One-Pot Synthesis of Sustainable Aviation Fuel from Brown Grease Using Multifunctional Zeolite-Supported Catalysts

Abstract

The most viable way to decarbonize aviation in the near term is through Sustainable Aviation Fuel (SAF), most of which is currently produced via the deoxygenation of fats, oils, and greases (FOG) followed by a separate isomerization step. Multifunctional zeolite-supported catalysts offer several advantages over existing formulations, such as enabling the use of waste FOG streams, performing their deoxygenation via decarboxylation/decarbonylation (deCO_x), and effecting the synthesis of SAF in one-pot. Previous work has shown that while supported Ni-Cu catalysts can afford excellent results in the conversion of waste FOG to fuel-like hydrocarbons via deCO_x, zeolitic materials represent promising supports in formulations employed for the synthesis of SAF. In this contribution, catalysts involving different zeolitic supports and the same Ni-Cu active phase were prepared, characterized, and tested in the conversion of brown grease to SAF to identify the carrier affording the best results. A Ni-Cu/ZSM-5 catalyst displayed the highest conversion and yield of SAF-like hydrocarbons relative to formulations supported on ZSM-22, SAPO-11, or SAPO-34 (these catalysts being referred to herein as NCZSM-5, NCZSM-22, NCSAPO-11, and NCSAPO-34).

1. Introduction

The demand for aviation fuel is expected to more than double by 2050 and triple by 2070 [1], which is of concern since aviation already accounts for 2.5% of global energy-related CO₂ emissions [2]. Given that the electrification of aircraft is in its infancy, the most promising way to decarbonize aviation in the near to medium term is through the use of Sustainable Aviation Fuel (SAF), which can be produced from several feedstocks via different technologies [3]. However, the International Air Transport Association (IATA) has reported that >80% of SAF is produced via the hydroprocessing of esters and fatty acids (HEFA) [4]. Processes utilizing fats, oils, and greases (FOG) as feedstocks—such as HEFA—are expected to remain the primary way to produce SAF for the foreseeable future given the maturity and widespread use of this technology [5].

Current HEFA processes are mostly based on the hydrodeoxygenation (HDO) reaction, which displays several shortcomings. Indeed, HDO-based HEFA requires large amounts and high pressures of hydrogen, which are typically only available at centralized facilities where this gas is produced from natural gas via steam reforming. In addition, HDO-based HEFA typically employs sulfided catalysts, which risk contaminating the product with sulfur [6,7], deactivate in the presence of water (a product of HDO) [7,8], are not regenerated online [9], and require FOG feedstocks low in impurities like phosphorus, metals, and free fatty acids (FFAs) [10]. Finally, SAF production via HDO-based HEFA involves separate deoxygenation and isomerization steps that result mainly in normal and branched alkanes but not in other SAF components, namely, cycloalkanes and aromatics. (Jet fuel—a mixture of C8–C16 hydrocarbons—comprises ~20 vol.% n-alkanes, ~33 vol.% iso-alkanes, ~27 vol.% cycloalkanes, and ~20 vol.% aromatics. These hydrocarbon families, respectively, confer jet fuel a high energy density, good cold flow properties, low freezing point, and seal-swelling properties [11]). This leads to SAF production facilities having separate units to (1) pre-treat the feedstock to remove impurities; (2) hydroprocess the pre-treated feedstock to yield n-alkanes; (3) isomerize the n-alkanes to iso-alkanes; and (4) separate the SAF from other fuel fractions via distillation, all of which adds complexity and cost.

Converting FOG to fuel-like hydrocarbons via decarbonylation/decarboxylation (deCO_x) can avoid the drawbacks of HDO-based HEFA since deCO_x reactions require smaller amounts and lower pressures of hydrogen and deCO_x catalysts are typically supported metals that can be simple (non-sulfided), inexpensive (non-precious), and robust (resistant to deactivation). Moreover, these formulations can be regenerated online and upgrade waste FOG streams rich in impurities and/or composed mainly of FFAs such as brown grease (BG) [12]—i.e., FOG collected from wastewater streams—all of which is currently incinerated or landfilled [13,14]. Another advantage of deCO_x formulations is that they can involve zeolitic materials as the support of multifunctional catalysts capable of performing deoxygenation and isomerization (but also cyclization and aromatization) reactions in “one-pot” [15]. Therefore, deCO_x-based technology has the potential to convert currently unutilized waste FOG streams to a product mixture containing all families of hydrocarbons comprising aviation fuel and to do so in less steps or units, which can reduce both complexity and cost.

