Pre-publication: New and updated processes for non-fertilizer related chemicals

Date: March 20, 2025

Summary

This pre-publication presents the proposed changes to the non-fertilizer related chemicals processes in the Fuel Life Cycle Assessment Model (the Model). It provides an opportunity to stakeholders to provide their comments on the proposed changes and to Environment and Climate Change Canada to consider them for the next formal version of the Model.

1 Purpose

The purpose of this pre-publication is to present the proposed changes to non-fertilizer related chemicals processes in the Fuel Life Cycle Assessment (LCA) Model (the Model) for the next formal publication in 2026.

With this pre-publication, Environment and Climate Change Canada (ECCC) provides a description of the proposed changes to the Model, including a description of the methodology, data sources and assumptions for the processes of chemicals that were not updated as part of the June 2024 publication, as well as for a number of new chemicals. It also provides the carbon intensities (CIs) associated with the proposed changes. Stakeholders are invited to provide comments on this pre-publication.

Pre-publications are not be used for compliance with the Clean Fuel Regulations or other regulations or used to support eligibility for programs such as the Clean Hydrogen Investment Tax Credit unless otherwise specified.

The CIs presented in the pre-publication may differ from those included in the next formal publication of the Model based on comments received and other changes implemented in the Model.

This pre-publication includes this descriptive document and an openLCA module.  The module includes, for each new or updated process, a system process located in the Data Library folder of the Model Database and a unit process located in a new folder named Background Modelling. The system processes contain the rolled-up CI, while the unit processes showcase the inputs and outputs for each process. The unit processes are included for transparency to provide disaggregated data used in the modeling.

2 Context

The current Model Database includes the following non-fertilizer related chemical processes. These processes are available in the folder Data Library\Chemical inputs\Chemicals:

• Acetic acid (CH3COOH)
• Alpha amylase
• Calcium carbonate (CaCO3)
• Cellulase
• Cellulase protein
• Citric acid (C6H8O7)
• Corn steep liquor
• Gluco amylase
• Glucose
• Hexane (n-hexane)
• Hydrochloric acid (HCl)
• Lime (CaO)
• Methanol (CH3OH), from natural gas
• Nitrogen gas (N2), gaseous, from natural gas
• Potassium hydroxide (KOH)
• Sodium hydroxide (NaOH)
• Sodium methoxide (CH3ONa)
• Yeast
• Yeast extract

These processes are modelled using cradle-to-user life cycle emission factors from R&D GREET 2018.

The Model also includes the following chemicals related to fertilizers available in the folder Data Library\Chemical inputs\Chemicals:

• Ammonia from SMR (NH3)
• Ammonium nitrate (NH4NO3)
• Ammonium sulfate ((NH4)2SO4)
• Hydrogen production, at producer
• Monoammonium phosphate (NH4H2PO4)
• Monoammonium phosphate (NH4H2PO4), as N
• Monoammonium phosphate (NH4H2PO4), as P2O5
• Nitric acid (HNO3)
• Diammonium phosphate ((NH4)2HPO4))
• Diammonium phosphate ((NH4)2HPO4)), as N
• Diammonium phosphate ((NH4)2HPO4)), as P2O5
• Phosphoric acid (H3PO4)
• Sulfuric acid (H2SO4)
• Urea-Ammonium nitrate (UAN)

Since these processes were already updated in the June 2024 publication of the Model, this pre-publication does not propose any changes to these processes.

The Model also includes fertilizer processes bundled for each nutrient category (Nitrogen, Phosphorus, Potassium and Sulfur) that are available in the following folder: Data library\Chemical inputs\Agrochemicals. They are modelled using CIs from a 2020 ChemInfo report for Canadian Roundtable for Sustainable Crops (CRSC). This pre-publication does not propose any changes to these processes.

Finally, the Model contains three types of predefined chemical mixes to represent the chemicals used in the production of three types of fuels: conventional bioethanol, cellulosic ethanol, and biodiesel. These can be found in the folder Data Library\Chemical inputs\Predefined chemical mixes. This pre-publication does not propose any changes to these processes.

For information on the current modelling methodology, see Chapter 3.1.1 of the Fuel LCA Model Methodology.

