Cost and Carbon Intensity Implications of Coprocessing Sustainable Aviation Fuel at Petroleum Refineries

Abstract

Sustainable aviation fuel (SAF) will play a critical role in decarbonizing the aviation industry. Among SAF production pathways, alcohol-to-jet (ATJ) stands out for its scalability, supported by abundant feedstock availability and a well-established bioethanol industry. However, significant reductions in SAF carbon intensity (CI) require the use of future feedstocks (e.g., cellulosic) whose adoption is hindered by high capital costs for feedstock processing and ethanol upgrading. Here, we evaluate the financial viability and environmental implications of integrating an ATJ SAF biorefinery within a petroleum refinery, utilizing miscanthus and switchgrass as example feedstocks. Three scenarios are evaluated: standalone (benchmark), colocated, and repurposing (coprocessing SAF within the petroleum refinery). Results show repurposing reduces baseline capital costs by 36% and SAF minimum selling price (MSP) by 12% to 8.14 USD·gal^–1; the superior performance of repurposing is consistent across both feedstocks. Integration has a limited effect on SAF CI, which remains stable across scenarios, whereas using cellulosic feedstocks reduces CI by over 70% relative to corn, with baseline values of 17.01 g CO₂e·MJ^–1 for miscanthus and 12.23 g CO₂e·MJ^–1 for switchgrass. Global sensitivity analysis reveals MSP declines with greater coprocessing levels. Overall, this work demonstrates the potential of cellulosic ATJ coprocessing to enable cost-effective, low-carbon aviation fuels.

Introduction

Though essential for global connectivity, the aviation sector is responsible for approximately 11% of transport-related carbon dioxide (CO₂) emissions. When non-CO₂ greenhouse gas (GHG) emissions (such as contrails and nitrogen oxides) are included, aviation contributes around 1.7 Gt of CO₂ equivalents (CO₂e) assuming a 100 year time horizon. − With jet fuel demand expected to double by 2050 and triple by 2070, sustained efforts to decarbonize the aviation sector are critical to mitigating future climate impacts. − However, decarbonizing aviation remains particularly challenging as medium- to long-haul flights require high-density energy sources that cannot yet be directly supplied by electricity or hydrogen, necessitating the ongoing use of energy-dense liquid hydrocarbon fuels. ,,

Sustainable aviation fuel (SAF) is derived from waste streams or renewable feedstocksincluding first-generation sources (e.g., corn and sugar cane), second-generation sources (e.g., cellulosic biomass and agricultural residues) and e-fuels (electro-fuels)and can serve as a drop-in aviation fuel with a lower carbon footprint than traditional jet fuel. − Depending on the feedstock and production pathway, 100% SAF use can enable large reductions in life cycle carbon intensity (CI; e.g., 94% using municipal solid waste, 91% using agricultural residues). , To expedite SAF deployment, in 2021, the United States announced a target of producing 3 billion gallons of SAF annually by 2030. This initiative is supported by a diverse portfolio of SAF pathways, 11 of which have been approved by ASTM for use in aviation. , As of 2025, several regionally tailored SAF production projects are currently operational or under development in the United States, each leveraging locally available biomass sources. For example, SAF production from oilseeds via the hydroprocessed esters and fatty acids (HEFA) pathway is underway in Great Falls (Montana), while woody biomass-based facilities employing gasification and Fischer–Tropsch (FT) synthesis are being developed in Louisiana and East Texas. , In parallel, LanzaJet is now producing SAF from ethanol in Georgia, representing the world’s first commercial alcohol-to-jet (ATJ) facility.

Within the broader strategy of aligning SAF production with regional feedstock availability, ethanol emerges as a flexible, widely distributed SAF platform intermediate via the ATJ pathway, which is now being commercialized (notably at LanzaJet in Georgia). Ethanol is already produced commercially from a variety of feedstocks such as corn (Midwestern United States) and sugar cane (Brazil), while cellulosic-ethanol projects utilizing perennial cellulosic biomass in diverse U.S. regions remain under development. Among these, perennial biomass has been identified as a promising choice for bioenergy production due to its high yields, low input, and a capability to sequester soil organic carbon (SOC)a benefit attributed to extensive root and rhizome biomass, the absence of annual tillage, and the accumulation of litter such as fallen leaves. − Perennial crops are projected to contribute approximately 2.52 billion gallons and roughly 65% of total SAF production by 2030, through Fischer–Tropsch synthesis and ATJ pathways. Notably, the use of cellulosic feedstocks for ATJ is essential to meet emerging global policies, such as the EU’s Renewable Energy Directive II.