Previous work has shown that supported metal catalysts comprising Ni and Cu in a 4:1 proportion (by weight), can afford excellent results in the conversion of BG to fuel-like hydrocarbons via deCO_x [12,16]. In addition, several zeolitic materials have been studied as supports for Ni-based catalysts used in the synthesis of SAF from oleaginous biomass, promising results being obtained using SAPO-11 [17], SAPO-34 [18], ZSM-5 [19,20], and ZSM-22 [21]. However, the literature lacks systematic studies in which the nature of the active phase and the reaction conditions employed are fixed and the zeolitic support is varied to identify the carrier(s) leading to the best results in the conversion of FOG to SAF. A notable exception is the work of Wang et al., who upgraded vegetable oil over Ni-based catalysts supported on SAPO-11, ZSM-5, ZSM-22, ZSM-23, or β-zeolites and observed quantitative conversion, high isomerization activity, and decreased cracking over 8 wt.% Ni/SAPO-11, which was found to be the most promising among all catalysts studied [17]. The present contribution aimed to further address the aforementioned gap in the literature by supporting Ni and Cu in a 4:1 proportion (by weight) on several zeolitic materials and testing the resulting catalysts in the conversion of BG to SAF-like hydrocarbons via deCO_x. The characterization of both fresh and spent catalysts was also undertaken to rationalize trends and elucidate structure–activity relationships.

2. Results and Discussion

2.1. Catalyst Synthesis and Fresh Catalyst Characterization

A series of catalysts involving 12 wt.% Ni-3 wt.% Cu as the supported metals and different zeolitic materials—namely, SAPO-11, SAPO-34, ZSM-5, and ZSM-22—as the carrier were prepared via incipient wetness impregnation employing the materials and methods described in Section 3.1. The synthesized catalysts were analyzed by means of inductively coupled plasma mass spectrometry (ICP-MS) to experimentally determine their metal loadings. The results of these tests, which are summarized in

Table A1. Experimental loading determined via ICP-MS of the catalysts used in this study.

Catalyst	Ni (%)	Cu (%)
NCSAPO-11	12.8 ± 0.4	3.3 ± 0.1
NCSAPO-34	12.0 ± 0.1	3.0 ± 0.0
NCZSM-5	12.8 ± 0.1	3.2 ± 0.0
NCZSM-22	13.2 ± 0.1	2.9 ± 0.1

within Appendix A, show that the experimental loadings obtained are close to the nominal loadings targeted.

The textural properties of the bare supports employed and of the catalysts synthesized were determined by means of N₂-physisorption, the results of these measurements being included in

Table 1. Textural properties and average NiO particle size of the catalysts used in this study.

Catalyst	Surface Area (m2/g)	Pore Volume (cm3/g)	Avg. Pore Diameter (nm)	Avg. NiO Particle Size (nm)
SAPO-11	158.5	0.06	2.1	-
NCSAPO-11	76.9	0.02	1.8	25.0
SAPO-34	525.7	0.26	1.8	-
NCSAPO-34	427.9	0.20	3.0	28.4
ZSM-5	300.4	0.12	2.0	-
NCZSM-5	181.4	0.07	2.1	25.1
ZSM-22	59.6	0.01	6.0	-
NCZSM-22	52.9	0.07	6.2	21.9

. The data in Table 1 clearly show that the textural properties of the bare zeolitic supports vary widely. For instance, surface area, which follows the trend ZSM-22 < SAPO-11 < ZSM-5 < SAPO-34, ranges from 59.6 to 525.7 m²/g. Unsurprisingly, the surface area of the supported metal catalysts is lower than that of the bare supports, since metal deposition causes some degree of pore filling.

The X-ray diffractograms of the catalysts synthesized are included in Figure A1 within Appendix A. The diffraction peaks so denoted are attributed to NiO, peaks at 37.3°, 43.3°, and 62.9°, respectively, representing its (111), (200), and (220) planes [22]. The fact that no peaks associated with Cu species were observed can be attributed to the lower Cu loading and/or to Cu species being highly dispersed, which is in agreement with previous reports on NCSAPO-11 [15]. The Scherrer equation was applied to the diffraction peak at 43.3° to determine the average NiO particle size, which afforded the values included in Table 1. The average NiO particle size ranges from 21.9 to 28.4 nm and it follows the trend NCZSM-22 < NCSAPO-11 ≈ NCSZM-5 < NCSAPO-34. Notably, the latter trend is the same as that observed for the surface area of the bare supports, which is at odds with the fact that carriers with a higher surface area would be expected to stabilize supported metals in the form of smaller particles and suggests the presence of amorphous and/or sub-nanometric Ni species not detected by means of XRD. Additional information regarding metal dispersion was acquired through H₂ pulse chemisorption measurements, which resulted in the data included in

Table 2. H₂ uptake, Ni-specific surface area, and total acidity of the catalysts used in this study.

Catalyst	H2 Uptake (cm3/g)	Metal-Specific Surface Area (m2/g)	Total Acidity (mmol of NH3/g)
NCSAPO-11	0.040	0.140	1.56
NCSAPO-34	0.005	0.018	1.98
NCZSM-5	0.405	1.414	1.44
NCZSM-22	1.250	4.360	0.63

. The trend observed for H₂ uptake and metal-specific surface area—namely, NCSAPO-34 < NCSAPO-11 < NCSZM-5 < NCZSM-22—is also indicative of the metal dispersion trend followed by these catalysts.

More detailed information regarding the elemental composition and chemical environment of the surface of the catalysts surveyed was acquired via X-ray photoelectron spectroscopy (XPS) measurements, which were performed after reducing the catalysts using the conditions and approach described in Section 3.2.

Table 3. Surface concentration and oxidation state (in at.%) of supported metals as determined via XPS.

