on the acidity of the catalyst. For example, Hengst et al.[33] found enhanced deoxygenation activity of free fatty acids for diesel fuel production using an acidic catalyst (Pd/Pural SB1-derived Al2O3). The NH3 TPD of fresh catalysts is shown in Figure to determine the acidity of the investigated catalysts. The peak at around 100–350 °C is assigned to the Lewis acid sites and the peak at 500 °C represents the Bronsted acid sites, respectively, in the TPD profiles.[41−43] The NH3 TPD profiles show that all catalysts contain mostly Lewis acid sites. The total amounts of ammonia desorbed by Mo–Al2O3, Ni–Al2O3, and MgO–Al2O3 are 33.5, 25, and 24.3 mL STP/g. The value of NH3 uptake indicates that the overall acidity of the Mo–Al2O3 catalyst is comparatively higher than that of the Ni–Al2O3 and MgO–Al2O3 catalysts.

Figure 1

NH3 TPD of fresh catalysts.

Crystallinity of a metal-supported catalyst is usually measured by XRD. Figure (i) shows the XRD patterns of all fresh catalysts with support. Pure γ-Al2O3 phases at 2θ = 35.2, 47.2, and 67.6° are found in Figure (i)a.[38] Mostly, MoO2 phases were found in the XRD pattern of fresh 10 wt % Mo–Al2O3 catalyst [Figure (i)b], which implies that Mo is a hard to reduce metal, even using a reduction temperature of 950 °C. Very weak reflections of pure Mo are observed at 2θ = 42 and 61°. Fresh 10 wt % MgO–Al2O3 and 10 wt % Ni–Al2O3 catalysts show mainly MgAl2O4 and NiAl2O4 phases, which indicates that all MgO and Ni particles react with Al2O3 to form MgAl2O4 and NiAl2O4, respectively [Figure (i)d,f]. Spinel MgAl2O4 phase was found at 2θ = 37.1, 45.3, and 66.1°. Spinel NiAl2O4 phase was found at 2θ = 37, 45, and 66°. No XRD peaks of MgO and Ni were found for the fresh catalysts, which may be due to the overlapping by reflections of MgAl2O4 and NiAl2O4 peaks or enhanced MgO or Ni dispersion on the catalyst surface.[44,45] Higher temperature for reduction or calcination enhances the formation of MgAl2O4 and NiAl2O4.[46] No MgSO4 or S peaks were found in fresh 10 wt % MgO–Al2O3 catalyst as MgSO4·7H2O was used as the MgO precursor. S usually shows sharp peaks between 15 and 60°.[47] The sharp peak intensity of fresh 10 wt % Ni–Al2O3 catalyst compared to fresh 10 wt % MgO–Al2O3 and 10 wt % Ni–Al2O3 catalysts indicates the smaller metal particle phases present on the surface. Moreover, no reflections for θ-Al2O3 (25.6 and 43.3°) or α-Al2O3 (31.2 and 36.6°) were evident on the reduced catalysts. This indicates that phase changes of Al2O3 were not enhanced by the chosen reduction/calcination temperature.

Figure 2

(i) XRD patterns of fresh and spent catalysts and (ii) TPR profiles of fresh catalysts.

H2-TPR experiments were conducted to determine the optimum reduction temperature of the investigated catalysts. H2-TPR profiles of fresh catalysts are shown in Figure (ii). Four reduction peaks are observed in the TPR profile of fresh 10 wt % Mo–Al2O3 at 475, 666, 925, and 1065 °C, respectively. It has been reported that Mo species reduction is a two-step process such as MoO3 to MoO2 and then MoO2 to Mo.[48] Different reduction temperatures obtained in the TPR profile indicate the formation of different Mo species. It was previously reported that four types of Mo phases were formed during the reduction of Mo-based catalyst. Two forms of Mo species were reported by Ma et al.[49] One was polynuclear (located on the external surface of the support), either in a square-pyramidal coordination as MoO form or in an octahedral coordination as MoO3 crystallite form. The second one was related to Al atoms in the lattice channel of support. The reduction peak at 475 °C of fresh 10 wt % Mo–Al2O3 is ascribed to the reduction of MoO3 to MoO2.[50] The strong peak assigned to 1065 °C indicates further reduction of MoO2 to form metallic Mo.[51] The reduction peaks at 666 and 925 °C may be ascribed to the initial and the further reduction of the aggregative MoO3 species.[48]

Fresh 10 wt % MgO–Al2O3 catalyst shows its only reduction peak at 475 °C, which represents the reduction of MgAl2O4. Usually two major reduction peaks at 250–350 and 600 −850 °C are observed for Ni–Al2O3 catalyst, corresponding to the easily reducible NiO and hard to reduce NiO, respectively.[52] Because there is no peak observed around 250–350 °C in this study, all NiO reacts with Al2O3 to form NiAl2O4. The reduction peak of NiAl2O4 was observed at 750 °C for the fresh 10 wt % Ni–Al2O3. Reducibility of nickel catalysts largely depends on the wt % of nickel, calcination or reduction temperature, and the interaction between the metal and support.[53] Calcination or reduction at high temperature increases the metal–support interaction and the formation of NiAl2O4, which results in greater difficulty of nickel catalysts in reduction.

The metal–support interaction and the ease of reduction are determined from both the peak width and the reduction temperature. The wide peak represents a strong metal–support interaction and the higher temperature for reduction indicates a hard to reduce metal. It is observed from the TPR-H2 profile that both fresh 10 wt % Mo–Al2O3 and 10 wt % Ni–Al2O3 were reduced at higher temperature compared to fresh 10 wt % MgO–Al2O3 catalyst. Comparing the fresh Mo- and Ni-based catalysts, the Mo catalyst was reduced at higher temperatures. However, the TPR-H2 peaks for the MgO catalyst are comparatively narrower than those for the other two catalysts, indicating a weak interaction of MgO with Al2O3. The H2-TPR results are consistent with the XRD results.

For a more detailed investigation of the surface structure, XPS spectra of the fresh 10 wt % Mo–Al2O3 are presented in Figure . The spectra were curve-fit using the screened and unscreened peak-fitting parameters [peak positions, full width at half-maximum (fwhm), and area ratios] for Mo(IV) as outlined by Scanlon et al.[54] The parameters obtained for fresh 10 wt % Mo–Al2O3 catalyst are provided in Table S1. An Mo 3d5/2–Mo 3d3/2 spin–orbit double for Mo(VI) was constrained to have an area ratio of 3:2, equal fwhms, a doublet spacing of 3.13 eV, and an Mo 3d5/2 peak position ranging from 232.2 to 232.6 eV. Peak fitting showed that a small component attributable to Mo(V) was also present. As such, a spin–orbit doublet, constrained to have a similar fwhm as that for the Mo(VI) component, was added with an Mo 3d5/2 peak position ranging from 231.6 to 231.9 eV.[55,56] It should be noted that the presence of the component attributed to Mo(V) in this fitting may be due to some portion of the Mo(IV) unscreened peak structure not accounted for by Scanlon et al.[54] fitting, possibly due to spectrometer differences. Mo(V) may also be present because of the X-ray reduction of MoO3 during the analysis.[57]

Figure 3

XPS high-resolution spectra of Mo (3d) for 10 wt % Mo/Al2O3 catalyst (detail peak assignment shown in Tables S1 and S7).

