Characterizing the Potential for Sustainable Azelaic Acid Production from High-Oleic Vegetable Oil Using Two-Step Oxidative Cleavage

Abstract

Azelaic acid is a renewable monomer conventionally produced via the energy-intensive ozonolysis of oleic acid. Recent advancements have enabled the use of high-oleic vegetable oils (rather than tallow-derived oleic acid) and replaced ozonolysis with two-step oxidative cleavage using hydrogen and oxygen. Although this shift would improve process safety, the financial viability and environmental implications remain uncertain. In this study, we characterized the sustainability of azelaic acid production from high-oleic vegetable oil using two-step oxidative cleavage. Process design, simulation, technoeconomic analysis (TEA), and life cycle assessment (LCA) were executed under uncertainty using BioSTEAM. The modeled system produces azelaic acid at a market-competitive minimum selling price (MSP) of 8.32 [4.93–13.34] $ kg^–1 (median 5th–95th percentiles), below the minimum estimated market price of 9.93 $ kg^–1. Further, it has the potential to approach carbon neutrality (0.0 [−5.5 to 5.6] kg of CO₂-eq kg^–1) under displacement allocation. Improvements to dihydroxylation (86 to 99%) and oxidative cleavage conversions (93 to 99%) would reduce MSP to $5.24 kg^–1 and carbon intensity to −1.90 kg of CO₂-eq kg^–1 (displacement). Additionally, increasing the feedstock triolein content (75 to 85%) lowers MSP by $0.82 kg^–1. Overall, this research demonstrates the potential for financially viable production of azelaic acid from vegetable oils and the utility of agile TEA/LCA.

Introduction

Azelaic acid, a C9 dicarboxylic acid derived from oleic acid, is used in lubricants, polymeric materials, and skincare products. The global azelaic acid market is projected to expand from $219 million in 2022 to $422 million in 2032 with a compound annual growth rate (CAGR) of 7.0%. In 2022, North America held 39% of the market share valued at $84 million, but capitalizing on this growth requires safer and more sustainable alternatives to the conventional ozonolysis process. , Although ozone effectively cleaves oleic acid into azelaic and pelargonic acid, ozonolysis poses safety risks and requires capital-intensive on-site ozone production. − Some laboratory-scale studies have investigated enzymatic, chemoenzymatic, and catalytic pathways to substitute ozone with more benign oxidants to enhance process safety, − but the financial viability and environmental implications of such alternatives remain unclear. Despite prior efforts to evaluate economic feasibility, ,, the environmental sustainability of azelaic acid production remains unexplored, underscoring the critical need for a comprehensive life cycle assessment.

In parallel with the process safety challenges of ozonolysis, limited access to affordable oleic acid hinders the sustainable growth of azelaic acid production. Tallow fats (∼45 wt % oleic acid) were once a low-cost oleic acid source in the United States. However, its diversion toward biofuel production and supply inelasticity has increased price and constrained availability for specialty oleochemical production like azelaic acid. − Vegetable oils have emerged as an alternative for oleic acid production, with the added benefit of reducing production costs from over 2280 $ ton^–1 for tallow-derived oleic acid to 980–1020 $ ton^–1 for vegetable oil-derived oleic acid (prices indexed to 2023 dollars using Table S7, Supporting Information A (SI-A): Additional process design details, methods, and results). High-oleic varieties of vegetable oil, with a minimum of 75 wt % oleic acid, have particularly strong potential to serve as an economically favorable substitute for tallow-based oleic acid because they can be used for azelaic acid production without further purification. − High-oleic soybean, sunflower, and canola oils are among the primary sources of high-oleic vegetable oils. ,, Additionally, continued progress in engineering oil accumulation in crops such as sugarcane and sorghum could enable the use of oilcane and oil sorghum as future feedstocks for azelaic acid production. , However, further research is needed to evaluate how specific triacylglyceride (TAG) compositions in these engineered feedstocks may affect downstream separation and the efficiency of azelaic acid recovery.

A commercial process developed by Novamont, a European bioproduct company, utilizes hydrogen peroxide and air, in a two-step oxidative cleavage (TSOC) process to convert unsaturated fatty acid compounds in high-oleic vegetable oils, including oleic acid, linoleic acid, and linolenic acid, into azelaic acid and several coproducts. − One published study demonstrated that high-oleic sunflower oil and palm oil feedstocks have the potential to be financially viable, but this work relied on a single set of assumptions for mass and energy balances and did not mechanistically link feedstock properties and process assumptions to the complex reaction pathways involved in the TSOC conversion process and downstream separation. This past work also assessed the sensitivity of net present value (NPV) using a correlation matrix informed by expert judgment, but it focused primarily on major cost and price variables without addressing the complex interdependencies among technological parameters, contextual parameters, and decision variables in process design. While this study offers valuable preliminary insight into the economic feasibility of azelaic acid production, a comprehensive understanding of sustainability drivers requires the integration of process design, simulation, and sustainability assessments within uncertainty and global sensitivity analysis frameworks. This integration would enable a systematic evaluation of the impact of individual decisions (e.g., reaction mechanisms and catalyst recovery system design) and sources of uncertainty (e.g., catalyst longevity, feedstock triacylglycerol (TAG) composition, and unit costs of materials) on key sustainability indicators such as the minimum selling price (MSP) and life cycle carbon intensity (CI) of azelaic acid.

