A techno-economic and life cycle assessment for the production of green methanol from CO2: Catalyst and process bottlenecks

: The success of catalytic schemes for the large-scale valorization of CO 2 does not only depend on the development of active, selective and stable catalytic materials but also on the overall process design. Here we present a multidisciplinary study (from catalyst to plant and techno-economic/lifecycle analysis) for the production of green methanol from renewable H 2 and CO 2 . We combine an in-depth kinetic analysis of one of the most promising recently reported methanol-synthesis catalysts (InCo) with a thorough process simulation and techno-economic assessment. We then perform a life cycle assessment of the simulated process to gauge the real environmental impact of green methanol production from CO 2 . Our results indicate that up to 1.75 ton of CO 2 can be abated per ton of produced methanol only if renewable energy is used to run the process, while the sensitivity analysis suggest that either rock-bottom H 2 prices (1.5 $ kg −1 ) or severe CO 2 taxation (300 $ per ton) are needed for a profitable methanol plant. Besides, we herein highlight and analyze some critical bottlenecks of the process. Especial attention has been paid to the contribution of H 2 to the overall plant costs, CH 4 trace formation, and purity and costs of raw gases. In addition to providing important information for policy makers and industrialists, directions for catalyst (and therefore process) improvements are outlined.


Introduction
The high environmental impact of greenhouse gas emissions requires the development of technologies for the capture, storage and valorization of CO 2 .In this scenario, the catalytic conversion of captured CO 2 has attracted a great deal of attention over the last few decades.Production of methanol, dimethyl ether and other hydrocarbons is amidst the most studied CO 2 hydrogenation processes [1,2].Among these possibilities, methanol stands out [3,4], thanks to its versatility as commodity chemical, potential as a marine fuel, and as feedstock for the manufacture of intermediate chemicals and even more complex road transport fuels [5].Not surprisingly, the large-scale production of green methanol is being sought after by both industry and academia.
Within the strategies for CO 2 valorization through catalytic technologies, insightful results have been reported for its electrochemical [6] and photochemical transformation [7].In that sense, recent publications have reported the potential of electrofuels from CO 2 .Brynolf et al. [8] reviewed the production cost estimates of this technology, which could potentially be competitive by 2030 according to the developments of technology.Regarding life cycle assessments (LCAs) of electrofuels, there is no doubt that greenhouse gas emissions can be clearly decreased by circa 90%, values that change depending on the considered boundary conditions [9].Nevertheless, the uncertainty of calculations was also highlighted, as well as the high costs associated with these new technologies, mainly caused by electrolizer capital costs and stack life [8].
For these reasons, when considering the current technology readiness level (TRL) of these technologies, thermocatalytic processes seem to be far ahead in terms of industrial implementation.In the field of thermocatalytic CO 2 hydrogenation to methanol, in addition to a large number of studies focusing on the classical syngas methanol synthesis catalyst, Cu-ZnO-Al 2 O 3 , several new systems have been recently reported, with ZnO-ZrO 2 [10], Pt-based [11] and In 2 O 3 /ZrO 2 catalysts being among the most promising solids in terms of methanol selectivity under mild reaction conditions [12,13].Indium oxide has been reported as a highly active catalyst to produce methanol.
The activation of CO 2 takes place over the active sites generated by In 2 O 3 and an oxygen vacancy derived from the ZrO 2 , and the stability of the catalyst depends on the In reducibility [13].In a previous work, an In 2 O 3 /Co catalyst was presented as a promising alternative to the Zr-based catalyst [14].Computational catalysis proved that the reaction is driven by the same mechanism, with oxygen vacancies being derived from Co.The obtained results in terms of methanol selectivity (>80%) and productivity (0.86 g CH3OH g catalyst −1 h −1 ) at near thermodynamic equilibrium conversion place this new catalyst among the state of the art [14].
Catalyst development is, however, only one (although very important) hurdle to overcome when improving green methanol production; equally important is the overall process that determines the economic feasibility of the manufacture route.In order to gain insight into process performance, understanding (and modelling) reaction kinetics is a must.Vanden Bussche and Froment [15] proposed a steady-state kinetic model for the CO 2 hydrogenation to methanol considering all the individual steps involved in the process, including thermodynamic considerations.After simplifications, the system was described by apparent kinetic and adsorption constants.More recently, Seidel et al. [16] reported a comparison of a complex model with a Langmuir-Hinshelwood model, resulting in a very accurate estimation of experimental results in both cases.Indeed, Langmuir-Hinshelwood models have shown reliable overall fitting of experimental data for CO 2 hydrogenation using different catalysts and conditions [17][18][19][20].
Using similar kinetic models, albeit based on outdated data for the commercial Cu-ZnO-Al 2 O 3 catalyst, several groups have carried out steady-state simulations of a methanol plant using Aspen Plus [21][22][23][24].Most of these studies propose a similar plant outline, incorporating compression and separation stages and reactant recirculation along with the catalytic reactor.These simulations can be used to predict the amount of CO 2 abated by the production plant [21], estimate the viability of the process as a function of reactant price fluctuation [23], and calculate the carbon tax that would make the process economically viable [24].However, as recently highlighted by Nestler et al. [19], in most cases these simulations have relied either on outdated Cu-ZnO-Al 2 O 3 kinetic models or on models derived under conditions far from those of the actual plant operation.Moreover, these studies provide limited insights into the process bottlenecks.Indeed, CH 4 trace formation, the raw gas source (and therefore purity and cost), the influence of the energy source in the final CO 2 abatement, the impact of the different process costs (i.e., CAPEX, raw materials, utilities) on the breakeven MeOH price, or the maximum O 2 content in the system (which can alter the catalyst performance, especially for the Cu-ZnO-Al 2 O 3 commercial system) have been largely overlooked in the open literature.
Hence, in order to address the issues highlighted above, we present a full multidisciplinary approach which should give the most updated economic analysis on green methanol synthesis.We first describe the catalyst material and perform the corresponding kinetic fitting.This is followed by detailed process modeling using the developed realistic kinetics.We then study how process variables impact the technoeconomics of green methanol production.Lastly, we perform a life cycle assessment of the process to assess the real environmental impact and determine the main process bottlenecks.

