Techno-economic assessment of atmospheric CO₂-based carbon fibre production enabling negative emissions

Abstract

The fight against global warming requires novel approaches for the defossilisation of industrial processes, and the limitation of global warming requires options for negative carbon dioxide (CO₂) emissions. The production of carbon fibre (CF) is an energy-intensive chain of processes which cause CO₂ emissions. Having in mind the high market growth for CF composite materials, CF production might stand against the fight against global warming. CF also offers a huge mitigation opportunity, as CF contain up to 95–98wt% of pure carbon. This study investigates possible ways to link CF production to atmospheric CO₂, enabling negative CO₂ emissions through CF manufacturing. Production value chains for CF based on poly(acrylonitrile) (PAN) and pitch, the two most important CF precursor materials, are developed and analysed regarding their energy and mass balances. The PAN value chain is further assessed regarding a first economic estimation of CF production cost with atmospheric CO₂ as carbon source. The results show that production costs per ton CO₂ removed might be unattractive at 2949 €/tCO₂ in 2050. However, from a CF perspective, production cost of 10.3 €/kgCF in 2050 might enable a business case for electricity-based CF production from atmospheric CO₂ in the future. Each ton of CF produced can store about 3.5 tCO₂ due to a very high carbon share in the final product. With an increasing market for CF, a total negative emission potential of at least 0.7 GtCO₂ per year can be enabled by 2050. Further research opportunities are discussed.

1 Introduction

Limiting global warming to 1.5 °C above the pre-industrial level as agreed upon in the Paris Agreement (UNFCCC 2015) requires an estimated 150–1180 GtCO₂ negative emissions within the twenty-first century for different scenarios (Rueda et al. 2021) as defined by the Intergovernmental Panel on Climate Change (IPCC). If aiming for a rebalancing of earth’s climate and limiting global warming to about 1.0 °C, which correlates to reduce the concentration of carbon dioxide (CO₂) in the atmosphere back to about 350 ppm, a total requirement for negative emissions of about 1483 GtCO₂ can be estimated (Breyer et al. 2021). Such a rebalancing will only be possible by massively scaling up carbon dioxide removal (CDR) via negative emission technologies (NET).

Research in possible NET portfolios has gained interest in recent years (Rueda et al. 2021). Commonly discussed technologies are afforestation and/or reforestation, biochar, bioenergy carbon capture and storage (BECCS), direct air capture and carbon storage (DACCS), enhanced weathering, ocean fertilisation, and soil carbon sequestration (Fuss et al. 2016; Minx et al. 2018; Osman et al. 2021; Park et al. 2022). By qualifying possible long-term and highly secure storage options for CO₂, three major requirements shall be fulfilled: conversion of gaseous CO₂ into a solid-state material; high chemical inertia to avoid decomposition followed by a potential release of CO₂ into the atmosphere, as well as a high combustion point to avoid the possibility of burning the material in which the CO₂ is bonded, either by accident or on purpose. The only storage option for which these fundamental requirements have been confirmed is the mineralisation of CO₂ to solid rock (Baciocchi and Costa 2021), usually applied by BECCS and DACCS, as shown on the case of DACCS (Ratouis et al. 2022).

However, there are also carbon-based materials used in industry containing carbon and showing high chemical inertia as well as high combustion points. Most interesting materials which currently see a high market growth with a compound annual growth rate (CAGR) of approximately 6% and a global market potential of almost 200,000 kt/a until 2050 (Zhang et al. 2020) are carbon fibre composite materials (CFCM). Their basic material carbon fibres (CF) gain high interest, having a widespread area for potential use due to its low weight, high chemical resistance, and high temperature resistance, among others (Bhatt and Goe 2017; Zhang et al. 2020). From a negative emission point of view, CFs are very interesting due to their high density of carbon atoms, as modern CFs are able to be made out of almost 100wt% carbon (Frank et al. 2014). In the context of this study, unwoven basic CF is reflected upon.

One argument against CF is their energy-intensive production process and carbon footprint. Production processes for CF can be characterised by long processing times and processes requiring high-temperature heat (Dér et al. 2021). Nevertheless, the gate-to-gate energy intensity of CF manufacturing in literature underlies extreme variations of 2.1–132.8 kWh per kg CF produced, with extreme outliers of up to 319.6 kWh per kg CF (Dér et al. 2021), most likely due to quite different system boundaries. Especially due to a high demand of natural gas (NG) for exhaust gas cleaning, the carbon footprint of each kg CF produced can be 24.4–31.0 kgCO₂/kg CF (Tapper et al. 2020).

Arnold et al. (2018a, 2018b) made a first attempt to model the production of CF via algae-based carbon fixation. In contrast, the aim of this paper is to adapt common production routes of CFs to enable negative emissions by using atmospheric CO₂ as carbon source, captured via direct air capture (DAC) units (Fasihi et al. 2019). Part of the investigation is to define energy and mass balances of the processes and to estimate the cost per ton CO₂ removed from the atmosphere and per kg of produced CF. Results are presented for a system setup in the years 2030, 2040, and 2050 including respective techno-economic input data. The insights presented can be further used in climate energy system modelling to account for negative CO₂ emissions from the atmosphere by CF production, as well as an alternative production route for carbon–neutral electricity-based CF (e-CF) solely based on air, water, and electricity, leading to a power-to-CF (PtCF) transformation process, as a specialised power-to-X conversion (Sterner and Specht 2021).

2 Value chains of conventional precursor and carbon fibre production

Currently, CF are produced by two main precursors, which are poly(acrylonitrile) (PAN) and pitch (Frank et al. 2014). Other precursors gaining interest are biomass-based CF obtained from cellulose or lignin, or plastics like poly(ethylene) (PE) (Frank et al. 2014). Even though other process routes apart from PAN are gaining interest, at least 90% of the current CF production is based on PAN precursor fibres (Nunna et al. 2019). CF based on cellulose, lignin, PE, or other precursor fibres might show a gain in interest in recent years; however, the quality of the CF produced and the availability of data and literature limit the possibility for more detailed investigations. An overview of conventional production routes for CF can be seen in Fig. 1.

