Utilising carbon dioxide for transport fuels: The economic and environmental sustainability of diﬀerent Fischer-Tropsch process designs

Producing fuels and chemicals from carbon dioxide (CO 2 ) could reduce our dependence on fossil resources and help towards climate change mitigation. This study evaluates the sustainability of utilising CO 2 for production of transportation fuels. The CO 2 feedstock is sourced from anaerobic digestion of sewage sludge and the fuels are produced in the Fischer-Tropsch (FT) process. Using life cycle assessment, life cycle costing and proﬁtability analysis, the study considers four diﬀerent process designs and a range of plant capacities to explore the eﬀect of the economies of scale. For large-scale plants (1,670t/day), the FT fuels outperform fossil diesel in all environmental impacts across all the designs, with several impacts being net-negative. The only exceptions are ozone depletion, for which fossil diesel is the best option, and global warming potential (GWP), which is lower for fossil diesel for some process designs. Optimising the systems reduces the GWP of FT fuels in the best case by 70% below that of fossil diesel. Assuming a replacement of 9.75–12.4% of fossil diesel consumed in the UK by 2,032, as stipulated by policy, would avoid 2–8Mt of CO 2 eq./yr, equivalent to 2–8% of annual emissions from transportation. However, these fuels are not economically viable and matching diesel pump price would require subsidies of 35–79% per litre. Optimising production yields would allow decreasing the subsidies to 8%. Future research should be aimed at technology improvements to optimise these systems as well as evaluating diﬀerent policy mechanisms needed to stimulate markets for CO 2 -derived fuels.


Introduction
Our dependence on fossil fuels is depleting abiotic resources and contributing to climate change [1]. While substantial reductions in consumption are paramount, research efforts to produce fuels from alternative feedstocks must continue. One such feedstock is carbon dioxide (CO 2 ). While CO 2 is already being utilised for production of chemicals, such as urea, methanol and polyols [2,3], the overall consumption of CO 2 as a feedstock by the chemicals industry is below 1% of the 49.3 Gt/yr of anthropogenic CO 2 emitted globally [4]. Therefore, CO 2 utilisation strategies should encourage and prioritise the production of products with high demand, such as transportation fuels [5]. This paper focuses on utilisation of CO 2 to produce liquid transport fuels via Fischer-Tropsch (FT) synthesis with the aim of finding out if such fuels would be environmentally and economically sustainable on a life cycle basis relative to fossil and renewable fuels.
Production of liquid fuels via FT synthesis is a well-established practice in industry. However, the feedstocks are largely derived from fossil resources, such as coal or natural gas, particularly in countries with limited reserves of crude oil [6,7]. For example, Sasol in South Africa produces synthetic liquid fuels from coal [8], while ExxonMobil in Qatar [7] and Shell in Malaysia rely on their large natural gas reserves for the production of liquid fuels [5]. In FT synthesis, syngas (carbon monoxide and hydrogen) is converted into a mixture of gaseous and liquid hydrocarbon fuels through a catalyst-aided polymerisation reaction [9]. Although syngas is commonly produced from coal gasification or steam reforming of natural gas, alternative feedstocks, such as biomass or CO 2 , can be used for the production of syngas [5,6,9].

Literature review
Over the last decade, several studies evaluated techno-economic feasibility and/or environmental performance of the production of FT liquid fuels from different fossil and renewable feedstocks. For an overview of these studies, see Table 1. Dry [6] and Jaramillo et al. [7] compared the economic viability of different pathways for producing FT fuels from coal and natural gas in South Africa and the United States, respectively. Both studies concluded that the profitability of FT plants depended highly on crude oil prices, carbon taxes and feedstock prices. For example, it was found that FT fuels from natural gas could only compete with conventional fossil fuels at low gas prices ($0.5 per million cubic feet) and high crude oil prices ($120 per barrel). Conversely, at low crude oil prices ($40 per barrel), neither coal-nor gas-derived FT fuels were economically competitive. On the other hand, Jaramillo et al. [7] found that both types of FT fuels could achieve greenhouse gas emissions (GHG) comparable to fossil fuels (87 vs 85 g CO 2 eq./MJ), but only under very optimistic scenarios, involving carbon capture and storage and low-carbon electricity sources. Most recently, Liu et al. [10] evaluated the economic and GHG emissions of coal and biomass to liquid (CBTL) processes with carbon capture and storage (CCS) for the production of liquid fuels. According to this study, both price and GHG emissions highly depended on regional biomass availability as this determined the optimum biomass-to-coal ratio. It was determined that the ratio of 8/92 provided the best balance between price and GHG reductions estimated at 56 pence per litre and 93.6 g CO 2 eq./MJ, respectively [10].
Although FT fuels from renewable feedstocks were found generally to be more favourable than from fossil resources with respect to life cycle GHG emissions, the latter ranged widely from −20 to 150 g CO 2 eq./MJ of fuel, depending on the unit of analysis, system boundaries, assumptions and allocation methods. For example, some studies excluded transport and pre-treatment of the feedstock [11], while others excluded the use (combustion) of fuels [12].
Some studies considered specifically techno-economic feasibility and/or environmental implications of CO 2 -derived liquid fuels. Those most relevant to this work are summarised in Table 2; for further details on the production pathways and costs, see Graves et al. [13] and Brynolf et al. [14], respectively. For example, Abanades et al. [15] assessed the avoided CO 2 emissions and related mitigation costs of a CO 2 -tomethanol process. The study showed that this particular carbon capture and utilisation (CCU) system was no better than producing methanol from fossil resources in terms of climate change mitigation and was even more costly. On the other hand, Deutz et al. [16] found that blending diesel with CO 2 -derived oxymethylene ethers could reduce the GHG emissions from transport fuels by 22% if hydrogen was produced via water electrolysis using renewable energy. In addition to GHG emissions, this study also estimated the emissions of NOx and particulates from fuel combustion. Hombach et al. [17] also looked at the production of hydrogen via two different water electrolysis technologies using fossil and renewable energy for the production of FT liquid fuel from CO 2 captured from air considering current (2015) and future projections (2030). The estimated global warming potential (GWP) varied between 64 and 440 g CO 2 eq. per MJ of FT liquid fuels for the current situation, and 7-150 g CO 2 eq./MJ for the future scenario, depending on the electrolysis technology and energy source. The production costs for all cases were found to be 3-5 times higher than the pump price of fossil diesel in Germany. Van der Giesen et al. [18] also estimated the GWP of the production of FT liquid fuels from CO 2 captured from three different sources: combustion of woodchips, combustion of natural gas and direct air capture. Similarly to Hombach et al. [17], the results ranged from 30 to 290 g CO 2 eq. per MJ of fuels, depending on the CO 2 and energy sources. Overall, the direct air capture process powered by solar electricity was the best option with respect to the GWP. Solar energy was also considered in a techno-economic analysis of the production of FT liquid fuels in comparison with production of methanol from fossil-derived CO 2 [19]. The results indicated that the capital investment and operating and maintenance costs required for the FT plant were 2.3 times lower than for the methanol plant. Furthermore, three studies assessed the economic viability of using biogas from sewage sludge to produce FT liquid fuels [5,20,21]. However, they varied in scope. For example, Badgett et al. [20] estimated the potential feedstock prices of various organic waste sources, including sewage sludge, based on the associated costs of their treatment and disposal. The study concluded that sewage sludge could be available at low or negative prices depending on local regulations. Hernandez and Martin [21] went beyond the feedstocks to determine the costs of producing FT fuels from sewage-derived biogas. The authors suggested that the production costs could potentially compete with current selling prices of fossil fuels if 'sustainability tax allowances' were introduced. Dimitriou et al. [5] also considered costs of producing FT fuels from biogas, comparing different process configurations on fuel yields, energy requirements, capital investment and production costs. The results suggested that the production of liquid fuels via FT was not economically feasible, mainly because of the low CO 2 separation and conversion efficiencies as well as the high energy requirements [5]. However, the study did not consider full life cycle costs, focusing only on the production costs.
Similarly, none of the previous studies considered the full life cycle costs or environmental impacts either. Furthermore, no study of FT fuels derived from CO 2 considered any other environmental impacts beyond the GWP (Table 2). This work goes beyond the state-of-the-art to provide a comprehensive evaluation of the life cycle economic and environmental sustainability of the production and consumption of FT liquid fuels derived from CO 2 , with the aim of identifying the most promising designs for potential deployment. Life cycle assessment (LCA), life cycle costing (LCC) and profitability analysis have been used for these purposes. The last goes beyond production costs to estimate potential selling prices of fuels. Four representative FT process designs are considered, using CO 2 emitted in anaerobic digestion of sewage sludge during production of biogas, therefore, exploring an alternative and possibly a more profitable application of sludge, which is currently used only as fertiliser. This is described in the next section, alongside the data and assumptions used in the study. The results of the environmental and economic evaluations are presented in Section 4, together with the scale-up of the plants and various sensitivity analyses. Possible policy implications related to these fuels are also considered in this section. The conclusions and recommendations for policy and future work are given in Section 5.