Percent metal dispersion is an important parameter for a metal-supported catalyst, which enhances greatly the catalytic reaction. Pulse chemisorption experiments were conducted to quantify the % metal dispersion, the active metal surface area, and the active particle size of metal crystals on the alumina surface. Table shows the CO pulse chemisorption data for 10 wt % Mo–Al2O3 and 10 wt % Ni–Al2O3 catalysts. Metal dispersion (2.49%) was found for fresh 10 wt % Mo–Al2O3 catalyst, whereas the value is smaller for fresh10 wt % Ni–Al2O3 catalyst. Higher metallic surface area per gram of sample and per gram of metal were also obtained for fresh 10 wt % MgO–Al2O3 catalysts compared to fresh 10 wt % Ni–Al2O3. Smaller active metal particle size was observed in fresh 10 wt % Mo–Al2O3 catalyst. The obtained larger metal particle size from the Ni-based catalyst is attributed to the lower dispersion and metallic surface area.

Table 3

CO Pulse Chemisorption Data of Fresh and Spent Catalysts

sample	fresh/spent	% metal dispersion	metallic surface area (m2/g sample)	metallic surface area (m2/g metal)	active particle diameter (nm)
10 wt % Mo–Al2O3	fresh	2.49	1.34	9.81	10.1
	spent	1.68	0.74	6.11	90.7
10 wt % Ni–Al2O3	fresh	1.94	0.81	6.74	95.8
	spent	0.45	0.32	3.05	200.3

The morphology of the fresh catalysts was observed using TEM. The 10 wt % Mo–Al2O3 catalyst shows small metal particles distributed over the surface, whereas the 10 wt % Ni–Al2O3 and 10 wt % MgO–Al2O3 catalysts have larger metal particles on their surfaces (Figure ). Small Mo particles improve the % metal dispersion and reduce the active particle diameter, whereas the larger metal particles of Ni and MgO catalysts reduce the % metal dispersion. Smaller particle size provides more active sites for the decarboxylation reaction compared to larger particle size. The TEM results are consistent with the CO pulse chemisorption results.

Figure 4

TEM images of fresh catalysts: (a) 10 wt % Mo–Al2O3; (b) 10 wt % Ni–Al2O3; and (c) 10 wt % MgO–Al2O3.

Decarboxylation of OA

Screening of Decarboxylation Catalyst

The initial experiments compared the catalytic activities for OA decarboxylation in a continuous flow reactor using 10 wt % loading of Ni, Mo, and MgO on γ-Al2O3 at 375 °C, space time of 4 h, and ratio (v/v) of water-to-OA = 5:1 using 5 g of catalyst. Figure depicts the attenuated total reflection–Fourier transform infrared (ATR–FTIR) spectra of the decarboxylated liquid products with the degree of decarboxylation being calculated by eq . The results show that the degree of decarboxylation of OA varied strongly using the investigated catalysts, being 67, 65, and 92% using a 10 wt % MgO–Al2O3, 10 wt % Ni–Al2O3, and 10 wt % Mo–Al2O3 catalyst, respectively. The obtained liquid and gaseous product yields using these three catalysts are liquid yield = 65, 30, and 71% and gaseous yields are 35, 70, and 29%, for the 10 wt % MgO–Al2O3, 10 wt % Ni–Al2O3, and 10 wt % Mo–Al2O3, respectively. The low liquid and high gaseous product yields indicate that the 10 wt % Ni–Al2O3 catalyst is primarily a gasification catalyst, consistent with our previous work on supercritical water gasification.[58−61]

Figure 5

ATR–FTIR spectra of (a) OA and the formed products using different catalysts such as (b) 10 wt % MgO–Al2O3, (c) 10 wt % Ni–Al2O3, and (d) 10 wt % Mo–Al2O3.

Although all catalysts have similar BET surface area, higher acidity and metal dispersion of 10 wt % Mo–Al2O3 catalysts are favored for a higher degree of decarboxylation (92%). Higher metal dispersion, metallic surface area, and smaller metal particle sizes of the 10 wt % Mo–Al2O3 catalyst provide more active sites for decarboxylation of fatty acids. Hengst et al.[33] showed that an acidic catalyst enhanced the deoxygenation and cracking of free fatty acids into diesel fuels.

Because the molar mass of Mo (96 g) is higher than those of Ni (58.7 g) and MgO (40.3 g), the molar percentages of Ni and MgO in 10 wt % Ni–Al2O3 and 10 wt % MgO–Al2O3 catalysts are much higher than that of Mo in the 10 wt % Mo–Al2O3 catalyst at the same loading wt %. Hence, the catalytic performance of the 10 wt % Mo–Al2O3 catalyst is much better than the other two.

The gaseous products CO and CO2 found in the product streams are compared in Figure S3. The Ni–Al2O3 and MgO–Al2O3 catalysts gave a higher yield of CO compared to CO2, whereas the Mo–Al2O3 catalyst provided a high yield of CO2 compared to CO. This indicates that the Mo-based catalyst dominates the decarboxylation of OA under the chosen reaction conditions at which the best catalyst was determined for this study. Significant amounts of H2 and CH4 and lighter fractions of hydrocarbons (C2 to C4) were found using the Ni–Al2O3 catalyst (data not shown). On the basis of these initial ATR–FTIR and gas chromatography–thermal conductivity detection (GC-TCD) results, the 10 wt % Mo/Al2O3 catalyst was chosen for the subsequent parametric study examining the experimental conditions.

Table S2 shows carbon balance using 10 wt % MgO/Al2O3, 10 wt % Ni/Al2O3, and 10 wt % Mo/Al2O3 catalysts for the decarboxylation of OA at 375 °C, space time of 4 h, and water-to-OA (v/v) ratio of 5:1 using 5 g of catalyst. The carbon balance in all cases is found to be 97 to 99%.