Cellulosic SAF remains capital-intensive, but coprocessing biobased intermediates at petroleum refineries offers a cost-effective strategy to enable large-scale deployment. Depending on the refinery configuration, the intermediate compound’s property, and the target products, potential insertion points include fluid catalytic cracking (FCC) units, diesel hydrotreaters, and hydrocrackers. , HEFA and FT coprocessing pathways have demonstrated technical feasibility and are approved under ASTM D1655 with coprocessing limits of 5%. ATJ coprocessing presents a complementary route. While it has not yet received regulatory approval, this reflects a policy gap rather than a technical barrier: ethanol-derived oligomers can be inserted into diesel hydrotreaters. This approach not only utilizes existing refinery assets but also benefits from compositional synergy, as the petroleum fraction supplies aromatics that complement the SAF blendstock. Previous studies have highlighted the potential of retrofitting idle or decommissioned refineries for SAF coprocessing, with estimated reductions in SAF selling price by 16% to 34%. , Distinct from reliance on shuttered or idled facilities and wholesale replacement of petroleum-based jet fuel production, this work advances a rapidly actionable, lower-disruption strategy that integrates SAF production into operating petroleum refinery units. Central to this approach are two parameters that regulate coprocessing degree: ethanol split ratio, defined as the fraction of ethanol routed to on-site upgrading (through coprocessing) versus for sale; and coprocessing ratio, defined as the volumetric fraction of SAF intermediates (i.e., oligomers) sent to insertion points (i.e., diesel hydrotreaters) relative to the combined throughput capacity of those units (Supporting Information eq S1). Together, these parameters govern coprocessing performance: the ethanol split ratio provides operational flexibility, thereby regulating the coprocessing level, while the coprocessing ratio reflects how this design choice translates into refinery-level outcomes and determines regulatory eligibility under coprocessing limits.

The objective of this work is to investigate the cost-reduction potential and GHG implications of integrating SAF production via the ATJ pathway within operating petroleum refineries. Using cellulosic feedstock (miscanthus and switchgrass) as illustrative examples, we examine three integration scenarios: (i) standalone as a benchmark; (ii) colocated, where the SAF biorefinery is situated adjacent to a petroleum refinery to share part of the outside battery limits (OSBL) infrastructure (e.g., wastewater treatment system, steam and power generation) and service facilities (e.g., buildings, land); and (iii) repurposing or coprocessing, where SAF is processed alongside the petroleum products, utilizing OSBL, inside battery limits (ISBL) units (e.g., ethanol fermentation, ethanol dehydration), and service facilities (Figure ). Leveraging BioSTEAM, an open-source platform in Python, we perform biorefinery design, simulation, TEA, and LCA in each scenario under uncertainty. − Overall, this work provides critical insights for prioritizing SAF adoption and supporting the aviation industry’s transition toward net-zero emissions.

1.

(A) Simplified block diagram of the designed SAF biorefinery system via the ATJ pathway, including (I) ethanol production and (II) ethanol upgrading. Part of the ethanol produced is sold as coproducts, while the remaining portion (defined as ethanol split ratio) undergoes further on-site upgrading processesdehydration, oligomerization, hydrogenation, and fractionationyielding SAF as the main product and gasoline and diesel as coproducts. In the repurposing scenario, oligomers are routed to the petroleum refinery for hydrogenation and distillation. (B) Schematic configuration of three integration scenarios for SAF integration within petroleum refineries: (i) the standalone scenario where a new SAF biorefinery is constructed independently; (ii) the colocated scenario where the SAF biorefinery is situated adjacent to an existing petroleum refinery to share part of the OSBL units and service facilities; (iii) the repurposing scenario where the SAF intermediates are coprocessed within the petroleum refinery, with the SAF biorefinery sharing ISBL, more OSBL units, and service facilities. OSBL units and service facilities are not extensively shown for clarity but detailed in Supporting Information Table S6.

Methods

System Design

The SAF biorefinery system was designed in two stages: ethanol production and upgrading via the ATJ pathway (Figure A). The baseline capacity of the biorefinery was 2000 dry metric tons of miscanthus (or switchgrass) per day. The process design of ethanol production was based on a previous study that evaluated energycane (0% triacylglycerides) as the feedstock, and was dedicated solely to ethanol production rather than biodiesel. Feedstock was assumed to have been preprocessed to meet key quality specifications, such as particle size, ash content, and moisture levels. Then it underwent steam pretreatment, simultaneous saccharification and cofermentation, and ethanol recovery. The molecular sieve was excluded from ethanol purification to reduce costs, as the subsequent dehydration process could handle the residue moisture.

A portion of the ethanol was diverted to the second stage for on-site upgrading to SAF (termed the ethanol split ratio), while the remainder was sold externally as coproducts to regulate the coprocessing level (the justification for this treatment is detailed in the subsequent section of integration scenarios). The upgrading stage consists of four steps: ethanol dehydration to ethylene, ethylene oligomerization to long-chain oligomers, hydrogenation of oligomers to liquid hydrocarbons, and fractionation of resulting hydrocarbons. , A two-stage oligomerization process was designed to enhance the yield of SAF-range hydrocarbons (C9 to C16). The biorefinery included two types of units: ISBL, which was involved in the ethanol production and upgrading process, and OSBL, which encompassed utilities such as steam and power generation (boiler turbogenerator or BT) and wastewater treatment (WWT). Details on process parameters can be found in Supporting Information Section S1.1.