Catalyst preparation
The catalyst was prepared via co-precipitation method following the procedure detailed elsewhere [14].Briefly, 3.49 g of Co(CH 3 COO) 2 6H 2 O and 1.02 g of In(CH 3 COO) 3 were dissolved in 35 mL of H 2 O.Separately, 1.39 mg of H 2 O 2 and 35 mL of 28% ammonia hydroxide solution were diluted with H 2 O to 315 mL.Then, the saltscontaining solution was added dropwise to the stirred ammonia hydroxide solution with a flow rate of 2 mL min −1 .The formed powder, denoted as InCo, was recovered and washed via centrifuging and subsequently dried at 60 C.The washing protocol was repeated for three times.

Catalytic tests
Catalytic tests were carried out in a 16 channel Flowrence® from Avantium.Samples of the InCo catalyst were loaded in 15 of the channels, using the remaining one as a blank without catalyst.Prior to the reactions, the tubes were pressurized to 50 bar using a membrane-based pressure controller and the temperature was set to 300 °C.After 24 h of pretreatment under the reaction flows, experimental runs were carried out in the pressure range of 30-50 bar at 250-350 °C, using space time values up to 11 g h mol CO2 −1 , a total flow of 13 cm 3 (STD) min −1 per channel, and H 2 /CO 2 molar ratios of 3-4.A small constant concentration of He was mixed with the feed as internal standard.
,R He,R CO ,blk CO ,R He,blk He,R 100, where C He,blk , C He,R , , are the concentrations of He in the blank, He in the 2 CO ,blk

C
reactor effluent, CO 2 in the blank, and CO 2 in the reactor effluent, respectively, and is the concentration of each i product in the reactor effluent with n i carbon atoms.The carbon balance closure was higher than 97.5% in all cases.Each experiment was repeated four times, and the composition was computed by average values of the results.
Please note that negligible conversion of CO 2 to MeOH, CO or CH 4 was observed in all blank tests.

Kinetic modeling methodology
The kinetic model for the conversion of CO 2 into methanol was developed using the convection-dispersion-reaction equation, accounting for the mass conservation equation of each component of the reaction medium.Several simplifications can be assumed due to the characteristics of the system.The catalytic tests were performed during 180 h in a sequence that includes repeated control experiments.Deactivation was not observed, and the catalyst activity remained constant during the duration of the experiments.This avoids the need for a time-dependent expression of the equation and the intrinsic kinetic model can be extracted from these steady-state experiments.Moreover, the reactor inner diameter is 2 mm, which completely avoids radial dispersion and permits modeling the flow as purely convective transport.Axial dispersion is assumed negligible due to the high enough gas flows in the small-diameter reactor, with Re up to 10 5 .At these values, dispersion coefficient can be assumed as low as 0.2 m 2 s −1 [25], with Pe reactor number values of 10 2 .The pressure in all channels was accurately controlled by a membrane system and temperature was maintained constant in each set of channels (temperature controllers for each group of 4 reactors).Therefore, isothermal and isobaric conditions can also be assumed.In summary, experimental runs were carried out at suitable conditions that allow the intrinsic kinetic parameters to be calculated using the steadystate design equation for a packed bed reactor.Then, for the molar fraction of each i component of the reaction medium: is defined for a catalytic bed length within 0<l<L, where F is the molar flow rate, S is the reactor section, ρ is the catalytic bed density and r i is the formation rate of each i compound.Although the contribution of axial dispersion is low, a negligible diffusion term was also used aiming at the stabilization of the mathematical method, as we explained elsewhere [26].For that reason, the above mentioned Pe number was considered.
These formation rates are calculated from the reaction rate of each j step of the reaction network, described by the reverse water gas-shift reaction, the production of methanol from CO 2 and CO and the undesirable formation of methane (heavier hydrocarbons are not observed in the product stream), The kinetic equations follow Langmuir-Hinshelwood expressions, defined as with kinetic constants (k j ) being defined by a reparametrized form of the Arrhenius equation: where k j * is the kinetic constant at the reference temperature T * , E j is the activation energy and R is the universal gas constant.The equilibrium constants (K j ) are estimated as a function of the temperature using the empirical correlations reported elsewhere [27].The adsorption constants for the CO 2 and water ( and , respectively) follow the van't Hoff equation: where K * is the adsorption constant at the reference temperature T * and ΔH is the enthalpy of adsorption.Note that, despite methanol adsorption could also limit reaction rates, it was not considered in the model because a significant improvement of experimental data was not observed probably due to its lower partial pressure in the reaction medium.
The system is solved using a Runge-Kutta method of orders 1-5, with k j * , E j , K * and ΔH being the parameters to be optimized.With this aim, an objective function based on the sum of square errors (SSE) between the calculated and experimental values of molar fractions (y i and y i e , respectively) is defined as where ω i is the weight factor of each i compound and n c and n e are the number of compounds and experiments, respectively