Effect of Reaction Parameters on Degree of Decarboxylation

Reaction parameters including temperature, water-to-OA (v/v) ratio, and space time (τ) are the three most important parameters for a continuous hydrothermal decarboxylation reaction. ATR–FTIR spectra of the decarboxylated liquid products obtained under different experimental conditions (T = 325 to 400 °C, ratio (v/v) of water to OA = 2:1 to 5:1 & space time = 0.5 to 4 h) using 5 g of 10 wt % Mo–Al2O3 catalyst are presented in Figure , whereas Figure S4 shows the corresponding degree of decarboxylation. The results show that increasing temperature from 325 to 375 °C enhanced the percent decarboxylation (from 55 to 92%) (Figure S4a). Further increasing of temperature from 375 to 400 °C improved the degree of decarboxylation by only 0.3%. The results show that the degree of decarboxylation of OA is less sensitive to temperature above 375 °C. Therefore, the maximized temperature chosen for this study was 375 °C.

Figure 6

ATR–FTIR spectra of (a) OA and the products formed using water-to-OA (v/v) ratio of 5:1 and 4 h of reaction time at (b) 325, (c) 350, (d) 375, and (e) 400 °C.

The effects of process parameters of water-to-OA ratio (v/v) and space time at 375 °C on the formed decarboxylated products are presented in Figure . Figure S4b,c shows their corresponding degree of decarboxylation. The results indicate that increasing the ratio (v/v) of water-to-OA from 2:1 to 5:1 and space time from 0.5 to 4 h enhanced the percent decarboxylation. The maximum degree of decarboxylation (92%) was obtained at 375 °C, ratio (v/v) of water-to-OA = 5:1, and space time = 4 h. On the basis of these results, the maximized process conditions for maximum removal of −COOH peak using the 10 wt % Mo–Al2O3 catalyst for the current study were 375 °C, ratio of water-to-OA (v/v) = 5:1, and space time = 4 h, respectively.

Figure 7

ATR–FTIR spectra of (a) OA and the products formed at 375 °C for different reaction times using different ratios of water-to-OA (v/v): (b) 0.5 h and ratio of 5:1; (c) 1 h and ratio of 5:1; (d) 2 h and ratio of 5:1; (e) 4 h and ratio of 5:1; (f) 4 h and ratio of 4:1; (g) 4 h and ratio of 3:1; and (h) 4 h and ratio of 2:1.

The peak for alkenyl =CH stretching at 3004 cm–1 (Figures and 7) disappeared for all products obtained using catalyst. This indicates that the 10 wt % Mo–Al2O3 catalyst is not only a decarboxylation but also a hydrogenation catalyst (C=C to C–C). There are some other peaks such as C–O stretching and O–H deformation (combined) at 1412 cm–1 and out-of-plane O–H bending at 934 cm–1, which also significantly decreased in the liquid product. This corroborates that the Mo–Al2O3 catalyst enhances the decarboxylation of OA.

Gas chromatography–Flame Ionization Detection Analysis of Liquid Products

Use of liquid products as commercial fuel largely depends on an understanding of controlling the molecular fingerprint. Figure displays the hydrocarbons identified in the liquid decarboxylated products at the studied experimental conditions using 5 g of 10 wt % Mo–Al2O3 catalyst. Selectivity of hydrocarbon varies with temperature, as presented in Figure a. The selectivity of tetradecane increased from 0 to 34.8% and pentadecane selectivity decreased from 49.3 to 24% with increasing temperature from 325 to 375 °C. At the same time, the selectivity of hexadecane decreased from 39 to 22.9% and the selectivity of hetadecane slightly increased from 10.7 to 18.3%. Lower heptadecane and higher tetradecane selectivities at 375 °C indicate that the Mo-based catalyst cracks OA into smaller hydrocarbons.

Figure 8

Hydrocarbons present in the liquid products at different (a) temperatures, (b) water-to-OA ratios (v/v), and (c) space times.

Selectivities of hydrocarbon compounds vary with the space time and water-to-OA (v/v) ratio. Figure b,c shows the effect of water-to-OA ratio and space time on the distribution of hydrocarbon compounds in the liquid decarboxylated products. The selectivity of tetradecane increased (from 34.8 to 77%) with decreasing water-to-OA ratio (v/v) from 5:1 to 2:1 and decreased (from 34.8 to 14.5) with decreasing space time from 4 to 0.5 h. The selectivity of pentadecane increased (from 7.2 to 24% and 9.2 to 24%) with increasing ratio (v/v) from 2:1 to 5:1 and space time from 0.5 to 4 h, respectively. The selectivity of hexadecane slightly increased with increasing ratio (v/v) of water-to-OA and space time. The selectivity of heptadecane increased with increasing ratio and decreased with increasing space time. The results conclude that higher temperature and lower water-to-OA ratio (v/v) and space time favor higher selectivity of hepatadecane (73.6%), whereas higher temperature, water-to-OA ratio (v/v), and space time provide higher tetradecane selectivity (34.8%). Lower selectivity of heptadecene for all catalytic experiments indicates a higher degree of saturation of C=C to C–C. This result is consistent with the ATR–FTIR results.

Mass balances for all experiments in this study were found to be in the range of 97 to 99%. Figure S5 shows the mass balance of the experiment conducted at the maximized reaction conditions. Comparing the conversion and liquid yield or selectivity of the current study and the literature data (Table ) shows that the decarboxylation study using different reaction conditions such as mode of operation, operating conditions, and catalyst provides various results. Most of the studies reported in the literature have been conducted under an H2 atmosphere to enhance the decarboxylation reaction. The maximum conversion (∼100%) was obtained using NiMoS2/γ-Al2O3, Ni/ZrO2, and Pd/SiO2 catalysts using hydrogen, whereas the current study showed 91% conversion without adding any hydrogen. Table S3 shows the selectivity of heptadecane between the current study and the literature data, which was obtained from OA conversion using different catalysts. The highest heptadecane selectivity reported is 89.3% from OA using activated carbon as the catalyst.[24] The objective of the current study was to find a cost effective metal-supported catalyst for decarboxylation of fatty acids and their derivatives with no added H2. Our study provided up to 91% conversion of OA (degree of decarboxylation) with a 71% liquid yield at the maximized reaction conditions using the Mo/Al2O3 catalyst.