Description of Petroleum Refinery

The design of the petroleum refinery for integration was based on a configuration reported in a previous study that adopted a heavy-coking refinery setup and provided the capacities of the refinery units (Supporting Information Table S1). It was assumed to have sufficient remaining service life to match the 30 year SAF project horizon, consistent with the typical 30 to 50 year operational lifespan of petroleum refineries. It was further assumed to be situated in the U.S. rainfed regions to capitalize on the regional abundance of cellulosic biomass, while also aligning with regional refinery capacity in the Gulf Coast and Midwest. , It had a crude distillation capacity of 120,000 barrels per day and accommodated an atmospheric distillation column to produce gas, naphtha, gas oil, and heavy bottoms. Naphtha and gas oil underwent further processing through hydrotreatment, hydrocracking, isomerization, and reforming to produce jet fuel, gasoline, and diesel. Heavy bottoms were processed through vacuum distillation, fluid catalytic cracking (FCC), coking, and hydrotreatment. ,,

Integration Scenarios

We evaluated three scenarios of increasing integration levels (Figure B). The standalone scenario serves as the benchmark, representing a greenfield SAF biorefinery requiring complete construction for ISBL, OSBL, and service facilities.

In the colocated scenario, the SAF biorefinery was adjacent to the petroleum refinery, sharing parts of its infrastructure while ensuring the petroleum refinery remained unaffected. This configuration reduced the costs from buildings, yard improvements, and service facilities compared to standalone construction (Supporting Information Table S6). ,

In contrast to the colocated scenario, a distinct feature of the repurposing scenario was the insertion of SAF intermediates into the petroleum refinery, replacing an equivalent volume of petro-derived stream for subsequent processing. In this work, diesel hydrotreaters were chosen as the insertion points due to their mild hydrogenation activity, which was compatible with the hydrogenation of oligomers from the ATJ process (further justification of the insertion point is provided in Supporting Information Section S1.2.2; process flow diagram for integration is provided in Supporting Information Figure S1). , Repurposing also allowed for a broader range of shared assets, such as on-site hydrogen production, steam and power generation, and wastewater treatment systems (Supporting Information Table S6). However, this required a capital cost of 88.3 million USD to expand WWT and BT capacity to accommodate SAF production (utility consumption comparisons between a petroleum refinery and SAF production are listed in Supporting Information Table S2; facility expansion cost estimates are detailed in Supporting Information Section S1.4.2). While repurposing enabled the sharing of facilities, we did account for the additional material inputs associated with SAF production (i.e., hydrogenation catalyst and natural gas required for hydrogen production). Additionally, the core management group and maintenance supervisor of the petroleum refinery could share their management and technical skills for SAF production (Supporting Information Table S7).

To quantify the extent of coprocessing in the repurposing scenario, we applied the concept of coprocessing ratio, as defined in ASTM Standard D1655-19, which was calculated in this work as the ratio of the volumetric flow of oligomers to the combined throughput capacity of diesel hydrotreaters (5595 m³·day^–1). The coprocessing ratio was approximately 10% at baseline feedstock flow; to align with the current regulatory limit of 5% and explore a potential higher limit, the coprocessing ratios in this work ranged from 4% to 15% (primarily determined by feedstock flow and ethanol split ratio). Introducing oligomers, on the other hand, displaced the space that would otherwise be occupied by crude intermediates, thereby reducing petroleum product yield. Therefore, we included the opportunity cost associated with this replacement in the annual operating cost (AOC) in the repurposing scenario (Supporting Information Section S1.4.4).

Process Model

Open-source process models for all scenarios were developed in BioSTEAM to dynamically connect system design with predictions of biorefinery performance. , After each set of decision variables, technological parameters, and contextual parameters were defined, BioSTEAM was used to simulate the mass and energy flows through the biorefinery until reaching convergence. The influent and effluent streams from each unit were linked to cost algorithms that determined the unit sizing and cost. Baseline values and parameter distributions for this system are provided in Supporting Information Table S9. Design specifications of the major units and equipment are outlined in Supporting Information Table S10. All Python scripts for biorefinery setup, simulation, and evaluation are available online.

Techno-Economic Analysis (TEA) and Life Cycle Assessment (LCA)

We performed TEA and LCA for the biorefinery system across all scenarios in BioSTEAM, focusing exclusively on the bioderived SAF stream. The functional unit for TEA was 1 gallon of SAF to align with industrial practice. Material and energy data from system simulations were used to estimate capital and operating costs (TEA framework is shown in Supporting Information Figure S2; cost item inputs for each scenario are listed in Table S6). Standard nth-plant assumptions were applied, indicating that the technology is mature and that multiple biorefinery plants using the same technology have been developed and are operating. A discounted cash flow rate of return analysis was used to calculate the net present value (NPV), and the minimum selling price (MSP) of SAF was determined as the price required to achieve an NPV of zero at a 10% internal rate of return (IRR). MSP was calculated at the refinery gate, assuming ethanol (not upgraded on-site) and gasoline and diesel were sold as coproducts at historical market prices. Feedstock was priced at its delivered cost, including the breakeven price at the farm gate (Supporting Information Table S3) and round-trip transportation costs. Transportation distance was estimated from 10,000 farmnearest petroleum refinery pairs (Supporting Information Section 1.3). All costs and prices are presented in 2023 U.S. dollars (USD).