Process simulation
Process simulations were carried out with steady-state simulation models developed in Aspen Plus ® V8.8 software.The selected property method was SR-POLAR, based on an equation-of-state model by Schwarzentruber and Renon, which is an extension of the Redlich-Kwong-Soave equation of state.The Wegstein method was used for flowsheet convergence with a mass balance closure of the system was better than 99.99%.In line with the standard conventional plants, the annual productivity target of the system is 275 kton year −1 of 99.5% purity MeOH.Economic analysis was carried out with the Economics Solver extension of Aspen Plus.The electricity price was set to 0.0775 k$ kWh −1 .For the CO 2 emissions associated with the electricity and steam utilities, Natural gas was selected as fuel source with a CO 2 emission data source from US-EPA-RULE-E9-5711 (5.59 10 −5 kg CO 2 per kJ).CO 2 emissions using these estimates were only considered for the conventional energy sources scenario.The CO 2 abated was then calculated as the total CO 2 reacted in the plant minus the CO 2 emitted by the process utilities following the above CO 2 emission data source.
The techno-economic assessment of a plant using the conventional Cu/ZnO/Al 2 O 3 methanol catalyst was also carried out as a comparison scenario.In particular, the plant proposed by Szima et al. [28] was scaled in order to match the target output of 275 kton year −1 of 99.5% purity MeOH.It should be noted that the comparison scenario was performed without accounting for detailed kinetics of the Cu/ZnO/Al 2 O 3 catalyst.

Life cycle assessment
The LCA simulation of the methanol plant was carried out using the software GaBi ® Pro.9.5 together with the databases associated to this program.Processes located in the European Union were considered.Two study cases were evaluated: the first study case considers all energy supplies for the reaction and electrolysis to be provided by a mix of renewable sources (47% photovoltaic, 30% wind energy, and 27% hydraulic [29,30]) and thermal energy supplied by the combustion of hydrogen from water electrolysis (hydrogen combustion has been selected as the best way in terms of environmental impact for thermal energy production [31]); the second study case considers thermal energy to be supplied by the combustion of natural gas, electricity from a current standard mix and hydrogen from natural gas steam cracking.A cement plant was considered as the source of CO 2 with amine based capture located next to the MeOH plant.In case 1 the H 2 electrolyzer was also considered to be close to the MeOH plant.
Both cases were evaluated using the environmental indicators recommended by The European Platform on Life Cycle Assessment [32].Detailed information on these indicators and the assumptions of each study cases can be found in the Supplementary Material.

Experimental data fitting and kinetic modeling
Experimental runs were carried out across a wide range of reaction conditions to gather information on the effect of the main variables for the development of the kinetic model.Fig. 1 shows the evolution with temperature of the selectivity to products (methanol, CO, methane and hydrocarbons) and CO 2 conversion at 50 bar, H 2 /CO 2 molar ratio of 4 and using different space time values of 3.7, 7.3 and 11.0 g h mol CO2 −1 (Fig. 1a-c, respectively).Each run was repeated at least three times, with the corresponding error bars being displayed in the conversion symbols; the reproducibility of the results is shown to be high.From the experimental points depicted in Fig. 1 we can observe that CO 2 conversion increases with temperature and space time and reaches a maximum value of ca.25% at the extreme study case (350 °C and 11.0 g h mol CO2 −1 ).
Nevertheless, the product distribution is drastically modified as the reaction conditions are varied.Without considering the reaction at 250 °C and 3.7 g h mol CO2 −1 , which shows negligible conversion values, methanol selectivity significantly declines as temperature increases.This presumably occurs due to the disfavored equilibrium reaction of CO 2 conversion at higher temperatures [18].Consequently, the selectivity to CO is increased with temperature.The concentration of CH 4 is also higher at 325 and 350 °C, which could be related to the promoted methanation reaction over the Co catalyst.A similar trend is observed with space time.The increase in the amount of catalyst leads to a higher extent of reaction, which decreases the selectivity to methanol and increases that of CO.Aiming for maximum production of methanol in the simulated process, a compromise between conversion and selectivity must be reached.The evolution of methanol productivity in each of these experimental runs is depicted in Fig. S1.In all cases, a maximum of productivity is observed, which is displaced towards lower temperatures upon increasing the space time value.The overall maximum methanol production exhibited by the InCo catalyst (0.42 g g −1 h −1 ) is located at the central experimental data point, at 50 bar, H 2 /CO 2 molar ratio of 4, 300 °C and 7.3 g h molCO 2 −1 .Note that runs at different pressures and H 2 /CO 2 molar ratios were also used for the kinetic model computation, but methanol productivity was lower in all cases.
The kinetic modeling methodology (see Methods and Supplementary Material) and the collected experimental data were used in order to calculate the kinetic parameters that describe the catalytic system.The best fitting of the experimental data is obtained with the parameters listed in Table 1.The 95% confidence intervals, also shown in Table 1, were calculated from the discrepancies between experimental and calculated data.The value of the SSE obtained is ca.0.04 and the standard deviation is ca.0.02, which gives a very narrow normality plot for the kinetic parameter computation.With the computed kinetic parameters, the initial reaction rates for CO and MeOH formation are estimated to be 5.1 10 −3 and 1.7 10 −2 mol g −1 h −1 , respectively.This is in agreement with most of the reported mechanistic studies, which suggest the prevalence of the direct mechanism of CO 2 to MeOH via formate and methoxy intermediates [33].
The order of magnitude of the computed constants is similar to that reported for a similar catalytic system with CO 2 conversions higher than 15% [34].In terms of energy requirements, the values of activation energy calculated for all the steps of the reaction network are within the 47-65 kJ mol −1 range, characteristic values of the kinetic regime.
Especially, the value of the activation energy for the direct hydrogenation of CO 2 to methanol is within the range of those previously proposed by other authors with similar catalytic systems [18].At our experimental conditions, the apparent activation energy for the rWGS reaction is slightly lower than that of methanol formation, which can also be explained by equilibrium-limited reaction at the integral regime used in this work.
The adsorption heats for CO 2 and water are, as expected for an exothermic process, negative in value.The suitable fitting of the experimental data is more easily observed in Fig. 1(d-g), where the evolution with space time of the molar fractions of Ccontaining products is illustrated.The model predicts well the CO 2 conversion into CO and methanol.Moreover, the maximum concentrations of methanol previously observed at 300 and 325 °C are also estimated by the model.Experimental data fitting for reactions with lower H 2 /CO 2 ratio and lower pressure is also shown in Fig. S2.
Although Fig. 1(d-g) show the evolution of C-containing products, the model also accounts for the consumption of H 2 and formation of water.Indeed, the partial pressures of both components play a key role in the reaction network (see Equations ( 8)- (11) and experimental data fitting in Figs.S3-S6).As expected, the concentration of H 2 decreases along with that of CO 2 and water is produced as a function of the conversion.
Then, the higher the extent of the reaction, the more water is produced through the reverse water-gas shift reaction (Equation ( 4)) and the direct formation of methanol (Equation ( 5)).Management of these two carbon-free compounds is of a paramount importance for the process, as H 2 should be recirculated and H 2 O should be condensed in order to maximize methanol production.Likewise, although its concentration is the lowest one at most conditions, the production of methane increases with space time, temperature, total pressure and H 2 /CO 2 ratio.At this point, this experimental observation (well predicted by the model) should be considered for the simulation of the plant since recirculation and scaling-up will lead to an accumulation of this undesirable byproduct in the system, ultimately causing a pronounced decrease in methanol productivity.At the optimal conditions that will be simulated in the plant, the net formation rates of CO, methanol and methane are compared in Fig. S7.Despite the rates of methanol and CO formation being considerably higher (0.75 times lower that of CO at this temperature), methane is present in the product stream, with its formation rate being 0.14 times lower than that of methanol.
As an overall fitting result, the parity plot in Fig. 1(h) shows the comparison of the experimental and calculated data for the 90 non-repeated runs fitted for the kinetic computation (from which the 95% confidence interval was estimated).Most of the dots are located at the diagonal and only few outliers are identified, validating the model.