Catalyst	Ni (Ni0)	Cu (Cu0)
NCSAPO-11	1.05 (1.05)	0.59 (0.10)
NCSAPO-34	5.2 (NAD)	1.39 (0.57)
NCZSM-5	1.02 (1.02)	0.46 (0.34)
NCZSM-22	5.17 (NAD)	1.17 (0.80)

NAD: Not accurately determinable.

summarizes the results of these analyses in terms of the surface concentration and oxidation state of the supported metals, while

Table A2. Elemental surface composition of the catalysts employed in this study determined via XPS (at%).

Catalyst	C	O	P	Si	Al	Ni	Cu
NCSAPO-11	7.61	40.38	3.41	14.14	32.83	1.05	0.59
NCSAPO-34	2.87	18.69	2.87	0	68.99	5.2	1.39
NCZSM-5	3.21	46.85	0	31.9	16.55	1.02	0.46
NCZSM-22	7.07	15.86	3.44	10.11	57.17	5.17	1.17

and

Table A3. Area of XPS peaks representing supported metal species (at.%) and their respective binding energies.

Catalyst	Metal Species (%) [Binding Energy (eV)]
	Ni0	NiO	Cu0	CuO	Cu2O	Cu (Cu0)
NCSAPO-11	100 [852.64]	0 [NA]	18.1 [934.23]	80.6 [936.96]	0 [N/A]	1.29 [946.77]
NCSAPO-34	NAD [852.81]	NAD [NA]	41.28 [932.51]	0 [N/A]	58.72 [937.22]	0 [N/A]
NCZSM-5	100 [852.67]	0 [NA]	73.83 [933.76]	0 [N/A]	26.17 [934.24]	0 [N/A]
NCZSM-22	NAD [852.72]	NAD [NA]	68.29 [931.79]	0 [N/A]	31.71 [934.22]	0 [N/A]

NAD: Not accurately determinable. NA: Not applicable.

in Appendix A provide more detailed information vis-à-vis the elemental surface composition and the supported metal species of the catalysts as determined via XPS. Full scans as well as Ni and Cu 2p regions of X-ray photoelectron spectra are included in Figure A2, Figure A3 and Figure A4 within Appendix A. The XPS analysis of NCSAPO-11 and NCZSM-5 clearly showed that the entirety of Ni is present in the metallic state in these two catalysts. Similarly, the low intensity of the Ni 2p_3/2 satellite peak in the spectra of NCSAPO-34 and NCZSM-22 suggests that Ni is mostly reduced in these two formulations. However, the inhomogeneous character of these two samples precludes the neutralization of charging effects, which cause peak broadening and the appearance of a supplemental peak in the spectra of NCZSM-22 and NCSAPO-34, respectively, and make it impossible to accurately determine Ni⁰ concentration.

The most salient feature of the data in Table 3 is the surface concentration of the metals, which follows the trend NCSAPO-34 > NCZSM-22 >> NCSAPO-11 > NCZSM-5. However, the surface concentration of Cu⁰ shows a slightly different trend—namely, NCZSM-22 > NCSAPO-34 > NCZSM-5 > NCSAPO-11—which confirms that the reducibility of the surface metals varies across these catalysts. Indeed, the trend observed vis-à-vis the fraction that Cu⁰ represents of the total surface metal is NCSZM-5 > NCSZM-22 > NCSAPO-34 > NCSAPO-11. Admittedly, it is difficult to relate particle size information estimated from XRD, metal-specific surface area results from H₂ pulse chemisorption, and surface concentration data on the supported metals derived from XPS. This can be attributed to the heterogeneity of Ni and Cu species expected to result from the deposition of these metals via impregnation on the zeolitic materials employed, since Ni and Cu can be expected to be present not only as nanoparticles but also as amorphous clusters and/or sub-nanometric species at different locations of the carriers. For instance, the fact that the Cu₂O signal is observed at much higher binding energy than usual—see Table A3 and Figure A4 in Appendix A—may suggest an ionic Cu species exchanged on zeolitic Brønsted acid sites. Nevertheless, the results above provide valuable insights regarding the size of metal nanoparticles, metal-specific surface area, as well as the surface concentration and oxidation state of the supported metals.

The acidity of the catalysts used in this study was assessed by means of ammonia temperature-programmed desorption (NH₃-TPD) given the importance of surface acidity to several reactions and phenomena of interest, including isomerization and cracking as well as coke-induced deactivation. In terms of total acidity, the catalysts followed the trend NCSAPO-34 > NCSAPO-11 > NCZSM-5 > NCZSM-22 (see Table 2), which is in overall agreement with previously reported acidity trends for the bare supports [23,24]. In contrast with previous reports focused on the bare supports, most acidity was found to be associated with weakly acidic sites (from which ammonia desorbs below 250 °C) albeit a smaller number of sites of medium acidity (from which ammonia desorbs between 250 and 400 °C) were also found to be present (see Figure 1). The fact that strongly acidic sites (displaying higher NH₃ desorption temperatures) were not observed can be attributed to the fact that these sites would represent the main adsorption and anchor sites for metal precursors and particles, respectively. Thus, any strongly acidic sites—along with most sites of medium acidity—would be occupied during catalyst synthesis, leaving only weakly acidic sites and the relatively small number of sites of medium acidity observed.