3 Description of the proposed changes to the Model

For the next formal publication of the Model, it is proposed to update the following chemical processes:

• Acetic acid (CH3COOH)
• Alpha amylase
• Calcium carbonate (CaCO3)
• Lime (CaO)
• Cellulase
• Cellulase protein
• Citric acid (C6H8O7)
• Corn steep liquor
• Gluco amylase
• Glucose
• Hexane (n-hexane)
• Hydrochloric acid (HCl)
• Methanol (CH3OH), from natural gas
• Nitrogen gas (N2), gaseous^{Footnote 1} 
• Potassium hydroxide (KOH)
• Sodium hydroxide (NaOH)
• Sodium methoxide (CH3ONa), dry^{Footnote 2} 
• Yeast
• Yeast extract

In addition, it is proposed to introduce the following new processes:

• Chlorine (Cl2)
• Glycerin, crude
• Glycerin, refined
• Sodium chloride (NaCl)
• Sodium methoxide (CH3ONa), in solution
• Starch

It is also proposed to establish a consistent approach for determining the grid electricity and transport inputs for the chemical processes and implement it for the chemical processes updated in this pre-publication. This approach is presented in Annex B.

The updated CIs for these processes are available for public review in Annex A of this descriptive document and in a module that can be uploaded in openLCA.

The following sections provide an overview of the proposed modelling approach for the chemicals. More details are provided in Annex B.

3.1 Chemicals modelled primarily with GREET 2023

The following processes are proposed to be updated using material inputs, energy inputs, process emissions, and transportation data from R&D GREET 2023 Revision 1 (R&D GREET 2023):

• Acetic acid (CH3COOH)
• Alpha amylase
• Calcium carbonate (CaCO3)
• Lime (CaO)
• Cellulase
• Cellulase protein
• Corn steep liquor
• Gluco amylase
• Glucose
• Hydrochloric acid (HCl)
• Nitrogen gas (N2), gaseous
• Potassium hydroxide (KOH)
• Yeast extract

3.2 Methanol

The proposed updates to the methanol process are based primarily on data from a Studio Gear Up report (Hamelinck et al., 2022), as recommended by the Stakeholder Technical Advisory Committee (STAC). Data on transportation of methanol from the end of the production facility gate to the end user was taken from R&D GREET 2023, while data on the splits of natural gas used as a feedstock and as an energy source was calculated using data from a 2017 report on methanol production (Blumberg et al., 2017).

3.3 Sodium hydroxide and chlorine

The proposed updates to the modelling of the process for the coproduction of sodium hydroxide (NaOH) and chlorine (Cl₂) uses input and output data from a 2014 publication on the chlor-alkyl production process (BAT chlor-alkali, 2014). Assumptions for transportation modes and distances for the products from the production facility to the end user are from R&D GREET 2023.

3.4 Citric acid

The proposed updates to the modelling of the process for citric acid uses data from a 2020 publication on citric acid production (Wang et al., 2020). Assumptions were made for transportation modes and transportation distances for both the transport of the corn feedstock to the citric acid production facility and for the transport of citric acid from the end of the production facility gate to the end user.

3.5 Sodium methoxide

The proposed updates to the modelling of the process for sodium methoxide (CH3ONa) use data from a 2015 publication analyzing the production of sodium methoxide (Granjo et al., 2015). Currently, sodium methoxide production in the Model is represented by the Process “Sodium methoxide (CH3ONa)”. In the proposed update, two processes will be available: “Sodium methoxide (CH3ONa), dry” and “Sodium methoxide (CH3ONa), in solution”. “Sodium methoxide (CH3ONa), dry” is equivalent in scope to the existing “Sodium methoxide (CH3ONa)” process, while “Sodium methoxide (CH3ONa), in solution” includes the methanol used to dilute sodium methoxide in its CI. Users can choose the process that best fits their modelling needs. Assumptions for transportation modes and distances for the sodium methoxide from the end of the production facility gate to the end user are from R&D GREET 2023.

3.6 Hexane and yeast

There is no change to the methodology for the following processes:

• Hexane (n-hexane)
• Yeast

The proposed changes are associated with the data sources used to determine the life cycle emission factors. In the current Model, R&D GREET 2018 life cycle emission factors are used, while the proposed update uses more recent data from R&D GREET 2023.