Process description
This study considers four process designs (PD) for the production of FT fuels from sewage sludge proposed previously by the authors [5]. All designs rely on the best available and proven technologies for CCU and are ready for deployment. They differ in the methods used to separate and capture the CO 2 present in the biogas, as follows: (i) PD-MEA: CO 2 capture from raw ('unsweetened') biogas by absorption in monoethanolamine (MEA); (ii) PD-CHP1: combustion of unsweetened biogas in a combined heat and power (CHP) and utilisation of CO 2 present in flue gases; (iii) PD-CHP2 (combination of PD-MEA and PD-CHP1): MEA CO 2 capture from unsweetened biogas and utilisation of CO 2 from flue gases generated in combustion of sweetened biogas in a CHP plant; and (iv) PD-CHP3: MEA CO 2 capture from flue gas generated by combustion of unsweetened biogas in a CHP plant.  A schematic overview of each design is given in Figs. 1-4. All four systems comprise an anaerobic digester to produce unsweetened biogas from sewage sludge. The anaerobic digester is operated at 35°C and 1 bar, with a residence time of 15 days. This is followed by CO 2 capture, syngas production and low-temperature FT synthesis, respectively to produce a mixture of gaseous and liquid fuels. Due to a higher selectivity for heavier hydrocarbons in low-temperature FT synthesis [6], diesel is the predominant liquid fuel produced. The four conceptual designs, simulated through Aspen Plus [24], are described briefly in the following sections; for a more detailed description, see Dimitriou et al. [5].

PD-MEA
As shown in Fig. 1, in this design, CO 2 present in the biogas is captured by absorption in MEA. The CO 2 is then released from the amine solution using a stripping column and the regenerated MEA is recycled to the absorber for re-use, with 99.997% of MEA recovered. The upgraded biogas and water react in a steam-methane reformer to produce syngas at 850°C and 25 bar. The excess amount of hydrogen produced in the reformer is recovered using a pressure swing adsorption (PSA) unit similar to the Linde PSA technology. The process operates at 30°C and 40 bar, which achieves an 85% hydrogen recovery with a purity of 99.999%, whilst the adsorbent is regenerated by lowering the pressure to near atmospheric [5]. The recovered hydrogen is reacted with CO 2 in a reverse-water gas shift (RWGS) reactor to produce syngas at 650°C and 1 bar. The excess water is removed from the syngas in a condensation unit placed before PSA (in all four designs). However, the nitrogen is not removed from the syngas as the FT process can cope with high nitrogen content [25,26]. All syngas is converted into a mixture of gaseous and unrefined liquid fuels via FT synthesis which takes place at 220°C and 30 bar. Such relatively low temperatures and high operating pressures, along with a H 2 :CO molar ratio of 2:1, favour the production of FT liquid fuels [5]. The gaseous fuels are combusted to produce low-pressure steam, which is used to heat the anaerobic digester.

PD-CHP1
In PD-CHP1, untreated biogas is used in a CHP plant to produce electricity and heat which are used within the CCU process (Fig. 2). The flue gases, containing CO 2 , air and water vapour, react with hydrogen in a RWGS reactor to produce syngas at 650°C and 1 bar. As in PD-MEA, the excess hydrogen is recovered via PSA and reused in the RWGS reactor. The syngas is converted into gaseous and liquid fuels in the FT unit and the gaseous fuels are used to heat the anaerobic digester.