Aspen Plus ® process simulation
In our process simulation we targeted an annual productivity of 275 kton year −1 of 99.5% purity MeOH, in line with standard conventional plants [35].A simplification of the plant is illustrated in Fig. 2(a), whereas the detailed process model flowsheet is summarized in Fig. 2(b) and contains three main parts: the feeding stage, the reaction stage and the separation stage.

The feeding stage
To meet the desired product capacity, the process has a CO 2 feed of 1750 kmol h −1 and a H 2 feed of 5500 kmol, close to the 1:3 stoichiometric ratio.Both the CO 2 and H 2 feed are considered to be produced in external facilities.Taking the reactants recirculation into account, the CO 2 -to-H 2 molar ratio in the reactor section is 1:4, as was previously observed to be the optimal one.The inlet conditions for both gases were assumed to be 1 bar and 25 °C.Both streams were compressed by 4 isentropic compressors (C1 to C4 for CO 2 and C5 to C8 for H 2 ) with intercooling heat exchangers (HX1 to HX6) to cool down the pressurized streams.The isentropic efficiency was set to 0.8 and the mechanical efficiencies to 0.95.Once pressurized at 50 bar, the streams were mixed (M1) together with the recycle stream.

The reaction stage
The pressurized mixed feed is first sent to a heat exchanger (HX7) to preheat the

The separation stage
The reactor outlet at 300 °C and 50 bar is first cooled down to 30 °C in a heat exchanger (HX8).The cooled stream is then sent to a flash drum unit (Flash1) operating at 30 °C and 50 bar to remove the light gasses (CO, CH 4 , and H 2 ).This gas stream is recycled back and mixed to the reactor feed with a 2.5% purge (Purge1).Based on our experimental observations and model predictions (Fig. 1), avoiding the production of methane is not completely possible at all conditions, which motivates the necessity of including this purge stream.The liquid stream is sent through a pressure relief valve to reduce the pressure to 1 bar and then to another flash drum (Flash2), operating at 25 °C and 1 bar to remove traces of CO 2 dissolved in water (Purge2).The liquid stream, rich in water and methanol, is sent to a RadFrac rigorous distillation column operating at 1 bar with 30 equilibrium stages final product purification.The condenser type is total condenser and the reboiler type is kettle.The reflux ratio is set to 2 and the final purity of the methanol stream is 99.5%.

Techno-economic evaluation (TEA)
The mass balance and composition of each stream of the process depicted in Fig. S2.In total, 32.1 ton h −1 of methanol with a purity of 99.5% is produced in the plant.Assuming 8500 h year −1 of operation, the plant will generate a total of 272850 ton year −1 of liquid methanol, close to the design value of the plant.The total thermal duty was −74 MW (34 MW of heating duty and 108 MW of cooling duty, mostly for cooling water) and total electrical power was 31.5 MW.The recycle of the light gases stream would allow an overall CO 2 conversion of 72.5% to be obtained in the process, almost 5 times higher than the experimental conversion per pass.