The objective of this work was to characterize the cost and life cycle CI of industrial-grade (90 wt %) azelaic acid production from high-oleic soybean oil (HOSoy) using the TSOC method and to elucidate key drivers and sources of uncertainty governing system performance. To this end, we coded all relevant unit operations using BioSTEAM, an open-source Python-based software package enabling biorefinery design, simulation, and evaluation under uncertainty, including via technoeconomic analysis (TEA) and life cycle assessment (LCA). We model biodiesel production from HOSoy oil as a prerequisite step to convert TAGs into methyl esters, enabling the use of HOSoy oil as a compatible feedstock for the patented TSOC process, which operates on methyl esters rather than on TAGs. The resulting methyl esters are subsequently valorized into azelaic acid and coproducts as described in the TSOC process. Additionally, we build upon prior literature describing the recovery and purification of azelaic acid, coproducts, and catalysts to model downstream separation and purification. − Using this integrated framework, we developed a complete biorefinery design, available in the BioIndustrial Park (BIP) repository (installation guidelines available in Section S6 of SI-A), to produce azelaic acid from HOSoy oil. The design includes two-step oxidative cleavage including dihydroxylation of unsaturated esters present in biodiesel (namely, oleic acid, linoleic acid, and linolenic acid), recovery of cobalt acetate and tungstic acid catalysts, and separation and purification of azelaic acid and coproducts such as glycerol, C5–C9 monocarboxylic acids (MCA) (including methyl–oxo nonanoic acid), pelargonic acid, methanol, and a fatty acid blend. To characterize the impacts of vegetable oil composition on the MSP of azelaic acid, the biorefinery was simulated across a spectrum of TAG compositions based on data from the food chemical codex (FCC), varying triolein, trilinolenin, trilinolein, tristearin, and tripalmitin concentrations. Overall, the conclusions from this study represent the first (to the authors’ knowledge) CI characterization of bioderived azelaic acid. The findings offer insights into critical drivers of system performance and demonstrate a pathway for developing sustainable azelaic acid production from a broader portfolio of high-oleic-content vegetable oils and agricultural feedstocks.

Methods

System Description

Biorefinery Process Data: Mass Balance, Energy Balance, and Code Availability

Information on reaction chemistry, pathways, and system design is available in Section S1 of SI-A. The baseline system report (Excel file provided in the folder Supporting Information B (SI-B): System Design Report and Uncertainty & Sensitivity Analysis Results) includes system-wide diagrams, mass balance stream table, itemized costs and utility requirements, design requirements, and stoichiometric reaction conversions. Python scripts for system design, TEA, LCA, global sensitivity analysis, and uncertainty evaluation are available on BioIndustrial Park (folder name oleochemicals), a GitHub repository for biorefineries modeled using BioSTEAM (version 2.38.3; installation guidance in Section S6). BioSTEAM solves iteratively to converge the mass and energy balance of the entire biorefinery below a specified tolerance. This study used a relative molar tolerance of 0.01% and an absolute molar tolerance of 0.0001 km h^–1.

Reaction Pathways for Production of Monomethyl Azelate and Other Coproducts

In this study, the biorefinery is designed to utilize the TSOC process to meet an annual azelaic acid production capacity of 11,000 MT (biorefinery capacity estimation in Section S2.4). HOSoy oil (baseline with a Plenish variety; Section S1.1) is converted into biodiesel and glycerol (simplified process diagram of biodiesel production illustrated in Figure S1; detailed process description provided in Section S1.2). The two-step oxidative cleavage pathway begins with dihydroxylation of the unsaturated esters to form vicinal diols, followed by oxidative cleavage of these diols, along with other side reactions, to yield azelaic acid and coproducts. Methyl oleate, methyl linoleate, and methyl linolenate esters in biodiesel are dihydroxylated using tungstic acid and hydrogen peroxide to yield diols. These diols are oxidatively cleaved using air and a cobalt acetate catalyst. During this conversion process, the methyl oleate diol produces nonanal and methyl oxo-nonanoic intermediates and also undergoes esterification with azelaic acid and pelargonic acid to produce several mono- and di-mixed esters. Methyl oxo-nonanoic acid and nonanal intermediates are further oxidized into monomethyl azelate and pelargonic acid, respectively. In another parallel reaction, methyl oleate diol is decarboxylated to produce monomethyl suberate and caprylic acid (see the reaction scheme in Figure ; detailed process description in Section S1.3).

1.

Simplified reaction scheme for the production of azelaic acid, pelargonic acid, and other coproducts from triolein. Balanced chemical equations and molar conversions across all of the conversion processes for all triacylglycerides (TAGs) including triolein, trilinolein, trilinolenin, tristearin, and tripalmitin, and intermediates are provided in Table S2. Baseline moles and mass balances for the conversion of methyl oleate are discussed in Table S3.