We conducted LCA by leveraging simulated streams (inputs and emissions) and utility inventories. The system boundary spanned cradle-to-grave processes for all inputs and utilities, including feedstock farming, harvesting, transportation, fuel production, and combustion phase (Supporting Information Figure S4 and Table S3). The feedstock itself was considered carbon-neutral, indicating the biogenic carbon taken up during plant growth equals the biogenic carbon released from the biorefinery and fuel combustion. The farming and harvesting stages accounted for SOC change, emissions from nitrogen, phosphorus, potassium fertilizers, and chemicals (i.e., herbicides), including upstream production and nitrogen-fertilizer-induced nitrous oxide emissions through soil microbial processes. , Notably, the substantial negative contribution from SOC sequestration was the primary factor lowering the feedstock CI (Supporting Information Table S4). , Feedstock transportation emissions were estimated based on the transport distance (Supporting Information Section 1.3), and a one-way trip was counted to be consistent with other studies. − The functional unit was defined as 1 megajoule (MJ) of SAF on a lower heating value basis. Two allocation methods were applied to evaluate coproduct impacts: energy allocation and displacement (or system expansion), the latter of which aligns with ISO 14040/14044 , and biofuel policies such as the Renewable Fuel Standard. Environmental impacts were characterized by CI, which was quantified as the global warming potential over 100 years (GWP₁₀₀) using characterization factors from the Intergovernmental Panel on Climate Change (IPCC). Life cycle inventory data were primarily gathered from ecoinvent v3.8 and GREET 2020 (Supporting Information Table S8). ,

Uncertainty and Sensitivity Analysis

To quantify model uncertainties arising from the variability of input parameters, such as technological parameters, material cost, and life cycle inventory data, we performed uncertainty analyses for each scenario using Monte Carlo simulation with Latin Hypercube Sampling (2000 simulations) for 49 uncertain parameters (Supporting Information Table S9). Price distributions for coproducts, natural gas, and electricity were modeled from historical trends provided by the U.S. Energy Information Administration. , Parameter ranges for pretreatment, fermentation, and upgrading processes were supported by techno-economic studies, reports, and patents. ,− Plant capacity, on-stream days, and IRR followed previous studies. , For parameters lacking strong literature support, uniform distributions were applied.

The sensitivity of MSP and CI to all uncertain inputs was determined via Spearman’s rank order correlation coefficients (Spearman’s ρ) using Monte Carlo simulation results. The parameters to which results were most sensitive (p-value <0.05 and absolute value of Spearman’s ρ ≥ 0.1) were also identified for each scenario.

Results and Discussion

Material and Energy Flows

First, it should be noted again that all results in this work focus exclusively on green stream (SAF products) and exclude petroleum-derived products. We examined the baseline carbon and energy flows for SAF production to assess process efficiency, normalizing all flows to 100% of the input feedstock. Under baseline assumptions, the SAF biorefinery processes 2000 dry metric tons of feedstock daily. Results showed in the ethanol production process, 35.7% of the carbon and 52.2% of the energy in feed miscanthus were converted to ethanol, yielding 203 g·kg^–1 dry miscanthus (Figure A,B), which fell within the reported experimental range of 185 to 253 g·kg^–1. Notably, miscanthus-derived ethanol achieved a higher yield (0.44 L·kg^–1 dry miscanthus) than reported cellulosic ethanol conversion (0.38 L·kg^–1 dry biomass) due to the higher fermentable sugar content in miscanthus. In contrast, switchgrass produced a lower ethanol yield of 0.38 L·kg^–1 dry biomass and thus resulted in a reduced SAF yield. A substantial portion of carbon and energy remained in the yeast residue after fermentation, which was subsequently diverted to the boiler for energy recovery. At baseline, 90% of the ethanol produced was assumed to be upgraded on-site (i.e., the ethanol split ratio of 90%), with the remaining 10% sold to other SAF producers as a coproduct. The final fuel products, obtained through oligomerization, hydrogenation, and fractionation, included SAF as the main product with an annual yield of 30 million gallons and gasoline and diesel as coproducts. SAF accounted for 23.6% of the original carbon and 35.3% of the chemical energy of feedstock, constituting 74.7% of the energy and 74.6% of the volume within the fuel products. This process yielded 0.44 gallons of SAF per gallon of ethanol, consistent with simulation inventories reported in a previous study.

2.

Sankey diagrams of (A) carbon mass flow and (B) energy flow through each process at baseline (i.e., 10% of ethanol is sold as a coproduct). All flows are scaled to 100% of the input feedstock. The wastewater treatment system (WWT) and boiler turbogenerator (BT) are not shown for clarity. A detailed carbon analysis that considers all biogenic carbon uptake and emissions in different chemical forms is presented in Supporting Information Figure S5.