Fig. 1

Simplified conventional production value chains for CF based on fossil propane or propene via PAN precursor (left) and fossil ethene via pitch precursor (right). The carbonisation process includes all steps of thermal fibre treatment (stabilisation/oxidation, carbonisation/pyrolysis)

PAN is produced in two basic steps. Step one is the vapor-phase propylene ammoxidation process (further referred to as ammoxidation) developed by SOHIO (Centi et al. 1992; Brazdil 2012; Cespi et al. 2014) of propene (C₃H₆) or propane (C₃H₈), ammonia (NH₃), and oxygen (O₂) to acrylonitrile (AN, C₃H₃N). However, the oxygen is only required for propane ammoxidation. Within a fluidised bed reactor, the reactants contact a solid catalyst at a temperature of 400–510 °C and a pressure of 30–200 bar (Brazdil 2012; Cespi et al. 2014). Apart from AN as an intermediate substance and PAN as the desired precursor substance, the ammoxidation also produces coproducts, which are useful for further applications in the chemical industry, as a solvent or fertiliser (Brazdil 2012). However, the coproducts are not of further interest in the context of this study. In a second step, the produced AN is polymerised in a free radical emulsion copolymerisation of AN and methyl acrylate (C₄H₆O₂) as copolymer with 97/3wt% shares (Morris et al. 2014). The process includes stirring at 25 °C and 65 °C, as well as drying at 100 °C (Morris et al. 2014). Carbonisation of the spun filament, which consists of a first thermal stabilisation at 200–300 °C for 0.5 to 2 h and in the second stage of a short high-temperature treatment at 1000–1400 °C for several minutes (Das 2011; Nunna et al. 2019) is followed by some additional surface treatment. As literature is generally scarce in techno-economic specifications of the CF production from PAN, this study focuses on the production on high tenacity CF that are relevant to industry and are produced at a rather moderate carbonisation temperature at 1200–1400 °C (Arnold et al. 2018a; Groetsch et al. 2023; Nunna et al. 2019).

The pitch route also consists of two basic steps. In the first step, ethene (C₂H₄) is fed to the chlorination process, where chlorine gas (Cl₂) and oxygen is added (Saeki and Emura 2002). At 500 °C, a vinyl chloride monomer (VCM) is formed. This monomer is further processed to polyvinyl chloride (PVC, (C₂H₂Cl)_n) via suspension or bulk polymerisation at 50–70°C (Saeki and Emura 2002). Currently, suspension polymerisation is the industry standard covering about 80% of PVC produced globally (Saeki and Emura 2002). The PVC is further processed to a (mesophase) PVC-based pitch by pyrolysis and further heat treatment (Qiao et al. 2004). In order to initiate the pyrolysis, the PVC is treated at 260 °C for 2 h, then the heat treatment at 430 °C for 1–2 h transfers the pitch into mesophase. After spinning to a fibre, stabilisation at up to 320 °C, carbonisation at up to 900 °C, and activation (surface treatment) with steam at 900 °C for 0.5–1.5 h follows (Qiao et al. 2004).

In order to deal with toxic and harmful exhaust gases occurring in the stabilisation and carbonisation of the fibres based on PAN, such as hydrogen cyanide (HCN), methane (CH₄), carbon monoxide (CO), ammonia, and different hydrocarbons, the exhaust gases are processed in a regenerative thermal oxidation (RTO) unit (Arnold et al. 2018a, b; Dér et al. 2021). Additionally, construction materials should be considered carefully for the pitch route in case of chlorine containing process streams. Via combustion of NG, those substances are neutralised via oxidation in the combustion process. Pyrolytic treatment of PVC also causes polluting substances, such as hydrochloric acid (HCl) and sulfur dioxide (SO₂) (Qiao et al. 2004, 2006).

3 Production of electricity-based carbon fibres from atmospheric CO₂

Principally, the value chains for precursor production stay the same. However, the feedstock hydrocarbons propene and ethene can be supplied in a novel way. In the following, the principle of e-CF production from atmospheric CO₂ shall be outlined and techno-economic parameters estimated. For both procedures, the carbon source is atmospheric CO₂, which can be captured from the air by a DAC unit (Fasihi et al. 2019). All techno-economic calculation steps are available in detail in the supplementary material for the PAN-based production route.

3.1 CO₂ and electricity-based carbon fibre production via PAN precursor

Gaseous CO₂ can be transformed to methanol (MeOH) by a methanol synthesis unit (IRENA and Methanol Institute 2021; Chen and Yang 2021). In addition to CO₂, hydrogen is needed, which can be provided by an electrolyser (Fasihi and Breyer 2020). Furthermore, the required ammonia for the ammoxidation step can be produced based on renewable electricity by a Haber–Bosch synthesis unit (Fasihi et al. 2021). At a temperature of 350–550 °C and a pressure of 100–250 bar, ammonia is synthesised from nitrogen (N₂) and hydrogen, where nitrogen (N₂) can also be captured from the air by an air separation unit (ASU) (Fasihi et al. 2021).

Afterwards, there are two different ways of producing propene as source material for further utilisation: by a methanol-to-olefine (MTO) process (Tian et al. 2015; Dimian and Bildea 2018; Dutta et al. 2019; Zhao et al. 2021), or by a methanol-to-propylene (MTP) process (Jasper and El-Halwagi 2015; Zhao et al. 2021). Both options are principally the same process, but vary in the ratio of propylene (also known as propene) and sort of coproducts produced (Zhao et al. 2021). MTO is an exothermal reaction at around 495 °C and produces the coproducts liquified petroleum gas (LPG), gasoline, ethylene, and water (Tian et al. 2015; Dimian and Bildea 2018; Dutta et al. 2019; Zhao et al. 2021). Commercial MTP process (Lurgi MTP™) from methanol to propene is via dimethyl ether using the zeolite catalyst, and MTP has a process temperature of 400–500 °C (Zhao et al. 2021). As propene is the substance of interest in this case, the MTP process is chosen for further consideration. If the MTP process is seen as part of the CF production with methanol as the basic source material, the whole process including MTP, ammoxidation (SOHIO), polymerisation, and CF manufacturing can be further simplified and condensed to a methanol-to-CF process with the PAN precursor, more simplified written as (MeOH-to-CF)_PAN (cf. Figure 2).