Separation Processes for Azelaic Acid Recovery

The oxidative cleavage reaction produces a reaction mixture with an aqueous phase containing the dissolved catalysts and an organic phase containing the intermediates, side products, coproducts, and final products (simplified process design diagram illustrating conversion processes shown in Figure , Process A). The aqueous phase undergoes catalyst recovery based on a previous experimental study and is recycled back into the dihydroxylation and oxidative cleavage reactors (simplified process diagram illustrating catalyst recovery processes shown in Figure , Process B; detailed process description provided in Section S1.4). The organic mixture is first stripped of moisture and volatiles and then enters downstream separation (simplified process diagram illustrating downstream separation processes shown in Figure , Process C; detailed description of the downstream separation processes provided in Section S1.5). ,, A vacuum distillation column operating at 1000 Pa recovers a C5–C9 MCA fraction top product containing 42 wt % hexanoic acid, 25 wt % monomethyl suberate, and 19 wt % caprylic acid (the stream compositions presented here are rounded for clarity; full compositions can be found in the system_report Excel file under the sheet name Stream table in SI-B). The bottom product enters a second vacuum distillation column (5000 Pa), yielding a top product with 97 wt % pelargonic acid. The bottom contains 58 wt % monomethyl azelate and ∼28 wt % of other unreacted fatty acid methyl esters. Monomethyl azelate and other C8–C10 and C16–C18 monomethyl esters are hydrolyzed using a series of three continuous hydrolysis reactors, each with a conversion efficiency of 30 wt % to produce crude azelaic acid. , Methanol and water vapor are vented from each hydrolysis reactor, and the methanol is then condensed to recover biobased methanol (purity of 98 wt %). The unreacted diols and complex esters present in the crude azelaic acid stream are separated using a multieffect evaporator. Using a stream splitter, 50 wt % of the recovered diols and mixed esters are recycled back to the dihydroxylation reactor and the remaining is combined with another diol-rich stream obtained later in the process to make the fatty acid blend coproduct. The diol-free stream obtained from the multieffect evaporator undergoes a liquid–liquid extraction (LLE) using a counterflow multistage mixer-settlers to primarily extract azelaic acid, monomethyl azelate, and monomethyl suberate, in hot water (90 °C) and recover remaining monocarboxylic acids in the organic phase using heptane as the solvent (partition coefficients in Table S4). Heptane is distilled from the organic phase as the top product and recycled back to multistage mixers, while the bottom product is combined with the second stream obtained from an upstream diol splitter to make the fatty acid blend coproduct. The water in the aqueous phase is evaporated; 75 wt % of it is recycled back to the counterflow multistage extractor, and the rest is sent to the wastewater treatment facility. The resulting moisture-free stream contains azelaic acid (48 wt %), monomethyl azelate (49 wt %), and water (∼3 wt %), which is vacuum-distilled to yield an azelaic acid-rich stream (90 wt %), which is flaked. The bottom stream containing monomethyl azelate (95 wt %) and azelaic acid (5 wt %) is recycled back to the third hydrolysis reactor. Additionally, the biorefinery also includes facilities such as wastewater treatment, a boiler turbogenerator, a cooling tower, and air distribution systems (more details on the facilities and additional auxiliaries can be found in Sections S1.6 and S1.7).

2.

Simplified block flow diagram illustrating (Process A) TSOC conversion processes, (Process B) catalyst recovery steps, and (Process C) separation and recovery of azelaic acid and coproducts. The block flow diagram for biodiesel production and glycerol recovery from high-oleic soybean oil (HOSoy) is provided in Figure S1. Wastewater treatment is denoted as WWT. For clarity, some streams are omitted; the complete process flow diagram is available in the Excel file named system_report, under the sheet name Flowsheet in SI-B.

Technoeconomic Analysis (TEA) and Life Cycle Assessment (LCA)

Previous TEA literature on azelaic acid production uses a modified Petley’s correlation to estimate capital cost for the n ^th plant (i.e., a successful industry has been established with mature technologies). The correlation only considers the maximum temperature, number of process steps, and maximum pressure used in a process. In contrast, our study utilizes a more detailed factorial cost estimation approach. − The installed equipment cost (IEC) for each unit is estimated individually based on the purchase cost of each unit priced using BioSTEAM’s cost decorators and equipment cost adjustment factors (eq S3) upon simulation. BioSTEAM’s TEA framework was then employed to add cost factors to IEC for the estimation of total depreciable capital (TDC), fixed capital investment (FCI), and fixed operating cost (FOC). These factors were obtained from textbooks on plant design and previously published literature and are provided in Table S9. The default TEA functions in BioSTEAM were used to estimate the total capital investment (TCI), return on investment (ROI), and annual operating cost (AOC) and conduct cash flow analysis to estimate the minimum selling price (MSP) of azelaic acid at a net present value (NPV) of zero and an internal rate of return (IRR) of 15%, consistent with the average IRR reported for comparable biobased products. Even though our previous works have assumed a lower IRR of 10%, , in this study, we adopt a higher IRR of 15% to reflect the elevated risk associated with both the novelty of the feedstock and the process for producing azelaic acid.

Maximum feedstock purchase price (MFPP) was also calculated at NPV = 0, but at constant prices of all products (including azelaic acid) and costs of other inputs. A fixed IRR of 15% was used for both the MSP and MFPP estimation. Feedstock purchase price was estimated based on processor premiums imposed on high-oleic feedstock. All material costs, utility costs, and product prices are discussed in Section S2.1 and are available in Table S8. All equipment, material, and labor costs are indexed to 2023 dollars.

To systematically evaluate the environmental impacts of the proposed system, we performed a cradle-to-grave life cycle assessment (LCA) using simulated material and utility flows. The functional unit was set to 1 kg of produced azelaic acid (90 wt % purity), consistent with the TEA. Although the LCA framework is setup to characterize a range of impacts, the focus of our discussion will be on carbon intensity (CI) [kg of CO₂-eq] with a 100–year time horizon. The CI characterization factors for individual materials and utilities were obtained from GREET.net (2023), IPCC 2013 impact factors from Ecoinvent (version 3.6), and past literature (values are provided in Table S8). Following standards of practice in biofuel/bioproduct LCAs and guidelines in the U.S. Renewable Fuel Standard (RFS), the LCA does not include impacts resulting from the construction and demolition of the necessary biorefinery infrastructure (i.e., only operation and maintenance of the biorefinery are included). Since the HOSoy varieties and commodity soybean varieties share the same growing conditions, fertilizer use, and pesticide use and have an identical total lipid content resulting in the same oil and meal distribution, the life cycle inventory was assumed to be consistent with assumptions made in a past study.