PD-CHP2
PD-CHP2 combines the previous two designs, utilising CO 2 from two sources: unsweetened biogas and flue gas generated in combustion of sweetened biogas (Fig. 3). The CO 2 in the raw biogas is first captured in MEA and then released using the same process as in PD-MEA. The upgraded biogas is then combusted in a CHP plant and the CO 2 in the resulting flue gas reacted with hydrogen to produce syngas following the method in PD-CHP1.

PD-CHP3
This design assumes that the untreated biogas is combusted in a CHP plant and the CO 2 in the flue gas is then captured in the MEA unit. After recovery in a stripping column, CO 2 is used in the RWGS reactor to produce syngas (Fig. 4). The rest of the process is identical to PD-CHP1.

Study scope and inventory data
The following life cycle stages are considered in both the environmental and cost analyses ( Fig. 5): transportation of sewage sludge and other raw materials to the FT plant, capture and utilisation of CO 2 from biogas to produce liquid fuels in the four designs described previously, use of fuels for road transport and plant construction and decommissioning at the end of its lifetime. The latter is assumed at 20 years in the base case but the effect of different plant lifetimes is explored through a sensitivity analysis. The process for refining the produced liquid fuels is not considered as it is the same across the different process designs considered. This is not expected to influence the outcomes of the study as previous studies [11,22,23] found that the contribution of energy consumption in the refining process was below 1%.
The unit of analysis (functional unit) is defined as 'production and consumption of 1 litre of FT liquid fuels derived from CO 2 '. All four process designs are assumed to use the same amount of sewage sludge to obtain the CO 2 , but they produce different amounts of liquid fuels due to different configurations and efficiencies ( Table 3).
The inventory data are summarised in Tables 4 and 5, expressed per litre of liquid fuels. These have been estimated from the material and energy balances obtained through Aspen simulation, based on the annual operation of the plant [5]. The transport distances for the sludge have been assumed at 100 km and for hydrogen and MEA at 50 km. The life cycle inventory data for the materials, energy, transport and other background systems used for LCA modelling have been sourced from Ecoinvent 2.2 [27]. The production process is assumed to be based in the UK.
As can be observed in Table 4, PD-MEA and PD-CHP3 produce enough hydrogen to satisfy own demand, while the other two designs require additional hydrogen. This has been assumed to be produced elsewhere by steam reforming of natural gas [28]. The heat demand is met via the CHP plant, combustion of off-gases and steam from natural gas. Electricity is sourced from the CHP plant and the grid, except for PD-MEA, where the electricity demand is fully met by the grid as this option does not include a CHP plant. The UK electricity mix in 2017 [29] has been used to estimate the life cycle impacts of grid electricity; for details, see Figs. S1 and S2 in the SI.
As indicated in Table 5, all four systems produce different co-products for which they have been credited for avoiding environmental impacts from products made in other production systems. It has been assumed that the nutrients in the digestate displace an equivalent amount of industrially-produced fertilisers (see Section S2 in the SI). For the three CHP-based systems, the excess steam and electricity produced in the CHP plant is assumed to replace the equivalent amounts of steam from natural gas and grid electricity. The systems have also been credited for the recycling of construction materials remaining after the decommissioning of the plant. The quantities of the end-of-life materials for different plant capacities are specified in  Table 3 The amounts of liquid fuel produced in different designs [5]. Production PD-MEA PD-CHP1 PD-CHP2 PD-CHP3 Hourly production (kg/h) 43 a Estimated using fuel densities obtained from Aspen simulations, as follows: PD-MEA: 0.849 kg/l; PD-CHP1: 0.817 kg/l; PD-CHP2: 0.819 kg/l; PD-CHP3: 0.832 kg/l. The densities differ across the designs due to slight differences in the composition of the produced liquid fuels. b Based on 8000 h of operation. c Based on the plant lifetime of 20 years.

Table 4
Materials and energy used in the production of Fischer-Tropsch liquid fuels.  Table 6, estimated using data in Ecoinvent as explained in Section 3.3.5. Given that diesel is the predominant liquid fuel produced, all the impacts and costs have been allocated to it. In reality, other liquid fuels would also be produced in the refining process, including gasoline; however, due to a lack of data, it is not possible to speculate on their respective amounts.
The CO 2 emitted during the combustion of the fuel has not been considered as it is biogenic in nature (derived from sewage sludge). This is congruent with the ISO 14067 standard for estimating greenhouse gas emissions [30]. The impacts and costs of sewage sludge are also excluded as it is considered a waste material. However, the impacts of its transport are considered, together with the related economic costs, as explained in the next section.

Cost data and assumptions
The life cycle costs considered in this work are outlined in Fig. 5 and described in more detail in the following sections. All the data represent the 'overnight' costs with no discounting applied, following the method in Peters et al. [31].

Sewage sludge costs
The sludge is assumed to be free of charge but its transportation costs have been taken into account. These comprise fuel and labour costs given in Table 7. The fuel costs have been calculated based on the total amount of diesel used by a 27 t truck over 100 km, the distance assumed between the wastewater treatment and the FT plants. It has been assumed that 1.58 litre of diesel is needed to transport 1 t of sludge over 100 km [27] and the average price of diesel in 2017 was 120 pence per litre [32].
The costs of labour for transporting the sludge comprise wages and non-wage costs, such as sick leave, employer's pension contributions, holidays and other employment overheads. The wages have been estimated based on the hourly rate for a large goods vehicle (LGV) driver [33] and the total hours worked over 20 years. An LGV driver earns on average £22,750/yr, equivalent to an hourly rate of £10.42 [33]. Nonwage costs represent 27% of the total labour costs [34] and are shown in Table 7, together with the total sludge costs.

Fuel production costs
The production costs comprise capital, operating and maintenance costs as summarised in Table 8 for the four process designs [5]. The unit production costs, also shown in the table, are based on the total production of liquid fuels over the lifetime of the plant (for the latter, see Table 3).

Basic price of fuels and the payback period
To ensure the profitability of the plant and a minimum expected return on investment, it is necessary to determine the minimum price at which fuels need to be sold. This excludes the excise duty and VAT applied to road fuels [35] and is normally referred to as the 'basic price'. The basic price has been estimated using the modified payback period (PBP) method that incorporates the minimum annual rate of return (MARR). The PBP method is widely used to determine the Table 5 Co-products and waste generated in the production of CO 2 Fischer-Tropsch liquid fuels.