Table 4

Comparison of Conversion and Liquid Yield between the Literature Data and the Current Studya

catalyst	feedstock	mode of operation	operating conditions	conversion (%)	overall liquid yield (%)	specific product yield or selectivity (%)
NiMoS2/γ-Al2O3[62]	refined palm kernel oil	continuous	330 °C, H2 atmosphere, 1 h–1 of LHSV	100	∼92	58 (selectivity of C10–C12)
NiMoS2/γ-Al2O3[62]	refined palm oil	continuous	330 °C, H2 atmosphere, 1 h–1 of LHSV	100	∼98	58 (selectivity of C10–C12)
NiMo/γ-Al2O3 and NiMoS2/γ-Al2O3[63]	crude palm oil	continuous	350 °C, H2 atmosphere, 2 h–1 of LHSV (pilot scale)	100	∼100	n/a
NiMoS2/γ-Al2O3[64]	refined palm kernel oil	continuous	330 °C, H2 atmosphere, 1 h–1 of LHSV	100	∼89	>95.5 (n-alkane yield)
NiMoS2/γ-Al2O3[37]	food grade rapeseed oil	continuous	280 °C, H2 atmosphere, 0.25–4 h of contact time (V/F)	80–100	>90	n/a
commercial NiMo/γ-Al2O3 and conventional in situ sulfidation by DDS[65]	waste cooking oil (mainly sunflower oil)	continuous	350–390 °C, H2 atmosphere, 0.5–2 h–1 of LHSV	>95	73–82	96.8–97.9 (diesel selectivity)
Pd/mesoporous C[21]	tall oil fatty acid	batch	350 °C, H2 atmosphere, 5.5 h of reaction time	59	n/a	91 (selectivity of heptadecane and heptadecene)
Pd/SiO2[66]	lauric acid	batch	300 °C, H2 atmosphere, 4 h of reaction time	100	n/a	96 (n-undecane yield)
Pd/Al2O3[66]	lauric acid	batch	300 °C, H2 atmosphere, 4 h of reaction time	100	n/a	94 (n-undecane yield)
Pd/C[67]	lauric acid	batch	300 °C, H2 atmosphere, 5 h of reaction time	65	58.4	91 (n-undecane selectivity)
Pt/ZIF-67/zeolite 5A[18]	lauric acid	batch	320 °C, 2 h of reaction time, CO2 atmosphere	95	n/a	93.5 (n-undecane selectivity)
Ni/HZSM[68]	methyl laurate	batch	280 °C, H2 atmosphere, 5 h of reaction time	69–86	n/a	27–68 (yield of C11 to C12)
Pt/ZIF-67/zeolite 5A[18]	palmitic acid	batch	320 °C, 2 h of reaction time, CO2 atmosphere	95	n/a	91.7 (pentadecane selectivity)
Ni/ZrO2[34]	palmitic acid	batch	300 °C, H2 atmosphere in the presence of H2O, 6 h of reaction time	88.2	66.8	30.2 (pentadecane yield)
Ni/ZrO2[34]	palmitic acid	batch	300 °C, H2 atmosphere, 6 h of reaction time	88	61	30.2 (pentadecane yield)
Pd/CNT[69]	palmitic acid	batch	260 °C, H2 atmosphere, 4 h of reaction time	93.3	n/a	85.4 (pentadecane selectivity)
MoO2/CNT[69]	palmitic acid	batch	260 °C, H2 atmosphere, 4 h of reaction time	100	n/a	15.4 (pentadecane selectivity)
Pt/C[8]	palmitic acid	batch	290 °C, hydrothermal conditions, 6 h of reaction time	90	90	98 (pentadecane selectivity)
Pd/C[8]	palmitic acid	batch	370 °C, hydrothermal conditions, 3 h of reaction time	n/a	n/a	63 ± 5 (pentadecane yield)
AC[7]	palmitic acid	batch	370 °C, hydrothermal conditions, 3 h of reaction time	33 ± 13	n/a	58 ± 4 (pentadecane selectivity)
NiMCF(9.2T-3D) (R)[70]	palmitic acid	batch	300 °C and 6 h of reaction time	86.4	n/a	31.8 (pentadecane yield)
AC[7]	OA	batch	370 °C, hydrothermal conditions, 3 h of reaction time	80 ± 4	n/a	7 ± 1 (heptadecane selectivity)
Pt/zeollite 5A[19]	OA	batch	320 °C, 2 h of reaction time, H2 atmosphere	98.7	∼100	72.6 ± 2 (heptadecane selectivity)
Pt/ZIF-67/zeollite 5A[19]	OA	batch	320 °C, 2 h of reaction time, H2 atmosphere	98.7	∼100	81.5 ± 3 (heptadecane selectivity)
Pt/ZIF-67/zeollite 5A[20]	OA	batch	320 °C, 2 h of reaction time, CO2 atmosphere	100	∼100	90.5 ± 1.3 (heptadecane yield)
Pt-Ga-MOF[71]	OA	batch	320 °C, 2 h of reaction time, H2 atmosphere	92	∼84	21.5 (heptadecane selectivity)
Ga-MOF[71]	OA	batch	320 °C, 2 h of reaction time, H2 atmosphere	66	∼72.4	5.7 (heptadecane selectivity)
Pt/SAPO[72]	OA	batch	325 °C, 2 h of reaction time, H2 atmosphere	98	n/a	32 (heptadecane yield)
Pt-SAPO-34[73]	OA	batch	325 °C, 2 h of reaction time, H2 atmosphere	98	91	30 (heptadecane selectivity)
Mo/γ-Al2O3 (this study)	OA	continuous	375 °C, hydrothermal conditions, 4 h of space time	91	71	18.3 (heptadecane selectivity)

n/a: data is not available in the cited references.

Comparison between the current study and the literature data in terms of selectivity of heptadecane from the catalytic conversion of OA (see Table S3) indicates that the selectivity of heptadecane varies with the process parameters and types of catalysts used. Our study showed 73.6% selectivity for heptadecane using a water-to-oil ratio of 5:1 (v/v) and 0.5 h of space time at 375 °C, with a degree of decarboxylation of 58.6%. On the other hand, the selectivity of heptadecane was 18.3% at 4 h of space time with a degree of decarboxylation of 91%. The results indicate that the Mo catalyst is cracking higher-length hydrocarbon chains at longer space times, increasing the degree of decarboxylation. Although the liquid product in this study contains a broad distribution of hydrocarbons, the liquid product falls within the range of commercial fuels such as kerosene, diesel, and jet fuels (discussed in the following section).

Fuel Quality

Density is an important physical characteristic of liquid fuel, which determines whether it can be used as diesel, kerosene, or jet fuel. For liquids, temperature is an important factor that can affect the density of a liquid with density being expressed at a given temperature for comparison purposes. The values of density at different temperatures of the product obtained at maximized conditions are compared with commercial fuels and listed in Table . Comparing the experimental data with conventional fuels indicates that the decarboxylated product falls within the typical diesel range.