Fig. 2

Simplified value chain and process scheme of carbon fibre production based on atmospheric CO₂ and intermediate methanol via PAN as precursor

All processes except the RTO require electricity as input. DAC additionally requires low-temperature heat at about 100 °C. Within the (MeOH-to-CF)_PAN process cluster, the ammoxidation produces heat at ~ 450 °C, which is assumed to be able to fully cover the heat demand of the MTP process at the same temperature level. Furthermore, the remaining heat surplus is used to partially cover the low-temperature heat demand required by the polymerisation of acrylonitrile. In theory, waste heat from the RTO, which would be available at around 850 °C, could be used to cover the heat demand of the thermal stabilisation of the CF. However, due to the lack of specific data on this opportunity, the full process heat, mainly the carbonisation heat at around 1200–1400 °C is covered by electricity via electric heaters. Low-temperature heat of up to 100 °C is assumed to be covered by a heat pump.

Mass and energy balances of each process are obtained either from available process data in literature or from the theoretical reactions via molar masses of the substances. A final line-up of the separate processes including energy requirement and mass balances per unit output can be found in Table 2 for the auxiliary processes and in Table 3 for the subprocesses of the (MeOH-to-CF)_PAN process cluster in the appendix.

3.2 CO₂ and electricity-based carbon fibre production via pitch precursor

Pitch can also be produced from methanol as the basic material. In this case, the MTO process is the more favourable option over MTP, as for the pitch production, ethylene (also known as ethene) is the substance of interest. The whole pitch-based process including MTO, chlorination, polymerisation, pyrolysis, and heat treatment, and finally CF manufacturing can be further simplified written as (MeOH-to-CF)_pitch process (cf. Figure 3). Oxygen required for the chlorination can be covered by available oxygen from the electrolysis. Pitch offers a second value chain based on another carbon carrier. Instead of methanol as the carbon carrier, methane can be used as the basic material. To use methane, though, an additional process step, the so-called oxidative coupling, is necessary, converting methane to ethene (Zhao et al. 2021). This step requires additional oxygen, which also can be covered by the available oxygen from the electrolysis. The process requires heat at 650–950 °C (Ren et al. 2008; Wang et al. 2017). Oxidative coupling, however, has a major drawback, which is the low conversion yield of methane to ethylene of 25–50% besides high amount of CO₂ and CO produced, e.g., due to poor selectivity (Ren et al. 2008; Alvarez-Galvan et al. 2011; Hammond et al. 2012). Currently, higher temperatures enable higher conversion efficiencies that increase the energy consumption (Ren et al. 2008). The whole value chain of the pitch-based CF production based on methanol and methane can be seen in Fig. 3.

Fig. 3

Simplified value chain and process scheme of carbon fibre production based on atmospheric CO₂ and intermediate methanol or methane via pitch as precursor

DAC, electrolysis, methanation, and methanol synthesis are in principle equal to the PAN route explained in the previous subsection. In case of CF production based on pitch as a precursor, the information density about process details is relatively weak. Additionally, no specifics about the waste treatment of stabilisation and carbonisation residues are available. Substances like water, CO₂, carbon monoxide, nitrogen, methane, and sulfur dioxide (Huson 2017) may occur in addition to hydrochloric acid (HCl), which is a result of pyrolytic treatment of PVC (Qiao et al. 2004, 2006). Electric heaters provide the heat for the carbonisation of the fibres with up to 900°C.

Apart from the oxidative coupling for the (CH₄-to-CF)_pitch route, all process steps are equal to the conventional route as described in Section. 2. Similar to the (MeOH-to-CF)_PAN route, low-temperature heat for the suspension polymerisation of VCM to PVC can be provided by a heat pump. A possible re-use of heat from exothermic reactions within the (MeOH-to-CF)_pitch and (CH₄-to-CF)_pitch process clusters cannot be quantified as well at this point. Mass and energy balances for the (MeOH-to-CF)_pitch and (CH₄-to-CF)_pitch routes can be found in Table 4 in the appendix. However, data availability for pitch-based CF manufacturing is scarce, especially for the pitch-based CF production specifics, and the full process portfolio data could not be collected.

3.3 Economics of electricity-based carbon fibre production

Economic data are only comprehensively available for CF production via PAN precursor to allow an estimation of the production cost for e-CF, in contrast to pitch-based CF production. Equation (1) is used to calculate the total levelised cost of carbon dioxide removal LCOCDR

$$LCOCDR=\sum_p^{proc}\left[\left({LCOP}_p\cdot m_p\right)+H_{LT,p}\cdot{LCOH}_{LT,p}+\left(E_{el,p}+\eta_{HT}\cdot H_{HT,p}\right)\cdot{cost}_{el}\right]$$

(1)

including the total mass output of each process m_p required to remove 1 tCO₂ from the atmosphere, low-temperature heat H_LT to be multiplied by the levelised cost of low temperature heat LCOH_LT occurring by using a heat pump, as well as the electricity E_el and the high-temperature heat H_HT to be multiplied with the electricity cost cost_el. High-temperature heat is assumed to be covered by direct electric heating. The cost for the heaters themselves is assumed to be included in the capital expenditures capex of the process equipment, and the efficiency of the electric heater η_HT is assumed to be 100%. Costs for compressors of processes requiring high pressures are also assumed to be included. If this cost calculation would be made for the pitch route, a material factor for chlorine-resistant materials would have to be included in the capex. The capital recovery factor crf is calculated by Eq. (2):

$$crf=\frac{WACC\cdot\left(1+WACC\right)^N}{\left(1+WACC\right)^N-1}$$

(2)

where N is the lifetime of each technology. The weighted average cost of capital WACC is set to 7% for all processes and years based on a 70/30 debt to equity ratio, 4% interest on debt and 14% return on equity (Vartiainen et al. 2019). Costs for processes with a given capex are based on the annual output capacity, the levelised cost of process LCOP is calculated according to Eq. (3):

$$LCOP=\frac{capex\cdot\left(crf+{opex}_{fix}\right)\cdot capacity}{{out}_p}+{opex}_{var}$$

(3)

where opex_fix are the fixed operational expenditures and opex_var are the variable operational expenditures per unit output. The output of the process out_p is defined by Eq. (4) including the annual capacity and availability factor τ:

$${out}_p=capacity\cdot\tau$$

(4)

In the base case, τ is set to 0.95, representing a baseload operation of all processes at 95% of the year. A respective sensitivity analysis assessing the impact of τ on the total process cost is carried out and presented in the results.