3.7 New chemicals

The following new chemical processes are proposed to be added to the Model:

• Chlorine (Cl2)
• Glycerin, crude
• Glycerin, refined
• Sodium chloride (NaCl)
• Starch

The proposed modeling approach for the chlorine process uses input and output data from BAT chlor-alkali, 2014. Assumptions related to transportation modes and distances for the products from the production facility gate to the end user are from R&D GREET 2023.

The proposed modeling approach for the refined glycerin process uses crude glycerin as an input, and the refinement process uses data from a publication on crude glycerol purification (Bansod et al., 2024).

The proposed modeling approach for the sodium chloride and starch processes uses material inputs, energy inputs, process emissions, and transportation data from R&D GREET 2023. Crude glycerin is modelled using lifecycle emission factors from R&D GREET 2023 with energy allocation between itself and biodiesel.

4 Instructions for importing the module

The module is available in the folder New and updated processes for non-fertilizer related chemicals of the ECCC Data Catalogue and can be imported either into an empty database or into the Model Database.

Importing the module into an empty database allows users to see only the new and revised processes. No CI calculations can be performed when importing the module into an empty database.

Importing the module into the Model Database allows users to recalculate CIs without any additional steps.  However, it is important to note that the import of the module in the Model Database will update the values of the existing processes and the changes are irreversible. Consequently, users should always import the module into a copy of their original database.

For more information, please refer to the Instructions on how to import a module into openLCA.

5 How to submit comments on this pre-publication

Stakeholders are invited to review this pre-publication and provide comments to ECCC within 30 days following the pre-publication at modeleacvcarburant-fuellcamodel@ec.gc.ca.

Please indicate the following in the subject line: Comments on the pre-publication: New and updated processes for non-fertilizer related chemicals

Comments submitted will be considered for the development of the next formal version of the Fuel LCA Model.

For any questions related to this pre-publication, please contact modeleacvcarburant-fuellcamodel@ec.gc.ca with the following subject line: Questions on the pre-publication: New and updated processes for non-fertilizer related chemicals.

Annex A– CI Comparison

The CI presented in this Annex use the global warming potential (GWP) for the 100-year time horizon of the Intergovernmental Panel on Climate Change (IPCC) 5th Assessment Report (AR5). For more information, please refer to chapter 2.8 of the Fuel LCA Model Methodology.

Annex B – Proposed revised methodology

The methodology for the chemicals in the current version of the Model can be found in Chapter 3.1.1 of the Fuel Life Cycle Assessment Model Methodology.

The following sections provide the proposed changes to the methodology that will be reflected in the next formal version of the Model. Note that the section numbers and text could change for the next formal version of the Fuel LCA Model Methodology.

3.1 Chemical inputs

Chemicals used throughout the production processes of LCIF pathways include enzymes, acids, fertilizers, and catalysts, and others. The functional unit for each chemical is 1 kg of product, unless otherwise specified. The methodology for determining the CI for each of these chemicals included in the Model is described in the sub-sections below, and the methodology used depends on Canadian data availability. 

For most chemicals found in the folder Data Library\Chemical inputs\Chemicals, the following approach is used for the selection of the grid electricity process and transport process. If the approach chosen for a specific chemical is different than the one below, it will be specified in the specific chemical subsection.

Selection of grid electricity process:

• If the chemical is known to be produced solely in one specific location (i.e. in a specific state or province), that jurisdiction’s electricity grid will be selected 
• If the chemical is known to be produced solely in one country (either Canada or the U.S.), but the specific location is either unknown or it is produced across many provinces/states, the country’s average electricity grid will be selected
• If the chemical’s production location is either unknown or if the chemical is produced by both Canada and the U.S., the U.S. grid will be selected as a conservative approach 

Selection of the transport process and distance:

• Transportation was omitted if a chemical is known to typically be produced onsite
• The transportation mode was chosen based on available information for typical transport scenarios. If unavailable, 25-tonne truck transport was selected as a conservative approach to representing transportation in North America, due to chemicals typically being transported in smaller quantities
• The transportation distance was chosen based on available information for typical transport scenarios. If unavailable, the transport distance was assumed to be 50 miles

3.1.1 Chemicals related to fertilizers

Modelling approach for chemicals related to fertilizers

The following processes are modelled using the R&D GREET 2022 model’s process energy, material inputs and process emissions (R&D GREET 2022). The functional units are on a mass of product basis, unless otherwise specified.