Table 5

Density of the Decarboxylated Product and Some Commercial Fuels

compounds	temperature (°C)	density (kg/m3)
decarboxylated producta	15.6	0.85
	21.6	0.85
	25	0.85
	40	0.84
kerosene[74]	15.6	0.78–0.82
jet fuel[75]	15	0.78–0.84
diesel[74]	15.6	0.80–0.96

Maximized conditions.

High heating value (HHV) is the most important parameter for any fuel, which determines the tendency to produce more energy in the engine. Fuels with lower HHV tend to burn inadequately, initiating air pollution.[76]Table shows the HHVs of the obtained decarboxylated products with some commercial fuels. The results indicate that the HHV of the decarboxylated product is similar to jet fuel, kerosene, and diesel.

Table 6

HHVs of Feed, Product, and Commercial Fuels

compounds	HHVs (MJ/kg)
OA	39.2
decarboxylated producta	44.7
jet fuel[77]	43.5
kerosene[74]	46.2
diesel[74]	44.8

Maximized conditions.

GC-TCD Analysis of Gaseous Products

Decarboxylation of OA was confirmed by analyzing the gaseous products using GC-TCD. The results obtained at different reaction temperatures are presented in Figure S6. The number of moles of CO and CO2 were decreased and increased, respectively, with increasing reaction temperature from 325 to 375 °C. This result shows that decarboxylation is the dominating reaction at 375 °C, although 10 wt % Mo–Al2O3 catalyst slightly cracks the reactant molecules into smaller hydrocarbons. The GC-TCD results help confirm the ATR–FTIR results. Lighter fraction of hydrocarbons was found in the gaseous products, accounting for about 1.75 wt % of the total products (Figure S5).

Decarboxylation of Fatty Acid Derivatives

Hydrothermal decarboxylation of three nonedible feedstocks including castor oil, palm oil, and frying oil (compositions will be found in Tables S4–S6) was performed at the maximized conditions achieved for OA decarboxylation. The maximized conditions are: T = 375 °C, water-to-oil ratio = 5:1(v/v), space time = 4 h, and amount of catalyst = 5 g (10 wt % Mo–Al2O3). The ATR–FTIR results of the formed products from castor oil, palm oil, and frying oil are shown in Figure .

Figure 9

ATR–FTIR spectra of (a) castor oil, (b) the formed product from castor oil, (c) palm oil, (d) the formed product from palm oil, (e) frying oil, and (f) the formed product from frying oil.

In the spectrum of castor oil (Figure a), there are several absorbance peaks at 3388, 3007, 2923, 2853, 1742, 1653, 1456, 1240, 1161, 1032, and 723 cm–1 resulting from hydroxyl, alkenyl, methylene, and carboxylic ester groups of ricinoleic acid contained in the castor oil. Alcoholic O–H and C–O stretching peaks are observed at 3388 and 1032 cm–1, whereas the peaks at 3007 and 1653 cm–1 are ascribed to alkenyl =CH & C=C stretching, respectively. The peaks at 2923, 2853, and 723 cm–1 are attributed to methylene C–H asymmetric stretching, symmetric stretching, and rocking modes, respectively. The peak at 1456 cm–1 is assigned to methylene C–H scissoring and methyl C–H asymmetric bending modes. The peaks at 1742, 1240, and 1161 cm–1 are attributable to carboxylic ester C=O stretching and C–O stretching modes of the glycerol ester, respectively. After the decarboxylation reaction, C=O, C–O, O–H, C=C, and =C–H peaks disappear, whereas a few new peaks appear with the peak at 1711 cm–1 being ascribed to the carboxylic C=O stretching mode (Figure b). These results confirm the conversion of carboxylic ester (C(=O)–O), alcoholic OH, and alkenyl (C=C) groups and formation of small amount of carboxylic acid (−COOH).

In comparison to the spectrum of castor oil (Figure a), the spectrum of palm oil (Figure c) shows an additional absorbance peak at 1708 cm–1, which is ascribed to the C=O stretching of the carboxylic acid group contained in palm oil. Similar to that seen for castor oil (Figure b), after the decarboxylation reaction, C=O, C–O, C=C, and =C–H peaks disappear, whereas the peak at 1708 cm–1 decreases significantly. These results confirm the conversion of carboxylic ester (C(=O)–O), carboxylic acid (−COOH), and alkenyl (C=C) groups.

The spectrum of frying oil is displayed in Figure e. Similar to that of castor oil (Figure b), after the decarboxylation reaction, C=O, C–O, C=C, and =C–H peaks disappear, whereas a few new peaks appear with the peak at 1711 cm–1 being ascribed to carboxylic C=O stretching mode (Figure f). These results confirm the conversion of carboxylic ester (C(=O)–O), alcoholic OH, and alkenyl (C=C) groups and formation of small amount of carboxylic acid (COOH).

Decarboxylation of castor, palm, and frying oil was further confirmed by GC-TCD analysis of gaseous products formed during the reaction. Figure S7 shows the number of moles of CO or CO2 present in the gaseous products. The amounts of CO2 were found to be 0.88, 0.86, and 0.77 mol in the gaseous products formed during the decarboxylation of castor, frying, and palm oil. On the other hand, the quantities of CO were found to be 0.05, 0.13, and 0.23 mol, respectively. Although complete decarboxylation was not achieved for the above three nonedible feedstocks, our hydrothermal decarboxylation process shows significant feedstock feasibility to decarboxylate any fatty acid or its derivatives without adding any external hydrogen source or hydrogen donor solvent. These results are potentially attractive to implement the process for commercial production of liquid hydrocarbons from various feed sources, which can help lower our dependency on fossil fuels.

Catalyst Deactivation Studies

A metal-supported catalyst has a tendency for deactivation when it is exposed to harsh (high temperature or pressure) hydrothermal environments. Deactivation of a metal catalyst can occur from several factors including adsorption of impurities from the feed/product streams, coke deposition on the catalyst surface, oxidation of metal, metallic surface area reduction from sintering/leaching, and a drop in surface area from pore blockage.[78] XRD analysis was performed on all three spent catalysts (Figure ). Metal-supported spent catalysts usually show the peaks for graphitic coke formation at 2θ = 62° and atomic coke formation at 2θ = 30°, respectively.[5,58] From the XRD pattern, no such peak was found for all three catalysts (Figure c,e,g). There were no significant differences observed in the XRD patterns for fresh and spent 10 wt % MgO/Al2O3 and 10 wt % Ni/Al2O3 catalysts. On the other hand, peak intensities for the spent 10 wt % Mo/Al2O3 catalyst were found to be larger than those of its fresh one, indicating catalyst agglomeration from the decarboxylation reaction. Agglomeration of Mo particles may be the main reason for the observed reduction of the BET surface area of spent Mo catalyst compared to the spent MgO and Ni catalysts (Table and Figure S2i). The lower BET surface area observed for the spent Mo catalyst may be due to pore blockage by produced hydrocarbon molecules which did not wash out with hexanes or catalyst drying step. Although all spent catalysts have lower BET surface areas than their fresh ones, they still maintain the mesoporous pore size distributions (Figure S2ii).