LCOP for transformers with a given capex based on installed capacity or energy unit output are calculated via the full load hours FLH with Eq. (5):

$$LCOP=\frac{capex\cdot\left(crf+{opex}_{fix}\right)}{FLH\cdot\tau}+{opex}_{var}$$

(5)

In case of the electrolyser, the storage cost for hydrogen is added to the opex_var (cf. Table 1). Furthermore, the LCOH_LT of the heat pump are calculated with Eq. (6):

$${LCOH}_{LT}=\frac{capex\cdot\left(crf+{opex}_{fix}\right)}{FLH\cdot\tau}+{opex}_{var}+\frac{{cost}_{el}}{COP}$$

(6)

and includes the cost for electricity to power the heat pump, divided by the coefficient of performance COP.

Likewise, the levelised cost of carbon production LCOCF can be calculated similarly to the LCOCDR using Eq. (7):

$$LCOCF=\frac{LCOCDR}{{m}_{p,CF,1tCO2}}$$

(7)

The LCOR are normalised to 1 tCF produced by dividing the necessary mass output of each process by the mass of CF produced per 1 tCO₂ removed m_p,CF,1tCO2. The relation given in Eq. (8) is the amount or potential of CO₂ that can be stored in one unit of CF produced (CDR potential) CDRpot,_CF.

$${CDR}_{pot,CF}=\frac{1}{{m}_{p,CF,1tCO2}}$$

(8)

The economic analysis is made for the years 2030, 2040, and 2050, as the production of CF by the novel approach does not yet exist. Numbers for all subprocesses are not always available presenting current costs. It is assumed that established chemical processes are mature technologies with no further cost reductions until 2050. Numbers taken from literature are converted to €₂₀₂₀ based on the year and currency used in the respective publication (Alioth Finance 2022). An overview of all cost input data and necessary technological specifications can be found in Table 1. The table only contains relevant cost numbers for PAN-based CF production. While a bigger variety of input sources would be beneficial for a higher confidence on input data for the (MeOH-to-CF)_PAN processes, a general lack of data does not allow for more varied references. The available energy and mass balances and cost numbers especially for the CF manufacturing process based on a pitch precursor do not allow a proper estimation of production cost. Nevertheless, available numbers that could be obtained from literature for the sub-processes can be found in the appendix in Table 5 for general information.

Table 1 Available economic input data for all auxiliary and sub-processes of the (MeOH-to-CF)_PAN, process route for the years 2030, 2040, and 2050. All cost numbers are inflation-adjusted to €₂₀₂₀ real values

Furthermore, the cost per produced kgCF and tCO₂ removed from the atmosphere is assessed in a cost sensitivity analysis. Numbers provided in Table 1 are the reference values. For hydrogen the higher heating value (HHV) of 39.41 kWh_H2,HHV/kgH₂ and the lower heating value (LHV) of 33.33 kWh_H2,LHV/kgH₂ is used for conversions from energetic content of the produced hydrogen to its mass equivalent. The cost assumptions for electricity are based on the fact that large-scale e-CF manufacturers, similar to other electricity-based product manufacturers in highly competitive markets, will most likely obtain their energy from direct contracting instead of volatile and unpredictable wholesale markets. Especially large-scale electricity-based chemical plants will either use direct contracting or will be built completely or partially off-grid, as assumed for example in Fasihi et al. (2021). Very low-cost solar PV based electricity can be observed in more and more countries around the world (IEA-PVPS 2022), while the PV learning curve remains stable leading to ongoing cost reductions (OPIS 2023).

4 Results and discussion

First, the energy and mass balances for the PAN value chain are presented in Section. 4.1. In Section. 4.2, the cost and cost sensitivities of the value chain are assessed. Section 4.3 finally discusses results and further research opportunities.

4.1 Energy and mass balances

The energy and mass balance for PAN-based e-CF manufacturing is presented based on Fig. 2 for the (MeOH-to-CF)_PAN value chain. Figure 4 shows the balance overview for the e-CF production with PAN as precursor based on methanol for the year 2030.

Fig. 4

Simplified value chain, process scheme and energy and mass balances of e-CF production based on atmospheric CO₂ and intermediate methanol via PAN as precursor in the year 2030. Energy and mass balances are normalised to 1 tCO₂ removed from the atmosphere. Co-products that are not related to the production of e-CF are left out from the value chain for simplification

Normalised to 1 tCO₂ removed from the atmosphere, 0.287 tCF can be produced. Applying Eq. (8), this indicates a total of about 3.5 tCO₂ which can be removed from the atmosphere per ton e-CF produced. Due to the very high carbon share of up to 95–98 wt% in CF, this value is very close to the theoretical maximum of 3.66 tCO₂ per ton carbon. Presuming about 200,000 ktCF/a market potential until 2050 (Zhang et al. 2020), this would indicate a possible 0.7 GtCO₂/a negative emission potential until mid-century. In case markets would grow stronger due to a faster increase in demand this negative emission potential could be further increased.