2(b) can be found in Table
Nevertheless, attempts to increase this conversion by reducing the purge ratio below 2.5% were fruitless as CH 4 accumulated in the system to concentrations above 15%.
Moreover, since the heating/cooling/compression steps are also an indirect source of CO 2 (see experimental section), the real CO 2 conversion (i.e., total CO 2 abated) decreases to 28.9% (purge ratio of 2.5%), with abated CO 2 quantities of 0.77 tons per ton of methanol produced (see Table 2).Hence, these numbers highlight the importance of using renewable energy to power up the plant as 55% of the CO 2 converted in the reactor will be offset if traditional fossil energy sources are used.In contrast, if only renewable energy sources are used and all these indirect sources of CO 2 are obviated, the CO 2 abated will jump to 1.75 tons per ton of methanol, a 225% increase.To provide further insight, we estimated the size of the renewable energy source depending on its type and the associated CAPEX needed.As discussed above, the total electrical power is estimated in 32.5 MW.For solar photovoltaic cells (PV), on a capacity basis, the total-area capacity-weighted average is ca.28 MW km −2 [36,37].For typical average wind installation, this requirement is 5 MW km −2 [38].Therefore, only considering the electricity, the methanol plant will require 1.16 km 2 of land if PV is used or 6.5 km 2 of land if wind is used.This later value can be further reduced by circa 40% if offshore wind farms are used due to their higher capacity density.Assuming now that CAPEX costs of PV and wind farms can be estimated in 500 and 1500 $ kW −1 , respectively [39], the solar PV farm will cost ca.16 M$ and the wind farm will cost ca.
48 M$.Both values are less than the methanol plant CAPEX.However, only approx.
40% of the CO 2 emissions of the plant come from the electricity used for the compressors (31.5 MW).The remaining 60% is due to the heating thermal duty (34.5 MW), meaning that virtually two times bigger PV or wind plants will be required, with a resulting CAPEX rivaling the one of the methanol plant alone.Hence, here we need to ponder that the integration of renewable energy comes with a huge associated cost that can double the initial CAPEX investing.Additionally, we also need to consider the intermittency of the renewable energy [40].While the intermittency can be partially offset by combining different sources that complement each other (i.e., continental wind that tends to peak at night and solar that peaks with daylight), it might require external storage systems that can impact the initial investing.This could even lead to prohibitive MeOH costs above 1500 $ ton −1 if intermittency is addressed solely by energy storage [41].A much more reasonable approach is the flexible process with a variable load able to meet the annual target [41].Hence, integration of renewable energy is not straightforward and individual studies for each particular plant capacity and/or location are necessary.
Following, the estimated equipment costs required for the operation are shown in Fig. 3(a).In addition, the values of individual costs can also be found in Table S3.As observed, two compressors account for more than 65% of the total equipment costs.
Specifically, C5 and C6 compressors deal with H 2 compression from 1 bar to only 10 bar (please see the effect of final pressure on energy requirement for compression on Fig. S9).Hence, if green electrochemically generated hydrogen could be produced at least at 10 bar, the CAPEX would be reduced from 94 to 34 M$.At this point, it is worth mentioning that most of the previous process simulation studies in the state of the art do not consider the H 2 compression [21][22][23][24], despite its obvious influence on CAPEX.Most of them only consider the simulation of the Cu-ZnO-Al 2 O 3 commercial catalyst with the feed stream directly at 30 bars, thereby painting an unrealistic picture of the overall process.Only the work of Szima et al. [28] included the compression of reactants from atmospheric to 80 bar, which was also predicted as one of the most expensive operations of the process.Next, we evaluated the economic viability of the proposed MeOH plant.For the initial analysis, the following material costs were used: 550 $ ton MeOH −1 , 50 $ ton CO2 −1 and 3.5 $ kg H2 −1 .The MeOH price was set based on the guidance provided by the Methanol Institute in their 2019 report on green methanol [42] and on the work of Bergins et al. from Mitsubishi Hitachi Power Systems [43].A 20 year timeframe was considered for CAPEX amortization with a linear depreciation method.The results of overall costs are listed in Table 3 and depicted in Fig. 3(b).H 2 and CO 2 costs (325 M$ year −1 ) are two times the product sales (154 M$ year −1 ), making this process unfeasible.
This high influence of the raw materials can be easily observed in Fig. 3(c), where the contribution of the different annualized process costs on the final MeOH price is evaluated.Up to 67.5% of the final MeOH price comes only from H 2 , while CO 2 contributes to 16.5%.Similarly, the utilities account for 12.4% of the MeOH price of which almost 85% comes from the compression steps (see Fig. S10).Last, the CAPEX contribution to final MeOH price is only 2.2%.These numbers highlight the importance of including gas compression in the TEA and gives fair warning to the optimistic MeOH prices provided by current simulations in the state of the art.However, it could be reasonable to consider that grey and blue hydrogen can be produced and delivered at high pressure.This assumption should be avoided for green hydrogen produced in current electrolyzers, which work close to near atmospheric pressures, and therefore, this extra expense needs to be taken into consideration.In summary, the overall process can be considered low capital intensive, highly dependent on the raw material cost, and hindered by the energy costs required for H 2 compression.

Variable Value
Total capital cost ($ ton MeOH −1 ) 347.6 Equipment installed cost ($ ton MeOH −1 ) 209.6 Total operating cost ($ ton MeOH −1 year −1 ) 1408.7 Raw materials cost ($ ton MeOH −1 year −1 ) 1192.0 Utilities cost ($ ton MeOH −1 year −1 ) 96.5 Total product sales ($ ton MeOH −1 year −1 ) 567.9 To better evaluate the raw materials effect on the net revenue (i.e., product salesoperating costs), a sensitivity analysis was performed focusing on H 2 and CO 2 (see Fig. 3d and e, respectively).As expected, the feasibility of the MeOH plant largely depends on H 2 and CO 2 prices.In particular, with the current MeOH plant design, the process would be economically viable with H 2 prices lower than 1.5 $ kg −1 with a CO 2 price of 50 $ ton −1 .Similarly, with the current H 2 price fixed at 3.5 $ kg −1 , a taxation of CO 2 of at least 300 $ ton −1 would be needed to obtain a positive revenue.The low price range of 1.5 $ kg −1 is currently only feasible for blue/gray hydrogen from fossil fuels [44], while according to the recent Monte Carlo study of Yates et al. [45], the high range of 3.5 $ kg −1 might not even be feasible for current PV electrolysis depending on the electrolyzer technology or location.To validate our simulations, TEA on the commercial Cu-ZnO-Al 2 O 3 catalyst was also performed, scaling the model developed by Szima et al. [28] to match our system productivity target.Similar CAPEX, H 2 prices and CO 2 taxations are required to reach a MeOH breakeven of 550 $ ton −1 (see section S7 in Supplementary Material), corroborating therefore our simulations and the high influence of the raw material costs.