The surface morphology of the synthesized catalysts were examined by SEM imaging both before and after the decarboxylation reaction (Figure S8). All fresh catalysts showed uniform metal/metal oxide particle distribution on the surface of the catalysts, indicating better catalytic properties. Although the Ni catalyst has a morphology similar to that of the Mo and MgO catalyst, the Ni catalyst was found to gasify the feedstock at the chosen reaction conditions instead of decarboxylation (explained earlier). The agglomerated structure of the spent 10 wt % Mo–Al2O3 catalyst was confirmed from the SEM images. There were no morphological differences observed for the fresh and spent 10 wt % MgO–Al2O3 and 10 wt % Ni–Al2O3 catalysts. These results are consistent with the XRD and BET results.

Elemental analysis was performed for all three spent catalysts to see the composition differences between their fresh and spent states (Figure S9). The weight percentages of Mo and Ni were found to be 9.58 and 9.23 in the spent catalysts. Slightly lower percentage of Mo and Ni in the spent catalyst does not indicate metal leaching during the decarboxylation reaction. Deposition of impurities (Figure S10) on the catalyst surface may dilute the concentration of Mo and Ni. As mentioned earlier, C found on the fresh catalysts originates from the sample holder. However, the peak intensities of C in all three spent catalysts are significantly larger those of the fresh ones, indicating that the catalysts have carbon deposition from the decarboxylation reaction. Because XRD did not detect any crystalline carbon (atomic or coke), the carbon identified by EDXS analysis is likely amorphous carbon. Amorphous carbon can be simply removed by calcining the catalyst above the reaction temperature under an inert atmosphere.

Thermogravimetry/differential thermal analysis (TG-DTA) was conducted to quantify the deposited carbon on the catalyst surfaces based on the weight loss with increasing temperature (Figure a–c). The weight loss and the corresponding DTA peaks in TG-DTA spectra represent burning of three types of carbon such as atomic carbon, amorphous carbon, and graphitic carbon. These three carbons show DTA peaks at three temperature regions such as <250, 250–600, and >600 °C, respectively.[79] No peak was observed for all spent catalysts at >600 °C, which confirms the absence of graphitic coke on the spent catalyst surfaces. The weight losses for spent Mo–Al2O3, Ni–Al2O3, and MgO–Al2O3 are 30, 15, and 18, respectively. These weight losses during TG-DTA correspond to only the formation of atomic and amorphous carbon. Higher amount of carbon deposition on spent Mo–Al2O3 catalyst is due to the higher acidity of Mo–Al2O3 compared to Ni–Al2O3 and MgO–Al2O3 catalysts. Higher acidity of catalyst is favorable for higher degree of carbon deposition.[80,81]

Figure 10

(a–c) TG-DTA of spent catalysts and (d) XPS survey spectra of fresh and spent Mo/Al2O3 catalyst.

XPS survey spectra of fresh and spent 10 wt % Mo–Al2O3 catalyst were obtained to provide insight into the Mo catalyst deactivation process (Figure d). Carbon in the fresh 10 wt % Mo–Al2O3 catalyst corresponds to adventitious carbon which is typically detected in samples that have been exposed to the atmosphere or generated during the analysis. The relative content of C in the fresh catalyst is 16.5%, whereas the amount in the spent catalyst was 43.6%. The difference between the quantities of carbon in fresh and spent catalysts is 27.1%, which is closer to the amount quantified by TG-DTA. Thus, large increase in C on the spent catalyst indicates the deposition from the decarboxylation reaction. This C potentially may be bound with multiple Mo ions, which is hard to remove by simple hexane washing or vacuum drying.[82] The existence of different states of Mo was also confirmed in the spent 10 wt % Mo–Al2O3 catalyst using high-resolution XPS spectra (Figure b and Table S7). A higher percentage of Mo(IV) (77%) was found in spent sample compared to the fresh one (40%), whereas the Mo(VI) component was decreased from 51 to 11% and the Mo(V) amount remained relatively constant at 9–12%, respectively. Higher percentage of Mo(VI) state compared to other oxidation states on the fresh catalyst surface may be the active catalytic phase of decarboxylation. The results also indicates that reduction of Mo (VI) to Mo(IV) was occurring by in situ hydrogen during decarboxylation of fatty acids.

Catalyst Stability, Reusability, and Regeneration

Catalyst stability is an important factor for commercializing any catalytic process. Formation of graphitic coke on the metal-supported catalyst surface mainly poisons the active sites of the catalyst and reduces activity during the reaction. Graphitic coke formation also prevents the reusability of any catalyst as it is very hard to burn off the graphitic carbon during regeneration step. Stability study of Mo–Al2O3 was conducted at the maximized reaction conditions until the catalyst starts to deactivate. ATR–FTIR spectra of decarboxylated liquid products at the maximized reaction conditions at 132 h (5.5 days) time on stream are shown in Figure . Deactivation of the catalyst was measured by the degree of decarboxylation obtained at different time of streams. Figure shows that the degree of decarboxylation (∼92%) was consistent over 72 h time on stream. This is the maximum degree of decarboxylation obtained for this study. The degree of decarboxylation started decreasing afterward. The values of degree of decarboxylation obtained after 96 and 132 h time on stream were 60 and 48%, respectively. Decreasing trend of degree of decarboxylation indicates that the catalyst is losing its activity during the reaction. EDXS, XPS, and TG-DTA of spent catalysts showed that a significant amount of carbon (atomic and amorphous) was deposited on the catalyst surface during decarboxylation reaction which is deactivating the catalyst. Deposited carbon is blocking the actives sites of catalysts, which is retarding the decarboxylation of OA. The disappearance of =C–H peak (∼3004 cm–1) during 132 h time on stream indicates that the catalyst was active to saturate the final products.

Figure 11

FTIR spectra of (a–e) the decarboxylated products at 132 h time on stream: (a) 24; (b) 48; (c) 72; (d) 96; (e) 132 h; and (f) OA.