Cost Savings Potential for Integration

All scenarios utilizing miscanthus as feedstock resulted in MSPs (not accounting for financial incentives for low-CI fuels) higher than the U.S. jet fuel wholesale price of 2.12 to 2.48 USD·gal^–1 (with the crude oil cost of 80 USD·barrel^–1), regarded as the refinery gate price (Figure A). On the other hand, these MSPs remained competitive with current cellulosic ATJ technology, not exceeding the reported production cost range of 7.08 to 11.82 USD·gal^–1 (at 0% IRR). Notably, the MSP values reported in this work represented a more conservative cost estimate than literature production costs, as they explicitly accounted for the biomass logistics costs and incorporated a 10% IRR to capture investment returns. For miscanthus, logistics contributed 21.69% [19.13% to 25.03%] of the delivered feedstock cost (Supporting Information Figure S7B). In the standalone scenario, the MSP was estimated to be 9.26 USD·gal^–1 at baseline (equivalent to a crude oil cost of 431 USD·barrel^–1; Supporting Information Section 1.8), with a range of 8.50 to 11.06 USD·gal^–1 [5th to 95th percentiles are presented in brackets hereinafter]. When integration strategies were implemented, the cost advantage was further enhanced. In the colocated scenario, the MSP of 8.97 [8.22 to 10.71] USD·gal^–1 (equivalent crude oil cost of 416 USD·barrel^–1) was lower than in the standalone scenario across all simulations, with a reduction of [0.29 to 0.41] USD·gal^–1. The repurposing scenario achieved the most substantial cost reduction of 1.12 [0.97 to 1.96] USD·gal^–1 relative to the standalone scenario, with the MSP of 8.14 [7.45 to 9.60] USD·gal^–1 (equivalent crude oil cost of 374 USD·barrel^–1). Comparable outcomes were observed for switchgrass-derived scenarios (Supporting Information Figure S10A).

3.

Cost performance of standalone, colocated, and repurposing scenarios from 2000 Monte Carlo simulations. (A) Box and whiskers plot of MSP. Results are presented with baseline values (diamonds), median values (solid lines), 25th to 75th percentiles (shaded region), and 5th to 95th percentiles (whiskers). U.S. jet fuel wholesale price by refiners is considered refinery gate price, aligning with the definition of MSP in this work. All prices and costs have been adjusted to 2023 dollars for consistency. (B) Stacked bar plot of installed equipment cost for inside-battery limit (ISBL) units and outside-battery limit (OSBL) units. Results are shown with mean values (solid lines) and standard deviation (error bars).

Capital expenditure constitutes a significant portion of the total production cost and is directly influenced by the level of integration between two facilities. The repurposing scenario achieved the most substantial reduction in installed equipment cost, with a 47% decrease, averaging 215.1 million USD compared to 404.9 million USD in the standalone scenario (Figure B). The most significant cost savings were attributed to OSBL expenditures, primarily due to the shared use of the wastewater treatment system and steam and power generation, the most capital-intensive components with baseline installed costs of 72 million and 111.7 million USD, respectively (Supporting Information Figure S7A and Table S11). Additionally, sharing process units, including diesel hydrotreaters and distillation columns, reduced ISBL costs by approximately 6 million USD, bringing them down from 217 million to 210.7 million USD. In contrast, the colocated scenario achieved only a modest reduction in OSBL costs, resulting from sharing storage units, with a decrease in total installed equipment cost of less than 1.1%. For switchgrass-derived SAF production, the total installed equipment cost (averaging 395.5 million USD in the standalone scenario) was lower than that of miscanthus, as the lower fuel conversion required a smaller capacity. However, the relative cost reductions in the colocated and repurposing scenarios followed trends similar to those for miscanthus (Supporting Information Figure S10B).

Total capital investment (TCI, estimated using ISBL and OSBL costs) and annual operating costs (AOC) represent the two key cost categories for facilities. Compared to the TCI of 879.9 million USD [649.5 million to 1068.2 million USD] in the standalone scenario, the repurposing scenario, despite requiring an additional capital cost of 88.3 million USD for expansion of WWT and steam and power generation, showed a significant reduction of 36.2% at baseline (Supporting Information Figure S7A), with TCI of 561 million USD [437.5 million to 662.1 million USD]. This reduction was primarily due to sharing OSBL units, service facilities, land, buildings, and yard work (Supporting Information Table S11). The colocated scenario reduced the TCI by only 7.9% at baseline, as the cost savings were limited to buildings, yard work, the electric substation, distribution, and product storage.

AOC, which includes fixed operating costs (FOC, covering insurance, maintenance, and labor costs) and variable operating costs (VOC, covering material and utility costs), remained unchanged in the colocated scenario compared to the standalone, as colocation did not affect unit operations or material flows in SAF production. Across all scenarios, the largest contribution to AOC was associated with feedstock (including transport cost) and hydrolysis enzymes, accounting for 56.8% [52.1% to 64.4%] and 23.5% [16.9% to 31.0%] of total material costs, respectively (Supporting Information Figure S7B). The former was driven by the midsize ethanol facility with a yield of 75.8 [48.7 to 98.9] million gal·yr^–1; the latter was driven by the high price of hydrolysis enzymes (8.03 USD·kg^–1), and the parameter change of the enzyme loading caused the broad range. However, three notable differences emerge in the repurposing scenario. First, the repurposing scenario avoided the 1.69 million USD annual cost of purchasing hydrogen but instead incurred 1.19 million USD·yr^–1 for natural gas to produce hydrogen in situ via the SMR unit. Second, labor costs decreased from 4.53 million to 4.12 million USD due to the shared operation personnel, including the core management team and maintenance supervisor, with the petroleum refinery. Third, the opportunity cost was introduced as a VOC term to reflect the profit reduction of the petroleum refinery caused by the decreased product yield due to the intervention. The unit opportunity cost was quantified by the foregone profit per volume of petroleum products (Supporting Information Section S1.4.4). This additional cost resulted in a 7.3% increase in AOC compared to the standalone scenario.