A total of 56.9 MWh_el is needed to remove 1 tCO₂ or 198.3 MWh_el per 1 tCF produced (198.3 kWh_el/kgCF). The whole value chain for e-CF production has been electrified. With the change from fossil fuel-based CF to electricity-based CF a value chain for power-to-carbon fibre (PtCF) to produce e-CF has been created. This includes all necessary energy, which means direct electricity demand, electricity for direct electric heaters, heat pump and electricity for e-methane production used in the exhaust gas treatment. Therefore, the total energy intensity of this novel approach lies within a common range of up to 319.6 kWh/kgCF reported by Dér et al. (2021), though higher than the majority of available literature values up to 132.8 kWh/kgCF. Notably, the use of heat pumps to provide low temperature heat for the polymerisation of AN can be seen as a huge efficiency improvement compared to direct heating with electricity; however, it induced a higher capex. It is also estimated that a total of 1368.4 kWh_th/tCO₂ of high-temperature heat from the exothermic ammoxidation process can be used for the MTP and partially the polymerisation process. In 2040, the total electricity demand decreases to 55.2 MWh_el/tCO₂ (192.1 kWh_el/kgCF), and in 2050 to 53.7 MWh_el/tCO₂ (186.8 kWh_el/kgCF). At this point of research, the decrease in energy intensity is caused by efficiency improvements of auxiliary processes solely. The total low-temperature heat for DAC decreases from 6203.2 kWh_th/tCO₂ in 2030 to 5318.2 kWh_th/tCO₂ in 2040 and to 4557.3 kWh_th/tCO₂ in 2050, while the heat demand in general for the (MeOH-to-CF)_PAN process cluster stays the same. The respective electricity consumption of the heat pump in 2030 of 4037.4 kWh_el/tCO₂ decreases to 3600.2 kWh_el/tCO₂ (2040) and 3280.9 kWh_el/tCO₂ (2050).

Besides, the (MeOH-to-CF)_PAN value chain itself, the most energy or rather electricity-intensive process is the electrolysis. The electrolyser feeds three different processes with necessary hydrogen. For e-methane, a relatively high amount of hydrogen is needed, as each methane molecule comprises four hydrogen atoms for each carbon atom. It is similar for e-ammonia, where each ammonia molecule comprises three hydrogen atoms per nitrogen atom. The chemical formula of methanol can be either CH₃OH or CH₄O. Therefore, again four hydrogen atoms are needed to synthesise one methanol molecule. The energy equivalent of the total hydrogen produced is 9673.8 kWh_H2,HHV/tCO₂ in 2030, 2040 and 2050, even though the electricity demand to produce this amount of hydrogen decreases from 33.1 MWh_el/tCO₂ in 2030 to 31.8 MWh_el/tCO₂ in 2040 and to 30.7 MWh_el/tCO₂ in 2050 due to increasing electrolyser efficiency.

To remove 1 tCO₂ from the atmosphere, a total of 4.135 tCO₂ must be captured by DAC units. However, 1.544 tCO₂ that go to the methanation and further to the RTO are not to be counted as direct emissions, as the e-methane cycle can be assumed to be fully carbon neutral.

4.2 Cost structure for carbon removal and electricity-based carbon fibre production

First, the cost structure per tCO₂ captured and kgCF produced is analysed without energy cost in order to assess the plain process cost. The respective cost structure for the years 2030, 2040 and 2050 is shown in Fig. 5.

Fig. 5

Process cost structure excluding energy cost for PAN-based e-CF production via (MeOH-to-CF)_PAN process cluster normalised to 1 tCO₂ removed from the atmosphere (left) and 1 kgCF produced (right) for the years 2030 (top), 2040 (centre), and 2050 (bottom). Abbreviation: synth.: synthesis

In case of the auxiliary processes, the electrolysis is the most expensive process in the value chain, as already mentioned in Section. 4.1 due to the relatively high amount of hydrogen needed for three different synthesis processes. It is also the process with one of the highest cost reduction potentials, reducing its cost by 27% from 2030 to 2050. As no energy costs are included this is due to a favourable capex development of electrolysers in the decades to come. The highest relative cost reduction can be noticed for the DAC unit, with a reduction by 44% from 2030 to 2050, caused by a combination of capex development and lifetime improvement. As no changes in economic values for the (MeOH-to-CF)_PAN process cluster has been implemented, the plain process cost stays the same in all cases. In total, the process cost decrease from 2507 €/tCO₂ (8.73 €/kgCF) in 2030 to 2364 €/tCO₂ (8.23 €/kgCF) in 2040 and to 2316 €/tCO₂ (8.06 €/kgCF) in 2050. Therefore, the total process cost reduction within the investigated time frame is 7.6%.

For all three years, the (MeOH-to-CF)_PAN process costs have a higher share than the auxiliary processes. In 2030, the share of those process costs in the total process cost is 74.9%. Naturally, this number increases until 2050 as no cost reduction for (MeOH-to-CF)_PAN processes are assumed and auxiliary processes see decreasing capex and partial increase in lifetime over the years. In 2040, the share is 79.5% and 81.2% in 2050.

The costs to produce the PAN precursor are also of particular interest. For this purpose, only the cost share for producing PAN is considered, leaving out processes not directly included in the PAN production chain, which include DAC for methanation, methanation itself, and electrolyser cost for methanation. The total PAN production cost is 1710 €/tCO₂ or 5.95 €/kgCF at a share of 68.2% in 2030, 1623 €/tCO₂ or 5.90 €/kgCF at a share of 71.7% in 2040, and 1591 €/tCO₂ or 5.88 €/kgCF at a share of 72.9%.

Figure 6 shows the full cost structure including all energy costs occurring, as well as the cost balance after accounting for e-CF sales value and CO₂ compensation for negative emissions.

Fig. 6

Process cost structure including energy cost, CF sales, and CO₂ compensation for PAN-based e-CF production via (MeOH-to-CF)_PAN process cluster normalised to 1 tCO₂ removed from the atmosphere (left) and 1 kgCF produced (right) for the years 2030 (top), 2040 (centre), and 2050 (bottom). Abbreviations: synth.: synthesis; comp.: compensation

From a negative emission point of view, the production of e-CF seems to be unattractive at first sight. The cost for sequestering 1 tCO₂ from the atmosphere in e-CF amounts to 3767 €/tCO₂ in 2030, 3299 €/tCO₂ in 2040, and 2949 €/tCO₂ in 2050. The cost increase compared to the process cost without accounting for energy is 50.3% in 2030, 39.6% in 2040 and 27.3% in 2050. Energy has a share in total production cost of 33.5% in 2030, 28.3% in 2040 and 21.5% in 2050.