Given that the production process generates several multifunctional products, including crude glycerol, C5–C9 MCA fraction, pelargonic acid, methanol, fatty acid blend, and azelaic acid, selection of the allocation procedure used in LCA may be an important decision. To understand the implications of this methodological choice, we repeated the full analysis with three different allocation methods: (i) mass-based allocation, (ii) economic allocation, and (iii) system expansion through displacement. We did not consider energy allocation since all of the products (except the fatty acid blend, which can be used for biodiesel production) are not energy products. We excluded hybrid LCA methods due to their complexity, which could hinder interpretability. Mass-based allocation is used in LCAs of bioproducts with the implicit assumption that larger mass flows through the engineered process will result in larger emissions (scope 1, 2, and 3), raw material requirements, and energy consumption. , Alternatively, economic allocation, in essence, attributes the greatest environmental burdens to the products yielding the greatest revenue. Because the economic allocation method is sensitive to price fluctuations, the uncertainty in CI scores stemming from this analysis is also influenced by uncertainty in financial inputs used in the TEA. With regard to system expansion, it is important to ensure that the coproducts (i.e., the system outputs displacing designated products on the market) have large enough market sizes and possess future growth potential so that all the coproducts can be consumed. If their market size is small, then it can become quickly saturated and might make the allocation method unreliable. , Similar to azelaic acid, coproducts have a multitude of uses that are expected to experience continued market growth (Table S10), with forecast CAGRs of 6.7% (pelargonic acid), 4.9% (fatty acid blend), 8.1% for caproic acid (the main component of the C5–C9 fraction, making up 42 wt %), 1.9% (crude glycerol), and 5.06% (methanol).

In cases where the CI of the conventional method of production was not available or the coproduct did not exactly match an existing product on the market, alternative compounds (for which the coproduct could serve as a substitute) were considered (details in Table S8). For instance, the C5–C9 MCA product stream is rich in caproic acid (42 wt %) and methyl caprylate (19 wt %), and we thus simulated the displacement of an equivalent mass of caproic acid produced using its conventional method (via coconut oil) and displacement of an equivalent mass of caprylic acid produced using its conventional method (via palm oil). In the case of pelargonic acid, life cycle CI was not available in existing databases. However, pelargonic acid has been proposed as a more sustainable alternative to glyphosate, the latter of which has received worldwide attention for its harmful effects. , Thus, pelargonic acid production was assumed to offset glyphosate production with the system expansion allocation approach. Lastly, biobased methanol produced in the system can be used to displace conventional fossil-derived methanol, and crude glycerol produced in the system can be used to displace conventionally produced glycerin from epichlorohydrin. Even though the fatty acid blend coproduct has several applications such as biofuels, waxes, and lubricants, , the TSOC method suggests using it for biodiesel production and therefore was assumed to displace conventional diesel.

Uncertainty and Sensitivity Analyses

To address the uncertainty associated with process design, simulation, and TEA/LCA assumptions, we conducted a Monte Carlo analysis with Latin hypercube sampling. A total of 59 parameters related to technological decisions, life cycle inventory assumptions, and material costs and prices varied across 2000 and 10,000 Monte Carlo simulations. Selection of distributions used in uncertainty analysis is discussed in Section S2.3, and distribution values are provided in Table S8. The corresponding data sets are available as Excel files, uncertainty_analysis_2000 and uncertainty_analysis_10000 in SI-B. Percentile values (5th, 50th, and 95th) for key metrics such as MSP, mass allocation (CI_mass), economic allocation (CI_economic), and system expansion via displacement (CI_displacement) are summarized in Table S21. Differences across all metrics are ≤1% and do not affect the conclusions of this study; therefore, we report and discuss the results from the 2000-run simulations. Spearman’s rank order correlation coefficients (Spearman’s ρ), which are a measure of monotonicity between input and output variables, were used to characterize the sensitivity of parameters to several sustainability indicators. These included economic metrics (MSP, MFPP, TCI, ROI, AOC, and ROI), process-related metrics (product yield, product purity, heating and cooling demand, and electricity consumption), and CI_mass, CI_economic, and CI_displacement to each individual uncertain parameter. To understand the potential return for investors, we also estimated the IRR at NPV = 0 while keeping the prices of all products (including azelaic acid) and the costs of other inputs constant. The calculated IRR represented an additional economic metric of potential investment performance. The impacts of feedstock flow and IRR on MSP were quantified separately as shown in Figure S2 (data set provided in Table S20) and Figure S3 (data sets provided in SI-B under subfolder name MSP_at_different_IRRs).

Results and Discussion

Financial Viability under Uncertainties

Benchmarking the Azelaic Acid MSP

The MSP of azelaic acid is estimated to be 6.62 $ kg^–1 at the baseline with a median value of 8.32 [4.93–13.34] [hereinafter, 5th–95th percentiles are shown in brackets] (Figure A). Estimates of the market price of azelaic acid ranged from 9.93 $ kg^–1 (based on a cost ratio to sebacic acid, a competing dicarboxylic acid) to 12.10 $ kg^–1 (based on bulk pricing estimates), as detailed in Section S2.1 of the SI. The MSP was below the low end of the market price range in 72% of the simulations and below the high end of the market price range in 90% of the simulations (Figure A). In comparison to the market price range of sebacic acid, estimated at 6.42–6.82 $ kg^–1 (Section S2.1), the MSP of azelaic acid falls within the sebacic acid market price range in 5% of simulations and entirely below the market range in 22% of simulations. These results support the conclusion that the TSOC method has the potential to be financially viable with HOSoy as a feedstock despite uncertainties associated with reaction conversions, technological decisions associated with the separation process, and material costs and prices.

3.

Uncertainty analysis and baseline breakdowns for (A) minimum selling price (MSP) and (B, C) carbon intensity (CI). Market price ranges in (A) extend from 9.93 to 12.10 $ kg^–1 for azelaic acid and 6.42 to 6.82 $ kg^–1 for sebacic acid (details in Section S2.1 of the SI). CI distributions in (B) were estimated using (1) system expansion via displacement, (2) mass-based allocation, and (3) economic allocation. Whiskers, boxes, and center lines correspond to the 5th/95th, 25th/75th, and 50th percentiles, respectively, from 2000 Monte Carlo simulations. Diamonds and stacked bars report results for baseline values. Stacked bar plots further detail (A) the cost and utility breakdown (in %) and (C) the CI breakdown for each LCA method. Direct emissions include emissions from both the production process and end-of-life (which were assumed to be entirely from passive oxidation to CO₂). Electricity purchased reflects net electricity use, accounting for electricity generated on-site via a boiler turbogenerator system powered by process waste-derived steam.