Other process bottlenecks
An important factor that strongly affects the process that has not been considered up to now and is usually neglected in the literature is the inevitable formation of CH 4 during the process.Surprisingly, most of the reported kinetic models and process simulations for any of the MeOH catalysts on the state of the art do not take this constraint into account, modeling close to 100% recycle streams.Nonetheless, up to now, there is no reported catalyst able to totally suppress the formation of CH 4 .By way of example, additional CO 2 hydrogenation reaction tests with the commercial Cu-ZnO-Al 2 O 3 catalyst were conducted in a wide variety of conditions (see Table S4).Although low concentrations are achieved, the benchmark Cu-ZnO-Al 2 O 3 catalyst produces CH 4 even under the most favorable reaction conditions.The InCo catalyst studied herein, which demonstrated a significant improvement of the process selectivity [14], also produces CH 4 (Fig. 1a-c).As a product whose separation is not considered in the plant, it must be taken into account in order to obtain realistic results of the operation.
First, the formation of CH 4 must be included as an individual step of the reaction network, which is commonly avoided aiming for computational simplifications.In this case, we performed experiments in a very wide range of temperatures, even higher than the optimal one.The goal was to obtain CH 4 concentrations in the effluent high enough to calculate kinetic parameters with mathematical significance.Then, when CH 4 is included in the catalytic system, it directly affects the design of the plant, because its accumulation limits the maximum recycle stream.
In order to illustrate this effect, additional simulations were performed for our InCo catalyst considering the real CH 4 production, half of the CH 4 and no CH 4 production.For this, the kinetic constant of reaction 4 (Equation 11) was adapted in each case.The resulting MeOH prices for the 3 cases as a function of the CO 2 and H 2 prices are depicted in Fig. S11.Indeed, the lower the CH 4 formation, the lower the MeOH final price can be, independently of the H 2 or CO 2 prices.For instance, with no CH 4 formation, MeOH prices close to the target of 550 $ ton −1 can be achieved with a H 2 price of 2 $ kg −1 (a very realistic approach) and without any CO 2 taxation.The reason behind this is the ability to increase the recycle stream by reducing the CH 4 formation, as its accumulation in the plant will be greatly suppressed.In our particular case, by reducing the CH 4 formation by half, we can increase the recycle from 97.5 to 99% without having a CH 4 accumulation above 15%.Consequently, the MeOH productivity of the process increases by 15%.In case of no CH 4 formation, the recycle stream can be further increased to 99.5%, thereby boosting MeOH productivity up to 30%.
Apart from CH 4 , other undesired compounds can also affect the process in multiple ways rarely evaluated in most studies.For instance, the source of both CO 2 and H 2 will determine its purity and its price.Obviously, the highest purity means the highest production cost (see Table 4 [46]).This is especially important due to the already demonstrated high influence of the raw materials in the final MeOH cost (see Fig. 3c).Moreover, these traces can also affect the performance of the catalyst.As it has been reported, the commercial Cu-ZnO-Al 2 O 3 cannot tolerate high concentrations of O 2 (300 ppm) or Cl (10 ppm) due to metal reoxidation and Cu metal poisoning (CuCl is formed), respectively [47].To shed light on this hurdle, additional simulations were performed to investigate the accumulation of O 2 in the system when ppm of O 2 is fed with the reactants (see Fig. S12).Again, recirculation of gases (assumed at 99% for the simulations) leads to a magnification of the O 2 concentration in the system, especially when O 2 traces are considered in the H 2 feed stream.The 300 ppm limit established for the commercial Cu-ZnO-Al 2 O 3 catalyst is obtained feeding only 145 ppm of O 2 in the H 2 stream.As the recycle stream increases, the process requires even lower O 2 concentrations in the feed stream, thereby increasing drastically the H 2 and CO 2 purity needed and hindering the positive recycle productivity.One possible solution could be to include a deoxygenation (deoxo) catalytic reactor [48] in the system after the first compression steps (see Fig. S13).The rationale of this location is to take advantage of the high temperature of the feed after compression (ca.200 °C), enough to avoid any additional heat supply to the reactor, while still being below the 10 bar limit operation pressure of the deoxo process.However, these reactors cannot eliminate completely the O 2 in the system (ca.10 ppm [49]), do not tolerate inlet O 2 concentrations higher than 0.8%, and are based on expensive Pt or Pd catalysts that need to be replaced every ca. 5 years [50], which will drastically increase the OPEX costs. of the process will depend on non-catalytic variables as the reactant prices and their compression to the reaction pressure.Moreover, developing a catalyst that does not produce CH 4 or can withstand large oxygen impurities could have the greatest impact in the process.In particular, the absence of CH 4 could boost the productivity by minimizing the purges, whereas the increase in oxygen tolerance would allow low-cost feed gases to be used in the process, whose huge positive effects in the final MeOH breakeven price have been already discussed.For instance, in the particular case of green H 2 produced via water electrolysis, a great fraction of the costs comes from the oxygen removal after the electrolyzer [51,52].Consequently, catalyst development, if done in the right direction, could avoid purification costs and greatly increase the viability of the plant.