Outputs
Displaced product PD-MEA PD-CHP1 PD-CHP2 PD-CHP3   All process designs are assumed to consume the same amount of sludge but produce different amounts of liquid fuels (see Table 3).  [5], adjusted for costs of sludge transport not considered there. b Includes costs for sludge transport (see Table 7). c Based on the amount of liquid fuels produced over 20 years given in Table 3.

R. Cuéllar-Franca, et al.
A p p lie d E n e r g y 2 5 3 ( 2 0 1 9 ) 1 1 3 5 6 0 8 economic feasibility and compare different investment opportunities [31,36], which is the aim of this part of the study. The application of the PBP method is described below. An initial payback period (PBP) is determined from a chosen minimum acceptable rate of return (MARR): where: PBP payback period (years). MARR minimum acceptable rate of return (%/yr). N evaluation period (yr).
Four values of MARR have been considered here (8%, 16%, 20% and 24%) to cover a range of risk levels, from low to high.
Using the PBP equation based on fixed capital investment and annual cash flow, the basic price can be estimated as follows: where: V manufacturing fixed capital investment (£). A x non-manufacturing fixed capital investment (£). A j, avg average annual cash flow (£/y).
The sum of the manufacturing fixed-capital investment (V) and nonmanufacturing fixed-capital investment (A x ) represents the total fixed capital investment, assumed to be 95% of the TCI.
The average annual cash flow is equal to: where: S avg average annual revenue from fuel sales over the lifetime of the plant (£/yr). C O&M avg average annual operating and maintenance costs (£/yr).
The average annual revenue (S avg ) is equal to: where: p basic price of liquid fuels (£/l). m annual production of liquid fuels (l/yr).
Therefore, by substituting Eqs. (3) and (4) into Eq. (2), and rearranging to solve for price (p), the basic price of fuels can be estimated as: where: 100 factor to convert £ to pence.
The annual depreciation has been estimated using the straight-line method, as this is often used in profitability analyses that do not consider the time value of money [31]: where: n length of the straight-line recovery period (9.5 years for petrochemical plants).
The variables in Eq. (5) are summarised in Table 9, based on the data in Dimitriou et al. [5].

Decommissioning costs
Decommissioning and closure costs include plant dismantling, site decontamination and remediation. It has been assumed that these costs correspond to 4.3% of the plant's tangible fixed assets, such as machinery, buildings and land [37]. The tangible fixed assets costs are equivalent to 57% of the fixed capital investment shown in Table 9 [31]. Both the tangible fixed assets and the related decommissioning costs are presented in Table 10.
A percentage of the initial investment can also be recouped from the recovery of the working capital [38] and from the sales of land and used equipment [31]. It has been assumed that the salvage value related to the latter amounts to 20% of the fixed-capital investment [24]. It has also been assumed that all the working capital is recovered and that the price of land corresponds to 2% of the total capital investment. The endof-life revenue can be found in Table 11.

Table 9
Data for the estimation of the basic price of liquid fuels [5]. a Annual production based on the amount of liquid fuels produced over 20 years given in Table 3.

Table 10
End-of-life costs as a function of tangible fixed assets. Assets

Plant scale-up
The production rates considered in this study correspond to a daily production of < 1 t of liquid fuels (see Table 12), which is significantly lower than the production rates of actual FT plants. For example, the Bintulu GTL plant in Malaysia produces 1670 t/day of liquid fuels [39]. Therefore, to provide more realistic estimates of costs and profitability, the production capacity of the process designs considered here have been scaled up to a medium (850 t/day) and large (1670 t/day) capacity plants using the six-tenths factor rule as follows [5]: where: TCI 1 total capital investment for a base-case plant (£). TCI 2 total capital investment for a scaled-up plant (£). C 1 capacity of a base-case plant (t/day). C 2 capacity of a scaled-up plant (t/day).
The scaled-up TCI costs of the four process designs are shown in Table 12, along with the operating and maintenance costs. As can be seen, all the scaled-up designs are assumed to have the same capacity but they retain the respective ratios between the parameters, including the feedstock requirements and fuels produced as in the base case.
The six-tenths factor rule has also been used to estimate the environmental impacts of the scaled-up designs, replacing the TCI in Eq. (9) with the life cycle environmental impacts associated with the construction of the production plants. As the impacts of constructing the FT plants are not available, data for constructing an organic chemical plant have been used instead, with an annual production capacity of 50,000 t. These data have been sourced from Ecoinvent [27] and the impacts of construction scaled to the small, medium and large capacities of the FT plants considered here. The estimated data for the amounts of construction materials needed for different capacities can be found in Table 6.
The results for the four process designs are shown in Figs. 6 and 7. The GWP is discussed first as mitigating climate change is the main driver for CCU. This is followed by a discussion of the other environmental impacts. The contributions of different life cycle stages  [5]. They refer to the production of liquid fuels. Annual operation: 8,000 h. b Includes costs of sludge transport. mentioned below refer to the total impacts before the system credits.

Global warming potential
As can be seen in Fig. 6, the lowest GWP is estimated for PD-MEA (3.2 kg CO 2 eq./l) and the highest for PD-CHP2 and PD-CHP1 (17.2 and 16.4 kg CO 2 eq./l, respectively). The impact from PD-CHP3 is around two times lower than from the other two CHP-based designs (7.6 kg CO 2 eq./l) as all its hydrogen and electricity are produced internally (see Table 4).
The reason for the MEA-based design being the best option for this impact is a much lower requirement for the raw materials relative to the other three designs (see Table 4). These consumables, particularly steam and hydrogen, are the main hotspots for the CHP-based designs, contributing 60% to the GWP of PD-CHP1 and PD-CHP2 and 40% for PD-CHP3. Process energy adds a further 25-30% to the total for these three designs. It can be noted in Fig. 6 that the energy consumption is slightly greater in the PD-CHP1 and PD-CHP2 systems because of the need to heat and compress the large volume of flue gasses produced in the CHP plant that are fed downstream (see Figs. 2 and 3). Even though PD-MEA has the lowest energy requirements (see Table 4), its impact related to the energy consumption is similar to PD-CHP3 because all the electricity is supplied from the grid. For that reason, and due to a low input of raw materials, the process energy contributes 58% to the total GWP of PD-MEA.
The contribution of transport is also significant, ranging from 14% for PD-CHP1 and PD-CHP2 to 31% for PD-MEA. On the other hand, the shares of process waste management and use of fuels are negligible across the designs (< 1%). It should be noted that the impact from the use of fuels shown in Fig. 6 is related to the emissions of nitrous oxide from the car exhaust catalyst and volatile organic compounds due to incomplete combustion [42]. As mentioned earlier, the emissions of CO 2 from combustion are not considered, since they are of biogenic origin.
As can also be observed in Fig. 6, the system credits for the various co-products and recycling of end-of-life materials have a significant effect on the total GWP. PD-CHP3 has the greatest avoided GHG emissions of 8.3 kg CO 2 eq./l, largely due to the amount of fertiliser and grid electricity it displaces. The avoided impact is the lowest for the MEA-based design (5.1 kg CO 2 eq./l) due to a lack of electricity or heat generation in this system.