2.2. BG Conversion to SAF

The BG sample used in this study was analyzed via gas chromatography (GC) and found to be composed entirely of C18 and C16 FFAs as shown in

Table A4. Composition of the BG used in this study.

Fatty Acid	Concentration (%)
Oleic acid (C18:1)	61.1
Stearic acid (C18:0)	22.5
Palmitic acid (C16:0)	15.1
Linoleic acid (C18:2)	1.2

in Appendix B. This feed was upgraded over the catalysts synthesized using the equipment and methods described in Section 3.3, the experimental approach employed used for the GC analysis of both the feed and the reaction products being described in Section 3.4. The results of these upgrading experiments and analyses are summarized in

Table 4. Results of BG conversion to SAF over the catalysts used in this study.

Catalyst	Conversion (%)	C8–C16 Hydrocarbons (%)	Selectivity (%)
NCSAPO-11	73.8 ± 14.1	80.5 ± 12.3	C8–C16 n-alkane: 88.4 ± 5.6
			C8–C16 iso-alkane: 7.6 ± 4.6
			C8–C16 cycloalkane: 2.6 ± 0.6
			C8–C16 aromatic: 1.4 ± 0.4
NCSAPO-34	98.5 ± 0.1	82.9 ± 6.5	C8–C16 n-alkane: 88.5 ± 4.6
			C8–C16 iso-alkane: 7.5 ± 2.7
			C8–C16 cycloalkane: 2.5 ± 1.9
			C8–C16 aromatic: 1.5 ± 0.1
NCZSM-5	99.9 ± 0.0	87.3 ± 0.9	C8–C16 n-alkane: 49.4 ± 6.3
			C8–C16 iso-alkane: 22.3 ± 6.7
			C8–C16 cycloalkane: 4.0 ± 1.3
			C8–C16 aromatic: 24.4 ± 1.5
NCZSM-22	99.9 ± 0.0	88.7 ± 1.9	C8–C16 n-alkane: 74.5 ± 2.3
			C8–C16 iso-alkane: 17.3 ± 1.1
			C8–C16 cycloalkane: 1.2 ± 0.5
			C8–C16 aromatic: 7.0 ± 1.7

. Saliently, both BG conversion and the fraction of aviation fuel-range alkanes (C8–C16 hydrocarbons) in the reaction products follow the trend NCSAPO-11 < NCSAPO-34 < NCZSM-5 ≈ NCZSM-22, which suggests that 12 wt.% Ni-3 wt.% Cu catalysts supported on the socony mobil zeolites tested afford better results that the corresponding formulations involving the silicoaluminophospate zeotypes essayed. Parenthetically, the inferior conversion of NCSAPO-11 is likely not due to its relatively low concentration of surface Ni⁰ and Cu⁰ (see Table 3), which represent the active sites for the deCO_x reactions. Indeed, NCZSM-5 affords quantitative conversion despite displaying a slightly lower concentration of Ni⁰ than NCSAPO-11 and the second lowest concentration of Cu⁰ of all catalysts tested. Similarly, differences in conversion cannot be assigned to the surface Ni⁰/Cu⁰ concentrations of these catalysts, since the ratios for NCSAPO-11 and NCZSM-5 are 10.5 and 3.0, respectively. Instead, differences in conversion can be attributed—at least in part—to the propensity of these formulations towards fouling and coking (see Section 2.3). ZSM-supported catalysts also afford better results than SAPO-supported formulations in terms of the selectivity to the hydrocarbon families comprising aviation fuel, since the catalyst supported on SAPO-11 and SAPO-34 displayed almost identical selectivity of ~88.5% n-alkanes, ~7.5% iso-alkanes, ~2.5% cycloalkanes, and ~1.5% aromatics. In contrast, the catalysts supported on ZSM-5 and ZSM-22 showed lower selectivity to n-alkanes and higher selectivity to the other hydrocarbon families comprising aviation fuel. Among these, NCZSM-5 afforded the highest selectivity to iso-alkanes and cycloalkanes, as well as a yield of aromatics within the 8–25% specified for aviation fuel [25].

To elucidate the fate of the oxygen atoms in the BG feed, the gas stream produced during the upgrading of BG over NCZSM-5 was analyzed with a refinery gas analyzer (RGA) and found to contain 0.45% CO and 0.26% CO₂ as the only oxygen-bearing compounds, which is unsurprising since deoxygenation proceeds via deCO_x. The other compounds detected by the RGA were hydrogen (the reaction atmosphere) and C1–C6 hydrocarbons resulting from CO_x methanation and/or the cracking of alkyl chains, which is consistent with previous reports [26].