Environmental Implications of Integration

The CI of SAF was highly consistent across all scenarios, and only the standalone scenario results were presented in the main text (full results for miscanthus as feedstock are detailed in Supporting Information Figure S8 and Table S12, and results for switchgrass as feedstock are detailed in Supporting Information Figure S11 and Table S13). The only distinction across scenarios that would drive operational GHG emissions was the hydrogen source: purchased hydrogen for the standalone and colocated, versus on-site generation via natural gas SMR for repurposing. This variation in hydrogen sourcing and associated natural gas had only a minor effect on SAF CI (Supporting Information Figure S9). Overall, SAF CI was found to be significantly lower than the global jet fuel CI benchmark of 89 g CO₂e·MJ^–1 (with the reduction of 80.9% [73.1% to 97.7%] using the displacement method) and below the threshold for qualifying SAF for tax credits, which mandates a 50% reduction from the jet benchmark (Figure A; switchgrass-derived SAF exhibited a similar trend shown in Figure S12A). For comparison, corn-derived SAF via the ATJ pathway had the CI benchmark of 67.2 g CO₂e·MJ^–1 (using energy allocation). The lower CI in this work was primarily due to the use of cellulosic biomass as a feedstock, which yielded negative feedstock CIs (baseline values of −139 kg CO₂e·dry metric ton^–1 for miscanthus and −58.73 kg CO₂e·dry metric ton^–1 for switchgrass) through SOC sequestration. The displacement method mostly resulted in the lower CI values of 17.01 [2.06 to 23.93] g CO₂e·MJ^–1 than 17.48 [7.91 to 21.94] g CO₂e·MJ^–1 using the energy allocation method. This difference between the calculated CI via energy allocation vs the displacement method arose from fuel coproducts, including ethanol for sale, and final diesel and gasoline outputs (which altogether resulted in a CI benefit of −8.97 [−12.06 to −6.05] g CO₂e·MJ^–1).

4.

Environmental performance of standalone and colocated scenarios (repurposing exhibits highly similar patterns and is not displayed separately, with detailed CI results provided in Supporting Information Figure S8 and Table S12). (A) Carbon intensity (CI) results from 2000 Monte Carlo simulations using displacement and energy allocation methods. Results are presented with baseline values (diamond), median values (solid lines), 25th to 75th percentiles (shaded region), and 5th to 95th percentiles (whiskers). (B) Stacked bar plot of CI breakdown under the displacement method at baseline. Coproducts (i.e., ethanol for sale, gasoline, and diesel) provide carbon credits and feedstock (i.e., miscanthus and switchgrass) production was carbon negative due to changes in soil organic carbon. Heating, cooling, and electricity demand constitute steam and electricity demand met by the boiler turbogenerator, with the total emission equaling natural gas input and nonbiogenic emissions from the boiler turbogenerator. Other materials referred to material inputs such as enzyme, diammonium phosphate, corn steep liquor, and hydrogen. Waste emissions, including the disposal of spent catalysts, were not shown as they contributed less than 1% of emissions.

Examining the CI breakdown under the displacement method, other materials acquisition (e.g., caustic, enzyme, corn steep liquor, catalysts) dominated across the three scenarios, contributing 41.9% [36.2% to 56.3%] of positive emissions, primarily from the acquisition of caustic (29.8% [26% to 39.7%]), which was used in wastewater treatment to neutralize acidity during the aerobic digestion process (Figure B). The biorefinery’s heating demand was the second largest contributor (28.3% [11.7% to 34.1%]) to positive emissions, from the associated acquisition (4.6% [1.6% to 6.7%]) and combustion (23.7% [10.1% to 27.4%]; we assumed complete oxidation to CO₂) of natural gas in the boiler. In contrast, the use of cellulosic biomass provided carbon credits, offsetting 31.6% [26.9% to 46.7%] of the emissions, with biomass logistics then accounting for 4.2% [3.8% to 5.9%] of the total positive emissions. Fuel coproducts also played a role under the displacement method, offering an additional offset of 17.6% [12.7% to 26.3%] of the emissions. Other process inputs (i.e., enzyme, diammonium phosphate, corn steep liquor, catalysts, and hydrogen) did not contribute substantially to CI (<15% of total contributions). Notably, although switchgrass exhibited overall higher CIs than miscanthus, SAF derived from switchgrass achieved lower CI scores. This outcome was driven by the feedstock composition: switchgrass resulted in lower fuel yields (26.55 million gallons of SAF per year at baseline compared with 30.42 million gallons using miscanthus), which subsequently resulted in more process waste being routed to the boiler and more on-site energy production, thereby reducing natural gas consumption to satisfy the system’s steam and power demands (Supporting Information Figure S12B).