Life cycle assessment
Due to the above discussed key contribution of reactants and the renewable energetic requirements to the process, a LCA of the MeOH plant was also carried out (see Methods and Supplementary Material).Diagrams for the two study cases are provided in Fig. 4(a) (renewable energy sources, case 1) and Fig. 4(b) (non-renewable sources, case 2).The functional unit of both cases studied is the production of 2.75 10 8 kg of methanol, which is the production of the plant along one year with 8500 h working per year.A summary of the TEA estimation for the two study cases is summarized in Table S5.Fig. 5(a) shows the comparison of both study cases, with the values of each indicator being summarized in Tables S7 and S8.As observed, 10 of the  The relative contribution of the process variables to each indicator in both study cases is detailed in Fig. 5(b and c), respectively.As expected, a significantly higher contribution of electricity production (yellow bars) from the non-renewable sources can be easily observed for the study case 2. We can also observe that the contribution of CO 2 capture is negative on the GWP due to its consumption in the process and that, in the second case, the emission related with the energy supply and hydrogen production  Following, we have compared the MeOH price and the normalized GWP per ton of MeOH produced of our process with the ones obtained in the production of MeOH via conventional sources, i.e., natural gas and coal [53,54] (see Fig. 6).As expected, the MeOH prices that can be obtained from coal or natural gas are lower than the 550 $ target of the CO 2 valorization process (which is even hard to reach as the above TEA revealed).Nevertheless, when looking at the GWP, the picture changes and now the lowest value can be observed for case 1.The normalized GWP of case 1 is −1.5 kg CO 2 /kg MeOH, very close to the 1.75 maximum value predicted by the TEA.However, for case 2 the GWP is still slightly higher than in the particular case of MeOH from natural gas, making this process no sense from both an environmental and economic point of view.These results are in line with recent TEA/LCA work that report similar CO 2 abatement numbers and the importance of the energy mixt to the overall process [55][56][57].Case 2 uses traditional energy sources.Case 3 corresponds to case 1 plus the impact of the cement plant.MeOH from coal and natural gas was obtained using data from [53,54].
Lastly, we have also evaluated a third scenario where the impact of the cement plant normal operation is considered, being the energy of the MeOH plant still provided by renewable energy sources.While the selection of the source of CO 2 will largely impact the results (i.e., from a cement plant to direct air capture), it is also paramount to assess to total impact of the overall process.The results of this new scenario can be found in Table S9 of the Supplementary Material.A direct comparison between the 2 scenarios is also depicted in Fig. S14.We can observe that, by including the cement plant in the analysis and as expected, most of the indicators deteriorate.Especially severe is the deterioration of the respiratory inorganics (RI), photochemical ozone formation (POF) and acidification potential (AC).Similarly, the normalized GWP increases from the −1.5 kg CO 2 /kg MeOH of case 1 to 1.2 kg CO 2 /kg MeOH (Case 3 in Fig. 6), implying that the MeOH plant cannot fully offset the emissions of the highly contaminating cement plant.

Conclusions
A multidisciplinary study of a methanol production plant from CO 2 has been carried out, assessing from the possibilities of a novel catalyst to the economic viability and the main operational issues coupled with a life cycle assessment.Using a catalyst and a kinetic model made in-house, we have analyzed some relevant parameters for the implementation of the plant from the process and economic point of views, such as CH 4 trace formation, purity, and costs of raw gases, O 2 tolerance in the system, or influence of energy sources.
From experimental results at laboratory scale, an intrinsic kinetic model for the valorization of CO 2 into methanol has been developed using packed bed reactors and an InCo catalyst, which has been demonstrated to be highly active and selective.A comparison of the experimental and calculated data indicated an accurate prediction of the catalytic system by the intrinsic kinetic model and demonstrated that considering the CH 4 formation rate is essential for a truly accurate model.
The implementation of the above kinetic model in the process simulation allowed us to develop a realistic methanol production plant using Aspen Plus ® software, taking not only the catalyst CH 4 formation into account but also the important H 2 and CO 2 compression steps.
Based on our results, a successful scale-up of the process will depend on noncatalytic variables, mainly the reactant prices with H 2 itself accounting for most of the final MeOH costs.Our sensitivity analysis confirms that the viability of the plant will largely depend on the fluctuation of H 2 and CO 2 prices, requiring H 2 prices lower than 1.5 $ kg −1 or a taxation of CO 2 of at least 300 $ ton −1 to reach breakeven.Nevertheless, our TEA suggests that developing a catalyst that does not produce CH 4 or can withstand large oxygen impurities could have the greatest impact in the process.In particular, the absence of CH 4 could boost the productivity by minimizing the purges, whereas the increase in oxygen tolerance would allow low-cost feed gases to be used in the process, boosting the profitability of the system.Consequently, catalyst development in this particular direction, paying special attention to the CH 4 formation rate and the oxygen tolerance, could affect the viability of the plant more than the current trend of developing more and more active catalyst.
The evaluation of the plant energy sources indicates that, if only renewable energy sources are used and all the indirect sources of CO 2 are obviated, the maximum CO 2 abated in the plant will jump to 1.75 tons per ton of methanol, a 225% increase versus using traditional energy sources.The LCA supports this conclusion with 10 of the 16 indicators studied being lower for the case where only renewable energy is used to run the plan, with especial emphasis to the GWP indicator.
Hence, from an environmental point of view, our results suggest that it does not make sense to produce MeOH from CO 2 if all the energy requirements are also not renewable.Finally, when taking into account the impact of the CO 2 source (a cement plant in our case), most of the LCA indicators are deteriorated and the MeOH plant cannot fully offset the vast emissions of the cement plant.These results highlight the importance of looking at the overall picture when developing a catalyst/process for CO 2 conversion to methanol and, we hope that they help to pave the route for a truly carbon neutral economy.the financial support received from the Spanish Ministry of Science and Innovation with some ERDF funds (CTQ2016-77812-R) and the Basque Government (IT1218-19).T.
Cordero-Lanzac also acknowledges the Spanish Ministry of Education, Culture and Sport for the award of his FPU grant (FPU15-01666). A. Navajas and L.M. Gandía gratefully acknowledge the financial support from Spanish Ministerio de Ciencia, Innovación y Universidades, and the European Regional Development Fund (ERDF/FEDER) (grant RTI2018-096294-B-C31).L.M. Gandía also thanks Banco de Santander and Universidad Pública de Navarra for their financial support under "Programa de Intensificación de la Investigación 2018" initiative.