(MeOH-to-CF)_PAN processes have a lower share in total production cost if energy is considered. The reasons are the efficiencies of the electrolyser and synthesis units, as well as the heat of the polymerisation, which can be supplied very efficiently via the heat pump. In 2030, the share in total production cost of the mentioned processes is 56.6%, in 2040 63.0%, and in 2050 68.9%.

However, the main purpose of e-CF production might not be to create negative emissions, but to use e-CF for various applications. Therefore, the market value of CF must be considered. If the sales income of CF is considered, the production of e-CF by the presented novel approach is already economically attractive in 2030, creating a net profit with higher sales profits than cost. Enabling negative emissions might also be compensated in the future by paying a fee for negative emissions. If those are also included, the presented approach leads to a net benefit of 557 €/tCO₂ in 2030. Assuming a relatively stable CF market price, this net benefit can be increased to 1111 €/tCO₂ in 2040 and more than doubled from 2030 to 2050 to 1461 €/tCO₂.

Having a look at the cost structure from a e-CF production point of view and re-normalising the cost to €/kgCF, the total production cost in 2030 can be estimated to 13.1 €/kgCF. This can be reduced to 11.5 €/kgCF in 2040. Until 2050, CF can be produced almost at cost of 10 €/kgCF with 10.3 €/kgCF. The portion of CO₂ compensation for each kgCF is 0.47 €/kgCF in 2030 and 0.77 €/kgCF in 2040 and 2050. The total net benefit for produced CF is 1.94 €/kgCF in 2030, 3.87 €/kgCF in 2040, and 5.08 €/kgCF in 2050. If carbon compensation is included, the total profit can therefore be as high as 49.3% of the total production cost at the given CF value.

The total PAN precursor production cost including energy for each kg of e-CF amounts to 1852 €/tCO₂ or 6.44 €/kgCF at a share of 49.1% in 2030, 1740 €/tCO₂ or 6.06 €/kgCF at a share of 52.8% in 2040, and 1687 €/tCO₂ or 5.87 €/kgCF at a share of 57.2% in 2050. The total production costs for the PAN precursor are 14.2 €/kgPAN in 2030, 13.3 €/kgPAN in 2040, and 12.9 €/kgPAN in 2050.

4.3 Sensitivity analysis

The sensitivity of the LCOCDR and LCOCF are of high interest, as some parameters, especially for the (MeOH-to-CF)_PAN process cluster, may be of higher uncertainty. Figure 7 shows a sensitivity analysis for changing four different parameters relative to the reference case as used in the default analysis presented in previous sections and the respective relative change for LCOCDR and LCOCF.

Fig. 7

Sensitivities of LCOCDR and LCOCF relative to the reference value for changing electricity cost (top left), changing capex of the (MeOH-to-CF)_PAN cluster processes MTP, ammoxidation, polymerisation and CF manufacturing (top right), changing lifetime of the (MeOH-to-CF)_PAN cluster processes (bottom left) and availability factor of all processes (bottom right)

The process is most vulnerable to electricity cost changes in 2030. However, this does not come as a surprise, as 2030 has the highest electricity cost and less efficient processes compared to the following time steps. If the electricity cost would be halved the LCOCDR and LCOCF would be about 87% of the reference value in 2030, 89% in 2040, and 93% in 2050. The correlation is linear, so the increase of LCOCDR and LCOCF for 50% higher electricity cost would be 113%, 111%, and 107% in 2030, 2040, and 2050, respectively. Chances are high that the electricity cost in combination with electrolyser cost reductions can outperform the assumed reference cost in this study, as dedicated solar hydrogen may reduce the hydrogen cost faster than projected (Vartiainen et al. 2021).

Capital expenditures of the (MeOH-to-CF)_PAN process cluster do not yet underly changes within the investigated time frame, but might change until 2050. The sensitivity analyses show an inverted characteristic as for the electricity cost change, i.e., the 2050 results are more vulnerable to the changes of capex and the 2030 results the least. By halving the capex of the (MeOH-to-CF)_PAN processes, the LCOCDR and LCOCF can be reduced to 73%, 70%, and 67% in 2030, 2040, and 2050, respectively. The electricity cost changes also have a linear relation, meaning that 150% of the capex leads to LCOCDR and LCOCF of 127%, 130%, and 133% in 2030, 2040, and 2050, respectively.

Noticeably different is the relation between the lifetime of the (MeOH-to-CF)_PAN processes and the results. The relation shows an exponential relation, so the effect of lower lifetimes is higher than for higher lifetimes. If the lifetime would only be half of the ones currently used in the reference case, the production costs would increase to 121% in 2030, 124% in 2040, and 127% in 2050, respectively. However, for higher lifetimes, the cost decrease is smaller. In 2030, the cost could be reduced to 94% of the reference case, to 93% in 2040, and 92% in 2050.

Finally, the availability of the process components has been subject to the sensitivity analysis. In the reference case, it was assumed that all processes are available 95% of the year, corresponding to a baseload operation. By changing the availability factor, a very high vulnerability of the cost results can be noticed. If all processes would run only 50% of the year the price per tCO₂ or per kgCF would increase to 169%, 173%, and 178% in 2030, 2040, and 2050, respectively. Even small changes already demonstrate a significant impact. Only 10% less availability causes a cost increase of 8–9%. As typical for industrial processes, most of the units will have to be used in baseload operation. Electrolysers on the other hand will be able to act as a flexibility option in future energy systems with a high share of renewable energy. Respective investigation will have to be part of future research. The variability of the input electricity from solar photovoltaics and wind power is fully mitigated by flexible electrolysers and a hydrogen buffer storage (Fasihi and Breyer 2020), so that all subsequent processes can be run for baseload conditions.

Important for the assertiveness of the novel approach is the profitability of the product. Figure 8 shows a sensitivity analysis for the profitability of the final product if the sales prices for the CF varies, or rather if the product is able to generate profit when sold.