Figure 2 shows the carbon number distribution of the product mixtures resulting from upgrading experiments. Irrespective of the catalyst support employed, most hydrocarbon products were found to be within the C8–C16 range characteristic of aviation fuel (with the remainder being ≥C17 hydrocarbons). However, SAPO-supported catalysts show noticeable differences in terms of the carbon number of the reaction products relative to those obtained over ZSM-supported formulations. Indeed, NCSAPO-11 and NCSAPO-34 yield a considerable and comparable amount of C15 and C11 hydrocarbons, these two carbon numbers representing the most abundant among the products obtained over these catalysts. In contrast, NCZSM-5 and NCZSM-22 afford much lower amounts of C15, since lighter (≤C13) hydrocarbons are favored among which C10 hydrocarbons are the most abundant.

These results cannot be readily explained by the relative acidity of the supports favoring cracking reactions—since ZSM-supported catalysts have a total acidity that is either slightly (NCZSM-5) or much lower (NCZSM-22) than SAPO-supported formulations—albeit this can be attributed to the fact that most of the acid sites in these supports are either weakly or moderately acidic (see Figure 1) with a relatively low activity towards C-C cleavage [27,28]. Instead, these results can be rationalized by invoking the metal-specific surface area of ZSM-supported catalysts, which is much higher than that of their SAPO-supported counterparts (see Table 2). Indeed, it can be argued that a higher concentration of surface metal sites would increase the metal-catalyzed hydrogenolysis of C–C bonds, which in turn would favor the formation of lighter—at the expense of heavier—hydrocarbons. The relatively low concentration of surface Ni⁰ within NCSAPO-11 can only partially explain the lower cracking activity of this catalyst, since NCZSM-5 displays a slightly lower concentration of surface Ni⁰ (see Table 4). In addition to the aforementioned heterogeneity of Ni and Cu species, the lack of a consistent trend in this regard can also be attributed to the relative distribution of these metals on the catalyst surface, since C–C hydrogenolysis reactions require contiguous Ni atoms to be catalyzed [29] and Cu can disrupt Ni adjacency [30], curbing cracking reactions through both geometric and electronic effects [12,30,31,32]. However, the ability of Cu to disrupt Ni adjacency and cracking reactions appears to be limited in these catalysts, since NCSAPO-11 displays a lower cracking activity than NCZSM-5 despite having a much higher surface Ni⁰/Cu⁰ concentration ratio (10.5 vs. 3.0).

Given that the results in Table 4 preclude the identification of NCSZM-5 or NCZSM-22 as the most active catalyst (since both formulations afforded quantitative conversion), upgrading experiments were performed in which the amount of BG employed was doubled relative to that specified in Section 3.3. The results of these experiments are summarized in

Table A5. Results of BG conversion to SAF over the catalysts used in this study when twice the amount of BG relative to that specified in Section 3.3 was employed.

Catalyst	Conversion (%)	C8–C16 Hydrocarbons (%)	Selectivity (%)
NCZSM-5	99.9 ± 0.0	79.1 ± 3.7	C8–C16 n-alkane: 58.4 ± 17.1
			C8–C16 iso-alkane: 10.7 ± 2.2
			C8–C16 cycloalkane: 0.4 ± 0.5
			C8–C16 aromatic: 28.5 ± 14.1
NCZSM-22	99.8 ± 0.3	79.2 ± 4.1	C8–C16 n-alkane: 71.3 ± 10.4
			C8–C16 iso-alkane: 8.3 ± 2.0
			C8–C16 cycloalkane: 2.1 ± 1.6
			C8–C16 aromatic: 17.2 ± 9.3

in Appendix B. Surprisingly, both catalysts displayed quantitative BG conversion and practically identical selectivity to C8–C16 hydrocarbons under these conditions. An additional effort was made to identify the most active formulation between NCSZM-5 or NCZSM-22 by halving the amount of catalyst relative to the amount used in the experiments that afforded the data in Table A5. However, conversion remained quantitative, which suggests that testing these catalysts in a fixed-bed reactor, which would allow for additional variables (such as space velocity and time on stream) to be investigated—represents a better approach to enable this comparison. Previous reports on the upgrading of BG over supported Ni-Cu catalysts indicate that fixed-bed reactor experiments are also the best was to study catalyst stability and recyclability [12]; however, those experiments fall outside of the scope of this work and will be the focus of a future contribution.

2.3. Spent Catalyst Characterization

In an effort to rationalize the results of BG upgrading to SAF, the spent catalysts were analyzed by means of thermogravimetric analysis (TGA). Indeed, TGA offers a way to determine the amount of type of the carbonaceous deposits that accumulate on the catalyst surface in the course of the reaction, which can in turn serve as a gauge for the catalyst deactivation caused by fouling and coking. Parenthetically, coking and fouling represent the main deactivation mechanism for supported Ni-Cu catalysts used for the conversion of oleaginous biomass to fuel-like hydrocarbons, since metal loss due to leaching or even the formation of volatile carbonyls has been found to be negligible [12,15]. The results of these TGA measurements are summarized in Figure 3. Saliently, the spent NCSAPO-11 catalyst displayed the most significant weight loss (~18%), which suggests that fouling and coking could explain (at least in part) the fact that it shows the lowest BG conversion among all catalysts tested. NCSAPO-11 is unique in the sense that it represents the only spent formulation displaying considerable weight losses below 400 °C—which can be assigned to the removal of residual solvent, reactants, products as well as soft coke—as well as weight losses above the same temperature that can be attributed to the combustion of hard coke or graphitic carbon [33]. Indeed, spent NCSAPO-34 displays the vast majority of its weight loss below 400 °C, whereas spent NCZSM-5 and NCZSM-22 undergo most of their weight loss above this temperature.