Technological Hotspots at the Current State of Technology

According to the cost and utility breakdown (Figure ; absolute and relative contributions for cost and utility by biorefinery area are provided in Table S12), facilities, including the boiler turbogenerator, wastewater treatment, and other facilities, account for 49.3% [45.2–54.3%] of the total installed equipment cost (IEC). Among these, the boiler turbogenerator is the largest single contributor, comprising 34.9% [32.9–37.7%] of total IEC, primarily due to the large system heating demand.

The separation of methanol, azelaic acid, and the fatty acid blend contributes 17.5% [16.3–18.5%] to total IEC. Within this area, the hydrolysis units are the most capital-intensive, accounting for 8.9% [8.0–9.9%] of total IEC. This step could be bypassed entirely if high-purity oleic acid is used as a feedstock, yielding azelaic acid directly rather than the monomethyl azelate ester generated during the TSOC process, thereby reducing capital expenditures. Biodiesel production and glycerol recovery collectively contribute 17.6% [16.1–19.2%] to total IEC. The remaining biorefinery areas cumulatively contribute 15.3% [14.1–17.1%]. The biodiesel production and glycerol recovery area is the dominant driver of annual material costs, accounting for 83.1% [78.1–87.1%], nearly all of which is due to the purchase of high-oleic soybean (HOSoy) oil (representing 82.9% [77.8–86.9%] of annual material cost). The high purchase cost of HOSoy oil stems from a premium (0.66 $ kg^–1) over refined, bleached, and deodorized soybean oil. The dihydroxylation section is the second largest consumer of materials, primarily due to hydrogen peroxide, which contributes 13.3% [9.9–17.6%] to the annual material cost. Despite this, hydrogen peroxide is a more economical and versatile oxidant than ozone. Both processes require stoichiometrically equivalent oxidants for the initial conversion of the feedstock to diol intermediates, followed by an equimolar quantity of oxygen for the subsequent oxidation to azelaic acid and coproducts. , These processes result in complete oxidant consumption, with any trace ozone decomposing to oxygen and all residual hydrogen peroxide decomposing during the cobalt-catalyzed step. , Ultimately, the cost of ozone per kilogram of azelaic acid (indexed to 2023 dollars) is 1.04 $ kg^–1, including operating, maintenance, and utility expenses for the required on-site ozone generation. In comparison, hydrogen peroxide can be directly purchased and delivered to the facility, resulting in a hydrogen peroxide material purchase cost of $0.85 per kilogram of 90% pure azelaic acid produced under the current TSOC process. An additional consideration is that ozone is less effective for oxidizing polyunsaturated fatty acids such as linoleic and linolenic acids, often requiring higher consumption and a preceding separation step.

Azelaic acid recovery is the most heat-intensive process area, responsible for 51.9% [38.7–63.4%] of total heating demand and 31.3% [24.7–36.3%] of total cooling demand. The counter-current LLE using multistage mixer-settlers is the main driver due to the heating and cooling of water, solvents, and organic streams, as well as downstream water recovery and recycling. LLE contributes 10.3% [6.4–14.7%] to heating demand and 13.5% [8.5– 8.7%] to cooling demand. Although energy-intensive, water-based extraction has been proposed as an industrially viable method for separating azelaic acid from mixtures containing monocarboxylic acids, with no other feasible alternative identified to date. ,,,, Separation of monomethyl azelate from azelaic acid at 3000 Pa contributes 10.1% [7.2–13.3%] to heating demand and 3.5% [2.6–4.6%] to cooling demand. This separation step could be eliminated if oleic acid were used directly as a feedstock. All distillation columns involved in recovering the C5–C9 MCA fraction, pelargonic acid, fatty acid blend, and azelaic acid operate at pressures between 50 and 3000 Pa to avoid thermal degradation. ,

Additionally, distillation columns for the separation of pelargonic acid and the C5–C9 MCA fraction impose significant energy burdens: 31.2% [21.0–46.6%] of total heating and 12.4% [7.0–22.5%] of total cooling demand. However, this step is essential to reduce MCA impurities to <0.5 wt %, thereby preventing termination of azelaic acid chain growth polymerization. MCA recovery also requires considerable electricity, particularly for the second distillation column used in pelargonic acid recovery (20.5% [16.9–24.7%] of total electricity use). Finally, the azelaic acid separation area further contributes 24.5% [21.9–28.0%] of electricity demand, largely due to heptane recovery from the organic phase following counter-current LLE, which alone accounts for 15.0% [13.2–17.3%]. Overall, counter-current LLE of azelaic acid emerged as a major contributor to capital cost, heating and cooling demand, and electricity consumption.

Drivers of MSP and System Performance

Sensitivity analysis results (Figure ) indicate that the uncertainty in MSP is significantly influenced by oxidative cleavage of diols to intermediates (DI conversion; Spearman’s ρ = −0.5) and moderately influenced by dihydroxylation (D conversion; ρ = −0.4) and oxidation of intermediates to products (IP conversion; ρ = −0.3). In addition to conversions, the uncertainty in MSP is also driven by feedstock cost (ρ = 0.4), pelargonic acid price (ρ = −0.4), and fatty acid price (ρ = −0.3). The signs of these ρ values indicate that improvements in reaction conversions, increases in the product prices, and reductions in feedstock cost will all decrease the MSP. The D and DI conversions are also key drivers of total product yield and MFPP, with Spearman’s ρ values ≥ 0.4. Detailed sensitivity analysis results are available in the Excel file sensitivity_analysis_2000 in SI-B.