Fig. 8

Sensitivities of the net profit relative to the reference value for changing the CF sales price in 2030, 2040, and 2050 with and without included carbon compensation. Abbreviations: comp.: compensation

The analysis shows that the e-CF produced with the novel approach is most vulnerable to CF sales prices in 2030. If carbon compensation is not paid, the e-CF production is unprofitable for a CF sales price 10% below the reference sales price assumed in this study. In 2040, the CF sales price could be reduced by about 22%, and in 2050 by 29%. However, if carbon compensation is paid for negative emissions, the CF sales price could drop by about 13% in 2030, 27% in 2040, and almost 35% in 2050, still generating net profit.

4.4 Discussion and research outlook

The scarcity of detailed process data remains a challenge for a detailed techno-economic modelling of the whole value chain. Although data are available for the e-CF production via PAN, currently the main precursor used in industry, and the following CF manufacturing, a more diverse literature would be very beneficial. For the nowadays less interesting production routes such as CF produced based on the pitch precursor, the availability of detailed data is insufficient to allow a techno-economic modelling with respective implications. The scarcity of available data relates both to the technical data for the processes, meaning detailed energy demand and mass balances of the real processes including real conversion rates and efficiencies, as well as economic data for the processes steps. The defossilisation of the chemical industry will play a major role in fighting climate change (Kätelhön et al. 2019; Galán-Martín et al. 2021; Lopez et al. 2023), and thorough techno-economic modelling of alternative production routes is indispensable for the contribution of the chemical industry to the energy transition in general.

Whether the production of e-CF is beneficial as a negative emission technology depends largely on the CF market price. At current prices of about 14.2 €/kgCF, the novel e-CF production route proves to enable an attractive production option. Comprehensive estimations of conventional CF production cost in the future based on fossil feedstock are not available in literature. However, Nunna et al. (2019) present possible production cost of as less as 10.9 USD/kgCF (ca. 9.1 €/kgCF), which is insignificantly lower than the production cost of 10.3 €/kgCF achievable for e-CF as presented in this study. Choi et al. (2019) expect a necessary CF production cost of 11–15.4 USD/kgCF (ca. 9.2–12.8 €/kgCF) to successfully introduce CF as alternative material in the automotive industry, a level that is almost in reach in 2030 for the upper end of the range and in 2050 for the lower end of the range. However, having the ever-increasing prices for fossil fuels in mind, such cost figures might be hard to achieve for conventionally produced CF on the longer term, while increasing CO₂ emission cost will further contribute. This leads to an open research question on the volume and price elasticity of substituting steel, aluminium, and other materials in industries such as automotive, aviation, shipbuilding, construction, and others. Its superior properties including high chemically inertness, tensile strength, fire resistance, and thermal conductivity qualifies CF for further applications, while showing a long-term stability without dissolving or releasing carbon (Bhatt and Goe 2017). The growth potential for e-CF with net cost of less than 10 €/kgCF may be substantial, driven by potentially lower e-CF production cost and higher CO₂ emission pricing, which is a second revenue stream as the once sequestered carbon is permanently stored in the material. However, Choi et al. (2019) also present further lower cost options for CF production based on PAN at 15.7–17.7 USD/kgCF (ca. 13.1–14.8 €/kgCF), based on lignin at 15.2 USD/kgCF (ca. 12.7 €/kgCF) or polyethylene (PE) at 16 USD/kgCF (ca. 13.3 €/kgCF). None of the mentioned options can undercut e-CF at these production costs in 2040. It has to be mentioned, though, that two cost factors are not yet included in the e-CF value chain: Cost for the copolymer for the AN polymerisation, for which no cost numbers are available but can be considered as negligible low due to the low demand, and the eventually needed nitrogen for the carbonisation of the fibre, which takes place in an inert gas environment to avoid the CF from dissolving in high temperatures. The latter is assumed to happen without nitrogen loss.

e-CF production promises to be an additional value add for the currently discussed NET portfolio. On the contrary to, e.g., biomass-based options like BECCS requiring massive land area if scaled up and thus cause more environmental problems as can be solved (Creutzig et al. 2019; Rueda et al. 2021), e-CF production runs only on air, water, and electricity. The key technology is DAC, which effectively captures CO₂ directly from the atmosphere and is expected to play a major role for future climate change mitigation (Breyer et al. 2019). Technologies for carbon capture and storage (CCS) and carbon capture and utilisation (CCU) should be clearly separated based on the different motivation and logics (Bruhn et al. 2016); however, e-CF production is one of the very few examples that can combine CCS and CCU in one application. The commonly discussed NET options of afforestation and reforestation, BECCS, biochar, enhanced weathering, DACCS, ocean fertilisation, and soil carbon sequestration are estimated to be able to contribute between 0.5 and 5 GtCO₂/a each for a total estimated negative emission requirement of 21 GtCO₂/a by 2050 (Fuss et al. 2018). With the market volume assumed in the present studies for CF until 2050, the production of e-CF would be able to contribute a negative emission potential within the same range of the listed conventional options, or almost 10% of the total CDR requirement. Another interesting CDR option similar to e-CF is electricity-based silicon carbide (e-SiC) production, adding another 13.6 GtCO₂/a negative emission potential to the portfolio in 2050, assuming that 50% of the global demand for construction sand might be substituted by e-SiC (Mühlbauer et al. 2023). However, e-CF can be realised with a lower cost risk as the final product will be able to be sold and create net profit even without an additional carbon compensation. This research contributes to new research demands for a more holistic view how energy-industry-CDR systems and their integrated modelling can foster climate change mitigation (Breyer et al. 2022). A potential limitation of the CDR potential made is the suitability of the produced CF as modelled in this study for various markets, as the tensile strength and modulus of the CF depends on the carbonisation temperature.