In short, it can be argued that coking and fouling can only be invoked to explain catalytic activity when the catalyst displays both a relatively low surface area and a high propensity for coking as is the case for NCSAPO-11. All other formulations display a moderate degree of coking (like NCZSM-22), a relatively high surface area (like NCSAPO-34), or both (as is the case for NCZSM-5). The analysis by means of TGA of the catalysts spent in the experiments resulting in the data within Table A5 seems to corroborate this conclusion, since the weight loss spent NCZSM-22 and NCZSM-5 displayed, respectively, remained unchanged and increased only slightly (see Figure A5 in Appendix B) with no impact on conversion. Lastly, it should be noted that the TGA results can only be partially explained invoking acidity. Indeed, the least acidic catalysts—namely, NCZSM-22—displays the lowest degree of fouling and coking as would be expected. However, the extent of coking and fouling observed for the other catalysts does not show a linear relationship with acidity, which suggests that other variables (like the concentration and relative distribution of Ni and Cu on the catalyst surface) are at play.

2.4. Catalyst Recycling Studies

The recyclability of the catalyst that afforded the most promising results—i.e., NCZSM-5 (see Section 2.2)—was assessed by testing this catalyst in five sequential BG upgrading reactions. Between reactions, the spent catalyst was regenerated through a procedure comprising a calcination step under static air (to remove carbonaceous deposits from the spent catalyst surface) and a reduction step identical to the one used before the first upgrading reaction (to re-reduce the supported metals oxidized during the calcination step). The results of these experiments are summarized in Figure 4.

Notably, BG conversion remained quantitative across all reaction cycles. Interestingly, the selectivity to C8–C16 hydrocarbons was 100% in cycles 1 and 2, dropped to 86.5 and 82.7% in cycles 3–4 (due to the formation of oxygenates and C19 hydrocarbons) and increased to 98% in cycle 5 (since much less oxygenates and no C19 hydrocarbons were produced in this last cycle). Parenthetically, the oxygenates detected in some products mixtures were aldehydes, alcohols, ethers, and ketones, which are produced in small amounts via the incomplete deoxygenation of FFAs in BG and through condensation reactions involving oxygenated intermediates. Finally, Figure 4 also shows how the selectivity to the different hydrocarbon families comprising jet fuel (n-alkanes, iso-alkanes, cyclo-alkanes, and aromatics) varied significantly across cycles.

To rationalize these results, the post-mortem of the NCZSM-5 catalyst recovered at the end of the recycling experiments was performed focusing on Ni-specific metal active sites and on surface acid sites associated with the zeolitic support. H₂ pulse chemisorption results showed that the metal-specific surface area of the catalyst dropped from 1.414 m²/g in the fresh catalyst to 0.004 m²/g in the catalyst recovered at the end of the recycling experiments, which clearly indicates a loss of metal sites likely due to metal sintering and coking. Similarly, NH₃-TPD measurements indicate that the total acidity of the catalyst dropped from 1.44 mmol NH₃/g on the fresh catalyst to 0.59 mmol NH₃/g on the catalyst recovered at the end of the recycling experiments, which is indicative of a loss of acid sites attributable to the accumulation of recalcitrant coke deposits.

The results of recycling studies and of the post-mortem catalyst characterization suggests that metal sites within NCZSM-5 have a very high intrinsic activity, since conversion remained quantitative despite the loss of most metal sites. Moreover, changes in the amount of metal active sites and surface acid sites are in line with the changes in selectivity observed during the recycling studies, since structural changes are bound to affect the performance of the catalyst in cracking, isomerization, cyclization, and aromatization reactions. In short, although these studies show the recycling stability of the catalyst in terms of conversion, the results also show variance in terms of selectivity, clearly indicating the direction of future work. However, the latter falls outside of the scope of the present contribution and will be the focus of future studies.

3. Materials and Methods

3.1. Catalyst Synthesis

Catalysts were prepared via incipient wetness impregnation—using Ni(NO₃)₂·6H₂O (Alfa Aesar, Ward Hill, MA, USA) and Cu(NO₃)₂·3H₂O (Sigma Aldrich, St. Louis, MO, USA) as metal precursors and SAPO-11 (6% SiO₂, 48% Al₂O₃; ACS MATERIAL, Pasadena, CA, USA), SAPO-34 (~0.5 SiO₂/Al₂O₃ molar ratio; ACS MATERIAL), ZSM-5 (30 SiO₂/Al₂O₃ molar ratio; Zeolyst, Kansas City, KS, USA), or ZSM-22 (65–80 SiO₂/Al₂O₃ molar ratio; ACS MATERIAL) as the support—targeting a Ni loading of 12 wt.% and a Cu loading of 3.0 wt.%. After drying under vacuum overnight at 60 °C, the catalysts were calcined at 500 °C for 3 h in static air.