Drivers of Economic and Environmental Performance

In the repurposing scenario, MSP was found to be most sensitive to hydrolysis enzyme loading with Spearman’s ρ of 0.51, feedstock price with Spearman’s ρ of 0.46, feedstock flow with Spearman’s ρ of −0.35 (global sensitivity analysis results are provided in Supporting Information Figure S9). Similar effects were shown in the switchgrass-derived SAF system (Supporting Information Figure S13). For both feedstocks, the feedstock flow and the split of ethanol for upgrading exhibited negative correlations with MSP, influencing the SAF yield. These two parameters directly determined the coprocessing ratio, i.e., the volumetric flow of oligomers to the total throughput capacity of the diesel hydrotreaters which was assumed to remain unchanged in the repurposed refinery. The opportunity cost, reflecting foregone profits associated with the SAF stream, positively correlated with MSP (Spearman’s ρ of 0.2). While hydrolysis enzyme loading and reaction conversion showed strong impacts on MSP, they were not readily improved in practice due to current technical constraints, highlighting the degree of coprocessing as a more tractable strategy for reducing SAF production costs.

Across the sampling space, the coprocessing ratio increased from 0.05 to 0.15, corresponding to a SAF annual production range of 19 to 39 million gallons, resulting in a decrease in MSP from 9.45 USD·gal^–1 to 7.30 USD·gal^–1 (50th percentiles; Figure A). This decline was because the cost savings from reduced infrastructure investments outweighed the opportunity costs, and this benefit was enhanced at higher coprocessing ratios. However, while the overall MSP trended downward for the median MSP values, slight fluctuations were observed in other percentiles because the coprocessing ratio was obtained as a model output, reflecting incomplete coverage of parameter combinations at specific ratios within the sampling space. The same trend was observed with switchgrass as the feedstock, but it yielded a lower coprocessing ratio of 0.04 to 0.13 with the SAF annual yield ranging from 17 to 34 million gallons (Figure B).

5.

Minimum fuel selling price (MSP) as a function of the coprocessing ratio in the repurposing scenario using (A) miscanthus and (B) switchgrass as the feedstock. The solid line indicates the 50th percentile of the MSP, the shaded area indicates the 25th to 75th percentiles, and the dotted lines indicate the 5th and 95th percentiles.

Spearman’s ρ between the input parameters and CI (displacement method) were similar across the three scenarios (global sensitivity analysis results for miscanthus as feedstock provided in Supporting Information Figures S9). For miscanthus-derived SAF system, boiler efficiencylargely fixed by boiler design and difficult to improvewas the most influential parameter with a Spearman’s ρ of −0.62, which stemmed from the energy-intensity of the ATJ pathway (heat utility demand of 133 MJ·gal^–1 SAF; Supporting Information Figure S6) and the impact of efficiency on the amount of natural gas required for steam and power generation. The ethanol split ratio for upgrading had the second largest magnitude rank correlation coefficient (Spearman’s ρ = 0.39) because it affected the carbon credits from ethanol sales. The feedstock CI was also significant (Spearman’s ρ = 0.21) due to its substantial volumetric flow and considerable unit impact. Switchgrass-derived SAF CI showed similar dominant parameters (Supporting Information Figure S13). Examining the impacts of the ethanol split ratio and miscanthus CI, all resulting CIs were well below the 50% reduction threshold from jet fuel baseline (89 g CO₂e·MJ^–1; >50% reduction requiring <44.5 g CO₂e·MJ^–1) under two LCA accounting methods (Figure ; Supporting Information Figure S14 for switchgrass-derived SAF). Although lower feedstock CIs and ethanol splits both led to lower SAF CIs, the displacement method resulted in even lower SAF CIs than the energy allocation method when feedstock CI was low and a greater portion of ethanol was sold rather than upgraded to SAF on-site. For instance, at the ethanol split of 0.8, the displacement method yielded −0.17 to 11.33 g CO₂e·MJ^–1 across miscanthus CIs of −181.68 to −103.21 kg CO₂e·dry metric ton^–1 (using 25th to 75th percentiles from Supporting Information Table S3), compared to that of 7.85 to 14.75 g CO₂e·MJ^–1 using the energy allocation method. However, this difference narrowed and even reversed when less ethanol was credited (i.e., ethanol split ratio was increased to yield less ethanol coproduct) and feedstock CI increased, indicating that higher coproduct fuel yields (>25% of total energy exported) would lead to a greater reduction in SAF CI under energy allocation than credit by the displacement method. When 100% of ethanol was upgraded on-site, the displacement could yield SAF CI of 13.35 to 24.81 g CO₂e·MJ^–1 (gasoline and diesel credit of 4.7 g CO₂e·MJ^–1) whereas the energy allocation led to 13.45 to 22.01 g CO₂e·MJ^–1.

6.

Contour plots of carbon intensity (CI) of SAF under the repurposing scenario at varying miscanthus CI and ethanol split ratios to upgrading by using (A) displacement method and (B) energy allocation method.