Deactivation of the catalyst necessitates to regeneration step because the degree of decarboxylation obtained using spent catalyst was only 25% (Figure S11). Because only atomic and amorphous carbon deposition were observed, it is easy to remove these types of carbon by burning them off in the presence of air. Spent Mo–Al2O3 catalyst (from stability study) was regenerated by simply burning with air followed by reduction afterward and reused second time for decarboxylation of OA at the maximized reaction conditions. ATR–FTIR spectra of decarboxylated liquid products using regenerated Mo–Al2O3 catalyst are presented in Figure S11 along with the spectra of fresh and spent catalyst for comparison. The degree of decarboxylation obtained using regenerated Mo–Al2O3 catalyst was 70%, whereas the value was 92% using fresh catalyst. The regenerated catalyst was not able to achieve the same degree of decarboxylation which may be due to the agglomeration of metal particles after first use. Agglomeration of Mo particles in the spent catalyst was confirmed by SEM imaging and CO pulse chemisorption experiment (Figure S8 and Table ) measuring the percent metal dispersion and active particle size. The degree of decarboxylation using regenerated catalyst was 70% during 72 h time on stream and started decreasing afterward.

XRD patterns of regenerated and spent regenerated catalysts have the similar trend except the intensity difference (Figure S12a). The broad peaks of spent regenerated catalyst indicate that the metal particles are agglomerating to make bigger particle size during decarboxylation reaction. The formation of deposited carbon on the surface of spent regenerated catalyst is quantified by TG-DTA, which is about 12 wt % (Figure S12b). The formation of carbon on the catalyst surface during decarboxylation indicates that it is hard to prevent. The BET surface area and pore volume of regenerated catalyst are 102 m2/g and 0.30 cm3/g, whereas the values are 70 m2/g and 0.22 m3/g, respectively. N2 adsorption–desorption isotherms of regenerated and spent regenerated Mo–Al2O3 catalyst have the similar pattern as fresh and spent catalysts. BJH pore size distributions in both cases slightly shift to large pore diameter range but still maintain the mesoporosity (Figure S12c). The surface morphology of spent regenerated catalyst also indicates the agglomerated structure of the catalyst (Figure S12d).

Conclusions

This work investigated three different catalysts including 10 wt % Mo–Al2O3, 10 wt % MgO–Al2O3, and 10 wt % Ni–Al2O3 and showed that the Mo catalyst is an efficient catalyst for decarboxylation of OA and their derivatives in subcritical water. The maximized conditions for higher degree of decarboxylation of OA (92%) and liquid yield (71%) were found to be 375 °C, water-to-OA ratio of 5:1 (v/v), and space time of 4 h using 5 g of 10 wt % Mo–Al2O3 catalyst. The Mo catalyst was found to crack the OA into smaller hydrocarbon molecules. The selectivities of OA to hydrocarbons obtained at maximized reaction conditions were 34.8% tetradecane, 24% pentadecane, 22.9% hexadecane, and 18.3% heptadecane. The Mo catalyst was also found to hydrogenate C=C to C–C with no added hydrogen. This hydrothermal approach for decarboxylating real feedstocks such as castor oil, palm oil, and frying oil is found to be promising to produce oxygen-free liquid hydrocarbons.

The deactivation studies of catalysts showed that the Mo catalyst was slightly agglomerated compared to Ni and MgO catalysts. No graphitic coke was found in the three evaluated catalysts but higher amount of amorphous coke was detected in the Mo catalyst surface because of its high acidity. Amorphous coke is easy to remove by simple calcination and the catalyst can be reused.

Experimental Section

Materials

OA (90%), castor oil, hexanes (ACS Grade), nickel nitrate hexahydrate [Ni(NO3)2·6H2O], ammonium molybdate tetrahydrate [(NH4)6Mo7O24·4H2O], and magnesium sulfate heptahydrate [MgSO4·7H2O] were obtained from Sigma-Aldrich, Canada, and are used as received. Alumina (γ-Al2O3) powder (Catalox SSCa 5/200) was obtained from SASOL. Frying oil (used canola oil) was obtained from author’s home and filtered to remove solid particles before use. Palm oil was obtained from Malaysia. A compact ultrapure water system (EASY pure LF, Mandel Scientific Co., model BDI-D7381) was used to obtain deionized water (18.2 MΩ).

Catalyst Synthesis

Ni, Mo, and MgO supported on γ-Al2O3 catalyst were synthesized using an incipient impregnation method.[61,83] γ-Al2O3 was chosen as the support because it has larger surface area and pore volume and higher acidity compared to other phases of Al2O3 (results will be found in Supporting Information, Figure S13). Larger surface area provides more active sites and a higher acidic catalyst[33] is favorable for decarboxylation. For a standard synthesis, the desired amount of metal precursor for 10 wt % loading was dissolved in deionized water equivalent to 120 vol % of pore volume of alumina (0.50 cm3/gm). For example, 1 g of 10 wt % Ni–Al2O3 catalyst requires 0.50 g of Ni(NO3)2·6H2O and 0.9 g of Al2O3, 1 g of 10 wt % Mo catalyst requires 0.18 g of (NH4)6Mo7O24·4H2O and 0.9 g of Al2O3, 1 g of 10 wt % MgO requires 0.64 g of MgSO4·7H2O and 0.9 g of Al2O3, respectively. The required support (alumina) was immersed into the metal solution at once for better dispersion. The wet catalyst was placed into a vacuum oven at 80 °C overnight. The dry powder was then calcined into a muffle furnace at 600 °C @ 5 °C/min for 4 h. Hydrogen reduction with 5 vol % H2 balanced with N2 at 950 °C @ 3 °C/min for 2 h was performed afterward for Ni and Mo catalysts only. The actual metal loadings were confirmed by EDXS and XPS analysis.

Catalyst Testing

Decarboxylation of OA and its derivatives was performed in a benchtop reaction system (BTRSJR, Autoclave Engineers, Erie, PA) with a maximum allowable working pressure of 2900 psi @ 650 °C. Figure represents the detailed reactor setup. The system has a 10 mL tubular reactor (316 stainless steel reactor tube with type 316 stainless steel fittings) assembled with a furnace. The furnace temperature was set 20 °C higher than the reaction temperature. The reactor is connected with four feed lines for feeding either gas or liquids. All feeds were mixed and vaporized into a mixer vaporizer before feeding to the reactor. The feed stream then passed through the reactor. The mixer is placed into an oven to preheat the feed mixture, whereas the maximum oven temperature is 250 °C. A gas–liquid separator is placed next to the reactor outside the oven. The separator is surrounded by a cooling arrangement connected with a chiller, with the chiller temperature maintained at 6 °C during the reaction to separate the gas and liquid phases.

Figure 12

Schematic of continuous reactor setup.