Other environmental impacts
PD-MEA has the lowest ADP fossil, HTP, ODP, POCP and TETP while CHP-3 is the best option for the remaining five categories (Fig. 7). The main hotspots are the plant construction, raw materials and their transport. Plant construction is particularly significant for ADP elements, HTP and TETP, contributing 50-90% to the total across the designs due to the use of metals for the plant infrastructure and emissions of heavy metals during their production. Raw materials are dominant for the ADP fossil of the CHP-based designs, causing 40-60% of the impact, largely due to the use of hydrogen and steam. They also contribute 17-35% to the AP, HTP and TETP of these designs. Finally, sludge transport causes 30-60% of AP, EP, ODP and POCP across the designs.
The credits for the co-products are significant for most categories and process designs. For AP, EP and MAETP, the credits are greater than the impacts caused, resulting in net-negative overall values. This is mainly due to the avoided production of mineral fertilisers and grid [The values on top of the bars represent the total impact after the system has been credited for the co-products. The values of some impacts have been scaled to fit. The original values can be obtained by multiplying the value shown on top of the bars by the scaling factor given on the x-axis. Impact categories: ADP elements: abiotic depletion potential of elements; ADP fossil: abiotic depletion potential of fossil resources; AP: acidification potential; EP: eutrophication potential; FAETP: fresh water aquatic ecotoxicity potential; HTP: human toxicity potential; MAETP: marine aquatic ecotoxicity potential; ODP: ozone depletion potential; POCP; photochemical oxidants creation potential; TETP: terrestrial ecotoxicity potential. DCB: dicholorbenzene.]

R. Cuéllar-Franca, et al.
A p p lie d E n e r g y 2 5 3 ( 2 0 1 9 ) 1 1 3 5 6 0 1 1 electricity. As for the GWP, the share of process waste management is negligible across the designs and impacts.

Comparison with other fuels
With the exception of fossil diesel, comparison of CO 2 -derived FT fuels with other fuels is only possible for GWP as data for the other impacts are not available in the literature. Since different fuels have differing energy content, for the purposes of comparison, the GWP values are expressed per MJ instead of per litre.
As indicated in Fig. 8, of the four designs, only PD-MEA can compete with fossil diesel as their GWP is almost identical (87 and 88 g CO 2 eq./ MJ, respectively). Compared to the 1 st and 2 nd generation biodiesel, the CO 2 -derived fuel has at best 30% higher (PD-MEA) and at worst seven times greater impact (PD-CHP2). However, in comparison to the 3 rd generation (algal) biofuels, PD-MEA has on average a 2.5 times lower GWP. PD-MEA also has a lower impact than the other FT fuels shown in Fig. 8, except for diesel from biogas and MSW. The latter is the best option overall, with a net negative value of −22 g CO 2 eq./MJ due to the credits for co-products [22].

Fig. 8. Global warming potential of Fischer-Tropsch fuels produced from CO 2 in comparison with diesel from fossil and renewable resources.
[Data for the other fuels sourced from: Jaramillo et al. [7], Azapagic and Stichnothe [43], Defra [44], Jeswani and Azapagic [45], Ecoinvent Centre [27], Wang et al. [12], Pressley et al. [22], Ahmadi Moghaddam et al. [11], Liu et al. [10], van der Giesen et al. [18], Hombach et al. [17]. Error bars indicate the range of results reported in the literature. MSW: municipal solid waste.]  Fig. 7. Data for the impacts from fossil diesel sourced from Ecoinvent [27]. For GWP, see Fig. 8.] It can also been seen in Fig. 8 that all CHP-based designs have a much higher impact than any other fuels, with the exception of biodiesel from algae and FT diesel derived from direct air capture of CO 2 , with which these designs are comparable. Therefore, from a climate change perspective, these designs are not viable for the production of FT liquid fuels.
The picture is less clear for the other environmental impacts, with the FT fuels from CO 2 having lower ADP fossil, AP, EP, MAETP and POCP and fossil diesel being a better option for ADP elements, FAETP, HTP, ODP and TETP (Fig. 9). For example, fossil diesel has 35-50 times lower ADP elements and 2-7 times lower FAETP, HTP, ODP and TETP. However, its ADP fossil is 220 times higher. FT fuels are also significantly better for AP, EP and MAETP, which are net-negative due to the system credits.
The next section considers the economic viability of FT liquid fuels in comparison with conventional diesel.

Life cycle economic sustainability
The estimated life cycle costs, revenue and profits for different MARR values are shown in Fig. 10. As can be observed, over the 20-year lifetime, operation and maintenance (O&M) are the main contributors (78%) to the total costs, with the latter ranging from £133 million for PD-MEA to £169 million for PD-CHP3. The revenues over the period vary from £211-£390 million (PD-MEA) to £262-£477 million (PD-CHP3) across the MARR values. This includes the revenue from the recovery of land, salvage and working capital of £7.8-9.4 million ( Table 11). The resulting total profits are estimated in the range of £66 million for PD-MEA to £79 million for PD-CHP3 for the lowest MARR of 8%, increasing to £244 million and £293 million, respectively, for the MARR of 24%. Assuming the latter, the payback period across the designs is 3 years, while for the lowest MARR considered (8%), this goes up to seven years (Table 13).
Therefore, the results suggest that PD-MEA is the least costly but PD-CHP3 is the most profitable design, despite the latter having high O&M costs. This is due to the highest estimated basic price of liquid fuels for this design, which ranges from 4,736 to 8,609 pence per litre (£47.36-86.09/l), depending on the MARR (Table 13). The fuel prices for the other designs range from 2,601-4,808 pence/l (PD-MEA) to 3,875-6,952 pence/l (PD-CHP2). These prices are 60-190 times higher than the average basic price of fossil diesel in 2017 of 44.4 pence/l (120 pence at pump [32]). These prices are so much higher than for diesel because of the small production scale considered. For that reason, the effect of scaling-up the plants to capacities comparable to existing fuel production facilities is explored further on, in Section 4.4.
Prior to that, the next section considers through a sensitivity analysis if the economic as well as the environmental sustainability of the four designs could be improved by targeting the respective hotspots.