• Ammonium nitrate (NH4NO3)^{Footnote 4} 
• Ammonium sulfate ((NH4)2SO4)
• Diammonium phosphate ((NH4)2HPO4))^{Footnote 5} 
• Diammonium phosphate ((NH4)2HPO4)), as N
• Diammonium phosphate ((NH4)2HPO4)), as P2O5
• Monoammonium phosphate (NH4H2PO4)^{Footnote 6} 
• Monoammonium phosphate (NH4H2PO4), as N
• Monoammonium phosphate (NH4H2PO4), as P2O5
• Nitric acid (HNO3)
• Sulfuric acid (H2SO4)^{Footnote 7} 
• Urea-Ammonium nitrate (UAN)^{Footnote 8}

The following process is modelled using R&D GREET 2022 life cycle emission factors:

• Phosphoric acid (H3PO4)

In the case of monoammonium phosphate (MAP) and diammonium phosphate (DAP), the processes for both allocated nutrient categories belonging to a multinutrient fertilizer (N and P₂O₅) must always be used together. The user must ensure that the quantities of both components of the multinutrient fertilizers are correctly reported.

The CI of the following chemical processes are based on the Canadian production data (feedstock and energy requirements) collected from the Greenhouse Gases Reporting Program for the year 2019-2020-2021-2022 (GHGRP 2019-2022):

• Ammonia from SMR (NH3)
• Urea (CH4N2O)

The modelling of ammonia and urea considers that ammonia from steam methane reforming (SMR) and urea are co-products. Urea production combines two molecules of ammonia with one molecule of carbon dioxide to form urea and water in solution. With this dual process, a portion of the CO₂ that would otherwise be emitted to the atmosphere is recovered by the urea process. A feedstock ratio of 0.567 kg NH3/kg urea is used to calculate the mass balance of the net ammonia production (stoichiometric mass ratio for 2NH₃ + CO₂ –˃ CH₄N₂O + H₂O). Allocation procedures based on nitrogen content was used for the ammonia and urea co-products. The nitrogen content used were 82.2% and 46.6% for ammonia and urea respectively.

A four-year combined average, from 2019 to 2022, for production, natural gas feedstock and energy requirements data was used. Plant activities considered include flaring, on-site transportation, steam generation, and other stationary combustion. Only plants that produce ammonia from steam methane reforming were used.

Geographical scope for chemicals related to fertilizers

The processes are modelled using Canadian and American data. They represent the production in North America.

Allocation for chemicals related to fertilizers

For Ammonia from SMR and Urea:  Allocation based on nitrogen nutrient content was used for the ammonia and urea co-products in the background modelling. The nitrogen content used were 82.2% and 46.6% for ammonia and urea respectively.

For MAP and DAP:  For the per-mass of nutrient basis processes, energy requirements, taken from R&D GREET 2022, are allocated to specific nutrient based on factors taken from a 2007 life cycle inventory report (Nemecek & Kägi, 2007). Material inputs are entirely allocated to the nutrient category they represent. 

No allocation procedures were performed for the other chemicals.

3.1.2 Enzymes

Modelling approach for enzymes

The following enzymes are included as part of the Model, and use energy inputs, material inputs, process emissions, and transportation data from R&D GREET 2023 (R&D GREET 2023):

• Alpha amylase
• Cellulase
• Cellulase protein
• Gluco amylase
• Yeast extract^{Footnote 9}

Using the electricity grid selection approach described in Chapter 3.1, the average U.S. electricity grid was used in the modelling of each chemical in this section.

The following process is modelled using R&D GREET 2023 life cycle emission factors:

• Yeast

Geographical scope for enzymes

The processes are modelled using Canadian and American data. They are representative of the production in North America.

Allocation for enzymes

No allocation procedures were performed for the enzymes.

3.1.3 Sodium hydroxide and chlorine

Modelling approach for sodium hydroxide and chlorine

The following chemicals are modelled based on the membrane cell and diaphragm cell techniques for the chlor-alkali process, found in two reports on the chlor-alkali process (Lee et al., 2017; BAT chlor-alkali, 2014):

• Chlorine (Cl2)
• Sodium hydroxide (NaOH)

Energy inputs (electricity and natural gas) from Lee et al., 2017 were used to model the process. The sodium chloride input was pulled from BAT chlor-alkali, 2014. The process is modelled as a production-weighted average of both techniques (0.45 for the membrane cell technique and 0.55 for the diaphragm cell technique). This is based off Lee et al., 2017, which analyzed the prevalence of each technique for the chlor-alkali process in the U.S.  2015 was used as the reference year.