Substituting steel probably shows the highest potential for e-CF. Road and other transportation had a share of about 17% of the global steel demand in 2017 (Lopez et al. 2022), the cost range for a successful introduction of CF in the automotive industry has already been discussed (Choi et al. 2019). Automotive steel demand was about 248 Mt_steel in 2015 and is projected to grow to 328 Mt_steel by 2050 in a business-as-usual scenario (Lopez et al. 2022). McKinsey (2012) estimates that a conventional car, on average, uses about 875 kg_steel, representing 63% of its total weight. However, an extremely lightweight car based largely on CF would reach a weight of about 900 kg in total, thereof 36% of CF, and still 16% of steel. This leads to a substitution potential of 83% of all steel used for cars and a ratio of 2.26 kg_steel replaced by 1 kgCF. This may lead to a significant CF market in the automotive industry of about 120 MtCF in 2050, which might increase the assumed market volume in 2050 as mentioned by Zhang et al. (2020) by about 50%. However, the highest steel market share of 51% falls on the construction sector (Lopez et al. 2022). Research on steel substitution with CF currently focuses on textile-reinforced concrete. CF as concrete reinforcement especially might play a major role in construction in the future, enabling thinner structures (Kalaimathi and Shanmugam 2022) and thus, reducing cement demand, which is a substantial source of CO₂ emissions and limited reduction options for the limestone-related emissions (Farfan et al. 2019). Textile-reinforced concrete (Scholzen et al. 2012; Halvaei et al. 2020) could use woven carbon fibre mats. The negative emission potential of CF in the construction sector via substitution of construction steel in CF-reinforced concrete seems very promising as recent life cycle assessments point out (Backes et al. 2023; Mostert et al. 2022). However, CF might face a challenging situation in construction due to the strong competition with cheap steel-enforced concrete, which is used en masse as of today.

Further research is necessary to obtain current economic input data for the chemical processes. All techno-economic input data will have to be scaled to respective production capacities and trajectories until 2050 and beyond will have to be made to allow a thorough investigation on the role of e-CF as a cost-competitive production alternative, as well as a negative CO₂ emission technology contributing to a stabilisation of earth’s climate and a limitation of global warming. Further energy efficiency options must be investigated as well. The United States Department of Energy expects a major possible reduction for the energy intensity of CF production (US DoE 2017). More emphasis must be drawn to the total net water demand of the process to avoid further water stress in already affected regions. A net nitrogen demand for the inert gas nitrogen of the carbonisation process must be identified and included in the nitrogen balance of the ASU. More detailed investigations regarding technology variations, sensitivities and dynamics may be a value add for the techno-economic assessment of this novel e-CF approach in the future.

5 Conclusion

Negative CO₂ emissions are necessary to limit anthropogenic global warming, besides a thorough defossilisation of the industry sector. This paper presented a novel approach of linking the production of CF to atmospheric CO₂ as carbon source, enabling negative emissions while decarbonising the production route of CF. Three possible production routes have been shown for the two most common precursors of CF production poly(acrylonitrile) and pitch. Literature research for available techno-economic input data of all sub-processes has been performed and respective gaps in data availability have been shown.

The proposed production value chains for CF estimate the possibility of a full defossilisation of the CF production based on e-ammonia, e-methanol, and e-methane. Proposed efficiency measures such as heat pumps for low temperature heat and heat recovery for high temperature heat proved to have a promising impact on the overall energy intensity of CF production. A total of about 186.8 kWh/kgCF can be estimated, being solely electricity creating a production route for electricity-based CF. A ton of CO₂ removed from the atmosphere will therefore require a total of about 53.7 MWh_el/tCO₂. The very high carbon share in carbon fibres of up to 95–98wt% and the very light weight of the final product increases the number when re-normalised from a unit of CF produced to a unit of CO₂ removed from the atmosphere. Via the poly(acrylonitrile) precursor route, it is estimated that each ton of CF produced will be able to remove about 3.5 tCO₂ from the atmosphere, enabling about 0.7 GtCO₂/a negative emissions at a respective market volume of 200,000 ktCF/a until 2050. This positions electricity-based carbon fibre as a relevant negative CO₂ emission option, with further upside potential in case of substituting other materials due to improved relative attractiveness.

Total production cost of electricity-based CF can be reduced from 13.1 to 10.3 €/kgCF between 2030 and 2050. Normalised to a ton CO₂ removed from the atmosphere, the total cost in 2030 amount to 3767 €/tCO₂ which can be decreased to 2949 €/tCO₂ until 2050. Producing CF only for the purpose of enabling negative CO₂ emissions is therefore economically not beneficial at a carbon compensation of 135 €/tCO₂ in 2030 and 220 €/tCO₂ in 2040 and 2050. Nevertheless, if a CF price of 14.2 €/kgCF is assumed, electricity-based CF are competitive already in 2030 even without carbon compensation. Some safety against decreasing CF market prices until 2050 is present, as the production cost can be decreased to 10.3 €/kgCF.

Further research opportunities have been identified. The techno-economic modelling of the chemical processes requires more diverse input data with current numbers. Some technological assumptions should be further investigated and updated if necessary. Especially efficiency measures for the chemical processes and exhaust gas treatment may further reduce the energy demand, which would have a direct impact on the production cost of carbon fibre characterised by a share of energy-related cost of 33.5–21.5% from 2030 to 2050. Electricity-based CF could be proven as an interesting production option for a very interesting and promising product with many fields of application, by both enabling the defossilisation of the production route and enabling negative CO₂ emissions by using atmospheric CO₂ as the carbon source.

6 Appendix

Tables 2, 3, 4 and 5

Table 2 Energy and mass balances for auxiliary processes of CF production normalised to 1 ton output of the desired material

Table 3 Energy and mass balances for sub-processes of PAN-based CF production route (MeOH-to-CF)_PAN normalised to 1 ton output of the desired material. Numbers are kept the same for 2030, 2040, and 2050. If not further specified, numbers are average values obtained from references

Table 4 Energy and mass balances for sub-processes of pitch-based CF production routes (MeOH-to-CF)_pitch and (CH₄-to-CF)_pitch normalised to 1 ton output of the desired material (current estimation). If not further specified, numbers are average values obtained from references

Table 5 Cost information for pitch-based CF production routes (MeOH-to-CF)_pitch and (CH₄-to-CF)_pitch