Sensitivity analysis
The key parameters that could affect the environmental sustainability of FT fuels are displacement of mineral fertilisers, energy consumption, sludge transport distance, production yield and plant  lifetime. The last two, as well as the total capital investment, could also influence the economic sustainability. Therefore, these parameters are considered within the sensitivity analysis, with each being varied in turn by ± 30% around their original values; for the latter, see Sections 3.2 and 3.3.

Effect on the environmental sustainability
The effect of different parameters on the GWP of the four designs is shown in Fig. 11; the results for the other impacts can be found in Figs. S3-S6 in the SI. As indicated in Fig. 11, the GWP of PD-MEA is more sensitive to the changes in the displacement of mineral fertilisers and energy consumption than to the other three parameters. For example, if the system can displace an additional 30% of mineral fertilisers resulting from variations in the NPK content in the digestate, its GWP would be nearly two times lower than in the base case and, as a consequence, almost half the GWP of fossil diesel. The same reductions would be achieved if the energy consumption was reduced by 30%. Furthermore, shortening the sludge transport distance by 30%, would reduce the impact by a further 25%, while increasing the yield and the plant lifetime by the same percentage would lead to the GWP reductions of 18% and 6%, respectively. Therefore, optimising these parameters simultaneously could lead to significant reductions in GHG emissions from the PD-MEA system, making it much more competitive with fossil diesel. It would also mean that FT liquid fuels from CO 2 would be able to meet the requirement of the EU Renewable Energy Directive (RED) that alternative diesel fuels must have at least 60% lower life cycle GHG emissions than conventional diesel [46].
For the other three designs, a very different pattern can be observed in Fig. 11. For instance, the GWP of PD-CHP1 and PD-CHP2 is most sensitive to the changes in process yields and least affected by the variation in the plant lifetime. This is due to the raw materials being the main hotspot in both designs. The impact of PD-CHP3 is also most affected by the yield but also by the fertiliser displacement, energy consumption and sludge transportation distances. However, from a climate change perspective, none of these variations achieves significant reductions to make liquid fuels from these designs competitive with fossil diesel. For example, the minimum GWP achieved for PD-CHP1 and PD-CHP2 is 12.8 kg CO 2 eq./l (362 g CO 2 eq./MJ) and 13.38 kg CO 2 eq./l (378 g CO 2 eq./MJ), respectively, which is still four times higher than the GWP of fossil diesel. The GWP of PD-CHP3 drops to 5.48 kg CO 2 eq./l (152 g CO 2 eq./MJ) when the displacement of mineral fertilisers is increased by 30%, but this is still nearly two times higher than the impact of diesel.
All process designs show significant variations of ± (40%-120%) in most other impact categories with the changes in the displacement of mineral fertilisers (Figs. S3-S6). ADP fossil and ODP are the only exceptions, which are mostly affected by changes in the yield, as well as energy consumption and sludge transport distances in the case of PD-MEA. The second most influential parameter is the process yield, with the impacts changing by ± (20-75%) across most categories and designs. The only exceptions are POCP of PD-MEA and HTP of PD-CHP1 and PD-CHP2 which are not affected by the yield as they are largely related to the plant infrastructure and sludge transport, respectively. It can also be observed that the lifetime of the plant affects ADP elements, FAETP, HTP, MAETP and TETP, which vary by ± (20-145%). Changes in the sludge transportation distances lead to a ± (20-40%) variation in AP, FAETP and POCP. Finally, ADP fossil, ODP and POPC of PD-MEA, as well as AP and FAETP of PD-CHP1 and PD-CHP2, are sensitive to the changes in energy consumption ( ± 20-30%).
Compared to fossil diesel, at the most favourable conditions, the FT liquid fuels from some process designs become a better option for HTP, FAETP and TETP than fossil diesel (Tables S4 and S5); this is in addition [The vertical lines in the graphs denote the GWP values in the base case. For PD-MEA, GWP of 3.23 kg CO 2 eq./l is equal to the GWP of fossil diesel. For the other three designs, the equivalent values equal to the GWP of fossil diesel are: 3.11 kg CO 2 eq./l (PD-CHP1), 3.12 kg CO 2 eq./l (PD-CHP2) and 3.17 kg CO 2 eq./l (PD-CHP3) kg CO 2 eq./l.] to the other five impacts for which they were better in the base case (Fig. 9). For example, increasing the displacement of mineral fertilisers by 30% reduces the HTP of PD-MEA and PD-CHP3 by 31% and 36% relative to diesel and TETP by 18% and 32%, respectively. In addition, FAETP of all process designs is also reduced significantly on the base case, with PD-CHP3 becoming net-negative for this impact and a better option than fossil diesel. However, ADP elements and ODP still remain higher across the designs than the equivalent impacts of diesel. Even by extending the plant lifetime by 30%, the ADP elements is still 16-36 times higher than for diesel. Finally, ODP remains 25-30% higher than that of diesel for the best values of any of the parameters considered.