4.

Spearman’s ρ values between input parameters and MSP, CI_displacement, CI_mass, and CI_economic. In the plot, D stands for dihydroxylation conversion, DI stands for oxidative cleavage of diol to intermediates, IP stands for oxidation of intermediates to products, and OD stands for oxidative decarboxylation of the diol to produce shorter-chain compounds (reactions discussed in more detail in Section S1.3). Bullet points indicate the parameters that affect biorefinery mass and energy (M & E) flows, annual azelaic acid production, capital cost, and operating cost. A total of 59 parameters were selected for sensitivity analysis (full list of distributions included in Table S8 in SI-A), and only 17 parameters with absolute values of ρ ≥ 0.09 for MSP, CI_displacement, CI_mass, and CI_economic are shown here. Crosses indicate that MSP, CI_displacement, CI_mass, and CI_economic were not appreciably (absolute value of ρ < 0.05) affected by the parameters. The entire list of all the parameters with their corresponding Spearman’s ρ values is provided in the Excel file.

Among other parameters examined, the amount of water in the multistage counter-current extraction of azelaic acid (labeled as “water ratio” in Figure ) has a substantial impact on capital investment, energy demands, and electricity consumption (all with Spearman’s ρ ≥ 0.4). The analysis underscores the need for innovative, less energy-intensive purification strategies to enable the cost-effective production of high-purity azelaic acid.

Effect of Feedstock Price and Reaction Conversions on the Minimum Selling Price

At a baseline feedstock purchase cost of 2.13 $ kg^–1, which includes 1.47 $ kg^–1 for soybean oil and a 0.66 $ kg^–1 premium, improving the dihydroxylation (D) reaction conversion from 86% to 99% and oxidative cleavage (DI) conversion from 93% to 99% reduces the MSP from 6.62 to 5.24 $ kg^–1 (Figure A). Across the full distribution range, an increase in feedstock purchase cost from the lower end (1.60 $ kg^–1) to the upper end (2.66 $ kg^–1) raises the baseline MSP from 5.30 to 7.95 $ kg^–1. At the higher feedstock purchase cost of 2.66 $ kg^–1, even with maximum proposed conversion efficiencies (99% for both D and DI conversions), the MSP only decreases to 6.28 $ kg^–1, down from 7.95 $ kg^–1. In contrast, at the lower feedstock cost of 1.60 $ kg^–1, the MSP drops from 5.30 to 4.20 $ kg^–1 at 99% conversion in both D and DI conversions (Figure S4). While the absolute MSP is lower at reduced feedstock costs, remaining below the average market price of sebacic acid (6.62 $ kg^–1) in both baseline and improved scenarios, the marginal benefit of improving process conversions is greater at higher feedstock prices. Specifically, the reduction in MSP at 2.66 $ kg^–1 feedstock cost is 1.67 $ kg^–1, compared to a 1.09 $ kg^–1 reduction at a feedstock cost of 1.60 $ kg^–1. These findings highlight the increasing importance of targeting conversion efficiencies at elevated feedstock prices, particularly given that high-oleic vegetable oils typically carry an ∼$0.66 kg^–1 premium over conventional soybean oil.

5.

(A) Minimum selling price (MSP) and (B–D) carbon intensity (CI) as functions of dihydroxylation and oxidative cleavage conversion efficacy. CIs are calculated under (B) displacement allocation (CI_displacement), (C) mass allocation (CI_mass), and (D) economic allocation (CI_displacement). Data for all the contour plots are available in Tables S14, S15, S16, and S17 in SI-A.

Benchmarking the Azelaic Acid CI and the Impacts of Improved Conversion Efficiency

There is substantial variability in the estimated CI of azelaic acid depending on the allocation method used (Figure B): system expansion through displacement (CI_displacement = 0.0 [−5.5 to 5.6] kg of CO₂-eq kg^–1), mass-based allocation (CI_mass = 3.3 [3.2–3.4] kg of CO₂-eq kg^–1), and economic allocation (CI_economic = 7.1 [5.5–9.3] kg of CO₂-eq kg^–1). The negative value of azelaic acid CI_displacement (Figure ) results from coproducts that collectively offset 105.9% of the total emissions at the baseline, with a median offset of 99.3% [58.2–155.1%]. Excluding pelargonic acid, the remaining coproducts offset 22.5% of total emissions at the baseline with a median offset of 25.9% [19.9–33.7%]. Among all coproducts, pelargonic acid provides the largest offset, accounting for 83.4% at the baseline and a median offset of 73.3% [38.3–121.4%]. This significant reduction in CI_displacement is based on the assumption that biobased pelargonic acid can be used as a substitute for glyphosate in herbicide applications. , Glyphosate has a CI of 11 kg of CO₂-eqkg^–1, making CI_displacement highly sensitive to pelargonic acid CI (Spearman’s ρ = −0.9). Additionally, azelaic acid CI_displacement is also weakly correlated with D conversion (Spearman’s ρ = −0.2) and DI conversion (Spearman’s ρ = −0.3; Figure ). Due to this strong dependence on pelargonic acid CI, the influence of reaction conversions on CI_displacement was not apparent in the global sensitivity analysis. To better understand the direct effect of improvements on conversions, we intentionally varied D from 86% to 99% and DI from 93% to 99% and observed the CI_displacement reduced from −0.6 to < −1.9 (Figure B).