Before starting any experiment, the reactor was washed thoroughly with soapy water, clean water, and then hexanes to remove any residuals from previous experiments. Air was passed through the reactor to remove any water or hexane sticking to the walls. Catalyst (5 g) was loaded for each experiment into the reactor with quartz wool being placed at the top and bottom of the reactor to maintain the catalyst inside. All fittings were attached to the reactor, which was then placed into the furnace. The main power of the system was turned on and the oven temperature was set to the desired reaction temperature. The reaction temperature was varied from 325 to 400 °C according to the experimental methodology explained below. During the reactor heating, N2 gas was flowed through the reactor to remove any oxygen. N2 flow was then stopped and the reactor outlet was opened to remove any N2 from the system. Two reactants (OA/its derivatives) and water were then fed to the reactor continuously. Each experiment was run for a minimum of 24 h and repeated at least three times in this study and the average results are reported. The space time (τ) was calculated from the volume of the catalyst (amount of catalyst/packed density of catalyst) divided by the reactant flow rate. Both reactants (fatty acid/derivatives and water) were accounted to calculate the flow rates at ambient T and P. The liquid and gaseous products were then collected and analyzed continuously from the gas–liquid separator until the catalyst started to deactivate. The liquid product was stored in glass vials. An air tight Tedler gas bag (SKC Inc., PA) was used to store the gaseous product for analysis. The spent catalyst collected for each run was thoroughly washed with excess hexanes. Hexanes were discarded from the solid catalyst. Liquid products which were stuck into catalyst bed were collected afterward during hexane evaporation. The wet solid catalyst was then dried at 80 °C for 12 h in a vacuum oven. The liquid product yield is calculated as the weight of the liquid product after reaction divided by the amount of OA fed to the reactor, whereas the gaseous product yield is (1—liquid product yield).

Product Analysis

Types of hydrocarbons present in the liquid product was quantified using a Shimadzu, GC-2014 equipped with a flame ionization detector and a capillary DB WAX column (Agilent Technologies, Santa Clara, CA, USA) (dimension: 30 m × 0.250 mm × 0.25 μm, temperature limit: 20 to 260 °C). Helium, hydrogen, and helium-air were used as the carrier, flame, and makeup gas, respectively. The oven temperature programming was followed from Hossain et al.[84] Retention times of known standards such as C8–C20 saturated hydrocarbons and heptadecene (Sigma-Aldrich, Oakville, ON) were used to identify the types of hydrocarbons in the liquid product. Sample (1 μL) with a 10:1 split ratio was injected manually into the column. Repetitive injection for each sample was performed until the similar peak area of interested compounds was obtained. Selectivity of hydrocarbons present in the liquid product was calculated from the following equation[85−87]

The injector and detector temperatures were retained at 200 and 250 °C, respectively.

Infrared spectra (600–4000 cm–1 with a resolution of 4 cm–1 over 64 scans) of feed and liquid products were collected using an ATR–FTIR spectroscope (Nicolet 6700 FTIR, Thermo Scientific). Degree of decarboxylation was calculated using the following equation

An Eagle Eye SG-Ultra Max Hydrometer (Density meter) [dimension = 5.5″ W × 5.5″ D × 1″ H (outside)] and IKA C2000 bomb calorimeter were used to measure density and HHVs of the liquid products.

Gaseous products formed during decarboxylation reaction were measured using a mixture of standard calibration gases (H2, N2, O2, CH4, CO, and CO2). Gas sample (1 mL) using an SGE gas tight syringe (model number 008100, Reno, NV USA) was injected manually into a nickel packed column (120/80 Hayesep D stainless steel 3.18 mm ID, 6.2 m L) of Shimadzu, GC-2014 (He as the carrier gas) which is connected with a thermal conductivity detector. The oven temperature was held at 35 °C for 6 min followed by 25 °C/min ramp and 1 min hold at 200 °C. The injector and detector temperatures were maintained at 200 and 250 °C, respectively. Injection was repeated multiple times for accuracy. Lighter fractions of hydrocarbons (C2 to C4) were quantified as the total no of moles of gas produced minus the number of moles of known gas.

Catalyst Characterization

N2-Physisorption

A Tristar II 3020 (Micromeritics Instrument Corporation, Norcross, GA, USA) instrument was used to determine the BET surface area and BJH pore size distributions of the fresh and spent catalysts at −193 °C using 99.995% pure N2 gas obtained from Praxair (Oakville, Canada). Approximately, 80 mg of each catalyst sample was degassed under an N2 atmosphere at 150 °C overnight before measurements. Degassing enhances the removal of the adsorbed moisture from the catalyst surface.

XRD Analysis

Crystallinity of catalyst samples was analyzed using a Bruker D2 Phaser powder diffractometer over 2θ = 10–80° using a scan rate of 0.1°/min and Cu Kα radiation (λ for Kα is equal to 1.54059 Å).

H2-TPR Analysis

An Autochem II 2920 analyzer (Micromeritics Instrument Corporation, Norcross, GA, USA) equipped with a thermal conductivity detector was used to obtain the TPR profiles of fresh catalysts by flowing 10% H2 balanced Ar at a rate of 50 mL/min. A minimum of 140–150 mg of the catalyst sample was used for each analysis. The temperature was increased from room temperature to 1100 °C @ 10 °C/min.

CO Pulse Chemisorption

An Autochem II 2920 analyzer was used to determine the active particle diameter, the % metal dispersion, and the active metal surface area of catalyst samples using a series of 1 mL of CO pulses (10% CO balanced with Ar) injected into the catalyst sample at 40 °C. The catalyst sample was pretreated with a stream of Ar (50 mL/min) before CO pulses were injected.

NH3-TPD Analysis

NH3 TPD analysis was conducted using a Micromeritics Autochem II 2920 analyzer coupled with a thermal conductivity detector. Samples were pretreated at 400 °C @ 15 °C/min for 1 h under helium. The sample was then cooled from 400 to 100 °C using 10% NH3 balanced with He (50 cm3/min) for another hour. NH3 flow was then stopped and He (50 cm3/min) was flowed at 100 °C for 1 h to remove any physically adsorbed NH3. The temperature was then increased to 850 °C @ 15°/min for desorption of NH3 from the catalyst surface. The holding time at 850 °C was 1 h. The acidity of the catalysts was calculated from the amount of NH3 desorbed from the acid sites of the catalyst during the desorption process which was monitored by the thermal conductivity detector.

TEM and SEM Imaging

The morphologies of both the fresh and spent catalysts were obtained from TEM (model JEOL 2010F) and SEM imaging (LEO 1530).

EDXS Analysis

Elemental composition of fresh and spent catalysts was confirmed and quantified by using the EDXS feature of the scanning electron microscope.

XPS Analysis

XPS analysis of fresh and spent catalyst samples was conducted with a Kratos Axis Ultra spectrometer using a monochromatic Al Kα source (15 mA, 14 kV).