Effect on the economic sustainability
The effect of the plant lifetime on the LCC and profitability of the different designs is shown in Fig. 10. Increasing the lifetime from 20 to 26 years, increases the LCC by 24%, while reducing it to 14 years lowers the costs by 24% across the designs. It can also be observed that the LCC of PD-MEA over 26 years are comparable to the CHP-based plants if their lifetime is kept at 20 years. This would also lead to a competitive profitability of this process compared to the rest whilst maintaining lower fuel prices. For example, for a MARR of 8%, PD-CHP1 would earn a profit of £70 million over 20 years at a basic price of fuel of 3,645 pence per litre. By comparison, PD-MEA would gain £71 million over 26 years for a price of fuel of 2,466 pence per litre.
The basic fuel price would also be reduced with an increase in the yield and a reduction in TCI. For a MARR of 8%, increasing the yield by 30% would decrease the fuel price across the designs by 23% (Fig. 12). However, if the yield was 30% lower than in the base case, the fuel price would increase by 43%. Increasing the TCI by 30% would increase the basic prices by 20-30% while reducing the TCI by 30% would decrease them by 33% across the designs. Similar trends are found for the other MARR values (Figs. S7-S9 in the SI). However, under the best conditions, these prices would still be much higher than the price of fossil diesel. As mentioned earlier, this is largely due to the small-scale production considered so far. The issue of scale is explored further in the next section.

Plant scale-up
This section examines the effect of the economies of scale on the environmental and economic sustainability of the four process designs considered in the previous sections. Two production capacities are considered: medium and large, producing 850 and 1,670 tonnes of liquid fuels per day, respectively. The scaling-up has been carried out following the methodology explained in Section 3.3.5.

Life cycle environmental sustainability
Only the impacts of the large-scale plants are discussed here as the trends are similar to the medium scale. The results for the latter can be found in Figs. S10 and S11 in the SI.
As shown in Fig. 13, the FT fuels from the scaled-up plants outperform fossil diesel in all impact categories across all the designs. The only exceptions are the ODP, for which fossil diesel is the best option, and the GWP, which is lower for fossil diesel than for the CHP-based designs. However, PD-MEA now has a 24% lower GWP than fossil diesel and is the best option from a climate change perspective. Nevertheless, it still has a higher impact than the 1 st or 2 nd generation biofuels. For the other impacts, PD-CHP3 now emerges as the most sustainable alternative among the four designs, except for ADP fossil and POCP, where the MEA design is better and, as previously mentioned, ODP for which fossil diesel is preferred.
It is notable that, with the scaling-up, ADP elements, FAETP, HTP and TETP improve substantially and become net-negative. This is due to a much lower requirements for the plant infrastructure per unit of FT fuels produced. For that reason, the reduction in ADP elements is particularly significant -from being 35-50 times higher than that of fossil diesel in the base case (Fig. 9), it is now around 25-50 times lower (Fig. 13). The change in the remaining impacts (ADP fossil, AP, EP, MAETP, ODP and POCP) with the scaling-up is negligible. The reason for this is that the contribution of the plant infrastructure to these impacts is minimal, both in the base case (Fig. 7) and scaled-up plants (Fig. S12). However, as larger plants are typically more resource efficient, it is likely that all these and the remaining impacts would improve further. In the absence of real data, it is not possible to consider fully the effect of the economies of scale on the resource consumption and the related reductions in impacts. Instead, this is explored in a sensitivity analysis, focusing on PD-MEA, which outperforms fossil diesel for all impacts but ODP.
Varying the same parameters as in the base-case sensitivity analysis (Section 4.3.1) by ± 30% would change the GWP of the scaled-up PD-MEA as shown in Fig. 14; the effect on the other impacts can be seen in Fig. S13 in the SI. As in the base case, the amount of mineral fertilisers displaced and the process energy consumption have the greatest effect on the GWP. In the best case, the GWP reduces to 1 kg CO 2 eq./l or 27 g CO 2 eq./MJ, which is around 70% lower than the impact of fossil diesel. However, in the worst case, the GWP of FT liquid fuels increases to 3.95 kg CO 2 eq./l (107 g CO 2 eq./MJ), 18% higher than for fossil diesel. Decreasing sewage transport distances and increasing the yield by 30% each reduces the impact to 1.68 and 1.92 kg CO 2 eq./l, which is 50% and 40% lower than for fossil diesel, respectively. The equivalent changes in the opposite direction lead to a 1-8% higher impact than that of diesel. The GWP is unaffected by the changes in the plant lifetime as the infrastructure plays a much smaller role in this impact for large plants. The above results for can be used to estimate the potential for mitigating the climate change impact from the transport sector by displacing a certain share of fossil diesel with FT liquid fuels. In 2017, 30.4 billion tonnes of diesel were consumed in the UK [47], emitting 95 Mt CO 2 eq. of GHG emissions [48]. The revised Renewable Transport Fuels Obligation (RTFO) stipulates through the fuel blending obligation that 9.75% of fossil fuels should be replaced by alternative fuels by 2020 and 12.4% by 2032 [49]. Assuming hypothetically that all of this replacement is met through CO 2 -derived fuels from large-scale PD-MEA plants, gives the maximum GHG mitigation potentials in Fig. 15. This has been calculated considering the base-case and the improved GWP values shown in Fig. 14, assuming that the actual plants will be optimised and hence their impacts reduced rather than increased. In that case, FT liquid fuels could avoid around 2-8 Mt CO 2 eq. annually by displacing fossil diesel, with 2 and 2.5 Mt CO 2 eq. avoided in the base case for the displacements of 9.75% and 12.4% of fossil diesel, respectively. Therefore, there is a clear potential for FT fuels produced in large-scale PD-MEA plants to contribute towards mitigation of the climate change impact from transport. This would also lead to reductions in all other life cycle impacts considered here, ranging from 8% and 13% for POCP and ADP fossil to 170-480% for MAETP and ADP elements for the displacement of 9.75% and 12.4% of fossil diesel, respectively. The only exception is ODP which would increase by 7% and 8% (Table S6).
Their economic potential at large-scale production is explored next.