In contrast, the CI_mass of azelaic acid is most influenced by the feedstock CI (Spearman’s ρ = 0.7). It is not sensitive to D conversion (Spearman’s ρ = 0), but it is weakly correlated with reaction conversions, particularly the DI conversion (Spearman’s ρ = −0.3), IP conversion (Spearman’s ρ = −0.2), and OD conversion (Spearman’s ρ = 0.3). Additional factors such as tungstic acid reusability (Spearman’s ρ = −0.2) and the amount of water used in LLE (Spearman’s ρ = 0.3) further influence CI_mass (Figure ). Increasing D and DI to 99% can reduce CI_mass from 3.3 to 3.2 kg of CO₂-eqkg^–1 (Figure C). Compared to CI_displacement and CI_economic, CI_mass is more stable as increases in reaction conversions scale up the mass of azelaic acid and coproducts in parallel, preserving their relative impact fractions. However, caution is warranted when interpreting results from the mass-based allocation method: since the fatty acid blend is assumed to be used for energy production in this study, mass-based allocation may not be appropriate. Still, this approach may be appropriate under alternative fatty acid blend use-case scenarios, as outlined in Table S10.

Lastly, the CI_economic of azelaic acid is also sensitive to reaction conversions, with Spearman’s ρ values of −0.5 for DI, −0.4 for D, and −0.2 for IP conversions. CI_economic is strongly correlated with azelaic acid market price (Spearman’s ρ = 0.5) and weakly correlated with pelargonic acid price (ρ = −0.3), and fatty acid blend price (ρ = −0.3) (Figure ). Improving D and DI from the baseline to 99% can reduce the CI_economic of azelaic acid from 6.8 to 5.8 (Figure D). Although coproduct prices have limited influence on CI_economic, it is important to note that the entire C5–C9 MCA fraction lacks a well-established commercial market. As modeled here, caproic acid is used as a proxy, but this may not be a reliable representation of the market conditions.

Influence of Triacylglyceride (TAG) Composition on Financial Viability

The TAG composition of high-oleic vegetable oils is determined by the proprietary seed varieties used for their production. ,, Understanding the influence of individual TAGs, specifically the mass fractions of triolein (O), trilinolein (L), trilinolenin (Ln), tripalmitin (P), and tristearin (S), on the MSP of azelaic acid can inform feedstock selection for the TSOC process. In this process, only fatty acids produced from O, L, and Ln in the biodiesel are converted into value-added products, while fatty acids produced from P and S remain unconverted and contribute to the fatty acid blend coproduct, as discussed in previous sections. To assess how TAG composition affects MSP and to identify HOSoy oil profiles that optimize economic performance, we analyzed a data set of 1252 feedstock compositions generated using TAG percentage ranges reported by the FCC for HOSoy oil (available in SI-B under subfolder name feedstock_composition_heatmap_data). Compositions varied across the following ranges: L (1–10%), Ln (1–6%), O (75–85%), P (4–8%), and S (2–6%). All feedstocks were constrained to contain a 98% TAG content by total mass, with the remaining 2% consisting of water and free fatty acids, consistent with the baseline composition discussed in Section S1.1 (SI-A). To ensure a consistent comparison across feedstock compositions, all analyses in this section were conducted using baseline conditions, with a feedstock purchase cost of 2.13 $ kg^–1 and all other input costs and coproduct prices held constant. Additionally, to account for potential price variation among different HOSoy varieties, we estimated MSP values at feedstock purchase prices of 1.60 and 2.66 $ kg^–1 across 1252 feedstock TAG compositions, representing a ±25% variation around the baseline feedstock unit cost (results provided in SI-B under subfolder feedstock_composition_heatmap_data).

To identify the TAGs with the strongest influence on the MSP, we estimated MSP values against the concentration of each individual TAG across the spectrum of potential values (Figure S5). The slopes in Figure S5 reflect the absolute rate and direction of change in median MSP ($ kg^–1) per unit (%) change in TAG content. Positive slopes (e.g., for P, S, L, and Ln) indicate that higher levels of these TAGs raise the MSP, while negative slopes (e.g., for O) indicate an inverse relationship. Although the slope magnitudes are modest (approximately ±0.03–0.06 $ kg^–1 per % TAG), the maximum difference in MSP is 0.82 $ kg^–1 across the full composition range. Increasing the O (%) from 75% to 85%, specifically, reduces the median MSP from 6.84 [6.79–6.93] to 6.24 [6.17 – 6.29] $ kg^–1. Because the total TAG content is fixed at 98%, increasing one TAG necessarily reduces the proportions of others. To minimize confounding effects of this compositional interdependence, we examined the combined influence of L and Ln on MSP while holding the values for O and P constant and specifying discrete values for S (Figure ). As shown, increasing O consistently reduces MSP, indicated by a shift from darker bluish gray (higher MSP) to lighter green (lower MSP). At each fixed O level, increasing P results in a higher MSP. For example, at 85% O, increasing P from 4% to 6% modestly raises the median MSP from 6.24 to 6.26 $ kg^–1. Additionally, when O is held at 85%, P at 4%, and S at 2%, increasing the amount of Ln from 1% to 5% and reducing the amount of L from 6% to 2% raise the MSP from 6.18 to 6.26 $ kg^–1 (Figure ). These trends indicate that, at fixed O, P, and S levels, a higher Ln content increases MSP, while a lower L content can partially offset this effect.

6.

In this figure, %O represents the mass percentage of triolein in HOSoy oil, which is fixed at three levels, 75%, 80%, and 85%, corresponding to each row of subplots. The columns represent fixed values of %P (mass percentage of tripalmitin), set at 4%, 6%, and 8%. Within each subplot, %S (mass percentage of tristearin) is annotated inside the cells. The y-axis of each subplot represents %L (trilinolein), while the x-axis represents %Ln (trilinolenin). Together, the plots illustrate how variation in the TAG composition affects the MSP of azelaic acid.