3.2. Fresh Catalyst Characterization

The fresh catalysts prepared were characterized using previously described equipment and methods for ICP [34], N₂ physisorption [35], XRD [32], NH₃ TPD [36], H₂ pulse chemisorption [36], and XPS [15,32,37]. Briefly, ICP was performed using a Varian 720-ES analyzer (Agilent Technologies, Santa Clara, CA, USA), while XRD was carried out on a Phillips X’Pert diffractometer (Philips Analytical, Almelo, The Netherlands) with Cu Kα radiation at 0.02° step size. NH₃ TPD and H₂ pulse chemisorption experiments were performed on an Autochem II analyzer (Micromeritics, Norcross, GA, USA) outfitted with a thermal conductivity detector (TCD) and a Thermostar mass spectrometer (Pfeiffer, Aßlar, Germany). For each experiment, 250 mg of catalyst was loaded into a U-tube quartz reactor connected to the instrument and heated by a clamshell oven. Prior to analysis, the catalyst was reduced at 350 °C for 1 h under 10% H₂/Ar. For NH₃ TPD, after cooling the reduced catalyst to 30 °C, the sample was exposed to a flow (100 sccm) of 1% NH₃/N₂ for an hour before purging with a flow (100 sccm) of Ar for another hour to remove physisorbed NH₃. The sample was then heated to 500 °C at a rate of 10 °C/min while NH₃ desorption was monitored using the mass spectrometer signal at m/z = 16. For H₂ pulse chemisorption, after reduction the reactor was purged with Ar at 450 °C for 30 min and cooled to 45 °C. Pulses of 0.025mL (STP) H₂ were then introduced to the reactor in 3 min intervals using Ar (50 sccm) as carrier gas, the TCD being employed to monitor H₂ breakthrough. Pulses were continued until the amount of breakthrough H₂ detected via TCD became constant. For XPS analyses, catalysts were reduced under a flow of H₂ at 350 °C for 3 h in a pre-treatment chamber connected to the instrument and allowing the transfer of pre-treated samples into the XPS analysis chamber without exposure to air. XPS was performed using a PHI 5000 Versaprobe—Physical Electronics (Chanhassen, MN, USA)–apparatus using a monochromatic Al Kα1 X-Ray source with energy of 1486.6 eV, accelerating voltage of 15 kV, power of 50 W, and spot size diameter of 200 μm. Spectra were processed using the CasaXPS software package (version 2.3.25PR1.0).

3.3. Catalyst Testing for BG Conversion to SAF

Catalyst testing was conducted in a 100 mL stainless steel reactor—Autoclave Engineers (Erie, PA, USA)—operated in semi-batch mode. After placing 0.5 g of the formulation to be tested in the reactor, closing the latter, and purging the system with Argon, the catalyst was reduced at 350 °C for 3 h under 100 psi and a flow (60 sscm) of 10% H₂ in N₂. Subsequently, the reactor was allowed to cool down to room temperature and the system was purged with Argon. Then, a solution comprising 1.8 g of BG—which was produced by “Greasezilla™” a product of Downey Ridge Environmental Company (Lansing, WV, USA) and has the composition in Table A4 within Appendix B—dissolved in 22.75 g of dodecane (99+% Sigma Aldrich) was introduced to the reactor through a port in the reactor head using a needle connected to a syringe containing the feed. After sealing the reactor, Argon was replaced with H₂, which was used to pressurize the system to 580 psi and establish a flow of 60 sccm using a backpressure regulator and a mass flow controller, respectively. The reaction mixture was then heated to 375 °C and kept at said temperature for 3 h under constant mechanical stirring (1000 rpm). All experiments were performed in duplicate to ascertain both reproducibility and the statistical significance of the results.

3.4. Analysis of BG and Its Catalytic Upgrading Products

The identity and quantity of all compounds comprising the BG feed employed and the liquid products stemming from its catalytic upgrading were determined via dual detection GC using instrumentation and methods described in detail in a previous contribution [38]. Briefly, analyses were performed using an Agilent 7890A GC (Santa Clara, CA, USA) equipped with both MS and flame ionization detectors. The inlet was operated in split mode (split ratio 25:1; split flow 50 mL/min) using an initial temperature of 100 °C. The inlet temperature was increased immediately upon injection (at a rate of 8 °C/min) to a final temperature of 320 °C, which was maintained for the duration of the analysis. The initial oven temperature (45 °C) was increased upon injection first to 325 °C (at a rate of 4 °C/min) and then to 400 °C (at a rate of 10 °C/min). This temperature was then maintained for 12.5 min, making the total run time 90 min. An Agilent J&W DB-5HT column (30 m × 250 μm × 0.1 μm) rated to 450 °C was employed along with a constant He flow of 2 mL/min. Quantification was performed using cyclohexanone as internal standard. Agilent Chemstation and Separation Systems Inc. (Gulf Breeze, FL, USA). SimDis Expert 9 software were used to perform chromatographic programming and to process the chromatographic data acquired. Solvents (i.e., chloroform and dodecane) were quenched and subtracted prior to data processing. A representative chromatogram—corresponding to the liquid products stemming from the upgrading of BG over NCZSM-5—is provided in Appendix C as Figure A6. In addition, a table including the retention time (RT) of each peak in this chromatogram along with its corresponding compound name, compound carbon number (CN), and percent area under the curve (% AUC) is provided in the same appendix as

Table A6. Retention time (RT), compound name, compound carbon number (CN), and percent area under the curve (% AUC) of the chromatogram in Figure A6.