E j
Activation energy, kJ mol  From experimental results at laboratory scale, an intrinsic kinetic model for the valorization of CO 2 into methanol has been developed using experimental data of an InCo catalyst.
 The implementation of the kinetic model allowed us to simulate a realistic methanol production plant using Aspen Plus® software.
 An economically viable scale-up of the process will depend on non-catalytic variables, mainly the H 2 prices that accounts for most of the final MeOH costs.
 Developing a catalyst that does not produce CH 4 or can withstand large oxygen impurities could have the greatest impact in the process.
The reaction products were analyzed online using an Agilent 7890B GC with two sample loops.One sample loop goes to a TCD channel with 2 Haysep pre-column and MS5A, where He, H 2 , CH4 and CO are separated.Gases that have longer retention times than CO 2 on the Haysep column (Column 4 Haysep Q 0.5 m G3591-80023) are back-flushed.Further separation of permanent gases is done on another Haysep column (Column 5 Haysep Q 6 Ft G3591-80013) to separate CO 2 before going to MS5A.The second sample loop goes to a FID with an Innowax pre-column (5 m, 0.20 mm OD, 0.4 µm film), a Gaspro column (Gaspro 30 M, 0.32 mm OD) and another Innowax column (45 m, 0.2 mm OD, 0.4 µm).Gaspro column separates C 1 -C 8 , paraffins and olefins.Innowax separates oxygenates and aromatics.The CO 2 conversion ( , %) and selectivity of each i product (S i , %)

Fig. 1 .
Fig. 1.Evolution with temperature of the CO 2 conversion and products selectivity at 50 bar, H 2 /CO 2 =4 and using space time values of (a) 3.7, (b) 7.3 and (c) 11.0 g h mol CO2 −1 ; comparison of the experimental data (symbols) with those estimated by the model (line) for the evolution

Fig. 3 .
Fig. 3. (a) Equipment cost distribution for the MeOH process; (b) overview of the main costs for the MeOH process.H 2 price fixed to 3.5 $ kg −1 and CO 2 price to 50 $ ton −1 ; (c) contribution

Fig. 4 .
Fig. 4. Flow chart for the study cases (a) 1 and (b) 2. The impact of the operation of cement via steam cracking can actually overcome the CO 2 consumption.Those indicators with higher values in the case 1 study are mainly associated with the renewable source requirements.H 2 is the main contributor to ODP, LU and WU indicators, which is explained by photovoltaic and hydraulic energy production.Some indicators as HTC (human toxicity potential, cancer effects) and RDM (resource use of minerals and metals) are mainly related to the catalyst production and are unavoidable.The CO 2 capture process via amines (green bars) also negatively impacts most of the indicators expect the GWP.Nevertheless, although the different distribution on each indicator, we can consider that, overall, H 2 production (red bars) is the most environmentally impactful variable of the process for both cases, as well as the most expensive resource as predicted by the TEA (Fig.3).

Fig. 5 .
Fig. 5. (a) Comparison of the total environmental impact between study cases 1 and 2;

− 1 F 1 ij 1 S 1 ρ
Total molar flow rate, mol h −Compound i Reaction step j K j , K j * Equilibrium constant and equilibrium constant at the reference temperature of each j reaction step, respectively K CO2 , K H2O Adsorption constant for CO 2 and water, respectively, bar −1 k j , k j * Kinetic constant and kinetic constant at the reference temperature of each j reaction step, respectively L Catalytic bed length, m l Longitudinal position of the reactor n e , n c Number of experiments and compounds, respectively P R Partial pressure of the reactant R r i Formation rate of each I compound, mol kg −1 h −1 r j Reaction rate of each j reaction step, mol kg −1 h −Reactor Calculated molar fraction of each i compound y i e Experimental molar fraction of each i compound Greek symbols ΔH Adsorption heat, kJ mol −Catalytic bed density, kg m −3 ω i Weight factor of each i compound  A multidisciplinary study of a methanol production plant from CO 2 has been carried out, assessing from the possibilities of a novel InCo catalyst to the economic viability and the main operational issues coupled with a life cycle assessment.

Table 1 .
Kinetic and adsorption constants at the reference temperature (300 °C), activation energies and adsorption heats for the InCo intrinsic kinetic model.

Table 2 .
CO 2 mass balance for the MeOH process.If only renewable energy sources are used, impact of the CO 2 source emissions not included. a

Table 3 .
Summary of the cost evaluation for the MeOH process.
16 evaluated indicators are lower in the study case 1, highlighting the global warming potential (GWP) indicator that is reduced from 1.7 to −4.2 kg 10 8 CO 2 eq.when renewable energies are used to run the plant.Hence, in line with the above TEA discussion, we can resolve that it does not make sense to produce MeOH from CO 2 if all the energy requirements are also not renewable.Nevertheless, we also need to highlight that some indicators, specially the ozone depletion potential (ODP), land use (LU) and water use (WU) are higher in case 1.