Given that oleic acid (in triolein, O) contains one double bond, linoleic acid (in trilinolein, L) contains two, and linolenic acid (in trilinolenin, Ln) contains three, the cost of oxidatively cleaving Ln using the TSOC process is expected to be higher than those of O and L. In summary, selecting feedstocks with a high triolein (O) content and a low trilinolenin content (Ln) results in a lower MSP for azelaic acid production. This finding aligns with the development of high-oleic, low-linolenic oilseed varieties, which are favored in the food industry for their enhanced oxidative stability. Accordingly, feedstock traits selected for food applications can also simultaneously improve the economic viability of bioproducts, such as azelaic acid, produced via the TSOC process. The trends in MSP in response to TAG composition discussed in this section are robust and hold true across the full range of evaluated feedstock prices (results provided in SI-B under the subfolder feedstock_composition_heatmap_data). However, the open-source model can be readily adapted to reflect the price and TAG composition of any oil variety containing oleic, linoleic, and linolenic acids, enabling a customized evaluation of feedstock performance.

Conclusions

In this study, we present the first comprehensive technoeconomic analysis (TEA) and life cycle assessment (LCA) of azelaic acid production via the TSOC process using RBD HOSoy oil as a feedstock (Plenish variety), implemented using the BioSTEAM platform. The modeled process includes transesterification of HOSoy oil to produce biodiesel and glycerol, oxidative conversion of biodiesel via TSOC, catalyst recovery and recycle, and downstream separation of coproducts including the C5–C9 MCA fraction, pelargonic acid, methanol, a fatty acid blend, and azelaic acid.

Simulation and TEA results suggest that azelaic acid can be produced sustainably and cost-effectively from HOSoy oil, with a baseline MSP of 6.62 $ kg^–1, comparable to the average market price of sebacic acid (∼6.62 $ kg^–1), a structurally similar dicarboxylic acid. Uncertainty analysis further supports the financial viability of the TSOC process, with the MSP falling below estimated market values in over 70% of the simulations.

Key process hotspots identified include the counter-current liquid–liquid extraction (LLE) and vacuum distillation units used for coproduct recovery and monomethyl azelate isolation, all of which significantly impact cost, energy use, and electricity demand. The LLE system is required for extraction of azelaic acid, and the vacuum distillation columns (operating at 50–3000 Pa) are necessary to prevent thermal degradation and achieve the ∼90% purity required for industrial-grade azelaic acid, highlighting the trade-off between separation efficiency, resource use, and final product quality. Future work could focus on process intensification through extractive distillation and membrane-based separation technologies, which might offer improved performance.

This study also provides the first estimates of the CI of azelaic acid production. Results vary substantially depending on the LCA allocation method: system expansion via displacement yields a median CI of 0.0 [−5.5 to 5.6] kg of CO₂-eq kg^–1, indicating potential for net-negative emissions when and if coproducts displace conventional products, e.g., if pelargonic acid successfully displaces glyphosate, if biobased methanol successfully displaces fossil-derived methanol, and if the fatty acid blend successfully displaces conventional diesel. In contrast, mass-based and economic allocation methods produce higher CI values of 3.3 and 7.1 kg of CO₂-eq kg^–1, respectively, with economic allocation dependent on assumptions regarding coproduct usage and market prices. These findings underscore the importance of the allocation methodology, coproduct adoption in the marketplace, and the need for the transparent documentation of assumptions in sustainability assessments.

Feedstock composition also plays a critical role in the process performance. Among the TAGs present in HOSoy oil, triolein (O) most favorably influences MSP, while trilinolenin (Ln) and tripalmitin (P) increase it. Increasing the O content from 75% to 85% reduces the median MSP from 6.84 to 6.24 $ kg^–1. These insights offer clear guidance for selecting or engineering high-oleic feedstocks optimized for TSOC-based azelaic acid production. While HOSoy exhibits a fatty acid composition that is highly favorable for industrial production, it holds significant value within the food industry due to its oxidative stability and potentially results in competition with food supply chains. Beyond HOSoy, the modeling framework developed in this study (along with biorefinery models from our previous work) can be applied to other high-oleic vegetable oil varieties such as sunflower and canola as their production quantities increase as they may offer comparable compositional advantages. Additionally, the model can be utilized to evaluate the sustainability of azelaic acid production from nonfood lipid feedstocks such as transgenic oilcane, which also offer favorable fatty acid profiles for oleochemical production. At present, purchase cost data for such nonfood lipid feedstocks are unavailable, limiting direct comparison of their sustainability with HOSoy; the analysis of emerging feedstocks may be the focus of future work.

In conclusion, the TSOC process represents a promising route for more sustainable and economically competitive azelaic acid production from high-oleic-grade vegetable oils. The modeling framework developed here enables rigorous assessment across diverse lipid feedstocks, including other high-oleic oil varieties and nonfood lipid crops such as transgenic oilcane and oil sorghum. This work identifies key technological bottlenecks, prioritizes areas for future research, and provides a transparent evaluation of the environmental impacts associated with azelaic acid production via the TSOC pathway.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsestengg.5c00853.

• Process description and analysis methods; methodology for estimating costs, prices, carbon intensity, and parameter distributions for uncertainty analysis; supplementary figures; supplementary equations; supplementary tables; installation guideline for setting up the oleochemicals biorefinery; data sets supporting all figures (PDF)

• Detailed system reports, feedstock composition heat map data, minimum selling prices (MSP) at different internal rates of return (IRRs), Spearman’s rank order correlation results, and uncertainty analysis results (ZIP)

a.

Present address: Department of Chemical and Biological Engineering, University of Wisconsin-Madison, 1415 Engineering Drive, Madison, Wisconsin 53706, United States

J.S.G. and L.P.K. conceived the research. L.P.K. built the models with input from Y.R.C.-P. and S.S.B., and L.P.K performed the analyses. L.P.K. and J.S.G. interpreted the results and wrote the paper. All authors edited the paper.

The authors declare no competing financial interest.