RT	Compound Name	CN	% AUC
2.134	3,5 dimethyl cyclohexene	C8	1.022
2.164	1 methyl 4 methylene cyclohexane	C8	0.414
2.206	methyl ethyl cyclopentene	C8	2.231
2.545	4 methyl octane	C9	4.259
2.615	p-xylene	C8	8.029
3.371	3,3,5 trimethyl cyclohexane	C9	1.214
3.878	propyl benzene	C9	1.499
4.071	1 ethyl 3 methyl benzene	C9	6.17
4.142	3,5 dimethyl octane	C10	1.158
4.267	3 methyl nonane	C10	1.684
4.687	1,2,3 trimethyl benzene	C9	2.235
4.903	decane	C10	35.813
5.989	1,2 diethyl benzene	C10	0.437
6.071	1 methyl 3 propyl benzene	C10	1.219
6.18	1 methyl 4 propyl benzene	C10	1.954
6.335	5 methyl decane	C11	0.441
6.433	4 methyl decane	C11	0.558
6.559	2 methyl decane	C11	0.599
6.71	3 methyl decane	C11	1.116
6.943	2 ethyl 1,4 dimethyl benzene	C10	1.296
7.556	undecane	C11	7.825
9.208	5 ethyl decane	C12	4.543
9.336	4 methyl undecane	C12	3.241
9.499	2 methyl undecane	C12	3.657
9.679	3,8 dimethyl decane	C12	3.803
11.092	2,6 dimethyl undecane	C13	1.63
13.135	2 methyl naphtalene	C11	0.476
13.766	tridecane	C13	0.473
18.569	4 methyl tetradecane	C15	0.268
19.799	pentadecane	C15	0.515
25.319	hexadecane	C16	0.221

.

3.5. Spent Catalyst Characterization

Spent catalysts were subjected to TGA using a Discovery Q500 thermogravimetric analyzer—TA Instruments (New Castle, DE, USA)—after being dried in a vacuum at 60 °C overnight to remove water. Dried spent catalysts held in a platinum TGA pan were heated from room temperature to 800 °C at a rate of 10 °C/min under a flow (50 sccm) of air.

3.6. Catalyst Recycling Studies

Recycling studies were performed using NCZSM-5, the most promising formulation identified in catalyst testing runs. After each upgrading reaction, the spent catalyst was recovered via filtration, rinsed with chloroform, dried at 60 °C overnight in a vacuum oven, and calcined under static air at 450 °C for 5 h using a muffle furnace to remove carbonaceous deposits (see Figure 3). The calcined catalyst was then subjected to the same catalyst testing procedure described in Section 3.3, adjusting the amount of feed employed to keep the catalyst–feed ratio constant for five reaction cycles.

4. Conclusions

In this study, catalysts comprising Ni and Cu supported on various zeolitic carriers were synthesized, characterized, and evaluated for the conversion of BG to SAF. Among the tested formulations, a 12.8% Ni-3.2% Cu/ZSM-5 (NCZSM-5) catalyst exhibited the highest BG conversion and selectivity to SAF-like hydrocarbons. Specifically, the NCZSM-5 catalyst achieved a BG conversion of 99.9% and a jet fuel yield of 87.3%. Moreover, NCZSM-5 afforded SAF-range products containing n-alkanes, iso-alkanes, cycloalkanes, and aromatics, all integral components of aviation fuel. The results obtained compare favorably to those recently reported by Ritmun. Indeed, this author also found that a 12.5% Ni-2.5% Cu/ZSM-5 catalyst displayed the best results among a series of Ni-Cu formulations supported on various zeolites—including SAPO-11 and ZSM-5—in the conversion of palm oil to SAF [39]. A notable difference is that while palm oil is an edible feedstock composed almost entirely of triglycerides, BG is a waste stream that is typically incinerated or landfilled and is 100% FFAs. The latter is significant, as current HEFA technology typically avoids feedstocks with FFA levels exceeding 25% [40]. This highlights the potential of BG as a viable and underutilized resource for SAF production.

Additionally, this study addressed a gap in the literature by systematically varying the zeolitic support while using the same active phase and reaction conditions in the conversion of BG to SAF. The findings demonstrate that ZSM-5 offers superior performance relative to other supports tested, including SAPO-11, SAPO-34, and ZSM-22. This contrasts with the results of Wang et al., who found that SAPO-11 led to the best results when they upgraded vegetable oil over Ni-based catalysts supported on SAPO-11, ZSM-5, ZSM-22, ZSM-23, or β-zeolites [17].

Overall, the results presented herein provide valuable insights into catalyst design for SAF production from FOG. Particularly, they underscore the role of zeolitic supports in optimizing performance and suggest that the latter may be dependent on the nature of the active phase and/or the feedstock employed.