Assumptions regarding transportation modes and distances for each product from the production facility to the end user are taken from R&D GREET 2023. Additionally, the average U.S. electricity grid was used in the modelling as a conservative approach to be representative of production in North America. The process inputs and outputs are tabulated below:

Geographical scope for sodium hydroxide and chlorine

The processes are modelled using Canadian, American, and foreign data. They are representative of the production in North America.

Allocation for sodium hydroxide and chlorine

Each process input is allocated to a specific product (chlorine or sodium hydroxide) if it is only used for that product’s after-treatment; otherwise, the inputs are allocated on a mass basis.

3.1.4 Methanol

Modelling approach for methanol

The modelling of methanol production from natural gas uses data related to conventional technology from a 2022 Studio Gear report (Hamelinck et al., 2022). The data used includes material and energy inputs for the production of methanol, including total combined volume of natural gas used as a material input and as an energy input and grid electricity consumption. This information is summarized in Table 3 below. Since methanol is mainly produced in Alberta, its provincial grid was used in the modelling.

For methanol, modes of transportation and distance from the production facility to the end user, were taken from R&D GREET 2023.

The total natural gas consumption was divided into a material input and an energy input to account for combustion emissions from natural gas. The split between the two inputs was calculated using data from a 2017 report on methanol production (Blumberg et al., 2017), summarized below in Table 4, resulting in approximately 39% of natural gas being modelled as a material input.

Geographical scope for methanol

The process is modelled using Canadian, American, and foreign data. It is representative of the production in North America.

Allocation for methanol

No allocation procedures were performed for methanol.

3.1.5 Citric acid

Modelling approach for citric acid

The modelling of citric acid uses data from the ion exchange recovery method (the method used for the only citric acid producer in Canada) as modelled in a 2020 report on citric acid production (Wang et al., 2020). The data used includes the quantities of corn, amylase, urea, hydrochloric acid, grid electricity, and steam.

Assumptions were made with respect to the transportation of the corn feedstock to the citric acid plant, as well as the transportation of citric acid to the end user. Additionally, the average U.S. electricity grid was used in the modelling as a conservative approach to be representative of production in North America. The production data and assumptions used for modelling are summarized in Table 5.

Geographical scope for citric acid

The process is modelled using Canadian, American, and foreign data. It is representative of the production in North America.

Allocation for citric acid

No allocation procedures were performed for citric acid.

3.1.6 Refined glycerin

Modelling approach for refined glycerin

The modelling for refined glycerin consists of refining crude glycerin, co-produced from biodiesel production. Vacuum distillation process data is used from a 2024 study on glycerin production (Bansod et al., 2024) due to the prevalence of this process in glycerin purification. Specifically, data for the quantities of crude glycerin, hydrochloric acid, steam, and grid electricity were used.

The average U.S. electricity grid was used as a conservative approach to be representative of production in North America. Additionally, the transportation mode and distance from the end of the production facility gate to the end user were assumed based off commonly used transportation data in R&D GREET 2023 for other similar chemicals.

The data used to model refined glycerin is summarized below:

Geographical scope for refined glycerin

The process is modelled using Canadian, American, and foreign data. It is representative of production in North America.

Allocation for refined glycerin

No allocation procedures were performed for refined glycerin.

3.1.7    Sodium methoxide

Modelling approach for sodium methoxide

Dry sodium methoxide and sodium methoxide in solution are both modelled using data from Process III in a 2015 report on sodium methoxide production (Granjo et al., 2015), which models production with sodium hydroxide as the feedstock. This process was chosen over other production methods due to its prevalence in industry and its higher stability compared to production with a sodium metal feedstock. In the modeling, sodium hydroxide in the output was assumed to be a minor impurity and was omitted.

Sodium methoxide is generally produced in a 30 wt.% solution with methanol. This excess methanol is modelled differently in the two processes; sodium methoxide in solution includes it in its CI, whereas dry sodium methoxide omits the excess methanol entirely. If users wish to include methanol in their subsequent modelling themselves, dry sodium methoxide should be used, and methanol must be added in their process to properly account for its CI. For every kilogram of dry sodium methoxide, users should add 2.344 kg of methanol to their modelled process.