Prioritizing Paths Forward for a Low-Carbon Aviation Future

This work evaluates the economic and environmental implications of integrating SAF production utilizing two perennial cellulosic feedstocks within established petroleum refineries. By leveraging shared OSBL units and process units, the repurposing scenario achieves a substantial 36% reduction in capital costs relative to greenfield construction of separate ethanol biorefineries and ATJ facilities, yielding the lowest MSP of 8.14 [7.45 to 9.60] USD·gal^–1 (reduced by 12.10% [10.58% to 18.28%]). This trend is also applied to switchgrass as the feedstock. From an environmental perspective, however, integration strategies have a limited effect on SAF CI, which remains stable across scenarios. Importantly, using cellulosic biomass reduces SAF CI by more than 70% compared with corn, with baseline values of 17.01 g CO₂e·MJ^–1 for miscanthus and 12.23 g CO₂e·MJ^–1 for switchgrass (using the displacement method). This remarkably low carbon intensity underscores the advantage of perennial feedstocks for sustainable biofuel production.

Blending conventional jet fuel with SAF offers a critical near-term decarbonization strategy, enabling gradual adoption while requiring minimal or no modification to existing aircraft or fueling infrastructure. However, this approach introduces a trade-off between CI and MSP, as deeper emission reductions (achieved through higher SAF blend ratios) are associated with increased production costs (Figure ; parallel results for switchgrass as feedstock are shown in Supporting Information Figure S15). Across the three integration levels, the standalone incurs the highest cost penalty for decarbonization (approximately 0.09 USD per g CO₂e·MJ^–1 reduction), whereas the repurposing achieves the least cost penalty for approximately 0.07 USD per g CO₂e·MJ^–1 reduction. These findings underscore the potential of refinery repurposing as the most cost-effective and scalable strategy to accelerate SAF deployment and support the transition to a low-carbon aviation future.

7.

Estimated economic and environmental outcomes of blending jet fuel with SAF produced under standalone (red), colocated (blue), and repurposing (green) scenarios (using miscanthus as feedstock). The gray rectangle indicates the jet fuel benchmark with MSP of 2.12 to 2.48 USD·gal^–1 and CI of 84 to 89.5 g CO₂e·MJ^–1. Shading in the kernel density plots for 100% and 50% SAF represents the density distribution from 2000 Monte Carlo simulations (darker regions have higher density of results). The kernel density plots for 50% SAF (the current blending limit approved for the ATJ pathway) were generated using Latin hypercube sampling from SAF and jet fuel ranges. Box and whisker plots represent the same data as the kernel density plots and show median, 25th/75th, 10th/90th percentiles with the center line, bottom/top of the box, and lower/upper whiskers, respectively.

Ultimately, although ethanol is an intermediate for the ATJ pathway, most commercial production still relies on starch-based corn using relatively simple conversion processes. In contrast, cellulosic feedstocks require intensive pretreatment and enzyme hydrolysis to release fermentable sugarscapabilities that conventional dry mill ethanol plants lack. Realizing the full potential of cellulosic ATJ SAF will therefore depend not only on downstream integration with petroleum refineries, but also on coordinated upstream transitions in ethanol production, supported by targeted policies and sustained investment. In parallel, technical limitations associated with the ATJ pathway must also be addressed. Current ATJ technologies convert ethanol into only two of the four hydrocarbon classes required for fully infrastructure-compatible, 100% drop-in SAFspecifically lacking aromatics and cycloparaffins. As a result, standalone ATJ SAF necessitates additional blending, storage, and certification steps, increasing cost and complexity. This challenge is compounded by a regulatory hurdle: unlike standalone ATJ SAF, which is certificated under ASTM D7566, coprocessed fuels intended for use as jet fuel must meet ASTM D1655 specifications. To date, no commercial efforts have been made to certify ATJ coprocessing under D1655. Resolving this certification challenge could significantly expand deployment opportunities and reduce costs by enabling a broader integration of SAF into existing refinery infrastructure. Additionally, tracking the renewable fraction is essential for CI determination and regulatory crediting, but it is technically challenging. Current approaches (e.g., energy/mass balance or yield allocation) are sensitive to feedstock variability and refinery operating conditions. Radiocarbon analysis offers a direct means of quantifying biogenic carbon, yet its accuracy diminishes at low coprocessing levels. , Hybrid frameworks that pair routine operational accounting with periodic carbon-tracking are therefore expected to provide more verifiable attribution of the renewable fraction. Finally, while the integration scenarios capture cost savings from shared refinery infrastructure, the financial benefits of process integration reported in this work are likely to be underestimated, as the cost estimates for retrofits (which were represented as additional capital costs) were conservative and likely exceed the actual requirements for facility expansions. A more detailed cost estimation of such expansions and upgrades remains an important area for future work.

All data that support the findings of this study are available in the Supporting Information, and Python scripts for biorefinery models are publicly available on GitHub. The source data underlying figures can be provided by the corresponding author upon request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.5c17540.

• Process design and parameter assumptions for SAF production; miscanthus and switchgrass cost and carbon intensity; TEA structure and results; LCA scope and results; sensitivity analyses results; parallel results for switchgrass as an additional feedstock (PDF)

J.S.G., W.G. conceived of the research. W.G. built the models with input from L.P.K., S.S.B. and M.F.B., and W.G. performed the analyses. W.G. and J.S.G. interpreted the results and wrote the paper. All authors edited the paper.

The authors declare no competing financial interest.