Life cycle economic sustainability
The life cycle costs and profits of the scaled-up process designs are presented in Fig. 16 for a MARR of 8%; the profit values for the other MARRs can be found in Fig. S13 in the SI. As can be seen in Fig. 16, the LCC for the medium-size plants are 80-100 times higher than in the base case, while the larger plants have 130-160 times greater costs, depending on the design. However, the larger plants are more profitable due to the economies of scale. In the worst case, a medium-scale PD-MEA would earn around £3.7 billion in profits over 20 years, while in the best case, large-scale PD-CHP3 would have a profit of around £8.4 billion. [The values for some impacts have been scaled to fit. The original values can be obtained by multiplying the value shown on top/bottom of the bars by the scaling factor given on the x-axis. GWP data for fossil diesel sourced from: Azapagic and Stichnothe [43], Defra [44], Jeswani and Azapagic [45], and Ecoinvent Centre [27]; data for other environmental impacts sourced from Ecoinvent [27]. See Fig. 7 for impacts acronyms.] [The vertical lines in the graphs denote the original GWP value (2.47 kg CO 2 eq./l). GWP of 3.23 kg CO 2 eq./l is equal to the GWP of fossil diesel.]  [48] and RAC [47]. RTFO: Renewable Transport Fuels Obligation [49].] Both the costs and profits are affected by the plant lifetime. For example, extending the lifetime of the PD-MEA from 20 to 26 years would increase its profits by 8% while decreasing it to 14 years would lead to 8% lower net earnings (Fig. 16).
Based on the profitability analysis of the scaled-up plants, the estimated basic prices of FT fuels decrease considerably with the increase in capacity (Fig. 17). For the large-scale plant (1,670 t/day), PD-MEA has the lowest basic price of 185 pence/l at 8% MARR, increasing to 300 pence/l for the MARR of 24%. Thus, even in the best case, this is still four times higher than the basic price of fossil diesel considered here (44.4 pence/l). However, it should be noted that a biodiesel was also significantly more expensive than fossil diesel a few years ago. For example, in 2013, their respective wholesale prices in the EU were €0.85 and €0.47/l [50] but the gap has closed since due to a wider deployment of biodiesel, as mandated by the RED, and the increasing prices of diesel.
The basic prices for the other three designs range from 221 to 275 pence/l for 8% MARR and 357-447 pence/l for the MARR of 24%. Thus, in the worst case for the large plant (PD-CHP3 @ 24% MARR), FT fuels are ten times more expensive than fossil diesel. For the mediumscale capacity, the basic prices range from 226 to 561 pence/l across the designs and MARRs; this is 5-13 times greater than the price of diesel.
These prices are also unfavourable in comparison to other FT liquid fuels reported in the literature. For example, the basic price of FT diesel produced from natural gas ranges from 9 to 47 pence/l [7]. For coalderived diesel, it is 30 pence/l [7] and around 56 pence for fuels produced in a CBTL plant [10]. However, CO 2 -derived FT fuels compare more favourably with the basic price of algae biodiesel, which can be as high as 2,300 pence/l depending on estimates [51].
Assuming the same taxation for CO 2 -derived FT fuels as for the conventional road fuels, with 63% of the pump price being the excise duty and VAT, the pump prices of these fuels would range from 500 pence (large-scale PD-MEA @ 8% MARR) to 1,516 pence (medium-scale PD-CHP3 @ 24% MARR), compared to the assumed 120 pence for a litre of fossil diesel (Table 14). Therefore, these fuels would not be viable without either reducing the taxation, providing subsidies and/or mandating their use through policy. Similar applies to biofuels, which are still not economically viable and would not be produced without strong policy incentives [52]. In addition to the fuel blending obligation mandated through the RTFO, this also included a 20 pence/l excise duty exemption until 2012.
The issue of taxation and incentives is considered in Fig. 18, which shows that subsidies would be required in all cases, ranging from 35% to 79% of the pump price of 120 pence/l, depending on the process design, production scale and the MARR.
The tax margins and subsidies are explored further through a sensitivity analysis by varying the TCI and production yields by ± 30%. The results in Fig. 19 suggest that in the best case, subsidies could be reduced to 8% but in the worst case, they would need to be at the level of 70-85%.

Conclusions
This paper has evaluated the life cycle environmental and economic sustainability of FT liquid fuels produced from CO 2 generated during anaerobic digestion of sewage sludge. Four process configurations have been considered and the sustainability evaluated for a range of plant capacities to explore the effect of the economies of scale. For a small scale production (< 1 t fuels/day), the lowest global warming potential (GWP) of 87 g CO 2 eq./MJ estimated for PD-MEA matches the GWP of fossil diesel. However, the highest impact of 484 g CO 2 eq./MJ (PD-CHP2) is 5.5 times higher than that of diesel. Liquid fuels from PD-MEA also have the lowest other environmental impacts compared to the fuels from the other three designs. In comparison with diesel, PD-MEA fuels are better for five impact categories while diesel is environmentally more sustainable for the remaining five impacts.
The results suggest that PD-MEA is the least costly but PD-CHP3 is the most profitable design. However, the latter has the highest estimated basic price, ranging from 4,736 to 8,609 pence per litre, depending on the MARR. The lowest fuel prices of 2,601-4,808 pence/l are found for PD-MEA. These are 60-190 times higher than the average basic price of fossil diesel in 2017 of 44.4 pence per litre.
Both the environmental and economic sustainability improve significantly with the economies of scale. For the large-scale plants (1,670 t/day), the FT fuels outperform fossil diesel in all impacts across all the designs, with several impacts being net-negative. The only exceptions are the ODP, for which fossil diesel is still the best option, and the GWP, which is lower for fossil diesel than for the CHP-based designs. Optimising the key parameters for the scaled-up plants leads to the GWP of FT fuels from PD-MEA being up to 70% lower than that of diesel. Therefore, FT fuels from CO 2 have a potential to reduce GHG emissions from the transport sector. Following the RTFO fuel blending obligation, and assuming a hypothetical replacement of 9.75-12.4% of diesel by FT fuels from CO 2 , would mitigate 2-8 Mt of CO 2 eq./yr, equivalent to around 2-8% of emissions from the transport sector.
The fuel prices are also significantly lower for the large plants, reducing to 185 pence per litre in the best case (PD-MEA at 8% MARR), but this is still four times higher than the price of diesel. Thus, these fuels would not be economically viable without either reducing the taxation, providing subsidies and/or mandating their use through policy, in a similar manner the biofuels markets have been stimulated. The results obtained in this work suggest that matching the diesel pump price of 120 pence/litre would require subsidies between 35% and 79%, depending on the MARR and the scale of production. If the production yields can be improved by 30%, subsidies could be reduced to 8%. While the economies of scale and policy will help to improve the competitiveness of these fuels, further technological developments are also essential to optimise key process parameters. These include more efficient CO 2 capture and conversion processes as well as novel sorbents and catalysts aimed at reducing energy consumption and increasing fuel production yields. Future research should also explore different policy mechanisms for stimulating development of the markets for CO 2derived fuels.