The average U.S. electricity grid was used in the modelling. Additionally, the transportation mode and distance from the end of the production facility gate to the end user were assumed based off commonly used transportation data in R&D GREET 2023 for other similar chemicals. The process data used to model both sodium methoxide processes are provided in Table 7.

Geographical scope for sodium methoxide

The processes are modelled using Canadian, American, and foreign data. They are representative of production in North America.

Allocation for sodium methoxide

No allocation procedures were performed for sodium methoxide.

3.1.8 Other chemicals

Modelling approach for other chemicals

All other chemical processes found in the folder Data Library/Chemical inputs/Chemicals were modelled using energy inputs, material inputs, process emissions, and transportation data from R&D GREET 2023. Using the electricity grid selection approach described in Chapter 3.1, the average U.S. electricity grid was used in the modelling of each chemical in this section.

The following chemicals fall under this category, with specific assumptions listed below:

• Acetic acid (CH3COOH)^{Footnote 10} 
• Calcium carbonate (CaCO3)
• Corn steep liquor^{Footnote 11} 
• Glucose
• Hydrochloric acid (HCl)
• Lime (CaO)
• Nitrogen gas (N2), gaseous^{Footnote 12} 
• Potassium hydroxide (KOH)
• Sodium chloride (NaCl)
• Starch^{Footnote 13}

The following processes were modelled using R&D GREET 2023 life cycle emission factors:

• Glycerin, crude^{Footnote 14} 
• Hexane (n-hexane)^{Footnote 15}

Geographical scope for other chemicals

The processes are modelled using Canadian and American data. They can be used to represent the production in North America. 

Allocation for other chemicals

No allocation procedures were performed for the other chemicals.

References

Bansod, Y., Crabbe, B., Forster, L., Ghasemzadeh, K., & D’Agostino, C. (2024). Evaluating the environmental impact of crude glycerol purification derived from biodiesel production: A comparative life cycle assessment study. Journal of Cleaner Production, 437, 140485.

Brinkmann T, Giner Santonja G, Schorcht F, Roudier S, Delgado Sancho L. Best Available Techniques (BAT) Reference Document for the Production of Chlor-alkali. Industrial Emissions Directive 2010/75/EU (Integrated Pollution Prevention and Control). EUR 26844. Luxembourg (Luxembourg): Publications Office of the European Union; 2014. JRC91156

Blumberg, Timo, Morosuk, Tatiana, & Tsatsaronis, George. A Comparative Exergoeconomic Evaluation of the Synthesis Routes for Methanol Production from Natural Gas. Appl. Sci. 2017, 7(12), 1213.

Environment and Climate Change Canada. Greenhouse Gas Reporting Program (GHGRP) - Facility Greenhouse Gas (GHG) Data – Ammonia production (2019-2020-2021-2022).

Granjo, J. F., & Oliveira, N. M. (2015). Process simulation and techno-economic analysis of the production of sodium methoxide. Industrial & Engineering Chemistry Research, 55(1), 156–167.

Hamelinck, Carlo & Bunse, Mark. January 2022. Carbon Footprint of Methanol (Underlying Data). Studio Gear Up, prepared for The Methanol Institute Carbon Footprint of Methanol.

Lee, D-Y., Elgowainy, A., & Dai, Q. (2017). Life Cycle Greenhouse Gas Emissions of By-product Hydrogen from Chlor-Alkali Plants. Argonne National Laboratory.

Nemecek, Thomas & Kägi, Thomas. Agrosope Rechenholtz Tänikon Research Station (ART). Life Cycle Inventories of Agricultural Production Systems (data v2.0 [2007]). Ecoinvent Report No. 15. December 2007.

The R&D Greenhouse gases, Regulated Emissions, and Energy use in Technologies Model.  Argonne National Laboratory. (R&D GREET 2022). Argonne GREET Model (anl.gov).^{Footnote 5}

The R&D Greenhouse gases, Regulated Emissions, and Energy use in Technologies Model.  Argonne National Laboratory. (R&D GREET 2023 Revision 1). Argonne GREET Model (anl.gov).^{Footnote 16}

Wang, J., Cui, Z., Li, Y., Cao, L., & Lu, Z. (2020). Techno-Economic Analysis and environmental impact assessment of citric acid production through different recovery methods. Journal of Cleaner Production, 249, 119315.